Datasets:

alexshengzhili

/

0612_output_processed_final_human_written

like 0

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Reindeer (Rangifer tarandus) have lengthy seasonal migrations on land and their feet possess excellent locomotor characteristics that can adapt to complex terrains. In this study, the kinematics and vertical ground reaction force (GRF) of reindeer forelimb joints (interphalangeal joint b, metacarpophalangeal joint c, and wrist joint d) under walk, trot 1, and trot 2 were measured using a motion tracking system and Footscan pressure plates.</ns0:p><ns0:p>Significant differences among different locomotor activities were observed in the joint angles, but not in changes of the joint angles (&#945; b , &#945; c , &#945; d ) during the stance phase. Peak vertical GRF increased as locomotor speed increased. Net joint moment, power, and work at the forelimb joints were calculated via inverse dynamics. The peak joint moment and net joint power related to the vertical GRF increased as locomotor speed increased. The feet absorbed and generated more energy at the joints. During different locomotor activities, the contribution of work of the forelimbs changed with both gait and speed. In the stance phase, the metacarpophalangeal joint absorbed more energy than the other two joints while trotting and thus performed better in elastic energy storage. The joint angles changed very little (~5&#176;) from 0 to 75% of the stance phase, which reflected the stability of reindeer wrist joints. Compared to typical ungulates, reindeer toe joints are more stable and the stability and energy storage of forelimb joints contribute to locomotor performance in reindeer.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Large animals like ungulates and humans exhibit better locomotor efficiency than small animals like mice <ns0:ref type='bibr' target='#b1'>[1,</ns0:ref><ns0:ref type='bibr' target='#b2'>2]</ns0:ref>. This is because large animals use the ground reaction force (GRF) to store mechanical energy in their elastic feet to drive the trunk forward in a locomotor cycle. The storage and generation of elastic energy in the feet is an efficient way to reduce metabolic energy cost during locomotion <ns0:ref type='bibr' target='#b4'>[3]</ns0:ref>. Stretching of compliant tendons also allows limb muscles to save energy by isometric contraction under load <ns0:ref type='bibr' target='#b5'>[4]</ns0:ref>. Thus, the muscle-tendon units of foot joints in large animals serve important functions in energy storage, stabilization, and shock absorption.</ns0:p><ns0:p>In horses, during different locomotor activities, such as walking and trotting, the long digital flexor tendons stretch and recoil from metacarpophalangeal (MCP) dorsiflexion and plantarflexion, leading to elastic energy storage and energy generation at the joints <ns0:ref type='bibr' target='#b7'>[5]</ns0:ref>. This action is like a passive spring and benefits the forelimbs during locomotion <ns0:ref type='bibr' target='#b8'>[6]</ns0:ref>. The MCP joints are controlled by long tendons, superficial and deep digital flexor tendons, accessory ligaments, and muscles. These tendons are ideal structures for energy storage and generation <ns0:ref type='bibr' target='#b9'>[7]</ns0:ref>.</ns0:p><ns0:p>The stability of foot joints is associated with animal locomotion. For a trotting horse, both the elbows and shoulders of the forelimbs have net extension moments, but there is little joint movement when the moments are maximized. These joints are relatively rigid, which allow the trunk muscles to absorb and transmit energy <ns0:ref type='bibr' target='#b11'>[8]</ns0:ref>. The soft tissues at animal joints have important functions during locomotion. Additionally, kinematic parameters, such as joint angle, joint speed, and plantar pressure, all differ depending on locomotor speed and gait <ns0:ref type='bibr' target='#b12'>[9,</ns0:ref><ns0:ref type='bibr' target='#b14'>10]</ns0:ref>.</ns0:p><ns0:p>Gait refers to the pattern and order of limb locomotion. Most animals use different gaits based on their terrain and locomotor speeds <ns0:ref type='bibr' target='#b16'>[11,</ns0:ref><ns0:ref type='bibr' target='#b17'>12]</ns0:ref>. Gaits are categorized by order of ground contact into walking, trotting, and galloping <ns0:ref type='bibr' target='#b18'>[13]</ns0:ref>. Quadrupeds use different gaits depending on their locomotor speeds and select walking, trotting, or galloping during low-, moderate-, and highspeed locomotion activities, respectively. While walking and trotting, the limbs are in symmetrical gaits, and the left and right limbs are almost under constant relative phases with at least one limb in the stance phase. While galloping, the limbs are in asymmetrical gaits. As the left and right limbs change the relative phase with locomotor speed, a swing phase exists in which the four limbs are simultaneously in the air <ns0:ref type='bibr' target='#b19'>[14]</ns0:ref>. Studies on the gaits of horses, deer, and cheetahs in terms of mechanics, energy, kinematics, and dynamics have clarified that the mechanism of gait selection corresponding to locomotor speed is related to animal balance, speed, and energy saving <ns0:ref type='bibr' target='#b20'>[15]</ns0:ref><ns0:ref type='bibr' target='#b21'>[16]</ns0:ref><ns0:ref type='bibr' target='#b22'>[17]</ns0:ref><ns0:ref type='bibr' target='#b24'>[18]</ns0:ref><ns0:ref type='bibr' target='#b26'>[19]</ns0:ref>.</ns0:p><ns0:p>Based on locomotion speed and GRF, researchers have used various methods for studying the dynamics of animal limbs in detail. Pandy et al. calculated the inter-articular force, joint moment, and power of goats during different locomotion activities by using GRF and limb movement, and found that the foot inertia was small and negligible relative to the trunk inertia <ns0:ref type='bibr' target='#b16'>[11]</ns0:ref>. <ns0:ref type='bibr'>Dutto et al.</ns0:ref> measured the GRF, joint angle, moment, and power while trotting, and analyzed the kinematics and dynamics of the four limbs, as well as the energy storage and consumption of tendons <ns0:ref type='bibr' target='#b27'>[20,</ns0:ref><ns0:ref type='bibr' target='#b28'>21]</ns0:ref>. Moreover, the muscle stress of horse limbs while galloping was 200-400 kpa, and long tendons and extremely short pinnate muscle fibers allowed force production to be economical and enhanced the storage of tendon elastic energy <ns0:ref type='bibr' target='#b29'>[22]</ns0:ref>.</ns0:p><ns0:p>Load bearing and locomotion differ between the forelimbs and hindlimbs of animals. While trotting, the maximum vertical GRFs in the forelimbs and hindlimbs of German Shepherds were ~63% and ~37% of their body weight, respectively, and the impact on the forelimbs was significant <ns0:ref type='bibr' target='#b30'>[23]</ns0:ref>. While walking and trotting, the maximum vertical GRFs of the forelimbs were ~1.7 and ~1.4-times those of the hindlimbs, respectively <ns0:ref type='bibr' target='#b32'>[24]</ns0:ref>. The functions of horse forelimbs also differ depending on the GRFs, as the forelimbs mainly exert a braking effect and decrease the speed and kinetic energy, while the hindlimbs mainly play propulsive roles <ns0:ref type='bibr' target='#b33'>[25,</ns0:ref><ns0:ref type='bibr' target='#b34'>26]</ns0:ref>. While trotting, the maximum vertical GRFs of the forelimbs and hindlimbs are ~10-times the horizontal reaction and lateral reaction forces, while the vertical GRFs are much larger than the other component forces <ns0:ref type='bibr' target='#b34'>[26]</ns0:ref>. The GRFs of limbs in walking cows and gibbons are the same <ns0:ref type='bibr' target='#b36'>[27,</ns0:ref><ns0:ref type='bibr' target='#b37'>28]</ns0:ref>.</ns0:p><ns0:p>Reindeer, a typical Arctic migratory animal, have a limb structure suitable for migration in complex environments <ns0:ref type='bibr' target='#b39'>[29]</ns0:ref>. They can adapt to various terrains, such as ice, snow, wetland, and sand <ns0:ref type='bibr' target='#b41'>[30]</ns0:ref>. Reindeer seasonally migrate long distances on land and some populations migrate farther than other terrestrial mammals <ns0:ref type='bibr' target='#b42'>[31]</ns0:ref>. Additionally, all fibers in reindeer skeletal muscles have a high oxidative capacity, which may be related to endurance activity <ns0:ref type='bibr' target='#b44'>[32]</ns0:ref>. The sizes and structures of foot soles differ between forelimbs and hindlimbs <ns0:ref type='bibr' target='#b46'>[33]</ns0:ref>. The foot soles of the forelimbs are longer than the hindlimbs (87.0 &#177; 1.6 vs. 74.6 &#177; 1.0 cm). We speculate that this difference may be attributed to the different functions between reindeer forelimbs and hindlimbs during long migrations.</ns0:p><ns0:p>The toe and wrist joints of reindeer forelimbs are more stable than typical ungulates, and the stability of the wrist joint is higher. Reindeer MCP joints play the same energy storage role as in typical ungulates. Additionally, the work contribution from the forelimbs changed with gait and speed. In this study, the plantar pressure, kinematics, net joint power, and locomotor strategy of reindeer forelimbs were investigated using the vertical GRFs and limb movements during different locomotion activities. Based on previous studies, we used inverse dynamics and the static approach to explore the functions of the main forelimb joints, including energy saving and stabilization. We investigated whether the functions of interphalangeal joint b, MCP joint c, and wrist joint d in reindeer forelimbs were related to energy saving and stabilization during different locomotion activities. Depending on the locomotor posture and speed, reindeer locomotion was classified as walk, trot 1, or trot 2. In different locomotion activities, the temporal changes of plantar pressure and joint angles in the right forelimbs of four healthy adult male reindeer were measured. Based on inverse dynamics, we calculated the net joint moment and net joint power of the right forelimbs and analyzed the energy absorption and generation of the limbs at the joints.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Samples</ns0:head><ns0:p>Fifteen healthy 8-year-old adult reindeer, including seven males and eight females, were selected from the Evenki ethnic group in Genhe City, China. The female reindeer were all excluded to rule out sex differences. Three males rushed to the fence and suffered foot injuries during training, and thus were not used in the experiment. Finally, four other easily-trained and healthy male reindeer were selected as experimental subjects. The animal laboratory of Jilin University has granted ethical approval (No. 3130068) to this experiment. And the field experiments were approved by the Breeding Garden of Reindeer. The masses, shoulder widths, and body lengths of the subjects were 118.75&#177;14.93 kg, 1.22&#177;0.51 m, and 1.89&#177;0.83 m, respectively (mean&#177;standard deviation). All subjects were in healthy condition and had not undergone any surgical treatment or other invasive procedures. The reindeer were kept in an outdoor fenced area (2&#215;10 3 m 2 ) during the day with adequate supply of food and water, which was close to a semi-wild state, and were released into a wild forest at night. Before data collection, each reindeer was trained to walk and trot on the runway, for no less than 30 minutes, twice a day for one month. Manuscript to be reviewed of naturally dead reindeer were purchased and amputated from the wrist joints. The lower half of each forelimb was saved and sent for CT scans.</ns0:p></ns0:div> <ns0:div><ns0:head>Preparing the field</ns0:head><ns0:p>The field was 80 m long surrounded by 1.5-m-high fences with a 3-m-long and 1.5-m-wide data acquisition area in the middle (Fig. <ns0:ref type='figure'>1</ns0:ref>). The outside of the data acquisition area was a 77-m-long and 2-m-wide hard ground runway. The stones, weeds, and other debris were removed to ensure that the runway was even and there were places saved for reindeer to rest and eat at both ends. A pressure plate (2096&#215;472 mm 2 , 500 Hz sampling, 16384 sensors with 0.5&#215;0.7 cm 2 , USBII interface; Olen, Belgium) was placed on the runway and its position was adjusted to ensure that it was on the same plane as the runway. The pressure plate was connected via a signal conditioner (National Instruments, Austin TX, USA) to a computer (Dell, Xiamen, China) to record the data. One camera was placed on one side of the data acquisition area and another two cameras were placed on the other side. A high-speed camera system involving three synchronous digital cameras (Casio Exilim EX-FH25, Tokyo, Japan; 120 frame&#8226;s -1 ) was established. Prior to the experiment, a 36-point, three-dimensional calibration frame, located in the plane of movement over the force platform, was recorded for calibration.</ns0:p><ns0:p>During the procedures, the feeder used food or training instructions to guide the reindeer to walk and trot steadily on the runway. Adequate rest and food were provided for the animals within this period to prevent fluctuations in the test data. The locomotion of reindeer was divided by the gaits and speeds into walk (u=0.44&#177;0.08), trot 1 (u=0.95&#177;0.15), and trot 2 (u=1.46&#177;0.24).</ns0:p></ns0:div> <ns0:div><ns0:head>Markers and joint angles</ns0:head><ns0:p>We tested the three-dimensional (3D) coordinates of the five joints (a, b, c, d, e (Fig. <ns0:ref type='figure' target='#fig_7'>2A</ns0:ref>)) in the right forelimbs and three joint angles (&#945; b , &#945; c , &#945; d ) by using a three-camera motion tracking system (Simi Motion 2D/3D &#174; 7.5 software, SIMI Reality Motion Systems GmbH, Germany). The right forelimbs of the four adult reindeer underwent CT scans, and a 3D geometric model (Fig. <ns0:ref type='figure' target='#fig_7'>2C</ns0:ref>) of metacarpal, the second, third, fourth and fifth digits was established. Markers a (the dorsal of the hoof at the third digit), b (the proximal phalanx and the middle phalanx of the third digit at the joint), c (MCP joint), and d (wrist joint) were located according to the 3D limb model. The location of e (elbow joint) was determined based on the joint anatomy. Manuscript to be reviewed Regular circular reflective stickers (R=1.5 cm) were used as markers which were attached to the main joints of their right forelimbs (Fig. <ns0:ref type='figure' target='#fig_7'>2A</ns0:ref>). Researchers found that the relative locations of the distal phalanx and the middle phalanx were almost on one straight line <ns0:ref type='bibr' target='#b47'>[34]</ns0:ref>. Since the distal phalanx inside the hoof was hard to measure, the hoof and distal phalanx were taken as one part ( thepart surrounded by dotted lines in Fig. <ns0:ref type='figure' target='#fig_7'>2C</ns0:ref>) and marker a on the hoof was considered as the joint of the middle phalanx and the distal phalanx.</ns0:p><ns0:p>We defined three joint angles (Fig. <ns0:ref type='figure' target='#fig_7'>2E</ns0:ref>), including the joint angles between the middle phalanx and the proximal phalanx (&#945; b ), between the proximal phalanx of the third digit and the metacarpal (&#945; c ), and between the metacarpal and radius (&#945; d ).</ns0:p></ns0:div> <ns0:div><ns0:head>Vertical GRF</ns0:head><ns0:p>The vertical GRF of each right forelimb was measured by the pressure plate. Before the measurement, the subject's weight was input into the computer and then the subject moved on the pressure plate to complete calibration. The pressure data for the right forelimb were collected and analyzed, using Footscan 7Gait 2nd generation (RSscan International, Oren, Belgium).</ns0:p><ns0:p>Footscan was used to export the fore-aft coordinates of the COP during the full stance phase duration. We calculated the average path of the centre of pressure (COPy; fore-aft component) for a series of forefoot sequences within the same speed range. In terms of the kinematic data, when the displacement of the ungula cusps on Z-axis changed from a negative value to 0, it was identified as the touch-down moment of pressure plate data. When the displacement of the ungula cusps on Z-axis changed from 0 to a positive value, it was identified as the lift-off moment of pressure plate data (Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>). Taking the ungula cusps as the origin, the coordinates of the COP and the joints relative to the ungula cusp were calculated, and then a global coordinates of the kinematic data and the COP was established. The relationship between GRF and time was drawn and normalized to the sample mass. Angular velocity, stance phase, vertical GRF, net joint moment, power, and work were calculated on Origin Pro 2015 (OriginLab Corporation, Northampton, MA, USA) based on the data from the joint 3D coordinates.</ns0:p></ns0:div> <ns0:div><ns0:head>Net joint moment</ns0:head><ns0:p>The mass of the animal foot is small and the toe joints (the joint of the middle phalanx and the distal phalanx, the joint of the proximal phalanx and the middle phalanx) and wrist joint during the stance phase displaced less than other proximal joints <ns0:ref type='bibr' target='#b48'>[35,</ns0:ref><ns0:ref type='bibr' target='#b49'>36]</ns0:ref>. Therefore, we applied a static Manuscript to be reviewed approach regardless of gravity and inertia. The net joint moment (M m ) was determined by vertical GRF and joint position (Fig. <ns0:ref type='figure' target='#fig_7'>2D</ns0:ref>) and was equal to the product of vertical GRF (averaged from the fourreindeer at walk, trot 1, and trot 2) and L (vertical distance vector from the joint marker to the GRF) <ns0:ref type='bibr' target='#b50'>[37]</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_0'>) 1 ( m L GRF M &#61655; &#61501;</ns0:formula><ns0:p>We have defined the positive direction of the forelimb joint moment of the reindeer (Fig. <ns0:ref type='figure' target='#fig_7'>2D</ns0:ref>):</ns0:p><ns0:p>&#8226; For the wrist joint (d), net extension moment is positive (produced by extensor muscle), and net flexion moment is negative (produced by flexor muscle);</ns0:p><ns0:p>&#8226; For the joints of the toes (b, c), net flexion moment is positive (produced by the plantar flexor muscle) and net extension moment is negative (produced by the plantar extensor muscle).</ns0:p></ns0:div> <ns0:div><ns0:head>Net joint power and work</ns0:head><ns0:p>To estimate the energy absorbed and generated by the interphalangeal joint b, the MCP joint c, and the wrist joint d, we calculated the net joint power. Joint angular velocity was calculated from the joint angle versus the time derivative by using a differential function (the central difference method). The positive direction of angular velocity is the same as that of the joint moment (Fig. <ns0:ref type='figure' target='#fig_7'>2E</ns0:ref>). The net joint power (P m) of the joint equals the product of net joint moment (M m ) and joint angular velocity (&#969;), where &#969; is averaged from the four subjects at walk, trot 1, and trot 2 <ns0:ref type='bibr' target='#b48'>[35]</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_1'>) 2 ( m m &#969; M P &#61655; &#61501;</ns0:formula><ns0:p>When the directions of the joint moment and joint angular velocity are the same, the net joint power is positive, and otherwise it is negative. The work done at the joint is the integral of the net joint power with respect to time. Positive work and negative work represent the energy generated and absorbed by muscle-tendon units respectively <ns0:ref type='bibr' target='#b52'>[38]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Gait and speed</ns0:head><ns0:p>Each reindeer completed at least five groups of tests (walk, trot 1, and trot 2) on the hard ground.</ns0:p><ns0:p>We combined the research on the gaits of other animals (e.g. goats and horses <ns0:ref type='bibr' target='#b17'>[12,</ns0:ref><ns0:ref type='bibr' target='#b26'>19]</ns0:ref>) and reindeer's locomotor postures and then sorted out the gaits and order of the footprints (Fig. <ns0:ref type='figure'>1B</ns0:ref>). The reindeer's postures of the right forelimbs at walk, trot 1, and trot 2 during the stance phase were shown in Fig. <ns0:ref type='figure' target='#fig_8'>3</ns0:ref>. The moments of touch-down, mid-stance, and lift-off are 0%, 50% and 100% of the stance phase respectively.</ns0:p><ns0:p>&#8226; Walk: Symmetrical gait. At any time during the stance phase, at least two limbs are on the ground and four limbs leave the ground in sequence (e.g. the leaving sequence of left rear -left front -right rear -right front) (Fig. <ns0:ref type='figure' target='#fig_7'>2B</ns0:ref>).</ns0:p><ns0:p>&#8226; Trot: Symmetrical gait. Each forelimb and its diagonal hindlimb move in the same phase, and only two limbs are in the stance phase (sometimes all four legs are in the swing phase at the same time, e.g. the leaving sequence of left rear and right front-right rear and left front) (Fig. <ns0:ref type='figure' target='#fig_7'>2B</ns0:ref>).</ns0:p><ns0:p>Speed data were normalized by Froude number (u), where v is the average velocity, l is the height of the shoulder joint from touch-down to lift-off, and g is the acceleration of gravity:</ns0:p><ns0:formula xml:id='formula_2'>) 3 ( l g v u &#61655; &#61501;</ns0:formula><ns0:p>In order to examine changes with speed, relevant variables from all trotting trials were divided into two bins indicative of trot 1 (range of u: 0.8-1.1) and trot 2 (range of u: 1.1-1.7). The speeds of reindeer at walk, trot 1, and trot 2, after nondimensionalization, are 0.44&#177;0.08, 0.95&#177;0.15, and 1.46&#177;0.24 (mean&#177;S.D.).</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Vertical GRF, net joint moment and net joint work were normalized to stance duration and to reindeer's mass. Statistical analyses were conducted to examine the differences in different gaits and speeds, three key indicators (joint angles at touch-down, mid-stance and lift-off) between slow walking and trotting gaits/ trot 1 and trot 2 speeds by using Origin Pro 2015 (OriginLab Corporation, Northampton, MA, USA). We used one-way ANOVA statistical technique to analyze the effect of locomotor gait and speed on each joint angle indicator. F-test was conducted to examine whether these two variations are significantly different. Statistical significance level was considered as P &lt; 0.05.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Joint angles</ns0:head><ns0:p>During walk, trot 1, and trot 2 in the stance phase, the reindeer interphalangeal joint angle &#945; b (Fig. <ns0:ref type='figure' target='#fig_9'>4A-C</ns0:ref>), MCP joint angle &#945; c (Fig. <ns0:ref type='figure' target='#fig_9'>4D-F</ns0:ref>), and wrist joint angle &#945; d (Fig. <ns0:ref type='figure' target='#fig_9'>4G-I</ns0:ref> During different activities, the maximum and minimum values of &#945; b and &#945; c during the stance phase and the corresponding time points differed. Also the joint range of motion (ROM) was larger at the trotting gaits than at the walking gait. The ROMs of &#945; d among the three joint angles were the greatest, around 29&#176;, 30&#176; and 35&#176; during walk, trot 1 and trot 2, respectively. The ROMs of &#945; b were the smallest, around 26&#176;, 27&#176; and 33&#176; during walk, trot 1 and trot 2, respectively. Therefore, the relationships of joint angles with gaits and speeds showed that the joints can adapt to different gaits and locomotor speeds.</ns0:p></ns0:div> <ns0:div><ns0:head>Vertical GRF</ns0:head><ns0:p>The forelimbs have different vertical GRFs at different time points during the stance phase.</ns0:p><ns0:p>According to the time corresponding to the peak vertical GRF, the forelimb locomotion can be divided into a braking phase and a propulsive phase <ns0:ref type='bibr' target='#b53'>[39]</ns0:ref>. The vertical GRF increased with time during the braking phase, and decreased with time during the propulsive phase (Fig. <ns0:ref type='figure' target='#fig_11'>6</ns0:ref>).</ns0:p><ns0:p>The peak vertical GRFs (normalized to body mass) during walk, trot 1, and trot 2 were 8.95, 11.33, and 12.80 times the body mass, respectively, and the corresponding peak time was 57.03%, 50.45%, and 47.78% of the stance phase, respectively. The gaits and locomotor speeds of reindeer affect the vertical GRF. As for different gaits, the peak vertical GRF at trot was larger than that at walk. At the same gait, the peak vertical GRF at trot 2 was larger than that at trot 1.</ns0:p></ns0:div> <ns0:div><ns0:head>Net joint moment</ns0:head><ns0:p>In different activities, the forelimb joints b and c of reindeer in the stance phase (about 0 to 100%) produced positive net flexion moment by the plantar flexor (Fig. <ns0:ref type='figure' target='#fig_12'>7</ns0:ref>). Joint d in the early stance Manuscript to be reviewed phase (about 0 to 75%) and late stance phase (about 75 to 100%) generated the negative net flexion moment and positive net extension moment respectively by the flexor and extensor muscles (Fig. <ns0:ref type='figure' target='#fig_13'>8</ns0:ref>). Reindeer and horses have similar net joint moment curves for joints c and d when trotting on hard ground <ns0:ref type='bibr' target='#b27'>[20]</ns0:ref>.</ns0:p><ns0:p>In different activities, reindeer have different peak net joint moments at the forelimb joints.</ns0:p><ns0:p>Joints b, c, and d reached the peak net joint moments at about 45%, 50% and 30% of the stance phase, respectively. The peak net joint moments at walk, trot 1, and trot 2 were 0.28, 0.37, and 0.42 Nm&#8226;kg -1 at joint b, 0.55, 0.79, and 0.93 Nm&#8226;kg -1 at joint c, and -0.95, -1.35, and -1.78 Nm&#8226;kg -1 , at joint d, respectively. The vertical GRF of reindeer forelimbs increased with the rising locomotor speed and, accordingly, the peak joint moment also rose. Since the vertical distance vector of vertical GRF from joint d was the largest, joint d had the greatest peak joint moment, followed successively by joint c and joint b.</ns0:p></ns0:div> <ns0:div><ns0:head>Net joint power and work</ns0:head><ns0:p>The net joint moment reflects the activity (extension and flexion) of muscles (extensors and flexors) , but not the changes of energy in the muscle-tendon units at the joints. Net joint power and work, which are directly related to the energy absorption and generation at the limb joints <ns0:ref type='bibr' target='#b54'>[40]</ns0:ref>, at the forelimb joints were shown in Figs. <ns0:ref type='figure' target='#fig_13'>7 and 8</ns0:ref>. As mentioned above, the net joint moment and angular velocity at the forelimb joints increased with the rise of locomotor speed.</ns0:p><ns0:p>Therefore, the net joint power at the joints increased accordingly. The net joint power ranges at joint c were the largest and were -0.37 to 0.06, -0.19 to 0.21, and -4.37 to 2.46 W&#8226;kg -1 at walk, trot 1, and trot 2, respectively. As the locomotor speed was accelerated, the net joint power range was enlarged and thus the feet need to absorb and generate more energy at the joints.</ns0:p><ns0:p>In different activities, reindeer had similar work patterns at the same joint. From about 0 to 55% of the stance phase, the dorsiflexion of joint c produced a net flexion moment and the foot absorbed energy at the joint. From about 55% to 100% of the stance phase, joint c plantarflexed and the plantar flexor and extensor muscles generated and absorbed energy, respectively (details of energy absorption and generation at each joint are shown in Table <ns0:ref type='table'>1</ns0:ref>). The energy changes at the limb joints are related to joint functions, such as energy storage and stabilization <ns0:ref type='bibr' target='#b55'>[41,</ns0:ref><ns0:ref type='bibr' target='#b56'>42]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>We investigated whether the functions of interphalangeal joint b, MCP joint c, and wrist joint d in the forelimbs correlated with energy saving and stability. Depending on the locomotor postures and speeds, reindeer locomotor activities were divided into walk, trot 1, or trot 2. In different locomotion activities, we measured the temporal changes of plantar pressure and joint angles in the right forelimbs of four healthy adult male reindeer. Based on inverse dynamics, we calculated the net joint moment and net joint power of the right forelimbs, and the energy absorption and generation by the limb joints.</ns0:p><ns0:p>Locomotor strategies change depending on the speed and gait adopted by the animal. For example, goat limbs adjust the work of muscles and tendons to adapt to a walk or trot locomotion on slope-variable surfaces <ns0:ref type='bibr' target='#b57'>[43]</ns0:ref>. Horses rely on the minimization of metabolic costs and change gaits based on a range of different locomotion speeds, including the most energy-efficient trot gait <ns0:ref type='bibr' target='#b24'>[18]</ns0:ref>. In our study, during different gaits (walk or trot) and speeds (trot 1 or trot 2), significant differences were detected in the reindeer joint angles at the moments of touch-down, mid-stance, and lift-off (Fig. <ns0:ref type='figure'>9</ns0:ref>), which may be associated with reindeer locomotor strategies.</ns0:p><ns0:p>Although reindeer changed locomotor strategies during different gaits and speeds, we still found some similarities, such as the elastic energy saving function of joint b and the effect of joint d.</ns0:p></ns0:div> <ns0:div><ns0:head>Contribution of work change with gaits and speeds</ns0:head><ns0:p>Most animals use the inverted pendulum model in their walking gaits and restore mechanical energy via the periodic conversion of kinetic and potential energies <ns0:ref type='bibr' target='#b58'>[44,</ns0:ref><ns0:ref type='bibr' target='#b59'>45]</ns0:ref>. While trotting, the spring-mass system and inverted pendulum model are used, wherein the limbs act as springs that store and generate energy, which is characterized by a significant reduction in the difference between the potential and kinetic energies during the stance phase <ns0:ref type='bibr' target='#b60'>[46,</ns0:ref><ns0:ref type='bibr' target='#b61'>47]</ns0:ref>. In our study, significant differences were detected in joint angles &#945; b and &#945; c between walking and trotting gaits, and in &#945; b and &#945; d between trot 1 and trot 2 (Fig. <ns0:ref type='figure'>9</ns0:ref>). In the trotting gait, the MCP joint absorbed more energy than the other two joints (Table <ns0:ref type='table'>1</ns0:ref>), but in the walking gait, the MCP joint absorbed less energy than the wrist joint. This may be attributed to the preference of animals over the inverted pendulum gait at low speeds and over the mass spring inverted gait at high speeds, which both enhanced locomotor performance and energy saving <ns0:ref type='bibr' target='#b62'>[48,</ns0:ref><ns0:ref type='bibr' target='#b64'>49]</ns0:ref>. The different motion patterns among different gaits and speeds in reindeer forelimbs may be caused by the more efficient energy mechanism.</ns0:p><ns0:p>Reindeer have enhanced MCP joints. While walking, the proximal phalanx pivots about joint b (stance phase of 0-10%) with slight downward and upward movements (Fig. <ns0:ref type='figure' target='#fig_10'>5A</ns0:ref>). However, while trotting, the distal phalanx moves downward (joint b plantarflexion) for a prolonged period of time (stance phase of 0-20%) (Fig. <ns0:ref type='figure' target='#fig_10'>5A</ns0:ref>). Owing to the stretching and recoiling of plantar flexor tendons, plantarflexion and dorsiflexion of interphalangeal joint b are typical in loading and rebounding patterns. This indicates that the elastic elements at the toe joints offset the GRF and the trotting gait can reduce pressure, as well as protect the soft tissues of the toes by prolonging the foot-to-ground contact time ratio.</ns0:p><ns0:p>As previously reported in horses, as locomotor speed increased, the positive and negative work done by MCP joint c increased significantly, and elastic energy storage and generation also increased <ns0:ref type='bibr' target='#b27'>[20]</ns0:ref>. This finding is consistent with our results. When the ROM of MCP joints and net flexion moment increased as locomotor speed increased, the foot work of reindeer at the joint also increased. The difference in the ROM of wrist joint d was small among different activities (~5&#176;), but as locomotor speed increased, the vertical GRF and angular velocity also increased, thereby increasing the net joint moment and net joint power. Compared to the trotting gait, the power of the wrist joint was smaller, but fluctuated more severely in the walking gait (Fig. <ns0:ref type='figure' target='#fig_13'>8A</ns0:ref>).</ns0:p><ns0:p>Thus, slow locomotor activities may require a higher level of neural control <ns0:ref type='bibr' target='#b65'>[50]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Comparison of reindeer forelimb joints to typical ungulates</ns0:head><ns0:p>Ungulate locomotion has evolved in vastly different patterns depending on the specific habitat of a given species <ns0:ref type='bibr' target='#b66'>[51]</ns0:ref>. Compared to horse forelimbs, reindeer toe joints are stable, as the ROMs between the interphalangeal and MCP joints are smaller. The ROMs of interphalangeal joint b in reindeer and horses are ~30&#176; and ~40&#176;, respectively, and the ROMs of MCP joint c in reindeer and horses are ~31&#176; and ~40&#176;, respectively <ns0:ref type='bibr' target='#b27'>[20]</ns0:ref>. These differences indicate that reindeer forelimbs have stable toe joints. Measurements obtained using a linked segment model and spring coefficients from a spring model demonstrated that the stiffness of goat limbs was twice that of dog limbs during different activities, suggesting that goats have adapted to a rougher and steeper terrain <ns0:ref type='bibr' target='#b67'>[52]</ns0:ref>. Wrist joint d flexion produced a net flexion moment, which generated propulsion in the middle stance phase (45-75%) <ns0:ref type='bibr' target='#b68'>[53]</ns0:ref>. In the last 20% of the stance phase, the long digital flexor tendons at interphalangeal joint b and MCP joint c recoiled, and joint plantarflexion produced a net flexion moment; however, the net flexion moment and propulsion were small (Fig. <ns0:ref type='figure' target='#fig_12'>7</ns0:ref>). Wrist joint d changed between 0 and 75% of the stance phase (~5&#176;). While trotting, the change trend of wrist joint d in horses was similar to reindeer. The wrist joint angle of horses was maintained at 180-190&#176; within 0-60% of the stance phase, then gradually decreased <ns0:ref type='bibr' target='#b27'>[20]</ns0:ref>. Wrist joint d in reindeer was maintained and stable over time. Reportedly, horse knees during the stance phase produce a net flexion moment and the flexor muscles assist foot movement, wherein the extensor muscles stabilize the joints <ns0:ref type='bibr' target='#b65'>[50]</ns0:ref>. Similarly, reindeer wrist joint d displayed this stabilizing ability during the early and middle stance phases (0-75%). The small change in the joint angle (~5&#176;) indicated that the wrist joint plays a role in stabilizing foot locomotion. Thus, the low flexibility and high stability of the forelimb joints may be beneficial during long distance migrations.</ns0:p></ns0:div> <ns0:div><ns0:head>MCP joints as energy storage devices</ns0:head><ns0:p>The MCP joints of most animals elastically store and generate energy because they are mainly composed of small muscles, short pinnate muscle fibers, and long tendons <ns0:ref type='bibr' target='#b11'>[8,</ns0:ref><ns0:ref type='bibr' target='#b69'>54,</ns0:ref><ns0:ref type='bibr' target='#b70'>55]</ns0:ref>. Ligaments have a protective effect on joints <ns0:ref type='bibr' target='#b71'>[56]</ns0:ref>, and tendons also provide an energy advantage in highspeed locomotion <ns0:ref type='bibr' target='#b65'>[50]</ns0:ref>.</ns0:p><ns0:p>Reportedly, the relatively short muscle fibers and long tendons in turkey hindlimbs act like springs <ns0:ref type='bibr' target='#b72'>[57]</ns0:ref>, as the short muscle fibers contribute to more economical muscular energy and stretching of the long tendons allows muscle fibers to generate energy with little change in length, thus decreasing metabolic costs <ns0:ref type='bibr' target='#b73'>[58,</ns0:ref><ns0:ref type='bibr' target='#b74'>59]</ns0:ref>. The MCP joints in reindeer feet also absorb and generate energy during different locomotor activities (Table <ns0:ref type='table'>1</ns0:ref>), which are manifested by an elastic system for energy storage and generation (Fig. <ns0:ref type='figure' target='#fig_12'>7</ns0:ref>). The distal joints in horse forelimbs recover 40% of the energy during the stance phase <ns0:ref type='bibr' target='#b55'>[41]</ns0:ref>. Furthermore, 70-80% of the plantar flexors stretched at the metatarsal joints during the stance phase. The Achilles tendons, long plantarflexion tendons, and plantar connective tissues of the feet absorb energy and convert it into elastic potential <ns0:ref type='bibr' target='#b53'>[39]</ns0:ref>. Reindeer MCP joints at the same position in the forelimbs also performed well in energy storage during different locomotor activities and absorbed 6.42 &#215; 10 -2 , 19.71 &#215; 10 -2 , and 33.03 &#215; 10 -2 J&#8226;kg -1 of energy (negative power) in walk, trot 1, and trot 2, respectively. During different locomotor activities, the joint angles significantly differed at the touch-down, mid-stance, and lift-off moments. In the trotting gaits, the MCP joint absorbed more energy than the other two joints, but in the walking gaits, it absorbed less energy than the wrist joint. Across different gaits and locomotor speeds, the forelimbs adopted different locomotor strategies to improve locomotor performance and save energy.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>As reindeer speed increased, the peak joint moment and net joint power both increased. The feet absorbed and generated more energy at the joints. The feet first absorbed energy, then generated energy at the MCP joint during the stance phase, thus performing well in elastic energy storage.</ns0:p><ns0:p>In the middle stance phase (45-75%), the feet exerted a propulsive effect during the flexion of the wrist joint. In the early and middle stance phases (0-75%), the joint angle changed very little (~5&#176;) and the wrist joint stabilized the feet. Clearly, during long-distance migration, forelimbs play stability maintenance and energy storage roles.</ns0:p></ns0:div> <ns0:div><ns0:head>Forecast</ns0:head><ns0:p>The kinematics of hindlimb and the coordination of hindlimb and forelimb of reindeer would be analyzed to research the effect of reindeer foot joint on movement in a follow-up study. Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Four right forelimbsPeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>) showed similar patterns and ranges. Therefore, the data of joint angles of the four reindeer were combined to analyze the temporal variation of joint angles during the stance phase. The stick figure of the PeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020) Manuscript to be reviewed forelimbs during different locomotor stance phases was shown in Fig. 4J-L. At the moment of touch-down, the limb joints (b, c, d) first moved toward the ground and then left the ground after touching the lowest point. This pattern of motion may be related to energy saving. The integrated data correspond to the means and variances of &#945; b , &#945; c and &#945; d (Fig. 5A-C) at walk, trot 1, and trot 2 in the stance phases. The joint angles (&#945; b , &#945; c , &#945; d ) displayed similar patterns during different locomotor stance phases. The &#945; b increased (joint plantarflexion) during the early stance phase (about 0-30%), decreased (joint dorsiflexion) in the mid-stance phase (about 30%-80%), and rose (joint plantarflexion) in the late stance phase (about 80%-100%). The interphalangeal joint b plantarflexed in the late stance phase, and the hoof gradually lifted off the ground with the tip still in contact with the ground.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020) Manuscript to be reviewed The forelimb joint angles of reindeer (&#945; b , &#945; c , &#945; d ) changed in similar patterns during different locomotor stance phases. The peak vertical GRF increased as locomotor speed increased. The peak vertical GRFs (normalized to body mass) in walk, trot 1, and trot 2 were 8.95, 11.33, and 12.80-times the body mass, respectively.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1 Schematic 2 )</ns0:head><ns0:label>12</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2 Reindeer</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 3 Locomotor</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 4 The</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 5 Means</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Figure 6 Means</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>Figure 7 Means</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head>Figure 8 Means</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,70.87,525.00,408.75' type='bitmap' /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:04:47437:2:0:NEW 24 Sep 2020)</ns0:note> </ns0:body> "

"Rebuttal Letter Dear Editors and Reviewers, Thank you for your email dated 28 August 2020 with the reviewers’ comments concerning our manuscript entitled “Forelimb Joints Contributing to Locomotor Performance in Reindeer (Rangifer Tarandus) by Maintaining Stability and Energy Storage” (Article: 47437). After reviewing the comments, we revised the manuscript accordingly. Point by point responses to the editors and reviewers’ comments are listed below. We would like to express our sincere gratitude to the editors and reviewers for your constructive comments. Yours sincerely, Rui Zhang Replies to Editors: Field Permit: Comment 1: Thanks for supplying your field permit. Using this document, we have drafted this Field Studies statement: 'The Breeding Garden of Reindeer approved field experiments.' At the next revision, please include this information in your Methods section and re-upload the manuscript. Answer: Thank you for the comments. We have added relevant content according to the editor’s suggestion. Please refer to Samples (Line 118). Affiliations: Comment 2: We notice that the author affiliations you have provided in the system are slightly different to those in the document. System version: Jilin University, Key Laboratory of Bionic Engineering, Ministry of Education Jilin University, Changchun, China Manuscript version: Key Laboratory of Bionic Engineering, Ministry of Education, Jilin University, Changchun, PR China Answer: Thank you for pointing out the inconsistency. Now we unified the affiliation as “Key Laboratory of Bionic Engineering, Ministry of Education, Jilin University, Changchun, People’s Republic of China”. Please refer to Line 8. Figures: Comment 3: Apologies for not catching this earlier. Figure 4 also has multiple parts. Each figure with multiple parts should have alphabetical (e.g. A, B, C) labels on each part and all parts of each single figure should be submitted together in one file. In this case: the 12 parts of Figure 4 should be labeled A-L. Answer: Thank you for the comments. We have revised relevant content according to the suggestion. Please refer to Figure 4. Legends: Comment 4: Thank you for labeling the figures with multiple parts. Please be aware that the legends should also be edited to reflect the lettered parts. This can be done at the next round of revision or at the proof stage if this is accepted for publication. Answer: Thank you for the comments. We have revised the legends according to the editor’s suggestion. Please refer to the legends. Replies to Reviewer 1: Basic reporting: Comment 1: There remain a few wording issues throughout the manuscript. For example: Line 99. I would change the question to something like 'Does the work done by or on the forelimb change with speed?' Line 319. A suggestion would be to change this line to 'Locomotor strategies change depending on the speed and/or gait adopted by the animal.' Trying to avoid using 'different' twice in the sentence. This appears a couple of other times in the manuscript. It is a small issue, but relates to the readability of the doc. Answer: Thank you for the comments. We’ve revised related sentences and applied for the language polishing service provided by LetPub. Comment 2: I would have still liked to see the same axis scale (maximum, minimum and intervals) used in Figures 6,7, and 8. This would allow for a more visual comparison of the presented data in addition to the interpretation as to the values. Answer: Thank you for the comments. In Figures 6, 7, and 8, we used the same axis scale (maximum, minimum and intervals) for all values, except for the power values in Figures 7 and 8. That’s because power varies greatly in different motions. Please refer to Figures 6,7, and 8. Experimental design: Comment 3: Improved and sufficient. Answer: Thank you for your approval. Validity of the findings: Comment 4: Explanation is sufficient. Answer: Thank you for your approval. Comments for the Author: Comment 5: Thank you to the authors for the extensive revisions to the document. I appreciate the time and effort that are associated with doing these revisions. Answer: We are grateful for all the comments from you, which help immensely with the quality improvement of our article. Thank you so much for taking time to review and comment on the manuscript. The above revised contents are only a point-by-point response to the reviews. In order to fully understand the revised contents, please read the revised manuscript. Replies to Reviewer 2: Basic reporting: Comment 1: As before, in general, the grammar and writing are acceptable. There are a few instances where corrections are advisable, primarily in sentences added to this revision. The authors have met criteria for references, background, professional article structure, figures and tables. They have also provided the raw data. The results are self-contained and relevant to the introduction in the article. In response to reviewer comments, the authors have added questions to frame what they did in their study. I would have preferred specific hypotheses, but these are acceptable. Answer: Thank you for the comments. Experimental design: Comment 2: This is original primary research that is within the aims and scope of the journal. The research questions are relevant and meaningful. The authors state how this study fills a gap in the literature. The authors provided a rigorous investigation of reindeer forelimb mechanics using inverse dynamics and appear to have met institutional ethical requirements. The methods have been described sufficiently to allow for replication and improvements were made in their description in response to reviewer comments. Answer: Thank you for the comments. Validity of the findings: Comment 3: The findings in this study are valid, robust, and are sound. The authors have indicated what statistical methods they used (F-test), but did not offer a lot of detail. The authors conclusions are now connected to specific research questions posed in the introduction. Answer: We used the statistical methods of F-test. With regard to the method, please refer to another paper of ours, entitled “Zhang R , Ji, Q L , Luo G , et al. Phalangeal joints kinematics during ostrich (struthio camelus) locomotion. Peerj, 2017, 5(1), e2857”. Comments for the Author: Comment 4: Overall, I feel that the authors have improved their manuscript with the suggested changes from the reviewers. However, there are a number of grammatical issues, primarily with the sentences that were added in this revision which will require some proofreading. I have made a few comments by referencing line numbers in the manuscript that I added for easier reference. I will upload this file. Abstract [Line 22] Unnecessary ‘the’. Should read ‘increased with rising locomotor speed’ [Line 24] Missing ‘the’. Should read ‘…net joint power related to the vertical GRF’ Introduction [Line 46] Trunk shouldn’t be plural. Should read ‘…in their elastic feet to drive the trunk forward…’ There are still some grammatical and proofreading issues with this manuscript. I have pointed out the first few but many are in sentences that have been added to this revision. [Line 167] This is likely a translation issue, but I think it would be more appropriate to say that the females were ‘excluded’ rather than ‘abandoned’. [Line 447] The ‘Wrist joints as a stabilizer and a pusher’ section didn’t need to be completely removed, I just suggested removing references to human locomotion since I didn’t feel it was relevant. Answer: Thank you so much for the detailed suggestions. We’ve revised related contents and applied for LetPub proofreading service. The above revised contents are only a point-by-point response to the reviews. In order to fully understand the revised contents, please read the revised manuscript. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Vitamin D 1&#945;-hydroxylase CYP27B1 is the key factor in the vitamin D pathway. Previously, we analyzed the expression of CYP27B1 in human gingival fibroblasts in vitro. In the present study, we analyzed the gingival expression of CYP27B1 in vivo. Methods. Forty-two patients with periodontitis Stage IV Grade C and 33 controls were recruited. All patients with periodontitis had unsalvageable teeth and part of the wall of the periodontal pocket was resected and obtained after tooth extraction. All controls needed crown-lengthening surgery, and samples of gingiva resected during surgery were also harvested. All the individuals' gingivae were used for immunohistochemistry and immunofluorescence. In addition, gingivae from seventeen subjects of the diseased group and twelve subjects of the control group were analyzed by real-time PCR.</ns0:p><ns0:p>Results. Expression of CYP27B1 was detected both in gingival epithelia and in gingival connective tissues, and the expression in connective tissues colocalized with vimentin, indicating that CYP27B1 protein is expressed in gingival fibroblasts. The expression of CYP27B1 mRNA in gingival connective tissues and the CYP27B1 staining scores in gingival fibroblasts in the diseased group were significantly higher than those in the control group.</ns0:p><ns0:p>Conclusions. Expression of CYP27B1 in human gingival tissues was detected, not only in the fibroblasts of gingival connective tissues, but also in the gingival epithelial cells, and might be positively correlated with periodontal inflammation.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Vitamin D 3 is of great importance in regulating calcium and phosphorus metabolism and immunological responses <ns0:ref type='bibr' target='#b27'>(Sassi et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b22'>Medrano et al., 2018)</ns0:ref>. The active hormonal metabolite of vitamin D 3 , 1&#945;,25-dihydroxyvitamin D 3 (1,25OH 2 D 3 ), is formed by two-step hydroxylations <ns0:ref type='bibr' target='#b22'>(Medrano et al., 2018)</ns0:ref>: &#9312; from vitamin D 3 to 25-hydroxyvitamin D 3 (25OHD 3 ) by vitamin D 25-hydroxylase in the liver, followed by &#9313; from 25OHD 3 to 1,25OH 2 D 3 by vitamin D 1&#945;-hydroxylase in the kidney. Vitamin D 1&#945;-hydroxylase CYP27B1 was first detected in the kidney <ns0:ref type='bibr' target='#b30'>(Takeyama et al., 1997)</ns0:ref>, but subsequently extra-renal sites of 1,25OH 2 D 3 synthesis were also verified, including the skin <ns0:ref type='bibr' target='#b3'>(Bikle &amp; Christakos, 2020)</ns0:ref>, prostate <ns0:ref type='bibr' target='#b4'>(Capiod et al., 2018)</ns0:ref>, bone <ns0:ref type='bibr' target='#b33'>(van Driel et al., 2006)</ns0:ref>, eye <ns0:ref type='bibr' target='#b0'>(Alsalem et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b17'>Markiewicz et al., 2019)</ns0:ref>, blood vessels <ns0:ref type='bibr'>(Somjen et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zehnder et al., 2002)</ns0:ref>, human periodontal ligament cells and human gingival fibroblasts (hGFs) <ns0:ref type='bibr' target='#b12'>(Liu et al., 2012a;</ns0:ref><ns0:ref type='bibr' target='#b13'>Liu et al., 2012b)</ns0:ref>.</ns0:p><ns0:p>The vitamin D pathway, including connected reactions from the activation of toll-like receptors to the expression of the human cationic antimicrobial protein of 18 kDa (hCAP18) in monocytes, was first proposed in 2006 <ns0:ref type='bibr' target='#b14'>(Liu et al., 2006)</ns0:ref>. hCAP18 is the precursor of the important antimicrobial peptide, cathelicidin (composed of 37 amino acids, also called LL37), which is the end product of the vitamin D pathway <ns0:ref type='bibr' target='#b14'>(Liu et al., 2006)</ns0:ref>. LL37 has a broad-spectrum antibacterial effect, and has a regulatory effect on the immuno-inflammatory response <ns0:ref type='bibr'>(Teles et al., 2013;</ns0:ref><ns0:ref type='bibr'>Xhindoli et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b34'>Wang et al., 2019)</ns0:ref>. A similar pathway also exists in keratinocytes <ns0:ref type='bibr' target='#b28'>(Schauber et al., 2007)</ns0:ref>. In our previous study <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>, the vitamin D pathway was detected in hGFs, and vitamin D 1&#945;-hydroxylase CYP27B1 was demonstrated to be the key factor in the pathway. Our results suggested that the vitamin D pathway might be important in periodontal immune defense <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>, which was in line with another research group <ns0:ref type='bibr' target='#b39'>(Zhou et al., 2018)</ns0:ref>. As the key factor in the vitamin D pathway <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>, CYP27B1 is worthy of further research. To our knowledge, however, the in vivo expression of CYP27B1 in hGFs has not been reported.</ns0:p><ns0:p>Although CYP27B1 expressed in hGFs in vitro is the same as that in kidney, its regulation is different: periodontitis-related inflammatory stimuli interleukin-1&#946; (IL-1&#946;), sodium butyrate and Porphyromonas gingivalis lipopolysaccharide (Pg-LPS) induce significant up-regulation of CYP27B1, while regulators of 1&#945;-hydroxylase in kidney (parathyroid hormone, calcium and 1,25OH 2 D 3 ) do not significantly influence the expression of CYP27B1 in hGFs in vitro <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b13'>Liu et al., 2012b)</ns0:ref>. However, the actual situation in vivo is much more complicated than that simulated in vitro. Previously, our group reported that IL-1&#946; and butyric acid, which are both up-regulators of CYP27B1, could be detected in the gingival crevicular fluids of patients with periodontitis, and the concentrations were positively correlated with periodontal inflammation <ns0:ref type='bibr' target='#b10'>(Liu et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Lu et al., 2014)</ns0:ref>. Thus, based on our previous data, it might be hypothesized that CYP27B1 is expressed in hGFs in vivo and patients with periodontitis might have stronger expression. The aim of this study was to test the above hypothesis and to elucidate the features of CYP27B1 expression in hGFs in vivo. In addition, the expression of CYP27B1 in human gingival epithelial cells (hGEs) was analyzed.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Tissue sampling</ns0:head><ns0:p>The institutional review board of Peking University School and Hospital of Stomatology <ns0:ref type='bibr'>Peri-Implant Diseases and Conditions (Tonetti et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b8'>Lang &amp; Bartold, 2018;</ns0:ref><ns0:ref type='bibr'>Papapanou et al., 2018)</ns0:ref>, diagnosis was made for each individual after complete periodontal examination. The inclusion criteria were as follows. Periodontitis: at least eight teeth with probing depth (PD) &#8805; 7 mm and evidence of alveolar bone loss on radiographs; at least four teeth with mobility II or III; at least one unsalvageable tooth with mobility III and alveolar bone resorption close to the root apex, needing to be extracted. Healthy controls: no site with attachment loss (AL); no site with PD &gt; 3 mm after supragingival scaling; no radiographic evidence of alveolar bone loss; less than 10% of sites with bleeding on probing (BOP); at least one tooth needing crown-lengthening surgery. Any subjects with systemic diseases or pregnancy were excluded. All 75 subjects enrolled were non-smokers.</ns0:p><ns0:p>The PD of all the enrolled subjects' teeth and AL of each unsalvageable tooth or each control tooth needing crown-lengthening surgery were recorded at six sites (mesial, distal, and middle sites of facial and lingual sides). Bleeding index (BI) <ns0:ref type='bibr' target='#b19'>(Mazza et al., 1981)</ns0:ref> was also recorded for each tooth of each individual. The mean PD, AL and BI were calculated for each analyzed tooth.</ns0:p><ns0:p>The percentage of surfaces (facial and lingual) with BOP was also calculated and recorded as BOP%.</ns0:p><ns0:p>All unsalvageable teeth were extracted before periodontal treatment. After extraction of the unsalvageable teeth, part of the wall of the periodontal pocket was resected and harvested. The gingiva resected during the crown-lengthening surgery of the controls was also collected.</ns0:p><ns0:p>Gingival samples from 17 subjects of the diseased group and twelve subjects of the control group were divided into part 1 and part 2. Gingival connective and epithelial tissues were obtained from part 1 using sharp tissue scissors, and then were stored in RNAwait (Solarbio Science &amp; Technology Co., Beijing, China) at -80&#61616;C until RNA extraction. Part 2 and gingival samples from the other subjects were dehydrated and embedded in paraffin and serial sections were cut with the microtome set at 5 &#956;m. One section of each sample was examined after staining with hematoxylin and eosin (H&amp;E). Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Detection of CYP27B1 expression</ns0:head><ns0:p>RNA was extracted using Trizol (Invitrogen, Carlsbad, CA, US) and was reverse transcribed to cDNA using a reverse transcription kit (TOYOBO Life Science (Shanghai), Shanghai, China).</ns0:p><ns0:p>Real-time PCR reactions were performed using Faststart Universal SYBR Green Master Mix (Roche, Basel, Switzerland) in a real-time Thermocycler (Applied Biosystems, Foster City, CA, US) in triplicate. &#946;-actin was used as an internal control (Forward primer: 5'-GCCGTGGTGGTGAAGCTGT-3' and reverse primer: 5'-ACCCACACTGTGCCCATCTA-3').</ns0:p><ns0:p>The forward primer for detection of CYP27B1 was 5'-ACGGTGTCCAACACGCTCT-3' and the reverse primer was 5'-AACAGTGGCTGAGGGGTAGG-3'. Data are presented as relative mRNA levels calculated by the equation 2 -&#916;Ct (&#916;Ct = Ct of target gene minus Ct of &#946;-actin) <ns0:ref type='bibr' target='#b15'>(Livak &amp; Schmittgen, 2001)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Immunohistochemistry</ns0:head><ns0:p>Immunohistochemistry was performed according to previously described methods <ns0:ref type='bibr' target='#b9'>(Li et al., 2017)</ns0:ref>. Briefly, selected sections were transferred onto adhesive slides (Zhongshan Golden Bridge Biotechnology Co., Beijing, China), deparaffinized with xylene and rehydrated with descending concentrations of ethanol, then digested with 1 g/L trypsin at 37&#176;C for 10 min for antigen retrieval. Endogenous peroxidase blocking was achieved by treatment with 3% H 2 O 2 for 10 min at room temperature, then sections were incubated with primary sheep polyclonal antibody against CYP27B1 (1:100; The Binding Site Ltd., Birmingham, UK) at 4&#176;C for 12 h. This was followed by incubation with an anti-sheep secondary antibody (1:500; EarthOx Life Sciences, San Francisco, CA, US) at 37&#176;C for 30 min. The PV-9000 Polymer Detection System and a diaminobenzidine (DAB) kit (both from Zhongshan Golden Bridge Biotechnology Co., Beijing, China) were used for immunohistochemical staining of CYP27B1. The DAB staining time was 150 s for each section. Finally, sections were counterstained with hematoxylin.</ns0:p></ns0:div> <ns0:div><ns0:head>Immunofluorescence</ns0:head><ns0:p>After deparaffinization of the sections, antigen retrieval was accomplished by boiling in citric acid-sodium citrate buffer (0.01 M, pH 6.0) for 15 min, and endogenous peroxidase blocking was performed using the same method as described above. Then, sections were incubated with a PeerJ reviewing <ns0:ref type='table'>PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:ref> Manuscript to be reviewed working solution of primary rabbit anti-vimentin monoclonal antibody (ZA-0511; Zhongshan Golden Bridge Biotechnology Co., Beijing, China) at 4&#176;C for 12 h. Next, sections were incubated with an anti-rabbit secondary antibody (ZF-0511, diluted 1:400; Zhongshan Golden Bridge Biotechnology Co., Beijing, China) at 37&#176;C for 1 h. Then sections were incubated with primary mouse monoclonal antibody against CYP27B1 (sc-515903, diluted 1:50; Santa Cruz Biotechnology, Santa Cruz, CA, US) at 4&#176;C for another 12 h. The anti-mouse secondary antibody (sc-516141, diluted 1:50; Santa Cruz Biotechnology, CA, US) was added and incubated at 37&#176;C for 1 h, then nuclei were counterstained with DAPI (Neobioscience Biological Technology Co., Shenzhen, China), and sections were observed using immunofluorescence microscopy (Nikon, Tokyo, Japan).</ns0:p></ns0:div> <ns0:div><ns0:head>Image analysis</ns0:head><ns0:p>Image evaluation of the immunohistochemical results was performed by two experienced pathologists, who were unaware to which group the histological sections belonged. The CYP27B1 staining of each sample was rated as one of the following four grades: negative (-), weak (+), moderate (++) or strong (+++), translated as 0, 1, 2 and 3 for statistical analysis, respectively. The staining intensity of each sample was the consensus of the opinions of the two pathologists.</ns0:p><ns0:p>Each pathologist chose five non-overlapping 40&#215; microscope fields of each section for evaluation of hGFs. The total number of hGFs and the number of immunohistochemically CYP27B1-positive ones in each chosen microscope field were recorded and staining of gingival fibroblasts were rated as negative (-), weak (+), moderate (++) or strong (+++). The percentage of +, ++ or +++ hGFs was calculated and CYP27B1 staining score was calculated using the following formula: CYP27B1 staining score = (percentage of +++ cells) &#215; 3 + (percentage of ++ cells) &#215; 2 + (percentage of + cells). The method for calculating CYP27B1 staining score was previously reported by Yoon et al. <ns0:ref type='bibr' target='#b36'>(Yoon et al., 2010</ns0:ref>) and a higher score indicated stronger staining intensity. Cell counting and rating were performed by the two pathologists using a multi-person sharing microscope (Olympus, Tokyo, Japan) at the same time. They calculated the PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed average CYP27B1 staining score of each section, and the mean of the two average CYP27B1 staining scores for each section was used for analysis.</ns0:p><ns0:p>Since immunofluorescence was usually used for qualitative analysis, the results of immunofluorescence staining were only used for the colocalization of vimentin and CYP27B1 in hGFs.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical methods</ns0:head><ns0:p>The Mann-Whitney U Test was used to compare AL, BI, BOP%, the relative mRNA expression of CYP27B1 in gingival epithelia and the staining grades of the two groups since normal distribution was not assumed. All the other comparisons between the two groups were carried out using Independent-samples T Test. Statistical analyses were carried out using the SPSS 11.5 software package (SPSS Inc., Chicago, IL, US). A P value &lt; 0.05 was considered statistically significant.</ns0:p><ns0:p>All the parameters were used to calculate power values using PASS 2008 (NCSS Inc., Kaysville, UT, US) and each power value was over 0.99.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The characteristics of the two groups are shown in Table <ns0:ref type='table'>1</ns0:ref>. Significantly higher PD and AL were observed in the periodontitis group than in the control group. BI and BOP% of all teeth analyzed in the periodontitis group were 4 and 100%, respectively. BI of the teeth analyzed in the control group was 0 or 1, so none had BOP.</ns0:p><ns0:p>The mRNA expressions of CYP27B1 in gingival connective tissues of patients with periodontitis were significantly higher than those in the gingival connective tissues of controls (Fig. <ns0:ref type='figure'>1A</ns0:ref>). In contrast, there was no significant difference in CYP27B1 mRNA expression between the gingival epithelia of the two groups (Fig. <ns0:ref type='figure'>1B</ns0:ref>).</ns0:p><ns0:p>The gingiva of one patient with periodontitis (Fig. <ns0:ref type='figure'>2A-C</ns0:ref>) and one control (Fig. <ns0:ref type='figure'>2D-F</ns0:ref> Manuscript to be reviewed connective tissues were CYP27B1 positive, and the expression of CYP27B1 was also detected in gingival epithelia. In the periodontitis group, the expression of CYP27B1 was detected in all epithelial layers, but expression was stronger in the superficial layer than in the deep layer of the epithelia in the control group. As shown in Fig. <ns0:ref type='figure' target='#fig_9'>3 (A-F</ns0:ref>), the expressions of CYP27B1 and vimentin were colocalized, indicating that in gingival connective tissues, the cells positive for CYP27B1 expression were hGFs. Statistical analysis indicated that CYP27B1 staining intensities of the gingiva of patients with periodontitis [3.00, (3.00 to 3.00)] were significantly higher than those of the controls [1.00, (1.00 to 2.00)] (Fig. <ns0:ref type='figure' target='#fig_10'>4A</ns0:ref>). In all the 40&#215; microscopic fields chosen for analysis, almost 100% of the hGFs were CYP27B1 positive and the CYP27B1 staining scores of hGFs of patients with periodontitis (2.49 &#177; 0.08) were significantly higher than those of controls (1.84 &#177; 0.12) (Fig. <ns0:ref type='figure' target='#fig_10'>4B</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our previous experiments in vitro verified that CYP27B1 is expressed in hGFs and that expression is up-regulated by the inflammatory stimuli, IL-1&#946; and sodium butyrate <ns0:ref type='bibr' target='#b13'>(Liu et al., 2012b)</ns0:ref>. In the present study, we demonstrated the expression of CYP27B1 in gingival connective tissues in vivo. Since there are several types of cells in gingival connective tissues, immunofluorescence experiments were performed and demonstrated that CYP27B1 colocalized with vimentin, a marker of fibroblasts, indicating that the CYP27B1-positive cells in gingival connective tissues were hGFs. However, it should be pointed out that endothelial cells also express vimentin (Piera-Velazquez &amp; Jimenez, 2019), as shown in Fig. <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>. Additionally, as shown in Fig. <ns0:ref type='figure'>2</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>, endothelial cells were also found to be CYP27B1 positive, which was in line with the results of Zehnder et al. <ns0:ref type='bibr' target='#b37'>(Zehnder et al., 2002)</ns0:ref>. Because hGFs and endothelial cells can easily be distinguished by pathologists, our morphological analysis of hGFs was not influenced by endothelial cells. Although the actual situation in vivo is much more complex than that simulated in the laboratory, our observations that patients with periodontitis had higher mRNA expression of CYP27B1 and higher CYP27B1 staining scores were in line with our results in Manuscript to be reviewed vitro. Therefore, our hypothesis that 'CYP27B1 is expressed by hGFs in vivo and the expression might be positively associated with periodontal inflammation' was verified.</ns0:p><ns0:p>Although we focused on hGFs, epithelial tissues were also observed in the present study, and gingival epithelia were found to be CYP27B1 positive in both groups, which was in line with the findings of other researchers <ns0:ref type='bibr' target='#b20'>(McMahon et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b23'>Menzel et al., 2019)</ns0:ref>. What should be pointed out is that the distribution of CYP27B1 expression differed between the two groups (Fig. <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>). When obtaining the gingival epithelial tissues for analysis of mRNA expression, it was impossible to obtain the entire epithelium clinically. Therefore, only the superficial layer was obtained to avoid contamination of connective tissues. Since CYP27B1 expression was relatively strong in the superficial layers in both groups, the lack of a significant difference in CYP27B1 expression between the epithelia of the two groups could be explained.</ns0:p><ns0:p>It has been demonstrated that a vitamin D pathway exists in hGFs <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref> and hGEs <ns0:ref type='bibr' target='#b20'>(McMahon et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b23'>Menzel et al., 2019)</ns0:ref>. The pathway might be involved in periodontal immune defense for the following reasons. (1) 25OHD 3 alleviates experimental periodontitis in diabetic mice via the vitamin D pathway <ns0:ref type='bibr' target='#b39'>(Zhou et al., 2018)</ns0:ref>. ( <ns0:ref type='formula'>2</ns0:ref>) In hGFs, the pathway is activated by the periodontal inflammatory stimulus Pg-LPS, and suppresses the expression of some inflammatory chemokines such as IL-8 and MCP-1 <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>, indicating that the pathway might play a role in immune defense in periodontal soft tissues. (3) 25OHD 3 is an important part of the vitamin D pathway, and higher 25OHD 3 concentrations were detected in both the gingival crevicular fluids and the plasma of aggressive periodontitis patients compared to those of healthy controls <ns0:ref type='bibr' target='#b11'>(Liu et al., 2009)</ns0:ref>. Moreover, after periodontal inflammation is reduced by initial periodontal therapy, the 25OHD 3 levels in gingival crevicular fluids and plasma significantly drop <ns0:ref type='bibr' target='#b10'>(Liu et al., 2010)</ns0:ref>, indicating that activity of the vitamin D pathway might be positively associated with periodontal inflammation. As the key factor in the pathway <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>, CYP27B1 is worthy of further research. In the present study, the finding that the in vivo gingival CYP27B1 expression was higher in the periodontitis group than in the control group could provide new evidence of the involvement of the vitamin D pathway in Manuscript to be reviewed periodontal immune defense. According to the study by <ns0:ref type='bibr' target='#b29'>Tada et al. (Tada et al., 2016)</ns0:ref>, 1,25OH 2 D 3 stimulation resulted in over 70-fold up-regulation of hCAP-18/LL-37 in an hGE cell line (Ca9-22). In contrast, 25OHD 3 or 1,25OH 2 D 3 stimulation only resulted in 3-to 4-fold enhancement of expression of hCAP-18/LL-37 in hGFs <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>. As the forefront of periodontal immune defense, it is reasonable that hGEs had a more active vitamin D pathway than hGFs. However, the relatively less active vitamin D pathway in hGFs is still worth studying, because once hGE, as the first line of periodontal defense, is breached and periodontal inflammation exacerbates to include gingival connective tissues, hGFs could play their role in periodontal immune defense through the vitamin D pathway. Thus, the present study is of biological significance, although more mechanisms via which the vitamin D pathway impacts gingival health in periodontitis still need to be elucidated.</ns0:p><ns0:p>Reasons for the higher expression of CYP27B1 in the periodontitis group might be as follows:</ns0:p><ns0:p>(1) 25OHD 3 is an up-regulator of CYP27B1 in hGFs <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref>, and 25OHD 3 levels in gingival crevicular fluids of patients with periodontitis before initial periodontal therapy were significantly higher than those after therapy <ns0:ref type='bibr' target='#b10'>(Liu et al., 2010)</ns0:ref>; (2) Periodontal inflammation results in higher concentrations of IL-1&#946; and butyric acid in gingival crevicular fluids <ns0:ref type='bibr' target='#b10'>(Liu et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Lu et al., 2014)</ns0:ref>, which also induces the expression of CYP27B1 in hGFs <ns0:ref type='bibr' target='#b13'>(Liu et al., 2012b)</ns0:ref>.</ns0:p><ns0:p>Our previous studies <ns0:ref type='bibr' target='#b11'>(Liu et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b10'>Liu et al., 2010)</ns0:ref> indicated that systemic and local 25OHD 3 levels in patients with aggressive periodontitis were positively associated with periodontal inflammation. However, several existing studies <ns0:ref type='bibr' target='#b5'>(Dietrich et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b7'>Jimenez et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b38'>Zhan et al., 2014)</ns0:ref> suggested that vitamin D deficiency is associated with an increased risk of periodontal disease. What should be pointed out is that in these studies, the participants were about 50 years of age or older, an age range that did not overlap with that of the population in our previous studies (younger than 30 years old). Additionally, no correlation between plasma 25OHD 3 levels and periodontal health was found in another large cross-sectional study <ns0:ref type='bibr' target='#b2'>(Antonoglou et al., 2015)</ns0:ref>, and the participants in that study were 30-49 years old. Thus, the Manuscript to be reviewed relationship between 25OHD 3 and periodontitis in people of different ages might be different.</ns0:p><ns0:p>In studies investigating the association between 25OHD 3 and periodontal health in large samples <ns0:ref type='bibr' target='#b5'>(Dietrich et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b7'>Jimenez et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b38'>Zhan et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b2'>Antonoglou et al., 2015)</ns0:ref>, the participants were from the general population. However, in our previous study <ns0:ref type='bibr' target='#b11'>(Liu et al., 2009)</ns0:ref>, only patients with aggressive periodontitis had higher plasma 25OHD 3 levels, and the patients had much more severe periodontal inflammation than the other participants. In the special group, it is unclear whether the higher plasma 25OHD 3 level is the reason for or the result of severe periodontal inflammation. Our previous study <ns0:ref type='bibr' target='#b6'>(Gao et al., 2018)</ns0:ref> suggested that 25OHD 3 activates the vitamin D pathway, which participates in periodontal immune defense. Therefore, it is possible that, due to severe periodontal inflammation, more LL37 is needed for antibacterial and anti-inflammatory function, and more 25OHD 3 is synthesized for the more active vitamin D pathway in periodontium. This possibility could help to explain why patients with severe periodontitis had higher systemic and local 25OHD3 levels. In the present study, the results that patients with periodontitis had higher CYP27B1 expression in hGFs indicated that patients with periodontitis had a more active vitamin D pathway in hGFs, which further supported this possibility.In the present study, subjects matched by age and gender were included in the two groups and all were non-smokers, in order to minimize the influence of potential confounding factors. To analyze the typical inflammatory situation in vivo, all the patients enrolled were diagnosed with periodontitis Stage IV Grade C, the most severe periodontitis in the new classification scheme for periodontal diseases <ns0:ref type='bibr' target='#b25'>(Tonetti et al., 2018;</ns0:ref><ns0:ref type='bibr'>Papapanou et al., 2018)</ns0:ref>. In addition, all gingival tissues of patients with periodontitis were obtained around unsalvageable teeth, which had not received any periodontal therapy so that periodontal inflammation was serious enough and was not influenced by periodontal treatments. The PD and AL of the unsalvageable teeth analyzed were high and BOP was positive at all surfaces of the teeth. In contrast, all the teeth analyzed in the control group had PD less than 3 mm and had no AL or BOP, indicating that these teeth were clinically healthy. It should be pointed out that all the teeth analyzed in the control group needed crown-lengthening surgery because of excessive Manuscript to be reviewed gingival display or subgingival location of fracture lines or carious lesions. When parts of the teeth were subgingival, accumulation of dental plaque was often detected. Thus, the BI of some teeth in the control group was 1 and mild inflammation of the gingiva could be detected.</ns0:p><ns0:p>Similarly, it was reported that 'healthy' gingiva might also harbor inflammatory cellular infiltrates, indicating that subclinical gingivitis might exist <ns0:ref type='bibr' target='#b8'>(Lang &amp; Bartold, 2018)</ns0:ref>. Thus, CYP27B1 staining intensities of two of the 33 teeth in the control group were strong (+++) and the mild inflammation of the gingiva might be the reason for high expression of CYP27B1 in the control group.</ns0:p><ns0:p>Immunohistochemistry is of course a highly subjective method. We tried to objectively evaluate CYP27B1 expression in gingiva in vivo by letting two experienced pathologists perform the evaluation in a blinded manner. However, the subjectivity of the evaluation was inevitable, which is a limitation of the present study.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, CYP27B1 expression was detected in hGFs in vivo, and this expression might be induced by periodontal inflammation. These results validated our previous in vitro findings, and indicated the potential involvement of the vitamin D pathway in periodontal immune defense.</ns0:p><ns0:p>The present study can help lay the foundation for using vitamin D pathway in the treatment of periodontitis via vitamin D supplement. Manuscript to be reviewed Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref> Evaluation of CYP27B1 protein expressions in gingiva (A) CYP27B1 staining intensities of gingival connective tissues of the diseased group were significantly higher than those of the control group. (B) CYP27B1 staining scores of gingival fibroblasts of the diseased group were significantly higher than those of the control group.</ns0:p><ns0:note type='other'>Figure 3</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>approved the study protocol (PKUSSIRB-2011007). Written informed consent was obtained from each participant in accordance with the Declaration of Helsinki. Forty-two patients with periodontitis Stage IV Grade C and 33 healthy controls were enrolled from the clinic of the Periodontology Department, Peking University School and Hospital of PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020) Manuscript to be reviewed Stomatology. On the basis of the 2017 World Workshop on the Classification of Periodontal and</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>) are shown. Negative controls are shown in Fig. 2 (G-I). The black frame indicates the epithelial tissue, and the blue frame indicates the connective tissue. As shown in the figure, the gingival PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49910:1:1:NEW 27 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 1 Fig. 1</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 2 Fig. 2</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Fig. 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig.3 Colocalization of CYP27B1 and vimentin in hGFs</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> </ns0:body> "

"Dear Dr. Verghese, Thank you very much for your consideration. We greatly appreciated the 2 reviewers’ valuable advice. The advice has been read word for word and the manuscript has been revised accordingly. Below are the responses to the reviewers’ comments (sentences in bold italics are the reviewers’ comments). Each change in the manuscript was highlighted in yellow. Reviewer 1 (Gill Diamond) Basic reporting The article is clearly written, providing sufficient literature references for background, methodology and discussion. The structure of the article is appropriate with figures and tables properly designed.  One issue is that the data in figures 1 and 2 are presented as relative mRNA expression, apparently each is relative to b-actin levels. However, figure 2 is simply a reshowing of figure 1. Is this just the controls or the periodontitis? Regardless, it would be much more appropriate to make the control set to 1.0 for all cases in figure 1, with the periodontitis shown relative to that amount. Then, it would make sense to show the data in figure 2, setting either one as the 1.0. As it stands, there is no reason for figure 2, and the data are hard to understand in figure 1. Response: The advice of revising Figure 1 and 2 is so valuable. According to the advice, Figure 1 was revised and Figure 2 was deleted. In Figure 1, the relative mRNA expression of CYP27B1 is relative to beta-actin levels. After revision, the control in Figure 1 was set to 1.0. After the data transformation, the relative mRNA expression of CYP27B1 in gingival epithelia didn’t conform to the normal distribution. Thus, Figure 1B was revised as a box-plot. The revised Figure 1 is as follows. Experimental design The authors have previously shown that CYP27B1 is expressed in vitro, so it is important to show this expression in vivo, especially where it is difficult to distinguish cell type (fibroblast vs. epithelium). Thus, the immunohistochemistry data are important. The results are clearly shown and support the in vitro data. All of the methods are sufficiently detailed. That said, the ultimate significance of the fact that CYP27B1 is expressed in these tissues is not well elaborated. It would strengthen the manuscript to explain to the reader why it is important to know A) that CYP27B1 is expressed in both (or either) cell type, and B) that CYP27B1 expression apparently differs in hGF vs. hGE in periodontal disease. Response: This is really a thoughtful comment. Liu et al. (Liu et al., 2006) first proposed that the vitamin D pathway includes a sequence of reactions from 25(OH)D3 to the production of hCAP18/LL37, and reported that TLR2 agonists show synergistic effects with 25(OH)D3 in the production of antimicrobial peptides in monocytes. It has been demonstrated that the vitamin D pathway also exists in hGFs (Gao et al., 2018) and hGEs (McMahon et al., 2011; Menzel et al., 2019). The pathway might be involved in periodontal immune defense for the following reasons. (1) 25OHD3 could alleviate experimental periodontitis in diabetic mice via the vitamin D pathway (Zhou et al., 2018). (2) In hGFs, the pathway could be activated by periodontal inflammatory stimulus Pg-LPS, and the pathway could suppress the expression of some inflammatory chemokines such as IL-8 and MCP-1 (Gao et al., 2018), indicating that the pathway might play a role in immune defense in periodontal soft tissues. (3) 25OHD3 is an important part of the vitamin D pathway, and higher 25OHD3 concentrations were detected in both the gingival crevicular fluids and the plasma of aggressive periodontitis patients compared to those of healthy controls (Liu et al., 2009). Moreover, after the periodontal inflammation was reduced by initial periodontal therapy, the 25OHD3 levels in gingival crevicular fluids and plasma significantly dropped (Liu et al., 2010), indicating that activity of the vitamin D pathway might be positively associated with periodontal inflammation. As the key factor in the pathway (Gao et al., 2018), CYP27B1 is worthy of further research. In the present study, the finding that the in vivo gingival CYP27B1 expression was higher in the periodontitis group than in the control group could provide new evidence of the involvement of the vitamin D pathway in periodontal immune defense. According to the study by Tada et al. (Tada et al., 2016), 1,25OH2D3 stimulation resulted in over 70-fold up-regulation of hCAP-18/LL-37 in a hGEs cell line (Ca9-22). In contrast, 25OHD3 or 1,25OH2D3 stimulation only resulted in 3- to 4-fold enhancement of expression of hCAP-18/LL-37 in hGFs (Gao et al., 2018). As the forefront of periodontal immune defense, it is reasonable that hGEs had more active vitamin D pathway than hGFs. But the relatively less active vitamin D pathway in hGFs is still worth studying, because once the first line of periodontal defense hGEs is breached and periodontal inflammation exacerbates to gingival connective tissues, hGFs could play their role in periodontal immune defense through the vitamin D pathway. Thus, the present study is of biological significance. These sentences have been added into the third paragraph of the Discussion. Validity of the findings The data appear to be fairly well presented. It would strengthen the rigor of the study to present the PCR data with each data point (either as a box plot or to keep the column graph, but to indicate each data point). Response: In the revised Figure 1, the PCR data with each data point were presented to strengthen the rigor of the study. The revised Figure 1 is as follows. The conclusions are presented, but some speculation on the further utility of this information would truly strengthen the manuscript. Response: The present study may help lay the foundation for using vitamin D pathway in the treatment of periodontitis via vitamin D supplement. The sentence has been added at the end of the manuscript. Comments for the Author Overall, this is a straightforward study demonstrating the expression of CYP27B1 in gingival tissue in vivo. It is important to the field to know this information, but the overall significance of the findings should be elaborated upon. Response: It has been demonstrated that a vitamin D pathway exists in hGFs (Gao et al., 2018) and hGEs (McMahon et al., 2011; Menzel et al., 2019): a sequence of reactions from 25(OH)D3 to the production of hCAP18/LL37. The pathway might be involved in periodontal immune defense for the following reasons. (1) 25OHD3 could alleviate experimental periodontitis in diabetic mice via the vitamin D pathway (Zhou et al., 2018). (2) In hGFs, the pathway could be activated by periodontal inflammatory stimulus Pg-LPS, and the pathway could suppress the expression of some inflammatory chemokines such as IL-8 and MCP-1 (Gao et al., 2018), indicating that the pathway might play a role in immune defense in periodontal soft tissues. (3) 25OHD3 is an important part of the vitamin D pathway, and higher 25OHD3 concentrations were detected in both the gingival crevicular fluids and the plasma of aggressive periodontitis patients compared to those of healthy controls (Liu et al., 2009). Moreover, after the periodontal inflammation was reduced by initial periodontal therapy, the 25OHD3 levels in gingival crevicular fluids and plasma significantly dropped (Liu et al., 2010), indicating that activity of the vitamin D pathway might be positively associated with periodontal inflammation. As the key factor in the pathway (Gao et al., 2018), CYP27B1 is worthy of further research. In the present study, the finding that the in vivo gingival CYP27B1 expression was higher in the periodontitis group than in the control group could provide new evidence of the involvement of the vitamin D pathway in periodontal immune defense. According to the study by Tada et al. (Tada et al., 2016), 1,25OH2D3 stimulation resulted in over 70-fold up-regulation of hCAP-18/LL-37 in a hGEs cell line (Ca9-22). In contrast, 25OHD3 or 1,25OH2D3 stimulation only resulted in 3- to 4-fold enhancement of expression of hCAP-18/LL-37 in hGFs (Gao et al., 2018). As the forefront of periodontal immune defense, it is reasonable that hGEs had more active vitamin D pathway than hGFs. But the relatively less active vitamin D pathway in hGFs is still worth studying, because once the first line of periodontal defense hGEs is breached and periodontal inflammation exacerbates to gingival connective tissues, hGFs could play their role in periodontal immune defense through the vitamin D pathway. Thus, the present study is of biological significance. These sentences have been added into the third paragraph of the Discussion. Reviewer 2 (Anonymous) Basic reporting Literature analysis is incomplete, own works are over cited. Response: This comment is so valuable. We re-analyzed the literature and the following 12 articles have been added in the Introduction and Discussion. Antonoglou et al., 2015. Journal of Periodontology 86:755-65. Dietrich et al., 2004. American Journal of Clinical Nutrition 80:108-113. Jimenez et al., 2014. Public Health Nutrition 17:844-852. Liu et al., 2006. Science 311:1770-1773. Schauber et al., 2007. Journal of Clinical Investigation 117:803-811. Tada et al., 2016. Biomedical Research 37:199-205. Takeyama et al., 1997. Science 277: 1827-1830. Teles et al., 2013. Science 339:1448-1453. Wang et al., 2019. Advances in Experimental Medicin and Biology 1117:215-240. Xhindoli et al., 2016. Biochimica et Biophysica Acta 1858:546-66. Zhan et al., 2014. Journal of Dental Research 93:639-44. Zhou et al., 2018. Journal of Nutritional Science and Vitaminology 64:307-315. Some figures need to be improved. Response: According to the 2 Reviewers’ valuable advice, Figure 1 was revised and Figure 2 was deleted. After the revision, the control in Figure 1 was set to 1.0. After the data transformation, the relative mRNA expression of CYP27B1 in gingival epithelia didn’t conform to the normal distribution. Thus, Figure 1B was revised as a box-plot. In addition, in the revised Figure 1, the PCR data with each data point were also presented. Figure 2 and 3 have been revised. Arrows have been added in the revised Figure 2, and scale bars have been added in the revised Figure 3. Figure 4 has also been revised, and CYP27B1 staining intensity data with each data point were added in the revised Figure 4A. Experimental design There are some questions regarding patient selection and tissue sampling. Response: The questions regarding patient selection and tissue sampling have been answered under “the primary concern” and “the second critical point”. Validity of the findings The number of replicates is not mentioned for some quantitative data. Response: We apologize for not mentioning the number of replicates. Real-time PCR was performed in triplicate. In the uploaded raw data with the file name “mRNA expression of CYP27B1”, “qpcrnum” indicated the serial number of each replicate. The second sentence in “Detection of CYP27B1 expression” in M&M has been revised as follows: “Real-time PCR reactions were performed using Faststart Universal SYBR Green Master Mix (Roche, Basel, Switzerland) in a real-time Thermocycler (Applied Biosystems, Foster City, CA, US) in triplicate.” Comments for the Author In the present study, the expression of CYP27B1 in the gingival soft tissue of periodontitis patients and healthy controls is compared. Vitamin D3 seems to play an essential role in periodontitis. Several previous studies of this group and other groups show that vitamin D3 could be locally converted into 25(OH)D3 and subsequently into bioactive 1,25(OH)2D3 by several cells of periodontal tissue.  The primary concern is the sampling of gingival tissue. In periodontitis patients, samples were taken from “part of the wall of periodontal pocket”, whereas, in the healthy subjects, the gingiva resected during crown lengthening procedure was used. First, how these anatomically different tissues could be compared? Second, the wall of the periodontal pocket could be the alveolar bone. Response: In order to analyze the typical inflammatory situation in vivo, all the patients enrolled were diagnosed with periodontitis Stage IV Grade C, the most severe periodontitis in the new classification scheme for periodontal diseases. In addition, all the participants in the periodontitis group had at least one unsalvageable tooth with mobility III and alveolar bone resorption close to the root apex. All unsalvageable teeth were extracted before periodontal treatment. All samples obtained in the periodontitis group were obtained around unsalvageable teeth, which had not received any periodontal therapy so that periodontal inflammation was severe enough and was not influenced by periodontal treatments. After extraction of the unsalvageable teeth, part of the wall of the periodontal pocket was resected and harvested. Since the remaining alveolar bone was close to the root apex (e.g. the tooth 25 shown in the following X-ray), we were able to harvest the coronal part of the wall of the periodontal pocket, and only gingiva but no alveolar bone was included. In the control group, the samples obtained were gingiva resected in the crown-lengthening surgery around periodontally healthy teeth. Thus, the samples in both groups were gingiva and could be compared. Figure 1 Example of unsalvageable teeth The second critical point is the control of serum vitamin D level or at least vitamin D supplementation by the study participants.  Response: In the present study, the serum vitamin D levels of the participants’ were not detected, however, we could make sure that none of the participants enrolled took vitamin D supplementation, since according to the Inclusion/Exclusion criteria, participants taking vitamin D supplementation would not be enrolled. The following sentence in Chinese is from the research plan submitted to the institutional review board of Peking University School and Hospital of Stomatology: “全身健康，无糖尿病、心血管疾病等全身系统病，女性非妊娠期、哺乳期或服用避孕药，未服用维生素D、碳酸钙、乳酸钙、葡萄糖酸钙等常见补充维生素D或钙的药物.” After translation into English, the sentence is “Systemically healthy, without systemic diseases such as diabetes or cardiovascular disease; women who are not pregnant, breastfeeding or taking contraceptives; not taking vitamin D, calcium carbonate, calcium lactate, calcium gluconate and other common vitamin D or calcium supplements.” Thus, the influence of vitamin D supplementation could be eliminated in the present study. Further comments Lines. 58-61. The conversion of vitamin D3 into 25(OH)D3 by the liver and subsequently into 1,25(OH)2D3 by the kidney is well known. However, the studies mentioned here (Liu et al. 2012a,b) refers only to the vitamin D3 conversion by gingival fibroblasts/periodontal ligament cells. Response: The comment is so valuable. The activation of vitamin D3 in the liver and the kidney should be mentioned. After revision, the second sentence of the Introduction has been revised as follows: “The active hormonal metabolite of vitamin D3, 1α,25-dihydroxyvitamin D3 (1,25OH2D3), is formed by two-step hydroxylations (Medrano et al., 2018): ① from vitamin D3 to 25-hydroxyvitamin D3 (25OHD3) by vitamin D 25-hydroxylase in the liver, followed by ② from 25OHD3 to 1,25OH2D3 by vitamin D 1α-hydroxylase in the kidney.” Line 62. 1-alpha hydroxylase CYP27B1 expression in the kidney was first detected before the study of Nykjaer et al. Response: Thank you so much for your reminding. 1α-hydroxylase CYP27B1 expression in the kidney was first reported in the study by Takeyama et al. in 1997. The reference in the manuscript has been revised accordingly. Line 85. No analysis of CYP27B1 expression in endothelial cells was performed. Endothelial cells should be counterstained with CD31 or von Willebrand factor. Periodontal inflammation is characterized by increased vascularization and, therefore, the increased number of endothelial cells. Thus, the possibility that the increased CYP27B1 expression is due to higher endothelial cell numbers should be excluded. Response: This is really a thoughtful comment. We really agree that periodontal inflammation is characterized by increased vascularization, and number of endothelial cells increased, which are also CYP27B1 expression positive. Thus, in the analysis of CYP27B1 mRNA expression in gingival connective tissues, the influence of endothelial cells did exist. The influence was difficult to exclude, because different cells could not be distinguished in RNA extraction. However, in the morphological analysis, hGFs and endothelial cells can easily be distinguished by pathologists, and endothelial cells did not influence the CYP27B1 staining scores of hGFs in the present study. The number of replicates should be shown for each experiment.  Response: We apologize for not mentioning the number of replicates. Real-time PCR was performed in triplicate. In the uploaded raw data with the file name “mRNA expression of CYP27B1”, “qpcrnum” indicated the serial number of each replicate. The second sentence in “Detection of CYP27B1 expression” in M&M has been revised as follows: “Real-time PCR reactions were performed using Faststart Universal SYBR Green Master Mix (Roche, Basel, Switzerland) in a real-time Thermocycler (Applied Biosystems, Foster City, CA, US) in triplicate.” In the calculation of staining scores of hGFs, each pathologist chose five non-overlapping 40× microscope fields of each section to exclude the influence of the differences among different microscope fields. In the uploaded raw data with the file name “Scope scores of CYP27B1 in hGFs”, the score of each field was shown. The mean scores of different microscope fields were used for analysis. Gene expression analysis was done only for some 17 periodontitis patients and 12 healthy controls. How were these patients selected? What were the demographic and clinical data of these patients?  Response: The gingival samples for analysis of RNA expression should be divided into connective tissues and epithelia. To avoid mutual contamination of the two kinds of tissues, only the surface layers could be obtained from the resected gingiva using sharp tissue scissors. But if the gingiva was too thin, it would be very difficult to finish the job, and such gingiva could not be used for analysis of gene expression. Thus, gene expression analysis was done only for 17 periodontitis patients and 12 healthy controls. Although samples for analysis of mRNA expression were fewer, the power value was over 0.99, indicating that the sample size was still large enough for the analysis. The demographic and clinical data of these participants were shown in the following table. Table 1 Demographic data and clinical parameters of the participants supplying samples for the detection of the mRNA expression of CYP27B1 Parameters Periodontitis (n=17) Controls (n=12) Age (years) 35.9 ± 7.8 31.9 ± 11.3 Gender (male/female) 7/10 9/3 PD (mm) 7.1 ± 0.4* 2.0 ± 0.6 AL (mm) 5.7 ± 0.4* 0 BI 4* 0 (0 to 0.5) BOP% 100%* 0 Data are presented as mean ± SD or median (lower to upper quartile) or number of subjects as indicated * Compared to the control group (p <0.05) Immunohistochemistry, Figure 3. The expression of CYP27B1 should be shown by arrows so that inexperienced readers can recognize it. Response: In the revised Figure 2 (the previous Figure 3), the expression of CYP27B1 in hGFs has been shown by arrows for the convenience of readers. Figure 3. Scale bars should be used instead of providing the magnifications because the pictures could be scaled during the publication process.  Response: In the revised Figure 2 (the previous Figure 3), scale bars have been used. Figure 4. Any scale bar is absent. Response: In the revised Figure 3 (the previous Figure 4), scale bars have been used. Fig 5A. What kind of plot is it? Is it a box-plot? In this case, the median and quartiles should be visible. Otherwise, some graphical presentations, for example, histogram, should be used.  Response: This figure is a box-plot. As shown in the raw data file uploaded with the name “Overall staining intensity of CYP27B1 in gingival connective tissues”, “intofct” was the intensity of CYP27B1 staining in gingival connective tissues, and the intensity was 1, 2 or 3. The maximum, P75, median, P25 and minimum in the periodontitis group was 3, 3, 3, 3 and 2, respectively. The maximum, P75, median, P25 and minimum in the control group was 3, 2, 1, 1 and 1, respectively. Thus, the figure didn’t look like a typical box-plot. To make the figure more readable, CYP27B1 staining intensity data with each data point were added in the revised Figure 4A. The revised Figure 4 is as follows. Discussion; lines 246-255. How could the increased level of 25(OH)D3 in the gingival crevicular fluid be related to the CYP27B1 expression? CYP 27B1 converts 25(OH)D3 into biologically active 1,25(OH)2D3 but is hardly connected to the local 25(OH)D3 levels.  Response: In the manuscript, the aim of lines 246-255 was explaining why the vitamin D pathway might be involved in periodontal immune defense. The second reason (after revision, this was the third reason) was: as an important part in the vitamin D pathway, 25OHD3 was associated with periodontal inflammation (Liu et al., 2010; Liu et al., 2009), indicating that the vitamin D pathway might be involved in periodontal immune defense. Thus, as the key factor in the vitamin D pathway, CYP27B1 was worth studying. Besides, according to our previous results (Gao et al., 2018), 25OHD3 could significantly induce the expression of CYP27B1 mRNA in hGFs. Thus, increased level of 25OHD3 in the gingival crevicular fluids might induce the expression of CYP27B1 in hGFs. This point has been added in the fourth paragraph of Discussion. The fourth paragraph of Discussion has been revised as follows: “Reasons for the higher expression of CYP27B1 in the periodontitis group might be as follows: (1) 25OHD3 was an up-regulator of CYP27B1 in hGFs (Gao et al., 2018), and 25OHD3 levels in gingival crevicular fluids of patients with periodontitis before initial periodontal therapy were significantly higher than those after therapy (Liu et al., 2010); (2) Periodontal inflammation could result in the higher concentrations of IL-1β and butyric acid in gingival crevicular fluids (Liu et al., 2010; Lu et al., 2014), which could also induce the expression of CYP27B1 in hGFs (Liu et al., 2012b).” Most existing literature suggests that vitamin D deficiency is associated with an increased risk of periodontal disease. This fact should be discussed in connection with the obtained data. Response: Our previous studies (Liu et al., 2009; Liu et al., 2010) indicated that the systemic and local 25OHD3 levels in patients with aggressive periodontitis were positively associated with periodontal inflammation. However, several existing studies (Dietrich et al., 2004; Jimenez et al., 2014; Zhan et al., 2014) suggested that vitamin D deficiency is associated with an increased risk of periodontal disease. What should be pointed out is that in these studies, the participants were about 50 years or older, an age range that didn’t overlap with that of the population in our previous studies (younger than 30 years old). Additionally, no correlation between plasma 25OHD3 levels and periodontal health was found in another large cross-sectional study (Antonoglou et al., 2015), and the participants in the study was 30-49 years. Thus, the relationship between 25OHD3 and periodontitis in people of different ages might be different. In studies investigating the association between 25OHD3 and periodontal health in large samples (Dietrich et al., 2004; Jimenez et al., 2014; Zhan et al., 2014; Antonoglou et al., 2015), the participants were from the general population. However, in our previous study (Liu et al., 2009), only patients with aggressive periodontitis had higher plasma 25OHD3 levels, and the patients had much more severe periodontal inflammation than the other participants. In the special group, the question of whether the higher plasma 25OHD3 level is the reason or the result of severe periodontal inflammation is unclear. Our previous study (Gao et al., 2018) suggested that 25OHD3 could activate the vitamin D pathway, which participates in periodontal immune defense. Therefore, there is the following possibility: due to severe periodontal inflammation, more LL37 is needed for antibacterial and anti-inflammatory function, and more 25OHD3 is synthesized for the more active vitamin D pathway in periodontium. This possibility could help to explain why patients with severe periodontitis had higher systemic and local 25OHD3 levels. In the present study, the results that patients with periodontitis had higher CYP27B1 expression in hGFs indicated that patients with periodontitis had more active vitamin D pathway in hGFs, which further supported this possibility. These 2 paragraphs have been added in the revised manuscript as the 5th and 6th paragraphs in Discussion. Thank you again and we hope that the revision can meet the reviewers’ requirements. We are looking forward to your reply. Yours sincerely, Huanxin Meng, Kaining Liu Department of Periodontology Peking University School and Hospital of Stomatology、National Engineering Laboratory for Digital and Material Technology of Stomatology、Beijing Key Laboratory of Digital Stomatology on behalf of all authors "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Previous studies have shown that RNA Polymerase III Subunit G (POLR3G) has oncogenic effects in cultured cells and mice. However, the role of POLR3G in transitional cell carcinoma (TCC) has not been reported. This study explores the potential of POLR3G as a novel molecular marker for TCC. Methods. The RNA sequencing data and clinical information of patients with TCC were downloaded from The Cancer Genome Atlas official website. Transcriptome analysis was performed as implemented in the edgeR package to explore whether POLR3G was up-regulated in TCC tissues compared to normal bladder tissues. The expression of POLR3G in bladder cancer cell line T24 and human uroepithelial cell line SV-HUC-1 were detected via quantitative real time polymerase chain reaction (qRT-PCR). Correlations between POLR3G expression and clinicopathological characteristics were analyzed using Mann-Whitney U test or Kruskal-Wallis H test.</ns0:p><ns0:p>Clinicopathological characteristics associated with overall survival were explored using the Kaplan-Meier method and Cox regression analyses. Gene set enrichment analysis (GSEA) was performed to explore the associated gene sets enriched in different POLR3G expression phenotypes and the online tool Tumor IMmune Estimation Resource (TIMER) was used to explore the correlation between POLR3G expression and tumor immune infiltration in TCC. Results. Transcriptome analysis showed that POLR3G was significantly up-regulated in TCC tissues compared to normal bladder tissues. Furthermore, qRT-PCR revealed high expression of POLR3G in T24 cells compared to SV-HUC-1 cells. Overall, POLR3G expression was associated with race, tumor status, tumor subtype, T classification, and pathological stage. Kaplan-Meier survival analysis revealed that higher POLR3G expression was associated with lower overall survival. The univariate Cox regression model revealed that age at diagnosis, pathological stage, and POLR3G expression were associated with prognosis of TCC patients. Further multivariate analyses identified these three clinicopathological characteristics as independent prognostic factors for overall survival. GSEA analysis showed that several gene sets associated with tumor</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Bladder cancer is the 7th most commonly diagnosed cancer in males and the 11th most commonly diagnosed cancer when both genders are considered <ns0:ref type='bibr' target='#b11'>(Ferlay et al., 2013)</ns0:ref>. In 2020, an estimated 81,400 new cases of bladder cancer (62,100 men; 19,300 women) will be diagnosed in the United States of America and approximately 17,980 deaths (13,050 men; 4,930 women) will occur during the same period of time <ns0:ref type='bibr' target='#b29'>(Siegel et al., 2020)</ns0:ref>. Transitional cell carcinoma (TCC) is the most common histological type of bladder cancer, contributing to more than 90% of all bladder cancer cases <ns0:ref type='bibr' target='#b38'>(Witjes et al., 2020)</ns0:ref>. Approximately 75% of patients with bladder cancer are present with non-muscle-invasive bladder cancer (NMIBC) at the initial diagnosis, while the remaining 25% of patients are present with muscle-invasive bladder cancer (MIBC) or metastatic bladder cancer <ns0:ref type='bibr' target='#b38'>(Witjes et al., 2020)</ns0:ref>. The standard treatment for NMIBC is trans-urethral resection of bladder tumor (TURBT) followed by intravesical chemotherapy or bacillus Calmette-Gu&#233;rin (BCG) immunotherapy depending on risk stratifications. For MIBC, on the other hand, neoadjuvant chemotherapies followed by radical cystectomy are first line recommendations. However, the prognosis for MIBC is poor even with effective treatments. The five-year recurrence-free survival rate was 89% for patients with T2 tumors, 50% for patients with T4 tumors, and 35% for patients with lymph node metastasis respectively <ns0:ref type='bibr' target='#b31'>(Stein et al., 2001)</ns0:ref>. Patients with NMIBCs have better prognoses, however, high grade NMIBC has a 70% recurrence rate with a 15%-40% risk of progression after five years (Kashif <ns0:ref type='bibr' target='#b14'>Khan et al., 2014)</ns0:ref>.</ns0:p><ns0:p>For TCC, the most important histopathological prognostic variables are tumor stage and lymph node status <ns0:ref type='bibr' target='#b10'>(Dutta et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b31'>Stein et al., 2001)</ns0:ref>. However, no predictive molecular markers are routinely used in clinical practice. Thus, identifying effective markers is essential for predicting prognoses and directing treatments for patients with TCC. Although previous studies have revealed that RNA Polymerase III Subunit G (POLR3G) overexpression can have oncogenic effects in cultured cells and mice <ns0:ref type='bibr' target='#b13'>(Haurie et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b15'>Khattar et al., 2016)</ns0:ref>, its role in TCC has still not been reported. Herein, the aim of this study is to evaluate the correlation between POLR3G and prognoses of patients with TCC.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data acquisition</ns0:head><ns0:p>The RNA sequencing (RNA-seq) data (Workflow types: HTSeq-FPKM; HTSeq-Counts) and corresponding clinical information of patients with TCC (Project: TCGA-BLCA, Disease type: transitional cell papillomas and carcinomas) were downloaded from GDC Data Portal (https://portal.gdc.cancer.gov/). Counts data were used for transcriptome analysis. TPM data were calculated based on FPKM data, and were used for further analysis. Only patients with both RNA-seq data and survival information were included in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Cell lines and cell culture</ns0:head><ns0:p>Human bladder cancer cell line T24 and immortalized normal urothelial cell line SV-HUC-1 were purchased from National Infrastructure of Cell Line Resource (Beijing, China). Cell lines were maintained in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco) supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Gibco). All cells were maintained in a humidified atmosphere with 5% CO2 at 37 &#176;C.</ns0:p></ns0:div> <ns0:div><ns0:head>RNA extraction and qRT-PCR</ns0:head><ns0:p>Total RNA was extracted from cells using the RNA simple Total RNA Kit (Tiangen). FastQuant RT Kit (Tiangen) was used for cDNA synthesis. The quantitative real time polymerase chain reactions (qRT-PCR) were performed using KAPA SYBR FAST Universal q-PCR Kit (KAPA). The relative mRNA levels of genes were calculated using cycle threshold (CT) methods, and &#946;actin was used as an endogenous control. Three replicate samples were studied for detection of mRNA expression. The primers were listed below: POLR3G (forward): 5'-CGCAGGCAAAGGCACAC-3'; POLR3G (reverse): 5'-CCTCTTTTTTCCAATTCCTCCA-3'; &#946;-actin (forward): 5'-CCAACCGCGAGAAGATGA-3'; &#946;-actin (reverse): 5'-CCAGAGGCGTACAGGGATAG-3'.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene set enrichment analysis</ns0:head><ns0:p>Gene set enrichment analysis (GSEA) is widely applied to determine whether predefined gene sets are differentially expressed in different phenotypes. <ns0:ref type='bibr' target='#b32'>(Subramanian et al., 2005)</ns0:ref> To identify signaling pathways that are differentially activated in TCCs, we conducted GSEA analysis between high and low POLR3G expression groups. GSEA analysis was performed using GSEA software (version 4.0.3). The h.all.v7.1.symbols.gmt (hallmark) dataset was obtained from the Molecular Signatures Database (MsigDB) <ns0:ref type='bibr' target='#b19'>(Liberzon et al., 2015)</ns0:ref>. Enrichment analysis was performed by default weighted enrichment statistics, with the random combinatorial count set as 1,000. Gene sets were judged as significantly enriched by P &lt; 0.05 as well as false discovery rates (FDR) &lt; 0.25.</ns0:p></ns0:div> <ns0:div><ns0:head>Tumor infiltrating immune cells and immune checkpoint molecule expression analysis</ns0:head><ns0:p>Tumor IMmune Estimation Resource (TIMER) is a web server for comprehensive analysis of tumor-infiltrating immune cells, which applies a previously published statistical approach called the deconvolution that uses the gene expression profiles to produce an inference on the number of tumor-infiltrating immune cells <ns0:ref type='bibr' target='#b16'>(Li et al., 2016)</ns0:ref>. Survival module was used to explore the association between immune cell infiltration (B cells, CD4+ T cells, CD8+ T cells, neutrophils, macrophages, and dendritic cells) and clinical outcome in bladder cancer. Gene module was used to explore the correlations between the expression of POLR3G and immune cell infiltration. Correlation module was used to explore the correlations between POLR3G expression and immune checkpoint molecule expression, including PDCD1 (also known as PD1), CD274 (also known as PD-L1), PDCD1LG2 (also known as PD-L2), CTLA4, LAG3, HAVCR2, and TIGIT.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>The data processing and further statistical analyses were performed using R (v.3.6.3). Transcriptome analysis of the differentially expressed genes (DEGs) in the TCC tissues and normal bladder tissues was performed using edgeR package <ns0:ref type='bibr' target='#b27'>(Robinson et al., 2010)</ns0:ref>. The differential expression of POLR3G between T24 cells and SV-HUC-1 cells was analyzed via independent t test. The relationships between clinicopathological characteristics and POLR3G expression were analyzed using the Mann-Whitney U test or Kruskal-Wallis H test. The Kaplan-Meier method and Cox regression models were used to explore the influence of POLR3G expression on overall survival along with other clinicopathological characteristics (age at diagnosis, gender, race, and pathological stage). The cut-off value of POLR3G expression was determined by its median value. P &lt; 0.05 was considered to indicate a statistically significant difference.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Clinical characteristics</ns0:head><ns0:p>The clinical characteristics of 404 patients with TCC in The Cancer Genome Atlas (TCGA) are presented in Table <ns0:ref type='table'>1</ns0:ref>. There were 299 (74.01%) male patients and 105 (25.99%) female patients with a median age of 69 years old at the time of diagnosis. Race information was available for 388 patients, among which 322 (82.99%) patients were white, 43 (11.08%) were Asian, and 23 (5.93%) were black or African American. At least 202 (63.72%) patients were tumor free while 115 (36.28%) patients still had tumors. Overall, 2 (0.50%) of the patients showed stage I, 129 (32.09%) stage II, 137 (34.08%) stage III, and 134 (33.33%) stage IV. Among 363 patients diagnosed with clear N stage, 129 (35.54%) had lymph node metastases. From 205 patients diagnosed with clear M stage, 11 (5.37%) had distant metastases. Median follow-up for subjects alive at last contact was 21.3 months (range 0-169 months).</ns0:p></ns0:div> <ns0:div><ns0:head>POLR3G was up-regulated in multiple cancer types including TCC</ns0:head><ns0:p>TIMER analyses revealed that POLR3G expression was up-regulated in multiple cancer types, including cholangiocarcinoma, colon adenocarcinoma, esophageal carcinoma, kidney renal clear cell carcinoma. (Fig. <ns0:ref type='figure' target='#fig_0'>1A</ns0:ref>). RNA-Seq differential expression analysis revealed that 2,211 genes were up-regulated (log2FC &gt; 1, and FDR &lt; 0.01) in TCC tissues compared to normal bladder tissues, including POLR3G (log2FC = 1.038, FDR = 0.006), and 1,853 genes were downregulated (log2FC &lt; -1, and FDR &lt; 0.01). (Fig. <ns0:ref type='figure' target='#fig_0'>1B</ns0:ref> and Table <ns0:ref type='table'>S1</ns0:ref>). Furthermore, qRT-PCR results showed higher expression levels of POLR3G in T24 cells compared to SV-HUC-1 cells (P = 0.004; Fig. <ns0:ref type='figure' target='#fig_0'>1C</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Relationship between POLR3G expression and clinicopathological characteristics</ns0:head><ns0:p>Expression levels of POLR3G were strongly correlated with T classification (T1-2 vs. T3-4, P = 0.001; Fig. <ns0:ref type='figure'>2A</ns0:ref>), pathological stage (stage I-II vs. stage III-IV, P = 0.005; Fig. <ns0:ref type='figure'>2B</ns0:ref>), tumor status (tumor free vs. with tumor, P = 0.001; Fig. <ns0:ref type='figure'>2C</ns0:ref>), tumor subtype (papillary vs. non-papillary, P &lt; 0.001; Fig. <ns0:ref type='figure'>2D</ns0:ref>), and race (P &lt; 0.001; Fig. <ns0:ref type='figure'>2E</ns0:ref>). No statistically significant differences were observed between groups stratified by age (&#8804; 60 years old vs. &gt; 60 years old, P = 0.113; Fig. <ns0:ref type='figure'>2F</ns0:ref>), gender (female vs. male, P = 0.072; Fig. <ns0:ref type='figure'>2G</ns0:ref>), lymph node metastasis (positive vs. negative, P = 0.201; Fig. <ns0:ref type='figure'>2H</ns0:ref>), and distant metastasis (positive vs. negative, P = 0.056; Fig. <ns0:ref type='figure'>2I</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Survival outcomes and Cox regression analysis</ns0:head><ns0:p>Kaplan-Meier survival analysis revealed that higher expression of POLR3G was associated with worse prognoses (P = 0.002; Fig. <ns0:ref type='figure' target='#fig_0'>1D</ns0:ref>). The univariate Cox regression model revealed that age at diagnosis (HR = 1.03, 95% CI = 1.02 to 1.05, P &lt; 0.001), pathological stage (HR = 1.71, 95% CI = 1.41 to 2.08, P &lt; 0.001), and POLR3G expression (HR = 1.04, 95% CI = 1.01 to 1.07, P = 0.02) were associated with overall survival of patients with TCC. Furthermore, multivariate Cox regression after adjustment indicated that age at diagnosis (HR = 1.03, 95% CI = 1.02 to 1.05, P &lt; 0.001), pathological stage (HR = 1.77, 95% CI = 1.45 to 2.17, P &lt; 0.001), and POLR3G expression (HR = 1.05, 95% CI = 1.02 to 1.08, P = 0.001) were independent prognostic factors for overall survival in patients with TCC (Table 2, Fig. <ns0:ref type='figure'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gene sets enriched in POLR3G high expression phenotype</ns0:head><ns0:p>In the hallmark dataset, 41 gene sets were significantly enriched in POLR3G high expression phenotype (Table <ns0:ref type='table'>3</ns0:ref>). Several of the gene sets are associated with oncogenesis, progression, and metastasis of cancer, such as mitotic spindle, hypoxia, Kras signaling up, PI3K-AKT-mTOR signaling, IL6-JAK-STATS3 signaling, mTORC1 signaling, TNF-&#61537; signaling via NF-&#61547;B, inflammatory response, and Myc targets v1(Fig. <ns0:ref type='figure'>4A-I</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>POLR3G expression was associated with levels of immune cell infiltration and immune checkpoint molecule expression</ns0:head><ns0:p>Analysis of TIMER survival module indicated that the infiltration of CD8+ T cells is related to the cumulative survival rate in TCC (P = 0.006, Fig. <ns0:ref type='figure'>5A</ns0:ref>). Gene module analysis revealed that POLR3G expression was negatively correlated with tumor purity (cor = &#8722;0.178, P = 6.02e&#8722;04) and positively correlated with infiltrating levels of CD8+ T cell (cor = 0.317, P = 5.70e&#8722;10), neutrophil cells (cor = 0.237, P = 5.12e&#8722;06), and dendritic cells (cor = 0.399, P = 2.15e&#8722;15) in TCC (Fig. <ns0:ref type='figure'>5B</ns0:ref>). Moreover, Correlation module analysis revealed that the expression of POLR3G was significantly correlated with the expression of immune checkpoint molecules including PDCD1 (cor = 0.227, P = 3.57e&#8722;06), CD274 (cor = 0.455, P = 3.18e&#8722;22), PDCD1LG2 (cor = 0.399, P = 4.60e&#8722;17), CTLA-4(cor = 0.279, P = 9.87e&#8722;09), LAG3 (cor = 0.344, P = 8.32e&#8722;13 ), HAVCR2 (cor = 0.312, P = 1.18e&#8722;10 ), and TIGIT(cor = 0.255, P = 1.78e&#8722;07) (Fig. <ns0:ref type='figure'>5C</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Several studies have investigated prognostic biomarkers for TCC, such as FGF2 <ns0:ref type='bibr' target='#b28'>(Shariat et al., 2010)</ns0:ref>, UHRF1 <ns0:ref type='bibr' target='#b36'>(Unoki et al., 2009)</ns0:ref>, and GRIA1 <ns0:ref type='bibr' target='#b34'>(Tilley et al., 2017)</ns0:ref>. However, there are still no ideal predictive molecule for clinical application. This study demonstrates that POLR3G is a potentially useful biomarker for predicting prognosis of TCC.</ns0:p><ns0:p>POLR3G is an RNA polymerase III peripheral subunit that synthesizes small RNAs, such as 5S rRNA, tRNAs, and some microRNAs <ns0:ref type='bibr' target='#b13'>(Haurie et al., 2010)</ns0:ref>. POLR3G plays a role in sensing and limiting infection by intracellular bacteria and DNA viruses, acts as a nuclear and cytosolic DNA sensor involved in innate immune responses, and is also essential for the maintenance of stem cell state <ns0:ref type='bibr' target='#b0'>(Ablasser et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chiu et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b20'>Lund et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b39'>Wong et al., 2011)</ns0:ref>. Several studies have described the links between POLR3G and cancer. For example, Durrieu-Gaillard et al. reported that POLR3G expression was strongly up-regulated during the process of tumoral transformation in the human lung fibroblast cell line IMR90 model system <ns0:ref type='bibr' target='#b9'>(Durrieu-Gaillard et al., 2018)</ns0:ref>. <ns0:ref type='bibr'>Haurie et al.</ns0:ref> showed that overexpression of POLR3G in IMR90 increased the expression of genes associated with tumor growth and metastasis, including S100A4, RFC2, EZR, and RAC1, and reduced the expression of tumor-suppressing genes, such as PFDN5 and KLF6 <ns0:ref type='bibr' target='#b13'>(Haurie et al., 2010)</ns0:ref>. Another study <ns0:ref type='bibr' target='#b24'>(Petrie et al., 2019)</ns0:ref> found that the expression of POLR3G was up to three-fold higher in prostate tumors compared to normal adjacent samples. Similar results were observed at the cellular level; POLR3G expression was elevated in the prostate cancer cell line PC-3 compared to the immortalized healthy prostate epithelium cell line PNT2C2. In addition, knockdown of POLR3G triggered the proliferative arrest of PC-3.</ns0:p><ns0:p>Results of this study revealed that POLR3G was highly expressed in multiple cancer types, including TCC, and qRT-PCR further confirmed that POLR3G was elevated in T24 cells compared to SV-HUC-1 cells. These findings are indicative of a cumulative alteration of POLR3G expression during TCC tumorigenesis. The traditional perspective of TCC tumorigenesis postulates that TCCs arise via two different but overlapping pathways: papillary pathway and non-papillary pathway <ns0:ref type='bibr' target='#b8'>(Dinney et al., 2004)</ns0:ref>. We found that non-papillary TCCs exhibited higher POLR3G expression compared to papillary TCCs. Therefore, POLR3G might play different roles in these two pathways. We also found that high POLR3G expression was positively correlated with high T classification, advanced clinical stage, and tumor recurrence, which are strongly correlated with poor prognosis in patients with TCC. More importantly, further univariate and multivariate analysis identified POLR3G expression as an independent prognostic factor for overall survival.</ns0:p><ns0:p>We conducted GSEA analysis to investigate the relationship between POLR3G and gene signatures in TCCs. Our results showed that 41 gene sets were significantly enriched in the POLR3G high expression group, including mitotic spindle, Inflammatory response, TGF-&#946; signaling, epithelial mesenchymal transition, PI3K-AKT-mTOR signaling, and IL-6-JAK-STATS3 signaling. Several of these pathways are associated with oncogenesis, progression, and metastasis of cancer, suggesting that POLR3G expression contributes to the development, progression, and prognosis of TCC. However, the regulatory mechanism needs to be further elucidated.</ns0:p><ns0:p>Immunotherapy is a key treatment approach for TCC. Intravesical BCG immunotherapy has been used to treat superficial TCC for over 40 years and still represents the first-line adjuvant treatment for superficial TCC after TURBT to prevent tumor recurrence <ns0:ref type='bibr' target='#b1'>(Babjuk et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b22'>Morales et al., 1976)</ns0:ref>. Over the past decade, immune checkpoint inhibitor (ICI) immunotherapy breakthroughs have enriched the available treatment modalities for advanced TCCs. Atezolizumab and Pembrolizumab have been approved for first-line systemic therapy for cisplatin-ineligible patients with local advanced or metastatic TCC whose tumors express PD-L1 <ns0:ref type='bibr' target='#b2'>(Balar et al., 2017a;</ns0:ref><ns0:ref type='bibr' target='#b4'>Balar et al., 2017b)</ns0:ref>. However, the objective response rate (ORR) to ICIs in bladder cancer patients was only ~20% <ns0:ref type='bibr' target='#b2'>(Balar et al., 2017a;</ns0:ref><ns0:ref type='bibr' target='#b4'>Balar et al., 2017b)</ns0:ref>. Thus, identifying reliable biomarkers to distinguish which patients are more likely to respond to ICI immunotherapy is crucial for the successful treatment. Previous studies have demonstrated that the level of the immune infiltration within tumors correlates with bladder cancer prognosis and is a positive prognostic indicator of response to immunotherapy <ns0:ref type='bibr' target='#b12'>(Fridman et al., 2012;</ns0:ref><ns0:ref type='bibr'>Pfannstiel et al., 2019)</ns0:ref>. Furthermore, the expression of immune checkpoint PD-L1 on tumors correlates with unfavorable prognosis, but can also predict the immunotherapy reactivity of patients <ns0:ref type='bibr' target='#b33'>(Thompson et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b35'>Topalian et al., 2015)</ns0:ref>. Therefore, we explored the potential role of POLR3G in immune cell infiltration within TCCs using TIMER.</ns0:p><ns0:p>First, we used the survival module to explore the association between immune infiltrate abundance and clinical outcome. Previous studies have reported that CD8+ T cells infiltration might play a positive role in the prognosis of colorectal cancer <ns0:ref type='bibr' target='#b23'>(Naito et al., 1998)</ns0:ref>, triplenegative breast cancer <ns0:ref type='bibr' target='#b37'>(Vihervuori et al., 2019)</ns0:ref>, and pancreatic cancer <ns0:ref type='bibr' target='#b21'>(Masugi et al., 2019)</ns0:ref>. However, our results suggest that CD8+ T cell infiltration was negatively correlated with cumulative survival in TCC. Second, we used the gene module to explore the correlation between POLR3G expression and immune infiltrate abundance. Results from this analysis showed that POLR3G expression was significantly correlated with the level of infiltrating immune cells in TCC. More specifically, POLR3G expression was negatively correlated with tumor purity, and positively correlated with the infiltrating levels of CD8+ T cells, neutrophil cells, and dendritic cells in TCC. We further explored the correlations between POLR3G and immune checkpoint molecules in TCC via the correlation module. Results revealed that POLR3G expression was significantly correlated with several immune checkpoint molecules, including PDCD1, CD274, PDCD1LG2, CTLA4, LAG3, HAVCR2, and TIGIT. Taken together, these findings suggest that POLR3G contributes to the regulation of immune cell infiltration and immune checkpoint molecule expression, resulting in the suppression of anti-tumor immunity. These results provide a possible mechanistic explanation for the worse prognosis observed in patients with higher POLR3G expression.</ns0:p><ns0:p>To the best of our knowledge, this is the first study investigating the role of POLR3G in TCC. We found that POLR3G expression was an independent prognostic factor for overall survival and can potentially be used as a prognostic biomarker in TCC. However, there were some limitations to this study. First, this study was conducted using data from the public database TCGA, and the clinical information was incomplete for some patients. Further investigation with a larger sample size is needed to validate our findings. Second, the relationships between POLR3G and immune cell infiltration were analyzed using online tools, which need to be further elucidated via molecular experiments.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, POLR3G expression was up-regulated in TCC and can potentially be used as a prognostic marker. In addition, the expression of POLR3G was associated with levels of immune cell infiltration and the expression of immune checkpoint molecules in TCC, suggesting potential value for predicting patient response to ICI immunotherapy. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 POLR3G</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,42.52,199.12,525.00,306.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,275.62,525.00,279.00' type='bitmap' /></ns0:figure> <ns0:note place='foot' n='1'>PeerJ reviewing PDF | (2020:06:50549:1:1:NEW 21 Sep 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Dear Professor Verger, We thank you for giving us this opportunity to revise our manuscript, and thank the reviewers for their constructive comments and valuable suggestions on our manuscript entitled 'Increased expression of POLR3G predicts poor prognosis in transitional cell carcinoma' (ID: 50549). We have studied the reviewer’s comments carefully and made revisions according to the comments. We would like to resubmit the revised version for your kindly consideration. Look forward to hearing from you soon. Thank you and best regards. Yours sincerely, Xianhui Liu Response to Reviewer 1:  Basic reporting Overall manuscript is looking good, still need to improve the English. Some sentences are unclear. Reply: Thank you for your positive comments on our manuscript. We have rephrased the sentences those were not expressed clearly and had our manuscript polished with the help of editing service. Experimental design Overall experimental and statistical analysis is good. Validity of the findings no comment Comments for the author The manuscript 'Increased expression of POLR3G predicts poor prognosis in transitional cell carcinoma' is well written by Liu et al. I have few comments: 1) 'Data acquisition' section is too short and unclear. Reply: We rewrote the “Data acquisition” section to be clearer and more specific as followed: The RNA sequencing (RNA-seq) data (Workflow types: HTSeq-FPKM; HTSeq-Counts) and corresponding clinical information of patients with TCC (Project: TCGA-BLCA, Disease type: transitional cell papillomas and carcinomas) were downloaded from GDC Data Portal (https://portal.gdc.cancer.gov/). Counts data were used for transcriptome analysis. TPM data were calculated based on FPKM data, and were used for further analysis. Only patients with both RNA-seq data and survival information were included in this study. 2) capitalize the 'cell lines and cell culture '. Reply: Corrected. 3) First sentence of the 'Gene set enrichment analysis' is similarity with the GSEA-MSigDB GitHub, reframe the sentence. Reply: The sentence was reframed as followed: Gene set enrichment analysis (GSEA) is widely applied to determine whether predefined gene sets are differentially expressed in different phenotypes. 4) 'Statistical downloaded from TCGA were merged and conducted using R (v.3.6.3). '; sentence is unclear. Reply: We apologize that we didn’t demonstrate this clearly. We downloaded the RNA-seq data and clinical data from GDC data portal(https://portal.gdc.cancer.gov/), where the RNA-seq data or clinical data of each patient was uploaded as an individual file. Therefore, we firstly used R (v.3.6.3) to merge the separate RNA-seq data or clinical data of all TCC patients for further analysis. We reframed this sentence as followed: The data processing and further statistical analysis were performed using R (v.3.6.3). 5) Why they used TPM in Figure 1a and FPKM in Figure 1c? ： Reply: Thank you for pointing out the inconsistency in our manuscript. Figure 1A was analyzed by the online tool TIMER based on TPM data, Figure 1B (According to our manuscript, we think the reviewer actually meant Figure 1B instead of Figure 1C) was initially analyzed based on FPKM data downloaded from TCGA website. In the revised manuscript, we calculated TPM data based on FPKM data firstly, and then performed further analysis using TPM data. Therefore, the results in the revised manuscript were slightly different from the original manuscript, including: (1) P value of KM analysis. (2) P values in the analysis of the associations between POLR3G expression and clinicopathological characteristics. (3) P values and HR values in univariate and multivariate Cox regression analyses. (4) The results of GSEA analysis. But the conclusions were basically the same. Figures and Tables were updated and all the revisions were marked red in the revised manuscript with tracked changes. In addition, the other reviewer suggested that DESeq2 or edgeR are typically better for dealing with count data coming from RNA-seq experiments. We replaced Figure 1B by the results of transcriptome analysis using edgeR package, which also indicated POLR3G was significantly up-regulated in TCC tissues compared to normal bladder tissues (log2FC = 1.038, FDR = 0.006). Detailed results were added in Table S1. 6) I didn't find Fig. 5. in the manuscript. Reply: We apologize for referring to Fig. 5A-C as Fig. 3A-C in the original manuscript (Line 204, line 208, and line 212) mistakenly and has corrected these typos in the revised manuscript with tracked changes. 7) Need professional editing. Reply: We have rephrased the sentences those were not expressed clearly and had our manuscript polished. Response to Reviewer 2:  Basic reporting The structure of the manuscript is suitable and contains all necessary sections (Abstract, introduction, methods, results, discussion and references). Some raw data are shared, although the tables provided are not referred in the text, not labelled and potentially problematic (all 'standard deviations' are zero?). Reply: Thank you for your valuable comments concerning on our manuscript. We apologize for the confusions. As PeerJ staff required, we provided the raw data of qRT-PCR only for editor and reviewers to review, not for publication. Therefore, the data were not referred in the text. In the raw data of qRT-PCR, each row demonstrated the result of one single test of each sample. Thus, the Cq mean was equal to Cq value and the Cq standard deviation was zero in all rows. We performed three independent qRT-PCR experiments, and run three duplicate tests of each sample in each experiment. The results of relative POLR3G expression in T24 cells and SV-HUC-1 cells were presented by means with standard deviations in Fig.1C. To avoid confusions to editor and reviewers, we re-edited the supplemental data as followed. The figures need to be better labelled, with clear legends (for instance, explicit what is blue and what is red in Figure 1A. Reply: Thank you for pointing out the deficiencies of figures included in the manuscript. We have re-edited the figures with clear legends in the revised manuscript. Few English errors or typos: line 52-53 'may of great value' should be 'may be of great value' in multiple places, 'There’re' should be replaced by 'There are'. Reply: Corrected. Experimental design The authors of the manuscript investigated the expression level of POLR3G in the TCGA dataset of Transitional cell carcinoma. They identified systematic overexpression of the gene in tumors samples compared to normal samples and a potential association of POLR3G level with survival. They then explored the association of POLR3G expression with clinicopathologic characteristics, with gene sets (using GSEA) and with immune infiltration (using TIMER). The importance of POLR3G overexpression might be overstated here, because the test used may not be suitable for this context (DESeq2 or edgeR are typically better for dealing with count data coming from RNASeq experiments). The authors do not indicate how many genes would be 'overexpressed' with the test they used, or how much overexpression there is in tumor samples. In addition, it is not surprising that many genes are differentially expressed in tumors compared to adjacent normal cells, because of all the molecular processes occurring in tumors and the different cell composition in the samples. Reply: We agree that DESeq2 or edgeR are typically better for dealing with count data coming from RNA-seq experiments. We downloaded the count data of TCC patients from TCGA website and performed transcriptome analysis using edgeR package. Results showed that 2211 genes were up-regulated (log2FC > 1, and FDR < 0.01) in TCC tissues compared to normal bladder tissues, including POLR3G (log2FC = 1.038, FDR = 0.006), and 1853 genes were down-regulated (log2FC < -1, and FDR < 0.01). We have added the result in the revised manuscript with tracked changes (Figure 1B and Table S1). We also agree that the importance of POLR3G overexpression might be overstated here, so more molecular experiments will be performed to validate the function of POLR3G. Validity of the findings The results are presented with adequate support. Conclusions are derived from the results, and limitations of the study are indicated. Thank you once again. We studied your comments and suggestions carefully and made revisions which we hope meet with your approval. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. In order to regulate the water flow hydraulic structures such as weirs or checks, frequently equipped with gates, are used. Water can flow below or over the gate or, simultaneously, over and below the gate. Both diversifications of hydraulic gradient, being an effect of damming up a river by the structure and shear stresses at the bed, which exceeds the critical shear stress value, invoke the local scouring downstream the structure. This phenomenon has been studied in laboratory and field conditions for many years, however Researchers do not agree on the parameters that affect the size of the local scour and the intensity of its formation. There are no universal methods for estimating its magnitude However, solutions are sought in the form of calculation formulas typical for the method of flow through the structure, taking into account the parameters that characterize a given structure. These formulas are based on factors that affect the size of the local scours, that is, their dimensions and location. Examples of such formulas are those contained in this article: Franke (1960), <ns0:ref type='bibr' target='#b54'>Straube (1963</ns0:ref><ns0:ref type='bibr' target='#b61'>), Tarajmovi&#269; (1966)</ns0:ref>, <ns0:ref type='bibr' target='#b50'>Rossinski &amp; Kuzmin (1969)</ns0:ref> equations. The need to study this phenomenon results from the prevalence of hydrotechnical structures equipped with gates (from small gated checks to large weirs) and from potential damage that may be associated with excessive development of local erosion downstream, including washing of foundations and, consequently, loss of stability of the structure.</ns0:p><ns0:p>Methods. This study verifies empirical formulas applied to estimate the geometry parameters of a scour hole on a laboratory model of a structure where water is conducted downstream the gate with bottom reinforcements of various roughness. A specially designed remote-controlled measuring device, equipped with laser scanner, was applied to determine the shape of the sandy bottom. Then the formula optimization is conducted, using Monte Carlo sampling method, followed by verification of field conditions.</ns0:p><ns0:p>Results. The suitability of a specially designed device, equipped with laser scanner for measuring the bottom shape in laboratory conditions was demonstrated. Simple formula describing local scour geometry in laboratory conditions was derived basing on the Straube formula. The optimized formula was verified in field conditions giving very good comparative results. Therefore, it can be applied in engineering and designing practices.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The article addresses the subject of hydraulic structures, understood as structures for water management, shaping water resources and water use <ns0:ref type='bibr' target='#b11'>(Chen 2015)</ns0:ref>. The analyzed case is a damming structure i.e. a structure enabling permanent or periodical maintenance of a water surface level elevation above the adjacent land or body of water. There is a strong link between water and environment, as well as between water and development, both with opposite aims. While the willingness to develop and expand the hydrotechnical infrastructure is a drive for economic development and thus for urbanization to flourish, the environmental aspects direct the designer rather towards sustainable solutions <ns0:ref type='bibr' target='#b31'>(Koskinen, Leino &amp; Riipinen 2008;</ns0:ref><ns0:ref type='bibr' target='#b22'>Jordaan 2009;</ns0:ref><ns0:ref type='bibr'>Rasekh, Afshar &amp; Afshar 2010)</ns0:ref> and, as engineering practice demonstrated, to reject a project at an early stage. It should be denoted that a failure of a hydrotechnical structure does not only concern dams of significant sizes but also small hydraulic structures (usually not taken into account by ecologists) which are a potential source of disaster, affecting the life and economy of a given region <ns0:ref type='bibr' target='#b20'>(Hossain 1992;</ns0:ref><ns0:ref type='bibr' target='#b39'>Lopardo &amp; Seoane 2004)</ns0:ref>.</ns0:p><ns0:p>There are many criteria in the design of hydraulic structures that must be optimized at the same time. Construction, operating and maintenance costs, construction reliability, environmental impact, social disruption and potential loss of life can be identified as one of the most noticeable of a large number of criteria <ns0:ref type='bibr'>(Rasekh, Afshar &amp; Afshar 2010)</ns0:ref>. Therefore, a comprehensive research on the environmental risk of hydrotechnical structures failure is necessary. One aspect of this type of research is a proper recognition of hydraulic and morphodynamic processes that accompany the construction of water structures using physical and mathematical models <ns0:ref type='bibr' target='#b39'>(Lopardo &amp; Seoane 2004;</ns0:ref><ns0:ref type='bibr' target='#b56'>Syvitski et al. 2010)</ns0:ref>.</ns0:p><ns0:p>Damming up rivers by implementation of a hydraulic structure unavoidably influences the stream course and valley morphology. Upstream the structure, due to the water surface level increment the water can flood the adjacent valley and the reservoir may form. Due to the crosssection enhancement and, therefore, stream velocity reduction, sediment sedimentation and accumulation may occur when the weight of the ground particles will outweigh the transport capacity of the water stream <ns0:ref type='bibr'>(Graf 1989</ns0:ref>). Simultaneously, erosion can intensify downstream the structure, especially in the case of very low water surface level and the stream velocity rapid enhancement. The debris-free stream leaving the dammed structure, with additional kinetic energy and increased turbulence, has a high eroding capacity <ns0:ref type='bibr' target='#b58'>(Szyd&#322;owski &amp; Zima 2006;</ns0:ref><ns0:ref type='bibr'>Pagliara &amp; Kurdinstani 2013;</ns0:ref><ns0:ref type='bibr' target='#b66'>Zobeyer et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b35'>Lee &amp; Hong 2019)</ns0:ref>. The effect of intensified erosion, which takes place downstream the structure, is mainly local scouring and gradual lowering of the bottom on an increasingly longer section of the river. The increased erosion of a riverbed is unfavorable and undesirable not only due to slow degradation of the riverbed, but also because of occurrence of rapid morphodynamic processes. It is commonly assumed that the most intensive transformation of riverbeds takes place during catastrophic flooding when basic hydrodynamic parameters of the stream increase many times. Excessive development of a scour hole directly behind the structure, such as weir or sluice, poses a threat to its safety as it may lead to washing away the foundation, embankments damages and loss of stability <ns0:ref type='bibr' target='#b4'>(Bajkowski, Siwicki, Urba&#324;ski 2002)</ns0:ref>. Removing and repairing these undesirable effects is troublesome and expensive. Therefore, technical solutions are needed to reduce scour hole dimensions.</ns0:p><ns0:p>Due to a focus on ecological changes in aquatic environment and the adjacent area in recent decades, it is worth considering what environmental benefits of local scouring can be. In order to specify them, it is necessary to identify this phenomenon more precisely i.e. to know the causes of its occurrence, its characteristics and the process over time <ns0:ref type='bibr' target='#b52'>(Siwicki, Urba&#324;ski 2004)</ns0:ref>.</ns0:p><ns0:p>The specific character of the flow within a scour area is a certain diversity of conditions in comparison with a river in its natural course. Stream velocity and associated physical forces constitute the most important environmental factor affecting organisms living in watercourses <ns0:ref type='bibr' target='#b12'>(Cullen 1991;</ns0:ref><ns0:ref type='bibr' target='#b53'>Smith, Goodwin &amp; Nestler 2014)</ns0:ref>.</ns0:p><ns0:p>A significant reduction of stream velocity in a bottom local scour area leads to a specific flow region formation named the wall-adjacent boundary layer. This layer can serve as a shelter for organisms from turbulence and high water velocities. In well-developed scour holes, the nearzero stream velocity in the wall-adjacent boundary layer can form local areas of still water <ns0:ref type='bibr' target='#b40'>(Lupandin 2005;</ns0:ref><ns0:ref type='bibr' target='#b38'>Liao &amp; Cotel 2013;</ns0:ref><ns0:ref type='bibr' target='#b19'>Hockley et al. 2013)</ns0:ref>. The adaptation of fish to living in still and flowing water has been extensively studied and it was found that one of the most important environmental factors is the dissolved oxygen content of the water. The higher the stream velocity and turbulence intensity, the more oxygenated the water is. Therefore, the water after passing through the structure is aerated as evidenced by the measured increases in speed and turbulence of the stream in the position downstream compared to the upper structure <ns0:ref type='bibr'>(Kobus &amp; Koschitzky 2014)</ns0:ref>.</ns0:p><ns0:p>As a result of the process of sorting and armoring, i.e. washing out finer particles and leaving thicker fractions at the bottom, the material is sorted on the scour hole bottom surface and a layer made of thicker fractions is formed. These are the factors that determine the structure of the velocity field at the bottom. This type of scouring process creates quite favourable environmental conditions for many species of fish, i.e. a well aerated stream, thicker material on the bottom and a smooth type of stream in the wall-adjacent boundary layer area <ns0:ref type='bibr' target='#b52'>(Siwicki, Urba&#324;ski 2004;</ns0:ref><ns0:ref type='bibr' target='#b45'>Ochman &amp; Kaszubkiewicz 2004;</ns0:ref><ns0:ref type='bibr' target='#b18'>Hauer et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Local scouring in the aquatic environment may be particularly desirable during the season of low dischagres. It can then serve as a reservoir in which particular species that require a certain water depth, can survive. Lowering of the river bed may also be beneficial in the construction of a fish pass. The inlet and outlet of fishponds must be submerged deep enough under the water surface level, and this could be difficult in the structure stand downstream, where the depth of the stream is normally lower than in upstream stand area <ns0:ref type='bibr' target='#b52'>(Siwicki, Urba&#324;ski 2004)</ns0:ref>. At the design stage, it is important to develop a reliable forecast of a size, shape and position of the local scour, both in dams and small hydraulic structures, such as gated checks, weirs or sluices. Gated checks are small river training structures, applied for limiting channel incision, bed stabilization, reducing flow velocity and raising upstream water level. These structures are often used in channels where the adjustment of water level is required more frequently or where higher cost compared to stop-logs, are justified (e.g. saving of labour). Gated checks are usually equipped with hand-operated slide gates of various types, from simple wooden shutters to handwheel operated adjustable orifice type gates <ns0:ref type='bibr' target='#b32'>(Kraatz &amp; Mahajan 1975)</ns0:ref>.</ns0:p><ns0:p>Although many studies on local scouring downstream hydraulic structures can be found in the literature in the recent years (for example <ns0:ref type='bibr' target='#b55'>Sun, Wang &amp; Wang 2012;</ns0:ref><ns0:ref type='bibr' target='#b48'>Pagliara et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b23'>Khaple et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b1'>Al-Husseini, Al-Madhhachi &amp; Naser 2019;</ns0:ref><ns0:ref type='bibr' target='#b51'>Singh, Devi &amp; Kumar 2020;</ns0:ref><ns0:ref type='bibr'>Li el al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b60'>Taha, El-Feky &amp; Fathy 2020;</ns0:ref><ns0:ref type='bibr' target='#b64'>Wang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b65'>Yan, Rennie, Mohammadian 2020)</ns0:ref>, among them only a few focus on small hydraulic structures <ns0:ref type='bibr' target='#b39'>(Lopardo &amp; Seoane 2004;</ns0:ref><ns0:ref type='bibr' target='#b25'>Kiraga &amp; Popek 2016;</ns0:ref><ns0:ref type='bibr' target='#b46'>Odgaard 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Al-Suhaili, Abbood &amp; Samir Saleh 2017;</ns0:ref><ns0:ref type='bibr'>Kiraga &amp; Popek 2018)</ns0:ref>. Especially gated checks are a rare object of research, which, given quite high prevalence of this type of structures, determines the appropriateness of the undertaken subject.</ns0:p><ns0:p>In recent years, the role of small retention of valley areas was emphasized <ns0:ref type='bibr' target='#b7'>(Boix et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b41'>Mioduszewski 2014)</ns0:ref>. A significant role in maintaining and increasing retention is played by small water structures <ns0:ref type='bibr' target='#b39'>(Lopardo &amp; Seoane 2004)</ns0:ref> whichare less popular in the research subject matter than bridges or larger weirs with controlled closures.</ns0:p><ns0:p>It should also be noted that hydraulic instrumentation has advanced significantly, so fundamental flow parameters or river bed shape can be measured with greater precision. For instance, the PIV imaging anemometry system can be used to describe the distribution of velocity fields in the area of water structure and the description of a bottom shape can be perform using an echo sounder, e.g. fixed on a boat, or by laser scanning technique <ns0:ref type='bibr'>(Hager &amp; Boes 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Killinger 2014</ns0:ref>).</ns0:p><ns0:p>Most of the research samples in order to identify the factors influencing the amount of local scouring to the greatest extent below water structures were carried out in laboratory conditions. The following factors that influence a scour hole shape and location are: --related to the flume geometry (width, depth, bed inclination); --related to the type and geometry of the structure (type of structure, reinforcement construction, dimensions of upstream and downstream part of the structure elements); --characterizing water flow conditions (flow rate, average speed, hydraulic gradient, bed shear stress, flow resistance), --water properties (density, viscosity), --characterizing the flume material (grain size, grain distribution, density, porosity, roughness), --characterizing the conditions of sediment transport (critical velocity, critical shear stress, sediment transport intensity), --time <ns0:ref type='bibr'>(Graf 1989;</ns0:ref><ns0:ref type='bibr' target='#b9'>Breusers &amp; Raudkivi 1991;</ns0:ref><ns0:ref type='bibr' target='#b28'>Kiraga &amp; Popek 2019)</ns0:ref>.</ns0:p><ns0:p>In spite of many experimental works carried out under various constructional conditions and high variability of hydraulic conditions, the universal principles of calculating the local scour dimensions and transferring it to field conditions are still unknown. Solutions are sought, involving different coefficients, which characterize a given construction, based on identified factors that influence scour size and position <ns0:ref type='bibr' target='#b15'>(Franke 1960;</ns0:ref><ns0:ref type='bibr' target='#b54'>Straube 1963;</ns0:ref><ns0:ref type='bibr' target='#b61'>Tarajmovi&#269; 1966;</ns0:ref><ns0:ref type='bibr' target='#b50'>Rossinski &amp; Kuzmin 1969)</ns0:ref>. The formation and expansion of local scouring that results from time-varying, two-phase movement of water and sediment is one of the most undiscovered processes in hydrotechnical engineering <ns0:ref type='bibr' target='#b17'>(Graf 1998;</ns0:ref><ns0:ref type='bibr' target='#b44'>Nouri Imamzadehei et al. 2016)</ns0:ref>. Despite numerous studies carried out since the first decades of last century (for example <ns0:ref type='bibr' target='#b33'>Lacey 1946;</ns0:ref><ns0:ref type='bibr' target='#b0'>Ahmad 1953;</ns0:ref><ns0:ref type='bibr' target='#b9'>Breusers &amp; Raudkivi 1991;</ns0:ref><ns0:ref type='bibr' target='#b36'>Lenzi, Marion &amp; Comiti 2003;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ben Meftah &amp; Mossa 2006;</ns0:ref><ns0:ref type='bibr' target='#b25'>Kiraga &amp; Popek 2016;</ns0:ref><ns0:ref type='bibr' target='#b48'>Pagliara et al. 2016;</ns0:ref><ns0:ref type='bibr'>Kiraga &amp; Popek 2018;</ns0:ref><ns0:ref type='bibr' target='#b1'>Al-Husseini, Al-Madhhachi &amp; Naser 2019)</ns0:ref>, there is no sufficient and unquestionable basis for the mathematical description of the process of local erosion, and thus for a development of forecasts of scour holes that will occur during the design of structures. Also, it is not always possible to predict fully reliable estimation based on the results of laboratory tests, because in laboratories the researchers are usually unable to lead to the occurrence of the so-called final scour, i.e. to a state in which the extension of the duration of the experiment does not cause changes in the dimensions and location of sandy bottom and banks <ns0:ref type='bibr' target='#b10'>(Chabert &amp; Engeldinger 1956;</ns0:ref><ns0:ref type='bibr' target='#b5'>Barbhuiya &amp; Dey 2004)</ns0:ref>. Moreover, designers find it difficult to choose those that give reliable results. Due to the diversity of applied constructions of structures and the variability of hydraulic conditions, it is difficult to generalize the derived formulas <ns0:ref type='bibr' target='#b17'>(Graf, 1998;</ns0:ref><ns0:ref type='bibr' target='#b5'>Barbhuiya &amp; Dey 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Ben Meftah &amp; Mossa 2006)</ns0:ref>.</ns0:p><ns0:p>The absence of a forecast of the effects of local erosion makes it impossible to rationally assess the degree of safety and certainty of the use of a hydraulic structure, since the scour poses a similar threat to the structure, such as insufficient structure capacity, too low structural stability coefficient, ground strength exceeding, etc.</ns0:p><ns0:p>The complexity of a local scouring process means that only fragments of the problem are usually examined with limited objectives, such as: --explanation of the influence of factors, e.g. the construction of downstream part of the structure, length and roughness of the reinforcements, etc. <ns0:ref type='bibr' target='#b50'>(Rossinski &amp; Kuzmin 1969</ns0:ref> --formulation of dependencies, formulas, etc. to determine scour forecasting for assumed geometrical, hydraulic and ground conditions <ns0:ref type='bibr' target='#b16'>(Gaudio &amp; Marion 2003;</ns0:ref><ns0:ref type='bibr' target='#b28'>Kiraga &amp; Popek 2019)</ns0:ref>.</ns0:p><ns0:p>Additionally, the results of tests carried out in a laboratory are difficult to translate directly into field conditions due to the scale effect <ns0:ref type='bibr' target='#b14'>(Farhoudi &amp; Smith, 1985)</ns0:ref>, whereas during field tests problems result mainly from the lack of knowledge of the initial conditions, i.e. the shape of the bottom before disturbing the existing dynamic balance in the channel <ns0:ref type='bibr' target='#b36'>(Lenzi, Marion &amp; Comiti, 2003;</ns0:ref><ns0:ref type='bibr' target='#b48'>Pagliara et al., 2016)</ns0:ref>.</ns0:p><ns0:p>Researchers agree that regardless of a construction of ahydraulic structure, the depth of the scour hole is influenced by length, roughness and height position of the fortifications downstream. <ns0:ref type='bibr' target='#b50'>(Rossinski &amp; Kuzmin 1969;</ns0:ref><ns0:ref type='bibr' target='#b62'>Urba&#324;ski 2008;</ns0:ref><ns0:ref type='bibr' target='#b60'>Taha, El-Feky &amp; Fathy 2020)</ns0:ref>. Difficulties in explaining and presenting the influence of factors on the process of scouring and the lack of perspectives for establishing universal relations between the complicated flow system and the sediment transport, lead to use simple, in fact intuitive, relationships that allow to determine the depth of local scour.</ns0:p><ns0:p>The estimation of the maximal scour hole depth and the channel reach infested by extensive erosion allows for a proper design of the downstream of hydraulic structure, ensuring safety and stability, as well as reducing the construction and subsequent operation cost. Therefore, the estimation of the geometry of forecasted scour should be an integral part of the design stage of hydrotechnical structures <ns0:ref type='bibr' target='#b8'>(Brandimarte, Paron &amp; Di Baldasarre, 2012;</ns0:ref><ns0:ref type='bibr' target='#b49'>Prendergast &amp; Gavin, 2014)</ns0:ref>.</ns0:p><ns0:p>Difficulties of local scouring investigations result primarily from the multitude of factors influencing its shape and dimensions. The following factors can be mentioned among them <ns0:ref type='bibr' target='#b15'>(Franke 1960;</ns0:ref><ns0:ref type='bibr' target='#b54'>Straube 1963;</ns0:ref><ns0:ref type='bibr' target='#b61'>Tarajmovi&#269; 1966;</ns0:ref><ns0:ref type='bibr' target='#b50'>Rossinski &amp; Kuzmin 1969)</ns0:ref> This paper comprises the identification, verification and validation of chosen empirical formulas: Shalash &amp; Franke, Straube, M&#252;ller and Tarajmovi&#269; serving to estimate the scour dimensions in local scour process forming due to damming up the flume by the gate, equipped with downstream stage embankment. For formula optimization the Monte Carlo sampling method was applied. Laboratory research was performed as a first part of the studies, then the formula best describing flume experiment was verified in field conditions.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>Research based on laboratory studies was performed in a 11-m long hydraulic flume with 0.58m width. No bed inclination downstream was introduced, however it should be noted that lowland rivers that formed in alluvial depositions usually have gradients of 0.5 -3%. If such a slope were to be reproduced in the present laboratory conditions, the difference in elevation of the bottom below the water structure would be 1 mm to a maximum of 1 cm. Data were collected as previously described in the publication 'Bed shear stress influence on local scour geometry properties in various flume development conditions' in Water by Marta <ns0:ref type='bibr'>Kiraga and Zbigniew Popek (2019)</ns0:ref>. Specifically, the research assumed bed shape measurements during local scouring formation, both using pin gauge, laser scanning of the surface and water surface level examination in presumed hydraulic conditions. However, the examined flume development differs from the mentioned publication in Water. Namely, two gated check models assumed slide gate introduction, which was constantly raised to 5 cm of height to ensure invariable flow area of 0.029 m 2 (Figure <ns0:ref type='figure'>1 a, b</ns0:ref>). In the vicinity of the damming structure, the bottom was solid on the length of L 1 = 0.30 m upstream the gate and L 2 = 0.80 m downstream, designed to imitate bed reinforcement typical for weirs or other river training structures, often made of concrete or rip rap. The reinforcement downstream the check was made of plain slab working as a reinforcement within the model I of flume development (Figure <ns0:ref type='figure'>2</ns0:ref> a, 3 a) (4), whereas model II assumed stone riprap reinforcement made of rocks (8) whose medium height was 1.5 cm (Figure <ns0:ref type='figure'>2</ns0:ref> b, 3 b). A scour hole formed inside a sandy part downstream the check with a length L 3 of 2.20 m (2). The applied constructional solution of the model assumed representation of water flow under a partially open valve of the gated check with lowered reinforced bottom below with variable roughness where sediment transport takes place through the gate -not held by any weir. According to <ns0:ref type='bibr' target='#b32'>Kraatz &amp; Mahajan (1975)</ns0:ref> the length of the reinforcement downstream the gated check should be ca. 1.5 times longer than the width of the gate -and the model applied within present experiment ensure the elements' dimensions close to this ratio.</ns0:p><ns0:p>Bottom shape was investigated at all flume lengths. After each measurement series, the sand in the flume was dried for ca. 13 hours -the outflowing water was removed by using drainage pipe (5 in Figure <ns0:ref type='figure'>2</ns0:ref> a and b). The measurement schedule was similar as previously described in 'Bed shear stress influence on local scour geometry properties in various flume development conditions' in Water by Marta <ns0:ref type='bibr'>Kiraga and Zbigniew Popek (2019)</ns0:ref>. Namely, pin water gauges were used in order to measure water surface elevation at the intake part and along the flume in the central axis (1). The water surface level was regulated with an outlet gate (6). Before introducing water into the flume and after draining the sand the final level of sandy bottom was measured with a laser scanner ( <ns0:ref type='formula'>7</ns0:ref>) and with a moving disc probe (1) as a supportive device in presumed time steps (0.5 -2 h).</ns0:p><ns0:p>Before starting each measurement series and introducing water stream into the flume, the sandy bed was uniformly adjusted to a constant level and compacted with load of 2.5 kg dropped to the bed surface with an energy of about 5 J. Then, the position of the bottom was measured with a disc probe with presumed mesh 5 x 7 cm to 20 x 10 cm both upstream and downstream the hydraulic structure.</ns0:p><ns0:p>As described in 'Bed shear stress influence on local scour geometry properties in various flume development conditions' in Water by Marta <ns0:ref type='bibr'>Kiraga and Zbigniew Popek (2019)</ns0:ref> flume side walls were made of glass with a roughness coefficient n w = 0.010 s&#8226;m -1/3 . The soil used during the study was uniform coarse sand with medium diameter d 50 = 0.91 and d 95 = 1.2 mm and roughness coefficient n b = 0.028 s&#8226;m -1/3 . Experiments were performed in the scope of steady water flow discharge within the following range Q w = 0.010 -0.045 m 3 &#8226;s, water depth downstream the structure h = 0.05 -0.26 m and Froude number Fr &lt;1. 29 measurement series were performed, each lasting 8 hours (9 measurement series on model I and 20 on model II) (Table <ns0:ref type='table'>1 and 2)</ns0:ref>. No sediment feeding system was adopted. Bedload transport conditions were assured by specific set of hydraulic conditions that invokes particle movement from upstream towards downstream. Therefore, the experiment was carried out in 'live-bed' conditions, where soil leaving the scour hole is substituted by approaching load from the upstream. It is worthy to notice that for typical lowland rivers both bedload load and suspended load are present in various relations. For example, the suspended load constitutes 60-70% of the whole sediment load transported by Vistula River in Poland and 50-90% of that is transported by its tributaries <ns0:ref type='bibr' target='#b34'>(Lajczak 1996)</ns0:ref>, although only bedload was investigated in this study.</ns0:p><ns0:p>A group of experiments carried out in a hydraulic laboratory, the results of which were published, among others, in Water or IEEE Access journals <ns0:ref type='bibr' target='#b25'>(Kiraga and Popek 2016;</ns0:ref><ns0:ref type='bibr' target='#b28'>Kiraga and Popek 2019;</ns0:ref><ns0:ref type='bibr' target='#b29'>Kiraga and Miszkowska 2020)</ns0:ref>, concerned the phenomenon of formation of local scouring as a result of not only varied roughness of the materials building the bed, but also the restriction of the flow field by inserting a model of damming structures into the flume. Namely, due to the flow resistance increment along the whole flume resulting from varied roughness of solid and sandy bottom, the hydraulic gradient increases causing the intensification of shear stress at the bottom. After exceeding the critical shear stress, the motion of sediment grains starts, followed by gradual scouring of the bed. Maximal scour depth z max , scour length L s and the distance between the deepest point of the hole and the end of reinforcement L e were examined (Figure <ns0:ref type='figure'>4</ns0:ref>) during each measurement.</ns0:p><ns0:p>In order to investigate the final scour shape both a device equipped with laser scanner and a disc probe were applied as supportive devices. Data were collected as previously described in 'Bed shear stress influence on local scour geometry properties in various flume development conditions' in Water by Marta <ns0:ref type='bibr'>Kiraga and Zbigniew Popek (2019)</ns0:ref>. Prototype A1 of the device was engineered in 2016 by Marta Kiraga and Matvey Razumnik within the university grant for young researchers 'The influence of small hydraulic structures on sediment transport conditions' <ns0:ref type='bibr'>(Kiraga et al. 2018)</ns0:ref>.</ns0:p><ns0:p>Prototype A1 (Figure <ns0:ref type='figure'>5</ns0:ref>) is equipped with a laser rangefinder and automatic movement system embedded on guides along the flume, scanning the bottom area with a demanded grid (every 1 mm in case of present experiments). Grid density alteration possibility gives a far greater accuracy of measurement than the disc probe. The use of the device ensures data transmission directly in digital form, so that the coordinates can be easily processed to obtain the desired scour hole geometrical parameters.</ns0:p><ns0:p>Laser scanning, also known as LiDAR (Light Detection and Ranging) is an active tele-detection method which uses electromagnetic waves sent by the emitter. The result is point cloud with coordinates (x, y, z) <ns0:ref type='bibr' target='#b21'>(Jaboyedoff et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b24'>Killinger 2014</ns0:ref>). The measuring system (LiDAR) consists mainly of a transmitter i.e. a module generating laser light (diodes), an optical telescope focusing the returning reflected radiation and a detector converting light energy into an impulse recorded in the module that records the acquired data. The prototype's supporting elements are made of biodegradable polyactide (PLA) and are printed on a 3D printer. Using Raspberry Pi microcomputer allows simultaneous computations and data collecting by the beam. The device is fully automated which was achieved by the application of a single board computer, dedicated software and the set of stepper motors, which results in measurements repeatability, constant accuracy on demand and fast execution of results. The obtained coordinates mesh is characterized by high resolution: 1 mm by 1 mm-therefore bottom shape is described very precisely, both in numerical form and as a graphical tracing. Numerical cloud can be easily transformed thence scour hole dimensions such as length or depth can be estimated. LiDAR technology application in scour shape and its volume in flume experiments is based on the introduction of an automatic measuring module which, placed above the bottom on a specially prepared controllable system of guides, describes its shape by creating a point cloud.</ns0:p><ns0:p>Deriving from a statement that scouring process stops when stream velocity v is equal to nonscouring velocity v n Rossinski <ns0:ref type='bibr' target='#b50'>(Rossinski &amp; Kuzmin 1969)</ns0:ref> stated that water depth above the local scour can be calculated as:</ns0:p><ns0:formula xml:id='formula_0'>&#119867; = &#119911; &#119898;&#119886;&#119909; + &#8462; = &#119896; 1 1.2 &#119902; &#119907; &#119899;1 (1)</ns0:formula><ns0:p>where z max is local scour depth, [m]; h is water elevation before scour formation (See Figure <ns0:ref type='figure'>4</ns0:ref>), [m]; k 1 is a dimensionless coefficient, describing intensified turbulence of the stream, <ns0:ref type='bibr'>[-]</ns0:ref>; q is unit discharge, [m 3 &#8226;s -1 &#8226;m -1 ]; v n1 is non-scouring velocity for water depth of 1 m, depending to soil properties, [m 2 &#8226;s -1 ] calculated as following:</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_1'>&#119907; &#119899;1 = 2&#119892;(&#120574; &#119903; -&#120574; &#119908; )/1.75&#120574; &#119908;&#119889; 50 &#119897;&#119900;&#119892;(8.8/&#119889; 95 ) (2)</ns0:formula><ns0:p>in which g is gravity acceleration, g = 9.81 m&#8226;s -2 ; and are specific weights, sediment and &#120574; &#119903; &#120574; &#119908; water, respectively [N&#8226;m -3 ]; d 50 and d 95 are diameters that correspond to 50% and 95% of particles finer than the reported particle size.</ns0:p><ns0:p>The k 1 value in the formula ( <ns0:ref type='formula'>1</ns0:ref>) is an empirical coefficient, dependent on downstream development conditions. Based on practice experiences k 1 takes the value of 1.70 when the reinforcement downstream the gate is not deepened and sheet piling, palisade or other vertical securing element make an additional protection. Due to the stream energy enhancement in the region of the gate outlet without any energy dissipating device local scouring process is intensified. When transverse trench is dug downstream the reinforcement, of depth equal to the expected depth of the scour; and the slope of this trench is no more than 1:4, then k 1 =1.05 should be assumed (Figure <ns0:ref type='figure'>6</ns0:ref> a, b).</ns0:p><ns0:p>Experimental case is referred to the conditions when coarse sandy bed is preceded by deepened reinforcement downstream (Figure <ns0:ref type='figure'>6 c</ns0:ref>) therefore, empirical studies on k 1 parameter were needed.</ns0:p><ns0:p>The difficulty of explaining and presenting the impact of factors influencing local scouring process in large scale hydraulic structures with the lack of perspectives for establishing relations between a complicated flow system and sediment transport, forces to apply simple, intuitive relations allowing for the determination of the depth of scour holes. Scour length L s and the distance between the deepest point of the hole and the end of reinforcement L e where the stream comes out from under the gate were determined by several authors: --According to Shalash and Franke <ns0:ref type='bibr' target='#b15'>(Franke 1960)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_2'>&#119871; &#119904; &#119886;&#119899;&#119889; &#119871; &#119890; = &#119891;(&#119911; &#119898;&#119886;&#119909; ) (3) &#119871; &#119904; = 11 &#8226; &#119911; &#119898;&#119886;&#119909; (4) &#119871; &#119890; = 6.6 &#8226; &#119911; &#119898;&#119886;&#119909;<ns0:label>(5)</ns0:label></ns0:formula><ns0:p>--According to M&#252;ller <ns0:ref type='bibr' target='#b15'>(Franke 1960;</ns0:ref><ns0:ref type='bibr' target='#b54'>Straube 1963)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_3'>&#119871; &#119904; &#119886;&#119899;&#119889; &#119871; &#119890; = &#119891;(&#119911; &#119898;&#119886;&#119909; ) (6) &#119871; &#119904; = (9.9 &#247; 0.8) &#8226; &#119911; &#119898;&#119886;&#119909; (7) &#119871; &#119890; = (4.9 &#247; 0.5) &#8226; &#119911; &#119898;&#119886;&#119909; (8) or &#119871; &#119904; &#119886;&#119899;&#119889; &#119871; &#119890; = &#119891;(&#119911; &#119898;&#119886;&#119909; ,&#8462;) (9) &#119871; &#119904; = (6.0 &#247; 1.22) &#8226; (&#119911; &#119898;&#119886;&#119909; + &#8462;) (10) &#119871; &#119890; = (2.94 &#247; 0.59) &#8226; (&#119911; &#119898;&#119886;&#119909; + &#8462;)<ns0:label>(11)</ns0:label></ns0:formula><ns0:p>--According to Straube <ns0:ref type='bibr' target='#b54'>(Straube 1963)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_4'>&#119871; &#119904; = &#119891;(&#119911; &#119898;&#119886;&#119909; , &#119902;,&#8462;,&#119889; 50 )<ns0:label>(12)</ns0:label></ns0:formula><ns0:p>&#119871; &#119904; = 8.0&#119902; 0.36 (&#119911; &#119898;&#119886;&#119909; + &#8462;)&#119889; 50 -0.14 &#8462; -0.40 (13) &#119871; &#119890; = &#119891;(&#119871; &#119904; ,&#8462;,&#119889; 50 ) (14) &#119871; &#119890; = 0.39&#119871; &#119904; &#119889; 50 0.12 &#8462; -0.12 (15)</ns0:p><ns0:p>The formulas (3-15) are recommended for systems in which the outflow from the gate goes directly onto unreinforced ground. For constructions equipped with reinforcement the Tajarmovi&#269; formula is recommended <ns0:ref type='bibr' target='#b61'>(Tarajmovi&#269; 1966)</ns0:ref>:</ns0:p><ns0:p>&#119871; &#119890; = &#119891;(&#119911; &#119898;&#119886;&#119909; ) (16) &#119871; &#119890; = 12.75&#119911; &#119898;&#119886;&#119909; 0.5 <ns0:ref type='bibr'>(17)</ns0:ref> Monte-Carlo integration works by comparing random samples with the function value. Straube equations can be described generally in the following forms: Using a random number generator in the assumed ranges of values, 6000 combinations of parameters a, b, c, d for equation ( <ns0:ref type='formula'>18</ns0:ref>) and 6000 combinations of parameters k, m, p for equation (19) were selected. The average relative error &#948; for all 29 series of measurements was chosen as a criterion for evaluation of the formula described by a given combination of parameters.</ns0:p><ns0:formula xml:id='formula_5'>&#119871; &#119904; = &#119886;&#119902; &#119887; (&#119911; &#119898;&#119886;&#119909; + &#8462;)&#119889; 50 -&#119888; &#8462; -&#119889; (18) &#119871; &#119890; = &#119896;&#119871; &#119904; &#119889; 50 &#119898; &#8462; -&#119901; (19)</ns0:formula><ns0:p>The key to the accuracy and correctness of the Monte Carlo method is a random number generator. The method presents a solution to a problem as a parameter of a hypothetical population. Using a sequence of random numbers, it creates a population sample from which estimated values of the sought parameters can be obtained <ns0:ref type='bibr' target='#b42'>(Niederreiter 1992</ns0:ref>).</ns0:p><ns0:p>Next step was to verify the optimized formula on independent measurement results published in 2010 by Gaudio and Marion, performed on a flume with sandy bottom with the hydraulic structure represented by the cascade of transversal sills.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Basic geometric parameters of observed scour during 29 measurement series, each characterized by unit discharge q (9 on Model I and 20 on model II) are presented in Table <ns0:ref type='table' target='#tab_3'>3 and 4</ns0:ref>. Nonscouring velocity for water depth of 1 m v n1 in presumed grain conditions was equal to 0.502 m/s. Maximal scour depth ranged from 1 to 10 cm. The criterion for the reach infested by the scour is bed level, i.e. scour, is recognized within an area in which the depth of the bottom after 8-hour measurement series exceeds 10% of the maximum hole depth <ns0:ref type='bibr' target='#b29'>(Kiraga and Miszkowska 2020)</ns0:ref>.</ns0:p><ns0:p>Within the scope of the assumed measurement schedule each measurement series were carried out with unique, within each model, variable combinations of input flow rate and water level. Manuscript to be reviewed However, by means of variability of those combinations, the same values of unit flow rate q (per unit width) were obtained, which leads to a conclusion that in respect of unit flows, the repeatability of the experiments was ensured. Moreover, during the laboratory tests it was necessary to repeat some measurement series several times, e.g. due to faulty transfer of numerical data from the microcomputer used, which made it possible to check the repeatability of test results. The repetition of the tests was performed assuming the measurement series duration and under the same hydraulic conditions. Differences in the bottom formation were shown, described by means of basic geometrical parameters of the scouring in the range of 0.3 -1.9% in the maximum depth of the scour hole and 2.2-4% in the range of the average depth of the scour, which indicates high repeatability of test results (Figure <ns0:ref type='figure'>7 a, b, c</ns0:ref>). Slightly more significant deviation was connected with scour parameters connected with its length: relative error ranging within 0.5 -13.6% was met in total scour length and 3.0 -16.6% in the case of the distance from the end of reinforcement to the deepest scour point.</ns0:p><ns0:p>The Rossinski formula (1) parameters were identified for investigated test stand due to lack of the present gate check structure construction analyses so far. Parameters identification was performed on the basis of mean relative error between observed scour depth and calculations &#120575; results (Parameter k 1 was tested in the range 0.00 to 2.00. With k 1 equal to 1.10, the mean relative error reached the minimum value. For the entire tested range of k1 values, errors in the range of 15 -100% were achieved (Figure <ns0:ref type='figure'>8</ns0:ref>).</ns0:p><ns0:p>Parameters of 4,5; 7,8; 10,11; 13; 15 and 17 formulas were verified for two models of gated check development. Calculated parameters of observed scour were examined in comparison with the measured ones. The criterion of comparison evaluation was mean relative error of each scour parameter estimation (Table <ns0:ref type='table' target='#tab_5'>5</ns0:ref>) calculated for each group of 29 measurements.</ns0:p></ns0:div> <ns0:div><ns0:head>&#120575;</ns0:head><ns0:p>The limitation in determining the range of a scour hole was the length of sandy part (bottom edge) L 3 , which was 2.20 m. In field studies, the length is long enough for the full scour length development -there is no sandy bed length limitation. As mentioned above, the field of the scour was considered to be an area where the bottom lowering exceeded 10% of its maximum depth in presumed time step (Figure <ns0:ref type='figure'>9</ns0:ref>). If another criterion was to be adopted, for example, a consideration of the scour hole area within the region where a bottom lowering exceeds 15 or more % of the maximum scour depth, a limiting effect of the sandy bottom downstream the structure could be avoided in some measurement series. Investigated equations were divided into two groups: --simple formulas based only on maximal local scour depth z max or on z max and water depth h before scour formation (Eq. 4, 7, 10 for total scour length estimation and <ns0:ref type='bibr'>Eq. 5,</ns0:ref><ns0:ref type='bibr'>8,</ns0:ref><ns0:ref type='bibr'>11,</ns0:ref><ns0:ref type='bibr'>17</ns0:ref> for the distance between the deepest point of scour and the end of reinforcement estimation) --formulas involving not only local scour hole depth z max and the depth of water above the unwashed bottom h, but also grain characteristics, represented by d 50 diameter and hydraulic parameter, i.e. unit water discharge q (the Straube formula -Eq. 13, 15) Formulas that depend only on the local scour hole depth z max <ns0:ref type='bibr'>5;</ns0:ref><ns0:ref type='bibr'>8)</ns0:ref> or on the local scour hole depth z max and the depth of water above the unwashed bottom h (M&#252;ller -Eq. 10, 11) demonstrated mean relative error of 56.9-72.8% in the scope of total scour length L s and 38.3-57.0% for the distance between the deepest point of scour and the end of reinforcement L e . The Tajarmovi&#269; equation (Eq. 17) indicates a 392.7% error. Medium relative error for simple formulas (4,7,10) was equal to 66.5% and for formulas <ns0:ref type='bibr'>(5,8,11 and 17)</ns0:ref> was equal to 133.3%.</ns0:p><ns0:p>Calculations using the formula, involving not only local scour hole depth z max and the depth of water above the unwashed bottom h, but also grain characteristics, represented by d 50 diameter and hydraulic parameter, i.e. unit water discharge q (the Straube formula -Eq. 13, 15) provide the best fit to the measurement data. The relative error was 34.2 % for total scour length and 32.1% for the distance between the deepest point of scour and the end of reinforcement. The Figures <ns0:ref type='figure'>10 and 11</ns0:ref> demonstrate the results of calculations in relation to the measured values.</ns0:p><ns0:p>One combination of parameters a, b, c, d and one combination of parameters k, m, p were selected basing on the presumed criteria: the formulas described by these parameters were characterized by the lowest average relative error for all 29 test series. The best data explanation &#120575; for laboratory database of I and II model was achieved in the following parameters values: a = 7.41; b = 0.38; c = -0.10; d = -0.45; k = 0.34; m = 0.01; p = -0.01, thence the identified formulas, can be described as:</ns0:p><ns0:formula xml:id='formula_6'>&#119871; &#119904; = 7.41&#119902; 0.38 (&#119911; &#119898;&#119886;&#119909; + &#8462;)&#119889; 50 -0.10 &#8462; -0.45<ns0:label>(20)</ns0:label></ns0:formula><ns0:formula xml:id='formula_7'>&#119871; &#119890; = 0.34&#119871; &#119904; &#119889; 50 0.01 &#8462; -0.01<ns0:label>(21)</ns0:label></ns0:formula><ns0:p>Optimization revealed a diminished error, both in the case of total scour length L s (10.1%) and for the distance between the deepest point of scour and the end of reinforcement L e (18.2%).</ns0:p><ns0:p>Verification of the optimized Straube formula was performed on independent data published in 2010 by <ns0:ref type='bibr'>Gaudio and Marion. In 1998, research</ns0:ref> was carried out in the Wallingford Ltd., a hydraulic laboratory, on the evolution of local scouring downstream the bed sills cascade. The flume consists of a 60 cm-wide, 24.5 cm-high and 5.57-m working section with rectangular cross-section. For the full description of the duct, see <ns0:ref type='bibr' target='#b16'>Gaudio, Marion (2003)</ns0:ref>. The bed sills used in all experiments were 25 mm-thick by 15 cm-high wooden plates, with the same width as the transversal section. The sediment used in all tests was sand with median diameter d 50 = 1.8 mm. No sediment recirculating system was adopted.</ns0:p><ns0:p>The similarity of Gaudio and Marion test stand and the flume, where present research was performed, come down to used bed material (sand), the shape of the flume (rectangular, 60-cm width), the same order or magnitude of unit discharges and the transversal type of water structure. The main difference is duration of each experimental series: in Gaudio and Marion, experiment series last much longer than the present one: from 45 to 90 hours. Gaudio carried out 12 series of measurements in order to obtain the geometric dimensions of local scour holes formed under given hydraulic conditions. The maximal depth z max and the total length of the scour L s was studied. The hydraulic parameters of each measurement series and the geometric properties of the local scour are summarized in Table <ns0:ref type='table'>6</ns0:ref>. Based on the results received, Straube formula was verified in its original and optimized in form by Monte Carlo sampling procedure (Eq. 13 and 20). A mean relative error &#948; = 49.2% was obtained for the original form, whereas the application of the optimized Straube formula demonstrated a better description of the data obtained in the laboratory, characterized by an error mean relative &#948; equal to 18.1%.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The optimized Straube formula demonstrated very accurate laboratory dataset description, whereas the remaining equations analysis showed a relative error ranging up to more than 390%. The Straube equations forms, which have been optimized for the laboratory workstation, have been validated for field data. Zago&#380;d&#380;onka River in Czarna (Poland) was built in the fifties of last century as a concrete hydraulic structure to store up the water and to use its energy to drive the mill wheel. The total width of the spill is divided by I-beam guides into 3 clear spans: 1.16 m wide outermost spans and 1.22 m center span (Figure <ns0:ref type='figure'>12 a, b</ns0:ref>). In the guides, a measuring sharp-crested triangular weir was installed. The height and shape of the weir edges were developed to ensure non-submergence weir working conditions at the highest possible flow rates.</ns0:p><ns0:p>The downstream part of the structure consists of 8.80 m long concrete reinforcement with a longitudinal slope of 1%, followed by a 0.60 m drop, so it could be recognized as similar to laboratory condition test models. The river bed, directly below the drop, is partially covered with a stone riprap over on a reach of about 1.0 m, and in a further section it is scourable, made of sand, with d 50 diameter of 0.42 mm and d 90 diameter of 0.74 mm.</ns0:p><ns0:p>On June 11 2013, a flood occurred. The flow rate in the hydrograph peak reached 5.06 m 3 s -1 . This event resulted in local scour formation downstream the weir, whose dimensions were measured, analyzed and published by <ns0:ref type='bibr'>Urba&#324;ski and Hejduk (14)</ns0:ref>. Field measurements performed the following local scour dimensions:</ns0:p><ns0:p>--water depth above the deepest scour point H max = z max + h = 2.43 m; --local scour length L s = 13.8 m; --the distance between the deepest point of scour and the end of reinforcement L e = 5.20 m.</ns0:p><ns0:p>In the case of water depth H max calculations an error of 39.5% was achieved using the Rossinski formula with a k 1 parameter equal to 1.70 (Eq.1) (Table <ns0:ref type='table' target='#tab_6'>7</ns0:ref>). The best fit of the measurement and calculations was obtained for M&#252;ller equations, where measured scour length and the distance between the deepest point of scour from the end of reinforcement rare within the ranges described in equations no. 8 and 10 (an error of 0%). In the case of default form of Straube equations, 57.2 % of an error was achieved for scour length L s analysis and 7.7 % for the L e distance.</ns0:p><ns0:p>An optimized form of Straube equations ( <ns0:ref type='formula' target='#formula_6'>20</ns0:ref>), ( <ns0:ref type='formula' target='#formula_7'>21</ns0:ref>) were checked on the field measurements. Calculations using the Straube's optimized formula showed excellent adherence for the measured and calculated value of the local bottom scour length (an error equal to 0.2%). However, the distance of the maximum hole depth from the end of the reinforcement was underestimated and the underestimation amounted to 16.6% of this value. A common observation for laboratory and field tests is the overestimation of both parameters using the Tajarmovi&#269; formula.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Two gated check models were investigated in water discharge flowing out from under the gate, characterized by different roughness of the reinforcement downstream, followed by scourable bed. 29 measurement series were performed in total, each lasting 8 hours. The basic geometrical parameters of local scour hole, resulting from the disturbance of hydrodynamic balance of the system were examined using autonomic remote-controlled measuring unit. The construction of the tested models was chosen due to the prevalence of such solutions among real objects. 10 computational formulas, used for many years in the water engineering practice, were verified for laboratory data. It was stated that functions based only on one (z max ) or two (z max , h) parameters provide weaker adjustment between calculations results and laboratory measurements. The Straube's formula, assuming that geometric parameters follow up on not only maximal scour depth and water level, but also granulometric parameters, represented by medium grain diameter d 50 and hydraulic properties of experiment, such as unit discharge q was distinguished as the best description of laboratory test results.</ns0:p><ns0:p>The Straube function demonstrated the mean relative error of 34.2% in the case of comparing the measurement and calculation result of the local scour depth and an error of 32.1% for the distance of the deepest point from the end of the reinforcement, while medium error for all the rest of formulas was 67% for L s and 133% for L e .</ns0:p><ns0:p>The Monte Carlo sampling method allowed the original formulas to be adapted to calculate the local scour geometrical parameters downstream the model of the gated check. In optimizing the parameters, the criterion of minimizing the relative error was applied. The Monte Carlo sampling procedure resulted in a much better match between the calculation results and the dimensions measured in the laboratory: Straube function optimized in this way demonstrated an error of 10.1% in the case of comparing the measurement and the calculation of the local scour length and an error of 18.2% for the distance of the deepest point from the end of the reinforcement. The Straube's formula, chosen as the best describing laboratory results, was verified on independent dataset, whose main features in common with present experiment characteristics are: used bed material (sand), the shape of the flume, the same order or magnitude of unit discharges and the transversal type of water structure. A mean relative error &#948; = 49.2% was obtained for the original form of the Straube formula, and 18.1% for optimized formula using the Monte Carlo sampling method. Due to the data availability, only the total length of the scour was compared.</ns0:p><ns0:p>The optimized for laboratory measurements equation was checked for the real object, which was selected on the basis of the similarity of the downstream reinforcement, and of the data availability. It should be emphasized that field measurements of the bottom shape after the formation of local scour hole are often difficult to access due to the imperfection of measuring instruments and lack of data before the formation of a local scour. The optimization led to obtain an error of 0.2% for scour length and an error of 16.6% for the distance of the deepest point from the end of the reinforcement.</ns0:p><ns0:p>The extension of the optimized Straube formula verification to other hydro-technical field objects is necessary for the applicability of the investigation, however it has to be stated that very high degree of adjustment of calculation results to field data (especially local scour length) provide an encouraging premise for further research. Manuscript to be reviewed Model I -scour geometry parameters summary table</ns0:p><ns0:p>Where: q -unit water flow discharge; z max -maximal scour depth; L e -the distance between the deepest point of the hole and the end of reinforcement; L s -scour length.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Model II -scour geometry parameters summary table</ns0:p><ns0:p>Where: q -unit water flow discharge; z max -maximal scour depth; L e -the distance between the deepest point of the hole and the end of reinforcement; L s -scour length.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Field measurements and calculations results summary table -Czarna Gauge z max -maximal scour depth; h -water depth before scour formation; L s -scour length; L e -the distance between the deepest point of the hole and the end of reinforcement.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>; Al-Mohammed, Jassin &amp; Abbas 2019); --highlighting the effectiveness of various design solutions to prevent excessive erosion (Epely-Chauvin, De Cesare &amp; Schwindt 2015; Taha, El-Feky &amp; Fathy 2020);</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>: --relation to the flume or channel geometry (e.g. the Shalash and Franke, M&#252;ller, Tajarmovi&#269; formula); --relation to the type and geometry of the structure (e.g. the Rossinski formula); --relation to water flow conditions, such as flow rate, average speed or flow resistance); --water physical properties; --relation to bed material (e.g. the Straube method).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>where a, b, c, d, k, m, p are function parameters which were sampled in the following ranges: --a &#61646; &lt;5.0, 8.0&gt;; --b &#61646; &lt;0.24, 0.40&gt;; --c &#61646; &lt;0.10, 0.20&gt;; --d&#61646; &lt;0.35, 0.45&gt;; --k &#61646; &lt;0.30, 0.60; --m &#61646; &lt;0.01, 0.13&gt;; --p &#61646; &lt;0.01, 0.20&gt;.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,280.87,525.00,303.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,232.42,525.00,264.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,178.87,525.00,296.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='32,42.52,201.82,525.00,355.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='33,42.52,206.92,525.00,213.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='34,42.52,181.57,525.00,342.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='35,42.52,252.82,525.00,342.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='36,42.52,229.87,525.00,428.25' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>1</ns0:head><ns0:label /><ns0:figDesc>No of measurement seriesq [m 3 &#8901; s -1 &#8901; m -1 ]</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>z max</ns0:cell><ns0:cell>H max</ns0:cell><ns0:cell>L e</ns0:cell><ns0:cell>L s</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0431</ns0:cell><ns0:cell cols='3'>0.0201 0.1451 0.59</ns0:cell><ns0:cell>2.10</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0345</ns0:cell><ns0:cell cols='3'>0.0911 0.1411 0.66</ns0:cell><ns0:cell>2.10</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0397</ns0:cell><ns0:cell cols='3'>0.0532 0.1532 0.68</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0517</ns0:cell><ns0:cell cols='3'>0.0821 0.1621 0.67</ns0:cell><ns0:cell>2.18</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0431</ns0:cell><ns0:cell cols='3'>0.1020 0.1600 0.78</ns0:cell><ns0:cell>2.01</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0517</ns0:cell><ns0:cell cols='3'>0.0672 0.1772 0.78</ns0:cell><ns0:cell>2.18</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0483</ns0:cell><ns0:cell cols='3'>0.0511 0.1611 0.76</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0448</ns0:cell><ns0:cell cols='3'>0.0630 0.1630 0.66</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>0.0500</ns0:cell><ns0:cell cols='3'>0.0772 0.1572 0.71</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row></ns0:table><ns0:note>2PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head /><ns0:label /><ns0:figDesc>No of measurement seriesq [m 3 &#8901; s -1 &#8901; m -1 ]</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>z max</ns0:cell><ns0:cell>H max [m]</ns0:cell><ns0:cell>L e</ns0:cell><ns0:cell>L s</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>[m]</ns0:cell><ns0:cell /><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>0.0345</ns0:cell><ns0:cell>0.0700</ns0:cell><ns0:cell>0.1200</ns0:cell><ns0:cell>0.80</ns0:cell><ns0:cell>2.19</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>0.0397</ns0:cell><ns0:cell>0.0143</ns0:cell><ns0:cell>0.1143</ns0:cell><ns0:cell>0.56</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell>0.0517</ns0:cell><ns0:cell>0.0287</ns0:cell><ns0:cell>0.1087</ns0:cell><ns0:cell>0.61</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell>0.0431</ns0:cell><ns0:cell>0.1020</ns0:cell><ns0:cell>0.1600</ns0:cell><ns0:cell>0.61</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell>0.0517</ns0:cell><ns0:cell>0.0487</ns0:cell><ns0:cell>0.1587</ns0:cell><ns0:cell>0.79</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell>0.0483</ns0:cell><ns0:cell>0.0610</ns0:cell><ns0:cell>0.1710</ns0:cell><ns0:cell>0.27</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell>0.0448</ns0:cell><ns0:cell>0.0313</ns0:cell><ns0:cell>0.1313</ns0:cell><ns0:cell>0.79</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>8</ns0:cell><ns0:cell>0.0500</ns0:cell><ns0:cell>0.0412</ns0:cell><ns0:cell>0.1212</ns0:cell><ns0:cell>0.63</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell>9</ns0:cell><ns0:cell>0.0414</ns0:cell><ns0:cell>0.0410</ns0:cell><ns0:cell>0.1210</ns0:cell><ns0:cell>0.31</ns0:cell><ns0:cell>1.77</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0500</ns0:cell><ns0:cell>0.0175</ns0:cell><ns0:cell>0.1175</ns0:cell><ns0:cell>0.58</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0220</ns0:cell><ns0:cell>0.0220</ns0:cell><ns0:cell>0.0820</ns0:cell><ns0:cell>0.33</ns0:cell><ns0:cell>1.00</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0231</ns0:cell><ns0:cell>0.0321</ns0:cell><ns0:cell>0.0871</ns0:cell><ns0:cell>0.41</ns0:cell><ns0:cell>1.10</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0240</ns0:cell><ns0:cell>0.0430</ns0:cell><ns0:cell>0.1030</ns0:cell><ns0:cell>0.51</ns0:cell><ns0:cell>1.50</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0360</ns0:cell><ns0:cell>0.0673</ns0:cell><ns0:cell>0.1273</ns0:cell><ns0:cell>0.53</ns0:cell><ns0:cell>1.98</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0385</ns0:cell><ns0:cell>0.0244</ns0:cell><ns0:cell>0.1144</ns0:cell><ns0:cell>0.51</ns0:cell><ns0:cell>1.60</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0375</ns0:cell><ns0:cell>0.0873</ns0:cell><ns0:cell>0.1573</ns0:cell><ns0:cell>0.66</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0465</ns0:cell><ns0:cell>0.0530</ns0:cell><ns0:cell>0.1230</ns0:cell><ns0:cell>0.61</ns0:cell><ns0:cell>2.20</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0475</ns0:cell><ns0:cell>0.0510</ns0:cell><ns0:cell>0.1310</ns0:cell><ns0:cell>0.56</ns0:cell><ns0:cell>2.13</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0415</ns0:cell><ns0:cell>0.0511</ns0:cell><ns0:cell>0.1211</ns0:cell><ns0:cell>0.55</ns0:cell><ns0:cell>2.01</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>0.0510</ns0:cell><ns0:cell>0.0271</ns0:cell><ns0:cell>0.0971</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell>1.70</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 5 (on next page)</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Formulas verification summary tableWhere: (1 -17) -number of formula; d Hmax -an error of the depth of the water above the deepest point of the scour calculation; ; d Ls -an error of the scour length calculation; d Le -an error of the distance between the deepest point of the scour and the end of reinforcement calculation.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Straube</ns0:cell><ns0:cell>Straube</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Straube</ns0:cell><ns0:cell>Straube</ns0:cell><ns0:cell>original</ns0:cell><ns0:cell>optimized</ns0:cell></ns0:row><ns0:row><ns0:cell>No of test</ns0:cell><ns0:cell>q</ns0:cell><ns0:cell>h</ns0:cell><ns0:cell>z max</ns0:cell><ns0:cell>L s</ns0:cell><ns0:cell>original (Eq. 13)</ns0:cell><ns0:cell>optimized (Eq. 20)</ns0:cell><ns0:cell>(Eq. 13) relative</ns0:cell><ns0:cell>(Eq. 20) relative</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>L s</ns0:cell><ns0:cell>L s</ns0:cell><ns0:cell>error &#948;</ns0:cell><ns0:cell>error &#948;</ns0:cell></ns0:row><ns0:row><ns0:cell>-</ns0:cell><ns0:cell>m 2 &#8901; s -1</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[m]</ns0:cell><ns0:cell>[%]</ns0:cell><ns0:cell>[%]</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>0.020</ns0:cell><ns0:cell>0.050</ns0:cell><ns0:cell>0.084</ns0:cell><ns0:cell>1.25</ns0:cell><ns0:cell>2.10</ns0:cell><ns0:cell>1.63</ns0:cell><ns0:cell>68.4%</ns0:cell><ns0:cell>30.1%</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>0.032</ns0:cell><ns0:cell>0.065</ns0:cell><ns0:cell>0.097</ns0:cell><ns0:cell>1.65</ns0:cell><ns0:cell>2.71</ns0:cell><ns0:cell>2.09</ns0:cell><ns0:cell>64.5%</ns0:cell><ns0:cell>26.6%</ns0:cell></ns0:row><ns0:row><ns0:cell>3</ns0:cell><ns0:cell>0.021</ns0:cell><ns0:cell>0.050</ns0:cell><ns0:cell>0.071</ns0:cell><ns0:cell>1.35</ns0:cell><ns0:cell>1.93</ns0:cell><ns0:cell>1.50</ns0:cell><ns0:cell>43.3%</ns0:cell><ns0:cell>10.8%</ns0:cell></ns0:row><ns0:row><ns0:cell>4</ns0:cell><ns0:cell>0.025</ns0:cell><ns0:cell>0.062</ns0:cell><ns0:cell>0.068</ns0:cell><ns0:cell>1.47</ns0:cell><ns0:cell>2.03</ns0:cell><ns0:cell>1.56</ns0:cell><ns0:cell>38.1%</ns0:cell><ns0:cell>6.1%</ns0:cell></ns0:row><ns0:row><ns0:cell>5</ns0:cell><ns0:cell>0.021</ns0:cell><ns0:cell>0.050</ns0:cell><ns0:cell>0.058</ns0:cell><ns0:cell>1.43</ns0:cell><ns0:cell>1.73</ns0:cell><ns0:cell>1.34</ns0:cell><ns0:cell>20.7%</ns0:cell><ns0:cell>6.6%</ns0:cell></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell>0.028</ns0:cell><ns0:cell>0.071</ns0:cell><ns0:cell>0.083</ns0:cell><ns0:cell>1.25</ns0:cell><ns0:cell>2.37</ns0:cell><ns0:cell>1.81</ns0:cell><ns0:cell>89.9%</ns0:cell><ns0:cell>45.1%</ns0:cell></ns0:row><ns0:row><ns0:cell>7</ns0:cell><ns0:cell>0.030</ns0:cell><ns0:cell>0.070</ns0:cell><ns0:cell>0.095</ns0:cell><ns0:cell>1.90</ns0:cell><ns0:cell>2.62</ns0:cell><ns0:cell>2.01</ns0:cell><ns0:cell>38.0%</ns0:cell><ns0:cell>5.7%</ns0:cell></ns0:row><ns0:row><ns0:cell>8</ns0:cell><ns0:cell>0.021</ns0:cell><ns0:cell>0.050</ns0:cell><ns0:cell>0.087</ns0:cell><ns0:cell>1.50</ns0:cell><ns0:cell>2.19</ns0:cell><ns0:cell>1.69</ns0:cell><ns0:cell>46.0%</ns0:cell><ns0:cell>12.9%</ns0:cell></ns0:row><ns0:row><ns0:cell>9</ns0:cell><ns0:cell>0.024</ns0:cell><ns0:cell>0.060</ns0:cell><ns0:cell>0.102</ns0:cell><ns0:cell>1.55</ns0:cell><ns0:cell>2.53</ns0:cell><ns0:cell>1.94</ns0:cell><ns0:cell>63.0%</ns0:cell><ns0:cell>25.3%</ns0:cell></ns0:row><ns0:row><ns0:cell>10</ns0:cell><ns0:cell>0.030</ns0:cell><ns0:cell>0.070</ns0:cell><ns0:cell>0.110</ns0:cell><ns0:cell>1.63</ns0:cell><ns0:cell>2.86</ns0:cell><ns0:cell>2.19</ns0:cell><ns0:cell>75.5%</ns0:cell><ns0:cell>34.4%</ns0:cell></ns0:row><ns0:row><ns0:cell>11</ns0:cell><ns0:cell>0.027</ns0:cell><ns0:cell>0.065</ns0:cell><ns0:cell>0.084</ns0:cell><ns0:cell>1.98</ns0:cell><ns0:cell>2.35</ns0:cell><ns0:cell>1.80</ns0:cell><ns0:cell>18.6%</ns0:cell><ns0:cell>9.0%</ns0:cell></ns0:row><ns0:row><ns0:cell>12</ns0:cell><ns0:cell>0.020</ns0:cell><ns0:cell>0.060</ns0:cell><ns0:cell>0.069</ns0:cell><ns0:cell>1.52</ns0:cell><ns0:cell>1.88</ns0:cell><ns0:cell>1.44</ns0:cell><ns0:cell>23.9%</ns0:cell><ns0:cell>5.1%</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Mean relative error</ns0:cell><ns0:cell>49.2%</ns0:cell><ns0:cell>18.1%</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 7 (on next page)</ns0:head><ns0:label>7</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head /><ns0:label /><ns0:figDesc>Geometric scour parametersH max = z max + h [m] L s [m] L e [m]</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Field measurements</ns0:cell><ns0:cell>2.43</ns0:cell><ns0:cell>13.8</ns0:cell><ns0:cell>5.2</ns0:cell></ns0:row><ns0:row><ns0:cell>Author:</ns0:cell><ns0:cell cols='3'>Calculations results using Eqs. (1), (7)-(21) (error %)</ns0:cell></ns0:row><ns0:row><ns0:cell>Rossinski</ns0:cell><ns0:cell>(1) 3.39</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>(39.5 %)</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>M&#252;ller</ns0:cell><ns0:cell /><ns0:cell>(7) 10.3 -12.1</ns0:cell><ns0:cell>(8) 5.0-6.1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(12.3%)</ns0:cell><ns0:cell>(0%)</ns0:cell></ns0:row><ns0:row><ns0:cell>M&#252;ller</ns0:cell><ns0:cell /><ns0:cell>(10) 11.6 -17.5</ns0:cell><ns0:cell>(11) 5.7-8.6</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(0%)</ns0:cell><ns0:cell>(9.6%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Straube</ns0:cell><ns0:cell /><ns0:cell>(13) 21.7</ns0:cell><ns0:cell>(15) 4.8</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(57.2%)</ns0:cell><ns0:cell>(7.7%)</ns0:cell></ns0:row><ns0:row><ns0:cell>(Optimized) Straube</ns0:cell><ns0:cell /><ns0:cell>(20) 13.8</ns0:cell><ns0:cell>(21) 4.3</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(0.2%)</ns0:cell><ns0:cell>(16.6%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Tajarmovi&#269;</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>(17) 4.8</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>(159.6%)</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)Manuscript to be reviewed</ns0:note> <ns0:note place='foot' n='2'>PeerJ reviewing PDF | (2020:03:46700:1:1:NEW 3 Sep 2020)</ns0:note> </ns0:body> "

" 159 Nowoursynowska Street 02-776 Warsaw telephone: 22 59 35010 e-mail address: [email protected] August 28th, 2020 Dear Editors We thank all the Reviewers and the Academic Editor for having read our text carefully and for the accurate comments that made it possible for us to improve our paper. All comments have been considered by us, opening a scientific discussion and enhancing the authors view on the subject matter. The effect of our considerations is included in this rebuttal letter. We believe that the manuscript is now suitable for publication in PeerJ. DSc. Marta Kiraga Doctor of Technical Sciences in the field of Civil Engineering On behalf of all authors. Editor comments (Timothy Scheibe) Dear Editor, This rebuttal letter will repeatedly refer to the manuscript with tracked changes. To locate the phrases this text refers to, we kindly ask you to display a view of all the changes, including formatting, as shown in the figure below: It is important to show all revisions inline, not in balloons. Editors remarks: We have received three significant reviews, all of which recommend major revision. In my opinion, although there are many comments that require attention, none should be problematic to the authors. Please provide detailed responses to each reviewer comment, indicating how the comment was addressed in the revised manuscript. One theme to pay particular attention to is the generality of the results; multiple reviewers raised questions along these lines and the authors should pay attention to properly framing their conclusions accordingly. Thank you for the opportunity to improve our article and to resubmit it. Within the revised article we take into account all the comments of the reviewers. We paid special attention to the reconstruction of the article's structure, suggested by one of the reviewers (Michał Szydłowski). The original layout consisted of two parts: identification of formula parameters and its validation on the field data, while in its current form the article also includes the validation of obtained formulas for independent data. Also conclusions part was rebuilt and new insight was made. [# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.  It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter.  Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #] We have referred in the article to all the comments made in the reviews. I kindly ask you to check the detailed annotations, included within the present rebuttal letter, in relation to the comments of each of the reviewers. [# PeerJ Staff Note: The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title) #] We have sent our article for linguistic proofreading, done by an outer company. A professional translator made a correction of the article. Reviewer 1 (Michał Szydłowski) Dear Reviewer, This rebuttal letter will repeatedly refer to the manuscript with tracked changes. To locate the phrases this text refers to, we kindly ask you to display a view of all the changes, including formatting, as shown in the figure below: It is important to show all revisions inline, not in balloons. Reviewer remarks: Basic reporting The article must be written in English and must use clear, unambiguous, technically correct text. The article must conform to professional standards of courtesy and expression. After reading the article, in my opinion the English language in the text is far from perfect. Although English is not my native language, I have the impression that the sentence structure and some expressions used in the article correspond to the structure of the Polish language. Undoubtedly, the text requires proofreading and the naming of technical terms by a native speaker. The parts of text and phrases that I have doubts about are marked in yellow in the text (PDF). We have sent our article for linguistic proofreading, done by an outer company. A professional translator made a correction of the article. The pdf file you sent us has been carefully studied and the comments marked in yellow have been considered. The correct vocabulary has been found for the technical expressions with support of the professional translator. You can find attached the effect of proofreading works within the comparative document (revised manuscript with tracked changes). The article should include sufficient introduction and background to demonstrate how the work fits into the broader field of knowledge. Relevant prior literature should be appropriately referenced. The subject matter of the publication concerns the technical aspects of the operation of hydro-engineering structures. It discussed the impact of a small weir on the stability of the river bed downstream the water structure and the formation of local scour. The article was submitted to the Environmental Science section of PeerJ Journal. The authors noted that the content of the article falls within the scope of the following topics: Coupled Natural and Human Systems, Natural Resource Management, Ecohydrology, Environmental Impacts Food, Water and Energy Nexus. In my opinion, the introduction lacks an analysis of the relationship between the main topic of the article (local scouring) and the listed topics of the Environmental Science section. It would be necessary to do additional research showing the relationships between them. Research in world literature should be significantly deepened in this area. Thank you for your comment, the complementary rebuilding of the Introduction part creates an opportunity to improve the reader's introduction to the subject of water construction and its role in shaping the environment. At first sight, the subject matter does not seem to coincide. Sometimes the links between scientific areas are not obvious. However, we have tried to demonstrate the environmental aspect of the hydrotechnical structures introducing, especially small ones (lines 77-89; 124-155; 174-177 of the manuscript with tracked changes). We have reviewed the literature of additional sources, enabling the technical and natural aspects of hydraulic engineering to be linked together, supplementing the Introduction section with the information gathered during the literature review, for example Cullen 1991; Hossain 1992; Lopardo & Seoane 2004; Ochman & Kaszubkiewicz 2004; Siwicki, Urbański 2004; Lupandin 2005; Szydłowski & Zima 2006; Koskinen, Leino & Riipinen 2008; Jordaan 2009; Rasekh, Afshar & Afshar 2010; Syvitski et al. 2010; Zobeyer et al. 2010; Boix et al. 2012; Hockley et al. 2013; Liao & Cotel 2013; Pagliara & Kurdinstani 2013; Mioduszewski 2014; Smith, Goodwin & Nestler 2014; Hauer et al. 2018; Lee & Hong 2019. The structure of the submitted article should conform to an acceptable format of ‘standard sections’. Significant departures in structure should be made only if they significantly improve clarity or conform to a discipline-specific custom. The article structure is correct. Thank you, this is a good guide for our further publication work. Figures should be relevant to the content of the article, of sufficient resolution, and appropriately described and labeled. The quality of the drawings is acceptable. The charts 9 and 10 are too small. Horizontal axes should be definitely longer than vertical. In addition, the information on these charts is not clear. For example, it is not known what the fields colored blue and red are. It is not known how to interpret the black horizontal dotted line (bottom edge, Fig. 9). In my opinion the figures 9 and 10 should be improved. Figures 9 and 10 have been enlarged and their height-to-width ratios were changed as indicated. Additionally, the information presented in the figure has been adjusted to the reviewer's comments, that is: - the black horizontal dotted line was described on the figure - the blue and red fields have been removed, because of ambiguity they rose. Experimental design The submission must describe original primary research within the Aims & Scope of the Journal. The article describes the original results of laboratory tests in the field of water engineering. The authors did not explicitly indicate the relationship of their research with the domain of Environmental Sciences. I leave the decision to the Editor whether the topic of the work falls within the journal area of interest (Aims & Scope). In my opinion it would be needed to do additional research showing the relationships of authors’ research with the field of Environmental Sciences before acceptance. As mentioned above, an effort has been made to demonstrate the environmental aspect of the construction of water structures, emphasizing their role in shaping the retention of the valley areas (lines 77-89; 174-177 of the manuscript with tracked changes). Additionally, the natural effects of the local scour formation were wider described (lines 124-155 of the manuscript with tracked changes). As the literature review demonstrates publications usually focus on the negative consequences of this process (for example cited in the present article: Ślizowski and Radecki-Pawlik 2003; Barbhuiya and Dey 2004; Zobeyer et al. 2010; Urbański and Hejduk 2014; Singh, Devi and Kumar 2020), whereas present article, in the corrected form, also deals with the positive aspects of local scouring, which, given the small number of publications on the subject (the widest description in Siwicki and Urbański 2004), is a valuable research direction in environmental sciences field. The submission should clearly define the research question, which must be relevant and meaningful. The knowledge gap being investigated should be identified, and statements should be made as to how the study contributes to filling that gap. The research question is clearly defined. The paper focuses on the verification of some empirical formulas to estimate the scour dimensions in the case of local scouring processes downstream the small hydro-engineering structure (weir). The knowledge gap being investigated was identified as a uncertainty of the most often used formulas. Thank you for this remark, we agree about knowledge gap identification, however the importance of our considerations is increased by the verification procedure on independent data and, finally, validation for field conditions. The investigation must have been conducted rigorously and to a high technical standard. Laboratory experiments were carried out using well known measurement methods with appropriate measurement procedures. High-class measuring equipment with sufficiently good accuracy was used. Thank you, this is a promising remark for our further experimental works. Methods should be described with sufficient information to be reproducible by another investigator. The test stand and the course of the experiments were described in detail. It is possible to repeat it by other researchers. Thank you for this remark, we tried to describe the position and procedures as precisely and faithfully as possible. Validity of the findings The data should be robust, statistically sound, and controlled. The article presents the results of 29 measurement series. The way experiments are carried out and their results are beyond doubt. However, according to the research description, each experiment was carried out only once. This means that the authors have not examined (or do not mention in this article) the issue of repeatability of the results obtained. If the experiments were not repeated, then there is doubt as to their statistical quality. This requires explanation and commentary from the authors. We agree with this remark. The material presented does not give rise to a statement of repeatability of results obtained in a single measurement series. However, the construction works on the of the prototype of the laser scanner device, included the procedure of device calibration, as well as many trial, partial, incomplete measurements or those rejected for other reasons. Therefore, we dispose unpublished data from, among others, three eight-hour measurement series, each performed three times, on different days, assuming the same hydraulic conditions. The procedure is fully described in the improved version of the article. Please see the lines 457-472 of the manuscript with tracked changes. Differences in the bottom formation were shown in the figures 7 a,b,c. Quickly summarizing the obtained results, included in improved manuscript, differences in geometric parameters of local scour range of 0.3 - 1.9% in the maximum depth of the scour hole and 2.2-4% in the range of the average depth of the scour. Slightly more significant deviation was connected with scour parameters connected with its length: relative error ranging within 0.5 – 13.6% was met in total scour length and 3.0 – 16.6% in the case of the distance from the end of reinforcement to the deepest scour point. The data on which the conclusions are based must be provided or made available in an acceptable discipline-specific repository. The authors have included all measurement data in a shared database. However, in some places there are descriptions in Polish, which makes data analysis difficult. All descriptions should be in English. Descriptions in the shared database were translated into English. The conclusions should be appropriately stated, should be connected to the original question investigated, and should be limited to those supported by the results. In particular, claims of a causative relationship should be supported by a well-controlled experimental intervention. Correlation is not causation. There is some doubt about the conclusion about the optimized Straube formula. All available (made by the authors) measurement data were used to determine (optimize) the parameters of this formula. This process can be called identification. Next, validation of the optimized formula was performed based on field test (Czarna Weir). Unfortunately, the research process lacked an intermediate step - verification of the obtained formula. Such verification should be carried out before validation, using independent measurement results other than those used for the identification. Authors should fill this gap in the research procedure. In fact, the original version of the article lacked a verification procedure for independent data, which, as the reviewer points out, should appear before validating the formula on field data. Such a verification procedure undoubtedly reinforces the conclusion about the applicability of the formula and the appropriateness of the conducted optimization of its parameters. Therefore, following the valuable reviewer's remark, the methodology was supplemented by the verification of optimized formulas for independent data (please see the lines 445 – 447; 539 – 559; 624 – 629 of the manuscript with tracked changes. Verification of the optimized Straube formula was performed on independent data published in 2010 by Gaudio and Marion. The similarity of Gaudio and Marion test stand and the flume come down to used bed material (sand), the shape of the flume (rectangular, 60-cm width), the same order or magnitude of unit discharges and the transversal type of water structure. Full description of the test stand properties is available within the manuscript. A mean relative error δ = 49.2% was obtained for the original form of Straube formula for Gaudio and Marion data, whereas the application of the optimized Straube formula demonstrated a better description of the data obtained in the laboratory, characterized by an error mean relative δ equal to 18.1%. Comments for the Author Some remarks As a result of optimization (identification) of the Straube formula, the authors received positive values of parameters c, d and p (equations 18, 19 and 20, 21). According to the original formula (equations 13.15), their value is negative. This seems a mistake. Thank you for catching the error. We agree, there was a mistake in the formulas 18, 19, 20 and 21. The obtained parameters should be negative. Fortunately, calculations were made using the correct formula, which can be checked in the submitted raw data. All units should be written using the exponential form. There is a lot of wrong forms in the paper. All the wrong forms are revised and corrected using exponential forms. Authors should avoid citing Polish literature if it does not concern publication of research carried out by Polish researchers. It is not allowed to cite and place in the bibliography publications with titles in Polish - Dabkowski et al., 1982. For classic formulas, the first sources should be given. DSc. Adam Carpenter (Reviewer 2) had the same observation. So we give you a collective answer. Your remark has mobilized us to extensive research and to study archival sources. Many of them come from the sixties, seventies of the last century, however they have neither ISSN, nor DOI or even English language versions of the titles required today. Therefore, each of the formulas was addressed to the first source, without the mediation of Dąbkowski's publication, cited in the first version of the present manuscript. Annotated manuscript The reviewer has also provided an annotated manuscript as part of their review. Thank you for your effort aiming at improving the quality of our article. All marked errors have been corrected. The translator has been informed to pay special attention to the marked fragments. Reviewer 2 (Adam Carpenter) Dear Reviewer, This rebuttal letter will repeatedly refer to the manuscript with tracked changes. To locate the phrases this text refers to, we kindly ask you to display a view of all the changes, including formatting, as shown in the figure below: It is important to show all revisions inline, not in balloons. Reviewer remarks: Basic reporting The overall reporting within this paper appears sound. There are some clarifications can be made throughout the text to improve clarity and understandability of the work. Thank you for your effort aiming at improving the quality of our article. All marked errors have been corrected and ambiguities you pointed are clarified in the present form of the article. Please see the following explanations and the version of the article with tracked change. The abstract and body of the paper both reference “many years” of disagreement on parameters that influence size and depth of scouring but does not provide sufficient evidence of this ongoing disagreement (whether through published papers or through reporting other evidence). A discussion of why these previous papers could not come to the same conclusions as this paper would be helpful Your guidance led us to an in-depth review of the literature in this area and to formulate valuable comments in our opinion. New excerpts from the text were introduced to the Introduction, please take a look at lines 185-233 and 238 - 244 of the manuscript with tracked changes). In order to broaden the subject matter additional sources were reviewed: Lacey 1946; Ahmad 1953; Chabert & Engeldinger 1956; Breusers & Raudkivi 1991; Gaudio & Marion 2003; Barbhuiya & Dey 2004; Urbański 2008; Epely-Chauvin, De Cesare & Schwindt 2015; Kiraga & Popek 2016; Nouri Imamzadehei et al. 2016; Al-Husseini, Al-Madhhachi & Naser 2019; Al-Mohammed, Jassin & Abbas 2019; Kiraga & Popek 2019; Taha, El-Feky & Fathy 2020 which full reference data are given within the present version of the article. Please also see comments to the authors. Thank you for all the comments, we tried to develop the article comprehensively to meet the requirements. Each comment for the authors has been marked by us with a notice below, in the part Comments to the authors. Experimental design The experimental design of the paper is clear, although would benefit from being introduced earlier in the text and the abstract and introduction referencing the intended work more specifically. Given the several components (laboratory work, field work, and several stages of analysis, a clear description of the procedures from beginning to end would make the work more easily replicated. Thank you for the remark. We want our procedures descriptions to be understandable and the experiments to be readily repeatable. This creates later opportunities to establish cooperation, including international partnership, and the conclusions drawn during the research and verified by other research teams become more relevant. In the original version of the article we tried to provide as much data as possible, including both the geometry of the flume itself, granulometry of the bedload, as well as detailed characteristics of the hydraulic structure models. We have described in detail the hydraulic specifications, including the flow regime, described by the Froude number, and the conditions under which each test series is carried out, providing details of the discharges, unit discharges, duration of the test series, water level in a specific reference section. Also the procedures of test stand preparation is fully described. Suggesting your comment, the part Materials & Methods was developed and rebuilt. We have supplemented the description with details, which we feel are important to repeat the experiments - check please lines 270 – 280; 291 – 313; 309-313; 445-447. Validity of the findings The paper makes many good points and has a good experimental design with the overall framework supported by the material presented. However, there are some items presented that could benefit from additional clarification, to help prevent unintentionally misrepresenting the paper as more than it is. As discussed in “comments to the author,” the field evaluation of these equations would be greatly improved by either adding evaluation of additional structures and events, or by caveating the conclusions that these conclusions were tested against a single data point. As the first reviewer, Michał Szydłowski, writes, in the original version of the article there was no verification of optimized formulas prior to their validation in the field. Due to Mr. Szydłowski's remark, the procedure of verifying formulas on independent data was conducted. I believe this may be a reference to your comment as well – in fact, the inclusion of independent data is very advantageous for the reception of the article and its scientific value. Allow me to partially repeat here the response to Professor Szydłowski's remark, as your remark concerns the same issue. The methodology was supplemented by the verification of optimized formulas for independent data (please see the lines 445 – 447; 539 – 556; 624 – 629 of the manuscript with tracked changes. Verification of the optimized Straube formula was performed on independent data published in 2010 by Gaudio and Marion. The similarity of Gaudio and Marion test stand and the flume come down to used bed material (sand), the shape of the flume (rectangular, 60-cm width), the same order or magnitude of unit discharges and the transversal type of water structure. Full description of the test stand properties is available within the manuscript. A mean relative error δ = 49.2% was obtained for the original form of Straube formula for Gaudio and Marion data, whereas the application of the optimized Straube formula demonstrated a better description of the data obtained in the laboratory, characterized by an error mean relative δ equal to 18.1%. Comments for the Author - Overall, my opinion is that this is a generally well-written paper with the potential to add to the literature. With revision, I believe this paper will be suitable for publication in this journal. Thank you for this remark, it is very encouraging to continue the experiments related to the small water structures subject - inconspicuous and essentially ignored, yet shaping such valuable valley retention, especially in areas affected by periodic or permanent drought. - Line 68: This first sentence of the introduction is confusing as written. It refers to “the river” several times but does not indicate which river (to speak generally, likely the correct phrasing would be by discussing “rivers” broadly). There also appears to either be a word missing or some other issues, as “structures unavailable influences” is likely supposed to be either “structures causes unavoidable influences on…” or perhaps “structures unavoidably influences”. A significant part of the Introduction has been rebuilt, taking into consideration all reviewers' suggestions. The paragraph to which this remark applies has also been rebuilt. I hope that its current form will meet your expectations. Please check the lines 98-111 of the manuscript with tracked changes. A lot of ambiguity in the text was caused by the language barrier – we believe that ordering a professional check and translation of problematic parts of the text solves this problem. - Lines 69-71: These lines discuss several impacts of these structures, and are written as absolutes (that is to say, that these impacts *always* occur). It is not clear if that was intended or if these are meant to be common or likely (but not certain) impacts. Additionally, this list of impacts is not common knowledge and should be properly cited. Thank you for your remark, we want our articles to be based on reliable sources of scientific information. However, as you noticed, a simplifying shortcut has crept in here. As a matter of fact, the fact is that the damming of the river with a hydraulic structure will change the flow regime, will impound the water level upstream and will pass the discharge only partially through the span of the structure. However, morphodynamical processes (erosion, accumulation) will not necessarily be triggered. The sediment transport will be started only when the critical shear stress at the bed (Shields number) is exceeded. The material transported with the approaching flow will be accumulated when the force of gravity exceeds the force carrying the grain. A number of factors, widely described within the article, will influence the local scour formation downstream the structure (within which turbulence forces are highlighted). In connection with the unclearness, a part of the description has been reworked, please check the lines 98-111 of the manuscript with tracked changes, which explain occurring processes. The following sources were used to build the description: Graf 1989; Szydłowski & Zima 2006; Pagliara & Kurdinstani 2013; Zobeyer et al. 2010; Lee & Hong 2019, which full bibliographic data can be read in the publication. - Lines 71-72: Especially at its first use, it may be preferable to clarify that “directly below” the structure refers to the area just downstream, as opposed to the area underneath the structure. We agree with the remark, this is not the best expression. “Directly below” may indicate a process taking place underneath the closure, gate etc. Therefore, we used just “downstream” instead of “directly below”. - Line 73: Is “permanently” supposed to be “potentially” or some other word here? It would appear that although this damage can be considerable, isn’t necessarily permanent. Again, a linguistic awkwardness. Fortunately, our translator has managed to choose the right vocabulary (please check the line 110 of the manuscript with tracked changes). - Line 89: Please define “recognition laws” and if appropriate provide a reference. The sentence referred to in this comment in the original version of the manuscript is too simplistic, and we agree that it should be clarified. Therefore, it has been rebuilt and extended. Please check the lines 165 – 175; 186-196 of the manuscript with tracked changes. Factors that influence a scour hole shape, that could be defined as mentioned “recognition laws”, are listed: -- related to the flume geometry (width, depth, bed inclination); -- related to the type and geometry of the structure (type of structure, reinforcement construction, dimensions of upstream and downstream part of the structure elements); -- characterizing water flow conditions (flow rate, average speed, hydraulic gradient, bed shear stress, flow resistance), -- water properties (density, viscosity), -- characterizing the flume material (grain size, grain distribution, density, porosity, roughness), -- characterizing the conditions of sediment transport (critical velocity, critical shear stress, sediment transport intensity), -- time. Abovementioned list was built basing on author’s studies within the doctoral thesis of one of the authors (Marta Kiraga) on the ground of following sources: Graf 1989; Breusers & Raudkivi 1991; and finally Kiraga & Popek 2019, which full bibliographic description could be found in the text. - Line 90-92: This section refers to seeking “universal principles” but does not acknowledge that the variety of hydraulic and built-environment conditions means that it’s possible that universal parameters may not be appropriate to this concern. These lines also discuss “many experiments” related to these concerns but does not cite said experiments In response to this comment, an extensive review of the literature has been carried out, emphasizing that the diversity of research conditions, taking into account the differences in the construction not only in the model but also in the field conditions, leads to the search for solutions typical for a given hydraulic structure. This diversity may be based on the introduction of dimensionless coefficients, typical for an investigated structure. Please check the lines 197-219 of the manuscript with tracked changes. Following sources were used for built the description: Lacey 1946; Ahmad 1953; Chabert & Engeldinger 1956; Franke 1960; Straube 1963; Tarajmovič 1966; Rossinski & Kuzmin 1969; Breusers & Raudkivi 1991; Graf 1998; Lenzi, Marion & Comiti 2003; Barbhuiya & Dey 2004; Ben Meftah & Mossa 2006; Kiraga & Popek 2016; Nouri Imamzadehei et al. 2016; Pagliara et al. 2016; Kiraga & Popek 2018; Al-Husseini, Al-Madhhachi & Naser 2019 which full bibliographic description could be found in the text. - Line 98: The use of “on the other hand” should have a phrase before it with “on one hand”. Otherwise, consider different phrasing. Thank you for your comments, actually the sentence needs to be corrected, so the form was changed with the help of a professional translator to make it grammatically correct (line 234-235 of the manuscript with tracked changes). - Line 117: This sentence discusses “chosen empirical formulas” but does not state how said formulas were chosen. With reference to this commentary, the information has been completed. Please check the line 262 of the manuscript with tracked changes. - Line 125: The phrase “basically comes down to” is a very casual phrase. Consider replacing with a more formal description of measurements made. Thank you for this comment, with such remarks we improve our vocabulary, suitable for scientific articles - not only in the context of this article, but also for future scientific papers. In connection with the recommendation to use a more formal vocabulary, the whole article has been reviewed, but in this particular case, the sentence has been reworked into a formula that can be seen in line 276 – 280 of manuscript with tracked changes. - Lines 128-131: This section describes a series of parameters for the structure of the experimental model water structure. Although these descriptions are welcome, they do not explain how the various parameters were chosen. The paper would be made stronger by providing these reasonings. We have supplemented this paragraph with a reference to the conditions prevailing in the field, i.e. to real hydraulic structures, observed in nature. Please check the lines 291-297 of the manuscript with tracked changes. There was also a spontaneous vision in the field to expand knowledge towards gated checks, done by Marta Kiraga. To be honest, the session was done during the holiday trip in the region of Polesie National Park in Poland, so I did not have any measuring equipment. However, we do not include them in the article because the field conditions and the lack of equipment did not allow for a precise determination of the bottom structure, that could bring scientific information for this article. Photo 1. Gated checks used on drainage ditches for agricultural purposes in Polesie. - Lines 168-171: This section provides a good description of LiDAR technology, but does not cite any references. I recommend referencing relevant work. Additionally, I believe that with appropriate references to information about LiDAR, Figure 6 will no longer be necessary and can be removed. The description was supplemented with references Jaboyedoff et al. 2010; Killinger 2014 which full bibliographic description could be found in the text. In the cited sources a description of the LiDAR system and its variety of applications could be found, therefore we agree that in the present form Figure 6 will no longer be necessary and was removed. - Lines 176-177: Although the description of the sustainable properties of the polyesters used is intellectually interesting, I’m not sure this description is relevant to the research and may be more distracting than helpful. As a matter of fact, the aspect of material science in the construction of hydraulic models and research stations is interesting to consider as a separate article, perhaps as a case study, in cooperation with scientists in the field. However, it is not necessarily relevant in this article. Therefore, this part of the text has been shortened, as you suggest (please check the lines 373-374 of the manuscript with tracked changes). - Lines 213-217: This section lays out a series of formulas. Although various formulas were noted as being from different authors (e.g. Shalash and Franke) a single source is cited for all equations. I recommend citing the original source for each of these groups of equations. Professor Szydłowski (Reviewer 1) had the same observation. So we give you a collective answer. Your remark has mobilized us to extensive research and to study archival sources. Many of them come from the sixties, seventies of the last century, however they have neither ISSN, nor DOI or even English language versions of the titles required today. Therefore, each of the formulas was addressed to the first source, without the mediation of Dąbkowski's publication, cited in the first version of the present manuscript. - Lines 292-310: It appears that this paper’s laboratory findings were validated against a single high flow event at a single water structure to demonstrate the correlation between modeled and measured scouring length and depth. Because the laboratory data is only compared to a single field example, the authors should consider reviewing all of their conclusions to assure they are not portrayed as being broadly applicable. Before the optimized equations could be considered to be representative, they would need to be compared to many events at many facilities at many events. That's right, paper’s laboratory findings were validated against a single high flow event at a single water structure. The conclusions concerning verification in field conditions apply only to this one event. We tried not to exaggerate the achievements, but to give and comment only the results of our calculations. Generalizing conclusion, the final one (lines 637 – 640 of the manuscript with tracked changes) concerns only that very high degree of adjustment of calculation results to field data (especially local scour length) provide an encouraging premise for further research. Reviewer 3 (Anonymous) Dear Reviewer, This rebuttal letter will repeatedly refer to the manuscript with tracked changes. To locate the phrases this text refers to, we kindly ask you to display a view of all the changes, including formatting, as shown in the figure below: It is important to show all revisions inline, not in balloons. Reviewer remarks: Basic reporting Basically, this study presents an optimization of current gate scour prediction formula coupling with the Monte Carlo method. The presentation is proper and clearly understood. Besides, the authors themselves developed a laser scanning to detect the shape of the scoured hole downstream the gate, which is interesting. Thank you for your critical review and appreciation of our own construction of the measuring device. Experimental design no comment Validity of the findings no comment Comments for the Author I think that the authors need to elaborate how the install the laser scanning technique. I personally recommend the publication as long as the authors can commit a major revision with respect to the following comments. Thank you for your comment. The article describes all the elements of the laser scanner - beginning from the general (the very principle of the scanner and data collection, referring to the sources: Jaboyedoff et al. 2010; Killinger 2014 whose full bibliographic data are included in the text of the article) to the details (movement system, material of the supports, data transmission system). Please check the lines 349-369 of the manuscript with tracked changes. The device itself has already been described in another authors' publication (Kiraga et al. 2018), which we also quote. We wanted to avoid duplication of content already published, limiting ourselves to the most important information about the device, while emphasizing the authors' own achievements in this area (lines 345-348 of the manuscript with tracked changes). - Please cite the recent-5-year studies in the introduction to show the progresses of the local scour due to a gated check. Thank you for this comment, it led us to an in-depth review of scientific sources. Literature concerning hydraulic structures subject is rich, however covers mostly the subject of large objects, such as dams or groynes and objects significantly affect the change of flow regime, i.e. weirs with gates, or without gates, but with highly positioned overflow edge. Therefore, an extensive search was made among the sources of the subject matter of small hydraulic structures, among which the greatest attention was paid to gated checks. Please check the lines 165-177 of the manuscript with tracked changes. A literature review demonstrated that also many studies on the subject of local scouring downstream hydraulic structures can be found in the literature in the recent years, but among them only a few focus on small hydraulic structures. Especially gated checks are a rare object of research, which, having regard to its role in shaping the valley retention and in agricultural applicability, is to the insufficiently covered knowledge gap. Following sources were used to build up the description: Lopardo & Seoane 2004; Boix et al. 2012; Sun, Wang & Wang 2012; Mioduszewski 2014; Pagliara et al. 2016; Khaple et al. 2017; Al-Husseini, Al-Madhhachi & Naser 2019; Singh, Devi & Kumar 2020; Li el al. 2020; Taha, El-Feky & Fathy 2020; Wang et al. 2020; Yan, Rennie, Mohammadian 2020), (Lopardo & Seoane 2004; Kiraga & Popek 2016; Odgaard 2017; Al-Suhaili, Abbood & Samir Saleh 2017; Kiraga & Popek 2018, which full bibliographic description could be found in the text. - Line 44-47: Please rephrase the statement. It is not clearly stated. All the reviewers pointed to many uncertainties connected with the wording. For our part, we could add that most of them are due to our far from ideal language skills. Fortunately, these problems were solved by hiring a professional translator. This was also the case here. The sentence referred to in the comment has been rewritten. Please check the lines 45-49 of the manuscript with tracked changes. We hope that in its current form it can be embedded in the article. - Line 48: Please rephrase “researchers still disagree on the parameters that influence its size and intensity”. The sentence referred to in the comment has been checked by the translator. Please check the lines 50-51 of the manuscript with tracked changes. We hope that in its current form it can be embedded in the article. - Line 49: the authors say that there are no universal methods for estimating its magnitude. Does this mean that universal methods have been developed in this study? Thank you, actually, this sentence may indicate that we have discovered universal principles that govern the phenomenon of local scouring, and this was not our intention at all. We wanted to prove something completely different - that there are no universal rules, but the use of formulas that take into account dimensionless coefficients, depending on the structure of the building, gives good results. That's why we have made this phrase more precise in the text. Please check the lines 51-57 in the abstract and 200-215 in the Introduction of the manuscript with tracked changes. - Line 51: what is the “great dams”? In order to avoid the embarrassing wording, we have replaced great dams with large weirs. We wanted to emphasize the range of objects that can be hidden under the concept of hydraulic structures. - Line 58: “was applied to describe the shape”. “describe” is not proper here. Please revise. “Describe” was replaced with “determine” here (line 66 of the manuscript with tracked changes). Yet another linguistic calque. - Line 63-64: “basing on Straube formula” should be “basing on the Straube formula”. Also, please pay attention to any place of the same issue. Thank you for this comment, this phrase has been repeated many times throughout the article. When checking the linguistic correctness of the article we paid special attention to it. - Line 68: “unavoidable influences” should be “unavoidably influences”. The formulation has been changed as suggested (line 98 of the manuscript with tracked changes). - Line 70: “Simultaneously erosion” should be “Simultaneously, erosion”. Similarly, please check the similar issues through the entire context. The formulation has been changed as suggested (lines 44, 103 of the manuscript with tracked changes). A professional translator also checked the text for grammatical correctness. - Line 74: “The increased erosion of a riverbed is an unfavorable and undesirable phenomenon not only due to the slow degradation of the riverbed”. Not only … but also… Please rephrase. This phrase is repeated many times throughout the text, however, care has been taken to ensure that it is used correctly in each case, that is in the form of “not only … but also…”. - Line 124-125: Why the flume is with no bed inclination downstream? This actually is not true for the real case. Please clarify the intention of this design consideration. A good point. The experimental flume on which we work is tiltable, that is, we can achieve a certain inclination, as is the case with natural watercourses. As a last option, we can align the sand level in such a way as to achieve the same inclination every time. I make it a point, however, that the second way would be much more laborious and time consuming. Already in the present form of conducting experiments, the test flume preparation, i.e. bedload levelling and its compaction was connected with works that took an hour, sometimes even two. This procedure was included into the present form in the article. Please check the lines 309 - 313 of the manuscript with tracked changes. To justify the lack of a fall in the bottom of the trough, we can add that lowland rivers that formed in alluvial depositions usually have gradients of 0.5 - 3%. If such a slope were to be reproduced in the present laboratory conditions, the difference in elevation of the bottom below the water structure would be 1 mm to a maximum of 1 cm (lines 272 - 275 of the manuscript with tracked changes). - Line 129-131: What’s the consideration of “the bottom was non-washable”? Is there any existing study for reference? In this case, “solid” is the better phrase. We just wanted to emphasize the various character of reinforcement material (line 284 of the manuscript with tracked changes), which is also shown in figures 2a, b; 3a, b. - The flume configuration: the sandy bed which the authors used the water to scour has a limiting downstream length. However, for a real case, the length should be long enough for scour. The downstream should have no impact on the scour pattern. So what’s the authors’ consideration? Thank you for this comment, however we were afraid of getting such. The length of the bottom of the scour was the biggest problem while writing this article. During the experiments, visually, the scour seemed to be clearly limited within the sandy part, which seemed to be long enough. Only when numerical results were delivered, determining the shape of the bottom, that clearly indicated that, if the length of the sandy bed had been increased, it is possible that the length of the local scour would have increased as well. We even thought to abandon its submission because of this. However, instead of postponing its submission, we started to think about the question - what is the criterion of whether the local scour in a given place occurs or not. The discussion led us to the statement: the field of the scour was considered to be an area where the bottom lowering exceeded 10% of its maximum depth in presumed time step. If another criterion was to be adopted, for example, a consideration of the scour hole area within the region where a bottom lowering exceeds 15 or more % of the maximum scour depth, a limiting effect of the sandy bottom downstream the structure could be avoided in some measurement series. Please check the lines 486-492 of the manuscript with tracked changes. Also, additional figure was prepared (figure 9). - What’s the drainage function with a pipe for? After each measurement series, the sand in the flume was dried for ca. 13 hours - the outflowing water was removed by using drainage pipe. This information was inserted into the Material and Methods part in lines 298-300 of the manuscript with tracked changes. - Line 135: “A pin water gauges were used” Grammar error. The authors need do proof-reading for the entire manuscript. That's right, we admit, here - and not only, as other reviewers have also pointed out - there was a grammatical error. Proof-reading of the entire manuscript was made. The number of grammatical errors is evidenced by a significant number of corrections made by a professional translator, which you can see in the manuscript with tracked changes. We believe that after corrections the text is suitable for publication. - Line 137: what’s the function of the moving disc probe besides the laser scanner device The moving disc probe was an additional measuring device. The probe was used to take measurements every set time step, while the device equipped with a laser scanner measured the shape of the bottom before water was introduced into the flume and after it was dried (using a drainage pipe). Measurement using the 'traditional' method, by this we mean a disc probe and an 'innovative' one (here laser scanning) gave the possibility of mutual control of results. In the past, the result of the comparison of the geometric parameters of the local scour using both methods was also published: Kiraga MJ, Razumnik M, Popek Z, Chmielewski L. 2018. Applying laser scanning technology to studying alluvial flume-bed topography in laboratory conditions. Acta Scientiarum Polonorum Formatio Circumiectus 17:69–84. Amendments are included in the lines 342-343 of the manuscript with tracked changes. - Line 137: “sandy bottom was measured with laser scanner device” should be “sandy bottom was measured with a laser scanner device”. Also revise the similar issues for the entire context. Proof-reading of the entire manuscript was made, including this phrase (line 306 of the manuscript with tracked changes). - Line 139: the unit of roughness coefficient is problematic. The unit in the present form of the article is corrected (line 318 of the manuscript with tracked changes). - Line 140: “during the studies” should be “during the study”. Proof-reading of the entire manuscript was made, including this linguistic awkwardness (line 317 of the manuscript with tracked changes). - Line 142: “the following range Qw = 0,010 – 0,045” should be “0.010 - 0.045”. The separating sign was changed from comma to dot. - Line 150: should be “due to the flow resistance”. Again, please pay attention to the use of “a, an, the” for the entire manuscript. This wording has been corrected (line 336 of the manuscript with tracked changes). - Line 188: zmax should be the elevation not the depth. Confused. To differentiate the meanings of the scour depth and the position of the water surface above it, two terms were introduced: -- local scour depth zmax -- water elevation before scour formation h However, it should be remembered that the water elevation is also related to the water depth (line 385 of the manuscript with tracked changes - see also Figure 4).. - Line 195: should be “finer than the reported particle size” not “under the reported particle size”. This wording has been corrected (line 393 of the manuscript with tracked changes). - Line 214: “ ” this way is not scientific for a research paper. I follow your comments and my submitted text carefully. However, in this case I don't really know what your comment is referring to. It may be a problem of not displaying some character you are quoting or quotation marks, but the line 214 of the original version of the manuscript contains neither an unusual character nor a quotation mark: I suspect here the problem of simply losing the character in your comment while copying to a shared message by the Editors. The only thing that doesn't really suit me here and what you could mean is a comma between Ls and Le. I changed it for the word 'and' - it doesn't change the meaning, but sounds much more appropriate for this type of publication (line 414 of the revised manuscript with tracked changes). - Line 220-227: the description of the Monte Carlo method to evaluate the optimal parameters of the prediction formula is poorly stated. The authors need to detail how the authors used the Monte Carlo method for their intention, for instance the procedures. That's right, the originally submitted version of the article lacks a description of the Monte Carlo sampling method, especially in the Material and Methods section. It is more extended in terms of the results obtained. However, we agree that the description needs to be completed and rebuilt. Therefore, we have rethought the structure of the article in terms of the description of this methodology. Please check the lines 422-444; 517-522; 617-619 of the manuscript with tracked changes. Again, I would like to thank all the reviewers for their critical comments and effort put into reviewing the article. I hope that we met the requirements. DSc. Marta Kiraga Doctor of Technical Sciences in the field of Civil Engineering On behalf of all authors. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>High impacts of COVID-19 are apparent in some countries with large tropical peatland areas, some of which are relatively poorly resourced to tackle disease pandemics. Despite this, no previous investigation has considered tropical peatlands in the context of</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The COVID-19 pandemic has caused unprecedented global disruption, at the time of writing infecting millions and killing hundreds of thousands of people across the globe <ns0:ref type='bibr' target='#b50'>(Dong, Du &amp; Gardner, 2020)</ns0:ref>. These health impacts, plus lockdowns and other measures to control the pandemic, have resulted in reduced economic activity and job losses, leading to potentially the worst global recession since the Great Depression <ns0:ref type='bibr'>(IMF, 2020)</ns0:ref>. While these global economic and social disruptions have had a positive, albeit likely temporary, impact on global carbon emissions (Le <ns0:ref type='bibr' target='#b123'>Qu&#233;r&#233; et al., 2020)</ns0:ref>, negative outcomes are widely expected for biodiversity conservation, research <ns0:ref type='bibr' target='#b35'>(Corlett et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b61'>Evans et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b85'>Hockings et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b130'>Lindsey et al., 2020)</ns0:ref>, and indigenous communities (IUCN, 2020; UN/DESA, 2020; UN/EMRIP, 2020).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed COVID-19 cases and deaths have been recorded for most tropical countries, with high numbers in some <ns0:ref type='bibr' target='#b50'>(Dong, Du &amp; Gardner, 2020)</ns0:ref>, and with testing shortfalls likely partially accounting for low reported case numbers for some other tropical nations, particularly in Africa <ns0:ref type='bibr' target='#b46'>(Ditekemena, 2020;</ns0:ref><ns0:ref type='bibr' target='#b159'>Nordling, 2020)</ns0:ref> . Many tropical nations are low-or middle-income countries, with weaker health systems and fewer resources to tackle the pandemic, generating further concerns. Some of the countries with high infection and mortality rates also have large remaining areas of tropical peatland, including Brazil, Peru, Ecuador and Indonesia <ns0:ref type='bibr' target='#b50'>(Dong, Du &amp; Gardner, 2020;</ns0:ref><ns0:ref type='bibr' target='#b166'>Page, Rieley &amp; Banks, 2011)</ns0:ref>. While currently reporting relatively few cases, the Congo Basin contains the world's largest tropical peatland area <ns0:ref type='bibr' target='#b41'>(Dargie et al., 2017)</ns0:ref>, is among the most poorly resourced to tackle disease pandemics in general <ns0:ref type='bibr' target='#b164'>(Oppenheim et al., 2019)</ns0:ref> and COVID-19 in particular (WB, 2020a), and the Democratic Republic of Congo (DRC) is projected to suffer a substantial number of cases and deaths <ns0:ref type='bibr' target='#b19'>(Cabore et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The importance of healthy tropical peatlands for carbon storage and emission mitigation, conserving biodiversity and providing ecosystem services for local communities is increasingly recognised <ns0:ref type='bibr' target='#b8'>(Baker et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b37'>Crump, 2017;</ns0:ref><ns0:ref type='bibr' target='#b41'>Dargie et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b49'>Dommain et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b86'>Hooijer et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b88'>Husson et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b166'>Page, Rieley &amp; Banks, 2011;</ns0:ref><ns0:ref type='bibr' target='#b175'>Posa, Wijedasa &amp; Corlett, 2011)</ns0:ref>, but to our knowledge no published study has specifically considered tropical peatlands and their inhabitants in the context of emerging infectious disease (EID), although some infectious disease studies have been conducted in tropical peatland areas (e.g., <ns0:ref type='bibr' target='#b229'>Vittor et al., 2006)</ns0:ref>. Addressing this gap is important because ongoing land-use change is reducing tropical peat-swamp forest (TPSF) coverage, while bringing an increasing number of human communities in close contact with peatlands <ns0:ref type='bibr' target='#b67'>(Field, van der Werf &amp; Shen, 2009;</ns0:ref><ns0:ref type='bibr' target='#b168'>Parish et al., 2008)</ns0:ref> and thus their biodiversity, although to date this has occurred far less in South American and African peatlands, compared to South-east Asia <ns0:ref type='bibr' target='#b40'>(Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017)</ns0:ref>. Such an assessment is made more urgent considering the ongoing COVID-19 pandemic and its potential impacts, and the global trend for increased EID event incidence <ns0:ref type='bibr' target='#b98'>(Jones et al., 2008)</ns0:ref>. It is also relevant for understanding and reducing the potential for emergence of, and impacts arising from, any future EIDs in tropical peatland nations.</ns0:p><ns0:p>Our goals in this paper are thus to present a preliminary synthesis of: (i) the potential for future EID (re-)emergence from tropical peatlands; (ii) potential threats to tropical peatland conservation and local communities from the current COVID-19 pandemic; and (iii) potential steps to help mitigate these risks. These goals cover a very broad range of potential topic areas, ranging from local livelihoods and food security, to habitat conservation efforts and scientific research, among others. A summary illustrating some key features relating to (i) and (ii) is provided in Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>. Although focused on tropical peatlands, many of the issues discussed and concerns raised will also be relevant to non-peatland areas in the tropics. This synthesis should thus be of interest to researchers, conservation/restoration/community project proponents, land managers and policy makers in the tropics, especially in but not restricted to peatland areas. While framed in the context of COVID-19, it is also pertinent to note that our discussion and many of the issues raised are in reality not uniquely linked to COVID-19, but rather relate more generally to (pandemic-related) sudden socio-economic shocks (e.g., economic recessions, border closures due to other causes, or extreme events related to climate change) that may occur in future.</ns0:p></ns0:div> <ns0:div><ns0:head>Survey Methodology</ns0:head><ns0:p>We conducted a scoping review of relevant literature in relation to the above goals. First, we conducted a structured search of scientific databases using search terms related to these topics (Table <ns0:ref type='table'>S1</ns0:ref>). This approach reduces potential for bias in awareness among our author team, but yielded very few potentially relevant studies (total n = 6, Table <ns0:ref type='table'>S2</ns0:ref>), owing to a lack of past studies concerning EIDs, including COVID-19, in the context of tropical peatlands. These results alone were insufficient to draw meaningful conclusions relating to any one of our goals, whereas conducting similarly structured database searches around every potential topic of relevance to EIDs/COVID-19 in the context of tropical peatlands would have been impractical, given the huge variety of potentially relevant topics relating to our goals, large number of countries containing tropical peatlands and the fact that these countries are not entirely covered in tropical peatlands, thus limiting potential to use individual tropical peatland nation names as search criteria. Consequently, we also conducted a less formally structured literature review, drawing on the subject knowledge, awareness of formal and informal literature sources, personal experiences and networks of our author team. In line with our scoping aims and very broad line of questioning, this did not employ strict inclusion and exclusion criteria, asides from excluding studies of no conceivable relevance to EIDs/COVID-19 and tropical peatlands. Such an approach allows us to draw relevant information from the many studies and reports of potential relevance that we were aware of that do not specifically concern EIDs/COVID-19 in the context of tropical peatlands and that did not therefore appear in our structured searches. The remainder of this manuscript is therefore focused on this much more informative less formally structured review, though findings from the structured search are integrated in relevant sections of the text.</ns0:p><ns0:p>To help ensure comprehensive coverage of the literature, and minimise geographic and subject bias in this review, our author team was developed to include natural and social scientists with substantial direct experience of, and familiarity with the literature in relation to, tropical peatland research in South-east Asia, Africa and South America (see, e.g., past reviews: <ns0:ref type='bibr' target='#b40'>Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b81'>Harrison et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b165'>Page &amp; Hooijer, 2016;</ns0:ref><ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017)</ns0:ref>. While we attempt to provide context across all four continents on which tropical peatlands are found, we acknowledge some bias towards South-east Asia, for which a far greater volume of information is available. Where possible we draw upon peer-reviewed and other highly reputable sources (e.g., UN reports), but owing to the COVID-19 pandemic's recent emergence and consequent paucity of such literature relating specifically to it in the context of the issues considered, we also draw upon pre-prints and media reports where peer-reviewed sources are unavailable or just provide general (rather than COVID-19 specific) support for a statement. We attempt to indicate such cases clearly and verify these from multiple sources, while noting that this reflects the uncertainty and rapid evolution of the pandemic and associated public debate.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Are Tropical Peatlands a Potential Source Habitat for Disease Pandemics?</ns0:head><ns0:p>Most EID events are dominated by zoonoses (60.3%), with the majority of these (71.8%) originating in wildlife, including Acquired Immunodeficiency Syndrome (AIDS), Severe Acute Respiratory virus (SARS), Middle East Respiratory Syndrome (MERS) and Ebola virus <ns0:ref type='bibr' target='#b98'>(Jones et al., 2008)</ns0:ref>, plus the novel COVID-2019, an ongoing global pandemic as we write this paper <ns0:ref type='bibr'>(Li et al., 2020)</ns0:ref>. In Africa, for example, 25 types of parasites, nine main types of viruses and eight types of bacteria have been reported as present in wild meat and communicable to humans <ns0:ref type='bibr' target='#b225'>(van Vliet et al., 2017)</ns0:ref>. The joint-first reported case of Ebola in 1976 is from a peatland area (Yambuku, DRC: <ns0:ref type='bibr' target='#b153'>Muyembe-Tamfum et al., 2012)</ns0:ref>, as is the most recent outbreak in May 2020 (Mbandaka, DRC: WHO, 2020c; Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>), and the cradle of the HIV/AIDS pandemic is believed to be around Kinshasa, DRC, another area with extensive peatlands <ns0:ref type='bibr' target='#b197'>(Sharp &amp; Hahn, 2011;</ns0:ref><ns0:ref type='bibr' target='#b252'>Worobey et al., 2008)</ns0:ref>. The risk of zoonotic EID emergence is positively correlated with high human population density and wildlife host richness <ns0:ref type='bibr' target='#b1'>(Allen et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b98'>Jones et al., 2008)</ns0:ref>; wild animal harvesting and/or movement of animals or body parts, leading to increased contact between wildlife vectors and humans <ns0:ref type='bibr' target='#b9'>(Bengis et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b97'>Johnson et al., 2020)</ns0:ref>; biodiversity loss <ns0:ref type='bibr' target='#b109'>(Keesing et al., 2010)</ns0:ref>; spread of non-indigenous vectors and pathogens; plus habitat encroachment, fragmentation and alteration <ns0:ref type='bibr' target='#b1'>(Allen et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b97'>Johnson et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b174'>Pongsiri et al., 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>INSERT FIG. 2 AROUND HERE</ns0:head><ns0:p>The natural habitat of tropical peatlands, tropical peat-swamp forest (TPSF), possesses a rich fauna and flora, including numerous vertebrate taxa known to represent zoonotic EID risk, such as bats, rodents, pangolins and primates <ns0:ref type='bibr' target='#b88'>(Husson et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b90'>Inogwabini et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b175'>Posa, Wijedasa &amp; Corlett, 2011)</ns0:ref>. Indeed, previous studies on primates and small mammals in Southeast Asian TPSF areas have recorded numerous parasite species found in humans and that are of medical concern <ns0:ref type='bibr' target='#b83'>(Hilser, 2011;</ns0:ref><ns0:ref type='bibr' target='#b84'>Hilser et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b139'>Madinah et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b139'>Madinah et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b160'>Nurcahyo, Konstanzov&#225; &amp; Foitov&#225;, 2017)</ns0:ref>, and surveys of bats from TPSF areas in Peru have detected high rates of Bartonella bacteria infection <ns0:ref type='bibr' target='#b7'>(Bai et al., 2012)</ns0:ref>, suggesting potential for disease transmission from humans to wildlife and zoonotic transmission from wildlife to humans in TPSF areas. Studies conducted in non-TPSF areas on species that are also found in TPSF support this conclusion (e.g., chimpanzee Pan troglodytes deaths from human paramyxoviruses in Ivory Coast: <ns0:ref type='bibr' target='#b111'>K&#246;ndgen et al., 2008)</ns0:ref>.</ns0:p><ns0:p>TPSF conversion, plus fire and wildlife harvesting brings more people into closer contact with peatland biodiversity. In South-east Asia, degradation, fragmentation and conversion of TPSF to agriculture has been particularly widespread, with the area of peatland in Malaysia, Sumatra and Kalimantan covered by TPSF declining from 76% (11.9 Mha) in 1990 to 29% (4.6 Mha) in 2015, with a concomitant increase from 11% (1.7Mha) to 50% (7.8Mha) of the area covered by agriculture over the same time period <ns0:ref type='bibr' target='#b145'>(Miettinen, Shi &amp; Liew, 2016)</ns0:ref>. This near doubling of agricultural extent has been driven by small-scale farmers (43-44%), plus industrial oil palm (39%) and paper pulp (11-26%) <ns0:ref type='bibr' target='#b145'>(Miettinen, Shi &amp; Liew, 2016;</ns0:ref><ns0:ref type='bibr' target='#b248'>Wijedasa et al., 2018)</ns0:ref>. Further forest habitat fragmentation is expected if all planned road and rail infrastructure development projects on Kalimantan proceed as proposed, with a projected reduction in landscape connectivity from 89% to 55% <ns0:ref type='bibr' target='#b0'>(Alamgir et al., 2019)</ns0:ref>. Such threats are not limited to South-east Asia. For example, although there is currently limited encroachment into the TPSFs of the Peruvian Amazon from industrial agriculture and infrastructure development <ns0:ref type='bibr' target='#b128'>(Lilleskov et al., 2019)</ns0:ref>, these threats are present and considerable <ns0:ref type='bibr' target='#b8'>(Baker et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017)</ns0:ref>, and could increase the likelihood of EID emergence from the Amazon in the near future <ns0:ref type='bibr' target='#b58'>(Ellwanger et al., 2020)</ns0:ref>. Likewise, projected mining permits, gas and oil exploration, timber and palm oil concessions, associated road construction and changing rainfall patterns due to global warming pose potential risks to the relatively undisturbed TPSF of the Congo Basin and its wildlife <ns0:ref type='bibr' target='#b40'>(Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b77'>Haensler, Saeed &amp; Jacob, 2013;</ns0:ref><ns0:ref type='bibr' target='#b146'>Miles et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b242'>Wich et al., 2014)</ns0:ref>. In the Republic of Congo (RoC), health authority reports claim that several recent malaria infections appeared in swampy forests for the first time following opening of logging trails (R&#233;daction, 2020), highlighting the potential for increased EID risk should these peatlands experience greater future encroachment.</ns0:p><ns0:p>Patterns of human-wildlife contact and wild meat hunting in tropical peatlands provide further potential for disease spill-over events (during which a pathogen from one species moves into another) from wildlife to humans to occur. Human population density in tropical peatland areas is typically not high (e.g., Peru: <ns0:ref type='bibr' target='#b128'>Lilleskov et al., 2019;</ns0:ref><ns0:ref type='bibr'>Congo Basin: Dargie et al., 2019)</ns0:ref>, but some large population centres are found close to TPSF (e.g., Jambi and Palangka Raya, Indonesia; North and South Selangor, Malaysia; Iquitos, Peru; Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>), and certain human activities increase contact between people and potential animal vectors in TPSF. Wildlife harvesting for consumption and trade is common in tropical forest nations <ns0:ref type='bibr' target='#b64'>(Fa, Currie &amp; Meeuwig, 2003;</ns0:ref><ns0:ref type='bibr' target='#b156'>Nielsen et al., 2018)</ns0:ref>, including in TPSF areas. For example, in Central Kalimantan, Indonesia, Pteropus vampyrus (Linn.) fruit bats are captured in TPSF areas and transported to local markets for sale as wild meat <ns0:ref type='bibr' target='#b80'>(Harrison et al., 2011)</ns0:ref>. High contact levels between bats, hunters and vendors occur, with hunters and vendors frequently bitten and most bites drawing blood, raising concerns regarding potential zoonotic disease transmission <ns0:ref type='bibr' target='#b80'>(Harrison et al., 2011)</ns0:ref>, though recent anecdotal observations suggest that local demand and trade has decreased as a likely result of the pandemic (R. <ns0:ref type='bibr'>Dwi et al., 2020, pers. comm.)</ns0:ref>. Other species are commonly harvested for commercial trade in tropical peatland nations, often for sale in dense population areas, including turtles <ns0:ref type='bibr' target='#b191'>(Schoppe, 2009)</ns0:ref> and primates <ns0:ref type='bibr' target='#b158'>(Nijman et al., 2015)</ns0:ref> in Indonesia; plus pangolins and numerous other species from across South-east Asia <ns0:ref type='bibr' target='#b157'>(Nijman, 2010)</ns0:ref>. Similar unsustainable hunting has been reported in the Congo Basin <ns0:ref type='bibr' target='#b64'>(Fa, Currie &amp; Meeuwig, 2003;</ns0:ref><ns0:ref type='bibr' target='#b176'>Poulsen et al., 2009)</ns0:ref>. In the Peruvian Amazon TPSF, tapirs (Tapirus terrestris, Linn.), primates, rodents and other mammals are commonly hunted by local communities <ns0:ref type='bibr' target='#b193'>(Schulz et al., 2019a)</ns0:ref>. Although there is some wildlife export to local market centres (e.g. Iquitos: <ns0:ref type='bibr' target='#b15'>Bodmer &amp; Lozano, 2001;</ns0:ref><ns0:ref type='bibr' target='#b141'>Mayor et al., 2019)</ns0:ref>, most is consumed at a household level and constitutes an important protein source <ns0:ref type='bibr' target='#b193'>(Schulz et al., 2019a)</ns0:ref>, providing higher potential to limit wildlife export from peatland areas than may be the case in South-east Asia and Africa.</ns0:p><ns0:p>High densities of domestic and semi-wild animals reared on peatlands could also serve as a direct or indirect zoonotic EID vector to humans. For instance, in Indonesia, over 1.8 million chickens are kept in the predominantly peatland municipality of Palangka Raya (BPS Palangka Raya, 2018), while large numbers of naturally cave-roosting edible-nest swiftlets (mostly PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Aerodramus spp.) are reared in special buildings in many peatland areas, with most nests exported to China <ns0:ref type='bibr' target='#b88'>(Husson et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b112'>Koon &amp; Cranbrook, 2014;</ns0:ref><ns0:ref type='bibr' target='#b211'>Thorburn, 2014)</ns0:ref>.</ns0:p><ns0:p>In summary, the combination of high native faunal diversity, habitat encroachment and fragmentation, plus trade in native and non-native fauna in many tropical peatland areas appears to represent a suitable set of conditions under which zoonotic EIDs could potentially (re-)emerge in future. Given that TPSFs represent the largest remaining blocks of lowland forest in some areas (e.g., lowland Borneo: <ns0:ref type='bibr' target='#b248'>Wijedasa et al., 2018;</ns0:ref><ns0:ref type='bibr'>Congo Basin: Dargie et al., 2017</ns0:ref>; parts of the Amazon: <ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017)</ns0:ref>, attention should be paid to conserving and sustainably managing these TPSFs and their wildlife to reduce the likelihood of potential future zoonotic EID pandemics arising.</ns0:p></ns0:div> <ns0:div><ns0:head>What are the Potential Immediate Impacts of the COVID-19 Pandemic in Tropical Peatland Areas?</ns0:head><ns0:p>The impacts of the COVID-19 pandemic will depend heavily on its length and severity, and evolving government and societal responses, which will vary between and potentially even within tropical peatland nations. This unpredictability notwithstanding, we nevertheless outline some areas of potential concern relating to tropical peatlands specifically, while referring readers to the more generic issues raised by <ns0:ref type='bibr' target='#b35'>Corlett et al. (2020)</ns0:ref>, <ns0:ref type='bibr' target='#b61'>Evans et al. (2020)</ns0:ref> and <ns0:ref type='bibr' target='#b85'>Hockings et al. (2020)</ns0:ref>, many of which will apply to tropical peatlands as much as to other habitats, and to any potential future pandemics causing similar levels of socio-economic disruptions to COVID-19. It is also pertinent to note that these issues occur on top of pre-existing challenges for tropical peatland conservation and sustainable management (see <ns0:ref type='bibr' target='#b40'>Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b81'>Harrison et al., 2020;</ns0:ref><ns0:ref type='bibr'>Roucoux et al., 2017 for reviews)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Public Health and Potential Combined Impacts from Haze Pollution</ns0:head><ns0:p>Disadvantaged populations are expected to be disproportionately affected by pandemics, further exacerbating existing social and economic inequalities <ns0:ref type='bibr' target='#b124'>(Lee, Rogers &amp; Braunack-Mayer, 2008;</ns0:ref><ns0:ref type='bibr'>WHO, 2009)</ns0:ref>. With the exception of Reunion, Brunei, Puerto Rico and Australia, which contain only small peatland areas, all tropical peatland nations listed by <ns0:ref type='bibr' target='#b166'>Page, Rieley &amp; Banks (2011)</ns0:ref> are classified as low or middle income (OECD, 2020), with many also being considered relatively under-prepared to cope with disease pandemics <ns0:ref type='bibr' target='#b164'>(Oppenheim et al., 2019)</ns0:ref>. Tropical peatland communities are often relatively remote and (agricultural) conditions marginal, with lower access to public health and other services, no or poor medical insurance, fewer formal employment opportunities and higher poverty rates than non-peatland areas (e.g., Kalimantan: <ns0:ref type='bibr' target='#b143'>Medrilzam et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b212'>Thornton, 2017;</ns0:ref><ns0:ref type='bibr'>van Beukering et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b250'>Wiseman et al., 2018;</ns0:ref><ns0:ref type='bibr'>DRC: C. Ewango &amp; G. Dargie, pers. obs.;</ns0:ref><ns0:ref type='bibr' /> Peru: E. Honorio, pers. obs.; Fig. <ns0:ref type='figure'>3</ns0:ref>). The rural populations of several tropical peatland nations have disproportionate numbers of people with underlying health conditions and/or malnutrition (e.g., <ns0:ref type='bibr' target='#b106'>Kandala et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b154'>Nair, Wares &amp; Sahu, 2010)</ns0:ref>, and many do not have access to formal healthcare, or the running water, good sanitation and hygiene systems required to implement the recommended WASH approach to COVID-19 (WHO &amp; PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed UNICEF, 2020). For example, in Central Kalimantan, less than 40% of people have access to improved sanitation (WHO, 2017) and the province tends to perform poorly in healthcare provision evaluations <ns0:ref type='bibr' target='#b206'>(Suparmi et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b250'>Wiseman et al., 2018)</ns0:ref>. Disseminating COVID-19 health guidance information will likely also be more difficult in rural tropical peatland areas with poor communications infrastructure, further reducing the probability that risk reduction behaviours will be followed. It is therefore possible that the health impacts of the COVID-19 pandemic may be relatively severe and/or prolonged in tropical peatland nations.</ns0:p></ns0:div> <ns0:div><ns0:head>INSERT FIG. 3 AROUND HERE</ns0:head><ns0:p>While their remoteness and low population densities may reduce the potential for COVID-19 to reach and spread between some rural tropical peatland communities, evidence to support this supposition is limited, since many of these communities-like many others on resource frontiers-are deeply embedded in market relations (e.g., <ns0:ref type='bibr' target='#b126'>Li, 2014;</ns0:ref><ns0:ref type='bibr' target='#b143'>Medrilzam et al., 2014)</ns0:ref> and local to international value chains bring them into regular contact with outsiders <ns0:ref type='bibr' target='#b52'>(Dove, 2011;</ns0:ref><ns0:ref type='bibr' target='#b192'>Schreer, 2016)</ns0:ref>. Even among those communities living relatively autonomously, there are still levels of contact with non-community actors, including NGOs, researchers, public agencies and commercial organisations. Indeed, media reports from the Americas and Africa indicate that COVID-19 has already spread into remote indigenous communities <ns0:ref type='bibr' target='#b18'>(Brito, 2020;</ns0:ref><ns0:ref type='bibr' target='#b230'>Wallace, 2020)</ns0:ref>. In DRC, this comes on top of existing Ebola and measles epidemics <ns0:ref type='bibr' target='#b11'>(Blomfield, 2020)</ns0:ref>, and in Iquitos, Peru, on top of dengue fever and leptospirosis outbreaks <ns0:ref type='bibr' target='#b32'>(Collyns, 2020b)</ns0:ref>. Once introduced, the possibility of COVID-19 cases going undetected and unreported, owing to a lack of testing, limited awareness plus an element of fatalism arising from the commonness of disease in such communities (e.g., Borneo: L. Chua, pers. obs.), combined with the important role that communal activities and high levels of social interaction play in many peatland communities, raises the risk of infection spreading. The more limited public health resources in rural tropical peatland areas means that the health impacts arising should an outbreak occur are potentially more serious than in less remote and more affluent communities. Furthermore, development and medical assistance in tropical peatland areas may be temporarily halted due to financial and organisational challenges (e.g., mobile medical teams in Peruvian Amazon: <ns0:ref type='bibr'>(Vine Trust, 2020)</ns0:ref>. These considerations place an onus on businesses, governments, NGOs and researchers working with such communities to take measures to reduce the chances of introducing and spreading the disease between communities.</ns0:p><ns0:p>Given that elderly people are more vulnerable to COVID-19 <ns0:ref type='bibr' target='#b255'>(Zhou et al., 2020)</ns0:ref>, one impact of outbreaks in tropical peatland areas may be a loss of traditional local knowledge regarding these ecosystems. Another important knock-on impact highlighted by the WHO is potential for disrupted responses to other major public health issues that may risk reversing gains made against these. This includes malaria and dengue fever, which are endemic in many tropical peatland areas and exhibit several symptoms similar to <ns0:ref type='bibr'>COVID-19 (PAHO, 2020;</ns0:ref><ns0:ref type='bibr'>WHO, 2020d)</ns0:ref>; and immunisations for diseases such as diphtheria, measles and polio (WHO, 2020a). In addition, co-infection of COVID-19 with other diseases and increased COVID-19 mortality rates in co-infected patients has been reported <ns0:ref type='bibr' target='#b119'>(Lansbury et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b255'>Zhou et al., 2020)</ns0:ref>, as has a case of co-infection of COVID-19 and Plasmodium vivax malaria <ns0:ref type='bibr' target='#b189'>(Sardar et al., 2020</ns0:ref>; following hospitalisation and discharge, the patient in this case tested negative for both diseases). Important questions therefore exist regarding the potential for (increased impacts from) COVID-19 co- <ns0:ref type='table'>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:ref> infection in tropical peatland areas, especially for the most vulnerable members of these communities. Measures to tackle these pre-existing diseases must therefore continue to remain a priority in tropical peatland areas.</ns0:p><ns0:p>Tropical peatland degradation and drainage increase fire risk. In Indonesia, peatland fires and their associated haze (smoke pollution) now occur almost annually <ns0:ref type='bibr' target='#b71'>(Gaveau et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b165'>Page &amp; Hooijer, 2016;</ns0:ref><ns0:ref type='bibr'>Fig. 4</ns0:ref>), leading to high carbon emissions, forest and biodiversity losses, and major public health impacts from inhalation of the toxic haze (see <ns0:ref type='bibr' target='#b81'>Harrison et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b165'>Page &amp; Hooijer, 2016;</ns0:ref><ns0:ref type='bibr' target='#b218'>Uda, Hein &amp; Atmoko, 2019</ns0:ref> for reviews). The haze contains high small particulate concentrations and several toxic compounds, including CO 2 , CO, CH 4 , NH 3 , HCN, NO X , OCS and HCl <ns0:ref type='bibr' target='#b203'>(Stockwell et al., 2016)</ns0:ref>. Haze exposure during the prenatal period has been linked to decreased adult height attainment (Tan-Soo &amp; Pattanayak, 2019), and short-term exposure during the severe 2015 fires is estimated to have caused 100,300 or more premature mortalities in Equatorial Asia <ns0:ref type='bibr' target='#b113'>(Koplitz et al., 2016</ns0:ref>; see also <ns0:ref type='bibr' target='#b36'>Crippa et al., 2016)</ns0:ref>. This is important in light of recent reports that increased air pollution may elevate COVID-19 case numbers, hospital admissions and mortality <ns0:ref type='bibr' target='#b30'>(Cole et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b33'>Conticini, Frediani &amp; Caro, 2020;</ns0:ref><ns0:ref type='bibr' target='#b162'>Ogen, 2020;</ns0:ref><ns0:ref type='bibr' target='#b253'>Wu et al., 2020)</ns0:ref>, and has led to concerns being raised by both regional think tanks <ns0:ref type='bibr' target='#b69'>(Gan et al., 2020)</ns0:ref> and relevant experts in media reports regarding peatland fires and COVID-19 <ns0:ref type='bibr' target='#b100'>(Jong, 2020b;</ns0:ref><ns0:ref type='bibr' target='#b132'>Listiyorini, 2020;</ns0:ref><ns0:ref type='bibr' target='#b226'>Varkkey, 2020)</ns0:ref>, though under COVID-19 lockdown conditions such impacts may be at least partially mitigated by general shutdowns of anthropogenic activities <ns0:ref type='bibr' target='#b107'>(Kanniah et al., 2020)</ns0:ref>. In particular, in a pre-print article, <ns0:ref type='bibr' target='#b30'>Cole et al. (2020)</ns0:ref> and <ns0:ref type='bibr' target='#b253'>Wu et al. (2020)</ns0:ref> report that an increase in PM2.5 (small particulate matter) of just 1 &#956;g/m 3 is associated with a 8-16.6% increase in COVID-19 death rate, whereas in Central Kalimantan, PM2.5 levels have been reported to exceed 1,500 &#956;g/m 3 during severe fire periods <ns0:ref type='bibr' target='#b6'>(Atwood et al., 2016)</ns0:ref> and average mean exposures between 2011-2015 have been estimated at 26 &#956;g/m 3 , over double the recommended WHO exposure limit <ns0:ref type='bibr' target='#b218'>(Uda, Hein &amp; Atmoko, 2019)</ns0:ref>. Similar findings were reported in relation to the impacts of air pollution on fatalities from the earlier SARS epidemic in China <ns0:ref type='bibr' target='#b38'>(Cui et al., 2003)</ns0:ref>, suggesting that high levels of air pollution may increase vulnerability of populations exposed to haze from peatland fires to future pandemics. Observations that SARS-CoV-2 RNA can be present on particulate matter <ns0:ref type='bibr' target='#b196'>(Setti et al., 2020)</ns0:ref>, and suggestions that the ability of fine particulates, such as PM2.5, to penetrate deep inside the lungs and remain in the air for long periods of time <ns0:ref type='bibr' target='#b68'>(Frontera et al., 2020)</ns0:ref> lead to further concern that haze pollution from peat fires may increase COVID-19 transmission. Finally, some symptoms of haze exposure are also similar to those of COVID-19 (e.g., dry cough, weakness), which may lead to complications with regards to COVID-19 testing and case identification. While extensive tropical peatland fires are currently mainly limited to Indonesia <ns0:ref type='bibr' target='#b40'>(Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b128'>Lilleskov et al., 2019)</ns0:ref>, increasing pressure to develop African and South American peatlands could elevate their fire risks if preventative measures are not implemented <ns0:ref type='bibr' target='#b186'>(Roucoux et al., 2017)</ns0:ref>, with consequent potential impacts on the susceptibility of their populations to future respiratory EID pandemics. Continuing and amplifying measures to avoid and control fires in tropical peatlands <ns0:ref type='bibr' target='#b165'>(Page &amp; Hooijer, 2016;</ns0:ref><ns0:ref type='bibr' target='#b48'>Dohong et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b248'>Wijedasa et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b81'>Harrison et al., 2020)</ns0:ref> is therefore of heightened importance.</ns0:p></ns0:div> <ns0:div><ns0:head>INSERT FIG. 4 AROUND HERE</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Economy and Livelihoods</ns0:head><ns0:p>The COVID-19 pandemic and associated response measures are likely to have deep and longlasting adverse global and local economic impacts, with the worst case scenario being that poverty alleviation achieved in recent decades might be reversed, earnings reduced and our ability to meet the UN Sustainable Development Goal of ending poverty by 2030 compromised <ns0:ref type='bibr'>(ASEAN, 2020;</ns0:ref><ns0:ref type='bibr' target='#b134'>Lucas, 2020;</ns0:ref><ns0:ref type='bibr' target='#b205'>Sumner, Hoy &amp; Ortiz-Juarez, 2020;</ns0:ref><ns0:ref type='bibr'>WB, 2020b)</ns0:ref>. Although tropical peatland communities are less likely to face the types of direct interruptions to livelihood resulting from strict lockdowns, adverse economic impacts linked to the pandemic may nevertheless have important consequences for them. Indeed, low incomes in many tropical peatland households, and weak social and food security safety nets in many tropical peatland countries <ns0:ref type='bibr' target='#b66'>(FAO et al., 2019)</ns0:ref>, may make them particularly vulnerable. Anticipating the nature and the severity of these effects is complicated by the diversity of peatland communities and their livelihood heterogeneity <ns0:ref type='bibr' target='#b96'>(Jelsma et al., 2017)</ns0:ref>. In Indonesia, for example, economic activities include self-employment in commodity tree crop, short-term vegetable crop and livestock production (especially chicken), swiftlet nest farming, hunting, fishing, gathering non-timber forest products, logging, artisanal mining, employment in construction and other industries, trading, remittances from family members employed elsewhere, and employment by government services and NGOs <ns0:ref type='bibr' target='#b192'>(Schreer, 2016;</ns0:ref><ns0:ref type='bibr' target='#b212'>Thornton, 2017)</ns0:ref>. The intensity of direct economic impacts arising from the COVID-19 pandemic are likely to be linked to the way communities are integrated into wider trade and resource allocation networks. Here we consider some of these potential impacts.</ns0:p><ns0:p>The anticipated global recession is expected to have a generally negative impact on agricultural prices, with prevalent low incomes in peatland areas likely to amplify the impact of price falls. Indeed, the UN and IUCN have flagged the expected negative impact of the COVID-19 pandemic on the livelihoods of remote indigenous groups and called for special measures to restore and support traditional indigenous economies (IUCN, 2020; UN/DESA, 2020; UN/EMRIP, 2020). Tropical peatland communities are commonly not food self-sufficient and tend to be close to poverty at the best of times <ns0:ref type='bibr' target='#b249'>(Wildayana &amp; Armanto, 2018)</ns0:ref>. Global food shortages due to COVID-19 may have implications for the price of imported food, increasing pressure on these communities and potentially forcing some into poverty (as feared more generally by key international institutions: (S&#225;nchez-P&#225;ramo, 2020; WFP, 2020). Small-scale settler communities cultivating short-lifecycle food crops and staples serving local markets are likely to be less adversely affected, because demand for their output is relatively inelastic and close market proximity means they are better placed to respond to changes in local demand by rapidly switching production. To date, prices of such crops are indeed remaining relatively stable <ns0:ref type='bibr' target='#b134'>(Lucas, 2020)</ns0:ref>, although some increases are anticipated (e.g., <ns0:ref type='bibr' target='#b2'>Amanta &amp; Aprilianti, 2020)</ns0:ref>. For such producers, the greater problem may be changes in the price of agricultural inputs (e.g., fertilisers) and other imported goods.</ns0:p><ns0:p>Producers of commodities such as palm oil, rubber and beef on tropical peatlands are likely to experience more severe adverse impacts due to falling prices resulting from changes in international demand. For example, the value of palm oil fell by over 20% from December 2019 to April 2020 (Index Mundi, 2020), and declining palm oil export volumes and domestic consumption have already been noted in Indonesia <ns0:ref type='bibr' target='#b190'>(Sarkar, 2020)</ns0:ref>, leading to reported concerns regarding lower demand and falling prices <ns0:ref type='bibr' target='#b69'>(Gan et al., 2020)</ns0:ref>. Any sustained fall in palm oil price is likely to more severely impact peatland-based production, which is less productive than cultivation on mineral soil, and small-scale producers, for whom yields are typically lower and who will have less resources to sustain them through difficult periods <ns0:ref type='bibr' target='#b60'>(Euler et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b204'>Sumarga et al., 2016)</ns0:ref>. These problems are exacerbated by the long-term commitment and substantial investments that these crops require, which restrict the capacity of farmers to rapidly change to other crops, thus limiting their resilience. Prolonged low prices, combined with longstanding issues such as land titling, are thus likely to lead to increased hardship among these small-scale commodity producers, with previous commodity price falls being linked to increased poverty, mental health issues and suicide among small scale producers <ns0:ref type='bibr' target='#b217'>(Tyson, Varkkey &amp; Choiruzzad, 2018)</ns0:ref>. Conversely, decisions of better resourced, larger plantation companies regarding expansion or contraction of operations will be more heavily influenced by predictions of the longer-term impact of the COVID-19 pandemic, rather than immediate price changes. Current thinking in the palm oil industry appears to be that, while the short-term impact may be significant, longer-term effects are uncertain, given the nature of the product and its market <ns0:ref type='bibr' target='#b190'>(Sarkar, 2020)</ns0:ref>, reducing the likelihood of substantial longer-term reductions in oil palm expansion by large operators. These price changes may nevertheless mean that alternative economic and livelihood options may compete more favourably economically with palm oil in tropical peatland areas (cf. <ns0:ref type='bibr' target='#b241'>Wich et al., 2011)</ns0:ref>, possibly encouraging their uptake. However, evidence of land holding in tropical peatlands strongly suggests that financial problems linked to poor revenues and limited capital may contribute to plot abandonment <ns0:ref type='bibr' target='#b254'>(Yusoff, Muharam &amp; Khairunniza-Bejo, 2017)</ns0:ref>.</ns0:p><ns0:p>It is likely that many employed and casual workers in tropical peatland cities will lose their immediate income sources owing to lockdowns and other disruptions, leading to migration of people from cities to home communities, with media reports suggesting this may already be happening in Indonesia <ns0:ref type='bibr' target='#b133'>(Listori, 2020)</ns0:ref>, Malaysia <ns0:ref type='bibr' target='#b177'>(Radhi, 2020)</ns0:ref> and Peru <ns0:ref type='bibr' target='#b31'>(Collyns, 2020a</ns0:ref>; E. N. Honorio Coronado, pers. obs.). There is concern that this 'returning home' may bring sources of infection to vulnerable communities, leading to media reports of the strict closure of some indigenous territories <ns0:ref type='bibr' target='#b198'>(Sierra Praeli, 2020)</ns0:ref>. This situation may impose a triple strain on tropical peatland communities by increasing the risk of infection via returning community members, imposing additional burdens on household resources and depriving households of external income sources. Furthermore, during previous periods of economic instability in Indonesia, such as following the 1998 financial crash, illegal land uses increased, including illegal logging (EIA-TELAPAK, 1999), as people sought alternative economic means. Economic disruptions are also expected in communities dependent upon eco-tourism <ns0:ref type='bibr' target='#b61'>(Evans et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b85'>Hockings et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b130'>Lindsey et al., 2020)</ns0:ref>. There is thus a risk that the financial burden caused by the COVID-19 pandemic could increase unlawful exploitation of natural resources (timber, wild animals), as well as increasing other marginal livelihood practices (e.g. artisanal mining).</ns0:p></ns0:div> <ns0:div><ns0:head>Food Security</ns0:head><ns0:p>The UN World Food Programme has predicted that the COVID-19 pandemic will lead to over a quarter of a billion people suffering acute hunger by the end of 2020, owing to increased conflict, reduced aid and trade, price fluctuations and lost incomes <ns0:ref type='bibr' target='#b3'>(Anthem, 2020)</ns0:ref>. Concerns have been raised that the pandemic is affecting all four pillars of food security (food availability, utilisation/nutrient intake, stability and particularly food access) and is leading to forced cutbacks in nutrient-rich non-staple foods towards starchy staples, with consequent long-term adverse health impacts <ns0:ref type='bibr' target='#b118'>(Laborde et al., 2020)</ns0:ref>. We have already suggested that, given existing relatively high poverty levels and frequent lack of food self sufficiency, plus positions at the end of long supply chains for many external commodities, tropical peatland communities may be particularly vulnerable to food supply problems and price rises. Furthermore, while statutory and NGO food assistance programmes are operating and good transport links exist in some tropical peatland areas, these are designed to deal with relatively small numbers of clients and could be quickly overwhelmed if demand increases dramatically, whilst other areas have poor assistance infrastructure and poor, or often no, road access (e.g., Fig. <ns0:ref type='figure'>3</ns0:ref>). For indigenous communities, such issues may be exacerbated by a common lack of visibility in public policy <ns0:ref type='bibr' target='#b219'>(UN, 2009)</ns0:ref>. Ensuring food security and the provision of food assistance to tropical peatland areas thus presents logistical, organisational and political challenges, and development of more localised food supply chains may therefore be favourable in post-pandemic responses (as recommended in a general context by <ns0:ref type='bibr' target='#b170'>Pearson et al., 2020)</ns0:ref>.</ns0:p><ns0:p>As previously mentioned, tropical peatland community livelihood strategies are often based on a range of economic activities, whose relative importance varies temporally. For example, rural communities in Borneo have for centuries been fairly entrepreneurial and responsive to changing global/regional trade patterns <ns0:ref type='bibr' target='#b4'>(Arenz et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b52'>Dove, 2011)</ns0:ref>. From their perspective, the COVID-19 pandemic may therefore not be viewed exceptionally, but rather as yet another development to which they must respond. Consequently, any potential loss of economic opportunities and reduced external demand for, or price of, locally produced commodities in such communities may lead to individuals reconsidering their livelihood strategies, resulting in changes in natural resource exploitation. For example, any price declines in swiftlet nests produced in Indonesian peatland areas following potential economic slowdowns in China (the main market), may lead swiftlet keepers to search for alternative incomes from harvesting forest resources. While such shifts may not necessarily result in increased pressure on natural resources, concern nevertheless appears warranted, as recent reviews suggest that increased overall local environmental resource pressures may be expected under conditions of rural economic hardship <ns0:ref type='bibr' target='#b142'>(Robinson, 2016)</ns0:ref> and that, where poverty exists, households are more likely to pursue economic activities that improve their family's short-term livelihoods, regardless of environmental impacts (South-east Asia: <ns0:ref type='bibr' target='#b51'>Douglas, 2006;</ns0:ref><ns0:ref type='bibr'>Peru: Schulz et al., 2019b)</ns0:ref>. From a positive perspective, any such changes may present an opportunity to work productively with affected communities to innovate and find more sustainable, locally rooted ways of responding to their needs.</ns0:p><ns0:p>Nevertheless, research suggests that experience of adverse shocks leads to increased risk aversion <ns0:ref type='bibr' target='#b73'>(Gloede, Menkhoff &amp; Waibel, 2015)</ns0:ref> and that communities may be less willing to engage in alternative 'sustainable' livelihood activities if these are perceived to carry increased risks <ns0:ref type='bibr' target='#b183'>(Rodriguez et al., 2009)</ns0:ref>. Perceived threats to health, income and food security related to COVID-19 may therefore reduce uptake or outcomes of revitalisation and community development activities in tropical peatland areas. Indeed, work in Indonesia has demonstrated that ecosystem restoration activities are unlikely to be adopted unless accompanied by assurances concerning food and income security <ns0:ref type='bibr' target='#b21'>(Carmenta &amp; Vira, 2018)</ns0:ref>. The post-(peak) pandemic pressures governments face to 'rescue the economy', may lead governments and companies in tropical peatland areas to reduce (or at least withhold attempting to improve) social and environmental standards (e.g., media reports: <ns0:ref type='bibr' target='#b22'>Carrington, 2020;</ns0:ref><ns0:ref type='bibr' target='#b87'>Hurowitz, 2020)</ns0:ref>, in an effort to reduce commodity prices and thus increase sales to rescue or expand industries such as palm oil. While evidence is currently limited, some media reports indicate that the pandemic may be employed to chip away at existing environmental measures, such as the EU's palm oil ban <ns0:ref type='bibr' target='#b110'>(Kobo, 2020)</ns0:ref>. Such trends could be exacerbated if economic restrictions and priority shifts lead to reduced enforcement of environmental and social regulations concerning extractive industries, or reduced emphasis on conservation and restoration, in tropical peatland areas.</ns0:p></ns0:div> <ns0:div><ns0:head>Land Conflicts</ns0:head><ns0:p>Land conflicts are one of the most important and complicated problems in land-use management in many countries with large peatland areas <ns0:ref type='bibr' target='#b27'>(Colchester, 2010;</ns0:ref><ns0:ref type='bibr' target='#b29'>Colchester et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b155'>Nesadurai, 2013;</ns0:ref><ns0:ref type='bibr' target='#b195'>Scullion et al., 2014)</ns0:ref>. In Peru, 24% of the estimated peatlands of the Pastaza-Mara&#241;on overlap with oil-extraction or exploration concessions, national reserves and territories of indigenous communities <ns0:ref type='bibr' target='#b186'>(Roucoux et al., 2017)</ns0:ref>. In Indonesia, 807,178 ha are in land tenure conflict, with 73% of this contested land in the plantation sector (KPA, 2019). Increasing economic pressures and food insecurity associated with the COVID-19 pandemic may potentially aggravate this situation, with any increases in land tenure conflict likely to lead to increased peat fire incidence, given previously reported links <ns0:ref type='bibr' target='#b143'>(Medrilzam et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b207'>Suyanto, 2007)</ns0:ref>, and potentially therefore to further disenfranchisement and displacement of the poorest households and communities. This, in turn, may trigger further habitat encroachment and increased human-wildlife contact, whilst both food insecurity and haze pollution in degraded tropical peatland areas may increase COVID-19 impacts and susceptibility, as we have previously outlined. A glimpse of the complex and uneven effects that can result from such large-scale disruptions is found in Schreer's (2016) ethnography of a Central Kalimantan village's experiences and aspirations in the early-to mid-2010s. Here, the end of the logging boom-precipitated in by rapid governmental decentralisation-in the mid-2000s had highly variable effects in her fieldsite, with migrants and non-migrants reacting differently, and other industries and initiatives bringing in new opportunities, risks and land conflicts, as well as widespread disillusionment and resentment at various external parties <ns0:ref type='bibr' target='#b192'>(Schreer, 2016)</ns0:ref>.</ns0:p><ns0:p>The return migration of casual and temporary workers from cities in to rural tropical peatland communities and the continued presence of migrant labour (working in industries disrupted by COVID-19), as noted above, constitutes a further factor that may exacerbate pressure on land resources and thus increase conflicts, as well as acting as a direct vector for disease transmission. Supporting local communities in and around tropical peatland areas to resolve land tenure conflicts and promote inclusive, sustainable community-led management will likely prove important in mitigating these risks surrounding land conflict issues.</ns0:p></ns0:div> <ns0:div><ns0:head>Unequal Community and Gender Impacts</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020) Importantly, any negative economic, social, and health impacts for communities will not occur in a vacuum, and the social and ecological impacts of COVID-19 will affect different communities and community members unequally. Regional inequalities, and differences in access to health facilities between rural and urban areas, are likely to impact (efforts to curb) COVID-19 impacts. For example, in Indonesia, COVID-19 testing rates in Java have only surpassed WHO minimum recommendations in the capital, Jakarta (WHO, 2020d), and other provinces with large peatland areas generally lag behind in terms of health facility availability and access (MoH RoI, 2019). Furthermore, there is evidence that legal and policy frameworks support development patterns that marginalise peatland communities, resulting in the neglect of their interests by policy makers <ns0:ref type='bibr' target='#b142'>(McCarthy &amp; Robinson, 2016)</ns0:ref>. The COVID-19 crisis may exacerbate these trends, contributing to increased levels of economic <ns0:ref type='bibr' target='#b169'>(Payne &amp; Bradley, 2020)</ns0:ref> and intersectional inequality <ns0:ref type='bibr' target='#b16'>(Bowleg, 2020)</ns0:ref>. It is being increasingly recognised that tropical peatland communities are not homogenous in relation to economic status, ethnicity, access to facilities, etc., (e.g., <ns0:ref type='bibr' target='#b212'>Thornton, 2017)</ns0:ref> and the interaction of these factors means that the impact of COVID-19 is therefore highly unlikely to affect all members of tropical peatland communities equally, with impacts likely varying between households.</ns0:p><ns0:p>In addition, studies have highlighted how economic, food and health crises often put women in a more vulnerable position compared to men, despite shrinking options available for both genders <ns0:ref type='bibr' target='#b172'>(Pitkin &amp; Bedoya, 1997;</ns0:ref><ns0:ref type='bibr' target='#b173'>Pollock &amp; Lin Aung, 2010)</ns0:ref>. For example, previous studies on other diseases such as tuberculosis have demonstrated context-specific gender-related differences in the barriers to access diagnostic and treatment services, especially in developing countries <ns0:ref type='bibr' target='#b117'>(Krishnan et al., 2014)</ns0:ref>, while the greater role of women as primary care givers may increase their probability of infection <ns0:ref type='bibr' target='#b42'>(Davies &amp; Bennet, 2016;</ns0:ref><ns0:ref type='bibr' target='#b232'>Wenham et al., 2020)</ns0:ref>. In line with this,international organisations have already highlighted the key role that women play in food provision in indigenous rural communities and the likely heightening of this and other burdens on women, including the loss of childcare and other supports services, plus surges in domestic violence, during the COVID-19 pandemic (FAO, 2020; UN, 2020; UN/DESA, 2020). For example, the closure of schools is likely to disproportionately increase the childcare burden on mothers and older daughters within families <ns0:ref type='bibr' target='#b43'>(de Paz et al., 2020)</ns0:ref>. Recognising and accounting for such gender differences will be important for mitigating the impact of COVID-19 in tropical peatland areas. Related to this, care should also be taken in pursuing novel economic developments, such as the expansion of independent oil palm smallholdings, which may risk exacerbating or even producing new gender inequalities <ns0:ref type='bibr' target='#b59'>(Elmhirst et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b105'>Julia &amp; White 2012)</ns0:ref> that may in turn compound gender inequalities relating to COVID-19.</ns0:p></ns0:div> <ns0:div><ns0:head>Research, Training and Education</ns0:head><ns0:p>The COVID-19 pandemic is significantly impacting field research globally, with travel restrictions and social distancing, plus (potential) reductions in research funding, leading to the adjustment, postponement or cancellation of many ongoing and planned field activities <ns0:ref type='bibr' target='#b35'>(Corlett et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b61'>Evans et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b130'>Lindsey et al., 2020)</ns0:ref>. Among other impacts, research disruptions could potentially delay or prevent key output production (with possible knock-on effects on policy development), cause data gaps (particularly for long-term ecological data collection) and equipment supply issues. Social research, especially that involving bringing people together PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020) physically, will be particularly impacted <ns0:ref type='bibr' target='#b35'>(Corlett et al., 2020)</ns0:ref>. For example, at the time of writing, the Indonesian Ministry of Research and Technology has prohibited new foreign researchers from entering and conducting research activities until the pandemic has ended (RISTEK-BRIN, 2020). In DRC, the UKRI-funded CongoPeat project is continuing to operate virtually but field activities and face-to-face initiatives and meetings have been paused, with field researchers required to depart suddenly and leave collected samples behind, thus delaying or potentially even risking completion of their analysis (C. E. N. Ewango, G. <ns0:ref type='bibr'>Dargie &amp; D. Kopansky, pers. obs.)</ns0:ref>. Media reports indicate that some indigenous communities have effectively closed off to visitors to protect themselves from the potentially devastating impacts of COVID-19 on their communities (Peruvian Amazon, including the peatland region of Loreto: Sierra <ns0:ref type='bibr' target='#b198'>Praeli, 2020;</ns0:ref><ns0:ref type='bibr'>Papua: Milko, 2020)</ns0:ref>. Similar observations have also been made for remote villages in Central and West Kalimantan, Sabah and Sarawak (L. Chua &amp; M. A. Imron, pers. obs.), with such measures representing a traditional mechanism among rural Borneo communities for protecting themselves from the spiritual and other effects of dangerous events like deaths or disease outbreaks <ns0:ref type='bibr' target='#b24'>(Chua, 2012)</ns0:ref>. Even after the COVID-19 pandemic (peak), such communities may retain their wariness, remaining less welcoming or even hostile to outside researchers, who have an ethical obligation to ensure that they do not inadvertently transmit the disease to these communities. Research on great apes and other primates is also at high risk of disruption, as measures are put in place to prevent reverse zoonotic transmission of SARS-CoV-2, given the perceived vulnerability of these species <ns0:ref type='bibr' target='#b39'>(Damas et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b72'>Gillespie et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b94'>IUCN SSC WHSG &amp; PSG SGA, 2020;</ns0:ref><ns0:ref type='bibr' target='#b144'>Melin et al., 2020)</ns0:ref>. This has led some to question whether all primate field research should be cancelled for 2020 <ns0:ref type='bibr' target='#b181'>(Reid, 2020)</ns0:ref>, with others highlighting the potential negative impacts of such a move and the potential primate conservation and research opportunities that may arise from the pandemic <ns0:ref type='bibr' target='#b121'>(Lappan et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b214'>Trivedy, 2020)</ns0:ref>. Ultimately, the risks of any research in relation to the COVID-19 threat will need to be carefully balanced against the counter-risks of not conducting or restricting research in terms of addressing other important conservation and community wellbeing issues. From a positive perspective, some of these developments may facilitate the empowerment of local scientists and thus help strengthen in-country scientific research in tropical peatland areas.</ns0:p><ns0:p>Research and training meetings are being cancelled or postponed around the globe, including rescheduling of the International Peatland Congress from June 2020 to May 2021, which will negatively impact both networking opportunities for peat scientists and short-term income for the International Peatland Society. Disruptions to teaching, research and networking activities, and related publications, will potentially have reverberating impacts on the careers and (short-term) incomes of students and junior researchers, with possible overall negative impacts on future scientific expertise <ns0:ref type='bibr' target='#b35'>(Corlett et al., 2020)</ns0:ref>. For example, Borneo Nature Foundation (BNF) and Universitas Gadjah Mada (UGM) had been developing plans for BNF to support selected UGM students to conduct research dissertations on tropical peatlands in 2020. These have now been converted to desk-based projects, which cannot provide students with vital first-hand perspectives of the tropical peatland environment. More widely, impacts are likely to be especially severe and potentially irreversible for students and researchers already facing (economic) disadvantages, acting to further entrench existing biases and low representation of black, Asian and minority ethnic (BAME) groups in environmental organisations <ns0:ref type='bibr' target='#b209'>(Taylor, 2015)</ns0:ref>, and women in tropical peatland research <ns0:ref type='bibr' target='#b213'>(Thornton et al., 2019)</ns0:ref>. Parents, and particularly mothers, are also likely to be disproportionately affected <ns0:ref type='bibr' target='#b200'>(Staniscuaski et al., 2020)</ns0:ref>. Some positive responses are, however, already evident in tropical peatland nations, with for example UGM instituting a series of online expert talks, BNF conducting live children's education sessions through Facebook Live (BNF, 2020a) and some schools in remote areas of Kalimantan without internet using radio broadcasts to deliver teaching during the pandemic <ns0:ref type='bibr' target='#b95'>(Jakarta Post, 2020)</ns0:ref>.</ns0:p><ns0:p>There is also the risk that research and development funds will be redirected towards COVID-19related projects, increasing the difficulty of accessing funding for other research and conservation priorities <ns0:ref type='bibr' target='#b61'>(Evans et al., 2020)</ns0:ref>. Funding shortfalls for tropical peatland research may lead to local research and other staff temporarily losing income tied to specific projects, or even redundancies of permanent staff. While some international NGOs may be able to furlough some staff to receive government funding support, this is unlikely to be an option in most tropical peatland countries. In addition to negatively impacting local economies (tropical peatland research is often conducted in remote rural areas with limited economic opportunities), redundancies of permanent research staff would also represent a consequent loss in local research project capacity if staff move to work in other sectors, which may be difficult and require a long time to replace. On the plus side, prolonged international travel restrictions may help promote the role of local researchers in multi-national research projects, while also reducing the carbon footprint of research involving international flights. Whilst it may be possible to maintain communications and coordination of many ongoing projects using online communication tools, this may be limited by lack of or slow internet connections in more remote tropical peatland areas. Equipping remote communities with such communication infrastructure would help minimise this and other wider impacts resulting from reduced potential for outsider travel to these areas. While researchers may be able to help contribute some individual-based assistance (e.g., laptop provision, booster aerials), provided sufficient flexibility exists in research grant provisions, larger-scale upgrades in communications infrastructure would require in-country government support. Potential may also exist for international travel funds to be reallocated to support online learning or other training for local students and researchers, helping to build local capacity in the long term, although in the shorter-term international travel to undertake studentships and receive training may be curtailed.</ns0:p></ns0:div> <ns0:div><ns0:head>Conservation and Restoration</ns0:head><ns0:p>In addition to potential funding impacts discussed in the previous section, field activity disruptions, delays and cancelations will likely detrimentally impact many tropical peatland conservation, restoration, community development and outreach projects (e.g., Indonesia: <ns0:ref type='bibr' target='#b69'>Gan et al., 2020)</ns0:ref>. This is particularly so when activities are time sensitive (e.g., planting seedlings during periods of optimal peat water level), require continued maintenance (e.g., tending seedlings in nurseries that may otherwise die, requiring much time and funds to replace), or typically involve large teams or in-person gatherings. For example, BNF temporarily halted all its in-person childrens' education and other activities involving groups of people, including community development, sustainable livelihood and fire-fighting training (BNF, 2020b). Some organisations have initiated self-isolation protocols before and after field visits as a COVID-19 precaution (e.g., Frankfurt Zoological Society in Jambi, Sumatra: Lestari, 2020, pers. comm.), PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed which although potentially effective in relation to reducing transmission risk, may place additional demands and stresses upon staff.</ns0:p><ns0:p>Halting dialogue with tropical peatland communities, who are frequently remote and have limited or no internet access, may lead to reduced uptake or failure of alternative livelihood development activities, if critical implementation periods are missed (e.g. a planting season), enthusiasm wanes, community members seek alternative opportunities and become tied to these, or if initiatives are incorrectly implemented and so fail because of reduced training provision, thus potentially decreasing local enthusiasm for such initiatives in future. Even activities that can be postponed may still suffer from negative impacts, if this means that calendar-based targets are not met, community engagement drops, or consequent additional damage occurs (e.g., dams to restore peatland hydrology are damaged owing to lack of project presence in the area). As noted in the previous section, equipping rural communities with improved communications infrastructure would help minimise such impacts.</ns0:p><ns0:p>As is the case more generally <ns0:ref type='bibr' target='#b35'>(Corlett et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b61'>Evans et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b85'>Hockings et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b130'>Lindsey et al., 2020)</ns0:ref>, TPSF encroachment, timber and wildlife harvesting is likely to increase if conservation agencies are less active or visible, or enforcement reduced, particularly if rural communities do increasingly fall back on exploiting natural resources owing to economic shock or food insecurity (see above). For example, media reports quoting conservation organisations have already indicated increased poaching in the Leuser Ecosystem, Sumatra, which includes TPSF areas <ns0:ref type='bibr' target='#b79'>(Hanafiah, 2020)</ns0:ref>. Importantly, if work restrictions or funding shortfalls lead to reduced fire-fighting capabilities in degraded tropical peatland areas, this may result in increased fire incidence and severity, as highlighted by reputed researchers and institutions in media reports <ns0:ref type='bibr' target='#b20'>(Cannon, 2020;</ns0:ref><ns0:ref type='bibr' target='#b210'>Taylor, 2020)</ns0:ref>. Fire-fighting in South-east Asian peatlands is often undertaken by local community members or, in plantations, by company teams. In these situations, <ns0:ref type='bibr' target='#b151'>Moore et al. (2020)</ns0:ref> identify COVID-19 infection risks either from a home setting to fire teams and, given the difficulties in social distancing during fire-fighting, from the fire team back to the home and community. This could impact on the effectiveness of fire-fighting, as well as on individuals and their families. As noted above, any resulting increases in TPSF degradation, fragmentation, wildlife harvesting, or haze pollution will be expected to increase the immediate impacts of the COVID-19 pandemic and/or future potential for zoonotic EID emergence in tropical peatland areas.</ns0:p><ns0:p>As a result of the above, plus the potential for SARS-CoV-2 transmission to susceptible wildlife species, there is a risk of significant negative impacts on tropical peatland biodiversity. This is likely to be particularly important for peatland specialists and already threatened species. This includes, for example, the orangutan (Pongo spp.): the two species occurring on tropical peatland are Critically Endangered, with low and rapidly declining populations, have limited distribution on only one island each (Sumatra: P. abelii, Lesson; Borneo: P. pygmaeus, Linn.), have significant proportions of their populations in tropical peatland, are unable to persist in completely deforested landscapes and are at high risk from hunting owing to their slow reproductive rates <ns0:ref type='bibr' target='#b224'>(Utami-Atmoko et al., 2017)</ns0:ref>. Orangutans and other non-human primates <ns0:ref type='bibr' target='#b39'>(Damas et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b144'>Melin et al., 2020)</ns0:ref> are also potentially at risk of spill-over of the virus, including from asymptomatic human carriers, leading to recommendations for ape-based ecotourism, field research and non-essential habitat conservation activities to be reduced, and impacting the activities of rescue, rehabilitation and release organisations (BOSF, 2020; <ns0:ref type='bibr' target='#b72'>Gillespie et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b94'>IUCN SSC PSG SGA, 2020;</ns0:ref><ns0:ref type='bibr' target='#b94'>IUCN SSC WHSG &amp; PSG SGA, 2020;</ns0:ref><ns0:ref type='bibr' target='#b181'>Reid, 2020)</ns0:ref>. The crisis thus increases the urgency to fully implement the IUCN best practice guidelines on ape tourism and health monitoring/disease control <ns0:ref type='bibr' target='#b72'>(Gillespie et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b94'>IUCN SSC WHSG &amp; PSG SGA, 2020)</ns0:ref>. While eco-tourism is not currently commonplace in tropical peatlands, disruptions to ape-based eco-tourism may potentially threaten the viability of ape/habitat conservation and livelihood initiatives in those tropical peatland areas relying heavily on this source of income, and result in shifts back towards more destructive economic activities (cf. <ns0:ref type='bibr' target='#b61'>Evans et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b85'>Hockings et al., 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>How Might COVID-19 Impact Future Tropical Peatland Conservation, and What Repercussions Might This Have in Relation to Disease Pandemics?</ns0:head><ns0:p>Recent comprehensive reviews have highlighted the multiple, inter-linked threats and challenges already facing tropical peatland conservation and restoration, and provided suggestions to help tackle these <ns0:ref type='bibr' target='#b40'>(Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b81'>Harrison et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017)</ns0:ref>. The additional short-to medium-term considerations arising in the context of the COVID-19 pandemic that we outline above may not all come to light, may create only short-term issues if the pandemic proves to be relatively short lived and recovery is relatively rapid, or may create longer-term changes if the pandemic is long-lasting or creates irreversible shifts. The severity and duration of impacts experienced will inevitably also vary between and within tropical peatland nations. We cannot therefore provide firm predictions regarding longer-term impacts of the pandemic on tropical peatlands and their communities, and consequently on changes in the potential of tropical peatlands to act as a source of future zoonotic EID pandemics. This caveat notwithstanding, we highlight some potential longer-term changes that may be important for policy makers, scientists and other tropical peatland stakeholders to consider going forward. Some tropical peatland nations have already indicated that development projects will be paused or cancelled, and that economic growth may be prioritised above sustainability concerns in postpandemic recovery phases, which could have long-term repercussions. For example, media reports indicate that Indonesia's planned capital city relocation may be postponed owing to the pandemic <ns0:ref type='bibr' target='#b45'>(Demetriadi, 2020;</ns0:ref><ns0:ref type='bibr' target='#b74'>Gokkon, 2020)</ns0:ref>, whereas Padat Karya (Labour Intensive) government projects involving large numbers of local community members are still proceeding and being prioritised as an attempt to mitigate the economic impacts of the crisis <ns0:ref type='bibr' target='#b91'>(Iskandar, 2020;</ns0:ref><ns0:ref type='bibr' target='#b243'>Widodo, 2020)</ns0:ref>, as are 89 major infrastructure development projects, including potential development of large peatland areas for food production in Central Kalimantan with the goal to improve food security <ns0:ref type='bibr' target='#b99'>(Jong, 2020a)</ns0:ref>. A history of prioritising economic growth over public health in the context of haze from peatland fires has been suggested in Malaysia <ns0:ref type='bibr' target='#b228'>(Varkkey &amp; Copeland, 2020)</ns0:ref>, while some recent media reports indicate that Indonesia will prioritise economic growth following the pandemic, with the consequence that emission reduction targets will not be increased <ns0:ref type='bibr' target='#b101'>(Jong, 2020c)</ns0:ref> and environmental protections potentially rolled back <ns0:ref type='bibr' target='#b104'>(Jong, 2020d)</ns0:ref>. In Peru, the largest threat is the (re-)activation of regional economies to promote Manuscript to be reviewed agricultural development <ns0:ref type='bibr' target='#b55'>(El Peruano, 2020)</ns0:ref>, which with the current lack of any documents to guide land use planning, may increase deforestation and access to remote forests (e.g. potential road construction that will affect peatlands in Loreto: <ns0:ref type='bibr' target='#b8'>Baker et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The race to return to economic growth and stability could also drive increased fossil fuel extraction. This is a risk in the Congo Basin <ns0:ref type='bibr' target='#b40'>(Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b146'>Miles et al., 2017)</ns0:ref> and Peruvian Amazon <ns0:ref type='bibr' target='#b128'>(Lilleskov et al., 2019)</ns0:ref>, where many oil concessions coincide with peatland areas. In Peru, oil extraction and transportation has led to many spills from poorly maintained pipes that have often been slow to be remediated (e.g., <ns0:ref type='bibr' target='#b184'>Rodr&#237;guez Mega, 2016)</ns0:ref>. Expanding oil exploration and bringing new concessions into production involves infrastructure creation, clearing seismic survey lines, and building access roads and pipelines <ns0:ref type='bibr' target='#b186'>(Roucoux et al., 2017)</ns0:ref>, typically causing habitat loss and degradation <ns0:ref type='bibr' target='#b122'>(Laurance, Goosem &amp; Laurance, 2009;</ns0:ref><ns0:ref type='bibr' target='#b140'>M&#228;ki, Kalliola &amp; Vuorinen, 2001)</ns0:ref>. We note, however, that conversely and at least in the short-term, an alternative outcome of the pandemic could be lower global oil demand and production through at least 2020 (OGJ Editors, 2020), which may result in oil exploration becoming a lower immediate priority. Indeed, it has been indicated by the Ministry of Environment and Tourism that oil exploration in RoC peatland areas will not proceed in accordance with the country's international climate commitments (Arlette Soudan-Nonault, 2020, pers. comm.; see also <ns0:ref type='bibr' target='#b233'>Weston, 2020)</ns0:ref>.</ns0:p><ns0:p>Concerns have been raised regarding the intersection of the COVID-19 outbreak and public health responses with the ongoing increase in climate hazards associated with climate change <ns0:ref type='bibr' target='#b171'>(Phillips et al., 2020)</ns0:ref>. Tropical peatlands are vulnerable to climate change, which will likely result in increased drought frequency, duration and/or severity, with consequent peat water table lowering, and increased peat oxidation and fire incidence, as already seen extensively in Southeast Asia <ns0:ref type='bibr' target='#b215'>(Turetsky et al., 2015)</ns0:ref>. There thus exists a risk of a positive-feedback loop development (cf. <ns0:ref type='bibr' target='#b81'>Harrison et al., 2020)</ns0:ref>, whereby any decisions or events related to the pandemic that lead to increased peatland degradation and fire also then lead to further carbon emissions and peatland degradation, thus increasing future fire prevalence and the various negative impacts associated with this, including in relation to COVID-19 and potential future disease pandemics.</ns0:p><ns0:p>While peatland rewetting and revegetation may help mitigate some impacts of potential increases in the areas of degraded and burned peatland, these processes are unavoidably slow in nature, with peat burned in a single fire event taking potentially hundreds of years to re-form, generating concerns regarding the political tractability of some tropical peatland restoration goals <ns0:ref type='bibr' target='#b81'>(Harrison et al., 2020)</ns0:ref>. Coupled with the existing large areas of degraded peatland already requiring restoration, particularly in South-east Asia, any potential increase in the extent of degraded and burned peatland as an indirect consequence of the COVID-19 pandemic would be expected to have long-term negative repercussions for tropical peatland management, community livelihoods, public health, carbon emissions and biodiversity. Reversing such changes, and enhancing the sustainability of tropical peatland management in general, may also become more challenging in the foreseeable future if, as discussed above, the current pandemic leads to loss of skilled, in-country peatland scientists and conservationists who would take time to replace. Any future pandemics may obviously exacerbate this.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions and Recommendations</ns0:head><ns0:p>Our review indicates that sustainable management and conservation of tropical peatlands is important for mitigating the impacts of the current COVID-19 pandemic, while avoiding further encroachment into these ecosystems and intensive harvesting of their wildlife could reduce the potential for future zoonotic EID emergence and severity (summarised in Fig 1 <ns0:ref type='figure'>.</ns0:ref>). Importantly, because many of the potential impacts identified in relation to COVID-19 arise through socioeconomic disruptions relating to our response to the pandemic, rather than unique characteristics of the SARS-CoV-2 virus itself, this conclusion should be applicable more widely to pandemic situations and similar sudden socio-economic shocks. To our knowledge, this link between tropical peatland ecosystems and zoonotic EIDs has not been specifically noted previously and this therefore represents an important finding in terms of our understanding of the benefits that these ecosystems provide, thus strengthening the argument for their conservation and sustainable use. In addition, and recognising that our assessment is both preliminary and likely incomplete owing to the recent emergence and rapid evolution of the COVID-19 pandemic, we highlight the potential additional threats and challenges towards sustainable management of tropical peatlands and their wildlife posed by this pandemic, and the potential knock-on effects of these for tropical peatland conservation and local communities going forward.</ns0:p><ns0:p>These conclusions support and strengthen previous recommendations regarding the sustainable management of tropical peatlands, including regarding hydrological functions, fire prevention, avoiding encroachment and habitat fragmentation, plus maintaining and where necessary restoring healthy peatland ecosystems that support diverse biological communities, sequester carbon and provide socio-economic benefits to local communities (see, e.g., <ns0:ref type='bibr' target='#b40'>Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b48'>Dohong, Abdul Aziz &amp; Dargusch, 2018;</ns0:ref><ns0:ref type='bibr' target='#b76'>Graham, Giesen &amp; Page, 2017;</ns0:ref><ns0:ref type='bibr' target='#b81'>Harrison et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b149'>Mizuno, Fujita &amp; Kawai, 2016;</ns0:ref><ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b193'>Schulz et al., 2019a;</ns0:ref><ns0:ref type='bibr' target='#b194'>Schulz et al., 2019b;</ns0:ref><ns0:ref type='bibr' target='#b248'>Wijedasa &amp; al., 2018</ns0:ref> and references therein for detailed recommendations). They also support previous recommendations to carefully manage terrestrial wildlife harvesting in and trade from tropical peatland areas <ns0:ref type='bibr' target='#b15'>(Bodmer &amp; Lozano, 2001;</ns0:ref><ns0:ref type='bibr' target='#b80'>Harrison et al., 2011)</ns0:ref> and more generally (e.g., <ns0:ref type='bibr' target='#b34'>Corlett, 2007;</ns0:ref><ns0:ref type='bibr' target='#b82'>Harrison et al., 2016)</ns0:ref>. Further to these generic recommendations, we offer the following specific suggestions for researchers and practitioners relating to tropical peatland sustainable management, COVID-19 and potential future zoonotic EIDs:</ns0:p><ns0:p>1. Field projects should support frontline team members to reduce their COVID-19 exposure risk and, thus, the risk of them infecting other community members and of work being impacted through worker infection. In addition to ensuring and facilitating adherence to current government and WHO guidelines, specific recommendations have been proposed for fire-fighters <ns0:ref type='bibr' target='#b151'>(Moore et al., 2020)</ns0:ref>, and work with great apes (IUCN SSC WHSG &amp; PSG SGA, 2020) and in their habitats (IUCN PSG SGA, 2020). Development of similar specific guidelines for other sectors relating to tropical peatland management could be beneficial. 2. All actors should support local communities in and around tropical peatland areas to resolve land tenure conflicts and promote sustainable community-led management <ns0:ref type='bibr' target='#b10'>(Blackman et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b188'>Santika et al., 2017)</ns0:ref>, facilitate locally-derived sustainable economy development <ns0:ref type='bibr' target='#b170'>(Pearson et al., 2020)</ns0:ref>, and prevent external users from unsustainably exploiting these ecosystems. This includes encouraging, and where necessary opening up new, discussions regarding the relationships between use of and encroachment into tropical peatlands, Manuscript to be reviewed commercial vs. subsistence wildlife harvesting, and zoonotic EID risk, and considering these holistically in the context of community wellbeing and aspirations (cf. <ns0:ref type='bibr' target='#b108'>Kavousi et al., 2020)</ns0:ref>.</ns0:p><ns0:p>It should also include re-orientating discussions, from conservation and development organisations approaching with a perspective of 'how can we help them' to an approach of 'we're all in this together, so what can we each bring to the table to help?'. Such reorientations should ideally be informed by fresh social research on how communities' lives, relations and experiences have been impacted by COVID-19, as these are unlikely to simply revert to a pre-pandemic 'normal'. Incorporating and centering local knowledge and priorities from the outset has the additional benefit of increasing the success of conservation and development interventions <ns0:ref type='bibr' target='#b148'>(Mistry &amp; Berardi, 2016)</ns0:ref>. COVID-19 has forced increased use of online communications tools and we hope this can help foster new patterns of international collaboration that further empower local researchers. See <ns0:ref type='bibr' target='#b136'>Lupton (2020)</ns0:ref> for an extensive list of potential approaches to reducing COVID-19 risk during social science fieldwork. 4. Wherever possible and working within all relevant government and sector COVID-19 guidelines, research, conservation and community projects should continue collection of long-term monitoring data and initiate repeat data collection to provide before-and-after comparisons, in order to help assess the impacts of the COVID-19 pandemic in relation to the issues discussed in this paper. This may include, for example, patrol team or satellite data monitoring to assess if forest incursions or fire incidence have increased, or community livelihood data to assess if livelihood choices have changed and if/how this has impacted local wellbeing. It should also include collecting baseline human and wildlife community health information to better understand, monitor and thus mitigate zoonotic EID risk in tropical peatland areas. 5. Policy makers in tropical peatland areas should ensure that their policy decisions do not amplify the risks of the COVID-19 or potential future pandemics. In particular, any weakening of environmental and social standards and regulations relating to tropical peatlands should be avoided (see also <ns0:ref type='bibr' target='#b63'>Evers et al., 2017;</ns0:ref><ns0:ref type='bibr'>Wijedasa et al., 2017)</ns0:ref>. Instead, we recommend development and implementation of a 'One Health / Ecohealth' approach in tropical peatland nations, that recognises and supports the complex, mutually-beneficial interconnections among the health of people, animals, plants and our shared environment <ns0:ref type='bibr' target='#b57'>(El Zowalaty &amp; J&#228;rhult, 2020;</ns0:ref><ns0:ref type='bibr' target='#b85'>Hockings et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b185'>Roger et al., 2016;</ns0:ref><ns0:ref type='bibr'>UNEP &amp; ILRI, 2020)</ns0:ref>.</ns0:p><ns0:p>We further recommend strengthening communications infrastructure in (remote) tropical peatland areas, to help minimise any impacts of reduced travel or on-the-ground activities in these areas. 6. Funders should exercise maximum possible flexibility to tropical peatland research, conservation and community projects impacted by the COVID-19 pandemic, including regarding achieving deliverable targets and deadlines, where possible offering additional support to maintain local staff employment during periods of enforced down time instead of offering zero-cost funding extensions, and not penalising unavoidable 'failures' by current projects in future funding application rounds. This will be important in minimising potential PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020) project closures, plus staff redundancies or lost income, and associated risk of loss of expertise. Opportunities to diversify revenue sources to increase the resilience of conservation efforts should also be explored <ns0:ref type='bibr' target='#b130'>(Lindsey et al., 2020)</ns0:ref>. 7. All actors should recognise the role that tropical peatlands, their conservation, sustainable management and restoration will play in both the current COVID-19 pandemic and in the potential for future zoonotic EID emergence; and actively promote this message in fundraising, education, outreach, community and government engagement, while taking care to tailor messages and associated recommendations appropriately to local audiences <ns0:ref type='bibr' target='#b25'>(Chua, 2020;</ns0:ref><ns0:ref type='bibr' target='#b108'>Kavousi et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b137'>MacFarlane &amp; Rocha, 2020</ns0:ref>; see also <ns0:ref type='bibr' target='#b23'>Charania &amp; Tsuji, 2012;</ns0:ref><ns0:ref type='bibr' target='#b26'>Chua et al., 2020)</ns0:ref>.</ns0:p><ns0:p>As is the case more generally (e.g., <ns0:ref type='bibr' target='#b35'>Corlett et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b61'>Evans et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b85'>Hockings et al., 2020</ns0:ref>; UN/DESA, 2020; UN/EMRIP, 2020; UNEP &amp; ILRI, 2020), our review indicates that the relationships and reciprocal impacts between tropical peatland ecosystems and communities, COVID-19 and disease pandemics are inter-linked, multi-faceted, and likely to vary over both space and time. This adds an extra pandemic-related dimension to the increasingly complex picture that is emerging regarding the challenges and opportunities for conservation and sustainable management of tropical peatlands <ns0:ref type='bibr' target='#b40'>(Dargie et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b81'>Harrison et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b186'>Roucoux et al., 2017)</ns0:ref>. At the same time, this also suggests that potential for win-win solutions exists to simultaneously address the challenges identified herein relating to COVID-19 and disease pandemics in tropical peatland areas, alongside the previously identified challenges facing these ecosystems and their resident human communities. This is particularly pertinent in light of recent analyses estimating that global pandemic prevention costs associated with reducing deforestation, wildlife trade and farmed animal spill-over, and early zoonotic disease detection and control over the course of ten years is equivalent to only about 2% of the costs of the COVID-19 pandemic <ns0:ref type='bibr' target='#b47'>(Dobson et al., 2020)</ns0:ref>. Moving beyond these conclusions linked to our review, and while we hope that the COVID-19 pandemic is rapidly mitigated and thus many of the potential issues discussed in this paper fail to (fully) materialise, we nevertheless trust that our consideration of these issues and recommendations provided helps improve our ability to anticipate and prevent potential negative impacts that may arise from the pandemic. In addition, and regardless of the pandemic, we contend that it helps to foster the inter-connected thinking that will be required to ensure the future health and wellbeing of tropical peatlands and their human communities alike. In (b) and (c) the satellite image is a false colour composite of Landsat 7 and 8 imagery from 2016 to 2017 created in GoogleEarth Engine. Human population density per square kilometre <ns0:ref type='bibr' target='#b70'>(Gaughan et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b129'>Linard et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b251'>WorldPop, 2020)</ns0:ref>, is overlaid with colours ranging from pink with a density of one person per square kilometre to greater than 50 in red. Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 4</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>3. The peatland research community should rethink approaches to research in at-risk communities, to potentially incorporate more remote and online working, plus enhanced roles for local research teams. This may require re-appraising how likely increasingly limited research resources are distributed, and particularly supporting local research capacity development and local researcher empowerment within international collaborations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>Peatland boundaries are indicated by yellow lines and purple indicates treeless areas. Most of the Congo Basin peatlands have low population density, but there is Mbandaka city in the centre (red), plus a growing population in the periphery on the northeast and east. In Riau, there are large population centres, plus growing populations and plantations directly adjacent to the peatlands. Map data: Google, USGS;Gaughanet al., (2013); Linardet al. (2012); Leifeld &amp; Menicetti (2018); WorldPop (2020).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 3 Figure 3 .</ns0:head><ns0:label>33</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Peatland fire encroaching into forest (a) and local fishers working under thick haze conditions from peatland fires (b) in Central Kalimantan, Indonesia.</ns0:figDesc></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50176:1:1:NEW 10 Sep 2020)</ns0:note> </ns0:body> "

" The Editor PeerJ Centre for Ecology and Conservation College of Life and Environmental Sciences University of Exeter Penryn Campus, TR10 9FE Cornwall, UK t +44 (0) 1326 255 974 e [email protected] w www.exeter.ac.uk 19 January 2023 Dear Dr Roberts, Re: Manuscript resubmission Many thanks for forwarding your and the two reviewers’ constructive feedback regarding our manuscript “Tropical peatland conservation is important in the context of COVID-19 and potential future (zoonotic) disease pandemics”, submitted to PeerJ. We appreciate the comments provided and have revised our manuscript in response to these, as detailed in our point-by-point reply below (our responses in italics) and the resubmitted manuscript files, which include track change and compare documents, in addition to the clean version. We feel that taking this feedback on board has helped improve our manuscript and so hope that you now consider it to be acceptable for publication in PeerJ. Yours sincerely, Mark E. Harrison BSc PhD CBiol MRSB (on behalf of all authors) Postdoctoral Research Fellow [email protected] Editor comments (David Roberts) Thank you for what is an interesting and timely paper that places the importance of tropical peatlands in the context of zoonotic disease. The reviewers have provided what I believe is useful feedback on your manuscript and I look forward to reading the corrections and seeing your rebuttal. One of the reviewers noted that the last paragraph is an opinion and not a summary as it is a literature review not an opinion piece. I agree with this, however I believe these opinions are important and would suggest providing a strong summary before moving to what is clearly an opinion e.g. saying 'Moving beyond the findings of this literature review...' (or some similar wording) so that it is distinct and therefore disassociates what is the fact based literature review from the author's opinions. Thank you for this positive assessment and advice regarding our final paragraph, which we have edited as suggested. Reviewer 1 (Anonymous) Basic reporting - The topics covered in the article are of great interest in the current and future scenario; No changes required. - Abstract, line 63: I suggest: “many potential vertebrate and invertebrate vectors”; Changed as suggested. - Introduction: The introduction is well structured and the objectives of the review are well defined; No changes required. - Figure 1: Please, remove the Earth illustration from the center of the virus. This is unnecessary. I also suggest connecting the green box “increased biodiversity loss” (cause) with the box “increased human-wildlife contact” (consequence); While we did initially consider an alternative Earth-free illustration inset in the EID circle for this figure, we discarded for submission in favour of the current “Earth illustration”. We believe that this Earth symbol serves an important purpose, as it illustrates that EIDs emerging anywhere in the world can potentially impact tropical peatlands (as we demonstrate is the case with COVID-19) and that EIDs that may arise in tropical peatlands in future should likewise be of global concern. We have therefore not made any changes to this part of the figure, but will be happy to show the previous version of this figure to the Editor and discuss this further if required. Connection added between the increased human-wildlife contact and biodiversity loss boxes, as suggested. - This is a cross-disciplinary review within the scope of the journal. No changes required. - The article is written in good English; No changes required. - The revised literature is updated. We have updated the manuscript with the most highly relevant citations that have emerged since our initial submission, including discussion of debate within the primatological community regarding cancelling fieldwork and opportunities arising (Lappan et al., 2020; Reid, 2020; Trivedy, 2020), plus citation of new IUCN recommendations regarding activities in ape habitats, new findings on vulnerability of different animal species (Damas et al., 2020), fire risk and other conservation issues in Indonesia (Gan et al., 2020), the intersection of climate hazards and COVID-19 (Phillips et al, 2020), COVID-19 and conservation in Africa (Lindsey et al., 2020), food security (Laborde et al., 2020), indigenous communities (IUCN, 2020), air pollution (Cole et al., 2020; Setti et al, 2020; Frontera et al., 2020), compound risks associated with climate change (Phillips et al. 2020), and the economics of pandemic prevention vs. response (Dobson et al., 2020). The Akebe (2000) citation for Africa case incidence in the Introduction has also been replaced with a newer, peer-reviewed citation (Nordling, 2020). Experimental design - The methodology is interesting (it looks “sincere” and consistent) and adequately described. In brief, the methodology is well structured for a review article; No changes required. - Table S1: The authors state “non-zoonotic diseases (e.g. flu)”. Flu is not a good example since some authors consider flu a zoonotic disease (usually human influenza strains is derived from non-human animals). Please, change to another example to avoid confusion; Example changed to “diseases caused by malnutrition”. - Table S2: Okay; No changes required. - Sources are adequately cited and the text is organized logically. No changes required. Validity of the findings - Although the article often has very opinionated parts, references were used appropriately; No changes required. - The text is drafted in an appropriate manner and respecting a logical structure; No changes required. - Figures 2-6: Okay; No changes required. - Lines 864-869 (conclusion section): I suggest removing or modifying it for a more 'pragmatic' paragraph. This final paragraph is full of expressions like 'we hope'. The authors must not forget that this is not an opinion article, but a review article, and therefore the final paragraph should represent the main conclusions about the works that have been reviewed throughout the text. Please see response above to the Editor regarding this comment. Comments for the Author - Regarding the topic “Are Tropical Peatlands a Potential Source Habitat for Disease Pandemics?”: I suggest to include a paragraph (maybe second paragraph, after line 200) discussing the risk of spillover events associated with human contact with bushmeat and wildlife in forest environments; We agree with the reviewer regarding the high importance of this topic. We note, however, that we already designate two paragraphs in this section to discussing the role of human-wildlife contact and bushmeat (which we now refer to as “wild meat”, because many dictionaries refer to “bushmeat” as being just from Africa) hunting in this context (L242-269). In the absence of any identified omissions from these existing paragraphs, we would therefore consider these to be sufficient and have not therefore added any additional paragraphs in relation to this. We concur that mention of the term “spill-over event” would be beneficial in this context, however, and have therefore added the following sentence to the start of the two paragraphs mentioned above: “Patterns of human-wildlife contact and wild meat hunting in tropical peatlands provide further indication of the potential for disease spill-over events from wildlife to humans to occur.” - Lines 285-286: Please, remove “It is therefore possible – and indeed we hope – that many of the concerns raised here will never be realised in the context of this pandemic.” This sentence is speculative; Sentence deleted, as suggested. We feel that some kind of qualifier/disclaimer is still needed, however, to make clear that we do not expect our predictions to be perfect, and have therefore added the following underlined text to the start of the next sentence: “This unpredictability notwithstanding, we nevertheless outline some areas of potential concern…”. - Line 843: Please, correct “Wijedasa & al., 2017”. Thanks for the spot! Corrected. Reviewer 2 (Valéria Kaminski) Basic reporting This review presents broad and cross-disciplinary interests that fall within the scope of the journal. The authors present recent and very important issues that are indeed accessible to researchers of different areas of knowledge. Besides, the authors raised important and current concerns. The Introduction section is clear. However, I strongly suggest some recommendations to better align the Introduction with the text body with the proposal, such as outlined as follows: # The present paper raised very important questions regarding the impacts of the current global scenario caused by Covid-19 pandemic. The authors indeed show the importance to keep the existing conservation strategies in peatland areas at this moment and elegantly expose the likely consequences of a halt in investments in training scientists, people education, and conservatory demands along with the need of economic improvements to avoid community crisis. However, the paper is not really focused on direct links between Covid-19 and tropical peatlands. Considering this, the conclusion and introduction should be constructed into the environmental health of peatland areas (and its communities' livelihoods) and their connected concerns in times of pandemics instead of a direct consequence of the pandemics or risk for EID. Of note, these last issues should appear as they are in the text, but the reader should be aware of the actual discussions raised through the text. We completely agree with the reviewer regarding this assertion. Some previous hints towards this were included in our original submission, and we have now strengthened these, and added additional text in the Abstract, Introduction and Conclusion to similar effect. # Regarding the title, some adjustments could be make it more suitable to the text body. Recalling the comments above and the elegant issues raised mainly in the sections 'Economy and Livelihoods', 'Food Security and Land Conflicts', and 'Unequal Community Impacts', I encourage the authors to reformulate the review title, raising a different call similar as 'COVID-19 AND TROPICAL PEATLAND AREAS: LESSONS ABOUT VULNERABLE COMMUNITIES, CONSERVATION STRATEGIES, AND RESEARCH THREATS IN TIMES OF PANDEMIC'. Because the COVID-19 pandemic is so new and our current assessment of its impacts is thus unavoidably based on reasoned predictions and short-term observations only, rather than the detailed research quantifying the longer-term impacts of the pandemic that will doubtlessly emerge over the coming years, we consider that it is premature to refer to “lessons learned” in the title. Nevertheless, we recognise that our original title was very narrowly focussed and have thus broadened it through addition of the underlined to: “Tropical peatlands and their conservation are important in the context of COVID-19 and potential future (zoonotic) disease pandemics”. We consider that this simple change fits better with the two-way nature of the relationship (i.e. that COVID-19 is having impacts on tropical peatlands, in addition to tropical peatland conservation impacting COVID-19 and potential for future pandemics) while retaining the paper’s main take-home message in the title, which we feel will attract readers’ attention more readily. Experimental design The review content is within the Aims and Scope of the Journal and the authors did present adequate investigation with high technical and ethical standards. The methods used by the authors are descriptive and did present sufficient and detailed information to be confirmed or replicated (obviously considering the possible changes in the scenario outlined in the time of writing). The information provided did present a comprehensive and unbiased coverage along with an important and neglected issue in the context of the current global scenario. The references were adequately chosen and cited. Nevertheless, I have some recommendations regarding the paragraphs and subsections as follows: # Within the topic 'Public Health', from line 344 to line 350, there is a brief discussion regarding different pathogens and infections. This is an important and likely neglected issue, especially considering peatland communities and Sars-CoV-2. Considering this, it would be interesting to introduce a paragraph raising the possible co-infections and their complications to such communities in vulnerable areas. The paragraph should answer questions such as: Is there such a relevant co-infection risk in peatlands communities? How severe this risk would be? What are the most common infections that could represent complications for people affected by Covid-19? What are the strategies to manage such risks? The reviewer poses some highly relevant questions here that are important to attempt to answer, both within and beyond the context of our paper. To our knowledge definitive answers to these questions for the diseases common in tropical peatland areas do not (yet) exist, but we have nevertheless highlighted these questions in two new sentences added at the end of this paragraph, which includes reference to studies that have shown increased mortality from COVID-19 in co-infected patients, plus an observation of co-infection with malaria. # In the 'Public Health' section, from line 352, there is a discussion more related to environmental aspects than public health per se. I suggest including a topic named 'Environmental Aspects' in which the text from lines 352 to 381 could be replaced. Alternatively, the authors could rename the topic to 'Public Health and Environmental Aspects'. Also, despite the great information provided in this section, it would be adequate to provide or briefly cite the alternatives to deal with the problems stated (similar to the strategies suggested for food security in lines 482-485 in the topic 'Food Security and Land Conflicts'). We have renamed this section “Public Health and Potential Combined Impacts from Haze Pollution”, noting that the text referred to discusses public health impacts in relation to this particular environmental aspect, rather than a more general/isolated discussion of environmental aspects (cf. in the Conservation and Restoration section). Potential solutions to the problems described were already included at the end of the second paragraph in this section, and have now been added to the end of the third and fourth paragraphs, in response to this comment. # Considering that 'land conflicts are one of the most important and complicated problems in land-use management in many communities with large peatland areas, affecting a large number of people', this issue deserves a separate topic with a more profound discussion and evaluation. Besides, considering the specific interest of this review, the Covid-19 pandemic should be addressed in the context of land conflicts in a perspective of how it would make communities more vulnerable to the disease and, again, authors may present strategies to avoid or minimize such effects. We agree and have now expanded this text along the lines suggested, more than doubling the previous length of this text and positioning it inside a new “Land Conflicts” sub-heading. # Authors should be congratulated for including the topic 'Unequal Community Impacts'. Moreover, this is a very important and still neglected aspect of the impacts of Covid-19 pandemic in overall aspects. Raising this issue greatly increases the coverage and insight of the review work carried out, adding unprecedented value to the article and to the Journal. Of note, a more profound discussion about how women on peatland communities are differently affected by the pandemic should be included. Finally, I suggest change the section title to 'Impacts of Unequal Communities and Gender Bias'. We agree with the reviewer regarding the importance of the gender bias issue and have therefore substantially expanded this discussion, as requested. Section heading changed to “Unequal Community and Gender Impacts”. # The topic 'Research, Training, and Education' is pretty welcome. Over-funding focused only on Covid-19 research is quite dangerous indeed and the authors elegantly provide examples of the consequences of short and long-term replacement of resources directed only to Sars-CoV-2 research in detriment of other scientific areas. To better finish this, it would be wonderful if authors suggest strategies to better assist the remote peatland areas with adequate technology such as internet support, remote censoring, and all the needs for the viability of future (without physical contact) researches and collaborations worldwide. Such strategies, once previously presented, could be recalled in line 646 of the topic 'Conservation and Restoration'. As suggested, sentence has been added to this effect in the final paragraph of this section, and to the second paragraph of the Conservation and Restoration section. # Line 695: 'How Might COVID-19 Impact Future Tropical Peatland Conservation and Sustainable Management, and What Knock-on Effects Might This Have on the COVID-19 and Possible Future Pandemics?' is quite a long title. Please consider separating it in two 'title questions' and adequate the text to respond to each one of them, to better highlight each of the questions addressed.  We agree with the reviewer that our current title for this section is rather long, but consider that splitting the section would create disjoint between these two inter-related questions. We have therefore kept this as one section, but have shortened the section heading to: “How Might COVID-19 Impact Future Tropical Peatland Conservation, and What Repercussions Might This Have in Relation to Disease Pandemics?”. # Recommendations and Conclusion could be separated sections. Also, even presenting recommendations at the end of the article, I reinforce to include or cite the possible solutions just after a problem is presented in the text body. As it is only the first paragraph of this section that is purely conclusions, which then flows into generic recommendations in the second paragraph, before presenting our specific recommendations list, we feel that the article flows better with this as one combined section and so have not separated. As noted in responses to some of this reviewer’s other comments above, we have, however, added more possible solutions to issues raised within the body text. # Line 826: online communications/ technological resources for distance communication are recalled. Please make it an additional item in the recommendations section. A sentence to this effect has been added to Recommendation #5. As the numbered recommendations are each directed at different actors (with #5 directed at policy makers), we consider that this is more appropriate than adding an additional recommendation just on this point (which would then give two numbered recommendations to policy makers and one to all other actors, though in reality many of these recommendations incorporate more than one suggestion). Validity of the findings This review presents an important perspective, which considers various logistical aspects of both research and access to vulnerable populations in peatland areas so that they have adequate assistance. Still, the economic and environmental aspects revisited, in parallel with social issues of paramount importance, raise issues that, although urgent, are still neglected. However, the issue of the risk of emergence of infectious diseases and a direct link with Covid-19 are aspects presented in a secondary way throughout the text, thus some changes must be made in the Introduction and Abstract of the article, as outlined previously. I reiterate that such changes do not make the article less relevant. It is a question of adequacy with the ideas outlined in the text that, indeed, in some points are affected by the pandemic or that raise the discussion about EID. Finally, the conclusion is adequate and quite complete and presents strategies for mitigating the problems presented. Thank you for this positive appraisal. Please see responses to the point regarding direct links with COVID-19 and changes to the Introduction and Abstract in our response to this reviewer’s and the Editor’s earlier comment on this above. Comments for the Author Dear Author(s), I must congratulate you on the initiative in writing a paper addressing the conservation of peatland areas and the impacts that these strategies may suffer in the current pandemic scenario. This is a very important perspective, which considers various logistical aspects of both research and access to vulnerable populations so that they have adequate assistance. Still, the economic and environmental aspects revisited, in parallel with social issues of paramount importance, raise issues that, although urgent, are still neglected. This is the overview I get from reading your manuscript. The risk of emergence of infectious diseases and a direct link with Covid-19 are aspects presented in a secondary way throughout the text, thus some changes must be made in the Introduction and Abstract of the article. I reiterate that such changes do not make the article less relevant. It is a question of adequacy with the ideas outlined in the text that, indeed, in some points are affected by the pandemic or that raise the discussion about EID. Please see response to previous and earlier comments from this reviewer and the Editor. # It sounds redundant when the authors use the term 'emergence' and the abbreviation EID together. I suggest revising this throughout the text, with abstract as a starting point. Good point. Edits made as suggested in the Abstract and other relevant places throughout the text. # Line 682: once transmission of Sars-CoV-2 from humans to other primates is mentioned, please use the term 'spillover'. Once the authors decide to not explain this term previously, it would be interesting to emphasize its meaning at this moment. This term has been added to the quoted sentence, as suggested, and in the “Are Tropical Peatlands a Potential Source Habitat for Disease Pandemics?” section, to which a definition of “spill-over” event has been added in parentheses. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. It is well-documented that the (bio)chemical reaction capacity of raw potato starch depends on the crystallinity, morphology and other chemical and physical properties of starch granules, and the properties of these granules are closely related to gene functions. Preparative yield, amylose/amylopectin content, and phosphorylation of potato tuber starch are starch-related traits studied at the genetic level. In this paper, we perform a genome-wide association study using a 22K SNP potato array to identify, for the first time, genomic regions associated with starch granule morphology and to increase the number of known genome loci associated with potato starch yield.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods.</ns0:head><ns0:p>A set of 90 potato (Solanum tuberosum L.) varieties from the ICG 'GenAgro' collection (Novosibirsk, Russia) were harvested, 90 samples of raw tuber starch were obtained, and DNA samples were isolated from the skins of the tubers. The morphology of potato tuber starch granules was evaluated by optical microscopy and subsequent computer image analysis. A set of 15,214 scorable SNPs was used for the genome-wide analysis. In total, 53 SNPs were found to be significantly associated with potato starch morphology traits (aspect ratio, roundness, circularity, and the 1 st bicomponent) and yieldrelated traits.</ns0:p><ns0:p>Results. A total of 53 novel SNPs were identified on potato chromosomes 1, 2, 4, 5, 6, 7, 9, 11, and 12; these SNPs are associated with tuber starch preparative yield and granule morphology. Eight SNPs are situated close to each other on the chromosome 1 and nineteen SNPs -on the chromosome 2, forming two DNA regions -potential QTLs, regulating Aspect ratio and Roundness of the starch granules. Thirtyseven of 53 SNPs are located in protein-coding regions. There are indications that granule shape may depend on starch phosphorylation processes. The GWD gene, which is known to regulate starch phosphorylation -dephosphorylation, participates in the regulation of a number of morphological traits, rather than one specific trait. Some significant SNPs are associated with membrane and plastid proteins, as well as DNA transcription and binding regulators. Other SNPs are related to low-molecular-weight metabolite synthesis, may be associated with flavonoid biosynthesis and circadian rhythm-related metabolic processes. The preparative yield of tuber starch is a polygenic trait that is associated with a number of SNPs from various regions and chromosomes in the potato genome.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Starch is one of the most important renewable and economically notable organic resources of humankind <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>. Simplicity of potato growing and ease of industrial production of pure potato starch makes it a necessary component and starting compound in food and chemical industries and determines valuability of an exhaustive study of all aspects of manufacturing and commercial application of potato starch is necessary. There are many publications related to chemical and biochemical transformations of starch of diverse botanical origin and differences properties of various starches. Surprisingly, little attention is paid to difference in starch properties at the level of different cultivars, such as potato cultivars. Commercial starch is still subdivided as 'potato starch', 'rice starch', 'corn starch', and so on, despite it is well-known that there are significant differences in the properties of starch of the same botanical origin. Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref> demonstrates handdrawn in 1935-1939 scientific table, which is located in Vavilov memorial room of the N.I. Vavilov All-Russian Institute of Plant Genetic Resources (St. Petersburg, Russia). The table relates to the expeditions to Southern America in 1932-1933 and illustrates intervarietal diversity of potato starch granule morphology. From general considerations, it is evident that chemical and biochemical reaction ability of the raw potato starch must depend on its granules crystallinity and morphology. If so, starches from different potato varieties with different granules' properties should fit different applications better. (Bio)chemical reaction capacity of raw potato starch depends on the crystallinity, morphology and other chemical and physical properties of starch granules. For example, after chemical acylation of the raw potato starch and fractionation of resulting acetylated starch according to granule size, amylose and amylopectin may be isolated and characterized by degree of substitution (DS) and degradability with &#945;-amylase, &#946;-amylase and amyloglucosidase. In contrast to amylose, the DS of the amylopectin from the differently sized granules increased with decreasing granule size. The acetyl groups of the amylose molecules from small granules are more heterogeneously distributed and located more closely to the non-reducing ends compared to amylose from larger granules.</ns0:p><ns0:p>The amylose populations from small granules of the acetylated starches were less susceptible to all the enzyme degradation reactions than the amylose from the large granules, even though the DS was similar. Additionally, the acetyl group distributions were different for amylopectin from different granule size fractions. <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b3'>3]</ns0:ref>. Some benefits of small potato starch granules are summarized in European patent <ns0:ref type='bibr' target='#b4'>[4]</ns0:ref>. The large size of the potato granules often hampers the utilization of potato starches, for example, in printing ink formulations, wherein potato starch granules may obstruct the printing apertures. Smaller granules facilitate chemical modification of the starch, since the surface-volume ratio is important for the accessibility of the modifying compound to the starch chains. In textile printing, for instance, carboxymethylated and roll dried potato starch of small granule size enables the production of finer prints. Furthermore, in adhesives, particularly in highly concentrated bag adhesives, crosslinked fine potato starches may be used as fillers and in drilling fluids, crosslinked and modified fine starches are expected to reduce fluid loss. In addition, smaller granule potato starches may be used as filler/structurant in soap, since this starch provides a pleasant sensation to the skin. Small granular potato starches are particularly suitable for applications in the food industry. Small granular potato starch will provide such advantages as lower starch dosage level, better taste profile, and smooth but not excessively swollen granules <ns0:ref type='bibr' target='#b5'>[5]</ns0:ref>. Other practical and important properties that are sensitive to starch granule size are liquid composite viscosity <ns0:ref type='bibr' target='#b6'>[6]</ns0:ref>, chemical modification and flowability <ns0:ref type='bibr' target='#b7'>[7]</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>It is evident that the shape of potato starch granules is closely related to genes function. Amylopectin polysaccharide is predominantly responsible for the smooth granule morphology.</ns0:p><ns0:p>Even low-amylose starch granules still maintain a smooth shape, whereas starch with reduced amylopectin content (by antisense knockout of SBE and GWD genes) is associated with more oddly fissured granules <ns0:ref type='bibr' target='#b8'>[8]</ns0:ref>. The role of starch synthases and related genes (GBSS, SSI -SSIII) in potato starch granule morphology has been discussed elsewhere <ns0:ref type='bibr' target='#b9'>[9]</ns0:ref>. In general, the role of certain genes in starch biosynthesis and starch granule morphology in particular is summarized in the review <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref>.</ns0:p><ns0:p>Traditionally, the preparative yield of potato tuber starch is the only starch-related trait generally accepted as important for potato breeding. In the past decade, starch phosphorylation and amylose/amylopectin content were also added to the short list of the traits studied at the genetic level by modern molecular biology methods. Thus, forward genetic tools, such as QTL analysis <ns0:ref type='bibr' target='#b11'>[11,</ns0:ref><ns0:ref type='bibr' target='#b12'>12]</ns0:ref> and association mapping <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>, were used to reveal loci associated with potato traits. The tuber starch characteristics, such as starch granular size and shape, and chemical-thermal properties of 21 potato varieties were determined and associated with genetic diversity through SSR markers. SSR-based cluster analysis revealed that varieties with interesting quality attributes were distributed among all clusters and subclusters, suggesting that the genetic basis of analyzed traits might differ among the varieties <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>.</ns0:p><ns0:p>Recently developed 22K SNP potato array is characterized by a high average density of markers, one locus per 40 kb (in the abovementioned studies <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref><ns0:ref type='bibr' target='#b11'>[11]</ns0:ref><ns0:ref type='bibr' target='#b12'>[12]</ns0:ref><ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>, this value does not exceed one marker per 4 Mbp). We used the chip to identify eight novel genomic regions on the chromosomes 1, 4, 5, 7, 8, 10, and 11 associated with starch phosphorylation. Some of the identified SNPs were located in noncoding genomic regions <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref>.</ns0:p><ns0:p>In this paper, for the first time we perform a genome-wide association study using a 22K SNP potato array to locate genomic regions associated with starch granule morphology and to increase the number of known genomic loci associated with potato starch yield. Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Plant material</ns0:head><ns0:p>The same set of 90 potato (Solanum tuberosum L.) varieties from the ICG 'GenAgro' collection (Novosibirsk, Russia) described in <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> was used in this paper as well. Complete list of the accessions used in the study and their origin is presented in Supplemental data S1 (Excel file).</ns0:p></ns0:div> <ns0:div><ns0:head>Starch isolation</ns0:head><ns0:p>Potato starch was isolated from tubers according to the typical procedure described elsewhere (for example, see <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>DNA isolation and genotyping</ns0:head><ns0:p>DNA was isolated from tuber skin using DNeasy Plant Mini Kit (Qiagen) according to the standard procedure. &#1040; set of 15, 214 (71.7%) scorable SNPs <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> was used for GWAS analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Microscopy and image processing</ns0:head><ns0:p>Sample preparation, microscopic image acquisition and processing were performed according to a previously developed procedure <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>. Five milligrams of raw starch were suspended in 1 mL of distilled water and dyed while shaking with 50 &#181;L of iodine solution. Twenty microliters of suspended dyed raw starch granules were placed on a microplate and covered with glass.</ns0:p><ns0:p>At least four pictures of every sample were acquired (250-300 granules in every image) in transmitted light mode with bright-field technique. Micro images of starch granules obtained with the research optical microscope Axio Scope A1 (Carl ZEISS), objective -A-Plan 10x/0.25, CCD camera -AxioCam ICc 3, adaptor -TV 2/3&#8242;&#8242;C 0.63x, software ZEN, total magnification 10 (objective) &#215;10 (ocular) &#215; 0.63 (adaptor).</ns0:p><ns0:p>Automatic image processing and analysis were performed in the freely distributed ImageJ program. Seven morphological parameters were analyzed (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_0'>S1</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Principal components and 2B PLS analysis</ns0:head><ns0:p>Two sets of principal components were calculated for both phenotypic morphological traits of starch granules, as well as for the genotyping data for potato varieties, through the distance matrix using the JACOBI 4 software package <ns0:ref type='bibr' target='#b16'>[16]</ns0:ref>. To calculate the distance matrix between the potato varieties, we recoded the tetraploid potato genome from four-letter codes to numerical codes, taking into account the dose of a certain allele. After the recoding, 0 was assigned to the effector allele, and 1 was assigned as the non-effector allele, and their intermediate forms were coded as 0.75, 0.5 and 0.25. For example, the AAAA allele is reflected as 1, AAAG -as 0.75, AAGG -as 0.5, AGGG -as 0.25, and GGGG -as 0.</ns0:p><ns0:p>Both sets of principal components were applied as blocks for two-block partial least squares (2B-PLS) analysis, where the first block related to phenotypic traits and the second block related to genotypic.</ns0:p><ns0:p>Population structure matrix (Q-matrix) and the genotyping data were analyzed by Bayesian cluster analysis in STRUCTURE v.2.3.4 <ns0:ref type='bibr' target='#b17'>[17]</ns0:ref>. Cluster analysis of the same population has been previously discussed in the authors' previous paper <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Association analysis</ns0:head><ns0:p>Association analysis was performed with the TASSEL 5 package <ns0:ref type='bibr' target='#b18'>[18]</ns0:ref>. Four different statistical models were tested to identify significant marker associations with potato starch yield and granule morphology: (1) general linear model (GLM) without taking into account population structure, (2) GLM using a Q-matrix of population membership (GLM+Q) taking into account the population structure, (3) GLM taking into account population membership estimates derived from principal components analysis (GLM + PCA), and (4) a composite approach that combines both Q-matrix and the average relationship between individuals or lines (null matrix), represented in TASSEL as mixed linear model (MLM). Adaptation of MLM for GWAS has been Manuscript to be reviewed discussed in <ns0:ref type='bibr' target='#b19'>[19]</ns0:ref>. But MLM approach did not result into valuable results and no significant SNPs were identified with this tool.</ns0:p><ns0:p>To identify significant SNPs, two corrections were used: (i) the Bonferroni correction, where the significant threshold (0.05) is divided by the total number of tests; in this work, the total number of markers (15,214) yields a threshold of 3.29&#215;10 -6 , and (ii) the false discovered rate (FDR) <ns0:ref type='bibr' target='#b20'>[20]</ns0:ref>, which was calculated for each isolate in each model. Some markers did not exceed the threshold but still possessed low p-value (&gt;10 -4 ). We referred to the markers as 'suggestive'.</ns0:p><ns0:p>The percentage of random was &lt;10%. Belonging of identified SNPs to genes and their association with certain proteins were confirmed on the site https://plants.ensembl.org.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head></ns0:div> <ns0:div><ns0:head>Phenotyping</ns0:head><ns0:p>Our initial study <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref> on potato tuber starch granule morphology provided a reliable method for granule shape evaluation. This work also indicated that granule shape is specific for different potato varieties and represents a set of quantitative traits that may be applied in modern potato breeding for production starch best suitable for industrial processing. Thus, 90 varieties of potato were harvested in 2017, and 90 samples of raw tuber starch and related DNA were isolated. Seven morphological parameters were automatically captured and evaluated for the 90 varieties: area (area of granule projection in microscope visible bright field), Feret's diameter, minimal Feret's diameter, aspect ratio (AR), roundness, circularity and solidity. Exact explanations and formulas for the parameters are provided in Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_0'>S1</ns0:ref>.</ns0:p><ns0:p>The preparative yield of starch was also captured. It varied significantly from 7.4% (variety Agata) to 18.8% (variety Tango) (Supplemental data S1, Excel file).</ns0:p><ns0:p>The process of capturing starch micro images, treatment and evaluation were discussed in our previous study <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>. For GWAS, we used average values of all the morphological parameters for every cultivar.</ns0:p></ns0:div> <ns0:div><ns0:head>Genome-wide association study</ns0:head><ns0:p>Principal components analysis and 2B PLS analysis 'Phenotype -genotype' covariation was calculated as a set of linear bicomponents (Table <ns0:ref type='table'>S2</ns0:ref>). It was shown that the first three linear bicomponents capture 92.7% of the total covariation. All the phenotypic traits studied, as well as preparative yield, correlate with the first bicomponent. The second component showed correlations with morphology traits and no correlations with the 'preparative yield of starch'. 'Preparative yield of starch' is the only trait that correlates significantly with the third bicomponent (Table <ns0:ref type='table'>S3</ns0:ref>). Plotting of the first and the third pairs of bicomponents showed good positive correlation between genotypes and phenotypes in both planes (Fig. <ns0:ref type='figure' target='#fig_3'>S2</ns0:ref>). The haplotypes most interesting for further selection are situated at the opposite corners of the plots and are highlighted by the circles (Fig. <ns0:ref type='figure'>S3</ns0:ref>). Despite the close values of the 'Preparative yield of starch' in varieties Udacha and Svitanok Kievsky, they are genetically highly different.</ns0:p><ns0:p>Analysis of genotypic data gives three clusters (Fig. <ns0:ref type='figure'>S4</ns0:ref>). The first and the second clusters comprise single trait -preparative yield of starch and aspect ratio, respectively. All other morphological traits studied formed the third cluster. Preparative yield of starch and aspect ratio are opposite traits on the plate. Therefore, the breeding process for optimizing the traits should be performed in opposite directions in genetic coordinates.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Quantile-quantile plots</ns0:head><ns0:p>For all the traits studied, we evaluated whether the analysis of the population structure possesses additional accuracy in finding significant SNPs. First, the GLM analysis data were compared in the QQ-plots (quantile-quantile plots). They show that only for traits 'circularity' and 'Feret's diameter' one may expect correct evaluation of noticeable SNPs with the GLM method. For the other traits, GLM demonstrates inflation (overestimated p-values) of noticeable SNP evaluation (Fig. <ns0:ref type='figure'>S5</ns0:ref>).</ns0:p><ns0:p>The GLM+PCA model takes into account the population structure and returns results of significantly higher quality (Fig. <ns0:ref type='figure'>S6</ns0:ref>): calculated p-values are closer to the expected ones even in the region of high values. Thus, calculated data for 'Circularity' and 'Area' traits are close to theoretical ones, for 'Solidity', 'Feret's diameter' and 'Minimal Feret's diameter', (-lg p)-values are somewhat lower. For other traits, obtained (-lg p)-values are slightly higher than expected. A similar QQ correlation for the 1 st PLS bicomponent, which includes all the genotypic traits, shows significant deviations from theoretical data and thus a lower quality of p-value evaluation (Fig. <ns0:ref type='figure'>S6</ns0:ref>).</ns0:p><ns0:p>To account for the population structure, the GLM+Q statistical model was applied. The model worked well for 'Feret's diameter', 'Minimal Feret's diameter', 'Circularity' and 'Area' traits giving a good conformity, but did not give (-lg p)-value &gt; 4, which indicates the absence of significant markers. The 'Preparative tuber starch' and 'Roundness' traits showed a number of false positive SNPs (Fig. <ns0:ref type='figure'>S7</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Table 1. Significant SNPs associated with preparative yield of starch and morphological traits.</ns0:head></ns0:div> <ns0:div><ns0:head>Manhattan plots</ns0:head><ns0:p>In total, 53 significant SNPs were associated with morphological and starch yield traits (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Recently, several genome association studies with starch granules size distribution were published for maize <ns0:ref type='bibr' target='#b21'>[21]</ns0:ref>, wheat <ns0:ref type='bibr' target='#b22'>[22,</ns0:ref><ns0:ref type='bibr' target='#b23'>23]</ns0:ref> and Japonica rice <ns0:ref type='bibr' target='#b24'>[24]</ns0:ref>. In the present study, no noticeable SNPs with p-values higher than Bonferroni or FDR levels were found for average values for the 'Area', 'Feret's diameter', 'Minimal Feret's diameter' and 'Solidity' traits. This result is surprising, since three of the traits are related to the starch granule size, which is expected to be genotypedependent. Indeed, if we apply ANOVA to the granules' sizes of 90 varieties involved in the study, we see that factor 'variety' clearly influences the size-related traits ('Area', 'Feret's diameter' or 'Minimal Feret's diameter'). Building a 'tree' of varieties' starch granule size identifies five clusters of varieties (genotypes) with relatively close granule size, and ANOVA within the clusters enables us to conclude that the size of the granules in the clusters is dependent on the genotype (Fig. <ns0:ref type='figure'>S8</ns0:ref>).</ns0:p><ns0:p>It appears that granule size is not single-or oligogenic but is a polygenic trait. To determine the SNP pattern associated with this trait, analysis of p-values and their comparison with Bonferroni or FDR levels is not sufficient, and more advanced and complicated analysis is warranted.</ns0:p><ns0:p>Preparative Tuber Starch Yield. Association of SNPs with phenotypic data for the 'preparative tuber starch yield' trait revealed 10 significant SNPs when the GLM model was applied. The pvalue of one SNP exceeded the Bonferroni level, and the p-values of the other nine exceeded the FDR (Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>). The SNPs were assigned to chromosomes 4, 5, 6, 7, 9, and 11. For the 'preparative tuber starch yield' trait, GLM+PCA appeared to be the best model according to the QQ plot (Fig. <ns0:ref type='figure'>S6</ns0:ref>), but for the model (-lg p)-values are lower than for GLM (Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>). No significant SNPs were revealed, but two SNPs on chromosomes 4 and 5 may be referred to as 'suggestive'. Nevertheless, the GLM+Q model confirmed a significant SNP on chromosome 4. Some significant regions on chromosomes 4, 5, 9, and 11 were revealed with GLM and confirmed by GLM+Q (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Taking into account possible false-positive SNPs predicted by the QQ-plot, PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed we lowered the FDR level following the suggestion of <ns0:ref type='bibr' target='#b25'>[25]</ns0:ref>. Setting the FDR level at the10.05 percentile of p-values gave a reasonable number of detectable SNPs (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p><ns0:p>Aspect Ratio. The most promising results obtained for morphological traits related not to size but to shape of the starch granules. GLM without population structure analysis yielded five SNPs with (-lg p)-values exceeding the Bonferroni level (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, Fig. <ns0:ref type='figure'>3</ns0:ref>). All of these SNPs are tightly grouped on chromosome 2, forming qualitative trait loci (QTL). Four other noticeable SNPs determined by GLM have (-lg p)-values exceeding FDR level and are located on chromosomes 1, 7, 11 and 12. The same three SNPs were confirmed by the GLM+PCA method and are probably parts of appropriate QTLs. Belonging to QTLs increases the probability that certain SNPs are associated with the trait studied.</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 3. Manhattan plots for &#171;Aspect Ratio&#187; trait. A) GLM; B) GLM+PCA; C) GLM+Q.</ns0:head><ns0:p>Roundness. Analysis of SNPs for the 'Roundness' trait in detail reproduced the results obtained for the 'Aspect ratio'. Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> contains SNPs associated with the 'Aspect ratio' and includes SNPs for 'Roundness' as well. Manhattan plots are highly similar for the two traits because both describe the shape of potato starch granules.</ns0:p><ns0:p>Circularity. Three SNPs associated with the 'Circularity' trait were revealed by GWAS using the statistical models GLM+PCA and GLM+Q (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, Fig. <ns0:ref type='figure'>S9</ns0:ref>).</ns0:p><ns0:p>SNPs correlated with the first phenotypic bicomponent. Since the 'Preparative yield', 'Area', 'Circularity', 'Feret's diameter', 'Minimal Feret's diameter' and 'Solidity' traits mutually contribute the first principal component significantly (Table <ns0:ref type='table'>S3</ns0:ref>), it was logical to analyze SNPs' associations with the first bicomponent to determine if there were any SNPs associated with all of the traits simultaneously.</ns0:p><ns0:p>One of the SNPs found for the first bicomponent (solcap_snp_c1_15462 on the 7th chromosome) fits the SNP, which is significant for the 'Circularity' trait. The 'Circularity' trait contributes the first bicomponent (-0.67992). The other four significant SNPs are unique and not associated with any Manuscript to be reviewed single trait. All of these SNPs are a part of appropriate QTLs ('supported' with other SNPs with lower (-lg p)-values) (Fig. <ns0:ref type='figure' target='#fig_0'>S10</ns0:ref>). Genes around the SNPs may be responsible for the general starch carbohydrate polymer arrangement in the starch granule.</ns0:p><ns0:p>No significant SNPs associated with the second and third bicomponents were revealed.</ns0:p></ns0:div> <ns0:div><ns0:head>SNPs and Related Genes/Proteins</ns0:head><ns0:p>In total, 53 SNPs were found to be significant in the association study of the potato genotypes and starch morphology/yield-related traits. (-lg p)-Values of one SNP were associated with 'preparative yield of starch', five SNPs -with 'aspect ratio' and 'roundness' traits, and one SNP -with the 1 st bicomponent exceeded strict Bonferroni criteria. Nevertheless, some associated SNPs overcame 0.1 and 0.05 FDR (false discovery rate) and suggestive levels. Most of the SNPs are located on the 1 st and 2 nd chromosomes, the least on chromosomes 4, 5, 6, 7, 9, 11, and 12.</ns0:p><ns0:p>Thirty-seven of 53 SNPs are located in protein coding regions. Not one of the SNPs studied in this paper coincides with the ones associated with covalently bound phosphorus content in starch <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> (Table <ns0:ref type='table'>S4</ns0:ref>). </ns0:p></ns0:div> <ns0:div><ns0:head>Preparative yield of starch</ns0:head></ns0:div> <ns0:div><ns0:head>Aspect ratio and Roundness</ns0:head><ns0:p>These traits are well-correlated with each other and describe the shape of starch granules in a similar way; thus, the traits are associated with the same SNPs. There is a locus on chromosome 1 containing eight SNPs in the same DNA region. Some of these proteins are related to circadian rhythm-regulating proteins. The circadian clock regulates numerous plant developmental and metabolic processes. In crop species, the circadian clock contributes significantly to plant performance and productivity and to the adaptation and geographical range over which crops can be grown. Other SNPs are related to the phosphorylation of important biochemical intermediates and plastid organization. On chromosome 2, a total of 19 significant SNPs were identified. The SNPs are narrowly situated, forming a single DNA region, a potential trait-related QTL. Most of the SNPs are located in noncoding regions or related to proteins with unknown functions. Two SNPs are related to WPP domain-associated protein-encoding genomic regions, one to plastid high chlorophyll fluorescence 136, another to pentatricopeptide repeat-containing protein, and one SNP is related to DNA binding protein. Two SNPs on chromosomes 7 and 11 were found in the genes encoding proteins with unknown functions.</ns0:p></ns0:div> <ns0:div><ns0:head>Circularity</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Only three SNPs associated with the circularity trait were identified. The first SNP is related to a noncoding region of chromosome 12, the second is located on chromosome 7 in the minichromosome maintenance 5 protein coding sequence, and the third is located on chromosome 11 in the flavonoid 3',5'-hydroxylase-encoding region.</ns0:p></ns0:div> <ns0:div><ns0:head>First 'phenotype' bicomponent</ns0:head><ns0:p>Six SNPs were found to be associated with the first bicomponent from 2B-PLS, which is a complex component comprising all of the traits studied. Thus, one SNP was the same for the 1st bicomponent and Preparative yield (chromosome 7), while another SNP was the same for the first bicomponent and Circularity (chromosome 5). Among the four other unique 1st bicomponent associated SNPs, two are related to the GWD gene, one of the key starch biosynthesis genes <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref>, which is responsible for the phosphorylation-dephosphorylation of glucans (chromosome 9).</ns0:p><ns0:p>The other two SNPs are associated with cytochrome B561 family protein (chromosome 5) and Pto-interacting protein 1 (chromosome 12). In summary, chromosomes 1 and 2 contain important regions responsible for the roundness and aspect ratio of tuber starch granules. There are indications that granule shape may depend on circadian rhythm-related metabolic processes and starch phosphorylation processes. The GWD gene, which is known to regulate phosphorylation and dephosphorylation participates in the regulation of a whole number of morphological traits, rather than a single certain one.</ns0:p><ns0:p>Nevertheless, some other mechanisms and proteins located on chromosome 2 influence the granule formation process. The preparative yield of tuber starch is probably a polygenic trait, regulated by a number of proteins that are encoded by sequences in various parts and chromosomes of the potato genome.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>A genome-wide association study using a 22K SNP potato array enabled 53 novel SNPs to be identified on chromosomes 1, 2, 4, 5, 6, 7, 9, 11, and 12; these SNPs are associated with tuber starch preparative yield and with starch granule morphology (aspect ratio, roundness, circularity, Manuscript to be reviewed Significant SNPs associated with preparative yield of starch and morphological traits.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Hand-drawn table entitled 'Starch granules size of various potato varieties'. Located in Vavilov memorial room of the N.I. Vavilov All-Russian Research Institute of Plant Genetic Resources (VIR), St. Petersburg, Russia. Three rows (of 3 small images each) are entitled 'Small starch granules', 'Middle starch granules', 'Large starch granules' from top to down. Every small image contains scale in microns, sign 'Magnification 280', and the name of wild potato variety under it.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Manhattan plots for 'Preparative tuber starch yield' trait, A) GLM B) GLM+PCA C) GLM+Q.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Three out of four SNPs found on chromosome 4 are located in noncoding sequences and relate to two different DNA regions. There are some other related SNPs with considerably lower significance in these regions, indicating possible QTLs. The fourth SNP is situated in the different region of the gene, coding for a DNA transcription regulator consisting of 95 amino acid residues.All three SNPs found on chromosome 5 are related to the sequences coding different proteins.One SNP corresponds to coiled coil protein with unclear function; two others are related to lowmolecular-weight organic metabolite conversion, specifically to the enzymes aldehyde dehydrogenase and phenyl alanine lyase, which are responsible for alcohol-aldehyde equilibria in the cell and for flavonoid synthesis. Taking into account SNPs that are less significant but still associated with starch yield, we may speak about two trait-associated DNA regions that arePeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)Manuscript to be reviewed potential QTLs. The significant SNP on chromosome 6 seems also be a part of a QTL and relates to plastid transcriptionally active protein 16 (PTAC16), which is suspected to be involved in the regulation of plastid gene expression. The significant SNP on chromosome 7 encodes the protein translation factor SUI1. The significant SNP on chromosome 9 encodes a protein with unknown function. A group of five SNPs located within 2855044 to 3572445 bp on chromosome 11 belongs to the same QTL, and the SNPs are included in the coding regions of several proteins: DEGP10, methylenetetrahydrofolate reductase, acetolactate synthase, and two proteins of unknown function. In general, most of the proteins associated with potato starch preparative yield variations are involved in plastid activity and low-molecular-weight metabolite biosynthesis.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)Manuscript to be reviewed and the 1 st bicomponent). Some of the SNPs observed in this study are located in noncoding regions. The coding regions are associated with membrane and plastid proteins, DNA transcription and binding regulators, low-molecular-weight metabolite synthesis as well as flavonoid biosynthesis. The information on significant regions can be used to convert SNPs to PCR-markers, convenient for screening breeding material in programs aimed on development of potato varieties with desired starch properties.PeerJ reviewing PDF | (2020:05:48564:1:0:NEW 5 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> </ns0:body> "

"Dear Editor, Authors thank all the Reviewers for their notes and suggestions. We carefully worked through each of them and included corrections into the main text where needed. Below we gave our answers to the Reviewers’ questions and suggestions (blue). Thank you, Vadim K. Khlestkin, on behalf of the authors. Reviewer 1 (Anonymous) Comments for the Author The paper is on the genetic loci determination related to starch granules in potato. A total of 90 potato samples were used. GWD gene is related to the results. I believe that the paper is interesting and showed new insights in to starch biosynthesis in potatoes. i only have few comments and rather suggest the paper to be acceptance after making comments to mine. There are ;already papers on the genes responsible for the granule morphology in different crops such as arabidopsis. I hope the authors may comment on how the findings from other crops can be related to the potatoes. Arabidopsis is not an important starch crop, so, its STARCH GRANULES MORPHOLOGY (NOT morphology of seeds or leaves) at genetic level was not yet explored. We know about quite recent study (Natural Polymorphisms in Arabidopsis Result in Wide Variation or Loss of the Amylose Component of Starch David Seung, Alberto Echevarría-Poza, Burkhard Steuernagel, Alison M. Smith Published February 2020. DOI: https://doi.org/10.1104/pp.19.01062), but it does not deal with starch granules. But for some industrial crops some genetic associations studies are performed. We added a few references on wheat, rice and maize starch granules size distribution, refs 21-24. Figure 1, the captions should be translated as most of the readers don't learn Russian. also, which year is this drawn? We have added a few lines, related to the history of the picture and translation of the signs in it, in the text and Figure 1 comment. GWAS in the title, this may be spelled out in full. Done. How about the number of starch granules in the amyloplasts? That point is out of the scope of our current study. We are focused on isolated starch, taking into account its potential use for further industrial treatment. However, the parameter suggested by reviewer may be an interesting direction for subsequent research. There are still some errors in your writings, Please double check your paper again to make sure not many mistakes are there. Done. Reviewer 2 (Anonymous) Basic reporting 1. As I am from a non-English speaking country, it is difficult to judge the English of the manuscript. We double checked our text, thank you. 2. I feel that in the section Results and Discussion, the literature referencing is insufficient. For instance, the authors discuss the connection of identified SNPs to certain proteins associated with studied traits. However, no references were provided neither for proteins itself nor for the hypothesis behind putative associations. Phrase about annotation of SNPs on https://plants.ensembl.org was added in Materials and Methods. That is the only source used for association of SNPs with proteins. 3. In the section Materials, it is not clear why the authors did not provide a full list of accessions in the study? A simple reference to another published work is not sufficient. Please provide a separate file with the list of accessions, their origin, and other details. The complete list of accessions is presented in Supplemental data S1 Excel file. It also contains average values of morphological parameters for each accession. Their origin is added to the file. We added reference to that file in the main text in Plant material section. 4. It is not clear why the MLM method was not applied in the GWAS, although the method was stated in the section “Materials and Methods”? A small-sized population and the GLM (alone or with a combination of matrices) are often lead to the detection of false-positive signals due to inflated p-value. Indeed, MLM did not result in any significant SNPs, that’s why it was not shown in pictures and in tables. We added the explanation phrase in Materials and Methods. 5. I think that Table S4 should be moved to the text itself. The Table itself should have more information, such as QTL effect and trait heritability values. In figures with Manhattan plots, I think that all figs should be given with side-by-side QQ plots. Thank you for the suggestion. In the first version of the manuscript we had some more tables and pictures incorporated in the main text. We were asked to remove some information to Supplemental materials. So, for now we leave it as it is. But in case Editor strongly recommend to add the table and pictures back, we will do that. Experimental design I have no question to this section, except the fact that the studied population is rather small-sized, which may lead to potential false statements in QTN identification for studied traits. The population is not too wide, but even that number of varieties allows to discover some SNPs and QTLs. Taking onto account that potato species are clones within variety, the number of varieties is acceptable. Of course, the results will be confirmed with other types of markers (KASP, for example), which is mentioned in the Conclusion. Validity of the findings The results of the study are fairly novel. However, I recommend the authors use not only GLM but also the MLM method, which is a more reliable approach to avoid the detection of possible false-positive associations. Conclusions should be reassessed after the application of MLM methods in the GWAS. We used MLM with no significant result (see one of our answers above). Comments for the Author Table S4 is missing information for the QTL effects of the associations. Also, please provide heritability values for analyzed traits in the study? Done. Reviewer 3 (Anonymous) Basic reporting Ok Experimental design OK Validity of the findings OK Comments for the Author 1. Fig.1 should have english note beside the Russian; Done. Added in main text and in Figure note. 2. Only one environment for the trait phenotyping, can this make the significant SNP reliable; SNPs’ associations with traits will be double check and confirmed with other type of markers in future works. 3. Move Some figures and tables to main text from the supplimentary file; Table 1 moved back to the main text. 4. Is there a candidate gene on Chr.2H for the siginficant SNP or a published QTL for this region; We found several significant SNPs on Chr2. Some of them do not belong to a known genes. But some of them belong to different genes including ones with unknown function. That information may be found in Table S5. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. It is well-documented that (bio)chemical reaction capacity of raw potato starch depends on crystallinity, morphology and other chemical and physical properties of starch granules, and these properties are closely related to gene functions. Preparative yield, amylose/amylopectin content, and phosphorylation of potato tuber starch are starch-related traits studied at the genetic level. In this paper, we perform a genome-wide association study using a 22K SNP potato array to identify for the first time genomic regions associated with starch granule morphology and to increase number of known genome loci associated with potato starch yield.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods.</ns0:head><ns0:p>A set of 90 potato (Solanum tuberosum L.) varieties from the ICG 'GenAgro' collection (Novosibirsk, Russia) was harvested, 90 samples of raw tuber starch were obtained, and DNA samples were isolated from the skin of the tubers. Morphology of potato tuber starch</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. It is well-documented that (bio)chemical reaction capacity of raw potato starch depends on crystallinity, morphology and other chemical and physical properties of starch granules, and these properties are closely related to gene functions. Preparative yield, amylose/amylopectin content, and phosphorylation of potato tuber starch are starch-related traits studied at the genetic level. In this paper, we perform a genome-wide association study using a 22K SNP potato array to identify for the first time genomic regions associated with starch granule morphology and to increase number of known genome loci associated with potato starch yield.</ns0:p><ns0:p>Methods. A set of 90 potato (Solanum tuberosum L.) varieties from the ICG 'GenAgro' collection (Novosibirsk, Russia) was harvested, 90 samples of raw tuber starch were obtained, and DNA samples were isolated from the skin of the tubers. Morphology of potato tuber starch granules was evaluated by optical microscopy and subsequent computer image analysis. A set of 15,214 scorable SNPs was used for the genome-wide analysis. In total, 53 SNPs were found to be significantly associated with potato starch morphology traits (aspect ratio, roundness, circularity, and the 1 st bicomponent) and starch yield-related traits.</ns0:p><ns0:p>Results. A total of 53 novel SNPs was identified on potato chromosomes 1, 2, 4, 5, 6, 7, 9, 11, and 12; these SNPs are associated with tuber starch preparative yield and granule morphology. Eight SNPs are situated close to each other on the chromosome 1 and nineteen SNPs -on the chromosome 2, forming two DNA regions -potential QTLs, regulating Aspect ratio and Roundness of the starch granules. Thirtyseven of 53 SNPs are located in protein-coding regions. There are indications that granule shape may depend on starch phosphorylation processes. The GWD gene, which is known to regulate starch phosphorylation -dephosphorylation, participates in the regulation of a number of morphological traits, rather than one specific trait. Some significant SNPs are associated with membrane and plastid proteins, as well as DNA transcription and binding regulators. Other SNPs are related to low-molecular-weight metabolite synthesis, and may be associated with flavonoid biosynthesis and circadian rhythm-related metabolic processes. The preparative yield of tuber starch is a polygenic trait that is associated with a number of SNPs from various regions and chromosomes in the potato genome.</ns0:p></ns0:div> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Starch is one of the most important renewable and economically notable organic resources of humankind <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>. Simplicity of potato growing and ease of industrial production of pure potato starch makes it a necessary component and starting compound in food and chemical industries and determines valuability of an exhaustive study of all aspects of manufacturing and commercial application of potato starch is necessary. There are many publications related to chemical and biochemical transformations of starch of diverse botanical origin and differences properties of various starches. Surprisingly, little attention is paid to difference in starch properties at the level of different cultivars, such as potato cultivars. Commercial starch is still subdivided as 'potato starch', 'rice starch', 'corn starch', and so on, despite it is well-known that there are significant differences in the properties of starch of the same botanical origin. From general considerations, it is evident that chemical and biochemical reaction ability of the raw potato starch must depend on its granules crystallinity and morphology. If so, starches from different potato varieties with different granules' properties should fit different applications better. (Bio)chemical reaction capacity of raw potato starch depends on the crystallinity, morphology and other chemical and physical properties of starch granules. For example, after chemical acylation of the raw potato starch and fractionation of resulting acetylated starch according to granule size, amylose and amylopectin may be isolated and characterized by degree of substitution (DS) and degradability with &#945;-amylase, &#946;-amylase and amyloglucosidase. In contrast to amylose, the DS of the amylopectin from the differently sized granules increased with decreasing granule size. The acetyl groups of the amylose molecules from small granules are more heterogeneously distributed and located more closely to the non-reducing ends compared to amylose from larger granules. The amylose populations from small granules of the acetylated starches were less susceptible to all the enzyme degradation reactions than the amylose from the large granules, even though the DS was similar. Additionally, the acetyl group distributions were different for amylopectin from different granule size fractions. <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b3'>3]</ns0:ref>. Some benefits of small potato starch granules are summarized in European patent <ns0:ref type='bibr' target='#b4'>[4]</ns0:ref>. The large size of the potato granules often hampers the utilization of potato starches, for example, in printing ink formulations, wherein potato starch granules may obstruct the printing apertures. Smaller granules facilitate chemical modification of the starch, since the surface-volume ratio is important for the accessibility of the modifying compound to the starch chains. In textile printing, for instance, carboxymethylated and roll dried potato starch of small granule size enables the production of finer prints. Furthermore, in adhesives, particularly in highly concentrated bag adhesives, crosslinked fine potato starches may be used as fillers and in drilling fluids, crosslinked and modified fine starches are expected to reduce fluid loss. In addition, smaller granule potato starches may be used as filler/structurant in soap, since this starch provides a pleasant sensation to the skin. Small granular potato starches are particularly suitable for applications in the food industry. Small granular potato starch provides such advantages as lower starch dosage level, better taste profile, and smooth but not excessively swollen granules <ns0:ref type='bibr' target='#b5'>[5]</ns0:ref>. Other practical and important properties that are sensitive to starch granule size are liquid composite viscosity <ns0:ref type='bibr' target='#b6'>[6]</ns0:ref>, chemical modification and flowability <ns0:ref type='bibr' target='#b7'>[7]</ns0:ref>.</ns0:p><ns0:p>It is evident that the shape of potato starch granules is closely related to genes function. Amylopectin polysaccharide is predominantly responsible for the smooth granule morphology.</ns0:p><ns0:p>Even low-amylose starch granules still maintain a smooth shape, whereas starch with reduced amylopectin content (by antisense knockout of SBE and GWD genes) is associated with more oddly fissured granules <ns0:ref type='bibr' target='#b8'>[8]</ns0:ref>. The role of starch synthases and related genes (GBSS, SSI -SSIII) in potato starch granule morphology has been discussed elsewhere <ns0:ref type='bibr' target='#b9'>[9]</ns0:ref>. In general, the role of certain genes in starch biosynthesis and starch granule morphology in particular is summarized in the review <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref>.</ns0:p><ns0:p>Traditionally, the preparative yield of potato tuber starch is the only starch-related trait generally accepted as important for potato breeding. In the past decade, starch phosphorylation and Manuscript to be reviewed amylose/amylopectin content were also added to the short list of the traits studied at the genetic level by modern molecular biology methods. Thus, forward genetic tools, such as QTL analysis <ns0:ref type='bibr' target='#b11'>[11,</ns0:ref><ns0:ref type='bibr' target='#b12'>12]</ns0:ref> and association mapping <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>, were used to reveal loci associated with potato traits. The tuber starch characteristics, such as starch granular size and shape, and chemical-thermal properties of 21 potato varieties were determined and associated with genetic diversity through SSR markers. SSR-based cluster analysis revealed that varieties with interesting quality attributes were distributed among all clusters and subclusters, suggesting that the genetic basis of analyzed traits might differ among the varieties <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>.</ns0:p><ns0:p>Recently developed 22K SNP potato array is characterized by a high average density of markers, one locus per 40 kb (in the abovementioned studies <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref><ns0:ref type='bibr' target='#b11'>[11]</ns0:ref><ns0:ref type='bibr' target='#b12'>[12]</ns0:ref><ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>, this value does not exceed one marker per 4 Mbp). We used the chip to identify eight novel genomic regions on the chromosomes 1, 4, 5, 7, 8, 10, and 11 associated with starch phosphorylation. Some of the identified SNPs were located in noncoding genomic regions <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref>.</ns0:p><ns0:p>In this paper, for the first time we perform a genome-wide association study using a 22K SNP potato array to locate genomic regions associated with starch granule morphology and to increase the number of known genomic loci associated with potato starch yield.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Plant material</ns0:head><ns0:p>The same set of 90 potato (Solanum tuberosum L.) varieties from the ICG 'GenAgro' collection (Novosibirsk, Russia) described in <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> was used in this paper as well. Complete list of the accessions used in the study and their origin is presented in Supplemental data S1 (Excel file).</ns0:p><ns0:p>All plants are grown during the period May to October 2017 in the same field in Novosibirsk (54&#176;52' N and 83&#176;00' E). Seed tubers of all cultivars were planted in two rows with 0.75 m spacing and 0.3 m distance between the plants on the rows. In total, 10 plants were planted in Manuscript to be reviewed the row; the length of each row was 10 m. Each cultivar was planted in three replicates; distances between the replicates' plots were 2 m. Sowing: the first decade of May. Harvesting: the 3rd decade of September. After harvesting tubers were stored for three weeks at +4&#176;C before starch isolation.</ns0:p></ns0:div> <ns0:div><ns0:head>Starch isolation</ns0:head><ns0:p>Potato starch was isolated from tubers according to the typical procedure described elsewhere (for example, see <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>DNA isolation and genotyping</ns0:head><ns0:p>DNA was isolated from tuber skin using DNeasy Plant Mini Kit (Qiagen) according to the standard procedure. &#1040; set of 15, 214 (71.7%) scorable SNPs <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> was used for GWAS analysis.</ns0:p><ns0:p>Genotyping information is available in Supplemental data S2 (Excel file).</ns0:p><ns0:p>.</ns0:p></ns0:div> <ns0:div><ns0:head>Microscopy and image processing</ns0:head><ns0:p>Sample preparation, microscopic image acquisition and processing were performed according to a previously developed procedure <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>. Five milligrams of raw starch were suspended in 1 mL of distilled water and dyed while shaking with 50 &#181;L of iodine solution. Twenty microliters of suspended dyed raw starch granules were placed on a microplate and covered with glass.</ns0:p><ns0:p>At least four pictures of every sample were acquired (250-300 granules in every image) in transmitted light mode with bright-field technique. Micro images of starch granules obtained with the research optical microscope Axio Scope A1 (Carl ZEISS), objective -A-Plan 10x/0.25, CCD camera -AxioCam ICc 3, adaptor -TV 2/3&#8242;&#8242;C 0.63x, software ZEN, total magnification 10 (objective) &#215;10 (ocular) &#215; 0.63 (adaptor).</ns0:p><ns0:p>Automatic image processing and analysis were performed in the freely distributed ImageJ program. Seven morphological parameters were analyzed (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>, Fig. <ns0:ref type='figure'>S1</ns0:ref>). Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Principal components and 2B PLS analysis</ns0:head><ns0:p>Two sets of principal components were calculated for both phenotypic morphological traits of starch granules, as well as for the genotyping data for potato varieties, through the distance matrix using the JACOBI 4 software package <ns0:ref type='bibr' target='#b16'>[16]</ns0:ref>. To calculate the distance matrix between the potato varieties, we recoded the tetraploid potato genome from four-letter codes to numerical codes, taking into account the dose of a certain allele. After the recoding, 0 was assigned to the effector allele, and 1 was assigned as the non-effector allele, and their intermediate forms were coded as 0.75, 0.5 and 0.25. For example, the AAAA allele is reflected as 1, AAAG -as 0.75, AAGG -as 0.5, AGGG -as 0.25, and GGGG -as 0.</ns0:p><ns0:p>Both sets of principal components were applied as blocks for two-block partial least squares (2B-PLS) analysis, where the first block related to phenotypic traits and the second block related to genotypic.</ns0:p><ns0:p>Population structure matrix (Q-matrix) and the genotyping data were analyzed by Bayesian cluster analysis in STRUCTURE v.2.3.4 <ns0:ref type='bibr' target='#b17'>[17]</ns0:ref>. Cluster analysis of the same population has been previously discussed in the authors' previous paper <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Association analysis</ns0:head><ns0:p>Association analysis was performed with the TASSEL 5 package <ns0:ref type='bibr' target='#b18'>[18]</ns0:ref>. Four different statistical models were tested to identify significant marker associations with potato starch yield and granule morphology: (1) general linear model (GLM) without taking into account population structure, (2) GLM using a Q-matrix of population membership (GLM+Q) taking into account the population structure, (3) GLM taking into account population membership estimates derived from principal components analysis (GLM + PCA), and (4) a composite approach that combines both Q-matrix and the average relationship between individuals or lines (null matrix), represented in TASSEL as mixed linear model (MLM). Adaptation of MLM for GWAS has been Manuscript to be reviewed discussed in <ns0:ref type='bibr' target='#b19'>[19]</ns0:ref>. But MLM approach did not result into valuable results and no significant SNPs were identified with this tool.</ns0:p><ns0:p>To identify significant SNPs, two corrections were used: (i) the Bonferroni correction, where the significant threshold (0.05) is divided by the total number of tests; in this work, the total number of markers (15,214) yields a threshold of 3.29&#215;10 -6 , and (ii) the false discovered rate (FDR) <ns0:ref type='bibr' target='#b20'>[20]</ns0:ref>, which was calculated for each isolate in each model. Some markers did not exceed the threshold but still possessed low p-value (&gt;10 -4 ). We referred to the markers as 'suggestive'.</ns0:p><ns0:p>The percentage of random was &lt;10%. Belonging of identified SNPs to genes and their association with certain proteins were confirmed on the site https://plants.ensembl.org.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head></ns0:div> <ns0:div><ns0:head>Phenotyping</ns0:head><ns0:p>Our initial study <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref> on potato tuber starch granule morphology provided a reliable method for granule shape evaluation. This work also indicated that granule shape is specific for different potato varieties and represents a set of quantitative traits that may be applied in modern potato breeding for production starch best suitable for industrial processing. Thus, 90 varieties of potato were harvested in 2017, and 90 samples of raw tuber starch and related DNA were isolated. Seven morphological parameters were automatically captured and evaluated for the 90 varieties: area (area of granule projection in microscope visible bright field), Feret's diameter, minimal Feret's diameter, aspect ratio (AR), roundness, circularity and solidity. Exact explanations and formulas for the parameters are provided in Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref> and Fig. <ns0:ref type='figure'>S1</ns0:ref>.</ns0:p><ns0:p>The preparative yield of starch was also captured. It varied significantly from 7.4% (variety Agata) to 18.8% (variety Tango) (Supplemental data S1, Excel file).</ns0:p><ns0:p>The process of capturing starch micro images, treatment and evaluation were discussed in our previous study <ns0:ref type='bibr' target='#b15'>[15]</ns0:ref>. For GWAS, we used average values of all the morphological parameters for every cultivar.</ns0:p></ns0:div> <ns0:div><ns0:head>Genome-wide association study</ns0:head><ns0:p>Principal components analysis and 2B PLS analysis 'Phenotype -genotype' covariation was calculated as a set of linear bicomponents (Table <ns0:ref type='table'>S2</ns0:ref>). It was shown that the first three linear bicomponents capture 92.7% of the total covariation. All the phenotypic traits studied, as well as preparative yield, correlate with the first bicomponent. The second component showed correlations with morphology traits and no correlations with the 'preparative yield of starch'. 'Preparative yield of starch' is the only trait that correlates significantly with the third bicomponent (Table <ns0:ref type='table'>S3</ns0:ref>). Plotting of the first and the third pairs of bicomponents showed good positive correlation between genotypes and phenotypes in both planes (Fig. <ns0:ref type='figure'>S2</ns0:ref>). The haplotypes most interesting for further selection are situated at the opposite corners of the plots and are highlighted by the circles (Fig. <ns0:ref type='figure'>S3</ns0:ref>). Despite the close values of the 'Preparative yield of starch' in varieties Udacha and Svitanok Kievsky, they are genetically highly different.</ns0:p><ns0:p>Analysis of genotypic data gives three clusters (Fig. <ns0:ref type='figure'>S4</ns0:ref>). The first and the second clusters comprise single trait -preparative yield of starch and aspect ratio, respectively. All other morphological traits studied formed the third cluster. Preparative yield of starch and aspect ratio are opposite traits on the plate. Therefore, the breeding process for optimizing the traits should be performed in opposite directions in genetic coordinates. Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Quantile-quantile plots</ns0:head><ns0:p>For all the traits studied, we evaluated whether the analysis of the population structure possesses additional accuracy in finding significant SNPs. First, the GLM analysis data were compared in the QQ-plots (quantile-quantile plots). They show that only for traits 'circularity' and 'Feret's diameter' one may expect correct evaluation of noticeable SNPs with the GLM method. For the other traits, GLM demonstrates inflation (overestimated p-values) of noticeable SNP evaluation (Fig. <ns0:ref type='figure'>S5</ns0:ref>).</ns0:p><ns0:p>The GLM+PCA model takes into account the population structure and returns results of significantly higher quality (Fig. <ns0:ref type='figure'>S6</ns0:ref>): calculated p-values are closer to the expected ones even in the region of high values. Thus, calculated data for 'Circularity' and 'Area' traits are close to theoretical ones, for 'Solidity', 'Feret's diameter' and 'Minimal Feret's diameter', (-lg p)-values are somewhat lower. For other traits, obtained (-lg p)-values are slightly higher than expected. A similar QQ correlation for the 1 st PLS bicomponent, which includes all the genotypic traits, shows significant deviations from theoretical data and thus a lower quality of p-value evaluation (Fig. <ns0:ref type='figure'>S6</ns0:ref>).</ns0:p><ns0:p>To account for the population structure, the GLM+Q statistical model was applied. The model worked well for 'Feret's diameter', 'Minimal Feret's diameter', 'Circularity' and 'Area' traits giving a good conformity, but did not give (-lg p)-value &gt; 4, which indicates the absence of significant markers. The 'Preparative tuber starch' and 'Roundness' traits showed a number of false positive SNPs (Fig. <ns0:ref type='figure'>S7</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Table 1. Significant SNPs associated with preparative yield of starch and morphological traits.</ns0:head></ns0:div> <ns0:div><ns0:head>Manhattan plots</ns0:head><ns0:p>In total, 53 significant SNPs were associated with morphological and starch yield traits (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p><ns0:p>Recently, several genome association studies with starch granules size distribution were published for maize <ns0:ref type='bibr' target='#b21'>[21]</ns0:ref>, wheat <ns0:ref type='bibr' target='#b22'>[22,</ns0:ref><ns0:ref type='bibr' target='#b23'>23]</ns0:ref> and Japonica rice <ns0:ref type='bibr' target='#b24'>[24]</ns0:ref>. In the present study, no noticeable SNPs with p-values higher than Bonferroni or FDR levels were found for average values for the 'Area', 'Feret's diameter', 'Minimal Feret's diameter' and 'Solidity' traits. This result is surprising, since three of the traits are related to the starch granule size, which is expected to be genotypedependent. Indeed, if we apply ANOVA to the granules' sizes of 90 varieties involved in the study, we see that factor 'variety' clearly influences the size-related traits ('Area', 'Feret's diameter' or 'Minimal Feret's diameter'). Building a 'tree' of varieties' starch granule size identifies five clusters of varieties (genotypes) with relatively close granule size, and ANOVA within the clusters enables us to conclude that the size of the granules in the clusters is dependent on the genotype (Fig. <ns0:ref type='figure'>S8</ns0:ref>).</ns0:p><ns0:p>It appears that granule size is not single-or oligogenic but is a polygenic trait. To determine the SNP pattern associated with this trait, analysis of p-values and their comparison with Bonferroni or FDR levels is not sufficient, and more advanced and complicated analysis is warranted.</ns0:p><ns0:p>Preparative Tuber Starch Yield. Association of SNPs with phenotypic data for the 'preparative tuber starch yield' trait revealed 10 significant SNPs when the GLM model was applied. The pvalue of one SNP exceeded the Bonferroni level, and the p-values of the other nine exceeded the FDR (Fig. <ns0:ref type='figure'>1</ns0:ref>). The SNPs were assigned to chromosomes 4, 5, 6, 7, 9, and 11.</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 1. Manhattan plots for 'Preparative tuber starch yield' trait, A) GLM B) GLM+PCA C) GLM+Q.</ns0:head><ns0:p>For the 'preparative tuber starch yield' trait, GLM+PCA appeared to be the best model according to the QQ plot (Fig. <ns0:ref type='figure'>S6</ns0:ref>), but for the model (-lg p)-values are lower than for GLM (Fig. <ns0:ref type='figure'>1</ns0:ref>). No significant SNPs were revealed, but two SNPs on chromosomes 4 and 5 may be referred to as 'suggestive'. Nevertheless, the GLM+Q model confirmed a significant SNP on chromosome 4. Some significant regions on chromosomes 4, 5, 9, and 11 were revealed with GLM and confirmed by GLM+Q (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Taking into account possible false-positive SNPs predicted by the QQ-plot, we lowered the FDR level following the suggestion of <ns0:ref type='bibr' target='#b25'>[25]</ns0:ref>. Setting the FDR level at the10.05 percentile of p-values gave a reasonable number of detectable SNPs (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p><ns0:p>Aspect Ratio. The most promising results obtained for morphological traits related not to size but to shape of the starch granules. GLM without population structure analysis yielded five SNPs with (-lg p)-values exceeding the Bonferroni level (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, Fig. <ns0:ref type='figure'>2</ns0:ref>). All of these SNPs are tightly grouped on chromosome 2, forming qualitative trait loci (QTL). Four other noticeable SNPs determined by GLM have (-lg p)-values exceeding FDR level and are located on chromosomes 1, 7, 11 and 12. The same three SNPs were confirmed by the GLM+PCA method and are probably parts of appropriate QTLs. Belonging to QTLs increases the probability that certain SNPs are associated with the trait studied.</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 2. Manhattan plots for &#171;Aspect Ratio&#187; trait. A) GLM; B) GLM+PCA; C) GLM+Q.</ns0:head><ns0:p>Roundness. Analysis of SNPs for the 'Roundness' trait in detail reproduced the results obtained for the 'Aspect ratio'. Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> contains SNPs associated with the 'Aspect ratio' and includes SNPs for 'Roundness' as well. Manhattan plots are highly similar for the two traits because both describe the shape of potato starch granules.</ns0:p><ns0:p>Circularity. Three SNPs associated with the 'Circularity' trait were revealed by GWAS using the statistical models GLM+PCA and GLM+Q (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, Fig. <ns0:ref type='figure'>S9</ns0:ref>).</ns0:p><ns0:p>SNPs correlated with the first phenotypic bicomponent. Since the 'Preparative yield', 'Area', 'Circularity', 'Feret's diameter', 'Minimal Feret's diameter' and 'Solidity' traits mutually contribute the first principal component significantly (Table <ns0:ref type='table'>S3</ns0:ref>), it was logical to analyze SNPs' associations with the first bicomponent to determine if there were any SNPs associated with all of the traits simultaneously.</ns0:p><ns0:p>One of the SNPs found for the first bicomponent (solcap_snp_c1_15462 on the 7th chromosome) fits the SNP, which is significant for the 'Circularity' trait. The 'Circularity' trait contributes the first bicomponent (-0.67992). The other four significant SNPs are unique and not associated with any single trait. All of these SNPs are a part of appropriate QTLs ('supported' with other SNPs with lower (-lg p)-values) (Fig. <ns0:ref type='figure'>S10</ns0:ref>). Genes around the SNPs may be responsible for the general starch carbohydrate polymer arrangement in the starch granule.</ns0:p><ns0:p>No significant SNPs associated with the second and third bicomponents were revealed.</ns0:p></ns0:div> <ns0:div><ns0:head>SNPs and Related Genes/Proteins</ns0:head><ns0:p>In total, 53 SNPs were found to be significant in the association study of the potato genotypes and starch morphology/yield-related traits. (-lg p)-Values of one SNP were associated with 'preparative yield of starch', five SNPs -with 'aspect ratio' and 'roundness' traits, and one SNP -with the 1 st bicomponent exceeded strict Bonferroni criteria. Nevertheless, some associated SNPs overcame 0.1 and 0.05 FDR (false discovery rate) and suggestive levels. Most of the SNPs are located on the 1 st and 2 nd chromosomes, the least on chromosomes 4, 5, 6, 7, 9, 11, and 12.</ns0:p><ns0:p>Thirty-seven of 53 SNPs are located in protein coding regions. Not one of the SNPs studied in this paper coincides with the ones associated with covalently bound phosphorus content in starch <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> (Table <ns0:ref type='table'>S4</ns0:ref>). </ns0:p></ns0:div> <ns0:div><ns0:head>Preparative yield of starch</ns0:head></ns0:div> <ns0:div><ns0:head>Aspect ratio and Roundness</ns0:head><ns0:p>These traits are well-correlated with each other and describe the shape of starch granules in a similar way; thus, the traits are associated with the same SNPs. There is a locus on chromosome 1 containing eight SNPs in the same DNA region. Some of these proteins are related to circadian rhythm-regulating proteins. The circadian clock regulates numerous plant developmental and metabolic processes. In crop species, the circadian clock contributes significantly to plant performance and productivity and to the adaptation and geographical range over which crops can be grown. Other SNPs are related to the phosphorylation of important biochemical intermediates and plastid organization. On chromosome 2, a total of 19 significant SNPs were identified. The SNPs are narrowly situated, forming a single DNA region, a potential trait-related QTL. Most of the SNPs are located in noncoding regions or related to proteins with unknown functions. Two SNPs are related to WPP domain-associated protein-encoding genomic regions, one to plastid high chlorophyll fluorescence 136, another to pentatricopeptide repeat-containing protein, and one SNP is related to DNA binding protein. Two SNPs on chromosomes 7 and 11 were found in the genes encoding proteins with unknown functions.</ns0:p></ns0:div> <ns0:div><ns0:head>Circularity</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Only three SNPs associated with the circularity trait were identified. The first SNP is related to a noncoding region of chromosome 12, the second is located on chromosome 7 in the minichromosome maintenance 5 protein coding sequence, and the third is located on chromosome 11 in the flavonoid 3',5'-hydroxylase-encoding region.</ns0:p></ns0:div> <ns0:div><ns0:head>First 'phenotype' bicomponent</ns0:head><ns0:p>Six SNPs were found to be associated with the first bicomponent from 2B-PLS, which is a complex component comprising all of the traits studied. Thus, one SNP was the same for the 1st bicomponent and Preparative yield (chromosome 7), while another SNP was the same for the first bicomponent and Circularity (chromosome 5). Among the four other unique 1st bicomponent associated SNPs, two are related to the GWD gene, one of the key starch biosynthesis genes <ns0:ref type='bibr' target='#b10'>[10]</ns0:ref>, which is responsible for the phosphorylation-dephosphorylation of glucans (chromosome 9).</ns0:p><ns0:p>The other two SNPs are associated with cytochrome B561 family protein (chromosome 5) and Pto-interacting protein 1 (chromosome 12). In summary, chromosomes 1 and 2 contain important regions responsible for the roundness and aspect ratio of tuber starch granules. There are indications that granule shape may depend on circadian rhythm-related metabolic processes and starch phosphorylation processes. The GWD gene, which is known to regulate phosphorylation and dephosphorylation participates in the regulation of a whole number of morphological traits, rather than a single certain one.</ns0:p><ns0:p>Nevertheless, some other mechanisms and proteins located on chromosome 2 influence the granule formation process. The preparative yield of tuber starch is probably a polygenic trait, regulated by a number of proteins that are encoded by sequences in various parts and chromosomes of the potato genome.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>A genome-wide association study using a 22K SNP potato array enabled 53 novel SNPs to be identified on chromosomes 1, 2, 4, 5, 6, 7, 9, 11, and 12; these SNPs are associated with tuber starch preparative yield and with starch granule morphology (aspect ratio, roundness, circularity, and the 1 st bicomponent). Some of the SNPs observed in this study are located in noncoding regions. The coding regions are associated with membrane and plastid proteins, DNA transcription and binding regulators, low-molecular-weight metabolite synthesis as well as flavonoid biosynthesis. The information on significant regions can be used to convert SNPs to PCR-markers, convenient for screening breeding material in programs aimed on development of potato varieties with desired starch properties. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Three out of four SNPs found on chromosome 4 are located in noncoding sequences and relate to two different DNA regions. There are some other related SNPs with considerably lower significance in these regions, indicating possible QTLs. The fourth SNP is situated in the different region of the gene, coding for a DNA transcription regulator consisting of 95 amino acid residues.All three SNPs found on chromosome 5 are related to the sequences coding different proteins.One SNP corresponds to coiled coil protein with unclear function; two others are related to lowmolecular-weight organic metabolite conversion, specifically to the enzymes aldehyde dehydrogenase and phenyl alanine lyase, which are responsible for alcohol-aldehyde equilibria in the cell and for flavonoid synthesis. Taking into account SNPs that are less significant but still associated with starch yield, we may speak about two trait-associated DNA regions that arePeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)Manuscript to be reviewed potential QTLs. The significant SNP on chromosome 6 seems also be a part of a QTL and relates to plastid transcriptionally active protein 16 (PTAC16), which is suspected to be involved in the regulation of plastid gene expression. The significant SNP on chromosome 7 encodes the protein translation factor SUI1. The significant SNP on chromosome 9 encodes a protein with unknown function. A group of five SNPs located within 2855044 to 3572445 bp on chromosome 11 belongs to the same QTL, and the SNPs are included in the coding regions of several proteins: DEGP10, methylenetetrahydrofolate reductase, acetolactate synthase, and two proteins of unknown function. In general, most of the proteins associated with potato starch preparative yield variations are involved in plastid activity and low-molecular-weight metabolite biosynthesis.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Significant SNPs associated with preparative yield of starch and morphological traits.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:05:48564:2:2:NEW 26 Sep 2020)</ns0:note></ns0:figure> </ns0:body> "

"Dear Editor, Authors thank all the Reviewers and the Editor for their notes and suggestions. We carefully worked through each of them and included corrections into the main text as much as pdf file allowed (hope, technical editor will them properly). Below we gave our answers to the Editor’s comments (blue). Thank you very much, Vadim K. Khlestkin, on behalf of the authors. Editor comments (Ruslan Kalendar) MINOR REVISIONS Authors need to provide the following information on the material under study: - The genotype information for each line needs to be made available as a supplemental data set. Dear Editor, thank you so much for the corrections. We carefully examined PeerJ practice related to GWAS raw data publishing. We discovered that in recent papers, the data either are not disclosed (https://peerj.com/articles/8572/), either are stored in authors’ own repositories (https://peerj.com/articles/7259/). In our case, we plan to continue using genotype information for looking for genotype associations with commercially viable traits of potato starch. Publishing of the genotyping data (payed by authors) may result in competitive situation with other researchers working in the similar area. Nevertheless, we plan to publish the genotype information after we complete the studies. For now, the genotype information is available from authors directly upon request, which is now mentioned in the main text after line 139. - There is little information about the experimental design. - How were the plants grown? field or greenhouse? what was the layout? how old when harvested? what time of year? what was the replication? Experimental design details are added in section Plant material, after line 133. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>We present a robustness analysis of an inter-cities mobility complex network, motivated by the challenge of the COVID-19 pandemic and the seek for proper containment strategies.</ns0:p><ns0:p>Brazilian data from 2016 are used to build a network with more than five thousand cities and twenty-seven states with the edges representing the weekly flow of people between cities via terrestrial transports. Regions are systematically isolated (removed from the network) either at random (failures) or guided by specific strategies (targeted attacks), and the impacts are assessed with three metrics: the number of components, the size of the giant component, and the total remaining flow. We propose strategies to identify which regions should be isolated first, their impact on people mobility, and how they compare to the so-called reactive strategy, which consists of isolating regions ordered by the date the first case of COVID-19 appeared. The municipal and state initiatives are here abstracted as nodes' failures since there is no well defined country-level organization, and the federal actions are the targeted attacks. The results reveal that individual municipalities' initiatives do not cause a high impact on mobility restraint since they tend to be disconnected from the country's global interventions. Oppositely, the coordinated isolation of specific cities is crucial to detach entire network areas and thus prevent spreading.</ns0:p><ns0:p>Besides, the targeted attacks pose better results than the reactive strategy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Since early 2020, the SARS-CoV-2 quickly spread to the entire world and became a pandemic in a short time. As of September 2nd, 2020, the virus has reached more than 180 countries, with more than 26,065,382 confirmed cases of COVID-19, the disease caused by the virus, and about 863,826 deaths, globally <ns0:ref type='bibr'>(JHU Database, 2020)</ns0:ref>. In Brazil, there are more than 4,003,441 confirmed cases and nearly 123,926 deaths, with the first documented case located in the city of S&#227;o Paulo on February 25th, 2020 <ns0:ref type='bibr' target='#b6'>(Cota, 2020)</ns0:ref>.</ns0:p><ns0:p>The design of containment strategies promoted in federal, state and municipal actions became an enormous challenge to prevent community transmission. In this context, the analysis of the inter-cities terrestrial mobility network is useful for decision making since the coordinated isolation of specific cities and states is crucial to spreading prevention.</ns0:p><ns0:p>The complex networks <ns0:ref type='bibr' target='#b8'>(Estrada, 2012)</ns0:ref> emerge as a natural mechanism to treat mobility data, taking areas as nodes and movements between origins and destinations as edges <ns0:ref type='bibr' target='#b1'>(Barbosa et al., 2018)</ns0:ref>. A complex network is a graph (set of nodes and relations between them) that represents a complex system.</ns0:p><ns0:p>A mobility network is a set of areas connected by the flow of people, and, unlike physical networks (such PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed as transportation infrastructures), they are social networks <ns0:ref type='bibr' target='#b22'>(Santos et al., 2019a)</ns0:ref>.</ns0:p><ns0:p>The structure of the underlying network of a system reveals its ability to survive to random failures and coordinated attacks. Knowing which and how many nodes can be removed until the network completely fragments into small pieces is of great importance <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>. In this paper, we present a robustness analysis <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b4'>Callaway et al., 2000)</ns0:ref> on Brazilian mobility networks, motivated by the challenge of the COVID-19 pandemic and the seek for proper containment strategies. We propose strategies to identify which regions should be isolated first, their impact on people mobility, and how they compare to the so-called reactive strategy, which consists of isolating regions ordered by the date the first case of COVID-19 appeared.</ns0:p><ns0:p>We effectively damage the network structure through different strategies by systematically removing the cities (or states) that have more impact on mobility. Within the context of robustness analysis, a failure is the random removal of a node, and an attack is the removal of a node based on a specific strategy. The local initiatives are here modeled as random failures because there is no central/global orchestration. It is possible that some cities (states) start to care about an epidemic before the others and/or before the country itself, either because their mayors (governors) have more political influence than the others, or due to local popular pressure. In both cases, the outcome for the city (state) is likely to diverge from the announced measures for the country at the federal level. Contrarily, the cooperation between cities, states, and the federal government characterize the targeted attacks, so that a federal level scheme guides the isolation process.</ns0:p><ns0:p>The present study employs the IBGE data from 2016 (IBGE -Instituto Brasileiro de Geografia e Estat&#237;stica, 2017), which contains the flow of people between cities, considering only terrestrial vehicles from companies that sell tickets to passengers. Another data source, commonly used, is the pendular travels <ns0:ref type='bibr' target='#b3'>(Brasil, 2020)</ns0:ref> of people moving from home to work/study. Yet, the former is more recent and captures the flows of people between all pairs of Brazilian cities in a more general scenario. The data we use concerns the flow of people and does not cover the transport of supplies. The isolation of a region consists of closing the borders to the flow of people to/from other regions, as performed in Wuhan, China. <ns0:ref type='bibr' target='#b18'>(Li et al., 2020a)</ns0:ref>.</ns0:p><ns0:p>Our contributions are the robustness analysis of the Brazilian inter-cities mobility network, where random failures abstract local actions from cities or states, and the targeted attacks are the federal's. We assess the impacts of nodes' removal with three metrics: the size of the giant component, the number of components, and the total remaining flow within the network. Strategies based on centrality measures such as degree, betweenness, and topological vulnerability guide the targeted attacks. Lastly, we compare both the random failures and the targeted attacks with the reactive strategy.</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head><ns0:p>The complex network approach is often applied to treat mobility data, taking areas as nodes and movements between origins and destinations as edges. Formally, a network is defined as an undirected graph G(V, E), consisting of the set V of vertices (or nodes) and set E of edges, with the total number of nodes N = |V | and the total number of edges |E|. The edges' weights are represented as the matrix W = {w i j }, for i, j = 1, &#8226; &#8226; &#8226; , N, so that w i j is the weight between edges i and j. The mean value and standard deviation of this matrix are w and &#963; , respectively.</ns0:p><ns0:p>The network flows (weights) (IBGE -Instituto Brasileiro de Geografia e Estat&#237;stica, 2017) are here aggregated within the round trip, which means that the number of travels from city A to city B is the same as from B to A. We produce three types of undirected networks with a different number N of nodes to capture actions in distinct scales (country and state):</ns0:p><ns0:p>1. N = 5420 -Brazil (BR): nodes are cities and edges are the flow of direct travels between them. The dataset encompasses almost all Brazilian cities.</ns0:p><ns0:p>2. N = 620 -S&#227;o Paulo state (SP): a subset of the previous network, containing only cities within the S&#227;o Paulo state, the first Brazilian state with a confirmed case.</ns0:p><ns0:p>3. N = 27 -Brazilian states (BS): in contrast with the others, in this network, each state is a node, and the edges are the accumulated flows between them.</ns0:p><ns0:p>Several networks are analyzed from the three models (BR, SP, and BS), with flow thresholds employed in three levels: i) original data with all recorded flow, ii) only edges of at least an average flow, and</ns0:p></ns0:div> <ns0:div><ns0:head>2/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed iii) a more restricted topology with the higher flows. The chosen thresholds are &#951; 0 = 0, &#951; 1 = w and &#951; 2 = w + &#963; . Edges with flows below these values are discarded. We thus end up with nine networks in total, as described in Table <ns0:ref type='table'>1</ns0:ref>, where N is the size of the network, and |E| is the number of edges/links.</ns0:p><ns0:p>The motivation behind the threshold levels is the fact that most centrality measures we investigated do not account for the flows and thus consider all edges with the same importance. Besides, neglecting some small flow connections may help to approximate the network measures to the real spreading dynamics of SARS-CoV-2 <ns0:ref type='bibr' target='#b9'>(Freitas et al., 2020)</ns0:ref>.</ns0:p><ns0:p>Table <ns0:ref type='table'>1</ns0:ref>. Networks' statistics. The Brazilian (BR), S&#227;o Paulo state (SP), and Brazilian states (BS) networks, with three flow thresholds: &#951; 0 = 0, &#951; 1 = w and &#951; 2 = w + &#963; , where w is the average flow and &#963; is the standard deviation. </ns0:p></ns0:div> <ns0:div><ns0:head>Network</ns0:head></ns0:div> <ns0:div><ns0:head>Measures of complex networks</ns0:head><ns0:p>The degree k is the number of cities (or sates) that a city (state) is connected to, showing the number of possible destinations for the SARS-CoV-2. The betweenness centrality b considers the entire network to depict the topological importance of a city in the routes more likely to be used. The vulnerability V accounts for the impact in the network efficiency when a particular city (state) is isolated. Lastly, the strength s captures the total number of people that travel to (or come from) such places in a week. From a probability perspective, the cities that receive more flow of people are more vulnerable to SARS-CoV-2.</ns0:p><ns0:p>The topological degree k of a node presents its connectivity: it is the number of edges it has to other nodes. The networks are undirected with no distinction between incoming and outgoing edges. On the other hand, the betweenness centrality captures the importance of a node. Between any pairs of nodes l and m of a connected network, there is at least one shortest path, and the betweenness b i is the rate of such paths that pass through i ( <ns0:ref type='bibr' target='#b2'>Barth&#233;lemy, 2004)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_0'>b i = &#8721; l =m =i g lm (i) g lm ,<ns0:label>(1)</ns0:label></ns0:formula><ns0:p>in which l, m, i &#8712; V , g lm is the total number of shortest paths (or geodesic paths) between l and m, and g lm (i) are those that pass through i.</ns0:p><ns0:p>The efficiency e i j in the communication between a pair of nodes i and j can be defined as the inverse of the shortest path length between them, and the network efficiency E <ns0:ref type='bibr' target='#b11'>(Goldshtein et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b27'>Wang et al., 2017)</ns0:ref> is</ns0:p><ns0:formula xml:id='formula_1'>E = &#8721; i = j e i j N(N &#8722; 1) ,<ns0:label>(2)</ns0:label></ns0:formula><ns0:p>the average of all efficiencies, with i, j &#8712; V . The vulnerability index V i <ns0:ref type='bibr' target='#b23'>(Santos et al., 2019b)</ns0:ref>, quantifies how vulnerable to the removal of node i a network is:</ns0:p><ns0:formula xml:id='formula_2'>V i = E &#8722; E * i E ,<ns0:label>(3)</ns0:label></ns0:formula></ns0:div> <ns0:div><ns0:head>3/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed in which E * i is the average network efficiency after the removal of node i. In brief, the flow of information is considered more efficient in networks with small shortest path lengths.</ns0:p><ns0:p>The strength s i of a node is the accumulated flow from incident edges:</ns0:p><ns0:formula xml:id='formula_3'>s i = N &#8721; j=1 w i j .<ns0:label>(4)</ns0:label></ns0:formula></ns0:div> <ns0:div><ns0:head>Robustness</ns0:head><ns0:p>The robustness of a network is its capacity to keep connected even after the removal of nodes and/or edges <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>. A breakdown (for example, an energy drop) of some computers in computer networks, or a car accident on an important road, are usually unpredictable events that depend on several internal and/or external causes, thus characterizing a system failure. Conversely, an intentionally removed node to disrupt the network structure typifies an attack <ns0:ref type='bibr' target='#b25'>(Schneider et al., 2011)</ns0:ref>. We propose strategies to identify the municipalities (states) that play a key role in mobility. Our motivation is the fact that real networks are robust to random failures but are fragile to attacks <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b4'>Callaway et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b5'>Cohen et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b16'>Iyer et al., 2013)</ns0:ref>. The main question is to figure out how many and which nodes must be removed until the network collapses. Understanding which cities are important for mobility to know exactly which node to isolate in a disease outbreak is of major interest.</ns0:p><ns0:p>We keep track of three measures to quantify the network response to both random failures and targeted attacks when a rate f of nodes are removed: the number of nodes in the giant component P &#8734; ( f ), the total number of components C( f ), and the total remaining flow W ( f ) = &#8721; i j w i j . Within this framework, whether a single node or a small group is isolated from the rest, it is considered a component itself.</ns0:p><ns0:p>There are different ways to choose which node to remove. Random failures are the trivial case for which nodes are randomly selected. However, targeted attacks demand some strategy like always removing the nodes with higher degrees. We propose four strategies: deleting nodes with a higher degree (max k), betweenness (max b), vulnerability (max V ), and strength (max s). Attacks oriented by higher degrees are effective to reduce the size of the giant component and produce better results than non-local measures in most cases <ns0:ref type='bibr' target='#b16'>(Iyer et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The BR network (N = 5420) has a degree distribution that follows a power-law with a coefficient &#947; = 2.57, which characterizes a scale-free topology. This means that, under random failures, the critical threshold f c = 0.9911, for f c = 1 &#8722; (1/(&#954; &#8722; 1)) with &#954; = k 2 / k , gives the exact fraction of random node removals that break the network. This structure is strongly robust to failures, i.e., almost all nodes must be removed before the giant component takes apart <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>. On the other hand, such networks are vulnerable to attacks, especially when they target higher degree nodes (hubs).</ns0:p><ns0:p>Robustness is measured by <ns0:ref type='bibr' target='#b16'>(Iyer et al., 2013</ns0:ref>)</ns0:p><ns0:formula xml:id='formula_4'>R = 1 N N &#8721; i=1 &#915;(i/N) &#915;(0) ,<ns0:label>(5)</ns0:label></ns0:formula><ns0:p>for R &#8712; (0, 1/2) and &#915;( f ) is the network response function after removing a fraction f of its nodes. The higher the R, the more robust the network is according to the function &#915;, which could be either P &#8734; or W .</ns0:p><ns0:p>Note that the normalization factor 1/N allows the comparison of networks of different sizes. For P &#8734; , the star-like topology reaches the minimum value R = 1/N, and the complete graph achieves the maximum</ns0:p><ns0:formula xml:id='formula_5'>R = 1 2 (1 &#8722; 1/N).</ns0:formula><ns0:p>The R measure cannot be computed from C( f ), since this function does not always decrease like in P &#8734; and W . The number of components and their number of participants may oscillate instead. Two components with dozens of nodes each or two components with a single node are evaluated alike with C( f ), thus not giving a direct notion of connectivity or flow.</ns0:p><ns0:p>The simulations of the next section were carried on an Intel(R) Core(TM) i5-4210U CPU 1.70GHz</ns0:p><ns0:p>&#215; 4, with 8 GB Ram, using Python programming language. The respective data and source code are available at https://github.com/vanderfreitas/network robustness.</ns0:p></ns0:div> <ns0:div><ns0:head>4/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>The measures related to the BR, BS, and SP networks for each flow threshold are summarized in Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>, and Fig. <ns0:ref type='figure'>1</ns0:ref> presents a sketch of the national network with two different flow thresholds. Material).</ns0:p><ns0:p>The reactive strategy (COVID-19 curve of Fig. <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>) reaches an intermediate performance, with both R values and C curves amid random failures and targeted attacks. Despite not better, the results for the reactive strategy are comparable to the targeted attacks when the remaining flow W ( f ) is at stake (bottom of Fig. <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>). However, the targeted attacks are more effective when it comes to P &#8734; ( f ).</ns0:p><ns0:p>We normalized C( f ) according to the initial number of components (before removing nodes). There is about 25 times the number of components observed in C(0), when half of the nodes ( f = 0.5) are removed under the guidance of the betweenness centrality in the BR network with &#951; 0 (blue curve of Fig.</ns0:p></ns0:div> <ns0:div><ns0:head>5/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed . The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table <ns0:ref type='table'>1</ns0:ref>. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:p></ns0:div> <ns0:div><ns0:head>2A</ns0:head><ns0:p>). The C(0) does not equal 1 (one) necessarily since some cities are either isolated or compose small components that do not have terrestrial flows of people to the rest.</ns0:p><ns0:p>The number of components C( f ) increases almost linearly under random failures for BR with &#951; 0 (black curve of Fig. <ns0:ref type='figure' target='#fig_0'>2A</ns0:ref>) and only decreases in the end, with f &#8776; 0.9. The giant component for the same network is initially well connected and does not break easily, then the number of components remains closely the same. On the other hand, C( f ) only decreases for &#951; 1 and &#951; 2 , due to the lower number of links.</ns0:p><ns0:p>This results in a maximum number of components that is smaller than in &#951; 0 , since the initial number of clusters is higher in the former cases. The same is observed in Fig. <ns0:ref type='figure' target='#fig_1'>3</ns0:ref> and 4 for SP and BS network, respectively.</ns0:p><ns0:p>Attack-wise, the degree is more well-succeeded in decreasing the size of the giant component, and strength performs better regarding the total remaining flow in both BR and SP networks. The degree (yellow curves) indeed decreases the size of the components, because it targets the most connected nodes.</ns0:p><ns0:p>The betweenness, on the other hand, generates a larger number of components (blue curves), since it detects the shortest paths between groups of well-connected nodes, which coincide with their bridges.</ns0:p><ns0:p>Some methods for community detection -like the Girvan-Newman -systematically remove the edges with higher betweenness <ns0:ref type='bibr' target='#b7'>(Easley and Kleinberg, 2010)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>6/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='figure'>E,H</ns0:ref>) &#951; 1 ; and C,F,I) &#951; 2 , as in Table <ns0:ref type='table'>1</ns0:ref>. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:p><ns0:p>Although some cities have not reported COVID-19 cases until September 2nd, 2020, we computed the R related to the reactive strategy, since the number of remaining cities is negligible: 61 for BR (1.1% of its nodes). Note that Eq. ( <ns0:ref type='formula' target='#formula_4'>5</ns0:ref>) takes into account the entire curve of &#915;( f ) for f &#8712; [1/N, 1]. We verified that the remaining nodes of the BR network would impact in fluctuations of a maximum of 10 &#8722;2 in R for P &#8734; ( f ), and 10 &#8722;3 for W ( f ).</ns0:p><ns0:p>The reactive strategy has a low impact on the number of connected cities in the giant component, but has a strong effect in the remaining flow in BR. There is an important feedback mechanism in this case: the emergence of COVID-19 cases is possibly associated with both imported cases and community transmission between cities in the country. Thus, the flow of people is on both sides of this relation.</ns0:p><ns0:p>The S&#227;o Paulo mobility network (SP) produces similar results as the BR, but the topological vulnerability starts to play a more significant role than in BR, being the second-best under &#951; 0 and P &#8734; .</ns0:p><ns0:p>The differences between failures and attacks are only noticeable for higher thresholds in the network formed by the Brazilian states (BS) -see Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>. Removing nodes with the attacking strategies does not cause much more impact than picking by chance under &#951; 0 and P &#8734; . The results differ for other thresholds when the shortest paths between nodes increase.</ns0:p></ns0:div> <ns0:div><ns0:head>7/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='figure'>E,H</ns0:ref>) &#951; 1 ; and C,F,I) &#951; 2 , as in Table <ns0:ref type='table'>1</ns0:ref>. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:p><ns0:p>Notice that some plateaus represent regions where the removal of nodes does not impact on robustness.</ns0:p><ns0:p>An example is the interval f &#8712; [0.2, 0.75] of Fig. <ns0:ref type='figure' target='#fig_2'>4F</ns0:ref>, where attacking nodes under the betweenness guidance does not cause any harm, because the referred nodes do not belong to the giant component.</ns0:p><ns0:p>Interestingly, the attacks and failures perform similarly, and sometimes the failures are even more effective (Fig. <ns0:ref type='figure' target='#fig_2'>4E and G</ns0:ref>). The strategies follow the same order of efficacy for W under all thresholds: strength, degree, vulnerability, and betweenness, with strength being the best and betweenness the worst. The reactive strategy is even better than betweenness for &#951; 0 .</ns0:p><ns0:p>Regarding P &#8734; , there is an increasing importance of the vulnerability measure from BR to BS. Besides, while the degree is the best measure to guide the attacks for the National and S&#227;o Paulo networks, it is not for the BS, where vulnerability and betweenness have more importance. Similarly, in BR and SP, for W , the strength is the leading measure for attacks, and vulnerability is the worst. Conversely, although strength is also the best for BS, betweenness is the worst.</ns0:p></ns0:div> <ns0:div><ns0:head>8/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>As expected <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>, random failures do not break the network until almost all nodes are removed, due to its scale-free structure, and all targeted attacks dismantle the networks for small f , except for the reactive strategy. The higher the threshold, the fewer nodes must be removed to break the network structure since the giant component is initially smaller than the observed for &#951; 0 . The R measure shows that the more effective attack strategy for P &#8734; is guided by degree, and by strength for W for all thresholds. The smaller the R, the more destructive the corresponding attack strategy is. The maximum number of components arises in targeted attacks guided by the betweenness centrality for BR and SP networks. When it comes to the BS network, the same happens for &#951; 0 , but other measures also hit the maximum for other thresholds.</ns0:p><ns0:p>The reactive strategy produces an impact similar to that of targeted attacks on decreasing the flow of people, although slightly worse. The number of remaining connected cities is always higher. Therefore, despite reacting to the disease spreading is a valid action, targeted attacks provide better results in terms of the size of the giant component and remaining flow in the system.</ns0:p><ns0:p>The cities from the state of S&#227;o Paulo that have higher values are also cited in recent studies <ns0:ref type='bibr' target='#b9'>(Freitas et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b12'>Guimar&#227;es Jr et al., 2020)</ns0:ref> on the most vulnerable cities to COVID-19 due to the intensive traffic of people.</ns0:p><ns0:p>Quickly breaking the transmission network is vital to contain any highly contagious disease, which demands the rapid implementation of control measures such as travel restrictions. Cities that preemptively adhered to the measures reported fewer cases than the others, and the virus reached them later <ns0:ref type='bibr' target='#b26'>(Tian et al., 2020)</ns0:ref>. The city of Wuhan was the main focus in China, and the complete isolation of the area was essential to mitigate the virus spreading <ns0:ref type='bibr' target='#b18'>(Li et al., 2020a)</ns0:ref>. On the other hand, the rest of the world received the SARS-CoV-2 concurrently at different places and had to divide efforts to restrain it.</ns0:p><ns0:p>The targeted attacks are especially relevant in areas where people are not sufficiently tested for COVID-19 since the reactive strategy strongly depends on effective epidemic surveillance. Li et al.</ns0:p><ns0:p>(2020b) estimate that 86% of infections were undocumented in China before the travel restrictions of January 23rd, 2020, and the undocumented infections were the source of 79% of the documented cases.</ns0:p><ns0:p>Underreporting is also present in Brazil, as <ns0:ref type='bibr' target='#b14'>Hallal et al. (2020)</ns0:ref> pointed out in a nationwide seroprevalence survey they conducted.</ns0:p><ns0:p>The US response to COVID-19 is mostly guided by Governors and Mayors primarily because of their political system. Korea and Taiwan implemented a centralized national strategy instead <ns0:ref type='bibr' target='#b13'>(Haffajee and Mello, 2020)</ns0:ref>. Canada has the Health Portfolio Operations Centre (HPOC), which concentrates the operations at different levels of government. The Organisation for Economic Co-operation and Development OECD (2020) argues that coordinated response across regions and states minimize coordination failures since they avoid the 'pass the buck' behavior.</ns0:p><ns0:p>Federal initiatives towards SARS-CoV-2 containment are more effective in breaking the transmission network than leaving the cities (or states) on their own. We have shown that random failures usually take longer to dismantle the networks than choosing the nodes with some criteria. Besides, network measures provide a good approximation for the emergence of COVID-19 <ns0:ref type='bibr' target='#b9'>(Freitas et al., 2020)</ns0:ref>. With a coordinated operation scheme, the different organizational levels of government can implement the isolation or more rigid physical distancing policies in specific cities/states that are key to restrain the virus spreading.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Based on the robustness analysis, the reactive strategy is not effective in reducing the size of the giant component nor breaking the mobility network into disconnected groups when compared to the targeted attacks. Moreover, the federal actions have a substantial impact on the network, while the local ones usually do not break it before almost all cities are isolated. Choosing the cities with higher degrees for the targeted attacks is the best option in most cases, considering the size of the largest component, especially for the two largest networks (N = 5420 and N = 620 cities). However, there is a transition, showing that the vulnerability index performs nearly the same as the degree for the S&#227;o Paulo State network, and it is the best choice for the network of the Brazilian states (N = 27 nodes) under most threshold levels. The total flow of the network is affected similarly by both the targeted attacks and the reactive strategy, but the former is more well succeeded when guided by the strength measure. Lastly, the removal of regions ordered by their betweenness centrality generates a higher number of disconnected islands in the mobility</ns0:p></ns0:div> <ns0:div><ns0:head>9/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed network, which ensures the containment of the disease within small isolated groups.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2. Robustness analysis for the Brazilian mobility network (BR). Attack strategies: max s (strength), max k (degree), max V (vulnerability), max b (betweenness), and the reactive (COVID-19 curve). The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table1. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure3. Robustness analysis for the S&#227;o Paulo mobility network (SP). Attack strategies: max s (strength), max k (degree), max V (vulnerability), max b (betweenness), and the reactive (COVID-19 curve). The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table1. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Robustness analysis for the Brazilian states' mobility network (BS). Attack strategies: max s (strength), max k (degree), max V (vulnerability), max b (betweenness), and the reactive (COVID-19 curve). The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table1. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Networks' measures. Average degree k , average betweenness b , average strength s and average vulnerability V for the Brazilian (BR), S&#227;o Paulo state (SP), and Brazilian states (BS) networks under flow thresholds &#951; 0 , &#951; 1 and &#951; 2 . The larger nodes are state capitals. Nodes with smaller degrees are blue, with higher degrees are red and intermediate values are dark red. The figure for &#951; 0 is not properly visible due to its 65254 edges.All measures (degree, betweenness, vulnerability, and strength) for the BR network under &#951; 0 exhibit the cities of S&#227;o Paulo and Belo Horizonte within the top-five higher values and most present Campinas and Bras&#237;lia. Concerning the SP network, the measures rank the cities of S&#227;o Paulo, Campinas, S&#227;o Jos&#233; do Rio Preto, and Ribeir&#227;o Preto within the top-five values as well. Differently, the BS network does not display a clear pattern for the degrees, but the states of S&#227;o Paulo and Minas Gerais come out in the first positions for betweenness, vulnerability, and strength (see the corresponding Tablesin the Supplemental</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Network</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>&#951; 0</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>&#951; 1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>&#951; 2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell></ns0:row><ns0:row><ns0:cell>k</ns0:cell><ns0:cell>24.08</ns0:cell><ns0:cell>15.47</ns0:cell><ns0:cell>17.56</ns0:cell><ns0:cell>5.72</ns0:cell><ns0:cell>4.21</ns0:cell><ns0:cell>4.0</ns0:cell><ns0:cell>1.56</ns0:cell><ns0:cell>1.22</ns0:cell><ns0:cell>1.63</ns0:cell></ns0:row><ns0:row><ns0:cell>b</ns0:cell><ns0:cell cols='2'>5574.09 504.24</ns0:cell><ns0:cell>4.56</ns0:cell><ns0:cell cols='3'>4828.08 397.67 13.04</ns0:cell><ns0:cell cols='2'>2177.91 125.7</ns0:cell><ns0:cell>6.41</ns0:cell></ns0:row><ns0:row><ns0:cell>s</ns0:cell><ns0:cell cols='4'>1156.86 1132.35 35677.91 845.23</ns0:cell><ns0:cell cols='3'>813.66 30162.76 504.93</ns0:cell><ns0:cell>473.9</ns0:cell><ns0:cell>21043.46</ns0:cell></ns0:row><ns0:row><ns0:cell>V</ns0:cell><ns0:cell cols='3'>4.18E-4 3.62E-3 7.57E-2</ns0:cell><ns0:cell cols='3'>4.97E-4 4.6E-3 8.11E-2</ns0:cell><ns0:cell cols='3'>6.95E-4 6.53E-3 0.1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>A</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>B</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>Figure 1. Brazilian mobility network (BR) under A) &#951; 1 and B) &#951; 2 .</ns0:note></ns0:figure> <ns0:note place='foot' n='11'>/11 PeerJ reviewing PDF | (2020:06:49958:1:0:NEW 3 Sep 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"September 3, 2020 Vander L. S. Freitas Department of Computing Federal University of Ouro Preto Ouro Preto, MG, Brazil 35400-000 e-mail: [email protected] Dear Editor PeerJ We would like to submit our revised manuscript “Robustness analysis in an inter-cities mobility network: modeling municipal, state and federal initiatives as failures and attacks toward SARS-CoV-2 containment” by Vander L. S. Freitas, Gladston J. P. Moreira and Leonardo B. L. Santos for consideration for publication at PeerJ. The manuscript was carefully revised according to the comments and recommendations presented by the reviewers and editor, and we hope you will find it suitable for publication. Summary of Changes We thank the editor and the reviewers for their attention with this submission, as well as for their insightful comments and suggestions. The current version of our manuscript addresses all the comments. The changes in the manuscript are highlighted with the latexdiff tool and can be summarized as follows: • We have performed a complete revision in the paper structure, improving the language and presentation. The sections are now more precise. • We mention the GitHub repository with data and source code in the text. • We provided supplemental material with the tables of cities ordered according to the computed metrics. • Another metric to assess the impact of nodes’ removals was included: the number of components C(f ). The detailed responses to each point raised by the reviewers are provided below. Response to Editor Comment #1: I apologize for the delay in sending this decision. I was able to secure two good reviews of your manuscript. It’s my opinion that authors must do a much better job in improving the flow of ideas throughout the whole paper. The way to structure paragraphs is not adequate, you very often split sentences about the same subject into two or more paragraphs. You also keep restating the motivation of the study over and over again in several parts of the Methods, this makes the reading a little bit chunky. Try to improve the flow of ideas. Almost all Results section need to be reformulated for clarity. Avoid repeating what is already written in the figure and table legends. Instead, call attention to any specific detail or summarize the pattern shown in them, and cite them at the end of the sentence. The conclusions section actually doesn’t provide conclusions and need to be reformulated entirely. See my comments on that in the PDF attached. English language needs to be carefully revised. Also, pay special attention to R1 suggestion on how to improve figures. Also, you definitely need to clarify the assumptions for the working hypothesis and make them understandable to a broader audience. What do you mean by attack and failure in this context? I second R2 about the necessity to provide raw data used for analysis. Provide more details in the M&M section that allow people to reproduce your analysis. For example, which software was used for analysing the data? Overall, the manuscript lacks citations, for the metrics (and formulae) you used, some concepts, as a support for assumptions, I have made several comments and corrections directly in the attached PDF. Please, remember to refer to it when revising your manuscript. Answer: • We thank you for handling our manuscript in such a careful way. We have performed a complete revision of the paper and improved the text structure as required, including the addition of more references for the metrics, formulae, and concepts. • We addressed all points raised by R1 and improved the Figures’ visualization. • We addressed all points raised by R2 and included the URL of the GitHub repository (https: // github. com/ vanderfreitas/ network_ robustness ) with all the data and source code to allow the reproduction of our work. ??? Response to Reviewer #1 Basic reporting: Comment #1: In relation to the figures, I have two comments: the visualization of Figure 1 may be improved using blue color for the nodes with smaller degrees and light gray for the links; the captions of Figures 2, 3 and 4 should include that the procedure of removed nodes is based on different strategies as described on “Robustness” sub-section (deleting nodes with a higher degree, vulnerability, betweenness and strength). Moreover it is not clear in the caption how the COVID-19 cases (what means “temporal sequence”?) and “failure” curves are constructed; it is not clear the meaning of giant component and remaining flow for those curves. Finally I did not figure out the reason that there is a value of robustness measure for COVID-19 curve in Figure 4 but not in Figures 2 and 3. Answer: • We have improved the visualization of Figure 1: the nodes with smaller degrees are blue, higher degrees are red and intermediate values are dark red. • We re-wrote the caption of all Figures for clarity. • The failure curve is associated with an average behavior for 50 simulations, where we randomly remove the nodes, one-by-one. • The COVID-19 curve is formed from the removal of cities (states) according to the date of its first confirmed case. The first positive case in Brazil occurred in the city of Sao Paulo, followed by Barra Mansa, and Feira de Santana. This is exactly the order of removal: first Sao Paulo, then Barra Mansa, Feira de Santana in third place, and so on. • Eventually, there may exist many disjoint groups of connected nodes in the network. The number of nodes in the larger group is the size of the giant component. Whether all nodes are connected in a single group, the size of the giant component is the total number of nodes. Contrarily, if there are no edges, the number of the giant component is one, since each isolated node is a component itself. • The computation of the robustness measure R demands the availability of the entire curve, i.e., all nodes should be removed either by attack or failure. By May, when we submitted the manuscript, there were still many cities without reported COVID-19 patients both in BR (Figure 2) and SP (Figure 3), making impossible the computation of R. All states, on the other hand, had at least one case and this is why Figure 4 had the R for the COVID-19 curve. • By the time this new version of the manuscript is being submitted, almost all Brazilian cities have confirmed cases, which allowed us to include the referred R in Figures 2, 3, and 4. ??? Comment #2: I think that the authors should present in a supplementary material a more complete list of results obtained from their analysis. For instance, in discussion section, the authors refer the two cities (or states) with the higher values of network measures in relation to the 3 selected networks: Brazilian cities (BR), Brazilian states (BS) and São Paulo state (SP). I suggest the authors include a complete list (in supplementary material) that reveal the order of states (BS) in relation to the higher values of the considered network measures; for BR network, the authors may select the capitals of the states, and for SP network, they may select, for example, 30 cities with highest number of cases. Answer: We have included the requested lists of cities and states in the Supplemental Material. ??? Experimental design: Comment #3: The authors set up a methodological procedure assuming the local initiatives as failures and the federal initiatives as attacks on the networks. I have some questions related to those assumptions as well as some conclusions associated with the obtained results. First of all, it would be reasonable to provide an explanation to those assumptions; for instance, in conclusions section the authors affirm that the attacks are isolation measures determined by the federal government and applied at a municipality level. As far as I know, there is a legal debate in Brazil about the assignment in relation to isolation of the municipalities. Is there a final decision about it? In case of affirmative answer, I suggest to include a reference about it; in case of negative answer, it would be interesting to justify the assumptions. Answer: Attacks are isolation measures determined in agreement with the different levels of administrative organization (federal, state, and municipal) and applied at a municipality level - so, its a local (municipal) action but based on a global (country level) analysis. Currently, the Brazilian states and municipalities have the power to choose which policies, when and how they are applied. Considering that they all act differently, regarding timing and method, we model them as random failures, i.e., nodes’ removals lack specific strategies. ??? Comment #4: In order to highlight the results, it is essential to clarify how the disease information concerning the temporal sequence of COVID-19 cases is related to the procedure of removed nodes that is fundamental for the methodology. Answer: We improved the explanation within the manuscript. The reactive strategy consists of the removal of cities/nodes according to the date they documented the first case of COVID-19 (see Comment #1). This ordering may not impact the mobility like attacking the nodes according to their corresponding centrality measures (targeted attacks - starting with the node with higher centrality value towards the nodes with smaller). Three metrics evaluate the impact of the reactive strategy and the targeted attacks in mobility: the number of components C, the size of the giant component P∞ , and the total remaining flow of people kW k. ??? Validity of the findings: Comment #5: The authors perform a comparative analysis of the network measures, pointing out which one is better or worse for each network. In general they conclude that the strength is a more suitable measure to guide the attacks due to the smaller value of R. As far as I understood, the attack is more effective than the failure for BR and SP networks since the curves fit better for the flow than for the giant component; however, both of them fits well for BS network. Do the authors agree with that summarization? Does that result provide any suggestion about the relevance of the strategies of the state government strategies? Answer: • Attacks usually cause more impact on the networks than both the reactive strategy and random failures. The only exception is when one evaluates the size of the giant component in the BS network under η0 , where all the strategies perform roughly the same. We observed that attacks guided by strength are more effective to diminish the total remaining flow of people; those guided by degree usually result in smaller giant components and; the ones guided by betweenness generate more isolated components. • The results indeed do not provide any clue of what could be done regarding the BS network under η0 (with all flows being considered), especially regarding the size of the giant component. In this case, all states have terrestrial connections to the others and the centrality measures for the nodes are qualitatively similar. The results start to change for increasing η since only the edges with higher flows are kept and some states start to play a key role in connectivity. ??? Comment #6: Undoubtedly the findings are relevant since it assigns a robustness analysis based on mobility network of a large country with a significative number of COVID-19 cases. The analysis may be useful for other transmitted diseases in the country. However, since it is a complex system, other factors get influence in the government decision of isolation; for instance it is very hard to isolate some large cities that are essential for the distribution of food and drugs. In this sense, it is not easy to determine which cities should be isolated. How do the authors make compatible their findings with that actual scenario? Answer: We thank the reviewer for the opportunity to make it clearer. The data we use concerns the flow of people and does not cover the transport of supplies (such as gas, food, and medicines). The isolation of a region consists of closing the borders to the flow of people to/from other regions. ??? Comments for the Author: Comment #7: I think that the manuscript may be published in PeerJ after clarifying the arisen points that I have classified as minor revision, since I suppose the authors had obtained the results suggested for supplementary material. Answer: Thank you for your careful revision. We included the requested data in the supplemental material and carefully answered your concerns. Response to Reviewer #2 Basic reporting: Comment #1: The authors report a timely study of a Brazilian transport network to identify potential strategies assisted by network analyses to contain the spread of Covid-19. The manuscript is clear, structured and easy to read. However, some notations could be more intuitively defined, for instance, why are the number of nodes in the giant component called P∞ (f ), I don’t understand what infinite has to do with it? More importantly, regarding the reproducibility of the results and access to raw data, I did not find any Git repository or an open server where the data and associated scripts to reproduce the results are deposited. I strongly suggest the authors to make their code and data available in an open platform. Answer: • We have created a repository on GitHub (https: // github. com/ vanderfreitas/ network_ robustness ) and added this information to the manuscript. The provided raw data and source code allow for the complete reproduction of our results. • The notation P∞ (f ) comes from the Percolation Theory. Consider the inclusion of edges with probability p in a d-dimensional regular lattice. There is a threshold pc such that when p > pc the so-called percolating cluster emerges with high probability. Considering that the lattice is infinite, the size of the largest cluster becomes infinite as well at pc , allowing it to percolate the whole lattice. In this case, P∞ is the probability of a randomly selected node belong to the largest cluster (percolating cluster). Assessing the robustness of a complex network is the opposite process: instead of adding edges with probability p, we remove either nodes or edges with a probability f . The two processes map into each other by f = 1 − p. Therefore, P∞ (f ) is the size of the giant component after removing a fraction f of the nodes. Although the size of the investigated networks are not infinite, the notation is the same. Source: Chapter 8 of the traditional “Network science” book, available online: http: // networksciencebook. com/ chapter/ 8# percolation-theory Barabási, A.-L. et al. (2016).Network science. Cambridge university press. ??? Experimental design: Comment #2: Overall, I am convinced with the technicality of the analyses design. However, a main concern I have is using the size of giant component to assess the performance of the targeted attack. May be a complimentary metric such as the actual number of connected components (or clusters) for each node removal might indicate how the network is effectively destabilized, even when the giant component is intact. This might mean certain transportation among major cities are very strongly connected, however, many rural cities could be disconnected. So, in addition to the 2 metrics the authors have used to assess the network performance after node perturbations, I would like to see a metric that captures how many components or clusters remain after each perturbation. This might give a picture of how the cities are actually disconnected after node removals. I can see that fraction of remaining flow within the network could be useful but, still I suspect a few large components in the network might dominate the flow. Another strategy to choose node(s) for removal could be to employ minimum cut sets? i.e. the idea is to identify a minimal sets of nodes that disrupt the entire (or major) topology of the network? More details about this could be found here: Klamt and Gilles, 2004, Bioinformatics Answer: • We included the suggested metric (the number of components C(f )) and observed that betweennessguided attacks usually lead to the emergence of more disconnected components than other strategies. This happens because nodes with higher betweenness are usually bridges between components. Most shortest paths between nodes of different “communities” pass through them. • Although it is possible that some large components dominate the flow, the tracking of the total remaining flow already gives the intuition of how many people are still able to move around. • The graphs we use allow only 1:1-relations. The Minimum cut sets (MCS) of Klamt and Gilles (2004) are usually employed to hypergraphs, where a single edge may link k nodes with l nodes. They graphically present this limitation in Figure 2. ??? Validity of the findings: Comment #3: Although I could partly agree with the author’s claim that global (or Federal) actions are more effective compared to local (or municipal) actions to disrupt the transportation network, I am not entirely sure if this can be concluded from the results presented as it is. As I mentioned in the earlier paragraph, I would like to see more complementary metrics to assess the network disruption both locally and globally. And may be an intermediate strategy where both local and global policies must be coordinated could also emerge if more metrics are included in the study. Answer: • We now evaluate the number of components as nodes are removed. • The global strategy is already an agreement between the federal government, states, and municipalities. The actions are orchestrated by the country and executed by the smaller instances. Besides, as our results suggest, an intermediate strategy would be the reactive strategy, since its impact in network robustness is usually amid targeted attacks and random failures in almost all cases. The only exception is when one evaluates the size of the giant component in the BS network under η0 , where all strategies perform roughly the same. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>We present a robustness analysis of an inter-cities mobility complex network, motivated by the challenge of the COVID-19 pandemic and the seek for proper containment strategies.</ns0:p><ns0:p>Brazilian data from 2016 are used to build a network with more than five thousand cities (nodes) and twenty-seven states with the edges representing the weekly flow of people between cities via terrestrial transports. Nodes are systematically isolated (removed from the network) either at random (failures) or guided by specific strategies (targeted attacks), and the impacts are assessed with three metrics: the number of components, the size of the giant component, and the total remaining flow of people. We propose strategies to identify which regions should be isolated first and their impact on people mobility. The results are compared with the so-called reactive strategy, which consists of isolating regions ordered by the date the first case of COVID-19 appeared. We assume that the nodes' failures abstract individual municipal and state initiatives that are independent and possess a certain level of unpredictability. Differently, the targeted attacks are related to centralized strategies led by the federal government in agreement with municipalities and states. Removing a node means completely restricting the mobility of people between the referred city/state and the rest of the network. Results reveal that random failures do not cause a high impact on mobility restraint, but the coordinated isolation of specific cities with targeted attacks is crucial to detach entire network areas and thus prevent spreading.</ns0:p><ns0:p>Moreover, the targeted attacks perform better than the reactive strategy for the three analyzed robustness metrics.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Since early 2020, the SARS-CoV-2 quickly spread to the entire world and became a pandemic in a short time. As of September 2nd, 2020, the virus has reached more than 180 countries, with more than 26,065,382 confirmed cases of COVID-19, the disease caused by the virus, and about 863,826 deaths, globally <ns0:ref type='bibr'>(JHU Database, 2020)</ns0:ref>. In Brazil, there are more than 4,003,441 confirmed cases and nearly 123,926 deaths, with the first documented case located in the city of S&#227;o Paulo on February 25th, 2020 <ns0:ref type='bibr' target='#b6'>(Cota, 2020)</ns0:ref>.</ns0:p><ns0:p>The design of containment strategies promoted in federal, state and municipal actions became an enormous challenge to prevent community transmission. In this context, the analysis of the inter-cities terrestrial mobility network is useful for decision making since the coordinated isolation of specific cities and states is crucial to spreading prevention.</ns0:p><ns0:p>The complex networks <ns0:ref type='bibr' target='#b8'>(Estrada, 2012)</ns0:ref> emerge as a natural mechanism to treat mobility data, taking areas as nodes and movements between origins and destinations as edges <ns0:ref type='bibr' target='#b1'>(Barbosa et al., 2018)</ns0:ref>. A PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed complex network is a graph (set of nodes and relations between them) that represents a complex system.</ns0:p><ns0:p>A mobility network is a set of areas connected by the flow of people, and, unlike physical networks (such as transportation infrastructures), they are social networks <ns0:ref type='bibr' target='#b24'>(Santos et al., 2019a)</ns0:ref>.</ns0:p><ns0:p>The structure of the underlying network of a system reveals its ability to survive to random failures and coordinated attacks. Knowing which and how many nodes can be removed until the network completely fragments into small pieces is of great importance <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>. In this paper, we present a robustness analysis <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b4'>Callaway et al., 2000)</ns0:ref> on Brazilian mobility networks, motivated by the challenge of the COVID-19 pandemic and the seek for proper containment strategies. We propose strategies to identify which regions should be isolated first, their impact on people mobility, and how they compare to the so-called reactive strategy, which consists of isolating regions ordered by the date the first case of COVID-19 appeared.</ns0:p><ns0:p>We effectively damage the network structure through different strategies by systematically removing the cities (or states) that have more impact on mobility. Within the context of robustness analysis, a failure is the random removal of a node, and a targeted attack is the removal of a node based on a specific strategy.</ns0:p><ns0:p>The local initiatives are here modeled as random failures because there is no central/global orchestration.</ns0:p><ns0:p>It is possible that some cities (states) start to care about an epidemic before the others and/or before the country itself, either because their mayors (governors) have more political influence than the others, or due to local popular pressure. In both cases, the outcome for the city (state) is likely to diverge from the announced measures for the country at the federal level. Contrarily, the cooperation between cities, states, and the federal government characterize the targeted attacks, so that a federal level scheme guides the isolation process.</ns0:p><ns0:p>The present study employs the IBGE data from 2016 (IBGE -Instituto Brasileiro de Geografia e Estat&#237;stica, 2017), which contains the flow of people between cities, considering only terrestrial vehicles from companies that sell tickets to passengers. Another data source, commonly used, is the pendular travels <ns0:ref type='bibr' target='#b3'>(Brasil, 2020)</ns0:ref> of people moving from home to work/study. Yet, the former is more recent and captures the flows of people between all pairs of Brazilian cities in a more general scenario. The data we use concerns the flow of people and does not cover the transport of supplies. The isolation of a region consists of closing the borders to the flow of people to/from other regions, as performed in Wuhan, China. <ns0:ref type='bibr' target='#b21'>(Li et al., 2020a)</ns0:ref>.</ns0:p><ns0:p>Our contributions are the robustness analysis of the Brazilian inter-cities mobility network, where random failures abstract local actions from cities or states, and the targeted attacks are the federal's. We assess the impacts of nodes' removal with three metrics: the size of the giant component, the number of components, and the total remaining flow within the network. Strategies based on centrality measures such as degree, betweenness, and topological vulnerability guide the targeted attacks. Lastly, we compare both the random failures and the targeted attacks with the reactive strategy. While the nodes' removals through targeted attacks follow the sorted values of the centrality measures, from the higher value to the lower, in the reactive strategy, the removal starts from the first node that notified COVID-19 in its territory, followed by the next, until the last node in a temporal order.</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head><ns0:p>The complex network approach is often applied to treat mobility data, taking areas as nodes and movements between origins and destinations as edges. Formally, a network is defined as an undirected graph G(V, E), consisting of the set V of vertices (or nodes) and set E of edges, with the total number of nodes N = |V | and the total number of edges |E|. The edges' weights are represented as the matrix W = {w i j }, for i, j = 1, &#8226; &#8226; &#8226; , N, so that w i j is the weight between edges i and j. The mean value and standard deviation of this matrix are w and &#963; , respectively.</ns0:p><ns0:p>The network flows (weights) (IBGE -Instituto Brasileiro de Geografia e Estat&#237;stica, 2017) are here aggregated within the round trip, which means that the number of travels from city A to city B is the same as from B to A. We produce three types of undirected networks with a different number N of nodes to capture actions in distinct scales (country and state):</ns0:p><ns0:p>1. N = 5420 -Brazil (BR): nodes are cities and edges are the flow of direct travels between them. The dataset encompasses almost all Brazilian cities.</ns0:p><ns0:p>2. N = 620 -S&#227;o Paulo state (SP): a subset of the previous network, containing only cities within the S&#227;o Paulo state, the first Brazilian state with a confirmed case.</ns0:p></ns0:div> <ns0:div><ns0:head>2/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed 3. N = 27 -Brazilian states (BS): in contrast with the others, in this network, each state is a node, and the edges are the accumulated flows between them.</ns0:p><ns0:p>Several networks are analyzed from the three models (BR, SP, and BS), with flow thresholds employed in three levels: i) original data with all recorded flow, ii) only edges of at least an average flow, and iii) a more restricted topology with the higher flows. The chosen thresholds are &#951; 0 = 0, &#951; 1 = w and &#951; 2 = w + &#963; . Edges with flows below these values are discarded. We thus end up with nine networks in total, as described in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>, where N is the size of the network, and |E| is the number of edges/links.</ns0:p><ns0:p>The motivation behind the threshold levels is the fact that most centrality measures we investigated do not account for the flows and thus consider all edges with the same importance. Besides, neglecting some small flow connections may help to approximate the network measures to the real spreading dynamics of SARS-CoV-2 <ns0:ref type='bibr' target='#b9'>(Freitas et al., 2020)</ns0:ref>. </ns0:p></ns0:div> <ns0:div><ns0:head>Measures of complex networks</ns0:head><ns0:p>The degree k is the number of cities (or sates) that a city (state) is connected to, showing the number of possible destinations for the SARS-CoV-2. The betweenness centrality b considers the entire network to depict the topological importance of a city in the routes more likely to be used. The vulnerability V accounts for the impact in the network efficiency when a particular city (state) is isolated. Lastly, the strength s captures the total number of people that travel to (or come from) such places in a week. From a probability perspective, the cities that receive more flow of people are more vulnerable to SARS-CoV-2.</ns0:p><ns0:p>The topological degree k of a node presents its connectivity: it is the number of edges it has to other nodes. The networks are undirected with no distinction between incoming and outgoing edges. On the other hand, the betweenness centrality captures the importance of a node. Between any pairs of nodes l and m of a connected network, there is at least one shortest path, and the betweenness b i is the rate of such paths that pass through i ( <ns0:ref type='bibr' target='#b2'>Barth&#233;lemy, 2004</ns0:ref>):</ns0:p><ns0:formula xml:id='formula_0'>b i = &#8721; l =m =i g lm (i) g lm ,<ns0:label>(1)</ns0:label></ns0:formula><ns0:p>in which l, m, i &#8712; V , g lm is the total number of shortest paths (or geodesic paths) between l and m, and g lm (i) are those that pass through i.</ns0:p><ns0:p>The efficiency e i j in the communication between a pair of nodes i and j can be defined as the inverse of the shortest path length between them, and the network efficiency E <ns0:ref type='bibr' target='#b12'>(Goldshtein et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b29'>Wang et al., 2017)</ns0:ref> is</ns0:p><ns0:formula xml:id='formula_1'>E = &#8721; i = j e i j N(N &#8722; 1) ,<ns0:label>(2)</ns0:label></ns0:formula></ns0:div> <ns0:div><ns0:head>3/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed the average of all efficiencies, with i, j &#8712; V . The vulnerability index V i <ns0:ref type='bibr' target='#b25'>(Santos et al., 2019b)</ns0:ref>, quantifies how vulnerable to the removal of node i a network is:</ns0:p><ns0:formula xml:id='formula_2'>V i = E &#8722; E * i E ,<ns0:label>(3)</ns0:label></ns0:formula><ns0:p>in which E * i is the average network efficiency after the removal of node i. In brief, the flow of information is considered more efficient in networks with small shortest path lengths.</ns0:p><ns0:p>The strength s i of a node is the accumulated flow from incident edges:</ns0:p><ns0:formula xml:id='formula_3'>s i = N &#8721; j=1 w i j .<ns0:label>(4)</ns0:label></ns0:formula></ns0:div> <ns0:div><ns0:head>Robustness</ns0:head><ns0:p>The robustness of a network is its capacity to keep connected even after the removal of nodes and/or edges <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>. A breakdown (for example, an energy drop) of some computers in computer networks, or a car accident on an important road, are usually unpredictable events that depend on several internal and/or external causes, thus characterizing a system failure. Conversely, an intentionally removed node to disrupt the network structure typifies an attack <ns0:ref type='bibr' target='#b27'>(Schneider et al., 2011)</ns0:ref>. We propose strategies to identify the municipalities (states) that play a key role in mobility. Our motivation is the fact that real networks are robust to random failures but are fragile to attacks <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b4'>Callaway et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b5'>Cohen et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b18'>Iyer et al., 2013)</ns0:ref>. The main question is to figure out how many and which nodes must be removed until the network collapses. Understanding which cities are important for mobility to know exactly which node to isolate in a disease outbreak is of major interest.</ns0:p><ns0:p>We keep track of three measures to quantify the network response to both random failures and targeted attacks when a rate f of nodes are removed: the number of nodes in the giant component P &#8734; ( f ), the total number of components C( f ), and the total remaining flow W ( f ) = &#8721; i j w i j . Within this framework, whether a single node or a small group is isolated from the rest, it is considered a component itself.</ns0:p><ns0:p>There are different ways to choose which node to remove. Random failures are the trivial case for which nodes are randomly selected. However, targeted attacks demand some strategy like always removing the nodes with higher degrees. We propose four strategies: deleting nodes with a higher degree (max k), betweenness (max b), vulnerability (max V ), and strength (max s). Attacks oriented by higher degrees are effective to reduce the size of the giant component and produce better results than non-local measures in most cases <ns0:ref type='bibr' target='#b18'>(Iyer et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The BR network (N = 5420) has a degree distribution that follows a power-law with a coefficient &#947; = 2.57, which characterizes a scale-free topology. This means that, under random failures, the critical threshold f c = 0.9911, for f c = 1 &#8722; (1/(&#954; &#8722; 1)) with &#954; = k 2 / k , gives the exact fraction of random node removals that break the network. This structure is strongly robust to failures, i.e., almost all nodes must be removed before the giant component takes apart <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>. On the other hand, such networks are vulnerable to attacks, especially when they target higher degree nodes (hubs).</ns0:p><ns0:p>Robustness is measured by <ns0:ref type='bibr' target='#b18'>(Iyer et al., 2013</ns0:ref>)</ns0:p><ns0:formula xml:id='formula_4'>R = 1 N N &#8721; i=1 &#915;(i/N) &#915;(0) ,<ns0:label>(5)</ns0:label></ns0:formula><ns0:p>for R &#8712; (0, 1/2) and &#915;( f ) is the network response function after removing a fraction f of its nodes. The higher the R, the more robust the network is according to the function &#915;, which could be either P &#8734; or W .</ns0:p><ns0:p>Note that the normalization factor 1/N allows the comparison of networks of different sizes. For P &#8734; , the star-like topology reaches the minimum value R = 1/N, and the complete graph achieves the maximum</ns0:p><ns0:formula xml:id='formula_5'>R = 1 2 (1 &#8722; 1/N).</ns0:formula><ns0:p>The R measure cannot be computed from C( f ), since this function does not always decrease like in P &#8734; and W . The number of components and their number of participants may oscillate instead. Two components with dozens of nodes each or two components with a single node are evaluated alike with C( f ), thus not giving a direct notion of connectivity or flow.</ns0:p></ns0:div> <ns0:div><ns0:head>4/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The simulations of the next section were carried on an Intel(R) Core(TM) i5-4210U CPU 1.70GHz</ns0:p><ns0:p>&#215; 4, with 8 GB Ram, using Python programming language. The respective data and source code are available at https://github.com/vanderfreitas/network robustness.</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>The measures related to the BR, BS, and SP networks for each flow threshold are summarized in Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>, and Fig. <ns0:ref type='figure'>1</ns0:ref> presents a sketch of the national network with two different flow thresholds. Material).</ns0:p><ns0:p>The reactive strategy (COVID-19 curve of Fig. <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>) reaches an intermediate performance, with both R values and C curves amid random failures and targeted attacks. Despite not better, the results for the reactive strategy are comparable to the targeted attacks when the remaining flow W ( f ) is at stake (bottom of Fig. <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>). However, the targeted attacks are more effective when it comes to P &#8734; ( f ).</ns0:p></ns0:div> <ns0:div><ns0:head>5/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed . The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:p><ns0:p>We normalized C( f ) according to the initial number of components (before removing nodes). There is about 25 times the number of components observed in C(0), when half of the nodes ( f = 0.5) are removed under the guidance of the betweenness centrality in the BR network with &#951; 0 (blue curve of Fig. <ns0:ref type='figure' target='#fig_0'>2A</ns0:ref>). The C(0) does not equal 1 (one) necessarily since some cities are either isolated or compose small components that do not have terrestrial flows of people to the rest.</ns0:p><ns0:p>The number of components C( f ) increases almost linearly under random failures for BR with &#951; 0 (black curve of Fig. <ns0:ref type='figure' target='#fig_0'>2A</ns0:ref>) and only decreases in the end, with f &#8776; 0.9. The giant component for the same network is initially well connected and does not break easily, then the number of components remains closely the same. On the other hand, C( f ) only decreases for &#951; 1 and &#951; 2 , due to the lower number of links.</ns0:p><ns0:p>This results in a maximum number of components that is smaller than in &#951; 0 , since the initial number of clusters is higher in the former cases. The same is observed in Fig. <ns0:ref type='figure' target='#fig_2'>3 and 4</ns0:ref> for SP and BS network, respectively.</ns0:p><ns0:p>Attack-wise, the degree is more well-succeeded in decreasing the size of the giant component, and strength performs better regarding the total remaining flow in both BR and SP networks. The degree (yellow curves) indeed decreases the size of the components, because it targets the most connected nodes.</ns0:p><ns0:p>The betweenness, on the other hand, generates a larger number of components (blue curves), since it</ns0:p></ns0:div> <ns0:div><ns0:head>6/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='figure'>E,H</ns0:ref>) &#951; 1 ; and C,F,I) &#951; 2 , as in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:p><ns0:p>detects the shortest paths between groups of well-connected nodes, which coincide with their bridges.</ns0:p><ns0:p>Some methods for community detection -like the Girvan-Newman -systematically remove the edges with higher betweenness <ns0:ref type='bibr' target='#b7'>(Easley and Kleinberg, 2010)</ns0:ref>.</ns0:p><ns0:p>Although some cities have not reported COVID-19 cases until September 2nd, 2020, we computed the R related to the reactive strategy, since the number of remaining cities is negligible: 61 for BR (1.1% of its nodes). Note that Eq. ( <ns0:ref type='formula' target='#formula_4'>5</ns0:ref>) takes into account the entire curve of &#915;( f ) for f &#8712; [1/N, 1]. We verified that the remaining nodes of the BR network would impact in fluctuations of a maximum of 10 &#8722;2 in R for P &#8734; ( f ), and 10 &#8722;3 for W ( f ).</ns0:p><ns0:p>The reactive strategy has a low impact on the number of connected cities in the giant component, but has a strong effect in the remaining flow in BR. There is an important feedback mechanism in this case: the emergence of COVID-19 cases is possibly associated with both imported cases and community transmission between cities in the country. Thus, the flow of people is on both sides of this relation.</ns0:p><ns0:p>The S&#227;o Paulo mobility network (SP) produces similar results as the BR, but the topological vulnerability starts to play a more significant role than in BR, being the second-best under &#951; 0 and P &#8734; .</ns0:p><ns0:p>The differences between failures and attacks are only noticeable for higher thresholds in the network</ns0:p></ns0:div> <ns0:div><ns0:head>7/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='figure'>E,H</ns0:ref>) &#951; 1 ; and C,F,I) &#951; 2 , as in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:p><ns0:p>formed by the Brazilian states (BS) -see Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>. Removing nodes with the attacking strategies does not cause much more impact than picking by chance under &#951; 0 and P &#8734; . The results differ for other thresholds when the shortest paths between nodes increase.</ns0:p><ns0:p>Notice that some plateaus represent regions where the removal of nodes does not impact on robustness.</ns0:p><ns0:p>An example is the interval f &#8712; [0.2, 0.75] of Fig. <ns0:ref type='figure' target='#fig_2'>4F</ns0:ref>, where attacking nodes under the betweenness guidance does not cause any harm, because the referred nodes do not belong to the giant component.</ns0:p><ns0:p>Interestingly, the attacks and failures perform similarly, and sometimes the failures are even more effective (Fig. <ns0:ref type='figure' target='#fig_2'>4E and G</ns0:ref>). The strategies follow the same order of efficacy for W under all thresholds: strength, degree, vulnerability, and betweenness, with strength being the best and betweenness the worst. The reactive strategy is even better than betweenness for &#951; 0 .</ns0:p><ns0:p>Regarding P &#8734; , there is an increasing importance of the vulnerability measure from BR to BS. Besides, while the degree is the best measure to guide the attacks for the National and S&#227;o Paulo networks, it is not for the BS, where vulnerability and betweenness have more importance. Similarly, in BR and SP, for W , the strength is the leading measure for attacks, and vulnerability is the worst. Conversely, although strength is also the best for BS, betweenness is the worst.</ns0:p></ns0:div> <ns0:div><ns0:head>8/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>As expected <ns0:ref type='bibr' target='#b0'>(Barab&#225;si et al., 2016)</ns0:ref>, random failures do not break the network until almost all nodes are removed, due to its scale-free structure, and all targeted attacks dismantle the networks for small f , except for the reactive strategy. The higher the threshold, the fewer nodes must be removed to break the network structure since the giant component is initially smaller than the observed for &#951; 0 . The R measure shows that the more effective attack strategy for P &#8734; is guided by degree, and by strength for W for all thresholds. The smaller the R, the more destructive the corresponding attack strategy is. The maximum number of components arises in targeted attacks guided by the betweenness centrality for BR and SP networks. When it comes to the BS network, the same happens for &#951; 0 , but other measures also hit the maximum for other thresholds.</ns0:p><ns0:p>The reactive strategy produces an impact similar to that of targeted attacks on decreasing the flow of people, although slightly worse. The number of remaining connected cities is always higher. Therefore, despite reacting to the disease spreading is a valid action, targeted attacks provide better results in terms of the size of the giant component and remaining flow in the system.</ns0:p><ns0:p>The cities from the state of S&#227;o Paulo that have higher values are also cited in recent studies <ns0:ref type='bibr' target='#b9'>(Freitas et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b13'>Guimar&#227;es Jr et al., 2020)</ns0:ref> on the most vulnerable cities to COVID-19 due to the intensive traffic of people.</ns0:p><ns0:p>Quickly breaking the transmission network is vital to contain any highly contagious disease, which demands the rapid implementation of control measures such as travel restrictions. Cities that preemptively adhered to the measures reported fewer cases than the others, and the virus reached them later <ns0:ref type='bibr' target='#b28'>(Tian et al., 2020)</ns0:ref>. The city of Wuhan was the main focus in China, and the complete isolation of the area was essential to mitigate the virus spreading <ns0:ref type='bibr' target='#b21'>(Li et al., 2020a)</ns0:ref>. On the other hand, the rest of the world received the SARS-CoV-2 concurrently at different places and had to divide efforts to restrain it.</ns0:p><ns0:p>The targeted attacks are especially relevant in areas where people are not sufficiently tested for COVID-19 since the reactive strategy strongly depends on effective epidemic surveillance. Li et al.</ns0:p><ns0:p>(2020b) estimate that 86% of infections were undocumented in China before the travel restrictions of January 23rd, 2020, and the undocumented infections were the source of 79% of the documented cases.</ns0:p><ns0:p>Underreporting is also present in Brazil, as <ns0:ref type='bibr' target='#b15'>Hallal et al. (2020)</ns0:ref> pointed out in a nationwide seroprevalence survey they conducted.</ns0:p><ns0:p>The US response to COVID-19 is mostly guided by Governors and Mayors primarily because of their political system. Korea and Taiwan implemented a centralized national strategy with the support of other government instances <ns0:ref type='bibr' target='#b14'>(Haffajee and Mello, 2020;</ns0:ref><ns0:ref type='bibr' target='#b20'>Kim et al., 2020)</ns0:ref>. Canada has the Health Portfolio Operations Centre (HPOC), which concentrates the operations at different levels of government. While in UK the response to the crisis was diverse in the different regions, in Switzerland, the communication and agreement between all levels of government was strong from the very beginning, based on mutual learning and integration <ns0:ref type='bibr' target='#b11'>(Gaskell and Stoker, 2020)</ns0:ref>. The Organisation for Economic Co-operation and Development OECD (2020) argues that coordinated response across regions and states minimize coordination failures since they avoid the 'pass the buck' behavior.</ns0:p><ns0:p>We assume that the targeted attacks represent the centralization of the efforts to isolate municipalities and states following specific and well-engineered strategies. Such coordination is only possible with a consensus at federal, state, and municipal levels since systematized isolation rely on the adherence of all involved parts. The random failures, on the other hand, abstract independent and decentralized actions.</ns0:p><ns0:p>Within this framework, the federal initiatives towards SARS-CoV-2 containment are more effective in breaking the transmission network than leaving the cities (or states) on their own. We have shown that random failures usually take longer to dismantle the networks than choosing the nodes with some criteria.</ns0:p><ns0:p>However, both the targeted attacks and the reactive strategy are possibly not feasible in some regions due to the widely divergent kinds of issues they may face <ns0:ref type='bibr' target='#b11'>(Gaskell and Stoker, 2020)</ns0:ref>, where the authorities must tailor specific strategies at the local level. The conducted robustness analysis points out the more central cities according to the network metrics and how their isolation impacts in connectivity and the flow of people. We thus present action plans that depend on cooperation and could conceivably rearrange in real-world scenarios.</ns0:p></ns0:div> <ns0:div><ns0:head>9/11</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2. Robustness analysis for the Brazilian mobility network (BR). Attack strategies: max s (strength), max k (degree), max V (vulnerability), max b (betweenness), and the reactive (COVID-19 curve). The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table1. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure3. Robustness analysis for the S&#227;o Paulo mobility network (SP). Attack strategies: max s (strength), max k (degree), max V (vulnerability), max b (betweenness), and the reactive (COVID-19 curve). The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table1. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Robustness analysis for the Brazilian states' mobility network (BS). Attack strategies: max s (strength), max k (degree), max V (vulnerability), max b (betweenness), and the reactive (COVID-19 curve). The failure curve is the average behavior for 50 random simulations. Three connection thresholds are considered: A,D,G) &#951; 0 ; B,E,H) &#951; 1 ; and C,F,I) &#951; 2 , as in Table1. Functions to evaluate the impact of removing a fraction f of nodes: the normalized number of connected components C( f )/C(0), the normalized size of the giant component P &#8734; ( f )/P &#8734; (0), and the normalized remaining flow in the system W ( f )/ W (0).</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Networks' statistics. The Brazilian (BR), S&#227;o Paulo state (SP), and Brazilian states (BS) networks, with three flow thresholds: &#951; 0 = 0, &#951; 1 = w and &#951; 2 = w + &#963; , where w is the average flow and &#963; is the standard deviation.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Network</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell></ns0:row><ns0:row><ns0:cell>N = |V |</ns0:cell><ns0:cell>5420</ns0:cell><ns0:cell>620</ns0:cell><ns0:cell>27</ns0:cell></ns0:row><ns0:row><ns0:cell>w</ns0:cell><ns0:cell>48.04</ns0:cell><ns0:cell>73.20</ns0:cell><ns0:cell>2032.29</ns0:cell></ns0:row><ns0:row><ns0:cell>&#963;</ns0:cell><ns0:cell cols='3'>100.21 122.79 4397.86</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>|E| for &#951; 0 65264 9592</ns0:cell><ns0:cell>474</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>|E| for &#951; 1 15505 2610</ns0:cell><ns0:cell>108</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>|E| for &#951; 2 4217</ns0:cell><ns0:cell>758</ns0:cell><ns0:cell>44</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Networks' measures. Average degree k , average betweenness b , average strength s and average vulnerability V for the Brazilian (BR), S&#227;o Paulo state (SP), and Brazilian states (BS) networks under flow thresholds &#951; 0 , &#951; 1 and &#951; 2 . Brazilian mobility network (BR) under A) &#951; 1 and B) &#951; 2 . The larger nodes are state capitals. Nodes with smaller degrees are blue, with higher degrees are red and intermediate values are dark red. The figure for &#951; 0 is not properly visible due to its 65254 edges.All measures (degree, betweenness, vulnerability, and strength) for the BR network under &#951; 0 exhibit the cities of S&#227;o Paulo and Belo Horizonte within the top-five higher values and most present Campinas and Bras&#237;lia. Concerning the SP network, the measures rank the cities of S&#227;o Paulo, Campinas, S&#227;o Jos&#233; do Rio Preto, and Ribeir&#227;o Preto within the top-five values as well. Differently, the BS network does not display a clear pattern for the degrees, but the states of S&#227;o Paulo and Minas Gerais come out in the first positions for betweenness, vulnerability, and strength (see the corresponding Tablesin the Supplemental</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Network</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>&#951; 0</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>&#951; 1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>&#951; 2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell><ns0:cell>BR</ns0:cell><ns0:cell>SP</ns0:cell><ns0:cell>BS</ns0:cell></ns0:row><ns0:row><ns0:cell>k</ns0:cell><ns0:cell>24.08</ns0:cell><ns0:cell>15.47</ns0:cell><ns0:cell>17.56</ns0:cell><ns0:cell>5.72</ns0:cell><ns0:cell>4.21</ns0:cell><ns0:cell>4.0</ns0:cell><ns0:cell>1.56</ns0:cell><ns0:cell>1.22</ns0:cell><ns0:cell>1.63</ns0:cell></ns0:row><ns0:row><ns0:cell>b</ns0:cell><ns0:cell cols='2'>5574.09 504.24</ns0:cell><ns0:cell>4.56</ns0:cell><ns0:cell cols='3'>4828.08 397.67 13.04</ns0:cell><ns0:cell cols='2'>2177.91 125.7</ns0:cell><ns0:cell>6.41</ns0:cell></ns0:row><ns0:row><ns0:cell>s</ns0:cell><ns0:cell cols='4'>1156.86 1132.35 35677.91 845.23</ns0:cell><ns0:cell cols='3'>813.66 30162.76 504.93</ns0:cell><ns0:cell>473.9</ns0:cell><ns0:cell>21043.46</ns0:cell></ns0:row><ns0:row><ns0:cell>V</ns0:cell><ns0:cell cols='3'>4.18E-4 3.62E-3 7.57E-2</ns0:cell><ns0:cell cols='3'>4.97E-4 4.6E-3 8.11E-2</ns0:cell><ns0:cell cols='3'>6.95E-4 6.53E-3 0.1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>A</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>B</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Figure 1.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='11'>/11 PeerJ reviewing PDF | (2020:06:49958:2:0:NEW 2 Oct 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"October 2, 2020 Vander L. S. Freitas Department of Computing Federal University of Ouro Preto Ouro Preto, MG, Brazil 35400-000 e-mail: [email protected] Dear Editor PeerJ We would like to submit our revised manuscript (v2) “Robustness analysis in an inter-cities mobility network: modeling municipal, state and federal initiatives as failures and attacks toward SARS-CoV-2 containment” by Vander L. S. Freitas, Gladston J. P. Moreira and Leonardo B. L. Santos for consideration for publication at PeerJ. The manuscript was carefully revised according to the comments and recommendations presented by the reviewers and editor, and we hope you will find it suitable for publication. Summary of Changes We thank the editor and the reviewers for their attention with this submission, as well as for their insightful comments and suggestions. The current version of our manuscript addresses all the comments. The changes in the manuscript are highlighted with the latexdiff tool and can be summarized as follows: • We have improved the abstract and the discussion section to explain the expressions “random failures” and “targeted attacks”. • The second part of the cover letter contains the last round of review with the corresponding line numbers so that the reviewer can track the changes. We apologize for sending the revisions without it the last time. The detailed responses to each point raised by the reviewers are provided below. Response to Editor Comment #1: Please, pay attention to R1 comments on some specific aspects of the manuscript that still deserve the authors’ attention. I second their claim about the attack/failure terms that still appear in the abstract and are extremely confusing. Answer: • We replied to all the reviewer concerns and have improved some parts of the abstract to explain the expressions “targeted attack” and “random failure” [lines 24 to 26]. ??? Response to Reviewer #1 Basic reporting Comment #1: The authors present a revised version of the manuscript with huge changes of the manuscript related to the text structure and, at least some added results since they extend the period of their analysis until September 2020. In some sense it is a “new” manuscript and due to that, in my point of view, it would be reasonable the authors point out, at least, the number of lines where they have introduced the changes that were required by us in the previous report. Following the tracked-changes version of the manuscript, I tried to catch the required changes. Another relevant point is to highlight in the cover letter the implications of extending the time interval of analysis for the discussion of the results. Summarizing the authors tried to answer the editor’s criticisms about the text – that was relevant for improving the quality of the manuscript – but do not take care to make clear in the letter what was done point by point, at least when the changes are related to the content of the manuscript (for instance, it was not necessary to do it for English revision). The authors have improved the quality of the figures and its description in figure captions. Now, based on an extended data bank (until September), it is possible to a value of robustness measure for COVID-19 curve in Figures 2 and 3 as in Figure 4. As I have suggested, the authors also present in a supplementary material a detailed analysis that is useful for reproduction as well as relevant for comparison with other scenarios. Answer: • We apologize for not indicating the line numbers where modifications took place. The second part of the letter contains the line numbers of the previous revisions. • Implications of extending the time interval of the analysis: we calculate the R value after attacking all the network nodes. The computation of the corresponding R values of the reactive strategy (COVID-19 curves) is only possible if all cities (states) have confirmed cases. In the first version of the manuscript, we had data available until May 26th, 2020, when all states had already registered at least one patient, but only 71% of Brazilian cities and nearly 81% from Sao Paulo State had. In the second version, we considered data until September 2nd, 2020, and these numbers increased to 98.9% for the Brazilian cities, and 100% for the Sao Paulo State, which allowed the computation of R in all cases. The remaining 1.1% for BR is not relevant as pointed out in the manuscript [lines 204 to 208]. The implication of extending the analysis period is that now we can directly compare the random failures, targeted attacks, and the reactive strategy since they all have associated R values. ??? Validity of the findings Comment #2: Concerning the summarization of the results, do the authors summarize it as below (which lines??) Authors: “Attacks usually cause more impact on the networks than both the reactive strategy and random failures. The only exception is when one evaluates the size of the giant component in the BS network under η0 , where all the strategies perform roughly the same. We observed that attacks guided by strength are more effective to diminish the total remaining flow of people; those guided by degree usually result in smaller giant components and; the ones guided by betweenness generate more isolated components.” In my point of view, it is reasonable to explain to the reviewers whether and which results were changed taking into account a longer time interval besides the value of R in Figure 2 and 3. Due to the changes in the manuscript, I think that the discussion about federal and municipal actions, at least in the context of Brazil and Brazilian networks analysed in the manuscript was diminished in the discussion section of the revised version but it is still in the abstract and also in the introduction. In my opinion, in the discussion it is important to emphasize how the results reveal what is affirmed in the abstract below – which is not so direct - or reduce the emphasis in the last two sentences of the abstract. Authors: “The municipal and state initiatives are here abstracted as nodes’ failures since there is no well defined country-level organization, and the federal actions are the targeted attacks. The results reveal that individual municipalities’ initiatives do not cause a high impact on mobility restraint since they tend to be disconnected from the country’s global interventions. Oppositely, the coordinated isolation of specific cities is crucial to detach entire network areas and thus prevent spreading. Besides, the targeted attacks pose better results than the reactive strategy.” For instance, to introduce in the abstract the idea that the association is a reasonable assumption. Reviewer: “Assuming the municipal ... the targeted attacks, the results reveal that individual ...” Answer: • We mention the quoted text you pointed out in lines 197 to 203 and lines 215 to 218. • The results for Figures 2 and 3 followed the trend they indicated in the last version of the manuscript. Previously, one could visually tell that the curves corresponding to the reactive strategy (COVID-19 data) presented worse results in harming the network, when compared to other types of attack. The interpretation of the results remains the same as before, but it is now supported by the R measure. • We have improved the assumptions in the abstract [lines 24 to 26] and the discussion [line 269 to 273]. ??? Comments for the Author Comment #3: I think that the manuscript may be published in PeerJ after clarifying some points in the letter for reviewers since there are many changes in the revised manuscript that in some features lead to, at least, modified emphasis of the manuscript, in my point of view. Please see the comments introduced in the above sections: basic reporting, experimental design and validity of the findings. Answer: • Thank you for the careful revision. ??? Response to Reviewer #2 Comment #1: The authors have addressed all the points raised. This manuscript can be published in its revised state. Answer: • Thank you for the careful revision. ??? Find below the revisions of the last round. We now indicate the manuscript line numbers of where we introduce changes. Response to Editor Comment #1: I apologize for the delay in sending this decision. I was able to secure two good reviews of your manuscript. It’s my opinion that authors must do a much better job in improving the flow of ideas throughout the whole paper. The way to structure paragraphs is not adequate, you very often split sentences about the same subject into two or more paragraphs. You also keep restating the motivation of the study over and over again in several parts of the Methods, this makes the reading a little bit chunky. Try to improve the flow of ideas. Almost all Results section need to be reformulated for clarity. Avoid repeating what is already written in the figure and table legends. Instead, call attention to any specific detail or summarize the pattern shown in them, and cite them at the end of the sentence. The conclusions section actually doesn’t provide conclusions and need to be reformulated entirely. See my comments on that in the PDF attached. English language needs to be carefully revised. Also, pay special attention to R1 suggestion on how to improve figures. Also, you definitely need to clarify the assumptions for the working hypothesis and make them understandable to a broader audience. What do you mean by attack and failure in this context? I second R2 about the necessity to provide raw data used for analysis. Provide more details in the M&M section that allow people to reproduce your analysis. For example, which software was used for analysing the data? Overall, the manuscript lacks citations, for the metrics (and formulae) you used, some concepts, as a support for assumptions, I have made several comments and corrections directly in the attached PDF. Please, remember to refer to it when revising your manuscript. Answer: • We thank you for handling our manuscript in such a careful way. We have performed a complete revision of the paper and improved the text structure as required, including the addition of more references for the metrics, formulae, and concepts. • We addressed all points raised by R1 and improved the Figures’ visualization. • We addressed all points raised by R2 and included the URL of the GitHub repository (https: // github. com/ vanderfreitas/ network_ robustness ) with all the data and source code to allow the reproduction of our work [line 170]. ??? Response to Reviewer #1 Basic reporting: Comment #1: In relation to the figures, I have two comments: the visualization of Figure 1 may be improved using blue color for the nodes with smaller degrees and light gray for the links; the captions of Figures 2, 3 and 4 should include that the procedure of removed nodes is based on different strategies as described on “Robustness” sub-section (deleting nodes with a higher degree, vulnerability, betweenness and strength). Moreover it is not clear in the caption how the COVID-19 cases (what means “temporal sequence”?) and “failure” curves are constructed; it is not clear the meaning of giant component and remaining flow for those curves. Finally I did not figure out the reason that there is a value of robustness measure for COVID-19 curve in Figure 4 but not in Figures 2 and 3. Answer: • We have improved the visualization of Figure 1: the nodes with smaller degrees are blue, higher degrees are red and intermediate values are dark red [between lines 173 and 174]. • We re-wrote the caption of all Figures for clarity. • The failure curve is associated with an average behavior for 50 simulations, where we randomly remove the nodes, one-by-one (it is in the captions of the Figures 2, 3, and 4). • The COVID-19 curve is formed from the removal of cities (states) according to the date of its first confirmed case. The first positive case in Brazil occurred in the city of Sao Paulo, followed by Barra Mansa, and Feira de Santana. This is exactly the order of removal: first Sao Paulo, then Barra Mansa, Feira de Santana in third place, and so on. • Eventually, there may exist many disjoint groups of connected nodes in the network. The number of nodes in the larger group is the size of the giant component. Whether all nodes are connected in a single group, the size of the giant component is the total number of nodes. Contrarily, if there are no edges, the number of the giant component is one, since each isolated node is a component itself. • The computation of the robustness measure R demands the availability of the entire curve, i.e., all nodes should be removed either by attack or failure. By May, when we submitted the manuscript, there were still many cities without reported COVID-19 patients both in BR (Figure 2) and SP (Figure 3), making impossible the computation of R. All states, on the other hand, had at least one case and this is why Figure 4 had the R for the COVID-19 curve. • By the time this new version of the manuscript is being submitted, almost all Brazilian cities have confirmed cases, which allowed us to include the referred R in Figures 2, 3, and 4 [lines 204 to 208]. ??? Comment #2: I think that the authors should present in a supplementary material a more complete list of results obtained from their analysis. For instance, in discussion section, the authors refer the two cities (or states) with the higher values of network measures in relation to the 3 selected networks: Brazilian cities (BR), Brazilian states (BS) and São Paulo state (SP). I suggest the authors include a complete list (in supplementary material) that reveal the order of states (BS) in relation to the higher values of the considered network measures; for BR network, the authors may select the capitals of the states, and for SP network, they may select, for example, 30 cities with highest number of cases. Answer: We have included the requested lists of cities and states in the Supplemental Material. ??? Experimental design: Comment #3: The authors set up a methodological procedure assuming the local initiatives as failures and the federal initiatives as attacks on the networks. I have some questions related to those assumptions as well as some conclusions associated with the obtained results. First of all, it would be reasonable to provide an explanation to those assumptions; for instance, in conclusions section the authors affirm that the attacks are isolation measures determined by the federal government and applied at a municipality level. As far as I know, there is a legal debate in Brazil about the assignment in relation to isolation of the municipalities. Is there a final decision about it? In case of affirmative answer, I suggest to include a reference about it; in case of negative answer, it would be interesting to justify the assumptions. Answer: Attacks are isolation measures determined in agreement with the different levels of administrative organization (federal, state, and municipal) and applied at a municipality level - so, its a local (municipal) action but based on a global (country level) analysis [lines 63 to 65]. Currently, the Brazilian states and municipalities have the power to choose which policies, when and how they are applied. Considering that they all act differently, regarding timing and method, we model them as random failures, i.e., nodes’ removals lack specific strategies. ??? Comment #4: In order to highlight the results, it is essential to clarify how the disease information concerning the temporal sequence of COVID-19 cases is related to the procedure of removed nodes that is fundamental for the methodology. Answer: We improved the explanation within the manuscript [lines 52 to 55, lines 77 to 82, lines 206 to 209, lines 241 to 244, and lines 254 and 259]. The reactive strategy consists of the removal of cities/nodes according to the date they documented the first case of COVID-19 (see Comment #1). This ordering may not impact the mobility like attacking the nodes according to their corresponding centrality measures (targeted attacks starting with the node with higher centrality value towards the nodes with smaller). Three metrics evaluate the impact of the reactive strategy and the targeted attacks in mobility: the number of components C, the size of the giant component P∞ , and the total remaining flow of people kW k. ??? Validity of the findings: Comment #5: The authors perform a comparative analysis of the network measures, pointing out which one is better or worse for each network. In general they conclude that the strength is a more suitable measure to guide the attacks due to the smaller value of R. As far as I understood, the attack is more effective than the failure for BR and SP networks since the curves fit better for the flow than for the giant component; however, both of them fits well for BS network. Do the authors agree with that summarization? Does that result provide any suggestion about the relevance of the strategies of the state government strategies? Answer: • Attacks usually cause more impact on the networks than both the reactive strategy and random failures. The only exception is when one evaluates the size of the giant component in the BS network under η0 , where all the strategies perform roughly the same. We observed that attacks guided by strength are more effective to diminish the total remaining flow of people; those guided by degree usually result in smaller giant components and; the ones guided by betweenness generate more isolated components [lines 197 to 203, and lines 215 to 225]. • The results indeed do not provide any clue of what could be done regarding the BS network under η0 (with all flows being considered), especially regarding the size of the giant component. In this case, all states have terrestrial connections to the others and the centrality measures for the nodes are qualitatively similar. The results start to change for increasing η since only the edges with higher flows are kept and some states start to play a key role in connectivity [lines 215 to 218]. ??? Comment #6: Undoubtedly the findings are relevant since it assigns a robustness analysis based on mobility network of a large country with a significative number of COVID-19 cases. The analysis may be useful for other transmitted diseases in the country. However, since it is a complex system, other factors get influence in the government decision of isolation; for instance it is very hard to isolate some large cities that are essential for the distribution of food and drugs. In this sense, it is not easy to determine which cities should be isolated. How do the authors make compatible their findings with that actual scenario? Answer: We thank the reviewer for the opportunity to make it clearer. The data we use concerns the flow of people and does not cover the transport of supplies (such as gas, food, and medicines). The isolation of a region consists of closing the borders to the flow of people to/from other regions [lines 27 to 28, and lines 71 to 73]. ??? Comments for the Author: Comment #7: I think that the manuscript may be published in PeerJ after clarifying the arisen points that I have classified as minor revision, since I suppose the authors had obtained the results suggested for supplementary material. Answer: Thank you for your careful revision. We included the requested data in the supplemental material and carefully answered your concerns. Response to Reviewer #2 Basic reporting: Comment #1: The authors report a timely study of a Brazilian transport network to identify potential strategies assisted by network analyses to contain the spread of Covid-19. The manuscript is clear, structured and easy to read. However, some notations could be more intuitively defined, for instance, why are the number of nodes in the giant component called P∞ (f ), I don’t understand what infinite has to do with it? More importantly, regarding the reproducibility of the results and access to raw data, I did not find any Git repository or an open server where the data and associated scripts to reproduce the results are deposited. I strongly suggest the authors to make their code and data available in an open platform. Answer: • We have created a repository on GitHub (https: // github. com/ vanderfreitas/ network_ robustness ) and added this information to the manuscript. The provided raw data and source code allow for the complete reproduction of our results [line 170]. • The notation P∞ (f ) comes from the Percolation Theory. Consider the inclusion of edges with probability p in a d-dimensional regular lattice. There is a threshold pc such that when p > pc the so-called percolating cluster emerges with high probability. Considering that the lattice is infinite, the size of the largest cluster becomes infinite as well at pc , allowing it to percolate the whole lattice. In this case, P∞ is the probability of a randomly selected node belong to the largest cluster (percolating cluster). Assessing the robustness of a complex network is the opposite process: instead of adding edges with probability p, we remove either nodes or edges with a probability f . The two processes map into each other by f = 1 − p. Therefore, P∞ (f ) is the size of the giant component after removing a fraction f of the nodes. Although the size of the investigated networks are not infinite, the notation is the same. Source: Chapter 8 of the traditional “Network science” book, available online: http: // networksciencebook. com/ chapter/ 8# percolation-theory Barabási, A.-L. et al. (2016).Network science. Cambridge university press. ??? Experimental design: Comment #2: Overall, I am convinced with the technicality of the analyses design. However, a main concern I have is using the size of giant component to assess the performance of the targeted attack. May be a complimentary metric such as the actual number of connected components (or clusters) for each node removal might indicate how the network is effectively destabilized, even when the giant component is intact. This might mean certain transportation among major cities are very strongly connected, however, many rural cities could be disconnected. So, in addition to the 2 metrics the authors have used to assess the network performance after node perturbations, I would like to see a metric that captures how many components or clusters remain after each perturbation. This might give a picture of how the cities are actually disconnected after node removals. I can see that fraction of remaining flow within the network could be useful but, still I suspect a few large components in the network might dominate the flow. Another strategy to choose node(s) for removal could be to employ minimum cut sets? i.e. the idea is to identify a minimal sets of nodes that disrupt the entire (or major) topology of the network? More details about this could be found here: Klamt and Gilles, 2004, Bioinformatics Answer: • We included the suggested metric (the number of components C(f )) and observed that betweennessguided attacks usually lead to the emergence of more disconnected components than other strategies [Figures 2, 3 and 4, lines 142 to 145, and lines 185 to 193]. This happens because nodes with higher betweenness are usually bridges between components. Most shortest paths between nodes of different “communities” pass through them. • Although it is possible that some large components dominate the flow, the tracking of the total remaining flow already gives the intuition of how many people are still able to move around. • The graphs we use allow only 1:1-relations. The Minimum cut sets (MCS) of Klamt and Gilles (2004) are usually employed to hypergraphs, where a single edge may link k nodes with l nodes. They graphically present this limitation in Figure 2. ??? Validity of the findings: Comment #3: Although I could partly agree with the author’s claim that global (or Federal) actions are more effective compared to local (or municipal) actions to disrupt the transportation network, I am not entirely sure if this can be concluded from the results presented as it is. As I mentioned in the earlier paragraph, I would like to see more complementary metrics to assess the network disruption both locally and globally. And may be an intermediate strategy where both local and global policies must be coordinated could also emerge if more metrics are included in the study. Answer: • We now evaluate the number of components as nodes are removed. • The global strategy is already an agreement between the federal government, states, and municipalities. The actions are orchestrated by the country and executed by the smaller instances. Besides, as our results suggest, an intermediate strategy would be the reactive strategy, since its impact in network robustness is usually amid targeted attacks and random failures in almost all cases. The only exception is when one evaluates the size of the giant component in the BS network under η0 , where all strategies perform roughly the same. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Bacterial biofilms have become a major threat to human health. The objective of this study was to isolate amylase-producing bacteria from soil to determine the overall inhibition of certain pathogenic bacterial biofilms. Methods. We used serial dilution and the streaking method to obtain a total of 75 positive amylase isolates. The starch-agar plate method was used to screen the amylolytic activities of these isolates, and we used morphological and biochemical methods to characterize the isolates. Optimal conditions for amylase production and purification using Sephadex G-200 and SDS-PAGE were monitored. We screened these isolates' antagonistic activities and the purified amylase against pathogenic and multi-drug-resistant human bacteria using the agar disk diffusion method. Some standard antibiotics were controlled according to their degree of sensitivity. Finally, we used spectrophotometric methods to screen the antibiofilm 24 and 48 h after application of filtering and purifying enzymes in order to determine its efficacy at human pathogenic bacteria. . Results. The isolated Bacillus species were Bacillus megaterium (26.7%), Bacillus subtilis (16%), Bacillus cereus (13.3%), Bacillus thuringiesis (10.7%), Bacillus lentus (10.7%), Bacillus mycoides (5.3%), Bacillus alvei (5.3%), Bacillus polymyxa (4%), Bacillus circulans (4%), and Micrococcus roseus (4%). Interestingly, all isolates showed a high antagonism to target pathogens. B. alevi had the highest recorded activity (48 mm) and B. polymyxa had the lowest recorded activity (12 mm) against Staphylococcus aureus (MRSA) and Escherichia coli, respectively. On the other hand, we detected no antibacterial activity for purified amylase. The supernatant of the isolated amylase-producing bacteria and its purified amylase showed significant inhibition for biofilm: 93.7% and 78.8%, respectively. This suggests that supernatant and purified amylase may be effective for clinical and environmental biofilm control. Discussion. Our results showed that soil bacterial isolates such as Bacillus sp. supernatant and its purified amylase are good antibiofilm tools that can inhibit multidrug-resistant former strains. They could be beneficial for pharmaceutical use. While purified amylase was effective as an antibiofilm, the isolated supernatant showed better results.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Bacterial biofilms have increasingly become a serious threat to human health <ns0:ref type='bibr' target='#b21'>(Hall-Stoodley et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b59'>Saber et al., 2017)</ns0:ref>. These substances have a high level of antibiotic resistance and are hosts to immune response stimulants <ns0:ref type='bibr' target='#b57'>(Rodrigues et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b68'>Sharma et al., 2019)</ns0:ref>. They also play an essential role in the pathogenicity of several chronic human infections <ns0:ref type='bibr'>(Parsek &amp; Singh, 2003)</ns0:ref>. Biofilm removal is a particularly difficult task. The principal method for preventing biofilm formation is applying chemicals or antimicrobials, such as chemical biocides, detergents, and surfactants. Biofilm destruction and prevention are effective methods, as are mechanical removal techniques such as shredding, sonication, freezing, and thawing <ns0:ref type='bibr' target='#b15'>(de Carvalho, 2007;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kalpana et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b16'>Elamary et al., 2020)</ns0:ref>. However, because the exopolysaccharide biofilm cells are protected <ns0:ref type='bibr' target='#b27'>(Kalpana et al., 2012)</ns0:ref>, it is difficult to completely remove biofilms using these methods. Using enzymes is also a good strategy for biofilm removal because enzymes are rabidly biodegradable and environmentally harmless <ns0:ref type='bibr' target='#b79'>(Xavier et al., 2005)</ns0:ref>. Amylase is a member of the glycosidic hydrolases, which are digestive enzymes that hydrolyze starch glycosidic bonds <ns0:ref type='bibr' target='#b30'>(Kaur et al., 2012)</ns0:ref>. This family also includes maltotriotic glucose, dextrin, and maltose. Amylase has exhibited excellent antibiotic activity against Pseudomonas aeruginosa and Staphylococcus aureus marine-derived biofilm-forming bacteria <ns0:ref type='bibr'>(Vaikundamoorthy et al., 2018)</ns0:ref>. Soil is the main part of the terrestrial environment, which is compared with aquatic environments with a large association of microorganisms. Among terrestrial bacteria, Bacillus sp. is the best source of amylase producers, including Bacillus subtilis, Bacillus cereus, and Bacillus polymyxa <ns0:ref type='bibr' target='#b17'>(El-Fallal et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b14'>Dash et al., 2015)</ns0:ref>. Bacillus amylase is thermostable, and retains a high pH, osmolarity, and high pressure, which are important for manufacturing <ns0:ref type='bibr' target='#b23'>(Islam et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Antibiotics produced by Bacillus sp. such as bacitracin, gramicidin S, polymyxin, and tyrotricidin have exhibited great efficacy against gram-positive and gram-negative bacteria PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr' target='#b45'>(Perez et al., 1992;</ns0:ref><ns0:ref type='bibr' target='#b46'>1993;</ns0:ref><ns0:ref type='bibr'>Yilmaz et al., 2006)</ns0:ref>. In this study, we identified and isolated Bacillus spp. from soil using morphological and biochemical assays. We compared the antimicrobial activity of these isolates against five human pathogenic strains. We optimized and purified the amylase after determining the optimal temperature, pH, incubation period, and starch levels needed for the greatest purification. Finally, we monitored the antibiofilm activity of the filtrate and purified amylase from these isolates.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>1-Soil sample collection. We collected 100 soil samples during January 2019 from different sites across the Luxor governorate (Monshaat Al Amari, 25&#176;41&#8242;14&#8243;N32&#176;41&#8242;40&#8243;E, 16.2 km), Egypt. Samples were collected in sterile plastic bags under aseptic conditions and were transported to the laboratory <ns0:ref type='bibr' target='#b55'>(Reed and Rigney, 1947)</ns0:ref>. We added 1 gram of soil to 5 ml of tryptic soy broth (Oxoid, Hampshire, United Kingdom), which we modified with 1% starch to make enrichment broth. Samples were incubated at 37 o C for 24 h. The landowner, Mohamed El sanousy, approved field sampling.</ns0:p></ns0:div> <ns0:div><ns0:head>2-Screening and isolation of amylase-producing bacteria.</ns0:head><ns0:p>Serial dilution techniques are one of the most precise methods for isolating bacteria from soil <ns0:ref type='bibr' target='#b25'>(Jamil et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b54'>Rasooli et al., 2008)</ns0:ref>. We performed serial dilutions up to 10 -7 . We aseptically transferred 100 &#181;l from each dilution, which we spread into tryptic soy agar media fortified with 1% starch. The plates were incubated at 37 o C for 24 h to determine the colony-forming unit (CFU)/ml. The plates were then flooded with iodine that turns blue when it reacts to unhydrolyzed starch. If the starch was hydrolyzed, a clear halo zone would appear against a dark blue background around the colonies that produce amylase <ns0:ref type='bibr' target='#b20'>(Gupta et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b1'>Abd-Elhahlem et al., 2015)</ns0:ref>. We further subcultured bacterial isolates to obtain a pure culture and identified isolates using standard morphological techniques based on colony shape, Gram's staining, spore formation, and biochemical Manuscript to be reviewed characterization <ns0:ref type='bibr' target='#b12'>(Cruickshank et al., 1975;</ns0:ref><ns0:ref type='bibr' target='#b11'>Collins and Lyne, 1984;</ns0:ref><ns0:ref type='bibr' target='#b31'>Koneman et al., 1992;</ns0:ref><ns0:ref type='bibr'>)</ns0:ref>.</ns0:p><ns0:p>Isolates were then maintained in a 70% sterilized glycerol stock at -70 o C for further use.</ns0:p><ns0:p>3-Selecting isolates for amylase purification. We selected isolates for amylase extraction and purification, as well as for comparing the purified amylase's antibiofilm activity against some human pathogenic bacteria, according to the starch hydrolysis ratio (SHR) that we calculated using the following equation <ns0:ref type='bibr' target='#b51'>(Pranay et al., 2019)</ns0:ref>:</ns0:p><ns0:p>SHR= clear halo zone diameter (mm) / colony growth diameter (mm).</ns0:p><ns0:p>Isolates were subcultured on starch agar plates, which were incubated for 24 h at 37 o C. After incubation, the plates were flooded with iodine. Finally, we calculated SHR using the equation above.</ns0:p></ns0:div> <ns0:div><ns0:head>4-Optimization of amylase production</ns0:head><ns0:p>a-Effect of temperature and incubation periods. The starch nutrient medium was prepared and the pH was adjusted to 7.5. We then inoculated the medium with the tested isolates. The culture was allowed to grow on a rotatory shaker (250 revs/min) at temperatures ranging from 15 to 65 o C over 48 h. We took 20 ml from each culture at all temperatures and time intervals <ns0:ref type='bibr'>(18, 24, 48, 72, 96, and 120 h)</ns0:ref> and centrifuged them to remove the bacterial cells. Finally, the supernatant was collected to assay the amylase activity <ns0:ref type='bibr' target='#b42'>(Nimisha et al., 2019)</ns0:ref>. b-Effect of pH. We prepared the starch nutrient medium and adjusted the pH to different values <ns0:ref type='bibr'>(5, 6, 7, 8, 9, and 10)</ns0:ref>. Each isolate was inoculated into a portion of this medium and were grown at 50 o C for 24 h. We then collected 20 ml from each isolate and applied the same treatment as above to determine amylase activity <ns0:ref type='bibr' target='#b42'>(Nimisha et al., 2019)</ns0:ref>. Manuscript to be reviewed quantities were added to fresh medium to give final concentrations of 0.1, 0.5, 1, 1.5, 2, 2.5, and 3%. We inoculated each isolate in this medium at 50 o C for 24 h to determine their amylase activity <ns0:ref type='bibr' target='#b42'>(Nimisha et al., 2019)</ns0:ref>.</ns0:p><ns0:p>5-Determining amylase activity under optimum conditions. The assay mixture contained 2 ml of a solution made up of 1% starch in 50 mM sodium phosphate buffer (pH 7) and 0.1 ml of enzyme solution. After 10 min. of incubation at 40 o C, we stopped the reaction by adding 2 ml of 3,5 dinitrosalicylic acid (DNS) reagent, and heated the tubes at 100 o C for 5 minutes. The absorbance was measured spectrophotometrically at 540 nm using a blank containing buffer instead of the culture supernatant. We calculated the amount of reduced sugars from a maltose standard curve <ns0:ref type='bibr' target='#b37'>(Meyer et al., 1951)</ns0:ref>. Protein was determined using <ns0:ref type='bibr' target='#b9'>Bradford's (1976)</ns0:ref> method.</ns0:p></ns0:div> <ns0:div><ns0:head>6-Enzyme purification</ns0:head><ns0:p>a-Ammonium sulfate precipitation. The crude amylase enzyme was brought to 45% saturation with ammonium sulfate and was kept overnight in a cold room at 4 o C. We removed the precipitate, brought the supernatant to 85% saturation with ammonium sulfate, and centrifuged it at 8,000 rpm for 10 min at 4 o C. After collecting the precipitate during this step, we stored it at 4 o C <ns0:ref type='bibr' target='#b69'>(Shinde &amp; Soni, 2014)</ns0:ref>. b-Dialysis. This step was conducted to exclude the ammonium sulfate remains and to concentrate the enzyme. We used the dialysis tubes, which were previously soaked in 0.1 M phosphate buffer (pH 6.2), for precipitate dialysis. The precipitate was dissolved in 0.1 M phosphate buffer and was dialyzed against the same buffer <ns0:ref type='bibr' target='#b58'>(Roe, 2001)</ns0:ref>. Manuscript to be reviewed 7), the flow rate was adjusted to 1 ml per 1 min., and 200 ml of eluents were collected into 40 tubes (1x7 cm) using an automatic circular fraction collector. We determined enzyme activity and protein concentration in each fraction using the described assay method. Fractions of the highest specific activity were pooled together and kept for further study.</ns0:p></ns0:div> <ns0:div><ns0:head>d-SDS-PAGE</ns0:head><ns0:p>We carried out polyacrylamide gel electrophoresis according to <ns0:ref type='bibr' target='#b32'>Laemmli's (1970)</ns0:ref> method using 10% polyacrylamide gel. Purified B. alvei and B. cereus amylase was loaded into wells parallel to the standard protein markers. The protein bands were stained with Coomassie brilliant blue (Sigma, St. Louis, MO, USA). We estimated the enzyme's relative molecular weight by comparing it to molecular mass standard markers (Fermentas, Vilnius, Lithuania).</ns0:p></ns0:div> <ns0:div><ns0:head>7-Antibacterial activities</ns0:head><ns0:p>7.1. Antagonistic efficacy of the isolated bacteria. We compared the antagonistic efficacy of all isolates against five human pathogenic strains (Escherichia coli, P. aeruginosa, S. aureus (MRSA), Klebsiella pneumoniae, and Acinetobacter baumanii). Strains were kindly provided by the International Luxor Hospital in Luxor Governorate, Egypt. We performed screening using the disc diffusion method. All bacteria were cultured on TSB modified with 1% starch, adjusted to OD 595 = 0.01, and incubated at 37 o C at 24 h. The isolated bacterial cultures were centrifuged to exclude the cell debris (6,000 rm for 15 min., Biofuge). We then modified 20 ml of TSA with 1% starch, and poured it in a sterile Petri plate (100 mm diameter). We streaked 100 &#181;l of the five tested pathogens on the plates and punched 6-mm wells in the plates using a sterile borer.</ns0:p><ns0:p>The wells were then filled with 100 &#181;l of the isolated bacteria filtrate, and the plates were incubated at 37 o C for 24 h. The inhibition zone was measured using a ruler <ns0:ref type='bibr' target='#b56'>(Reinheimer et al., 1990)</ns0:ref>. Standard antibiotics were used as the controls according to the Kirby Bauer disk diffusion PeerJ reviewing <ns0:ref type='table' target='#tab_6'>PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:ref> Manuscript to be reviewed method <ns0:ref type='bibr' target='#b6'>(Bauer et al., 1966)</ns0:ref>. The antibiotics were chloramphenicol (C; 30 &#181;g, Oxoid), oxacillin (OX; 1 mcg, Bioanalyse&#174;), vancomycin (VA; 30 mcg, Bioanalyse&#174;), ampicillin/sulbactam (SAM; 10/10 mcg, Bioanalyse&#174;), penicillin G (P; 10 U; Bioanalyse&#174;), erythromycin (E; 15 mcg, Bioanalyse&#174;), sulfamethoxazole/trimethoprim (SXT; 23.75/1.25 &#181;g, BBL&#8482;), cefotaxime (CTX; 30 mcg, Bioanalyse&#174;), gentamycin (GM; 10 &#181;g, Bioanalyse&#174;), meropenem (MEM; 10&#181;g, Bioanalyse&#174;), piperacillin (PIP; 100 &#181;g, Bioanalyse&#174;), and piperacillin-tazobactam (PTZ; 100/10 &#181;g, Bioanalyse&#174;). We interpreted the results using the Clinical Laboratory Standard Institute guidelines (CLSI, 2017) to determine whether the tested pathogens were resistant, intermediate, or sensitive against the antibiotics. 7.2. Antibacterial activity of purified amylase enzyme from the isolated Bacillus. We placed 100 &#181;l of purified amylase from the selected isolates according to their SHR in the wells of the agar plates inoculated with the target strains . The plates were incubated at 37 o C for 24 h. The halo zone was measured using a ruler.</ns0:p></ns0:div> <ns0:div><ns0:head n='7.3.'>Biofilm formation assay.</ns0:head><ns0:p>We determined the biofilm formation ability of the tested pathogens (E. coli, P. aeruginosa, S. aureus (MRSA), K. pneumoniae, and A. baumanii) using 96-well polystyrene plates <ns0:ref type='bibr' target='#b66'>(Seper et al., 2011)</ns0:ref> and the methods described by <ns0:ref type='bibr'>Salem et al. (2015)</ns0:ref>:</ns0:p><ns0:p>isolates were subcultured on tryptic soy agar for 24 h at 37 o C, suspended in tryptic soy broth, and adjusted to an OD 595 of 0.02. We placed 130 &#181;l of each adjusted isolate culture in the microtitre plate (U bottom, Sterilin) for 24 and 48 h at 37 o C. After incubation, the wells were washed with distilled water (six times) and were stained with 0.1% crystal violet for 10 min. The wells were then washed again with distilled water (four times) to remove excess stain. Finally, the wells were destained using 210 &#181;l of ethanol 96%, and the OD 595 was read using an Infinite &#174; F50 Robotic (Ostrich) Microplate Plate to quantify the amount of biofilm. <ns0:ref type='table' target='#tab_6'>PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:ref> Manuscript to be reviewed 7.4. Antibiofilm activity of the isolated Bacillus sp. filtrate and its purified amylase enzyme.</ns0:p></ns0:div> <ns0:div><ns0:head>PeerJ reviewing</ns0:head><ns0:p>We compared the antibiofilm effects of the isolated bacteria filtrate and the purified amylase from the selected isolates against the five human biofilm pathogenic bacteria using the following spectrophotometric methods: a fresh isolate culture was prepared and adjusted to 0.5 McFarland (10 6 CFU/ml), and 30 &#181;l (this volume was selected according to a preliminary experiment) of these cultures and purified amylase enzyme were added to 130 &#181;l of the tested pathogens at an OD 595 of 0.02 after 24 h of incubation at 37 o C to allow biofilm formation. The plates were then incubated for 24 and 48 h and stained with crystal violet. Wells without isolated cultures or amylase served as controls.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis.</ns0:head><ns0:p>The variability degree of the results was expressed in the form of mean &#177; standard deviation (mean &#177;SD) based on three independent determinations (n=3). We statistically analyzed the data by one-way ANOVA analysis and compared the control and treatment groups using the least significant difference (LSD) test at 1% (*) levels <ns0:ref type='bibr'>(Snedecor &amp; Cochran, 1980)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head><ns0:p>1-Screening and isolating amylase-producing bacteria. Microorganisms that produce amylase are generally isolated from soil and other sources <ns0:ref type='bibr' target='#b19'>(Fossi et al., 2005)</ns0:ref>. Our study explored the isolation of amylase-producing bacteria from soil using the serial dilution spread plate technique. <ns0:ref type='bibr' target='#b70'>Singh &amp; Kumari (2016)</ns0:ref> used a similar method by diluting soil samples on starch agar plates and flooding the plates with an iodine solution. The presence of a halo zone around certain colonies indicated amylase production, and a total of 75 bacterial isolates showed a zone of clearance with a starch agar medium. Bacterial isolates were selected according to their amylolytic activity (Table <ns0:ref type='table'>1</ns0:ref>). A similar method was also employed by Magalhaes (2010). We further characterized isolates using morphological and biochemical tests shown in Tables <ns0:ref type='table' target='#tab_6'>1 and PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:ref> Manuscript to be reviewed CFU/ml (Table <ns0:ref type='table'>1</ns0:ref>). <ns0:ref type='figure'>and B</ns0:ref>. lentus with SHRs of 6.0, 5.67, 5.33, 5.0, 4.0, and 3.5 mm, respectively, for amylase purification.</ns0:p></ns0:div> <ns0:div><ns0:head>2-Optimizing amylase production. Using the starch hydrolysis rates shown in</ns0:head><ns0:p>a-Effect of temperature and time intervals. All isolates showed maximum amylase production after 24 h. Similar results were obtained by <ns0:ref type='bibr' target='#b70'>Singh &amp; Kumari (2016)</ns0:ref>, who observed that the highest amylase activity of some Bacillus sp. occurred at 24 h of incubation and that the activity began to decrease after 48 and 72 h of incubation time (Fig. <ns0:ref type='figure'>1B</ns0:ref>). B. megaterium, B. subtilis, and B. cereus showed maximum amylase production at 45 o C, while other isolates showed maximum amylase production at 55 o C. <ns0:ref type='bibr' target='#b38'>Mohamed et al. (2009)</ns0:ref> similarly reported that some amylase were stable at 40 o C and some at 50 o C (Fig. <ns0:ref type='figure'>1A</ns0:ref>).</ns0:p><ns0:p>b-Effect of pH. All Bacillus isolates showed maximum amylase production at a pH of 8, except for B. subtilis which maximally produced amylase at a pH of 7. A previous study by <ns0:ref type='bibr' target='#b7'>Behal et al. (2006)</ns0:ref> found that the optimum pH for amylase production was 8. Another study by <ns0:ref type='bibr'>Singh &amp; Kumar (2016)</ns0:ref> reported that while amylase activity was recorded at different pH levels from 5 to 10, maximum activity was observed at pH 7 (Fig. <ns0:ref type='figure'>1C</ns0:ref>). Manuscript to be reviewed c-Effect of substrate concentration. Our results showed that B. subtilis and B. cereus had maximum amylase production at 1.5% soluble starch concentration. The remaining isolates showed maximum amylase production at 2.0 % soluble starch concetration (Fig. <ns0:ref type='figure'>1D</ns0:ref>). <ns0:ref type='bibr'>Mishra &amp; Behera (2008)</ns0:ref> reported that Bacillus strains produced the maximum yield of amylase at a starch concentration of 2%.</ns0:p><ns0:p>d-Enzyme activity. We purified extracellular amylase from the Bacillus isolated from soil to homogeneity using 45-85% ammonium sulfate precipitation and Sephadex G-200 (Fig. <ns0:ref type='figure'>2</ns0:ref>). As shown in Table <ns0:ref type='table'>3</ns0:ref>, the highest amylase activity was found in B. alvei (96.02 U/ml), followed by B. thuringiesis (88.64 U/ml). B. megaterium, B. subtilis, and B. cereus showed amylase activities of 80.03, 76.0, and 55.9 U/ ml, respectively. B. lentus showed the lowest amylase activity of 45.69 U/ml. e-SDS-PAGE. After purification, the SDS-PAGE profile showed a single protein band of amylase for each bacteria, confirming that the enzyme has been purified to homogeneity. The molecular weights of B. alvei and B. cereus were 60 KDa and 43KDa, respectively (Fig. <ns0:ref type='figure'>3</ns0:ref>). The molecular weight of B. alvei was similar to that of the amylase isolated from B. subtilis (56 KDa and 55 KDa, respectively) <ns0:ref type='bibr' target='#b5'>(Bano et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b74'>Takkinen et al., 1983)</ns0:ref>. The molecular weight of B. cereus was equal to the molecular weight of amylase from B. cereus and B. subtilis (42 KDa) <ns0:ref type='bibr' target='#b2'>(Annamalai et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Das et al., 2004)</ns0:ref>. Previous studies have reported different molecular weights for amylase isolated from Bacillus sp. <ns0:ref type='bibr' target='#b35'>Lin et al. (1998)</ns0:ref> found that Bacillus sp. can produce five different forms of amylase. Manuscript to be reviewed and purified amylase with standard antibiotics against human pathogens (Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). The standard antibiotics served as the control group since pathogenic bacteria can become extremely resistant to widely-used antibiotics. The pharmaceutical industry is in need of new and natural antimicrobials that can overcome the problem of multidrug-resistant strains <ns0:ref type='bibr' target='#b65'>(Schmidt, 2004;</ns0:ref><ns0:ref type='bibr'>Salem et al., 2015;</ns0:ref><ns0:ref type='bibr'>2017)</ns0:ref>. Several soil organisms can produce antibiotics using a survival mechanism that can eliminate their competition <ns0:ref type='bibr' target='#b75'>(Talaro &amp; Talaro, 1996;</ns0:ref><ns0:ref type='bibr' target='#b26'>Jensen &amp; Wright, 1997)</ns0:ref>. The Bacillus genus is a terrestrial strain that can produce inhibitory compounds from peptide-derivative and lipopolypeptide antibiotics <ns0:ref type='bibr' target='#b36'>(Mannanov &amp; Sattarova, 2001;</ns0:ref><ns0:ref type='bibr' target='#b76'>Tamehiro et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b72'>Stein, 2005)</ns0:ref>. <ns0:ref type='bibr'>Oscaiz et al. (1999) and</ns0:ref><ns0:ref type='bibr'>Yilmaz et al. (2006)</ns0:ref> found that isolated bacteriocin-producing strains such as Bacillus sp. were active against gram-negative and grampositive bacteria. We compared the antimicrobial activity of the isolated amylase-producing bacteria and purified amylase against five human pathogenic bacteria (Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). We found that E. coli was resistant to sulfamethoxazole-trimethoprim (23.75/1.25 mcg), gentamycin (10&#181;g), cefotaxime (30&#181;g), piperacillin (100&#181;g), and piperacillin-tazobactam (100/10&#181;g); showed intermediate sensitivity to ampicillin-sulbactam (10/10 mcg); and was sensitive to chloramphenicol (30 &#181;g) and meropenem (10 &#181;g). Notably, we observed that all isolated bacteria showed high antimicrobial activity in response to E. coli, with B. polymyxa showing the most activity (36 mm) and B. subtilis and B. cereus showing the least (12 mm). B. mycoides and M. roseus showed no antimicrobial activity in response to E. coli. Our results were consistent with the results of <ns0:ref type='bibr' target='#b40'>Moshafi et al. (2011)</ns0:ref>, who observed that one soil bacterial isolate, identified as Bacillus sp., was found to inhibit six pathogenic bacteria, namely E. coli, K. pneumoniae, S. typhi, P. aeruginosa, S. aureus, and S. epidermidis. When examining the antimicrobial activity in response to K. pneumoniae, we found that although K. pneumoniae was resistant to all tested <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). In contrast, B. mycoides and M. roseus were resistant to K. pneumoniae. In a recent study, <ns0:ref type='bibr' target='#b44'>Ozabor &amp; Fadahunsi (2019)</ns0:ref> reported that B. subtilis metabolites inhibited K. pneumoniae, P. aeruginosa, S. aureus, E. coli, P. mirabilis, and other bacteria. A. baumanii was resistant to all tested antibiotics and was sensitive only to chloramphenicol (30 &#181;g). However, all isolated bacteria showed improved antibacterial effects against the tested pathogens, with B. alvei and B. cirulans showing the greatest inhibitory effects (39 mm), and B. subtilis and B. thuringiesis showing the lowest (21 mm) (Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). We found that B. mycoides and Micrococcus roseus displayed no inhibitory effects against the tested pathogens. <ns0:ref type='bibr' target='#b53'>Ramachandran et al. (2014)</ns0:ref> reported that B. subtilis showed antimicrobial activity against A. baumanii, E. coli, K. pneumoniae, P. aeruginosa, and S. aureus. The susceptibility level of P. aeruginosa indicated that it was resistant to sulfamethoxazole-trimethoprim (23.75/1.25 &#181;g), cefotaxime (30 mcg), gentamycin (10 &#181;g), meropenem (10 &#181;g), and piperacillin (100 &#181;g). It had intermediate sensitivity to chloramphenicol (30 &#181;g) and ampicillin-sulbactam (10/10 mcg), and was sensitive to piperacillin-tazobactam (100/10 &#181;g). We noted that all isolates showed great antibacterial effects against the tested pathogens, with B. lentus and B. cirulans having the greatest effects (32 mm) and B. subtilis having the least (15 mm). <ns0:ref type='bibr'>Salem et al. (2015)</ns0:ref> similarly reported that Bacillus strains exhibited antimicrobial activity against P. aeruginosa, E. coli, and S. typhi. <ns0:ref type='bibr' target='#b45'>Perez et al. (1992)</ns0:ref> and <ns0:ref type='bibr' target='#b3'>Aslim et al. (2002)</ns0:ref> found that B. subtilis, B. thuringiesis, and B. megaterium showed antibacterial activity against E. coli and P. aeruginosa. We found that S. aureus (MRSA) was resistant to oxacillin (1 mcg), vancomycin (30 mcg), penicillin G (10 U), cefotaxime (30 mcg), <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). However, B. mycoides and M. roseus did not affect S. aureus. Similar results were obtained by <ns0:ref type='bibr' target='#b40'>Moshafi et al. (2011)</ns0:ref> and <ns0:ref type='bibr' target='#b53'>Ramachandran et al. (2014)</ns0:ref>. In contrast to the high antimicrobial activity observed in the isolated soil bacteria, the purified amylase from the selected isolates had very little effect on E. coli and K. pneumoniae (the highest inhibition diameter was 7.5 mm), and no recorded effect in response to the other tested pathogens (Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). This result is similar to that of <ns0:ref type='bibr' target='#b27'>Kalpana et al, (2012)</ns0:ref>, who confirmed that amylase enzyme has no antibacterial effect.</ns0:p></ns0:div> <ns0:div><ns0:head>3-</ns0:head></ns0:div> <ns0:div><ns0:head n='3.2.'>Biofilm formation assay.</ns0:head><ns0:p>We quantitatively determined the amount of biofilm (OD 595 ) in the tested pathogens and designated the 24 and 48 h treatments as the control groups <ns0:ref type='bibr'>(Figs. 4,</ns0:ref><ns0:ref type='bibr'>5,</ns0:ref><ns0:ref type='bibr'>6,</ns0:ref><ns0:ref type='bibr'>and 7)</ns0:ref>. Using the OD 595 nm mean values, we defined the pathogens as low, moderate, or high bacterial biofilm formers when the OD 595 nm was &lt;1, 1 -2.9, and &gt; 2.9, respectively. A. baumanii and Klebsiella pneumoniae were high biofilm formers while E. coli, P. aeruginosa, and S. aureus (MRSA) were low biofilm formers.</ns0:p></ns0:div> <ns0:div><ns0:head>3.3.</ns0:head><ns0:p>Antibiofilm activity of isolated bacterial filtrate and purified amylase enzyme from selected Bacillus isolates. In a natural ecosystem, bacteria can be exist in two forms: planktonic cells, which are susceptible to antibiotics and other antimicrobial agents, and biofilm, which are resistant to antibiotics and disinfectants <ns0:ref type='bibr' target='#b34'>(Limoli et al., 2015)</ns0:ref>. A biofilm is a complex community of bacteria attached to a surface or interface enclosed in an exopolysaccharide matrix, protected from unfavorable antibiotics, host defenses, or oxidative stresses <ns0:ref type='bibr' target='#b67'>(Shakibaie, 2018)</ns0:ref>. Microbial Manuscript to be reviewed biofilms have created huge problems in the treatment of both community and hospital infections.</ns0:p><ns0:p>Most antimicrobial agents are unable to penetrate biofilm due to its extracellular polymeric substances (EPS), which act as a barrier protecting the bacterial cells within the biofilm.</ns0:p><ns0:p>Therefore, we must use compounds that have the potential to degrade the biofilm's EPS.</ns0:p><ns0:p>Enzymes have proven to be effective in EPS degradation <ns0:ref type='bibr'>(Kalapna et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b33'>Lequette et al., 2010)</ns0:ref>. In our study, we compared the antibiofilm activity of the Bacillus sp. that we isolated from soil (supernatant) and the purified amylase from these isolates against five human pathogenic biofilm former strains. Our study has reported that Bacillus supernatant and amylase enzyme can inhibit the biofilm formation in various pathogens. We confirmed the ability of pathogenic bacterial strains to form biofilm formation using spectrophotometric methods before applying the antibiofilm treatments of bacterial filtrate and purified amylase enzyme (Fig. <ns0:ref type='figure' target='#fig_13'>4</ns0:ref>, 5, 6, and 7). The antibiofilm activity was screened using a spectrophotometric method with crystal violet staining. Our results showed that the bacteria isolated from soil exhibited significant antibiolfilm effects against the tested pathogenic strains after 24 h of treatment. The percentage of inhibition significantly increased after 48 h of treatment. The highest percentage of inhibition was recorded for B. circulans against K. pneumonia: 93.7 % after 48 h of treatment (Fig. <ns0:ref type='figure' target='#fig_14'>5D, T8</ns0:ref>). We also monitored the efficacy of the purified amylase enzyme as an antibiofilm against the same tested pathogens. Our results revealed that the purified amylase showed significant antibiofilm effects after 24 h of treatment. The percentages of inhibition significantly increased after 48 h of treatment. We observed the highest percentage for B. alvei against K. pneumonia:</ns0:p><ns0:p>78.8% after 48 h of treatment (Fig. <ns0:ref type='figure'>7D, T6</ns0:ref>). Our results also showed the greatest enzyme activity Manuscript to be reviewed an enzyme activity of 55.9 U/ml, and B. lentus, with an enzyme activity of 45.69 U/ml. <ns0:ref type='bibr' target='#b27'>Kalpana et al., (2012)</ns0:ref> first reported that purified amylase enzyme from B. subtilis was a good antibiofilm agent against biofilm-forming clinical pathogens. The purified enzyme caused 68.33%, 64.84%, 61.81%, and 59.2% of inhibition in V. cholerae (VC5, VC26), MRSA (102), and P. aeruginosa ATCC10145, respectively. Another study by <ns0:ref type='bibr'>Vaikundamoorthy et al., (2018)</ns0:ref> confirmed the antibiofilm efficacy of the thermostable amylase enzyme from B. cereus.</ns0:p><ns0:p>It is worth noting that the isolated bacteria filtrate showed great antibiofilm activity compared to the purified amylase enzyme from the selected isolates. This may be due to the accumulation of some extracellular and intracellular metabolites in the medium, which is further explained by the metabolic overflow theory <ns0:ref type='bibr' target='#b48'>(Pinu et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b47'>Pinu et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b22'>Horak et al., 2019)</ns0:ref>. Bacillus also showed great efficacy in the production of carbohydrate-active enzymes and bioactive compounds, as well as the secretion of a variety of extracellular metabolites and lytic enzymes (Abdel-Aziz, 2013). Additionally, Bacillus species are the most efficient at producing peptide antibiotic compounds such as polymyxin, colistin, and circulin <ns0:ref type='bibr'>(Katz &amp; Domain, 1997;</ns0:ref><ns0:ref type='bibr' target='#b4'>Atanasova-Pancevska et al., 2016)</ns0:ref>. Our results also indicated a great inhibition of biofilm from the amylase enzymes of B. alvei (96.02 U/ml), followed by B. thuringiesis (88.64 U/ml), B. megaterium (80.03 U/ml), B. subtilis (76.0 U/ml), B. cereus (55.9 U/ml), and B. lentus (45.69 U/ml) (Table <ns0:ref type='table'>3</ns0:ref>). This may be due to the increased enzyme activity in each species.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Our results indicated that the ability of Bacillus sp. to produce extracellular and intracellular metabolites, lytic enzymes, and some peptide antibiotics directly affects the antimicrobial functions of various Bacillus sp. (amylase producers) in the soil. We observed the highest inhibition rate (93.7%) when comparing the species' antibiofilm effects against five human pathogenic strains. We observed an inhibition rate of 78.8% when comparing the antibiotic Manuscript to be reviewed biofilm activity of purified amylase against the strains. Our study showed that Bacillus filtrate is an effective clinical antibiofilm. Futher studies are being conducted to determine the exact composition of the filtrate and its active agents. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>.25&#181;g) R R R R I Cefotaxime (30 mcg) R R R R R Gentamycin (10 &#181;g) R R R R R Meropenem (10&#181;g) S R R R NA Piperacillin (100 &#181;g) R R R R NA Piperacillin-tazobactam (100/10 &#181;g) R R R S NA</ns0:formula><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Optimization and purification conditions of amylase enzyme from selected Bacillus sp. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>c-Effect of starch concentration. All Bacillus isolates were grown on nutrient broth medium with a pH of 9, except B. subtilis which was grown at a pH of 7. Different soluble starch PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>The crude enzyme preparations of the six culture filtrates were applied separately to a column of DEAE-Sephadex G-200. The enzyme was eluted with a linear gradient of sodium chloride (0 -0.4 M) in 200 ml of sodium phosphate buffer (0.05 M and pH PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>2. Our results showed that the 75 isolates were comprised of 19 B. megaterium, 12 B. subtilis, 10 B. cereus, eight B. thuringiesis, eight B. lentus, four B. mycoides, four B. alvei, three B. polymyxa, three B. circulans, and three Micrococcus roseus. B. megaterium had the highest recorded prevalence (26.7%) and B. circulans and Micrococcus roseus had the lowest (4%). The CFU of the amylase-producing bacteria in our 100 soil samples ranged from 115x10 3 -198x10 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>Antibacterial activity 3.1. Antagonistic efficacy of the isolated bacteria and purified amylase enzyme from selected isolates. In this study, we compared the antimicrobial activity of Bacillus supernatant PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020) Manuscript to be reviewed antibiotics, it showed intermediate sensitivity to ampicillin-sulbactam (10/10 mcg). It is worth mentioning that all isolates had great antimicrobial effects, with B. megaterium showing the highest inhibition (26 mm) and B. polymyxa showing the lowest (17 mm) (Table</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020) Manuscript to be reviewed and gentamycin (10 &#181;g), and showed intermediate sensitivity to chloramphenicol (30 &#181;g), erythromycin (15 mcg), and sulfamethoxazole-trimethoprim (23.75/1.25 &#181;g). The isolated amylase-producing bacteria showed better antibacterial effects on the tested pathogens, with the greatest effect shown by B. alvei (48 mm) and the least effect shown by B. cereus (14 mm) (Table</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>from B. alvei (96.02 U/ ml), followed by B. thuringiesis (88.64 U/ ml), B. megaterium (80.03 U/ ml), and B. subtilis (76.0 U/ml). The lowest antibiofilm efficacy was recorded in B. cereus, with PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>trimethoprim (23.75/1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A): The effect of temperature. B. megaterium, B. subtilis, and B. cereus showed maximum amylase production at 45 o C, while other isolates showed maximum amylase production at 55 o C. (B): The effect of incubation time. All isolates showed maximum amylase production after 24 h incubation. (C): The effect of pH. All Bacillus isolates showed maximum amylase production at a pH of 8.0 except B. subtilis, which maximally produced amylase at a pH of 7.0. (D): The effect of starch concentration. B. subtilis and B. cereus had maximum amylase production at 1.5% soluble starch concentration. The remaining isolates had maximum amylase production at 2.0% soluble starch concentration.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='37,42.52,70.87,525.00,440.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='38,42.52,255.37,525.00,294.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='43,42.52,428.62,525.00,280.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='44,42.52,428.62,525.00,275.25' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 ,</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>we selected the six isolates with the highest hydrolysis rates, namely B. alvei, B. thuringiesis, B.</ns0:figDesc><ns0:table /><ns0:note>megaterium, B. subtilis, B. cerus,</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table ( 1</ns0:head><ns0:label>(</ns0:label><ns0:figDesc>): Prevalence of Bacillus species isolated from soil.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Isolates a</ns0:cell><ns0:cell /><ns0:cell cols='3'>No. of isolates b Percentage c (%) CFUml -1d</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Parameters</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Bacillus megaterium</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>26.7</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Bacillus mycoides Bacillus alvei Bacillus polymyxa Bacillus circulans Bacillus subtilis Bacillus cereus Bacillus thuringiesis Bacillus lentus</ns0:cell><ns0:cell>12 10 8 8 4 4 3 3</ns0:cell><ns0:cell>16 13.3 10.7 10.7 5.3 4 4 5.3</ns0:cell><ns0:cell>115&#215;10 3 -198&#215;10 5</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Micrococcus roseus</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>4</ns0:cell></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>75</ns0:cell><ns0:cell>100</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>a: The amylase-producing bacteria isolated from soil. b: Number of each isolated type from the</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>total number of the positive isolated sample, c: Percentage of each isolate, d: Average colony-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='5'>forming unit of amylase-producing bacteria per ml of 100 g soil samples (highest value to lowest</ns0:cell></ns0:row><ns0:row><ns0:cell>value).</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Biochemical activities of amylase-producing bacterial isolates and their starch hydrolysis rates.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>a: Morphological and biochemical tests used for identifying isolated bacteria. +: Positive, -:</ns0:cell></ns0:row><ns0:row><ns0:cell>Negative, b: starch hydrolysis rate.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table ( 2</ns0:head><ns0:label>(</ns0:label><ns0:figDesc>): Biochemical activities of amylase-producing bacterial isolates and their starch hydrolysis rates.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell>Isolate</ns0:cell><ns0:cell>Gram</ns0:cell><ns0:cell cols='3'>Motility Catalase Egg yolk</ns0:cell><ns0:cell>Nitrate</ns0:cell><ns0:cell>Vogas</ns0:cell><ns0:cell>Citrate</ns0:cell><ns0:cell>Gelatin</ns0:cell><ns0:cell>Starch</ns0:cell><ns0:cell>Indole</ns0:cell></ns0:row><ns0:row><ns0:cell>Tests a</ns0:cell><ns0:cell>reaction</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>lecithinase</ns0:cell><ns0:cell>reduction</ns0:cell><ns0:cell>proskauer</ns0:cell><ns0:cell>utilization</ns0:cell><ns0:cell>hydrolysis</ns0:cell><ns0:cell>hydrolysis</ns0:cell><ns0:cell>production</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus megaterium</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus subtilis</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus cereus</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus thuringiesis</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus lentus</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus mycoides</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus alvei</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus polymyxa</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus circulans</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Micrococcus roseus</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='3'>Starch hydrolysis rate (mm) b</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Halo zone (mm)</ns0:cell><ns0:cell /><ns0:cell cols='3'>Diameter of colony (mm)</ns0:cell><ns0:cell /><ns0:cell>SHR</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus megaterium</ns0:cell><ns0:cell /><ns0:cell>16</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>5.33</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus subtilis</ns0:cell><ns0:cell /><ns0:cell>10</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>5.0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus cereus</ns0:cell><ns0:cell /><ns0:cell>12</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>4.0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus thuringiesis</ns0:cell><ns0:cell /><ns0:cell>17</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>5.67</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus lentus</ns0:cell><ns0:cell /><ns0:cell>14</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3.5</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus mycoides</ns0:cell><ns0:cell /><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>2.0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus alvei</ns0:cell><ns0:cell /><ns0:cell>18</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>6.0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus polymyxa</ns0:cell><ns0:cell /><ns0:cell>7</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>1.4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacillus circulans</ns0:cell><ns0:cell /><ns0:cell>16</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3.2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Micrococcus roseus</ns0:cell><ns0:cell /><ns0:cell>10</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>3</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>4</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>5</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>6</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>7</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table ( 3</ns0:head><ns0:label>(</ns0:label><ns0:figDesc>): Purification profile of amylase produced from different Bacillus sp. isolates</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Isolates a</ns0:cell><ns0:cell>Bacillus</ns0:cell><ns0:cell>Bacillus</ns0:cell><ns0:cell>Bacillus</ns0:cell><ns0:cell>Bacillus</ns0:cell><ns0:cell>Bacillus</ns0:cell><ns0:cell>Bacillus</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Purification step b</ns0:cell><ns0:cell>megaterium</ns0:cell><ns0:cell>subtilis</ns0:cell><ns0:cell>cereus</ns0:cell><ns0:cell>thuringiesis</ns0:cell><ns0:cell>lentus</ns0:cell><ns0:cell>alvei</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>TP(mg/ml)</ns0:cell><ns0:cell>0.66</ns0:cell><ns0:cell>0.96</ns0:cell><ns0:cell>0.81</ns0:cell><ns0:cell>0.54</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>0.86</ns0:cell></ns0:row><ns0:row><ns0:cell>Crude</ns0:cell><ns0:cell>TA(U)</ns0:cell><ns0:cell>6630</ns0:cell><ns0:cell>5000</ns0:cell><ns0:cell>3932</ns0:cell><ns0:cell>7890</ns0:cell><ns0:cell>2780</ns0:cell><ns0:cell>10040</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>EA(U/ml)</ns0:cell><ns0:cell>33.15</ns0:cell><ns0:cell>25.0</ns0:cell><ns0:cell>19.66</ns0:cell><ns0:cell>39.45</ns0:cell><ns0:cell>13.9</ns0:cell><ns0:cell>50.2</ns0:cell></ns0:row><ns0:row><ns0:cell>Ammonium</ns0:cell><ns0:cell>TP(mg/ml)</ns0:cell><ns0:cell>0.58</ns0:cell><ns0:cell>0.88</ns0:cell><ns0:cell>0.75</ns0:cell><ns0:cell>0.39</ns0:cell><ns0:cell>0.79</ns0:cell><ns0:cell>0.77</ns0:cell></ns0:row><ns0:row><ns0:cell>sulfate</ns0:cell><ns0:cell>TA(U)</ns0:cell><ns0:cell>701.0</ns0:cell><ns0:cell>640</ns0:cell><ns0:cell>461.0</ns0:cell><ns0:cell>820</ns0:cell><ns0:cell>308</ns0:cell><ns0:cell>1029</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>EA(U/ml)</ns0:cell><ns0:cell>35.06</ns0:cell><ns0:cell>32.0</ns0:cell><ns0:cell>23.08</ns0:cell><ns0:cell>41.0</ns0:cell><ns0:cell>15.4</ns0:cell><ns0:cell>51.47</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>TP(mg/ml)</ns0:cell><ns0:cell>0.50</ns0:cell><ns0:cell>0.82</ns0:cell><ns0:cell>0.71</ns0:cell><ns0:cell>0.37</ns0:cell><ns0:cell>0.60</ns0:cell><ns0:cell>0.70</ns0:cell></ns0:row><ns0:row><ns0:cell>Dialysis</ns0:cell><ns0:cell>TA(U)</ns0:cell><ns0:cell>718</ns0:cell><ns0:cell>700</ns0:cell><ns0:cell>502</ns0:cell><ns0:cell>805</ns0:cell><ns0:cell>365</ns0:cell><ns0:cell>591</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>EA(U/ml)</ns0:cell><ns0:cell>35.9</ns0:cell><ns0:cell>35.0</ns0:cell><ns0:cell>25.12</ns0:cell><ns0:cell>40.26</ns0:cell><ns0:cell>18.26</ns0:cell><ns0:cell>29.56</ns0:cell></ns0:row><ns0:row><ns0:cell>Sephadex G-</ns0:cell><ns0:cell>TP(mg/ml)</ns0:cell><ns0:cell>0.45</ns0:cell><ns0:cell>0.76</ns0:cell><ns0:cell>0.64</ns0:cell><ns0:cell>0.29</ns0:cell><ns0:cell>0.48</ns0:cell><ns0:cell>0.59</ns0:cell></ns0:row><ns0:row><ns0:cell>200</ns0:cell><ns0:cell>TA(U)</ns0:cell><ns0:cell>800</ns0:cell><ns0:cell>760</ns0:cell><ns0:cell>559</ns0:cell><ns0:cell>886</ns0:cell><ns0:cell>456.9</ns0:cell><ns0:cell>960</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>EA(U/ml)</ns0:cell><ns0:cell>80.03</ns0:cell><ns0:cell>76.0</ns0:cell><ns0:cell>55.9</ns0:cell><ns0:cell>88.64</ns0:cell><ns0:cell>45.69</ns0:cell><ns0:cell>96.02</ns0:cell></ns0:row></ns0:table><ns0:note>a: Isolated Bacillus sp. selected for amylase purification according to SHR, b: Different purification steps of amylase purification, TP: total protein, TA: total activity, EA: enzyme activity. PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Comparing the antibacterial activity of Bacillus sp. and purified amylase enzyme to different standard antibiotics. Comparing the antimicrobial susceptibility of a group of standard antibiotics(CLSI, 2017) against five human pathogenic strains (control). R= Resistant, S= sensitive, I= intermediate, NA= Not applicable (for antibiotics that were not specific to the bacterial strains), b: antimicrobial activity of the isolated Bacillus sp., NI= No inhibition, c: antimicrobial activity of purified amylase from some isolated Bacillus sp., ABM = amylase purified from Bacillus</ns0:figDesc><ns0:table /><ns0:note>a: megaterium, ABS amylase purified from Bacillus subtilis, ABC = amylase purified from purified Bacillus cereus, ABT= amylase purified from Bacillus thuringiesis, ABL= amylase purified from Bacillus lentus, ABA= amylase purified from Bacillus alvei. Values expressed as mean&#177; SD. PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table ( 4</ns0:head><ns0:label>(</ns0:label><ns0:figDesc>): Comparing the antibacterial activity of Bacillus sp. and purified amylase enzyme to different standard antibiotics.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head /><ns0:label /><ns0:figDesc>Inhibition zone in mm b R= Resistant, S= sensitive, I= intermediate, NA= Not applicable (for antibiotics that were not specific to the bacterial strains), b: antimicrobial activity of the isolated Bacillus sp., NI= No inhibition, c: antimicrobial activity of purified amylase from</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Bacillus megaterium</ns0:cell><ns0:cell>21&#177;1.5</ns0:cell><ns0:cell>26&#177;1</ns0:cell><ns0:cell>36&#177;1</ns0:cell><ns0:cell>24&#177;1</ns0:cell><ns0:cell>31&#177;1</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus subtilis</ns0:cell><ns0:cell>12&#177;1</ns0:cell><ns0:cell>18&#177;2</ns0:cell><ns0:cell>21&#177;1</ns0:cell><ns0:cell>15&#177;1.5</ns0:cell><ns0:cell>20&#177;2</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus cereus</ns0:cell><ns0:cell>12&#177;1.2</ns0:cell><ns0:cell>21&#177;1</ns0:cell><ns0:cell>36&#177;1</ns0:cell><ns0:cell>18&#177;1</ns0:cell><ns0:cell>14&#177;1.5</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus thuringiesis</ns0:cell><ns0:cell>14&#177;1.5</ns0:cell><ns0:cell>21&#177;1</ns0:cell><ns0:cell>21&#177;1</ns0:cell><ns0:cell>22&#177;1</ns0:cell><ns0:cell>18&#177;0.6</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus lentus</ns0:cell><ns0:cell>22&#177;1.6</ns0:cell><ns0:cell>22&#177;1</ns0:cell><ns0:cell>24&#177;0.6</ns0:cell><ns0:cell>32&#177;1.5</ns0:cell><ns0:cell>31&#177;3</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus mycoides</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus alvei</ns0:cell><ns0:cell>34&#177;1.5</ns0:cell><ns0:cell>21&#177;1</ns0:cell><ns0:cell>39&#177;1.5</ns0:cell><ns0:cell>29&#177;2.5</ns0:cell><ns0:cell>48&#177;2</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus polymyxa</ns0:cell><ns0:cell>36&#177;2.5</ns0:cell><ns0:cell>17&#177;0.6</ns0:cell><ns0:cell>38&#177;1</ns0:cell><ns0:cell>17&#177;4</ns0:cell><ns0:cell>20&#177;0.6</ns0:cell></ns0:row><ns0:row><ns0:cell>Bacillus circulans</ns0:cell><ns0:cell>21&#177;1.5</ns0:cell><ns0:cell>23&#177;1</ns0:cell><ns0:cell>39&#177;1</ns0:cell><ns0:cell>32&#177;1</ns0:cell><ns0:cell>32&#177;2</ns0:cell></ns0:row><ns0:row><ns0:cell>Micrococcus roseus</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Inhibition zone in mm c</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ABM</ns0:cell><ns0:cell>7.2+0.3</ns0:cell><ns0:cell>7.5+0</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell>ABS</ns0:cell><ns0:cell>7.3+0.3</ns0:cell><ns0:cell>7+0</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell>ABC</ns0:cell><ns0:cell>7+0.3</ns0:cell><ns0:cell>7.3+0.3</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell>ABT</ns0:cell><ns0:cell>7.5+0</ns0:cell><ns0:cell>7.2+0.3</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell>ABL</ns0:cell><ns0:cell>7.3+0</ns0:cell><ns0:cell>7.5+0</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell>ABA</ns0:cell><ns0:cell>7.5+0.5</ns0:cell><ns0:cell>7.3+0.3</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell><ns0:cell>NI</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>a: Comparing the antimicrobial susceptibility of a group of standard antibiotics (CLSI, 2017) against five human pathogenic strains</ns0:cell></ns0:row><ns0:row><ns0:cell>(control).</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:03:46812:3:1:NEW 17 Sep 2020)</ns0:note></ns0:figure> </ns0:body> "

"South Valley University Faculty of Science South Valley University Botany and Microbiology Department Revision 3 Dear Prof. Chris Yeager The Academic Editor, PeerJ We thank the reviewers for their generous comments on the manuscript and we have edited the manuscript to address their concerns. In particular, all of the code we wrote is available and I have included multiple links throughout the paper to the appropriate code repositories. We believe that the manuscript is now suitable for publication in peer J. Editor comments (Chris Yeager) MINOR REVISIONS The listing of enzyme activities alongside discussion of the anti-biofilm efficacy of the various amylase extracts is sufficient to meet reviewer #1's concerns in the editors opinion. However, English throughout the manuscript still needs improvement so that clarity of the message is not lost on the readership. PeerJ does offer an English editing service (for a fee) if a strong English editor is not available to the research team. [# PeerJ Staff Note: The Academic Editor has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title) #] The manuscript has been completely edited by an English language service provided from PeerJ Staff. Word document with tracked changes is submitted and the clean manuscript as well. We checked the scientific meaning throughout. All my regards,,, Prof. WESAM SALEM Figure/Table Citation Citations must be organized, and cited for the first time, in ascending numerical order, meaning Figure 1 must always be cited first, Figure 2 must always be cited second, and so on. The same rule applies to Tables. In this case, Table 4 (line 249) is cited for the first time before Table 3 (line 265). The citations of both figures and tables were checked and corrected. All my regards,,, Prof. WESAM SALEM "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Background</ns0:head><ns0:p>Although integrated pest management (IPM) is essential for conservation agriculture, this method can be inadequate for severely infected fields. The ability to predict the potential occurrence of severe infestation of soil-borne disease would enable farmers to adopt suitable methods for high-risk areas, such as soil disinfestation, and apply other options for lower risk areas. Recently, researchers have used species distribution modeling (SDM) to predict the occurrence of target plant and animal species based on various environmental variables. In this study, we applied this technique to predict and map the occurrence probability of a soil-borne disease, Verticillium wilt, using cabbage as a case study.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods</ns0:head><ns0:p>A disease survey assessing the distribution of Verticillium wilt in cabbage fields in Tsumagoi village (central Honshu, Japan) was conducted two or three times annually from 1997 to 2013. Road density, elevation and topographic wetness index (TWI) were selected as explanatory variables for disease occurrence potential. A model of occurrence probability of Verticillium wilt was constructed using the MaxEnt software for SDM analysis. As the disease survey was mainly conducted in an agricultural area, the area was weighted as 'Bias Grid' and area except for agricultural area was set as the background.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Grids with disease occurrence showed a high degree of coincidence with those with a high probability of such occurrence. The highest contribution to the prediction of disease occurrence</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background Although integrated pest management (IPM) is essential for conservation agriculture, this method can be inadequate for severely infected fields. The ability to predict the potential occurrence of severe infestation of soil-borne disease would enable farmers to adopt suitable methods for high-risk areas, such as soil disinfestation, and apply other options for lower risk areas. Recently, researchers have used species distribution modeling (SDM) to predict the occurrence of target plant and animal species based on various environmental variables. In this study, we applied this technique to predict and map the occurrence probability of a soil-borne disease, Verticillium wilt, using cabbage as a case study.</ns0:p><ns0:p>Methods A disease survey assessing the distribution of Verticillium wilt in cabbage fields in Tsumagoi village (central Honshu, Japan) was conducted two or three times annually from 1997 to 2013. Road density, elevation and topographic wetness index (TWI) were selected as explanatory variables for disease occurrence potential. A model of occurrence probability of Verticillium wilt was constructed using the MaxEnt software for SDM analysis. As the disease survey was mainly conducted in an agricultural area, the area was weighted as 'Bias Grid' and area except for agricultural area was set as the background.</ns0:p><ns0:p>Results Grids with disease occurrence showed a high degree of coincidence with those with a high probability of such occurrence. The highest contribution to the prediction of disease occurrence was the variable road density at 97.1%, followed by TWI at 2.3%, and elevation at 0.5%. The highest permutation importance was road density at 93.0%, followed by TWI at 7.0%, while the variable elevation at 0.0%. The constructed model accurately represented the occurrence of Verticillium wilt of cabbage, as demonstrated by the high value, 0.861, of the area under curve average of the receiver operating characteristic curve. This method of predicting disease probability occurrence can help with disease monitoring in grids with high probability occurrence and inform farmers about the selection of control measures.</ns0:p></ns0:div> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Intensive agriculture has supported world food production since World War II <ns0:ref type='bibr' target='#b30'>(Moffatt 2020)</ns0:ref>. However, the sustainability of intensive agriculture is often questioned because of negative consequences such as soil and water pollution, risks to human health from pesticides and fertilizer, and water resource depletion by over-exploitation of water resources <ns0:ref type='bibr' target='#b35'>(Oosterbaan 1989;</ns0:ref><ns0:ref type='bibr' target='#b28'>Matson et al. 1997;</ns0:ref><ns0:ref type='bibr' target='#b52'>Tilman et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b0'>Antonio and Hernandeza 2006)</ns0:ref>. To address these problems, a shift to conservation agriculture is required <ns0:ref type='bibr' target='#b46'>(Reicosky 2003;</ns0:ref><ns0:ref type='bibr' target='#b18'>Kassam et al. 2009</ns0:ref>). On the other hand, measures to increase food supply to maintain food security are also needed because the world population is expected to reach over 9.7 billion by 2050 (United Nations 2015). These competing demands for sustainability and higher yields presents a complex problem facing modern intensive agriculture <ns0:ref type='bibr' target='#b5'>(Brussaard et al. 2010</ns0:ref>), but it is a challenge that must be met.</ns0:p><ns0:p>To cope with the ever growing global food demand, intensive agriculture primarily revolves around large-scale monoculture, the continuous or consecutive growth of the same crop over large areas <ns0:ref type='bibr' target='#b6'>(Cook and Weller, 2004)</ns0:ref>. Monocultures carry a heavy risk of soil-borne disease because pathogens have a continuous supply of host plant and are thus able to persist or even accumulate in the soil <ns0:ref type='bibr' target='#b33'>(Newton et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b17'>Jenking et al. 2010)</ns0:ref>. Soil disinfestation by fumigation with chemical control agents can be effective in the management of soil-borne disease <ns0:ref type='bibr' target='#b58'>(Wilhelm and Paulus 1980;</ns0:ref><ns0:ref type='bibr' target='#b20'>Koike et al. 2003)</ns0:ref>, but also carries the risk of negative impacts on the environment and human health <ns0:ref type='bibr' target='#b0'>(Antonio and Hernandeza 2006;</ns0:ref><ns0:ref type='bibr' target='#b47'>Sande et al. 2011</ns0:ref>; United States Environmental Protection Agency 2017). In addition, the cost of applying chemical control to vast farmland areas is enormous <ns0:ref type='bibr'>(Ladis 1987;</ns0:ref><ns0:ref type='bibr' target='#b23'>Labrada et al;</ns0:ref><ns0:ref type='bibr'>2001, Koike et al;</ns0:ref><ns0:ref type='bibr' target='#b10'>2006)</ns0:ref>. So, although effective, chemical disinfestation to reduce soil-borne disease can be difficult to adopt when the risks and costs cannot be contained within acceptable levels.</ns0:p><ns0:p>Integrated pest management (IPM), which is defined as the long-term prevention of pests or their damage through a combination of techniques, is essential for sustainable agriculture <ns0:ref type='bibr' target='#b1'>(Apple et al. 1976</ns0:ref><ns0:ref type='bibr' target='#b55'>, UC IPM 2020)</ns0:ref>. The control of soil-borne disease in IPM typically involves the application of resistant cultivars, crop rotation, exclusion and prevention of pathogen's inoculum source <ns0:ref type='bibr' target='#b54'>(Tsushima 2014;</ns0:ref><ns0:ref type='bibr' target='#b15'>Ikeda et al. 2015)</ns0:ref>. However, such methods are inadequate to bring about control in severely infested fields. Therefore, intensive continuous soil fumigation may be justified in severely infested fields. If the potential occurrence of severe infection of soil-borne disease could be predicted, farmers could adopt soil fumigation for high-risk areas and apply other options for areas of lower risk. Furthermore, evaluating infestation risk in a giving area can help growers to prevent the spread to unaffected fields or re-introduction in previously controlled fields. This approach is in accordance with the IPM framework for sustainable, lowenvironmental impact, and cost-effective agriculture <ns0:ref type='bibr' target='#b53'>(Tsushima and Yoshida 2012)</ns0:ref>. Control measures for soil-borne diseases should be selected according to the disease potential occurrence of each field. However, the development of a practical method to predict such disease potential occurrence has not previously been successful in the field of plant pathology.</ns0:p><ns0:p>Ecological science uses species distribution modeling (SDM) (e.g. <ns0:ref type='bibr' target='#b38'>Osawa 2011;</ns0:ref><ns0:ref type='bibr' target='#b37'>Osawa 2015)</ns0:ref> for predicting the probability of occurrence and sometimes also abundance of target plant and animal species based on environmental variables such as terrain and climate factors <ns0:ref type='bibr' target='#b13'>(Guisan and Zimmermann 2000;</ns0:ref><ns0:ref type='bibr' target='#b50'>Sober&#243;n and Peterson 2005;</ns0:ref><ns0:ref type='bibr' target='#b49'>Sober&#243;n J. 2007;</ns0:ref><ns0:ref type='bibr' target='#b41'>Peterson et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b40'>Peterson 2014)</ns0:ref>. SDM can predict the distribution of a target species using survey data obtained in the area. Additionally, the respective contribution, or importance, of environmental variables on species distribution can be calculated. Maximum entropy algorithm is one of the most famous SDM approach and can be used to predict species potential distribution by analyzing only presence records with environmental variables <ns0:ref type='bibr' target='#b42'>(Phillips et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b43'>Phillips et al. 2006)</ns0:ref>. MaxEnt software has been successfully used in predicting potential species distribution <ns0:ref type='bibr' target='#b36'>(Ortega-Huerta et al. 2008)</ns0:ref>. Recently, MaxEnt has been used for a prediction of plant disease distribution <ns0:ref type='bibr' target='#b7'>(Cunniffe et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b26'>Macedo et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b31'>Narouei-Khandan et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Berthon et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b57'>Wang et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b32'>Narouei-Khandan et al., 2020)</ns0:ref>. MaxEnt can predict plant disease potential, essential information for farmers to enact IPM. In the present study, we used MaxEnt to predict the occurrence probability of Verticillium wilt, a soil-borne disease, as a case study on cabbage field in Japan.</ns0:p><ns0:p>Tsumagoi village in the Gunma prefecture of Japan intensively produces cabbage, as one of the main production sites for this crop in Japan. This is an ideal site to test the application of the SDM technique, given its large-scale area of consolidated monoculture and years of field monitoring, which enable a detailed analysis of the occurrence of Verticillium wilt. Verticillium wilt in cabbage has been a big problem in this region's cabbage production since 1994 <ns0:ref type='bibr' target='#b48'>(Shiraishi et al., 2000)</ns0:ref>. This is a soil-borne disease caused by Verticillium dahliae and V. longisporum <ns0:ref type='bibr' target='#b1'>(Banno et al., 2011)</ns0:ref>. The cabbage farmers in Tsumagoi village have suffered severe economic losses from this disease.</ns0:p><ns0:p>The objective of the study was to construct an occurrence probability map of Verticillium wilt of cabbage in Tsumagoi village in Gunma prefecture in Japan to test the application of SDM to soilborne plant disease. The findings aim to inform farmers to choose adequate control measures, such as introducing resistant cultivars or changing cultivation period. This trial can contribute to the development of IPM by offering both environmental and economic improvements to soilborne disease management in agricultural production.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Study area</ns0:head><ns0:p>The distribution of Verticillium wilt of cabbage was surveyed in Tsumagoi village (36&#176;51'17'N, 138&#176;53'00'E) in central Honshu, the largest of Japan's four main islands (Fig. <ns0:ref type='figure'>1</ns0:ref>). The climate is subarctic humid climate (Dfb) in K&#246;ppen climate classification with an elevation ranging between 700 and 1500 m above sea level and a mean annual temperature of 7.2 &#176;C (Japan Meteorological Agency 2017, Village Office of Tsumagoi 2017). Cabbages are produced on approximately 2790 ha of the 3200 ha agricultural area. This is a remarkably large degree of monoculture for a horticultural area of this size in Japan. Thus, the village is renowned in Japan as a production site for cabbage. Cabbages are harvested there from April (early spring) to October (autumn). However, potato, maize and scarlet runner bean are also grown in this area, in significantly lower proportions than those of cabbage.</ns0:p></ns0:div> <ns0:div><ns0:head>Target disease and disease survey</ns0:head><ns0:p>Cabbage production in the area has suffered from Verticillium wilt since 1994 <ns0:ref type='bibr' target='#b48'>(Shiraishi et al. 2000)</ns0:ref>. Verticillium wilt is a soil-borne disease associated with Verticillium dahliae and V. longisporum <ns0:ref type='bibr' target='#b1'>(Banno et al. 2011)</ns0:ref>, which can survive for years using microsclerotia as a resting structure. V. dahliae has a board range of hosts, mainly plants from the Solanaceae and Brassicaceae families <ns0:ref type='bibr' target='#b39'>(Pegg and Brady, 2002)</ns0:ref>, while V. longisporum infects mainly members of the Brassicaceae family <ns0:ref type='bibr' target='#b14'>(Inderbitzin et al., 2013)</ns0:ref>. In our study site, V. dahliae was found to be the dominant species causing Vertillium wilt in cabbage <ns0:ref type='bibr' target='#b2'>(Banno et al., 2015)</ns0:ref>. In recent years, resistant to Verticillium wilt have been planted in the village. The disease survey was conducted two or three times per year every year from 1997 to 2013, 38 times in total. In every survey, official and private agricultural extension workers and plant pathologists capable of distinguish the disease all fields of the village. This survey and field collections was approved by Gunma Prefectural Office under permission document no. H28.114.30. Disease occurrence was identified by external symptoms such as outer leaf yellowing and wilting (Fig. <ns0:ref type='figure'>2A, B</ns0:ref>). When a suspect field was found, the observers examined the vascular system of selected cabbages to identify browning and thus confirm the existence of the disease (Fig. <ns0:ref type='figure'>2C</ns0:ref>). The locations of fields where the disease occurred in field survey were marked on a 1:5000 map. When new records were found, investigators took cabbage samples for isolation and identification of the pathogens. The study area was divided into a grid of 500 &#215; 500 m cells. The grid cell size was selected according to both the degree of accuracy of the disease survey and the scale of land ownership. Disease occurrence (a binary variable) of both pathogen species and the values of the selected explanatory variables were recorded for each grid cell (Fig. <ns0:ref type='figure'>3A, B</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Explanatory variables</ns0:head><ns0:p>Geographic Information System software ArcMap (ver. 10.7.1, ESRI, Redlands, CA, USA) was used for analysis of the data set. We selected road density, elevation and topographic wetness index (TWI) as possible explanatory variables that may affect the disease potential occurrence. From the beginning, the disease had been thought that the disease spread from one field to another by the infested soil because the pathogen was soil-borne, hence we selected the variable road density. Additionally, cabbage field in our study site from 700 to 1500 m above sea level. Given the relationship between elevation and temperature, we included this as one of the variables in the model. Finally, we also selected TWI, which indicates the degree of soil moisture <ns0:ref type='bibr' target='#b12'>(Gallant 2000)</ns0:ref>, the latter is known influence the development of soil-borne disease like Verticillium wilt <ns0:ref type='bibr' target='#b39'>(Pegg and Brady, 2002)</ns0:ref>. Road data were downloaded as polyline data from Fundamental Geospatial Data (Geographical Survey Institute; http://www.gsi.go.jp/kiban/, accessed on 24 June 2020). The number of new road had not increased dramatically from 2013 to 2016 in the study area. The total length of road in each grid cell was calculated from the polyline data as 'road density.' Elevation and slope were downloaded from the National Land Numerical Information download service (Ministry of Land, Infrastructure, Transport and Tourism, http://nlftp.mlit.go.jp/ksj/, accessed on 24 June 2020). Slope data are required to calculate TWI <ns0:ref type='bibr' target='#b12'>(Gallant 2000)</ns0:ref>, which is defined as follows:</ns0:p><ns0:formula xml:id='formula_0'>TWI = ln &#119886;</ns0:formula><ns0:p>tan &#119887; where a is local catchment area (i.e., the local upslope area draining per unit) and b is the local slope. Values for elevation and TWI were applied to each grid along with the road density variable. The agricultural area shown in Fig. <ns0:ref type='figure'>3A</ns0:ref> were also downloaded from the National Land Numerical Information download service described above. There were 1392 grids cells in our study area, we confirmed disease occurrence in 194 grid cells (Fig. <ns0:ref type='figure'>3B</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Model construction</ns0:head><ns0:p>The model of probability of disease occurrence was constructed SDM software MaxEnt ver. 3.3.3k (downloaded at: https://biodiversityinformatics.amnh.org/open_source/maxent/) <ns0:ref type='bibr' target='#b42'>(Phillips et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b43'>Phillips et al. 2006)</ns0:ref>. This software was chose as it can analyze presence-only data efficiently with a low number of occurrences <ns0:ref type='bibr'>(Elith 2006)</ns0:ref>. MaxEnt is used for predicting the probability of presence of target species, but here we applied it to the prediction of probability of soil-borne disease occurrence. The disease occurrence probability (0-1) in each grid cell, percent contribution and permutation importance of each explanatory variables were calculated by MaxEnt. We conducted the analysis using the whole village area, given that this was the scale relevant for our study. However, the original disease survey was mainly conducted within the agricultural areas (Figure <ns0:ref type='figure'>3A</ns0:ref>), hence we used a 'Bias Grid' weighting and set the remaining 'non-agricultural area' as background in MaxEnt setting. To avoid overfitting, only 'Liner' and 'Quadractic' were on and the value for 'Regularization multiplier' was changed '1' to '2' in MaxEnt setting <ns0:ref type='bibr' target='#b29'>(Merow et al., 2013;</ns0:ref><ns0:ref type='bibr'>Radosavlijevic and Adnderson, 2014;</ns0:ref><ns0:ref type='bibr' target='#b51'>Syfert et al., 2013)</ns0:ref>.</ns0:p><ns0:p>Cross-validation was selected in 'Replicated run type' and repeated 20 times. The model of the probability of disease occurrence was also estimated by means of receiver operating characteristic (ROC) analysis. Furthermore, correctly classified instances (CCI), sensitivity, specificity, true skill statistics (TSS) were calculated based on the 'threshold by maximum training sensitivity plus specificity' in MaxEnt.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The SDM model constructed with MaxEnt was used to map the probability of disease occurrence and points of disease occurrence (Fig. <ns0:ref type='figure'>4</ns0:ref>). Threshold by Maximum training sensitivity plus specificity from MaxEnt analysis was 0.543. The disease occurrence probability map was created based on this threshold. Grids in this map were colored when the grids had a higher probability compared to the threshold (in total, 556 / 1392 grids in this study area). The grids with disease occurrence had a high degree of coincidence with those with a high probability of such occurrence.</ns0:p><ns0:p>Percent contribution and permutation importance of predictor variables are shown in Table <ns0:ref type='table'>1</ns0:ref>. The highest contribution to the prediction was road density at 97.1%, followed by TWI at 2.3%, and elevation at 0.5%. Permutation importance was as follow: road density at 93.0%, TWI at 7.0%, and elevation at 0.0%. Both percent contribution and permutation importance indicated that road density was the best predictor variable: disease probability occurrence was highest when road density was about 4000 m/grid (Fig. <ns0:ref type='figure'>5</ns0:ref>).</ns0:p><ns0:p>Estimation indices for accuracy of MaxEnt modeling of cabbage Verticillium wilt was shown in Table <ns0:ref type='table'>2</ns0:ref>. The constructed model accurately represented the occurrence of Verticillium wilt of cabbage, as demonstrated by the high value, 0.861, of the AUC average of the ROC. The calculated values for CII, sensitivity, specificity, and TSS using the threshold were 0.705, 0.876, 0.678 and 0.544, respectively.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In this study, we applied SDM to predict the occurrence probability of Verticillium wilt in cabbage as a case study. Our model provided good explanation of the occurrence of Verticillium wilt of cabbage in Tsumagoi village. To the best of our knowledge, this is the first time SDM has been used to estimate occurrence probability of soil-borne disease. Our findings indicated that this method can successfully predict disease occurrence probability using survey data and selected environmental variables in a monoculture area.</ns0:p><ns0:p>Our result showed that road density significantly influenced disease occurrence probability (Table <ns0:ref type='table'>1</ns0:ref>), with the highest probability of disease at around 4000 m/grid (Fig. <ns0:ref type='figure'>5</ns0:ref>). <ns0:ref type='bibr' target='#b34'>Numminen and Laine (2020)</ns0:ref> showed that road networks played an important role in the spread of plant diseases. This result was consistent with the previous report of pathogens spreading through road networks. It was thought that the pathogen was locally transmitted between fields through the soil, as they are soil-borne. Soil can be transferred from infested fields to road surfaces on the wheels of tractors or other vehicles and even the footwear of farmworkers. Additionally, Tsumagoi village is located in mountainous area with steep slopes, which facilitates soil transfer between fields (e.g., by wind and water). Both natural and human-mediated processes can influence the spread of Verticillium wilt in our study area.</ns0:p><ns0:p>On the other hand, our model suggested that the influence of TWI was quite low. The relationship between Verticillium wilt in cabbage and soil moisture had not yet been investigated before this study; however, it was reported that soil moisture was not significant factor for disease development in oilseed rape, Brassicaceae family <ns0:ref type='bibr' target='#b22'>(Kn&#252;fer, 2013)</ns0:ref>. Consequently, the low impact shown by TWI on disease occurrence probability concurs with previous research, although more research is required to conclusively rule out this relationship. Elevation also showed a very low contribution to the probability of disease occurrence. Given that the optimal temperature for cabbage Verticillium wilt was between 19&#176;C and 23&#176;C, we predicted that temperature during the harvest period would have a strong influence on disease occurrence <ns0:ref type='bibr' target='#b19'>(Kemmochi et al 2001)</ns0:ref>. Cabbage is normally cultivated temperature between 4&#176;C and 24&#176;C <ns0:ref type='bibr' target='#b4'>(Bradley and Courtier, 2006)</ns0:ref>. Cabbage cultivation was conducted following temperature change in each field, despite decreasing temperatures with increasing elevation. Given our results, the elevation factor might not influence the probability of disease occurrence in our study area.</ns0:p><ns0:p>Our model was evaluated using estimation indices based on the defined threshold (Table <ns0:ref type='table'>2</ns0:ref>) and the disease occurrence probability map (Fig. <ns0:ref type='figure'>4</ns0:ref>). It appeared that many grids where disease occurrence probability was recorded had a high probability. The AUC, CCI, sensitivity, specificity and TSS obtained from our analysis were all sufficiently high, indicated that our model was accurate in its evaluation. Importantly, the sensitivity should be regarded as the most important for our objective. Because sensitivity in our study meant that how exactly the model could predict the disease occurrence grids. The sensitivity of the model constructed was shown to be able to predict the disease occurrence grids at about 88% accuracy. As a result, the model constructed using MaxEnt was sufficiently effective for prediction of the disease occurrence. Some Unifested grid cells where in which Verticillium wilt of cabbage had never been recorded had were with a high probability of disease occurrence, suggesting that the disease is likely to develop in these high probability grids in the near future. Our findings indicate that road density was the most significant predictor of disease occurrence (Table <ns0:ref type='table'>1</ns0:ref>). Despite a level indicating the potential that road density is high in, the following reason is thought about not becoming the disease, 1) the pathogen had not yet invaded that a grid, 2) there were no or almost no cabbage fields, 3) there were unknown factors suppressing disease occurrence. In other factors expected for using this study, it might have the possibility that the factors would become novel control measures and elucidation of the disease life cycle.</ns0:p><ns0:p>The disease occurrence probability map helps disease monitoring in high probability grids and selective practical control measures, such as fumigation, planting disease-resistant cultivars <ns0:ref type='bibr' target='#b19'>(Kemmochi et al. 2001)</ns0:ref>, and changing the timing of cabbage cultivation. Although it is not possible to change the road density, it is possible to more apply soil erosion and sediment transport control technologies such as the establishment of cover crops <ns0:ref type='bibr' target='#b25'>(Lawson et al. 2015)</ns0:ref> to decrease the potential occurrence of the disease, especially in those grid cells where there are preexisting records. In addition, we recommend applying methods to the farmers and extension workers for reducing spread potential, such as portable vehicle washing equipment and changing footwear at the site are required.</ns0:p><ns0:p>This study shows that SDM analysis using MaxEnt can be used to predict the occurrence and spread of plant pathogens using survey data and environmental variables with a high level of accuracy. Our study showed that this model could be applied to area under intensive crop cultivation in other area if accurate long-term disease survey data and GIS data of potential influencing environmental variables are available.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In this study, we applied SDM to predict the occurrence probability of Verticillium wilt of cabbage, soil-borne plant disease. The SDM model constructed from three explanatory variables provided good prediction of the occurrence of Verticillium wilt of cabbage in Tsumagoi village, Japan. The model developed from the field survey data showed that road density played important roles in the occurrence of the disease. The following three points are particularly important and the highlight novel results of this study: 1) we find a measure to predict and visualize the potential occurrence of soil-borne disease. 2) predicting the potential occurrence was conducted using SDM : and 3) by applying geographical information system (GIS), we could visualize the distribution of the potential occurrence. Predicting the probability of disease occurrence can help farmers to select suitable control methods for high and low-risk areas. That soil-borne disease occurrence could be predicted by survey data and environmental variables without plant pathological analysis is a novel finding.</ns0:p></ns0:div><ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='14,42.52,178.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='15,42.52,204.37,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='16,42.52,331.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,331.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,224.62,525.00,363.00' type='bitmap' /></ns0:figure> </ns0:body> "

"Editor comments (Victoria Sosa) The four reviewers that checked your manuscript had conflicting opinions on your paper. I suggest that you consider all of them. Many of the issues raised by reviewers were in relation to methods and suggested additional analyses for improving results. Others are related to sampling and more detail is needed to understand results. I think this is novel research that uses SDMs to predict disease occurrence, thus I recommend taking carefully every issue by reviewers. Res: Thank you for your positive comments on our manuscript. We performed re-analysis with MaxEnt in the revised version of our manuscript, for reducing survey bias and overfitting. We have also provided more detail regarding the sampling design, analytical procedure, and the obtained results. The manuscript was revised according to the suggestions provided by the 4 reviewers. We have also revised the URLs in this manuscript. ****************************************** Reviewer 1 (Anonymous) Thank you for the helpful comments and suggestions. We have revised our manuscript according to your suggestion. We have re-analyzed the new parameters and evaluated the accuracy of the model using the estimation indices. Our responses to your comments are appended hereafter. <Basic reporting> 1) A major shortcoming of this study is that it is a case study. While this is not a major problem, it is not mentioned anywhere in the manuscript, and all analyses proceeded as if the study could be replicated. Nor is this issue much discussed in the Discussion. While lack of replication is often unavoidable in SDMs studies at a low spatial scale particularly using survey data, this has to be reported as a limitation of the study. Res: As per your suggestion, we mentioned that this study is a case study in the Introduction and Discussion sections (L24-26, L104-106, and L230-231). 2) The abstract is not well balanced. The authors present too much background and just a few details on the results, mentioning mainly how the model was evaluated. Here the authors need to highlight their findings and the relevance of their study. Res: We revised the abstract as per your suggestion, with special emphasis on the background section, to ensure that the abstract appear balanced and concise. The abstract was revised in accordance with the suggestions provided by you and the other reviewers. 3) The introduction reads well and has all the important topics. Although, I suggest adding more information regarding the uses and importance of SDMs. Here, it is important that the authors differentiate between MaxEnt the algorithm and the software. Res: We have revised this and provided more details regarding the importance and significance of SDM (L95-97). We have also clearly differentiated between the MaxEnt algorithm and software (L32-33, L97, and L99-100). 4) Material and Methods need more information. For example, authors need to include a justification of the selected variables. A simple description is given for TWI in L157 (A high TWI indicates that a site is moist) but it is not enough. Why the authors selected those variables? How do they influence Verticillium wilt occurrence? Because there is no justification for variable selection, the information on variable contribution feels weak later in the discussion. Also, I am curious to know why no climate predictors were used to model potential occurrence. Res: The reason why the 3 explanatory variables were selected have been described in L165-171. We did not use any climate variables because the temperature was strongly correlated with the elevation, which we also used as a variable. Also there was no large variation in precipitation as Tsumagoi Village has an area of only 337.58 km2. The TWI and slope were strongly correlated as the slope was used for calculating the TWI. Therefore, the slope was removed during model reconstruction. 5) In L190, the authors provide the percent contribution of each explanatory variable. I suggest to report instead, or at least include also in the manuscript, the permutation importance of each variable, as this parameter is more relevant and informative within MaxEnt outputs. Res：The percent contribution and the permutation importance are enlisted in Table 1 (L216). 6) L192-193. The authors must justify why they used Maxent's default setting. This point is very important since the authors are basing their results with this configuration without trying to optimize the model. There are many articles that have explored this problem (see Anderson and Gonzalez 2012; Warren and Seifert 2011; Merow et al. 2013; Radosavljevic and Anderson 2013; Syfert et al. 2013; Halvorsen et al. 2016). I would have expected that at the very least and to increase the quality of the models, the ‘hinge’ and ‘threshold’ features would have been deactivated to avoid overfitting response curves. Res: The default setting of MaxEnt were altered for correct evaluation of the survey bias and for avoiding overfitting. The changes in the setting are mentioned in the Materials and Methods section (L195-200). 7) Regarding the evaluation of model performance, the authors only used the AUC value, which although it is high, does not show much, only that the model is not bad. Getting high values of AUC is very common for models based on presence-only data such as the one developed in this work. If the model is evaluated with cross-validation (for example, using spatial blocks, as in Roberts et al. 2017, http://onlinelibrary.wiley.com/doi/10.1111/ecog.02881/full), a high value of AUC would be more convincing. AUC levels for good performance are only relevant for presence/absence data. Since the authors base all the performance of the model on this statistic, it is necessary to justify it, especially since problems with AUC are well documented (see Lobo et al. 2007). Another option would be to calculate an additional statistic, such as ‘true skill statistics (TSS)’, to corroborate model performance. Res: The AUC was obtained by cross-validation, using 20 replicated runs. Additionally, the CCI, sensitivity, specificity, and TSS were calculated from the threshold maximum training sensitivity plus specificity using MaxEnt (Table 2, L202-204, and L223-227). 8) How was the 'background' selected for the model? There is no mention of this anywhere in Methods and it is important to describe it. Res: The disease survey was mainly conducted in the agricultural area (Fig3A). The agricultural area was therefore weighted as a “Bias Grid”, and the area except the agricultural area comprised the background in MaxEnt (L195-200). 9) As the results are just one paragraph, I suggest combining these with the Discussion. Although, I am not sure if the journal accepts a section combining “Results & Discussion”. Res: The permutation importance and the model evaluation indices are provided in the Tables and additionally described in the Results section (L216-221). 10) In the results, I don’t understand why the authors did not project their model to the whole agricultural area in Tsumagoi village (i.e. all grid cells in Figure 3A, including outside the agricultural area). This would be easy to do and informative considering agricultural expansion. This will provide information on which areas are more vulnerable to the disease in potential new locations. Res: As depicted in Fig. 3A, we wanted to predict the area, barring the agricultural region that had been stored by a national survey in 2010. Therefore, the MaxEnt analysis was conducted for the entire area of the village (L194-195). 11) In L223-224 the authors say 'Some grid cells where Verticillium wilt of cabbage had never been recorded had a high probability of disease occurrence'; here, the authors need to discuss these findings. This is very important, is it because these grid cells have the optimal conditions for the occurrence of the disease? This is the place to speculate and discuss in more detail the response of the disease to the selected variables and the conditions across the landscape. Res: This has been discussed in L272-280. 12) The authors also need to provide a better discussion on the contribution to the prediction for the variables. Specifically, including why TWI (7.7%) and slope (7.1%) had the lowest values. This has to be connected to the justification of the variable selection (see comment above) Res: The contribution of the variables has been mentioned in L237-260. 13) There’s no need to have two panels in Figure 4. Just present Figure 4B. Res: Figure 4B was deleted as per your suggestion. 14) There’s no need to include in figure 3 and 4 legends 'These maps in this figure were arranged by ArcMap (ver. 10.2.2, ESRI, Redlands, CA, USA)' and 'The occurrence probability calculated in each grid was arranged by ArcMap', respectively. Res: We deleted this sentence from the legends to Figures 3 and 4, as per your suggestion. Additionally, all revisions suggested by you has been directly revised in the main text with no comments. ********************************** Reviewer 1 (Anonymous) Comments in main text <Abstract> L1 Deleted: in Res: This was deleted as per your suggestion. L39 Commented [ME1]: Algorithm? Res: We have clearly differentiated between the MaxEnt algorithm and software (L33, L97, and L99-100). L45 Commented [ME2]: I wouldn’t give such a big importance to this value in your results in the abstract. This number only indicates that your model is mathematical good but it is more important to assess model performance for its capacity to spatially predict suitable habitat based on occurrences Res: We have provided more details of the results (L38-46). L52：Commented [ME3]: Relative importance? Res: This was deleted when the abstract was revised. <Introduction> Commented [ME4]: Reference? Res: We have provided the corresponding reference (L50). Commented [ME5]: To connect with the previous paragraph, perhaps you could start this paragraph mentioning that one of the consequences to keep the growing food demand is the intensive agriculture…. Commented [ME6]: Reference? Res: The corresponding references were added, and the sentence was revised as per your suggestion (L62-64). Commented [ME7]: This is quite ambiguous. Provide a number with the reverence to have a better context of the economic implications Res: The corresponding references have been provided (L70-71). Commented [ME8]: Sustainable? Res: This has been changed in L76. Commented [ME9]: What protocol? You only mentioned one potential tool (i.e. prediction of infestation) Res: This has been changed to 'approach' (L85). Commented [ME10]: Do you mean disease potential occurrence at each field? Res: Yes, it means 'potential of disease occurrence'. This term has been mentioned in L86-88. Commented [ME11]: Same as above, potential occurrence Res: This has been revised in L88-89. Commented [ME12]: Recommend deleting this Res: We agree with your suggestion and have deleted the sentence accordingly. Commented [ME13]: I recommend describing for your readers in more detail how SDMs work and how they predict probability of occurrence Res: We have described these in details, with explanations, as suggested (L95-99). Commented [ME14]: Most used? Res: Changed to “most famous” (L97). Commented [ME15]: Is it none or few? You need to make a better background search. If there are studies then add “but see [add references]” Res: Thank you for your suggestion. We apologize for not providing sufficient background information. We have included some relevant references as per your suggestion (L101-103). Commented [ME16]: This reads odd, perhaps change to: “we used MaxEnt to predict the occurrence probability of a soil-borne disease, Verticillium wilt of cabbage, using disease field survey data” Res: This has been revised as per your suggestion (L104-106). Commented [ME17]: Strongly suggest changing this to: “This trial can contribute to the development…… Res: This has been revised as per your suggestion (L120-122). Commented [ME18]: Explain for your readers some characteristics of this disease and what the implications are. Why is important to model this disease? Res: The characteristics and implications of this disease has been described in detail in the Introduction section (L108-115). Commented [ME19]: Include total number Res: The total number has been mentioned as suggested (L146-147). Commented [ME20]: Perhaps simplify to: “When new records were found, investigators” Res: This has been revised as per your suggestion (L154-155). Commented [ME21]: Does this figure relate to the locations of the disease mentioned in lines 145-146? (The locations of fields where the disease occurred were marked on a 1:5000 map). If yes, you need to cite this figure in those lines Res: We apologized for the confusion. The map that has been provide here is intended for field survey. We have explained this in L156-159. Commented [ME22]: Why is this important to model potential occurrence? Res: We have explained the variable selection in L165-171. Commented [ME23]: How much could these data have changed between 2016 (when data were created) and 1994-2913 (the survey period)? Res: The number of road have not increased dramatically from 2013 to 2016 in the village. We have explained this in L173-174. Commented [ME24]: Provide details on this calculation. Formula? Res: We have provided the formula that was used for calculating TWI (L177-181). Commented [ME25]: What are these data? Provide more details. Res: These data refer to the agricultural area in the village (Fig. 3A). This has been revised as per your suggestion (L182-183). Commented [ME26]: Change to: “We selected MaxEnt because is a presence-only data that works efficiently with low number of occurrences (Elith 2006).” Res: This has been revised as per your suggestion (L189-190). Commented [ME27]: I supposed you try to say that model performance was estimated by calculating the average test AUC (the area under the receiver operating characteristic curve; ROC) Res: The AUC average at cross-validation has been mentioned in the Results section (L225). Commented [ME28]: Was this analysis done through fold cross-validation? How many? Provide more details Res: The details of the setting used for the MaxEnt analysis has been provided as per your suggestion (L195-200). Commented [ME29]: You may want to include in Methods the thresholds for considering AUC indicative of good model performance Res: The threshold and the other estimated are mentioned in the Results section (L225-227) and Table2. Commented [ME30]: This sentence is not clear. I don’t understand what “distribution rate of agricultural area” mean. Please, clarify Commented [ME31]: Same as above, it is not clear how this conclusion was made Res: These sentences were deleted because the contribution of each variable was altered by re-analysis, which was performed using user-defined settings. Commented [ME32]: Is it because these grid cells have the optimal conditions for the occurrence of the disease? Res: We have provided a description of the grid cells in L272-280. Commented [ME33]: Don’t you think monitoring is the　first measure to consider in those sites? Res: Yes, we agree with your suggestion, and have mentioned 'monitoring' in the sentence (L282). Commented [ME34]: Not sure about using this word here Res: This has been deleted as per your suggestion. Commented [ME35]: Is this feasible for the land owners? Res: This is not feasible the land owners, such as the extension workers. This has been revised accordingly (L288-290). Commented [ME36]: This is not discussion. It is part of the conclusions. Merge with the paragraph below and remove repetitive information Res: The paragraph was merged with conclusion as suggested (L309-311). Commented [ME37]: What do you mean by “other options”? Clarify Res: The 'other options' refer to the control measures that were used for the fields with low infestations, such as the fields of resistant cultivars and the fields where the cultivation periods are altered. These have been mentioned in L308-309. Commented [ME38]: Change to: “Predicting potential occurrence of severe infestation of soil-borne disease can help farmers to select more suitable ways for high-risk areas identification and apply other options for areas of lower risk” Res: Thank you for your suggestion. This has been revised as suggested (L308-309). *************************************************** Reviewer 2 (Anonymous) Thank you for the helpful comments and suggestions. We have revised our manuscript according to your suggestions. We re-analyzed using new parameters and evaluated the accuracy of the model using the estimation indices. The response to your comments are appended hereafter. <Basic reporting> Literature reference is not sufficient, because different sections such as Introduction and Discussion are without a background of the disease, niche theory and on species distribution models. All results are inconsistent, since they were obtained of a weak and unclear modelling approach and a limited data set. A manuscript should show a relevant objective. Species distributions models are not only maps. You need to improve this manuscript a lot with all suggestions included in this document. Res: Thank you for your suggestions. We agree that our manuscript lacked suitable references and that a more detailed description of the disease background was necessary. We have revised these accordingly. We have provided a description of the SDM in the Introduction and Discussion section. The modeling approach has also been revised. The contribution of the explanatory variables has been discussed in the revised version of the manuscript. 75-78 1) Integrated pest management (IPM), which is defined as the long-term prevention of pests or their damage through a combination of techniques, is essential for conservation agriculture (Apple et al. 1976). The control of soil-borne disease in IPM typically involves the application of resistant cultivars and crop rotation (Tsushima 2014; Ikeda et al. 2015). R: In general, you can refer separately to pest and disease in a scientific language. Integrated pest management (IPM) is associate with pests, but it is not a great problem. Also, you should to read and include recent references on this aspect. Res: The term 'pest' in 'Integrated pest management (IPM)' does not refer to insect pests. The definition of the term “pest” according to the IPM website (available at: https://www2.ipm.ucanr.edu/What-is-IPM/#DEFINITION) is as follows: 'A pest can be a plant (weed), vertebrate (bird, rodent, or other mammal), invertebrate (insect, tick, mite, or snail), nematode, pathogen (bacteria, virus, or fungus) that causes disease, or other unwanted organism that may harm water quality, animal life, or other parts of the ecosystem.' We have cited this definition in the text and provide the appropriate reference (L77). 89-92 2) Ecological science, especially as it applies to biodiversity conservation, uses species distribution modeling (SDM) (e.g. Osawa 2011; Osawa 2015) for predicting the probability of occurrence and sometimes also abundance of target plant and animal species based on environmental variables such as terrain and climate factors (Guisan and Zimmermann 2000) R: Please, you have to define better this concept to avoid misunderstandings. Also, there are many reference that were not included in this manuscript. Different subjects can use this approach such as Biogeography, species interaction and evolution. I listed some reference here: Res: Thank you for the suggestion regarding the non-inclusion of relevant references. We have cited the references you suggested and have explain the importance and implications of using SDM (L93-95). 93-95 3) MaxEnt is one of the major SDM software tools available and can be used to predict species occurrence distribution by analyzing only presence records and environmental variables (Phillips et al. 2004; Phillips et al. 2006). R: There many reasons to use or not a MaxEnt model or you need to show different performances of models to evaluate what is the best model. Sometimes, ensemble are good options too. You need to included advantages and disadvantages of the models. What is the best model? MaxEnt? No. You need to include context and a good model can change according to your objective, data and hypothesis. Also, all models show uncertainty. You can see important information here: Res: Thank you for the suggestion and the references supplied. As per your suggestion, we have provided an explanation in the Results section (Model construction, L186-204), and described the results of re-analysis using the modified setting. 96 - 97 4) There are no or few known reports in the literature of its application in the field of plant pathology in our knowledge R: Please, there are good papers on this subject. You need to improve all references and include the most relevant references in the manuscript. Also, this is not a good sentence: “our knowledge”. You can see here several examples: Res: We have included these references in the text (L101-103) and deleted the term “our knowledge” in the text, as per your suggestion. 101-103 5) The objective of the study was to construct an occurrence probability map of Verticillium wilt of cabbage in Tsumagoi village in Gunma prefecture, Japan to test the application of SDM to plant pathology R: This is a great problem. A manuscript should show a relevant objective. Species distributions models are not only maps. You need to improve this manuscript a lot with all suggestions included in this document. Res: The disease potential map was constructed with MaxEnt and has been depicted. We have also described the contribution of the variables in details (L237-260). Additionally, the other objective of the study has been described in the Introduction section (L119-120). 103 -104 6) Tsumagoi village is an ideal site to test the application of the SDM technique because it is a large-scale area of consolidated monoculture cabbage production in Japan where years of field monitoring has established a good record of occurrence of Verticillium wilt of cabbage. R: Verticillium wilt is a global disease with many hosts and can be caused by different species. According to theory of species distribution models, you should consider a global sampling (or a disease/pathogen area) to create the model, after that, you can study a small region. You can create a regional model, but you need to justify. It is very important to show a reasonable result, with less bias. Please, you check here a potential distribution (we can find in all continents): Res: The objective of this study using SDM was to select reasonable variables and to construct the disease potential map for selecting a suitable control methods to Verticillium wilt in cabbage. In order to model this area, we used a 'Bias Grid' and altered some of the settings for the parameters (L195-200). <Materials & Methods> 121-124 7) Cabbage production in the area has suffered from Verticillium wilt since 1994 (Shiraishi et al. 2000). It is a soil-borne disease associated with Verticillium dahliae and V. longisporum (Banno et al. 2011) R: You can indicate here what species are being modelled in the manuscript. You need to model each one, or you can model the occurrence of the disease. This disease is a species complex, then you must indicate in manuscript. Res: The model was constructed for modeling the occurrence of disease associated with both pathogens. We have explained this in detail in L158. 127-128 8) Disease occurrence was identified by external symptoms such as outer leaf yellowing and wilting R: You did not indicate a confirmation of absence in grids. Res: We have described the confirmation of the about absence of disease occurrence in the grids. In this study, the absence (absence of disease occurrence) could not be confirmed. We therefore selected MaxEnt, as it does not require the data regarding the absence of the target species (L187-190). 136-139 9) The study area was divided into a grid of 500 × 500 m cells. The grid cell size was selected according to both the degree of accuracy of the disease survey and the scale of land ownership. Disease occurrence (a binary yes/no variable) and values of the potential explanatory variables were stored for each grid cell (Fig. 3A, B). R: This part should be included in “Target disease and disease survey” section. I did not understand how you did this analysis. Accuracy is a performance metric with bias and you need to include different metrics such as f1-score, recall and precision to evaluate better your survey. Res: We shifted this portion to the “Target disease and disease survey” section. The details of the analysis and estimation of the accuracy of the constructed model have been provided in the Materials and Methods (L202-204) and Results sections (L223-224). 140-141 10) We selected road density, elevation, slope, and topographic wetness index (TWI) as possible explanatory variables that may affect the disease occurrence potential. R: The modeling is showing serious problems. All results may be influenced by overfitting of explanatory variables and there is no a justification (biology, ecology, epidemiology) to these chosen variables in the manuscript. A distribution of a specie is associated with the prevalence, that is, if a specie is present in a region, climate and soil are suitable to local establishment. Direct variables are the most indicated variables in niche modelling. For example, elevation is associated directly to temperature. Res: We have explained about the variables selection in the revised manuscript (L165-171). Additionally, the settings were altered to avoid overfitting. These have been described in the section on model construction. 155-157 11) Although we used only presence data without absence data and the data set was small, this method provides more robust results than alternative methods R: There are many models that can be used utilizing presence and absence data. Sometimes, pseudo absence is an option (a correction of an imperfect detection). Overfitting is a well-known problem in Maxent. See: Res: Strictly speaking, the absence of disease (disease not confirmed) could not be confirmed. We therefore selected MaxEnt, as it does not require the absence. Some of the settings the parameters were altered to avoid overfitting (L195-200). 161-165 12) The disease occurrence probability (0-1) in each grid cell and percent contribution of each explanatory variables were calculated by the MaxEnt software. In addition, the model of the probability of disease occurrence was also estimated by means of receiver operating characteristic (ROC) analysis. We used the default settings of the MaxEnt software to predict potential occurrence. R: Why default? You must explain each chosen parameter of this model. There is many parameters that can changed and the result may change too. Res: Some of the parameters were altered during re-analysis (L195-200). We observed that results also changed following re-analysis. <Discussion> 169-174 13) The SDM model constructed with MaxEnt was used to map the probability of disease occurrence (Fig. 4A). The cells with disease occurrence had a high degree of coincidence with those with a high probability of such occurrence (Fig. 4B). The constructed model accurately represented the occurrence of Verticillium wilt of cabbage, as demonstrated by the high value, 0.903, of the area under the curve of the receiver operating characteristic. The highest contribution to the prediction was road density at 60.7%, followed by elevation at 24.4%, TWI at 7.7%, and slope at 7.1%. R: Results may be different considering all changes indicated until here. The chosen model, parametrization, data base, variable overfitting and uncertainty analysis are some problems. Res: The results changed entirely after changing the settings of the parameters. 205-210 14) In the present study, we applied SDM to predict the occurrence probability of Verticillium wilt of cabbage. That soil-borne disease occurrence could be predicted by environmental variables without plant pathological analysis is a novel finding. Use of the model will promote the adoption of effective and sustainable management that considers the costs and life cycle of the disease. The results of this study may help reconcile the competing demands for conservation agriculture and food security. R: The discussion section is very weak. I found two references in the whole section. Also, all results can show inconsistency. Res: The Discussion section was changed entirely as the Results section changed following re-analysis. <Conclusion> 217-222 15) The model developed from the field survey data showed that road density and elevation played important roles in the occurrence of the disease. The following three points are particularly important and novel results of this study: 1) we could find the measure to predict and visualize the occurrence potential of soil-borne disease. 2) Predicting the occurrence potential was conducted by SDM that was used in the ecological science. 3) By applying geographical information system (GIS) R: Point one: all results are not robust to be associated with soil-borne diseases. Point two: it is not a new approach and is already being used in the scientific literature. Point three: thousands of papers showed similar approaches. This paper does not have adequate modeling and this distribution of the disease is being affected by problems of the method and theory, indicating a lack of ecological, biological and epidemiological bases. Res: Thank you for your suggestion. This manuscript has been revised according to these three points. ******************************************** Reviewer 3 (Anonymous) Thank you for the helpful comments and suggestions. We have revised our manuscript according to your suggestions. We have revised the descriptions from the phytopathological viewpoint. The responses to your comments are appended hereafter. <Basic reporting> Lines 23-24: Please review the use of infected (the host) vs. infested (soil, inanimate objects) here and throughout (Lines 79, 80, I would tend to write this sentence as “…IPM methods are sometimes inadequate to bring about control in severely infested fields.” since the field is infested, unless you are talking about the plants in the field. Res: All instance of 'infest' and 'infect' were revised as per your suggestion. Line 26: do you mean disinfestation? Res: Yes, we meant disinfestation. We have revised this accordingly (L22). Line 29: “…to a soilborne…” Res: This was deleted in the revised version of the Abstract. Lines 66 and 72: change disinfection to disinfestation Res: This has been revised accordingly (L66, L72). Lines 122-124: I think more information on the pathogens could be provided here (e.g. how long it survives in soil, host range – wide in V .dahliae vs. brassicas in V. longisporum. What is there relative abundance in relation to each other? This can be important depending on what other crops (hosts vs non-hosts) are rotated in the fields. Res: We have provided an explanation regarding the pathogens as per your suggestion (L140-145). <Experimental design> Lines 112-119: what other crops are grown here? Are other hosts grown in rotation? Does infected seed play a role in the introduction of the pathogen into fields? Res: Although potato, maize, and scarlet runner bean are grown in this area, the cultivation of these crops in markedly lower than that of cabbage. We have explained this explanation in L135-136. Crop rotation is not practiced in this area, as the farmers are separated. There are no reports on whether infected cabbage seed play a role in the introduction of Verticillum in this area. Line 121: are certain cultivars of cabbage grown that are resistant or susceptible? Verticillium may go undetected in fields planted to resistant cultivars or detected more frequently in fields planted more often to susceptible cultivars. Res: Some seed companies in Japan have developed numerous cultivars of cabbage. Owing to the tractability system, it is possible to be certain whether the cultivars planted in the fields are devoid of disease. Resistant cultivars are being used in the recent years. We have described this in L145-146. Line 125: Were fields sampled more or less equally, or is there any potential bias for certain cells? Res: We investigated all the agricultural area during survey, as depicted in Fig 3A. Therefore, there is a potential survey bias for this agricultural area and other grid cells. In order to avoid this bias, the bias grid was used during re-analysis with MaxEnt analysis (L195-197). Line 135: Do some fields or areas have longer histories of cabbage production than others? Since Verticillium can survive in the soil and buildup over time, would fields with longer histories of cabbage production be identified as infected more instances over time than newer fields? Res: Thank you for the comment. It could have been possible as there were some grids in which disease occurrence was not recorded despite the high potential of the disease occurrence. It was assumed that the cabbage fields in these girds were newly infected. We have described this in the Discussion section (L272-280). Lines 157-158: what was the sample size (n)? Res: The number of grid cells in which the disease occurred and the whole area of the village has been mentioned in the revised manuscript (L183-184). <Validity of the findings> Lines 77-78: what about exclusion and prevention? Res: We too agree that these are important, and have described there in L78. Lines 80-82: it could also be mentioned that knowledge of infested and non-infested fields could help growers prevent infesting new fields or re-introducing the pathogen to fields that have been fumigated. Res: Thank you for the suggestion. We have described this accordingly (L83-85). ****************************************** Reviewer 4 (Anonymous) Thank you for the helpful comments and suggestions. We have revised our manuscript according to your suggestions. We have revised some of the descriptions from a phytopathological viewpoint. Our response to your comments area appended hereafter. <Basic reporting> 1) The manuscript is clearly written. There are some relevant references missing. Please note the following references when you state that there is no previous studies utilizing SDMs in plant pathogen epidemics. Res: Thank you for the suggestion. We have incorporated the references as per your suggestion (L101-103). 2) I also suggest you discuss more on the road network in disease occurrence in the light of other studies eg. The spread of a wild plant pathogen is driven by the road network Res: We have described the contribution of the road density in disease occurrence, and have also incorporated some relevant reference (L237-246). <Validity of the findings> 3) The authors should discuss how their findings could be extended to wider areas. Is it possible that this model can be used in other Japanese or global areas and why. Res: We have discussed how the findings of our study can be extended to the wider area in the revised manuscript (L292-296). "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background Although integrated pest management (IPM) is essential for conservation agriculture, this method can be inadequate for severely infected fields. The ability to predict the potential occurrence of severe infestation of soil-borne disease would enable farmers to adopt suitable methods for high-risk areas, such as soil disinfestation, and apply other options for lower risk areas. Recently, researchers have used species distribution modeling (SDM) to predict the occurrence of target plant and animal species based on various environmental variables. In this study, we applied this technique to predict and map the occurrence probability of a soil-borne disease, Verticillium wilt, using cabbage as a case study.</ns0:p><ns0:p>Methods A disease survey assessing the distribution of Verticillium wilt in cabbage fields in Tsumagoi village (central Honshu, Japan) was conducted two or three times annually from 1997 to 2013. Road density, elevation and topographic wetness index (TWI) were selected as explanatory variables for disease occurrence potential. A model of occurrence probability of Verticillium wilt was constructed using the MaxEnt software for SDM analysis. As the disease survey was mainly conducted in an agricultural area, the area was weighted as 'Bias Grid' and area except for the agricultural area was set as background.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Grids with disease occurrence showed a high degree of coincidence with those with a high probability occurrence. The highest contribution to the prediction of disease occurrence was the variable road density at 97.1%, followed by TWI at 2.3%, and elevation at 0.5%. The highest permutation importance was road density at 93.0%, followed by TWI at 7.0%, while the variable elevation at 0.0%. This method of predicting disease probability occurrence can help with disease monitoring in areas with high probability occurrence and inform farmers about the selection of control measures.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Intensive agriculture has supported world food production since World War II <ns0:ref type='bibr' target='#b31'>(Moffatt 2020</ns0:ref>). However, the sustainability of intensive agriculture is often questioned because of negative consequences such as soil and water pollution, risks to human health from pesticides and fertilizer, and water resource depletion by over-exploitation of water resources <ns0:ref type='bibr' target='#b36'>(Oosterbaan 1989;</ns0:ref><ns0:ref type='bibr' target='#b29'>Matson et al. 1997;</ns0:ref><ns0:ref type='bibr' target='#b53'>Tilman et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b13'>Hern&#225;ndez et al. 2006</ns0:ref>). To address these problems, a shift to conservation agriculture is required <ns0:ref type='bibr' target='#b47'>(Reicosky 2003;</ns0:ref><ns0:ref type='bibr'>Kassam et al. 2009</ns0:ref>). On the other hand, measures to increase food supply to maintain food security are also needed because the world population is expected to reach over 9.7 billion by 2050 (United Nations 2015). These competing demands for sustainability and higher yields presents a complex problem facing modern intensive agriculture <ns0:ref type='bibr' target='#b5'>(Brussaard et al. 2010</ns0:ref>), but it is a challenge that must be met.</ns0:p><ns0:p>To cope with the ever growing global food demand, intensive agriculture primarily revolves around large-scale monoculture, the continuous or consecutive growth of the same crop over large areas <ns0:ref type='bibr' target='#b6'>(Cook and Weller, 2004)</ns0:ref>. Monocultures carry a heavy risk of soil-borne disease because pathogens have a continuous supply of host plant and are thus able to persist or even accumulate in the soil <ns0:ref type='bibr' target='#b34'>(Newton et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b17'>Jenking et al. 2010)</ns0:ref>. Soil disinfestation by fumigation with chemical control agents can be effective in the management of soil-borne disease <ns0:ref type='bibr' target='#b59'>(Wilhelm and Paulus 1980;</ns0:ref><ns0:ref type='bibr' target='#b19'>Koike et al. 2003)</ns0:ref>, but also carries the risk of negative impacts on the environment and human health <ns0:ref type='bibr' target='#b13'>(Hern&#225;ndez et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b48'>Sande et al. 2011</ns0:ref>; United States Environmental Protection Agency 2017). In addition, as more fields require chemical control, the costs associated therewith also increase <ns0:ref type='bibr' target='#b24'>(Landis 1987;</ns0:ref><ns0:ref type='bibr' target='#b23'>Labrada et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b20'>Koike et al. 2006)</ns0:ref>. So, although effective, chemical disinfestation to reduce soil-borne disease can be difficult to adopt when the costs and consequences outweigh the benefits.</ns0:p><ns0:p>Integrated pest management (IPM), which is defined as the long-term prevention of pests or their damage through a combination of techniques, is essential for sustainable agriculture <ns0:ref type='bibr' target='#b0'>(Apple et al. 1976</ns0:ref><ns0:ref type='bibr' target='#b56'>, UC IPM 2020)</ns0:ref>. The control of soil-borne disease in IPM typically involves the application of resistant cultivars, crop rotation, exclusion and prevention of pathogen's inoculum source <ns0:ref type='bibr' target='#b55'>(Tsushima 2014;</ns0:ref><ns0:ref type='bibr' target='#b15'>Ikeda et al. 2015)</ns0:ref>. However, such methods are inadequate to bring about control in severely infested fields. Therefore, intensive continuous soil fumigation may be justified in severely infested fields <ns0:ref type='bibr' target='#b22'>(Krikun et al. 1974)</ns0:ref>. If the potential occurrence of severe infection of soil-borne disease could be predicted, farmers could adopt soil fumigation for high-risk areas and apply other options for areas of lower risk. Furthermore, evaluating infestation risk in a giving area can help growers to prevent the spread to unaffected fields or re-introduction in previously managed fields. This approach is in accordance with the IPM framework for sustainable, low-environmental impact, and cost-effective agriculture <ns0:ref type='bibr' target='#b54'>(Tsushima and Yoshida 2012)</ns0:ref>. Control measures for soil-borne diseases should be selected according to the disease potential occurrence of each target area. However, the development of a practical method to predict such disease potential occurrence has not previously been given in the field of plant pathology.</ns0:p><ns0:p>Ecological science uses species distribution modeling (SDM) (e.g. <ns0:ref type='bibr' target='#b39'>Osawa 2011;</ns0:ref><ns0:ref type='bibr' target='#b38'>Osawa 2015)</ns0:ref> for predicting the probability of occurrence and sometimes also abundance of target plant and animal species based on environmental variables such as terrain and climate factors <ns0:ref type='bibr' target='#b12'>(Guisan and Zimmermann 2000;</ns0:ref><ns0:ref type='bibr' target='#b51'>Sober&#243;n and Peterson 2005;</ns0:ref><ns0:ref type='bibr' target='#b50'>Sober&#243;n J. 2007;</ns0:ref><ns0:ref type='bibr' target='#b42'>Peterson et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b41'>Peterson 2014)</ns0:ref>. SDM can predict the distribution of a target species using survey data obtained in the area (McCune 2016). Additionally, the respective contribution, or importance, of environmental variables on species distribution can be calculated. Maximum entropy algorithm is one of the most famous SDM approaches and can be used to predict species potential distribution by analyzing presence-only data with environmental variables <ns0:ref type='bibr' target='#b43'>(Phillips et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Phillips et al. 2006)</ns0:ref>. MaxEnt software has been successfully used in predicting potential species distribution <ns0:ref type='bibr' target='#b37'>(Ortega-Huerta et al. 2008)</ns0:ref>. Recently, MaxEnt has been used to predict plant disease distribution, such as Fusarium dry root rot in common beans, Phomopsis vaccinii in Vaccinium species, myrtle rust in Myrtaceae family and Pseudomonas syringae pv. actinidiae in kiwifruit <ns0:ref type='bibr' target='#b7'>(Cunniffe et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b27'>Macedo et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b32'>Narouei-Khandan et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Berthon et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b58'>Wang et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b33'>Narouei-Khandan et al. 2020</ns0:ref>). In the present study, we used MaxEnt to predict the occurrence probability of Verticillium wilt, a soil-borne disease, as a case study on cabbage field in Japan.</ns0:p><ns0:p>Tsumagoi village in the Gunma prefecture of Japan intensively produces cabbage as one of the main production sites for this crop in Japan. This is an ideal site to test the application of the SDM technique, given its large-scale area of consolidated monoculture and years of field monitoring, which enable a detailed analysis of the occurrence of Verticillium wilt. Verticillium wilt in cabbage has been a big problem in this region's cabbage production since 1994 <ns0:ref type='bibr' target='#b49'>(Shiraishi et al., 2000)</ns0:ref>. This is a soil-borne disease caused by Verticillium dahliae and V. longisporum <ns0:ref type='bibr' target='#b1'>(Banno et al., 2011)</ns0:ref>. The cabbage farmers in Tsumagoi village have suffered severe economic losses from this disease <ns0:ref type='bibr' target='#b49'>(Shiraishi et al. 2000)</ns0:ref>.</ns0:p><ns0:p>The objective of the study was to construct an occurrence probability map of Verticillium wilt of cabbage in Tsumagoi village in Gunma prefecture in Japan to test the application of SDM to soilborne plant disease. The findings aim to inform farmers to choose adequate control measures, such as introducing resistant cultivars or changing cultivation period. This trial can contribute to the development of IPM by offering both environmental and economic improvements to soilborne disease management in agricultural production.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Study area</ns0:head><ns0:p>The distribution of Verticillium wilt of cabbage was surveyed in Tsumagoi village (36&#176;51'17'N, 138&#176;53'00'E) in central Honshu, the largest of Japan's four main islands (Fig. <ns0:ref type='figure'>1</ns0:ref>). The climate is subarctic humid climate (Dfb) in K&#246;ppen climate classification with an elevation ranging between 700 and 1500 m above sea level and a mean annual temperature of 7.2 &#176;C (Japan Meteorological Agency 2017, Village Office of Tsumagoi 2017). Cabbages are produced on approximately 2790 ha of the 3200 ha agricultural area. This is a remarkably large monoculture site for cabbage production in Japan. Thus, the village is renowned in Japan as a production site for cabbage. Cabbages are harvested there from April (early spring) to October (autumn) (Village Office of Tsumagoi 2017). However, potato, maize and scarlet runner bean are also grown in this area, in significantly lower proportions than those of cabbage (Village Office of Tsumagoi 2017).</ns0:p></ns0:div> <ns0:div><ns0:head>Target disease and disease survey</ns0:head><ns0:p>Cabbage production in the area has suffered from Verticillium wilt since 1994 <ns0:ref type='bibr' target='#b49'>(Shiraishi et al. 2000)</ns0:ref>. Verticillium wilt is a soil-borne disease associated with Verticillium dahliae and V. longisporum <ns0:ref type='bibr' target='#b1'>(Banno et al. 2011)</ns0:ref>, which can survive for years using microsclerotia as a resting structure. V. dahliae has a board range of hosts, mainly plants from the Solanaceae and Brassicaceae families <ns0:ref type='bibr' target='#b40'>(Pegg and Brady, 2002)</ns0:ref>, while V. longisporum infects mainly members of the Brassicaceae family <ns0:ref type='bibr' target='#b14'>(Inderbitzin et al., 2013)</ns0:ref>. In our study site, V. dahliae was found to be the dominant species causing Vertillium wilt in cabbage <ns0:ref type='bibr' target='#b2'>(Banno et al., 2015)</ns0:ref>. In recent years, cultivars resistant to Verticillium wilt have been planted in the village. The disease survey was conducted two or three times per year annually from 1997 to 2013, 38 times in total by official and private agricultural extension workers and plant pathologists capable of distinguishing the disease in all fields of the village. This survey and field collections were approved by Gunma Prefectural Office under permission document no. H28.114.30. Disease occurrence was identified by external symptoms such as outer leaf yellowing and wilting (Fig. <ns0:ref type='figure'>2A, B</ns0:ref>). When a cabbage suspected to be infected was found, the observers examined the vascular system of selected cabbages to identify browning and thus confirm the existence of the disease (Fig. <ns0:ref type='figure'>2C</ns0:ref>). The locations of fields where the disease occurred in field survey were marked on a 1:5000 map. When new records were found, investigators took cabbage samples for isolation and identification of the pathogens. The study area was divided into a grid of 500 &#215; 500 m cells. The grid cell size was selected according to both the degree of accuracy of the disease survey and the scale of land ownership. Disease occurrence (a binary variable) of both pathogen species and the values of the selected explanatory variables were recorded for each grid cell (Fig. <ns0:ref type='figure'>3A, B</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Explanatory variables</ns0:head><ns0:p>Geographic Information System software ArcMap (ver. 10.7.1, ESRI, Redlands, CA, USA) was used for analysis of the data set. We selected road density, elevation and topographic wetness index (TWI) as possible explanatory variables that may affect the disease potential occurrence. In Tsumagoi village, the pathogens of cabbage Vertillium wilt were detected in field soil <ns0:ref type='bibr' target='#b1'>(Banno et al. 2011</ns0:ref>) and there was serious soil erosion from cabbage fields to the roads <ns0:ref type='bibr' target='#b8'>(Deb 2006)</ns0:ref>. This suggested to us that the disease spread from one field to another by pathogen-infested soil, which prompted us to investigated road density as an explanatory variables. Furthermore, we included elevation as an explanatory variables because of relationship between elevation and temperature: for every 100m increase in elevation, temperature rises by about 0.6&#176;C. Finally, we also selected TWI, which indicates the degree of soil moisture <ns0:ref type='bibr' target='#b11'>(Gallant 2000)</ns0:ref>, the latter is known to influence the development of soil-borne disease like Verticillium wilt <ns0:ref type='bibr' target='#b40'>(Pegg and Brady, 2002)</ns0:ref>. Road data were downloaded as polyline data from Fundamental Geospatial Data (Geographical Survey Institute; http://www.gsi.go.jp/kiban/, accessed on 24 June 2020). The number of new roads had not increased dramatically from 2013 to 2016 in the study area. The total length of road in each grid cell was calculated from the polyline data as 'road density.' Elevation and slope were downloaded from the National Land Numerical Information download service (Ministry of Land, Infrastructure, Transport and Tourism, http://nlftp.mlit.go.jp/ksj/, accessed on 24 June 2020). Slope data are required to calculate TWI <ns0:ref type='bibr' target='#b11'>(Gallant 2000)</ns0:ref>, which is defined as follows:</ns0:p><ns0:formula xml:id='formula_0'>TWI = ln &#119886;</ns0:formula><ns0:p>tan &#119887; where a is local catchment area (i.e., the local upslope area draining per unit) and b is the local slope. Values for elevation and TWI were applied to each grid cell along with the road density variable. The agricultural area shown in Fig. <ns0:ref type='figure'>3A</ns0:ref> were also downloaded from the National Land Numerical Information download service described above.</ns0:p></ns0:div> <ns0:div><ns0:head>Model construction</ns0:head><ns0:p>The model of probability of disease occurrence was constructed using SDM software MaxEnt ver. 3.3.3k (downloaded at: https://biodiversityinformatics.amnh.org/open_source/maxent/) <ns0:ref type='bibr' target='#b43'>(Phillips et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Phillips et al. 2006)</ns0:ref>. This software was chosen as it can analyze presenceonly data efficiently with a low number of occurrences <ns0:ref type='bibr' target='#b9'>(Elith 2006)</ns0:ref>. MaxEnt is used for predicting habitat suitability of target species, but here we applied it to the prediction of probability of soil-borne disease occurrence. The disease occurrence probability (0-1) in each grid cell, percent contribution and permutation importance of each explanatory variables were calculated by MaxEnt. We conducted the analysis using the whole village area, given that this was the scale relevant for our study. However, the original disease survey was mainly conducted within the agricultural areas (Figure <ns0:ref type='figure'>3A</ns0:ref>), hence we used a 'Bias Grid' weighting and set the remaining 'non-agricultural area' as background in MaxEnt setting. To avoid overfitting, only 'Liner' and 'Quadractic' functions were used and the value for 'Regularization multiplier' was changed from '1' to '2' in MaxEnt settings <ns0:ref type='bibr' target='#b30'>(Merow et al., 2013;</ns0:ref><ns0:ref type='bibr'>Radosavlijevic and Adnderson, 2014;</ns0:ref><ns0:ref type='bibr' target='#b52'>Syfert et al., 2013)</ns0:ref>. Cross-validation was selected in 'Replicated run type' and repeated 20 times. The model of the probability of disease occurrence was also estimated by means of receiver operating characteristic (ROC) analysis. Furthermore, correctly classified instances (CCI), sensitivity, specificity, true skill statistics (TSS) were calculated based on the 'threshold by maximum training sensitivity plus specificity' in MaxEnt to evaluate model performance.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>There were 1392 grids cells in our study area, we confirmed disease occurrence in 194 grid cells (Fig. <ns0:ref type='figure'>3B</ns0:ref>). The SDM model constructed with MaxEnt was used to map the probability of disease occurrence and points of disease occurrence (Fig. <ns0:ref type='figure'>4</ns0:ref>). Threshold by Maximum training sensitivity plus specificity from MaxEnt analysis was 0.543. The disease occurrence probability map was created based on this threshold. Grids in this map were colored when the grids had a higher probability compared to the threshold (in total, 556 / 1392 grids in this study area). The grids with disease occurrence had a high degree of coincidence with those with a high probability of such occurrence.</ns0:p><ns0:p>Percent contribution and permutation importance of predictor variables are shown in Table <ns0:ref type='table'>1</ns0:ref>. The highest contribution to the prediction was road density at 97.1%, followed by TWI at 2.3%, and elevation at 0.5%. Permutation importance was as follow: road density at 93.0%, TWI at 7.0%, and elevation at 0.0%. Both percent contribution and permutation importance indicated that road density was the best predictor variable: disease probability occurrence was highest when road density was about 4000 m/grid (Fig. <ns0:ref type='figure'>5</ns0:ref>).</ns0:p><ns0:p>Estimation indices for accuracy of MaxEnt modeling of cabbage Verticillium wilt is shown in Table <ns0:ref type='table'>2</ns0:ref>. The constructed model represented the occurrence of Verticillium wilt of cabbage moderately well, as demonstrated by the value, 0.861, of the AUC average of the ROC. The calculated values for CCI, sensitivity, specificity, and TSS using the threshold were 0.705, 0.876, 0.678 and 0.544, respectively.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In this study, we applied SDM to predict the occurrence probability of Verticillium wilt in cabbage as a case study. Our model could provide the occurrence of Verticillium wilt of cabbage in Tsumagoi village to some degree. To the best of our knowledge, this is the first time SDM has been used to estimate occurrence probability of soil-borne disease. Our findings indicated that this method can successfully predict disease occurrence probability using survey data and selected environmental variables in a monoculture area.</ns0:p><ns0:p>Our result showed that road density significantly influenced disease occurrence probability (Table <ns0:ref type='table'>1</ns0:ref>), with the highest probability of disease at around 4000 m/grid (Fig. <ns0:ref type='figure'>5</ns0:ref>). <ns0:ref type='bibr' target='#b35'>Numminen and Laine (2020)</ns0:ref> showed that road networks played an important role in the spread of plant diseases, which is consistent with our findings, suggested that the pathogen spread via road networks. It was thought that the pathogen was locally transmitted between fields through the soil, as they are soil-borne <ns0:ref type='bibr' target='#b1'>(Banno et al. 2011)</ns0:ref>. Soil can be transferred from infested fields to road surfaces on the wheels of tractors or other vehicles and even the footwear of farmworkers. Additionally, Tsumagoi village is located in mountainous area with steep slopes, which facilitates soil transfer between fields (e.g., by wind and water). Both natural and human-mediated processes can influence the spread of Verticillium wilt in our study area.</ns0:p><ns0:p>On the other hand, our model suggested that the influence of TWI was quite low, even though soil moisture has previously been known to influence the development of soil-borne disease like Verticillium wilt <ns0:ref type='bibr' target='#b40'>(Pegg and Brady, 2002)</ns0:ref>. The relationship between Verticillium wilt in cabbage and soil moisture had not yet been investigated before this study; however, it was reported that soil moisture was not significant factor for disease development in oilseed rape, Brassicaceae family <ns0:ref type='bibr' target='#b21'>(Kn&#252;fer, 2013)</ns0:ref>. Consequently, the low impact shown by TWI on disease occurrence probability concurs with previous research, although more research is required to conclusively rule out this relationship. Elevation also showed a very low contribution to the probability of disease occurrence. In general, there are a correlation between elevation and temperature, as temperature rises about 0.6 &#176;C for every 100m increase in elevation. Given that the optimal temperature for cabbage Verticillium wilt was between 19&#176;C and 23&#176;C, temperature during the harvest period is regarded as an especially important <ns0:ref type='bibr' target='#b18'>(Kemmochi et al 2001)</ns0:ref>. Therefore, we predicted that elevation could have an influence on disease occurrence. Cabbage is normally cultivated between 4 and 24&#176;C <ns0:ref type='bibr' target='#b4'>(Bradley and Courtier, 2006)</ns0:ref>. Cabbage cultivation period was determined according to the elevation of each field because temperature decreases with increasing elevation. Given our results, the elevation factor might not influence the probability of disease occurrence in our study area.</ns0:p><ns0:p>Our model was evaluated using statistic metrics based on the defined threshold (Table <ns0:ref type='table'>2</ns0:ref>) and the disease occurrence probability map (Fig. <ns0:ref type='figure'>4</ns0:ref>). It appeared that many grids where disease occurrence was recorded had a high probability. However, the AUC, CCI, sensitivity, specificity and TSS obtained from our analysis were moderate compared with other studies <ns0:ref type='bibr' target='#b58'>(Wang et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b33'>Narouei-Khandan et al., 2020)</ns0:ref>. The sensitivity of the model should be regarded as the most important for our objective because we hope use it to monitor grids with a high disease occurrence probability. The sensitivity of the constructed model was 0.88, which indicates that the model can predict the disease occurrence grids at about 88% accuracy. As a result, the model constructed using MaxEnt was sufficiently effective for prediction of the disease occurrence.</ns0:p><ns0:p>The disease occurrence probability map helps disease monitoring in high probability grids and selected area, and can inform the application of measures such as fumigation, planting diseaseresistant cultivars, and changing the timing of cabbage cultivation <ns0:ref type='bibr' target='#b18'>(Kemmochi et al. 2001</ns0:ref>).</ns0:p><ns0:p>Although it is not possible to change the road density, it is possible to apply soil erosion and sediment transport control technologies such as the establishment of cover crops <ns0:ref type='bibr' target='#b25'>(Lawson et al. 2015)</ns0:ref> to decrease the potential occurrence of the disease, especially in those grid cells where there are preexisting records and those grid cells predicted to have high occurrence probability. In addition, we recommend that the farmers and extension workers apply methods to reduce potential spread of infested soil, such as using portable vehicle-washing equipment and changing footwear at the site are required.</ns0:p><ns0:p>This study shows that SDM analysis using MaxEnt can be used to predict the occurrence using survey data and environmental variables. This approach could be applied to other intensive crop cultivation areas with accurate disease survey data and environmental variables, available as GIS data, that potentially influence disease occurrence.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In this study, we applied SDM to predict the occurrence probability of Verticillium wilt of cabbage, soil-borne plant disease. The SDM model constructed from three explanatory variables provided prediction of the occurrence of Verticillium wilt of cabbage in Tsumagoi village, Japan. The model developed from the field survey data showed that road density played an important roles in the occurrence of the disease. Key points of this study are; 1) SDM enabled us to predict the probability of disease occurrence in a relatively narrow area; 2) potential explanatory variables for disease occurrence were predicted based on SDM using field survey; and 3) predicting the probability of disease occurrence can help farmers to select suitable control methods for high and low-risk areas.</ns0:p></ns0:div><ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='14,42.52,229.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='15,42.52,204.37,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='16,42.52,331.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,331.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,224.62,525.00,363.00' type='bitmap' /></ns0:figure> </ns0:body> "

"Editor comments (Victoria Sosa) I agree with the two reviewers in that it is necessary an English editorial review, and that the Discussion needs clarity. One of the reviewers included comments in the attached file, please consider them. Res: Thank you for your revise again on our manuscript. We altered some sentences and Discussion part for reducing confusion of PeerJ readers. Thus the manuscript was revised according to the suggestions provided by the 2 reviewers. In addition, we have adjusted reference style according to the guideline of PeerJ. [# PeerJ Staff Note: The Academic Editor has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title) #] Res: We asked proofreading service to check our English language. We attached certification of the service, please confirm it. ****************************************** Reviewer 1 (Anonymous) Thank you for your revise again on our manuscript. We have revised our manuscript according to your suggestion. We have mainly altered some sentences and a clarity of Discussion part that you pointed out. Our responses to your comments are appended hereafter. <Comments for the Author> Dear authors, I appreciate your effort addressing all my previous comments; however, the new version of your manuscript has some issues that need to be clarify and corrected before acceptance. English requires a review, as there are sections that are not easy to follow or understand the ideas. Res: We altered some sentences and Discussion part for reducing confusion. In addition, we asked proofreading service to check our English language. We attached a certification of the service, please confirm it. <Attached PDF file> We indicated line numbers of previous manuscript and your comments firstly, and then those of new version in parentheses in our responses. L35 agricultural area was set as the background. Res: This has been revised as suggested (L34-35) L39 delete Res: This has been deleted as suggested. L42-44 as I mentioned in my previous review, AUC values do not say much, consider deleting this. Res: This has been deleted as suggested. L45 areas? Res: Yes, this has been changed to “areas” (L43). L71 quite subjective Res: This sentence has revised not to become subjective (L67-68). L73 What is “acceptable”? Res: This sentence has been changed without “acceptable” (L70) L81 reference? Res: We have provided the corresponding reference (L77-78). L84 managed? Res: Yes, this has been changed to “managed” (L82). L88 target area Res: This has been revised as suggested (L85). L96 references? Res: We have provided the corresponding reference (L93-94). L99 presence-only data Res: This has been revised as suggested (L97). L101 to predict Res: This has been revised as suggested (L99). L101 perhaps provide more details Res: We provided more detailed description (L99-100) L103 repetitive Res: This sentence has been deleted. L108 delete Res: This has been deleted. L115 reference Res: We provided the corresponding reference (L11). L132 ? Res: We apologize for confusing you. This words “degree of” were deleted (L131-132). L135, 136 reference? Res: We have provided the corresponding references (L134, 135-136). L142 Verticillium dahliae Res: This has not been changed, because “Verticillium” was spelled out in line 140. L146 resistant what? Res: We inserted the word “cultivar” in front of “resistant” (L146). L147 annually Res: This has been changed as suggested (L147). L148 distinguishing Res: This has been revised as suggested (L148). L149 in all fields Res: This has been revised as suggested (L149). L149 were Res: This has been changed (L149). L152 what do you mean with “suspected field”? Res: This sentence has been changed (L151-152). L165-167 beginning of what?, not clear this justification, also, too weak Res: We revised this part to justify selection of road density as potential explanatory variable (L166-168). L167 this sentence seems lost here Res: This sentence has been delete. L168 but why not temperature over elevation? did you actually run a correlation analysis to select variables? Res: As there is only one site where measures temperature automatically in Tsumagoi village, strictly speaking, we cannot run actually correlation between temperature and elevation. However, it is generally known that temperature rises about 0.6 °C per 100 m elevation rise. Therefore, we have added sentence for explanation (L168-170). L168 which one? Res: We have changed this sentence entirely (L168-170). L169 what model? you haven't mentioned this Res: This has been deleted, because model construction was explained in next paragraph. L170 to influence Res: We have revised as suggested (L171). L173 roads Res: We have revised as suggested (L174). L181 grid cell Res: “cell” has been added (L182). L184 results Res: This sentence has been transferred to “Results” section (L208-209). L187 using SDM… Res: We have revised as suggested (L187). L189 chosen Res: This has been changed as suggested (L189). L191 no, Maxent doesn't predict presence, it predicts habitat suitability that can be related to species' presence. Res: We appreciated your comment. We revised this sentence as suggested (L190-191). L198 functions were used and Res: This has been revised as suggested (L198). L198 changed from “1” to… Res: This has been revised (L199). L199 settings Res: This has been revised (L199). L204 to evaluate model performance Res: This phase has been added as suggested (L204). L208-212 Methods Res: This paragraph has explained the threshold obtained, therefore we have not transferred this sentences to “Materials and Methods” section. L223 is Res: This has been altered (L224). L226 not a very good model Res: We understood your comment in here. Therefor related description hereafter has been revised entirely. In this part, “accurately” has been altered to “moderately” and the word “high” has been deleted (L225-228). L231 based on what? Res: We have also revised here as well as the Results section. In here, the word “good” has been changed to “moderate” (L232-233). L241 reference? Which report? Res: The report was shown in just before, (Numminen and Laine 2020). We have changed the sentence (L239-241). L242 reference? Res: The corresponding references has been added (L243). L250 then, why did you use it? Res: Soil moisture is known to influence the development of soil-borne disease like Verticillium wilt (Pegg and Brady, 2002). However, the relationship between Verticillium wilt in cabbage and soil moisture had not yet been investigated before this study. Therefore, we selected TWI as potential explanatory variables. These have been described here and “Explanatory variables” (L252-254). L254-258 but you used elevation and not temperature, and you haven't mentioned anything about how this two variables correlated, therefore, this discussion is weak. Also, don't you mean temperature during the 'survey' period'? not sure why harvest period is important here Res: We have added explanation about relationship between elevation and temperature in here (L259-263) and “Explanatory variables” (L169-171). Kemmochi et al.(2001) reported temperature in harvest period influenced the disease occurrence. Therefore, we mentioned temperature in harvest period. We have inserted description here (L258-262). L258 what do you mean? Res: We apologized confusing you. We asked the proofreading service to check our English. Please check this sentence again (L262-263). L272-280 This paragraph has no relevance and repetitive information, consider deleting it Res: We agreed your suggestion, this paragraph has been deleted. L283 your map doesn't help to select practical control measures; it can be used to select areas to apply those measures Res: This has been changed as suggested (L277-279). L284 reference? Res: This reference has been shifted. This reference described about resistant cultivars and changing the timing of the cultivation (L279). L285 delete Res: This has been deleted. L288 and those grid cells predicted to have high occurrence probability Res: We have added this phrase as suggested (L283). L288 apply methods to the farmers and workers?? don't you mean farmers and workers should apply methods on..... Res: We apologized confusing you. This has been changed (L284-286). L293 you didn’t show this Res: This has been deleted. L293 delete Res: We have deleted this phrase. L295 check grammar, not sure what you mean. Res: We apologized confusing you. We asked the proofreading service to check our English. Please check this sentence again (L289-291). L302 not really, looking at figure 4, the probability is low, below 70% Res: We have revised as suggested, the word “good” has been deleted (L297) L304 an important role Res: This has been revised (L298-299). L304 Key point of this study are: Res: We have revised this sentence as suggested (L299). L305-308 These three points are very weak, you must conclude with the applicability and potential uses of your study, not your methods Res: As per your suggestion, we have revised these three point entirely. Applicability and potential uses have been picked up here (L299-303). L309-311 delete, let your reader develop this idea, as a conclusion is weak. Res: This sentence has been deleted as suggested. Our ideas, originality of this study, have been mentioned above. ******************************************** Reviewer 3 (Anonymous) Thank you for the helpful comments and suggestions. We have revised our manuscript according to your suggestions. The responses to your comments are appended hereafter. < Basic reporting > line 89 - change to 'given' line 98 - change to 'approaches' line 145 - change to 'In recent years, cultivars...' line 147-148 - change to '...times in total by official and private.....capable of distinguishing...' line 149 - change to 'The surveys and field collections were approved...' line 173 - change to 'roads' line 285 - delete 'more' Res: As per your suggestion, we have revised all mentions of the above. line 258-259 - this sentence is not clear to me. Res: We apologized confusing you. We asked the proofreading service to check our English. Please check this sentence again (L262-263). line 272- change to 'uninfested' line 276 - should 'becoming' be changed to 'detecting'? This sentence is not clear. line 278-280 - This sentence is not clear. Do the authors mean that the disease was present but not detected due to some unknown control measures that were being used? Res: This paragraph has been deleted as suggestion of other reviewer. < Experimental design > Is there a(n) (auto)correlation between road density and agricultural production that would confound these results (i.e. there are more roads in agricultural areas, and more Verticillium in agricultural areas, hence road density and Verticillium appear to be correlated)? Res: In this study, correlation between road density and agricultural area was not detected, as roads have distributed not only agricultural area but also urban area. On the other hand, degree of road density within agricultural area was different. Therefore, probability of disease occurrence seem to become highest at a certain road density, such as 4000m/grid. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>High nitrogen (N) external inputs can maximize maize yield but subsequent with a notably reduction in N use efficiency (NUE). Thus, it is necessary to identify the minimum effective N fertilizer input that does not affect maize grain yield (GY) and to investigate photosynthetic and root attributes basis of this optimal dose. We conducted a 4-year field experiment from 2014 to 2017 with four N application rates: 300 (N 300 ), 225 (N 225 ) 150 (N 150 ), and 0 kg ha &#8722;1 (N 0 ) in the Northwest of China. Lower GY was assessed by measuring photosynthetic and root system. Grain yield decreased by -3%, 7.7% and 21.9% with the N application rates decreased by 25%, 50% and 100% from 300 kg ha -1</ns0:p><ns0:p>. We found that the main reason for yield reduction driven by N reduction was the decreased RUE, WUE instead of intercepted photosynthetically active radiation and evapotranspiration. In the N 225 treatment, GY, water use efficiency (WUE), and radiation use efficiency (RUE) were not significantly reduced, or in some cases, were greater than those of the N 300 treatment, together with relevant photosynthetic and root attributes i.e. high net photosynthetic rate, stomatal conductance and root weight as well as deep root distribution. Our results concluded that application of nitrogen 225 kg ha &#8722;1 can increased yield by improving the RUE, WUE, and NUE under semi-arid regions.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>In the past four decades, global maize production has greatly increased <ns0:ref type='bibr' target='#b16'>(FAO, 2018)</ns0:ref> mainly due to application of nitrogen (N) fertilizer. Worldwide, excessive N fertilizer have been widely applied to achieve higher grain yield <ns0:ref type='bibr' target='#b12'>(Cui et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Meng et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b30'>Liang et al., 2020)</ns0:ref>. For example, the average dose of N fertilizer applied by the farmers was greater than 300 kg ha &#8722;1 (288&#177;113kg ha -1 ), which far exceeds the maize optimal N rates demonstrated in field experiments <ns0:ref type='bibr' target='#b63'>(Zhang et al., 2015a;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chang et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b61'>Yang et al., 2017)</ns0:ref>. N fertilizer was applied lavishly (350-600 kg ha &#8722;1 year &#8722;1 , <ns0:ref type='bibr'>Mueller et al., 2013)</ns0:ref> in an attempt to maximize yields in North China Plain. However, excessive application of N fertilizer has negative effects on crops, greatly reduces N use efficiency (NUE), causing great nitrate leaching losses (more than 50% N to the environment) and contamination of groundwater <ns0:ref type='bibr'>(Erisman et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b55'>Wang et al.,2014;</ns0:ref><ns0:ref type='bibr' target='#b34'>McBratney and Field, 2015;</ns0:ref><ns0:ref type='bibr' target='#b1'>Ahmad et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b51'>Suchy et al., 2018;</ns0:ref><ns0:ref type='bibr'>Wang et al., 2019)</ns0:ref>. Reducing N input rates from this level to 'moderate' levels in maize fields may improve NUE, maintain a fair level of grain yield of maize <ns0:ref type='bibr' target='#b67'>(Zhao et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b13'>Dai et al., 2015</ns0:ref><ns0:ref type='bibr' target='#b45'>, Qiang et al., 2019)</ns0:ref> as well as for its beneficial environmental impacts. Therefore, it is necessary to assess the extent to which the N fertilizer application rate is consistent with crop N requirements to maximize resource utilization and maintain relatively high grain yields <ns0:ref type='bibr' target='#b48'>(Robertson and Vitousek, 2009;</ns0:ref><ns0:ref type='bibr' target='#b64'>Zhang et al., 2015b)</ns0:ref>. Radiation interception and radiation use efficiency (RUE) form the basic framework for analyzing crop yield constraints. Variations in crop biomass due to abiotic factors may be attributed to intercepted photosynthetically active radiation (IPAR), RUE, or the combination of both <ns0:ref type='bibr'>IPAR and RUE (Fletcher et al., 2013)</ns0:ref>. Reduced leaf growth under low-N conditions is accompanied by a reduction in radiation interception <ns0:ref type='bibr' target='#b33'>(Massignam et al., 2012)</ns0:ref>. Low-N conditions primarily decrease the photosynthetic rate per unit area <ns0:ref type='bibr' target='#b54'>(Vos, Van Der Putten and Birch, 2005)</ns0:ref>, indicating that a low N effect on both leaf growth and photosynthetic rate may affect the final grain yield. Understanding that how maize production and resource use are affected by varying N application rates in favor of improving nitrogen fertilizer management to achieve optimal grain yield and resource use efficiency. The appropriate amount of N fertilizer input can improve utilization of precipitation during the crop season <ns0:ref type='bibr' target='#b13'>(Dai et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b17'>Herrera et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b61'>Yang et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Qiang et al., 2019)</ns0:ref>. Previous studies indicated that N fertigation effect on grain yield and resource use efficiency that often describe the effects of single resource utilization at the leaf or plant level <ns0:ref type='bibr' target='#b2'>(Brown, Jamieson and Moot, 2012)</ns0:ref>. The preceding study described the relationship between maize water use efficiency (WUE) and NUE in pot conditions <ns0:ref type='bibr'>(Wang et al., 2019)</ns0:ref>. However, there are few comprehensive studies on the effects of N fertilizer on the utilization of radiation, water, and N resources in field crops.</ns0:p><ns0:p>Enhancing photosynthesis is extensively accepted as critical to advancing the crop yield <ns0:ref type='bibr' target='#b37'>(Mu et al., 2017)</ns0:ref>. In order to understand the underlying mechanisms and differences in photosynthetic capacity, it is necessary to determine the relevant photosynthetic parameters of ear leaves <ns0:ref type='bibr' target='#b25'>(Li et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b66'>Zhang et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b22'>Lamptey et al., 2017)</ns0:ref>. Root morphology and distribution also play key roles in the acquisition of soil resources such as nutrients and water <ns0:ref type='bibr' target='#b46'>(Ristova and Busch, 2014;</ns0:ref><ns0:ref type='bibr' target='#b32'>Lynch, 2013;</ns0:ref><ns0:ref type='bibr' target='#b62'>Yu et al., 2014)</ns0:ref>. <ns0:ref type='bibr' target='#b36'>Mi et al. (2010)</ns0:ref> proposed a hypothesis about the ideotype root architecture of 'high yield and high N efficiency' in maize, which provided references for root research. In field conditions, the complexity of root sampling has limited efforts to understand the effects of N on roots to shoots. Such research has been conducted under pot conditions <ns0:ref type='bibr'>(Wang et al., 2019)</ns0:ref>; thus, it is not possible to assess variation due to differences in light and temperature conditions as well as nitrate N leaching, as occurs in field conditions. Limited knowledge about shoot and root traits related to maize grain yield, NUE, RUE and WUE under field conditions was investigated. Therefore, exploring the response of photosynthetic parameters and root development to grain yield reduction due to N reduction will provide an important reference for management of N fertilizer inputs.</ns0:p><ns0:p>In the current, we determined that N reduction results in maintenance of photosynthetic activity and root development to maximize grain yield and radiation, water, and N use by maize crops. The objectives of this study were to 1) investigate the effects of N reduction on water, radiation, and N use efficiencies in maize crops, and 2) determine the effects of N reduction on photosynthesis activity and root development in maize crops under semi-arid regions. Our finding provides a key data on enhanced maize production and resource use efficiency.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Field experiments</ns0:head><ns0:p>A 4-year field experiment was conducted on the same land at the agricultural experimental station of Northwest A&amp;F University in Yangling (34&#176;20&#8242;N, 108&#176;04&#8242;E, elevation, 455 m), <ns0:ref type='bibr'>Shaanxi province, China in 2014</ns0:ref><ns0:ref type='bibr'>, 2015</ns0:ref><ns0:ref type='bibr'>, 2016</ns0:ref><ns0:ref type='bibr'>and 2017.</ns0:ref> The experimental site experienced an annual average daylight of 2150 h, an annual average temperature of 12&#176;C to 14&#176;C, and an average annual precipitation of 581 mm. The annual mean temperature and total precipitation of the experimental area were shown in Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>. The soil of the experimental site is classified as a dark loess soil, and the former crop was winter wheat. Before sowing, soil chemical properties were analyzed in the top 60 cm of soil for organic matter content, nitrogen (N), phosphorus (P), and potassium (K) (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>The research work was carried out with a randomized complete block design having three replications. The total sub plot size was 39 m 2 (7.8 m long and 5 m wide). The row to row spacing was maintained at 60 cm and plant to plant at 25 cm. The maize seeds were planted manually and in each hill two seeds were sown at a depth of 5 cm in the middle of June during each growing seasons. Plants were thinned manually for normal plant population densities in the area of (67,500 pl. ha) at V3 (three leaf stage) <ns0:ref type='bibr' target='#b47'>(Ritchie and Hanway, 1982)</ns0:ref>. Plots were kept free of weeds, insects, and diseases; and the weeding was controlled by hand and hoe during each growing season. The N treatments applied in this study were: 1) no N (N 0 , 100% reduction from N 300 ); 2) 150 kg N ha &#8722;1 (N 150 , 50% reduction from N 300 ); 3) 225 kg N ha &#8722;1 (N 225 , 25% reduction from N 300 ); and 4) 300 kg N ha &#8722;1 (N 300, the traditional N dose applied by farmers in the Loess Plateau of China). Fertilizer N was sourced from urea (46% N), evenly split in the fractions of 1/2 at pre-sowing and side-banded deep (5cm) into the soil on the sowing rows of 1/2 at twelveleaf stage. A total of 150 kg of phosphorus (calcium superphosphate, P 2 O 5 16%) ha &#8722;1 and 150 kg of potassium (as potassium sulfate, K 2 O 45%) ha &#8722;1 were applied 5 days before sowing. Irrigation was applied at twelve-leaf stage (75mm). The amount of irrigation is controlled by a water meter (Zhejiang Ningbo Water Meter Co., Ltd.).</ns0:p></ns0:div> <ns0:div><ns0:head>Sampling and measurements Intercepted Photosynthetically Active Radiation (IPAR)</ns0:head><ns0:p>During the V6-R3, IPAR was measured every 5-7 days (on clear and sunny days between 11:00 AM to 2:00 PM) by using the SunScan Canopy Analyzer (Delta, UK). In each plot, three points were chosen. When measuring, the line sensor was placed horizontally between the two ridges (5 cm from the soil surface) and was used to collect three consecutive readings of transmitted photosynthetically active radiation (PAR) <ns0:ref type='bibr'>(Chen et al., 2016)</ns0:ref>. Round-trip observations were used to minimize error.</ns0:p></ns0:div> <ns0:div><ns0:head>Soil Water Content</ns0:head><ns0:p>At sowing and maturity, soil water content from the 0-200 cm layer were determined using a hand-held soil iron drill <ns0:ref type='bibr' target='#b65'>(Zhang et al., 2019)</ns0:ref>. In Between 0 and 20 cm, samples were collected every 10 cm, and between 20 and 200 cm, samples were collected every 20 cm. Soil samples were stored in a closed aluminum box and weighed, then oven-dried at 105&#176;C for 24h, and weighed separately. The soil water content of each plot was calculated from the average of three random soil core samples. The soil water content equals the difference between the fresh soil weight minus the dry soil weight divided by the dry soil weight.</ns0:p></ns0:div> <ns0:div><ns0:head>Photosynthetic Parameters and Leaf Area Index (LAI)</ns0:head><ns0:p>In 2016 and 2017, 20 plants were marked prior to the eight-leaf stage (V8). At V8, the tasseling stage (VT), milking stage (R3), and physiological maturity (R6), the three marked plants per plot were selected to measure net photosynthetic rate (Pn), intercellular CO 2 concentration (Ci), and stomatal conductance (Gs) on the ear leaves ( at VT, R3 and R6) or fully expanded leaves at the top of a plant (at V8) using a photosynthesis analyzer system LI-6400 (LI-COR, Lincoln, Nebraska, USA) on a clear sunny day between 9:00 AM to 11:00 AM <ns0:ref type='bibr' target='#b66'>(Zhang et al., 2017)</ns0:ref>. For measuring the green leaf area at V8, VT, R3, and R6 the same plants were used. The green leaf area index (LAI) was calculated as follows <ns0:ref type='bibr' target='#b3'>(Birch, Vos and Van Der Putten, 2003)</ns0:ref>: LAI=0.75&#215;Leaf length &#215;maximum width &#215;number of plants within a unit area of land /area of land</ns0:p></ns0:div> <ns0:div><ns0:head>Root System</ns0:head><ns0:p>Three plants root were sampled using the soil profile method <ns0:ref type='bibr' target='#b18'>(Holanda et al., 1998)</ns0:ref> at V8, VT, R3 and R6 stages of maize. Each root system was excavated from an area of 0.15 m 2 (0.6 m&#215; 0.25 m) around the center of the plant. Root sampling was conducted at depth intervals of 0-30 cm (surface soil layer), 30-60 cm (middle soil layer), and 60-90 cm (deep soil layer) in each plot. Excavated roots were immersed overnight in a plastic container filled with water and washed with tap water on a 0.25-mm screen until the roots were free of soil. Roots were scanned using an HP Scanjet 8200 scanner, and each root image was analyzed using a root analysis program (Regent Instruments Inc. WinRhizo Provision 5.0, Canada) to obtain the root surface area (cm 2 plant -1 ) and root length (mm). Root volume was measured by the drainage method. Root length density was calculated as the average of three plants' root lengths divided by the soil volume <ns0:ref type='bibr' target='#b27'>(Li et al., 2010)</ns0:ref>. Root samples were dried 48h at 70&#61616;C in an oven to obtain the root dry weight per plant.</ns0:p></ns0:div> <ns0:div><ns0:head>Biomass Yield, Shoot N Content, and Grain Yield</ns0:head><ns0:p>Four central rows were harvested randomly at 20 m 2 to measure the grain yield at harvest in each plot. Ten ears were randomly selected from each plot and threshed separately to determine moisture content and kernel number. Grain yield was estimated based on kernel weight and water content and expressed as 14% (w/w) moisture content. Six plants were sampled at each plot and were divided into leaves, stems, and grains. Before determining the N concentration, all plant parts were dried at 70&#61616;C for 48h and weighed. After weighed, the samples were ground into powder using a willey-type mill (&lt;1mm mesh), weigh a certain amount of sample (0.3-0.4 g) and were mineralized using H 2 SO 4 -H 2 O 2 and then total N concentration was obtained by an automatic Kjeldhal microdistillation analyzer (FOSS, Sweden, <ns0:ref type='bibr' target='#b41'>Nelson and Sommers, 1973)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Daily intercepted solar radiation was calculated using the following equation <ns0:ref type='bibr' target='#b31'>(Liu et al., 2014)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_0'>(1) LT = &#119875;&#119860;&#119877; &#119871; &#119875;&#119860;&#119877; &#119879;</ns0:formula><ns0:p>Where LT is the light transmission ratio, PAR L is the intercepted photosynthetically active radiation at the bottom of the canopy, and PAR T is the intercepted photosynthetically active radiation at the top of the canopy.</ns0:p><ns0:p>The measured intercepted photosynthetically active radiation value of the bottom layer was analyzed by two-dimensional interpolation over one day to obtain the intercepted photosynthetically active radiation rate of the entire canopy. Then, the intercepted photosynthetically active radiation rate obtained by interpolation analysis is multiplied by the incident PAR measured on the corresponding date by the meteorological observatory to determine the amount of canopy IPAR. Radiation use efficiency (RUE) was calculated using the following equation:</ns0:p><ns0:p>(</ns0:p><ns0:formula xml:id='formula_1'>2) &#119877;&#119880;&#119864; = &#119872;X&#8462; &#8721;&#119876; &#215; 10 -7 &#215; 100%</ns0:formula><ns0:p>Where &#8721;Q (MJ m -2 ) is the accumulated intercepted solar radiation and h (kJ kg -1 ) is the heat energy released of per kg of grain yield, M is the grain yield (kg ha -1 ) Water use efficiency (WUE) was calculated as:</ns0:p><ns0:formula xml:id='formula_2'>(3) WUE = &#119866;&#119884; &#119864;&#119879;</ns0:formula><ns0:p>Where GY (kg ha -1 ) is the grain yield, ET (mm) is the evapotranspiration was calculated as by soil water balance equation <ns0:ref type='bibr' target='#b19'>(Huang et al., 2005)</ns0:ref>. The following equations were also used:</ns0:p><ns0:formula xml:id='formula_3'>Internal N efficiency, (4) INE = &#119866;&#119884; &#119878;&#119873;&#119862; Agronomic N use efficiency, (5) ANE = (&#119866;&#119884; &#119873;&#119894; -&#119866;&#119884; &#119873;0 ) &#119873;&#119894; Apparent N recovery efficiency, (6) REN (%) = (&#119878;&#119873;&#119862; &#119873;&#119894; -&#119878;&#119873;&#119862; &#119873;0 ) &#119873;&#119894; &#215; 100 N harvest index, (7) NHI (%) = &#119866;&#119873;&#119862; &#119878;&#119873;&#119862; &#215; 100</ns0:formula><ns0:p>where SNC (kg ha -1 )is the shoot N content calculated as biomass (kg ha -1 ) multiplied by shoot N concentration (kg kg -1 ), i (N rates, kg ha -1 ) is 150, 225, or 300, GY Ni (kg ha -1 ) is grain yield in the N application plots, GY N0 (kg ha -1 ) is grain yield in the no-N application plots, and Ni is the N application rate. SNC Ni (kg ha -1 ) is the shoot N content in N application plots, SNC N0 (kg ha -1 ) is the shoot N content in the no-N application plots. GNC is grain N content (kg ha -1 ). The experimental data were organized and processed using Microsoft and are presented with standard error. SPSS18.0 (SPSS Institute Inc.) statistical analysis software was used for variance analysis. The data was checked for normality (Kolmogorov-Smirnov test) and homogeneity of variance (Bartlett-Box test). The effects of N rates, years and their interactions on the measured variables were tested using one-and two-way ANOVAs. To identify significant treatments effects, multiple comparisons among different treatments were performed using Duncan's multiple range test. Differences with P &lt; 0.05 were considered statistically significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Grain yield, Biomass Yield, and Crop Resource Utilization</ns0:head><ns0:p>Our results revealed that the year and N application rates showed significant effects on grain yield (GY), biomass yield (BY), grain N content (GNC), agronomic N use efficiency (ANE), apparent N recovery efficiency (REN), and N harvest index (NHI) (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). The interaction between N application rate and year had no significant effect on the above parameters. We observed that no significant differences in GY, BY, and GNC between the N 225 and N 300 treatments in all growing seasons. Compared with N 300 , the GY of N 150 and N 0 decreased by 4.7% to 13.6% and 19.7% to 22.8%, respectively, while the grain yield of treatment N 225 increased by 1.5% to 3.7%. Treatment N 300 increased biomass yield by 28% to 34%, compared with N 0 , while N 225 increased biomass yield by 27% to 32%. The GY difference among years may be due to the rainfall amount and seasonal distribution. Rainfall was 390 mm, 284 mm, 311 mm, and 371 mm in <ns0:ref type='bibr'>2014, 2015, 2016, and 2017</ns0:ref>, respectively (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Among the experimental years, GY was higher in 2016 and 2017 compared to 2014 and 2015. 2017 was a more suitable year for maize growth. Although the rainfall in 2016 was not much, the distribution was relatively uniform throughout the growth period. The early rainfall ensured the regularity and vegetative growth of maize seedlings. The lower GY in 2015 can be explained by the less rainfall. In 2014, the higher rainfall was mainly due to the large amount of rainfall occurring 70 days after sowing. Continuous rainfall from the silking to the flowering stage affected maize pollination, and severe stalk rot disease occurred during the grain filling stage, which caused prosenescence (Fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>), causing lower GY. Optimum rate of N application increased the AEN, REN, and NHI but except AEN of N 150 during the 2016 growing season. The NHI was highest for the N 225 and N 150 <ns0:ref type='bibr'>(2014, 2015 and 2017)</ns0:ref> or N 225 (2016) treatments.</ns0:p><ns0:p>Throughout the crop growth cycle, differences in ET among N treatments were significant (P&lt;0.05, Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). During growing seasons 2015, 2016, and 2017 we observed that the rate of N application showed no significant effect on IPAR. IPAR of N 0 was smaller than N 300 in the 2014 growing season. SNC was increased by increasing the rate of N application. For N 300 , SNC increased by from 73% to 92% compared with N 0 , and N 225 application increased SNC by from 68% to 81% compared with the N 0 treatment. IEN was increased with low rate N application. Application of nitrogen rate 225 kg ha -1 increased IEN of by from 8.5% to 12.3% compared with N 300 , while N 150 decreased IEN by 9.7% to 20.7% compared with N 300 . Nitrogen application at the rate of N 300 and N 225 were significantly similar, while N 300 increased RUE by 19% to 26%, whereas N 225 increased RUE by 23% to 29% compared with N 0 . Reduced N application was associated with reduced WUE. Our results showed that N 225 increased grain WUE by 25% to 26% compared to N 0 . WUE was also significantly increased in the N 300 treatment but to a lesser extent (19% -22%). RUE and NUE exhibited non-linear responses to the N application rate, indicating that the maximum grain yield can occur with N 150 or N 225 treatments. The differential grain yield between N 225 and N 300 was relatively small.</ns0:p></ns0:div> <ns0:div><ns0:head>Photosynthetic parameters and LAI</ns0:head><ns0:p>N application rates and year had significant effects on Pn, Gs, and Ci, but the interaction between N application rates and year did not display significant effects (Fig. <ns0:ref type='figure'>2</ns0:ref>). In both 2016 and 2017, the N 0 and N 150 treatments resulted in lower values of Pn and Gs when compared with N 225 and N 300 , however, the effect of N 225 was relatively greater at R6 than that of the N 300 treatment. This finding indicates that N fertilizer inputs can increase Gs and improve the photosynthetic capacity of maize crops. Conversely, the Pn and Gs of the N 300 treatment were reduced compared with N 225 . Pn of N 150 , N 225 and N 300 increased by 16%-81% across growth stages compared with N 0 .</ns0:p><ns0:p>N application rates significantly affected 1-Ci/Ca (P &lt; 0.05). The 1-Ci/Ca of the N 0 treatment was significantly lower than that of N 300 in all measurements (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>Overall, the leaf area index (LAI) was unaffected (P = 0.55) by N application rates with an average of 2.6 and 3.7 at the V8 and VT stages (Fig. <ns0:ref type='figure'>2</ns0:ref>). N application rates significantly affected LAI at the R3 and R6 stages. Compared with N 300 , the LAI of N 225 , N 150 and N 0 decreased by 1%, 6%, 11%, respectively, and 1%, 7%, and 13% at the R3 and R6 stages.</ns0:p></ns0:div> <ns0:div><ns0:head>Root System</ns0:head><ns0:p>The root dry weight reached a maximum value with the N 225 treatment. N 225 increased root dry weight by 35% to 67% compared with N 0 , while N 300 increased the root dry weight by 29% to 53% compared with N 0 at the V8, VT, R3, and R6 stages (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>). We observed clear differences in vertical distributions of roots as well as root morphology between N application rates, including root volume, root surface area, and root length density, which varied N application rates and years (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). The root system distributed in the surface layer (0-30 cm) accounted for more than 88 % of the root system after the VT stage for all treatments. At the V8 stage, the root ratio was proportional to the N application rate and increased significantly in the middle layer (30-60 cm). In the N 0 treatment, almost all roots were concentrated in the surface soil layer. The N 0 treatment exhibited significant increases in the deep-layer root ratio at the VT stage. The root ratio in the deep soil layer decreased with increasing N application after the VT stage. At the R3 and R6 stages, the surface root ratio increased with increasing N application. For the N 0 treatment, the deep-layer root ratio decreased after the VT stage. Similar to the dry root weight, root morphology indexes (root volume, surface area, and length density) increased initially and then decreased with increasing N application rate. These indexes reached their maximum values under the N 225 treatment. At the V8 stage, root morphology indexes were in the order of N 225 &gt; N 300 &gt; N 150 &gt; N 0 in the surface soil layer, which differed from the middle soil layer (N 300 &gt; N 225 &gt; N 150 &gt; N 0 ). Larger roots appeared in deep soil layers at the VT stage when compared with the V8 stage, and the maximum value for the N 0 treatment was observed in the middle and deep soil layer. At R3 and R6 stages, trends in root morphology indexes varied between soil layers; in the surface layer, the indexes followed the order N 225 &gt; N 300 &gt; N 150 &gt; N 0 ; in the middle layer, N 225 &gt; N 150 &gt; N 300 &gt; N 0 ; and in the deep layer, N 225 &gt; N 150 &gt; N 0 &gt; N 300 .</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Optimum rate of N application can produce a balance between crop demand and N supply and ensure maximum crop production while conserving resources and protecting against environmental damage <ns0:ref type='bibr' target='#b9'>(Ciampitti and Vyn, 2011;</ns0:ref><ns0:ref type='bibr' target='#b43'>Peng, Li and Fritschi, 2014)</ns0:ref>. Our results showed that during 2017 growing season, even when photosynthesis was significantly affected by a 50% reduction in N application (N 150 ), grain yield was not significantly decreased. In 2016, photosynthetic parameters decreased further for the N 150 and N 0 treatments, and the grain yield also decreased significantly. Over four years, we observed no significant reduction in grain yield for the N 225 treatment (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>), which even displayed a grain yield higher than that of the N 300 treatment, although the difference was not significant. Our research indicated that we could decrease the N dose by at least 25% without compromising grain production. Our results are similar to those of <ns0:ref type='bibr' target='#b29'>Li et al. (2007)</ns0:ref> and <ns0:ref type='bibr' target='#b22'>Lamptey et al. (2017)</ns0:ref>, who reported an optimal N fertilization range for summer maize of 200 to 300 kg N ha -1 . Under the N 0 and N 150 treatments, dry matter accumulation was limited, while the difference between N 300 and N 225 treatments was not significant. The reduction in crop yield induced by N reduction can be explained by various factors as following.</ns0:p><ns0:p>Regarding the mechanism by which maize yield is affected by N application rates, our results demonstrated that IPAR was not affected by N fertilizer application rates for most years, the finding similar to that described in previous studies <ns0:ref type='bibr' target='#b54'>(Vos, Van Der Putten and Birch, 2005;</ns0:ref><ns0:ref type='bibr' target='#b33'>Massignam et al., 2012)</ns0:ref>. The N application rates significantly affected RUE (P&lt;0.05). For N 0 and N 150 , the RUE was significantly less than that of N 300 . However, the RUE of the N 225 -treated crop was greater than that of the N 300 -treated crop. In this case, lower grain yield driven by lower N treatment corresponded to lower RUE. These results indicate that under such production conditions, N application rate mainly affects grain yield by affecting RUE rather than IPAR. <ns0:ref type='bibr' target='#b54'>Vos, Van Der Putten and Birch (2005)</ns0:ref> also found that maize tends to sacrifice specific leaf nitrogen and RUE while maintaining leaf area (small changes to leaf area index (LAI, Fig. <ns0:ref type='figure'>2</ns0:ref>) in comparison with the large decrease in Pn (Fig. <ns0:ref type='figure'>2</ns0:ref>) at low N application rates). Therefore, based on maintaining high IPAR, improving the RUE could be a valid way to achieve high grain yield of maize.</ns0:p><ns0:p>The actual ET involves two components, crop transpiration and soil evaporation. N application can increase ET during the reproductive period because of high leaf transpiration under high N conditions <ns0:ref type='bibr' target='#b22'>(Lamptey et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b49'>Rudnick et al., 2017)</ns0:ref>. The relatively low sensitivity of IPAR to N supply in maize maybe also consistent with low sensitivity of soil evaporation. In this case, the lower grain yields associated with the N 0 treatments corresponded to lower ET and WUE values. In addition, WUE of the N 300 -treated crop was lower than that of the N 225 -treated crop. This due to a higher sensitivity to soil water by the plant at higher N application rates. An increase in N application rates is usually accompanied by a decrease in NUE <ns0:ref type='bibr' target='#b21'>(Ju et al., 2015)</ns0:ref>. Significant effects of N were also observed on N use efficiency <ns0:ref type='bibr'>(NUE including AEN, REN, NHI and INE)</ns0:ref> in our study. Conversely, reducing the N application rate can achieve a balance between crop demand and N supply <ns0:ref type='bibr' target='#b22'>(Lamptey et al., 2017)</ns0:ref>. The desired N concentration of the plant, under N 0 treatment, has a high INE value, indicating that the biomass N concentration or plant yield are low and that the amount of N in the plant absorbed from the soil is small. N accumulation increased with increasing N application, but N accumulation in the grain was not significantly different between the N 225 and N 300 treatments, and the value of NHI at N 225 was greater than that at N 300 , which indicated that increased N in the plant did not transfer to the grain, resulting in residual N in the vegetative organs and excessive N absorption.</ns0:p><ns0:p>The plasticity of root morphology allows it to respond to soil mineral nutrients <ns0:ref type='bibr' target='#b44'>(Peng, Li and Li, 2012;</ns0:ref><ns0:ref type='bibr' target='#b62'>Yu et al., 2014)</ns0:ref>. We found that root dry weight reached a maximum with the N 225 treatment, which suggested that the relationship between N input and the root system is not linear and positive; N input may even have a negative impact on root growth and development. In the present study, application of N fertilizer promoted growth of roots in the 0-60 cm soil layer and increased the proportion of roots in this layer, indicating that N application improved growth of the surface and middle layer roots. In the late growth stage (after VT), N 0 treatment exhibited a negative effect on the root dry weight and proportion in the 30-60 cm and 60-90 cm soil layers, indicating that N deficiency would be detrimental during the accelerated aging of deep layer roots. Not only are root morphology and nutrient absorption closely related, but the spatial distribution of roots is also closely connected to crop growth and productivity <ns0:ref type='bibr' target='#b36'>(Mi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b32'>Lynch, 2013)</ns0:ref>. We observed that differences in root spatial distribution among N treatments. In both years, the root system exhibited the optimal distribution under N 225 treatment, with a higher root length density in the observed soil layer, resulting in larger and deeper infiltration scales. Slower root senescence in the N 225 treatment is also a major contributor to N rate-induced increases in grain yield. Studies have shown that the relative stability of deep root environment is beneficial in promoting the buffer capacity of the root system in adverse soil environments and achieving high grain yield and resource use efficiency <ns0:ref type='bibr' target='#b7'>(Chen et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b36'>Mi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b58'>Wasson et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b50'>Saengwilai et al., 2014)</ns0:ref>. The results of our study during 2016 and 2017 growing seasons showed that excessive N (N 300 ) application negatively affects early deep root growth compared with N 225 . High external N input also appeared to generate an overall inhibitory effect on later root growth. There are many reasons for the observed reduction in crop yield. Slight reductions in crop yield induced by excessive N application may be due to negative impacts on root growth during the early growth stage or may be caused by differing mechanisms of aging leading to N loss and relative N deficiency during the reproductive stage. N deficiency induces root thinning and increases longitudinal expansion by promoting root growth in the underlying soil, while high N inhibits vertical expansion of roots <ns0:ref type='bibr' target='#b53'>(Trachsel et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b39'>Mu et al., 2015)</ns0:ref>. This study explained the effect of excessive N and N deficiency on yield from the perspective of root morphology and growth. The role of N in grain formation is mainly explained by photosynthesis and N reduction usually negatively impacts photosynthetic performance in maize <ns0:ref type='bibr' target='#b33'>(Massignam et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b42'>Olszewski et al., 2014)</ns0:ref>. Although N 150 treatment significantly reduced Pn in the 2017 growing season, the grain yield for the N 150 treatment was not significantly reduced, indicating that the plant's transient photosynthetic parameters were more sensitive than the dry mass factors in responding to changes in N application rates. Our results illustrate the physiological basis for utilizing the 225 kg ha -1 N rate could improve the stress resistance of summer maize plants in the Loess Plateau. For the N 300 treatment, decreased Pn was due to lower Gs, which may be due to Nassociated increased sensitivity to soil water and results in lower WUE than the N 225 treatment at the R6 stages. In the low N treatment, maize plants showed fewer leaf rolling under soil waterstress compared with the high N treatment, and thus obtained higher WUE <ns0:ref type='bibr'>(Wang et al., 2019)</ns0:ref>.</ns0:p><ns0:p>At different measurement dates, the maximum value of Gs was not always associated with the N 300 treatment (Fig. <ns0:ref type='figure'>2</ns0:ref>). In general, plants' water requirements are expected be greater under high N conditions. Thus, high-N fertilizer plots are more likely to be water-deficient if the soil moisture is inadequate, which would aggravate the plant's stomatal limitations (1-Ci/Ca) and reduce Gs and Pn. Similar findings have been described in wheat <ns0:ref type='bibr' target='#b66'>(Zhang et al., 2017)</ns0:ref>. Also, previous studies pointed out that under soil water-stress, ABA (as the signal carrier) transmitted to the shoot, reducing the Gs and increasing stomatal limitations in the initial drought <ns0:ref type='bibr' target='#b23'>(Larcher, 2003;</ns0:ref><ns0:ref type='bibr' target='#b28'>Li and Xu, 2014;</ns0:ref><ns0:ref type='bibr' target='#b59'>Yan et al., 2017)</ns0:ref>. The inadequate soil moisture of N 300 treatment is not serious enough to affect non-stomatal factors in restricting photosynthetic carbon assimilation. The Pn of the N 225 treatment was significantly similar to that of the N 300 treatment during the VT stage, but it was less than that of the N 225 treatment in R6. Although the N 225 treatment showed a higher photosynthetic rate in the R6 than the N 300 treatment, but the ratio did not make it achieve significantly higher yield, which may be caused by the small contribution rate of high photosynthetic capacity to a grain yield <ns0:ref type='bibr' target='#b0'>(Acciaresi et al., 2014)</ns0:ref>. In addition, the photosynthetic capacity may also be related to the differences of N nutrition characteristics among different maize genotypes <ns0:ref type='bibr' target='#b55'>(Chen et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b26'>Li et al., 2015)</ns0:ref>. Thus, the application rate of 225 kg N ha &#8722;1 was maybe still high, the current N rate was relatively effective in improving resource use efficiencies. Under production conditions, large amounts of N input is often used as an 'insurance' against higher yields to ensure further increases in maize production <ns0:ref type='bibr' target='#b20'>(Huang et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b6'>Chen et al., 2012)</ns0:ref>. However, this behavior results in a significant reduction in NUE <ns0:ref type='bibr' target='#b7'>(Chen et al., 2010)</ns0:ref> and a slight reduction in RUE and WUE. Previously, Chen et al. ( <ns0:ref type='formula'>2016</ns0:ref>) and <ns0:ref type='bibr' target='#b40'>Mu et al. (2018)</ns0:ref> reported that N is used primarily for cell morphogenesis, that N in leaves is N redundant, and excess N is mainly stored in soluble protein and light-harvesting pigment-protein complexes. Therefore, in our study, the grain yield of N 225 treatment did not display significantly different results compared to the N 300 treatment, and the root system and photosynthetic capacity showed certain advantages. From these findings, we can conclude that it is achievable to improve resource use efficiencies while ensuring grain yield. Actually, maize genotypes and soil moisture also affect GY and physiological characteristics <ns0:ref type='bibr' target='#b10'>(Ciampitti et al., 2013)</ns0:ref>. In the current experiment, these factors are not taken into consideration. So in the future study we should focus on the effects of genotype aiming to maximize GY and resource use efficiency.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Decreased grain yield due to N reduction was driven by reduced radiation utilization efficiency and water use efficiency; the impact of radiation interception and total water evapotranspiration were relatively small. An application rate of 225 kg N ha &#8722;1 could be used as a reference for optimal N application in the Loess Plateau of China. This N application rate optimized the ecophysiological responses of the plant, a finding which was confirmed by measuring photosynthetic activity and root system. This response to optimizing N input resulted in higher grain yield, RUE, WUE, and NUE. Reducing N application rates has strong recoverability in maize production and can maximize the capture and utilization of resources, increasing the maize grain yield. Layer (cm) Organic matter (g kg -1 )</ns0:p><ns0:p>Total nitrogen (g kg -1 )</ns0:p><ns0:p>Alkaline nitrogen (mg kg -1 )</ns0:p><ns0:p>Available phosphorus (mg kg -1 ) Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Effect of nitrogen application rates on root volume, root surface and root length density Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Means followed by different lowercase letters within each column indicate significantly different (p &lt; 0.05). N 300 , N 225 , N 150 and N 0 represent application of nitrogen at a rate of 300</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Effects of nitrogen application rates on grain yield (GY), biomass yield (BY), grain nitrogen content (GNC), N harvest index (NHI) and agronomic N use efficiency (AEN),</ns0:figDesc><ns0:table><ns0:row><ns0:cell>apparent N recovery efficiency (RNE)</ns0:cell></ns0:row><ns0:row><ns0:cell>Means followed by different lowercase letters indicate significantly different (p &lt; 0.05) within</ns0:cell></ns0:row><ns0:row><ns0:cell>the same column and the same year. N 300 , N 225 , N 150 and N 0 represent application of nitrogen</ns0:cell></ns0:row><ns0:row><ns0:cell>at a rate of 300, 225, 150 and 0 kg ha -1 . * Significant at the 0.05 probability level. **</ns0:cell></ns0:row><ns0:row><ns0:cell>Significant at the 0.01 probability level. ns, non-significant</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Effect of nitrogen application rates on evapotranspiration (ET), the accumulated intercepted solar radiation (IPAR), shoot nitrogen content (SNC), internal N use efficiency (INE), radiation use efficiencies (RUE), and water use efficiency(WUE) </ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='3'>Means followed by different lowercase letters within each column indicate significantly</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>different (p &lt; 0.05). N 300 , N 225 , N 150 and N 0 represent application of nitrogen at a rate of 300,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>225, 150 and 0 kg ha -1</ns0:cell><ns0:cell>. ET, total evapotranspiration (mm); IPAR, the accumulated intercepted</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>solar radiation (MJ m -2 ), SNC, shoot nitrogen content (kg ha</ns0:cell><ns0:cell>-1 ), IEN, internal N use efficiency</ns0:cell></ns0:row><ns0:row><ns0:cell>(kg kg -1</ns0:cell><ns0:cell cols='2'>); RUE, radiation use efficiency; WUE, water use efficiency. * Significant at the 0.05</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>probability level.</ns0:cell></ns0:row></ns0:table><ns0:note>** Significant at the 0.01 probability level. ns, non-significant.PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Effect of nitrogen application rates on root dry matter and root ratio</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Means followed by different lowercase letters within each column indicate significantly</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>different (p &lt; 0.05). N 300 , N 225 , N 150 and N 0 represent application of nitrogen at a rate of 300,</ns0:cell></ns0:row><ns0:row><ns0:cell>225, 150 and 0 kg ha -1</ns0:cell><ns0:cell>.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 5 (on next page)</ns0:head><ns0:label>5</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head /><ns0:label /><ns0:figDesc>Root volume (cm 3 plant -1 ) Root surface (cm 2 plant -1 ) Root length density (mm cm -3 )</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Year Soil layer</ns0:cell><ns0:cell>Nitrogen rate</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>V8</ns0:cell><ns0:cell>VT</ns0:cell><ns0:cell>R3</ns0:cell><ns0:cell>R6</ns0:cell><ns0:cell>V8</ns0:cell><ns0:cell>VT</ns0:cell><ns0:cell>R3</ns0:cell><ns0:cell>R6</ns0:cell><ns0:cell>V8</ns0:cell><ns0:cell>VT</ns0:cell><ns0:cell>R3</ns0:cell><ns0:cell>R6</ns0:cell></ns0:row><ns0:row><ns0:cell>2016</ns0:cell><ns0:cell>0-30</ns0:cell><ns0:cell>N</ns0:cell><ns0:cell>31.2a</ns0:cell><ns0:cell>80.7ab</ns0:cell><ns0:cell>80.7a</ns0:cell><ns0:cell>68.6b</ns0:cell><ns0:cell>306.19a</ns0:cell><ns0:cell>536.5b</ns0:cell><ns0:cell>695.8b</ns0:cell><ns0:cell>569.3b</ns0:cell><ns0:cell>0.86a</ns0:cell><ns0:cell>1.24b</ns0:cell><ns0:cell>0.82b</ns0:cell><ns0:cell>1.52b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>37.7a</ns0:cell><ns0:cell>91.2a</ns0:cell><ns0:cell>89.9a</ns0:cell><ns0:cell>76.6a</ns0:cell><ns0:cell>339.78a</ns0:cell><ns0:cell>611.6a</ns0:cell><ns0:cell>744.9a</ns0:cell><ns0:cell>639.5a</ns0:cell><ns0:cell>0.98a</ns0:cell><ns0:cell>1.54a</ns0:cell><ns0:cell>2.30a</ns0:cell><ns0:cell>1.78a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>17.3b</ns0:cell><ns0:cell>86.8b</ns0:cell><ns0:cell>64.0b</ns0:cell><ns0:cell>52.4c</ns0:cell><ns0:cell>167.56b</ns0:cell><ns0:cell>652.4a</ns0:cell><ns0:cell>563.5c</ns0:cell><ns0:cell>472.6c</ns0:cell><ns0:cell>0.50b</ns0:cell><ns0:cell>1.23b</ns0:cell><ns0:cell>1.54c</ns0:cell><ns0:cell>1.11c</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N 0</ns0:cell><ns0:cell>10.7c</ns0:cell><ns0:cell>54.0c</ns0:cell><ns0:cell>31.7c</ns0:cell><ns0:cell>26.9d</ns0:cell><ns0:cell>132.53c</ns0:cell><ns0:cell>405.1a</ns0:cell><ns0:cell>281.8d</ns0:cell><ns0:cell>228.5d</ns0:cell><ns0:cell>0.40b</ns0:cell><ns0:cell>0.90c</ns0:cell><ns0:cell>0.70d</ns0:cell><ns0:cell>0.53d</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>30-60</ns0:cell><ns0:cell>N</ns0:cell><ns0:cell>1.9a</ns0:cell><ns0:cell>6.7b</ns0:cell><ns0:cell>6.2a</ns0:cell><ns0:cell>2.9b</ns0:cell><ns0:cell>10.53a</ns0:cell><ns0:cell>101.4b</ns0:cell><ns0:cell>73.9b</ns0:cell><ns0:cell>51.1b</ns0:cell><ns0:cell>0.05a</ns0:cell><ns0:cell>0.34bc</ns0:cell><ns0:cell>0.48a</ns0:cell><ns0:cell>0.18b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>0.8b</ns0:cell><ns0:cell>9.0a</ns0:cell><ns0:cell>6.7a</ns0:cell><ns0:cell>5.5a</ns0:cell><ns0:cell>8.94a</ns0:cell><ns0:cell>117.4a</ns0:cell><ns0:cell>138.0a</ns0:cell><ns0:cell>100.2a</ns0:cell><ns0:cell>0.03a</ns0:cell><ns0:cell>0.39b</ns0:cell><ns0:cell>0.54a</ns0:cell><ns0:cell>0.4a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>0.1c</ns0:cell><ns0:cell>4.2c</ns0:cell><ns0:cell>3.3b</ns0:cell><ns0:cell>2.6b</ns0:cell><ns0:cell>0.97b</ns0:cell><ns0:cell>88.5b</ns0:cell><ns0:cell>90.8b</ns0:cell><ns0:cell>66.5b</ns0:cell><ns0:cell>0.003b</ns0:cell><ns0:cell>0.31c</ns0:cell><ns0:cell>0.44b</ns0:cell><ns0:cell>0.35a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N 0</ns0:cell><ns0:cell>0.06c</ns0:cell><ns0:cell>7.1b</ns0:cell><ns0:cell>2.9b</ns0:cell><ns0:cell>1.9c</ns0:cell><ns0:cell>0.17c</ns0:cell><ns0:cell>112.5a</ns0:cell><ns0:cell>14.7c</ns0:cell><ns0:cell>36.5c</ns0:cell><ns0:cell>0.000b</ns0:cell><ns0:cell>0.46a</ns0:cell><ns0:cell>0.22c</ns0:cell><ns0:cell>0.13b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>60-90</ns0:cell><ns0:cell>N</ns0:cell><ns0:cell /><ns0:cell>1.1b</ns0:cell><ns0:cell>2.2b</ns0:cell><ns0:cell>0.9c</ns0:cell><ns0:cell /><ns0:cell>11.1c</ns0:cell><ns0:cell>44.1c</ns0:cell><ns0:cell>26.2c</ns0:cell><ns0:cell /><ns0:cell>0.01c</ns0:cell><ns0:cell>0.19c</ns0:cell><ns0:cell>0.07b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell /><ns0:cell>3.2a</ns0:cell><ns0:cell>5.2a</ns0:cell><ns0:cell>3.9a</ns0:cell><ns0:cell /><ns0:cell>37.4b</ns0:cell><ns0:cell>118.0a</ns0:cell><ns0:cell>87.9a</ns0:cell><ns0:cell /><ns0:cell>0.17b</ns0:cell><ns0:cell>0.39a</ns0:cell><ns0:cell>0.24a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell /><ns0:cell>1.5b</ns0:cell><ns0:cell>3.1b</ns0:cell><ns0:cell>2.6b</ns0:cell><ns0:cell /><ns0:cell>26.1b</ns0:cell><ns0:cell>91.0b</ns0:cell><ns0:cell>72.1b</ns0:cell><ns0:cell /><ns0:cell>0.07c</ns0:cell><ns0:cell>0.25b</ns0:cell><ns0:cell>0.21a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N 0</ns0:cell><ns0:cell /><ns0:cell>4.6a</ns0:cell><ns0:cell>2.0b</ns0:cell><ns0:cell>1.0c</ns0:cell><ns0:cell /><ns0:cell>118.0a</ns0:cell><ns0:cell>47.9c</ns0:cell><ns0:cell>31.3c</ns0:cell><ns0:cell /><ns0:cell>0.35a</ns0:cell><ns0:cell>0.12c</ns0:cell><ns0:cell>0.09b</ns0:cell></ns0:row><ns0:row><ns0:cell>2017</ns0:cell><ns0:cell>0-30</ns0:cell><ns0:cell>N</ns0:cell><ns0:cell>36.8a</ns0:cell><ns0:cell>92.3b</ns0:cell><ns0:cell>85.6b</ns0:cell><ns0:cell>71.6b</ns0:cell><ns0:cell>428.83b</ns0:cell><ns0:cell>624.4b</ns0:cell><ns0:cell>726.8a</ns0:cell><ns0:cell>611.8b</ns0:cell><ns0:cell>1.08b</ns0:cell><ns0:cell>1.37b</ns0:cell><ns0:cell>1.77b</ns0:cell><ns0:cell>1.29b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>42.9a</ns0:cell><ns0:cell>112.9a</ns0:cell><ns0:cell>103.8a</ns0:cell><ns0:cell>94.3a</ns0:cell><ns0:cell>520.53a</ns0:cell><ns0:cell>683.4a</ns0:cell><ns0:cell>779.8a</ns0:cell><ns0:cell>690.1a</ns0:cell><ns0:cell>1.12a</ns0:cell><ns0:cell>1.74a</ns0:cell><ns0:cell>3.05a</ns0:cell><ns0:cell>2.67a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell cols='2'>21.2b 100.4a</ns0:cell><ns0:cell>73.9b</ns0:cell><ns0:cell>63.5b</ns0:cell><ns0:cell>248.31c</ns0:cell><ns0:cell>690.1a</ns0:cell><ns0:cell>609.6b</ns0:cell><ns0:cell>487.3c</ns0:cell><ns0:cell>0.56c</ns0:cell><ns0:cell>1.43b</ns0:cell><ns0:cell>1.92b</ns0:cell><ns0:cell>1.54b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N 0</ns0:cell><ns0:cell>11.8c</ns0:cell><ns0:cell>58.4c</ns0:cell><ns0:cell>38.4c</ns0:cell><ns0:cell>32.6c</ns0:cell><ns0:cell>166.37d</ns0:cell><ns0:cell>560.2c</ns0:cell><ns0:cell>345.4c</ns0:cell><ns0:cell>286.7d</ns0:cell><ns0:cell>0.52d</ns0:cell><ns0:cell>1.00c</ns0:cell><ns0:cell>0.97c</ns0:cell><ns0:cell>0.58c</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>30-60</ns0:cell><ns0:cell>N</ns0:cell><ns0:cell>2.7a</ns0:cell><ns0:cell>8.3b</ns0:cell><ns0:cell>7.2b</ns0:cell><ns0:cell>3.3b</ns0:cell><ns0:cell>11.37a</ns0:cell><ns0:cell>115.7a</ns0:cell><ns0:cell>100.0b</ns0:cell><ns0:cell>64.2c</ns0:cell><ns0:cell>0.05a</ns0:cell><ns0:cell>0.49b</ns0:cell><ns0:cell>0.57a</ns0:cell><ns0:cell>0.33b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>1.2b</ns0:cell><ns0:cell>11.0a</ns0:cell><ns0:cell>8.4a</ns0:cell><ns0:cell>6.6a</ns0:cell><ns0:cell>14.15a</ns0:cell><ns0:cell>133.0b</ns0:cell><ns0:cell>168.5a</ns0:cell><ns0:cell>146.1a</ns0:cell><ns0:cell>0.06a</ns0:cell><ns0:cell>0.53a</ns0:cell><ns0:cell>0.60a</ns0:cell><ns0:cell>0.52a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell>0.4c</ns0:cell><ns0:cell>5.5c</ns0:cell><ns0:cell>4.4bc</ns0:cell><ns0:cell>3.4b</ns0:cell><ns0:cell>1.91b</ns0:cell><ns0:cell>110.4a</ns0:cell><ns0:cell>110.2b</ns0:cell><ns0:cell>96.0b</ns0:cell><ns0:cell>0.02b</ns0:cell><ns0:cell>0.44b</ns0:cell><ns0:cell>0.53b</ns0:cell><ns0:cell>0.48a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N 0</ns0:cell><ns0:cell>0.3c</ns0:cell><ns0:cell>8.5b</ns0:cell><ns0:cell>3.8c</ns0:cell><ns0:cell>2.6b</ns0:cell><ns0:cell>0.34c</ns0:cell><ns0:cell>145.5b</ns0:cell><ns0:cell>29.0c</ns0:cell><ns0:cell>54.9c</ns0:cell><ns0:cell>0.03b</ns0:cell><ns0:cell>0.58a</ns0:cell><ns0:cell>0.33c</ns0:cell><ns0:cell>0.25c</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>60-90</ns0:cell><ns0:cell>N</ns0:cell><ns0:cell /><ns0:cell>1.4b</ns0:cell><ns0:cell>3.2b</ns0:cell><ns0:cell>1.1c</ns0:cell><ns0:cell /><ns0:cell>29.3d</ns0:cell><ns0:cell>76.2b</ns0:cell><ns0:cell>49.1c</ns0:cell><ns0:cell /><ns0:cell>0.09c</ns0:cell><ns0:cell>0.29c</ns0:cell><ns0:cell>0.10bc</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell /><ns0:cell>4.8a</ns0:cell><ns0:cell>6.3a</ns0:cell><ns0:cell>4.3a</ns0:cell><ns0:cell /><ns0:cell>84.3b</ns0:cell><ns0:cell>145.7a</ns0:cell><ns0:cell>126.3a</ns0:cell><ns0:cell /><ns0:cell>0.30b</ns0:cell><ns0:cell>0.47a</ns0:cell><ns0:cell>0.31a</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N</ns0:cell><ns0:cell /><ns0:cell>2.3b</ns0:cell><ns0:cell>3.7b</ns0:cell><ns0:cell>3.1b</ns0:cell><ns0:cell /><ns0:cell>46.2c</ns0:cell><ns0:cell>101.1b</ns0:cell><ns0:cell>87.9b</ns0:cell><ns0:cell /><ns0:cell>0.13c</ns0:cell><ns0:cell>0.36b</ns0:cell><ns0:cell>0.30b</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>N 0</ns0:cell><ns0:cell /><ns0:cell>5.5a</ns0:cell><ns0:cell>3.0b</ns0:cell><ns0:cell>2.0c</ns0:cell><ns0:cell /><ns0:cell>145.1a</ns0:cell><ns0:cell>60.2c</ns0:cell><ns0:cell>36.3c</ns0:cell><ns0:cell /><ns0:cell>0.44a</ns0:cell><ns0:cell>0.20c</ns0:cell><ns0:cell>0.10c</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:50071:1:1:NEW 19 Aug 2020)</ns0:note></ns0:figure> </ns0:body> "

"Chinese Institute of Water-saving Agriculture Northwest A&F University Yangling 712100, Shaanxi, China. E-Mail: [email protected] August 31th, 2020 Dear editor, We appreciate being given the opportunity to revise our manuscript (Article title: Nitrogen fertigation affects maize grain yield through regulating nitrogen uptake, radiation and water use efficiency, photosynthetic and root distribution, Article ID: 50071). We would like to express our sincere thanks to the reviewers for your constructive and valuable comments, which are very helpful for improving the quality of our manuscript, and for guiding our future research. We have revised the text with the changes described below, point by point in response to each comment. We included the raw data in the supplemental file and adjusted the figures’ format, in addition, we merged Figure 2 and 3 as Figure 2 (there are two Figures in the manuscript). And we added Table 1 to show the soil chemical properties (0–60cm depth) more clearly. We sincerely hope that the revised version of the manuscript is now suitable for publication in your journal. If you have any queries, please do not hesitate to contact us, and we will do our best to revise the manuscript until you are satisfied. Qingfang Han On behalf of all authors. Editor comments: MINOR REVISIONS This paper is about N and water use. Authors have presented some valuable findings, while more details are needed in several places before it can be accepted. A general introduction to the N fertilizer in the world, not just in Loess Plateau is required. Some details about photosynthetic characteristics and N leaching are also needed. Regarding the differences between years, the potential effect of precipitation should be discussed. Response: Many thanks for your kind advice, as you suggested that we focus on an issue on worldwide, not just in Loess Plateau in Introduction section. We have added the reasons for the difference between years in the Results section and photosynthetic characteristics related to soil moisture and maize varieties in the Discussion section. Please check it at your convenience. Once again, we would like to express our sincere thanks for your constructive and valuable comments. Reviewer 1: 1. Title, please change fertigation to fertilization. Response: Agreed. We replaced ‘fertigation’ with ‘fertilization’. 2. Abstract, please revise the first sentence. Response: We apologize for the confusion caused by the inappropriate description. We have revised the sentences to ‘High nitrogen (N) external inputs can maximize maize yield but subsequent with a notably reduction in N use efficiency (NUE).’ 3. IN, Line 35-47 as a manuscript for shooting a international journal, I would suggest not using an local issue to start your introduction, you have to focus on an issue worldwide or regional problem in some countries at least. So please rewrite this part. Response: Many thanks for your kind advice, as you suggested we rewrite this part and focus on an issue on worldwide. 4. MM, Line 115-137 please add refs for these indicators measurements. Response: Many thanks for your kind advice, we added references for these indicators measurements in the Material and Methods and References sections. 5. Results and discussion were ok with the description. Response: Special thanks to you for your good comments and affirmations for this manuscript. 6. Other revisions. We checked through the manuscript carefully and have made some additional revisions in our revised manuscript. For example, we adjusted the Figure 1 format and merged Figure 2 and 3 as Figure 2. And we added Table 1 to show the soil chemical properties (0–60cm depth) more clearly. Besides, we have added some details in Material and Methods in our revised manuscript. Please check it at your convenience. Once again, we would like to express our sincere thanks to the reviewers for your constructive and valuable comments. Reviewer 2: Comments for the Author This is a well-organized article. I have a little doubt that photosynthetic characteristics may be related to soil moisture and maize varieties. Please add this part to the discussion. Response: Many thanks for your kind advice and affirmation of this manuscript. The questions you raised are indeed what we have overlooked. According to your suggestions, we have added corresponding content in the Discussion section. Again, thanks for your comments which are very helpful for guiding our manuscript and future research. Other revisions. We checked through the manuscript carefully and have made some additional revisions in our revised manuscript. For example, we adjusted the figure 1 format and merged Figure 2 and 3 as Figure 2. And we added Table 1 to show the soil chemical properties (0–60cm depth) more clearly. Besides, we have added some details in Material and Methods in our revised manuscript and added some references in the Material and Methods. Please check it at your convenience. Reviewer 3: 1. Basic reporting (1) Line 37, although leaching is not a focus in this study, but I feel like it is important to point it out as the excessive N fertilizer not only reducing NUE but causing great leaching (belowground water contamination) Response: Thank you for your valuable advice. According to your suggestions, we have added relevant research progress in Introduction section. (2) Line 39, omit 'application rates vary greatly'? since the point of this sentence is to emphasize the excessive N input. Response: Thank you very much for your reminder. We quite agree with you and delete 'and application rates vary greatly'. (3) Line 41, what is approximate level of maize demand? Response: The average dose of N fertilizer applied by the farmers far exceeds the maize optimal N rates demonstrated by field experiments, Yang et al. (2017) suggested proper N application rates for summer maize of 180−200 kg N ha −1. N aboveground uptake requirement per 1 Mg grain yield, with optimized N treatment, was on 17.0 to 19.8 kg Mg −1 grain for maize (Meng et al., 2016). (4) Line 43, probably also mention about reducing the N leaching, which is a big problem in Ag science. Response: Thank you for your valuable advice. According to your suggestions, we have added relevant research progress in Introduction section. (5) Line 89, Line 92, '455 m' and '581 mm' are good enough, no need to have one more digit. Response: Thank you for your pointing out the inappropriate digital form. (6) Line 94-97, different font size Response: Many thanks for your kind reminder. We adjusted the font size. (7) Line 99, m2, superscript Response: Many thanks for your kind reminder. (8) Line 249-263, please keep consistent about the format of 'N0, N150, N225, N300' throughout the manuscript Response: We are very sorry for our carelessness. (9) Figure 1, possibly move to the Appendix since this was a background information, and was not used to explain Results in the Discussion. And only one legend is needed instead of four in four panels. Response: Many thanks for your kind advice, as you suggested that we adjust the Figure 1 and in order to explain the difference among years we used to explain in the Results, so we do not move to the Appendix. Thanks for your comments which are very helpful for guiding our manuscript. (10) Figure 2 and 3 can be possibly merged since they are sharing the same experiential settings. Response: Agreed. We have merged Figure 2 and 3 as Figure 2. Please check it. (11) Table 4, last row has five cells displayed in square? Response: We are very sorry, this is an error when changing the format, we have made changes, thank you for your reminder. 2. Experimental design (1) Line 95, were the top 60 cm all organic soils? or having a layer of mineral soils (A horizon)? Response: Thank you for pointing out the confusion caused by our unclear description. In order to show the soil chemical properties more clearly, an analysis of soil samples (0–60cm depth) were showed in Table 1. (2) Line 160, better to write out the method Response: Agree. (3) Line 163, no need to mention Excel. Response: Agree. We delete it. (4) Line 163, 'presented as SE?' should be 'presented with SE'? Response: Agree. (5) Line 164-166, these sentences about ANOVA tests should be moved down to line 195, after calculation of all the index. And the description needs more details, for example, authors also included year as an independent variable in the test, which was not described here. Response: Agree. Many thanks for your kind reminder. We moved these sentences to the end of the Statistical analysis section. “The experimental data were organized and processed using Microsoft and are presented with standard error. SPSS18.0 (SPSS Institute Inc.) statistical analysis software was used for variance analysis. The data was checked for normality (Kolmogorov–Smirnov test) and homogeneity of variance (Bartlett–Box test). The effects of N rates, years and their interactions on the measured variables were tested using one- and two-way ANOVAs. To identify significant treatments effects, multiple comparisons among different treatments were performed using Duncan’s multiple range test. Differences with P < 0.05 were considered statistically significant.” 3. Validity of the findings (1) Line 201, were the following sentences saying there was an interaction impact? as the treatment effect was not consistent across years. Response: We are sorry that these statements have caused a misunderstanding. We performed an analysis of variance on the data, and the results showed that the interaction was not significant. We re-describe the experimental results here to eliminate misunderstandings. (2) Line 228-235, were these statements true in all four years? Response: These statements are the result of 2016 and 2017. (3) Line 243 and line 246, were these two sentences conveying the same idea? omit the second sentence? Response: Yes. We rewrite these two sentences. (4) Line 265, this first (topic) sentence was too long and included two messages. I suggested starting from 'Optimum rate...' Response: Agree. (5) Line 282, reconstruct the sentences. Technically, the N0 is all one type of N application rates. Response: Agree. Rewrite it. (6) Line 293, line 301, these two topic sentences read like a Result, please edit them. Response: Agree. (7) What are the reasons of seeing different treatment effects across years? Probably due to the variations in precipitation (amount and seasonal distribution) and temperature that shown in Figure 1? Authors should add possible explanation in the Discussion. Response: Many thanks for your kind advice, as you suggested that we added the reasons for the difference in precipitation (amount and seasonal distribution) between the years in the Results. 4. Other revisions. We checked through the manuscript carefully and have made some additional revisions in our revised manuscript. For example we have added some details in Material and Methods in our revised manuscript and added some references in the Material and Methods. Please check it at your convenience. Again, special thanks to you for your good comments. We tried our best to improve the manuscript and made some changes in the manuscript. We sincerely hope that the revised version of the manuscript is now acceptable for publication in your journal. If you have any queries, please don’t hesitate to contact us. We will try our best to revise the manuscript till you are satisfied. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background Pancreatic cancer (PC) has much weaker prognosis, which can be divided into diabetes and non-diabetes. PC patients with diabetes mellitus will have more opportunities for physical examination due to diabetes, while pancreatic cancer patients without diabetes tend to have higher risk. Identification of prognostic markers for diabetic and non-diabetic pancreatic cancer can improve the prognosis of patients with both types of pancreatic cancer. Methods Both types of PC patients perform differently at the clinical and molecular levels. The Cancer Genome Atlas (TCGA) is employed in this study. The gene expression of the PC with diabetes and non-diabetes is used for predicting their prognosis by LASSO (Least Absolute Shrinkage and Selection Operator) Cox regression.</ns0:p><ns0:p>Furthermore, the results are validated by exchanging gene biomarker with each other and verified by independent Gene Expression Omnibus (GEO) and international cancer genome consortium (ICGC). The prognostic index (PI) is generated by a combination of genetic biomarkers that are used to rank the patient's risk ratio. Survival analysis is applied to test significant difference between high-risk group and low-risk group. Results An integrated gene prognostic biomarker consisted by 14 low-risk genes and 6 high-risk genes in PC with non-diabetes. Meanwhile, and another integrated gene prognostic biomarker consisted by 5 low-risk genes and 3 high-risk genes in PC with diabetes. Therefore, the prognostic value of gene biomarker in PC with non-diabetes and diabetes are all greater than clinical traits (HR=1.102, P-value &lt;0.0001; HR=1.212, P-value &lt;0.0001). Gene signature in PC with nondiabetes was validated in two independent datasets Conclusions The conclusion of this study indicated that the prognostic value of genetic biomarkers in PCs with non-diabetes and diabetes. The gene signature was validated in two independent database. Therefore, this study is expected to provide a novel gene biomarker for predicting prognosis of PC with non-diabetes and diabetes and improving clinical decision.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PC is an aggressive cancer of the digestive system, which is becoming a serious health problem worldwide. Overall survival for patients with pancreatic cancer is poor, mainly due to a lack of biomarkers to enable early diagnosis and a lack of prognostic markers that can inform decisionmaking, facilitating personalized treatment and an optimal clinical outcome <ns0:ref type='bibr' target='#b0'>(1)</ns0:ref>. In most cases, type-II diabetes frequently occurs in patients with PC .Thus, it is considered to be an important risk factor for malignancy of PC <ns0:ref type='bibr' target='#b1'>(2)</ns0:ref>. However, non-diabetes PC patients have no early diagnosis indicator, which makes it more difficult to diagnose. In addition, PC with diabetes and without diabetes are very different in histopathology <ns0:ref type='bibr' target='#b2'>(3)</ns0:ref> and molecular levels. Currently, many studies do not consider the difference between PC with diabetes and non-diabetes. They just considered that diabetes was a risk factor in PC development <ns0:ref type='bibr' target='#b3'>(4)</ns0:ref>. With the deeper understanding of the relationship between PC patient with diabetes and non-diabetes, recent data suggests that diabetes and altered in glucose metabolism are the consequence of PC, and yet, the clinical presentation of the altered glucose metabolism in these patients vary considerably <ns0:ref type='bibr' target='#b4'>(5)</ns0:ref>. So, PC patients with diabetes and nondiabetes may represent two types of PC. Therefore, we predict that PC patients with diabetes and non-diabetes are also different in their prognostic biomarkers. The different prognostic biomarkers indicate that they should be treated respectively via their own different ways.</ns0:p><ns0:p>Generally, patients with diabetes have more opportunities to detect the potential risk of pancreatic cancer, while patients without diabetes often lack indicators for early diagnosis and miss the best opportunity for pancreatic cancer treatment. Furthermore, good prognostic markers can also be targeted at two types of pancreatic cancer patients to propose better treatment options, improve the prognosis.</ns0:p><ns0:p>In this study, The Cancer Genomic Atlas (TCGA) database, Gene Expression Omnibus (GEO) database and international cancer genome consortium (ICGC)were employed to investigate and validate gene biomarker for prognosis in PC with or without diabetes. By characterizing genetic alterations, TCGA project has provided a large number of comprehensive genomic cancer data and corresponding clinical data that we can be used to figure out the relationship between them, which allows us to understand PC better and more accurate. However, high through-put genomic data (microarray or High seq V2) may encounter the problem in statistics which called 'curse of dimensionality'' <ns0:ref type='bibr' target='#b5'>(6)</ns0:ref>. Due to this problem, ordinary regression is subject to over-fitting and instable coefficients and stepwise variable selection methods do not scale well <ns0:ref type='bibr' target='#b6'>(7)</ns0:ref>. Therefore, the least absolute shrinkage and selection operator (LASSO) method is employed to resolve this problem <ns0:ref type='bibr' target='#b7'>(8,</ns0:ref><ns0:ref type='bibr' target='#b8'>9)</ns0:ref>. Through adjusting the coefficient of Cox regression, LASSO can penalize the regression in high dimensionality and collinearity to solve 'curse of dimensionality'' <ns0:ref type='bibr' target='#b9'>(10,</ns0:ref><ns0:ref type='bibr' target='#b10'>11)</ns0:ref>. Least Absolute Shrinkage and Selection Operator (LASSO) regression and a hybrid of these (elastic net regression); all three methods are based on penalizing the L1 norm, the L2 norm, and both the L1 norm and L2 norm with tuning parameters. Although the traditional Cox proportional hazards model is widely used to discover cancer prognostic factors, it is not appropriate for the genomic setting due to the high dimensionality and collinearity. Several groups have proposed to combine the Cox regression model with the elastic net dimension reduction method to select survivalcorrelated genes within a high-dimensional expression dataset and have made available the associated computation procedures. Many studies have adopted elastic-net regression to screen genes, in order to predict survival of patients. In the current study, we are going to subject the integrated mRNA and clinical factors profiles of PC patients&#65292;aiming to identify and analyze gene biomarker that can predict the overall survival (OS) in the diabetes and non-diabetes of PC patients by LASSO.</ns0:p><ns0:p>Recently, many studies employed TCGA (TCGA-PAAD) and GEO dataset (GSE62452) to identify useful gene biomarker which can predict prognosis in many various cancer patients <ns0:ref type='bibr' target='#b11'>(12,</ns0:ref><ns0:ref type='bibr' target='#b12'>13)</ns0:ref>. In this study, ICGC dataset was also employed to validate prognostic gene signature.</ns0:p><ns0:p>Along with the increasing genomic data of PC patients, lots of corresponding studies begin to analyze the genomic data and try their best to explore interesting and meaningful but extremely Manuscript to be reviewed difficult problems <ns0:ref type='bibr' target='#b13'>(14,</ns0:ref><ns0:ref type='bibr' target='#b14'>15)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Information of Patients</ns0:head><ns0:p>All related studies about diabetic and non-diabetic patients with PC were identified and collected by carefully searching from the online TCGA (TCGA: GDC TCGA Pancreatic Cancer) databases (http://tcga-data.nci.nih.gov/tcga/). The following combination of keywords was simultaneously applied for the literature search according to the requirement of this study 'pancreatic cancer' or 'PC' or 'pancreatic tumor' or 'pancreatic malignancy' and 'diabetes' and 'non-diabetes'. In addition, the following research feature criteria are used to further improve and screen the desired search samples: <ns0:ref type='bibr' target='#b0'>(1)</ns0:ref> researches that concentrated on patients with diabetes and non-diabetes were selected; (2) survival time involved of patients was more than 30 days; <ns0:ref type='bibr' target='#b2'>(3)</ns0:ref> patients who didn't receive any adjuvant therapy before. (4) all tissues that were from patients must be the primary tumor. After filtering and screening the data by these above criteria, 136 samples were selected from TCGA databases, which included 99 non-diabetic patients and 37 diabetic patients with PC.</ns0:p></ns0:div> <ns0:div><ns0:head>RNA data Gathering and Filtering</ns0:head><ns0:p>The data of mRNA expression was downloaded from TCGA database. And the IIIumina HiSeq RNASeqV2 platform is selected.</ns0:p></ns0:div> <ns0:div><ns0:head>Clinical factors and survival analysis</ns0:head><ns0:p>Clinical factors for the both diabetic and non-diabetic patients with PC are listed exhaustively in supplementary table1. For the correlation between RNA expression and OS was carried out by forthputting univariate Cox regression (the two-sided log-rank test). In the present meta-analysis, HRs and corresponding 95% CIs were combined to estimate the value of cancer prognosis. The hazard ratio (HR) was calculated from exp (&#946;) and &#946; was the coefficient from Cox regression. Manuscript to be reviewed regarded as an important indicator of diabetic and non-diabetic patient prognosis.</ns0:p></ns0:div> <ns0:div><ns0:head>The Expression of mRNA associated with Survival Analysis</ns0:head><ns0:p>The relationship between patient survival and mRNA expression was analyzed through drawing on the univariate Cox proportional hazard regression. The null-selected RNA is calculated again and again. P-value&#8804;0.05 screened for mRNA (P &#8804; 0.05). In normal conditions, RNAs that had a HR&gt;1 and P value &#8804;0.05 were considered to be a risky gene while HR&lt;1 is seen as an improved low-risky gene. In diabetic patients with PC, we reached a conclusion that 64 mRNAs are significantly associated with overall survival time (p&lt;0.05) by univariate Cox regression. In nondiabetic patients with PC, we acknowledged that 1,559 mRNAs are obvious significantly associated with overall survival time (p&lt;0.05). In data of high dimension gene expression, the coefficients (&#946;) of Cox regression model needs to be penalized in order that it can fit better and minimize errors as much as possible. Therefore, elastic net-regulated Cox regression method is applied to calculate the results from univariate Cox regression. The penalized log-likelihood function is defined as following:</ns0:p><ns0:formula xml:id='formula_0'>&#119897; &#119901; ( &#120573;,&#119883; ) = &#119897; ( &#120573;,&#119883; ) -&#955; &#119901; &#8721; &#119895; = 1 |&#120573; &#119895; |</ns0:formula><ns0:p>With the value of increasing, value of would be decreased. Then, some coefficients (&#946;)</ns0:p><ns0:formula xml:id='formula_1'>&#955; &#8721; &#119901; &#119895; = 1 |&#120573; &#119895; |</ns0:formula><ns0:p>of RNAs would be changed into 0. This result was analyzed by selecting the LASSO-adjusted Cox regression coefficient &#8800;0 mRNA. These steps are carried out by R package 'glmnet'. Finally, we obtained eight mRNAs in diabetic patient with PC and 20 mRNAs in non-diabetic patients with PC.</ns0:p></ns0:div> <ns0:div><ns0:head>Prognosis index construction</ns0:head><ns0:p>PI is calculated from linear combination of candidate RNAs and their expression for each PC patient. We defined a weighted prognostic index (WPI) <ns0:ref type='bibr' target='#b15'>(16)</ns0:ref> for integrating indicators of RNAs for each PC patient, as following: Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_3'>(2) WPI = &#119875;&#119868; -mean(&#119875;&#119868;) &#119878;&#119863;(&#119875;&#119868;)</ns0:formula><ns0:p>Where &#946; i represents the coefficient in Cox regression of the ith variable. And V i signifies the value of the ith variable. Mean (PI) and SD (PI) stand for the mean value and standard deviation of the PI, respectively. Where V i is the expression value of each mRNA (log2-transformed expression value) and &#946; i is the LASSO regulated Cox proportional hazards regression coefficient of the ith RNA or clinical traits.</ns0:p></ns0:div> <ns0:div><ns0:head>Risk stratification and ROC curves</ns0:head><ns0:p>The capacity of the integrated RNA and clinical model to predict clinical outcome was evaluated by comparing the analysis of area under curve (AUC) of the receiver operation characteristic (ROC) curves. AUC for the ROC curve was applied to the 'survival ROC' package in R software <ns0:ref type='bibr' target='#b16'>(17)</ns0:ref>. The higher AUC is considered as a better model performance and range of AUC value is from 0.5 to 1. The AUC range from 0.80-0.90 is treated as good performance. And the range from 0.90-1.00 was considered to be excellent performance. The risk of patient group was classified into two groups based on the median of WPIs: high-risk and a low-risk. Survival analysis is forthputting Kaplan-Meier curves. Statistical analysis and graph in this study were performed using the software of R software <ns0:ref type='bibr' target='#b17'>(18)</ns0:ref>, version 3.2.4 and Bioconductor, version 2.15 <ns0:ref type='bibr' target='#b18'>(19)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene Ontology and Pathway Enrichment</ns0:head><ns0:p>Gene ontology (GO) functional enrichment analysis was performed to RNAs which classified as low-risk and high-risk group by making use of the online tool of the DAVID (version 6.8). We chose 'Homo sapiens' as the background in order to search terms 'GO_TERM_BP_FAT' for further analysis. And these genes are also enriched in Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway for analysis <ns0:ref type='bibr' target='#b19'>(20)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Validation data of patient information collection</ns0:head><ns0:p>In this study, we selected two independent datasets to validation. An independent mRNA expression data of PC patients with 65 PC patients was downloaded from Gene Expression Omnibus(GEO: GSE62452) database (https://www.ncbi.nlm.nih.gov/geo/ ). The clinical traits and Manuscript to be reviewed expression were all downloaded from GSE62452. And the mRNA expression data were generated by Affymetrix Human Genome U133A Array. Data from GEO was analyzed using the updated July 26, 2018.</ns0:p><ns0:p>Another database was downloaded from ICGC database (https://dcc.icgc.org/). We selected </ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Clinical traits</ns0:head><ns0:p>In the TCGA PC cohort of the 136 patients, 99 patients are non-diabetic PC patients and 37 patients are diabetic PC patients. We calculated the clinical factors by adopting univariate survival analysis and multivariable Cox regression analysis. We selected nine clinical variables including age, gender, tumor status, alcohol history, history of chronic pancreatitis, number of lymph nodes positive, maximum tumor dimension, neoplasm histologic grade and pathologic stage. And these data are summarized in table1. In pancreatic patients without a diabetes cohort, tumor status was significantly associated with overall survival by long-rank and multivariate Cox regression analysis. This result indicated that tumor status is an independent factor correlated with overall survival. In pancreatic patients with diabetes cohort, gender is significantly associated with overall survival time. But this factor is not an independent factor by multivariate Cox regression analysis (Figure <ns0:ref type='figure'>1</ns0:ref>, Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gene signature analysis in PC cohort</ns0:head><ns0:p>By Manuscript to be reviewed which were significantly associated with overall survival. Among these genes, the values of HR&lt; 1 and P value &lt;0.01 were considered as protective RNAs and otherwise the values of HR &gt; 1 were risky RNAs (Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>, 3). And the graph for elastic net Cox regression is listed in supplementary file (supplementary 1 and supplementary 2).</ns0:p><ns0:p>The PI was significantly associated with pancreatic patient survival. After normalized PI to WPI, the median value of WPI is acted as cutoff threshold to classify low-risk and high-risk patient cohort (Figure <ns0:ref type='figure'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Validation of the prognostic gene signature</ns0:head><ns0:p>The results were employed in two different ways to verify its stability and reliability. Firstly, we used the gene biomarker in PC patients with diabetes (8 mRNAs) to test the survival curve in PC patients with non-diabetes. Secondly, we used the gene biomarker in PC patients with non-diabetes (20 mRNAs) to swap above calculation.</ns0:p><ns0:p>The validated results showed that the gene biomarker in two groups performed poor result after exchange (Figure <ns0:ref type='figure'>2</ns0:ref>). The results indicated that the gene biomarker in different groups has specificity in each condition.</ns0:p><ns0:p>For validation result, independent mRNA expression data and corresponding clinical information of PC patient with non-diabetes is downloaded from GEO database to estimate the reproducibility and robustness of the results from TCGA database.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene Ontology Enrichment</ns0:head><ns0:p>The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 was employed to discover the function of genes both in PC patient with diabetes and non-diabetes. The eight genes in PC with diabetes were associated with regulation of transcription with a Benjamini-Hochberg correction P-value&lt;0.05. And many genes had DNA binding function. For 20 genes identified in PC without diabetes were not enriched statistically significant association.</ns0:p></ns0:div> <ns0:div><ns0:head>Comparison of clinical traits and gene biomarker for predicting prognosis</ns0:head><ns0:p>We integrated clinical traits that significantly associated with survival and PI of gene biomarker Manuscript to be reviewed that significantly associated with survival to analyze the pancreatic cancer in diabetic and nondiabetic individuals. After multivariate Cox regression analysis, the results showed that PI of gene biomarker performed greatest P-value (Table <ns0:ref type='table' target='#tab_5'>4</ns0:ref>). We filtered the clinical factors that significantly associated with survival by log-rank test into integrative model. In PC with non-diabetes, tumor status, number of lymph nodes positive, stage G2, G3 and G4 were significantly associated with survival (Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>). And in PC with diabetes, gender, stage G2 and G3 were significantly associated with survival by log-rank test (Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>).</ns0:p><ns0:p>From the table 4, we find PI of gene biomarker have smallest P-value after multivariable Cox regression. Although HR is not the highest among clinical traits, P-value is the smallest. Besides, we can find that tumor status is another significant risk factor in PC with non-diabetes.</ns0:p></ns0:div> <ns0:div><ns0:head>Independent data validation for PC with non-diabetes</ns0:head><ns0:p>For further validation result, independent mRNA expression data and corresponding clinical information of PC patient with non-diabetes is downloaded from GEO database (GSE62452) to estimate the reproducibility and robustness of the results from TCGA database. The results showed that the gene signature from TCGA data could be validated in GEO database (n=65). PI was calculated from gene signature can effectively predict survival of PC with non-diabetes. The median of PI value divided 65 patients into high-risk group and low-risk group (HR=3.006, P-value&lt;0.001). And results of ROC showed that AUC=0.828. The results indicated that the gene signature from TCGA could be validated in independent dataset (Figure <ns0:ref type='figure'>3</ns0:ref>).</ns0:p><ns0:p>Pancreatic cancer data was downloaded from ICGC database. This data included 92 patients with genomic data and clinical information. The gene signature was matched ICGC database and constructed PI model. The results showed that the PI from gene signature can divided patients into high-risk and low-risk groups significantly (HR=2.84, P-value&lt;0.001) in ICGC data. ROC showed that AUC=0.74, which indicated that the gene signature also validated in ICGC and predict performance well in 3 years (Figure <ns0:ref type='figure' target='#fig_9'>4</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>PC patients showed different prognostic gene signature in diabetes and non-diabetes. Identification special gene signature in different types of PC patients would provide precise medicine for different patients. We identified and verified specific high-risk genes for PC patients without diabetes. And these genes have not been reported before. These gene targets may be potential therapeutic targets for pancreatic cancer.</ns0:p><ns0:p>In this study, we proposed two classes of gene biomarkers in PC patients with and without diabetes which can guide us to predict PC patient survival better and more accurate. To a large extent, PC patients with and without diabetes have quite different gene biomarker for predicting prognosis.</ns0:p><ns0:p>After a series of studies, we not only find that genes candidate in both PC patient groups have no overlapping but also figure out that gene biomarker in non-diabetes PC patients is validated by GEO and ICGC datasets. The result indicated that the two sets of gene biomarker in both groups have been very specified. Therefore, they have their own gene biomarker for predicting their prognosis. Because the differences between diabetic and non-diabetic pancreatic cancer patients are often ignored, we only got two types of patients in TCGA database. Other validation databases contained only non-diabetic patients. Furthermore, non-diabetic patients with pancreatic cancer are more likely to be ignored in the diagnosis, leading to a higher risk of such patients. Thus, we validated gene biomarker in non-diabetes PC patients in more datasets. Although a large number of studies have reported some biomarkers in PC patients, many genes have been identified</ns0:p><ns0:p>primarily in PC patients without diabetes. We identified and compared the gene signature that predict both types of PC patients. And many genes have not been reported yet so far. Among the high risk prognostic genes, CRCT1, MUC20, RTP1, C10orf111, SPACA5 and FZD10 have high level of HR. MUC20, FZD10 have been identified in PC patients <ns0:ref type='bibr' target='#b20'>(21,</ns0:ref><ns0:ref type='bibr' target='#b21'>22)</ns0:ref> and these two genes play a vital role in two important pathways associated with cancer. MUC20 is involved in MET (Mesenchymal-Epithelial transitions) process which is a common process in many tumors <ns0:ref type='bibr' target='#b22'>(23)</ns0:ref>. Manuscript to be reviewed</ns0:p><ns0:p>And it may regulate MET signaling cascade. It appears to decrease hepatocyte growth factor (HGF)-induced transient MAPK activation <ns0:ref type='bibr' target='#b23'>(24)</ns0:ref>. FZD10 is associated with WNT signaling pathway which is implicated in embryogenesis as well as in carcinogenesis <ns0:ref type='bibr' target='#b24'>(25)</ns0:ref>. Other genes were not reported in PC patients, but only SPACA5 is reported in bladder cancer <ns0:ref type='bibr' target='#b25'>(26)</ns0:ref>. Although many genes have not been reported before, we find that these combinations of these genes can greatly distinguish high-risk and low-risk PC patients with non-diabetes. In addition, these genes were validated in an independent GEO database and ICGC database. The results of GSE62452 in the GEO database indicated that these genes were stably expressed and the gene biomarker could distinct between high-risk and low-risk gene greatly.</ns0:p><ns0:p>The gene biomarker in PC patients with diabetes, three genes are high-risk genes. We can find that the production of these three genes (ZNF793, GBP6, FOSL1) are binding function proteins. Thus, we infer that they are all transcription factors. Of the three genes, FOSL1 has been reported to be closely associated with PC <ns0:ref type='bibr' target='#b26'>(27)</ns0:ref><ns0:ref type='bibr' target='#b27'>(28)</ns0:ref><ns0:ref type='bibr' target='#b28'>(29)</ns0:ref>. But these studies have not reported that this high-risk gene is associated with PC with diabetes yet. Only one study reported that FOSL1 is closely associated with diabetes mellitus <ns0:ref type='bibr' target='#b29'>(30)</ns0:ref>. And this gene has not been identified in PC with non-diabetes. GBP6</ns0:p><ns0:p>is reported in diabetes(31) but is not reported in PC patients with diabetes. ZNF793 is not identified in both PC and diabetes. Thus, we infer that the gene is a potential risk factor in PC patients with diabetes.</ns0:p><ns0:p>Through multivariate Cox regression analysis, it is interesting to note that tumor status is an independent predictor of prognosis in non-diabetes PC patients. Gender is an independent predictor of prognosis in patients with diabetes in PC. Tumor status is a vital clinical factor for predicting the prognosis in many cancers.</ns0:p><ns0:p>From the results, we find that there was no overlapping of both groups. Thus, we conclude that two types of PC vary greatly at the molecular level. Prognostic gene signature in non-diabetes PC patients showed robustness among two datasets (GEO and ICGC). Many genes have not reported in publication and we hope that these genes can predict prognosis for improving clinical decision. Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Pancreatic cancer patients with diabetes and without diabetes have different gene signature for predicting their respective prognosis. The results indicated that Gene signature of pancreatic cancer patients without diabetes has been validated in two independent datasets. Thus, the different gene marker might be as an useful tool for the clinical decision in future.</ns0:p></ns0:div> <ns0:div><ns0:head>Acknowledgement</ns0:head><ns0:p>This project was supported by the National Natural Science Foundation of China (Grant No.</ns0:p></ns0:div> <ns0:div><ns0:head>81660581).</ns0:head></ns0:div> <ns0:div><ns0:head>Ethical Policies and Standards</ns0:head><ns0:p>Conflict of Interest: The authors declare that they have no conflict of interest.</ns0:p><ns0:p>Ethical approval: This article does not contain any studies with human participants or animals performed by any of the authors.</ns0:p><ns0:p>of the Wnt/beta-catenin pathway and its target genes for the N-and C-terminal domains of parathyroid hormone-related protein in bone from diabetic mice. FEBS Lett. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Clinical variables from univariate Cox proportional hazards regression P-value&#8804;0.05 were PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PI = &#8721;( &#120573;&#119894; * &#119881;&#119894; ) PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Number of figures: 4 Figure 1 .</ns0:head><ns0:label>41</ns0:label><ns0:figDesc>Figure 1. WPI analysis of the integrated gene-and-clinical model for 136 TCGA PC patients. (A) Survival analysis in PC patient with non-diabetes. (B) WPI distribution in the TCGA PC cohort without diabetes. The dash line represents the cutoff used to categorize patients into the low-risk group or the high-risk group. (C) Survival analysis in PC patient with diabetes. (D) WPI distribution in the TCGA PC cohort with diabetes.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 2 .Figure 3 .</ns0:head><ns0:label>23</ns0:label><ns0:figDesc>Figure 2. Exchange gene biomarker to cross-validate in two groups.(A) Using gene biomarker of PC with diabetes to test in PC with non-diabetes. (B) Using gene biomarker of PC with nondiabetes to test in PC with diabetes Figure 3. Kaplan-Meier curves and ROC curves for validation PC patients in GEO database. (A)The gene biomarker can greatly classify PC patients into high-risk and low-risk groups (p&lt;0.001). (B)The AUC of ROC is 0.828, which represent that the gene biomarker model is very good.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Gene signature validation in Pancreatic cancer from ICGC database. (A) High-risk and low-risk groups showed significantly difference (HR=2.84, P-value&lt;0.001) in ICGC PC data. (B) ROC curve showed gene signature performance well in 3 years in ICGC PC data..</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,178.87,525.00,342.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,178.87,525.00,356.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='28,42.52,178.87,525.00,347.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,178.87,525.00,347.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,178.87,525.00,392.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='31,42.52,178.87,525.00,387.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='32,42.52,199.12,525.00,384.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='33,42.52,178.87,525.00,373.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='34,42.52,178.87,525.00,223.50' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>Pancreatic Cancer -AU data for further validation. This dataset included 92 PC patients with RNAseq and clinical information. The genomic data of this dataset uses the technology of next generation sequencing. This gene data contained 56,026 RNAs and 92 patients' follow-up data.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>We extracted gene signature from 56,026 RNAs for verification prognosis. (All raw data and code</ns0:cell></ns0:row><ns0:row><ns0:cell>was listed in supplementary file 1)</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>2010;584(14):3095. 31. O'Tierney PF, Lewis RM, Mcweeney SK, et al. Immune Response Gene Profiles in the Term Placenta Depend Upon Maternal Muscle Mass. Reprod. Sci. 2012;19(10):1041. Clinical traits in PC patients with non-diabetes and diabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='3'>Manuscript to be reviewed</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Non-diabetes PC(n=99)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Diabetes PC(n=37)</ns0:cell></ns0:row><ns0:row><ns0:cell>Factors</ns0:cell><ns0:cell /><ns0:cell>Death/patients</ns0:cell><ns0:cell>Log-rank</ns0:cell><ns0:cell>Multivariat</ns0:cell><ns0:cell>Death/patients</ns0:cell><ns0:cell>Log-rank</ns0:cell><ns0:cell>Multivariate</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>e Cox P</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Cox P</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.051</ns0:cell><ns0:cell>0.496</ns0:cell><ns0:cell /><ns0:cell>0.959</ns0:cell><ns0:cell>0.446</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;=64</ns0:cell><ns0:cell /><ns0:cell>22/52</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>7/16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;64</ns0:cell><ns0:cell /><ns0:cell>31/47</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>8/21</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gender</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.402</ns0:cell><ns0:cell>0.172</ns0:cell><ns0:cell /><ns0:cell>0.001*</ns0:cell><ns0:cell>0.340</ns0:cell></ns0:row><ns0:row><ns0:cell>Female</ns0:cell><ns0:cell /><ns0:cell>27/50</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>7/12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Male</ns0:cell><ns0:cell /><ns0:cell>26/49</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>8/25</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Tumor Status</ns0:cell><ns0:cell /><ns0:cell>9.3e-06*</ns0:cell><ns0:cell>0.0004*</ns0:cell><ns0:cell /><ns0:cell>0.005*</ns0:cell><ns0:cell>0.513</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>With Tumor</ns0:cell><ns0:cell>42/57</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>10/17</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Tumor Free</ns0:cell><ns0:cell>6/35</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>2/15</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Unknown</ns0:cell><ns0:cell /><ns0:cell>7/7</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3/5</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Alcohol history</ns0:cell><ns0:cell /><ns0:cell>0.537</ns0:cell><ns0:cell>0.144</ns0:cell><ns0:cell /><ns0:cell>0.599</ns0:cell><ns0:cell>0.638</ns0:cell></ns0:row><ns0:row><ns0:cell>Yes</ns0:cell><ns0:cell /><ns0:cell>40/68</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>10/27</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>No</ns0:cell><ns0:cell /><ns0:cell>12/39</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>5/10</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Unknown</ns0:cell><ns0:cell /><ns0:cell>1/2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>-</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>History</ns0:cell><ns0:cell>of</ns0:cell><ns0:cell /><ns0:cell>0.597</ns0:cell><ns0:cell>0.998</ns0:cell><ns0:cell /><ns0:cell>0.273</ns0:cell><ns0:cell>0.998</ns0:cell></ns0:row><ns0:row><ns0:cell>chronic</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>pancreatitis</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Yes</ns0:cell><ns0:cell /><ns0:cell>4/8</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3/4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>No</ns0:cell><ns0:cell /><ns0:cell>48/86</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>10/31</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Unknown</ns0:cell><ns0:cell /><ns0:cell>1/5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>2/2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Number</ns0:cell><ns0:cell>of</ns0:cell><ns0:cell /><ns0:cell>0.003*</ns0:cell><ns0:cell>0.396</ns0:cell><ns0:cell /><ns0:cell>0.480</ns0:cell><ns0:cell>0.533</ns0:cell></ns0:row><ns0:row><ns0:cell>lymph</ns0:cell><ns0:cell>nodes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>positive by he</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>&lt;3</ns0:cell><ns0:cell /><ns0:cell>22/52</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>7/20</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;=3</ns0:cell><ns0:cell /><ns0:cell>30/45</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>8/16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Maximum</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.394</ns0:cell><ns0:cell>0.216</ns0:cell><ns0:cell /><ns0:cell>0.147</ns0:cell><ns0:cell>0.279</ns0:cell></ns0:row><ns0:row><ns0:cell>tumor</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>dimension</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;3.5</ns0:cell><ns0:cell /><ns0:cell>27/44</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>9/16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&lt;=3.5</ns0:cell><ns0:cell /><ns0:cell>26/51</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>6/20</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Neoplasm</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.039*</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.004*</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>histologic grade</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>G1</ns0:cell><ns0:cell /><ns0:cell>4/16</ns0:cell><ns0:cell /><ns0:cell>-</ns0:cell><ns0:cell>2/7</ns0:cell><ns0:cell /><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell /><ns0:cell>31/52</ns0:cell><ns0:cell /><ns0:cell>0.606</ns0:cell><ns0:cell>6/20</ns0:cell><ns0:cell /><ns0:cell>0.998</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell /><ns0:cell>17/29</ns0:cell><ns0:cell /><ns0:cell>0.202</ns0:cell><ns0:cell>7/10</ns0:cell><ns0:cell /><ns0:cell>0.308</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Gene biomarker in PC patients with non-diabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Hazard</ns0:cell><ns0:cell>CI</ns0:cell><ns0:cell>P value</ns0:cell><ns0:cell>Description</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Low Risk genes</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>TTTY9B</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell cols='2'>0.000-0.028 0.0102</ns0:cell><ns0:cell>testis-specific transcript, Y-linked 9B</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>(non-protein coding)</ns0:cell></ns0:row><ns0:row><ns0:cell>RNF121</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell cols='2'>0.000-0.260 0.0142</ns0:cell><ns0:cell>RING finger protein 121</ns0:cell></ns0:row><ns0:row><ns0:cell>FHAD1</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell cols='3'>0.001-0.051 3.60E-06 Forkhead-associated domain-containing</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>protein 1</ns0:cell></ns0:row><ns0:row><ns0:cell>GTF2F2</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell cols='2'>0.000-0.516 0.0235</ns0:cell><ns0:cell>General transcription factor IIF subunit 2</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>ADAMTS19 0.009</ns0:cell><ns0:cell cols='2'>0.001-0.113 0.0002</ns0:cell><ns0:cell>A disintegrin and metalloproteinase with</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>thrombospondin motifs 19</ns0:cell></ns0:row><ns0:row><ns0:cell>LHFPL1</ns0:cell><ns0:cell>0.024</ns0:cell><ns0:cell cols='2'>0.002-0.283 0.0031</ns0:cell><ns0:cell>Lipoma HMGIC fusion partner-like 1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>protein</ns0:cell></ns0:row><ns0:row><ns0:cell>DHDH</ns0:cell><ns0:cell>0.05</ns0:cell><ns0:cell cols='3'>0.013-0.191 1.16E-05 Trans-1,2-dihydrobenzene-1,2-diol</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>dehydrogenase</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>LOC256880 0.062</ns0:cell><ns0:cell cols='2'>0.006-0.600 0.0164</ns0:cell></ns0:row><ns0:row><ns0:cell>SLC25A41</ns0:cell><ns0:cell>0.093</ns0:cell><ns0:cell cols='2'>0.022-0.392 0.001</ns0:cell><ns0:cell>Solute carrier family 25 member 41</ns0:cell></ns0:row><ns0:row><ns0:cell>ZNF233</ns0:cell><ns0:cell>0.095</ns0:cell><ns0:cell cols='2'>0.017-0.516 0.0060</ns0:cell><ns0:cell>Zinc finger protein 233</ns0:cell></ns0:row><ns0:row><ns0:cell>C6orf195</ns0:cell><ns0:cell>0.129</ns0:cell><ns0:cell cols='2'>0.024-0.695 0.0171</ns0:cell></ns0:row><ns0:row><ns0:cell>PCDHA11</ns0:cell><ns0:cell>0.144</ns0:cell><ns0:cell cols='3'>0.050-0.419 0.00037 Proto cadherin alpha-11</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>LOC401127 0.146</ns0:cell><ns0:cell cols='2'>0.022-0.969 0.0463</ns0:cell></ns0:row><ns0:row><ns0:cell>TUBBP5</ns0:cell><ns0:cell>0.303</ns0:cell><ns0:cell cols='2'>0.139-0.663 0.0028</ns0:cell><ns0:cell>tubulin beta pseudo gene 5</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>High risk genes</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>CRCT1</ns0:cell><ns0:cell>2.107</ns0:cell><ns0:cell cols='2'>1.154-3.847 0.0152</ns0:cell><ns0:cell>Cysteine-rich C-terminal protein 1</ns0:cell></ns0:row><ns0:row><ns0:cell>MUC20</ns0:cell><ns0:cell>14.76</ns0:cell><ns0:cell cols='3'>4.387-49.66 1.37E-05 Mucin-20</ns0:cell></ns0:row><ns0:row><ns0:cell>RTP1</ns0:cell><ns0:cell>18.01</ns0:cell><ns0:cell cols='2'>1.075-301.8 0.0444</ns0:cell><ns0:cell>Receptor-transporting protein 1</ns0:cell></ns0:row><ns0:row><ns0:cell>C10orf111</ns0:cell><ns0:cell>23.6</ns0:cell><ns0:cell cols='2'>1.314-423.9 0.0319</ns0:cell></ns0:row><ns0:row><ns0:cell>SPACA5</ns0:cell><ns0:cell>23.83</ns0:cell><ns0:cell cols='2'>1.821-311.7 0.0156</ns0:cell><ns0:cell>Sperm acrosome-associated protein 5</ns0:cell></ns0:row><ns0:row><ns0:cell>FZD10</ns0:cell><ns0:cell>26.54</ns0:cell><ns0:cell cols='3'>5.142-136.9 9.02E-05 Frizzled-10</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>403 *p&lt;0.05, statistically significant</ns0:cell><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Gene biomarker in PC patients with diabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell cols='2'>Hazard CI(95%)</ns0:cell><ns0:cell>p-value</ns0:cell><ns0:cell>Description</ns0:cell></ns0:row><ns0:row><ns0:cell>Low Risk genes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SYS1-</ns0:cell><ns0:cell>0.347</ns0:cell><ns0:cell>0.909-1.815</ns0:cell><ns0:cell>0.0020</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>DBNDD2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>NCRNA00167 0.231</ns0:cell><ns0:cell>0.978-1.719</ns0:cell><ns0:cell>0.0015</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>IRX5</ns0:cell><ns0:cell>0.473</ns0:cell><ns0:cell>0.282-1.185</ns0:cell><ns0:cell>0.0012</ns0:cell><ns0:cell>Iroquois-class homeodomain protein</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>IRX-5</ns0:cell></ns0:row><ns0:row><ns0:cell>ZNF77</ns0:cell><ns0:cell>0.244</ns0:cell><ns0:cell>0.770-1.801</ns0:cell><ns0:cell>0.0040</ns0:cell><ns0:cell>Zinc finger protein 77</ns0:cell></ns0:row><ns0:row><ns0:cell>CATSPERG</ns0:cell><ns0:cell>0.296</ns0:cell><ns0:cell>0.651-0.991</ns0:cell><ns0:cell>0.0029</ns0:cell><ns0:cell>Cation channel sperm-associated</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>protein subunit gamma</ns0:cell></ns0:row><ns0:row><ns0:cell>High Risk genes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ZNF793</ns0:cell><ns0:cell>2.968</ns0:cell><ns0:cell>0.358-1.978</ns0:cell><ns0:cell>0.0063</ns0:cell><ns0:cell>Zinc finger protein 793</ns0:cell></ns0:row><ns0:row><ns0:cell>GBP6</ns0:cell><ns0:cell>1.744</ns0:cell><ns0:cell>0.342-1.207</ns0:cell><ns0:cell>0.0011</ns0:cell><ns0:cell>Guanylate-binding protein 6</ns0:cell></ns0:row><ns0:row><ns0:cell>FOSL1</ns0:cell><ns0:cell>2.306</ns0:cell><ns0:cell cols='2'>0.9601-1.051 0.0091</ns0:cell><ns0:cell>Fos-related antigen 1</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>405 *p&lt;0.05, statistically significant</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Multivariate Cox regression analysis of prognosis index and clinical traits PC with Nondiabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>HR</ns0:cell><ns0:cell>CI</ns0:cell><ns0:cell>Multivariate Cox</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>P-value</ns0:cell></ns0:row><ns0:row><ns0:cell>PI</ns0:cell><ns0:cell>1.102</ns0:cell><ns0:cell>1.070-1.136</ns0:cell><ns0:cell>2.68e-10*</ns0:cell></ns0:row><ns0:row><ns0:cell>Tumor Status</ns0:cell><ns0:cell>0.117</ns0:cell><ns0:cell>0.298-1.924</ns0:cell><ns0:cell>0.0005*</ns0:cell></ns0:row><ns0:row><ns0:cell>Number of lymph</ns0:cell><ns0:cell>1.589</ns0:cell><ns0:cell>0.907-2.783</ns0:cell><ns0:cell>0.106</ns0:cell></ns0:row><ns0:row><ns0:cell>nodes positive by</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>he</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell>2.103</ns0:cell><ns0:cell>0.187-5.400</ns0:cell><ns0:cell>0.123</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell>2.036</ns0:cell><ns0:cell>0.739-5.613</ns0:cell><ns0:cell>0.169</ns0:cell></ns0:row><ns0:row><ns0:cell>G4</ns0:cell><ns0:cell>2.215</ns0:cell><ns0:cell>0.257-19.087</ns0:cell><ns0:cell>0.469</ns0:cell></ns0:row><ns0:row><ns0:cell>PC with Diabetes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>PI</ns0:cell><ns0:cell>1.212</ns0:cell><ns0:cell>1.108-1.327</ns0:cell><ns0:cell>2.83e-05*</ns0:cell></ns0:row><ns0:row><ns0:cell>Gender</ns0:cell><ns0:cell>0.173</ns0:cell><ns0:cell>0.053-0.564</ns0:cell><ns0:cell>0.004*</ns0:cell></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell>0.897</ns0:cell><ns0:cell>0.168-4.775</ns0:cell><ns0:cell>0.898</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell>5.310</ns0:cell><ns0:cell>0.892-31.616</ns0:cell><ns0:cell>0.067</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>407 *p&lt;0.05, statistically significant</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>1 Table 1</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Clinical traits in PC patients with non-diabetes and diabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='4'>Non-diabetes Pancreatic Cancer(n=99)</ns0:cell><ns0:cell /><ns0:cell cols='3'>Diabetes Pancreatic Cancer(n=37)</ns0:cell></ns0:row><ns0:row><ns0:cell>Factors</ns0:cell><ns0:cell /><ns0:cell>Death/patients</ns0:cell><ns0:cell>Log-rank</ns0:cell><ns0:cell>Multivariat</ns0:cell><ns0:cell>Death/patients</ns0:cell><ns0:cell>Log-rank</ns0:cell><ns0:cell>Multivariate</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>e Cox P</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Cox P</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.051</ns0:cell><ns0:cell>0.496</ns0:cell><ns0:cell /><ns0:cell>0.959</ns0:cell><ns0:cell>0.446</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;=64</ns0:cell><ns0:cell /><ns0:cell>22/52</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>7/16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;64</ns0:cell><ns0:cell /><ns0:cell>31/47</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>8/21</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gender</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.402</ns0:cell><ns0:cell>0.172</ns0:cell><ns0:cell /><ns0:cell>0.001*</ns0:cell><ns0:cell>0.340</ns0:cell></ns0:row><ns0:row><ns0:cell>Female</ns0:cell><ns0:cell /><ns0:cell>27/50</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>7/12</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Male</ns0:cell><ns0:cell /><ns0:cell>26/49</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>8/25</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Tumor Status</ns0:cell><ns0:cell /><ns0:cell>9.3e-06*</ns0:cell><ns0:cell>0.0004*</ns0:cell><ns0:cell /><ns0:cell>0.005*</ns0:cell><ns0:cell>0.513</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>With Tumor</ns0:cell><ns0:cell>42/57</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>10/17</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Tumor Free</ns0:cell><ns0:cell>6/35</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>2/15</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Unknown</ns0:cell><ns0:cell /><ns0:cell>7/7</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3/5</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Alcohol history</ns0:cell><ns0:cell /><ns0:cell>0.537</ns0:cell><ns0:cell>0.144</ns0:cell><ns0:cell /><ns0:cell>0.599</ns0:cell><ns0:cell>0.638</ns0:cell></ns0:row><ns0:row><ns0:cell>Yes</ns0:cell><ns0:cell /><ns0:cell>40/68</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>10/27</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>No</ns0:cell><ns0:cell /><ns0:cell>12/39</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>5/10</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Unknown</ns0:cell><ns0:cell /><ns0:cell>1/2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>-</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>History</ns0:cell><ns0:cell>of</ns0:cell><ns0:cell /><ns0:cell>0.597</ns0:cell><ns0:cell>0.998</ns0:cell><ns0:cell /><ns0:cell>0.273</ns0:cell><ns0:cell>0.998</ns0:cell></ns0:row><ns0:row><ns0:cell>chronic</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>pancreatitis</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Yes</ns0:cell><ns0:cell /><ns0:cell>4/8</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3/4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>No</ns0:cell><ns0:cell /><ns0:cell>48/86</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>10/31</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Unknown</ns0:cell><ns0:cell /><ns0:cell>1/5</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>2/2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Number</ns0:cell><ns0:cell>of</ns0:cell><ns0:cell /><ns0:cell>0.003*</ns0:cell><ns0:cell>0.396</ns0:cell><ns0:cell /><ns0:cell>0.480</ns0:cell><ns0:cell>0.533</ns0:cell></ns0:row><ns0:row><ns0:cell>lymph</ns0:cell><ns0:cell>nodes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>positive by he</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>&lt;3</ns0:cell><ns0:cell /><ns0:cell>22/52</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>7/20</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;=3</ns0:cell><ns0:cell /><ns0:cell>30/45</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>8/16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Maximum</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.394</ns0:cell><ns0:cell>0.216</ns0:cell><ns0:cell /><ns0:cell>0.147</ns0:cell><ns0:cell>0.279</ns0:cell></ns0:row><ns0:row><ns0:cell>tumor</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>dimension</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;3.5</ns0:cell><ns0:cell /><ns0:cell>27/44</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>9/16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&lt;=3.5</ns0:cell><ns0:cell /><ns0:cell>26/51</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>6/20</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Neoplasm</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.039*</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.004*</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>histologic grade</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>G1</ns0:cell><ns0:cell /><ns0:cell>4/16</ns0:cell><ns0:cell /><ns0:cell>-</ns0:cell><ns0:cell>2/7</ns0:cell><ns0:cell /><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell /><ns0:cell>31/52</ns0:cell><ns0:cell /><ns0:cell>0.606</ns0:cell><ns0:cell>6/20</ns0:cell><ns0:cell /><ns0:cell>0.998</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell /><ns0:cell>17/29</ns0:cell><ns0:cell /><ns0:cell>0.202</ns0:cell><ns0:cell>7/10</ns0:cell><ns0:cell /><ns0:cell>0.308</ns0:cell></ns0:row><ns0:row><ns0:cell>G4</ns0:cell><ns0:cell /><ns0:cell>1/2</ns0:cell><ns0:cell /><ns0:cell>0.757</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell /><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>TNM stage</ns0:cell><ns0:cell /><ns0:cell>0.100</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.431</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Gene biomarker in PC patients with non-diabetes</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Gene signature in PC patients with non-diabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Hazard</ns0:cell><ns0:cell>95%CI</ns0:cell><ns0:cell>P-value</ns0:cell><ns0:cell>Description</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Low Risk genes</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>TTTY9B</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell cols='3'>0.000-0.028 0.0102* testis-specific transcript, Y-linked 9B</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>(non-protein coding)</ns0:cell></ns0:row><ns0:row><ns0:cell>RNF121</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell cols='3'>0.000-0.260 0.0142* RING finger protein 121</ns0:cell></ns0:row><ns0:row><ns0:cell>FHAD1</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell cols='3'>0.001-0.051 &lt;0.001* Forkhead-associated domain-containing</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>protein 1</ns0:cell></ns0:row><ns0:row><ns0:cell>GTF2F2</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell cols='3'>0.000-0.516 0.0235* General transcription factor IIF subunit</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>ADAMTS19 0.009</ns0:cell><ns0:cell cols='3'>0.001-0.113 0.0002* A disintegrin and metalloproteinase</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>with thrombospondin motifs 19</ns0:cell></ns0:row><ns0:row><ns0:cell>LHFPL1</ns0:cell><ns0:cell>0.024</ns0:cell><ns0:cell cols='3'>0.002-0.283 0.0031* Lipoma HMGIC fusion partner-like 1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>protein</ns0:cell></ns0:row><ns0:row><ns0:cell>DHDH</ns0:cell><ns0:cell>0.05</ns0:cell><ns0:cell cols='3'>0.013-0.191 &lt;0.001* Trans-1,2-dihydrobenzene-1,2-diol</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>dehydrogenase</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>LOC256880 0.062</ns0:cell><ns0:cell cols='2'>0.006-0.600 0.0164*</ns0:cell></ns0:row><ns0:row><ns0:cell>SLC25A41</ns0:cell><ns0:cell>0.093</ns0:cell><ns0:cell cols='2'>0.022-0.392 0.001*</ns0:cell><ns0:cell>Solute carrier family 25 member 41</ns0:cell></ns0:row><ns0:row><ns0:cell>ZNF233</ns0:cell><ns0:cell>0.095</ns0:cell><ns0:cell cols='3'>0.017-0.516 0.0060* Zinc finger protein 233</ns0:cell></ns0:row><ns0:row><ns0:cell>C6orf195</ns0:cell><ns0:cell>0.129</ns0:cell><ns0:cell cols='2'>0.024-0.695 0.0171*</ns0:cell></ns0:row><ns0:row><ns0:cell>PCDHA11</ns0:cell><ns0:cell>0.144</ns0:cell><ns0:cell cols='3'>0.050-0.419 &lt;0.001* Proto cadherin alpha-11</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>LOC401127 0.146</ns0:cell><ns0:cell cols='2'>0.022-0.969 0.0463*</ns0:cell></ns0:row><ns0:row><ns0:cell>TUBBP5</ns0:cell><ns0:cell>0.303</ns0:cell><ns0:cell cols='3'>0.139-0.663 0.0028* tubulin beta pseudo gene 5</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>High risk genes</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>CRCT1</ns0:cell><ns0:cell>2.107</ns0:cell><ns0:cell cols='3'>1.154-3.847 0.0152* Cysteine-rich C-terminal protein 1</ns0:cell></ns0:row><ns0:row><ns0:cell>MUC20</ns0:cell><ns0:cell>14.76</ns0:cell><ns0:cell cols='3'>4.387-49.66 &lt;0.001* Mucin-20</ns0:cell></ns0:row><ns0:row><ns0:cell>RTP1</ns0:cell><ns0:cell>18.01</ns0:cell><ns0:cell cols='3'>1.075-301.8 0.0444* Receptor-transporting protein 1</ns0:cell></ns0:row><ns0:row><ns0:cell>C10orf111</ns0:cell><ns0:cell>23.6</ns0:cell><ns0:cell cols='2'>1.314-423.9 0.0319*</ns0:cell></ns0:row><ns0:row><ns0:cell>SPACA5</ns0:cell><ns0:cell>23.83</ns0:cell><ns0:cell cols='3'>1.821-311.7 0.0156* Sperm acrosome-associated protein 5</ns0:cell></ns0:row><ns0:row><ns0:cell>FZD10</ns0:cell><ns0:cell>26.54</ns0:cell><ns0:cell cols='3'>5.142-136.9 &lt;0.001* Frizzled-10</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>*p&lt;0.05, statistically significant</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_10'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Gene biomarker in PC patients with diabetes</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_11'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc><ns0:ref type='bibr' target='#b2'>3</ns0:ref> Gene signature in PC patients with diabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell cols='2'>Hazard 95%CI</ns0:cell><ns0:cell>P-value</ns0:cell><ns0:cell>Description</ns0:cell></ns0:row><ns0:row><ns0:cell>Low Risk genes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>SYS1-DBNDD2 0.347</ns0:cell><ns0:cell>0.909-1.815</ns0:cell><ns0:cell>0.0020*</ns0:cell></ns0:row><ns0:row><ns0:cell>NCRNA00167</ns0:cell><ns0:cell>0.231</ns0:cell><ns0:cell>0.978-1.719</ns0:cell><ns0:cell>0.0015*</ns0:cell></ns0:row><ns0:row><ns0:cell>IRX5</ns0:cell><ns0:cell>0.473</ns0:cell><ns0:cell>0.282-1.185</ns0:cell><ns0:cell cols='2'>0.0012* Iroquois-class homeodomain protein</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>IRX-5</ns0:cell></ns0:row><ns0:row><ns0:cell>ZNF77</ns0:cell><ns0:cell>0.244</ns0:cell><ns0:cell>0.770-1.801</ns0:cell><ns0:cell cols='2'>0.0040* Zinc finger protein 77</ns0:cell></ns0:row><ns0:row><ns0:cell>CATSPERG</ns0:cell><ns0:cell>0.296</ns0:cell><ns0:cell>0.651-0.991</ns0:cell><ns0:cell cols='2'>0.0029* Cation channel sperm-associated</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>protein subunit gamma</ns0:cell></ns0:row><ns0:row><ns0:cell>High Risk genes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>ZNF793</ns0:cell><ns0:cell>2.968</ns0:cell><ns0:cell>0.358-1.978</ns0:cell><ns0:cell cols='2'>0.0063* Zinc finger protein 793</ns0:cell></ns0:row><ns0:row><ns0:cell>GBP6</ns0:cell><ns0:cell>1.744</ns0:cell><ns0:cell>0.342-1.207</ns0:cell><ns0:cell cols='2'>0.0011* Guanylate-binding protein 6</ns0:cell></ns0:row><ns0:row><ns0:cell>FOSL1</ns0:cell><ns0:cell>2.306</ns0:cell><ns0:cell cols='3'>0.9601-1.051 0.0091* Fos-related antigen 1</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>*p&lt;0.05, statistically significant</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_12'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Multivariate Cox regression analysis of prognosis index and clinical traits</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2019:05:37344:1:2:NEW 22 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_13'><ns0:head>1 Table 4 .</ns0:head><ns0:label>14</ns0:label><ns0:figDesc>Multivariate Cox regression analysis of prognosis index and clinical traits PC with Nondiabetes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>HR</ns0:cell><ns0:cell>CI</ns0:cell><ns0:cell>Multivariate Cox</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>P-value</ns0:cell></ns0:row><ns0:row><ns0:cell>PI</ns0:cell><ns0:cell>1.102</ns0:cell><ns0:cell>1.070-1.136</ns0:cell><ns0:cell>2.68e-10*</ns0:cell></ns0:row><ns0:row><ns0:cell>Tumor Status</ns0:cell><ns0:cell>0.117</ns0:cell><ns0:cell>0.298-1.924</ns0:cell><ns0:cell>0.0005*</ns0:cell></ns0:row><ns0:row><ns0:cell>Number of lymph</ns0:cell><ns0:cell>1.589</ns0:cell><ns0:cell>0.907-2.783</ns0:cell><ns0:cell>0.106</ns0:cell></ns0:row><ns0:row><ns0:cell>nodes positive by</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>he</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell>2.103</ns0:cell><ns0:cell>0.187-5.400</ns0:cell><ns0:cell>0.123</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell>2.036</ns0:cell><ns0:cell>0.739-5.613</ns0:cell><ns0:cell>0.169</ns0:cell></ns0:row><ns0:row><ns0:cell>G4</ns0:cell><ns0:cell>2.215</ns0:cell><ns0:cell>0.257-19.087</ns0:cell><ns0:cell>0.469</ns0:cell></ns0:row><ns0:row><ns0:cell>PC with Diabetes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>PI</ns0:cell><ns0:cell>1.212</ns0:cell><ns0:cell>1.108-1.327</ns0:cell><ns0:cell>2.83e-05*</ns0:cell></ns0:row><ns0:row><ns0:cell>Gender</ns0:cell><ns0:cell>0.173</ns0:cell><ns0:cell>0.053-0.564</ns0:cell><ns0:cell>0.004*</ns0:cell></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell>0.897</ns0:cell><ns0:cell>0.168-4.775</ns0:cell><ns0:cell>0.898</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell>5.310</ns0:cell><ns0:cell>0.892-31.616</ns0:cell><ns0:cell>0.067</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>2 *p&lt;0.05, statistically significant</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> </ns0:body> "

"Dear Editors and Reviewers: Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Gene signature for prognosis in comparison of pancreatic cancer patient with diabetes and non-diabetes” (ID: 37344). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion are marked in red in the paper. The main corrections in the paper and the responds to the reviewer’s comments are as following: The manuscript needs extensive English editing because there are several typos and grammatical errors. The reviewers indicated that the language used in the manuscript should be improved. I agree with this evaluation and I would, therefore, request for the manuscript to be revised accordingly especially interpreting statistical analysis results. In addition to this: 1. In Line 83 and Line 153, the access date of database and accession numbers of datasets must be given. Response: Thanks for your comment. The access date of database and accession numbers of datasets have been added in manuscript. The data of TCGA is analyzed after updating in April 2019. And GEO dataset is analyzed after updating July 26, 2018. 2. In Line 134, provide more details according to the LASSO regulated Cox proportional hazards regression. Response: We would like to provide more details to the LASSO Cox regression in introduction and method section. Least Absolute Shrinkage and Selection Operator (LASSO) regression and a hybrid of these (elastic netregression); all three methods are based on penalizing the L1 norm, the L2 norm, and both the L1 norm and L2 norm with tuning parameters. Although the traditional Cox proportional hazards model is widely used to discover cancer prognostic factors, it is not appropriate for the genomic setting due to the high dimensionality and colinearity. Several groups have proposed to combine the Cox regression model with the elastic net dimension reduction method to select survival-correlated genes within a high-dimensional expression dataset and have made available the associated computation procedures. 3. In Line 177, it is not clear why the authors prefer to use the median value as cut-off? Response: Thanks for your suggestion. There are many methods to distinguish patients into high risk and low risk by prognostic index, including median method, quartile method and slope maximum method. Median is a common method to distinguish high-risk and low-risk patients by prognostic index. In this paper, we also refer to the methods in other papers, which use the median method to divide patients into high-risk and low-risk. Some references were listed as following: [1] Zhitong Bing, Zhiyuan Cheng, Danfeng Shi, et.al. Investigate the mechanisms of Chinese medicine Fuzhengkangai towards EGFR mutation-positive lung adenocarcinomas by network pharmacology. BMC Complementary and Alternative Medicine. 2018,18: 293 [2] Xinkui Liu, Jiarui Wu, Dan Zhang, et.al. Identification of Potential Key Genes Associated With the Pathogenesis and Prognosis of Gastric Cancer Based on Integrated Bioinformatics Analysis. Front Genet. 2018,9: 265 4. In Line 193, Is that “Benjamini correction” refer to “Benjamini-Hochberg correction” or “Bonferroni correction”? Please revise this. Response: We are sorry for our careless. This manuscript refers to Benjamini-Hochberg correction. We have revised this error. 5. The authors did not provide detailed results of ROC analysis and not discussed these results with p-values and AUC values. Provide details of their findings, and add these figures in the manuscript. Response: We are sorry for our careless. We have added details of ROC analysis in Results section. The details were listed as follows. The results showed that the gene signature from TCGA data could be validated in GEO database (n=65). PI was calculated from gene signature can effectively predict survival of PC with non-diabetes. The median of PI value divided 65 patients into high-risk group and low-risk group (HR=3.006, P-value<0.001). And results of ROC showed that AUC=0.828. The results indicated that the gene signature from TCGA could be validated in independent dataset. For all tables with p-values, please use * for p<0.05 and write the phrase “*p<0.05, statistically significant” in the footnote of the table. Tables are not readable. Improve readability. Response: Thanks for your suggestion. We have revised this problem. [# PeerJ Staff Note: The Editor has identified that the English language needs to be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title). #] Reviewer 1 (Anonymous) Basic reporting I would recommend that the authors seek professional language services to improve their writing. The terms ‘protective’ and ‘risky’ genes may be replaced with more suitable terms. Please refer to other articles describing prognostic biomarkers. The use of ‘generally speaking’ should be avoided. Response: Thanks for your suggestion. We would like to revised these problems. Experimental design Experimental design appears to be appropriate. Validity of the findings The authors identified a set of prognostic biomarkers for pancreatic cancer patients from TCGA. They have validated these markers using a dataset from GEO. It would be useful to validate the markers in another dataset (with more patient numbers) from ICGC. Response: Thanks for your suggestion. We would like to use ICGC database for further validation the results from TCGA. Pancreatic cancer data was downloaded from ICGC database. This data included 92 patients with genomic data and clinical information. The gene signature was matched ICGC database and constructed PI model. The results showed that the PI from gene signature can divided patients into high-risk and low-risk groups significantly (HR=2.84, P-value<0.001) in ICGC data. ROC showed that AUC=0.74, which indicated that the gene signature also validated in ICGC and predict performance well in 3 years. Reviewer 2 (Anonymous) Basic reporting The English language should be improved to ensure that an International audience can clearly understand your text. Response: Thanks for your suggestion. We try our best to improve our manuscript. The authors do not discuss the clinical significance of their findings and the utility and nature of markers they found. The markers are correlative and not necessarily causative. Response: Thanks for your suggestion. This is true of the problems mentioned by reviewers. These genetic markers are clinically relevant, but not necessarily causal. Therefore, we will further discuss this point fully in our discussion. The conclusion section needs to be improved and the authors need to elaborate what are the main findings of their study. Response: Thanks for your suggestion. We have been rewritten the conclusion section. Specific comments are added in the annotated PDF file. Experimental design no comment Validity of the findings Impact and novelty not discussed Response: Thanks for your suggestions. We have added impact and novelty to discussion section. Conclusions are not well stated and need to be rewritten Response: Thanks for your suggestion. We have rewritten the conclusion section. Annotated manuscript The reviewer has also provided an annotated manuscript as part of their review Reviewer 3 (Anonymous) Basic reporting The manuscript is unfortunately poorly written. The use of Scientific English and the grammar is poor. Throughout the manuscript the given information, findings and discussion of the data is not intelligible. Response: We are sorry for our careless. We tried our best to correct the mistakes in English and grammar. And the result part is modified to further reflect the rationality of the manuscript. Writers do not link their hypothesis and their findings in the text and often make confusing jumps to other findings. The coherence is severely lacking. Response: We are sorry for our careless. Thanks for your suggestion. We have revised our manuscript and enhanced our hypothesis. We continue to adjust the word order in the manuscript to make it read more smoothly. Experimental design Clinical findings on tumor state and histological grading is not novel, what authors here try to show is the difference in diabetes groups however there is no highlight to diabetes and in the manuscript it feels like they are repetitively mentioning well established findings. Response: Thanks for your suggestion. We try our best to revised this problem. In this manuscript, we revisited the findings of the paper. The results and their discussion should be focused on the novel findings. Validity of the findings. Response: Thanks for your comment. We focused on our novel findings and validity the finding. There are mislabeled figures and table numbers in the main text body, which is misleading and confusing (e.g in lines 203 and in 204 the label says Table 2 however it must be Table 1) Some findings that are essential to the hypothesis are completely untouched, not mentioned. Authors should give more conscience explanation of each result. Response: Thanks for your comment. (1) We are sorry for our careless. The mislabeled figures and tables have been revised in manuscript. (2) We have revised manuscript about hypothesis. (3) We are sorry for our careless. We added more explanation for each result. Comments for the Author I advise for thorough editing of the use of English. From the introduction to the discussion it is very hard to understand and follow the logic of the findings also more detailed explanation of the rationale of the study and the findings must be given. Response: Thanks for your comment. We try our best to revise manuscript. Comments for the Author The authors reported that the integrated gene prognostic biomarker systems are identified in PC with non-diabetes or diabetes 1. Are there relation or differences between two prognostic biomarker systems？ Response: Thanks for your suggestion. Two prognostic biomarkers systems showed that two types of pancreatic cancer have different prognostic gene signature. The biomarkers in PC with diabetes mainly enriched in regulation of transcription biological process such zinc proteins. The biomarkers in PC with non-diabetes are not enriched into one biological function. However, these high-risk biomarkers in PC with non-diabetes are related with regulation transcription factor, proto cadherin and transporter protein. 2. The diabetes were newly discovered in PC patients？was there any difference between the newly and old diabetes？ Response: Thanks for your suggestion. In this work, PC patients with diabetes were not newly discovery. However, we discovered PC patients with diabetes and non-diabetes were very different in prognostic gene signature. And the gene signature was identified in independent database. 3. TNM staging is most important for prognosis, why the author did not include TNM staging? Response: Thanks for your suggestion. In fact, we have considered TNM staging in Table 1. In previous version, we include TNM stage as Pathologic stage. In this revised version, we have revised as TNM stage. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. In 2011, the United Nations set a target to reduce premature mortality from noncommunicable diseases (NCDs) by 25% by 2025. While studies have reported the target in some countries, no studies have been done in China. This study aims to project the ability to reach the target in Hunan Province, China, and establish the priority for future interventions.</ns0:p><ns0:p>Methods. We conducted the study during 2019-2020. From the Global Burden of Disease Study 2016, we extracted death data for Hunan during 1990-2016 for four main NCDs, namely cancer, cardiovascular disease (CVD), chronic respiratory diseases, and diabetes. We generated estimates for 2025 by fitting a linear regression to the premature mortality over the most recent trend identified by a joinpoint regression model. We also estimated excess premature mortality attributable to unfavorable changes over time.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Premature death from noncommunicable diseases (NCDs) remains a major global development challenge in the 21st century. Each year, a total of 15 million people around the world die from NCDs between the ages of 30 and 70. <ns0:ref type='bibr' target='#b0'>1</ns0:ref> As the most populous country in the world, China is PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed particularly affected by this challenge. A combination of market globalization, rapid urbanization, modifiable risk factors, and population aging over the past decades has led to an NCD epidemic in China: NCDs, mainly including cardiovascular disease (CVD), cancer, diabetes and chronic respiratory diseases, account for 70% of the disease burden and are responsible for 89% of all deaths in the Chinese population. <ns0:ref type='bibr' target='#b0'>1,</ns0:ref><ns0:ref type='bibr' target='#b1'>2</ns0:ref> They have rapidly become the top killer in the country. <ns0:ref type='bibr' target='#b2'>3</ns0:ref> Meanwhile, the high burden of NCDs reduces effective labor supply and productivity; it also increases treatment costs, thus lowering the accumulation of physical capital and impeding economic growth. <ns0:ref type='bibr' target='#b4'>4</ns0:ref> In response to the global NCD epidemic, the United Nations (UN) set a target in 2011 for member countries to achieve a relative 25% reduction from the 2010 level in premature mortality from NCDs by 2025 (referred to as the 25 by 25 target). <ns0:ref type='bibr' target='#b6'>5</ns0:ref> Although studies have estimated the 25 by 25 targets in some countries, <ns0:ref type='bibr' target='#b6'>[5]</ns0:ref><ns0:ref type='bibr' target='#b8'>[6]</ns0:ref><ns0:ref type='bibr' target='#b9'>[7]</ns0:ref> no reports for China have been produced. In addition, because the benefits of controlling NCDs produce are realized gradually, it is urgent to ascertain the target feasibility in China to identify essential efforts for future, more effective interventions.</ns0:p><ns0:p>Our study therefore projects whether the UN target can be met by 2025 in Hunan Province, China, and when and how much excess premature mortality from NCDs due to unfavorable changes occurred to establish the priority for future, more effective interventions. Additionally, we hope that our research provides a useful reference for countries or regions that are also working to better reduce the risk of premature deaths from NCDs.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data source</ns0:head><ns0:p>We conducted the projection in Hunan Province, Central China, where both the burden of years of lost (YLLs) and the ratio of observed to expected disability-adjusted life-years (DALYs) are significantly higher than the national average. <ns0:ref type='bibr' target='#b10'>8</ns0:ref> Based on the Global Burden of Disease Study (GBD) 2016 for China, we extracted death data from 1990 to 2016 for the above four main PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed NCDs for Hunan by age, sex, year, and underlying causes. The GBD study was a collaboration between the Institute for Health Metrics and Evaluation, University of Washington, and the Chinese Center for Disease Control and Prevention (CDC). With a highly standardized method, the mortality was estimated for China based on multisource death surveillance systems or surveys conducted in the country, which mainly consisted of national disease surveillance points system, population death information registry and management system, national maternal and child health surveillance system, local cancer registry and some other mortality reports. <ns0:ref type='bibr' target='#b14'>11</ns0:ref> According to the International Classification of Diseases tenth revision (ICD-10), we distributed the ICD-10 codes for the four NCDs as follows: cancer: C00-C97; CVD: I00-I99; chronic respiratory diseases: J30-J98; and diabetes: E10-E14.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>We set premature mortality from NCDs (also written as premature NCD mortality) as the key indicator for the analysis, with age-standardized rates (ASRs) standardized to the 2010 China census population as a minor indicator. Premature mortality is defined by the WHO as the probability of dying between the age of 30 and 70 from NCDs and is calculated by age-specific death rates with a life table method in the following manner: 2) The was translated into age-specific probability of death : . *5 1 * 2.5</ns0:p><ns0:formula xml:id='formula_0'>x x x M q M &#61501; &#61483;</ns0:formula><ns0:p>3) The probability of death for persons aged 30-70 was calculated last: .</ns0:p></ns0:div> <ns0:div><ns0:head n='1'>()</ns0:head><ns0:p>x q q &#61501; &#61485; &#61485;</ns0:p></ns0:div> <ns0:div><ns0:head>&#61653;</ns0:head><ns0:p>We performed a joinpoint regression to examine trends in premature NCD mortality, with a maximum of three joinpoints set for the analysis. The joinpoint regression describes continuous </ns0:p><ns0:formula xml:id='formula_1'>i i i w b AAPC w &#61676; &#61692; &#61679; &#61679; &#61501; &#61485; &#61620; &#61677; &#61693; &#61679; &#61679; &#61678; &#61694; &#61669; &#61669;</ns0:formula><ns0:p>where represents the slope coefficients for each segment in the years studied, and This allowed us to determine if the relative reduction would be greater than 25%.</ns0:p><ns0:p>To identify excess premature NCD mortality due to unfavorable changes (slowed, stalled or reversed), <ns0:ref type='bibr' target='#b16'>13</ns0:ref> we performed a three-step estimation: First, each most significant APC for the four NCDs was selected as a projection point to find the expected premature mortality, assuming it would continue to decline to 2025 at the same level as the selected APCs. Second, we compared the total differences among observed (1990-2016)-projected (2017-2025) premature mortality with the expected ones to obtain the absolute excess premature mortality. Third, the differences PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed were divided by expected premature mortality to obtain the relative change in the excess premature mortality.</ns0:p><ns0:p>Experimental verification was carried out to evaluate the prediction accuracy of the Joinpoint model. Death data of all NCDs combined during 1990-2011 from the GBD was selected as a sample, to project premature mortality rate for 2012-2016. The results were then compared with the real data with three metrics: Mean Square Error (MSE), Percentage Error (PE) and Mean Absolute Percentage Error (MAPE). <ns0:ref type='bibr' target='#b18'>14</ns0:ref> Among them, , ^2</ns0:p><ns0:p>) </ns0:p><ns0:formula xml:id='formula_3'>n i MSE i n y y &#61501; &#61501; &#61485; &#61669; , ,</ns0:formula></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Temporal trends during 1990-2016 are shown as the AAPC, APC and overall percent change (Table <ns0:ref type='table'>S1</ns0:ref>). Premature mortality from all NCDs combined was projected to be 19.5% (95% CI 19.0%-20.1%). The top contributor to premature mortality was CVD (8.2%, 95% CI: 7.9%-8.5%), followed by cancer (7.9%, 95% CI 7.5%-8.3%). The premature mortality rates for chronic respiratory diseases and diabetes were 1.2% (95% CI 1.2%-1.3%) and 0.6% (95% CI 0.5%-0.6%), respectively. Except for a narrow difference in diabetes, men had greater premature mortality from NCDs than women, with an approximately two-fold difference. Regarding the PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed ASRs in 2025, there will be 377.7 deaths (95% CI 367.5-387.8) per 100,000 persons for all NCDs, with the main contributors being cancer (ASR: 152.5 deaths per 100,000 persons) and CVD (ASR: 143.7 deaths per 100,000 persons) (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_7'>1</ns0:ref> presents the temporal trends for NCDs and the ability to reach the 25 by 25 target in Hunan. A similar trend between the premature mortality and ASRs can be observed. With all NCDs combined, it is not possible to achieve the UN target, as the relative reduction is 16.4%.</ns0:p><ns0:p>Among the subcategories, cancer is the least likely to reach the target, with the smallest relative reduction (11.8%). Another disease failing to meet the target would be CVD, with a 22.1% relative reduction. Both chronic respiratory diseases and diabetes shared a more than 25% relative reduction in premature mortality, with the former showing the greater reduction, at 44.0%.</ns0:p><ns0:p>A difference was seen in the distribution of the top two NCDs for both sexes (Figure <ns0:ref type='figure' target='#fig_8'>2</ns0:ref>). In men, CVD remained the top contributor to premature mortality over time, followed by cancer. In women, cancer will take this position in 2025 due to a faster decline in CVD than in cancer.</ns0:p><ns0:p>Another difference was the ability to reach the target: a relative reduction of 31.6% in women but only 7.8% in men was projected. In women, except for a slightly smaller reduction for cancer (23.0%), the three other subcategories all showed a greater than 25% reduction. The situation is grim, however, for men, in whom only chronic respiratory diseases achieved a greater than 25% reduction (29.5%), and the result for diabetes even showed a 15.8% increase.</ns0:p><ns0:p>During1990-2025, a total absolute excess premature NCD mortality rate of 55.4% and relative excess change of 19.4% were estimated (Figure <ns0:ref type='figure' target='#fig_12'>3A</ns0:ref>). These unfavorable changes mostly occurred from 2008-2009. Among the subcategories (Figure <ns0:ref type='figure' target='#fig_9'>3</ns0:ref>: B to D), CVD showed both higher absolute excess premature mortality (29.6%) and relative excess change (22.8%) than cancer (absolute excess premature mortality: 14.8%; relative change: 11.4%). The greatest excess change (42.1%) was estimated for chronic respiratory disease, despite its much lower absolute excess premature</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed mortality. The absolute excess premature mortality from diabetes was estimated at only 0.8%, whereas its relative excess reached 11.8% during the same period.</ns0:p><ns0:p>The Joinpoint model verification showed that for all NCDs combined projected for 2012-2016 (Table <ns0:ref type='table' target='#tab_5'>2</ns0:ref>), the MSE was estimated to be 0.476. A range of 0.70% to 4.48% was estimated for PE, resulting in a MAPE of 2.79%. Among men, the MSE would be 0.721, and the PE would be from 0.65% to 4.61%, resulting in the MAPE at 2.65%. Among women, the matching values were projected to be: MSE at 0.303, PE ranging from 0.99% to 4.68%, and MAPE at 3.26%.</ns0:p><ns0:p>These results indicated a good prediction accuracy for the Joinpoint model.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Although previous studies have estimated premature mortality from NCDs, each had a different focus. For instance, in the study by Kontis V et al., <ns0:ref type='bibr' target='#b6'>5</ns0:ref> NCDs have experienced unfavorable declines. A possible reason is that the Chinese population has experienced adverse changes in both diet and lifestyle over the past decades. According to the national NCD Risk Factor Surveillances Reports, <ns0:ref type='bibr' target='#b23'>17,</ns0:ref><ns0:ref type='bibr'>18</ns0:ref> an increased prevalence of unhealthy diets and physical inactivity has been seen in China. These factors may contribute to the high of 19&#8226;5% for premature mortality from total NCDs by 2025, with the major contributors CVD (8.2%) and cancer (7.9%). Considering the baseline in 2010, although it is highly likely that chronic respiratory disease and diabetes will achieve the target, it will be very difficult for CVD and cancer, causing the total NCD to also be unlikely to meet the target. These results indicate that premature NCD deaths remain an urgent heath challenge in Hunan and across the country and that both cancer and CVD are the priority NCDs that need to be immediately addressed.</ns0:p><ns0:p>We also found that men had much higher premature mortality than women, with only chronic respiratory diseases expected to reach the target. One reason for the substantial gender differences is men's higher prevalence of major NCDs: The prevalence of obesity among men rose more significantly than among women in the decade 2004-2013, rising from 6.1% to 14.0% versus 7.9% to 14.1% among women. Men had a higher prevalence of hypertension and diabetes, but lower performance in awareness, treatment, or management of both diseases. <ns0:ref type='bibr'>18,</ns0:ref><ns0:ref type='bibr'>19</ns0:ref> In addition, men are more likely than women to be exposed to key risk factors for NCDsin China, also contributing to the difference. Manuscript to be reviewed interventions targeting men to tangibly reduce the number of premature NCD deaths.</ns0:p><ns0:p>Most premature NCD deaths can be prevented or delayed by addressing global health risks.</ns0:p><ns0:p>Among the modifiable risk factors shared by individuals with NCDs, high blood pressure, smoking, a high-salt diet, and ambient particulate matter pollution (PM, mainly PM2.5) exposure are the four leading factors in China. <ns0:ref type='bibr' target='#b10'>8</ns0:ref> Previous studies have shown that premature mortality from NCDs will not show the most favorable decline unless such factors are simultaneously brought under control. <ns0:ref type='bibr' target='#b6'>5,</ns0:ref><ns0:ref type='bibr' target='#b24'>20</ns0:ref> Therefore, a multipronged approach is needed to address the above problems;</ns0:p><ns0:p>specifically, an integrated strategy combining a population-wide intervention targeting the above factors with a strengthened health care system is urgently needed, because the benefits of reducing NCD risk factors are produced gradually.</ns0:p></ns0:div> <ns0:div><ns0:head>Modified high blood pressure control</ns0:head><ns0:p>Every 10 mm Hg reduction in systolic blood pressure significantly lowered the risk of major CVD events (relative risk 0.80, 95% CI 0.77-0.83), resulting in a 13% reduction in all-cause mortality. <ns0:ref type='bibr' target='#b25'>21</ns0:ref> However, high blood pressure management, from awareness to treatment or control, is poor in China. <ns0:ref type='bibr' target='#b2'>3</ns0:ref> A fact is that only 45% of Chinese adults with hypertension were aware of their condition, only 30% were taking anti-hypertensive drugs, and just 7% had achieved normal blood pressure levels. <ns0:ref type='bibr' target='#b26'>22</ns0:ref> A comprehensive, multistage strategy is needed that involves a diet low in salt and rich in polyunsaturated fatty acids, adequate physical activity (no less than a metabolic equivalent of 600 minutes per week), and an improved primary health-care system. Manuscript to be reviewed implementing community-based hypertension screening, as it could have a significant long term impact on systolic blood pressure at the population level. <ns0:ref type='bibr' target='#b27'>23</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>More ambitious tobacco control measures</ns0:head><ns0:p>Although the smoking rate has fallen in many high-income countries, it is rising rapidly in China, with a prevalence of 50.5% in male adults. <ns0:ref type='bibr' target='#b28'>24</ns0:ref> It would impose a high macroeconomic burden of tobacco-attributable NCDs for China: Tobacco-attributable NCDs would cost China 16.7 trillion yuan (US$2.3 trillion in constant 2018 prices) from 2015 to 2030, equivalent to a 0.9% annual tax on aggregate income. Secondhand smoke exposure would contribute to 14% of the burden. <ns0:ref type='bibr' target='#b29'>25</ns0:ref> As the hometown for of the two best-selling brands of cigarettes in China, Hunan has been slow</ns0:p><ns0:p>to take measures for tobacco control. Many successful policies for tobacco control in other countries have shown that a 50% reduction in smoking is feasible. <ns0:ref type='bibr'>26,27-</ns0:ref>Such feasibility could be achieved by the following actions: First, raise cigarette taxes. Although China implemented a tax linkage in 2015, raising the wholesale and price taxes on cigarettes from 5% to 11%, <ns0:ref type='bibr' target='#b33'>28</ns0:ref> cigarettes are much more affordable here than in other countries, with most costing 10 CNY (1.4 USD) a package. It is estimated that a 50% price increase in cigarettes due to taxes in China would yield an additional 231 million years of life. <ns0:ref type='bibr' target='#b34'>29</ns0:ref> Therefore, a higher cigarette tax rate should be the first action. Second, regulate smoking with reference to practices in developed cities such as Beijing, Shanghai, and Shenzhen by comprehensively enacting smoking bans in indoor workplaces, indoor public places, and public transportation. Third, enforce strict bans on tobacco advertising, sponsorship, or any other activity that may weaken smoking control. Fourth, reform the design of cigarette packaging. Cigarette packages in China are all beautifully designed due to a deeply rooted smoking culture, and the warning occupies only a small part of the design space.</ns0:p><ns0:p>Packaging should be redesigned to feature the warning in text and graphics. Fifth, conduct targeted health education. The public's awareness is often confined to 'smoking is harmful to health', and many people do not know exactly how smoking is harmful. Health education could be conducted by combining traditional media with new media (such as the Internet or WeChat)</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed or through health campaigns hosted by professional doctors.</ns0:p></ns0:div> <ns0:div><ns0:head>A stepwise reduction in salt intake</ns0:head><ns0:p>High salt consumption is the leading cause of hypertension and is strongly tied to stroke in China. <ns0:ref type='bibr' target='#b36'>30</ns0:ref> Chinese people have a daily average salt intake of 12-14 g, <ns0:ref type='bibr' target='#b36'>30,</ns0:ref><ns0:ref type='bibr' target='#b38'>31</ns0:ref> much greater than the WHO recommendation of &lt; 5 g/day. Due to the characteristics of Hunan cuisine, residents' diets are particularly high in salt. We suggest a shift strategy for salt reduction involving both commercial foods and consumer behaviors. First, reduce salt in commercial or processed foods through an industry-wide shift. The key is gradual salt reduction in small steps. Following the UK's practice, <ns0:ref type='bibr' target='#b39'>32</ns0:ref> gradually lower salt targets (such as a 20% decrease) can be set in high-salt categories for the food industry. It is also encouraged that alternatives with the same or better taste be proposed, such as 'less salt, more spices' or a 'more potassium, less sodium' diet, which have been shown to be helpful in reducing blood pressure and CVD mortality. <ns0:ref type='bibr' target='#b40'>33</ns0:ref> Second, shift consumer awareness to action. National salt campaigns can not only raise consumer awareness but also have a remarkable impact on salt intake in the population. In 2017, the Action on Salt China (ASC) program was established with four cluster randomized controlled trial packages. <ns0:ref type='bibr' target='#b41'>34</ns0:ref> Hunan participated in the program at five county-level locations, but with 130 county-level areas in the province, actual participation was low. Thus, we suggest expanding salt-reducing interventions from the ASC to the whole province. The salt industry should also take responsibility, for instance, by developing saltshakers with smaller holes or convenient salt intake calculators to help consumers make essential behavior changes.</ns0:p></ns0:div> <ns0:div><ns0:head>Reduce harmful alcohol intake</ns0:head><ns0:p>Health issues attributable to alcohol use, such as CVD and cancer, have been largely underemphasized in China. During the past 30 years, a striking increase has been seen in alcohol use among Chinese men, greater than that in most other countries, <ns0:ref type='bibr' target='#b42'>35</ns0:ref> and this trend is forecasted to continue. <ns0:ref type='bibr' target='#b43'>36</ns0:ref> This increase is strongly associated with robust economic development and a</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed deeply rooted alcohol culture. However, the government can play a substantial role in developing alcohol use policies to alter drinking levels. A good example would be Russia, where WHO's recommended best buys interventions for alcohol use, including taxation, availability restrictions, and bans on marketing, were effectively carried out, leading to remarkable changes in both alcohol use and the burden of alcohol-related disease. <ns0:ref type='bibr' target='#b44'>37</ns0:ref> For Hunan, similar interventions are also needed and should be broadened through strict restrictions on alcohol advertising on television, legally binding regulations on alcohol sponsorship, and heavy punishments for drunk drivers.</ns0:p></ns0:div> <ns0:div><ns0:head>Multisector cooperation to control PM pollution</ns0:head><ns0:p>Although the WHO has not set targets for environmental risk factors, PM pollution should be seriously addressed due to its striking position as the fourth-leading risk factor for death in China. <ns0:ref type='bibr' target='#b10'>8</ns0:ref> PM pollution causes 3&#8226;3 million premature deaths worldwide each year, with China being the largest contributor. <ns0:ref type='bibr' target='#b45'>38</ns0:ref> Meanwhile, PM pollution has a huge macroeconomic impact on NCD in China, where total losses from NCDs associated with air pollution were estimated to be $499 billion (constant 2010 USD) from 2015-2030. <ns0:ref type='bibr' target='#b46'>39</ns0:ref> As in many other regions in China, PM2. Manuscript to be reviewed hospitals. <ns0:ref type='bibr' target='#b50'>42,</ns0:ref><ns0:ref type='bibr' target='#b51'>43</ns0:ref> However, these measures impose costs, involving human resources, economic losses, public engagement, coordinated governance structures at the national and local levels, etc.</ns0:p><ns0:p>Thus we need to think about long-term plans for epidemic control. In the post lock-down COVID-19 era in China, the management of NCDs becomes especially important as people with underlying chronic diseases are more likely to die from COVID-19. <ns0:ref type='bibr' target='#b52'>44</ns0:ref> Reducing the risk factors in NCDs such as smoking and air pollution can also lower the risk of death for COVID-19. <ns0:ref type='bibr' target='#b53'>45,</ns0:ref><ns0:ref type='bibr' target='#b54'>46</ns0:ref> Our study is subject to some limitations. First, the data were derived from the GBD 2016, and all the limitations in the GBD study are also applicable to this study. Second, only five risk factors were addressed in the recommended interventions because the effects of other factors can be partially replaced, but this may affect the most effective control of future premature NCD deaths.</ns0:p><ns0:p>Third, we conducted the projection under the current trend, with no consideration for possible greater efforts to reduce NCD risk factors in the future. However, the benefits of lowering the risk factors are produced gradually, which should have a small impact on the results of the present study. Additionally, our projection did not consider the change of age structure of the population during the period studied. This may weaken to some extent, the extrapolation of the model presentation. Manuscript to be reviewed or not (narrower than the chart area).</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:note type='other'>Figure legends</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020) Manuscript to be reviewed changes by connecting several line segments on a log scale and identifies statistically significant changes with a Monte Carlo permutation test. 11,12 The three following indicators, annual percentage change (APC), average APC (AAPC) and overall percent change, reflected different changes in temporal trends. The former two indicators were calculated by = {exp(b i ) -1} x</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>of each segment in the intervals. The final indicator was calculated with an AAPC-based exponentiation function by first converting the AAPC to the predicted single-year change and then exponentiating to the number of study years minus one to produce the overall change and its magnitude, which was finally converted to a percent change.<ns0:ref type='bibr' target='#b14'>11</ns0:ref> We projected premature mortality from NCDs (as well as ASRs) with 95% confidence intervals (CIs) for 2025 by fitting a linear regression over the most recent trend identified by the joinpoint model. To estimate the ability to meet the UN target in Hunan, we compared the projected premature mortality from NCDs in the province in 2025 with the level in 2010 to find the relative reduction with the following formula:</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>the authors highlighted the impacts of achieving the WHO agreed six risk factors (tobacco and alcohol use, salt intake, obesity, raised blood pressure and glucose) targets on reaching the 25 by 25 target. The authors calculated a time-based population impact fraction to identify relative reductions in the premature mortality through reanalyses and meta-analyses of epidemiological studies. In another study by Norheim OF et al., 16 the authors proposed a more ambitious goal of avoiding 40% of premature deaths from all causes globally by 2030, beyond the current UN sustainable development goal (reducing premature mortality from NCDs by one-third by 2030). They reviewed the UN-based overall 1970-2010 mortality and WHO-based cause-specific 2000-2010 mortality. They concluded such a target could be achieved by moderately accelerating the current mortality decrease during 2000-2010. Unlike these two studies, our study estimated the 25 by 25 target's feasibility at the local rather than the global level. As NCDs account for high proportions of the disease burden and all deaths, PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020) Manuscript to be reviewed as mentioned above, we emphasized premature deaths from NCDs rather than all causes to set future control priorities. Based on a Joinpoint regression model, we derived 1990-2016 mortality for Hunan from the GBD 2016 to project premature mortality from NCDs and the excess situation for 2025, assessing the ability to meet the target here. Through the projection, we found that although premature mortality from NCDs in Hunan has continuously declined since 1990, this decline is insufficient to reach the 25 by 25 target. Especially since 2008-2009, almost all</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Health authorities and professional</ns0:head><ns0:label /><ns0:figDesc>institutions need to work together to promote an integrated prevention-control-treatment model at the community level leveraging the internet and health information technology. Within such a model, a hypertension outpatient service in community medical institutions and a family doctor contracting service are required to provide individuals with regular hypertension management with respect to screening, essential anti-hypertensive medications, health counseling, follow-up services, etc. It is particularly crucial for PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>5 reduction</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>control in Hunan has just begun. The local government released in 2018 a 3-year action plan to reduce the annual PM2.5 concentration to less than 40 &#956;g/m 3 by 2020, 40 but this plan still falls far short of the WHO guidelines. 41 Thus, intersector collaboration in public-private partnerships should be encouraged by setting PM2.5 levels as an assessment indicator for local government development, supervising pollution from industrial enterprises, establishing rewardand-punishment mechanisms, prompting responses to heavy pollution weather, etc. When discussing NCDs control, we must recognize that the COVID-19 is now capturing the world focus. As the COVID-19 pandemic is raging worldwide and spreading fast in many countries, China has largely controlled the epidemic. The Chinese government has spent substantial efforts in controlling the COVID-19 epidemic--such as lockdown, extension of the lunar new year holiday, and facility isolation of mild to moderate cases using Fangcang shelter PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Despite a continuous decline</ns0:head><ns0:label /><ns0:figDesc>in premature mortality from NCDs in Hunan, China, the decline slowed ten years ago. Premature NCD deaths remain high and are unlikely, particularly in men, to reach the 25 by 25 target by 2025. More bold actions combining population-wide interventions for key risk factors with improved health-care systems are urgently needed.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Premature mortality from NCDs and their ASRs in Hunan, China, from observed years (1990-2016) to projected years (2017-2025). The ability to meet the 25 by 25 target differed across the total NCDs and subcategories: (A) to (C) and (F) will not meet the target with reductions less than 25%, while (D) and (E) will with reductions greater than 25%.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Observed and projected premature NCD mortality by sex in Hunan, China, 1990-2025.Women will meet the target by 2025 except with respect to cancer, while men will not except with respect to chronic respiratory disease. The black dashed line is used to distinguish observed years (1990-2016) from projected years (2017-2025). Five colored lines matching each chart area are shown for the target range and are used to test whether the target will be met (wider than</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Excess premature mortality from NCDs in Hunan, China, 1990-2025. Yellow line: observed premature mortality from 1990 to 2016 and projected values from 2017 to 2025; grey line: favorable premature mortality from 1990 to 2025; blue area: total absolute excess premature mortality from the NCDs during the period.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 1 Premature</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>Observed and projected premature NCD mortality by sex in Hunan,China, 1990-2025 The black dashed line is used to distinguish observed years (1990-2016) from projected years (2017-2025). Five colored lines matching each chart area are shown for the target range and are used to test whether the target will be met (wider than the chart area in 2025)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>Figure 3 Excess</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc><ns0:ref type='bibr' target='#b13'>10</ns0:ref> </ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>1 Table 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Observed premature mortality from NCDs, ASRs in 2016 and predicted values for 2025, Hunan Province, China Premature mortality was defined as the probability (%) of dying aged 30-70 from NCDs. Rates standardized to the 2010 China census population with age groups 30-34, 35-39&#8230; and 65-79 years, in per 100000 populations.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>CVD</ns0:cell><ns0:cell>6.3</ns0:cell><ns0:cell>108.3</ns0:cell><ns0:cell>4.8(4.5-5.0)</ns0:cell><ns0:cell>80.6(77.3-84.0)</ns0:cell><ns0:cell>-23.8</ns0:cell></ns0:row><ns0:row><ns0:cell>Sex</ns0:cell><ns0:cell>Diabetes Diseases Chronic respiratory disease Other NCDs</ns0:cell><ns0:cell cols='2'>Observed 2016 0.5 8.5 premature mortality rate a ASR b 0.9 14.9 2.2 39.1</ns0:cell><ns0:cell>0.4(0.4-0.5) Predicted 2025 premature mortality rate (95% CI) 0.6(0.4-0.7) 1.9(1.8-1.9)</ns0:cell><ns0:cell>6.8(6.5-7.0) 8.5(7.7-9.4) ASR (95% CI) 32.5(32.1-33.0)</ns0:cell><ns0:cell>-20.0 Percent Change in premature -33.3 mortality rate -13.6</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>Abbreviations: NCDs, non-communicable diseases; ASRs, age-standardized rates; CI, Confidential Interval; CVD, cardiovascular</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>disease.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>a b PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Verification results from Joinpoint model: Based on observed premature mortality from NCDs during 1990-2011 and projected values for 2012-2016</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Verification results from Joinpoint model: Based on observed premature mortality from NCDs during 1990-2011 and projected values for 2012-2016</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Sex</ns0:cell><ns0:cell>Items</ns0:cell><ns0:cell>2012</ns0:cell><ns0:cell>2013</ns0:cell><ns0:cell>2014</ns0:cell><ns0:cell>2015</ns0:cell><ns0:cell>2016</ns0:cell></ns0:row><ns0:row><ns0:cell>male</ns0:cell><ns0:cell cols='2'>projected data 1 28.19</ns0:cell><ns0:cell>27.88</ns0:cell><ns0:cell>27.58</ns0:cell><ns0:cell>27.28</ns0:cell><ns0:cell>26.98</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>real data 2</ns0:cell><ns0:cell>28.37</ns0:cell><ns0:cell>28.62</ns0:cell><ns0:cell>28.91</ns0:cell><ns0:cell>28.26</ns0:cell><ns0:cell>27.51</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>PE(%)</ns0:cell><ns0:cell>0.65</ns0:cell><ns0:cell>2.59</ns0:cell><ns0:cell>4.61</ns0:cell><ns0:cell>3.47</ns0:cell><ns0:cell>1.93</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>MAPE(%)</ns0:cell><ns0:cell>2.65</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>MSE</ns0:cell><ns0:cell>0.721</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>female</ns0:cell><ns0:cell cols='2'>projected data 16.03</ns0:cell><ns0:cell>15.49</ns0:cell><ns0:cell>14.96</ns0:cell><ns0:cell>14.46</ns0:cell><ns0:cell>13.97</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>real data</ns0:cell><ns0:cell>16.19</ns0:cell><ns0:cell>15.75</ns0:cell><ns0:cell>15.64</ns0:cell><ns0:cell>15.16</ns0:cell><ns0:cell>14.66</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>PE(%)</ns0:cell><ns0:cell>0.99</ns0:cell><ns0:cell>1.67</ns0:cell><ns0:cell>4.32</ns0:cell><ns0:cell>4.64</ns0:cell><ns0:cell>4.68</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>MAPE(%)</ns0:cell><ns0:cell>3.26</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>MSE</ns0:cell><ns0:cell>0.303</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>both</ns0:cell><ns0:cell cols='2'>projected data 22.56</ns0:cell><ns0:cell>22.15</ns0:cell><ns0:cell>21.74</ns0:cell><ns0:cell>21.34</ns0:cell><ns0:cell>20.94</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>real data</ns0:cell><ns0:cell>22.72</ns0:cell><ns0:cell>22.66</ns0:cell><ns0:cell>22.76</ns0:cell><ns0:cell>22.18</ns0:cell><ns0:cell>21.53</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>PE(%)</ns0:cell><ns0:cell>0.70</ns0:cell><ns0:cell>2.24</ns0:cell><ns0:cell>4.48</ns0:cell><ns0:cell>3.79</ns0:cell><ns0:cell>2.74</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>MAPE(%)</ns0:cell><ns0:cell>2.79</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>MSE</ns0:cell><ns0:cell>0.476</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='1'>Projected data=premature mortality rate from all all NCDs combined projected for 2012-2016.2 Real data=observed premature mortality rate from all all NCDs during 1990-2011. Abbreviations: MSE=Mean Square Error, PE=Percentage Error, MAPE=Mean Absolute Percentage Error. PeerJ reviewing PDF | (2020:07:50885:1:1:NEW 9 Oct 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Replies to editor and reviewer comments Note: The following new line represents the line number in the resubmitted manuscript with no tracked changes. Re: means replies. Eidtor: Please check the statistics as suggested by reviewer #4 and comparison to previously published methods as suggested by reviewer #2. Re: The statistics verification suggested by reviewer #4 has been added in method section, new lines 135-142, and results section, lines 188-193, respectively. The comparison suggested by reviewer #2 were added in the first three paragraphs of discussion, in new lines 196-238. [# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. Re: It was shown in the following replies to reviewers in the point-by-point structure . [# PeerJ Staff Note: The review process has identified that the English language must be improved. Re: Once the resubmitted manuscript has been reviewed by you and the reviewers, we will order your journal language editing services for the manuscript. Reviewer 1: p6. line 44 what are the NCDs on China? What diseases are included on NCD list? clarify on text. Re: Four main NCDs (cardiovascular disease, cancer, diabetes and chronic respiratory diseases) were added in new lines 57-58. p.7 line 69. clarify what systems or surveys Re: The multi-source death surveillance systems or surveys were added in details in new lines 86-88. p. 12 line 192: refference. Re: A reference was added in new line 254. p. 15 Study limitations can exist due to constraints on research design or methodology, and these factors may impact the findings of your study. 'estimations should be interpreted with caution': this sentence minimizes the importance and reliability of your study. So, remove this topic and dicuss this on first paragraph of discussion. Re: The sentence 'estimations should be interpreted with caution' was removed. According to the discussion guide in the Peer J manuscript template, the limitation topic was discussed and placed on the last rather than the first paragraph of discussion. Reviewer 2: Basic reporting -- line 50: ... by 25 targets in some countries, no reports for China have been produced. Please cite previous studies estimating 25*25 targets. Re: The related citations were added in new line 67. -- line 80: citation for life table method is wrong. Re: Revised in new line 99. Experimental design 1. Please use the most up-to-date GBD data. As far as I know, there is already GBD 2017, the authors should double-check and use the latest data source. Re: At the beginning of our study, the latest GBD data source obtained from the Chinese CDC was the 2016 version. Months later, we got GBD 2017 and also analyzed its data. However, we found the two GBD datasets were contradictory in premature mortality from NCDs. From 1990 to 2008 or so, both datasets showed a continuously downward trend in the premature mortality. After that, GBD 2016 maintained the trend, while GBD 2017 turned to an upward trend. We considered various possibilities for this upward trend, but could not find a reasonable explanation. China has made substantial progress in reducing the burden of many diseases and disabilities. According to earlier GBD-based studies by Li Y et al. (doi:10.1186/s12916-017-0894-5), Zeng X et al. (doi:10.3760/cma.j.issn.0253-9624.2017.03.004) and Zhou M et al.(doi: 10.3760/cma.j.issn.0254-6450.2016.11.001), premature mortality from NCDs in China experienced a continuously downward trend from 1990 to 2013, and from 1990 to 2015. Also, life expectancy and healthy life expectancy of the Chinese increased continuously from 1990 to 2015. Besides, since a death surveillance points system, both nationally and provincially representative, was established in late 2013, death surveillance data for Hunan in the past years have also declined continuously in the premature mortality. We discussed these unusual results from the GBD 2017 with the relevant colleagues from the Chinese CDC. They also found these data unusual. Therefore, we didn’t consider it appropriate to extend our study period from 2016 to 2017. We can review this topic once the latest dataset is available. 2. The health statistical year book (2020) in China could have the most up-to-date death data for Hunan. If so, it would be better to use the health statistical year book rather than GBD data. Re: Health statistical yearbook (2020) in China has not yet been published. In the 2019 edition, there were only overall national crude death rates from sample areas, by urban and rural regions in four given years (2005, 2010, 2015, 2018). No death data was found there for Hunan. 3. line 82: better to write the denominator as 'Mid-year population aged(x, x+5)' Re: Revised in new line 101. 4. better to discuss the differences in methods and findings with previous studies on this topic: -- Kontis V, Mathers CD, Rehm J, Stevens GA, Shield KD, Bonita R, Riley LM, Poznyak V, Beaglehole R, Ezzati M. Contribution of six risk factors to achieving the 25× 25 non-communicable disease mortality reduction target: a modelling study. The Lancet. 2014 Aug 2;384(9941):427-37. -- Avoiding 40% of the premature deaths in each country, 2010–30: review of national mortality trends to help quantify the UN Sustainable Development Goal for health Re: The differences with two previous studies, as you listed, were added in the first three paragraphs of discussion, in new lines 196-238. Validity of the findings The discussions are well-written. I think the paper can be stronger by adding more discussions on 1) the broader social and economic impact of NCDs and related risk factors in China, 2) effective prevention strategies, and 3) the covid-19 pandemic, and 4) the assumptions in this study and how the assumptions may affect the results. 1) The broader social and economic impact of NCDs and related risk factors: -- Could further expand the discussion of the NCDs on the macroeconomic impact: the high burden of NCDs reduces effective labor supply--both through mortality (reduction in quantity) and morbidity (reduction in quality); it also increases treatment costs, thus lowering the accumulation of physical capital and impede economic growth. Could cite: Bloom DE, Chen S, Kuhn M, McGovern ME, Oxley L, Prettner K. The economic burden of chronic diseases: Estimates and projections for China, Japan, and South Korea. J Econ Ageing 2018; [Epub ahead of print]; Re: To better describe the NCD epidemic situation, we put the macroeconomic impact of NCDs in the first paragraph of the introduction, new lines 60-62, rather than in the discussion section. -- As for PM pollution, better to add discussions on the total losses from NCDs associated with air pollution in China-- $499 billion (constant 2010 USD) from 2015–2030. Could cite: the Chen S, Bloom DE. The macroeconomic burden of noncommunicable diseases associated with air pollution in China. PLoS One 2019; 14(4): e0215663. Re: Added in new lines 331-333. -- Tobacco-attributable NCDs affect China’s productive capacity--impose a total cost of 16.7 trillion yuan (US$2.3 trillion, in constant 2018 prices) in the period 2015–30, which corresponds to an annual tax of 0.9 percent on aggregate income. Secondhand smoke exposure accounts for 14 percent of the burden. Could cite: Chen S, Kuhn M, Prettner K, Bloom DE. Noncommunicable Diseases Attributable To Tobacco Use In China: Macroeconomic Burden And Tobacco Control Policies. Health Affairs 2019. Re: Added in new lines 271-273. 2) The prevention strategies that can help in reaching the target: --When discussing the high blood pressure control issues, the author can add that community based hypertension screening and encouraging people with raised blood pressure to seek care and adopt blood pressure lowering behaviour changes could have important long term impact on systolic blood pressure at the population level. Could cite: -- Ettehad D, Emdin CA, Kiran A, et al. Blood pressure lowering for prevention of cardiovascular disease and death: a systematic review and meta-analysis. Lancet 2016;387:957-67. doi:10.1016/S0140- 6736(15)01225-8; Chen S, Sudharsanan N, Huang F, Liu Y, Geldsetzer P, Bärnighausen T. Impact of community based screening for hypertension on blood pressure after two years: Regression discontinuity analysis in a national cohort of older adults in China. BMJ 2019; 366: l4064.; Sudharsanan N, Chen S, Garber M, Bärnighausen T, Geldsetzer P. The effect of home-based hypertension screening on blood pressure change over time in South Africa. Health Affairs 2020; 39(1): 124-32. Re: In terms of community-based hypertension screening and the correlation between systolic blood pressure lowering and the risk of CVD events, the relevant references except for the study by Sudharsanan N et al., were cited in new lines 251-253 and 264-266. We did not cite this study because it focused on a home-based screening, which was beyond the scope of our discussions. Additionally, it showed reductions in systolic blood pressure only for women and younger men but not for older men, presenting an incomplete finding for both sexes. -- And the factor that only 45% of Chinese adults with hypertension were aware of their condition, only 30% were taking antihypertensive drugs, and just 7% had achieved normal blood pressure levels. Could cite: Lu J, Lu Y, Wang X, et al. Prevalence, awareness, treatment, and control of hypertension in China: data from 1·7 million adults in a population-based screening study (China PEACE Million Persons Project). Lancet 2017;390:2549-58. doi:10.1016/S0140- 6736(17)32478-9 Re: The situation of awareness, treatment, and control of hypertension was added in new lines 254-256. 3) Adding some discussions on the importance of managing chronic diseases during the covid-19 pandemic China has spent substantial efforts in controlling the epidemic-- such as lockdown, extension of lunar new year holiday, and facility isolation of mild cases using Fangcang shelter hospitals, however, these measures impose costs (can cite:--Chen S, Yang J, Yang W, Wang C, Bärnighausen T. COVID-19 control in China during mass population movements at New Year. The Lancet 2020; 395(10226): 764-6. --Chen S, Zhang Z, Yang J, et al. Fangcang shelter hospitals: a novel concept for responding to public health emergencies. The Lancet 2020; 395(10232): 1305-14.). Thus we need to think about long-term plans for epidemic control. In the post lock-down covid-19 era in China, management of chronic diseases becomes especially important as people with underlying chronic diseases are more likely to die from covid-19 (cite: Wu Z, McGoogan JM. Characteristics of and important lessons from the coronavirus disease 2019 (COVID-19) outbreak in China: summary of a report of 72 314 cases from the Chinese Center for Disease Control and Prevention. Jama. 2020 Apr 7;323(13):1239-42) and reducing the risk factors in chronic diseases such as smoking and air pollution can also lower the risk of death for covid-19. (could cite: Chen K, Wang M, Huang C, Kinney PL, Anastas PT. Air pollution reduction and mortality benefit during the COVID-19 outbreak in China. The Lancet Planetary Health. 2020 Jun 1;4(6):e210-2.) Re: That’s a good point. Added in new lines 341-351. 4) better to add discussions on the assumptions in this study and how the assumptions may affect the results Re: The assumptions and the possible affects were put in the study limitation section, new lines 357-362. Reviewer 4: 1. Raw data not supplied. Re: Provided in the resubmitted supplementary files. 2. When determining mortality in 5-year age groups (lines 81 and 82), the terms “age grops”, “Total death ... between age (x) and exact age (x + 5)”, “Total population between age (x ) and exact age (x + 5) '. However, in medical reporting age is usually given as “full years,” as a result, the group “55 to 60 years” would include 6 years instead of 5. The term 'exact age' used in the definition is not unambiguous. Therefore, it is desirable to give an example of the definition of age groups. A more unambiguous definition of the concept of “age” used in the article is also required, since age can be determined both by the year of birth and at the time of the origin of the event under study. In this case, the population is usually determined as at the beginning of the year, and for death - the age at the time of death. If the age in determining the population size and in the deceased is determined differently, this leads to a systematic error in determining the dependence of mortality on age, and it is advisable to make adjustments for this bias effect (by an average of 0.5 years). Re: An example of the definition of age groups was added in new line 100. The 5-year age group was defined as [x, x+5) (e.g. [30, 35) means 30-34). The term “exact age” was replaced with “age” (new line 101), which was defined as the full years from birth date to death date. 3. When predicting the dynamics of mortality, the authors use statistical methods that not only give a forecast, but also estimate its expected accuracy. However, these methods evaluate the accuracy of the forecast using a number of assumptions that may not be met. In such cases, it is advisable not to restrict ourselves to the obtained estimate of the forecast accuracy, but also to carry out experimental verification. In this regard, I would recommend the authors (if technically feasible) to carry out such a check and provide data on its results in the article. This can be done, for example, in the following way. The work uses data for 1970-2016. It is possible to forecast the data for 1970-2011 and compare the results with the actual data for 2012-2016. Re: No death data were available until 1990s. We used the GBD data from 1990 to 2011 to verify the accuracy of the Joinpoint model prediction (2012-2016). Detailed descriptions and results were added in new lines 135-142 and lines 188-193, respectively. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Non-specific low back pain (LBP) is the leading cause of disability worldwide. The primary physiotherapeutic treatment for LBP is physical exercise, but evidence suggesting a specific exercise as most appropriate for any given case is limited.</ns0:p><ns0:p>Objective. To determine if specific stabilization exercise (SSE) is more effective than traditional trunk exercise (TTE) in reducing levels of pain, disability and inflammation in women with non-specific low back pain (LBP).</ns0:p></ns0:div> <ns0:div><ns0:head>Design.</ns0:head><ns0:p>A pilot randomized controlled trial was conducted in Rovira i Virgili University, Catalonia.</ns0:p><ns0:p>Methods. 39 females experiencing non-specific LBP were included in two groups: the TTE program and SSE program, both were conducted by a physiotherapist during twenty sessions. The primary outcome was pain intensity (10-cm Visual Analogue Scale). Secondary outcomes were disability (Roland Morris Disability Questionnaire), and inflammation (IL-6 and TNF-&#945; plasma levels). Measurements were taken at baseline, at half intervention, at post-intervention, and a month later.</ns0:p><ns0:p>Results: Mean group differences in change from baseline to post-intervention for TTE were: -4.5 points (CI 3.3 to 5.6) for pain, -5.1 points (CI 3.0 to 7.3) for disability, 0.19 pg/mL (95% CI -1.6 to 1.2) for IL-6 levels, and 46.2 pg/mL (CI 13.0 to 85.3) for TNF-&#945; levels. For SSE, differences were: -4.3 points (CI 3.1 to 5.6) for pain, -6.1 points (CI 3.7 to 8.6) for disability, 1.1 pg/mL (CI 0.0 to 2.1) for IL-6 levels , and 12.8 pg/mL (95% CI -42.3 to 16.7) for TNF-&#945; levels. There were an insignificant effect size and no statistically significant overall mean differences between both groups.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion:</ns0:head><ns0:p>This study suggests that both interventions (traditional trunk and specific stabilization exercises) are effective in reducing pain and disability in non-specific LBP patients, but the two programs produce different degrees of inflammation change.</ns0:p><ns0:p>Clinical trial registration number: NCT02103036.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Common low back pain (LBP) is defined as pain between the costal margins and the inferior gluteal folds, usually accompanied by painful limitation of movement, often influenced by physical activities and posture, and which may be associated with referred pain in the leg <ns0:ref type='bibr' target='#b25'>(Kovacs et al., 2006)</ns0:ref>. LBP represents an important public health problem, because worldwide prevalence of the condition ranges from 12% to 33% <ns0:ref type='bibr' target='#b59'>(Walker, 2000)</ns0:ref>. It is also known that LBP is more prevalent in females than males; for example, in 2015 in Catalonia, 30.1% of females suffered LBP, as compared to 18.7% of males <ns0:ref type='bibr' target='#b10'>(Garcia et al., 2016)</ns0:ref>. LBP remains a common disabling condition <ns0:ref type='bibr' target='#b60'>(Walker et al., 2004)</ns0:ref> and is associated with high costs for medical health and social care <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b64'>Wieser et al., 2011)</ns0:ref>.</ns0:p><ns0:p>One of the difficulties in reducing the burden of spinal disorders is the wide and heterogeneous range of specific diseases and non-specific musculoskeletal disorders that can involve the spinal column, most of which manifest pain <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012)</ns0:ref>. Despite this factor, or perhaps because of its impact on individuals, their families, and the healthcare systems, spinal disorders remain one of the most controversial and challenging conditions <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b18'>D. Hoy et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Back pain is sometimes associated with a likely aetiology (e.g.radiculopathy or spinal stenosis), but most LBP cases are of unknown origin and are classified as non-specific, which has also been described as mechanical pain or strain, account for 90% or more of all people experiencing spinal pain <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b57'>M. W. Van Tulder et al., 1997)</ns0:ref>. For all these reasons, it is necessary to find an effective treatment for LBP. Unfortunately, the scientific literature does not offer relevant conclusions, due in part to the poor methodology employed in many published studies -e.g., short follow-up periods, population heterogeneity and non-validated measurements <ns0:ref type='bibr' target='#b1'>(Atlas &amp; Nardin, 2003)</ns0:ref>.</ns0:p><ns0:p>According to current clinical reviews and guides, first-line treatments for LBP pathology focus on analgesic measures <ns0:ref type='bibr' target='#b22'>(Koes et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b55'>M. Van Tulder et al., 2006)</ns0:ref>. Conservative physiotherapeutic treatments for LBP also exist <ns0:ref type='bibr' target='#b4'>(Bordas et al., 2004)</ns0:ref>, including advice and postural education, electrotherapy, manual therapy, and physical exercises <ns0:ref type='bibr' target='#b66'>(Williams et al., 2014)</ns0:ref>. Exercise is one of the chief recommendations for pain reduction, mobility increase, improvement of physical and psychological abilities and anxiety reduction <ns0:ref type='bibr' target='#b58'>(Waddell &amp; Burton, 2005)</ns0:ref>. The problem with exercise programs lies in the fact that rationales for choosing the appropriate exercise for an individual case are very weak. Controversy arises because the types of exercise programs for LBP vary considerably, as do the types of patients. This makes it very unlikely that a particular program will be equally effective in all cases <ns0:ref type='bibr' target='#b28'>(Macedo et al., 2008)</ns0:ref>.</ns0:p><ns0:p>One exercise option is the back school program, a therapeutic program including information on the anatomy of the back, biomechanics, optimal posture, ergonomics, and back exercises <ns0:ref type='bibr' target='#b35'>(Parreira et al., 2016)</ns0:ref>, which has proven effective in reducing pain and disability <ns0:ref type='bibr' target='#b42'>(Sahin et al., 2011)</ns0:ref>. There are many types of back school exercise programs, but considerable discussion has centered on the question of whether the specific stabilization exercise program (SSEP) -in which the deep muscles are the protagonists -is preferable to the traditional trunk exercise program (TTEP), which includes more exercises for strengthening abdominal and back muscles. There are systematic reviews that support the idea that SSEP is superior to TTEP <ns0:ref type='bibr' target='#b27'>(Lederman, 2010)</ns0:ref>, but there are also studies that have found both approaches equally effective for improvement of LBP in terms of pain and disability <ns0:ref type='bibr' target='#b44'>(Shamsi et al., 2015)</ns0:ref>.</ns0:p><ns0:p>On the other hand, pro-inflammatory cytokines levels were detected on assessing local tissue in adults with LBP <ns0:ref type='bibr' target='#b39'>(Queiroz et al., 2015)</ns0:ref> and evidence shows that physical exercise therapy decreases systemic inflammatory mediators production, this demonstrates its clinical relevance <ns0:ref type='bibr' target='#b37'>(Pereira et al., 2013)</ns0:ref>.</ns0:p><ns0:p>After years of research into LBP treatment, and taking into account the variety of treatment options, exercise continues to be accepted as an effective approach. The question remains, however, of which type of exercise is most effective in treating various patient subgroups <ns0:ref type='bibr' target='#b1'>(Atlas &amp; Nardin, 2003;</ns0:ref><ns0:ref type='bibr' target='#b43'>Saner et al., 2011)</ns0:ref>. There is limited evidence that the specific stabilization exercise program is more effective than the traditional trunk exercise for patients with nonspecific LBP. Therefore, the following were the research questions this study sought to answer:</ns0:p><ns0:p>1.</ns0:p><ns0:p>Is the specific stabilization exercise program more effective than the traditional trunk exercise program in reducing levels of pain and disability in women with non-specific LBP? 2.</ns0:p><ns0:p>Which type of back school exercise produces different degrees of inflammation change in women with non-specific LBP?</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Ethics</ns0:head><ns0:p>The Clinical Ethics Committee of University Hospital Sant Joan of Reus approved this study (12-06-28/6proj4). All participants gave written informed consent before data collection began.</ns0:p></ns0:div> <ns0:div><ns0:head>Study design</ns0:head><ns0:p>A pilot randomized trial was conducted in Catalonia from February 2013 to February 2015 (NCT02103036). Participants, diagnosed by medical practitioners and referred for treatment of non-specific LBP, were randomized using computer-generated random number tables into two treatment groups: a TTEP group and a SSEP group. Afterwards, measurements were taken at baseline (session 0), at half intervention (session 10), post-intervention (session 20) and one month later. A single-blind study was conducted, due to the impossibility of achieving double blindness -the physiotherapist performing the intervention had to know which treatment each participant was to receive. The therapist who performed the intervention was a qualified health professional with 5 years' experience in the field; a different professional took all the study measurements, he was blinded to the participant's assignment. Participants were blinded to their group allocation, design and hypotheses <ns0:ref type='bibr' target='#b33'>(Page &amp; Persch, 2013)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Participants</ns0:head><ns0:p>Participants entering the trial were required to meet the following inclusion criteria: females aged between 18 and 70 years; diagnosed with non-specific LBP (fewer than 6 weeks of pain duration) by a specialist doctor who used imaging such as magnetic resonance, radiographic or computed axial tomography to rule out other spinal disorders, and under no pharmacological treatment for pain. Exclusion criteria were: diagnosed with other spinal disorders and/or any other serious co-morbidities (e.g., cancer, severe lung pathology); presence of cognitive impairment; inability to perform exercises; having followed a specific training program with a physiotherapist in the previous three months; having been treated with analgesic infiltration in the previous 6 weeks, or failure to follow their 20-treatment schedule exactly. All study participants were volunteers, and all underwent intervention under the Faculty of Medicine and Health Sciences of University Rovira i Virgili.</ns0:p></ns0:div> <ns0:div><ns0:head>Intervention</ns0:head><ns0:p>Both treatment groups (TTEP and SSEP) underwent 20 sessions of treatment at a frequency of 3 to 5 sessions per week <ns0:ref type='bibr' target='#b43'>(Saner et al., 2011)</ns0:ref>, as follows:</ns0:p><ns0:p>In the first 5 sessions, the only treatments were application of an infrared lamp and transcutaneous electrical nerve stimulation (TENS), since these have been proven to reduce pain in both acute and chronic LBP <ns0:ref type='bibr' target='#b3'>(Bertalanffy et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b21'>Jauregui et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b56'>M. Van Tulder et al., 2000)</ns0:ref>. Patients did not perform any exercise in these first sessions, since there is strong evidence that exercise is not effective in relieving acute pain, and can even worsen symptoms (M. <ns0:ref type='bibr' target='#b56'>Van Tulder et al., 2000)</ns0:ref>. TENS therapy was applied with a multichannel portable TENS unit (Megasonic 313 P4, Carin) on the lumbar spine. Biphasic square wave impulses at a frequency of 100 Hz and pulse duration of 70 &#181;s were used for a total duration of 20 minutes. Four rectangular 90 x 45-mm electrodes were applied on the fascia thoracolumbaliis and approximately 10 cm proximal to this, along the midline of the muscle (i.e. directly over the site of pain) <ns0:ref type='bibr' target='#b23'>(Kofotolis et al., 2008)</ns0:ref>.</ns0:p><ns0:p>In sessions 6 through 20, each group engaged in its respective back school exercise program in regular sessions (Figure <ns0:ref type='figure'>1</ns0:ref>); they followed 30-minute protocols of 10 exercises, with 10 repetitions of each. The TTEP group performed exercises from the LBP protocol developed by the Physiotherapy and Rehabilitation Service at Sant Joan University Hospital (Reus, Spain). The SSEP group performed exercises gathered in a search of literature on core stability exercises <ns0:ref type='bibr' target='#b24'>(Koumantakis et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b44'>Shamsi et al., 2015)</ns0:ref>. Before beginning exercises, participants received education on the anatomy of the back (members of the SSEP received a simplified explanation of core musculature), correct posture, and spinal alignment, as part of a back school. The same physiotherapist supervised all classes. Upon treatment completion, he gave each participant a home exercise program which included the exercises from their treatment sessions.</ns0:p></ns0:div> <ns0:div><ns0:head>Outcome measurements</ns0:head><ns0:p>The primary outcome was pain, measured with a 10-cm Visual Analogue Scale (VAS), which has been shown valid and reliable. VAS is a numerical rating scale (0 = no pain to 10 = worst imaginable pain) which represents the intensity of the current pain and allows the evaluator to compare it with previous or later evaluations. <ns0:ref type='bibr' target='#b17'>(Hawker et al., 2011)</ns0:ref> VAS was used to measure pain at baseline session 0 and at sessions 10 and 20. Pain measurement one month after the final treatment session was done by telephone, using the 11-point Numerical Rating Scale (NRS). During this phone conversation, each patient also performed a verbal, subjective assessment. Both these final assessments are considered sufficiently sensitive to detect clinically relevant pain changes <ns0:ref type='bibr' target='#b17'>(Hawker et al., 2011)</ns0:ref>; they are even considered interchangeable for calculating pain in lumbar pathologies <ns0:ref type='bibr' target='#b17'>(Hawker et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b48'>Thong et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b54'>van Tubergen et al., 2002)</ns0:ref>. One secondary outcome was disability, measured using the Roland Morris Disability Questionnaire (RMDQ). The questionnaire asks 24 questions related to the participant's current functional status. Different studies have shown RMDQ to be a useful and reliable instrument for evaluating participants with LBP <ns0:ref type='bibr' target='#b36'>(Payares et al., 2015)</ns0:ref>. RMDQ was used to measure pain at baseline session 0, at sessions 10 and 20, and one month after the final treatment session.</ns0:p><ns0:p>Another secondary outcome was degree of inflammation, as measured by blood-sample levels of the cytokines interleukin 6 (IL-6) and tumor necrosis factor alpha (TNF-&#945;). These markers were used because they have been found to play significant roles in relation to back pain <ns0:ref type='bibr' target='#b8'>(De Queiroz et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kraychete et al., 2010)</ns0:ref>. The presence of some inflammatory mediators might be associated with pain and disability in patients with LBP, since pro-inflammatory cytokines such as IL-6 or TNF-&#945; contribute to the activation of nociceptors that generate potential of action and pain hyper sensibility <ns0:ref type='bibr' target='#b7'>(Cui et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b39'>Queiroz et al., 2015)</ns0:ref>. Degree of inflammation was measured at the beginning and end of the treatment program by enzyme-linked immunosorbent assay (ELISA) method. Blood samples were collected by a qualified doctor (blinded to group allocation) and will be obtained from the antecubital vein <ns0:ref type='bibr' target='#b49'>(Tomazoni et al., 2019)</ns0:ref>.</ns0:p><ns0:p>The study recorded additional factors, including anthropometric characteristics (age, height, weight, body mass index [BMI]), and degree of physical activity using Quick Classifier of Physical Activity (ClassAF) in Metabolic Equivalent of Tasks (METs). ClassAF is a global questionnaire which classifies people as physically active or inactive using a corresponding qualitative formula <ns0:ref type='bibr' target='#b51'>(Vallbona et al., 2007)</ns0:ref>. All these data were collected before the intervention began by the trained physiotherapist.</ns0:p></ns0:div> <ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>Groups were compared with respect to change, from baseline (session 0) to half-intervention (session 10), baseline to post-intervention (session 20), and baseline to 1 month after the intervention concluded; from session 10 to session 20 and session 10 to one month post intervention; and finally from session 20 to one month post intervention.</ns0:p></ns0:div> <ns0:div><ns0:head>SPSS program version 23</ns0:head><ns0:p>Windows was used to analyze the data. A descriptive analysis was made of the study sample, with standard averages, deviations and percentages of the different variables collected. The Kolmog&#243;rov-Smirnov test was applied to assess data distribution in each group. A Student-t test was done to assess differences between the two treatments, and effect size was calculated to measure the magnitude of the experimenter effect, using the standardized mean difference (SMD) for variables normally distributed and the effect size of Mann-Whitney's U test for variables not normally distributed <ns0:ref type='bibr' target='#b9'>(Field, 2005)</ns0:ref>. Two-way repeated ANOVA analyses were used to examine differences over time. Assessments were carried out using non-parametric tests for variables that did not present normal distributions. The level of statistical significance for the study was established at p &lt;0.05.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Flow of participants and therapists through the trial 59 potential participants were referred to the research team. Of those referred, 20 were not included, for various reasons (Figure <ns0:ref type='figure'>2</ns0:ref>). 20 participants were placed in the traditional trunk exercise group (TTEP); the remaining 19 were placed the specific stabilization exercise group (SSEP). Of these 39 participants, 30 completed their course of treatment (15 from each group). Table <ns0:ref type='table'>1</ns0:ref> shows participants' baseline characteristics: age, height, weight, BMI, physical activity level, pain or disability. According to the CONSORT statement, significance testing of baseline differences in randomized controlled trials were not performed <ns0:ref type='bibr' target='#b29'>(Moher et al., 2010)</ns0:ref>.Two members of the SSEP group could not be reached for the one-month follow-up telephone call.</ns0:p><ns0:p>Compliance with the trial method 30 (76,9%) participants attended all 20 intervention sessions. Once the intervention was completed, the physiotherapist advised participants to repeat their exercises at home, three times a week for one month. 15 (50%) participants reported performing the exercises as advised; 9 (30%) reported doing their exercises occasionally; 4 (13.3%) did not perform exercises at home; the remaining 2 (6,67%) were unreachable.</ns0:p></ns0:div> <ns0:div><ns0:head>Effect of intervention</ns0:head><ns0:p>Data on pain and disability are shown in Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>; data on degree of inflammation are in Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_0'>4</ns0:ref>.</ns0:p><ns0:p>Results show an insignificant effect size and no significant differences between groups in terms of current pain intensity, or for any outcome measure. At the end of intervention (session 20), pain intensity for the TTEP group had decreased by 0.33 cm (95% CI -1.7 to 1.0, p=0.615) more than in the SSEP group. Both back school treatments showed positive results for pain reduction from baseline to end of treatment, and baseline to one month post-intervention. In the TTEP group, pain (baseline to final session) reduced by 4.6 cm (95% CI 3.3 to 5.8); the SSEP group's reduction was 4.3 cm (95% CI 3.0 to 5.6).</ns0:p><ns0:p>Similarly, there were an insignificant effect size and no significant differences between groups in terms of change in disability. At post-intervention (session 20), disability levels in the SSEP group had decreased by 0.40 points (95% CI -1.7 to 2.5, p=0.701) more than the TTEP group. Both back school treatments yielded positive results in disability reduction, baseline to end of treatment, and baseline to one month post-treatment. In the TTEP group, RMDQ scores reduced by 5.1 points (95% CI 3.0 to 7.3) from baseline to post-intervention. In the SSEP group, RMDQ reduction for the same interval was 6.1 points (95% CI 3.7 to 8.6).</ns0:p><ns0:p>Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure' target='#fig_0'>4</ns0:ref> show outcomes for inflammation. TNF-&#945; showed higher values for the TTEP group than the SSEP group at the two visits where TNF-&#945; was measured. Significant differences were observed between the groups, baseline and post-treatment. In the first case, a difference of 66.97 pg/mL (95% CI 6.3 to 139.5) was recorded; in the second case the difference was 128.94 pg/mL (95% CI 52.8 to 205.0).</ns0:p><ns0:p>In contrast, IL-6 levels were found to be similar between the two treatment groups, with no significant differences observed. At baseline the difference was 2.51 pg/mL (95% CI -2.3 to 7.3); at post-intervention the difference was 1.25 pg/mL (95% CI -3.9 to 6.4).</ns0:p><ns0:p>In reference to the evolution of inflammatory biomarkers between baseline and post-treatment, the results for participants who practiced traditional TTEP indicate an increase in TNF-&#945; levels of 46.16 pg/mL (95% CI 13.0 to 85.3) and a tendency toward decreased levels of IL-6, 0.19 pg/mL (95% CI -1.6 to 1.2).</ns0:p><ns0:p>In contrast, the results in the group that practiced SSEP are the other way around: there was an increase in IL-6 levels of 1.06 pg/mL (95% CI 0.03 to 2.1) and a tendency toward decrease in TNF-&#945; levels of 12.81 pg/mL (95% CI -42.3 to 16.7).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study hypothesis suggested that treatment with SSEP would be found to decrease pain and disability more effectively than TTEP, in women with non-specific LBP. We found this hypothesis not entirely true -although the effectiveness of SSEP was apparently demonstrated, the effectiveness of TTEP was found to be quite similar in our study group. The literature documents study results confirming those of our own study: Shamsi et al., also concluded that the two types of exercise provide improvement in LBP, but found no evidence as to which type might be more effective <ns0:ref type='bibr' target='#b44'>(Shamsi et al., 2015)</ns0:ref>. The literature also includes meta-analyses comparing back schools for chronic LBP. These found deep-muscle exercises more effective in reducing short-term pain and disability <ns0:ref type='bibr' target='#b5'>(Chang et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b30'>Niederer &amp; Mueller, 2020;</ns0:ref><ns0:ref type='bibr' target='#b62'>Wang et al., 2012)</ns0:ref>, though they found no significant differences in long-term improvement. In our study, however, after the half-way point in treatment (session 10) we observed pain reduction by both modalities. We believe this was most likely due to the fact that participants in the meta-analyzed clinical trials suffered from chronic lower back pain, which has a worse prognosis than non-specific LBP in an early phase <ns0:ref type='bibr' target='#b52'>(Van Den Hoogen et al., 1998)</ns0:ref>. Contrary to our results, in a recent systematic review, a meta-analysis of 8 studies indicated that stabilization exercises were more effective than general exercises in reducing pain. Five studies demonstrated a significant improvement in disability between patients treated with stabilization exercises compared with those treated with general exercises <ns0:ref type='bibr' target='#b14'>(Gomes-Neto et al., 2017)</ns0:ref>. In our case, the SSEP and TTEP seem to be effective in reducing pain and improving disability. The mean of pain in the analyzed studies was 6.01 at baseline, being 2.1 at the end of the stabilization exercises on a 0-10 pain scale <ns0:ref type='bibr' target='#b14'>(Gomes-Neto et al., 2017)</ns0:ref>. The SSEP results of our trial are consistent with these findings: 6.53 at baseline and 2.2 at the end of the intervention. There are authors who have found that stabilizing treatment shows no significant advantage over traditional treatment <ns0:ref type='bibr' target='#b24'>(Koumantakis et al., 2005)</ns0:ref>; some of these authors believe that where there appears to be such an advantage, it is due to certain characteristics of the LBP patients involved, such as segmental instability of the column, or the size of the multifidus muscles. One of the exclusion criteria in our study was diagnosis of other spinal disorders, so our sample was more homogeneous. Our results with regard to inflammation indicate that, following TTEP, TNF-&#945; levels had increased; when SSEP was used, IL-6 levels had increased by the end of our 20-session course of treatment. Al-Obaid et al., recently reported on a study with characteristics similar to ours, which found increased production of pro-inflammatory cytokine TNF-&#945; after treatment, but no change in IL-6 production <ns0:ref type='bibr' target='#b0'>(Al-Obaidi &amp; Mahmoud, 2014)</ns0:ref>. The author justified this result by stating that overexpression of TNF-&#945; and other pro-inflammatory cytokines occurs in many studies of lowback pathologies <ns0:ref type='bibr' target='#b46'>(Takahashi et al., 1996)</ns0:ref>. In addition, he explained, IL-6 cytokine levels are not altered because IL-6 has both pro-inflammatory and anti-inflammatory properties <ns0:ref type='bibr' target='#b32'>(Opal &amp; Depalo, 2000)</ns0:ref>. Various studies claim that IL-6 acts predominantly as an anti-inflammatory cytokine, regulating the synthesis of pro-inflammatory cytokines IL-1 and TNF-&#945; and stimulating the appearance, in circulation, of anti-inflammatory cytokines such as <ns0:ref type='bibr'>IL-10 (Opal &amp; Depalo, 2000;</ns0:ref><ns0:ref type='bibr' target='#b40'>Saavedra Ram&#237;rez et al., 2011)</ns0:ref>. One study goes further <ns0:ref type='bibr' target='#b38'>(Petersen &amp; Pedersen, 2005)</ns0:ref>, claiming that IL-6 stimulates lipolysis and oxidation of fats, as well as producing anti-inflammatory effects during exercise -and therefore may offer protection against TNF-&#945;. Relating this information to our own findings, we could say that treatment with SSEP aims to be more effective because, in our case, TNF-&#945; levels were maintained while IL-6 increased. On the other hand, with TTEP the reverse was true: the cytokine found to have increased in the plasma was TNF-&#945;. We believe this is due to the nature of the exercises. In SSEP, deep muscle exercise is the basis of lumbar and segmental control stabilization. TTEP, on the other hand, focuses on building overall muscle resistance, strength and flexibility, being a more dynamic and intense activity. The literature includes findings that lower-intensity exercises are more effective than those of greater intensity, when it comes to reducing inflammation <ns0:ref type='bibr' target='#b12'>(Ghafourian et al., 2016)</ns0:ref>.</ns0:p><ns0:p>One of our study's limitations is its sample size, but we also prioritized for this pilot trial the homogeneity of our patients through strict inclusion and exclusion criteria, for example we only studied women due their physiological characteristics such as less muscle and bone mass as well as psychological factors <ns0:ref type='bibr' target='#b20'>(Damian Hoy et al., 2012)</ns0:ref>. A study design with larger samples would allow a greater effect size between groups and the creation of subgroups according to age, degree of physical activity, or BMI -facilitating more definitive conclusions regarding these factors. Further, we believe it would be interesting to add another follow-up, beyond this study's onemonth-post-intervention evaluation. Further follow-up (at six months, for example) would reveal any difference between the treatments in terms of long-term clinical improvement, although the results of the current literature suggest that SSEP improves pain and functional status at 3 months but not at 6 or 12 months <ns0:ref type='bibr' target='#b6'>(Coulombe et al., 2017)</ns0:ref>. In summary, this study suggests that any type of back school exercise is highly effective in reducing pain and reducing disability in women with non-specific LBP. Further, it showed that SSEP seems to have an anti-inflammatory effect in such patients, potentially offering protection against chronic diseases associated with low-grade inflammation <ns0:ref type='bibr' target='#b38'>(Petersen &amp; Pedersen, 2005)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This study adds to the literature the finding that both back school exercise program are apparently effective and equivalent in reducing pain and improving disability in women with non-specific LBP, from the tenth treatment session to one month after intervention. Moreover, it demonstrates the influence of each back school in the degree of inflammation, concluding that SSEP seems to increase production of anti-inflammatory biomarkers, while TTEP increases proinflammatory biomarker production. A large, adequately powered study is recommended to determine if the results from this pilot study can be duplicated. Manuscript to be reviewed Mean (SD) for outcomes reported at all study visits for total and each group, significant differences between visits within groups, p values and mean (95% CI) difference effect size between groups for pain intensity and disability </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>TTEP= traditional trunk exercise program, SSEP= specific stabilization exercise program, VAS: visual analogue scale, NRS: numerical rating scale, RMDQ= Roland Morris disability questionnaire, p&lt;0.05= significant difference between groups,</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> </ns0:body> "

"Dear Editors We thank the reviewers for their generous comments on the manuscript and have edited the manuscript to address their concerns. In particular all of the code we wrote is available and I have included multiple links throughout the paper to the appropriate code repositories. We believe that the manuscript is now suitable for publication in PeerJ. Dr. Eduard Minobes-Molina Department of Basic Health Sciences director Faculty of Health Sciences and Welfare University of Vic-Central University of Catalonia (UVic-UCC) On behalf of all authors Editor comments (Justin Keogh) The three reviewers and I have identified a number of important issues that the authors need to address before this manuscript can be more strongly considered for publication in PeerJ. Therefore, please make sure you look to address all of the comments from the three reviewers in your revised manuscript. In particular, due to the small sample size, it is suggested that the authors rephrase this as a pilot/feasibility study. Response: We thank the editor and the reviewers for their generous comments on the manuscript and have edited the manuscript to address their concerns. We want to highlight that following your advice, we have a considered our study as a pilot randomized controlled trial. Reviewer 1 Specific comments 1- Materials & Methods: The authors may need explain the DETAILS of the randomization method (e.g., allocation concealment mechanism, implementation, and blinding). Response: We thank the reviewer for this helpful suggestion and have explained the details on the manuscript into the “Study design” section. 2- The authors used p-values to assess balance in baseline characteristics between two groups. There are many published references that avoid researchers to use of statistical testing for baseline comparison (de Boer et al., ‎2015). P-values are affected by sample size. If the sample size is large enough, even very small differences may be statistically significant. On the other hand, even large differences may lead to non-significant results if the sample is too small (less than 20) (du Prel et al., 2009). For quantitative variables, other approaches such as difference of <0.25 standard deviation (SD) in baseline characteristics can be used. Response: We thank the reviewer for this suggestion and have avoided to use of statistical testing for baseline comparison (we have removed p-values in Table 1). De Boer et al., discussed that testing for baseline differences serves no purpose and can be misleading, especially because some researchers still think that these tests are the basis for choosing covariates in their analyses. This practice however ignores the prognostic strength of covariates, which is a more important characteristic to take into account, because adjustment for prognostic factors can also increase precision of effect estimates. Finally, based on these arguments, they proposed that journals adopt the CONSORT 2010 statement on this topic by not publishing these tests anymore (de Boer MR, Waterlander WE, Kuijper LD, Steenhuis IH, Twisk JW. Testing for baseline differences in randomized controlled trials: an unhealthy research behavior that is hard to eradicate. Int J Behav Nutr Phys Act. 2015;12:4. Published 2015 Jan 24. doi:10.1186/s12966-015-0162-z). 3- The authors may need to calculate effect size measures (e.g., the standardized mean difference [SMD]) with associate confidence intervals in order to compare between two groups. The major benefit of determining effect size is that unlike test statistics, the effect size is not greatly influenced by sample size, thus reducing problems of power associated with large and small samples. Response: We thank the reviewer for his insightful comments and have incorporated the standardized mean difference (SMD) with associate confidence intervals in order to compare between two groups in the data analysis. 4- For variables that were not normally distributed, the authors can use transformation to calculate SMD. Response: As the reviewer advises we have incorporated the SMD for variables normally distributed and the Mann-Whitney's test for variables not normally distributed in the data analysis. 5- The authors may need to present the significance level sign (*) in figures 2 and 3 for within-group comparison. Response: We thank the reviewer for this suggested addition. However, we want to highlight the different kind of significant differences of the following way: A= Significant difference between baseline and post-intervention in TTEP group, B= Significant difference between baseline and post-intervention in SSEP group, b= Significant difference between post-intervention groups. 6- Is was not clear whether confounding variables (e.g., Smoke and depression) were controlled in patients with NLBP. Response: We thank the reviewer for this question. We controlled some confounding factors using the inclusion and exclusion criteria and not as a variable due the intervention was short (20 sessions). Participants entering the trial were required to be under no pharmacological treatment for pain. Besides other exclusion criteria were: diagnosed with other spinal disorders and/or any other serious co-morbidities (e.g., cancer, severe lung pathology); presence of cognitive impairment; having followed a specific training program with a physiotherapist in the previous three months and having been treated with analgesic infiltration in the previous 6 weeks. All of them could be confounding factors. 7- It is highly recommended that the authors compare the effect of intervention on the outcome measures with their MCID values to assess whether the effect is substantially beneficial. Response: We thank the reviewer for raising this point. The minimal clinically important difference score (MCID) is an interesting concept. However, there is no standard method on how to calculate it, and this has led to or resulted in a number of methodological or interpretation problems. There also problems in defining a MCID due to issues such as patients’ inability to understand the context of improvement. MCID usually considers patient and/or clinician perception/perspective and this information was not collected in our study (Cook CE. Clinimetrics Corner: The Minimal Clinically Important Change Score (MCID): A Necessary Pretense. J Man Manip Ther. 2008;16(4):E82-E83. doi:10.1179/jmt.2008.16.4.82E; Wells G, Beaton D, Shea B, et al. Minimal clinically important differences: review of methods. J Rheumatol. 2001;28(2):406-412. ). Therefore, calculating and informing MCIDs is not possible at this stage. 8- In addition to statistical tests, the authors may need to assess the normal distribution with histograms. Response: We thank the reviewer for this suggested addition. Nevertheless, we followed the CONSORT Statement. In this standard way for authors to prepare reports of trial findings, histograms with normal distribution are not usual. Within the manuscript we have already added 4 figures and 2 tables and we believe that adding the histograms would be having too many figures. Statistical variables that did not follow normal distribution are indicated in the tables according to the parametric or non-parametric tests used. In any case, we can send the histograms to you if you require them. 9- The authors mentioned that “One of our study’s limitations is its sample size…” Why did the authors state this sentence whereas they calculated the sample size? Response: Finally, following the advice of the reviewers, we have a considered our study as a pilot randomized controlled trial. As we said into the “Discussion”, larger samples would allow a greater effect size between groups and the creation of subgroups according to age, degree of physical activity, or BMI – facilitating more definitive conclusions regarding these factors. Probably the writing of this section generates confusion and we have considered to detail it better. 10- The authors may need to report of the mode of TENS (conventional, low rate, etc.) with its details (i.e., frequency, electrode placement, pulse width, time, etc). Response: We thank the reviewer for this suggested addition and can confirm that the information has been added to the paper. 11- How did the physiotherapist progress the exercise program? Response: We thank the reviewer for this observation and have added the Figure 4 (as is suggested in the comment 12) where the exercise progress can be observed. 12- The figures of exercises should be presented. Response: We thank the reviewer for this observation and have added the Figure 4 accordingly. 13- Did the participants use analgesics during the treatment period? If yes, how did the authors manage this issue? Response: As we explain into the inclusion criteria, participants entering the trial were required to be under no pharmacological treatment for pain. Reviewer 2 Basic reporting The english writing is acceptable The literature review content is very cursory around the issue of comparing different exercise types, as the evidence is overwhelming that no one type of exercise is superior to another (incl stabilization exercise). There is next to no content to justify the inflammatory measures used and they should be removed moving forward. This article should be scoped as a feasibility study and described as such. Response: We thank the reviewer for these observations. We have justified the inflammatory measures accordingly into the “Outcome measurements” section. On the other hand, our work cannot be classified as a feasibility study. A feasibility study asks whether something can be done, should we proceed with it, and if so, how. Therefore, this type of studies should include information such as willingness of participants to be randomized, follow-up rates, adherence/compliance rates, time needed to collect data, etc. Our study did not have the main objective of determining if the intervention, study design and procedures could be successfully executed. In this sense, Back School exercises are not a novel intervention and their application has been done in several trials for the last decades. Unlike what was conducted in our work, feasibility studies often include qualitative methods to assess the perspectives of the participants (Eldridge SM, Lancaster GA, Campbell MJ, et al. Defining Feasibility and Pilot Studies in Preparation for Randomised Controlled Trials: Development of a Conceptual Framework. PLoS One. 2016;11(3):e0150205. Published 2016 Mar 15. doi:10.1371/journal.pone.0150205; Feeley N, Cossette S, Côté J, et al. The importance of piloting an RCT intervention. Can J Nurs Res. 2009;41(2):85-99.). However, as the editor and other reviewers suggested, we have considered this article as a pilot study, considering that a large, adequately powered study is recommended to determine if the results from this pilot study can be duplicated. Experimental design The risk of bias is unacceptable for this study with the single physiotherapist conducting both interventions, thus we have no faith either intervention is valid. The sample size estimation to justify such a small sample is extremely flawed, and scoping the study as a feasibility study is best. However the single physiotherapist as the treating professional is unacceptable to have any confidence in the validity of the intervention. Validity of the findings This section is a fail as per the above critical error in study design. Comments for the Author This could be reframed as a clinical feasibility study, but the issue of a single clinician providing the two fundamentally different types of exercise is a major bias that cannot be corrected or avoided here. There can be no confidence in the validity of the interventions or the data. Response: We thank the reviewer for this helpful comment and have added the details of the randomization method into the “Study design” section. As you say, a single physiotherapist conducted both interventions and we will consider to change it for future trials, but we followed some strategies for blinding key players for the trial that the literature explains (Page SJ, Persch AC. Recruitment, retention, and blinding in clinical trials. In: American Journal of Occupational Therapy. ; 2013. doi:10.5014/ajot.2013.006197): the health care provider was blinded to design, hypotheses, eligibility criteria, and outcomes, he was trained according to manual of procedures and implemented fidelity measures. Besides the data collector was blinded to the main information and he used objective outcome measures (e.g., VAS or RMDQ). We used similar strategies with the other key players as the participants and the principal investigator. Reviewer 3 Basic reporting The article in general uses clear, concise, unambiguous and technically correct text. The introduction and background sufficiently describes the impact of LBP. Relevant prior literature is appropriately referenced. The article structure, figures and tables are all of professional level, however some specific comments for the Results, Discussion and figures sections are provided below. Raw data were not shared. Response: We thank the reviewer for the specific comments. In reference of the “raw data” file, it was shared as an additional document during the submission of the manuscript. On the other hand, the datasets generated and analyzed during the current study will be available from the corresponding author on reasonable request. The structure of the article conforms to an acceptable format of ‘standard sections’. The article is generally self-contained with relevant results to hypotheses. However, some issues require a clearer explanation and should be added. Experimental design Original primary research with research question well defined, relevant & meaningful, stating how research fills an identified knowledge gap. It is interesting that the authors wished to explore inflammatory markers in LBP. A well-conducted investigation performed to a sufficient technical & ethical standard. Methods not described with sufficient detail & information to replicate in some respect, and this is pointed out to the authors. Response: We thank the reviewer for this helpful suggestion and we have explained the details into the “Materials & Methods” section (e.g., randomization method, TENS parameters or the Figure 4 with the exercises details). Validity of the findings The validity of the findings is not sound, as results are representative of a small group of LBP participants and only female. Therefore this study can be characterized as a pilot study. Response: We thank the reviewer for this suggested advice and can confirm that this study has been characterized as a pilot study. Pilot studies are often conducted when the intervention is novel and the sample size calculation is limited due to the scarce literature (Lancaster GA, Dodd S, Williamson PR. Design and analysis of pilot studies: recommendations for good practice. J Eval Clin Pract. 2004;10(2):307-312. doi:10.1111/j..2002.384.doc.x). We only studied women due their higher prevalence because their physiological characteristics such as less muscle and bone mass as well as psychological factors. In this case, future powered studies are required to confirm the results of our pilot study. Conclusions are adequately stated, linked to original research question & limited to supporting results. Comments for the Author 1. In my opinion, TNF-a and IL-6 values in the Abstract should be made for both groups. Please amend. Response: We thank the reviewer for this comment and have amended the Abstract results section. 2. Follow up of 1 month post-exercise is a rather limited timeframe. However, within that limited time, pain and disability levels in both groups have increased. Possible implications of a non-long-lasting effect of both exercise programs on pain & disability levels should be commented on by the authors. Response: We thank the reviewer for his insightful comments and have incorporated the suggestions into the discussion. 3. The authors seem to over-discuss the significance of their secondary measures (not the Disability one) over their primary one. However, for such a small study, and probably underpowered for those particular measures, such discussion is somewhat redundant. For instance, TNF-a unequal baselines, is probably due to small sample size. On the other hand, IL-6 baseline levels were equal between-groups. Which renders this latter measure the one that confers an equal starting point. For TNF-a levels, an ANCOVA analysis might have been better suited than an ANOVA. Response: We have developed the discussion of our main results comparing with a current systematic review. We agree with the reviewer that secondary measures are widely discussed; the reason is that we think that this study adds to the literature the influence of each back school in the grade of inflammation, apparently the SSEP increases the anti-inflammatory biomarkers production and TTEP increases the pro-inflammatory biomarkers production. Therefore we believe that these results should be highlighted and promote future research. 4. Why did TNF-a values increase in the TTE group to the authors opinion? If this is due to exercise intensity of the TTE, some more information on the particular exercises used should be given. However, it seems that there was no impact of TNF-a levels in either Pain or Disability levels. Response: We thank the reviewer for this question. As we explain into the discussion we believe this is due to the nature of the exercises. In SSEP, deep muscle exercise is the basis of lumbar and segmental control stabilization. TTEP, on the other hand, focuses on building overall muscle resistance, strength and flexibility, being a more dynamic and intense activity. The literature includes findings that lower-intensity exercises are more effective than those of greater intensity, when it comes to reducing inflammation (Ghafourian M, Ashtary-Larky D, Chinipardaz R, Eskandary N, Mehavaran M. Inflammatory Biomarkers’ Response to Two Different Intensities of a Single Bout Exercise Among Soccer Players. Iran Red Crescent Med J. Published online 2016. doi:10.5812/ircmj.21498). We have added the Figure 4 where the exercise details can be observed. 5. Lines 81: Correct typo. Response: We thank the reviewer for this observation and have amended the typo accordingly. 6. Line 121 & 133 & 138: Why are the 2 programs referred to as ‘back schools’? Please provide reasoning and reference(s). In no other spinal stabilization study, however, have they been referred as such. Response: Back School is one treatment that provides both exercise and education for the treatment of people with LBP. It was designed to reduce pain and prevent recurrences of LBP episodes (Forssell MZ. The Swedish Back School. Physiotherapy. 1980;66(4):112-114.; Forssell MZ. The back school. Spine (Phila Pa 1976). 1981;6(1):104-106. doi:10.1097/00007632-198101000-00022). Back School is a therapeutic program including information on the anatomy of the back, biomechanics, optimal posture, ergonomics, and back exercises (Parreira P, Heymans MW, van Tulder MW, et al. DOES NOT RE-ESTABLISH WALKING IN NON-WALKING SUBJECTS WITH. Cochrane database Syst Rev. Published online 2016. doi:10.2340/16501977-2508). In our intervention, participants received education on the anatomy of the back (members of the SSEP received a simplified explanation of core musculature), correct posture, and spinal alignment. It is detailed into the “Intervention” section and in the Figure 1. 7. Line 135: The study does not seem to be blind, as the authors imply. Please explain if otherwise. How were the purposes of the study conveyed to the participants, for instance. Response: We thank the reviewer for this comment and we have explained in more detail the strategies for blinding key players that we have used into the “Study design” section. 8. Lines 159-161: The authors contend that initial treatment was not in the form of exercise, as the participants were in the acute stage of symptoms. However, from the inclusion criteria this cannot be surmised. Response: We thank the reviewer for this comment and have adjusted the inclusion criteria to ensure this point is clearer. 9. Line 180 & 181: The NRS used in the follow up is not quite the same as the VAS. It does not account for in-between decimal values, usually. For example, a patient might have indicated a pain level of 4 in the NRS, whereas in the VAS, a value of 3.6 might have been indicated. I understand this was the best available option over the telephone, but it is not correct to change the pain scale administration format as a trial progresses. Response: We thank the reviewer for his insightful comments and we thought about the question. Finally, we chose to use this option during the follow-up (by telephone) because we found in the literature a very strong association between the NRS and VAS (r=0.93), indicating that they measure essentially the same thing (Thong ISK, Jensen MP, Miró J, Tan G. The validity of pain intensity measures: What do the NRS, VAS, VRS, and FPS-R measure? Scand J Pain. Published online 2018. doi:10.1515/sjpain-2018-0012). 10. From lines 209-213 it is gathered that the study was not adequately powered, as a between-group difference in the VAS scale was not reached in either of the 3 follow up time points. Response: Results show no significant differences between groups in terms of current pain intensity, for any of the 3 follow up time points. Into the “Discussion” section we comment that the literature confirms our results: “the two types of exercise provide improvement in LBP, but found no evidence as to which type might be more effective” (Shamsi M, Sarrafzadeh J, Jamshidi A. Comparing core stability and traditional trunk exercise on chronic low back pain patients using three functional lumbopelvic stability tests. Physiother Theory Pract. Published online 2015. doi:10.3109/09593985.2014.959144). In any case, we have considered this article as a pilot study, considering that a large, adequately powered study is recommended to determine if the results from this pilot study can be duplicated. 11. Lines 218-9: The term ‘Repeated ANOVA analyses’ does not specify which particular ANOVA type was used. A 2 x 4 perhaps for pain & disability? Response: We used a two-way repeated measures ANOVA to determine whether any change in pain is the result of the interaction between the 'type of treatment' (SSEP or TTEP, which is one of our two factors) and 'time' (our second factor). We used the same for disability. 12. Line 237: From figure 1 it can be seen that 5 & 4 participants left the TTE & SSE Groups respectively, therefore all participants did not complete the 20 sessions as mentioned in that particular text line. In fact, in line 228, this is clearly stated. Response: We thank the reviewer for this comment and have amended the results accordingly. We wanted to explain that all participants had to attended all 20 intervention sessions. Otherwise, participants were excluded from the study. 13. Line 294: One can have non-specific chronic LBP, also. Response: We thank the reviewer for raising this point. It’s linked to your comment 8. As we said before, we have clarified the inclusion criteria detailing them and all the participants had fewer than 6 weeks of pain duration, therefore we didn’t include participants suffering non-specific chronic LBP. 14. Figure 2 & 3: SD depiction seems not symmetrical around the mean value. Please amend or explain why. Response: We thank the reviewer for this comment and have amended the Figures 2 and 3 accordingly. 15. Significant differences in Figure 3 are not depicted within the graph (by letter b, as indicated in Figure 2). Response: We thank the reviewer for this comment and have amended the Figures 2 and 3 accordingly. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Non-specific low back pain (LBP) is the leading cause of disability worldwide. The primary physiotherapeutic treatment for LBP is physical exercise, but evidence suggesting a specific exercise as most appropriate for any given case is limited.</ns0:p><ns0:p>Objective. To determine if specific stabilization exercise (SSE) is more effective than traditional trunk exercise (TTE) in reducing levels of pain, disability and inflammation in women with non-specific low back pain (LBP).</ns0:p></ns0:div> <ns0:div><ns0:head>Design.</ns0:head><ns0:p>A pilot randomized controlled trial was conducted in Rovira i Virgili University, Catalonia.</ns0:p><ns0:p>Methods. 39 females experiencing non-specific LBP were included in two groups: the TTE program and SSE program, both were conducted by a physiotherapist during twenty sessions. The primary outcome was pain intensity (10-cm Visual Analogue Scale). Secondary outcomes were disability (Roland Morris Disability Questionnaire), and inflammation (IL-6 and TNF-&#945; plasma levels). Measurements were taken at baseline, at half intervention, at post-intervention, and a month later.</ns0:p><ns0:p>Results: Mean group differences in change from baseline to post-intervention for TTE were: -4.5 points (CI 3.3 to 5.6) for pain, -5.1 points (CI 3.0 to 7.3) for disability, 0.19 pg/mL (95% CI -1.6 to 1.2) for IL-6 levels, and 46.2 pg/mL (CI 13.0 to 85.3) for TNF-&#945; levels. For SSE, differences were: -4.3 points (CI 3.1 to 5.6) for pain, -6.1 points (CI 3.7 to 8.6) for disability, 1.1 pg/mL (CI 0.0 to 2.1) for IL-6 levels , and 12.8 pg/mL (95% CI -42.3 to 16.7) for TNF-&#945; levels. There were an insignificant effect size and no statistically significant overall mean differences between both groups.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion:</ns0:head><ns0:p>This study suggests that both interventions (traditional trunk and specific stabilization exercises) are effective in reducing pain and disability in non-specific LBP patients, but the two programs produce different degrees of inflammation change.</ns0:p><ns0:p>Clinical trial registration number: NCT02103036.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Common low back pain (LBP) is defined as pain between the costal margins and the inferior gluteal folds, usually accompanied by painful limitation of movement, often influenced by physical activities and posture, and which may be associated with referred pain in the leg <ns0:ref type='bibr' target='#b25'>(Kovacs et al., 2006)</ns0:ref>. LBP represents an important public health problem, because worldwide prevalence of the condition ranges from 12% to 33% <ns0:ref type='bibr' target='#b59'>(Walker, 2000)</ns0:ref>. It is also known that LBP is more prevalent in females than males; for example, in 2015 in Catalonia, 30.1% of females suffered LBP, as compared to 18.7% of males <ns0:ref type='bibr' target='#b10'>(Garcia et al., 2016)</ns0:ref>. LBP remains a common disabling condition <ns0:ref type='bibr' target='#b60'>(Walker et al., 2004)</ns0:ref> and is associated with high costs for medical health and social care <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b64'>Wieser et al., 2011)</ns0:ref>.</ns0:p><ns0:p>One of the difficulties in reducing the burden of spinal disorders is the wide and heterogeneous range of specific diseases and non-specific musculoskeletal disorders that can involve the spinal column, most of which manifest pain <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012)</ns0:ref>. Despite this factor, or perhaps because of its impact on individuals, their families, and the healthcare systems, spinal disorders remain one of the most controversial and challenging conditions <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b18'>D. Hoy et al., 2010)</ns0:ref>.</ns0:p><ns0:p>Back pain is sometimes associated with a likely aetiology (e.g.radiculopathy or spinal stenosis), but most LBP cases are of unknown origin and are classified as non-specific, which has also been described as mechanical pain or strain, account for 90% or more of all people experiencing spinal pain <ns0:ref type='bibr' target='#b15'>(Haldeman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b57'>M. W. Van Tulder et al., 1997)</ns0:ref>. For all these reasons, it is necessary to find an effective treatment for LBP. Unfortunately, the scientific literature does not offer relevant conclusions, due in part to the poor methodology employed in many published studies -e.g., short follow-up periods, population heterogeneity and non-validated measurements <ns0:ref type='bibr' target='#b1'>(Atlas &amp; Nardin, 2003)</ns0:ref>.</ns0:p><ns0:p>According to current clinical reviews and guides, first-line treatments for LBP pathology focus on analgesic measures <ns0:ref type='bibr' target='#b22'>(Koes et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b55'>M. Van Tulder et al., 2006)</ns0:ref>. Conservative physiotherapeutic treatments for LBP also exist <ns0:ref type='bibr' target='#b4'>(Bordas et al., 2004)</ns0:ref>, including advice and postural education, electrotherapy, manual therapy, and physical exercises <ns0:ref type='bibr' target='#b66'>(Williams et al., 2014)</ns0:ref>. Exercise is one of the chief recommendations for pain reduction, mobility increase, improvement of physical and psychological abilities and anxiety reduction <ns0:ref type='bibr' target='#b58'>(Waddell &amp; Burton, 2005)</ns0:ref>. The problem with exercise programs lies in the fact that rationales for choosing the appropriate exercise for an individual case are very weak. Controversy arises because the types of exercise programs for LBP vary considerably, as do the types of patients. This makes it very unlikely that a particular program will be equally effective in all cases <ns0:ref type='bibr' target='#b28'>(Macedo et al., 2008)</ns0:ref>.</ns0:p><ns0:p>One exercise option is the back school program, a therapeutic program including information on the anatomy of the back, biomechanics, optimal posture, ergonomics, and back exercises <ns0:ref type='bibr' target='#b35'>(Parreira et al., 2016)</ns0:ref>, which has proven effective in reducing pain and disability <ns0:ref type='bibr' target='#b42'>(Sahin et al., 2011)</ns0:ref>. There are many types of back school exercise programs, but considerable discussion has centered on the question of whether the specific stabilization exercise program (SSEP) -in which the deep muscles are the protagonists -is preferable to the traditional trunk exercise program (TTEP), which includes more exercises for strengthening abdominal and back muscles. There are systematic reviews that support the idea that SSEP is superior to TTEP <ns0:ref type='bibr' target='#b27'>(Lederman, 2010)</ns0:ref>, but there are also studies that have found both approaches equally effective for improvement of LBP in terms of pain and disability <ns0:ref type='bibr' target='#b44'>(Shamsi et al., 2015)</ns0:ref>.</ns0:p><ns0:p>On the other hand, pro-inflammatory cytokines levels were detected on assessing local tissue in adults with LBP <ns0:ref type='bibr' target='#b39'>(Queiroz et al., 2015)</ns0:ref> and evidence shows that physical exercise therapy decreases systemic inflammatory mediators production, this demonstrates its clinical relevance <ns0:ref type='bibr' target='#b37'>(Pereira et al., 2013)</ns0:ref>.</ns0:p><ns0:p>After years of research into LBP treatment, and taking into account the variety of treatment options, exercise continues to be accepted as an effective approach. The question remains, however, of which type of exercise is most effective in treating various patient subgroups <ns0:ref type='bibr' target='#b1'>(Atlas &amp; Nardin, 2003;</ns0:ref><ns0:ref type='bibr' target='#b43'>Saner et al., 2011)</ns0:ref>. There is limited evidence that the specific stabilization exercise program is more effective than the traditional trunk exercise for patients with nonspecific LBP. Therefore, the following were the research questions this study sought to answer:</ns0:p><ns0:p>1.</ns0:p><ns0:p>Is the specific stabilization exercise program more effective than the traditional trunk exercise program in reducing levels of pain and disability in women with non-specific LBP? 2.</ns0:p><ns0:p>Which type of back school exercise produces different degrees of inflammation change in women with non-specific LBP?</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Ethics</ns0:head><ns0:p>The Clinical Ethics Committee of University Hospital Sant Joan of Reus approved this study (12-06-28/6proj4). All participants gave written informed consent before data collection began.</ns0:p></ns0:div> <ns0:div><ns0:head>Study design</ns0:head><ns0:p>A pilot randomized trial was conducted in Catalonia from February 2013 to February 2015 (NCT02103036). Participants, diagnosed by medical practitioners and referred for treatment of non-specific LBP, were randomized using computer-generated random number tables into two treatment groups: a TTEP group and a SSEP group (Figure <ns0:ref type='figure'>1</ns0:ref>). Afterwards, measurements were taken at baseline (session 0), at half intervention (session 10), post-intervention (session 20) and one month later. A single-blind study was conducted, due to the impossibility of achieving double blindness -the physiotherapist performing the intervention had to know which treatment each participant was to receive. The therapist who performed the intervention was a qualified health professional with 5 years' experience in the field; a different professional took all the study measurements, he was blinded to the participant's assignment. Participants were blinded to their group allocation, design and hypotheses <ns0:ref type='bibr' target='#b33'>(Page &amp; Persch, 2013)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Participants</ns0:head><ns0:p>Participants entering the trial were required to meet the following inclusion criteria: females aged between 18 and 70 years; diagnosed with non-specific LBP (fewer than 6 weeks of pain duration) by a specialist doctor who used imaging such as magnetic resonance, radiographic or computed axial tomography to rule out other spinal disorders, and under no pharmacological treatment for pain. Exclusion criteria were: diagnosed with other spinal disorders and/or any other serious co-morbidities (e.g., cancer, severe lung pathology); presence of cognitive impairment; inability to perform exercises; having followed a specific training program with a physiotherapist in the previous three months; having been treated with analgesic infiltration in the previous 6 weeks, or failure to follow their 20-treatment schedule exactly. All study participants were volunteers, and all underwent intervention under the Faculty of Medicine and Health Sciences of University Rovira i Virgili.</ns0:p></ns0:div> <ns0:div><ns0:head>Intervention</ns0:head><ns0:p>Both treatment groups (TTEP and SSEP) underwent 20 sessions of treatment at a frequency of 3 to 5 sessions per week <ns0:ref type='bibr' target='#b43'>(Saner et al., 2011)</ns0:ref>, as follows:</ns0:p><ns0:p>In the first 5 sessions, the only treatments were application of an infrared lamp and transcutaneous electrical nerve stimulation (TENS), since these have been proven to reduce pain in both acute and chronic LBP <ns0:ref type='bibr' target='#b3'>(Bertalanffy et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b21'>Jauregui et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b56'>M. Van Tulder et al., 2000)</ns0:ref>. Patients did not perform any exercise in these first sessions, since there is strong evidence that exercise is not effective in relieving acute pain, and can even worsen symptoms (M. <ns0:ref type='bibr' target='#b56'>Van Tulder et al., 2000)</ns0:ref>. TENS therapy was applied with a multichannel portable TENS unit (Megasonic 313 P4, Carin) on the lumbar spine. Biphasic square wave impulses at a frequency of 100 Hz and pulse duration of 70 &#181;s were used for a total duration of 20 minutes. Four rectangular 90 x 45-mm electrodes were applied on the fascia thoracolumbaliis and approximately 10 cm proximal to this, along the midline of the muscle (i.e. directly over the site of pain) <ns0:ref type='bibr' target='#b23'>(Kofotolis et al., 2008)</ns0:ref>.</ns0:p><ns0:p>In sessions 6 through 20, each group engaged in its respective back school exercise program in regular sessions (Figure <ns0:ref type='figure'>2</ns0:ref>); they followed 30-minute protocols of 10 exercises, with 10 repetitions of each. The TTEP group performed exercises from the LBP protocol developed by the Physiotherapy and Rehabilitation Service at Sant Joan University Hospital (Reus, Spain). The SSEP group performed exercises gathered in a search of literature on core stability exercises <ns0:ref type='bibr' target='#b24'>(Koumantakis et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b44'>Shamsi et al., 2015)</ns0:ref>. Before beginning exercises, participants received education on the anatomy of the back (members of the SSEP received a simplified explanation of core musculature), correct posture, and spinal alignment, as part of a back school. The same physiotherapist supervised all classes. Upon treatment completion, he gave each participant a home exercise program which included the exercises from their treatment sessions.</ns0:p></ns0:div> <ns0:div><ns0:head>Outcome measurements</ns0:head><ns0:p>The primary outcome was pain, measured with a 10-cm Visual Analogue Scale (VAS), which has been shown valid and reliable. VAS is a numerical rating scale (0 = no pain to 10 = worst imaginable pain) which represents the intensity of the current pain and allows the evaluator to compare it with previous or later evaluations. <ns0:ref type='bibr' target='#b17'>(Hawker et al., 2011)</ns0:ref> VAS was used to measure pain at baseline session 0 and at sessions 10 and 20. Pain measurement one month after the final treatment session was done by telephone, using the 11-point Numerical Rating Scale (NRS). During this phone conversation, each patient also performed a verbal, subjective assessment. Both these final assessments are considered sufficiently sensitive to detect clinically relevant pain changes <ns0:ref type='bibr' target='#b17'>(Hawker et al., 2011)</ns0:ref>; they are even considered interchangeable for calculating pain in lumbar pathologies <ns0:ref type='bibr' target='#b17'>(Hawker et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b48'>Thong et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b54'>van Tubergen et al., 2002)</ns0:ref>. One secondary outcome was disability, measured using the Roland Morris Disability Questionnaire (RMDQ). The questionnaire asks 24 questions related to the participant's current functional status. Different studies have shown RMDQ to be a useful and reliable instrument for evaluating participants with LBP <ns0:ref type='bibr' target='#b36'>(Payares et al., 2015)</ns0:ref>. RMDQ was used to measure pain at baseline session 0, at sessions 10 and 20, and one month after the final treatment session.</ns0:p><ns0:p>Another secondary outcome was degree of inflammation, as measured by blood-sample levels of the cytokines interleukin 6 (IL-6) and tumor necrosis factor alpha (TNF-&#945;). These markers were used because they have been found to play significant roles in relation to back pain <ns0:ref type='bibr' target='#b8'>(De Queiroz et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kraychete et al., 2010)</ns0:ref>. The presence of some inflammatory mediators might be associated with pain and disability in patients with LBP, since pro-inflammatory cytokines such as IL-6 or TNF-&#945; contribute to the activation of nociceptors that generate potential of action and pain hyper sensibility <ns0:ref type='bibr' target='#b7'>(Cui et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b39'>Queiroz et al., 2015)</ns0:ref>. Degree of inflammation was measured at the beginning and end of the treatment program by enzyme-linked immunosorbent assay (ELISA) method. Blood samples were collected by a qualified doctor (blinded to group allocation) and will be obtained from the antecubital vein <ns0:ref type='bibr' target='#b49'>(Tomazoni et al., 2019)</ns0:ref>.</ns0:p><ns0:p>The study recorded additional factors, including anthropometric characteristics (age, height, weight, body mass index [BMI]), and degree of physical activity using Quick Classifier of Physical Activity (ClassAF) in Metabolic Equivalent of Tasks (METs). ClassAF is a global questionnaire which classifies people as physically active or inactive using a corresponding qualitative formula <ns0:ref type='bibr' target='#b51'>(Vallbona et al., 2007)</ns0:ref>. All these data were collected before the intervention began by the trained physiotherapist.</ns0:p></ns0:div> <ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>Groups were compared with respect to change, from baseline (session 0) to half-intervention (session 10), baseline to post-intervention (session 20), and baseline to 1 month after the intervention concluded; from session 10 to session 20 and session 10 to one month post intervention; and finally from session 20 to one month post intervention.</ns0:p></ns0:div> <ns0:div><ns0:head>SPSS program version 23</ns0:head><ns0:p>Windows was used to analyze the data. A descriptive analysis was made of the study sample, with standard averages, deviations and percentages of the different variables collected. The Kolmog&#243;rov-Smirnov test was applied to assess data distribution in each group. A Student-t test was done to assess differences between the two treatments, and effect size was calculated to measure the magnitude of the experimenter effect, using the standardized mean difference (SMD) for variables normally distributed and the effect size of Mann-Whitney's U test for variables not normally distributed <ns0:ref type='bibr' target='#b9'>(Field, 2005)</ns0:ref>. Two-way repeated ANOVA analyses were used to examine differences over time. Assessments were carried out using non-parametric tests for variables that did not present normal distributions. The level of statistical significance for the study was established at p &lt;0.05.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Flow of participants and therapists through the trial 59 potential participants were referred to the research team. Of those referred, 20 were not included, for various reasons (Figure <ns0:ref type='figure'>1</ns0:ref>). 20 participants were placed in the traditional trunk exercise group (TTEP); the remaining 19 were placed the specific stabilization exercise group (SSEP). Of these 39 participants, 30 completed their course of treatment (15 from each group). Table <ns0:ref type='table'>1</ns0:ref> shows participants' baseline characteristics: age, height, weight, BMI, physical activity level, pain or disability. According to the CONSORT statement, significance testing of baseline differences in randomized controlled trials were not performed <ns0:ref type='bibr' target='#b29'>(Moher et al., 2010)</ns0:ref>.Two members of the SSEP group could not be reached for the one-month follow-up telephone call.</ns0:p><ns0:p>Compliance with the trial method 30 (76,9%) participants attended all 20 intervention sessions. Once the intervention was completed, the physiotherapist advised participants to repeat their exercises at home, three times a week for one month. 15 (50%) participants reported performing the exercises as advised; 9 (30%) reported doing their exercises occasionally; 4 (13.3%) did not perform exercises at home; the remaining 2 (6,67%) were unreachable.</ns0:p></ns0:div> <ns0:div><ns0:head>Effect of intervention</ns0:head><ns0:p>Data on pain and disability are shown in Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>; data on degree of inflammation are in Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure'>4</ns0:ref>.</ns0:p><ns0:p>Results show an insignificant effect size and no significant differences between groups in terms of current pain intensity, or for any outcome measure. At the end of intervention (session 20), pain intensity for the TTEP group had decreased by 0.33 cm (95% CI -1.7 to 1.0, p=0.615) more than in the SSEP group. Both back school treatments showed positive results for pain reduction from baseline to end of treatment, and baseline to one month post-intervention. In the TTEP group, pain (baseline to final session) reduced by 4.6 cm (95% CI 3.3 to 5.8); the SSEP group's reduction was 4.3 cm (95% CI 3.0 to 5.6).</ns0:p><ns0:p>Similarly, there were an insignificant effect size and no significant differences between groups in terms of change in disability. At post-intervention (session 20), disability levels in the SSEP group had decreased by 0.40 points (95% CI -1.7 to 2.5, p=0.701) more than the TTEP group. Both back school treatments yielded positive results in disability reduction, baseline to end of treatment, and baseline to one month post-treatment. In the TTEP group, RMDQ scores reduced by 5.1 points (95% CI 3.0 to 7.3) from baseline to post-intervention. In the SSEP group, RMDQ reduction for the same interval was 6.1 points (95% CI 3.7 to 8.6).</ns0:p><ns0:p>Figure <ns0:ref type='figure'>3</ns0:ref> and Figure <ns0:ref type='figure'>4</ns0:ref> show outcomes for inflammation. TNF-&#945; showed higher values for the TTEP group than the SSEP group at the two visits where TNF-&#945; was measured. Significant differences were observed between the groups, baseline and post-treatment. In the first case, a difference of 66.97 pg/mL (95% CI 6.3 to 139.5) was recorded; in the second case the difference was 128.94 pg/mL (95% CI 52.8 to 205.0).</ns0:p><ns0:p>In contrast, IL-6 levels were found to be similar between the two treatment groups, with no significant differences observed. At baseline the difference was 2.51 pg/mL (95% CI -2.3 to 7.3); at post-intervention the difference was 1.25 pg/mL (95% CI -3.9 to 6.4).</ns0:p><ns0:p>In reference to the evolution of inflammatory biomarkers between baseline and post-treatment, the results for participants who practiced traditional TTEP indicate an increase in TNF-&#945; levels of 46.16 pg/mL (95% CI 13.0 to 85.3) and a tendency toward decreased levels of IL-6, 0.19 pg/mL (95% CI -1.6 to 1.2).</ns0:p><ns0:p>In contrast, the results in the group that practiced SSEP are the other way around: there was an increase in IL-6 levels of 1.06 pg/mL (95% CI 0.03 to 2.1) and a tendency toward decrease in TNF-&#945; levels of 12.81 pg/mL (95% CI -42.3 to 16.7).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our study hypothesis suggested that treatment with SSEP would be found to decrease pain and disability more effectively than TTEP, in women with non-specific LBP. We found this hypothesis not entirely true -although the effectiveness of SSEP was apparently demonstrated, the effectiveness of TTEP was found to be quite similar in our study group. The literature documents study results confirming those of our own study: Shamsi et al., also concluded that the two types of exercise provide improvement in LBP, but found no evidence as to which type might be more effective <ns0:ref type='bibr' target='#b44'>(Shamsi et al., 2015)</ns0:ref>. The literature also includes meta-analyses comparing back schools for chronic LBP. These found deep-muscle exercises more effective in reducing short-term pain and disability <ns0:ref type='bibr' target='#b5'>(Chang et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b30'>Niederer &amp; Mueller, 2020;</ns0:ref><ns0:ref type='bibr' target='#b62'>Wang et al., 2012)</ns0:ref>, though they found no significant differences in long-term improvement. In our study, however, after the half-way point in treatment (session 10) we observed pain reduction by both modalities. We believe this was most likely due to the fact that participants in the meta-analyzed clinical trials suffered from chronic lower back pain, which has a worse prognosis than non-specific LBP in an early phase <ns0:ref type='bibr' target='#b52'>(Van Den Hoogen et al., 1998)</ns0:ref>. Contrary to our results, in a recent systematic review, a meta-analysis of 8 studies indicated that stabilization exercises were more effective than general exercises in reducing pain. Five studies demonstrated a significant improvement in disability between patients treated with stabilization exercises compared with those treated with general exercises <ns0:ref type='bibr' target='#b14'>(Gomes-Neto et al., 2017)</ns0:ref>. In our case, the SSEP and TTEP seem to be effective in reducing pain and improving disability. The mean of pain in the analyzed studies was 6.01 at baseline, being 2.1 at the end of the stabilization exercises on a 0-10 pain scale <ns0:ref type='bibr' target='#b14'>(Gomes-Neto et al., 2017)</ns0:ref>. The SSEP results of our trial are consistent with these findings: 6.53 at baseline and 2.2 at the end of the intervention. There are authors who have found that stabilizing treatment shows no significant advantage over traditional treatment <ns0:ref type='bibr' target='#b24'>(Koumantakis et al., 2005)</ns0:ref>; some of these authors believe that where there appears to be such an advantage, it is due to certain characteristics of the LBP patients involved, such as segmental instability of the column, or the size of the multifidus muscles. One of the exclusion criteria in our study was diagnosis of other spinal disorders, so our sample was more homogeneous. Our results with regard to inflammation indicate that, following TTEP, TNF-&#945; levels had increased; when SSEP was used, IL-6 levels had increased by the end of our 20-session course of treatment. Al-Obaid et al., recently reported on a study with characteristics similar to ours, which found increased production of pro-inflammatory cytokine TNF-&#945; after treatment, but no change in IL-6 production <ns0:ref type='bibr' target='#b0'>(Al-Obaidi &amp; Mahmoud, 2014)</ns0:ref>. The author justified this result by stating that overexpression of TNF-&#945; and other pro-inflammatory cytokines occurs in many studies of lowback pathologies <ns0:ref type='bibr' target='#b46'>(Takahashi et al., 1996)</ns0:ref>. In addition, he explained, IL-6 cytokine levels are not altered because IL-6 has both pro-inflammatory and anti-inflammatory properties <ns0:ref type='bibr' target='#b32'>(Opal &amp; Depalo, 2000)</ns0:ref>. Various studies claim that IL-6 acts predominantly as an anti-inflammatory cytokine, regulating the synthesis of pro-inflammatory cytokines IL-1 and TNF-&#945; and stimulating the appearance, in circulation, of anti-inflammatory cytokines such as <ns0:ref type='bibr'>IL-10 (Opal &amp; Depalo, 2000;</ns0:ref><ns0:ref type='bibr' target='#b40'>Saavedra Ram&#237;rez et al., 2011)</ns0:ref>. One study goes further <ns0:ref type='bibr' target='#b38'>(Petersen &amp; Pedersen, 2005)</ns0:ref>, claiming that IL-6 stimulates lipolysis and oxidation of fats, as well as producing anti-inflammatory effects during exercise -and therefore may offer protection against TNF-&#945;. Relating this information to our own findings, we could say that treatment with SSEP aims to be more effective because, in our case, TNF-&#945; levels were maintained while IL-6 increased. On the other hand, with TTEP the reverse was true: the cytokine found to have increased in the plasma was TNF-&#945;. We believe this is due to the nature of the exercises. In SSEP, deep muscle exercise is the basis of lumbar and segmental control stabilization. TTEP, on the other hand, focuses on building overall muscle resistance, strength and flexibility, being a more dynamic and intense activity. The literature includes findings that lower-intensity exercises are more effective than those of greater intensity, when it comes to reducing inflammation <ns0:ref type='bibr' target='#b12'>(Ghafourian et al., 2016)</ns0:ref>.</ns0:p><ns0:p>One of our study's limitations is its sample size, but we also prioritized for this pilot trial the homogeneity of our patients through strict inclusion and exclusion criteria, for example we only studied women due their physiological characteristics such as less muscle and bone mass as well as psychological factors <ns0:ref type='bibr' target='#b20'>(Damian Hoy et al., 2012)</ns0:ref>. A study design with larger samples would allow a greater effect size between groups and the creation of subgroups according to age, degree of physical activity, or BMI -facilitating more definitive conclusions regarding these factors. Further, we believe it would be interesting to add another follow-up, beyond this study's onemonth-post-intervention evaluation. Further follow-up (at six months, for example) would reveal any difference between the treatments in terms of long-term clinical improvement, although the results of the current literature suggest that SSEP improves pain and functional status at 3 months but not at 6 or 12 months <ns0:ref type='bibr' target='#b6'>(Coulombe et al., 2017)</ns0:ref>. In summary, this study suggests that any type of back school exercise is highly effective in reducing pain and reducing disability in women with non-specific LBP. Further, it showed that SSEP seems to have an anti-inflammatory effect in such patients, potentially offering protection against chronic diseases associated with low-grade inflammation <ns0:ref type='bibr' target='#b38'>(Petersen &amp; Pedersen, 2005)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>This study adds to the literature the finding that both back school exercise program are apparently effective and equivalent in reducing pain and improving disability in women with non-specific LBP, from the tenth treatment session to one month after intervention. Moreover, it demonstrates the influence of each back school in the degree of inflammation, concluding that SSEP seems to increase production of anti-inflammatory biomarkers, while TTEP increases proinflammatory biomarker production. A large, adequately powered study is recommended to determine if the results from this pilot study can be duplicated. Manuscript to be reviewed Mean (SD) for outcomes reported at all study visits for total and each group, significant differences between visits within groups, p values, mean difference (95% CI) and effect size (95% CI) between groups for pain intensity and disability </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>TTEP= traditional trunk exercise program, SSEP= specific stabilization exercise program, VAS: visual analogue scale, NRS: numerical rating scale, RMDQ= Roland Morris disability questionnaire, p&lt;0.05= significant difference between groups,</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> </ns0:body> "

"Dear Editors We thank the reviewers for their generous comments on the manuscript and have edited the manuscript to address their concerns. In particular we have added some minors changes in Table 2. We believe that the manuscript is now suitable for publication in PeerJ. Dr. Eduard Minobes-Molina Department of Basic Health Sciences director Faculty of Health Sciences and Welfare University of Vic-Central University of Catalonia (UVic-UCC) On behalf of all authors Editor comments (Justin Keogh) I congratulate the authors for attending to all of the reviewers comments, with the very minor exception of the inclusion of some 95% confidence intervals is highlighted by reviewer one. Please ensure these confidence intervals are included so that this manuscript can be accepted for publication in PeerJ. Response: We thank the editor and the reviewers for their generous comments on the manuscript and have edited the manuscript to address their concerns. Your help during the process has been a rich learning for us. We want to highlight that following your advice, we have improved the quality of the manuscript. The minor changes related with the confidence intervals have been added in Table 2. The following are the answers we have made in response to the reviewer’s comments: Reviewer 1 Specific comments 1- I want to thank the authors for their hard work to revise the manuscript. The quality of the manuscript is far more improved now. I have just one minor comment. The authors mentioned that they have presented the standardized mean difference (SMD) with associate confidence intervals. However, I did not find the 95%CIs of the reported effect sizes in table 2. Please report them in Table 2. Please rename the label in the 6th column as “Mean difference (95% CI)”. In addition, in column #7, please rename the label as “Effect size (95% CI)” Response: We thank the reviewer for this helpful suggestion and have reported the effect size (95% CI) in the Table 2 accordingly. In addition we have renamed labels in the columns 6 and 7 as recommended by the reviewer. Reviewer 2 Basic reporting Acceptable. Experimental design Reframed appropriately. Validity of the findings The scope of the article will ensure findings are considered appropriately moving forward. Comments for the Author The authors have reframed the study to an appropriate scope. Response: We thank the reviewer for the valuable comments. We have reframed this article as a pilot study, considering that a large, adequately powered study is recommended to determine if the results from this pilot study can be duplicated. Thanks for positive remarks. Reviewer 3 Basic reporting No comment Experimental design No comment Validity of the findings No comment Comments for the Author The authors have thoroughly revised the manuscript and it is now suitable for publication to Peer J. Response: We thank the reviewer for their generous comments. Your help during the process has been useful to improve the quality of the manuscript. We thank you and reviewer’s time spent on this manuscript, and their suggestions have improved the paper, and we hope the manuscript is now in a form acceptable for publication in PeerJ. Sincerely, Eduard Minobes-Molina, Ph.D. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Individual animals experience different costs and benefits associated with group living, which may impact on their foraging efficiency in ways not yet well specified. This study investigated associations between social dominance, body condition and interruptions to foraging behaviour in a cross-sectional study of 116 domestic horses and ponies, kept in 20 discrete herds. Social dominance was measured for each individual alongside observations of winter foraging behaviour. During bouts of foraging, the duration, frequency and category (vigilance, movement, social displacements given and received, scratching and startle responses) of interruptions were recorded, with total interruption time taken as a proxy measure of foraging efficiency. Total foraging time was not influenced by body condition or social dominance. Body condition was associated with social dominance, but more strongly associated with foraging efficiency. Specifically, lower body condition was associated with greater vigilance. This demonstrates that factors other than social dominance can result in stable differences in winter body condition.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Social behaviour can influence energetic reserves and subsequent body condition. Previous modelling studies have outlined the potential importance of social effects on foraging behaviour (bouts of biting, chewing and swallowing interrupted by relocation movements) in determining body condition in group living animals <ns0:ref type='bibr' target='#b26'>(Houston and McNamara, 1999;</ns0:ref><ns0:ref type='bibr' target='#b43'>Rands et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b44'>2004;</ns0:ref><ns0:ref type='bibr' target='#b45'>2006;</ns0:ref><ns0:ref type='bibr'>2008)</ns0:ref> and also the role of dominance behaviours in determining resource access and subsequent body condition <ns0:ref type='bibr' target='#b2'>(Clark and Ekman, 1995;</ns0:ref><ns0:ref type='bibr' target='#b50'>Stillman et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b45'>Rands et al., 2006)</ns0:ref>.</ns0:p><ns0:p>Thus, the foraging success of individual animals in social groups may be partly influenced by their social status. However, few of these predictions have been investigated empirically in socially-foraging herbivores and the relationship between herd behaviours, dominance and body condition is not fully understood.</ns0:p><ns0:p>In a socially foraging herbivore the benefits of group living outweigh the costs <ns0:ref type='bibr' target='#b30'>(Krause and Ruxton, 2002)</ns0:ref>. Individual animals living within groups follow behavioural rules which allow them to function as a social unit <ns0:ref type='bibr' target='#b22'>(Hemelrijk, 2002;</ns0:ref><ns0:ref type='bibr'>Rands, 2011a,b)</ns0:ref>. These rules are likely to depend upon both aspects of their own body condition (such as energetic reserves) and also the actions of other individuals within the group <ns0:ref type='bibr' target='#b26'>(Houston and McNamara, 1999;</ns0:ref><ns0:ref type='bibr' target='#b43'>Rands et al., 2003;</ns0:ref><ns0:ref type='bibr'>2008)</ns0:ref>. Rules governing social interaction (e.g. dominance) may be important for a wellfunctioning group in terms of minimising costly conflict over resources <ns0:ref type='bibr' target='#b30'>(Krause and Ruxton, 2002)</ns0:ref>. <ns0:ref type='bibr' target='#b42'>Rands et al. (2011b)</ns0:ref> considered a game theoretical framework to explore how the rules used by individuals with different dominance ranks could evolve, assuming these individuals paid attention to the ranks and energetic state of both themselves and the individual that they were interacting with. This model, and a companion simulation exploring the rules of thumb generated <ns0:ref type='bibr' target='#b41'>(Rands 2011a</ns0:ref>) demonstrated that both energetic state and social status are important for determining the behaviour of co-foraging individuals. Furthermore, individual-based simulations <ns0:ref type='bibr' target='#b44'>(Rands et al 2004</ns0:ref><ns0:ref type='bibr' target='#b45'>(Rands et al , 2006</ns0:ref>) demonstrated that including an additional effect of dominance that led to subordinates having reduced access to food could lead not only to dominant individuals PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020) Manuscript to be reviewed having higher energetic reserves than subordinates, but also subordinate individuals increasing their activity.</ns0:p><ns0:p>We aimed to assess whether this framework was useful in understanding the foraging behaviour of the horse. We were particularly interested to determine whether dominant animals had higher body condition and whether subordinate individuals showed increased activity in line with model predictions. Horses are generalist herbivores with sophisticated social capacities. Free-ranging horses spend between 9 and 16h each day foraging <ns0:ref type='bibr' target='#b8'>(Ellis, 2010)</ns0:ref> maintaining a high daily intake of plant material by grazing (or browsing) interrupted by frequent walking. Horses form strong affiliative bonds with familiar companions, but aggressive encounters and subtle threats, are also a common feature of equine social structure, particularly when resources are limited <ns0:ref type='bibr'>(Mills and Redgate, 2010)</ns0:ref>. The current study was conducted under winter conditions where pasture availability was limited and a degree of competition for supplementary forage was evident. s situation applies commonly for domestic horses (kept for a variety of reasons including as companion animals or as conservation grazers <ns0:ref type='bibr' target='#b12'>(Gilhaus and Hoelzel, 2016)</ns0:ref> during winter periods within temperate zones). Understanding the factors that drive large inter-individual differences in body condition when group-living horses are kept during winter (e.g. <ns0:ref type='bibr' target='#b28'>Ing&#243;lfsd&#243;ttir and Sigurj&#243;nsd&#243;ttir, 2008;</ns0:ref><ns0:ref type='bibr' target='#b14'>Giles et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b56'>Yngvesson et al., 2019)</ns0:ref> is an important goal. It has been estimated that around a third of outdoor living horses and ponies within the UK are obese <ns0:ref type='bibr' target='#b13'>(Giles et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b48'>Robin et al., 2015)</ns0:ref> but rates of obesity can reach 70% in some populations <ns0:ref type='bibr' target='#b34'>(Menzies-Gow et al., 2017)</ns0:ref>. It is timely to study the social factors influencing body condition in horses to reduce obesity prevalence and associated metabolic disease.</ns0:p><ns0:p>Previous empirical studies in horses have demonstrated that higher ranking individuals spend more time eating hay and have a higher body condition during the winter <ns0:ref type='bibr' target='#b28'>(Ing&#243;lfsd&#243;ttir and Sigurj&#243;nsd&#243;ttir, 2008;</ns0:ref><ns0:ref type='bibr' target='#b14'>Giles et al., 2015)</ns0:ref> but have not examined the mechanisms behind this association.</ns0:p><ns0:p>This study advanced our previous work by examining situations where bouts of foraging on supplementary forage were interrupted for reasons including anti-predator vigilance and startle responses <ns0:ref type='bibr' target='#b15'>(Goodwin, 1999)</ns0:ref>, displacement interruptions directed towards or received from other group members <ns0:ref type='bibr' target='#b0'>(Appleby, 1980;</ns0:ref><ns0:ref type='bibr' target='#b45'>Rands et al., 2006)</ns0:ref> or short movements between foraging locations <ns0:ref type='bibr' target='#b4'>(Duncan, 1980)</ns0:ref>. We examined the duration, frequency and type of interruption to the foraging behaviour of individual horses and ponies (hereafter termed 'horses') living in social herds. The total time attributed to interrupted foraging was considered as a proxy measure of foraging efficiency (the ratio of energy gained over energy expended during foraging).</ns0:p><ns0:p>An important precursor to analysing foraging efficiency was understanding any differences in overall time spent foraging. We measured overall time spent foraging to check that individuals with a lower foraging efficiency didn't simply compensate by spending more time foraging. A unique feature of the study was the inclusion of measures of social status and body condition, enabling the assessment of associations not previously examined in foraging herbivores.</ns0:p><ns0:p>Predictions suggest that subordinate individuals may suffer more displacement than dominant conspecifics <ns0:ref type='bibr' target='#b17'>(Goss-Custard et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b50'>Stillman et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b20'>2000;</ns0:ref><ns0:ref type='bibr' target='#b45'>Rands et al., 2006)</ns0:ref>, reflected in increased displacement interactions and subsequent movement within foraging bouts. Dominant animals may also force subordinate conspecifics into more exposed foraging positions <ns0:ref type='bibr' target='#b5'>(Ekman, 1987;</ns0:ref><ns0:ref type='bibr' target='#b44'>Rands et al., 2004)</ns0:ref> leading to a reduction in foraging efficiency due to a greater requirement for vigilance. In contrast, models predict that dominant individuals will be more efficient foragers, feeding in positions with lower interference, potentially leading to a greater energetic intake and overall body condition <ns0:ref type='bibr' target='#b5'>(Ekman, 1987;</ns0:ref><ns0:ref type='bibr' target='#b49'>Schneider, 1984;</ns0:ref><ns0:ref type='bibr' target='#b45'>Rands et al., 2006)</ns0:ref>.</ns0:p><ns0:p>A greater body condition may in turn allow a subsequent competitive advantage <ns0:ref type='bibr'>(Rands, 2011;</ns0:ref><ns0:ref type='bibr' target='#b45'>Rands et al., 2006)</ns0:ref>. Our aims were to: i) Confirm an association between dominance rank (adjusted for herd size, see Methods) and body condition. ii)</ns0:p><ns0:p>Assess whether adjusted dominance rank is associated with interruptions to foraging (as a proxy for foraging efficiency). iii)</ns0:p><ns0:p>Assess whether body condition is associated with interruptions to foraging (as a proxy for foraging efficiency). iv)</ns0:p><ns0:p>Use multivariate analysis to investigate the contextual factors (age, breed, sex, height, supplementary feeding) that might influence these associations.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed v) Consider the applied implications of our findings for the management of domestic horses. We predicted that foraging interruptions would be associated with both body condition and dominance status, and that subordinate individuals would, overall, have a reduced foraging efficiency compared with more dominant conspecifics and a lower body condition, as indicated in a previous study <ns0:ref type='bibr' target='#b14'>(Giles et al., 2015)</ns0:ref>. This study goes beyond previous research to assess whether differences in foraging efficiency could plausibly be the mechanism linking dominance to body condition.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>a) Animals and Ethical Statement</ns0:head><ns0:p>The work was approved by the University of Bristol Animal Welfare and Ethical Review Board (University Investigation Number UB/10/049) and all methods were carried out in accordance with relevant guidelines.</ns0:p><ns0:p>The study sample was drawn from a population of outdoor, group-living horses based at Redwings Horse Sanctuary (UK), that had been living together for at least three months and had established social relationships. All of the individual animals were managed similarly, fed forage from identical sources, lived in outdoor environments and were not ridden, meaning that structured exercise could be removed as a potential confounding factor. Herds that included pregnant or lactating mares were not considered for the study. Twenty study herds were selected randomly from all remaining suitable herds within the sampling frame.</ns0:p><ns0:p>The policy of the sanctuary was to house horses in relatively compatible groups with shared characteristics. Thus, larger horses were housed in separate herds from smaller ponies, all stallions were housed in one 'bachelor' herd, while youngsters were also housed together, with the few horses under 1 year of age (three individuals) accompanied by older 'nanny' mares.</ns0:p><ns0:p>Herd size was 2-10 (mean 6 &#177; 0.56 individuals). 116 individuals (84 ponies of height &lt;148cm, and 32 horses of height &#8805; 148cm) from within these herds were observed between 2 December, PeerJ reviewing <ns0:ref type='table' target='#tab_4'>PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:ref> Manuscript to be reviewed 2013 and 23 January, 2014. Ages ranged from 5 months to 32 years (11.83 &#177; 0.63 years). Breeds were native ponies (51.72%), native cobs (17.24%), lightweight horses (12.07%), heavy horses (5.17%), sports horse breeds (5.17%) and other (8.62%).</ns0:p></ns0:div> <ns0:div><ns0:head>b) Study period and horse management</ns0:head><ns0:p>The winter months were chosen for observation as natural food resources were at their minimum and therefore food based social interactions were likely at their highest due to the close proximity of individuals. All horses lived in an outdoor paddock environment for 24 hours a day and were fed from circular hay feeders provided at a fixed ratio of feeder space per animal.</ns0:p><ns0:p>Horses were fed twice daily with fresh hay replenished once at the start of morning observation (between 08:00 and 09:00) and once at the start of afternoon observation (between 11.30 and 13:00). Any uneaten hay remained in the hay feeder throughout the day. Twelve study horses received additional supplementary feed from a bucket once a day, and this was recorded as a potential confounder.</ns0:p></ns0:div> <ns0:div><ns0:head>c) Time spent foraging</ns0:head><ns0:p>Each study herd was observed for six hours to assess overall time spent foraging, and interruptions occurring during foraging bouts, once during a three hour morning session (08:00-09:00 until 11:00-12:00) and once during a three hour afternoon session (11:30-13:00 until 14:30-16:00) on a different day within the same week, by a single trained observer. Due to the time of year, these times were chosen based on daylight hours. Time spent foraging was recorded using scan sampling at five minute intervals throughout each three-hour observation period. A random number generator was used to determine the order in which individuals were observed. Once this order was determined, all individuals were observed in sequence, in five-minute intervals. At each interval, it was recorded which individuals were foraging and which were not. Foraging was defined as the horse ingesting either hay or grass, with intermittent periods of the head down ingesting forage and the head up chewing this forage material. The horse could be foraging from either the hay feeder or eating grass (although the PeerJ reviewing <ns0:ref type='table' target='#tab_4'>PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:ref> Manuscript to be reviewed latter was rare as there was little grass available). The percentage of time spent foraging was then calculated based on the number of intervals that each individual was foraging within the full six hours of observation per herd.</ns0:p><ns0:p>Alongside this, continuous five minute focal animal observations were scheduled for each horse during each three hour recording period. Each individual animal was independently observed for at least 20 minutes (4 &#215; 5-minutes) in total. These observations were predominantly used to record foraging interruptions and social interactions (as detailed in sections d and e below), however they were also used to more accurately estimate the total foraging time for each individual. If an individual was not foraging for more than one minute during the five-minute observation period, it was considered to have stopped foraging. The number of minutes it had stopped foraging for were then subtracted from the total five minutes.</ns0:p></ns0:div> <ns0:div><ns0:head>d) Foraging efficiency -duration and frequency of foraging interruptions</ns0:head><ns0:p>During the continuous five-minute focal animal observations, described above, observations relating to foraging interruptions were also conducted. Interruption to foraging was defined as an activity that was short in duration (less than one minute) and prevented the individual from selecting, biting or chewing hay or grass. Both the frequency and overall duration of any interruption was recorded and interruptions were categorised as one of the following:</ns0:p><ns0:p>Vigilance: Head raised from foraging and ears pricked in the direction of interest, the head is higher and the ears upright distinguishing vigilance from raising the head to chew. Manuscript to be reviewed If any interruption lasted for over one minute then the individual was classed as having stopped foraging. Note that individuals were only observed in detail when they were foraging, if an individual was not foraging when it was due to be observed, this was recorded (to calculate total foraging time, as described in section a) and but also counted as 'missed' in terms of recording interruptions. Once a missed individual was foraging again it was observed next as a priority (only if it had not yet already been observed for 20 minutes), but just for a single five-minute interval, before resuming the original order. This was to maximise the collection of data on foraging efficiency for each individual.</ns0:p></ns0:div> <ns0:div><ns0:head>Movement whilst foraging</ns0:head><ns0:p>The frequency of foraging interruption (a proxy for foraging efficiency) was calculated as the number of instances of all interruptions per minute foraging. Separate frequencies were also determined for each interruption category (Table <ns0:ref type='table'>1</ns0:ref>). The duration of interrupted foraging referred to the total percentage of time spent interrupted per individual.</ns0:p></ns0:div> <ns0:div><ns0:head>e) Dominance rank</ns0:head><ns0:p>Although the concept of dominance lacks universal explanatory power in describing social structure, it is a useful construct when considering the specific context of competition for a limited food resource. Under such conditions, horses generally follow a linear ranking hierarchy, with occasional triangles and some influence of third-party interactions <ns0:ref type='bibr' target='#b25'>(Houpt et al., 1978;</ns0:ref><ns0:ref type='bibr' target='#b55'>van Dierendonck et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b19'>Hartmann et al., 2017</ns0:ref>).</ns0:p><ns0:p>Here we defined dominance 'an asymmetry in the outcome of dyadic interactions between individuals, or a priority of access to resources' <ns0:ref type='bibr' target='#b3'>(Drews, 1993)</ns0:ref> and assessed it by measuring outcomes between dyadic pairs when feeding from hay feeders. Agonistic interactions were recorded continuously throughout the three-hour observation period (these were easily measurable alongside other observations). An agonistic interactionwas defined as one individual approaching or displaying to another with the neck outstretched and ears back flat against the head and, crucially, the second individual moving away. Dominance rank was then calculated PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed using the methods described by <ns0:ref type='bibr' target='#b0'>Appleby (1980)</ns0:ref>. The number of agonistic interactions both given and received was recorded for each herd individual, and then the number of other individuals that a focal individual both dominated and was dominated by was calculated.</ns0:p><ns0:p>Once an Appleby rank had been given, this was then adjusted to take into account herd size (as in <ns0:ref type='bibr' target='#b14'>Giles et al., 2015)</ns0:ref>. Adjusted dominance rank was calculated as 1 -(a -1)/(h -1), where a is the Appleby rank and h is the herd size. Where dominance rank or dominance status is referred to in this manuscript, this refers to this adjusted dominance rank.</ns0:p><ns0:p>f) Body condition score Measurements were taken immediately after the second set of observations on the herd had been completed. All study animals were accustomed to being handled. Body condition score was measured using the Henneke nine-point scale <ns0:ref type='bibr' target='#b23'>(Henneke et al., 1983</ns0:ref>) by a single trained observer (SLG). Six areas of the horse were scored between 1 and 9 and then averaged and rounded to the nearest 0.5, to obtain a single score. A score of five on the scale was taken to indicate an ideal body condition.</ns0:p></ns0:div> <ns0:div><ns0:head>g) Statistical analyses</ns0:head><ns0:p>Results were analysed using Stata 12.1 (Statacorp, Texas). Univariable relationships were assessed using mixed effects linear regression, the clustered study design was controlled for by including herd group and herd size as a random effects, on the basis that herd size or other herd specific factors such as environment could plausibly have some influence on foraging and interactive behaviours. Univariable relationships of primary interest were:</ns0:p><ns0:p>1) The relationship between dominance rank (adjusted for herd size) and body condition score 2) The relationship between dominance rank (adjusted for herd size) and interruptions to foraging (as a proxy for foraging efficiency)</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>3) The relationship between body condition and interruptions to foraging (as a proxy for foraging efficiency)</ns0:p><ns0:p>Following an initial univariable exploration of these relationships, relationships between the separate foraging interruption variables were also considered. In addition, breed, age, height, sex and whether or not the individual received supplementary feed were recorded as potential confounding variables. To be considered a potential confounder the variable had to be associated with both the explanatory and outcome variable, and not on the causal pathway between the two <ns0:ref type='bibr' target='#b39'>(Petrie and Sabin, 2009)</ns0:ref>. Statistical significance was defined using p&#8804;0.05 with a screening pvalue for multivariable models of p &#8804; 0.07.</ns0:p><ns0:p>Mixed effects multivariable linear regression was then used to build a best-fit explanatory model for both adjusted dominance rank and body condition. The foraging interruption variables (see Table <ns0:ref type='table'>1</ns0:ref> for list) were added to the model one at a time, based on the strength of univariable association, starting with a minimal model. A likelihood ratio test was used to assess the contribution of each variable to the model fit and variables were retained on the basis of this and the adjusted p value.</ns0:p><ns0:p>Multivariable analysis using a mixed effects linear regression model was also used to make predictions regarding interruptions to foraging -to explore whether this could be a possible mechanism linking dominance status and body condition. Duration of foraging interruption was associated with both dominance status and body condition, therefore this was added to a model containing adjusted dominance rank and body condition. Its explanatory contribution to the model was then assessed using both the adjusted p and estimates and a likelihood ratio test.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>During 120h of observation, the amount of time that individual animals spent foraging averaged 76.4% SD 0.17. Values per herd are given in Table <ns0:ref type='table'>S1</ns0:ref>. Figure <ns0:ref type='figure' target='#fig_2'>1</ns0:ref> shows that there was no significant correlation between adjusted dominance rank and foraging time (r 2 = 0.004, n = 116, p = 0.51) or between body condition score (range 4 to 8.5) and total foraging time (r 2 = 0.016; n = 116, p = 0.182). This is important in the interpretation of subsequent results.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>a) Univariable Analysis</ns0:head><ns0:p>The relationship between adjusted dominance rank and body condition score</ns0:p><ns0:p>Adjusted dominance rank was positively associated with body condition score within our study population (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Foraging Efficiency</ns0:head><ns0:p>During approximately 92h of the 120h total observation period, horses were foraging. During this time, the observed frequencies of each type of interruption contributing to foraging efficiency were: vigilance 2518; movement whilst foraging 454; displacements given 198; displacements received 222; scratching 65; startle responses 5.</ns0:p></ns0:div> <ns0:div><ns0:head>The relationship between dominance rank and foraging efficiency</ns0:head><ns0:p>Although the frequency of foraging interruptions did not show evidence of association with adjusted dominance rank (Z=-1.55, p=0.12, Table <ns0:ref type='table' target='#tab_0'>S2</ns0:ref>), the total duration of interruptions decreased as adjusted dominance rank increased (Table <ns0:ref type='table'>1</ns0:ref>). An increase in adjusted dominance rank was also associated with a decrease in some specific interruption behaviours, namely instances of movement whilst foraging, displacements given, and displacements received (Table <ns0:ref type='table'>1</ns0:ref>). Figure <ns0:ref type='figure' target='#fig_2'>1</ns0:ref> shows that the reduced foraging efficiency of subordinate individuals is not compensated for by an increase in total foraging time.</ns0:p></ns0:div> <ns0:div><ns0:head>The relationship between body condition score and foraging efficiency</ns0:head><ns0:p>The number of incidences (frequency) of foraging interruptions occurring during foraging bouts was lower for animals with higher body condition scores. Vigilance decreased with an increase in body condition (Table <ns0:ref type='table'>1</ns0:ref>), but none of the other separately defined foraging interruptions showed any association with body condition (Supplementary Information, Table <ns0:ref type='table' target='#tab_0'>S2</ns0:ref>). Figure <ns0:ref type='figure' target='#fig_2'>1</ns0:ref> shows that the reduced foraging efficiency of individuals with lower body condition is not compensated for by an increase in total foraging time.</ns0:p></ns0:div> <ns0:div><ns0:head>Associations between the individual foraging interruption variables and consideration of potential confounders</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Frequency of 'displacements received' was strongly associated with 'moving whilst foraging' and 'displacements given'. Frequency of 'displacements given' was also associated with 'moving whilst foraging' (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>In this study, none of the potential confounder variables (breed, age, height, sex) were associated with body condition score, adjusted dominance rank or any category of interrupted foraging, and there were no biologically plausible interactions, therefore adjusted estimates were not required.</ns0:p><ns0:p>This also included whether or not a horse received additional supplementary feed, which showed no evidence of association with either adjusted dominance rank (Z = -0.50, p = 0.61) or body condition ( = 12.40, p = 0.19).</ns0:p><ns0:p>&#120568; 2 9 b) Multivariable analysis</ns0:p></ns0:div> <ns0:div><ns0:head>Model for adjusted dominance rank</ns0:head><ns0:p>Controlling for other model variables, frequency of 'displacements received', 'displacements given' and body condition score were associated with adjusted dominance rank (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Model for body condition score</ns0:head><ns0:p>Controlling for other model variables, vigilance frequency and adjusted dominance rank were strongly associated with body condition score (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The relationship between body condition score and adjusted dominance rank when taking into account interruptions to foraging</ns0:head><ns0:p>The association between body condition score and adjusted dominance rank was weaker when total duration of foraging interruptions (or time spent interrupted) was included in the model (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>, p = 0.06, as opposed to p = 0.03 in the univariable model). The effect size also reduced slightly (from a 0.66 increase in adjusted dominance rank per half unit of body condition score to 0.55). The likelihood ratio test results (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>) indicate that duration of foraging interruptions has a more significant contribution to the model fit (p = 0.04) than adjusted dominance rank (p = 0.06).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The study explored the inter-relationships between foraging interruptions, dominance and body condition, controlling for herd size and herd identity effects. No effects of age, sex or height were detected in our study. Clearly, large horses have differing energy requirements from smaller ponies, whilst growing youngsters and older horses with reduced digestive efficiency (e..g <ns0:ref type='bibr'>Ralston et al., 1989)</ns0:ref> will also differ from young but mature adults. However, the horses in our study were housed in herds that contained animals of similar characteristics (see Methods and Supplementary Table <ns0:ref type='table'>)</ns0:ref>. For example, heavy horses were housed separately from lighter Thoroughbreds and smaller ponies. Although this policy greatly reduces or eliminates our ability to detect age and sex effects on foraging, it enhances our ability to detect the relative effects of dominance and body condition within herds. Importantly, our analysis showed that the relationships we detected applied across all herd types.</ns0:p><ns0:p>Within this study population, dominance status was positively associated with body condition, although this relationship was weaker when foraging efficiency was included in the multivariate model (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>). In addition, the association between body condition and foraging efficiency was stronger than that between body condition and dominance. Thus, whilst dominance explains some variation in body condition, our results highlight the potential role of factors other than social dominance that could influence foraging efficiency. Factors such as a tendency to show vigilance behaviour have been little explored to date but have the potential to greatly influence the ratio of energy gained vs energy expended during bouts of foraging.</ns0:p><ns0:p>There was no evidence that subordinate or low body condition individuals compensated for less efficient foraging by increasing total foraging time. Another recent study found that horses with low body condition tend to adopt more passive behaviour <ns0:ref type='bibr' target='#b29'>(Jorgensen et al., 2016)</ns0:ref>. Potentially such results may be due to a strong motivation to feed as a group in this species and thus synchronise feeding and resting behaviour <ns0:ref type='bibr'>(Rands et al., 2008)</ns0:ref>. Subordinate or lower body score individuals were unlikely to remain foraging when conspecifics were not, supporting suggestions that social factors may result in stable differences in body condition within group living animals <ns0:ref type='bibr'>(Rands, 2011;</ns0:ref><ns0:ref type='bibr'>Rands et al., 2010)</ns0:ref>. Indeed the tendency to synchronous feeding and resting (as in sheep, McDougall and Ruckstuhl, 2018) may be hard-wired as an optimal behaviour.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The lack of a compensatory change in total foraging time means that any variation observed in foraging efficiency could plausibly have an effect on body condition.</ns0:p><ns0:p>Given these results and previous theoretical predictions, an association between foraging efficiency, dominance and overall body condition was expected <ns0:ref type='bibr' target='#b32'>(McNamara and Houston, 1990;</ns0:ref><ns0:ref type='bibr' target='#b51'>Stillman et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b45'>Rands et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b47'>Rands and Whitney, 2008)</ns0:ref> but our study is the first to explore the role of the different components of foraging efficiency, such as movement, social displacement or vigilance.</ns0:p></ns0:div> <ns0:div><ns0:head>Vigilance and body condition</ns0:head><ns0:p>Vigilance frequency was the individual interruption behaviour most strongly associated with body condition score -it showed a strong negative association. However vigilance was not associated with dominance status. These results suggest that certain individuals may be more likely to conduct vigilance, perhaps on behalf of the group, regardless of their social status.</ns0:p><ns0:p>These results do seem to support the suggestion that vigilance is an inherently costly activity <ns0:ref type='bibr' target='#b7'>(Elgar, 1989;</ns0:ref><ns0:ref type='bibr' target='#b10'>Fritz et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b9'>Fattorini and Ferretti, 2019)</ns0:ref> as demonstrated by the negative association with body condition. However, lower body condition individuals may also be more stressed or nervous individuals, which would also explain the association with increased vigilance.</ns0:p><ns0:p>The complexity of vigilance as a single trait may somewhat explain the lack of observed association with dominance status. Vigilance may serve a range of functions in group living animals <ns0:ref type='bibr' target='#b9'>(Fattorini and Ferretti, 2019)</ns0:ref>, including anti-predatory behaviour <ns0:ref type='bibr' target='#b7'>(Elgar, 1989;</ns0:ref><ns0:ref type='bibr' target='#b27'>Hunter and Skinner, 1998)</ns0:ref>, monitoring of other herd members and scanning the environment for resources <ns0:ref type='bibr' target='#b54'>(Underwood, 1982)</ns0:ref>. Ungulate mammals that are unexposed to predation have been observed to greatly reduce their vigilance behaviour <ns0:ref type='bibr' target='#b27'>(Hunter and Skinner, 1998)</ns0:ref>. Horses, unexposed to predation, may therefore show relatively low levels of vigilance, with reasons other than anti-predatory vigilance having a proportionally larger role.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Alongside the association between dominance status and body condition, the association between body condition and vigilance provides evidence of two separate behavioural traits associated with body condition in group living animals. Behavioural predictors of body condition have so far received little attention in horses (for exceptions, see <ns0:ref type='bibr' target='#b28'>Ing&#243;lfsd&#243;ttir and Sigurj&#243;nsd&#243;ttir, 2008;</ns0:ref><ns0:ref type='bibr' target='#b14'>Giles et al., 2015)</ns0:ref> and may warrant continued investigation, especially as obese horses (BCS &gt;7) may show differences in activity and eating behaviour when compared to lean horses (BCS 4-5) <ns0:ref type='bibr' target='#b37'>(Moore et al., 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Dominance status, movement during foraging and displacement interactions</ns0:head><ns0:p>Subordinate horses showed more movement whilst foraging, and were (as expected) more likely to receive displacements. Indeed, statistical analysis revealed that displacement was strongly associated with movement during foraging in our study population, with subordinate animals forced to move foraging location. Theoretical models and empirical studies have proposed that subordinate individuals may be forced to foraging positions carrying a greater risk of predation <ns0:ref type='bibr' target='#b18'>(Hamilton, 1971;</ns0:ref><ns0:ref type='bibr' target='#b20'>Hemelrijk, 2000)</ns0:ref>. Future studies could examine whether subordinate animals showed increased vigilance specifically when in displaced locations, and during non-foraging periods.</ns0:p><ns0:p>Overall our results therefore appear to support predictions that displacement reduces foraging efficiency for the recipient <ns0:ref type='bibr' target='#b1'>(Bautista et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b52'>Stillman et al., 2002)</ns0:ref>. Valuable foraging time is wasted not only over the initial dispute, but also in relocating to a new foraging location. In contrast, dominant horses tended to interrupt their own foraging to displace others, but these interruptions tended to be of short duration, allowing the dominant animal to return quickly to foraging. As our study herds were feeding from hay feeders, potentially displacement and movement occurred more often than would occur during foraging on pasture, due to the artificially close proximity of herd members <ns0:ref type='bibr'>(Hoffman et al., 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>These results are novel and exciting in that they present the first behavioural evidence Manuscript to be reviewed 2010; <ns0:ref type='bibr' target='#b43'>Rands et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b45'>2006;</ns0:ref><ns0:ref type='bibr'>Rands 2011;</ns0:ref><ns0:ref type='bibr' target='#b53'>Sueur et al., 2013)</ns0:ref> linking condition and behaviour in a group-living species. Our results suggest (in line with model predictions) that differences in energetic reserves (body condition) can emerge simply via a reduction in energetic intake by subordinates when dominants are present. This hypothesis could be further tested in a future prospective study. If easily-observable individual behavioural differences such as high social dominance can influence body condition, then there is a potential to put measures in place to prevent dominant animals becoming obese. Horses, like other companion animals <ns0:ref type='bibr' target='#b11'>(German et al., 2019)</ns0:ref> and like humans, are experiencing an obesity crisis <ns0:ref type='bibr' target='#b13'>(Giles et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b48'>Robin et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b34'>Menzies-Gow et al., 2017)</ns0:ref>. Consideration of the social factors driving this could improve the overall health of herd-living animals.</ns0:p><ns0:p>Table <ns0:ref type='table'>1</ns0:ref>. Statistically significant univariable associations (p &#8804; 0.05) using mixed effects linear regression, controlling for herd group and herd size as a random effects. Non-significant associations are given in the supplementary material, Table <ns0:ref type='table' target='#tab_0'>S2</ns0:ref>. Manuscript to be reviewed The final multivariable explanatory model for body condition score, using mixed effects linear regression, controlling for herd group and herd size as random effects. Manuscript to be reviewed Multivariable linear regression model showing the effect of foraging efficiency (total duration of foraging interruptions) upon the relationship between dominance status and body condition.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>: a short movement resulting in a change in foraging location, either following a displacement by another individual or simply changing location at a walk. Displacements given: interaction directed towards another individual, with the head outstretched and ears flat back against the head resulting in recipient raising head, or taking a step away in any direction. Displacements received: interaction received from another individual defined as above, causing recipient to raise head, move sideways or take a step away in any direction. Scratching: Using either the mouth or the hoof to scratch the body Startle response: A quick reaction to an unexpected stimulus, the startle usually involved a quick movement, either jump backwards or sideways followed by looking up with ears pricked PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>confirming a broad body of influential theoretical work (e.g. Marshall et al., 2012; Petit and Bon, PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>The final multivariable explanatory model for adjusted dominance rank, using mixed effects linear regression, controlling for herd group and herd size as random effects.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Interruption behaviour variables</ns0:cell><ns0:cell>&#946;</ns0:cell><ns0:cell>S.E.</ns0:cell><ns0:cell>95% CI</ns0:cell><ns0:cell>Z</ns0:cell><ns0:cell>p</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>The final multivariable explanatory model for body condition score, using mixed effects linear regression, controlling for herd group and herd size as random effects.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Explanatory variable</ns0:cell><ns0:cell>&#946;</ns0:cell><ns0:cell>SE</ns0:cell><ns0:cell>95% CI</ns0:cell><ns0:cell>Z</ns0:cell><ns0:cell>p</ns0:cell></ns0:row><ns0:row><ns0:cell>Vigilance frequency</ns0:cell><ns0:cell>-0.89</ns0:cell><ns0:cell>0.30</ns0:cell><ns0:cell>-1.48 --0.31</ns0:cell><ns0:cell>-3.01</ns0:cell><ns0:cell>0.003</ns0:cell></ns0:row><ns0:row><ns0:cell>Adjusted dominance rank</ns0:cell><ns0:cell>0.63</ns0:cell><ns0:cell>0.29</ns0:cell><ns0:cell>0.06 -1.18</ns0:cell><ns0:cell>2.19</ns0:cell><ns0:cell>0.03</ns0:cell></ns0:row><ns0:row><ns0:cell>Constant</ns0:cell><ns0:cell>6.14</ns0:cell><ns0:cell>0.23</ns0:cell><ns0:cell>5.68 -6.59</ns0:cell><ns0:cell>26.55</ns0:cell><ns0:cell>&lt;0.001</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Multivariable linear regression model showing the effect of foraging efficiency (total duration of foraging interruptions) upon the relationship between dominance status and body condition.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Likelihood Ratio Test</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46192:1:1:NEW 12 Aug 2020)</ns0:note></ns0:figure> </ns0:body> "

"Dear Editor Many thanks for allowing us the opportunity to improve our manuscript. We are pleased that all three reviewers found our study interesting and we are very grateful for the useful comments that they have provided. We have improved the manuscript by better matching our aims with our discussion/conclusions, clarifying our sometimes inconsistent or confusing use of terminology/definitions, and by adding important contextual information about herd composition. Specific responses are outlined below. Reviewer 1 Thank you for a very nice study in an interesting area! Thank you. Reviewer 2 L 57 One bracket seems to be missing The bracket has been added. I miss information about the behavioural ecology of horses. Horses are used as a model species for a social herbivore in this study, but herbivores differ substantially in their foraging strategies. E.g. horses graze and move more or less constantly whereas cows graze and ruminate (lying down) intermittently. Hence, a few sentences about the social structure of horses and their natural foraging behaviour would help the reader to understand how they fit into the theoretical framework described in the introduction. A few sentences have been added about the social structure and natural foraging behaviour of horses (lines 66-71 on clean revised copy). L 106-107 the first sentence starting with ”The study explored…” This sentence does not belong in the Materials & Methods section but rather as an early sentence in the Discussion section. This sentence moved as suggested L 140-143 This paragraph does not belong in the Materials & Methods, but rather in the Introduction. This paragraph moved as suggested L 174 What is an ”energetic reward”? This was over-complicated. Simplified (line 198 on clean revised copy) L 205 Dominance hierarchy in horses is very complex and right now debated. I think the MS would be improved if you were very clear that this is dominance in relation to the resource forage when this resource is limited (usually by people). We agree with this, and have clarified the specific context in which we consider dominance rank (lines 229-233 on clean revised copy) L 341-343 I agree that vigilance is very interesting. Could vigilance behaviour be correlated to any of the factors age, breed, herd size etc? Maybe this will be published in another paper? – We did analyse whether any of the behaviours that contributed to “interrupted foraging” (including vigilance) were associated with breed, age, height or sex (previous lines 309-311). Now clarified (lines 336-338; and discussion 364-373 on clean revised copy) L 381-383 Is this really likely. I would predict that even unexposed prey animals would have an innate level of vigilance behaviour. – Hunter and Skinner emphasize that indeed vigilance in ungulates can fall to very low levels in the absence of direct predatory threat. However, we make a slightly broader point (explained lines 418-421). L 410 What are your ”physiological evidence”? If you mean the body condition score I think this is stretching the definition of physiology too far. – Agreed, and deleted. The Conclusions do not correspond with the Aims. The Conclusions with the sudden reasoning about welfare implications is not connected to the rest of the MS. If you want to draw conclusions about welfare you need to give the reader a context of horse welfare earlier in the MS. Welfare needs to be defined. I can, in the light of the rest of the MS reccomend the definition by Smulders 2006 which contains the different welfare aspects health, behaviour, physiology and production/reproduction. Here the body condition would fit with the production-aspects of welfare. We now introduce the applied aspects of the work more fully (lines 75-81) and have added an applied aim (line 122). We have also altered our concluding section to refer back to both fundamental and applied aims (lines 452-457). These sections now fit together in a more coherent way. We have removed reference to the broad concept of animal welfare but make brief mention that there are health consequences of obesity in horses. L 411 A “broad” body needs more than one reference. Further references have been added (lines 453-454) L 417-418 I do not agree with you that it is easy for the general horse owner to estimate either the dominance in a group of horses nor the vigilance level. A far more easy measure is the actual body condition scoring. But this is not novel and it was not the scope of your paper? I do not understand the ”novel tool with broad welfare implications”. Either re-write the text or explain the tool more clearly. We have reconsidered our conclusions. There is evidence that many horse owners are not very good at body condition scoring but this is not a key point for this paper and so we have omitted reference to this. We do consider that estimating dominance in the specific context of food competition is relatively easy and feasible for horse owners and so we have highlighted this aspect (lines 458-460). If dominant animals can be identified, then they can be managed so as to prevent the development of obesity. Reviewer 3 Line 37: definition of foraging would be appreciated to make the introduction clearer a definition of foraging has been added (line 34 on clean revised copy) Line 49: a more detailed description of the modelling framework proposed by Rands et al. in their different papers will be important to recall here. Indeed, the impact of social status on foraging that will be investigated in this manuscript is not stated in the current explanation. – a more detailed description of the modelling framework has been presented and the aspect to be investigated has been clarified (lines 52-66) Lines 87 to 93 – Aims: if a correlation is found between dominance rank and body condition (which is actually the case, see line 277) (i), the two following aims (ii and iii) need to be reconsidered. As a consequence, only one association with interruption of foraging (dominance rank or body condition*) has to be investigated contrary to the analyses that have been conducted (lines 287 to 296). We have considered this suggestion carefully and also consulted with a statistician. We have been advised that all three variables (dominance, body condition and interruption to foraging) potentially affect the other two differently. Although dominance and body condition are correlated, they may not have similar effects on interruption. As an analogy, in humans, height and weight are strongly correlated but they don’t have equal effects on a third variable such as risk of diabetes – weight is the better predictor. We therefore believe it is correct to examine the effects of both dominance and body condition, initially separately in the univariable models) and then together in the multivariable models. These analyses reveal that the relationships between body condition and interrupted foraging are not quite the same as the relationships between dominance and interrupted foraging. Body condition score has a stronger relationship with vigilance, whilst dominance rank has a stronger relationship with movements and displacements. These subtleties would be lost were we to omit the full set of statistical analyses (that take full account of correlations between variables). Experimental design - Lines 106-107: on my opinion, the sentence “The study explored the inter-relationships between foraging interruptions, dominance and body condition.” is unnecessary. agreed and removed Moreover, the next sentences lines 107 to 110) will be better placed in the “Dominance rank” paragraph (line 203) agreed and moved - Herd size between 2-10 individuals: the pressure from dominant individuals will be quite different in a group of 2 individuals (only one animal might displace the subordinate animal) than in a group with many individuals, where the pressure will be more important (a subordinate animal will be the target of more displacements/threats). This pressure might be estimated by a coefficient depending on herd size and then applied in the analysis. We agree that pressure from dominant animals will differ in small and larger groups. We therefore corrected for herd size throughout, and included herd size in all of the models explicitly as a random effect (original paper Lines 236-238; now line 263). Thus, all of the statistics we present are adjusted to take this into account and all of the associations we see occur despite differences in herd size. This important point was rather “tucked away” and so we have now mentioned it again in the first sentence of the discussion, along with the fact that herd identify was also included as a random factor (line 364). - Line 124: the range of age raises some problems: young animals are not well situated in the hierarchy of a group. Moreover, these animals are in development and their body condition/physiology is certainly different from adults. The analysis will probably gain in robustness without these young individuals. Or their case might be considered separately. Moreover, older horses have a reduced digestive efficiency (See Ralston, S.L., Squires, E.L., Nockels, C.F. (1989). Digestion in the aged horse. J Equine Vet. Sci. 8(4), 203-205) that could influence the time spent foraging. It might be interesting to investigate this factor per se. - What about the sex? If mares are pregnant or lactating, this will influence their body condition and energetic needs. More details are needed here. We thank the reviewer for making these important points about the differing energy and body condition considerations for horses with different age/sex/reproductive characteristics. The first point we make is that we did run statistical analyses to check for any effects of age or sex within our study (see original paper lines 309-311; now line 336-338) but we found no significant effects. Given the reviewer’s well-argued points on this matter, we have now added more contextual information about this to our methods and discussion. (i) We have explained that we did not include any herds containing pregnant or lactating mares in our study (lines 143-144) (ii) The lack of significant age effects may have been because of very low numbers of young horses in our study (for example, only 3 (weaned) individuals were <1 year of age) (line 149) (iii) The lack of significant age or sex effects may have been due to the sanctuary policy of keeping horses in compatible groups that would have minimised such influences. For example, all 9 stallions were housed together in one bachelor herd, whilst the 3 weaned foals were housed together with some small elderly mares in a separate small herd. This policy means that the chances of us detecting age and sex effects are greatly diminished. The housing policy of the sanctuary does, however, reflect “typical” domestic horse management where attempts are made to house similar horses together. The effects of dominance, body condition and interruptions to foraging can thus be considered as relative effects acting within herds of mainly compatible animals (herd identity included as a random factor in our analysis). An important point is that the relationships we found applied across herds housing all types of horses (added to discussion, lines 364-373). (iv) We signpost the reader to the full table of information about the age, sex, height, and herd composition of every horse in our study where data are available for further analysis (line 369). -When I read the definition of Movement whilst foraging, I’m concerned by the fact that the two origins of such movement (displacement by another individual or simply changing of location at a walk) were not distinguished for subsequent analysis. Moreover, if a Movement whilst foraging is due to a displacement by another individual, it has to be recorded as a Displacements received. If this is the case, such event is accounted in two categories (and thus twice), which is incorrect. Logically, in the results section, the authors reported that they found a correlation between the frequency of ‘displacements received’ and ‘moving whilst foraging’ (line 304). I suggest reconsidering the classification of both behaviours. The reviewer is concerned about our definition of “movement whilst foraging” and whether it might be double counted if it is due to displacements received. We have clarified this to explain that there is no double counting. We have corrected our definition of displacement received (original line 182; lines 204-208 in clean, revised copy) to remove mention of movement. We classified displacements as interactions that resulted in a head up posture or a small step away. These steps were not counted as movement. Movement whilst foraging could follow a displacement received (and this was relatively common) but it could also occur for other reasons (lines 202-203 in clean, revised copy). Dominance rank: I have no problem to infer Dominance rank from displacements emitted and received which is a classical method for measuring dominance in social groups but I don’t understand the justification (lines 212 to 217, see after in italics) that renders the understanding confusing: The number of displacements both given and received was recorded for each herd individual, and then the number of other individuals that a focal individual both dominated and was dominated by was calculated (see Appleby, 1980). Note that frequency of displacements is not the same as dominance rank. Herds may have very interactive individuals who are not particularly high or low ranking, displacing others and being displaced often, whereas the highest-ranking individual may actually displace others infrequently. Dominance rank (original lines 212-217). This was not intended as a justification, more as a general comment but we agree that it is not relevant within our own methods section. We have omitted this. Validity of Findings – I’ve previously commented some results in the experimental design section. Lines 317-318: If I well understand, the Model for adjusted dominance rank investigates whether the frequency of ‘displacements received’ and ‘displacements given’ are linked to the adjusted dominance rank. However, dominance rank was determined thanks to these two behaviours. This analysis sounds odd for me. This query arises because of terminology that we have now improved. We assessed dominance rank in a separate 3h period which involved looking at all agonistic interactions during both foraging and non-foraging periods. This was quite separate from the recording of displacements given and received during the 5min foraging bouts which was the outcome variable in the statistical models. The confusion has arisen because we mistakenly used the term “displacement” in describing the calculation of dominance rank. We now use the term “agonistic interaction” in describing how we calculated rank (lines 237-242 on clean, revised copy) and retain the term displacement for the foraging-bout observations only. It is difficult to comment this part because the concerns I’ve raised above will have consequences on the results and thus, on the interpretations. This is illustrated in the first paragraph of the Discussion. The discussion has been re-written to reflect the clarifications made in response to reviewers’ comments. Concerning the interpretation between vigilance and body condition, I think that recording vigilance when animals were not foraging to compare their rates of vigilance when they forage would have allowed to verify whether some animals were more naturally vigilant than others. Unfortunately, in the protocol, the observer stopped observing an animal when this latter stopped to forage for more than one minute. This point needs to be discussed. Vigilance outside of foraging bouts – now added as a suggestion for future study (lines 436-440). Lines 375-377: the authors wrote: Theoretical models and empirical studies have proposed increased vigilance in subordinate individuals due to more risky foraging positions (Hamilton, 1971; Hemelrijk, 2000) but this was not supported here. I could be wrong, but I didn't find nor in the methods neither in the results section any record and analysis of these risky foraging position. Line 375-377 – now clarified and added as a suggestion for future study (lines 436-440). Conclusions: On my opinion, this section is more perspectives than a conclusion about the study. Conclusions – have been rewritten. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Individual animals experience different costs and benefits associated with group living, which may impact on their foraging efficiency in ways not yet well specified. This study investigated associations between social dominance, body condition and interruptions to foraging behaviour in a cross-sectional study of 116 domestic horses and ponies, kept in 20 discrete herds. Social dominance was measured for each individual alongside observations of winter foraging behaviour. During bouts of foraging, the duration, frequency and category (vigilance, movement, social displacements given and received, scratching and startle responses) of interruptions were recorded, with total interruption time taken as a proxy measure of foraging efficiency. Total foraging time was not influenced by body condition or social dominance. Body condition was associated with social dominance, but more strongly associated with foraging efficiency. Specifically, lower body condition was associated with greater vigilance. This demonstrates that factors other than social dominance can result in stable differences in winter body condition.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Social behaviour can influence energetic reserves and subsequent body condition. Previous modelling studies have outlined the potential importance of social effects on foraging behaviour (bouts of biting, chewing and swallowing interrupted by relocation movements) in determining body condition in group living animals <ns0:ref type='bibr' target='#b26'>(Houston and McNamara, 1999;</ns0:ref><ns0:ref type='bibr' target='#b48'>Rands et al., 2003;</ns0:ref><ns0:ref type='bibr'>2004;</ns0:ref><ns0:ref type='bibr' target='#b51'>2006;</ns0:ref><ns0:ref type='bibr'>2008)</ns0:ref> and also the role of dominance behaviours in determining resource access and subsequent body condition <ns0:ref type='bibr' target='#b3'>(Clark and Ekman, 1995;</ns0:ref><ns0:ref type='bibr' target='#b56'>Stillman et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b51'>Rands et al., 2006)</ns0:ref>.</ns0:p><ns0:p>Thus, the foraging success of individual animals in social groups may be partly influenced by their social status. However, few of these predictions have been investigated empirically in socially-foraging herbivores and the relationship between herd behaviours, dominance and body condition is not fully understood.</ns0:p><ns0:p>In a socially foraging herbivore the benefits of group living outweigh the costs <ns0:ref type='bibr' target='#b30'>(Krause and Ruxton, 2002)</ns0:ref>. Individual animals living within groups follow behavioural rules which allow them to function as a social unit <ns0:ref type='bibr' target='#b21'>(Hemelrijk, 2002;</ns0:ref><ns0:ref type='bibr'>Rands, 2011a,b)</ns0:ref>. These rules are likely to depend upon both aspects of their own body condition (such as energetic reserves) and also the actions of other individuals within the group <ns0:ref type='bibr' target='#b26'>(Houston and McNamara, 1999;</ns0:ref><ns0:ref type='bibr' target='#b48'>Rands et al., 2003;</ns0:ref><ns0:ref type='bibr'>2008)</ns0:ref>. Rules governing social interaction (e.g. dominance) may be important for a wellfunctioning group in terms of minimising costly conflict over resources <ns0:ref type='bibr' target='#b30'>(Krause and Ruxton, 2002)</ns0:ref>. <ns0:ref type='bibr' target='#b47'>Rands et al. (2011b)</ns0:ref> considered a game theoretical framework to explore how the rules used by individuals with different dominance ranks could evolve, assuming these individuals paid attention to the ranks and energetic state of both themselves and the individual that they were interacting with. This model, and a companion simulation exploring the rules of thumb generated <ns0:ref type='bibr' target='#b46'>(Rands 2011a</ns0:ref>) demonstrated that both energetic state and social status are important for determining the behaviour of co-foraging individuals. Furthermore, individual-based simulations <ns0:ref type='bibr' target='#b50'>(Rands et al 2004</ns0:ref><ns0:ref type='bibr' target='#b51'>(Rands et al , 2006</ns0:ref>) demonstrated that including an additional effect of dominance that led to subordinates having reduced access to food could lead not only to dominant individuals PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020) Manuscript to be reviewed having higher energetic reserves than subordinates, but also subordinate individuals increasing their activity.</ns0:p><ns0:p>We aimed to assess whether this framework was useful in understanding the foraging behaviour of the horse. We were particularly interested to determine whether dominant animals had higher body condition and whether subordinate individuals showed increased activity in line with model predictions. Horses are generalist herbivores with sophisticated social capacities. Free-ranging feral and primitive Przewalksi's horses spend a high proportion of each day foraging (52%, <ns0:ref type='bibr' target='#b2'>Berger et al., 1999;</ns0:ref><ns0:ref type='bibr'>68% Lamoot and Hoffman 2004</ns0:ref>; up to 75% daylight and 53% nocturnal, <ns0:ref type='bibr' target='#b38'>Mayes and Duncan, 1986</ns0:ref>)maintaining a high daily intake of plant material by grazing (or browsing) interrupted by frequent walking <ns0:ref type='bibr' target='#b24'>(Houpt, 2005)</ns0:ref>. Accelerometry studies find similar proportions of time spent foraging by domestic horses kept on pasture (61% daylight, 47% nocturnal, <ns0:ref type='bibr' target='#b32'>Maisonpierre et al., 2019)</ns0:ref>. Horses form strong affiliative bonds with familiar companions, but aggressive encounters and subtle threats, are also a common feature of equine social structure, particularly when resources are limited <ns0:ref type='bibr'>(Mills and Redgate, 2010)</ns0:ref>. The current study was conducted under winter conditions where pasture availability was limited and a degree of competition for supplementary forage was evident. Thi situation applies commonly for domestic horses (kept for a variety of reasons including as companion animals or as conservation grazers <ns0:ref type='bibr' target='#b11'>(Gilhaus and Hoelzel, 2016)</ns0:ref> during winter periods within temperate zones).</ns0:p><ns0:p>Understanding the factors that drive large inter-individual differences in body condition when group-living horses are kept during winter (e.g. <ns0:ref type='bibr' target='#b28'>Ing&#243;lfsd&#243;ttir and Sigurj&#243;nsd&#243;ttir, 2008;</ns0:ref><ns0:ref type='bibr' target='#b13'>Giles et al., 2015;</ns0:ref><ns0:ref type='bibr'>Yngvesson et al., 2019)</ns0:ref> is an important goal. It has been estimated that around a third of outdoor living horses and ponies within the UK are obese <ns0:ref type='bibr' target='#b12'>(Giles et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b54'>Robin et al., 2015)</ns0:ref> but rates of obesity can reach 70% in some populations <ns0:ref type='bibr' target='#b39'>(Menzies-Gow et al., 2017)</ns0:ref>. It is timely to study the social factors influencing body condition in horses to reduce obesity prevalence and associated metabolic disease.</ns0:p><ns0:p>Previous empirical studies in horses have demonstrated that higher ranking individuals spend more time eating hay and have a higher body condition during the winter <ns0:ref type='bibr' target='#b28'>(Ing&#243;lfsd&#243;ttir and Sigurj&#243;nsd&#243;ttir, 2008;</ns0:ref><ns0:ref type='bibr' target='#b13'>Giles et al., 2015)</ns0:ref> but have not examined the mechanisms behind this association.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>This study advanced our previous work by examining situations where bouts of foraging on supplementary forage were interrupted for reasons including anti-predator vigilance and startle responses <ns0:ref type='bibr' target='#b14'>(Goodwin, 1999)</ns0:ref>, displacement interruptions directed towards or received from other group members <ns0:ref type='bibr' target='#b0'>(Appleby, 1980;</ns0:ref><ns0:ref type='bibr' target='#b51'>Rands et al., 2006)</ns0:ref> or short movements between foraging locations <ns0:ref type='bibr' target='#b6'>(Duncan, 1980)</ns0:ref>. We examined the duration, frequency and type of interruption to the foraging behaviour of individual horses and ponies (hereafter termed 'horses') living in social herds. The total time attributed to interrupted foraging was considered as a proxy measure of foraging efficiency (the ratio of energy gained over energy expended during foraging).</ns0:p><ns0:p>An important precursor to analysing foraging efficiency was understanding any differences in overall time spent foraging. We measured overall time spent foraging to check that individuals with a lower foraging efficiency didn't simply compensate by spending more time foraging. A unique feature of the study was the inclusion of measures of social status and body condition, enabling the assessment of associations not previously examined in foraging herbivores.</ns0:p><ns0:p>Predictions suggest that subordinate individuals may suffer more displacement than dominant conspecifics <ns0:ref type='bibr' target='#b16'>(Goss-Custard et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b56'>Stillman et al., 1997;</ns0:ref><ns0:ref type='bibr' target='#b20'>2000;</ns0:ref><ns0:ref type='bibr' target='#b51'>Rands et al., 2006)</ns0:ref>, reflected in increased displacement interactions and subsequent movement within foraging bouts. Dominant animals may also force subordinate conspecifics into more exposed foraging positions <ns0:ref type='bibr' target='#b7'>(Ekman, 1987;</ns0:ref><ns0:ref type='bibr' target='#b50'>Rands et al., 2004)</ns0:ref> leading to a reduction in foraging efficiency due to a greater requirement for vigilance. In contrast, models predict that dominant individuals will be more efficient foragers, feeding in positions with lower interference, potentially leading to a greater energetic intake and overall body condition <ns0:ref type='bibr' target='#b7'>(Ekman, 1987;</ns0:ref><ns0:ref type='bibr' target='#b55'>Schneider, 1984;</ns0:ref><ns0:ref type='bibr' target='#b51'>Rands et al., 2006)</ns0:ref>.</ns0:p><ns0:p>A greater body condition may in turn allow a subsequent competitive advantage <ns0:ref type='bibr'>(Rands, 2011;</ns0:ref><ns0:ref type='bibr' target='#b51'>Rands et al., 2006)</ns0:ref>. Our aims were to: i) Confirm an association between dominance rank (adjusted for herd size, see Methods) and body condition. ii)</ns0:p><ns0:p>Assess whether adjusted dominance rank is associated with interruptions to foraging (as a proxy for foraging efficiency).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed iii) Assess whether body condition is associated with interruptions to foraging (as a proxy for foraging efficiency). iv)</ns0:p><ns0:p>Use multivariate analysis to investigate the contextual factors (age, breed, sex, height, supplementary feeding) that might influence these associations. v)</ns0:p><ns0:p>Consider the applied implications of our findings for the management of domestic horses. We predicted that foraging interruptions would be associated with both body condition and dominance status, and that subordinate individuals would, overall, have a reduced foraging efficiency compared with more dominant conspecifics and a lower body condition, as indicated in a previous study <ns0:ref type='bibr' target='#b13'>(Giles et al., 2015)</ns0:ref>. This study goes beyond previous research to assess whether differences in foraging efficiency could plausibly be the mechanism linking dominance to body condition.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>a) Animals and Ethical Statement</ns0:head><ns0:p>The work was approved by the University of Bristol Animal Welfare and Ethical Review Board (University Investigation Number UB/10/049) and all methods were carried out in accordance with relevant guidelines.</ns0:p><ns0:p>The study sample was drawn from a population of outdoor, group-living horses based at Redwings Horse Sanctuary (UK), that had been living together for at least three months and had established social relationships. All of the individual animals were managed similarly, fed forage from identical sources, lived in outdoor environments and were not ridden, meaning that structured exercise could be removed as a potential confounding factor. Herds that included pregnant or lactating mares were not considered for the study. Twenty study herds were selected randomly from all remaining suitable herds within the sampling frame.</ns0:p><ns0:p>The policy of the sanctuary was to house horses in relatively compatible groups with shared characteristics. Thus, larger horses were housed in separate herds from smaller ponies, all stallions were housed in one 'bachelor' herd, while youngsters were also housed together, with the few horses under 1 year of age (three individuals) accompanied by older 'nanny' mares.</ns0:p><ns0:p>Herd size was 2-10 (mean 6 &#177; 0.56 individuals). 116 individuals (84 ponies of height &lt;148cm, and 32 horses of height &#8805; 148cm) from within these herds were observed between 2 December, 2013 and 23 January, 2014. Ages ranged from 5 months to 32 years (11.83 &#177; 0.63 years). Breeds were native ponies (51.72%), native cobs (17.24%), lightweight horses (12.07%), heavy horses (5.17%), sports horse breeds (5.17%) and other (8.62%).</ns0:p></ns0:div> <ns0:div><ns0:head>b) Study period and horse management</ns0:head><ns0:p>The winter months were chosen for observation as natural food resources were at their minimum and therefore food based social interactions were likely at their highest due to the close proximity of individuals. All horses lived in an outdoor paddock environment for 24 hours a day and were fed from circular hay feeders provided at a fixed ratio of feeder space (30cm) per animal. Horses were fed twice daily with fresh hay replenished once at the start of morning observation (between 08:00 and 09:00) and once at the start of afternoon observation (between 11.30 and 13:00). Any uneaten hay remained in the hay feeder throughout the day. Twelve study horses received additional supplementary feed from a bucket once a day, and this was recorded as a potential confounder.</ns0:p></ns0:div> <ns0:div><ns0:head>c) Time spent foraging</ns0:head><ns0:p>Each study herd was observed for six hours to assess overall time spent foraging, and interruptions occurring during foraging bouts, once during a three hour morning session (08:00-09:00 until 11:00-12:00) and once during a three hour afternoon session (11:30-13:00 until 14:30-16:00) on a different day within the same week, by a single trained observer. Due to the time of year, these times were chosen based on daylight hours. Time spent foraging was recorded using scan sampling at five minute intervals throughout each three-hour observation period. A random number generator was used to determine the order in which individuals were observed. Once this order was determined, all individuals were observed PeerJ reviewing <ns0:ref type='table' target='#tab_2'>PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:ref> Manuscript to be reviewed in sequence, in five-minute intervals. At each interval, it was recorded which individuals were foraging and which were not. Foraging was defined as the horse ingesting either hay or grass, with intermittent periods of the head down ingesting forage and the head up chewing this forage material. The horse could be foraging from either the hay feeder or eating grass (although the latter was rare as there was little grass available). The percentage of time spent foraging was then calculated based on the number of intervals that each individual was foraging within the full six hours of observation per herd.</ns0:p><ns0:p>Alongside this, continuous five minute focal animal observations were scheduled for each horse during each three hour recording period. Each individual animal was independently observed for at least 20 minutes (4 &#215; 5-minutes) in total. These observations were predominantly used to record foraging interruptions and social interactions (as detailed in sections d and e below), however they were also used to more accurately estimate the total foraging time for each individual. If an individual was not foraging for more than one minute during the five-minute observation period, it was considered to have stopped foraging. The number of minutes it had stopped foraging for were then subtracted from the total five minutes.</ns0:p></ns0:div> <ns0:div><ns0:head>d) Foraging efficiency -duration and frequency of foraging interruptions</ns0:head><ns0:p>During the continuous five-minute focal animal observations, described above, observations relating to foraging interruptions were also conducted. Interruption to foraging was defined as an activity that was short in duration (less than one minute) and prevented the individual from selecting, biting or chewing hay or grass. Both the frequency and overall duration of any interruption was recorded and interruptions were categorised as one of the following:</ns0:p><ns0:p>Vigilance: Head raised from foraging and ears pricked in the direction of interest, the head is higher and the ears upright distinguishing vigilance from raising the head to chew. Manuscript to be reviewed Displacements received: interaction received from another individual defined as above, causing recipient to raise head, move sideways or take a step away in any direction.</ns0:p></ns0:div> <ns0:div><ns0:head>Movement whilst foraging</ns0:head><ns0:p>Scratching: Using either the mouth or the hoof to scratch the body Startle response: A quick reaction to an unexpected stimulus, the startle usually involved a quick movement, either jump backwards or sideways followed by looking up with ears pricked If any interruption lasted for over one minute then the individual was classed as having stopped foraging. Note that individuals were only observed in detail when they were foraging, if an individual was not foraging when it was due to be observed, this was recorded (to calculate total foraging time, as described in section a) and but also counted as 'missed' in terms of recording interruptions. Once a missed individual was foraging again it was observed next as a priority (only if it had not yet already been observed for 20 minutes), but just for a single five-minute interval, before resuming the original order. This was to maximise the collection of data on foraging efficiency for each individual.</ns0:p><ns0:p>The frequency of foraging interruption (a proxy for foraging efficiency) was calculated as the number of instances of all interruptions per minute foraging. Separate frequencies were also determined for each interruption category (Table <ns0:ref type='table'>1</ns0:ref>). The duration of interrupted foraging referred to the total percentage of time spent interrupted per individual.</ns0:p></ns0:div> <ns0:div><ns0:head>e) Dominance rank</ns0:head><ns0:p>Although the concept of dominance lacks universal explanatory power in describing social structure, it is a useful construct when considering the specific context of competition for a limited food resource. Under such conditions, horses generally follow a linear ranking hierarchy, with occasional triangles and some influence of third-party interactions <ns0:ref type='bibr' target='#b25'>(Houpt et al., 1978;</ns0:ref><ns0:ref type='bibr'>van Dierendonck et al., 1995;</ns0:ref><ns0:ref type='bibr' target='#b18'>Hartmann et al., 2017</ns0:ref>).</ns0:p><ns0:p>Here we defined dominance 'an asymmetry in the outcome of dyadic interactions between individuals, or a priority of access to resources' <ns0:ref type='bibr' target='#b5'>(Drews, 1993)</ns0:ref> Manuscript to be reviewed outcomes between dyadic pairs when feeding from hay feeders. Agonistic interactions were recorded continuously throughout the three-hour observation period (these were easily measurable alongside other observations). An agonistic interaction was defined as one individual approaching or displaying to another with the neck outstretched and ears back flat against the head and, crucially, the second individual moving away. Dominance rank was then calculated using the methods described by <ns0:ref type='bibr' target='#b0'>Appleby (1980)</ns0:ref>. The number of agonistic interactions both given and received was recorded for each herd individual, and then the number of other individuals that a focal individual both dominated and was dominated by was calculated.</ns0:p><ns0:p>Once an Appleby rank had been given, this was then adjusted to take into account herd size (as in <ns0:ref type='bibr' target='#b13'>Giles et al., 2015)</ns0:ref>. Adjusted dominance rank was calculated as 1 -(a -1)/(h -1), where a is the Appleby rank and h is the herd size. Where dominance rank or dominance status is referred to in this manuscript, this refers to this adjusted dominance rank.</ns0:p><ns0:p>f) Body condition score Measurements were taken immediately after the second set of observations on the herd had been completed. All study animals were accustomed to being handled. Body condition score was measured using the Henneke nine-point scale <ns0:ref type='bibr' target='#b22'>(Henneke et al., 1983</ns0:ref>) by a single trained observer (SLG). Six areas of the horse were scored between 1 and 9 and then averaged and rounded to the nearest 0.5, to obtain a single score. A score of five on the scale was taken to indicate an ideal body condition.</ns0:p></ns0:div> <ns0:div><ns0:head>g) Statistical analyses</ns0:head><ns0:p>Results were analysed using Stata 12.1 (Statacorp, Texas). Univariable relationships were assessed using mixed effects linear regression, the clustered study design was controlled for by including herd group and herd size as a random effects, on the basis that herd size or other herd specific factors such as environment could plausibly have some influence on foraging and interactive behaviours. Univariable relationships of primary interest were:</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed 1) The relationship between dominance rank (adjusted for herd size) and body condition score 2) The relationship between dominance rank (adjusted for herd size) and interruptions to foraging (as a proxy for foraging efficiency) 3) The relationship between body condition and interruptions to foraging (as a proxy for foraging efficiency)</ns0:p><ns0:p>Following an initial univariable exploration of these relationships, relationships between the separate foraging interruption variables were also considered. In addition, breed, age, height, sex and whether or not the individual received supplementary feed were recorded as potential confounding variables. To be considered a potential confounder the variable had to be associated with both the explanatory and outcome variable, and not on the causal pathway between the two <ns0:ref type='bibr' target='#b44'>(Petrie and Sabin, 2009)</ns0:ref>. Statistical significance was defined using p&#8804;0.05 with a screening pvalue for multivariable models of p &#8804; 0.07.</ns0:p><ns0:p>Mixed effects multivariable linear regression was then used to build a best-fit explanatory model for both adjusted dominance rank and body condition. The foraging interruption variables (see Table <ns0:ref type='table'>1</ns0:ref> for list) were added to the model one at a time, based on the strength of univariable association, starting with a minimal model. A likelihood ratio test was used to assess the contribution of each variable to the model fit and variables were retained on the basis of this and the adjusted p value.</ns0:p><ns0:p>Multivariable analysis using a mixed effects linear regression model was also used to make predictions regarding interruptions to foraging -to explore whether this could be a possible mechanism linking dominance status and body condition. Duration of foraging interruption was associated with both dominance status and body condition, therefore this was added to a model containing adjusted dominance rank and body condition. Its explanatory contribution to the model was then assessed using both the adjusted p and estimates and a likelihood ratio test.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed During 120h of observation, the amount of time that individual animals spent foraging averaged 76.4% SD 0.17. Values per herd are given in Table <ns0:ref type='table'>S1</ns0:ref>. Figure <ns0:ref type='figure' target='#fig_5'>1</ns0:ref> shows that there was no significant correlation between adjusted dominance rank and total foraging time (r 2 = 0.004, n = 116, p = 0.51) and Figure <ns0:ref type='figure' target='#fig_6'>2</ns0:ref> shows that there was no significant correlation between body condition score (range 4 to 8.5) and total foraging time (r 2 = 0.016; n = 116, p = 0.182). This is important in the interpretation of subsequent results.</ns0:p></ns0:div> <ns0:div><ns0:head>a) Univariable Analysis</ns0:head><ns0:p>The relationship between adjusted dominance rank and body condition score</ns0:p><ns0:p>Adjusted dominance rank was positively associated with body condition score within our study population (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Foraging Efficiency</ns0:head><ns0:p>During approximately 92h of the 120h total observation period, horses were foraging (total across all horses). During this time, the observed total numbers of each type of interruption contributing to foraging efficiency were: vigilance 2518; movement whilst foraging 454; displacements given 198; displacements received 222; scratching 65; startle responses 5.</ns0:p></ns0:div> <ns0:div><ns0:head>The relationship between dominance rank and foraging efficiency</ns0:head><ns0:p>Although the frequency of foraging interruptions did not show evidence of association with adjusted dominance rank (Z=-1.55, p=0.12, Table <ns0:ref type='table'>S2</ns0:ref>), the total duration of interruptions decreased as adjusted dominance rank increased (Table <ns0:ref type='table'>1</ns0:ref>). An increase in adjusted dominance rank was also associated with a decrease in some specific interruption behaviours, namely instances of movement whilst foraging, displacements given, and displacements received (Table <ns0:ref type='table'>1</ns0:ref>). Figure <ns0:ref type='figure' target='#fig_5'>1</ns0:ref> shows that the reduced foraging efficiency of subordinate individuals is not compensated for by an increase in total foraging time.</ns0:p></ns0:div> <ns0:div><ns0:head>The relationship between body condition score and foraging efficiency</ns0:head><ns0:p>The number of incidences (frequency) of foraging interruptions occurring during foraging bouts was lower for animals with higher body condition scores. Vigilance decreased with an increase</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed in body condition (Table <ns0:ref type='table'>1</ns0:ref>), but none of the other separately defined foraging interruptions showed any association with body condition (Supplementary Information, Table <ns0:ref type='table'>S2</ns0:ref>). Figure <ns0:ref type='figure' target='#fig_5'>1</ns0:ref> shows that the reduced foraging efficiency of individuals with lower body condition is not compensated for by an increase in total foraging time.</ns0:p></ns0:div> <ns0:div><ns0:head>Associations between the individual foraging interruption variables and consideration of potential confounders</ns0:head><ns0:p>Frequency of 'displacements received' was strongly associated with 'moving whilst foraging' and 'displacements given'. Frequency of 'displacements given' was also associated with 'moving whilst foraging' (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>In this study, none of the potential confounder variables (breed, age, height, sex) were associated with body condition score, adjusted dominance rank or any category of interrupted foraging, and there were no biologically plausible interactions, therefore adjusted estimates were not required.</ns0:p><ns0:p>This also included whether or not a horse received additional supplementary feed, which showed no evidence of association with either adjusted dominance rank (Z = -0.50, p = 0.61) or body condition ( = 12.40, p = 0.19).</ns0:p><ns0:p>&#120568; 2 9 b) Multivariable analysis</ns0:p></ns0:div> <ns0:div><ns0:head>Model for adjusted dominance rank</ns0:head><ns0:p>Controlling for other model variables, frequency of 'displacements received', 'displacements given' and body condition score were associated with adjusted dominance rank (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Model for body condition score</ns0:head><ns0:p>Controlling for other model variables, vigilance frequency and adjusted dominance rank were strongly associated with body condition score (Table <ns0:ref type='table' target='#tab_0'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The relationship between body condition score and adjusted dominance rank when taking into account interruptions to foraging</ns0:head><ns0:p>The association between body condition score and adjusted dominance rank was weaker when total duration of foraging interruptions (or time spent interrupted) was included in the model (Table <ns0:ref type='table' target='#tab_1'>4</ns0:ref>, p = 0.06, as opposed to p = 0.03 in the univariable model). The effect size also reduced Manuscript to be reviewed slightly (from a 0.66 increase in adjusted dominance rank per half unit of body condition score to 0.55). The likelihood ratio test results (Table <ns0:ref type='table' target='#tab_1'>4</ns0:ref>) indicate that duration of foraging interruptions has a more significant contribution to the model fit (p = 0.04) than adjusted dominance rank (p = 0.06).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The study explored the inter-relationships between foraging interruptions, dominance and body condition, controlling for herd size and herd identity effects. No effects of age, sex or height were detected in our study. Clearly, large horses have differing energy requirements from smaller ponies, whilst growing youngsters and older horses with reduced digestive efficiency (e..g <ns0:ref type='bibr'>Ralston et al., 1989)</ns0:ref> will also differ from young but mature adults. However, the horses in our study were housed in herds that contained animals of similar characteristics (see Methods and Supplementary Table <ns0:ref type='table'>)</ns0:ref>. For example, heavy horses were housed separately from lighter Thoroughbreds and smaller ponies. Although this policy greatly reduces or eliminates our ability to detect age and sex effects on foraging, it enhances our ability to detect the relative effects of dominance and body condition within herds. Importantly, our analysis showed that the relationships we detected applied across all herd types.</ns0:p><ns0:p>Within this study population, dominance status was positively associated with body condition, although this relationship was weaker when foraging efficiency was included in the multivariate model (Table <ns0:ref type='table' target='#tab_1'>4</ns0:ref>). In addition, the association between body condition and foraging efficiency was stronger than that between body condition and dominance. Thus, whilst dominance explains some variation in body condition, our results highlight the potential role of factors other than social dominance that could influence foraging efficiency. Factors such as a tendency to show vigilance behaviour have been little explored to date but have the potential to greatly influence the ratio of energy gained vs energy expended during bouts of foraging.</ns0:p><ns0:p>There was no evidence that subordinate or low body condition individuals compensated for less efficient foraging by increasing total foraging time. Another recent study found that horses with low body condition tend to adopt more passive behaviour <ns0:ref type='bibr' target='#b29'>(Jorgensen et al., 2016)</ns0:ref>. Potentially</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed such results may be due to a strong motivation to feed as a group in this species and thus synchronise feeding and resting behaviour <ns0:ref type='bibr'>(Rands et al., 2008)</ns0:ref>. Subordinate or lower body score individuals were unlikely to remain foraging when conspecifics were not, supporting suggestions that social factors may result in stable differences in body condition within group living animals <ns0:ref type='bibr'>(Rands, 2011;</ns0:ref><ns0:ref type='bibr'>Rands et al., 2010)</ns0:ref>. Indeed the tendency to synchronous feeding and resting (as in sheep, <ns0:ref type='bibr' target='#b35'>McDougall and Ruckstuhl, 2018</ns0:ref>) may be hard-wired as an adaptivebehaviour.</ns0:p><ns0:p>The lack of a compensatory change in total foraging time means that any variation observed in foraging efficiency could plausibly have an effect on body condition.</ns0:p><ns0:p>Given these results and previous theoretical predictions, an association between foraging efficiency, dominance and overall body condition was expected <ns0:ref type='bibr' target='#b36'>(McNamara and Houston, 1990;</ns0:ref><ns0:ref type='bibr' target='#b57'>Stillman et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b51'>Rands et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b53'>Rands and Whitney, 2008)</ns0:ref> but our study is the first to explore the role of the different components of foraging efficiency, such as movement, social displacement or vigilance.</ns0:p></ns0:div> <ns0:div><ns0:head>Vigilance and body condition</ns0:head><ns0:p>Vigilance frequency was the individual interruption behaviour most strongly associated with body condition score -it showed a strong negative association. However, vigilance was not associated with dominance status. These results suggest that certain individuals may be more likely to conduct vigilance, perhaps on behalf of the group, regardless of their social status.</ns0:p><ns0:p>These results do seem to support the suggestion that vigilance is an inherently costly activity <ns0:ref type='bibr' target='#b8'>(Elgar, 1989;</ns0:ref><ns0:ref type='bibr' target='#b10'>Fritz et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b9'>Fattorini and Ferretti, 2019;</ns0:ref><ns0:ref type='bibr'>Pacheco and Herrera, 1999)</ns0:ref> as demonstrated by the negative association with body condition. However, lower body condition individuals may also be more stressed or nervous individuals, which would also explain the association with increased vigilance.</ns0:p><ns0:p>The complexity of vigilance as a single trait may somewhat explain the lack of observed association with dominance status. Vigilance may serve a range of functions in group living animals <ns0:ref type='bibr' target='#b9'>(Fattorini and Ferretti, 2019)</ns0:ref>, including anti-predatory behaviour <ns0:ref type='bibr' target='#b8'>(Elgar, 1989;</ns0:ref><ns0:ref type='bibr' target='#b27'>Hunter and Skinner, 1998)</ns0:ref>, monitoring of other herd members and scanning the environment for Manuscript to be reviewed resources <ns0:ref type='bibr' target='#b60'>(Underwood, 1982)</ns0:ref>. Ungulate mammals that are unexposed to predation have been observed to greatly reduce their vigilance behaviour <ns0:ref type='bibr' target='#b27'>(Hunter and Skinner, 1998)</ns0:ref>. Horses, unexposed to predation, may therefore show relatively low levels of vigilance, with reasons other than anti-predatory vigilance having a proportionally larger role.</ns0:p><ns0:p>Alongside the association between dominance status and body condition, the association between body condition and vigilance provides evidence of two separate behavioural traits associated with body condition in group living animals. Behavioural predictors of body condition have so far received little attention in horses (for exceptions, see <ns0:ref type='bibr' target='#b28'>Ing&#243;lfsd&#243;ttir and Sigurj&#243;nsd&#243;ttir, 2008;</ns0:ref><ns0:ref type='bibr' target='#b13'>Giles et al., 2015)</ns0:ref> and may warrant continued investigation, especially as obese horses (BCS &gt;7) may show differences in activity and eating behaviour when compared to lean horses (BCS 4-5) <ns0:ref type='bibr' target='#b41'>(Moore et al., 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Dominance status, movement during foraging and displacement interactions</ns0:head><ns0:p>Subordinate horses showed more movement whilst foraging, and were (as expected) more likely to receive displacements. Indeed, statistical analysis revealed that displacement was strongly associated with movement during foraging in our study population, with subordinate animals forced to move foraging location. Theoretical models and empirical studies have proposed that subordinate individuals may be forced to foraging positions carrying a greater risk of predation <ns0:ref type='bibr' target='#b17'>(Hamilton, 1971;</ns0:ref><ns0:ref type='bibr' target='#b20'>Hemelrijk, 2000)</ns0:ref>. Future studies could examine whether subordinate animals showed increased vigilance specifically when in displaced locations, and during non-foraging periods.</ns0:p><ns0:p>Overall our results therefore appear to support predictions that displacement reduces foraging efficiency for the recipient <ns0:ref type='bibr' target='#b1'>(Bautista et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b58'>Stillman et al., 2002)</ns0:ref>. Valuable foraging time is wasted not only over the initial dispute, but also in relocating to a new foraging location. In contrast, dominant horses tended to interrupt their own foraging to displace others, but these interruptions tended to be of short duration, allowing the dominant animal to return quickly to foraging. As our study herds were feeding from hay feeders, potentially displacement and Manuscript to be reviewed movement occurred more often than would occur during foraging on pasture, due to the artificially close proximity of herd members <ns0:ref type='bibr'>(Hoffman et al., 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>These results are novel and exciting in that they present the first behavioural evidence confirming a broad body of influential theoretical work (e.g. <ns0:ref type='bibr'>Marshall et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b43'>Petit and Bon, 2010;</ns0:ref><ns0:ref type='bibr' target='#b48'>Rands et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b51'>2006;</ns0:ref><ns0:ref type='bibr'>Rands 2011;</ns0:ref><ns0:ref type='bibr' target='#b59'>Sueur et al., 2013)</ns0:ref> linking condition and behaviour in a group-living species. Our results suggest (in line with model predictions) that differences in energetic reserves (body condition) can emerge simply via a reduction in energetic intake by subordinates when dominants are present. This hypothesis could be further tested in a future prospective study. One application of our work is that information on individual horse dominance status could be included as a relevant factor when addressing health problems associated with equine obesity <ns0:ref type='bibr' target='#b12'>(Giles et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b54'>Robin et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b39'>Menzies-Gow et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Table <ns0:ref type='table'>1</ns0:ref>. Statistically significant univariable associations (p &#8804; 0.05) using mixed effects linear regression, controlling for herd group and herd size as a random effects. Non-significant associations are given in the supplementary material, Table <ns0:ref type='table'>S2</ns0:ref>. Manuscript to be reviewed Multivariable linear regression model showing the effect of foraging efficiency (total duration of foraging interruptions) upon the relationship between dominance status and body condition.</ns0:p><ns0:formula xml:id='formula_0'>Interruption</ns0:formula><ns0:p>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>: a short movement resulting in a change in foraging location, either following a displacement by another individual or simply changing location at a walk. Displacements given: interaction directed towards another individual, with the head outstretched and ears flat back against the head resulting in recipient raising head, or taking a step away in any direction. PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>and assessed it by measuring PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>The final multivariable explanatory model for body condition score, using mixed effects linear regression, controlling for herd group and herd size as random effects.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>behaviour variables</ns0:cell><ns0:cell>&#946;</ns0:cell><ns0:cell>S.E.</ns0:cell><ns0:cell>95% CI</ns0:cell><ns0:cell>Z</ns0:cell><ns0:cell>p</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Multivariable linear regression model showing the effect of foraging efficiency (total duration of foraging interruptions) upon the relationship between dominance status and body condition.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Likelihood Ratio Test</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:02:46192:2:0:NEW 11 Oct 2020)</ns0:note></ns0:figure> <ns0:note place='foot'>Netherlands Journal of Zoology 45: </ns0:note> </ns0:body> "

"Dear Editor Many thanks for allowing us the opportunity to make further minor improvements to our manuscript. Your help is very much appreciated. Our response to the minor revisions requested is outlined below. Editor Comments Agreement with reviewer that ….the aims are very clear and defined whereas the conclusions are more speculative and include a statement on the obesity issue in pets, horses and humans. We have removed the more speculative parts of the conclusions. However, one of our stated aims was to consider potential applied implications and so this is now mentioned more briefly. line 239: two words appear to be unintentionally joined here. *These have been separated. Reference list: there are a few places where the author name has not been bolded - please correct this. *Corrected Tables - please ensure you use the same level of precision (number of decimal places) within each column (i.e. your p-values) *Corrected – revised tables have been uploaded. Figures - please present axis labels without underscores, and indicate units of measurement (e.g. Foraging_time should be Foraging time (mins)). Figure headings are not required, so please remove these. Figure captions should be stand-alone (e.g. can be interpreted by someone who has not read your article) so please revise the figure captions so they are more informative. *The figures have been redrawn to specify that foraging time is presented as a proportion, labels without underscores, headings removed and captions more informative. Reviewer Comments L 67 the reference by Ellis (2010) does not support your stating in the text that free living horses spend from 9 h foraging. You need to cite the original reference and give the correct results. Ellis mentions 10-14 hours and (though difficult to see from the abstract) does not seem to be the original source but to be reviewing other research. L 68 the end of the same sentence lack references. *Reference to Ellis has been removed and original source references are now presented on lines 67 to 72. L 72 At the end of the line something is missing. L 160 What was the fixed feeder space/ horse? *Missing information added. L309 Interesting result, they forage 77% of the observed time. I guess the 92 h is the total observed foraging time for all horses observed? *This is correct – we have added “total” for clarity L 311-312 Are these figures really frequencies? Or are they the total number of observed events? To me frequencies are figures by something e.g. time or number of horses, though I could be wrong here. *“Total number” is better – changed on line 314 Tinbergian context or at least add words like ”adaptive” or ”evolutionary”. * “adaptive” added line 396 I couldn’t see if you had tested for a relationship between vigilance and group size. There is a paper on feral horses https://www.jstor.org/stable/1382634?origin=crossref&seq=1#metadata_info_tab_contents and there seems to be an effect of group size on vigilance for other species *We did test for a relationship between vigilance and group size (this was reported in manuscript, lines 266-269). Herd identity and herd group size were both included in all statistical models as random effects. We have added the reference suggested by the reviewer on effects of group size on vigilance in horses. We had already included references to general effects of group size on vigilance (e.g. Elgar, 1989). "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: Altered glycosylation of proteins contributes to tumor progression. Dolichol phosphate mannose synthase (DPMS), an essential mannosyltransferase, plays a central role in post-translational modification of proteins, including N-linked glycoproteins, O-mannosylation, C-mannosylation and glycosylphosphatidylinositol anchors synthesis. Little is known about the function of DPMS in liver cancer. Methods: The study explored the roles of DPMS in the prognosis of hepatocellular carcinoma using UALCAN, Human Protein Atlas, GEPIA, cBioPortal and Metascape databases. The mRNA expressions of DPM1/2/3 also were detected by quantitative real-time PCR experiments in vitro. Results:The transcriptional and proteinic expressions of DPM1/2/3 were both over-expressed in patients with hepatocellular carcinoma. Over expressions of DPMS were discovered to be dramatically associated with clinical cancer stages and pathological tumor grades in hepatocellular carcinoma patients. In addition, higher mRNA expressions of DPM1/2/3 were found to be significantly related to shorter overall survival in liver cancer patients. Futhermore, high genetic alteration rate of DPMS (41%) was also observed in patients with liver cancer, and genetic alteration in DPMS was associated with shorter overall survival in hepatocellular carcinoma patients. We also performed quantitative real-time PCR experiments in human normal hepatocytes and hepatoma cells to verify the expressions of DPM1/2/3 and results showed that the expression of DPM1 was significantly increased in hepatoma cells SMMC-7721 and HepG2.</ns0:p><ns0:p>Conclusions: Taken together, these results suggested that DPM1 could be a potential prognostic biomarker for survivals of hepatocellular carcinoma patients.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Hepatocellular carcinoma (HCC) is one of the most frequently and commonly occurring malignant tumors worldwide. The global incidence and mortality rate of HCC are ranked 5th and 3rd among all types of cancers <ns0:ref type='bibr' target='#b0'>(1,</ns0:ref><ns0:ref type='bibr' target='#b1'>2)</ns0:ref>. Despite making remarkable advances in new technologies for diagnosis and treatment, the incidence and mortality of HCC still continue to growth because of the poorest prognosis <ns0:ref type='bibr' target='#b2'>(3,</ns0:ref><ns0:ref type='bibr' target='#b3'>4)</ns0:ref>. Therefore, it is urgently needed to determine reliable predictive biomarkers for early diagnosis and accurate prognosis, and to develop new molecular targeted therapeutic strategies.</ns0:p><ns0:p>The occurrence and development of several cancer types are closely associated with aberrant protein glycosylation <ns0:ref type='bibr' target='#b5'>(5,</ns0:ref><ns0:ref type='bibr' target='#b6'>6)</ns0:ref>. Studies have suggested that altered glycosylation of proteins has been observed in liver cancer <ns0:ref type='bibr' target='#b7'>(7)</ns0:ref>. Although mounting evidence has reported the role of glycosylation in tumor progression <ns0:ref type='bibr' target='#b8'>(8)</ns0:ref><ns0:ref type='bibr' target='#b9'>(9)</ns0:ref><ns0:ref type='bibr' target='#b10'>(10)</ns0:ref>, there is limited information on how glycosylation affects the liver cancer development. Recent studies have focused on glycosylation crosstalks with cellular metabolism and related kinases <ns0:ref type='bibr' target='#b11'>(11)</ns0:ref><ns0:ref type='bibr' target='#b13'>(12)</ns0:ref><ns0:ref type='bibr' target='#b14'>(13)</ns0:ref><ns0:ref type='bibr' target='#b15'>(14)</ns0:ref>. Dolichol phosphate mannose synthase (DPMS), an essential mannosyltransferase, plays a central role in post-translational modification of proteins, including N-linked glycoproteins, Omannosylation, C-mannosylation and glycosylphosphatidylinositol (GPI) of proteins <ns0:ref type='bibr' target='#b16'>(15)</ns0:ref>. It has three subunits containing DPM1, DPM2 and DPM3 in human. DPM1, a mainly catalytic component of DPMS, is composed of 260 amino acids without any transmembrane domain region <ns0:ref type='bibr' target='#b18'>(16,</ns0:ref><ns0:ref type='bibr' target='#b20'>17)</ns0:ref>. DPM2 and DPM3 are regulatory subunits that help DPM1 localize on the endoplasmic reticulum membrane and enable it to exert catalytic activity <ns0:ref type='bibr' target='#b22'>(18)</ns0:ref>. The most reported about DPMS gene is that its absence activity is associated with congenital diseases of glycosylation (CDG) and a defect in DPM1 has been indentified to cause CDG-Ie <ns0:ref type='bibr' target='#b23'>(19,</ns0:ref><ns0:ref type='bibr' target='#b25'>20)</ns0:ref>. In addition to this, studies have reported that abnormal expression or altered enzymatic activity of DPMS was related to cell proliferation and angiogenesis. Increased DPMS activity in bovine capillary endothelial cells correlated with rised cellular proliferation <ns0:ref type='bibr' target='#b27'>(21)</ns0:ref>. Moreover, previous studies also reported that overexpressing DPMS in capillary endothelial cells significantly enhanced angiogenesis and strengthened wound healing <ns0:ref type='bibr' target='#b28'>(22)</ns0:ref>. DPMS activity however, was lacking and subsquently led to cell cycle arrest and induction of apoptosis in tunicamycin-treated capillary endothelial cells <ns0:ref type='bibr' target='#b29'>(23)</ns0:ref>. Reduced gene expression of DPMS also decreased the cellular angiogenic potential <ns0:ref type='bibr' target='#b31'>(24)</ns0:ref>. These research results indicate that the genes encoding DPMS and its protein activity may be positively related to tumor progression. However, the specific role of DPMS remains unclear in the development and progression of liver cancer. In this present work, we solved this problem by analyzing the expressions and genetic alterations of three subunits of DPMS and their association with clinical parameters in HCC patients. Furthermore, we also analyzed the predicted functions and pathways of DPMS as well as their similar genes. Materials and methods UALCAN UALCAN (http://ualcan.path.uab.edu) is a comprehensive, user-friendly, and interactive web resource and provides data online analysis and mining based on cancer OMICS data (TCGA and MET500). It is designed to analyze relative transcriptional expression of potential genes of interest between tumor and normal samples and association of the transcriptional expression with relative clinicopathologic parameters. In addition, it is also used to evaluate epigenetic regulation of gene expression and pan-cancer gene expression <ns0:ref type='bibr' target='#b32'>(25)</ns0:ref>. In our study, UALCAN was used to analyze the mRNA expressions of three subunits of DPMS in HCC samples and their relationship with clinicopathologic parameters. Difference of transcriptional expression or pathological stage analysis was compared by students' t test and p &lt;0.05 was considered as statically significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Human Protein Atlas</ns0:head><ns0:p>The Human Protein Atlas (https://www.proteinatlas.org) is a website that provides human proteins data in cells, tissues and organs, including immunohistochemistry-based expression data for near 20 common kinds of cancers <ns0:ref type='bibr' target='#b34'>(26)</ns0:ref>. The database can be conveniently used to compare the protein differential expressions of interest genes in tumors and normal tissues. In this study, direct comparison of protein expression of three subunits of DPMS between human normal and HCC tissues was performed by immunohistochemistry image. GEPIA Gene Expression Profiling Interactive Analysis (GEPIA) is a database developed and built by the team of professor Zhang of Peking University based on the data of the UCSC Xena project. It is an interactive web server that can dynamically analyze and visualize TCGA (The Cancer Genome Atlas) gene expression profile data. It can provide customizable and powerful functions, including differential expression analysis between tumor and normal samples, profiling plotting, survival analysis, similar gene detection, and so on <ns0:ref type='bibr' target='#b35'>(27)</ns0:ref>. In the current study, we operated correlative prognostic analysis and similar gene detection of DPM1, DPM2 and DPM3, respectively. p &lt;0.05 was considered as statically significant. The significance of expression analysis was completed using student's t-test. Kaplan-Meier curve was used to accomplish prognostic analysis. cBioPortal cBioPortal (www.cbioportal.org), an online open-access website resource, can display multidimensional cancer genomics data in a visual form. It can also help researchers explore the genetic changes between samples, genes and pathways, and combine them with clinical results <ns0:ref type='bibr' target='#b36'>(28)</ns0:ref>. In this experiment, we studied the genomic profiles of DPMS three subunits, which included putative copy-number alterations (CNAs) from genomic identification of significant targets in cancer (GISTIC) and mRNA Expression z-Scores (RNASeq V2 RSEM) were gained with a z-score threshold &#177;1.8. Genetic alterations in DPMS and their association with overall survival (OS) and disease free survival (DFS) of HCC patients were exhibited as Kaplan-Meier plots and log-rank test was implemented to confirm the significance of the difference between the survival curves, and when a p value &lt;0.05, the difference was statically significant. Metascape Metascape (http://metascape.org), a free and credible gene-list analysis device, can be used for gene annotation analysis and function analysis. It is a mechanized meta-analysis device that can realize habitual and different pathways in a set of orthogonal target-discovery studies <ns0:ref type='bibr' target='#b37'>(29)</ns0:ref>. In this work, Metascape was used to implement function and pathway enrichment analysis of DPMS members and their similar genes that acquired using GEPIA. Statistically significant difference was p &lt; 0.05 and minimum enrichment number was 3. Databases containing OmniPath and BioGrid were used for protein-protein interactions enriched analysis. Futhermore, Molecular Complex Detection (MCODE) was supposed to recognize closely related protein components.</ns0:p></ns0:div> <ns0:div><ns0:head>Cell Culture</ns0:head><ns0:p>The human hepatoma cells SMMC-7721, HepG2 and immortal hepatic cell QSG-7701 involved in the experiment were gained from Institute of Cell Biology (Shanghai, China). All cell lines were cultured in RPMI-1640 or DMEM medium (Gibco/Invitrogen, Camarillo, CA, UNITED STATES) supplied with 10 % fetal bovine serum (PAN-Biotech, Aidenbach, Germany), and then all cells were incubated at 37 &#176;C in a 5% CO2 environment.</ns0:p></ns0:div> <ns0:div><ns0:head>RT-qPCR</ns0:head><ns0:p>TRIeasy&#8482; Total RNA Extraction Reagent (Yeasen, Shanghai, China) was used for total RNA extraction, and then the total RNA was reverse transcribed to cDNA with the Hifair&#174; &#8545; 1st Strand cDNA Synthesis Kit (Yeasen, Shanghai, China) according to the product instruction. Hieff UNICON&#174; Power qPCR SYBR Green Master Mix (Yeasen, Shanghai, China) was used to conduct RT-qPCR experiment on a Bio-Rad CFX96 System (Bio-Rad, Hercules, CA, USA). The reaction conditions were as follows: pre-denaturation at 95&#8226;C for 30 s, followed by 40 cycles of amplification at 95&#8226;C for 10 s and 60&#8226;C for 30 s. Relative mRNA expression levels of DPM1/2/3 were measured based on the 2 &#8722;&#9651;&#9651;Ct method with 18S used for normalization. Table <ns0:ref type='table'>1</ns0:ref> showed the primers we used in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Results 1. Transcriptional levels of DPMS in liver cancer</ns0:head><ns0:p>In order to explore the gene expressions of three subunits of DPMS in different types of cancer, mRNA expressions of DPM1, DPM2 and DPM3 were analyzed by UALCAN. As was shown in Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>, we observed that DPM1, DPM2 and DPM3 had higher mRNA expressions for most kinds of tumor samples compared to normal samples, respectively. For example, mRNA expression levels of DPM1 and DPM2 were very highly expressed in colon adenocarcinoma (COAD) (DPM1, p=1.62E-12; DPM2, p&lt;1E-12 ), head and neck squamous cell carcinoma (HNSC) (DPM1, p &lt;1E-12; DPM2, p=1.62E-12 ), esophageal carcinoma (ESCA) (DPM1, p=1.22E-07; DPM2, p=2.30E-02 ), liver hepatocellular carcinoma (LIHC) (DPM1, p=1.62E-12; DPM2, p&lt;1E-12 ), rectum adenocarcinoma (READ) (DPM1, p=4.07E-09; DPM2, p=1.62E-12 ) and so on (Figure <ns0:ref type='figure' target='#fig_8'>1A,B</ns0:ref>). Similarly, DPM3 gene was particularly highly expressed in breast invasive carcinoma (BRCA) (p=1.62E-12), ESCA (p=8.22E-10), LIHC (p=1.11E-16) and glioblastoma multiforme (GBM) (p=1.53E-05) (Figure <ns0:ref type='figure' target='#fig_8'>1C</ns0:ref>). Thus, our results showed that transcriptional expressions of DPMS were significantly over-expressed in many different types of cancer. In particular, all three subunits of DPMS were expressed highly in LIHC and ESCA. Next, we examined the specific mRNA expressions of DPM1, DPM2 and DPM3 in liver tumor using UALCAN database. As was shown in Figure <ns0:ref type='figure' target='#fig_2'>2A</ns0:ref>,B and C, mRNA expressions of three genes were all found significantly up-regulated in HCC tissues compared to normal samples (all p&lt;0.001). We next performed the protein expression levels of DPMS in HCC using Human Protein Atlas database. Results indicated that medium and low protein expressions of DPM1 and DPM3 were expressed in normal liver tissues, while high protein expressions of them were showed in HCC tissues (Figure <ns0:ref type='figure' target='#fig_2'>2D,F</ns0:ref>). In addition, DPM2 protein were not detected in normal liver tissues, whereas medium expression of DPM2 were observed in HCC tissues (Figure <ns0:ref type='figure' target='#fig_2'>2E</ns0:ref>). In general, the results indicated that transcriptional and proteinic expressions of DPMS were both over-expressed in patients with HCC.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.'>Relationship between the mRNA levels of DPMS and the clinicopathological parameters in liver cancer patients</ns0:head><ns0:p>Because we observed mRNA and protein levels of DPMS were over-expressed in HCC patients, we subsequently investigated the connection between mRNA expressions of DPMS members with clinicopathological features of HCC patients with UALCAN, containing tumor grades and patients' individual cancer stages. As presented in Figure <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>, mRNA expressions of DPMS members were significantly associated with tumor grades, and the mRNA expressions of DPMS headed to be higher with tumor grade elevated. The maximum mRNA expressions of DPM1/2 were showed in tumor grade 4 (Figure <ns0:ref type='figure' target='#fig_3'>3A,B</ns0:ref>), whereas the supreme mRNA expression of DPM3 was found in tumor grade 3 (Figure <ns0:ref type='figure' target='#fig_3'>3C</ns0:ref>). The reason why mRNA expression of DPM3 in grade 3 seemed to be higher than that in grade 4 may be attributed to the small sample size (only 12 HCC patients at grade 4). Similarly, the mRNA expressions of DPMS were noticeably related to the cancer stage of patients so, the patients with more advanced cancer, the higher in mRNA expressions of DPMS. The highest mRNA expressions of DPM1/2 were observed in tumor stage 3 (Figure <ns0:ref type='figure' target='#fig_3'>3D,E</ns0:ref>), while the maximum DPM3 mRNA expression was noticed in stage 4 (Figure <ns0:ref type='figure' target='#fig_3'>3F</ns0:ref>). Briefly, the results above indicated that mRNA expressions of DPMS were obviously associated with pathological parameters in HCC patients.</ns0:p></ns0:div> <ns0:div><ns0:head n='3.'>Prognostic value of mRNA expression of DPMS in liver cancer patients</ns0:head><ns0:p>To assess the value of differentially expressed DPMS in the progression of HCC, we used GEPIA to evaluate the relationship between differentially expressed DPMS and clinical outcome. OS curves were presented in Figure <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>. We detected that liver cancer patients with low transcriptional levels of DPM1 (p=0.007), DPM2 (p=0.0032) and DPM3 (p=0.029), were significantly connected with longer OS (Figure <ns0:ref type='figure' target='#fig_4'>4A,B and C</ns0:ref>). The worth of differentially expressed DPMS in the DFS of HCC patients was also estimated. Noteworthy, the longer DFS indicated to the HCC patients with lower DPM2 transcriptional levels (p=0.049) (Figure <ns0:ref type='figure' target='#fig_4'>4E</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head n='4.'>DPMS genetic alteration and similar gene network in patients with HCC</ns0:head><ns0:p>Next, we implemented a universal analysis of the molecular characteristics of differentially expressed DPMS. Genetic variations of differentially expressed DPMS in HCC was analyzed utilizing cBioPortal. A total of 366 samples from TCGA pan cancer database were studied, and altered gene set or pathway was detected in 151queried samples (alteration rate was 41%). The alteration rates of DPM1, DPM2, and DPM3 were 19%, 6% and 24%, respectively ((Figure <ns0:ref type='figure' target='#fig_5'>5A,B</ns0:ref>). The most prevalent change in these samples was enhanced mRNA expression. The Kaplan-Meier plotter results and log-rank test presented a considerable difference in OS (p=0.0264), but no remarkable difference in DFS (p=0.0841) between the samples with changes in one of the target genes and those without variations in any target genes (Figures <ns0:ref type='figure' target='#fig_5'>5C,D</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head n='5.'>Functional enrichment analysis of DPMS in patients with HCC</ns0:head><ns0:p>Top 50 genes similar to DPM1, DPM2 and DPM3 respectively (a total of 150 genes) were searched by GEPIA (Supplementary Table <ns0:ref type='table'>1</ns0:ref>). Next, the functions of DPMS and their similar genes were predicted by analyzing GO and KEGG in Metascape. The top 20 GO enrichment items were classified into three functional groups: biological process group, molecular function group, and cellular component group (Figures <ns0:ref type='figure' target='#fig_6'>6A,B</ns0:ref> and Table <ns0:ref type='table'>2</ns0:ref>). The DPMS members and their similar genes were mainly enriched in biological processes such as ncRNA processing, DNA repair, viral gene expression, deoxyribonucleoside triphosphate metabolic process and so on. The molecular functions regulated by DPMS and their similar genes were snRNP binding, ubiquitin binding, nucleotidyltransferase activity and ubiquitin-like protein transferase activity. The cellular components affected by DPMS and their similar genes were involved in transferase complex, methyltransferase complex, chromosomal region and nucleolar part.</ns0:p><ns0:p>The 6 most significant KEGG pathways for the DPMS and their similar genes were displayed in Figures <ns0:ref type='figure' target='#fig_6'>6C,D</ns0:ref> and Table <ns0:ref type='table'>3</ns0:ref>. These pathways comprised pyrimidine metabolism, RNA transport, ubiquitin mediated proteolysis, mTOR signaling pathway and so on. Moreover, for more comprehending the relationship between DPMS and HCC, we performed enrichment analysis of protein-protein interaction with Metascape. Figures <ns0:ref type='figure' target='#fig_6'>6E and F</ns0:ref> Manuscript to be reviewed interaction correlation and important MCODE components. The top 3 essential MCODE components were achieved from the protein-protein interaction network. After function and pathway enrichment analysis for each MCODE constituents respectively, the results demonstrated that biological functions regulated by DPMS and their similar genes were mainly related to mRNA and RNA splicing, protein export form nucleus and nucleocytoplasmic transport.</ns0:p></ns0:div> <ns0:div><ns0:head n='6.'>The mRNA expression levels of DPM1/2/3 in vitro</ns0:head><ns0:p>We evaluated DPM1, DPM2 and DPM3 expression levels in a panel of three cell lines: two hepatoma cells (HepG2 and SMMC-7721) and one normal liver cell line (QSG-7701). The mRNA expression measured by RT-qPCR revealed that DPM1 transcription levels in cancerous cell lines were higher than that in normal liver cells (Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>) and the result was consistent with our prediction. Moreover, the expression of DPM2 and DPM3 in SMMC-7721 cell was significantly increased, while those expression did not change significantly in HepG2 cell (Figure <ns0:ref type='figure' target='#fig_7'>7B,C</ns0:ref>). This discrepancy may be due to a number of differences between cell types and more cell and tissue samples are needed to validate the results. Therefore, DPM1 could be the most potential prognostic biomarker for survivals of HCC patients.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Abnormal glycosylation has been found in human cancer cells decades ago, and more and more researchers have discovered that protein glycosylation contributed to tumor metastasis, angiogenesis and progression <ns0:ref type='bibr' target='#b38'>(30,</ns0:ref><ns0:ref type='bibr' target='#b40'>31)</ns0:ref>. Being an essential component of glycosyltransferase complex, DPMS protein is involved in multiple protein glycosylation process, including Nglycosylation, O-glycosylation, C-mannosylation and GPI anchors synthesis <ns0:ref type='bibr' target='#b16'>(15)</ns0:ref>. Many studies have reported that overexpressed DPMS promoted cell proliferation and angiogenesis <ns0:ref type='bibr' target='#b28'>(22)</ns0:ref>, and silencing DPMS with shRNA significantly reduced cell growth <ns0:ref type='bibr' target='#b31'>(24)</ns0:ref>. Moreover, increased DPMS activity also accelerated cellular growth <ns0:ref type='bibr' target='#b27'>(21,</ns0:ref><ns0:ref type='bibr' target='#b29'>23)</ns0:ref>. In view of the above results, we speculated that DPMS may be related to tumorigenesis and progression. To confirm this hypothesis, we predicted the expression of DPMS in cancer through bioinformatics methods, especially in liver cancer. In addition, genetic alteration and prognostic values of three subunits of DPMS in HCC were also analyzed.</ns0:p><ns0:p>Results from our study showed that the transcriptional levels of DPMS were highly expressed in different types of cancer. Moreover, over-expressions of mRNA and protein were both found in three subunits of DPMS, and mRNA expressions of DPMS were significantly associated with patients' individual cancer stages and tumor grades in HCC patients. Besides, higher mRNA expressions of DPM1/2/3 were significantly associated with shorter OS in liver cancers patients. Meanwhile, higher mRNA expression of DPM2 was significantly associated with shorter DFS in liver cancers samples. These data demonstrated that differentially expressed DPMS may play a significant role in HCC. Since three subunits of DPMS were significantly differentially expressed in HCC and closely related to liver tumor prognosis, we next explored their molecular characteristics in HCC. High alteration rate (41%) of DPMS was observed in HCC patients and the genetic alteration in DPMS was associated with shorter OS in HCC patients. Tumorigenesis and development of HCC is sophisticated and various, and genetic alteration exerts an important function among this process <ns0:ref type='bibr' target='#b41'>(32)</ns0:ref>. Among the genetic alteration elevated mRNA expression and gene amplification were the most common changes. Gene amplification, or genomic DNA copy number aberration, is frequently observed in some solid tumors and has been thought to contribute to tumor evolution <ns0:ref type='bibr' target='#b42'>(33)</ns0:ref><ns0:ref type='bibr' target='#b43'>(34)</ns0:ref><ns0:ref type='bibr' target='#b44'>(35)</ns0:ref>. Therefore, the high alteration of gene amplification in DPMS may be related to liver cancer progression. However, the specific function of gene amplification of DPMS in liver cancer need to be further studied. Finally, functions and pathways of DPM1/2/3 and their total 150 similar genes in HCC patients were analyzed. Biological processes such as ncRNA processing and DNA repair, cellular components such as transferase complex, molecular functions snRNP binding and ubiquitin binding, signal pathways such as RNA transport were remarkably regulated by DPMS and their similar genes in HCC. Our findings that DPMS was highly expressed in tumor cells are consistent with the conclusion that overexpression of DPMS in capillary endothelial cells promoted cell proliferation <ns0:ref type='bibr' target='#b28'>(22)</ns0:ref>. In addition, a paper noted that upregulation of DPMS activity may involve in angiogenesis for breast and other solid tumor proliferation and metastasis and identified DPMS as a potential 'angiogenic switch' <ns0:ref type='bibr' target='#b27'>(21)</ns0:ref>. Another report related to prostate tumor invasion pointed out DPM3 was a invasion suppressor using microarray expression analysis of the transcription levels in prostate cancer sublines <ns0:ref type='bibr' target='#b45'>(36)</ns0:ref>. This result is inconsistent with our conclusion that DPM3 was over-expressed in liver cancer cells, and the relationship between DPM3 and the invasion ability in liver cancer cells is worth further study. In addition to the above, the abnormal expressions of DPMS have been reported to be associated with human health, such as aging <ns0:ref type='bibr' target='#b48'>(37)</ns0:ref>, Thy-1 lymphoma <ns0:ref type='bibr' target='#b49'>(38)</ns0:ref> and CDG <ns0:ref type='bibr' target='#b23'>(19,</ns0:ref><ns0:ref type='bibr' target='#b50'>39)</ns0:ref>. These findings may help us to deepen our understanding for the role of DPMS in tumorigenesis and specific action mechanism among cancers.</ns0:p><ns0:p>It is known that HCC generally occurs in patients with chronic liver disease (CLD) as a result of hepatitis B virus (HBV ) and hepatitis C virus (HCV) infections, nonalcoholic fatty liver disease and alcohol-use disorder <ns0:ref type='bibr' target='#b51'>(40)</ns0:ref>. The occurrence of CLD caused by above factors is related to the glycosylation changes of key proteins <ns0:ref type='bibr' target='#b52'>(41)</ns0:ref><ns0:ref type='bibr' target='#b53'>(42)</ns0:ref><ns0:ref type='bibr' target='#b54'>(43)</ns0:ref><ns0:ref type='bibr' target='#b55'>(44)</ns0:ref><ns0:ref type='bibr' target='#b56'>(45)</ns0:ref>. For example, hepatocytes in transgenic mice that specifically expressed N-acetylglucosaminyltransferase III (GnT-III) had a swollen oval-like morphology and many lipid droplets <ns0:ref type='bibr' target='#b52'>(41,</ns0:ref><ns0:ref type='bibr' target='#b53'>42)</ns0:ref>. GnT-III was also likely to play essential roles in the change of glycosylation in viral infected people with liver diseases. DPMS is upstream of GnT-III and whether DPMS participates in the regulation process of this enzyme is worth further studying. In addition, ethanol oxidation products such as acetaldehyde interfered with the N-glycan biosynthesis and/or transfer by binding the involved enzymes in patients with liver disease. Modified glycosylation influenced proteins and receptors binding of the sinusoidal and cell surfaces of the liver in diverse CLD. Main membrane receptors glycosylation orchestrated their function in controlling tumor cell adhesion, motility and invasiveness <ns0:ref type='bibr' target='#b54'>(43)</ns0:ref>. Furthermore, modification in glycosylated receptor assignment and concentration led to glycoproteins accumulation, which were associated with the tumor size in HCC patients <ns0:ref type='bibr' target='#b55'>(44,</ns0:ref><ns0:ref type='bibr' target='#b56'>45)</ns0:ref>. Hence, the etiology of liver cancer due to chronic liver disease is perhaps attributed to the major membrane receptors and DPMS as an essential mannosyltransferase may be involved in glycosylation of major membrane receptors in liver cancer. Meanwhile, alterations in glycosylation are a common feature of cancer cells, and the complexity in protein glycosylation improves cell molecules functional diversity <ns0:ref type='bibr' target='#b57'>(46)</ns0:ref>. Many glycosyltransferases such as N-acetylglucosaminyltransferase V (GnT-V), N-acetylglucosaminyltransferase III (GnT-III) and &#945;1-6 fucosyltransferase (FUT8) have been considered to be related to the development of HCC. Genomic analysis of HCC patients inspired that overexpressed of FUT8 gene, the cause of core fucosylation, indicated that these glycan changes promoted hepatocarcinogenis, letting them potential tumor biomarkers and therapeutic targets <ns0:ref type='bibr' target='#b58'>(47)</ns0:ref>. Studies have shown that expression changes of fucosyltransferase 1 and &#946;-1,3galactosyltransferase 5 led to the occurrence of HCC <ns0:ref type='bibr' target='#b60'>(48)</ns0:ref>. High expression of these enzymes in liver cancer patients was closely linked to shorter survival times of HCC patients <ns0:ref type='bibr' target='#b61'>(49)</ns0:ref>. DPMS is upstream of these enzymes and the expression of DPMS is closely related to the expression of these enzymes. Therefore, DPMS may influence prognosis of HCC via affecting these related enzymes or similar mechanisms with these enzymes.</ns0:p><ns0:p>So far alpha-fetoprotein (AFP), des-&#947;-carboxy-prothrombin (DCP) and glypican3 (GPC3) are the major already-existed cancer biomarkers for HCC <ns0:ref type='bibr' target='#b62'>(50)</ns0:ref>. These biomarkers could be used for early detection of HCC and as markers of recurrence in the follow-up of HCC patients. AFP is more sensitive to the diagnosis of HCC, but its specificity is lower than that of DCP <ns0:ref type='bibr' target='#b63'>(51,</ns0:ref><ns0:ref type='bibr' target='#b64'>52)</ns0:ref>. Soluble GPC3 is more sensitive than AFP in monitoring highly or moderately differentiated HCC. Simultaneous detection of two or more markers increases the overall sensitivity from 50% to 72% <ns0:ref type='bibr' target='#b66'>(53)</ns0:ref>. However, about 30% of HCC patients are still negative for these traditional tumor markers. In our study, DPM1 could be a potential prognostic biomarker for survivals of HCC patients. Therefore, it is possible to use DPM1 as an effective supplemental biomarker of liver cancer. The combined application of DPM1 and other already-existed biomarkers would greatly improve the early diagnosis and accurate prognosis of liver cancers. Our study also has some limitations. First, despite mRNA expressions of DPM1/2/3 were related to the prognosis of HCC, all the data performed in our research were obtained from the online website, further studies containing larger sample sizes are needed to confirm our results and to explore the clinical application of the DPMS in HCC treatment. Second, we did not assess the potential diagnostic and therapeutic roles of DPMS in HCC, so future studies are required to explore whether DPMS could be used as diagnostic markers or as therapeutic targets. Finally, we did not explore the potential mechanisms of DPMS in HCC. Future studies are worth to investigate the detailed mechanism between DPMS expression and HCC.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In this paper, we studied the expressions of DPM1/2/3 in tumor cells and its relationship with tumorigenesis for the first time. Our results showed that over-expressions of DPM1/2/3 were significantly associated with clinical cancer stages and pathological tumor grades in HCC patients. Besides, higher mRNA expressions of DPM1/2/3 were found to be significantly connected with OS in HCC patients. Moreover, high genetic alteration rate of DPM1/2/3 (41%) was also observed, and genetic alteration in DPM1/2/3 was associated with shorter OS in HCC patients, which provide a better understanding of molecular targets for improved liver cancer therapeutic strategies in the future. DPM1 was the most potential prognostic biomarker for liver cancer via cell experiment verified. To sum up, these results indicated that DPM1 could be a prognostic biomarker for survivals of HCC patients. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.</ns0:p></ns0:div> <ns0:div><ns0:head>Figure legend</ns0:head></ns0:div> <ns0:div><ns0:head>Figure 6</ns0:head><ns0:p>The enrichment analysis of DPMS and their similar genes in HCC (Metascape). </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>exhibited the protein PeerJ reviewing PDF | (2020:06:50511:1:1:NEW 28 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure legend Figure 1. Transcriptional expressions of (A) DPM1, (B) DPM2 and (C) DPM3 in different types of cancer diseases (UALCAN database). Blue: Normal; Red: Tumor. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.Figure 2. mRNA and protein expressions of DPMS in HCC and normal liver tissues. (A-C) mRNA expressions of DPM1, DPM2 and DPM3 in HCC tissues compared to normal samples (UALCAN database). *** p&lt;0.001. (D-F) Representative immunohistochemistry images of DPM1, DPM2 and DPM3 in HCC tissues and normal liver tissues (Human Protein Atlas).Figure 3. Association of mRNA expressions of DPMS with tumor grades and patients' individual cancer stages in HCC patients (UALCAN). (A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure legend Figure 1. Transcriptional expressions of (A) DPM1, (B) DPM2 and (C) DPM3 in different types of cancer diseases (UALCAN database). Blue: Normal; Red: Tumor. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.Figure 2. mRNA and protein expressions of DPMS in HCC and normal liver tissues. (A-C) mRNA expressions of DPM1, DPM2 and DPM3 in HCC tissues compared to normal samples (UALCAN database). *** p&lt;0.001. (D-F) Representative immunohistochemistry images of DPM1, DPM2 and DPM3 in HCC tissues and normal liver tissues (Human Protein Atlas).Figure 3. Association of mRNA expressions of DPMS with tumor grades and patients' individual cancer stages in HCC patients (UALCAN). (A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure legend Figure 1. Transcriptional expressions of (A) DPM1, (B) DPM2 and (C) DPM3 in different types of cancer diseases (UALCAN database). Blue: Normal; Red: Tumor. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.Figure 2. mRNA and protein expressions of DPMS in HCC and normal liver tissues. (A-C) mRNA expressions of DPM1, DPM2 and DPM3 in HCC tissues compared to normal samples (UALCAN database). *** p&lt;0.001. (D-F) Representative immunohistochemistry images of DPM1, DPM2 and DPM3 in HCC tissues and normal liver tissues (Human Protein Atlas).Figure 3. Association of mRNA expressions of DPMS with tumor grades and patients' individual cancer stages in HCC patients (UALCAN). (A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>The prognostic value of different expressed DPM1, DPM2 and DPM3 in HCC patients (GEPIA). (A-C) Overall survival curves of DPM1, DPM2 and DPM3. (D-F) Disease free survival curves of DPM1, DPM2 and DPM3.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Genetic alterations in DPMS and their association with OS and DFS in HCC patients (cBioPortal). (A) Summary of alterations in DPMS. (B) OncoPrint visual summary of alteration on a query of DPMS. (C) Kaplan-Meier plots comparing OS in cases with/without DPMS gene alterations. (D) Kaplan-Meier plots comparing DFS in cases with/without DPMS gene alterations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>The enrichment analysis of DPMS and their similar genes in HCC (Metascape). (A) Heatmap of Gene Ontology (GO) enriched terms colored by p-values. (B) Network of GO enriched terms colored by p-value, where terms containing more genes tend to have a more significant pvalue. (C) Heatmap of Kyoto Encyclopedia of Genes and Genomes (KEGG) enriched terms colored by p-values. (D) Network of KEGG enriched terms colored by p-value, where terms containing more genes tend to have a more significant p-value. (E) Protein-protein interaction (PPI) network and three most significant MCODE components form the PPI network. (F) Independent functional enrichment analysis of three MCODE components.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>The mRNA expression levels of (A) DPM1, (B) DPM2 and (C) DPM3 in normal liver cells and hepatoma cell lines. *P &lt;0.05, ***P &lt;0.001.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 1 The</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Heatmap of Gene Ontology (GO) enriched terms colored by p-values. (B) Network of GO enriched terms colored by p-value, where terms containing more genes tend to have a more significant p-value. (C) Heatmap of Kyoto Encyclopedia of Genes and Genomes (KEGG) enriched terms colored by p-values. (D) Network of KEGG enriched terms colored by p-value, where terms containing more genes tend to have a more significant p-value. (E) Protein-protein interaction (PPI) network and three most significant MCODE components form the PPI network. (F) Independent functional enrichment analysis of three MCODE components.</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,280.87,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,275.62,525.00,320.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,250.12,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,275.62,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,229.87,525.00,170.25' type='bitmap' /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50511:1:1:NEW 28 Aug 2020)</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50511:1:1:NEW 28 Aug 2020)Manuscript to be reviewed</ns0:note> <ns0:note place='foot'>Manuscript to be reviewed</ns0:note> </ns0:body> "

"Dear Editor, Thank you very much for giving us the opportunity to revise our manuscript entitled 'DPM1 Expression as a Potential Prognostic Tumor Marker in Hepatocellular Carcinoma'. We have carefully revised the paper according to the good suggestions of the reviewers. The corrections have been highlighted in “Red” in the text. Our answers to the reviewers’ critiques are listed below. Answers to reviewer 1: 1. Because HCC usually develops from chronic injured liver, the cause of chronic liver disease (CLD), such as viral infection, alcohol, fatty liver, is of importance. They should analyze the relationship between the expression of DPM1/2/3 and the etiology for CLD. Thank the reviewer for the valuable comments, we have analyzed the relationship between the expression of DPM1/2/3 and the etiology for CLD in discussion. That is “It is known that HCC generally occurs in patients with chronic liver disease (CLD) as a result of hepatitis B virus (HBV) and hepatitis C virus (HCV) infections, nonalcoholic fatty liver disease and alcohol-use disorder (40). The occurrence of CLD caused by above factors is related to the glycosylation changes of key proteins (41-45). For example, hepatocytes in transgenic mice that specifically expressed N-acetylglucosaminyltransferase III (GnT-III) had a swollen oval-like morphology and many lipid droplets (41,42). GnT-III was also likely to play essential roles in the change of glycosylation in viral infected people with liver diseases. DPMS is upstream of GnT-III and whether DPMS participates in the regulation process of this enzyme is worth further studying. In addition, ethanol oxidation products such as acetaldehyde interfered with the N-glycan biosynthesis and/or transfer by binding the involved enzymes in patients with liver disease. Modified glycosylation influenced proteins and receptors binding of the sinusoidal and cell surfaces of the liver in diverse CLD. Main membrane receptors glycosylation orchestrated their function in controlling tumor cell adhesion, motility and invasiveness (43). Furthermore, modification in glycosylated receptor assignment and concentration led to glycoproteins accumulation, which were associated with the tumor size in HCC patients (44,45). Hence, the etiology of liver cancer due to chronic liver disease is perhaps attributed to the major membrane receptors and DPMS as an essential mannosyltransferase may be involved in glycosylation of major membrane receptors in liver cancer.” Please see Page 8-9/Line 304-321 in revised manuscript. 2. There is no comparative analysis between DPM1/2/3 and already-existed markers such as AFP, DCP, and GPC3. To highlight the significance of DPM1/2/3, additional analyses would be necessary. In response to your excellent suggestion, we have done the comparative analysis among DPM1 and already-existed markers such as AFP, DCP, and GPC3 in discussion. That is “So far alpha-fetoprotein (AFP), des-γ-carboxy-prothrombin (DCP) and glypican3 (GPC3) are the major already-existed cancer biomarkers for HCC (50). These biomarkers could be used for early detection of HCC and as markers of recurrence in the follow-up of HCC patients. AFP is more sensitive to the diagnosis of HCC, but its specificity is lower than that of DCP (51,52). Soluble GPC3 is more sensitive than AFP in monitoring highly or moderately differentiated HCC. Simultaneous detection of two or more markers increases the overall sensitivity from 50% to 72% (53). However, about 30% of HCC patients are still negative for these traditional tumor markers. In our study, DPM1 could be a potential prognostic biomarker for survivals of HCC patients. Therefore, it is possible to use DPM1 as an effective supplemental biomarker of liver cancer. The combined application of DPM1 and other already-existed biomarkers would greatly improve the early diagnosis and accurate prognosis of liver cancers.” Please see Page 9/Line 335-345 in revised manuscript. 3. The authors documented that HCC cases with low DPM1/2/3 expression showed significantly longer overall survival. From a perspective of its biological function (various protein glycosylation), please discuss why the DPM1/2/3 overexpression contributes to favorable prognosis. In agreement with your opinion, we have described the relationship between DPM1/2/3 overexpression and shorter survival in discussion. That is “Meanwhile, alterations in glycosylation are a common feature of cancer cells, and the complexity in protein glycosylation improves cell molecules functional diversity (46). Many glycosyltransferases such as N-acetylglucosaminyltransferase V (GnT-V), N-acetylglu-cosaminyltransferase III (GnT-III) and α1-6 fucosyltransferase (FUT8) have been considered to be related to the development of HCC. Genomic analysis of HCC patients inspired that overexpressed of FUT8 gene, the cause of core fucosylation, indicated that these glycan changes promoted hepatocarcinogenis, letting them potential tumor biomarkers and therapeutic targets (47). Studies have shown that expression changes of fucosyltransferase 1 and β-1,3-galactosyltransferase 5 led to the occurrence of HCC (48). High expression of these enzymes in liver cancer patients was closely linked to shorter survival times of HCC patients (49). DPMS is upstream of these enzymes and the expression of DPMS is closely related to the expression of these enzymes. Therefore, DPMS may influence prognosis of HCC via affecting these related enzymes or similar mechanisms with these enzymes.” Please see Page 9/Line 322-334. 4. It is hard to realize Fig.5. There is no information about genetic mutations. Amplification is one of the genetic alteration, but not mutation. Sorry for the incorrect understanding of amplification, we have changed “genetic mutation” into “genetic alteration”. 5. They conducted functional enrichment analyses. However, they analyzed 50 genes similar to DPM1/2/3, but not DPM1/2/3. Please clarify this issue.？？？ We believe that genes similar to DPM1/2/3 may perform the same functions as DPM1/2/3. Based on this, we used GEPIA to search for similar genes of DPM1/2/3, and then carried out enrichment analysis to deduce the functions and pathways of possible enrichment of similar genes. Therefore, we can further understand the function of DPM1/2/3 in liver cancer. In addition, DPM1/2/3 three genes cannot proceed for the functional enrichment analyses due to the less available genes. Minor comments: 1. Please spell out all abbreviations when they were firstly written. Additionally, names of cancer types in Fig.1 were spelled out in its legend. Thank the reviewer for the helpful comments, we have spelt out all abbreviations when they were firstly written and added the relevant contents to the figure legend of the manuscript. Please see Page 16/Line 597-606. 2. Please confirm that there is no problem to use images downloaded from Human Protein Atlas in your manuscript. The reviewer has never experienced such a case. According to many similar studies, the authors have used images downloaded from Human Protein Atlas. Therefore, we tried to use them to enrich our data. 3. Labels in Fig. 3 was too unclear to recognize. These should be modified using drawing software. Sorry for the unclear picture, we have changed the blurry image for a clear one. Answers to reviewer 2: This paper presents the results of gene expression analysis of DPM1, DPM2, and DPM3 as a potential prognostic biomarker of hepatocellular cancer patients. It describes interesting data on expression of these genes in retrospective data sets. This research contains interesting findings with a focus on DPMS, but there are several questions and comments. Thank the reviwer for the positive evaluation. 1. The authors represent the relationship between the DPMS expression profile and survival based on online database. UALCAN, GEPIA, and cBioportal are analytical tools based on TCGA database and others. However, cBioportal uses data of gene amplification as well as gene expression. Although results by each tool were similar, the role of gene amplification could be discussed in cBioportal. Thank the reviewer for the good advice, we have discussed the role of gene amplification in discussion. Please see Page 8/Line 280-285. 2. Please provide sufficient information in Figure 1. Do the blue and red labels indicate the normal and cancerous samples, respectively? Sorry for the unclearness. We have explained the meaning of the blue and red labels in Figure 1 legend. As you can see, blue label indicates normal samples, red label refers to tumor samples. Please see Page 16/Line 597 3. In Figure 1, statistical significance of the difference between normal and cancerous samples should be stated. According to your good suggestion, we have added the statistical significance of the Figure 1 in revised manuscript. Please see Page 5/Line 167-172. 4. Please provide sufficient information in Figure 3. What meaning did the colors refer to? Sorry for the unclearness, we have added the sufficient information in revised manuscript. Blue, orange, green, brown and red labels indicate normal, grade 1 grade 2, grade 3 and grade 4 samples in Figure 3A-C, respectively. In Figure 3D-F, these labels refer to normal and stage 1-4 samples. 5. Genetic alteration should be replaced to genomic alteration in Figure 5. Thank the reviewer for the suggestion, we have confirmed the analyzed results of cBioPortal database and ensured that the “genetic alteration” in Figure 5 is not “genomic alteration”. Answers to editor: Questions about image format of Figure 1: The images of Figure 1 are downloaded from UALCAN database directly and we could not change the Y-axes optionally. In summary, we have revised our manuscript carefully according to the good suggestions and recommendations by the reviewers. We hope that you find our resubmitted manuscript acceptable for publication at Peer J. Thank you, Sincerely, Ping Shi, Ph.D. Professor State Key Laboratory of Bioreactor Engineering East China University of Science and Technology 130 Meilong Road, Shanghai 200237, China +86-21-64251655 (phone), +86-21-64252920 (fax) "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: Altered glycosylation of proteins contributes to tumor progression. Dolichol phosphate mannose synthase (DPMS), an essential mannosyltransferase, plays a central role in post-translational modification of proteins, including N-linked glycoproteins, O-mannosylation, C-mannosylation and glycosylphosphatidylinositol anchors synthesis. Little is known about the function of DPMS in liver cancer.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods:</ns0:head><ns0:p>The study explored the roles of DPMS in the prognosis of hepatocellular carcinoma using UALCAN, Human Protein Atlas, GEPIA, cBioPortal and Metascape databases. The mRNA expressions of DPM1/2/3 also were detected by quantitative real-time PCR experiments in vitro.</ns0:p></ns0:div> <ns0:div><ns0:head>Results:</ns0:head><ns0:p>The transcriptional and proteinic expressions of DPM1/2/3 were both over-expressed in patients with hepatocellular carcinoma. Over expressions of DPMS were discovered to be dramatically associated with clinical cancer stages and pathological tumor grades in hepatocellular carcinoma patients. In addition, higher mRNA expressions of DPM1/2/3 were found to be significantly related to shorter overall survival in liver cancer patients. Futhermore, high genetic alteration rate of DPMS (41%) was also observed in patients with liver cancer, and genetic alteration in DPMS was associated with shorter overall survival in hepatocellular carcinoma patients. We also performed quantitative real-time PCR experiments in human normal hepatocytes and hepatoma cells to verify the expressions of DPM1/2/3 and results showed that the expression of DPM1 was significantly increased in hepatoma cells SMMC-7721 and HepG2.</ns0:p><ns0:p>Conclusions: Taken together, these results suggested that DPM1 could be a potential prognostic biomarker for survivals of hepatocellular carcinoma patients.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Hepatocellular carcinoma (HCC) is one of the most frequently and commonly occurring malignant tumors worldwide. The global incidence and mortality rate of HCC are ranked 5th and 3rd among all types of cancers <ns0:ref type='bibr' target='#b0'>(1,</ns0:ref><ns0:ref type='bibr' target='#b1'>2)</ns0:ref>. Despite making remarkable advances in new technologies for diagnosis and treatment, the incidence and mortality of HCC still continue to growth because of the poorest prognosis <ns0:ref type='bibr' target='#b2'>(3,</ns0:ref><ns0:ref type='bibr' target='#b3'>4)</ns0:ref>. Therefore, it is urgently needed to determine reliable predictive biomarkers for early diagnosis and accurate prognosis, and to develop new molecular targeted therapeutic strategies.</ns0:p><ns0:p>The occurrence and development of several cancer types are closely associated with aberrant protein glycosylation <ns0:ref type='bibr' target='#b5'>(5,</ns0:ref><ns0:ref type='bibr' target='#b6'>6)</ns0:ref>. Studies have suggested that altered glycosylation of proteins has been observed in liver cancer <ns0:ref type='bibr' target='#b7'>(7)</ns0:ref>. Although mounting evidence has reported the role of glycosylation in tumor progression <ns0:ref type='bibr' target='#b8'>(8)</ns0:ref><ns0:ref type='bibr' target='#b9'>(9)</ns0:ref><ns0:ref type='bibr' target='#b10'>(10)</ns0:ref>, there is limited information on how glycosylation affects the liver cancer development. Recent studies have focused on glycosylation crosstalks with cellular metabolism and related kinases <ns0:ref type='bibr' target='#b11'>(11)</ns0:ref><ns0:ref type='bibr' target='#b13'>(12)</ns0:ref><ns0:ref type='bibr' target='#b14'>(13)</ns0:ref><ns0:ref type='bibr' target='#b15'>(14)</ns0:ref>. Dolichol phosphate mannose synthase (DPMS), an essential mannosyltransferase, plays a central role in post-translational modification of proteins, including N-linked glycoproteins, Omannosylation, C-mannosylation and glycosylphosphatidylinositol (GPI) of proteins <ns0:ref type='bibr' target='#b16'>(15)</ns0:ref>. It has three subunits containing DPM1, DPM2 and DPM3 in human. DPM1, a mainly catalytic component of DPMS, is composed of 260 amino acids without any transmembrane domain region <ns0:ref type='bibr' target='#b18'>(16,</ns0:ref><ns0:ref type='bibr' target='#b20'>17)</ns0:ref>. DPM2 and DPM3 are regulatory subunits that help DPM1 localize on the endoplasmic reticulum membrane and enable it to exert catalytic activity <ns0:ref type='bibr' target='#b22'>(18)</ns0:ref>. The most reported about DPMS gene is that its absence activity is associated with congenital diseases of glycosylation (CDG) and a defect in DPM1 has been indentified to cause CDG-Ie <ns0:ref type='bibr' target='#b23'>(19,</ns0:ref><ns0:ref type='bibr' target='#b25'>20)</ns0:ref>. In addition to this, studies have reported that abnormal expression or altered enzymatic activity of DPMS was related to cell proliferation and angiogenesis. Increased DPMS activity in bovine capillary endothelial cells correlated with rised cellular proliferation <ns0:ref type='bibr' target='#b27'>(21)</ns0:ref>. Moreover, previous studies also reported that overexpressing DPMS in capillary endothelial cells significantly enhanced angiogenesis and strengthened wound healing <ns0:ref type='bibr' target='#b28'>(22)</ns0:ref>. DPMS activity however, was lacking and subsquently led to cell cycle arrest and induction of apoptosis in tunicamycin-treated capillary endothelial cells <ns0:ref type='bibr' target='#b29'>(23)</ns0:ref>. Reduced gene expression of DPMS also decreased the cellular angiogenic potential <ns0:ref type='bibr' target='#b31'>(24)</ns0:ref>. These research results indicate that the genes encoding DPMS and its protein activity may be positively related to tumor progression. However, the specific role of DPMS remains unclear in the development and progression of liver cancer. In this present work, we solved this problem by analyzing the expressions and genetic alterations of three subunits of DPMS and their association with clinical parameters in HCC patients. Furthermore, we also analyzed the predicted functions and pathways of DPMS as well as their similar genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div> <ns0:div><ns0:head>Datasets</ns0:head><ns0:p>Datasets used for correlation analysis between DPM1/2/3 and chronic liver disease (CLD) were obtained from GEO database (http://www.ncbi.nlm.nih.gov/geo/) after searching for keywords related to CLD. We selected three separate gene expression profiles (GSE114783, GSE128726 and GSE89632) for our study and the detailed information of the datasets was shown in Table <ns0:ref type='table'>1</ns0:ref>. The data used for ROC curve plotted were collected from TCGA LIHC datasets. The figures about the relationship between DPM1/2/3 expression and CLD were drawn using R package, ggplot2 v3.3.2. The significance of DPM1/2/3 expressions between normal and CLD samples were analyzed via unpaired Student's t-test. The ROC curves were created by R package, pROC v1.16.2.</ns0:p></ns0:div> <ns0:div><ns0:head>UALCAN</ns0:head><ns0:p>UALCAN (http://ualcan.path.uab.edu) is a comprehensive, user-friendly, and interactive web resource and provides data online analysis and mining based on cancer OMICS data (TCGA and MET500). It is designed to analyze relative transcriptional expression of potential genes of interest between tumor and normal samples and association of the transcriptional expression with relative clinicopathologic parameters. In addition, it is also used to evaluate epigenetic regulation of gene expression and pan-cancer gene expression <ns0:ref type='bibr' target='#b32'>(25)</ns0:ref>. In our study, UALCAN was used to analyze the mRNA expressions of three subunits of DPMS in HCC samples and their relationship with clinicopathologic parameters. Difference of transcriptional expression or pathological stage analysis was compared by Student's t-test and p &lt;0.05 was considered as statically significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Human Protein Atlas</ns0:head><ns0:p>The Human Protein Atlas (https://www.proteinatlas.org) is a website that provides human proteins data in cells, tissues and organs, including immunohistochemistry-based expression data for near 20 common kinds of cancers <ns0:ref type='bibr' target='#b34'>(26)</ns0:ref>. The database can be conveniently used to compare the protein differential expressions of interest genes in tumors and normal tissues. In this study, direct comparison of protein expression of three subunits of DPMS between human normal and HCC tissues was performed by immunohistochemistry image. GEPIA Gene Expression Profiling Interactive Analysis (GEPIA) is a database developed and built by the team of professor Zhang of Peking University based on the data of the UCSC Xena project. It is an interactive web server that can dynamically analyze and visualize TCGA (The Cancer Genome Atlas) gene expression profile data. It can provide customizable and powerful functions, including differential expression analysis between tumor and normal samples, profiling plotting, survival analysis, similar gene detection, and so on <ns0:ref type='bibr' target='#b35'>(27)</ns0:ref>. In the current study, we operated correlative prognostic analysis and similar gene detection of DPM1, DPM2 and DPM3, respectively. p &lt;0.05 was considered as statically significant. The significance of expression analysis was completed using Student's t-test. Kaplan-Meier curve was used to accomplish prognostic analysis. cBioPortal cBioPortal (www.cbioportal.org), an online open-access website resource, can display multidimensional cancer genomics data in a visual form. It can also help researchers explore the genetic changes between samples, genes and pathways, and combine them with clinical results <ns0:ref type='bibr' target='#b36'>(28)</ns0:ref>. In this experiment, we studied the genomic profiles of DPMS three subunits, which included putative copy-number alterations (CNAs) from genomic identification of significant targets in cancer (GISTIC) and mRNA Expression z-Scores (RNASeq V2 RSEM) were gained with a z-score threshold &#177;1.8. Genetic alterations in DPMS and their association with overall survival (OS) and disease free survival (DFS) of HCC patients were exhibited as Kaplan-Meier plots and log-rank test was implemented to confirm the significance of the difference between the survival curves, and when a p value &lt;0.05, the difference was statically significant. Metascape Metascape (http://metascape.org), a free and credible gene-list analysis device, can be used for gene annotation analysis and function analysis. It is a mechanized meta-analysis device that can realize habitual and different pathways in a set of orthogonal target-discovery studies <ns0:ref type='bibr' target='#b37'>(29)</ns0:ref>. In this work, Metascape was used to implement function and pathway enrichment analysis of DPMS members and their similar genes that acquired using GEPIA. Statistically significant difference was p &lt; 0.05 and minimum enrichment number was 3. Databases containing OmniPath and BioGrid were used for protein-protein interactions enriched analysis. Futhermore, Molecular Complex Detection (MCODE) was supposed to recognize closely related protein components.</ns0:p></ns0:div> <ns0:div><ns0:head>Cell Culture</ns0:head><ns0:p>The human hepatoma cells SMMC-7721, HepG2 and immortal hepatic cell QSG-7701 involved in the experiment were gained from Institute of Cell Biology (Shanghai, China). All cell lines were cultured in RPMI-1640 or DMEM medium (Gibco/Invitrogen, Camarillo, CA, UNITED STATES) supplied with 10 % fetal bovine serum (PAN-Biotech, Aidenbach, Germany), and then all cells were incubated at 37 &#176;C in a 5% CO2 environment.</ns0:p></ns0:div> <ns0:div><ns0:head>RT-qPCR</ns0:head><ns0:p>TRIeasy&#8482; Total RNA Extraction Reagent (Yeasen, Shanghai, China) was used for total RNA extraction, and then the total RNA was reverse transcribed to cDNA with the Hifair&#174; &#8545; 1st Strand cDNA Synthesis Kit (Yeasen, Shanghai, China) according to the product instruction. Hieff UNICON&#174; Power qPCR SYBR Green Master Mix (Yeasen, Shanghai, China) was used to conduct RT-qPCR experiment on a Bio-Rad CFX96 System (Bio-Rad, Hercules, CA, USA). The reaction conditions were as follows: pre-denaturation at 95&#8226;C for 30 s, followed by 40 cycles of amplification at 95&#8226;C for 10 s and 60&#8226;C for 30 s. Relative mRNA expression levels of DPM1/2/3 were measured based on the 2 &#8722;&#9651;&#9651;Ct method with 18S used for normalization. The significance of expression analysis was completed using Student's t-test. Table <ns0:ref type='table'>2</ns0:ref> showed the primers we used in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head n='1.'>Transcriptional levels of DPMS in liver cancer</ns0:head><ns0:p>In order to explore the gene expressions of three subunits of DPMS in different types of cancer, mRNA expressions of DPM1, DPM2 and DPM3 were analyzed by UALCAN. As was shown in Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, we observed that DPM1, DPM2 and DPM3 had higher mRNA expressions for most kinds of tumor samples compared to normal samples, respectively. For example, mRNA expression levels of DPM1 and DPM2 were very highly expressed in colon adenocarcinoma (COAD) (DPM1, p=1.62E-12; DPM2, p&lt;1E-12 ), head and neck squamous cell carcinoma (HNSC) (DPM1, p &lt;1E-12; DPM2, p=1.62E-12 ), esophageal carcinoma (ESCA) (DPM1, p=1.22E-07; DPM2, p=2.30E-02 ), liver hepatocellular carcinoma (LIHC) (DPM1, p=1.62E-12; DPM2, p&lt;1E-12 ), rectum adenocarcinoma (READ) (DPM1, p=4.07E-09; DPM2, p=1.62E-12 ) and so on (Figure <ns0:ref type='figure' target='#fig_9'>1A,B</ns0:ref>). Similarly, DPM3 gene was particularly highly expressed in breast invasive carcinoma (BRCA) (p=1.62E-12), ESCA (p=8.22E-10), LIHC (p=1.11E-16) and glioblastoma multiforme (GBM) (p=1.53E-05) (Figure <ns0:ref type='figure' target='#fig_9'>1C</ns0:ref>). Thus, our results showed that transcriptional expressions of DPMS were significantly over-expressed in many different types of cancer. In particular, all three subunits of DPMS were expressed highly in LIHC and ESCA. Next, we examined the specific mRNA expressions of DPM1, DPM2 and DPM3 in liver tumor using UALCAN database. As was shown in Figure <ns0:ref type='figure' target='#fig_1'>2A,B and C</ns0:ref>, mRNA expressions of three genes were all found significantly up-regulated in HCC tissues compared to normal samples (all p&lt;0.001). We next performed the protein expression levels of DPMS in HCC using Human Protein Atlas database. Results indicated that medium and low protein expressions of DPM1 and DPM3 were expressed in normal liver tissues, while high protein expressions of them were showed in HCC tissues (Figure <ns0:ref type='figure' target='#fig_1'>2D,F</ns0:ref>). In addition, DPM2 protein were not detected in normal liver tissues, whereas medium expression of DPM2 were observed in HCC tissues (Figure <ns0:ref type='figure' target='#fig_1'>2E</ns0:ref>). In general, the results indicated that transcriptional and proteinic expressions of DPMS were both over-expressed in patients with HCC.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.'>Relationship between the mRNA levels of DPMS and the clinicopathological parameters in liver cancer patients</ns0:head><ns0:p>Because we observed mRNA and protein levels of DPMS were over-expressed in HCC patients, we subsequently investigated the connection between mRNA expressions of DPMS members with clinicopathological features of HCC patients with UALCAN, containing tumor grades and patients' individual cancer stages. As presented in Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>, mRNA expressions of DPMS members were significantly associated with tumor grades, and the mRNA expressions of DPMS headed to be higher with tumor grade elevated. The maximum mRNA expressions of DPM1/2 were showed in tumor grade 4 (Figure <ns0:ref type='figure' target='#fig_10'>3A,B</ns0:ref>), whereas the supreme mRNA expression of DPM3 was found in tumor grade 3 (Figure <ns0:ref type='figure' target='#fig_10'>3C</ns0:ref>). The reason why mRNA expression of DPM3 in grade 3 seemed to be higher than that in grade 4 may be attributed to the small sample size (only 12 HCC patients at grade 4). Similarly, the mRNA expressions of DPMS were noticeably related to the cancer stage of patients so, the patients with more advanced cancer, the higher in mRNA expressions of DPMS. The highest mRNA expressions of DPM1/2 were observed in tumor stage 3 (Figure <ns0:ref type='figure' target='#fig_10'>3D,E</ns0:ref>), while the maximum DPM3 mRNA expression was noticed in stage 4 (Figure <ns0:ref type='figure' target='#fig_10'>3F</ns0:ref>). Briefly, the results above indicated that mRNA expressions of DPMS were obviously associated with pathological parameters in HCC patients. Moreover, HCC usually developed from CLD caused by hepatitis B virus (HBV) and hepatitis C virus (HCV) infection, fatty liver and so on. The relationships between the expressions of DPM1/2/3 and CLD including HBV-related liver cirrhosis, HCV-related liver cirrhosis and non-alcoholic steatohepatitis were also analyzed. Three suitable datasets (GSE114783, GSE128726 and GSE89632) were chosen for verifying the expression of DPM1/2/3 in CLD. We found that the expressions of DPM1/2/3 in HBV and HCV-related liver cirrhosis and non-alcoholic steatohepatitis samples were more or less higher than normal samples (Figure <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>). Therefore, the expressions of DPM1/2/3 were also related to disease development of HCC.</ns0:p></ns0:div> <ns0:div><ns0:head n='3.'>Prognostic value of mRNA expression of DPMS in liver cancer patients</ns0:head><ns0:p>To assess the value of differentially expressed DPMS in the progression of HCC, we used GEPIA to evaluate the relationship between differentially expressed DPMS and clinical outcome. OS curves were presented in Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>. We detected that liver cancer patients with low transcriptional levels of DPM1 (p=0.007), DPM2 (p=0.0032) and DPM3 (p=0.029), were significantly connected with longer OS (Figure <ns0:ref type='figure' target='#fig_13'>5A,B and C</ns0:ref>). The worth of differentially expressed DPMS in the DFS of HCC patients was also estimated. Noteworthy, the longer DFS indicated to the HCC patients with lower DPM2 transcriptional levels (p=0.049) (Figure <ns0:ref type='figure' target='#fig_13'>5E</ns0:ref>). The receiver operating characteristic (ROC) curves were used to detect the prediction accuracy of DPM1/2/3 in distinguishing the HCC from the normal samples compared with existed liver tumor markers containing alpha-fetoprotein (AFP), glypican-3 (GPC-3) and transforming growth factor-&#946;1 (TGF&#946;1). Our results indicated that DPM1/2/3 had a better performance than AFP and TGF&#946;1 for the diagnosis of HCC (Figure <ns0:ref type='figure' target='#fig_5'>6</ns0:ref>). Area under the curve (AUC) of the DPM1/2/3 were 0.709, 0.860 and 0.746, respectively (Figure <ns0:ref type='figure' target='#fig_5'>6A,B and C</ns0:ref>). AUC of the existed tumor markers including AFP, GPC-3 and TGF&#946;1 were 0.679, 0.879 and 0.577, respectively (Figure <ns0:ref type='figure' target='#fig_5'>6D,E and F</ns0:ref>). Taken together, DPM1/2/3 may be also potential biomarkers for diagnosis or screening of HCC besides AFP, GPC-3 and TGF&#946;1.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.'>DPMS genetic alteration and similar gene network in patients with HCC</ns0:head><ns0:p>Next, we implemented a universal analysis of the molecular characteristics of differentially expressed DPMS. Genetic variations of differentially expressed DPMS in HCC was analyzed utilizing cBioPortal. A total of 366 samples from TCGA pan cancer database were studied, and altered gene set or pathway was detected in 151queried samples (alteration rate was 41%). The alteration rates of DPM1, DPM2, and DPM3 were 19%, 6% and 24%, respectively ((Figure <ns0:ref type='figure' target='#fig_16'>7A,B</ns0:ref>). The most prevalent change in these samples was enhanced mRNA expression. The Kaplan-Meier plotter results and log-rank test presented a considerable difference in OS (p=0.0264), but no remarkable difference in DFS (p=0.0841) between the samples with changes in one of the target genes and those without variations in any target genes (Figures <ns0:ref type='figure' target='#fig_16'>7C,D</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head n='5.'>Functional enrichment analysis of DPMS in patients with HCC</ns0:head><ns0:p>Top 50 genes similar to DPM1, DPM2 and DPM3 respectively (a total of 150 genes) were searched by GEPIA (Supplementary Table <ns0:ref type='table'>1</ns0:ref>). Next, the functions of DPMS and their similar genes were predicted by analyzing GO and KEGG in Metascape. The top 20 GO enrichment items were classified into three functional groups: biological process group, molecular function group, and cellular component group (Figures <ns0:ref type='figure' target='#fig_7'>8A,B</ns0:ref> and Table <ns0:ref type='table'>3</ns0:ref>). The DPMS members and their similar genes were mainly enriched in biological processes such as ncRNA processing, DNA repair, viral gene expression, deoxyribonucleoside triphosphate metabolic process and so on. The molecular functions regulated by DPMS and their similar genes were snRNP binding, ubiquitin binding, nucleotidyltransferase activity and ubiquitin-like protein transferase activity. The cellular components affected by DPMS and their similar genes were involved in transferase complex, methyltransferase complex, chromosomal region and nucleolar part.</ns0:p><ns0:p>The 6 most significant KEGG pathways for the DPMS and their similar genes were displayed in Figures <ns0:ref type='figure' target='#fig_7'>8C,D and Table 4</ns0:ref>. These pathways comprised pyrimidine metabolism, RNA transport, ubiquitin mediated proteolysis, mTOR signaling pathway and so on. Moreover, for more comprehending the relationship between DPMS and HCC, we performed enrichment analysis of protein-protein interaction with Metascape. Figures <ns0:ref type='figure' target='#fig_7'>8E and F</ns0:ref> exhibited the protein interaction correlation and important MCODE components. The top 3 essential MCODE components were achieved from the protein-protein interaction network. After function and pathway enrichment analysis for each MCODE constituents respectively, the results demonstrated that biological functions regulated by DPMS and their similar genes were mainly related to mRNA and RNA splicing, protein export form nucleus and nucleocytoplasmic transport.</ns0:p></ns0:div> <ns0:div><ns0:head n='6.'>The mRNA expression levels of DPM1/2/3 in vitro</ns0:head><ns0:p>We evaluated DPM1, DPM2 and DPM3 expression levels in a panel of three cell lines: two hepatoma cells (HepG2 and SMMC-7721) and one normal liver cell line (QSG-7701). The mRNA expression measured by RT-qPCR revealed that DPM1 transcription levels in cancerous cell lines were higher than that in normal liver cells (Figure <ns0:ref type='figure' target='#fig_8'>9A</ns0:ref>) and the result was consistent with our prediction. Moreover, the expression of DPM2 and DPM3 in SMMC-7721 cell was significantly increased, while those expression did not change significantly in HepG2 cell (Figure <ns0:ref type='figure' target='#fig_8'>9B,C</ns0:ref>). This discrepancy may be due to a number of differences between cell types and more cell and tissue samples are needed to validate the results. Therefore, DPM1 could be the most potential prognostic biomarker for survivals of HCC patients.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Abnormal glycosylation has been found in human cancer cells decades ago, and more and more researchers have discovered that protein glycosylation contributed to tumor metastasis, angiogenesis and progression <ns0:ref type='bibr' target='#b38'>(30,</ns0:ref><ns0:ref type='bibr' target='#b40'>31)</ns0:ref>. Being an essential component of glycosyltransferase complex, DPMS protein is involved in multiple protein glycosylation process, including Nglycosylation, O-glycosylation, C-mannosylation and GPI anchors synthesis <ns0:ref type='bibr' target='#b16'>(15)</ns0:ref>. Many studies have reported that overexpressed DPMS promoted cell proliferation and angiogenesis <ns0:ref type='bibr' target='#b28'>(22)</ns0:ref>, and silencing DPMS with shRNA significantly reduced cell growth <ns0:ref type='bibr' target='#b31'>(24)</ns0:ref>. Moreover, increased DPMS activity also accelerated cellular growth <ns0:ref type='bibr' target='#b27'>(21,</ns0:ref><ns0:ref type='bibr' target='#b29'>23)</ns0:ref>. In view of the above results, we speculated that DPMS may be related to tumorigenesis and progression. To confirm this hypothesis, we predicted the expression of DPMS in cancer through bioinformatics methods, especially in liver cancer. In addition, genetic alteration and prognostic values of three subunits of DPMS in HCC were also analyzed.</ns0:p><ns0:p>Results from our study showed that the transcriptional levels of DPMS were highly expressed in different types of cancer. Moreover, over-expressions of mRNA and protein were both found in three subunits of DPMS, and mRNA expressions of DPMS were significantly associated with patients' individual cancer stages and tumor grades in HCC patients. Besides, higher mRNA expressions of DPM1/2/3 were significantly associated with shorter OS in liver cancers patients. Meanwhile, higher mRNA expression of DPM2 was significantly associated with shorter DFS in liver cancers samples. These data demonstrated that differentially expressed DPMS may play a significant role in HCC. Since three subunits of DPMS were significantly differentially expressed in HCC and closely related to liver tumor prognosis, we next explored their molecular characteristics in HCC. High alteration rate (41%) of DPMS was observed in HCC patients and the genetic alteration in DPMS was associated with shorter OS in HCC patients. Tumorigenesis and development of HCC is sophisticated and various, and genetic alteration exerts an important function among this process <ns0:ref type='bibr' target='#b41'>(32)</ns0:ref>. Among the genetic alteration elevated mRNA expression and gene amplification were the most common changes. Gene amplification, or genomic DNA copy number aberration, is frequently observed in some solid tumors and has been thought to contribute to tumor evolution <ns0:ref type='bibr' target='#b42'>(33)</ns0:ref><ns0:ref type='bibr' target='#b43'>(34)</ns0:ref><ns0:ref type='bibr' target='#b44'>(35)</ns0:ref>. Therefore, the high alteration of gene amplification in DPMS may be related to liver cancer progression. However, the specific function of gene amplification of DPMS in liver cancer need to be further studied. Finally, functions and pathways of DPM1/2/3 and their total 150 similar genes in HCC patients were analyzed. Biological processes such as ncRNA processing and DNA repair, cellular components such as transferase complex, molecular functions snRNP binding and ubiquitin binding, signal pathways such as RNA transport were remarkably regulated by DPMS and their similar genes in HCC. Our findings that DPMS was highly expressed in tumor cells are consistent with the conclusion that overexpression of DPMS in capillary endothelial cells promoted cell proliferation <ns0:ref type='bibr' target='#b28'>(22)</ns0:ref>. In addition, a paper noted that upregulation of DPMS activity may involve in angiogenesis for breast and other solid tumor proliferation and metastasis and identified DPMS as a potential 'angiogenic switch' <ns0:ref type='bibr' target='#b27'>(21)</ns0:ref>. Another report related to prostate tumor invasion pointed out DPM3 was a invasion suppressor using microarray expression analysis of the transcription levels in prostate cancer sublines <ns0:ref type='bibr' target='#b45'>(36)</ns0:ref>. This result is inconsistent with our conclusion that DPM3 was over-expressed in liver cancer cells, and the relationship between DPM3 and the invasion ability in liver cancer cells is worth further study. In addition to the above, the abnormal expressions of DPMS have been reported to be associated with human health, such as aging <ns0:ref type='bibr' target='#b48'>(37)</ns0:ref>, Thy-1 lymphoma <ns0:ref type='bibr' target='#b49'>(38)</ns0:ref> and CDG <ns0:ref type='bibr' target='#b23'>(19,</ns0:ref><ns0:ref type='bibr' target='#b50'>39)</ns0:ref>. These findings may help us to deepen our understanding for the role of DPMS in tumorigenesis and specific action mechanism among cancers.</ns0:p><ns0:p>It is known that HCC generally occurs in patients withCLD as a result of HBV and HCV infections, nonalcoholic fatty liver disease and alcohol-use disorder <ns0:ref type='bibr' target='#b51'>(40)</ns0:ref>. The occurrence of CLD caused by above factors is related to the glycosylation changes of key proteins <ns0:ref type='bibr' target='#b52'>(41)</ns0:ref><ns0:ref type='bibr' target='#b53'>(42)</ns0:ref><ns0:ref type='bibr' target='#b54'>(43)</ns0:ref><ns0:ref type='bibr' target='#b55'>(44)</ns0:ref><ns0:ref type='bibr' target='#b56'>(45)</ns0:ref>. For example, hepatocytes in transgenic mice that specifically expressed Nacetylglucosaminyltransferase III (GnT-III) had a swollen oval-like morphology and many lipid droplets <ns0:ref type='bibr' target='#b52'>(41,</ns0:ref><ns0:ref type='bibr' target='#b53'>42)</ns0:ref>. GnT-III was also likely to play essential roles in the change of glycosylation in viral infected people with liver diseases. DPMS is upstream of GnT-III and whether DPMS participates in the regulation process of this enzyme is worth further studying. In addition, ethanol oxidation products such as acetaldehyde interfered with the N-glycan biosynthesis and/or transfer by binding the involved enzymes in patients with liver disease. Modified glycosylation influenced proteins and receptors binding of the sinusoidal and cell surfaces of the liver in diverse CLD. Main membrane receptors glycosylation orchestrated their function in controlling tumor cell adhesion, motility and invasiveness <ns0:ref type='bibr' target='#b54'>(43)</ns0:ref>. Furthermore, modification in glycosylated receptor assignment and concentration led to glycoproteins accumulation, which were associated with the tumor size in HCC patients <ns0:ref type='bibr' target='#b55'>(44,</ns0:ref><ns0:ref type='bibr' target='#b56'>45)</ns0:ref>. Hence, the etiology of liver cancer due to chronic liver disease is perhaps attributed to the major membrane receptors and DPMS as an essential mannosyltransferase may be involved in glycosylation of major membrane receptors in liver cancer.</ns0:p><ns0:p>Meanwhile, alterations in glycosylation are a common feature of cancer cells, and the complexity in protein glycosylation improves cell molecules functional diversity <ns0:ref type='bibr' target='#b57'>(46)</ns0:ref>. Many glycosyltransferases such as N-acetylglucosaminyltransferase V (GnT-V), N-acetylglucosaminyltransferase III (GnT-III) and &#945;1-6 fucosyltransferase (FUT8) have been considered to be related to the development of HCC. Genomic analysis of HCC patients inspired that overexpressed of FUT8 gene, the cause of core fucosylation, indicated that these glycan changes promoted hepatocarcinogenis, letting them potential tumor biomarkers and therapeutic targets <ns0:ref type='bibr' target='#b58'>(47)</ns0:ref>. Studies have shown that expression changes of fucosyltransferase 1 and &#946;-1,3galactosyltransferase 5 led to the occurrence of HCC <ns0:ref type='bibr' target='#b60'>(48)</ns0:ref>. High expression of these enzymes in liver cancer patients was closely linked to shorter survival times of HCC patients <ns0:ref type='bibr' target='#b61'>(49)</ns0:ref>. DPMS is upstream of these enzymes and the expression of DPMS is closely related to the expression of these enzymes. Therefore, DPMS may influence prognosis of HCC via affecting these related enzymes or similar mechanisms with these enzymes.</ns0:p><ns0:p>So far AFP, des-&#947;-carboxy-prothrombin (DCP) ,GPC3 and TGF&#946;1 are the major alreadyexisted cancer biomarkers for HCC <ns0:ref type='bibr' target='#b62'>(50)</ns0:ref>. These biomarkers could be used for early detection of HCC and as markers of recurrence in the follow-up of HCC patients. AFP is more sensitive to the diagnosis of HCC, but its specificity is lower than that of DCP <ns0:ref type='bibr' target='#b63'>(51,</ns0:ref><ns0:ref type='bibr' target='#b64'>52)</ns0:ref>. Soluble GPC3 is more sensitive than AFP in monitoring highly or moderately differentiated HCC. Simultaneous detection of two or more markers increases the overall sensitivity from 50% to 72% <ns0:ref type='bibr' target='#b66'>(53)</ns0:ref>. However, about 30% of HCC patients are still negative for these traditional tumor markers. In our study, DPM1 could be a potential prognostic biomarker for survivals of HCC patients. Therefore, it is possible to use DPM1 as an effective supplemental biomarker of liver cancer. The combined application of DPM1 and other already-existed biomarkers would greatly improve the early diagnosis and accurate prognosis of liver cancers. Our study also has some limitations. First, despite mRNA expressions of DPM1/2/3 were related to the prognosis of HCC, all the data performed in our research were obtained from the online website, further studies containing larger sample sizes are needed to confirm our results and to explore the clinical application of the DPMS in HCC treatment. Second, we did not assess the potential diagnostic and therapeutic roles of DPMS in HCC, so future studies are required to explore whether DPMS could be used as diagnostic markers or as therapeutic targets. Finally, we did not explore the potential mechanisms of DPMS in HCC. Future studies are worth to investigate the detailed mechanism between DPMS expression and HCC.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In this paper, we studied the expressions of DPM1/2/3 in tumor cells and its relationship with tumorigenesis for the first time. Our results showed that over-expressions of DPM1/2/3 were significantly associated with clinical cancer stages and pathological tumor grades in HCC patients. Besides, higher mRNA expressions of DPM1/2/3 were found to be significantly connected with OS in HCC patients. Moreover, high genetic alteration rate of DPM1/2/3 (41%) was also observed, and genetic alteration in DPM1/2/3 was associated with shorter OS in HCC patients, which provide a better understanding of molecular targets for improved liver cancer therapeutic strategies in the future. DPM1 was the most potential prognostic biomarker for liver cancer via cell experiment verified. To sum up, these results indicated that DPM1 could be a prognostic biomarker for survivals of HCC patients. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>The mRNA expressions of DPM1/2/3 in normal liver samples and CLD samples. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 6</ns0:note><ns0:p>The assessment of the diagnosis effect among DPM1/2/3 and existing markers in normal and HCC using the ROC curve. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 8</ns0:note><ns0:p>The enrichment analysis of DPMS and their similar genes in HCC (Metascape). </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure legend Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure legend Figure 1. Transcriptional expressions of (A) DPM1, (B) DPM2 and (C) DPM3 in different types of cancer diseases (UALCAN database). Blue: Normal; Red: Tumor. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.Figure 2. mRNA and protein expressions of DPMS in HCC and normal liver tissues. (A-C) mRNA expressions of DPM1, DPM2 and DPM3 in HCC tissues compared to normal samples (UALCAN database). ***p&lt;0.001. (D-F) Representative immunohistochemistry images of DPM1, DPM2 and DPM3 in HCC tissues and normal liver tissues (Human Protein Atlas).Figure 3. Association of mRNA expressions of DPMS with tumor grades and patients' individual cancer stages in HCC patients (UALCAN). (A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure legend Figure 1. Transcriptional expressions of (A) DPM1, (B) DPM2 and (C) DPM3 in different types of cancer diseases (UALCAN database). Blue: Normal; Red: Tumor. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.Figure 2. mRNA and protein expressions of DPMS in HCC and normal liver tissues. (A-C) mRNA expressions of DPM1, DPM2 and DPM3 in HCC tissues compared to normal samples (UALCAN database). ***p&lt;0.001. (D-F) Representative immunohistochemistry images of DPM1, DPM2 and DPM3 in HCC tissues and normal liver tissues (Human Protein Atlas).Figure 3. Association of mRNA expressions of DPMS with tumor grades and patients' individual cancer stages in HCC patients (UALCAN). (A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure legend Figure 1. Transcriptional expressions of (A) DPM1, (B) DPM2 and (C) DPM3 in different types of cancer diseases (UALCAN database). Blue: Normal; Red: Tumor. Abbreviations: BLCA, Bladder urothelial carcinoma; BRCA, Breast invasive carcinoma; CESC, Cervical squamous cell carcinoma; CHOL, Cholangiocarcinoma; COAD, Colon adenocarcinoma; ESCA, Esophageal carcinoma; GBM, Glioblastoma multiforme; HNSC, Head and Neck squamous cell carcinoma; KICH, Kidney chromophobe; KIRC, Kidney renal clear cell carcinoma; KIRP, Kidney renal papillary cell carcinoma; LIHC, Liver hepatocellular carcinoma; LUAD, Lung adenocarcinoma; LUSC, Lung squamous cell carcinoma; PAAD, Pancreatic adenocarcinoma; PRAD, Prostate adenocarcinoma; PCPG, Pheochromocytoma and Paraganglioma; READ, Rectum adenocarcinoma; SARC, Sarcoma; SKCM, Skin cutaneous melanoma; THCA, Thyroid carcinoma; THYM, Thymoma; STAD, Stomach adenocarcinomna; UCEC, Uterine corpus endometrial carcinoma.Figure 2. mRNA and protein expressions of DPMS in HCC and normal liver tissues. (A-C) mRNA expressions of DPM1, DPM2 and DPM3 in HCC tissues compared to normal samples (UALCAN database). ***p&lt;0.001. (D-F) Representative immunohistochemistry images of DPM1, DPM2 and DPM3 in HCC tissues and normal liver tissues (Human Protein Atlas).Figure 3. Association of mRNA expressions of DPMS with tumor grades and patients' individual cancer stages in HCC patients (UALCAN). (A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>The mRNA expressions of DPM1/2/3 in normal liver samples and CLD samples. (A-C) The expressions of DPM1/2/3 in normal and HBV-related liver cirrhosis. (D-F) The expressions of DPM1/2/3 in normal and HCV-related liver cirrhosis. (G-I) The expressions of DPM1/2/3 in normal and non-alcoholic steatohepatitis. Abbreviations: HBV-LC, HBV-related liver cirrhosis; HCV-LC, HCV-related liver cirrhosis, NASH, non-alcoholic steatohepatitis.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>The prognostic value of different expressed DPM1, DPM2 and DPM3 in HCC patients (GEPIA). (A-C) Overall survival curves of DPM1, DPM2 and DPM3. (D-F) Disease free survival curves of DPM1, DPM2 and DPM3.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>The assessment of the diagnosis effect among DPM1/2/3 and existing markers in normal and HCC using the ROC curve. (A-C) ROC curves and AUC values of DPM1, DPM2 and DPM3 respectively. (D-F) ROC curves and AUC values of AFP, GPC3 and TGF&#946;1 respectively.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Genetic alterations in DPMS and their association with OS and DFS in HCC patients (cBioPortal). (A) Summary of alterations in DPMS. (B) OncoPrint visual summary of alteration on a query of DPMS. (C) Kaplan-Meier plots comparing OS in cases with/without DPMS gene alterations. (D) Kaplan-Meier plots comparing DFS in cases with/without DPMS gene alterations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 8 .</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>The enrichment analysis of DPMS and their similar genes in HCC (Metascape). (A) Heatmap of Gene Ontology (GO) enriched terms colored by p-values. (B) Network of GO enriched terms colored by p-value, where terms containing more genes tend to have a more significant pvalue. (C) Heatmap of Kyoto Encyclopedia of Genes and Genomes (KEGG) enriched terms colored by p-values. (D) Network of KEGG enriched terms colored by p-value, where terms containing more genes tend to have a more significant p-value. (E) Protein-protein interaction (PPI) network and three most significant MCODE components form the PPI network. (F) Independent functional enrichment analysis of three MCODE components.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 9 .</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>The mRNA expression levels of (A) DPM1, (B) DPM2 and (C) DPM3 in normal liver cells and hepatoma cell lines. *p &lt;0.05, ***p &lt;0.001.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 1 Transcriptional</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 3 Association</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-C) Association of mRNA expressions of DPM1, DPM2 and DPM3 with tumor grades in HCC patients. (D-F) Relationship between mRNA expressions of DPM1, DPM2 and DPM3 and individual cancer stages of HCC patients. *p&lt;0.05, **p&lt;0.01, ***p&lt;0.001. PeerJ reviewing PDF | (2020:06:50511:2:0:CHECK 9 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-C) The expressions of DPM1/2/3 in normal and HBV-related liver cirrhosis. (D-F) The expressions of DPM1/2/3 in normal and HCV-related liver cirrhosis. (G-I) The expressions of DPM1/2/3 in normal and non-alcoholic steatohepatitis. Abbreviations: HBV-LC, HBV-related liver cirrhosis; HCV-LC, HCV-related liver cirrhosis; NASH, non-alcoholic steatohepatitis.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head>Figure 5 The</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-C) Overall survival curves of DPM1, DPM2 and DPM3. (D-F) Disease free survival curves of DPM1, DPM2 and DPM3. PeerJ reviewing PDF | (2020:06:50511:2:0:CHECK 9 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_15'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A-C) ROC curves and AUC values of DPM1, DPM2 and DPM3 respectively. (D-F) ROC curves and AUC values of AFP, GPC3 and TGF&#946;1 respectively. PeerJ reviewing PDF | (2020:06:50511:2:0:CHECK 9 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_16'><ns0:head>Figure 7 Genetic</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_17'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Summary of alterations in DPMS. (B) OncoPrint visual summary of alteration on a query of DPMS. (C) Kaplan-Meier plots comparing OS in cases with/without DPMS gene alterations. (D) Kaplan-Meier plots comparing DFS in cases with/without DPMS gene alterations. PeerJ reviewing PDF | (2020:06:50511:2:0:CHECK 9 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_18'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) Heatmap of Gene Ontology (GO) enriched terms colored by p-values. (B) Network of GO enriched terms colored by p-value, where terms containing more genes tend to have a more significant p-value. (C) Heatmap of Kyoto Encyclopedia of Genes and Genomes (KEGG) enriched terms colored by p-values. (D) Network of KEGG enriched terms colored by p-value, where terms containing more genes tend to have a more significant p-value. (E) Protein-protein interaction (PPI) network and three most significant MCODE components form the PPI network. (F) Independent functional enrichment analysis of three MCODE components.</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,280.87,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='30,42.52,224.62,525.00,170.25' type='bitmap' /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50511:2:0:CHECK 9 Oct 2020)Manuscript to be reviewed</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50511:2:0:CHECK 9 Oct 2020)</ns0:note> </ns0:body> "

"Dear Editor, Thank you very much for giving us the opportunity to revise our manuscript entitled 'DPM1 Expression as a Potential Prognostic Tumor Marker in Hepatocellular Carcinoma (Article ID: 50511) '. We have carefully revised the paper according to the good suggestions of the reviewers. The corrections have been highlighted in “Red” in the text. Our answers to the reviewers’ critiques are listed below. The comments of reviewer 1: The reviewer considered that major comments 1 (analysis between the relationship between the expression of DPM1/2/3 and the etiology for CLD) and 2 (analysis the relationship between DPM1/2/3 and already-existed markers such as AFP, DCP, and GPC3) was extremely important to improve the quality of your paper. Therefore, additional data based on further analyses, but not discussion, was requested. In addition, they could not respond a major comment 4 (improvement of Fig.5). Alteration of DPM1/2/3 genes would be divided into (Amplification, deletion, somatic mutation etc). Answers to reviewer 1: 1. Because HCC usually develops from chronic injured liver, the cause of chronic liver disease (CLD), such as viral infection, alcohol, fatty liver, is of importance. They should analyze the relationship between the expression of DPM1/2/3 and the etiology for CLD. Thank the reviewer for the good idea. Although it is difficult for us to obtain enough effective clinical data on the relationship between DPMS expression and the etiology of CLD, we have tried our best to analyze their relationships. The results are shown in Figure 4. 2. There is no comparative analysis between DPM1/2/3 and already-existed markers such as AFP, DCP, and GPC3. To highlight the significance of DPM1/2/3, additional analyses would be necessary. According to the reviewer’s suggestion, we have done the comparative analysis between DPM1/2/3 and already-existed markers such as AFP, GPC3, and TGFβ1. The results are shown in Figure 6. 3. It is hard to realize Fig.5. There is no information about genetic mutations. Amplification is one of the genetic alteration, but not mutation. Alteration of DPM1/2/3 genes would be divided into (Amplification, deletion, somatic mutation etc). We agree with the reviewer. As shown in Figure 7 ( original Figure 5), the genetic alteration of DPM1/2/3 mainly includes amplification, elevated mRNA and decreased mRNA. Genetic mutations of DPM1/2/3 account for a small proportion in genetic changes, therefore we mainly discussed the role of amplification not mutation of DPM1/2/3 in discussion (Please see Page 8/Line 310-315). In summary, we have revised our manuscript carefully according to the good suggestions and recommendations by the editor and reviewer. We hope that you find our resubmitted manuscript acceptable for publication at Peer J. Thank you, Sincerely, Ping Shi, Ph.D. Professor State Key Laboratory of Bioreactor Engineering East China University of Science and Technology 130 Meilong Road, Shanghai 200237, China +86-21-64251655 (phone), +86-21-64252920 (fax) "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Portable chest x-ray (pCXR) has become an indispensable tool in the management of Coronavirus Disease 2019 (COVID-19) lung infection. This study employed deep-learning convolutional neural networks to classify COVID-19 lung infections on pCXR from normal and related lung infections to potentially enable more timely and accurate diagnosis. This retrospect study employed deep-learning convolutional neural network (CNN) with transfer learning to classify based on pCXRs COVID-19 pneumonia (N=455) on pCXR from normal (N=532), bacterial pneumonia (N=492), and non-COVID viral pneumonia (N=552). The data was randomly split into 75% training and 25% testing, randomly. A five-fold crossvalidation was used for the testing set separately. Performance was evaluated using receiver-operating curve analysis. Comparison was made with CNN operated on the whole pCXR and segmented lungs. CNN accurately classified COVID-19 pCXR from those of normal, bacterial pneumonia, and non-COVID-19 viral pneumonia patients in a multiclass</ns0:p><ns0:p>model. The overall sensitivity, specificity, accuracy, and AUC were 0.79, 0.93, and 0.79, 0.85 respectively (whole pCXR), and were 0.91, 0.93, 0.88, and 0.89 (CXR of segmented lung). The performance was generally better using segmented lungs. Heatmaps showed that CNN accurately localized areas of hazy appearance, ground glass opacity and/or consolidation on the pCXR. Deep-learning convolutional neural network with transfer learning accurately classifies COVID-19 on portable chest x-ray against normal, bacterial pneumonia or non-COVID viral pneumonia. This approach has the potential to help radiologists and frontline physicians by providing more timely and accurate diagnosis.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Coronavirus Disease 2019 (COVID-19) is a highly infectious disease that causes severe respiratory illness <ns0:ref type='bibr' target='#b13'>(Hui et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b21'>Lu et al. 2020)</ns0:ref>. It was first reported in Wuhan, China in December 2019 <ns0:ref type='bibr' target='#b20'>(Li et al. 2020c)</ns0:ref> and was declared a pandemic on Mar 11, 2020 . The first confirmed case of coronavirus disease 2019 in the United States was reported from Washington State on January 31, 2020 <ns0:ref type='bibr' target='#b11'>(Holshue et al. 2020)</ns0:ref>. Soon after, Washington, California and New York reported outbreaks. COVID-19 has already infected 10 million, killed more than 0.5 million people, and the United States has become the worst-affected country, with more than 2.4 million diagnosed cases and at least 122,796 deaths (https://coronavirus.jhu.edu, assessed Jun 28, 2020). There are recent spikes of COVID-19 infection cases across many states and around the world and there will likely be second waves and recurrence.</ns0:p><ns0:p>A definitive test of COVID-19 infection is the reverse transcription polymerase chain reaction (RT-PCR) of a nasopharyngeal or oropharyngeal swab specimen <ns0:ref type='bibr' target='#b31'>(Tang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b32'>Wang et al. 2020)</ns0:ref>. Although RT-PCR has high specificity, it has low sensitivity, high false negative rate, and long turn-around time <ns0:ref type='bibr' target='#b31'>(Tang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b32'>Wang et al. 2020</ns0:ref>) (currently ~4 days although improvement and other tests are becoming available (CDC). By contrast, portable chest X-rays (pCXR) is convenient to perform, has a fast turnaround, and is well suited for imaging contagious Manuscript to be reviewed which compromises imaging quality. Thus, pCXR data under the COVID-19 pandemic circumstances are suboptimal and, thus, may be more challenging to interpret. Moreover, pCXR is increasingly read by non-chest radiologists in some hospitals due to increasing demands, resulting in reduced accuracy and efficiency. pCXR images contain important clinical features that could be easily missed by the naked eyes. Computer-aided methods can improve efficiency and accuracy of pCXR interpretations, which in turn provides more timely and relevant information to frontline physicians. Deep-learning artificial intelligence (AI) is increasingly used to analyze diagnostic images <ns0:ref type='bibr' target='#b6'>(Ehteshami Bejnordi et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b17'>LeCun et al. 2015)</ns0:ref>. AI has the potential to facilitate disease diagnosis, staging of disease severity and longitudinal monitoring of disease progression.</ns0:p><ns0:p>One common machine-learning algorithm is the convolutional neural network (CNN) <ns0:ref type='bibr' target='#b15'>(Krizhevsky et al. 2012)</ns0:ref>, which takes an input image, learns important features in the image such as size or intensity, and saves these parameters as weights and bias to differentiate types of images <ns0:ref type='bibr' target='#b30'>(Song et al. 2020)</ns0:ref>. CNN architecture is ideally suited for analyzing images. Moreover, many of the machine learning algorithms are trained to solve specific tasks, where models need to be rebuilt from scratch if the feature changes. Transfer learning overcomes such drawback by utilizing knowledge acquired for one task to solve related ones. Transfer learning is useful when dealing with small sample size data because the pre-trained weights improve efficiency and performance <ns0:ref type='bibr' target='#b3'>(Byra et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b28'>Samala et al. 2017</ns0:ref>). Many artificial intelligence (AI) algorithms based on deep-learning convolutional neural networks have been deployed for pCXR applications <ns0:ref type='bibr' target='#b9'>(Harris et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b10'>Heo et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b22'>Mekov et al. 2020</ns0:ref>) and these algorithms can be readily repurposed for COVID-19 pandemic circumstances. While there are already many papers describing prevalence and radiographic</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed features on pCXR of COVID-19 lung infection (see reviews <ns0:ref type='bibr' target='#b2'>(Bao et al. 2020</ns0:ref>)), there is a few AI papers <ns0:ref type='bibr' target='#b1'>(Apostolopoulos &amp; Mpesiana 2020;</ns0:ref><ns0:ref type='bibr' target='#b4'>Cohen et al. 2020a;</ns0:ref><ns0:ref type='bibr' target='#b7'>Elaziz et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b14'>Hurt et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b23'>Murphy et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b24'>Ozturk et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b26'>Pereira et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zhu et al. 2020a)</ns0:ref> COVID-19 infection on pCXR. The performance was evaluated using receiver-operating curve (ROC) analysis. Heatmaps were also generated to visualize and assessment the performance of the AI algorithm. sublayers. A sublayer is another layer within a convolutional layer that further filters images and passes information down to the next sublayers. The information collected from all the sublayers is compiled and sent to the next layer to make a cohesive prediction. A max-pooling layer was used after each step in the model to down sample the input and identify its important features based on the methods described in <ns0:ref type='bibr' target='#b27'>(Ren et al. 2020)</ns0:ref>. A max-pooling layer reduces the dimensionality of the image and allows for assumptions to be made about features contained in the sub regions <ns0:ref type='bibr' target='#b27'>(Ren et al. 2020)</ns0:ref>. All convolutional layers used rectified linear units (ReLUs) as an activation function because it adds a small number of learnable parameters <ns0:ref type='bibr' target='#b27'>(Ren et al. 2020)</ns0:ref>. Three fully connected layers were used, each having 4096 nodes. Fully connected layers compose some of the last few layers in a model and connect all the inputs from each layer to the activation unit of the next layer <ns0:ref type='bibr' target='#b27'>(Ren et al. 2020)</ns0:ref>. Dropout layers were used, along with the Softmax function, to prevent overfitting. Dropout layers work by randomly setting the edges of hidden neurons to 0 at each update of the training phase. The softmax function turns all the scores from the images into a normalized probability distribution, which helps make the final prediction <ns0:ref type='bibr' target='#b27'>(Ren et al. 2020)</ns0:ref>. For data analysis, batch sizes of 32 were used to limit computational expense and trained for 50 epochs.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data</ns0:head><ns0:p>Epochs can be thought of as iterations. Several optimizers were tested and Adams optimization function was found to yield the lowest validation loss. The learning rate was lowered from the recommended 0.01 to 0.001 to prevent overshooting the global minimum loss. Categorical cross</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed entropy was used as a loss function since the loss value decreases as the predicted probability converges to the actual label. CNN analysis was performed on the whole pCXR as well as virtually segmented lungs.</ns0:p><ns0:p>Lung segmentation was performed using a CNN architecture with 22 convolutional layers, 4 max- shown in Table <ns0:ref type='table' target='#tab_3'>1</ns0:ref>. The precision, recall, and F1 scores for the whole pCXR (Table <ns0:ref type='table' target='#tab_4'>2</ns0:ref>) showed good to excellent performance. For CNN with transfer learning performed on the whole pCXR, the overall sensitivity, specificity, accuracy, and AUC were 0.79, 0.93, and 0.79, .84 respectively.</ns0:p><ns0:p>For CNN performed on segmented lungs, the overall sensitivity, specificity, accuracy, and AUC were 0.91, 0.93, 0.88, 0.89 respectively. The performance was generally better using segmented lungs.</ns0:p><ns0:p>To visualize the spatial location on the images that the CNN networks were paying attention to for classification, heatmaps of the COVID-19 versus normal pCXR are shown in Manuscript to be reviewed of ground glass opacities and/or consolidations, but some hot spots were located outside the lungs.</ns0:p><ns0:p>For CNN performed on segmented lungs, the majority of the hot spots were reasonably localized to regions of ground glass opacities and/or consolidations, mostly as expected. There were a few pixels outside the lung that the algorithm paid attention to. These 'errors' could be due to small sample sizes. It learned from the training dataset and there may be information that the algorithm might consider important. Large sample size usually minimizes such 'error.'</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>This <ns0:ref type='bibr'>)</ns0:ref>. This method achieved accuracy rates of 96.09% and 98.09% for each of the respective datasets <ns0:ref type='bibr' target='#b7'>(Elaziz et al. 2020)</ns0:ref>. Note that patient</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed AUC and accuracy were not reported. AI has also been employed to stage pCXR disease severity against radiologist scores <ns0:ref type='bibr' target='#b4'>(Cohen et al. 2020a;</ns0:ref><ns0:ref type='bibr' target='#b35'>Zhu et al. 2020a)</ns0:ref>. Our study had one of the larger cohorts, balanced sample sizes, and multi-class model. Our approach is also amongst the simplest AI models with comparable performance index, likely facilitate immediate clinical translation.</ns0:p><ns0:p>Together, these studies indicate that AI has the potential to assist frontline physicians in distinguishing COVID-19 infection based on CXRs. Heatmaps are informative tools to visualize regions that CNN algorithm pays attention to for detection. This is particular important given AI operates on high dimensional space. Such heatmaps enable reality checks and make AI interpretable with respect to clinical findings. Our algorithm showed that the majority of the hotspots were highly localized to abnormalities within the lungs, i.e., ground glass opacity and/or consolidation, albeit imperfect. The majority of the above-mentioned machine learning studies to classify COVID-19 CXRs did not provide heatmaps.</ns0:p><ns0:p>We also noted that CNN on whole pCXR image resulted in some hot spots located outside the lungs. CNN of segmented lungs solved this problem. Another advantage of using segmented lung is reduced computational cost during training. Transfer learning also reduced computational cost, making this algorithm practical. The performance is generally better using segmented lungs.</ns0:p><ns0:p>Most COVID-19 positive patients showed significant abnormalities on pCXR. Some early studies have even suggested that pCXR could be used as a primary tool for COVID-19 screening in epidemic areas <ns0:ref type='bibr' target='#b0'>(Ai et al. 2020)</ns0:ref>, which could complement swab testing which still has long turnaround time and non-significant false positive rate. In some cases, imaging revealed chest abnormalities even before swab tests confirm infection <ns0:ref type='bibr'>(Fang et al. 2020, in press;</ns0:ref><ns0:ref type='bibr' target='#b18'>Li et al. 2020a</ns0:ref>).</ns0:p><ns0:p>In addition, pCXR can detect superimposed bacteria pneumonia, which necessitates urgent antibiotic treatment. pCXR can also suggest acute respiratory distress syndrome, which is Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Precision and recall rate and F1 score (whole CXR).</ns0:p><ns0:note type='other'>Figure legends</ns0:note><ns0:p>Table <ns0:ref type='table' target='#tab_4'>2</ns0:ref> shows the precision and recall rate and F1 score (whole CXR).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>patients and longitudinal monitoring of critically ill patients in the intensive care units because the equipment can be readily disinfected, preventing cross-infection. pCXR of COVID-19 infection has certain unique characteristics, such as predominance of bilateral, peripheral, and low lobes involvement, with ground-glass opacities with or without airspace consolidations as the disease progresses. These characteristics generally differ from other lung pathologies, such as bacterial pneumonia or other viral (non-COVID-19) lung infection. Based on CXR and laboratory findings, clinicians might start patients on empirical treatment before the RT-PCR results become available or even if the RT-PCR come back negative due to high false negative rate of RT-PCR. Early treatment in COVID-19 patients is associated with better clinical outcomes. Similarly, computed tomography (CT), which offers relatively more detailed features (such as subtle ground-glass opacity<ns0:ref type='bibr' target='#b19'>(Li et al. 2020b;</ns0:ref><ns0:ref type='bibr' target='#b33'>Xu et al. 2020</ns0:ref>)), has also been used in the context of COVID-19. However, CT suite and equipment are more challenging to disinfect, and thus it is much less suitable for examining patients suspected of or confirmed with contagious diseases in general and COVID-19 in particular. Longitudinal CT monitoring of critically ill patients in the intensive care units is also challenging. In short, pCXR has become an indispensable imaging tool in the management of COVID-19 infection, is often one of the first examinations a patient suspected of COVID-19 infection receives in the emergency room, and ideally used for longitudinal monitoring of critically ill patients in the intensive care units.The usage of pCXR under the COVID-19 pandemic circumstances is unusual in many aspects. For instance, pCXR is preferred as it can be used at the bedside without moving the patients, but the imaging quality is not as good as conventional CXR. In addition, COVID-19 patients may not be able to take full inspirations during the examination, obscuring possible pathology, especially in the lower lung fields. Many sicker patients may be positioned on the side PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>sources: This retrospective study used publicly available pCXR of i) COVID-19 infection, ii) non-COVID-19 viral infection, iii) bacterial pneumonia, and iv) normal subjects. The COVID-19 pCXR were downloaded from on May 27th, 2020<ns0:ref type='bibr' target='#b5'>(Cohen et al. 2020b</ns0:ref>). The original download contained 673 CT or pCXR images of COVID-19, SARS, acute respiratory distress syndromes, pneumocystis, streptococcus, legionella, Chlamydophila, E Coli, Klebsiella, lipoid, Varicella, and influenza. The labels for the data came from a metadata file associated with the open dataset. The final sample size for COVID-19 patients was 455 pCXR from 197 patients. We recognized that this dataset was a public, community-driven dataset and there are potential selection biases. A radiologist (B.S.) evaluated all images for quality and relevance and each case was COVID-19 positive based on available data. As a result of this evaluation, a few images that were deemed to be of poor quality, were excluded.The other datasets were taken from the established Kaggle chest X-ray image (pneumonia) dataset (https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia). Although the Kaggle database has a large sample size, we randomly selected a sample size comparable to that of COVID-19 to avoid asymmetric sample size bias that could skew sensitivity and specificity.The sample sizes chosen for bacterial pneumonia, non-COVID-19 viral pneumonia, and normal pCXR were 492, 552 and 532 patients, respectively. Similarly, a chest radiologist evaluated all images for quality. CNN: A CNN, a type of neural network, is ideally suited for analyzing images. In a standard CNN model, a filter (window) travels over each region of an image and looks for different features such as edges, colorations, patches, and more in order to classify an image into a certain category.Our CNN architecture was based on VGG16 (Figure1), a convolutional neural network<ns0:ref type='bibr' target='#b29'>(Simonyan &amp; A. 2014)</ns0:ref>, architecture was utilized for computation efficiency and ease to implement, for immediate translation potential. Our VGG16 architecture had 13 convolution layers that each run a series of filters over the image to extract important features. The VGG16 model was used because it was pretrained on the ImageNet database and properly employs transferlearning which makes the training process efficient. In other words, instead of having to learn all the relationships in an image from scratch, the model is already familiar with that when transfer learning is employed. The data was normalized first by transforming all files into RGB images and resizing them into 224x224 pixels to make them compatible with the VGG16 framework. Next, the images were one-hot-encoded and randomly split into 75% training and 25% testing. One hot encoding means to turn all the categorical labels into numerical values containing zeroes and ones to make it much easier for the computer to read. VGG16 implements 13 convolutional layers: 5 Max Pooling layers and 3 Dense layers which sum up to 21 layers and 16 weight layers (Ren et al. 2020). Conv 1 has 64 filters while Conv 2 has 128 filters, Conv 3 has 256 filters while Conv 4 and Conv 5 have 512 filters. The first two layers have 2 sublayers while the 4th and 5th layers have 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Figure2shows examples of pCXR from a normal subject, patients with bacterial</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. The CNN algorithm was able to localize the area of pathology on pCXR. For CNN</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>cohorts were highly asymmetric. Murphy et al. used an artificial intelligence to classify COVID-19 CXRs (N=223) from non-COVID-19 CXRs (N=231) with an 0.81 AUC and they also showed that AI outperformed expert readers (Murphy et al. 2020). Ozturk et al. used an AI model to perform multiclass classification for COVID-19 (N=127) vs. No-Findings (N=500) vs. Pneumonia (N=500) as well as a binary classification for COVID vs. No-Findings which achieved 87.02%and 98.08% accuracies, respectively<ns0:ref type='bibr' target='#b24'>(Ozturk et al. 2020)</ns0:ref>. Pereira et al. performed a multiclass classification and a hierarchical classification for COVID-19 vs pneumonia vs no-finding using resampling algorithms, texture descriptors, and CNN. This model achieved a F1-Score of 0.65 for the multiclass approach and F1 score of 0.89 for the hierarchical classification<ns0:ref type='bibr' target='#b26'>(Pereira et al. 2020</ns0:ref>).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>associated with severe negative outcomes and necessitates immediate treatment. Together with the potential widespread shortage of intensive care units and mechanical ventilators in many hospitals, pCXR may play a critical role in decision-making. A timely implementation of AI methods could help to realize the full potential of pCXR in this COVID-19 pandemic. This pilot proof-of-principal study has several limitations. This is a retrospective study with a small sample size and the data sets used for training had limited alternative diagnoses. Although the Kaggle database has a large sample size for non-COVID-19 CXR, we chose the sample sizes to be comparable to that of COVID-19 to avoid asymmetric sample sizes that could skew sensitivity and specificity. Future studies will need to increase the COVID-19 sample size and include additional lung pathologies. The spatiotemporal characteristics on pCXR of COVID-19 infection and its relation to clinical outcomes are unknown. Future endeavors could include PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)Manuscript to be reviewed developing AI algorithms to stage severity, and predict progression, treatment response, recurrence, and survival, to inform and advise risk management and resource allocation associated with the COVID-19 pandemic, with inclusion of clinical variables in predictive models<ns0:ref type='bibr' target='#b16'>(Lam et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b34'>Zhao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b36'>Zhu et al. 2020b)</ns0:ref>.In conclusion, deep learning convolutional neural networks with transfer learning accurately classify COVID-19 pCXR from pCXR of normal, bacterial pneumonia, and non-COVID viral pneumonia patients in a multi-class neural nwork model. This approach has the potential to help radiologists and frontline physicians by providing efficient and accurate diagnosis.PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. VGG16 architecture with 16 weighted layers including 3 fully connected layers.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Examples of chest radiographs (a) normal, (b) COVID-19 viral pneumonia, (c) non-</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: CNN (a) training and (b) validation loss and accuracy. Loss decreases and accuracy</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: (a) pCXR from a COVID-19 patient, (b) the corresponding segmented lung, (c) heatmap</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Figure 2 CXRFigure 2 :</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>Figure 4 HeatmapFigure 4 :</ns0:head><ns0:label>44</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1 (on next page) Confusion tableTable 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Confusion table showing the multiclass CNN classification (whole CXR)</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Confusion table showing the multiclass CNN classification (whole CXR)</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Normal</ns0:cell><ns0:cell>COVID-19</ns0:cell><ns0:cell>Non-COVID-pneumonia 19 viral</ns0:cell><ns0:cell>Bacterial pneumonia</ns0:cell></ns0:row><ns0:row><ns0:cell>Normal</ns0:cell><ns0:cell>122</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>2</ns0:cell></ns0:row><ns0:row><ns0:cell>Covid19</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>102</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>6</ns0:cell></ns0:row><ns0:row><ns0:cell>Non-COVID-19 viral</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>94</ns0:cell><ns0:cell>20</ns0:cell></ns0:row><ns0:row><ns0:cell>Pneumonia</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacterial pneumonia</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>85</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:52307:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>shows the precision and recall rate and F1 score (whole CXR).</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Precision</ns0:cell><ns0:cell>Recall</ns0:cell><ns0:cell>F1 -score</ns0:cell></ns0:row><ns0:row><ns0:cell>Normal</ns0:cell><ns0:cell>0.82</ns0:cell><ns0:cell>0.85</ns0:cell><ns0:cell>0.84</ns0:cell></ns0:row><ns0:row><ns0:cell>Covid19</ns0:cell><ns0:cell>0.94</ns0:cell><ns0:cell>0.87</ns0:cell><ns0:cell>0.91</ns0:cell></ns0:row><ns0:row><ns0:cell>Non-covid19 viral</ns0:cell><ns0:cell>0.65</ns0:cell><ns0:cell>0.71</ns0:cell><ns0:cell>0.68</ns0:cell></ns0:row><ns0:row><ns0:cell>pneumonia</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacterial pneumonia</ns0:cell><ns0:cell>0.75</ns0:cell><ns0:cell>0.71</ns0:cell><ns0:cell>0.73</ns0:cell></ns0:row></ns0:table></ns0:figure> </ns0:body> "

"Dear Editors and reviewers, Thank you for your helpful and constructive comments. Please find the point-by-point responses and major changes are tracked in the manuscript. We believe we have responded to all comments positively. Reviewer: Yara Seif Basic reporting no comment Experimental design It is unclear whether the labels they used originated from the metadata associated with the publicly available pCXR images. Yes. The labels for the data came from a metadata file associated with the open dataset. The authors mention that “a chest radiologist evaluated all [] images for quality”. Were any images filtered out or relabeled as a result of this step? A few images were excluded if image quality is poor. No other filter or relabeling were done. Did the radiologist re-classify the images? No. If that’s the case, how does their input compare with the starting classification? N/A. Were there other factors that were used to diagnose each disease type? No. The labels for the data came from a metadata file associated with the open dataset. For example, in reference 19, were symptoms and/or RT-PCR used in conjunction with the examination of pCXR lung images? We only used RT-PCR results as covid19 positivity The authors mention splitting their data set into training and testing. It would benefit the reader to know how, for example, did the authors use random sampling, stratified sampling, etc. ? Randomly sampled. In addition, was five-fold cross-validation implemented for the test set separately? Yes. Randomly sampled and executed 5 times. Validity of the findings no comment Comments for the author I have read the study entitled “Deep-learning convolutional neural networks with transfer learning 2 accurately classify COVID19 lung infection on portable chest 3 radiographs” by Kikkisetti et al. In this study, the authors develop a deep learning convolutional neural network able to classify pCXR lung images into 1) COVID-19 infections, 2) other viral infections, 3) bacterial infections, and 4) normal. They report encouraging quality metrics, specifically for the CNN that was trained on segmented lung pCXRs. I especially commend the authors for clearly outlining their methodology as well as the limitations of their study in the discussion section.  With that said, one minor comment I have regards the training data. It is unclear whether the labels they used originated from the metadata associated with the publicly available pCXR images. The authors mention that “a chest radiologist evaluated all [] images for quality”. Were any images filtered out or relabeled as a result of this step? Did the radiologist re-classify the images? If that’s the case, how does their input compare with the starting classification? Were there other factors that were used to diagnose each disease type? For example, in reference 19, were symptoms and/or RT-PCR used in conjunction with the examination of pCXR lung images? The authors mention splitting their data set into training and testing. It would benefit the reader to know how, for example, did the authors use random sampling, stratified sampling, etc. ? In addition, was five-fold cross-validation implemented for the test set separately? These comments are addressed above. Thank you. Reviewer 2 Basic reporting No comments Experimental design no comments Validity of the findings no comments Comments for the author This manuscript by Kikkisetti et al. reported their study on Portable chest x-ray (pCXR) combined with deep-learning convolutional neural networks (CNN) to classify COVID-19 lung infections. Segmented lungs showed better performance compared to the whole pCXR. In general, the manuscript is clearly presented and written, and contains appropriate introductory material, methods, statistics and reasonable. As noted above, the entire study sounds entirely plausible and provides useful information for further COVID19 lung infection classify at this special period. However, there are several minor concerns as described as follows. I would suggest the paper being accepted for publication if the authors can incorporate these comments in their revision. There are several concerns proposed as follows: Minor concerns:  1. The abstract is very good and has clear logical analysis, but the results of the paper are less clearly described. For example, In the abstract, there are many data COVID-19 pneumonia (N=455), normal (N=532), bacterial pneumonia (N=492), and non-COVID viral pneumonia (N=552). But how these data used to analyze and got the results should clearly present in the RESULTS part and also labeled in the table legend. The results part not just describe the figures and tables, it should follow your research logics and provided the information why you did this experiment and how you did it and what result you got. We is now revised. 2. On line 205, “but some hot spots were located outside the lungs”, could you explain why this situation appears and how to improve this? The following was added. There were a few pixels outside the lung that the algorithm paid attention to. Theese “error” could be due to small sample sizes. It learned from the training dataset and there may be information that the algorithm might consider important. Large sample size usually minimizes such “error.” 3. For Figure 2 and Figure 4, you just give some examples, that’s fine. But would you give a semiquantitative analysis to show how many samples are represented by the images in each figure? We appreciate the comment. The extent of opacity is highly heterogeneous and varied greatly across patients. It is generally not possible to subjectively categorize. We made the data public so readers can evaluate. Thank you. 4. You may move Figure 1 to supplemental materials. Although we agree, the paper is not very long and there are not too many figures. If that is ok, we would like to leave it as is. Reviewer 3 Basic reporting Only minor english language use/punctuation issues e.g. lines 52,110,111,112. Thank you. They are fixed. In the few months since the manuscript was written, the number of US COVID-19 related deaths has almost doubled. In the revision this should be updated. Lines 55 and 56. We understand but we would like to keep as is for readers to know when this study was done. Lines 60-62 the authors state that PRC has a low sensitivity, high false negative rate and long turn-around time of 4 days. The authors should double check if this is uniformly still the case or is unique to the USA. We qualified that this was the case in early pandemic. I agree this was vague. Experimental design Description of deep learning convolutional neural networks. For readers not familiar with convolutional neural networks there needs to be more detail explaining how it all works. The authors also need to explain what the following jargon means: VGG16 model one-hot-encoded convolutional layers and sublayers filters max-pooling dropout layers  soft max function etc, etc Thank you. More information have been provided. Validity of the findings No comment Comments for the author This timely paper nicely demonstrates how with relative ease, CNN and be used on portable chest x-rays for assessment and analysis of a new disease when clinical experience is limited. This could change management in sick patients and be life saving. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Gymnosperms such as ginkgo, conifers, cycads, and gnetophytes are vital components of land ecosystems, and they have significant economic and ecologic value, as well as important roles as forest vegetation. In this study, we investigated the structural variation and evolution of chloroplast transfer RNAs (tRNAs) in gymnosperms. Chloroplasts are important organelles in photosynthetic plants. tRNAs are key participants in translation where they act as adapter molecules between the information level of nucleic acids and functional level of proteins. The basic structures of gymnosperm chloroplast tRNAs were found tohave family-specific conserved sequences. The tRNA &#936;-loop was observed to contain a conforming sequence, i.e., U-U-C-X-A-X2. In gymnosperms, tRNA Ile was found to encode a 'CAU' anticodon, which is usually encoded by tRNA Met . Phylogenetic analysis suggested that plastid tRNAs have a common polyphyletic evolutionary pattern, i.e., rooted in abundant common ancestors. Analyses of duplication and loss events inchloroplast tRNAs showed that gymnosperm tRNAs have experienced little more gene loss than gene duplication. Transition and transversion analysis showed that the tRNAs are iso-acceptor specific and they have experienced unequal evolutionary rates. These results provide new new insights into the structural variation and evolution of gymnosperm chloroplast tRNAs, which may improve our comprehensive understanding of the biological characteristics of the tRNA family.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Gymnosperms originated in the Paleozoic Devonian Period (about 385 million years ago), and they are key groups in terms of the transformation from spore reproduction to seed reproduction in higher plants <ns0:ref type='bibr' target='#b14'>(Gerrienne et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Crisp &amp; Cook, 2011)</ns0:ref>. According to the latest phylogenetic classification, gymnosperm species aredivided into eight orders, 12 families, 84 genera, and more than 1,000 species <ns0:ref type='bibr' target='#b64'>(Wang &amp; Ran, 2014)</ns0:ref>. Gymnosperms include ginkgo, cycads, conifers, and gnetophytes, which are grown in forests as important timber speciesand they provide raw materials for human usage,such as fiber, resin, and tannin <ns0:ref type='bibr' target='#b3'>(Christenhusz et al., 2010)</ns0:ref>. In addition, gymnosperms include some important threatened plants, where 40% are at high risk of extinction <ns0:ref type='bibr' target='#b13'>(Forest et al., 2018)</ns0:ref>. Recent phylogenetic and evolutionary studies ofgymnosperms have demonstrated the rapid evolution of mitochondrial (mt) genes and provided further evidence of sister relationship between conifers and Gnetales <ns0:ref type='bibr' target='#b51'>(Ran, Gao &amp; Wang, 2010)</ns0:ref>. The high levels of genetic diversity and population differentiation among the Pinus species in gymnosperms have been studied based on plastid DNA markers <ns0:ref type='bibr' target='#b40'>(Liu et al., 2014)</ns0:ref>. Other studies have indicated patterns related to the physiological ecology, phylogenetic relationships, and population genetic structure of gymnosperm species <ns0:ref type='bibr' target='#b67'>(Yu et al., 2014;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b8'>Dong et al., 2016)</ns0:ref>. However, these studies mainly considered the phylogeny and evolution at the whole populations level. Thus, the detailed evolutionary characteristics of gymnosperms still need to be elucidated.</ns0:p><ns0:p>Chloroplasts are the site of photosynthesis and of various essential metabolic pathways, e.g., fatty acid and amino acid biosynthesis and the assimilation of nitrogen, sulfur, and selenium <ns0:ref type='bibr'>(Wise &amp; Hoober, 2006;</ns0:ref><ns0:ref type='bibr'>Des Marais,2000;</ns0:ref><ns0:ref type='bibr' target='#b31'>Knorr &amp; Heimann, 2001;</ns0:ref><ns0:ref type='bibr' target='#b49'>Pilon-Smits et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b17'>Guo et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b33'>Kretschmer, Croll &amp; Kronstad, 2017)</ns0:ref>. It is generally recognized that chloroplasts are derived from proto-eukaryotic symbiotic cyanobacteria that internalized in eukaryotic cells <ns0:ref type='bibr' target='#b21'>(Hiroki &amp; Daisuke, 2018</ns0:ref>) and evolved into central organelles. Chloroplasts have their own genome encoding about 100 proteins and they are maternally inherited organelles in most angiosperm plants <ns0:ref type='bibr' target='#b1'>(Abdallah, Salamini &amp; Leister, 2000;</ns0:ref><ns0:ref type='bibr' target='#b20'>Heuertz et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b4'>Civan et al., 2014)</ns0:ref>. Among gymnosperms, paternal plastid inheritance is the typical characteristic of conifers <ns0:ref type='bibr'>(Faur&#233; et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kaundun &amp; Matsumoto, 2011)</ns0:ref>. Studies have shown that the chloroplast genome is quite conserved with an average evolutionary rate of 0.2-1.0&#61620;10 -9 per site per year, which is only one-fifth of that for the nuclear genome <ns0:ref type='bibr' target='#b9'>(Drouin, Daoud &amp; Xia, 2008;</ns0:ref><ns0:ref type='bibr' target='#b10'>Duchene &amp; Bromham, 2013)</ns0:ref>. The chloroplast genome is a covalently closed circular structure with four parts comprising the large single copy region small single copy region, inverted repeat region A (IRa), and inverted repeat region B (IRb). The two IRs have the same sequence but in the opposite direction <ns0:ref type='bibr' target='#b63'>(Wang et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b41'>Logacheva et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b19'>Hereward et al., 2018)</ns0:ref>. Due to the independent evolution of the chloroplast genome, it is possible to construct a molecular phylogenetic tree using the chloroplast genome and without requiring any other data. Data analysis based on the conserved evolution of plastids is highly valuable for phylogenetic studies <ns0:ref type='bibr' target='#b28'>(Kim &amp; Suh, 2013)</ns0:ref> because it can provide reliable and useful phylogenetic information.</ns0:p><ns0:p>The relative completeness and independence of the chloroplast genome means that it can provide valuable material for researchpurposes.</ns0:p><ns0:p>The universal genetic code and transfer RNAs (tRNAs) were discovered over 50 years ago, thereby transforming the life science by providing an analytical framework for understanding living systems <ns0:ref type='bibr' target='#b55'>(Schimmel, 2017)</ns0:ref>. tRNAs undergo numerous post-transcriptional nucleotide modifications and they exhibit abundant chemical diversity where the bases experience methylation, formylation, and other modifications <ns0:ref type='bibr' target='#b59'>(Suzuki &amp; Suzuki, 2014)</ns0:ref>. Chemical nucleotide modifications are frequent in tRNAs and they are important for the structure, stability, correct folding, aminoacylation, and decoding. For example, a previous analysis of the chemically synthesized f 5 C34-modified anticodon loop of human mt-tRNA Met showed that f 5 C34 contributes to the anticodon domain structure of the mt-tRNA <ns0:ref type='bibr' target='#b44'>(Lusic et al., 2008)</ns0:ref>.</ns0:p><ns0:p>tRNAs comprise sequences of less than 100 polynucleotides that fold into a clover-type secondary structure and then into an L-shaped tertiary structure <ns0:ref type='bibr' target='#b66'>(Wilusz, 2015)</ns0:ref>. The secondary structure of tRNAs comprises different arms as well as loops, i.e., the D-arm, acceptor arm, anticodon arm, pseudouridine arm (&#936;-arm), D-loop, variable arm, anticodon loop, and pseudouridine loop (&#936;-loop) <ns0:ref type='bibr' target='#b16'>(Gieg&#233;, Puglisi &amp; Florentz, 1993;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mizutani &amp; Goto, 2000)</ns0:ref>. This unique structure allows tRNA to act as important bridges between the information level of nucleic acids and functional level of proteins. The vital components of tRNAs comprise an anticodon region that discerns the messenger RNA carried by the specific codons, a 3&#8242;-CCA tail for attaching to the cognate amino acid, the &#936;-arm , and a &#936;-loop that has a relationship with the ribosome machinery <ns0:ref type='bibr' target='#b30'>(Kirchner &amp; Ignatova, 2014)</ns0:ref>. The different length of tRNAs are due todifferences in the composition of the D-loop and variable loop. Asymmetric combinations and the divided segments in tRNA genes allow us to understand the diversity of tRNA molecules. Previous studies have shown that tRNA species fulfill various functions in cellular homeostasis, the adaptation of cellular functions to changing environments, regulation of gene expression and epigenetics, biogenesis, and even biological disease <ns0:ref type='bibr' target='#b52'>(Ribasd &amp; Dedon, 2014;</ns0:ref><ns0:ref type='bibr' target='#b25'>Kanai, 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Schimmel, 2017)</ns0:ref>. These diverse functional tRNA genes were formed via several mutation, reorganization, and duplication events that during their evolution. The evolutionary relationships determined between cyanobacteria and monocots show that tRNAs evolved polyphyletically and they originated from multiple common ancestors with a high rate of gene loss <ns0:ref type='bibr'>(Mohanta et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b47'>Mohanta et al., 2019)</ns0:ref>. Nevertheless, the basic details of the tRNAs in plant chloroplasts still need to be elucidated and on the diverse evolutionary features of gymnosperm tRNAs are still unclear.</ns0:p><ns0:p>In this study, we assessed all of the chloroplast genomes in 12 families of gymnosperms from eight orders. The main aims of this study were as follows: (1) to determine the diversification of nucleotides in the secondary structure of gymnosperm tRNAs; (2) to identify the detailed genomic features of chloroplast tRNAs; (3) to assess the evolutionary relationships among different chloroplast tRNAs; and (4) to evaluate the duplication or loss events that occurred in all of the tRNAs considered. Our findings provide important insights into the biological characteristics and evolutionary variation of the tRNA family.</ns0:p><ns0:p>Structural analysis of chloroplast tRNAs ARAGORN <ns0:ref type='bibr' target='#b37'>(Laslett &amp; Canback, 2004)</ns0:ref> and tRNAScan-SE software <ns0:ref type='bibr' target='#b42'>(Lowe &amp; Eddy, 1997)</ns0:ref> were employed to analyze the sequences and the secondary structure of tRNAs in the chloroplast genomes of the selected gymnosperm plants. The default parameters were set in ARAGORN software. The parameters for tRNAScan-SE were set as: sequence source, bacterial; search mode, default; query sequences, formatted (FASTA); and genetic code for tRNA isotype prediction, universal. Phylogenetic tree construction A phylogenetic tree was constructed for all of the tRNAs using MEGA7.0 software <ns0:ref type='bibr' target='#b34'>(Kumar et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b35'>Kumar, Stecher &amp; Tamura, 2016)</ns0:ref>. To study the evolutionary details of chloroplast tRNAs in gymnosperm species, an alignment file for tRNAs was achieved by clustal omega software before the phylogenetic tree was constructed. MEGA7 software was obtained with CLUSTAL Omega software before constructing the phylogenetic tree. MEGA7 software was used to transform the alignment file into MEGA format. The phylogenetic tree was constructed with the following parameters: phylogeny reconstruction of analysis, maximum likelihood model, bootstrap method in phylogeny test, 1000 bootstrap replicates, nucleotides type, gamma distributed with invariant sites (G+I) model, five discrete gamma categories, partial deletion for gaps/missing data treatment, 95 % site coverage cut-off, and very strong for branch swap filter.</ns0:p><ns0:p>Transition/transversion analysis The sequences of the tRNA isotypes were aligned to determine the transition and transversion rates for chloroplast tRNAs in gymnosperm plants. The files covering all 20 types of tRNAs were transformed into the MEGA file format and analyzed separately using MEGA7.0 software <ns0:ref type='bibr' target='#b36'>(Kumar, Tamura &amp; Nei, 1994)</ns0:ref>. The transition and transversion rates were analyzed for tRNAs with the following parameters: substitution pattern estimation (ML) analysis, automatic (neighbor-joining tree), maximum likelihood statistical method, nucleotide substitution type, Kimura two-parameter model, gamma distributed (G) site rates, five discrete gamma categories, partial deletion of gaps/missing data treatment, 95% of site coverage cut-off, and very strong branch swap filter.</ns0:p><ns0:p>Loss and duplication events analysis for tRNA genes In order to investigate the duplication or loss events in tRNA genes, the NCBI taxonomy browser was utilized to construct the whole species tree for the 12 gymnosperm species considered. The phylogenetic tree conducted in the evolutionary study was employed as gene tree. The gene tree for the tRNAs and species tree for the gymnosperm species were submitted to Notung 2.9 software <ns0:ref type='bibr' target='#b2'>(Chen, Durand &amp; Farach-Colton, 2000)</ns0:ref>, and then reconciled to discover duplicated and lost tRNA genes in the chloroplast genomes of gymnosperms.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Genomic features of gymnosperm chloroplast tRNAs</ns0:head><ns0:p>Sequences were analyzed to identify the genomic tRNAs in the chloroplast genomes of 12 gymnosperm species comprising C. <ns0:ref type='bibr'>debaoensis, D. spinulosum, G. biloba, C. deodara, W. nobilis, R. piresii, S. verticillata, C. lanceolate, T. mairei, W. mirabilis, G. gnemon, and E. equisetina</ns0:ref>, which were obtained from the NCBI database (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The results showed that the length of the chloroplast tRNAs vary from the smallest with 64 nucleotides (nt) (tRNA Met -CAU inT. mairei) to the largest with 96 nt (tRNA Tyr -AUA in W. nobilis, C. deodara, and G. biloba) (Data S1). We found that the chloroplast genomes of gymnosperm plants encode 28 to 33 tRNAs (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>), where D. spinulosum, C. deodara, and S. verticillate encode 31 anticodons, W. nobilis, R. piresii, and G. gnemon encode 32 tRNA isotypes, and G. biloba, C. lanceolate, and W. mirabilis encode 33 tRNAs. Other species comprising T. mairei, E. equisetina and C. debaoensis encode 28, 28, 30 tRNA isotypes, respectively (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). tRNA Ala was not found in R. piresii and T. mairei, and tRNA Val was not detected in T. mairei. We also observed that all of the species do not encode selenocysteine and its suppressor tRNA (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Overall, tRNA Ser (in W. nobilis) and tRNA Arg (in W. mirabilis) are the most abundant (four types) followed by tRNA Leu (three types) (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p><ns0:p>Variations in structures of chloroplast tRNAs Some tRNAs with a loop structure in the variable region were found to be encoded in the gymnosperm chloroplast genomes (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>). A novel tRNA lacking the D-arm was found in tRNA Gly in W. nobilis (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>). As shown in Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>, tRNA Leu , tRNA Ser , and tRNA Tyr contain a loop structure that is similar to the anti-codon loop in the variable region of tRNAs. In these tRNAs, the anticodon loop of tRNA Ser contains the conserved consensus sequence X-U-X-G-A-A-X and tRNA Leu has the consensus sequence G-U-U-A-X2-A. The variable loop region contains 3 to 7 nucleotides nt in tRNA Ser , tRNA Tyr to tRNA Leu . The stem of the variable loop region contains 0 to 7 nt. tRNA Ser and tRNA Leu have variable loop regions with 6 nt and 3 nt, respectively (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>). The novel loop structures identified in this study may play important functions during the protein translation process in chloroplasts.</ns0:p></ns0:div> <ns0:div><ns0:head>Chloroplast tRNAs contain 25 to 31 anticodons</ns0:head><ns0:p>The genomes of the species analyzed were found to contain at least one tRNA Met -CAU and tRNA fMet -CAU anticodon. All of the gymnosperm chloroplast tRNAs encode 25 to 31 anticodons (Table <ns0:ref type='table'>3</ns0:ref>), whereE. equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S. verticillate, and C. lanceolate encode 28 anticodens, and D. spinulosum, C. deodara, W. mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis and G. gnemon encode 30 anticodons, and G. biloba encodes 31anticodons (Table <ns0:ref type='table'>3</ns0:ref>). tRNA Arg -CCG was present in the genomes of nine gymnosperm species but absent from C. lanceolate, T. mairei, and E. equisetina, while tRNA Gly -UCC was lacking from C. debaoensis, S. verticillate, D. spinulosum, C. lanceolate, T. mairei, and E. equisetina (Table <ns0:ref type='table'>3</ns0:ref>).The most abundant anticodons found in thechloroplast genomes were tRNA Ala -UGC, tRNA Gly -GCC, tRNA Gly -UCC, tRNA Pro -UGG, tRNA Pro -GGG, tRNA Thr -GGU, tRNA Thr -UGU, tRNA Val -GAC, tRNA Val -UAC, tRNA Ser -GGA, tRNA Ser -UGA, tRNA Ser -GCU, tRNA Arg -ACG, tRNA Arg -CCG, tRNA Arg -UCU, tRNA Leu -UAG, tRNA Leu -CAA, tRNA Leu -UAA, tRNA Phe -GAA, tRNA Asn -GUU, tRNA Lys -UUU, tRNA Asp -GUC, tRNA Glu -UUC, tRNA His -GUG, tRNA Gln -UUG, tRNA Ile -CAU, tRNA Met -CAU, tRNA Tyr -GUA, tRNA Cys -GCA, and tRNA Trp -CCA (Table <ns0:ref type='table'>3</ns0:ref>). Two tRNA Trp isoacceptors present in E. equisetina was one more than other gymnosperm species analyzed in this study..</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved gymnosperm chloroplast tRNAs</ns0:head><ns0:p>The clover leaf-like secondary structure of a tRNA is shown in Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>. In the study, we found that most tRNAs contain a 'G' asthe first nucleotide in the D-arm, except for tRNA Lys , tRNA Met , tRNA Pro , tRNA Thr , tRNA Tyr , and tRNA Val . 'A' is present in the first and the last position of the D-loop. In addition, in the final two positions of the &#936;-arm, all of the tRNAs were found to have conserved 'G-G' nucleotides, except for tRNA Arg (Table <ns0:ref type='table' target='#tab_6'>4</ns0:ref>). Small conserved consensus sequences were found in the &#936; region according to multiple sequence alignment in 20 members of the tRNA gene family. The &#936;-loop in tRNAs was found to contain a conserved sequence comprising U-U-C-X-A-X2 (Table <ns0:ref type='table' target='#tab_6'>4</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Diversification of tRNAs structures</ns0:head><ns0:p>PeerJ reviewing <ns0:ref type='table' target='#tab_8'>PDF | (2019:11:43190:1:1:NEW 25 May 2020)</ns0:ref> Manuscript to be reviewed</ns0:p><ns0:p>The diverse arms and loops of tRNAs allow the regulation and control of protein translation. Each arm and loop has a specific nucleotide composition. Our analysis based on 373 tRNAs showed that the acceptor arm of chloroplast tRNAs contains 1 nt to 7 nt, where 359 have 7 nt, 12 have 6 nt, and the remaining tRNAs contain less than 5 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The D-arms were found to contain 3 or 4 nt, with a stable 'G' in the initial position and 'C' in the last position. The Dloops usually contain 7 nt to 11 nt with conserved 'A' nucleotides at the two end locations. The anticodon arms of chloroplast tRNAs mainly contain 5 nt (91.69%). We found that 371 (99.46%) tRNAs contain 7 nt in their anticodon loop, thereby indicating that the sequence of the anticodon loop is highly conserved (Table <ns0:ref type='table' target='#tab_6'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The variable loops of different tRNAs contain 4 nt to 23 nt, wherethose in tRNA Ala , tRNA Asn , tRNA Asp , TtRNA His , tRNA Met , tRNA Phe , tRNA Pro , tRNA Leu and tRNA Trp contain 5 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The &#936;-arm contains 5 nt in most of the gymnosperm chloroplast tRNAs, except for tRNA Ala and some of the tRNA Trp , tRNA Gly and tRNA Arg chloroplast tRNAs. The &#936;-loops of most tRNAs contain 7 nt, apart from tRNA Ala (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gymnosperm chloroplast tRNAs derived from multiple common ancestors</ns0:head><ns0:p>The phylogenetic tree demonstrated the presence of three major clusters covering 64 groups and the different types of all tRNAs (as shown by the different strings in Fig. <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>). We detected 37 groups in cluster I, five in cluster II, and 22 groups in cluster III. Cluster I contains tRNA Ser , tRNA Tyr , tRNA His , tRNA Gln , tRNA Thr , tRNA Pro ,tRNA Gly , tRNA Met , tRNA Asp , tRNA Arg , tRNA Ala ,tRNA Cys , tRNA Lys , tRNA Glu , tRNA Ile , tRNA Asn , tRNA Val , tRNA Leu , and tRNA Trp . Cluster II contains tRNA His , tRNA Ser , tRNA Tyr , and tRNA Leu . Cluster III contains tRNA Leu , tRNA Ile , tRNA Gly , tRNA Thr , tRNA Ser , tRNA Val , tRNA Glu , tRNA Lys , tRNA Cys , tRNA Gln , tRNA His , tRNA Arg , tRNA Phe , tRNA Ala , and tRNA Met (Fig. <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>). tRNA Ser , tRNA His , and tRNA Leu are present in cluster I but also in cluster II and cluster III, thereby suggesting that these tRNAs evolved from multiple lineages. Most of the tRNAs were found to form more than one group in the phylogenetic tree. In cluster I, the tRNAs that formed two groups in the phylogenetic tree were identified as tRNA Tyr , tRNA Gln , tRNA Met , tRNA Asp , tRNA Ala , tRNA Lys , tRNA Ile , and tRNA Trp , whereas those that clustered to form three groups were determined as tRNA Ser , tRNA Pro , tRNA Arg , tRNA Glu , tRNA Asn , tRNA Val , and tRNA Leu . Moreover, tRNA Thr clustered into four groups. In cluster II, tRNA Ser was found to form two groups. In cluster III, tRNA Leu and tRNA Val were found to form two groups, whereas tRNA Ile , tRNA Gly , and tRNA Thr formed three groups. Some tRNAs in cluster III were found to form one group, where these tRNAs containing the anticodons C-G-A in tRNA Ser , U-U-C in tRNA Glu , U-U-U in tRNA Lys , G-C-A and A-C-A in tRNA Cys , U-U-G in tRNA Gln , G-U-G in tRNA His , U-C-U in tRNA Arg , G-A-A in tRNA Phe , U-G-C in tRNA Ala , and C-A-U in tRNA Met all grouped separately (Fig. <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>). The multiple groupings of different tRNAs suggest that they evolved from multiple common ancestors. Furthermore, the tRNAs presented in cluster III, i.e., tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Thr (GGU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU), tended to be the most basic tRNAs and they had undergone gene duplication and diversification to generate other tRNA molecules.</ns0:p></ns0:div> <ns0:div><ns0:head>C-A-U anticodon in tRNA Ile</ns0:head><ns0:p>Our detailed genomic study showed that tRNA Ile also encodes a C-A-U anticodon in addition to the presence of this typical anticodon in tRNA Met . In general, the C-A-U anticodon is recognized as a typical characteristic of tRNA Met and there is only one isoacceptor. In particular, we found that the tRNA Ile in T. mairei encodes two C-A-U anticodons, and C. debaoensis, S. verticillate, D. spinulosum, C. lanceolate, G. biloba, C. deodara, W. mirabilis, G. gnemon, R. piresii, E. equisetina, and W. nobilis also encode a C-A-U anticodon (Table <ns0:ref type='table'>3</ns0:ref>, Data S1).</ns0:p><ns0:p>Transition/transversion of tRNAs A previous study <ns0:ref type='bibr' target='#b47'>(Mohanta et al., 2019)</ns0:ref> showed that the evolutionary rates are almost equal for tRNAs with respect to transition and transversion despite the low probability of transition or transversion events in tRNAs. In this study, we we identified several interesting substitutions of gymnosperm chloroplast tRNAs. Overall, our analysis of the substitution rates detected using the whole set of chloroplast tRNAs showed that average transition rate (15.38) was significantly larger than the average transversion rate (4.81) with a ratio of 3:1 (Table <ns0:ref type='table' target='#tab_8'>5</ns0:ref>). The same transition:transversion ratio bias was found in all the set of tRNAs for tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , and tRNA Leu . The ratio was over 6:1 for tRNA Cys and tRNA Arg .The transition rates for tRNA Trp , tRNA Val , and tRNA Gly were about 10 times higher than their transversion rates. These findings suggest that tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , tRNA Leu , tRNA Cys , tRNA Arg , tRNA Trp , tRNA Val and tRNA Gly underwent transition substitutions more readily than transversion substitutions during their evolution in gymnosperm chloroplast genomes. In addition,the transition rates in tRNA Lys and tRNA Pro were about 15 times higher than their transversion rates. The transition rates in tRNA Asn , tRNA Phe , and tRNA His were about 20 times higher than their transversion rates. These results indicate that tRNAs are much more likely to have undergone transition events rather than transversion events. The highest transition rate of 25.00 was found in tRNA Asp and the highest transversion rate of 0.00 was found in tRNA Ala and the lowest transversion rate of 0.00 in tRNA Asp (Table <ns0:ref type='table' target='#tab_8'>5</ns0:ref>). Similarly, tRNA Ala was found to have a transversion high rate but no significant transitions (Table <ns0:ref type='table' target='#tab_8'>5</ns0:ref>).</ns0:p><ns0:p>tRNA duplication/loss events In addition to transition and transversion events, gene duplication and loss events have played improtant roles in gene evolution. Our analysis of duplication and loss events indicated that 153 duplication events (duplication and conditional duplication) have occurred in all of the gymnosperm chloroplast tRNA genes investigated in this study (Fig. <ns0:ref type='figure' target='#fig_2'>S2</ns0:ref>). In addition, 220 gymnosperm chloroplast tRNA gene loss events were detected (Table <ns0:ref type='table' target='#tab_3'>S2</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_2'>S2</ns0:ref>). Thus, the loss of genes was slightly more frequent than their duplication for gymnosperm chloroplast tRNA genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>tRNAs are major genetic components of semi-autonomous chloroplasts and our analysis of gymnosperm chloroplast genomes showed that they have several basic conserved genomic features. The gymnosperm chloroplast genomes investigated in the present study were found to encode 28 to 33 tRNA isotypes, thereby indicating that there is substantial variation in the quantity of tRNAs in gymnosperm chloroplast genomes. The lack of tRNA Ala in R. piresii and T. mairei, and the absence of tRNA Val in T. mairei were interesting. Thus, it is necessary to understand how the translation process is conducted in chloroplasts without these crucial tRNAs. According to previous studies <ns0:ref type='bibr' target='#b62'>(Treangen &amp; Rocha, 2011;</ns0:ref><ns0:ref type='bibr' target='#b47'>Mohanta et al., 2019)</ns0:ref>, it is likely that the deficiency of these tRNAs is compensated for by the tRNAs in other organelle genomes, which may have multiple functions during the translation process in chloroplasts. In addition to the absence of tRNA Ala and tRNA Val , all of the gymnosperm plants were shown to not encode selenocysteine tRNA and its suppressor tRNA in their chloroplast genomes (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Selenocysteine tRNA and its suppressor tRNA were also not detected in the chloroplast in Oryza sativa <ns0:ref type='bibr' target='#b46'>(Mohanta &amp; Bae, 2017)</ns0:ref>.</ns0:p><ns0:p>In addition to the presence of C-A-U anticodon in tRNA Met , we found that tRNA-CAU is present in tRNA Ile (Table <ns0:ref type='table'>3</ns0:ref>). Similarly, the C-A-U anticodon was detected in tRNA Ile in Bacillus subtilis (Ehrenberg) Cohn and spinach <ns0:ref type='bibr' target='#b26'>(Kashdan &amp; Dudock, 1982;</ns0:ref><ns0:ref type='bibr' target='#b32'>K&#246;hrer et al., 2014)</ns0:ref>. The possible mechanism that governs the specificity of this amino acid may involve modification of the wobble position in the anticodon by a tRNA-modifying enzyme. Chloroplasts originate from bacteria so the tRNA modifications found in bacteria may also occur in chloroplast tRNAs. In bacteria, the tRNA-modifying enzyme TilS can convert the 5&#61602;-C residue in the CAU anticodon of specific tRNA Ile molecules into lysidine to decode 5&#61602;-AUA (Ile) codons instead of 5&#61602;-AUG (Met) codons <ns0:ref type='bibr' target='#b56'>(Soma et al., 2003)</ns0:ref>. The absence of tRNA Ile -lysidine synthetase leads to a failure to modify C34 to lysidine in tRNA Ile (LAU) (i.e., the synthesis of CAU-tRNA Ile ) and this inactivates the translation of AUA codons <ns0:ref type='bibr' target='#b32'>(K&#246;hrer et al., 2014)</ns0:ref>. Moreover, the conformational preferences of modified nucleosides may also affect the identification of methionine and isoleucine. The tautomer form of lysidine may provide compatible hydrogen bond donoracceptor sites to allow base pairing with 'A' and this may lead to the recognition of the codon AUA instead of AUG <ns0:ref type='bibr' target='#b57'>(Sonawane &amp; Tewari, 2008;</ns0:ref><ns0:ref type='bibr' target='#b54'>Sambhare et al., 2014)</ns0:ref>.</ns0:p><ns0:p>During protein coding, a certain species or gene tends to use one or more specific synonym codons, which is referred to as codon usage bias <ns0:ref type='bibr'>(Comeron &amp; Aguad&#233;, 1998;</ns0:ref><ns0:ref type='bibr' target='#b53'>Rota-Stabelli et al., 2012)</ns0:ref>. In the present study, tRNA Arg -CCG was found to be present in the genomes of nine species but absent from C. lanceolate, T. mairei, and E. equisetina. Similarly, tRNA Gly -UCC was shown to be absent from the chloroplast genomes of C. debaoensis, S. verticillate, D. spinulosum, C. lanceolate, T. mairei, and E. equisetina (Table <ns0:ref type='table'>3</ns0:ref>). These results suggest that gymnosperm chloroplast tRNA genes are characterized by codon usage bias <ns0:ref type='bibr' target='#b65'>(Wei &amp; Jin, 2017;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In general, the secondary structure of tRNAs is characterized as clover leaf-like, except for a few tRNAs with unusual secondary structures <ns0:ref type='bibr' target='#b24'>(J&#252;hling et al., 2018)</ns0:ref>. In our study, we identified clover leaf-like tRNAs with loop structures in their variable region (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_2'>2</ns0:ref>). In particular, a novel tRNA structure lacking the D arm was found in tRNA Gly in W. nobilis (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>). Numerous tRNA Leu , tRNA Ser , and tRNA Tyr were identified with a loop configuration in their variable region, thereby suggesting the existence of significant variation in chloroplast tRNAs. These variant and novel tRNAs require further investigation to determine whether they have biological importance. Most tRNAs have a clover-like structure formed by complementary base pairing between small segments <ns0:ref type='bibr' target='#b23'>(Hubert et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b12'>Florentz, 2002)</ns0:ref>. Previous studies have showed that the acceptor arm of tRNAs in chloroplasts contain 7 nt to 9 nt, the D-arm contains 3nt to 4 nt, the D-loop has 4 nt to 12 nt, the anticodon arm has 5 nt, the anticodon loopcontains 7 nt, the variable region comprises 4 nt to 23 nt, and &#936;-arm contains 5 nt, and the &#936;-loop has 7 nt <ns0:ref type='bibr' target='#b66'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b48'>Mohanta &amp; Hanhong, 2017;</ns0:ref><ns0:ref type='bibr' target='#b47'>Mohanta et al., 2019)</ns0:ref>. In the present study, we found that the acceptor arm of chloroplast tRNAs contains 2 nt to 7 nt in 373 tRNAs, where the D-arm has 3 nt or 4 nt and the D-loop usually contains 7 nt to 11 nt. The anticodon loop of gymnosperm chloroplast tRNAs generally contains 7 nt, and thus the sequence of the anticodon loop is typically highly conserved (Table <ns0:ref type='table' target='#tab_6'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The variable loop of different tRNAs contain 4 nt to 23 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The &#936;-arm of gymnosperm chloroplast tRNAs generally contains 5 nt and the &#936;-loop has 7 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). Our results are consistent with previous findings <ns0:ref type='bibr' target='#b66'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b48'>Mohanta &amp; Hanhong, 2017)</ns0:ref> and they suggest that chloroplast RNAs are significantly conserved. In addition, small conserved consensus sequences were found in the &#936; region with a composition of 'U-U-C-X-A-X2' (Table <ns0:ref type='table' target='#tab_6'>4</ns0:ref>). A previous study also reported the existence of a similar sequence in the &#936;-loop of tRNAs <ns0:ref type='bibr' target='#b46'>(Mohanta &amp; Bae, 2017)</ns0:ref>.</ns0:p><ns0:p>Our phylogenetic analysis detected three clear clusters and many tRNA groups. Some tRNAs (tRNA Ser , tRNA His , and tRNA Leu ) in cluster I and cluster II were also in cluster III, thereby indicating that these tRNAs evolved from multiple lineages by gene duplication and gene divergence. Moreover, anticodon types comprising CGA, UUC, UUU, GCA, ACA, UUG, GUG, UCU, GAA, UGC, and CAU appeared severally at phylogenetic tree, and thus the corresponding tRNAs evolved from multiple common ancestors. The overlapping of tRNAs groups demonstrates that these tRNAs might have diverse common ancestors in the evolutionary process <ns0:ref type='bibr' target='#b46'>(Mohanta &amp; Bae, 2017)</ns0:ref>. Phylogenetic analysis also showed that tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Thr (GGU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU) in cluster III tended to be the most basic tRNAs, whereas tRNA Met tended to be the most original tRNA. Overall, the results clearly indicate that the tRNAs encoded in gymnosperm chloroplast genomes have multiple common evolutionary ancestors.</ns0:p><ns0:p>Our results also provided insights into the gene substitution rates in gymnosperm chloroplast tRNAs. Overall, the average transition rate for tRNAs was greater than the transversion rate, where the relationship was about 3:1 (Table <ns0:ref type='table' target='#tab_8'>5</ns0:ref>). In all of the chloroplast tRNAs, the average transition rate was slightly higher than the average transversion rate, thereby indicating that chloroplast tRNAs have unequal substitution rates.</ns0:p><ns0:p>In addition to the transition and transversion events in tRNAs, loss and duplication events have played significant roles in the evolution of tRNAs in gymnosperm chloroplast genomes <ns0:ref type='bibr' target='#b18'>(He &amp; Zhang, 2006;</ns0:ref><ns0:ref type='bibr'>Magadum et al., 2013)</ns0:ref>. In general, the gene loss events tended to occur after whole genome duplication events. We found 153 duplication events and 220 loss events in gymnosperm chloroplast tRNAs, and thus loss events have occurred slightly more frequently than duplication events (Table <ns0:ref type='table' target='#tab_3'>S2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our basic structure analysis showed that gymnosperm chloroplast tRNAs encode 25 to 31 anticodons. The acceptor arm of chloroplast tRNA contains 2 nt to 7 nt, the D-arm has 3 nt or 4 nt, the D-loop contains 7 nt to 11 nt, and the anticodon loop usually contains 7 nt. In different tRNAs, the variable loop contains 4 nt to 23 nt. The &#936;-arm contains a conserved sequence comprising U-U-C-X-A-X2. In addition, tRNA Ala was found to be absent from R. piresii and T. mairei, and tRNA Val was lacking in T. mairei. Gymnosperm chloroplasts do not encode selenocysteine tRNA and its suppressor tRNA in their genomes. tRNA-CAU is also present in tRNA Ile as well as tRNA Met . A novel tRNA structure lacking the D arm was identified in tRNA Gly in W. nobilis. Numerous tRNA Leu , tRNA Ser , and tRNA Tyr types were found to have a loop configuration similar to the anticodon loop in the variable region of these tRNAs. Phylogenetic analysis showed that tRNAs such as tRNA Ser , tRNA His , and tRNA Leu in cluster I and cluster II were also present in cluster III, thereby indicating that these tRNAs have multiple evolutionary origins. The overlapping of tRNA groups demonstrates that tRNAs might have multiple common ancestors in the evolutionary process. tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Ile (CAU) and others tended to be the most basic tRNAs, whereas tRNA Met tended to be the most original tRNA. Transition and transversion analyses of tRNAs indicated that tRNAs are iso-acceptor specific and the transition rate was generally higher than the transversion rate. Furthermore, gene loss events (220) have occurred slightly more frequently than gene duplication events (153) in gymnosperm chloroplast tRNAs. Our results provide new insights into the evolution of gymnosperm chloroplast tRNAs and their diverse roles.</ns0:p></ns0:div> <ns0:div><ns0:head>Table 1(on next page)</ns0:head><ns0:p>Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> A view of the selected gymnosperms in analysis.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:11:43190:1:1:NEW 25 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div> <ns0:div><ns0:head>Met/fMet</ns0:head><ns0:p>Tyr Cys Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>1 1 2 1 1 1 1 1 1 1 1 Trp 1 1 1 1 1 1 1 2 1 1 1 Selenocystei ne 0 0 0 0 0 0 0 0 0 0 0 Suppressor 0 0 0 0 0 0 0 0 0 0 0</ns0:formula></ns0:div> <ns0:div><ns0:head>Table 3(on next page)</ns0:head><ns0:p>Table <ns0:ref type='table'>3</ns0:ref> Distribution of anti-codons in the chloroplast genome of selected gymnosperms.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:11:43190:1:1:NEW 25 May 2020)</ns0:p><ns0:p>Manuscript to be reviewed 1 Table <ns0:ref type='table'>3</ns0:ref>. Distribution of anti-codons in the chloroplast genome of selected gymnosperms. Manuscript to be reviewed Manuscript to be reviewed The consensus sequence in chloroplast tRNAs of selected gymnosperms.</ns0:p><ns0:formula xml:id='formula_1'>U Threonine G-C-X 2 -G-X 2 X-C-U- C A-G-X-G-G-U-X 0-1 - A X-C-G- C-X X-U-X 3 -A-A X 2 -G-U-C X-U-X- G-G U-U-C-X-A- X 2 Tryptopha n G-C-G-C-U- C-U G-U- U-X A-G-X 3 -G-G-U-A X 2 -G-G- U C-U-C-C-A- A-A A-U-G-X-C G-U-A- G-G U-U-C-A-A- A-U Tyrosine X-G-G-X-C- X-A X-C-X 2 A-G-X 6 -A ***** C-U-G-U-A- A-A **** X 3 -G-G U-U-C-X-A- X 2 Valine A-G-G-G-X- U-A X-C-U- C A-G-X 4-6 -A U-C-X- C-X U-U-X-A-C- X 2 X 2-3 -G-U-C X 2 -C-X-G U-U-C-X-A- X 2</ns0:formula><ns0:p>Note, the consensus sequence was investigated from 5' pole of the tRNA. The asterisk mark (*) show the absence of conserved nucleotide consensus sequence in respective region of chloroplast tRNAs. 5' AC-arm, 5' Acceptor arm; ANC-arm, Anti-codon arm; ANC-loop, Anti-codon loop; &#936;-arm, Pseudouridine arm; &#936;-loop, Pseudouridine loop. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note></ns0:div><ns0:figure xml:id='fig_1'><ns0:head>Fig. 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig. 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Fig. 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Fig. 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Fig. 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig. 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Fig. 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Fig. 4</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>A view of the selected gymnosperms in analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:1:1:NEW 25 May 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Distribution of tRNA isotypes in chloroplast genome of selected gymnosperms.</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2019:11:43190:1:1:NEW 25 May 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Distribution of tRNA isotypes in chloroplast genome of selected gymnosperms.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>tRNA Isotypes</ns0:cell><ns0:cell>debaoensis C.</ns0:cell><ns0:cell>D.</ns0:cell><ns0:cell>Number of tRNAs</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Transition and transversion rate of chloroplast tRNA.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Alanine</ns0:cell></ns0:row></ns0:table><ns0:note>1 PeerJ reviewing PDF | (2019:11:43190:1:1:NEW 25 May 2020)</ns0:note></ns0:figure> <ns0:note place='foot' n='2'>PeerJ reviewing PDF | (2019:11:43190:1:1:NEW 25 May 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Dear Dr. Sun, Thank you very much for kindly returning our manuscript. We are very grateful to you for your time and patience, as well as your and reviewers’ constructive and thoughtful comments, which have been valuable and very helpful for revising and improving our MS. In the revised version, we have made thorough corrections, and responded point by point to the comments as itemized below (our responses are in bold characters). If you have any questions regarding the manuscript, please feel free to contact the corresponding authors: Zhong-Hu Li, Jian-Ni Liu E-mail: [email protected]; [email protected] We are looking forward to hearing from you at your earliest convenience. Yours sincerely, Drs. Zhonghu Li & Jian-Ni Liu Authors reply to editors’ comments: Reviewer 1 (Anonymous) The structure of the manuscript is well organized and easy to follow, but the English language is not of sufficient quality. Reply: Thank you very much for your kindly comment. In the revised manuscript, we have thoroughly revised the manuscript in the help of a native English speaker: Dr. Duncan Jackson (A senior English editor) and we hope this version could be more suitable. Experimental design The authors used the species tree to reconcile the tRNA phylogeny to identify gene loss and gain events. No figure legend was provided for the figure s1. What is the difference between the 'D' and 'cD' on the nodes of the phylogeny? Reply: Thank you very much for your helpful suggestion. We have added the figure legend of Figure S1 (reorded as Figure S2) in the revised manuscript, and the difference of “D” nodes and “cD” nodes of the phylogeny was: “D” means Duplication, and “cD” means Conditional Duplication. Both these two types of nodes represent duplication nodes, so in the study, we put these nodes as duplication in general terms. Validity of the findings The authors claimed the phylogenetic tree of tRNAs showed the presence of three major clusters covering 64 groups of all tRNAs. How the three major clusters are recognized? This conclusion could not be obtained basing on the current form of the figure 3. The authors should provide more details about getting this conclusion. Reply: Many thanks to you for your kindly advice. The three major clusters were recognized by the clustering of the phylogenetic tree, and we have improved the Fig. S1 to provide a clearer appearance of the three major clusters of the phylogenetic tree in the revised manuscript. Thank you again for your suggestion. The authors found that 'tRNAAla has been found to be absent in R. piresii and T. mairei while tRNAVal has been found with inexistence in T. mairei'. It would be interesting to investigate whether the absent of these tRNAs affected the codon usage of the protein coding genes in the chloroplast of these species. Reply: Thank you very much for your helpful suggestion. We have thought about the issue carefully and consulted relevant literatures. From literatures we understood that maybe genomic tRNA compensated for the absence of these tRNAs or perhaps other tRNAs from the organellar genome perform multiple functions to conduct related protein translation. We therefore considered that this compensation mechanism probably presented in the chloroplast and supplemented codon usage of the protein coding genes in the chloroplast. And in the Discussion, we have added these contents. Comments for the Author This study aimed: (1) to determine nucleotide diversification of secondary structure of gymnosperm tRNAs; (2) to identify the detailed genomic aspects of chloroplast tRNAs; (3) to assess the evolutionary relationship between different chloroplast tRNAs; (4) to evaluate the duplication or loss on all involved tRNAs. This appeared to be original. There are a few points that need to be further clarified as mentioned above. Reply: Thank you very much for your kindly advice. In the revised version, we have thoroughly revised these contents accordingly. Reviewer 2 (Anonymous) Basic reporting Basically, it is interesting to study the evolution of tRNAs encoded in the chloroplast genomes of gymnosperms, considering that they are maternally or paternally inherited, with evolution rates about five times slower than that of nuclear genomes. However, the manuscript is immature in its present form. Major concerns: a) The English of the manuscript needs to be improved substantially; the sense of many sentences is unclear, cryptic or misleading. Reply: Thank you very much for your helpful and constructive comment. In the revised manuscript, we have made every effort to improve the English to make expression clearer. b) The resolution/quality of Figure 1 is too low. I am also not convinced to classify the tRNAs shown in Fig. 1 as “non-cloverleaf-like structures”. They adopt cloverleaf structures, but additionally have more or less rigid variable arms. Reply: Thanks very much for your helpful comment. We highly agree with your proposal. We have improved the quality of Figure 1 and reordered Figures. We also have corrected the description of tRNAs shown in Fig. 1 in the revised manuscript. c) Table 4: it is not immediately clear which part of the tRNA these consensus sequences represent; for example, in the column “Ψ arm”, the reader is wondering if this is the sequence of the T stem 5’- or 3’-strand? Indicate this more clearly, also give sequence polarity. The authors should include a tRNA secondary structure where the mark the sequence elements specified in Table 4. Reply: Thanks a lot for your kindly suggestions. We have carefully revised these contents to make it clearer and indicated sequence polarity in Table 4. We also have supplemented a tRNA secondary structure (Fig. 4) where marks the sequence elements specified in Table 4, according to your advice. d) These endless listings of tRNAs (for example, in lines 265 to 313) makes the manuscript hard to digest. The authors should illustrate their findings in additional summary figures. The clustering of tRNAs is not comprehensible for the reader. The authors should exemplarily delineate the sequence criteria underlying the clustering of the respective tRNA; explain why some tRNAs appear in more than one group. The evolutionary clustering of gymnosperm chloroplast tRNAs, cumulating in the phylogenetic tree of Figure 3, is at the heart of this story and should thus be more intelligible to the reader. Reply: Thank you very much for your constructive comments. We have rewritten these contents in order to make it easy to be understand in the revised manuscript. The groups of tRNAs were identified by the clustering of the phylogenetic tree, and we have marked these groups in Fig. S1 (strings of different lengths) to assist explanation. Multiple grouping of different tRNAs suggested the evolution of chloroplast tRNAs were from multiple common ancestors. And we also discussed these findings in Discussion. Thank you again for your helpful advice. e) The CAU anticodon of tRNA-Ile: it is known from bacteria that the enzyme TilS converts the 5’-C residue in the CAU anticodon of specific tRNA-Ile molecules to lysidine (2-lysyl cytidine; abbreviated as L or k2C ) to decode 5’-AUA (Ile) codons instead of 5’-AUG (Met) codons (Reference: Soma A, Ikeuchi Y, Kanemasa S, Kobayashi K, Ogasawara N, Ote T, Kato J, Watanabe K, Sekine Y, Suzuki T: An RNA-modifying enzyme that governs both the codon and amino acid specificities of isoleucine tRNA. Mol Cell 2003, 12(3):689–698). This mechanism may also be operational in chloroplasts. The authors should discuss this issue and should check the literature if there is evidence for the same mechanism in chloroplasts. Reply: Many thanks to you for your kindly advice. Chloroplast is of bacterial origin. Therefore, the mechanism of tRNA-modification in bacteria might explain the CAU anti-codon of tRNAIle in chloroplast. We also have consulted relevant literatures and discussed this content in the revised manuscript according to your suggestion. f) Figure 2: describe this predicted W. nobilis RNA structure more cautiously as a tRNA-Gly-like structure, as it has a (most likely) non-functional acceptor stem that is not charged with glycine. Reply: Thank you so much for your helpful suggestion. We have carefully checked this part and identified the secondary structure again using tRNAScan-SE software. The perfect match for tRNAGly-UCC of W. nobilis is Oscillatoria nigroviridis_PCC_7112_Osci_nigr_PCC_7112_tRNA-Gly-TCC (please see below partial results). Thus, we considered the tRNA in W. nobilis was tRNAGly. And it lacks D-arm compared with the typical secondary structure of tRNA. In the revised manuscript, we have corrected the description of this tRNA structure. Reviewer 3 Comments for the author Here are the comments which can be useful but not limited in improving the structure and quality of the manuscript. 1. Authors did not talk about the presence of modified bases in tRNA? Reply: Thank you very much for your valuable advice. Modification is important for tRNAs in their structure, stability, correct folding and aminoacylation and so on. And in the introduction and discussion of revised manuscript, we have supplemented this content. 2. The English used in the manuscript is very poor and needs to be improved heavily so as to reach an average science audience. It was very hard to understand at multiple places that the reader loses connectivity with the theme of manuscript. Use of few words (e.g. besides) at unnecessary places needs to be avoided. Please avoid long and complex sentences such as line no. 76-80. Also, multiple sentences are repeating in every section. Because of poor English, at many places there arises a doubt on the experiments performed. So the whole manuscript has to be restructured. Reply: Many thanks to you for your kindly comments. We have carefully and thoroughly improved the English in the manuscript, avoided using few words at unnecessary places, and rewritten long and complex sentences in order to express clearly. And we also have modified the repeated sentences in each section according to your advice. 3. All the figures need to be submitted online with proper fig captions. Please rename fig 1b as 2 and so on if possible. Fig. 3 for the phylogenetic tree is not at all clear hence I could not validate the results based on it. The manuscript contains most of the data based on that figure. Hence, at this stage the results cannot be validated. Reply: Thank you very much for your helpful and constructive suggestions. We have supplemented fig captions for every Figure and reordered figures in the revised manuscript. We have marked tRNA groups in Fig. S1 (strings of different lengths) to offer a detailed explanation. Thank you again for your helpful advice. 4. It will be useful to include a line or two in the abstract or introduction stating the importance of this study or how these results are promising over the past studies. Reply: Thank you so much for your kindly advice. We have supplemented this content in the revised manuscript. 5. The experimental protocol needs to be explored in more precise stepwise manner (probably English improvisation could help). Reply: Thanks a lot for your helpful suggestion. In the revised manuscript, we have carefully and made every effort to improve the description of experimental methods. 6. Line no. 138 says bacterial parameter in tRNAScan-SE. Why this was used? Reply: Thank you for your valuable comment. In the study, tRNAScan-SE was used to investigate the secondary structure of chloroplast tRNAs. The reason we used bacterial parameter is that chloroplast was generally convinced to be the origin of cyanobacteria. And based on referencing to other relevant literatures, we decided to use this parameter. 7. Protocol described for phylogenetic tree construction from line no. 144 is unclear and its present written form will be difficult to reproduce the results. Also from line no. 161 various methods were used to filter the data. What is the rationale behind such multiple filtering? How each filter can be correlated in case of this gymnosperm tRNA analysis? Reply: Thank you so much for your helpful suggestion. Sequences need to be aligned before constructing phylogenetic tree. We aligned tRNA sequences and turned the sequences’ format to MEGA format to construct the phylogenetic tree of tRNAs. And we used MEGA to calculate the Transition/transversion rate of each kind of tRNA. Each kind of tRNA sequences should be aligned before. In Methods, we have modified these contents in order to express more clearly. 8. Please include the results of gene manipulation (line no. 169) in the main document as they are much essential to interpret the results stated in the manuscript. Reply: Thank you very much for your constructive advice. We highly agree with your suggestion, but the results of gene manipulation (Fig. S1) is too large to place in the main manuscript, so we have improved it and reordered it as Fig. S2. 9. Line no. 265-313 & 424-438 is totally un-interpretable because of poor resolution of fig 3. Also, line no. 349-355 correspond to fig S1 and table S2. Please make them available in the main text. Reply: Many thanks for your helpful comment. We have modified these contents in order to make it easy to be understand in the revised manuscript. We also optimized Fig. S1 (the reordered figure) to show phylogenetic tree detailedly and to offer a clearer appearance. And Fig. S1 (Fig. S2 in the revised manuscript) and table S2 are too long, so we put it in the Supplementary materials. 10. For this line, “Notwithstanding, it is still unclear that how the tRNA captured with CAU anti-codon to distinguish Methionine and Isoleucine and carry them respectively” authors may look into these references; Nucleosides, Nucleotides and Nucleic Acids. 27, 1158-1174, 2008; RSC Advances, 4, 14176- 14188, 2014; Reply: Thank you very much for your valuable advice. We have looked into these references carefully and supplemented possible reasons of distinguishing Methionine and Isoleucine (the modification of tRNA-modifying enzyme and conformational preferences of modified nucleosides) in the revised manuscript according to these literatures and other related literatures. Thank you again for your suggestion. 11. Please provide a reference to the statement made in line no. 377. Reply: Thank you so much for your helpful advice. We have corrected this content and supplemented relevant literatures. 12. Line no. 397 mentions a novel tRNA structure lacking the acceptor arm. I would like to know how the authors have re-validated its occurrence. Please explain. Reply: Thank you very much for your kindly suggestion. We identified the secondary structure using tRNAScan-SE software. And it lacks D-arm compared with the traditional secondary structure of tRNA. In the revised manuscript, we have corrected the description of this novel tRNA. 13. The conclusions do not completely reflect the findings. Please fortify this section by correlating the key findings from results sections. Reply: Thanks a lot to you for your valuable comment. We have modified this part in the revised manuscript according to your suggestion. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Gymnosperms such as ginkgo, conifers, cycads, and gnetophytes are vital components of land ecosystems, and they have significant economic and ecologic value, as well as important roles as forest vegetation. In this study, we investigated the structural variation and evolution of chloroplast transfer RNAs (tRNAs) in gymnosperms. Chloroplasts are important organelles in photosynthetic plants. tRNAs are key participants in translation where they act as adapter molecules between the information level of nucleic acids and functional level of proteins. The basic structures of gymnosperm chloroplast tRNAs were found to have family-specific conserved sequences. The tRNA &#936;-loop was observed to contain a conforming sequence, i.e., U-U-C-N-A-N 2 . In gymnosperms, tRNA Ile was found to encode a 'CAU' anticodon, which is usually encoded by tRNA Met . Phylogenetic analysis suggested that plastid tRNAs have a common polyphyletic evolutionary pattern, i.e., rooted in abundant common ancestors. Analyses of duplication and loss events in chloroplast tRNAs showed that gymnosperm tRNAs have experienced little more gene loss than gene duplication. Transition and transversion analysis showed that the tRNAs are isoacceptor specific and they have experienced unequal evolutionary rates. These results provide new insights into the structural variation and evolution of gymnosperm chloroplast tRNAs, which may improve our comprehensive understanding of the biological characteristics of the tRNA family.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Gymnosperms originated in the Paleozoic Devonian Period (about 385 million years ago), and they are key groups in terms of the transformation from spore reproduction to seed reproduction in higher plants <ns0:ref type='bibr' target='#b14'>(Gerrienne et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Crisp &amp; Cook, 2011)</ns0:ref>. According to the latest phylogenetic classification, gymnosperm species are divided into eight orders, 12 families, 84 genera, and more than 1,000 species <ns0:ref type='bibr'>(Wang &amp; Ran, 2014)</ns0:ref>. Gymnosperms include ginkgo, cycads, conifers, and gnetophytes, which are grown in forests as important timber species and they provide raw materials for human usage, such as fiber, resin, and tannin <ns0:ref type='bibr' target='#b3'>(Christenhusz et al., 2010)</ns0:ref>. In addition, gymnosperms include some important threatened plants, where 40% are at high risk of extinction <ns0:ref type='bibr' target='#b13'>(Forest et al., 2018)</ns0:ref>. Recent phylogenetic and evolutionary studies of gymnosperms have demonstrated the rapid evolution of mitochondrial (mt) genes and provided further evidence of sister relationship between conifers and Gnetales <ns0:ref type='bibr' target='#b51'>(Ran, Gao &amp; Wang, 2010)</ns0:ref>. The high levels of genetic diversity and population differentiation among the Pinus species in gymnosperms have been studied based on plastid DNA markers <ns0:ref type='bibr' target='#b40'>(Liu et al., 2014)</ns0:ref>. Other studies have indicated patterns related to the physiological ecology, phylogenetic relationships, and population genetic structure of gymnosperm species <ns0:ref type='bibr'>(Yu et al., 2014;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b8'>Dong et al., 2016)</ns0:ref>. However, these studies mainly considered the phylogeny and evolution at the whole populations level. Thus, the detailed evolutionary characteristics of gymnosperms still need to be elucidated.</ns0:p><ns0:p>Chloroplasts are the site of photosynthesis and of various essential metabolic pathways, e.g., fatty acid and amino acid biosynthesis and the assimilation of nitrogen, sulfur, and selenium <ns0:ref type='bibr'>(Wise &amp; Hoober, 2006;</ns0:ref><ns0:ref type='bibr'>Des Marais,2000;</ns0:ref><ns0:ref type='bibr' target='#b30'>Knorr &amp; Heimann, 2001;</ns0:ref><ns0:ref type='bibr' target='#b49'>Pilon-Smits et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b16'>Guo et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b33'>Kretschmer, Croll &amp; Kronstad, 2017)</ns0:ref>. It is generally recognized that chloroplasts are derived from proto-eukaryotic symbiotic cyanobacteria that internalized in eukaryotic cells <ns0:ref type='bibr' target='#b21'>(Hiroki &amp; Daisuke, 2018</ns0:ref>) and evolved into central organelles. Chloroplasts have their own genome encoding about 100 proteins and they are maternally inherited organelles in most angiosperm plants <ns0:ref type='bibr' target='#b0'>(Abdallah, Salamini &amp; Leister, 2000;</ns0:ref><ns0:ref type='bibr' target='#b20'>Heuertz et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b4'>Civan et al., 2014)</ns0:ref>. Among gymnosperms, paternal plastid inheritance is the typical characteristic of conifers <ns0:ref type='bibr'>(Faur&#233; et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kaundun &amp; Matsumoto, 2011)</ns0:ref>. Studies have shown that the chloroplast genome is quite conserved with an average evolutionary rate of 0.2-1.0&#61620;10 -9 per site per year, which is only one-fifth of that for the nuclear genome <ns0:ref type='bibr' target='#b9'>(Drouin, Daoud &amp; Xia, 2008;</ns0:ref><ns0:ref type='bibr' target='#b10'>Duchene &amp; Bromham, 2013)</ns0:ref>. The chloroplast genome is a covalently closed circular structure with four parts comprising the large single copy region (LSC), small single copy region (SSC), inverted repeat region A (IRa), and inverted repeat region B (IRb). The two IRs have the same sequences but in the opposite direction <ns0:ref type='bibr'>(Wang et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b41'>Logacheva et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b19'>Hereward et al., 2018)</ns0:ref>. Due to the independent evolution of the chloroplast genome, it is possible to construct a molecular phylogenetic tree using the chloroplast genome and without requiring any other data. Data analysis based on the conserved evolution of plastids is highly valuable for phylogenetic studies <ns0:ref type='bibr' target='#b28'>(Kim &amp; Suh, 2013)</ns0:ref> because it can provide reliable and useful phylogenetic information. The Manuscript to be reviewed relative completeness and independence of the chloroplast genome means that it can provide valuable material for research purposes.</ns0:p><ns0:p>Transfer RNAs (tRNAs) undergo numerous post-transcriptional nucleotide modifications and they exhibit abundant chemical diversity where the bases experience methylation, formylation, and other modifications <ns0:ref type='bibr' target='#b58'>(Suzuki &amp; Suzuki, 2014)</ns0:ref>. Chemical nucleotide modifications are frequent in tRNAs and they are important for the structure, stability, correct folding, aminoacylation, and decoding. For example, a previous analysis of the chemically synthesized f 5 C34-modified anticodon loop of human mt-tRNA Met showed that f 5 C34 contributes to the anticodon domain structure of the mt-tRNA <ns0:ref type='bibr' target='#b43'>(Lusic et al., 2008)</ns0:ref>. tRNAs comprise sequences of less than 100 polynucleotides that fold into a clover-type secondary structure and then into an L-shaped tertiary structure <ns0:ref type='bibr'>(Wilusz, 2015)</ns0:ref>. The secondary structure of tRNAs comprises different arms as well as loops, i.e., the D-arm, acceptor arm, anticodon arm, pseudouridine arm (&#936;-arm), D-loop, variable arm, anticodon loop, and pseudouridine loop (&#936;loop) <ns0:ref type='bibr' target='#b15'>(Gieg&#233;, Puglisi &amp; Florentz, 1993;</ns0:ref><ns0:ref type='bibr' target='#b44'>Mizutani &amp; Goto, 2000)</ns0:ref>. This unique structure allows tRNA to act as important bridges between the information level of nucleic acids and functional level of proteins. The vital components of tRNAs comprise an anti-codon region that discerns the messenger RNA carried by the specific codons, a 3&#61602;-CCA tail for attaching to the cognate amino acid, the &#936;-arm, and a &#936;-loop that has a relationship with the ribosome machinery <ns0:ref type='bibr' target='#b29'>(Kirchner &amp; Ignatova, 2014)</ns0:ref>. Asymmetric combinations and the divided segments in tRNA genes allow us to understand the diversity of tRNA molecules. tRNA species fulfill various functions in cellular homeostasis, regulation of gene expression and epigenetics, biogenesis, and even biological disease <ns0:ref type='bibr' target='#b52'>(Ribasd &amp; Dedon, 2014;</ns0:ref><ns0:ref type='bibr' target='#b25'>Kanai, 2015;</ns0:ref><ns0:ref type='bibr' target='#b55'>Schimmel, 2017)</ns0:ref>. The evolutionary relationships determined between cyanobacteria and monocots show that tRNAs evolved polyphyletically and they originated from multiple common ancestors with a high rate of gene loss <ns0:ref type='bibr'>(Mohanta et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b47'>Mohanta et al., 2019)</ns0:ref>. Nevertheless, the basic details of the tRNAs in plant chloroplasts still need to be elucidated and on the diverse evolutionary features of gymnosperm tRNAs are still unclear.</ns0:p><ns0:p>In this study, we assessed all of the chloroplast genomes in 12 families of gymnosperms from eight orders. The main aims of this study were as follows: (1) to determine the diversification of nucleotides in the secondary structure of gymnosperm tRNAs; (2) to identify the detailed genomic features of chloroplast tRNAs; (3) to assess the evolutionary relationships among different chloroplast tRNAs; and (4) to evaluate the duplication or loss events that occurred in all of the tRNAs considered. Our findings provide important insights into the biological characteristics and evolutionary variation of the tRNA family.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>Annotation and identification of chloroplast tRNA sequences in gymnosperms We downloaded complete chloroplast genomes for 12 representative gymnosperms in eight orders from the National Center for Biotechnology Information database (NCBI, https://www.ncbi.nlm.nih.gov/). The gymnosperm species investigated were: Cycas debaoensis <ns0:ref type='formula'>NC_011954</ns0:ref>).The gymnosperm tRNA genomes were annotated using GeSeq-Annotation of Organellar Genomes tool <ns0:ref type='bibr' target='#b59'>(Tillich et al., 2017)</ns0:ref> where the parameters were set as: circular sequence(s), chloroplast of sequence source, generate multi FASTA; BLAT protein search identity 25% for annotating plastid IR, 85% identity for BLAT rRNA, tRNA and DNA search, Embryophyta chloroplast (CDS+rRNA), third party tRNA annotator ARAGORN v1.2.38, ARWEN v1.2.3, tRNAScan-SE v2.0, and without Refseq choice.</ns0:p><ns0:p>Structural analysis of chloroplast tRNAs ARAGORN <ns0:ref type='bibr' target='#b37'>(Laslett &amp; Canback, 2004)</ns0:ref> and tRNAScan-SE software <ns0:ref type='bibr' target='#b42'>(Lowe &amp; Eddy, 1997)</ns0:ref> were employed to analyze the sequences and the secondary structure of tRNAs in the chloroplast genomes of the involved gymnosperm plants. The default parameters were set in ARAGORN software. The parameters for tRNAScan-SE were set as: sequence source, bacterial; search mode, default; query sequences, formatted (FASTA); and genetic code for tRNA isotype prediction, universal. Phylogenetic tree construction A phylogenetic tree was constructed for all of the tRNAs using MEGA7.0 software <ns0:ref type='bibr' target='#b34'>(Kumar et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b35'>Kumar, Stecher &amp; Tamura, 2016)</ns0:ref>. To study the evolutionary details of chloroplast tRNAs in gymnosperm species, an alignment file for tRNAs was achieved by CLUSTAL Omega software before the phylogenetic tree was constructed. MEGA7 software was used to transform the alignment file into MEGA format. The phylogenetic tree was constructed with the following parameters: phylogeny reconstruction of analysis, maximum likelihood model, bootstrap method in phylogeny test, 1000 bootstrap replicates, nucleotides type, gamma distributed with invariant sites (G+I) model, five discrete gamma categories, partial deletion for gaps/missing data treatment, 95 % site coverage cut-off, and very strong for branch swap filter.</ns0:p><ns0:p>Transition/transversion analysis The sequences of the tRNA isotypes were aligned to determine the transition and transversion rates for chloroplast tRNAs in gymnosperm plants. The files covering all 20 types of tRNAs were transformed into the MEGA file format and analyzed separately using MEGA7.0 software <ns0:ref type='bibr' target='#b36'>(Kumar, Tamura &amp; Nei, 1994)</ns0:ref>. The transition and transversion rates were analyzed for tRNAs with the following parameters: substitution pattern estimation (ML) analysis, automatic (neighbor-joining tree), maximum likelihood statistical method, nucleotide substitution type, Kimura two-parameter model, gamma distributed (G) site rates, five discrete gamma categories, partial deletion of gaps/missing data treatment, 95% of site coverage cut-off, and very strong branch swap filter.</ns0:p><ns0:p>Loss and duplication events analysis for tRNA genes In order to investigate the duplication or loss events in tRNA genes, the NCBI taxonomy browser was utilized to construct the whole species tree for the 12 gymnosperm species considered. The phylogenetic tree conducted in the evolutionary study was employed as gene tree. The gene tree for the tRNAs and species tree for the gymnosperm species were submitted to Notung 2.9 software <ns0:ref type='bibr' target='#b1'>(Chen, Durand &amp; Farach-Colton, 2000)</ns0:ref>, and then reconciled to discover duplicated and lost tRNA genes in the chloroplast genomes of gymnosperms.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Genomic features of gymnosperm chloroplast tRNAs Sequences were analyzed to identify the genomic tRNAs in the chloroplast genomes of 12 gymnosperm species comprising C. <ns0:ref type='bibr'>debaoensis, D. spinulosum, G. biloba, C. deodara, W. nobilis, R. piresii, S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon, and E. equisetina</ns0:ref>, which were obtained from the NCBI database (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The results showed that the length of the chloroplast tRNAs vary from the smallest with 64 nucleotides (nt) (tRNA Met -CAU in T. mairei) to the largest with 96 nt (tRNA Tyr -AUA in W. nobilis, C. deodara, and G. biloba) (Data S1). We found that the chloroplast genomes of gymnosperm plants encode 28 to 33 tRNAs (Table <ns0:ref type='table'>2</ns0:ref>), where D. spinulosum, C. deodara, and S. verticillata encode 31 anticodons, W. nobilis, R. piresii, C. lanceolata, and G. gnemon encode 32 tRNA isotypes, G. biloba, and W. mirabilis encode 33 tRNAs. Other species comprising T. mairei, E. equisetina and C. debaoensis encode 28, 28, 30 tRNA isotypes, respectively (Table <ns0:ref type='table'>2</ns0:ref>). tRNA Ala was not found in R. piresii and T. mairei, and tRNA Val was not detected in T. mairei (Fig. <ns0:ref type='figure'>S3</ns0:ref>). We also observed that all of the species do not encode selenocysteine and its suppressor tRNA (Table <ns0:ref type='table'>2</ns0:ref>). Overall, tRNA Ser (in W. nobilis) and tRNA Arg (in W. mirabilis) are the most abundant (four types) followed by tRNA Leu (three types) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>Variations in structures of chloroplast tRNAs Some tRNAs with a loop structure in the variable region were found to be encoded in the gymnosperm chloroplast genomes (Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>). A novel tRNA lacking the D-arm was found in tRNA Gly in W. nobilis (Fig. <ns0:ref type='figure'>3</ns0:ref>). As shown in Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>, tRNA Leu , tRNA Ser , and tRNA Tyr contain expanded variable stem/loops. In these tRNAs (except for tRNA Ser -GCU of D. spinulosum), the anticodon loop of tRNA Ser contains the conserved consensus sequence N-U-N-G-A-A-N, and tRNAs Leu have the consensus sequence C-U-N-A-N 2 -A. The variable loop region is predicted to fold into stem-loop structures with apical loops of 3 to 7 nt in tRNA Ser and several tRNA Leu variants. The stems contain up to 7 bp (Fig. <ns0:ref type='figure' target='#fig_7'>1 and 2</ns0:ref>). The expanded variable loop structures may play important functions during the protein translation process in chloroplasts.</ns0:p></ns0:div> <ns0:div><ns0:head>Chloroplast genomes contain 25 to 30 anticodon-specific tRNAs</ns0:head><ns0:p>The genomes of the species analyzed were found to code for at least two copies of tRNA Met -CAU/tRNA fMet -CAU. Each of the gymnosperm chloroplast genomes encodes 25 to 30 anticodon-specific tRNAs (Table <ns0:ref type='table' target='#tab_2'>2 and Table 3)</ns0:ref>, where E. equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S. verticillata, and C. lanceolata encode 28 anticodons, and D. spinulosum, C. deodara, W. mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis, G. gnemon and G. biloba encodes 30 anticodons (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). tRNA Arg -CCG was present in the genomes of nine gymnosperm species but absent from C. lanceolata, T. mairei, and E. equisetina, while tRNA Gly -UCC was lacking from C. debaoensis, S. verticillata, D. spinulosum, C. lanceolata, T. mairei, and E. equisetina (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). The most abundant anticodons found in the chloroplast genomes were tRNA Gly -GCC, tRNA Pro -UGG, tRNA Ser -UGA, tRNA Ser -GCU, tRNA Arg -ACG, tRNA Arg -UCU, tRNA Leu -UAG, tRNA Leu -CAA, tRNA Phe -GAA, tRNA Asn -GUU, tRNA Lys -UUU, tRNA Asp -GUC, tRNA Glu -UUC, tRNA His -GUG, tRNA Gln -UUG, tRNA Ile -CAU, tRNA Met -CAU, tRNA Tyr -GUA, tRNA Cys -GCA, and tRNA Trp -CCA (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). Two tRNA Trp iso-acceptors are present in E. equisetina chloroplasts, compared with a single one in the other gymnosperm species analyzed in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved gymnosperm chloroplast tRNAs</ns0:head><ns0:p>The clover leaf-like secondary structure of a tRNA is shown in Fig. <ns0:ref type='figure'>4</ns0:ref>. In the study, we found that most tRNAs contain a 'G' as the first nucleotide in the D-arm, except for tRNA Lys , tRNA Met , tRNA Pro , tRNA Thr , tRNA Tyr , and tRNA Val . 'A' is present in the first and the last position of the D-loop apart from tRNA Gly , tRNA Ile , tRNA Leu , tRNA Met , and tRNA Ser . In addition, in the final two positions of the &#936;-arm, all of the tRNAs were found to have conserved 'G-G' nucleotides, except for tRNA Arg , tRNA Cys , tRNA Phe , and tRNA Val (Table <ns0:ref type='table'>4</ns0:ref>). Small conserved consensus sequences were found in the &#936; region. To be specific, except for tRNA Ser , the &#936;-loop in tRNAs was found to contain a conserved sequence comprising U-U-C-N-A-N 2 according to a multiple sequence alignment of 20 members of the tRNA gene family (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Diversification of tRNAs structures</ns0:head><ns0:p>The diverse arms and loops of tRNAs allow the regulation and control of protein translation. Each arm and loop has a specific nucleotide composition. Our analysis based on 373 tRNAs showed that the acceptor arm of chloroplast tRNAs contains 3 bp to 7 bp, where 357 have 7 bp, 13 have 6 bp, and the remaining tRNAs contain no more than 5 bp (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The D-arms were found to contain 3 or 4 bp generally, with a stable 'G' in the initial position and 'C' in the last position of most tRNAs (such as tRNA Ala , tRNA Asn , tRNA Asp , tRNA Cys , tRNA Glu , tRNA His , tRNA Ile , and tRNA Phe ). The most D-loops usually contain 7 bp to 11 bp with conserved 'A' nucleotides at the two end locations. The anticodon arms of chloroplast tRNAs mainly contain 5 bp (90.35%). We found that 367 (about 99.01%) tRNAs contain 7 nt in their anticodon loop, thereby indicating that the sequence of the anticodon loop is highly conserved (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The variable loops of different tRNAs contain 3 bp to 23 bp, where those in tRNA Ala , tRNA Asp , tRNA His , tRNA Phe , and tRNA Pro contain 5 bp (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The &#936;-arm contains 5 bp in most of the gymnosperm chloroplast tRNAs, except for tRNA Ala and some of the tRNA Trp , tRNA Gly , tRNA Thr , and tRNA Arg in chloroplast. The &#936;-loops of most tRNAs contain 7 bp, apart from tRNA Ala and several of tRNA Cys and tRNA Thr (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gymnosperm chloroplast tRNAs derived from multiple common ancestors</ns0:head><ns0:p>The phylogenetic tree demonstrated the presence of three major clusters covering 64 groups and the different types of all tRNAs (as shown by the different strings in Fig. <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>). We detected 37 groups in cluster I, five in cluster II, and 22 groups in cluster III. Cluster I contains tRNA Ser , tRNA Tyr , tRNA His , tRNA Gln , tRNA Thr , tRNA Pro , tRNA Gly , tRNA Met , tRNA Asp , tRNA Arg , tRNA Ala , tRNA Cys , tRNA Lys , tRNA Glu , tRNA Ile , tRNA Asn , tRNA Val , tRNA Leu , and tRNA Trp . Cluster II contains tRNA His , tRNA Ser , tRNA Tyr , and tRNA Leu . Cluster III contains tRNA Leu , tRNA Ile , tRNA Gly , tRNA Thr , tRNA Ser , tRNA Val , tRNA Glu , tRNA Lys , tRNA Cys , tRNA Gln , tRNA His , tRNA Arg , tRNA Phe , tRNA Ala , and tRNA Met (Fig. <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>). tRNA Ser , tRNA His , and tRNA Leu are present in cluster I but also in cluster II and cluster III, thereby suggesting that these tRNAs evolved from multiple lineages. Most of the tRNAs were found to form more than one group in the phylogenetic tree. In cluster I, the tRNAs that formed two groups in the phylogenetic tree were identified as tRNA Tyr , tRNA Gln , tRNA Met , tRNA Asp , tRNA Ala , tRNA Lys , tRNA Ile , and tRNA Trp , whereas those that clustered to form three groups were determined as tRNA Ser , tRNA Pro , tRNA Arg , tRNA Glu , tRNA Asn , tRNA Val , and tRNA Leu . Moreover, tRNA Thr clustered into four groups. In cluster II, tRNA Ser was found to form two groups. In cluster III, tRNA Gly and tRNA Val were found to form two groups, whereas tRNA Thr formed three groups, tRNA Ile formed four groups. Some tRNAs in cluster III were found to group individually, where these tRNAs containing the anticodons C-G-A in tRNA Ser , U-U-C in tRNA Glu , U-U-U in tRNA Lys , G-C-A in tRNA Cys , U-U-G in tRNA Gln , G-U-G in tRNA His , U-C-U in tRNA Arg , G-A-A in tRNA Phe , U-G-C in tRNA Ala , and C-A-U in tRNA Met all grouped separately (Fig. <ns0:ref type='figure' target='#fig_3'>S1</ns0:ref>). The multiple groupings of different tRNAs suggest that they evolved from multiple common ancestors. Furthermore, the tRNAs presented in cluster III, i.e., tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA His (GUG), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Ile (UAU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Gly (UCC), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU), tended to be the most basic tRNAs and they had undergone gene duplication and diversification to generate other tRNA molecules.</ns0:p></ns0:div> <ns0:div><ns0:head>C-A-U anticodon in tRNA Ile</ns0:head><ns0:p>Our detailed genomic study showed that tRNA Ile also encodes a C-A-U anticodon in addition to the presence of this typical anticodon in tRNA Met . In general, the C-A-U anticodon is recognized as a typical characteristic of tRNA Met and there is only one iso-acceptor. In particular, we found that the tRNA Ile in T. mairei encodes two C-A-U anticodons, and C. debaoensis, <ns0:ref type='bibr'>S. verticillata, D. spinulosum, C. lanceolata, G. biloba, C. deodara, W. mirabilis, G. gnemon, R. piresii, E. equisetina, and W.</ns0:ref> nobilis also encode a C-A-U anticodon (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>, Data S1, Fig. <ns0:ref type='figure'>S3</ns0:ref>).</ns0:p><ns0:p>Transition/transversion of tRNAs A previous study <ns0:ref type='bibr' target='#b47'>(Mohanta et al., 2019)</ns0:ref> showed that the evolutionary rates are almost equal for tRNAs with respect to transition and transversion despite the low probability of transition or transversion events in tRNAs. In this study, we identified several intriguing substitutions of gymnosperm chloroplast tRNAs. Overall, our analysis of the substitution rates detected using the whole set of chloroplast tRNAs showed that average transition rate (15.38) was significantly larger than the average transversion rate (4.81) with a ratio of 3:1 (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). The same transition: transversion ratio bias was found in all the set of tRNAs for tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , and tRNA Leu . The ratio was over 6:1 for tRNA Cys and tRNA Arg . The transition rates for tRNA Trp , tRNA Val , and tRNA Gly were about 10 times higher than their transversion rates. These findings suggest that tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , tRNA Leu , tRNA Cys , tRNA Arg , tRNA Trp , tRNA Val and tRNA Gly underwent transition substitutions more readily than transversion substitutions during their evolution in gymnosperm chloroplast genomes. In addition, the transition rates in tRNA Lys and tRNA Pro were about 15 times higher than their transversion rates. The transition rates in tRNA Asn , tRNA Phe , and tRNA His were about 20 times higher than their transversion rates. These results indicate that tRNAs are much more likely to have undergone transition events rather than transversion events. The highest transversion rate of 12.50 was found in tRNA Ala and the lowest transversion rate of 0.00 in tRNA Asp (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). Correspondingly, tRNA Ala lacks any transitions (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>).</ns0:p><ns0:p>tRNA duplication/loss events In addition to transition and transversion events, gene duplication and loss events have played important roles in gene evolution. Our analysis of duplication and loss events indicated that 153 duplication events (duplication and conditional duplication) have occurred in all of the gymnosperm chloroplast tRNA genes investigated in this study (Fig. <ns0:ref type='figure' target='#fig_7'>S2</ns0:ref>). In addition, 220 gymnosperm chloroplast tRNA gene loss events were detected (Table <ns0:ref type='table'>S2</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_7'>S2</ns0:ref>). Thus, the loss of genes was slightly more frequent than their duplication for gymnosperm chloroplast tRNA genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>tRNAs are major genetic components of semi-autonomous chloroplasts and our analysis of gymnosperm chloroplast genomes showed that they have several basic conserved genomic features. The gymnosperm chloroplast genomes investigated in the present study were found to encode 28 to 33 tRNA isotypes, thereby indicating that there is substantial variation in the quantity of tRNAs in gymnosperm chloroplast genomes. The lack of tRNA Ala in R. piresii and T. mairei, and the absence of tRNA Val in T. mairei were interesting. Thus, it is necessary to understand how the translation process is conducted in chloroplasts without these crucial tRNAs. According to previous studies <ns0:ref type='bibr' target='#b61'>(Treangen &amp; Rocha, 2011;</ns0:ref><ns0:ref type='bibr' target='#b47'>Mohanta et al., 2019)</ns0:ref>, it is likely that the deficiency of these tRNAs is compensated for by the tRNAs transferred from other organelle genomes such as nuclear genome and mitochondrial genome, which may have multiple functions through modification during the translation process in chloroplasts. In addition to the absence of tRNA Ala and tRNA Val , all of the gymnosperm plants were shown to not encode selenocysteine tRNA and its suppressor tRNA in their chloroplast genomes (Table <ns0:ref type='table'>2</ns0:ref>). Selenocysteine tRNA and its suppressor tRNA were also not detected in the chloroplast of Oryza sativa <ns0:ref type='bibr' target='#b45'>(Mohanta &amp; Bae, 2017)</ns0:ref>.</ns0:p><ns0:p>In addition to the presence of C-A-U anticodon in tRNA Met , we found that tRNA-CAU is present in tRNA Ile (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). Similarly, the C-A-U anticodon was detected in tRNA Ile in Bacillus subtilis (Ehrenberg) Cohn and spinach <ns0:ref type='bibr' target='#b26'>(Kashdan &amp; Dudock, 1982;</ns0:ref><ns0:ref type='bibr' target='#b31'>K&#246;hrer et al., 2014)</ns0:ref>. The possible mechanism that governs the specificity of this amino acid may involve modification of the wobble position in the anticodon by a tRNA-modifying enzyme. Chloroplasts originate from bacteria so the tRNA modifications found in bacteria may also occur in chloroplast tRNAs. In bacteria, the tRNA-modifying enzyme TilS can convert the 5&#61602;-C residue in the CAU anticodon of specific tRNA Ile molecules into lysidine to decode 5&#61602;-AUA (Ile) codons instead of 5&#61602;-AUG (Met) codons <ns0:ref type='bibr' target='#b56'>(Soma et al., 2003)</ns0:ref>. In addition, when lysidine decodes isoleucine, the tautomer form of lysidine provides compatible hydrogen bond donor-acceptor sites to allow base pairing with 'A' and this may help to the recognition of the codon AUA instead of AUG <ns0:ref type='bibr' target='#b57'>(Sonawane &amp; Tewari, 2008;</ns0:ref><ns0:ref type='bibr' target='#b54'>Sambhare et al., 2014)</ns0:ref>. The absence of tRNA Ile -lysidine synthetase leads to a failure to modify C34 to lysidine in tRNA Ile (LAU) (i.e., the synthesis of CAU-tRNA Ile ) and this inactivates the translation of AUA codons <ns0:ref type='bibr' target='#b31'>(K&#246;hrer et al., 2014)</ns0:ref>.</ns0:p><ns0:p>During protein coding, a certain species or gene tends to use one or more specific synonym codons, which is referred to as codon usage bias <ns0:ref type='bibr'>(Comeron &amp; Aguad&#233;, 1998;</ns0:ref><ns0:ref type='bibr' target='#b53'>Rota-Stabelli et al., 2012)</ns0:ref>. In the present study, tRNA Arg -CCG was found to be present in the genomes of nine species but absent from C. lanceolata, T. mairei, and E. equisetina. Similarly, tRNA Gly -UCC was shown to be absent from the chloroplast genomes of C. debaoensis, S. verticillata, D. spinulosum, C. lanceolata, T. mairei, and E. equisetina (Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>). These results suggest that gymnosperm chloroplast tRNA genes are characterized by codon usage bias <ns0:ref type='bibr'>(Wei &amp; Jin, 2017;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In general, the secondary structure of tRNAs is characterized as clover leaf-like, except for a few tRNAs with unusual secondary structures <ns0:ref type='bibr' target='#b24'>(J&#252;hling et al., 2018)</ns0:ref>. In our study, we identified clover leaf-like tRNAs with expanded variable loop regions (Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref>). Numerous tRNA Leu , tRNA Ser , and tRNA Tyr were found to have specific variable loop configurations in terms of length and structure, suggesting significant structural variation among chloroplast tRNAs. It is interesting to note that there were also stem-loop structures in variable regions of certain tRNAs in cyanobacteria. This might indicate that similar structural variations exist between chloroplast tRNAs and cyanobacterial tRNAs <ns0:ref type='bibr'>(Mohanta et al., 2017)</ns0:ref>. Future studies will have to determine the biological importance of these variant tRNAs. Noteworthy, a novel tRNA structure lacking the D arm was found for tRNA Gly in W. nobilis (Fig. <ns0:ref type='figure'>3</ns0:ref>). Most tRNAs have a clover-like structure formed by complementary base pairing between small segments <ns0:ref type='bibr' target='#b23'>(Hubert et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b12'>Florentz, 2002)</ns0:ref>. Previous studies have showed that the acceptor arm of tRNAs in chloroplasts contain 7 bp to 9 bp, the D-arm contains 3 bp to 4 bp, the D-loop has 4 nt to 12 nt, the anticodon arm has 5 bp, the anticodon loop contains 7 nt, the variable region comprises 4 nt to 23 nt, and &#936;-arm contains 5 bp, and the &#936;-loop has 7 nt <ns0:ref type='bibr'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b48'>Mohanta &amp; Hanhong, 2017;</ns0:ref><ns0:ref type='bibr' target='#b47'>Mohanta et al., 2019)</ns0:ref>. In the present study, we found that the acceptor arm of chloroplast tRNAs contains 3 bp to 7 bp in 373 tRNAs, where the D-arm has 3 bp or 4 bp and the D-loop usually contains 7 nt to 11 nt. The anticodon loop of gymnosperm chloroplast tRNAs generally contains 7 nt, and thus the sequence of the anticodon loop is typically conserved (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The variable loop of different tRNAs contain 3 nt to 23 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The &#936;-arm of gymnosperm chloroplast tRNAs generally contains 5 bp and the &#936;loop has 7 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). Our results are consistent with previous findings (Wilusz, 2015; <ns0:ref type='bibr' target='#b48'>Mohanta &amp; Hanhong, 2017)</ns0:ref> and they suggest that chloroplast RNAs are significantly conserved. The consensus sequence 'U-U-C-N-A-N 2 ' was found in the &#936; region (Table <ns0:ref type='table'>4</ns0:ref>). Previous studies also reported the existence of a similar sequence in the &#936;-loop of tRNAs in Oryza sariva and Cyanobacterial <ns0:ref type='bibr' target='#b45'>(Mohanta &amp; Bae, 2017;</ns0:ref><ns0:ref type='bibr'>Mohanta et al., 2017)</ns0:ref>. This suggested the short and conserved motif was universal in &#936; loop. Our study as well as previous researches revealed that the consensus 'U-U-C-N-A-N 2 ' motif in the &#936; region might be a general consensus motif of canonical tRNAs.</ns0:p><ns0:p>Our phylogenetic analysis detected three clear clusters and many tRNA groups. Some tRNAs (tRNA Ser , tRNA His , and tRNA Leu ) in cluster I and cluster II were also in cluster III, thereby indicating that these tRNAs evolved from multiple lineages by gene duplication and gene divergence. Moreover, anticodon types comprising CGA, UUC, UUU, GCA, UUG, GUG, UCU, UGC, and CAU appeared several times in the phylogenetic tree, and thus the corresponding tRNAs evolved from multiple common ancestors. The overlapping of tRNAs groups demonstrates that these tRNAs might have diverse common ancestors in the evolutionary process <ns0:ref type='bibr' target='#b45'>(Mohanta &amp; Bae, 2017)</ns0:ref>. Phylogenetic analysis also showed that tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA His (GUG), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Ile (UAU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Gly (UCC), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU) in cluster III tended to be the most basic tRNAs, whereas tRNA Met tended to be the most original tRNA. Overall, the results clearly indicate that the tRNAs encoded in gymnosperm chloroplast genomes have multiple common evolutionary ancestors.</ns0:p><ns0:p>Our results also provided insights into the gene substitution rates in gymnosperm chloroplast tRNAs. Overall, the average transition rate for tRNAs was greater than the transversion rate, where the relationship was about 3:1 (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). In all of the chloroplast tRNAs, the average transition rate was slightly higher than the average transversion rate, thereby indicating that chloroplast tRNAs have unequal substitution rates.</ns0:p><ns0:p>In addition to the transition and transversion events in tRNAs, loss and duplication events have played significant roles in the evolution of tRNAs in gymnosperm chloroplast genomes <ns0:ref type='bibr' target='#b17'>(He &amp; Zhang, 2006;</ns0:ref><ns0:ref type='bibr'>Magadum et al., 2013)</ns0:ref>. In general, the gene loss events tended to occur after whole genome duplication events. We found 153 duplication events and 220 loss events in gymnosperm chloroplast tRNAs, and thus loss events have occurred slightly more frequently than duplication events (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our basic structure analysis showed that gymnosperm chloroplast genomes encode 25 to 30 anticodon-specific tRNAs. The acceptor arm of chloroplast tRNA contains 3 bp to 7 bp, the Darm has 3 bp or 4 bp, the D-loop contains 7 nt to 11 nt mainly, and the anticodon loop usually contains 7 nt. In different tRNAs, the variable loop contains 3 nt to 23 nt. The &#936;-arm contains a conserved sequence comprising U-U-C-N-A-N 2 . In addition, tRNA Ala was found to be absent from R. piresii and T. mairei, and tRNA Val was lacking in T. mairei. Gymnosperm chloroplasts do not encode selenocysteine tRNA and its suppressor tRNA in their genomes. A CAU anticodon is encoded in tRNA Met as well as in tRNA Ile . A novel tRNA structure lacking the D arm was identified for the chloroplast tRNA Gly of W. nobilis. Numerous tRNA Leu , tRNA Ser , and tRNA Tyr types were found to have expanded variable regions, forming stem-loop structures with up to 7 bp in tRNAs Ser . Phylogenetic analysis showed that tRNAs such as tRNA Ser , tRNA His , and tRNA Leu in cluster I and cluster II were also present in cluster III, thereby indicating that these tRNAs have multiple evolutionary origins. The overlapping of tRNA groups demonstrates that tRNAs might have multiple common ancestors in the evolutionary process. tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Ile (CAU) and others tended to be the most basic tRNAs, whereas tRNA Met tended to be the most original tRNA. Transition and transversion analyses of tRNAs indicated that different tRNAs harbored their own transition/transversion rates, i.e., it was isoacceptor specific. And the transition rate was generally higher than the transversion rate. Furthermore, gene loss events (220) have occurred slightly more frequently than gene duplication events (153) in gymnosperm chloroplast tRNAs. Our results provide new insights into the evolution of gymnosperm chloroplast tRNAs and their diverse roles. acuteserrata forests? PloS one 9:e89886. DOI: 10.1371/journal.pone.0089886</ns0:p></ns0:div> <ns0:div><ns0:head>Figure Legends</ns0:head><ns0:p>Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref> Certain tRNAs in C. debaoensis, D. spinulosum, G. biloba, C. deodara, and R. piresii contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from C. debaoensis, D. spinulosum, G. biloba, C. deodara, R. piresii were observed to contain an expanded variable stem and loop. The anti-codon loop of tRNA Ser (except for tRNA Ser -GCU of D. spinulosum) was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and tRNA Leu harbor the consensus sequence C-U-N-A-N 2 -A. Fig. <ns0:ref type='figure' target='#fig_7'>2</ns0:ref> Certain tRNAs in S. verticillata, C. lanceolata, T. mairei, W. mirabilis, and G. gnemon contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon were observed to contain a variable stem and loop. The anti-codon loop of tRNA Ser was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and the consensus sequence was C-U-N-A-N 2 -A for tRNA Leu . Fig. <ns0:ref type='figure'>3</ns0:ref> An abnormal tRNA structure lacking the D-arm found in W. nobilis. The tRNA Gly with anti-codon UCC was found lacking D-arm. Fig. <ns0:ref type='figure'>4</ns0:ref> Clover leaf-like structure of gymnosperms tRNA. The tRNA contains the Acceptor arm (3-7 bp, green, &gt;95% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region (3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, green, &gt;95% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). Several tRNAs harbor the nucleotides of C-C-A tail.</ns0:p></ns0:div> <ns0:div><ns0:head>Supplementary materials</ns0:head><ns0:p>Data S1 tRNA sequences of gymnosperms chloroplast genome conducted in the study.</ns0:p></ns0:div> <ns0:div><ns0:head>Table S1</ns0:head><ns0:p>Nucleotide composition in different parts of clover-structure of chloroplast genome tRNA. AC-arm, Acceptor arm; ANC-arm, Anti-codon arm; ANC-loop, Anti-codon loop; &#936;-arm, Pseudouridine arm; &#936;-loop, Pseudouridine loop. The base pairs in the AC-arm were counted according to the predicted clover structures of tRNAs. Table <ns0:ref type='table'>S2</ns0:ref> Loss events of chloroplast genomic tRNAs. Manuscript to be reviewed A view of the gymnosperms in analysis.</ns0:p><ns0:p>Statistics of the 12 gymnosperms in the study.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed 1 Table <ns0:ref type='table' target='#tab_2'>3</ns0:ref>. Distribution of anti-codons in the chloroplast genome of gymnosperms. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>An abnormal tRNA structure lacking the D-arm found in W. nobilis.</ns0:p><ns0:p>The tRNA Gly with anti-codon UCC was found lacking D-arm.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Clover leaf-like structure of gymnosperms tRNA.</ns0:p><ns0:p>The tRNA contains the Acceptor arm (3-7 bp, green, &gt;95% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved ), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region</ns0:p><ns0:p>(3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, green, &gt;95% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). Several tRNAs harbor the nucleotides of C-C-A tail.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>&#61623;</ns0:head><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Y. C. Zhong &amp; C. J. Chen (KM459003), Dioon spinulosum Dyer ex Eichler (NC_027512), Ginkgo biloba L. (NC_016986), Cedrus deodara (Roxb.) G. Don (NC_014575), Wollemia nobilis W. G. Jones, K. D. Hill &amp; J. M. Allen (NC_027235), Retrophyllum piresii Silba C. N. (KJ017081), Sciadopitys verticillata (Thunb.) Sieb. et Zucc. (NC_029734), Cunninghamia lanceolata (Lamb.) Hook. (NC_021437), Taxus mairei (Lemee et Levl.) Cheng et L. K. Fu (KJ123824), Welwitschia mirabilis Hook.f. (EU342371), Gnetum gnemon L. (KR476377), and Ephedra equisetina Bge. (</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Wang RJ, Cheng CL, Chang CC, Wu CL, Su TM, Chaw SM. 2008. Dynamics and evolution of the inverted repeat-large single copy junctions in the chloroplast genomes of monocots. BMC evolutionary biology 8:36. DOI: 10.1186/1471-2148-8-36 Wang XQ, Ran JH. 2014. Evolution and biogeography of gymnosperms. Molecular phylogenetics and evolution 75:24-40. DOI: 10.1016/j.ympev.2014.02.005 Wei QK, Jin HY. 2017. The complete chloroplast genome sequence of Morus cathayana and Morus multicaulis, and comparative analysis within genus Morus L. Peer J 5:e3037. DOI: 10.7717/peerj.3037 Wilusz JE. 2015. Controlling translation via modulation of tRNA levels. Wiley interdisciplinary reviews. RNA 6:453-470. DOI: 10.1002/wrna.1287 Yu F, Wang DX, Yi XF, Shi XX, Huang YK, Zhang HW, Zhang XP. 2014. Does animalmediated seed dispersal facilitate the formation of Pinus armandii-quercus aliena var.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Fig. S1</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>The phylogenetic tree constructed by 373 tRNAs. Multiply tRNAs are shown by different colors. Different groups are marked by different strings. The phylogenetic clades with low bootstrap replicates were collapsed with 50% cutoff values. Phylogenetic analysis illustrates that Gymnosperm chloroplast tRNA derived from common multiple ancestors. Fig. S2 The loss and duplication tree. Blue: Duplication events; Gray: Loss events; D: Duplication node; cD: Conditional Duplication node. Fig. S3 tRNA gene content in analyzed gymnosperms chloroplast genome. The tRNA genes are shown in the left (top to bottom). Boxes in light green, dark green, and white represent one copy of tRNA genes, two copies of tRNA genes, and the absence of tRNA genes. PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>1</ns0:head><ns0:label /><ns0:figDesc>Table 1. A view of the gymnosperms in analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Order</ns0:cell><ns0:cell>Family</ns0:cell><ns0:cell /><ns0:cell>Subfamily</ns0:cell><ns0:cell /><ns0:cell>Genus</ns0:cell><ns0:cell>Species</ns0:cell><ns0:cell cols='2'>NCBI Locus</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cycadales</ns0:cell><ns0:cell cols='2'>Cycadaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Cycas</ns0:cell><ns0:cell>debaoensis</ns0:cell><ns0:cell>KM459003</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Zamiaceae</ns0:cell><ns0:cell>Diooideae</ns0:cell><ns0:cell /><ns0:cell>Dioon</ns0:cell><ns0:cell>spinulosum</ns0:cell><ns0:cell cols='2'>NC_027512</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Ginkgoales</ns0:cell><ns0:cell cols='2'>Ginkgoaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Ginkgo</ns0:cell><ns0:cell>biloba</ns0:cell><ns0:cell cols='2'>NC_016986</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Pinales</ns0:cell><ns0:cell cols='2'>Pinaceae</ns0:cell><ns0:cell>Abieteae</ns0:cell><ns0:cell /><ns0:cell>Cedrus</ns0:cell><ns0:cell>deodara</ns0:cell><ns0:cell cols='2'>NC_014575</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Araucariales</ns0:cell><ns0:cell cols='2'>Araucariaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Wollemia</ns0:cell><ns0:cell>nobilis</ns0:cell><ns0:cell cols='2'>NC_027235</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Podocarpaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Retrophyllum</ns0:cell><ns0:cell>piresii</ns0:cell><ns0:cell>KJ017081</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cupressales</ns0:cell><ns0:cell cols='2'>Sciadopityaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Sciadopitys</ns0:cell><ns0:cell>verticillata</ns0:cell><ns0:cell cols='2'>NC_029734</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Cupressaceae</ns0:cell><ns0:cell>Cunninghami a</ns0:cell><ns0:cell /><ns0:cell>Cunninghamia</ns0:cell><ns0:cell>lanceolata</ns0:cell><ns0:cell cols='2'>NC_021437</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Taxaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Taxus</ns0:cell><ns0:cell>mairei</ns0:cell><ns0:cell>KJ123824</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Welwitschiales</ns0:cell><ns0:cell cols='2'>Welwitschiaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Welwitschia</ns0:cell><ns0:cell>mirabilis</ns0:cell><ns0:cell>EU342371</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gnetales</ns0:cell><ns0:cell cols='2'>Gnetaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Gnetum</ns0:cell><ns0:cell>gnemon</ns0:cell><ns0:cell>KR476377</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Ephedrales</ns0:cell><ns0:cell cols='2'>Ephedraceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Ephedra</ns0:cell><ns0:cell>equisetina</ns0:cell><ns0:cell cols='2'>NC_011954</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Met/fMet</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Tyr</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cys</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Trp</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Selenocystei ne</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Suppressor</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>28</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Distribution of anti-codons in the chloroplast genome of gymnosperms.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Each of the gymnosperm chloroplast genomes encodes 25 to 30 anticodon-specific tRNAs. E.</ns0:cell></ns0:row><ns0:row><ns0:cell>equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S.</ns0:cell></ns0:row><ns0:row><ns0:cell>verticillata, and C. lanceolata encode 28 anticodons, and D. spinulosum, C. deodara, W.</ns0:cell></ns0:row><ns0:row><ns0:cell>mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis, G.</ns0:cell></ns0:row><ns0:row><ns0:cell>gnemon and G. biloba encodes 30 anticodons.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Transition and transversion rate of chloroplast tRNA.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020) Manuscript to be reviewed 1 PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)Manuscript to be reviewed</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43190:2:0:NEW 16 Jul 2020)</ns0:note> </ns0:body> "

" Dear Editor, Thank you very much for returning our manuscript. We are very grateful to you for your time and patience, as well as the reviewers’ generous and thoughtful comments. These suggestions and comments are all valuable and very helpful for revising and improving our MS. In the revised manuscript, we have studied all these comments carefully and have made thoroughly corrections, and responded point by point to the comments as itemized below (our responses are in bold characters). If you have any questions regarding the manuscript, please feel free to contact the corresponding author: Zhong-Hu Li Key Laboratory of Resource Biology and Biotechnology in Western China (Ministry of Education), College of Life Sciences, Northwest University, Xi’an 710069, China Fax: +86 29 888302411; E-mail: [email protected] We shall look forward to hearing from you at your earliest convenience. Yours sincerely, Zhonghu Li, Ph. D E-mail: [email protected] Editor and Reviewer comments: Editor comments (Genlou Sun) MAJOR REVISIONS Major revision is still needed to improve the manuscript. Reply: Thank you very much for your comments. We are so grateful to you for providing an opportunity for revising the manuscript. According to these helpful suggestions and comments, we have carefully and thoroughly revised the manuscript and hope to this version is suitable to the journal. Reviewer 2 (Anonymous) 1. Basic reporting: The manuscript has improved to some extent, but there are still some unclear major and minor points. As this is a detailed bioinformatic study, the authors have to convince the reader that the analyses were meticulously carried out, taking into account that the reader cannot examine the correctness of every detail. However, at several positions in the text doubts are created whetherthe study was performed with maximum diligence, as some of the described findings and conclusions are either misleading or not comprehensible. In my previous review, I already criticized the endless listings of tRNAs in the text. The authors should illustrate their findings in additional summary figures in the main manuscript. I have made proposals for Fig. 4 and Table 4 (see below). Very helpful would also be a sequence/structure alignment of all analyzed gymnosperm chloroplast tRNAs (e.g. as in the tRNAdb). Reply: Thank you very much for your valuable and constructive comments. We have carefully checked and improved the unclear major and minor points in order to provide meticulous analyses in the revised manuscript with our maximum diligence. We have corrected the misleading or incomprehensible description of findings and conclusions. In line 201 to 202, we have rewritten “…, and tRNATyr contain a loop structure that is similar to the anti-codon loop in the variable region of tRNAs.” to “…, and tRNATyr contain expanded variable stem/loops.” according to your suggestions. In line 240 to 243, we have rewritten “Small conserved consensus sequences were found in the Ψ region according to multiple sequence alignment in 20 members of the tRNA gene family. The Ψ-loop in tRNAs was found to contain a conserved sequence comprising U-U-C-X-A-X2.” To “Small conserved consensus sequences were found in the Ψ region. To be specific, except for tRNASer, the Ψ-loop in tRNAs was found to contain a conserved sequence comprising U-U-C-N-A-N2 according to a multiple sequence alignment of 20 members of the tRNA gene family.” We have re-predicted and improved the incorrect variable loop structures in Fig. 1 and 2. And we have supplemented the exceptions to better differentiate the features of stem-loop elements in D loop, D-arm, Ψ-arm, anticodon region, and variable loop region. We have added the summary figure (Fig. S3) to emphasize our findings in the revised manuscript. We have improved Fig. 4 by integrating into the tRNA secondary structure the conservations (different colors represent different conservation of base identities) discovered in gymnosperm chloroplast tRNA set, and we have improved Table 4 by supplementing the sequence of both strands of the stems. We have aligned all analyzed gymnosperm chloroplast tRNAs sequences before constructing the phylogenetic tree. And through the tree, we have obtained the phylogenetic relationship of gymnosperm tRNAs (please see Fig. S1 and Fig. S3). Major concerns: a) At multiple positions in the text, gaps between separate words are lacking. Reply: Many thanks to you for your kindly comment. We have checked carefully throughout the manuscript and corrected these mistakes. b) for censensus sequence motifs, use “N” instead of “X” to indicate any of the four bases, or further differentiate (R, Y, S, W, K and so forth). Reply: Thank you very much for your helpful suggestions. In the revised manuscript, we have corrected “X” to “N” for consensus sequence motifs. c) I find the Introduction somewhat too long in the second part describing the basics of tRNA sequence and structural elements and their functions. Reply: Agreed. We have deleted the superfluous relevant contents in the Introduction according to your advice. d) lines 199/200, “… contain a loop structure that is similar to the anti-codon loop in the variable region of tRNAs.” This is misleading. What you want to say is that there are tRNAs with expanded variable loops, some of which form extended stem-loop structures, some shorter stem-loop structures and others are void of base-pairing within the variable loop. This is also a feature of corresponding bacterial tRNA species. My proposal for rewriting in line 199: “As shown in Fig. 1 and Fig. 2, tRNALeu, tRNASer, and tRNATyr contain expanded variable (stem-)loops.” I does not make sense to say that these variable loops are similar to the anticodon loop. Also, some of the variable loop structures in Fig. 1 and 2 are incorrect: in Fig. 1F, a loop with a single bp in the stem is unliekley to form; likewise, in Fig. 2F, a stem without apical loop does not form. In Fig. 2B, the variable loop stem can be extended to 4 bp. Reply: We strongly agree with your brilliant view. In line 199 to 201, we have rewritten the sentence “As shown in Fig. 1 and Fig. 2, tRNALeu, tRNASer, and tRNATyr contain a loop structure that is similar to the anti-codon loop in the variable region of tRNAs.” to “ As shown in Fig. 1 and Fig. 2, tRNALeu, tRNASer, and tRNATyr contain expanded variable stem/loops.” according to your helpful proposal. We have checked and re-predicted the structures of tRNAs in Fig. 1 and Fig. 2, and corrected the inappropriate variable loop structures of Fig. 1F, Fig. 2B, and Fig. 2F. Thank you again for your helpful comments. e) lines 201/202: tRNASer-GCU of D. spinulosum does not follow the consensus N-U-N-G-A-A-N. In lines 202-206, the authors should better differentiate between variable loop region (the entire sequence), and stem-loop elements in the variable loop region. I propose to rewrite from line 202 on: “The variable loop region is predicted to fold into stem-loop structures with apical loops of 3 to 7 nt in tRNASer and several tRNALeu variants. The stems contain up to 7 bp (Fig. 1 and 2). The expanded variable loop structures may play important functions during the protein translation process in chloroplasts.” At some point, these structural features should be compared with those of bacterial counterparts. Reply: Thanks so much for your kindly suggestions. In the revised manuscript, we have supplemented the exceptional case of tRNASer-GCU of D. spinulosum in the description of consensus N-U-N-G-A-A-N. We have corrected the sentence “The variable loop region contains 3 to 7 nucleotides nt in tRNASer, tRNATyr to tRNALeu. The stem of the variable loop region contains 0 to 7 nt. tRNASer and tRNALeu have variable loop regions with 6 nt and 3 nt, respectively (Fig. 1, Fig. 2). The novel loop structures identified in this study may play important functions during the protein translation process in chloroplasts.” to “The variable loop region is predicted to fold into stem-loop structures with apical loops of 3 to 7 nt in tRNASer and several tRNALeu variants. The stems contain up to 7 bp (Fig. 1 and 2). The expanded variable loop structures may play important functions during the protein translation process in chloroplasts.” according to your suggestion. And in Discussion part, we have supplemented related discussions including the comparison of structural variation between chloroplast tRNA and bacterial tRNAs. f) line 224, Tables 2 and 3: for C. lanceolata, two tRNATrp isoacceptors are indicated in Table 2, but only 1 in Table 3. Reply: Thanks a lot for your helpful comments. We have checked carefully and altered this mistake in the revised manuscript. g) lines 230-235 (“A” is present in the first and the last position of the D-loop): this is not true for tRNAGln, Gly, Ile, Leu, Met. / (In addition, in the final two positions of the Ψ-arm, all of the tRNAs were found to have conserved “G-G” nucleotides, except for tRNAArg): according to Table 4, this is also not the case for tRNACys, Phe and Val. Reply: We are very grateful to you for your helpful and constructive advice. We have improved the sentences in line 230-235 to “ “A” is present in the first and the last position of the D-loop apart from tRNAGln, tRNAGly, tRNAIle, tRNALeu, and tRNAMet.” and “In addition, in the final two positions of the Ψ-arm, all of the tRNAs were found to have conserved “G-G” nucleotides, except for tRNAArg, tRNACys, tRNAPhe, and tRNAVal (Table 4).” in the revised manuscript correspondingly. h) lines 233-235: it is unclear here if you found the consensus U-U-C-X-A-X2 only in a multiple sequence alignment of 20 members of the tRNA gene family or in all tRNAs analyzed here; please clarify. Reply: Thank you so much for your kindly advice. The consensus U-U-C-X-A-X2 was found through a multiple sequence alignment of 20 members of the tRNA gene family. And we have improved this sentence to make it more explicit. i) lines 238-251: generally use “bp” when describing the length of stems; line 240: an acceptor stem of 1 or 2 bp is non-existent. In Fig. 3, the acceptor stem of W. nobilis tRNAGly likely forms a labile 7-bp acc. stem with two non-canonical A-C bp. This should also be changed in Table S1, replace “2” by “7” or “(7)” in the column “AC-arm” for W. nobilis tRNAGly. How does the AC-arm for tRNAThr of Sciadopitys_verticillata_94443 (1 bp) look like?. Provide a legend to Table S1 where you define how you counted the base pairs in the AC-arm; how do the AC-arms with 3 or 6 bp look like? Reply: Thank you very much for your valuable comments. In the revised manuscript, we have corrected “cn” to “bp” when describing the length of stems. We have corrected the statistics of the length of acceptor stem. We have corrected “2” to “7” in the column “AC-arm” for W. nobilis tRNAGly in Table S1 according to Fig. 3. We have re-predicted the clover structure of tRNAThr in Sciadopitys verticillata_94443 (please see below result, A) and modified the previous statistics of this tRNA. The base pairs in the AC-arm were counted according to the predicted clover structures of tRNAs, and we have added this legend to Table S1. The structures with the 3-bp AC-arm (Please see B, tRNAMet in Taxus mairei_119804) and 6-bp AC-arm (Please see C, tRNAMet in Retrophyllum piresii_6159) are as follows: Thank you again for your suggestions. j) line 242, C in the last position of the D-arm: is this true for all tRNAs? Reply: Many thanks to you for your helpful suggestion. We have checked this content carefully. Here for most tRNAs such as tRNAAla, tRNAAsn, tRNAAsp, tRNACys, tRNAGlu, tRNAHis, tRNAIle, and tRNAPhe but not for all tRNAs, there are “G” in the initial position and “C” in the last position. And we have clarified this in the manuscript. k) Table 4: here, the sequence of both strands of the stems should be shown; I assume that the entire sequences of the loops are listed in this table. Use a presentation such as in the tRNAdb. Change the title to “Conserved sequence motifs in chloroplast tRNAs from gymnosperms”; what did you mean by “selected gymnosperms? All those anlyzed here? Reply: We thank you very much for your constructive comment. We have listed both strands of the stems in Table 4 using a presentation such as in the tRNAdb. And we have changed the title to “Conserved sequence motifs in chloroplast tRNAs from gymnosperms” as your suggestion. “selected gymnosperms” here refered to the gymnosperm species investigated in this study (C. debaoensis, D. spinulosum, G. biloba, C. deodara, W. nobilis, R. piresii, S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon, and E. equisetina.). All chloroplast genomic tRNAs of these plant species were analyzed. Thanks again for your comments. l) line 307, correct: “… the highest transversion rate of 12.5 was found in tRNAAla and the lowest transversion rate of in tRNAAsp (Table 5). Correspondingly, tRNAAla lacks any any transitions (Table 5).” Reply: Thank you very much for your helpful advice. We have corrected the sentence “The highest transition rate of 25.00 was found in tRNAAsp and the highest transversion rate of 0.00 was found in tRNAAla and the lowest transversion rate of 0.00 in tRNAAsp (Table 5). Similarly, tRNAAla was found to have a transversion high rate but no significant transitions (Table 5).” To “The highest transversion rate of 12.5 was found in tRNAAla and the lowest transversion rate of 0.00 in tRNAAsp (Table 5). Correspondingly, tRNAAla lacks any transitions (Table 5).” according to your suggestion. m) line 328-330: it is unclear what you mean here; are you alluding to the possibility that these missing tRNAs are encoded in the nuclear genome and are imported into the chloroplast? Might there be tRNA editing in chloroplasts? Reply: Thanks a lot for your helpful comment. Yes, the missing tRNAs may be made up by tRNAs transferred from nuclear genome and mitochondrial genome according to the relevant literatures. We have improved this content to make it clearer. We have rewritten the sentence in 328-330 to “According to previous studies (Treangen & Rocha, 2011; Mohanta et al., 2019), it is likely that the deficiency of these tRNAs is compensated for by the tRNAs transferred from other organelle genomes such as nuclear genome and mitochondrial genome, which may have multiple functions through modification during the translation process in chloroplasts.” There might be tRNA editing in chloroplasts. And in the second paragraph of Discussion, we have discussed this possibility. Thank you again for your suggestions. n) lines 345-349: please rewrite, not clear what you want to express here beyond what is expressed before. Reply: Thank you very much for your kindly advice. In the revised manuscript, we have rewritten these sentences to “In addition, when lysidine decodes isoleucine, the tautomer form of lysidine provides compatible hydrogen bond donor-acceptor sites to allow base pairing with “A” and this may help to the recognition of the codon AUA instead of AUG (Sonawane & Tewari, 2008; Sambhare et al., 2014). The absence of tRNAIle-lysidine synthetase leads to a failure to modify C34 to lysidine in tRNAIle (LAU) (i.e., the synthesis of CAU-tRNAIle) and this inactivates the translation of AUA codons (Köhrer et al., 2014).” o) line 371: there is no tRNA with a 2-bp acceptor stem (see above); please rewrite. Reply: Done. p) lines 378/379: the consensus U-U-C-X-A-X2” motif in the Ψ region: isn’t this a general consensus motif of canonical tRNAs? Please discuss this point. Reply: Thank you very much for your helpful comment. We hold that the consensus “U-U-C-X-A-X2” motif in the Ψ region is a general consensus motif of canonical tRNAs of gymnosperms. And previous researchers found similar consensus motifs in Oryza sariva and Cyanobacterial (Mohanta 2017; Mohanta et al., 2017). We have discussed this point in the revised manuscript according to your suggestion. q) line 423: what do you mean by “Transition and transversion analyses of tRNAs indicated that tRNAs are iso-acceptor specific”? Do you mean that transition/transversion patterns were isoacceptor-specific? Reply: Many thanks to you for your kindly suggestion. From transition and transversion analyses of tRNAs, we could know that different tRNA isotypes have their own transition/transversion rates, thus we described it is iso-acceptor specific. And on the whole, the transition rate was higher than the transversion rate. And we have improved this sentence to make it specific and clear. r) Fig. 4 is too trivial: here the authors should integrate into the tRNA 2D structure the conservations they found in their gymnosperm chloroplast tRNA set, differentiating between base identities that are e.g. >50%, > 75% and 100% conserved. Reply: Thank you very much for your helpful advice. We have improved Fig. 4 and integrated into the tRNA secondary structure the conservations discovered in gymnosperm chloroplast tRNAs according to your suggestion. In Fig. 4, gymnosperm chloroplast tRNA components were shown in different colors to differentiate their identities and conservation. s) add legends to supplementary figures and tables with explanations for the reader. Reply: Done. Minor comments: - line 209, rewrite: “The genomes of the species analyzed were found to code for at least two copies of tRNAMet-CAU/tRNAfMet-CAU. Each of the gymnosperm chloroplast genomes encodes 25 to 31 anticodon-specific tRNAs (Tables 2 and 3), where E. equisetina …” Reply: Thanks a lot for your useful comments. We have rewritten this sentence in the revised manuscript. - line 224, rephrase: “Two tRNATrp isoacceptors are present in E. equisetina chloroplasts, compared with a single one in the other gymnosperm species analyzed in this study.” Reply: Thank you so much for your kindly advice. We have rephrased it. - line 228: separate paragraph heading from text. Reply: Done. - Table 4, footnote 2, rephrase: “Note that the consensus sequences are shown from 5’ to 3’.” Reply: Thanks very much for your helpful suggestions. We have rephrased this in the revised manuscript. - line 270, rewrite: “… tRNAs in cluster III were found to group individually, where these tRNAs …” Reply: Many thanks to you for your kindly comment. We have rewritten this sentence. - line 333_ “… in the chloroplast of Oryza …” Reply: Done. - line 359 ff., rewrite: “In our study, we identified cloverleaf-like tRNAs with expanded variable loop regions (Fig. 1 and 2). Numerous tRNALeu, tRNASer, and tRNATyr were found to have specific variable loop configurations in terms of length and structure, suggesting significant structural variation among chloroplast tRNAs. Future studies will have to determine the biological importance of these variant tRNAs. Noteworthy, a novel tRNA structure lacking the D arm was found for tRNAGly in W. nobilis (Fig. 3).” (As said before, compare with bacterial tRNAs) Reply: We are very grateful to you for your helpful comments and suggestions. In the revised manuscript, we have rewritten these sentences according to your advice, and we have supplemented relevant discussions involving the comparison of structural variation between chloroplast tRNA and bacterial tRNAs. We have added “It is interesting to note that there were also stem-loop structures in variable regions of certain tRNAs in cyanobacteria. This might indicate that similar structural variations exist between chloroplast tRNAs and cyanobacterial tRNAs (Mohanta et al., 2017).” Thank you again for your advice. - line 378, rewrite: “The consensus sequence “U-U-C-X-A-X2” was found in the Ψ region (Table 4).” Reply: Thank you so much for your helpful suggestion. We have rewritten this sentence. - line 385: “… appeared several times in the phylogenetic tree, and thus …” Reply: Thanks a lot for your kindly comment. We have rephrased this sentence in the revised manuscript. - line 414, rewrite: “A CAU anticodon is encoded in tRNAMet as well as in tRNAIle. A novel tRNA structure lacking the D arm was identified for the chloroplast tRNAGly of W. nobilis. Numerous tRNALeu, tRNASer, and tRNATyr types were found to have expanded variable regions, forming stem-loop structures with up to 7 bp in tRNAsSer.” Reply: Thank you very much for your valuable advice. In the revised manuscript, we have rewritten the sentence in line 414 to “A CAU anticodon is encoded in tRNAMet as well as in tRNAIle. A novel tRNA structure lacking the D arm was identified for the chloroplast tRNAGly of W. nobilis. Numerous tRNALeu, tRNASer, and tRNATyr types were found to have expanded variable regions, forming stem-loop structures with up to 7 bp in tRNAsSer.” according to your suggestion. Experimental design Requires improved illustration of results Reply: Thank you so much for your helpful comment. We have carefully and thoroughly improved the results and we hope this revised manuscript is more suitable. Validity of the findings expand comparison of gymnosperm chloroplast tRNA features with those of bacterial tRNAs in the discussion Reply: Thank you a lot for your helpful suggestion. In the Discussion, we have supplemented comparison of gymnosperm chloroplast tRNA features with those of bacterial tRNAs according to your comment. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Gymnosperms such as ginkgo, conifers, cycads, and gnetophytes are vital components of land ecosystems, and they have significant economic and ecologic value, as well as important roles as forest vegetation. In this study, we investigated the structural variation and evolution of chloroplast transfer RNAs (tRNAs) in gymnosperms. Chloroplasts are important organelles in photosynthetic plants. tRNAs are key participants in translation where they act as adapter molecules between the information level of nucleic acids and functional level of proteins. The basic structures of gymnosperm chloroplast tRNAs were found to have family-specific conserved sequences. The tRNA &#936;-loop was observed to contain a conforming sequence, i.e., U-U-C-N-A-N 2 . In gymnosperms, tRNA Ile was found to encode a 'CAU' anticodon, which is usually encoded by tRNA Met . Phylogenetic analysis suggested that plastid tRNAs have a common polyphyletic evolutionary pattern, i.e., rooted in abundant common ancestors. Analyses of duplication and loss events in chloroplast tRNAs showed that gymnosperm tRNAs have experienced little more gene loss than gene duplication. Transition and transversion analysis showed that the tRNAs are isoacceptor specific and they have experienced unequal evolutionary rates. These results provide new insights into the structural variation and evolution of gymnosperm chloroplast tRNAs, which may improve our comprehensive understanding of the biological characteristics of the tRNA family.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Gymnosperms originated in the Paleozoic Devonian Period (about 385 million years ago), and they are key groups in terms of the transformation from spore reproduction to seed reproduction in higher plants <ns0:ref type='bibr' target='#b14'>(Gerrienne et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Crisp &amp; Cook, 2011)</ns0:ref>. According to the latest phylogenetic classification, gymnosperm species are divided into eight orders, 12 families, 84 genera, and more than 1,000 species <ns0:ref type='bibr' target='#b62'>(Wang &amp; Ran, 2014)</ns0:ref>. Gymnosperms include ginkgo, cycads, conifers, and gnetophytes, which are grown in forests as important timber species and they provide raw materials for human usage, such as fiber, resin, and tannin <ns0:ref type='bibr' target='#b2'>(Christenhusz et al., 2010)</ns0:ref>. In addition, gymnosperms include some important threatened plants, where 40% are at high risk of extinction <ns0:ref type='bibr' target='#b13'>(Forest et al., 2018)</ns0:ref>. Recent phylogenetic and evolutionary studies of gymnosperms have demonstrated the rapid evolution of mitochondrial (mt) genes and provided further evidence of sister relationship between conifers and Gnetales <ns0:ref type='bibr' target='#b50'>(Ran, Gao &amp; Wang, 2010)</ns0:ref>. The high levels of genetic diversity and population differentiation among the Pinus species in gymnosperms have been studied based on plastid DNA markers <ns0:ref type='bibr' target='#b39'>(Liu et al., 2014)</ns0:ref>. Other studies have indicated patterns related to the physiological ecology, phylogenetic relationships, and population genetic structure of gymnosperm species <ns0:ref type='bibr'>(Yu et al., 2014;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b8'>Dong et al., 2016)</ns0:ref>. However, these studies mainly considered the phylogeny and evolution at the whole populations level. Thus, the detailed evolutionary characteristics of gymnosperms still need to be elucidated.</ns0:p><ns0:p>Chloroplasts are the site of photosynthesis and of various essential metabolic pathways, e.g., fatty acid and amino acid biosynthesis and the assimilation of nitrogen, sulfur, and selenium <ns0:ref type='bibr'>(Wise &amp; Hoober, 2006;</ns0:ref><ns0:ref type='bibr'>Des Marais,2000;</ns0:ref><ns0:ref type='bibr' target='#b29'>Knorr &amp; Heimann, 2001;</ns0:ref><ns0:ref type='bibr' target='#b47'>Pilon-Smits et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b16'>Guo et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kretschmer, Croll &amp; Kronstad, 2017)</ns0:ref>. It is generally recognized that chloroplasts are derived from proto-eukaryotic symbiotic cyanobacteria that internalized in eukaryotic cells <ns0:ref type='bibr' target='#b20'>(Hiroki &amp; Daisuke, 2018</ns0:ref>) and evolved into central organelles. Chloroplasts have their own genome encoding about 100 proteins and they are maternally inherited organelles in most angiosperm plants <ns0:ref type='bibr' target='#b0'>(Abdallah, Salamini &amp; Leister, 2000;</ns0:ref><ns0:ref type='bibr' target='#b19'>Heuertz et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b3'>Civan et al., 2014)</ns0:ref>. Among gymnosperms, paternal plastid inheritance is the typical characteristic of conifers <ns0:ref type='bibr'>(Faur&#233; et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kaundun &amp; Matsumoto, 2011)</ns0:ref>. Studies have shown that the chloroplast genome is quite conserved with an average evolutionary rate of 0.2-1.0&#61620;10 -9 per site per year, which is only one-fifth of that for the nuclear genome <ns0:ref type='bibr' target='#b9'>(Drouin, Daoud &amp; Xia, 2008;</ns0:ref><ns0:ref type='bibr' target='#b10'>Duchene &amp; Bromham, 2013)</ns0:ref>. The chloroplast genome is a covalently closed circular structure with four parts comprising the large single copy region (LSC), small single copy region (SSC), inverted repeat region A (IRa), and inverted repeat region B (IRb). The two IRs have the same sequences but in the opposite direction <ns0:ref type='bibr' target='#b61'>(Wang et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b40'>Logacheva et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b18'>Hereward et al., 2018)</ns0:ref>. Due to the independent evolution of the chloroplast genome, it is possible to construct a molecular phylogenetic tree using the chloroplast genome and without requiring any other data. Data analysis based on the conserved evolution of plastids is highly valuable for phylogenetic studies <ns0:ref type='bibr' target='#b27'>(Kim &amp; Suh, 2013)</ns0:ref> because it can provide reliable and useful phylogenetic information. The Manuscript to be reviewed relative completeness and independence of the chloroplast genome means that it can provide valuable material for research purposes.</ns0:p><ns0:p>Transfer RNAs (tRNAs) undergo numerous post-transcriptional nucleotide modifications and they exhibit abundant chemical diversity where the bases experience methylation, formylation, and other modifications <ns0:ref type='bibr' target='#b57'>(Suzuki &amp; Suzuki, 2014)</ns0:ref>. Chemical nucleotide modifications are frequent in tRNAs and they are important for the structure, stability, correct folding, aminoacylation, and decoding. For example, a previous analysis of the chemically synthesized f 5 C34-modified anticodon loop of human mt-tRNA Met showed that f 5 C34 contributes to the anticodon domain structure of the mt-tRNA <ns0:ref type='bibr' target='#b42'>(Lusic et al., 2008)</ns0:ref>. tRNAs comprise sequences of less than 100 polynucleotides that fold into a clover-type secondary structure and then into an L-shaped tertiary structure <ns0:ref type='bibr' target='#b64'>(Wilusz, 2015)</ns0:ref>. The secondary structure of tRNAs comprises different arms as well as loops, i.e., the D-arm, acceptor arm, anticodon arm, pseudouridine arm (&#936;-arm), D-loop, variable arm, anticodon loop, and pseudouridine loop (&#936;loop) <ns0:ref type='bibr' target='#b15'>(Gieg&#233;, Puglisi &amp; Florentz, 1993;</ns0:ref><ns0:ref type='bibr' target='#b43'>Mizutani &amp; Goto, 2000)</ns0:ref>. This unique structure allows tRNA to act as important bridges between the information level of nucleic acids and functional level of proteins. The vital components of tRNAs comprise an anti-codon region that discerns the messenger RNA carried by the specific codons, a 3&#61602;-CCA tail for attaching to the cognate amino acid, the &#936;-arm, and a &#936;-loop that has a relationship with the ribosome machinery <ns0:ref type='bibr' target='#b28'>(Kirchner &amp; Ignatova, 2014)</ns0:ref>. Asymmetric combinations and the divided segments in tRNA genes allow us to understand the diversity of tRNA molecules. tRNA species fulfill various functions in cellular homeostasis, regulation of gene expression and epigenetics, biogenesis, and even biological disease <ns0:ref type='bibr' target='#b51'>(Ribasd &amp; Dedon, 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Kanai, 2015;</ns0:ref><ns0:ref type='bibr' target='#b54'>Schimmel, 2017)</ns0:ref>. The evolutionary relationships determined between cyanobacteria and monocots show that tRNAs evolved polyphyletically and they originated from multiple common ancestors with a high rate of gene loss <ns0:ref type='bibr'>(Mohanta et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mohanta et al., 2019)</ns0:ref>. Nevertheless, the basic details of the tRNAs in plant chloroplasts still need to be elucidated and on the diverse evolutionary features of gymnosperm tRNAs are still unclear.</ns0:p><ns0:p>In this study, we assessed all of the chloroplast genomes in 12 families of gymnosperms from eight orders. The main aims of this study were as follows: (1) to determine the diversification of nucleotides in the secondary structure of gymnosperm tRNAs; (2) to identify the detailed genomic features of chloroplast tRNAs; (3) to assess the evolutionary relationships among different chloroplast tRNAs; and (4) to evaluate the duplication or loss events that occurred in all of the tRNAs considered. Our findings provide important insights into the biological characteristics and evolutionary variation of the tRNA family.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>Annotation and identification of chloroplast tRNA sequences in gymnosperms We downloaded complete chloroplast genomes for 12 representative gymnosperms in eight orders from the National Center for Biotechnology Information database (NCBI, https://www.ncbi.nlm.nih.gov/). The gymnosperm species investigated were: Cycas debaoensis <ns0:ref type='formula'>NC_011954</ns0:ref>). The gymnosperm tRNA genomes were annotated using GeSeq-Annotation of Organellar Genomes tool <ns0:ref type='bibr' target='#b58'>(Tillich et al., 2017)</ns0:ref> where the parameters were set as: circular sequence(s), chloroplast of sequence source, generate multi FASTA; BLAST protein search identity 25% for annotating plastid IR, 85% identity for BLAST rRNA, tRNA and DNA search, Embryophyta chloroplast (CDS+rRNA), third party tRNA annotator ARAGORN v1.2.38, ARWEN v1.2.3, tRNAScan-SE v2.0, and without Refseq choice.</ns0:p><ns0:p>Structural analysis of chloroplast tRNAs ARAGORN <ns0:ref type='bibr' target='#b36'>(Laslett &amp; Canback, 2004)</ns0:ref> and tRNAScan-SE software <ns0:ref type='bibr' target='#b41'>(Lowe &amp; Eddy, 1997)</ns0:ref> were employed to analyze the sequences and the secondary structure of tRNAs in the chloroplast genomes of the involved gymnosperm plants. The default parameters were set in ARAGORN software. The parameters for tRNAScan-SE were set as: sequence source, bacterial; search mode, default; query sequences, formatted (FASTA); and genetic code for tRNA isotype prediction, universal. Phylogenetic tree construction A phylogenetic tree was constructed for all of the tRNAs using MEGA7.0 software <ns0:ref type='bibr' target='#b32'>(Kumar et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b33'>Kumar, Stecher &amp; Tamura, 2016)</ns0:ref>. To study the evolutionary details of chloroplast tRNAs in gymnosperm species, an alignment file for tRNAs was achieved by CLUSTAL Omega software before the phylogenetic tree was constructed. MEGA7 software was used to transform the alignment file into MEGA format. The phylogenetic tree was constructed with the following parameters: phylogeny reconstruction of analysis, maximum likelihood model, bootstrap method in phylogeny test, 1000 bootstrap replicates, nucleotides type, gamma distributed with invariant sites (G+I) model, five discrete gamma categories, partial deletion for gaps/missing data treatment, 95 % site coverage cut-off, and very strong for branch swap filter.</ns0:p><ns0:p>Transition/transversion analysis The sequences of the tRNA isotypes were aligned to determine the transition and transversion rates for chloroplast tRNAs in gymnosperm plants. The files covering all 20 types of tRNAs were transformed into the MEGA file format and analyzed separately using MEGA7.0 software <ns0:ref type='bibr' target='#b35'>(Kumar, Tamura &amp; Nei, 1994)</ns0:ref>. The transition and transversion rates were analyzed for tRNAs with the following parameters: substitution pattern estimation (ML) analysis, automatic (neighbor-joining tree), maximum likelihood statistical method, nucleotide substitution type, Kimura two-parameter model, gamma distributed (G) site rates, five discrete gamma categories, partial deletion of gaps/missing data treatment, 95% of site coverage cut-off, and very strong branch swap filter.</ns0:p><ns0:p>Loss and duplication events analysis for tRNA genes In order to investigate the duplication or loss events in tRNA genes, the NCBI taxonomy browser was utilized to construct the whole species tree for the 12 gymnosperm species considered. The phylogenetic tree conducted in the evolutionary study was employed as gene tree. The gene tree for the tRNAs and species tree for the gymnosperm species were submitted to Notung 2.9 software <ns0:ref type='bibr' target='#b1'>(Chen, Durand &amp; Farach-Colton, 2000)</ns0:ref>, and then reconciled to discover duplicated and lost tRNA genes in the chloroplast genomes of gymnosperms.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Genomic features of gymnosperm chloroplast tRNAs Sequences were analyzed to identify the genomic tRNAs in the chloroplast genomes of 12 gymnosperm species comprising C. <ns0:ref type='bibr'>debaoensis, D. spinulosum, G. biloba, C. deodara, W. nobilis, R. piresii, S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon, and E. equisetina</ns0:ref>, which were obtained from the NCBI database (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). The results showed that the length of the chloroplast tRNAs vary from the smallest with 64 nucleotides (nt) (tRNA Met -CAU in T. mairei) to the largest with 96 nt (tRNA Tyr -AUA in W. nobilis, C. deodara, and G. biloba) (Data S1). We found that the chloroplast genomes of gymnosperm plants encode 28 to 33 tRNAs (Table <ns0:ref type='table'>2</ns0:ref>), where D. spinulosum, C. deodara, and S. verticillata encode 31 anticodons, W. nobilis, R. piresii, C. lanceolata, and G. gnemon encode 32 tRNA isotypes, G. biloba, and W. mirabilis encode 33 tRNAs. Other species comprising T. mairei, E. equisetina and C. debaoensis encode 28, 28, 30 tRNA isotypes, respectively (Table <ns0:ref type='table'>2</ns0:ref>). tRNA Ala was not found in R. piresii and T. mairei, and tRNA Val was not detected in T. mairei (Fig. <ns0:ref type='figure' target='#fig_10'>S3</ns0:ref>). We also observed that all of the species do not encode selenocysteine and its suppressor tRNA (Table <ns0:ref type='table'>2</ns0:ref>). Overall, tRNA Ser (in W. nobilis) and tRNA Arg (in W. mirabilis) are the most abundant (four types) followed by tRNA Leu (three types) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>Variations in structures of chloroplast tRNAs Some tRNAs with a loop structure in the variable region were found to be encoded in the gymnosperm chloroplast genomes (Fig. <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>). A novel tRNA lacking the D-arm was found in tRNA Gly in W. nobilis (Fig. <ns0:ref type='figure' target='#fig_10'>3</ns0:ref>). As shown in Fig. <ns0:ref type='figure' target='#fig_3'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>, tRNA Leu , tRNA Ser , and tRNA Tyr contain expanded variable stem/loops. In these tRNAs (except for tRNA Ser -GCU of D. spinulosum), the anticodon loop of tRNA Ser contains the conserved consensus sequence N-U-N-G-A-A-N, and tRNAs Leu have the consensus sequence C-U-N-A-N 2 -A. The variable loop region is predicted to fold into stem-loop structures with apical loops of 3 to 7 nt in tRNA Ser and several tRNA Leu variants. The stems contain up to 7 bp (Fig. <ns0:ref type='figure' target='#fig_9'>1 and 2</ns0:ref>). The expanded variable loop structures may play important functions during the protein translation process in chloroplasts.</ns0:p></ns0:div> <ns0:div><ns0:head>Chloroplast genomes contain 25 to 30 anticodon-specific tRNAs</ns0:head><ns0:p>The genomes of the species analyzed were found to code for at least two copies of tRNA Met -CAU/tRNA fMet -CAU. Each of the gymnosperm chloroplast genomes encodes 25 to 30 anticodon-specific tRNAs (Table <ns0:ref type='table' target='#tab_3'>2 and Table 3)</ns0:ref>, where E. equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S. verticillata, and C. lanceolata encode 28 anticodons, and D. spinulosum, C. deodara, W. mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis, G. gnemon and G. biloba encodes 30 anticodons (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>). tRNA Arg -CCG was present in the genomes of nine gymnosperm species but absent from C. lanceolata, T. mairei, and E. equisetina, while tRNA Gly -UCC was lacking from C. debaoensis, S. verticillata, D. spinulosum, C. lanceolata, T. mairei, and E. equisetina (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>). The most abundant anticodons found in the chloroplast genomes were tRNA Gly -GCC, tRNA Pro -UGG, tRNA Ser -UGA, tRNA Ser -GCU, tRNA Arg -ACG, tRNA Arg -UCU, tRNA Leu -UAG, tRNA Leu -CAA, tRNA Phe -GAA, tRNA Asn -GUU, tRNA Lys -UUU, tRNA Asp -GUC, tRNA Glu -UUC, tRNA His -GUG, tRNA Gln -UUG, tRNA Ile -CAU, tRNA Met -CAU, tRNA Tyr -GUA, tRNA Cys -GCA, and tRNA Trp -CCA (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>). Two tRNA Trp iso-acceptors are present in E. equisetina chloroplasts, compared with a single one in the other gymnosperm species analyzed in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved gymnosperm chloroplast tRNAs</ns0:head><ns0:p>The clover leaf-like secondary structure of a tRNA is shown in Fig. <ns0:ref type='figure'>4</ns0:ref>. In the study, we found that most tRNAs contain a 'G' as the first nucleotide in the D-arm, except for tRNA Lys , tRNA Met , tRNA Pro , tRNA Thr , tRNA Tyr , and tRNA Val . 'A' is present in the first and the last position of the D-loop apart from tRNA Gly , tRNA Ile , tRNA Leu , tRNA Met , and tRNA Gln . In addition, in the final two positions of the &#936;-arm, all of the tRNAs were found to have conserved 'G-G' nucleotides, except for tRNA Arg , tRNA Cys , tRNA Phe , and tRNA Val (Table <ns0:ref type='table'>4</ns0:ref>). Small conserved consensus sequences were found in the &#936; region. To be specific, except for tRNA Ser , the &#936;-loop in tRNAs was found to contain a conserved sequence comprising U-U-C-N-A-N 2 according to a multiple sequence alignment of 20 members of the tRNA gene family (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Diversification of tRNAs structures</ns0:head><ns0:p>The diverse arms and loops of tRNAs allow the regulation and control of protein translation. Each arm and loop has a specific nucleotide composition. Our analysis based on 373 tRNAs showed that the acceptor arm of chloroplast tRNAs contains 6 bp to 7 bp (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The D-arms were found to contain 3 or 4 bp generally, with a stable 'G' in the initial position and 'C' in the last position of the D-stem 5' strand in most tRNAs (such as tRNA Ala , tRNA Asn , tRNA Asp , tRNA Cys , tRNA Glu , tRNA His , tRNA Ile , and tRNA Phe ). Most D-loops usually contain 7 to 11 nt with conserved 'A' nucleotides at the two end locations. The anticodon arms of chloroplast tRNAs mainly contain 5 bp (90.4%). We found that 367 (about 99%) tRNAs contain 7 nt in their anticodon loop, thereby indicating that the sequence of the anticodon loop is highly conserved (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The variable loops of different tRNAs contain 3 to 23nt, where those in tRNA Ala , tRNA Asp , tRNA His , tRNA Phe , and tRNA Pro contain 5 bp (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The &#936;-arm contains 5 bp in most of the gymnosperm chloroplast tRNAs, except for tRNA Ala and some of the tRNA Trp , tRNA Gly , tRNA Thr , and tRNA Arg in chloroplast. The &#936;-loops of most tRNAs contain 7 nt, apart from tRNA Ala and several of tRNA Cys and tRNA Thr (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gymnosperm chloroplast tRNAs derived from multiple common ancestors</ns0:head><ns0:p>The phylogenetic tree demonstrated the presence of three major clusters covering 64 groups and the different types of all tRNAs (as shown by the different strings in Fig. <ns0:ref type='figure' target='#fig_4'>S1</ns0:ref>). We detected 37 groups in cluster I, five in cluster II, and 22 groups in cluster III. Cluster I contains tRNA Ser , tRNA Tyr , tRNA His , tRNA Gln , tRNA Thr , tRNA Pro , tRNA Gly , tRNA Met , tRNA Asp , tRNA Arg , tRNA Ala , tRNA Cys , tRNA Lys , tRNA Glu , tRNA Ile , tRNA Asn , tRNA Val , tRNA Leu , and tRNA Trp . Cluster II contains tRNA His , tRNA Ser , tRNA Tyr , and tRNA Leu . Cluster III contains tRNA Leu , tRNA Ile , tRNA Gly , tRNA Thr , tRNA Ser , tRNA Val , tRNA Glu , tRNA Lys , tRNA Cys , tRNA Gln , tRNA His , tRNA Arg , tRNA Phe , tRNA Ala , and tRNA Met (Fig. <ns0:ref type='figure' target='#fig_4'>S1</ns0:ref>). tRNA Ser , tRNA His , and tRNA Leu are present in cluster I but also in cluster II and cluster III, thereby suggesting that these tRNAs evolved from multiple lineages. Most of the tRNAs were found to form more than one group in the phylogenetic tree. In cluster I, the tRNAs that formed two groups in the phylogenetic tree were identified as tRNA Tyr , tRNA Gln , tRNA Met , tRNA Asp , tRNA Ala , tRNA Lys , tRNA Ile , and tRNA Trp , whereas those that clustered to form three groups were determined as tRNA Ser , tRNA Pro , tRNA Arg , tRNA Glu , tRNA Asn , tRNA Val , and tRNA Leu . Moreover, tRNA Thr clustered into four groups. In cluster II, tRNA Ser was found to form two groups. In cluster III, tRNA Gly and tRNA Val were found to form two groups, whereas tRNA Thr formed three groups, tRNA Ile formed four groups. Some tRNAs in cluster III were found to group individually, where these tRNAs containing the anticodons C-G-A in tRNA Ser , U-U-C in tRNA Glu , U-U-U in tRNA Lys , G-C-A in tRNA Cys , U-U-G in tRNA Gln , G-U-G in tRNA His , U-C-U in tRNA Arg , G-A-A in tRNA Phe , U-G-C in tRNA Ala , and C-A-U in tRNA Met all grouped separately (Fig. <ns0:ref type='figure' target='#fig_4'>S1</ns0:ref>). The multiple groupings of different tRNAs suggest that they evolved from multiple common ancestors. Furthermore, the tRNAs presented in cluster III, i.e., tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA His (GUG), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Ile (UAU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Gly (UCC), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU), tended to be the most basic tRNAs and they had undergone gene duplication and diversification to generate other tRNA molecules.</ns0:p></ns0:div> <ns0:div><ns0:head>C-A-U anticodon in tRNA Ile</ns0:head><ns0:p>Our detailed genomic study showed that tRNA Ile also encodes a C-A-U anticodon in addition to the presence of this typical anticodon in tRNA Met . In general, the C-A-U anticodon is recognized as a typical characteristic of tRNA Met and there is only one iso-acceptor. In particular, we found that the tRNA Ile in T. mairei encodes two C-A-U anticodons, and C. debaoensis, <ns0:ref type='bibr'>S. verticillata, D. spinulosum, C. lanceolata, G. biloba, C. deodara, W. mirabilis, G. gnemon, R. piresii, E. equisetina, and W.</ns0:ref> nobilis also encode a C-A-U anticodon (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>, Data S1, Fig. <ns0:ref type='figure' target='#fig_10'>S3</ns0:ref>).</ns0:p><ns0:p>Transition/transversion of tRNAs A previous study <ns0:ref type='bibr' target='#b45'>(Mohanta et al., 2019)</ns0:ref> showed that the evolutionary rates are almost equal for tRNAs with respect to transition and transversion despite the low probability of transition or transversion events in tRNAs. In this study, we identified several intriguing substitutions of gymnosperm chloroplast tRNAs. Overall, our analysis of the substitution rates detected using the whole set of chloroplast tRNAs showed that average transition rate (15.38) was significantly larger than the average transversion rate (4.81) with a ratio of 3:1 (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). The same transition: transversion ratio bias was found in all the set of tRNAs for tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , and tRNA Leu . The ratio was over 6:1 for tRNA Cys and tRNA Arg . The transition rates for tRNA Trp , tRNA Val , and tRNA Gly were about 10 times higher than their transversion rates. These findings suggest that tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , tRNA Leu , tRNA Cys , tRNA Arg , tRNA Trp , tRNA Val and tRNA Gly underwent transition substitutions more readily than transversion substitutions during their evolution in gymnosperm chloroplast genomes. In addition, the transition rates in tRNA Lys and tRNA Pro were about 15 times higher than their transversion rates. The transition rates in tRNA Asn , tRNA Phe , and tRNA His were about 20 times higher than their transversion rates. These results indicate that tRNAs are much more likely to have undergone transition events rather than transversion events. The highest transversion rate of 12.50 was found in tRNA Ala and the lowest transversion rate of 0.00 in tRNA Asp (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). Correspondingly, tRNA Ala lacks any transitions (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>).</ns0:p><ns0:p>tRNA duplication/loss events In addition to transition and transversion events, gene duplication and loss events have played important roles in gene evolution. Our analysis of duplication and loss events indicated that 153 duplication events (duplication and conditional duplication) have occurred in all of the gymnosperm chloroplast tRNA genes investigated in this study (Fig. <ns0:ref type='figure' target='#fig_5'>S2</ns0:ref>). In addition, 220 gymnosperm chloroplast tRNA gene loss events were detected (Table <ns0:ref type='table'>S2</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_5'>S2</ns0:ref>). Thus, the loss of genes was slightly more frequent than their duplication for gymnosperm chloroplast tRNA genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>tRNAs are major genetic components of semi-autonomous chloroplasts and our analysis of gymnosperm chloroplast genomes showed that they have several basic conserved genomic features. The gymnosperm chloroplast genomes investigated in the present study were found to encode 28 to 33 tRNA isotypes, thereby indicating that there is substantial variation in the quantity of tRNAs in gymnosperm chloroplast genomes. The lack of tRNA Ala in R. piresii and T. mairei, and the absence of tRNA Val in T. mairei were interesting. Thus, it is necessary to understand how the translation process is conducted in chloroplasts without these crucial tRNAs. According to previous studies <ns0:ref type='bibr' target='#b60'>(Treangen &amp; Rocha, 2011;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mohanta et al., 2019)</ns0:ref>, it is likely that the deficiency of these tRNAs is compensated for by the tRNAs transferred from other organelle genomes such as nuclear genome and mitochondrial genome. In addition to the absence of tRNA Ala and tRNA Val , all of the gymnosperm plants were shown to not encode selenocysteine tRNA and its suppressor tRNA in their chloroplast genomes (Table <ns0:ref type='table'>2</ns0:ref>). Selenocysteine tRNA and its suppressor tRNA were also not detected in the chloroplast of Oryza sativa <ns0:ref type='bibr' target='#b44'>(Mohanta &amp; Bae, 2017)</ns0:ref>.</ns0:p><ns0:p>In addition to the presence of C-A-U anticodon in tRNA Met , we found that tRNA-CAU is present in tRNA Ile (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>). Similarly, the C-A-U anticodon was detected in tRNA Ile in Bacillus subtilis (Ehrenberg) Cohn and spinach <ns0:ref type='bibr' target='#b25'>(Kashdan &amp; Dudock, 1982;</ns0:ref><ns0:ref type='bibr' target='#b30'>K&#246;hrer et al., 2014)</ns0:ref>. The possible mechanism that governs the specificity of this amino acid may involve modification of the wobble position in the anticodon by a tRNA-modifying enzyme. Chloroplasts originate from bacteria so the tRNA modifications found in bacteria may also occur in chloroplast tRNAs. In bacteria, the tRNA-modifying enzyme TilS can convert the 5&#61602;-C residue in the CAU anticodon of specific tRNA Ile molecules into lysidine to decode 5&#61602;-AUA (Ile) codons instead of 5&#61602;-AUG (Met) codons <ns0:ref type='bibr' target='#b55'>(Soma et al., 2003)</ns0:ref>. In addition, when lysidine decodes isoleucine, the tautomer form of lysidine provides compatible hydrogen bond donor-acceptor sites to allow base pairing with 'A' and this may help to the recognition of the codon AUA instead of AUG <ns0:ref type='bibr' target='#b56'>(Sonawane &amp; Tewari, 2008;</ns0:ref><ns0:ref type='bibr' target='#b53'>Sambhare et al., 2014)</ns0:ref>. The absence of tRNA Ile -lysidine synthetase leads to a failure to modify C34 to lysidine in tRNA Ile (LAU) (i.e., the synthesis of CAU-tRNA Ile ) and this inactivates the translation of AUA codons <ns0:ref type='bibr' target='#b30'>(K&#246;hrer et al., 2014)</ns0:ref>.</ns0:p><ns0:p>During protein coding, a certain species or gene tends to use one or more specific synonym codons, which is referred to as codon usage bias <ns0:ref type='bibr'>(Comeron &amp; Aguad&#233;, 1998;</ns0:ref><ns0:ref type='bibr' target='#b52'>Rota-Stabelli et al., 2012)</ns0:ref>. In the present study, tRNA Arg -CCG was found to be present in the genomes of nine species but absent from C. lanceolata, T. mairei, and E. equisetina. Similarly, tRNA Gly -UCC was shown to be absent from the chloroplast genomes of C. debaoensis, S. verticillata, D. spinulosum, C. lanceolata, T. mairei, and E. equisetina (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>). These results suggest that gymnosperm chloroplast tRNA genes are characterized by codon usage bias <ns0:ref type='bibr' target='#b63'>(Wei &amp; Jin, 2017;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In general, the secondary structure of tRNAs is characterized as clover leaf-like, except for a few tRNAs with unusual secondary structures <ns0:ref type='bibr' target='#b23'>(J&#252;hling et al., 2018)</ns0:ref>. In our study, we identified clover leaf-like tRNAs with expanded variable loop regions (Fig. <ns0:ref type='figure' target='#fig_3'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>). Numerous tRNA Leu , tRNA Ser , and tRNA Tyr were found to have specific variable loop configurations in terms of length and structure, suggesting significant structural variation among chloroplast tRNAs. It is interesting to note that there were also stem-loop structures in variable regions of certain tRNAs in cyanobacteria. This might indicate that similar structural variations exist between chloroplast tRNAs and cyanobacterial tRNAs <ns0:ref type='bibr'>(Mohanta et al., 2017)</ns0:ref>. Future studies will have to determine the biological importance of these variant tRNAs. The novel tRNA structure lacking the D arm might play some other significative functions in the translation progress and additional research is necessary to elucidate its exact function and mechanisms. Most tRNAs have a clover-like structure formed by complementary base pairing between small segments <ns0:ref type='bibr' target='#b22'>(Hubert et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b12'>Florentz, 2002)</ns0:ref>. Previous studies have showed that the acceptor arm of tRNAs in chloroplasts contain 7 bp to 9 bp, the D-arm contains 3 bp to 4 bp, the D-loop has 4 nt to 12 nt, the anticodon arm has 5 bp, the anticodon loop contains 7 nt, the variable region comprises 4 nt to 23 nt, and &#936;-arm contains 5 bp, and the &#936;-loop has 7 nt <ns0:ref type='bibr' target='#b64'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b46'>Mohanta &amp; Hanhong, 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mohanta et al., 2019)</ns0:ref>. In the present study, we found that the acceptor arm of chloroplast tRNAs contains 6 bp to 7 bp in 373 tRNAs, where the D-arm has 3 bp or 4 bp and the D-loop usually contains 7 nt to 11 nt. The anticodon loop of gymnosperm chloroplast tRNAs generally contains 7 nt, and thus the sequence of the anticodon loop is typically conserved (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The variable loop of different tRNAs contain 3 nt to 23 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). The &#936;-arm of gymnosperm chloroplast tRNAs generally contains 5 bp and the &#936;loop has 7 nt (Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref>). Our results are consistent with previous findings <ns0:ref type='bibr' target='#b64'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b46'>Mohanta &amp; Hanhong, 2017)</ns0:ref> and they suggest that chloroplast RNAs are significantly conserved. The consensus sequence 'U-U-C-N-A-N 2 ' was found in the &#936; region (Table <ns0:ref type='table'>4</ns0:ref>). Previous studies also reported the existence of a similar sequence in the &#936;-loop of tRNAs in Oryza sariva and Cyanobacteria <ns0:ref type='bibr' target='#b44'>(Mohanta &amp; Bae, 2017;</ns0:ref><ns0:ref type='bibr'>Mohanta et al., 2017)</ns0:ref>. This suggests that the consensus 'U-U-C-N-A-N 2 ' motif of the &#936; region, identified here and in previous analyses, is a general consensus motif of canonical tRNAs.</ns0:p><ns0:p>Our phylogenetic analysis detected three clear clusters and many tRNA groups. Some tRNAs (tRNA Ser , tRNA His , and tRNA Leu ) in cluster I and cluster II were also in cluster III, thereby indicating that these tRNAs evolved from multiple lineages by gene duplication and gene divergence. Moreover, anticodon types comprising CGA, UUC, UUU, GCA, UUG, GUG, UCU, UGC, and CAU appeared several times in the phylogenetic tree, and thus the corresponding tRNAs evolved from multiple common ancestors. The overlapping of tRNAs groups demonstrates that these tRNAs might have diverse common ancestors in the evolutionary process <ns0:ref type='bibr' target='#b44'>(Mohanta &amp; Bae, 2017)</ns0:ref>. Phylogenetic analysis also showed that tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA His (GUG), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Ile (UAU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Gly (UCC), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU) in cluster III tended to be the most basic tRNAs, whereas tRNA Met tended to be the most original tRNA. Overall, the results clearly indicate that the tRNAs encoded in gymnosperm chloroplast genomes have multiple common evolutionary ancestors.</ns0:p><ns0:p>Our results also provided insights into the gene substitution rates in gymnosperm chloroplast tRNAs. Overall, the average transition rate for tRNAs was greater than the transversion rate, where the relationship was about 3:1 (Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>). In all of the chloroplast tRNAs, the average transition rate was slightly higher than the average transversion rate, thereby indicating that chloroplast tRNAs have unequal substitution rates.</ns0:p><ns0:p>In addition to the transition and transversion events in tRNAs, loss and duplication events have played significant roles in the evolution of tRNAs in gymnosperm chloroplast genomes <ns0:ref type='bibr' target='#b17'>(He &amp; Zhang, 2006;</ns0:ref><ns0:ref type='bibr'>Magadum et al., 2013)</ns0:ref>. In general, the gene loss events tended to occur after whole genome duplication events. We found 153 duplication events and 220 loss events in gymnosperm chloroplast tRNAs, and thus loss events have occurred slightly more frequently than duplication events (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our basic structure analysis showed that gymnosperm chloroplast genomes encode 25 to 30 anticodon-specific tRNAs. The acceptor arm of chloroplast tRNA contains 6 bp to 7 bp, the Darm has 3 bp or 4 bp, the D-loop contains 7 nt to 11 nt mainly, and the anticodon loop usually contains 7 nt. In different tRNAs, the variable loop contains 3 nt to 23 nt. The &#936;-arm contains a conserved sequence comprising U-U-C-N-A-N 2 . tRNA Ala was absent from R. piresii and T. mairei, and tRNA Val was lacking in T. mairei. Gymnosperm chloroplasts do not encode selenocysteine tRNA and its suppressor tRNA in their genomes. A CAU anticodon is encoded in tRNA Met as well as in tRNA Ile . A novel tRNA structure lacking the D arm was identified for the chloroplast tRNA Gly of W. nobilis. Numerous tRNA Leu , tRNA Ser , and tRNA Tyr types were found to have expanded variable regions. Phylogenetic analysis showed that tRNAs might have multiple common ancestors in the evolutionary process. Different tRNAs harbored their own transition/transversion rates, i.e., it was iso-acceptor specific. And the transition rate was generally higher than the transversion rate. Furthermore, gene loss events (220) have occurred slightly more frequently than gene duplication events (153) in gymnosperm chloroplast tRNAs. Our results provide new insights into the evolution of gymnosperm chloroplast tRNAs and their diverse roles. <ns0:ref type='bibr'>. debaoensis, D. spinulosum, G. biloba, C. deodara, and R. piresii</ns0:ref> contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from C. debaoensis, D. spinulosum, G. biloba, C. deodara, R. piresii were observed to contain an expanded variable stem and loop. The anti-codon loop of tRNA Ser (except for tRNA Ser -GCU of D. spinulosum) was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and tRNA Leu harbor the consensus sequence C-U-N-A-N 2 -A. Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref> Certain tRNAs in S. verticillata, C. lanceolata, T. mairei, W. mirabilis, and G. gnemon contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon were observed to contain a variable stem and loop. The anti-codon loop of tRNA Ser was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and the consensus sequence was C-U-N-A-N 2 -A for tRNA Leu . Fig. <ns0:ref type='figure' target='#fig_10'>3</ns0:ref> An abnormal tRNA structure lacking the D-arm found in W. nobilis. The tRNA Gly with anti-codon UCC was found lacking D-arm. Fig. <ns0:ref type='figure'>4</ns0:ref> Clover leaf-like structure of gymnosperms tRNA. The tRNA contains the Acceptor arm (6-7 bp, green, &gt;95% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region (3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, green, &gt;95% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). Several tRNAs harbor the nucleotides of C-C-A tail.</ns0:p><ns0:note type='other'>Figure Legends</ns0:note></ns0:div> <ns0:div><ns0:head>Supplementary materials</ns0:head><ns0:p>Data S1 tRNA sequences of gymnosperms chloroplast genome conducted in the study. Table <ns0:ref type='table' target='#tab_0'>S1</ns0:ref> Nucleotide composition in different parts of clover-structure of chloroplast genome tRNA. AC-arm, Acceptor arm; ANC-arm, Anti-codon arm; ANC-loop, Anti-codon loop; &#936;-arm, Pseudouridine arm; &#936;-loop, Pseudouridine loop. The base pairs in the AC-arm were counted according to the predicted clover structures of tRNAs. Table <ns0:ref type='table'>S2</ns0:ref> Loss events of chloroplast genomic tRNAs. Manuscript to be reviewed A view of the gymnosperms in analysis.</ns0:p><ns0:p>Statistics of the 12 gymnosperms in the study.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:11:43190:3:0:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed 1 Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>. Distribution of anti-codons in the chloroplast genome of gymnosperms. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>Met/fMet Tyr Cys 1 1 2 1 1 1 1 1 1 1 1 Trp 1 1 1 1 1 1 1 1 1 1 1 Selenocystei ne 0 0 0 0 0 0 0 0 0 0 0 Suppressor 0 0 0 0 0 0 0 0 0 0 0</ns0:formula><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Clover leaf-like structure of gymnosperms tRNA.</ns0:p><ns0:p>The tRNA contains the Acceptor arm (6-7 bp, green, &gt;95% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved ), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region</ns0:p><ns0:p>(3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, green, &gt;95% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). Several tRNAs harbor the nucleotides of C-C-A tail.</ns0:p></ns0:div><ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Y. C. Zhong &amp; C. J. Chen (KM459003), Dioon spinulosum Dyer ex Eichler (NC_027512), Ginkgo biloba L. (NC_016986), Cedrus deodara (Roxb.) G. Don (NC_014575), Wollemia nobilis W. G. Jones, K. D. Hill &amp; J. M. Allen (NC_027235), Retrophyllum piresii Silba C. N. (KJ017081), Sciadopitys verticillata (Thunb.) Sieb. et Zucc. (NC_029734), Cunninghamia lanceolata (Lamb.) Hook. (NC_021437), Taxus mairei (Lemee et Levl.) Cheng et L. K. Fu (KJ123824), Welwitschia mirabilis Hook.f. (EU342371), Gnetum gnemon L. (KR476377), and Ephedra equisetina Bge. (</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Yu F, Wang DX, Yi XF, Shi XX, Huang YK, Zhang HW, Zhang XP. 2014. Does animalmediated seed dispersal facilitate the formation of Pinus armandii-quercus aliena var. acuteserrata forests? PloS one 9:e89886. DOI: 10.1371/journal.pone.0089886</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Fig. 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig.1Certain tRNAs in C. debaoensis, D. spinulosum, G. biloba, C. deodara, and R. piresii contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from C. debaoensis, D. spinulosum, G. biloba, C. deodara, R. piresii were observed to contain an expanded variable stem and loop. The anti-codon loop of tRNA Ser (except for tRNA Ser -GCU of D. spinulosum) was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and tRNA Leu harbor the consensus sequence C-U-N-A-N 2 -A. Fig.2Certain tRNAs in S. verticillata, C. lanceolata, T. mairei, W. mirabilis, and G. gnemon contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon were observed to contain a variable stem and loop. The anti-codon loop of tRNA Ser was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and the consensus sequence was C-U-N-A-N 2 -A for tRNALeu . Fig.3An abnormal tRNA structure lacking the D-arm found in W. nobilis. The tRNA Gly with anti-codon UCC was found lacking D-arm. Fig.4Clover leaf-like structure of gymnosperms tRNA. The tRNA contains the Acceptor arm (6-7 bp, green, &gt;95% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region (3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, green, &gt;95% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). Several tRNAs harbor the nucleotides of C-C-A tail.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Fig. S1</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>The phylogenetic tree constructed by 373 tRNAs. Multiple tRNAs are shown by different colors. Different groups are marked by different strings. The phylogenetic clades with low bootstrap replicates were collapsed with 50% cutoff values. Phylogenetic analysis illustrates that Gymnosperm chloroplast tRNA derived from common multiple ancestors.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Fig. S2</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Fig.S2The loss and duplication tree. Blue: Duplication events; Gray: Loss events; D: Duplication node; cD: Conditional Duplication node. Fig.S3tRNA gene content in analyzed gymnosperms chloroplast genome. The tRNA genes are shown in the left (top to bottom). Boxes in light green, dark green, and white represent one copy of tRNA genes, two copies of tRNA genes, and the absence of tRNA genes.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>1</ns0:head><ns0:label /><ns0:figDesc>Table 1. A view of the gymnosperms in analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Order</ns0:cell><ns0:cell>Family</ns0:cell><ns0:cell>Subfamily</ns0:cell><ns0:cell>Genus</ns0:cell><ns0:cell>Species</ns0:cell><ns0:cell>NCBI Locus</ns0:cell></ns0:row><ns0:row><ns0:cell>Cycadales</ns0:cell><ns0:cell>Cycadaceae</ns0:cell><ns0:cell /><ns0:cell>Cycas</ns0:cell><ns0:cell>debaoensis</ns0:cell><ns0:cell>KM459003</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Zamiaceae</ns0:cell><ns0:cell>Diooideae</ns0:cell><ns0:cell>Dioon</ns0:cell><ns0:cell>spinulosum</ns0:cell><ns0:cell>NC_027512</ns0:cell></ns0:row><ns0:row><ns0:cell>Ginkgoales</ns0:cell><ns0:cell>Ginkgoaceae</ns0:cell><ns0:cell /><ns0:cell>Ginkgo</ns0:cell><ns0:cell>biloba</ns0:cell><ns0:cell>NC_016986</ns0:cell></ns0:row><ns0:row><ns0:cell>Pinales</ns0:cell><ns0:cell>Pinaceae</ns0:cell><ns0:cell>Abieteae</ns0:cell><ns0:cell>Cedrus</ns0:cell><ns0:cell>deodara</ns0:cell><ns0:cell>NC_014575</ns0:cell></ns0:row><ns0:row><ns0:cell>Araucariales</ns0:cell><ns0:cell>Araucariaceae</ns0:cell><ns0:cell /><ns0:cell>Wollemia</ns0:cell><ns0:cell>nobilis</ns0:cell><ns0:cell>NC_027235</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Podocarpaceae</ns0:cell><ns0:cell /><ns0:cell>Retrophyllum</ns0:cell><ns0:cell>piresii</ns0:cell><ns0:cell>KJ017081</ns0:cell></ns0:row><ns0:row><ns0:cell>Cupressales</ns0:cell><ns0:cell>Sciadopityaceae</ns0:cell><ns0:cell /><ns0:cell>Sciadopitys</ns0:cell><ns0:cell>verticillata</ns0:cell><ns0:cell>NC_029734</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Cupressaceae</ns0:cell><ns0:cell>Cunninghami a</ns0:cell><ns0:cell>Cunninghamia</ns0:cell><ns0:cell>lanceolata</ns0:cell><ns0:cell>NC_021437</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Taxaceae</ns0:cell><ns0:cell /><ns0:cell>Taxus</ns0:cell><ns0:cell>mairei</ns0:cell><ns0:cell>KJ123824</ns0:cell></ns0:row><ns0:row><ns0:cell>Welwitschiales</ns0:cell><ns0:cell>Welwitschiaceae</ns0:cell><ns0:cell /><ns0:cell>Welwitschia</ns0:cell><ns0:cell>mirabilis</ns0:cell><ns0:cell>EU342371</ns0:cell></ns0:row><ns0:row><ns0:cell>Gnetales</ns0:cell><ns0:cell>Gnetaceae</ns0:cell><ns0:cell /><ns0:cell>Gnetum</ns0:cell><ns0:cell>gnemon</ns0:cell><ns0:cell>KR476377</ns0:cell></ns0:row><ns0:row><ns0:cell>Ephedrales</ns0:cell><ns0:cell>Ephedraceae</ns0:cell><ns0:cell /><ns0:cell>Ephedra</ns0:cell><ns0:cell>equisetina</ns0:cell><ns0:cell>NC_011954</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Distribution of anti-codons in the chloroplast genome of gymnosperms.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Each of the gymnosperm chloroplast genomes encodes 25 to 30 anticodon-specific tRNAs. E.</ns0:cell></ns0:row><ns0:row><ns0:cell>equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S.</ns0:cell></ns0:row><ns0:row><ns0:cell>verticillata, and C. lanceolata encode 28 anticodons, and D. spinulosum, C. deodara, W.</ns0:cell></ns0:row><ns0:row><ns0:cell>mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis, G.</ns0:cell></ns0:row><ns0:row><ns0:cell>gnemon and G. biloba encodes 30 anticodons.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:3:0:NEW 6 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Transition and transversion rate of chloroplast tRNA.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:3:0:NEW 6 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2019:11:43190:3:0:NEW 6 Sep 2020) Manuscript to be reviewed 1 PeerJ reviewing PDF | (2019:11:43190:3:0:NEW 6 Sep 2020)Manuscript to be reviewed</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43190:3:0:NEW 6 Sep 2020)</ns0:note> </ns0:body> "

"Dear Editor, Thank you very much for kindly returning our manuscript. We are very grateful to you for your time and patience, and the reviewers’ professional and thoughtful comments. These suggestions and opinions are all precious and very helpful for revising and improving our MS. In the revised manuscript, we have studied all these suggestions carefully and have made thoroughly corrections, and responded point by point to the comments as itemized below (our responses are in green characters). If you have any questions regarding the manuscript, please feel free to contact the corresponding author: Zhong-Hu Li Key Laboratory of Resource Biology and Biotechnology in Western China (Ministry of Education), College of Life Sciences, Northwest University, Xi’an 710069, China Fax: +86 29 888302411; E-mail: [email protected] We shall look forward to hearing from you at your earliest convenience. Yours sincerely, Zhonghu Li, Ph. D E-mail: [email protected] Reviewer comments: Reviewer 2 (Anonymous) 1. Basic reporting: The manuscript has improved substantially. Before publication, a few points should be addressed. Reply: Thank you very much for your positive and valuable comments. We have checked carefully and modified these points in the revised manuscript accordingly. 1) related to the second part of my previous comment d): in the new Fig. 2B, the variable stem- loop should be as predicted by RNAfold: Reply: Thank you very much for your kindly and constructive comments. In the revised manuscript, we have checked the Figure carefully and improved Fig. 2B by adopting optimal variable stem- loop structure as predicted by RNAfold. 2) related to my previous comment g): in line 220, replace tRNASer with tRNAGln Reply: Thank you very much for your kindly suggestion. In line 220, we have corrected “tRNASer ” to “tRNAGln ”. 3) related to my previous comment i), paragraph ' Diversification of tRNAs structures' (lines 227-242): - line 232, rewrite: '... with a stable “G” in the initial position and “C” in the last position of the D-stem 5' strand in most tRNAs ...' Reply: Thanks for your kindly comments. We have done it accordingly. - line 234, rewrite: ' Most D-loops usually contain 7 to 11 nt with conserved “A” nucleotides at the two end locations.' Reply: We have revised it accordingly. - line 236: simplify 90.35% and 99.01% to 90.4% and 99% Reply: We have revised it accordingly. - line 238: 3 to 23 nt ('bp' only for stems!) Reply: Thank you for your comments. We have revised it accordingly. • line 241: 'The Ψ-loops of most tRNAs contain 7 nt, apart ...' Reply: Many thanks for your kindly comments. We have revised it accordingly. Table S1, counting of base pairs in the AC stem: for W. nobilis tRNAGly, you now indicate 7 bp for the AC stem which includes 2 A-C pairs; however, for Sciadopitys verticillata_94443 tRNAThr (structure A below), you indicate 6 bp in Table S1, ignoring the A-C pair; A-C pairs can form with one hydrogen bond at neutral pH and two hydrogen bonds at slightly acidic pH (N.B. Leontis et al., 2002, Nucleic Acids Res. 30, 3497-3531); so the AC stem in structure A below has likely 7 bp, but the stem end is weak; I would prefer this perspective rather than counting only Watson-Crick and G-U wobble base pairs; in any case, your counting of AC stem base pairs should be unifiorm. I do not believe the AC stem length of structure B below; please add a few genome-encoded nucleotides to the 5'- and 3'-end and see if the AC stem can be extendend including AC or other Non-Watson-Crick pairs. Also in line 363 of the revised text, I am not happy with your statement that Chloroplast tRNAs have AC stems of 3 to 7 bp; a tRNA with a 3-bp AC stem is not a functional tRNA; the only exceptions are tRNAsMet with 6 bp in the AC stem carrying a terminal non-Watson-Crick pair. Reply: Thanks a lot for your helpful and valuable comments. We have rewritten and modified the sentences in line 232, line 234, line 236, line 238, and line 241 of the revised manuscript. We have checked carefully and corrected “bp” to “nt” for loops. We quite agree with your sagacious comments and suggestions, A-C pairs could form with one hydrogen bond or two hydrogen bonds at different pH, and in Table S1, we have corrected “6” to “7” bp for the AC stem of Sciadopitys verticillata_94443 tRNAThr. We have added contiguous genome-encoded nucleotides to the 5'- and 3'-end of structure B (tRNAMet in Taxus mairei_119804) and extended the AC stem (please see below result). And we have modified the sentence in line 363, we have rewritten “…the acceptor arm of chloroplast tRNAs contains 3 bp to 7 bp…” to “…the acceptor arm of chloroplast tRNAs contains 6 bp to 7 bp…” in the revised manuscript according to your suggestion. Thank you again for your helpful advices. 4) lines 319/320: I do not understand what you mean by '... which may have multiple functions through modification during the translation process in chloroplasts.' Reply: Thank you very much for your kindly comments. The sentence in line319/320 was to illustrate the complementary source of these tRNAs. And we have improved this sentence to make it clear. 5) line 371: was the previously reported conserved Ψ-loop motif (for Oryza sariva and Cyanobacteria) similar or identical to the one found here? Reply: Thanks so much for your helpful suggestions. Our study found the conserved motif, U-U-C-N-A-N2 was present in Ψ loop of chloroplast tRNA of Gymnosperm. And the conserved Ψ-loop motif reported in Oryza sariva and Cyanobacteria was U-U-C-X-A and U-U-C-X-A-X-U (Mohanta & Bae, 2017, Frontiers in genetics, 8:90; Mohanta et al., 2017, Frontiers in genetics, 8:200). So we hold that the previously reported conserved Ψ-loop motif (for Oryza sariva and Cyanobacteria) was similar to the one found in this study. 6) line 372, rewrite: '... Cyanobacteria' (Mohanta & Bae, 2017; Mohanta et al., 2017). This suggests that the consensus “U-U-C-N-A-N2” motif of the Ψ region, identified here and in previous analyses, is a general consensus motif of canonical tRNAs.' Reply: Many thanks to you for your kindly suggestion. We have rewritten the sentence “This suggested the short and conserved motif was universal in Ψ loop. Our study as well as previous researches revealed that the consensus “U-U-C-N-A-N2” motif in the Ψ region might be a general consensus motif of canonical tRNAs.” to “This suggests that the consensus “U-U-C-N-A-N2” motif of the Ψ region, identified here and in previous analyses, is a general consensus motif of canonical tRNAs.” according to your advice. 7) related to my previous comment r): what I want to see in Fig. 4 is not a colouring of the structural tRNA elements, but a presentation of conserved base identities, differentiated between '>50%, > 75% and 100% conserved'. Reply: Thanks a lot for your valuable comment. We have tried to show a presentation of conserved base identities. But we observed that different tRNA isotypes harbored their own conserved base identities. Thus, we counted the conservative rate at the level of gymnosperm chloroplast tRNA set. And different colors here were to illustrate different conservation. And we have supplemented the legend to explain conservative rate in Fig. 4. 8) in Table 4, indicate examplarily for the first tRNAAla the polarity of sequence elements, such as 5'-GGGGAUA-3' ||||||| 3'-CCCCUAU-5' for the AC arm, 5'-AGUUGGUA-3' for the D-loop etc. Reply: Thank you so much for your helpful suggestions. In the revised manuscript, we have supplemented the polarity of sequence elements for the first tRNAAla in Table 4. Reviewer 3 (Anonymous) Basic reporting 1. In figure 1 and 2; it would be good to label the expanded variable loop of tRNAs by making coloured box just to differentiate with other loop regions. Reply: Thank you for your helpful suggestions. In Figure 1 and 2, we have added colored box (yellow box) to the expanded variable loop of tRNAs to differentiate with other loop regions. 2. In fig 3; at the hinge region it looks that there is one extra ‘A’ or its counterpart base is missing. Just check? Reply: Thank you very much for your kindly advice. We have checked Fig. 3 and have modified it. 6. In the caption of Fig S1; it should be multiple instead of multiply. Reply: Done. 7. In table 4; There must be some modified bases in the tRNA; especially in the anticodon loop of tRNAs; so it would be good to mention ‘symbols’ of modified bases at 32, 34 and 37 as well as other positions if they are present. Reply: Many thanks to you for your helpful suggestion. In Table 4, we have added short lines to the bases in anticodon loop of tRNAIle to indicate its possible modification because possible base modification at anticodon loop (34, 35, and 36) of tRNAIle has been reported before (Soma et al., 2003, Molecular cell 12:689-698; Sonawane & Tewari, 2008, Nucleotides and Nucleic Acids 27:1158-1174). The anti-codon nucleotides in the anti-codon loop of tRNA is always numbered 34, 35, and 36 (Mohanta et al., 2017, Frontiers in genetics, 8:200). And in Discussion part, we have also discussed the possible modification mechanism of tRNAIle in the wobble position of its anticodon. Experimental design 9. Overall, the study is good and can be published in Peer J with some minor revisions. Reply: Thank you very much for your positive and kindly comments. We have carefully checked and revised the manuscript. And we hope this version would be suitable for the journal. Comments to authors (Manuscript ref Number: #43190-v2) 1. In figure 1 and 2; it would be good to label the expanded variable loop of tRNAs by making coloured box just to differentiate with other loop regions. Reply: Thank you for your valuable suggestion. In Fig. 1 and Fig. 2, we have supplemented colored box (yellow box) to the expanded variable loop of tRNAs to differentiate with other loop regions. 2. In fig 3; at the hinge region it looks that there is one extra ‘A’ or its counterpart base is missing. Just check? Reply: Thank you a lot for your helpful advice. In revised version, we have checked Fig. 3 and have modified it. 3. In the material & method section; line no. 127-128; please mention BLAST properly. Reply: Thank you so much for your kindly suggestion. We have corrected this mistake in line 127-128. 4. In these lines; 136-138; authors have mentioned default parameters of ARGAON; in which they have mentioned as; taken sequence source as a bacterial? Why? Reply: Thank you for your valuable comment. In the study, we used bacterial parameter because that chloroplast was generally convinced to be the origin of cyanobacteria. And based on referencing to other relevant literatures, we decided to use this parameter. 5. Authors sat that they have observed a novel tRNAGly in W. nobilis (Fig. 3) lacking DHU loop; it would be good to speculate the probable significance of this tRNA for the general readers. Reply: Thank you very much for your constructive advice. The novel tRNA structure maybe play some other significative functions in the translation progress and additional research is demanded to explain its exact function and mechanisms. And we have added this speculation in the Discussion part accordingly. 6. In the caption of Fig S1; it should be multiple instead of multiply. Reply: Many thanks to you for your suggestion. We have modified it according to your suggestion. 7. In table 4; There must be some modified bases in the tRNA; especially in the anticodon loop of tRNAs; so it would be good to mention ‘symbols’ of modified bases at 32, 34 and 37 as well as other positions if they are present. Reply: Thank you very much for your helpful comment and suggestion. In Table 4, we have supplemented short lines under the bases in anticodon loop of tRNAIle to demonstate its possible modification because possible base modification at anticodon loop (i.e., 34, 35, and 36) of tRNAIle has been reported before (Soma et al., 2003, Molecular cell 12:689-698; Sonawane & Tewari, 2008, Nucleotides and Nucleic Acids 27:1158-1174). The anti-codon nucleotides in the anti-codon loop of tRNA is always numbered 34, 35, and 36 (Mohanta et al., 2017, Frontiers in genetics, 8:200). And in Discussion part, we have also discussed the possible modification mechanism of tRNAIle in the wobble position of its anticodon. 8. In the conclusion section few results are mentioned, so it would be good to remove these results part and keep only conclusion of this study. By doing this authors can also reduce the size of the conclusion which is unnecessarily large and mention. Reply: Thank you a lot for your helpful suggestion. We have removed unnecessary results in Conclusion part to reduce the size of the conclusion according to your advice. 9. Overall, the study is good and can be published in Peer J with some minor revisions. Reply: Thank you so much for your kindly and positive comment. We have carefully and thoroughly revised the manuscript. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Gymnosperms such as ginkgo, conifers, cycads, and gnetophytes are vital components of land ecosystems, and they have significant economic and ecologic value, as well as important roles as forest vegetation. In this study, we investigated the structural variation and evolution of chloroplast transfer RNAs (tRNAs) in gymnosperms. Chloroplasts are important organelles in photosynthetic plants. tRNAs are key participants in translation where they act as adapter molecules between the information level of nucleic acids and functional level of proteins. The basic structures of gymnosperm chloroplast tRNAs were found to have family-specific conserved sequences. The tRNA &#936;-loop was observed to contain a conforming sequence, i.e., U-U-C-N-A-N 2 . In gymnosperms, tRNA Ile was found to encode a 'CAU' anticodon, which is usually encoded by tRNA Met . Phylogenetic analysis suggested that plastid tRNAs have a common polyphyletic evolutionary pattern, i.e., rooted in abundant common ancestors. Analyses of duplication and loss events in chloroplast tRNAs showed that gymnosperm tRNAs have experienced little more gene loss than gene duplication. Transition and transversion analysis showed that the tRNAs are isoacceptor specific and they have experienced unequal evolutionary rates. These results provide new insights into the structural variation and evolution of gymnosperm chloroplast tRNAs, which may improve our comprehensive understanding of the biological characteristics of the tRNA family.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Gymnosperms originated in the Paleozoic Devonian Period (about 385 million years ago), and they are key groups in terms of the transformation from spore reproduction to seed reproduction in higher plants <ns0:ref type='bibr' target='#b14'>(Gerrienne et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Crisp &amp; Cook, 2011)</ns0:ref>. According to the latest phylogenetic classification, gymnosperm species are divided into eight orders, 12 families, 84 genera, and more than 1,000 species <ns0:ref type='bibr' target='#b62'>(Wang &amp; Ran, 2014)</ns0:ref>. Gymnosperms include ginkgo, cycads, conifers, and gnetophytes, which are grown in forests as important timber species and they provide raw materials for human usage, such as fiber, resin, and tannin <ns0:ref type='bibr' target='#b2'>(Christenhusz et al., 2010)</ns0:ref>. In addition, gymnosperms include some important threatened plants, where 40% are at high risk of extinction <ns0:ref type='bibr' target='#b13'>(Forest et al., 2018)</ns0:ref>. Recent phylogenetic and evolutionary studies of gymnosperms have demonstrated the rapid evolution of mitochondrial (mt) genes and provided further evidence of sister relationship between conifers and Gnetales <ns0:ref type='bibr' target='#b50'>(Ran, Gao &amp; Wang, 2010)</ns0:ref>. The high levels of genetic diversity and population differentiation among the Pinus species in gymnosperms have been studied based on plastid DNA markers <ns0:ref type='bibr' target='#b39'>(Liu et al., 2014)</ns0:ref>. Other studies have indicated patterns related to the physiological ecology, phylogenetic relationships, and population genetic structure of gymnosperm species <ns0:ref type='bibr'>(Yu et al., 2014;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b8'>Dong et al., 2016)</ns0:ref>. However, these studies mainly considered the phylogeny and evolution at the whole populations level. Thus, the detailed evolutionary characteristics of gymnosperms still need to be elucidated.</ns0:p><ns0:p>Chloroplasts are the site of photosynthesis and of various essential metabolic pathways, e.g., fatty acid and amino acid biosynthesis and the assimilation of nitrogen, sulfur, and selenium <ns0:ref type='bibr'>(Wise &amp; Hoober, 2006;</ns0:ref><ns0:ref type='bibr'>Des Marais,2000;</ns0:ref><ns0:ref type='bibr' target='#b29'>Knorr &amp; Heimann, 2001;</ns0:ref><ns0:ref type='bibr' target='#b47'>Pilon-Smits et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b16'>Guo et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b31'>Kretschmer, Croll &amp; Kronstad, 2017)</ns0:ref>. It is generally recognized that chloroplasts are derived from proto-eukaryotic symbiotic cyanobacteria that internalized in eukaryotic cells <ns0:ref type='bibr' target='#b20'>(Hiroki &amp; Daisuke, 2018</ns0:ref>) and evolved into central organelles. Chloroplasts have their own genome encoding about 100 proteins and they are maternally inherited organelles in most angiosperm plants <ns0:ref type='bibr' target='#b0'>(Abdallah, Salamini &amp; Leister, 2000;</ns0:ref><ns0:ref type='bibr' target='#b19'>Heuertz et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b3'>Civan et al., 2014)</ns0:ref>. Among gymnosperms, paternal plastid inheritance is the typical characteristic of conifers <ns0:ref type='bibr'>(Faur&#233; et al., 1994;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kaundun &amp; Matsumoto, 2011)</ns0:ref>. Studies have shown that the chloroplast genome is quite conserved with an average evolutionary rate of 0.2-1.0&#61620;10 -9 per site per year, which is only one-fifth of that for the nuclear genome <ns0:ref type='bibr' target='#b9'>(Drouin, Daoud &amp; Xia, 2008;</ns0:ref><ns0:ref type='bibr' target='#b10'>Duchene &amp; Bromham, 2013)</ns0:ref>. The chloroplast genome is a covalently closed circular structure with four parts comprising the large single copy region (LSC), small single copy region (SSC), inverted repeat region A (IRa), and inverted repeat region B (IRb). The two IRs have the same sequences but in the opposite direction <ns0:ref type='bibr' target='#b61'>(Wang et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b40'>Logacheva et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b18'>Hereward et al., 2018)</ns0:ref>. Due to the independent evolution of the chloroplast genome, it is possible to construct a molecular phylogenetic tree using the chloroplast genome and without requiring any other data. Data analysis based on the conserved evolution of plastids is highly valuable for phylogenetic studies <ns0:ref type='bibr' target='#b27'>(Kim &amp; Suh, 2013)</ns0:ref> because it can provide reliable and useful phylogenetic information. The Manuscript to be reviewed relative completeness and independence of the chloroplast genome means that it can provide valuable material for research purposes.</ns0:p><ns0:p>Transfer RNAs (tRNAs) undergo numerous post-transcriptional nucleotide modifications and they exhibit abundant chemical diversity where the bases experience methylation, formylation, and other modifications <ns0:ref type='bibr' target='#b57'>(Suzuki &amp; Suzuki, 2014)</ns0:ref>. Chemical nucleotide modifications are frequent in tRNAs and they are important for the structure, stability, correct folding, aminoacylation, and decoding. For example, a previous analysis of the chemically synthesized f 5 C34-modified anticodon loop of human mt-tRNA Met showed that f 5 C34 contributes to the anticodon domain structure of the mt-tRNA <ns0:ref type='bibr' target='#b42'>(Lusic et al., 2008)</ns0:ref>. tRNAs comprise sequences of less than 100 polynucleotides that fold into a clover-type secondary structure and then into an L-shaped tertiary structure <ns0:ref type='bibr' target='#b64'>(Wilusz, 2015)</ns0:ref>. The secondary structure of tRNAs comprises different arms as well as loops, i.e., the D-arm, acceptor arm, anticodon arm, pseudouridine arm (&#936;-arm), D-loop, variable arm, anticodon loop, and pseudouridine loop (&#936;loop) <ns0:ref type='bibr' target='#b15'>(Gieg&#233;, Puglisi &amp; Florentz, 1993;</ns0:ref><ns0:ref type='bibr' target='#b43'>Mizutani &amp; Goto, 2000)</ns0:ref>. This unique structure allows tRNA to act as important bridges between the information level of nucleic acids and functional level of proteins. The vital components of tRNAs comprise an anti-codon region that discerns the messenger RNA carried by the specific codons, a 3&#61602;-CCA tail for attaching to the cognate amino acid, the &#936;-arm, and a &#936;-loop that has a relationship with the ribosome machinery <ns0:ref type='bibr' target='#b28'>(Kirchner &amp; Ignatova, 2014)</ns0:ref>. Asymmetric combinations and the divided segments in tRNA genes allow us to understand the diversity of tRNA molecules. tRNA species fulfill various functions in cellular homeostasis, regulation of gene expression and epigenetics, biogenesis, and even biological disease <ns0:ref type='bibr' target='#b51'>(Ribasd &amp; Dedon, 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Kanai, 2015;</ns0:ref><ns0:ref type='bibr' target='#b54'>Schimmel, 2017)</ns0:ref>. The evolutionary relationships determined between cyanobacteria and monocots show that tRNAs evolved polyphyletically and they originated from multiple common ancestors with a high rate of gene loss <ns0:ref type='bibr'>(Mohanta et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mohanta et al., 2019)</ns0:ref>. Nevertheless, the basic details of the tRNAs in plant chloroplasts still need to be elucidated and on the diverse evolutionary features of gymnosperm tRNAs are still unclear.</ns0:p><ns0:p>In this study, we assessed all of the chloroplast genomes in 12 families of gymnosperms from eight orders. The main aims of this study were as follows: (1) to determine the diversification of nucleotides in the secondary structure of gymnosperm tRNAs; (2) to identify the detailed genomic features of chloroplast tRNAs; (3) to assess the evolutionary relationships among different chloroplast tRNAs; and (4) to evaluate the duplication or loss events that occurred in all of the tRNAs considered. Our findings provide important insights into the biological characteristics and evolutionary variation of the tRNA family.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>Annotation and identification of chloroplast tRNA sequences in gymnosperms We downloaded complete chloroplast genomes for 12 representative gymnosperms in eight orders from the National Center for Biotechnology Information database (NCBI, https://www.ncbi.nlm.nih.gov/). The gymnosperm species investigated were: Cycas debaoensis <ns0:ref type='formula'>NC_011954</ns0:ref>). The gymnosperm tRNA genomes were annotated using GeSeq-Annotation of Organellar Genomes tool <ns0:ref type='bibr' target='#b58'>(Tillich et al., 2017)</ns0:ref> where the parameters were set as: circular sequence(s), chloroplast of sequence source, generate multi FASTA; BLAST protein search identity 25% for annotating plastid IR, 85% identity for BLAST rRNA, tRNA and DNA search, Embryophyta chloroplast (CDS+rRNA), third party tRNA annotator ARAGORN v1.2.38, ARWEN v1.2.3, tRNAScan-SE v2.0, and without Refseq choice.</ns0:p><ns0:p>Structural analysis of chloroplast tRNAs ARAGORN <ns0:ref type='bibr' target='#b36'>(Laslett &amp; Canback, 2004)</ns0:ref> and tRNAScan-SE software <ns0:ref type='bibr' target='#b41'>(Lowe &amp; Eddy, 1997)</ns0:ref> were employed to analyze the sequences and the secondary structure of tRNAs in the chloroplast genomes of the involved gymnosperm plants. The default parameters were set in ARAGORN software. The parameters for tRNAScan-SE were set as: sequence source, bacterial; search mode, default; query sequences, formatted (FASTA); and genetic code for tRNA isotype prediction, universal. Phylogenetic tree construction A phylogenetic tree was constructed for all of the tRNAs using MEGA7.0 software <ns0:ref type='bibr' target='#b32'>(Kumar et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b33'>Kumar, Stecher &amp; Tamura, 2016)</ns0:ref>. To study the evolutionary details of chloroplast tRNAs in gymnosperm species, an alignment file for tRNAs was achieved by CLUSTAL Omega software before the phylogenetic tree was constructed. MEGA7 software was used to transform the alignment file into MEGA format. The phylogenetic tree was constructed with the following parameters: phylogeny reconstruction of analysis, maximum likelihood model, bootstrap method in phylogeny test, 1000 bootstrap replicates, nucleotides type, gamma distributed with invariant sites (G+I) model, five discrete gamma categories, partial deletion for gaps/missing data treatment, 95 % site coverage cut-off, and very strong for branch swap filter.</ns0:p><ns0:p>Transition/transversion analysis The sequences of the tRNA isotypes were aligned to determine the transition and transversion rates for chloroplast tRNAs in gymnosperm plants. The files covering all 20 types of tRNAs were transformed into the MEGA file format and analyzed separately using MEGA7.0 software <ns0:ref type='bibr' target='#b35'>(Kumar, Tamura &amp; Nei, 1994)</ns0:ref>. The transition and transversion rates were analyzed for tRNAs with the following parameters: substitution pattern estimation (ML) analysis, automatic (neighbor-joining tree), maximum likelihood statistical method, nucleotide substitution type, Kimura two-parameter model, gamma distributed (G) site rates, five discrete gamma categories, partial deletion of gaps/missing data treatment, 95% of site coverage cut-off, and very strong branch swap filter.</ns0:p><ns0:p>Loss and duplication events analysis for tRNA genes In order to investigate the duplication or loss events in tRNA genes, the NCBI taxonomy browser was utilized to construct the whole species tree for the 12 gymnosperm species considered. The phylogenetic tree conducted in the evolutionary study was employed as gene tree. The gene tree for the tRNAs and species tree for the gymnosperm species were submitted to Notung 2.9 software <ns0:ref type='bibr' target='#b1'>(Chen, Durand &amp; Farach-Colton, 2000)</ns0:ref>, and then reconciled to discover duplicated and lost tRNA genes in the chloroplast genomes of gymnosperms.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Genomic features of gymnosperm chloroplast tRNAs Sequences were analyzed to identify the genomic tRNAs in the chloroplast genomes of 12 gymnosperm species comprising C. <ns0:ref type='bibr'>debaoensis, D. spinulosum, G. biloba, C. deodara, W. nobilis, R. piresii, S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon, and E. equisetina</ns0:ref>, which were obtained from the NCBI database (Table <ns0:ref type='table'>1</ns0:ref>). The results showed that the length of the chloroplast tRNAs vary from the smallest with 64 nucleotides (nt) (tRNA Met -CAU in T. mairei) to the largest with 96 nt (tRNA Tyr -AUA in W. nobilis, C. deodara, and G. biloba) (Data S1). We found that the chloroplast genomes of gymnosperm plants encode 28 to 33 tRNAs (Table <ns0:ref type='table'>2</ns0:ref>), where D. spinulosum, C. deodara, and S. verticillata encode 31 anticodons, W. nobilis, R. piresii, C. lanceolata, and G. gnemon encode 32 tRNA isotypes, G. biloba, and W. mirabilis encode 33 tRNAs. Other species comprising T. mairei, E. equisetina and C. debaoensis encode 28, 28, 30 tRNA isotypes, respectively (Table <ns0:ref type='table'>2</ns0:ref>). tRNA Ala was not found in R. piresii and T. mairei, and tRNA Val was not detected in T. mairei (Fig. <ns0:ref type='figure'>S3</ns0:ref>). We also observed that all of the species do not encode selenocysteine and its suppressor tRNA (Table <ns0:ref type='table'>2</ns0:ref>). Overall, tRNA Ser (in W. nobilis) and tRNA Arg (in W. mirabilis) are the most abundant (four types) followed by tRNA Leu (three types) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>Variations in structures of chloroplast tRNAs Some tRNAs with a loop structure in the variable region were found to be encoded in the gymnosperm chloroplast genomes (Fig. <ns0:ref type='figure' target='#fig_4'>1</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). A novel tRNA lacking the D-arm was found in tRNA Gly in W. nobilis (Fig. <ns0:ref type='figure'>3</ns0:ref>). As shown in Fig. <ns0:ref type='figure' target='#fig_4'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>, tRNA Leu , tRNA Ser , and tRNA Tyr contain expanded variable stem/loops. In these tRNAs (except for tRNA Ser -GCU of D. spinulosum), the anticodon loop of tRNA Ser contains the conserved consensus sequence N-U-N-G-A-A-N, and tRNAs Leu have the consensus sequence C-U-N-A-N 2 -A. The variable loop region is predicted to fold into stem-loop structures with apical loops of 3 to 7 nt in tRNA Ser and several tRNA Leu variants. The stems contain up to 7 bp (Fig. <ns0:ref type='figure' target='#fig_5'>1 and 2</ns0:ref>). The expanded variable loop structures may play important functions during the protein translation process in chloroplasts.</ns0:p></ns0:div> <ns0:div><ns0:head>Chloroplast genomes contain 25 to 30 anticodon-specific tRNAs</ns0:head><ns0:p>The genomes of the species analyzed were found to code for at least two copies of tRNA Met -CAU/tRNA fMet -CAU. Each of the gymnosperm chloroplast genomes encodes 25 to 30 anticodon-specific tRNAs (Table <ns0:ref type='table' target='#tab_2'>2 and Table 3)</ns0:ref>, where E. equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S. verticillata, and C. lanceolata encode 28 anticodons, and D. spinulosum, C. deodara, W. mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis, G. gnemon and G. biloba encodes 30 anticodons (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). tRNA Arg -CCG was present in the genomes of nine gymnosperm species but absent from C. lanceolata, T. mairei, and E. equisetina, while tRNA Gly -UCC was lacking from C. debaoensis, S. verticillata, D. spinulosum, C. lanceolata, T. mairei, and E. equisetina (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). The most abundant anticodons found in the chloroplast genomes were tRNA Gly -GCC, tRNA Pro -UGG, tRNA Ser -UGA, tRNA Ser -GCU, tRNA Arg -ACG, tRNA Arg -UCU, tRNA Leu -UAG, tRNA Leu -CAA, tRNA Phe -GAA, tRNA Asn -GUU, tRNA Lys -UUU, tRNA Asp -GUC, tRNA Glu -UUC, tRNA His -GUG, tRNA Gln -UUG, tRNA Ile -CAU, tRNA Met -CAU, tRNA Tyr -GUA, tRNA Cys -GCA, and tRNA Trp -CCA (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). Two tRNA Trp iso-acceptors are present in E. equisetina chloroplasts, compared with a single one in the other gymnosperm species analyzed in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved gymnosperm chloroplast tRNAs</ns0:head><ns0:p>The clover leaf-like secondary structure of a tRNA is shown in Fig. <ns0:ref type='figure'>4</ns0:ref>. In the study, we found that most tRNAs contain a 'G' as the first nucleotide in the D-arm, except for tRNA Lys , tRNA Met , tRNA Pro , tRNA Thr , tRNA Tyr , and tRNA Val . 'A' is present in the first and the last position of the D-loop apart from tRNA Gly , tRNA Ile , tRNA Leu , tRNA Met , and tRNA Gln . In addition, in the final two positions of the &#936;-arm, all of the tRNAs were found to have conserved 'G-G' nucleotides, except for tRNA Arg , tRNA Cys , tRNA Phe , and tRNA Val (Table <ns0:ref type='table'>4</ns0:ref>). Small conserved consensus sequences were found in the &#936; region. To be specific, except for tRNA Ser , the &#936;-loop in tRNAs was found to contain a conserved sequence comprising U-U-C-N-A-N 2 according to a multiple sequence alignment of 20 members of the tRNA gene family (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Diversification of tRNAs structures</ns0:head><ns0:p>The diverse arms and loops of tRNAs allow the regulation and control of protein translation. Each arm and loop has a specific nucleotide composition. Our analysis based on 373 tRNAs showed that the acceptor arm of chloroplast tRNAs contains 6 bp to 7 bp (Table <ns0:ref type='table'>S1</ns0:ref>). The D-arms were found to contain 3 or 4 bp generally, with a stable 'G' in the initial position and 'C' in the last position of the D-stem 5' strand in most tRNAs (such as tRNA Ala , tRNA Asn , tRNA Asp , tRNA Cys , tRNA Glu , tRNA His , tRNA Ile , and tRNA Phe ). Most D-loops usually contain 7 to 11 nt with conserved 'A' nucleotides at the two end locations. The anticodon arms of chloroplast tRNAs mainly contain 5 bp (90.4%). We found that 367 (about 99%) tRNAs contain 7 nt in their anticodon loop, thereby indicating that the sequence of the anticodon loop is highly conserved (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The variable loops of different tRNAs contain 3 to 23nt, where those in tRNA Ala , tRNA Asp , tRNA His , tRNA Phe , and tRNA Pro contain 5 bp (Table <ns0:ref type='table'>S1</ns0:ref>). The &#936;-arm contains 5 bp in most of the gymnosperm chloroplast tRNAs, except for tRNA Ala and some of the tRNA Trp , tRNA Gly , tRNA Thr , and tRNA Arg in chloroplast. The &#936;-loops of most tRNAs contain 7 nt, apart from tRNA Ala and several of tRNA Cys and tRNA Thr (Table <ns0:ref type='table'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gymnosperm chloroplast tRNAs derived from multiple common ancestors</ns0:head><ns0:p>The phylogenetic tree demonstrated the presence of three major clusters covering 64 groups and the different types of all tRNAs (as shown by the different strings in Fig. <ns0:ref type='figure' target='#fig_2'>S1</ns0:ref>). We detected 37 groups in cluster I, five in cluster II, and 22 groups in cluster III. Cluster I contains tRNA Ser , tRNA Tyr , tRNA His , tRNA Gln , tRNA Thr , tRNA Pro , tRNA Gly , tRNA Met , tRNA Asp , tRNA Arg , tRNA Ala , tRNA Cys , tRNA Lys , tRNA Glu , tRNA Ile , tRNA Asn , tRNA Val , tRNA Leu , and tRNA Trp . Cluster II contains tRNA His , tRNA Ser , tRNA Tyr , and tRNA Leu . Cluster III contains tRNA Leu , tRNA Ile , tRNA Gly , tRNA Thr , tRNA Ser , tRNA Val , tRNA Glu , tRNA Lys , tRNA Cys , tRNA Gln , tRNA His , tRNA Arg , tRNA Phe , tRNA Ala , and tRNA Met (Fig. <ns0:ref type='figure' target='#fig_2'>S1</ns0:ref>). tRNA Ser , tRNA His , and tRNA Leu are present in cluster I but also in cluster II and cluster III, thereby suggesting that these tRNAs evolved from multiple lineages. Most of the tRNAs were found to form more than one group in the phylogenetic tree. In cluster I, the tRNAs that formed two groups in the phylogenetic tree were identified as tRNA Tyr , tRNA Gln , tRNA Met , tRNA Asp , tRNA Ala , tRNA Lys , tRNA Ile , and tRNA Trp , whereas those that clustered to form three groups were determined as tRNA Ser , tRNA Pro , tRNA Arg , tRNA Glu , tRNA Asn , tRNA Val , and tRNA Leu . Moreover, tRNA Thr clustered into four groups. In cluster II, tRNA Ser was found to form two groups. In cluster III, tRNA Gly and tRNA Val were found to form two groups, whereas tRNA Thr formed three groups, tRNA Ile formed four groups. Some tRNAs in cluster III were found to group individually, where these tRNAs containing the anticodons C-G-A in tRNA Ser , U-U-C in tRNA Glu , U-U-U in tRNA Lys , G-C-A in tRNA Cys , U-U-G in tRNA Gln , G-U-G in tRNA His , U-C-U in tRNA Arg , G-A-A in tRNA Phe , U-G-C in tRNA Ala , and C-A-U in tRNA Met all grouped separately (Fig. <ns0:ref type='figure' target='#fig_2'>S1</ns0:ref>). The multiple groupings of different tRNAs suggest that they evolved from multiple common ancestors. Furthermore, the tRNAs presented in cluster III, i.e., tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA His (GUG), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Ile (UAU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Gly (UCC), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU), tended to be the most basic tRNAs and they had undergone gene duplication and diversification to generate other tRNA molecules.</ns0:p></ns0:div> <ns0:div><ns0:head>C-A-U anticodon in tRNA Ile</ns0:head><ns0:p>Our detailed genomic study showed that tRNA Ile also encodes a C-A-U anticodon in addition to the presence of this typical anticodon in tRNA Met . In general, the C-A-U anticodon is recognized as a typical characteristic of tRNA Met and there is only one iso-acceptor. In particular, we found that the tRNA Ile in T. mairei encodes two C-A-U anticodons, and C. debaoensis, <ns0:ref type='bibr'>S. verticillata, D. spinulosum, C. lanceolata, G. biloba, C. deodara, W. mirabilis, G. gnemon, R. piresii, E. equisetina, and W.</ns0:ref> nobilis also encode a C-A-U anticodon (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>, Data S1, Fig. <ns0:ref type='figure'>S3</ns0:ref>).</ns0:p><ns0:p>Transition/transversion of tRNAs A previous study <ns0:ref type='bibr' target='#b45'>(Mohanta et al., 2019)</ns0:ref> showed that the evolutionary rates are almost equal for tRNAs with respect to transition and transversion despite the low probability of transition or transversion events in tRNAs. In this study, we identified several intriguing substitutions of gymnosperm chloroplast tRNAs. Overall, our analysis of the substitution rates detected using the whole set of chloroplast tRNAs showed that average transition rate (15.38) was significantly larger than the average transversion rate (4.81) with a ratio of 3:1 (Table <ns0:ref type='table' target='#tab_3'>5</ns0:ref>). The same transition: transversion ratio bias was found in all the set of tRNAs for tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , and tRNA Leu . The ratio was over 6:1 for tRNA Cys and tRNA Arg . The transition rates for tRNA Trp , tRNA Val , and tRNA Gly were about 10 times higher than their transversion rates. These findings suggest that tRNA Ser , tRNA Glu , tRNA Tyr , tRNA Ile , tRNA Met , tRNA Gln , tRNA Thr , tRNA Leu , tRNA Cys , tRNA Arg , tRNA Trp , tRNA Val and tRNA Gly underwent transition substitutions more readily than transversion substitutions during their evolution in gymnosperm chloroplast genomes. In addition, the transition rates in tRNA Lys and tRNA Pro were about 15 times higher than their transversion rates. The transition rates in tRNA Asn , tRNA Phe , and tRNA His were about 20 times higher than their transversion rates. These results indicate that tRNAs are much more likely to have undergone transition events rather than transversion events. The highest transversion rate of 12.50 was found in tRNA Ala and the lowest transversion rate of 0.00 in tRNA Asp (Table <ns0:ref type='table' target='#tab_3'>5</ns0:ref>). Correspondingly, tRNA Ala lacks any transitions (Table <ns0:ref type='table' target='#tab_3'>5</ns0:ref>).</ns0:p><ns0:p>tRNA duplication/loss events In addition to transition and transversion events, gene duplication and loss events have played important roles in gene evolution. Our analysis of duplication and loss events indicated that 153 duplication events (duplication and conditional duplication) have occurred in all of the gymnosperm chloroplast tRNA genes investigated in this study (Fig. <ns0:ref type='figure' target='#fig_3'>S2</ns0:ref>). In addition, 220 gymnosperm chloroplast tRNA gene loss events were detected (Table <ns0:ref type='table'>S2</ns0:ref>, Fig. <ns0:ref type='figure' target='#fig_3'>S2</ns0:ref>). Thus, the loss of genes was slightly more frequent than their duplication for gymnosperm chloroplast tRNA genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>tRNAs are major genetic components of semi-autonomous chloroplasts and our analysis of gymnosperm chloroplast genomes showed that they have several basic conserved genomic features. The gymnosperm chloroplast genomes investigated in the present study were found to encode 28 to 33 tRNA isotypes, thereby indicating that there is substantial variation in the quantity of tRNAs in gymnosperm chloroplast genomes. The lack of tRNA Ala in R. piresii and T. mairei, and the absence of tRNA Val in T. mairei were interesting. Thus, it is necessary to understand how the translation process is conducted in chloroplasts without these crucial tRNAs. According to previous studies <ns0:ref type='bibr' target='#b60'>(Treangen &amp; Rocha, 2011;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mohanta et al., 2019)</ns0:ref>, it is likely that the deficiency of these tRNAs is compensated for by the transfer of corresponding tRNAs from the nucleus or mitochondria. In addition to the absence of tRNA Ala and tRNA Val , all of the gymnosperm plants were shown to not encode selenocysteine tRNA and its suppressor tRNA in their chloroplast genomes (Table <ns0:ref type='table'>2</ns0:ref>). Selenocysteine tRNA and its suppressor tRNA were also not detected in the chloroplast of Oryza sativa <ns0:ref type='bibr' target='#b44'>(Mohanta &amp; Bae, 2017)</ns0:ref>.</ns0:p><ns0:p>In addition to the presence of C-A-U anticodon in tRNA Met , we found that tRNA-CAU is present in tRNA Ile (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). Similarly, the C-A-U anticodon was detected in tRNA Ile in Bacillus subtilis (Ehrenberg) Cohn and spinach <ns0:ref type='bibr' target='#b25'>(Kashdan &amp; Dudock, 1982;</ns0:ref><ns0:ref type='bibr' target='#b30'>K&#246;hrer et al., 2014)</ns0:ref>. The possible mechanism that governs the specificity of this amino acid may involve modification of the wobble position in the anticodon by a tRNA-modifying enzyme. Chloroplasts originate from bacteria so the tRNA modifications found in bacteria may also occur in chloroplast tRNAs. In bacteria, the tRNA-modifying enzyme TilS can convert the 5&#61602;-C residue in the CAU anticodon of specific tRNA Ile molecules into lysidine to decode 5&#61602;-AUA (Ile) codons instead of 5&#61602;-AUG (Met) codons <ns0:ref type='bibr' target='#b55'>(Soma et al., 2003)</ns0:ref>. In addition, when lysidine decodes isoleucine, the tautomer form of lysidine provides compatible hydrogen bond donor-acceptor sites to allow base pairing with 'A' and this may help to the recognition of the codon AUA instead of AUG <ns0:ref type='bibr' target='#b56'>(Sonawane &amp; Tewari, 2008;</ns0:ref><ns0:ref type='bibr' target='#b53'>Sambhare et al., 2014)</ns0:ref>. The absence of tRNA Ile -lysidine synthetase leads to a failure to modify C34 to lysidine in tRNA Ile (LAU) (i.e., the synthesis of CAU-tRNA Ile ) and this inactivates the translation of AUA codons <ns0:ref type='bibr' target='#b30'>(K&#246;hrer et al., 2014)</ns0:ref>.</ns0:p><ns0:p>During protein coding, a certain species or gene tends to use one or more specific synonym codons, which is referred to as codon usage bias <ns0:ref type='bibr'>(Comeron &amp; Aguad&#233;, 1998;</ns0:ref><ns0:ref type='bibr' target='#b52'>Rota-Stabelli et al., 2012)</ns0:ref>. In the present study, tRNA Arg -CCG was found to be present in the genomes of nine species but absent from C. lanceolata, T. mairei, and E. equisetina. Similarly, tRNA Gly -UCC was shown to be absent from the chloroplast genomes of C. debaoensis, S. verticillata, D. spinulosum, C. lanceolata, T. mairei, and E. equisetina (Table <ns0:ref type='table' target='#tab_1'>3</ns0:ref>). These results suggest that gymnosperm chloroplast tRNA genes are characterized by codon usage bias <ns0:ref type='bibr' target='#b63'>(Wei &amp; Jin, 2017;</ns0:ref><ns0:ref type='bibr'>Li et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In general, the secondary structure of tRNAs is characterized as clover leaf-like, except for a few tRNAs with unusual secondary structures <ns0:ref type='bibr' target='#b23'>(J&#252;hling et al., 2018)</ns0:ref>. In our study, we identified clover leaf-like tRNAs with expanded variable loop regions (Fig. <ns0:ref type='figure' target='#fig_4'>1</ns0:ref> and Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). Numerous tRNA Leu , tRNA Ser , and tRNA Tyr were found to have specific variable loop configurations in terms of length and structure, suggesting significant structural variation among chloroplast tRNAs. It is interesting to note that there were also stem-loop structures in variable regions of certain tRNAs in cyanobacteria. This might indicate that similar structural variations exist between chloroplast tRNAs and cyanobacterial tRNAs <ns0:ref type='bibr'>(Mohanta et al., 2017)</ns0:ref>. Future studies will have to determine the biological importance of these variant tRNAs. The novel tRNA structure lacking the D arm might play some other significative functions in the translation progress and additional research is necessary to elucidate its exact function and mechanisms. Most tRNAs have a clover-like structure formed by complementary base pairing between small segments <ns0:ref type='bibr' target='#b22'>(Hubert et al., 1998;</ns0:ref><ns0:ref type='bibr' target='#b12'>Florentz, 2002)</ns0:ref>. Previous studies have showed that the acceptor arm of tRNAs in chloroplasts contain 7 bp to 9 bp, the D-arm contains 3 bp to 4 bp, the D-loop has 4 nt to 12 nt, the anticodon arm has 5 bp, the anticodon loop contains 7 nt, the variable region comprises 4 nt to 23 nt, and &#936;-arm contains 5 bp, and the &#936;-loop has 7 nt <ns0:ref type='bibr' target='#b64'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b46'>Mohanta &amp; Hanhong, 2017;</ns0:ref><ns0:ref type='bibr' target='#b45'>Mohanta et al., 2019)</ns0:ref>. In the present study, we found that the acceptor arm of chloroplast tRNAs contains 6 bp to 7 bp in 373 tRNAs, where the D-arm has 3 bp or 4 bp and the D-loop usually contains 7 nt to 11 nt. The anticodon loop of gymnosperm chloroplast tRNAs generally contains 7 nt, and thus the sequence of the anticodon loop is typically conserved (Table <ns0:ref type='table'>4</ns0:ref>, Table <ns0:ref type='table'>S1</ns0:ref>). The variable loop of different tRNAs contain 3 nt to 23 nt (Table <ns0:ref type='table'>S1</ns0:ref>). The &#936;-arm of gymnosperm chloroplast tRNAs generally contains 5 bp and the &#936;loop has 7 nt (Table <ns0:ref type='table'>S1</ns0:ref>). Our results are consistent with previous findings <ns0:ref type='bibr' target='#b64'>(Wilusz, 2015;</ns0:ref><ns0:ref type='bibr' target='#b46'>Mohanta &amp; Hanhong, 2017)</ns0:ref> and they suggest that chloroplast RNAs are significantly conserved. The consensus sequence 'U-U-C-N-A-N 2 ' was found in the &#936; region (Table <ns0:ref type='table'>4</ns0:ref>). Previous studies also reported the existence of a similar sequence in the &#936;-loop of tRNAs in Oryza sariva and Cyanobacteria <ns0:ref type='bibr' target='#b44'>(Mohanta &amp; Bae, 2017;</ns0:ref><ns0:ref type='bibr'>Mohanta et al., 2017)</ns0:ref>. This suggests that the consensus 'U-U-C-N-A-N 2 ' motif of the &#936; region, identified here and in previous analyses, is a general consensus motif of canonical tRNAs.</ns0:p><ns0:p>Our phylogenetic analysis detected three clear clusters and many tRNA groups. Some tRNAs (tRNA Ser , tRNA His , and tRNA Leu ) in cluster I and cluster II were also in cluster III, thereby indicating that these tRNAs evolved from multiple lineages by gene duplication and gene divergence. Moreover, anticodon types comprising CGA, UUC, UUU, GCA, UUG, GUG, UCU, UGC, and CAU appeared several times in the phylogenetic tree, and thus the corresponding tRNAs evolved from multiple common ancestors. The overlapping of tRNAs groups demonstrates that these tRNAs might have diverse common ancestors in the evolutionary process <ns0:ref type='bibr' target='#b44'>(Mohanta &amp; Bae, 2017)</ns0:ref>. Phylogenetic analysis also showed that tRNA Met (CAU), tRNA Thr (UGU, GGU), tRNA Val (UAC), tRNA Ala (UGC), tRNA Phe (GAA), tRNA Arg (UCU), tRNA His (GUG), tRNA Gln (UUG), tRNA Cys (GCA), tRNA Lys (UUU), tRNA Glu (UUC), tRNA Ile (UAU), tRNA Val (GAC), tRNA Leu (CAA), tRNA Gly (UCC), tRNA Ser (CGA), tRNA Gly (GCC), and tRNA Ile (CAU) in cluster III tended to be the most basic tRNAs, whereas tRNA Met tended to be the most original tRNA. Overall, the results clearly indicate that the tRNAs encoded in gymnosperm chloroplast genomes have multiple common evolutionary ancestors.</ns0:p><ns0:p>Our results also provided insights into the gene substitution rates in gymnosperm chloroplast tRNAs. Overall, the average transition rate for tRNAs was greater than the transversion rate, where the relationship was about 3:1 (Table <ns0:ref type='table' target='#tab_3'>5</ns0:ref>). In all of the chloroplast tRNAs, the average transition rate was slightly higher than the average transversion rate, thereby indicating that chloroplast tRNAs have unequal substitution rates.</ns0:p><ns0:p>In addition to the transition and transversion events in tRNAs, loss and duplication events have played significant roles in the evolution of tRNAs in gymnosperm chloroplast genomes <ns0:ref type='bibr' target='#b17'>(He &amp; Zhang, 2006;</ns0:ref><ns0:ref type='bibr'>Magadum et al., 2013)</ns0:ref>. In general, the gene loss events tended to occur after whole genome duplication events. We found 153 duplication events and 220 loss events in gymnosperm chloroplast tRNAs, and thus loss events have occurred slightly more frequently than duplication events (Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our basic structure analysis showed that gymnosperm chloroplast genomes encode 25 to 30 anticodon-specific tRNAs. The acceptor arm of chloroplast tRNA contains 6 bp to 7 bp, the Darm has 3 bp or 4 bp, the D-loop contains 7 nt to 11 nt mainly, and the anticodon loop usually contains 7 nt. In different tRNAs, the variable loop contains 3 nt to 23 nt. The &#936;-arm contains a conserved sequence comprising U-U-C-N-A-N 2 . tRNA Ala was absent from R. piresii and T. mairei, and tRNA Val was lacking in T. mairei. Gymnosperm chloroplasts do not encode selenocysteine tRNA and its suppressor tRNA in their genomes. A CAU anticodon is encoded in tRNA Met as well as in tRNA Ile . A novel tRNA structure lacking the D arm was identified for the chloroplast tRNA Gly of W. nobilis. Numerous tRNA Leu , tRNA Ser , and tRNA Tyr types were found to have expanded variable regions. Phylogenetic analysis showed that tRNAs might have multiple common ancestors in the evolutionary process. Different tRNAs harbored their own transition/transversion rates, i.e., it was iso-acceptor specific. And the transition rate was generally higher than the transversion rate. Furthermore, gene loss events (220) have occurred slightly more frequently than gene duplication events (153) in gymnosperm chloroplast tRNAs. Our results provide new insights into the evolution of gymnosperm chloroplast tRNAs and their diverse roles.</ns0:p><ns0:p>Yu F, Wang DX, Yi XF, Shi XX, Huang YK, Zhang HW, Zhang XP. 2014. Does animalmediated seed dispersal facilitate the formation of Pinus armandii-quercus aliena var.</ns0:p><ns0:p>acuteserrata forests? PloS one 9:e89886. DOI: 10.1371/journal.pone.0089886</ns0:p></ns0:div> <ns0:div><ns0:head>Figure Legends</ns0:head><ns0:p>Fig. <ns0:ref type='figure' target='#fig_4'>1</ns0:ref> Certain tRNAs in C. debaoensis, D. spinulosum, G. biloba, C. deodara, and R. piresii contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from C. debaoensis, D. spinulosum, G. biloba, C. deodara, R. piresii were observed to contain an expanded variable stem and loop. The anti-codon loop of tRNA Ser (except for tRNA Ser -GCU of D. spinulosum) was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and tRNA Leu harbor the consensus sequence C-U-N-A-N 2 -A. Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref> Certain tRNAs in S. verticillata, C. lanceolata, T. mairei, W. mirabilis, and G. gnemon contain expanded variable stem and loops. tRNA Ser , tRNA Leu , and tRNA Tyr from S. verticillata, C. lanceolata, T. mairei, W. mirabilis, G. gnemon were observed to contain a variable stem and loop. The anti-codon loop of tRNA Ser was made up of seven nucleotides with the conservative N-U-N-G-A-A-N consensus sequence, and the consensus sequence was C-U-N-A-N 2 -A for tRNA Leu . Fig. <ns0:ref type='figure'>3</ns0:ref> An abnormal tRNA structure lacking the D-arm found in W. nobilis. The tRNA Gly with anti-codon UCC was found lacking D-arm. Fig. <ns0:ref type='figure'>4</ns0:ref> Clover leaf-like structure of gymnosperms tRNA. The tRNA contains the Acceptor arm (6-7 bp, dark green, &gt;96% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region (3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, light purple, &gt;97% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). '% conservation' means the conservative ratio of base identities in each stem and loop structure of the whole set of gymnosperm tRNAs. Several tRNAs harbor the nucleotides of C-C-A tail.</ns0:p></ns0:div> <ns0:div><ns0:head>Supplementary materials</ns0:head><ns0:p>Data S1 tRNA sequences of gymnosperms chloroplast genome conducted in the study. Table <ns0:ref type='table'>S1</ns0:ref> Nucleotide composition in different parts of clover-structure of chloroplast genome tRNA. AC-arm, Acceptor arm; ANC-arm, Anti-codon arm; ANC-loop, Anti-codon loop; &#936;-arm, Pseudouridine arm; &#936;-loop, Pseudouridine loop. The base pairs in the AC-arm were counted according to the predicted clover structures of tRNAs. Table <ns0:ref type='table'>S2</ns0:ref> Loss events of chloroplast genomic tRNAs. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>An abnormal tRNA structure lacking the D-arm found in W. nobilis.</ns0:p></ns0:div> <ns0:div><ns0:head>The tRNA</ns0:head><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Clover leaf-like structure of gymnosperms tRNA.</ns0:p><ns0:p>The tRNA contains the Acceptor arm (6-7 bp, dark green, &gt;96% conserved), D-arm (3-4 bp, light blue, &gt;65% conserved), D-loop (7-11 nt, purple, &gt;80% conserved), Anti-codon arm (5 bp, dark blue, &gt;75% conserved), anti-codon loop (7 nt, gray, &gt;99% conserved), variable region (3-23 nt, orange, &gt;45% conserved), &#936;-arm (5 bp, light purple, &gt;97% conserved), and &#936;-loop (7 nt, green, &gt;95% conserved). '% conservation' means the conservative ratio of base identities in each stem and loop structure of the whole set of gymnosperm tRNAs.</ns0:p><ns0:p>Several tRNAs harbor the nucleotides of C-C-A tail. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Y. C. Zhong &amp; C. J. Chen (KM459003), Dioon spinulosum Dyer ex Eichler (NC_027512), Ginkgo biloba L. (NC_016986), Cedrus deodara (Roxb.) G. Don (NC_014575), Wollemia nobilis W. G. Jones, K. D. Hill &amp; J. M. Allen (NC_027235), Retrophyllum piresii Silba C. N. (KJ017081), Sciadopitys verticillata (Thunb.) Sieb. et Zucc. (NC_029734), Cunninghamia lanceolata (Lamb.) Hook. (NC_021437), Taxus mairei (Lemee et Levl.) Cheng et L. K. Fu (KJ123824), Welwitschia mirabilis Hook.f. (EU342371), Gnetum gnemon L. (KR476377), and Ephedra equisetina Bge. (</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Fig. S1</ns0:head><ns0:label>S1</ns0:label><ns0:figDesc>The phylogenetic tree constructed by 373 tRNAs. Multiple tRNAs are shown by different colors. Different groups are marked by different strings. The phylogenetic clades with low bootstrap replicates were collapsed with 50% cutoff values. Phylogenetic analysis illustrates that Gymnosperm chloroplast tRNA derived from common multiple ancestors.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Fig. S2</ns0:head><ns0:label>S2</ns0:label><ns0:figDesc>Fig.S2The loss and duplication tree. Blue: Duplication events; Gray: Loss events; D: Duplication node; cD: Conditional Duplication node. Fig.S3tRNA gene content in analyzed gymnosperms chloroplast genome. The tRNA genes are shown in the left (top to bottom). Boxes in light green, dark green, and white represent one copy of tRNA genes, two copies of tRNA genes, and the absence of tRNA genes.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,70.87,525.00,567.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>1</ns0:head><ns0:label /><ns0:figDesc>Table 1. A view of the gymnosperms in analysis.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Order</ns0:cell><ns0:cell>Family</ns0:cell><ns0:cell /><ns0:cell>Subfamily</ns0:cell><ns0:cell /><ns0:cell>Genus</ns0:cell><ns0:cell>Species</ns0:cell><ns0:cell cols='2'>NCBI Locus</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cycadales</ns0:cell><ns0:cell cols='2'>Cycadaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Cycas</ns0:cell><ns0:cell>debaoensis</ns0:cell><ns0:cell>KM459003</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Zamiaceae</ns0:cell><ns0:cell>Diooideae</ns0:cell><ns0:cell /><ns0:cell>Dioon</ns0:cell><ns0:cell>spinulosum</ns0:cell><ns0:cell cols='2'>NC_027512</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Ginkgoales</ns0:cell><ns0:cell cols='2'>Ginkgoaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Ginkgo</ns0:cell><ns0:cell>biloba</ns0:cell><ns0:cell cols='2'>NC_016986</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Pinales</ns0:cell><ns0:cell cols='2'>Pinaceae</ns0:cell><ns0:cell>Abieteae</ns0:cell><ns0:cell /><ns0:cell>Cedrus</ns0:cell><ns0:cell>deodara</ns0:cell><ns0:cell cols='2'>NC_014575</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Araucariales</ns0:cell><ns0:cell cols='2'>Araucariaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Wollemia</ns0:cell><ns0:cell>nobilis</ns0:cell><ns0:cell cols='2'>NC_027235</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Podocarpaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Retrophyllum</ns0:cell><ns0:cell>piresii</ns0:cell><ns0:cell>KJ017081</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cupressales</ns0:cell><ns0:cell cols='2'>Sciadopityaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Sciadopitys</ns0:cell><ns0:cell>verticillata</ns0:cell><ns0:cell cols='2'>NC_029734</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Cupressaceae</ns0:cell><ns0:cell>Cunninghami a</ns0:cell><ns0:cell /><ns0:cell>Cunninghamia</ns0:cell><ns0:cell>lanceolata</ns0:cell><ns0:cell cols='2'>NC_021437</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Taxaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Taxus</ns0:cell><ns0:cell>mairei</ns0:cell><ns0:cell>KJ123824</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Welwitschiales</ns0:cell><ns0:cell cols='2'>Welwitschiaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Welwitschia</ns0:cell><ns0:cell>mirabilis</ns0:cell><ns0:cell>EU342371</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Gnetales</ns0:cell><ns0:cell cols='2'>Gnetaceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Gnetum</ns0:cell><ns0:cell>gnemon</ns0:cell><ns0:cell>KR476377</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Ephedrales</ns0:cell><ns0:cell cols='2'>Ephedraceae</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Ephedra</ns0:cell><ns0:cell>equisetina</ns0:cell><ns0:cell cols='2'>NC_011954</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Met/fMet</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Tyr</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Cys</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Trp</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Selenocystei ne</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Suppressor</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>28</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>32</ns0:cell><ns0:cell>28</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Distribution of anti-codons in the chloroplast genome of gymnosperms.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Each of the gymnosperm chloroplast genomes encodes 25 to 30 anticodon-specific tRNAs. E.</ns0:cell></ns0:row><ns0:row><ns0:cell>equisetina encodes 25 anticodons, T. mairei encodes 26 anticodons, C. debaoensis, S.</ns0:cell></ns0:row><ns0:row><ns0:cell>verticillata, and C. lanceolata encode 28 anticodons, and D. spinulosum, C. deodara, W.</ns0:cell></ns0:row><ns0:row><ns0:cell>mirabilis, and R. piresii encode 29 anticodons. Other species comprising W. nobilis, G.</ns0:cell></ns0:row><ns0:row><ns0:cell>gnemon and G. biloba encodes 30 anticodons.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:4:0:NEW 5 Oct 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Distribution of anti-codons in the chloroplast genome of gymnosperms.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>tRNA Isotypes</ns0:cell><ns0:cell>Isoacceptors</ns0:cell><ns0:cell /><ns0:cell>tRNA Isotypes</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>Isoacceptors</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>C. debaoensis (28)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='2'>S. verticillata (28)</ns0:cell></ns0:row><ns0:row><ns0:cell>Ala</ns0:cell><ns0:cell>AGC: GGC: 0 CGC: 0 UGC: 1</ns0:cell><ns0:cell /><ns0:cell>Ala</ns0:cell><ns0:cell cols='3'>AGC: 0 GGC: 0 CGC: 0 UGC: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Gly</ns0:cell><ns0:cell>ACC: GCC: 1 CCC: 0 UCC: 0</ns0:cell><ns0:cell /><ns0:cell>Gly</ns0:cell><ns0:cell cols='3'>ACC: 0 GCC: 1 CCC: 0 UCC: 0</ns0:cell></ns0:row><ns0:row><ns0:cell>Pro</ns0:cell><ns0:cell>AGG: GGG: 1 CGG: 0 UGG: 1</ns0:cell><ns0:cell /><ns0:cell>Pro</ns0:cell><ns0:cell cols='3'>AGG: 0 GGG: 0 CGG: 0 UGG: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Thr</ns0:cell><ns0:cell>AGU: GGU: 0 CGU: 0 UGU: 1</ns0:cell><ns0:cell /><ns0:cell>Thr</ns0:cell><ns0:cell>AGU: 0</ns0:cell><ns0:cell>GGU: 1</ns0:cell><ns0:cell>CGU: 0 UGU: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Val</ns0:cell><ns0:cell>AAC: GAC: 1 CAC: 0 UAC: 1</ns0:cell><ns0:cell /><ns0:cell>Val</ns0:cell><ns0:cell cols='3'>AAC: 0 GAC: 1 CAC: 0 UAC: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Ser</ns0:cell><ns0:cell>AGA: GGA: 1 CGA: 0 UGA: 1 ACU: 0</ns0:cell><ns0:cell>GCU: 1</ns0:cell><ns0:cell>Ser</ns0:cell><ns0:cell cols='3'>AGA: 0 GGA: 1 CGA: 0 UGA: 1 ACU: 0 GCU: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Arg</ns0:cell><ns0:cell>ACG: GCG: 0 CCG:1 UCG: 0 CCU: 0</ns0:cell><ns0:cell>UCU: 1</ns0:cell><ns0:cell>Arg</ns0:cell><ns0:cell cols='3'>ACG: 1 GCG: 0 CCG:1 UCG: 0 CCU: 0 UCU: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Leu</ns0:cell><ns0:cell>AAG: GAG: 0 CAG: 0 UAG: 1 CAA: 1</ns0:cell><ns0:cell>UAA: 1</ns0:cell><ns0:cell>Leu</ns0:cell><ns0:cell cols='3'>AAG: 0 GAG: 0 CAG: 0 UAG: 1 CAA: 1 UAA: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Phe</ns0:cell><ns0:cell>AAA: GAA: 1</ns0:cell><ns0:cell /><ns0:cell>Phe</ns0:cell><ns0:cell>AAA: 0</ns0:cell><ns0:cell>GAA: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Asn</ns0:cell><ns0:cell>AUU: GUU: 1</ns0:cell><ns0:cell /><ns0:cell>Asn</ns0:cell><ns0:cell>AUU: 0</ns0:cell><ns0:cell>GUU: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Lys</ns0:cell><ns0:cell>CUU: UUU: 1</ns0:cell><ns0:cell /><ns0:cell>Lys</ns0:cell><ns0:cell>CUU: 0</ns0:cell><ns0:cell>UUU: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Asp</ns0:cell><ns0:cell>AUC: GUC: 1</ns0:cell><ns0:cell /><ns0:cell>Asp</ns0:cell><ns0:cell cols='2'>AUC: 0 GUC: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Glu</ns0:cell><ns0:cell>CUC: UUC: 2</ns0:cell><ns0:cell /><ns0:cell>Glu</ns0:cell><ns0:cell>CUC: 0</ns0:cell><ns0:cell>UUC: 2</ns0:cell></ns0:row><ns0:row><ns0:cell>His</ns0:cell><ns0:cell>AUG: GUG: 1</ns0:cell><ns0:cell /><ns0:cell>His</ns0:cell><ns0:cell>AUG: 0</ns0:cell><ns0:cell>GUG: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Gln</ns0:cell><ns0:cell>CUG: UUG: 1</ns0:cell><ns0:cell /><ns0:cell>Gln</ns0:cell><ns0:cell>CUG: 0</ns0:cell><ns0:cell>UUG: 2</ns0:cell></ns0:row><ns0:row><ns0:cell>Ile</ns0:cell><ns0:cell>AAU: GAU: 0 CAU: 1 UAU: 0</ns0:cell><ns0:cell /><ns0:cell>Ile</ns0:cell><ns0:cell>AAU: 0</ns0:cell><ns0:cell>GAU: 0</ns0:cell><ns0:cell>CAU: 1 UAU: 0</ns0:cell></ns0:row><ns0:row><ns0:cell>Met</ns0:cell><ns0:cell>CAU:</ns0:cell><ns0:cell /><ns0:cell>Met</ns0:cell><ns0:cell>CAU: 2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Tyr</ns0:cell><ns0:cell>AUA: GUA: 1</ns0:cell><ns0:cell /><ns0:cell>Tyr</ns0:cell><ns0:cell>AUA: 0</ns0:cell><ns0:cell>GUA: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Cys</ns0:cell><ns0:cell>ACA: GCA: 1</ns0:cell><ns0:cell /><ns0:cell>Cys</ns0:cell><ns0:cell cols='2'>ACA: 0 GCA: 1</ns0:cell></ns0:row><ns0:row><ns0:cell>Trp</ns0:cell><ns0:cell>CCA:</ns0:cell><ns0:cell /><ns0:cell>Trp</ns0:cell><ns0:cell>CCA: 1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Supressor</ns0:cell><ns0:cell>CUA: UUA: 0 UCA: 0</ns0:cell><ns0:cell /><ns0:cell>Supressor</ns0:cell><ns0:cell>CUA: 0</ns0:cell><ns0:cell>UUA: 0</ns0:cell><ns0:cell>UCA: 0</ns0:cell></ns0:row><ns0:row><ns0:cell>Sec</ns0:cell><ns0:cell>UCA:</ns0:cell><ns0:cell /><ns0:cell>Sec</ns0:cell><ns0:cell>UCA: 0</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2019:11:43190:4:0:NEW 5 Oct 2020) Manuscript to be reviewed D. spinulosum (29) PeerJ reviewing PDF | (2019:11:43190:4:0:NEW 5 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Transition and transversion rate of chloroplast tRNA.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Alanine</ns0:cell></ns0:row></ns0:table><ns0:note>1 PeerJ reviewing PDF | (2019:11:43190:4:0:NEW 5 Oct 2020)Manuscript to be reviewed</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2019:11:43190:4:0:NEW 5 Oct 2020)</ns0:note> </ns0:body> "

"Dear Editor, Thank you so much for returning our manuscript and for your time and patience. Your suggestions and reviewer’s comments are professional, and they are all precious and very helpful for improving our MS. In the revised manuscript, we have perused all these suggestions and have made thoroughly corrections, and responded to the comments as itemized below point by point (our responses are in green characters). If you have any questions regarding the manuscript, please feel free to contact the corresponding author: Zhong-Hu Li Key Laboratory of Resource Biology and Biotechnology in Western China (Ministry of Education), College of Life Sciences, Northwest University, Xi’an 710069, China Fax: +86 29 888302411; E-mail: [email protected] We shall look forward to hearing from you at your earliest convenience. Yours sincerely, Zhonghu Li, Ph. D E-mail: [email protected] Editor and Reviewer comments: Editor comments (Genlou Sun) MINOR REVISIONS Please make changes according to the comments from reviewer. Reply: Thank you very much for your kindly comments. We are so grateful to you for providing us opportunity for revising the manuscript. According to these helpful suggestions and comments, we have carefully and thoroughly revised the manuscript and hope to this version is suitable to the journal. Reviewer 2 (Anonymous) Basic reporting Three minor points should be changed before publication. Reply: Thank you very much for your valuable comments. In the revised manuscript, we have checked carefully and improved these points accordingly. a) line 316: If I understand it correctly the sentence should be changed to: 'According to previous studies (Treangen & Rocha, 2011; Mohanta et al., 2019), it is likely that the deficiency of these tRNAs is compensated for by the transfer of corresponding tRNAs from the nucleus or mitochondria.' Reply: Thank you very much for your kindly suggestion. In line 316, we have rewritten the sentence “According to previous studies (Treangen & Rocha, 2011; Mohanta et al., 2019), it is likely that the deficiency of these tRNAs is compensated for by the tRNAs transferred from other organelle genomes such as nuclear genome and mitochondrial genome.” to “According to previous studies (Treangen & Rocha, 2011; Mohanta et al., 2019), it is likely that the deficiency of these tRNAs is compensated for by the transfer of corresponding tRNAs from the nucleus or mitochondria.” according to your helpful proposal. b) in Table 4, first lines, Alanine tRNA: the 3' or 5' should not be moved to a next line; it should remain attached to the sequence; if this is a space problem, the authors may also consider to indicate only 5'- or 3'-end. Reply: Thanks so much for your helpful advice. In Table 4, the reason why 3' or 5' appeared in a next line is that there is a space problem. And we have improved this content by indicating only 5'- or 3'-end according to your helpful suggestions. c) Fig. 4: in the legend to the figure, the authors should explain more clearly what '% conservation' means. If you consider the whole set of gymnosperm tRNAs, there cannot be > 95% conservation (green) at all positions in the acceptor stem and T arm; or is this the conservation you find for the individual gymnosperm tRNA isotypes? Please clarify. Reply: We strongly agree with your thoughtful view. In Fig. 4, '% conservation' means the conservative ratio of base identities in each stem and loop of the whole set of gymnosperm tRNAs. And we have modified the legend of Fig. 4 by supplementing the explanation of conservative ratio. We have also checked carefully and recalculated the conservative ratio of the acceptor stem and T arm. And we have corrected the conservative ratio of the acceptor stem and T arm to > 96% and > 97% in Fig. 4. Thank you very much for it. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Recent discussions in the sport and exercise science community have focused on the appropriate use and reporting of effect sizes. Sport and exercise scientists often analyze repeated-measures data, from which mean differences are reported. To aid the interpretation of these data, standardized mean differences (SMD) are commonly reported as description of effect size. The use of SMDs has been criticized by sport and exercise scientists as well as those in other disciplines. However, we believe, when thoughtfully applied, SMDs can be a useful analytical tool for sport and exercise scientists. In this manuscript, we hope to alleviate some confusion by demonstrating how effect sizes-primarily SMDs-are calculated, and describing their statistical properties. Finally, we provide high-level recommendations for how sport and exercise scientists can thoughtfully report raw effect sizes, SMDs, or other effect sizes for their own studies.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Effect sizes are a family of descriptive statistics used to communicate the magnitude or strength of a quantitative research finding. Many forms of effect sizes exist, ranging from mean raw values to correlation coefficients. In sport and exercise science, a standardized mean difference (SMD) is commonly reported in studies that observe changes from pre-to post-intervention, and for which units may vary from study-to-study (e.g., muscle thickness vs. cross-sectional area vs. volume). Put simply, an SMD is any mean difference or change score that is divided, hence standardized, by a standard deviation or combination of standard deviations. Thus, even among SMDs, there exist multiple calculative approaches <ns0:ref type='bibr' target='#b27'>(Lakens, 2013;</ns0:ref><ns0:ref type='bibr' target='#b3'>Baguley, 2009)</ns0:ref>. A scientist must therefore decide which SMD is most appropriate to report for their particular study, or if to report one at all. In this manuscript, we will exclusively be focusing on SMD calculations for studies involving repeated-measures since this is a common feature of sport and exercise science studies; other study designs (i.e., between-subjects) have already been extensively covered elsewhere <ns0:ref type='bibr' target='#b3'>(Baguley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kelley and Preacher, 2012;</ns0:ref><ns0:ref type='bibr' target='#b21'>Hedges, 2008)</ns0:ref>.</ns0:p><ns0:p>Different forms of SMDs communicate unique information and have distinct statistical properties.</ns0:p><ns0:p>Yet, some authors in sport and exercise science have staunchly advocated for specific SMD calculations, and in doing so, outright rebuke other approaches <ns0:ref type='bibr' target='#b8'>(Dankel and Loenneke, 2018)</ns0:ref>. While we appreciate that previous discussions of effect sizes have brought this important topic to the forefront, we wish to expand on their work by providing a deeper philosophical and mathematical discussion of SMD choice.</ns0:p><ns0:p>In doing so, we suggest that the choice of an SMD should be based on the objective of each study and therefore is likely to vary from study-to-study. Scientists should have the intellectual freedom to choose whatever statistics are needed to appropriately answer their question. Importantly, this freedom should not be encroached on by broad recommendations that ignore the objectives of an individual scientist. To facilitate these reporting decisions, it is imperative to understand what to report and why.</ns0:p><ns0:p>In this paper, we broadly focus on three things to consider when reporting an SMD. First, before choosing an SMD, a scientist must decide if one is necessary. When making this decision, it is prudent to consider arguments for and against reporting SMDs, in addition to why one should be reported. Second, we broadly categorize repeated-measures SMDs into two categories: signal-to-noise and magnitude-based SMDs. This dichotomy provides scientists with a philosophical framework for choosing an SMD. Third, we describe the statistical properties of SMDs, which we believe scientists should try to understand if they are to report them. We relate these perspectives to previous discussions of SMDs, make general recommendations, and conclude by urging scientists to think carefully about what effect sizes they are reporting and why.</ns0:p></ns0:div> <ns0:div><ns0:head>SHOULD I REPORT A STANDARDIZED MEAN DIFFERENCE?</ns0:head><ns0:p>Before reporting an SMD-or any statistic for that matter-a researcher should first ask themselves whether it is necessary or informative. When answering this, one may wish to consider arguments both for and against SMDs, in addition to field standards. Here, we briefly detail these arguments, in addition to SMD reporting within sport and exercise science.</ns0:p></ns0:div> <ns0:div><ns0:head>Proponents and Opponents of Standardized Effect Sizes</ns0:head></ns0:div> <ns0:div><ns0:head>Opponents</ns0:head><ns0:p>Although SMDs may be useful in some contexts, they are far from a panacea. Arguments against the use of SMDs, including those by prominent statisticians, are not uncommon. These arguments should be considered when choosing whether or not to report an SMD. In particular, the evidentiary value of reporting an SMD must be considered relative to the strength of the general arguments against SMDs.</ns0:p><ns0:p>Below, we have briefly summarized some of the major arguments against the use of SMDs.</ns0:p><ns0:p>As far back as 1969, the use of standardized effect sizes-and by proxy, SMDs-has been heavily criticized. The eminent statistician John Tukey stated that 'only bad reasons seem to come to mind' for using correlation coefficients instead of unstandardized regression coefficients to interpret data. To put it simply, scientists should not assume that standardized effect sizes will make comparisons meaningful <ns0:ref type='bibr' target='#b45'>(Tukey, 1969)</ns0:ref>. This same logic can also be applied to qualitative benchmarks (e.g., Cohen's d = 0.2 is 'small'); we believe it is likely that Cohen would also argue against the broad implementation of these arbitrary benchmarks in all areas of research. Similar arguments against the misuse of standardized effect sizes have been echoed elsewhere <ns0:ref type='bibr' target='#b29'>(Lenth, 2001;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kelley and Preacher, 2012;</ns0:ref><ns0:ref type='bibr' target='#b3'>Baguley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b41'>Robinson et al., 2003)</ns0:ref>.</ns0:p><ns0:p>Others have outright argued against the use of standardized effect sizes because they oversimplify the analysis of, and distort the conclusions derived from, data. In epidemiology, <ns0:ref type='bibr' target='#b17'>Greenland et al. (1986)</ns0:ref> provided a damning indictment of the use of standardized coefficients; namely, because they are largely determined by the variance in the sample, which is heavily influenced by the study design. In psychology, <ns0:ref type='bibr' target='#b3'>Baguley (2009)</ns0:ref> offers a similarly bleak view of standardized effect sizes. He argues that the advantages of standardized effect sizes are far outweighed by the difficulties that arise from the standardization process.</ns0:p><ns0:p>In particular, scientists tend to ignore the impact of reliability and range restriction on effect size estimates, in turn overestimating the generalizability of standardized effect sizes to wider populations and other study designs <ns0:ref type='bibr' target='#b3'>(Baguley, 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Proponents</ns0:head><ns0:p>Conversely, prominent statisticians have also argued in favor of standardized effect sizes, especially for facilitating meta-analysis <ns0:ref type='bibr' target='#b21'>(Hedges, 2008)</ns0:ref>. <ns0:ref type='bibr' target='#b7'>Cohen (1977)</ns0:ref> was the first to suggest the use of standardized effect sizes to be useful for power analysis purposes. This is because, unlike the t-statistic, (bias-corrected) standardized effect sizes are not dependent on the sample size. Similarly, while p-values indicate the compatibility of data with some test hypothesis (e.g., the null hypothesis) <ns0:ref type='bibr' target='#b16'>(Greenland, 2019)</ns0:ref>, SMDs provide information about the 'effect' itself <ns0:ref type='bibr' target='#b39'>(Rhea, 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Thomas et al., 1991)</ns0:ref>. Thus, p-values and t-statistics provide information about the estimate of the mean relative to some test hypothesis and thus are sensitive to sample size, while SMDs strictly pertain to the size of the effect and thus are insensitive to sample size. Moreover, any linear transformation of the data will still yield the exact same standardized effect size <ns0:ref type='bibr' target='#b20'>(Hedges, 1981)</ns0:ref>. The scale invariance property of a standardized effect size theoretically allows them to be compared across studies, various outcomes, and incorporated into a meta-analysis. Therefore,</ns0:p></ns0:div> <ns0:div><ns0:head>2/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed scientists can measure a phenomenon across many different scales or measurement tools and standardized effect sizes should, in theory, be unaffected. Finally, SMDs can provide a simple way to communicate the overlap of two distributions (https://rpsychologist.com/d3/cohend/).</ns0:p></ns0:div> <ns0:div><ns0:head>Comments on Standardized Mean Differences in Sport and Exercise Science</ns0:head><ns0:p>Sport and exercise scientists have also commented on the use of standardized effect sizes <ns0:ref type='bibr' target='#b9'>(Dankel et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b8'>Dankel and Loenneke, 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Rhea, 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Thomas et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b12'>Flanagan, 2013)</ns0:ref>. The discussion has focused on the need to report more than just p-values, emphasizing that scientists have to discuss the magnitude of their observed effects. <ns0:ref type='bibr' target='#b39'>Rhea (2004)</ns0:ref> also provided new benchmarks for SMDs specific to strength and conditioning research, which is certainly an improvement from just using Cohen's benchmarks.</ns0:p><ns0:p>If SMDs are to be reported, they should not be done so in lieu of understanding effects on their natural scales. To this end, we agree with the laments of <ns0:ref type='bibr' target='#b45'>Tukey (1969)</ns0:ref>: too often, standardized effects sizes, particularly SMDs, are relied upon to provide a crutch for interpreting the meaningfulness of results.</ns0:p><ns0:p>Default and arbitrary scales, such as 'small' or 'large' based on those proposed by <ns0:ref type='bibr' target='#b7'>Cohen (1977)</ns0:ref>, should generally be avoided. SMDs should be interpreted on a scale calibrated to the outcome of interest. For example, <ns0:ref type='bibr' target='#b39'>Rhea (2004)</ns0:ref> or <ns0:ref type='bibr' target='#b37'>Quintana (2016)</ns0:ref> have demonstrated how to develop scales of magnitude for a specific area of research. When possible, it is best practice to interpret the meaningfulness of effects in their raw units, and in the context of the population and the research question being asked. For example, a 5 mmHg decrease in systolic blood pressure may be hugely important or trivial, depending on the context-here, the SMD alone cannot communicate clinical relevance.</ns0:p><ns0:p>In our opinion, standardized effect sizes can be useful tools for interpreting data when thoughtfully employed by the scientists reporting them. However, sport and exercise scientists should be careful when selecting the appropriate SMD or effect size, and ensure that their choice effectively communicates the effect of interest <ns0:ref type='bibr' target='#b19'>(Hanel and Mehler, 2019)</ns0:ref>. Herein, we will discuss things to consider when reporting an SMD, and we will close by providing general recommendations and examples that we believe sport and exercise scientists will find useful.</ns0:p></ns0:div> <ns0:div><ns0:head>WHICH STANDARDIZED MEAN DIFFERENCE SHOULD I REPORT?</ns0:head><ns0:p>To facilitate a fruitful discussion of SMDs, here, we categorize them based on the information they convey.</ns0:p><ns0:p>We contend that there are two primary categories of SMDs that sport and exercise scientists will encounter in the literature and use for their own analyses. The first helps to communicate the magnitude of an effect (magnitude-based SMD), and the second is more related to the probability that a randomly selected individual experiences a positive or negative effect (signal-to-noise SMD). These categories serve distinct purposes, and they should be used in accordance with the information a scientist is trying to convey to the reader. We will contrast these SMD categories in terms of the information that they communicate and when scientists may wish to choose one over the other. In doing so, we will show that both approaches to calculating the SMD are distinctly valuable. Finally, we demonstrate that, when paired with background information and other statistics-whether they be descriptive or inferential-each SMD can assist in telling a unique, meaningful story about the reported data.</ns0:p></ns0:div> <ns0:div><ns0:head>Signal-to-noise Standardized Mean Difference</ns0:head><ns0:p>The first category of SMDs can be considered a signal-to-noise metric: it communicates the average change score in a sample relative to the variability in change scores. This is called Cohen's d z , and it is an entirely appropriate way to describe the change scores in paired data. The Z subscript refers to the difference being compared is no longer between the measurements (X or Y ) but the difference (Z = Y &#8722; X).</ns0:p><ns0:p>This SMD is directly estimating the change standardized to the variation in this response, making it a mathematically natural signal-to-noise statistic.</ns0:p><ns0:p>Cohen's d z can be calculated with the mean change, &#948; , and the standard deviation of the difference, &#963; &#948; ,</ns0:p><ns0:formula xml:id='formula_0'>d z = &#948; &#963; &#948; .<ns0:label>(1)</ns0:label></ns0:formula><ns0:p>Alternatively, for convenience, can be calculated from the t-statistic and the number of pairs (n),</ns0:p><ns0:formula xml:id='formula_1'>d z = t &#8730; n .<ns0:label>(2)</ns0:label></ns0:formula><ns0:p>In Eq. 2, one can see that d z is closely related to the t-statistic. Specifically, the t-statistic is a signal-tonoise metric for the mean (i.e., using its sampling distribution), while d z is a signal-to-noise metric for the entire sample. This means that the t-statistic will tend to increase with increases in sample size, since the estimate of the mean becomes more precise, while (bias-corrected) Cohen's d z will not change with sample size.</ns0:p><ns0:p>Although Cohen's d z may be useful to describe the change in a standardized form, it is typically not reported in meta-analyses since it cannot be used to compare differences across between-and withinsubjects designs (see SMDs below). It is difficult to interpret the value of this type of SMD; that is, since the signal-to-noise ratio itself is more related to the consistency of a change, one can wonder how much consistency constitutes a 'large' effect? This is in contrast to other types of SMDs, wherein the statistic conveys information about the distance between two central tendencies (mean) relative to the dispersion of the data (standard deviation). Moreover, it appears that, to sport and exercise scientists, the value of this SMD is measuring the degree of the change in comparison to the variability of the change scores <ns0:ref type='bibr' target='#b8'>(Dankel and Loenneke, 2018)</ns0:ref>. Therefore, scientists' intent on using d z should consider reporting the common language effect size (CLES) <ns0:ref type='bibr' target='#b32'>(McGraw and Wong, 1992)</ns0:ref>, also known as the probability of superiority <ns0:ref type='bibr' target='#b18'>(Grissom, 1994)</ns0:ref>. In contrast to d z , CLES communicates the probability of a positive (CLES &gt; 0.5) or negative (CLES &lt; 0.5) change occurring in a randomly sampled individual (see below).</ns0:p></ns0:div> <ns0:div><ns0:head>Alternative to the signal-to-noise Standardized Mean Difference</ns0:head><ns0:p>The information gleaned from the signal-to-noise SMD (Cohen's d z ) can also be captured with the CLES <ns0:ref type='bibr' target='#b32'>(McGraw and Wong, 1992;</ns0:ref><ns0:ref type='bibr' target='#b18'>Grissom, 1994)</ns0:ref>. In paired samples, the CLES conveys the probability of a randomly selected person's change score being greater than zero. The CLES is easy to obtain; it is simply the Cohen's d z (SMD) converted to a probability (CLES = &#934;(d z ), where &#934; is the standard normal cumulative distribution function). Importantly, CLES can be converted back to a Cohen's d z with the inverse normal cumulative distribution function (d z = &#934; &#8722;1 (CLES)). CLES is particularly useful because it directly conveys the direction and variability of change scores without suggesting that the mean difference itself is small or large. Further, current evidence would suggest that the CLES is easier for readers to comprehend than a signal-to-noise SMD <ns0:ref type='bibr' target='#b19'>(Hanel and Mehler, 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Magnitude-based Standardized Mean Difference</ns0:head><ns0:p>The second category of SMDs can be considered a magnitude-based metric: it communicates the size of an observed effect relative to spread of the sample. The simplest and most understood magnitude-based SMD is Glass's &#8710;, which is used to compare two groups, and is standardized to the standard deviation of one of the groups. However, a conceptually similar version of Glass's &#8710;, which we term Glass's &#8710; pre , can also be employed for repeated-measures. In &#8710; pre the mean change change is standardized by the pre-intervention standard deviation. 1 For basic pre-post study designs, Glass's &#8710; pre is fairly straightforward; mean change is simply standardized to the standard deviation of the pre-test responses. There are other effect sizes for repeated measures designs such as Cohen's d av and d rm , but for brevity's sake these are described in the appendix. Of note, &#8710; pre , d av , and d rm are identical when pre-and post-intervention variances are the same (see Appendix).</ns0:p><ns0:formula xml:id='formula_2'>&#8710; pre = &#948; &#963; pre (3)</ns0:formula><ns0:p>Importantly, &#8710; pre is well-described <ns0:ref type='bibr' target='#b36'>(Morris and DeShon, 2002;</ns0:ref><ns0:ref type='bibr' target='#b34'>Morris, 2000;</ns0:ref><ns0:ref type='bibr' target='#b4'>Becker, 1988)</ns0:ref> and can also be generalized to parallel-group designs; in particular, when there are 2 groups, typically a control and treatment group, being compared over repeated-measurements <ns0:ref type='bibr' target='#b35'>(Morris, 2008)</ns0:ref>. Typically, in these cases, a treatment and control group are being directly compared in a 'pretest-posttest-control design' (PPC). A simple version of the PPC-adapted &#8710; pre is</ns0:p><ns0:formula xml:id='formula_3'>&#8710; ppc = &#8710; T &#8722; &#8710; C (4)</ns0:formula><ns0:p>where &#8710; T and &#8710; C are the &#8710; pre from the treatment and control groups, respectively. There are several other calculative approaches which should be considered for comparing SMDs in a parallel-group designs. We highly encourage further reading on this topic if this type of design is of interest to readers <ns0:ref type='bibr' target='#b35'>(Morris, 2008;</ns0:ref><ns0:ref type='bibr' target='#b4'>Becker, 1988;</ns0:ref><ns0:ref type='bibr' target='#b47'>Viechtbauer, 2007)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Summary of Standardized Mean Differences</ns0:head><ns0:p>Our distinction between signal-to-noise (namely, Cohen's d z ) and magnitude-based SMDs (including Glass's &#8710; pre , Cohen's d av , and Cohen's d rm ) provides a conceptual dichotomy to assist researchers in picking an SMD (summarized in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). However, along with the conceptual distinctions, researchers should also consider the the properties of these SMDs. In the following section, we briefly go over the math underlying each SMD and its implications. The properties that follow from the math complement the conceptual framework we just presented, in turn providing researchers with a theoretical, mathematical basis for choosing and justifying their choice of an SMD. </ns0:p></ns0:div> <ns0:div><ns0:head>WHAT ARE THE STATISTICAL PROPERTIES OF STANDARDIZED MEAN DIFFERENCES?</ns0:head><ns0:p>An SMD is an estimator. Estimators, including SMDs, have basic statistical properties associated with them that can be derived mathematically. From a high level, grasping how an estimator behaveswhat makes it increase or decrease and to what extent-is essential for interpretation. In addition, one should have a general understanding of the statistical properties of an estimator they are using; namely, its bias and variance, which together determine the accuracy of the estimator (mean squared error, MSE = Bias( &#952; , &#952; ) 2 + Var &#952; ( &#952; ), for some true parameter, &#952; , and its estimate, &#952; ). These properties depend on the arguments used in the estimator. As a result, signal-to-noise and magnitude-based SMDs are not only distinct in terms of their interpretation, but also their statistical properties. Although these properties have been derived elsewhere (e.g., <ns0:ref type='bibr' target='#b20'>Hedges (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b36'>Morris and DeShon (2002)</ns0:ref>; <ns0:ref type='bibr' target='#b34'>Morris (2000)</ns0:ref>; <ns0:ref type='bibr' target='#b13'>Gibbons et al. (1993)</ns0:ref>; <ns0:ref type='bibr' target='#b4'>Becker (1988)</ns0:ref>), their implications are worth repeating. In particular, there are several salient distinctions between the properties of each of these metrics, which we will address herein. Although this section is more technical, we will return to a higher-level discussion of SMDs in the next section.</ns0:p></ns0:div> <ns0:div><ns0:head>Estimator Components</ns0:head><ns0:p>Before discussing bias and variance, we will briefly discuss the components of the formulae and their implications. Of course, all SMDs contain the mean change score, &#948; , in the numerator, and thus increase linearly with mean change (all else held equal). Since this is common to all SMDs, we will not discuss it further.</ns0:p><ns0:p>More interestingly, the signal-to-noise and magnitude-based SMDs contain very different denominators. To simplify matters, let us assume the pre-and post-intervention standard deviations are equal (&#963; pre = &#963; post = &#963; ). This assumption is reasonable since pre-and post-intervention standard deviations typically do not substantially differ in sports and exercise science. In this case, the standard deviation of change scores can be found simply:</ns0:p><ns0:formula xml:id='formula_4'>&#963; &#948; = 2&#963; 2 (1 &#8722; r).</ns0:formula><ns0:p>With these assumptions, d rm = d av = &#8710; pre for &#8722;1 &lt; r &lt; 1, where r is the observed pre-post correlation (Appendix). Greater pre-post correlations, r, are indicative of more homogeneous change scores. This makes the behavior of the magnitude-based SMDs fairly straightforward; that is, the estimates themselves will not be affected by the correlation between pre-and post-intervention scores. Their dependence on &#963; means that the magnitude-based SMD will blow up as &#963; &#8594; 0. This is in contrast to d z , whose denominator contains both &#963; and r (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref> (Bottom) The standard error of each estimator increases as &#963; pre &#8594; 0. Importantly, &#8710; pre has lower or similar standard errors as r &#8594; 1, whereas d z has greater standard errors as r &#8594; 1. Additional simulations, including those of other SMDs, can be found in online supplemental material https://www.doi.org/10.17605/OSF.IO/FC5XW sample. On the other hand, when breaking d z down into its constituent parts, it depends on the mean change score, the spread of scores in the sample, and the correlation between pre-and post-intervention scores-the latter two will create &#963; &#948; . These sensitivities should be understood before implementing an SMD.</ns0:p></ns0:div> <ns0:div><ns0:head>Bias</ns0:head><ns0:p>Bias means that, on average, the estimate of a parameter ( &#952; ) differs from the 'true' parameter being estimated (&#952; ). Most SMDs follow a non-central t-distribution, allowing the bias to be easily assessed and corrected. As shown by <ns0:ref type='bibr' target='#b20'>Hedges (1981)</ns0:ref>, SMDs are generally biased upwards with small sample sizes; that is, with smaller samples, SMDs are overestimates of the true underlying SMD ( &#952; &gt; &#952; ). This bias is a function of both the value of the SMD obtained and the sample size:</ns0:p><ns0:formula xml:id='formula_5'>E[d] = d = d c(n &#8722; 1) (5) =&#8658; Bias[ d, d] = d &#8722; d = d ( 1 /c(n&#8722;1) &#8722; 1) , (<ns0:label>6</ns0:label></ns0:formula><ns0:formula xml:id='formula_6'>)</ns0:formula><ns0:p>where d is the 'true' parameter being estimated, d is its estimate, and c(m) = 1 &#8722; 3 4(m)&#8722;1 is Hedges' bias-correction factor <ns0:ref type='bibr' target='#b20'>(Hedges, 1981)</ns0:ref> and m = n &#8722; 1 is the degrees of freedom for a paired sample. Please note that this degrees of freedom will differ for different study designs and standard deviations. For example, with two groups and a pooled standard deviation, m = n 1 + n 2 &#8722; 2. We have noticed the incorrect use of degrees of freedom in some published papers within sport and exercise science, so we urge authors to be cautious.</ns0:p><ns0:p>Because SMDs are biased, especially in small samples, it is advisable to correct for this bias. Thus, when using Cohen's d in small sample settings, most sport and exercise scientists should apply a Hedges' correction to adjust for bias. A bias-corrected d is typically referred to as Hedges' g:</ns0:p><ns0:formula xml:id='formula_7'>g = d &#8226; c(n &#8722; 1),<ns0:label>(7)</ns0:label></ns0:formula></ns0:div> <ns0:div><ns0:head>6/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed where d can represent any of the SMD estimates outlined above. This correction decreases the SMD by about 10 and 5% with 10 and 15 participants, respectively; corrections are negligible with larger sample sizes. Bias correction can also be applied via bootstrapping <ns0:ref type='bibr' target='#b42'>(Rousselet and Wilcox, 2019)</ns0:ref>.</ns0:p><ns0:p>More generally, we stress to readers that bias per se is not a bad thing or undesirable property.</ns0:p><ns0:p>Especially in multidimensional cases, bias can improve the accuracy of an estimate by decreasing its variance-this is known as Stein's paradox <ns0:ref type='bibr' target='#b11'>(Efron and Morris, 1977)</ns0:ref>. Indeed, biased (shrunken) estimators of SMDs have been suggested which may decrease MSE <ns0:ref type='bibr' target='#b22'>(Hedges and Olkin, 1985)</ns0:ref>. However, these are not commonly employed. Having said this, the upward bias of SMDs is generally a bad thing. As will be discussed in the next subsection, by correcting for the upward bias, we also improve (decrease) the variance of the SMD estimate, in turn decreasing MSE via both bias and variance <ns0:ref type='bibr' target='#b20'>(Hedges, 1981;</ns0:ref><ns0:ref type='bibr' target='#b22'>Hedges and Olkin, 1985)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Variance</ns0:head><ns0:p>While bias tells us about the extent to which an estimator over-or underestimates the value of a true parameter, variance tells us how variable the estimator is. Estimators that are more precise (less variable)</ns0:p><ns0:p>will have tighter standard errors and thus confidence intervals, allowing us to make better judgments as to the 'true' magnitude of the SMD.</ns0:p><ns0:p>By looking at formulae for variance and its arguments, we can gain a better understanding of what affects its statistical properties. Below are the variance formulae for Cohen's d z and Glass's &#8710; pre , which are the two best understood SMDs for paired designs <ns0:ref type='bibr' target='#b4'>(Becker, 1988;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gibbons et al., 1993;</ns0:ref><ns0:ref type='bibr' target='#b15'>Goulet-Pelletier and Cousineau, 2018;</ns0:ref><ns0:ref type='bibr' target='#b34'>Morris, 2000;</ns0:ref><ns0:ref type='bibr' target='#b36'>Morris and DeShon, 2002)</ns0:ref>.</ns0:p><ns0:formula xml:id='formula_8'>Var[d z ] = n &#8722; 1 n(n &#8722; 3) (1 + d 2 z n) &#8722; d 2 z c(n &#8722; 1) 2 (8) Var[&#8710; pre ] = n &#8722; 1 n(n &#8722; 3) 2(1 &#8722; r) + &#8710; 2 pre n &#8722; &#8710; 2 pre c(n &#8722; 1) 2<ns0:label>(9)</ns0:label></ns0:formula><ns0:p>Variances for the biased SMDs (above) can be easily converted to variances for the bias-corrected SMDs by multiplying each formula by c(n&#8722;1) 2 , which is guaranteed to decrease variance since c(&#8226;) &lt; 1 <ns0:ref type='bibr' target='#b20'>(Hedges, 1981)</ns0:ref>.</ns0:p><ns0:p>Each variance formula contains the SMD itself, meaning that variance will tend to increase with an increasing SMD. This also complicates matters for d z ; since &#963; &#948; can increase from a smaller &#963; pre or greater r, d z 's variance explodes with homogeneous populations or change scores (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Such a quality is not very desirable, as typically, we would like more precision as effects become more homogeneous; this property is a further indication that d z is not a measure of effect magnitude. This is in contrast to the magnitude-based SMDs, which become more precise as the effect becomes more homogeneous (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Of note, these differences in variance behaviors do not reflect differences in statistical efficiency; after adjusting for scaling, all are unbiased and equally efficient.</ns0:p><ns0:p>By investigating and understanding the statistical properties of a statistic-here, the SMDs-we can gain a better understanding of what we should and should not expect from an estimate. These properties provide us with an intuitive feel for the implications of the mathematical machinery underlying each SMD, in turn helping us choose and justify an SMD.</ns0:p></ns0:div> <ns0:div><ns0:head>Considering Previous Arguments for Signal-to-noise Standardized Mean Differences</ns0:head><ns0:p>There have been arguments against SMDs-at least certain calculative approaches-with one particular article claiming that magnitude-based SMDs are flawed <ns0:ref type='bibr' target='#b8'>(Dankel and Loenneke, 2018)</ns0:ref>. Specifically, <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> profess the superiority of Cohen's d z over magnitude-based SMDsspecifically, Glass's &#8710; pre -because of its statistical properties 2 and its relationship with the t-statistic.</ns0:p><ns0:p>Regarding the former, <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> opine that the magnitude-based SMD is 'dependent 2 <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> suggest that, '... normalizing effect size values to the pre-test SD will enable the calculation of a confidence interval before the intervention is even completed ... This again also points to the flaws of normalizing effect sizes to the pretest SD because the magnitude of the effect ... is dependent on the individuals recruited rather than the actual effectiveness of the intervention' (p. 4). This is of course not the case, since the variance of the SMD will depend on, among other things, the change scores themselves. Thus, the confidence interval of the magnitude-based SMD estimate cannot be calculated a priori.</ns0:p></ns0:div> <ns0:div><ns0:head>7/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed on the individuals recruited rather than the actual effectiveness of the intervention.' We do not find this to be a compelling argument against magnitude-based SMDs for several reasons. First, it is in no way specific to magnitude-based SMDs; all descriptive statistics are always specific to the sample. Second, if the data are randomly sampled (a necessary condition for valid statistical inference), then the sample should, on average, be representative of the target population. If imbalance in some relevant covariate is a concern, then an analysis of covariance, and the effect size estimate from this statistical model, should be utilized <ns0:ref type='bibr' target='#b40'>(Riley et al., 2013)</ns0:ref>.</ns0:p><ns0:p>It is certainly the case that Cohen's d z has a natural relationship with the t-statistic. Stemming from this relationship, <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> suggest that it is a more appropriate effect size statistic for repeated-measures designs. Although it is true that d z is closely related to the t-statistic, this does not imply that d z is the most appropriate SMD to report. First, the t-statistic and degrees of freedom (which should be reported) together provide the required information to calculate a Cohen's d z , meaning d z may contain purely redundant information. Second, although Cohen's d z has a clear relationship with the statistical power of a paired t-test, we want to emphasize that utilizing an observed effect size in power analyses is an inappropriate practice. Performing such power analyses to justify sample sizes of future work implicitly assumes that 1) the observed effect size is the true effect size; 2) follow-up studies will require this observed SMD; and 3) this effect size is what is of interest (rather than one based on theory or practical necessity). In most cases, observed effect sizes do not provide accurate estimates of the population-level SMD, and utilizing the observed SMD from a previous study will likely lead to an underpowered follow-up study <ns0:ref type='bibr' target='#b0'>(Albers and Lakens, 2018)</ns0:ref>, and moreover, relying on previously reported effect sizes ignores the potential heterogeneity of observed effect sizes between studies (McShane and B&#246;ckenholt, 2014). Rather, there exist alternative approaches to justifying sample sizes (Appendix 2).</ns0:p><ns0:p>In general, and in contrast to <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref>, we believe that SMDs can be used for different purposes-whether to communicate the size of an effect, calculate power, or some other purpose-and what is best for one objective is not necessarily what is best for the others. Furthermore, we want to emphasize that these are not arguments against the use of signal-to-noise SMDs, but rather a repudiation of arguments meant to discourage the use of magnitude-based SMD by sport and exercise scientists.</ns0:p></ns0:div> <ns0:div><ns0:head>RECOMMENDATIONS FOR REPORTING EFFECT SIZES</ns0:head><ns0:p>In most cases, sport and exercise scientists are strongly encouraged to present and interpret effect sizes in their raw or unstandardized form. As others previously discussed, journals should require authors to report some form of an effect size, along with interpretations of its magnitude, instead of only reporting p-values <ns0:ref type='bibr' target='#b39'>(Rhea, 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Thomas et al., 1991)</ns0:ref>. However, an SMD, along with other standardized effect sizes, do not magically provide meaning to meaningless values. They are simply a convenient tool that can provide some additional information and may sometimes be helpful to those performing meta-analyses or who are unfamiliar with the reported measures. Specifically, there are situations where the outcome measure may be difficult for readers to intuitively grasp (e.g., a psychological survey, arbitrary units from Western Blots, moments of force). In such cases, a magnitude-based SMD-in which the SD of preand/or post-intervention measures is used in the denominator-can be used to communicate the size of the effect relative to the heterogeneity of the sample. In other words, a magnitude-based SMD represents the expected number of sample SDs (not the change due to the intervention) by which the participants and the second would imply that she moves to the 60th percentile. Clearly, both of these interpretations are wrong. As opposed to a magnitude-based SMD, Cohen's d z is a signal-to-noise statistic that is related to the probability of a randomly sampled individual experiencing an effect rather than its magnitude alone.</ns0:p><ns0:p>In our opinion, Cohen's d z does not provide any more information than that which is communicated by the t-statistic and the associated degrees of freedom (which should be reported regardless of the effect size).</ns0:p><ns0:p>Instead, if the signal-to-noise is of interest, a CLES may provide the information a sport and exercise scientist is interested in presenting. Going back to our earlier example (d z = 11.62 and 0.25 respectively), the CLES would be approximately &gt; 99% and 59.9%, or the probability of a randomly sample individual undergoing an improvement is &gt; 99% or 59.9% for intervention 1 and 2, respectively. As <ns0:ref type='bibr' target='#b19'>Hanel and Mehler (2019)</ns0:ref> demonstrated, the CLES may be a more intuitive description of the signal-to-noise SMD.</ns0:p><ns0:p>While our personal recommendation leans towards the use of magnitude-based SMDs and CLES, it is up to the individual sport and exercise scientist to decide what effect size they feel is most appropriate for the data they are analyzing and point they are trying to communicate <ns0:ref type='bibr' target='#b25'>(H&#246;nekopp et al., 2006)</ns0:ref>.</ns0:p><ns0:p>In choosing an SMD, we also sympathize with <ns0:ref type='bibr' target='#b27'>Lakens (2013)</ns0:ref>, '... to report effect sizes that cannot be calculated from other information in the article, and that are widely used so that most readers should understand them. Because Cohen's d z can be calculated from the t-value and the n, and is not commonly used, my general recommendation is to report Cohen's d av or Cohen's d rm .' Along these same lines, if scientists want to present an SMD, it should not exist in isolation. It is highly unlikely that a single number will represent all data in a meaningful way. We believe that data are often best appreciated when presented in multiple ways. The test and inferential statistics (p-values and t-statistics) should be reported alongside an effect size that provides some type of complementary information. This effect size can be standardized (e.g., &#8710; pre ) or unstandardized (raw), and should be reported with a confidence interval. Confidence intervals (CI) of a magnitude-based SMD will provide readers with information concerning both the magnitude and uncertainty of an effect size; CIs can be calculated using formulae, or perhaps more easily, using the bootstrap. In situations where the measurements are directly interpretable, unstandardized estimates are generally preferable. The CLES can also be reported when the presence of a change or difference between conditions is of interest.</ns0:p></ns0:div> <ns0:div><ns0:head>Percent Changes</ns0:head><ns0:p>It is not uncommon for sport and exercise scientists to report their data using percentages (e.g., percent change). While this is fine if it supplements the reporting of their data in raw units, it can be problematic if it is the only way the data are presented or if the statistics are calculated based on the percentages. In the case of SMDs, an SMD calculated using a percent change is not the same as an SMD calculated using raw units. More importantly, the latter-which is often of greater interest to readers or those performing meta-analysis-cannot be back-calculated from the former. It is imperative that authors consider the properties of the values that they report and what readers can glean from them.</ns0:p></ns0:div> <ns0:div><ns0:head>Data Sharing</ns0:head><ns0:p>To facilitate meta-analysis, we suggest that authors upload their data to a public repository such as the Open Science Framework, FigShare, or Zenodo <ns0:ref type='bibr' target='#b5'>(Borg et al., 2020)</ns0:ref>. This ensures that future metaanalysis or systematic reviews efforts have flexibility in calculating effect sizes since there are multitude of possible calculative approaches, designs, and bias corrections (see <ns0:ref type='bibr' target='#b3'>Baguley (2009)</ns0:ref>). When data sharing is not possible, we highly encourage sport and exercise scientists to upload extremely detailed descriptive statistics as supplementary material (i.e., sample size per group, means, standard deviations, and correlations), or alternatively, a synthetic dataset that mimics the properties of the original <ns0:ref type='bibr' target='#b38'>(Quintana, 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Examples</ns0:head><ns0:p>In the examples below, we have simulated data and analyzed it in R (see supplementary material) to demonstrate how results from a study in sport and exercise science could be interpreted with the appropriate application of SMDs. For those unfamiliar with R, there is an online web application (https://</ns0:p></ns0:div> <ns0:div><ns0:head>9/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed doomlab.shinyapps.io/mote/) and extensive documentation (https://www.aggieerin.</ns0:p><ns0:p>com/shiny-server/introduction/) to simplify the process of calculating SMDs for those without R programming experience <ns0:ref type='bibr' target='#b6'>(Buchanan et al., 2019)</ns0:ref>. Now, let us imagine a study trying to estimate the effect of cold water immersion on muscle soreness. For this hypothetical study, muscle soreness is measured on a visual analog scale before and after cold water immersion following a muscle damaging exercise. The muscle soreness score would be represented by cm on the scale measured left-to-right. Because sensations tend to be distributed lognormal <ns0:ref type='bibr' target='#b30'>(Mansfield, 1974)</ns0:ref>-and are multiplicative rather than additive-it is sensible to work with the logarithm of the reported soreness levels. Since these logged scores are not directly interpretable, it is sensible to use an SMD to help interpret the change scores. The hypothetical study could be written up as follows:</ns0:p><ns0:p>Muscle soreness was lower after cold water immersion (mean = 27, SD = 7) compared to before (mean = 46, SD = 11) cold water immersion, t(9) = -6.90, p &lt; .001, Glass's &#8710; pre = -2.2 95% CI <ns0:ref type='bibr'>[-3.2, 1.3]</ns0:ref>. The CLES indicates that the probability of a randomly selected individual experiencing a reduction in muscle soreness after cold water immersion is 99%.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSION</ns0:head><ns0:p>We contend that the reporting of effect sizes should be specific to the research question in conjunction with the narrative that a scientist wants to convey. In this context, pooled pre-and/or post-study SDs are viable choices for the SMD denominator. This approach provides insight into the magnitude of a given finding, and thus can have important implications for drawing practical inferences. Moreover, the values of this approach are distinct and, in our professional opinion, potentially more insightful than signal-to-noise SMDs, which essentially provide information that is redundant with the t-statistic. At the very least, there is no one-size-fits-all solution to reporting an SMD, or any other statistics for that matter.</ns0:p><ns0:p>Despite our personal preference towards other effect sizes, a sport and exercise scientist may prefer a signal-to-noise SMD (d z ) and could reasonably justify this decision. We urge sport and exercise scientists to avoid reporting the same default effect size and interpreting them based on generalized, arbitrary scales.</ns0:p><ns0:p>Rather, we strongly encourage sport and exercise scientists justify which SMD is most appropriate and provide qualitative (i.e., small, medium, or large effect) interpretations that are specific to that outcome and study design. Also, sport and exercise scientists should be careful to report the rationale for using an SMD over simply presenting raw mean differences. Lastly, the creation of statistical rituals wherein a single statistic, by default, is used to interpret the data is likely to result in poor statistical analyses rather than informative ones <ns0:ref type='bibr' target='#b14'>(Gigerenzer, 2018)</ns0:ref>. As J.M. Hammersley once warned, 'There are no routine statistical questions; only questionable statistical routines' <ns0:ref type='bibr' target='#b43'>(Sundberg, 1994)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>APPENDIX 1: STANDARDIZED MEAN DIFFERENCE CALCULATIVE AP-</ns0:head></ns0:div> <ns0:div><ns0:head>PROACHES</ns0:head><ns0:p>Throughout the text, we use Glass's &#8710; pre as our token magnitude-based SMD. However, there exist other approaches to calculating magnitude-based SMDs. Here, we briefly discuss two other common calculations of magnitude-based SMDs. Of note, these two other calculative approaches may contain some 'effects' (variance) from the intervention in the denominator, arguably making Glass's a more 'pure' (in the sense that the denominator is uncontaminated by intervention effects) magnitude-based SMD.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure1. Standardized mean differences for a range of pre-post correlations and pre-intervention standard deviations. Standardized mean differences (SMD) were calculated for a pre-post design study with 20 participants to depict the different properties of the different SMDs. We calculated SMDs for a range of pre-post correlations (r) and pre-intervention standard deviations (&#963; pre ), each with a mean change score of 1. (Top) Magnitude-based SMDs have similar estimates across the range of pre-post correlations and largely only vary as a function of &#963; pre , whereas signal-to-noise SMDs are a function of both &#963; pre and r. Note, d z blows up as r &#8594; 1, and all SMDs blow up as &#963; pre &#8594; 0. (Bottom) The standard error of each estimator increases as &#963; pre &#8594; 0. Importantly, &#8710; pre has lower or similar standard errors as r &#8594; 1, whereas d z has greater standard errors as r &#8594; 1. Additional simulations, including those of other SMDs, can be found in online supplemental material https://www.doi.org/10.17605/OSF.IO/FC5XW</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>examples presented previously in the sport and exercise science literature. The examples presented in Figure 1 by Dankel and Loenneke (2018), in which both interventions have a preintervention SD pre = 6.05 and undergo a change of &#916;= 3.0 (SMD = 3.0 6.05 = 0.5). This can be interpreted simply: the expected change is 0.5 standard deviation units relative to the measure in the sample. Put differently, if the person with the median score (50th percentile) were to improve by the expected change, she would move to the 69th percentile. 3 Like a mean change, this statistic is not intended to provide information about the variability of change scores. The magnitude-based SMD simply provides a unitless, interpretable value that indicates the magnitude of the expected change relative to the between-subject standard deviation. Of course, it can be complemented with a standard error or confidence interval if one is interested in the uncertainty around this estimate. The above can be contrasted with Cohen's d z , which uses the SD of change scores. Again, using the examples presented in Figure 1 of Dankel and Loenneke (2018), Cohen's d z of 11.62 and 0.25 are reported for interventions 1 and 2, respectively. If one tries to interpret these SMDs in a way that magnitude-based SMDs are interpreted, he will undoubtedly come to incorrect conclusions. The first would suggest that a person with the median score who experiences the expected change would move to &gt;99.99th percentile,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Scenario 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Interpretable Raw Differences In the first hypothetical example, let us imagine a study trying to estimate the change in maximal oxygen consumption ( VO 2 ; L&#8226;min -1 ) in long-distance track athletes before and after a season of training. For this study, maximal VO 2 was measured during a Bruce protocol with a Parvomedics 2400 TrueOne Metabolic System. The results of this hypothetical outcome could be written up as the following: VO 2 after a season of training with the track team (mean = 4.13 L&#8226;min -1 , SD = 0.25) increased compared to when they joined the team (M = 3.89 L&#8226;min -1 , SD = 0.21), t(7) = 3.54, p = 0.009, &#948; = 0.23 L&#8226;min -1 95% C.I. [0.07, 0.38]. The CLES indicates that the probability of a randomly selected individual's V O 2 increasing after their first season with the team is 89%. Scenario 2: Uninterpretable Raw Differences</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Types of Standardized Mean Differences for Pre-Post Designs Magnitude-based Glass's &#8710; pre , Cohen's d av , Cohen's d rm Signal-to-noise Cohen's d z</ns0:figDesc><ns0:table /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:note> <ns0:note place='foot' n='1'>Although conceptually similar, Glass's &#8710; and &#8710; pre have different distributional properties<ns0:ref type='bibr' target='#b4'>(Becker, 1988)</ns0:ref>.4/13PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:note> <ns0:note place='foot' n='3'>Assumes a normal distribution. We note that<ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> data vignettes are approximately uniformly distributed which is an odd assumption to make about theoretical data, but nonetheless, sufficiently conveys the point.8/13PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:note> <ns0:note place='foot' n='10'>/13 PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)</ns0:note> <ns0:note place='foot' n='13'>/13 PeerJ reviewing PDF | (2020:05:49464:1:2:NEW 8 Sep 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"We thank the reviewers for their time, effort, and feedback. We apologize for the length of time it has taken us to respond. Since submitting our manuscript we have received comments from the statisticians and exercise scientists (many noted in the acknowledgements). Therefore, we have made extensive changes to the manuscript in order to enhance the readability of the manuscript, and address both the reviewers and colleagues concerns. We would like to highlight some of the major changes we have made to the manuscript since receiving our reviews. 1. The introduction has been extensively edited to help improve readability. 2. The manuscript now presents two SMDs (δpre and dz as the primary examples of our two types of SMDs (others have been included in a appendix. 3. The Figure in the estimator section now shows the mathematical behavior (rather than simulations) of the two SMDs. Simulations are now included in the supplement. 4. The interim summary section has been combined with the ”conventions in exercise and sport science” section in order to improve the flow of this section into ”choosing an effect size” section. 5. More details and suggestions have been added to the ”Recommendations” section. 6. In order to avoid confusion, we have limited our criticisms of Dankel Leonneke to one specific section of the manuscript. Below, we address each of the points of the reviewers. 1 1.1 Reviewer: Ivan Jukic Basic reporting Line 29 - Please replace “observing” with “observe”. Changed. Proponents section - While the entire sub-section is well presented and already contains valuable information and a nice small overview, I think that it would also be beneficial to include a very brief discussion regarding p values and moving away from binary thinking regarding the differences between the xy and z, for instance. I saw that you mentioned it in the “Effect size conventions in exercise and sport science” sub-section but I believe that it would add to the story here as well. We politely disagree that SMDs play such a role in moving away from dichotomizations. However, we have rewritten this section and mentioned in the recommendations section that journals should enforce policies that help encourage the reporting of effect sizes. Interim section - While I agree with your summary here, don’t you think that it would also be beneficial for our field to consider smallest detectable changes/differences for their outcomes 1 before attempting to answer their research question with a study and then quantify the size of the effect based on how much it exceeds (or how much it doesn’t exceed) a previously defined change that should be considered relevant/important for practitioners/clinicians in our field. In this way, qualitative descriptors would not be as important and the readers would be “forced” to interpret the results in a way that relates to their own practice. This would generally be a push toward interpreting the data on an individual basis (i.e., what’s meaningful for me, for my context, as it might (or should) differ from one scenario/situation/context etc. to the next). We (ARC and ADV) in fact have differing views on this topic, so we tried to incorporate both perspectives. ARC is of the opinion that SESOI is a pragmatic approach to thinking through your problem, measures, and justifying your sample size, while ADV is of the opinion that it is ontologically baseless, creates a false dichotomy in the effect size domain, and ignores context and decision theory. We present both perspectives. Along with this, we touch on the dichotomization inherent in NHST. We hope you find this insightful and ample. Lines 145:156 - Please amend this sentence “The Z subscript refers to the difference being compared is no longer between the measurements (X or Y ) but the difference (Z = Y − X)” so that future readers can relate to it better. Changed. Thank you. Line 151 - Please add “that” between “appears” and “the value”. Added. Formula (3) - I believe that it would be beneficial, for the sake of comparison and the greater picture, to put Cohen’s ds formula as well (with spooled in the denominator). We will have to politely disagree again. There are other papers with very extensive descriptions of SMDs for fully between subjects designs (we have cited them within the paper). Our intent within this manuscript is to specifically focus on issues in within subjects (repeated measures) designs which we believe have not received as much attention in the applied statistics literature. Lines 157:159 - I think that this part is a little bit unclear and the readers might have difficulties interpreting what are you trying to say, so I suggest expanding/re-writing this portion of the paragraph. In addition, can you please expand on this topic further, especially on standardised mean changes when a researcher wants to compare the “effectiveness” of the two different interventions. This is something people frequently overlook in their studies, is something that is now increasingly being done in sport and exercise science intervention studies, and is a common issue for meta-analysts looking to compare the effects of two different interventions on a given outcome. I believe it would add a lot to your story here and since you already 2 mentioned the relevant literature that covers it (i.e., Morris and DeShon 2002; Morris, 2008; Becker, 1988; potentially you could check Viechtbauer, 2007), I believe that I won’t take a lot of your time and space in the manuscript. The manuscript has undergone extensive revisions and restructuring in order to improve the clarity of our message. We have added a brief mention of comparing two different interventions, and now included one calculative approach for this comparison. Line 169 - While I understand that you are trying to be cautious with your statements here, I would still change “it need not exist in isolation” to “it should not exist in isolation”. Changed. Lines 169:179 - Since you are already covering “the reporting” in this sub-section, I would recommend to add a few lines highlighting the importance of “holistic” and accurate (i.e., exact numbers) reporting which will also facilitate more accurate meta-analyses. Essentially, people in sport science field should be better at reporting overall (i.e., not just effect sizes). We have added a note to this effect added to our recommendations section (See ”Data Sharing”). Line 194 - There might be a typo regarding “the denominator” and “the numerator”? Thanks! Yes, too much discussion of denominators made it an easy mistake to make. Figure 1 - In my opinion, I think you should try to pick a better palette with RColorBrewer. This current figure doesn’t really communicate what it intends since you really have to pay attention to see the difference. This is especially important since, in the next line or so, you said that one can “clearly see”. I actually asked for an opinion of several colleagues on this figure and they thought the same, so I believe its not just me. This is just for the benefit of the communication because I really like the figure even in its current form. We agree. We have updated the figure aide the visual representation. Data is also available at our OSF repository. We have also simplified the figure to only include the two primary effect sizes discussed throughout the manuscript. Bias sub-section - When you mention small sample sizes correction, can you please add the recommendation from the literature how small sample size (e.g. 20) should be before needing to apply Hedges’ g correction. Perhaps, you can just put this information in the parentheses where you feel it’s appropriate. We’ve added, “This correction decreases the SMD by about 10 and 5% with 10 and 15 participants, respectively; corrections are negligible with larger sample sizes.” 3 261-262 - While I entirely agree with you here, it is important to note that although the design of randomised trials aims to ensure that baseline characteristics are balanced, imbalance may arise by chance (especially if a trial’s sample size is small) not only through an inadequate randomisation strategy (Riley, 2013). Perhaps, you can add this statement as its relevant for sport science field. While we believe this is orthogonal to our point that SMD estimate from a sample on average represents the population parameter, we have added some text citing Riley for when an analysis of covariance may be warranted “If imbalance in some relevant covariate is a concern, then an analysis of covariance, and the effect size estimate from this statistical model, should be utilized ” Line 276 - Please flip the numbers in the SMD formula (i.e., current denominator should be numerator and vice versa). Good catch. Thanks. Fixed. Line 284:296 - Please consider giving a concrete example of reporting the dz effect size, similar to the one above (lines 274:283). You could add that after outlining the wrong way of interpreting dz effect size (lines 290:291). Perhaps, you can also add a statement here that people should actually start reporting t values and associated degrees of freedom as this is rarely the case in our field and potentially causes this “drama” around dz effect size. We believe we have provided a concrete example of how it is commonly (sometimes inappropriately) interpreted in the paragraph. We are not against reported Cohen’s dz per se, but do not find it personally useful in interpreting results. If authors find it useful we have no problem with its utilization so long as they can justify that decision. We have added this clarification to multiple parts of the manuscript to make it clear to readers that we are not against reporting dz . 1.2 Experimental design No comment 1.3 Validity of the findings No comment 4 1.4 Comments for the author This article was very well-written, and I commend the authors for maintaining a balanced, even tone throughout the piece despite the obvious disagreement with the other research group. I believe that this article facilitates a deeper understanding of different effect sizes (as well as pros and cons of each) which frequently create confusion among researchers regardless of their field of expertise. Therefore, I only have minor suggestions aiming to improve the manuscript further. 5 2 2.1 Reviewer: Anthony Ciccone Basic reporting Line 36: Elaborate on the specific suggestions/claims the Dankel study makes and move this paragraph to be melded with the fourth paragraph in the introduction, where specific examples are first identified. This is a great manuscript, but the intro didn’t really draw me in. I feel like you could create a bit more interest by detailing the potential issue(s) with the Dankel recommendation. We have restructured our introduction accordingly. 2.2 Experimental design Line 88: In the current manuscript, authors also make this suggestion. I am not quite sure where in the 1977 Cohen textbook he suggests what is stated in this manuscript, that “standardized effect sizes are not dependent on the sample size”. Similaraly, Dankel and Loenneke suggest: “Importantly, the effect size is supposed to provide information about the statistical test while removing the influence of the sample size”. . . a page later “Thus, a larger effect size can be obtained by one of 2 ways: (a) a larger mean change from pre to post; or (b) a smaller magnitude of variability with respect to how people responded to the intervention” and attribute this to the supposed ability of dz to negate the effect of sample size, which is easily proven false by doubling any repeated measures data and calculating dz. The mean change will remain and the SD of the change scores will decrease, causing dz to decrease. Unless I am missing something, the math does not support this statement in either the current manuscript or the Dankel article. This issue needs to be addressed somewhere in the manuscript. The following articles may be a place to start. Negative correlation between sample size and effect size (Slavin and Smith, 2009) We have trouble following this assertion, “[. . . ] which is easily proven false by doubling any repeated measures data and calculating dz. The mean change will remain and the SD of the change scores will decrease, causing dz to decrease.” √ ), the effect of sample While standard error scales substantially with sample size (∝ n−1/2 since SE = SD n size on sample standard deviation is relatively small. That is, sample standard deviation is a biased estimate of population standard deviation, but this bias is a small overestimate that quickly decays after ∼4–5 subjects. On average, sample standard deviations only have negligible changes with sample size (https: //en.wikipedia.org/wiki/Unbiased_estimation_of_standard_deviation). It is therefore unclear as to why the sample standard deviation would appreciably decrease with increases in sample size. The effects of bias on the SMDs themselves are well-studied and understood (e.g., Hedges, Becker, etc.). Effects due to publication bias, such as those in the study you linked, are orthogonal to this discussion. Line 117-130: I think authors are inappropriately over-interpreting the Dankel paper. The Dankel paper really doesn’t say much other than to stop using pooled SDs in repeatedmeasures analyses ES and power calculations, and provides numerous methods to calculate dz 6 if change SD is not reported. This leaves point 1 in this section irrelevant and distracting. I am fine with point two (the argument against observed effect sizes in power analyses), but I don’t think it is fair to link the whole list to the Dankel paper. Further, if authors are going to make such an argument against observed effect sizes for power analyses, they should provide a more acceptable alternative. We should note that our paper started as a response to the Dankel & Loenneke paper, but after it grew in length, we decided to give it a higher-level, educational twist so that it would be more constructive. However, the issues with Dankel & Loenneke’s assertions need to be addressed; they should have never been published since they are mathematically wrong and philosophically questionable. Moreover, our concern over Dankel & Leonneke’s paper has only grown as we have seen it cited during the review of manuscripts to criticize the use of the other SMDs (personal correspondence with colleagues). Having said that, we kindly disagree. Dankel & Loenneke (2018) wholly and unambiguously advocate for Cohen’s dz to be calculated for pre-post designs in lieu of other SMDs, especially ∆pre . Effect size calculations that are made on paired data should be made relative to the SD of the change score because this provides the information of the statistical test while removing the influence of the sample size. [. . . ] the purpose of this article is to provide examples detailing why effect size calculations on paired data should be made relative to the variability of the change score as opposed to the variability of the pre-test. [. . . ] the most appropriate way to calculate an effect size when examining the effect of an intervention would be to examine the mean difference divided by the SD of the mean difference (i.e., Cohen’s dz). In addition to the above, Dankel & Loenneke also brazenly condemn what we term ∆pre in favor of dz : When using within-subjects designs, normalizing effect size values to the pre-test SD will enable the calculation of a confidence interval before the intervention is even completed. This again does not make sense because the confidence interval should provide information on how accurate the estimated effect will be, and this can only be calculated once the variability of the intervention is known. This again also points to the flaws of normalizing effect sizes to the pretest SD because the magnitude of the effect (i.e., the effectiveness of the intervention) is dependent on the individuals recruited rather than the actual effectiveness of the intervention. [. . . ] future studies should not detail the effects of an intervention in pre-test SD units, but should rather detail the effect in change score SD units. The appropriate effect size (Cohen’s dz) can be calculated as the mean change divided by the SD of the change score (i.e., the pre-test to post-test difference). However, we have condensed the majority of our arguments against Dankel & Loenneke to one section. In addition, we have added a brief paragraph about sample size justifications. Line 150: “It is difficult to interpret the value of this type of SMD (e.g., how many standard deviations of the change score is large?).” This is a very tough argument to make, as it could be 7 made for many effect sizes where the raw units are absent. I suggest removing this statement from the manuscript as authors provide a more applied reason to not use this SMD towards the end of the manuscript. We are not against reported Cohen’s dz per se, but do not find it personally useful in interpreting results. If authors find it useful we have no problem with its utilization so long as they can justify that decision. We have added clarifying statements throughout (i.e., ”In our professional opinion...”) in order to make our points clearer and provide more direction for our readers. Line 155: It seems like CLES and dz are different ways of conveying the same information and I, again, don’t think it is appropriate to be tough on the Dankel paper here since CLES is a mathematical transformation of dz. However, I do think the authors should describe the potential usefulness of CLES and how it may be a more intuitive than dz. This is indeed the case. We expanded on this to try to make it clearer. Line 175: the 2013 Lakens paper states that CIs can be calculated for dz. Thus, I believe this argument should be removed from the manuscript We have trouble understanding why this point relates to the ability of CIs to be calculated for dz . We certainly agree with this—indeed, we provide the formula for dz ’s variance in our paper—but this does not enable the reader to tease apart the contributions of high (low) signal or low (high) noise to a high (low) dz . The CI itself is estimated from the magnitude of dz and the sample size; the raw effect and its variance are not independent components of the CI calculation. In other words, dz has a variance that depends on its magnitude. This does not mean it’s not a viable option, but these are properties of the estimator that should be known and understood by those looking to apply it. Line 194: I believe authors meant to say numerator instead of denominator. This should be changed. Good catch! Thank you. Line 200 and : This is a subtly strong statement that sounds like you are suggesting dz is less reliable than other effect sizes, which would require a citation. However, it seems like the more concise statement is that pre-post correlation affects dz, but not other ES calculations. This sentence should be clarified, preferably without using the word, “unstable”. We disagree that unstable is inappropriate, given that instability is a property that means the output will be sensitive to small changes in inputs and has been discussed in relation to estimators (e.g. see https://www.jstor.org/stable/1912329), we have changed this to “blows up” and “explodes,” which is more common in mathematics while instability is more from dynamics. Note, all of these properties are related to the fact these estimators have singularities. The estimators and their variance blowing up is evident from the math and is incontrovertible. 8 Line 234-240: See my concerns for Line 200. Again, these statements either need a simulation or citations to show that dz variance is less “stable”. Lakens seems to see nothing wrong with dz in his 2013 article, and seems to suggest it is an accurate representation of an effect size for repeated measures data. Both our simulations (now in supplemental) and the closed-form mathematics (Figure 1) support this assertion. In the left panel of Figure 1 one can see that there is unfavorable properties of dz as ρ approaches 1. While we personally do not find much utility in reporting Cohen’s dz , we do not oppose its use by thoughtful exercise and sport scientists. We simply are explaining our professional opinion via how these statistics can be used (interpretation) and their estimator properties (mathematics). 2.3 Validity of the findings Line 245: At this point, readers are aware that the Dankel/Loenneke article is subsumed by this manuscript. This whole section could be removed, and the manuscript would still convey the same information. I suggest removing the whole Previous Arguments section and incorporating lines 257-262 into a much earlier section, closer to page 3. We have restructured the manuscript in order to limit our criticisms of Dankel & Leonneke to one section. Line 292: I think this idea should be expanded to include the mean difference confidence interval. We respectfully disagree. The use of confidence (compatibility) intervals can be an effective tool for communicating a reasonable range of effect sizes (i.e., answering the question: what effect size, or mean difference, is compatible with data and was are the extremes of that compatibility?). While the t-statistic and degrees of freedom can be used to derive this information they do not provide this information directly. Whereas, the t-statistic and degrees of freedom do accomplish the stated goals of what Dankel & Leonneke described as an appropriate effect size (sans being sample size invariant). Somewhere in this article, authors should include a bit about journals needing to revisit their statistical requirements, as many journals require the inclusion of effect sizes, which can be an issue for authors or reviewers who may deem effect sizes to be distracting or unproductive for the data at hand. We have added text to our recommendations section that journals start enforcing, and encouraging, their policies on reporting effect sizes 9 Effect sizes calculated using percent change data are incompatible with any raw unit effect size. Thus, it is inappropriate to ONLY report percent change mean and SD, as those data are not usable in a meta analysis with other typical effect size data. Thus, the Cohen proponent paragraph is also false, as not all data are compatible as effect sizes.This should be addressed as it is definitely an issue in our field. We absolutely agree and have added a couple of sentences about the dangers of using percentage change. Below is a quote from the 2013 Lakens article. I think authors should provide readers with this argument with or without an argument against it. It would probably fit best towards the end of the manuscript. Lakens, 2013: “I believe this discussion is currently biased by what could be called designism, a neologism to refer to the implicit belief that between-subjects designs are the default experimental design, and that effect sizes calculated from betweensubjects designs are more logical or natural. The defense for designism is as follows. It is desirable to be able to compare effect sizes across designs, regardless of whether the observations originate from a within or between-subjects design. Because it is not possible to control for individual differences in between-subject designs, we therefore should consider the effect size that does not control for individual differences as the natural effect size. As a consequence, effect sizes that control for individual differences are “inflated” compared to the “default” (e.g., Dunlap et al., 1996)” . . . “When empirical questions can only be examined in withinsubjects designs (such as in the case of post-error slowing), effect sizes that control for intra-subjects variability (ηp2 and ωp2 ), or that take the correlation between measurements into account (Cohen’s dz ) are a reasonable statistic to report” We sympathize, yet do not wholly agree, with Lakens’ perspective of “designism”. That is to say, we agree that one effect size should not be reported strictly on the basis of its ability to be employed across both within- and between-subject designs. This rationale is likely a pragmatic one born from meta-analysis (i.e., if meta-analysts wish to include studies that employ both types of designs). Our perspective is tangential to the idea of designism; rather than favoring one type of SMD based on whether it can be related to between-subject designs, we take a higher-level perspective. That is, scientists should choose an ES based on the information that the SMD communicates and its statistical properties. For this reason, we do not think adding Lakens’ perspective, and our response to it, furthers the point we’re trying to convey in the manuscript. As we state multiple times, a scientist has every right to report whatever effect size they deem adequate (so long as it is calculated correctly). Our professional opinion and recommendations that tend to favor magnitude-based SMDs are based on the fact that we find them easier to interpret, and we are generally more interested in magnitudes than signal-to-noise characteristics. Together with its CI, the magnitudebased SMD arguably adequately communicates both the magnitude of the effect and its uncertainty, while these cannot be teased apart with the signal-to-noise SMD without knowing its components. We have added emphasis throughout the manuscript to make the distinction between our recommendation/opinion and freedom authors should have in reporting effect sizes. 10 2.4 Comments for the author This manuscript provides sport an exercise scientists and practitioners with thoughts and suggestions regarding the use of effect sizes that will probably be novel to most readers. Authors did a great job of gathering and disseminating the effect size knowledge in a relatively easy to digest text. My suggestions do not require major changes, but aim to balance the manuscript as to minimize the chance of reader-perceived bias. Thanks so much! 11 "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Recent discussions in the sport and exercise science community have focused on the appropriate use and reporting of effect sizes. Sport and exercise scientists often analyze repeated-measures data, from which mean differences are reported. To aid the interpretation of these data, standardized mean differences (SMD) are commonly reported as description of effect size. The use of SMDs has been criticized by sport and exercise scientists as well as those in other disciplines. However, we believe, when thoughtfully applied, SMDs can be a useful analytical tool for sport and exercise scientists. In this manuscript, we hope to alleviate some confusion by demonstrating how effect sizes-primarily SMDs-are calculated, and describing their statistical properties. Finally, we provide high-level recommendations for how sport and exercise scientists can thoughtfully report raw effect sizes, SMDs, or other effect sizes for their own studies.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Effect sizes are a family of descriptive statistics used to communicate the magnitude or strength of a quantitative research finding. Many forms of effect sizes exist, ranging from mean raw values to correlation coefficients. In sport and exercise science, a standardized mean difference (SMD) is commonly reported in studies that observe changes from pre-to post-intervention, and for which units may vary from study-to-study (e.g., muscle thickness vs. cross-sectional area vs. volume). Put simply, an SMD is any mean difference or change score that is divided, hence standardized, by a standard deviation or combination of standard deviations. Thus, even among SMDs, there exist multiple calculative approaches <ns0:ref type='bibr' target='#b27'>(Lakens, 2013;</ns0:ref><ns0:ref type='bibr' target='#b2'>Baguley, 2009)</ns0:ref>. A scientist must therefore decide which SMD is most appropriate to report for their particular study, or if to report one at all. In this manuscript, we will exclusively be focusing on SMD calculations for studies involving repeated-measures since this is a common feature of sport and exercise science studies; other study designs (i.e., between-subjects) have already been extensively covered elsewhere <ns0:ref type='bibr' target='#b2'>(Baguley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kelley and Preacher, 2012;</ns0:ref><ns0:ref type='bibr' target='#b21'>Hedges, 2008)</ns0:ref>.</ns0:p><ns0:p>Different forms of SMDs communicate unique information and have distinct statistical properties.</ns0:p><ns0:p>Yet, some authors in sport and exercise science have staunchly advocated for specific SMD calculations, and in doing so, outright rebuke other approaches <ns0:ref type='bibr' target='#b8'>(Dankel and Loenneke, 2018)</ns0:ref>. While we appreciate that previous discussions of effect sizes have brought this important topic to the forefront, we wish to expand on their work by providing a deeper philosophical and mathematical discussion of SMD choice.</ns0:p><ns0:p>In doing so, we suggest that the choice of an SMD should be based on the objective of each study and therefore is likely to vary from study-to-study. Scientists should have the intellectual freedom to choose whatever statistics are needed to appropriately answer their question. Importantly, this freedom should not be encroached on by broad recommendations that ignore the objectives of an individual scientist. To facilitate these reporting decisions, it is imperative to understand what to report and why.</ns0:p><ns0:p>In this paper, we broadly focus on three things to consider when reporting an SMD. First, before PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed choosing an SMD, a scientist must decide if one is necessary. When making this decision, it is prudent to consider arguments for and against reporting SMDs, in addition to why one should be reported. Second, we broadly categorize repeated-measures SMDs into two categories: signal-to-noise and magnitude-based SMDs. This dichotomy provides scientists with a philosophical framework for choosing an SMD. Third, we describe the statistical properties of SMDs, which we believe scientists should try to understand if they are to report them. We relate these perspectives to previous discussions of SMDs, make general recommendations, and conclude by urging scientists to think carefully about what effect sizes they are reporting and why.</ns0:p></ns0:div> <ns0:div><ns0:head>SHOULD I REPORT A STANDARDIZED MEAN DIFFERENCE?</ns0:head><ns0:p>Before reporting an SMD-or any statistic for that matter-a researcher should first ask themselves whether it is necessary or informative. When answering this, one may wish to consider arguments both for and against SMDs, in addition to field standards. Here, we briefly detail these arguments, in addition to SMD reporting within sport and exercise science.</ns0:p></ns0:div> <ns0:div><ns0:head>Proponents and Opponents of Standardized Effect Sizes</ns0:head></ns0:div> <ns0:div><ns0:head>Opponents</ns0:head><ns0:p>Although SMDs may be useful in some contexts, they are far from a panacea. Arguments against the use of SMDs, including those by prominent statisticians, are not uncommon. These arguments should be considered when choosing whether or not to report an SMD. In particular, the evidentiary value of reporting an SMD must be considered relative to the strength of the general arguments against SMDs.</ns0:p><ns0:p>Below, we have briefly summarized some of the major arguments against the use of SMDs.</ns0:p><ns0:p>As far back as 1969, the use of standardized effect sizes-and by proxy, SMDs-has been heavily criticized. The eminent statistician John Tukey stated that 'only bad reasons seem to come to mind' for using correlation coefficients instead of unstandardized regression coefficients to interpret data. To put it simply, scientists should not assume that standardized effect sizes will make comparisons meaningful <ns0:ref type='bibr' target='#b45'>(Tukey, 1969)</ns0:ref>. This same logic can also be applied to qualitative benchmarks (e.g., Cohen's d = 0.2 is 'small'); we believe it is likely that Cohen would also argue against the broad implementation of these arbitrary benchmarks in all areas of research. Similar arguments against the misuse of standardized effect sizes have been echoed elsewhere <ns0:ref type='bibr' target='#b28'>(Lenth, 2001;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kelley and Preacher, 2012;</ns0:ref><ns0:ref type='bibr' target='#b2'>Baguley, 2009;</ns0:ref><ns0:ref type='bibr' target='#b41'>Robinson et al., 2003)</ns0:ref>.</ns0:p><ns0:p>Others have outright argued against the use of standardized effect sizes because they oversimplify the analysis of, and distort the conclusions derived from, data. In epidemiology, <ns0:ref type='bibr' target='#b17'>Greenland et al. (1986)</ns0:ref> provided a damning indictment of the use of standardized coefficients; namely, because they are largely determined by the variance in the sample, which is heavily influenced by the study design. In psychology, <ns0:ref type='bibr' target='#b2'>Baguley (2009)</ns0:ref> offers a similarly bleak view of standardized effect sizes. He argues that the advantages of standardized effect sizes are far outweighed by the difficulties that arise from the standardization process.</ns0:p><ns0:p>In particular, scientists tend to ignore the impact of reliability and range restriction on effect size estimates, in turn overestimating the generalizability of standardized effect sizes to wider populations and other study designs <ns0:ref type='bibr' target='#b2'>(Baguley, 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Proponents</ns0:head><ns0:p>Conversely, prominent statisticians have also argued in favor of standardized effect sizes, especially for facilitating meta-analysis <ns0:ref type='bibr' target='#b21'>(Hedges, 2008)</ns0:ref>. <ns0:ref type='bibr' target='#b7'>Cohen (1977)</ns0:ref> was the first to suggest the use of standardized effect sizes to be useful for power analysis purposes. This is because, unlike the t-statistic, (bias-corrected) standardized effect sizes are not dependent on the sample size. Similarly, while p-values indicate the compatibility of data with some test hypothesis (e.g., the null hypothesis) <ns0:ref type='bibr' target='#b16'>(Greenland, 2019)</ns0:ref>, SMDs provide information about the 'effect' itself <ns0:ref type='bibr' target='#b39'>(Rhea, 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Thomas et al., 1991)</ns0:ref>. Thus, p-values and t-statistics provide information about the estimate of the mean relative to some test hypothesis and thus are sensitive to sample size, while SMDs strictly pertain to the size of the effect and thus are insensitive to sample size. Moreover, any linear transformation of the data will still yield the exact same standardized effect size <ns0:ref type='bibr' target='#b20'>(Hedges, 1981)</ns0:ref>. The scale invariance property of a standardized effect size theoretically allows them to be compared across studies, various outcomes, and incorporated into a meta-analysis. Therefore, scientists can measure a phenomenon across many different scales or measurement tools and standardized</ns0:p></ns0:div> <ns0:div><ns0:head>2/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed effect sizes should, in theory, be unaffected. Finally, SMDs can provide a simple way to communicate the overlap of two distributions (https://rpsychologist.com/d3/cohend/).</ns0:p></ns0:div> <ns0:div><ns0:head>Comments on Standardized Mean Differences in Sport and Exercise Science</ns0:head><ns0:p>Sport and exercise scientists have also commented on the use of standardized effect sizes <ns0:ref type='bibr' target='#b9'>(Dankel et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b8'>Dankel and Loenneke, 2018;</ns0:ref><ns0:ref type='bibr' target='#b39'>Rhea, 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Thomas et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b12'>Flanagan, 2013)</ns0:ref>. The discussion has focused on the need to report more than just p-values, emphasizing that scientists have to discuss the magnitude of their observed effects. <ns0:ref type='bibr' target='#b39'>Rhea (2004)</ns0:ref> also provided new benchmarks for SMDs specific to strength and conditioning research, which is certainly an improvement from just using Cohen's benchmarks.</ns0:p><ns0:p>If SMDs are to be reported, they should not be done so in lieu of understanding effects on their natural scales. To this end, we agree with the laments of <ns0:ref type='bibr' target='#b45'>Tukey (1969)</ns0:ref>: too often, standardized effects sizes, particularly SMDs, are relied upon to provide a crutch for interpreting the meaningfulness of results.</ns0:p><ns0:p>Default and arbitrary scales, such as 'small' or 'large' based on those proposed by <ns0:ref type='bibr' target='#b7'>Cohen (1977)</ns0:ref>, should generally be avoided. SMDs should be interpreted on a scale calibrated to the outcome of interest. For example, <ns0:ref type='bibr' target='#b39'>Rhea (2004)</ns0:ref> or <ns0:ref type='bibr' target='#b37'>Quintana (2016)</ns0:ref> have demonstrated how to develop scales of magnitude for a specific area of research. When possible, it is best practice to interpret the meaningfulness of effects in their raw units, and in the context of the population and the research question being asked. For example, a 5 mmHg decrease in systolic blood pressure may be hugely important or trivial, depending on the context-here, the SMD alone cannot communicate clinical relevance.</ns0:p><ns0:p>In our opinion, standardized effect sizes can be useful tools for interpreting data when thoughtfully employed by the scientists reporting them. However, sport and exercise scientists should be careful when selecting the appropriate SMD or effect size, and ensure that their choice effectively communicates the effect of interest <ns0:ref type='bibr' target='#b19'>(Hanel and Mehler, 2019)</ns0:ref>. Herein, we will discuss things to consider when reporting an SMD, and we will close by providing general recommendations and examples that we believe sport and exercise scientists will find useful.</ns0:p></ns0:div> <ns0:div><ns0:head>WHICH STANDARDIZED MEAN DIFFERENCE SHOULD I REPORT?</ns0:head><ns0:p>To facilitate a fruitful discussion of SMDs, here, we categorize them based on the information they convey.</ns0:p><ns0:p>We contend that there are two primary categories of SMDs that sport and exercise scientists will encounter in the literature and use for their own analyses. The first helps to communicate the magnitude of an effect (magnitude-based SMD), and the second is more related to the probability that a randomly selected individual experiences a positive or negative effect (signal-to-noise SMD). These categories serve distinct purposes, and they should be used in accordance with the information a scientist is trying to convey to the reader. We will contrast these SMD categories in terms of the information that they communicate and when scientists may wish to choose one over the other. In doing so, we will show that both approaches to calculating the SMD are distinctly valuable. Finally, we demonstrate that, when paired with background information and other statistics-whether they be descriptive or inferential-each SMD can assist in telling a unique, meaningful story about the reported data.</ns0:p></ns0:div> <ns0:div><ns0:head>Signal-to-noise Standardized Mean Difference</ns0:head><ns0:p>The first category of SMDs can be considered a signal-to-noise metric: it communicates the average change score in a sample relative to the variability in change scores. This is called Cohen's d z , and it is an entirely appropriate way to describe the change scores in paired data. The Z subscript refers to the difference being compared is no longer between the measurements (X or Y ) but the difference (Z = Y &#8722; X).</ns0:p><ns0:p>This SMD is directly estimating the change standardized to the variation in this response, making it a mathematically natural signal-to-noise statistic.</ns0:p><ns0:p>Cohen's d z can be calculated with the mean change, &#948; , and the standard deviation of the difference, &#963; &#948; ,</ns0:p><ns0:formula xml:id='formula_0'>d z = &#948; &#963; &#948; .<ns0:label>(1)</ns0:label></ns0:formula><ns0:p>Alternatively, for convenience, can be calculated from the t-statistic and the number of pairs (n),</ns0:p><ns0:formula xml:id='formula_1'>d z = t &#8730; n .</ns0:formula><ns0:p>(2)</ns0:p></ns0:div> <ns0:div><ns0:head>3/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>In Eq. 2, one can see that d z is closely related to the t-statistic. Specifically, the t-statistic is a signal-tonoise metric for the mean (i.e., using its sampling distribution), while d z is a signal-to-noise metric for the entire sample. This means that the t-statistic will tend to increase with increases in sample size, since the estimate of the mean becomes more precise, while (bias-corrected) Cohen's d z will not change with sample size.</ns0:p><ns0:p>Although Cohen's d z may be useful to describe the change in a standardized form, it is typically not reported in meta-analyses since it cannot be used to compare differences across between-and withinsubjects designs (see SMDs below). It is difficult to interpret the value of this type of SMD; that is, since the signal-to-noise ratio itself is more related to the consistency of a change, one can wonder how much consistency constitutes a 'large' effect? This is in contrast to other types of SMDs, wherein the statistic conveys information about the distance between two central tendencies (mean) relative to the dispersion of the data (standard deviation). Moreover, it appears that, to sport and exercise scientists, the value of this SMD is measuring the degree of the change in comparison to the variability of the change scores <ns0:ref type='bibr' target='#b8'>(Dankel and Loenneke, 2018)</ns0:ref>. Therefore, scientists' intent on using d z should consider reporting the common language effect size (CLES) <ns0:ref type='bibr' target='#b32'>(McGraw and Wong, 1992)</ns0:ref>, also known as the probability of superiority <ns0:ref type='bibr' target='#b18'>(Grissom, 1994)</ns0:ref>. In contrast to d z , CLES communicates the probability of a positive (CLES &gt; 0.5) or negative (CLES &lt; 0.5) change occurring in a randomly sampled individual (see below).</ns0:p></ns0:div> <ns0:div><ns0:head>Alternative to the signal-to-noise Standardized Mean Difference</ns0:head><ns0:p>The information gleaned from the signal-to-noise SMD (Cohen's d z ) can also be captured with the CLES <ns0:ref type='bibr' target='#b32'>(McGraw and Wong, 1992;</ns0:ref><ns0:ref type='bibr' target='#b18'>Grissom, 1994)</ns0:ref>. In paired samples, the CLES conveys the probability of a randomly selected person's change score being greater than zero. The CLES is easy to obtain; it is simply the Cohen's d z (SMD) converted to a probability (CLES = &#934;(d z ), where &#934; is the standard normal cumulative distribution function). Importantly, CLES can be converted back to a Cohen's d z with the inverse normal cumulative distribution function (d z = &#934; &#8722;1 (CLES)). CLES is particularly useful because it directly conveys the direction and variability of change scores without suggesting that the mean difference itself is small or large. Further, current evidence would suggest that the CLES is easier for readers to comprehend than a signal-to-noise SMD <ns0:ref type='bibr' target='#b19'>(Hanel and Mehler, 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Magnitude-based Standardized Mean Difference</ns0:head><ns0:p>The second category of SMDs can be considered a magnitude-based metric: it communicates the size of an observed effect relative to spread of the sample. The simplest and most understood magnitude-based SMD is Glass's &#8710;, which is used to compare two groups, and is standardized to the standard deviation of one of the groups. However, a conceptually similar version of Glass's &#8710;, which we term Glass's &#8710; pre , can also be employed for repeated-measures. In &#8710; pre the mean change is standardized by the pre-intervention standard deviation. 1 For basic pre-post study designs, Glass's &#8710; pre is fairly straightforward; mean change is simply standardized to the standard deviation of the pre-test responses. There are other effect sizes for repeated measures designs such as Cohen's d av and d rm , but for brevity's sake these are described in the appendix. Of note, &#8710; pre , d av , and d rm are identical when pre-and post-intervention variances are the same (see Appendix).</ns0:p><ns0:formula xml:id='formula_2'>&#8710; pre = &#948; &#963; pre (3)</ns0:formula><ns0:p>Importantly, &#8710; pre is well-described <ns0:ref type='bibr' target='#b36'>(Morris and DeShon, 2002;</ns0:ref><ns0:ref type='bibr' target='#b34'>Morris, 2000;</ns0:ref><ns0:ref type='bibr' target='#b4'>Becker, 1988)</ns0:ref> and can also be generalized to parallel-group designs; in particular, when there are 2 groups, typically a control and treatment group, being compared over repeated-measurements <ns0:ref type='bibr' target='#b35'>(Morris, 2008)</ns0:ref>. Typically, in these cases, a treatment and control group are being directly compared in a 'pretest-posttest-control design' (PPC). A simple version of the PPC-adapted &#8710; pre is</ns0:p><ns0:formula xml:id='formula_3'>&#8710; ppc = &#8710; T &#8722; &#8710; C (4)</ns0:formula><ns0:p>where &#8710; T and &#8710; C are the &#8710; pre from the treatment and control groups, respectively. There are several other calculative approaches which should be considered for comparing SMDs in a parallel-group designs. We highly encourage further reading on this topic if this type of design is of interest to readers <ns0:ref type='bibr' target='#b35'>(Morris, 2008;</ns0:ref><ns0:ref type='bibr' target='#b4'>Becker, 1988;</ns0:ref><ns0:ref type='bibr' target='#b47'>Viechtbauer, 2007)</ns0:ref>.</ns0:p><ns0:p>1 Although conceptually similar, Glass's &#8710; and &#8710; pre have different distributional properties <ns0:ref type='bibr' target='#b4'>(Becker, 1988)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>4/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Summary of Standardized Mean Differences</ns0:head><ns0:p>Our distinction between signal-to-noise (namely, Cohen's d z ) and magnitude-based SMDs (including Glass's &#8710; pre , Cohen's d av , and Cohen's d rm ) provides a conceptual dichotomy to assist researchers in picking an SMD (summarized in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). However, along with the conceptual distinctions, researchers should also consider the the properties of these SMDs. In the following section, we briefly go over the math underlying each SMD and its implications. The properties that follow from the math complement the conceptual framework we just presented, in turn providing researchers with a theoretical, mathematical basis for choosing and justifying their choice of an SMD. should have a general understanding of the statistical properties of an estimator they are using; namely, its bias and variance, which together determine the accuracy of the estimator (mean squared error, MSE = Bias( &#952; , &#952; ) 2 + Var &#952; ( &#952; ), for some true parameter, &#952; , and its estimate, &#952; ). These properties depend on the arguments used in the estimator. As a result, signal-to-noise and magnitude-based SMDs are not only distinct in terms of their interpretation, but also their statistical properties. Although these properties have been derived elsewhere (e.g., <ns0:ref type='bibr' target='#b20'>Hedges (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b36'>Morris and DeShon (2002)</ns0:ref>; <ns0:ref type='bibr' target='#b34'>Morris (2000)</ns0:ref>; <ns0:ref type='bibr' target='#b13'>Gibbons et al. (1993)</ns0:ref>; <ns0:ref type='bibr' target='#b4'>Becker (1988)</ns0:ref>), their implications are worth repeating. In particular, there are several salient distinctions between the properties of each of these metrics, which we will address herein. Although this section is more technical, we will return to a higher-level discussion of SMDs in the next section.</ns0:p></ns0:div> <ns0:div><ns0:head>Estimator Components</ns0:head><ns0:p>Before discussing bias and variance, we will briefly discuss the components of the formulae and their implications. Of course, all SMDs contain the mean change score, &#948; , in the numerator, and thus increase linearly with mean change (all else held equal). Since this is common to all SMDs, we will not discuss it further.</ns0:p><ns0:p>More interestingly, the signal-to-noise and magnitude-based SMDs contain very different denominators. To simplify matters, let us assume the pre-and post-intervention standard deviations are equal (&#963; pre = &#963; post = &#963; ). This assumption is reasonable since pre-and post-intervention standard deviations typically do not substantially differ in sports and exercise science. In this case, the standard deviation of change scores can be found simply:</ns0:p><ns0:formula xml:id='formula_4'>&#963; &#948; = 2&#963; 2 (1 &#8722; r).</ns0:formula><ns0:p>With these assumptions, d rm = d av = &#8710; pre for &#8722;1 &lt; r &lt; 1, where r is the observed pre-post correlation (Appendix). Greater pre-post correlations, r, are indicative of more homogeneous change scores. This makes the behavior of the magnitude-based SMDs fairly straightforward; that is, the estimates themselves will not be affected by the correlation between pre-and post-intervention scores. Their dependence on &#963; means that the magnitude-based SMD will blow up as &#963; &#8594; 0. This is in contrast to d z , whose denominator contains both &#963; and r (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>, top), making it blow up if either &#963; &#8594; 0 or r &#8594; 1.</ns0:p><ns0:p>The parsimonious nature of magnitude-based SMDs arguably makes their interpretation easier; with reasonable assumptions, they only depend on the mean change score and the spread of scores in the sample. On the other hand, when breaking d z down into its constituent parts, it depends on the mean change score, the spread of scores in the sample, and the correlation between pre-and post-intervention scores-the latter two will create &#963; &#948; . These sensitivities should be understood before implementing an SMD.</ns0:p></ns0:div> <ns0:div><ns0:head>5/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed (Bottom) The standard error of each estimator increases as &#963; pre &#8594; 0. Importantly, &#8710; pre has lower or similar standard errors as r &#8594; 1, whereas d z has greater standard errors as r &#8594; 1. Additional simulations, including those of other SMDs, can be found in online supplemental material https://www.doi.org/10.17605/OSF.IO/FC5XW</ns0:p></ns0:div> <ns0:div><ns0:head>Bias</ns0:head><ns0:p>Bias means that, on average, the estimate of a parameter ( &#952; ) differs from the 'true' parameter being estimated (&#952; ). Most SMDs follow a non-central t-distribution, allowing the bias to be easily assessed and corrected. As shown by <ns0:ref type='bibr' target='#b20'>Hedges (1981)</ns0:ref>, SMDs are generally biased upwards with small sample sizes; that is, with smaller samples, SMDs are overestimates of the true underlying SMD ( &#952; &gt; &#952; ). This bias is a function of both the value of the SMD obtained and the sample size:</ns0:p><ns0:formula xml:id='formula_5'>E[d] = d = d c(n &#8722; 1) (5) =&#8658; Bias[ d, d] = d &#8722; d = d ( 1 /c(n&#8722;1) &#8722; 1) ,<ns0:label>(6)</ns0:label></ns0:formula><ns0:p>where d is the 'true' parameter being estimated, d is its estimate, and c(m) = 1 &#8722; 3 4(m)&#8722;1 is Hedges' bias-correction factor <ns0:ref type='bibr' target='#b20'>(Hedges, 1981)</ns0:ref> and m = n &#8722; 1 is the degrees of freedom for a paired sample. Please note that this degrees of freedom will differ for different study designs and standard deviations. For example, with two groups and a pooled standard deviation, m = n 1 + n 2 &#8722; 2. We have noticed the incorrect use of degrees of freedom in some published papers within sport and exercise science, so we urge authors to be cautious.</ns0:p><ns0:p>Because SMDs are biased, especially in small samples, it is advisable to correct for this bias. Thus, when using Cohen's d in small sample settings, most sport and exercise scientists should apply a Hedges' correction to adjust for bias. A bias-corrected d is typically referred to as Hedges' g:</ns0:p><ns0:formula xml:id='formula_6'>g = d &#8226; c(n &#8722; 1),<ns0:label>(7)</ns0:label></ns0:formula><ns0:p>where d can represent any of the SMD estimates outlined above. This correction decreases the SMD by about 10 and 5% with 10 and 15 participants, respectively; corrections are negligible with larger sample sizes. Bias correction can also be applied via bootstrapping <ns0:ref type='bibr' target='#b42'>(Rousselet and Wilcox, 2019)</ns0:ref>.</ns0:p><ns0:p>More generally, we stress to readers that bias per se is not a bad thing or undesirable property.</ns0:p><ns0:p>Especially in multidimensional cases, bias can improve the accuracy of an estimate by decreasing its</ns0:p></ns0:div> <ns0:div><ns0:head>6/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed variance-this is known as Stein's paradox <ns0:ref type='bibr' target='#b11'>(Efron and Morris, 1977)</ns0:ref>. Indeed, biased (shrunken) estimators of SMDs have been suggested which may decrease MSE <ns0:ref type='bibr' target='#b22'>(Hedges and Olkin, 1985)</ns0:ref>. However, these are not commonly employed. Having said this, the upward bias of SMDs is generally a bad thing. As will be discussed in the next subsection, by correcting for the upward bias, we also improve (decrease) the variance of the SMD estimate, in turn decreasing MSE via both bias and variance <ns0:ref type='bibr' target='#b20'>(Hedges, 1981;</ns0:ref><ns0:ref type='bibr' target='#b22'>Hedges and Olkin, 1985)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Variance</ns0:head><ns0:p>While bias tells us about the extent to which an estimator over-or underestimates the value of a true parameter, variance tells us how variable the estimator is. Estimators that are more precise (less variable)</ns0:p><ns0:p>will have tighter standard errors and thus confidence intervals, allowing us to make better judgments as to the 'true' magnitude of the SMD.</ns0:p><ns0:p>By looking at formulae for variance and its arguments, we can gain a better understanding of what affects its statistical properties. Below are the variance formulae for Cohen's d z and Glass's &#8710; pre , which are the two best understood SMDs for paired designs <ns0:ref type='bibr' target='#b4'>(Becker, 1988;</ns0:ref><ns0:ref type='bibr' target='#b13'>Gibbons et al., 1993;</ns0:ref><ns0:ref type='bibr' target='#b15'>Goulet-Pelletier and Cousineau, 2018;</ns0:ref><ns0:ref type='bibr' target='#b34'>Morris, 2000;</ns0:ref><ns0:ref type='bibr' target='#b36'>Morris and DeShon, 2002)</ns0:ref>.</ns0:p><ns0:formula xml:id='formula_7'>Var[d z ] = n &#8722; 1 n(n &#8722; 3) (1 + d 2 z n) &#8722; d 2 z c(n &#8722; 1) 2 (8) Var[&#8710; pre ] = n &#8722; 1 n(n &#8722; 3) 2(1 &#8722; r) + &#8710; 2 pre n &#8722; &#8710; 2 pre c(n &#8722; 1) 2 (9)</ns0:formula><ns0:p>Variances for the biased SMDs (above) can be easily converted to variances for the bias-corrected SMDs by multiplying each formula by c(n&#8722;1) 2 , which is guaranteed to decrease variance since c(&#8226;) &lt; 1 <ns0:ref type='bibr' target='#b20'>(Hedges, 1981)</ns0:ref>.</ns0:p><ns0:p>Each variance formula contains the SMD itself, meaning that variance will tend to increase with an increasing SMD. This also complicates matters for d z ; since &#963; &#948; can increase from a smaller &#963; pre or greater r, d z 's variance explodes with homogeneous populations or change scores (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Such a quality is not very desirable, as typically, we would like more precision as effects become more homogeneous; this property is a further indication that d z is not a measure of effect magnitude. This is in contrast to the magnitude-based SMDs, which become more precise as the effect becomes more homogeneous (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Of note, these differences in variance behaviors do not reflect differences in statistical efficiency; after adjusting for scaling, all are unbiased and equally efficient.</ns0:p><ns0:p>By investigating and understanding the statistical properties of a statistic-here, the SMDs-we can gain a better understanding of what we should and should not expect from an estimate. These properties provide us with an intuitive feel for the implications of the mathematical machinery underlying each SMD, in turn helping us choose and justify an SMD.</ns0:p></ns0:div> <ns0:div><ns0:head>Considering Previous Arguments for Signal-to-noise Standardized Mean Differences</ns0:head><ns0:p>There have been arguments against SMDs-at least certain calculative approaches-with one particular article claiming that magnitude-based SMDs are flawed <ns0:ref type='bibr' target='#b8'>(Dankel and Loenneke, 2018)</ns0:ref>. Specifically, <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> profess the superiority of Cohen's d z over magnitude-based SMDsspecifically, Glass's &#8710; pre -because of its statistical properties 2 and its relationship with the t-statistic.</ns0:p><ns0:p>Regarding the former, <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> opine that the magnitude-based SMD is 'dependent on the individuals recruited rather than the actual effectiveness of the intervention.' We do not find this to be a compelling argument against magnitude-based SMDs for several reasons. First, it is in no way specific to magnitude-based SMDs; all descriptive statistics are always specific to the sample. Second, if the data are randomly sampled (a necessary condition for valid statistical inference), then the sample should, on average, be representative of the target population. If imbalance in some relevant covariate is a 2 <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> suggest that, '... normalizing effect size values to the pre-test SD will enable the calculation of a confidence interval before the intervention is even completed ... This again also points to the flaws of normalizing effect sizes to the pretest SD because the magnitude of the effect ... is dependent on the individuals recruited rather than the actual effectiveness of the intervention' (p. 4). This is of course not the case, since the variance of the SMD will depend on, among other things, the change scores themselves. Thus, the confidence interval of the magnitude-based SMD estimate cannot be calculated a priori.</ns0:p></ns0:div> <ns0:div><ns0:head>7/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed concern, then an analysis of covariance, and the effect size estimate from this statistical model, should be utilized <ns0:ref type='bibr' target='#b40'>(Riley et al., 2013)</ns0:ref>.</ns0:p><ns0:p>It is certainly the case that Cohen's d z has a natural relationship with the t-statistic. Stemming from this relationship, <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> suggest that it is a more appropriate effect size statistic for repeated-measures designs. Although it is true that d z is closely related to the t-statistic, this does not imply that d z is the most appropriate SMD to report. First, the t-statistic and degrees of freedom (which should be reported) together provide the required information to calculate a Cohen's d z , meaning d z may contain purely redundant information. Second, although Cohen's d z has a clear relationship with the statistical power of a paired t-test, we want to emphasize that utilizing an observed effect size in power analyses is an inappropriate practice. Performing such power analyses to justify sample sizes of future work implicitly assumes that 1) the observed effect size is the true effect size; 2) follow-up studies will require this observed SMD; and 3) this effect size is what is of interest (rather than one based on theory or practical necessity). In most cases, observed effect sizes do not provide accurate estimates of the population-level SMD, and utilizing the observed SMD from a previous study will likely lead to an underpowered follow-up study <ns0:ref type='bibr' target='#b0'>(Albers and Lakens, 2018)</ns0:ref>, and moreover, relying on previously reported effect sizes ignores the potential heterogeneity of observed effect sizes between studies (McShane and B&#246;ckenholt, 2014). Rather, there exist alternative approaches to justifying sample sizes (Appendix 2).</ns0:p><ns0:p>In general, and in contrast to <ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref>, we believe that SMDs can be used for different purposes-whether to communicate the size of an effect, calculate power, or some other purpose-and what is best for one objective is not necessarily what is best for the others. Furthermore, we want to emphasize that these are not arguments against the use of signal-to-noise SMDs, but rather a repudiation of arguments meant to discourage the use of magnitude-based SMD by sport and exercise scientists.</ns0:p></ns0:div> <ns0:div><ns0:head>RECOMMENDATIONS FOR REPORTING EFFECT SIZES</ns0:head><ns0:p>In most cases, sport and exercise scientists are strongly encouraged to present and interpret effect sizes in their raw or unstandardized form. As others previously discussed, journals should require authors to report some form of an effect size, along with interpretations of its magnitude, instead of only reporting p-values <ns0:ref type='bibr' target='#b39'>(Rhea, 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Thomas et al., 1991)</ns0:ref>. However, an SMD, along with other standardized effect sizes, do not magically provide meaning to meaningless values. They are simply a convenient tool that can provide some additional information and may sometimes be helpful to those performing meta-analyses or who are unfamiliar with the reported measures. Specifically, there are situations where the outcome measure may be difficult for readers to intuitively grasp (e.g., a psychological survey, arbitrary units from Western Blots, moments of force). In such cases, a magnitude-based SMD-in which the SD of preand/or post-intervention measures is used in the denominator-can be used to communicate the size of the effect relative to the heterogeneity of the sample. In other words, a magnitude-based SMD represents the expected number of sample SDs (not the change due to the intervention) by which the participants improve. and the second would imply that she moves to the 60th percentile. Clearly, both of these interpretations are wrong. As opposed to a magnitude-based SMD, Cohen's d z is a signal-to-noise statistic that is related to the probability of a randomly sampled individual experiencing an effect rather than its magnitude alone.</ns0:p><ns0:p>In our opinion, Cohen's d z does not provide any more information than that which is communicated by the t-statistic and the associated degrees of freedom (which should be reported regardless of the effect size).</ns0:p><ns0:p>Instead, if the signal-to-noise is of interest, a CLES may provide the information a sport and exercise scientist is interested in presenting. Going back to our earlier example (d z = 11.62 and 0.25 respectively), the CLES would be approximately &gt; 99% and 59.9%, or the probability of a randomly sample individual undergoing an improvement is &gt; 99% or 59.9% for intervention 1 and 2, respectively. As <ns0:ref type='bibr' target='#b19'>Hanel and Mehler (2019)</ns0:ref> demonstrated, the CLES may be a more intuitive description of the signal-to-noise SMD.</ns0:p><ns0:p>While our personal recommendation leans towards the use of magnitude-based SMDs and CLES, it is up to the individual sport and exercise scientist to decide what effect size they feel is most appropriate for the data they are analyzing and point they are trying to communicate <ns0:ref type='bibr' target='#b25'>(H&#246;nekopp et al., 2006)</ns0:ref>.</ns0:p><ns0:p>In choosing an SMD, we also sympathize with <ns0:ref type='bibr' target='#b27'>Lakens (2013)</ns0:ref> reported alongside an effect size that provides some type of complementary information. This effect size can be standardized (e.g., &#8710; pre ) or unstandardized (raw), and should be reported with a confidence interval. Confidence intervals (CI) of a magnitude-based SMD will provide readers with information concerning both the magnitude and uncertainty of an effect size; CIs can be calculated using formulae, or perhaps more easily, using the bootstrap. In situations where the measurements are directly interpretable, unstandardized estimates are generally preferable. The CLES can also be reported when the presence of a change or difference between conditions is of interest.</ns0:p></ns0:div> <ns0:div><ns0:head>Percent Changes</ns0:head><ns0:p>It is not uncommon for sport and exercise scientists to report their data using percentages (e.g., percent change). While this is fine if it supplements the reporting of their data in raw units, it can be problematic if it is the only way the data are presented or if the statistics are calculated based on the percentages. In the case of SMDs, an SMD calculated using a percent change is not the same as an SMD calculated using raw units. More importantly, the latter-which is often of greater interest to readers or those performing meta-analysis-cannot be back-calculated from the former. It is imperative that authors consider the properties of the values that they report and what readers can glean from them.</ns0:p></ns0:div> <ns0:div><ns0:head>Data Sharing</ns0:head><ns0:p>To facilitate meta-analysis, we suggest that authors upload their data to a public repository such as the Open Science Framework, FigShare, or Zenodo <ns0:ref type='bibr' target='#b5'>(Borg et al., 2020)</ns0:ref>. This ensures that future metaanalysis or systematic reviews efforts have flexibility in calculating effect sizes since there are multitude of possible calculative approaches, designs, and bias corrections (see <ns0:ref type='bibr' target='#b2'>Baguley (2009)</ns0:ref>). When data sharing is not possible, we highly encourage sport and exercise scientists to upload extremely detailed descriptive statistics as supplementary material (i.e., sample size per group, means, standard deviations, and correlations), or alternatively, a synthetic dataset that mimics the properties of the original <ns0:ref type='bibr' target='#b38'>(Quintana, 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Examples</ns0:head><ns0:p>In the examples below, we have simulated data and analyzed it in R (see supplementary material) to demonstrate how results from a study in sport and exercise science could be interpreted with the appropriate application of SMDs. For those unfamiliar with R, there is an online web application (https:// doomlab.shinyapps.io/mote/) and extensive documentation (https://www.aggieerin.</ns0:p><ns0:p>com/shiny-server/introduction/) to simplify the process of calculating SMDs for those without R programming experience <ns0:ref type='bibr' target='#b6'>(Buchanan et al., 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>9/13</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Now, let us imagine a study trying to estimate the effect of cold water immersion on muscle soreness. For this hypothetical study, muscle soreness is measured on a visual analog scale before and after cold water immersion following a muscle damaging exercise. The muscle soreness score would be represented by cm on the scale measured left-to-right. Because sensations tend to be distributed lognormal <ns0:ref type='bibr' target='#b30'>(Mansfield, 1974)</ns0:ref>-and are multiplicative rather than additive-it is sensible to work with the logarithm of the reported soreness levels. Since these logged scores are not directly interpretable, it is sensible to use an SMD to help interpret the change scores. The hypothetical study could be written up as follows:</ns0:p><ns0:p>Muscle soreness was lower after cold water immersion (mean = 27, SD = 7) compared to before (mean = 46, SD = 11) cold water immersion, t(9) = -6.90, p &lt; .001, Glass's &#8710; pre = -2.2 95% CI <ns0:ref type='bibr'>[-3.2, 1.3]</ns0:ref>. The CLES indicates that the probability of a randomly selected individual experiencing a reduction in muscle soreness after cold water immersion is 99%.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSION</ns0:head><ns0:p>We contend that the reporting of effect sizes should be specific to the research question in conjunction with the narrative that a scientist wants to convey. In this context, pooled pre-and/or post-study SDs are viable choices for the SMD denominator. This approach provides insight into the magnitude of a given finding, and thus can have important implications for drawing practical inferences. Moreover, the values of this approach are distinct and, in our professional opinion, potentially more insightful than signal-to-noise SMDs, which essentially provide information that is redundant with the t-statistic. At the very least, there is no one-size-fits-all solution to reporting an SMD, or any other statistics for that matter.</ns0:p><ns0:p>Despite our personal preference towards other effect sizes, a sport and exercise scientist may prefer a signal-to-noise SMD (d z ) and could reasonably justify this decision. We urge sport and exercise scientists to avoid reporting the same default effect size and interpreting them based on generalized, arbitrary scales.</ns0:p><ns0:p>Rather, we strongly encourage sport and exercise scientists justify which SMD is most appropriate and provide qualitative (i.e., small, medium, or large effect) interpretations that are specific to that outcome and study design. Also, sport and exercise scientists should be careful to report the rationale for using an SMD over simply presenting raw mean differences. Lastly, the creation of statistical rituals wherein a single statistic, by default, is used to interpret the data is likely to result in poor statistical analyses rather than informative ones <ns0:ref type='bibr' target='#b14'>(Gigerenzer, 2018)</ns0:ref>. As J.M. Hammersley once warned, 'There are no routine statistical questions; only questionable statistical routines' <ns0:ref type='bibr' target='#b43'>(Sundberg, 1994)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>APPENDIX 1: STANDARDIZED MEAN DIFFERENCE CALCULATIVE AP-</ns0:head></ns0:div> <ns0:div><ns0:head>PROACHES</ns0:head><ns0:p>Throughout the text, we use Glass's &#8710; pre as our token magnitude-based SMD. However, there exist other approaches to calculating magnitude-based SMDs. Here, we briefly discuss two other common calculations of magnitude-based SMDs. Of note, these two other calculative approaches may contain some 'effects' (variance) from the intervention in the denominator, arguably making Glass's a more 'pure' (in the sense that the denominator is uncontaminated by intervention effects) magnitude-based SMD.</ns0:p><ns0:p>Cohen's d av : Some have argued that Cohen's d z is an overestimate of the SMD, and instead advocate for reporting an SMD very similar to the Cohen's d s typically utilized for between-subjects (independent samples) designs <ns0:ref type='bibr' target='#b10'>(Dunlap et al., 1996)</ns0:ref>. The only difference between Cohen's d av and Cohen's d s is that</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure1. Standardized mean differences for a range of pre-post correlations and pre-intervention standard deviations. Standardized mean differences (SMD) were calculated for a pre-post design study with 20 participants to depict the different properties of the different SMDs. We calculated SMDs for a range of pre-post correlations (r) and pre-intervention standard deviations (&#963; pre ), each with a mean change score of 1. (Top) Magnitude-based SMDs have similar estimates across the range of pre-post correlations and largely only vary as a function of &#963; pre , whereas signal-to-noise SMDs are a function of both &#963; pre and r. Note, d z blows up as r &#8594; 1, and all SMDs blow up as &#963; pre &#8594; 0. (Bottom) The standard error of each estimator increases as &#963; pre &#8594; 0. Importantly, &#8710; pre has lower or similar standard errors as r &#8594; 1, whereas d z has greater standard errors as r &#8594; 1. Additional simulations, including those of other SMDs, can be found in online supplemental material https://www.doi.org/10.17605/OSF.IO/FC5XW</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Let us consider examples presented previously in the sport and exercise science literature. The examples presented in Figure 1 by Dankel and Loenneke (2018), in which both interventions have a preintervention SD pre = 6.05 and undergo a change of &#916;= 3.0 (SMD = 3.0 6.05 = 0.5). This can be interpreted simply: the expected change is 0.5 standard deviation units relative to the measure in the sample. Put differently, if the person with the median score (50th percentile) were to improve by the expected change, she would move to the 69th percentile. 3 Like a mean change, this statistic is not intended to provide information about the variability of change scores. The magnitude-based SMD simply provides a unitless, interpretable value that indicates the magnitude of the expected change relative to the between-subject standard deviation. Of course, it can be complemented with a standard error or confidence interval if one is interested in the uncertainty around this estimate. The above can be contrasted with Cohen's d z , which uses the SD of change scores. Again, using the examples presented in Figure 1 of Dankel and Loenneke (2018), Cohen's d z of 11.62 and 0.25 are reported for interventions 1 and 2, respectively. If one tries to interpret these SMDs in a way that magnitude-based SMDs are interpreted, he will undoubtedly come to incorrect conclusions. The first would suggest that a person with the median score who experiences the expected change would move to &gt;99.99th percentile,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Scenario 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Interpretable Raw Differences In the first hypothetical example, let us imagine a study trying to estimate the change in maximal oxygen consumption ( VO 2 ; L&#8226;min -1 ) in long-distance track athletes before and after a season of training. For this study, maximal VO 2 was measured during a Bruce protocol with a Parvomedics 2400 TrueOne Metabolic System. The results of this hypothetical outcome could be written up as the following: VO 2 after a season of training with the track team (mean = 4.13 L&#8226;min -1 , SD = 0.25) increased compared to when they joined the team (M = 3.89 L&#8226;min -1 , SD = 0.21), t(7) = 3.54, p = 0.009, &#948; = 0.23 L&#8226;min -1 95% C.I. [0.07, 0.38]. The CLES indicates that the probability of a randomly selected individual's V O 2 increasing after their first season with the team is 89%.Scenario 2: Uninterpretable Raw Differences</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Types of Standardized Mean Differences for Pre-Post Designs</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Magnitude-based Glass's &#8710; pre , Cohen's d av , Cohen's d rm</ns0:cell></ns0:row><ns0:row><ns0:cell>Signal-to-noise</ns0:cell><ns0:cell>Cohen's d z</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>WHAT ARE THE STATISTICAL PROPERTIES OF STANDARDIZED MEAN</ns0:cell></ns0:row><ns0:row><ns0:cell>DIFFERENCES?</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>An SMD is an estimator. Estimators, including SMDs, have basic statistical properties associated with them that can be derived mathematically. From a high level, grasping how an estimator behaveswhat makes it increase or decrease and to what extent-is essential for interpretation. In addition, one</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>, '... to report effect sizes that cannot be calculated from other information in the article, and that are widely used so that most readers should understand them. Because Cohen's d z can be calculated from the t-value and the n, and is not commonly used, my general recommendation is to report Cohen's d av or Cohen's d rm .' Along these same lines, if scientists want to present an SMD, it should not exist in isolation. It is highly unlikely that a single number will represent all data in a meaningful way. We believe that data are often best appreciated when presented in multiple ways. The test and inferential statistics (p-values and t-statistics) should be</ns0:figDesc><ns0:table /></ns0:figure> <ns0:note place='foot' n='3'>Assumes a normal distribution. We note that<ns0:ref type='bibr' target='#b8'>Dankel and Loenneke (2018)</ns0:ref> data vignettes are approximately uniformly distributed which is an odd assumption to make about theoretical data, but nonetheless, sufficiently conveys the point.8/13PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:note> <ns0:note place='foot' n='10'>/13PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)</ns0:note> <ns0:note place='foot' n='13'>/13 PeerJ reviewing PDF | (2020:05:49464:2:1:NEW 14 Oct 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"1 Reviewer: Ivan Jukic The authors should be applauded for the great work with this manuscript which will, undoubtedly, facilitate a deeper understanding of different effect sizes commonly used in sport and exercise science. I appreciate the responses of the authors. The changes that the authors have made, based on the suggestions of their colleagues as well as the reviewers, just improved the overall readability of the manuscript and made the message that they wanted to convey more clear. Congratulations! Thank you for the kinds words. We hope future readers feel the same way and this leads to improved presentations of effect sizes in the literature. 1 2 Anthony Ciccone The proponents section should reflect the following: 1) the fact that small sample sizes do in fact affect ALL effect size metrics, including Hedges g. When I triple the following dataset: [pre:1,2,2,3] [post: 2,2,3,4], ANY effect size is around 0.1 larger. Obviously, this effect approaches zero as sample size increases, but it is misleading to continue to perpetuate the idea that effect sizes are not affected by sample size. We have trouble understanding this comment. While sample size will affect the variance of these estimators, as we describe in our paper, it is unclear why it should bias these estimators beyond the bias described by Hedges and others. If the reviewer has a reference, mathematical proof, or even Monte Carlo simulation, we will be happy to consider this point further. To concretely address the example provided, when duplicating the sample provided three times, the effect size increases slightly, as the reviewer suggests. However, after duplicating the sample 100x, there is no indication that the effect size approaches zero; rather, it stabilizes close to where it is when the sample is duplicated 3x (Figure 1). Mathematically, we see no reason why the effect size would approach zero and the reviewer has neither provided a mathematical reason (proof) nor reference. Proofs, which we reference throughout our paper, suggest the opposite. value 1.6 type 1.4 Delta dz 1.2 1.0 0 25 50 75 100 reps Figure 1: Duplicating the sample provided up to 100x produces a nonzero asymptotic result. Code to reproduce included at the end. While we appreciate the attempt in these comments to improve the quality of the manuscript, we do not believe this statement is an accurate reflection on bias corrections. Bias refers to a property of a statistic to over- or under-estimate of a population parameter on average. This means, in the long run, Cohen’s d (a biased estimator) would overestimate the standardized mean difference. To illustrate this point, we’ve simulated 1000 samples of n ∈ {5, 10, 15, ..., 100}. Each sample has a preintervention score of 0 ± 1, post-intervention score of 1 ± 1, and a pre-post correlation of 0.5. We’ve presented the uncorrected and bias-corrected results in Figure 2, where the red line is a generalized additive model (GAM). The SMDs perform as the math would suggest; on average, uncorrected ESs are upwardly biased with small samples but quickly converge to the true SMD = 1. Bias-corrected SMDs are unbiased for all sample sizes. Sample size affects the second moment, but not the first moment. 2 Uncorrected Bias−corrected 2.0 1.5 delta 1.0 SMD 0.5 0.0 2.0 1.5 dz 1.0 0.5 0.0 0 25 50 75 100 0 25 50 75 100 n Figure 2: Simulations of standardized mean differences (uncorrected and bias-corrected) across a range of sample sizes. Note, the red lines (GAMs) follow what is expected from Hedges (1981); that is, the uncorrected SMDs are upwardly biased with small sample sizes, but converge to the true SMD (1) quickly. Bias-corrected SMDs produce an unbiased estimate with small sample sizes. The variance depicted is due to sampling error, which is well-described in our paper. 3 2) percent change effect sizes cannot be linearly transformed to non-percent change effect sizes We currently discuss the limitations of percent changes in the Percent Changes section. In the Proponents section specifically, when we state that “[...] standardized effect sizes do not depend on the scale of the underlying outcome variable, and any linear transformation of the data will still yield the exact same standardized effect size,” we are not referring to percent change. We’ve not seen percent change be referred to as a standardized effect size, so we do not think this point is appropriate for this section. 4 1 2 3 library ( ggplot2 ) library ( tidyr ) library ( pbmcapply ) 4 5 6 7 # #### # TOY EXAMPLE # DUPLICATES SIZE 8 9 10 pre <- c (1 ,2 ,2 ,3) post <- c (2 ,2 ,3 ,4) 11 12 13 14 15 SMDs <- sapply (1:100 , function ( x ) { cbind ( mean ( rep ( post , x ) - rep ( pre , x ) ) / sd ( rep ( pre , x ) ) , mean ( rep ( post , x ) - rep ( pre , x ) ) / sd ( rep ( post , x ) - rep ( pre , x ) ) ) }) 16 17 18 19 20 SMDs <- as . data . frame ( t ( SMDs ) ) colnames ( SMDs ) <- c ( ' Delta ' ,' dz ' ) SMDs $ reps <- 1:100 SMDs <- gather ( SMDs , type , value , 1:2) 21 22 23 24 25 26 ggplot ( data = SMDs , aes ( x = reps , y = value , color = type ) ) + geom _ point () 27 28 ggsave ( ' sample _ dup . pdf ' , width = 6 , height = 3) 29 30 31 32 # #### # SIMULATE SAMPLE SIZES 5 -100 33 34 35 36 37 38 SMDsim <- lapply ( seq (5 ,100 , by =5) , function ( x ) { sims <- pbmclapply (1:1000 , function ( i ) { df <- MASS :: mvrnorm (x , c (0 ,1) , matrix ( c (1 ,0.5 ,0.5 ,1) , nrow =2) ) pre <- df [ ,1] post <- df [ ,2] 39 cbind ( delta = mean ( post - pre ) / sd ( pre ) , dz = mean ( post - pre ) / sd ( post - pre ) ) 40 41 }) sims <- do . call ( rbind . data . frame , sims ) sims $ n <- x sims <- gather ( sims , type , value , 1:2) 42 43 44 45 46 sims 47 48 }) 49 50 51 52 SMDsim <- do . call ( rbind . data . frame , SMDsim ) SMDsim $ g <- SMDsim $ value * ( 1 - 3 / (4 * ( SMDsim $n -1) -1) ) SMDsim <- gather ( SMDsim , corrected , value , 3:4) 53 54 55 56 SMDsim $ corrected <- factor ( SMDsim $ corrected , levels = unique ( SMDsim $ corrected ) , labels = c ( ' Uncorrected ' ,' Bias - corrected ' ) ) 57 58 59 60 61 62 63 64 65 ggplot ( data = SMDsim , aes ( x = n , y = value ) ) + geom _ jitter ( alpha = 0.01) + facet _ grid ( type ~ corrected ) + geom _ smooth ( color = ' red ' ) + ylab ( ' SMD ' ) + coord _ cartesian ( ylim = c (0 ,2) ) 66 5 67 ggsave ( ' simulated . pdf ' , width = 6 , height = 4) 6 "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Adaptive behavior emerges through a dynamic interaction between cognitive agents and changing environmental demands. The investigation of information processing underlying adaptive behavior relies on controlled experimental settings in which individuals are asked to accomplish demanding tasks whereby a hidden regularity or an abstract rule has to be learned dynamically. Although performance in such tasks is considered as a proxy for measuring high-level cognitive processes, the standard approach consists in summarizing observed response patterns by simple heuristic scoring measures. With this work, we propose and validate a new computational Bayesian model accounting for individual performance in the Wisconsin Card Sorting Test (WCST), a renowned clinical tool to measure set-shifting and deficient inhibitory processes on the basis of environmental feedback. We formalize the interaction between the task's structure, the received feedback, and the agent's behavior by building a model of the information processing mechanisms used to infer the hidden rules of the task environment. Furthermore, we embed the new model within the mathematical framework of the Bayesian Brain Theory (BBT), according to which beliefs about hidden environmental states are dynamically updated following the logic of Bayesian inference. Our computational model maps distinct cognitive processes into separable, neurobiologically plausible, information-theoretic constructs underlying observed response patterns. We assess model identification and expressiveness in accounting for meaningful human performance through extensive simulation studies. We then validate the model on real behavioral data in order to highlight the utility of the proposed model in recovering cognitive dynamics at an individual level.</ns0:p><ns0:p>We highlight the potentials of our model in decomposing adaptive behavior in the WCST into several information-theoretic metrics revealing the trial-by-trial unfolding of information processing by focusing on two exemplary individuals whose behavior is examined in depth. Finally, we focus on the theoretical implications of our computational model by discussing the mapping between BBT constructs and functional neuroanatomical</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION 37</ns0:head><ns0:p>Computational models of cognition provide a way to formally describe and empirically account for mech-38 anistic, process-based theories of adaptive cognitive functioning <ns0:ref type='bibr' target='#b59'>(Sun (2009)</ns0:ref>; <ns0:ref type='bibr' target='#b14'>Cooper et al. (1996)</ns0:ref>; Lee 39 and Wagenmakers (2014)). A foundational theoretical framework for describing functional characteristics 40 of neurocognitive systems has recently emerged under the hood of Bayesian brain theories <ns0:ref type='bibr'>(Knill and</ns0:ref> 41 <ns0:ref type='bibr' target='#b32'>Pouget (2004)</ns0:ref>; <ns0:ref type='bibr' target='#b23'>Friston (2010)</ns0:ref>). Bayesian brain theories owe their name to their core assumption that beliefs, about the causal structure of events in the environment <ns0:ref type='bibr' target='#b22'>(Friston (2005)</ns0:ref>; <ns0:ref type='bibr' target='#b32'>Knill and Pouget (2004))</ns0:ref> and forms a basis for adaptive behavior. It is assumed that internal beliefs are constantly updated and refined to match the current state of the world as new observations become available. The core idea behind the Bayesian brain hypothesis is that computational mechanisms underlying such an internal belief updating follow the logic of Bayesian probability theory. In this respect, information about the external world provided by sensory inputs is represented as a conditional probability distribution over a set of environmental states. Consequently, the brain relies on this probabilistic representation of the world to infer the most likely environmental causes (states) which generate those inputs, and such a process follows the computational principles of Bayesian inference <ns0:ref type='bibr' target='#b26'>(Friston and Kiebel (2009)</ns0:ref>; <ns0:ref type='bibr' target='#b23'>Friston (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b12'>Buckley et al. (2017)</ns0:ref>).</ns0:p><ns0:p>To clarify this concept, consider a simple example of a perceptual task in which a cognitive agent is required to judge whether an item depicted on a flat plane is concave or convex. Its judgment is based solely on the basis of a set of observed perceptual features, such as, shape, orientation, texture and brightness. Here, the concave-to-convex gradient entails the set of environmental states which must be inferred. The internal generative model of the agent codifies beliefs about how different degrees of convexity might give rise to certain configurations of perceptual inputs. From a Bayesian perspective, the problem is solved by inverting the generative model of the environment in order to turn assumptions about how environmental states generate sensory inputs into beliefs about the most likely states (e.g., degree of convexity) given the available sensory information.</ns0:p><ns0:p>Potentially, there are no limitations regarding the complexity of environmental settings (e.g., items and rules in experimental tasks) and cognitive processes to be described in light of the Bayesian brain framework. Indeed, the latter has proven to be a consistent computational modeling paradigm for the investigation of a variety of neurocognitive mechanisms, such as motor control (?), oculomotor dynamics (?), object recognition (?), attention (?), perceptual inference (? <ns0:ref type='bibr' target='#b32'>Knill and Pouget (2004)</ns0:ref>), multisensory integration (?), as well as for providing a foundational theoretical account of general neural systems' functioning (? <ns0:ref type='bibr' target='#b22'>Friston (2005</ns0:ref><ns0:ref type='bibr' target='#b21'>Friston ( , 2003))</ns0:ref>) and complex clinical scenarios such as Schizophrenia (?), and Autistic Spectrum Disorder (??). For this reason, such a modeling approach might provide a comprehensive and unified framework under which several cognitive impairments can be measured and understood in the light of a general process-based theory of neural functioning.</ns0:p><ns0:p>In this work, we address the challenging problem of modeling adaptive behavior in a dynamic environment. The empirical assessment of adaptive functioning often relies on dynamic reinforcement learning scenarios which require participants to adapt their behavior during the unfolding of a (possibly) demanding task. Typically, these tasks are designed with the aim to figure out how adaptive behavior unfolds through multiple trials as participants observe certain environmental contingencies, take actions, and receive feedback based on their actions. From a Bayesian theoretical perspective, optimal performance in such adaptive experimental paradigms require that agents infer the probabilistic model underlying the hidden environmental states. Since these models usually change as the task progresses, agents, in turn, need to adapt their inferred model, in order to take optimal actions.</ns0:p><ns0:p>Here, we propose and validate a computational Bayesian model which accounts for the dynamic behavior of cognitive agents in the Wisconsin Card Sorting Test (WCST; <ns0:ref type='bibr' target='#b6'>Berg (1948)</ns0:ref>; <ns0:ref type='bibr' target='#b29'>Heaton (1981)</ns0:ref>), which is perhaps the most widely adopted neuropsychological setting employed to investigate adaptive functioning, due to its specificity in accounting for executive components underlying observed behavior, such as set-shifting, cognitive flexibility and impulsive response modulation <ns0:ref type='bibr' target='#b8'>(Bishara et al. (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b0'>Alvarez and Emory (2006)</ns0:ref>). For this reason, we consider the WCST as a fundamental paradigm for investigating adaptive behavior from a Bayesian perspective.</ns0:p><ns0:p>The environment of the WCST consists of a target and a set of stimulus cards with geometric figures which vary according to three perceptual features. The WCST requires participants to infer the correct classification principle by trial and error using the examiner's feedback. The feedback is thought to carry a positive or negative information signaling the agent whether the immediate action was appropriate or not. Modeling adaptive behavior in the WCST from a Bayesian perspective is straightforward, since observable actions emerge from the interaction between the internal probabilistic model of the agent and a set of discrete environmental states.</ns0:p><ns0:p>Performance in WCST is usually measured via a rough summary metric such as the number of correct/incorrect responses or pre-defined psychological scoring criteria (see for instance <ns0:ref type='bibr' target='#b29'>Heaton (1981)</ns0:ref>).</ns0:p><ns0:p>These metrics are then used to infer the underlying cognitive processes involved in the task. A major shortcoming of this approach is that it simply assumes the cognitive processes to be inferred without specifying an explicit process model. Moreover, summary measures do not utilize the full information present in the data, such as trial-by-trial fluctuations or various interesting agent-environment interactions.</ns0:p><ns0:p>For this reason, crude scoring measures are often insufficient to disentangle the dynamics of the relevant cognitive (sub)processes involved. Consequently, an entanglement between processes at the metric level can prevent us from answering interesting research questions about aspects of adaptive behavior.</ns0:p><ns0:p>In our view, a sound computational account for adaptive behavior in the WCST needs to provide at least a quantitative measure of effective belief updating about the environmental states at each trial. This measure should be complemented by a measure of how feedback-related information influences behavior. The first measure should account for the integration of meaningful information. In other words, it should describe how prior beliefs about the current environmental state change after an observation has been made. The second measure should account for signaling the (im)probability of observing a certain environmental configuration (e.g., an (un)expected feedback given a response) <ns0:ref type='bibr' target='#b53'>(Schwartenbeck et al. (2016)</ns0:ref>).</ns0:p><ns0:p>Indeed, recent studies suggest that the meaningful information content and the pure unexpectedness of an observation are processed differently at the neural level. Moreover, such disentanglement appears to be of crucial importance to the understanding of how new information influences adaptive behavior <ns0:ref type='bibr' target='#b42'>(Nour et al. (2018)</ns0:ref>; <ns0:ref type='bibr' target='#b53'>Schwartenbeck et al. (2016);</ns0:ref><ns0:ref type='bibr' target='#b44'>O'Reilly et al. (2013)</ns0:ref>). Inspired by these results and previous computational proposals <ns0:ref type='bibr' target='#b33'>(Koechlin and Summerfield (2007)</ns0:ref>), we integrate these different information processing aspects into the current model from an information-theoretic perspective.</ns0:p><ns0:p>Our computational cognitive model draws heavily on the mathematical frameworks of Bayesian probability theory and information theory <ns0:ref type='bibr' target='#b51'>(Sayood (2018)</ns0:ref>). First, it provides a parsimonious description of observed data in the WCST via two neurocognitively meaningful parameters, namely, flexibility and information loss (to be motivated and explained in the Model section). Moreover, it captures the main response patterns obtainable in the WCST via different parameter configurations. Second, we formulate a functional connection between cognitive parameters and underlying information processing mechanisms related to belief updating and prediction formation. We formalize and distinguish between Bayesian surprise and Shannon surprise as the main mechanisms for adaptive belief updating. Moreover, we introduce a third quantity, which we named predictive Entropy and which quantifies an agent's subjective uncertainty about the current internal model. Finally, we propose to measure these quantities on a trial-bytrial basis and use them as a proxy for formally representing the dynamic interplay between agents and environments.</ns0:p><ns0:p>The rest of the paper is organized as follows. First, the WCST is described in more detail and a mathematical representation of the new Bayesian computational model is provided. Afterwards, we explore the model's characteristics through simulations and perform parameter recovery on simulated data using a powerful Bayesian deep neural network method <ns0:ref type='bibr' target='#b47'>(Radev et al. (2020)</ns0:ref>). We then apply the model to real behavioral data from an already published dataset. Finally, we discuss the results as well as the main strengths and limitations of the proposed model.</ns0:p></ns0:div> <ns0:div><ns0:head>THE WISCONSIN CARD SORTING TEST</ns0:head><ns0:p>In a typical WCST <ns0:ref type='bibr' target='#b29'>(Heaton (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b6'>Berg (1948)</ns0:ref>), participants learn to pay attention and respond to relevant stimulus features, while ignoring irrelevant ones, as a function of experimental feedback. In particular, Individuals are asked to match a target card with one of four stimulus cards according to a proper sorting principle, or sorting rule. Each card depicts geometric figures that vary in terms of three features, namely, color (red, green, blue, yellow), shape (triangle, star, cross, circle) and number of objects (1, 2, 3 and 4). For each trial, the participant is required to identify the sorting rule which is valid for that trial, that is, which of the three feature has to be considered as a criterion to matching the target card with the right stimulus card (see Figure <ns0:ref type='figure'>1</ns0:ref>). Noticed that both features and sorting rules refer to the same concept. However, the feature still codifies a property of the card, whilst the sorting rule refers to the particular feature which is valid for the current trial.</ns0:p></ns0:div> <ns0:div><ns0:head>3/24</ns0:head><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref>. Suppose that the current sorting rule is the feature shape. The target card in the first trial (left box) contains two blue triangles. A correct response requires that the agent matches the target card with the stimulus card containing the single triangle (arrow represents the correct choice), regardless of the features color and number. The same applies for the second trial (right box) in which matching the target card with the stimulus card containing three yellow crosses is the correct response.</ns0:p><ns0:p>Each response in the WCST is followed by a feedback informing the participant if his/her response is correct or incorrect. After some fixed number of consecutive responses, the sorting rule is changed by the experimenter without warning, and participants are required to infer the new sorting rule. Clearly, the most adaptive response would be to explore the remaining possible rules. However, participants sometimes would persist responding according to the old rule and produce what is called a perseverative response.</ns0:p></ns0:div> <ns0:div><ns0:head>METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>The Model</ns0:head><ns0:p>The core idea behind our computational framework is to encode the concept of belief into a generative probabilistic model of the environment. Belief updating then corresponds to recursive Bayesian updating of the internal model based on current and past interactions between the agent and its environment.</ns0:p><ns0:p>Optimal or sub-optimal actions are selected according to a well specified or a misspecified internal model and, in turn, cause perceptible changes in the environment.</ns0:p><ns0:p>We assume that the cognitive agent aims to infer the true hidden state of the environment by processing and integrating sensory information from the environment. Within the context of the WCST, the hidden environmental states might change as a function of both the structure of the task and the (often suboptimal) behavioral dynamics, so the agent constantly needs to rely on environmental feedback and own actions to infer the current state. We assume that the agent maintains an internal probability distribution over the states at each individual trial of the WCST. The agent then updates this distribution upon making new observations. In particular, the hidden environmental states to be inferred are the three features, s t &#8712; {1, 2, 3}, which refer the three possible sorting rules in the task environment such that 1: color, 2: shape and 3: number of objects. The posterior probability of the states depends on an observation vector x t = (a t , f t ), which consists of the pair of agent's response a t &#8712; {1, 2, 3, 4}, codifying the action of choosing deck 1, 2, 3 or 4, and received feedback f t &#8712; {0, 1}, referring to the fact that a given response results in a failure (0) or in a success (1), in a given trial t = 0, ..., T . The discrete response a t represents the stimulus card indicator being matched with a target card at trial t. We denote a sequence of observations as x 0:t = (x 0 , x 1 , ..., x t ) = ((a 0 , f 0 ), (a 1 , f 1 ), (a 2 , f 2 ), ..., (a t , f t )) and set x 0 = &#8709; in order to indicate that there are no observations at the onset of the task. Thus, trial-by-trial belief updating is recursively computed according to Bayes' rule:</ns0:p><ns0:formula xml:id='formula_0'>p(s t |x 0:t ) = p(x t |s t , x 0:t&#8722;1 )p(s t |x 0:t&#8722;1 ) p(x t |x 0:t&#8722;1 )<ns0:label>(1)</ns0:label></ns0:formula><ns0:p>Accordingly, the agent's posterior belief about the task-relevant features s t after observing a sequence of response-feedback pairs x 0:t is proportional to the product of the likelihood of observing a particular 4/24 response-feedback pair and the agent's prior belief about the task-relevant feature in the current trial. The likelihood of an observation is computed as follows:</ns0:p><ns0:formula xml:id='formula_1'>p(x t |s t , x 0:t&#8722;1 ) = f t p(a t |s t = i) + (1 &#8722; f t )(1 &#8722; p(a t |s t = i)) f t &#8721; j p(a t |s t = j) + (1 &#8722; f t ) &#8721; j (1 &#8722; p(a t |s t = j))</ns0:formula><ns0:p>(2)</ns0:p><ns0:p>where j = 1, 2, 3 and p(a t |s t = i) indicates the probability of a matching between the target and the stimulus card assumed that the current feature is i. Here, we assume the likelihood of a current observation to be independent from previous observations without loss of generality, that is:</ns0:p><ns0:formula xml:id='formula_2'>p(x t |s t , x 0:t&#8722;1 ) = p(x t |s t )</ns0:formula><ns0:p>The prior belief for a given trial t is computed based on the posterior belief generated in the previous trial, p(s t&#8722;1 |x 0:t&#8722;1 ), and the agent's belief about the probability of transitions between the hidden states, p(s t |s t&#8722;1 ). The prior belief can also be considered as a predictive probability over the hidden states.</ns0:p><ns0:p>The predictive distribution for an upcoming trial t is computed according to the Chapman-Kolmogorov equation:</ns0:p><ns0:formula xml:id='formula_3'>p(s t+1 = k|x 0:t ) = 3 &#8721; i=1 p(s t+1 = k|s t = i, &#915;(t))p(s t = i|x 0:t )<ns0:label>(3)</ns0:label></ns0:formula><ns0:p>where &#915;(t) represents a stability matrix describing transitions between the states (to be explained shortly).</ns0:p><ns0:p>Thus, the agent combines information from the updated belief (posterior distribution) and the belief about the transition properties of the environmental states to predict the most probable future state. The predictive distribution represents the internal model of the cognitive agent according to which actions are generated.</ns0:p><ns0:p>The stability matrix &#915;(t) encodes the agent's belief about the probability of states being stable or likely to change in the next trial. In other words, the stability matrix reflects the cognitive agent's internal representation of the dynamic probabilistic model of the task environment. It is computed on each trial based on the response-feedback pair, x t , and a matching signal, m t , which are observed.</ns0:p><ns0:p>The matching signal m t is a vector informing the cognitive agent which features are currently relevant (meaningful), such that m (i) t = 1 when a positive feedback is associated with a response implying feature s t = i, and m (i) t = 0 otherwise. Note, that the matching signal is not a free parameter of the model, but is completely determined by the task contingencies. The matching signal vector allows the agent to compute the state activation level &#969; (i)</ns0:p><ns0:p>t &#8712; [0, 1] for the hidden state s t = i, which provides an internal measure of the (accumulated) evidence for each hidden state at trial t. Thus, the activation levels of the hidden states are represented by a vector &#969; t . The stability matrix is a square and asymmetric matrix related to hidden state activation levels such that:</ns0:p><ns0:formula xml:id='formula_4'>&#915;(t) = &#63726; &#63727; &#63727; &#63727; &#63728; &#969; (1) t 1 2 (1 &#8722; &#969; (1) t ) 1 2 (1 &#8722; &#969; (1) t ) 1 2 (1 &#8722; &#969; (2) t ) &#969; (2) t 1 2 (1 &#8722; &#969; (2) t ) 1 2 (1 &#8722; &#969; (3) t ) 1 2 (1 &#8722; &#969; (3) t ) &#969; (3) t &#63737; &#63738; &#63738; &#63738; &#63739; (4)</ns0:formula><ns0:p>where the entries &#915; ii (t) in the main diagonal represent the elements of the activation vector &#969; t , and the non-diagonal elements are computed so as to ensure that rows sum to 1. The state activation vector is computed in each trial as follows:</ns0:p><ns0:formula xml:id='formula_5'>&#63726; &#63727; &#63728; &#969; (1) t &#969; (2) t &#969; (3) t &#63737; &#63738; &#63739; = f t &#969; &#948; t&#8722;1 &#63726; &#63727; &#63728; m (1) t m (2) t m (3) t &#63737; &#63738; &#63739; + &#955; &#63726; &#63727; &#63728;(1 &#8722; f t )&#969; &#948; t&#8722;1 &#63726; &#63727; &#63728; 1 &#8722; m (1) t 1 &#8722; m (2) t 1 &#8722; m (3) t &#63737; &#63738; &#63739; &#63737; &#63738; &#63739; &#63726; &#63727; &#63728; &#969; (1) t&#8722;1 &#969; (2) t&#8722;1 &#969; (3) t&#8722;1 &#63737; &#63738; &#63739; .</ns0:formula><ns0:p>(5)</ns0:p></ns0:div> <ns0:div><ns0:head>5/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>This equation reflects the idea that state activations are simultaneously affected by the observed feedback, f t , and the matching signal vector, m t . However, the matching signal vector conveys different information based on the current feedback. Matching a target card with a stimulus card makes a feature (or a subset of features) informative for a specific state. The vector m t contributes to increase the activation level of a state if the feature is informative for that state when a positive feedback is received, as well as to decrease the activation level when a negative feedback is received.</ns0:p><ns0:p>The parameter &#955; &#8712; [0, 1] modulates the efficiency to disengage attention to a given state-activation configuration when a negative feedback is processed. We therefore term this parameter flexibility. We also assume that information from the matching signal vector can degrade by slowing down the rate of evidence accumulation for the hidden states. This means that the matching signal vector can be re-scaled based on the current state activation level. The parameter &#948; &#8712; [0, 1] is introduced to achieve this re-scaling. When &#948; = 0, there is no re-scaling and updating of the state activation levels relies on the entire information conveyed by m t . On the other extreme, when &#948; = 1, several trials have to be accomplished before converging to a given configuration of the state activation levels. Equivalently, higher values of &#948; affect the entropy of the distribution over hidden states by decreasing the probability of sampling of the correct feature. We therefore refer to &#948; as information loss.</ns0:p><ns0:p>The free parameters &#955; and &#948; are central to our computational model, since they regulate the rate at which the internal model converges to the true task environmental model. Eq. ( <ns0:ref type='formula'>5</ns0:ref>) can be expressed in compact notation as follows:</ns0:p><ns0:formula xml:id='formula_6'>&#969; t = f t &#969; &#948; t&#8722;1 m t + &#955; (1 &#8722; f t )&#969; &#948; t&#8722;1 (1 &#8722; m t ) &#969; t&#8722;1 (6)</ns0:formula><ns0:p>Note that the information loss parameter &#948; affects the amount of information that a cognitive agent acquires from environmental contingencies, irrespective of the type of feedback received. Global information loss thus affects the rate at which the divergence between the agent's internal model and the true model is minimized. Figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref> illustrates these ideas.</ns0:p><ns0:p>The probabilistic representation of adaptive behaviour provided by our Bayesian agent model allows us to quantify latent cognitive dynamics by means of meaningful information-theoretic measures. Information theory has proven to be an effective and natural mathematical language to account for functional integration of structured cognitive processes and to relate them to brain activity <ns0:ref type='bibr' target='#b33'>(Koechlin and Summerfield (2007)</ns0:ref>; <ns0:ref type='bibr' target='#b24'>Friston et al. (2017)</ns0:ref>; <ns0:ref type='bibr' target='#b13'>Collell and Fauquet (2015)</ns0:ref>; <ns0:ref type='bibr' target='#b58'>Strange et al. (2005)</ns0:ref>; <ns0:ref type='bibr' target='#b21'>Friston (2003)</ns0:ref>). In particular, we are interested in three key measures, namely, Bayesian surprise, B t , Shannon surprise, I t , and entropy, H t . The subscript t indicates that we can compute each quantity on a trial-by-trial basis. Each quantity is amenable to a specific interpretation in terms of separate neurocognitive processes. Bayesian surprise B t quantifies the magnitude of the update from prior belief to posterior belief. Shannon surprise I t quantifies the improbability of an observation given an agent's prior expectation. Finally, entropy H t measures the degree of epistemic uncertainty regarding the true environmental states. Such measures are thought to account for the ability of the agent to manage uncertainty as emerging as a function of competing behavioral affordances <ns0:ref type='bibr' target='#b30'>(Hirsh et al. (2012)</ns0:ref>). We expect an adaptive system to attenuate uncertainty over environmental states (current features) by reducing the entropy of its internal probabilistic model.</ns0:p><ns0:p>Bayesian surprise can be computed as the Kullback-Leibler (KL) divergence between prior and posterior beliefs about the environmental states. Thus, Bayesian surprise accounts for the divergence between the predictive model for the current trial and the updated predictive model for the upcoming trial.</ns0:p><ns0:p>It is computed as follows:</ns0:p><ns0:formula xml:id='formula_7'>B t = KL[p(s t+1 |x 0:t )||p(s t |x 0:t&#8722;1 )] = 3 &#8721; i=1 p(s t+1 = i|x 0:t ) log p(s t+1 = i|x 0:t ) p(s t = i|x 0:t&#8722;1 )<ns0:label>(7)</ns0:label></ns0:formula><ns0:p>The Shannon surprise of a current observation given a previous one is computed as the conditional Manuscript to be reviewed </ns0:p><ns0:formula xml:id='formula_8'>6/24</ns0:formula><ns0:formula xml:id='formula_9'>(i) t &#8712; [0, 1]</ns0:formula><ns0:p>State activation level Agent's internal measure of the accrued evidence for the hidden environmental state i in trial t.</ns0:p><ns0:formula xml:id='formula_10'>B t &#8712; R +</ns0:formula><ns0:p>Bayesian surprise Kullback-Leibler divergence between prior and posterior beliefs about hidden environmental states in trial t.</ns0:p><ns0:formula xml:id='formula_11'>I t &#8712; R +</ns0:formula><ns0:p>Shannon surprise Information-theoretic surprise encoding the improbability or unexpectedness of an observation in trial t.</ns0:p><ns0:formula xml:id='formula_12'>H t &#8712; R +</ns0:formula><ns0:p>Entropy Degree of epistemic uncertainty in the internal model of the environment in trial t.</ns0:p><ns0:p>Table <ns0:ref type='table'>1</ns0:ref>. Descriptive summary of all quantities involved in our model representation.</ns0:p><ns0:p>information content of the observation:</ns0:p><ns0:formula xml:id='formula_13'>I t = &#8722; log p(x t |x 0:t&#8722;1 ) = &#8722; log 3 &#8721; i=1 [p(x t |s t = i)p(s t = i|x 0:t&#8722;1 )]<ns0:label>(8)</ns0:label></ns0:formula><ns0:p>Finally, the entropy is computed over the predictive distribution in order to account for the uncertainty in the internal model of the agent in trial t as follows:</ns0:p><ns0:formula xml:id='formula_14'>H t = E [&#8722; log p(s t |x 0:t&#8722;1 )] = &#8722; 3 &#8721; i=1 p(s t = i|x 0:t&#8722;1 ) log p(s t = i|x 0:t&#8722;1 )<ns0:label>(9)</ns0:label></ns0:formula><ns0:p>Once the flexibility (&#955; ) and information loss (&#948; ) parameters are estimated from data, the informationtheoretic quantities can be easily computed and visualized for each trial of the WCST (see Figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>).</ns0:p><ns0:p>This allows to rephrase standard neurocognitive constructs in terms of measurable information-theoretic quantities. Moreover, the dynamics of these quantities, as well as their interactions, can be used for formulating and testing hypotheses about the neurcognitive underpinnings of adaptive behavior in a principled way, as discussed later in the paper. A summary of all quantities relevant for our computational model is provided in Table <ns0:ref type='table'>1</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>8/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Simulations</ns0:head><ns0:p>In this section we evaluate the expressiveness of the model by assessing its ability to reproduce meaningful behavioral patterns as a function of its two free parameters. We study how the generative model behaves when performing the WCST in a 2-factorial simulated Monte Carlo design where flexibility (&#955; ) and information loss (&#948; ) are systematically varied.</ns0:p><ns0:p>In this simulation, the Heaton version of the task <ns0:ref type='bibr' target='#b29'>(Heaton (1981)</ns0:ref>) is administered to the Bayesian cognitive agent. In this particular version, the sorting rule (true environmental state) changes after a fixed number of consecutive correct responses. In particular, when the agent correctly matches the target card in 10 consecutive trials, the sorting rule is automatically changed. The task ends after completing a maximum of 128 trials.</ns0:p></ns0:div> <ns0:div><ns0:head>Generative Model</ns0:head><ns0:p>The cognitive agent's responses are generated at each time step (trial) by processing the experimental feedback. Its performance depends on the parameters governing the computation of the relevant quantities.</ns0:p><ns0:p>The generative algorithm is outlined in Algorithm 1.</ns0:p><ns0:p>Algorithm 1 Bayesian cognitive agent 1: Set parameters &#952; = (&#955; , &#948; ).</ns0:p><ns0:p>2: Set initial activation levels &#969; 0 = (0.5, 0.5, 0.5).</ns0:p><ns0:p>3: Set initial observation x 0 = &#8709; and p(s 1 |x 0 ) = p(s 1 ). 4: for t = 1, ..., T do 5:</ns0:p><ns0:p>Sample feature from prior/predictive internal model s t &#8764; p(s t |x 0:t&#8722;1 ).</ns0:p></ns0:div> <ns0:div><ns0:head>6:</ns0:head><ns0:p>Obtain a new observation x t = (a t , f t ).</ns0:p></ns0:div> <ns0:div><ns0:head>7:</ns0:head><ns0:p>Compute state posterior p(s t |x 0:t ).</ns0:p></ns0:div> <ns0:div><ns0:head>8:</ns0:head><ns0:p>Compute new activation levels &#969; t . 9:</ns0:p><ns0:p>Compute stability matrix &#915;(t).</ns0:p><ns0:p>10:</ns0:p><ns0:p>Update prior/predictive internal model to p(s t+1 |x 0:t ). 11: end for Simulation 1: Clinical Assessment of the Bayesian Agent Ideally, the qualitative performance of the Bayesian cognitive agent will resemble human performance. To this aim, we adopt a metric which is usually employed in clinical assessment of test results in neurological and psychiatric patients <ns0:ref type='bibr' target='#b10'>(Braff et al. (1991)</ns0:ref>; <ns0:ref type='bibr' target='#b64'>Zakzanis (1998)</ns0:ref>; <ns0:ref type='bibr' target='#b5'>Bechara and Damasio (2002)</ns0:ref>; Landry and Al-Taie ( <ns0:ref type='formula'>2016</ns0:ref>)). Thus, agent performance is codified according to a neuropsychological criterion <ns0:ref type='bibr' target='#b29'>(Heaton (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b19'>Flashman et al. (1991)</ns0:ref>) which allows to classify responses into several response types. These response types provide the scoring measures for the test.</ns0:p><ns0:p>Here, we are interested in: 1) non-perseverative errors (E); 2) perseverative errors (PE); 3) number of trials to complete the first category (TFC); and 4) number of failures to maintain set (FMS). Perseverative errors occur when the agent applies a sorting rule which was valid before the rule has been changed.</ns0:p><ns0:p>Usually, detecting a perseveration error is far from trivial, since several response configurations could be observed when individuals are required to shift a sorting rule after completing a category (see <ns0:ref type='bibr' target='#b19'>Flashman et al. (1991)</ns0:ref> for details). On the other hand, non-perseverative errors refer to all errors which do not fit the above description, or in other words, do not occur as a function of changing the sorting rule, such as casual errors.</ns0:p><ns0:p>The number of trials to complete the first category tells us how many trials the agent needs in order to achieve the first sorting principle, and can be seen as an index of conceptual ability <ns0:ref type='bibr' target='#b2'>(Anderson (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b54'>Singh et al. (2017)</ns0:ref>). Finally, a failure to maintain a set occurs when the agent fails to match cards according to the sorting rule after it can be determined that the agent has acquired the rule. A given sorting rule is assumed to be acquired when the individual correctly sorts at least five cards in a row <ns0:ref type='bibr' target='#b29'>(Heaton (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b17'>Figueroa and Youmans (2013)</ns0:ref>). Thus, a failure to maintain a set arises whenever a participant suddenly changes the sorting strategy in the absence of negative feedback. Failures to maintain a set are mostly attributed to distractibility. We compute this measure by counting the occurrences of first errors after the acquisition of a rule.</ns0:p><ns0:p>We run the generative model by varying flexibility across four levels, &#955; &#8712; {0.3, 0.5, 0.7, 0.9}, and information loss across three levels, &#948; &#8712; {0.4, 0.7, 0.9}. We generate data from 150 synthetic cognitive agents The simulated performance of our Bayesian cognitive agents demonstrates that different parameter combinations capture different meaningful behavioral patterns. In other words, flexibility and information loss seem to interact in a theoretically meaningful way.</ns0:p><ns0:p>First, overall errors increase when flexibility (&#955; ) decreases, which is reflected by the inverse relation between the number of casual, as well as perseverative, errors and the values of parameter &#955; . Moreover, this pattern is consistent across all the levels of parameter &#948; . More precisely, information loss (&#948; ) seems to contribute to the characterization of the casual and the perseverative components of the error in a different way. Perseverative errors are likely to occur after a sorting rule has changed and reflect the inability of the agent to use feedback to disengage attention from the currently attended feature. They therefore result from local cognitive dynamics conditioned on a particular stage of the task (e.g., after completing a series of correct responses).</ns0:p><ns0:p>Second, information loss does not interact with flexibility when perseverative errors are considered. This is due to the fact that high information loss affects general performance by yielding a dysfunctional response strategy which increases the probability of making an error at any stage of the task. The lack of such interaction provides evidence that our computational model can disentangle between error patterns due to perseveration and those due to general distractibility, according to neuropsychological scoring criteria.</ns0:p><ns0:p>However, in our framework, flexibility (&#955; ) is allowed to yield more general and non-local cognitive dynamics as well. Indeed, &#955; plays a role whenever belief updating is demanded as a function of negative feedback. An error classified as non-perseverative (e.g., casual error) by the scoring criteria might still be processed as a feedback-related evidence for belief updating. Consistently, the interaction between &#955; and &#948; in accounting for causal errors shows that performance worsens when both flexibility and information loss become less optimal, and that such pattern becomes more pronounced for lower values of &#948; .</ns0:p><ns0:p>On the other hand, a specific effect of information loss (&#948; ) can be observed for the scoring measures related to slow information processing and distractibility. The number of trials to achieve the first category reflects the efficiency of the agent in arriving at the first true environmental model. Flexibility does not contribute meaningfully to the accumulation of errors before completing the first category for some levels of information loss. This is reflected by the fact that the mean number of trials increases as a function of &#948; , and do not change across levels of &#955; for low and mid values of &#948; . A similar pattern applies for failures to maintain a set. Both scoring measures index a deceleration of the process of evidence accumulation for Manuscript to be reviewed a specific environmental configuration, although the latter is a more exhaustive measures of dysfunctional adaptation.</ns0:p><ns0:p>Therefore, an interaction between parameters can be observed when information loss is high. A slow internal model convergence process increases the amount of errors due to improper rule sampling from the internal environmental model. However, internal model convergence also plays a role when a new category has to be accomplished after completing an older one. On the one hand, compromised flexibility increases the amount of errors due to inefficient feedback processing. This leads to longer trial windows needed to achieve the first category. On the other hand, when information loss is high, belief updating upon negative feedback is compromised due to high internal model uncertainty. At this point, the probability to err due to distractibility increases, as accounted by the failures to maintain a set measures.</ns0:p><ns0:p>Finally, the joint effect of &#948; and &#955; for high levels of information loss suggests that the roles played by the two cognitive parameters in accounting for adaptive functioning can be entangled when neuropsychological scoring criteria are considered.</ns0:p><ns0:p>Simulation 2: Information-Theoretic Analysis of the Bayesian Agent</ns0:p><ns0:p>In the following, we explore a different simulation scenario in which information-theoretic measures are derived to assess performance of the Bayesian cognitive agent. In particular, we explore the functional relationship between cognitive parameters and the dynamics of the recovered information-theoretic measures by simulating observed responses by varying flexibility across three levels, &#955; &#8712; {0.1, 0.5, 0.9}, and information loss across three levels, &#948; &#8712; {0.1, 0.5, 0.9}.</ns0:p><ns0:p>For this simulation scenario, we make no prior assumptions about sub-types of error classification.</ns0:p><ns0:p>Instead, we investigate the dynamic interplay between Bayesian surprise, B t , Shannon surprise, I t , and entropy, H t over the entire course of 128 trials in the WCST. In general, low information loss (&#948; ) ensures optimal behavior by speeding up internal model convergence by decreasing the number of trials needed to minimize uncertainty about the environmental states.</ns0:p><ns0:p>Low uncertainty reflects two main aspects of adaptive behavior. On the one hand, the probability that a response occurs due to sampling of improper rules decreases, allowing the agent to prevent random responses due to distractibility. On the other hand, model convergence entails a peaked Shannon surprise when a negative feedback occurs, due to the divergence between predicted and actual observations. Flexibility (&#955; ) plays a crucial role in integrating feedback information in order to enable belief updating. The first row depicted in Figure <ns0:ref type='figure' target='#fig_3'>4</ns0:ref> shows cognitive dynamics related to low information loss, across the levels of flexibility. As can be noticed, there is a positive relation between the magnitude of the Bayesian surprise and the level of flexibility, although unexpectedness yields approximately the same amount of signaling, as accounted by peaked Shannon surprise. From this perspective, surprise and belief updating can be considered functionally separable, where the first depends on the particular internal model probability configuration related to &#948; , whilst the second depends on flexibility &#955; .</ns0:p><ns0:p>However, more interesting patterns can be observed when information loss increases. In particular, model convergence slows down and several trials are needed to minimize predictive model entropy.</ns0:p><ns0:p>Casual errors might occur within trial windows characterized by high uncertainty, and interactions between entropy and Shannon surprise can be observes in such cases. In particular, Shannon surprise magnitude increases when model's entropy decreases, that is, during task phases in which the internal model has already converged. As a consequence, negative feedback could be classified as informative or uninformative, based on the uncertainty in the current internal model. This is reflected by the negative relation between entropy and Shannon surprise, as can be noticed by inspecting the graphs depicted in the third row of Figure <ns0:ref type='figure' target='#fig_3'>4</ns0:ref>. Therefore, the magnitude of belief updating depends on the interplay between entropy and Shannon surprise, and can differ based on the values of the two measures in a particular task phase.</ns0:p><ns0:p>In sum, both simulation scenarios suggest that the simulated behavior of our generative model is in accord with theoretical expectations. Moreover, the flexibility and information loss parameters can account for a wide range of observed response patterns and inferred dynamics of information processing.</ns0:p></ns0:div> <ns0:div><ns0:head>Parameter Estimation</ns0:head><ns0:p>In this section, we discuss the computational framework for estimating the parameters of our model from observed behavioral data. Parameter estimation is essential to inferring the cognitive dynamics underlying observed behavior in real-world applications of the model. This section is slightly more technical and can be skipped without significantly affecting the flow of the text.</ns0:p></ns0:div> <ns0:div><ns0:head>Computational Framework</ns0:head><ns0:p>Rendering our cognitive model suitable for application in real-world contexts also entails accounting for uncertainty about parameter estimates. Indeed, uncertainty quantification turns out to be a fundamental and challenging goal when first-level quantities, that is, cognitive parameter estimates, are used to recover (second-level) information-theoretic measures of cognitive dynamics. The main difficulties arise when model complexity makes estimation and uncertainty quantification intractable at both analytical and numerical levels. For instance, in our case, probability distributions for the hidden model are generated at each trial, and the mapping between hidden states and responses changes depending on the structure of the task environment.</ns0:p><ns0:p>Identifying such a dynamic mapping is relatively easy from a generative perspective, but it becomes challenging, and almost impossible, when inverse modeling is required. Generally, this problem arises when the likelihood function relating model parameters to the data is not available in closed-form or too complex to be practically evaluated <ns0:ref type='bibr' target='#b55'>(Sisson and Fan (2011)</ns0:ref>). To overcome these limitations, we apply the first version of the recently developed BayesFlow method (see <ns0:ref type='bibr' target='#b47'>Radev et al. (2020)</ns0:ref> for mathematical details). At a high-level, BayesFlow is a simulation-based method that estimates parameters and quantifies estimation uncertainty in a unified Bayesian probabilistic framework when inverting the generative model is intractable. The method is based on recent advances in deep generative modeling and makes no assumptions about the shape of the true parameter posteriors. Thus, our ultimate goal becomes to</ns0:p></ns0:div> <ns0:div><ns0:head>13/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed ). Importantly, we can apply the same pre-trained inference network to an arbitrary number of real or simulated data sets (i.e., the training effort amortizes over multiple evaluations of the network).</ns0:p><ns0:p>For our purposes of validation and application, we train the network for 50 epochs which amount to 50000 forward simulations. As a prior, we use a bivariate continuous uniform distribution p(&#952;</ns0:p><ns0:formula xml:id='formula_15'>) &#8764; U ([0, 0], [1, 1]).</ns0:formula><ns0:p>We then validate performance on a separate validation set of 1000 simulated data sets with known ground-truth parameter values. Training the networks took less than a day on a single machine with an NVIDIA &#174; GTX1060 graphics card (CUDA version 10.0) using TensorFlow (version 1.13.1) (?).</ns0:p><ns0:p>In contrast, obtaining full parameter posteriors from the entire validation set took approximately 1.78</ns0:p><ns0:p>seconds. In what follows, we describe and report all performance validation metrics.</ns0:p></ns0:div> <ns0:div><ns0:head>Performance Metrics and Validation Results</ns0:head><ns0:p>To assess the accuracy of point estimates, we compute the root mean squared error (RMSE) and the coefficient of determination (R 2 ) between posterior means and true parameter values. To assess the quality of the approximate posteriors, we compute a calibration error <ns0:ref type='bibr' target='#b47'>(Radev et al. (2020)</ns0:ref>) of the empirical coverage of each marginal posterior Finally, we implement simulation-based calibration (SBC, Talts et al.</ns0:p><ns0:p>(2018)) for visually detecting systematic biases in the approximate posteriors.</ns0:p><ns0:p>Point Estimates. Point estimates obtained by posterior means as well as corresponding RMSE and R 2 metrics are depicted in Figure <ns0:ref type='figure' target='#fig_7'>5A-B</ns0:ref>. Note, that point estimates do not have any special status in Bayesian inference, as they could be misleading depending on the shape of the posteriors. However, they are simple to interpret and useful for ease-of-comparison. We observe that pointwise recovery of &#955; is better than that of &#948; . This is mainly due to suboptimal pointwise recovery in the lower (0, 0.1) range of &#948; . This pattern is evident in Figure <ns0:ref type='figure' target='#fig_7'>5A</ns0:ref> Manuscript to be reviewed </ns0:p></ns0:div> <ns0:div><ns0:head>APPLICATION</ns0:head><ns0:p>In this section we fit the Bayesian cognitive model to real clinical data. The aim of this application is to evaluate the ability of our computational framework to account for dysfunctional cognitive dynamics of information processing in substance dependent individuals (SDI) as compared to healthy controls.</ns0:p></ns0:div> <ns0:div><ns0:head>Rationale</ns0:head><ns0:p>The advantage of modeling cognitive dynamics in individuals from a clinical population is that model predictions can be examined in light of available evidence about individual performance. For instance, SDIs are known to demonstrate inefficient conceptualization of the task and dysfunctional, error-prone response strategies. This has been attributed to defective error monitoring and behavior modulation systems, which depend on cingulate and frontal brain regions functionality <ns0:ref type='bibr' target='#b34'>(K&#252;bler et al. (2005)</ns0:ref>; Willuhn et al. ( <ns0:ref type='formula'>2003</ns0:ref>)). On the other hand, the WCST should be a rather easy and straightforward task for healthy participants to obtain excellent performance. Therefore, we expect our model to consistently capture such characteristics. To test these expectations, we estimate the two relevant parameters &#955; and &#948; from both clinical patients and healthy controls from an already published dataset <ns0:ref type='bibr' target='#b5'>(Bechara and Damasio (2002)</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The Data</ns0:head><ns0:p>The dataset used in this application consists of responses collected by administering the standard Heaton version of the WCST <ns0:ref type='bibr' target='#b29'>(Heaton (1981)</ns0:ref>) to healthy participants and SDIs. In this version of the task, the sorting rule changes when a participant collects a series of 10 consecutive correct responses, and the task ends when this happens for 6 times. Participants in the study consisted of 39 SDIs and 49 healthy individuals. All participants were adults (&gt; 18 years old) and gave their informed consent for inclusion which was approved by the appropriate human subject committee at the University of Iowa. SDIs were diagnosed as substance dependent based on the Structured Clinical Interview for DSM-IV criteria <ns0:ref type='bibr' target='#b18'>(First (1997)</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>15/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Model Fitting</ns0:head><ns0:p>We fit the Bayesian cognitive agent separately to data from each participant in order to obtain individuallevel posterior distributions. We apply the same BayesFlow network trained for the previous simulation studies, so obtaining posterior samples for each participant is almost instant (due to amortized inference).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The means of the joint posterior distributions are depicted for each individual in Figure <ns0:ref type='figure'>6</ns0:ref>, and provide a complete overview of the heterogeneity in cognitive sub-components at both individual and group levels (individual-level full joint posterior distributions can be found in the SI Appendix).</ns0:p><ns0:p>Figure <ns0:ref type='figure'>6</ns0:ref>. Joint posterior mean coordinates of the cognitive parameters, flexibility (&#955; ) and information loss (&#948; ), estimated for each individual. We observe a great heterogeneity in the distribution of posterior means, most pronouncedly for the flexibility parameter. However, a moderate between-subject variability in information loss can still be observed in both groups.</ns0:p><ns0:p>The estimates reveal a rather interesting pattern across both healthy and SDI participants. In particular, in both clinical and control groups, individuals with a poor flexibility (e.g., low values of &#955; ) can be detected. However, the group parameter space appears to be partitioned into two main clusters consisting of individuals with high and low flexibility, respectively. As can be noticed, the majority of SDIs belongs to the latter cluster, which suggests that the model is able to capture error-related defective behavior in the</ns0:p></ns0:div> <ns0:div><ns0:head>16/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed clinical population and attribute it specifically to the flexibility parameter. On the other hand, individual performance seems hardly separable along the information loss parameter dimension.</ns0:p><ns0:p>As a further validation, we compare the classification performance of two logistic regression models.</ns0:p><ns0:p>The first uses the estimated parameter means as inputs and the participants' binary group assignment (patient vs. control) as an outcome. The second uses the four standard clinical measures (non-perseverative errors (E), perseverative errors (PE), number of trials to complete the first category (TFC), number of failures to maintain set (FMS) computed from the sample as inputs and the same outcome. Since we are interested solely in classification performance and want to mitigate potential overfitting due to small sample size, we compute leave-one-out cross-validated (LOO-CV) performance for both models.</ns0:p><ns0:p>Interestingly, both logistic regression models achieve the same accuracy of 0.70, with a sensitivity of 0.71 and specificity of 0.70. Thus, it appears that our model is able to differentiate between SDIs and healthy individuals as good as the standard clinical measures.</ns0:p><ns0:p>However, as pointed out in the previous sections, estimated parameters serve merely as a basis to reconstruct cognitive dynamics by means of the trial-by-trial unfolding of information-theoretic measures.</ns0:p><ns0:p>Moreover, cognitive dynamics can only be analysed and interpreted by relying on the joint contribution of both estimated parameters and individual-specific observed response patterns.</ns0:p><ns0:p>To further clarify this concept, we investigate the reconstructed time series of information-theoretic quantities based on the response patterns of two exemplary individuals (Figure <ns0:ref type='figure' target='#fig_8'>7</ns0:ref>). In particular, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref> depicts the behavioral outcomes of a SDI with sub-optimal performance where the information-theoretic trajectories are reconstructed by taking the corresponding posterior means ([ &#955; = 0.07, &#948; = 0.82]), thus representing compromised flexibility and high information loss. Differently, Figure <ns0:ref type='figure' target='#fig_8'>7B</ns0:ref> shows the information-theoretic path related to response dynamics of an optimal control participant, according to the parameter set [ &#955; = 0.60, &#948; = 0.35], representing relatively high flexibility, and low information loss.</ns0:p><ns0:p>Note, that in both cases, the reconstructed information-theoretic measures are based on the estimated posterior means for ease of comparison (see SI Appendix for the full joint posterior densities of the two exemplary individuals and the rest of the sample). 17/24 Processing unexpected observations is accounted by the quantification of surprise upon observing a response-feedback pair which is inconsistent with the current internal model of the task environment.</ns0:p><ns0:p>Negative feedback is maximally informative when errors occur after the internal model has converged to the true task model (grey area, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref>), or the entropy approaches zero (grey line, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref>). The Shannon surprise (orange line) is maximal when errors occur within trial windows in which the agent's uncertainty about environmental states is minimal (orange areas, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref>). However, internal model updates following an informative feedback are not optimally performed, which is reflected by very small</ns0:p></ns0:div> <ns0:div><ns0:head>18/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Bayesian surprise (blue line, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref>). This can be attributed to impaired flexibility and reflects the fact that after internal model convergence, informative feedback is not processed adequately and the internal model becomes impervious to change.</ns0:p><ns0:p>Conversely, errors occurring when the agent is uncertain about the true environmental state carry no useful information for belief updating, since the system fails to conceive such errors as unexpected and informative. The information loss parameter plays a crucial role in characterizing this cognitive behavior.</ns0:p><ns0:p>The slow convergence to the true environmental model, accompanied by the slow reduction of entropy in the predictive model, leads to a large number of trials required to achieve a good representation of the current task environment (white areas, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref>). Errors occurring within trial windows with large predictive model entropy (green area, Figure <ns0:ref type='figure' target='#fig_8'>7A</ns0:ref>) do not affect subsequent behavior, and feedback is maximally uninformative.</ns0:p><ns0:p>Rather different cognitive dynamics can be observed in Figure <ns0:ref type='figure' target='#fig_8'>7B</ns0:ref>, accounting for a typical optimal behavior where the errors produced fall within the trial windows which follow a rule completion (e.g. when the individual completes a sequence of 10 consecutive correct responses), and, thus, the environmental model becomes obsolete. However, the high flexibility, &#955; , allows to rely on local feedback-related information to suddenly update beliefs about the hidden states, that is, the most appropriate sorting rule. In this case, negative feedback become maximally informative after model convergence (grey area, Figure <ns0:ref type='figure' target='#fig_8'>7B</ns0:ref>) and the process of entropy reduction (green line, Figure <ns0:ref type='figure' target='#fig_8'>7B</ns0:ref>) is faster (e.g. less trials are needed) compared to the sub-optimal behavior scenario. Since uncertainty about the environmental states decreases faster, the Shannon surprise is always highly peaked when errors occur (orange line, Figure <ns0:ref type='figure' target='#fig_8'>7B</ns0:ref>), thus ensuring an efficient employment of the local feedback-related information. Accordingly, higher values of Bayesian surprise are observed (blue line, Figure <ns0:ref type='figure' target='#fig_8'>7B</ns0:ref>), revealing optimal internal model updating.</ns0:p><ns0:p>In general, the role that predictive (internal) model uncertainty plays in characterizing the way the agent processes feedback allows to disentangle sub-types of errors based on the information they convey for subsequent belief updating. From this perspective, error classification is entirely dependent on the status of the internal environmental model across task phases. Identifying such a dynamic latent process is therefore fundamental, since the error codification criterion evolves with respect to the internal information processing dynamics. Otherwise, the problem of inferring which errors are due to perseverance in maintaining an older (converged) internal model and which due to uncertainty about the true environmental state becomes intractable, or even impossible.</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Investigating information processing related to changing environmental contingencies is fundamental to understanding adaptive behavior. For this purpose, cognitive scientists mostly rely on controlled settings in which individuals are asked to accomplish (possibly) highly demanding tasks whose demands are assumed to resemble those of natural environments. Even in the most trivial cases, such as the WCST, optimal performance requires integrated and distributed neurocognitive processes. Moreover, these processes are unlikely to be isolated by simple scoring or aggregate performance measures.</ns0:p><ns0:p>In the current work, we developed and validated a new computational Bayesian model which maps distinct cognitive processes into separable information-theoretic constructs underlying observed adaptive behavior. We argue that these constructs could help describe and investigate the neurocognitive processes underlying adaptive behavior in a principled way.</ns0:p><ns0:p>Furthermore, we couple our computational model with a novel neural density estimation method for simulation-based Bayesian inference <ns0:ref type='bibr' target='#b47'>(Radev et al. (2020)</ns0:ref>). Accordingly, we can quantify the entire information contained in the data about the assumed cognitive parameters via a full joint posterior over plausible parameter values. Based on the joint posterior, a representative summary statistic can be computed to simulate the most plausible unfolding of information-theoretic quantities on a trial-by-trial basis.</ns0:p><ns0:p>Several computational models have been proposed to describe and explain performance in the WCST, ranging from behavioral <ns0:ref type='bibr' target='#b8'>(Bishara et al. (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b28'>Gl&#228;scher et al. (2019)</ns0:ref>; <ns0:ref type='bibr' target='#b56'>Steinke et al. (2020)</ns0:ref>) to neural network models <ns0:ref type='bibr' target='#b16'>(Dehaene and Changeux (1991)</ns0:ref>; <ns0:ref type='bibr' target='#b1'>Amos (2000)</ns0:ref>; <ns0:ref type='bibr' target='#b37'>Levine et al. (1993)</ns0:ref>; <ns0:ref type='bibr' target='#b41'>Monchi et al. (2000)</ns0:ref>).</ns0:p><ns0:p>These models aim to provide psychologically interpretable parameters or biologically inspired network Manuscript to be reviewed learning <ns0:ref type='bibr' target='#b56'>(Steinke et al. (2020)</ns0:ref>)) and disentangle psychological sub-processes explaining observed task performance. However, the main advantage of our Bayesian model is that it provides both a cognitive and a measurement model which coexist within the overarching theoretical framework of Bayesian brain theories. More precisely, the presented model is specifically designed to capture trial-by-trial fluctuations in information processing as described by second-order information-theoretic quantities. The latter can be seen as a multivariate quantitative account of the interaction between the agent and its environment.</ns0:p><ns0:p>Moreover, it is worth noting that such a model representation might not be applicable outside a Bayesian theoretical framework.</ns0:p><ns0:p>Even though our computational model is not a neural model, it might provide a suitable description of cognitive dynamics at a representational and/or a computational level <ns0:ref type='bibr' target='#b39'>(Marr (1982)</ns0:ref>). This description can then be related to neural functioning underlying adaptive behavioral. Indeed, there is some evidence to suggest that neural processes related to belief maintenance/updating and unexpectedness are crucial for performance in the WCST. In particular, brain circuits associated with cognitive control and belief formation, such as the parietal cortex and prefrontal regions, seem to share a functional basis with neural substrates involved in adaptive tasks <ns0:ref type='bibr' target='#b42'>(Nour et al. (2018)</ns0:ref>). Prefrontal regions appear to mediate the relation between feedback and belief updating <ns0:ref type='bibr' target='#b38'>(Lie et al. (2006)</ns0:ref>) and efficient functioning in such brain structures seems to be heavily dependent on dopaminergic neuromodulation <ns0:ref type='bibr' target='#b46'>(Ott and Nieder (2019)</ns0:ref>). Moreover, the dopaminergic system plays a role in the processing of salient and unexpected environmental stimuli, in learning based on error-related information, and in evaluating candidate actions <ns0:ref type='bibr' target='#b42'>(Nour et al. (2018)</ns0:ref>; <ns0:ref type='bibr' target='#b15'>Daw et al. (2011);</ns0:ref><ns0:ref type='bibr' target='#b27'>Gershman (2018)</ns0:ref>). Accordingly, dopaminergic system functioning has been put in relation with performance in the WCST <ns0:ref type='bibr' target='#b31'>(Hsieh et al. (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b49'>Rybakowski et al. (2005)</ns0:ref>) and shown to be critical for the main executive components involved in the task, that is, cognitive flexibility and set-shifting <ns0:ref type='bibr' target='#b7'>(Bestmann et al. (2014)</ns0:ref>; <ns0:ref type='bibr' target='#b57'>Stelzel et al. (2010)</ns0:ref>). Further, neural activity in the anterior cingulate cortex (ACC) is increased when a negative feedback occurs in the context of the WCST <ns0:ref type='bibr' target='#b38'>(Lie et al. (2006)</ns0:ref>).</ns0:p><ns0:p>This finding corroborates the view that the ACC is part of an error-detection network which allocates attentional resources to prevent future errors. The ACC might play a crucial role in adaptive functioning by encoding error-related or, more generally, feedback-related information. Thus, it could facilitate the updating of internal environmental models <ns0:ref type='bibr' target='#b48'>(Rushworth and Behrens (2008)</ns0:ref>).</ns0:p><ns0:p>The neurobiological evidence suggests that brain networks involved in the WCST might endow adaptive behavior by accounting for maintaining/updating of an internal model of the environment and efficient processing of unexpected information. Is it noteworthy, that these processing aspects are incorporated into our computational framework. At this point, we briefly outline the empirical and theoretical potentials of the proposed computational framework for investigating adaptive functioning and discuss future research vistas.</ns0:p><ns0:p>Model-Based Neuroscience. Recent studies have pointed out the advantage of simultaneously modeling and analyzing neural and behavioral data within a joint modeling framework. In this way, the latter can be used to provide information for the former, as well as the other way around <ns0:ref type='bibr' target='#b61'>(Turner et al. (2017</ns0:ref><ns0:ref type='bibr' target='#b62'>(Turner et al. ( , 2013))</ns0:ref>; <ns0:ref type='bibr' target='#b20'>Forstmann et al. (2011)</ns0:ref>). This involves the development of joint models which encode assumptions about the probabilistic relationships between neural and cognitive parameters.</ns0:p><ns0:p>Within our framework, the reconstruction of information-theoretic discrete time series yields a quantitative account of the agent's internal processing of environmental information. Event-related cognitive measures of belief updating, epistemic uncertainty and surprise can be put in relation with neural measurements by explicitly providing a formal account of the statistical dependencies between neural and cognitive (information-theoretic) quantities. In this way, latent cognitive dynamics can be directly related to neural event-related measures (e.g., fMRI, EEG). Applications in which information-theoretic measures are treated as dependent variables in standard statistical analysis are also possible.</ns0:p><ns0:p>Neurological Assessment. Although neuroscientists have considered performance in the WCST as a proxy for measuring high-level cognitive processes, the usual approach to the analysis of human adaptive behavior consists in summarizing response patterns by simple heuristic scoring measures (e.g., occurrences of correct responses and sub-types of errors produced) and classification rules <ns0:ref type='bibr' target='#b19'>(Flashman et al. (1991)</ns0:ref>).</ns0:p><ns0:p>However, the theoretical utility of such a summary approach remains questionable. Indeed, adaptive behavior appears to depend on a complex and intricate interplay between multiple network structures <ns0:ref type='bibr' target='#b3'>(Barcelo et al. (2006)</ns0:ref>; <ns0:ref type='bibr' target='#b40'>Monchi et al. (2001)</ns0:ref>; <ns0:ref type='bibr' target='#b38'>Lie et al. (2006)</ns0:ref>; <ns0:ref type='bibr' target='#b4'>Barcel&#243; and Rubia (1998)</ns0:ref>; <ns0:ref type='bibr' target='#b11'>Buchsbaum et al. (2005)</ns0:ref>). This posits a great challenge for disentangling high-level cognitive constructs at a model level and further investigating their relationship with neurobiological substrates. It appears that standard scoring</ns0:p></ns0:div> <ns0:div><ns0:head>20/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed measures might not be able to fulfil these tasks. Moreover, there is a pronounced lack of anatomical specificity in previous research concerning the neural and functional substrates of the WCST <ns0:ref type='bibr' target='#b43'>(Nyhus and Barcel&#243; (2009)</ns0:ref>).</ns0:p><ns0:p>Thus, there is a need for more sophisticated modeling approaches. For instance, disentangling errors due to perseverative processing of previously relevant environmental models from those due to uncertainty about task environmental states, is important and nontrivial. Sparse and distributed error patterns might depend on several internal model probability configurations. Such internal models are latent, and can only be uncovered through cognitive modeling. Therefore, information-based criteria to response (error) classification can enrich clinical evaluation beyond heuristically motivated criteria.</ns0:p><ns0:p>Generalizability. Another important advantage of the proposed computational framework is that it is not solely confined to the WCST. In fact, one can argue that the seventy-year old WCST does not provide the only or even the most suitable setting for extracting information about cognitive dynamics from general populations or maladaptive behavior in clinical populations. One can envision tasks which embody probabilistic (uncertain) or even chaotic environments (for instance with partially observable or unreliable feedback or partially observable states) and demand integrating information from different modalities <ns0:ref type='bibr' target='#b44'>(O'Reilly et al. (2013)</ns0:ref>; <ns0:ref type='bibr' target='#b42'>Nour et al. (2018)</ns0:ref>). These settings might prove more suitable for investigating changes in uncertainty-related processing or cross-modal integration than deterministic and fully observable WCST-like settings.</ns0:p><ns0:p>Despite these advantages, our proposed computational framework has certain limitations. A first limitation might concern the fact that the new Bayesian cognitive model accounts for the main dynamics in adaptive tasks by relying on only two parameters. Although such a parsimonious proposal suffices to disentangle latent data-generating processes, a more exhaustive formal description of cognitive subcomponents might be envisioned. However, parameter estimation can become challenging in such a scenario, especially when one-dimensional response data is used as a basis for parameter recovery. Second, the information loss parameter appears to be more challenging to estimate than the flexibility parameter in some datasets. There are at least two possible remedies for this problem. On the one hand, global estimation of information loss might be hampered due to the model's current functional (algorithmic)</ns0:p><ns0:p>formulation and can therefore be optimized via an alternative formulation/parameterization. On the other hand, it might be the case that the data obtainable in the simple WCST environment is not particularly informative about this parameter and, in general, not suitable for modeling more complex and non-linear cognitive dynamics in general. Future works should therefore focus on designing and exploring more data-rich controlled environments which can provide a better starting point for investigating complex latent cognitive dynamics in a principled way. Additionally, the information loss parameter seems to be less effective in differentiating between substance abusers and healthy controls in the particular sample used in this work. Thus, further model-based analyses on individuals from different clinical populations are needed to fully understand the potential of our 2-parameter model as a clinical neuropsychological tool.</ns0:p><ns0:p>Finally, in this work, we did not perform formal model comparison, as this would require an extensive consideration of various nested and non-nested model within the same theoretical framework and between different theoretical frameworks. We therefore leave this important endeavor for future research.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>In conclusion, the proposed model can be considered as the basis for a (bio)psychometric tool for measuring the dynamics of cognitive processes under changing environmental demands. Furthermore, it can be seen as a step towards a theory-based framework for investigating the relation between such cognitive measures and their neural underpinnings. Further investigations are needed to refine the proposed computational model and systematically explore the advantages of the Bayesian brain theoretical framework for empirical research on high-level cognition.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Suppose the correct sorting rule is the feature shape. The figure shows the rate of convergence of the predictive distributions to the true task environmental model. The predictive distributions at trial t + 1 depends on the sorting action a t (first row) and the received feedback f t (second row). Two examples of updating a predictive distribution are shown: one in which information loss is high (&#948; = 0.7, third row), and one in which information loss is low (&#948; = 0.3, fifth row). High information loss slows down the convergence of the internal model to the true environmental model. The gray bar plots represent the predictive probability distribution over the rules from which an action is sampled at each trial. Dotted bars represent the updated predictive distribution after the feedback observation. For each scenario, trial-by-trial information-theoretic measures are shown.</ns0:figDesc><ns0:graphic coords='9,143.80,149.77,409.44,374.89' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020) Manuscript to be reviewed per parameter combination and compute standard scoring measures for each of the agents simulated responses. Results from the simulation runs are depicted in</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Clinical scoring measures as functions of flexibility (&#955; ) and information loss (&#948; ) -simulated scenarios. The different cells show the violin plots for the estimated distribution densities of the scoring measures obtained from the group of synthetic individuals, for the levels of &#955; across different levels of &#948; . In particular, they show the distribution of non-perseverative errors (E), perseverative errors (PE), number of trials to complete the first category (TFC), number of failures to maintain set (FMS) obtained from 150 synthetic agent's response simulations for each cell of the factorial design.</ns0:figDesc><ns0:graphic coords='13,143.80,129.10,409.44,452.10' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Information-theoretic measures varying as a function of flexibility &#955; and information loss &#948; across 128 trials of the WCST. Optimal belief updating and uncertainty reduction are achieved with low information loss and high flexibility (first row, third column).</ns0:figDesc><ns0:graphic coords='14,143.80,351.76,409.42,233.36' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4 depicts results from the nine simulation scenarios. Although an exhaustive discussion on cognitive dynamics should couple information-theoretic measures with patterns of correct and error responses, we focus solely on the information-theoretic time series for illustrative purposes. We refer to the Application section for a more detailed description of the relation between observed responses and estimated information-theoretic measures in the context of data from a real experiment. Again, simulated performance of the Bayesian cognitive agent shows that different parameter combinations yield different patterns of cognitive dynamics. Observed spikes and their related magnitudes</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>approximate and analyze the joint posterior distribution over the model parameters. The parameter posterior is given via an application of Bayes' rule: p(&#952; |x 0:T , m 0:T ) = p(x 0:T , m 0:T |&#952; )p(&#952; ) p(x 0:T , m 0:T |&#952; )p(&#952; )d&#952; (10) where we set &#952; = (&#955; , &#948; ) and stack all observations and matching signals into the vectors x 0:T = (x 0 , x 1 , ..., x T ) and m 0:T = (m 0 , m 1 , ..., m T ), respectively. The BayesFlow method uses simulations from the generative model to optimize a neural density estimator which learns a probabilistic mapping between raw data and parameters. It relies on the fact that data can easily be simulated by repeatedly running the generative model with different parameter configurations &#952; sampled from the prior. During training, the neural network estimator iteratively minimizes the divergence between the true posterior and an approximate posterior. Once the network has been trained, we can efficiently obtain samples from the approximate joint posterior distribution of the cognitive parameters of interest, which can be further processed in order to extract meaningful summary statistics (e.g., posterior means, medians, modes, etc.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>-B and is due to the fact that &#948; values in this range produce almost indistinguishable data patterns. Bootstrap estimates yielded an average RMSE of 0.155 (SD = 0.004) and an average R 2 of 0.708 (SD = 0.015) for the &#948; parameter. An average RMSE of 0.094 (SD = 0.002) and an average R 2 of 0.895 (SD = 0.007) were obtained for the &#955; parameter. These results suggest good global pointwise recovery but also warrant the inspection of full posteriors, especially in the low ranges of &#948; .Full Posteriors. Average bootstrap calibration error was 0.011 (SD = 0.005) for the marginal posterior of &#948; and 0.014 (SD = 0.007) for the marginal posterior of &#955; . Calibration error is perhaps the most important metric here, as it measures potential under-or overconfidence across all confidence intervals of the approximate posterior (i.e., an &#945;-confidence interval should contain the true posterior with a probability of &#945;, for all &#945; &#8712; (0, 1)). Thus, low calibration error indicates a faithful uncertainty representation of the approximate posteriors. Additionally, SBC-histograms are depicted in Figure5C-D. As shown by (Talts et al. (2018)), deviations from the uniformity of the rank statistic (also know as a PIT histogram) indicate systematic biases in the posterior estimates. A visual inspection of the histograms reveals that the posterior means slightly overestimate the true values of &#948; . This corroborates the pattern seen in Figure 5A-B for the lower range of &#948; . Finally, Figure 5E-H depicts the full marginal posteriors on two example validation sets. Even on these two data sets, we observe strikingly different posterior shapes. The marginal posterior of &#948; obtained from the first data set is slightly left-skewed and has its density concentrated over the (0.8, 1.0) range. On the other hand, the marginal posterior of &#948; from the second data set is noticeably right-skewed and peaked across the lower range of the parameter. The marginal posteriors of &#955; appear more symmetric and warrant 14/24 PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Parameter recovery results on validation data; (A and B) Posterior means vs. true parameter values; (C and D) Histograms of the rank statistic used for simulation-based calibration; (E-H) Example full posteriors for two validation data sets; (I and J) Example information-theoretic dynamics recovered from the parameter posteriors.</ns0:figDesc><ns0:graphic coords='17,141.73,63.78,413.57,212.57' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Recovered cognitive dynamics of two exemplary individuals. (A) Trial-by-trial information-theoretic measures of a SDI characterized by very low flexibility and very high information loss; (B) Trial-by-trial information-theoretic measures of a healthy individual characterized by relatively high flexibility and low information loss. Labels C and E indicate correct and error responses.</ns0:figDesc><ns0:graphic coords='20,141.73,63.78,413.58,334.87' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,143.80,187.77,409.44,409.44' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Table 2 and a graphical representation is provided in Figure3. Mean clinical scoring measures as functions of flexibility (&#955; ) and information loss (&#948; ). Cells show the average scores across simulated agents (standard deviation is shown in parenthesis).</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Scoring Measure</ns0:cell><ns0:cell>Info. Loss (&#948; )</ns0:cell><ns0:cell>&#955; = 0.3</ns0:cell><ns0:cell cols='2'>Flexibility (&#955; ) &#955; = 0.5 &#955; = 0.7</ns0:cell><ns0:cell>&#955; = 0.9</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell>9.07 (2.68)</ns0:cell><ns0:cell>7.95 (2.07)</ns0:cell><ns0:cell>7.50 (2.13)</ns0:cell><ns0:cell>6.85 (1.75)</ns0:cell></ns0:row><ns0:row><ns0:cell>Casual Errors (E)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell>10.84 (2.35)</ns0:cell><ns0:cell>9.60 (2.2)</ns0:cell><ns0:cell>8.25 (2.23)</ns0:cell><ns0:cell>7.37 (1,74)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell cols='3'>12.75 (2.96) 11.25 (2.43) 9.12 (2.09)</ns0:cell><ns0:cell>7.79 (1.73)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell cols='4'>20.81 (2.27) 18.18 (1.88) 14.99 (1.88) 12.37 (1.12)</ns0:cell></ns0:row><ns0:row><ns0:cell>Perseverative Errors (PE)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell cols='4'>19.77 (2.55) 17.65 (2.26) 15.42 (1.94) 12.39 (1.47)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell cols='4'>18.56 (2.76) 16.58 (2.53) 14.49 (2.03) 12.33 (1.44)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell cols='4'>12.20 (1.46) 11.91 (1.35) 11.83 (1.24) 11.67 (1.04)</ns0:cell></ns0:row><ns0:row><ns0:cell>Trials to First Category (TFC)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell cols='4'>13.82 (2.76) 13.32 (2.52) 12.97 (2.13) 12.29 (1.53)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell cols='4'>17.27 (4.21) 16.63 (4.04) 14.39 (3.58) 12.91 (1.91)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell>0.11 (0.31)</ns0:cell><ns0:cell>0.09 (0.31)</ns0:cell><ns0:cell>0.05 (0.32)</ns0:cell><ns0:cell>0.02 (0.14)</ns0:cell></ns0:row><ns0:row><ns0:cell>Failures to Maintain Set (FMS)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell>1.65 (1.4)</ns0:cell><ns0:cell>1.41 (1.3)</ns0:cell><ns0:cell>0.84 (0.91)</ns0:cell><ns0:cell>0.35 (0.69)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell>4.44 (1.96)</ns0:cell><ns0:cell>3.88 (1.86)</ns0:cell><ns0:cell>2.79 (1.56)</ns0:cell><ns0:cell>1.54 (1.25)</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='2'>/24 PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:note> <ns0:note place='foot' n='10'>/24 PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:note> <ns0:note place='foot' n='12'>/24 PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:note> <ns0:note place='foot' n='24'>/24 PeerJ reviewing PDF | (2020:07:50686:1:2:NEW 26 Sep 2020)</ns0:note> </ns0:body> "

"RESPONSE LETTER We thank the editor and the reviewers for the positive reception of our ideas and the concrete and helpful suggestions for improving the manuscript. Below you can find a pointby-point summary of the changes introduced the new version of the manuscript. Editor (E) Comments E_1: “I believe that a model comparison would strengthen your conclusions”. A_1: We also agree that our modeling work would benefit from a model comparison. Although it is a best practice in this kind of work, we preferred to omit comparing our model with already available computational models for the WCST for two main reasons: 1) In the context of the WCST, our model seems to be one of a kind, from both coneptual and mathematical perspectives. The main purpose of the model is to recover second-order quantities, such as the information-theoretic measures on the basis of the observed sequence of responses, and the estimeable cognitive parameters flexibility and information loss serve as a proxy to accomplish this task. According to our knowledge, there are not current proposals in the literature which provide an exhaustive probabilistic framework to represent cognitive dynamics, that is, there are no competitors which allow to recover consistent probabilistic measures from behavioral outcomes. This provides a limitation when we want to compare models from a qualitative perspective, since comparisons would be based on totally different metrics (e.g,. Shannon surprise computed from mathematically consistent probabilistic object such as the model evidence should be compared with a different quantity such as the prediction error of a standard reinforcement learning model). On the other hand, standard model comparison techniques, based on information criteria (e.g. AIC and BIC) and the marginal likelihood (e.g. the Bayes factor) can be used if we focus on the model when no second-order quantities are taken into account. However, this practice might be problematic (see point 2). 2) Likelihood-based model camparisons are problematic in our case since, as discussed in lines 424-432, the likelihood function which relates model parameters and observed data is not available. This is the reason why we relied on a likelihoodfree deep learning-based method for parameter estimation. Competing models already available in the literature have either a computable likelihood, or are based on neural network architectures. Providing a framework to compare such an heterogeneous set of models is not straightforward, and would need a dedicated and rather expensive section in the paper, which, in our opinion, is outside the purpose of the current work. However, we are currently engaged in a project dealing with the specific problem of providing a comprehensive WCST model comparison. E_2: “it would be useful to assess the performance in healthy controls”. A_2: We requested and obtained the full dataset of patients (SDIs) and controls and provided a novel application section including healthy individuals model-based analysis. E_3: “the manuscript should be more accessible to a wider audience”. A_3: We tried to accomplish this goal by rewriting some key paragraphs in the introduction. The manuscript has now a revised description of the Bayesian brain hypothesis (lines 50-65), together with a simple example which might clarify the main concepts of the theory (lines 66-77). Besides, we enriched the corpus of references related to the employment of our computational framework in many neuropsychological settings. Rewiever 1 (R1) Comments R1_1: “provide a more detailed abstract that better clarify the methodology and the conclusions of the article”. A_1: We provided an extended abstract as requested. R1_2: “I think that the Authors should make an effort to be more understandable for the clinical and experimental neuropsychologist, with further examples concerning the Bayesian brain theories (e.g., they can report a classic example concerning sensory processing), and which (if any) benefits the neuropsychologist could take from using this perspective.” A_2: We fully agree with this point. The manuscript has now a revised description of the Bayesian brain hypothesis (lines 50-65), together with a simple example which might clarify the main concepts of the theory (lines 66-77). Besides, we enriched the corpus of references related to the employment of our computational framework in many neuropsychological settings in order to provide some insight on the usefulness of the framework for researchers and clinicians (lines 78-90). R1_3: “I would avoid the term ‘dub’”. A_3: The term “dub” has been removed. R1_4: “In the general description of the Winsconsin card sorting test, you should report which is the version that you are using, if a validation and normative data eixst, how many cards there are, after how many consecutive responses the sorting rule changes, etc.” A_4: We reported the typical version of the task we referred to (line 157) and provided a more precise description of the task administered to the participants in the study (lines 514-516). R1_5: “when you refer to BayesFlow, add the version number or the date of the version”. A_5: We adopted the first version of BayesFlow as now reported in the text. R1_6: “In Figure 4 and 5, the labels are too small.” A_6: We increase the label font size from 12 to 16. R1_7: “In the caption of Figure 3 please report what the violin plots and the boxplot are showing. Please, make more visible the boxplot because it is very hard to see” A_7: We improved the caption of Figure 3 as well as the entire figure. R1_8: “Why are you reporting only the graphics card of the computer? If you have used a CUDA-based computing approach you should explicitly state it.” A_8 The text now reports the version of CUDA and TensorFlow. R1_9: “More specifically, what does it mean s_t = 1, or 2, or 3? This is very confusing because all the three features (colour, shape and number of objects) can change among 4 different states. So I would expect something like s_t =(c_t, sh_t, n_t), where is the colour, is the shape, is the number of symbols. You are probably referring to the fact that the sorting rule can be shape, colour or number, but this is not specified anywhere. Please be clearer and use a mathematical symbology that is related to the task. If this is the case, you should call s “sorting rule”, and clearly state what s_t = 1, or 2, or 3 mean. Please, also specify that is the action of choosing desk 1, 2, 3 or 4, and that is the outcome failure (0) or success (1).” A_9: The manuscript now provided clearer descriptions of s_t, a_t and f_t in the Model section, as requested. Therefore, in order to avoid confusion, we improved the description of the task (lines 162-167) by discussing the relationship (overlapping) between the sorting rule and the card features. R1_10: “Please, make it clearer that “the hidden environmental states might change at a non-constant rate” means that they change each action of the agent.” A_10: The term “non-constant rate” seemed ambigous at a second inspection of the test, so we removed it from the manuscript and replaced the sentence. R1_11: “In eq. (2) I suppose that the j in the summation should be within {1, 2, 3}. Please, if I am correct, add it.” A_11: You are correct. We added the summation space. R1_12: “what do you mean with resp. decrease and resp. negative? You never introduced the “resp.” abbreviation.” A_12: The abbreviation for “respectively” (“resp.”) has been removed and the sentence has been made more explicit. R1_13: “ simulation 1. Please, report central tendency indexes and dispersion parameters of your choice in a table to numerically describe the results obtained from the simulations shown in Figure 3.” A_13: A table of mean and standard deviation of scoring measures for each cell has been added in the Simulation 1 section. R1_14: “ think that it might be interesting to see when E, PE, TCG and FMS in one condition are different from the other conditions. Would you mind to add statistical inference about that (better if done in the Bayesian framework)?” A_14: This is indeed an interesting idea, however, as usual for modeling works, we would prefer to remain on the level of the qualitative comparison when simulating data from the generative model. We think that performing a statistical test on simulated data is not a suitable practice in our case, especially when the sample size can be manipulated without a pre-defined criterion. Theoretically, we could always obtain a significance in the differences between conditions of a target scoring measure if we increase the sample size of simulated agents in order to approach infinity. R1_15: “Also in simulation 2, a statistical comparison among the different results would be nice.” A_15: We adopt the same argument here. In addition, it is likely that testing differences in time series of information-theoretic quantities would require nontrivial nonparametric techniques for time-series data, which, in our opinion, might be hard to describe and justify. R1_16: “Results. The results are very interesting, however, I am wondering why we do not have the results also from the Healthy population to make a comparison for the flexibility, information loss parameters, Bayesian surprise, Shannon surprise, and entropy.” A_16: We requested and obtained the full dataset of patients (SDIs) and controls and provided a novel application section including model-based analysis of healthy individuals. We also discuss the ability of the model to discrminate between partients and controls in comparison to discrmination with standard scoring measures. Rewiever 2 (R2) Comments R2_1: “both in the figures and text the authors could write what the parameters stand for instead of just the parameter itself”. A_1: as requested, we provided the name of the parameter together with the mathemtical symbol throughout the text, where possible. R2_2: “the authors could include a table with the parameter, its name, and what it measures”. A_2: This is a really good idea and we provided a table of all the model-related quantities introduced in the manuscript. R2_3: “it would be nice to also have examples of figures from control participants.” A_3: We provided the analysis of healthy individuals and introduced a new figure with information-theoretic trajectories recovered from an optimal control participant and compared it to a sub-optimal individual. R2_4: “The authors fit their model to actual data of substance-dependent individuals (SDI). This allowed them to see that SDIs generally had low flexibility and high information loss (as per the parameters of their model). It would be useful to see how their model performs in healthy controls, and what the model parameters are in this group.” A_4: We requested and obtained the full dataset of patients (SDIs) and controls and provided a novel application section including healthy individuals model-based analysis. Full posteriors for each participant are also available in an SI Appendix. R2_5: “it would be interesting to see if the model could have a direct clinical utility in terms of distinguishing between SDIs and healthy controls” A_5: We agree that this is an interesting question. We added such an analysis in the new application section. R2_6: “How does the model compare to existing models in the literature (e.g. a Reinforcement learning model)? However, it is perfectly ok for the authors to decide that model comparison is not an objective of this paper, in which case this could be stated directly” A_6: We also agree that our modeling work would benefit from a model comparison. Although it is a best practice in this kind of work, we preferred to omit comparing our model with already available computational models for the WCST for two main reasons: 1) In the context of the WCST, our model seems to be one of a kind, from both coneptual and mathematical perspectives. The main purpose of the model is to recover second-order quantities, such as the information-theoretic measures, on the basis of the observed sequence of responses, and the estimeable cognitive parameters flexibility and information loss serve as a proxy to accomplish this task. According to our knowledge, there are not current proposals in the literature which provide an exhaustive probabilistic framework to represent cognitive dynamics, that is, there are no competitors which allow to recover consistent probabilistic measures from behavioral outcomes. This provides a limitation when we want to compare models from a qualitative perspective, since comparisons would be based on totally different metrics (e.g. Shannon surprise computed from mathematically consistent probabilistic object such as the model evidence should be compared with a different quantity such as the prediction error of a standard reinforcement learning model). On the other hand, standard model comparison techniques, based on information criteria (e.g. AIC and BIC) and the marginal likelihood (e.g. the Bayes factor) can be used if we focus on the model when no second-order quantities are taken into account. However, this practice might be problematic (see point 2). 2) Likelihood-based model camparisons are problematic in our case since, as discussed in lines 424-432, the likelihood function which relates model parameters and observed data is not available. This is the reason why we relied on a likelihoodfree deep learning-based method for parameter estimation. Competing models already available in the literature have either a computable likelihood, or are based on neural network architectures. Providing a framework to compare such an heterogeneous set of models is not straightforward, and would need a dedicated and rather expensive section in the paper, which, in our opinion, is outside the purpose of the current work. However, we are currently engaged in a project dealing with the specific problem of providing a comprehensive WCST model comparison. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Adaptive behavior emerges through a dynamic interaction between cognitive agents and changing environmental demands. The investigation of information processing underlying adaptive behavior relies on controlled experimental settings in which individuals are asked to accomplish demanding tasks whereby a hidden regularity or an abstract rule has to be learned dynamically. Although performance in such tasks is considered as a proxy for measuring high-level cognitive processes, the standard approach consists in summarizing observed response patterns by simple heuristic scoring measures. With this work, we propose and validate a new computational Bayesian model accounting for individual performance in the Wisconsin Card Sorting Test (WCST), a renowned clinical tool to measure set-shifting and deficient inhibitory processes on the basis of environmental feedback. We formalize the interaction between the task's structure, the received feedback, and the agent's behavior by building a model of the information processing mechanisms used to infer the hidden rules of the task environment. Furthermore, we embed the new model within the mathematical framework of the Bayesian Brain Theory (BBT), according to which beliefs about hidden environmental states are dynamically updated following the logic of Bayesian inference. Our computational model maps distinct cognitive processes into separable, neurobiologically plausible, information-theoretic constructs underlying observed response patterns. We assess model identification and expressiveness in accounting for meaningful human performance through extensive simulation studies. We then validate the model on real behavioral data in order to highlight the utility of the proposed model in recovering cognitive dynamics at an individual level.</ns0:p><ns0:p>We highlight the potentials of our model in decomposing adaptive behavior in the WCST into several information-theoretic metrics revealing the trial-by-trial unfolding of information processing by focusing on two exemplary individuals whose behavior is examined in depth. Finally, we focus on the theoretical implications of our computational model by discussing the mapping between BBT constructs and functional neuroanatomical</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION 37</ns0:head><ns0:p>Computational models of cognition provide a way to formally describe and empirically account for mech-38 anistic, process-based theories of adaptive cognitive functioning <ns0:ref type='bibr' target='#b70'>(Sun (2009)</ns0:ref>; <ns0:ref type='bibr' target='#b15'>Cooper et al. (1996)</ns0:ref>; Lee 39 and Wagenmakers (2014)). A foundational theoretical framework for describing functional characteristics 40 of neurocognitive systems has recently emerged under the hood of Bayesian brain theories <ns0:ref type='bibr'>(Knill and</ns0:ref> 41 <ns0:ref type='bibr' target='#b38'>Pouget (2004)</ns0:ref>; <ns0:ref type='bibr' target='#b26'>Friston (2010)</ns0:ref>). Bayesian brain theories owe their name to their core assumption that beliefs, about the causal structure of events in the environment <ns0:ref type='bibr' target='#b24'>(Friston (2005)</ns0:ref>; <ns0:ref type='bibr' target='#b38'>Knill and Pouget (2004))</ns0:ref> and forms a basis for adaptive behavior. It is assumed that internal beliefs are constantly updated and refined to match the current state of the world as new observations become available. The core idea behind the Bayesian brain hypothesis is that computational mechanisms underlying such an internal belief updating follow the logic of Bayesian probability theory. In this respect, information about the external world provided by sensory inputs is represented as a conditional probability distribution over a set of environmental states. Consequently, the brain relies on this probabilistic representation of the world to infer the most likely environmental causes (states) which generate those inputs, and such a process follows the computational principles of Bayesian inference <ns0:ref type='bibr' target='#b29'>(Friston and Kiebel (2009)</ns0:ref>; <ns0:ref type='bibr' target='#b26'>Friston (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b13'>Buckley et al. (2017)</ns0:ref>).</ns0:p><ns0:p>To clarify this concept, consider a simple example of a perceptual task in which a cognitive agent is required to judge whether an item depicted on a flat plane is concave or convex. Its judgment is based solely on the basis of a set of observed perceptual features, such as, shape, orientation, texture and brightness. Here, the concave-to-convex gradient entails the set of environmental states which must be inferred. The internal generative model of the agent codifies beliefs about how different degrees of convexity might give rise to certain configurations of perceptual inputs. From a Bayesian perspective, the problem is solved by inverting the generative model of the environment in order to turn assumptions about how environmental states generate sensory inputs into beliefs about the most likely states (e.g., degree of convexity) given the available sensory information.</ns0:p><ns0:p>Potentially, there are no limitations regarding the complexity of environmental settings (e.g., items and rules in experimental tasks) and cognitive processes to be described in light of the Bayesian brain framework. Indeed, the latter has proven to be a consistent computational modeling paradigm for the investigation of a variety of neurocognitive mechanisms, such as motor control <ns0:ref type='bibr' target='#b30'>(Friston et al. (2010)</ns0:ref>), oculomotor dynamics <ns0:ref type='bibr' target='#b27'>(Friston et al. (2012)</ns0:ref>), object recognition <ns0:ref type='bibr' target='#b37'>(Kersten et al. (2004)</ns0:ref>), attention <ns0:ref type='bibr' target='#b18'>(Feldman and Friston (2010)</ns0:ref>), perceptual inference <ns0:ref type='bibr' target='#b57'>(Petzschner et al. (2015)</ns0:ref>; <ns0:ref type='bibr' target='#b38'>Knill and Pouget (2004)</ns0:ref>), multisensory integration <ns0:ref type='bibr' target='#b40'>(K&#246;rding et al. (2007)</ns0:ref>), as well as for providing a foundational theoretical account of general neural systems' functioning <ns0:ref type='bibr' target='#b45'>(Lee and Mumford (2003)</ns0:ref>; <ns0:ref type='bibr' target='#b24'>Friston (2005</ns0:ref><ns0:ref type='bibr' target='#b23'>Friston ( , 2003))</ns0:ref>) and complex clinical scenarios such as Schizophrenia <ns0:ref type='bibr' target='#b68'>(Stephan et al. (2006)</ns0:ref>), and Autistic Spectrum Disorder <ns0:ref type='bibr' target='#b33'>(Haker et al. (2016)</ns0:ref>; <ns0:ref type='bibr' target='#b43'>Lawson et al. (2014)</ns0:ref>). For this reason, such a modeling approach might provide a comprehensive and unified framework under which several cognitive impairments can be measured and understood in the light of a general process-based theory of neural functioning.</ns0:p><ns0:p>In this work, we address the challenging problem of modeling adaptive behavior in a dynamic environment. The empirical assessment of adaptive functioning often relies on dynamic reinforcement learning scenarios which require participants to adapt their behavior during the unfolding of a (possibly) demanding task. Typically, these tasks are designed with the aim to figure out how adaptive behavior unfolds through multiple trials as participants observe certain environmental contingencies, take actions, and receive feedback based on their actions. From a Bayesian theoretical perspective, optimal performance in such adaptive experimental paradigms require that agents infer the probabilistic model underlying the hidden environmental states. Since these models usually change as the task progresses, agents, in turn, need to adapt their inferred model, in order to take optimal actions.</ns0:p><ns0:p>Here, we propose and validate a computational Bayesian model which accounts for the dynamic behavior of cognitive agents in the Wisconsin Card Sorting Test (WCST; <ns0:ref type='bibr' target='#b7'>Berg (1948)</ns0:ref>; <ns0:ref type='bibr' target='#b34'>Heaton (1981)</ns0:ref>), which is perhaps the most widely adopted neuropsychological setting employed to investigate adaptive functioning, due to its specificity in accounting for executive components underlying observed behavior, such as set-shifting, cognitive flexibility and impulsive response modulation <ns0:ref type='bibr' target='#b9'>(Bishara et al. (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b1'>Alvarez and Emory (2006)</ns0:ref>). For this reason, we consider the WCST as a fundamental paradigm for investigating adaptive behavior from a Bayesian perspective.</ns0:p><ns0:p>The environment of the WCST consists of a target and a set of stimulus cards with geometric figures which vary according to three perceptual features. The WCST requires participants to infer the correct classification principle by trial and error using the examiner's feedback. The feedback is thought to carry a positive or negative information signaling the agent whether the immediate action was appropriate or not. Modeling adaptive behavior in the WCST from a Bayesian perspective is straightforward, since observable actions emerge from the interaction between the internal probabilistic model of the agent and a set of discrete environmental states.</ns0:p><ns0:p>Performance in WCST is usually measured via a rough summary metric such as the number of</ns0:p></ns0:div> <ns0:div><ns0:head>2/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed correct/incorrect responses or pre-defined psychological scoring criteria (see for instance <ns0:ref type='bibr' target='#b34'>Heaton (1981)</ns0:ref>).</ns0:p><ns0:p>These metrics are then used to infer the underlying cognitive processes involved in the task. A major shortcoming of this approach is that it simply assumes the cognitive processes to be inferred without specifying an explicit process model. Moreover, summary measures do not utilize the full information present in the data, such as trial-by-trial fluctuations or various interesting agent-environment interactions.</ns0:p><ns0:p>For this reason, crude scoring measures are often insufficient to disentangle the dynamics of the relevant cognitive (sub)processes involved. Consequently, an entanglement between processes at the metric level can prevent us from answering interesting research questions about aspects of adaptive behavior.</ns0:p><ns0:p>In our view, a sound computational account for adaptive behavior in the WCST needs to provide at least a quantitative measure of effective belief updating about the environmental states at each trial. This measure should be complemented by a measure of how feedback-related information influences behavior. The first measure should account for the integration of meaningful information. In other words, it should describe how prior beliefs about the current environmental state change after an observation has been made. The second measure should account for signaling the (im)probability of observing a certain environmental configuration (e.g., an (un)expected feedback given a response) <ns0:ref type='bibr' target='#b63'>(Schwartenbeck et al. (2016)</ns0:ref>).</ns0:p><ns0:p>Indeed, recent studies suggest that the meaningful information content and the pure unexpectedness of an observation are processed differently at the neural level. Moreover, such disentanglement appears to be of crucial importance to the understanding of how new information influences adaptive behavior <ns0:ref type='bibr' target='#b52'>(Nour et al. (2018)</ns0:ref>; <ns0:ref type='bibr' target='#b63'>Schwartenbeck et al. (2016);</ns0:ref><ns0:ref type='bibr' target='#b54'>O'Reilly et al. (2013)</ns0:ref>). Inspired by these results and previous computational proposals <ns0:ref type='bibr' target='#b39'>(Koechlin and Summerfield (2007)</ns0:ref>), we integrate these different information processing aspects into the current model from an information-theoretic perspective.</ns0:p><ns0:p>Our computational cognitive model draws heavily on the mathematical frameworks of Bayesian probability theory and information theory <ns0:ref type='bibr' target='#b62'>(Sayood (2018)</ns0:ref>). First, it provides a parsimonious description of observed data in the WCST via two neurocognitively meaningful parameters, namely, flexibility and information loss (to be motivated and explained in the Model section). Moreover, it captures the main response patterns obtainable in the WCST via different parameter configurations. Second, we formulate a functional connection between cognitive parameters and underlying information processing mechanisms related to belief updating and prediction formation. We formalize and distinguish between Bayesian surprise and Shannon surprise as the main mechanisms for adaptive belief updating. Moreover, we introduce a third quantity, which we named predictive Entropy and which quantifies an agent's subjective uncertainty about the current internal model. Finally, we propose to measure these quantities on a trial-bytrial basis and use them as a proxy for formally representing the dynamic interplay between agents and environments.</ns0:p><ns0:p>The rest of the paper is organized as follows. First, the WCST is described in more detail and a mathematical representation of the new Bayesian computational model is provided. Afterwards, we explore the model's characteristics through simulations and perform parameter recovery on simulated data using a powerful Bayesian deep neural network method <ns0:ref type='bibr' target='#b58'>(Radev et al. (2020)</ns0:ref>). We then apply the model to real behavioral data from an already published dataset. Finally, we discuss the results as well as the main strengths and limitations of the proposed model.</ns0:p></ns0:div> <ns0:div><ns0:head>THE WISCONSIN CARD SORTING TEST</ns0:head><ns0:p>In a typical WCST <ns0:ref type='bibr' target='#b34'>(Heaton (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b7'>Berg (1948)</ns0:ref>), participants learn to pay attention and respond to relevant stimulus features, while ignoring irrelevant ones, as a function of experimental feedback. In particular, Individuals are asked to match a target card with one of four stimulus cards according to a proper sorting principle, or sorting rule. Each card depicts geometric figures that vary in terms of three features, namely, color (red, green, blue, yellow), shape (triangle, star, cross, circle) and number of objects (1, 2, 3 and 4). For each trial, the participant is required to identify the sorting rule which is valid for that trial, that is, which of the three feature has to be considered as a criterion to matching the target card with the right stimulus card (see Figure <ns0:ref type='figure'>1</ns0:ref>). Notice that both features and sorting rules refer to the same concept.</ns0:p><ns0:p>However, the feature still codifies a property of the card, whilst the sorting rule refers to the particular feature which is valid for the current trial.</ns0:p></ns0:div> <ns0:div><ns0:head>3/24</ns0:head><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref>. Suppose that the current sorting rule is the feature shape. The target card in the first trial (left box) contains two blue triangles. A correct response requires that the agent matches the target card with the stimulus card containing the single triangle (arrow represents the correct choice), regardless of the features color and number. The same applies for the second trial (right box) in which matching the target card with the stimulus card containing three yellow crosses is the correct response.</ns0:p><ns0:p>Each response in the WCST is followed by a feedback informing the participant if his/her response is correct or incorrect. After some fixed number of consecutive responses, the sorting rule is changed by the experimenter without warning, and participants are required to infer the new sorting rule. Clearly, the most adaptive response would be to explore the remaining possible rules. However, participants sometimes would persist responding according to the old rule and produce what is called a perseverative response.</ns0:p></ns0:div> <ns0:div><ns0:head>METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>The Model</ns0:head><ns0:p>The core idea behind our computational framework is to encode the concept of belief into a generative probabilistic model of the environment. Belief updating then corresponds to recursive Bayesian updating of the internal model based on current and past interactions between the agent and its environment.</ns0:p><ns0:p>Optimal or sub-optimal actions are selected according to a well specified or a misspecified internal model and, in turn, cause perceptible changes in the environment.</ns0:p><ns0:p>We assume that the cognitive agent aims to infer the true hidden state of the environment by processing and integrating sensory information from the environment. Within the context of the WCST, the hidden environmental states might change as a function of both the structure of the task and the (often suboptimal) behavioral dynamics, so the agent constantly needs to rely on environmental feedback and own actions to infer the current state. We assume that the agent maintains an internal probability distribution over the states at each individual trial of the WCST. The agent then updates this distribution upon making new observations. In particular, the hidden environmental states to be inferred are the three features, s t &#8712; {1, 2, 3}, which refer the three possible sorting rules in the task environment such that 1: color, 2: shape and 3: number of objects. The posterior probability of the states depends on an observation vector x t = (a t , f t ), which consists of the pair of agent's response a t &#8712; {1, 2, 3, 4}, codifying the action of choosing deck 1, 2, 3 or 4, and received feedback f t &#8712; {0, 1}, referring to the fact that a given response results in a failure (0) or in a success (1), in a given trial t = 0, ..., T . The discrete response a t represents the stimulus card indicator being matched with a target card at trial t. We denote a sequence of observations as x 0:t = (x 0 , x 1 , ..., x t ) = ((a 0 , f 0 ), (a 1 , f 1 ), (a 2 , f 2 ), ..., (a t , f t )) and set x 0 = &#8709; in order to indicate that there are no observations at the onset of the task. Thus, trial-by-trial belief updating is recursively computed according to Bayes' rule:</ns0:p><ns0:formula xml:id='formula_0'>p(s t |x 0:t ) = p(x t |s t , x 0:t&#8722;1 )p(s t |x 0:t&#8722;1 ) p(x t |x 0:t&#8722;1 )<ns0:label>(1)</ns0:label></ns0:formula><ns0:p>Accordingly, the agent's posterior belief about the task-relevant features s t after observing a sequence of response-feedback pairs x 0:t is proportional to the product of the likelihood of observing a particular 4/24 response-feedback pair and the agent's prior belief about the task-relevant feature in the current trial. The likelihood of an observation is computed as follows:</ns0:p><ns0:formula xml:id='formula_1'>p(x t |s t , x 0:t&#8722;1 ) = f t p(a t |s t = i) + (1 &#8722; f t )(1 &#8722; p(a t |s t = i)) f t &#8721; j p(a t |s t = j) + (1 &#8722; f t ) &#8721; j (1 &#8722; p(a t |s t = j))</ns0:formula><ns0:p>(2)</ns0:p><ns0:p>where j = 1, 2, 3 and p(a t |s t = i) indicates the probability of a matching between the target and the stimulus card assumed that the current feature is i. Here, we assume the likelihood of a current observation to be independent from previous observations without loss of generality, that is:</ns0:p><ns0:formula xml:id='formula_2'>p(x t |s t , x 0:t&#8722;1 ) = p(x t |s t )</ns0:formula><ns0:p>The prior belief for a given trial t is computed based on the posterior belief generated in the previous trial, p(s t&#8722;1 |x 0:t&#8722;1 ), and the agent's belief about the probability of transitions between the hidden states, p(s t |s t&#8722;1 ). The prior belief can also be considered as a predictive probability over the hidden states.</ns0:p><ns0:p>The predictive distribution for an upcoming trial t is computed according to the Chapman-Kolmogorov equation:</ns0:p><ns0:formula xml:id='formula_3'>p(s t+1 = k|x 0:t ) = 3 &#8721; i=1 p(s t+1 = k|s t = i, &#915;(t))p(s t = i|x 0:t )<ns0:label>(3)</ns0:label></ns0:formula><ns0:p>where &#915;(t) represents a stability matrix describing transitions between the states (to be explained shortly).</ns0:p><ns0:p>Thus, the agent combines information from the updated belief (posterior distribution) and the belief about the transition properties of the environmental states to predict the most probable future state. The predictive distribution represents the internal model of the cognitive agent according to which actions are generated.</ns0:p><ns0:p>The stability matrix &#915;(t) encodes the agent's belief about the probability of states being stable or likely to change in the next trial. In other words, the stability matrix reflects the cognitive agent's internal representation of the dynamic probabilistic model of the task environment. It is computed on each trial based on the response-feedback pair, x t , and a matching signal, m t , which are observed.</ns0:p><ns0:p>The matching signal m t is a vector informing the cognitive agent which features are currently relevant (meaningful), such that m (i) t = 1 when a positive feedback is associated with a response implying feature s t = i, and m (i) t = 0 otherwise. Note, that the matching signal is not a free parameter of the model, but is completely determined by the task contingencies. The matching signal vector allows the agent to compute the state activation level &#969; (i)</ns0:p><ns0:p>t &#8712; [0, 1] for the hidden state s t = i, which provides an internal measure of the (accumulated) evidence for each hidden state at trial t. Thus, the activation levels of the hidden states are represented by a vector &#969; t . The stability matrix is a square and asymmetric matrix related to hidden state activation levels such that:</ns0:p><ns0:formula xml:id='formula_4'>&#915;(t) = &#63726; &#63727; &#63727; &#63727; &#63728; &#969; (1) t 1 2 (1 &#8722; &#969; (1) t ) 1 2 (1 &#8722; &#969; (1) t ) 1 2 (1 &#8722; &#969; (2) t ) &#969; (2) t 1 2 (1 &#8722; &#969; (2) t ) 1 2 (1 &#8722; &#969; (3) t ) 1 2 (1 &#8722; &#969; (3) t ) &#969; (3) t &#63737; &#63738; &#63738; &#63738; &#63739; (4)</ns0:formula><ns0:p>where the entries &#915; ii (t) in the main diagonal represent the elements of the activation vector &#969; t , and the non-diagonal elements are computed so as to ensure that rows sum to 1. The state activation vector is computed in each trial as follows:</ns0:p><ns0:formula xml:id='formula_5'>&#63726; &#63727; &#63728; &#969; (1) t &#969; (2) t &#969; (3) t &#63737; &#63738; &#63739; = f t &#969; &#948; t&#8722;1 &#63726; &#63727; &#63728; m (1) t m (2) t m (3) t &#63737; &#63738; &#63739; + &#955; &#63726; &#63727; &#63728;(1 &#8722; f t )&#969; &#948; t&#8722;1 &#63726; &#63727; &#63728; 1 &#8722; m (1) t 1 &#8722; m (2) t 1 &#8722; m (3) t &#63737; &#63738; &#63739; &#63737; &#63738; &#63739; &#63726; &#63727; &#63728; &#969; (1) t&#8722;1 &#969; (2) t&#8722;1 &#969; (3) t&#8722;1 &#63737; &#63738; &#63739; .</ns0:formula><ns0:p>(5)</ns0:p></ns0:div> <ns0:div><ns0:head>5/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>This equation reflects the idea that state activations are simultaneously affected by the observed feedback, f t , and the matching signal vector, m t . However, the matching signal vector conveys different information based on the current feedback. Matching a target card with a stimulus card makes a feature (or a subset of features) informative for a specific state. The vector m t contributes to increase the activation level of a state if the feature is informative for that state when a positive feedback is received, as well as to decrease the activation level when a negative feedback is received.</ns0:p><ns0:p>The parameter &#955; &#8712; [0, 1] modulates the efficiency to disengage attention to a given state-activation configuration when a negative feedback is processed. We therefore term this parameter flexibility. We also assume that information from the matching signal vector can degrade by slowing down the rate of evidence accumulation for the hidden states. This means that the matching signal vector can be re-scaled based on the current state activation level. The parameter &#948; &#8712; [0, 1] is introduced to achieve this re-scaling. When &#948; = 0, there is no re-scaling and updating of the state activation levels relies on the entire information conveyed by m t . On the other extreme, when &#948; = 1, several trials have to be accomplished before converging to a given configuration of the state activation levels. Equivalently, higher values of &#948; affect the entropy of the distribution over hidden states by decreasing the probability of sampling of the correct feature. We therefore refer to &#948; as information loss.</ns0:p><ns0:p>The free parameters &#955; and &#948; are central to our computational model, since they regulate the rate at which the internal model converges to the true task environmental model. Eq. ( <ns0:ref type='formula'>5</ns0:ref>) can be expressed in compact notation as follows:</ns0:p><ns0:formula xml:id='formula_6'>&#969; t = f t &#969; &#948; t&#8722;1 m t + &#955; (1 &#8722; f t )&#969; &#948; t&#8722;1 (1 &#8722; m t ) &#969; t&#8722;1 (6)</ns0:formula><ns0:p>Note that the information loss parameter &#948; affects the amount of information that a cognitive agent acquires from environmental contingencies, irrespective of the type of feedback received. Global information loss thus affects the rate at which the divergence between the agent's internal model and the true model is minimized. Figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref> illustrates these ideas.</ns0:p><ns0:p>The probabilistic representation of adaptive behaviour provided by our Bayesian agent model allows us to quantify latent cognitive dynamics by means of meaningful information-theoretic measures. Information theory has proven to be an effective and natural mathematical language to account for functional integration of structured cognitive processes and to relate them to brain activity <ns0:ref type='bibr' target='#b39'>(Koechlin and Summerfield (2007)</ns0:ref>; <ns0:ref type='bibr' target='#b28'>Friston et al. (2017)</ns0:ref>; <ns0:ref type='bibr' target='#b14'>Collell and Fauquet (2015)</ns0:ref>; <ns0:ref type='bibr' target='#b69'>Strange et al. (2005)</ns0:ref>; <ns0:ref type='bibr' target='#b23'>Friston (2003)</ns0:ref>). In particular, we are interested in three key measures, namely, Bayesian surprise, B t , Shannon surprise, I t , and entropy, H t . The subscript t indicates that we can compute each quantity on a trial-by-trial basis. Each quantity is amenable to a specific interpretation in terms of separate neurocognitive processes. Bayesian surprise B t quantifies the magnitude of the update from prior belief to posterior belief. Shannon surprise I t quantifies the improbability of an observation given an agent's prior expectation. Finally, entropy H t measures the degree of epistemic uncertainty regarding the true environmental states. Such measures are thought to account for the ability of the agent to manage uncertainty as emerging as a function of competing behavioral affordances <ns0:ref type='bibr' target='#b35'>(Hirsh et al. (2012)</ns0:ref>). We expect an adaptive system to attenuate uncertainty over environmental states (current features) by reducing the entropy of its internal probabilistic model.</ns0:p><ns0:p>Bayesian surprise can be computed as the Kullback-Leibler (KL) divergence between prior and posterior beliefs about the environmental states. Thus, Bayesian surprise accounts for the divergence between the predictive model for the current trial and the updated predictive model for the upcoming trial.</ns0:p><ns0:p>It is computed as follows:</ns0:p><ns0:formula xml:id='formula_7'>B t = KL[p(s t+1 |x 0:t )||p(s t |x 0:t&#8722;1 )] = 3 &#8721; i=1 p(s t+1 = i|x 0:t ) log p(s t+1 = i|x 0:t ) p(s t = i|x 0:t&#8722;1 )<ns0:label>(7)</ns0:label></ns0:formula><ns0:p>The Shannon surprise of a current observation given a previous one is computed as the conditional </ns0:p><ns0:formula xml:id='formula_8'>6/24</ns0:formula></ns0:div> <ns0:div><ns0:head>7/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed information content of the observation:</ns0:p><ns0:formula xml:id='formula_9'>I t = &#8722; log p(x t |x 0:t&#8722;1 ) = &#8722; log 3 &#8721; i=1 [p(x t |s t = i)p(s t = i|x 0:t&#8722;1 )]<ns0:label>(8)</ns0:label></ns0:formula><ns0:p>Finally, the entropy is computed over the predictive distribution in order to account for the uncertainty in the internal model of the agent in trial t as follows:</ns0:p><ns0:formula xml:id='formula_10'>H t = E [&#8722; log p(s t |x 0:t&#8722;1 )] = &#8722; 3 &#8721; i=1 p(s t = i|x 0:t&#8722;1 ) log p(s t = i|x 0:t&#8722;1 )<ns0:label>(9)</ns0:label></ns0:formula><ns0:p>Once the flexibility (&#955; ) and information loss (&#948; ) parameters are estimated from data, the informationtheoretic quantities can be easily computed and visualized for each trial of the WCST (see Figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>).</ns0:p><ns0:p>This allows to rephrase standard neurocognitive constructs in terms of measurable information-theoretic quantities. Moreover, the dynamics of these quantities, as well as their interactions, can be used for formulating and testing hypotheses about the neurcognitive underpinnings of adaptive behavior in a principled way, as discussed later in the paper. A summary of all quantities relevant for our computational model is provided in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>8/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Simulations</ns0:head><ns0:p>In this section we evaluate the expressiveness of the model by assessing its ability to reproduce meaningful behavioral patterns as a function of its two free parameters. We study how the generative model behaves when performing the WCST in a 2-factorial simulated Monte Carlo design where flexibility (&#955; ) and information loss (&#948; ) are systematically varied.</ns0:p><ns0:p>In this simulation, the Heaton version of the task <ns0:ref type='bibr' target='#b34'>(Heaton (1981)</ns0:ref>) is administered to the Bayesian cognitive agent. In this particular version, the sorting rule (true environmental state) changes after a fixed number of consecutive correct responses. In particular, when the agent correctly matches the target card in 10 consecutive trials, the sorting rule is automatically changed. The task ends after completing a maximum of 128 trials.</ns0:p></ns0:div> <ns0:div><ns0:head>Generative Model</ns0:head><ns0:p>The cognitive agent's responses are generated at each time step (trial) by processing the experimental feedback. Its performance depends on the parameters governing the computation of the relevant quantities.</ns0:p><ns0:p>The generative algorithm is outlined in Algorithm 1.</ns0:p><ns0:p>Algorithm 1 Bayesian cognitive agent 1: Set parameters &#952; = (&#955; , &#948; ).</ns0:p><ns0:p>2: Set initial activation levels &#969; 0 = (0.5, 0.5, 0.5).</ns0:p><ns0:p>3: Set initial observation x 0 = &#8709; and p(s 1 |x 0 ) = p(s 1 ). 4: for t = 1, ..., T do 5:</ns0:p><ns0:p>Sample feature from prior/predictive internal model s t &#8764; p(s t |x 0:t&#8722;1 ).</ns0:p></ns0:div> <ns0:div><ns0:head>6:</ns0:head><ns0:p>Obtain a new observation x t = (a t , f t ).</ns0:p></ns0:div> <ns0:div><ns0:head>7:</ns0:head><ns0:p>Compute state posterior p(s t |x 0:t ).</ns0:p></ns0:div> <ns0:div><ns0:head>8:</ns0:head><ns0:p>Compute new activation levels &#969; t . 9:</ns0:p><ns0:p>Compute stability matrix &#915;(t).</ns0:p><ns0:p>10:</ns0:p><ns0:p>Update prior/predictive internal model to p(s t+1 |x 0:t ). 11: end for Simulation 1: Clinical Assessment of the Bayesian Agent Ideally, the qualitative performance of the Bayesian cognitive agent will resemble human performance. To this aim, we adopt a metric which is usually employed in clinical assessment of test results in neurological and psychiatric patients <ns0:ref type='bibr' target='#b11'>(Braff et al. (1991)</ns0:ref>; <ns0:ref type='bibr' target='#b75'>Zakzanis (1998)</ns0:ref>; <ns0:ref type='bibr' target='#b6'>Bechara and Damasio (2002)</ns0:ref>; Landry and Al-Taie ( <ns0:ref type='formula'>2016</ns0:ref>)). Thus, agent performance is codified according to a neuropsychological criterion <ns0:ref type='bibr' target='#b34'>(Heaton (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b21'>Flashman et al. (1991)</ns0:ref>) which allows to classify responses into several response types. These response types provide the scoring measures for the test.</ns0:p><ns0:p>Here, we are interested in: 1) non-perseverative errors (E); 2) perseverative errors (PE); 3) number of trials to complete the first category (TFC); and 4) number of failures to maintain set (FMS). Perseverative errors occur when the agent applies a sorting rule which was valid before the rule has been changed.</ns0:p><ns0:p>Usually, detecting a perseveration error is far from trivial, since several response configurations could be observed when individuals are required to shift a sorting rule after completing a category (see <ns0:ref type='bibr' target='#b21'>Flashman et al. (1991)</ns0:ref> for details). On the other hand, non-perseverative errors refer to all errors which do not fit the above description, or in other words, do not occur as a function of changing the sorting rule, such as casual errors.</ns0:p><ns0:p>The number of trials to complete the first category tells us how many trials the agent needs in order to achieve the first sorting principle, and can be seen as an index of conceptual ability <ns0:ref type='bibr' target='#b3'>(Anderson (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b64'>Singh et al. (2017)</ns0:ref>). Finally, a failure to maintain a set occurs when the agent fails to match cards according to the sorting rule after it can be determined that the agent has acquired the rule. A given sorting rule is assumed to be acquired when the individual correctly sorts at least five cards in a row <ns0:ref type='bibr' target='#b34'>(Heaton (1981)</ns0:ref>; <ns0:ref type='bibr' target='#b19'>Figueroa and Youmans (2013)</ns0:ref>). Thus, a failure to maintain a set arises whenever a participant suddenly changes the sorting strategy in the absence of negative feedback. Failures to maintain a set are mostly attributed to distractibility. We compute this measure by counting the occurrences of first errors after the acquisition of a rule.</ns0:p><ns0:p>We run the generative model by varying flexibility across four levels, &#955; &#8712; {0.3, 0.5, 0.7, 0.9}, and information loss across three levels, &#948; &#8712; {0.4, 0.7, 0.9}. We generate data from 150 synthetic cognitive agents</ns0:p></ns0:div> <ns0:div><ns0:head>9/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed per parameter combination and compute standard scoring measures for each of the agents simulated responses. Results from the simulation runs are depicted in Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref> and a graphical representation is provided in Figure <ns0:ref type='figure' target='#fig_1'>3</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Scoring Measure</ns0:head><ns0:p>Info. Loss (&#948; ) Flexibility (&#955; ) &#955; = 0.3 &#955; = 0.5 &#955; = 0.7 &#955; = 0.9</ns0:p><ns0:p>Casual Errors (E) The simulated performance of our Bayesian cognitive agents demonstrates that different parameter combinations capture different meaningful behavioral patterns. In other words, flexibility and information loss seem to interact in a theoretically meaningful way.</ns0:p><ns0:formula xml:id='formula_11'>&#948; = 0.4 9.</ns0:formula><ns0:p>First, overall errors increase when flexibility (&#955; ) decreases, which is reflected by the inverse relation between the number of casual, as well as perseverative, errors and the values of parameter &#955; . Moreover, this pattern is consistent across all the levels of parameter &#948; . More precisely, information loss (&#948; ) seems to contribute to the characterization of the casual and the perseverative components of the error in a different way. Perseverative errors are likely to occur after a sorting rule has changed and reflect the inability of the agent to use feedback to disengage attention from the currently attended feature. They therefore result from local cognitive dynamics conditioned on a particular stage of the task (e.g., after completing a series of correct responses).</ns0:p><ns0:p>Second, information loss does not interact with flexibility when perseverative errors are considered. This is due to the fact that high information loss affects general performance by yielding a dysfunctional response strategy which increases the probability of making an error at any stage of the task. The lack of such interaction provides evidence that our computational model can disentangle between error patterns due to perseveration and those due to general distractibility, according to neuropsychological scoring criteria.</ns0:p><ns0:p>However, in our framework, flexibility (&#955; ) is allowed to yield more general and non-local cognitive dynamics as well. Indeed, &#955; plays a role whenever belief updating is demanded as a function of negative feedback. An error classified as non-perseverative (e.g., casual error) by the scoring criteria might still be processed as a feedback-related evidence for belief updating. Consistently, the interaction between &#955; and &#948; in accounting for causal errors shows that performance worsens when both flexibility and information loss become less optimal, and that such pattern becomes more pronounced for lower values of &#948; .</ns0:p><ns0:p>On the other hand, a specific effect of information loss (&#948; ) can be observed for the scoring measures related to slow information processing and distractibility. The number of trials to achieve the first category reflects the efficiency of the agent in arriving at the first true environmental model. Flexibility does not contribute meaningfully to the accumulation of errors before completing the first category for some levels of information loss. This is reflected by the fact that the mean number of trials increases as a function of &#948; , and do not change across levels of &#955; for low and mid values of &#948; . A similar pattern applies for failures to maintain a set. Both scoring measures index a deceleration of the process of evidence accumulation for Manuscript to be reviewed a specific environmental configuration, although the latter is a more exhaustive measures of dysfunctional adaptation.</ns0:p><ns0:p>Therefore, an interaction between parameters can be observed when information loss is high. A slow internal model convergence process increases the amount of errors due to improper rule sampling from the internal environmental model. However, internal model convergence also plays a role when a new category has to be accomplished after completing an older one. On the one hand, compromised flexibility increases the amount of errors due to inefficient feedback processing. This leads to longer trial windows needed to achieve the first category. On the other hand, when information loss is high, belief updating upon negative feedback is compromised due to high internal model uncertainty. At this point, the probability to err due to distractibility increases, as accounted by the failures to maintain a set measures.</ns0:p><ns0:p>Finally, the joint effect of &#948; and &#955; for high levels of information loss suggests that the roles played by the two cognitive parameters in accounting for adaptive functioning can be entangled when neuropsychological scoring criteria are considered.</ns0:p><ns0:p>Simulation 2: Information-Theoretic Analysis of the Bayesian Agent</ns0:p><ns0:p>In the following, we explore a different simulation scenario in which information-theoretic measures are derived to assess performance of the Bayesian cognitive agent. In particular, we explore the functional relationship between cognitive parameters and the dynamics of the recovered information-theoretic measures by simulating observed responses by varying flexibility across three levels, &#955; &#8712; {0.1, 0.5, 0.9}, and information loss across three levels, &#948; &#8712; {0.1, 0.5, 0.9}.</ns0:p><ns0:p>For this simulation scenario, we make no prior assumptions about sub-types of error classification.</ns0:p><ns0:p>Instead, we investigate the dynamic interplay between Bayesian surprise, B t , Shannon surprise, I t , and entropy, H t over the entire course of 128 trials in the WCST. In general, low information loss (&#948; ) ensures optimal behavior by speeding up internal model convergence by decreasing the number of trials needed to minimize uncertainty about the environmental states.</ns0:p><ns0:p>Low uncertainty reflects two main aspects of adaptive behavior. On the one hand, the probability that a response occurs due to sampling of improper rules decreases, allowing the agent to prevent random responses due to distractibility. On the other hand, model convergence entails a peaked Shannon surprise when a negative feedback occurs, due to the divergence between predicted and actual observations. Flexibility (&#955; ) plays a crucial role in integrating feedback information in order to enable belief updating. The first row depicted in Figure <ns0:ref type='figure' target='#fig_2'>4</ns0:ref> shows cognitive dynamics related to low information loss, across the levels of flexibility. As can be noticed, there is a positive relation between the magnitude of the Bayesian surprise and the level of flexibility, although unexpectedness yields approximately the same amount of signaling, as accounted by peaked Shannon surprise. From this perspective, surprise and belief updating can be considered functionally separable, where the first depends on the particular internal model probability configuration related to &#948; , whilst the second depends on flexibility &#955; .</ns0:p><ns0:p>However, more interesting patterns can be observed when information loss increases. In particular, model convergence slows down and several trials are needed to minimize predictive model entropy.</ns0:p><ns0:p>Casual errors might occur within trial windows characterized by high uncertainty, and interactions between entropy and Shannon surprise can be observes in such cases. In particular, Shannon surprise magnitude increases when model's entropy decreases, that is, during task phases in which the internal model has already converged. As a consequence, negative feedback could be classified as informative or uninformative, based on the uncertainty in the current internal model. This is reflected by the negative relation between entropy and Shannon surprise, as can be noticed by inspecting the graphs depicted in the third row of Figure <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>. Therefore, the magnitude of belief updating depends on the interplay between entropy and Shannon surprise, and can differ based on the values of the two measures in a particular task phase.</ns0:p><ns0:p>In sum, both simulation scenarios suggest that the simulated behavior of our generative model is in accord with theoretical expectations. Moreover, the flexibility and information loss parameters can account for a wide range of observed response patterns and inferred dynamics of information processing.</ns0:p></ns0:div> <ns0:div><ns0:head>Parameter Estimation</ns0:head><ns0:p>In this section, we discuss the computational framework for estimating the parameters of our model from observed behavioral data. Parameter estimation is essential to inferring the cognitive dynamics underlying observed behavior in real-world applications of the model. This section is slightly more technical and can be skipped without significantly affecting the flow of the text.</ns0:p></ns0:div> <ns0:div><ns0:head>Computational Framework</ns0:head><ns0:p>Rendering our cognitive model suitable for application in real-world contexts also entails accounting for uncertainty about parameter estimates. Indeed, uncertainty quantification turns out to be a fundamental and challenging goal when first-level quantities, that is, cognitive parameter estimates, are used to recover (second-level) information-theoretic measures of cognitive dynamics. The main difficulties arise when model complexity makes estimation and uncertainty quantification intractable at both analytical and numerical levels. For instance, in our case, probability distributions for the hidden model are generated at each trial, and the mapping between hidden states and responses changes depending on the structure of the task environment.</ns0:p><ns0:p>Identifying such a dynamic mapping is relatively easy from a generative perspective, but it becomes challenging, and almost impossible, when inverse modeling is required. Generally, this problem arises when the likelihood function relating model parameters to the data is not available in closed-form or too complex to be practically evaluated <ns0:ref type='bibr' target='#b65'>(Sisson and Fan (2011)</ns0:ref>). To overcome these limitations, we apply the first version of the recently developed BayesFlow method (see <ns0:ref type='bibr' target='#b58'>Radev et al. (2020)</ns0:ref> for mathematical details). At a high-level, BayesFlow is a simulation-based method that estimates parameters and quantifies estimation uncertainty in a unified Bayesian probabilistic framework when inverting the generative model is intractable. The method is based on recent advances in deep generative modeling and makes no assumptions about the shape of the true parameter posteriors. Thus, our ultimate goal becomes to</ns0:p></ns0:div> <ns0:div><ns0:head>13/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed ). Importantly, we can apply the same pre-trained inference network to an arbitrary number of real or simulated data sets (i.e., the training effort amortizes over multiple evaluations of the network).</ns0:p><ns0:p>For our purposes of validation and application, we train the network for 50 epochs which amount to 50000 forward simulations. As a prior, we use a bivariate continuous uniform distribution p(&#952;</ns0:p><ns0:formula xml:id='formula_12'>) &#8764; U ([0, 0], [1, 1]).</ns0:formula><ns0:p>We then validate performance on a separate validation set of 1000 simulated data sets with known ground-truth parameter values. Training the networks took less than a day on a single machine with an NVIDIA &#174; GTX1060 graphics card (CUDA version 10.0) using TensorFlow (version 1.13.1) <ns0:ref type='bibr' target='#b0'>(Abadi et al. (2016)</ns0:ref>). In contrast, obtaining full parameter posteriors from the entire validation set took approximately 1.78 seconds. In what follows, we describe and report all performance validation metrics.</ns0:p></ns0:div> <ns0:div><ns0:head>Performance Metrics and Validation Results</ns0:head><ns0:p>To assess the accuracy of point estimates, we compute the root mean squared error (RMSE) and the coefficient of determination (R 2 ) between posterior means and true parameter values. To assess the quality of the approximate posteriors, we compute a calibration error <ns0:ref type='bibr' target='#b58'>(Radev et al. (2020)</ns0:ref>) of the empirical coverage of each marginal posterior Finally, we implement simulation-based calibration (SBC, Talts et al.</ns0:p><ns0:p>(2018)) for visually detecting systematic biases in the approximate posteriors.</ns0:p><ns0:p>Point Estimates. Point estimates obtained by posterior means as well as corresponding RMSE and R 2 metrics are depicted in Figure <ns0:ref type='figure' target='#fig_6'>5A-B</ns0:ref>. Note, that point estimates do not have any special status in Bayesian inference, as they could be misleading depending on the shape of the posteriors. However, they are simple to interpret and useful for ease-of-comparison. We observe that pointwise recovery of &#955; is better than that of &#948; . This is mainly due to suboptimal pointwise recovery in the lower (0, 0.1) range of &#948; . This pattern is evident in Figure <ns0:ref type='figure' target='#fig_6'>5A</ns0:ref> Manuscript to be reviewed </ns0:p></ns0:div> <ns0:div><ns0:head>APPLICATION</ns0:head><ns0:p>In this section we fit the Bayesian cognitive model to real clinical data. The aim of this application is to evaluate the ability of our computational framework to account for dysfunctional cognitive dynamics of information processing in substance dependent individuals (SDI) as compared to healthy controls.</ns0:p></ns0:div> <ns0:div><ns0:head>Rationale</ns0:head><ns0:p>The advantage of modeling cognitive dynamics in individuals from a clinical population is that model predictions can be examined in light of available evidence about individual performance. For instance, SDIs are known to demonstrate inefficient conceptualization of the task and dysfunctional, error-prone response strategies. This has been attributed to defective error monitoring and behavior modulation systems, which depend on cingulate and frontal brain regions functionality <ns0:ref type='bibr' target='#b41'>(K&#252;bler et al. (2005)</ns0:ref>; Willuhn et al. ( <ns0:ref type='formula'>2003</ns0:ref>)). On the other hand, the WCST should be a rather easy and straightforward task for healthy participants to obtain excellent performance. Therefore, we expect our model to consistently capture such characteristics. To test these expectations, we estimate the two relevant parameters &#955; and &#948; from both clinical patients and healthy controls from an already published dataset <ns0:ref type='bibr' target='#b6'>(Bechara and Damasio (2002)</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The Data</ns0:head><ns0:p>The dataset used in this application consists of responses collected by administering the standard Heaton version of the WCST <ns0:ref type='bibr' target='#b34'>(Heaton (1981)</ns0:ref>) to healthy participants and SDIs. In this version of the task, the sorting rule changes when a participant collects a series of 10 consecutive correct responses, and the task ends when this happens for 6 times. Participants in the study consisted of 39 SDIs and 49 healthy individuals. All participants were adults (&gt; 18 years old) and gave their informed consent for inclusion which was approved by the appropriate human subject committee at the University of Iowa. SDIs were diagnosed as substance dependent based on the Structured Clinical Interview for DSM-IV criteria <ns0:ref type='bibr' target='#b20'>(First (1997)</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>15/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Model Fitting</ns0:head><ns0:p>We fit the Bayesian cognitive agent separately to data from each participant in order to obtain individuallevel posterior distributions. We apply the same BayesFlow network trained for the previous simulation studies, so obtaining posterior samples for each participant is almost instant (due to amortized inference).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The means of the joint posterior distributions are depicted for each individual in Figure <ns0:ref type='figure'>6</ns0:ref>, and provide a complete overview of the heterogeneity in cognitive sub-components at both individual and group levels (individual-level full joint posterior distributions can be found in the SI Appendix).</ns0:p><ns0:p>Figure <ns0:ref type='figure'>6</ns0:ref>. Joint posterior mean coordinates of the cognitive parameters, flexibility (&#955; ) and information loss (&#948; ), estimated for each individual. We observe a great heterogeneity in the distribution of posterior means, most pronouncedly for the flexibility parameter. However, a moderate between-subject variability in information loss can still be observed in both groups.</ns0:p><ns0:p>The estimates reveal a rather interesting pattern across both healthy and SDI participants. In particular, in both clinical and control groups, individuals with a poor flexibility (e.g., low values of &#955; ) can be detected. However, the group parameter space appears to be partitioned into two main clusters consisting of individuals with high and low flexibility, respectively. As can be noticed, the majority of SDIs belongs to the latter cluster, which suggests that the model is able to capture error-related defective behavior in the</ns0:p></ns0:div> <ns0:div><ns0:head>16/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed clinical population and attribute it specifically to the flexibility parameter. On the other hand, individual performance seems hardly separable along the information loss parameter dimension.</ns0:p><ns0:p>As a further validation, we compare the classification performance of two logistic regression models.</ns0:p><ns0:p>The first uses the estimated parameter means as inputs and the participants' binary group assignment (patient vs. control) as an outcome. The second uses the four standard clinical measures (non-perseverative errors (E), perseverative errors (PE), number of trials to complete the first category (TFC), number of failures to maintain set (FMS) computed from the sample as inputs and the same outcome. Since we are interested solely in classification performance and want to mitigate potential overfitting due to small sample size, we compute leave-one-out cross-validated (LOO-CV) performance for both models.</ns0:p><ns0:p>Interestingly, both logistic regression models achieve the same accuracy of 0.70, with a sensitivity of 0.71 and specificity of 0.70. Thus, it appears that our model is able to differentiate between SDIs and healthy individuals as good as the standard clinical measures.</ns0:p><ns0:p>However, as pointed out in the previous sections, estimated parameters serve merely as a basis to reconstruct cognitive dynamics by means of the trial-by-trial unfolding of information-theoretic measures.</ns0:p><ns0:p>Moreover, cognitive dynamics can only be analysed and interpreted by relying on the joint contribution of both estimated parameters and individual-specific observed response patterns.</ns0:p><ns0:p>To further clarify this concept, we investigate the reconstructed time series of information-theoretic quantities based on the response patterns of two exemplary individuals (Figure <ns0:ref type='figure' target='#fig_7'>7</ns0:ref>). In particular, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref> depicts the behavioral outcomes of a SDI with sub-optimal performance where the information-theoretic trajectories are reconstructed by taking the corresponding posterior means ([ &#955; = 0.07, &#948; = 0.82]), thus representing compromised flexibility and high information loss. Differently, Figure <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref> shows the information-theoretic path related to response dynamics of an optimal control participant, according to the parameter set [ &#955; = 0.60, &#948; = 0.35], representing relatively high flexibility, and low information loss.</ns0:p><ns0:p>Note, that in both cases, the reconstructed information-theoretic measures are based on the estimated posterior means for ease of comparison (see SI Appendix for the full joint posterior densities of the two exemplary individuals and the rest of the sample). 17/24 Processing unexpected observations is accounted by the quantification of surprise upon observing a response-feedback pair which is inconsistent with the current internal model of the task environment.</ns0:p><ns0:p>Negative feedback is maximally informative when errors occur after the internal model has converged to the true task model (grey area, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>), or the entropy approaches zero (grey line, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>). The Shannon surprise (orange line) is maximal when errors occur within trial windows in which the agent's uncertainty about environmental states is minimal (orange areas, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>). However, internal model updates following an informative feedback are not optimally performed, which is reflected by very small</ns0:p></ns0:div> <ns0:div><ns0:head>18/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Bayesian surprise (blue line, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>). This can be attributed to impaired flexibility and reflects the fact that after internal model convergence, informative feedback is not processed adequately and the internal model becomes impervious to change.</ns0:p><ns0:p>Conversely, errors occurring when the agent is uncertain about the true environmental state carry no useful information for belief updating, since the system fails to conceive such errors as unexpected and informative. The information loss parameter plays a crucial role in characterizing this cognitive behavior.</ns0:p><ns0:p>The slow convergence to the true environmental model, accompanied by the slow reduction of entropy in the predictive model, leads to a large number of trials required to achieve a good representation of the current task environment (white areas, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>). Errors occurring within trial windows with large predictive model entropy (green area, Figure <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>) do not affect subsequent behavior, and feedback is maximally uninformative.</ns0:p><ns0:p>Rather different cognitive dynamics can be observed in Figure <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref>, accounting for a typical optimal behavior where the errors produced fall within the trial windows which follow a rule completion (e.g. when the individual completes a sequence of 10 consecutive correct responses), and, thus, the environmental model becomes obsolete. However, the high flexibility, &#955; , allows to rely on local feedback-related information to suddenly update beliefs about the hidden states, that is, the most appropriate sorting rule. In this case, negative feedback become maximally informative after model convergence (grey area, Figure <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref>) and the process of entropy reduction (green line, Figure <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref>) is faster (e.g. less trials are needed) compared to the sub-optimal behavior scenario. Since uncertainty about the environmental states decreases faster, the Shannon surprise is always highly peaked when errors occur (orange line, Figure <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref>), thus ensuring an efficient employment of the local feedback-related information. Accordingly, higher values of Bayesian surprise are observed (blue line, Figure <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref>), revealing optimal internal model updating.</ns0:p><ns0:p>In general, the role that predictive (internal) model uncertainty plays in characterizing the way the agent processes feedback allows to disentangle sub-types of errors based on the information they convey for subsequent belief updating. From this perspective, error classification is entirely dependent on the status of the internal environmental model across task phases. Identifying such a dynamic latent process is therefore fundamental, since the error codification criterion evolves with respect to the internal information processing dynamics. Otherwise, the problem of inferring which errors are due to perseverance in maintaining an older (converged) internal model and which due to uncertainty about the true environmental state becomes intractable, or even impossible.</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Investigating information processing related to changing environmental contingencies is fundamental to understanding adaptive behavior. For this purpose, cognitive scientists mostly rely on controlled settings in which individuals are asked to accomplish (possibly) highly demanding tasks whose demands are assumed to resemble those of natural environments. Even in the most trivial cases, such as the WCST, optimal performance requires integrated and distributed neurocognitive processes. Moreover, these processes are unlikely to be isolated by simple scoring or aggregate performance measures.</ns0:p><ns0:p>In the current work, we developed and validated a new computational Bayesian model which maps distinct cognitive processes into separable information-theoretic constructs underlying observed adaptive behavior. We argue that these constructs could help describe and investigate the neurocognitive processes underlying adaptive behavior in a principled way.</ns0:p><ns0:p>Furthermore, we couple our computational model with a novel neural density estimation method for simulation-based Bayesian inference <ns0:ref type='bibr' target='#b58'>(Radev et al. (2020)</ns0:ref>). Accordingly, we can quantify the entire information contained in the data about the assumed cognitive parameters via a full joint posterior over plausible parameter values. Based on the joint posterior, a representative summary statistic can be computed to simulate the most plausible unfolding of information-theoretic quantities on a trial-by-trial basis.</ns0:p><ns0:p>Several computational models have been proposed to describe and explain performance in the WCST, ranging from behavioral <ns0:ref type='bibr' target='#b9'>(Bishara et al. (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b32'>Gl&#228;scher et al. (2019)</ns0:ref>; <ns0:ref type='bibr' target='#b66'>Steinke et al. (2020)</ns0:ref>) to neural network models <ns0:ref type='bibr' target='#b17'>(Dehaene and Changeux (1991)</ns0:ref>; <ns0:ref type='bibr' target='#b2'>Amos (2000)</ns0:ref>; <ns0:ref type='bibr' target='#b46'>Levine et al. (1993)</ns0:ref>; <ns0:ref type='bibr' target='#b50'>Monchi et al. (2000)</ns0:ref>).</ns0:p><ns0:p>These models aim to provide psychologically interpretable parameters or biologically inspired network structures, respectively, accounting for specific qualitative patterns of observed data. Behavioral models, in particular, abstract the main cognitive features underlying individual performance in the WCST according to different theoretical frameworks (e.g., attentional updating <ns0:ref type='bibr' target='#b9'>(Bishara et al. (2010)</ns0:ref>), or reinforcement 19/24 learning (Steinke et al. ( <ns0:ref type='formula'>2020</ns0:ref>))) and disentangle psychological sub-processes explaining observed task performance. However, the main advantage of our Bayesian model is that it provides both a cognitive and a measurement model which coexist within the overarching theoretical framework of Bayesian brain theories. More precisely, the presented model is specifically designed to capture trial-by-trial fluctuations in information processing as described by second-order information-theoretic quantities. The latter can be seen as a multivariate quantitative account of the interaction between the agent and its environment.</ns0:p><ns0:p>Moreover, it is worth noting that such a model representation might not be applicable outside a Bayesian theoretical framework.</ns0:p><ns0:p>Even though our computational model is not a neural model, it might provide a suitable description of cognitive dynamics at a representational and/or a computational level <ns0:ref type='bibr' target='#b48'>(Marr (1982)</ns0:ref>). This description can then be related to neural functioning underlying adaptive behavioral. Indeed, there is some evidence to suggest that neural processes related to belief maintenance/updating and unexpectedness are crucial for performance in the WCST. In particular, brain circuits associated with cognitive control and belief formation, such as the parietal cortex and prefrontal regions, seem to share a functional basis with neural substrates involved in adaptive tasks <ns0:ref type='bibr' target='#b52'>(Nour et al. (2018)</ns0:ref>). Prefrontal regions appear to mediate the relation between feedback and belief updating <ns0:ref type='bibr' target='#b47'>(Lie et al. (2006)</ns0:ref>) and efficient functioning in such brain structures seems to be heavily dependent on dopaminergic neuromodulation <ns0:ref type='bibr' target='#b56'>(Ott and Nieder (2019)</ns0:ref>). Moreover, the dopaminergic system plays a role in the processing of salient and unexpected environmental stimuli, in learning based on error-related information, and in evaluating candidate actions <ns0:ref type='bibr' target='#b52'>(Nour et al. (2018)</ns0:ref>; <ns0:ref type='bibr' target='#b16'>Daw et al. (2011);</ns0:ref><ns0:ref type='bibr' target='#b31'>Gershman (2018)</ns0:ref>). Accordingly, dopaminergic system functioning has been put in relation with performance in the WCST <ns0:ref type='bibr' target='#b36'>(Hsieh et al. (2010)</ns0:ref>; <ns0:ref type='bibr' target='#b60'>Rybakowski et al. (2005)</ns0:ref>) and shown to be critical for the main executive components involved in the task, that is, cognitive flexibility and set-shifting <ns0:ref type='bibr' target='#b8'>(Bestmann et al. (2014)</ns0:ref>; <ns0:ref type='bibr' target='#b67'>Stelzel et al. (2010)</ns0:ref>). Further, neural activity in the anterior cingulate cortex (ACC) is increased when a negative feedback occurs in the context of the WCST <ns0:ref type='bibr' target='#b47'>(Lie et al. (2006)</ns0:ref>).</ns0:p><ns0:p>This finding corroborates the view that the ACC is part of an error-detection network which allocates attentional resources to prevent future errors. The ACC might play a crucial role in adaptive functioning by encoding error-related or, more generally, feedback-related information. Thus, it could facilitate the updating of internal environmental models <ns0:ref type='bibr' target='#b59'>(Rushworth and Behrens (2008)</ns0:ref>).</ns0:p><ns0:p>The neurobiological evidence suggests that brain networks involved in the WCST might endow adaptive behavior by accounting for maintaining/updating of an internal model of the environment and efficient processing of unexpected information. Is it noteworthy, that these processing aspects are incorporated into our computational framework. At this point, we briefly outline the empirical and theoretical potentials of the proposed computational framework for investigating adaptive functioning and discuss future research vistas.</ns0:p><ns0:p>Model-Based Neuroscience. Recent studies have pointed out the advantage of simultaneously modeling and analyzing neural and behavioral data within a joint modeling framework. In this way, the latter can be used to provide information for the former, as well as the other way around <ns0:ref type='bibr' target='#b72'>(Turner et al. (2017</ns0:ref><ns0:ref type='bibr' target='#b73'>(Turner et al. ( , 2013))</ns0:ref>; <ns0:ref type='bibr' target='#b22'>Forstmann et al. (2011)</ns0:ref>). This involves the development of joint models which encode assumptions about the probabilistic relationships between neural and cognitive parameters.</ns0:p><ns0:p>Within our framework, the reconstruction of information-theoretic discrete time series yields a quantitative account of the agent's internal processing of environmental information. Event-related cognitive measures of belief updating, epistemic uncertainty and surprise can be put in relation with neural measurements by explicitly providing a formal account of the statistical dependencies between neural and cognitive (information-theoretic) quantities. In this way, latent cognitive dynamics can be directly related to neural event-related measures (e.g., fMRI, EEG). Applications in which information-theoretic measures are treated as dependent variables in standard statistical analysis are also possible.</ns0:p><ns0:p>Neurological Assessment. Although neuroscientists have considered performance in the WCST as a proxy for measuring high-level cognitive processes, the usual approach to the analysis of human adaptive behavior consists in summarizing response patterns by simple heuristic scoring measures (e.g., occurrences of correct responses and sub-types of errors produced) and classification rules <ns0:ref type='bibr' target='#b21'>(Flashman et al. (1991)</ns0:ref>).</ns0:p><ns0:p>However, the theoretical utility of such a summary approach remains questionable. Indeed, adaptive behavior appears to depend on a complex and intricate interplay between multiple network structures <ns0:ref type='bibr' target='#b4'>(Barcelo et al. (2006)</ns0:ref>; <ns0:ref type='bibr' target='#b49'>Monchi et al. (2001)</ns0:ref>; <ns0:ref type='bibr' target='#b47'>Lie et al. (2006)</ns0:ref>; <ns0:ref type='bibr' target='#b5'>Barcel&#243; and Rubia (1998)</ns0:ref>; <ns0:ref type='bibr' target='#b12'>Buchsbaum et al. (2005)</ns0:ref>). This posits a great challenge for disentangling high-level cognitive constructs at a model level and further investigating their relationship with neurobiological substrates. It appears that standard scoring</ns0:p></ns0:div> <ns0:div><ns0:head>20/24</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed measures might not be able to fulfil these tasks. Moreover, there is a pronounced lack of anatomical specificity in previous research concerning the neural and functional substrates of the WCST <ns0:ref type='bibr' target='#b53'>(Nyhus and Barcel&#243; (2009)</ns0:ref>).</ns0:p><ns0:p>Thus, there is a need for more sophisticated modeling approaches. For instance, disentangling errors due to perseverative processing of previously relevant environmental models from those due to uncertainty about task environmental states, is important and nontrivial. Sparse and distributed error patterns might depend on several internal model probability configurations. Such internal models are latent, and can only be uncovered through cognitive modeling. Therefore, information-based criteria to response (error) classification can enrich clinical evaluation beyond heuristically motivated criteria.</ns0:p><ns0:p>Generalizability. Another important advantage of the proposed computational framework is that it is not solely confined to the WCST. In fact, one can argue that the seventy-year old WCST does not provide the only or even the most suitable setting for extracting information about cognitive dynamics from general populations or maladaptive behavior in clinical populations. One can envision tasks which embody probabilistic (uncertain) or even chaotic environments (for instance with partially observable or unreliable feedback or partially observable states) and demand integrating information from different modalities <ns0:ref type='bibr' target='#b54'>(O'Reilly et al. (2013)</ns0:ref>; <ns0:ref type='bibr' target='#b52'>Nour et al. (2018)</ns0:ref>). These settings might prove more suitable for investigating changes in uncertainty-related processing or cross-modal integration than deterministic and fully observable WCST-like settings.</ns0:p><ns0:p>Despite these advantages, our proposed computational framework has certain limitations. A first limitation might concern the fact that the new Bayesian cognitive model accounts for the main dynamics in adaptive tasks by relying on only two parameters. Although such a parsimonious proposal suffices to disentangle latent data-generating processes, a more exhaustive formal description of cognitive subcomponents might be envisioned. However, parameter estimation can become challenging in such a scenario, especially when one-dimensional response data is used as a basis for parameter recovery. Second, the information loss parameter appears to be more challenging to estimate than the flexibility parameter in some datasets. There are at least two possible remedies for this problem. On the one hand, global estimation of information loss might be hampered due to the model's current functional (algorithmic)</ns0:p><ns0:p>formulation and can therefore be optimized via an alternative formulation/parameterization. On the other hand, it might be the case that the data obtainable in the simple WCST environment is not particularly informative about this parameter and, in general, not suitable for modeling more complex and non-linear cognitive dynamics in general. Future works should therefore focus on designing and exploring more data-rich controlled environments which can provide a better starting point for investigating complex latent cognitive dynamics in a principled way. Additionally, the information loss parameter seems to be less effective in differentiating between substance abusers and healthy controls in the particular sample used in this work. Thus, further model-based analyses on individuals from different clinical populations are needed to fully understand the potential of our 2-parameter model as a clinical neuropsychological tool.</ns0:p><ns0:p>Finally, in this work, we did not perform formal model comparison, as this would require an extensive consideration of various nested and non-nested model within the same theoretical framework and between different theoretical frameworks. We therefore leave this important endeavor for future research.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>In conclusion, the proposed model can be considered as the basis for a (bio)psychometric tool for measuring the dynamics of cognitive processes under changing environmental demands. Furthermore, it can be seen as a step towards a theory-based framework for investigating the relation between such cognitive measures and their neural underpinnings. Further investigations are needed to refine the proposed computational model and systematically explore the advantages of the Bayesian brain theoretical framework for empirical research on high-level cognition.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Suppose the correct sorting rule is the feature shape. The figure shows the rate of convergence of the predictive distributions to the true task environmental model. The predictive distributions at trial t + 1 depends on the sorting action a t (first row) and the received feedback f t (second row). Two examples of updating a predictive distribution are shown: one in which information loss is high (&#948; = 0.7, third row), and one in which information loss is low (&#948; = 0.3, fifth row). High information loss slows down the convergence of the internal model to the true environmental model. The gray bar plots represent the predictive probability distribution over the rules from which an action is sampled at each trial. Dotted bars represent the updated predictive distribution after the feedback observation. For each scenario, trial-by-trial information-theoretic measures are shown.</ns0:figDesc><ns0:graphic coords='9,143.80,149.77,409.44,374.89' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Clinical scoring measures as functions of flexibility (&#955; ) and information loss (&#948; ) -simulated scenarios. The different cells show the violin plots for the estimated distribution densities of the scoring measures obtained from the group of synthetic individuals, for the levels of &#955; across different levels of &#948; . In particular, they show the distribution of non-perseverative errors (E), perseverative errors (PE), number of trials to complete the first category (TFC), number of failures to maintain set (FMS) obtained from 150 synthetic agent's response simulations for each cell of the factorial design.</ns0:figDesc><ns0:graphic coords='13,143.80,129.10,409.44,452.10' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Information-theoretic measures varying as a function of flexibility &#955; and information loss &#948; across 128 trials of the WCST. Optimal belief updating and uncertainty reduction are achieved with low information loss and high flexibility (first row, third column).</ns0:figDesc><ns0:graphic coords='14,143.80,351.76,409.42,233.36' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4 depicts results from the nine simulation scenarios. Although an exhaustive discussion on cognitive dynamics should couple information-theoretic measures with patterns of correct and error responses, we focus solely on the information-theoretic time series for illustrative purposes. We refer to the Application section for a more detailed description of the relation between observed responses and estimated information-theoretic measures in the context of data from a real experiment. Again, simulated performance of the Bayesian cognitive agent shows that different parameter combinations yield different patterns of cognitive dynamics. Observed spikes and their related magnitudes</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>approximate and analyze the joint posterior distribution over the model parameters. The parameter posterior is given via an application of Bayes' rule: p(&#952; |x 0:T , m 0:T ) = p(x 0:T , m 0:T |&#952; )p(&#952; ) p(x 0:T , m 0:T |&#952; )p(&#952; )d&#952; (10) where we set &#952; = (&#955; , &#948; ) and stack all observations and matching signals into the vectors x 0:T = (x 0 , x 1 , ..., x T ) and m 0:T = (m 0 , m 1 , ..., m T ), respectively. The BayesFlow method uses simulations from the generative model to optimize a neural density estimator which learns a probabilistic mapping between raw data and parameters. It relies on the fact that data can easily be simulated by repeatedly running the generative model with different parameter configurations &#952; sampled from the prior. During training, the neural network estimator iteratively minimizes the divergence between the true posterior and an approximate posterior. Once the network has been trained, we can efficiently obtain samples from the approximate joint posterior distribution of the cognitive parameters of interest, which can be further processed in order to extract meaningful summary statistics (e.g., posterior means, medians, modes, etc.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>-B and is due to the fact that &#948; values in this range produce almost indistinguishable data patterns. Bootstrap estimates yielded an average RMSE of 0.155 (SD = 0.004) and an average R 2 of 0.708 (SD = 0.015) for the &#948; parameter. An average RMSE of 0.094 (SD = 0.002) and an average R 2 of 0.895 (SD = 0.007) were obtained for the &#955; parameter. These results suggest good global pointwise recovery but also warrant the inspection of full posteriors, especially in the low ranges of &#948; .Full Posteriors. Average bootstrap calibration error was 0.011 (SD = 0.005) for the marginal posterior of &#948; and 0.014 (SD = 0.007) for the marginal posterior of &#955; . Calibration error is perhaps the most important metric here, as it measures potential under-or overconfidence across all confidence intervals of the approximate posterior (i.e., an &#945;-confidence interval should contain the true posterior with a probability of &#945;, for all &#945; &#8712; (0, 1)). Thus, low calibration error indicates a faithful uncertainty representation of the approximate posteriors. Additionally, SBC-histograms are depicted in Figure5C-D. As shown by (Talts et al. (2018)), deviations from the uniformity of the rank statistic (also know as a PIT histogram) indicate systematic biases in the posterior estimates. A visual inspection of the histograms reveals that the posterior means slightly overestimate the true values of &#948; . This corroborates the pattern seen in Figure 5A-B for the lower range of &#948; . Finally, Figure 5E-H depicts the full marginal posteriors on two example validation sets. Even on these two data sets, we observe strikingly different posterior shapes. The marginal posterior of &#948; obtained from the first data set is slightly left-skewed and has its density concentrated over the (0.8, 1.0) range. On the other hand, the marginal posterior of &#948; from the second data set is noticeably right-skewed and peaked across the lower range of the parameter. The marginal posteriors of &#955; appear more symmetric and warrant 14/24 PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Parameter recovery results on validation data; (A and B) Posterior means vs. true parameter values; (C and D) Histograms of the rank statistic used for simulation-based calibration; (E-H) Example full posteriors for two validation data sets; (I and J) Example information-theoretic dynamics recovered from the parameter posteriors.</ns0:figDesc><ns0:graphic coords='17,141.73,63.78,413.57,212.57' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Recovered cognitive dynamics of two exemplary individuals. (A) Trial-by-trial information-theoretic measures of a SDI characterized by very low flexibility and very high information loss; (B) Trial-by-trial information-theoretic measures of a healthy individual characterized by relatively high flexibility and low information loss. Labels C and E indicate correct and error responses.</ns0:figDesc><ns0:graphic coords='20,141.73,63.78,413.58,334.87' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,143.80,187.77,409.44,409.44' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Descriptive summary of all quantities involved in our model representation.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Observation</ns0:cell><ns0:cell>Pair of action and feedback which constitutes the agent's</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>observation in trial t.</ns0:cell></ns0:row><ns0:row><ns0:cell>&#915;(t)</ns0:cell><ns0:cell>Stability matrix</ns0:cell><ns0:cell>Matrix encoding the agent's beliefs about state transi-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>tions from trial t to the next trial t + 1.</ns0:cell></ns0:row><ns0:row><ns0:cell>&#955; &#8712; [0, 1]</ns0:cell><ns0:cell>Flexibility</ns0:cell><ns0:cell>Parameter encoding the efficiency to disengage attention</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>from a currently attended hidden state when signaled by</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>the environment.</ns0:cell></ns0:row><ns0:row><ns0:cell>&#948; &#8712; [0, 1]</ns0:cell><ns0:cell>Information loss</ns0:cell><ns0:cell>Parameter encoding how efficiently the agent's internal</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>model converges to the true environmental model based</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>on experience.</ns0:cell></ns0:row><ns0:row><ns0:cell>m (i)</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>Expression Name Description s t &#8712; {1, 2, 3} Sorting rule Card feature relevant for the sorting criterion in trial t. a t &#8712; {1, 2, 3, 4} Choice action Action of choosing one of the four stimulus cards in trial t. f t &#8712; {0, 1} Feedback Indicates whether the action of matching a stimulus to a target card is correct or not in trial t. x t = (a t , f t ) t &#8712; {0, 1} Matching signal Signal indicating whether feature i is relevant in trial t based on the feedback received. &#969; (i) t &#8712; [0, 1] State activation level Agent's internal measure of the accrued evidence for the hidden environmental state i in trial t. B t &#8712; R + Bayesian surprise Kullback-Leibler divergence between prior and posterior beliefs about hidden environmental states in trial t. I t &#8712; R + Shannon surprise Information-theoretic surprise encoding the improbability or unexpectedness of an observation in trial t. H t &#8712; R + Entropy Degree of epistemic uncertainty in the internal model of the environment in trial t.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Mean clinical scoring measures as functions of flexibility (&#955; ) and information loss (&#948; ). Cells show the average scores across simulated agents (standard deviation is shown in parenthesis).</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>07 (2.68)</ns0:cell><ns0:cell>7.95 (2.07)</ns0:cell><ns0:cell>7.50 (2.13)</ns0:cell><ns0:cell>6.85 (1.75)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell>10.84 (2.35)</ns0:cell><ns0:cell>9.60 (2.2)</ns0:cell><ns0:cell>8.25 (2.23)</ns0:cell><ns0:cell>7.37 (1,74)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell cols='3'>12.75 (2.96) 11.25 (2.43) 9.12 (2.09)</ns0:cell><ns0:cell>7.79 (1.73)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell cols='4'>20.81 (2.27) 18.18 (1.88) 14.99 (1.88) 12.37 (1.12)</ns0:cell></ns0:row><ns0:row><ns0:cell>Perseverative Errors (PE)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell cols='4'>19.77 (2.55) 17.65 (2.26) 15.42 (1.94) 12.39 (1.47)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell cols='4'>18.56 (2.76) 16.58 (2.53) 14.49 (2.03) 12.33 (1.44)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell cols='4'>12.20 (1.46) 11.91 (1.35) 11.83 (1.24) 11.67 (1.04)</ns0:cell></ns0:row><ns0:row><ns0:cell>Trials to First Category (TFC)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell cols='4'>13.82 (2.76) 13.32 (2.52) 12.97 (2.13) 12.29 (1.53)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell cols='4'>17.27 (4.21) 16.63 (4.04) 14.39 (3.58) 12.91 (1.91)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.4</ns0:cell><ns0:cell>0.11 (0.31)</ns0:cell><ns0:cell>0.09 (0.31)</ns0:cell><ns0:cell>0.05 (0.32)</ns0:cell><ns0:cell>0.02 (0.14)</ns0:cell></ns0:row><ns0:row><ns0:cell>Failures to Maintain Set (FMS)</ns0:cell><ns0:cell>&#948; = 0.7</ns0:cell><ns0:cell>1.65 (1.4)</ns0:cell><ns0:cell>1.41 (1.3)</ns0:cell><ns0:cell>0.84 (0.91)</ns0:cell><ns0:cell>0.35 (0.69)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&#948; = 0.9</ns0:cell><ns0:cell>4.44 (1.96)</ns0:cell><ns0:cell>3.88 (1.86)</ns0:cell><ns0:cell>2.79 (1.56)</ns0:cell><ns0:cell>1.54 (1.25)</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='10'>/24PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:note> <ns0:note place='foot' n='12'>/24PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:note> <ns0:note place='foot' n='24'>/24PeerJ reviewing PDF | (2020:07:50686:2:0:NEW 10 Oct 2020)</ns0:note> </ns0:body> "

"RESPONSE LETTER We thank the editor and the reviewers for the positive reception of our ideas and the concrete and helpful suggestions for improving the manuscript. Below you can find a point-by-point summary of the changes introduced the new version of the manuscript. Rewiever 1 (R1) Comments R1_1: “Everything is fine by my side, please only check your references. Probably you used an old bibtex file because there are a lot of missing references (e.g. page 2, lines 79-82).”. A_1: Thanks for the comment. We fixed the problems with the bibliography. Rewiever 2 (R2) Comments R2_1: “In the example the authors give of a perceptual task, the authors could have chosen a more intuitive/more tangible example (e.g. judging if an item in the sky is a plane or a bird; the weight of a mug of coffee and the force needed to grab it without spilling; etc.). This is, of course, a personal preference, and the authors are free to just keep the example as is.”. A_1: We really thank you for the advice, which we consider valuable in general, but we prefer to keep the (technical) perception example since it is more popular in this particular kind of modeling papers. R2_2: “There are a few question marks instead of possible citations (e.g. neurocognitive mechanisms, such as motor control (?), oculomotor dynamics70 (?), object recognition (?), attention (?) ). This is probably a reference manager/compilation issue but should be corrected.”. A_2: Thanks for the comment. We fixed the problems with the bibliography. R2_3: “Just before Figure 1 – replace “noticed” by “notice” (“Noticed that both features and sorting rules refer to the same” – line 148 of the clean manuscript)” A_3: We fixed the typo. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Introduction: Despite careful patient selection for subthalamic nucleus deep brain stimulation (STN DBS), some Parkinson's disease patients show limited improvement of motor disability. Innovative predictive analysing methods hold potential to develop a tool for clinicians that reliably predicts individual postoperative motor response, by only regarding clinical preoperative variables. The main aim of preoperative prediction would be to improve preoperative patient counselling, expectation management, and postoperative patient satisfaction.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods:</ns0:head><ns0:p>We developed a machine learning logistic regression prediction model which generates probabilities for experiencing weak motor response one year after surgery. The model analyses preoperative variables and is trained on 89 patients using a five-fold cross-validation. Imaging and neurophysiology data are left out intentionally to ensure usability in the preoperative clinical practice. Weak responders (n = 30) were defined as patients who fail to show clinically relevant improvement on Unified Parkinson Disease Rating Scale II, III or IV.</ns0:p></ns0:div> <ns0:div><ns0:head>Results:</ns0:head><ns0:p>The model predicts weak responders with an average area under the curve of the receiver operating characteristic of 0.79 (standard deviation: 0.08), a true positive rate of 0.80 and a false positive rate of 0.24, and a diagnostic accuracy of 78%. The reported influences of individual preoperative variables are useful for clinical interpretation of the model, but cannot been interpreted separately regardless of the other variables in the model.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion:</ns0:head><ns0:p>The model's diagnostic accuracy confirms the utility of machine learning based motor response prediction based on clinical preoperative variables.After reproduction and validation in a larger and prospective cohort, this prediction model holds potential to support clinicians during preoperative patient counseling.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Subthalamic nucleus deep brain stimulation (STN DBS) is a widely accepted therapy for Parkinson's disease (PD) patients in which dopaminergic replacement therapy is unsatisfactory. <ns0:ref type='bibr' target='#b6'>(Deuschl et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b15'>Limousin et al. 1995;</ns0:ref><ns0:ref type='bibr' target='#b19'>Odekerken et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b24'>Schuepbach et al. 2013)</ns0:ref> In the majority of these patients, DBS can reduce motor symptoms or their fluctuations and thereby improve quality of life. <ns0:ref type='bibr' target='#b26'>(Williams et al. 2010</ns0:ref>) Despite careful patient selection, some patients still show limited or no improvement of motor fluctuations and quality of life. <ns0:ref type='bibr' target='#b26'>(Williams et al. 2010)</ns0:ref> Since the introduction of STN DBS, clinicians aimed to determine reliable predictors. <ns0:ref type='bibr' target='#b22'>(Pinter et al. 1999)</ns0:ref> Preoperative levodopa responsiveness of motor symptoms, severity of motor symptoms, and younger age are repeatedly reported as positive predictive factors for postoperative (Movement Disorders Society -) Unified Parkinson's Disease Rating Scale ((MDS-)UPDRS) motor improvement. <ns0:ref type='bibr' target='#b13'>(Kleiner-Fisman et al. 2006</ns0:ref>) Contrarily, preoperative levodopa responsiveness is also reported to not predict STN DBS outcome. <ns0:ref type='bibr' target='#b25'>(Schuepbach et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b28'>Zaidel et al. 2010)</ns0:ref> Preoperative severe quality of life (QoL) impairment, more time spent in off-condition of dopaminergic medication, levodopa responsiveness, and low BMI are reported as positive predictive factors on postoperative QoL. <ns0:ref type='bibr' target='#b0'>(Abboud et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b5'>Daniels et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b8'>Frizon et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b25'>Schuepbach et al. 2019)</ns0:ref> Reports on the predictive value of disease duration, daily levodopa dosage, postural and gait impairment, and non-motor symptoms all show conflicting results. <ns0:ref type='bibr' target='#b4'>(Dafsari et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b8'>Frizon et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kleiner-Fisman et al. 2006;</ns0:ref><ns0:ref type='bibr' target='#b17'>Liu et al. 2018)</ns0:ref> Comparison of reported motor outcome is hampered due to variance in assessment scales and assessments during varying dopaminergic states. <ns0:ref type='bibr' target='#b9'>(Goetz et al. 2008)</ns0:ref> These non-conclusive results maintain the need for a simple tool which neurologists can use in clinical practice to predict motor outcome after STN DBS for individual patients. To realize a usable and representative tool for the preoperative setting, our approach is limited to preoperative clinical variables. Preoperative prediction will always lack surgical information such as lead placement. This lack of information is inherent to any approach that aims to contribute to a better preoperative counselling. Machine learning methods are increasingly used in medical practice to unravel patterns to improve understanding of clinical data. <ns0:ref type='bibr' target='#b18'>(Meyer et al. 2018</ns0:ref>) Predictive machine learning models can be distinguished from traditional statistics by generating outcome predictions for new, individual patients, instead of correlations between pre-and postoperative variables on a group level. To ensure practical usability, clinical relevance, and interpretable results, the development and implementation of these models requires statistical, programming, and clinical expertise.(Pieter Kubben 2019) To add value to PD care, predictive analysis should improve challenging clinical decision making instead of reproduce valid clinical decisions. <ns0:ref type='bibr' target='#b1'>(Ballarini et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b2'>Cerasa 2016</ns0:ref>) Here, we report the development and proof-of-concept of a prediction model that generates probabilities for weak and strong motor response one year after STN DBS for individual PD patients based on preoperative clinical variables.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Study population</ns0:head><ns0:p>We considered patients who underwent STN DBS for PD in our academic neurosurgical centre between 2004 and March 2018. The surgical procedure is described in the Supplemental Material. We included 127 patients who completed one-year postoperative follow up during this period. We excluded patients who had missing UPDRS-III scores in their preoperative onmedication condition, or postoperative on-medication, on-stimulation condition. The Medical Ethical Committee of Maastricht UMC+ approved this study (2018-0739). Informed consent was not obtained since the retrospective data was collected coded.</ns0:p></ns0:div> <ns0:div><ns0:head>Pre-and postoperative variables</ns0:head><ns0:p>All available preoperative demographic data, disease specific data (disease onset, disease duration, levodopa equivalent daily dosage (LEDD)), <ns0:ref type='bibr' target='#b7'>(Esselink et al. 2004</ns0:ref>) clinical performance scores ((MDS)-UPDRS, and Hoehn &amp; Yahr (H&amp;Y) scores), as well as relevant neuropsychological scores assessing executive functioning, in particular verbal fluency (semantic and lexical) and response inhibition (based on the interference score of the Stroop Colour Word Test) were incorporated. We left out imaging and neurophysiology data, to ensure the userfriendliness and accessibility in clinical practice during preoperative counselling. No analyses are required which ask software, hardware, or analysing knowledge. All included preoperative clinical and neuropsychological scores were assessed in the onmedication condition and the available (MDS-)UPDRS III and H&amp;Y scores in the off-medication condition were also included. Preoperative motor levodopa-responsiveness was calculated by subtracting UPDRS III scores in the off-medication condition with UPDRS III scores in the onmedication condition. Postoperative collected variables consist of UPDRS I, II, III and IV and H&amp;Y scores in on-medication and on-stimulation conditions, UPDRS III in on-stimulation and off-medication conditions, and the performance on the verbal fluency and Stroop tests in onstimulation and on-medication conditions. Both MDS-UPDRS and UPDRS scores were collected due to the variation in surgery dates among the population. To create uniform UPDRS scores, all MDS-UPDRS scores were recalculated to UPDRS scores. <ns0:ref type='bibr' target='#b9'>(Goetz et al. 2008</ns0:ref>) Pre-and postoperative differences for UPDRS scores I until IV, H&amp;Y scores, LEDD, and neuropsychological scores were calculated. Furthermore, we registered applied DBS voltage, frequency, and pulse width at one-year follow up. To compare DBS-settings, we computed the mean total electrical energy delivered (TEED). <ns0:ref type='bibr' target='#b14'>(Koss et al. 2005)</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Prediction model</ns0:head><ns0:p>The machine learning prediction model uses multivariate logistic regression analyses. This logistic regression model distinguishes itself from (univariate) correlative regression models by generating outcome probabilities for individual patients (fig. <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). We focused on motor response as outcome and differentiated between 'strong responders' and 'weak responders'. To capture a wide spectrum of motor responders, improvement on UPDRS II, III and IV was evaluated. A strong responder is defined as a patient who showed a minimal clinically important difference (MCID) on UPDRS II, III, or IV in on-medication and on-stimulation condition one-year postoperative vs. preoperative on-medication condition (see fig. <ns0:ref type='figure'>2</ns0:ref>). MCID was defined as more than 3, 5, and 3 points improvement for UPDRS II, III and IV respectively, based on a literature review (see Supplemental Material). Patients who improved more than the MCID on UPDRS II or IV, but showed a deterioration on UPDRS III of more than the MCID and the yearly natural disease progression together 7 points, were defined as weak responders <ns0:ref type='bibr' target='#b10'>(Holden et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b11'>Horvath et al. 2015)</ns0:ref>. The prediction model uses the following available preoperative variables to generate an outcome probability: gender, age at DBS, PD duration at DBS, age at PD onset, UPDRS I, II, III and IV in on-medication condition, motor levodopa response, H&amp;Y scale in on-and off-condition, the Stroop interference score, the verbal fluency scores, and the LEDD. The logistic regression model was fitted, i.e. trained, on the relation between preoperative variables and postoperative outcome categorization. <ns0:ref type='bibr' target='#b20'>(Pedregosa et al. 2011)</ns0:ref> We evaluated the trained model with a 5-fold cross-validation. This cross-validation fits, i.e. trains, the model on 80% of the patients, the 'training data'. During this 'training phase', a weight, '&#946;', is assigned to every single preoperative variable, 'x'. The fitted model was then evaluated, i.e. tested, on the remaining 20% of patients in the database, the 'test data'. During this 'test phase', the preoperative variables of every individual patient in the test data were inserted in the model separately. The model generates an outcome probability to become a weak responder for every individual patient. This probability was generated by a calculation of all 'x' values of the inserted patient with the corresponding weights (&#946;) using the logistic function 1 / (1 + exp(-&#946; * x)). The generated probabilities from the test data are compared with the actual outcome to test predictive accuracy. The 5-fold cross-validation repeats these phases 5 times until every patient was used for testing exactly once. The cross-validation leads to less limitations in sample size regarding number of considered predictive variables. <ns0:ref type='bibr' target='#b27'>(Wynants et al. 2015)</ns0:ref> Still, the small number of patients on which the trained model is tested during every iteration in this 5-fold cross-validation is a limitation of this approach. Evaluating the average performance over the 5 iterations gives the best assumption of the predictive performance of the model. We chose logistic regression as a prediction model instead of a deep learning-based model due to the relatively small database size and the fact that the weight, or influence, of every preoperative variable can be interpret easily. This interpretation helps to generate an intuition what the prediction is based on. <ns0:ref type='bibr' target='#b23'>(Rudin 2019)</ns0:ref> To use a certain prediction model in clinical practice, a threshold should be chosen to accept a probability. This means every probability above the threshold is regarded to be true (weak response in this model), and every probability below the threshold is regarded to be false (strong response in this model). The accuracy of the model is strongly dependent on the threshold. A common way to evaluate the overall performance of a prediction model is to plot the receiver operating characteristic (ROC). The ROC visualizes for different thresholds between 0 and 1 the corresponding true positive and false positive rates (fig <ns0:ref type='figure'>3A</ns0:ref>). Performance of prediction models is often expressed as the area under the curve (AUC) of the ROC (fig. <ns0:ref type='figure'>3A</ns0:ref>). In clinical practice, a threshold should be selected before the model can be used as a prospective application. To understand which variables are important in the prediction model, we can explore the importance of every separate preoperative variable. Variable importance is expressed as 'weights'. To make these weights interpretable, they are converted to Odds Ratios by calculating exp(&#946;), and normalized afterwards. These normalized Odds Ratios are called 'relative influences', and they denote the change in probability to be a weak responder when the respective variable increases 1 unit, and all other variables stay equal (fig <ns0:ref type='figure'>3A</ns0:ref>). Comparative descriptive analysis between preoperative and postoperative variables and between weak and strong responders are performed with Mann-Whitney-U-tests. To facilitate prediction models, we imputed missing data-points in preoperative variables. (For further explanation on the Random Forest imputations applied on preoperative variables, please see Supplementary Material). To prevent imputations of variables that are the target of prediction, we did not impute postoperative variables. Analysis is performed in Python Jupyter Notebook 3 (Jupyter Team, https://jupyter.org, revision fe7c2909) using packages pandas (version 1.0.4), Numpy (version 1.16.4), scikit-learn (version 0.21.2), and Scipy (version 1.3.0). We report our findings according to the TRIPOD Checklist for Prediction Model Development. <ns0:ref type='bibr' target='#b3'>(Collins et al. 2015</ns0:ref>)</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Preoperative and postoperative variables</ns0:head><ns0:p>We included 89 patients with a well-documented one year follow up after STN DBS, 37 patients were excluded due to missing data points in UPDRS III score in preoperative on-medication condition, or postoperative on-medication and on-stimulation condition. We report descriptive statistics containing the original data (no imputed preoperative data). The total group showed statistically significant postoperative improvements in UPDRS III scores, both compared with preoperative on-and off-medication conditions, and in UPDRS IV scores. We observed a significant decrease in LEDD (table 1). Further, there was a significant deterioration in neuropsychological scores on a group level. 60 out of 89 patients were categorized as strong responders, 30 patients were categorized as weak responders (fig. <ns0:ref type='figure'>2</ns0:ref>). Postoperative clinical records until one-year follow-up were evaluated and surgical factors explaining weak response were ruled out for all weak responders. The groups had significant differences on all postoperative UPDRS scores and differences, except for the UPDRS III during on-stimulation and off-medication state (see table 2). We observed no significant or relevant differences between the groups regarding neuropsychological scores, LEDD, or TEED.</ns0:p></ns0:div> <ns0:div><ns0:head>Performance of the prediction model</ns0:head><ns0:p>The prediction model has a good general performance with an average AUC of the ROC of 0.79 (standard deviation: 0.08) (figure <ns0:ref type='figure'>3A</ns0:ref>). When 0.29 is chosen as a threshold for accepting probabilities to become a weak responder, this leads to a true positive rate of 0.80 and a false positive rate of 0.24 (fig. <ns0:ref type='figure'>3A-B</ns0:ref>). This corresponds to a positive predictive value of 0.63 and a negative predictive value of 0.88. Selecting 0.29 as the threshold for probability acceptance leads to a classification accuracy of 78%, since 69 out of 89 patients are predicted correctly.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:03:46750:1:2:NEW 5 Aug 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The relative influence values represent the influence, or weight, of each preoperative variable in the prediction model (fig. <ns0:ref type='figure'>3C</ns0:ref>). Older age at PD onset has the strongest relative influence for becoming a weak responder. High preoperative UPDRS III and IV scores in the on-medication condition are the strongest predictors for becoming a strong responder (figure <ns0:ref type='figure'>3C</ns0:ref>). Additionally, a high preoperative UPDRS II score, high scores on the categorical fluency and Stroop interference test, and higher H&amp;Y score in the on-condition were moderate predictors for becoming a strong responder.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head></ns0:div> <ns0:div><ns0:head>Proof of concept of machine learning prediction in preoperative DBS outcome counselling</ns0:head><ns0:p>The presented machine learning model differentiated between individual weak and strong motor responders one-year after STN DBS for PD with a good overall predictive performance, the AUC of the ROC was 0.78, and the classification accuracy was 0.78% (fig. <ns0:ref type='figure'>3A-B</ns0:ref>). These results contribute to a proof-of-concept of machine learning prediction of individual postoperative motor outcome, solely based on preoperative clinical variables. We want to stress that these results are the first step towards the clinical utilization of smart supportive computational models in the delicate, multifactorial decision-making process of DBS therapy counselling. To increase the likelihood of creating a beneficial clinical impact for the patient, a model should be interpretable for clinicians, generalizable over the aimed patient population, and the effect of a utilization on the quality of clinical care should be investigated. <ns0:ref type='bibr' target='#b12'>(Kelly et al. 2019</ns0:ref>) Interpretation of the predictive performance and the clinical utilization A predictive machine learning model for clinical support generates individual outcome probabilities range from 0 to 1, rather than binary classes. The presented confusion matrix is an example of a clinical utilization where probabilities to become a weak responder higher than 0.29 were accepted (see fig. <ns0:ref type='figure'>3A-B</ns0:ref>). The selection of this threshold will eventually determine the model's clinical behaviour, usefulness, and its potential clinical impact on patient care. The value of this threshold leads to a different balance between false positive and false negative predictions (fig. <ns0:ref type='figure'>3B</ns0:ref>), and should be validated on an external cohort. The presented threshold is chosen to realize a good accuracy (78%) and to fit to the intended clinical utilization of this model. Since the majority of STN DBS candidates will experience a strong response, it is important that the clinician can trust a strong response prediction (negative predictive rate (0.88). Also, the model should create awareness about the chance of becoming a weak responder in case of increased risk. This requires a true positive rate, here 0.80. Further, the confusion matrix shows that most incorrect predictions are actual strong responders who get a weak responder prediction. The final decision will be accurately guided by the experience of the DBS team and will overrule the majority of these predictive inaccuracies. Therefore, the actual clinical usefulness and impact should be investigated in a prospective clinical study. Moreover, these numbers and considerations emphasize that a clinical decision support tool in a precarious setting as preoperative counselling for DBS therapy should have a warning role, instead of a directive role. The goal should be to support the clinician with validated numerical expectations, and ensure her or his awareness in case of a patient with a higher than average chance on suboptimal therapeutic effect.</ns0:p></ns0:div> <ns0:div><ns0:head>The clinical value of predicting STN DBS motor response in the preoperative phase</ns0:head><ns0:p>Establishing an accurate prediction tool for motor outcome after STN DBS facilitates the clinician to improve patient counselling, expectation management, postoperative patient satisfaction, and potentially even patient selection. <ns0:ref type='bibr' target='#b16'>(Lin et al. 2019</ns0:ref>) Due to the complexity and heterogeneity of individual STN DBS candidates, outcome prediction needs to be accompanied by a clinical expert's appraisal. Moreover, the accuracy of a prediction model solely regarding clinical preoperative factors will always be limited due to the influence of surgical factors. Nevertheless, we intentionally chose to leave pre-, intra-and postoperative imaging and neurophysiology variables out of our model. This way, we ensure the model's accessibility and usability in clinical practice. We aim to provide the clinician during preoperative counselling with numerical support regarding the most probable motor outcome for an individual patient. Further it is important to underline this model's target patient population and clinical utilization. The model is designed for, and tested on, PD patients who were included for STN DBS implementation. This means the model should be applied to patients which are highly likely to be included for STN DBS implementation in the current care practices. In this population, the model is aimed to inform the clinician, and indirectly the patient, about a potential increased risk on a suboptimal motor response. This means the model is not developed to identify optimal STN DBS candidates from a general PD population.</ns0:p></ns0:div> <ns0:div><ns0:head>The additive value of machine learning methods for clinical decision support tools</ns0:head><ns0:p>The applied predictive multivariate logistic regression model was chosen to overcome limitations inherent to conventional (univariate) logistic regression models. <ns0:ref type='bibr' target='#b5'>(Daniels et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b8'>Frizon et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b25'>Schuepbach et al. 2019)</ns0:ref> Traditional predictive or correlative analyses mainly result in a correlation between one preoperative variable and a postoperative outcome, while controlling for several confounding preoperative variables. The absence of confounders and predictive variable selection in machine learning models, makes them less limited by sample size than traditional correlative analyses. <ns0:ref type='bibr' target='#b27'>(Wynants et al. 2015)</ns0:ref> The presented prediction model distinguishes itself by evaluating all available variables simultaneously. The applied cross-validation decreases the restriction due to sample size and leads to less a-priori selection-bias. Nevertheless, the advantages of machine learning predictive models come with specific analysing risks. For example, an external validation is required to evaluate under-or overfitting of the model, and validation of the threshold for accepting probabilities. Further, we stress the importance of using interpretable predictive machine learning models. In contrast to more complicated models such as deep neural networks, interpretable machine learning models remain explainable. This is essential in evaluating clinical validity and creating clinical confidence in a supportive decision tool which are both important in realizing actual clinical impact. <ns0:ref type='bibr' target='#b23'>(Rudin 2019)</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Interpretation of the preoperative predictive variables in this model</ns0:head><ns0:p>This overview of interpretable weight of each predictive variable is an advantage of the applied logistic regression in the prediction model (fig. <ns0:ref type='figure'>3C</ns0:ref>). This advantage enables clinicians to verify whether the ratio behind the predictions is clinically valid or whether predictions are based on unexpected variables. The reported large influence of higher age at PD onset on becoming a weak responder contradicts a finding of a meta-analysis that report younger age to be a positive predictor for a favourable outcome. Contrarily, the same meta-analysis reports longer PD duration as a predictor for favourable outcome. <ns0:ref type='bibr' target='#b13'>(Kleiner-Fisman et al. 2006)</ns0:ref> Preoperative UPDRS III and IV scores in the on-medication condition have the largest relative influence values for becoming a strong responder in this model. High preoperative motor severity increasing the chance to become a strong responders is in line with the findings of a meta-analysis, although most included studies describe preoperative severity in the offmedication condition. <ns0:ref type='bibr' target='#b13'>(Kleiner-Fisman et al. 2006</ns0:ref>) Evidence on the predictive value of symptom severity in on-medication condition is limited. The finding that H&amp;Y scores do not majorly influence outcome probabilities is in line with previous literature. This literature describes that disease severity positively influences the chance on strong motor response, while axial and balance problems negatively influence this chance. <ns0:ref type='bibr' target='#b13'>(Kleiner-Fisman et al. 2006</ns0:ref>) Since H&amp;Y severity is based on both these factors, an inconclusive effect is expected. Furthermore, there is literature on predictive or correlative variables and QoL outcome after STN DBS. We cannot compare these findings one-on-one with our findings. However, our holistic outcome classification aims to cover multiple aspects which influence QoL. High preoperative UPDRS III scores, and high UPDRS III levodopa response, are identified as important predictors of good QoL outcome, and motor outcome. <ns0:ref type='bibr' target='#b5'>(Daniels et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b8'>Frizon et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kleiner-Fisman et al. 2006)</ns0:ref> Conversely, recent studies have failed to replicate this positive predictive value of UPDRS III severity, or UPDRS III levodopa response on QoL outcome, or motor outcome. <ns0:ref type='bibr' target='#b0'>(Abboud et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b25'>Schuepbach et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b28'>Zaidel et al. 2010)</ns0:ref> Our findings are in line with some of these findings, since the absolute UPDRS III score showed a relevant influence, while the UPDRS III difference between on-vs. off-medication condition did not have a relevant influence. Regarding the reported influence of levodopa responsiveness, one should consider that LEDD is expressed in milligrams, which means that the relative influence of a unit increase (1 milligram) is not a clinically relevant increase. A high score on categorical Fluency is a small contributor to becoming a strong responder (fig. <ns0:ref type='figure'>3C</ns0:ref>). A high categorical Fluency score corresponds to better neuropsychological functioning. The contribution of the Stroop interference score is very small. Thus, there is no large influence of neuropsychological tests in our prediction model. Our lack of QoL scores prevented replication of previous findings which suggested that an impaired preoperative QoL-functionality predicts a large postoperative QoL improvement. <ns0:ref type='bibr' target='#b17'>(Liu et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b25'>Schuepbach et al. 2019</ns0:ref>) Likewise, the absence of a proper non-motor symptom scale hampered potential reproduction of the recently described importance of non-motor symptoms. <ns0:ref type='bibr' target='#b4'>(Dafsari et al. 2018)</ns0:ref> The reported influences of the preoperative variables on the outcome probability are mainly consistent with the literature, and are partly contradicting literature. We stress that the reported influences of this model cannot be seen outside the scope of this model. They are only reported to gain insight in the underlying weights which determine the probabilities. They cannot be interpreted on their own within individual patients when other variables in the model are disregarded.</ns0:p></ns0:div> <ns0:div><ns0:head>Limitations</ns0:head><ns0:p>Our study is limited by its retrospective character. Missing preoperative data points were overcome by imputations. Outcome values were not imputed to prevent training of the model based on imputed self-generated data. Even though the imputation method was sound, the imputed values will never reach true values and will influence outcomes. Second, the internal consensus on the applied categorization for motor outcome is based on scientific grounds, but can always be disputed. The holistic approach including UPDRS II, III and IV, aims to cover aspects of daily life activities, motor symptoms, and adverse effects of treatment. Future work should include QoL metrics and investigate the correlation between (QoL) and the presented classification. We argue our approach in the Supplementary Material. Lastly, the accuracy of a preoperative prediction model will always be limited and contain variance due to the lack of surgical variables.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>The presented prediction model identified strong vs. weak responders one-year after STN DBS for PD with a good classification accuracy. The potential distribution of predictive inaccuracies was in line with the aimed clinical utilization. These findings contribute to the proof-of-concept of machine learning prediction of individual motor outcome after STN DBS based on preoperative clinical variables. The reported preoperative variables cannot be interpreted separately outside the scope of this prediction model, but endorse the clinical reliability of the applied method. These results and considerations support the potential and the timely relevance of predictive clinical support tools for DBS outcome, and advocate further reproduction and validation in a representative, multicenter cohort. The optimal clinical utilization should be refined and the clinical additional value and impact should be clarified before a predictive clinical support tool can be applied during individual preoperative DBS counseling. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>Prediction model performance and importance per predictive variable A: Visualization of the performance of our prediction model. Our prediction model performs with an average area under the curve (AUC) of the receiver operating curve (ROC, blue line) of 0.78 (standard deviation: 0.08). All the dots on the ROC represent a threshold between 0 and 1 for accepting a probability to be a weak responder to be true. Every threshold leads to a different true positive rate and false positive rate. The red circle represents the threshold corresponding with pane B. The orange line represents chance level in which true positive rates equal true negative rates. B: Matrix of the example when 0.29 is chosen as a threshold for accepting the probability to be a weak responder (red circle in pane A). The true positive rate of 0.80 results in 24 out of 30 true weak responders getting a true weak prediction. The false positive rate of 0.24 results in 14 out of 59 true strong responders getting a false weak prediction. The classification accuracy is 0.78 with 69 out of 89 correct predicted patients.</ns0:p><ns0:p>C: Relative influence of all preoperative predictive variables. The blue bars represent the normalized Odds Ratios. The heights represent the effect on prediction outcome of a 1 unit increase in the specific variable, while all other variables stay equal. preoperative: off-medication, postoperative: off-medication and on-stimulation &#182; &#182;: &#8224;: percentage of Hoehn and Yahr scales are relative based on the number of available data (pre: n = 85, post: n = 71) *: significant difference with p-value &lt; 0.05, calculated with Mann Whitney-U test &#8224;: on-stimulation, on-medication at one-year follow up &#182;: difference between on-medication and on-stimulation vs. preoperative on-medication &#8225;: on-stimulation and off-medication at one-year follow up &#177;: difference between on-stimulation and off-medication vs. preoperative off-medication *: significant difference with p-value &lt; 0.05, calculated with Mann Whitney-U test</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>AUC: area under the curve, DBS: deep brain stimulation, H&amp;Y: Hoehn &amp; Yahr scale, LEDD: levodopa equivalent daily dosage, Levodopa response: difference between UPDRS III off-medication minus UPDRS III on-medication; off: off-medication, on: on-medication, ROC: receiver operate characteristic, TEED: total electrical energy delivered, UPDRS: Unified Parkinson Disease Rating Scale, PD: Parkinson's disease</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,229.87,525.00,322.50' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 . Preoperative and postoperative variables of total population</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>H&amp;Y: Hoehn &amp; Yahr scale, LEDD: levodopa equivalent daily dosage, off-/on-med: off-/onmedication, off-/on-stim: off-/on-stimulation, TEED: total electrical energy delivered, UPDRS: Unified Parkinson Disease Rating Scale $: values are given as mean and standard deviation of the mean.preoperative: on-medication, postoperative: on-medication and on-stimulation &#182;:</ns0:figDesc><ns0:table><ns0:row><ns0:cell>UPDRS II &#182;</ns0:cell><ns0:cell>09.8 (6.6)</ns0:cell><ns0:cell>09.6 (5.5)</ns0:cell></ns0:row><ns0:row><ns0:cell>UPDRS III &#182;</ns0:cell><ns0:cell>21.9 (12.5)</ns0:cell><ns0:cell>16.4 (9.9) *</ns0:cell></ns0:row><ns0:row><ns0:cell>UPDRS III &#182; &#182;</ns0:cell><ns0:cell>39.1 (13.1)</ns0:cell><ns0:cell>16.4 (9.9) *</ns0:cell></ns0:row><ns0:row><ns0:cell>UPDRS IV &#182;</ns0:cell><ns0:cell>05.5 (4.0)</ns0:cell><ns0:cell>02.8 (2.4) *</ns0:cell></ns0:row><ns0:row><ns0:cell>H&amp;Y 1 &#182;</ns0:cell><ns0:cell>02 (02%) &#8224;</ns0:cell><ns0:cell>04 (03%)</ns0:cell></ns0:row><ns0:row><ns0:cell>H&amp;Y 1.5 &#182;</ns0:cell><ns0:cell>02 (02%)</ns0:cell><ns0:cell>01 (01%)</ns0:cell></ns0:row><ns0:row><ns0:cell>H&amp;Y 2 &#182; H&amp;Y 2.5 &#182; H&amp;Y 3 &#182; H&amp;Y 4 &#182; H&amp;Y 5 &#182;</ns0:cell><ns0:cell>13 (15%) 34 (40%) 25 (29%) 09 (11%) -</ns0:cell><ns0:cell>21 (30%) 24 (34%) 19 (27%) 02 (03%) -</ns0:cell></ns0:row><ns0:row><ns0:cell>Fluency total</ns0:cell><ns0:cell>39.7 (9.4)</ns0:cell><ns0:cell>33.6 (9.8) *</ns0:cell></ns0:row><ns0:row><ns0:cell>categories &#182;</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Fluency total</ns0:cell><ns0:cell>35.5 (10.8)</ns0:cell><ns0:cell>33.6 (11.9) *</ns0:cell></ns0:row><ns0:row><ns0:cell>letters &#182;</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Stroop</ns0:cell><ns0:cell>56.1 (35.1)</ns0:cell><ns0:cell>76.7 (63.1) *</ns0:cell></ns0:row><ns0:row><ns0:cell>interference &#182;</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>LEDD (milligrams)</ns0:cell><ns0:cell>1187 (619)</ns0:cell><ns0:cell>656 (510) *</ns0:cell></ns0:row><ns0:row><ns0:cell>TEED</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>134 (130)</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 . Comparison of postoperative variables in groups with strong responders and weak responders</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>LEDD: levodopa equivalent daily dosage, off-/on-med: off-/on-medication, off-/on-stim: off-/on-stimulation, TEED: total electrical energy delivered, UPDRS: Unified Parkinson Disease Rating Scale $: mean (standard deviation)</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:03:46750:1:2:NEW 5 Aug 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Maastricht, July 6, 2020 Dear professor Black, We thank the two reviewers and you for the time and effort leading to the high-quality review of our work, and the possibility to improve our work. In general, we are very happy about the favourable comments by both reviewers and especially thank reviewer 2 for his kind praise, describing the study as “excellent”. We could use most of the received comments to improve the quality of our manuscript. In the attached pages, we respond point-wise on all received comments, and elaborate where we adjusted the manuscript if required. Two main points require specific highlighting here: First, we want to mention a comment on our outcome-classification method. We are very thankful to Reviewer 1 for the critical appraisal of how we classified strong motor responders based on improvement on UPDRS II, III or IV. We did adjust our classification following the reviewer’s suggestion and re-classified 3 people as bad responders. One patient got excluded because of a converted diagnosis. One comment we did not integrate in our manuscript is the suggestion for an external validation cohort. We fully agree with reviewer 1 that external validation of the prediction model is the designated next step in the validation process. However, we see this proof-of-concept paper as an essential step prior to the collection of a large multicenter cohort of PD patients treated with STN DBS. By showing a proof-of-concept of the development of a machine learning prediction model for DBS outcome, including discussing the essential limitations and trade-offs to consider, we want to pave the way for the next step in assessing whether this approach can lead to a clinical useful and relevant decision support tool. Again, we are grateful with the qualitative review that you offered us, and are looking forward to your and the reviewers’ opinion on our argumentation. Best regards, on behalf of all the authors, Jeroen Habets, Christian Herff, Marcus Janssen Attachment: Pointwise response on reviewer comments Reviewer 1 Basic reporting 176 ‘can be interpret easily’: typo. Corrected. Figure 3B: please use different colours (lowerright corner is unreadable). We changed the colour and improved the readability. Table 1: legend is incomplete. Adjsuted. Experimental design See below. Validity of the findings See below. Comments for the Author Habets and colleagues describe an interesting study on a machine learning based prediction model to model success after DBS surgery (or rather: failure of DBS surgery). Their use of an advanced modelling approach is commended, as well as their use of preoperative clinical features to enhance the generalizability. The authors rightly state that this model is a proof-of-concept and should not be interpreted as a ‘final model’ for clinical practice. There are however, several limitations with regard to insufficient information on variable-importance, a small sample size which is (incorrectly) stated to be overcome through cross-validation, and lack of external validation whereas more cohorts with these data are available. We thank the reviewer for the kind and correct comments, we will discuss how we improved the paper based on these comments in the following points. 73. “… limited or no improvement of motor disability”: do you intend motor fluctuations or severity of motor symptoms? We reformulated to ‘motor fluctuations and quality of life according to the referenced paper’. (line 91 – 92) 100-102 “To add value… clinical decisions”: you correctly state that challenging cases should be improved rather than valid decisions. However, given that you intend to develop a generic model the case-mix you require should resemble the case-mix from the actual population in order to prevent overfitting / underfitting. We strived to collect a data set which is as representative as possible for the intended use of the prediction model as a decision tool. Therefore, we decided to include patients which are selected to undergo STN DBS implementation, and we test whether our prediction model could identify the ‘weak-responding’ patients within this group. Since the current clinical practice results in an unbalanced case-mix of eventually weak and strong responders after STN DBS-implementation, it is natural that our population consists of more strong responders. This means that the majority of patients is correctly included for STN DBS in the current practice, without a decision tool. To avoid creating uncertainty for those cases, we aim to utilize the suggested model with a high negative predictive value (negative is strong response). This is also addressed in the adjusted lines 275 – 285. 111-113 “We included all… on-stimulation condition”. How many patients were initially considered (i.e. underwent STN DBS in the given time period) and how many patients were excluded for the mentioned reasons? We added the total amount of considered patients 127 patients. (line 132) 114-115: Was this study approved in the absence of informed consent, or was formal evaluation waived given the retrospective nature? The authorized medical ethical committee waived the requirement for explicit informed consent based on a formal evaluation of the study protocol. Two main reasons were the retrospective nature, and the ‘no-objection statement’ these patients gave during their outpatient clinical visits. 117-123: Please provide a list with the exact tests that were incorporated in the model (‘categorical, and letter tests’ are relatively vaguely descripted). We deleted a comma, to indicate that the ‘categorical and letter tests’ are two tests within the Verbal Fluency. (lines 144 – 145) 150-151: The reason for labelling a patient as a ‘strong responder’ in case of MCID improvement in any of the three domains can be understood, however if a patient improves in one domain and deteriorates in the two others this label may be subject for debate. As mentioned in the beginning, this comment improved our methodology substantial. We reviewed all Strong Responders, and found that 3 people who improved on UPDRS II or IV, showed a clinically relevant deterioration on UPDRS III, which was confirmed during medical charts review. Based on these findings, we redefined these patients as weak responders and redefined our outcome classification to make it robust for similar cases. We defined an allowed non-clinically relevant UPDRS III deterioration for patient who clinically relevant improved on UPDRS II and/or IV, also keeping the natural disease progression in mind. We elaborate on this revised outcome classification from line 171. Three patients improved on UPDRS III or IV, but deteriorated on UPDRS II. Review of the medical charts by a neurologist confirmed the status as strong responders. One patient’s diagnosis was found to be converted to multiple systemic atrophy and was excluded. These adjustments resulted in a distribution of 59 strong responders and 30 weak responders. In the revision of the analysis we conducted a 5-fold cross-validation to prevent unbalanced folds. The revised analysis did not change our results significantly and we adjusted all findings throughout the manuscript accordingly. 172-173: Using this approach (i.e. cross-validation) the test-set remains relatively small and you run the risk of testing on outliers. I agree with the merits of this approach (definitely better than split-sample validation) but the limitations should be discussed as well. We agree with the reviewer’s comments on the applied validation method. We elaborated more on the limitations of the cross-validation as most suitable approach, in lines 208 – 212. 173: ‘predictive variables or confounders’: please remove the word ‘confounders’, there are no confounders in prediction modelling (as opposed to causal research). Adjusted. 176 ‘can be interpret easily’: typo. Adjusted. (line 213) 221 “63 out of 90 patients were categorized as strong responders”. See earlier comment, please consider at least reporting the MCID improvements per domain separately as well rather than joined together. We refer to figure 2 where we visualize the distribution of improvements on all the different UPDRS scales. From the total of 60 strong responders, e.g. 11 patients show the MCID improvement on all 3 domains, 3 showed MCID improvement on UPDRS II and IV, and so on. We hope this yields a satisfying answer on the reviewer’s comment. 231 Why choose 0.24 as a threshold? Even if the argumentation holds true, are you confident that this threshold would hold up in a prospective study of large sample size and therefore should be utilized in clinical practice? 0.24 was chosen as a threshold based to realize a high negative predictive value and a high sensitivity. Due to the revised outcome classification the exact numbers in the results slightly changed, and we chose the threshold again to ensure a high negative predictive value and a high sensitivity. This led to a threshold for probability acceptance of 0.29. Prospective studies have to investigate whether this high negative predictive value and sensitivity holds. If prospective research shows that the clinically ‘straight-forward’ do not suffer from false positive predictions, and a sufficient amount on eventual weak responders gets a preoperative warning, the model might be used in clinical practice. We elaborated on these thoughts in lines 275 – 290. Discussion: I do find the discussion to be somewhat overoptimistic of the actual performance. It states that in 78% of cases the predicted class would be correct, however there are more outcomes relevant to PD care than just the UPDRS-classes (e.g. quality-of-life, satisfaction with surgery etc.). Moreover, the sample-size is by far too small to make any definitive claims on its utility. Although the authors have stressed this in the final paragraphs of the discussion, I would appreciate some more elaboration on the 'next required steps' before this research can proceed to clinical implementation. We agree with the reviewer that the results should be presented more modestly. We rephrased several sentences in the discussion to find the right tone for our message. In lines 327 – 337 we elaborated on next required steps. 279-285: see previous comments: there is no such thing as confounding variables in prediction modelling. We agree with the reviewer. Here, ‘confounding’ is referring to the way variables in traditional correlative analyses are regarded. We rephrased lines 522 – 524 to makes this more clear. 286-289: with such a small sample size: overfitting is still an issue. I do agree that cross-validation circumvents this issue, but definitely not entirely and the statement that cross-validation ‘decreases restriction due to sample-size’ is untrue. Moreover, cross-validation certainly does not prevent selection-bias as this is inherent to your study-design. An option is to provide ROC curves per cross-validation run to assess the dispersity of results (although this may not be very visually attractive). Why was external validation not performed? There are a lot more similar cohorts available with which to increase your sample size and overcome several of the limitations mentioned in your discussion. We thank the reviewer for this justly critique and suggestion for improvement. As mentioned in the discussion and underlined by our next study protocol registration (NCT NCT04093908), we see external validation as the next step. To enquire and collect a large enough multicenter cohort of PD patients who underwent STN DBS, we need to show this proof-of-concept. We agree with the reviewer that visualizing the different ROC’s is not contributing to the understandability of performance or nuances of our model. We elaborated more on the limitations of this first proof-of-concept study and it’s cross-validation in the methods section, and address the limitations and nuances as well in the adjusted lines 331 – 337, and 356 – 361. A rather large limitation of this study is that this model is based only on those patients that have been selected for surgery. Cases on which there is ‘preoperative’ doubt may or may not have been rejected for surgery and therefore there is a large likelihood of those patients being underrepresented in the training-set, thereby reducing its validity in a clinical setting. Parallel to the previous comment, we recognize the line of thinking of the reviewer from our own thoughts during experimental design and interpretation. We assume that including not-operated patients will provide answers to a different research question, since there are per definition no postoperative scores to label the patients as weak or strong. We agree with the reviewer that this nuance is essential for a proper interpretation of our findings, and made the target population clearer in lines 508 – 514. A prospective validation in a later stage could collect clinicians doubts and considerations during surgical counselling. These data would contribute to insights and potentially enable the development of a variation on this model, applicable on a more general PD patient population which is considered for DBS therapy. We emphasis this last argument in lines 332 – 361 Figure 3B: please use different colours (lowerright corner is unreadable). Adjusted. Figure 3C: please provide exact numbers rather than visualization. Since we use normalized odds ratio, we only can provide the displayed relative influence percentages. For clarification, we added the direction of attribution to weak and strong response. Table 1: legend is incomplete. Adjusted. Reviewer 2 (Anonymous) Basic reporting The basic reporting is excellent: The submission is ‘self-contained,’ represents an appropriate ‘unit of publication’, and includes all results relevant to the hypothesis. Coherent bodies of work are not inappropriately subdivided merely to increase publication count. Experimental design The experimental design is conceptualised and conducted very well: This is original primary research within the Aims and Scope of the journal. The Research question is clear, relevant, timely and meaningful. It is stated how research fills an identified knowledge gap. The investigation is conducted rigorously and to a high technical standard. The research is conducted in conformity with the prevailing ethical standards in the field. Methods described with great detail. The raw data is provided to replicate the results. Validity of the findings The validity of the findings is excellent: The raw data is provided to replicate the results. The benefit to literature is clearly stated. All underlying data have been provided; they are robust, statistically sound, and controlled. Conclusions are well stated, linked to original research question & limited to supporting results. Comments for the Author In this is a proof-of-principle study in 90 patients with Parkinson’s disease Habets et al. implement a machine learning logistic regression prediction model of “weak” and “strong” response to bilateral subthalamic stimulation. The model input parameters include a wide range of preoperative variables. Not abiding to the “rule-of-ten” is feasible because the authors use this machine learning prediction with a 10-fold cross-validation. The model predicts weak responders with a good C-statistic (0.88) and has 78% diagnostic accuracy. This is a timely study by an experienced group, the statistical approach is sound and the primary research question of preoperative prediction of postoperative motor DBS outcomes is of key clinical importance. Overall, the authors have done an excellent job in explaining the advantages of the implemented machine learning logistic regression prediction model as opposed to the traditional univariate approach implement in a number of earlier studies. We thank the reviewer for these kind and very positive comments. In the following points we will elaborate how we used the comments to improve our paper. Minor points: - In 4 patients the motor improvement was predicted to be strong but was observed in truth to be weak. It is understandable to include only preoperative clinical parameters for the model. However, the authors may add that STN targeting was confirmed (e.g. based on postoperative imaging or intraoperative electrophysiology – at least in these 4 patients) so that it is clear that there were no lead misplacements which could account for false predictions in these 4 patients. We agree on the importance of ruling out surgical factors which could have explained a weak response. We added this in lines 275 – 277. - The preoperative motor response was quantified by subtracting UPDRS-III ON – UPDRS-III OFF. Other studies have often used a ratio and reported the % improvement. Could the authors explain why they preferred a subtraction? We have chosen for the absolute difference since this variable is translatable over patients. Especially since we evaluate 3 different UPDRS scales, low baseline values are possible over the patient population. In a patient with a low baseline score, a small improvement would lead to a very high percentage-improvement. Therefore, we preferred the absolute values to give a better comparison over the group. - Line 329 and 330: the term neuropsychological functioning is very broad and may be specified (executive functions). We added additional information about the applied neuropsychological tests in lines 144 – 145. - Line 332: higher preoperative QoL-scores predict larger QoL postoperative improvement We stated the sentence clearer according to the two references mentioned (Liu 2018, Schuepbach 2019), line 644. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>When people are confronted with health proposals during the coronavirus disease 2019 (COVID-19) pandemic, it has been suggested that fear of COVID-19 can serve protective functions and ensure public health compliance. However, health proposal repetition and its perceived efficacy also influence the behavior intention toward the proposal, which has not yet been confirmed in the COVID-19 context. The present study aims to examine whether the extended parallel process model (EPPM) can be generalized to a naturalistic context like the COVID-19 pandemic. Additionally, we will explore how repetition of a health proposal is involved with the EPPM. In this study, two groups of participants are exposed to the same health proposal related to COVID-19, where one group is exposed once and another group twice. They then fill out a questionnaire consisting of items concerning behavior intention and adapted from the Risk Behavior Diagnosis Scale. Structural equation modeling will be used to determine the multivariate associations between the variables. We predict that repetition of the health proposal will associate with response efficacy (i.e., a belief about the effectiveness of the health proposal in deterring the threat) and perceived susceptibility (i.e., a belief about the risk of experiencing the threat). It is also predicted that following the EPPM, behavior intention will associate with both perceived efficacy of the health proposal, which underlies response efficacy, and perceived threat of COVID-19, which underlies perceived susceptibility. We will discuss the process, based on the model, where health message repetition affects behavior intention during the COVID-19 pandemic.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>1) What is the main question being addressed in your study?</ns0:head><ns0:p>At the time of writing, it has been seven months since the outbreak of the COVID-19 pandemic. During this period, while a heavy loss of both life and economy has been caused, some countries and regions have achieved staged success in the fight against this disease.</ns0:p><ns0:p>There is no doubt that public compliance with effective health proposals plays a crucial role in achieving this success.</ns0:p><ns0:p>It is suggested that functional fear, which is explained as one of the negative emotions serving protective functions in certain contexts, has promoted public health compliance during the COVID-19 pandemic <ns0:ref type='bibr' target='#b10'>(Harper et al., 2020)</ns0:ref>. Nevertheless, this needs to be explored further, especially considering previous studies on health communication and the features of information dissemination in real life. The main question being addressed in our study is to examine how people's health compliance intention is influenced by various factors in the COVID-19 context in an exhaustive way.</ns0:p><ns0:p>It is known that fear has an impact on health compliance. In fact, fear appealing communication is considered an effective way to promote health campaigns and has been widely investigated for promoting health awareness related to various topics, including smoking <ns0:ref type='bibr' target='#b18'>(Leventhal &amp; Watts, 1966)</ns0:ref>, alcohol use <ns0:ref type='bibr' target='#b34'>(Wolburg, 2001;</ns0:ref><ns0:ref type='bibr' target='#b22'>Moscato et al., 2001)</ns0:ref>, AIDS <ns0:ref type='bibr' target='#b30'>(Treise &amp; Weigold, 2001)</ns0:ref> and so forth. After several initial studies, inconsistent results indicated that a simple monotonic function of fear may not be expected in persuasive health communication. Specifically, despite a large number of studies indicating the positive main effect of fear on persuasion <ns0:ref type='bibr' target='#b18'>(Leventhal &amp; Watts, 1966;</ns0:ref><ns0:ref type='bibr' target='#b7'>Dabbs &amp; Leventhal, 1966;</ns0:ref><ns0:ref type='bibr'>Leventhal &amp; Singer, 1966)</ns0:ref>, relatively few report the negative main effect of fear <ns0:ref type='bibr' target='#b12'>(Janis &amp; Feshbach, 1953;</ns0:ref><ns0:ref type='bibr' target='#b13'>Janis &amp; Terwilliger, 1962)</ns0:ref>. Later research confirmed that the effects of fear appeal interact with various source variables, message variables, and receiver variables, and thus cannot be described easily <ns0:ref type='bibr' target='#b21'>(Miller &amp; Hewgill, 1966)</ns0:ref>.</ns0:p><ns0:p>One of the most recent and prevalent theories on fear appealing communication is the extended parallel process model (EPPM: <ns0:ref type='bibr' target='#b31'>Witte, 1992</ns0:ref><ns0:ref type='bibr' target='#b32'>Witte, , 1994))</ns0:ref>, which is based on former frameworks including the Parallel Response Model <ns0:ref type='bibr' target='#b19'>(Leventhal, 1970)</ns0:ref> and protection motivation theory <ns0:ref type='bibr' target='#b25'>(Rogers, 1975)</ns0:ref>. In the EPPM, there are four main factors, which influence the prediction of certain communication outcomes: perceived susceptibility and severity composing perceived threat, self-efficacy and response efficacy composing perceived efficacy. Perceived susceptibility refers to a belief about the risk of experiencing a threat, whereas severity refers to a belief about the magnitude of the threat. On the other hand, selfefficacy is defined as a belief about the ability to perform a recommended proposal to avert the threat; response efficacy is a belief about the effectiveness of the recommended proposal in deterring the threat <ns0:ref type='bibr' target='#b33'>(Witte, 1996)</ns0:ref>. Concretely, when both perceived threat and perceived efficacy are high, people are most likely to engage in a danger control process, which means conforming to the recommended health proposal. Nevertheless, when perceived threat and perceived efficacy are high and low, respectively, people turn to a fear control process leading to coping responses that reduce fear and danger control responses. Meta-analyses on the results of fear appeal research confirm the validity of the EPPM <ns0:ref type='bibr' target='#b8'>(Floyd, Prentice-Dunn &amp; Rogers, 2000;</ns0:ref><ns0:ref type='bibr'>Peter, Ruiter &amp; Kok, 2013)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Currently, we are exposed to health proposals concerning COVID-19, in all likelihood, more than once in our daily lives, which makes message repetition an essential factor to be taken into consideration. Although research concerning stimulus repetition first emerged in the 1960s <ns0:ref type='bibr' target='#b35'>(Zajonc, 1968)</ns0:ref>, there was no study on message repetition in persuasive communication until ten years later. It is suggested that under moderate repetition (less than 3 times), agreement toward persuasive messages rise <ns0:ref type='bibr' target='#b2'>(Cacioppo &amp; Petty, 1979)</ns0:ref>. The rationale is explained as follows: Scrutiny is reinforced through a moderate level of message repetition, which enhances the understanding of message content and the merits advocated by it, consequently improving supportive attitudes toward the message. Later research revealed that the mechanism mentioned above is applicable only when arguments in the message are perceived as strong, and when the issue is of high personal relevance <ns0:ref type='bibr' target='#b3'>(Cacioppo &amp; Petty, 1989;</ns0:ref><ns0:ref type='bibr' target='#b5'>Claypool et al., 2004)</ns0:ref>.</ns0:p><ns0:p>In health communication on COVID-19, we assume that the interpretation of the content of a certain health proposal message may be related to factors in the EPPM.</ns0:p><ns0:p>Concretely, response efficacy and perceived susceptibility may be perceived when reading a health proposal message. If the content of a health proposal is supported, high response efficacy will be found, resulting in a decrease in perceived susceptibility. Given the connection between the EPPM and message repetition in persuasive health communication and the actual state we have been through when confronted with health proposals during the COVID-19 pandemic, we built an integrative model to investigate the factors and their associations concerning an individual's behavior intention to conduct effective health proposals to prevent the infection. In the model, we will first focus on the influence of message repetition on the response efficacy of a certain health proposal and the perceived susceptibility of COVID-19. We will then elaborate on the change in behavior intention toward the proposal due to the variation in perceived efficacy and perceived threat.</ns0:p><ns0:p>Furthermore, we will confirm whether the underlying roles of perceived efficacy and perceived threat hold true in the COVID-19 pandemic. The present study is unique and necessary in several aspects.</ns0:p><ns0:p>One, considering that COVID-19 is a real-life and ongoing public health emergency, some of its properties are fixed and thus cannot be manipulated. Take perceived threat for example: In previous research, it was always considered together with fear because without extra fear-arousing materials (e.g., explanation of a certain disease in text or video), the perceived threat may not be notable enough to act as an independent variable. However, the COVID-19 situation is different from all the previous topics. With the World Health Organization (WHO) declaring COVID-19 a pandemic on March 11, 2020, it has now affected over 200 countries and caused changes to most people's lives. Unlike other unfamiliar diseases (e.g., melanoma), even though the degree to which one is influenced by the pandemic may vary, it is easy to understand the health threat of COVID-19 without extra explanation. Considering that fear-arousing information is no longer needed in the message, we are interested in exploring how perceived threat alone for the COVID-19 affects behavior intention when there is no intentional manipulation of fear arousal. Two, there is still no research to test a model that combines the EPPM with message repetition. Although studies on similar topics have been conducted, the results have not been analyzed in an integrated way <ns0:ref type='bibr' target='#b29'>(Skilbeck, Tulips &amp; Ley, 1977;</ns0:ref><ns0:ref type='bibr' target='#b30'>Treise &amp; Weigold, 2001;</ns0:ref><ns0:ref type='bibr' target='#b28'>Shi &amp;</ns0:ref> PeerJ reviewing <ns0:ref type='table'>PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p><ns0:p>Smith, 2016). To be specific, when supportive arguments toward the content of a recommended proposal are enhanced due to moderate repetition, it can also be interpreted as the change in response efficacy and perceived susceptibility in the EPPM. Nevertheless, this connection between two research topics has rarely been made clear, causing difficulty in making an exact prediction of the results.</ns0:p><ns0:p>Three, it remains essential to confirm the validity of EPPM's established construction, namely the sub-dimensions of perceived efficacy and perceived threat, to enhance compliance with public health guidelines in the COVID-19 pandemic. We assume better health compliance is due to higher perceived efficacy, but what does perceived efficacy mean? By learning more about its two indicators, self-efficacy and response efficacy, we can better understand why people choose to conform or not to certain health proposals. For instance, even though social distancing is considered efficient in preventing infection (high response efficacy), the difficulty in conducting it may vary between people based on their socializing needs (divergent self-efficacy), which results in different levels of health compliance.</ns0:p><ns0:p>Similarly, if we are aware that perceived threat is indicated by perceived susceptibility and severity, we may know which properties to emphasize in education on COVID-19 to boost public health compliance.</ns0:p><ns0:p>In summary, the present study is of comprehensive significance in revealing the mechanism in the purview of public compliance with effective health proposals in the COVID-19 pandemic.</ns0:p><ns0:p>2) Describe the key independent and dependent variable(s), specifying how they will be measured.</ns0:p><ns0:p>We will conduct a two-wave survey for two groups of participants: One is a norepetition group and the other is a repetition group. In the first wave, two groups will first answer the same dummy questionnaire of the Need for Cognition Scale <ns0:ref type='bibr' target='#b15'>(Kouyama &amp; Fujihara, 1991)</ns0:ref>. At the end of the questionnaire, only the repetition group will be exposed to the target health proposal message, which is written in Japanese as:</ns0:p><ns0:formula xml:id='formula_0'>&#12450;&#12523;&#12467;&#12540;&#12523;&#12399;&#28040;&#27602;&#12539;&#27578;&#33740;&#21177;&#26524;&#12364;&#12354;&#12427;&#12392;&#35328;&#12431;&#12428;&#12390;&#12356;&#12414;&#12377;&#12290;&#19968;&#26041;&#12289;&#23569;&#37327;&#12398;&#12450;&#12523;&#12467;&#12540;&#12523;&#28040; &#27602;&#28082;&#12434;&#25163;&#12395;&#21462;&#12387;&#12383;&#22580;&#21512;&#12289;&#28040;&#27602;&#21177;&#26524;&#12364;&#20302;&#19979;&#12375;&#12390;&#12375;&#12414;&#12356;&#12414;&#12377;&#12290;&#12381;&#12371;&#12391;&#12289;&#12450;&#12523;&#12467;&#12540;&#12523;&#28040;&#27602; &#28082;&#12391;&#25163;&#12434;&#28040;&#27602;&#12377;&#12427;&#38555;&#12399;&#12289;&#23569;&#12394;&#12367;&#12392;&#12418;&#12509;&#12531;&#12503;&#37096;&#20998;&#12434;&#19979;&#12414;&#12391;&#12422;&#12387;&#12367;&#12426;&#25276;&#12375;&#12390;&#12289;&#12383;&#12387;&#12407;&#12426; &#25163;&#12395;&#21462;&#12387;&#12390;&#20351;&#12358;&#12371;&#12392;&#12434;&#12362;&#12377;&#12377;&#12417;&#12375;&#12414;&#12377;&#12290;</ns0:formula><ns0:p>The English translation of the message is as follows: 'While alcohol-based hand sanitizer is useful for preventing COVID-19 infection, the effect will be discounted if not enough amount is used. Therefore, it is recommended to press the pump to the bottom every time to get enough amount of hand sanitizer.' We selected this health proposal message according to a pilot study to examine how much the message is known and agreed on. As per the results, the knowledge rate of the message was low (40.8%). The attitudes to the message were measured using a 7-point scale (ranging from 1 = strongly disagree and 7 = strongly agree). The attitude to the message score was considerably favorable (i.e., the scores were significantly higher than 4 on the scale, M = 5.35, SD = 1.29, t (200) = 5.52, p &lt; .001, PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Cohen's dz = 1.047). Considering the definition by <ns0:ref type='bibr' target='#b3'>Cacioppo &amp; Petty (1989)</ns0:ref> that a strong argument refers to the one toward which favorable thoughts are generated predominantly, the above message is appropriate in the present study in two ways. One is that as mentioned above (i.e., 1. What is the main question being addressed in your study?). The supportive arguments on a message can be improved through moderate repetition only when the message is believed to be a strong one; therefore, our manipulation on message repetition will be meaningful. The other reason is that since the scores on favorable thought are moderately high, there is still space for supportive arguments to increase, preventing ceiling effect.</ns0:p><ns0:p>The second wave will be conducted 24-72 hours after the first wave. The interval range is set because we cannot control the precise timepoint at which participants answer the second questionnaire even if we invite them to do it on time. In the second wave, both groups of participants will be exposed to the same message, which is identical to the message shown to the repetition group in the first wave. They will then fill out the same questionnaire containing 13 items (Table <ns0:ref type='table'>1</ns0:ref>). Twelve items are adapted from the Risk Behavior Diagnosis Scale <ns0:ref type='bibr' target='#b33'>(Witte, 1996)</ns0:ref>, and 1 item enquires behavior intention toward the target health proposal.</ns0:p><ns0:p>All items will be scored on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). In the preliminary experiment, which had the same design as the main experiment (more details in Supplementary 1), we checked the convergent validity and discriminant validity for items 1-12. The results of our calculations are as follows: (1) average variance extracted (AVE) &gt; .5; and (2) square root of AVE &gt; inter-construct correlations are both met, confirming the validity of the items <ns0:ref type='bibr' target='#b9'>(Hair et al., 2010)</ns0:ref>. The scores of the 13 items in Table <ns0:ref type='table'>1</ns0:ref> will be treated as the key dependent variables.</ns0:p><ns0:p>We aim to test the model (Figure <ns0:ref type='figure'>1</ns0:ref>) which combines the EPPM with message repetition, further exploring the factors that associate with behavior intention. As described above, there will be two conditions with various frequencies of exposure to the message. Each participant will be assigned to one of the conditions. Thus, repetition, as a binary variable in the model, will be adopted as the key independent variable.</ns0:p></ns0:div> <ns0:div><ns0:head>3) What are your hypotheses?</ns0:head><ns0:p>The hypotheses of this research are:</ns0:p><ns0:p>&#61599; Hypothesis 1: Perceived efficacy has a positive effect on self-efficacy. H0: Perceived efficacy has no effect on self-efficacy.</ns0:p><ns0:p>&#61599; Hypothesis 2: Perceived efficacy has a positive effect on response efficacy. H0: Perceived efficacy has no effect on response efficacy.</ns0:p><ns0:p>&#61599; Hypothesis 3: Perceived threat has a positive effect on perceived susceptibility. H0: Perceived threat has no effect on perceived susceptibility.</ns0:p><ns0:p>&#61599; Hypothesis 4: Perceived threat has a positive effect on severity. H0: Perceived threat has no effect on severity.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>&#8226; Hypothesis 5: Repetition has a positive effect on response efficacy. H0: Repetition has no effect on response efficacy.</ns0:p><ns0:p>&#8226; Hypothesis 6a: Repetition has a positive effect on perceived susceptibility.</ns0:p><ns0:p>H6b: Repetition has a negative effect on perceived susceptibility. H0: Repetition has no effect on perceived susceptibility.</ns0:p><ns0:p>&#8226; Hypothesis 7: Perceived efficacy has a positive effect on behavior intention. H0: Perceived efficacy has no effect on behavior intention.</ns0:p><ns0:p>&#8226; Hypothesis 8a: Perceived threat has a positive effect on behavior intention.</ns0:p><ns0:p>H8b: Perceived threat has a negative effect on behavior intention. H0: Perceived threat has no effect on behavior intention.</ns0:p></ns0:div> <ns0:div><ns0:head>4) How many and which conditions will participants/samples be assigned to?</ns0:head><ns0:p>As mentioned in Section 5), our planned maximum sample size is N = 602. There will be two conditions (i.e., being exposed to the message once meaning no-repetition condition and being exposed to the message twice meaning repetition condition). We will recruit 301 participants for each condition via Yahoo! Crowdsourcing (http://crowdsourcing.yahoo.co.jp/).</ns0:p></ns0:div> <ns0:div><ns0:head>5) How many observations will be collected and what rule will you use to terminate data collection?</ns0:head><ns0:p>We will perform structural equation modeling (SEM) with the data acquired from participants in the two conditions in the second wave. Based on a preliminary experiment, the minimum sample size for RMSEA-based SEM in the present experiment was found to be N = 301 in total, using the findRMSEAsamplesize function in R <ns0:ref type='bibr' target='#b20'>(MacCallum, Browne &amp; Sugawara, 1996)</ns0:ref>, with &#945; = .05, power = .95, rmsea0 = .05, rmseaA = .01, df = 70. Given there are two conditions in the present study, we set N = 151 as the required sample size for each condition.</ns0:p><ns0:p>We doubled the required sample size as the maximum sample size (i.e., N = 602 in total and 301 per condition) because we have to plan to collect data from a larger sample for the following two reasons: (a) A large amount of data may have to be excluded based on the criteria detailed below (i.e., 7. What are your data exclusion criteria?). b) The dropout rate from the second wave may be high because it is hard to ensure all the participants in the first wave will take part in the second wave voluntarily within the requested time range, which is 24-72 hours after the first wave.</ns0:p><ns0:p>If the collected data does not reach the required sample size (i.e., N = 301 in total and 151 per condition) after excluding the data meeting the criteria in Section 7, we will collect PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed data from additional participants to reach the required sample size. However, if the number of participants exceeds 602, we will choose the data of the first 602 participants based on the time stamp and use those for analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>6) What are your study inclusion criteria?</ns0:head><ns0:p>Participants will be recruited via Yahoo! Crowdsourcing (http://crowdsourcing.yahoo.co.jp/) and should possess Japanese nationality. We will indicate this criterion to potential participants in the instruction and invite them to participate only when they fulfill this criterion. Besides, a question on nationality will be asked before the main questionnaire (i.e., the items of the Risk Behavior Diagnosis Scale and behavior intention toward the target health proposal).</ns0:p></ns0:div> <ns0:div><ns0:head>7) What are your data exclusion criteria?</ns0:head><ns0:p>We will apply six criteria to perform data exclusion. 1) To identify distracted respondents or satisficers <ns0:ref type='bibr' target='#b4'>(Chandler et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b23'>Oppenheimer et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b26'>Sasaki &amp; Yamada, 2019)</ns0:ref>, we will insert a simple question as an attention check question (ACQ) in the middle of the questionnaire in the second wave. The reason why we do not add an ACQ in the first wave is that the first wave is conducted only to control the frequency of exposure to the message, and thus the data in the first wave will not be analyzed. The ACQ in the second wave will be: Please choose number '2' from below.</ns0:p><ns0:p>Consistent with other items, the ACQ will also use a 7-point scale. The data from participants who choose 1 or 3-7 will be excluded.</ns0:p><ns0:p>2) To ensure that our manipulation on the frequency of exposure to the health proposal message is valid, the data from those who have seen this proposal before will be excluded.</ns0:p><ns0:p>We will ask all participants the following question: 'Have you seen this message before?' right after they are exposed to the message for the first time; we will ask the question in the first wave of the repetition condition and in the second wave of the no-repetition condition.</ns0:p><ns0:p>The data from participants whose answer is 'Yes' will be excluded.</ns0:p><ns0:p>3) For participants in the first wave of the repetition condition, a multiple-choice question concerning the content of the message ('What the message is about') will be asked to confirm whether they read the message carefully enough to capture its meaning. Data from participants who give a false answer will be excluded. 4) Before analyzing the data, we will calculate the standard deviation (SD) of each participant's scores on the 13 items in the second wave and exclude data from participants whose SD is zero. We plan to perform this data exclusion because an SD equal to zero means the same score on different items measuring divergent properties, which is strange and not possible. 5) As stated in Section 6, data of participants whose nationality is not Japanese will be excluded based on their answers to the relevant question on nationality.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 6) There will be an open-ended question on the experiment's real purpose at the end of the questionnaire under both conditions in the second wave. The question will be optional, and the data from participants who give correct answers (consistent with the experiment's real purpose) will be excluded.</ns0:p></ns0:div> <ns0:div><ns0:head>8)</ns0:head><ns0:p>What positive controls or quality checks will confirm that the obtained results are able to provide a fair test of the stated hypothesis?</ns0:p><ns0:p>Regarding the experiment design, to avoid unwanted interference from experiment materials (a health proposal message in our study), we conducted a pilot study to investigate whether it has been heard before and evaluated the agreement on the message. Considering that we want to discuss the effect of message repetition on the EPPM, a message already heard by most people is not appropriate. However, the COVID-19 pandemic's impact is so broad that it is unlikely health proposals have not been heard. We can only try to find one that is known to relatively few people and exclude participants who have heard it before in the main experiment. Besides, a strongly supported message is prone to ceiling effects because there is no more space for improvement in the evaluation. As a result, we selected a message heard by 40.8% of the participants in the pilot study with an average agreement score of 5.35 (SD = 1.29, 7-point scale), which is not too high to induce ceiling effects. Moreover, our preliminary experiment also showed that there were no floor or ceiling effects (for more details, please refer to Supplement 1).</ns0:p><ns0:p>During data collection and management, two questions will serve as manipulation checks on message repetition. The first question, which is for all participants when they are exposed to the message for the first time, is: 'Have you seen this message before?' If the answer is 'Yes,' it means there is interference from participants' previous experience on our manipulation on message repetition. Thus, their data will be excluded. The other question, which will be asked only in the repetition condition in the first wave to ensure that the message's first presentation is valid, is: 'What is the message about?' Data from participants with false answers will be excluded.</ns0:p></ns0:div> <ns0:div><ns0:head>9) Specify exactly which analyses you will conduct to examine the main question/hypothesis(es)</ns0:head><ns0:p>Since SEM will be used to analyze the data, we will evaluate our model's fit before proceeding to the hypotheses examination. It is reported that chi-square is sensitive to sample size <ns0:ref type='bibr'>(Schlermelleh-Engel et al. 2003)</ns0:ref>, and therefore, we will not rely on it as a basis for acceptance or rejection of the model. Instead, we will use root mean square error of approximation (RMSEA), comparative fit index (CFI), and Tucker-Lewis index (TLI) to evaluate the model's fit. If RMSEA &lt; .08, CFI &gt; .9, and TLI &gt; .9 <ns0:ref type='bibr' target='#b14'>(Kline, 2005;</ns0:ref><ns0:ref type='bibr' target='#b1'>Bentler &amp; Bonett, 1980)</ns0:ref> are all met, we will consider the model's fit acceptable. If RMSEA &lt; .06, CFI &gt; .95, and TLI &gt; .95 <ns0:ref type='bibr' target='#b11'>(Hu &amp; Bentler, 1999)</ns0:ref> are all met, we will consider the model's fit good. A detailed design planner on the present study is provided in Table <ns0:ref type='table'>2</ns0:ref>. The present study received approval from the psychological research ethics committee of the Faculty of Human-Environment Studies at Kyushu University (approval number: 2019-034). We will collect new data, and thus there will be no interference from existing data. This study consist of a series of anonymous online surveys. By participating in the surveys, participants consent to data collection. </ns0:p></ns0:div> <ns0:div><ns0:head>Self-Efficacy</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree)</ns0:p></ns0:div> <ns0:div><ns0:head>1</ns0:head><ns0:p>I am able to perform the underlined proposal to prevent the infection of COVID-19. 2</ns0:p><ns0:p>It is easy to perform the underlined proposal to prevent the infection of COVID-19. 3 I can perform the underlined proposal to prevent the infection of COVID-19.</ns0:p></ns0:div> <ns0:div><ns0:head>Response Efficacy</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree) 4</ns0:p><ns0:p>Performing the underlined proposal prevents the infection of COVID-19. 5</ns0:p><ns0:p>Performing the underlined proposal works in deterring COVID-19. 6</ns0:p><ns0:p>Performing the underlined proposal is effective in getting rid of COVID-19.</ns0:p></ns0:div> <ns0:div><ns0:head>Perceived Susceptibility</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree)</ns0:p></ns0:div> <ns0:div><ns0:head>7</ns0:head><ns0:p>I am at risk of being infected with COVID-19. 8</ns0:p><ns0:p>It is possible that I will get infected with COVID-19. 9</ns0:p><ns0:p>I am susceptible to COVID-19 infection. Severity</ns0:p><ns0:p>(1-strongly disagree, 7-strongly agree) 10 COVID-19 is a serious threat. 11 COVID-19 is harmful. 12 COVID-19 is a severe threat.</ns0:p></ns0:div> <ns0:div><ns0:head>Behavior Intention</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree) 13</ns0:p><ns0:p>In the future, when sanitizing my hands with alcohol-based hand sanitizer, I will press the pump slowly to the bottom to get a sufficient amount. As there are no other details to supplement, please refer to Section 5 for the sampling plan of the present study.</ns0:p><ns0:p>We will analyze relevant indexes in the model using SEM. Specifically, we will use the false discovery rate <ns0:ref type='bibr' target='#b0'>(Benjamini &amp; Hochberg, 1995)</ns0:ref> to adjust the p values of the coefficients of concern and then compare the adjusted p values with .05 to decide whether each coefficient is significant or not.</ns0:p><ns0:p>There is no evidence showing that the perceived threat of COVID-19 underlies its perceived susceptibility in the Japanese context.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)Manuscript to be reviewed10) Are you proposing to collect new data or analyze existing data?</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>1</ns0:figDesc></ns0:figure> <ns0:note place='foot' n='2'>PeerJ reviewing PDF | (2020:07:50866:1:2:CHECK 2 Sep 2020)</ns0:note> </ns0:body> "

"Graduate School of Human-Environment Studies Kyushu University 744 Motooka, Nishiku, Fukuoka 819-0395, Japan Tel/Fax: +81-92-642-2418 August 28, 2020 Dear Editors: We appreciate the generous comments of the reviewers on the manuscript and have edited it to address their concerns. In particular, we have uploaded another supplementary file that describes the preliminary experiment. We believe that the manuscript is now suitable for publication in PeerJ. Jingwen Yang Graduate School of Human-Environment Studies, Kyushu University On behalf of all authors. Reviewer 1 (Hamid Hassen) Basic reporting • Line 2 - 4: Your abstract needs modification. How you relate fear with health messages is not clear. Health messages are not necessarily be fearful I suggest that you improve the description at lines 2-4 to provide more description for rationale. Reply: We have modified the first two sentences in the abstract (lines 2–6) and made it clear that the target of fear is not health proposals but the health threat, which in this study, is the coronavirus disease 2019 (COVID-19). The fear of COVID-19 as a health threat can predict health compliance with related health proposals, which do not necessarily induce fear. • Line 5 - 6: The title doesn't tell about the first aim (“EPPM in naturalistic context…”)...rather it focuses only on the second objective. Reply: We have modified the title to include the focus on EPPM as well. • You described the association between health proposal on Susceptibility. However, it is not clear what kind of you referring for, physiological or perceived. I think you mean perceived susceptibility. Be explicit in description. Reply: We have changed all instances of “susceptibility” in the manuscript to “perceived susceptibility.” • Your description of the rationale in Line 35-36, you mentioned as there are inconsistent findings. It needs more elaboration on the inconsistent findings by providing necessary studies. Reply: We have added relevant studies with inconsistent findings for reference in lines 40–43 of the manuscript. For example, only a small number of studies report the negative main effect of fear on persuasion (Janis & Feshbach, 1953; Janis & Terwilliger, 1962) out of a large number of studies indicating the positive main effect of fear (Leventhal & Watts, 1966; Dabbs & Leventhal, 1966; Leventhal & Singer, 1966). • In line 84-85 of the introduction, you mentioned that “Even though the degree to which one is influenced by the pandemic may vary, the cognition of its perceived threat can be developed spontaneously.” However, the cognition of perceived threat also vary to a large extent as the influence varies. Reply: I agree. We do not assume that everyone’s cognition of perceived threat stays at the same level, and as in the experimental design, will measure the degree of perceived threat individually. What we mean here is that while people may not be familiar with some diseases like melanoma, they can understand the health threat of COVID-19 without extra explanation. The relevant clarification has been added to the manuscript in lines 98–101. • The way you write the hypothesis needs modification. It looks, in any way the hypothesis is true regardless of the data. I suggest to formulate it like; H1: Perceived efficacy has a positive/negative/ (depending on whether one- or twosided test) effect on self-efficacy. H0: Perceived efficacy has no effect on self-efficacy. Reply: We have modified the eight hypotheses of this study based on your comments. Positive effects are hypothesized in Hypotheses 1–5 and 7. Considering the difficulty in addressing the exact direction (positive/negative) of the effect in Hypotheses 6 and 8, we adopted the formulation of “H1a - positive, H1b - negative” to describe them. • Table: you frequently mentioned “Based on a preliminary experiment….”. Could you cite the results of the preliminary experiment or upload as supplementary material? Reply: We have uploaded a supplementary material in which a brief introduction to the preliminary experiment is provided. • The interpretation column of the table need to be revised. The one written here is a general rule not specific to your study and context. It is also a redundancy from the hypothesis section. Reply: We have revised the interpretation column and deleted redundant content. In particular, we specified the target of perceived threat and perceived efficacy for hypotheses 1–4 to avoid generality in interpretation. • The table in the analysis plan is not relevant. All are redundancy from the hypothesis section and sampling. You can summarize it using a single paragraph or simple table provided that there is no specific analysis for each hypothesis. Reply: We have combined the cells into one in which the model’s analysis plan is summarized. Experimental design • In your research question, line 28-29 you described “to examine how people health compliance intention is influenced by various factors…….”. This research question is not reflected in the title. Your title needs modification to accommodate most of the research aims. Reply: The title has been modified. Given that the main research aims are to explore the association of message repetition with factors in the EPPM and its influence on behavior intention toward COVID-19 health proposals, the modified title should fit these aims. • Line 12-13: The study aimed to compare one vs two-time messages. However, the description in line 12-13 (‘the times of exposure’) and also in the main manuscript looks as you have multiple exposure times. It is not possible to assess the association of the ‘times’ of exposure with response efficacy. The difference between one vs two messages might not be necessarily equal to the difference in two vs three messages. Your interpretation seems continuous exposure variable but it is binary. Reply: We have changed “the times of exposure” in line 13–14 to “repetition of the health proposal” to avoid misunderstanding. We agree with you that instead of number of times of exposure, the variable to be assessed in the present study should be whether there is message repetition or not. However, the remaining instances of “the times of exposure” in the manuscript are to describe details of our manipulation, not to define the variable (e.g., lines 171, 234, etc.). We consider these descriptions necessary and free of vagueness, and thus have decided to keep them. Furthermore, we changed “times of exposure” to “frequency of exposure” based on another reviewer’s comments. • In this study, a number of hypotheses are being tested. Multiple comparison is still an issue. As you are evaluating the statistical significance of multiple parameters in a model, the type I error is expected to inflate. You should describe how to control the multiplicity issues. What type of statistical adjustment will you use? Reply: We will use the false discovery rate (FDR) proposed by Benjamini and Hochberg (1995) to perform multiplicity control for the parameters in the SEM. Relevant descriptions have been added to the “Analysis Plan” column in the table in Section 9 of the manuscript. • In line 197-199 you described how you determine sample size. However, the sample size calculation is not clear. The ‘findRMSEAsamplesize’ function of R requires a degree of freedom to calculate sample size and the sample size varies accordingly. What degree of freedom did you used? How much is the expected dropout rate and you need to adjust for it as well. Reply: The degree of freedom in the present study is 70, which is determined by the model shown in Figure 1. We have added the information in line 206. Regarding the dropout rate, it should not be a problem since we plan to collect data from additional participants if the required sample size is not met. This is mentioned in lines 216–218 of the manuscript. • In your inclusion criteria, in line 222-224: Will you use the nationality as screening or they will fill out the questionnaire regardless of their nationality? Reply: Participants will fill out the questionnaire regardless of their nationality, because we are not able to screen participants according to their nationality. We can only remind them before they respond to the questionnaire. There will be a question on nationality before the target 13 items in Table 1, and the answer to this question will determine whether the participant’s data will be included or not. • The data exclusion and sample size adjustment is not clear. Needs more elaboration. Reply: We have elaborated Section 7 concerning data exclusion as follows: a) We changed the ACQ’s formulation in criterion 1 to avoid confusion, b) modified the descriptions in criterion 4, and c) added 2 data exclusion criteria (namely 5 and 6). We also modified Section 5 concerning sample size adjustment by deleting the contents repeating what is stated in Section 7 and adding extra descriptions. • In line 246-247 of the data exclusion section you mentioned that participants with SD of 0 will be excluded. I didn’t get the reason to exclude the participants with SD of 0. The measurement is only twice, it is highly possible to have same score i.e. SD of 0. Why you exclude them? Reply: The SD here refers to the SD of the 13 items in Table 1. Since the 13 items measure different properties, it is strange that they have the same score, which is indicated by the statistical result of “SD = 0.” Relevant descriptions have been modified in the manuscript. There may be some misunderstanding. There are two conditions (i.e., “no repetition” vs. “repetition”) in our study, and thus the participants in each condition will only be measured once. The SD is limited to the only-once result of each participant. • The quality check and control needs descriptions in detail what quality checks will be performed during the design, data collection, and management. Reply: We have elaborated Section 8 by categorizing quality checks according to the stage to which they belong (i.e., design, data collection, and management) and adding more details. Analysis plan • It is recommended to use Tucker–Lewis index in addition to RMSEA and CFI to assess the model fit. Reply: We have added the Tucker–Lewis index to assess the model fit (lines 287–291). • Page 24: The items looks similar. Any validity test for these items? How is the interitem correlation of the questionnaire? In general, the psychometric validity of all the tools to be used need to be described in detail. Reply: We performed validity tests for items 1–12 using the data from the preliminary experiment and verified the items’ validity. Relevant descriptions have been added in the manuscript (lines 163–167) and Supplementary 1. Validity of the findings • As a protocol, I couldn’t assess the validity of the findings comparing with the data. • You mentioned about the preliminary experiment you conducted, however, neither the data nor the results are provided. Reply: We have uploaded supplementary material (Supplementary 1) in which a brief introduction to the preliminary experiment is provided. Comments for the author • Line 11-12: It says “SEM is used to....” As it is a protocol, it needs to be corrected to “SEM will be used….” Reply: Thank you for your advice. We have corrected this sentence to “SEM will be used…” (line 12). • You repeatedly used the phrase “we predicted” for instance line 12, 15… You mean hypothesized? These two terms are different. Reply: In the abstract, we use the word “predict” to make predictions of what will happen under our experiment design based on the hypotheses in Section 3. As far as we are concerned, “hypothesize” is used to put forward testable explanations for the observed phenomena, while “predict” is used to forecast future events. We thus consider “predict” more appropriate in the given context. • Line 108-111: Put the direct translation of the message in English. Reply: We have revised this part and used the direct English translation of the message instead (lines 138–141). • Line 254-256: This sentence should be part of the exclusion not the quality control Reply: Both the exclusion and quality control sections have been edited. • The description written in the “Sampling plan” column of the Table is repetition. You should remove it. Reply: We have removed the descriptions in the “Sampling plan” column and replaced it with a guiding sentence to Section 5. • RMSEA, CFI – Put the long form at first instance - root mean square error of approximation - RMSEA, comparative fit index - CFI. Reply: Thank you for pointing this out. We have added their long forms where they are first mentioned in the manuscript. Reviewer 2 Basic reporting Yang and colleagues seek to probe factors relating to the EPPM model of fearmotivated health behaviors as well as frequency of measure exposure that may affect behavior intention in the context of the Covid-19 pandemic. Such a study may identify features that have led to relative success in managing the pandemic in some countries relative to others. The proposed study would certainly be beneficial in our attempt to understand and combat the pandemic. Further, the proposed study benefits from a clean and elegant experiment, that can be easily understood by respondents and then further operationalized for analysis. One omission is of preliminary analyses of the pilot data. The study also does not appear to involve any neuroscience and therefore should not be listed as such . Reply: Thank you for your advice. We have reedited the tags and deleted the “neuroscience” tag. Furthermore, we uploaded supplementary material (Supplementary 1) for reference, which provides a brief introduction to the preliminary experiment along with the analyses. Experimental design • 2. Experimental design 1. I recommend changing the ACQ to something less subjective. You risk biasing your sample towards those that already formulate their beliefs based on evidence, which does not comprise the general population, or at the very least, lose participants who find it humorous to answer > 1. Perhaps a simple request to select a number on the scale could serve as an ACQ. Reply: We have revised the ACQ based on your advice. Please refer to lines 235–237. 2. There is also a concern regarding the nationality restriction, given that the question regards a global pandemic. There may be IRB limits to your participant pool but it may be advantageous to expand the survey population, if possible, to other nationalities. Reply: Thank you for your proposal. We agree with your idea that it is of significance to expand the study to other nationalities considering that the pandemic is global. However, as you mention, the present study was only approved by the IRB at Kyushu University, Japan; therefore, the participants are limited to those who are Japanese. We would like to first check the hypothesized effects in the present study. Based on the results, we will then decide whether to conduct further research on the broader population in another study. 3. 1. In line 23, the message to be repeated is first assessed according to favorability but it is unclear what is meant by this measure. Is it assessing how much respondents: 1) like the message; 2) agree with the message? Reply: Here, favorability means the degree to which one agrees with the message. To avoid ambiguity, we added clarification on favorability in the manuscript (line 142, 144–145). Validity of the findings N/A Comments for the author Major comments It is a bit unclear what the specific aim /question of the study is. In the abstract, I can identify 4: response efficacy ~ frequency of exposure; susceptibility ~ frequency of exposure; behavior intention ~ perceived efficacy; behavior intention ~ perceived threat. The stated aims in the rest of the report are fuzzier. There are some attempts, in lines 77-78; 96-98; 101-103, but they are vague. This part of the report needs to be tightened. With regards to the specific hypotheses listed, there are too many and some may not be pertinent, notably hypotheses 1-4. Perhaps providing a real-world translation in the interpretation column could help. But it is not clear why finding a relationship between perceived and self-efficacy is of interest or utility in the context of the proposed study. The same criticism underlies hypothesis 2. With hypotheses 3 and 4, one can argue that it is highly likely that perceived threat (i.e. some average value of susceptibility and severity) is related to susceptibility and severity. Reply: Thank you for the advice. We have tightened the four aims you refer to (i.e., response efficacy – repetition; susceptibility – repetition; behavior intention – perceived efficacy; behavior intention – perceived threat) and clarified the aims of confirming the underlying roles of perceived efficacy and perceived threat as second-order factors in the EPPM in the COVID-19 pandemic (lines 84–89). Furthermore, we explained why hypotheses 1–4 are of interest in dealing with real-world problems concerning the COVID-19 pandemic (lines 112–123) by citing concrete examples (e.g., the health proposal of social distancing). Minor comments: 1. In the opening paragraph of the introduction, you use colloquial and vague language (“It has been a while…”). Please be more specific, so as to allow the reader to contextualize the problem (e.g. “As of writing, it has been x months since the outbreak (…)”. Second, the impact on life and financial well-being are not incalculable, as both life and economic loss can be counted. Reply: Thank you for your correction. We have revised these two sentences to “At the time of writing, it has been seven months…” and “heavy loss of both life and economy…” 2. Some language throughout the report is a bit colloquial (e.g. “sum up” instead of “in summary) and there are a few typos (Hypothesis 4 is listed twice in the table). A revision of the writing would benefit the report. Reply: Thank you for your advice. We have corrected the typos you pointed out and revised the manuscript. 3. Please change “times of exposure” to “frequency of exposure” to avoid confusion. Reply: We have changed all instances of “times of exposure” to “frequency of exposure” according to your advice. Reviewer 3 Basic reporting The project meets basic report standards. The table is somewhat large and difficult to read, however, as the cells aren't always aligned well. (eg., sampling plan doesn't track all the way down with the other cells) Reply: We have reedited the table and ensured there is no redundancy. There are a few minor English errors in the manuscript and it should be thoroughly proofread. Reply: Thank you for your advice. We have revised the manuscript and again had it proofread by a native speaker of English. Experimental design The experimental design is poor. The primary manipulation isn't very strong (1 vs 2 repetitions) and there is no control group. Most of the hypotheses are regarding structural relationships that are well-established and unlikely to be impacted by repetition very extensively. There aren't safe guards to ensure that any changes aren't due to demand characteristics. Reply: Thank you for your comments. As mentioned in lines 67–74, previous research has indicated that agreement on a message, which is perceived as strong and of high personal relevance, is enhanced when the frequency of exposure to the message increases from once to twice. Strong manipulation does not necessarily mean more repetition. Given that COVID-19 has already broken out and time cannot flow backward, it is difficult to add a control group. The present study is both confirmatory and exploratory in design. We believe it necessary to confirm the validity of established theory on health communication in the COVID-19 pandemic, because exceptions and uncertainty exist and must be checked. Furthermore, it would benefit our understanding of health emergencies if the established model can be expanded and incorporate more factors like message repetition, which is the exploratory aspect of our study. To prevent changes due to demand characteristics, we have added an open-ended question on the experiment’s real purpose at the end of the questionnaire in the second wave for both conditions. Participants’ data will be excluded if their answers are consistent with the experiment’s real purpose. Please refer to lines 255–258 in the manuscript. Validity of the findings I doubt the current study will lead to robust repetition effects, and it is lacking evidence that it should. Reply: We have conducted a preliminary experiment (Supplementary 1), in which there is a tendency for repetition effects, with the same design as the main experiment. Considering that the statistical power of the main experiment is increased, it is promising that the robustness of the repetition effects is confirmed. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>When people are confronted with health proposals during the coronavirus disease 2019 (COVID-19) pandemic, it has been suggested that fear of COVID-19 can serve protective functions and ensure public health compliance. However, health proposal repetition and its perceived efficacy also influence the behavior intention toward the proposal, which has not yet been confirmed in the COVID-19 context. The present study aims to examine whether the extended parallel process model (EPPM) can be generalized to a naturalistic context like the COVID-19 pandemic. Additionally, we will explore how repetition of a health proposal is involved with the EPPM. In this study, two groups of participants are exposed to the same health proposal related to COVID-19, where one group is exposed once and another group twice. They then fill out a questionnaire consisting of items concerning behavior intention and adapted from the Risk Behavior Diagnosis Scale. Structural equation modeling will be used to determine the multivariate associations between the variables. We predict that repetition of the health proposal will associate with response efficacy (i.e., a belief about the effectiveness of the health proposal in deterring the threat) and perceived susceptibility (i.e., a belief about the risk of experiencing the threat). It is also predicted that following the EPPM, behavior intention will associate with both perceived efficacy of the health proposal, which will underlie response efficacy, and perceived threat of COVID-19, which will underlie perceived susceptibility. We will discuss the process, based on the model, where health message repetition affects behavior intention during the COVID-19 pandemic.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>1) What is the main question being addressed in your study?</ns0:head><ns0:p>At the time of writing, it has been seven months since the outbreak of the COVID-19 pandemic. During this period, while a heavy loss of both life and economy has been caused, some countries and regions have achieved staged success in the fight against this disease. There is no doubt that public compliance with effective health proposals plays a crucial role in achieving this success.</ns0:p><ns0:p>It is suggested that functional fear, which is explained as one of the negative emotions serving protective functions in certain contexts, has promoted public health compliance during the COVID-19 pandemic <ns0:ref type='bibr' target='#b17'>(Harper et al., 2020)</ns0:ref>. Nevertheless, this needs to be explored further, especially considering previous studies on health communication and the features of information dissemination in real life. The main question being addressed in our study is to examine how people's health compliance intention is influenced by various factors in the COVID-19 context in an exhaustive way.</ns0:p><ns0:p>It is known that fear has an impact on health compliance. In fact, fear appealing communication is considered an effective way to promote health campaigns and has been widely investigated for promoting health awareness related to various topics, including smoking <ns0:ref type='bibr' target='#b25'>(Leventhal &amp; Watts, 1966)</ns0:ref>, alcohol use <ns0:ref type='bibr' target='#b44'>(Wolburg, 2001;</ns0:ref><ns0:ref type='bibr' target='#b29'>Moscato et al., 2001)</ns0:ref>, AIDS <ns0:ref type='bibr' target='#b38'>(Treise &amp; Weigold, 2001)</ns0:ref> and so forth. After several initial studies, inconsistent results indicated that a simple monotonic function of fear may not be expected in persuasive health communication. Specifically, despite a large number of studies indicating the positive main effect of fear on persuasion <ns0:ref type='bibr' target='#b25'>(Leventhal &amp; Watts, 1966;</ns0:ref><ns0:ref type='bibr' target='#b14'>Dabbs &amp; Leventhal, 1966;</ns0:ref><ns0:ref type='bibr'>Leventhal &amp; Singer, 1966)</ns0:ref>, relatively few report the negative main effect of fear <ns0:ref type='bibr' target='#b19'>(Janis &amp; Feshbach, 1953;</ns0:ref><ns0:ref type='bibr' target='#b20'>Janis &amp; Terwilliger, 1962)</ns0:ref>. Later research confirmed that the effects of fear appeal interact with various source variables, message variables, and receiver variables, and thus cannot be described easily <ns0:ref type='bibr' target='#b28'>(Miller &amp; Hewgill, 1966)</ns0:ref>.</ns0:p><ns0:p>One of the most recent and prevalent theories on fear appealing communication is the extended parallel process model (EPPM: <ns0:ref type='bibr' target='#b40'>Witte, 1992</ns0:ref><ns0:ref type='bibr' target='#b42'>Witte, , 1994))</ns0:ref>, which is based on former frameworks including the Parallel Response Model <ns0:ref type='bibr' target='#b26'>(Leventhal, 1970)</ns0:ref> and protection motivation theory <ns0:ref type='bibr' target='#b33'>(Rogers, 1975)</ns0:ref>. In the EPPM, there are four main factors, which influence the prediction of certain communication outcomes: perceived susceptibility and severity composing perceived threat, self-efficacy and response efficacy composing perceived efficacy. Perceived susceptibility refers to a belief about the risk of experiencing a threat, whereas severity refers to a belief about the magnitude of the threat. On the other hand, self-efficacy is defined as a belief about the ability to perform a recommended proposal to avert the threat; response efficacy is a belief about the effectiveness of the recommended proposal in deterring the threat <ns0:ref type='bibr' target='#b43'>(Witte, 1996)</ns0:ref>. Concretely, when both perceived threat and perceived efficacy are high, people are most likely to engage in a danger control process, which means conforming to the recommended health proposal. Nevertheless, when perceived threat and perceived efficacy are high and low, respectively, people turn to a fear control process leading to coping responses that reduce fear and danger control responses. Meta-analyses on the results of fear appeal research confirm the validity of the EPPM <ns0:ref type='bibr' target='#b15'>(Floyd, Prentice-Dunn &amp; Rogers, 2000;</ns0:ref><ns0:ref type='bibr'>Peter, Ruiter &amp; Kok, 2013)</ns0:ref>.</ns0:p><ns0:p>Currently, we are exposed to health proposals concerning COVID-19, in all likelihood, more than once in our daily lives, which makes message repetition an essential factor to be taken into consideration. Although research concerning stimulus repetition first emerged in the 1960s <ns0:ref type='bibr' target='#b45'>(Zajonc, 1968)</ns0:ref>, there was no study on message repetition in persuasive communication until ten years later. It is suggested that under moderate repetition (less than 3 times), agreement toward persuasive messages rise <ns0:ref type='bibr' target='#b9'>(Cacioppo &amp; Petty, 1979)</ns0:ref>. The rationale is explained as follows: Scrutiny is reinforced through a moderate level of message repetition, which enhances the understanding of message content and the merits advocated by it, consequently improving supportive attitudes toward the message. Later research revealed that the mechanism mentioned above is applicable only when arguments in the message are perceived as strong, and when the issue is of high personal relevance <ns0:ref type='bibr' target='#b10'>(Cacioppo &amp; Petty, 1989;</ns0:ref><ns0:ref type='bibr' target='#b12'>Claypool et al., 2004)</ns0:ref>.</ns0:p><ns0:p>In health communication on COVID-19, we assume that the interpretation of the content of a certain health proposal message may be related to factors in the EPPM. Concretely, response efficacy and perceived susceptibility may be perceived when reading a health proposal message. If the content of a health proposal is supported, high response efficacy will be found, resulting in a decrease in perceived susceptibility. Given the connection between the EPPM and message repetition in persuasive health communication and the actual state we have been through when confronted with health proposals during the COVID-19 pandemic, we built an integrative model to investigate the factors and their associations concerning an individual's behavior intention to conduct effective health proposals to prevent the infection. In the model, we will first focus on the influence of message repetition on the response efficacy of a certain health proposal and the perceived susceptibility of COVID-19. We will then elaborate on the change in behavior intention toward the proposal due to the variation in perceived efficacy and perceived threat. Furthermore, we will confirm whether the underlying roles of perceived efficacy and perceived threat hold true in the COVID-19 pandemic. The present study is unique and necessary in several aspects.</ns0:p><ns0:p>One, considering that COVID-19 is a real-life and ongoing public health emergency, some of its properties are fixed and thus cannot be manipulated. Take perceived threat for example: In previous research, it was always considered together with fear because without extra fear-arousing materials (e.g., explanation of a certain disease in text or video), the perceived threat may not be notable enough to act as an independent variable. Unlike other unfamiliar diseases (e.g., melanoma), even though the degree to which one is influenced by the pandemic may vary, COVID-19 should be one of immediate health threats in several countries and regions including Japan, where the present study will be conducted. Considering that fear-arousing information is no longer needed in the message, we are interested in exploring how perceived threat alone for the COVID-19 affects behavior intention when there is no intentional manipulation of fear arousal. Two, there is still no research to test a model that combines the EPPM with message repetition. Although studies on similar topics have been conducted, the results have not been analyzed in an integrated way <ns0:ref type='bibr' target='#b37'>(Skilbeck, Tulips &amp; Ley, 1977;</ns0:ref><ns0:ref type='bibr' target='#b38'>Treise &amp; Weigold, 2001;</ns0:ref><ns0:ref type='bibr' target='#b36'>Shi &amp; Smith, 2016)</ns0:ref>. To be specific, when supportive arguments toward the content of a recommended proposal are enhanced due to moderate repetition, it can also be interpreted as the change in response efficacy and perceived susceptibility in the EPPM. Nevertheless, this connection between two research topics has rarely been made clear, causing difficulty in making an exact prediction of the results.</ns0:p><ns0:p>Three, it remains essential to confirm the validity of EPPM's established construction, namely the sub-dimensions of perceived efficacy and perceived threat, to enhance compliance with public health guidelines in the COVID-19 pandemic. We assume better health compliance is due to higher perceived efficacy, but what does perceived efficacy mean? By learning more about its two indicators, self-efficacy and response efficacy, we can better understand why people choose to conform or not to certain health proposals. For instance, even though physical distancing is considered efficient in preventing infection (high response efficacy), the difficulty in conducting it may vary between people based on their socializing needs (divergent selfefficacy), which results in different levels of health compliance. Similarly, if we are aware that perceived threat is indicated by perceived susceptibility and severity, we may know which properties to emphasize in education on COVID-19 to boost public health compliance.</ns0:p><ns0:p>In summary, the present study is of comprehensive significance in revealing the mechanism in the purview of public compliance with effective health proposals in the COVID-19 pandemic.</ns0:p><ns0:p>2) Describe the key independent and dependent variable(s), specifying how they will be measured.</ns0:p><ns0:p>We will conduct a two-wave survey for two groups of participants: One is a no-repetition group and the other is a repetition group. In the first wave, two groups will first answer the same dummy questionnaire of the Need for Cognition Scale <ns0:ref type='bibr' target='#b22'>(Kouyama &amp; Fujihara, 1991)</ns0:ref>. At the end of the questionnaire, only the repetition group will be exposed to the target health proposal message, which is written in Japanese as:</ns0:p><ns0:formula xml:id='formula_0'>&#12450;&#12523;&#12467;&#12540;&#12523;&#12399;&#28040;&#27602;&#12539;&#27578;&#33740;&#21177;&#26524;&#12364;&#12354;&#12427;&#12392;&#35328;&#12431;&#12428;&#12390;&#12356;&#12414;&#12377;&#12290;&#19968;&#26041;&#12289;&#23569;&#37327;&#12398;&#12450;&#12523;&#12467;&#12540;&#12523;&#28040;&#27602;&#28082; &#12434;&#25163;&#12395;&#21462;&#12387;&#12383;&#22580;&#21512;&#12289;&#28040;&#27602;&#21177;&#26524;&#12364;&#20302;&#19979;&#12375;&#12390;&#12375;&#12414;&#12356;&#12414;&#12377;&#12290;&#12381;&#12371;&#12391;&#12289;&#12450;&#12523;&#12467;&#12540;&#12523;&#28040;&#27602;&#28082;&#12391;&#25163;&#12434; &#28040;&#27602;&#12377;&#12427;&#38555;&#12399;&#12289;&#23569;&#12394;&#12367;&#12392;&#12418;&#12509;&#12531;&#12503;&#37096;&#20998;&#12434;&#19979;&#12414;&#12391;&#12422;&#12387;&#12367;&#12426;&#25276;&#12375;&#12390;&#12289;&#12383;&#12387;&#12407;&#12426;&#25163;&#12395;&#21462;&#12387;&#12390;&#20351; &#12358;&#12371;&#12392;&#12434;&#12362;&#12377;&#12377;&#12417;&#12375;&#12414;&#12377;&#12290;</ns0:formula><ns0:p>The English translation of the message is as follows: 'While alcohol-based hand sanitizer is useful for preventing COVID-19 infection, the effect will be discounted if not enough amount is used. Therefore, it is recommended to press the pump to the bottom every time to get enough amount of hand sanitizer.' We selected this health proposal message according to a pilot study to examine how much the message is known and agreed on. As per the results, the knowledge rate of the message was low (40.8%). The attitudes to the message were measured using a 7-point scale (ranging from 1 = strongly disagree and 7 = strongly agree). The attitude to the message score was considerably favorable (i.e., the scores were significantly higher than 4 on the scale, M = 5.35, SD = 1.29, t (200) = 5.52, p &lt; .001, Cohen's dz = 1.047). Considering the definition by <ns0:ref type='bibr' target='#b10'>Cacioppo &amp; Petty (1989)</ns0:ref> that a strong argument refers to the one toward which favorable thoughts are generated predominantly, the above message is appropriate in the present study in two ways. One is that as mentioned above (i.e., 1. What is the main question being addressed in your study?). The supportive arguments on a message can be improved through moderate repetition only when the message is believed to be a strong one; therefore, our manipulation on message repetition will be meaningful. The other reason is that since the scores on favorable thought are moderately high, there is still space for supportive arguments to increase, preventing ceiling effect.</ns0:p><ns0:p>The second wave will be conducted 24-72 hours after the first wave. The interval range is set because we cannot control the precise timepoint at which participants answer the second questionnaire even if we invite them to do it on time. In the second wave, both groups of participants will be exposed to the same message, which is identical to the message shown to the repetition group in the first wave. They will then fill out the same questionnaire containing 13 items (Table <ns0:ref type='table'>1</ns0:ref>). Twelve items are adapted from the Risk Behavior Diagnosis Scale <ns0:ref type='bibr' target='#b43'>(Witte, 1996)</ns0:ref>, and 1 item enquires behavior intention toward the target health proposal. All items will be scored on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). In the preliminary experiment, which had the same design as the main experiment (more details in Supplementary 1), we checked the convergent validity and discriminant validity for items 1-12. The results of our calculations are as follows: (1) average variance extracted (AVE) &gt; .5; and (2) square root of AVE &gt; inter-construct correlations are both met, confirming the validity of the items <ns0:ref type='bibr' target='#b16'>(Hair et al., 2010)</ns0:ref>. The scores of the 13 items in Table <ns0:ref type='table'>1</ns0:ref> will be treated as the key dependent variables.</ns0:p><ns0:p>We aim to test the model (Figure <ns0:ref type='figure'>1</ns0:ref>) which combines the EPPM with message repetition, further exploring the factors that associate with behavior intention. As described above, there will be two conditions with various frequencies of exposure to the message. Each participant will be assigned to one of the conditions. Thus, repetition, as a binary variable in the model, will be adopted as the key independent variable.</ns0:p></ns0:div> <ns0:div><ns0:head>3) What are your hypotheses?</ns0:head><ns0:p>The hypotheses of this research are: &#61599; Hypothesis 1: Perceived efficacy has a positive effect on self-efficacy. H0: Perceived efficacy has no effect on self-efficacy.</ns0:p><ns0:p>&#61599; Hypothesis 2: Perceived efficacy has a positive effect on response efficacy. is the number of observed variables (i.e., 14 in the model in Figure <ns0:ref type='figure'>1</ns0:ref>), and q is the number of parameters to be estimated (i.e., 35 in the model). Given that there are two conditions in the present study, we set N = 151 as the required sample size for each condition.</ns0:p><ns0:p>We doubled the required sample size as the maximum sample size (i.e., N = 602 in total and 301 per condition) because we have to plan to collect data from a larger sample for the following two reasons: (a) A large amount of data may have to be excluded based on the criteria detailed below (i.e., 7. What are your data exclusion criteria?). b) The dropout rate from the second wave may be high because it is hard to ensure all the participants in the first wave will take part in the second wave voluntarily within the requested time range, which is 24-72 hours after the first wave.</ns0:p><ns0:p>If the collected data does not reach the required sample size (i.e., N = 301 in total and 151 per condition) after excluding the data meeting the criteria in Section 7, we will collect data from additional participants to reach the required sample size. However, if the number of participants exceeds 602, we will choose the data of the first 602 participants based on the time stamp and use those for analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>6) What are your study inclusion criteria?</ns0:head><ns0:p>Participants will be recruited via Yahoo! Crowdsourcing (http://crowdsourcing.yahoo.co.jp/). They should possess Japanese nationality and be over 18 years old. We will indicate the criteria to potential participants in the instruction and invite them to participate only when they fulfill the criteria. Besides, questions on nationality and age will be asked before the main questionnaire (i.e., the items of the Risk Behavior Diagnosis Scale and behavior intention toward the target health proposal).</ns0:p></ns0:div> <ns0:div><ns0:head>7) What are your data exclusion criteria?</ns0:head><ns0:p>We will apply six criteria to perform data exclusion. 1) To identify distracted respondents or satisficers <ns0:ref type='bibr' target='#b11'>(Chandler et al., 2014;</ns0:ref><ns0:ref type='bibr' target='#b31'>Oppenheimer et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b34'>Sasaki &amp; Yamada, 2019)</ns0:ref>, we will insert a simple question as an attention check question (ACQ) in the middle of the questionnaire in the second wave. The reason why we do not add an ACQ in the first wave is that the first wave is conducted only to control the frequency of exposure to the message, and thus the data in the first wave will not be analyzed. The ACQ in the second wave will be: Please choose number '2' from below. Consistent with other items, the ACQ will also use a 7-point scale. The data from participants who choose 1 or 3-7 will be excluded.</ns0:p><ns0:p>2) To ensure that our manipulation on the frequency of exposure to the health proposal message is valid, the data from those who have seen this proposal before will be excluded. We will ask all participants the following question: 'Have you seen this message before?' right after PeerJ reviewing PDF | (2020:07:50866:2:0:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed they are exposed to the message for the first time; we will ask the question in the first wave of the repetition condition and in the second wave of the no-repetition condition. The data from participants whose answer is 'Yes' will be excluded.</ns0:p><ns0:p>3) For participants in the first wave of the repetition condition, a multiple-choice question concerning the content of the message ('What the message is about') will be asked to confirm whether they read the message carefully enough to capture its meaning. Data from participants who give a false answer will be excluded.</ns0:p><ns0:p>4) Before analyzing the data, we will calculate the standard deviation (SD) of each participant's scores on the 13 items in the second wave and exclude data from participants whose SD is zero. We plan to perform this data exclusion because an SD equal to zero means the same score on different items measuring divergent properties, which is strange and not possible. 5) As stated in Section 6, data of participants whose nationality is not Japanese and whose age is under 18 will be excluded based on their answers to the relevant questions.</ns0:p><ns0:p>6) There will be an open-ended question on the experiment's real purpose at the end of the questionnaire under both conditions in the second wave. The question will be optional, and the data from participants who give correct answers (consistent with the experiment's real purpose) will be excluded.</ns0:p></ns0:div> <ns0:div><ns0:head>8) What positive controls or quality checks will confirm that the obtained results are able to provide a fair test of the stated hypothesis?</ns0:head><ns0:p>Regarding the experiment design, to avoid unwanted interference from experiment materials (a health proposal message in our study), we conducted a pilot study to investigate whether it has been heard before and evaluated the agreement on the message. Considering that we want to discuss the effect of message repetition on the EPPM, a message already heard by most people is not appropriate. However, the COVID-19 pandemic's impact is so broad that it is unlikely health proposals have not been heard. We can only try to find one that is known to relatively few people and exclude participants who have heard it before in the main experiment. Besides, a strongly supported message is prone to ceiling effects because there is no more space for improvement in the evaluation. As a result, we selected a message heard by 40.8% of the participants in the pilot study with an average agreement score of 5.35 (SD = 1.29, 7-point scale), which is not too high to induce ceiling effects. Moreover, our preliminary experiment also showed that there were no floor or ceiling effects (for more details, please refer to Supplement 1).</ns0:p><ns0:p>During data collection and management, two questions will serve as manipulation checks on message repetition. The first question, which is for all participants when they are exposed to the message for the first time, is: 'Have you seen this message before?' If the answer is 'Yes,' it means there is interference from participants' previous experience on our manipulation on message repetition. Thus, their data will be excluded. The other question, which will be asked only in the repetition condition in the first wave to ensure that the message's first presentation is valid, is: 'What is the message about?' Data from participants with false answers will be excluded. 9) Specify exactly which analyses you will conduct to examine the main question/hypothesis(es) Since SEM will be used to analyze the data, we will evaluate our model's fit before proceeding to the hypotheses examination. It is reported that chi-square is sensitive to sample size <ns0:ref type='bibr'>(Schlermelleh-Engel et al. 2003)</ns0:ref>, and therefore, we will not rely on it as a basis for acceptance or rejection of the model. Instead, we will use root mean square error of approximation (RMSEA), comparative fit index (CFI), and Tucker-Lewis index (TLI) to evaluate the model's fit. If RMSEA &lt; .08, CFI &gt; .9, and TLI &gt; .9 <ns0:ref type='bibr' target='#b21'>(Kline, 2005;</ns0:ref><ns0:ref type='bibr' target='#b8'>Bentler &amp; Bonett, 1980)</ns0:ref> are all met, we will consider the model's fit acceptable. If RMSEA &lt; .06, CFI &gt; .95, and TLI &gt; .95 <ns0:ref type='bibr' target='#b18'>(Hu &amp; Bentler, 1999)</ns0:ref> are all met, we will consider the model's fit good. More details are described in Table <ns0:ref type='table'>2</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>10) Are you proposing to collect new data or analyze existing data?</ns0:head><ns0:p>The present study received approval from the psychological research ethics committee of the Faculty of Human-Environment Studies at Kyushu University (approval number: 2019-034). We will collect new data, and thus there will be no interference from existing data. This study consist of a series of anonymous online surveys. By participating in the surveys, participants consent to data collection. </ns0:p></ns0:div> <ns0:div><ns0:head>Self-Efficacy</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree)</ns0:p></ns0:div> <ns0:div><ns0:head>1</ns0:head><ns0:p>I am able to perform the underlined proposal to prevent the infection of COVID-19. 2</ns0:p><ns0:p>It is easy to perform the underlined proposal to prevent the infection of COVID-19. 3 I can perform the underlined proposal to prevent the infection of COVID-19.</ns0:p></ns0:div> <ns0:div><ns0:head>Response Efficacy</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree) <ns0:ref type='bibr' target='#b6'>4</ns0:ref> Performing the underlined proposal prevents the infection of COVID-19. 5</ns0:p><ns0:p>Performing the underlined proposal works in deterring COVID-19. 6</ns0:p><ns0:p>Performing the underlined proposal is effective in getting rid of COVID-19.</ns0:p></ns0:div> <ns0:div><ns0:head>Perceived Susceptibility</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree)</ns0:p></ns0:div> <ns0:div><ns0:head>7</ns0:head><ns0:p>I am at risk of being infected with COVID-19. 8</ns0:p><ns0:p>It is possible that I will get infected with COVID-19. 9</ns0:p><ns0:p>I am susceptible to COVID-19 infection. Severity</ns0:p><ns0:p>(1-strongly disagree, 7-strongly agree) 10 COVID-19 is a serious threat. 11 COVID-19 is harmful. 12 COVID-19 is a severe threat.</ns0:p></ns0:div> <ns0:div><ns0:head>Behavior Intention</ns0:head><ns0:p>(1-strongly disagree, 7-strongly agree) 13</ns0:p><ns0:p>In the future, when sanitizing my hands with alcohol-based hand sanitizer, I will press the pump slowly to the bottom to get a sufficient amount. As there are no other details to supplement, please refer to Section 5 for the sampling plan of the present study.</ns0:p><ns0:p>We will analyze relevant indexes in the model using SEM. Specifically, we will use the false discovery rate <ns0:ref type='bibr'>(Benjamini &amp; Hochberg, 1995)</ns0:ref> to adjust the p values of the coefficients of concern and then compare the adjusted p values with .05 to decide whether each coefficient is significant or not.</ns0:p><ns0:p>There is no evidence showing that the perceived threat of COVID-19 underlies its perceived susceptibility in the Japanese context.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>1</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='15,42.52,178.87,525.00,409.50' type='bitmap' /></ns0:figure> <ns0:note place='foot' n='2'>PeerJ reviewing PDF | (2020:07:50866:2:0:NEW 3 Oct 2020)</ns0:note> </ns0:body> "

"Graduate School of Human-Environment Studies Kyushu University 744 Motooka, Nishiku, Fukuoka 819-0395, Japan Tel/Fax: +81-92-642-2418 October 3, 2020 Dear Editors: Thank you very much for your thoughtful and constructive comments again regarding my manuscript titled, “Changing health compliance through message repetition based on the extended parallel process model in the COVID-19 pandemic.” I am very pleased with the reviewers’ constructive comments. I revised the manuscript based on all the comments. Thanks to this revision, the manuscript has improved in its more precise and complete description of details including data inclusion criteria, degrees of freedom and so on, and the quality has considerably increased. The individual responses to each of the comments are listed below. We hope that the manuscript is now suitable for publication in PeerJ. Jingwen Yang Graduate School of Human-Environment Studies, Kyushu University On behalf of all the authors. Reviewer 1 (Hamid Hassen) Basic reporting • I still have some concerns on the description in line 96-103. The declaration of WHO as pandemic alone could not be considered as the health treat by all the people to the same extent. The way people react to the this health treat (COVID-19) considerably vary according to the type of health messages communicated by each country and even specific contexts. For instance, the health messages in Belgium are different from the messages in Uganda or Ethiopia, which in turn leads variation in the influence on peoples. I agree that the government measures could be the same as WHO standard. However, the health messages and respective reactions significantly varies within and across countries. Reply: Thank you for your comments. We agree with you in that people may react to the COVID-19 pandemic in various ways, depending on the health messages given by each country in specific contexts. Therefore, we have modified lines 95–98 in the manuscript to take this premise into account by pointing out that COVID-19 is considered immediate health threat only in part of the countries or areas including Japan, where we will conduct the study. • The formulation of the hypothesis is improved. I still have minor issues. The H1 is one sided while the Ho is two sided. For instance, if perceived efficacy has a negative effect on self-efficacy, what conclusions would be made? This is the same for some of the hypotheses. Reply: Based on the findings and theories of previous research, it is rather unlikely that the other direction (i.e., negative effects in Hypothesis 1–5, 7) would happen, and thus it is difficult to consider the explanations or conclusions in these occasions. If there turns out to be any negative effect, we will discuss other possibilities according to the results of the post hoc analyses and the actual infection rate during the questionnaire period. Of course, we will never make a conclusion based on the post hoc analyses and observations. Experimental design • Good to add the degree of freedom. However, not clear why select 70 as df? Since it is a protocol, it is better if you add a description to justify it. Reply: The degree of freedom (df) is established based on the model to test. Since we will test the model in Figure 1, the df is fixed to be 70. We have added a brief explanation on the calculation of df after the sentence where it is mentioned in the manuscript. Please refer to lines 204–207 for details. • I suggest to put age as relevant characteristics as inclusion criteria. Reply: We have added age as one of the inclusion criteria. According to the protocol approved by the ethics committee, the participants in our study have to be male / female over 18 years old. Thus, we will only include those who meet this criterion for data analysis. For related revision, please refer to lines 223–228, 254–255. Validity of the findings Not applicable Comments for the author Thank you for the opportunity to review this manuscript once again. Overall, the manuscript has significantly improved. The preliminary experiment substantiates the need for the present study. I have a few minor comments. Reply: Thanks to your valuable comments, our manuscript has been considerably improved. Reviewer 2 Basic reporting n/a Experimental design Overview: Thank you for revising your report. I still maintain that H1-4 are not true hypotheses, as the outcome variables are constructed from the dependent variables. Related coefficients are not included in supplementary 1, Figure 1’. I leave it to the authors to consider when submitting their results. Some examples of this relationship: Abstract: Here, you claim that “perceived threat of COVID-19, which underlies perceived susceptibility”, so why test this in H4? Lines 64-65: If perceived efficacy is composed of self-efficacy and response efficacy, I still don’t understand why their relationships is being tested in Hypotheses 1 & 2. You could use these hypotheses as positive controls however. Reply: Thank you for your suggestions. We have added the coefficients regarding H1– 4 in supplementary 1, Figure 1’. It may seem that H1–4 are the basis of the present study and should not be tested again. However, we include them in the hypotheses for the following reason. As is shown in Table 1, the contents of the Risk Behavior Diagnosis Scale (RBDS) vary according to the disease and the corresponding health proposal which are used in the study. As far as we know, there is still no research using the RBDS to investigate neither the attitudes towards COVID-19 nor certain health proposals to prevent it up to now. Since we are not sure whether there will be some properties of COVID-19 making it different from other diseases so that the RBDS is not applicable anymore, we consider it appropriate and safe to test the scale’s validity again to ensure that COVID-19 fits in the RBDS as well. Validity of the findings n/a Comments for the author Line 187: While the term “ social distancing” has gained ground, it is at base incorrect. Please change the term to “physical distancing”. Reply: Thanks for pointing it out. We have corrected “social distancing” to “physical distancing”. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: Lung cancer is the leading cause of cancer-related deaths worldwide. Lung adenocarcinoma (LUAD) is one of the main subtypes of lung cancer. Hundreds of metabolic genes are altered consistently in LUAD; however, their prognostic role remains to be explored. This study aimed to establish a molecular signature that can predict the prognosis in patients with LUAD based on metabolic gene expression. Methods: The transcriptome expression profiles and corresponding clinical information of LUAD were obtained from the Cancer Genome Atlas and Gene Expression Omnibus databases. The differentially expressed genes (DEGs) between LUAD and paired non-tumor samples were identified by the Wilcoxon rank sum test. Univariate Cox regression analysis and the lasso Cox regression model were used to construct the best-prognosis molecular signature. A nomogram was established comprising the prognostic model for predicting overall survival. To validate the prognostic ability of the molecular signature and the nomogram, the Kaplan-Meier survival analysis, Cox proportional hazards model, and receiver operating characteristic analysis were used. Results: The six-gene molecular signature (PFKP, PKM, TPI1, LDHA, PTGES, and TYMS) from the DEGs was constructed to predict the prognosis. The molecular signature demonstrated a robust independent prognostic ability in the training and validation sets. The nomogram including the prognostic model had a greater predictive accuracy than previous systems. Furthermore, a gene set enrichment analysis revealed several significantly enriched metabolic pathways, which suggests a correlation of the molecular signature with metabolic systems and may help explain the underlying mechanisms. Conclusions: Our study identified a novel six-gene metabolic signature for LUAD prognosis prediction. The molecular signature could reflect the dysregulated metabolic microenvironment, provide potential biomarkers for predicting prognosis, and indicate potential novel metabolic molecular-targeted therapies.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Lung cancer is the leading cause of cancer-related deaths worldwide, accounting for nearly 20% of all cancer deaths <ns0:ref type='bibr' target='#b4'>(Bray et al., 2018)</ns0:ref>. Lung adenocarcinoma (LUAD) is one of the main subtypes of lung cancer <ns0:ref type='bibr' target='#b42'>(Travis, 2020)</ns0:ref>, accounting for more than 40% of lung cancer cases <ns0:ref type='bibr' target='#b19'>(Hutchinson et al., 2019)</ns0:ref>, and its relative frequency is increasing <ns0:ref type='bibr' target='#b44'>(Twardella et al., 2018)</ns0:ref>. Despite great improvements in the treatment of LUAD, the prognosis in patients with LUAD remains poor owing to the lack of early detection and effective individual therapies <ns0:ref type='bibr' target='#b10'>(Dolly et al., 2017)</ns0:ref>. Therefore, exploring prognostic biomarkers is a critical need to help predict prognosis in LUAD and to design individual therapies. Until now, most prognostic models were based on clinical characteristics (e.g., age, sex, TNM stage, vascular tumor invasion, and organization classification) or a single molecular biomarker, such as carcinoembryonic antigen and epidermal growth factor receptor. However, these prognostic models have limited power for predicting prognosis because of the complicated molecular mechanisms of LUAD development and progression. Therefore, it is important to explore the mechanism of LUAD pathology in more depth using bioinformatics to construct prognostic models that predict the patient' prognosis more accurately.</ns0:p><ns0:p>Metabolic reprogramming is one of the hallmarks of cancer <ns0:ref type='bibr' target='#b14'>(Faubert et al., 2020)</ns0:ref>, which takes place from the onset and throughout the development of cancer <ns0:ref type='bibr' target='#b6'>(Chang, Fang &amp; Gu, 2020)</ns0:ref>. It plays an important role in the progression, metastasis, depressed immunity, and therapy resistance of cancer <ns0:ref type='bibr' target='#b27'>(Lane et al., 2019)</ns0:ref>. Metabolic reprogramming has been widely accepted as the basis for the discovery of novel tumor biomarkers. <ns0:ref type='bibr' target='#b35'>Satriano et al. (2019)</ns0:ref> observed that metabolic rearrangement played an important role in predicting the prognosis in patients with primary liver cancers. <ns0:ref type='bibr' target='#b7'>Chen et al. (2019)</ns0:ref> revealed that reprogrammed tumor glucose metabolism could promote cancer stemness and result in poor prognosis in breast cancer patients. There are hundreds of metabolic genes that consistently have an altered expression in LUAD <ns0:ref type='bibr' target='#b1'>(Asavasupreechar et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b46'>Vanhove et al., 2019)</ns0:ref>; however, their roles and mechanisms of action remain unclear. This study investigated the role of abnormal metabolism in predicting the prognosis in patients with LUAD.</ns0:p><ns0:p>With the development of genome sequencing and bioinformatics, new data have emerged. Prognosis-related gene signatures that were constructed using these new tools have made great contributions to tumor prognosis prediction. This study aimed to use bioinformatic methods to establish a prognostic metabolic-gene molecular model that can predict prognosis in patients with LUAD. This model could potentially guide personalized therapy for such patients.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data expression datasets</ns0:head><ns0:p>The transcriptome expression profiles and corresponding clinical information for LUAD were downloaded from the Cancer Genome Atlas (TCGA; http://portal.gdc.cancer.gov/) and Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) databases. From the TCGA, gene expression data were of the HTSeq-FPKM type, obtained from 497 LUAD and 54 nontumor samples. From the GEO, the GSE68465 dataset included 443 LUAD and 19 non-tumor samples, using the GPL96 platform (Affymetrix Human Genome U133A Array). The metabolic genes in the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway were extracted from Gene Set Enrichment Analysis (GSEA) (http://software.broadinstitute.org/gsea/index.jsp/), and the overlapping metabolism-related genes were identified from TCGA and GSE68465 <ns0:ref type='bibr' target='#b33'>(Possemato et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b54'>Zhu et al., 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Construction and validation of the prognostic metabolic gene signature</ns0:head><ns0:p>The clinical cases from the TCGA database were used to assess the prognostic associations of the metabolic genes with clinical outcomes. The differentially expressed genes (DEGs) between LUAD and paired non-tumor samples were obtained by the Wilcoxon rank sum test using the R package called 'limma', and the adjusted P-value &lt; 0.05 and absolute log2 fold change (FC) &gt;1 were considered as the selection criterion. Univariate Cox regression analysis was used to identify prognosis-related metabolic genes, and adjusted P-values &lt; 0.001 were considered statistically significant. The lasso penalty for Cox proportional hazards model (1,000 iterations) was used to construct the prognostic gene-expression signature utilizing an R package called 'glmnet.' The prognostic gene-expression signature was designed using a risk scoring method with the following formula:</ns0:p><ns0:formula xml:id='formula_0'>&#119929;&#119946;&#119956;&#119948; &#119956;&#119940;&#119952;&#119955;&#119942; = &#119951; &#8721; &#119946; (&#119961; &#119946; * &#120631; &#119946; )</ns0:formula><ns0:p>where x i indicates the expression of gene i and &#946; i indicates the coefficient of gene i generated from the Cox multivariate regression. The R package 'survminer' was used to explore the cutoff point of the risk score, which divided patients into high-and low-risk groups. The R package 'survival' was used to draw the Kaplan-Meier survival curves to demonstrate the overall survival (OS) in the high-and low-risk groups. The R package 'survival ROC' was used to evaluate the prognostic value of the gene-expression signature.</ns0:p></ns0:div> <ns0:div><ns0:head>Independence of the prognostic gene signature from other clinical characteristics</ns0:head><ns0:p>To determine whether the predictive power of the prognostic gene-expression signature could be independent from other clinicopathological variables in patients with LUAD (including age, sex, TNM stage, T stage, N stage, and M stage), univariate and multivariate Cox regression analyses were performed. The hazard ratio (HR), 95% confidence intervals (Cis), and P-values were calculated.</ns0:p></ns0:div> <ns0:div><ns0:head>Construction and validation of a predictive nomogram</ns0:head><ns0:p>The nomogram was constructed using all the independent prognostic factors of the Cox regression analyses using R package 'rms.' Validation of the nomogram was assessed by discrimination and calibration using the concordance index (C-index) by <ns0:ref type='bibr'>Harrell et al. (1996)</ns0:ref> (bootstraps with 1,000 resamples) and the calibration plot, respectively.</ns0:p></ns0:div> <ns0:div><ns0:head>External validation of the prognostic metabolic gene signature</ns0:head><ns0:p>To verify the prognostic metabolic-gene molecular signature in the GEO dataset, the risk score of patients was calculated directly with the gene-expression signature constructed from the TCGA dataset for further analysis. The receiver operating characteristic (ROC) and Kaplan-Meier analyses were performed identically with the gene signature in the TCGA dataset. The mRNA expression levels of the signature genes were analyzed further using online databases (the Oncomine database [http://www.oncomine.org/] and TIMER database [http://cistrome.shinyapps.io/timer/]). The protein expression levels associated with the signature genes were validated using the Human Protein Atlas database (http://www.proteinatlas.org/). The known genetic alterations of the signature genes were investigated using cBioPortal for Cancer Genomics (http://www.cbioportal.org/).</ns0:p></ns0:div> <ns0:div><ns0:head>Gene Set Enrichment Analysis</ns0:head><ns0:p>Enrichment analysis of the KEGG pathways of the signature genes was performed using GSEA on the TCGA dataset. The nominal (NOM) P-value &lt; 0.05 and the False Discovery Rate (FDR) q-value &lt; 0.25 indicated statistical significance.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>All analyses were performed using R software v3.6.3 (R Foundation for Statistical Computing, Vienna, Austria). Two-tailed P-values &lt; 0.05 were considered statistically significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Clinical characteristics</ns0:head><ns0:p>The TCGA dataset included 486 patients with LUAD (Table <ns0:ref type='table'>S1</ns0:ref>). The GEO dataset included 443 patients with LUAD (Table <ns0:ref type='table'>S1</ns0:ref>). Patients with a survival time of less than 30 days were omitted. For the study, 454 and 439 patients remained in the TCGA and GEO datasets, respectively. The detailed clinical characteristics of all patients are listed in Table <ns0:ref type='table'>1</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Building and validation of the prognostic metabolic gene signature</ns0:head><ns0:p>To clarify our study design, a flow chart of the analysis procedure is presented in Fig. <ns0:ref type='figure'>1</ns0:ref>. A list of 994 genes in the KEGG pathway was identified from GSEA (Table <ns0:ref type='table'>S2</ns0:ref>), and 633 overlapping metabolism-related genes were abstracted from TCGA and GSE68465 (Table <ns0:ref type='table'>S3</ns0:ref>). The 96 DEGs (72 up-regulated genes and 24 download genes) between LUAD and paired non-tumor samples were identified from the further analysis (Fig. <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>; Table <ns0:ref type='table'>S4</ns0:ref>). Seven significant genes associated with OS were identified using univariate analysis (Table <ns0:ref type='table'>S4</ns0:ref>). Furthermore, six genes were selected to build the prognostic model using a lasso-penalized Cox analysis (Table <ns0:ref type='table'>2</ns0:ref>). The six genes were phosphofructokinase platelet (PFKP), pyruvate kinase muscle (PKM), triosephosphate isomerase 1 (TPI1), lactate dehydrogenase A (LDHA), prostaglandin E synthase (PTGES), and thymidylate synthase (TYMS). Risk score = (0.00005&#215; PFKP mRNA level) + (0.00173&#215; PKM mRNA level) + (0.00038&#215; TPI1 mRNA level) + (0.00379&#215; LDHA mRNA level) + (0.00292 &#215; PTGES mRNA level) + (0.02490&#215; TYMS mRNA level).</ns0:p><ns0:p>The 445 patients with LUAD were divided into the high-risk or low-risk group based on the median risk score of 0.861 in the TCGA dataset. Patients in the high-risk group had significantly poorer OS than those in the low-risk group (P &lt; 0.001; Fig. <ns0:ref type='figure' target='#fig_1'>3A</ns0:ref>). The distribution of the risk score and survival status of the patients is presented in Fig. <ns0:ref type='figure' target='#fig_1'>3C</ns0:ref>, which showed a higher mortality in the high-risk group than in the low-risk group. The expression of the six prognostic genes is shown in the heatmap. All the six genes had a significant positive correlation with the high-risk group (Fig. <ns0:ref type='figure' target='#fig_1'>3E</ns0:ref>). The area under the curve (AUC) of the time-dependent ROC curve was used to identify the prognostic ability of the six-gene molecular signature. The AUCs of the six-gene signature model were 0.693, 0.655, and 0.565 for the 1-, 3-, and 5-year OS, respectively, suggesting that the prediction model had a good performance in predicting the OS in patients with LUAD (Fig. <ns0:ref type='figure' target='#fig_1'>3G</ns0:ref>).</ns0:p><ns0:p>The prognostic model was validated in the GSE68465 dataset. The 439 patients with LUAD were divided into the high-risk or low-risk group based on the median risk score of 0.861. Patients in the high-risk group had a poor OS compared with those in the low-risk group (P &lt; 0.001; Fig. <ns0:ref type='figure' target='#fig_1'>3B</ns0:ref>). The distribution of the risk score and survival status showed a higher mortality in the high-risk group than in the low-risk group (Fig. <ns0:ref type='figure' target='#fig_1'>3D</ns0:ref>). The expression heatmap of the six prognostic genes showed that all the six genes had a significant positive correlation with the high-risk group (Fig. <ns0:ref type='figure' target='#fig_1'>3F</ns0:ref>). The AUCs of the six-gene signature model were 0.728, 0.654, and 0.618 for the 1-, 3-, and 5-year OS, respectively (Fig. <ns0:ref type='figure' target='#fig_1'>3H</ns0:ref>). Taken together, these results suggested that the prognostic model had a high sensitivity and specificity in predicting the OS in patients with LUAD.</ns0:p><ns0:p>The prognostic gene signature was independent from other clinicopathological factors Univariate and multivariate Cox regression analyses were conducted to assess the independent predictive value of the six-gene prognostic signature. In the TCGA dataset, univariate Cox regression analysis demonstrated that the prognostic model (HR: 2.845, P &lt; 0.001), TNM stage (HR: 1.666, P &lt; 0.001), T stage (HR: 1.605, P &lt; 0.001), and N stage (HR: 1.806, P &lt; 0.001) had a prognostic value for OS (Fig. <ns0:ref type='figure' target='#fig_2'>4A</ns0:ref>). Multivariate Cox regression analysis demonstrated that the only prognostic model (HR: 2.448, P &lt; 0.001) and TNM stage (HR: 1.950, P &lt; 0.01) were independent prognostic factors for OS (Fig. <ns0:ref type='figure' target='#fig_2'>4A</ns0:ref>). In the GSE68465 dataset, the prognostic model, T stage, N stage, and age had a prognostic value in the univariate and multivariate Cox regression analyses (Fig. <ns0:ref type='figure' target='#fig_2'>4B</ns0:ref>). Gender was the only independent prognostic factor for OS in the univariate Cox regression analysis (Fig. <ns0:ref type='figure' target='#fig_2'>4B</ns0:ref>).</ns0:p><ns0:p>In addition, the time-dependent ROC curve was used to identify the predictive ability of the prognostic model compared with the other clinicopathological characteristics. In the TCGA dataset, the AUCs of the prognostic model were 0.693, 0.655, and 0.565 for the 1-, 3-, and 5-year OS, respectively, which were higher than most of the other clinicopathological characteristics including age (0.498, 0.511, 0.485), gender (0.579, 0.485, 0.451), T stage (0.673, 0.613, 0.608), N stage (0.685, 0.666, 0.628), and M stage (0.508, 0.527, 0.530) (Fig. <ns0:ref type='figure' target='#fig_3'>5A</ns0:ref>). Furthermore, in the GSE68465 dataset, the AUCs of the prognostic model were 0.728, 0.654, and 0.618 for the 1-, 3-, and 5-year OS, respectively, which were higher than most of the other clinicopathological characteristics including age (0.593, 0.568, 0.581), gender (0.539, 0.549, 0.547), grade (0.580, 0.571, 0.548), T stage (0.647, 0.606, 0.606), and N stage (0.690, 0.680, 0.655) (Fig. <ns0:ref type='figure' target='#fig_3'>5B</ns0:ref>). The prognostic model had a larger AUC value compared with other clinicopathological characteristics. These results indicated that the model was an excellent prognostic model for LAUD patients, especially for the 1-and 3-year OS.</ns0:p><ns0:p>These results suggested that our prognostic model could be an independent predictor of prognosis in patients with LAUD.</ns0:p></ns0:div> <ns0:div><ns0:head>Building and validating a predictive nomogram</ns0:head><ns0:p>A nomogram was built to predict the survival probability in patients with LAUD in the TCGA dataset. The nomogram was constructed using four prognostic factors (the TNM stage, T stage, N stage, and prognostic model; Fig. <ns0:ref type='figure' target='#fig_5'>6A</ns0:ref>). The C-index was calculated to evaluate the predictive ability of the nomogram for OS. The C-index for the nomogram was 0.754 (95% CI: 0.561-0.947). Calibration plots indicated that the nomogram had a good accuracy in predicting the 1and 3-year OS (Fig. <ns0:ref type='figure' target='#fig_5'>6B</ns0:ref>).</ns0:p><ns0:p>To predict the survival probability more accurately, the combined prognostic model was built based on the nomogram. The combined prognostic model consisted of the TNM stage, T stage, N stage, and prognostic model. A time-dependent ROC curve was used to identify the predictive ability of the combined prognostic model. The AUCs of the combined prognostic models were 0.782, 0.717, and 0.688 for the 1-, 3-, and 5-year OS, respectively, which were higher than other clinical models including the TNM stage model (0.732, 0.687, 0.681), T stage model (0.671, 0.612, 0.613), N stage model (0.686, 0.661, 0.648), and the prognostic model (0.692, 0.634, 0.576). The combined model had the largest AUC value compared with other factors, which indicated that the combined model had a good predictive accuracy for survival. These results suggested that the predictive ability of the combined model built with the nomograms is better than other models, especially for predicting 1-and 3-year survival (Fig. <ns0:ref type='figure' target='#fig_5'>6C</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Gene Set Enrichment Analysis</ns0:head><ns0:p>To recognize signaling pathways that are differentially activated in LUAD, a GSEA was used, and a total of 49 significantly enriched KEGG pathways were found in the high-risk group and low-risk group (Table <ns0:ref type='table'>S5</ns0:ref>) of the TCGA dataset (FDR q-val &lt; 0.25, NOM p-val &lt; 0.05). Among them, many enriched pathways were related to metabolism and some highly dysregulated pathways including cell cycle, p53 signaling pathway, and basal transcription factors were also contained in these results (Table <ns0:ref type='table'>S5</ns0:ref>). We chose the top five significantly enriched metabolismsignaling pathways depending on the normalized enrichment score from the high-risk group or low-risk group. We found that the top five most significantly enriched metabolism-related pathways of the high-risk group were the cysteine and methionine, fructose and mannose, glyoxylate and dicarboxylate, purine, and pyrimidine pathways (Fig. <ns0:ref type='figure'>7A</ns0:ref>). The top five most significantly enriched metabolism-related pathways of the low-risk group were the alpha linolenic acid, arachidonic acid, ether lipid, glycerophospholipid, and linoleic acid pathways (Fig. <ns0:ref type='figure'>7B</ns0:ref>). Most of the metabolism-related pathways in the high-risk group mainly focused on amino acid and glycolysis metabolism, while the pathways in the low-risk group mainly focused on lipid metabolism. The results of the ten representative enriched metabolism-related KEGG pathways are given in Table <ns0:ref type='table'>3</ns0:ref>. Furthermore, all the six metabolic genes of the prognostic model enriched these metabolism pathways significantly. LDHA enriched the cysteine and methionine pathway (Table <ns0:ref type='table'>S6</ns0:ref>); PFKP and TPI1 enriched the fructose and mannose pathway (Table <ns0:ref type='table'>S6</ns0:ref>); PKM enriched the purine pathway (Table <ns0:ref type='table'>S6</ns0:ref>); TYMS enriched the pyrimidine pathway (Table <ns0:ref type='table'>S6</ns0:ref>); and PTGES enriched the arachidonic acid pathway (Table <ns0:ref type='table'>S6</ns0:ref>). The results further elucidated the role of metabolism in LUAD and the value of the six-gene signature in predicting the prognosis of LUAD.</ns0:p></ns0:div> <ns0:div><ns0:head>External validation using online databases</ns0:head><ns0:p>To further identify the role of the six metabolic genes in LUAD, we compared the mRNA expression levels of the six metabolic genes (PFKP, PKM, TPI1, LDHA, PTGES, and TYMS) in the LAUD tissues with those in the normal lung tissues using data from the Oncomine database (Fig. <ns0:ref type='figure' target='#fig_7'>8</ns0:ref>). Obviously, all the six genes were overexpressed in lung cancer in all the datasets from the Oncomine database with the threshold of fold change=2, P-value=0.001(Fig. <ns0:ref type='figure' target='#fig_7'>8A</ns0:ref>). Furthermore, the mRNA levels of all the six genes in LUAD were significantly upregulated than those in normal tissues in the combined LUAD datasets from the Oncomine database (Fig. <ns0:ref type='figure' target='#fig_7'>8B</ns0:ref>; Table <ns0:ref type='table'>4</ns0:ref>). To further validate the overexpression of the six genes in LUAD, we analyzed the expression of the six genes using TIMER databases (Fig. <ns0:ref type='figure' target='#fig_8'>9</ns0:ref>). The results revealed that all the mRNA expression of the six genes in LUAD were significantly higher than in normal tissues. All the results from the Oncomine and TIMER databases were consistent with our results for the TCGA and GEO datasets. In addition, the mRNA expression of the six genes was also higher in esophageal carcinoma, head and neck squamous cell carcinoma, lung squamous cell carcinoma, and stomach adenocarcinoma from the TIMER databases (Fig. <ns0:ref type='figure' target='#fig_8'>9</ns0:ref>). The protein expressions of these six genes were analyzed using clinical specimens from the Human Protein Profiles (Fig. <ns0:ref type='figure' target='#fig_9'>10A and 10B</ns0:ref>; Table <ns0:ref type='table'>5</ns0:ref>). The representative images of the six gene protein levels from the Human Protein Profiles are shown in Fig. <ns0:ref type='figure' target='#fig_9'>10A</ns0:ref>. Compared with the expression level in normal lung tissue, LDHA (100%, n=7) and TYMS (80%, n=5) showed a significantly higher percentage of high/medium expression levels in the LAUD tissue (Fig. <ns0:ref type='figure' target='#fig_9'>10B</ns0:ref>; Table <ns0:ref type='table'>5</ns0:ref>). PKM (50%, n=6), PFKP (33.33%, n=6), and PTGES (16.67%, n=6) showed a significant moderate percentage of high/medium expression levels in the LAUD tissue (Fig. <ns0:ref type='figure' target='#fig_9'>10B</ns0:ref>; Table <ns0:ref type='table'>5</ns0:ref>). However, TPI1 showed no detected expression both in the LAUD and normal lung tissue (Fig. <ns0:ref type='figure' target='#fig_9'>10B</ns0:ref>; Table <ns0:ref type='table'>5</ns0:ref>). The genetic alterations were explored in the cBioPortal database. Amplifications and mutations were the most common alterations in the six metabolic genes (Fig. <ns0:ref type='figure' target='#fig_9'>10C</ns0:ref>). The aberrant genetic alterations might elucidate the overexpression of these six genes in LUAD.</ns0:p><ns0:p>Altogether, the correlation of the aberrant expression of these six genes with LAUD cancer was further validated using multiple online databases.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>LUAD is the most common histological subtype of primary lung cancer. The incidence of LUAD has been increasing rapidly, and mortality has not significantly decreased despite great improvements in research and treatment. Therefore, exploring the molecular mechanisms of LUAD progression and constructing a valid and accurate molecule-based tool for evaluating the prognosis in patients is urgently needed. This could help design more efficient therapeutic strategies for LUAD. Metabolic reprogramming in cancers could lead to their development and progression <ns0:ref type='bibr' target='#b32'>(Nwosu et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b28'>Liu et al., 2020)</ns0:ref>. Characterization of the changes in metabolic gene expression in LUAD would allow development of novel prognostic biomarkers. However, a single biomarker is not a robust measure for predicting patient prognosis. Thus, constructing a robust multiple-biomarker signature for predicting the prognosis in cancer patients is necessary.</ns0:p><ns0:p>We identified and designed a novel six-gene prognostic molecular signature based on the TCGA database and validated its efficiency in the GSE68465 dataset. The results indicated that the molecular signature was significantly associated with OS in patients with LUAD in the training and validation sets. These results indicate that the molecular signature has a robust prognostic value, especially for predicting short-term survival in patients with LUAD. These results also demonstrated that the prognostic signature was independent of other clinicopathological characteristics, which further supports the prognostic value of this signature.</ns0:p><ns0:p>To increase the accuracy of the prediction of prognosis, we constructed a nomogram built with the combination of genetic and clinically related variables of patients with LUAD. The nomogram included the prognostic model, TNM, T stage, and N stage. Its predictive accuracy was verified using calibration plots, the C-index, and the AUC, which indicated that the nomogram had a greater predictive value than the previous systems. The Gene Set Enrichment Analysis showed that many significantly enriched pathways were metabolism-related pathways. The different risk groups possessed different metabolic pathway features. The metabolismrelated pathways in the high-risk group were mainly associated with amino acid and glycolysis metabolism, while the pathways in the low-risk group were mainly associated with lipid metabolism. These results revealed that the different risk groups possessed the different metabolic features, which might provide the underlying metabolic mechanisms of promoting the prognosis of LUAD. All these results further suggest a strong association between the molecular signature and metabolic systems and might reflect the dysregulated metabolic microenvironment of cancers.</ns0:p><ns0:p>Most of the six genes in our prognostic signature are suggested to be related to cancer development. PFKP is a major isoform of cancer-specific phosphofructokinase-1, an enzyme that catalyzes the phosphorylation of fructose-6-phosphate to form fructose-1,6-bisphosphate. Recently, PFKP was noted to have an aberrant upregulation in many cancers, such as breast cancer, prostate cancer, and glioblastoma. The dynamic upregulation of PFKP promotes metabolic reprogramming and cancer cell survival <ns0:ref type='bibr' target='#b3'>(Bjerre et at., 2019;</ns0:ref><ns0:ref type='bibr' target='#b21'>Kim et al., 2017)</ns0:ref>. As a key regulator enzyme in glycolysis, PFKP enriched the fructose and mannose metabolism pathway. Recent studies showed that PFKP is highly expressed in lung cancer and promotes lung cancer development via fructose and mannose metabolism <ns0:ref type='bibr' target='#b39'>(Shen et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b49'>Wang, et al., 2015)</ns0:ref>. PKM is a rate-limiting enzyme in the final step of glycolysis, that is considered as one of the metabolic hallmarks of cancer <ns0:ref type='bibr' target='#b34'>(Prakasam et al., 2017)</ns0:ref>. The abnormal expression of PKM promoted cancer growth, invasion, and metastasis by governing aerobic glycolysis <ns0:ref type='bibr' target='#b34'>(Prakasam et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b53'>Zahra et al., 2020)</ns0:ref> and induced cancer treatment resistance <ns0:ref type='bibr' target='#b5'>(Calabretta et al., 2016)</ns0:ref>. Furthermore, PKM is overexpressed in non-small cell lung cancer (NSCLC) and involved in the development and prognosis of NSCLC <ns0:ref type='bibr' target='#b30'>(Luo et al., 2018)</ns0:ref>. TPI1 is a crucial enzyme in carbohydrate metabolism, catalyzing the interconversion of dihydroxyacetone phosphate and d-glyceraldehyde-3-phosphate during glycolysis and gluconeogenesis. TPI1 is abnormally expressed in different kinds of cancers, such as breast cancer, gastric cancer, and lymphoma and is associated with a poor prognosis in patients with neuroblastoma and pancreatic cancer through dysregulating glycometabolism <ns0:ref type='bibr' target='#b29'>(Ludvigsen et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b0'>Applebaum et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b15'>Follia et al., 2019)</ns0:ref>. LDHA is an enzyme that catalyzes the interconversion of pyruvate and lactate. LDHA was enriched in cysteine and methionine metabolism, and its aberrant metabolism regulation promoted many pathological processes in tumors, such as cell proliferation, survival, invasion, metastasis, and immunity <ns0:ref type='bibr'>(Dorneburg et al., 2018)</ns0:ref>. Overexpressed LDHA is associated with poor prognosis in many tumors, including NSCLC, breast cancer, gallbladder carcinoma, and gastrointestinal cancer <ns0:ref type='bibr' target='#b31'>(Mizuno et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b17'>Guddeti et al., 2019)</ns0:ref>. PTGES is a key enzyme in the arachidonic acid metabolism pathway. An abnormally high expression of PTGES is correlated with proliferation, invasion, and metastasis in many cancer cells <ns0:ref type='bibr' target='#b23'>(Kim et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b8'>Delgado-Go&#241;i et al., 2020)</ns0:ref>. The dysregulated PTGES promoted tumor migration and metastasis of lung cancer cells and played an important role in lung cancer progression <ns0:ref type='bibr' target='#b48'>(Wang et al., 2019)</ns0:ref>. TYMS is a rate-limiting enzyme, which plays an important role in regulating the pyrimidine metabolism signaling pathway <ns0:ref type='bibr' target='#b52'>(Yeh et al., 2017)</ns0:ref>. TYMS is overexpressed frequently in different kinds of cancers, such as NSCLC, pancreatic, colorectal, and breast cancers, and it has resulted in a poor cancer prognosis and chemotherapy resistance via dysregulating pyrimidine metabolism <ns0:ref type='bibr' target='#b43'>(Troncarelli Flores et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b50'>Wu et al., 2019)</ns0:ref>. In our study, we constructed a six-gene signature for a prognostic model based on the TCGA database. This novel six-gene signature had a higher survival prediction, and the predictive ability of this signature was further validated by the GSE68465 dataset and multiple online databases. To our knowledge, the six-gene signature for prognosis prediction in LUAD has not been reported yet. Compared with the traditional prognostic models such as clinical characteristics (e.g., TNM stage, vascular tumor invasion, and organization classification) or a single molecular biomarker, a multi-gene signature can predict the prognosis more accurately and provide a clearer molecular mechanism for personalized LUAD therapy.</ns0:p><ns0:p>There are limitations in our study. First, our nomogram was not validated further in the GEO database because the GSE68465 lacked detailed TNM stage data. Thus, the nomogram should be externally validated using larger datasets from multicenter clinical trials and perspective studies. Second, functional experiments should be further performed to explore the molecular mechanisms predicted by the metabolic gene expression.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>We concluded from our research results that the six-gene metabolic prognostic signature could accurately predict the prognosis in patients with LUAD. The molecular signature may provide potential biomarkers for metabolic therapy and prognosis prediction of LUAD.</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 1</ns0:head><ns0:p>Overall flowchart of steps used in the construction of the prognostic metabolic gene signature.</ns0:p><ns0:p>The TCGA dataset was utilized to construct the prognostic metabolic gene signature. The TCGA clinical information, GSE68465 dataset and online databases from international platforms were further utilized to validate the prognostic model. TCGA, The Cancer Genome Atlas; OS, overall survival; ROC, the receiver operating characteristic.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51200:1:0:NEW 10 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Owing to only one dataset meeting the screening criteria, the comparison of PKM or TPI1 expression in LUAD and normal has not been built based on the combined LUAD datasets. P value &lt; 0.001 and fold change &gt; 2 were utilized for screening. LUAD, lung adenocarcinoma.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>(A) The time-dependent ROC curves of risk score, age, gender, TNM stage, T stage, N stage, and M stage in the TCGA dataset. (B) The time-dependent ROC curves of risk score, age, gender, grade, T stage, and N stage in the GEO dataset. LUAD, lung adenocarcinoma.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) The nomogram was built in the TCGA dataset. (B) Calibration plots revealed the nomogram-predicted survival probabilities. (C) The time-dependent ROC analysis evaluated the accuracy of the nomogram. TCGA, The Cancer Genome Atlas; ROC, receiver operating characteristic; LUAD, lung adenocarcinoma.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 8 mRNA</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 9 mRNA</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 10 Protein</ns0:head><ns0:label>10</ns0:label><ns0:figDesc>Figure 10</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,301.12,525.00,320.25' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>cbioportal.org/ ). PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES,</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>Characteristics</ns0:cell><ns0:cell cols='2'>TCGA (n, %)</ns0:cell><ns0:cell /><ns0:cell cols='2'>GSE68465 (n, %)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(n=454)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>(n=439)</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#65308;60</ns0:cell><ns0:cell cols='2'>133 (29.3%) LUAD (n=45) vs. Normal (n=65)</ns0:cell><ns0:cell>7.946</ns0:cell><ns0:cell>2.536</ns0:cell><ns0:cell cols='2'>128 (29.2%) 3.39E-11 Hou et al., 2010</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8805;60</ns0:cell><ns0:cell cols='2'>321 (70.7%) LUAD (n=58) vs. Normal (n=58)</ns0:cell><ns0:cell cols='2'>10.910 2.883</ns0:cell><ns0:cell cols='2'>311 (70.8%) 1.64E-17 Selamat et al., 2012</ns0:cell></ns0:row><ns0:row><ns0:cell>NA</ns0:cell><ns0:cell cols='2'>0 (0.0%) LUAD (n=20) vs. Normal (n=19)</ns0:cell><ns0:cell>5.277</ns0:cell><ns0:cell>3.149</ns0:cell><ns0:cell>0 (0.0%) 9.37E-6</ns0:cell><ns0:cell>Stearman et al., 2005</ns0:cell></ns0:row><ns0:row><ns0:cell>Gender</ns0:cell><ns0:cell cols='2'>Comparison of PFKP expression across 5</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>1.64E-17</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Female Analysis between LUAD and Normal 248 (54.6%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>218 (49.7%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Male PKM</ns0:cell><ns0:cell cols='2'>206 (45.4%) LUAD (n=58) vs. Normal (n=58)</ns0:cell><ns0:cell cols='2'>12.037 2.551</ns0:cell><ns0:cell cols='2'>221 (50.3%) 3.56E-20 Selamat et al., 2012</ns0:cell></ns0:row><ns0:row><ns0:cell>NA TPI1</ns0:cell><ns0:cell cols='2'>0 (0.0%) LUAD (n=40) vs. Normal (n=6)</ns0:cell><ns0:cell>4.929</ns0:cell><ns0:cell>2.283</ns0:cell><ns0:cell>0 (0.0%) 4.03E-4</ns0:cell><ns0:cell>Garber et al., 2001</ns0:cell></ns0:row><ns0:row><ns0:cell>Grade LDHA</ns0:cell><ns0:cell cols='2'>LUAD (n=9) vs. Normal (n=3)</ns0:cell><ns0:cell>4.502</ns0:cell><ns0:cell>4.037</ns0:cell><ns0:cell>6.29E-4</ns0:cell><ns0:cell>Yamagata et al. 2003</ns0:cell></ns0:row><ns0:row><ns0:cell>G1</ns0:cell><ns0:cell cols='2'>0 (0.0%) LUAD (n=58) vs. Normal (n=58)</ns0:cell><ns0:cell cols='2'>11.533 2.179</ns0:cell><ns0:cell cols='2'>60 (13.7%) 1.59E-19 Selamat et al., 2012</ns0:cell></ns0:row><ns0:row><ns0:cell>G2</ns0:cell><ns0:cell cols='2'>0 (0.0%) Comparison of LDHA expression across</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell cols='2'>206 (46.9%) 3.15E-4 -</ns0:cell></ns0:row><ns0:row><ns0:cell>G3</ns0:cell><ns0:cell cols='2'>0 (0.0%) 2 Analysis between LUAD and Normal</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>166 (37.8%)</ns0:cell></ns0:row><ns0:row><ns0:cell>NA PTGES</ns0:cell><ns0:cell cols='2'>454 (100%) LUAD (n=20) vs. Normal (n=19)</ns0:cell><ns0:cell>9.332</ns0:cell><ns0:cell>5.883</ns0:cell><ns0:cell>7 (1.6%) 1.54E-11</ns0:cell><ns0:cell>Stearman et al., 2005</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>TNM stage LUAD (n=40) vs. Normal (n=6)</ns0:cell><ns0:cell>6.690</ns0:cell><ns0:cell>4.969</ns0:cell><ns0:cell>1.12E-6</ns0:cell><ns0:cell>Garber et al., 2001</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8544;</ns0:cell><ns0:cell cols='2'>243 (53.5%) LUAD (n=58) vs. Normal (n=58)</ns0:cell><ns0:cell cols='2'>10.267 2.179</ns0:cell><ns0:cell>5.58E-16</ns0:cell><ns0:cell>Selamat et al., 2012</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8545;</ns0:cell><ns0:cell cols='2'>105 (23.1%) LUAD (n=45) vs. Normal (n=65)</ns0:cell><ns0:cell>6.513</ns0:cell><ns0:cell>2.170</ns0:cell><ns0:cell>6.22E-9</ns0:cell><ns0:cell>Hou et al., 2010</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8546;</ns0:cell><ns0:cell cols='2'>74 (16.3%) Comparison of PTGES expression across</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>5.62E-7</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8547;</ns0:cell><ns0:cell cols='2'>24 (5.3%) 4 Analysis between LUAD and Normal</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>NA TYMS</ns0:cell><ns0:cell cols='2'>8 (1.8%) LUAD (n=45) vs. Normal (n=65)</ns0:cell><ns0:cell>9.322</ns0:cell><ns0:cell>3.929</ns0:cell><ns0:cell cols='2'>439(100%) 6.92E-15 Hou et al., 2010</ns0:cell></ns0:row><ns0:row><ns0:cell>T stage</ns0:cell><ns0:cell cols='2'>LUAD (n=27) vs. Normal (n=30)</ns0:cell><ns0:cell>7.395</ns0:cell><ns0:cell>3.016</ns0:cell><ns0:cell>2.40E-9</ns0:cell><ns0:cell>Su et al., 2007</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>T1 prostaglandin E synthase; TYMS, thymidylate synthase; TCGA, The Cancer Genome Atlas; 156 (34.4%) 150 (34.2%) LUAD (n=58) vs. Normal (n=49) 11.169 2.797 9.86E-20 Landi et al., 2008 T2 240 (52.9%) 248 (56.5%) LUAD (n=20) vs. Normal (n=19) 6.509 2.118 2.18E-7 Stearman et al., 2005</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>T3 LUAD, lung adenocarcinoma. LUAD (n=86) vs. Normal (n=10) 37 (8.1%) T4 18 (4.0%) LUAD (n=58) vs. Normal (n=58)</ns0:cell><ns0:cell>4.191 8.565</ns0:cell><ns0:cell>2.158 2.040</ns0:cell><ns0:cell cols='2'>28 (6.4%) 3.05E-4 Beer et al., 2002 11 (2.5%) 3.35E-13 Selamat et al., 2012</ns0:cell></ns0:row><ns0:row><ns0:cell>Tx</ns0:cell><ns0:cell cols='2'>3 (0.7%) Comparison of TYMS expression across</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>2 (0.5%) 1.09E-7</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>N stage</ns0:cell><ns0:cell cols='2'>6 Analysis between LUAD and Normal</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>N0</ns0:cell><ns0:cell /><ns0:cell>291 (64.1%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>297 (67.7%)</ns0:cell></ns0:row><ns0:row><ns0:cell>N1</ns0:cell><ns0:cell /><ns0:cell>86 (18.9%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>87 (19.8%)</ns0:cell></ns0:row><ns0:row><ns0:cell>N2</ns0:cell><ns0:cell /><ns0:cell>64 (14.1%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>52 (11.8%)</ns0:cell></ns0:row><ns0:row><ns0:cell>N3</ns0:cell><ns0:cell /><ns0:cell>2 (0.4%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0 (0.0%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Nx</ns0:cell><ns0:cell /><ns0:cell>11 (2.4%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>3 (0.7%)</ns0:cell></ns0:row><ns0:row><ns0:cell>M stage</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>M0</ns0:cell><ns0:cell /><ns0:cell>305 (67.2%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>439 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>M1</ns0:cell><ns0:cell /><ns0:cell>23 (5.1%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0 (0.0%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Mx</ns0:cell><ns0:cell /><ns0:cell>126 (27.8%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0 (0.0%)</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Survival status</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Alive</ns0:cell><ns0:cell /><ns0:cell>300 (66.1%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>206 (46.9%)</ns0:cell></ns0:row><ns0:row><ns0:cell>Dead</ns0:cell><ns0:cell /><ns0:cell>154 (33.9%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell cols='2'>233 (53.1%)</ns0:cell></ns0:row></ns0:table><ns0:note>Note:</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51200:1:0:NEW 10 Oct 2020)</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51200:1:0:NEW 10 Oct 2020)Manuscript to be reviewed</ns0:note> <ns0:note place='foot'>Manuscript to be reviewed</ns0:note> </ns0:body> "

"October 10, 2020 Dear Editor and Reviewers: Thank you for all your comments on our manuscript titled “Identification of a Six-Gene Metabolic Signature Predicting Overall Survival for Patients with Lung Adenocarcinoma” (ID: 51200). These comments are all valuable and very helpful for improving our paper, as well as giving more significance to our research. We have studied your comments carefully, and we have revised our manuscript accordingly. We hope the revisions addressed all your concerns fully. In this letter, we have provided a point-by-point response to all your comments. For clarity, your comments are in italic, followed by our responses. Find all the changes in the revised manuscript highlighted in yellow. We have also provided the line numbers in the manuscript corresponding to the sections we have revised. Responses to the editor’s comments: Editor (Mirna Mourtada-Maarabouni) MINOR REVISIONS You have used a few databases and bioinformatics tools to identify a six-gene metabolic signature predicting overall survival for patients with lung adenocarcinoma. The paper is well written and present interesting findings. However, I agree with both reviewers, the presentation of some of the results needs to be improved before the paper is published. You will need to address the recommendations of the reviewers and the list below before the paper is accepted for publication: 1) State the names of the genes in the abstract. Response: Thank you for this valuable suggestion. We have stated the names of the genes in the abstract (line 52). 2) Fully describe your figures/graphs in the results section. Response: Thank you for this great suggestion on improving our manuscript. We have fully described Figure 3 (line 185-205), Figure 4 (line 208-217), Figure 5 (line 218-230), Figure 6 (line 235-251), Figure 7/Table 3 (line 254-275), and Figure 8-10/Table 4-5 (line 278-302) in the results. 3) The table showing the six genes and their expression values in normal and cancer samples should be included in the paper and not as supplementary table. Response: This is quite correct. The six genes and their expression values in the normal and cancer samples have been provided as Table 2 in the manuscript. 4) Figure 1 needs legend Response: Thank you for this helpful remark. We have provided a legend for Figure 1 as follows: Figure 1 Overall flowchart of steps used in the construction of the prognostic metabolic gene signature. The TCGA dataset was utilized to construct the prognostic metabolic gene signature. The TCGA clinical information, GSE68465 dataset, and online databases from international platforms were further utilized to validate the prognostic model. TCGA, The Cancer Genome Atlas; OS, overall survival; ROC, receiver operating characteristic. 5) Figure 3 legend is not satisfactory. You should give provide an explanatory legend and explain what each of this graphs/figure show. Panel A and B consist of five graphs each, and it is not clear what each of these figures is showing. Response: Thank for your great suggestion on improving Figure 3. We have provided a more detailed explanatory legend for Figure 3. We have adjusted the display order of ten graphs and relabeled each graph to make each of them clear in Figure 3. We have also fully described Figure 3 (line 185-205) as follows: Figure 3 Identification of the prognostic model in lung adenocarcinoma. (A, B) Kaplan-Meier curves of overall survival of the high-risk and low-risk groups stratified by the six-gene signature-based risk score in the TCGA or GEO dataset. (C, D) Risk score distribution, survival status distribution in the TCGA or GEO dataset. (E, F) The expression heatmap of the six prognostic genes in the TCGA or GEO dataset. (G, H) Time-dependent ROC curves of the six-gene signature in the TCGA or GEO dataset. TCGA, The Cancer Genome Atlas; GEO, Gene Expression Omnibus; ROC, receiver operating characteristic. 6) Figure 7 provide more explanation in the legend and briefly describe what you are showing. Response: Thank you for this helpful suggestion. We have provided a more detailed explanatory legend for Figure 7. Moreover, we have deleted the right panel of Figure 7 and shown the related data in Table 3. We have fully described Figure 7and Table 3 (line 254-275) as follows: Figure 7 The representative enriched metabolism-related KEGG pathways in the TCGA dataset by GSEA. (A) The top five significantly representative enriched metabolism-related KEGG pathways in the high-risk patients. (B) The top five significantly representative enriched metabolism-related KEGG pathways in the low-risk patients. Related parameters for the ten representative enriched metabolism-related KEGG pathways are given in Table 3. GSEA, Gene Set Enrichment Analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; TCGA, The Cancer Genome Atlas. Table 3: The results of the ten representative enriched metabolism-related KEGG pathways analysed by GSEA. 7) Figure 8: legend needs more details for both panels. Oncomine results should be presented in more accessible format not just screenshots. Can these data be extracted and presented in graphs? Response: Thank you for this valuable suggestion. We have provided a more detailed explanatory legend for Figure 8. The right panel of Figure 8A has been edited, and labeled as Figure 8B. Moreover, those comparing the mRNA data between LUAD and normal tissues from the Oncomine database have been extracted and presented in Table 4. In addition, the Figure 8B has been moved to Figure 10C for elucidating the overexpression of these six genes in LUAD more clearly. Find the relevant legend below for your quick reference: Figure 8 mRNA expression levels of the six prognostic genes from online databases. (A) mRNA expression levels of the six genes in the Oncomine database. The threshold is shown at the bottom (P value < 0.001 and fold change > 2 were utilized for screening). The figure in the colored cell represents the number of datasets complying with the threshold. The red cells indicate that the genes were overexpressed in the cancer cells, while the blue cells indicate that the genes were overexpressed in the normal tissues. (B) Comparisons of the mRNA expression levels of the six genes between LUAD and normal tissues in the combined LUAD datasets from the Oncomine database. PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES, prostaglandin E synthase; TYMS, thymidylate synthase; LUAD, lung adenocarcinoma. 8) Figure 9 legend needs more detail. You should try to extract the relevant data from this figure or try to compare the expression of these six genes to other cancers. Is the increase expression of these six genes specific to lung adenocarcinoma? Response: Thank you for this helpful suggestion. We have provided a more detailed explanatory legend for Figure 9. We have compared the expression of these six genes to other cancers and fully described the results in Figure 9 (line 285-291) in the manuscript. Fine the relevant legend below for your quick reference: Figure 9 mRNA expression levels of the six prognostic genes extracted from online database. The mRNA expression levels of the six genes in different tumour types from the TIMER database (http://cistrome.shinyapps.io/timer/) (*P < 0.05, **P < 0.01, ***P < 0.001). PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES, prostaglandin E synthase; TYMS, thymidylate synthase. 9) Figure 10: provide more details to the legend. Extract data and show in form of graph. Response: Thank you for this helpful suggestion. We have provided a more detailed explanatory legend for Figure 10. We have extracted the data on the expression levels of proteins from the Human Protein Profiles and presented them in Figure 10B and Table 5. In addition, we have moved the Figure 8B to Figure 10C for elucidating the overexpression of these six genes in LUAD clearly. Find the relevant legend below for your quick reference: Figure 10 Protein expression levels and genetic alterations of the corresponding six prognostic genes obtained from online databases. (A) The representative immunohistochemistry images of the protein expression of the six genes in the normal lung tissues and LUAD tissues from the Human Protein Atlas database. (http://www.proteinatlas.org/) (B) The percentage of protein expression levels analysed based on the Human Protein Atlas database. Anti-PFKP antibody is HPA018257; anti-PKM antibody is CAB019421; anti-TPI1 antibody is HPA053568; anti-LDHA antibody is CAB069404; anti-PTGES is HPA045064; anti-TYMS antibody is CAB002784. (C) Genetic alterations in the six genes in 230 LUAD patients/samples (TCGA, Firehose Legacy). Data were obtained from the cBioportal for Cancer Genomics (http://www.cbioportal.org/). PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES, prostaglandin E synthase; TYMS, thymidylate synthase; TCGA, The Cancer Genome Atlas; LUAD, lung adenocarcinoma. 10) Provide more details about the pathways and the role of these genes in these pathways. Response: Thank you for this helpful suggestion. We have provided more details on the pathways and the role of the genes in these pathways in the results (line 254-275) and discussion (line 328-372). Responses to the reviewers’ comments: Reviewer 1 (Anonymous) Basic reporting No comment Experimental design No comment Validity of the findings The figures are a bit crowded and would benefits of a bit more explanation and additional legend descriptions. Response: Thank you for this great suggestion on improving our Figures. We have adjusted the display order of ten graphs and provided a more detailed explanatory legend for Figure 3. We have also fully described Figure 3 (line 185-205). We have provided a more detailed explanatory legend for Figure 7. Moreover, we have deleted the right panel of Figure 7 and shown the related data in Table 3. We have fully described Figure 7 and Table 3 (line 254-275). We have provided a more detailed explanatory legend for Figure 8. The right panel of Figure 8A has been edited, and labeled as Figure 8B. Moreover, those comparing the mRNA data between LUAD and normal tissues from the Oncomine database have been extracted and presented in Table 4. In addition, the Figure 8B has been moved to Figure 10C for elucidating the overexpression of these six genes in LUAD more clearly. We have provided a more detailed explanatory legend for Figure 10. We have extracted the data on the expression levels of proteins from the Human Protein Profiles and presented them in Figure 10B and Table 5. We have fully described Figure 8-10 and Table 4-5 (line 278-302). For improved legibility, we have made some changes as follows: the resolution of Figures 1-10 has been adjusted; and the regular letters in Figure 1 and Figures 3-10 have been put in bold. Some additional files could be merge as some just contains 3 or 4 entries on a excel sheet. Response: Thank you for this helpful suggestion. We have merged Table S1-2 into Table S1, Table S5-6 into Table S4, and Table S8-9 into Table S5. Table S3 has been labeled as Table S2 and Table S4 has been labeled as Table S3. Comments for the Author The paper is well written and provide analysing supporting the conclusions. This work tries fo fill a gap in the study of LUAD prognosis and thus propose to face a valable research question. The authors made a good use of available databases and performed relevant analysis in order to identify these genes. More details about the metabolic pathways involved and how these 6 genes could interfere would bring more strength to this manuscript. Response: Thank you for this valuable suggestion. We have provided more details on the pathways and the role of the genes in these pathways in the results (line 254-275) and discussion (line 328-372). Reviewer 2 (Anonymous) Basic reporting No comment. Experimental design No comment. Validity of the findings No comment. Comments for the Author In the manuscript ‘Identification of a six-gene metabolic signature predicting overall survival for patients with lung adenocarcinoma’, using a bioinformatics approach the authors provide some interesting insights concerning the role of a six-gene metabolic signature in the prediction of survival in LUAD patients. Overall, it is a robustly conducted study and a well-written manuscript. However, some minor corrections need to be made before it is accepted to be published. 1. The world ‘prognoses’ should be substituted by ‘prognosis’ throughout the text. Response: Thank for your helpful suggestion. We have replaced prognoses with prognosis throughout the manuscript. 2. In the ‘Gene Set Enrichment Analysis’ section of the results, the metabolism-related enriched pathways should be described further, including some of the key genes involved. Response: Thank you for this helpful suggestion. We have provided more details on the pathways and the role of the genes in these pathways in the results (line 254-275) and discussion (line 328-372). 3. Tables S6 and S7 should be moved in the main text, potentially as one merged table. Response: This is quite correct. We have merged Table S6 and S7 into Table 2. 4. Figure 1 should have at least a brief legend. Response: Thank you for this helpful suggestion. We have provided a legend for Figure 1. The relevant legend is provided below for your quick reference: Figure 1 Overall flowchart of steps used in the construction used in the construction of the prognostic metabolic gene signature. The TCGA dataset was utilized to construct the prognostic metabolic gene signature. The TCGA clinical information, GSE68465 dataset, and online databases from international platforms were further utilized to validate the prognostic model. TCGA, The Cancer Genome Atlas; OS, overall survival; ROC, receiver operating characteristic 5. The legend in figure 3 should provide some more details, i.e. what each graph corresponds to in each panel (A and B). A clearer alignment of the graphs would be useful, as well. Response: Thank you for this great suggestion to improve Figure 3. We have provided a more detailed explanatory legend fore Figure 3. We have adjusted the display order of ten graphs and relabeled each graph to make each of them clear in Figure 3. We have also fully described Figure 3 (line 185-205). The legend is provided below for your quick reference: Figure 3 Identification of the prognostic model in lung adenocarcinoma. (A, B) Kaplan-Meier curves of overall survival of the low-risk and high-risk groups stratified by the six-gene signature-based risk score in the TCGA or GEO dataset. (C, D) Risk score distribution, survival status distribution in the TCGA or GEO dataset. (E, F) The expression heatmap of the six prognostic genes in the TCGA or GEO dataset. (G, H) Time-dependent ROC curves of the six-gene signature in the TCGA or GEO dataset. TCGA, The Cancer Genome Atlas; GEO, Gene Expression Omnibus; ROC, receiver operating characteristic. 6. The quality of figure 8 should be improved. The right panel is almost impossible to read without zooming in. Alternatively, some of the data could be presented in the form of a table. Response: Thank you for this valuable suggestion. For improved legibility, the resolution of Figure 8 has been adjusted. We have provided a more detailed explanatory legend for Figure 8. The right panel of Figure 8A has been edited, and labeled as Figure 8B. Moreover, those comparing the mRNA data between LUAD and normal tissues from the Oncomine database have been extracted and presented in Table 4. In addition, the Figure 8B has been moved to Figure 10C for elucidating the overexpression of these six genes in LUAD more clearly. The relevant legend of Figure 8 is provided below for your quick reference: Figure 8 mRNA expression levels of the six prognostic genes from online databases. (A) mRNA expression levels of the six genes in the Oncomine database. The threshold is shown at the bottom (P value < 0.001 and fold change > 2 were utilized for screening). The figure in the colored cell represents the number of datasets complying with the threshold. The red cells indicate that the genes were overexpressed in the cancer cells, while the blue cells indicate that the genes were overexpressed in the normal tissues. (B) Comparisons of the mRNA expression levels of the six genes between LUAD and normal tissues in the combined LUAD datasets from the Oncomine database. PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES, prostaglandin E synthase; TYMS, thymidylate synthase; LUAD, lung adenocarcinoma. 7. Figure 9 is redundantly complicated. It would be clearer if the figure was focused only on LUAD, instead of incorporating so many different cancer types. It may be worth keeping only some relevant cancer types, e.g. other lung cancers or cancers where LUAD could metastasise to. Response: Thank you for this helpful suggestion. We have provided a more detailed explanatory legend for Figure 9. We have compared the expression of these six genes to other cancers and fully described the results of Figure 9 (line 285-291). The relevant legend is provided below for your quick reference: Figure 9 mRNA expression levels of the six prognostic genes extracted from online database. The mRNA expression levels of the six genes in different tumour types from the TIMER database (http://cistrome.shinyapps.io/timer/) (*P < 0.05, **P < 0.01, ***P < 0.001). PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES, prostaglandin E synthase; TYMS, thymidylate synthase. 8. The legend of figure 10 needs elaboration/details. Show which protein shows elevated/ decreased expression for each corresponding transcript of the six-gene signature. Response: Thank you for this helpful suggestion. We have provided a more detailed explanatory legend for Figure 10. We have extracted the data on the expression levels of the proteins from the Human Protein Profiles and presented them in Figure 10B and Table 5. In addition, we have moved the Figure 8B to Figure 10C for elucidating the overexpression of these six genes in LUAD clearly. Find the relevant legend below for your quick reference: Figure 10 Protein expression levels and genetic alterations of the corresponding six prognostic genes obtained from online databases. (A) The representative immunohistochemistry images of the protein expression of the six genes in the normal lung tissues and LUAD tissues from the Human Protein Atlas database. (http://www.proteinatlas.org) (B) The percentage of protein expression levels analysed based on the Human Protein Atlas database. Anti-PFKP antibody is HPA018257; anti-PKM antibody is CAB019421; anti-TPI1 antibody is HPA053568; anti-LDHA antibody is CAB069404; anti-PTGES is HPA045064; anti-TYMS antibody is CAB002784. (C) Genetic alterations of the six genes in 230 LUAD patients/samples (TCGA, Firehose Legacy). Data were obtained from the cBioportal for Cancer Genomics (http://www.cbioportal.org/). PFKP, phosphofructokinase platelet; PKM, pyruvate kinase muscle; TPI1, triosephosphate isomerase 1; LDHA, lactate dehydrogenase A; PTGES, prostaglandin E synthase; TYMS, thymidylate synthase; TCGA, The Cancer Genome Atlas; LUAD, lung adenocarcinoma. Thank you for giving us this great opportunity to improve our manuscript. We have done our best in revising the manuscript to correctly address your concerns. We appreciate the Editor and Reviewers earnestly, and we hope that you will approve our revisions. Once again, thank you very much for your comments. Sincerely, The authors "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. The persistence of the latent HIV-1 reservoir is a major obstacle to cure HIV-1 infection. HIV-1 integrates into the cellular genome and some targeted genomic loci are frequently detected in clonally expanded latently HIV-1 infected cells, for instance, the gene BTB domain and CNC homology 2 (BACH2).</ns0:p><ns0:p>Methods. We investigated HIV-1 promoter activity after integration into specific sites in BACH2 in Jurkat T-cells. The HIV-1-based vector LTatCL[M] contains two fluorophores: 1.) Cerulean, which reports the activity of the HIV-1 promoter, and 2.) mCherry driven by a constitutive promotor and flanked by genetic insulators. This vector was inserted into introns 2 and 5 of BACH2 of Jurkat T-cells via CRISPR/Cas9 technology in the same and convergent transcriptional orientation of BACH2, and into the genomic safe harbour AAVS1. Single cell clones representing active (Cerulean + /mCherry + ) and inactive (Cerulean -/mCherry + ) HIV-1 promoters were characterized.</ns0:p><ns0:p>Results. Upon targeted integration of the 5.3 kb vector LTatCL[M] into BACH2, the HIV-1 promoter was gradually silenced as reflected by the decrease in Cerulean expression over a period of 162 days. Silenced HIV-1 promoters could be reactivated by TNF-&#945; and Romidepsin. This observation was independent of the targeted intron and the transcriptional orientation. BACH2 mRNA and protein expression was not impaired by mono-allelic integration of LTatCL[M].</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion.</ns0:head><ns0:p>Successful targeted integration of the HIV-1-based vector LTatCL[M] allows longitudinal analyses of HIV-1 promoter activity.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Antiretroviral therapy (ART) blocks efficiently virus replication, however, does not cure HIV-1 infection due to the presence of replication-competent but silenced proviruses preferentially integrated in long-lived resting CD4 + T-cells <ns0:ref type='bibr' target='#b4'>(Finzi et al. 1997;</ns0:ref><ns0:ref type='bibr' target='#b42'>Wong et al. 1997)</ns0:ref>. Various factors and molecular mechanisms that result in HIV-1 latency have been proposed <ns0:ref type='bibr' target='#b28'>(Ruelas &amp; Greene 2013)</ns0:ref>. One such factor might be the integration site of the provirus, which has been suggested to not only be responsible for silencing the provirus, but also supporting cell expansion, thus maintaining the size of the HIV-1 latent reservoir <ns0:ref type='bibr' target='#b20'>(Maldarelli et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wagner et al. 2014)</ns0:ref>. Integration of HIV-1 into the human genome is not random. In vivo and ex vivo HIV-1 integration site analyses revealed that HIV-1 favours integration into introns of active transcription units in gene-dense regions, although a minority of integration events outside of these regions have consistently been observed <ns0:ref type='bibr' target='#b2'>(Ciuffi &amp; Bushman 2006;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kok et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b22'>Mitchell et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b32'>Schroder et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b37'>Stevens &amp; Griffith 1996)</ns0:ref>. Furthermore, HIV-1 appears to target active transcription units that are in close proximity to the nuclear pore <ns0:ref type='bibr' target='#b21'>(Marini et al. 2015)</ns0:ref>. On the population level, intragenic HIV-1 does not have a preference for either transcriptional orientation of the targeted gene <ns0:ref type='bibr' target='#b3'>(Cohn et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b32'>Schroder et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b37'>Stevens &amp; Griffith 1996)</ns0:ref>. HIV-1-infected cells can undergo clonal expansion and increase over time <ns0:ref type='bibr' target='#b1'>(Cesana et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Cohn et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b20'>Maldarelli et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b26'>Rezaei &amp; Cameron 2015;</ns0:ref><ns0:ref type='bibr' target='#b31'>Satou et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wagner et al. 2014)</ns0:ref>, and the proviruses in these clonally expanded cells are often located in specific regions of the human genome. One recurrent integration gene (RIG) that has been observed across patients'cells in numerous independent studies is the gene BACH2, in which the provirus is almost exclusively found in intron 5 and in the same transcriptional orientation as BACH2 <ns0:ref type='bibr' target='#b1'>(Cesana et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b9'>Ikeda et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b10'>Imamichi et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b18'>Mack et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b20'>Maldarelli et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wagner et al. 2014)</ns0:ref>. Since these BACH2 integration sites were identified in HIV-1-infected individuals who have been on ART for several years, it is conceivable that these proviruses are inactive, although it remains unknown whether this presumed inactivity is a result of integration site-dependent silencing of replication-competent proviruses or due to defective proviruses. To address the question of whether the HIV-1 promoter would be silenced upon integration into intron 5 of BACH2 in the same transcription orientation, we employed a modified version of our dual-fluorophore HIV-1-based vector, LTatC <ns0:ref type='bibr'>[M]</ns0:ref>, which reproduces features of active and latent HIV-1 infections <ns0:ref type='bibr' target='#b12'>(Kok et al. 2018)</ns0:ref>. This vector comprises two fluorescent reporter genes: 1.) Cerulean, which reports the activity of the HIV-1 promoter, and 2.) mCherry, the expression of which is driven by a constitutive promoter and further protected from position-effect variegation by a pair of flanking genetic insulators to identify cells harbouring an integrated vector <ns0:ref type='bibr' target='#b38'>(Uchida et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b39'>Villemure et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b43'>Yahata et al. 2007)</ns0:ref>. In this study, we investigate whether CRISPR/Cas9-mediated targeted HIV-1 integration in BACH2 is feasible and would lead to inactivation of the HIV-1 promoter over time, and if so, whether it is locus and/or transcriptional orientation dependent.</ns0:p></ns0:div> <ns0:div><ns0:head>Material and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Generation of LTatCL[M] with target locus homologous arms and Cas9/guide RNA-encoding plasmids</ns0:head><ns0:p>In LTatC[M] the 3'LTR is located downstream of the second fluorophore mCherry to enable retrovirus production and subsequent infection of target cells <ns0:ref type='bibr' target='#b12'>(Kok et al. 2018)</ns0:ref>. LTatC[M] was modified to LTatCL[M], i.e., the 3&#61602;LTR (L) was inserted between Cerulean (C) and the insulator cHS4 (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>) to further enhance the transcriptional independence of the HIV-1 promoter controlled Cerulean. For targeted integration of our HIV-1-based, dual-fluorophore vector, retrovirus production is not required. Thus, the HIV-1 3'LTR was relocated immediately downstream of Cerulean. Additionally, a polyA signal was inserted between mCherry ([M]) and the second insulator sMAR8 (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>). The homologous regions on both sides of the targeted HIV-1 integration site in the human genome, were obtained from NCBI GenBank: BACH2 intron 5 <ns0:ref type='bibr'>(Accession No: NT_007299.13;</ns0:ref>, and AAVS1 (Accession No: NC_000019.10; 5' arm nucleotides 1'399-2'218, 3' arm nucleotides 2'219-3'051). Targeted integration sites are depicted in Figure <ns0:ref type='figure' target='#fig_3'>1B</ns0:ref>. Primers to amplify the respective target locus homologous arms are listed in Supplementary Table <ns0:ref type='table'>1</ns0:ref>. Each PCR reaction contained 100 ng of human genomic DNA (Sup-T1 cell line; obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH, from Dr. Dharam Ablashi), 1x Platinum Taq PCR buffer (ThermoFisher), 2 mM MgCl 2 (ThermoFisher), 0.2 mM dNTP (NEB), 0.4 &#956;M of each forward and reverse primer, and 1 U Platinum Taq polymerase (ThermoFisher) in a total volume of 50 &#956;L. The PCR cycling conditions were as follows: 94&#176;C for 2 min; 35 cycles of (94&#176;C for 30 s, 55&#176;C for 30 s, 68&#176;C for 1 min); 68&#176;C for 5 min; 4&#176;C hold. The respective pair of target locus homologous arms were cloned into pGEM-T Easy vector (Promega). Subsequently, the dual-fluorophore vector LTatCL[M] was cloned into each plasmid in the same or convergent orientation via blunt-end cloning. The 6 plasmids pBACH2_i5-, pBACH2_i2-, and pAAVS1-LTatCL[M] were generated, containing the homologous arms of the targeted HIV-1 integration loci in BACH2 intron 5, BACH2 intron 2, and in the safe harbour loci AAVS1, respectively, containing LTatCL[M] in both transcriptional orientations (s and c) (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>). Plasmids encoding Cas9 and guide RNAs were based on pX458 (Addgene plasmid 48138) (Supplementary Table <ns0:ref type='table'>1</ns0:ref>). Annealing of 100 &#956;M 5&#61602; phosphorylated primers for the guide RNAs was performed using the following conditions: 80&#176;C for 5 min, 65&#176;C for 7 min, 60&#176;C for 7 min, 55&#176;C for 7 min, 50&#176;C for 7 min, 45&#176;C for 7 min, 40&#176;C for 7 min, 35&#176;C for 7 min, 30&#176;C for 7 min, 25&#176;C for 7 min, and 4&#176;C hold. The annealed guide RNA primers were separately cloned into pX458. All plasmid sequences were confirmed by next-generation sequencing. Briefly, plasmids were gel purified, diluted to 0.2 ng/&#956;l, and 1 ng of the plasmids were processed using the Illumina Nextera XT DNA library Prep kit. The DNA libraries were subsequently paired-end or singleend sequenced with an Illumina MiSeq using the MiSeq Reagent Kit v3 (150-cycle). Plasmid maps are shown in Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Transfection of Jurkat T-cells and single cell sorting</ns0:head><ns0:p>Prior to transfection, the six LTatCL[M] vectors (pBACH2_i5-, pBACH2_i5c-, pBACH2_i2s-, pBACH2_i2c-, pAAVS1_s-, and pAAVS1_c-LTatCL[M]) were linearized using 100 units of NsiI (NEB)/20 &#181;g DNA. For transfection, 1 million Jurkat T-cells were resuspended in 100 &#956;L Nucleofector solution (Cell Line Nucleofector&#8482; Kits, Lonza) and combined with 2 &#956;g of the linearized dual-fluorophore vector and 2 &#956;g of the corresponding gRNA/Cas9 plasmid: pX458_gBACH2 94,355-94,356 (intron 5), pX458_gBACH2 340,186-340,187 (intron 2), and pX458_gAAVS1 2218-2219 (AAVS1). pSpCas9(BB)-2A-GFP (pX458) was a gift from Feng Zhang (Addgene plasmid # 48138) <ns0:ref type='bibr' target='#b25'>(Ran et al. 2013)</ns0:ref>. Nucleofection was performed using the program X-001 for Jurkat T-cells (Amaxa&#8482; Nucleofector&#8482; II, Lonza). Cells were cultured in RPMI-1640 medium media supplemented with 10% fetal bovine serum (FBS) and 1% Penicillin Streptomycin (10'000 units/ml Penicillin, 10 mg/ml Streptomycin). At day 9 post transfection, cells were analysed using fluorescence-activated cell sorting (FACS) and sorted at 20 cells per well in 96-well plates using a BD FACSAria&#8482; III (BD Biosciences). Two cell phenotypes were sorted: 1. Cerulean + /mCherry + and 2. Cerulean -/mCherry + . This was done for each nucleofected cell population: BACH2_i5s, BACH2_i5c, BACH2_i2s, BACH2_i2c, AAVS1_s, and AAVS1_c. After 50 days, expanded cell cultures were analysed by flow cytometry and subsequently single cell sorted. Longitudinal flow cytometric analysis of the cells were done with the LSRFortessa II (BD Biosciences) and data were analysed using the FlowJo Software v.10.0.8. (FLOWJO, LLC) (Figure <ns0:ref type='figure' target='#fig_3'>1C</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Activation of silenced HIV-1 promoter in cells with integrated LTatCL[M] in BACH2</ns0:head><ns0:p>Over time silenced HIV-1 promoter in sorted Cerulean + /mCherry + monoclonal cell lines were reactivated using 10 ng/&#956;L Tumor Necrosis Factor Alpha (TNF-and 4 nM Romidepsin (Ro) &#120572;) (Selleckchem). After 24 h the cells were analysed by flow cytometry.</ns0:p></ns0:div> <ns0:div><ns0:head>Amplifying the junctions of LTatCL[M] integration into BACH2 and AAVS1</ns0:head><ns0:p>Genomic DNA was extracted from 5 million transfected Jurkat T-cells using the DNeasy Blood and Tissue kit (Qiagen) and quantified by Quant-iT&#8482; PicoGreen&#8482; dsDNA Assay kit (ThermoFisher Scientific). To verify the targeted integration of the dual-fluorophore vector LTatCL[M] into BACH2 or AAVS1, the junctions of targeted integration were amplified by (semi-)nested PCR using 300 ng genomic DNA containing 1x PCR buffer (Sigma Aldrich), 1.5 mM MgCl 2 , 0.2 mM dNTPs, 0.5 &#956;M of each respective forward and reverse primer (Supplementary Table <ns0:ref type='table'>1</ns0:ref>), and 0.5 U JumpStart Taq DNA polymerase (Sigma-Aldrich) in a total volume of 25 &#956;L. The PCR cycling conditions were as follows: 94&#176;C for 2 min; 35-40 cycles of (94&#176;C for 30 s, 63&#176;C for 30 s, 72&#176;C for 2 min); 72&#176;C for 5 min; 4&#176;C hold. Mono-allelic or bi-allelic integration of LTatCL[M] into BACH2 and AAVS1 was verified by PCR containing 100 ng genomic DNA, 1x PCR buffer, 1.5 mM MgCl 2 , 0.2 mM dNTPs, 0.5 &#956;M of each respective forward and reverse primers (Supplementary Table <ns0:ref type='table'>1</ns0:ref>), and 0.5 U JumpStart Taq DNA polymerase in a total volume of 25 &#956;L. The PCR cycling conditions were as follows: 94&#176;C for 2 min; 35 cycles (94&#176;C for 30 s, 63&#176;C for 30 s, 72&#176;C for 2 min); 72&#176;C for 5 min; 4&#176;C hold. Amplicons were sequenced by Sanger sequencing. The sequencing reaction contains 1x Seqmix (Big Dy Termination v1.1, 1x Dilution Buffer (ThermoFisher)), and 100 ng DNA in a total volume of 40 &#956;L. 0.2 &#956;M primer was added (Supplementary Table <ns0:ref type='table'>1</ns0:ref>). The PCR cycling condition were as follows: 40 cycles (96&#176;C for 30 s, 50&#176;C for 30 s, 60&#176;C for 4 min); 4&#176;C hold. Sequencing was performed following the manufacturer's instruction on an 3130xl Genetic Analyzer (Applied Biosystems).</ns0:p></ns0:div> <ns0:div><ns0:head>Near full-length amplification of integrated vector LTatCL[M]</ns0:head><ns0:p>Genomic DNA was extracted from 3 million transfected cells using the DNeasy Blood and Tissue kit and quantified by Quant-iT&#8482; PicoGreen&#8482; dsDNA Assay kit. To investigate the integrity of the integrated vector LTatCL[M], near full-length amplification of the vector was performed, using genomic 300 ng DNA, 1x Long Amp Taq Reaction Buffer (NEB), 0.4 mM dNTPs, 0.5 &#956;M of each forward and reverse primer (Supplementary Table <ns0:ref type='table'>1</ns0:ref>) and 2.5 U Long Amp Taq Polymerase (NEB) in a total volume of 25 &#956;L. The PCR cycling conditions were as follows: 94&#176;C for 30 min; 30 cycles of (94&#176;C for 20 s, 58-61&#176;C for 30 s, 72&#176;C for 5 min); 72&#176;C for 10 min; 4&#176;C hold. 1 ng purified amplicons were processed with the Nextera XT DNA Library Preparation Kit (Illumina) and subsequently sequenced using the MiSeq reagent Kit v2 as described above.</ns0:p></ns0:div> <ns0:div><ns0:head>Quantification of BACH2 mRNA expression</ns0:head><ns0:p>RNA was extracted from 3 million transfected cells using the All Prep DNA/RNA kit (Qiagen) and quantified by Nanodrop 1000. 800ng RNA was reverse transcribed with Prime Script Reverse Transcriptase (Takara) according to manufacturer's instructions. To exclude contamination of plasmid and genomic DNA contamination, reactions lacking reverse transcriptase were included for each RNA sample. To quantify the mRNA level of BACH2, qPCR reactions were performed in triplicates for each sample. cDNA was diluted 1:10 mixed with 1x PCR buffer (Sigma Aldrich), 1.5 mM MgCl 2 , 0.2 &#956;M dNTP (NEB), 0.5 &#956;M of each Fw and Rv primers, 1x SYBR&#174; Green, 50 nM Rox, 0.5 U JumpStart Taq Polymerase (Sigma-Aldrich) in a total volume of 25 &#956;l. For each reaction, primer pair was chosen to amplify a specific region of the BACH2 mRNA, spanning exon 7-8. As a mRNA expression control glycerinaldehyd-3-phosphat-dehydrogenase (GAPDH) was amplified with GAPDH-specific primers (Supplementary Table <ns0:ref type='table'>1</ns0:ref>). The qPCR cycling conditions were as follows: 94&#176;C for 2 min; 40 cycles (94&#176;C for 30 s, 55&#176;C for 30 s, 63&#176;C for 1 min (collect data)); 72&#176;C for 2 min; 95&#176;C for 15 s; 4&#176;C hold. Melt curves were collected to analyze specificity of the amplification. The qPCR was performed in the ABI 7800 real-time PCR thermos-cycler then analysed using the 7500 Software v2.0.4, using the comparative C t method <ns0:ref type='bibr' target='#b17'>(Livak &amp; Schmittgen 2001)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Quantification of BACH2 protein expression</ns0:head><ns0:p>Cells were lysed using assay lysis buffer (50 mM Tris-HCL pH8, 150 mM NaCl, 0.5% Na deoxycholate, 0.5% Triton X-100, 1x Protease inhibitor, ddH 2 O) and protein was quantified using the Bicinchoninic acid assay (BCA) (Pierce&#8482; BCA Protein Assay Kit). 40 &#956;g protein were mixed with loading dye and loaded on a Bolt&#8482; 4-12% Bis-Tris Plus Gels (ThermoFisher Scientific). Subsequently, proteins were transferred onto a nitrocellulose membrane (iBlot&#8482; 2 Transfer Stacks, ThermoFisher Scientific) blocked for 1h at RT with 5% Top-Block wt/vol in PBS supplemented with Tween-20 (10x PBS pH7.4, 0.1% Tween 20, and ddH 2 O). For protein detection, the membrane was incubated with primary antibody (1/400 diluted rabbit anti-BACH2 antibody (PA5-23642, ThermoFisher Scientific) and 1/5'000 diluted rabbit-anti-GAPDH antibody (ab9485, Abcam)) overnight at 4&#176;C. After washing the membrane four times with 1x PBS-T, the secondary antibody, 1/10'000 diluted goat anti-rabbit IgG H&amp;L (HRP) (ab97051, Abcam) was added and the membrane was incubated at RT for 1 h. The membrane was then washed four times with 1x PB-T. To visualize the protein, Immobilon Crescendo Western HRP substrate (Merck) was added to the membrane. Chemiluminescence was captured with the Stella system (model 3200, Matlab Group Companies).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Targeted integration of the 5.3 kb HIV-1-based, dual-fluorophore vector LTatCL[M] into BACH2 and AAVS1 via CRISPR/Cas9-mediated homology directed repair</ns0:head><ns0:p>To investigate specific effects of HIV-1 integration into BACH2, four constructs were generated containing the vector LTatCL[M] flanked by BACH2 homologous arms (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>) and integrated in Jurkat T-cells by means of the CRISPR/Cas9 technology. The constructs were integrated into intron 5 of BACH2 in 1.) the in vivo observed same (s) and 2.) the convergent (c) transcriptional orientation as BACH2 to investigate the potential impact of transcriptional orientation on HIV-1 promoter activity. Furthermore, LTatCL[M] was integrated into intron 2 of BACH2 in 3.) the same and 4.) the convergent transcriptional orientation to investigate the position effect of HIV-1 integration into BACH2. Intron 2 was chosen as it has not yet been observed in HIV-1 infected patients, however, it was frequently observed as HIV-1 integration locus in human CD34 + hematopoietic stem cells infected in vitro with HIV-1 <ns0:ref type='bibr' target='#b20'>(Maldarelli et al. 2014</ns0:ref>). In addition, LTatCL[M] was integrated into intron 1 of PPP1R12C (protein phosphatase 1 regulatory subunit 12C) in again both transcriptional orientations. This genomic locus was previously identified as preferred site for integration of adeno-associated virus (AAV) DNA and designated AAVS1 (AAV integration site) <ns0:ref type='bibr' target='#b14'>(Kotin et al. 1992)</ns0:ref>. AAVS1 is widely used as genomic safe harbour (GSH) althought not without some limitations particularly in the context of future use in gene therapy trials <ns0:ref type='bibr' target='#b23'>(Papapetrou &amp; Schambach 2016;</ns0:ref><ns0:ref type='bibr' target='#b29'>Sadelain et al. 2011</ns0:ref>). Nevertheless, for research purposes it is considered a GSH and we expect that the HIV-1 promoter will not be silenced when integrated in this GSH. These six vector constructs and subsequently generated Jurkat T-cell clones were named BACH2_i5s, BACH2_i5c, BACH2_i2s, BACH2_i2c, AAVS1_c, and AAVS1_s (Figure <ns0:ref type='figure' target='#fig_3'>1A and B</ns0:ref>). Jurkat T-cells were nucleofected with linearized plasmids of the six LTatCL[M] constructs. Nine days post transfection, the frequencies of stably transfected Cerulean + /mCherry + and single mCherry + cells were between 0.1 and 1.8% (Figure <ns0:ref type='figure' target='#fig_3'>1C</ns0:ref>). Cells were sorted to enrich Cerulean + /mCherry + (i.e., active HIV-1 promoter) and single mCherry + (i.e., inactive HIV-1 promoter) cells (Figure <ns0:ref type='figure' target='#fig_3'>1D</ns0:ref>). To obtain monoclonal cell lines, a sequential flow cytometric sorting strategy was employed. For all targeted integration variants, Cerulean + /mCherry + and single mCherry + cells were first sorted at 20 cells per well in 96-well plates and expanded in culture up to 50 days post transfection. We analysed by means of flow cytometry a total of 92 cell populations, and selected 29 for further expansion. These cells were single-cell sorted to obtain monoclonal cell lines and expanded in culture for at least 25 days prior to following further characterization. In total, 43 single cells expanded resulting in 43 monoclonal cell lines. For each targeted integration variant (BACH2_i5s, BACH2_i5c, BACH2_i2s, BACH2_i2c, AAVS1_s, and AAVS1_c), at least one monoclonal cell line for each of the two phenotypes (Cerulean + /mCherry + and single mCherry + ) were obtained with the exception of AAVS1_c, single mCherry + (Table <ns0:ref type='table'>1</ns0:ref>). The monoclonal cell lines were characterized as follows. First, targeted integration of the HIV-1-based, dual-fluorophore vector LTatCL[M] was verified by amplification of a fragment spanning the targeted gene up-or downstream of the 5' or 3' homologous arms, respectively, and LTatCL[M]. Targeted integration was confirmed in 39/43 monoclonal cell lines representing all three genomic loci in both orientations and for both fluorescence phenotypes, except for AAVS1_c single mCherry + ( Table <ns0:ref type='table'>1</ns0:ref>). Second, targeted integration of LTatCL[M] in those 39 monoclonal cell lines was verified to be mono-allelic by a PCR strategy that allows sufficient amplification only of the not targeted BACH2 or AAVS1 allele by using primers up-and downstream of the BACH2 or AAVS1 homologeous arms (Figure <ns0:ref type='figure' target='#fig_3'>1A</ns0:ref>) and short PCR extension times. All monoclonal cell lines contained a not targeted allele of the respective targeted integration site (Table <ns0:ref type='table'>1</ns0:ref>). Third, BACH2 mRNA expression was quantified at 120 days post sorting in 11 Cerulean + /mCherry + monolonal cell lines, representing each targeted integration variant. A decrease of BACH2 mRNA expression could not be observed in any Cerulean + /mCherry + monolonal cell line (Figure <ns0:ref type='figure'>2A</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>). Forth, the potential impact on BACH2 protein expression after targeted integration of LTatCL[M] into BACH2 was investigated by Western blot analyses at 120 days post sorting. In line with unaltered BACH2 mRNA expression, BACH2 protein expression was not impaired compared to non-transfected Jurkat T-cells in all 35 monoclonal cell lines tested (Figure <ns0:ref type='figure'>2B</ns0:ref>, Table <ns0:ref type='table'>1</ns0:ref>). Fifth, HIV-1 integration can affect various aspects of cellular physiology, for instance, cell proliferation. To test whether targeted integration of LTatCL[M] into BACH2 would lead to an evolutionary advantage in cell growth, a cell-growth competition assay was performed. The stably Cerulean and mCherry expressing BACH2_i5s_1.1 Cerulean + /mCherry + and AAVS1_s_2.1 Cerulean + /mCherry + monoclonal cell lines were combined in an approximately 1:1 ratio with the parental Jurkat T-cell line and cocultured for 25 days. A comparable decrease of Cerulean + /mCherry + expression was observed for both the BACH2_i5s_1.1 Cerulean + /mCherry + and the AAVS1_s_2.1 Cerulean + /mCherry + monoclonal cell line (Supplementary Figure <ns0:ref type='figure'>5</ns0:ref>).</ns0:p><ns0:p>In summary, we successfully performed targeted integration of the 5.3 kb HIV-1-based, dualfluorophore vector LTatCL[M] into various loci of the human genome via CRISPR/Cas9mediated homology-directed repair. We subsequently generated monoclonal cell lines modelling actively and latently HIV-1-infected cells with integrated LTatCL[M] in either transcriptional orientation in the targeted genomic loci.</ns0:p></ns0:div> <ns0:div><ns0:head>Silencing of the HIV-1 promoter in Jurkat T-cells with integrated LTatCL[M] in BACH2</ns0:head><ns0:p>To study the HIV-1 promoter activity over time in Cerulean + /mCherry + monoclonal cell lines, i.e., active HIV-1 promoter, the fluorescence profile was frequently monitored for 162 days. In the majority of Cerulean + /mCherry + monoclonal cell lines with integrated LTatCL[M] in BACH2, the HIV-1 promoter was gradually silenced as observed by the decline of the frequency of Cerulean + /mCherry + cells to &lt;50% in 16/18 monoclonal cell lines within 24-162 days and to &lt;10% in 10/18 monoclonal cell lines within 58-162 days (Figure <ns0:ref type='figure'>3A</ns0:ref> and Supplementary Figure <ns0:ref type='figure'>2</ns0:ref>). This was observed independent of the BACH2 introns 2 or 5 chosen for targeted integration of LTatCL[M] and the transcriptional orientation of LTatCL[M] relative to BACH2. In contrast, in all 5 AAVS1_s and AAVS1_c Cerulean + /mCherry + monoclonal cell lines the frequency of Cerulean + /mCherry + cells remained relatively stable for 162 days (&gt;80% in 4/5 cell clones) as observed for one BACH2_i5s clone (Figure <ns0:ref type='figure'>3A</ns0:ref>). Overall, the HIV-1 promoter when integrated into the BACH2 gene is silenced over time in the majority of Cerulean + /mCherry + monoclonal cell lines whereas it remains active when integrated into AAVS1.</ns0:p></ns0:div> <ns0:div><ns0:head>Silenced HIV-1 promoters can be reactivated by TNF-&#945; and Romidepsin</ns0:head><ns0:p>To evaluate the reactivation of silenced HIV-1 promoters in Cerulean + /mCherry + monoclonal cell lines, which gradually lost Cerulean expression, we performed an activation assay using TNF-&#945; and Romidepsin. TNF-&#945; and Romidepsin in combination have been shown to activate the silenced HIV-1 promoter <ns0:ref type='bibr' target='#b12'>(Kok et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b36'>Sogaard et al. 2015)</ns0:ref>. These compounds were used to evaluate the reactivability of the silenced HIV-1 promoter in 8 Cerulean + /mCherry + monoclonal cell lines with integrated LTatCL[M] in BACH2. After 99 days, 1.6-48.2% of cells were Cerulean + /mCherry + . Upon treatment with TNF-&#945; and Romidepsin, the frequencies of Cerulean + /mCherry + cells increased to 61.6-89.8% (Figure <ns0:ref type='figure'>3B</ns0:ref>), demonstrating that the silenced HIV-1 promoter was reactivatable in those monoclonal cell lines.</ns0:p></ns0:div> <ns0:div><ns0:head>Monoclonal single mCherry + cell lines harbour large internal deletions in LTatCL[M]</ns0:head><ns0:p>A total of 17 single mCherry + monoclonal cell lines, presumably representing latently HIV-1infected cells, silenced at an early time point upon targeted HIV-1 integration, were treated with TNF-&#945; and Romidepsin to activate the HIV-1 promoter. Cerulean expression was not induced by TNF-&#945; and Romidepsin in any of these 17 monoclonal cell lines (Supplementary Figure <ns0:ref type='figure'>3</ns0:ref>). To further investigate this, the whole vector LTatCL[M] was amplified and sequenced in 8 of these 17 monoclonal cell lines. Surprisingly, the integrated vector LTatCL[M] in these monoclonal single mCherry + cell lines harboured large internal deletions in the 5'HIV-1 LTR, tat, and/or Cerulean cassette (Figure <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>). The monoclonal BACH2_i5s cell lines 3.1 and 3.2, which were derived from the same cell culture after the 1 st sort, showed the same deletion in the IRES-Cerulean region (Figure <ns0:ref type='figure' target='#fig_2'>4A</ns0:ref>). A similar observation was made in the monoclonal BACH2_i2c cell lines 13.1 and 13.3, also derived from the same 1 st sorted cell population: The same deletion spanning from the 5'LTR to the Cerulean cassette was observed in both monoclonal cell lines (Figure <ns0:ref type='figure' target='#fig_2'>4B</ns0:ref>). All independent single mCherry + monoclonal cell lines contained large deletions in different regions of the vector suggesting that the plasmids were not the source of those deletions. This was confirmed by next-generation sequencing of the plasmids not showing any evidence for large deletions. Next, the integrated vector LTatCL[M] was sequenced in a cell population of single mCherry + cells immediately after the 1 st sort, 9 days post transfection (Supplementary Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Evidence for numerous different large deletions were observed in the 5'HIV-1 LTR, Tat, IRES and/or Cerulean (Supplementary Figure <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>), indicating that those deletions occurred early during targeted integration.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>We developed a novel model to study HIV-1 promoter activity based on CRISPR/Cas9-mediated targeted integration of an 5'337 bp HIV-1-based, dual-fluorophore vector into selected sites in the human genome. In our previous study, we have shown that the HIV-1-based, dualfluorophore vector LTatC[M] reproduces features of active and latent HIV-1 infections. <ns0:ref type='bibr' target='#b12'>(Kok et al. 2018</ns0:ref>). Here, we studied the fate of the HIV-1 promoter in specific HIV-1 integration loci/sites by targeted integration of the vector LTatCL[M] into BACH2 and AAVS1. In Jurkat T-cells, HIV-1-based vector integration into BACH2 led initially to an active HIV-1 promoter as shown by the expression of HIV-1 LTR controlled Cerulean. Over time those monoclonal cell lines showed a gradual silencing of the HIV-1 promoter. This might be due to transcriptional interference, which can occur in two ways: Either through promotor occlusion or convergent transcription, in which the transcription from the host genes interferes with HIV-1 transcription <ns0:ref type='bibr' target='#b5'>(Han et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b16'>Lenasi et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b34'>Shan et al. 2011</ns0:ref>). In our study, however, HIV-1 integration into the two target loci of BACH2 intron 5 and intron 2 in both orientations showed similar silencing of the HIV-1 promoter over time. Differences were observed for the safe harbour locus AAVS1, in which no substantial silencing of the HIV-1 promoter was observed over a time period of 162 days. This raises the question, whether BACH2 is an exceptional HIV-1 integration site promoting HIV-1 promoter silencing or, vice versa, AAVS1 is an exceptional HIV-1 integration site preventing HIV-1 promoter silencing. In a future study, we will be expanding the repertoire of investigated HIV-1 integration sites. In virally suppressed HIV-1-infected individuals, the provirus integrated into BACH2 has been found predominantly in intron 5 in the same orientation as the gene and has been linked with clonal expansion. However, in in vitro in HIV-1 infected cell lines, integration has been found to occur randomly in the BACH2 gene, indicating that the integration selection observed in vivo cannot be fully recapitulated <ns0:ref type='bibr' target='#b20'>(Maldarelli et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wagner et al. 2014)</ns0:ref>. The preference of HIV-1 integration into BACH2 intron 5 in the same orientation in vivo could be caused by: 1.) Selection of BACH2 intron 5 integrants over other integrants over time <ns0:ref type='bibr' target='#b7'>(Hughes &amp; Coffin 2016)</ns0:ref>, 2.) BACH2 intron 5 is preferentially targeted by HIV-1 in primary CD4 + T-cells, presumably due to the spatial location of this locus in the nucleus of target cells <ns0:ref type='bibr' target='#b21'>(Marini et al. 2015)</ns0:ref>, or 3.) distinct alterations in BACH2 expression levels caused by different HIV-1 integration patterns. BACH2 expression levels vary between T-cells in distinct differentiation stages <ns0:ref type='bibr' target='#b27'>(Richer et al. 2016</ns0:ref>). An increase of BACH2 transcripts in regulatory and effector T-cells were observed when HIV-1 was integrated in BACH2 <ns0:ref type='bibr' target='#b1'>(Cesana et al. 2017)</ns0:ref>. Enhanced expression of the wild-type BACH2 in regulatory T-cells leads to increased proliferation capacity without affecting the cells' phenotype <ns0:ref type='bibr' target='#b1'>(Cesana et al. 2017)</ns0:ref>. Alteration of BACH2 expression might cause HIV-1 persistence in BACH2 intron 5 and expansion of the cell. In Jurkat T-cells, as compared to primary CD4 + Tcells, the BACH2 expression levels might be different <ns0:ref type='bibr' target='#b33'>(Shan et al. 2017</ns0:ref>). We measured BACH2 expression 120 days post sorting. At this time point the majority of the transfected cell clones harboured silenced HIV-1 promoters. We could not observe an inhibitory effect on BACH2 mRNA or protein expression when BACH2 introns 5 and 2 were targeted for integration with our HIV-1-based vector. This indicates that other factors might lead to the persistence of HIV-1 in this locus in Jurkat T-cells. HIV-1 infection into a cellular gene can affect various aspects of cellular physiology as for example proliferation or a longer half-life of the cell <ns0:ref type='bibr' target='#b26'>(Rezaei &amp; Cameron 2015;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wagner et al. 2014</ns0:ref>). However, there was no evidence that targeted integration of LTatCL[M] into BACH2 had an evolutionary advantage in cell growth in competition with the parental Jurkat T-cell line. The expression of Cerulean decreased over time in almost all BACH2 monoclonal cell clones independent of the targeted intron and the orientation of HIV-1-based vector integration. The slow decline and the presence of Cerulean positive cells for up to 162 days in some of those monoclonal cell clones shows that targeted integration of LTatCL[M] into BACH2 does not immediately and not completely lead to HIV-1 promoter silencing in Jurkat Tcells. However, insulator-protected mCherry expression also declined in some of those cell clones, suggesting a gradual and strong silencing effect at the BACH2 loci. This silencing was reversible as shown by the treatment of cell clones harbouring silenced HIV-1 promoters with TNF-&#945; and Romidepsin that led to HIV-1 promoter reactivation. These observations are partly in line with <ns0:ref type='bibr'>Lange et al.(32)</ns0:ref>: They observed a comparable reversible silencing of the HIV-1 promoter in BACH2 within 20-40 days targeting two sites within BACH2 intron 5 via CRISPR/Cas9 and inserting one reporter gene under the control of the HIV-1 LTR-promoter in the same transcriptional orientation as BACH2 <ns0:ref type='bibr' target='#b15'>(Lange et al. 2018</ns0:ref>). Here, we observed the same phenotypes in another intron of BACH2 and independent of the orientation of the HIV-1 integration. Furthermore, targeted integration of LTatCL[M] into the safe harbour site AAVS1 did not lead to silencing of the HIV-1 promoter. This model can be further expanded by insertion of, for instance, certain HIV-1 genes, which will allow to study the impact of viral genes on the host's expression profile and cell cycle. Mono-allelic integration of 5.3 kb long LTatCL[M] into BACH2 and AAVS1 has been confirmed in monoclonal cell lines by amplifying and sequencing junctions of integration. So far, the longest fragment inserted in human lymphocytes was 1.5 kb long <ns0:ref type='bibr' target='#b8'>(Hung et al. 2018</ns0:ref>), hence, our findings show that insertion of a 3.5x larger fragment is possible in Jurkat T-cells. Fragments of similar length (5.5 kb and 7.4 kb) have been successfully integrated in other cell types (embryonic stem cells and zygotes) <ns0:ref type='bibr' target='#b41'>(Wang et al. 2015)</ns0:ref>. Monoclonal cell lines expressing initially only mCherry, i.e., supposed to model latently HIV-1 infected cells, were found to contain large deletions in the LTR-tat-Cerulean cassette. Similar deletions were already observed by us when we used a variant of our vector to generate retroviruses for infection of cells <ns0:ref type='bibr' target='#b12'>(Kok et al. 2018</ns0:ref>). There, we speculated that the deletions were caused by error-prone HIV-1 reverse transcription <ns0:ref type='bibr' target='#b0'>(Bebenek et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b24'>Patel &amp; Preston 1994)</ns0:ref> or copy-choice recombination <ns0:ref type='bibr' target='#b30'>(Sanchez et al. 1997;</ns0:ref><ns0:ref type='bibr' target='#b35'>Simon-Loriere &amp; Holmes 2011)</ns0:ref>, which is also observed in cells from HIV-1-infected individuals <ns0:ref type='bibr' target='#b6'>(Ho et al. 2013</ns0:ref>). However, our current system does not require reverse transcriptase prior to integration since we are using CRISPR/Cas9-mediated targeted integration. Therefore, the deletions might be due to recombination events triggered by the LTRs flanking the Cerulean cassette <ns0:ref type='bibr' target='#b19'>(Mager &amp; Goodchild 1989)</ns0:ref>. Together with our previous observation <ns0:ref type='bibr' target='#b12'>(Kok et al. 2018)</ns0:ref>, our results underline the importance to further investigate single fluorescent cell populations (representing latently HIV-1 infected cells) in latency models using HIV-1-based vector systems. Beyond HIV-1, these findings might also have implications for lentiviral vectors used for gene therapy.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Using the CRISPR/Cas9-technology, stable targeted integration of the 5.3 kb long HIV-1 dualfluorphore vector LTatCL[M] was successful, enabling longitudinal studies on effects of selected genomic sites on HIV-1 promoter activity and cellular phenotypes. Targeting the BACH2 gene revealed that loci within this gene are capable of supporting an active HIV-1 promoter upon integration but its activity diminishes over time in Jurkat T-cells. On the contrary, the HIV-1 promoter was not silenced when integrated into the genomic safe harbour AAVSI of Jurkat Tcells. (Cerulean and mCherry) to distinguish between inactive and active HIV-1 promoters, i.e., modelling latently and actively HIV-1 infected cells. The Cerulean cassette is under the control of the HIV-1 LTR whereas mCherry is under the control of an independent constitutive promoter (heIF4A1) and flanked by two insulators (cHS4 and sMAR8). LTR, long terminal repeat; tat, HIV-1 transactivator; IRES, internal ribosome entry site; cHS4, chicken hypersensitive site 4; tetO, tetracycline operator sequences; heIF4A1, human eukaryotic initiation factor 4A1, gene promoter; sMAR8, synthetic matrix attachment region 8. (B) Scheme of the targeted HIV-1 integration sites in BACH2 and AAVS1. Some described HIV-1 integration sites in vivo are marked by red arrows <ns0:ref type='bibr' target='#b20'>(Maldarelli, et al. 2014</ns0:ref><ns0:ref type='bibr' target='#b40'>, Wagner, et al. 2014)</ns0:ref> </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>.</ns0:head><ns0:label /><ns0:figDesc>post transduction of Jurkat T-cells targeting the different loci in BACH2 and AAVS1. The means and standard deviations of 3 independent experiments are depicted. (D) Flow chart to generate monoclonal cell lines. Jurkat T-cells were separately transfected with the vectors shown in A and the corresponding gRNA/Cas9 plasmid. Nine days post transfection, the six different targeted HIV-1 integration variants were each sorted by 20 After at least 25 days post 2 nd sorting, cells were further characterized. (A-D). The in vivo observed preferential HIV-1 integration</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Table 1 (</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>on next page) Characteristics of monoclonal Jurkat T-cell lines after targeted integration of LTatCL[inactive HIV-1 promoters in LTatCL[M] transfected cells. The in vivo observed preferential HIV-1 integration loci in BACH2, BACH2_i5s, is highlighted by a red box. PeerJ reviewing PDF | (2020:08:51698:1:1:NEW 13 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,403.12,525.00,246.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,70.87,525.00,438.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,70.87,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>43 (90.7%) 39/39 (100%) 11/11 (100%) 35/35 (100%)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>BACH2_i5s</ns0:cell><ns0:cell /><ns0:cell>.3%)</ns0:cell><ns0:cell>5/5 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>5/5 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>inactive</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>5/7 (71.4%)</ns0:cell><ns0:cell>5/5 (100%)</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>active</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>BACH2_i5c</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>inactive</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1/1 (100%)</ns0:cell><ns0:cell>1/1 (100%)</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>1/1 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>active</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>7/7 (100%)</ns0:cell><ns0:cell>7/7 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>7/7 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>BACH2_i2s</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>inactive</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>3/4 (75%)</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>active</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>4/4 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>BACH2_i2c</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>inactive</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>active</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>1/1 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>AAVS1_s</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>inactive</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>n.a.</ns0:cell></ns0:row><ns0:row><ns0:cell>active</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell><ns0:cell>2/2 (100%)</ns0:cell><ns0:cell>3/3 (100%)</ns0:cell></ns0:row><ns0:row><ns0:cell>AAVS1_c</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>inactive</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>n.a.</ns0:cell><ns0:cell>n.a.</ns0:cell></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell>43</ns0:cell><ns0:cell>39/</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51698:1:1:NEW 13 Oct 2020)</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51698:1:1:NEW 13 Oct 2020)Manuscript to be reviewed</ns0:note> <ns0:note place='foot' n='1'>1 active= Cerulean + /mCherry + , inactive = single mCherry + n.a. = not available PeerJ reviewing PDF | (2020:08:51698:1:1:NEW 13 Oct 2020)</ns0:note> </ns0:body> "

" Zurich, 7th October 2020 Dear Prof. Gwyn Gould, We are grateful for the editor’s and reviewers’ very favorable assessment of our study and deeply appreciate all thoughtful comments and constructive suggestions, which we have fully addressed in this revision and believe that this has strengthened the manuscript significantly. Our point-by-point reply contains our answers in red and changes are highlighted in the manuscript. We hope that the manuscript is now suitable for publication in PeerJ. Yours sincerely, Karin J. Metzner Reviewer 1 (Brendan Bell) Basic reporting In their manuscript entitled “HIV-1 promoter is gradually silenced when integrated into BACH2 in Jurkat T-cells”, Inderbitzin and colleagues report the development of initial characterization of a system to investigate the impact of integration site choice and HIV gene expression (latency). I found the introduction to be clearly written and to explain the knowledge gap being addressed. The Figures were easy to follow and the Materials and Methods section is detailed enough, in my opinion, to allow independent replication of the experiments. Overall the English is high quality and accessible for international readers. Experimental design The authors employ dual reporter viruses that have previously been published from their group. These constructs were integrated into the genome of Jurkat T cells using CRISPR/Cas9 technology via the addition of arms for homologous recombination into the known BACH2 gene that has been shown by several independent groups to be a “hot spot” for HIV integration. The work certainly suits the broad scope of the journal and will be of use to others working in the field. One question for the authors, in the section “Availability of materials” the authors have marked not applicable. Given the use of new viral constructs and the creation of useful cell lines in the work, why is this the case? We thank the reviewer for noticing this mistake. All constructs and cell clones are available upon request, we further clarified this in line 469. “All constructs and cell clones are available upon request.” Validity of the findings The experiments are in my view well controlled and the results are compelling. The authors show that HIV expression is gradually reduced in the BACH2 locus over time in an orientation-independent fashion, without significantly disturbing the expression of BACH2 protein levels. These results contrast with the integration into AAVS1 “safe haven” genomic site where HIV remains active over time. It should be fascinating to use the system in the future to further explore other genomic sites. Although I found the authors discussion of the results to be reasonable, I had two minor points that should be addressed for clarity to readers. Potential limitations of the system include the fact that an immortalized cancer cell line was employed. Given the fact that genomic preferences are only found in primary cells, certain elements of the system may not recapitulate the selection seen in vivo. A second limitation that should be mentioned is the fact that other viral genes such as vpr are not in these constructs. Since cell cycle impacts HIV gene expression and latency, and vpr manipulates the cell cycle, the absence of these genes could have importance on the future findings. We are thankful for the positive evaluation of our model. We regret that our model based on the Jurkat T-cell line was not designed to study the impact of HIV-1 infection on host proliferation or apoptosis, and therefore, is not suited to recapitulate the selection seen in vivo. The experimental approach using our dual-fluorophore HIV-1-based vector, LTatC[M], was meant to specifically measure the HIV-1 LTR activity at targeted genetic loci such as the clinically relevant BACH2 intron 5 over time. We too agree that it would be interesting to study the relationship between host cell cycle and HIV-1 infection. A full-length HIV-1 construct, for example, using our established experimental approaches would be more suited to address this question. Nevertheless, our model is a useful tool, which can be further modified to include specific viral gene(s) to study the combinatorial impacts of HIV-1 infection in conjunction with integration sites. We further specified this in lines 365-367. “However, in in vitro HIV-1 infected cell lines, integration has been found to occur randomly in the BACH2 gene, indicating that the integration selection observed in vivo cannot be fully recapitulated (Maldarelli et al. 2014; Wagner et al. 2014).” And lines 404-406. “This model can be further expanded by insertion of, for instance, certain HIV-1 genes, which will allow to study the impact of viral genes on the host’s expression profile and cell cycle.” Comments for the Author Overall, I commend the authors on a significant amount of careful work and a useful contribution to the field. Reviewer 2 (Anonymous) Basic reporting Inderbitzin et al. presents a study of HIV-1 integration into introns of the BACH2 gene in Jurkat cells. The author’s approach is using CRISPR insertion of a dual HIV-1 reporter, one gene reporting HIV-1 LTR activity, and another constitutively active. Authors insert this vector into two introns of BACH2 in sense and antisense orientation (relative to the gene) as well as an alternate locus in Jurkat cells. Authors then iteratively sort integrated cells to obtain monogenic clones. Authors observe a limited impact on the gene, BACH2, in terms of RNA and protein expression. Authors also observe a gradual decline in the expression of the HIV specific reporter. While a very interesting study, the approach is limiting to the general applicability of the results and interpretation must be made with caution. The authors should address that it is possible that the finding is because of insulator insertion rather than HIV insertion into BACH2 (if not experimentally justified). Experimental design Overall the authors spent respectful effort in inserting this cassette into the BACH2 site with several controls. Validity of the findings 1. Investigators have used a reporter with genomic insulating elements flanking promoters. While desirable for generating a dual reporter, authors have not demonstrated that these do not impact the ability of the HIV-1 LTR from impacting the integrated gene. The cHS4 element is specifically investigated in viral vector gene therapy because of its ability to block chromatin modifications and enhancer promoter interactions (see Emery et al. PNAS 2000). Enhancer blocking activity was later discovered to be due to a CTCF binding site (Bell et al. Cell 1999), which has been implicated gene expression changes in other human retroviruses (Satou et al. PNAS 2016, Melamed et al. eLife 2018). Given these known impacts, it is unclear that authors have demonstrated the impact of the HIV-1 LTR on the inserted gene. To make this claim, authors would need to generate clones without these insulator elements. Further, the vector used in this study contains elements that would likely interfere with read through HIV-1 driven transcription, such as the additional reporter promoter and the polyadenylation site. Or, alternatively, can the authors address these limitations of having significant transcriptional interference by the insulator cHS4, that the results of may be caused by insulator insertion? We thank the reviewer for this important remark. We do not think that the insulators would have a confounding effect on the HIV-1 LTR activity based on the study by Tian et al. (2009, Gene Therap., PMID: 19440229) and our previous study (Kok et al. Scientific Reports. 2018, PMID: 29977044). In Tian et al., they showed that - even without a polyA signal between two fluorescent reporter cassettes, having this combination of insulators and in this specific orientation - the expression levels of the individual fluorescent genes are comparable to constructs having only any one of the two respective reporter genes. Furthermore, two features in our construct would conceivably minimize interference on the HIV-1 LTR activity: (i) our construct has a polyA signal between the two fluorescent reporter gene cassettes to reduce crosstalks between them and (ii) the insulators are downstream of the gene cassette that is regulated by the HIV-1 LTR. In our first study using a similar construct (LTatC[M]), we have shown that the ratio of emergence of active and latent cells and vector integration site patterns between this construct and one without any genetic insulators (LTatCM) are comparable within the first 10 days after transduction using retroviral vectors (Supplementary Figure 1, Kok et al. Scientific Reports, 2018). Since our cells were sorted at these earlier time points, it is reasonable that the insulators did not confound our observations. The main function of the insulators is to prevent the mCherry cassette from being silenced as a result of position-effect variegation and transcriptional readthrough from the upstream HIV-1 LTR-driven Cerulean cassette, thus enabling the active identification of latent infections either in the early or late timepoints. Supplementary Fig. S1: 10-day fluorescence profile of SUP-T1 cells transduced with LTatC[M] or LTatCM. SUP-T1 cells were transduced with LTatC[M] (left panel) or a vector variant with no genetic insulators in the mCherry cassette (LTatCM) (right panel) at 15% (top panel) or 5% (bottom panel) transduction efficiencies. The fluorescence profiles of the four cell populations arose from each transduction: double negative (DN), single mCherry positive (M+), double positive (DP), and single Cerulean positive (C+), were measured with flow cytometry daily for 10 days. Each data point represents the mean of two independent transductions (n2) and error bars depict standard error means. Some error bars are within data points. (Kok et al. Spontaneous reactivation of latent HIV-1 promoters is linked to the cell cycle as revealed by a genetic-insulators-containing dual-fluorescence HIV-1-based vector, Scientific Reports, 2018, License under a CC BY 4.0 license) 2. Authors did not functionally validate the impact of HIV-1 integration into BACH2. While changes may not be appreciable at the qPCR and western blot level, integration into BACH2 may provide additional competitive advantages or clonal expansion. Authors should conduct competition experiments to determine if this is the case. Authors could perform these experiments either directly (by mixing parental cells with integrant cells and watching the ratio of parental to integrant cells overtime) or indirectly (by starting several wells of clones with the same number of cells and plotting expansion over time. Or, alternatively, can the authors address these caveats in their results? We thank the reviewer for this suggestion. We had in fact performed a cell-growth competition experiment mixing the cell clones BACH2_i5s_1.1 and AAVS1_s_2.1, both showing stable Cerulean and mCherry expression, with mock transfected parental Jurkat T-cells. We observed that the parental Jurkat T-cells outgrew both the BACH2_i5s and AAVS1_s cell clones in a highly similar dynamics. We included these results in a new supplementary figure (Supplementary Fig. 5) and added the following part in the result section in line 281-289. “Fifth, HIV-1 integration can affect various aspects of cellular physiology, for instance, cell proliferation. To test whether targeted integration of LTatCL[M] into BACH2 would lead to an evolutionary advantage in cell growth, a cell-growth competition assay was performed. The stably Cerulean and mCherry expressing BACH2_i5s_1.1 Cerulean+/mCherry+ and AAVS1_s_2.1 Cerulean+/mCherry+ monoclonal cell lines were combined in an approximately 1:1 ratio with the parental Jurkat T-cell line and cocultured for 25 days. A comparable decrease of Cerulean+/mCherry+ expression was observed for both the BACH2_i5s_1.1 Cerulean+/mCherry+ and the AAVS1_s_2.1 Cerulean+/mCherry+ monoclonal cell line (Supplementary Figure 5).” Supplementary Figure 5: Outgrowth of parental Jurkat T-cells within 25 days in a cell-growth competition experiment with Cerulean+/mCherry+ monoclonal cell lines. Five time points (0, 3, 10, 20, 25 days in culture) are depicted for exemplary Cerulean+/mCherry+ monoclonal cell lines, BACH2_i5s_1.1 and AAVS1_s_2.1, mixed in an approximately 1:1 ratio with the parental Jurkat T-cell line and BACH2_i5s_1.1 and AAVS1_s_2.1 without addition of the parental Jurkat T-cell line. Each data point represents the mean of three independent cell-growth competition assays (n=3) and error bars depict standard error means. Some error bars are within data points. The in vivo observed preferential HIV-1 integration loci in BACH2, BACH2_i5s, is highlighted by red boxes. Additionally, we added the following part in the discussion in lines 385-389: “HIV-1 infection into a cellular gene can affect various aspects of cellular physiology as for example proliferation or a longer half-life of the cell (Rezaei & Cameron 2015; Wagner et al. 2014). However, there was no evidence that targeted integration of LTatCL[M] into BACH2 had an evolutionary advantage in cell growth in competition with the parental Jurkat T-cell line.” 3. Authors could strengthen their claim that this system is physiologically relevant by showing their system recapitulates production of known chimeric RNAs such as those reported in Cesana et al. Nature Communications 2017 and Liu et a., Science Translational Medicien 2020. Or, alternatively, can the authors address this caveat, that without knowing whether this HIV inserts is still able to produce aberrant splicing to BACH2, it is unclear that whether the phenotype observed is because of HIV integration into BACH2 or the insulator insertion into BACH2? We thank the reviewer for raising this point. Our construct was not designed to study chimeric RNAs since it’s not a full-length HIV-1 plasmid, but to specifically study the impact of integration sites on the HIV-1 LTR activity. Nevertheless, our model provides insights into potential mechanisms of HIV-1 persistence. We reason that the phenotype we observe is not due to the insulator as explained in our answer to comment 1. 4. Investigators are unclear when they state “BACH2 mRNA nor BACH2 protein expression measured 120 days post sorting”. In figure 1D authors show these experiments occurring 25-30 days post second sort, which is 50 days post initial transfection which only adds to 75-80 days. One of the numbers included must be incorrect, and should be rectified for consistency and clarity. Can the authors revise? We thank the reviewer for carefully reading our manuscript and apologize for the confusion. The mentioned 25-30 days indicate the expansion time and from this time point on the different characterization experiments of single cell clones were performed, i.e., not all the characterization experiments were performed within 25-30 days post second sort. We clarified this in the results section in line 255-257 and revised it in Figure 1. “These cells were single-cell sorted to obtain monoclonal cell lines and expanded in culture for at least 25 days prior to following further characterization.” In brief, amplifications of the junctions and the Cerulean cassette were performed 25-40 days post 2nd sort, the activation assay was performed 99 days post 2nd sort, BACH2 protein expression was determined 120 days post 2nd sort , and longitudinal FACs analyses were performed up to 162 days post 2nd sort. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: In the absence of highly effective antiviral therapies against SARS-CoV-2, it is crucial to counter the known pathophysiological causes of severe COVID-19. Evaluating the efficacy existing drugs may expedite the development of such therapeutics. Severe COVID-19 is largely the result of a dysregulated immune response characterized by lymphocytopenia, neutrophilia and critical hypercytokinemia, or 'cytokine storm,' which is largely mediated by the cytokine interleukin-6 (IL-6). The IL-6 inhibitor tocilizumab (TCZ) could potentially suppress the effects of the pro-inflammatory cytokine and thereby lower mortality from the disease. This systematic analysis aimed to investigate and synthesize existing evidence for the efficacy of TCZ in reducing COVID-19 mortality. Methodology: PubMed and SearchWorks searches were performed to locate clinical studies with primary data on TCZ treatment for severe COVID-19. Sixteen case-control studies comparing mortality between TCZ and standard of care (SOC) were identified for quantitative synthesis. The systematic analysis was pre-approved through PROSPERO (CRD42020193479). Results: Combined mortality for the TCZ-treated and SOC groups were 26.0% and 43.4% respectively. In all but one of the studies, the odds ratio of mortality from COVID-19 pointed towards lower fatality with TCZ versus the SOC. A combined random effects odds ratio calculation yielded an odds ratio of 0.453 (95% CI 0.376-0.547, p&lt;0.001). Additionally, eighteen uncontrolled trials were identified for qualitative analysis producing a raw combined mortality rate of 16.0%. Conclusions: Important caveats to this research include the lack of prospective randomized control trials (RCTs) and the absence of data from the large COVATA study from the published literature. However, results from this systematic analysis of published research provide positive evidence for the potential efficacy of TCZ to treat severe COVID-19, validating the ethical basis and merit of ongoing randomized controlled clinical trials.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Coronavirus Disease 2019 (COVID-19) -caused by the novel Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) -manifests in a broad range of disease severity. Roughly 85% of confirmed cases present as a mild respiratory illness defined as minor fatigue, low-grade fever and dry cough, 15% develop severe pneumonia requiring hospitalization and 5% become critical indicated by acute respiratory distress syndrome (ARDS), septic shock, and multi-organ failure resulting in ICU admission, mechanical ventilation, and death. <ns0:ref type='bibr' target='#b0'>1</ns0:ref> The most significant known risk factors for COVID-19-related death are increasing age, chronic comorbidities including diabetes, cardiac disease, pulmonary and kidney dysfunction and male sex. <ns0:ref type='bibr' target='#b1'>2</ns0:ref> A dysregulated immune response -characterized by decreased T-cell counts, increased inflammatory cytokines and extra-pulmonary systemic hyperinflammation syndrome -is principally responsible for inducing critical pulmonary failure observed in COVID-19 and largely driven by interleukin-6 (IL-6). <ns0:ref type='bibr' target='#b2'>3</ns0:ref> This systematic review (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>) concerns the efficacy of an interleukin-6 inhibitor, tocilizumab (TCZ) in reducing severe COVID-19 mortality.</ns0:p></ns0:div> <ns0:div><ns0:head>COVID-19 Dysregulated Immune Response and the Role of IL-6</ns0:head><ns0:p>Severe COVID-19 is characterized by a dysregulated immune response to SARS-CoV-2 infection that is implicated in disease mortality even after viral load decreases. <ns0:ref type='bibr' target='#b3'>4</ns0:ref> The immune dysregulation presents with two sequential and diametrically opposed reactions that both instigate symptom aggravation. <ns0:ref type='bibr' target='#b4'>5</ns0:ref> The first pattern is impaired adaptive immune response with lymphocytopenia, which includes markedly reduced CD4+ and CD8+ T-cells, B-cells and natural killer (NK) cells. While T-cells are significantly decreased in all COVID-19 patients, reduction in B and NK cells are more affected in severe cases. <ns0:ref type='bibr' target='#b5'>6,</ns0:ref><ns0:ref type='bibr' target='#b6'>7</ns0:ref> Adaptive immune cell depletion impairs the body's ability to clear the virus and mitigate inflammatory reactions. <ns0:ref type='bibr' target='#b7'>8</ns0:ref> The second deleterious response is an over-activation of the innate immune system. This pathogenic response is characterized by an increase in neutrophils and pro-inflammatory cytokines including IL-6, IL-1&#946;, IL-2, IL-8, CCL3 and TNF-&#945;. The rapidly increasing cytokine levels -also known as a 'cytokine storm' -drives progression to septic shock, tissue damage and multiple organ failure (heart, liver, kidney, respiratory). <ns0:ref type='bibr' target='#b4'>5</ns0:ref> The effects are instigated by excessive NF-&#312;B and JAK/STAT pathway activation, alarmin release by damaged epithelial cells, neutrophil and macrophage infiltration, and alveolar damage by vessel permeability and alveolar wall thickening. <ns0:ref type='bibr' target='#b7'>8</ns0:ref> Roughly three-quarters of patients in severe condition present with IL-6 mediated respiratory failure while the other quarter occurs predominately through IL-1&#946; macrophage activation. <ns0:ref type='bibr' target='#b8'>9</ns0:ref> IL-6 concentration is thus a reliable predictor of COVID-19 severity as it is significantly elevated in fatal cases. <ns0:ref type='bibr' target='#b9'>10</ns0:ref> IL-6 has pleotropic effects including hematopoiesis, metabolic regulation, inflammation, autoimmunity and acute phase response. <ns0:ref type='bibr' target='#b10'>11</ns0:ref> Some IL-6dependent outcomes help stave off infections such as directing neutrophil migration to the infection site, increasing CD8+ T cell cytolytic capacity, and regulating antiviral thermostatic reactions. However, IL-6 is also implicated in viral infection disease progression as it leads to tissue permeability and edema, reduces IFN-&#947; production, drives anti-apoptotic molecules and promotes excessive neutrophil survival. The above adverse effects promote lethal inflammation and enable viral infiltration to distant organs. <ns0:ref type='bibr' target='#b11'>12</ns0:ref> Furthermore, elevated serum IL-6 is associated with impaired cytotoxic activity of NK cells, thus weakening their virus-killing capacity. <ns0:ref type='bibr' target='#b12'>13</ns0:ref> IL-6 is known to increase the rate of fibrotic clot formation, so it also may play a role in the thrombotic complications observed in COVID-19. <ns0:ref type='bibr' target='#b11'>12</ns0:ref> The renin-angiotensin system, which controls blood pressure and electrolyte balance, is an additional important factor in IL-6 modulation and COVID-19 pathology. As virus binds ACE2, thus reducing its availability, there is an increase of angiotensin II in COVID-19 patients, creating a positive feedback loop that advances proinflammatory signaling. <ns0:ref type='bibr' target='#b11'>12</ns0:ref> The responses and physiological effects of IL-6 release are summarized in Figure <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>COVID-19 Treatments Antivirals</ns0:head><ns0:p>Without a targeted drug for COVID-19, scientists and clinicians are attempting to rapidly find alternative treatments and solutions to combat the disease's lethal immunological effects. <ns0:ref type='bibr' target='#b13'>14</ns0:ref> One strategy is to utilize existing antiviral drugs with the expectation that they may exhibit similar effects against SARS-CoV-2. One of the more promising treatments is RNAdependent RNA polymerase inhibitor remdesivir which is shown to reduce COVID-19 mortality but is less effective in severe cases. <ns0:ref type='bibr' target='#b14'>15</ns0:ref> Other commonly used antivirals are hydroxychloroquine and chloroquine. Despite widespread utilization for COVID-19 patients, evidence is lacking for the clinical efficacy and prophylactic properties of hydroxychloroquine or chloroquine despite their in vitro antiviral and in vivo immunomodulatory properties. <ns0:ref type='bibr' target='#b15'>16</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Immune Suppressants</ns0:head><ns0:p>There are numerous investigations of existing immune suppressing and anti-cytokine interventions to counter the dysregulated, excessive immune response. After early evidence and recommendations against the use of corticosteroids to treat severe COVID-19, <ns0:ref type='bibr' target='#b16'>17,</ns0:ref><ns0:ref type='bibr' target='#b17'>18</ns0:ref> a large randomized evaluation of dexamethasone found that the drug significantly reduced 28-day mortality in patients included in the study (rate ratio 0.83; 95% CI 0.74-0.92; p&lt;0.001). However, mortality rate reductions varied depending on baseline respiratory demands upon randomization as there was reduction for patients on mechanical ventilation and oxygen but not for patients without respiratory support. <ns0:ref type='bibr' target='#b18'>19</ns0:ref> As of mid-August 2020, the WHO modified their recommendation against corticosteroids to include judicious administration under respiratory failure with ARDS. <ns0:ref type='bibr'>20</ns0:ref> Because of the systemic effects of corticosteroids, more options for targeted immune regulation is warranted. Common targets for inhibition include IL-6, the IL-1 family (IL-1&#946; and IL-18), TNF-&#945; and IFN-&#947; cytokines and the JAK/STAT pathway. <ns0:ref type='bibr' target='#b7'>8</ns0:ref> IL-6 is a particularly promising target due to its correlation with ARDS severity and mortality. <ns0:ref type='bibr' target='#b20'>21</ns0:ref> IL-6 inhibitors are already successfully utilized for other cytokine storm syndromes such as adverse T cell therapy reactions and Still's disease-associated hemophagocytic lymphohistiocytosis. <ns0:ref type='bibr' target='#b7'>8</ns0:ref> In COVID-19, IL-6 inhibitors should be carefully administered with appropriate timing due to its control of viral replication. <ns0:ref type='bibr' target='#b21'>22</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>IL-6 Inhibitor Tocilizumab</ns0:head><ns0:p>Tocilizumab (TCZ) [Actemra] is a recombinant monoclonal antibody with a humanized murine variable domain and a human IgG1 constant domain. TCZ binds to both membranebound and soluble IL-6 receptors, thus preventing IL-6 mediated signal transduction (Fig. <ns0:ref type='figure'>3</ns0:ref>). The drug was initially developed to treat rheumatoid arthritis and now it is also approved for giant cell arteritis and similar autoimmune ailments. Its safety profile was analyzed in a phase III double-blind placebo-controlled trial and it is reportedly effective in treating other cases severe cytokine release syndrome such as chimeric antigen receptor T-cell immunotherapy. <ns0:ref type='bibr' target='#b10'>11</ns0:ref> While TCZ is not yet approved for treatment of COVID-19, clinicians across the globe are utilizing the drug under emergency use authorization, including in the United States. One of the largest initial observational studies of TCZ evaluated 547 COVID-19 ICU patients in New Jersey comparing the survival rate of 134 individuals treated with standard of care (SOC) and TCZ with SOC controls finding a 46% and 56% mortality rate respectively and a 0.76 adjusted hazard ratio. <ns0:ref type='bibr' target='#b22'>23</ns0:ref> However, there was insufficient statistical power to conclude clinical efficacy of TCZ with the clinical data and they only focused cases that already progressed to a critical stage leading to an inflated mortality rate for both groups.</ns0:p><ns0:p>Randomized controlled trials (RCT) are the gold standard for evaluating the clinical efficacy of a drug. The first analyzed RCT for TCZ presented negative results. Genentech (Roche) -the producer of TCZ -discontinued their 450-participant phase III trial COVACTA because it failed to meet the primary endpoint of improved clinical status after 4 weeks with TCZ versus the SOC. <ns0:ref type='bibr'>24</ns0:ref> The disappointing outcome from COVACTA places serious doubt on the efficacy of TCZ against COVID-19. Additionally, another IL-6 inhibitor carlumab failed its Phase III RCT in the United States but still has some ongoing (NCT04322773, NCT04327388, NCT04412772). <ns0:ref type='bibr'>25</ns0:ref> Other IL-6 inhibitors under investigation include siltuximab (SYLVANT, NCT04322188) and fingolimod (NCT04280588) Nonetheless, several phase III and phase II RCTs remain in progress into at least September (REMDACTA, NCT04409262; NCT04372186, NCT04356937) and healthcare providers are still currently administering TCZ globally for advanced COVID-19 cases. However, the logic of suppressing IL-6 remains convincing. Therefore, it is essential that the known impact of TCZ is analyzed in a systematic manner to ascertain whether its continuing use is ethical, even in RCTs. Since dexamethasone is a broadly acting immunosuppressant, it should be noted that the use of this drug in clinical trials may obscure the effects of more targeted immunomodulatory drugs like TCZ.</ns0:p><ns0:p>The most recent systematic reviews on the use of TCZ for COVID-19 identified clinical trials without data synthesis <ns0:ref type='bibr' target='#b24'>26</ns0:ref> and performed a large meta-analysis of controlled trials but did not address issues with the individual studies. <ns0:ref type='bibr' target='#b25'>27</ns0:ref> Therefore, this systematic review will synthesize the evidence from individual case-control studies, analyze uncontrolled trials and evaluate their methods to determine whether the drug is potentially effective at reducing severe COVID-19-related mortality, thus corroborating the logic for continuing RCTs to evaluate the potential use of IL-6 inhibitors.</ns0:p></ns0:div> <ns0:div><ns0:head>Systematic Review Methods and Statistics</ns0:head><ns0:p>Articles utilized for the systematic review were selected from a PubMed search on August 4, 2020. Both authors screened and reviewed each paper, and disagreements were resolved by a third reader. Data was independently extracted by the readers. For the initial screening, the primary search terms were 'COVID-19' or 'SARS-CoV-2' and 'tocilizumab.' Papers with primary data for a case-control study comparing mortality rate from severe COVID-19 between TCZ and standard of care (SOC) were included for data synthesis. Uncontrolled studies on severe COVID-19 mortality with TCZ were reviewed separately without data synthesis. Exclusion criteria included papers without primary data, case reports, reviews, protocols, and studies without mortality numbers available or patient data that may be used in another studies. An additional search was performed on SearchWorks to identify case-control studies not found in PubMed.</ns0:p><ns0:p>For each study included in the synthesis, the mortality rate for the TCZ and SOC group were calculated. In the controlled studies, the odds ratio (RR) of mortality from COVID-19 with TCZ versus the SOC was determined followed by the 95% confidence interval (CI) and p-value calculation. The data from the individual controlled studies were synthesized by a random effects meta-analysis calculation using MedCalc software. MedCalc was also used to perform a sample size calculation with an alpha of 0.01 and power of 90% to detect a difference between the total crude TCZ and SOC mortality rates.</ns0:p><ns0:p>The systematic review protocol was pre-registered with PRISMA and approved on June 22, 2020 (CRD42020193479).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>A total of 314 articles were identified by the initial PubMed search and three additional case-control studies were found on a SearchWorks (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). <ns0:ref type='bibr' target='#b26'>28</ns0:ref> 38 articles were selected for fulltext review yielding 16 uncontrolled studies for qualitative analysis and 18 case-control studies for both quantitative synthesis and qualitative analysis. The study characteristics for the controlled studies are summarized in Supplementary Table <ns0:ref type='table'>1</ns0:ref> while the uncontrolled studies are outlined in Supplementary Table <ns0:ref type='table'>3</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Controlled Studies</ns0:head><ns0:p>The systematic review of sixteen controlled studies encompassed a total of 1008 TCZtreated and 1537 SOC control patients (Table <ns0:ref type='table'>1</ns0:ref>). 13 of the studies occurred in a single medical center while the remaining three aggregated data from multiple hospitals. The largest patient contributions to the analysis were from the multiple-hospital studies [Ip et <ns0:ref type='figure' target='#fig_5'>4</ns0:ref>). A random effects odds ratio analysis generated an odds ratio of 0.453 (95% CI 0.376-0.547) with a p-value less than 0.001. A sample size analysis with alpha of 0.01 and power of 90% determined that 392 total case and control patients are needed to detect a difference between 26.0% and 43.4% mortality. With the exception of Capra et al., the studies also scattered symmetrically around the overall odds ratio from the analysis suggesting a low likelihood of publication bias (Fig. <ns0:ref type='figure'>5</ns0:ref>). TCZ patients in five studies had significant secondary bacteremia but one reported a lower rate than the SOC (Rojas Marte et al.). Eight studies reported no adverse effects from TCZ.</ns0:p><ns0:p>No pattern was identified in terms of the length of hospitalization between TCZ and SOC. Nine studies found a lower rate of ICU admission with TCZ, four with statistical significance (Colaneri et Several variables varied between studies that could impact the mortality rate results. These include SOC, observation time, TCZ administration, treatment date, baseline clinical characteristics, geographic location and resources and mean/median age. Study design was also an important varying factor that may change results as nine of the sixteen studies matched cases with controls and thirteen studies were retrospective as opposed to prospective cohort.</ns0:p></ns0:div> <ns0:div><ns0:head>Uncontrolled Studies</ns0:head><ns0:p>The 18 uncontrolled trials encompassed 886 total patients who received TCZ. The mortality rate from severe COVID-19 ranged from 0% to 42.4% (SD 9.87%), although the two studies with 0% had relatively small sample sizes (20 for Xu et al. and 12 for Borku Uysal et al.). The raw overall mortality rate from the 12 studies is 16.0%. The initial patient severity level ranged from 'severe' -requiring supplemental oxygen -to ICU admission. One study only investigated ICU patients (Issa et al.) and others included as many as 77.7% on MV. SOC varied more widely in the uncontrolled trials than the controlled, but hydroxychloroquine was still the most common additional drug used. Few major side effects such as bacterial/fungal infections and increased hepatic enzymes were reported.</ns0:p><ns0:p>Comparing the combined data between uncontrolled (n=18) and controlled trials (n=14) with TCZ (Table <ns0:ref type='table'>2</ns0:ref>), excluding controlled studies with all the patients initially in the ICU, the mortality rate was 19.3% and 16.0% respectively (p=0.384).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head></ns0:div> <ns0:div><ns0:head>Controlled Studies</ns0:head><ns0:p>The purpose of this systematic analysis was to analyze and synthesize clinical data on the efficacy of TCZ treatment against severe COVID-19. In the sixteen existing case-control trials published as of August 4, 2020, combined mortality for the TCZ-treated and SOC groups were 26.0% and 43.4% respectively. All of the studies at least trended towards reduced mortality with TCZ -with one exception (Patel et al.) that showed no benefit -(Fig. <ns0:ref type='figure' target='#fig_5'>4</ns0:ref> ). After performing a quantitative synthesis, the random effects odds ratio of mortality with TCZ versus the SOC was 0.453 (95% CI 0.376-0.547, p&lt;0.001) illustrating a reported difference in patient outcomes associated with the use of TCZ. There was no indication of publication bias except for Capra et al. (Fig. <ns0:ref type='figure'>5</ns0:ref>) and the number of patients in the meta-analysis exceeded the 392 required total case and control patients to statistically detect the difference in mortality rate. Some studies also secondarily investigated the rate of ICU admission with nine finding a lower rate with TCZ versus SOC and four having statistical significance (Colaneri et <ns0:ref type='table'>1</ns0:ref>). Although the value of any single controlled clinical study does not hold definitive proof of efficacy, the consistent positive trend and statistical significance from the combined data in this analysis corroborates TCZ's potential positive effects.</ns0:p><ns0:p>TCZ also appears to be mostly benign as few serious side effects were reported in most of the studies. As expected from an immune suppressant, the most notable adverse event was secondary infection reaching as high as 32.4% of participants (Rossotti et al.) but not at a statistically greater rate than SOC.</ns0:p><ns0:p>It is important to acknowledge that only one of the studies (Guaraldi et al.) randomized who received TCZ which introduces the possibility for selection bias into each study's research methodology. Additionally, only one study provided both baseline and post-treatment values for important biomarkers, notably IL-6, CRP, neutrophils and lymphocytes. Specific strengths and shortcomings for the controlled trials are outlined in Supplementary Table <ns0:ref type='table'>2</ns0:ref>, which points towards additional methodological issues in each of the studies analyzed. Common themes include short observation timing, small sample size, difference in treatment time (later patients may benefit from better SOC), variation in disease severity within groups which can confound results, and changes to TCZ administration mid-study. Future clinical trials can be improved by addressing the methodological issues outlined above.</ns0:p><ns0:p>Additionally, it is possible that individual studies finding a negative effect of TCZ were not published and could not be accounted for in this systematic review. Most notably, the known public results from the COVACTA trial were not yet released in published format as of late August and were therefore not included in the quantitative synthesis. It is significant that the RCT COVACTA did not find reduced COVID-19 mortality with TCZ versus SOC, placing doubt as to the drug's potential efficacy and the ethical basis of continuing other trials. Nonetheless, there are still multiple RCTs in progress to evaluate the efficacy of TCZ (NCT04409262; NCT04372186, NCT04356937, NCT04412772). Data from this systematic review adds merit to clinical investigations despite the initial negative results from COVACTA.</ns0:p></ns0:div> <ns0:div><ns0:head>Uncontrolled Studies</ns0:head><ns0:p>Uncontrolled trials were analyzed separately to explore trends in treatment data. Recognizing the wide variation in patient outcomes between the studies, the combined mortality rate from 12 single-arm studies using TCZ against severe COVID-19 was 16.0% (SD 9.87%). However, without clinical trials, it is difficult to determine the baseline rate COVID-19 mortality for hospitalized patients to compare with the rate calculated from the uncontrolled trials. Clinical data reviews are subject to geographic and demographic differences, such as a 5700-person evaluation in New York from April 2020 that found a 21% fatality rate for hospitalized patients which is higher than other locations and later studies. <ns0:ref type='bibr' target='#b44'>44</ns0:ref> Participants must be matched to controls to eliminate bias and account for other confounding factors to draw definitive conclusions. Comparing the uncontrolled and controlled trials from this systematic review, excluding controlled trials with all the patients initially in the ICU (Ip et al. and Rojas-Marte et al.), there was no significant difference in mortality rate between the two experimental approaches (p=0.384). This observation supports the assertion that the combined reported results from the uncontrolled results from this systematic analysis are potentially accurate and corroborate the merit in evaluating TCZ efficacy. Nevertheless, the uncontrolled trials should still be evaluated with some degree of skepticism. Specific shortcomings for the individual uncontrolled trials are delineated in Supplementary Table <ns0:ref type='table'>4</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>TCZ Treatment Timing</ns0:head><ns0:p>There is a question of timing for the IL-6 blocking treatment. All of the studies evaluated included only patients who were already in a severe disease state. Given the patterns of COVID-19 pathology and immune dysregulation, it may be logical to defer TCZ until the inflammatory phase due to the positive effects of IL-6 release in the acute infection stage which theoretically prevents SARS-CoV-2 proliferation. Given the unique, aberrant immune reaction in COVID-19, some researchers argue that the optimal time to employ targeted immune suppressants such as TCZ, in order to curtail and not enhance mortality, is when patients begin to trend towards hypoxia and inflammation. <ns0:ref type='bibr' target='#b2'>3</ns0:ref> However, this timing is theoretical and must be demonstrated in a RCT.</ns0:p></ns0:div> <ns0:div><ns0:head>Limitations</ns0:head><ns0:p>There are notable limitations to this systematic analysis and qualitative synthesis of TCZ. First, only one of the studies presented randomized who received TCZ, opening the possibility for selection bias and confounding factors that cannot be accounted for statistically. This systematic analysis synthesizes data from studies with different SOCs, geographies, resources, demographics, and TCZ dosing amount, number and timing. It is not possible to conclude that TCZ is efficacious in reducing COVID-19 mortality, simply that the data trends towards a lower odds ratio for mortality with incomplete generalizability. Similarly, while patients across all of the studies were at least in severe condition, the combined data still represents individuals at various stages of COVID-19. Not all of the studies offered a longitudinal time component, so an overall hazard ratio or Kaplan-Meir survival curve cannot be produced. Additionally, many patients were still in the hospital at the end of the observation period potentially skewing the mortality rate. Importantly, in the random effects odds ratio calculation, there was no control for age, sex and baseline characteristics like individual studies were able to accomplish. As noted, the uncontrolled trials on TCZ cannot be adequately evaluated without direct comparison to a control group. Finally, it is possible that studies finding a negative effect of TCZ were not published and not accounted for in this systematic review.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>A systematic review of the clinical data of IL-6 inhibitor tocilizumab (TCZ) for severe COVID-19 points towards efficacy in reducing mortality from the disease. There are numerous notable methodological limitations in the studies analyzed including the lack of randomization in controlled trials and potential for an inadequate evaluation due to unpublished data. However, the results from this systematic analysis corroborate the logic and ethical basis for ongoing phase III RCTs on TCZ. In light of this analysis, several factors would facilitate the evaluation of TCZ as a therapeutic for the immune dysregulation associated with COVID-19: 1) Publication of the results from unpublished clinical trials, 2) Completion of additional randomized controlled trials especially where the potentially complicating effects of dexamethasone may be ruled out, 3) Comparison of TDZ findings with results assessing other IL-6 inhibitors such as sarilumab or siltuximab, and 4) Completion of additional metabolic studies measuring the levels of immune mediators and biomarkers in SARS-CoV-2 infected animal models and humans treated with TCZ. It will also be useful to assess the possibility of therapeutic synergy between antiviral agents such as remdesivir and TCZ. The use of TCZ outside the clinical trial context is discouraged until results from these clinical trials are released. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>al. and Guaraldi et al.] The baseline patient characteristic for all but one study [Guaraldi et al.] was severe COVID-19, generally qualified by oxygen supplementation needs. Ip et al., Eimer et al. and Potere et al. only analyzed patients who were already admitted into the ICU. 61% and 44.8% of both cases and controls in Rojas-Marte et al. and Roumier et al. respectively began the trial in critical condition. There was some variation in TCZ administration and SOC treatments, the most common being hydroxychloroquine -utilized in all but one study (Capra et al.) -and lopinavir/ritonavir. Length of observation ranged from 7 days to 30 days with endpoints of death or discharge. Mean age of participants in the treatment and control groups was 55.5 to 76.8 with no more than 6.1 years separating the two groups within one study. Combined mortality for the TCZ-treated and SOC groups were 26.0% and 43.4% respectively. All of the studies trended toward lower mortality from severe COVID-19 with TCZ versus the SOC with the exception of Patel et al. which showed no benefit. Six studies yielded a statistically significant result (Capra et al., Guaraldi et al., Gokhale et al., Rossotti et al., Somers et al., Potere et al.) (Fig.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>) including six with statistical significance (Capra et al., Guaraldi et al., Gokhale et al., Rossotti et al., Somers et al., Potere et al.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 1 PRISMA</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 2 Responses</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 3 IL- 6</ns0:head><ns0:label>36</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 4 Odds</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,229.87,525.00,393.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>al., Roumier et al., Guaraldi et al., Canziani et al.) (Supplementary Table 1). Only one study (Colaneri et al.) provided both baseline and post-treatment values for biomarkers such as IL-6, CRP, lymphocytes, neutrophils, ALT etc.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>al., Roumier et al., Guaraldi et al., Canziani et al.) (Supplementary Table</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>11 1 Table 1 :</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Quantitative Synthesis of Individual Case-Control Study Mortality Data</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Study</ns0:cell><ns0:cell>TCZ Mortality</ns0:cell><ns0:cell>Controls</ns0:cell><ns0:cell>Odds ratio</ns0:cell><ns0:cell>95% CI</ns0:cell><ns0:cell>Weight %</ns0:cell><ns0:cell>p</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>(Random</ns0:cell><ns0:cell>Value</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Effects)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Klopfenstein et al. 29</ns0:cell><ns0:cell>4/20 (20%)</ns0:cell><ns0:cell>12/25 (48%)</ns0:cell><ns0:cell>0.271</ns0:cell><ns0:cell>0.0704-</ns0:cell><ns0:cell>3.54</ns0:cell><ns0:cell>0.0575</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.042</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Campochiaro et</ns0:cell><ns0:cell>5/32 (15.6%)</ns0:cell><ns0:cell>11/33 (33.3%)</ns0:cell><ns0:cell>0.37</ns0:cell><ns0:cell>0.112-1.227</ns0:cell><ns0:cell>4.23</ns0:cell><ns0:cell>0.104</ns0:cell></ns0:row><ns0:row><ns0:cell>al. 30</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Capra et al. 31</ns0:cell><ns0:cell>2/62 (3.2%)</ns0:cell><ns0:cell>11/23 (47.8%)</ns0:cell><ns0:cell>0.0364</ns0:cell><ns0:cell>0.00713-</ns0:cell><ns0:cell>2.61</ns0:cell><ns0:cell>0.0001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.185</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Colaneri et al. 32</ns0:cell><ns0:cell>5/21 (23.8%)</ns0:cell><ns0:cell>6/21 (28.6%)</ns0:cell><ns0:cell>0.781</ns0:cell><ns0:cell>0.197-3.106</ns0:cell><ns0:cell>3.41</ns0:cell><ns0:cell>0.726</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Rojas-Marte et al. 33 50/96 (52.1%) 60/97 (61.9%)</ns0:cell><ns0:cell>0.67</ns0:cell><ns0:cell>0.378-1.189</ns0:cell><ns0:cell>9.7</ns0:cell><ns0:cell>0.171</ns0:cell></ns0:row><ns0:row><ns0:cell>Wadud et al. 34</ns0:cell><ns0:cell>17/44 (38.6%)</ns0:cell><ns0:cell>26/50 (52%)</ns0:cell><ns0:cell>0.581</ns0:cell><ns0:cell>0.255-1.323</ns0:cell><ns0:cell>6.9</ns0:cell><ns0:cell>0.196</ns0:cell></ns0:row><ns0:row><ns0:cell>Ip et al. 23</ns0:cell><ns0:cell>62/134</ns0:cell><ns0:cell>231/413</ns0:cell><ns0:cell>0.678</ns0:cell><ns0:cell>0.459-1.003</ns0:cell><ns0:cell>12.2</ns0:cell><ns0:cell>0.0519</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>(46.3%)</ns0:cell><ns0:cell>(55.9%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Roumier et al. 35</ns0:cell><ns0:cell>3/30 (10%)</ns0:cell><ns0:cell>9/30 (30%)</ns0:cell><ns0:cell>0.259</ns0:cell><ns0:cell>0.0623-</ns0:cell><ns0:cell>3.24</ns0:cell><ns0:cell>0.0635</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>1.079</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Guaraldi et al. 36</ns0:cell><ns0:cell cols='2'>13/179 (7.3%) 73/365 (20%)</ns0:cell><ns0:cell>0.309</ns0:cell><ns0:cell>0.2166-</ns0:cell><ns0:cell>9.11</ns0:cell><ns0:cell>0.0002</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.574</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Patel et al. 37</ns0:cell><ns0:cell cols='2'>11/42 (26.2%) 11/41 (28.6%)</ns0:cell><ns0:cell>0.968</ns0:cell><ns0:cell>0.355-2.565</ns0:cell><ns0:cell>5.62</ns0:cell><ns0:cell>0.947</ns0:cell></ns0:row><ns0:row><ns0:cell>Eimer et al. 38</ns0:cell><ns0:cell>4/22 (18.2%)</ns0:cell><ns0:cell>7/22 (31.8%)</ns0:cell><ns0:cell>0.476</ns0:cell><ns0:cell>0.116-1.94</ns0:cell><ns0:cell>3.31</ns0:cell><ns0:cell>0.301</ns0:cell></ns0:row><ns0:row><ns0:cell>Canziani et al. 39</ns0:cell><ns0:cell cols='2'>17/64 (26.6%) 24/64 (37.5%)</ns0:cell><ns0:cell>0.603</ns0:cell><ns0:cell>0.285-1.277</ns0:cell><ns0:cell>7.61</ns0:cell><ns0:cell>0.187</ns0:cell></ns0:row><ns0:row><ns0:cell>Gokhale et al. 40</ns0:cell><ns0:cell>33/70 (47.1%)</ns0:cell><ns0:cell>61/91 (67%)</ns0:cell><ns0:cell>0.345</ns0:cell><ns0:cell>0.185-0.644</ns0:cell><ns0:cell>8.85</ns0:cell><ns0:cell>0.0008</ns0:cell></ns0:row><ns0:row><ns0:cell>Rossotti et al. 41</ns0:cell><ns0:cell cols='2'>20/74 (26.4%) 86/146 (58.9%)</ns0:cell><ns0:cell>0.517</ns0:cell><ns0:cell>0.301-0.887</ns0:cell><ns0:cell>9.23</ns0:cell><ns0:cell>0.0166</ns0:cell></ns0:row><ns0:row><ns0:cell>Somers et al. 42</ns0:cell><ns0:cell cols='2'>14/78 (17.9%) 22/76 (35.5%)</ns0:cell><ns0:cell>0.379</ns0:cell><ns0:cell>0.189-0.836</ns0:cell><ns0:cell>7.67</ns0:cell><ns0:cell>0.0151</ns0:cell></ns0:row><ns0:row><ns0:cell>Potere et al. 43</ns0:cell><ns0:cell>2/40 (5%)</ns0:cell><ns0:cell>12/40 (30%)</ns0:cell><ns0:cell>0.123</ns0:cell><ns0:cell>0.0254-</ns0:cell><ns0:cell>2.76</ns0:cell><ns0:cell>0.009</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.593</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Total (random</ns0:cell><ns0:cell>262/1008</ns0:cell><ns0:cell>667/1537</ns0:cell><ns0:cell>0.453</ns0:cell><ns0:cell>0.376-0.547</ns0:cell><ns0:cell>100</ns0:cell><ns0:cell>&lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>effects)</ns0:cell><ns0:cell>(26.0%)</ns0:cell><ns0:cell>(43.4%)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:09:52525:1:1:NEW 1 Oct 2020)</ns0:note></ns0:figure> </ns0:body> "

"September 28, 2020 Corresponding Author: Avi Kaye 6069 S. Biscay St. Aurora, CO 80016 Email: [email protected] Phone: 303-919-3690 Dear PeerJ Editors, We greatly appreciate your exhaustive review of our manuscript and for providing constructive revision suggestions to strengthen it. We thoroughly considered your comments and addressed your concerns in the updated manuscript. The most significant changes include a figure to display TCZ’s mechanism of action, information on ICU admission in each study, statements clarifying questions from the editors and recommendations to improve future clinical trials. The lines referenced in the rebuttal letter refer to the review PDF, not the most recent manuscript. We believe that the manuscript now meets the standards for publication in the PeerJ journal. Avi Kaye, B.S Candidate in Human Biology at Stanford University Dr. Robert Siegel, Professor of Microbiology & Immunology and Human Biology at Stanford University Reviewer 1 (Manasi Kamat) Basic reporting 1. The language used is error-free and unambiguous, conveying the necessary scientific details. 2. Reference list is clear, correct and sufficient. Some comments listed below would need more references to be added, but the current ones are ideal for the written content. 3. The authors have mentioned how 3/4th of the severe cases present with IL6 mediated respiratory failure. Alluding to the remaining 1/4th cases (with appropriate references) would be helpful. According to the same source, the remaining ¼ of cases present with IL-1β mediated macrophage activation. This was addressed at the end of the sentence on lines 89-90. 4. A schematic showing the mechanism of action of TCZ and its downstream effects would be useful. We thought this was an excellent suggestion to clarify the logic of IL-6 inhibitors. A schematic (updated “Figure 2”) was added to illustrate the mechanism and downstream effects of TCZ. 5. Please clarify what “potentially repeating patient data” indicates. Phrasing changed to “patient data that may be used in another studies” on 185-186. Experimental design 6. This being a review article, it addresses all the relevant clinical trials and primary research articles to arrive to scientific conclusions about efficacy of TCZ in COVID-19 patients. 7. Are there any trials of TCZ on younger patients suffering from COVID-19? If so, since they weren’t in this review analysis, what were the results and maybe mentioning those in the discussion section would be useful. We did not identify any trials specifically investigating the efficacy of TCZ in younger patients with COVID-19. 8. Why did the one mentioned study derive results showing increased mortality with TCZ treatment compared to SOC? Please elaborate. The one study that did not find reduced mortality with TCZ demonstrated no benefit of the added treatment. Therefore, we clarified that Patel et al. “showed no benefit” in the results section (line 226) and the discussion (line 272). Validity of the findings COVID-19 affected individuals and results/conclusions arising from the studies are still in nascent phases and ever-changing based on recent research. Hence even though this review suggests that TCZ could be a potentially beneficial treatment for patients, the authors have highlighted the caveats of the research. Data is statistically analysed and speculations have been clearly outlined. Comments for the author The manuscript is a great attempt at bringing together all the relevant research for IL6 inhibitors, especially TCZ in COVID treatment. It would be of importance to the field, as we slowly know more about this virus and its deleterious effects. Reviewer 2 Basic reporting No comments Experimental design No comments Validity of the findings No comments Comments for the author Detailed systematic review! Reviewer 2 did not identify a concern that we needed to address. Reviewer 3 Basic reporting No comment Experimental design No comment Validity of the findings No comment Comments for the author This systematic review is based on efficacy of IL-6 inhibitor Tocilizumab in Covid-19 patients. The authors presented the review based on 16-case control studies gathered from PubMed search. Considering the ongoing pandemic COVID-19 and lack of effective therapies to treat patients affected with severe disease, careful evaluation of clinical data on existing/new anti-viral drugs is important and much needed. However, currently couple of studies are published on efficacy of Tocilizumab in COVID-19 patients as systematic review and meta-analysis (PMID: 32712333, 32713784 & Aziz M et al 2020). The two most recent systematic review and meta-analysis on TCZ (PMID 32713784 and Aziz et al.) were addressed and our paper’s specific contribution was clarified (lines 172-174). The authors can improve the article with following… 1) The introduction part of the review is somewhat lengthy and covered topics which doesn’t have scope in the present review. Authors can present a short and informative introduction that largely talks about IL-6 role in physiology, dysregulation of immune pathways, IL-6 role in COVID-19 pathology, mechanism of Tocilizumab action and known adverse effects (other IL-6 inhibitors). Instead of presenting the topics on other anti-viral drugs, authors can cite relevant existing literature. We appreciate the reviewer’s suggestion to reduce the scope of the paper’s introduction. It is our opinion that the information is important to justify the logic of using IL- 6 inhibitors and explaining why current therapeutic options are insufficient. We are willing to delete the discussion of antivirals (lines 114-120) if you believe that it does not fit the topic of our paper. 2) Authors did not talk about IL-6 levels and other biomarkers (post-treatment levels, if any)? We thoroughly investigated reported IL-6 levels in each of the studies we reviewed. Unfortunately, there was insufficient data provided to systematically analyze pre- and post-treatment IL-6 levels from the papers. Only one controlled study (Colaneri et al.) provided both baseline and post-treatment values for any biomarkers such as IL-6, CRP etc. Because we agree that these values are important, we note that tracking biomarkers before and after treatment is a necessary practice to improve ongoing clinical trials in our discussion and conclusions. 3) Discussion can be improved by adding points that can help ongoing/future clinical trials to better assess drug efficacy. We agree that this is an important point in our paper, so we clarified our recommendations in the discussion (line 292) and added explicit points into the conclusions: “In light of this analysis, several factors would facilitate the evaluation of TCZ as a therapeutic for the immune dysregulation associated with COVID-19: 1) Publication of the results from unpublished clinical trials, 2) Completion of additional randomized controlled trials especially where the potentially complicating effects of dexamethasone may be ruled out, 3) Comparison of TDZ findings with results assessing other IL-6 inhibitors such as sarilumab or siltuximab, and 4) Completion of additional metabolic studies measuring the levels of immune mediators and biomarkers in SARS-CoV-2 infected animal models and humans treated with TCZ. It will also be useful to assess the possibility of therapeutic synergy between antiviral agents such as remdesivir and TCZ.” "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Many teleost fishes undergo natural sex change, and elucidating the physiological and molecular controls of this process offers unique opportunities not only to develop methods of controlling sex in aquaculture settings, but to better understand vertebrate sexual development more broadly. Induction of sex change in some sequentially hermaphroditic or gonochoristic fish can be achieved in vivo through social manipulation, inhibition of aromatase activity, and steroid treatment. However, the induction of sex change in vitro has been largely unexplored. In this study, we established an in vitro culture system for ovarian explants in serum-free medium for a model sequential hermaphrodite, the New Zealand spotty wrasse (Notolabrus celidotus). This culture technique enabled evaluating the effect of various treatments with 17&#946;-estradiol (E 2 ), 11-ketotestosterone (11KT) or cortisol (CORT) on spotty wrasse ovarian architecture for 21 days. A quantitative approach to measuring the degree of ovarian atresia within histological images was also developed, using pixel-based machine learning software. Ovarian atresia likely due to culture was observed across all treatments including no-hormone controls, but was minimised with treatment of at least 10 ng/mL E 2 . Neither 11KT nor CORT administration induced proliferation of spermatogonia (i.e. sex change) in the cultured ovaries indicating culture beyond 21 days may be needed to induce sex change in vitro. The in vitro gonadal culture and analysis systems established here enable future studies investigating the paracrine role of sex steroids, glucocorticoids and a variety of other factors during gonadal sex change in fish.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>In most vertebrates, sex is fixed after birth and remains the same throughout life <ns0:ref type='bibr' target='#b27'>(Piferrer &amp; Guiguen, 2008;</ns0:ref><ns0:ref type='bibr' target='#b2'>Barske &amp; Capel, 2008;</ns0:ref><ns0:ref type='bibr' target='#b33'>Todd et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b19'>Liu et al., 2017)</ns0:ref>. Some teleost fishes, however, develop as one sex but remain able to change to the other during adulthood <ns0:ref type='bibr' target='#b11'>(Godwin, 2009;</ns0:ref><ns0:ref type='bibr' target='#b18'>Lamm et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b19'>Liu et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b10'>Gemmell et al., 2019)</ns0:ref>. In vitro culture manipulations enable precise monitoring of the effects of biotic and abiotic factors (e.g. temperature, pH, microplastics, etc.) on gonadal development, but are underutilised in aquaculture and fisheries research. Ex vivo approaches show particular potential to study responses of gonadal tissue to sex steroids and other hormones (e.g. glucocorticoids). Testicular and ovarian organ culture has been experimentally achieved with a handful of gonochoristic (fixed separate sexes) fish species <ns0:ref type='bibr' target='#b4'>(Carragher &amp; Sumpter, 1990;</ns0:ref><ns0:ref type='bibr' target='#b22'>Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., 2006</ns0:ref><ns0:ref type='bibr' target='#b25'>Ozaki et al., , 2019;;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>.</ns0:p><ns0:p>However, with the exception of a conference abstract on the effects of several hormones on the gonadal architecture of sex-changing three-spot wrasse (Halichoeres trimaculatus <ns0:ref type='bibr' target='#b34'>(Todo et al., 2008)</ns0:ref>), explant culture systems for the study of sex change in fish remain unreported. Gonadal remodelling in sex-changing fishes occurs in adulthood and, therefore, the study of sexdetermining mechanisms is not limited to egg or larval stages as in other vertebrates. Tissue and organoid in vitro culture allows evaluating multiple factors in a large number of replicates in wellcontrolled and standardised experimental conditions, while reducing animal usage and handling compared with in vivo manipulations <ns0:ref type='bibr' target='#b22'>(Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b30'>Schulz et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b25'>Ozaki et al., 2019)</ns0:ref>.</ns0:p><ns0:p>The aim of this study was to develop an organ culture system for a sequential hermaphrodite to establish a new model for investigating the molecular and cellular basis of sex reversal in the gonad. We chose the New Zealand spotty wrasse (Notolabrus celidotus), a common and widespread marine species, to evaluate the effects of a suite of hormonal factors on gonadal architecture. In this diandric protogynous hermaphroditic species, sex change is socially regulated, whereby the removal of the dominant male from the social group induces sex change in a resident female <ns0:ref type='bibr' target='#b32'>(Thomas et al., 2019)</ns0:ref>. Spotty wrasses exhibit sexual dimorphism with alternative male phenotypes. Initial phase (IP) individuals consist of females and primary males which develop directly from a juvenile stage with a bipotential gonad <ns0:ref type='bibr' target='#b6'>(Choat, 1965)</ns0:ref>. Both females and IP males can sex or role change, respectively, to become terminal phase (TP) males <ns0:ref type='bibr' target='#b6'>(Choat, 1965)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed While natural sex change involves modifications at the behavioural, physiological, molecular and anatomical levels <ns0:ref type='bibr' target='#b33'>(Todd et al., 2016)</ns0:ref>, herein, the term 'sex change' is used to refer to the development of testicular tissue in female gonads. The effect of different doses of steroid hormones 17&#946;-estradiol (E 2 ), 11-ketestosterone (11KT) or cortisol (CORT) on spotty wrasse ovary explants was evaluated in vitro to elucidate their effect on gonadal sex change in this species. Controls (C, no steroid) were used to determine the viability of the culture system and the potential triggering of gonadal sex change in the absence of exogenous factors. Treatment with E 2 was used to evaluate its role in the successful maintenance of spotty wrasse ovaries in vitro. 11KT, the most potent androgen in fish <ns0:ref type='bibr' target='#b12'>(Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>, was employed to explore its potential to induce sex change (i.e. proliferation of spermatogonia) in cultured spotty wrasse ovaries. In addition, treatment with the stress hormone CORT was predicted to induce oocyte degeneration through the inhibition of aromatase <ns0:ref type='bibr' target='#b9'>(Fernandino et al., 2013)</ns0:ref>, and it was hypothesised that this effect alone may trigger the female-to-male gonadal transformation. A strong effect of CORT, the main glucocorticoid in fish, has been linked to fish sex determination and sex change in several studies <ns0:ref type='bibr' target='#b13'>(Hayashi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b35'>Yamaguchi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Kitano et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b23'>Nozu &amp; Nakamura, 2015;</ns0:ref><ns0:ref type='bibr' target='#b21'>Miller et al., 2019)</ns0:ref>. In gonochoristic species such as the Japanese flounder (Paralichthys olivaceus) or medaka (Oryzias latipes), in vivo CORT treatment has been reported to mediate masculinisation of genetic females by promoting androgen production and apoptosis in the gonad, inducing maleness <ns0:ref type='bibr' target='#b13'>(Hayashi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b35'>Yamaguchi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Kitano et al., 2012)</ns0:ref>. Likewise, recent studies suggest a pivotal role of CORT in driving natural sex change in sequentially hermaphroditic teleosts <ns0:ref type='bibr' target='#b23'>(Nozu &amp; Nakamura, 2015;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chen et al., 2020)</ns0:ref>. It has been hypothesised that CORT, together with genetic (e.g. amh upregulation) and epigenetic (e.g. DNA-methylation) factors could lead to suppression of the female genetic network in favour of enhancing male pathway gene expression to trigger sex change <ns0:ref type='bibr' target='#b33'>(Todd et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b19'>Liu et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b10'>Gemmell et al., 2019)</ns0:ref>. This is the first study to evaluate the effects of CORT in an in vitro gonadal culture system of a sex-changing fish.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Sample collection</ns0:head><ns0:p>Five IP spotty wrasse individuals ranging from 160 -230 mm total length were captured by hook and line at the Portobello Marine Laboratory, Department of Marine Science, University of Otago, New Zealand during May 2019. Live fish were immediately transported to cell culture facilities at the Department of Zoology, University of Otago, where fish were euthanised by overdose in benzocaine (0.3 g/L) and gonads dissected within 1.5 h of capture. At this time, gonadal fragments from each fish were preserved for histological analysis as day 0 reference tissues (D0). Gonads fixed in 4% PFA were processed for routine embedding in paraffin (Otago Histology Services Unit, Department of Pathology, University of Otago). Fish were captured and manipulated with approval from the University of Otago Animal Ethics Committee (AUP-18-247), and in accordance with New Zealand National Animal Ethics Advisory Committee guidelines.</ns0:p></ns0:div> <ns0:div><ns0:head>Ovarian organ culture technique</ns0:head><ns0:p>In vitro culture of spotty wrasse ovary was carried out using the floating tissue culture method originally described by <ns0:ref type='bibr' target='#b22'>(Miura et al., 1991)</ns0:ref> with minor modifications. For each of 5 IP fish, freshly removed ovaries were cut into pieces of approximately 2 -3 mm and ovarian fragments were placed on floats of 1.5% agarose covered with a nitrocellulose membrane, each fragment in a well that further contained 500 &#956;l of medium in a 24-well plastic tissue culture dish. The basal medium consisted of Leibovitz L-15 medium supplemented with 0.5% bovine serum albumin, 10,000 U/L penicillin, 10 mg/L streptomycin and 1 &#181;g/mL porcine insulin adjusted to pH 7.4. Ovarian fragments were exposed to E 2 , 11KT or CORT for 21 days at 16 &#8451;. The effect of various concentrations of each hormone (1, 10 or 100 ng/mL) was examined. Steroids were first dissolved in EtOH and then diluted with the medium (0.1% EtOH), which was changed every 7 days. Controls (no steroid, 0.1% EtOH) were included. At the end of the experiment, tissues were fixed in 4% PFA and processed for routine paraffin embedding.</ns0:p></ns0:div> <ns0:div><ns0:head>Image analysis</ns0:head><ns0:p>PeerJ reviewing <ns0:ref type='table'>PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:ref> Manuscript to be reviewed An open source pixel-based machine learning trainable classifier, Ilastik (v. 1.3.2) (https://www.ilastik.org) <ns0:ref type='bibr' target='#b31'>(Sommer et al., 2011)</ns0:ref>, was employed to discriminate cell types in the gonadal images captured using brightfield light microscopy. Approximately 85 previtellogenic oocytes (PVO), from a subset of 5 images, were used to train Ilastik. Atretic oocytes, stromal tissue and background were labelled separately with a paintbrush-style tool using the Pixel Classification module. The training process was refined iteratively, updating pixel classifications until object types were accurately discriminated in the segmentation maps, as described in <ns0:ref type='bibr' target='#b20'>(Logan et al., 2016)</ns0:ref> (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). The Ilastik classifier files were extracted in HDF5 (Hierarchical Data Format 5) format</ns0:p><ns0:p>and loaded into open source image processing software ImageJ (v. 2.0.0-rc-69/1.52p), where the different object types were each assigned a different threshold. ImageJ 'Analyse particles' function was used to calculate the total relative surface area occupied by non-atretic PVO and measure the total surface of the tissue section in each image. 'Remove outliers' function was used to remove any outlier &lt; 10 pixels from analysis and 'Fill holes' function was used to mark PVO nuclei in the exceptional occasions when Ilastik failed to correctly recognise these.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Generalised linear mixed models were used to assess the effect of steroid (i.e. E 2 , 11KT or CORT) treatments on the proportion of total tissue surface occupied by non-atretic PVO ('PVO_area') in comparison to the total area of each ovarian section ('total_area'), using individual fish as a random effect. Models followed a negative binomial distribution and were performed using the following command line in R (v. 3.6.1): 'glmmPQL (PVO_area/total_area ~ Treatment, random = ~1|Fish, weights = total_area, family = negative.binomial (theta = 1), data = dat)'. If significant differences between treatments were found, Tukey's multiple comparison tests were performed to determine where the significance lay between treatments. All statistical analyses were performed using R <ns0:ref type='bibr'>(Core Team, 2013)</ns0:ref>. Results are expressed as means &#177; SEM of 5 biological replicates, except for the control (n = 15) or unless stated otherwise.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Histological analysis of D0 reference tissues confirmed that all 5 IP spotty wrasse captured were female. Visual inspection of gonadal sections under light microscopy revealed that oocyte degeneration was present in all experimental groups, including controls, but not the D0 reference tissues. However, initiation of spermatogenesis was not observed in any group, and no male structures were identified.</ns0:p><ns0:p>Atresia of PVO appeared to be a background effect of the in vitro culture <ns0:ref type='bibr' target='#b34'>(Todo et al., 2008)</ns0:ref>. This was reflected in reduced oocyte surface area in controls compared to D0 reference tissues, with the proportion of total tissue area filled with non-atretic PVO being significantly higher in D0 (73.6 &#177; 2.4 %) versus control ovaries (31.8 &#177; 3.7 %) (p-value &lt; 0.001) (Figure <ns0:ref type='figure' target='#fig_4'>2A</ns0:ref>). Estrogen treatment reduced the incidence of oocyte degeneration. A greater surface area covered by nonatretic PVO was observed in ovaries treated with the two highest E 2 doses (10 ng/mL, 51.1 &#177; 5.1 %; 100 ng/mL, 50.7 &#177; 9.8 %) (Figure <ns0:ref type='figure' target='#fig_4'>2A</ns0:ref>). Differences between the control ovaries and those treated with 100 ng/mL of E 2 were significant (p-value &lt; 0.05). Therefore, exogenous E 2 administration to cultured ovaries can effectively, albeit partially, prevent oocyte degeneration (i.e. increased surface area occupied by non-atretic PVO) associated with explant culture. This reinforces observations in female coho salmon (Oncorhynchus kisutch), in which in vivo fadrozole treatment (an aromatase inhibitor) induced reductions in E 2 and a higher incidence of associated atresia <ns0:ref type='bibr' target='#b0'>(Afonso et al., 1999)</ns0:ref>. In future culture systems, the addition of growth factors or the use of an artificial extracellular matrix <ns0:ref type='bibr' target='#b14'>(Kim et al., 2011)</ns0:ref> could be investigated to further mitigate oocyte degeneration as a consequence of culture.</ns0:p><ns0:p>Unexpectedly, there was no histological evidence that treatment with 11KT induced sex change in spotty wrasse ovaries in vitro. The proportion of non-atretic PVO found in the ovarian tissues was not dose-dependent and did not significantly vary between any of the 11KT-treated ovaries.</ns0:p><ns0:p>Although all 11KT-treated ovary explants showed a much lower percentage of PVO compared to total tissue area than D0 ovaries, values were comparable to the control (Figure <ns0:ref type='figure' target='#fig_4'>2B</ns0:ref>). The proportion of non-atretic PVO in 11KT-treated ovarian tissues was also unaffected by dose (1 ng/mL, 24.0 &#177; 10.7 %; 10 ng/mL, 25.6 &#177; 9.1 %; 100 ng/mL 21.6 &#177; 9.7 %).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Successful induction of spermatogenesis via in vitro androgen treatment in cultured testes has been demonstrated in Japanese eel (Anguilla japonica), in which Sertoli cells were activated by 11KT to stimulate the spermatogenic cascade <ns0:ref type='bibr' target='#b22'>(Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., 2006)</ns0:ref>. The role of 11KT as a promoter of spermatogenesis in ovaries of sex-changing fish has also been demonstrated <ns0:ref type='bibr' target='#b3'>(Bhandari et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b34'>Todo et al., 2008)</ns0:ref>. In the protogynous honeycomb grouper (Epinephelus merra), in vivo implantation of 11KT into pre-spawning females caused masculinisation in all fish <ns0:ref type='bibr' target='#b3'>(Bhandari et al., 2006)</ns0:ref>. Moreover, in vivo induction of sex change was also accomplished in the bidirectional sex changing Pseudolabrus sieboldi, both by administering sustained-release capsules containing 11KT to females and those with E2 to male individuals <ns0:ref type='bibr' target='#b24'>(Ohta et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Implants containing 11KT induced changes in body colour in the females too <ns0:ref type='bibr' target='#b24'>(Ohta et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Likewise, in vitro 11KT treatment of ovarian explants from protogynous three-spot wrasse induced the formation of presumed spermatogenic crypts and sperm within 2 -3 weeks <ns0:ref type='bibr' target='#b34'>(Todo et al., 2008)</ns0:ref>.</ns0:p><ns0:p>However, three-spot wrasses live in tropical waters, and sex change in captive populations has been observed in under 40 days <ns0:ref type='bibr' target='#b17'>(Kuwamura et al., 2007)</ns0:ref>, around half that required in spotty wrasse (60 -70 days <ns0:ref type='bibr' target='#b32'>(Thomas et al., 2019)</ns0:ref>). In the current study, it may be that spotty wrasse ovaries were not cultured long enough for 11KT to promote male germ cell proliferation, or that 11KT alone is insufficient to induce gonadal transformation in this species. It has been hypothesised that endogenous E 2 levels need to decrease below a physiological threshold for oocyte maintenance to become jeopardised and androgen levels to effectively induce proliferation of bipotential gonial stem cells in the ovary <ns0:ref type='bibr' target='#b3'>(Bhandari et al., 2006)</ns0:ref>. Co-treatment of 11KT with an aromatase inhibitor (e.g. fadrozole) should be performed in the future to further characterise the triggering of in vitro gonadal restructuring, potentially establishing this technique as a model for sex change research.</ns0:p><ns0:p>It was hypothesised that CORT treatment would promote oocyte degeneration, following evidence that CORT may mediate sex change in fish by inhibiting aromatase transcription and promoting androgen production <ns0:ref type='bibr' target='#b9'>(Fernandino et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b12'>Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>. However, no significant differences were observed in the proportion of non-atretic PVO between the controls and ovaries treated with 1 and 10 ng/mL of CORT (43.3 &#177; 7.7 % and 42.9 &#177; 6.2 %, respectively) (Figure <ns0:ref type='figure' target='#fig_4'>2C</ns0:ref>). While a significant increase in the percentage area of PVO was detected between the controls and 100 ng/mL CORT treatment group (39.5 &#177; 18.0 %) (p-value &lt; 0.005), this may reflect variability between fish and the effect of smaller sample size (n=4, as one biological replicate was PeerJ reviewing PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed lost during histological processing due to technical issues) on the statistical analysis. Therefore, these differences were not deemed a direct effect of CORT administration to the ovaries. Cortisol has been observed to negatively impact ovarian development by suppressing E 2 secretion in cultured rainbow trout (Oncorhynchus mykiss) follicles <ns0:ref type='bibr' target='#b4'>(Carragher &amp; Sumpter, 1990)</ns0:ref>. However, in that same study, <ns0:ref type='bibr' target='#b4'>Carragher &amp; Sumpter (Carragher &amp; Sumpter, 1990)</ns0:ref> argued that there may be a time lag for CORT suppressive effects to become evident during in vitro ovarian culture. Longerterm culture of spotty wrasse ovaries may be needed to observe any effects of CORT on ovarian structure, a finding in keeping with our findings for 11KT.</ns0:p><ns0:p>Cortisol also failed to initiate obvious sex change in cultured spotty wrasse ovaries, with no evidence of male tissues observed in our ovarian sections. It may be that CORT alone is ineffective in inducing spotty wrasse sex change, or that the experiment needs technical adjustments (e.g. longer duration) for the successful induction of sex change in spotty wrasse. Low doses of CORT (i.e. 0.01 -10 ng/mL) were reported to promote testicular development in Japanese eel and pejerrey (Odontesthes bonariensis) by increasing the production of 11KT, which, in turn, induced spermatogonial proliferation <ns0:ref type='bibr' target='#b26'>(Ozaki et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>. Interestingly, higher doses of CORT (i.e. 100 ng/mL) had the opposite effect and inhibited 11KT synthesis in these species <ns0:ref type='bibr' target='#b26'>(Ozaki et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>. Although measurement of 11KT concentration in the cultured media should be performed to evaluate such effects on spotty wrasse ovaries, absence of spermatogonia in the cultured ovaries suggests 11KT levels were not significantly affected. The possible metabolic conversion of CORT into 11KT and other 11oxygenated androgens has also been suggested in several teleosts <ns0:ref type='bibr' target='#b15'>(Kime, 1978;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., 2006)</ns0:ref>, as there is cross-talk between the glucocorticoid and androgen synthesis pathways <ns0:ref type='bibr' target='#b1'>(Arterbery, Deitcher &amp; Bass, 2010;</ns0:ref><ns0:ref type='bibr' target='#b12'>Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>. However, the mechanisms underlying 11KT synthesis are still not fully understood and a potential CORT-to-11KT direct conversion remains to be clarified <ns0:ref type='bibr' target='#b29'>(Schulz, 1986;</ns0:ref><ns0:ref type='bibr' target='#b12'>Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>. This work constitutes the first-ever in vitro organ culture system in a temperate sex-changing teleost, and opens doors for future work in a wide variety of hermaphroditic fish.</ns0:p><ns0:p>Although the current experiment could not induce any observable female-to-male restructuring of spotty wrasse gonads in culture, ovarian tissues were successfully maintained in culture for 21 days. Such long-term culture of fish organ explants has only been accomplished in a few instances <ns0:ref type='bibr' target='#b4'>(Carragher &amp; Sumpter, 1990;</ns0:ref><ns0:ref type='bibr' target='#b22'>Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., 2006</ns0:ref><ns0:ref type='bibr' target='#b25'>Ozaki et al., , 2019;;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Furthermore, this work validates the use of open-source machine learning software Ilastik to reliably recognise cell types in fish gonads, with enormous potential in future studies. The image analysis strategy presented here can be adapted and applied to other studies where accurate cell segmentation is required.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>This study developed an organ culture system for gonadal tissue explants in New Zealand spotty wrasse, providing an in vitro system for investigating control and perturbation of gonadal sex change in this and other species. Moreover, we successfully integrated Ilastik software to score ovarian architecture, an application which provides great scope for future studies on effects of steroids on ovarian development of spotty wrasse, or fish in general, in vitro. Although culture was associated with a degree of PVO degeneration, it was shown that administration of at least 10 ng/mL E 2 can partially mitigate tissue degeneration during culture. However, neither 11KT nor CORT treatment successfully accelerated ovarian atresia or induced spermatogenic proliferation.</ns0:p><ns0:p>Future in vitro manipulations should explore longer cultures and/or co-treatment of multiple steroids to analyse potential antagonistic or synergistic effects. This culture system will facilitate future investigations of ovarian and testicular trans-differentiation in sex-changing fish. For example, an in vitro system would be especially amenable to the use of small interfering RNAs to knock down expression of specific candidate genes or application of methylation inhibitors to further investigate the genetic and epigenetic cascade underlying sex change. In vitro approaches also offer valuable ethical benefits over experiments with live fish, enabling the application of the 3Rs guiding principles: reduce, replace and refine <ns0:ref type='bibr' target='#b28'>(Russell &amp; Burch, 1959)</ns0:ref>. Manuscript to be reviewed Bars show the surface area of all non-atretic PVO over total tissue area (expressed as %). Same day 0 reference tissues and controls (day 21) were used for all treatments. Results are shown as the mean &#177; SEM. Different letters denote a statistically significant difference between groups, while experimental groups with no significant differences share the same letter. Samples sizes: D0 n = 5; C (no steroid) n = 15; E 2 , 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; 11KT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; CORT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 4. Abbreviations: 11-ketotestosterone (11KT), control (C), cortisol (CORT), day 0 reference tissues (D0), 17&#946;-estradiol (E 2 ).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48900:1:1:NEW 25 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Bars show the surface area of all non-atretic PVO over total tissue area (expressed as %).</ns0:p><ns0:p>Same day 0 reference tissues and controls (day 21) were used for all treatments. Results are shown as the mean &#177; SEM. Different letters denote a statistically 508 significant difference between groups, while experimental groups with no significant 509 differences share the same letter. Samples sizes: D0 n = 5; C (no steroid) n = 15; E 2 , 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; 11KT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; CORT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 4. Abbreviations: 11-ketotestosterone (11KT), control (C), cortisol (CORT), day 0 reference tissues (D0), 17&#946;-estradiol (E 2 ).</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Abbreviations</ns0:head><ns0:label /><ns0:figDesc>Abbreviations</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> </ns0:body> "

"Editorial Comments Editor The paper focuses on a very interesting topic in order to try to establish an in vitro culture system for ovarian tissue of protogynous wrasse. The culture system is for 21 days (could be considered non-stable among the different specimens). Furthermore, only histological data have been used in this study and other techniques should be considered in order to better demonstrate the results. Besides this, some references are needed in order to reinforce many sentences of the manuscript. For all these reasons, this paper seems to be a preliminary version that needs to be deeply improved. To the editor, Thank you for inviting us to resubmit a revision of our manuscript to PeerJ. We are also grateful for the editor and two reviewers for their time in reviewing our manuscript and providing detailed and constructive feedback. All comments and suggestions have been carefully considered and addressed, which has improved the quality and clarity of our manuscript. Nevertheless, it is our impression that reviewers may have misunderstood the exact intent of the study. This is, to establish a gonadal explant culture system, seldom accomplished before, for the potential future application of this technique in the study of teleost sex change, among other fields, for example in the study of cell dedifferentiation and lineage reprogramming in sequential hermaphrodites. The inclusion of the cellular and hormonal data provided in this work solely aimed to validate the development of such culture technique, and not to provide a detailed account of the changes tissues undergo during culture. These data also support the successful integration of Ilastik software to score ovarian architecture, which provides great scope for future studies on effects of steroids on ovarian architecture of spotty wrasse, or fish in general, in vitro. Details of the revisions made are shown below each corresponding reviewer comment. The revised manuscript has been uploaded both with tracked changes and as a clean new version with no tracked changes. Thank you for considering our re-submission and we look forward to hearing from you. Kind regards, Alexander Goikoetxea Reviewers Comments Reviewer 1 Comments for the author: In this report, authors try to establish an in vitro culture system for ovarian tissue of protogynous wrasse. This culture system (for 21 days) is used to evaluate the effects of different steroids on female-to-male sex change. Unfortunately, only histological data used in this study. It is very difficult to observe the differences on cellular and molecular levels. We thank the reviewer for these comments. Unfortunately, we believe there may have been some misunderstanding as to the core purpose of our paper. The main objective of the work described in this manuscript was to establish a system in which fish ovaries can be maintained in culture to enable future manipulative experiments, not to document the cellular and molecular events that occur during this transformation. The histological and steroidal data we have provided in the paper prove we have successfully achieved that aim in our study. In addition, evidence from other hermaphroditic teleost species (bluehead wrasse, grouper, among others) of clear histological changes occurring during the different stages of sex change in these species, respectively, demonstrates the value, utility and suitability of using histology to document such events. 1. For cell proliferating activity, IHC staining with a proliferating marker PCNA or proliferating cells labeling with the BrdU could help authors to evaluate the effect of 11KT and cortisol. We thank the reviewer for this helpful suggestion. We agree that the applications of these additional techniques, including IHC staining, would be of value for documenting further the events that occur during gonadal transformation during sex change. However, documenting that process was not the central premise of this work, rather it was to show that in vitro organ culture was possible as described above. Further, the New Zealand spotty wrasse is an emerging model for sex change research relatively understudied. As such, unlike more established systems, there are currently no antibodies available for this species, making the additional experiments suggested impossible at this time. 2. For male differentiation, several sex-dimorphic expression genes (Dmrt1, Amh, Gsdf) could help authors to evaluate the effect of 11KT and cortisol. We acknowledge that the addition of molecular data, especially that of ‘stress and sex’ related genes, could provide further information on the sex-change process. However, the present manuscript focuses on the development of an organ culture system for a sequential hermaphrodite and establishment of an experimental system for in vitro manipulations of gonadal sex change. Therefore, the inclusion of gene expression analysis does not fit the scope of the present manuscript. It is however the next logical step once the ovary-to-testis in vitro transformation for New Zealand spotty wrasse is accomplished. 3. How authors to define the difference between the oogonia and spermatogonia in bisexual gonad? Authors have to add the figure to descript the characteristics of oogonia and spermatogonia in the early stage of sex change. We thank the reviewer for their thoughts on the histological analysis. A figure depicting spotty wrasse spermatogonia has not been included in the manuscript as it was considered it may lead to confusion to the readership of PeerJ, due to two main reasons detailed below: • As shown by our results, neither 11KT nor CORT treatment induced spermatogenic proliferation in the cultured ovaries, and therefore, presence of spermatogonia or ‘bisexual gonads’ per se was not detected among our samples during analysis. • In spite of this, as proliferation of spermatogenic cysts during spotty wrasse sex change is not visible until the mid-transitioning stage (see Thomas et al., 2019), individuals in an early stage of transition (ET) may exist that lack spermatogonia in their transforming gonads. However, there is strong evidence that ET gonads have a significantly higher degree of atresia of previtellogenic oocytes (PVO), often accompanied by melanomacrophages and yellow brown bodies, compared to females not undergoing sex change. Such differences were not observed between our control females and those treated with 11KT or CORT, which led us to believe that the process of sex change in these fish was not triggered by the hormonal treatments applied. 4. In Figure 2, the letters used to indicate significant difference is very difficult to understand. We thank the reviewer for highlighting this issue. The legend corresponding to Figure 2 has been updated to reflect this need for clarity, it now reads ‘Different letters denote a statistically significant difference between groups, while experimental groups with no significant differences share the same letter’. Reviewer 2 Basic reporting: 1) The reviewer feels that the cited references are slightly less. Adding more references might be informative for the Journal readers. For example, at line 58 and 60. Additional references have been added to the manuscript: • Line 58: Barske LA, Capel B. Blurring the edges in vertebrate sex determination. Curr Opin Genet Dev. 2008 Dec;18(6):499–505; Piferrer F, Guiguen Y. Fish gonadogenesis. Part II: Molecular biology and genomics of sex differentiation. Rev Fish Sci. 2008 Sep 15;16(sup1):35–55; Liu H, Todd E V, Lokman PM, Lamm MS, Godwin JR, Gemmell NJ. Sexual plasticity: a fishy tale. Mol Reprod Dev. 2017;84(2):171–94. • Line 60: Liu H, Todd E V, Lokman PM, Lamm MS, Godwin JR, Gemmell NJ. Sexual plasticity: a fishy tale. Mol Reprod Dev. 2017;84(2):171–94; Godwin JR. Social determination of sex in reef fishes. Semin Cell Dev Biol. 2009 May;20(3):264–70; Lamm MS, Liu H, Gemmell NJ, Godwin JR. The need for speed: neuroendocrine regulation of socially-controlled sex change. Integr Comp Biol. 2015;55(2):307–22. 2) line 74: The reference 11 is focused on in vitro culture for kidney development. And the reference 10 does not mention in vitro culture in detail. The reviewer feels that more suitable reference should be cited if possible. We thank the reviewer for their suggestion to cite more relevant references regarding the application of in vitro cultures for gonadal development and the advantages of this technique. Existing references have been replaced with the following: Ozaki Y, Damsteegt EL, Setiawan AN, Miura T, Lokman PM. Expressional regulation of gonadotropin receptor genes and androgen receptor genes in the eel testis. Gen Comp Endocrinol. 2019 Sep; 280:123–33; Miura T, Yamauchi K, Takahashi H, Nagahama Y. Hormonal induction of all stages of spermatogenesis in vitro in the male Japanese eel (Anguilla japonica). Proc Natl Acad Sci USA. 1991 Jul 1;88(13):5774–8. 3) line 83 and line 218: The reference 12 seems to be doctoral thesis. The reviewer is not sure that this reference is appropriate. As the relevant research has now been published, reference to the doctoral thesis has been replaced with the peer-reviewed publication: Thomas JT, Todd E V, Muncaster S, Lokman PM, Damsteegt EL, Liu H, et al. Conservation and diversity in expression of candidate genes regulating socially-induced female-male sex change in wrasses. PeerJ. 2019 Jun 11;7:e7032. 4) line 104: Your introduction regarding the roles of cortisol need more detail. The roles on sex differentiation in gonochoristic species and the roles on sex changing species should be introduced, briefly. This might be informative for the Journal readers. We thank the reviewer for this suggestion. A brief overview of the role of cortisol during sex differentiation of gonochoristic fish and sex change of hermaphroditic species has been added, and this section now reads: ‘In gonochoristic species such as the Japanese flounder (Paralichthys olivaceus) or medaka (Oryzias latipes), in vivo CORT treatment has been reported to mediate masculinisation of genetic females by promoting androgen production and apoptosis in the gonad, inducing maleness (Hayashi et al., 2010; Yamaguchi et al., 2010; Kitano et al., 2012). Likewise, recent studies suggest a pivotal role of CORT in driving natural sex change in sequentially hermaphroditic teleosts (Nozu & Nakamura, 2015; Chen et al., 2020). It has been hypothesised that CORT, together with genetic (e.g. amh upregulation) and epigenetic (e.g. DNA-methylation) factors could lead to suppression of the female genetic network in favour of enhancing male pathway gene expression to trigger sex change (Todd et al., 2016; Liu et al., 2017; Gemmell et al., 2019).’ 5) line 212: Regarding the effect of 11KT, information on the related wrasse (J Exp Zool A Ecol Genet Physiol. 2012: 317:552-60.) might be helpful for discussion. We are grateful to the reviewer for suggesting the inclusion of such an interesting study to our discussion, which has now been incorporated: ‘In the protogynous honeycomb grouper (Epinephelus merra), in vivo implantation of 11KT into pre-spawning females caused masculinisation in all fish (Bhandari et al., 2006). Moreover, in vivo induction of sex change was also accomplished in the bidirectional sex changing Pseudolabrus sieboldi, both by administering sustained-release capsules containing 11KT to females and those with E2 to male individuals (Ohta et al., 2012). Implants containing 11KT induced changes in body colour in the females too (Ohta et al., 2012). Likewise, in vitro 11KT treatment of ovarian explants from protogynous three-spot wrasse induced the formation of presumed spermatogenic crypts and sperm within 2 – 3 weeks (Todo et al., 2008).’ Experimental design 1) line 137: Your organ culture technique in materials and methods need more detailed explanation. Information on the amount of medium in each well and the frequency of medium exchange would be important information. Those should be mentioned. We thank the reviewer for pointing out the missing details in our ‘Materials and Methods’ section which have now been added. This section now reads: ‘For each of 5 IP fish, freshly removed ovaries were cut into pieces of approximately 2 – 3 mm and ovarian fragments were placed on floats of 1.5% agarose covered with a nitrocellulose membrane, each fragment in a well that further contained 500 μl of medium in a 24-well plastic tissue culture dish. […] Steroids were first dissolved in EtOH and then diluted with the medium (0.1% EtOH), which was changed every 7 days.’ 2) For image analysis using Ilastik, information on sizes of previtellogenic oocytes would be important. Usually, ovarian tissue has various sizes of previtellogenic oocytes. Therefore, the reviewer is wondering whether previtellogenic oocytes at smaller size can be distinguished by Ilastik software? The training process we followed while using Ilastik was based on iterative refinement, until the different cell types of interest were distinguished with accuracy in the segmentation maps. Therefore, Ilastik was trained to detect previtellogenic oocytes of different sizes, so that those of a smaller size would not be overlooked. Comments for the Author Some histological photos at high resolution might be helpful to see the size of cells. Those photos are also necessary to understand whether non atretic cells are normal or not. Photos of D0 tissue might be needed, too. The reviewer would like to know if the sizes of remaining oocytes differ between each steroid treatment. We thank the reviewer for this suggestion. Figure 1 has been updated and is now at a higher resolution (11965 × 24811px), allowing better visualisation of cell types and sizes analysed with Ilastik. As for the measurement of PVOs size, this had been previously measured but no significant differences in PVO size were found between treatments. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Many teleost fishes undergo natural sex change, and elucidating the physiological and molecular controls of this process offers unique opportunities not only to develop methods of controlling sex in aquaculture settings, but to better understand vertebrate sexual development more broadly. Induction of sex change in some sequentially hermaphroditic or gonochoristic fish can be achieved in vivo through social manipulation, inhibition of aromatase activity, and steroid treatment. However, the induction of sex change in vitro has been largely unexplored. In this study, we established an in vitro culture system for ovarian explants in serum-free medium for a model sequential hermaphrodite, the New Zealand spotty wrasse (Notolabrus celidotus). This culture technique enabled evaluating the effect of various treatments with 17&#946;-estradiol (E 2 ), 11-ketotestosterone (11KT) or cortisol (CORT) on spotty wrasse ovarian architecture for 21 days. A quantitative approach to measuring the degree of ovarian atresia within histological images was also developed, using pixel-based machine learning software. Ovarian atresia likely due to culture was observed across all treatments including no-hormone controls, but was minimised with treatment of at least 10 ng/mL E 2 . Neither 11KT nor CORT administration induced proliferation of spermatogonia (i.e. sex change) in the cultured ovaries indicating culture beyond 21 days may be needed to induce sex change in vitro. The in vitro gonadal culture and analysis systems established here enable future studies investigating the paracrine role of sex steroids, glucocorticoids and a variety of other factors during gonadal sex change in fish.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>In most vertebrates, sex is fixed after birth and remains the same throughout life <ns0:ref type='bibr' target='#b28'>(Piferrer &amp; Guiguen, 2008;</ns0:ref><ns0:ref type='bibr' target='#b2'>Barske &amp; Capel, 2008;</ns0:ref><ns0:ref type='bibr' target='#b34'>Todd et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b20'>Liu et al., 2017)</ns0:ref>. Some teleost fishes, however, develop as one sex but remain able to change to the other during adulthood <ns0:ref type='bibr' target='#b11'>(Godwin, 2009;</ns0:ref><ns0:ref type='bibr' target='#b19'>Lamm et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b20'>Liu et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b10'>Gemmell et al., 2019)</ns0:ref>. In vitro culture manipulations enable precise monitoring of the effects of biotic and abiotic factors (e.g. temperature, pH, microplastics, etc.) on gonadal development, but are underutilised in aquaculture and fisheries research. Ex vivo approaches show particular potential to study responses of gonadal tissue to sex steroids and other hormones (e.g. glucocorticoids). Testicular and ovarian organ culture has been experimentally achieved with a handful of gonochoristic (fixed separate sexes) fish species <ns0:ref type='bibr' target='#b4'>(Carragher &amp; Sumpter, 1990;</ns0:ref><ns0:ref type='bibr' target='#b23'>Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b27'>Ozaki et al., 2006</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., , 2019;;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>.</ns0:p><ns0:p>However, with the exception of a conference abstract on the effects of several hormones on the gonadal architecture of sex-changing three-spot wrasse (Halichoeres trimaculatus <ns0:ref type='bibr' target='#b35'>(Todo et al., 2008)</ns0:ref>), explant culture systems for the study of sex change in fish remain unreported. Gonadal remodelling in sex-changing fishes occurs in adulthood and, therefore, the study of sexdetermining mechanisms is not limited to egg or larval stages as in other vertebrates. Tissue and organoid in vitro culture allows evaluating multiple factors in a large number of replicates in wellcontrolled and standardised experimental conditions, while reducing animal usage and handling compared with in vivo manipulations <ns0:ref type='bibr' target='#b23'>(Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b31'>Schulz et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., 2019)</ns0:ref>.</ns0:p><ns0:p>The aim of this study was to develop an organ culture system for a sequential hermaphrodite to establish a new model for investigating the molecular and cellular basis of sex reversal in the gonad. We chose the New Zealand spotty wrasse (Notolabrus celidotus), a common and widespread marine species, to evaluate the effects of a suite of hormonal factors on gonadal architecture. In this diandric protogynous hermaphroditic species, sex change is socially regulated, whereby the removal of the dominant male from the social group induces sex change in a resident female <ns0:ref type='bibr' target='#b33'>(Thomas et al., 2019)</ns0:ref>. Spotty wrasses exhibit sexual dimorphism with alternative male phenotypes. Initial phase (IP) individuals consist of females and primary males which develop directly from a juvenile stage with a bipotential gonad <ns0:ref type='bibr' target='#b6'>(Choat, 1965)</ns0:ref>. Both females and IP males can sex or role change, respectively, to become terminal phase (TP) males <ns0:ref type='bibr' target='#b6'>(Choat, 1965)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48900:2:0:NEW 15 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed While natural sex change involves modifications at the behavioural, physiological, molecular and anatomical levels <ns0:ref type='bibr' target='#b34'>(Todd et al., 2016)</ns0:ref>, herein, the term 'sex change' is used to refer to the development of testicular tissue in female gonads. The effect of different doses of steroid hormones 17&#946;-estradiol (E 2 ), 11-ketestosterone (11KT) or cortisol (CORT) on spotty wrasse ovary explants was evaluated in vitro to elucidate their effect on gonadal sex change in this species. Controls (C, no steroid) were used to determine the viability of the culture system and the potential triggering of gonadal sex change in the absence of exogenous factors. Treatment with E 2 was used to evaluate its role in the successful maintenance of spotty wrasse ovaries in vitro. 11KT, the most potent androgen in fish <ns0:ref type='bibr' target='#b12'>(Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>, was employed to explore its potential to induce sex change (i.e. proliferation of spermatogonia) in cultured spotty wrasse ovaries. In addition, treatment with the stress hormone CORT was predicted to induce oocyte degeneration through the inhibition of aromatase <ns0:ref type='bibr' target='#b9'>(Fernandino et al., 2013)</ns0:ref>, and it was hypothesised that this effect alone may trigger the female-to-male gonadal transformation. A strong effect of CORT, the main glucocorticoid in fish, has been linked to fish sex determination and sex change in several studies <ns0:ref type='bibr' target='#b13'>(Hayashi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b36'>Yamaguchi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Kitano et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b24'>Nozu &amp; Nakamura, 2015;</ns0:ref><ns0:ref type='bibr' target='#b22'>Miller et al., 2019)</ns0:ref>. In gonochoristic species such as the Japanese flounder (Paralichthys olivaceus) or medaka (Oryzias latipes), in vivo CORT treatment has been reported to mediate masculinisation of genetic females by promoting androgen production and apoptosis in the gonad, inducing maleness <ns0:ref type='bibr' target='#b13'>(Hayashi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b36'>Yamaguchi et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b16'>Kitano et al., 2012)</ns0:ref>. Likewise, recent studies suggest a pivotal role of CORT in driving natural sex change in sequentially hermaphroditic teleosts <ns0:ref type='bibr' target='#b24'>(Nozu &amp; Nakamura, 2015;</ns0:ref><ns0:ref type='bibr' target='#b5'>Chen et al., 2020)</ns0:ref>. It has been hypothesised that CORT, together with genetic (e.g. amh upregulation) and epigenetic (e.g. DNA-methylation) factors could lead to suppression of the female genetic network in favour of enhancing male pathway gene expression to trigger sex change <ns0:ref type='bibr' target='#b34'>(Todd et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b20'>Liu et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b10'>Gemmell et al., 2019)</ns0:ref>. This is the first study to evaluate the effects of CORT in an in vitro gonadal culture system of a sex-changing fish.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48900:2:0:NEW 15 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Sample collection</ns0:head><ns0:p>Five IP spotty wrasse individuals ranging from 160 -230 mm total length were captured by hook and line at the Portobello Marine Laboratory, Department of Marine Science, University of Otago, New Zealand during May 2019. Live fish were immediately transported to cell culture facilities at the Department of Zoology, University of Otago, where fish were euthanised by overdose in benzocaine (0.3 g/L) and gonads dissected within 1.5 h of capture. At this time, gonadal fragments from each fish were preserved for histological analysis as day 0 reference tissues (D0). Gonads fixed in 4% PFA were processed for routine embedding in paraffin (Otago Histology Services Unit, Department of Pathology, University of Otago). Fish were captured and manipulated with approval from the University of Otago Animal Ethics Committee (AUP-18-247), and in accordance with New Zealand National Animal Ethics Advisory Committee guidelines.</ns0:p></ns0:div> <ns0:div><ns0:head>Ovarian organ culture technique</ns0:head><ns0:p>In vitro culture of spotty wrasse ovary was carried out using the floating tissue culture method originally described by <ns0:ref type='bibr' target='#b23'>(Miura et al., 1991)</ns0:ref> with minor modifications. For each of 5 IP fish, freshly removed ovaries were cut into pieces of approximately 2 -3 mm and ovarian fragments were placed on floats of 1.5% agarose covered with a nitrocellulose membrane, each fragment in a well that further contained 500 &#956;l of medium in a 24-well plastic tissue culture dish. The basal medium consisted of Leibovitz L-15 medium supplemented with 0.5% bovine serum albumin, 10,000 U/L penicillin, 10 mg/L streptomycin and 1 &#181;g/mL porcine insulin adjusted to pH 7.4. Ovarian fragments were exposed to E 2 , 11KT or CORT for 21 days at 16 &#8451;. The effect of various concentrations of each hormone (1, 10 or 100 ng/mL) was examined. Steroids were first dissolved in EtOH and then diluted with the medium (0.1% EtOH), which was changed every 7 days. Controls (no steroid, 0.1% EtOH) were included. At the end of the experiment, tissues were fixed in 4% PFA and processed for routine paraffin embedding (Figure <ns0:ref type='figure' target='#fig_1'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Image analysis</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48900:2:0:NEW 15 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>An open source pixel-based machine learning trainable classifier, Ilastik (v. 1.3.2) (https://www.ilastik.org) <ns0:ref type='bibr' target='#b32'>(Sommer et al., 2011)</ns0:ref>, was employed to discriminate cell types in the gonadal images captured using brightfield light microscopy. Approximately 85 previtellogenic oocytes (PVO), from a subset of 5 images, were used to train Ilastik. Atretic oocytes, stromal tissue and background were labelled separately with a paintbrush-style tool using the Pixel Classification module. The following feature selection was applied: colour/intensity, &#963; 3 = 1.60; edge, &#963; 3 = 1.60; texture, &#963; 3 = 1.60. The training process was refined iteratively, updating pixel classifications until object types were accurately discriminated in the segmentation maps, as described in <ns0:ref type='bibr' target='#b21'>(Logan et al., 2016)</ns0:ref> (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). The Ilastik classifier files were extracted in HDF5 (Hierarchical Data Format 5) format and loaded into open source image processing software ImageJ (v. 2.0.0-rc-69/1.52p), where the different object types were each assigned a different threshold. ImageJ 'Analyse particles' function was used to calculate the total relative surface area occupied by non-atretic PVO and measure the total surface of the tissue section in each image. 'Remove outliers' function was used to remove any outlier &lt; 10 pixels from analysis and 'Fill holes' function was used to mark PVO nuclei in the exceptional occasions when Ilastik failed to correctly recognise these.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Generalised linear mixed models were used to assess the effect of steroid (i.e. E 2 , 11KT or CORT) treatments on the proportion of total tissue surface occupied by non-atretic PVO ('PVO_area') in comparison to the total area of each ovarian section ('total_area'), using individual fish as a random effect. Models followed a negative binomial distribution and were performed using the following command line in R (v. 3.6.1): 'glmmPQL (PVO_area/total_area ~ Treatment, random = ~1|Fish, weights = total_area, family = negative.binomial (theta = 1), data = dat)'. If significant differences between treatments were found, Tukey's multiple comparison tests were performed to determine where the significance lay between treatments. All statistical analyses were performed using R <ns0:ref type='bibr'>(Core Team, 2013)</ns0:ref>. Results are expressed as means &#177; SEM of 5 biological replicates, except for the control (n = 15) or unless stated otherwise.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:05:48900:2:0:NEW 15 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Histological analysis of D0 reference tissues confirmed that all 5 IP spotty wrasse captured were female. Visual inspection of gonadal sections under light microscopy revealed that oocyte degeneration was present in all experimental groups, including controls, but not the D0 reference tissues. However, initiation of spermatogenesis was not observed in any group, and no male structures were identified.</ns0:p><ns0:p>Atresia of PVO appeared to be a background effect of the in vitro culture <ns0:ref type='bibr' target='#b35'>(Todo et al., 2008)</ns0:ref>. This was reflected in reduced oocyte surface area in controls compared to D0 reference tissues, with the proportion of total tissue area filled with non-atretic PVO being significantly higher in D0 (73.6 &#177; 2.4 %) versus control ovaries (31.8 &#177; 3.7 %) (p-value &lt; 0.001) (Figure <ns0:ref type='figure' target='#fig_4'>2A</ns0:ref>). Estrogen treatment reduced the incidence of oocyte degeneration. A greater surface area covered by nonatretic PVO was observed in ovaries treated with the two highest E 2 doses (10 ng/mL, 51.1 &#177; 5.1 %; 100 ng/mL, 50.7 &#177; 9.8 %) (Figure <ns0:ref type='figure' target='#fig_4'>2A</ns0:ref>). Differences between the control ovaries and those treated with 100 ng/mL of E 2 were significant (p-value &lt; 0.05). Therefore, exogenous E 2 administration to cultured ovaries can effectively, albeit partially, prevent oocyte degeneration (i.e. increased surface area occupied by non-atretic PVO) associated with explant culture. This reinforces observations in female coho salmon (Oncorhynchus kisutch), in which in vivo fadrozole treatment (an aromatase inhibitor) induced reductions in E 2 and a higher incidence of associated atresia <ns0:ref type='bibr' target='#b0'>(Afonso et al., 1999)</ns0:ref>. In future culture systems, the addition of growth factors or the use of an artificial extracellular matrix <ns0:ref type='bibr' target='#b14'>(Kim et al., 2011)</ns0:ref> could be investigated to further mitigate oocyte degeneration as a consequence of culture.</ns0:p><ns0:p>Unexpectedly, there was no histological evidence that treatment with 11KT induced sex change in spotty wrasse ovaries in vitro. The proportion of non-atretic PVO found in the ovarian tissues was not dose-dependent and did not significantly vary between any of the 11KT-treated ovaries.</ns0:p><ns0:p>Although all 11KT-treated ovary explants showed a much lower percentage of PVO compared to total tissue area than D0 ovaries, values were comparable to the control (Figure <ns0:ref type='figure' target='#fig_4'>2B</ns0:ref>). The proportion of non-atretic PVO in 11KT-treated ovarian tissues was also unaffected by dose (1 ng/mL, 24.0 &#177; 10.7 %; 10 ng/mL, 25.6 &#177; 9.1 %; 100 ng/mL 21.6 &#177; 9.7 %).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:48900:2:0:NEW 15 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Successful induction of spermatogenesis via in vitro androgen treatment in cultured testes has been demonstrated in Japanese eel (Anguilla japonica), in which Sertoli cells were activated by 11KT to stimulate the spermatogenic cascade <ns0:ref type='bibr' target='#b23'>(Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b27'>Ozaki et al., 2006)</ns0:ref>. The role of 11KT as a promoter of spermatogenesis in ovaries of sex-changing fish has also been demonstrated <ns0:ref type='bibr' target='#b3'>(Bhandari et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b35'>Todo et al., 2008)</ns0:ref>. In the protogynous honeycomb grouper (Epinephelus merra), in vivo implantation of 11KT into pre-spawning females caused masculinisation in all fish <ns0:ref type='bibr' target='#b3'>(Bhandari et al., 2006)</ns0:ref>. Moreover, in vivo induction of sex change was also accomplished in the bidirectional sex changing Pseudolabrus sieboldi, both by administering sustained-release capsules containing 11KT to females and those with E2 to male individuals <ns0:ref type='bibr' target='#b25'>(Ohta et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Implants containing 11KT induced changes in body colour in the females too <ns0:ref type='bibr' target='#b25'>(Ohta et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Likewise, in vitro 11KT treatment of ovarian explants from protogynous three-spot wrasse induced the formation of presumed spermatogenic crypts and sperm within 2 -3 weeks <ns0:ref type='bibr' target='#b35'>(Todo et al., 2008)</ns0:ref>.</ns0:p><ns0:p>However, three-spot wrasses live in tropical waters, and sex change in captive populations has been observed in under 40 days <ns0:ref type='bibr' target='#b18'>(Kuwamura et al., 2007)</ns0:ref>, around half that required in spotty wrasse (60 -70 days <ns0:ref type='bibr' target='#b33'>(Thomas et al., 2019)</ns0:ref>). In the current study, it may be that spotty wrasse ovaries were not cultured long enough for 11KT to promote male germ cell proliferation, or that 11KT alone is insufficient to induce gonadal transformation in this species. It has been hypothesised that endogenous E 2 levels need to decrease below a physiological threshold for oocyte maintenance to become jeopardised and androgen levels to effectively induce proliferation of bipotential gonial stem cells in the ovary <ns0:ref type='bibr' target='#b3'>(Bhandari et al., 2006)</ns0:ref>. Co-treatment of 11KT with an aromatase inhibitor (e.g. fadrozole) should be performed in the future to further characterise the triggering of in vitro gonadal restructuring, potentially establishing this technique as a model for sex change research.</ns0:p><ns0:p>It was hypothesised that CORT treatment would promote oocyte degeneration, following evidence that CORT may mediate sex change in fish by inhibiting aromatase transcription and promoting androgen production <ns0:ref type='bibr' target='#b9'>(Fernandino et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b12'>Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>. However, no significant differences were observed in the proportion of non-atretic PVO between the controls and ovaries treated with 1 and 10 ng/mL of CORT (43.3 &#177; 7.7 % and 42.9 &#177; 6.2 %, respectively) (Figure <ns0:ref type='figure' target='#fig_4'>2C</ns0:ref>). While a significant increase in the percentage area of PVO was detected between the controls and 100 ng/mL CORT treatment group (39.5 &#177; 18.0 %) (p-value &lt; 0.005), this may reflect variability between fish and the effect of smaller sample size (n=4, as one biological replicate was PeerJ reviewing PDF | (2020:05:48900:2:0:NEW 15 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed lost during histological processing due to technical issues) on the statistical analysis. Therefore, these differences were not deemed a direct effect of CORT administration to the ovaries. Cortisol has been observed to negatively impact ovarian development by suppressing E 2 secretion in cultured rainbow trout (Oncorhynchus mykiss) follicles <ns0:ref type='bibr' target='#b4'>(Carragher &amp; Sumpter, 1990)</ns0:ref>. However, in that same study, <ns0:ref type='bibr' target='#b4'>Carragher &amp; Sumpter (Carragher &amp; Sumpter, 1990)</ns0:ref> argued that there may be a time lag for CORT suppressive effects to become evident during in vitro ovarian culture. Longerterm culture of spotty wrasse ovaries may be needed to observe any effects of CORT on ovarian structure, a finding in keeping with our findings for 11KT.</ns0:p><ns0:p>Cortisol also failed to initiate obvious sex change in cultured spotty wrasse ovaries, with no evidence of male tissues observed in our ovarian sections. It may be that CORT alone is ineffective in inducing spotty wrasse sex change, or that the experiment needs technical adjustments (e.g. longer duration) for the successful induction of sex change in spotty wrasse. Low doses of CORT (i.e. 0.01 -10 ng/mL) were reported to promote testicular development in Japanese eel and pejerrey (Odontesthes bonariensis) by increasing the production of 11KT, which, in turn, induced spermatogonial proliferation <ns0:ref type='bibr' target='#b27'>(Ozaki et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>. Interestingly, higher doses of CORT (i.e. 100 ng/mL) had the opposite effect and inhibited 11KT synthesis in these species <ns0:ref type='bibr' target='#b27'>(Ozaki et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>. Although measurement of 11KT concentration in the cultured media should be performed to evaluate such effects on spotty wrasse ovaries, absence of spermatogonia in the cultured ovaries suggests 11KT levels were not significantly affected. The possible metabolic conversion of CORT into 11KT and other 11oxygenated androgens has also been suggested in several teleosts <ns0:ref type='bibr' target='#b15'>(Kime, 1978;</ns0:ref><ns0:ref type='bibr' target='#b27'>Ozaki et al., 2006)</ns0:ref>, as there is cross-talk between the glucocorticoid and androgen synthesis pathways <ns0:ref type='bibr' target='#b1'>(Arterbery, Deitcher &amp; Bass, 2010;</ns0:ref><ns0:ref type='bibr' target='#b12'>Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>. However, the mechanisms underlying 11KT synthesis are still not fully understood and a potential CORT-to-11KT direct conversion remains to be clarified <ns0:ref type='bibr' target='#b30'>(Schulz, 1986;</ns0:ref><ns0:ref type='bibr' target='#b12'>Goikoetxea, Todd &amp; Gemmell, 2017)</ns0:ref>. This work constitutes the first-ever in vitro organ culture system in a temperate sex-changing teleost, and opens doors for future work in a wide variety of hermaphroditic fish.</ns0:p><ns0:p>Although the current experiment could not induce any observable female-to-male restructuring of spotty wrasse gonads in culture, ovarian tissues were successfully maintained in culture for 21 days. Such long-term culture of fish organ explants has only been accomplished in a few instances <ns0:ref type='bibr' target='#b4'>(Carragher &amp; Sumpter, 1990;</ns0:ref><ns0:ref type='bibr' target='#b23'>Miura et al., 1991;</ns0:ref><ns0:ref type='bibr' target='#b27'>Ozaki et al., 2006</ns0:ref><ns0:ref type='bibr' target='#b26'>Ozaki et al., , 2019;;</ns0:ref><ns0:ref type='bibr' target='#b8'>Fernandino et al., 2012)</ns0:ref>.</ns0:p><ns0:p>Furthermore, this work validates the use of open-source machine learning software Ilastik to reliably recognise cell types in fish gonads, with enormous potential in future studies. The image analysis strategy presented here can be adapted and applied to other studies where accurate cell segmentation is required.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>This study developed an organ culture system for gonadal tissue explants in New Zealand spotty wrasse, providing an in vitro system for investigating control and perturbation of gonadal sex change in this and other species. Moreover, we successfully integrated Ilastik software to score ovarian architecture, an application which provides great scope for future studies on effects of steroids on ovarian development of spotty wrasse, or fish in general, in vitro. Although culture was associated with a degree of PVO degeneration, it was shown that administration of at least 10 ng/mL E 2 can partially mitigate tissue degeneration during culture. However, neither 11KT nor CORT treatment successfully accelerated ovarian atresia or induced spermatogenic proliferation.</ns0:p><ns0:p>Future in vitro manipulations should explore longer cultures and/or co-treatment of multiple steroids to analyse potential antagonistic or synergistic effects. This culture system will facilitate future investigations of ovarian and testicular trans-differentiation in sex-changing fish. For example, an in vitro system would be especially amenable to the use of small interfering RNAs to knock down expression of specific candidate genes or application of methylation inhibitors to further investigate the genetic and epigenetic cascade underlying sex change. In vitro approaches also offer valuable ethical benefits over experiments with live fish, enabling the application of the 3Rs guiding principles: reduce, replace and refine <ns0:ref type='bibr' target='#b29'>(Russell &amp; Burch, 1959)</ns0:ref>. Manuscript to be reviewed Bars show the surface area of all non-atretic PVO over total tissue area (expressed as %). Same day 0 reference tissues and controls (day 21) were used for all treatments. Results are shown as the mean &#177; SEM. Different letters denote a statistically significant difference between groups, while experimental groups with no significant differences share the same letter. Samples sizes: D0 n = 5; C (no steroid) n = 15; E 2 , 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; 11KT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; CORT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 4. Abbreviations: 11-ketotestosterone (11KT), control (C), cortisol (CORT), day 0 reference tissues (D0), 17&#946;-estradiol (E 2 ). Bars show the surface area of all non-atretic PVO over total tissue area (expressed as %).</ns0:p><ns0:p>Same day 0 reference tissues and controls (day 21) were used for all treatments. Results are shown as the mean &#177; SEM. Different letters denote a statistically 508 significant difference between groups, while experimental groups with no significant 509 differences share the same letter. Samples sizes: D0 n = 5; C (no steroid) n = 15; E 2 , 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; 11KT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 5; CORT, 1 ng/mL n = 5, 10 ng/mL n = 5, 100 ng/mL n = 4. Abbreviations: 11-ketotestosterone (11KT), control (C), cortisol (CORT), day 0 reference tissues (D0), 17&#946;-estradiol (E 2 ).</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Abbreviations</ns0:head><ns0:label /><ns0:figDesc>Abbreviations</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure S1 Figure 1</ns0:head><ns0:label>S11</ns0:label><ns0:figDesc>Figure S1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> </ns0:body> "

"Editorial Comments Editor Thank you for improving your manuscript. However, there is still a major concern related to the results of the Ilastik analysis for the small sizes of oocytes and the stages of atretic cells have not been clear. Therefore, some additional photos to show these histological and/or cellular information are needed. To the editor, Thank you for inviting us to resubmit a revision of our manuscript to PeerJ again. We are very grateful for the editor and reviewers for their time in reviewing our revised manuscript and providing valuable and constructive feedback. All comments and suggestions have been carefully considered and addressed. Additional information and photographs regarding the histological analysis and staging performed during the Ilastik analysis have been incorporated, which we hope have improved the quality and clarity of the manuscript. Details of the revisions made are shown below each corresponding reviewer comment. The revised manuscript has been uploaded both with tracked changes and as a clean new version with no tracked changes. Thank you for again considering our re-submission and we look forward to hearing from you. Kind regards, Alexander Goikoetxea Reviewers Comments Reviewer 1 Comments for the Author: In this report, authors try to establish an ovarian culture system of wrasse and then use for the future study. However, authors only used image software to validate the ovarian development. It is hard to understand the relation between the oocyte size and cellular situation. Thus, reviewer believes that cell proliferating assay (IHC staining with a proliferating marker PCNA or proliferating cells labelling with the BrdU) and apoptosis assay (TUNEL assay) would help to improve the quality of this study. In general, the commercial anti-PCNA antibody can widely use for many fish species. We truly thank the reviewer for their thoughts on the histological analysis. We completely agree that the use of the aforementioned techniques, such as immunohistochemical staining using PCNA antibody would be a good addition for the documentation of the events occurring at the cellular level during gonadal sex change. However, as outlined in our previous letter of rebuttal, it is impossible for us to perform such experiments at this time. Regarding the TUNEL assay, such an assay was tested for the in situ detection of apoptosis in New Zealand spotty wrasse gonadal samples from several different in vivo experiments (see manuscript in preparation https://www.biorxiv.org/content/ 10.1101/2020.08.28.271973v1.full.pdf) back in 2018, which proved unsuccessful. As New Zealand spotty wrasse is an emerging model for sex change research and relatively understudied, it may be that this kind of assays need to be further optimised for this species. Therefore, suggestions by Reviewer 1 will undoubtedly be considered in future experimental set-ups involving this species. Reviewer 2 Experimental design: The reviewer understands the Ilastik software can be used to distinguish the cells in gonads. This may be the first trial. Therefore, the reviewer would like you to recommend to mention about the data settings for Ilastik more carefully. Does the data vary when the different sizes of previtellogenic oocytes and/or the different stages of atretic cells are labelled to train Ilastik? It would be more helpful to understand the availability of Ilastik if you show these kinds of data. We thank the reviewer for highlighting that our description of the data settings for Ilastik as written was incomplete, and for the helpful suggestions for improvement. In Ilastik, features to be selected during the iterative training are grouped into three types: colour/intensity, edge and texture; the scales of which correspond to the sigma (σ) of the Gaussian. The manual selection of these training data enable the algorithm in Ilastik to correctly classify the different ‘objects’ and label these accordingly (Berg et al., 2019). This pixel classification assigns a class label (e.g. ‘PVO’, ‘background’…) defined by the user with the paintbrush tool to each pixel in an image. The algorithm estimates the probability that the pixel belongs to the assigned class, then providing the segmentations maps used for quantitative analysis (Berg et al., 2019). Hence, data settings are standardised for all defined objects and do not vary when, for example, different sizes of previtellogenic oocytes or different stages of atretic cells are labelled in Ilastik. The missing details in our ‘Materials and Methods’ section regarding these data have now been added. This section now reads: ‘Atretic oocytes, stromal tissue and background were labelled separately with a paintbrush-style tool using the Pixel Classification module. The following feature selection was applied: colour/intensity, σ3 = 1.60; edge, σ3 = 1.60; texture, σ3 = 1.60.’ Comments for the Author The reviewer would like to express his respect for your efforts to revise the manuscript. Overall, the manuscript seems to be much improved. We thank the reviewer for their positive feedback. On the other hand, information on the cellular and histological stages of oocytes may be still unclear, although the resolution of pictures has been improved. As the reviewer might say at the previous comments, the results of the Ilastik analysis for the small sizes of oocytes and the stages of atretic cells has not been clear. Therefore, it may be better to use some additional photos to show these histological and/or cellular information. Moreover, in this manuscript, it might be important to understand the differences of the cellular stages between in vivo and in vitro. The reader would like to know how different between the initial tissue and the cultured tissue. In addition, the analyses of histological and/or cellular differences of gonadal change between the present in vitro data and the previous in vivo data during gonadal sex change might be informative. We thank the reviewer for their comments. Additional photos depicting examples of ovarian tissue explants from day 0 (sampled as reference), treated with 100 ng/mL of 17β-estradiol, 10 ng/mL of 11KT-testosterone or 100 ng/mL of cortisol have been added as a supplementary figure (Figure S1) to show further histological information. As observed in these images, although the level of atresia is higher in all hormonallytreated cultures compared to the reference tissue, no signs of female-to-male gonadal transformation (e.g. presence of male cells) were observed. Thus, in reference to the differences observed at the cellular level between the initial tissue (day 0) and the cultured tissue, the most striking difference observed was that atresia of previtellogenic oocytes or PVO seemed to be directly linked to the in vitro culture itself. As shown by our results, there was a marked reduction in oocyte surface area in the ‘in vitro controls’ compared to the initial or reference tissues, in which a statistically significantly higher proportion of tissue area was filled with non-atretic PVO (p-value < 0.001). Unfortunately, the absence of transitional gonads obtained through the in vitro hormonal treatment of New Zealand spotty wrasse ovaries in this study prevented us from drawing comparisons between gonadal stages of sex change per se in our data. Furthermore, the aforementioned lack of gonadal sex change in our cultured ovaries also prevented comparisons of histological and/or cellular differences of gonadal sex change between the data from the present manuscript and that of some of our previous in vivo studies in which gonadal sex change was successfully induced. However, the culture system developed within this frame work will hopefully enable the future accomplishment of in vitro gonadal sex change of spotty wrasse ovaries and facilitate investigations of ovarian and testicular trans-differentiation in this and other sexchanging species. References: Berg S, Kutra D, Kroeger T, Straehle CN, Kausler BX, Haubold C, Schiegg M, Ales J, Beier T, Rudy M, Eren K, Cervantes JI, Xu B, Beuttenmueller F, Wolny A, Zhang C, Koethe U, Hamprecht FA, Kreshuk A. 2019. ilastik: interactive machine learning for (bio)image analysis. Nature Methods 16:1226–1232. DOI: 10.1038/s41592-0190582-9. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Access to the digital 'all-knowing cloud' has become an integral part of our daily lives. It has been suggested that the increasing offloading of information and information processing services to the cloud will alter human cognition and metacognition in the short and long term. A much-cited study published in Science in 2011 provided first behavioral evidence for such changes in human cognition. Participants had to answer difficult trivia questions, and subsequently showed longer response times in a variant of the Stroop task with internet-related words ('Google Stroop effect'). The authors of this study concluded that the concept of the Internet is automatically activated in situations where information is missing (e.g., because we might feel the urge to 'google' the information). However, the 'Google Stroop effect' could not be replicated in two recent replication attempts as part of a large replicability project. After the failed replication was published in 2018, the first author of the original study pointed out some problems with the design of the failed replication. In our study, we therefore aimed to replicate the 'Google Stroop effect' with a research design closer to the original experiment. Our results revealed no conclusive evidence in favor of the notion that the concept of the Internet or internet access (via computers or smartphones) is automatically activated when participants are faced with hard trivia questions. We provide recommendations for follow-up research</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>It seems intuitively plausible that today's ubiquitous 24/7 access to the Internet via smartphones and computers will affect our cognitive functioning and strategies. Specifically, it has been suggested that different types of 'cognitive offloading' (i.e., the use of our bodies, objects, and technology to alter the processing requirements of a task to reduce cognitive demand) may alter human cognition and metacognition in the short and long term <ns0:ref type='bibr'>(Risko &amp; Gilbert, 2016)</ns0:ref>. In the memory domain, one idea is that the Internet is taking the place not just of other humans as external sources of memory ('transactive memory'), but also of our own cognitive faculties <ns0:ref type='bibr'>(Ward, 2013;</ns0:ref><ns0:ref type='bibr'>Wegner &amp; Ward, 2013)</ns0:ref>. We increasingly offload information to 'the cloud', as almost all information today is readily available through a quick Internet search.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Evidence for such 'Google effects on memory' has been presented in a much-cited 1 and influential 2 paper published in Science <ns0:ref type='bibr'>(Sparrow, Liu &amp; Wegner, 2011)</ns0:ref>. Across four experiments, the authors showed that a) when people expect to have future access to information, they have lower rates of recall of the information itself and enhanced recall instead for where to access it, and b) when faced with difficult general knowledge (or, trivia) questions, people are primed to think about the Internet and computers. The latter effect was demonstrated in one experiment (Exp.1) where participants answered easy or difficult trivia questions, and then completed a variant of the Stroop task <ns0:ref type='bibr'>(MacLeod, 1991)</ns0:ref>. In a paradigm conceptually similar to the emotional Stroop paradigm <ns0:ref type='bibr' target='#b1'>(Algom, Chajut &amp; Lev, 2004)</ns0:ref>, participants responded to the ink color of written words, which were either related or unrelated to the Internet. Stroop-like interference from words relating to computers and Internet search engines was increased after participants answered difficult compared with easy questions, consistent with those terms being 'primed' in participants' minds <ns0:ref type='bibr' target='#b6'>(Doyen et al., 2014)</ns0:ref>. Briefly, the results seem to suggest that whenever information is needed and lacking, the concept of the Internet (including computer-related terms) is activated and can interfere with our behavior in subsequent tasks (e.g., because we might feel the urge to 'google' the information).</ns0:p><ns0:p>In 2018, Camerer and colleagues published a meta-analysis of 21 replications of social science and psychology experiments published in Science or Nature between 2010 and 2015 <ns0:ref type='bibr'>(Camerer et al., 2018)</ns0:ref>. This large replicability project included a replication of the 'Google Stroop effect' by <ns0:ref type='bibr'>Sparrow and colleagues (2011)</ns0:ref>. All materials and data from this replication -led by Holzmeister (see below), influenced the replication result. Therefore, it was our aim to investigate the 'Google Stroop effect' in a further study, based on the original experiment, the materials provided by Holzmeister &amp; Camerer, as well as the critical comments by the original authors.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Sample</ns0:head><ns0:p>This work is based on an undergraduate student research project ('Experimentell-Empirisches Praktikum -ExPra') at the Psychologische Hochschule Berlin (PHB). A total of 117 participants were tested. The sample consisted of students at the Psychologische Hochschule Berlin (PHB), as well as friends and families of the student experimenters (see acknowledgment). All participants provided written informed consent. The experiment was approved by the ethics committee of the PHB (approval number PHB10032019).</ns0:p></ns0:div> <ns0:div><ns0:head>Paradigm</ns0:head><ns0:p>Our version of the experiment was based on two previous studies: Sparrow et al.'s original Exp.1 <ns0:ref type='bibr'>(Sparrow, Liu &amp; Wegner, 2011)</ns0:ref>, and the replication study <ns0:ref type='bibr'>(Camerer et al., 2018)</ns0:ref>. We incorporated comments provided by the first author of the original study <ns0:ref type='bibr'>(Sparrow, 2018)</ns0:ref>, published in response to the failed replication 3 .</ns0:p><ns0:p>The original experiment tested the hypothesis that participants are primed to think about the Internet when faced with difficult trivia questions (e.g., 'Did Benjamin Franklin give piano lessons?'). Participants first had to answer a block of either hard or simple question followed by a modified Stroop Task. In this task, Internet-related and neutral words were presented in random order, and participants were instructed to indicate the word's color (blue or red) via button press.</ns0:p><ns0:p>RTs to the words were measured as the dependent variable. In order to manipulate cognitive load, a random six-digit number was presented, and participants were instructed to memorize it for delayed retrieval (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). After their response in the modified Stroop Task, participants were asked to enter the six-digit number. The results from the modified Stroop task showed the predicted pattern: RTs to computer terms (e.g., Google) were longer than RTs to neutral terms 3 Our own request for further details about the experimental design remained unanswered (email from March 2019).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed (e.g., Target), especially after participants were faced with difficult trivia questions ('question type x word type' interaction; F(1,66) = 5.02, p &lt; .03).</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref> Although <ns0:ref type='bibr'>Sparrow et al. (2011)</ns0:ref> reported that they used eight target words related to computers and search engines, and 16 unrelated words, the total number of trials presented in each 'hard question' and 'easy question' block remained unclear, and also whether words were repeated.</ns0:p><ns0:p>In their replication, Holzmeister &amp; Camerer used the 24 words originally reported by <ns0:ref type='bibr'>Sparrow et al. (2011)</ns0:ref>, and decided to run 48 trials per block, resulting in the repetition of words (i.e., each participant saw each word four times, twice in the 'hard question' block, and twice in the 'easy question' block.) In her response to the failed replication (Sparrow, 2018), Sparrow then strongly argues against the repetition of words, and reported the full set of 16 Internet-related words used in the original study.</ns0:p><ns0:p>The cognitive load manipulation is described as follows in the original study: 'Participants are presented with words in either blue or red, and were asked to press a key corresponding with the correct color. At the same time, they were to hold a 6 digit number in memory, creating cognitive load' <ns0:ref type='bibr'>(Sparrow et al., 2011, supplement)</ns0:ref>. In their replication, Holzmeister &amp; Camerer decided to manipulate cognitive load by presenting a six-digit number before each Stroop task block. After the block (involving several trials), participants were asked to enter the memorized number. In her response, Sparrow strongly argues against this block-wise procedure and provided more detail about the original study: '[&#8230;] before each word, they were shown a different six digit number, which was reported back by them after each Stroop word'.</ns0:p><ns0:p>Finally, Sparrow mentions that the original experiment was run in 2006, which is why she believes that the computer terms used originally are obsolete and should not have been used in a replication <ns0:ref type='bibr'>(Sparrow, 2018)</ns0:ref>. In her response, she writes that she 'would focus primarily on target words about phones (as they seem most ubiquitous), but would also be sure to pre-test many possible words to ensure their contextual relevance before putting them into the modified Stroop'.</ns0:p><ns0:p>For our replication study, we used the material provided by Holzmeister &amp; Camerer on the open science platform OSF (https://osf.io/wmgj9/), and revised it according to Sparrow's comments. All materials and data are available at OSF (https://osf.io/cjgea/). The experiment was implemented using oTree <ns0:ref type='bibr' target='#b5'>(Chen, Schonger &amp; Wickens, 2016)</ns0:ref>. We translated the trivia questions and Stroop words into the German language, and adjusted some questions to the German context (e.g., 'Was Cat in the Hat written by J.D. Salinger?'). The experiment consisted of two blocks of 16 either hard or easy questions, followed by 24 Stroop words of which eight were Internet-related (e.g., WLAN, Google, Website) and 16 were unrelated (e.g., frame, bottle, bamboo). The words were pre-tested for contextual relevance (see below), and randomly assigned to the easy and hard question condition. Words were presented in random order, and word color (blue or red) was randomly chosen on each trial. Participants were instructed to indicate the word color as quickly and as accurately as possible via button press ('e' for blue and 'i' for red, using the index fingers of both hands). Participants were asked to place their fingers on the keys of the computer keyboard before the start of the Stroop task. Before the presentation of each Stroop word, participants had to memorize a random six-digit number for delayed retrieval after their response in the color-naming task.</ns0:p><ns0:p>After the main experiment, participants provided information about their age, level of education, color blindness, and what they thought the purpose of the experiment was. Finally, participants</ns0:p><ns0:p>were asked what they would normally do when faced with a hard general knowledge question in their daily life: a) look up the answer on the Internet, b) ask someone who might know the answer, c) leave it at that. Including instructions, debriefing and signing the informed consent forms, the experiment lasted approximately 45 minutes in total.</ns0:p></ns0:div> <ns0:div><ns0:head>Word stimuli</ns0:head><ns0:p>In the original study, the color naming task contained '8 target words related to computers and search engines (e.g., Google, Yahoo, screen, browser, modem, keys, Internet, computer), and 16 unrelated words (e.g., Target, Nike, Coca Cola, Yoplait, table, telephone, book, hammer, nails, chair, piano, pencil, paper, eraser, laser, television)' (supplement, p.2). According to Sparrow (2018), the computer-related target words were selected from a larger set of the following 16</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed words: 'Google, Yahoo, mouse, keys, Internet, browser, computer, screen, Altavista, Wikipedia, disk, Lycos, Netscape, modem, router, online' (p.1). The full list of words included in the set of general terms is not provided. Thus, at least in the case of computer terms, the number of target words used in the experiment was smaller than the number of available target words (eight and 16, respectively). It remains unclear how the target words were chosen for each participant. What seems to be clear is that the words Target, Nike, Google, and Yahoo were chosen for each participant, because a within-subject comparison was calculated using this subset of words (see below). Furthermore, the authors of the original study write that the sets 'were matched for frequency to the target words ( <ns0:ref type='formula'>11</ns0:ref>)' (supplement, p.2). Without the full word sets, this claim is hard to evaluate. A brief search in the referenced word corpus (reference 11) revealed no hit for the computer terms 'Altavista' and 'Google' <ns0:ref type='bibr'>(Nelson, McEvoy &amp; Schreiber, 2004)</ns0:ref>.</ns0:p><ns0:p>Based on Sparrow's reply to the first replication (Sparrow, 2018), we prepared two new sets of German words, one set with general terms (32 words), and one set with computer terms ( <ns0:ref type='formula'>16</ns0:ref>words). As suggested by Sparrow ( <ns0:ref type='formula'>2018</ns0:ref>), we validated the contextual relevance of these words.</ns0:p><ns0:p>31 na&#239;ve participants (mostly undergraduate students who did not participate in the main experiment) took part in an online survey (https://www.surveymonkey.de/). For each of the 48 words, participants rated the contextual relevance (i.e., 'Internet-relatedness') on a 5-point scale (statement: 'I think of the Internet when reading this word'; Rating: 1 = Strongly agree; 2 = Agree; 3 = Neutral; 4 = Disagree; 5 = Strongly disagree). For computer words, the mean rating (i.e., the mean rating across the median ratings per participant) turned out to be 1.06, while it was 4.94 for the neutral terms.</ns0:p><ns0:p>Our list of computer (or, Internet-related) terms contained: website, data volume, email, search engine, Wikipedia, WLAN, app, Google, smartphone, hotspot, online, blog, Spotify, Firefox, Whatsapp, Chrome. The list of general (or, unrelated) terms contained words like nail or car, but also names of grocery stores (all words in German; full list on OSF). After completion of the study, we used the online DlexDB database to estimate word frequency in the two sets (www.dlexdb.de/query/kern/typposlem/). The median absolute type frequency (corpus frequency)</ns0:p><ns0:p>for the general terms was 374, while it was 23.5 for the computer terms. However, half of the words in the list of computer terms were not part of the data base (Wikipedia, WLAN, Google, smartphone, blog, Spotify, Firefox, Whatsapp). To what degree this difference in word frequency between the two sets could be problematic for the current experiment, is hard to say. Under the assumption that participants read the words when responding to their color, one would expect longer RTs for words in the set of computer terms due to the lower word frequency (Larsen, parsimoniously be explained by differences in word frequency, if word sets were not matched.</ns0:p><ns0:p>However, differences in word frequency between the two sets do not seem to preclude the proposed 'question type x word type' interaction, which is driven by priming (according to the dual-route model). In a control analysis, we did not find evidence for an effect of word frequency on RTs in our data set 4 .</ns0:p><ns0:p>As Sparrow (2018) in her reply to the replication strongly argues against presenting each target word more than once (referring to 'active thought suppression'), we decided to randomly select words for the easy and hard blocks. From our pool of 48 words, eight computer terms and 16 general terms were presented in each block, so that each word was presented only once in the experiment. The words were randomly chosen for each participant. Of note, each word was presented twice in the original study, once in the easy block, and once in the hard block. Since we did not intend to select single words for post hoc pairwise tests, we did consider a strict 'no word repetition' approach to be more in line with Sparrow's (2018) suggestions. Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> summarizes the known differences between the original study and the replication studies.</ns0:p><ns0:p>Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> about here </ns0:p></ns0:div> <ns0:div><ns0:head>Data preprocessing</ns0:head><ns0:p>We preregistered our analyses including confirmatory and exploratory statistical tests (https://aspredicted.org/z3xt4.pdf), and later decided to restrict the analysis to confirmatory tests.</ns0:p><ns0:p>All data are available on OSF (https://osf.io/cjgea/). Data were preprocessed and analyzed using R version 3.3.2 (www.r-project.org), and RStudio version 1.0.136 (www.rstudio.com). The original 'csv' data files were exported from oTree, imported into R, converted into the long format and merged using custom R scripts. Each student experimenter contributed one 'csv' file containing the data from multiple participants. The resulting data file in the long format thus contains blocks of data from different student experimenters, because we did not sort the data according to recording time and date.</ns0:p><ns0:p>The age distribution of participants turned out to be heavily skewed (mean age: 31 years; range:</ns0:p><ns0:p>18 -73). This was because the participant sample included students' families and friends, and we originally did not set a maximum age. Since the sample in the original paper (Sparrow, Liu &amp; Wegner, 2011) consisted of undergraduate students (supplement, p.2), we decided to deviate from our preregistered protocol, and set the maximum age as the 75% quantile of the age distribution (44 years). In the remaining sample of 89 participants, the mean age was 24 years.</ns0:p><ns0:p>A second deviation from the preregistered protocol was due to the fact that we originally assumed that trials with incorrect responses (i.e., when participants pressed the key for the wrong color) should be excluded from the analysis. According to the replication report (https://osf.io/84fyw/), which includes personal communication with the original authors, this was not the case in the original study. We therefore included trials with correct and incorrect responses in the colornaming task. (Note that trials with incorrect responses in the memory task were not excluded, either.)</ns0:p><ns0:p>Although not specified in the original study, and neither in the replication study, we preregistered an additional exclusion of trials with RT outliers and anticipatory responses (&lt; 0.1s). We defined RT outliers using the interquartile range (IQR) method <ns0:ref type='bibr'>(Tukey, 1977)</ns0:ref>, separately for each participant. These criteria resulted in the exclusion of 7 &#177; 3% of trials (mean percentage &#177; standard deviation). For our final analysis, we did not exclude RT outliers, but also report the results of the analysis including the exclusion of RT outliers.</ns0:p></ns0:div> <ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>For each participant, we calculated the performance in the easy and hard questionnaires, the performance in the number memory task, and the mean RT in each of the four conditions. The condition averages from all participants were then exported into JASP 0.10.2 (https://jaspstats.org/) for frequentist and Bayes Factor (BF) analysis, using default Cauchy priors (scale 0.707). To test the predicted 'question type x word type' interaction, we calculated the following difference and tested it against zero using a two-sided paired test: [RT(computer) -RT(general)] hard -[RT(computer) -RT(general)] easy . Performance in the number memory task was calculated for all trials (i.e., including trials with RT outliers in the color-naming task).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>Participants found the easy questions to be answerable (97 &#177; 4%, mean accuracy &#177; standard deviation), but had difficulty finding the correct answers to the hard questions (60 &#177; 11%). As in the original study (98% versus 47%), this difference was significant (t 88 = 48.31, p &lt; .001). Mean accuracies in the color-naming task and number memory task were high (98 &#177; 3% and 81 &#177; 16%, respectively).</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_1'>2A</ns0:ref> plots the RT data from the color-naming task. The dual-route model predicts that computer terms create more interference, and thus are associated with longer RTs than general terms, in particular in hard question blocks. The results show that in easy question blocks, mean RT was 830 &#177; 366 ms for general terms, and 901 &#177; 620 ms for computer terms (mean &#177; standard deviation). In hard question blocks, mean RT was 825 &#177; 453 ms for general terms, and 821 &#177; 379 ms for computer terms. Thus, the RT data do not show the predicted pattern. Accordingly, the sequential Bayes Factor (BF) analysis yielded a final BF 01 of 5.07 in favor of the null over the alternative hypothesis, after the data from N=89 participants were taken into account (Figure <ns0:ref type='figure' target='#fig_1'>2B</ns0:ref>).</ns0:p><ns0:p>The observed data are thus 5 times more likely under the null model than under the alternative model (i.e., the 'question type x word type' interaction). When the specific data pattern reported in the original study was considered as alternative model in an exploratory analysis, our data were 16 times more likely under the null model (BF 0+ = 16.43). The preregistered two-sided one-sample t-test was not significant (t 88 = -1.04, p = .301).</ns0:p><ns0:p>The pattern of results remained the same when incorrect responses in the color-naming task were excluded. When RT outliers were excluded from data analysis, overall mean RTs decreased, but the predicted interaction could still not be observed (BF 01 = 3.74; t 88 = 1.31, p = .194). In easy question blocks, mean RT was 737 &#177; 296ms for general terms, and 723 &#177; 295ms for computer terms (mean &#177; standard deviation). In hard question blocks, mean RT was 707 &#177; 268ms for general terms, and 712 &#177; 261ms for computer terms (data not plotted in a figure).</ns0:p><ns0:p>In the debriefing after the main experiment, 11 out of 89 participants did not provide a response (e.g., closed the browser before answering). 73% (57/78) of all responders said that they would look up the answer on the Internet when faced with a hard general knowledge question in their daily life; 18% (14/78) said that they would ask someone who might know the answer, and 9%</ns0:p><ns0:p>(7/78) responded that they would 'leave it at that'. When we restricted the RT data analysis to participants consulting the Internet (N=57), the pattern of results was very similar to the one PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed reported above (BF 01 = 5.01). Finally, when all participants (N=117) were included in the data analysis, the pattern of results turned out to be 9 times more likely under the null model than under the alternative model (BF 01 = 9.31). Sequential Bayes Factor (BF) analysis of the 'question type x word type' interaction. User prior refers to a Cauchy prior with scale 0.707, wide prior to a Cauchy prior with scale 1, and ultrawide prior to a Cauchy prior with scale &#8730;2.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical issues in the original study</ns0:head><ns0:p>During our work on this project, we noticed three statistical issues in the original paper by Sparrow et al. ( <ns0:ref type='formula'>2011</ns0:ref>). In the following paragraph, we address these observations in turn.</ns0:p><ns0:p>First, it remains unclear whether N=69 or N=46 participants were tested in Experiment 1 of the original paper <ns0:ref type='bibr'>(Sparrow, Liu &amp; Wegner, 2011)</ns0:ref>. In the supporting online material 5 , the authors write that 'Forty-six undergraduate students (28 female, 18 male) at Harvard University were tested in a within subjects experiment' (p.2). In their 2x2 within-subject design (easy/hard questions;</ns0:p><ns0:p>computer/general words), the correct degrees of freedom (df) would then be 45 for paired t-tests, as well as for main effects, the interaction and simple main effects in the rm-ANOVA. Under the assumption that N=69 participants were tested, the correct df would be 68. In the main text and supplement of the original paper, the reported df is 68 for paired t-tests, and 66 for the rm-ANOVA.</ns0:p><ns0:p>According to the authors of the replication, the original authors initially confirmed that the sample size was 46 participants, and that dfs were misreported in the paper, but after the publication of the replication the original authors pointed out that the number of participants was 69 (https://osf.io/84fyw/). While the reported t-, F-, and p-values appear to be in better agreement with the N=69 scenario 6 , the reported dfs are incorrect in both scenarios.</ns0:p><ns0:p>Second, two computer-related words (Google/Yahoo), and two unrelated words (Target/Nike)</ns0:p><ns0:p>were selected for analysis in the original paper. In the main text, RT data are reported only for this subset, together with the 'question type x word type' interaction (F(1,66) = 5.02, p &lt; .03). The Third, two paired t-tests on word type are reported for the complete data set ('hard questions' condition, computer words versus general words: t(68) = 3.26, p &lt; .003; 'easy questions' condition, t(68) = 2.98, p &lt; .005). The 'question type x word type' interaction is reported only for the 4-word-subset (F(1,66) = 5.02, p &lt; .03). Based on visual inspection of the reported average RTs, it seems plausible that the 'question type x word type' interaction should be smaller for the complete data set (Figure <ns0:ref type='figure' target='#fig_3'>3A</ns0:ref>) than for the 4-word-subset (Figure <ns0:ref type='figure' target='#fig_3'>3B</ns0:ref>). Sparrow et al. ( <ns0:ref type='formula'>2011</ns0:ref>) report the following result in the supplement: 'Taking out the 4 terms (Google/Yahoo and Target/Nike) which yielded an interaction with easy/hard questions (F(1,66) = 5.52 [sic], p &lt; .03), the interaction between computer and general terms and easy/hard questions remains significant F(1,66) = 9.49, p &lt; .004' (p.3, italics added). In fact, the interaction for the reduced data set (i.e., all data minus the 4-word-subset) turned out to be larger than the interaction reported for the subset (F-values 9.49 and 5.52, respectively). This result seems difficult to reconcile with the reported data, but access to the original data would be necessary to clarify this point. </ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Although the majority of participants in our study reported that they would normally look up the answer to hard general knowledge questions on the Internet, we did not find evidence for the have been discussed in much detail elsewhere <ns0:ref type='bibr'>(MacLeod, 1991;</ns0:ref><ns0:ref type='bibr' target='#b1'>Algom, Chajut &amp; Lev, 2004)</ns0:ref>. In Exp.1 by <ns0:ref type='bibr'>Sparrow et al. (2011)</ns0:ref>, there are two colors (red, blue), and two types of words: general terms (e.g., sport) and computer terms (e.g., Google). The authors propose a priming mechanism that specifically affects response times in trials with computer terms: 'not knowing the answer to general knowledge questions primes the need to search for the answer, and subsequently computer interference is particularly acute' (p.776).</ns0:p><ns0:p>According to this model, computer terms become more accessible when the concept of knowledge is activated, or when information necessary for answering a trivia question is lacking, which results in more interference. We think that at least two points need to be made about this model. First, what remains unspecified is the duration of the proposed priming effect (i.e., for how long the computer terms are more accessible than general terms due following the activation of concepts). Follow-up studies could focus more on the temporal dynamics of the proposed priming effect. Perceptual and semantic priming effects are typically short-lasting, within the range of hundreds of milliseconds, while priming effects from social psychology (e.g., the 'Florida effect') are substantially longer lasting, but have turned out to be not robust <ns0:ref type='bibr' target='#b8'>(Harris et al., 2013;</ns0:ref><ns0:ref type='bibr' target='#b6'>Doyen et al., 2014)</ns0:ref>. Second, the exact role of cognitive load (i.e., working memory load) for the 'Google Stroop effect' remains somewhat unclear. In her response to the failed replications, Sparrow (2018) links the working memory task to the paradigm of active thought suppression: 'when people are asked explicitly not to think about a single target word, they must engage in active suppression' (p.1). Sparrow ( <ns0:ref type='formula'>2018</ns0:ref>) cites an earlier study <ns0:ref type='bibr'>(Wegner &amp; Erber, 1992)</ns0:ref> in which participants performed a color-naming task similar to Exp.1 from <ns0:ref type='bibr'>Sparrow et al. (2011)</ns0:ref>. In this experiment, strongest Stroop-like interference was observed when participants were suppressing a specific target word under cognitive load, and when they were asked to name the color of this target word. In contrast to the 'Google Stroop effect', however, participants were asked to actively suppress a single word, so that the importance of high cognitive load in one task might not tell us much about the role of high cognitive load in the other task. Alternatively, it could be argued the word's meaning (in contrast to its color) acts as distracting information in the color-naming task, and that high working memory load increases distractor processing <ns0:ref type='bibr'>(Lavie, 2005)</ns0:ref>. Therefore, the manipulation of cognitive load might be crucial for the processing along the word reading pathway, and thus for the emergence of the 'Google Stroop effect'.</ns0:p><ns0:p>Our data, however, do not support the notion that the concept of the Internet (together with computer-related terms) becomes automatically activated when participants need to answer difficult general knowledge questions. We are aware that our study does not provide a definite answer, and the potential effects of 'cognitive offloading' on human cognition (Risko &amp; Gilbert, 2016) are definitely worth further attention and investigation. As mentioned above, the original study by <ns0:ref type='bibr'>Sparrow et al. (2011)</ns0:ref> avoided word repetitions, so that the RT data from each participant were based on single presentations of target words. Given such noisy RT measurements, in combination with the suboptimal analysis of variance on the sample mean (Whelan, 2008), post hoc selection of target words can easily lead to the wrong impression that an effect exists (e.g., we might find a 'Wikipedia Stroop effect' or 'Firefox Stroop effect' in our data set). Therefore, we recommend using the complete data set in future studies when testing for the crucial 'question type x word type' interaction, preferably with linear mixed-effects models that can better account for stimulus-driven variability in RTs than repeated measures ANOVA <ns0:ref type='bibr' target='#b2'>(Baayen, Davidson &amp; Bates, 2008)</ns0:ref>. Fitting more complex linear models, e.g., models with random slopes, would also require more data than in the present experimental design in which a limited set of words is presented only once per participant <ns0:ref type='bibr'>(Meteyard &amp; Davies, 2020)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Our results revealed no evidence in favor of the notion that the concept of the Internet or internet access (via computers or smartphones) becomes automatically activated whenever participants are faced with hard trivia questions. Thus, the 'Google Stroop effect' might be much smaller than previously thought, and less robust to variations in the experimental design. What else have we learned from this second replication of the 'Google Stroop effect'? Our work on this project revealed that the original research design might not be a good starting point to further investigate this effect. To study the effects of the digital 'all-knowing cloud' on our cognition, research designs with more precision and power are clearly necessary. Finally, as has been so adequately pointed out by Holzmeister &amp; Camerer, this set of replications 'illustrates the importance of open access to all of the materials of published studies for conducting direct replications and accumulating scientific knowledge' <ns0:ref type='bibr'>(Camerer et al., 2018)</ns0:ref>. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>about hereFigure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure1about here</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 2 about hereFigure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure2about here</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>reasoning behind this post hoc selection (i.e., selection of four words out of a pool of Stroop words) remains unclear. Without further information, it remains possible that the choice was made after seeing the data. Such post hoc choices and data-contingent analyses are misleading when they are not presented as exploratory analysis. The potential impact of this post hoc selection should not be underestimated. As pointed out by Sparrow in a reply to the failed replication, 'each Stroop word was seen only once by participants'(Sparrow, 2018). Hence, the RTs for Google/Yahoo and Target/Nike reported in the original paper are based on single trials per participant. Given such noisy measurements with low precision (Smith &amp; Little, 2018), the authors might have capitalized on chance by selecting a subset of words for their main analysis.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>FigureFigure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3 about here</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>'</ns0:head><ns0:label /><ns0:figDesc>Google Stroop effect', as originally published bySparrow et al. (2011). Thus, our data are more in line with the results fromCamerer et al. (2018) who failed to replicate the original effect in two independent experiments. Importantly, the design of our study considered the suggestions for improvement provided by Sparrow (2018) in response to the failed replications. Based on her commentary, we carefully updated and validated the computer-related terms, strictly avoided word repetitions, and manipulated cognitive load as in the original study.It seems worthwhile to take a closer look at the hypothetical cognitive model underlying the 'Google Stroop effect'. Sparrow et al. (2011) describe their Exp.1 as a modified Stroop task, as follows: 'People who have been disposed to think about a certain topic typically show slowed reaction times (RTs) for naming the color of the word when the word itself is of interest and is more accessible, because the word captures attention and interferes with the fastest possible color naming' (p.776). Classic and modified Stroop tasks, such as the emotional Stroop task,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Differences between the original study by Sparrow et al. (2011) and the replication studies. (*) Original sample size is based on the supplementary information. (**) Total number of trials is an informed guess based on the original study and the response to the failed replication (Sparrow, 2018</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,280.87,525.00,159.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,331.87,525.00,225.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,306.37,525.00,293.25' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Differences between the original study by Sparrow et al. (2011) and the replication studies. (</ns0:figDesc><ns0:table /><ns0:note>*) Original sample size is based on the supplementary information. (**) Total number of trials is an informed guess based on the original study and the response to the failed replication (Sparrow, 2018).</ns0:note></ns0:figure> <ns0:note place='foot'>&amp; Camerer -are freely available on OSF (https://osf.io/wmgj9/). In two separate experiments, the authors tested the hypothesis that, after answering a block of hard trivia questions, color-naming reaction times (RTs) are longer for computer-related terms than for general words. Neither experiment showed a significant effect despite adequate statistical power (see https://osf.io/4rfme/ for a short summary of this replication). However, as the original authors of Sparrow et al. (2011) did not provide Holzmeister &amp; Camerer with any materials or feedback on their inquiries, it was difficult to replicate the experimental design of the original study. After the replication had been completed and published, Sparrow noted some design differences compared to the original study(Sparrow, 2018). As Holzmeister &amp; Camerer point out(Camerer et al., 2018), it cannot be ruled out that these design differences, including the manipulation of cognitive load 1 1296 citations on Google Scholar, as of October 7 2020.2 The study is mentioned in a 2019 publication by the EU on harmful internet use: https://op.europa.eu/en/publication-detail/-/publication/fb2d58ea-8e58-11e9-9369-01aa75ed71a1/languageen/format-PDF/source-127484707PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:note> <ns0:note place='foot'>Mercer &amp; Balota, 2006). The main effect of 'word type' reported bySparrow et al. (2011) couldPeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:note> <ns0:note place='foot' n='4'>Using a Bayes Factor approach (https://richarddmorey.github.io/BayesFactor/), we found that a model not containing word frequency as predictor was preferred to a model containing word frequency by a factor of 16. Details can be found in the R analysis script on OSF.</ns0:note> <ns0:note place='foot' n='5'>https://science.sciencemag.org/content/suppl/2011/07/13/science.1207745.DC1 PeerJ reviewing PDF | (2020:08:51863:1:2:NEW 14 Oct 2020)</ns0:note> <ns0:note place='foot' n='6'>In the main text ofSparrow et al. (2011), the simple effect is reported as follows: F(1,66) = 4.44, p &lt; .04 (exact pvalue: .03891. For N=46, the result would be: F(1,45) = 4.44, p = .040712, and thus p &gt; .04.</ns0:note> </ns0:body> "

"We would like to thank both reviewers for their helpful and constructive comments. Below are our point-by-point responses (in blue font) to the reviewers’ comments (in black font). Reviewer 1: o Line 213 mentions that the author decided to skip the preregistered exploratory tests. Why? We preregistered two exploratory tests (https://aspredicted.org/z3xt4.pdf): “Following Sparrow et al. (2011), the analysis will be restricted to selected target words (e.g., “Google” versus “Nike”). In addition, the analysis will be restricted to participants who responded that the hard trivia questions motivated them to use the Internet.” We later dropped these exploratory tests mainly in an attempt not to overload the manuscript. Following the reviewer’s suggestion, we have included some of the results in the revised manuscript (see below). o Lines 227-233: in line with the original study, response times to incorrect RTs were analyzed as well. To me (and to the author as well, apparently), this seems very strange. Could the author perhaps confirm in the text body (or a footnote) whether or not the pattern of results is similar when only correct responses are analyzed? Somewhat related to this, was the accuracy on the color-naming task comparable to that of the original study? When all responses are included, the results are as follows (and as reported in the manuscript): in easy question blocks, mean RT was 830ms for general terms, and 901ms for computer terms; in hard question blocks, mean RT was 825ms for general terms, and 821ms for computer terms (BF01 = 5.07 for the predicted interaction). When incorrect RTs (i.e., incorrect responses in the color-naming task) are excluded, the results show the same pattern: 830, 906, 826, and 817ms (BF01 = 4.47). We have added this information to the main text (results section). Unfortunately, the accuracies in the color-naming task and in the number memory task are not reported in the original study. (As a side note: what was even more surprising to us was that no RT outlier removal procedures are reported in the original study. It seems obvious that RT outliers may have dramatic effects in an ANOVA based on a small sample of participants and stimuli.) o Lines 265-266 states that the current data were 5 times more likely under the null model than under THE alternative model. But the correct way to phrase this, given the statistical test that was conducted, is that the current data were 5 times more likely under the null model than under ANY alternative model (i.e., it’s a two-sided test). Of course, the author needed to conduct this two-sided test because it was pre-registered, but it would be valuable to also add the Bayes Factor for the null model relative to THE (i.e., expected) alternative model (i.e., BF0+), which shows how much more likely the null is compared to the specific pattern of findings observed in the original study. (Please do mention explicitly that this is a non-registered, exploratory analysis.) This is a good point, and we have added this exploratory result (BF0+ = 16.43). o Lines 273-277 describes that most participants (57 in total) would use the internet to look up the answer to difficult questions. As such, it could be worthwhile to explore whether the observed pattern of data (i.e., the null effect) also holds for this specific subgroup of participants. This was one of the preregistered exploratory tests. We restricted the sample to participants (N=57) who would use the internet to look up the answers to difficult questions. The observed pattern of data was very similar in this subgroup (BF01 = 5.01; (BF0+ = 11.80). We have added this result to the revised manuscript. o Lines 385 mentions “the effect” in “to what degree the effect depends on cognitive load”, but based on both the current replication and the previous one, there is no effect. We have removed this sentence from the manuscript. o In the Results and General Discussion sections, the author – rightfully – mentions the caveat of post-hoc selection of a subset of the data for making inferences from the data (as was done in the original study, by presenting the results of 4 words out of the full stimulus set). To drive the point home, it could be valuable to show examples of a selection of 4 words for which such an effect would have been statistically significant in the present dataset (if any), and/or quantify how many random selections of 4 words would have yielded any pattern of significant effects. This is a nice idea, and we have already explored our data set in that way. However, things get a little complicated here. In our study, restricting the analysis to a selection of words adds a between-subject factor to the experimental design. This is because we strictly avoided word repetitions, so that “Google” was presented in the hard questions block for some participants and in the easy questions block for others. In her response to the first failed replication, Sparrow argues against the repetition of words: “The procedure described in the […] replication study indicates that Stroop words were repeated four times and that one six digit number was given to participants to memorize across the entire block of trials. This is the procedure for active thought suppression: when people are asked explicitly not to think about a single target word, they must engage in active suppression. In such cases, one would expect post-suppression rebound, which would mean that there would be increased reaction times for repeated trials of the same target word” (Sparrow, 2018). We followed Sparrow’s advice and did not repeat any words in our replication study. We write in the methods section (line 202): “As Sparrow (2018) in her reply to the replication strongly argues against presenting each target word more than once (referring to “active thought suppression”), we decided to randomly select words for the easy and hard blocks. From our pool of 48 words, eight computer terms and 16 general terms were presented in each block, so that each word was presented only once in the experiment. The words were randomly chosen for each participant. Of note, each word was presented twice in the original study, once in the easy block, and once in the hard block. Since we did not intend to select single words for post hoc pairwise tests, we did consider a strict “no word repetition” approach to be more in line with Sparrow’s (2018) suggestions.” We hope that the reviewer agrees that adding an exploratory mixed between/within-subjects analysis of our data would be beyond the scope of our manuscript. Reviewer 2: 1. I struggled to understand the differences between the method used in the present study and the method used in the original study. I suggest that the author list every known difference in clear, concrete language, perhaps in a numbered list. We fully agree with the reviewer that the differences are not easy to follow. Therefore, we have added a table with all known differences (see next page). Table 1. Differences between the original study by Sparrow et al. (2011) and the replication studies. (*) Information about the sample size is based on the original supplement. (**) This is an informed guess based on the original study and the response to the failed replication (Sparrow, 2018). Sparrow et al. (2011), Exp.1 Camerer et al. (2018) Hesselmann (2020) Data collection 2006 2017 2019 Presentation software DirectRT oTree oTree Participant sample Undergraduate students (USA) Students and non-students (USA) Students and non-students (Germany) Sample size N=46* N=104 & N=130 N=117 Trivia questions Original set (16 easy, 16 hard) Original set (16 easy, 16 hard) Revised & translated set (16 easy, 16 hard) Stroop words Original set (incl. 16 internet words) Original subset (incl. 8 internet words) Revised, translated & validated set (incl. 16 internet words) Number memory task On each Stroop trial 1x per block of trials (easy/hard) On each Stroop trial Number of presentations per Stroop word 1x per block of trials (easy/hard)** 2x per block of trials (easy/hard) 1x per experiment Total number of trials 48 (24 easy, 24 hard)** 96 (48 easy, 48 hard) 48 (24 easy, 24 hard) Participant debriefing Unknown Unknown Yes 2. On lines 222 through 226, the author writes that he removed older subjects from the sample because the sample in the original study included only undergraduate students, and this exclusion criterion was not listed in the preregistration. The author deserves kudos for making this disclosure. Still, I think readers need to see both version of the results: that is, a) with the older subjects, and b) without the older subjects. In fact, maybe the author could add a figure with three panels: one showing the results of the original study, one showing the results of the present study with all subjects, and showing the results of the present study without the older subjects. As suggested by the reviewer, we have added this information to end of the revised results section: “Finally, when all participants (N=117) were included in the data analysis, the pattern of results turned out to be 9 times more likely under the null model than under the alternative model (BF01 = 9.31)”. However, we decided not to add a new figure panel showing the additional “null” results from the complete sample. 3. I don’t think the present study needs a theoretical story, especially since the observed finding is a null result, so I would recommend omitting Figure 4 and much of the discussion. I think it makes more sense to frame the study as a replication attempt. As suggested by the reviewer, we have omitted Figure 4 from the revised manuscript, as well as some parts of the discussion. However, we do believe that the priming model proposed by Sparrow et al. (2011) needs to be made explicit and deserves to be criticized, as it remains rather vague in the original publication. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: Deep sedation may hamper the detection of neurological deterioration in brain-injured patients. Impaired brainstem reflexes within the first 24 hours of deep sedation are associated with increased mortality in non-brain-injured patients. Our objective was to confirm this association in brain-injured patients. Methods:</ns0:p><ns0:p>This was an observational prospective multicenter cohort study involving four neurointensive care units. We included acute brain-injured patients requiring deep sedation, defined by a Richmond Assessment Sedation Scale (RASS) &lt; -3. Neurological assessment was performed at day 1 and included pupillary diameter, pupillary light, corneal and cough reflexes, and grimace and motor response to noxious stimuli. Pre-sedation Glasgow Coma Scale (GCS) and Simplified Acute Physiology Score (SAPS-II) were collected, as well as the cause of death in the Intensive Care Unit (ICU). Results: 137 brain-injured patients were recruited, including 70 (51%) traumatic brain-injured patients, 40 (29%) vascular (subarachnoid haemorrhage or intracerebral haemorrhage). Thirty patients (22%) died in</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Patients with severe brain injury frequently receive early deep sedation <ns0:ref type='bibr' target='#b18'>(Oddo et al., 2016)</ns0:ref>. Deep sedation may improve cerebral hemodynamics by reducing the rate of cerebral oxygen consumption and by decreasing intracranial pressure. However, deep sedation may also delay awakening, induce delirium <ns0:ref type='bibr' target='#b18'>(Oddo et al., 2016)</ns0:ref> and, increase mortality <ns0:ref type='bibr' target='#b25'>(Shehabi et al., 2018)</ns0:ref>. In addition, sedation hampers clinical assessment of neurological status in brain-injured patients, potentially masking acute neurological worsening. Deep sedation also compromises the assessment of the patient's prognosis, which then relies on the pre-sedation examination (such as Glasgow Coma Scale). It is therefore challenging for ICU-physicians to routinely assess the neurological status of severely brain-injured patients requiring deep sedation. Nonetheless, ICU physicians have at their disposal various tools, including neurological examination and electrophysiological testing. The main difficulty, then, is to distinguish the effects of sedation from the consequences of underlying brain injury on clinical signs and the activity of the electroencephalogram (EEG). This is a highly complex situation compounded by the existence of both primary and secondary brain insults. Clinical relevance of neurological examination -especially brainstem reflex assessment -has previously been demonstrated in non-brain-injured, critically-ill patients <ns0:ref type='bibr' target='#b9'>(Foo, Loan &amp; Brennan, 2019)</ns0:ref>, including deeply sedated patients <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a;</ns0:ref><ns0:ref type='bibr' target='#b1'>Rohaut et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b1'>Azabou et al., 2017</ns0:ref><ns0:ref type='bibr' target='#b2'>Azabou et al., , 2018))</ns0:ref>. Assessment of brainstem reflexes is feasible and reproducible, and constitutes an early independent predictor of ICU-mortality, after adjustment to critical illness severity, sedation level, and sedative doses in non-braininjured, critically ill patients <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a;</ns0:ref><ns0:ref type='bibr' target='#b1'>Rohaut et al., 2017;</ns0:ref><ns0:ref type='bibr'>Azabou et al.,</ns0:ref> PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>2017,</ns0:ref><ns0:ref type='bibr'>2018)</ns0:ref>. Our pathophysiological hypothesis is that critical illness may be associated with brainstem dysfunction, which might itself be caused by the combined effects of critical illness and sedation, and may contribute to mortality, notably via a central autonomic dysfunction <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a;</ns0:ref><ns0:ref type='bibr' target='#b1'>Rohaut et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b3'>Benghanem et al., 2020)</ns0:ref>.</ns0:p><ns0:p>A previousstudy has shown the prognostic value of early (day 1) assessment of brainstem reflexes in deeply sedated, non-brain-injured patients <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a)</ns0:ref>. The main objective of the present study was to extend these previous findings to brain-injured patients.</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIAL AND METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>Study design and setting</ns0:head><ns0:p>This was a prospective, multicenter, international observational cohort study, approved by the ethics committee of Paris (Ile de France IV), France (Approval number 2014-AO1102-45) and Brescia <ns0:ref type='bibr'>(NP 1840</ns0:ref><ns0:ref type='bibr'>(NP , 04-11-2014))</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Reporting Studies in Epidemiology (STROBE) guidelines were followed thoroughly <ns0:ref type='bibr' target='#b5'>(von Elm et al., 2007)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Participants</ns0:head><ns0:p>Adult patients were eligible if they were deeply sedated following a major brain injury: severe traumatic brain injury, subarachnoid or intraparenchymal cerebral haemorrhage, ischemic stroke, or following a complicated neurosurgical or endovascular procedure.</ns0:p><ns0:p>Deep sedation was defined by a Richmond Assessment Sedation Scale (RASS) score &lt;-3 <ns0:ref type='bibr' target='#b7'>(Ely et al., 2003)</ns0:ref>. Patients were included if sedation lasted between 12 and 24 hours, and could be neurologically assessed by one of the PI (SK, JA, CR, CV, FR, NH, GM, GS, NL and FB) at working hours.</ns0:p><ns0:p>Patients were excluded if they were affected by a peripheral neurologic disorder involving the cranial nerves (e.g., Guillain Barr&#233; syndrome), or had been referred following cardiac arrest. Based on our previous findings in non-brain-injured patients, we planned to include 150 patients in the present study <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a;</ns0:ref><ns0:ref type='bibr' target='#b1'>Rohaut et al., 2017)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Sedation</ns0:head><ns0:p>Decisions to initiate or to withdraw sedation as well as the level of sedation were made by the physician in charge of the patient. We recorded the time and reason for initiating deep sedation as well as the Glasgow Coma Scale (GCS) prior to sedation. Since sedation was administered as part of the treatment of the cerebral insult, no systematic interruptions of sedation or decreasing trials were performed during the first 24 hours.</ns0:p><ns0:p>Nevertheless, depth of sedation was monitored using the RASS every four hours. After the first 24 hours, the possibility of discontinuing sedation was assessed on a daily PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed basis. Titration of sedation was performed at least twice daily, targeting physician in charge-defined RASS levels. The date of awakening, defined by spontaneous eye opening and visual contact &gt;10 sec (i.e., RASS &#8805; -1), was recorded.</ns0:p><ns0:p>Sedation was obtained through a continuous infusion of midazolam and/or propofol, in combination with sufentanil or fentanyl. Total cumulative doses of administered drugs at the time of neurologic examination were collected.</ns0:p></ns0:div> <ns0:div><ns0:head>Neurologic examination</ns0:head><ns0:p>The detailed procedure pertaining to neurological examination has been previously described <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a)</ns0:ref>. Briefly, we assessed: 1) depth of sedation (RASS);</ns0:p><ns0:p>2) reactivity, using the motor and eye response components of the GCS; 3) brainstem reflexes, including pupil size (miosis, normal, or mydriasis), pupillary light reflex, corneal reflex, facial muscle movement in response to noxious stimulation of the temporomandibular joint, and the cough reflex in response to tracheal suctioning.</ns0:p><ns0:p>Neurologic examination was performed after 24 &#177; 12 hours of continuous sedation (day 1). Retaining the same definition used in our previous studies, corneal and pupillary reflexes were considered abolished only when both right and left side reflex were abolished. The oculocephalic response to lateral passive head rotation was not performed in traumatic brain injured patients. The Full Outline of Unresponsiveness (FOUR) score was recorded <ns0:ref type='bibr' target='#b29'>(Wijdicks et al., 2005)</ns0:ref>. At the time of neurological examination, intracranial pressure (ICP) was recorded if available. Intracranial hypertension was defined as ICP &gt; 25 mmHg lasting for more than 5 minutes or by the need of an additional treatment to control the intracranial pressure (additional sedative <ns0:ref type='table' target='#tab_0'>PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:ref> Manuscript to be reviewed agent, ventricular drainage, craniotomy with hematoma evacuation or decompressive craniectomy).</ns0:p></ns0:div> <ns0:div><ns0:head>Baseline clinical and biological and imaging data</ns0:head><ns0:p>Demographic characteristics, body weight, date, time and cause of ICU admission, comorbidities using McCabe score <ns0:ref type='bibr' target='#b17'>(McCabe &amp; Jackson, 1962)</ns0:ref>, date of invasive mechanical ventilation (MV) initiation and its duration, ICU length of stay, occurrence of microbiologically documented ventilator-associated pneumonia as well as the date and cause of death were recorded. The Simplified Acute Physiological Score II (SAPS-II) (Le <ns0:ref type='bibr' target='#b13'>Gall et al., 2005)</ns0:ref>, the Sequential Organ Failure Assessment (SOFA) <ns0:ref type='bibr' target='#b27'>(Vincent et al., 1996)</ns0:ref> as well as key interventions and standard biological tests needed to calculate these scores were recorded. Clinical, biological and neuroradiological data were collected as part of the routine care.</ns0:p><ns0:p>The presence or absence of infratentorial lesions, including brainstem lesions, was assessed on the first cerebral computed tomography scanner (CT-scan) performed within 24 hours following ICU admission, by the senior neuroradiologist and ICU physician in charge.</ns0:p></ns0:div> <ns0:div><ns0:head>Outcomes</ns0:head><ns0:p>Primary recorded outcome was ICU mortality. Secondary outcomes were the occurrence of delayed awakening and of delirium following sedation discontinuation.</ns0:p><ns0:p>Delayed awakening was defined by RASS&lt;-1 despite discontinuation of sedation for more than 72 hours. Delirium was assessed daily using the confusion assessment method for the ICU (CAM-ICU) <ns0:ref type='bibr' target='#b6'>(Ely et al., 2001)</ns0:ref> . Other secondary outcomes were the duration of mechanical ventilation, the length of stay in the ICU and the occurrence of Withdrawal of life-sustaining therapies (WLST) was determined according to applicable French and Italian Law. In current practice, no such decision involves the result of early neurological examination in a deeply-sedated patient.</ns0:p></ns0:div> <ns0:div><ns0:head>Bias and confounding factors</ns0:head><ns0:p>We sought to mitigate potential confounding factors that might influence neurological examination as well as relevant outcomes, namely mortality and the occurrence of delirium. Neurological examination was performed by senior ICU physicians who were either neurologists or neuro-intensivists, or were specifically trained by a senior neurologist. The investigator was different from the clinician in charge of the patient.</ns0:p><ns0:p>Inter-observer agreement for such an examination has been shown to be satisfactory (kappa scores ranged from 0.62 to 1) <ns0:ref type='bibr'>(Sharshar et al., 2011b)</ns0:ref>. Brainstem reflexes are routinely assessed in neuro-ICU patients. Management of deep sedation was assessed by collecting cumulative doses and duration of sedation as well as daily RASS. The cause of death and its main risk factors were also assessed, including the GCS prior to sedation, SAPS-II and the SOFA score as well as the cause of brain injury.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>Data are reported as numbers (percentage), mean (standard deviation), or median (inter-quartile range). Groups were compared using the Wilcoxon rank sum test or the PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed chi-square test. Logistic regression was used to explore the associations between the pre-sedation GCS, SAPS-II, RASS scores, sedation doses, cough reflex, corneal reflex, and ICU mortality. These variables were determined a priori according to our previous findings <ns0:ref type='bibr' target='#b1'>(Rohaut et al., 2017)</ns0:ref> . Since the number of events was limited, we used Firth penalized logistic regression for multivariable models, in order to limit small-sample bias. Discrimination of multivariable models was assessed using the concordance (c) index, which is equivalent to the area under the receiver operating characteristics curve.</ns0:p><ns0:p>It varies theoretically between 0.5 and 1.0, a value of 1.0 indicating a perfect discrimination. Missing data were handled by multiple imputation by chained equations, all variables considered for analysis being used in the imputation model. Since about half patients had at least one variable missing (35 patients had missing data for one variable, one had missing data for seven), 50 imputed datasets were generated <ns0:ref type='bibr' target='#b28'>(White, Royston &amp; Wood, 2011)</ns0:ref>. Each imputed dataset was analysed separately, and estimates were then pooled to provide point estimates and confidence intervals (CI) <ns0:ref type='bibr' target='#b15'>(Marshall et al., 2009)</ns0:ref>. P-values &lt;0.05 were considered as statistically significant. All analyses were carried out using the R statistical software version 3.4.1 (The R Foundation for Statistical Computing, Vienna, Austria).</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head></ns0:div> <ns0:div><ns0:head>Patient characteristics</ns0:head><ns0:p>Among 151 consecutive brain injured patients receiving deep sedation within 12 hours of admission, a total of 14 were excluded (Figure <ns0:ref type='figure'>1</ns0:ref>). Overall, 137 patients were enrolled; their baseline characteristics are presented in Manuscript to be reviewed (mainly post-surgical or endovascular procedure) in 27 (20%). Mechanical ventilation was initiated for neurological reasons (airway management in comatose patients) in 120 (88%) patients. Other reasons for intubation and mechanical ventilation were acute respiratory failure in 6 patients (4%), shock in one patient (1%), post-operative period in 6 patients (4%), and undetermined in 4 patients (3%). The median age was 50 [34 to 63] years, 80 (59%) patients were male, and the median SAPS II was 46 [36 to 55].</ns0:p></ns0:div> <ns0:div><ns0:head>Neurological features of the patients</ns0:head><ns0:p>Main neurological features are presented in Table <ns0:ref type='table'>2</ns0:ref>. The median GCS prior to sedation was 7 [4 to 11]. At the time of inclusion, all patients were deeply sedated (RASS &lt;-3), with 90 (66%) patients exhibiting RASS -5. Deep sedation was obtained using one (n=114, 83%), two (n=21, 15%) or three (n=2, 2%) hypnotic agents. One hundred twenty-two (89%) patients received midazolam, 35 (26%) propofol and 5 (4%) sodium thiopental. Sufentanil was administered to 107 (78%) patients. At inclusion, median FOUR score was 4 [2 to 5]. The most frequently preserved brainstem reflex was the corneal reflex, present in 103 patients (79%), and grimacing to pain was the most frequently abolished reflex (present in 32 patients, 30%). The cough reflex was present in 100 patients (74%). The proportion of patients with RASS -5 was greater in patients with abolished cough reflex (80% vs. 60%, p=0.039). Similarly, the cumulative doses administrated at time of clinical exam of midazolam and opioid agents were greater in patients with abolished cough reflex (1.2 vs. 1.9 mg/kg, p=0.005 and, 5.2 vs. 8.5 &#181;g/kg, p=0.009 respectively) (Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p><ns0:p>Brain imaging was performed for 133 patients (97%) by a CT scan, of whom 45 (34%) patients exhibited infra-tentorial lesions including brain stem lesions.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>ICU mortality</ns0:head><ns0:p>Thirty patients (22%) died in the ICU (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). Causes of death were multi-organ failure in 6 patients (20%), brain death in 5 patients (17%) and, withdrawal of life sustaining therapies (WLST) in 16 patients (53%). Cause of death was not reported for 3 patients (10%). WLST-related death occurred at day 8 [6-17] (ranging from day 3 to day 48 after admission). SAPSII (OR 1.06 per unit, p=0.001) was significantly greater in nonsurvivors (Table <ns0:ref type='table' target='#tab_1'>4</ns0:ref>). At admission, GCS were significantly lower (OR 0.80 per unit, p= 0.004) in non-survivors. The proportion of patients with RASS -5 (p=0.32) did not statistically differ between the two groups. The corneal (OR 2.69, p=0.034) and cough reflexes (OR 5.12, p=0.0003) were more frequently abolished in non-survivors. After adjustment to pre-sedation GCS, SAPS-II, and RASS (OR: 5.19, 95% CI: [1.92-14.1], p=0.001) or midazolam and sufentanyl doses (OR: 8.89, 95% CI: [2.64-30.0], p=0.0004), an abolished cough reflex was associated with ICU mortality.</ns0:p></ns0:div> <ns0:div><ns0:head>Other outcomes</ns0:head><ns0:p>Median duration of sedation was 5 [3 to 10] days while median duration of mechanical ventilation was 13 [8 to 24] days. Median time to awakening was 4 [1 to 12] days following discontinuation of sedation. Delayed awakening was observed in 67 patients (51%) and delirium in 66 patients (59%). Eighty-six patients (64%) developed ventilator associated pneumonia. Among the 103 patients in whom intracranial pressure was monitored, 50 patients (49%) suffered from at least one episode of elevated intracranial pressure.</ns0:p><ns0:p>PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>In this prospective, multicenter cohort study we found that, in brain injured patients requiring deep sedation, early abolition of the cough reflex (day 1) was associated with ICU mortality after adjustment for severity of illness (i.e. SAPS-II), brain injury (i.e. presedation GCS), depth of sedation (i.e. RASS) and sedative and analgesic doses.</ns0:p><ns0:p>Moreover, abolition of the cough reflex was independent of the type of injury.</ns0:p><ns0:p>These results extend our previous findings obtained in non-brain injured patients <ns0:ref type='bibr' target='#b23'>(Sharshar et al., 2011a;</ns0:ref><ns0:ref type='bibr' target='#b1'>Rohaut et al., 2017)</ns0:ref>. It supports our hypothesis that mortality could result from dysfunction of the brainstem, which controls vital functions via the autonomic nervous system <ns0:ref type='bibr' target='#b12'>(Heming et al., 2017)</ns0:ref>. Thus, abolition of the cough reflex could be a clinical marker of dysfunction in the medulla, which integrates cardiovascular and respiratory centers. Interestingly, critical illness is associated with a reduction in heart rate and tidal volume variability, which, as markers of impaired central control, are associated with organ failure, mortality, and weaning failure from MV <ns0:ref type='bibr' target='#b0'>(Annane et al., 1999;</ns0:ref><ns0:ref type='bibr' target='#b30'>Wysocki et al., 2006)</ns0:ref> .</ns0:p><ns0:p>The putative mechanisms of brainstem dysfunction can be related to three nonmutually exclusive mechanisms: 1) primary brain injury; 2) neuro-inflammation; 3) oversedation <ns0:ref type='bibr' target='#b3'>(Benghanem et al., 2020)</ns0:ref>.</ns0:p><ns0:p>(1) Cough reflex abolition may result from a direct brainstem injury related to haemorrhage, axonal lesions, herniation brain swelling etc. Specific assessment of brainstem injury would have required specific brain imaging (e.g. magnetic resonance imaging, diffusion tensor imaging) <ns0:ref type='bibr' target='#b8'>(Fischer et al., 2016)</ns0:ref> which was out of the scope of PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed this observational study. Patient management, including the choice of brain imaging within the first 24 hours, was under the responsibility of ICU-physicians. In this context, CT scan is the most commonly performed exam <ns0:ref type='bibr'>(Connolly et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b11'>Geeraerts et al., 2018)</ns0:ref>.</ns0:p><ns0:p>(2) In addition to the direct injury, systemic inflammation can cause neuro-inflammatory brainstem injuries. These lesions are usually radiologically undetectable. Circulating mediators can directly reach the brainstem through the area postrema. This zone's lack of a blood-brain barrier allows neuro-inflammatory insults and neuronal apoptosis, notably within the medullary autonomic and respiratory centers <ns0:ref type='bibr' target='#b22'>(Sharshar et al., 2002</ns0:ref><ns0:ref type='bibr' target='#b20'>(Sharshar et al., , 2003))</ns0:ref>. As described in septic acute encephalopathy or delirium in ICU, this mechanism could occur as early as 24 hours after brain injury <ns0:ref type='bibr' target='#b16'>(Mazeraud, Bozza &amp; Sharshar, 2018;</ns0:ref><ns0:ref type='bibr' target='#b26'>Slooter et al., 2020)</ns0:ref> (3) Finally, as respiratory centers are sensitive to sedation <ns0:ref type='bibr' target='#b1'>(Rohaut et al., 2017)</ns0:ref> especially opioids, cough reflex abolition could be a direct effect of sedation. The dose of sedatives and opioids and the depth of sedation (according to the RASS) were greater in patients with absent cough reflex, which can therefore be considered a marker of oversedation. However, abolition of cough reflex remains associated with mortality after adjustment on sedation levels as well as sedative and analgesic doses, suggesting that its abolition integrates other processes than sedation alone. It would be interesting to test the predictive value of cough reflex abolition against EEG, which may be more accurate to assess depth of sedation.</ns0:p><ns0:p>The conclusion of our study is pragmatic; severity and prognostic assessment of braininjured patients requiring deep sedation should integrate assessment of brainstem PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed reflexes, especially the cough reflex. This assessment can be performed by using the FOUR score <ns0:ref type='bibr' target='#b29'>(Wijdicks et al., 2005;</ns0:ref><ns0:ref type='bibr' target='#b9'>Foo, Loan &amp; Brennan, 2019)</ns0:ref>. It must be noted that information regarding the neurological components of the SAPS-II and SOFA severity scores, as well as of the RASS sedation scale, is limited. Indeed, the SAPS-II and SOFA score include the GCS value before sedation; while the RASS relies solely on the patient's motor reactivity to verbal or non-painful physical stimulation. Brainstem reflex assessment within the first 24 hours of sedation provides clinicians with a comprehensive and temporally integrative neurological evaluation.</ns0:p></ns0:div> <ns0:div><ns0:head>Limits</ns0:head><ns0:p>Our study has several limits. First, the causes of brain injuries are multiple.</ns0:p><ns0:p>Nevertheless, except for grimacing to pain and pupillary light reflexes, the neurological patterns were similar among the different subpopulations (Table <ns0:ref type='table'>2</ns0:ref>). Our results suggest that the association between cough reflex abolition and ICU mortality is independent from the cause of brain injury. Second, the optimal way to assess the depth of sedation remains a matter of debate. In accordance with our previous findings, we chose to assess the depth of sedation using a specifically designed clinical scale (the RASS; Table <ns0:ref type='table' target='#tab_1'>4</ns0:ref>, model 1). However, since doses of sedation were significantly different within patients with preserved and abolished cough reflex, we also decided to include this parameter in our multivariate analysis (Table <ns0:ref type='table' target='#tab_1'>4</ns0:ref>, model 2). Continuous EEG monitoring and/or determining plasma levels of sedatives could have provided more detail regarding the level of sedation, however these assessments are not routinely undertaken. Based on our results, and without measurement of plasmatic sedation levels, assessing the cause of cough reflex inhibition is not feasible. It would be PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed interesting to determine the prognostic value of the cough reflex against EEG, and to assess whether its abolition depends on circulating levels of sedatives. Third, as the main cause of death was withdrawal of life-sustaining therapies, our study could have been exposed to the self-fulfilling prophecy bias. However, the proportion of WLST in our study was comparable to other reports <ns0:ref type='bibr' target='#b14'>(Leblanc et al., 2018)</ns0:ref>. In addition, it is not a common clinical practice to base WLST solely on day 1 neurological examination. In our study, no WLST-related death occurred before day 3. Nevertheless, a neurological examination at day 1, 2, and 3 could have been more valuable than solely at day 1, and this must be further evaluated. It must also be noted that since the reason for ICU discharge was not prospectively assessed, we cannot rule out that patients have been discharged from ICU for palliation on the ward and have died on the ward. Only ICU mortality at day 28 was addressed. Fourth, our data does not allow any exploration of the causality of brainstem dysfunction. In addition to the proposed mechanisms (i.e primary injury, oversedation or neuroinflammation), the causal link between brainstem failure and multi-organ failure (which was responsible for 20% of the mortality in our study) has to be explored. Understanding to what extend brainstem dysfunction could be a cause or a consequence of multi-organ failure would need further investigation.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Taken together, our results suggest that an absence of cough reflex is an independent predictor of ICU-mortality in deeply sedated brain-injured patients. This is an important extension of our previous findings in deeply sedated non-brain-injured ICU patients. Manuscript to be reviewed </ns0:p></ns0:div> <ns0:div><ns0:head>Abolition of cough reflex possibly reflects a brainstem dysfunction that could</ns0:head></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>, Italy. Written informed consent was obtained from the patients' legal representative in France. In Italy, the Ethics Committee waived the requirement for consent because relatives are not regarded as legal representatives of the patient in the absence of a formal designation. The current study was a pilot study preceding the design of a larger prospective multicentre study assessing the prognostic value of brainstem dysfunction in sedated critically ill patients (ClinicalTrials.gov number: NCT02395861). Patients were recruited in four intensive care units, including one neuro ICU and three medical and surgical ICUs. Patients were recruited between December 2011 and February 2015. The Strengthening the PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020) microbiologically documented ventilator-associated pneumonia. Patients had follow-up until ICU discharge.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)Manuscript to be reviewed compromise vital functions. This brainstem dysfunction might result from over-sedation combined with primary and secondary cerebral insults. Assessment of brainstem reflexes may complete SAPS-II and RASS scores for evaluating the severity and depth of sedation, and help predict the prognosis of deeply sedated brain injured patients.Future research could aim at assessing the safety and efficacy of sedation titration aimed at preserving the cough reflex without compromising the control of intracranial pressure in brain-injured patients.PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Cause of brain injury was blunt trauma in 70 (51%) patients, vascular in 40 patients (29%) and miscellaneous</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2019:05:37617:1:2:NEW 6 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 4 . Association of neurological response with ICU death.</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>Unadjusted</ns0:cell><ns0:cell /><ns0:cell>Multiple model 1</ns0:cell><ns0:cell /><ns0:cell>Multiple model 2</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>OR (95% CI)</ns0:cell><ns0:cell>P</ns0:cell><ns0:cell>aOR (95% CI)</ns0:cell><ns0:cell>P</ns0:cell><ns0:cell>aOR (95% CI)</ns0:cell><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell>SAPS-II (per unit)</ns0:cell><ns0:cell cols='2'>1.06 (1.02-1.09) 0.002</ns0:cell><ns0:cell>1.04 (1.00-1.08)</ns0:cell><ns0:cell>0.086</ns0:cell><ns0:cell>1.04 (0.99-1.08)</ns0:cell><ns0:cell>0.092</ns0:cell></ns0:row><ns0:row><ns0:cell>GCS at admission (per unit)</ns0:cell><ns0:cell cols='2'>0.80 (0.69-0.93) 0.004</ns0:cell><ns0:cell>0.88 (0.74-1.04)</ns0:cell><ns0:cell>0.13</ns0:cell><ns0:cell>0.87 (0.72-1.04)</ns0:cell><ns0:cell>0.12</ns0:cell></ns0:row><ns0:row><ns0:cell>RASS -5</ns0:cell><ns0:cell cols='2'>1.58 (0.64-3.91) 0.32</ns0:cell><ns0:cell>0.74 (0.26-2.17)</ns0:cell><ns0:cell>0.59</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>Midazolam dose (mg/kg)</ns0:cell><ns0:cell cols='2'>0.69 (0.44-1.11) 0.12</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.42 (0.14-1.25)</ns0:cell><ns0:cell>0.12</ns0:cell></ns0:row><ns0:row><ns0:cell>Morphinic dose (&#61549;g/kg)*</ns0:cell><ns0:cell cols='2'>0.97 (0.88-1.08) 0.59</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>1.01 (0.82-1.25)</ns0:cell><ns0:cell>0.91</ns0:cell></ns0:row><ns0:row><ns0:cell>Absent cough reflex</ns0:cell><ns0:cell cols='2'>5.12 (2.13-12.4) 0.0003</ns0:cell><ns0:cell>5.19 (1.92-14.1)</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell>8.89 (2.64-30.0)</ns0:cell><ns0:cell>0.0004</ns0:cell></ns0:row><ns0:row><ns0:cell>Absent corneal reflex</ns0:cell><ns0:cell cols='2'>2.69 (1.08-6.68) 0.034</ns0:cell><ns0:cell>1.71 (0.57-5.10)</ns0:cell><ns0:cell>0.34</ns0:cell><ns0:cell>1.66 (0.54-5.08)</ns0:cell><ns0:cell>0.38</ns0:cell></ns0:row><ns0:row><ns0:cell>c index (95% CI)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.810 (0.726-0.893)</ns0:cell><ns0:cell /><ns0:cell>0.852 (0.773-0.931)</ns0:cell></ns0:row></ns0:table></ns0:figure> </ns0:body> "

" Montreal, July 24th 2020 Dear editors, on behalf of all the co-authors, I would like to thank the reviewers for their generous comments on the manuscript. We have edited the manuscript to address their concerns. In particular we have addressed the major concern of reviewer 2 and have acknowledged in the discussion the fact that our results did not enable any precision on how brainstem failure and multi-organ failure were interrelated. We hope that the manuscript is now suitable for publication in PeerJ. Best regards, Dr Stanislas Kandelman Assistant Professor, McGill University Health Center, Montreal, Canada (formerly and when the study was performed, Praticien hospitalier, Hôpital Beaujon, Clichy, France) Reviewer 1 (Ching Chung Foo) Basic reporting I would like to congratulate the authors for the interesting study. I commend the use of the STROBE reporting guidelines. We wish to thank the reviewers for the time they took to review our manuscript. I think the study design (cohort study) should be mentioned in the title or abstract and the study design section. Agreed. We add the fact that it was a cohort study in the abstract and the study design section. Follow-up / loss to follow-up have not been mentioned but presumably every participant had follow-up until they were discharged from ICU? This is correct, all the patients included in the study had follow up until ICU discharge: this information was added in the manuscript It has been mentioned in text that FOUR score was lower and SOFA and age were higher in non-survivors. I think Table 4 should include this data for clarity.  Thank you for this comment. Table 4 displays variables that were defined a priori for multivariate analysis according to our previous findings that is: pre-sedation GCS, SAPS-II, RASS, cough reflex, corneal reflex, and doses of sedation. The SOFA and FOUR scores and the age were not included a priori as defined, thus not displayed in Table 4. To make manuscript clearer, we have decided to remove the univariate results displayed in the RESULTS part, but not displayed in the table 4. We also corrected the statement that FOUR score would be part of the multivariate analysis. The p value for GCS (line 257) does not match the value in Table 4. Agreed, the manuscript was modified. Experimental design Recruitment method - were all patients admitted to the designated ICU recruited and assessed for eligibility or the investigators decided which patients to be recruited? All patients admitted at working hours in the four designated ICU during the inclusion period were assessed for eligibility at the condition that the designated PI (SK, JA, CR, CV, FR, NH, GM, GS, NL and FB) for each center was present on site to perform neurological assessment. Thank you for this question that lead to a precision in the manuscript. Study size - how was the size determined? Was power calculation done? Thank you for this question. There was no power calculation done for the study. We based the size of this study on a previous study (Sharshar et al. Crit Care Med 2011, reference number 4) that had included 144 patients to assess the association between ICU mortality and brainstem reflexes in medical ICU patients. Although the design of the study was different, we aimed at including a comparable number of patients.   - The efforts to mitigate potential confounding factors is laudable. Validity of the findings - I think the limitations of the study have been addressed adequately at the discussion. The conclusion states that early cough reflex abolition is an independent predictor of mortality in deeply sedated brain-injured patients. Perhaps it is more accurate to limit that to ICU mortality as patients did not get follow-up after they were discharged from ICU? Agreed. We have added this precision in the manuscript. Had there been any patients discharged from ICU for palliation on the ward and had these been included in the mortality count? Unfortunately, we have not prospectively assessed the reason for discharge from ICU, thus we cannot rule out that some patients have died in the ward after a palliation decision was made in the ICU. The mortality count only addresses the ICU mortality at day 28. This precision has been added to the manuscript. As you have mentioned in the limitations, doses of sedation were significantly different between those with preserved and abolished cough reflex. 20% of the ICU mortality died secondary to multi-organ failure. To what extent do you think the absence of cough reflex is a result of the high levels of sedative when the plasma level is unknown, and whether this might change the conclusion? We believe that over sedation may be a cause of abolished cough reflex or a marker of more severe patients (needing more sedation). Nevertheless, abolished cough reflex remains associated with ICU mortality after adjustment either on sedation levels or on sedative and analgesic doses. Although the conclusion remains unchanged, the causal mechanism still needs to be assessed more precisely. We have added this precision in the discussion Comments for the Author Some studies have shown that neurological assessment within the first day of brain injury may not be reliable / as good as those performed later in terms of predicting mortality. As you mentioned, it is not common practice to base decision for withdrawal of life supporting treatment solely on day 1 neurological examination findings. Would it be more valuable to assess the prognostication of cough reflex later (e.g. day 3)? We agree with the reviewer. As stated in the introduction, the main objective of the present study was to extend our previous findings concerning the prognostic value of early (day 1) assessment of brainstem reflexes in deeply sedated non-brain injured patients to brain-injured patients. In order to extend these findings, we have decided to keep the day 1 assessment, although assessing neurological examination later (e.g. day 2 or 3) would may be more valuable. The manuscript has been modified to clarify this point. As explained in the MATERIAL AND METHODS section, the present study is a pilot study for a currently ongoing larger study, that involves neurophysiological examinations (evoked potentials, EEG) in addition to the clinical examination (ClinicalTrial NCT0239586). In this ongoing trial, we have decided to assess neurological examination at day 1, 2 and 3. Reviewer 2 (Daniel Kondziella) Basic reporting Grammar and style are, by and large, acceptable but could be improved by having a native speaker read through the manuscript.  We thank the reviewer for this comment. A native speaker kindly accepted to read through the manuscript. We hope that the corrections he proposed concerning grammar and style improved the manuscript and that PeerJ publication standards are reached. Scientific background and context, text structure and literature references are adequate. Results are self-contained and relevant to the main hypothesis about brainstem dysfunction in the ICU. Experimental design This is an observational study with research questions that are within the aims and scope of the journal. Ethical and technical standards are explained in sufficient detail. Validity of the findings The findings are valid; some suggestions are provided below. Comments for the Author Kandelman et al. present data related to brainstem reflexes and clinical outcome from a binational, multicenter, prospective observational study on 137 brain-injured patients from the intensive care setting. Their main finding is that abolishment of the cough reflex within 24h after intubation predicted mortality, which remained significant after statistical adjustment for various confounding factors including sedation. This finding from brain-injured patients is in line with the same authors’ earlier data from non-brain injured patients.  The authors argue that, taken together, these results point towards brainstem dysfunction as an important and hitherto overlooked cause of mortality and morbidity in the intensive care unit. This is an intriguing and clinically highly relevant suggestion, which is likely to be picked-up by other researchers in this field.  The methods and statistics appear sound, the manuscript is well-structured (although the English could perhaps be improved somewhat), STROBE guidelines were followed, and the inferences made are sufficiently discussed, including most (but not all) limitations. The overall conclusion is therefore that the present manuscript adds an important novel aspect to the intensive care literature. However, I still have some points of criticism. MAJOR CONCERNS The authors discuss most limitations appropriately; the most important being lack of MRI assessing presence or absence of brainstem pathology and lack of EEG, respectively, sedative serum levels to assess the degree of sedation as the most significant confounder related to the abolishment of brainstem reflexes, including the cough reflex. However, although the possibility of brainstem dysfunction is certainly exciting, the data presented are inadequate to make inferences about causality – does brainstem dysfunction lead to multiorgan failure or vice versa? Obviously, these effects will almost certainly be bidirectional, but to which extent one or the other direction predominates cannot be answered by the data presented; this should be more clearly acknowledged. Agreed. Manuscript has been revised and this issue has been acknowledged in the discussion. Also, the authors speculate that neuroinflammation may predispose to brainstem dysfunction, but is inflammation not rather unlikely to be a prominent factor after just 24 hours? We thank the reviewer for this question. We believe that neuroinflammation could be responsible for brainstem dysfunction as early as day 1, as it is described for septic acute encephalopathy or delirium in ICU. The manuscript has been modified with corresponding references. MINOR ISSUES Although not strictly necessary, I think the manuscript would benefit from a detailed figure illustrating the concept of brainstem dysfunction, including the relationship between brainstem anatomy and pathophysiological mechanisms (and the bidirectional effects with systemic organ failure as outlined earlier). We are grateful to the reviewer for this remark. As explained in the MATERIAL AND METHODS section, the present study is a pilot study for a currently ongoing larger study, that involves neurophysiological examinations (evoked potentials, EEG) in addition to the clinical examination (ClinicalTrial NCT0239586). A detailed figure illustrating the concept of brainstem dysfunction will surely be part of this upcoming publication, with a more comprehensive part on the pathophysiology. The relationship with systemic organ failure will have to be displayed on this figure. Line 82-83: You may consider adding “… by reducing the rate of cerebral oxygen consumption and by decreasing intracranial pressure”. Agreed. Manuscript has been modified. Line 88: Should be “severely” Agreed. Manuscript has been modified. Line 89/90: Should be “…examination and electrophysiological testing” Agreed. Manuscript has been modified. Line 96: Should be “receiving deep sedation” or better “deeply sedated patients” Agreed. Manuscript has been modified. Lines 96-99: This statement must be backed up by a reference. Agreed, corresponding references have been added. Line 103-104: Was the main objective not to extend the findings from non-brain injured critically ill  patients (reference 3) to brain-injured patients, rather than a mere confirmation? Agreed, the manuscript has been modified. Line 118: ”December 2011 and February 2015” – why the delay in reporting? The manuscript was submitted to journals with longer than expected reviewing processes. We believe that our data is still relevant in 2020 and hope that the reviewers will believe so as well. Line 131: Should be “Decisions…. mere made…” (decisions are not managed) Agreed. Manuscript has been modified. Line 152-153: Your previous results were based on non-brain injured patients, whereas your present study involved brain-injured patients, including those with lateralized structural brainstem lesions – in these patients, absence of unilateral (i.e. ipsilateral) brainstem reflexes might carry a worse prognosis. Please comment. We thank the reviewer for this comment. We agree with the reviewer on the fact that in this brain injured population, absence of unilateral brainstem reflex could be observed more frequently than in non-brain injured population and that it could be associated with a worse prognosis. Nevertheless, in order to extend our findings to brain injured patients, and because the mechanism underlying the brainstem dysfunction is not focal, we decided to keep the bilateral absence as a definition for “absent reflex”. Statistically, this definition (instead of unilateral OR bilateral absence) can be underestimating rather than overestimating the association between ICU mortality and abolished cough reflex. A precision was made in the manuscript. Line 154: Why not in TBI patients? Provided cervical spine injury has been excluded, VOR assessment is safe. In one of the four centers involved in the study, VOR is not being assessed in TBI patients, even after cervical spine injury has been ruled out, cervical strain (causing cervical instability) remaining possible even after normal CT scan. Line 155: Should be “Unresponsiveness” Agreed, the manuscript has been modified. Line 226: “neurological reasons” – please be more specific, and also state what the remaining reasons for intubation were. We thank the reviewer for this remark. Neurological reasons implied airway management in comatose patients. Remaining reasons were acute respiratory failure (4%), shock (1%), post-operative period (4%), and undetermined (3%). The manuscript has been modified. Lines 246-248: How come that infratentorial lesions were associated with better survival? This seems counterintuitive, please comment. We thank the reviewer for this remark. In our early CT-scan based classification of lesions, cerebellum lesions which are not systematically associated with a poor prognosis, were mixed with brainstem lesions, all of those localizations being infra tentorial. This could be explaining the result of the univariate analysis. Since this parameter was not defined a priori for the multivariate analysis, since CT scan is not the reference imaging technique for brainstem lesions, and because our definition of infratentorial mixed very different types of lesions associated with different prognosis, we have decided to remove this result from the manuscript, hoping the reviewer will consider this modification as an improvement. Lines 325/326: Should be “limitations” Agreed, the manuscript has been modified Line 345: This section should be marked as “Conclusions”. Agreed, the manuscript has been modified. Line 349: Should be “over-sedation” Agreed, the manuscript has been modified. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Prochilodus magdalenae is a freshwater fish endemic to the Colombian hydrographic Magdalena-Cauca and Caribe basins. The genetic structure patterns of populations of different members of Prochilodus and the historic restocking of its depleted natural stocks suggest that P. magdalenae exhibits genetic stocks that coexist and co-migrate throughout the rivers Magdalena, Cauca, Cesar, Sin&#250;, and Atrato. To test this hypothesis and explore the levels of genetic diversity and population demography of 725 samples from the studied rivers, we developed a set of 11 species-specific microsatellite loci using next-generation sequencing, bioinformatics, and experimental tests of the levels of diversity of the microsatellite loci. The results evidenced that P. magdalenae exhibits high genetic diversity, significant inbreeding coefficient ranging from 0.162 to 0.202, and signs of erosion of the genetic pool. Additionally, the population genetic structure constitutes a mixture of genetic stocks heterogeneously distributed along the studied rivers, and moreover, a highly divergent genetic stock was detected in Chucur&#237;, Puerto Berr&#237;o, and Palagua that may result from restocking practices. This study provides molecular tools and a wide framework regarding the genetic diversity and structure of P. magdalenae, which is crucial to complement its baseline information, diagnosis and monitoring of populations, and to support the implementation of adequate regulation, management, and conservation policies.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>The family Prochilodontidae (Teleostei: Characiformes) comprises the genera Prochilodus, Semaprochilodus, and Ichthyoelephas, and encompasses 21 Neotropical freshwater fish species in the main river basins of South America <ns0:ref type='bibr'>(Castro &amp; Vari, 2004)</ns0:ref>. Most of the prochilodontids exhibit large body sizes, high fecundities, and abundances, representing around 50-80% of the biomass caught by the subsistence and commercial fisheries in some regions of their distribution area <ns0:ref type='bibr' target='#b2'>(Barroca et al., 2012b;</ns0:ref><ns0:ref type='bibr'>Melo et al., 2016a)</ns0:ref>. Furthermore, some members of Prochilodontidae constitute a potential resource for fish farming due to certain characteristics such as their fast growth and weight increase, rustic management, and high economic value <ns0:ref type='bibr'>(Flores-Nava &amp; Brown, 2010;</ns0:ref><ns0:ref type='bibr'>DellaRosa et al., 2014;</ns0:ref><ns0:ref type='bibr'>Roux et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In addition to the economic importance, Prochilodontidae plays an important trophic role in aquatic ecosystems. These detritivorous and migratory fishes contribute to the nutrient cycling, distribution, equilibrium, and maintenance of energetic flows and support a wide trophic network for a great number of predators <ns0:ref type='bibr'>(Flecker, 1996)</ns0:ref>. Hence, the adequate management of fisheries is crucial for the maintenance of high productivity and permanent resource availability, as well as to guarantee the stability and continuity of the aquatic ecosystems <ns0:ref type='bibr'>(Taylor, Flecker, &amp; Hall, 2006;</ns0:ref><ns0:ref type='bibr'>Batista &amp; Lima, 2010)</ns0:ref>.</ns0:p><ns0:p>The bocachico Prochilodus magdalenae Steindachner 1878 is the most representative endemic species of the Colombian ichthyofauna, considered the emblematic fishery resource of the Magdalena-Cauca Basin, with an estimated unload for the Magdalena Basin of 2,182.67 metric tons in 2013 (Colombian fishing statistical service: SEPEC). However, between 1978 and 2012, this species experienced drastic decreases in its population densities, catches (approx. 85%), and mean catch sizes. These effects resulted from overfishing during migratory periods, violations of legislation related to mean catch sizes, and habitat disturbances including deforesting, floodplain lake desiccations, agrochemical or chemical contamination The PCRs were conducted in a volume of 10 &#181;l, which contained 2-4 ng/&#181;l of template DNA isolated with the GeneJET Genomic DNA purification kit (Thermo Scientific, Karlsruhe, Germany) following the manufacturer&#180;s instructions, 1 &#215; buffer (Invitrogen, California, USA), 0.2 mM dNTPs (Thermo Scientific, Massachusetts, USA), 0.05 U/&#956;l Platinum&#8482; Taq DNA Polymerase (Invitrogen, , California, USA), 2.5 mM MgCl 2 , 2% formamide (Sigma-Aldrich, Steinheim, Germany), 0.35 pmoles/&#956;l labeled forward primer (either FAM6, VIC, NED, or PET, Applied Biosystems, California, USA), and 0.5 pmoles/&#956;l reverse primer <ns0:ref type='bibr'>(Macrogen, Seoul, Korea)</ns0:ref>. The PCRs were performed on a T100 thermocycler (BioRad, California, USA) with an initial denaturation step of 95 &#176;C for 3 min followed by 32 cycles consisting of a denaturation step of 90 &#176;C for 22 s and an annealing step for 18 s using the annealing temperatures described for each primer in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>. The extension step and a final elongation were absent in this thermal profile. Finally, the PCRs were submitted to electrophoresis on an automated sequencer ABI 3730 XL (Applied Biosystems, California, USA) using GeneScan 500 LIZ dye size standard (Applied Biosystems, California, USA) as the internal molecular size. Allelic fragments were denoted according to their molecular size and scored using GeneMapper v4.0 software (Applied Biosystems, California, USA; Table <ns0:ref type='table'>S2</ns0:ref>). Before the statistical analysis, <ns0:ref type='bibr'>Micro-Checker v.2.2.3 (van Oosterhout et al., 2004</ns0:ref>) was run to detect potential genotyping errors.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Tests for Hardy-Weinberg and Linkage equilibria, observed (H O ) and expected (H E ) heterozygosities and inbreeding coefficient (F IS ) were estimated using Arlequin v3.5.2.2 software <ns0:ref type='bibr'>(Excoffier, Laval, &amp; Schneider, 2005)</ns0:ref>. The sequential Bonferroni correction was applied to adjust the statistical significance in multiple comparisons <ns0:ref type='bibr' target='#b10'>(Holm, 1979;</ns0:ref><ns0:ref type='bibr'>Rice, 1989)</ns0:ref>. The average number of alleles per locus and the PIC <ns0:ref type='bibr'>(Botstein et al., 1980)</ns0:ref> for each microsatellite locus were calculated with GenAlEx v6.503 software <ns0:ref type='bibr'>(Peakall &amp; Smouse, 2006)</ns0:ref> and Cervus v3.0.7 software <ns0:ref type='bibr'>(Marshall et al., 1998)</ns0:ref>, respectively.</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The genetic differentiation among geographical samples was calculated using the standardized statistics F&#180;S T <ns0:ref type='bibr'>(Wright, 1943</ns0:ref><ns0:ref type='bibr'>(Wright, , 1965;;</ns0:ref><ns0:ref type='bibr'>Meirmans, 2006)</ns0:ref>, Jost&#180;s D&#180;e st <ns0:ref type='bibr'>(Jost, 2008;</ns0:ref><ns0:ref type='bibr'>Meirmans &amp; Hedrick, 2011)</ns0:ref> and analysis of molecular variance (AMOVA) <ns0:ref type='bibr'>(Meirmans, 2006)</ns0:ref> with 10,000 permutations and bootstraps included in GenAlEx v6.503 software <ns0:ref type='bibr'>(Peakall &amp; Smouse, 2006)</ns0:ref>. Furthermore, the diploid genotypes of 11 loci (22 variables) in 725 individuals were submitted to discriminant analysis of principal components (DAPC) using the R-package Adegenet <ns0:ref type='bibr'>(Jombart, 2008)</ns0:ref>.</ns0:p><ns0:p>To examine other groupings of the samples, genetic differentiation among samples was tested using the Bayesian analysis of population partitioning with Structure v2.3.4 software <ns0:ref type='bibr'>(Pritchard, Stephens &amp; Donnelly, 2000)</ns0:ref>. Parameters included 350,000 Monte Carlo Markov Chain steps and 50,000 iterations as burn-in, the admixture model, correlated frequencies, and the LOCPRIOR option for detecting relatively weak population structure <ns0:ref type='bibr' target='#b11'>(Hubisz et al., 2009)</ns0:ref>. Each analysis was repeated 20 times for each simulated K value, which ranged from 1 to n + 3 (n, number of populations compared). For a best estimation of genetic stocks (K), STRUCTURESELECTOR web-based software <ns0:ref type='bibr'>(Li &amp; Liu, 2018)</ns0:ref> was used to calculate the &#916;K ad hoc statistic <ns0:ref type='bibr'>(Evanno, Regnaut, &amp; Goudet, 2005)</ns0:ref>, the estimators MEDMEANK, <ns0:ref type='bibr'>MAXMEANK, MEDMEDK, and MAXMEDK (Puechmaille, 2016)</ns0:ref>, and to generate the graphical representation of results using the integrated Clumpak software <ns0:ref type='bibr'>(Kopelman et al., 2015)</ns0:ref>. Based on the coancestry coefficients provided by Structure and Clumpp, the individuals were reorganized by genetic stock in sample sites that showed multiple stocks and were later submitted to the genetic analyses described above.</ns0:p><ns0:p>Additionally, the occurrence of recent genetic bottlenecks of populations was evaluated by calculating the levels of heterozygosity and the M ratio using Bottleneck v1.2.02 software <ns0:ref type='bibr'>(Piry, Luikart, &amp;</ns0:ref><ns0:ref type='bibr'>Cornuet, 1999) and</ns0:ref><ns0:ref type='bibr'>Arlequin v3.5.2.2 (Excoffier, Laval, &amp;</ns0:ref><ns0:ref type='bibr'>Schneider, 2005)</ns0:ref>, respectively. Excess heterozygosity was assessed by employing the Wilcoxon sign-rank test <ns0:ref type='bibr'>(Luikart &amp; Cornuet, 1998)</ns0:ref>. The M ratio -the mean ratio of the number of alleles compared to the range of allele size -indicates that the population has PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed experienced a recent and severe reduction in population size when its values are smaller than 0.680 <ns0:ref type='bibr'>(Garza &amp; Williamson, 2001)</ns0:ref>.</ns0:p><ns0:p>To explore non-neutral evolutionary forces acting on the microsatellite loci, a scanning analysis was performed using the BayeScan v2.1 software <ns0:ref type='bibr'>(Foll &amp; Gaggiotti, 2008)</ns0:ref> to detect candidate loci under selection. Parameters for BayeScan analyses included 10:1 prior odds for the neutral model and 20 pilot runs consisting of 5,000 iterations each followed by 250,000 iterations with a burn-in length of 50,000 iterations <ns0:ref type='bibr'>(Foll &amp; Gaggiotti, 2008)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Phylogenetic relationships among genetic groups</ns0:head><ns0:p>To explore the phylogenetic relationships among individuals sampled along the basin, partial fragments of the mitochondrial cox1 gene (~650 bp) were amplified in a subset of samples using primers and PCR conditions previously described by <ns0:ref type='bibr'>Ward et al. (2005) and</ns0:ref><ns0:ref type='bibr'>Ivanova et al. (2007)</ns0:ref>. PCR products were sequenced by the Sanger method using an automated sequencer, ABI 3730 XL (Applied Biosystems). The best-fit evolutionary model was determined based on the Bayesian information criterion as implemented in the jModelTest v2.1.7 software <ns0:ref type='bibr'>(Posada &amp; Crandall, 1998)</ns0:ref>. Phylogenetic relationships were determined by Bayesian inference using the MrBayes v3.2.6 software <ns0:ref type='bibr'>(Ronquist &amp; Huelsenbeck, 2003)</ns0:ref>. For this purpose, we performed two independent runs of 20 million generations sampled each 1,000 generations using 25% as burn-in. The remaining values were left as default. The convergence of each parameter was checked based on a potential scale reduction factor nearing 1, an average standard deviation of the split frequencies lower than 0.010, and the visualization of the resulting trees was performed with FigTree v1.4.3 software (Rambaut, 2012). Finally, the pair-wise divergences of P. magdalenae and P. reticulatus haplotype sequences were estimated using the Kimura 2-parameters model in MEGA v10.1.8 software <ns0:ref type='bibr'>(Tamura et al., 2013)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head></ns0:div> <ns0:div><ns0:head>Microsatellite loci development</ns0:head><ns0:p>Genomic sequencing of the Illumina shotgun library of P. magdalenae (0.115 GB) generated 277,133 reads and 14,124 of 50,404 that contained microsatellite loci, were flanked by suitable PCR priming sites.</ns0:p><ns0:p>The dinucleotides (47.758%) were the most abundant repeat motifs, followed by tetranucleotide (28.353%), trinucleotide (16.193%), pentanucleotide (5.146%), and hexanucleotide (2.549%) repeats. The most common motifs found were AC (29.661%), TC (18.905%), ATT (4.811%), and AAAT (4.794%).</ns0:p><ns0:p>The sequences of contigs containing the microsatellite loci obtained in the present study are provided in supplementary files S3 and S4.</ns0:p><ns0:p>A total of 21 of the 52 microsatellite loci evaluated were polymorphic and showed Hardy-Weinberg disequilibrium (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>) and Linkage equilibrium (Table <ns0:ref type='table'>S5</ns0:ref>). The number of alleles per locus ranged from 11 to 37, with an average number of 20.619 alleles/locus, the average values of observed and expected heterozygosities were Ho = 0.589 and He = 0.876 and the PIC values ranged from 0.399 to 0.949 (average 0.867) (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>). A total of 10 loci failed to satisfy the selection criteria, showing low values of PIC (Pma32), dropout and stuttering (Pma32, Pma08), inconsistent amplifications (Pma17, Pma47, Pma57), or low-definition peaks (Pma42, Pma56, Pma26, Pma50). Consequently, only 11 (Pma39, Pma25, Pma02, Pma35, Pma01, Pma40, Pma46, Pma36, Pma18, Pma13, and Pma14) satisfied most of the parameters required to validate the new microsatellites primers described previously.</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic diversity, population demography, and outlier loci screening</ns0:head><ns0:p>Comparisons among rivers revealed that 8 of 11 loci satisfied the Hardy-Weinberg equilibrium expectations in at least one case (Table <ns0:ref type='table'>2</ns0:ref>). However, the analysis across loci showed significant departures from Hardy-Weinberg equilibrium expectations in all rivers evaluated (Table <ns0:ref type='table'>2</ns0:ref>). The average PeerJ reviewing <ns0:ref type='table' target='#tab_1'>PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:ref> Manuscript to be reviewed number of alleles per locus was higher in Cauca (22.455) and Magdalena (19.455), followed by Nare (15.636), Sin&#250; (15.273), the fish hatchery (14.818), and Atrato (14.636) and was lowest in San Jorge (13.545) and Cesar (13.364). Additionally, the highest values of observed and expected heterozygosities were found in San Jorge (Ho: 0.809; He: 0.884) and Cesar (Ho: 0.782; He: 0.873), followed by Sin&#250; (Ho: 0.767; He: 0.882), Magdalena (Ho: 0.758; He: 0.896), and Cauca (Ho: 0.725; He: 0.898) and were lowest in Atrato (Ho: 0.718; He: 0.879), the fish hatchery (Ho: 0.691; He: 0.880), and Nare (Ho: 0.659; He: 0.876) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>Furthermore, comparisons among sites within each river showed similar high levels of genetic diversity (Table <ns0:ref type='table'>3</ns0:ref>). The highest value of genetic diversity was found in the floodplain lake Palagua in the Magdalena River (Na: 17.182 alleles/locus; He: 0.895; Ho: 0.792), whereas the lowest was observed in Bet&#233;, a site of the Atrato River (Na: 9.273 alleles/locus; He: 0.791; Ho: 0.711). In addition, all sites exhibited a highly significant deficit of heterozygosity (Table <ns0:ref type='table'>3</ns0:ref>) with Mata de Palma and Saman&#225; Norte River showing the lowest and highest heterozygosity deficits, respectively. Inbreeding coefficients (F IS ) per site in main rivers of the different Colombian hydrographic areas were significant and ranged from 0.120 to 0.255 (Table <ns0:ref type='table'>3</ns0:ref>). Although decreased in magnitude, the inbreeding coefficients (Table <ns0:ref type='table'>3</ns0:ref>) remained significant even after comparing the genetic diversity according to genetic stocks in Chucur&#237;, Puerto Berr&#237;o, and Palagua and among the Magdalena River and tributaries.</ns0:p><ns0:p>Results of the genetic bottleneck (Table <ns0:ref type='table'>4</ns0:ref>) were significant for all populations under the infinite alleles model (IAM) and for most populations under the two-phase model (TPM), whereas they were nonsignificant under the stepwise mutation model (SMM). As it is thought that few loci follow the strict SMM <ns0:ref type='bibr'>(Piry, Luikart, &amp; Cornuet, 1999)</ns0:ref>, the best estimation of expected heterozygosity at mutation-drift equilibrium is expected under a combination of IAM and TPM. Additionally, all values of the M ratio were substantially smaller than 0.680, indicating that all populations have experienced recent and severe reductions in population size (Table <ns0:ref type='table'>4</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In contrast to other samples that did not show evidence of selection, BayeScan analysis revealed that 8 of 11 loci (Pma39, Pma25, Pma02, Pma35, Pma40, Pma36, Pma13, and Pma14) exhibit substantial evidence of selection in the Magdalena River (Table <ns0:ref type='table'>5</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic structure and phylogenetic relationships among the samples studied</ns0:head><ns0:p>At regional scale, the Bayesian analysis showed the presence of two genetic stocks (&#916;K = 2; MEDMEDK = 2; MEDMEANK = 2), one in the Magdalena River (Chucur&#237; + Puerto Berrio + Palagua) and the other one in the remaining evaluated rivers (Fig. <ns0:ref type='figure' target='#fig_0'>2A</ns0:ref>), which is concordant with DAPC (Fig. <ns0:ref type='figure' target='#fig_0'>2B</ns0:ref>) and AMOVA (F ST(7, 1407) = 0.009; P = 0.000). Together with Chucur&#237; + Puerto Berrio + Palagua, a predominant genetic stock with different levels of genetic admixture in Sin&#250; and Atrato rivers was revealed in the clusters suggested by the MAXMEAN and MAXMED statistics (K = 3). The additional clustering patterns (K = 4 -8; Fig. <ns0:ref type='figure' target='#fig_0'>2A</ns0:ref>) examined to compared them with other approaches, showed genetic admixture with other stocks absent in the examined rivers. Likewise, pairwise comparisons of the standardized statistics F&#180;S T <ns0:ref type='bibr'>(Meirmans, 2006)</ns0:ref> and Jost&#180;s D&#180;e st <ns0:ref type='bibr'>(Meirmans, &amp; Hedrick, 2011)</ns0:ref> showed genetic differences among Atrato, the fish hatchery, Sin&#250;, and the remaining rivers (Table <ns0:ref type='table'>6</ns0:ref>) as well as among the Magdalena River and its tributaries, Cauca and Nare.</ns0:p><ns0:p>However, excluding samples that exhibit loci putatively under selection (Chucur&#237; + Puerto Berrio + Palagua), comparisons among sites within each river revealed a genetic admixture of two stocks (&#916;K = 2; MEDMEDK = 2; MEDMEANK = 2) homogenously distributed in Magdalena River and its tributaries (Figs. <ns0:ref type='figure' target='#fig_1'>3A and 3B</ns0:ref>; Tables <ns0:ref type='table' target='#tab_1'>6, 7</ns0:ref>). Additionally, this analysis revealed a genetic substructure in Sin&#250; (&#916;K = 2; MEDMEDK = 2; MEDMEANK = 2; Figs. <ns0:ref type='figure' target='#fig_1'>3C and 3D</ns0:ref>; F ST(1, 67) = 0.033; P = 0.000; F&#180;S T = 0.027; P = 0.004; D&#180;est = 0.149; P = 0.005). In Atrato River, the Bayesian analysis showed a single genetic stock (&#916;K = 2; MEDMEDK = 1; MEDMEANK = 1; Fig. <ns0:ref type='figure' target='#fig_1'>3E</ns0:ref>) although remaining analysis showed genetic differentiation among sites (Fig. <ns0:ref type='figure' target='#fig_1'>3F</ns0:ref>; F ST(1, 57) = 0.045; P = 0.000; F&#180;S T = 0.047; P = 0.000; D&#180;est = 0.330; P = 0.000).</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Finally, the Bayesian tree using the cox1 gene clustered our samples (GenBank accession numbers MK330430 to MK330494; Supplementary file S6) with sequences of P. magdalenae and P. reticulatus deposited in public databases and in a different group, Prochilodus marie and Prochilodus nigricans (Fig. <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>). Moreover, Kimura-2-parameters genetic distances (Table <ns0:ref type='table' target='#tab_1'>S7</ns0:ref>) were larger among haplotypes of P. magdalenae (0.002 -0.014) than among P. magdalenae and P. reticulatus haplotypes (0.002 -0.010).</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head></ns0:div> <ns0:div><ns0:head>Microsatellite loci development</ns0:head><ns0:p>This work developed species-specific microsatellite loci using next-generation sequencing and bioinformatic analysis. Although a total of 21 of 52 microsatellite loci with tri-and tetra-nucleotide motifs were polymorphic in P. magdalenae, the consistency in the amplification in a larger sample, allelic size class distribution, and high definition peaks allowed the selection of only 11 microsatellite loci for further population genetic analysis. Most of the loci showed departures from Hardy-Weinberg equilibrium and significant heterozygosity deficit, which may be related to the significant levels of inbreeding as well as the genetic structure of the samples by the coexistence of two genetic stocks (see below).</ns0:p><ns0:p>Although the levels of genetic diversity measured by the expected heterozygosities were similar, the levels of observed heterozygosity as well as the average number of alleles per locus found in this study were substantially greater than those found by Orozco-Berdugo &amp; Narv&#225;ez-Barandica (2014). These results support the idea that the heterologous microsatellite loci used by these authors may be limited by the presence of null alleles or genotyping errors related to their dinucleotide motifs because a higher PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed variability is commonly expected in shorter motifs (e.g. 2mers, Orozco-Berdugo &amp; Narv&#225;ez-Barandica, 2014) compared with longer motifs (3mers and 4mers, this study). However, despite these differences, both heterologous and species-specific microsatellite loci revealed a general deficit of heterozygotes in all samples, indicating that its causes are biological rather than technical. In this context, the species-specific microsatellite loci developed in this study seem to provide a good approach to studying the population genetics of P. magdalenae considering that the levels of heterozygosity constitute a parameter used to estimate the genetic diversity of the populations.</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic diversity and population demography</ns0:head><ns0:p>Microsatellite data revealed average values of genetic diversity (He: 0.737) among the highest values found in other Prochilodontidae species, only surpassed by those reported for P. costatus <ns0:ref type='bibr'>(Melo et al., 2013)</ns0:ref> and P. argenteus (Coimbra et al., 2017) (0.747 and 0.753 respectively). Similarly, the average levels of expected heterozygosity were higher than that found in P. magdalenae measured by heterologous microsatellites (He: 0.877; Orozco-Berdugo &amp; Narv&#225;ez-Barandica, 2014) and Neotropical Characiforms (He: 0.675 &#177; 0.160; see review by <ns0:ref type='bibr' target='#b9'>Hilsdorf &amp; Hallerman, 2017)</ns0:ref>.</ns0:p><ns0:p>Additionally, this study found levels of observed heterozygosity higher than those found by Orozco-Berdugo &amp; Narv&#225;ez-Barandica (2014). However, the use of species-specific microsatellite loci developed in this study revealed similar values of expected heterozygosity among samples analyzed by Orozco-Berdugo &amp; Narv&#225;ez-Barandica (2014) and the remaining samples analyzed, indicating that differences between the two studies are related to the type of microsatellite loci utilized (heterologous vs. speciesspecific microsatellite loci).</ns0:p><ns0:p>The significant deficit of heterozygosity in all studied samples corroborates the previous findings for P. magdalenae from Magdalena River (Orozco-Berdugo &amp; Narv&#225;ez-Barandica, 2014); however, the magnitude of the heterozygosity deficit as well as the inbreeding coefficient were substantially lower (0.075 -0.239) than those previously reported (0.624 -0.788). Following <ns0:ref type='bibr'>Franklin (1980)</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>and Soul&#233;</ns0:head><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed (1980), the values above 10% of the inbreeding coefficient indicate that these populations require careful management to avoid future detrimental effects on its populations. This point is important since it has been recommended recently that any inbreeding coefficient higher than zero will usually have an adverse fitness effect <ns0:ref type='bibr'>(Frankham, Bradshaw, &amp; Brook, 2014)</ns0:ref>.</ns0:p><ns0:p>Another non-excluding alternative is plausible considering that the significant deficit of heterozygosity observed in all sites analyzed may be also explained by the coexistence of genetic stocks (Wahlund effect) as this was evidenced by the genetic structure analysis (see below). Another biological cause of heterozygosity deficit, assortative mating, does not seem to explain the results found in this study because P. magdalenae is iteroparous and characterized by total spawning (Jaramillo-Villa &amp; Jim&#233;nez-Segura, 2008) as described in its congeners, P. costatus <ns0:ref type='bibr'>(Carolsfield et al., 2004) and</ns0:ref><ns0:ref type='bibr'>P. lineatus (Roux et al., 2015)</ns0:ref>.</ns0:p><ns0:p>On the other hand, this study also provided evidence for a population bottleneck, suggesting that P. magdalenae shows signs of erosion of the genetic pool, likely by the constant pressure from fishing and other anthropogenic activities exerted on its populations. Although paradoxical to the heterozygosity deficit found in all populations evaluated, this outcome is plausible considering that the Bottleneck algorithm tests not for an excess of heterozygotes (Ho &gt; He) but rather for an excess of heterozygosity (He &gt; He at mutation-drift equilibrium) <ns0:ref type='bibr'>(Piry, Luikart, &amp; Cornuet, 1999)</ns0:ref>. Besides, the combination of a population bottleneck and a heterozygosity deficit may result from population growth in a closed system, population genetic structure, or admixture <ns0:ref type='bibr'>(Barson, Cable, &amp; Oosterhout, 2009)</ns0:ref>. Considering the lengths of the rivers studied, population growth in a closed system is unlikely but the last two alternatives may explain our results due to the coexistence of genetic stocks in the samples studied and the continuous restocking of natural stocks using juveniles from fish hatcheries, which may create an apparent excess of novel alleles and an incomplete allele frequency distribution. Similar results have also been found in guppies, Poecilia reticulata, in Trinidad and Tobago <ns0:ref type='bibr'>(Barson, Cable, &amp; Oosterhout, 2009)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div> <ns0:div><ns0:head>Genetic structure</ns0:head><ns0:p>This study tested the hypothesis that P. magdalenae exhibits genetic stocks that coexist and co-migrate along sections of the main channel and some tributaries of the Magdalena River (Cauca, San Jorge, and Cesar), Sin&#250;, and Atrato. Before testing this hypothesis, we compared the genetic structure at regional scale, finding two spatially structured populations: one in the Magdalena River (Puerto Berr&#237;o and the floodplains Chucur&#237; and Palagua) and the other in the remaining rivers evaluated.</ns0:p><ns0:p>The geographical genetic structure may result from taxonomic differences among stocks due to the lack of regulations on the restocking of natural stocks of P. magdalenae. The phylogenetic analysis using partial sequences of cox1 gene indicates that samples do not correspond to species such as P. marie or P. nigricans because this genetic stock is clustered with previously published sequences of P. magdalenae <ns0:ref type='bibr' target='#b0'>(Aguirre-Pab&#243;n, Narv&#225;ez-Barandica, &amp; Castro-Garc&#237;a, 2013)</ns0:ref>. However, it remains to be seen whether they represent artificial mixtures of P. magdalenae and P. reticulatus because the current mitochondrial phylogenetic analysis of Prochilodontidae does not allow the two species to be discriminated <ns0:ref type='bibr'>(Melo et al., 2016b</ns0:ref><ns0:ref type='bibr'>(Melo et al., , 2018))</ns0:ref>. Moreover, the morphological and molecular similitudes have led to the proposal that P. magdalenae and P. reticulatus represent only one species with probable allopatric differentiation resulting from the uplift of the Sierra del Perij&#225; <ns0:ref type='bibr'>(Melo et al., 2016b)</ns0:ref>. Thus, a separated clustering of mitochondrial sequences of those stocks is not expected in the phylogenetic analysis even though they represent allopatric populations. An alternative explanation is that the genetic differences result from eight outlier loci that are putatively under selection in three sites of the Magdalena River, suggesting that P. magdalenae experiences natural/artificial selection or local adaptation, although testing of these hypotheses is out of the scope of the present study. The explanation that outlier loci represent false positives resulting from the inclusion of severely bottlenecked populations <ns0:ref type='bibr'>(Teshima, Coop, &amp; Przeworski, 2006;</ns0:ref><ns0:ref type='bibr'>Foll &amp; Gaggiotti, 2008)</ns0:ref> seems unlikely because the significant excess of heterozygosity and small values of the M ratio were found even PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed in populations that do not exhibit outlier loci. Thus, considering that those sites have been exposed to restocking since 20 years ago and since most microsatellite loci are not transcriptionally active, the outlier loci found in this study may reflect hitchhiking selection resulting from restocking using juveniles selected artificially by fish hatcheries. Alternatively, the outlier loci may result from asymmetric gene flow by unidirectional migration from hatchery stocks to wild populations. Similar results were found in Denmark in populations of three brown trout, which have been significantly admixtured with stocked hatchery trout <ns0:ref type='bibr' target='#b4'>(Hansen, Meier, &amp; Mensberg, 2010)</ns0:ref>.</ns0:p><ns0:p>Although the above reasoning might explain the genetic differences between stocks, an additional justification is required to explain the restricted distribution of one genetic stock in only three sites of the Magdalena River considering the migratory abilities of these species/allopatric populations. Thus, this genetic structure seems to result from recent restocking before reproductive/feeding migrations, use of artificial barriers to avoid migration of the fish, clogging by sedimentation or vegetation, or the desiccation of access to floodplain lakes or may be a product of the intensive anthropic intervention in these territories characterized by the exploitation of hydrocarbons and livestock. This idea is concordant with the fact that degradation of preferred habitat and barriers that impede dispersal contribute to the degree of genetic differentiation among populations <ns0:ref type='bibr'>(Faulks, Gilligan, &amp; Beheregaray, 2011)</ns0:ref>. Furthermore, the results found here provide support for the hypothesis that P. magdalenae exhibits genetic stocks that coexist and co-migrate along sections of the rivers Magdalena, Cauca, Cesar (tributaries of the Magdalena River), Sin&#250;, and Atrato. Since similar patterns of genetic structure are found in P. reticulatus <ns0:ref type='bibr'>(L&#243;pez-Mac&#237;as et al., 2009)</ns0:ref>, P. marggravii <ns0:ref type='bibr' target='#b6'>(Hatanaka &amp; Galetti Jr., 2003)</ns0:ref>, P. argenteus <ns0:ref type='bibr'>(Sanches et al., 2012)</ns0:ref>, P. costatus <ns0:ref type='bibr' target='#b1'>(Barroca et al., 2012a)</ns0:ref>, P. magdalenae <ns0:ref type='bibr'>(Orozco Berdugo &amp; Narv&#225;ez Barandica, 2014;</ns0:ref><ns0:ref type='bibr' target='#b8'>Hern&#225;ndez, Navarro, &amp; Mu&#241;oz, 2017)</ns0:ref>, and Ichthyoelephas longirostris <ns0:ref type='bibr'>(Land&#237;nez-Garc&#237;a &amp; M&#225;rquez, 2016)</ns0:ref>, this outcome supports the idea that this genetic structure is a generalized tendency within the family Prochilodontidae.</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Excluding the genetic stock of Puerto Berr&#237;o and the floodplains Chucur&#237; and Palagua, each river showed the coexistence of at least two genetic stocks. Homogeneous and non-homogeneous distributions of these genetic stocks along the rivers explain similarities (Cauca, Magdalena, San Jorge, Cesar, and Nare) as well as geographical differences among the rivers analyzed (within Magdalena, including Puerto Berr&#237;o and the floodplains Chucur&#237; and Palagua, Sin&#250;, and Atrato). This genetic structure also explains the significant heterozygosity deficit observed in all sites analyzed (Wahlund effect) as discussed above.</ns0:p><ns0:p>Similar evidence of the Wahlund effect has been documented in the congener P. costatus, which exhibited genetic differences resulting from temporal isolation <ns0:ref type='bibr'>(Braga-Silva &amp; Galetti Jr., 2016)</ns0:ref>.</ns0:p><ns0:p>Although sampling in this study was not designed to detect temporal genetic structuring, genetic similarities among samples collected in different years suggest that the Wahlund effect must be more spatial than temporal. It remains to be seen whether this behavior is natural or artificial, considering that the restocking activities have been widely implemented along different Colombian rivers.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>This study provides evidence that P. magdalenae exhibits high genetic diversity, significant inbreeding levels between 0.162 to 0.202 per genetic stock, and signs of erosion of the genetic pool and conforms a mixture of genetic stocks heterogeneously distributed along the rivers studied. Additionally, this study developed a set of 11 microsatellite loci that allows the detection of reliable levels of genetic diversity, providing a tool for monitoring changes in the genetic diversity of the species, brood stocks, and juveniles used for supportive breeding and for measuring the efficacy of current population restocking activities.</ns0:p><ns0:p>Management and conservation strategies need to be implemented at the level of the basins Magdalena-Cauca, Sin&#250;, and Atrato concordantly with their genetic population structure.</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed S6a: Floodplain Lake Grande, S6b: Floodplain Lake Caimanera F, S6c: Guaranda. Values in bold denote statistical significance after Bonferroni correction (Cauca: P &lt; 0.0005; Magdalena: P &lt; 0.0001).</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Primer sequences, characteristics, polymorphism levels, and genetic diversity of 21 species-specific microsatellite loci in 88 individuals of Prochilodus magdalenae randomly chosen from the whole sample.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Manuscript to be reviewed</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>Ta: annealing temperature standardized in PCRs, Na: number alleles per locus; Ra: allelic size</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>range (bp); PIC: polymorphism information content; H O and H E : observed and expected</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>heterozygosities, respectively; P: statistical significance (values in bold represent significance</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>at P &lt; 0.05).</ns0:cell><ns0:cell>a</ns0:cell><ns0:cell cols='3'>Satisfied selection criteria,</ns0:cell><ns0:cell>b</ns0:cell><ns0:cell>inconsistent amplifications,</ns0:cell><ns0:cell>c</ns0:cell><ns0:cell>low definition peaks,</ns0:cell></ns0:row><ns0:row><ns0:cell>d</ns0:cell><ns0:cell>dropout,</ns0:cell><ns0:cell>e</ns0:cell><ns0:cell cols='3'>stuttering,</ns0:cell><ns0:cell>f</ns0:cell><ns0:cell>low value of PIC.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020) PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 7 (on next page)</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Pairwise Jost's D est (upper diagonal) and F' ST (below diagonal) of Prochilodus magdalenae samples among sites of the rivers Cauca and Magdalena.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2018:12:33631:1:1:NEW 7 Jul 2020)</ns0:note> </ns0:body> "

"Medellín, June 26th, 2020 Editor PeerJ Dear Editor, We appreciate the valuable comments and suggestions provided by the reviewers and editor, which have led to improve our paper. To address such comments and suggestions, we have edited the manuscript and the detailed answers are provided below. We hope that the manuscript is now suitable for publication in PeerJ. Sincerely, Edna J. Márquez On behalf of the all authors Editor I think the manuscript indeed suffers from serious flaws. Apart from the objections raised by the Reviewers I find the HW equilibrium analyses (e.g., l.227: how can allelic frequencies be concordant with expectations?) particularly disturbing. Done. The redaction of the sentences regarding this topic was edited. Also, the interpretation of heterozygote deficiency (l.297-305 ssq) needs deep reconsideration (inbreeding vs. silent alleles and inconsistency with the explanation for the results from other authors). Those aspects are discussed in the original version of the manuscript, section “Microsatellite loci development” (Lines 269 - 288). I do not understand the sentence containing: showing a single allele size class in more than 50% of alleles in the studied sample (l.220-221) which, in connection with the HW note above suggest the authors confuse allele and genotype. Now, that phrase was changed by “low values of PIC”. (High) PIC and and high polymorphism are redundant (l.136) Done. Now, we deleted “high polymorphism”. I do not understand the significance of the numbers at the first line of S2 Excel sheet. Now, this information was included in the legend of the Supplementary Table S2: “First row indicates respectively: Number of loci, Number of individuals, Number of sampling sites, Sample size for eight sites. Third row indicates respectively: River, Sampling site, Sampling site ID, Sample ID, Locus name for 11 loci”. Reviewer: Gabriel Yazbeck Basic reporting The study presents population genetics analyses on the migratory <Prochilodus magdalenae> fish species, endemic and highly conspicuous to the Magdalena-Cauca and Caribe Basins in Colombia, while presenting its first set of 11 specific microsatellite DNA markers. The study is notable for its pioneer quality and for a rather fair sample size, over 700 specimens. However, the manuscript emphasizes the use of Next-Generation Sequencing (NGS), whilst not disclosing any NGS data or even reporting on basic NGS results. These would not only be potentially helpful in a lot of predictable and unanticipated ways by the research community, but would also be consonant with the open data ethos. The text needs copy editing and verification. Specific spotted examples are highlighted directly in the annotated manuscript, along with many detailed punctual comments (signed in the PDF as 'Vera'). Now: “Genomic sequencing of the Illumina shotgun library of P. magdalenae (0.115 GB) generated 277,133 reads and 14,124 of 50,404 that contained microsatellite loci, were flanked by suitable PCR priming sites. The dinucleotides (47.758%) were the most abundant repeat motifs, followed by tetranucleotide (28.353%), trinucleotide (16.193%), pentanucleotide (5.146%), and hexanucleotide (2.549%) repeats. The most common motifs found were AC (29.661%), TC (18.905%), ATT (4.811%), and AAAT (4.794%). The sequences of contigs containing the microsatellite loci obtained in the present study are provided in supplementary files S3 and S4.” Experimental design The methods chosen for the development of new species-specific microsatellite markers were spot on, as there is little justification for using more traditional methods aiming this goal, currently. The population genetic analyses are also sound, pertinent and grounded on the precedent literature. Worth praising is the author's use of more current/update estimators for the genetic structure of populations, such as F'st, Jost's D and the application of Bayesian approaches. However, Table 1 is presented in the Methods section, while reporting on new results from this original research. It should be moved to the Results section and its findings about expected, observed heterozygosity, exact p and Fis values should be reconsidered, since it has been explained it was composed of a random sample of 88 fish. Given these fish are not from the same sampling effort or location, there is little utility for these results (unless the implications of its nature is stated, accompanied of its respective meaning in discussion, since there will be a tendency towards a deficit of observed heterozygosity and departure of HWE, due to Wahlund effect, as observed in these results). Maybe it could be used to reflect and support the idea the the species does not consist in a single panmitic deme throughout the studied geographic range. Done. Now this table was included in “Results”. However, we maintain the current information since it permits to discuss information requested by the editor (technical vs. biological causes of the heterozygosity deficit). Another issue worth raising awareness is the discussion of the inherent flaw in a specific feature of the Micro-Checker (MC) software, used here, presented in Waples, R.S., Journal of Heredity, 2015:106(1):1–19, doi:10.1093/jhered/esu062. The authors should reflect and consider if this could be of compromise to their intended goals ('detect potential genotyping errors') in the present study. I can't seem to locate the reporting of these results below. If it is the case, one could omit the use of this specific software, unless it supports/contradicts directly or indirectly the study findings. If appropriate, the authors could state that the MC analyses simply did not show evidence of genotyping errors. Done. MC was used uniquely for detecting genotyping errors such as null alleles, stutter bands or dropout effect. Biological causes of heterozygote deficit were discussed by using other reasoning. Validity of the findings The paper presents a thorough description of population genetics analyses for this species, the first using species-specific microsatellite markers and will be of great value as a departing port for guiding further studies, proposing testable hypothesis about this species biology and ultimately helping in policy guidance, environmental mitigation and management decisions, in face of the coming wave of population genomic applications, fostered by NGS. This could be made explicit in the discussion (i.e. the results inherent limitations on number of markers sampled in the genome), in the context that the paper forwards 11 useful loci (~11/27 chromosomes (?), or an approximate average of less than one marker per each two different chromossomes in this species complement) and that the current study conclusions or further inquiries would greatly benefit from further future NGS based genotyping such as RAD/GBS and similar approaches, now at hand. A very positive point is that there was a strong selection criteria for loci to be applied in the genetic structure diversity analyses. My main points to be discussed are: 1) No NGS results used for generating the data sifted to isolate microsatellites were reported (e.g. assembly results and statistics, depth of sequencing, number of contigs obtained/screened, number of microsatellites found, average PHRED quality, etc). No NGS data (e.g. FASTQ files, assemblies, etc.) was made public through adequate data repositories (e.g. NCBI's SRA). Ideally, each contig used for a validated microsatellite should be publicly revealed. Its average depth of coverage would be also a useful information. Still more preferable, the availability of alignments of short reads onto contigs for each locus should be made available or reported here. If not possible, the authors should disclose/discuss why. Now, we include basic NGS results: “Genomic sequencing of the Illumina shotgun library of P. magdalenae (0.115 GB) generated 277,133 reads and 14,124 of 50,404 that contained microsatellite loci, were flanked by suitable PCR priming sites. The dinucleotides (47.758%) were the most abundant repeat motifs, followed by tetranucleotide (28.353%), trinucleotide (16.193%), pentanucleotide (5.146%), and hexanucleotide (2.549%) repeats. The most common motifs found were AC (29.661%), TC (18.905%), ATT (4.811%), and AAAT (4.794%). The sequences of contigs containing the microsatellite loci obtained in the present study are provided in supplementary files S3 and S4.” Additionally, we provided as supplementary data, the sequences of contigs that contain the microsatellite loci evaluated in the present study. Remaining sequences are submitting to assemblage of mitochondrial genome and partial assemblage of some nuclear genes that will contribute to reconstruct the genome of this species. 2) What was the test significance (alpha value) after the Sequential Bonferroni procedure? Please, confirm if this was indeed a Holm-Bonferroni type multiple test correction or just a simple Bonferroni correction (less appropriate than the former) and clarify if it was used with all the HWE values from the whole sample, with or without the tests carried in the preliminary testing step, with the 88 fish (i.e. is there a single value for the study, or does the study show different values for Tables 1 and 2)? We made Sequential Bonferroni correction in each analysis as follows: the p values were ascendingly ordered (from smallest p-value to largest p-value). The test with the lowest probability was tested first with a Bonferroni correction involving all tests. The second test was tested with a Bonferroni correction involving one less test and so on for the remaining tests. Consequently, there are different values of “alpha per test (test wise)” for each analysis. To clarify this point, we include Holm, 1979 in the citation. Now: “The sequential Bonferroni correction was applied to adjust the statistical significance in multiple comparisons (Holm, 1979; Rice, 1989).” 3) Are the Fis values found actually accounted by breeding system endogamy? What would be the evidence supporting it, against the other concurrent hypothesis of sub-structuring and Wahlund effect? Even having the authors put it as non-excludent alternatives, in the light of the study's conclusion, I would tend to interpret these numbers as an effect of mixture of fish from different true stocks, so I advise caution with this conclusion. In the Wahlund effect, non-significant heterozygosity deficit is expected within each genetic stock that coexist in a site. Thus, we separated the genetic groups that circulate in the Magdalena and Cauca rivers based on its coancestry coefficients across 20 runs calculated by STRUCTURE software and summarized by Clummp software. Then, we calculated the genetic diversity and inbreeding coefficients in each stock (Please, see Table 3). Because inbreeding coefficients remained significant within each stock, the heterozygosity deficit cannot be explained uniquely by Wahlund effect. For clarity, we include in the text the following paragraph: “Based on the coancestry coefficients provided by Structure and Clumpp, the individuals were reorganized by genetic stock in sample sites that showed multiple stocks and were later submitted to the genetic analyses described above.” Same goes, in a less critical way, for disentangling effects from selection, from those of admixture. Selection was observed in the analysis that included samples of Chucurí, Puerto Berrío and Palagua but not in the analysis that excluded them. Since Fis values remain significant in all loci in other demes and genetic stocks that do not show evidence of selection their causes are attributable to other factors. Comments for the author The manuscript reports on a notable work in the area of Neotropical fish population genetics and adds to the expanding literature on prochilodontid species, with clear implications for production, management and conservation of an important fisheries resource and describes 11 new valid microsatellite markers isolated from <P. magdalanae>. It could be improved by a detailed copy editing (beyond this reviewer's pointed instances) and by addressing the specific questions raised, including why the NGS results were not presented and made available along the article, since they are an integral part of the works repeatability and could be used in other diverse potential works. Minor Changes L18: Capital letter in “basins” Suggestion not accepted. Here, “basins” is not a proper name. L23: I would recommend 'levels of diversity', since I tend to consider polymorphism to be a two state character: a locus is either polymorphic or monomorphic. The multiallelic quality of many microsatellite loci is better acknowledged as hipervariable, rather than more or less polymorphic... Done. We changed “levels of polymorphism” by ”levels of diversity” L24: Include “coefficient” after “inbreeding” Done L41: Include a comma in “sizes, high” Done L97: Include a space in “locidue” Done L121: This table presents some results not available before the study, so it should be relocated to the results section... Done L149: There are more recent editions to this classic text book, it should be prefered for the sake of the literature chronological. To attend a comment of the other reviewer, we deleted this sentence. L151: Maybe the authors should consider citing the original 2008 Jost paper (https://doi.org/10.1111/j.1365-294X.2008.03887.x) or a more recent review on the theme (https://doi.org/10.1111/eva.12590) Done. Now, we included: Jost, LOU. 2008. GST and its relatives do not measure differentiation. Molecular ecology 17(18), 4015-4026. Wright S. 1943. Isolation by distance. Genetics 28(2): 114. Wright S. 1965. The interpretation of population structure by F‐statistics with special regard to systems of mating. Evolution 19(3): 395-420. L209: What was the significance level considered (Table 2)? Now: “Values in bold represent significance at P < 0.05.” L211: For the sake of readability I would recommend rounding up to two decimal digits, these and all previous/subsequent parallels, including Tables. For consistence with values below 1 in parameters, we insist in including three decimal digits. L261: An useful result here would be the magnitude of the distance, such as kimura-2-parameter, to support your conclusions on the phylogenetic tree about the nature of the lack of genetic divergence from <P. reticulatus> Done. Now, we include a table showing Kimura 2-parameter distances. L269 - 270: Why were these results not presented? Now, we included: “Genomic sequencing of the Illumina shotgun library of P. magdalenae (0.115 Gb) generated 277,133 reads and 14,124 of 50,404 that contained microsatellite loci, were flanked by suitable PCR priming sites. More abundant repeat motifs were dinucleotides (47.758%), followed by tetranucleotide (28.353%), trinucleotide (16.193%), pentanucleotide (5.146%) and hexanucleotide (2.549%) repeats. The most common motifs found were AC (29,661%), TC (18,905%), ATT (4,811%) and AAAT (4,794%). The sequences of contigs containing the microsatellite loci obtained in the present study are provided in supplementary files S1 and S2.” L275: The most plausible explanation is the mixture one, since the fish are not from a single source Please, see the response above provided. L281-282: Some authors (e.g. doi: 10.1101/gr.075242.107) have found evidence that longer motifs tend to be more variable than shorter ones (i.e. number of alleles in average Tetra > Tri > Dinucleotides). Now: “…a higher variability is commonly expected in shorter motifs (e.g. 2mers, Orozco Berdugo & Narváez Barandica, 2014) compared with longer motifs (3mers and 4mers, this study)”. L285: I suggest considering, 'Still, …” instead “In this context…” This suggestion does not add clarity to the idea. L290-291: Yazbeck and Kalapothakis (2007, https://www.geneticsmr.org/articles/isolation-and-characterization-of-microsatellite-dna-in-the-piracema-fish-prochilodus-lineatus-characiformes.pdf) have implicitly reported an arithmetic average of 0.778 for He, in <Prochilodus lineatus> In this section, we are comparing studies in population genetics. Consequently, we do not include values derived from the development of microsatellite loci. L319: Change “exerted on” by “exerted over”? “exerted on” seems to be more commonly used. L341: Include “mitocondrial” in “current phylogenetic” Done. L355: Restocking or reinforcement, as the authors have argued? Done. For consistency, we use a single term. L355: This assumption is not always valid, mostly so for the trinucleotide loci (only Pma18?) (related to “since microsatellite loci are not transcriptionally active”). Now, we included “most”: “since most microsatellite loci are not transcriptionally active” L378: Although in a more superfical way, Yazbeck and Kalapothakis (2007), have also suggested substructuring in a sample of <P. lineatus>. We do not find some structure analysis that evidence the presence of different stocks in the sampled site. L390: Dito: restocking or reinforcement, as the authors have argued? Done. For consistency, we use a single term. L393-394: I would recommend caution with this conclusion, since I would favour the Wahlund explaneation for this apparent lack of heterozygotes. Now: “0.162 to 0.202 per genetic stock”. Please, see values of FIS of genetic stocks in Table 3. Table 1: What was the alpha value for test significance? Now: “P: statistical significance (values in bold represent significance at P < 0.05)”. Table 2: Dito: what was the alpha value? Done. Now “values in bold represent significance at P < 0.05” Table 7: What was the test significance (alpha value) after the Sequential Bonferroni procedure. Please confirm if this was indeed a Holm-Bonferroni type correction or just a simple Bonferroni corretion (less appropriate than the former). Done. Now: “Values in bold denote statistical significance after Bonferroni correction (Cauca: P < 0.0005; Magdalena: P < 0.0001).” Fig. 3: Labels are rather small, particularly in B, D and E. We made new figure 3 to summarize the structure analyses. Fig 4. Figure text, scale grades and legends are too small Done. Now we construct the new figure 3. Reviewer: Carlos Henrique Santos Basic reporting no comment Experimental design no comment Validity of the findings no comment Comments for the author This paper is effective, concise, and makes a compelling case for the study of the populations de Prochilodus magdalenae using species-specific microsatellite loci. I have no major issues with the theory or methods applied but would caution the authors to reconsider some details raised for the manuscript (see my comments below). I think it’s important to adhere to the most standardized, accessible terminology here, especially in the area of the conservation and population genetics (see my comments below). The structure of the paper is generally good, but the paragraph and sentence structure (particularly in the Introduction) needs to be better organized. While I recognize the importance of the data provided in the present study, I have detected a number of problems in the manuscript that make it unsuitable for publication the way it is. Therefore, in my opinion, there are some issues which need to be reviewed and/or answered in order for the manuscript to become suitable for publication. In addition, a review of English is required. I encourage the authors to review these issues and resubmit the manuscript for appreciation. Line-by-line revisions: Line 18: Change “The genetic structuring patterns...” to “The genetic structure patterns of populations....” Done Line 23: Withdraw “... next-generation sequencing, bioinformatics, and...” to “... next-generation sequencing, and...” We do not make this change because bioinformatics was used for assemblage of contigs and searching of loci microsatellite and e-PCR. Line 25: Withdraw “... and plausible signs of erosion...” to “... and signs of erosion...” Done. Line 29: Add “genetics” here: “... the genetics diversity and structure of P. Magdalenae...” We include the adjective “genetic” Line 39: Change “... fish species along the main river...” to “... fish species in the main river...” Done. Line 41-42: Change “... body sizes and high fecundities and abundances, representing around 50–80% of the biomass caught in artisanal and commercial fisheries throughout the distribution area...” to “... body sizes, high fecundities and abundances, representing around 50–80% of the biomass caught by the subsistence and commercial fisheries in some regions of your distribution area...” Artisanal fisheries is seen as subsistence fishery. However, care should be taken when generalizing a fisheries resource as representative for all your distribuition area. The information has to be very solid. Done. Now: “…body sizes, high fecundities and abundances, representing around 50–80% of the biomass caught by the subsistence and commercial fisheries in some regions of their distribution area…” Line 50: Add “ , ” here: “... permanent resource availability, as well as to guarantee...” Done. Line 54: Change “... of 2,182.67 metric tonin 2013...” to “...of 2,182.67 metric tons in 2013...” Done. Line 55: Add “the years of” here: “... between the years of 1978 and 2012...” It seems not necessary to include “the years of”. Line 56: Change “... population densities, catches (approx. 85%), and mean catch sizes...” to “... populations densities, mean catch size, and catch reduction around 85%...” Because this sentence includes “drastic decreases”, the suggestion would generate redundant information. Line 61: Change “... counteract its detrimental situation...” to “...counteract this detrimental situation...” Done. Line 63-64: Change “... was catalogued as under critical threat in 2002 and as vulnerable since 2012 in the Colombian Red List of freshwater fishes (Mojica et al., 2012)...” to “... was catalogued as critically endangered in 2002 and, in 2012 was considered as vulnerable for the Colombian Red List of freshwater fishes (Mojica et al., 2012)...” Done. Line 65-66: Change “... their efforts on population reinforcements (improperly called restocking) of natural stocks in...” to “... their efforts in the restocking population of natural stocks threatened in...” Done. Line 69: Change '... regulation of fish farming (Povh et al...” to “... regulation of the fish farm (Povh et al...” Suggestion not accepted. It refers to the regulation of an activity. Line 75: Change “... for population reinforcements of natural stocks...” to “... for population restocking of natural stocks...” Done. Line 75: Change “Hence, natural stocks...” to “Thus, natural stocks...” The sentence is correct. Line 81-82: Withdraw and Chance “... the observation that Prochilodus lineatus (Godoy, 1959) and Prochilodus argenteus (Godinho & Kynard, 2006) show fidelity to spawning sites (“homing”) suggests that P. magdalenae may exhibit...” to “... observations performed in the species Prochilodus lineatus (Godoy, 1959) and P. argenteus (Godinho & Kynard, 2006) show that these species present fidelity to the spawning sites (“homing”) suggesting that the species P. magdalenae can exhibit...” This suggestion does not add clarity to the idea. Line 84-85: Withdraw and Change “... Indeed, previous genetic studies have found the population structure and/or coexistence of multiples stocks along the Magdalena River and several tributaries...” to “... Previous genetic studies have found population structure and/or coexistence of multiples stocks along the Magdalena River and of several your tributaries...” Now: “Furthermore, previous genetic studies have found the population structure and/or coexistence of multiples stocks along the Magdalena River and several of its tributaries”. Line 84-92: Withdraw and Change “Although this structure may result from the unregulated population reinforcements of the natural stocks, it may also reflect a natural behavior of P. magdalenae since similar patterns of genetic population structure have been found in other congeners such as Prochilodus reticulatus (López-Macías et al., 2009), P. argenteus (Hatanaka & Galetti Jr., 2003; Hatanaka, Henrique-Silva, & Galetti Jr., 2006; Barroca et al., 2012a), P. lineatus (Ramella et al., 2006; Rueda et al., 2013; Gomes et al., 2017), and Prochilodus costatus (Barroca et al., 2012a,b)” to “Although the cause this structure can be the result of the restocking of populations unregulated for the natural stocks, it can also reflect in the natural behavior of P. magdalenae since similar patterns of genetic structure have been found in other congeners such as P. reticulatus (López-Macías et al., 2009), P. argenteus (Hatanaka & Galetti Jr., 2003; Hatanaka, Henrique-Silva, & Galetti Jr., 2006; Barroca et al., 2012a), P. lineatus (Ramella et al., 2006; Rueda et al., 2013; Gomes et al., 2017), and P. costatus (Barroca et al., 2012a,b), respectively” Now: “Although this structure may result from the unregulated restocking of the natural stocks, it could also reflect a natural behavior of P. magdalenae since similar patterns of genetic population structure have been found in other congeners such as P. reticulatus (López-Macías et al., 2009), P. argenteus (Hatanaka & Galetti Jr., 2003; Hatanaka, Henrique-Silva, & Galetti Jr., 2006; Barroca et al., 2012a), P. lineatus (Ramella et al., 2006; Rueda et al., 2013; Gomes et al., 2017), and P. costatus (Barroca et al., 2012a,b).” Line 95-97: Withdraw and Change “Likewise, we compare the genetic diversity and structure with those of five sites (Pijiño, Llanito, Mompox, Palomino, and San Marcos) previously studied by Orozco Berdugo & Narváez Barandica (2014)” to “Thus, we compare the genetic diversity and structure with the sites (Pijiño, Llanito, Mompox, Palomino, and San Marcos) previously studied by Orozco Berdugo & Narváez Barandica (2014)” Now “Additionally, we compare the genetic diversity and structure with those of five sites (Pijiño, Llanito, Mompox, Palomino, and San Marcos) previously studied by Orozco Berdugo & Narváez Barandica (2014).” Line 98: Add “study of” here: “... their advantages in study of population genetics...” Now: “…their advantages in the studies of population genetics” Line 104: Change “... the river mainstream and floodplain...” to “... the river main stream and floodplain...” Suggestion not accepted; “mainstream” seems to be correct. Line 104-106: Add “samples, (river name?), of, farm and (fish famr name?)” here: “... a total of 725 muscle tissues samples of P. magdalenae from the river main stream (river name?) and floodplain lakes along of the different Colombian hydrographic areas of the Magdalena-Cauca and Caribe (Fig. 1; Supplementary Information) and, 40 juveniles from a local fish farm hatchery (fish farm name?). Detailed information is provided in Fig. 1 and Supplementary information. Please, note that samples came from 8 rivers in both Colombian hydrographic areas: Magdalena-Cauca and Caribe. The name of fish farm is unknown. Blinded samples were provided to the laboratory by Integral S. A. Line 109-110: Change “... Territorial de Colombia #0155 on January 30, 2009 for Ituango hydropower...” to “... Territorial de Colombia #0155 (January 30, 2009) for Ituango hydropower...” Done. Line 112: Change and Add “respectively” here: “... permit #1293 of 2013 of the Universidad del Magdalena.” to “... permit #1293/2013 of the Universidad del Magdalena, respectively.” Done. For consistency with the previous suggestion, parentheses were used instead of a slash. Now: “…permit #1293 (2013) of the Universidad del Magdalena”. We do not include “respectively” because this sentence describes uniquely the samples used by Orozco Berdugo & Narváez Barandica (2014). Line 114-116: Withdraw “... P. magdalenae from the middle section of the Magdalena River was performed using the Illumina MiSeq v.2 instrument using the 'whole genome shotgun' strategy and the Nextera library preparation kits for the sequence...” to “... P. magdalenae for the region of middle Magdalena River was performed using the Illumina MiSeq v.2 platform (manufacturer, city, country). An alternative approach to shotgun whole genome sequencing (WGS) was applied using the Nextera Library preparation kit (specify the Nextera kit) (manufacturer, city, country) for the sequence...” We do not make this change because it distorts the original idea. Line 117: Change “... steps concerning the read cleaning, contig assemblage, identification...” to “... steps regarding the reads quality and filtering, contig assemblage, identification...” Now: “All steps regarding the read cleaning, contig assemblage, identification of microsatellite loci, primer design, in silico alignment of primers using electronic PCR (ePCR), PCR optimization, and polymorphism analysis of 50 microsatellites were performed following the methodology described by Landínez & Márquez (2016)” Line 120-126: Redo all this passage (see my comments on this topic – Major (1)). Please, see response to the comment on the topic- Major (1). Line 137-139: Withdraw and Change “... using LIZ500 (Applied Biosystems) as the internal molecular size. Allelic fragments were denoted according to their molecular size and scored using GeneMapper v.4.0 (Applied Biosystems)...” to “... using the GeneScan LIZ-500 standad size (Applied Biosystems, city, country) to determine fragment length. The alleles were scored based on the consistent pattern of their stutter peaks, and on the peak intensity corresponding to each individual at each locus using GeneMapper v4.0 (Applied Biosystems, city, country)...” In our study, the allele scoring was based on GeneScan 500 LIZ standard molecular size. Null alleles, stutter peaks or drop out effects were posteriorly examined with MicroChecker in raw data. Only loci with absence or low levels of stutter peaks were selected for further analyses. Line 142-144: Withdraw and Change “... for departures from Hardy–Weinberg linkage equilibria as well as the observed (HO) and expected (HE) heterozygosity and the inbreeding coefficient (FIS) were estimated using Arlequin v.3.5.2.2 (Excoffier, Laval, & Schneider...” to “... for Hardy–Weinberg equilibrium (HWE), observed (HO) and expected heterozygosity (HE) and, the inbreeding coefficient (FIS) were estimated using Arlequin v3.5.2.2 software (Excoffier, Laval, & Schneider...” Done. Line 146-147: Withdraw and Change “... for each marker were calculated with GenAlEx v.6.503 (Peakall & Smouse, 2006) and Cervus v.3.0.7 (Marshall et al., 1998), respectively” to ... for each SSR were calculated with GenAlEx v6.503 (Peakall & Smouse, 2006) and Cervus v3.0.7 software (Marshall et al., 1998), as well as to estimate the genetic diversity of P. magdalenae”. Now, we changed “marker” by “microsatellite loci”. The abbreviation SSR is not used along the manuscript. Line 148-150: Withdraw “The average number of alleles per locus, observed and expected average heterozygosities, and fixation index (Hartl & Clark, 1997) were calculated with GenAlex v.6.503 (Peakall & Smouse, 2006) to estimate the genetic diversity of P. magdalenae”. Done. Line 150-153: Withdraw and Change “... geographical samples was calculated using the standardized statistics F ́ST (Meirmans, 2006) and Jost ́s Dest (Meirmans & Hedrick, 2011) and analysis of molecular variance (AMOVA) (Meirmans, 2006) with 10,000 permutations and bootstraps included in GenAlex v.6.503...” to “... geographical samples were calculated using the standardized statistics F ́ST (Meirmans, 2006) and Jost ́s Dest (Meirmans & Hedrick, 2011) and, analysis of molecular variance (AMOVA) (Meirmans, 2006) with 10,000 permutations and bootstraps included in GenAlex v6.503 software...”. Please, note in the complete sentence that the verb to be is used accurately: “The genetic differentiation among geographical samples was calculated using the standardized statistics F´ST (Wright, 1943, 1965; Meirmans, 2006) and Jost´s D´est (Jost, 2008; Meirmans & Hedrick, 2011) and analysis of molecular variance (AMOVA) (Meirmans, 2006) with 10,000 permutations and bootstraps included in GenAlEx v6.503 software (Peakall & Smouse, 2006).” Line 156-167: Add 'software' and Withdraw “ . ” for all programs used. Example: Structure v.2.3.4 to Structure v2.3.4 software CLUMPP v.1.1.2b to CLUMPP v1.1.2b software Distruct v.1.1 to Distruct v1.1 software Done. Line 176: Withdraw and Change “... using the software BayeScan v.2.1 (Foll & Gaggiotti...” to “... using the BayeScan v2.1 software (Foll & Gaggiotti...” Done. Line 181-192: Add 'software' and Withdraw “ . ” for all programs used. Example: software jModelTest to jModelTest software software MrBayes v.3.2.6 to MrBayes v3.2.6 software Figtree v.1.4.3 to Figtree v1.4.3 software Done. Line 208-209: Change “... revealed that 8 of 11 loci exhibit allelic frequencies concordant with Hardy-Weinberg equilibrium expectations in at least one case (Table...” to ... revealed that 7 of 11 loci exhibit allelic frequencies concordant with Hardy-Weinberg equilibrium expectations in at least one case (see Table...”. Please, note in the P-values that 8 loci satisfied Hardy-Weinberg equilibrium expectations River Pma39 Pma25 Pma02 Pma35 Pma01 Pma40 Pma46 Pma36 Pma18 Pma13 Pma14 Across loci Cauca 0.000 0.000 0.002 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 Magdalena 0.000 0.001 0.510 0.000 0.000 0.000 0.000 0.000 0.000 0.058 0.002 0.000 San Jorge 0.650 0.299 0.645 0.000 0.638 0.002 0.531 0.307 0.009 0.318 0.000 0.000 Cesar 0.000 0.890 0.947 0.033 0.208 0.002 0.484 0.148 0.097 0.846 0.000 0.000 Nare 0.002 0.200 0.619 0.000 0.011 0.000 0.792 0.001 0.000 0.357 0.000 0.000 Sinú 0.000 0.064 0.129 0.000 0.000 0.004 0.074 0.004 0.143 0.089 0.036 0.000 Atrato 0.000 0.409 0.257 0.000 0.000 0.000 0.511 0.995 0.010 0.003 0.002 0.000 Fish Hatchery 0.030 0.014 0.625 0.000 0.007 0.000 0.001 0.000 0.000 0.137 0.000 0.000 Line 225: Withdraw “... magnitude, heterozygosity deficits and inbreeding coefficients...” to “... magnitude and inbreeding coefficients...” Done. Line 228-230: Change “... of the tests performed using Bottleneck (Table 4) were significant for all populations under the infinite alleles model (IAM) and for most populations under the two-phase model (TPM), whereas they were generally non-significant under the stepwise mutation model (SMM)” to “... of the genetic bottleneck were significant for all populations under the infinite alleles model (IAM) and for most populations under the two-phase model (TPM), however, was non-significant for the stepwise mutation model (SMM) (Table 4)”. Done. Now: “Results of the genetic bottleneck (Table 4) were significant for all populations under the infinite alleles model (IAM) and for most populations under the two-phase model (TPM), whereas they were non-significant under the stepwise mutation model (SMM).” Line 241: Change “... and AMOVA (F ́ST(7, 1407) = 0.009...” to “...and AMOVA (F ́ST(7, 1407) = 0.009...” Statistics are nor italicized along the text. Line 242-244: Change “... statistics F ́ST (Meirmans, 2006) and Jost ́s Dest (Meirmans, & Hedrick, 2011) showed additional genetic differences among Atrato, the fish hatchery, Sinú, and the remaining rivers (Table 6) as well as...” to “... statistics F ́ST (Meirmans, 2006) and Jost ́s Dest (Meirmans, & Hedrick, 2011) showed additional genetic differences among Atrato, the fish farm hatchery, Sinú, and the remaining rivers (Table 6), as well as...” “Fish hatchery” is wright. Major: (1) The authors discuss the parameters required to validate new microsatellite (SSR) primers, but it is not clear what these parameters. The criteria adopted by the authors to choose the 11 microsatellite loci, only two I consider important for validation of new microsatellite loci: (i) value of F or FIS (however, this depends directly on the pValue for the deviation of the Hardy-Weinberg equilibrium - HWE ) and (ii) the Polymorphic Information Content (PIC) - loci should be polymorphic. The authors speak of low levels of heterozygosity deficit, but when I see F or FIS values (see Table 1, HO < HE), I observe high levels of heterozygosity deficiency for all loci, as well as 16 loci with deviation HWE after Bonferroni correction (5%, p <= 0.05 / 21 = 0.00238). In addition, it is important to perform the linkage disequilibrium (LD) test for the validation of the 21 SSR loci. I suggest you take the LD test and add the information in the manuscript. We agree. However, we used 88 individuals chosen randomly from the whole sample. Consequently, values FIS and PIC are biased by the structure and genetic characteristics of the sample. In this context, some additional characteristics were considered such as clearly defined peaks, reproducibility and consistency of amplifications, bands concordant with motif sizes, absence or low presence of stutter bands, low levels of heterozygosity deficit and high levels of polymorphism. Information about LD was included in the text and the detailed information is now provided in the supplementary information (S5). 1.1. What are the criteria actually used for the choice of the 11 microsatellite loci for the present study? “I am not against the choice of 11 loci for this study, however, I would like this to be clear to me”. Clearly defined peaks, reproducibility and consistency of amplifications, specific bands. 1.2. What did mean by low levels of heterozygosity deficit? Now: “levels of heterozygosity”. 1.3. Why did not check the private alleles in the populations studied? I suggest adding the information from the private alleles in the manuscript. Although we found some private alleles, we consider that this information is still preliminary due to the differences among sample sizes and sampling periods. Besides, note that some private alleles are found in outlier loci (loci putatively under selection) in the Magdalena River. Sample site Pop Loci with private alleles Sample site Pop Loci with private alleles Cauca Pooled Pma01, 14, 25, 39, 40, 46 Magdalena Pooled Pma02, 13, 18, 36, 40 PuenteReal S1 Pma40 Mompox MP Pma40 PuenteReal S1 Pma46 Palomino PL Pma36 PuenteReal S1 Pma01 FLChucuri Ch Pma40 Pma13 Guriman S2 Pma46 FLChucuri Ch Pma40 Pma18 Pma13 Guamera S2 Pma14 FLChucuri Ch Pma02 Pma40 Pma18 Pma13 ValdiviaStream S3 Pma39 PuertoBerrio B Pma40 Pma18 Pma13 ManRiver S4 Pma01 PuertoBerrio B Pma40 Pma13 Margento S5 Pma14 FLPalagua P Pma40 Pma13 FLGrande S6a Pma25 FLPalagua P Pma40 Pma18 Pma13 Guaranda S6c Pma14 Sinú Pooled Pma01,40 Atrato Pooled Pma35, 40 CañoGrande SU Pma01 PaloBlanco AT Pma40 CañoGrande SU Pma40 PaloBlanco AT Pma35 Doctrina SU Pma40 (2) Why was not an analysis for isolation by distance (IBD) in the distribution area of the species studied? This would be interesting and would greatly contribute to the manuscript data. I suggest that this analysis be carried out. In isolation by distance model, a decrease in genetic similarity is expected among populations as the geographic distance between them increases. In contrast, we found evidence of gene flow, and consequently, the allele frequencies found in this study were not spatially autocorrelated. Despite, we performed the Mantel test requested to support our arguments. SSx SSy SPxy Rxy P(rxy-rand >= rxy-data) 2658194175,205 2257650,623 3101476,468 0,040 0,004 (3) I suggest the author better explain of the data information found in Figs. 2, 3 and 4 (structure), as well as the DAPC data. The current form as presented is very fragmented and can generate misinterpretation. What I observed in this part of the manuscript related to the structure and DAPC data: Figure 2A, two populations on the Magdalena River, and Fig. 2C two populations - Sinú and Atrato Rivers (Population 1) and the Cauca, Magdalena, Cesar, San Jorge and Nare Rivers (Population 2), respectively. The Fig. 3A shows that Magdalena (MG), Cesar (CS), San Jorge (SJ) and Nare (NA) rivers populations are more related than wtih the Cauca (CA) river populations. This is clearer with Fig. 3B and 3C, when we observed the Cauca (CA) and Magdalena (MG) rivers populations separately. Individuals from the fish farming hatchery, the author suggests that they originate from several rivers. It would be important for the authors to have this information about the origin of the individuals fish farm. Data on distance isolation (IBD) and a UPGMA tree would help to better understand Figs. 2, 3 and 4. As it was described above, IBD does not explain our results. Unfortunately, the origin of the individuals of the fish farm is not available. NOTE: I suggest the authors make a new run in the strucuture software up to a K = 8 (seven rivers + fish farm), adding each result of K to a single figure. Present the data for gene flow between the studied rivers. This information will be important to verify the level of reproductive isolation that these populations present. Now, we provided a new Fig. 2. Additionally, we include the Structure analyses within each river (new Fig. 3) because some authors have indicated that STRUCTURE software reveals only stocks related to the highest hierarchical grouping, limiting the nested fine substructure detection (Evanno, Regnaut & Goudet, 2005) and the cluster identification at low levels of genetic differentiation (Latch et al., 2006). Evanno G., Regnaut S., Goudet J. 2005. Detecting the number of clusters of individuals using the software structure: A simulation study. Molecular Ecology 14:2611–2620. DOI: 10.1111/j.1365-294X.2005.02553.x. Latch EK., Haverson LA., King JS., Hobson, MD., Rhodes OE. 2006. Assessing hybridization in wildlife populations using molecular markers: a case study in wild turkeys. Journal of Wildlife Management 70:485–492. Even with the existence of structure, the DAPC data show a genetic mix between the studied rivers, but this would be justified by the repopulation carried out in the region with individuals of the species Prochilodus magdalenae. We also discuss other alternatives that we can not exclude. The authors constructed a Bayesian tree (Fig. 5) using the cox1 gene. I suggest building a new tree containing the origin of each sample to understand the DAPC data. Done. Please see the new Fig. 4. (4) When checking Tables 2 and 3, it is observed that the majority of the populations of the target species are with values above 10% for FIS values. The FST values (Table 6) suggest that there is a good gene flow among the populations, however, the FIS values indicate the occurrence of mating with related individuals, what can be caused by the genetics bottleneck and/or restocking causing by the mixture of the population (the parents used in the restocking program probably originate from the same area of study). In addition, possible signs of local adaptation could have been verified (from high FST values, number of private and low alleles or absence of genetic flow). Please, note that we explored non-neutral evolutionary forces acting on the microsatellite loci, using the BAYESCAN v2.1 software (Foll & Gaggiotti, 2008) to detect candidate loci under selection. Detailed information was provided in the first version in the lines 175 – 179, section “Statistical analysis”; 235 – 237, section “Genetic diversity, population demography, and outlier loci screening” of “Results” and Table 5. The Wahlund effect on large rivers with the fragmentation of populations may be due to the construction of hydroelectric power plants (barrier to gene flow) e/or restocking program consequence performed in the rivers (different populations coexisting). As mentioned above, in addition to restocking, we explored different non-excluding alternatives that may explain the coexistence of different populations. To confirm the Wahlund effect due to the coexistence of genetic stocks in the study area, it will be important for the author to see the allelic frequencies that are different between populations and which has caused the heterozygosity deficit. In the Wahlund effect, non-significant heterozygosity deficit is expected within each genetic stock that coexist in a site. Thus, we separated the genetic groups that circulate in the Magdalena and Cauca Rivers based on its coancestry coefficients across 20 runs calculated by STRUCTURE software and summarized by Clummp software. Then, we calculated the genetic diversity and inbreeding coefficients in each stock (Please, see Table 3). Because inbreeding coefficients remained significant within each stock, the heterozygosity deficit can not be explained uniquely by Wahlund effect. In addition, there is no expectation that the Wahlund effect will be equal across loci, which is not supported by the observation of all loci showed significant FIS values. “Thus, it is lacking in the discussion a greater exploration of the consequences that this can bring if the management for species is not applied in the studied área”. Please, see lines 305 – 309 about this topic. Minor: (1) In the methodology nothing was found about of the DNA extraction step in the P. magdalenae samples. It is important that this information is contained in the material and methods. Please, add in the body of the text the method applied for DNA extraction. DNA isolation was described in the lines 128 – 129, “genotyping of samples” in “Materials and Methods” (2) “The extension step and a final elongation were absent in this thermal profile”. Justify why the absence of this step in the PCR? Did you use any method on this? If yes, it should be cited in the text. “…both annealing and extension temperatures can be combined into a single step called two-step PCR, instead of conventional three-step PCR. Two-step PCR shortens the time taken for the PCR process as there is no need for switching and stabilizing temperatures between annealing and extension. The extension time of PCR depends upon the synthesis rate of DNA polymerase and the length of target DNA.” (https://www.thermofisher.com/co/en/home/life-science/cloning/cloning-learning-center/invitrogen-school-of-molecular-biology/pcr-education/pcr-reagents-enzymes/pcr-cycling-considerations.html). The elongation step in the PCR, required for fragments >1 kb, is eliminated to avoid the formation of artifacts. Since our primers are designed for amplify fragments < 400 bp, we opted for the strategy Two-step PCR, considering that our annealing temperatures allowed efficiently amplify the amplicons. This approach is used by other authors: Henegariu, O., Heerema, N. A., Dlouhy, S. R., Vance, G. H., & Vogy, P. H. (1997). Multiplex PCR: Critical parameters and step-by-step protocol. Biotechniques, 23, 504–511. Vincent, I. R., Farid, A., & Otieno, C. J. (2003). Variability of thirteen microsatellite markers in American mink (Mustela vison). Canadian Journal of Animal Science, 83(3), 597–599. https://doi.org/10.4141/A03-001 (3) Caution: all chemicals and equipment must bear the following information - manufacturer, city and country. Example: GeneScan Liz-500 (-250) standard size (Applied Biosystems, Waltham, USA) or 96-well Veriti™ Thermal Cycler (Applied Biosystems, Waltham, USA). Done. (4) I am does not understand what the author meant by this passage - Line 197-198: 'A total of 21 of the 50 loci microsatellite evaluated were polymorphic and showed allelic frequencies that departed from Hardy-Weinberg equilibrium'. Now: “A total of 21 of the 52 microsatellite loci evaluated were polymorphic and showed Hardy-Weinberg disequilibrium (Table 1) and Linkage equilibrium (Table S5)”. How do you know that the allelic frequency departed from Hardy-Weinberg equilibrium? I suggest that add a supplementary table with the data of the allelic frequency. The redaction of the sentences regarding this topic was edited. Additionally, P- values of Hardy-Weinberg equilibrium were provided in all tables of the original manuscript. (5) Standardize in the manuscript: “Ho” to “HO”, “He” to “HE”, “F’ST” to “F’ST” and “F” to “FIS”, respectively. Done. (6) The fixation index (F) and inbreeding coefficient (FIS) they are not the same thing? Review table 3. Done. Now, we exclude the column “F” from all tables. (7) The genetic structure tends to decrease when populations are mixed, increasing or restoring the gene flow among individuals of different populations. Thus, I did not understand what the author wanted to say in the line 275-276: '... the genetic structure of the samples shaped by the mixture of two genetic stocks...' Now: “the genetic structure of the samples by the coexistence of two genetic stocks (see below).” "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Prochilodus magdalenae is a freshwater fish endemic to the Colombian Magdalena-Cauca and Caribbean hydrographic basins. The genetic structure patterns of populations of different members of Prochilodus and the historic restocking of its depleted natural populations suggest that P. magdalenae exhibits genetic stocks that coexist and comigrate throughout the rivers Magdalena, Cauca, Cesar, Sin&#250;, and Atrato. To test this hypothesis and explore the levels of genetic diversity and population demography of 725 samples of P. magdalenae from the studied rivers, we developed a set of 11 speciesspecific microsatellite loci using next-generation sequencing, bioinformatics, and experimental tests of the levels of diversity of the microsatellite loci. The results evidenced that P. magdalenae exhibits high genetic diversity, significant inbreeding coefficient ranging from 0.162 to 0.202, and signs of erosion of the genetic pool. Additionally, the population genetic structure constitutes a mixture of genetic stocks heterogeneously distributed along the studied rivers, and moreover, a highly divergent genetic stock was detected in Chucur&#237;, Puerto Berr&#237;o, and Palagua that may result from restocking practices.</ns0:p><ns0:p>This study provides molecular tools and a wide framework regarding the genetic diversity and structure of P. magdalenae, which is crucial to complement its baseline information, diagnosis and monitoring of populations, and to support the implementation of adequate regulation, management, and conservation policies.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>The family Prochilodontidae (Teleostei: Characiformes) comprises the genera Prochilodus, Semaprochilodus, and Ichthyoelephas, and encompasses 21 Neotropical freshwater fish species in the main river basins of South America <ns0:ref type='bibr'>(Castro &amp; Vari, 2004)</ns0:ref>. Most of the prochilodontids exhibit large body sizes, high fecundities, and abundances, representing around 50-80% of the biomass caught by the subsistence and commercial fisheries in some regions of their distribution area <ns0:ref type='bibr'>(Barroca et al., 2012b;</ns0:ref><ns0:ref type='bibr'>Melo et al., 2016a)</ns0:ref>. Furthermore, some members of Prochilodontidae constitute a potential resource for fish farming due to certain characteristics such as their fast growth and weight increase, rustic management, and high economic value <ns0:ref type='bibr'>(Flores-Nava &amp; Brown, 2010;</ns0:ref><ns0:ref type='bibr'>DellaRosa et al., 2014;</ns0:ref><ns0:ref type='bibr'>Roux et al., 2015)</ns0:ref>.</ns0:p><ns0:p>In addition to the economic importance, Prochilodontidae plays an important trophic role in aquatic ecosystems. These detritivorous and migratory fishes contribute to the nutrient cycling, distribution, equilibrium, and maintenance of energetic flows and support a wide trophic network for a great number of predators <ns0:ref type='bibr'>(Flecker, 1996)</ns0:ref>. Hence, the adequate management of fisheries is crucial for the maintenance of high productivity and permanent resource availability, as well as to guarantee the stability and continuity of the aquatic ecosystems <ns0:ref type='bibr'>(Taylor, Flecker, &amp; Hall, 2006;</ns0:ref><ns0:ref type='bibr'>Batista &amp; Lima, 2010)</ns0:ref>.</ns0:p><ns0:p>The bocachico Prochilodus magdalenae Steindachner 1878 is the most representative endemic species of the Colombian ichthyofauna, considered the emblematic fishery resource of the Magdalena-Cauca Basin, with an estimated unload for the Magdalena Basin of 2,182.67 metric tons in 2013 (Colombian fishing statistical service: SEPEC). However, between 1978 and 2012, this species experienced drastic decreases in its population densities, catches (approx. 85%), and mean catch sizes. These effects resulted from overfishing during migratory periods, violations of legislation related to mean catch sizes, and habitat disturbances including deforesting, floodplain lake desiccations, agrochemical or chemical contamination</ns0:p></ns0:div> <ns0:div><ns0:head>Genotyping of samples</ns0:head><ns0:p>The PCRs were conducted in a volume of 10 &#181;l, which contained 2-4 ng/&#181;l of template DNA isolated with the GeneJET Genomic DNA purification kit (Thermo Scientific, Karlsruhe, Germany) following the manufacturer&#180;s instructions, 1 &#215; buffer (Invitrogen, California, USA), 0.2 mM dNTPs (Thermo Scientific, Massachusetts, USA), 0.05 U/&#956;l Platinum&#8482; Taq DNA Polymerase (Invitrogen, , California, USA), 2.5 mM MgCl 2 , 2% formamide (Sigma-Aldrich, Steinheim, Germany), 0.35 pmoles/&#956;l labeled forward primer (either FAM6, VIC, NED, or PET, Applied Biosystems, California, USA), and 0.5 pmoles/&#956;l reverse primer <ns0:ref type='bibr'>(Macrogen, Seoul, Korea)</ns0:ref>. The PCRs were performed on a T100 thermocycler (BioRad, California, USA) with an initial denaturation step of 95 &#176;C for 3 min followed by 32 cycles consisting of a denaturation step of 90 &#176;C for 22 s and an annealing step for 18 s using the annealing temperatures described for each primer in Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>. The extension step and a final elongation were absent in this thermal profile. Finally, the PCRs were submitted to electrophoresis on an automated sequencer ABI 3730 XL (Applied Biosystems, California, USA) using GeneScan 500 LIZ dye size standard (Applied Biosystems, California, USA) as the internal molecular size. Allelic fragments were denoted according to their molecular size and scored using GeneMapper v4.0 software (Applied Biosystems, California, USA; Supplementary file S2). Before the statistical analysis, <ns0:ref type='bibr'>Micro-Checker v.2.2.3 (van Oosterhout et al., 2004</ns0:ref>) was run to detect potential genotyping errors.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Tests for Hardy-Weinberg and Linkage equilibria, observed (H O ) and expected (H E ) heterozygosities and inbreeding coefficient (F IS ) were estimated using Arlequin v3.5.2.2 software <ns0:ref type='bibr'>(Excoffier, Laval, &amp; Schneider, 2005)</ns0:ref>. The sequential Bonferroni correction was applied to adjust the statistical significance in multiple comparisons <ns0:ref type='bibr'>(Holm, 1979;</ns0:ref><ns0:ref type='bibr'>Rice, 1989)</ns0:ref>. The average number of alleles per locus and the PIC <ns0:ref type='bibr'>(Botstein et al., 1980)</ns0:ref> for each microsatellite locus were calculated with GenAlEx v6.503 software <ns0:ref type='bibr'>(Peakall &amp; Smouse, 2006)</ns0:ref> and Cervus v3.0.7 software <ns0:ref type='bibr'>(Marshall et al., 1998)</ns0:ref>, respectively. </ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The genetic differentiation among geographical samples was calculated using the standardized statistics F&#180;S T <ns0:ref type='bibr'>(Wright, 1943</ns0:ref><ns0:ref type='bibr'>(Wright, , 1965;;</ns0:ref><ns0:ref type='bibr'>Meirmans, 2006)</ns0:ref>, Jost&#180;s D&#180;e st <ns0:ref type='bibr'>(Jost, 2008;</ns0:ref><ns0:ref type='bibr'>Meirmans &amp; Hedrick, 2011)</ns0:ref> and analysis of molecular variance (AMOVA) <ns0:ref type='bibr'>(Meirmans, 2006)</ns0:ref> with 10,000 permutations and bootstraps included in GenAlEx v6.503 software <ns0:ref type='bibr'>(Peakall &amp; Smouse, 2006)</ns0:ref>. Furthermore, the diploid genotypes of 11 loci (22 variables) in 725 individuals were submitted to discriminant analysis of principal components (DAPC) using the R-package Adegenet <ns0:ref type='bibr'>(Jombart, 2008)</ns0:ref>.</ns0:p><ns0:p>To examine other groupings of the samples, genetic differentiation among samples was tested using the Bayesian analysis of population partitioning with Structure v2.3.4 software <ns0:ref type='bibr'>(Pritchard, Stephens &amp; Donnelly, 2000)</ns0:ref>. Parameters included 350,000 Monte Carlo Markov Chain steps and 50,000 iterations as burn-in, the admixture model, correlated frequencies, and the LOCPRIOR option for detecting relatively weak population structure <ns0:ref type='bibr'>(Hubisz et al., 2009)</ns0:ref>. Each analysis was repeated 20 times for each simulated K value, which ranged from 1 to n + 3 (n, number of populations compared). For a best estimation of genetic stocks (K), the web-based software STRUCTURESELECTOR <ns0:ref type='bibr'>(Li &amp; Liu, 2018)</ns0:ref> was used to calculate the &#916;K ad hoc statistic <ns0:ref type='bibr'>(Evanno, Regnaut, &amp; Goudet, 2005)</ns0:ref>, the estimators MEDMEANK, <ns0:ref type='bibr'>MAXMEANK, MEDMEDK, and MAXMEDK (Puechmaille, 2016)</ns0:ref>, and to generate the graphical representation of results using the integrated Clumpak software <ns0:ref type='bibr'>(Kopelman et al., 2015)</ns0:ref>. Based on the coancestry coefficients provided by Structure and Clumpp, the individuals were reorganized by genetic stock in sample sites that showed multiple stocks and were later submitted to the genetic analyses described above.</ns0:p><ns0:p>Additionally, the occurrence of recent genetic bottlenecks of populations was evaluated by calculating the levels of heterozygosity and the M ratio using Bottleneck v1.2.02 software <ns0:ref type='bibr'>(Piry, Luikart, &amp;</ns0:ref><ns0:ref type='bibr'>Cornuet, 1999) and</ns0:ref><ns0:ref type='bibr'>Arlequin v3.5.2.2 (Excoffier, Laval, &amp;</ns0:ref><ns0:ref type='bibr'>Schneider, 2005)</ns0:ref>, respectively. Excess heterozygosity was assessed by employing the Wilcoxon sign-rank test <ns0:ref type='bibr'>(Luikart &amp; Cornuet, 1998)</ns0:ref>. The M ratio -the mean ratio of the number of alleles compared to the range of allele size -indicates that the population has Manuscript to be reviewed experienced a recent and severe reduction in population size when its values are smaller than 0.680 <ns0:ref type='bibr' target='#b4'>(Garza &amp; Williamson, 2001)</ns0:ref>.</ns0:p><ns0:p>To explore non-neutral evolutionary forces acting on the microsatellite loci, a scanning analysis was performed using the BayeScan v2.1 software <ns0:ref type='bibr'>(Foll &amp; Gaggiotti, 2008)</ns0:ref> to detect candidate loci under selection. Parameters for BayeScan analyses included 10:1 prior odds for the neutral model and 20 pilot runs consisting of 5,000 iterations each followed by 250,000 iterations with a burn-in length of 50,000 iterations <ns0:ref type='bibr'>(Foll &amp; Gaggiotti, 2008)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Phylogenetic relationships among genetic groups</ns0:head><ns0:p>To explore the phylogenetic relationships among individuals sampled along the basin, partial fragments of the mitochondrial cox1 gene (~650 bp) were amplified in a subset of samples using primers and PCR conditions previously described by <ns0:ref type='bibr'>Ward et al. (2005) and</ns0:ref><ns0:ref type='bibr'>Ivanova et al. (2007)</ns0:ref>. PCR products were sequenced by the Sanger method using an automated sequencer, ABI 3730 XL (Applied Biosystems). The best-fit evolutionary model was determined based on the Bayesian information criterion as implemented in the jModelTest v2.1.7 software <ns0:ref type='bibr'>(Posada &amp; Crandall, 1998)</ns0:ref>. A Bayesian phylogenetic analysis was conducted in MrBayes v3.2.6 software <ns0:ref type='bibr'>(Ronquist &amp; Huelsenbeck, 2003)</ns0:ref> including GenBank sequences of Prochilodus magdalenae, Prochilodus reticulatus, Prochilodus mariae, Prochilodus nigricans and using Ichthyoelephas longirostris as outgroup. For this purpose, we performed two independent runs of 20 million generations sampled each 1,000 generations using 25% as burn-in. The remaining values were left as default. The convergence of each parameter was checked based on a potential scale reduction factor nearing 1, an average standard deviation of the split frequencies lower than 0.010, and the visualization of the resulting trees was performed with FigTree v1.4.3 software <ns0:ref type='bibr'>(Rambaut, 2012)</ns0:ref>. Finally, the pair-wise divergences of P. magdalenae and P. reticulatus haplotype sequences were estimated using the Kimura 2parameters model in MEGA v10.1.8 software <ns0:ref type='bibr'>(Tamura et al., 2013)</ns0:ref>. Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head></ns0:div> <ns0:div><ns0:head>Microsatellite loci development</ns0:head><ns0:p>Genomic sequencing of the Illumina shotgun library of P. magdalenae (0.115 GB) generated 277,133 reads and 14,124 of 50,404 that contained microsatellite loci, were flanked by suitable PCR priming sites.</ns0:p><ns0:p>The dinucleotides (47.758%) were the most abundant repeat motifs, followed by tetranucleotide (28.353%), trinucleotide (16.193%), pentanucleotide (5.146%), and hexanucleotide (2.549%) repeats. The most common motifs found were AC (29.661%), TC (18.905%), ATT (4.811%), and AAAT (4.794%).</ns0:p><ns0:p>The sequences of contigs containing the microsatellite loci obtained in the present study are provided in supplementary files S3 and S4.</ns0:p><ns0:p>A total of 21 of the 52 microsatellite loci evaluated were polymorphic and showed Hardy-Weinberg disequilibrium (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>) and Linkage equilibrium (Supplementary file S5). The number of alleles per locus ranged from 11 to 37, with an average number of 20.619 alleles/locus, the average values of observed and expected heterozygosities were Ho = 0.589 and He = 0.876 and the PIC values ranged from 0.399 to 0.949 (average 0.867) (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>). A total of 10 loci failed to satisfy the selection criteria, showing low values of PIC (Pma32), dropout and stuttering (Pma32, Pma08), inconsistent amplifications (Pma17, Pma47, Pma57), or low-definition peaks (Pma42, Pma56, Pma26, Pma50). Consequently, only 11 (Pma39, Pma25, Pma02, Pma35, Pma01, Pma40, Pma46, Pma36, Pma18, Pma13, and Pma14) satisfied most of the parameters required to validate the new microsatellites primers described previously.</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic diversity, population demography, and outlier loci screening</ns0:head><ns0:p>Comparisons among rivers revealed that 8 of 11 loci satisfied the Hardy-Weinberg equilibrium expectations in at least one case (Table <ns0:ref type='table'>2</ns0:ref>). However, the analysis across loci showed significant departures from Hardy-Weinberg equilibrium expectations in all rivers evaluated (Table <ns0:ref type='table'>2</ns0:ref>). The average PeerJ reviewing <ns0:ref type='table' target='#tab_1'>PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:ref> Manuscript to be reviewed number of alleles per locus was higher in Cauca (22.455) and Magdalena (19.455), followed by Nare (15.636), Sin&#250; (15.273), the fish hatchery (14.818), and Atrato (14.636) and was lowest in San Jorge (13.545) and Cesar (13.364). Additionally, the highest values of observed and expected heterozygosities were found in San Jorge (Ho: 0.809; He: 0.884) and Cesar (Ho: 0.782; He: 0.873), followed by Sin&#250; (Ho: 0.767; He: 0.882), Magdalena (Ho: 0.758; He: 0.896), and Cauca (Ho: 0.725; He: 0.898) and were lowest in Atrato (Ho: 0.718; He: 0.879), the fish hatchery (Ho: 0.691; He: 0.880), and Nare (Ho: 0.659; He: 0.876) (Table <ns0:ref type='table'>2</ns0:ref>).</ns0:p><ns0:p>Furthermore, comparisons among sites within each river showed similar high levels of genetic diversity (Table <ns0:ref type='table'>3</ns0:ref>). The highest value of genetic diversity was found in the floodplain lake Palagua in the Magdalena River (Na: 17.182 alleles/locus; He: 0.895; Ho: 0.792), whereas the lowest was observed in Bet&#233;, a site of the Atrato River (Na: 9.273 alleles/locus; He: 0.791; Ho: 0.711). In addition, all sites exhibited a highly significant deficit of observed heterozygosity (Table <ns0:ref type='table'>3</ns0:ref>) with Mata de Palma and Saman&#225; Norte River showing the lowest and highest observed heterozygosity deficits, respectively.</ns0:p><ns0:p>Inbreeding coefficients (F IS ) per site in main rivers of the different Colombian hydrographic areas were significant and ranged from 0.120 to 0.255 (Table <ns0:ref type='table'>3</ns0:ref>). Although decreased in magnitude, the inbreeding coefficients (Table <ns0:ref type='table'>3</ns0:ref>) remained significant even after comparing the genetic diversity according to genetic stocks in Chucur&#237;, Puerto Berr&#237;o, and Palagua and among the Magdalena River and tributaries.</ns0:p><ns0:p>Results of the genetic bottleneck tests (Table <ns0:ref type='table'>4</ns0:ref>) were significant for all populations under the infinite alleles model (IAM) and for most populations under the two-phase model (TPM), whereas they were nonsignificant under the stepwise mutation model (SMM). As it is thought that few loci follow the strict SMM <ns0:ref type='bibr'>(Piry, Luikart, &amp; Cornuet, 1999)</ns0:ref>, the best estimation of expected heterozygosity at mutation-drift equilibrium is expected under a combination of IAM and TPM. Additionally, all values of the M ratio were substantially smaller than 0.680, indicating that all populations have experienced recent and severe reductions in population size (Table <ns0:ref type='table'>4</ns0:ref>). </ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>In contrast to other samples that did not show evidence of selection, BayeScan analysis revealed that 8 of 11 loci (Pma39, Pma25, Pma02, Pma35, Pma40, Pma36, Pma13, and Pma14) exhibit substantial evidence of selection in the Magdalena River (Table <ns0:ref type='table'>5</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic structure and phylogenetic relationships among the samples studied</ns0:head><ns0:p>At regional scale, the Bayesian analysis showed the presence of two genetic stocks (&#916;K = 2; MEDMEDK = 2; MEDMEANK = 2), one in the Magdalena River (Chucur&#237; + Puerto Berrio + Palagua) and the other one in the remaining evaluated rivers (Fig. <ns0:ref type='figure' target='#fig_8'>2A</ns0:ref>), which is concordant with DAPC (Fig. <ns0:ref type='figure' target='#fig_8'>2B</ns0:ref>) and AMOVA (F ST(7, 1407) = 0.009; P = 0.000). Together with Chucur&#237; + Puerto Berrio + Palagua, a predominant genetic stock with different levels of genetic admixture in Sin&#250; and Atrato rivers was revealed in the clusters suggested by the MAXMEAN and MAXMED statistics (K = 3). The additional clustering patterns (K = 4 -8; Fig. <ns0:ref type='figure' target='#fig_8'>2A</ns0:ref>) examined to compared them with other approaches, showed genetic admixture with other stocks absent in the examined rivers. Likewise, pairwise comparisons of the standardized statistics F&#180;S T <ns0:ref type='bibr'>(Meirmans, 2006)</ns0:ref> and Jost&#180;s D&#180;e st <ns0:ref type='bibr'>(Meirmans, &amp; Hedrick, 2011)</ns0:ref> showed genetic differences among Atrato, the fish hatchery, Sin&#250;, and the remaining rivers (Table <ns0:ref type='table'>6</ns0:ref>) as well as among the Magdalena River and its tributaries, Cauca and Nare.</ns0:p><ns0:p>However, excluding samples that exhibit loci putatively under selection (Chucur&#237; + Puerto Berrio + Palagua), comparisons among sites within each river revealed a genetic admixture of two stocks (&#916;K = 2; MEDMEDK = 2; MEDMEANK = 2) homogenously distributed in Magdalena River and its tributaries (Figs. <ns0:ref type='figure' target='#fig_9'>3A and 3B</ns0:ref>; Tables <ns0:ref type='table' target='#tab_2'>6, 7</ns0:ref>). Additionally, this analysis revealed a genetic substructure in Sin&#250; (&#916;K = 2; MEDMEDK = 2; MEDMEANK = 2; Figs. <ns0:ref type='figure' target='#fig_9'>3C and 3D</ns0:ref>; F ST(1, 67) = 0.033; P = 0.000; F&#180;S T = 0.027; P = 0.004; D&#180;est = 0.149; P = 0.005). In Atrato River, the Bayesian analysis showed a single genetic stock (&#916;K = 2; MEDMEDK = 1; MEDMEANK = 1; Fig. <ns0:ref type='figure' target='#fig_9'>3E</ns0:ref>) although remaining analysis showed genetic differentiation among sites (Fig. <ns0:ref type='figure' target='#fig_9'>3F</ns0:ref>; F ST(1, 57) = 0.045; P = 0.000; F&#180;S T = 0.047; P = 0.000; D&#180;est = 0.330; P = 0.000).</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Finally, the Bayesian tree using the cox1 gene clustered our samples (GenBank accession numbers MK330430 to MK330494) with sequences of P. magdalenae and P. reticulatus deposited in public databases and in a different group, Prochilodus mariae and Prochilodus nigricans (Fig. <ns0:ref type='figure' target='#fig_10'>4</ns0:ref>). Moreover, Kimura-2-parameters genetic distances (Supplementary file S6) were larger among haplotypes of P. magdalenae (0.002 -0.010) than among P. magdalenae and P. reticulatus haplotypes (0.000 -0.005).</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head></ns0:div> <ns0:div><ns0:head>Microsatellite loci development</ns0:head><ns0:p>This work developed species-specific microsatellite loci using next-generation sequencing and bioinformatic analysis. Although a total of 21 of 52 microsatellite loci with tri-and tetra-nucleotide motifs were polymorphic in P. magdalenae, the consistency in the amplification in a larger sample, allelic size class distribution, and high definition peaks allowed the selection of only 11 microsatellite loci for further population genetic analysis. Most of the loci showed departures from Hardy-Weinberg equilibrium and significant observed heterozygosity deficit in the random sample. The observed heterozygosity deficit may be related to technical problems such as silent alleles; however, it remains to explore the potential variations in the primer alignment sequences, since this study sequenced the genome of a single specimen of P. magdalenae. Two non-excluding explanations may be related to the significant levels of inbreeding and the genetic structure of the samples by the coexistence of two genetic stocks (see below).</ns0:p><ns0:p>Although the levels of genetic diversity measured by the expected heterozygosities were similar, the levels of observed heterozygosity as well as the average number of alleles per locus found in this study Manuscript to be reviewed were substantially greater than those found in the same samples by Orozco-Berdugo &amp; Narv&#225;ez-Barandica (2014). However, despite these differences, both heterologous (Orozco-Berdugo &amp; Narv&#225;ez-Barandica, 2014) and species-specific microsatellite loci (this study) revealed a general deficit of heterozygotes in all samples. In this context, the species-specific microsatellite loci developed in this study seem to provide a good approach to study the population genetics of P. magdalenae considering that the levels of heterozygosity constitute a parameter used to estimate the genetic diversity of the populations. In addition to the applications in harvest management, stocking programmes, definition of conservation units, recovery of threatened species, and management of invasive species, these tools may be useful in forensic genetics since partially degraded DNA samples are often found in this area (see <ns0:ref type='bibr'>Bourret et al., 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic diversity and population demography</ns0:head><ns0:p>Microsatellite data revealed average values of genetic diversity (He: 0.737) among the highest values found in other Prochilodontidae species, only surpassed by those reported for P. costatus <ns0:ref type='bibr'>(Melo et al., 2013)</ns0:ref> and P. argenteus (Coimbra et al., 2017) (0.747 and 0.753 respectively). Similarly, the average levels of expected heterozygosity were higher than that found in P. magdalenae measured by heterologous microsatellites (He: 0.877; Orozco-Berdugo &amp; Narv&#225;ez-Barandica, 2014) and Neotropical Characiforms (He: 0.675 &#177; 0.160; see review by <ns0:ref type='bibr'>Hilsdorf &amp; Hallerman, 2017)</ns0:ref>.</ns0:p><ns0:p>Additionally, this study found levels of observed heterozygosity higher than those found by Orozco-Berdugo &amp; Narv&#225;ez-Barandica (2014). However, the use of species-specific microsatellite loci developed in this study revealed similar values of expected heterozygosity among samples analyzed by Orozco-Berdugo &amp; Narv&#225;ez-Barandica (2014) and the remaining samples analyzed, indicating that differences between the two studies are related to the type of microsatellite loci utilized (heterologous vs. speciesspecific microsatellite loci).</ns0:p><ns0:p>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>The significant deficit of observed heterozygosity in all studied samples corroborates the previous findings for P. magdalenae from Magdalena River (Orozco-Berdugo &amp; Narv&#225;ez-Barandica, 2014); however, the magnitude of the observed heterozygosity deficit as well as the inbreeding coefficient (0.075 -0.239) were substantially lower than those previously reported (0.624 -0.788). Following <ns0:ref type='bibr' target='#b3'>Franklin (1980)</ns0:ref> and <ns0:ref type='bibr'>Soul&#233; (1980)</ns0:ref>, the values above 10% of the inbreeding coefficient indicate that these populations require careful management to avoid future detrimental effects on its populations. This point is important since it has been recommended recently that any inbreeding coefficient higher than zero will usually have an adverse fitness effect <ns0:ref type='bibr' target='#b2'>(Frankham, Bradshaw, &amp; Brook, 2014)</ns0:ref>.</ns0:p><ns0:p>Another non-excluding alternative is plausible considering that the significant deficit of observed heterozygosity observed in all sites analyzed may be also explained by the coexistence of genetic stocks (Wahlund effect) as this was evidenced by the genetic structure analysis (see below). Another biological cause of observed heterozygosity deficit, assortative mating, does not seem to explain the results found in this study because P. magdalenae is iteroparous and characterized by total spawning (Jaramillo-Villa &amp; Jim&#233;nez-Segura, 2008) as described in its congeners, P. costatus <ns0:ref type='bibr'>(Carolsfield et al., 2004)</ns0:ref> and P. lineatus <ns0:ref type='bibr'>(Roux et al., 2015)</ns0:ref>. Even more, in this latter species, the genetic analysis based on microsatellite loci support polygamous mating in both sexes <ns0:ref type='bibr'>(Ribolli et al., 2020)</ns0:ref>.</ns0:p><ns0:p>On the other hand, this study also provided evidence for a population bottleneck, suggesting that P. magdalenae shows signs of erosion of the genetic pool, likely by the constant pressure from fishing and other anthropogenic activities exerted on its populations. Although paradoxical to the observed heterozygosity deficit found in all populations evaluated, this outcome is plausible considering that the Bottleneck algorithm tests not for an excess of heterozygotes (Ho &gt; He) but rather for an excess of heterozygosity (He &gt; He at mutation-drift equilibrium) <ns0:ref type='bibr'>(Piry, Luikart, &amp; Cornuet, 1999)</ns0:ref>. Besides, the combination of a population bottleneck and an observed heterozygosity deficit may result from population growth in a closed system, population genetic structure, or admixture <ns0:ref type='bibr'>(Barson, Cable, &amp; Oosterhout, 2009)</ns0:ref>. Considering the lengths of the rivers studied, population growth in a closed system is Manuscript to be reviewed unlikely but the last two alternatives may explain our results due to the coexistence of genetic stocks in the samples studied and the continuous restocking of natural stocks using juveniles from fish hatcheries, which may create an apparent excess of novel alleles and an incomplete allele frequency distribution.</ns0:p><ns0:p>Similar results have also been found in guppies, Poecilia reticulata, in Trinidad and Tobago <ns0:ref type='bibr'>(Barson, Cable, &amp; Oosterhout, 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Genetic structure</ns0:head><ns0:p>This study tested the hypothesis that P. magdalenae exhibits genetic stocks that coexist and co-migrate along sections of the main channel and some tributaries of the Magdalena River (Cauca, San Jorge, and Cesar), Sin&#250;, and Atrato rivers. Before testing this hypothesis, we compared the genetic structure at regional scale, finding two spatially structured populations: one in the Magdalena River (Puerto Berr&#237;o and the floodplains Chucur&#237; and Palagua) and the other in the remaining rivers evaluated.</ns0:p><ns0:p>The geographical genetic structure may result from taxonomic differences among stocks due to the lack of regulations on the restocking of natural stocks of P. magdalenae. The phylogenetic analysis using partial sequences of cox1 gene indicates that samples do not correspond to species such as P. marie or P. nigricans because this genetic stock is clustered with previously published sequences of P. magdalenae <ns0:ref type='bibr' target='#b0'>(Aguirre-Pab&#243;n, Narv&#225;ez-Barandica, &amp; Castro-Garc&#237;a, 2013)</ns0:ref>. However, it remains to be seen whether they represent artificial mixtures of P. magdalenae and P. reticulatus because the current mitochondrial phylogenetic analysis of Prochilodontidae does not allow the two species to be discriminated <ns0:ref type='bibr'>(Melo et al., 2016b</ns0:ref><ns0:ref type='bibr'>(Melo et al., , 2018))</ns0:ref>. Moreover, the morphological and molecular similitudes have led to the proposal that P. magdalenae and P. reticulatus represent only one species with probable allopatric differentiation resulting from the uplift of the Sierra del Perij&#225; <ns0:ref type='bibr'>(Melo et al., 2016b)</ns0:ref>. Thus, a separated clustering of mitochondrial sequences of those stocks is not expected in the phylogenetic analysis even though they represent allopatric populations. Manuscript to be reviewed An alternative explanation is that the genetic differences result from eight outlier loci that are putatively under selection in three sites of the Magdalena River, suggesting that P. magdalenae experiences natural/artificial selection or local adaptation, although testing of these hypotheses is out of the scope of the present study. The explanation that outlier loci represent false positives resulting from the inclusion of severely bottlenecked populations <ns0:ref type='bibr'>(Teshima, Coop, &amp; Przeworski, 2006;</ns0:ref><ns0:ref type='bibr'>Foll &amp; Gaggiotti, 2008)</ns0:ref> seems unlikely because the significant excess of heterozygosity and small values of the M ratio were found even in populations that do not exhibit outlier loci. Thus, considering that those sites have been exposed to restocking since 20 years ago and since most microsatellite loci are not transcriptionally active, the outlier loci found in this study may reflect hitchhiking selection resulting from restocking using juveniles selected artificially by fish hatcheries. Alternatively, the outlier loci may result from asymmetric gene flow by unidirectional migration from hatchery stocks to wild populations. Similar results were found in Denmark in populations of three brown trout, which have been significantly admixtured with stocked hatchery trout <ns0:ref type='bibr' target='#b7'>(Hansen, Meier, &amp; Mensberg, 2010)</ns0:ref>.</ns0:p><ns0:p>Although the above reasoning might explain the genetic differences between stocks, an additional justification is required to explain the restricted distribution of one genetic stock in only three sites of the Magdalena River considering the migratory abilities of these species/allopatric populations. Thus, this genetic structure seems to result from recent restocking before reproductive/feeding migrations, use of artificial barriers to avoid migration of the fish, clogging by sedimentation or vegetation, or the desiccation of access to floodplain lakes or may be a product of the intensive anthropic intervention in these territories characterized by the exploitation of hydrocarbons and livestock. This idea is concordant with the fact that degradation of preferred habitat and barriers that impede dispersal contribute to the degree of genetic differentiation among populations <ns0:ref type='bibr'>(Faulks, Gilligan, &amp; Beheregaray, 2011)</ns0:ref>. Furthermore, the results found here provide support for the hypothesis that P. magdalenae exhibits genetic stocks that coexist and co-migrate along sections of the rivers Magdalena, Cauca, Cesar (tributaries of the Magdalena River), Sin&#250;, and Atrato. Since similar patterns of genetic structure are Manuscript to be reviewed found in P. reticulatus <ns0:ref type='bibr'>(L&#243;pez-Mac&#237;as et al., 2009)</ns0:ref>, P. marggravii <ns0:ref type='bibr' target='#b9'>(Hatanaka &amp; Galetti Jr., 2003)</ns0:ref>, P.</ns0:p><ns0:p>argenteus <ns0:ref type='bibr'>(Sanches et al., 2012)</ns0:ref>, P. costatus <ns0:ref type='bibr' target='#b1'>(Barroca et al., 2012a)</ns0:ref>, P. magdalenae <ns0:ref type='bibr'>(Orozco Berdugo &amp; Narv&#225;ez Barandica, 2014;</ns0:ref><ns0:ref type='bibr'>Hern&#225;ndez, Navarro, &amp; Mu&#241;oz, 2017)</ns0:ref>, and Ichthyoelephas longirostris <ns0:ref type='bibr'>(Land&#237;nez-Garc&#237;a &amp; M&#225;rquez, 2016)</ns0:ref>, this outcome supports the idea that this genetic structure is a generalized tendency within the family Prochilodontidae.</ns0:p><ns0:p>Excluding the genetic stock of Puerto Berr&#237;o and the floodplains Chucur&#237; and Palagua, each river showed the coexistence of at least two genetic stocks. Homogeneous and non-homogeneous distributions of these genetic stocks along the rivers explain similarities (Cauca, Magdalena, San Jorge, Cesar, and Nare) as well as geographical differences among the rivers analyzed (within Magdalena, including Puerto Berr&#237;o and the floodplains Chucur&#237; and Palagua, Sin&#250;, and Atrato). This genetic structure also explains the significant heterozygosity deficit observed in all sites analyzed (Wahlund effect) as discussed above.</ns0:p><ns0:p>Similar evidence of the Wahlund effect has been documented in the congener P. costatus, which exhibited genetic differences resulting from temporal isolation <ns0:ref type='bibr'>(Braga-Silva &amp; Galetti Jr., 2016)</ns0:ref>.</ns0:p><ns0:p>Although sampling in this study was not designed to detect temporal genetic structuring, genetic similarities among samples collected in different years suggest that the Wahlund effect must be more spatial than temporal. It remains to be seen whether this behavior is natural or artificial, considering that the restocking activities have been widely implemented along different Colombian rivers.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>This study provides evidence that P. magdalenae exhibits high genetic diversity, significant inbreeding levels per genetic stock, and signs of erosion of the genetic pool and conforms a mixture of genetic stocks heterogeneously distributed along the rivers studied. Additionally, this study developed a set of 11 microsatellite loci that allows the reliable detection of levels of genetic diversity, providing a tool for monitoring changes in the genetic diversity of the species, brood stocks, and juveniles used for supportive Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)Manuscript to be reviewed breeding and for measuring the efficacy of current population restocking activities. Management and conservation strategies need to be implemented at the level of the basins Magdalena-Cauca, Sin&#250;, and Atrato concordantly with their genetic population structure.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Primer sequences, characteristics, polymorphism levels, and genetic diversity of 21 species-specific microsatellite loci in 88 individuals of Prochilodus magdalenae randomly chosen from the whole sample.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Manuscript to be reviewed</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>Ta: annealing temperature standardized in PCRs, Na: number alleles per locus; Ra: allelic size</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>range (bp); PIC: polymorphism information content; H O and H E : observed and expected</ns0:cell></ns0:row><ns0:row><ns0:cell cols='8'>heterozygosities, respectively; P: statistical significance (values in bold represent significance</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>at P &lt; 0.05).</ns0:cell><ns0:cell>a</ns0:cell><ns0:cell cols='3'>Satisfied selection criteria,</ns0:cell><ns0:cell>b</ns0:cell><ns0:cell>inconsistent amplifications,</ns0:cell><ns0:cell>c</ns0:cell><ns0:cell>low definition peaks,</ns0:cell></ns0:row><ns0:row><ns0:cell>d</ns0:cell><ns0:cell>dropout,</ns0:cell><ns0:cell>e</ns0:cell><ns0:cell cols='3'>stuttering,</ns0:cell><ns0:cell>f</ns0:cell><ns0:cell>low value of PIC.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020) PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 7 (on next page)</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Pairwise Jost's D est (upper diagonal) and F' ST (below diagonal) of Prochilodus magdalenae samples among sites of the rivers Cauca and Magdalena.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>S6a: Floodplain Lake Grande, S6b: Floodplain Lake Caimanera F, S6c: Guaranda. Values in</ns0:cell></ns0:row><ns0:row><ns0:cell>bold denote statistical significance after Bonferroni correction (Cauca: P &lt; 0.0005;</ns0:cell></ns0:row><ns0:row><ns0:cell>Magdalena: P &lt; 0.0001).</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2018:12:33631:2:1:NEW 30 Sep 2020)</ns0:note> </ns0:body> "

"Medellín, September 24th, 2020 Doctor Antonio Amorim PeerJ Dear Dr. Amorim, Newly, we appreciate the valuable comments and suggestions provided by the reviewers and editor, which have led to improve our paper. To address such comments and suggestions, we have edited the manuscript and the detailed answers are provided below. We hope that the manuscript is now suitable for publication in PeerJ. Sincerely, Edna J. Márquez On behalf of the all authors Editor comments (Antonio Amorim) MAJOR REVISIONS Some major issues remain to be solved, besides those presented by the reviewers: - I could not find data/results regarding the validation of the claimed species specificity of the markers; in case this was not properly assessed, as I fear, the claim should be omitted from the title and throughout the text. Our study analyzed only samples from Magdalena-Cauca and Caribbean basins, which were collected by Integral S.A. and the Universidad del Magdalena. The morphological diagnostic based on taxonomic keys of Castro and Vari (2004), and performed independently by both institutions, assigned all samples to Prochilodus magdalenae. The genomic library was constructed from a single specimen of P. magdalenae and sequenced using Illumina MiSeq v2 platform (Illumina, California, USA). Samples of other species of Prochilodus, were not included in our analysis of population genetics. Thus, the transferability of these new microsatellite markers developed de novo, remain to explore in samples of other congeners of P. magdalenae, including P. reticulatus. - I could not find any details on the taxonomic identification of the samples in M&M section. The morphological diagnostic based on taxonomic keys of Castro and Vari (2004), and performed independently by both institutions, assigned all samples to Prochilodus magdalenae. For clarity, now we included the following sentence: “The morphological diagnostic of individuals of P. magdalenae was conducted following the taxonomic keys of Castro & Vari (2004) and performed independently by both institutions.” Castro R., Vari RP. 2004. Detritivores of the South American fish family Prochilodontidae (Teleostei: Ostariophysi: Characiformes): A phylogenetic and revisionary study. Smithsonian Contributions to Zoology 622:1–190. DOI: 10.5479/si.00810282.622. - My reservations on HW analyses remain; in particular: how can you justify to have selected loci that «showed Hardy-Weinberg disequilibrium» (lines 226-227) and «the analysis across loci showed significant departures from Hardy-Weinberg equilibrium expectations in all rivers evaluated» (l. 238-9)? In the absence of a comparison of observed and expected genotype distributions (per maker and population) it is impossible to infer the possible causes Due to the microsatellite loci were assessed in a random sample, Hardy-Weinberg equilibrium is not expected. The departures from HWE are also expected in samples with observed heterozygosity significantly lower than expected heterozygosity in HWE. The Table 2 shows the statistical significance for tests of departure from Hardy-Weinberg equilibrium per locus, across loci and per population. The causes of deficit of observed heterozygosity may be technical (silent alleles) or biological (Wahlund effect, inbreeding coefficient, assortative mating). Because we sequenced the genome of a single specimen, we do not have direct evidences of the potential variations in the primer alignment sequences. Consequently, we modify the paragraph. Now: “Most of the loci showed departures from Hardy-Weinberg equilibrium and significant observed heterozygosity deficit in the random samples. The observed heterozygosity deficit may be related to technical problems such as silent alleles; however, it remains to explore the potential variations in the primer alignment sequences, since this study sequenced the genome of a single specimen of P. magdalenae. Two non-excluding explanations may be related to significant levels of inbreeding and the genetic structure of the samples by the coexistence of two genetic stocks (see below).” The microsatellite loci developed in this study allowed in short-term to study samples collected and preserved in alcohol before 2014, information that might be used as reference to monitor potential changes in the genetic diversity in a basin that experience strong anthropic activities. - I think when the «deficit of heterozygosity» (e.g.., l. 251) is mentioned, it should correctly be stated as «deficit of observed heterozygotes». Done - Phylogeny (based on mtDNA only, which means reflecting just maternal lineages) is also troubling: how come that some P. reticulatus are clustered with P. magdalenae (and that «genetic distances (Table S7) were larger among haplotypes of P. magdalenae (0.002 – 0.014) than among P. magdalenae and P. reticulatus haplotypes (0.002 – 0.010).»? This casts serious doubts on the species assignment among this group (N.B: in connection with my first criticism on the claimed specificity. ). Below it is stated that there is a «proposal that P. magdalenae and P. reticulatus represent only one species». How to conciliate/solve this issue? Hybridiztion?. We review the sequences, construct a new tree and recalculate the genetic distances. Now: K2p P. magdalenae = 0.002 – 0.010; P. magdalenae – P. reticulatus: 0.000 – 0.005. The results are similar to those previously found although now the genetic distances are lower. Prochilodus magdalenae and P. reticulatus are identified by morphological characteristic although they are not differentiable by the molecular analyses performed so far. The lack of molecular differentiation between P. magdalenae and P. reticulatus was previously evidenced by Melo et al., (2016) using three mitochondrial (16S, COI, cytb) and three nuclear (myh6, rag1, rag2) genes. These authors indicated “Though P. magdalenae and P. reticulatus differ in the ranges and modal numbers of scales and vertebrae and have allopatric distributions (Castro and Vari, 2004), the lack of molecular differentiation results in a paraphyletic P. magdalenae in our analysis. The substantial morphological and molecular similarities between these species and the limited scale of their morphological differences suggest that they may represent only one species with perhaps allopatric differentiation resulting from the uplift of the Sierra del Perijá which separates the populations”. The lack of differentiation at the nuclear and mitochondrial levels limit to test the hypothesis of hybridization. Additionally, highly diverse haplotypes of P. magdalenae were also previously found by Aguirre-Pabón et al., 2013, using control region (Please see Table 2, Aguirre-Pabón et al., 2013). Despite of this high diversity, the haplotypes of P. magdalenae are clustered in a single group. The lack of molecular differentiation between P. magdalenae and P. reticulatus a well as the high haplotypic diversity of P. magdalenae, evidence evolutionary process and taxonomic issues that deserves future additional studies that are out of the scope of the present study. Due to the restocking of natural populations of P. magdalenae have been conducted for more than 20 years with genetically unknown materials, we performed a phylogenetic analysis of our samples to clarify the role of the taxonomic differences in the genetic structure observed. Consequently, we compare our sequences of P. magdalenae with available GenBank sequences of congeners distributed in Colombia (P. reticulatus, P. marie and P. nigricans), which must have more probability to contaminate the natural stocks of P. magdalenae. The taxonomic causes might be plausible because the orange genetic stock masked the genetic differences among Magdalena, Atrato and Sinú rivers (Figure 2), isolated ca. 5.28 mya by the uplift of the Western Cordillera that caused the isolation of the trans-Andean Atrato-Pacific slope from the Magdalena basin during Late Miocene to Pliocene (Kellogg & Vega, 1995). Aguirre-Pabón J., Narváez Barandica J., Castro García L. 2013. Mitochondrial DNA variation of the bocachico Prochilodus magdalenae (Characiformes, Prochilodontidae) in the Magdalena River Basin, Colombia. Aquatic Conservation: Marine and Freshwater Ecosystems 23:594–605. DOI: 10.1002/aqc.2339. Castro R., Vari RP. 2004. Detritivores of the South American fish family Prochilodontidae (Teleostei: Ostariophysi: Characiformes): A phylogenetic and revisionary study. Smithsonian Contributions to Zoology 622:1–190. DOI: 10.5479/si.00810282.622. Kellogg J., Vega V. 1995. Tectonic development of Panamá, Costa Rica, and the Colombian Andes: Constraints from Global Positioning System geodetic studies and gravity. Special Papers - Geological Society of America 295:75–90. Melo BF., Sidlauskas BL., Hoekzema K., Frable BW., Vari RP., Oliveira C. 2016. Molecular phylogenetics of the Neotropical fish family Prochilodontidae (Teleostei: Characiformes). Molecular Phylogenetics and Evolution 102:189–201. DOI: 10.1016/j.ympev.2016.05.037. - I disagree with the conclusion: «both heterologous and species-specific microsatellite loci revealed a general deficit of heterozygotes in all samples, indicating that its causes are biological rather than technical.» In fact the Authors list a series of problems for the so-called ‘heterologous’ microsatellites which also apply to the claimed ‘species.-specific’. I think my previous objections were not removed, as I could not find any specific approach to elucidation of the cause of the deficit (inbreeding /substructure vs. silent alleles/technical problems). Now: “Most of the loci showed departures from Hardy-Weinberg equilibrium and significant observed heterozygosity deficit in the random samples. The observed heterozygosity deficit may be related to technical problems such as silent alleles; however, it remains to explore the potential variations in the primer alignment sequences, since this study sequenced the genome of a single specimen of P. magdalenae. Two non-excluding explanations may be related to the significant levels of inbreeding and the genetic structure of the samples by the coexistence of two genetic stocks (see below).” Minor corrections - Please confirm if what is meant by «Results of the genetic bottleneck (Table 4)» is: Results of the genetic bottleneck tests; if so, correct accordingly. Done. - In the supplemental file «Raw sequence contigs, longitude and some identification parameters of 52 microsatellite loci selected for Prochilodus magdalenae» by ‘longitude’ you mean ‘length’? If so, modify accordingly. Done. Finally, since - as he Authors mention - there is protective legislation regarding this animal, forensic cases may arise, for the solution of which this type of markers may be extremely valuable, provided they are validated. Accordingly, I advise the Authors to explicitly consider this application, consulting the relevant literature, namely from specialized journals such as Forensic Science International Genetics (and the Supplement Series), where relevant papers were recently published and recommendations/guidelines on animal forensic genetics issued. Now, in the discussion regarding microsatellite loci development: “In addition to the applications in harvest management, stocking programmes, definition of conservation units, recovery of threatened species, and management of invasive species, these tools may be useful in forensic genetics since partially degraded DNA samples are often found in this area (see Bourret et al., 2020).” Reviewer 1 (Gabriel Yazbeck) Basic reporting The authors have full, direct and satisfactorily addressed all remarks, questioning and suggestions I have made in the first round of revision. They have added a profuse and detailed suite of supplemental data that will certainly enrich the paper’s open virtual environment presentation. I also take the opportunity to thank back the authors for their acknowledgement and I hope they will forgive me, though, and cope with new little remarks and details I have freshly raised in this second round or review, of elements already present in the original manuscript, along one main important point: -Although there are no rules of name translation, throughout the text, tables and supplemental material, do prefer the use of the English term “the Caribbean” instead of “Caribe” (e.g. L18, L94 and Supplemental Table 1); Done. - Line 92 Since the objective deals with an atemporal aspect of the study (what it does) prefer the used of the present tense instead the past tense: 'this study tests the hypothesis'; Done. Experimental design No new comments; Validity of the findings No new comments; Comments for the Author My only point of concern is that the paper devalues itself while not making clear in the discussion and in the conclusion that these are indeed the FIRST species-specific microsatellite DNA markers published for this fish species (see instances on the annotated review manuscript), at least to the best of my knowledge. A search of GenBank revealed a dubious claim of 17 P. magdalenae species-specific microsatellites (implicit in the title of an unpublished manuscript – e.g. entries MH586829 through MH586842) for what actually seem to be other already published microsatellite loci for P. lineatus, P. argenteus and P. costatus. If the authors arrive to the conclusion this here is the first set of microsatellite markers for P. magdalenae, it certainly needs to be brought to light in the final work. The GenBank accessions MH586829 - MH586842 represent sequences obtained by Sanger method of clones named similarly to the microsatellite loci developed for Prochilodus species indicated by the reviewer (P. lineatus, P. argenteus and P. costatus). We have sent various E-mails to the first author Paul Bloor since July 26th to clarify this issue, but we have not got any response until now. Consequently, in the absence of a published work that support this assumption, we prefer conserve the current version of the discussion although our study probably it is the first set of species-specific microsatellite loci for P. magdalenae. Annotated manuscript The reviewer has also provided an annotated manuscript as part of their review: Download Annotated Manuscript Reviewer 2 (Carlos Henrique Santos) Basic reporting The present study shows great improvements after the suggested revisions and has relevant information for the conservation of the species Prochilodus magdalenae. The information presented here is clear and the article presents an adequate structure. Figures and tables (supplementary material) bring information that strengthens the content presented in the paper. The references are adequate for the discussion of the results, however, I felt a lack of more current references from the last 3 years (2017-2020). The paper presents only 7.8% of references for the years 2017-2020. We include new references: Bourret V, Albert V, April J, Côté G, Morissette O. 2020. Past, present and future contributions of evolutionary biology to wildlife forensics, management and conservation. Evolutionary Applications 13:1420–1434. DOI: 10.1111/eva.12977. Ribolli J, Miño CI, Scaranto BMS, Reynalte-Tataje DA, Zaniboni Filho E. 2020. Genetic evidence supports polygamous mating system in a wild population of Prochilodus lineatus (Characiformes: Prochilodontidae), a Neotropical shoal spawner fish. Neotropical Ichthyology 18(2):e190123. DOI: 10.1590/1982-0224-2019-0123. Experimental design The research presented here is adequate to the scope and objective of the journal. The paper's objective is well defined and relevant to the study area. The methodology applied is adequate and describes in detail the study. Validity of the findings The underlying data has been provided (supplementary material) and they are robust for a better understanding of the results. The conclusions are very clear from the study proposal and linked to the research question. Comments for the Author I congratulate the authors for the importance of the study and its contribution to the management and conservation of the species Prochilodus magdalenae. It would be interesting if the authors had applied a greater number of microsatellite loci, as well as some mitochondrial genes for further investigation. However, the work brings relevant information and contribution to the study area. I highlight here the sample size to the number of microsatellite markers used in the study. The review of the paper brought improvements to the study. We appreciate the suggestions for future directions. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Apoptosis accounts for eliminating damaged and virus-infected cells, normal cell turnover, proper development and functioning of the immune system. Caspases play a vital role in apoptosis in both mammals and invertebrates. Spodoptera littoralis is a generalist insect herbivore attacking a wide range of commercially important crops and is one of the most destructive pests in the tropical and subtropical areas of the world.</ns0:p><ns0:p>Spodoptera littoralis is a model organism in studying baculovirus infection. However, the apoptotic pathway in Spodoptera littoralis remains unclear. Methods. We cloned a new caspase gene named sldronc in Spodoptera littoralis using Rapid Amplification of cDNA Ends (RACE). Caspase activity on synthetic caspase substrates and effector caspase of SlDronc was measured. The function of SlDronc in apoptotic pathway and interaction with caspase inhibitors were also tested in SL2 cell. Results. We identified an initiator caspase, SlDronc, in Spodoptera littoralis. SlDronc cleaved and activated effector caspases.</ns0:p><ns0:p>Overexpression of SlDronc induced apoptosis in SL2 cell, and knockdown of Sldronc decreased apoptosis induced by UV irradiation in SL2 cell. The above results indicate that SlDronc is an apoptotic initiator caspase in Spodoptera littoralis. In addition, we also find that processed forms of SlDronc were increased with present of N-terminally truncated SlIAP and that SlDronc was inhibited by P49. The present study makes contributions to clearly elaborating apoptotic pathway in Spodoptera littoralis and may facilitate the study of baculovirus infection induced apoptosis.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Apoptosis is a form of strictly controlled programmed cell death which accounts for eliminating of needless cells including impaired and virus-infected cells <ns0:ref type='bibr' target='#b2'>(Clem 2001;</ns0:ref><ns0:ref type='bibr' target='#b16'>Duprez et al. 2009)</ns0:ref>. Apoptosis plays a vital role in normal cell turnover, proper development and functioning of the immune system <ns0:ref type='bibr' target='#b17'>(Elmore 2007)</ns0:ref>. Apoptosis mainly proceeds through intrinsic and extrinsic pathways, both of which converge on activation of the caspases <ns0:ref type='bibr' target='#b8'>(Degterev et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b21'>Hay &amp; Guo 2006;</ns0:ref><ns0:ref type='bibr' target='#b49'>Tait &amp; Green 2010)</ns0:ref>. Caspases are cysteinedependent aspartate-directed proteases which cleave numerous specific target sites in cellular proteins after activation by upstream signals <ns0:ref type='bibr' target='#b54'>(Wallach et al. 2016)</ns0:ref>. According to the localization in the apoptotic pathway and biological functions, caspases can be divided into PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed several groups, which includes initiator caspase, effector caspase, inflammatory caspase and caspase without known function <ns0:ref type='bibr' target='#b3'>(Cohen 1997;</ns0:ref><ns0:ref type='bibr' target='#b18'>Galluzzi et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b31'>Li &amp; Yuan 2008)</ns0:ref>.</ns0:p><ns0:p>Caspase is synthesized as inactive zymogen (pro-caspase), the pro-caspase has a prodomain in N-terminal and a conserved catalytic domain in C-terminal, which is consisted of a large subunit plus with a small subunit. During apoptosis, the pro-caspase is cleaved between prodomain and large subunit, the cleavage between large subunit and small subunit is not essential for activation of the caspase. A heterodimer was formed by a large subunit plus with a small subunit, and then an active unit of tetramer was formed by two heterodimers.</ns0:p><ns0:p>Initiator caspase generally has a long prodomain containing a CARD (caspase recruit domain) or DED(s) (death effector domain), CARD and DED can combine with adapter proteins that locates in the upstream of initiator caspase in apoptotic pathway through their homologous motifs. While effector caspase has a short prodomain without DED or CARD domain <ns0:ref type='bibr' target='#b33'>(Lo et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b43'>Salvesen &amp; Abrams 2004)</ns0:ref>. Initiator caspase is activated by dimerization that is facilitated through recruiting caspases to oligomeric activation platforms that assemble following an apoptotic signal <ns0:ref type='bibr' target='#b13'>(Donepudi et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b15'>Dorstyn &amp; Kumar 2008;</ns0:ref><ns0:ref type='bibr' target='#b39'>Pop et al. 2006)</ns0:ref>. The activated initiator caspase activate Effector caspase through proteolytic cleavage <ns0:ref type='bibr' target='#b37'>(Pop &amp; Salvesen 2009)</ns0:ref>.</ns0:p><ns0:p>In Drosophila melanogaster, a model organism in studying of insect apoptosis, seven caspases were reported. Dronc and Dredd are initiator caspases whose pro-domain have one CARD and two DEDs respectively, Drice, Dcp-1, Decay and Damm are effector caspases <ns0:ref type='bibr' target='#b4'>(Cooper et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b28'>Kumar &amp; Doumanis 2000)</ns0:ref>. Dronc acts as upstream caspase for the intrinsic pathway and the mechanism of its activation resembles the activation mechanism of PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed mammalian caspase-9 <ns0:ref type='bibr' target='#b14'>(Dorstyn et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b36'>Muro et al. 2004</ns0:ref>). Lepidopteran caspases are classified into 6 clades, among them Lep-Caspase-1, Lep-Caspase-2 and Lep-Caspase-3 are supposed effector caspases, Lep-Caspase-5 and Lep-Caspase-6 are supposed initiator caspases. Dronc belongs to the Lep-caspase-5 clade <ns0:ref type='bibr' target='#b5'>(Courtiade et al. 2011)</ns0:ref>. Dronc homologs from several Lepidopteran insects have been reported successively, including Bombyx mori (BmDronc), Lymantria dispar (LdDronc) and Spodoptera frugiperda (SfDronc) <ns0:ref type='bibr' target='#b23'>(Huang et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kitaguchi et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b47'>Suganuma et al. 2011)</ns0:ref>.</ns0:p><ns0:p>Apoptosis is regulated by multiple celluar proteins, among them IAPs function as a last line of defense against caspase-mediated apoptosis. IAPs can inhibit caspases by directly binding to them with BIR domains or ubiquitylating caspases with the RING domain after binding to them <ns0:ref type='bibr' target='#b9'>(Deveraux et al. 1997;</ns0:ref><ns0:ref type='bibr'>Ditzel et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b42'>Roy et al. 1997)</ns0:ref>. While IAPs need processing by caspases to be in working order, such as DIAP1 requires caspase-mediated cleavage for its functioning, drICE cleaves N-terminal 20 amino acids of DIAP1 to activate its ability to suppress apoptosis, and C-terminal of DIAP1 is degraded by N-end rule degradation pathway <ns0:ref type='bibr' target='#b12'>(Ditzel et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b56'>Yan et al. 2004</ns0:ref>). The N-end rule pathway is a proteolytic system depending on proteasome, it recognizes and degrades proteins which have N-degrons <ns0:ref type='bibr' target='#b19'>(Gibbs et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b51'>Tasaki et al. 2012)</ns0:ref>.</ns0:p><ns0:p>Apoptosis is also regulated by inhibitors in baculovirus, P35 in AcMNPV (Autographa californica multiple nucleopolyhedrovirus) and P49 in SpliNPV (Spodoptera littoralis nucleopolyhedrovirus) are two baculoviral apoptosis inhibitors, in generally, baculoviral apoptosis inhibitor P49 inhibits caspase activity of initiator caspases and baculoviral apoptosis inhibitor P35 inhibits caspase activity of effector caspases <ns0:ref type='bibr' target='#b25'>(Jabbour et al. 2002;</ns0:ref><ns0:ref type='bibr' /> PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr' target='#b58'>Zoog et al. 2002)</ns0:ref>.</ns0:p><ns0:p>Spodoptera littoralis is a generalist insect herbivore attacking a wide range of commercially important crops, including cotton, rice, maize and potato <ns0:ref type='bibr' target='#b29'>(Lee &amp; Anstee 1995)</ns0:ref>. Spodoptera littoralis distributes throughout Africa, the Mediterranean region and the Near East, and is one of the most destructive pests in the tropical and subtropical areas of the world <ns0:ref type='bibr' target='#b22'>(Hill 1987)</ns0:ref>. SL2 cell derived from Spodoptera littoralis and Sf9 cell derived from Spodoptera frugiperda are often used in studying baculovirus infection and apoptosis <ns0:ref type='bibr' target='#b34'>(Mialhe et al. 1984)</ns0:ref>. SL2 cell undergoes apoptosis and produce very low levels of polyhedrin in contrast to Sf9 cell when infected by AcMNPV <ns0:ref type='bibr' target='#b1'>(Chejanovsky &amp; Gershburg 1995)</ns0:ref>, which suggests that differences could be existed in the mechanism of apoptosis in SL2 cell and Sf9 cell.</ns0:p><ns0:p>However, several years have passed since the first effector caspase, Sl-caspase-1, and the inhibitor of apoptosis, SlIAP, from Spodoptera littoralis were reported <ns0:ref type='bibr' target='#b32'>(Liu et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b53'>Vilaplana et al. 2007)</ns0:ref>, none initiator caspase has been identified so far and a scanty fewarticles expounding mechanism of apoptosis in Spodoptera littoralis. In this study, we identified an initiator caspase, SlDronc, in Spodoptera littoralis. The amino acid sequences of the SlDronc was analysed and the biochemical character was tested, the function of SlDronc was verified in SL2 cell. This present study makes contributions to clearly elaborating apoptotic pathway in Spodoptera littoralis and may facilitate the study of baculovirus infection induced apoptosis.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Cells</ns0:head></ns0:div> <ns0:div><ns0:head>SL2 cells were kindly gifted by professor Nor Chejanovsky (Agricultural Research</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Organization, The Volcani Center, Israel). SL2 cells were cultured using Grace medium (Invitrogen) at 27 &#186;C in biochemical incubator, and heat-inactivated FBS (Gibco) was added into the Grace medium by 10% (v/v).</ns0:p></ns0:div> <ns0:div><ns0:head>Antibodies</ns0:head><ns0:p>Rat-derived monoclonal antibodies against His-tag, Flag-tag, HA-tag and &#946;-actin (Proteintech) were diluted by 1:5000 in block buffer when used in western blot analysis.</ns0:p><ns0:p>Rabbit-derived polyclonal antibody against Sf-caspase-1, which can recognize full length, large subunit of Sf-caspase-1 and Sl-caspase-1, obtained from professor Nor Chejanovsky, was diluted by 1:1000 in block buffer when used in western blotting. Polyclonal antibody against SlDronc, which can recognize full length and large subunit of SlDronc, was produced using a SlDronc fragment purified in E. coli as antigen to immunize rabbit. Polyclonal antibody against SfIAP, which can also recognize full length, cleaved SlIAP, was produced using a SfIAP fragment purified in E. coli as antigen to immunize rabbit.</ns0:p></ns0:div> <ns0:div><ns0:head>Cloning of Spodoptera littoralis dronc</ns0:head><ns0:p>Sldronc was firstly cloned as a partial sequence using the primers designed according to the alignment of dronc homologs from Spodoptera frugiperda (Sfdronc) and Spodoptera litura (Sl-caspase-5). The sequence contained an intact open reading frame (ORF) which was highly similar to Sfdronc. To obtain full length of Sldronc, primers were designed for Rapid Amplification of cDNA Ends (RACE) depending on the obtained partial sequence of Sldronc (Table <ns0:ref type='table'>1</ns0:ref>). SMARTer&#8482; RACE cDNA Amplification Kit (Clontech, CA, USA) was used in the 5' and 3' RACE to amplify coding sequence and untranslated regions (UTR) of Sldronc. Purified PCR products were ligated into the pCR-II vector (TA Cloning&#174; Kit;</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Invitrogen, CA, USA) and then plasmids extracted from several positive colonies were sequenced. Several sequencing results assemble a sequence having a 1338 bp ORF with 216 bp 5'UTR and 397 bp 3'UTR. This ORF was named Sldronc and was cloned into pCR-II vector from Spodoptera littoralis cDNA. Plasmids extracted from several positive colonies were sequenced, which confirmed the sequence information from RACE.</ns0:p></ns0:div> <ns0:div><ns0:head>Construction of plasmids</ns0:head><ns0:p>Plasmid pET28a-SlDronc-C-His expressing SlDronc with a His-tag at C-terminal was constructed by ligating the ORF of SlDronc into Nco I and Hind III sites in vector pET-28a.</ns0:p><ns0:p>Plasmids pET28a-Sl-caspase-1-C-His and pET28a-P49-(GS)3-C-His were used for expression of Sl-caspase-1 and P49 containing a C-terminal His-tag in E. coli. Plasmid pIE1-SlDronc-C-Flag was used for overexpression of SlDronc containing a C-terminal Flagtag in SL2 cells. Plasmid pIE1-N-HA-SlDronc was used for overexpression of SlDronc containing a N-terminal HA-tag in SL2 cells. Plasmid pIE1-P49-C-His was used for overexpression of P49 with a C-terminal His-tag in SL2 cells. Plasmids pIE1-SlIAP-C-Flag was used for overexpression of SlIAP with a C-terminal Flag-tag in SL2 cells. SlDronc C310A, Sl-caspase-1 C178A, SlIAP D87A and SlIAP N88G were constructed through introducing point mutation into wild type plasmids by site-directed mutagenesis PCR, the forward and reverse primers are listed in Table <ns0:ref type='table'>2</ns0:ref>. DNA sequencing was always used to confirm constructed plasmids.</ns0:p></ns0:div> <ns0:div><ns0:head>Purification of recombinant proteins</ns0:head><ns0:p>E. coli BL21 (DE3)/pLysS cells transfected by plasmid expressing interest protein or vector were cultured in LB medium with 50 &#956;g/mL kanamycin to a concentration that OD600 nm fall PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed in 0.4 and 0.6, expression of interest proteins were induced by 0.2 mM isopropyl-&#946;-Dthiogalactopyranoside (IPTG). and then cells were cultured for another 2 h at 200 rpm. 20 mM imidazole solution with 1% detergent Triton X-100 and protease inhibitor (Roche) was used to resuspend BL21 cells after centrifugation, then suspension was sonicated 4 second with 6 second interval for totally 30 min and 10-min interval between every 10 min. At 4 &#186;C, lysate of BL21 cells was centrifugated at 14,000 g for 30 min, Ni-NTA high affinity resin (Genscript) was incubated with the supernatant as protocol supplied manufacturer. 20 mM, 50 mM, 80 mM imidazole solution were used to wash the resin and 250 mM imidazole solution was used to elute interest protein. After quality analysis using western blotting and Coomassie staining, the purified interest proteins were stored in -80 &#186;C.</ns0:p></ns0:div> <ns0:div><ns0:head>SDS-PAGE and western blotting</ns0:head><ns0:p>Samples for SDS-PAGE were prepared by mixing purified proteins or cell lysates with 5&#215;SDS loading buffer, and then heated in boiling water for 10 min. Proteins were detected by Coomassie blue staining or western blotting using PVDF membrane (Merck Millipore).</ns0:p><ns0:p>5% milk or BSA in TBST was used to block the membranes for 1 h at room tempreture, and then primary antibodies diluted in Block buffer were incubated with the membranes for another one hour. After washing with TBST three times, HRP-conjugated secondary antibodies (Thermo) diluted in Block buffer were incubated with the membranes for 1 h. LAS 4000 (Fujifilm) was used to detect protein-primary antibody-secondary antibody-HRP complex after incubating the membranes with chemiluminescent substrate (Millipore).</ns0:p></ns0:div> <ns0:div><ns0:head>Caspase activity assay</ns0:head><ns0:p>In order to analysis substrate preference of recombinant SlDronc purified in E. coli, 13 kinds PeerJ reviewing <ns0:ref type='table'>PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:ref> Manuscript to be reviewed of fluorogenic synthetic caspase substrates were used, which are Ac-VEID-AFC, Ac-LETD-AFC, Ac-IETD-AFC, Ac-LEHD-AFC, Ac-AEVD-AFC, Ac-WEHD-AFC, Ac-DEVD-AFC, Ac-YVAD-AFC, Ac-DMQD-AFC, Ac-LEED-AFC, Ac-IEPD-AFC, Ac-VDVAD-AFC and Ac-LEVD-AFC (MP). The activities on Ac-DEVD-AFC of Sl-caspase-1 and the cell lysates were also measured. In brief, purified caspase, incubate or cell lysate was mixed relevant fluorogenic substrate in Na-Citrate buffer (1 M Na-Citrate, 50 mM Tris-base, 10 mM DTT, 0.05% CHAPS, pH 7.4) , and 20 &#956;M substrate was used in each 100 &#956;L test mixture <ns0:ref type='bibr' target='#b38'>(Pop et al. 2008)</ns0:ref>. Relative fluorescence unit was detected at 37 &#186;C every 2 min for 2 h after incubation at 37 &#186;C for 30 min. Maximum slope of each curve were calculated using the data, and the graph was generated by GraphPad Prism 6.</ns0:p></ns0:div> <ns0:div><ns0:head>Transfection</ns0:head><ns0:p>SL2 cells were transfected using the method previously described <ns0:ref type='bibr' target='#b48'>(Summers &amp; Smith 1987)</ns0:ref>.</ns0:p><ns0:p>Briefly, SL2 cells were seeded into a culture plate at a cell abundance of about 70%. After attachment period, the medium was removed and a mixture of plasmid, Transfection Buffer (25 mM HEPES, pH 7.1, 140 mM NaCl, 125 mM CaCl2) and Grace medium was added dropwise into the well with SL2 cells, then the cells were maintained at 27 &#186;C for 4 hours before replacement of the transfection mixture by fresh Grace medium. 3.0 &#956;g plasmid was used per well in 12-well plate.</ns0:p></ns0:div> <ns0:div><ns0:head>UV irradiation</ns0:head><ns0:p>SL2 cells were treated with UV irradiation by placing cell culture plate on a transilluminator for 45 min. 24 hours after UV treatment, cells were harvested and lysed for later analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Cell lysate preparation</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed SL2 cells were harvested by centrifugation of culture medium at 1,000 g for 10 min, cell lysates were prepared through following procedure. Lysis buffer (200 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP-40) with protease inhibitor (Roche) was used to suspend cells <ns0:ref type='bibr' target='#b57'>(Yang et al. 2016)</ns0:ref>. After 3 times of freeze and thaw, centrifuging suspension using 14,000 g at 4 &#186;C for 10 min. Taking the supernatants as cell lysates for analysis. When harvesting apoptosis SL2 cells, firstly cells were harvested by 3,000 g at 4 &#186;C for 10 min, and secondly supernatant was centrifuged to collect apoptosis bodies at 14,000 g at 4 &#186;C for 20 min. Harvested cells and apoptosis bodies were gathered to prepare cell lysate through the routine method.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene knockdown</ns0:head><ns0:p>DNA Fragments of interest gene were cloned utilizing primers (Table <ns0:ref type='table'>3</ns0:ref>) containing T7 promoter in 5' end and plasmids or cDNAs containing desire sequences as templates. RNAs were transcribed in vitro using the cloned DNA fragments. After removing of the template DNA by adding TURBO DNase, RNAs were extracted and purified by phenol-chloroform.</ns0:p><ns0:p>To obtained dsRNAs, RNA products were slowly cooled to room temperature after incubating at 95 &#186;C for 2 min. The NanoDrop one (Thermo Fisher Scientific Inc.) was used to determined qualities and concentrations of dsRNAs. 1 &#956;g dsRNA was used in each 6&#215;10 5 SL2 cells for knocking down desire gene. The dsRNA transcribed from gfp gene was utilized as a negative control.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Sequence analysis of SlDronc</ns0:head><ns0:p>A novel caspase cDNA supposed to be Sldronc was acquired by RT-PCR using total mRNA PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed from SL2 cells as template. Predicted protein encoded by the Sldronc ORF contains 445 amino acids (about 51 kDa) including a 15 kDa prodomain, a 24 kDa large subunit and a 12 kDa small subunit. Alignment of predicted amino acid sequence of SlDronc with Dronc homologs inSpodoptera frugiperda (SfDronc), Bombyx mori (BmDronc), Aedes aegypti (AeDronc), and Drosophila melanogaster (DmDronc) showed that SlDronc has 87%, 54%, 27%, 24% consistency ratio with SfDronc, BmDronc, AeDronc, DmDronc, respectively (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). Predicted SlDronc has the typical caspase catalytic site sequence Q 308 MCRG 312 (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>). Secondary structure of the predicted SlDronc contains a series of &#945;-helices and &#946;sheets which are highly conserved with Droncs in other insects <ns0:ref type='bibr' target='#b55'>(Watt et al. 1999</ns0:ref>) (Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>).</ns0:p><ns0:p>SlDronc possesses a long prodomain containing a CARD, E128 is predicted to be cleavage site between prodomain and large subunit and D338 is predicted to be the cleavage site between the large and small subunit referring to verified cleavage sites in other Droncs(Fig. <ns0:ref type='figure' target='#fig_2'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Phylogenetic analysis of SlDronc</ns0:head><ns0:p>Phylogenetic analysis of SlDronc and 30 chosen caspases in several insects (Table <ns0:ref type='table'>4</ns0:ref>) suggests that SlDronc vests in Dronc homolog clade, and Sl-caspase-5, Se-caspase-5, SfDronc and Ha-caspase-5 in lepidopteran insects are the closest relatives of SlDronc (Fig. <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>SlDronc underwent autocatalytic cleavage in E. coli</ns0:head><ns0:p>Caspases usually autoprocess themselves when they are brought into close proximity to each other in the 'induced-proximity' model <ns0:ref type='bibr' target='#b41'>(Riedl &amp; Shi 2004;</ns0:ref><ns0:ref type='bibr' target='#b44'>Salvesen &amp; Dixit 1999)</ns0:ref>.</ns0:p><ns0:p>Consisting with the 'induced-proximity' model, three major bands about 52 kDa, 38 kDa PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed and 13 kDa were detected during recombinant SlDronc with C-terminal His-tag was expressed and purified in E. coli and detected by western blotting (Fig. <ns0:ref type='figure' target='#fig_4'>3</ns0:ref>). The 52-kDa band is supposed to be the full length of SlDronc plus C-terminal His-tag (Full length+His). The 38-kDa band is supposed to be the fragment consisting of the prodomain and large subunit (Pro+LS). The 13-kDa band matched the supposed fragment consisting of the small subunit and the C-terminal His-tag (SS+His). The two forms of cleaved bands indicate that SlDronc underwent autocatalytic cleavage during expressed and purified in E. coli. In order to further test whether the detected autocleavage of wild type SlDronc relies on its caspase activity, cysteine (C) at 310 site was mutated to alanine (A) in the predicted catalytic site of SlDronc.</ns0:p><ns0:p>The catalytic site mutant SlDronc C310A was synchronously expressed and purified in E. coli and detected by western blotting (Fig. <ns0:ref type='figure' target='#fig_4'>3</ns0:ref>). Different from the wild type SlDronc, SlDronc C310A did not autoprocess between the large subunit and small subunit, one major band supposed to be full length of SlDronc was detected in purified SlDronc C310A boring a C-terminal His-tag. All above results indicate that caspase activity of SlDronc was essential for the observed autocatalytic cleavage. Taken together, SlDronc possessed caspase activity.</ns0:p></ns0:div> <ns0:div><ns0:head>Recombinant SlDronc possessed strong activity on substrates of initiator caspase</ns0:head><ns0:p>Synthetic caspase substrates have been widely used to identify caspase activity <ns0:ref type='bibr' target='#b38'>(Pop et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b50'>Talanian et al. 1997)</ns0:ref> Manuscript to be reviewed LEHD-AFC, Ac-AEVD-AFC, Ac-WEHD-AFC and Ac-DEVD-AFC (Fig. <ns0:ref type='figure'>4A</ns0:ref>), and SlDronc C310A with a C-terminal His-tag (Fig. <ns0:ref type='figure'>4B</ns0:ref>) showed no activity on the selected substrates.</ns0:p><ns0:p>Mutation of the catalytic site led to absolutely loss of SlDronc activity, which suggests that the activities on caspase substrates of the wild type SlDronc rely on its caspase activity.</ns0:p><ns0:p>Consisting with the fact that SlDronc shows a high sequence similarity with initiator caspase, SlDronc showed the strongest activity against Ac-VEID-AFC, a preferential substrate of initiator caspase. These above data suggest that SlDronc may be an initiator caspase.</ns0:p></ns0:div> <ns0:div><ns0:head>SlDronc cleaved and activated Sl-caspase-1 in vitro</ns0:head><ns0:p>As we all know, initiator caspase can cleave effector caspase to activate the effector caspase.</ns0:p><ns0:p>To investigate whether SlDronc could cleave recombinant Sl-caspase-1 (an effector caspase in Spodoptera littoralis), Sl-caspase-1 and Sl-caspase-1 C178A (catalytic site mutant) were expressed and purified in E. coli. In order to avoid autocatalytic cleavage of Sl-caspase-1, Sl-caspase-1 C178A was utilized in the experiment detecting cleavage by SlDronc.</ns0:p><ns0:p>Recombinant SlDronc was incubated with Sl-caspase-1 C178A at 37 &#186;C for 3 h, and incubation mixture was detected by western blotting using Anti-Sf-caspase-1 antibody. The result showed that Sl-caspase-1 C178A was cleaved by SlDronc, but not by SlDronc C310A (Fig. <ns0:ref type='figure' target='#fig_5'>5A</ns0:ref>), suggesting that SlDronc could cleave Sl-caspase-1 directly in vitro and the caspase activity of SlDronc is required. Also, incubation with SlDronc significantly increased the enzymatic activity of Sl-caspase-1 on Ac-DEVD-AFC (Fig. <ns0:ref type='figure' target='#fig_5'>5B</ns0:ref>), further confirming that SlDronc can activate Sl-caspase-1. Thus, SlDronc might function as an initiator caspase in the apoptotic pathway of Spodoptera littoralis.</ns0:p></ns0:div> <ns0:div><ns0:head>Overexpression of SlDronc induced apoptosis in SL2 cell</ns0:head><ns0:p>To examine whether SlDronc functioned as an apoptotic caspase, plasmids which express wild type SlDronc or the catalytic site mutant SlDronc C310A with a C-terminal Flag-tag were transfected into SL2 cells, and the cells were observed at multiple intervals. Untreated cells, and cells transfected with plasmid expressing C-terminally Flag-tagged GFP were used as controls. At 24 h post transfection, apoptosis was observed in SL2 cells expressed SlDronc, while only slight apoptosis was observed in SL2 cells expressed SlDronc C310A and no apoptosis was observed in mock treated cells or cells expressed GFP (Fig. <ns0:ref type='figure' target='#fig_6'>6A</ns0:ref>).</ns0:p><ns0:p>Moreover, lysate prepared from cells transiently expressed SlDronc showed significantly increased caspase activity on Ac-DEVD-AFC than cells transiently expressed GFP (Fig. <ns0:ref type='figure' target='#fig_6'>6B</ns0:ref>), that is consistent with the morphological result described above. Western blot analysis utilizing antibody against Flag-tag showed that the protein level of full length SlDronc was significantly less than that of SlDronc C310A, suggesting that wild type SlDronc was cleaved when transiently expressed in SL2 cells (Fig. <ns0:ref type='figure' target='#fig_6'>6C</ns0:ref>). Western blot analysis with antibody against Sf-caspase-1 showed that the protein level of full length Sl-caspase-1 was decreased in cells overexpressing SlDronc compared with that of cells overexpressing GFP (Fig. <ns0:ref type='figure' target='#fig_6'>6C</ns0:ref>). Taken together, these findings manifested that apoptosis in SL2 cells was induced by overexpression of SlDronc, and the caspase activity of SlDronc was required to induce apoptosis. Thus, SlDronc might be an apoptotic initiator caspase in SL2 cell.</ns0:p></ns0:div> <ns0:div><ns0:head>Knockdown of Sldronc decreased apoptosis induced by UV irradiation in SL2 cell</ns0:head><ns0:p>In order to verify the function of SlDronc in apoptosis, dsRNA was used to knock down Sldronc expression and the effect on apoptosis was investigated. First, the prepared Sldronc-PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed dsRNA was proved to knock down endogenous Sldronc expression at 24 h and 48 h after transfection successfully (Fig. <ns0:ref type='figure' target='#fig_7'>7A</ns0:ref>). After that, SL2 cells were transfected with Sldronc-dsRNA (dsRNA against Sldronc) for 24 h and followed by UV irradiation. SL2 cells transfected with gfp-dsRNA (dsRNA against gfp) and cells transfected with Sl-caspase-1-dsRNA (dsRNA against Sl-caspase-1) were used as controls. 24 h after UV treatment, less apoptosis was observed in cells transfected with Sldronc-dsRNA or Sl-caspase-1-dsRNA comparing with cells transfected with gfp-dsRNA (Fig. <ns0:ref type='figure' target='#fig_7'>7B</ns0:ref>). SL2 cells were harvested and then the cell lysates were subjected to caspase activity assay and western blotting.</ns0:p><ns0:p>Consisting with the morphologic result, caspase activity assay indicated that decreased activity on Ac-DEVD-AFC was shown in cells transfected with Sldronc-dsRNA or Slcaspase-1-dsRNA comparing with that of cells transfected with gfp-dsRNA when treated by UV irradiation (Fig. <ns0:ref type='figure' target='#fig_7'>7C</ns0:ref>). Western blotting results showed that knockdown of Sldronc significantly decreased cleavage of Sl-caspase-1 induced by UV irradiation treatment in SL2 cells (Fig. <ns0:ref type='figure' target='#fig_7'>7D</ns0:ref>). These results manifested that SlDronc was involved in apoptosis induced by UV irradiation in SL2 cell.</ns0:p></ns0:div> <ns0:div><ns0:head>Processed forms of SlDronc were increased with present of N-terminally truncated</ns0:head></ns0:div> <ns0:div><ns0:head>SlIAP</ns0:head><ns0:p>The inhibitor of apoptosis is a vital regulator of apoptosis, it is known to suppress caspasemediated apoptosis in Drosophila melanogaster, Bombyx mori and Spodoptera frugiperda <ns0:ref type='bibr' target='#b20'>(Hamajima et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kaiser et al. 1998;</ns0:ref><ns0:ref type='bibr' target='#b35'>Muro et al. 2002)</ns0:ref>. To test effect of SlIAP on SlDronc, SlIAP was coexpressed with SlDronc in SL2 cells. When SlDronc containing Cterminal Flag-tag was co-expressed with SlIAP, western blotting utilizing antibody against PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Flag-tag only detected decreased full length SlDronc, but no processed SlDronc. Meanwhile, in order to distinguish Pro+LS (39 kDa) and LS+SS (36 kDa) detected by western blotting utilizing anti-SlDronc antibody, N-terminally HA-tagged SlDronc was co-expressed with Cterminally Flag-tagged SlIAP in SL2 cells. Western blotting using antibody against SlDronc showed that the protein levels of processed SlDronc were increased while protein level of full length SlDronc was decreased when co-expressed with SlIAP, the processed form of 39 kDa corresponds to the form of SlDronc composed of the prodomain and large subunit (Pro+LS), and the processed form of 24 kDa corresponds to the large subunit of SlDronc (LS) (Fig. <ns0:ref type='figure' target='#fig_8'>8A</ns0:ref>, first panel). Western blotting using antibody against HA-tag also showed that the protein level of full length SlDronc was decreased and the protein levels of processed SlDronc were increased when co-expressed with SlIAP (Fig. <ns0:ref type='figure' target='#fig_8'>8A</ns0:ref>, second panel). Western blotting using antibody against SfIAP showed that SlIAP mainly exsit as cleaved SlIAP (about 40 kDa) when co-expressed with SlDronc in SL2 cells (Fig. <ns0:ref type='figure' target='#fig_8'>8A</ns0:ref>, third panel).</ns0:p><ns0:p>To further test which portion of SlIAP functioned on SlDronc, the predicted cleavage site D87 was replaced by Ala (A) to avoid N-terminally truncation, and destabilizing degron N88 at N-terminus of cleaved SlIAP was replaced by Gly (G), a stabilizing residue in the Nend rule degradation pathway. When expressed in SL2 cells, the D87A mutation blocked the cleavage indicated by the disappearance of cleaved SlIAP (40 kDa) and N88G mutation stabilized the cleaved SlIAP (Fig. <ns0:ref type='figure' target='#fig_8'>8B</ns0:ref>), which confirmed that SlIAP was cleaved at the site D87 in N-terminal and the cleaved SlIAP was degraded by the N-end rule degradation pathway. When co-expressed with SlDronc in SL2 cells, D87A mutant failed to increase the protein levels of processed SlDronc, N88G mutant increase the protein levels of processed PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed SlDronc more significantly (Fig. <ns0:ref type='figure' target='#fig_8'>8C</ns0:ref>). These results suggest that SlIAP was cleaved following aspartate residue D87 in N-terminal, truncated SlIAP was degraded by the N-end rule degradation pathway and processed forms of SlDronc were increased with present of Nterminally truncated SlIAP.</ns0:p></ns0:div> <ns0:div><ns0:head>SlDronc was inhibited by P49</ns0:head><ns0:p>In generally, P49, baculoviral apoptosis inhibitor in SpliNPV, inhibits caspase activity of initiator caspases, and P35, baculoviral apoptosis inhibitor in AcMNPV, inhibits caspase activity of effector caspases <ns0:ref type='bibr' target='#b25'>(Jabbour et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b58'>Zoog et al. 2002)</ns0:ref>. To determine whether SlDronc is regulated by P49, firstly recombinant SlDronc expressed and purified in E. coli was tested by incubating with increasing amount of P49, and then caspase activities were determined. The result showed that P49 inhibited caspase activity of SlDronc in a dose dependent manner (Fig. <ns0:ref type='figure' target='#fig_9'>9A</ns0:ref>). At a 1:1 ratio of P49:SlDronc, enzymatic activity of SlDronc was reduced by 54% and SlDronc activity was reduced by 97% when the ratio was 4:1. To further test the effect of P49 on SlDronc in SL2 cell, SlDronc was co-expressed with increasing amount of P49. Western blot analysis showed that the protein level of full length SlDronc was increased gradually along with the increasing amount of P49 co-expressing with SlDronc (Fig. <ns0:ref type='figure' target='#fig_9'>9B</ns0:ref>), and protein grayscale analysis of full length Sl-caspase-1 showed that the protein level of full length Sl-caspase-1 was also increased slightly (Fig. <ns0:ref type='figure' target='#fig_9'>9C</ns0:ref>). These data prove that SlDronc was inhibited by P49.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Dronc has been continuously proved to be an essential initiator caspase in apoptosis pathway, it cleaves and activates effector caspases to cleave downstream proteins in PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed apoptotic pathway that finally lead to apoptosis <ns0:ref type='bibr' target='#b6'>(Daish et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b14'>Dorstyn et al. 1999;</ns0:ref><ns0:ref type='bibr' target='#b40'>Quinn et al. 2000)</ns0:ref>. Dronc homologs from several Lepidopteran insects have been reported successively, including Bombyx mori (BmDronc), Lymantria dispar (LdDronc) and Spodoptera frugiperda (SfDronc) <ns0:ref type='bibr' target='#b23'>(Huang et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kitaguchi et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b47'>Suganuma et al. 2011)</ns0:ref>. In this study, the first Spodoptera littoralis initiator caspase, SlDronc, was identified.</ns0:p><ns0:p>SlDronc possesses typical features of initiator caspase and was proved to induce apoptosis in SL2 cell.</ns0:p><ns0:p>Both mammalian and Drosophila caspases are subject to regulation by IAPs, and ubiquitylation of caspases by IAPs can be degradative or nondegradative <ns0:ref type='bibr' target='#b7'>(Darding &amp; Meier 2012;</ns0:ref><ns0:ref type='bibr' target='#b35'>Muro et al. 2002;</ns0:ref><ns0:ref type='bibr' target='#b52'>Vaux &amp; Silke 2005)</ns0:ref>. In degradative ubiquitylation, caspases are degraded by proteasome, while nondegradative ubiquitylation does not lead to degradation of caspases, it is proposed that Ub sterically occlude substrate entry and cause a conformational change of the caspases reducing their catalytic processivity <ns0:ref type='bibr'>(Ditzel et al. 2008)</ns0:ref>. Several researches showed that ubiquitylated full length Dronc mediated by DIAP1 was degraded by the proteasome initially <ns0:ref type='bibr' target='#b0'>(Chai et al. 2003;</ns0:ref><ns0:ref type='bibr' target='#b35'>Muro et al. 2002)</ns0:ref>, other researches showed processed and activated Dronc in the Dark apoptosome was degraded in a DIAP1-dependent manner <ns0:ref type='bibr' target='#b30'>(Lee et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b45'>Shapiro et al. 2008)</ns0:ref>. Interestingly, both depletion and overexpression of cBm-IAP1 stabilized the cleaved form of Bm-Dronc despite of the opposite effect on apoptosis in BM-N cell <ns0:ref type='bibr' target='#b20'>(Hamajima et al. 2016)</ns0:ref>. In this present study, overexpression of SlIAP also increase the protein levels of cleaved SlDronc (Pro+LS and LS) and decreased protein level of full-length SlDronc (Fig. <ns0:ref type='figure' target='#fig_8'>8A</ns0:ref>). Decrease of full-length SlDronc may due to elevated ubiquitylation by overexpressed SlIAP, but the mechanism by was incubated respectively with 600 nM of SlDronc or SlDronc C310A with C-terminal His-tag at 37 &#186;C for 3 h in Na-Citrate buffer , and then the mixtures were detected by western blotting using Anti-Sf-caspase-1 antibody . (B) Sl-caspase-1 (5 nM) was incubated with buffer and SlDronc (240 nM) respectively. Sl-caspase-1 C178A (5 nM) was incubated with SlDronc (240 nM). Then caspase activity of the mixtures were measured using synthetic caspase substrate Ac-DEVD-AFC (20 &#956;M). Caspase activity was indicated relative to that of Slcaspase-1 (5 nM) or Sf-caspase-1 (5 nM) incubated with buffer. SD from three independent experiments were presented, and statistical significance was analysied by t test.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Caspase activity was indicated relative to that of SlDronc incubated with buffer. SD from three independent experiments were presented, and statistical significance was analysied by t test . (B) SlDronc was co-expressed with increasing amount of P49. Plasmid expressing GFP was used to make sure SL2 cells in each well were transfected with same amount of plasmid . At 24 h post transfection, SL2 cells were harvested and cell lysates were analysied by western blotting utilizing antibodies against Flag-tag, Sf-caspase-1, His-tag or &#946;-actin. (C) Protein grayscale analysis of full length Sl-caspase-1 was performed using Quantity One.</ns0:p><ns0:p>Band density was indicated relative to that of cells expressing SlDronc and GFP.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>. To verify SlDronc indeed has enzymatic activity, catalytic activities of recombinant SlDronc with a C-terminal His-tag on 13 kinds of caspase substrates were measured. Comparing with the control, SlDronc with a C-terminal His-tag showed dramatical activities on Ac-VEID-AFC, Ac-LETD-AFC, Ac-IETD-AFC, Ac-PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50779:1:1:CHECK 24 Aug 2020) Manuscript to be reviewed which increased the protein levels of cleaved SlDronc (Pro+LS and LS) requires further study. Identification and functional characterization of SlDronc make the apoptosis pathway in Spodoptera littoralis more clear. When received apoptotic signal, SlDronc functions as an initiator caspase cleaving and activating effector caspase, Sl-caspase-1, then activated Slcaspase-1 cleaves several celluar substrates serving as executioner of apoptosis in SL2 cell. SlIAP functions as the last line of defense against apoptosis. The identification of SlDronc will facilitate further studying mechanism of apoptosis in Spodoptera littoralis and of baculovirus infection induced apoptosis. Conclusions An initiator caspase, SlDronc, was identified in Spodoptera littoralis. SlDronc cleaved and activated effector caspases. Overexpression of SlDronc induced apoptosis in SL2 cell, and knockdown of Sldronc decreased apoptosis induced by UV irradiation. The above results indicate that SlDronc is an apoptotic initiator caspase in Spodoptera littoralis. In addition, we also find that processed forms of SlDronc were increased with present of N-terminally truncated SlIAP and that SlDronc was inhibited by P49. This study makes contributions to clearly elaborating apoptotic pathway in Spodoptera littoralis and may facilitate the study of baculovirus infection induced apoptosis.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 9</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure> </ns0:body> "

"Dear Editor Gianluca Tettamanti and Reviewers: Article ID: 50779 Thank you very much for taking your time to review this manuscript. We really appreciate all your comments and suggestions! Your suggestions have enabled us to improve our work, we have edited the manuscript to address your concerns. Appended to this letter is our point-by-point response to the comments raised by the reviewers. We hope that the manuscript is now suitable for publication in Peer J. Sincerely yours, Hao Liu State Key Laboratory of Virology, Modern Virology Research Center College of Life Sciences, Wuhan University, Wuhan 430072 P. R. China Phone: +86-27-6875-6369 Email: [email protected] On behalf of all authors. Responses to the comments of reviewers and a list of changes: The reviewer’s comments are in italics, which are followed by our responses. Responses to the comments of Reviewer #1: Basic reporting Authors of the manuscript clearly expounded the functions of SlDronc and the pathways involved in Spodoptera littoralis. In this study, they identified an initiator caspase, SlDronc, in Spodoptera littoralis, elaborated that SlDronc have caspase activity, can cleaved and activated effector caspases. Based on the result of overexpression of SlDronc induced apoptosis in SL2 cell and knockdown of SlDronc decreased apoptosis induced by UV irradiation in SL2 cell, indicate that SlDronc is an apoptotic initiator caspase in Spodoptera littoralis. And found that processed forms of SlDronc were increased with present of N-terminally truncated SlIAP and that SlDronc was inhibited by P49. Generally, the topic of this work is interesting, the experiments were performed in a technically sound fashion, and the obtained results seemed to be reasonable for the journal publication requirements. However, the authors should address the following concerns before it can be accepted. Experimental design In this study, they identified an initiator caspase, SlDronc, in Spodoptera littoralis, elaborated that SlDronc have caspase activity, can cleaved and activated effector caspases. Based on the result of overexpression of SlDronc induced apoptosis in SL2 cell and knockdown of SlDronc decreased apoptosis induced by UV irradiation in SL2 cell, indicate that SlDronc is an apoptotic initiator caspase in Spodoptera littoralis. And found that processed forms of SlDronc were increased with present of N-terminally truncated SlIAP and that SlDronc was inhibited by P49. Validity of the findings Generally, the topic of this work is interesting, the experiments were performed in a technically sound fashion, and the obtained results seemed to be reasonable for the journal publication requirements. Comments for the Author Authors of the manuscript clearly expounded the functions of SlDronc and the pathways involved in Spodoptera littoralis. In this study, they identified an initiator caspase, SlDronc, in Spodoptera littoralis, elaborated that SlDronc have caspase activity, can cleaved and activated effector caspases. Based on the result of overexpression of SlDronc induced apoptosis in SL2 cell and knockdown of SlDronc decreased apoptosis induced by UV irradiation in SL2 cell, indicate that SlDronc is an apoptotic initiator caspase in Spodoptera littoralis. And found that processed forms of SlDronc were increased with present of N-terminally truncated SlIAP and that SlDronc was inhibited by P49. Generally, the topic of this work is interesting, the experiments were performed in a technically sound fashion, and the obtained results seemed to be reasonable for the journal publication requirements. However, the authors should address the following concerns before it can be accepted. Answer: We thank the reviewer for his/her appreciation of our work. Major points: 1. Both result 1 and Result 2 are bioinformatics analysis, consider combining the results into Figure 1. Answer: We agree with the reviewer on this point and have tried to combine the results into Figure 1. However, the amino acids in result 1 became too small for read when combine the result 1 and result 2 into one figure, we have to divide bioinformatics analysis into two figures. 2. In Figure 6C, compared to the mutant group “SlDronc C310A-Flag”, there is no obvious cleaved band in wild group “SlDronc-Flag”? Answer: Yes, this question also confused us for a long time, we have tested several antibodies against Flag-tag, however all of them could not detect processed SlDronc. Full length of SlDronc and Sl-caspase-1 were significantly decreased in cells expressing SlDronc-Flag comparing with that in cells expressing SlDronc C310A-Flag, and overexpression of SlDronc-Flag induced apoptosis, so we speculate that cleaved active SlDronc participate in apoptosis and have a very short life time. 3. In western blotting, the endogenous antibody and anti-HA antibody cannot specifically bind to the target protein, how to determine the target band? And there is not protein standard marker. Answer: Apology for making you confused, in order to read clearly, protein standard marker was replaced by short horizontal lines, images of protein standard marker have been submitted in raw data. HA-SlDronc has an HA-tag at N terminal, anti-HA antibody can detect full-length and Pro+LS of SlDronc through this HA-tag, the endogenous antibody detected full-length and Pro+LS of recombinant SlDronc in Figure 3, and predicted molecular weight of target proteins were contrasted to protein standard marker, all above help us determine the target band. 4. In Figure 9B, in the first panel, P49 inhibit the caspase activity of SlDronc, as the dose of P49 increases, why does the cleaved SlDronc show no change? And the second panel, it is suggested to do protein grayscale analysis. Answer: Agree, we have done protein grayscale analysis of the second panel and added the result into Figure 9C. SlDronc-Flag in Figure 9B has a Flag-tag at C terminal, as for the reason why the cleaved SlDronc show no change. Firstly, result in figure 8 shows that SlDronc mainly existed as Pro+LS when inhibited by SlIAP. Secondly, anti-Flag antibody we have used is weak in detecting SS-Flag fragment of SlDronc-Flag. Minor points: 1. Line 223, does the 'boring' mean to express the meaning of “containing”? Answer: Yes, the “boring” means to express the meaning of “containing”, we have replaced the “boring” by “containing” at Line 229 and replaced the “possing” by “containing” at Line 208 in Revised Manuscript. 2. Line 256, “showed activity on none of the selected substrates” should be “showed no activity on the selected substrates”. Answer: Agree. We have replaced the “showed activity on none of the selected substrates” by “showed no activity on the selected substrates”. (Line 265 in Revised Manuscript) 3. Line 222, the “posses” and Line 362, the “possess” should be “possesses”, please check the spelling of words in the full text. Answer: We are very sorry for our incorrect writing and it is rectified at Line 229 and Line 379 in Revised Manuscript. Also, we checked the spelling of words in the full text, added a space symbol after “full length” at Line 119, added a space symbol after “harvested” at Line 203, replaced the “autocleave” by “autoprocess” at Line 239 and Line 253, replaced the “esstial” by “essential” at Line 256, rectified “experment” to “experiment” at Line 277, rectified “mutantion” to “mutation” at Line 346-347 in Revised Manuscript. 4. In Figure 6A, the 'SLDronc' group does not see obvious phenomena, it is suggested to use a narrow to mark the apoptotic cells or mark apoptotic body using DAPI; In Figure 6B, P<0.05 means the value is significant, why is there significant between the GFP and MOCK groups? And in Figure 6C, GFP is the most common reporter protein, it should not have two bands, please explain. Answer: We appreciate it very much for this good suggestion, and we have used arrows to mark the apoptotic cells. Cells subject to slightly apoptosis when cultured in medium. During transfection, the medium was removed in GFP group and MOCK group was untreated, which may lead to MOCK group retain more apoptosis bodies than GFP group. Agree with that GFP is the most common reporter protein and it should not have two bands. We have tested several antibodies against Flag-tag, most of them detected only one band of GFP-Flag, however they could not detect SlDronc-Flag. Finally, we found an antibody against Flag-tag which could detect SlDronc-Flag, this antibody detected two bands of GFP-Flag. We speculate the band between 15 kDa and 25 kDa could be product of initiating from inner ATG in ORF of GFP, when initiating from A234TG or A264TG, the products will also contain a Flag-tag at C terminal. Responses to the comments of Reviewer #2: Basic reporting This reports Sldronc as an iniater paspase and the results basically accord to previously published data as Drosophila droc. I found there is nothing wrong with the data but feel a little bit of frustration about the novelty. I wanted to learn more about specificity side of Spodoptera littralis since upon metamorphosis larval tissues undergo different modes of remodeling processes. In Drosophila most imaginal tissues arise from the imaginal discs but in Lepidoptera some imaginal tissues undergo transformation from larval tissues. Also, lepidoptera shows variability. The authors use cell line but we do not know what kind of developmental process the cells undergo in to to. Also, initiation of metamorphosis is controlled by ecdysone signaling pathway and the story should start here, although autocleavage is postulated. I would like the authors to reconsider the structure of manuscript to emphasize the focus of this report as the second report on droc after Drosophila and add some data on how the apoptosis proceeds in natural situationn and clarify the novelty of Spodoptera dronc from fly one. Answer: The cell line we used was established by Mialhe, E. and his colleagues from explants of ovarioles excised from chrysalids of the Spodoptera littoralis, and we have added related paper into reference. Initiation of metamorphosis is controlled by ecdysone signaling pathway, however, apoptosis can be induced through intrinsic pathway or extrinsic pathway, UV irradiation would be a convenient method to stimulate intrinsic pathway. We thank the reviewer for his/her suggestion, as the reviewer said Spodoptera littralis is different from Drosophila in development which is regulated by apoptosis, identification of the first initiator caspase would an important work for further study of development in Spodoptera littralis. In addition, processed forms of SlDronc were increased with present of N-terminally truncated SlIAP, that is different from the effect of DIAP on Dronc in Drosophila, which suggests that mechanism of Dronc regulation may be discrepant in Spodoptera littralis and Drosophila. Therefore, we would like to remain the structure of this manuscript to emphasize the identification of the first initiator caspase in Spodoptera littralis. Experimental design It should be started with ecdyson signaling cascade and we would like to see morphological changes after knock down or overexpression of the responsible genes. Answer: Agree with the point that ecdyson is a useful tool to study apoptosis in adults, however, Spodoptera littralis is listed as invasive species and it is not permissive to feed larvae or adults in China. So, we would like to identify the new caspase, SlDronc, firstly, and then to study more functions of this caspase in cooperation with laboratories abord after the end of influence of COVID-19. Validity of the findings Methodology and the execusion look fine. Answer: We thank the reviewer for his/her appreciation of our work. Comments for the Author The use of sell line alone is doubtful to understand the function of this gene. Answer: We agree with the reviewer that more study or more data would be useful to further verify the function of this gene, however, cell lines have been used to identify new genes in insects frequently, such as identification of Sl-caspase-1 in Spodoptera littralis and LdDronc in Lymantria dispar. The present data in our manuscript would be enough to verify sldronc as a new gene involved in apoptosis. Of course, more functions of this gene will be studied in our future work. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Chronic infection by Staphylococcus aureus drives pathogenesis in important clinical settings, such as recurrent pulmonary infection in cystic fibrosis and relapsing infection in osteomyelitis. Treatment options for intracellular S. aureus infection are limited. Rifampin, a lipophilic antibiotic, readily penetrates host cell membranes, yet monotherapy is associated with rapid antibiotic resistance and development of severe adverse events. Antibiotic cotreatment can reduce this progression, yet efficacy diminishes as antibiotic resistance develops. ML141 and simvastatin inhibit S. aureus invasion through host-directed rather than bactericidal mechanisms.</ns0:p><ns0:p>Objective. To determine whether cotreatment of ML141 or of simvastatin with rifampin would enhance rifampin efficacy.</ns0:p><ns0:p>Methods. Assays to assess host cell invasion, host cell viability, host cell membrane permeability, and bactericidal activity were performed using the human embryonic kidney (HEK) 293-A cell line infected with S. aureus (29213) and treated with vehicle control, simvastatin, ML141, rifampin, or cotreatment of simvastatin or ML141 with rifampin.</ns0:p><ns0:p>Results. We found cotreatment of ML141 with rifampin reduced intracellular infection nearly 85 % when compared to the no treatment control. This decrease more than doubled the average 40 % reduction in response to rifampin monotherapy. In contrast, cotreatment of simvastatin with rifampin failed to improve rifampin efficacy. Also, in contrast to ML141, simvastatin increased propidium iodide (PI) positive cells, from an average of 10 % in control HEK 293-A cells to nearly 20 % in simvastatin-treated cells, indicating an increase in host cell membrane permeability. The simvastatin-induced increase was reversed to control levels by cotreatment of simvastatin with rifampin.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion.</ns0:head><ns0:p>Taken together, rifampin efficacy is increased through host-directed inhibition of S. aureus invasion by ML141, while efficacy is not increased by simvastatin. Considerations regarding novel therapeutic approaches may be dependent on underlying differences in pharmacology.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Staphylococcus aureus is an important cause of both acute <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref> and chronic infections <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b2'>3]</ns0:ref>. Chronic infections contribute substantially to morbidity and mortality in certain settings, most notably in progressive lung disease characteristic of cystic fibrosis (CF) <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref> and in deterioration of bone and joint tissue in osteomyelitis <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. S. aureus is an initial pulmonary isolate in pediatric patients with CF <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref> and by adulthood the majority of CF patients remain chronically infected <ns0:ref type='bibr' target='#b4'>[5,</ns0:ref><ns0:ref type='bibr' target='#b5'>6]</ns0:ref>. Chronic S. aureus infection is an ongoing treatment challenge as indicated by the 2018 Cystic Fibrosis Foundation Patient Registry reporting an increase in the percentage of patients infected with S. aureus each year from 59 % in 2003 to 70 % in 2018 <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref>.</ns0:p><ns0:p>S. aureus also is the most common cause of acute and chronic osteomyelitis in children and adults <ns0:ref type='bibr' target='#b1'>[2,</ns0:ref><ns0:ref type='bibr' target='#b7'>8]</ns0:ref>. Treatment of staphylococcal osteomyelitis is complicated further by increased incidence of methicillin-resistant S. aureus (MRSA) infection. The predominance of S. aureus in initiating and sustaining chronic infection may be attributable to the capacity to invade host cells, to reemerge and invade adjoining cells and to undergo phenotypic differentiation within host cells enabling persistence within the intracellular environment <ns0:ref type='bibr' target='#b8'>[9,</ns0:ref><ns0:ref type='bibr' target='#b9'>10,</ns0:ref><ns0:ref type='bibr' target='#b10'>11]</ns0:ref>. <ns0:ref type='bibr' target='#b11'>[12]</ns0:ref>. On the bacterial surface, invasive S. aureus strains express fibronectin binding proteins (FnBPs) that bind fibronectin, a host extracellular matrix protein. Fibronectin, as it binds to the host cell receptor &#945;5&#946;1, triggers receptor-mediated endocytosis of the bacteria-fibronectin complex.</ns0:p></ns0:div> <ns0:div><ns0:head>S. aureus invades host cells by exploiting host endocytic mechanisms</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>During invasion, host cell actin stress fibers disassemble, potentially providing pulling forces needed for engulfment <ns0:ref type='bibr' target='#b12'>[13,</ns0:ref><ns0:ref type='bibr' target='#b13'>14]</ns0:ref>. CDC42, a member of the RHO GTPase family, regulates actin stress fiber dynamics and can function ahead of family members RAC and RHO in the mobilization of actin <ns0:ref type='bibr' target='#b14'>[15]</ns0:ref>. The apparent regulatory role and early CDC42 activation during S. aureus invasion <ns0:ref type='bibr' target='#b15'>[16]</ns0:ref> suggest host CDC42 plays a central role in this invasive mechanism.</ns0:p><ns0:p>Treatment options for intracellular S. aureus infection are limited as first-line antibiotics demonstrate limited membrane permeability <ns0:ref type='bibr' target='#b16'>[17]</ns0:ref>. Rifampin is a lipophilic antibiotic that demonstrates a propensity for intracellular uptake, enabling clearance of both extracellular and intracellular susceptible bacteria <ns0:ref type='bibr' target='#b17'>[18,</ns0:ref><ns0:ref type='bibr' target='#b18'>19]</ns0:ref>. However, rifampin monotherapy is associated with the development of cross resistance to vancomycin and daptomycin <ns0:ref type='bibr' target='#b19'>[20]</ns0:ref>, rapid rifampin resistance <ns0:ref type='bibr' target='#b21'>[21,</ns0:ref><ns0:ref type='bibr' target='#b22'>22]</ns0:ref>, and progression of severe adverse events <ns0:ref type='bibr' target='#b23'>[23,</ns0:ref><ns0:ref type='bibr' target='#b24'>24]</ns0:ref>. To limit the development of resistance and adverse events, the current standard of care is cotreatment of rifampin with an antibiotic cocktail <ns0:ref type='bibr' target='#b25'>[25]</ns0:ref>.</ns0:p><ns0:p>To circumvent antibiotic resistance, an emerging therapeutic approach is to target the host rather than bacterial cells <ns0:ref type='bibr' target='#b12'>[13,</ns0:ref><ns0:ref type='bibr' target='#b26'>26]</ns0:ref>. For nearly two decades, researchers have examined statin drugs as potential host-directed therapeutics for infection, including invasive infection by S. aureus <ns0:ref type='bibr' target='#b27'>[27,</ns0:ref><ns0:ref type='bibr' target='#b28'>28,</ns0:ref><ns0:ref type='bibr' target='#b29'>29,</ns0:ref><ns0:ref type='bibr' target='#b30'>30,</ns0:ref><ns0:ref type='bibr' target='#b31'>31]</ns0:ref>. Our work had indicated therapeutic benefit may include limiting the spread of infection by reducing host cell invasion <ns0:ref type='bibr' target='#b32'>[32,</ns0:ref><ns0:ref type='bibr' target='#b33'>33,</ns0:ref><ns0:ref type='bibr' target='#b34'>34]</ns0:ref>. We identified an underlying mechanism where simvastatin reduces invasion of S. aureus into host cells by sequestering small-GTPases CDC42, RAC, and RHO in the cytosol <ns0:ref type='bibr' target='#b12'>[13]</ns0:ref>. In turn, this sequestration limits actin stress fiber disassembly and reduces fibronectin binding at &#945;5&#946;1 <ns0:ref type='bibr' target='#b35'>[35]</ns0:ref>. We went on to investigate whether targeted inhibition of host CDC42 would be sufficient to limit invasion. We discovered that ML141, a small molecule inhibitor with specificity for host CDC42 <ns0:ref type='bibr' target='#b36'>[36,</ns0:ref><ns0:ref type='bibr' target='#b37'>37]</ns0:ref>, decreased host cell vehicle control, 0.4% dimethyl sulfoxide (Fisher Scientific, Waltham, MA; DMSO), or with increasing concentrations of rifampin (VWR International; structure provided in supplemental file) (1 hr, 37&#176;C, 5 % CO 2 ). Serial dilutions were incubated overnight on tryptic soy agar (Sigma-Aldrich; TSA; 37&#176;C) and colony counts performed to determine CFU/ml. Of note, for all experiments, 0.4% was the final DMSO solvent concentration (vol/vol) for each treatment and 0.4% DMSO served as the non-treatment, negative, vehicle control.</ns0:p><ns0:p>To determine the IC50 of rifampin for intracellular infection, 35 mm cell culture dishes (Fisher Scientific) were precoated with Attachment Factor (Fisher Scientific) prior to plating 3 x 10 5 HEK 293-A cells/ml. HEK 293-A cells were inoculated with S. aureus at multiplicity of infection (MOI) 2 (6 x 10 5 CFU/ml ) or at MOI 100 (3 x 10 7 CFU), 30 min or 1 hr, 37&#176;C, 5 % CO 2 ) in 10 % FBS/phosphate buffered saline (VWR International; PBS) followed by incubation with increasing concentrations of rifampin (1 hr, 37 &#176;C, 5 % CO 2 ). To remove extracellular bacteria, HEK 293-A cells were incubated with gentamicin (Sigma-Aldrich; 50 &#956;g/ml) and lysostaphin (Sigma-Aldrich, 20 &#956;g/ml) in DMEM (45 min, 37&#176;C, 5 % CO 2 ) following three 1X PBS washes. Intracellular S. aureus were harvested using 1 % saponin (20 min, 37&#176;C, 5 % CO 2 ) following three 1X PBS washes and serial dilutions of the supernatant were plated onto TSA, incubated overnight at 37&#176;C and CFU/ml quantified by colony counts.</ns0:p></ns0:div> <ns0:div><ns0:head>ML141 Pretreatment Assay</ns0:head><ns0:p>HEK 293-A cells were plated as described above. The following day, HEK 293-A cells were pretreated with DMSO (0.4 %) or with ML141 (10 &#956;M; 37&#176;C, 5% CO 2 , 24 h; structure provided in supplemental file). The next day, pretreated HEK 293-A cells were inoculated (3 x 10 7 CFU, 1 hr, 37&#176;C, 5 % CO 2 ) in 10 % FBS/PBS. Intracellular bacteria were isolated and quantified as described above.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head></ns0:div> <ns0:div><ns0:head>Cotreatment Assay</ns0:head><ns0:p>HEK 293-A cells were plated as described above. Treatments were performed as outlined in Table <ns0:ref type='table'>1</ns0:ref> at these concentrations: DMSO (0.4 %), ML141 (10 &#956;M), simvastatin (1 &#956;M; structure provided in supplemental file), rifampin (0.01 mg/L).</ns0:p><ns0:p>Following post-treatment, intracellular bacteria were isolated and quantified as described.</ns0:p></ns0:div> <ns0:div><ns0:head>Host Cell Viability Assay</ns0:head><ns0:p>HEK 293-A cells were plated at 5 x 10 4 cells/0.5 ml in 48-well cell culture plates (VWR International) precoated with Attachment Factor. To acquire uniform microplate readings, multiwell culture plates were used. This change from the 35 mm plates used for the invasion assay required a reduction in cell count. The following day, HEK 293-A cells were treated with DMSO (0.4 %), ML141 (10 &#956;M), rifampin (0.01 mg/L), or ML141 (10 &#956;M) with rifampin (0.01 mg/L; 24 hr, 37&#176;C, 5 % CO 2 ). Cell viability was measured using CellTiter 96&#174; Aqueous One Solution Reagent (Promega Corporation, Madison, WI). Absorbance was measured at 490 nm using a BioRad iMark microplate reader.</ns0:p></ns0:div> <ns0:div><ns0:head>Bactericidal Assay</ns0:head><ns0:p>S. aureus were harvested as described above and treated with DMSO (0.2 %), ML141 (10 &#956;M), rifampin (0.002 mg/L), or ML141 (10 &#956;M) with rifampin (0.002 mg/L; 1 hr, 37&#176;C, 5 % CO 2 ). Bacteria were serially diluted, plated on TSA, incubated overnight at 37&#176;C and CFU/ml determined from colony counts. Manuscript to be reviewed treated with DMSO (0.4 %), ML141 (10 &#956;M), simvastatin (1 &#956;M), rifampin (0.01 mg/L), ML141 (10 &#956;M) with rifampin (0.01 mg/L), or simvastatin (1 &#956;M) with rifampin (0.01 mg/L; 1 hr, 37&#176;C, 5 % CO 2 ). Immediately prior to performing the assay, HEK 293-A cells serving as the positive control were incubated with 70% ethanol (1 min). HEK 293-A cells were harvested by scrapping with cell lifters and washed in FACS buffer. Propidium iodide (Sigma Aldrich) was added to each sample immediately prior to measurement. Percentage of propidium iodide positive HEK 293-A cells was determined using a MACSQuant Analyzer 10 flow cytometer (Miltenyi Biotech Inc., Auburn, CA.).</ns0:p></ns0:div> <ns0:div><ns0:head>Propidium Iodide Flow Cytometry Assay</ns0:head></ns0:div> <ns0:div><ns0:head>Statistical Analysis</ns0:head><ns0:p>All data were analyzed using Prism software (GraphPad Software Inc., San Diego, CA).</ns0:p><ns0:p>Rifampin IC50 was determined by nonlinear regression analysis. Means between groups were compared by one-way ANOVA followed by Newman-Keuls or Holm-Sidak's post-hoc analysis.</ns0:p><ns0:p>The threshold for statistical significance was P &lt; 0.05.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Cotreatment of rifampin with ML141 reduces intracellular S. aureus infection more than rifampin alone</ns0:head><ns0:p>We first determined the half maximal inhibitory concentration (IC50) of rifampin in a model cell line, HEK 293-A. The rifampin IC50 was 0.004 mg/L when HEK 293-A cells were incubated with bacteria at MOI 2 and the IC50 increased to 0.01 mg/L when the MOI was increased to 100 (Fig. <ns0:ref type='figure' target='#fig_1'>1</ns0:ref>). Thus, a higher concentration of antibiotic was needed to clear a larger intracellular bacterial population. All subsequent experiments were performed at MOI 100 to ensure there were sufficient numbers of colonies in the cotreatment group for statistical analysis.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Consistent with what we had found in human umbilical vein endothelial cells <ns0:ref type='bibr' target='#b26'>[26]</ns0:ref>, pretreatment of HEK 293-A cells with 10 &#61549;M ML141 reduced the number of intracellular bacteria by an average of 40% when compared to pretreatment with the negative, vehicle control DMSO (Fig. <ns0:ref type='figure'>2</ns0:ref>, Panel A, P = 0.0169). Therefore, this concentration of ML141 served as the positive control for all subsequent experiments. Also consistent with our work and that of others <ns0:ref type='bibr' target='#b38'>[38,</ns0:ref><ns0:ref type='bibr' target='#b39'>39]</ns0:ref>, the experimental design used throughout this study is one of pretreatment. Pretreatment is not used as a model for prophylactic clinical applications, but rather, as a model for limiting the spread of infection post-onset and diagnosis, given the evidence that invasive S. aureus strains successfully exit infected host cells to initiate new infection <ns0:ref type='bibr' target='#b10'>[11,</ns0:ref><ns0:ref type='bibr' target='#b40'>40,</ns0:ref><ns0:ref type='bibr' target='#b41'>41]</ns0:ref>. We found cotreatment of 10 &#61549;M ML141 with rifampin at the IC50 decreased the number of intracellular bacteria more than ML141 or rifampin alone (Fig. <ns0:ref type='figure'>2</ns0:ref>, Panel B, P &lt; 0.0001). Thus, cotreatment of ML141 with rifampin appeared to enhance bacterial clearance.</ns0:p></ns0:div> <ns0:div><ns0:head>Underlying mechanisms for enhanced clearance are not associated with an appreciable loss in host cell viability or loss of host cell membrane integrity</ns0:head><ns0:p>We next assessed whether the reduction in the number of intracellular bacteria was due to decreased numbers of viable host cells rather than to enhanced bacterial clearance. HEK 293-A cells were incubated with the DMSO control, ML141 alone, rifampin alone, or ML141 combined with rifampin for the same length of time as had been used for the invasion assay. HEK 293-A metabolic activity was assayed by measuring the conversion of a tetrazolium compound to formazan, a colored product detectable by absorbance at 490 nm. Compared to the DMSO control group, no decrease in absorbance was detected in any treatment group, indicating PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed treatment did not diminish metabolic activity, an indicator of sustained cell viability (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>, Panel A).</ns0:p><ns0:p>We went on to assess whether clearance of intracellular infection might be attributable to increased host membrane permeability, allowing greater penetrance of antibiotic into the intracellular compartment. HEK 293-A cells were incubated with the vehicle control DMSO or with ML141, rifampin, or ML141 combined with rifampin for the same duration as had been used for the invasion assay. Host cell membrane permeability was assessed using propidium iodide uptake. The percentage of propidium iodide positive HEK 293-A cells, an indication of membrane permeability, remained at DMSO control levels in response to each treatment, indicating HEK 293-A host cell membrane integrity is maintained following incubation with all treatments (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>, Panels B and C).</ns0:p></ns0:div> <ns0:div><ns0:head>Underlying mechanisms for enhanced clearance by the co-treatment are not associated with appreciable ML141 bactericidal activity or with enhanced rifampin bactericidal activity</ns0:head><ns0:p>Although ML141 has been assessed repeatedly for bactericidal activity and no such activity has been detected toward multiple S. aureus or Streptococcus pyogenes strains <ns0:ref type='bibr' target='#b26'>[26,</ns0:ref><ns0:ref type='bibr' target='#b35'>35]</ns0:ref>,</ns0:p><ns0:p>we saw it as important to verify this under the experimental conditions of the rifampin studies.</ns0:p><ns0:p>We inoculated HEK 293-A media containing the DMSO control or 10 &#61549;M ML141 for the same length of time used in the invasion assay. We found that the number of viable bacteria following ML141 treatment was similar to the control treated group, indicating ML141 exhibited no bactericidal activity under the conditions used for the invasion assay (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>, Panel B).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>To assess whether co-treatment with ML141 might somehow enhance rifampin bactericidal activity, we first needed to determine the IC50 of rifampin on S. aureus ATCC 29213 growth in axenic (host cell-free) conditions. This was necessary because the rifampin IC50 for intracellular bacteria (0.01 mg/L, reported above) was expected to exceed the IC50 needed under axenic conditions. Consistent with this expectation, the IC50 of rifampin in the absence of host cells was 0.002 mg/L (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>, Panel A). We then found the number of viable bacteria following cotreatment of 10 &#61549;M ML141 with rifampin at the IC50 for axenic conditions was similar to the number of bacteria recovered following rifampin alone, indicating no detectable enhancement of rifampin bactericidal activity by ML141 at the concentrations used (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>, Panel C, P = 0.0005).</ns0:p></ns0:div> <ns0:div><ns0:head>Cotreatment of simvastatin with rifampin fails to improve rifampin efficacy</ns0:head><ns0:p>We went on to explore the hypothesis that cotreatment of rifampin with the host-directed therapeutic simvastatin might achieve similar enhancement in the clearance of intracellular infection. The hypothesis was based on our earlier findings that 1.0 &#61549;M simvastatin decreases intracellular infection in part by sequestering host GTPases, including CDC42 <ns0:ref type='bibr' target='#b12'>[13]</ns0:ref>, the hostdirected target of ML141 <ns0:ref type='bibr' target='#b36'>[36,</ns0:ref><ns0:ref type='bibr' target='#b37'>37]</ns0:ref>. Moreover, both ML141 and simvastatin disrupt actin dynamics necessary for host cell invasion by S. aureus and by Streptococcus pyogenes <ns0:ref type='bibr' target='#b26'>[26,</ns0:ref><ns0:ref type='bibr' target='#b35'>35]</ns0:ref>.</ns0:p><ns0:p>Contrary to the hypothesis, we found no enhancement of bacterial clearance was achieved by cotreatment of rifampin with simvastatin (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>, Panel A, P = 0.0124).</ns0:p></ns0:div> <ns0:div><ns0:head>Differential effect of simvastatin on HEK 293-A cell membrane permeability is reversed by cotreatment</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>We were curious to understand the differential effect between simvastatin and ML141. In earlier work, we <ns0:ref type='bibr' target='#b35'>[35]</ns0:ref> and others <ns0:ref type='bibr' target='#b38'>[38]</ns0:ref> had found that simvastatin can induce host cell membrane permeability in specific cell types. We examined the effect of simvastatin on HEK 293-A cell membrane permeability and the effect of rifampin cotreatment. We found in contrast to ML141 (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>, Panel B), simvastatin treatment increased HEK 293-A cell permeability (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>, Panel B, P = 0.0042). We also found the increase in membrane permeability in response to simvastatin was reversed by cotreatment with rifampin (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>, Panels B and C).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In this study, we report cotreatment of ML141 with rifampin decreases intracellular infection more than rifampin monotherapy (Fig. <ns0:ref type='figure'>2</ns0:ref>). Improved clearance is achieved through mechanisms that sustain host cell viability and host membrane integrity (Fig. <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>) in the absence of detectable improvement of rifampin bactericidal activity (Fig. <ns0:ref type='figure' target='#fig_4'>4</ns0:ref>).</ns0:p><ns0:p>In contrast, cotreatment of cholesterol-lowering simvastatin with rifampin yielded no detectable improvement in bacterial clearance (Fig. <ns0:ref type='figure' target='#fig_5'>5</ns0:ref>). To explore the differential response between ML141 and simvastatin, we compared effects on host membrane permeability. We found that in contrast to ML141, simvastatin increases membrane permeability. This finding is consistent with earlier reports by our group and others that simvastatin induces membrane permeability in specific cell types <ns0:ref type='bibr' target='#b35'>[35,</ns0:ref><ns0:ref type='bibr' target='#b38'>38]</ns0:ref>. Statins exert pleiotropic effects through inhibition of multiple intermediates within the cholesterol biosynthesis pathway, including loss of membrane integrity through decreased synthesis of the intermediate mevalonate <ns0:ref type='bibr' target='#b42'>[42,</ns0:ref><ns0:ref type='bibr' target='#b43'>43]</ns0:ref>. Yet, loss of membrane integrity is not a universal response to statins <ns0:ref type='bibr' target='#b12'>[13,</ns0:ref><ns0:ref type='bibr' target='#b44'>44]</ns0:ref>. Thus, it is plausible that PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed examination of simvastatin across multiple cell types, timepoints and dosages may have yielded a response similar to ML141.</ns0:p><ns0:p>We found the increase in membrane permeability by simvastatin was reversed by rifampin cotreatment. The return to baseline in response to cotreatment may be due to hostdirected effects of rifampin. Rifampin not only acts through inhibition of bacterial RNA polymerase but also through host-directed responses, including induction of members of a subclass of the mammalian ATP-binding cassette family, the multidrug resistance protein (MRP) transporters <ns0:ref type='bibr' target='#b45'>[45]</ns0:ref>. Statin drugs can undergo efflux from the cell via these MRP transporters <ns0:ref type='bibr' target='#b46'>[46]</ns0:ref>.</ns0:p><ns0:p>Our observation that membrane permeability returned to baseline in the cotreatment group would be consistent with statin efflux driven by rifampin induction of these transporters. Similarly, the failure of cotreatment to improve bacterial clearance also would be consistent with rifampindriven statin efflux. However, such conclusions await further experimental evidence.</ns0:p><ns0:p>Differences in the mode-of-action of simvastatin and ML141 also may contribute to the differences observed in response to cotreatment. Simvastatin inhibits 3-hydroxy-3methylglutaryl coenzyme A (HMG-CoA) reductase, the rate limiting enzyme in the cholesterol biosynthesis pathway <ns0:ref type='bibr' target='#b28'>[28]</ns0:ref>. Inhibition of HMG-CoA reductase decreases synthesis of cholesterol, as well as isoprenoid intermediates synthesized in the pathway <ns0:ref type='bibr' target='#b47'>[47]</ns0:ref>. Isoprenoid intermediates can serve as membrane anchors for CaaX domain containing proteins, including CDC42 <ns0:ref type='bibr' target='#b47'>[47]</ns0:ref>,</ns0:p><ns0:p>and decreased synthesis of isoprenoid intermediates limits CDC42 membrane localization <ns0:ref type='bibr' target='#b12'>[13]</ns0:ref>.</ns0:p><ns0:p>Although simvastatin inhibits membrane localization of CDC42, activation of CDC42 by GTP binding within the activation site is sustained <ns0:ref type='bibr' target='#b48'>[48]</ns0:ref>. This is in contrast to ML141, an allosteric inhibitor with specificity for human CDC42 <ns0:ref type='bibr' target='#b36'>[36,</ns0:ref><ns0:ref type='bibr' target='#b37'>37]</ns0:ref>. ML141 dissociates GTP within the activation site of CDC42, decreasing CDC42 activation <ns0:ref type='bibr' target='#b36'>[36]</ns0:ref>. Another distinction between simvastatin and ML141 is that simvastatin inhibits membrane localization of additional CaaXdomain containing proteins, including RAC and RHO <ns0:ref type='bibr' target='#b49'>[49]</ns0:ref>. Thus, simvastatin affects multiple small GTPases, whereas ML141 has demonstrated specificity for CDC42, with no inhibitory activity detected toward RAC or RHO <ns0:ref type='bibr' target='#b36'>[36]</ns0:ref>. These distinctions in the mode-of-action and underlying pharmacology of simvastatin and ML141 could contribute to the observed differences in response to cotreatment.</ns0:p><ns0:p>Our findings indicate not only the promise of host-directed therapeutic approaches such as ML141, but also potential limitations of combinatorial therapies, such as the use of simvastatin with rifampin. Targeting host cell invasion may indeed have therapeutic benefit when done in combination with antibiotics, but careful examination of underlying mechanisms continues to be warranted.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>We sought to determine whether cotreatment of the host-directed therapeutics ML141 or simvastatin with the lipophilic antibiotic rifampin enhances clearance of intracellular S. aureus.</ns0:p><ns0:p>We found cotreatment of ML141 with rifampin enhanced rifampin efficacy, while cotreatment of simvastatin with rifampin failed to improve rifampin efficacy. Simvastatin monotherapy increased host cell permeability, while ML141 monotherapy did not. Increases in host cell permeability in response to simvastatin were reversed by rifampin. Differences in the underlying pharmacology of simvastatin and ML141 may contribute to differences observed in response to cotreatment and should be considered when assessing the efficacy of use with antibiotics. Staphylococcus aureus colony forming units; CFU; 30 min, 37&#176;C, 5 % CO 2 ) then treated with dimethyl sulfoxide (DMSO), or 0.001, 0.003, 0.01, or 0.03 mg/L rifampin (1 hr, 37&#176;C, 5 % CO 2 ).</ns0:p><ns0:note type='other'>Figure Legends</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed HEK 293-A cells were treated with gentamicin (50 &#956;g/ml) and lysostaphin (20 &#956;g/ml; 45 min, 37&#176;C, 5 % CO 2 ) to remove extracellular bacteria and intracellular bacteria were harvested using 1 % saponin (20 min, 37&#176;C, 5 % CO 2 ). Serial dilutions were plated on tryptic soy agar (24 hr, 37&#176;C) and colony counts were performed. Data are pooled from two independent experiments and are represented as CFU/ml &#177; SEM (n = 4-8/treatment). Panel B HEK 293-A cells were plated as described above and infected at MOI of 100 (3x10 7 CFU; 1 hr, 37&#176;C, 5 % CO 2 ). HEK 293-A cells were treated with DMSO, or 0.003, 0.01, 0.03, or 0.1 mg/L rifampin (1 hr, 37&#176;C, 5 % CO 2 ). Extracellular bacteria were killed, intracellular bacteria harvested, and CFU/ml determined as described above. Data are represented as CFU/ml &#177; SEM (n = 4/treatment).</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 2 Pretreatment with ML141 and cotreatment of ML141 with rifampin reduces intracellular Staphylococcus aureus infection.</ns0:head><ns0:p>Panel A HEK 293-A cells were plated at 3 x 10 5 cells/ml in 35 mm cell culture dishes. The following day, cells were treated with dimethyl sulfoxide (DMSO) or ML141 (10 &#956;M; 24 hr, 37&#176;C, 5 % CO 2 ). HEK 293-A cells were infected at a multiplicity of infection (MOI) of 2 (30 min, 37&#176;C, 5 % CO 2 ) then were treated with gentamicin (50 &#956;g/ml) and lysostaphin (20 &#956;g/ml; 45 min, 37&#176;C, 5 % CO 2 ) to remove extracellular bacteria. Intracellular bacteria were harvested using 1 % saponin (20 min, 37&#176;C, 5 % CO 2 ) and serial dilutions were plated on tryptic soy agar (24 hr, 37&#176;C). Manuscript to be reviewed DMSO (green), ML141 (black), rifampin (purple), or rifampin with ML141 as cotreatment (red).</ns0:p><ns0:p>The overlay of histograms reveals nearly overlapping peaks for each treatment. under axenic conditions is 0.002 mg/L. 3x10 7 Staphylococcus aureus colony forming units (CFU) were incubated with increasing concentrations of rifampin or dimethyl sulfoxide (DMSO; 1 hr, 37&#176;C, 5 % CO 2 ). Serial dilutions were plated on tryptic soy agar (24 hr, 37&#176;C) and colony counts were performed. Data are represented as CFU/ml &#177; SEM (n = 3/treatment). Panel B ML141 bactericidal activity not detected at concentrations used. Following treatment of 3 x 10 7</ns0:p><ns0:p>Staphylococcus aureus colony forming units (CFU) with dimethyl sulfoxide (DMSO) or ML141 (10 &#956;M; 1 hr, 37&#176;C, 5 % CO 2 ), serial dilutions were plated onto tryptic soy agar (24 hr, 37&#176;C) and colony counts were performed. Data are represented by CFU/ml &#177; SEM (P &gt; 0.05 by oneway ANOVA followed by Holm-Sidak's post-hoc analysis, n = 3/treatment). Panel C No detectable enhancement of rifampin bactericidal activity by ML141 at concentrations tested.</ns0:p><ns0:p>Following treatment of 3 x 10 7 Staphylococcus aureus CFU with dimethyl sulfoxide (DMSO), ML141 (10 &#956;M), rifampin (0.002 mg/L), or ML141 (10 &#956;M) in combination with rifampin (0.002 mg/L; 1 hr, 37&#176;C, 5 % CO 2 ), bacteria were plated and quantified as described above. Data are represented by CFU/ml &#177; SEM (* less than DMSO, P &lt; 0.05 by one-way ANOVA followed by Holm-Sidak's post-hoc analysis, n = 3/treatment). Manuscript to be reviewed HEK 293-A cells were plated at 3 x 10 5 cells/ml in 35 mm cell culture dishes. The following day, HEK 293-A cells were treated with dimethyl sulfoxide (DMSO) or simvastatin (1 &#956;M; 24 hr, 37&#176;C, 5 % CO 2 ). The next day, HEK 293-A cells were infected with 3x10 7 Staphylococcus aureus colony forming units (CFU; 1 hr, 37&#176;C, 5 % CO 2 ) then were treated with DMSO, simvastatin (1 &#956;M), rifampin (0.01 mg/L), or cotreatment of simvastatin (1 &#956;M) with rifampin (0.01 mg/L; 1 hr, 37&#176;C, 5 % CO 2 ). Extracellular bacteria were killed using gentamicin (50 &#956;g/ml) and lysostaphin (20 &#956;g/ml; 45 min, 37&#176;C, 5 % CO 2 ) and intracellular S. aureus were harvested using 1 % saponin (20 min, 37&#176;C, 5 % CO 2 ). Serial dilutions were plated onto tryptic soy agar (24 hr, 37&#176;C) and colony counts were performed. Data are represented by CFU/ml &#177; SEM (* less than DMSO, P &lt; 0.05 by one-way ANOVA followed by Holm-Sidak's post-hoc analysis, n = 3/treatment). Data are pooled from two independent experiments. Panel B Increase in HEK 293-A cell membrane permeability by simvastatin is reversed by cotreatment. HEK 293-A cells were plated at 3 x 10 5 cells/ml in 35 mm cell culture dishes. HEK 293-A cells were treated with DMSO or simvastatin (1 &#956;M; 24 hr, 37&#176;C, 5 % CO 2 ). The next day, cells were treated with DMSO, simvastatin (1 &#956;M), rifampin (0.01 mg/L), or cotreatment of simvastatin (1 &#956;M) and rifampin (0.01 mg/L; 1 hr, 37&#176;C, 5 % CO 2 ). HEK 293-A cells were harvested and resuspended in FACS buffer, then stained with propidium iodide (PI). Samples were analyzed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>HEK 293-A cells were plated and incubated with DMSO (0.4 %), ML141 (10 &#956;M), or simvastatin (1 &#956;M) as described in the cotreatment assay. The following day, HEK 293-A were PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1 A</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1 A higher concentration of rifampin is required when there is an increase in the</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Intracellular bacteria were quantified by colony counts. Data are represented as colony forming units (CFU)/ml &#177; SEM (* less than DMSO, P &lt; 0.05 by Student's t-test, n = 4/treatment). Panel B HEK 293-A cells were plated and pretreated with ML141 or DMSO as described above. HEK 293-A cells were infected at MOI of 100 by incubating with 3 x 10 7 S. aureus CFU then were treated with DMSO, ML141 (10 &#956;M), rifampin (0.01 mg/L), or cotreatment of ML141 (10 &#956;M) with rifampin PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)Manuscript to be reviewed (0.01 mg/L; 1 hr, 37&#176;C, 5 % CO2). Extracellular bacteria were killed, and intracellular bacteria were harvested and plated as described above. Data are represented by CFU/ml &#177; SEM (* less than DMSO and ML141, ** less than DMSO, ML141, or rifampin, P &lt; 0.05 by one-way ANOVA followed by Holm-Sidak's post-hoc analysis, n = 4/treatment). Data are representative of three replicate experiments.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3 HEK 293-A cell viability and membrane integrity are maintained following cotreatment</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4 Rifampin bactericidal activity is not enhanced by ML141. Panel A Rifampin IC50</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5 Analysis of simvastatin cotreatment with rifampin yields differential results from those</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1 A 5 7 CFU; 1</ns0:head><ns0:label>1571</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2 with 3 x 10 7 S</ns0:head><ns0:label>27</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 3 HEK</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 4 Rifampin 7 Following treatment of 3 x 10 7 7</ns0:head><ns0:label>47107</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='31,42.52,70.87,220.24,672.95' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='33,42.52,70.87,245.52,672.95' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='36,42.52,70.87,247.97,672.95' type='bitmap' /></ns0:figure> <ns0:note place='foot' n='2'>Table 1. Treatment regimen for cotreatment assays. PeerJ reviewing PDF | (2020:06:50566:1:1:NEW 6 Sep 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Reviewer 1 (Anonymous) Basic reporting In the research article titled 'Differential effects of cotreatment of the antibiotic rifampin with host-directed therapeutics in reducing intracellular staphylococcus aureus infection', the authors have tried to study the impact of cotreatment of ML141/simvastatin with rifampin against S. aureus infections. Rifampin is a front line drug for treating S. aureus infections but is associated with bacterial resistance. the authors have explored a possibility of combinatorial therapy of using an anti-virulent molecule along with antibiotic rifampin to check for treating S. aureus. This concept is fast gaining attention for various infectious diseases.  Experimental design No comment  Validity of the findings No comment  Comments for the Author The manuscript is well written with supporting evidence, however the following points needs to be addressed. 1. One of my biggest concern is missing control for all the experiments. Neither positive nor negative controls were used in the experiments. using these controls can help us to understand how efficiently the experiment worked. Authors’ Response The authors apologize for the failure to have clarified the controls. Below are the sections where controls are clarified. Controls were used for all experiments. Lines 158-159 Immediately prior to performing the assay, HEK 293-A cells serving as the positive control were incubated with 70% ethanol (1 min). Lines 178-182 Consistent with what we had found in human umbilical vein endothelial cells [26], pretreatment of HEK 293-A cells with 10 M ML141 reduced the number of intracellular bacteria by an average of 40% when compared to pretreatment with the negative, vehicle control DMSO (Fig. 1, Panel A, P = 0.0169). Therefore, this concentration of ML141 served as the positive control for all subsequent experiments. Lines 113-115 Of note, for all experiments, 0.4% was the final DMSO solvent concentration (vol/vol) for each treatment and 0.4% DMSO served as the non-treatment, negative, vehicle control. 2. There is inconsistency in the number of cells used for various experiments which the authors have not reasoned. Similarly, the authors have used 3 X 105 cells/ml for all the experiments where 5 X 104 cells were used for Cell viability experiment in Figure 2A. It will be better if the authors reasons why different amount of cells were used for these experiments. The authors have added the following language to the Methods section to clarify the change in the number of cells used. Lines 141-143 To acquire uniform microplate readings, multi-well culture plates were used. This change from the 35 mm plates used for the invasion assay required a reduction in cell count. proper justification is required for why for Figure 1, HEK293-A cells were infected at a MOI of 2 for figure 1A and MOI of 100 for figure 1B. The authors apologize for not having made that rationale clear. The re-written section of the results are as follows: Lines 172-177 We first determined the half maximal inhibitory concentration (IC50) of rifampin in a model cell line, HEK 293-A. The rifampin IC50 was 0.004 mg/L when HEK 293-A cells were incubated with bacteria at MOI 2 and the IC50 increased to 0.01 mg/L when the MOI was increased to 100 (Fig. 1). Thus, a higher concentration of antibiotic was needed to clear a larger intracellular bacterial population. All subsequent experiments were performed at MOI 100 to ensure there were sufficient numbers of colonies in the cotreatment group for statistical analysis. 3.  The results could have been explained better. Details regarding the experiments/assay performed and what observations were made were missing in the Results section.   The authors apologize for the lack of explanation and details. The results have been rewritten, almost entirely. Please, see the Results section in its entirety below. Lines 170-251 Cotreatment of rifampin with ML141 reduces intracellular S. aureus infection more than rifampin alone We first determined the half maximal inhibitory concentration (IC50) of rifampin in a model cell line, HEK 293-A. The rifampin IC50 was 0.004 mg/L when HEK 293-A cells were incubated with bacteria at MOI 2 and the IC50 increased to 0.01 mg/L when the MOI was increased to 100 (Fig. 1). Thus, a higher concentration of antibiotic was needed to clear a larger intracellular bacterial population. All subsequent experiments were performed at MOI 100 to ensure there were sufficient numbers of colonies in the cotreatment group for statistical analysis. Consistent with what we had found in human umbilical vein endothelial cells [26], pretreatment of HEK 293-A cells with 10 M ML141 reduced the number of intracellular bacteria by an average of 40% when compared to pretreatment with the negative, vehicle control DMSO (Fig. 2, Panel A, P = 0.0169). Therefore, this concentration of ML141 served as the positive control for all subsequent experiments. Also consistent with our work and that of others [38,39], the experimental design used throughout this study is one of pretreatment. Pretreatment is not used as a model for prophylactic clinical applications, but rather, as a model for limiting the spread of infection post-onset and diagnosis, given the evidence that invasive S. aureus strains successfully exit infected host cells to initiate new infection [11,40,41]. We found cotreatment of 10 M ML141 with rifampin at the IC50 decreased the number of intracellular bacteria more than ML141 or rifampin alone (Fig. 2, Panel B, P < 0.0001). Thus, cotreatment of ML141 with rifampin appeared to enhance bacterial clearance. Underlying mechanisms for enhanced clearance are not associated with an appreciable loss in host cell viability or loss of host cell membrane integrity We next assessed whether the reduction in the number of intracellular bacteria was due to decreased numbers of viable host cells rather than to enhanced bacterial clearance. HEK 293-A cells were incubated with the DMSO control, ML141 alone, rifampin alone, or ML141 combined with rifampin for the same length of time as had been used for the invasion assay. HEK 293-A metabolic activity was assayed by measuring the conversion of a tetrazolium compound to formazan, a colored product detectable by absorbance at 490 nm. Compared to the DMSO control group, no decrease in absorbance was detected in any treatment group, indicating treatment did not diminish metabolic activity, an indicator of sustained cell viability (Fig. 3, Panel A). We went on to assess whether clearance of intracellular infection might be attributable to increased host membrane permeability, allowing greater penetrance of antibiotic into the intracellular compartment. HEK 293-A cells were incubated with the vehicle control DMSO or with ML141, rifampin, or ML141 combined with rifampin for the same duration as had been used for the invasion assay. Host cell membrane permeability was assessed using propidium iodide uptake. The percentage of propidium iodide positive HEK 293-A cells, an indication of membrane permeability, remained at DMSO control levels in response to each treatment, indicating HEK 293-A host cell membrane integrity is maintained following incubation with all treatments (Fig. 3, Panels B and C). Underlying mechanisms for enhanced clearance by the co-treatment are not associated with appreciable ML141 bactericidal activity or with enhanced rifampin bactericidal activity Although ML141 has been assessed repeatedly for bactericidal activity and no such activity has been detected toward multiple S. aureus or Streptococcus pyogenes strains [26,35], we saw it as important to verify this under the experimental conditions of the rifampin studies. We inoculated HEK 293-A media containing the DMSO control or 10 M ML141 for the same length of time used in the invasion assay. We found that the number of viable bacteria following ML141 treatment was similar to the control treated group, indicating ML141 exhibited no bactericidal activity under the conditions used for the invasion assay (Fig. 4, Panel B). To assess whether co-treatment with ML141 might somehow enhance rifampin bactericidal activity, we first needed to determine the IC50 of rifampin on S. aureus ATCC 29213 growth in axenic (host cell-free) conditions. This was necessary because the rifampin IC50 for intracellular bacteria (0.01 mg/L, reported above) was expected to exceed the IC50 needed under axenic conditions. Consistent with this expectation, the IC50 of rifampin in the absence of host cells was 0.002 mg/L (Fig. 4, Panel A). We then found the number of viable bacteria following cotreatment of 10 M ML141 with rifampin at the IC50 for axenic conditions was similar to the number of bacteria recovered following rifampin alone, indicating no detectable enhancement of rifampin bactericidal activity by ML141 at the concentrations used (Fig. 4, Panel C, P = 0.0005). Cotreatment of simvastatin with rifampin fails to improve rifampin efficacy We went on to explore the hypothesis that cotreatment of rifampin with the host-directed therapeutic simvastatin might achieve similar enhancement in the clearance of intracellular infection. The hypothesis was based on our earlier findings that 1.0 M simvastatin decreases intracellular infection in part by sequestering host GTPases, including CDC42 [13], the host-directed target of ML141 [36,37]. Moreover, both ML141 and simvastatin disrupt actin dynamics necessary for host cell invasion by S. aureus and by Streptococcus pyogenes [26,35]. Contrary to the hypothesis, we found no enhancement of bacterial clearance was achieved by cotreatment of rifampin with simvastatin (Fig. 5, Panel A, P = 0.0124). Differential effect of simvastatin on HEK 293-A cell membrane permeability is reversed by cotreatment We were curious to understand the differential effect between simvastatin and ML141. In earlier work, we [35] and others [38] had found that simvastatin can induce host cell membrane permeability in specific cell types. We examined the effect of simvastatin on HEK 293-A cell membrane permeability and the effect of rifampin cotreatment. We found in contrast to ML141 (Fig. 3, Panel B), simvastatin treatment increased HEK 293-A cell permeability (Fig. 5, Panel B, P = 0.0042). We also found the increase in membrane permeability in response to simvastatin was reversed by cotreatment with rifampin (Fig. 5, Panels B and C). Reviewer 2 (Anonymous) Basic reporting 1) Provide the structure of rifampin, simvastatin, and ML141 as a supplementary file.  These structures have been added to the manuscript (as supplementary files). 2) Describe what is ML141 and its mechanism of action in detail.  The authors apologize for not having provided these details in the original submission. The details are included in the following section: Lines 287-289 This is in contrast to ML141, an allosteric inhibitor with specificity for human CDC42 [36, 37]. ML141 dissociates GTP within the activation site of CDC42, decreasing CDC42 activation [36]. 3) Flow cytometer graphs should be provided and discussed.  Representative flow cytometer graphs have been added as: Figure 3 Panel C and Figure 5 Panel C and are discussed as follows: Lines 207-210 The percentage of propidium iodide positive HEK 293-A cells, an indication of membrane permeability, remained at DMSO control levels in response to each treatment, indicating HEK 293-A host cell membrane integrity is maintained following incubation with all treatments (Fig. 3, Panels B and C). Lines 250-251 We also found the increase in membrane permeability in response to simvastatin was reversed by cotreatment with rifampin (Fig. 5, Panels B and C). 4) Add literature regarding the use of these drugs against S. infections.  The authors have added the following to the Introduction regarding the use of these drugs against S. infections: Lines 78-89 For nearly two decades, researchers have examined statin drugs as potential host-directed therapeutics for infection, including invasive infection by S. aureus [27-31]. Our work had indicated therapeutic benefit may include limiting the spread of infection by reducing host cell invasion [32-34]. We identified an underlying mechanism where simvastatin reduces invasion of S. aureus into host cells by sequestering small-GTPases CDC42, RAC, and RHO in the cytosol [13]. In turn, this sequestration limits actin stress fiber disassembly and reduces fibronectin binding at α5β1 [35]. We went on to investigate whether targeted inhibition of host CDC42 would be sufficient to limit invasion. We discovered that ML141, a small molecule inhibitor with specificity for host CDC42 [36, 37], decreased host cell invasion [26]. Similar to simvastatin, we found ML141 inhibition of invasion is associated with diminished reordering of actin stress fibers and with decreased fibronectin binding at the host cell membrane. Experimental design 1) The experiments were performed only with one strain of S. aureus; please include a minimum 3 strains of S. aureus in the study to validate the findings. If possible, antibiotic rifampin-resistant strain should be included as a control.  The authors request leniency. The senior author has taken on an administrative role and closed the research lab in the summer of 2020. It is not feasible to conduct additional experiments. The authors acknowledge the importance of the Reviewer’s requirement. In previous work, the research group had investigated not only the methicillin-susceptible S. aureus (MSSA) strain used in the current study (ATCC 29213), but also MRSA252 NRS71, a hospital strain, and NRS123, a community-acquired methicillin resistant S. aureus (MRSA) strain. ML141 inhibition of host cell invasion by these clinically relevant, invasive MRSA strains was comparable to inhibition of the MSSA strain (Curr Pharm Biotechnol. 2014 ; 15(8): 727–737). Recognizing the earlier work did not include assessment of cotreatment efficacy, the earlier work demonstrated consistency across invasive strains in response to the same concentration of ML141 used in the current study. 2) What are the minimum inhibitory values of the rifampin, simvastatin, and ML141 against S. aureus in vitro? Please provide in vitro synergistic study data by incubating the bacteria with R, S, ML alone, and in combination against the S. aureus strains (minimum 3 strains should be included.). In earlier work, bactericidal activity by ML141 was not detected at the concentration used (Curr Pharm Biotechnol. 2014 ; 15(8): 727–737). An MIC was not determined. Of the structural analogs, only RSM26 demonstrated bactericidal activity at the concentration examined (same publication). Below are reported MIC values for rifampin and simvastatin in vitro. The authors request leniency in performing the synergistic study due to the lab closure. Minimal Inhibitory Concentration (MIC) toward S. aureus Reference Rifampin 0.004-0.03 g/ml (MIC reportedly is much lower than concentration needed for bactericidal activity) Antimicrob Agents Chemother 1983 Apr;23(4):571-6. Simvastatin 0.062 mg mL-1 (ATCC 29213) (135 M) Int J Nanomedicine 2019 Oct 3;14:7975-7985 3) For the co-treatment assay, what is the basis to select the following concentrations? ML141 (10 μM), simvastatin (1 μM), rifampin (0.01 mg/L), ML141 (10 μM) with rifampin (0.01 mg/L), or simvastatin (1 μM) with rifampin (0.01 mg/L) ? The authors apologize for not having clarified the rationale for the selection of each concentration. The rationales have been added to the manuscript in the following sections: ML141 (10 M) Lines 178-182 Consistent with what we had found in human umbilical vein endothelial cells [26], pretreatment of HEK 293-A cells with 10 μM ML141 reduced the number of intracellular bacteria by an average of 40% when compared to pretreatment with the negative, vehicle control DMSO (Fig. 2, Panel A, P = 0.0169). Therefore, this concentration of ML141 served as the positive control for all subsequent experiments. Simvastatin (1.0 M) Lines 236-238 The hypothesis was based on our earlier findings that 1.0 M simvastatin decreases intracellular infection in part by sequestering host GTPases, including CDC42 [13], the host-directed target of ML141 [36, 37]. Rifampin (0.01 mg/L) Lines 172-177 We first determined the half maximal inhibitory concentration (IC50) of rifampin in a model cell line, HEK 293-A. The rifampin IC50 was 0.004 mg/L when HEK 293-A cells were incubated with bacteria at MOI 2 and the IC50 increased to 0.01 mg/L when the MOI was increased to 100 (Fig. 1). Thus, a higher concentration of antibiotic was needed to clear a larger intracellular bacterial population. All subsequent experiments were performed at MOI 100 to ensure there were sufficient numbers of colonies in the cotreatment group for statistical analysis. 4) HEK Cells should be stained with fluorescent dyes and imaged and the live dead cells images should be provided in the support of the experimental claims.  Recognizing the validation this study would yield, the authors request leniency due to the lab closure. 5) In Rifampin bactericidal activity assay, different concentrations of antibiotics and the ML141 should be used to confirm that the bactericidal activity is not enhanced.  The authors recognize the contribution these data would make to fully characterize the response to cotreatment; yet, the authors request leniency due to the lab closure. Validity of the findings 1) Why simvastatin and rifampin yield differential results and rifampin-ML141 not? The authors should discuss this point in detail and provide multiple hypothesis against these results. These hypotheses should include literature citations.  The authors have added to the manuscript the following discussion of this point and have included literature citations: Lines 263-268 Statins exert pleiotropic effects through inhibition of multiple intermediates within the cholesterol biosynthesis pathway, including loss of membrane integrity through decreased synthesis of the intermediate mevalonate [42, 43]. Yet, loss of membrane integrity is not a universal response to statins [13, 44]. Thus, it is plausible that examination of simvastatin across multiple cell types, timepoints and dosages may have yielded a response similar to ML141. Reviewer 3 (Gunjan Arora) Basic reporting The manuscript is simple and well written.  Experimental design Research question addressed in this manuscript is provides new knowledge on host directed therapies.  Validity of the findings Overall the data is well represented.  Comments for the Author The manuscript by Evans et al., discusses the effect of rifampin therapy in combination with host directed ML141 therapy in S. aureus infection treatment. Overall the topic is interesting and the results are useful for further research. My specific comments are: 1. In the abstract, results should be more clearly written with exact changes observed with cotreatment as compared to mono-therapy or no treatment. The authors apologize for the lack of clarity. The following is the edited Results section of the abstract that includes exact changes observed with cotreatment compared to monotherapy or no treatment. Lines 33-40 Results. We found cotreatment of ML141 with rifampin reduced intracellular infection nearly 85 % when compared to the no treatment control. This decrease more than doubled the average 40 % reduction in response to rifampin monotherapy. In contrast, cotreatment of simvastatin with rifampin failed to improve rifampin efficacy. Also in contrast to ML141, simvastatin increased propidium iodide (PI) positive cells, from an average of 10 % in control HEK 293-A cells to nearly 20 % in simvastatin-treated cells, indicating an increase in host cell membrane permeability. The simvastatin-induced increase was reversed to control levels by cotreatment of simvastatin with rifampin. 2. In methods, the co-treatment method is not clear. The authors have discussed both pretreatment and cotreatment in this section, which is getting a bit confusing. Please modify. The authors apologize for not having made these treatment protocols clear. Below is the revision to the Methods section. Lines 127-137 ML141 Pretreatment Assay HEK 293-A cells were plated as described above. The following day, HEK 293-A cells were pretreated with DMSO (0.4 %) or with ML141 (10 μM; 37°C, 5% CO2, 24 h; structure provided in supplemental file). The next day, pretreated HEK 293-A cells were inoculated (3 x 107 CFU, 1 hr, 37°C, 5 % CO2) in 10 % FBS/PBS. Intracellular bacteria were isolated and quantified as described above. Cotreatment Assay HEK 293-A cells were plated as described above. Treatments were performed as outlined in Table 1 at these concentrations: DMSO (0.4 %), ML141 (10 μM), simvastatin (1 μM; structure provided in supplemental file), rifampin (0.01 mg/L). Following post-treatment, intracellular bacteria were isolated and quantified as described. 3. The authors have performed pretreatment assays with ML141. How is it relevant in real infection cases, would the patients will keep taking ML141 as precautionary measure? The authors apologize for not having clarified the treatment would not necessarily be used as a precautionary measure. The following has been added to the manuscript to clarify this point: Lines 183-186 Pretreatment is not used as a model for prophylactic clinical applications, but rather, as a model for limiting the spread of infection post-onset and diagnosis, given the evidence that invasive S. aureus strains successfully exit infected host cells to initiate new infection [11, 40, 41]. 4. Results, line 163-165, “We next assessed whether the reduction in the number of intracellular bacteria was due to…” The authors need to explain the results in more detail instead of directly writing the final results, how they address this question and what experiments were performed.  The authors apologize for the lack of explanation throughout the results and have rewritten the Results Section in its entirety. The edited version that corresponds to line 163-165 in the original is below: Lines 212-231 Underlying mechanisms for enhanced clearance by the co-treatment are not associated with appreciable ML141 bactericidal activity or with enhanced rifampin bactericidal activity Although ML141 has been assessed repeatedly for bactericidal activity and no such activity has been detected toward multiple S. aureus or Streptococcus pyogenes strains [26,35], we saw it as important to verify this under the experimental conditions of the rifampin studies. We inoculated HEK 293-A media containing the DMSO control or 10 M ML141 for the same length of time used in the invasion assay. We found that the number of viable bacteria following ML141 treatment was similar to the control treated group, indicating ML141 exhibited no bactericidal activity under the conditions used for the invasion assay (Fig. 4, Panel B). To assess whether co-treatment with ML141 might somehow enhance rifampin bactericidal activity, we first needed to determine the IC50 of rifampin on S. aureus ATCC 29213 growth in axenic (host cell-free) conditions. This was necessary because the rifampin IC50 for intracellular bacteria (0.01 mg/L, reported above) was expected to exceed the IC50 needed under axenic conditions. Consistent with this expectation, the IC50 of rifampin in the absence of host cells was 0.002 mg/L (Fig. 4, Panel A). We then found the number of viable bacteria following cotreatment of 10 M ML141 with rifampin at the IC50 for axenic conditions was similar to the number of bacteria recovered following rifampin alone, indicating no detectable enhancement of rifampin bactericidal activity by ML141 at the concentrations used (Fig. 4, Panel C, P = 0.0005). Article ID: 50566 "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Radon ( 222 Rn) and thoron ( 220 Rn) are radioactive gases emanating from geological materials.</ns0:p><ns0:p>Inhalation of these gases is closely related to an increase in the probability of lung cancer if the levels are high. The majority of studies focus on radon, and the thoron is normally ignored due to its short half-life (55.6 s). However, also the thoron decay products can cause a significant increase in dose. In buildings with high radon levels, the main mechanism for entry of radon is the pressuredriven flow of soil gas through cracks in the floor. Both radon and thoron can also be released from building materials to the indoor atmosphere.</ns0:p><ns0:p>In this study, the 222 Rn and 220 Rn exhalation rates (mass and surface) and the emanation coefficients of an extended variety of common building materials manufactured in the Iberian Peninsula (Portugal and Spain) but exported and used in all countries of the world were determined by using active measuring system. Radon and thoron emission from samples collected in the closed chamber was measured by an active method that uses a continuous radon/thoron monitor. The correlations between exhalation rates of these gases and their parent nuclide exhalation (radium/thorium) concentrations were examined. Finally, indoor radon and thoron and the annual effective dose were calculated from radon/thoron concentrations in the closed chamber. Zircon is the material with the highest concentration values of 226 Ra and 232 Th and the exhalation and emanation rates. Also in the case of zircon and some granite, the annual effective dose was higher than the annual exposure limit for the general public of 1 mSv y -1 , recommended by the European regulations.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Rn) are radioactive gases emanating from geological materials. Inhalation of these gases is closely related to an increase in the probability of lung cancer if the levels are high. The majority of studies focus on radon, and the thoron is normally ignored by its short half-life (55.6 s). However, also the thoron decay products can cause a significant increase in dose. In buildings with high radon levels, the main mechanism for entry of radon is pressure-driven flow of soil gas through cracks in the floor.</ns0:p><ns0:p>Both radon and thoron can also be released from building materials to the indoor atmosphere. In this work, we study the radon and thoron exhalation and emanation properties of an extended variety of common building materials manufactured in the Iberian Peninsula (Portugal and Spain) but exported and used in all countries of the world.</ns0:p><ns0:p>Radon and thoron emission from samples collected in the closed chamber was measured by an active method that uses a continuous radon/thoron monitor. The correlations between exhalation rates of these gases and their parent nuclide exhalation (radium/thorium) concentrations were examined. Finally, indoor radon and thoron and the annual effective dose were calculated from radon/thoron concentrations in the closed chamber. Zircon is the material with the highest concentration values of 226 Ra and 232 Th and the exhalation and emanation rates. Also in the case of zircon and some granites, the annual effective dose was higher than the annual exposure limit for the general public of 1 mSv y -1 , recommended by the European regulations.</ns0:p></ns0:div> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Radon and thoron are significant contributors to the average dose from natural background sources of radiation. They represent approximately half of the estimated dose from exposure to all natural sources of ionizing radiation <ns0:ref type='bibr'>(UNSCEAR, 2008)</ns0:ref>.</ns0:p><ns0:p>Inhalation of these radioactive gases and their decay products can cause health risks, especially in poor ventilate areas. Long-term exposure to high levels of radon/thoron in home and working area increases risk of developing lung cancer <ns0:ref type='bibr'>[1,</ns0:ref><ns0:ref type='bibr'>3]</ns0:ref>. Radon is the second leading cause of increase of the probability of lung cancer after tobacco smoke <ns0:ref type='bibr'>(World Health Organization, 2009)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed After its formation, these two radioisotopes are susceptible to escape, firstly from the grains constituting the material (known as emanation), and secondly, from the surface of the material (known as exhalation). These parameters depend, among other factors, on the half-life, consequently affecting the accumulation rate of these gaseous radioisotopes in indoor environments, and therefore, to the exposure of the human body to radiation. For radon, the halflife is 3.825 days while for thoron, just 55.6 s so, due to this difference, the effective dose from thoron and its progeny ( 212 Pb and 212 Bi) is estimated around of 10 % of that due to radon and its progeny ( 214 Pb and 214 Bi) in indoor environments (United Nations Scientific Committee on the Effects of Ionisin Radiation (UNSCEAR), Sources, Effects and Risks of Ionizing Radiation, Report to the General Assembly, 2016).</ns0:p><ns0:p>These factors lead to a complicated thoron measurement technique resulting in, the majority of the existing studies focus on the radon <ns0:ref type='bibr'>[2]</ns0:ref><ns0:ref type='bibr'>[3]</ns0:ref><ns0:ref type='bibr'>[4]</ns0:ref><ns0:ref type='bibr'>[5]</ns0:ref><ns0:ref type='bibr'>[6]</ns0:ref><ns0:ref type='bibr'>[7]</ns0:ref><ns0:ref type='bibr'>[8]</ns0:ref><ns0:ref type='bibr'>[9]</ns0:ref><ns0:ref type='bibr'>[10]</ns0:ref><ns0:ref type='bibr'>[11]</ns0:ref><ns0:ref type='bibr'>[12]</ns0:ref>. Many of these studies also include measures of 40 K, 226 Ra and 232 Th and risk indexes definitions trying to evaluate the radiological health hazards of these radionuclides <ns0:ref type='bibr' target='#b47'>(Turhan &amp; G&#252;nd&#252;z, 2008;</ns0:ref><ns0:ref type='bibr' target='#b52'>De With, De Jong &amp; R&#246;ttger, 2014;</ns0:ref><ns0:ref type='bibr' target='#b21'>Kumar et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b18'>Kayak&#246;k&#252;, Karatepe &amp; Do&#287;ru, 2016;</ns0:ref><ns0:ref type='bibr' target='#b25'>Madruga et al., 2018)</ns0:ref> or the effective dose due to radon and its progeny <ns0:ref type='bibr' target='#b42'>(Sabiha-Javied, Tufail &amp; Asghar, 2010)</ns0:ref>.</ns0:p><ns0:p>Nevertheless, despite thoron indoor concentration is generally lower than for the radon, the 212 Pb thoron progeny (half-life of 10.6 h) can accumulate to significant levels in breathable air, aggravating its inhalation risk <ns0:ref type='bibr'>(World Health Organization, 2009)</ns0:ref>. Some studies <ns0:ref type='bibr'>[23]</ns0:ref><ns0:ref type='bibr'>[24]</ns0:ref><ns0:ref type='bibr'>[25]</ns0:ref> have demonstrated that thoron concentrations can be comparable to radon and its progeny in some areas of elevated radiological risk. Furthermore, computational studies (de With &amp; de Jong, 2011) taking into account factors such as the ventilation and air exchange, the building dimensions, dispersion and deposition, mitigation measures, and material properties indicates that thoron effective doses can reach the 35 % of the total contribution. Therefore, these studies demonstrate the recent and growing interest that has emerged in recent decades by the study of thoron <ns0:ref type='bibr'>[27]</ns0:ref><ns0:ref type='bibr'>[28]</ns0:ref><ns0:ref type='bibr'>[29]</ns0:ref><ns0:ref type='bibr'>[30]</ns0:ref><ns0:ref type='bibr'>[31]</ns0:ref><ns0:ref type='bibr'>[32]</ns0:ref><ns0:ref type='bibr'>[33]</ns0:ref><ns0:ref type='bibr'>[34]</ns0:ref><ns0:ref type='bibr'>[35]</ns0:ref> in building materials <ns0:ref type='bibr'>[16,</ns0:ref><ns0:ref type='bibr'>19,</ns0:ref><ns0:ref type='bibr'>36,</ns0:ref><ns0:ref type='bibr'>37]</ns0:ref> although no further studies has been reported yet focusing in the assessment of the thoron risk index in the building materials used in buildings.</ns0:p><ns0:p>Among the methods to measure both exhalation rate and emanation factor of radon and thoron isotopes in building materials, passive methods, that use solid-state nuclear track detector, accumulation chamber methods and active methods with radon/thoron monitors, can be found <ns0:ref type='bibr' target='#b58'>(Zhang et al., 2012)</ns0:ref>.</ns0:p><ns0:p>In previous work, the gamma radiations emitted from 226 Ra, 232 Th and 40 K for some of these materials were studied, as well as the radiological health hazards associated with the external gamma radiation <ns0:ref type='bibr' target='#b25'>(Madruga et al., 2018)</ns0:ref>. In another study <ns0:ref type='bibr' target='#b11'>(Frutos-Puerto et al., 2018)</ns0:ref>, a technique of measurement of thoron had been developed and applied to the analysis of exhalation of 5 materials. In the present work, expanded with more materials, we study the radon and thoron exhalation and emanation properties of an extended variety of common building materials used in the Iberian Peninsula (Portugal and Spain). The correlations between exhalation rates of these gases and their parent nuclide exhalation (radium/thorium) concentrations were examined.</ns0:p><ns0:p>Furthermore, indoor radon/thoron and the annual effective dose were calculated from radon/thoron concentrations in the closed chamber. Measurements were carried out by an active method that uses a continuous radon/thoron monitor RTM1688-2 (SARAD GmbH, Dresden, Germany).</ns0:p></ns0:div> <ns0:div><ns0:head>Materials and methods</ns0:head></ns0:div> <ns0:div><ns0:head>Materials and sample preparation</ns0:head><ns0:p>Forty-one samples from quarries and suppliers of the most commonly used building materials manufactured in the Iberian Peninsula were collected. The mass of each sample ranged between 1 and 5 kg. Fig. <ns0:ref type='figure'>1</ns0:ref> shows the geographical origin of the materials. The materials were divided in two classes: materials coming from natural sources, NM, naturally occurring radioactive materials (NORM) incorporating waste after industrial processing, PM ('Directive. 2013/59/Euratom of 5 December 2013,' 2014). Within each classification of materials are found: Materials type NM:</ns0:p><ns0:p>&#8226; Concretes. Used in bulk amounts:</ns0:p><ns0:p>-Conventional.</ns0:p><ns0:p>-100% of the natural aggregate becomes electrical furnace slags.</ns0:p><ns0:p>-100% of the natural aggregate becomes blast furnace slags.</ns0:p><ns0:p>-Self-compacting.</ns0:p><ns0:p>-High-resistance.</ns0:p><ns0:p>-Mortars of resistance 5 and 7.5, respectively.</ns0:p><ns0:p>&#8226; Cements. Used in bulk amounts and superficial applications:</ns0:p><ns0:p>-Type I Portland cement with less than 3% fly ash.</ns0:p><ns0:p>-White cement.</ns0:p><ns0:p>-Cement glue.</ns0:p><ns0:p>-Rapid cement.</ns0:p><ns0:p>&#8226; Natural stones. Used as bulk and superficial products:</ns0:p><ns0:p>-Marble.</ns0:p><ns0:p>-Granite.</ns0:p><ns0:p>-Slate.</ns0:p><ns0:p>&#8226; Ceramic tiles as refractory and ceramic products to cover floors and walls, mainly:</ns0:p><ns0:p>-Tiles</ns0:p><ns0:p>&#8226; Raw materials of very different types and composition:</ns0:p><ns0:p>-Wood collected from Eucalyptus and Castahea Sativa trees.</ns0:p><ns0:p>-Aggregates as sand or clay bricks.</ns0:p><ns0:p>-Zircon.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials type PM:</ns0:head><ns0:p>&#8226; Industrial products resulting from the sulphates industry of the North of Spain:</ns0:p><ns0:p>-Gypsum -Plastic cement Sample preparation consisted in to crushing and drying building materials in an oven for 48 h at 105 o C, prior to its grounding and sieving (2 mm particle size).</ns0:p></ns0:div> <ns0:div><ns0:head>Gamma spectroscopic analysis</ns0:head><ns0:p>To carry out the &#61543;-emissions measurements, the milled samples were dried and placed in 160 cm 3 cylindrical containers made of plastic or in 1000 cm 3 Marinelli beakers, both, hermetically sealed for 28 or more days. This period is sufficient for equilibrium to occur between the radioisotopes of 226 Ra and 232 Th initially contained in the material and their decay products.</ns0:p><ns0:p>To obtain the 232 Th and 226 Ra content an HPGe semiconductor detector was employed according to the methodology followed by <ns0:ref type='bibr' target='#b25'>Madruga et al. (Madruga et al., 2018)</ns0:ref>. The 232 Th activity was determined by means of the &#947;-emissions of 228 Ac (911 keV) and 208 Tl (583.01 keV) and that of 226 Ra by means of those from 214 Bi (609.3 and 1764.5 keV) and 214 Pb (351.9 keV) assuming that both radioactive series are left in secular equilibrium.</ns0:p><ns0:p>A 50% relative efficiency broad energy HPGe detector (Canberra BEGe model BE5030), with an active volume of 150 cm 3 and a carbon window was used for the gamma spectrometry measurements. A lead shield with copper and tin lining shields the detector from the environmental radioactive background. Standard nuclear electronics was used and the software Genie 2000 (version 3.0) was employed for the data acquisition and spectral analysis. The detection efficiency was determined using NIST-traceable multi-gamma radioactive standards (Eckert &amp; Ziegler Isotope Products) with an energy range from 46.5 keV to 1836 keV and customized in a waterequivalent epoxy resin matrix (density of 1.15 g cm -3 ) to exactly reproduce the geometries of the samples. GESPECOR software (version 4.2) was used to correct for matrix (self-attenuation) and coincidence summing effects, as well as to calculate the efficiency transfer factors from the calibration geometry to the measurement geometry (whenever needed). The stability of the system (activity, FWHM, centroid) was checked at least once a week with a 152 Eu certified point source. <ns0:ref type='bibr' target='#b29'>(Mere&#353;ov&#225;, W&#228;tjen &amp; Altzitzoglou, 2012;</ns0:ref><ns0:ref type='bibr' target='#b57'>Xhixha et al., 2017)</ns0:ref>.In summary, the activity concentration for 232 Th and 226 Ra (A) was calculated by the following expression:</ns0:p><ns0:formula xml:id='formula_0'>&#119862; = &#119873; &#119905; &#119875; &#119872; &#120598; &#119891; (1)</ns0:formula><ns0:p>where N stands for net counts, t for data collection time, P for emission probability, M for mass of the sample and for efficiency of the detector for the corresponding peak. Besides, uncertainty &#120598; &#119891; in the yield is also include since several &#61543;-ray peaks were used for the calculation of 232 Th and 226 Ra activity.</ns0:p></ns0:div> <ns0:div><ns0:head>Determination of massic exhalation rate and emanation factor</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Exhalation is the amount of radon (radon activity) as obtained from a given layer (geological material on the surface/surface exposure) mainly the outer thinner part of the crust and it is given in Bq h -1 , according to the Netherlands Standardization Institute ('Netherlands Standardization Institute. Dutch Standard: Radioactivity measurement. Determination method of the rate of the radon exhalation of dense building materials,' 2001). Exhalation can be related to the mass of the samples (massic radon/thoron exhalation, and its value is expressed Bq kg -1 h -1 ). The method already referred <ns0:ref type='bibr'>[5,</ns0:ref><ns0:ref type='bibr'>39]</ns0:ref> and similar to that of other authors <ns0:ref type='bibr' target='#b14'>(Hassan et al., 2011)</ns0:ref> was employed to assess the massic exhalation of 222 Rn and 220 Rn and it is schematized in Fig. <ns0:ref type='figure' target='#fig_4'>3</ns0:ref>.</ns0:p><ns0:p>The calculation of 222 Rn and 220 Rn exhalation was carried out according to the expressions presented in the Mir&#243; et al. <ns0:ref type='bibr' target='#b31'>(Miro et al., 2014)</ns0:ref> from the formula of the temporal variation of the radon concentration C(t), in Bq m -3 :</ns0:p><ns0:formula xml:id='formula_1'>&#119889;&#119888; &#119889;&#119905; = &#119864;&#119872; &#119881; -&#120582;&#119862; -&#120572;&#119862;<ns0:label>(2)</ns0:label></ns0:formula><ns0:p>where E (Bq kg -1 h -1 ) is the radon-specific exhalation rate, M (kg) the mass of the sample, V (m 3 ) the air volume of the container, &#61548; (h -1 ) the 222 Rn or 220 Rn decay constant and &#945; (h -1 ) the leakage rate from the container. The bound exhalation rate determined by hermetically closing the sample in a container can be equal to the free exhalation corresponding to the actual room conditions only in the case that the sample volume would be less than the one-tenth of the container volume. Under these circumstances, the 'back diffusion' effect has no influence on exhalation rate measurements (Krisiuk, E.M., Tarasov, S.I., Seamov, V.P., Shalck, N.I., isachenko, E.P., <ns0:ref type='bibr' target='#b19'>Gomelsky, 1971)</ns0:ref>. The By solving Equation ( <ns0:ref type='formula' target='#formula_1'>2</ns0:ref>), the radon concentration growth as a function of time is given by:</ns0:p><ns0:formula xml:id='formula_2'>&#119862;(&#119905;) = &#119864;&#119872;[1 -&#119890; -(&#120582; + &#120572;)&#119905; ] (&#120582; + &#120572;)&#119881; + &#119862; 0 &#119890; -(&#120582; + &#120572;)&#119905; (3)</ns0:formula><ns0:p>being C 0 (Bq m -3 ) the radon concentration at t = 0.</ns0:p><ns0:p>The 222 Rn exhalation (E Rn222 ) and &#945; numeric calculation are made by adjusting by least-squares of the C vs t experimental data to the mathematical function given by equation (3).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed However, due to its short half-life, after the first cycle (2 hours) of measurements, the concentration of thoron in the container will reach its maximum value, remaining constant until the end of the measurements. So, from equation (3) the massic thoron exhalation, E Rn220 , can be calculated from the expression (4), which does not consider &#945; value because it is much smaller than the thoron decay constant, &#955; Rn220 :</ns0:p><ns0:formula xml:id='formula_3'>&#119864; &#119877;&#119899;220 = &#119862; &#119877;&#119899;220 &#120582; &#119877;&#119899;220 &#119881; M (4)</ns0:formula><ns0:p>where (Bq m -3 ) is the average concentration of thoron in the container during the interval &#119862; &#119877;&#119899;220 of measurement from the first cycle of 2 hours.</ns0:p><ns0:p>The emanation factor (amount of radon and thoron atoms that escape from the grains constituting the material into the interstitial space between the grains), &#949; Rn ,, was calculated by the following equation for both radioisotopes <ns0:ref type='bibr' target='#b46'>(Stoulos, Manolopoulou &amp; Papastefanou, 2003)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_4'>&#120576; &#119877;&#119899; = &#119864; &#119877;&#119899; &#119862; &#119894; &#955; &#119889;<ns0:label>(5)</ns0:label></ns0:formula><ns0:p>where is the 226 Ra or 232 Th content (Bq kg -1 ) of the sample for radon and thoron, respectively, &#119862; &#119894;</ns0:p><ns0:p>, the decay constant and the exhalation.</ns0:p><ns0:p>&#955; &#119889; &#119864; &#119877;&#119899; Equation ( <ns0:ref type='formula' target='#formula_4'>5</ns0:ref>) is applicable for all measured building materials, because the dimensions of the samples were chosen to be equal to the diffusion length of these gases for these materials, around 4 cm <ns0:ref type='bibr' target='#b46'>(Stoulos, Manolopoulou &amp; Papastefanou, 2003)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Determination of annual effective dose</ns0:head><ns0:p>The 222 Rn/ 220 Rn content accumulates in the surrounding air in a dwelling room, from building materials, depends on factors such as the room dimension, the parent element concentration, the subsequent exhalation directly from the soil and building materials in walls or soil (radon gain), the air exchange and the isotope radioactive decay. Therefore, building materials may cause an excess in the indoor 222 Rn or 220 Rn activity concentrations, which is described by the following equation <ns0:ref type='bibr' target='#b0'>(Amin, 2015)</ns0:ref>:</ns0:p><ns0:formula xml:id='formula_5'>&#119860; &#119877;&#119899; = &#119864; &#119860; &#119878; &#119881; &#119903; &#120582; &#119907; (6)</ns0:formula><ns0:p>where,</ns0:p><ns0:p>, is the 222 Rn or 220 Rn activity concentration (Bq m -3 ) in the air of the room; is the &#119860; &#119877;&#119899; &#119864; &#119860; surface exhalation rate (Bq m -2 h -1 ); S is the exhalation area (m 2 ); V r is the volume of the room (m 3 ) and is the ventilation rate of the room (h -1 ). Ratio S/V is taken to be 2 and , 0.5 h -1 ( &#120582; &#119907; &#120582; &#119907;</ns0:p><ns0:p>UNSCEAR, 2016). Considering the value of the sample emanation surface in the container (0.0078 m 2 ; circumference of 5 cm 2 ), and the mass of the sample (M), the surface exhalation rate for (&#119864; &#119860; ) the building materials can be calculated, using the following equation:</ns0:p><ns0:formula xml:id='formula_6'>&#119864; &#119860; = &#119864; &#119877;&#119899; &#119872; 0.0078<ns0:label>(7)</ns0:label></ns0:formula><ns0:p>This radon concentration model can then be used to determinate the annual effective doses of 222 Rn by equation ( <ns0:ref type='formula'>8</ns0:ref>), recommended by the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR, 2016):</ns0:p><ns0:formula xml:id='formula_7'>&#119863; &#119877;&#119899;222 = &#119860; &#119877;&#119899;222 &#119865; &#119890; &#119879; &#119886; &#119862;&#119865; &#119877;&#119899;222 (8)</ns0:formula><ns0:p>where D Rn222 is the annual effective dose of 222 Rn (Sv y -1 ); A Rn222 is the activity concentration for 222 Rn (Bq m -3 ); CF Rn222 is the dose conversion factor for 222 Rn progeny (Sv per Bq h m -3 ); F e is the equilibrium factor for 222 Rn and its progeny; and T a is the annual work time. The standard parameters were estimated using the RP 122 publication of <ns0:ref type='bibr'>EC 2002</ns0:ref><ns0:ref type='bibr'>(European Commission, 2002)</ns0:ref>. The values of CF Rn222 were assumed to be 9&#61620;10 -9 Sv per Bq h m -3 and the T a , 7000 h y -1 .</ns0:p><ns0:p>The value of F e was assumed to be 0.4 as reported in ( UNSCEAR, 2008).</ns0:p><ns0:p>Similarly, for 220 Rn:</ns0:p><ns0:formula xml:id='formula_8'>&#119863; &#119877;&#119899;220 = &#119860; &#119877;&#119899;220 &#119865; &#119890; &#119879; &#119886; &#119862;&#119865; &#119877;&#119899;220<ns0:label>(9)</ns0:label></ns0:formula><ns0:p>where, D Rn220 is the annual effective dose of 220 Rn (Sv y -1 ); A Rn220 is the activity concentration for 220 Rn (Bq m -3 ); CF Rn220 is the dose conversion factor for 220 Rn progeny (40&#61620;10 -9 Sv per Bq h m -3 ) and T a is the annual work time, 7000 h y -1 <ns0:ref type='bibr'>(European Commission, 2002)</ns0:ref>. F e is the equilibrium factor for 220 Rn and its progeny, 0.1 ( UNSCEAR, 2008).</ns0:p><ns0:p>However, since the diffusion length of 220 Rn is very short it is complex and ambiguous to calculate the internal exposure due to 220 Rn exhaling from the building material. The indoor thoron concentration in air depends on the distance from the wall <ns0:ref type='bibr'>[21,</ns0:ref><ns0:ref type='bibr'>23]</ns0:ref> as presented in the following equation:</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_9'>&#119860; &#119877;&#119899;220 (&#119883;) = &#119864; &#119860;&#119877;&#119899;220 &#120582; &#119877;&#119899;220 &#119863; &#119890;&#119891;&#119891; exp ( - &#120582; &#119877;&#119899;220 &#119863; &#119890;&#119891;&#119891; &#119883; ) (10)</ns0:formula><ns0:p>where, is the 220 Rn concentration at a distance, X, from the wall. is the 220 Rn &#119860; &#119877;&#119899;220 (&#119883;) &#119864; &#119860;&#119877;&#119899;220 estimated surface exhalation rate by equation ( <ns0:ref type='formula' target='#formula_6'>7</ns0:ref>), D ef is the effective diffusion coefficient herein</ns0:p><ns0:formula xml:id='formula_10'>taken as 1.8 m 2 h -1 [21],</ns0:formula><ns0:p>is the decay constant of 220 Rn, 45 h -1 . &#120582; &#119877;&#119899;220</ns0:p><ns0:p>It is reasonable to assume that the human respiratory organs are not more than 40 cm distance from the wall. Therefore, the 220 Rn concentration at the distance of 40 cm calculated by equation <ns0:ref type='bibr'>(10)</ns0:ref> , is used to determinate the annual effective doses of 220 Rn with equation ( <ns0:ref type='formula' target='#formula_8'>9</ns0:ref>). , &#119860; &#119877;&#119899;220</ns0:p></ns0:div> <ns0:div><ns0:head>Results and discussion</ns0:head><ns0:p>The results of activity concentration for 226 Ra, , massic exhalation, E Rn222 , and emanation &#119862; &#119877;a factor, &#949; Rn222 , for 222 Rn are summarised in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>.</ns0:p><ns0:p>In all samples, activity concentration for radium was above the detection limit (DL) except for the wood sample. In general, these results are comparable to those measured in a worldwide scale <ns0:ref type='bibr'>[2,</ns0:ref><ns0:ref type='bibr'>[45]</ns0:ref><ns0:ref type='bibr'>[46]</ns0:ref><ns0:ref type='bibr'>[47]</ns0:ref><ns0:ref type='bibr'>[48]</ns0:ref>. Thus, the values for radium content in building materials are less than the permissible value (370 Bq kg -1 ), which is acceptable as a safe limit (Group Experts of the OECD Nuclear Energy <ns0:ref type='bibr'>Agency, 1979)</ns0:ref>. The only exception was in the radium concentration in zircon, the highest value for the mean concentration was 2070 Bq kg -1 .</ns0:p><ns0:p>In many samples, the exhalation rate was lower than the DL (because of E Rn222 &lt; DL) with exception of all samples of slate, granite and zircon. The maximum value on average was obtained for zircon, 429 mBq kg -1 h -1 , which is much higher than that found for the aggregate and the granites. The values of exhalation rates reported in Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> correspond well with the values reported by other authors <ns0:ref type='bibr'>[15,</ns0:ref><ns0:ref type='bibr'>[50]</ns0:ref><ns0:ref type='bibr'>[51]</ns0:ref><ns0:ref type='bibr'>[52]</ns0:ref><ns0:ref type='bibr'>[53]</ns0:ref>. The variation in radon exhalation rates (one order of magnitude, in some cases) can be attributed to variations in radium concentrations, porosity, and surface crystallography. The emanation factor range from 0.2 % to 22.0 % for ceramic and aggregates respectively. These values are similar to the measured in worldwide scales <ns0:ref type='bibr'>[2,</ns0:ref><ns0:ref type='bibr'>15,</ns0:ref><ns0:ref type='bibr'>46,</ns0:ref><ns0:ref type='bibr'>49]</ns0:ref>.</ns0:p><ns0:p>The results of activity concentration for 232 Th, C Th , massic exhalation, E Rn220 , and emanation factor, &#949; Rn220 , for 220 Rn are summarised in Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>.</ns0:p><ns0:p>The highest mean value for 232 Th activity concentration is shown by zircon (340 Bq kg -1 ), and the lowest mean value is obtained for wood (0.6 Bq kg -1 ). The mean values of the 220 Rn massic exhalation rate range from 2.2 of the ceramic to 169 Bq kg -1 h -1 for zircon, respectively. The results show that the thoron exhalation rate is higher in zircon samples and lower in ceramic samples. This can presumably be explained by the different distributions of 224 Ra parent element in the different types of samples. It should be noted how the difference among the values of exhalation rate in granites (range from 2.6 to 144 Bq kg -1 h -1 ) reveal their different mineralogical composition.</ns0:p><ns0:p>The emanation factor range from 0.3 % to 29 % for ceramic and wood, respectively.</ns0:p><ns0:p>The ranges of results of all these parameters are in good agreement with the values reports by other authors <ns0:ref type='bibr'>[29,</ns0:ref><ns0:ref type='bibr'>54]</ns0:ref>.</ns0:p><ns0:p>A correlation study of 222 Rn mass exhalation rate with respect to 226 Ra content, as shown in Fig. <ns0:ref type='figure'>4a</ns0:ref>, showed a good linear correlation coefficient (R 2 = 0.9961). These results show that the 222 Rn mass exhalation rate increases as the 226 Ra content is higher in the samples. This good linear correlation has already been observed by other authors, some of them with values very close to 1 <ns0:ref type='bibr' target='#b0'>(Amin, 2015)</ns0:ref>. As could be expected (Fig. <ns0:ref type='figure'>4b</ns0:ref>), no correlation (R 2 = 0.0109) was found between the 222 Rn emanation factor and the 226 Ra content.</ns0:p><ns0:p>A similar correlation of 220 Rn mass exhalation rate with 232 Th content is shown in Fig. <ns0:ref type='figure'>5a</ns0:ref>, which shows a more weak correlation between the two quantities (R 2 = 0.8336). These results show that the 220 Rn mass exhalation rate increases for samples with higher 232 Th contents, as observed before for the 222 Rn exhalation rate and 226 Ra contents.</ns0:p><ns0:p>Moreover, as could be expected (Fig. <ns0:ref type='figure'>5b</ns0:ref>), no correlation (R 2 = 0.0115) was found between the 220 Rn emanation factor and the 232 Th content. Finally, no correlation (R 2 = 0.118) was found between the 222 Rn emanation factor and the 220 Rn emanation factor as shown in Fig. <ns0:ref type='figure'>5c</ns0:ref>.</ns0:p><ns0:p>The results obtained for indoor contribution, surface exhalation rate, activity concentration in the air of the room, and annual effective dose, for the different building materials had been shown in Tables <ns0:ref type='table' target='#tab_9'>3 and 4</ns0:ref> for 222 Rn and 220 Rn, respectively. Therefore, Table <ns0:ref type='table' target='#tab_6'>3</ns0:ref> shows that the mean values of 222 Rn surface exhalation rates varied from 9.2 to 3206 mBq m -2 h -1 for ceramic and zircon, respectively. The 222 Rn contribution of building materials to indoor 222 Rn considering the model room mentioned above, range from 0.04 for ceramic samples to 13 Bq m -3 for zircon. As a result of this, the annual effective dose ranged from 0.9 &#181;Sv y -1 for ceramic to 323 &#181;Sv y -1 for zircon.</ns0:p><ns0:p>These values are in agreement with the worldwide range <ns0:ref type='bibr'>[2,</ns0:ref><ns0:ref type='bibr'>55]</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>In the case of 220 Rn (see Table <ns0:ref type='table' target='#tab_9'>4</ns0:ref>), the surface exhalation rate average varied from 22 to 1264 Bq m -2 h -1 for cement and zircon respectively. Its contribution of building materials to indoor 220 Rn at 40 cm of the wall considering the model mentioned above, range from 2.0 for the cement to 112 Bq m -3 for zircon. Mean values of the annual effective dose ranged from 16 &#181;Sv y -1 for gypsum to 1300 &#181;Sv y -1 for zircon. These values are similar to those found by other authors for building materials <ns0:ref type='bibr' target='#b49'>(Uji&#263; et al., 2010)</ns0:ref>. However, estimation of annual effective dose from indoor thoron indicated the mean value of zircon and some values of granites had been higher than the annual exposure limit for the general public of 1 mSv y -1 , recommended by European Directive 2013/59/Euratom <ns0:ref type='bibr'>[40]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In this study, the radon and thoron exhalation and emanation properties of building materials commonly used in the Iberian Peninsula (Portugal and Spain) were measured by using an active method with a continuous radon/thoron monitor. The correlations between exhalation rates of these gases and their parent nuclide exhalation (radium/thorium) concentrations were examined. Finally, on estimation the indoor radon/thoron, the annual effective dose was calculated.</ns0:p><ns0:p>In general, 226 Ra content in building materials is less than the permissible value, 370 Bq kg -1 , except for zircon, which means value was 2100 Bq kg -1 . For this material the maximum value on average of 222 Rn massic exhalation rate (429 mBq kg -1 h -1 ) was also obtained. The emanation factor 222 Rn/ 226 Ra ranges from 0.2 % to 22.0 % for ceramic and aggregates, respectively. On average, the highest value for activity concentration of 232 Th and massic 220 Rn exhalation rate were showed by zircon, 340 Bq kg -1 and 169 Bq kg -1 h -1 , respectively. The emanation factor of 220 Rn/ 232 Th range from 0.3 % to 29 % for ceramic and wood, respectively. The correlation between the radon mass exhalation rate and the 226 Ra contents as well as the correlation between the thoron mass exhalation rate and 232 Th contents are in good agreement.</ns0:p><ns0:p>The mean values of 222 Rn surface exhalation rates varied from 9.2 to 3206 mBq m -2 h -1 for ceramic and zircon, respectively. The 222 Rn contribution of building materials to indoor 222 Rn considering the model room mentioned above, range from 0.04 for ceramic samples to 13 Bq m -3 for zircon. So, the annual effective dose ranged from 0.9 &#181;Sv y -1 for ceramic to 323 &#181;Sv y -1 for zircon.</ns0:p><ns0:p>In the case of 220 Rn, the surface exhalation rate average varied from 22 to 1264 Bq m -2 h -1 for cement and zircon respectively. Its contribution of building materials to indoor 220 Rn at 40 cm of the wall, range from 2.0 for cement samples to 112 Bq m -3 for zircon. Mean values of the annual effective dose ranged from 16 &#181;Sv y -1 for gypsum to 1300 &#181;Sv y -1 for zircon. Therefore, in the case of zircon and some granites, the annual effective dose was higher than the annual exposure limit for the general public of 1 mSv y -1 , recommended by the ICRP. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Rn surface exhalation rate, E A , activity concentrationi in the air of the room, A Rn222 , and annual effective dose, D Rn222 , for the different building materials.</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Rn surface exhalation rate, E A , activity concentration in the air of the room at 40 cm from the wall, A Rn220 , and annual effective dose, D Rn220 , for the different building materials.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>The acquisition time was set to 15 hours and the photopeaks used for the activity determination were: 295.2 keV (Pb-214), 351.9 keV (Pb-214) and 609.3 keV (Bi-214) for 226 Ra; 238.6 keV (Pb-212), 583.2 keV (Tl-208) and 911.2 keV (Ac-228) for 228 Ra and 1460.8 keV for K-40. Figure 2 presents as an example a gamma-ray spectrum for a granite sample. The overall quality control of the technique is guaranteed by the accreditation of the laboratory according to the ISO/IEC 17025:2005 standards and through the participation in intercomparison exercises organized by international organizations</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>numeric calculation are made by adjusting by least squares of the C vs t experimental data to the mathematical function given by equation (3). The &#945; values obtained range approximately from 0.009 to 0.04 h-1. For each material, such &#945; values were considered for the calculation of the 222Rn and 220Rn exhalation.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 1 Figure 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Gamma-ray spectrum of a granite sample.</ns0:figDesc><ns0:graphic coords='20,42.52,178.87,525.00,307.50' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Schematic experimental set-up for the radon/thoron concentration measurements.</ns0:figDesc><ns0:graphic coords='21,42.52,199.12,525.00,312.00' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure</ns0:head><ns0:label /><ns0:figDesc>Figure 4. Linear correlation analysis between 226 Ra content and (a) 222 Rn mass exhalation</ns0:figDesc><ns0:graphic coords='22,42.52,212.62,525.00,266.25' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure</ns0:head><ns0:label /><ns0:figDesc>Figure 5. Linear correlation analysis between 223 Th content and (a) 220 Rn mass exhalation</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Activity concentration for 226 Ra, C Ra , massic exhalation, E Rn222 , and emanation</ns0:figDesc><ns0:table><ns0:row><ns0:cell>factor, &#949; Rn222 , for</ns0:cell><ns0:cell>222</ns0:cell><ns0:cell>Rn of different building materials.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)Manuscript to be reviewed C</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Ra (Bq kg -1 ) E Rn222 (mBq kg -1 h -1 ) &#949; Rn222 (%) Building materials No. of samples (E Rn222 &gt; DL) Mean SD Range Mean SD Range Mean SD Range</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>Concrete</ns0:cell><ns0:cell>9 (7)</ns0:cell><ns0:cell>27.0</ns0:cell><ns0:cell>31.8</ns0:cell><ns0:cell>7.6 -87.3</ns0:cell><ns0:cell>12.2</ns0:cell><ns0:cell>8.7</ns0:cell><ns0:cell>4.3 -29.0</ns0:cell><ns0:cell>8.9</ns0:cell><ns0:cell cols='2'>6.7 1.5 -17.6</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Cement</ns0:cell><ns0:cell>5 (1)</ns0:cell><ns0:cell>28.2</ns0:cell><ns0:cell>25.1</ns0:cell><ns0:cell>21.5 -76.6</ns0:cell><ns0:cell>21.0</ns0:cell><ns0:cell>3.9</ns0:cell><ns0:cell>18.4 -23.8</ns0:cell><ns0:cell>11.2</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Marble</ns0:cell><ns0:cell>2 (1)</ns0:cell><ns0:cell>22.8</ns0:cell><ns0:cell>25.3</ns0:cell><ns0:cell>4.9 -40.7</ns0:cell><ns0:cell>26.3</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>8.6</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Slate</ns0:cell><ns0:cell>2 (2)</ns0:cell><ns0:cell>28.7</ns0:cell><ns0:cell>0.2</ns0:cell><ns0:cell>28.6 -28.9</ns0:cell><ns0:cell>16.0</ns0:cell><ns0:cell>97.4</ns0:cell><ns0:cell>10.4 -21.6</ns0:cell><ns0:cell>7.4</ns0:cell><ns0:cell>3.6</ns0:cell><ns0:cell>4.9 -9.9</ns0:cell></ns0:row><ns0:row><ns0:cell>NM</ns0:cell><ns0:cell>Granite</ns0:cell><ns0:cell>9 (9)</ns0:cell><ns0:cell>122.2</ns0:cell><ns0:cell>52.9</ns0:cell><ns0:cell>51.0 -239.1</ns0:cell><ns0:cell>70.3</ns0:cell><ns0:cell cols='2'>71.4 20.5 -221.4</ns0:cell><ns0:cell>8.5</ns0:cell><ns0:cell cols='2'>8.7 2.0 -24.9</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Ceramic</ns0:cell><ns0:cell>7 (1)</ns0:cell><ns0:cell cols='2'>126.4 105.8</ns0:cell><ns0:cell>49.9 -335.0</ns0:cell><ns0:cell>0.7</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.2</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Wood</ns0:cell><ns0:cell>1 (0)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Aggregate</ns0:cell><ns0:cell>2 (1)</ns0:cell><ns0:cell>69.9</ns0:cell><ns0:cell>39.7</ns0:cell><ns0:cell>41.8 -97.9</ns0:cell><ns0:cell>162.5</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>22.0</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Zircon</ns0:cell><ns0:cell>2 (2)</ns0:cell><ns0:cell>2070</ns0:cell><ns0:cell>14.4</ns0:cell><ns0:cell cols='4'>48.7 -4090.0 429.5 16.4 36.0 -823.0</ns0:cell><ns0:cell>6.2</ns0:cell><ns0:cell>5.0</ns0:cell><ns0:cell>2.7 -9.8</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>PM Gypsum</ns0:cell><ns0:cell>2 (1)</ns0:cell><ns0:cell>4.4</ns0:cell><ns0:cell>3.1</ns0:cell><ns0:cell>2.2 -6.6</ns0:cell><ns0:cell>1.4</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>142.6</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Activity concentration for Th , massic exhalation, E Rn220 , and emanation</ns0:figDesc><ns0:table><ns0:row><ns0:cell>factor, &#949; Rn220 , for</ns0:cell><ns0:cell>220</ns0:cell><ns0:cell>Rn of different building materials.</ns0:cell></ns0:row></ns0:table><ns0:note>232 Th, C PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>C Th (Bq kg -1 ) E Rn220 (Bq kg -1 h -1 ) &#949; Rn220 (%) Building materials No. of samples Mean SD Range Mean SD Range Mean SD Range</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>Concrete</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>9.8</ns0:cell><ns0:cell>3.9 -35</ns0:cell><ns0:cell>6.3</ns0:cell><ns0:cell>2.4</ns0:cell><ns0:cell>1.9 -10</ns0:cell><ns0:cell>1.2</ns0:cell><ns0:cell cols='2'>0.6 0.6 -2.1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Cement</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>9.2</ns0:cell><ns0:cell>5.5</ns0:cell><ns0:cell>1.1 -14</ns0:cell><ns0:cell>3.4</ns0:cell><ns0:cell>1.3</ns0:cell><ns0:cell>1.7 -5.4</ns0:cell><ns0:cell>1.6</ns0:cell><ns0:cell cols='2'>0.6 0.4 -5.9</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Marble</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2.9</ns0:cell><ns0:cell cols='2'>1.4 1.8 -3.9</ns0:cell><ns0:cell>3.5</ns0:cell><ns0:cell>0.3</ns0:cell><ns0:cell>3.3 -3.8</ns0:cell><ns0:cell>3.1</ns0:cell><ns0:cell cols='2'>1.3 2.2 -4.0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Slate</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>73</ns0:cell><ns0:cell>2.9</ns0:cell><ns0:cell>71 -75</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>2.7</ns0:cell><ns0:cell>20 -21</ns0:cell><ns0:cell>0.6</ns0:cell><ns0:cell cols='2'>0.1 0.6 -0.7</ns0:cell></ns0:row><ns0:row><ns0:cell>NM</ns0:cell><ns0:cell>Granite</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>51</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>10 -124</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>46</ns0:cell><ns0:cell>2.6 -144</ns0:cell><ns0:cell>1.1</ns0:cell><ns0:cell cols='2'>1.4 0.2 -4.8</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Ceramic</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>43</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>3.1 -80</ns0:cell><ns0:cell>2.2</ns0:cell><ns0:cell>1.6</ns0:cell><ns0:cell>1.5 -5.8</ns0:cell><ns0:cell>0.3</ns0:cell><ns0:cell cols='2'>0.4 0.0 -1.1</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Wood</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>0.6</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>78</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Aggregate</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>47</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>41 -54</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell>3.6</ns0:cell><ns0:cell>7.8 -13</ns0:cell><ns0:cell>2.4</ns0:cell><ns0:cell cols='2'>2.6 0.5 -4.2</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Zircon</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>340</ns0:cell><ns0:cell cols='2'>21 1.6 -676</ns0:cell><ns0:cell>169</ns0:cell><ns0:cell cols='2'>228 6.9 -330</ns0:cell><ns0:cell>5.4</ns0:cell><ns0:cell cols='2'>6.0 1.1 -9.6</ns0:cell></ns0:row><ns0:row><ns0:cell>PM</ns0:cell><ns0:cell>Gypsum</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1.4</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>2.7</ns0:cell><ns0:cell>0.3</ns0:cell><ns0:cell>2.5 -2.9</ns0:cell><ns0:cell>4.0</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row></ns0:table><ns0:note>1PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 3 : 222</ns0:head><ns0:label>3222</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_8'><ns0:head>1 E A (mBq m -2 h -1 ) (Bq m -3 ) &#119912; &#119929;&#119951;&#120784;&#120784;&#120784; D Rn222 (&#181;Sv y -1 ) Building materials No. of samples</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Mean</ns0:cell><ns0:cell>SD</ns0:cell><ns0:cell>Range</ns0:cell><ns0:cell>Mean</ns0:cell><ns0:cell>SD</ns0:cell><ns0:cell>Range</ns0:cell><ns0:cell>Mean</ns0:cell><ns0:cell>SD</ns0:cell><ns0:cell>Range</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Concrete</ns0:cell><ns0:cell>9 (7)</ns0:cell><ns0:cell>85</ns0:cell><ns0:cell>47</ns0:cell><ns0:cell>43 -169</ns0:cell><ns0:cell>0.34</ns0:cell><ns0:cell>0.19</ns0:cell><ns0:cell>0.17 -0.67</ns0:cell><ns0:cell>8.6</ns0:cell><ns0:cell>4.7</ns0:cell><ns0:cell>4.3 -17</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Cement</ns0:cell><ns0:cell>5 (1)</ns0:cell><ns0:cell>189</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.75</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>19</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Marble</ns0:cell><ns0:cell>2 (1)</ns0:cell><ns0:cell>212</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.85</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell /><ns0:cell>21</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Slate</ns0:cell><ns0:cell>2 (2)</ns0:cell><ns0:cell>162</ns0:cell><ns0:cell>48</ns0:cell><ns0:cell>127 -196</ns0:cell><ns0:cell>0.65</ns0:cell><ns0:cell>0.19</ns0:cell><ns0:cell>0.51 -0.78</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell>4.9</ns0:cell><ns0:cell>12.9 -20</ns0:cell></ns0:row><ns0:row><ns0:cell>NM</ns0:cell><ns0:cell>Granite</ns0:cell><ns0:cell>9 (9)</ns0:cell><ns0:cell>802</ns0:cell><ns0:cell>905</ns0:cell><ns0:cell>224 -2843</ns0:cell><ns0:cell>3.2</ns0:cell><ns0:cell>3.6</ns0:cell><ns0:cell>0.9 -11</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>91</ns0:cell><ns0:cell>23 -287</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Ceramic</ns0:cell><ns0:cell>7 (1)</ns0:cell><ns0:cell>9.2</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.04</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.9</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Wood</ns0:cell><ns0:cell>1 (0)</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Aggregate</ns0:cell><ns0:cell>2 (1)</ns0:cell><ns0:cell>1985</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>7.9</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>200</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Zircon</ns0:cell><ns0:cell>2 (2)</ns0:cell><ns0:cell>3206</ns0:cell><ns0:cell>75</ns0:cell><ns0:cell>219 -6193</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>0.9 -25</ns0:cell><ns0:cell>323</ns0:cell><ns0:cell>426</ns0:cell><ns0:cell>22 -624</ns0:cell></ns0:row><ns0:row><ns0:cell>PM</ns0:cell><ns0:cell>Gypsum</ns0:cell><ns0:cell>2 (1)</ns0:cell><ns0:cell>146</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.58</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>15</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>-</ns0:cell></ns0:row></ns0:table><ns0:note>2PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 4 (on next page) 220</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50636:1:1:NEW 8 Oct 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"ANALYTICAL CHEMISTRY DEPARTMENT Science Faculty Avda. de Elvas, s/n 06071 Badajoz, Spain Tfno.: +34 927-257152 [email protected] Badajoz, October 1, 2020 Dear Editor: We thank the reviewers for their generous comments on the manuscript entitled “Radon and thoron exhalation rate, emanation factor and radioactivity risks of building materials of the Iberian Peninsula”. We have edited the manuscript to address their concerns. We believe that the manuscript is now suitable for publication in PeerJ. Samuel Frutos Puerto Department of Analytical Chemistry Universidad de Extremadura Reviewer 1 (Anonymous) Basic reporting The manuscript has been presented in a Good manner by quoting some of the references of importance. The structure of the manuscript is also acceptable as per the quality of the work. Experimental design The authors should specifically mention the energy of the Gamma peaks used for the assessment of Radium and Thorium in the soil samples. I would suggest that the Radium equivalent activity must be mentioned instead of Radium and Thorium and the doses should be calculated by taking into consideration the Radium equivalent. We are agree with the Reviewer regarding the interest in the analysis of all these magnitudes from the radiological point of view to determinates the risks caused by the external radiation. For this reason, this issue was studied in detail in our previous work entitled “Radiations exposure from natural radionuclides in building materials”. M. J. Madruga, C. Miró, M. Reis and L. Silva. Rad. Prot. Dos. (2019), Vol. 185. N. 1, pp. 57-65. However, in this work we wanted to focus on the study of the dose due to the radon and thoron inhalation. In any case, in lines 140-143 of the tracked changes manuscript we include briefly the gamma peaks used for the assessment of Radium and Thorium. Validity of the findings I would like to see the graph of Gamma Spectrometry using HPGe and if possible, the FWHM values as it is very important that what software is used for the curve fitting and finding FWHM values for finding the areas under the peak. In lines 144-162 of manuscript (tracked changes), we have added more details regarding to the calibration of the gamma spectrometer, software, etc., including a new figure that show a gamma-ray spectrum of a granite sample. Comments for the Author Figure 1 is very vague, it must be drawn wrt (with respect to) dome coordinates. We have modified the figure 1 to show the dome coordinates. Reviewer 2 (Şeref Turhan) Basic reporting The article is well written. The language of the manuscript is fluent and understandable. But there are some writting errors. These errors were corrected on the manuscript (attached as revised manuscript pdf file). The introduction section reflects the general text of the article and adequately explains why the study was done. Figures and tables are relevant and well labeled and described. However, the article was not prepared according to the rules of the journal. The authors need to rearrange the article according to the instructions for authors. According to the Reviewer suggestions we have rearrange the article according to the instructions for authors. Experimental design Research question was well defined, relevant and meaningful. Methods were described with sufficient detail. The experimental methods used in this study are very well known. Validity of the findings In this study, (1) the activity concentrations of 226Ra and 232Th naturally occurring in 41 building materials (structural, covering and raw materials) manufactured in the Iberian Peninsula (Portugal and Spain), exported and used in all countries of the world were measured using a gamma-ray spectrometry with an HPGe detector, (2) radon (222Rn) and thoron (220Rn) exhalation rates (mass and surface) and emanation coefficients of these materials were determined using active radon and thoron monitor, and (3) the annual effective dose due to inhalation of radon and thoron was estimated to assess the radiological risks associated the use of these building materials. These findings are important for assessing the radon and thoron concentrations in buildings, environmental radioactivity, and health hazards of individuals. For this reason, the manuscript can be considered for publication in International PeerJ. Comments for the Author But there are some writting errors. These errors were corrected on the manuscript (attached as revised manuscript pdf file). The authors thank the reviewer for their suggestions and corrections, which have been taken into account in the new version of the article. Reviewer 3 (Dragoslav Nikezic) Basic reporting Specific comments Page in MS (Line) Written Change to Comments 1. L 48-49 Radon is the second leading cause of increase of the probability of lung cancer after tobacco smoke (Torres-Durán et al., 2014). This citation is not correct. This fact was established much before 2014. Find original reference 2. L.55 is 3.8 3.825 3. L.55 equivalent radiation dose Usually we talk about effective dose. 4. L62 to the detriment neglecting Sentence is too long and cumbersome 5. L70 progenies progeny There is not plural for progeny. 6. L99 materials was materials were 7. Eq2 Back diffusion was not taken into account in this Eq. Please explain why it was neglected 8. L. 167 which do not consider which does not consider 9. L. 219 from de wall from the wall Page in MS tracked changes (Line) Response 1. L.50 Citation changed. 2. L.56 Modified 3. L.57 Modified 4. L.61-62 We have shortened the sentence. 5. L.70 Ok, modified. 6. L.101 Changed 7. L.184-188 In accordance with the suggestion of the reviewer we have added an explanation. 8. L.200 Changed 9. L.252 Changed Experimental design Well done Validity of the findings Agree with world wide Comments for the Author Major revision is need for this ms. Experimental part of this ms is probably well done. Calibration of gamma spectrometer is missing. However, estimation of radon and thoron activity from some construction material obtained by Eq. 6 is not correct. It does not consider amount of material used in some room. So, it was obtained that effective dose (which is usually denoted with E) from zircon is larger than 1 mSv/year. My question, how much zircon can be found in some room, and how is realistic that all room is made by zircon. Formula in Eq.6 should be modified to take into account amount of material used in construction. Then, repeat calculation of effective dose In my, the best knowledge, zircon is used as coating material as very thin layer on ceramic tiles in bathroom or kitchen. Total mass of zircon is very small. It can be true or similar for some other material. In lines 144-162 of manuscript (tracked changes), we have added more details regarding to the calibration of the gamma spectrometer, including a new figure that show a gamma-ray spectrum of a granite sample. Eq. 6 follows the study by Amin (Amin, 2015) which is fully accepted by the scientific community. The mass quantity of building material does not appear explicitly in the equation but is considered as it would influence the value of the surface exhalation rate, which does appear in the equation. In Eq. 6 also appears the influence of the area and volume of the room based on the model proposed by the UNSCEAR 2016. Values for zirconium are indeed very high, but because the whole room built with this material would be considered. Obviously, we know that this material appears mixed with others. However, we have wanted to keep said models for rooms with unique materials. Logically, this model can be adapted later for mixing materials but it will depend on the final composition of each material whose casuistry is very large. It depends on the area where it is built, the type of house, etc. Reviewer 4 (Anonymous) Basic reporting The paper is interesting and contains data worth to be published. However the English should be substantially improved. I’m not a native English speaker and I don’t feel qualified to do a complete linguistic revision of the text. I just corrected some major mistakes (see the attached annotated pdf). However, in my opinion, the following sentences should be rewritten to improve clarity and comprehensibility: 1. Lines 60-62 2. Lines 67-69 3. Lines 151-153 4. Lines 164-168 5. Lines 184-186 (attending ??) 6. Lines 222-224 7. Line 236 (the maximum value on average: what do you mean?) Besides the language issue, the paper contains some problematic points, addressed more specifically in the next sections. These points need to be clarified before making the decision to publish or not The authors have accepted the suggestions of the reviewer. Experimental design In equation (1) (Material and methods), the mass m of the sample is missing in the denominator: please correct I have also some questions regarding the experimental set up described in figure 2. a) Do you have measured (or estimated) the leakage factor  in equation (2)? Its value is negligible for thoron but can substantially affect the radon exhalation estimation. b) Can you give the asymptotic activity concentration values (for both radon and thoron) reached in the cylindric (not cilindrical, please correct) sealed container? Eq. 1. Indeed, it was a mistake that have been corrected in the new version. a) The α numeric calculation are made by adjusting by least-squares of the C vs t experimental data to the mathematical function given by equation (3). We have included a new paragraph to explain this in lines 188-192 of the manuscript (tracked changes). b) We have not given these values because we thought it was not the most interesting in this work and we did not want this work to be too long. If the reviewer insists we would have no problem adding it. Validity of the findings My major concern regarding the findings is related to the use of equation (7) for the calculation of the surface exhalation rate from the measured mass exhalation rate. The relationship between these quantities can be highly affected by the diffusion length (l=√(D⁄lambda) ;lamba=decay constant; D=coefficiente of diffusion in the material) of radon and thoron in the different material. Moreover, the exhalation area of your experimental set up can be hardly compared to the exhalation area in real conditions. Actually, being the material crushed, you have increased the effective surface emission area compared to that of the real building material (bric, tile). This fact can affect your results. Some other minor questions should be addressed. In the tables you take mean values for some materials (zircon, for example) with only 2 samples which show a very large difference: it seems not very informative. Indeed, as the reviewer indicates, we are starting from crushed material to finally estimate the exhalation area in real conditions. However, we think that it can be a first approximation to determine the exhalation rate. We propose in future works to make the same estimation starting from non-crushed material. We agree with the reviewer's comment that indeed having little zircon material with its statistical result will be less significant than the rest of the materials. Comments for the Author The paper is interesting and worth publishing. I encourage the authors to answer to the above addressed points "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background Tuberculosis (TB) among migrants from high-risk countries and underling interventions were concerned for disease control. This study aimed to assess the TB trends among marriage-migrants with the 1-2-round vs. labor-migrants with the 4-round TB screenings in the period of the first four post-entry years; pre-entry screenings by an initial chest X-ray (CXR) were conducted during 2012-2015, and a friendly treatment policy was introduced in 2014. Methods TB data of migrants during 2012-2015 were obtained from the National TB Registry Database and analyzed. The incidences, clinical characteristics, and treatment outcomes were assessed to explore the impact of underlying interventions.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>During post-entry 0-4 years, the TB incidence rates among marriage-migrants ranged 11-90 per 100,000 person-years, with 60.8% bacteria-positive and 28.2% smearpositive cases. Whereas among labor migrants, the incidence rates ranged 67-120 per 100,000 person-years, with 43.6% bacteria-positive and 13.7% smear-positive cases. All migrants originated from Southeast Asia following pre-entry health screening in 2012-2015. The TB cases among marriage-migrants were with a higher proportion of sputum-smear-positivity (SS+) (OR: 4.82, 95% CI: 3.7-6.34) and CXR cavitation (OR: 2.90, 95% CI: 2.10-4.01). Marriage-migrants with TB had treatment completion rate of &gt; 90%, which was above the WHO target. For labor-migrants with TB, when compared the period of post-vs. pre-implementation of the friendly therapy policy that eliminated compulsory repatriation, the overall treatment completion rate of those who stayed in Taiwan improved by 30.9% (95% CI: 24.3-37.6) versus 6.7% (95% CI: 3.8-9.7), which exceeded a 4.88-fold (95% CI: 3.83-6.22) improvement. Additionally, the treatment initiation rate within 30 days of diagnosis for SS-TB and B-TB cases during post-vs. pre-implementation of the therapy policy was increased, i.e., 77.1% vs. 70.9% (OR: 1.38, 95% CI: 1.12-1.70) and 78% vs. 77% (OR:1.64, 95% CI 1.38-1.95). Conclusion Multiple CXR screenings could identify more TB cases with sputum-smear-negativity (SS-) TB at the early-stage,</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Taiwan, with a population of approximately 23.4 million, is a country with moderate TB incidence ranged 46-53 per 100,000 population in 2012-2015 <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref><ns0:ref type='bibr' target='#b1'>[2]</ns0:ref> and has gradually decreased over time. Nevertheless, TB disproportionately affects the foreign-born population from TB high-risk countries, with an incidence in this population that is several times higher than that among the Taiwanese-born population <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref><ns0:ref type='bibr' target='#b1'>[2]</ns0:ref><ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. The reasons for the TB burden in the migrant population are likely to be the reactivation of remotely acquired LTBI following migration from high TB burden countries to lower TB burden countries <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref><ns0:ref type='bibr' target='#b1'>[2]</ns0:ref><ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. TB control measures for the foreign-born population from high endemic countries in Taiwan occur within the scopes of the TB medical surveillance program. All newly arriving residents and temporary residents with high-risk undergo an immigration medical examination before arrival. This examination consists of a physical examination which includes chest X-ray (CXR), syphilis test, HIV serology and other routine tests. If there is a radiographic evidence of TB, three sputum smears and mycobacterial cultures are examined <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. Following the pre-entry screening, mandatory TB screening of either four rounds during the first three years <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref> or 1-2 rounds during the first 4 years <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref> are required post-entry for labor migrants or marriage migrants, respectively, who come from Southeast Asian (including Indonesia, Vietnam, Philippines, and Thailand) or China. Applicants with active TB are required to complete treatment before entering Taiwan. Before 2013, migrant workers with TB were repatriated; since 2014, these individuals are allowed to stay in Taiwan for treatment except multiple drug-resistant TB (MDR-TB) patients. Study showed that TB case with smear-positive sputum was more infectious than those with smearnegative sputum <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>. Our previous studies revealed that a 30.2% <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref> and 14.3% <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref> smearpositive sputum among marriage migrants vs. labor migrants in Taiwan. In term of TB control targeting high TB incidence migrants, it should take consideration including screening the differences in pathogen exposure due to their origins, the ethnic socio-economic disparities and the experience of migration itself <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref>; moreover, to overcome the cultural and structural barriers to accessing healthcare <ns0:ref type='bibr' target='#b6'>[7]</ns0:ref> is also critical. Thus, a friendly TB therapy policy may prompt access to treatment for a migrant in the new residing country. This study retrospectively analyzed a cohort from 2012-2015 aimed to analyze the impact on TB incidence and clinical characteristics following the enhanced screening interventions among these high-risk migrants. Also, the treatment outcomes following the concurrent TB intervention programs as well as the effectiveness of introducing a friendly TB therapy policy in 2014 targeting labor migrants was assessed.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIALS &amp; METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>Data Collection</ns0:head><ns0:p>TB data on migrants were retrieved from the following sources similar to previous studies <ns0:ref type='bibr' target='#b3'>[4,</ns0:ref><ns0:ref type='bibr' target='#b4'>5]</ns0:ref> :</ns0:p><ns0:p>(1) the national TB registry database, which was updated regularly by clinicians and local health divisions; (2) The populations of marriage migrants and labor migrants were obtained from the official publications of the Ministry of the Interior <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Definition of migrants with TB</ns0:head><ns0:p>Two strategies of post-entry TB screenings were targeted towards foreign migrants from highly endemic Southeast Asian countries (including Vietnam, Philippines, Thailand, and Malaysia) during the first 4 post-entry years following pre-entry screenings in Taiwan: 1) For labor migrants (foreign workers), 4 screenings at 0-3 days as well as at 6, 12, 18 and 30 months are conducted <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>. 2) For marriage migrants (foreign spouses), one to two mandatory TB screenings are conducted <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>. The definition of marriage migrants with TB <ns0:ref type='bibr' target='#b4'>[5]</ns0:ref> or labor migrants with TB <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref> was described as previously <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref><ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>. Treatment completion, according to the WHO guidelines, is defined as accomplished by a standard regimen, i.e., a 6-9-month regimen, or complete treatment by a longer regimen. The TB treatment completion rate was obtained as the number of TB patients who achieved treatment completion by the standard or longer therapy regimen divided by the number of TB patients staying in Taiwan.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistics and analysis</ns0:head><ns0:p>The TB incidence (per 100,000 population) was calculated as the number of newly detected labor migrants with TB during the study period <ns0:ref type='bibr'>(2012)</ns0:ref><ns0:ref type='bibr'>(2013)</ns0:ref><ns0:ref type='bibr'>(2014)</ns0:ref><ns0:ref type='bibr'>(2015)</ns0:ref> divided by the population per year. The yearly fluctuating trends in the TB incidence rates among labor migrants (X 1 ) and marriage migrants (X 2 ) vs. the WHO estimated TB incidence rates <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref> (Y), as well as the TB incidences (X 1 ) of the labor migrants vs. the marriage migrants (X 2 ), were detected with linear regression analysis. Furthermore, several odds ratio tests were performed: the studied TB case population was divided into two exposure categories: marriage migrants vs. labor migrants or postimplementation (2014-2015) vs. pre-implementation (2012-2013) of the friendly therapy policy. Categorical variables were subjected to binary analysis using a 2&#215;2 table to compute odds ratios (ORs) and 95% confidence intervals (CIs) by using R Programming or OpenEpi <ns0:ref type='bibr' target='#b7'>[8]</ns0:ref> for assessment of the difference between paired binary variable determinants; the control (i.e., reference) for each variable was the total number in each stratified category, excluding the detected items (i.e., independent variables). The TB treatment outcome was estimated by calculating the transferred-out rate and the treatment completion rate. To assess the impact of implementing the friendly therapy policy with no compulsory repatriation, the postimplementation (2014-2015) vs. pre-implementation (2012-2013) TB treatment completion rate, the percentage of sputum smear positivity of the transferred out cases vs. the cases staying in Taiwan for therapy, and the initiation of treatment within 30 days of diagnosis, i.e., the delay (in </ns0:p></ns0:div> <ns0:div><ns0:head>Ethics statement</ns0:head><ns0:p>This study was approved by the Institutional Review Board of the Taiwanese CDC underidentification No. TwCDCIRB106115.</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head></ns0:div> <ns0:div><ns0:head>TB incidence trends among migrants and TB screening impact</ns0:head><ns0:p>The TB surveillance strategy in Taiwan implemented a mandatory screening post-entry for immigrants at risk. This is either an 1-2 rounds or four rounds of screenings targeted to marriage migrants or labor migrants who come from high-burden countries, during the first 0-4 years or the first 0-3 years, respectively, following an initial pre-entry screening with CXR detection. For migrants from the same area of origin (Indonesia, Vietnam, Thailand, and the Philippines), the TB incidence rates of labor migrants were higher than those of marriage migrants, P &lt; 0.001 (Table <ns0:ref type='table'>1</ns0:ref>). Briefly, the TB incidence ranged 67.71-143.78 (per 100,000) among labor migrants, 11.3-99.74 among marriage migrants from Southeast Asia, and 11.4-15.66 among those from China during their first post-entry 0-4 years in 2012-2015 (Table <ns0:ref type='table'>1</ns0:ref>). Relationships among the fluctuation trends during the first 0-4 years of the post-entry period were detected. The TB incidence rates of the labor migrants (X 1 ) or of the marriage migrants (X 2 ) were positively correlated with those of WHO estimated (Y) for their countries of origin <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref> based on yearly surveillance by the Pearson's correlation or liner association in the manner of significant (n = 16, Pearson's correlation R X1:Y : 0.74, linear association R 2 X1:Y : 0.55, P &lt; 0.0001) or of less significant (n=14, Pearson's correlation R X2:Y :0.65, linear association R 2 X2:Y : 0.42, P = 0.0125). Furthermore, the TB incidence rates of the labor migrants (X 1 ) vs. those of the marriage migrants (X 2 ) identified were higher, but the association between these fluctuation trends appeared insignificant (i.e., n = 14, Pearson's correlation R:0.248; linear association: R 2 X2: X1 : 0.061, P = 0.393).</ns0:p></ns0:div> <ns0:div><ns0:head>Comparison of TB screening impact on clinical characteristics among TB cases</ns0:head><ns0:p>TB cases identified among marriage migrants were similar to those identified among labor migrants in terms of nationality, but the marriage migrants were more likely to be women (OR: 25.37, 95% CI 13.47-47.81), were more likely to be older than 45 years old (OR: 3.89, 95% CI 2.61-5.78) and had more severe infections (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>) at diagnosis. The effect of TB screenings on clinical characteristics among migrants with TB after post-entry screenings were assessed and revealed 60.8% bacteria-positive and 58.4% smear-positive cases among marriage migrants; 43.6% bacteria-positive and 42.6% smear-positive cases among labor migrants (Table <ns0:ref type='table'>1</ns0:ref>). The marriage migrants with 1-2 rounds vs. labor migrants with 4 rounds post-entry screenings had higher rates of sputum smear positivity (SS+) (OR: 4.82, 95% CI 3.7-6.34), higher rates of CXR Manuscript to be reviewed cavitation (OR: 2.90, 95% CI 2.10-4.01) and higher rates of B+ (OR: 2.1, 95% CI 1.61-2.51) (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Anti-TB treatment outcomes: post-vs. pre-implementation of the friendly therapy policy</ns0:head><ns0:p>The overall treatment completion rate was &gt;90% among both the labor and marriage migrants who stayed in Taiwan and accepted TB treatment (Table <ns0:ref type='table'>1</ns0:ref>), which was above the WHO target of &gt;85% <ns0:ref type='bibr' target='#b7'>[8]</ns0:ref>. Comparing outcome during post-(2014-2015) vs. pre-(2012-2013) implementation of the therapy policy with no compulsory repatriation for labor migrants, the treatment completion rate was increased by 4.88-fold (95% CI: 3.83-6.22), i.e., 30.9% (95% CI 24.3-37.6) vs. 6.7% (95% CI 3.8-9.7) (Table <ns0:ref type='table'>1</ns0:ref>). Higher completion rates for SS-TB cases OR: 7.45 (95% CI 5.44-10.2) or B-TB cases OR: 6.87 (95% CI 4.21-11.2); SS+ TB cases, OR:13.43 (95% CI 4.02-44.79) (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>), or B+TB cases OR: 5.83 (95% CI 3.62-9.42) were also observed.</ns0:p><ns0:p>After 2014 (post-) vs. before 2013 (pre) implementing friendly therapy being less in transfer out TB cases of OR: 0.21 (95% CI 0.16-0.26) (Table <ns0:ref type='table'>1</ns0:ref>) which was detailed by less transfer out cases with both SS-TB cases of OR: 0.61 (95% CI 0.38-0.97) or B-TB cases of OR: 0.58 (95% CI 0.49-0.69) and SS+TB cases of OR: 0.44 (95% CI 0.27-0.71) or B+ TB cases of OR: 0.66 (95% CI 0.56-0.79) (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>) as well as being no significant change in lost-to follow up TB cases of OR: 1.27 (95% CI 0.69-2.40) among labor migrants were exhibited (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>In terms of overall TB treatment initiation within 30 days of diagnosis, in both migrants with SS+ TB (96.7% (188/194) vs. 91.2% (290/318)) and SS-TB (81.8% (301/368) vs. 74.3% (1364/1834)), the marriage migrants were more compliant than labor migrants. Moreover, comparing the overall initiation rate of TB treatment among labor migrants with TB showed a significant increase of 73.7% vs. 68.2% (OR: 1.31, 95% CI 1.09-1.57) during the post-vs. preimplementation of the therapy policy. Furthermore, during the post-vs. pre-implementation of the therapy policy, the TB treatment initiation rate among labor migrants showed a significant increase of 77% vs. 71% (OR: 1.38, 95% CI 1.12-1.70) with SS-TB and a non-significant change of 89.4% vs. 93.00% (OR: 0.63, 95% CI: 0.29-1.39) with SS+ TB. Meanwhile, a significant increasing of 78% vs. 77% (OR:1.64, 95% CI 1.38-1.95) with B-TB and a nonsignificant change of 83% vs 85% (OR: 0.84, 95% CI: 0.70-1.01) with B+ TB was observed (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Multiple initial screenings with CXR could identify more SS-TB cases at an early stage with low infectivity and result in a higher incidence. For migrants who come from high TB burden regions, a pre-entry initial screening with a CXR is required and follow by mandatory 1-2 rounds of post-entry screening for marriage migrants or 4 rounds for labor migrants in Taiwan, respectively. This initiative resulted in a higher incidence with more SS-TB among labor migrants and a lower incidence but more SS+ TB or more CXR cavitation TB cases among Manuscript to be reviewed marriage-migrants (Table <ns0:ref type='table'>1</ns0:ref>, Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Moreover, the higher annual TB incidence rates in labor migrants were significantly (R 2 : 0.55, P &lt;0.0001) associated with the WHO estimated TB incidence rates <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref> of their original countries and there were significantly fewer severe cases, i.e., fewer SS+ TB cases or fewer CXR cavitation TB cases, among the labor migrants. On the other hand, relatively lower annual TB incidence rates among marriage migrants were less significantly associated with the WHO-estimated TB incidence rates <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref> of their original countries (R 2 = 0.42, P = 0.0125), and there were significantly more severe cases, i.e., more SS+ TB cases (OR: 4.82), or more CXR abnormal with cavitation TB (OR:2.90) (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Thus, these results indicated that the greater number (4 rounds) of mandatory post-entry screenings for labor migrants vs. the 1-2 rounds for marriage migrants could screen out more TB cases; although the incidence rate appeared higher, but more cases had early-stage disease with SS-TB. Also, these results corroborate previous findings that multiple TB screenings in individuals with initial abnormal CXRs result in the detection and identification of more SS-TB cases at an early disease stage <ns0:ref type='bibr' target='#b3'>[4]</ns0:ref>, which might benefit for timely therapy initiation; thus, might increase the success of treatment and reduce disease burden including blocking the potential disease dissemination. However, the high proportion of bacteriological negativity i.e., of &gt; 35% with B-TB in both groups, e.g., 56.4% in labor migrants vs. 39.2% in marriage migrants (Table <ns0:ref type='table'>1</ns0:ref>), has suggested the multiple TB screenings might be of over-detection <ns0:ref type='bibr' target='#b9'>[10,</ns0:ref><ns0:ref type='bibr' target='#b10'>11]</ns0:ref>. While the reasons for the high TB burden in the migrant population are likely to be the reactivation of remotely acquired LTBI following migration from high TB burden countries to lower TB burden countries <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref><ns0:ref type='bibr' target='#b1'>[2]</ns0:ref><ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. Therefore, it is important that applying a LTBI screening combined with prevention treatment (PT) at a pre-or post-entry in the very beginning will save a lot of efforts in later stages, e.g., multiple screenings <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref> and might meet the core of an earlier TB mitigation strategy for TB control intervention targeting at high-risk migrants even including BCG-vaccinated individuals <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>; e.g., a 2-step LTBI screening: performing an expensive interferon-gamma release assay (IGRA) after positivity on the economical tuberculin skin test (TST), then combined with PT <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref>. Thus, antibiotics can effectively and economically eliminate tuberculosis, before they become contagious, i.e., to treat individuals who are still invisibly sick. In several countries, such as the United States, Britain, and Canada, LTBI screenings combined with PT strategies have long since become public health norms for migrants <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>.</ns0:p><ns0:p>The repatriated labor migrants with TB were not traced in this study if they received any treatment when they returned home country. Nonetheless, both marriage migrants and labor migrants with TB who stayed in Taiwan and accepted either the standard (6-9 months) or longer regimen of TB therapy achieved a TB treatment completion rate of 87-99% (Table <ns0:ref type='table'>1</ns0:ref>), which was above the WHO target of &gt;85% <ns0:ref type='bibr' target='#b8'>[9]</ns0:ref> during 2012-2015. In terms of assessing the overall treatment outcome of TB cases, it was found that cases identified among labor migrants significantly had poorer treatment completion than those among marriage-migrants in Taiwan because of a higher percentage of transferred out cases in 2012-2015 (Table <ns0:ref type='table'>1</ns0:ref>). The treatment completion rate in Taiwan was high in marriage migrants (96-99%; Table <ns0:ref type='table'>1</ns0:ref>). As for labor migrants, after the implementation of the therapy policy which eliminated repatriation of TB cases and allowed for therapy in the host country, the completion rates raised to 24-37% from 3.8-9.7% (Table <ns0:ref type='table' target='#tab_3'>1 and Table 3</ns0:ref>) after the implementation of the therapy policy during 2012-2015.</ns0:p><ns0:p>Overall, the improvement of an increased 4.88-fold i.e., 30.9% in 2014-2015 versus 6.7% in 2012-2013 of the TB treatment completion among labor migrants during periods postimplementation (2014-2015). It was further observed that TB cases with more SS-(OR: 7.45) or B-44(OR:1.48) and more SS+ (OR: 13.43) or B+ (OR:5.83) was treatment completed since 2014 (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>) among labor-migrants, which demonstrates the benefit of implementing the friendly therapy policy. Also, the reducing potential structural barriers to TB treatment completion <ns0:ref type='bibr' target='#b15'>[14]</ns0:ref><ns0:ref type='bibr' target='#b16'>[15]</ns0:ref><ns0:ref type='bibr' target='#b17'>[16]</ns0:ref>, limited access to care <ns0:ref type='bibr' target='#b15'>[14]</ns0:ref>, and relocations of labor-migrants were conducted by introducing the therapy policy since 2014. Moreover, after 2014, a significant decline in transfer out TB cases was observed in both of SS-TB cases (OR: 0.61) or B-TB cases (OR: 0.58) and SS+TB cases (OR: 0.44) or B+ TB cases (OR: 0.66) among labor migrants during this study period (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>).</ns0:p><ns0:p>Additionally, because the higher number of SS-or B-TB cases among labor migrants than among marriage migrants was also worried as a potential risk of being delayed or untreated and then, in turn, developing TB dissemination. Since very few bacilli are sufficient to cause infection <ns0:ref type='bibr' target='#b18'>[17]</ns0:ref><ns0:ref type='bibr' target='#b19'>[18]</ns0:ref>, therapy should not be delayed in labor-migrants with SS-TB according to the WHO guidelines, especially those who were to provide long-term care for vulnerable people or the elderly; therefore, the initiation rate of anti-TB treatment within 30 days of diagnosis was concerned. After introducing the friendly therapy policy, the overall treatment initiation rate within 30 days of diagnosis was significantly increasing (OR: 1.31) among labor migrants. Furthermore, the overall initiation rate of treatment exhibited higher among labor migrants with SS+ TB (93.0%) or B+ TB (83.9%) than those with SS-TB (70.9%) or B-TB (77.4%) and implicated that relatively more TB cases with SS-or B-had somehow delayed treatment than those with SS+ or B+. Nevertheless, after introducing the therapy policy for labor migrants, showed a significant increase in the initiation rate of treatment both for SS-TB of 77% vs. 71% (OR: 1.38) or B-TB of 78% vs. 77% (OR:1.64) during the post vs. pre-implementation of the policy. Relatively, this implementation was not significantly impacted the treatment initiation rate among labor immigrants with SS+ TB of between 91%-89% (OR: 0.63) or B+ (OR: 0.84) (Table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>). Thus, the friendly therapy policy in Taiwan has resulted in an improved experience of reducing structural barriers to TB mitigation since its implementation in 2014, which has successfully promoted the anti-TB treatment outcomes including improvement in treatment initiation especially those who with S-TB or B-TB and increasing treatment completion for those migrants with TB stayed in Taiwan. Therefore, based on our observations, there is a need for intensifying health education that promotes TB therapy includes delivering the information of the ongoing availability of free, accessible health services for vulnerable groups such as highrisk migrants; this health education could also be a critical element in increasing treatment success <ns0:ref type='bibr' target='#b6'>[7,</ns0:ref><ns0:ref type='bibr' target='#b19'>[18]</ns0:ref><ns0:ref type='bibr' target='#b20'>[19]</ns0:ref> and early therapy for individuals with TB in receiving countries.</ns0:p></ns0:div> <ns0:div><ns0:head>Limitations</ns0:head><ns0:p>The post-entry screening frequency of 1-2 rounds for marriage migrants was not the same as that for labor migrants (4 rounds), which might cause an underestimation of TB incidence in the former. The authentic treatment completion rate may have been underestimated among labor migrants who opted to return to their original countries for treatment, as they were not or enrolled in or followed by this study. The proportions of migrants with TB who were due to TB reactivation or transmission were not defined by the molecular testing in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS</ns0:head><ns0:p>Multiple screenings following an initial abnormal CXR in migrants could detect early-stage TB cases; nevertheless, an improved therapy completion is a substantial step for TB elimination. The relatively higher odds of SS+ TB and bacterial negativity &gt;35% among migrants might be an index of persistent TB reactivation or over-diagnosis; therefore, it is recommended that adding LTBI screening combined with preventive treatment as an alternative approach might save multiple screening effort for high-risk migrants. The friendly treatment policy for migrants, which eliminated repatriation for labor migrants with TB, could benefit in anti-TB treatment outcomes include increasing therapy initiation in SS-or B-TB cases and treatment completion for those who stayed in receiving countries. Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020) days) of treatment initiation (standard or longer regimen) following diagnosis among labor migrants during post vs. pre-implementation of the therapy policy, were estimated by t-tests (2sides) or Chi-squared tests.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>a. B+: bacterial positivity, with positivity among 3 sputum smears or sputum cultures b. B-: bacterial negativity, with no positivity among 3 sputum smears or sputum cultures c. SS+: positivity among 3 sputum smears d. Staying in Taiwan cases = all TB cases -transferred-out cases, including compulsorily repatriated cases e. Treatment completion using 6-9-month regimens, i.e. a standard regimen or longer regimen f. The treatment completion rate of cases staying in Taiwan = cases staying in Taiwan with treatment completion by a standard or longer therapy regimen / cases staying in Taiwan g. Pearson's Chi-squared test h. Odds ratio of TB cases during 2014-2015 vs. 2012-2013 PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>1 Table 1 .</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Tuberculosis notification and treatment outcomes among labor and marriage migrants</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell cols='6'>Manuscript to be reviewed</ns0:cell></ns0:row><ns0:row><ns0:cell>treatment</ns0:cell><ns0:cell /><ns0:cell>Died</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell>0</ns0:cell><ns0:cell /><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Transferred out</ns0:cell><ns0:cell>426</ns0:cell><ns0:cell>526</ns0:cell><ns0:cell cols='2'>Labor migrants 457 380</ns0:cell><ns0:cell /><ns0:cell>9</ns0:cell><ns0:cell /><ns0:cell>2</ns0:cell><ns0:cell cols='2'>Marriage migrants 3 2</ns0:cell><ns0:cell>Chi-sq g</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Year</ns0:cell><ns0:cell cols='2'>2012 Transferred-out rate 0.903</ns0:cell><ns0:cell cols='2'>2013 0.905 0.707 2014</ns0:cell><ns0:cell cols='5'>2015 0.609 0.21(0.16-0.26) h 0.083 0.022 Sum 2012 2013</ns0:cell><ns0:cell cols='2'>2014 0.032 0.024 2015</ns0:cell><ns0:cell>Sum p-value</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>Notification TB cases Lost to follow up</ns0:cell><ns0:cell>472 8</ns0:cell><ns0:cell>581 9</ns0:cell><ns0:cell>646 11</ns0:cell><ns0:cell cols='2'>624 15 1.27(0.7-2.4) h 2323</ns0:cell><ns0:cell>109 0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>92</ns0:cell><ns0:cell>95 1</ns0:cell><ns0:cell>84 1</ns0:cell><ns0:cell>380</ns0:cell></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell cols='2'>&lt;=24 Staying in Taiwan (ST) d</ns0:cell><ns0:cell>101 46</ns0:cell><ns0:cell>102 55</ns0:cell><ns0:cell>129 189</ns0:cell><ns0:cell>120 244</ns0:cell><ns0:cell>452</ns0:cell><ns0:cell>6 100</ns0:cell><ns0:cell /><ns0:cell>9 90</ns0:cell><ns0:cell>4 92</ns0:cell><ns0:cell>4 82</ns0:cell><ns0:cell>23 &lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell cols='2'>25-44 Treat. completion (TC) e</ns0:cell><ns0:cell>363 46</ns0:cell><ns0:cell>464 48</ns0:cell><ns0:cell>489 178</ns0:cell><ns0:cell cols='2'>483 233 4.88(2.8-6.2) h 1799</ns0:cell><ns0:cell>91 99</ns0:cell><ns0:cell /><ns0:cell>72 89</ns0:cell><ns0:cell>80 90</ns0:cell><ns0:cell>72 80</ns0:cell><ns0:cell>315</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>&lt;=45 TC = 6-9 months e</ns0:cell><ns0:cell>8 42</ns0:cell><ns0:cell>15 48</ns0:cell><ns0:cell>28 174</ns0:cell><ns0:cell>21 232</ns0:cell><ns0:cell>72</ns0:cell><ns0:cell>12 99</ns0:cell><ns0:cell /><ns0:cell>11 89</ns0:cell><ns0:cell>11 88</ns0:cell><ns0:cell>8 79</ns0:cell><ns0:cell>42</ns0:cell></ns0:row><ns0:row><ns0:cell>Sex</ns0:cell><ns0:cell /><ns0:cell>Female TC &gt; 6-9 months e</ns0:cell><ns0:cell>295 0</ns0:cell><ns0:cell>352 0</ns0:cell><ns0:cell>361 4</ns0:cell><ns0:cell>370 1</ns0:cell><ns0:cell>1378</ns0:cell><ns0:cell>107 0</ns0:cell><ns0:cell /><ns0:cell>92 0</ns0:cell><ns0:cell>89 2</ns0:cell><ns0:cell>82 1</ns0:cell><ns0:cell>370 &lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Male STT-completion rate f 0.913 177</ns0:cell><ns0:cell>229 0.873</ns0:cell><ns0:cell>285 0.92</ns0:cell><ns0:cell>254 0.95</ns0:cell><ns0:cell>945</ns0:cell><ns0:cell cols='3'>2 0.96 0.989 0</ns0:cell><ns0:cell>6 0.96</ns0:cell><ns0:cell>2 0.96</ns0:cell><ns0:cell>10</ns0:cell></ns0:row><ns0:row><ns0:cell>Country</ns0:cell><ns0:cell /><ns0:cell>Indonesia</ns0:cell><ns0:cell>221</ns0:cell><ns0:cell>286</ns0:cell><ns0:cell>322</ns0:cell><ns0:cell>296</ns0:cell><ns0:cell>1125</ns0:cell><ns0:cell>13</ns0:cell><ns0:cell /><ns0:cell>6</ns0:cell><ns0:cell>7</ns0:cell><ns0:cell>9</ns0:cell><ns0:cell>35 &lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Vietnam</ns0:cell><ns0:cell>77</ns0:cell><ns0:cell>98</ns0:cell><ns0:cell>102</ns0:cell><ns0:cell>123</ns0:cell><ns0:cell>400</ns0:cell><ns0:cell>44</ns0:cell><ns0:cell /><ns0:cell>44</ns0:cell><ns0:cell>34</ns0:cell><ns0:cell>30</ns0:cell><ns0:cell>152</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Philippines</ns0:cell><ns0:cell>109</ns0:cell><ns0:cell>128</ns0:cell><ns0:cell>152</ns0:cell><ns0:cell>162</ns0:cell><ns0:cell>551</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell>4</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>16</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Thailand</ns0:cell><ns0:cell>65</ns0:cell><ns0:cell>69</ns0:cell><ns0:cell>68</ns0:cell><ns0:cell>43</ns0:cell><ns0:cell>245</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /><ns0:cell>2</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>6</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>China</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0</ns0:cell><ns0:cell>48</ns0:cell><ns0:cell /><ns0:cell>36</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>44</ns0:cell><ns0:cell>170</ns0:cell></ns0:row><ns0:row><ns0:cell>Population</ns0:cell><ns0:cell /><ns0:cell cols='5'>Indonesia 191127 213234 229491 236526</ns0:cell><ns0:cell cols='4'>870378 27684 27943</ns0:cell><ns0:cell cols='2'>28287 28699</ns0:cell><ns0:cell>112613 &lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='5'>Vietnam 100050 125162 150632 169981</ns0:cell><ns0:cell cols='4'>545825 87357 89042</ns0:cell><ns0:cell cols='2'>91004 93441</ns0:cell><ns0:cell>360844</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Philippines 86786</ns0:cell><ns0:cell cols='3'>89024 111533 123058</ns0:cell><ns0:cell>410401</ns0:cell><ns0:cell>7465</ns0:cell><ns0:cell cols='2'>7707</ns0:cell><ns0:cell>8021</ns0:cell><ns0:cell>8326</ns0:cell><ns0:cell>31519</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Thailand 67611</ns0:cell><ns0:cell cols='2'>61709 59933</ns0:cell><ns0:cell>58372</ns0:cell><ns0:cell>247625</ns0:cell><ns0:cell>8336</ns0:cell><ns0:cell cols='2'>8375</ns0:cell><ns0:cell>8467</ns0:cell><ns0:cell>8525</ns0:cell><ns0:cell>33703</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>306,51</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>China</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0</ns0:cell><ns0:cell /><ns0:cell cols='4'>315905 323,358 330069 1275846</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>4</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='5'>Indonesia 115.63 134.13 140.31 125.14</ns0:cell><ns0:cell /><ns0:cell cols='3'>46.96 21.47</ns0:cell><ns0:cell cols='2'>24.75 31.36</ns0:cell></ns0:row><ns0:row><ns0:cell>TB incidence</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>Vietnam 76.96</ns0:cell><ns0:cell cols='2'>78.3 67.71</ns0:cell><ns0:cell>72.36</ns0:cell><ns0:cell /><ns0:cell cols='3'>50.37 49.41</ns0:cell><ns0:cell cols='2'>37.36 32.11</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='5'>Philippines 125.6 143.78 136.28 131.65</ns0:cell><ns0:cell /><ns0:cell>53.58</ns0:cell><ns0:cell cols='2'>51.9</ns0:cell><ns0:cell>99.74</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='4'>Thailand 96.14 111.82 113.46</ns0:cell><ns0:cell>73.67</ns0:cell><ns0:cell /><ns0:cell cols='3'>0 23.88</ns0:cell><ns0:cell cols='2'>35.43 11.73</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>China</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>15.66</ns0:cell><ns0:cell cols='2'>11.4</ns0:cell><ns0:cell cols='2'>12.99 13.33</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>B+ a</ns0:cell><ns0:cell>205</ns0:cell><ns0:cell>260</ns0:cell><ns0:cell>271</ns0:cell><ns0:cell>276</ns0:cell><ns0:cell /><ns0:cell>67</ns0:cell><ns0:cell /><ns0:cell>59</ns0:cell><ns0:cell>64</ns0:cell><ns0:cell>41</ns0:cell></ns0:row><ns0:row><ns0:cell>Clinical</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>B-b</ns0:cell><ns0:cell>234</ns0:cell><ns0:cell>290</ns0:cell><ns0:cell>349</ns0:cell><ns0:cell>325</ns0:cell><ns0:cell /><ns0:cell>40</ns0:cell><ns0:cell /><ns0:cell>31</ns0:cell><ns0:cell>29</ns0:cell><ns0:cell>39</ns0:cell></ns0:row><ns0:row><ns0:cell>characteristics</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>SS+ c</ns0:cell><ns0:cell>77</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>75</ns0:cell><ns0:cell>85</ns0:cell><ns0:cell /><ns0:cell>31</ns0:cell><ns0:cell /><ns0:cell>32</ns0:cell><ns0:cell>31</ns0:cell><ns0:cell>23</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>MDR-TB</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell /><ns0:cell>0</ns0:cell><ns0:cell /><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell></ns0:row><ns0:row><ns0:cell>TB</ns0:cell><ns0:cell /><ns0:cell>DOST</ns0:cell><ns0:cell>297</ns0:cell><ns0:cell>398</ns0:cell><ns0:cell>457</ns0:cell><ns0:cell>484</ns0:cell><ns0:cell /><ns0:cell>101</ns0:cell><ns0:cell /><ns0:cell>92</ns0:cell><ns0:cell>92</ns0:cell><ns0:cell>80</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>The impact of TB screening on clinical characteristics among marriage migrants with TB subjected to 1-2 rounds of post-entry screening vs. labor migrants with TB subjected to 4 rounds of post-entry screening for TB in Taiwan, 2012-2015</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell cols='2'>Marriage migrant Labor migrant</ns0:cell><ns0:cell>Odds Ratio</ns0:cell></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell>N=380</ns0:cell><ns0:cell>N=2323</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Sex</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Female</ns0:cell><ns0:cell>370</ns0:cell><ns0:cell cols='2'>1378 25.37(13.47-47.81)*</ns0:cell></ns0:row><ns0:row><ns0:cell>Age at diagnosis, years</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>&gt;=45</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>72</ns0:cell><ns0:cell>3.89(2.61-5.78) *</ns0:cell></ns0:row><ns0:row><ns0:cell>44=&lt;</ns0:cell><ns0:cell>338</ns0:cell><ns0:cell>2251</ns0:cell><ns0:cell>0.26(0.17-0.38)*</ns0:cell></ns0:row><ns0:row><ns0:cell>CXR</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Normal</ns0:cell><ns0:cell>40</ns0:cell><ns0:cell>238</ns0:cell><ns0:cell>1.03(0.72-1.47)</ns0:cell></ns0:row><ns0:row><ns0:cell>Abnormal without cavitation</ns0:cell><ns0:cell>270</ns0:cell><ns0:cell>1922</ns0:cell><ns0:cell>0.51(0.40-0.66)*</ns0:cell></ns0:row><ns0:row><ns0:cell>Abnormal with cavitation</ns0:cell><ns0:cell>60</ns0:cell><ns0:cell>141</ns0:cell><ns0:cell>2.90(2.10-4.01) *</ns0:cell></ns0:row><ns0:row><ns0:cell>No record</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>260</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Sputum smear</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SS-</ns0:cell><ns0:cell>246(64.7%)</ns0:cell><ns0:cell cols='2'>1834(78.9%) 0.49(0.39-0.62)*</ns0:cell></ns0:row><ns0:row><ns0:cell>SS+</ns0:cell><ns0:cell>107(28.2%)</ns0:cell><ns0:cell>318(13.7%)</ns0:cell><ns0:cell>4.82(3.7-6.34)*</ns0:cell></ns0:row><ns0:row><ns0:cell>No record</ns0:cell><ns0:cell>27</ns0:cell><ns0:cell>171</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Sputum culture</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SC+</ns0:cell><ns0:cell>222</ns0:cell><ns0:cell>990</ns0:cell><ns0:cell>1.89(1.52-2.39)*</ns0:cell></ns0:row><ns0:row><ns0:cell>SC-</ns0:cell><ns0:cell>148</ns0:cell><ns0:cell>1222</ns0:cell><ns0:cell>0.57(0.46-0.71)*</ns0:cell></ns0:row><ns0:row><ns0:cell>No record</ns0:cell><ns0:cell>10</ns0:cell><ns0:cell>824</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Bacterial status</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>B+</ns0:cell><ns0:cell>231(60.8%)</ns0:cell><ns0:cell>1012(43.6%)</ns0:cell><ns0:cell>2.01(1.61-2.51) *</ns0:cell></ns0:row><ns0:row><ns0:cell>B-</ns0:cell><ns0:cell>149(39.2%)</ns0:cell><ns0:cell>1311(56.4%)</ns0:cell><ns0:cell>0.50(0.40-0.62)*</ns0:cell></ns0:row><ns0:row><ns0:cell>*Odds ratio: significant</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='2'>SS: sputum smear; SC :sputum culture</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>The tuberculosis treatment outcomes of labor migrants with TB post-implementation (2014-2015) vs. pre-implementation (2012-2013) of the friendly therapy (FT) policy with no compulsory repatriation in Taiwan Comparison of treatment outcomes post-vs. pre-implementation of friendly therapy, odds ratio test: significant; either N or n': calculate reference for odds ratio ** The proportions of cases of initiate treatment between post-FT and pre-FT was significantly different which was approved by a Fisher's exact test, P= 0.0296 (P&lt;0.05) i.e., to reject the null hypothesis; n : denominator for rate</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Anti-TB treatment TB cases</ns0:cell><ns0:cell>Pre-FT N=1053</ns0:cell><ns0:cell>Post-FT N=1270</ns0:cell><ns0:cell>Odds ratio</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Indonesia</ns0:cell><ns0:cell>n'=507</ns0:cell><ns0:cell>n'=618</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transf. out</ns0:cell><ns0:cell>450</ns0:cell><ns0:cell>394</ns0:cell><ns0:cell>0.22(0.16-0.31)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Longer reg.</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Stand. Reg.</ns0:cell><ns0:cell>49</ns0:cell><ns0:cell>206</ns0:cell><ns0:cell>4.81(3.43-6.75)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>No record</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Vietnam</ns0:cell><ns0:cell>n'=175</ns0:cell><ns0:cell>n'=225</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transf. out</ns0:cell><ns0:cell>158</ns0:cell><ns0:cell>153</ns0:cell><ns0:cell>0.23(0.13-0.40)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Longer reg.</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Outcome by nationality</ns0:cell><ns0:cell>Stand. Reg. No record Philippines Transf. out</ns0:cell><ns0:cell>15 98 n'=237 212</ns0:cell><ns0:cell>67 5 n'=314 206</ns0:cell><ns0:cell>4.52(2.48-8.25)* 0.22(0.14-0.36)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Longer reg.</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Stand. Reg.</ns0:cell><ns0:cell>24</ns0:cell><ns0:cell>106</ns0:cell><ns0:cell>4.52(2.79-7.33)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>No record</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Thailand</ns0:cell><ns0:cell>n'=134</ns0:cell><ns0:cell>n'=111</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transf. out</ns0:cell><ns0:cell>132</ns0:cell><ns0:cell>84</ns0:cell><ns0:cell>0.05(0.01-0.20)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Longer reg.</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Stand Reg.</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>26</ns0:cell><ns0:cell>19.88(4.6-85.95)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>No record</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>SS-/DOST</ns0:cell><ns0:cell>n=537</ns0:cell><ns0:cell>n=784</ns0:cell><ns0:cell>1.60(1.3-1.96)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Completion</ns0:cell><ns0:cell>50</ns0:cell><ns0:cell>334</ns0:cell><ns0:cell>7.45(5.44-10.2) *</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Refuse</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transfer out</ns0:cell><ns0:cell>250</ns0:cell><ns0:cell>225</ns0:cell><ns0:cell>0.61(0.38-0.97) *</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Lost to follow up</ns0:cell><ns0:cell>14</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Side effect</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Not bac</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>16</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>other</ns0:cell><ns0:cell>170</ns0:cell><ns0:cell>175</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>SS+/DOST</ns0:cell><ns0:cell>n=124</ns0:cell><ns0:cell>n=117</ns0:cell><ns0:cell>0.75(0.45-1.25)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Completion</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>13.43(4.02-44.79) *</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transfer out</ns0:cell><ns0:cell>62</ns0:cell><ns0:cell>45</ns0:cell><ns0:cell>0.44(0.27-0.71) *</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Lost to follow up</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Side effect</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>0</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Outcome of DOST</ns0:cell><ns0:cell>Not bac other B+/DOST</ns0:cell><ns0:cell>1 49 n=332</ns0:cell><ns0:cell>0 35 n=382</ns0:cell><ns0:cell>0.93(0.78-1.11)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Refuse</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Complete</ns0:cell><ns0:cell>20</ns0:cell><ns0:cell>129</ns0:cell><ns0:cell>5.83 (3.62-9.42)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Side effect</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Lost to follow up</ns0:cell><ns0:cell>11</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transfer out</ns0:cell><ns0:cell>438</ns0:cell><ns0:cell>408</ns0:cell><ns0:cell>0.66( 0.56-0.79)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>B-/DOST</ns0:cell><ns0:cell>n=346</ns0:cell><ns0:cell>n=534</ns0:cell><ns0:cell>1.48(1.25-1.76)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Refuse</ns0:cell><ns0:cell>8</ns0:cell><ns0:cell>4</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Complete</ns0:cell><ns0:cell>33</ns0:cell><ns0:cell>240</ns0:cell><ns0:cell>6.87 (4.21-11.2)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Side effect</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Lost to follow up</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>18</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Transfer out</ns0:cell><ns0:cell>458</ns0:cell><ns0:cell>392</ns0:cell><ns0:cell>0.58( 0.49-0.69)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>SS-</ns0:cell><ns0:cell>n=805</ns0:cell><ns0:cell>n=1029</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>&lt; 30 d</ns0:cell><ns0:cell>571</ns0:cell><ns0:cell>793</ns0:cell><ns0:cell>1.38(1.12-1.70)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&gt; 30 d</ns0:cell><ns0:cell>145</ns0:cell><ns0:cell>157</ns0:cell><ns0:cell>0.82(0.64-1.05)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>No record</ns0:cell><ns0:cell>89</ns0:cell><ns0:cell>79</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>&lt;n 30 d rate**</ns0:cell><ns0:cell>0.71</ns0:cell><ns0:cell>0.77</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Treatment Initiation day of diagnosis</ns0:cell><ns0:cell>SS+ &lt; 30 d &gt; 30 d No record</ns0:cell><ns0:cell>n=158 147 2 9</ns0:cell><ns0:cell>n=160 143 3 14</ns0:cell><ns0:cell>0.63(0.29-1.39) 2.99(0.50-18.08)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&lt; 30 d rate</ns0:cell><ns0:cell>0.93</ns0:cell><ns0:cell>0.89</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>B-</ns0:cell><ns0:cell>n=399</ns0:cell><ns0:cell>n=658</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>&lt;30 day</ns0:cell><ns0:cell>307</ns0:cell><ns0:cell>512</ns0:cell><ns0:cell>1.64(1.38-1.95)*</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>&gt;30day</ns0:cell><ns0:cell>92</ns0:cell><ns0:cell>146</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>&lt; 30d rate**</ns0:cell><ns0:cell>0.77</ns0:cell><ns0:cell>0.78</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020) Manuscript to be reviewed * PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:04:47696:1:2:NEW 26 Sep 2020)</ns0:note> </ns0:body> "

" Dear editor, I hereby express my greatest gratitude to the reviewers’ taking endeavor in the relative work, especially the reviewer 2 (Mesay Hailu Dangisso) who provided a lot of very critical suggestions that really help us improving this paper. And, I have done much to revise the manuscript according to reviewer 2’s concerns and offering expertise suggestions. Please see the following lists and the detail in the resubmitted manuscript. Also, I notice reviewer 1‘s concerns and I have already made a lot of progress. With best regards Sincerely yours Mei-Mei Kuan Annex Response to Reviewer 2 (Mesay Hailu Dangisso) Reviewer 1 (Anonymous) Basic reporting This statistical analyses and results presentation remain not reach the basic standard for a scientific manuscript, so I am not able to verify the validity of this study. I highly recommend the author to find a statistical consultant and go through the analytic plans and the result tables carefully. Experimental design NA Validity of the findings See basic report Reviewer 2 (Mesay Hailu Dangisso) Basic reporting Dear Academic editor, Thank you for inviting me to review this manuscript “Surveillance of tuberculosis and treatment outcomes following screening and therapy interventions among marriage migrants and labor migrants from high TB endemic countries in Taiwan (#47696)”. I have comments and questions for the authors to improve their manuscript. Review comments Major comments 1. The justification for comparing only treatment outcomes of marriage migrants and labour migrants from TB high endemic countries should explained. Why the authors did not compare the treatment outcome with Taiwanese population data? Ans: Because the population backgrounds of marriage migrants and labour migrants with similar origins which were different from Taiwanese population i.e., both of them were coming from highly TB endemic countries including Thailand, Vietnam, Indonesia and the Philippine. The issue hereby is for addressing the different interventions policy of TB regulations measurements and different screening measurements among different migrants this study, therefore highlight on data of migrants. 2. The study did not use the same assessment criteria to assess TB trends. They used 1-2 rounds of post-entry TB screening for marriage migrants and 4 rounds of Post-entry TB screening. The authors should justify this. Minor comments Ans: We use the same assessment criteria to assess the TB trends by calculate the TB incidence as TB cases newly detected per 105 migrant population per year during post-entry periods each with different scenario . Abstract 3.The authors used labor migrant as different comparison groups, how do the authors know marital status of labour migrants? Ans: In this study, the marriage migrants were defined as those who emigrant into Taiwan for fulfilling the purpose of their marriages with Taiwanese. Thought the marital status of labour migrants might be an interesting issue it was not intended to be addressed in this study. 4.… “with 61% bacteria-positive and 31% smear-positive cases..” what is the difference between smear-positive and bacteria-positive? Smear-positives are also bacteria positive even if they are culture positive. Ans: Agree. We defined the bacteria-positive cases as the cases of sputum smear-positive with sputum culture- positive or cultured negative as well as smear-negative with culture-positive. 5.“TB cure/treatment completion” use either treatment completion or cure rate. Cure rate and completion are different. Ans: Agree. It has revised according to reviewer’s concerns. In statistics and analysis section 6..…..“The TB incidence (per 100,000 population) was calculated as the number of newly detected 114 labor migrants with TB during the study period (2012-2015) divided by the population per year”…... Is the population (denominator) the total population of Taiwan or total labour migrant population? This needs to be clear for the readers Ans: The TB incidence (per 100,000 population) was calculated as the number of TB cases newly detected among labor migrants during the study period (2012-2015) divided by the population per year. The population (denominator) is the total number of labour migrants by each year. e.g., newly detected cases in 2012 divided by the each total entry into Taiwan population of labor migrants and so on. 7.Line # 125-126. The authors estimated the treatment outcomes by considering transferred out cases and treatment completion rate only. For international comparison why the authors did not estimated the lost-to-follow up cases? Ans: According to the reviewer’s concerns, the lost-to-follow up cases has been added for comparison, please see the listing items in the Table 3. The related issues have been addressed in the results and discussions for details. 8.Lines # 151, 168, . postentry…is this post-entry? if so ,correct this throughout the manuscript Ans: The wordings have revised, according to the reviewer’s comments. 9.Why the authors presented both Pearson’s correlation and linear association? Ans: Using two method to quantify the strength to present might be underlined this status. 10.Why the authors did not categorize TB cases as bacteriological confirmed and smear-or bacteriologically negative cases? Ans: The manuscript has been revised according to the reviewer’s suggestions to categorize TB cases as bacteriological confirmed and smear-or bacteriologically negative cases. Please see the content in the results, the discussion, Table 1, 2, 3. Discussion 11.General comments; no need of presenting the results here in the discussion. All the statistics were already presented in the results section Ans: According to reviewer’s suggestions, the statistics has revised not to repeat in the discussions since that had been presented in the results. 12.Line # 246. ….after 2014 vs. before 2013 being less transfer out.. isn’t this less transferred out? not clear Ans: Yes. There were less transferred out after 2014. A total odds ratio has added i.e., after 2014 vs. before 2013 being less transfer out, odds ratio 0.21 (IC 95% 0.16- 0.26) and the corresponding content has been revised according to the reviewer’s concerns. According to reviewer’s suggestions this data has been added into content, please see Tables, results and discussions. 13.The authors used only ss- and ss+ cases in the discussion of treatment outcomes. What about bacteriologically confirmed cases; culture positive cases. The authors could consider merging either ss+ and culture positive cases together or include bacteriologically confirmed cases in the discussion. Ans: According to reviewer’s concerns and suggestions in the discussion of treatment outcomes, 1)it has added merging either ss+ or sputum culture positive cases together as bacteriologically confirmed (B+) cases for further discussion. 2) the data has renewed to process for calculate the number of TB cases with B+ vs. B- and their treatment outcome, please also see the details in the revised Table 3. Limitation 14.“The proportions of migrants with TB who had reactivation or transmission cases were not defined by the lab”. This statement is confusing and not clear. Ans: According to reviewer’s concerns, this sentence has been revised as “The proportions of migrants with TB who were reactivation or transmission cases were not defined by the DNA sequencing in this study.”. 15.Table 1. In clinical characteristics, the authors analyzed and reported B+ and SS+ separately. The authors need to make clear whether SS+ cases were included in the B+ cases. Ans: “…whether SS+ cases were included in the B+ cases” The answer is “Yes.” Please see Table 1, which is presented in the post script. 16.Table 2. Under CXR; “Abnormal w/o cavitation” and “Abnormal w/ cavitation, should be written in full. Ans: According to reviewer’s comment. The wordings are revised in full as “Abnormal without cavitation” and “Abnormal with cavitation” 17.Under sputum culture. CC+, and CC- should be made clear for the readers new for the field Ans: According to reviewer’s concern the CC+ meaning as sputum cell culture positive, and has been change as SC+ and CC- also has been changed as SC- meaning sputum cell culture negative. Experimental design Original primary research within Aims and Scope of the journal. Research question well defined, relevant & meaningful. It is stated how research fills an identified knowledge gap. Yes The submission should clearly define the research question, which must be relevant and meaningful. The knowledge gap being investigated should be identified, and statements should be made as to how the study contributes to filling that gap. Rigorous investigation performed to a high technical & ethical standard. Yes The investigation must have been conducted rigorously and to a high technical standard. The research must have been conducted in conformity with the prevailing ethical standards in the field. Methods described with sufficient detail & information to replicate. Yes, with minor revision Methods should be described with sufficient information to be reproducible by another investigator. Validity of the findings Conclusions are well stated, linked to original research question & limited to supporting results. With modification https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4252125/pdf/pone.0114225.pdf https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4252125/ Trends of tuberculosis case notification and treatment outcomes in the Sidama Zone, southern Ethiopia: ten-year retrospective trend analysis in urban-rural settings. PLoS One 2014 2;9(12):e114225. Epub 2014 Dec 2. Center for International Health, Faculty of Medicine and Dentistry, University of Bergen, Bergen, Norway. Background: Ethiopia is one of the high tuberculosis (TB) burden countries. An analysis of trends and differentials in case notifications and treatment outcomes of TB may help improve our understanding of the performance of TB control services. Methods: A retrospective trend analysis of TB cases was conducted in the Sidama Zone in southern Ethiopia. We registered all TB cases diagnosed and treated during 2003-2012 from all health facilities in the Sidama Zone, and analysed trends of TB case notification rates and treatment outcomes. Results: The smear positive (PTB+) case notification rate (CNR) increased from 55 (95% CI 52.5-58.4) to 111 (95% CI 107.4-114.4) per 105 people. The CNRs of PTB+ in people older than 45 years increased by fourfold, while the mortality of cases during treatment declined from 11% to 3% for smear negative (PTB-) (X2trend, P<0.001) and from 5% to 2% for PTB+ (X2trend, P<0.001). The treatment success was higher in rural areas (AOR 1.11; CI 95%: 1.03-1.2), less for PTB- (AOR 0.86; CI 95%: 0.80-0.92) and higher for extra-pulmonary TB (AOR 1.10; CI 95%: 1.02-1.19) compared to PTB+. A higher lost-to-follow up was observed in men (AOR 1.15; CI 95%: 1.06-1.24) and among PTB- cases (AOR 1.14; CI 95%: 1.03-1.25). More deaths occurred in PTB-cases (AOR 1.65; 95% CI: 1.44-1.90) and among cases older than 65 years (AOR 3.86; CI 95%: 2.94-5.10). Lastly, retreatment cases had a higher mortality than new cases (6% vs 3%). Conclusion: Over the past decade TB CNRs and treatment outcomes improved, whereas the disparities of disease burden by gender and place of residence reduced and mortality declined. Strategies should be devised to address higher risk groups for poor treatment outcomes. Download full-text PDF Source http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0114225 PLOS http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4252125 PMC "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Patients with hemodialysis suffer with protein-energy wasting and uremic myopathy lead to lack of physical activity and poor functional performance. However, ventilation abnormality in patients undergone hemodialysis remains controversial regarding the respiratory impairment. Therefore, the study aimed to determine the effect of duration of dialysis on respiratory function. Methods: A multicenter study with crosssectional study was designed in four hemodialysis outpatient clinics. Respiratory muscle strength (i.e., maximal inspiratory pressure (MIP) and maximal expiratory pressure (MEP)) pulmonary function test (i.e., forced vital capacity (FVC), forced expiratory volume in one second (FEV 1 ) and FEV 1 /FVC ratio), functional capacity (6-minute walk test) and sensation of breathlessness were assessed prior to dialysis. Results: A total of 100 hemodialysis patients were recruited with 38 females and 62 males. An average of duration of hemodialysis was 5.93&#177;4.96 years. Decreased MIP values, FEV 1 values, FVC values, %FEV 1 and %FVC were noted in patients with long duration of dialysis (defined as &#8805; 5 years of dialysis) compared to those with short duration of dialysis (p s &lt;.05). In addition, increased sensation of breathlessness was observed in patients with long duration of dialysis (p&lt;.05). Furthermore, participants with long duration of dialysis had an increased risk of ventilatory restriction (OR 6.093, p = .007).</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Chronic renal failure (CRF) is not only impact on loss of renal function, it has been also presented in cardiorespiratory problems including shortness of breathing. The effect of muscle wasting and uremia in patients with CRF who are on hemodialysis resulting in the respiratory and peripheral muscle weakness due to impair oxygen uptake, consumption, transportation and decrease in protein synthesis <ns0:ref type='bibr' target='#b4'>(Bark et al., 1988)</ns0:ref>. Pulmonary complication is one of the major problems in patients with CRF that is because of fluid overload, pulmonary calcification and fibrosis and pulmonary edema <ns0:ref type='bibr' target='#b8'>(Craddock et al, 1977)</ns0:ref>. In addition, inflammation process and protein-energy wasting contributed to cardiovascular disease and these may develop pulmonary impairment in uremic patients <ns0:ref type='bibr' target='#b28'>(Yoon, Choi &amp; Yun, 2009)</ns0:ref>. A retrospective longitudinal study in Korea found that impairment of pulmonary function was associated with increase in the development of CRF <ns0:ref type='bibr' target='#b17'>(Kim et al., 2018)</ns0:ref>. The prevalence of obstructive pulmonary function increased across GFR from 6% in GER-2 to 11% in GER-5 <ns0:ref type='bibr' target='#b21'>(Mukai et al., 2018)</ns0:ref>. Compared to healthy individuals, patients with CRF shown decreased respiratory muscle strength and lung function <ns0:ref type='bibr' target='#b4'>(Bark et al., 1988;</ns0:ref><ns0:ref type='bibr' target='#b16'>Karacan et al., 2006)</ns0:ref>. Besides, a restrictive lung disease was found 36% of CRF patients with GFR &lt; 15 L/min/1.73m 2 which markedly higher as GFR declined <ns0:ref type='bibr' target='#b22'>(Murtagh et al., 2007)</ns0:ref>. Previous study reported the prevalence of lung dysfunction was related to declining glomerular filtration rate (GFR) and the prevalence of dyspnea symptom was 35% among patients with end-stage renal disease <ns0:ref type='bibr' target='#b22'>(Murtagh et al., 2007)</ns0:ref>. In addition, patients with CRF receiving hemodialysis or patients with end stage renal disease (i.e., a GFR below 15 ml/min/1.73 m 2 ) are associated with high prevalence of cardiovascular disease e.g., hemodynamic stress, myocardial stress and injury <ns0:ref type='bibr' target='#b0'>(Ahmadmehrabi &amp; Wilson Tang, 2018)</ns0:ref>. Therefore, patients treated with hemodialysis are related to the risk of poor cardio-respiratory function. However, results of the relationships CRF patients and cardio-respiratory performance are quite inconsistent, with some suggesting obstructive and other restrictive pulmonary impairment in CRF. It might be differences in sample size, and participants may account for such differences. Therefore, the present study aimed to explore the relationships between cardiorespiratory performance in end stage renal failure patients undergoing hemodialysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>The study was approved from the Ethics Human committee of Thammasat University, based on Declaration of Helsinki, the Belmont report, CIOMS guidelines, and the International practice (ICH-GCP) COA No. 334/2560, and also the Ethics in Human Research Committee of the Thammasat Hospital Medicine. All participants were informed and gave written consent prior to the study. A total of 100 patients with end stage renal disease on hemodialysis were recruited in four hemodialysis centers. Those underwent hemodialysis at least 3 times per weeks and more than 3 months. Participants aged between 30-75 years old both males and females were recruited. Individuals who had history of no-smoking or non-current smoker or ex-smoker are included. A current smoker is defined as at least one cigarette per day within 1 week and ex-smoker is defined as participants who had ceased cigarette smoking in 6 months ago prior to the test <ns0:ref type='bibr' target='#b26'>(Thornton, Lee &amp; Fry, 1994)</ns0:ref>. The participant who had neurological problems (e.g., stroke), chronic cardio-respiratory disease (e.g., chronic cough, obstructive sleep apnea, history of chronic obstructive pulmonary disease or restrictive pulmonary disease, chronic heart failure), mental health problems, uncontrollable blood pressure (e.g., resting systolic blood pressure over 200 mmHg or resting diastolic blood pressure more than 120 mmHg) were excluded.</ns0:p><ns0:p>Pulmonary function tests (i.e., forced vital capacity: FVC, and forced expiratory volume in one second: FEV 1 ) were assessed by the spirometeric meter (Carefusion MicroLab, United Kingdom) and a calibration was performed prior to the test. In addition, the respiratory muscle strength (inspiratory muscle and expiratory muscle strength) were assessed by respiratory pressure meter, which is a RPM01 (Micro Medical Ltd., United Kingdom). Individuals were asked to exhale slowly and completely and then inhale deeply and sustained pressure for 1.5 s to evaluate the maximal inspiratory pressure (MIP). Regarding to the maximal expiratory pressure (MEP), participants were instructed to exhale deeply and hold for 1.5 s. these individuals performed 3-5 MIP and MEP maneuvers, with the highest two within 10 cmH 2 O was recorded. These tests were followed by a recommendation of the American Thoracic Society/ European Respiratory Society (2002). Functional capacity is defined as 6-minute walk test (6MWT) and all individuals were asked to walk 30 m straight along a corridor in 6 minutes (American Thoracic Society, 2002). In addition, the participants were requested to rate the sensation of breathless during the past four weeks. Self-reported perceived breathlessness was measured using a numerical scale (1-5) with 1 corresponding to 'very difficult to breath' and 5 corresponding to 'not difficult to breath'; lower scores higher breathlessness. Furthermore, all participants were required to performed these tests prior to underwent hemodialysis. The determination of pulmonary ventilation, an obstructive lung disease is defined as FEV1/FVC &lt; 0.70 and restrictive impairment is defined as FEV 1 /FVC &#8805; 0.70, and %FVC &lt; 80 <ns0:ref type='bibr' target='#b15'>(Johnson &amp; Theurer, 2014)</ns0:ref>.</ns0:p><ns0:p>Data were displayed as mean and standard deviation, percentage, as appropriate. Statistical significance was set as the level of p &lt; .05. Normality was verified using the Komogorov Siminov Goodness of Fitness test A chi-square test was used to analyze differences between short and long duration of dialysis and type of respiratory function. Pearson correlation was conducted to determine whether the duration of hemodialysis was related to cardiorespiratory function and sensation of breathlessness. To examine association between these relationships after controlling for age and sex, partial correlation analysis was conducted. Comparisons between two groups by categorized dialysis duration (defined as &lt; 5years and &#8805;5 years) were assessed with unpaired t-test.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The mean age of patients with end stage renal disease was 51.54&#177;11.19 years. A total of 100 participants with 38 females and 62 males completed pulmonary function test respiratory muscle strength testing and 6MWT. A mean duration of hemodialysis was 5.93&#177;4.96 years. According to categorise the lung function into three groups; normal, obstructive and restrictive lung impairment, the result revealed that only 17 individuals (17.00%) were categorised as a normal pulmonary function, whereas 82 participants (82.00%) were categorised as a restrictive ventilatory impairment, none of participants was defined as obstructive lung function. <ns0:ref type='table' target='#tab_0'>1</ns0:ref> here minute walk distance (6MWD) and sensation of breathlessness (p &lt; .05). In addition, these relationships remained after adjusting for age and sex; individuals with long duration of hemodialysis still had a poor cardio-respiratory performance (i.e., pulmonary function, respiratory muscle strength and functional capacity) and high sensation of breathlessness.</ns0:p><ns0:p>Several studies have been proposed for choosing different cutpoints of short-term and long-term duration of HD. For example, Hou et al. ( <ns0:ref type='formula'>2014</ns0:ref>) defined a long-term of HD as at least one year of HD. Another study defined as an average 51 months for the long period of HD treatment <ns0:ref type='bibr' target='#b7'>(Chazot et al., 2001)</ns0:ref>. Here, the study is considering an equal number of the participants. Therefore, the categorization of period of hemodialysis is presented using cut point of five years. The study revealed that 50 out of 100 (50.00%) of the CRF patients were categorized as long duration of dialysis (&#8805; 5 years). Individuals with short duration of dialysis had higher FVC values, %FVC, FEV 1 values, %FEV 1 MIP values and also lower sensation of breathlessness than those long duration of dialysis (see table <ns0:ref type='table' target='#tab_8'>3</ns0:ref>). Individuals with long duration of hemodialysis (defined as a duration of hemodialysis &#61619; 5 years) had shorter duration of 6MWT than those with short duration of hemodialysis ; however, these did not reach the conventional criterial (p&gt; .05). Patients with CRF who had a long duration of hemodialysis had a high prevalence of restrictive pulmonary impairment at 47% (n = 47) compared to individuals with short duration of dialysis (3.00%, n = 3) (&#61539; 2 = 8.575, p = .003).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The present study examined the relationship between duration of hemodialysis and cardio-respiratory performance in patients with CRF. The study found the association of MIP values, FVC values, predicted FVC, FEV 1 values, predicted FEV 1 , 6MWT and the sensation of breathlessness. The study also shown that compared with short duration of hemodialysis (defined as &lt; 5 years), patients who had a long duration of hemodialysis displayed a reduction in of MIP values, FVC values, predicted FVC, FEV 1 values, predicted FEV 1 , 6MWT and increased in the sensation of breathlessness. <ns0:ref type='bibr' target='#b4'>(Bark et al., 1988)</ns0:ref>. Therefore, the respiratory muscle strength weakness was noted in patients with CRF on hemodialysis. This might be due to uremic myopathy that leads to reduced strength in skeletal muscles and diaphragm <ns0:ref type='bibr' target='#b25'>(Tarasuik, Heimer, &amp; Bark, 1992)</ns0:ref>. <ns0:ref type='bibr' target='#b25'>Tarasuik et al. (1992)</ns0:ref> reported that force and frequency of diaphragmatic muscle was decreased by 15% in the moderate uremia and by 45% in the severely uremic rats. In addition, they found the fatigability of diaphragm was increased in the moderately and severely uremic rats <ns0:ref type='bibr' target='#b25'>(Tarasuik et al., 1992)</ns0:ref>. Therefore, muscle- Manuscript to be reviewed related CRF complication may be explained by uremic myopathy which can be attributed structural changes (e.g., a decrease in excitability-contractility coupling of respiratory muscle) <ns0:ref type='bibr' target='#b4'>(Bark et al., 1988)</ns0:ref>. Further, a deficit of vitamin D, anemia, hypophosphatemia, and malnutrition have been reported in patients with CRF <ns0:ref type='bibr' target='#b16'>(Karacan et al., 2006)</ns0:ref>. One mechanism linking uremia to muscle weakness is altered active Calcium-transport of sarcoplasmic reticulum. Heimberg et al. found that the Calcium-transport system in uremic rabbits was changed; decrease in the calcium influx rate, increase in the calcium permeability and resulting decreased concentrating ability of sarcoplasmic reticulum <ns0:ref type='bibr' target='#b12'>(Heimberg et al., 1976</ns0:ref>). Those might be, at least in part, due to resulting in a reduction of respiratory muscle strength.</ns0:p><ns0:p>In the study, decrease in pulmonary function has been observed in patients receiving hemodialysis. Decreased spirometry parameters in patients undergoing hemodialysis was reported in several studies <ns0:ref type='bibr' target='#b18'>(Kovacevic et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b20'>Kovelis et al., 2008)</ns0:ref>. These CRF patients on hemodialysis showed a restrictive ventilatory defect (defined as FVC &lt; 80% and FEV 1 /FVC &#8805; 0.7) with have been similarly reported from the previous studies that is restrictive pulmonary impairment was common complication in patients with advanced CRF <ns0:ref type='bibr' target='#b16'>(Karacan et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b20'>Kovelis et al., 2008)</ns0:ref>. This is due to fluid overload, increasing interstitial oedema and bronchial wall decongestion resulting in decreased pulmonary function <ns0:ref type='bibr' target='#b20'>(Kovelis et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b27'>Yilmaz et al., 2016)</ns0:ref>. In addition, uremia has been reported to related pulmonary microcirculation dysfunction <ns0:ref type='bibr' target='#b9'>(Ewert et al., 2002)</ns0:ref>. Another is that the decrease in pulmonary compliance of the thoracic wall that might be a muscle wasting with protein-energy wasting and inflammation <ns0:ref type='bibr' target='#b21'>(Mukai et al., 2018)</ns0:ref>. Further, Bark et al. found that decreased respiratory muscle strength was related to a decrease in vital capacity among patients with CRF <ns0:ref type='bibr' target='#b4'>(Bark et al., 1988)</ns0:ref>. Therefore, impaired respiratory muscles strength and poor ventilator capacity or lung restriction were noted in patients with end stage renal disease on hemodialysis. Previous studies have been reported that the relationships between lung function and inflammation were negatively associations <ns0:ref type='bibr' target='#b6'>(Bolton et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b11'>Hancox et al., 2016)</ns0:ref>. It is possible that a restrictive lung function may be in part, increased systematic inflammation. Therefore, cardio-pulmonary-renal interactions have been reported in CRF can cause a respiratory restriction, impaired gas exchange and also decreased exercise capacity <ns0:ref type='bibr' target='#b14'>(Husain-Syed et al., 2015)</ns0:ref>. However, further studies need to explore the relationship with impaired cardiopulmonary function in CRF patients.</ns0:p><ns0:p>According to shortness of breathing, it was found that duration of hemodialysis was positively associated with breathlessness even after controlling for age and sex; longer duration of hemodialysis higher breathlessness in hemodialysis patients. The prevalence of dyspnea was also reported between 20% and 60%. <ns0:ref type='bibr' target='#b23'>Palamidas et al. (2014)</ns0:ref> found that 100% of hemodialysis patients had displayed mild to moderated degree of chronic dyspnea (measured by Modified Medical Research Council Dyspnea Scale) before hemodialysis. However, it should be noted that only 25 hemodialysis patients were included in the study. In addition, Palamidas et al. assumed that an accumulation of excess lung water and pulmonary edema in patients with hemodialysis leads to premature airway closure and gas trapping. Further, a high ventilatory drive appealed before the hemodialysis which reflects increased respiratory effort and work of breathing <ns0:ref type='bibr' target='#b23'>(Palamidas et al., 2014)</ns0:ref> Therefore, impaired respiratory muscle and ventilatory impairment might be perceived as shortness of breathing from the hemodialysis patients.</ns0:p><ns0:p>The reduction in functional capacity determined by the 6MWT has been observed in several studies <ns0:ref type='bibr' target='#b5'>(Barril et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b10'>Fassbinder et al., 2015)</ns0:ref>. Fassbinder et al. found the distance of 6MWT was 418.67&#177;117.3 meters in 27 hemodialysis patients with an average age was 58.15&#177;10.84 years old <ns0:ref type='bibr' target='#b10'>(Fassbinder et al., 2015)</ns0:ref>. <ns0:ref type='bibr' target='#b5'>Barril et al. (2018)</ns0:ref> found that the walking distance was 586.39&#177;155 meters in 108 patients with advanced chronic kidney disease, mean aged was 67.36&#177;12.97 years which was lower than the predicted values. Further, they reported the GFR was associated with the distance of 6MWT in patients with advanced chronic kidney disease <ns0:ref type='bibr' target='#b5'>(Barril et al., 2018)</ns0:ref>. In the present study, the distance of 6MWT was 373.04&#177;107.80 meters with a mean overall age of the participants was 51.54&#177;11.19 years and duration of hemodialysis was 5.93&#177;4.96 years. Thus, the impairment in functional capacity in the present study show a lower distance compared to the previous studies <ns0:ref type='bibr' target='#b5'>(Barril et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b10'>Fassbinder et al., 2015)</ns0:ref>. It has been already known that the demographic data such as age height, body weight and race are associated with distance of 6MWT (American Thoracic Society, 2002; <ns0:ref type='bibr' target='#b24'>Poh et al., 2006)</ns0:ref>. In addition, the duration of hemodialysis might be in part of the walking distance in patients with CRF. Therefore, a prospective cohort study should be explored.</ns0:p><ns0:p>A number of limitation should be a consideration. A cross sectional study was designed which could not be a causal relationship. A confounding factors such as volume overload, laboratory analysis of renal function, and comorbidity was also not recorded. Other investigations such as laboratory investigation (e.g., arterial blood gas) or chest x-ray (e.g., interstitial lung disease) were not reported. In addition, all participants were only clinically stable; therefore, the results might not have generalized to CRF population as a whole. Finally, the sample size of the study was not calculated; however, the results of the study have reanalyzed with statistical power to detect the effect. According to a total number of 100 participants, the effect size for FVC values was 0.56; therefore, the retrospective statistical power was 0.79 for a 2-tailed alpha was 0.05. Thus, the findings of study are commensurate adequate statistical power. However, the future study need to consider regarding the sample size calculation.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Patients with long duration of hemodialysis displayed a reduction of respiratory function, functional capacity and also increased breathlessness. In addition, these individuals shown a risk of restrictive ventilatory impairment. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>displays the descriptive data of CRF patients. Bivariate correlations between duration of hemodialysis and pulmonary function, respiratory muscle strength tests and 6MWT were presented in table 2. Duration of hemodialysis was negatively associated with FVC values, FEV 1 values, MIP values, %FVC, % FEV 1 , 6-</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Table</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>The patients with hemodialysis in the study had low mean MIP values (60.92&#177;28.55 cmH 2 O); indicating inspiratory muscle weakness (defined as the ATS/ERS guideline which MIP values of &lt; 80 cmH 2 O). Decreased respiratory muscle value was consistent with other studies (Bark et al., 1988; Fassbinder et al., 2015; Karacan et al., 2006). Karacan et al. (2006) reported inspiratory muscle was displayed 66.5 &#177; 23.4 cmH 2 O in 27 hemodialysis patients. In addition, Bark et al. found the MIP was 58.2&#177;24.9 cmH 2 O in 10 patients with CRF group</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 3 hereTable 2 here</ns0:head><ns0:label>32</ns0:label><ns0:figDesc>PeerJ</ns0:figDesc><ns0:table /><ns0:note>reviewing PDF | (2020:08:51935:1:2:NEW 9 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Demographic and characteristics of chronic renal failure patients</ns0:figDesc><ns0:table><ns0:row><ns0:cell>N (%)</ns0:cell><ns0:cell>mean&#177;SD</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51935:1:2:NEW 9 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Correlation between duration of hemodialysis and respiratory function in chronic renal failure patients</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Correlation between duration of hemodialysis and respiratory function in chronic renal failure patients</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Correlation between duration of hemodialysis and respiratory function in chronic renal</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>failure patients</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>MIP</ns0:cell><ns0:cell>MEP</ns0:cell><ns0:cell>6-MWD</ns0:cell><ns0:cell>FVC(L)</ns0:cell><ns0:cell>%FVC</ns0:cell><ns0:cell>FEV1</ns0:cell><ns0:cell>%FEV1</ns0:cell><ns0:cell>FEV1/FVC</ns0:cell><ns0:cell>PEFR</ns0:cell><ns0:cell>%PEFR</ns0:cell><ns0:cell>Sensation of</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>(p-value)</ns0:cell><ns0:cell>Breathlessness</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>value)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>(p-value)</ns0:cell></ns0:row><ns0:row><ns0:cell>HD</ns0:cell><ns0:cell>-0.2.75</ns0:cell><ns0:cell>-0.187</ns0:cell><ns0:cell>-.210</ns0:cell><ns0:cell>-0.296</ns0:cell><ns0:cell>-0.248</ns0:cell><ns0:cell>-0.307</ns0:cell><ns0:cell>-0.255</ns0:cell><ns0:cell>-0.025</ns0:cell><ns0:cell>-0.072</ns0:cell><ns0:cell>0.019</ns0:cell><ns0:cell>-0.220</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>(0.006)</ns0:cell><ns0:cell>(0.063)</ns0:cell><ns0:cell>(0.036)</ns0:cell><ns0:cell>(0.003)</ns0:cell><ns0:cell>(0.013)</ns0:cell><ns0:cell>(0.002)</ns0:cell><ns0:cell>(0.011)</ns0:cell><ns0:cell>(0.805)</ns0:cell><ns0:cell>(0.478)</ns0:cell><ns0:cell>(0.848)</ns0:cell><ns0:cell>(0.028)</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>HD # -0.229</ns0:cell><ns0:cell>-0.135</ns0:cell><ns0:cell>-0.200</ns0:cell><ns0:cell>-0.265</ns0:cell><ns0:cell>-0.238</ns0:cell><ns0:cell>-0.279</ns0:cell><ns0:cell>-0.232</ns0:cell><ns0:cell>-0.031</ns0:cell><ns0:cell>-0.007</ns0:cell><ns0:cell>0.032</ns0:cell><ns0:cell>-0.224</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>(0.024)</ns0:cell><ns0:cell>(0.186)</ns0:cell><ns0:cell>(0.048)</ns0:cell><ns0:cell>(0.008)</ns0:cell><ns0:cell>(0.018)</ns0:cell><ns0:cell>(0.005)</ns0:cell><ns0:cell>(0.021)</ns0:cell><ns0:cell>(0.759)</ns0:cell><ns0:cell>(0.947)</ns0:cell><ns0:cell>(0.756)</ns0:cell><ns0:cell>(0.027)</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'># Controlling for age and sex</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='11'>HD; hemodialysis, FVC; forced vital capacity, FEV1; forced expiratory volume in the first second, PEF; peak</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='9'>expiratory flow, MIP; maximal inspiratory pressure, MEP; maximal expiratory pressure</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:51935:1:2:NEW 9 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>displays the differences in pulmonary function and respiratory muscle strength by categorized duration of hemodialysis</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_10'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>displays the differences in pulmonary function and respiratory muscle strength by categorized duration of hemodialysis</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:08:51935:1:2:NEW 9 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_11'><ns0:head>Table 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>displays the differences in pulmonary function and respiratory muscle strength by categorized duration of hemodialysis</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Duration of HD &lt; 5yrs</ns0:cell><ns0:cell>Duration of HD &#8805;5years</ns0:cell><ns0:cell>t(98)</ns0:cell><ns0:cell>p-value</ns0:cell><ns0:cell>95% CI</ns0:cell><ns0:cell>Effect size</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>(n= 50)</ns0:cell><ns0:cell>(n=50)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>Cohen's d</ns0:cell></ns0:row><ns0:row><ns0:cell>MIP (cmH 2 O)</ns0:cell><ns0:cell>67.42&#177;25.47</ns0:cell><ns0:cell>55.42&#177;30.33</ns0:cell><ns0:cell>2.142</ns0:cell><ns0:cell>0.035</ns0:cell><ns0:cell>0.88 to 23.12</ns0:cell><ns0:cell>0.428</ns0:cell></ns0:row><ns0:row><ns0:cell>MEP(cmH 2 O)</ns0:cell><ns0:cell>71.10&#177;25.93</ns0:cell><ns0:cell>64.18&#177;33.91</ns0:cell><ns0:cell>1.146</ns0:cell><ns0:cell>0.254</ns0:cell><ns0:cell>-5.06 to 18.90</ns0:cell><ns0:cell>0.229</ns0:cell></ns0:row><ns0:row><ns0:cell>6-MWD (meters)</ns0:cell><ns0:cell>386.14&#177;122.29</ns0:cell><ns0:cell>359.94&#177;90.39</ns0:cell><ns0:cell>1.218</ns0:cell><ns0:cell>0.226</ns0:cell><ns0:cell>-16.48 to 68.88</ns0:cell><ns0:cell>0.244</ns0:cell></ns0:row><ns0:row><ns0:cell>FVC (L)</ns0:cell><ns0:cell>2.29&#177;0.74</ns0:cell><ns0:cell>1.89&#177;0.69</ns0:cell><ns0:cell>2.777</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell>0.11 to 0.68</ns0:cell><ns0:cell>0.559</ns0:cell></ns0:row><ns0:row><ns0:cell>FVC(%)</ns0:cell><ns0:cell>69.74&#177;18.27</ns0:cell><ns0:cell>59.68&#177;16.68</ns0:cell><ns0:cell>2.877</ns0:cell><ns0:cell>0.005</ns0:cell><ns0:cell>3.12 to 17.01</ns0:cell><ns0:cell>0.575</ns0:cell></ns0:row><ns0:row><ns0:cell>FEV 1 (L)</ns0:cell><ns0:cell>2.12&#177;0.66</ns0:cell><ns0:cell>1.77&#177;0.63</ns0:cell><ns0:cell>2.706</ns0:cell><ns0:cell>0.008</ns0:cell><ns0:cell>0.09 to 0.60</ns0:cell><ns0:cell>0.542</ns0:cell></ns0:row><ns0:row><ns0:cell>FEV 1 (%)</ns0:cell><ns0:cell>76.60&#177;21.56</ns0:cell><ns0:cell>65.46&#177;18.74</ns0:cell><ns0:cell>2.758</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell>3.13 to 19.16</ns0:cell><ns0:cell>0.552</ns0:cell></ns0:row><ns0:row><ns0:cell>FEV 1 /FVC</ns0:cell><ns0:cell>93.70&#177;5.78</ns0:cell><ns0:cell>94.54&#177;5.93</ns0:cell><ns0:cell>-0.717</ns0:cell><ns0:cell>0.475</ns0:cell><ns0:cell>-3.16 to 1.48</ns0:cell><ns0:cell>0.143</ns0:cell></ns0:row><ns0:row><ns0:cell>PEFR</ns0:cell><ns0:cell>278.48&#177;126.02</ns0:cell><ns0:cell>269.40&#177;120.31</ns0:cell><ns0:cell>0.369</ns0:cell><ns0:cell>0.713</ns0:cell><ns0:cell>-39.82 to 57.98</ns0:cell><ns0:cell>0.074</ns0:cell></ns0:row><ns0:row><ns0:cell>(%) PEFR</ns0:cell><ns0:cell>58.80&#177;24.63</ns0:cell><ns0:cell>58.57&#177;21.60</ns0:cell><ns0:cell>0.050</ns0:cell><ns0:cell>0.960</ns0:cell><ns0:cell>-8.96 to 9.43</ns0:cell><ns0:cell>0.010</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Sensation of breathlessness 4.08&#177;0.90</ns0:cell><ns0:cell>3.44&#177;1.28</ns0:cell><ns0:cell>2.892</ns0:cell><ns0:cell>0.005</ns0:cell><ns0:cell>0.20 to 1.08</ns0:cell><ns0:cell>0.578</ns0:cell></ns0:row></ns0:table><ns0:note>HD; hemodialysis, MIP; maximal inspiratory pressure, MEP; maximal expiratory pressure, FVC; forced vital capacity, FEV 1 ; forced expiratory volume in the first second, 6MWD; 6-minute walk distance</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:51935:1:2:NEW 9 Oct 2020)</ns0:note> </ns0:body> "

"Dear Editor/reviewers, Thank you very much for your letter and the reviewers’ reports. We deeply appreciate the valuable comments you have provided. We have highlighted the sections with major revision in red. Below is our response to their comments resulting in a number of clarifications. Entitled “Duration of hemodialysis associated with cardio-respiratory dysfunction and breathlessness: a multicenter study” Original comments of the reviewer Reply by the author(s) Changes done on page number and line number Reviewer #1: Basic reporting Writing generally good; needs slight English language polishing which will likely be done in the editing process. The manuscript has been edited by an English-speaking native In the abstract, the sentence 'These patients are also related to poor respiratory performance due to pulmonary impairment' should be removed, as it is presumptively asserting facts which this study is aiming to prove. We removed from the sentence. Background. Patients with hemodialysis suffer with protein-energy wasting and uremic myopathy lead to lack of physical activity and poor functional performance. These patients are also related to poor respiratory performance due to pulmonary impairment. Experimental design It should be stated explicitly in the abstract and in the methods that this was a prospective study The study was not a prospective study. it was a cross-sectional study; therefore, the design of the study has been added in the abstract part. Please see on page 1; abstract part A multicenter study with cross-sectional study was designed in four hemodialysis outpatient clinics. There needs to be more detail about what 'chronic cardio-respiratory conditions' were exclusion criteria; does this include any restrictive or obstructive respiratory pathology? Obstructive sleep apnoea? Heart failure? How about smoking history? Obstructive sleep apnoea and history of COPD or restrictive disease, chronic heart failure have been defined as a chronic cardiorespiratory conditions; therefore, we added in the method part, Please see on page 3, line 92-94. (e.g., chronic cough, obstructive sleep apnea, history of chronic obstructive pulmonary disease or restrictive pulmonary disease, chronic heart failure). We also added the inclusion criteria regarding the history of smoking; please see on page 3; line 89-91. Individuals who had history of no-smoking or non-current smoker or ex-smoker are included. A current smoker is defined as at least one cigarette per day within 1 week and ex-smoker is defined as participants who had ceased cigarette smoking in 6 months ago prior to the test (Thornton, Lee & Fry, 1994). How was duration of haemodialysis defined for the analysis- in years, or months? Several studies have been proposed for choosing different cutpoints of short-term and long-term duration of HD. For example, Hou et al. (2014) defined a long-term of HD as at least one year of HD. Another study defined as an average 51 months for the long period of HD treatment (Chazot et al., 2001) In addition, we are considering an equal number of the participants. Therefore, the duration of HD was categorized by using 5 years of HD; please see on page 4, line 150-154. Several studies have been proposed for choosing different cutpoints of short-term and long-term duration of HD. For example, Hou et al. (2014) defined a long-term of HD as at least one year of HD. Another study defined as an average 51 months for the long period of HD treatment (Chazot et al., 2001). Here, the study is considering an equal number of the participants. Where continuous data was expressed as mean and standard deviation, were tests of normality performed to check if this data was normally distributed? (If any of the continuous data variables were non-normally distributed, median +/- IQR would be more appropriate) Data was verified for normality of distribution; please see on page 4, line 118-119. Normality was verified using the Komogorov Siminov Goodness of Fitness test. Validity of the findings Line 142 to 144 do not make sense: 'Patients with CRF who had a long duration of HD had a high prevalence of restrictive pulmonary impairment at 47% (n = 47) compared to individuals with normal pulmonary function (3.00%, n = 3) (X2 = 8.575, p = .003).' This paragraph is concerned with comparisons between long duration of HD and short duration of HD groups, not Chi-Squared comparisons within the same group. We have changed from individuals with normal pulmonary function to individuals with short duration of dialysis. Please see on page 5; line 163. individuals with short duration of dialysis Table 4 is totally missing. Furthermore, even if Table 4 were included, it is unclear what is the purpose of the logistic regression analysis. Is it univariate or multivariate logistic regression analysis? We deleted the table 4. It appears that the binary dependent variables being studied in the regression are long duration of HD and short duration of HD. However, these are not outcomes that are influenced or predicted by other independent variables- they are simply a function of the passage of time since the patient started dialysis. Using clinical/respiratory parameters to 'predict' whether someone has been on dialysis for a long time has no practical utility. I would remove the logistic regression analysis altogether. We removed the logistic regression analysis on page 4-5. Finally, determine the risk of duration of HD, a logistic regression analysis for cardio-respiratory performance was performed. In the discussion, line 196/197 'The present study found that the association between FVC and inspiratory muscle strength was observed in HD (data was not shown)' is inadequate: if this is an interesting point, this data should be shown in the manuscript with the appropriate statistical analysis. Perhaps this data, comparing the association between different respiratory variables in this cohort, should be presented instead of the above-mentioned logistic regression The purposes of the study were only an examination of the effect of the duration of dialysis and the relationships between cardio-respiratory performance in end stage renal failure patients undergoing HD. Therefore, the study did not focus on the relationships between respiratory muscle strength and lung function. Therefore, we deleted the sentence. The present study found that the association between FVC and inspiratory muscle strength was observed in HD (data was not shown). Reviewer 2 -The number of literature references could be more extensive to provide suitable context. We have added other references. - The article structure is suitable however in this topics below I suggest: Introduction: Hypotheses should be described in this section. What is the rational for including the comments about GFR and this variable was not described in methods, tables or discussion? I think that GFR is a very important variable. The study explored the relationships between CRF and cardio-respiratory performance and because the study recruited only hemodialysis patients which means CRF stage 5 (GFR less than 15). Data also recruited in the outpatient clinics and all participants had been received hemodialysis 3 times / weeks. Therefore, the GFR values was not recorded in the OPD card. The hypothesis in the present study is that duration of hemodialysis might be associated with a poor cardio-respiratory performance. We added some details on page 3; line 69-73. In addition, patients with CRF receiving hemodialysis or patients with end stage renal disease (i.e., a GFR below 15 ml/min/1.73 m2) are associated with high prevalence of cardiovascular disease e.g., hemodynamic stress, myocardial stress and injury (Ahmadmehrabi & Tang, 2018). Therefore, patients treated with hemodialysis are related to the risk of poor cardio-respiratory function. Basic reporting The research question was well defined and relevant however there are many studies about these questions. Rigorous investigation of technical and ethical aspects. The methods section was sufficient but I have a question: • Why the authors did not use reference formulas to predict distance in TC6 and values of PImax and PEmax? These equations are essential to described and interpreted because comparing values against healthy adults. • There are several reference equations for 6-min walk distance and also PImax, PEmax. It depends on different age, sex, ethnic, for example. Therefore, the study did not interpret these values. However, the study compared the mean MIP values with ATS/ERS guideline and also 6MWD was compared with other studies. Please see on page 5-6. The patients with hemodialysis in the study had low mean MIP values (60.92±28.55 cmH2O); indicating inspiratory muscle weakness (defined as the ATS/ERS guideline which MIP values of < 80 cmH2O); page 5; line 174-176). In the present study, the distance of 6MWT was 373.04±107.80 meters with a mean overall age of the participants was 51.54±11.19 years and duration of hemodialysis was 5.93±4.96 years. Thus, the impairment in functional capacity in the present study show a lower distance compared to the previous studies (Barril et al., 2018; Fassbinder et al., 2015); page 6; line 238-242. Was the sample size estimated? We are not calculated the sample size; therefore, this is one of the limitation of the study; please see on page 7;line 254-260. Finally, the sample size of the study was not calculated; however, the results of the study have reanalyzed with statistical power to detect the effect. According to a total number of 100 participants, the effect size for FVC values was 0.56; therefore, the retrospective statistical power was 0.79 for a 2-tailed alpha was 0.05. Thus, the findings of study are commensurate adequate statistical power. However, the future study need to consider regarding the sample size calculation. Why the authors chose the categorized dialysis duration (defined as < 5years and ≥5 years)? Was based in previous literature? Several studies have been proposed for choosing different cutpoints of short-term and long-term duration of HD. For example, Hou et al. (2014) defined a long-term of HD as at least one year of HD. Another study defined as an average 51 months for the long period of HD treatment (Chazot et al., 2001) In addition, we are considering an equal number of the participants. Therefore, the duration of HD was categorized by using 5 years of HD; please see on page 4, line 150-154. Several studies have been proposed for choosing different cutpoints of short-term and long-term duration of HD. For example, Hou et al. (2014) defined a long-term of HD as at least one year of HD. Another study defined as an average 51 months for the long period of HD treatment (Chazot et al., 2001). Here, the study is considering an equal number of the participants. Validity of the findings The results showed a benefit to literature however, the external validity is hard to be concluded. The power of study was not showed without a priori calculate of sample size. There is a limitation of the study. It is because the sample size was not calculated; therefore, the power of study was not reported. However, we have reanalyzed with statistical power to detect the effect. According to 100 number of sample size, the effect size for FVC values was 0.56; therefore, the retrospective statistical power was 0.79 for a 2-tailed alpha was 0.05. Thus, the findings of study are commensurate adequate statistical power. Regarding the MIP values, the effect size was 0.43; therefore, the retrospective statistical power was 0.56 for a 2-tailed alpha = 0.05. We also added the limitations of the study regarding the calculation of sample size; please see on page 7; line 254-260. In addition, we have analyzed the effect size (Cohen’s d); please see table 3. Finally, the sample size of the study was not calculated; however, the results of the study have reanalyzed with statistical power to detect the effect. According to a total number of 100 participants, the effect size for FVC values was 0.56; therefore, the retrospective statistical power was 0.79 for a 2-tailed alpha was 0.05. Thus, the findings of study are commensurate adequate statistical power. However, the future study need to consider regarding the sample size calculation. Information about laboratory analysis of renal function would be very important in this population. We are considering the renal function laboratory; therefore, this is also one of the limitation. We have added in the limitation of the study; please see on page 7; line 251. laboratory analysis of renal function, Conclusions were linked to original questions. The conclusion was linked to the original questions; please see on page 7; line 263-266. Patients with long duration of hemodialysis displayed a reduction of respiratory function, functional capacity and also increased breathlessness. In addition, these individuals shown a risk of restrictive ventilatory impairment. General comments: Thank you very much for allowing me to review this manuscript. Thank you for giving us the feedback, we appreciate for all the comments. Comments for the Author Overall, this article is very interesting and does provide additive data to the existing published literature. Thank you for giving us the opportunity to improve and resubmit our manuscript. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Notch family proteins play a key role in a variety of developmental processes by controlling cell fate decisions and operating in a great number of biological processes in several organ systems, such as hematopoiesis, somatogenesis, vasculogenesis, neurogenesis and homeostasis. The Notch signaling pathway is crucial for the majority of developmental programs and regulates multiple pathogenic processes. Notch family receptors' activation has been largely related to its multiple effects in sustaining oncogenesis. The Notch signaling pathway constitutes an ancient and conserved mechanism for cell to cell communication. Much of what is known about Notch family proteins function comes from studies done in Caenorhabditis Elegans and Drosophila Melanogaster. Although, human Notch homologs had also been identified, the molecular mechanisms which modulate the Notch signaling pathway remained substantially unknown. In this study, an updated evolutionary analysis of the Notch family members among 603 different organisms of all kingdoms, from bacteria to humans, was performed in order to discover key regions that have been conserved throughout evolution and play a major role in the Notch signaling pathway. The major goal of this study is the presentation of a novel updated phylogenetic tree for the Notch family as a reliable phylogeny 'map', in order to correlate information of the closely related members and identify new possible pharmacological targets that can be used in pathogenic cases, including cancer.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The Notch gene was originally discovered by Dexter in 1914 and it was named from the irregular notched wing phenotype of Drosophila melanogaster, caused by the loss-of-function in the responsible gene's after a point mutation <ns0:ref type='bibr' target='#b4'>(Dexter, 1914)</ns0:ref>. Since then, Notch protein and its homologs, Notch1, Notch2, Notch3, Notch4, LIN-12 and GPL-1 have been identified in genomes from all kingdoms, indicating the progressive differentiation of Notch family. Their length varies from &#8776;110 amino acids in bacteria <ns0:ref type='bibr' target='#b6'>(Durieux, et al., 2019)</ns0:ref> to &#8776;4500 amino acids (aa) in animals <ns0:ref type='bibr' target='#b7'>(Fairclough, et al., 2013)</ns0:ref>. Notch family members are evolutionary conserved, type-1 transmembrane glycoproteins, that function both as transmembrane receptors for ligands and transcription factors <ns0:ref type='bibr' target='#b18'>(Kopan and Ilagan, 2009)</ns0:ref>. They regulate cell fate determination and promote cell differentiation, maintenance and survival. These proteins have either overlapping or unique cellular functions, but these functions remain quite unclarified, in the majority of the organisms found <ns0:ref type='bibr' target='#b12'>(Hogan and Bautch, 2004)</ns0:ref>.</ns0:p><ns0:p>In mammals, there are four genes that encode four paralogue Notch transmembrane receptors, Notch1 to 4. The Notch1 gene is essential for developmental processes, while loss-of-function mutations in result to early fetal death, due to dysfunctional angiogenesis, organogenesis and cardiogenesis. Moreover, it plays a key role in the definitive formation of Hematopoietic Stem Cells (HSCs), responsible for the production of all mature blood cells during adulthood <ns0:ref type='bibr' target='#b43'>(Tanigaki and Honjo, 2007)</ns0:ref>. Notch1 protein consists of approximately 2627 aa and its signaling pathway regulates the development of B and T lymphocytes <ns0:ref type='bibr' target='#b9'>(Gerhardt, et al., 2014)</ns0:ref>. The Notch2 gene encodes a &#8776;2471aa receptor and determines the cell fate in the heart, liver, kidneys, teeth, bones as well as other cell types that are being developed in the fetus. After birth, Notch2 signaling is involved in the immune system's function, tissue repair, homeostasis and bone reshaping when needed. The Notch3 gene encodes a &#8776;2321aa receptor, which determines the fate of Vascular Smooth Muscle Cells (VSMCs) in the arterial network of the brain and finally the Notch4 gene encodes a &#8776;2059aa receptor, that determines fetal vascular morphogenesis and remodeling <ns0:ref type='bibr' target='#b20'>(Krebs, et al., 2000;</ns0:ref><ns0:ref type='bibr' target='#b45'>Vlachakis, et al., 2014)</ns0:ref>.</ns0:p><ns0:p>Notch family protein domains have been conserved throughout evolutionary history, from invertebrates to humans <ns0:ref type='bibr' target='#b13'>(Hori, et al., 2013)</ns0:ref>. They are composed of an extracellular domain (Notch extracellular domain, NECD), a transmembrane domain and an intracellular domain (Notch intracellular domain, NICD) <ns0:ref type='bibr' target='#b48'>(Yavropoulou, et al., 2015)</ns0:ref>. The NECD contains 29 to 36 Epidermal Growth Factor-like Repeats (EGF-like domain), depending on the type of receptor and a Negative Regulatory Region (NRR). The NRR is composed of three cysteine-rich Notch / LIN-12 repeats (LNRs) and a Heterodimerization Domain (HD). Each EGF-like repeat has 6 cysteines, which form 3 disulfide bonds, contributing to the 3D structure of the protein <ns0:ref type='bibr' target='#b27'>(Mui&#241;o, et al., 2017)</ns0:ref>. The NICD has a RBPJ&#954;-associated molecule domain (RAM), nuclear localization sequences (NLS), 7 ankyrin repeats (ANK) domain, a transcriptional activation domain (TAD) and a C-terminal Pro Glu Ser Thr (PEST) domain <ns0:ref type='bibr' target='#b13'>(Hori, et al., 2013)</ns0:ref>. Although Notch receptors are highly conserved, they have some structural variation mainly in the number of EGF-like repeats, in the presence of the TAD domain, and the length of the segment between the ANK repeats and C-terminal <ns0:ref type='bibr' target='#b38'>(Sander, et al., 2006)</ns0:ref>. Notch signaling pathway is highly conserved, with a key role in cell-cell communication. It regulates the vascular development and physiology, as well as multiple developmental processes of the Central Nervous System (CNS). The pathway is involved in the regulation of Nerve Stem Cells' (NCSs) proliferation, survival, self-renewal and differentiation and has been associated with early neurodevelopment, learning and memory, as well as neurodegeneration <ns0:ref type='bibr' target='#b25'>(Mizutani, et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b33'>Polychronidou, et al., 2015)</ns0:ref>. Thus, mutations in Notch pathway participants, that lead to defective signaling, cause a variety of human diseases including neurodegenerative diseases, developmental disorders and cancer (Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Dataset collection and filtering</ns0:head><ns0:p>Data was collected from the NCBI database (ncbi.nlm.nih.gov) as previously described in <ns0:ref type='bibr'>Mitsis et al. (2020)</ns0:ref>, towards to extracting the amino acid sequences that are related to the Notch proteins using the keyword 'Notch' <ns0:ref type='bibr'>(Mitsis, et al., 2020)</ns0:ref>. Protein sequences that responded to the query but did not include the Notch family members were eliminated from the primary dataset, by using related keywords and regular expressions techniques in the header information, and local alignments with reference protein sequences. Furthermore, a final dataset for each species class was produced by using internal protein alignments and protein identity score. Duplicated protein sequences in each species that were found share 95% &gt; protein identity within the dataset were removed. In total, 25,761 Notch family related protein sequences were identified from several species, and a dataset containing 603 unique, non-duplicate protein sequences was created (Sup. Dataset 1).</ns0:p></ns0:div> <ns0:div><ns0:head>Multiple sequence alignment</ns0:head><ns0:p>Multiple sequence alignment (MSA) was executed using the MATLAB Bioinformatics Toolbox, utilizing a guide tree and the progressive MSA method as previously described in several studies <ns0:ref type='bibr'>(Mitsis, et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b30'>Papageorgiou, et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b40'>Sobie, 2011)</ns0:ref>. Pairwise distances among sequences were estimated based on the pairwise alignment with the 'Gonnet' method and followed by calculating the differences between each pair of sequences. The Neighbor-Joining method was used towards to estimating the guide tree by assuming equal variance and independence of evolutionary distance estimates (Sup. Dataset 2). Finally, consensus sequence was calculated and visualized through the JalView platform <ns0:ref type='bibr' target='#b46'>(Waterhouse, et al., 2009)</ns0:ref> using the multiple sequences alignment results and parameters including amino acid conservation. The commentary section of Jalview, which presents the amino-acid conservation using logos and histograms, was further observed to uncover innovative motifs.</ns0:p></ns0:div> <ns0:div><ns0:head>Notch family protein clusters</ns0:head><ns0:p>Notch family protein clusters were identified by using phylogenetic analysis. The phylogenetic analysis was performed using the MATLAB Bioinformatics Toolbox <ns0:ref type='bibr' target='#b21'>(Kufareva and Abagyan, 2012)</ns0:ref> utilizing the Unweighted Pair-Group Method (UPGMA) <ns0:ref type='bibr' target='#b22'>(Michener and Sokal, 1957;</ns0:ref><ns0:ref type='bibr' target='#b32'>Pavlopoulos, et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b39'>Sneath, 1973)</ns0:ref> while the matrix of the pairwise distances was calculated using the protein-adapted Jukes-Cantor statistical method <ns0:ref type='bibr' target='#b30'>(Papageorgiou, et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b47'>Yang and Zhang, 2008)</ns0:ref>. The constructed phylogenetic tree visualized using MEGA radiation option and the final Notch family protein clusters separated in different sub-datasets using a threshold. In total, eight major sub-clusters including Notch Bacteria, Notch_Plants, Notch_Invertebrates, Notch_Protist, Notch1, Notch2, Notch3 and Notch4 were identified (Sup. Dataset 3).</ns0:p></ns0:div> <ns0:div><ns0:head>Consensus sequences and a specialized phylogenetic analysis</ns0:head><ns0:p>Representative consensus protein sequences of the Notch family sub-clusters were estimated according to the clustering results of the Notch family protein clusters <ns0:ref type='bibr'>(Mitsis, et al., 2020)</ns0:ref>. The representative protein sequences for each sub-cluster were calculated using the MATLAB Bioinformatics Toolbox <ns0:ref type='bibr' target='#b28'>(Nanni, et al., 2014)</ns0:ref>. Two representative protein sequences of the GLP1 and LIN12 homologs of the Caenorhabditis elegans species <ns0:ref type='bibr' target='#b10'>(Girard, et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b41'>Sorkac, et al., 2018)</ns0:ref> were also included in the final dataset of the consensus protein sequences (Sup. Dataset 4). Last but not least, a specialized phylogenetic analysis was performed using the MATLAB Bioinformatics Toolbox <ns0:ref type='bibr' target='#b21'>(Kufareva and Abagyan, 2012)</ns0:ref> utilizing the Unweighted Pair-Group Method (UPGMA) <ns0:ref type='bibr' target='#b22'>(Michener and Sokal, 1957;</ns0:ref><ns0:ref type='bibr' target='#b32'>Pavlopoulos, et al., 2010;</ns0:ref><ns0:ref type='bibr' target='#b39'>Sneath, 1973)</ns0:ref> with 100 bootstrap replicates. Finally, the constructed topology of the phylogenetic tree was visualized with MEGA traditional option (Sup. Dataset 5).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Dataset</ns0:head><ns0:p>The primary NCBI dataset contained 26646 entries, related to Notch family members. This dataset consisted of not only Notch protein and its homologs, but also unrelated proteins, such as the example of strawberry notch homolog (SNO). These irrelevant proteins, along with synthetic, hypothetical, partial, low quality, predicted proteins and which were referred as to noisy data and were removed from our dataset. Since the NCBI database provides partial duplicates sequences, the sequences with greater than 95% similarity were also removed, retaining the one with the longer length. Thus, the final dataset involved 603 Notch proteins and specifically, 90% of these corresponded to Notch protein, 4% to Notch1, 3% to Notch2, 2% to Notch3, and 1% to Notch4 (Figure <ns0:ref type='figure' target='#fig_0'>1</ns0:ref>). Herein, representatives of all kingdoms from simpler organisms such as bacteria, to more complex organisms including Homo sapiens were detected. The length of the Notch family proteins ranges between &#8776;71 aa (bacteria's Notch-like protein) and &#8776;4835 aa (invertebrates Notch). As for the Notch1-4 paralogues, an exponential pattern of length was observed, with N4 being the shortest, followed by Notch3, Notch2 and finally Notch1 being the longest.</ns0:p></ns0:div> <ns0:div><ns0:head>Multiple sequence alignment and motifs</ns0:head><ns0:p>Multiple sequence alignment (MSA) of protein sequences from the Notch family was performed to identify highly conservative regions within all organisms, from monera to invertebrates. As known, Notch receptors consist of specific conservative domains including EGF-like domain, NRR, TMD, RAM, NLS, ANK, NOD/NODP and PEST, that could be identified in the MSA <ns0:ref type='bibr' target='#b1'>(Aster, et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b38'>Sander, et al., 2006)</ns0:ref> (Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>). Monitoring and analyzing the MSA using Jalview resulted in identifying EGF, LNR and NOD/NODP domains appearance in most kingdoms, though a variation of the number of repeats and length is noticeable <ns0:ref type='bibr' target='#b11'>(Gordon, et al., 2008)</ns0:ref>. Even bacteria Notch-like proteins share some of the characteristic motifs of those regions in their receptors; therefore, we can speculate that lateral gene transfer might have occurred via bacterial transfers <ns0:ref type='bibr' target='#b8'>(Gazave, et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Ponting, et al., 1999)</ns0:ref>. Two other members of the Notch pathway, including Fringe <ns0:ref type='bibr' target='#b14'>(Irvine and Wieschaus, 1994)</ns0:ref> and Strawberry Notch (Sno) <ns0:ref type='bibr' target='#b8'>(Gazave, et al., 2009)</ns0:ref> show the same scenario, in which the eventuality of lateral gene transfer cannot be excluded.</ns0:p><ns0:p>EGFs repeats are crucial components of the Notch signaling. Thus, it was expected to identify them as the most conserved site. EGF's 6 Cysteine residues, being the key element of the domain, are responsible for the formation of disulfide bonds and influence the native 3D structure of the Notch members (Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). The importance of these key residues is also apparent in the pathological phenotype. In the case of a Notch3 mutation, if any of the 6 Cys is mutated into another amino acid, it leads to a rare neurodegenerative syndrome, called CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy) <ns0:ref type='bibr' target='#b31'>(Papakonstantinou, et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b45'>Vlachakis, et al., 2014)</ns0:ref>. It is worth mentioning that the EGFdomain is also characterized by the conserved Glycine residues, as shown by the current multiple alignments (Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). Furthermore, the highly conserved EGFs, led us to suggest that two new motifs A and B are crucial for the Notch family evolution. The CXNGGXC motif (Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>/ Motif A) consisting of two highly conserved Gly residues and a second motif (Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref> / Motif B), CXCXXG[FY]XG, characterized by a conserved Cys and Gly and a non-polar aromatic amino acid (F/Y) among two conserved Gly (Figure <ns0:ref type='figure' target='#fig_2'>3</ns0:ref>). These motifs could possibly be &#917;GF-domain's precursors, since they are repeated with a different number of repetitions depending on the respective species, where the number of repetitions is directly related to the complexity of the organism <ns0:ref type='bibr' target='#b19'>(Kovall, et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Additionally, a segment of the LNR-domain appears to be highly conserved in organisms from all kingdoms and the characteristic motif of the LNR-repeat was also found in bacteria. Motif C could be considered as the LNRs' single unit of short peptide that has been repeated via internal tandem duplications through evolution in eukaryotes <ns0:ref type='bibr' target='#b2'>(Bjorklund, et al., 2006)</ns0:ref> (Figure <ns0:ref type='figure' target='#fig_3'>4</ns0:ref> / Motif C). NOD/NODP domain seems to have maintained throughout the evolutionary history of all eukaryotes and it contains two conserved motifs (Motifs D and E). Remarkably, the conserved motif D has been identified also in bacteria's Notch-like proteins, which are significantly smaller than the homolog proteins of the other species.</ns0:p></ns0:div> <ns0:div><ns0:head>Structural characterization of Notch Receptor</ns0:head><ns0:p>Despite the progress achieved recently, in the structure determination of limited fragments of the Notch receptor, its quaternary 3D structure remains undetermined (Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). Notch receptor 3D models are based on the structural features of the long NECD region, since it contains many calcium binding EGF-like domains (Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). The NECD differs between species. Drosophila and mammalian Notch receptors are much larger than their counterparts from other invertabrate species like Caenorhabditis elegans, although each invariably maintains the same molecular architecture <ns0:ref type='bibr' target='#b18'>(Kopan and Ilagan, 2009)</ns0:ref>. On the other hand, in monera like bacteria, and in protists, the NECD size is much shorter and compact (Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>). Based on previous studies it has been observed that the NECD region of the Notch receptor is expected to have a rigid near-linear 3D structure, however potential sites of flexibility may occur at the 3D structure of the EGF domain which is less conserved between species / kingdoms. (Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref> and 3) <ns0:ref type='bibr' target='#b18'>(Kopan and Ilagan, 2009;</ns0:ref><ns0:ref type='bibr' target='#b26'>Morgan, et al., 1999)</ns0:ref>.</ns0:p><ns0:p>At the N-terminal end, for most species, the Notch receptor contains 36 EGF-like domains, a region of containing calcium-binding sites. Next to the EGF region there are three LNR and a hydrophobic region which has been shown to mediate heterodimerization (HD) (Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). Together, the LNR repeats and the heterodimerization domain form the NRR, adjacent to the cell membrane (Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). This region prevents ligand-independent activation of the Notch receptor by concealing and protecting from metalloproteases. When Notch activation is achieved via ligand binding to repeats within the EGF-like domain, then two sequential proteolytic events termed S2 and S3 cleavages are induced <ns0:ref type='bibr' target='#b37'>(Sanchez-Irizarry, et al., 2004)</ns0:ref>. The S3 cleavage site lies within the transmembrane segment and is cleaved by the &#947;-secretase complex to liberate NICD. NICD comprises one RAM domain, seven ANK, one TAD and one PEST domain (Figure <ns0:ref type='figure' target='#fig_4'>5</ns0:ref>). Both the RAM domain and ANK repeats have been identified as regions involved in the interaction with CSL transcription factors <ns0:ref type='bibr' target='#b3'>(Chillakuri, et al., 2012)</ns0:ref>. The TAD region is found in Notch-1 and -2 but not in -3 and -4 in mammals <ns0:ref type='bibr' target='#b3'>(Chillakuri, et al., 2012)</ns0:ref>. The C-terminal PEST domain is involved in NICD degradation by proteolysis.</ns0:p><ns0:p>Recently, they have been determined the complete 3D structures of the extracellular domain (ECD) of the Drosophila Notch receptor and the human Notch1 receptor, by using single particle electron microscopy and antibody labelling <ns0:ref type='bibr' target='#b3'>(Chillakuri, et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b17'>Kelly, et al., 2010)</ns0:ref>. Since the inherent resolution of the method is poorly corresponding to the small size of the domains involved, these are significant challenging experiments.</ns0:p></ns0:div> <ns0:div><ns0:head>Ca +2 signaling activation and digital emulation</ns0:head><ns0:p>Notch family proteins, being capable of recognizing and binding to the neighbor cell's ligand, achieve selective cell-cell adhesion initiated by protein-protein interaction. They comprise an extracellular region, composed of EGF-repeats, LNR repeats and NOD/NODP. EGF-repeats act as transmitters that promote the signal to the LNRs and along those extracellular regions, intracellularly to the ANK repeats. In this process, calcium plays a crucial role and is required for the biological activation of transmembrane proteins that contain EGF and LNR repeats. More precisely, Ca +2 binds to cbEGFs, a large subgroup of EGFs. CbEGFs have essentially similar structure as EGFs, but they are so conserved, that they cluster together in EGF dendrograms <ns0:ref type='bibr' target='#b42'>(Stenflo, et al., 2000)</ns0:ref>. The central role of Ca +2 has been demonstrated through crystallization experiments, where Ca +2 seems to stabilize the N-terminal of tandem EGFs, establishing a stable surface for the protein-protein interaction <ns0:ref type='bibr' target='#b5'>(Downing, et al., 1996)</ns0:ref>. The Ca +2 binding coordination is related to cbEGFs' amino acid modifications, including hydroxylation and glycosylation of an asparagine or aspartic acid residue. Moreover, when a Ca +2 is binding, the extracellular structure is locally changing, making it versatile and difficult to model. Ca +2 role is also, observed in the LNR repeats, as each cysteine-rich Lin-12/Notch repeat binds one calcium ion to achieve the correct folding and maintenance of its 3-D structure. Mutations in calcium binding sites lead to dysregulated pathways, introducings pathogenic phenotypes. A substantial case of defective Ca +2 binding is the Marfan syndrome and Hemophilia B.</ns0:p><ns0:p>The extracellular region of a Notch family precursor is a mix of calcium-loaded and calcium-free sites, with each EGF repeat forming a beta-sheet and the LNR repeats forming loops. Simulating each beta-sheet as a hoop, the whole EGF region could be presented as a strongly-connected chain. Furthermore, emulating the chain into a digital format, each hoop, depending on its Ca+2 binding or Ca+2 binding disability, is emulated by a sequence of 1 and 0, respectively. In that way, simulating each Lin12/Notch repeat as a gear (are more specialized forms than EGF repeats), the complex of LNRs could be emulated by a compact gear mechanism, which produces a more specialized signal based on the calcium interactions (Figure <ns0:ref type='figure' target='#fig_5'>6</ns0:ref>). Combining all this mechanism together we can easily understand its specialization in producing different unique codes (signals) to the cell in order to start different biological tasks. This proposed emulation could represent a way to present if the signal is going to be promoted from the EGF repeats to the LNRs etc. and finally to the nucleus, completing a the signaling pathway. In case of mutation in these residues due to defective Ca+2 binding, the corresponding digit changes and an error in the code occurs, indicating signal transduction failure associated with the possibility of a functional abnormality (Figure <ns0:ref type='figure' target='#fig_5'>6</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Notch family protein clusters</ns0:head><ns0:p>Notch family protein clusters were estimated using a phylogenetic analysis (Figure <ns0:ref type='figure' target='#fig_6'>7</ns0:ref>). A phylogenetic tree could provide clustering information for the Notch family members, so that representative consensus sequences could be properly defined. Having such a large dataset with proteins of variant sequence length, we initiated the construction of an unrooted tree. Phylogenetic analysis revealed a distinct separation of Notch family members into eight monophyletic branches, the Notch bacteria cluster, the Notch plants cluster, the Notch protist cluster, the Notch invertebrates cluster, as well as the Notch1, Notch2, Notch3 and Notch4 clusters. Each cluster's content was carefully examined and reviewed (proteins, classes and kingdoms it contained), eventually leading to clustering the data into the eight groups (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Moreover, it was clear that the resulting clusters of the phylogenetic tree were formed by related protein members and not by classes/organisms, even if we included the same species in different clusters for the Notch1 to 4 homologs. Furthermore, considering the conservation and similarity of the characteristic domains presented in each Notch cluster, it was verified that the Notch family members classification was accurate because they belong in the same phylum / kingdom (Figure <ns0:ref type='figure' target='#fig_6'>7</ns0:ref>). In this study, using protein sequences from representatives of all kingdoms, it was necessary to observe the agreement of the emerging phylogenetic tree with the tree of life. Thus, a bibliographic examination was carried out, which confirmed the correct and proper positioning and clustering of organisms according to evolution, by examining the position of the invertebrates, such as cnidarians, flatworms, nematodes, arthropods, chordates, as well as the phylogenetic position of the protists and plants, in the tree of life <ns0:ref type='bibr' target='#b16'>(Keeling, 2019)</ns0:ref> <ns0:ref type='table' target='#tab_1'>(Table 2)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Notch family evolution</ns0:head><ns0:p>In this study, we identified novel Notch protein clusters from various kingdoms extending from bacteria to chordates and conducted a comprehensive phylogenetic analysis in order to confirm and expand the evolutionary history of the Notch family (Figure <ns0:ref type='figure' target='#fig_7'>8</ns0:ref>). Notch family evolution was examined in a more specialized phylogenetic analysis with 100 bootstrap replicates using representative consensus protein sequences of each identified cluster from the previous phylogenetic analysis. Each consensus sequence, with its amino acids corresponding to the most frequently encountered, fully represented the organisms and the members of the NOTCH family, accordingly. Hence, the specialized phylogenetic tree contained meaningful information, as needed for further evolutionary research <ns0:ref type='bibr' target='#b29'>(Papageorgiou, et al., 2018)</ns0:ref>. Based on results, both constructed phylogenetic trees of the Notch family members were found to share similar topology.</ns0:p><ns0:p>Due to the small number of sequences which were available the past decade, all previously performed evolutionary studies have a significantly smaller dataset. In this study, a more comprehensive phylogenetic analysis was achieved and it was important to compare the evolutionary scenarios presented here with the previous ones. The evolutionary distances between Notch2 and Notch3 groups have been found shorter than Notch, Notch1 and Notch4. This finding is correspondingly consistent with the evolutionary analysis performed by Gazave, E., et al. <ns0:ref type='bibr' target='#b8'>(Gazave, et al., 2009)</ns0:ref>. On the other hand, Notch4 group found to be evolutionary older than Notch1, Notch2 and Notch 3, and with this result being consistent with the evolutionary study performed by <ns0:ref type='bibr' target='#b44'>Theodosiou, A., et al. (Theodosiou, et al., 2009)</ns0:ref>. However, in this study we have introduced and present for the first time, the plant, protist and bacteria groups which are in the early Notch evolution (Figure <ns0:ref type='figure' target='#fig_7'>8</ns0:ref>). Another remarkable observation is that Notch protein groups are positioned before Notch 1-4 groups, indicating the origin of the Notch family. In consideration of the tree of life, bacteria, protists and plants, appeared before the rest of the eukaryotes and therefore their position in the constructed tree is validated. Bacteria belongs to prokaryotes, while protists and plants are eukaryotic organisms and thus, much more complex than prokaryotes. The fact that the latter two are eukaryotes, combined with the absence of evidence for the existence of Notch proteins in Archaea, partially modifies the obtained results, compared to the current evolutionary theories. In addition, bearing in mind the limited number of representatives given in the kingdom of plants, this slight divergence in the present evolutionary study of the Notch family is justified. However, future enrichment of the dataset, with additional Notch family protein sequences from other species, could provide an improved and even more distinct phylogenetic analysis. Lastly, in the phylogenetic tree, Notch paralogs of Caenorhabditis elegans appear distinctly before Notch1-4, as expected, given the fact of an independent genomic duplication, leading to two Notch genes <ns0:ref type='bibr' target='#b44'>(Theodosiou, et al., 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Central nervous system (CNS) has evolved in early metazoans (such as marine organisms) as a simple neural network. This precursor system, necessary for cells' electrical signaling and cell-tocell interaction, became more complicated through evolution <ns0:ref type='bibr' target='#b0'>(Akanuma, et al., 2002)</ns0:ref>. Notch family members form a key component of an evolutionarily conserved signaling mechanism, involved in the regulation of the CNS. Indeed, Notch signaling takes part in Neural Stem Cell (NSC) proliferation, survival, self-renewal, differentiation, apoptosis and cell fate choices <ns0:ref type='bibr' target='#b8'>(Gazave, et al., 2009)</ns0:ref>. In the CNS, Notch members are present during the entire lifetime, from embryonic stages to adult nervous system, controlling neurogenesis, axons' and dendrites' growth and CNS plasticity <ns0:ref type='bibr' target='#b36'>(Presente, et al., 2001)</ns0:ref>.</ns0:p><ns0:p>Several pathogenic mechanisms have been observed from different mutation cases in Notch receptor protein domains. Recent studies suggest that Notch receptors play correspondingly a crucial role in neurodegenerative disorders, including Alzheimer's disease, Down syndrome and CADASIL <ns0:ref type='bibr' target='#b33'>(Polychronidou, et al., 2015)</ns0:ref>. Notch family members are estimated that they existed a billion years before. Previous studies have shown that Notch signaling in Metazoan is a subsequent evolutionary result, as it uses ancient protein domains and mechanisms. Thus, it was assumed that this pathway may exist in Urmetazoa or even in more inferior organisms, such as bacteria which also have Notch-like proteins <ns0:ref type='bibr' target='#b8'>(Gazave, et al., 2009)</ns0:ref>. Therefore, new insights have been extracted from the present in silico study, where it could form the basis for detecting unrelated functions in this receptor family. So far, the evolutionary history of the Notch family seems to be directly linked to the tree of life. In the same direction, the complexity of these proteins could be proportional to the complexity of the species and the nervous system, as well. Last but not least, bacteria's undifferentiated Notch-like proteins seem to have been evolved and differentiated in all other types of paralogues Notch's as the complexity of the species increases. Similar evo-devo scenarios are claimed generally in proteins families, which are important for survival and evolution, such as the example of the nuclear receptor family <ns0:ref type='bibr' target='#b24'>(Mitsis, et al., 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Despite the enormous scientific interest in Notch family proteins, current knowledge of their involvement in biological pathways and their function is still quite limited. Certainly, this knowledge could be significantly enriched with the potential determination of the Notch receptor 3D structure. Considering the cell-fate-determination and the cell communication, enforced by these proteins as well as the proteolytic procedure that they undergo in the signaling pathway, Notch receptors could be used as promising therapeutic targets for several diseases, including Cancer. In the present study, a series of the most highly conserved motifs that have arisen through evolution is presented. These motifs could be used as innovative pharmacological targets, through the development of new technologies concerning the numerous critical pathways, affected by this protein family. All in all, useful beneficial insights are provided, concerning the Notch family's evolution. A comprehensive evolutionary analysis that confirms the existence of several kingdoms among Notch family members is provided. Notch genes were duplicated several times during evolution, leading to four genes in chordates, including Homo sapiens. If we accept that more than the half of our body is not human, then the most logical scenario for the ancient origin of Notch are the bacteria.</ns0:p></ns0:div> <ns0:div><ns0:head>Figures legend</ns0:head><ns0:p>Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref>: Diseases which are caused by NOTCH1-4 mutations. The research was done using the DisGeNET database (www.disgenet.org).</ns0:p><ns0:p>Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>: Categorization of the NOTCH family members using identified clusters, per kingdom and phylum. <ns0:ref type='bibr' target='#b23'>(Mitchell, et al., 2015)</ns0:ref> results for all identified kingdoms. Notch family receptors are represented with major domains annotated, including Notch extracellular domain (NECD), Notch intracellular domain (NICD), transmembrane domain (TM), Epidermal growth factor (EGF), Cysteine-rich LNR repeats (LNR), Notch domain present in many Notch proteins (NOD/NODP), RBPJ&#954;-associated molecule domain (RAM), Nuclear localization sequences (NLS), Ankyrin repeat domain (ANK), and the Domain rich in proline, glutamine, serine and threonine residues (PEST). Protein domains marked with (*) represent domains that have only been observed in the MSA, and those marked with (?) represent domains for which no other information is available. Eight distinct monophyletic branches are visible. The phylogenetic trees confidently separate the Notch1 cluster (branch colored Pink), the Notch2 cluster (branch colored orange), the Notch 3 cluster (branch colored red), the Notch 4 cluster (branch colored yellow), the Notch Bacteria cluster (branch colored light blue), the Notch Plants cluster (branch colored green), the Notch Protist cluster (branch colored dark blue) and the Notch invertebrates cluster (branch colored gray).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49209:1:1:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed The tree was constructed with UPGMA method in MATLAB for 100 bootstrap replicates and was visualized in MEGA. Manuscript to be reviewed</ns0:p><ns0:formula xml:id='formula_0'>&#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; - - - - - - &#10003; &#10003; NOTCH 40 8 5 1 - - - -&#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; - - NOTCH1 - - - -&#10003; &#10003; -&#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; - - &#10003; NOTCH2 1 - - - -&#10003; &#10003; &#10003; &#10003; &#10003; &#10003; &#10003; - &#10003; &#10003; &#10003; - - NOTCH3 - - - - -&#10003; -&#10003; &#10003; &#10003; - - &#10003; - &#10003; - -<ns0:label>NOTCH4</ns0:label></ns0:formula><ns0:note type='other'>Figure 2</ns0:note><ns0:p>Architecture of Notch family receptors based on the Interpro Database <ns0:ref type='bibr' target='#b23'>(Mitchell, et al., 2015)</ns0:ref> results for all identified kingdoms.</ns0:p><ns0:p>Architecture of Notch family receptors based on the Interpro Database <ns0:ref type='bibr' target='#b23'>(Mitchell, et al., 2015)</ns0:ref> results for all identified kingdoms. Notch family receptors are represented with major </ns0:p><ns0:note type='other'>Figure 8</ns0:note><ns0:p>The evolutionary history of Notch family members in a specialized phylogenetic tree.</ns0:p><ns0:p>The tree was constructed with UPGMA method in MATLAB for 100 bootstrap replicates and was visualized in MEGA.</ns0:p><ns0:p>The evolutionary history of Notch family members in a specialized phylogenetic tree. The tree was constructed with UPGMA method in MATLAB for 100 bootstrap replicates and was visualized in MEGA.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Statistical analysis of the Notch family members based on the sequence annotation.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: Architecture of Notch family receptors based on the Interpro Database<ns0:ref type='bibr' target='#b23'>(Mitchell, et al., 2015)</ns0:ref> results for all identified kingdoms. Notch family receptors are represented with major domains annotated, including Notch extracellular domain (NECD), Notch intracellular domain (NICD), transmembrane domain (TM), Epidermal growth factor (EGF), Cysteine-rich LNR repeats (LNR), Notch domain present in many Notch proteins (NOD/NODP), RBPJ&#954;-associated molecule domain (RAM), Nuclear localization sequences (NLS), Ankyrin repeat domain (ANK), and the Domain rich in proline, glutamine, serine and threonine residues (PEST). Protein domains marked with (*) represent domains that have only been observed in the MSA, and those marked with (?) represent domains for which no other information is available.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Highly conserved Notch family motifs based on the 603 Notch sequences MSA results. Motif-A and Motif-B have been highlighted (colored red) in the EGF-domain, based on the Notch consensus sequence from the MSA. The characteristic EGF-Cys-repeats are illustrated in with large red boxes.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: Highly conserved Notch family regions based on the 603 Notch sequences MSA results. (A) Conserved region of the LNR domain (B). Conserved regions of the NOD/NODP domain.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5: Structural architecture of Notch family receptors based on the available fragments from the Protein Data Bank (PDB: 5FM9, 3ETO, 3V79).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 6 :</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6: Structural visualization of the ENCD of a Notch family receptor and its digital format.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 7 :</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7:The unrooted phylogenetic tree of Notch family 603 members. The tree was constructed utilizing the UPGMA method in MATLAB and visualized using the TreeExplorer tool of MEGA. Eight distinct monophyletic branches are visible. The phylogenetic trees confidently separate the Notch1 cluster (branch colored Pink), the Notch2 cluster (branch colored orange), the Notch 3 cluster (branch colored red), the Notch 4 cluster (branch colored yellow), the Notch Bacteria cluster (branch colored light blue), the Notch Plants cluster (branch colored green), the Notch Protist cluster (branch colored dark blue) and the Notch invertebrates cluster (branch colored gray).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 8 :</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8: The evolutionary history of Notch family members in a specialized phylogenetic tree.The tree was constructed with UPGMA method in MATLAB for 100 bootstrap replicates and was visualized in MEGA.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>domains annotated, including Notch extracellular domain (NECD), Notch intracellular domain (NICD), transmembrane domain (TM), Epidermal growth factor (EGF), Cysteine-rich LNR repeats (LNR), Notch domain present in many Notch proteins (NOD/NODP), RBPJ&#954;-associated molecule domain (RAM), Nuclear localization sequences (NLS), Ankyrin repeat domain (ANK), and the Domain rich in proline, glutamine, serine and threonine residues (PEST). Protein domains marked with (*) represent domains that have only been observed in the MSA, and those marked with (?) represent domains for which no other information is available.</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,42.52,204.37,525.00,306.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,70.87,525.00,406.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,295.87,525.00,350.25' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='23,42.52,270.37,525.00,364.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='26,42.52,204.37,525.00,168.00' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Diseases which are caused by NOTCH1-4 mutations. The research was done using the DisGeNET database (www.disgenet.org).</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell cols='2'>NOTCH Proteins</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>NOTCH1</ns0:cell><ns0:cell>NOTCH2</ns0:cell><ns0:cell>NOTCH3</ns0:cell><ns0:cell>NOTCH4</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>o T-cell acute</ns0:cell><ns0:cell>o Hajdu-Cheney</ns0:cell><ns0:cell>o CADASIL</ns0:cell><ns0:cell>Unclarified</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>lymphoblastic</ns0:cell><ns0:cell>Syndrome</ns0:cell><ns0:cell>o Infantile Myofibromatosis</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>leukemia</ns0:cell><ns0:cell>o Alagille Syndrome</ns0:cell><ns0:cell>o Early-onset arteriopathy with</ns0:cell></ns0:row><ns0:row><ns0:cell>Diseases</ns0:cell><ns0:cell>o Adams-Oliver</ns0:cell><ns0:cell>o Cancer</ns0:cell><ns0:cell>cavitating leukodystrophy</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Syndrome</ns0:cell><ns0:cell /><ns0:cell>o Lateral meningocele</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>o Aortic Valve</ns0:cell><ns0:cell /><ns0:cell>Syndrome</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Disease</ns0:cell><ns0:cell /><ns0:cell>o Cancer</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>o Cancer</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Categorization of the NOTCH family members using identified clusters, per kingdom and phylum.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Eukaryotes</ns0:cell></ns0:row></ns0:table></ns0:figure> </ns0:body> "

" Athens, 28th August 2020 Dear Prof Ubersky & PeerJ editorial team, Please find attached our revised manuscript entitled: “An updated evolutionary study of the Notch family, reveals a new ancient origin and novel invariable motifs as potential pharmacological targets”. All of the respected reviewer comments have been addressed accordingly. A detailed point-by-point reply follows. I would like to take this opportunity to thank you for considering our manuscript, for your exceptional editorial support and for guiding us to revise our work, through the constructive comments of your reviewers’ team. Please do not hesitate to contact me should you require any further information or action from my side. Sincerely, Dimitrios Vlachakis, PhD MPhil, PhD, FBGS, FGS (UK), MRSC, MRSB, MRSMed Asst. Professor in Genetics & Computational Biology DarkDNA Group, Genetics Laboratory, Biotechnology Department School of Applied Biology & Biotechnology, Agr. University of Athens zoom: 427-048-4918, skype: naimenalla, url: http://darkdna.gr Researcher C on Precision Medicine & Genetics University Research Institute of Maternal and Child Health and Precision Medicine Medical School, National & Kapodistrian University of Athens  https://universityresearchinstitute.gr/en/institute/ Affiliated Researcher on Molecular Endocrinology Clinical, Experimental Surgery & Translational Research Biomedical Research Foundation of the Academy of Athens http://www.bioacademy.gr/faculty-details/HMiO/dimitrios Adjunct Investigator at the Qatar Genome Programme (QGP) Qatar Foundation & Pediatrics Dept. Weill Cornell Medical College Sidra Medicine, Sidra Medical & Research Center, Doha, Qatar _________________________________________________________________ “...not all those who wander are lost...”   J.R.R.Tolkien A Point-by-point respond follows this page: Point-by-point respond to the Editor: REVIEWERS’ S EVALUATION • Reviewer 1 – Yong Wang Basic reporting In this manuscript, the authors searched and analyzed Notch protein sequences from organisms of all kingdoms. As a result, a large number of Notch proteins are identified in bacteria, protists, fungi, plants and invertebrates in addition to the previously well characterized Notch1~4 proteins. Based on evolutionary analysis to 603 unique Notch protein sequences, they inferred existence of ancestral Notch sequences in bacteria, protists and plants. Furthermore, they successfully identified characteristic motifs A to E in EGF domain. These motifs have conserved glycine(s) in addition to conserved cysteines, which can be potential pharmacological targets for future studies. The results present in this paper enrich our knowledge on distribution, evolution and structure of Notch proteins in various organisms. I would recommend publication of it in PeerJ after the following minor revisions are made.  Thank you very much for your constructive comments and for considering our article. Lines 39 to 48: The abstract has too many words as ‘introduction’. Please reduce them and add words to describe results and significance of your work.  We have reduced the words in Abstract and more information related to the results and the significance of this study has been inserted. Line 109: Please specify the query item. Is it the word “Notch”?  Yes, that is true. This information has been included in Line 109. Line 161: The number is 613 here. It is 603 at other places (line 116 etc). Also, line 49 says “603 different organisms”. But, line 116 says “603 unique, non-duplicate protein sequences”. Please check.  The final number of the dataset is 603 samples. We have changed the number 613 to 603 in line 161. Lines 223 to 236: (Figure 5) should be (Figure 6), I guess.  Correct. This has been corrected in the revised manuscript. Line 278: (Figure 6) should be (Figure 7), I guess.  Correct. This has been corrected in the revised manuscript. Line 452: The reference is a duplicate of line 450.  Yes, that is true. The second reference has been removed. Table 2: It is better to replace the tick with number of Notch genes found in each phylum/class among the 603 unique sequences. Also in Table 2 (and in the whole manuscript), it is better to replace “Multicellular” with “Invertebrates”, because multicellular organisms include chordates.  That is a fair comment. All information requested by the respected reviewer has been added in the revised manuscript. Following your comment, we have replaced the term “Multicellular” with “Invertebrates” in whole manuscript, figures and Tables. Also, in Table 2 we have inserted the sample numbers in each phylum/class among the 603 unique sequences. Thank you. Figure 3: Please specify how many sequences are used to generate this graph. Is it the same with Figure 4 (603 Notch sequences)?  Yes, that is true. The Figure 3 title has been changed. Figure 7: It is better to put the color block beneath tree branches but not on them, so that the tree branches can be visualized. Besides, please indicate how many sequences were used for construction of this tree.  This has been addressed and more info about the tree has been added throughout the revised manuscript • Reviewer 2 Basic reporting In this paper, the authors aims to provide an updated evolutionary history of the Notch family members. This question is very interesting as the last paper dealing with it may be a bit old from now and new genomic data coming from various species may help to better understand and dissect the evolution of this family. However, the paper in this state does not allow to answer this question, due to several flaws in the methodology used (see below). Also the evolutionary considerations are not accurate (ex : l63 'indicating the progressive differentiation of Notch family'; l163 'simpler organisms', l349 'inferior organisms', L335 'the CNS has evolved in early metazoans, referring to ascidians which are far from early metazoans etc etc; ;phylogeny 'map' (abstract), separation of the animals into 'multicellular vs chordates' in the table 2 etc etc... .  In the present work we have studied the updated evolutionary scenario of Notch family members towards identifying new conserved motifs necessary for NOTCH signaling pathways. Several similar studies have been done and descripted in the past. Our novel study, provides additional information based on new identified members in NOTCH family. Our findings come to complement the prevailing theories about the evolution NOTCH family. Based on Reviewer 1 and 3 comments, many sentences have been changed in order to describe more efficiently the results of this study. I am also not sure the literature references provided are relevant and accurate, for ex in the introduction the authors cites 'Durieux et al 2019' and' Fairclough et al 2013' to describe the length of the notch proteins in bacteria and animals respectively. I did not see any mention of Notch protein in those papers and I do not think that notch proteins do exist in bacteria (maybe a specific domain of the notch protein, but not the protein itself).  Bacterial Notch protein sequences have been deposited and annotated in the NCBI or GenBank with the keywords “neurogenic locus notch” and “neurogenic locus notch like protein precursor”, the last four years. That’s why there are no available publications descripting in details the Bacteria NOTCH protein, although several studies hypothesize its existence. The two publications we have mentioned about bacterial data are some of the publications that are referred to the NCBI and described the genome exploration of some of the bacteria mentioned. In addition, other publications that examine the scenario of bacterial NOTCH proteins such as the publication of Gazave, E., et al. (2009) are provided. Finally, the link with pharmacological targets is not clear to me.  Based on the evolutionary analysis of the 603 unique Notch protein sequences we have provided, five new conserved motifs have been identified. As reviewer 1 says, “These motifs have conserved glycine(s) in addition to conserved cysteines, which can be potential pharmacological targets for future studies” Experimental design The most important issue of the paper is the selection of the proteins to be included in the analyses. The way to select/identify proteins in order to do comparative genomics is to blast a protein query (here probably human) against databases of interest. The key word search is not an alternative way of doing it, as it is biaised (it depends on the way the protein has been submitted to the database) and highly incomplete.  Yes, that is true. BLAST search is the easiest way to extract homolog protein sequence from biological databases. In the other hand, NOTCH family contains four related proteins from where in the NOTCH cluster homo sapiens doesn’t exist. That’s why we extracted the biological dataset using keyword/semantic search. As we described in the manuscript all the unrelative data have been removed from the primary dataset with several filtering techniques. 2nd, the authors needs to define what is the minimal domain composition of a protein to be considered as notch. The presence of one or few domains, not specific to a notch proteins is probably not sufficient. Short 'notch' sequences from bacteria such as Legionella X, have even no domain detected (search by NCBI CD search tool). How the authors deal with such different sizes of proteins ? how the alignment for the tree was done ? What about the missing data ? Did the author select just specific domains ? what about the variable number of EGF ? The notch protein phylogeny is quite hard due to all this specificities. Surprisingly, the authors used a very old method, not used any more to do their phylogenetic reconstructions (UPGMA), with no or few method of statistical support.  Protein sequences between NOTCH family members have a large variation in the length of the sequences. That’s why in the present study we have followed the descripted pipeline in methods. First of all, we have aligned the protein sequences using progressive methods and a guide tree. This technique allows us to identify the smallest possible common areas between the different proteins. The smallest possible area was defined based on the candidate conserved motifs where they are described. The consensus protein sequences of each major cluster were then determined based on the results of the unrooted tree. Thus, we keep all the representative characteristics of the protein regions of each cluster. Finally, we have examined the evolutionary history of NOTCH members using the UPGMA. This technique has been used for the last decades and is very popular and accepted by the entire scientific community. Validity of the findings Based on the flaws mentioned above, the results provided are not conclusive. The tree (figure 7) is non resolved. In the figure8, what are the number on the branches ? Why the authors made a separation between multicellular and chordates ? are chordates unicellular ?  Even though we see the reviewer’s point it is crucial to clarify here the main message of this article, the novelty it offers and the purpose it serves. As we mentioned in the manuscript, the unrooted tree in figure 7 was calculated towards identifying the several clusters of the NOTCH family. In the figure 8 is provided the final phylogenetic tree of the 100 bootstrap replicates (number on the branches) using the consensus protein sequences of each cluster. We have separated some of the multicellular organisms from the chordates based on the findings of the tree descripted in figure 7. Reviewer 1 suggest to rename the “multicellular” cluster to “Invertebrates” cluster to avoid misinterpretation by readers. All in all, we believe that our methodology provides enough proof to support our NOTCH evolutionary story, in a easy to follow but also scientifically sound pipeline. • Reviewer 3 – Sunil Shah Basic reporting Authors has made excellent efforts in compiling the MS with detailed, clearly, consistently communicated a rational design followed by their outcomes, interpretation and significance of the work.  Thank you very much for your excellent comments. Experimental design No changes required in the compiled version Validity of the findings Author has made great effort in contributing the new novelty in the present MS. My detail comments are in the author section.  Thank you very much for your excellent comments. Comments for the author The topic seems excellent addition in the current literature however, before accepting for the publication, it should advisable to refine further by making necessary minor corrections as suggested in the bottom which should provide more clarity to readers.  Thank you very much for your excellent comments. Comments on Manuscript Couple of Comments/suggestion should be incorporate to improve English, Formatting and Uniformity. 1. Title-Please include the before the Notch family  Yes, that is true. Title has been changed. 2. Affiliation 5, cedex should have ‘C’ in capital letter  Yes, that is true. Line has been changed. 3. Line 50, abstract, please add ‘s’ after human  Yes, that is true. Line has been changed. 4. Line 73, Introduction: replace ‘to’ with ‘in’  Yes, that is true. Line has been changed. 5. Line 76, provide symbol for approximation and replace the word amino acids with ‘aa’ for uniformity  Yes, that is true. Line has been changed. 6. Line 97, insert comma after TAD domain  Yes, that is true. Line has been changed. 7. Line 97, remove the word in before the length  Yes, that is true. Line has been changed. 8. Line 98, remove the before vascular…..  Yes, that is true. Line has been changed. 9. Line 101, add ‘the’ after involved in, so sentence looks like The pathway is involved in the….  Yes, that is true. Line has been changed. 10. Line 102, Insert comma, after learning  Yes, that is true. Line has been changed. 11. Line 113, insert ‘s’ after score; 114, replace which with that; Line 115, symbol will come before 95%  Yes, that is true. Lines have been changed. 12. Line 137, insert ‘a’ before threshold  Yes, that is true. Line has been changed. 13. Line 158, insert ‘to’ before as noisy….  Yes, that is true. Line has been changed. 14. Line 136, in the Figure 1, author should provide axis information along with as an insert total number of protein  Yes, that is true. Figure 1 has been changed. 15. Line 166, insert amino acids or aa after 4835  Yes, that is true. Line has been changed. 16. Line 185, replace a with the before most conserved site…..  Yes, that is true. Line has been changed. 17. Line 193, insert ‘s’ after alignment…  Yes, that is true. Line has been changed. 18. Line 196, delete ‘a’ before non-polar  Yes, that is true. Line has been changed. 19. Line 198, add they after since….are repeated….  Yes, that is true. Line has been changed. 20. Line 205, insert ‘the’ before evolutionary history……  Yes, that is true. Line has been changed. 21. Line 206, replace the It’s remarkable that with Remarkably,……  Yes, that is true. Line has been changed. 22. Line 207, insert ‘ly’ after significant  Yes, that is true. Line has been changed. 23. Line 217, Correct the spelling of Caenorabditis……  Yes, that is true. Line has been changed. 24. Line 221, insert ‘s’ after appear  Yes, that is true. Line has been changed. 25. Line 224, delete ‘of’ before containing calcium…..  Yes, that is true. Line has been changed. 26. Line 225 and 226, replace ‘s’ with ‘z’ for the word heterodimeri….  Yes, that is true. Lines have been changed. 27. Line 237, insert ‘the’ before human notch1  Yes, that is true. Line has been changed. 28. Line 243, charge on calcium should be superscript and similar mistakes should be corrected other places  Yes, that is true. Charge on calcium has been changed in whole manuscript. 29. Line 275, add with and delete to comes before the possibility,,,,,  Yes, that is true. Line has been changed. 30. Line 320, rewrite the sentence like be at the beginning of…  Yes, that is true. Sentence has been changed in “which are in the early Notch evolution” 31. Line 343, remove ‘A number of’ and start the statement with Several……., replace has with have  Yes, that is true. Line has been changed. 32. Line 347, insert ‘s’ after year… and replace showed with shown….  Yes, that is true. Line has been changed. 33. Line 374, remove ‘the’ before half of our…..  Yes, that is true. Line has been changed. 34. Line 375, remove are before the bacteria..  Yes, that is true. Line has been changed. 35. Line 450-452; 472-475; 530-533, remove the additional duplicate reference from the MS.  Yes, that is true. Second reference has been removed. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: Foam cells (FCs) play crucial roles in the process of all stages of atherosclerosis. Smooth muscle cells (SMCs) and macrophages are the major sources of FCs. This study aimed to identify the common molecular mechanism in these two types of FCs. Methods: GSE28829, GSE43292, GSE68021, and GSE54666 were included to identify the differentially expressed genes (DEGs) associated with FCs derived from SMCs and macrophages. Gene Ontology biological process (GO-BP) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed by using the DAVID database.</ns0:p><ns0:p>The co-regulated genes associated with the two origins of FCs were validated (GSE9874), and their expression in vulnerable AS plaques (GSE120521 and GSE41571) was assessed.</ns0:p><ns0:p>Results: A total of 432 genes associated with FCs derived from SMCs (SMC-FCs) and 81 genes associated with FCs derived from macrophages (M-FCs) were identified, and they were mainly involved in lipid metabolism, inflammation, cell cycle/apoptosis. Furthermore, three co-regulated genes associated with FCs were identified: GLRX, RNF13, and ABCA1. These three common genes showed an increased tendency in unstable or ruptured plaques, although in some cases, no statistically significant difference was found.</ns0:p><ns0:p>Conclusions: DEGs related to FCs derived from SMCs and macrophages have contributed to the understanding of the molecular mechanism underlying the formation of FCs and AS.</ns0:p></ns0:div> <ns0:div><ns0:head>GLRX, RNF13</ns0:head><ns0:p>, and ABCA1 might be potential targets for AS treatment.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Atherosclerosis is a complex chronic disease characterized by the thickening of the intima and the formation of atherosclerotic plaques, leading to asymmetric stenosis of the arterial The purpose of 'omics' research is to integrate and analyze a large amount of information that represents the overall biological parameters in a particular state through advanced computer technology to identify some important information related to phenotypes and/or diseases <ns0:ref type='bibr' target='#b38'>(Vernon et al. 2019)</ns0:ref>. A single biomarker may not be sufficient to represent the complex biological processes of the disease, and omics methods can be used to capture multiple variables and demonstrate the interrelationship between those variables in the disease process <ns0:ref type='bibr' target='#b38'>(Vernon et al. 2019)</ns0:ref>. In recent years, the expression profiles of atherosclerosis have been determined by omics approaches, such as microarrays and RNA-seq, and hundreds of differentially expressed genes (DEGs) involved in the development of atherosclerosis have been identified <ns0:ref type='bibr' target='#b21'>(Liu et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b36'>Tan et al. 2017)</ns0:ref>. Recently, many studies have attempted to explore the possible molecular mechanisms of macrophage-derived FCs (M-FCs) through microarray technology and bioinformatics analysis <ns0:ref type='bibr'>(H&#228;gg et al. 2008;</ns0:ref><ns0:ref type='bibr' target='#b14'>Huang et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b31'>Reschen et al. 2015)</ns0:ref>. However, few studies have systematically explored the molecular mechanism of foam cells derived from smooth muscle cells (SMC-FCs) through high-throughput methods. In this study, bioinformatics techniques were utilized to investigate the possible mechanisms of SMC-and macrophagederived FCs to identify the common molecular mechanisms of these two different sources of FCs.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods and materials</ns0:head></ns0:div> <ns0:div><ns0:head>Data resources</ns0:head><ns0:p>The GEO database (http://www.ncbi.nlm.nih.gov/geo) is an open public database established by Manuscript to be reviewed the National Center for Bioinformatics (NCBI), containing a large amount of data from microarrays, gene chips, and RNA-seq <ns0:ref type='bibr' target='#b32'>(Riksen &amp; Stienstra 2018)</ns0:ref>. To obtain the genes associated with the formation as well as the development of atherosclerosis, the dataset consisted of human atherosclerotic plaques and normal arteries or early and advanced atherosclerotic plaques that were searched in the GEO database, and two datasets (GSE28829 and GSE43292) were obtained. Then, the invitro dataset contained ox-LDL treated smooth muscle cells (SMC) or macrophages that were screened out to identify foam cell-related DEGs of the two cell types. Furthermore, because of the important role of foam cells in the development of vulnerable plaques <ns0:ref type='bibr' target='#b22'>(Liu et al. 2017)</ns0:ref>, the dataset composed of unstable vs. stable or ruptured vs. unruptured atherosclerotic plaques were screened to explore the expression of the foam cell-related genes in vulnerable plaques. The datasets of animal samples and the in-vivo human datasets of serum or plasma were excluded in this study.</ns0:p><ns0:p>The GSE28829 <ns0:ref type='bibr' target='#b8'>(D&#246;ring et al. 2012</ns0:ref>) dataset contains 13 early carotid atherosclerotic plaque samples and 16 advanced atherosclerotic plaque samples. GSE43292 <ns0:ref type='bibr' target='#b1'>(Ayari &amp; Bricca 2013)</ns0:ref> contains 32 normal carotid artery samples and 32 corresponding atherosclerotic plaque samples. GSE68021 <ns0:ref type='bibr' target='#b5'>(Dami&#225;n-Zamacona et al. 2016</ns0:ref>) contains the gene expression dataset of human vascular SMCs simulated with oxidized low-density lipoprotein (ox-LDL) for 0 h, 1 h, 5 h, and 24 h (3 technical replicates of each group). GSE54666 <ns0:ref type='bibr' target='#b31'>(Reschen et al. 2015</ns0:ref>) is a dataset of in vitro macrophage experiments including 6 samples of untreated macrophages and 6 samples of macrophage-derived FCs stimulated with ox-LDL for 48 h. GSE9874 <ns0:ref type='bibr'>(H&#228;gg et al. 2008</ns0:ref>) consists of the gene expression profiles of macrophages treated or untreated with ox-LDL from 15 healthy subjects and 15 atherosclerotic patients. GSE41571 <ns0:ref type='bibr' target='#b18'>(Lee et al. 2013)</ns0:ref> Manuscript to be reviewed macrophage-rich regions from 5 ruptured plaques and 6 stable plaques. GSE120521 <ns0:ref type='bibr' target='#b24'>(Mahmoud et al. 2019</ns0:ref>) consists of an RNA-seq profile of 4 stable and 4 unstable plaque samples. The analysis strategy is shown in Figure <ns0:ref type='figure' target='#fig_10'>1</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Identification of DEGs</ns0:head><ns0:p>The downloaded gene expression profiles and their matched platform files were loaded into R (version 3.6.1) software and converted into gene symbol expression profiles. The LIMMA package was used to identify differentially expressed genes (DEGs) between the two groups <ns0:ref type='bibr' target='#b33'>(Ritchie et al. 2015)</ns0:ref>. DEGs with adjusted p values &lt; 0.05 (adjusted by the Benjamini-Hochberg method) were considered significant. Co-DEGs were found in the overlap of different datasets as determined by an online web tool (http://jvenn.toulouse.inra.fr/app/example.html).</ns0:p><ns0:p>For the GSE68021 dataset, the DEGs of each time point (ox-LDL treated for 1 h, 5 h, 24 h) compared with negative control were identified. Furthermore, the Short Time-series Expression Miner (STEM) <ns0:ref type='bibr' target='#b9'>(Ernst &amp; Bar-Joseph 2006)</ns0:ref> program was used to analyze the temporal expression profiles of these DEGs. The significant clusters which manifested upregulated or downregulated characteristics over time were selected for the next analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Functional enrichment analyses</ns0:head><ns0:p>The purpose of gene function enrichment analysis is to determine the correlation between a group of genes and functional categories through a hypergeometric test. Gene Ontology (GO) is an international standardized gene functional classification system that covers three major Manuscript to be reviewed categories: biological process (BP), cellular component (CC), and molecular function (MF). The Kyoto Encyclopedia of Genes and Genomes (KEGG) database is the major public pathwayrelated database for helping researchers understand the advanced functions of genes in biological systems. DAVID (https://david.ncifcrf.gov/home.jsp) is a very useful bioinformatics data resource for scientists to find meaningful biological information on genes or proteins <ns0:ref type='bibr' target='#b13'>(Huang da et al. 2009</ns0:ref>). Herein, GO-BP and KEGG pathway analyses were performed using the DAVID database. The functional category with a p-value &lt; 0.05 <ns0:ref type='bibr' target='#b43'>(Xia et al. 2019</ns0:ref>) was considered significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Validation of the common FCs-related genes</ns0:head><ns0:p>For the validation of the common FCs-related genes, the expressions of these genes were extracted out from the microarray dataset and analyzed by student's t-test. For the RNA-seq dataset, the normalized gene expression matrix of FPKM (fragments per kilobase of exon model per million reads mapped) values was downloaded and transformed into TPM (transcripts per kilobase million) values for comparison <ns0:ref type='bibr' target='#b19'>(Li et al. 2010</ns0:ref>). P value &lt; 0.05 was defined as statistical significance.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Identification of DEGs associated with atherosclerosis</ns0:head><ns0:p>GSE28829 contained data on 13 early atherosclerotic plaques and 16 late atherosclerotic plaque samples. Through a LIMMA analysis, a total of 2,300 DEGs were identified in the advanced plaques, based on the gene expression in the early plaques (defined as atherosclerosis PeerJ reviewing <ns0:ref type='table'>PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:ref> Manuscript to be reviewed progression-related genes), including 1,150 upregulated genes and 1,150 downregulated genes (Figure <ns0:ref type='figure' target='#fig_12'>2A</ns0:ref>). GSE43292 contained information on 32 nonatherosclerotic plaque samples (control) and 32 atherosclerotic plaque samples, and 6,378 DEGs were obtained, including 3,066 up-and 3,312 downregulated DEGs in atherosclerosis, based on the gene expression in the control group (Figure <ns0:ref type='figure'>2B</ns0:ref>). After comparing the DEGs in the two datasets, 801 codownregulated and 849 co-upregulated genes were identified (Figure <ns0:ref type='figure'>2C, D</ns0:ref>). These coregulated DEGs were defined as atherosclerosis-related genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Identification of SMC-FCs-related genes and functional analysis</ns0:head><ns0:p>GSE68021 includes the gene expression data of human vascular SMCs treated with ox-LDL for 1 h, 5 h, and 24 h in vitro to simulate the status of SMC-FCs in atheroma plaques. In this study, a total of 3369, 4887, and 7170 DEGs were identified in 1 h, 5 h, and 24 h compared with control (Figure <ns0:ref type='figure' target='#fig_12'>3A</ns0:ref>). Moreover, these DEGs obtained from different time points were investigated their expressive trends by using STEM program. As shown in Figure <ns0:ref type='figure' target='#fig_11'>3B</ns0:ref>, 13 clusters with statistical significance were identified. And Cluster 1 (581 genes), 9 (601 genes), 12 (392 genes), 11 (413 genes), 26 (253 genes), 23 (240 genes) exhibited downregulated features over time, while Cluster 42 (630 genes), 48 (570 genes), 40 (493 genes), 29 (344 genes) exhibited upregulated trends over time. The expression of these genes with downregulated and upregulated trends were shown in Figure <ns0:ref type='figure' target='#fig_11'>3C</ns0:ref>. Then these genes were compared with AS-related genes, and a total of 432 common genes were obtained and defined as SMC-FCs-related genes (Figure <ns0:ref type='figure' target='#fig_11'>3D</ns0:ref>).</ns0:p><ns0:p>To clarify the biological functions of these SMC-FCs-related genes, the GO-BP and KEGG Manuscript to be reviewed pathway enrichment analyses were performed, and a total of 71 BP and 25 KEGG pathway functions were enriched. As shown in Figure <ns0:ref type='figure' target='#fig_11'>3E</ns0:ref> and supplementary Table <ns0:ref type='table'>S1</ns0:ref>, the enriched KEGG pathways were mainly involved in inflammation and contractive function pathways, such as Regulation of actin cytoskeleton (hsa04810), Vascular smooth muscle contraction (hsa04270), Chemokine signaling pathway (hsa04062), Leukocyte transendothelial migration (hsa04670).</ns0:p><ns0:p>Similarly, the results of GO-BP terms suggested these genes were mainly involved in the contraction function, inflammation, cell cycle/apoptosis, and substance uptake and intracellular transport, such as actin filament organization (GO:0007015), smooth muscle contraction (GO:0006939), T cell receptor signaling pathway (GO:0050852), inflammatory response (GO:0006954), apoptotic process (GO:0006915), cell cycle (GO:0007049), early endosome to late endosome transport (GO:0045022), lysosome organization (GO:0007040) (Figure <ns0:ref type='figure' target='#fig_11'>3F</ns0:ref> and supplementary Table <ns0:ref type='table'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Identification of DEGs associated with macrophage-derived foam cells (M-FCs) and their functional enrichment</ns0:head><ns0:p>GSE54666 was the gene expression profiling of macrophages stimulated with ox-LDL in vitro to simulate the status of M-FCs in atheroma plaques. A total of 484 differentially expressed genes were obtained, including 259 upregulated genes and 225 downregulated genes (Figure <ns0:ref type='figure' target='#fig_13'>4A</ns0:ref>). These DEGs and previously obtained AS-related genes were integrated, and the overlapping genes were defined as M-FC-related genes. As shown in Figure <ns0:ref type='figure' target='#fig_13'>4B</ns0:ref>, a total of 81 M-FC-related genes were identified. </ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>To explore the biological role of the 81 M-FCs-related genes, functional enrichment analysis was performed again using the DAVID database. A total of 30 GO-BP and 1 KEGG pathway were enriched. The GO-BP analysis mainly revealed lipid metabolism, foam cell differentiation, and immune response, such as negative regulation of cholesterol storage (GO:0010887), negative regulation of macrophage-derived foam cell differentiation (GO:0010745), response to low-density lipoprotein particle (GO:0055098), and negative regulation of interferon-gamma-mediated signaling pathway (GO:0060336). The enriched KEGG pathway was the PPAR signaling pathway (hsa03320) (Figure <ns0:ref type='figure' target='#fig_13'>4C</ns0:ref> and Supplementary Table <ns0:ref type='table'>S2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The common genes of SMC-FCs and M-FCs related DEGs</ns0:head><ns0:p>To identify common molecular mechanisms in the SMC-FCs and M-FCs, the common genes were screened out. As shown in Figure <ns0:ref type='figure' target='#fig_14'>5A</ns0:ref>, 15 common genes were identified: CTSD, HHEX, TNFRSF21, GLRX, EDEM2, CTSC, LAT2, SPOCD1, ABCA1, LST1, CD74, PLAUR, BRI3, EMB, and RNF13. Then, the expression trends of these 15 genes in AS tissue, SMC-FCs, and M-FCs were compared to that of the respective control. As shown in Figure <ns0:ref type='figure' target='#fig_14'>5 B-C</ns0:ref>, all of these 15 genes were upregulated in the atherosclerotic group compared to the non-atherosclerotic group or in advanced plaques compared with early plaques. However, only four genes (GLRX, ABCA1, HHEX, and RNF13) exhibited upregulated trends in ox-LDL treated SMCs (Figure <ns0:ref type='figure' target='#fig_14'>5D</ns0:ref>), while nine genes (SPOCD1, PLAUR, CTSD, RNF13, ABCA1, BRI3, GLRX, TNFRSF21, and EDEM2) were increased in ox-LDL treated macrophages (Figure <ns0:ref type='figure' target='#fig_14'>5E</ns0:ref>). Then, the co-upregulated and co- Manuscript to be reviewed downregulated genes were screened. The Venn diagrams showed the 3 (GLRX, RNF13, and ABCA1) co-upregulated genes; and no co-downregulated genes were found (Figure <ns0:ref type='figure' target='#fig_14'>5F</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Validation of the common FCs-related genes and their expression in vulnerable plaques</ns0:head><ns0:p>The expression of the 3 co-upregulated FC genes was then validated using another dataset.</ns0:p><ns0:p>GSE9874 contains the gene expression profile of peripheral blood monocyte-derived macrophages treated or untreated with ox-LDL. As shown in Figure <ns0:ref type='figure' target='#fig_15'>6A</ns0:ref>, all of the three genes were increased in the ox-LDL treated group (Foam cells group).</ns0:p><ns0:p>Previous studies have suggested that the formation and retention of the FCs could exacerbate atherosclerosis and fuel the development of the vulnerable plaques <ns0:ref type='bibr' target='#b2'>(B&#228;ck &amp; Hansson 2015;</ns0:ref><ns0:ref type='bibr' target='#b22'>Liu et al. 2017)</ns0:ref>. Therefore, the expressions of the three FCs-related genes were investigated in the vulnerable plaques. As shown in Figure <ns0:ref type='figure' target='#fig_15'>6B</ns0:ref>, only RNF13 was statistically significantly upregulated in unstable plaques, while ABCA1 and GLRX showed no statistical significance, though the means of ABCA1 were larger in unstable plaques than stable plaques. In ruptured plaques, GLRX and RNF13 were significantly upregulated compared with stable plaques, while ABCA1 showed a slight increase with no statistical significance (Figure <ns0:ref type='figure' target='#fig_15'>6C</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Previous studies have suggested the crucial role of foam cells (FCs) in the formation and development of atherosclerosis, and the molecular mechanisms of FCs have been the focus of attention in recent years <ns0:ref type='bibr' target='#b4'>(Chistiakov et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b30'>Poznyak et al. 2020)</ns0:ref>. Recently, advanced high- Manuscript to be reviewed throughput and bioinformatic technologies have been widely used for exploring the molecular mechanisms and predicting biomarkers for the diagnosis, treatment, and prognosis of the diseases <ns0:ref type='bibr' target='#b7'>(Dona et al. 2016)</ns0:ref>. Most of them focused on the mechanism of macrophages-derived FCs or atherosclerosis <ns0:ref type='bibr' target='#b1'>(Ayari &amp; Bricca 2013;</ns0:ref><ns0:ref type='bibr' target='#b31'>Reschen et al. 2015)</ns0:ref>. However, in-vivo/tissue studies could not reveal the molecular mechanism of foam cells (FCs), while in-vitro/cell studies maybe not completely simulated the molecular status of FCs in atherosclerotic plaques.</ns0:p><ns0:p>Moreover, few explored the common molecular mechanisms of the two main sources of FCs (SMCs and macrophages).</ns0:p><ns0:p>In the present study, two datasets of gene expression in AS tissues (GSE28829 and GSE43292) were analyzed, and 1650 co-regulated DEGs were identified. After screening these 1650 co-DEGs with DEGs of SMC-FCs (GSE68021) and M-FCs (GSE54666), a total of 432 SMC-FCs-related genes and 81 M-FCs-related genes were identified. To identify the common therapeutic targets of FCs, the common molecular mechanism between FCs derived from SMCs and macrophages was explored. Three co-upregulated genes were identified: GLRX, RNF13, and ABCA1. Furthermore, the expressions of these 3 genes in vulnerable AS plaques were explored, and all of them were increasingly expressed in unstable or ruptured plaques compared to their expression levels in stable plaques, although some of the increased levels were not statistically significant.</ns0:p><ns0:p>Based on murine models, it was long presumed that all FCs in human AS were derived from macrophages. However, subsequent studies demonstrated that resident cell types, especially SMCs, maybe the prominent originators of FCs <ns0:ref type='bibr' target='#b30'>(Poznyak et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wang et al. 2019b)</ns0:ref>. Manuscript to be reviewed However, macrophages are still the second most prolific sources of FCs and play crucial roles in the pathogenesis of AS, in both the early and advanced stages <ns0:ref type='bibr' target='#b28'>(Owsiany et al. 2019)</ns0:ref>. Therefore, macrophages remain the most representative model of FCs and will continue to be studied.</ns0:p><ns0:p>The process of lipid metabolism in macrophages is divided into three main stages: lipid uptake, esterification, and efflux <ns0:ref type='bibr' target='#b23'>(Maguire et al. 2019)</ns0:ref>. The lipid metabolism of macrophages in atheroma plaques gradually accelerates to be imbalanced, with excessive intracellular lipid deposition and leading to the formation of FCs <ns0:ref type='bibr' target='#b4'>(Chistiakov et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b23'>Maguire et al. 2019</ns0:ref>). In the present study, a total of 81 M-FCs-related genes were identified, and their functions were found to be mainly enriched in lipid metabolism and immune responses.</ns0:p><ns0:p>The final stage of cholesterol metabolism is the efflux of cholesterol, which is mediated by transporters that depend on multiple key transcription factors, such as PPAR and LXLR <ns0:ref type='bibr' target='#b3'>(Chistiakov et al. 2016;</ns0:ref><ns0:ref type='bibr' target='#b4'>Chistiakov et al. 2017</ns0:ref>). These transporters mainly include ATPbinding cassette transporter 1 (ABCA1), ATP-binding cassette subfamily member G1 (ABCG1), and SR-B1, which play crucial roles in preventing the excessive intracellular accumulation of cholesterol and the formation of FCs <ns0:ref type='bibr' target='#b6'>(DiMarco &amp; Fernandez 2015)</ns0:ref>. Similar to that of cholesterol internalization, a confusing relationship characterizes these efflux transporters during the progression of AS. For example, a knockout of ABCA1 aggravated the formation of FCs but did not promote AS plaque development <ns0:ref type='bibr' target='#b50'>(Zhao et al. 2011)</ns0:ref>. Moreover, studies of ABCG1 and SR-B1 have also revealed controversial results, as they have a promoting role in early atherogenesis but a protecting role in advanced AS <ns0:ref type='bibr' target='#b26'>(Meurs et al. 2012;</ns0:ref><ns0:ref type='bibr' target='#b37'>Van Eck et al. 2004;</ns0:ref><ns0:ref type='bibr' target='#b48'>Zhang et al. 2003)</ns0:ref>.</ns0:p><ns0:p>SMCs also play crucial roles in all stages of atherosclerosis, and they are considered the Manuscript to be reviewed leading sources of FCs <ns0:ref type='bibr' target='#b30'>(Poznyak et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b40'>Wang et al. 2019b)</ns0:ref>. During the process of AS, activated macrophages may secrete many inflammatory factors or cytokines that can promote the phenotypic transformation of SMCs from the resting systolic phenotype to the activated synthetic or 'macrophage-like' phenotype. These phenotypic changes in SMCs are generally associated with the downregulated expression of contractile proteins and increased proliferation <ns0:ref type='bibr' target='#b49'>(Zhang et al. 2012</ns0:ref>). In the present study, a total of 432 SMC-FC-related genes were identified, and they were mainly involved in contraction function, inflammation, cell cycle/apoptosis, and substance uptake and intracellular transport, reflecting the roles of SMCs in atherogenesis. Moreover, compared with M-FCs, the expression of the cholesterol efflux transporter ABCA1 in SMC-FCs is much lower. Furthermore, the ABCA1 level in SMCs in advanced AS was decreased compared to early AS; however, no changes in macrophages were observed <ns0:ref type='bibr' target='#b0'>(Allahverdian et al. 2014)</ns0:ref>. These results provide a plausible explanation for the high percentage of FCs derived from SMCs <ns0:ref type='bibr' target='#b0'>(Allahverdian et al. 2014)</ns0:ref>.</ns0:p><ns0:p>RNF13 is an E3 ubiquitin ligase that is widely expressed in various animals and tissues.</ns0:p><ns0:p>Previous studies have determined the roles of RNF13 in myogenesis, neuronal development, and tumorigenesis <ns0:ref type='bibr' target='#b45'>(Zhang et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b47'>Zhang et al. 2010)</ns0:ref>, in which it is mainly involved in cell proliferation <ns0:ref type='bibr' target='#b15'>(Jin et al. 2011)</ns0:ref>. But few studies investigated the role of RNF13 in atherosclerosis or FCs. GLRX is a member of the glutaredoxin family, which highly contributes to the antioxidant defense system. Overexpression of GLRX could attenuate H2O2-induced apoptosis of endothelial cells <ns0:ref type='bibr' target='#b20'>(Li et al. 2017)</ns0:ref>. However, GLRX inhibits endothelial cell angiogenic properties <ns0:ref type='bibr' target='#b25'>(Matsui et al. 2017)</ns0:ref>. Though increasing evidence has suggested the important roles of Manuscript to be reviewed oxidative stress in atherosclerosis <ns0:ref type='bibr' target='#b34'>(Sharif et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b44'>Zhang et al. 2020)</ns0:ref>, few studies have revealed the role of GLRX in atherosclerosis and the formation of FCs. Therefore, the roles of RNF13 and GLRX in the process of AS and FCs remain unknown, which needs further exploration.</ns0:p><ns0:p>The present study has several limitations. Firstly, the sample size of some datasets used in this study is small, which might result in some bias. Secondly, although this is an integrative microarray analysis, those external clinical traits (especially in clinical characteristics and sequencing related technical errors) have not been obtained in our study, which may affect the interpretation of the results. Thirdly, because of the tissue/cell heterogeneity and the potential interexperiment variability, only three co-up regulated genes were identified in this study.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In the present study, 432 SMC-FC-related genes and 81 M-FC-related genes were identified, and they were found to be mainly involved in lipid metabolism, inflammation, cell cycle/apoptosis. Furthermore, 3 co-regulated genes associated with FCs were identified: GLRX, RNF13, and ABCA1. Among these genes, the roles of ABCA1 in the atherosclerotic process have been widely described, but many contradictory results have been presented, suggesting that the roles of ABCA1 in AS need to be further explored. Moreover, the roles of RNF13 and GLRX in atherogenesis remain unknown, and more experiments in the future are needed to confirm their roles. These three genes have a common expression tendency in both SMCs and macrophages, which may have implications for the development of possible targeted therapeutic drugs in the </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50248:1:0:NEW 19 Sep 2020)Manuscript to be reviewed future.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 1 Flow</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>( A )</ns0:head><ns0:label>A</ns0:label><ns0:figDesc>Venn diagram showing the DEGs of each time point (ox-LDL treated for 1 h, 5 h, and 24 h) compared with control. (B) The expression trend clusters of all DEGs obtained above by STEM analysis. (C) Hierarchical cluster heatmap of DEGs with downregulated and upregulated trends overtime obtained by STEM analysis. (D) Venn diagram showing the overlap of DEGs in the ox-LDL-treated SMCs (with downregulated and upregulated trends overtime obtained by STEM analysis) and atherosclerosis-related genes, which were considered SMC-FCs-related DEGs. (E) and (F) The bubble diagram of the KEGG pathway and GO-BP enrichment analyses of SMC-FCs-related DEGs. In the bubble diagram, dot sizes represent counts of enriched DEGs, and dot colors represent negative Log 10 values (p values). SMC-FCs: smooth muscle cell derived foam cells; GO: Gene Ontology; BP: biological process; and KEGG: Kyoto Encyclopedia of Genes and Genomes; ox-LDL: oxidized low-density lipoprotein.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_15'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure> </ns0:body> "

"Dear Editor and Reviewers， We appreciate your generous comments concerning our manuscript entitled “Analysis of genes and underlying mechanisms involved in foam cells formation and atherosclerosis development”. Those comments are all valuable and very helpful for revising and improving our paper, as well as the critical guiding significance to our research. We have carefully revised the paper according to those instructive comments. We believe the paper has much improved after the revision. The revised parts are highlighted in red in the paper. Our point by point responses to your and the reviewers’ comments are attached. Sincerely, we wish you would consider the publication of our revised paper. Thank you! Sincerely, Dr Jian Zhuang, [email protected] Dr. Yueheng Wu, [email protected] Department of Cardiovascular Surgery, Guangdong Provincial Key Laboratory of South China Structural Heart Disease, Guangdong Provincial People's Hospital & Guangdong Academy of Medical Sciences, School of Medicine, South China University of Technology, Guangdong Cardiovascular Institute, Guangzhou, Guangdong, China On behalf of all authors. Responses to Editor: Comments to the Author The reviewers have raised many concerns about the study design (e.g. inclusion criteria, analysis strategy, and validation plan), and interpretation of the results. In addition, it's not clear what additional insights have been uncovered, compared to the original publications. I also have concerns about statistical methods. Specifically, (1) The sample characteristics of the samples are not given. It's unknown whether there are confounding factors that need to be adjusted. Answer: We appreciate the critical comment. This part is the core of the further analysis. Because we obtained the gene expression profiles from the public database, and many sample characteristics are not given. Therefore, we could not adjust the possible confounding factors. In addition, clinical ethics can’t be obtained in a short period of time, and further verification of atrial tissues can’t be carried out in this study. Revision: Limitation： Although this is an integrative microarray analysis, those external clinical traits (especially in clinical characteristics and sequencing related technical errors) have not been obtained in our study, which may affect the interpretation of the results. (2) The preprocessing of the expression data is completely missing. Answer: Thank you for your comments. The microarray data analyzed in this study were obtained from the GEO database “Series Matrix Files” which had been pre-processed, including background correction and normalization. For the RNA-seq dataset, the normalized gene expression matrix of FPKM (fragments per kilobase of exon model per million reads mapped) values was downloaded and transformed into TPM (transcripts per kilobase million) values for comparison(Li et al. 2010). Revision: For the RNA-seq dataset, the normalized gene expression matrix of FPKM (fragments per kilobase of exon model per million reads mapped) values was downloaded and transformed into TPM (transcripts per kilobase million) values for comparison(Li et al. 2010). P value < 0.05 was defined as statistical significance (line 176 to line 179 of revised paper). (3) Using LIMMA for the time-course data GSE68021 and the RNA-seq data is not appropriate. Answer: We appreciate this critical comment. GSE68021 contained the gene expression profile of smooth muscle cells (SMCs) treated with oxidized low-density lipoprotein (ox-LDL) for 0 h, 1 h, 5 h, and 24 h. In the revised manuscript, LIMMA was used to identify the DEGs of each time point (1 h, 5 h, and 24 h) compared with the time point of 0 h. The Short Time-series Expression Miner (STEM) program was used to investigate the expressive trends of the DEGs. While for the RNA-seq dataset, we only extracted out the expression of the three common genes for exploring the difference between stable and unstable plaques without using the LIMMA package. Moreover, the LIMMA package is also available for RNA-seq as well(Ritchie et al. 2015). Revision: For the GSE68021 dataset, the DEGs of each time point (ox-LDL treated for 1 h, 5 h, 24 h) compared with negative control were identified. Furthermore, the Short Time-series Expression Miner (STEM)(Ernst & Bar-Joseph 2006) program was used to analyze the temporal expression profiles of these DEGs. The significant clusters which manifested upregulated or downregulated characteristics over time were selected for the next analysis (line 152 to line 157 of revised paper). (4) The multiple testing correction method used is not stated. Answer: We appreciate this critical comment. In the present study, the Benjamini-Hochberg method was used to adjust p values. Revision: DEGs with adjusted p values < 0.05 (adjusted by the Benjamini-Hochberg method) were considered significant (line 148 to line 149 of revised paper). (5) it's unclear if the results are robust to some variation in the analysis. Answer: Thank you for your comments. The reviewers mainly questioned about the intersections between DEGs of different datasets of different tissues/cells. However, in-vivo/tissue studies could not reveal the molecular mechanism of foam cells (FCs), while in-vitro/cell studies maybe not completely simulated the molecular status of FCs in atherosclerotic plaques. Therefore, we took the intersection between DEGs of the in-vivo dataset and the in-vitro dataset. Moreover, the identification of the potential common molecular mechanisms between the two main sources of FCs (SMCs and macrophages) might provide a better therapeutic target for atherosclerosis. Thus, we also took an intersection between SMCs and macrophages derived FCs related genes. And 15 common genes were identified, though only three co-up regulated genes were found. We think the analysis process is rigorous. Responses to Reviewers: Reviewer #1: Basic reporting The manuscript is overall easy to understand. The introduction is clear, focusing on foam cells in the context of atherosclerosis. Flow chart of the analysis helps to follow the different comparisons. Major comments: (1)- The papers corresponding to the published datasets used in this paper have not been referenced and discussed. It would be important to explain how different this new analysis is compared to the ones done in the original papers. Answer: We appreciate the critical comment. We have cited the original research in our manuscript (line 128 to line 141), and the difference between our work and the original papers were discussed in the “Discussion” part. Revision: Most of them focused on the mechanism of macrophages-derived FCs or atherosclerosis(Ayari & Bricca 2013; Reschen et al. 2015). However, in-vivo/tissue studies could not reveal the molecular mechanism of foam cells (FCs), while in-vitro/cell studies maybe not completely simulated the molecular status of FCs in atherosclerotic plaques. Moreover, few explored the common molecular mechanisms of the two main sources of FCs (SMCs and macrophages) (line 282 to line 288). (2)- The discussion section does not correspond to a discussion of the results but instead to an extended introduction. For example, no mention of the results is done between lines 255 and 275. This needs to be modified and should include the following points. In particular, the authors should include a comparison to the previous analysis of the microarray/RNAseq and stating the novelty/difference of their approach. Answer: We appreciate these critical suggestions. We have removed the sections that do not correspond to the results. The difference between our work and the original papers has been discussed in the “Discussion” part, as mentioned above (the answer to question 1). (3) The detection of only a small overlap between SMC-FCs ad M-FCs should be discussed (due to cell type difference? experimental design?). The authors analysed data from in-vitro experiments as well as in-vivo/tissue. These analyses come with different limitations. Indeed, tissue heterogeneity will limit the interpretation of the result and this needs to be mentioned. Answer: We appreciate the comments. We agree with you that the tissue/cell heterogeneity is an important reason for the small overlap between the two types of FCs. The potential interexperiment variability may be another factor of the small size of overlap. Moreover, though we defined the DEGs by using almost the lowest threshold (adjusted p value < 0.05), only 484 DEGs were obtained in the in-vitro dataset of macrophage FCs, that was another crucial reason of the small overlap. Revision: Thirdly, because of the tissue/cell heterogeneity and the potential interexperiment variability, only three co-up regulated genes were identified in this study (line 360 to line 362). Minor comments: (4)- A few typos and unclear sentences can be found along the manuscript such as Line 82 “plague”; Line 206 “GSE5466 was expressed”; Line 244 “corelated”; Line 333 “compared its expressing in”. Answer: Thank you for your kind comments. We apologize for the previous mistakes. Revision: “plaque”; “GSE54666 was the gene expression profiling of macrophages”; “the common FCs-related genes”. (5)- Figures: Gene names in Heatmap should be removed if they are too small to be read. A title for the x-axis of the bubble diagram should be added and the size of the text should be increased. Answer: Thank you for your kind suggestions. We have made corrections in this revised paper, including removing the gene name in heatmaps, adding the x-axis’ titles of the bubble diagram, and changing the size of the text to 7 pt. (6)- The manuscript contains many abbreviations that make the reading of the manuscript difficult. For clarity, some abbreviations could be removed (“atherosclerosis” shouldn’t be abbreviated to AS). Answer: Thank you for your kind suggestions. We have removed some abbreviations with no much necessary. (7)- In Figure 6B, all significant comparisons are shown while only a few comparisons are described in the text. It might be better to show only the statistics for the relevant comparison. Answer: Thank you for your kind suggestions. In the revised version, the unnecessary comparisons were removed. Experimental design Overall, relevant questions and objectives are well defined with the flow chart helping to understand the analysis. Methods section is comprehensive. Major comments: (8)- The choice of dataset to include in this analysis has not been explained clearly. Does it correspond to all publicly available data on foam cells and atherosclerosis? If only some dataset were included, can the authors justify their selection criteria. Answer: We appreciate the critical comments. The criteria of this study were added in the revised manuscript (line 118 to line 127 of the revised paper). Revision: To obtain the genes associated with the formation as well as the development of atherosclerosis, the dataset consisted of human atherosclerotic plaques and normal arteries or early and advanced atherosclerotic plaques that were searched in the GEO database, and two datasets (GSE28829 and GSE43292) were obtained. Then, the in-vitro dataset contained ox-LDL treated smooth muscle cells (SMC) or macrophages that were screened out to identify foam cell-related DEGs of the two cell types. Furthermore, because of the important role of foam cells in the development of vulnerable plaques(Liu et al. 2017), the dataset composed of unstable vs. stable or ruptured vs. unruptured atherosclerotic plaques were screened to explore the expression of the foam cell-related genes in vulnerable plaques. The datasets of animal samples and the in-vivo human datasets of serum or plasma were excluded in this study. (9)- The choice and order of comparisons in this analysis are not completely claer for me. Indeed, the authors started by analysing 2 in-vivo dataset then 2 in-vitro datasets. Finally, they finish by analysing another in-vivo and in-vitro dataset, for validation. I do not understand why the 3 in-vivo datasets where not analysed and integrated together. What is the justification for choosing the dataset for target identification versus the dataset for validation? Answer: Thank you for your comments. Actually, we analyzed four in-vivo datasets in this study, including GSE28829, GSE43292, GSE120521, and GSE41571. The first two datasets were used to identify the differentially expressed genes (DEGs) associated with the formation as well as the development of atherosclerosis. The last two datasets were used to explore the expression of the foam cells (FCs)-related genes in vulnerable plaques. (10) Minor comment: The threshold (fold change and pvalue) to identify DEGs should be included in the Methods. Answer: Thank you for your kind suggestions. However, in the Methods section, we have defined the DEGs as “DEGs with adjusted p values < 0.05 (adjusted by the Benjamini-Hochberg method) were considered significant”, like previous studies(Huang et al. 2019; Wang et al. 2019). Validity of the findings While the comparison of in-vitro and in-vivo dataset to identify genes involved in foam cell formation in atherosclerosis is interesting, the interpretation of results and discussion is sometimes missing. (11) Major comment: As mentioned above, the Discussion of the manuscript should include the limitations of this type of analysis. Answer: Thank you for your kind suggestions. We have discussed the limitations in the revised manuscript (line 356 to line 362). Revision： The present study has several limitations. Firstly, the sample size of some datasets used in this study is small, which might result in some bias. Secondly, although this is an integrative microarray analysis, those external clinical traits (especially in clinical characteristics and sequencing related technical errors) have not been obtained in our study, which may affect the interpretation of the results. Thirdly, because of the tissue/cell heterogeneity and the potential interexperiment variability, only three co-up regulated genes were identified in this study. (12) Comments for the author In summary, this manuscript provides an interesting analysis by integrating different datasets to find relevant candidates involved in foam cell formation. The different steps of the analysis and results are clearly explained. However, the choice of data comparison is not justified and does not seem pertinent. The authors only find 4 candidates, probably due to the limitations of this kind of approach. Discussion of the results and limitation should be incorporated to this manuscript. Answer: Thank you for your critical comments. As mentioned above, we have discussed the limitations. Reviewer #2: Basic reporting Figures need to be remaking. Answer: Thank you for your kind comments. We have revised the figures. Experimental design no comment Validity of the findings no comment Comments for the Author This manuscript provided a list of genes that are functional related to the development of atherosclerosis. The idea to utilize multiple existing datasets is valuable, though I am not fully convinced by the technical details of the proposed approach. My comments are list as follow. Major comments: 1. For the DEGs associated with AS, the first gene sets come from DE analysis of early versus late atherosclerosis samples, and the second gene sets come from DE analysis of athrosclerotic versus nonathrosclerotic samples. It is not clear to me why you need to take the intersection of the intersection of the DEGs of the two comparisons. The first experiment design is comparing between patients with AS not health control. Theoretically, the second set of gene is enough. Please explain the specific purpose of including first gene sets. Answer: Thank you for your critical comments. In our work, we aimed to explore the possible common molecular mechanisms involved in the formation as well as the development of atherosclerosis. The first dataset, containing early and late atherosclerosis samples, was used to identify genes associated with the development of atherosclerosis. The second dataset was used to obtain the DEGs associated with the formation of atherosclerosis. Therefore, we took the intersection of the DEGs between the two datasets to identify the DEGs involved in the formation as well as the development of atherosclerosis. 2. To generate SMC-FCs related genes and M-FCs related genes. The authors simply take the overlap genes between studies. This approach ignores the difference of sample size across study designs and the effect size of each gene in DE analysis. It would be better to use a meta-analysis which adjusted for these effects. Answer: Thank you for your kind comments. Maybe the method of simple overlap is not strong to maintain the integrity of all valid information, but the results generated by using this method are robust enough, for these results are universal in all the different studies. And lots of works also took this simple approach to identify the common DEGs(Huang et al. 2019; Liu et al. 2020; Meng et al. 2019). 3. To validate the four coregulated genes associated with FCs, the authors used vulnerable AS plaques dataset. It seems to be a logic gap need to be filled. It is not clear to me why the vulnerability of AS can validate the role of FCs in AS. Please explain it in more details. Answer: Thank you for your critical comments. Previous studies have suggested that the formation and retention of the FCs could exacerbate atherosclerosis and fuel the development of the vulnerable plaques(Bäck & Hansson 2015; Liu et al. 2017) (line 123, line 266 to line268 of the revised manuscript). In addition to the synthesis of extracellular matrix, activated FCs can also secrete multiple matrix-degrading enzymes, leading to the instability and even rupture of plaques(Maguire et al. 2019). Therefore, we explored the expression of the FCs-related genes in vulnerable plaques. Minor comment: The gene names (row names) of Figure 2 and Figure 4’s heatmaps are unreadable. Please make a clearer figure or completely remove the gene name and put them in supplementary tables. Answer: Thank you for your kind suggestions. We have removed the gene name of the heatmaps. Reference Ayari H, and Bricca G. 2013. Identification of two genes potentially associated in iron-heme homeostasis in human carotid plaque using microarray analysis. J Biosci 38:311-315. 10.1007/s12038-013-9310-2 Bäck M, and Hansson GK. 2015. Anti-inflammatory therapies for atherosclerosis. Nat Rev Cardiol 12:199-211. 10.1038/nrcardio.2015.5 Ernst J, and Bar-Joseph Z. 2006. STEM: a tool for the analysis of short time series gene expression data. BMC Bioinformatics 7:191. 10.1186/1471-2105-7-191 Huang HM, Jiang X, Hao ML, Shan MJ, Qiu Y, Hu GF, Wang Q, Yu ZQ, Meng LB, and Zou YY. 2019. Identification of biomarkers in macrophages of atherosclerosis by microarray analysis. Lipids Health Dis 18:107. 10.1186/s12944-019-1056-x Li B, Ruotti V, Stewart RM, Thomson JA, and Dewey CN. 2010. RNA-Seq gene expression estimation with read mapping uncertainty. Bioinformatics 26:493-500. 10.1093/bioinformatics/btp692 Liu Y, Huan W, Wu J, Zou S, and Qu L. 2020. IGFBP6 Is Downregulated in Unstable Carotid Atherosclerotic Plaques According to an Integrated Bioinformatics Analysis and Experimental Verification. J Atheroscler Thromb. 10.5551/jat.52993 Liu Z, Zhu H, Dai X, Wang C, Ding Y, Song P, and Zou MH. 2017. Macrophage Liver Kinase B1 Inhibits Foam Cell Formation and Atherosclerosis. Circ Res 121:1047-1057. 10.1161/circresaha.117.311546 Maguire EM, Pearce SWA, and Xiao Q. 2019. Foam cell formation: A new target for fighting atherosclerosis and cardiovascular disease. Vascul Pharmacol 112:54-71. 10.1016/j.vph.2018.08.002 Meng LB, Shan MJ, Qiu Y, Qi R, Yu ZM, Guo P, Di CY, and Gong T. 2019. TPM2 as a potential predictive biomarker for atherosclerosis. Aging (Albany NY) 11:6960-6982. 10.18632/aging.102231 Reschen ME, Gaulton KJ, Lin D, Soilleux EJ, Morris AJ, Smyth SS, and O'Callaghan CA. 2015. Lipid-induced epigenomic changes in human macrophages identify a coronary artery disease-associated variant that regulates PPAP2B Expression through Altered C/EBP-beta binding. PLoS Genet 11:e1005061. 10.1371/journal.pgen.1005061 Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, and Smyth GK. 2015. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43:e47. 10.1093/nar/gkv007 Wang CH, Shi HH, Chen LH, Li XL, Cao GL, and Hu XF. 2019. Identification of Key lncRNAs Associated With Atherosclerosis Progression Based on Public Datasets. Front Genet 10:123. 10.3389/fgene.2019.00123 "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background: This study aimed to develop a deep-learning model and a risk-score system using clinical variables to predict intensive care unit (ICU) admission and in-hospital mortality in COVID-19 patients. Methods: This retrospective study consisted of 5766 persons-under-investigation for COVID-19 between February 7, 2020, and May 4, 2020. Demographics, chronic comorbidities, vital signs, symptoms, and laboratory tests at admission were collected. A deep neural network model and a risk-score system were constructed to predict ICU admission and in-hospital mortality. Prediction performance used the receiver operating characteristic area under the curve (AUC). Results: The top ICU predictors were procalcitonin, lactate dehydrogenase, C-reactive protein, ferritin, and oxygen saturation. The top mortality predictors were age, lactate dehydrogenase, procalcitonin, cardiac troponin, C-reactive protein, and oxygen saturation. Age and troponin were unique top predictors for mortality but not ICU admission. The deep-learning model predicted ICU admission and mortality with an AUC of 0.780 [95%</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Since the first reports of severe respiratory illness caused by coronavirus disease 2019 in Wuhan, China in mid-December 2019 <ns0:ref type='bibr'>(Huang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b28'>Zhu et al. 2020</ns0:ref>), over 6.2 million individuals have been infected, resulting in over 370,000 deaths worldwide <ns0:ref type='bibr'>(May 31, 2020)</ns0:ref>. The actual numbers are likely to be much higher due to testing shortages and under-reporting <ns0:ref type='bibr'>(Yelin et al. 2020)</ns0:ref>.</ns0:p><ns0:p>Many patients have mild or asymptomatic infections, while others deteriorate rapidly with multi-organ failure. There will likely be recurrence and secondary waves of this pandemic <ns0:ref type='bibr' target='#b14'>(Leung et al. 2020)</ns0:ref>.</ns0:p><ns0:p>A large array of clinical and demographic variables associated with COVID-19 infection have been identified (see reviews <ns0:ref type='bibr' target='#b1'>(Brown et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b2'>Cao et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b17'>Rodriguez-Morales et al. 2020)</ns0:ref>). A few of these have been associated with high likelihood of critical illness or mortality. There are however no established prognostic models that reliably predict the need for escalated (intensive care unit, ICU) care or mortality due to COVID-19 infection. Lacking this, effective triage of patients is challenging in a resource-constrained environment. The problem is further magnified by the poor sensitivity <ns0:ref type='bibr'>(Kim et al. 2020, in press</ns0:ref>) and a few day turnaround time <ns0:ref type='bibr'>(Yelin I 2020)</ns0:ref> of the most commonly used reversetranscriptase polymerase chain reaction (RT-PCR) test, during which time patients are assumed COVID-19 positive. This problem strains the resources of many hospitals and highlights the need for effective tools to anticipate patients' progression and properly triage patients.</ns0:p><ns0:p>The goal of this study was to develop a deep-learning algorithm (in contrast to previous methods) to identify the top, statistically significant predictors amongst the large array of clinical variables at admission to predict the likelihood of ICU admission and in-hospital mortality in COVID-19 patients. We further developed a simplified risk-score model to predict the likelihood of ICU admission and in-hospital mortality.</ns0:p></ns0:div> <ns0:div><ns0:head>METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>Study population</ns0:head><ns0:p>This retrospective study was approved by Institutional Review Board with exemption of informed consent and HIPAA waiver (Stony Brook University Hospital, IRB-2020-00207). Stony Brook University Hospital, the only academic hospital serving Suffolk county, about 40 miles east of New York City, was one of the hardest hit counties in the country at the time of this writing. The COVID-19 Persons Under Investigation (PUI) registry consisted of 5766 patients from February 7th, 2020 to May 4th, 2020. Only patients who were diagnosed by positive tests of real-time polymerase chain reaction (RT-PCR) for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) were included in the study.</ns0:p><ns0:p>Demographic information, chronic comorbidities, imaging findings, vital signs, symptoms, and laboratory tests at admission were collected. Imaging findings were extracted from patient chart review, which included information provided by radiology report as part of standard of care. The primary outcome was ICU admission versus general floor admission, and the secondary outcome was in-hospital mortality versus discharge. Mortality outside of hospital after discharge was not obtained.</ns0:p><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref> shows the flowchart of patient selection. Of the 2594 confirmed COVID-19 positive cases, all 1108 hospitalized COVID-19 positive patients were used in our analysis. patients were admitted to the ICU directly and an additional 194 patients were subsequently upgraded to an ICU from a general floor. Among these 271 ICU patients, 108 were discharged alive, 77 expired during the hospitalization and the other 86 are still in the hospital at the time of this analysis. Comparison was made to 837 general admissions who did not receive ICU care, among whom 772 patients were discharged alive and 65 expired during the hospitalization (none remained in the hospital).</ns0:p></ns0:div> <ns0:div><ns0:head>Data preprocessing</ns0:head><ns0:p>Two patients were excluded from machine learning analysis for missing categorical variables.</ns0:p><ns0:p>Brain natriuretic peptide (BNP) was missing from &gt;15% of patients, thus they were excluded from machine learning analysis. For the rest of the laboratory variables, missing data (in &lt;5% of patients) was imputed with predictive mean modeling using the Multivariate Imputation by Chained Equations in R (statistical analysis software, version 4.0) (van Buuren &amp; Groothuis-Oudshoorn 2011).</ns0:p></ns0:div> <ns0:div><ns0:head>Deep neural network prediction model</ns0:head><ns0:p>Ranking of clinical variables of categorical or numerical values were made using the Boruta, a statistical software <ns0:ref type='bibr' target='#b13'>(Kursa &amp; Rudnicki 2010)</ns0:ref>. Boruta ranks feature importance using the Random Forest method. In this decision tree-based method, the quantitative measure of importance is the Gini feature of importance, which counts the times that a feature is used to split a node of a decision tree, statistically weighted by the number of instances the node splits. In the DNN model, the top predictors were those that demonstrated statistical significance using built-in statistical methods within the Boruta algorithm.</ns0:p><ns0:p>A correlation coefficient &gt;0.5 from collinearity analysis was used to exclude correlated variables from machine learning analysis. Note that none of the top features we used in the final analysis demonstrated strong correlation with other features. Thus, no top features were removed as a result. A deep neural network (DNN) was constructed to predict ICU admission and mortality using five fully connected dense layers <ns0:ref type='bibr' target='#b3'>(Chen et al. 2020)</ns0:ref>. The top clinical predictors were input parameters, determined by testing subsets of these parameters, and ICU admission and mortality were outcome parameters. The DNN model used 5 hidden layers with 6, 8, 16, 8, 4 neurons respectively. We explored a few models using a range of number (3-7) of layers, and the 5-layer model yielded the optimal validation result. ReLu activation function for the hidden layers, the sigmoid activation function for the output layer, and the PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed 'he_normal' normalization scheme were applied. In the model training process, we used Adam optimizer, mean squared error as the cost function, a default learning rate of 0.01, and number of epochs of 100. The reported results yielded from the average of 5 consecutive runs. The dataset was randomly split into 90% training data and 10% testing data. ICU admission and mortality results were categorized using a binary classification. To minimize overfitting, we employed 5-fold cross-validation, ranked and removed less important features using correlation analysis and based on statistical significance by Boruta. We also employed regularization and stopped the training process at 100 epochs.</ns0:p></ns0:div> <ns0:div><ns0:head>Risk score model</ns0:head><ns0:p>Risk-score systems were constructed using the top independent clinical variables to predict ICU admission and mortality. For risk score, the mixed Generalized Additive Model was used to plot the probability of ICU admission and mortality for each clinical variable <ns0:ref type='bibr' target='#b23'>(Wood 2001)</ns0:ref>. Different cutoff points were evaluated where the chosen cutoff points yielded the optimal distribution (not skewed to high or low scores) of the risk score model. The corresponding numerical values of each top feature at probability of 0.3 for ICU and 0.2 for mortality were found to be the optimal cutoff values for the risk score model. Each of the top variables was assigned a weight of one point if the clinical measurement was above the probability cutoff. The risk score ranged from 0 to 5 for ICU admission and 0 to 6 for mortality (which were chosen based on statistical significance, see Results).</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis and performance evaluation</ns0:head><ns0:p>Statistical analysis was performed in SPSS v26 and in R (statistical analysis software 4.0). Group comparisons of categorical variables in frequencies and percentages used the chi-square test or Fisher exact test. Group comparison of continuous variables in medians and interquartile ranges (IQR) used the Mann-Whitney U test. A p value &lt; 0.05 was considered to be statistically significant. For performance evaluation, data were split 90% for training and 10% for testing. Prediction performance was evaluated by calculating the area under the curve (AUC) of the receiver operating characteristic (ROC) curve, accuracy, sensitivity, specificity, precision, recall, negative predictive value (NPV), positive predictive value (PPV) and F1 score (a harmonic mean of precision and recall). The average ROC analysis was repeated with five runs. In risk score models, SPSS was used to cross-check statistical significance of the top features, in which all top features used in the final analysis of risk score model had a p &lt; 0.001. <ns0:ref type='table' target='#tab_3'>PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head></ns0:div> <ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div> <ns0:div><ns0:head>Clinical variables associated with ICU admission</ns0:head><ns0:p>Table <ns0:ref type='table' target='#tab_0'>1</ns0:ref> summarizes the demographic characteristics, vital signs, comorbidities, and laboratory data for the ICU (n=271) and non-ICU (n=837) group. The median age of the ICU group was lower than that of the general admission group (59 years [IQR:49-71] versus 62 years [IQR:50-76], p=0.027).</ns0:p><ns0:p>Disproportionally more males were admitted to the ICU (67.5%% vs 32.5%, p&lt;0.001). History of cancer was the only comorbidity that was significantly associated with ICU admission (P=0.016).</ns0:p><ns0:p>All measured vital signs were significantly different between the ICU group and the non-ICU group. The ICU group had higher heart rate, respiratory rate and temperature, but lower systolic blood pressure and oxygen saturation (p&lt;0.05). The ICU group had higher alanine aminotransferase (ALT), Creactive protein (CRP), D-dimer, ferritin, lactate dehydrogenase (LDH), white blood cells (WBC), and procalcitonin (p&lt;0.05) and lower lymphocyte counts (p&lt;0.05). Cardiac troponin and BNP were not significantly different between groups (p&gt;0.05).</ns0:p><ns0:p>The symptom of dyspnea was significantly associated with ICU admission (p=0.001). Patients admitted to ICU were more likely to present with abnormal chest x-ray (p&lt;0.001), and more likely to have bilateral chest x-ray abnormalities on presentation, compared to that of general admission group (p&lt;0.001). </ns0:p></ns0:div> <ns0:div><ns0:head>Prediction models for ICU admission</ns0:head></ns0:div> <ns0:div><ns0:head>ICU admission for the testing set (Table 2).</ns0:head><ns0:p>A risk score system was constructed (training data set) using the top five statistically significant clinical variables, with 1 point given for each variable meeting the following criteria: procalcitonin&gt;0.5ng/mL, LDH &gt;487U/L and &lt;12586.7U/L, CRP&gt;14.2mg/dL, ferritin&gt;1250ng/mL and &lt;13080.5ng/mL, and SpO2&lt;88.8%. Odds ratios of procalcitonin, LDH, CRP, ferritin, and SpO2 for ICU admission were 3.062, 3.846, 3.001, 2.449, and 3.665, respectively. Figure <ns0:ref type='figure' target='#fig_6'>4</ns0:ref> shows the results for the testing data set using the risk score system. ICU admission rate increased with increasing risk scores. The performance of the risk score yielded an AUC of 0.728 [95% CI:0.726-0.729] for predicting ICU admission for the testing data set.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Clinical variables associated with mortality</ns0:head><ns0:p>Table <ns0:ref type='table' target='#tab_4'>3</ns0:ref> summarizes the demographic data, vital signs, comorbidities, and laboratory data for the non-survivors (n=142) and survivors (n=880) group. The median age of the non-survivor group was higher than that of the survivor group (76 years [IQR:66-84] versus 59 years [IQR:49-72], p&lt;0.001).</ns0:p><ns0:p>There was a disproportionally higher mortality rate in males (65.5% vs 34.5%, p=0.014). Of the comorbidities, hypertension, coronary artery disease, heart failure, chronic obstructive pulmonary disease, smoking history, and chronic kidney disease were significantly different between groups (p&lt;0.05).</ns0:p><ns0:p>Among vital signs, tachypnea and hypoxemia were significantly different between groups at presentation (p&lt;0.05). The expired cohort had higher BNP, CRP, D-dimer, ferritin, LDH, WBC, procalcitonin, and cardiac troponin but lower lymphocytes (p&lt;0.05). ALT was not significantly different between groups.</ns0:p><ns0:p>Among the symptoms, cough, myalgia, nausea or vomiting, chest discomfort, fatigue, fever, loss of taste, and headache were significantly different between groups (p&lt;0.05). There was no significant difference in x-ray findings between groups at presentation.</ns0:p></ns0:div> <ns0:div><ns0:head>Prediction models for mortality</ns0:head><ns0:p>The top 6 statistically significant predictors of mortality were age, LDH, procalcitonin, troponin, CRP, and SpO2 (Figure <ns0:ref type='figure' target='#fig_7'>5</ns0:ref> <ns0:ref type='table' target='#tab_7'>4</ns0:ref>). A risk score system was constructed (training data set) using the top 6 statistically significant clinical variables to predict mortality. The thresholds for the risk scores were: age &gt;71 years, LDH &gt;487U/L, procalcitonin &gt;1.1ng/mL, troponin &gt;0.03ng/mL, CRP &gt;17mg/dL, and SpO2 &lt;88%. Odds ratios of age, LDH, procalcitonin, troponin, CRP, and SpO2 for mortality were 4. <ns0:ref type='bibr'>301, 3.418, 6.232, 5.253, 4.240, and 3.750</ns0:ref>, respectively. Higher mortality rate was associated with higher risk scores for the testing set (Figure <ns0:ref type='figure' target='#fig_9'>7</ns0:ref>). The performance of the risk score yielded an AUC of 0.848 [95% CI:0.847-0.849] in predicting mortality for the testing set.</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Mining a large cohort of COVID-19 patients in the United States, deep-learning and resultant risk score models identified the top predictors of ICU admission in COVID-19 to be the admission levels of procalcitonin, LDH, CRP, ferritin, and SpO2; the top predictors of mortality were age, LDH, procalcitonin, cardiac troponin, CRP, and SpO2. Predictive models were developed using deep neural network of the top predictors, yielding an AUC of 0.779 and 0.882 for predicting ICU admission and mortality, respectively. The corresponding simplified risk scores yielded an AUC of 0.728 and 0.848, respectively.</ns0:p><ns0:p>The association between these biomarkers and poor outcomes in COVID-19 victims is biologically plausible: procalcitonin is elevated during bacterial infection, but less so during viral infection, suggesting that bacterial co-infection leads to worse outcome in COVID-19 patients <ns0:ref type='bibr' target='#b0'>(Assicot et al. 1993)</ns0:ref>. LDH reflects tissue damage <ns0:ref type='bibr'>(Huang et al. 2020;</ns0:ref><ns0:ref type='bibr' target='#b28'>Zhu et al. 2020</ns0:ref>), while CRP is indicative of inflammation <ns0:ref type='bibr' target='#b6'>(Gabay &amp; Kushner 1999)</ns0:ref>. Elevated ferritin is associated with acute respiratory distress syndrome (ARDS) <ns0:ref type='bibr' target='#b4'>(Connelly et al. 1997)</ns0:ref> and may be a marker of aberrant iron metabolism that could render the lungs susceptible to oxidative damage <ns0:ref type='bibr' target='#b16'>(Mumby et al. 2004</ns0:ref>). Ferritin may reflect hyperinflammation associated with a cytokine storm and multi-organ failure <ns0:ref type='bibr'>(Mehta et al. 2020)</ns0:ref>. Low SpO2 indicates failure of the lungs to oxygenate blood effectively, leading to tissue hypoxia <ns0:ref type='bibr' target='#b4'>(Connelly et al. 1997)</ns0:ref>. Elevated cardiac troponin indicates cardiac injury <ns0:ref type='bibr'>(Huang et al. 2020)</ns0:ref>. Although these variables have been previously associated with COVID-19 infection, most previous studies did not rank these clinical variables, or develop predictive models or risk scores to predict ICU admission or mortality. Not surprisingly, some of the same biomarkers in our study predicted both the need for ICU admission and likelihood of mortality. However, age and admission troponin level were uniquely predictive of mortality, indicating older age and cardiac issues are associated with higher rate of mortality in COVID-19 infection.</ns0:p><ns0:p>It is notable that individual comorbidities did not rank high in predicting ICU admission and mortality. Specifically, a history of heart failure, COPD, and coronary artery disease only ranked 7 th , 11 th and 14 th respectively for predicting mortality. Similarly, the patients' symptoms and vital signs (other than SpO2) at the time of admission were not found to be the top predictors of poor outcome. Although some comorbidities have been reported to be associated with critical illness and mortality, most previously studies did not rank their importance with respect to other laboratory variables.</ns0:p><ns0:p>Our predictive AUC performance for ICU admission was poorer than that for mortality. We speculate this might be due to variability in triage decision-making to send patients to ICU among frontline clinicians. For both predictions, precision, PPV and F1 scores were comparatively low, which was not unexpected due to the imbalanced sample sizes between the two groups as well as small sample sizes. Further studies are warranted. While a large number of studies have previously identified clinical variables associated with the severity of COVID-19 infection, only a few studies have attempted to develop a predictive or risk score model to predict mortality and disease severity. Jiang et al. used supervised learning (not deep learning) and found mildly elevated alanine aminotransferase, myalgias, and hemoglobin at presentation to be predictive of severe ARDS of COVID-19 with 70% to 80% accuracy. This study had small, non-uniform, heterogeneous clinical variables, obtained from different hospitals <ns0:ref type='bibr' target='#b10'>(Jiang et al. 2020</ns0:ref>). Ji et al. used logistic regression to predict stable versus progressive COVID-19 patients (n=208) based on whether their conditions worsened during hospitalization <ns0:ref type='bibr' target='#b9'>(Ji et al. 2020)</ns0:ref>. They reported comorbidities, older age, lower lymphocyte and higher lactate dehydrogenase at presentation to be independent high-risk factors for COVID-19 progression but did not develop a risk score. A nomogram of these 4 factors yielded a concordance index of 0.86. Yan et al. utilized supervised machine learning to predict critical COVID-19 at admission using presence of X-ray abnormality, cancer history, age, neutrophil/lymphocyte ratio, LDH, dyspnea, bilirubin, unconsciousness and number of comorbidities <ns0:ref type='bibr'>(Yan et al. 2020, in press</ns0:ref>). They reported an AUC of 0.88. Yuan et al. went one step further to predict mortality more than 12 days in advance with &gt;90% accuracy across all cohorts. Moreover, their Kaplan-Meier score shows that patients upon admission could clearly be differentiated into low, medium or high risk. They created a simple risk score system, and validated using multiple independent cohorts <ns0:ref type='bibr' target='#b28'>(Yuan et al. 2020)</ns0:ref>.</ns0:p><ns0:p>Our approach used a deep-learning algorithm which is novel and has distinct advantages over logistic regression and supervised learning approach. Deep learning is increasingly being used in medicine <ns0:ref type='bibr' target='#b5'>(Deo 2015;</ns0:ref><ns0:ref type='bibr' target='#b19'>Santos et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b21'>Tschandl et al. 2019)</ns0:ref>. In contrast to conventional analysis methods, which specify the relationships amongst data elements to outcomes, machine learning employs computer algorithms to identify relationships amongst different data elements to inform outcomes without the need to specify such relationships a priori. Deep learning can outperform human experts in performing many tasks in medicine <ns0:ref type='bibr' target='#b11'>(Killock 2020)</ns0:ref>. In addition to approximating physician skills, Deep learning can also detect novel relationships not readily apparent to human perception, especially in large, complex, and longitudinal datasets. Disadvantages of deep learning methods are that it requires comparatively large sample size, there is a potential of overfitting, and the complex relations could make deep learning results difficult to interpret, amongst others. In addition, we devised a simplified practical risk score adds practical utility to these findings. Although we ranked all variables and explicitly listed 10 or 15 top variables, we built the predictive model and risk score model using only the top 5 variables to simplify and increase translation potential in the clinical settings. The excellent prediction performances using a few clinical variables are encouraging.</ns0:p><ns0:p>This study has several limitations in addition to those mentioned above. This is a retrospective study carried out in a single hospital. These findings need to be replicated in large and multi-institutional settings for generalizability. We only analyzed clinical variables at admission. Longitudinal changes of these clinical variables need to be studied. As in all observational studies, other residual confounders may Manuscript to be reviewed exist that were not accounted for in our analysis. Future prospective studies validating our predictive models and scores are warranted.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSION</ns0:head><ns0:p>We implemented a deep-learning algorithm and a risk score model to predict the likelihood of ICU admission and mortality in COVID-19 patients. Our predictive model and risk score model can be easily retrained with additional data, new local data, as well as additional clinical variables. This approach has the potential to provide frontline physicians with a simple and objective tool to stratify patients based on risks so that COVID-19 patients can be triaged more effectively in time-sensitive, stressful and potentially resource-constrained environments. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2 shows the ranking of the clinical variables associated with ICU admission. The top 5 statistically significant predictors of ICU admission were procalcitonin, LDH, CRP, ferritin, and SpO2. A deep neural network predictive model for mortality was constructed using the top clinical variables and trained using the training dataset and tested on an independent testing dataset. The ROC and confusion matrix of the testing dataset are shown in Figure 3. The performance of the DNN model yielded an AUC = 0.780 [95% CI:0.760-0.785], sensitivity = 0.760, specificity = 0.709, and F1 score = 0.551 in predicting</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>). A deep neural network predictive model for mortality was constructed using the top clinical variables and trained using the training data set. The ROC and confusion matrix are shown in Figure 6. The performance of the DNN model yielded an AUC of 0.844 [95% CI:0.839-0.848], sensitivity = 0.750, specificity = 0.872, and F1 score = 0.616 for the testing dataset (Table</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure legends Figure 1 .</ns0:head><ns0:label>legends1</ns0:label><ns0:figDesc>Figure legends</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Ranking of clinical variables for predicting ICU admission by Boruta algorithm. The x-axis is attribute of level of importance, where a larger number indicates relatively higher importance. The y-axis are laboratory test variables. The top statistically significant predictors were: procalcitonin, LDH, CRP, ferritin, SpO2, lymphocytes, respiratory rate, systolic blood pressure, age and ALT. The top 10 variables were significant.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. ROC and confusion matrix for prediction of ICU admission of the DNN model.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. Risk score stratification for ICU admission. Scores ranged from 0 to 5, with 0 indicating the lowest risk and 5 being the highest risk of mortality. The numbers in the bar indicate the number of patients in the ICU (red) and non-ICU (blue) that were correctly predicted in the testing dataset.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Ranking of clinical variables for predicting mortality by Boruta algorithm. The x-axis is attribute of level of importance, where a larger number indicates relatively higher importance. The y-axis are laboratory test variables. The top statistically significant predictors were: age, LDH, procalcitonin, troponin, CRP, SpO2, history of heart failure, respiratory rate, lymphocytes, ferritin, history of COPD, Ddimer, ALT, history of coronary heart disease, and systolic blood pressure. The top 15 variables were significant.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. ROC and confusion matrix for prediction of mortality of the DNN model.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 7 .</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7. Risk score stratification for mortality. Scores ranged from 0 to 6, with 0 indicating the lowest risk and 6 being the highest risk of mortality. The numbers in the bar indicate the number of patients in the ICU (red) and non-ICU (blue) that were correctly predicted in the testing dataset.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Figure 6 ROC</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head /><ns0:label /><ns0:figDesc>Demographic characteristics, comorbidities, symptoms, imaging findings, vital signs, and laboratory findings of ICU versus non-ICU patients. Demographic characteristics, comorbidities, symptoms, imaging findings, vital signs, and laboratory findings of ICU versus non-ICU patients. Group comparison of categorical variables in frequencies and percentages used c 2 test or Fisher exact tests. Group comparison of continuous variables in medians and interquartile ranges (IQR) used the Mann-Whitney U test. PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,42.52,204.37,525.00,208.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,280.87,525.00,262.50' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Demographic characteristics, comorbidities, symptoms, imaging findings, vital signs, and 2 laboratory findings of ICU versus non-ICU patients. Group comparison of categorical variables in 3 frequencies and percentages used &#61539; 2 test or Fisher exact tests. Group comparison of continuous variables 4 in medians and interquartile ranges (IQR) used the Mann-Whitney U test.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Patients, No. (%)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>ICU</ns0:cell><ns0:cell>Non-ICU</ns0:cell><ns0:cell>p value</ns0:cell></ns0:row><ns0:row><ns0:cell>(n=271)</ns0:cell><ns0:cell>(n=837)</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Performance indices for predicting ICU admission of the testing dataset</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Performance indices for predicting ICU admission of the testing dataset. Abbreviations: area</ns0:cell></ns0:row><ns0:row><ns0:cell>under the curve (AUC), accuracy, sensitivity, specificity, precision, recall, negative predictive</ns0:cell></ns0:row><ns0:row><ns0:cell>value (NPV), positive predictive value (PPV) and F1 score (a harmonic mean of precision and</ns0:cell></ns0:row><ns0:row><ns0:cell>recall).</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Performance indices for predicting ICU admission of the testing dataset. Abbreviations: area under the curve (AUC), accuracy, sensitivity, specificity, precision, recall, negative predictive value (NPV), positive predictive value (PPV) and F1 score (a harmonic mean of precision and recall).</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>AUC</ns0:cell><ns0:cell cols='5'>Accuracy Sensitivity Specificity Precision NPV</ns0:cell><ns0:cell>PPV</ns0:cell><ns0:cell>F1</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Training 0.751</ns0:cell><ns0:cell>0.703</ns0:cell><ns0:cell>0.707</ns0:cell><ns0:cell>0.701</ns0:cell><ns0:cell>0.437</ns0:cell><ns0:cell>0.879</ns0:cell><ns0:cell>0.437</ns0:cell><ns0:cell>0.540</ns0:cell></ns0:row><ns0:row><ns0:cell>Testing</ns0:cell><ns0:cell>0.728</ns0:cell><ns0:cell>0.721</ns0:cell><ns0:cell>0.760</ns0:cell><ns0:cell>0.709</ns0:cell><ns0:cell>0.432</ns0:cell><ns0:cell>0.910</ns0:cell><ns0:cell>0.432</ns0:cell><ns0:cell>0.551</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Demographic characteristics, comorbidities, symptoms, imaging findings, vital signs, and laboratory findings of death versus non-death (discharged).</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='2'>Demographic characteristics, comorbidities, symptoms, imaging findings, vital signs, and</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>laboratory findings of death versus non-death (discharged). Group comparison of categorical</ns0:cell></ns0:row><ns0:row><ns0:cell>variables in frequencies and percentages used c 2</ns0:cell><ns0:cell>or Fisher exact tests. Group comparison of</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>continuous variables in medians and interquartile ranges (IQR) used the Mann-Whitney U</ns0:cell></ns0:row><ns0:row><ns0:cell>test.</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Demographic characteristics, comorbidities, symptoms, imaging findings, vital signs, and 2 laboratory findings of death versus non-death (discharged). Group comparison of categorical variables in 3 frequencies and percentages used &#61539; 2 or Fisher exact tests. Group comparison of continuous variables in 4 medians and interquartile ranges (IQR) used the Mann-Whitney U test.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Patients, No. (%)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Death</ns0:cell><ns0:cell>Non-death</ns0:cell><ns0:cell>p value</ns0:cell></ns0:row><ns0:row><ns0:cell>(n=142)</ns0:cell><ns0:cell>(n=880)</ns0:cell><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)Manuscript to be reviewed</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Performance indices for predicting mortality of the testing dataset. Abbreviations: area under the curve (AUC), accuracy, sensitivity, specificity, precision, recall, negative predictive value (NPV), positive predictive value (PPV) and F1 score (a harmonic mean of precision and recall).</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>AUC</ns0:cell><ns0:cell cols='5'>Accuracy Sensitivity Specificity Precision NPV</ns0:cell><ns0:cell>PPV</ns0:cell><ns0:cell>F1</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Training 0.852</ns0:cell><ns0:cell>0.892</ns0:cell><ns0:cell>0.706</ns0:cell><ns0:cell>0.922</ns0:cell><ns0:cell>0.589</ns0:cell><ns0:cell>0.952</ns0:cell><ns0:cell>0.589</ns0:cell><ns0:cell>0.642</ns0:cell></ns0:row><ns0:row><ns0:cell>Testing</ns0:cell><ns0:cell>0.844</ns0:cell><ns0:cell>0.853</ns0:cell><ns0:cell>0.750</ns0:cell><ns0:cell>0.872</ns0:cell><ns0:cell>0.522</ns0:cell><ns0:cell>0.949</ns0:cell><ns0:cell>0.522</ns0:cell><ns0:cell>0.616</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51182:1:1:CHECK 18 Sep 2020)</ns0:note> </ns0:body> "

"Dear editor and reviewers. We thank you for your reviews of this manuscript. We believe we have carefully and positively addressed your comments. The point by point responses are provided below and major changes were colored red for your references in the manuscript. Thank you again for your thoughtful and constructive feedbacks. Editor comments (Tuan Nguyen) MAJOR REVISIONS Thank you for allowing us to consider your manuscript which is handled by me (Tuan Nguyen). Your manuscript has now been reviewed by 3 experts, and their comments are attached for your perusal. As you will see from their comments, although the reviewers recognize the relevance of your work, they raise some methodological issues that I would like to invite you to comment on. As Academic Editor, I have also read you manuscript with great interest. I thought that the work is quite clinically relevant, but it should clarify a number of points as follows: 1. The ranking of clinical variables was done using a statistical software. Could you explain (preferably in simple language) what were the criteria for ranking. Was it based on some kind of statistical significance, variance explained, or the probability of being in a model? This needs more detailed explanation. The following was added: Ranking of clinical variables of categorical or numerical values were made using the Boruta, a statistical software (Kursa & Rudnicki 2010). Boruta ranks feature importance using the Random Forest method. In this decision tree-based method, the quantitative measure of importance is the Gini feature of importance, which counts the times that a feature is used to split a node of a decision tree, statistically weighted by the number of instances the node splits. In the DNN model, the top predictors were those that demonstrated statistical significance using built-in statistical methods within the Boruta algorithm. 2. How do you define 'statistical significance'. In the presence of multiple tests of association, the 0.05 significance level is perhaps not a good threshold. In the DNN model, the top predictors were those that demonstrated statistical significance using built-in statistical methods within the Boruta algorithm. In the risk score model, SPSS was used to cross-check statistical significance of the top features. All of the top features used in the final analysis had a p < 0.001. 3. I am not clear on the way you remove both variables. You state that a correlation coefficient >0.5 from collinearity analysis was used to exclude correlated variables from machine learning analysis. Why the threshold of 0.5 was chosen? The following was added: A correlation coefficient >0.5 from collinearity analysis was used to exclude correlated variables from machine learning analysis. Note that none of the top features we used in the final analysis demonstrated strong correlation with other features. Therefore, no features were removed as a result. 4. How did you define mortality? 30-day in hospital? What about mortality outside hospital? The following was added: The primary outcome was ICU admission versus general floor admission, and the secondary outcome was in-hospital mortality versus discharge. Mortality outside of hospital after discharge was not obtained. 5. Could you distinguish between surgical or medical ICU admission? The ICU admission refers to medicine ICU admission. Surgical ICU admission is not applicable in COVID-19 as there is no established surgical protocol that has been proved helpful in treating COVID patients and thus surgery is not routinely done. 6. I am a bit confused of terminologies. Did you use machine learning or Generalized Additive Model? It seems to me that you used the latter, and it should be stated so in the manuscript. We apologize for the confusion. Machine learning models (i.e. Boruta, DNN) were used to rank the top features and make prediction of ICU admission and mortality, while a mixed generalized additive model was used for the risk score model. 7. For transparency, I suggest that you provide the actual odds ratio for each variable (with the outcome being ICU admission and mortality). The following was added: Odds ratios of procalcitonin, LDH, CRP, ferritin, and SpO2 for ICU admission were 3.062, 3.846, 3.001, 2.449, and 3.665, respectively. Odds ratios of age, LDH, procalcitonin, troponin, CRP, and SpO2 for mortality were 4.301, 3.418, 6.232, 5.253, 2.240, and 3.750, respectively. [# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.  It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter.  Directions on how to prepare a rebuttal letter can be found at: https://nam03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpeerj.com%2Fbenefits%2Facademic-rebuttal-letters%2F&amp;data=02%7C01%7CXiaoran.Li%40stonybrookmedicine.edu%7C38323ff92d524c6fcef908d84c1636de%7Ceafa1b31b194425db36656c215b7760c%7C0%7C0%7C637343005688251892&amp;sdata=yhrj%2FPVPbxS0jZYNI%2FUhHgJDO9Hui4KC6tRmnb%2Fnx5w%3D&amp;reserved=0 #] Reviewer 1 (Anonymous) Basic reporting See below Experimental design See below Validity of the findings See below Comments for the Author This retrospective study consisted of 5766 persons-under-investigation for COVID-19 between February 7, 2020, and May 4, 2020. Demographics, chronic comorbidities, vital signs, symptoms, and laboratory tests at admission were collected. A deep neural network model and a risk score model were constructed to predict ICU admission and mortality. Prediction performance used area under the curve (AUC) of the receiver operating characteristic analysis (ROC). The top ICU predictors were procalcitonin, lactate dehydrogenase, C-reactive protein, ferritin, and SpO2. The top mortality predictors were age, lactate dehydrogenase, procalcitonin, cardiac troponin, C reactive protein, and SpO2. The paper is well-written and contains interesting results. Minor comments include: 1) it would be good to make the code available for readers to reproduce the analysis; Codes will be uploaded to GitHub after publication. 2) Method: A deep neural network (DNN) was constructed to predict ICU admission and mortality using five fully connected dense layers classification. Could the authors elaborate on why you choose the structure with 5 layers? How many neurons are there in each layer? The following was added: A correlation coefficient >0.5 from collinearity analysis was used to exclude correlated variables from machine learning analysis. Note that none of the top features we used in the final analysis demonstrated strong correlation with other features. Thus, no top features were removed as a result. A deep neural network (DNN) was constructed to predict ICU admission and mortality using five fully connected dense layers (Chen et al. 2020). The top clinical predictors were input parameters, determined by testing subsets of these parameters, and ICU admission and mortality were outcome parameters. The DNN model used 5 hidden layers with 6, 8, 16, 8, 4 neurons respectively. We explored a few models using a range of number (3-7) of layers, and the 5-layer model yielded the optimal validation result. ReLu activation function for the hidden layers, the sigmoid activation function for the output layer, and the “he_normal” normalization scheme were applied. In the model training process, we used Adam optimizer, mean squared error as the cost function, a default learning rate of 0.01, and number of epochs of 100. The reported results yielded from the average of 5 consecutive runs. The dataset was randomly split into 90% training data and 10% testing data. ICU admission and mortality results were categorized using a binary classification. To minimize overfitting, we employed 5-fold cross-validation, ranked and removed less important features using correlation analysis and based on statistical significance by Boruta. We also employed regularization and stopped the training process at 100 epochs. 3) It would be essential to compare/discuss the findings with other using standard machine learning methods, such as (but not limited to) https://nam03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.researchsquare.com%2Farticle%2Frs-41151%2Fv1&amp;data=02%7C01%7CXiaoran.Li%40stonybrookmedicine.edu%7C38323ff92d524c6fcef908d84c1636de%7Ceafa1b31b194425db36656c215b7760c%7C0%7C0%7C637343005688251892&amp;sdata=8H6VTYGqKN4WF2ckOIWF6A3FedfxPPOrjQ69Qw0okew%3D&amp;reserved=0. Thank you. It is cited and discussed as followed: Yuan et al. went one step further to predict mortality more than 12 days in advance with more than 90% accuracy across all cohorts. Moreover, the Kaplan-Meier score showed that patients upon admission can clearly be differentiated into low, medium or high risk. They created a simple risk score system, and validated using multiple independent cohorts (Yuan et al. 2020). Reviewer 2 (Anonymous) Basic reporting The paper proposes a method for utilising deep neural nets (DNNs) in predicting ICU admission likelihood for COVID-19 patients based on critical parameters. Despite the task being interesting, there are some major concerns:- 1. The ethical/research compliance approval pertains to 'AI of lung images of COVID-19'. The authors should clarify whether their study utilises lung images, or some numeric datasets (as the attached files in the raw data are .CSV and seem to pertain to numeric/categorical critical parameters). Please clarify this in the paper whether this data is derived based on the lung images or otherwise. Binary imaging findings of whether consolidation was found or not in the lung CXR images were extracted from patient chart review, which were provided by radiology report as part of standard of care. We did not use the ML model to analyze lung CXR images. The presence of consolidation was not ranked as one of the top features and thus it was not part of the final analysis. 2. The title of the paper is ambiguous. 'Deep-learning artificial intelligence prediction...'. It is a well known factor that deep learning is a type of AI technique, so either use AI or deep learning in the title. Using both does not make clear sense for the context. We have removed “artificial intelligence” in our title. 3. It would be better to have the figures in-text instead of appendices, which makes it difficult for the average reader of this journal. We believe figures and tables will be placed in-text during typesetting at proof by the journal. We will ensure this to be done. 4. The authors mention that the top 10 variables were significant in predicting ICU admission. Has the significance been medically correlated/verified or is it based on the experimental analysis by the authors? It is based on our experimental analysis. It is not trivial to validate these variables in the clinical settings as the disease is still new and a prospective study will be needed. Further studies are needed to independently confirm on a larger and multi-institutional data. 5. There are several technical details from the point of view of AI missing in the paper- especially in terms of the DNN architecture utilised and its hyper-parameters, making the paper lack robustness. Please see response to Reviewer #1, question #2. Experimental design The paper proposes utilisation of a deep neural network (DNN) for predicting ICU admission and mortality. However, their is insufficient details on the network choice and other hyperparameters:- 6. What was the basis to utilise 5 hidden layers for the DNN? Were smaller network architectures utilised? Please see response to Reviewer #1, question #2. 7. The experiments use a 90% training dataset and 10% testing dataset for the study. With such a large amount of training data and experiments being only carried on the smaller testing data (10%) - were more conventional train test splits used e.g. 70-30%, 80-20% etc.? If not, please provide a reference on the choice of train-test split or more details on this choice, as it is a vital characteristic in AI experimentation. In our reading of the literature, 70-30%, 80-20% and 90-10% are well used, and 70-30% and 80-20% are used more often. We feel that we need more training data due to small sample size. We cited this reference http://ee104.stanford.edu/lectures/validation.pdf which used 90-10%. 8. It is mentioned that correlated variables were excluded from analysis (with corr. coff > 0.5), however, generally for DNNs, feature selection and dimensionality reduction are automatic and manual feature engineering is not required? Please justify the basis for identifying top predictors through this technique. There can be other methods e.g. XGBoost which can provide top predictors during classification, so a justification of choosing this methodology will help make this paper more robust. The reviewer is correct that feature selection and dimension reduction are automatic in DNN. However, a DNN model is lack of interpretability, given its huge parameter space and non-linear operations. Therefore, we utilized DNN to build a model only for the purpose of achieving more accurate prediction. Other methods such as XGBoost, can be used to rank feature importance. We actually used another tree-based method, random forests, to provide top predictors by Boruta. 9. Many relevant details are missing for the DNN which affect reproducibility and transparency in the results- what was the optimiser utilised, number of epochs over which the network was trained, activation function, learning rate.....? Please see response to Reviewer #1, question #2. Validity of the findings 1. The approach presents an evaluation of the DNN on the test dataset for COVID-19 patients in the US with an AUC. The confusion which is present is regarding utilisation of simpler ML architectures or DNNs with fewer hidden layers- were these verified and experimented with? If so, please provide more information in the text. Please see response to Reviewer #1, question #2. 10. What were the key performance metrics alongside the AUC? True/False Positives/Negatives, F Score obtained for classification using the DNN, accuracy/precision/recall et al.? A confusion matrix here would be useful to clearly make inferences from the model's performance. These values are now provided in the manuscript. The ROC and confusion matrix are shown in Figure 3. The performance of the DNN model yielded an AUC = 0.780 [95% CI:0.760-0.785], sensitivity = 0.760, specificity = 0.709, and F1 score = 0.551 in predicting ICU admission for the testing set. The ROC and confusion matrix are shown in Figure 6. The performance of the DNN model yielded an AUC = 0.844 [95% CI:0.839-0.848], sensitivity = 0.750, specificity = 0.872, and F1 score = 0.616 in predicting mortality for the testing set. Comments for the Author The paper proposes an interesting application of DNNs for a real-world (and very significant problem) at the present time of COVID-19. However, there are insufficient technical details on the AI part of this work, making replication/reproducibility a concern. In addition, the results in the paper need more details and justifications to merit publication in this journal. Reviewer 3 (Anonymous) Basic reporting The authors' writing is clear and easy to follow. However, there are some points could be improved: 1. The AUCs of models for mortality should be consistent in the abstract (0.848 and 0.845) and the text (0.882 and 0.048). Corrected. 2. Authors mentioned that a deep-learning artificial intelligence algorithm is novel and has distinct advantages over logistic regression and supervised learning approach (line 237-238, 72-73) but did not give in any details of what advantages they are. The following is added: Our approach used a deep-learning algorithm which is novel and has distinct advantages over logistic regression and supervised learning approach. Deep learning is increasingly being used in medicine (Deo 2015; Santos et al. 2019; Tschandl et al. 2019). In contrast to conventional analysis methods, which specify the relationships amongst data elements to outcomes, machine learning employs computer algorithms to identify relationships amongst different data elements to inform outcomes without the need to specify such relationships a priori. Deep learning can outperform human experts in performing many tasks in medicine (Killock 2020). In addition to approximating physician skills, Deep learning can also detect novel relationships not readily apparent to human perception, especially in large, complex, and longitudinal datasets. Disadvantages of deep learning methods are that it requires comparatively large sample size, there is a potential of overfitting, and the complex relations could make deep learning results difficult to interpret, amongst others. In addition, we devised a simplified practical risk score adds practical utility to these findings. Although we ranked all variables and explicitly listed 10 or 15 top variables, we built the predictive model and risk score model using only the top 5 variables to simplify and increase translation potential in the clinical settings. The excellent prediction performances using a few clinical variables are encouraging. 3. Figure 2 &4: More information should be provided: What values shown in the x-axis, which method used to retrieve these values, what information indicated by different colour codes? The following was revised Figure 2. Ranking of clinical variables for predicting ICU admission by Boruta algorithm. The x-axis is attribute of level of importance, where a larger number indicates relatively higher importance. The y-axis are laboratory test variables. The top statistically significant predictors were: procalcitonin, LDH, CRP, ferritin, SpO2, lymphocytes, respiratory rate, systolic blood pressure, age and ALT. The top 10 variables were significant. 4. Figure 3: Legend indicates that scores ranged from 0 to 6, while the figure represents score range from 0 to 5. Which information do the numbers in each bar indicate? Figure 4. Risk score stratification for ICU admission. Scores ranged from 0 to 5, with 0 indicating the lowest risk and 5 being the highest risk of mortality. The numbers in the bar indicate the number of patients in the ICU (red) and non-ICU (blue) that were correctly predicted in the testing dataset. 5. Table 1: Race values should be aligned. There are two bold p values without any indication. Lack of explanation for the WBC abbreviation. Done 6. Table 2: Race values should be aligned. Lack of explanation for the WBC abbreviation. Done 7. Line 145: p values should be included. 8. Line 209: lack of references. 9. Minor typo error: line 223: identified f clinical. All errors are corrected. Thank you. Experimental design The research question is well-fined. The study settings is also clearly described. However, details of methods used in the study need to be provided: 10. Authors should explain in the text why Figure 2 mentions that top 10 variables were significant, but the models used only top 5 variables. Similarly, top 15 significant variables were mentioned in Figure 4 while the models used only top 6. If they were results from the collinearity analysis, please give details. The following was added: Although we ranked all variables and explicitly listed 10 or 15 top variables, we built the predictive model and risk score model using only the top 5 variables to simplify and increase translation potential in the clinical settings. The excellent prediction performances using a few clinical variables are encouraging. 11. What did authors do to avoid the overfitting issue? The following is added: To minimize overfitting, we employed 5-fold cross-validation, ranked and removed less important features using correlation analysis and based on statistical significance by Boruta. We also employed regularization and stopped the training process at 100 epochs. 12. Although authors cited several references, all algorithms and parameter tuning strategies used in the study should be briefly described in the text or supplemental document to be useful for future reader. There is lack of references for the collinearity analysis. Please see responses to Reviewer #1, Question #2 Validity of the findings Although your results are compelling, the data analysis should be improved in the following ways: 12. Provide ROC curves for training and test data sets. We have included the ROC curves in the revised manuscript (Figure 3 and 6). 13. Report performance of the models using other metrics, including sensitivity, specificity, accuracy, precision, and negative predictive value at the optional cut-off selected in training dataset because “change in AUC has little direct clinical meaning for clinicians” (Halligan et al, Eur Radiol, 2015, V25, p 932–939). These values need to be reported in both training and test set to see how balance they are. The following tables are provided. Table 3. Performance indices for predicting ICU admission of the testing dataset. Abbreviations: area under the curve (AUC), accuracy, sensitivity, specificity, precision, recall, negative predictive value (NPV), positive predictive value (PPV) and F1 score (a harmonic mean of precision and recall). AUC Accuracy Sensitivity Specificity Precision NPV PPV F1 Training 0.751 0.703 0.707 0.701 0.437 0.879 0.437 0.540 Testing 0.728 0.721 0.760 0.709 0.432 0.910 0.432 0.551 Table 4. Performance indices for predicting mortality of the testing dataset. Abbreviations: area under the curve (AUC), accuracy, sensitivity, specificity, precision, recall, negative predictive value (NPV), positive predictive value (PPV) and F1 score (a harmonic mean of precision and recall). AUC Accuracy Sensitivity Specificity Precision NPV PPV F1 Training 0.852 0.892 0.706 0.922 0.589 0.952 0.589 0.642 Testing 0.844 0.853 0.750 0.872 0.522 0.949 0.522 0.616 14. Compare your DNN model with models using logistic regression/supervised learning approach algorithms on your data to show the advantages. We did not compare our results using logistic regression as they are not trivial. We agree that these comparisons need to be made perhaps with a large dataset in future studies. We hope that the urgency of covid19 findings warrant publication. 15. To help the clinician visually interpret how the predictive score was generated by the deep-learning model, a monogram is recommended because DNN model is too complicated to interpret in practical settings. We generated the risk score system which is similar to monogram. 16. The simplified models use a cut-point for each top predictor. The underlying rational of these cut-points needs to be discussed in the discussion section. The following was added: Different cutoff points were evaluated where the chosen cutoff points yielded the optimal distribution (not skewed to high or low scores) of the risk score model. The corresponding numerical values of each top feature at probability of 0.3 for ICU and 0.2 for mortality were found to be the optimal cutoff values for the risk score model. Each of the top variables was assigned a weight of one point if the clinical measurement was above the probability cutoff. The risk score ranged from 0 to 5 for ICU admission and 0 to 6 for mortality (which were chosen based on statistical significance, see Results). Comments for the Author Xiaoran Li and colleagues conducted a retrospective study to develop a deep learning model and a risk score model to predict ICU admission and mortality in 1108 hospitalized COVID-19 positive patients using their clinical information. The models were built in 90% of data set and tested in the 10% left. They found that top 5 ICU predictors included procalcitonin, lactate dehydrogenase, C-reactive protein, ferritin, and SpO2, and top 6 mortality predictors included age, lactate dehydrogenase, procalcitonin, cardiac troponin, C-reactive protein, and SpO2. AUCs of the deep- learning model in test dataset were 0.780 for ICU admission and 0.882 for mortality. The simplified risk score model had slightly lower performance, yielding the AUC of 0.728 and 0.848, respectively. The findings of key predictors are not new as they have been reported elsewhere. The models need to be validated in external cohort before practical use. However, the models contribute to the growing need of quantitative tools to stratify hospitalized COVID-19 patients having need of ICU admission and high risk of mortality based on their clinical information. Although the analysis is interesting and meaningful at the population level, there are still some extra work need to be done. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Cervical squamous cancer (CESC) is an intractable gynecological malignancy because of its high mortality rate and difficulty in early diagnose. Several biomarkers have been found to predict the prognose of CESC using bioinformatics methods, but they are still lack of clinical effectiveness. Most of the existing bioinformatic studies only focus on the changes of oncogenes but neglect the differences on protein level and molecular biology validation are rarely conducted. Methods. Gene set data from NCBI-GEO database were used in this study to compare the differences of gene and protein levels between normal and cancer tissues through significant pathway selection and core gene signature analysis to screen potential clinical biomarkers of CESC.</ns0:p><ns0:p>Subsequently, the molecular and protein levels of clinical samples were verified by quantitative transcription PCR, Western Blot and immunohistochemistry. Results. Three differentially expressed genes (RFC4, MCM2, TOP2A) were found to have a significant survival (P&lt;0.05) and highly expressed in CESC tissues. Molecular biological verification using quantitative reverse transcribed PCR, western blotting and immunohistochemistry assays exhibited significant differences in the expression of RFC4 between CESC and paracancerous tissues (P&lt;0.05). Conclusion. This study identified three potential biomarkers (RFC4, MCM2, TOP2A) of CESC which may be useful to clarify the underlying mechanisms of CESC and predict the prognosis of CESC patients.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Cervical cancer now ranks fourth in the most prevalent cancers and it's the most common gynecological cancer in developing countries <ns0:ref type='bibr' target='#b33'>(Vu et al., 2018)</ns0:ref>. Despite the increase in the incidence of cervical adenocarcinoma, cervical squamous carcinoma (CESC) still stays as the commonest pathological type of cervical cancer <ns0:ref type='bibr' target='#b34'>(Wang et al., 2004;</ns0:ref><ns0:ref type='bibr' target='#b15'>Galic et al., 2012)</ns0:ref>. Currently, hundreds of gene mutations have been proved to be responsible for increase in the incidence of cervical cancer, which can be used as biomarkers for early detection, like DNA mutations occurring on the oncogenes tumor protein 53 (TP53) <ns0:ref type='bibr' target='#b10'>(Crook et al., 1992)</ns0:ref>, phosphatase and tensin homolog (PTEN) <ns0:ref type='bibr' target='#b0'>(2014)</ns0:ref>. However, the overall survival of CESC patients still stays weak due to the difficulties in early detection and treatment. Studies also showed that some biological markers can explain the pathogenesis of CESC and predict the consequences of this disease <ns0:ref type='bibr' target='#b24'>(Mao et al., 2019)</ns0:ref>. Therefore, more reliable biological markers should be explored to comprehensively understanding the pathogenesis of CESC and guide the treatment and prognosis of that.</ns0:p><ns0:p>With the developed bioinformatics and statistical analyses, the potential marker genes can be detected effectively, which shows great strength in the field of discovery and prediction of tumor markers, and plays a guiding role in the treatment and prognosis of the disease <ns0:ref type='bibr' target='#b4'>(Banwait &amp; Bastola, 2015)</ns0:ref>. Some biomarkers have been found in the field of cervical cancer, such as MicoRNA-425-5p and MicoRNA-489, which have been proposed for prognostic prediction <ns0:ref type='bibr' target='#b30'>(Sun et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b18'>Juan et al., 2018)</ns0:ref>.</ns0:p><ns0:p>However, the presented biomarkers for clinical application are far from enough, and in the previous bioinformatics studies, most studies only focus on the changes of oncogenes, which increases the possibility of clinical inefficacy. On the basis of learning the expression of differential genes between cancer tissues and normal tissues, this study analyzed and compared the difference in protein level between cancer tissue and normal tissue, which provides stronger evidence for the validity of biomarkers found in our bioinformatic research.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>Information of the microarray data NCBI-GEO (Gene Expression Omnibus) is known as a free public database of microarray cohort. The gene profiles of GSE27678, GSE39001 and GSE7803 were obtained in this study.</ns0:p><ns0:p>The three datasets were on the account of GPL570 platform, GPL201 platform and GPL96 platform, including 14 normal cervical tissues and 28 CESC tissues, 12 normal cervical tissues and 43 CESC tissues, 10 normal cervical tissues and 21 CESC tissues, respectively.</ns0:p></ns0:div> <ns0:div><ns0:head>Identification of differentially expressed genes</ns0:head><ns0:p>The differentially expressed genes (DEGs) were analyzed by GEO2Rto obtain the number of up-down-regulated genes <ns0:ref type='bibr' target='#b6'>(Barrett et al., 2013)</ns0:ref>. The genes with |log Fold Change|&#8805;2 and P &lt; 0. 05 were screened as differentially expressed genes.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway analyses</ns0:head><ns0:p>Gene Ontology (GO) is an international standardized classification system of gene function, which provides a dynamic updating database to describe the attributes of genes and gene products in organisms <ns0:ref type='bibr' target='#b2'>(Ashburner et al., 2000)</ns0:ref>. The main biological functions of differentially expressed genes could be determined by GO functional significance enrichment analysis. The GO items with q &lt; 0. 05 were considered to be significantly enriched in DEGs.</ns0:p><ns0:p>Kyoto Encyclopedia of Genes and Genomes (KEGG) is a bioinformatics resource for linking genomes to life and the environment <ns0:ref type='bibr' target='#b17'>(Goto et al., 1997)</ns0:ref>. Based on KEGG database, the enriched pathway analysis of DEGs was carried out to find out the important pathway.</ns0:p></ns0:div> <ns0:div><ns0:head>PPI &amp; module analysis</ns0:head><ns0:p>Cytoscape 3.8.0 is a software that was used for visualization and analyzation of complex network <ns0:ref type='bibr' target='#b28'>(Shannon et al., 2003)</ns0:ref>. Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) is an application that could conduct protein interaction group research, genome research and proteome research <ns0:ref type='bibr' target='#b11'>(Doncheva et al., 2019)</ns0:ref>. By mapping the information of DEGs to the STRING, we evaluated the protein-protein interaction (PPI) information of DEGs.</ns0:p><ns0:p>Interactions experimentally validated with combined score &gt; 0.4 and were selected.</ns0:p><ns0:p>Subsequently, we used another tool embedded in the Cytoscape named Molecular Complex Detection (MCODE) to clustering constructed functional module of PPI network <ns0:ref type='bibr' target='#b3'>(Bader &amp; Hogue, 2003)</ns0:ref>. The MCODE scores were set to be greater than 10 and nodes number more than 6.</ns0:p><ns0:p>Functional and pathway enrichment for DEGs in the modules were also conducted, P&lt;0.05 was considered to have significant difference.</ns0:p></ns0:div> <ns0:div><ns0:head>Survival analysis of significant genes in CESC and RNA expression of core genes</ns0:head><ns0:p>Kaplan-Meier (K-M) is a widely used method for estimating the survival rate of cancer patients and 'Survival' package was applied in the R studio software <ns0:ref type='bibr' target='#b27'>(Rich et al., 2010)</ns0:ref>. As for the compare of the magnitude of the difference in survival between the 2 groups, a Cox univariate hazard ratio (HR) was calculated. The clinical significance of each genes was also evaluated by performing the survival analysis of single gene in survival-related gene sets. A logrank test was used to calculate the statistical significance of the survival difference between these 2 groups mentioned above, and the P value set as 0.05 was considered to be significant.</ns0:p><ns0:p>Gene Expression Profiling Interactive Analysis (GEPIA) is visualization tool for gene research <ns0:ref type='bibr' target='#b32'>(Tang et al., 2017)</ns0:ref>. In this study, GEPIA was applied to analyze RNA expression of selected genes on the basis of thousands of simples from the TCGA database.</ns0:p></ns0:div> <ns0:div><ns0:head>Specimen collection</ns0:head><ns0:p>The tissues or cells of CESC patients were collected from Xiangya Hospital of Central South University in order to verify the high expression of RFC4 in tumor tissues for molecular and protein levels. This study was proved by Medical Ethics Committee of Xiangya Hospital <ns0:ref type='bibr'>(No.201912542)</ns0:ref>. CESC Patients and the kin have signed a consent form, agreeing to use cervical tissue for scientific research.</ns0:p></ns0:div> <ns0:div><ns0:head>Molecular biological verification of differences in gene expression</ns0:head><ns0:p>CESC tissues and para-cancerous tissues (para-CT) were selected from CESC patients to conduct the molecular validation of RFC4. The expression levels of RFC in CESC patients with different pathological stages were also compared. The pathological stage of &#8544; and &#8545; are PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed regarded as early stage which including 4 &#8544;B1 patients, 7 &#8544;B2 patients, 3 &#8544;B3 patients, 3 &#8545;A1 patients and 1 &#8545;A2 patient. Stage &#8546; are divided into advanced stage and 17 patients in &#8546;C1 stage were included. Total RNA was extracted from CESC tissues and para-CT using Trizol Reagent (RNAiso Plus, TaKaRa, 9109) according to the manufacturer's protocols, and reverse transcribed into cDNA using a PrimeScript&#8482; RT reagent Kit with gDNA Eraser (TaKaRa, RR047A-1). Gene expression levels were assessed by quantitative reverse transcribed PCR (qRT-PCR) with TB Green&#8482; Premix (Tli RNaseH Plus, TaKaRa, RR820A) and specific primers:</ns0:p><ns0:p>RFC4 forward: 5'-GGCAGCTTTAAGACGTACCATGG-3'; RFC4 reverse: 5'-TCTGACAGAGGCTTGAAGCGGA-3'.</ns0:p><ns0:p>The &#946;-actin expression was used as the normalization control. Relative mRNA levels are analyzed using 2 -&#916;&#916;Ct method.</ns0:p></ns0:div> <ns0:div><ns0:head>Verification of differences in protein expression</ns0:head><ns0:p>We adopted the cancerous tissues and para-CT of CESC patients to analyze the differences in protein expression by Western Blotting (WB) technology. The samples for WB analysis was separated using SDS-PAGE and transferred onto a PVDF membrane (Roche) which was blocked with 5% nonfat milk in Tris-buffered saline and incubated overnight at 4&#176;C with target antibodies against the following proteins: Anti-RFC4 antibody (ab156780, Abcam) and Anti-&#946;-Actin antibody (ab115777, Abcam). After three times washing with PBST (10 min for each time), the membrane was incubated with species-appropriate HRP-conjugated secondary antibodies, the fluorescent signals were detected using SageCapture&#8482; imaging system (SAGECREATION company).</ns0:p><ns0:p>Immunohistochemistry (IHC) assays were also performed to detected protein levels in CESC tissues and para-CT. The tissues were performed into 5-&#956;m-thick tissue sections with formalin fixed and paraffin embedded. Subsequently, there sections were deparaffinized and rehydrated with xylene and graded ethanol respectively, followed by heated in antigen retrieval PeerJ reviewing <ns0:ref type='table' target='#tab_0'>PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)</ns0:ref> Manuscript to be reviewed solution (EDTA, PH 9.0) and endogenous peroxidase inactivation with 3% H 2 O 2 . After blocking, the samples were incubated overnight at 4&#176;C with anti-RFC4 antibody (1:100, ab156780, Abcam). Then the slides were treated with the HRP-conjugated secondary antibody and stained with 3, 3'-diaminobenzidine until brown granules appeared in the membrane, cytoplasm, or nucleus. Finally, the sections were counterstained with hematoxylin at room temperature.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Screening for DEGs</ns0:head><ns0:p>Ninety-two cancer tissues and 36 normal tissues were selected from the three datasets in total, with the help of GEO2R tools, 211, 134 and 260 DEGs were extracted from GSE39001, GSE7803 and GSE27678. And Venn diagram was made by the Venn diagram software to investigate the commonly DEGs in all the three datasets. The results showed that there were 25 commonly DEGs in total and 18 of them were down-regulated while 7 were up-regulated (Figure <ns0:ref type='figure' target='#fig_1'>1</ns0:ref> and Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Significant pathways identified in CESC</ns0:head><ns0:p>We investigated upregulated and downregulated DEGs to identify the most significantly enriched pathways in each group by GO and KEGG pathway analysis (Figure <ns0:ref type='figure'>2 and Table2</ns0:ref>).</ns0:p><ns0:p>With GO analyzing, the results indicated that 1) for biology processes (BP) , the most Manuscript to be reviewed microtubule and kinesin complex.</ns0:p><ns0:p>The results of KEGG analysis demonstrated that the most significant signaling pathways of DEGs were cell cycle, pathways in cancer, ECM-receptor interaction, arrhythmogenic right ventricular cardiomyopathy (ARVC), melanoma, PI3K-Akt signaling pathway, focal adhesion, vascular smooth muscle contraction, DNA replication and oocyte meiosis (Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Systematic analysis of core genes by PPI network</ns0:head><ns0:p>PPI network investigated the systematic interaction between the DEGs we got above.</ns0:p><ns0:p>Twenty-five DEGs in total were mapped to the DEGs PPI network with 99 nodes and 270 edges.</ns0:p><ns0:p>Seven up-regulated DEGs and 18 down-regulated DEGs were included in the PPI network. And then Cytotype MCODE was applied for further analysis of the DEGs in PPI network, and we got a result of 15 particular nodes being identified which were all up-regulated DEGs (Figure <ns0:ref type='figure'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Analysis of core gene signature in CESC using K-M plotter and GEPIA</ns0:head><ns0:p>To investigate the survival data of the 15 genes we identified, K-M plotter indicated that three (TOP2A, RFC4, MCM2) of them had a significant survival rate while other 12 genes had not (P&gt;0.05) (Figure <ns0:ref type='figure'>4 and Table4</ns0:ref>). The expression of TOP2A, RFC4, MCM2 in normal tissue and CESC tissue was detected by GEPIA. The results showed that the expression of these three genes in CESC tissue was higher than that in normal tissue (P&lt;0.05) (Figure <ns0:ref type='figure'>5</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>RFC4 is validated to be overexpressed in CESC</ns0:head><ns0:p>By analyzing the data from the NCBI-GEO dataspace for mRNA expression in CESC patients, RFC4 gene was identified as an overexpressed gene in CESC patients. We collected 35 pairs of CESC patients for qPCR, the tissues of 6 pairs CESC patients were used for WB, 9 pairs CESC tissues and 4 normal cervical tissues for IHC. In order to validate our finding, total RNA was extracted from 35 paired CESC tissues and para-CT tissues, and qRT-PCR was conducted to measure the expression level of RFC4 gene. The result showed that the expression level of RFC4 on CESC tissues was significantly high compared with para-CT (P=0.0197) (Figure <ns0:ref type='figure'>6</ns0:ref>). And the expression of RFC4 in early stage CESC was significantly higher than that in advanced CESC (P=0.0314) (Figure <ns0:ref type='figure'>7</ns0:ref>). The same result was invested from WB. The results of WB analysis PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed indicated that the RFC4 was overexpressed in CESC tissues compared to para-CT tissues (Figure <ns0:ref type='figure' target='#fig_2'>8</ns0:ref>). A higher level of RFC4 expression on CESC tissues was observed from the result of IHC, and RFC4 protein was mainly concentrated in the nucleus (Figure <ns0:ref type='figure'>9</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In order to identify more effective prognostic biomarkers in CESC, we used different bioinformatics methods to analyze three data sets based on NCBI-GEO database, including 92 CESC tissues and 36 normal tissues. A total of 25 DEGs were selected by GEO2R and Venn software, including 7 up-regulated genes and 18 down-regulated genes. Then GO and KEGG pathway analysis were conducted, and the results of GO and KEGG indicated that the selected DEGs were significantly enriched in various cell pathways. Researches had reported that some genes from these pathways were associated with the pathogenesis and progression of cervical carcinoma. Nucleolar and spindle associated protein 1(NUSAP1) was a gene from spindle associated pathway, and it was reported to promote the metastasis of cervical cancer by activating Wnt/&#946;-catenin signaling <ns0:ref type='bibr' target='#b23'>(Li et al., 2019)</ns0:ref>. And studies showed that CXCL12/CXCR4 pathways was associated with HPV infection as a co-factor, which means a high risk to the incidence of cervical cancer <ns0:ref type='bibr' target='#b25'>(Meuris et al., 2016)</ns0:ref>. Genes involved epidermis development were also associated with the high-risk HPV infection <ns0:ref type='bibr' target='#b35'>(Zhang et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b9'>Chatterjee et al., 2019)</ns0:ref>.</ns0:p><ns0:p>After that PPI network was constructed using STRING software and MCODE analysis was conducted, and 15 particular DEGs were identified. Furthermore, by K-M plotter analysis we got 3 DEGs from the 15 which had a significantly better survival. The results of GEPIA showed that the expression levels of the 3 selected genes in CESC tissues were higher than that in normal tissues. To further validation, we performed RFC4 relevant molecule biological experiments and the results showed that compared with normal tissues, RFC4 was highly expressed in CESC tissues.</ns0:p><ns0:p>Being short for Replicant Factor C, RFC is a structure specific DNA-binding protein acting as a primer recognition factor for DNA polymerase <ns0:ref type='bibr' target='#b36'>(Zhou &amp; Hingorani, 2012)</ns0:ref>, which includes 5 subunits (RFC1-5). Among all 5 subunits of RFC complex, RFC4 has been reported to play an important role in DNA damage checkpoint and DNA replication pathways <ns0:ref type='bibr' target='#b12'>(Ellison &amp; Stillman, 2003)</ns0:ref>. In 2009, Arai M et al. reported that RFC4 was closely related to the prognosis of liver cancer <ns0:ref type='bibr' target='#b1'>(Arai et al., 2009)</ns0:ref>. Besides liver cancer, RFC4 has been reported to be associated with several types of cancer, including prostate cancer, colon cancer non-small cell lung cancer and leukemia <ns0:ref type='bibr' target='#b21'>(LaTulippe et al., 2002;</ns0:ref><ns0:ref type='bibr' target='#b19'>Jung et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b13'>Erdogan et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b5'>Barfeld et al., 2014)</ns0:ref>.</ns0:p><ns0:p>There were researches illustrated that up-regulated RFC4 expression was found in neck squamous cell carcinoma and it was 3.4-fold higher than that in normal tissues <ns0:ref type='bibr' target='#b29'>(Slebos et al., 2006)</ns0:ref>. Studies from <ns0:ref type='bibr'>Garnett et al.</ns0:ref> showed that RFC4 can be regulated by mutated RB1 in several types of cancers, suggesting that RFC4 could be a potential biomarker associated with the occurrence and prognosis of various cancers <ns0:ref type='bibr' target='#b16'>(Garnett et al., 2012)</ns0:ref>. Moreover&#65292;RFC4 was reported as an independent predictor of overall survival in breast cancer <ns0:ref type='bibr' target='#b14'>(Fatima et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b26'>Niu et al., 2017)</ns0:ref>.</ns0:p><ns0:p>In this study we observed RFC4 as a potential independent prognostic biomarker in CESC, and our results suggested that CESC patients with higher expression level of RFC4 may have a better overall survival. A possible reason might be that RFC4 was highly expressed throughout the cell circle process of proliferating cells, and tumor proliferation in situ will become slow with the development of the disease <ns0:ref type='bibr' target='#b31'>(Szymanska et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chaplain &amp; Sleeman, 1993)</ns0:ref>, which means a decrease in the expression of RFC4. Therefore, highly expressed RFC4 may suggest early stage CESC, which indicates better overall survival.</ns0:p><ns0:p>Several studies have proved that these three genes were associated with numerous types of cancer, but studies of RFC4 in CESC were rarely seen, and very few researches conducted molecule biology validation. Therefore, our study shows that RFC4 is a potential biomarker for the predicting the prognosis of CESC and provides a direction for further study of CESC. What should be noted is that there are some limitations in this study. Clinical samples from one hospital may have either region or race difference. The expression level of RFC4 in different stages of CESC and clinical investigations should be conducted in our future study to validate our results further.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In conclusion, by using bioinformatics analysis we identified three genes (TOP2A, RFC4, MCM2) based on three microarray datasets. These three genes were suggested to have a significant effect on the prognosis of CESC, which could be key factors in the occurrence and progression of CESC. And a high level expressed RFC4 was validated to be existed in CESC tissues using clinical samples. Although further investigation and experiments need conducting, the findings in our study could act as clinical biomarkers which would be helpful for us to better understand the pathological process and predict the prognostic of CESC. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>significantly enriched pathways of the DEGs were epidermis development, positive regulation of cell proliferation, peptide cross-linking, regulation of cell proliferation, positive regulation of cellular process, epidermal cell differentiation, skin development, keratinocyte differentiation, positive regulation of nuclear division, positive regulation of mitotic nuclear division; 2) for molecular function (MF), they were chemokine activity, chemokine receptor binding, calcium ion binding, collagen binding, CXCR chemokine receptor binding, growth factor activity, intergrin binding, cytokine activity, peptidase activity, acting on L-amino acid peptides, CCR chemokine receptor binding; 3) for cell component (CC), DEGs were significantly enriched in spindle, intercalated disc, intermediate filament, mitotic spindle, nuclear chromosome part, spindle midzone, condensed chromosome kinetochore, platelet alpha granule lumen, spindle PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 8 WB</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='17,42.52,70.87,326.28,672.95' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,275.62,525.00,273.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,70.87,426.20,672.95' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,229.87,525.00,279.75' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,70.87,373.92,672.95' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='27,42.52,70.87,525.00,552.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='29,42.52,70.87,360.54,672.95' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 . GO analysis of different expressed genes in CESC</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Expression</ns0:cell><ns0:cell>Category</ns0:cell><ns0:cell>Term</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020) Manuscript to be reviewed PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)Manuscript to be reviewed</ns0:note></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50933:1:0:NEW 2 Oct 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Reviewer 1 Response to 'Basic reporting' (1): Thanks for your kind suggestions. We have changed aspect ratio of Figure 1, and the Figure 2 is also revised. All the figures with low resolution are remade and improved in terms of pixels. Column table 1 has also been modified according to your suggestions. Thanks again. Response to 'Basic reporting' (2): Thanks for your comments. We have added our citations and made updates in the material and methods of the original text. Response to 'major comments' (1) of 'Experimental design': Thanks for your suggestions. We have to admitted that there may be a lack of innovation in our analysis methods, but we only regard it as a tool. The innovation of our article lies in the use of existing bioinformatics methods to find that RFC4 could be a potential biomarker of CESC and we performed validation process using clinical specimens. Response to 'major comments' (2) of 'Experimental design': Thanks for your precious comments. The tool we used here is GEO2R instead of DEGseq. I’m sorry for not modifying this in time when we submitted our manuscript. And we have revied the citation and corrected it in the original text (line 96-98). Thanks again. Response to 'minor issues' (1) of 'Experimental design': Thanks for your comments. We have adopted your advice and performed further validation. The result showed that the expression of RFC4 in early stage CESC was significantly higher than that in advanced CESC (P=0.0314) (Figure 7) (line145-149, 235-237). However, because of some limitations of the clinical specimens such as the lack of amount, there still needs more simples to validate this further. Response to 'Comments for the author': Thank you for providing us your kind comments, we have checked and revised our paper carefully based on your suggestions, and all the modifies have been marked in the manuscript. Because of the outbreak of COVID-19, further work may need to be done by us in the future. We hope you could review our revised manuscript again and look forward to the early publication of the article. Reviewer 2 Response to 'Basic reporting': Thanks for your kind suggestions to our manuscript. We have read your comments and revised our original text carefully. We have also checked our linguistic issues and revisions have been made. Response to 'Experimental design' (1): Thanks for your comments. The 'significant genes signatures' here means differentially expressed genes which associated with the prognosis of CESC patients and had the potential to be biomarkers to predict the overall survival of CESC patients. And the data from GEO is not original data, however, the data of the chip is reliable. In fact, we performed the analysis process on the basis of the bioinformatic related data, and the title has been changed to 'Identification of Significant Genes Signatures and Prognostic Biomarkers in Cervical Squamous Carcinoma via Bioinformatic Data'. Response to 'Experimental design' (2): Thanks. In this study, we discretized the continuous gene expression by sorting the continuous gene expression and taking intermediate values, and those with gene expression below the middle value were classified as 'low' while those gene expression above the middle value were divided into 'high'. The K-M survival analysis we conducted in our manuscript was based on the Cox PH model. We know that the time depend survival analysis would be needed if we didn’t calculate hazards radio based on Cox PH model. Thanks again. Response to 'Experimental design' (3): Thanks for your kind suggestions. We have not carried out a complete prognosis follow-up in this study. What we have done was to use the survival data already existed of CESC patients to find potential prognostic biomarkers. The findings of our study were based on the follow-up data of several datasets, which improved the reliability. And some patients have just undergone the surgery so our data is on the process on fellow-up. We have already considered about this and the investigation and improvement of patient follow-up data will be further completed in the future. Response to 'Experimental design' (4): Thanks for your precious comments. We have deleted the standard in the results section. And revision has been made in the original text (line 188-189). Response to 'Experimental design' (5): Thanks for your comments. We have explained the enrichment analysis and it has been updated in the original manuscript (line 249-257). Thanks again. Response to the 'Validity of the findings': Thanks for your kind suggestions. What we have done in this study was that we reviewed a lot of literature, and integrated analysis was also performed based on the bioinformatics data. Clinical specimen validation was also conducted on gene and protein levels while most of the existed bioinformatic research of RFC4 only focused on the gene level and clinical validation was rarely seen. And it’s the first time that we proposed the potential relationship between RFC4 and the prognosis of CESC and verified it with clinical samples. We used bioinformatic methods in this study which reduced the cost and improved the efficiency of scientific research. Because of the COVID-19, there may some work need to be done in the future. We would appreciate it if you could review our manuscript again and we are looking forward to the early publication of this article. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Multiple comparisons tests (MCTs) include the statistical tests used to compare groups (treatments) often following a significant effect reported in one of many types of linear models. Due to a variety of data and statistical considerations, several dozen MCTs have been developed over the decades, with tests ranging from very similar to each other to very different from each other. Many scientific disciplines use MCTs, including &gt;40,000 reports of their use in ecological journals in the last 60 years. Despite the ubiquity and utility of MCTs, several issues remain in terms of their correct use and reporting. In this study, we evaluated 17 different MCTs. We first reviewed the published literature for recommendations on their correct use. Second, we created a simulation that evaluated the performance of nine common MCTs. The tests examined in the simulation were those that often overlapped in usage, meaning the selection of the test based on fit to the data is not unique and that the simulations could inform the selection of one or more test when a researcher has choices. Based on the literature review and recommendations, planned comparisons are overwhelmingly recommended over unplanned comparisons, for planned non-parametric comparisons the Mann-Whitney-Wilcoxon U test is recommended, Scheff&#233;'s S test is recommended for any linear combination of (unplanned) means, Tukey's HSD and the Bonferroni or the Dunn-Sidak tests are recommended for pairwise comparisons of groups, and that may other test exists for particular types of data. All code</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Many of the popular and robust statistical techniques used in data analyses estimate group (or treatment or factor level) means. Data analyses are crowded with factors of interest from experiments and observations in which different groups show different effects and responsesand these significant results are what progress scientific knowledge. Models for evaluating the existence of differences among means include a wide range of linear models. The classic ANOVA (ANalysis Of Variance) is a general linear model that has been in use for over 100 years <ns0:ref type='bibr'>(Fisher 1918)</ns0:ref> and is often used when categorical or factor data need to be analyzed.</ns0:p><ns0:p>However, an ANOVA will only produce an F-statistic (and associated p-value) for the whole model. In other words, an ANOVA reports whether one or more significant differences among group levels exist, but it does not provide any information about specific group means compared to each other. Additionally, it is possible that group differences exist that ANOVA does not detect. For both of these reasons, a strong and defensible statistical method to compare groups is nearly a requirement for anyone analyzing data.</ns0:p><ns0:p>The lack of specifically being able to compare group means with ANOVA has long been known and a sub-field of multiple comparisons tests (MCTs) began to develop by the middle of the 20th century <ns0:ref type='bibr'>(Harter 1980)</ns0:ref>. Of course, when the analysis only includes two groups (as in a ttest), then a significant result from the model is consistent with a difference between groups.</ns0:p><ns0:p>However useful this approach may be, it is obviously very limiting and as Zar (2010) states:</ns0:p><ns0:p>'employing a series of two-sample tests to address a multisample hypothesis, is invalid.' What has developed over the last several decades has been a bounty of statistical procedures that can be applied to the evaluation of multiple comparisons. On the surface, this long list of options for MCTs is a good thing for researchers and data analysts; however, all the tests are unique, and</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed some are better suited to different data sets or circumstances. Put another way, some MCTs are questionable or invalid when applied to certain experimental designs and data sets. Additionally, because MCTs are applicable to general linear models, generalized linear models, hierarchical models, and other models, both the data and the model need to be considered when selecting an MCT.</ns0:p><ns0:p>Because a large proportion of scientists that use MCTs are not statisticians or otherwise versed in the details and nuance that inform their application, there are numerous cases where MCTs are used incorrectly. For example, <ns0:ref type='bibr'>Ruxton and Beauchamp (2008)</ns0:ref> reviewed 12 issues of Behavioral Ecology and reported on 70 papers that employed some type of multiple comparisons testing (or homogeneity of means, as they report). Their review found 10 different types of MCT used, including 'by eye' and 'undefined.' It would not be surprising to learn that similar inconsistencies in the application of MCTs exist in other fields, and it is clear that use and reporting of MCTs in many scientific disciplines is far from standardized. Although nonstatisticians need to use some criteria for selecting and reporting an MCT, some confusion may be expected as the statistical literature does not always provide non-technical recommendations for strengths, weaknesses, and applications of MCTs. Two factors motivated our study. First, scientists across disciplines are unlikely to use technical statistical literature to self-train on the correct use and best practices for MCTs. And second, scientists across disciplines are certain to continue using MCTs. Given these two conditions, we identified a need for non-statisticians to be better equipped to make decisions about which MCTs to use under different circumstances. Even when a correct MCT is used, a better understanding and clearer reporting of the application of the test may be warranted and improve the reporting.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The objectives of this study are to 1) Identify the most common MCTs used historically and currently in published ecological literature, 2) Conduct a simulation whereby the most commonly used tests are evaluated based on data sets with different (but common) attributes, 3) Evaluate data-independent considerations (e.g. planned vs. unplanned tests) for MCT selection, and 4) Combine the results of the simulation study with known best-practices for MCTs to arrive at recommendations for their selection and use.</ns0:p></ns0:div> <ns0:div><ns0:head>Background on the Conditions of Multiple Comparison</ns0:head><ns0:p>The development of MCTs has been undertaken to address the lack of specificity in comparing group means when using other statistical tools (e.g., ANOVA); however, all MCTs attempt to address the same inherent problem that stems from the propagation of statistical errors in hypothesis testing. Recall that the basic design of hypothesis testing yields one of four possible outcomes, which are the product of two possible states of the null hypothesis and two possible decisions about the null hypothesis (Table <ns0:ref type='table' target='#tab_6'>1</ns0:ref>). For the most part, we can ignore the two correct inferences that occur when the null hypothesis is true and we fail to reject it, or when the null hypothesis is false, and we correctly reject it. These are desired outcomes. The outcomes we need to be concerned with are commonly referred to as errors: a Type I error is a false positive, or rejecting a true null hypothesis, and a Type II error is a false negative, or failing to reject a false null hypothesis. Type I error is also commonly notated with &#945; and Type II error with &#946;, both of which can be thought of as probabilities.</ns0:p><ns0:p>Any test statistic and associated p-value is inherently providing an inference on an outcome. Because statistics involves describing quantities from distributions and their associated variability and uncertainty, statistical tests are not confirmatory by nature, and rather attempt to</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed provide a level of confidence about the outcome or estimate. Herein lies the challenge with MCTs. Any singular statistical outcome is compared against an &#945;, or a priori significance threshold. &#945; is often selected based on convention (e.g., 0.05), but also represents a compromise between a rigorous amount of evidence required (to avoid Type I error), but not so much evidence that we would never find a statistically significant result (to avoid Type II error). We know that &#945; will occasionally let a false positive go, but when conducting singular statistical tests, we accept this risk and have it quantified. With multiple comparisons, we are artificially increasing the number of tests (or test statistics), which greatly increases the chances of false positives if &#945; is not adjusted. For example, if we run just 15 MCTs (which represents all the pairwise combinations of 6 groups), and we do not adjust &#945;, our probability of a type I error is over 50%. The solution is to reduce &#945; such that there is a higher significance threshold that should reduce the false positives. However, when &#945; decreases, our &#946; will increase as we are Manuscript to be reviewed MCTs available. Additionally, MCTs that control the EER to below 5% (by strongly reducing PCERs) are known as conservative, while those with less strong adjustments of the PCERs which do not control the EER to at or below 5% are known as liberal. Finally, the term familywise error rate (FWER) is also commonly used to describe EER, and while this study uses them interchangeably, FWER and EER in some cases may be used to describe different collections of comparisons.</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIALS &amp; METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>Literature Review of Multiple Comparisons Test Usage</ns0:head><ns0:p>A literature search was conducted on 14 August 2019 to count the numbers of putative uses of 17 different MCTs reported in the ecological literature ( Although Google Scholar searches the whole text, there are some search limitations. For example, we were not able to search by discipline or category of journals. As a workaround, we specified in our search parameters that only journals with the term ecology in the title be searched. We recognize that this is an imperfect method to exhaustively search the full discipline of ecological journals; however, it was a bias imposed on all searches and the results likely included enough journals that we expect to have found the general trends. MCTs are also used in many fields beyond ecology; however, we wanted a field that was large enough to likely have all tests represented, while still confining our search to a specific scientific field. (And the authors PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed subsequent data set (e.g., large or small sample size). In many design cases there will be options for MCTs and understanding sensitivities about the tests may inform their selection.</ns0:p><ns0:p>Furthermore, test performance may be the determining factor when several tests are otherwise acceptable in a given situation. Therefore, to develop simulation-based recommendations for MCTs we evaluated the proportion of type I error, type II error, and the distribution of p-values for nine common MCTs under a range of data scenarios. Overall, we evaluated four main types of simulation study designs: balanced study designs for 1) type I error and 2) type II error, as well as unbalanced study designs for 3) type I error, and 4) type II error. Balanced study designs involved simulations with the same number of samples for each group, while unbalanced study designs had groups with varying numbers of samples. Furthermore, type I error was assessed by simulating data with the same mean and standard deviation between all groups while type II error designs contained one mean that was different among all groups. Finally, within each study design we evaluated four simulation treatments: 1) low sample size with few groups (LSFG), 2) low sample size with many groups (LSMG), 3) high sample size with few groups (HSFG), and 4) high sample size with many groups (HSMG) (see Figure <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The Data</ns0:head><ns0:p>A simulation iteration involved randomly drawing group samples from a normal distribution with a pre-specified number of samples, mean, and standard deviation. Given common practice for null hypothesis tests, we set group means to 0 except for type II study designs, where we set one group to have a mean of 1 rather than 0. Although in some datasets there may be multiple means that differ, our design was of a general and common scenario where a control group is</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed different from treatment groups. We systematically tested simulations with increasing standard deviation values and determined that a standard deviation of 3 was an approximate threshold for providing contrast between tests. Furthermore, maintaining the same standard deviation among groups permitted consistency in the results. We chose group number and sample size values based on values that seemed appropriate given our experience with real (ecological) datasets.</ns0:p><ns0:p>Low sample size simulations had 10 samples and high sample size simulations had 100 samples.</ns0:p><ns0:p>Unbalanced group study designs involved low sample size groups with 5, 10, or 15 samples and high sample size groups with 85, 100, or 115 samples. Finally, simulations with few groups had 3 groups, while simulations with many groups had 7 groups (see Figure <ns0:ref type='figure' target='#fig_3'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>The Simulations</ns0:head><ns0:p>Simulations involved 100 iterations, where all MCTs were run for each iteration. Nine parametric MCTs were used to test for differences between groups, 1) Scheff&#233;'s S test, 2) t-test with Bonferroni correction, 3) t-test with &#352;id&#225;k correction, 4) Tukey's HSD, 5) Fisher's LSD, 6)</ns0:p><ns0:p>Fisher's LSD with Bonferroni correction, 7) Fisher's LSD with &#352;id&#225;k correction, 8) Duncan's MRT, and 9) SNK. These nine tests were chosen based on their prevalence in the literature. We excluded the Dunnett's test because it is only applicable for special cases, and it would not have been appropriate to compare with other similar tests. Furthermore, we excluded Ryan's test due to the lack of readily available functions for its use, therefore limiting its applicability. We also excluded the Waller-Duncan k test because it does not use p-values to evaluate statistical significance and therefore would not be directly comparable to the other nine tests.</ns0:p><ns0:p>We systematically assessed results of our simulations with varied numbers of iterations.</ns0:p><ns0:p>Results with more than 100 iterations did not significantly vary from results with 1000 iterations.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>We therefore chose 100 iterations for the sake of computational efficiency. When running type I tests (no group differences), we extracted estimates and their associated statistics for all groups and all simulations. When running type II tests (group differences), we extracted estimates and their associated statistics for group comparisons between groups with a mean of 1 and 0. We then evaluated the proportion of type I or type II errors and the distribution of p-values by MCT across all simulations. All simulations were run in R ( <ns0:ref type='formula'>2020</ns0:ref>) and examples of functions used for calculating the various MCTs can be found in Table <ns0:ref type='table' target='#tab_10'>3</ns0:ref>. We have also included all the code needed to reproduce the results in this manuscript in Supplemental Code, found on the Github repository (https://github.com/stevemidway/MultipleComparisons) associated with this paper.</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head></ns0:div> <ns0:div><ns0:head>Literature Review of Multiple Comparisons Test Usage</ns0:head><ns0:p>Our literature review reported 41,561 instances of 17 different MCTs from published studies in the field of ecology (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>). However, use of the different MCTs was very unbalanced. For instance, the Bonferroni (and sequential Bonferroni) procedure accounted for nearly half (20,801) of all MCTs used while six other tests were reported over 1,000 times. On the other extreme, the Fligner-Policello test was only reported eight times and was one of three tests that were not even reported 100 times. Our investigation into select tests over time reveals an overall increase in the use of MCT, but this use is almost entirely explained by a small number of very popular tests-namely Bonferroni and Tukey's HSD (Figure <ns0:ref type='figure' target='#fig_4'>2</ns0:ref>). One hypothesis about the observed frequency of reported MCTs is that if we assume tests are applied correctly, then their usage reflects the types of data and analyses that researchers are performing. For example, the extreme use of Bonferroni should reflect the commonness of PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed is not likely to be understood from the literature; however, it might be expected when more than one test is appropriate. For instance, a researcher may select the more liberal test with the expectation that increased power will produce (more) significant comparisons. This factor is likely at play to some degree, judging by the nearly 3,000 reported instances of SNK, Fisher's LSD, and Duncan's MRT, which are all tests that are not recommended based on inadequate error rate adjustment.</ns0:p></ns0:div> <ns0:div><ns0:head>Simulation</ns0:head><ns0:p>In general, the type I error for treatments with small and large sample size showed the reverse trends from one another when using balanced or unbalanced data (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>). For balanced data, LSFG treatments had more type I PCER than HSFG treatments (Figure <ns0:ref type='figure' target='#fig_5'>3a</ns0:ref>). When there were many groups, type I PCER was greater with increased sample size for balanced designs, except for Duncan's MRT and Fisher's LSD. For unbalanced data, LSFG treatments had less type I PCER than HSFG treatments except for Scheff&#233;'s S and LSMG had more type I PCER than HSMG (Figure <ns0:ref type='figure' target='#fig_5'>3b</ns0:ref>). When comparing between unbalanced and balanced study designs, all HSMG and LSFG tests had lower type I PCER, Duncan's MRT and Fisher's LSD had lower</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed type I PCER for LSMG treatments, and the number of tests with lower type I PCER between balanced and unbalanced designs were split for HSFG treatments. The Duncan's MRT and unadjusted Fisher's LSD tests provided the greatest proportion of type I error regardless of the study design or treatment (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>). The SNK test provided an equal or higher type I PCER than all tests other than Duncan's MRT and the unadjusted Fisher's LSD, although it never exceeded a PCER of 0.05. This was most noticeable in the density plots, where the SNK test did not have a peak in p-value density near one but instead was relatively constant from zero to one (Figure <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>). The Scheff&#233;'s S test produced the least amount of type I error among all tests. The remaining five tests appeared almost identical in terms of proportion type I error allowed (Figure <ns0:ref type='figure' target='#fig_5'>3</ns0:ref>). All five tests had a large peak near one and a long tail to lower pvalues (Figure <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>). Somewhat surprisingly, both tests using Bonferroni correction had the largest density near one for type I study designs, yet the long tail from these peaks resulted in Bonferroni having more type I error than Scheffe's S test, which had a comparatively small peak near one. The density plots had similar trends between the balanced and unbalanced study designs, so only the balanced study design plots are shown (Figure <ns0:ref type='figure' target='#fig_6'>4</ns0:ref>). The patterns observed for proportion of type I error were approximately reversed for proportion of type II error due to the trade-off between these error rates (Figure <ns0:ref type='figure' target='#fig_7'>5</ns0:ref>). Having high sample size was more important for type II error than having more groups. However, the distribution of p-values appeared to favor more type II error when there were more groups (Figure <ns0:ref type='figure' target='#fig_8'>6</ns0:ref>). When sample sizes were large, the distribution of p-values was similar across all tests with peaks centered around or below 0.05 with tails of slightly varying sizes. Similar to the type I error tests, we have only shown the density plots for balanced study designs (Figure <ns0:ref type='figure' target='#fig_8'>6</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Our literature review and simulation were useful toward understanding how common MCTs performed under different-but realistic-data sets. The results from the simulation may be useful toward helping ecologists decide which MCT(s) is right for their data and analysis; however, independent of the actual data there are other criteria that may inform the choice of test. Here, we review and summarize suggestions for the application of MCTs based on published studies. The structure of this section follows delineations of whether data are parametric or not and whether comparisons are planned or unplanned, both of which are dichotomies that are useful in selecting an MCT (see Figure <ns0:ref type='figure'>7</ns0:ref> for a diagram on selecting an MCT).</ns0:p></ns0:div> <ns0:div><ns0:head>Parametric or Non-parametric data</ns0:head><ns0:p>The first delineation in MCTs is based on whether the data and model(s) are assumed to be parametric or non-parametric. As with several statistical decisions, knowing whether the data come from and exhibit parametric properties can greatly influence how the data need to be treated. Non-parametric data are immediately subjected to a different candidate list of MCTs than are parametric data, although decisions still need to be made as there is not a universally recommended non-parametric MCT.</ns0:p></ns0:div> <ns0:div><ns0:head>Non-parametric MCTs</ns0:head><ns0:p>For non-parametric planned comparisons, the common Mann-Whitney- Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Planned or Unplanned Tests</ns0:head><ns0:p>We will focus also on a delineation that many statistical texts emphasize regarding the right MCT-whether the comparisons are planned (i.e., a priori) or unplanned (i.e., post hoc). Planned comparisons are overwhelmingly recommended for several reasons. First, planned comparisons often result-but not always-in a number of comparisons that is lower than unplanned comparisons. In this case, fewer comparisons may mean (depending on the test) less adjustment to the EER and therefore a less strict threshold of significance. Planned comparisons also ensure that only meaningful and interesting hypotheses are entertained, and that EER is not being adjusted based on uninteresting or meaningless comparisons. Some sources even go so far as to remind us that all statistical designs under a null hypothesis significance testing paradigm should be planned (Kruschke 2013), and different p-values and inferences can be developed based on the same data set (through simply changing the sampling intentions and thus critical values used in test statistics).</ns0:p></ns0:div> <ns0:div><ns0:head>Unplanned Comparisons</ns0:head><ns0:p>Although planned comparisons are recommended and often provide benefits, the reality is that experimental design and data collection often have surprises and we cannot always plan everything perfectly. Unplanned comparisons may not include all pairwise combinations; however, they often do include all combinations. Unplanned comparisons for parametric data are by far the most commonly used MCTs, and also the category that offers the greatest variety of MCTs.</ns0:p><ns0:p>As a place to start with unplanned parametric MCTs, some recommend considering Scheff&#233;'s S test first <ns0:ref type='bibr'>(Ruxton and Beauchamp 2008)</ns0:ref>. Although Scheff&#233;'s S test is known to be conservative and it is 'entirely coherent with ANOVA results.' Scheff&#233;'s S test is also referred to as a protected test (protected from any differences or inconsistencies with ANOVA results), because a non-significant ANOVA will never produce a pairwise difference in a Scheff&#233;'s S test.</ns0:p><ns0:p>It is worth mentioning here that it may be surprising to know that the data used in a nonsignificant ANOVA could still produce a significant pairwise difference in a test other than Scheff&#233;'s S. <ns0:ref type='bibr'>Ruxton and Beauchamp (2008)</ns0:ref> <ns0:ref type='bibr'>2015,</ns0:ref><ns0:ref type='bibr'>2018)</ns0:ref>. Although in some cases, these interactions are better modeled by the use of random variables <ns0:ref type='bibr'>(Faraway, 2006;</ns0:ref><ns0:ref type='bibr'>Zuur et al., 2009)</ns0:ref>, in other cases, specific research hypotheses, specific models (e.g., multi-category logit), or insufficient sample sizes for parameterization of random effects result in situations where complex interactions of fixed effects are employed.</ns0:p><ns0:p>Interactions may be interpreted directly from parameter estimates, effect sizes estimated Manuscript to be reviewed an interaction, and for higher level interactions, lines may be plotted for two variables controlling for a third or more variables. Similar to interpreting linear models, the slope of the line and whether lines cross among levels of the variable meaning <ns0:ref type='bibr'>(Dowdy et al. 2003;</ns0:ref><ns0:ref type='bibr'>Kutner et al. 2004;</ns0:ref><ns0:ref type='bibr'>Mendenhall et al. 2013)</ns0:ref> and may be combined with information from p-values, effect sizes, or both. The disadvantages are few with the primary problems limited to potentially a large number of plots to examine and that if the magnitude of the difference is of interest, line plots will need to be complemented with parameter estimates criteria (e.g., p-values and objective functions)</ns0:p><ns0:p>and/or effect sizes. MCTs offer a straightforward method to interpret interactions providing a magnitude of difference in the original (linear models) or link transformed (generalized linear models) units and an adjusted p-value. Effect sizes may be readily estimated from the differences, if of interest, and lines may be plotted among the MCT estimated levels. Further, MCTs may be arranged in order of differences with lines or letters indicating significant differences. Therefore, the remainder of this section briefly reviews best practices for applying MCT in complex, multilevel interactions of fixed effects in a model. Interactions of fixed effects are sometimes seen as a nuisance, rather than as an opportunity for nuanced and detailed understanding of differing levels of categorical fixed effects (hereafter factors) and covariates. Often the simpler, main fixed effects and two-way interactions present attractive and simple interpretations. Yet, three-way and higher interactions may be fundamental to the study design or may account for important sources of variation. The lack of clarity in the interpretation of a complex interaction in a typical ANOVA-style output table in software (i.e., so which factor made it significant?) and often confusing presentations of individual parameter estimates that may differ in sum-to-zero or sum-to-last based on software choices (i.e., where is the parameter estimate for level j?) add to the view of complex interactions as frustrating. MCTs can be very useful in disentangling statistical significance and differences among parameter estimates.</ns0:p><ns0:p>The proper MCT implementation in the case of complex interactions is to perform the MCT on the most complicated statistically significant interaction. In other words, if the model includes three main fixed effects (X i1 , X i2 , and X i3 ), three two-way interactions (X i12 , X i13 , and X i23 ), and one three-way interaction (X i123 ) that is statistically significant, the MCT should be performed on the three-way interaction (X i123 ). This is because the parameter estimates for X i1 , X i2 , X i3 X i12 , X i13 , and X i23 are only meaningful as components of X i123 and are not interpretable on their own. If the interaction X i123 is not statistically significant, removing the interaction and refitting the model may be warranted (e.g., Faraway 2006) and interpreting the next most complicated interaction. The guiding principle should be outside-to-inside, if considering model notation, or bottom-to-top if considering an ANOVA-style output table.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Interpreting MCT in interactions can initially be intimidating, however, understanding the components makes the process easier and indicates where MCT choice impacts the interaction.</ns0:p><ns0:p>The output of MCT will generate an estimate based on the model for each factor (i) and level (j) of that factor. Although the term level is used here, level does not imply that the factor is ordinal or has a numeric value, rather, level is simply used as a convention as the level could be nominal as well. This estimate will be typically accompanied with a t-statistic and a p-value testing against a null estimate of 0. This p-value will be adjusted by the MCT, and if the p-value is used to make a determination of statistical significance of the estimate, the choice of MCT is important. If performed, for each pairwise comparison, a difference between estimates, test statistic, and an associated p-value are produced. In these comparisons as well, the choice of MCT will affect the test statistic and how the p-value is calculated. Sometimes, a comparison will be reported as non-estimable, which may mean that one combination of factor and level is missing or may be insufficiently replicated to generate an estimate. Therefore, when interpreting interactions, one should consider the appropriateness of the MCT for the data and model. Generally, for two-way interactions, MCT comparisons among the levels of each fixed effect are rather easy to follow. For example, a model with X i1 , X i2 , and X i12 is presented (Table <ns0:ref type='table' target='#tab_13'>4</ns0:ref>). The sign of the difference in MCT indicates which combination of variable and level is greater and illustrates the directionality of differences. The statistical significance of the MCT adjusted p-value indicates where the differences among variables and levels occurred, which is again not evident in ANOVA-style tables. It should be noted that the comparisons are interchangeable in two-way interactions (i.e., although presented as X i1 , j=1 -X i2 , j=1, to compare X i2 , j=1 -X i1 , j=1, one simply reverses the sign). Thus, a statistically significant twoway interaction of X i12 that has a statistically significant, positive X 1 , j=1 -X 2 , j=1 comparison</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed would be interpreted as the estimate at level 1 of X i1 is significantly greater than level 1 of X i2 .</ns0:p><ns0:p>One also can compare across levels; a statistically significant X i1 , j=1 -X i2 , j=2 would be interpreted as level 1 of X i1 differs from level 2 of X i2 . These comparisons do require an appreciation of the conditional nature of the comparison, for example, X i1 , j=1 -X i2 , j=2 does not mean that levels 1 and 2 are different always, they mean that levels 1 and 2 differ when comparing X i1 with X i2 and may not apply if the model includes another factor, X i3 . In higher level interactions, such as three-way or higher (e.g, X i123 ), the pairwise comparisons (e.g., X i2 , j=1 -X i1 , j=1) are conditional on other variables (e.g., X i3 ) in the interaction. Continuing with the example model with X i1 , X i2 , X i3 , X i12 , X i13 , X i23 , and X i123 , a three-way interaction is presented (Table <ns0:ref type='table' target='#tab_15'>5</ns0:ref>). Using the first comparison in the table, a</ns0:p><ns0:formula xml:id='formula_0'>statistically significant comparison of X i2 , j=1 | X i1 , j=1 -X i3 , j=1 | X i1 , j=1 would be interpreted as conditional on the first level of X i1 , level 1 of X i2 differs from level 1 of X i3 . If the variable X i1</ns0:formula><ns0:p>represents a sampling location, this pairwise comparison would be interpreted to indicate that X i2 level 1 differs from X i3 level 1 at the first sampling location. Higher level interactions of more than three variables follow the same logic, where each additional variable adds another conditional influence. Not all comparisons in MCT will be logical or relevant to the hypothesis being investigated. For example, the hypothesis could be investigating different food items in predator diets between seasons; however, the data were collected in different rivers, thus, resulting in a two-way interaction, river x season. In this case, river is included because there are inherent differences that cause variation. Ignoring this variation would be a mistake resulting in improper estimation of parameters and error, therefore, river is included in the interaction. However, only MCT results regarding season are relevant to the hypothesis. Although following Steegen et al.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed (2016) and Gelman (2017), there is value in examining in the 'multiverse' of multiple comparisons, it is still up to the investigator to focus on the relevant comparisons.</ns0:p></ns0:div> <ns0:div><ns0:head>GLMs</ns0:head><ns0:p>Generalized linear models (GLMs) present a highly useful class of models applicable in a number of situations, given that a link function describes the linear relationship between the observed mean with the mean of the linear combination of the model and the response is a random variable belonging to the exponential family of distributions. As a useful class of models, MCT techniques have been applied with GLMs. MCT is commonly and correctly used on the GLM estimates in their link transformed values; however, some confusion may occur when the MCT results are compared with the values in the original space (i.e., estimates inverse link transformed), particularly in cases where the interpretation based on visual inspection would differ (e.g., MCT on the GLM estimates suggests a difference not obvious in the observed data and vice versa). By the principle of invariant reparameterization of maximum likelihood estimates, statistical significance of MCT in GLM confers statistical significance in the observed data as well.</ns0:p></ns0:div> <ns0:div><ns0:head>False Discovery Rates</ns0:head><ns0:p>In addition to the previously mentioned alternatives to MCT (direct interpretation of parameter estimates, effect size estimation, and line plots), False Discover Rates (FDR) are another widely used method for interpreting meaningful statistical outcomes through an alternative process to control Type I error rates. Following <ns0:ref type='bibr' target='#b2'>Benjamini and Hochberg (1995)</ns0:ref> and Verhoeven et al.</ns0:p><ns0:p>(2005), MCT can be thought of as controlling the chance of at least one Type I error at a desired PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed level across all tests (V), whereas FDR can be thought of as controlling the proportion of Type I errors across all significant tests (i.e., V/r or desired rate/discoveries). FDR has the advantage of greater statistical power than traditional MCTs (i.e., keeps Type II level higher than a MCT across the same number of tests; <ns0:ref type='bibr'>Garcia 2004</ns0:ref><ns0:ref type='bibr'>Garcia , 2005))</ns0:ref> <ns0:ref type='bibr' target='#b3'>(Brinster et al. 2018;</ns0:ref><ns0:ref type='bibr'>White et al. 2019</ns0:ref><ns0:ref type='bibr'>). Verhoeven et al. (2005)</ns0:ref> suggest that FDR is relatively simple to implement, even in a spreadsheet, although computational mistakes occur in the literature <ns0:ref type='bibr'>(White et al. 2019)</ns0:ref>. Further, FDR introduces new considerations in reporting (e.g., adjusted p-value vs. q-value, with or without effect sizes and confidence intervals), and rates from FDR are not directly comparable to previously reported literature complicating comparisons. Therefore, although FDR offers numerous advantages, the goals of particular study may not be compatible for FDR alone suggesting a role for MCTs in studies.</ns0:p></ns0:div> <ns0:div><ns0:head>CONCLUSIONS Recommendations</ns0:head><ns0:p>The field of multiple comparisons includes a wide variety of very necessary procedures that often directly contribute to the results of scientific studies. Despite myriad test options, certain tests remain more popular than others, while some tests are rarely used. Due to the complexity of MCT choices and the increasing diversity of data and models that are being used, it is not</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed realistic to come up with a one-size-fits-all approach for their application. In many cases, identifying the basics of planned vs unplanned comparisons and parametric vs non-parametric data and models will narrow down MCT options (Figure <ns0:ref type='figure'>7</ns0:ref>). In addition to a decision tree approach for selecting MCTs, we have identified some broad recommendations from our own work and borrowing from others. 1. Know that you often have choices with it comes to MCTs. Although some data situations will leave you with only one test, many data and models will have more than one MCT to choose from. 2. Do not include more comparisons than you need. Consider each comparison to be a hypothesis. When extra and uninteresting comparisons are included, they not only provide no scientific progress, but they also add to the error rate adjustment by increasing the threshold for significance for other comparisons. 3. As noted by others, avoid Fisher's LSD, Duncan's MRT, and the SNK tests. These tests are very liberal as they do not make acceptable error rate adjustments. 4. For parametric situations, Scheff&#233;'s S test is coherent with ANOVA and especially recommended for linear combinations of means (not just pairwise comparisons). However, absent linear combinations of means, Tukey's HSD presents a robust and widely available test for a variety of situations. 5. When selecting an MCT, even the recommended MCTs perform differentially among studies with large and small observations with many and few groups. It may be necessary to compare among MCTs to determine the MCT that best suits the number of groups and number of observations within each group in a particular study.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed 6. Consider that MCT may not be the only option in a particular study. Other approaches, such as direct interpretation of the parameter estimates, effect size estimation, and plots covered in the Interaction section or FDR, discussed previously, may be complementary or alternative options. Also consider that MCTs are closely associated with p-values, and numerous critical evaluations of p-values suggest p-values alone (or at all) may be insufficient for interpretation. Manuscript to be reviewed Variable X i1 has 3 levels (j=3) and variable X i2 has 2 levels (j=2). Levels are presented as numbers in this example, but also may be words or characters. The method of estimating each variable-level combination (e.g., X i1 , X i2 ) depends on MCT, as does the test-statistic.</ns0:p></ns0:div> <ns0:div><ns0:head>Table 4. Example of MCT in a two-way interaction.</ns0:head><ns0:p>Variable X i1 has 3 levels (j=3) and variable X i2 has 2 levels (j=2). Levels are presented as numbers in this example, but also may be words or characters. The method of estimating each variable-level combination (e.g., X i1 , X i2 ) depends on MCT, as does the test-statistic.</ns0:p><ns0:p>Variable</ns0:p><ns0:formula xml:id='formula_1'>X i1 Level Variable X i2 Level Difference 1 1 Estimated X i1, j=1 -Estimated X i2, j=1 1 2 Estimated X i1, j=1 -Estimated X i2, j=2 2 1 Estimated X i1, j=2 -Estimated X i2, j=1 2 2 Estimated X i1, j=2 -Estimated X i2, j=2 3 1 Estimated X i1, j=3 -Estimated X i2, j=1 3 2 Estimated X i1, j=3 -Estimated X i2, j=2</ns0:formula><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Variable X i1 has 2 levels (j=2), variable B has 2 levels (j=2), and variable C has 2 levels (j=2).</ns0:p><ns0:p>The notation variable 1|variable 2 indicates the estimate is conditional on the second variable. For all situations, the test statistic and adjusted p-value depends on the choice of MCT.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Other MCT's excluded here (but listed in Table <ns0:ref type='table' target='#tab_6'>1</ns0:ref>) show relatively similar trends to SNK or were too infrequently reported to visualize on this figure.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:note type='other'>Figure 2</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 4</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:note type='other'>Figure 6</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>likely to misdiagnose null hypotheses that are false. At some level, all MCTs are an attempt to balance &#945; and &#946; based on various criteria, including aspects of the data and the number of comparisons to be conducted.The error-rate adjustments used in MCTs have their own terminology, which we will use in this study. Specifically, pairwise comparison error rate (PCER) is the probability of committing an error for an individual comparison and is often the error rate referred to in orthogonal comparisons (see below), or in cases where error-rate adjustments do not have to be made. The experimentwise type-I error rate (EER) is the error rate that reflects the probability of at least one type-I error occurring in a situation where several independent comparisons are made (the example in the above paragraph). The EER reflects the adjustment in PCER to account for multiple comparisons, and the variability in ways to adjust the PCER accounts for the variety of PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>parametric data and models, while the relatively less common nonparametric MCTs reflect fewer studies with nonparametric data and models. It might be expected that such large-scale interpretations are correct; however, the use and popularity of MCTs are likely influenced by several other factors, such as accessibility of tests in common statistical software, familiarity or understanding of tests by non-statisticians (perhaps based on their simplicity or exposure from previous uses), and the overall power of the test. The issue of selecting an MCT based on power</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>from parameter estimates(Cortina and Nouri 2000; Pituch and Stevens 2015), construction of line plots connecting effect means(Dowdy et al. 2003; Kutner et al. 2004; Mendenhall et al. 2013), MCTs, or by a combination of these approaches. Each approach has advantages and PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)Manuscript to be reviewed disadvantages, and a thorough review is outside of the scope of this effort. However, a few generalizations may help in placing MCTs in the context of these other approaches. Direct interpretation of parameter estimates uses the estimates themselves in a common scale (unit change in y given unit change in x) for comparison by taking advantage of the sum-to-zero principle in linear models (i.e., differences among levels within and among variables can be compared directly with a variable level set to 0), which is particularly useful for planned comparisons (K&#233;ry and Royle 2016). Determining whether the magnitude of the parameters estimate differences between or among the estimates within an interaction has meaning often uses criteria, such as p-values or may use an objective function (i.e., whether AIC suggests inclusion of the interaction term). Performance and concerns over the use of these criteria in this manner are well documented in the literature(Stephens et al. 2005(Stephens et al. , 2007; Murtaugh 2014; Wasserstein and Lazar 2016; Wasserstein et al. 2019). Effect sizes are estimated from the magnitude of differences between or among parameter estimates and offer an alternative criteria, often in units of standard deviation (Cohen's d) or variance (Nagelkerke's R 2 or &#544; 2 ) for assigning meaning to the interpretation of interactions either by the use of published guidelines (e.g., Cohen 1988) or comparison with discipline-specific literature values (Pituch and Stevens 2015).The use of effect sizes addresses some concerns about the use of p-values or objective functions (e.g., AIC) to assign meaning(Lenth 2001; Nakagwa and Cuthill 2007; Ellis 2010); however, some effect size estimators may not translate well across applications and are highly influenced by sample size, and importing guidelines across disciplines may be problematic(Osenberg et al. 1997; Nakagwa and Cuthill 2007; McCabe et al. 2012; Pituch and Stevens 2015; Pogrow 2019).Graphical interpretation of interactions by plotting lines connecting means is commonly available in software and has a long history of use. Lines may be plotted for any two variables in PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1. Diagram of multiple comparison simulations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Reported uses of the 3 most common parametric multiple comparisons tests (MCTs) by 5-year intervals.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3. Type I errors in simulations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4. p-values for type I errors in simulations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5. Type II errors in simulations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6. p-values for type II errors in simulations.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 7 Figure 7 .</ns0:head><ns0:label>77</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>). The search was conducted in Google Scholar because Web of Science (and comparable literature search programs) do not search the full text of articles, and MCTs are not commonly included in the title, abstract, or other searchable article information; MCTs are typically reported in the main body of the text.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>all operate within the domain of ecology, so we felt most comfortable working with this</ns0:figDesc><ns0:table><ns0:row><ns0:cell>literature.)</ns0:cell></ns0:row><ns0:row><ns0:cell>As stated earlier, we also recognize that some MCTs have different names or</ns0:cell></ns0:row><ns0:row><ns0:cell>abbreviations. For Tukey's Honest Significant Difference (HSD) and Fisher's Least Significant</ns0:cell></ns0:row><ns0:row><ns0:cell>Difference (LSD) we searched by abbreviation, under the assumption that while any first</ns0:cell></ns0:row><ns0:row><ns0:cell>mention of a test would include the whole test name, these specific abbreviations are well-</ns0:cell></ns0:row><ns0:row><ns0:cell>established and very common. Therefore, we examined different terms for test names (e.g.,</ns0:cell></ns0:row><ns0:row><ns0:cell>Tukey's HSD vs. Tukey's Test vs. Tukey) and ultimately searched the term that we thought was</ns0:cell></ns0:row><ns0:row><ns0:cell>the most descriptive of the test (although in instances where we searched multiple terms, the</ns0:cell></ns0:row><ns0:row><ns0:cell>search results were typically similar). We could not exclude any instances where a specific name</ns0:cell></ns0:row><ns0:row><ns0:cell>eponymous with an MCT (e.g., &#352;id&#225;k) could have been the name of an author or some other use</ns0:cell></ns0:row><ns0:row><ns0:cell>of that name unrelated to multiple comparisons. Despite this limitation, most of the tests we</ns0:cell></ns0:row><ns0:row><ns0:cell>searched included terms in addition to just a name and we expect any mentions of our search</ns0:cell></ns0:row><ns0:row><ns0:cell>terms to be almost entirely related to multiple comparisons. In other words, it is unlikely that a</ns0:cell></ns0:row><ns0:row><ns0:cell>person named after or sharing a name with an MCT would have published so much in the</ns0:cell></ns0:row><ns0:row><ns0:cell>ecological literature that they would overwhelm the search of a common statistical test. We were</ns0:cell></ns0:row><ns0:row><ns0:cell>interested in looking at the uses of certain MCTs over time, and as such, select tests terms were</ns0:cell></ns0:row><ns0:row><ns0:cell>searched for in the literature in 5-year blocks, starting in 1960 and going to 2019.</ns0:cell></ns0:row><ns0:row><ns0:cell>Simulation of Multiple Comparison Tests</ns0:cell></ns0:row><ns0:row><ns0:cell>Although there are certain recommendations that describe how different MCTs should be used in</ns0:cell></ns0:row><ns0:row><ns0:cell>different (experimental) design settings, there is not necessarily a correlation between</ns0:cell></ns0:row><ns0:row><ns0:cell>experimental design (e.g., planned vs. unplanned comparisons) and the attributes of the</ns0:cell></ns0:row><ns0:row><ns0:cell>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head /><ns0:label /><ns0:figDesc>Planned comparisons often represent the fewest number of comparisons, and if the comparisons are independent (i.e., orthogonal), allow a simple PCER (e.g., &#945;=0.05) because there is no required adjustment to the EER. A Student's t-test can simply be used when group variances are equal. A Behren's-Fisher t -test or Welch's t-test are recommended when group variances are unequal. The following sections focus primarily on MCT test options for unplanned parametric MCTs, although the conditions for planned and unplanned may not be very different in certain circumstances.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1963) is recommended for situations where all the data are ranked together (i.e., jointly), or</ns0:cell></ns0:row><ns0:row><ns0:cell>alternatively, the Steel-Dwass test (Steel 1960) is recommended for pairwise rankings, where</ns0:cell></ns0:row><ns0:row><ns0:cell>data are re-ranked for each pairwise comparison.</ns0:cell></ns0:row><ns0:row><ns0:cell>It should also be mentioned that many researchers will use a non-parametric MCT if they</ns0:cell></ns0:row></ns0:table><ns0:note>Wilcoxon U test is recommended (Day and Quinn 1989), and if the distributions are symmetrical the Fligner-Policello test provides a robust alternative (Fligner and Policello 1981). If non-parametric PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020) Manuscript to be reviewed comparisons are unplanned, up to four different tests may be available. The Dunn procedure (Dunn 1964) and Games-Howell Test (Games and Howell 1976) are commonly used. If the order of rankings is an important consideration, then the Nemeyni Joint-Rank test (Nemenyi think their residuals are not normally distributed. This could be a mistake because variances may still be unequal and because non-parametric tests are still valid for moderately non-normal samples. However, a test like the Mann-Whitney-Wilcoxon U test may be useful in a case of skewed data because the rank approach relaxes the effects of extreme values. For this reason, it is worth considering the features of a data set-and not just the outcome of a normality test-when considering MCTs.Parametric MCTsIf operating under a parametric assumption, planned comparisons may be the simplest option.PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head /><ns0:label /><ns0:figDesc>Perhaps most commonly used is Tukey HSD (honest significant difference) test widely applied for parametric unplanned comparisons (Tukey 1949), although the lesser known Tukey-Kramer test (Kramer 1956) should be used in cases of unequal sample size. Our simulations found Tukey's HSD test to be less conservative than the Dunn-&#352;id&#225;k test, and with lower Type II error rates than Bonferroni. Overall, the Tukey HSD test is a robust, Ryan 1960) are less common options, but may be attractive due to their ability to be modified for heterogeneity of variance.We have excluded recommending Fisher's LSD, Duncan's MRT, or SNK for unplanned comparisons as they are known for not adjusting the EER enough(Day and Quinn 1989; Ruxton and Beauchamp 2008). Specifically, with the SNK test, the EER can become greater than the chosen significance probability when there are two groups of means, such that the means within each group can be equal but the test will say that the groups differ.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Other Considerations</ns0:cell></ns0:row></ns0:table><ns0:note>go so far as to suggest that 'If any procedure other than Scheffe's is used, then it should be implemented regardless of the outcome of theANOVA.' Although there are reasons to consider Scheff&#233;'s S Test, it should be noted that the test is inherently conservative because it is designed for linear combinations of means and just pairwise comparisons. If no linear combinations are used, Scheff&#233;'s S Test may be more conservative than desired; however, if linear combinations of means are being compared, it is often an ideal MCT. If Scheff&#233;'s S test is not desired, other options exist. The Bonferroni and sequential Bonferroni tests are commonly used MCTs. Bonferroni is known to be very conservative, and the sequential Bonferroni (Holm 1979) was developed as an alternative approach which is to be no less conservative. The Dunn-&#352;id&#225;k test (or &#352;id&#225;k correction; &#352;id&#225;k 1967) is a recommended alternative to Scheff&#233;'s S test. PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020) Manuscript to be reviewed commonly available, and generally recommended test. Finally, the Waller-Duncan k-test (Waller and Duncan 1969) and Ryan's test (Interactions Frequently, research hypotheses may include the consideration of the interactions of fixed effects within a model, be structured as multilevel models, or require interpretation of fixed effects interactions in evaluating hypotheses of categorical data (e.g., extensions of contingency tables, counts or proportions of survey responses, relative abundances of taxa; Faraway 2006; Agresti</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head /><ns0:label /><ns0:figDesc>, being adaptive (i.e., rate is based on the number of discoveries, rather than number of overall tests; Garcia 2005; Verhoeven et al. 2005), and being consistent (also called scalable, i.e., rate has the same meaning, regardless of the number of discoveries). Recent comparisons of MCT (family-wise error rate methods) with FDR methods demonstrate the advantages of FDR; however, also demonstrated the importance of proper implementation</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 1 . Four outcomes of hypothesis testing.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>The two types of error are presented in boldface text.</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Null Hypothesis (H 0 ) True</ns0:cell><ns0:cell>Null Hypothesis (H 0 ) False</ns0:cell></ns0:row><ns0:row><ns0:cell>Fail to reject</ns0:cell><ns0:cell>Correct (true negative)</ns0:cell><ns0:cell>Type II error (false negative)</ns0:cell></ns0:row><ns0:row><ns0:cell>Reject</ns0:cell><ns0:cell cols='2'>Type I error (false positive) Correct (true positive)</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_8'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Multiple comparisons tests (MCTs) searched in the literature and total number of reported uses from 1960-2019.Tests are ordered by their general application and then by popularity-the number of times cited in ecological literature. Note: Terms like test and procedure have been removed where not necessary. Based on the literature, the bottom three tests are often not recommended, which is guidance we have adopted (and discuss in the study). Finally, we are not able to differentiate Bonferroni from sequential Bonferroni, but we expect that the number of reported citations captures most of both uses.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 2 . Multiple comparisons tests (MCTs) searched in the literature and total number of reported uses from 1960-2019.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Tests are ordered by their general application and then by popularity-the number of times cited in ecological literature. Note: Terms like test and procedure have been removed where not necessary. Based on the literature, the bottom three tests are often not recommended, which is guidance we have adopted (and discuss in the study). Finally, we are not able to differentiate Bonferroni from sequential Bonferroni, but we expect that the number of reported citations captures most of both uses.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Test</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_10'><ns0:head>Table 3 (on next page)</ns0:head><ns0:label>3</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_11'><ns0:head>Table 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Common multiple comparisons tests and their software implementations.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>This table is meant to serve as a reference for functions and is not meant to advocate for</ns0:cell></ns0:row><ns0:row><ns0:cell>particular packages and functions over others. Functions may give different results from one</ns0:cell></ns0:row><ns0:row><ns0:cell>another, and we recommend reading any instructions or helpfiles for details on specific test</ns0:cell></ns0:row><ns0:row><ns0:cell>implementations. Boldface functions indicate those used in the simulation component of this</ns0:cell></ns0:row><ns0:row><ns0:cell>manuscript.</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_12'><ns0:head>Table 3 . Common multiple comparisons tests and their software implementations.</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>This table is meant to serve as a reference for functions and is not meant to advocate for particular packages and functions over others. Functions may give different results from one another, and we recommend reading any instructions or helpfiles for details on specific test implementations. Boldface functions indicate those used in the simulation component of this manuscript.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Multiple comparisons test</ns0:cell><ns0:cell>R package::function</ns0:cell><ns0:cell>SAS Statements</ns0:cell><ns0:cell>SPSS Options</ns0:cell></ns0:row><ns0:row><ns0:cell>Tukey's HSD</ns0:cell><ns0:cell>stats::TukeyHSD*</ns0:cell><ns0:cell>MEANS / tukey;</ns0:cell><ns0:cell>Available by menu</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>agricolae::HSD.test</ns0:cell><ns0:cell>LSMEANS / adjust = 'tukey'*;</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>TukeyC::TukeyC</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>DescTools::PostHocTest</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>t-test with &#352;id&#225;k correction</ns0:cell><ns0:cell cols='2'>MHTdiscrete::Sidak.p.adjust MEANS / sidak;</ns0:cell><ns0:cell>EMMEANS ADJ(SIDAK)</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>mutoss::sidak</ns0:cell><ns0:cell>LSMEANS/ adjust = 'sidak'</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>t-test with Bonferroni correction</ns0:cell><ns0:cell>stats::p.adjust</ns0:cell><ns0:cell>MEANS / BON;</ns0:cell><ns0:cell>Available by menu</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>mutoss::bonferroni</ns0:cell><ns0:cell>LSMEANS/ adjust = 'Bon';</ns0:cell><ns0:cell>EMMEANS ADJ(BONFERRONI)</ns0:cell></ns0:row><ns0:row><ns0:cell>Scheff&#233;'s S</ns0:cell><ns0:cell>agricolae::scheffe.test</ns0:cell><ns0:cell>MEANS / scheffe;</ns0:cell><ns0:cell>Available by menu</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>DescTools::ScheffeTest</ns0:cell><ns0:cell>LSMEANS / adjust = 'scheffe';</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>GAD::snk.test</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>DescTools::PostHocTest</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Student-Neumen-Keul's Test</ns0:cell><ns0:cell>agricolae::SNK.test</ns0:cell><ns0:cell>MEANS / snk;</ns0:cell><ns0:cell>Available by menu</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>GAD::snk.test</ns0:cell><ns0:cell>LSMEANS /adjust = 'snk';</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>DescTools::PostHocTest</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Fisher's LSD</ns0:cell><ns0:cell>agricolae::LSD.test</ns0:cell><ns0:cell>MEANS / LSD</ns0:cell><ns0:cell>Available by menu</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>PMCMRplus::lsdTest</ns0:cell><ns0:cell>not available in LSMEANS</ns0:cell><ns0:cell>EMMEANS ADJ(LSD)</ns0:cell></ns0:row><ns0:row><ns0:cell>Fisher's LSD with &#352;id&#225;k correction</ns0:cell><ns0:cell>agricolae::LSD.test</ns0:cell><ns0:cell>Not available</ns0:cell><ns0:cell>Not available</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>MHTdiscrete::Sidak.p.adjust</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>mutoss::sidak</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Duncan's MRT</ns0:cell><ns0:cell>agricolae::duncan.test</ns0:cell><ns0:cell>MEANS / duncan</ns0:cell><ns0:cell>Available by menu</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>PMCMRplus::duncanTest</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>*as Tukey-Kramer</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_13'><ns0:head>Table 4 (on next page)</ns0:head><ns0:label>4</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_14'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Example of a MCT in a two-way interaction.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_15'><ns0:head>Table 5 (on next page)</ns0:head><ns0:label>5</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_16'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Example of a MCT in a three-way interaction.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_17'><ns0:head>Table 5 . Example of MCT in a three-way interaction.</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Variable X i1 has 2 levels (j=2), variable 2 B has 2 levels (j=2), and variable C has 2 levels (j=2). The notation variable 1|variable 2 3 indicates the estimate is conditional on the second variable. For all situations, the test statistic 4 and adjusted p-value depends on the choice of MCT. 5</ns0:figDesc><ns0:table><ns0:row><ns0:cell>1 Variable</ns0:cell><ns0:cell>Variable</ns0:cell><ns0:cell>Variable</ns0:cell><ns0:cell>Difference</ns0:cell></ns0:row><ns0:row><ns0:cell>X i1 Level</ns0:cell><ns0:cell>X i2 Level</ns0:cell><ns0:cell>X i3 Level</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i2, j=1 | X i1, j=1 -Estimated X i3, j=1 |X i1, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i2, j=1 | X i1, j=1 -Estimated X i3, j=2 |X i1, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i2, j=2 | X i1, j=1 -Estimated X i3, j=1 |X i1, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i2, j=2 | X i1, j=1 -Estimated X i3, j=2 |X i1, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i2, j=1 | X i1, j=2 -Estimated X i3, j=1 |X i1, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i2, j=1 | X i1, j=2 -Estimated X i3, j=2 |X i1, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i2, j=2 | X i1, j=2 -Estimated X i3, j=1 |X i1, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i2, j=2 | X i1, j=2 -Estimated X i3, j=2 |X i1, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i2, j=1 -Estimated X i3, j=1 |X i2, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i2, j=1 -Estimated X i3, j=2 |X i2, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i2, j=1 -Estimated X i3, j=1 |X i2, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i2, j=1 -Estimated X i3, j=2 |X i2, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i2, j=2 -Estimated X i3, j=1 |X i2, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i2, j=2 -Estimated X i3, j=2 |X i2, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i2, j=2 -Estimated X i3, j=1 |X i2, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i2, j=2 -Estimated X i3, j=2 |X i2, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i3, j=1 -Estimated X i2, j=1 |X i3, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i3, j=1 -Estimated X i2, j=2 |X i3, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i3, j=1 -Estimated X i2, j=1 |X i3, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i3, j=1 -Estimated X i2, j=2 |X i3, j=1</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i3, j=2 -Estimated X i2, j=1 |X i3, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=1 | X i3, j=2 -Estimated X i2, j=2 |X i3, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>1</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i3, j=2 -Estimated X i2, j=1 |X i3, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>2</ns0:cell><ns0:cell>Estimated X i1, j=2 | X i3, j=2 -Estimated X i2, j=2 |X i3, j=2</ns0:cell></ns0:row><ns0:row><ns0:cell>6</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:51035:1:1:NEW 13 Oct 2020)</ns0:note> </ns0:body> "

"Rebuttal letter to editorial and peer-review comments on: Comparing Multiple Comparisons: Practical Guidance for Choosing the Best Multiple Comparisons Test All comments/edits/requests have been reproduced in black font. Author team responses are provided in blue text. Editor Comments: Please make sure to pay particular attention to the Reviewers comments about discussing other approaches for making conclusions about meaningful effects, including focusing on parameter estimation rather than p-values and false discovery rate approaches. With respect to the latter, consider similar discussions of the use of FDR procedures in Biology (e.g.: Verhoeven et al. 2005 https://doi.org/10.1111/j.0030-1299.2005.13727.x) Response: We agree with the reviewer comments about including other approaches about concluding effects. While we were aware of these approaches from the start, we wanted to focus on multiple comparisons, but now see that some context for other approaches helps to contextualize multiple comparisons tests. To that end and in accordance with several reviewer comments, throughout the document we have added text on the topics of parameter estimates and false discovery rates, and other minor edits toward this end. Reviewer 1 (Anonymous) Comments The authors have reviewed the literature to show which types of multiple comparisons tests (MCTs) are being used by authors of papers in ecological journals (and just as important, which are not being used), and then present results of a simulation study to compare the performance of different MCTs. The literature review provides nice context and seems to be done in a reasonable way, but by itself would not provide sufficient content for an interesting publication. The simulation results are the more substantial part of the manuscript. I have looked at the authors’ code and results, and see no flaws. Their main results repeat the findings from some earlier papers; a few methods for MCTs are fundamentally flawed and should not be used, whereas a few others perform well. Nonetheless, I am not aware of any previous papers that are as thorough or clear. Thus, I think this work would provide a useful reference for many researchers, including those beyond the ecology researchers that are the focus of the literature review. The manuscript is well written. I found the text clear and free of grammatical errors. The authors did a good job proof-reading and preparing the manuscript, which made the task of reviewing simpler. Thanks. My main recommendation for improving the manuscript would be to expand the discussion of methods to at least touch upon the methods for false discovery rate (FDR) adjustments. The Benjamini-Hochberg method for family wise error rate adjustments (FWER, line 116) has become extremely common in other sub-disciplines of biology over the last couple decades, and is used at least as commonly as MCTs. I teach graduate biostatistics and draw students from a range of fields. Because of its importance in “-omics” analyses, I have had to add discussion of FDR approaches alongside older MCTs such as Bonferroni, Tukey HSD, Scheffé, and Dunn-Šidák (the only MCTs I covered the last time I taught biostatistics). Despite being dominant in a slightly different field, I think that most biology researchers and students are likely to be familiar with (or at least have seen reference to) the Benjamini-Hochberg FDR approach. Thus it would greatly improve the relevance and utility of the present manuscript to include some discussion of the relationship between FDR and MCT methods. Response: We agree with the reviewer and have added a subsection to the discussion that is devoted to FDR. This topic was omitted from the original submission because we thought it might be outside the scope of MCTs; however, we agree that some treatment of FDR and MCT improves the manuscript. Reviewer 2 (Anonymous) Basic reporting This is a useful study, but is narrowly focused, so that results are limited in scope. The English is mostly very good, but there are a few poorly phrased sentences where the meaning would not be clear to the intended reader, and there are also some erroneous statements that need to be corrected. A wider range of relevant literature could have been used. Response: This comment is somewhat broad, but we will attempt to reply. First, we think that the revision has produced a manuscript that is broader in scope than the original. Examples of the increase in breadth can be found in the responses to the reviewers. Second, we have revised any poorly phrased sentences as per reviewer suggestions. Finally, we have also addressed any erroneous statements and added several more citations. Experimental design The design of the simulations is adequate, but limited in scope, and as a result the results are not as general as they could be. Response: We made some revisions to the simulations, and those revisions are detailed in response to specific reviewer comments on this topic (found below). Validity of the findings The results are sometimes not interpreted well, with some minor errors. Much of the paper is advice drawn from previous literature, rather than based on results presented here. I consider that the advice should be more balanced, covering the various interpretations of best practice that are present in the literature. Response: Many of the conclusions from the literature review were in agreement with our simulations, and for whatever reason, it appears we lean more on the literature—perhaps because the totality of the literature is greater than our simulations. Regardless, the reviewer makes a good point, and our revision attempts to strike more of a balance of support between the literature and the simulation results. We have drawn on other specific comments (below) from the reviewer where the simulation results could be emphasized in the balance of how we develop our advice. Comments for the Author The paper first shows that despite good advice in published papers and data analysis texts written for ecologists, there are still many ecological papers published in which the multiple comparison tests used are poorly chosen. It then provides advice to ecologists as to which tests should be used. This is actually the most difficult part. Perhaps the visual presentation in the paper of the way these tests behave will grab attention to the issue and further research by future authors – I hope so. But to quote from Oaten (1995) Ecology 76: 2001-9: “The difficulties arise from the need to present the methods concisely and, as far as possible, painlessly. …. assertions tend to rely on proof by authority and to take the form of exhortations and instructions rather than statements of results…In time, some judgments become accepted as laws. These can be hard to challenge”. Further, simulations can be misleading, by showing what happens only under particular conditions and with particular types of data. Response: Although we understand where the reviewer is coming from and do not dispute the fact that “...some judgements become accepted as laws,” we would like to politely rebut this point. Perpetual mistakes do not make good practice; in fact, perpetual mistakes are perhaps worse than solitary mistakes because perpetual mistakes accumulate. If the reviewer is attempting to suggest that something can be acceptable if enough people do it, we would push back on that notion when what is being done is not a best practice. Yes, we recognize that mistakes are likely to happen when one group (ecologists) adopts methods from another group (statisticians); however, these mistakes are the primary motivator for writing this manuscript. If one manuscript can dispel “some judgements” and clearly explain best practices, then why would that be resisted or not encouraged? The best starting point is probably to note that most statisticians would encourage estimating confidence limits for the size of effects (e.g. differences between means), rather than use significance tests, because using tests encourages readers to mistakenly believe that effects are real if (and only if) the test is significant. MCTs can be seen as an attempt to maintain this belief even when researchers are actually data-snooping rather than testing pre-set hypotheses. Thus, I would have preferred to see this paper start with a brief summary to point out that ideally ecologists should be setting out to estimate the size of hypothesised effects, rather than testing whether their data provides a significant result for an effect. They should be especially cautious if the study design involves questions about which of several groups may differ, and try if possible to ask pre-set questions about particular differences, on the basis of the focus of the study or the theory that drives the study. Often these planned questions will not relate to differences between pairs of means, but more complex differences. The best approach is to estimate the size of these differences (with confidence limits). MCTs appear to be appropriate when either the investigator has little idea of which groups may differ when setting up the study, or the results of a study are surprising, so that new, unplanned questions arise after examining the data. At this point, one could enumerate the comparisons that have reasonable biological interpretations and set up MCTs based on this set of comparisons. However, an alternative view is that the PCER could be used for all comparisons, with a warning to the reader that the overall error rate would be >0.05 as each question is considered in isolation from the others. Response: Much like our adoption and inclusion of FDR, we have added text and discussion about effect sizes. We largely agree with the reviewer that effect sizes are another tool toward understanding differences between means. Text about effect sizes has been added in the subsections about interactions, FDRs, and even as a final (and additional) recommendation. My other general concerns with this paper are that: 1. The data used for simulations is all drawn from the normal distribution, with equal variances. This is a missed opportunity, as it is very unlikely that most data from ecological experiments (or in many other types of experiments or surveys) would be normally distributed (in many cases, not even very well behaved), or have equal variances in each treatment, as authors often do not consider the likely distribution of their data. A study that included the robustness of the MCTs to modest differences between variances and modest variations in data distribution (modest meaning those which would not be obvious in small samples) would have been more useful, as it would show the behaviour of tests in the context of the kind of data that they are used on in practice, and as previous studies have examined the error rates of most of these tests on normally distributed data with equal variances. Further, the issue of unequal variances is not covered in the recommendations, especially in regard to the use of non-parametric tests (see my specific comment below). Response: We understand the reviewer’s comment that ecological data are often messy and the perception of our simulations might look like a neat and normal approach. We would like to provide several responses that both address the reviewer’s concerns and also help explain our approach behind the design of the simulations. 1. Yes, we used equal variances in our simulations. This was a pragmatic decision because when variances are not equal, there are very few choices. The Waller-Duncan test can be used, although it does not use p-values and therefore would not be a viable comparison within our simulations (because nearly all MCTs use p-values). The Ryan test is another option. All of these points are made in the manuscript. 2. To address the concern of messy data, there are two potential responses. The first is the situation where the data do come from a normal distribution, but the sample is not sufficiently normal. We recognize this can be a common occurrence in ecological data, and to address that we included the low sample size groups in our simulation. The idea behind the low sample size groups was that it can be hard to represent a normal distribution with only 10 samples (the n we used) and by iterating this we are capturing the noise in small sample situations. The second concern is that the data are actually not normal; i.e., they come from another distribution. To that end, we have a subsection in the Discussion about generalized linear models (GLMs) that discusses the application of MCTs in instances where non-normal distributions are used in models. Although we do not see it as our job to review distributions and distributional assumptions of linear modeling, we do provide a full context of MCT approaches of the analyst who understands their data’s distributional assumptions. 3. To the argument that “authors often do not consider the likely distribution of their data,” this speaks to a larger issue of lack of command of statistical tools. We are not disagreeing with the reviewer, but we also do not see it as the scope of our manuscript to review the assumptions of linear modeling. 4. Finally, while we agree with the reviewer that there are extensions to the simulations that could include MCT robustness to non-normal data, etc., we see that as beyond the scope of this study. We set out to review and simulate reasonable and common data scenarios using randomized iterations of normal data and we think we have achieved that. We would look forward to future studies that dive deeper into specific data structures and their coherence with MCTs. Yes, data can get messy, but by studying data analysis under more constrained circumstances, we will gain better fundamental understanding of the characteristics that can then be used to study behaviors of messy data. 2. The authors essentially dodge the issue of what the criteria are for using non-parametric tests. Many researchers feel that if they are not sure if the data are normally distributed, they should use a non-parametric test, and this is not good practice, partly because the issue of unequal variances requires a different analysis process, and partly because parametric tests (which actually rely on the distribution of the sample mean being approximately normal) are valid for moderately non-normal samples even for small sample sizes. Tests on samples that have different skewed distributions are a case where rank-based tests such as the MWW-U test are useful because the use of ranks reduces the effects of extreme observations, but there are also other ways to reduce the effect of extreme observations when using parametric tests. The advantage of this is that estimates of differences are possible. Response: Because non-parametric analyses are much less common than parametric approaches, and there are fewer choices for non-parametric MCTs, we previously dedicated less text to the issue of non-parametric MCTs. However, we agree with the reviewer that non-parametric MCTs can be distinguished, and we have done so by adding the ideas that the reviewer provided as a small additional paragraph in the Parametric vs Non-parametric discussion section. 3. I initially found it hard to understand why the simulated type 1 error rate of the parametric tests in Figure 3 is far below the 5% level in all cases. Then I realized that the error rates shown are per comparison rates, although this is not specified in the text or in the figure captions. If you want to know how well a test procedure works, it is more useful to show how well it does what it is designed to do, which is to ensure the EER rate does not exceed the 5% rate. So I would want to see the Type 1 EER for each test. I suggest this ought to be shown, and in any case the caption should make clear which error rates are shown. Obviously, the tests control EER by reducing the PCER rate below 5%. The results report that the study designs with many groups had the lowest per comparison error rates, with no explanation. This is what the MCTs must do, to control the EER. Response: We agree with the reviewer that we should have specified that the figures represent PCER rather than EER and have modified the figure captions appropriately. In addition, we understand the reviewer’s interest in examining similar figures but for EER and have created a supplementary document that contains these figures. 4. The investigation of type 1 error only when a group of means are all equal misses the problem that occurs with the SNK test, where the EER becomes greater than the chosen significance probability when there are two groups of means, such that the means within each group are equal, but the groups differ (see Quinn, G.P. and Keough, M.J. (2002) Experimental design and data analysis for biologists. Cambridge University Press, Cambridge, p200). This leads to a false view of the performance of this test in Figures 3 and 4. As this was a known problem with this test, I suggest the authors should have demonstrated this problem. Response: Although we understand the reviewer’s comment, we do not think this needs a demonstration. First, as the reviewer states, this is a known problem and as such we do not see the need to recreate it. Second, we are not recommending SNK be used for the reasons discussed, and as such, we prefer to focus the manuscript on examining MCTs that are possible recommendations. Although we prefer not to demonstrate the issues with the SNK, we have added a few sentences regarding why SNK is not recommended and some of the caveats of its use. 5. It seems strange that the paper records the type 2 error rates of the various parametric MCTs (when one group mean differs, using normal distributions, for 4 study designs), but these results are not used when discussing which tests are recommended for various situations. The results (in Figure 5) show for example, that there is consistently less Type 2 error for Tukey’s HSD test than for the Bonferroni and Dunn-Sidak tests that are used for equivalent situations. While the difference is small, this suggests that the HSD tests will often be more powerful and thus a better choice. Exceptions might occur if the data are not of the type simulated, but this result conforms to that shown in Day and Quinn 1989, where the pattern of differences between groups was different, which suggests this is likely to be a consistent advantage of the HSD. On the other hand, the HSD can only be used with pairwise comparisons, while the Bonferroni and Dunn-Sidak methods were designed for any (potential) number of comparisons, whether pairwise or not. Response: This comment provides specifics related to a previous comment about balancing the guidance we arrive from both the literature and the simulations. We have sought to achieve this balance in our revision, which includes mention of the reviewer’s point about the Tukey HSD test. 6. The penultimate part of the discussion discusses using MCTs to interpret interactions in complex ANOVA designs. “MCTs can be very useful in disentangling statistical significance and differences among parameter estimates” (Lines 354-5). Many statistical texts would disagree (for example, Quinn and Keough 2002, p252), and I think the alternate approaches should be mentioned, rather than imply this is always the best or only way to untangle interaction effects. The problem here is that there will often be a large number of means (so that the MCTs are not powerful), and the results of the many MCTs can be ambiguous. A simple example is where there is a significant interaction shown by the ANOVA, yet no pairwise comparison is significant. This manuscript itself points out that the MCTs may not be consistent with the ANOVA (lines 311-3). I suggest it makes more sense to try to understand why the interactions occur using interaction plots of means (a line joining the means of level 1 of factor A at each level of factor B, another line joining the means of level 2 of A at each level of B, etc). Three way or more complex designs can be broken down - for example plots of each two-way interaction at each level of a third factor. Another method is to construct sums of squares to test the effect of factor A separately at each level of Factor B (or the effects of factors A and B at each level of C where there is a three-way interaction, etc). These tests would use the mean-square error of the original ANOVA, which is the best estimate of the error variance, with more degrees of freedom (and thus more power) than if the data were separated to run these analyses. Tests of fixed factors would normally be run at each level of a random factor (for example ‘river’ in the example in lines 403-7), but with two fixed factors one might test the effect of A at each level of B, and the effect of B at each level of A. Plots of the means for fixed factors at each level of the other factors would aid in interpretation. As these are exploratory analyses looking for significant effects in a series of tests (especially if there are many levels of a factor), a Bonferroni or Dunn-Sidak adjustment of the significance levels may be appropriate, depending on how the results are reported to the reader. But the number of tests involved would almost certainly be less than MCTs on all the means involved in an interaction. Response: We thank the reviewer for the detailed comment and the thought and effort put into this comment. Comparisons of MCTs with other approaches are lacking, or if available, are not in the same part of the text. Recent MCT information is lacking in general, which was a motivation for this paper. Therefore, we have added a paragraph to provide context and briefly compare MCTs with direct use of parameter estimates, effect sizes, and line plots and how these methods may be complementary. We did not address the construction of individual SS/MSE tests, because this information can be derived from categorical parameter estimates using the sum-to-zero principle. 7. The Abstract does not include some of the most useful conclusions presented in the discussion. I suggest these may include that Planned comparisons are overwhelmingly recommended (lines 262-3) and an unadjusted significance level can be used, that for planned non-parametric comparisons the Mann-Whitney -Wilcoxon U test is recommended, and that if planned comparisons are not used, Scheffe’s test may be used for any linear combination of the means, and Tukey’s HSD, the Bonferroni or the Dunn-Sidak tests are most commonly used for pairwise comparisons of groups. Other tests are recommended for particular types of data. Response: We have revised the abstract to include these useful conclusions, which we agree with the reviewer are important information. Specific comments. Abstract: Lines 19-20: The meaning of: “Due to the variable conditions of the data being analysed” is not clear, and this clause should be rephrased to make the meaning clear to readers. Response: We agree this statement is not as clear as it could be, and have rephrased it to “Due to a variety of data and statistical considerations,...” Line 22: “including >40,000 reports of their use in ecological journals” does not convey any information – there is no time period specified. Perhaps a statement such as: “and we have documented extensive use of MCTs in ecological research” would be better. Response: We agree that this finding needs a temporal component for context. Rather than revise the sentence to become more general, we added the time period information and think this should be a much clearer statement. Line 22: MCT should be MCTs. Response: Revised as suggested. Line 24: “We first reviewed the recommendations on their correct use.” There is nothing about this in the Results, Methods or the Background section of the Introduction. I assume that many of the recommendations in the discussion are taken from this review. I would have expected to see how this review was done – were a range of statistics texts consulted? Or were the initial descriptions of the tests checked to see what they were designed to do? Or what? And some MCTs were designed or recommended for particular situations. The only reference to this is the statement “Principles behind matching a MCT to an experimental design are discussed below.” on line 174 in the Methods. ‘Below’ turns out to be in the Discussion. But this should be in the results before the results of simulations, to help the reader evaluate what the simulation results indicate in terms of what to use. One example is Scheffe’s test – see my comment below. Another is the Tukey-Kramer test, used to replace Tukey’s test when the sample sizes are unequal – see my comment below. Response: The first statement that the reviewer references is simply summarizing that we did not use any a priori knowledge about tests and reviewed literature for MCT intention and application. To reduce any concern about ambiguity of the language, we revised this statement to read, “We first reviewed the published literature for recommendations on their correct use.” Regarding the second statement the reviewer questions, that sentence was removed because we thought it created confusion. We do not consider the principles we reviewed to be simulation results, which is why they were (and remain) outside the simulation results section. Based on a number of other comments, we have added text throughout the revision to better contextualize application of different MCTs. Also, while we are not sure if this was a reviewer comment, we think in the context of this comment is a good place to mention we did make some structural/organizational revisions to the manuscript. Specifically, the literature review results are now given their own subsection in the Results, and their previous omission was only because we considered the entire literature review to be preface to the simulations. However, we think the flow is better when treating the literature review and simulations as methodological investigations, and then evaluating the results of both simultaneously. Introduction: Line 34: The meaning of “Data analyses are crowded with factors of interest from experiments and observations” is not clear. Response: No change was made, because we are not sure what is unclear. This is not a technical statement, but an opening line to the Introduction, which is simply meant to suggest that many data analyses often have numerous factors in them, which suggests that multiple factors are being evaluated simultaneously. We are not resistant to changing this sentence, but without knowing what is unclear, we do not know what we should be changing. Line 51: “multiple comparisons evaluations” should be “the evaluation of multiple comparisons”. Response: Revised as suggested. Line 100: “creates more work for α” is poorly phrased. The non-statistician readers this paper is directed to will not understand what is meant. Response: We thought this phrasing would be more accessible to the non-statisticians the paper is directed at, but we take the reviewers point. We have rephrased it as “...which greatly increases the chances of false positives if α is not adjusted.” Lines 112-113: “The EER reflects the adjustment in error rates to account for multiple comparisons” should be “The EER reflects the adjustment in PCER to account for multiple comparisons” and “ ways to adjust for EER” should be “ways to adjust the PCER” (the EER is a chosen value, so “to adjust for EER” has no obvious meaning.) Response: We thank the reviewer for this clarification. The text has been revised exactly as suggested by the reviewer. Lines 114-5: The statement “EERs that reduce the power to detect differences are known as conservative, while those adjustments that are less strong are known as liberal” simply does not make sense. MCTs that control the EER to below 5% (by strongly reducing PCERs) are known as conservative, while those with less strong adjustments of the PCERs which do not control the EER to at or below 5% are known as liberal. Response: Although we thought we were expressing the same idea, the reviewer has suggested better, more accurate language to make this point, and we have adopted their suggested language. Methods: Lines 121-132: The search method is useful, although it has the limitations described by the authors, and the Table of search results interesting. However, there are other issues here. The frequencies with which these methods are used does not tell us whether they were used for the correct types of data. For example, it would be useful to know how often the 3 common methods used to compare adjusted means after an analysis of covariance, as this is a likely error. Response: The reviewer has identified an interesting point—but also something that is beyond the scope of our study. We agree it would be interesting to know if some of the more common MCTs are applied to the correct data types. However, the most common MCTs are used thousands of times and even reviewing a fraction of that literature would be a substantial undertaking. And while it would produce a very interesting result, we do not see any outcome of that investigation that would fundamentally change the objective of this paper, which is to review and recommend MCTs. Furthermore, we clearly include the results from Ruxton and Beauchamp (2008) that provide a small investigation and results of the type the reviewer has described. Lines 190-192: See my general comments on using normal distributions with equal variances, and on the misrepresentation of type 1 error rates for the SNK in practice, where means are not all equal. Response: Please see our detailed response to this comment above. Lines 192-193: Setting only 1 mean to be different from the rest to gauge type 2 error may not provide a good indication of the differences in power between some tests, as in practice several means may differ in an experiment. Response: We see the reviewer’s point; however, the simulations were designed to represent the common scenario where a control group is different from treatment groups. We have added language in the revision to justify this design. Line 207: “We excluded the Tukey-Kramer test and Dunnett’s test since they are only applicable for special cases”. But the Tukey-Kramer test is used in place of the Tukey test when sample sizes are unequal (‘unbalanced group study designs’). Does this mean that even for the unbalanced designs the authors evaluated Type 1 and 2 errors for the Tukey test? Response: The reviewer is correct to question this statement. We used the TukeyHSD function in R which automatically adjusts the Tukey test to the Tukey-Kramer equation when sample sizes are unequal (see https://github.com/SurajGupta/r-source/blob/master/src/library/stats/R/TukeyHSD.R). So our previous sentence was incorrect and we have modified the sentence to make clear that we only excluded the Dunnett’s test. We already had text discussing that Tukey-Kramer should be used with unequal sample sizes and have modified Table 3 to make clear that the TukeyHSD function does this on its own. Results: Lines 225-6: Treatment with higher sample sizes did not always have lower per comparison error rates than equivalent designs with lower sample sizes. This is not true of the LSMG versus HSMG for balanced designs, nor of the HSFG versus LSFG in unbalanced designs (for most tests). Nor is it true that unbalanced designs reduced the proportion of per comparison type 1 error for all tests – the reverse is true for many tests in the HSFG design and in the LSMG design. Response: The reviewer is correct that there were exceptions to some of the more general trends described. We have modified the text to more thoroughly describe differences between study designs and treatments. Lines 228-231: As I would expect, the SNK does not exceed an error rate of 0.05 in Figure 3. But Fig 3 shows it does not always have a higher error rate than the other tests as indicated here. Further, in the p-value density plots of Figure 4, it is Scheffe’s test that is relatively constant from zero to one (for some sample conditions), not the SNK test, which has a clear peak near 1. This error should be corrected. Response: The reviewer is correct that the SNK never exceeded a PCER of 0.05 and we have clarified this in the text. Further, we have specified that in some instances SNK had a PCER equal to other tests and always had a lower PCER than Duncan’s MRT and unadjusted Fisher’s LSD tests. Finally, the reviewer is correct about Figures 4 and 6. The y-axis labels appear to have been switched during final revisions and this has been corrected. SNK was in fact the test that was relatively constant from zero to one for certain treatments as had been stated in the text. Figure 4 and Figure 6: The four parts of each of these figures are labelled a, b, c, d and thus the caption statement in both figs that “Simulation group abbreviations can be found in the Figure 2 caption” is of no use to the reader. The simulation group abbreviations should be used in place of the a-d. Response: The reviewer is correct and we have added the simulation group abbreviations to the different parts of the figures. Line 231: Scheffe’s S test is designed to test any linear comparison of the means, not just pairwise comparisons, and this is why the test is conservative (reduces the EER below 5%) (and less powerful – see Fig 5) relative to tests designed for use when only pairwise comparisons are considered. It is also why it is coherent with ANOVA results, as ANOVA F tests depend on whether any linear combination of the group means is larger than expected by chance. The reader should be made aware of this in the ‘Background’ section, or when the selection of tests was described (line 203), but also reminded of this when discussing its selection for unplanned comparisons (lines 307-315). It is not clear if the use of this test in the surveyed literature was for its designed purpose, or if it has been frequently used for pairwise comparisons. I assume the frequencies shown in Table 2 represent the sum of either use of the test. Response: The reviewer makes a great point about Scheffe’s S test being designed for linear combinations (note that in the first line of the comment the reviewer uses the term linear comparisons, whereas we think they mean linear combinations). This is a relevant distinction, but one that we did not not make in the original submission. Although linear combinations vs pairwise comparisons may not be a relevant concern for many MCT applications, we do agree that this distinction needs to be made. As such, we have added a short paragraph to our discussion of Scheffe’s S test, in which we outline the conditions behind the test’s usage and generally pull back on our recommendation a little more than in the previous version. Discussion: Lines 254-259 (Parametric or non-parametric data). This section should also deal with the issue of unequal variances among the groups tested, This is a much more frequent problem in practice in terms of deciding whether a parametric general linear model analysis should be used, and non-parametric methods are also subject to an assumption that the distributions in the treatments compared are identical in shape and scale (i.e. except for the median or mean)- see Johnson, D.H. 1995 Ecology 76: 1998-2000. Thus the variances must be similar (on some scale) for the test to function correctly. In other words, the use of non-parametric tests may be useful when the data is far from normally distributed, but they do not solve the problem of unequal variances – unless some transform of the data would equalise the variances but make the distributions very non-normal. There are methods available that cope with unequal variances – there is mention of this later in the discussion, but no advice on when and why they should be used. Response: We have added text that reflects the good and useful comments provided by the reviewer. This text was combined with another revision based on the reviewer’s comment about context, usage, and support for non-parametric MCTs. Lines 272-3: Note that orthogonal contrasts are not completely independent statistically (see the Oaten, 1995 reference above). Response: The reviewer is correct, and we would just like to note that we were writing in generalities for a non-statistical audience. Given this comment—and the next two comments—we thought long and hard about how to deal with the issue of orthogonality. Ultimately, we decided to remove the few mentions of orthogonal contrasts from the manuscript. This was done for a few reasons. 1. While orthogonal is still a term used in the statistical literature, it is not a term widely used anymore among non-statisticians and therefore we thought it would not be a highly relevant term for our intended readership. 2. The concept of orthogonality is useful; however, we find the concept of planned vs. unplanned comparisons to be much more relevant and familiar to ecologists or others that would read this article. Ecologists can understand what a planned vs an unplanned comparison is, but we are less confident they would know or be able to recall the details of orthogonality in a way that would substantially enhance their decision of an MCT test. 3. Orthogonal vs non-orthogonal and planned vs unplanned comparisons are not the same thing. We recognize this and apologize for conflating the two. However, there can be substantial overlap between the principles and practice of both groups. We sought to focus on the similarities in the previous version but realize that we should not equate the two. However, because correctly not equating these groupings would require substantial technical text, we thought that the addition of this text would carry its weight in terms of adding to the manuscript. Again, this helped us determine that the concepts of orthogonality are not quite relevant to our discussion or readership, and we think this part of the manuscript reads much cleaner by removing it. Lines 273-4: Orthogonal contrasts are not necessarily comparisons that do not include the same group in more than one comparison. In the example of A, B, C and D used, A to B, C to D and (A+B)/2 to (C+D)/2 are all orthogonal contrasts. I cannot see the point of such a “loose definition made operational for non-statisticians” (lines 276-7) nor do I see the value of a conflation of orthogonal comparisons and planned comparisons, even if ‘other literature’ has adopted this (I note that no reference is provided here). Response: See the above comment. Our removal of any use and discussion of orthogonal contrasts has resolved this comment. Lines 282-3: “there is little practical difference between planned non-orthogonal comparisons and unplanned non-orthogonal comparisons”. I do see a difference between these, and the key issue is whether the comparisons are planned or not (see Ruxton and Beauchamp 2008) . First, if the limited set of questions that are of interest to the ecologist in setting up the study define comparisons that are not all orthogonal, I see no reason why they should not all be tested using the PCER, provided the logical links between non-orthogonal questions are pointed out to the reader. I see the examination of whether comparisons are orthogonal mainly as a useful check for the researcher, to point out to her/him that the answer to one question may be linked to the answer to others, and as a brake on the number of comparisons made, as the number of potential orthogonal contrasts is the number of groups minus 1. Unplanned comparisons, on the other hand, by definition, may arise from any of the potential ways to compare a set of groups with each other. Only considering pairwise comparisons is one way to reduce the potential number of these unplanned comparisons, yet gain information about where any differences arise among the groups. Response: See the above comment. Our removal of any use and discussion of orthogonal contrasts has resolved this comment. Lines 309-311: See my comments on Scheffe’s test above (Results, line 231). It would be useful here to explain when Scheffe’s test is designed to be useful – when a complex comparison is chosen for testing after the results have been examined. Response: We agree, and refer to our response to the above comment about adding more explanation about Scheffe’s S test. Lines 318-319: Simultaneous testing using the Bonferroni adjustment of the PCER is known to be conservative – that is, it reduces the EER below 5%, and thus it often reduces the PCER below the level that other tests do (see Figure 3). The sequential Bonferroni should be less conservative than the simultaneous test (and the same is true for the sequential Dunn-Sidak test). As shown in Figure 3 the Dunn-Sidak adjustment of PCER is slightly less conservative in some designs, and Tukey’s HSD is less conservative in some cases than the Dunn-Sidak tests. The power of tests is reduced if the tests are more conservative than they need to be, but this is not well shown in Figure 5, because the pattern of differences between the means that was simulated does not show these differences well. One can see however, that the HSD often has slightly lower type 2 error rates than the Bonferroni and Dunn-Sidak tests. Response: We would tend to generally agree with the reviewer on their comment here, but we also do not see a suggestion toward changing anything. We find the figures to be clear, and absent specific guidance on figure clarity, we have kept Figures 3 and 5 largely as is (although with some minor revisions independent of this comment). Table 4: The difference in the 3rd row of the table should have j=2 for the first variable. Response: This subscript has been corrected, and we thank the reviewer for their attention to detail. Table 5: The 9th row repeats the 8th row and should be deleted. There appears to be a 2-1-2 row missing below row 12. Response: Thank you for catching these errors. They have been corrected in the revision. Conclusions: Surely Figure 7 should be mentioned in these conclusions? Response: We agree it should be mentioned and have revised text in the Conclusions to reference Figure 7. I disagree with the thrust of the recommendation regarding Scheffe’s test, which is designed for an unusual situation of a complex comparison, selected from all such potential comparisons after the results are available, and thus constrains the PCER to very low levels and has very low power compared to other MCTs. Several studies show Tukeys HSD is the pairwise test with the best power, so this is the best recommendation for all pairwise contrasts, unless sequential testing is to be done. While the authors may not agree with my view, I think they should explain the ways the MCTs were designed for different types of comparisons. Response: We can agree with the reviewer on this. While we did not remove all recommendations of Scheffe’s S Test (it was kept when referencing linear combinations of means and ANOVA coherence), we reduced the emphasis on recommending this test. Simultaneously, we increased the emphasis on Tukey’s HSD test as a general, robust, all-purpose test that we do recommend. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Whole body vibrations have been used as an exercise modality or as a tool to study neuromuscular integration. There is increasing evidence that longer WBV exposures (up to 10 minutes) induce an acute impairment in neuromuscular function. However, the magnitude and origin of WBV induced fatigue is poorly understood.</ns0:p><ns0:p>Purpose. The study aimed to investigate the magnitude and origin of neuromuscular fatigue induced by half-squat long-exposure whole-body vibration intervention (WBV) with sets of different duration and compare it to non-vibration (SHAM) conditions.</ns0:p><ns0:p>Methods. Ten young, recreationally trained adults participated in six fatiguing trials, each consisting of maintaining a squatting position for several sets of the duration of 30, 60 or 180 seconds. The static squatting was superimposed with vibrations (WBV 30 , WBV 60 , WBV 180 ) or without vibrations (SHAM 30 , SHAM 60 , SHAM 180 ) for a total exercise exposure of 9-minutes in each trial. Maximum voluntary contraction (MVC), level of voluntary activation (%VA), low-(T 20 ) and high-frequency (T 100 ) doublets, low-to-highfrequency fatigue ratio (T 20/100 ) and single twitch peak torque (TW PT ) were assessed before, immediately after, then 15 and 30 minutes after each fatiguing protocol.</ns0:p><ns0:p>Result. Inferential statistics using RM ANOVA and post hoc tests revealed statistically significant declines from baseline values in MVC, T 20 , T 100 , T 20/100 and TW PT in all trials, but not in %VA. No significant differences were found between WBV and SHAM conditions. Magnitude based inference revealed a likely small to medium fatiguing effects in favour of WBV 30 for MVC. Possibly small to likely moderate fatiguing effect in favour of WBV 180 were observed for TW PT , T 20 and T 20/100 .</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion.</ns0:head><ns0:p>Our findings suggest that the origin of fatigue induced by WBV is not significantly different compared to control conditions without vibrations. The lack of significant differences in %VA and the significant decline in other assessed parameters suggest that fatiguing protocols used in this study induced peripheral fatigue of a similar magnitude in all trials. However, trials with longer sets duration (WBV 180 ) were likely to induce a possibly larger magnitude of fatigue compared to SHAM condition.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Whole body vibration (WBV) transfers sinusoidal oscillations into the human body, which inspired the use of this physical modality both as a tool to study the sensorimotor integration of the neuromuscular system and as an intervention stimulus with beneficial effects on performance <ns0:ref type='bibr'>(Rittweger, 2010)</ns0:ref>. Early studies have suggested that short single sessions of WBV of 3 to 5-minutes duration in a squat position immediately increase neuromuscular performance <ns0:ref type='bibr'>(Bosco et al., 2000;</ns0:ref><ns0:ref type='bibr'>Cardinale &amp; Bosco, 2003)</ns0:ref>, maximal voluntary contraction (MVC), jump performance and myoelectric activity <ns0:ref type='bibr'>(Alam, Khan &amp; Farooq, 2018)</ns0:ref>. The acute increase in neuromuscular performance after vibration is referred to as 'post-activation potentiation' (PAP) for short-lasting enhancements (less than 1 minute) and as 'post-activation performance enhancement' (PAPE) for more extended performance enhancement periods lasting up to several hours <ns0:ref type='bibr'>(Blazevich &amp; Babault, 2019)</ns0:ref>. Both phenomena are related to vibration-induced changes in the neuronal control of the affected skeletal muscles that encompass a facilitated central drive <ns0:ref type='bibr'>(Mileva, Bowtell &amp; Kossev, 2009;</ns0:ref><ns0:ref type='bibr'>Krause et al., 2017)</ns0:ref> concomitant with modified reflexive activation at the spinal level <ns0:ref type='bibr'>(Rittweger, Beller &amp; Felsenberg, 2000;</ns0:ref><ns0:ref type='bibr'>Ritzmann et al., 2018)</ns0:ref> persistent over a period of 15 minutes after vibration exposure <ns0:ref type='bibr'>(Krause et al., 2016;</ns0:ref><ns0:ref type='bibr'>Ritzmann et al., 2018)</ns0:ref>. In everyday practice, therapists and practitioners promote longer WBV exposures (up to 10 minutes), although the effects of such exercise modalities are mostly unknown <ns0:ref type='bibr'>(Torvinen et al., 2002;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref>. By increasing the WBV stimuli duration up to a cumulative total of 4 to 10 minutes, it has been suggested that WBV may acutely induce fatigue rather than potentiation <ns0:ref type='bibr'>(Torvinen et al., 2002;</ns0:ref><ns0:ref type='bibr'>de Ruiter et al., 2003;</ns0:ref><ns0:ref type='bibr'>Erskine et al., 2007;</ns0:ref><ns0:ref type='bibr'>Rittweger, 2010;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref>. For example, Torvinen et al. <ns0:ref type='bibr'>(2002)</ns0:ref> and <ns0:ref type='bibr'>de Ruiter et al. (2003)</ns0:ref> observed an immediate decrease of MVC after a 10&#215;1-minute WBV intervention. However, no changes in MVC were observed in the control condition without vibrations. Even though various studies have reported a fatigue-induced drop in neuromuscular performance, there have been contradictory findings regarding the underlying mechanisms which favour either a central or peripheral origin. Several authors investigated the effect of WBV on central fatigue <ns0:ref type='bibr'>(Jordan et al., 2010;</ns0:ref><ns0:ref type='bibr'>Maffiuletti et al., 2013;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref> and were unable to find any difference in the level of voluntary activation (%VA) between interventions with and without vibrations. To the best of our knowledge, the force-frequency fatigue-related mechanisms of WBVinduced peripheral fatigue have not been studied. By comparing the ratio of the electrically induced mechanical responses using low-frequency (below fusion frequency -20 Hz) and high-frequency (above fusion frequency -100 Hz) paired supramaximal electrical stimuli, peripheral fatigue can be subdivided into low-and high-frequency <ns0:ref type='bibr'>(Edwards, 2008;</ns0:ref><ns0:ref type='bibr'>Millet et al., 2011)</ns0:ref>. Analogous exercise-induced fatigue studies have demonstrated that prolonging exercise stimuli can shift the peripheral fatiguing mechanism towards low-frequency fatigue <ns0:ref type='bibr'>(Millet &amp; Lepers, 2004;</ns0:ref><ns0:ref type='bibr'>Tomazin et al., 2012)</ns0:ref>.</ns0:p><ns0:p>To better understand the intervention stimuli induced by WBV, it is crucial to establish which fatiguing mechanisms occur after a single session of WBV, and how different vibration parameters affect the magnitude and origin of neuromuscular fatigue. The scientific and practitioner choices for WBV intervention are motivated by achieving high superimposed effects throughout WBV to trigger physiological and neuromuscular adaptations and thus, WBV parameters are combined accordingly <ns0:ref type='bibr' target='#b0'>(Abercromby et al., 2007;</ns0:ref><ns0:ref type='bibr'>Ritzmann, Gollhofer &amp; Kramer, 2013)</ns0:ref>. Electromyography studies suggest that side-alternating vibration exposure driven by high amplitude and frequency cause the highest activation intensities in distal and proximal leg musculature <ns0:ref type='bibr' target='#b0'>(Abercromby et al., 2007;</ns0:ref><ns0:ref type='bibr'>Rittweger, 2010;</ns0:ref><ns0:ref type='bibr'>Ritzmann, Gollhofer &amp; Kramer, 2013)</ns0:ref>. In addition to vibrationassociated attributes, and in an analogy to strength training, the training load is mainly determined by intensity and volume <ns0:ref type='bibr'>(Baechle &amp; Earle, 2008)</ns0:ref>. Therefore, volume is subdivided into number of set and repetitions with defined set duration <ns0:ref type='bibr'>(Campbell et al., 2017)</ns0:ref>. In a similar manner, vibration amplitude and frequency define the training intensity in WBV interventions. However, to the best of our knowledge, there is a lack of studies investigating how WBV intervention volume (set numbers and set duration) affects the occurrence of neuromuscular fatigue. Therefore, the aim of the present study was to investigate the magnitude and origin of neuromuscular fatigue induced by long-exposure half-squat whole-body vibration intervention (WBV) with sets of different duration and compare it with non-vibration (SHAM) conditions. Thus, and with reference to, previous research involving longexposure WBV induced fatigue <ns0:ref type='bibr'>(Erskine et al., 2007;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref> we selected a long (cumulative exercise time of 9 minutes) static WBV fatiguing intervention divided into sets of different duration (30 s, 60 s or 180 s). In a series of MVC paradigms, we applied different peripheral nerve stimulation techniques, allowing us to distinguish the source of fatigue. We hypothesised that WBV exercise interventions would cause higher magnitudes of fatigue compared to non-vibration intervention <ns0:ref type='bibr'>(Erskine et al., 2007;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref>. We expected that fatigue magnitude would be dependent on the duration of exposure and would increase with set-duration. We hypothesised that predominantly peripheral, rather than central fatiguing mechanisms, would be causally involved <ns0:ref type='bibr'>(Jordan et al., 2010;</ns0:ref><ns0:ref type='bibr'>Maffiuletti et al., 2013;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods Study design</ns0:head><ns0:p>In a cross-over repeated measures design, each subject performed three different fatiguing exercise interventions with WBV and three exercise interventions in a sham condition without WBW (SHAM) to determinate the effect of WBV (Fig. <ns0:ref type='figure' target='#fig_5'>1A</ns0:ref>.). Each intervention comprised a cumulative exercise period with a duration of 9 minutes divided into different sets (either 18 x 30 s or 9 x 60 s or 3 x 180 s) with 120 s rest between sets (Fig <ns0:ref type='figure' target='#fig_5'>1A</ns0:ref>). The exercise interventions were performed on an activated vibration platform (WBV 30 , WBV 60 , WBV 180 ) and three on an inactive vibration platform (SHAM 30 , SHAM 60 , SHAM 180 ). Each intervention was executed on different visits with at least seven days rest in-between. The order was randomised. The subjects were not permitted to undertake explosive strength training or fatiguing workouts for 48 hours before each measuring day, in order to eliminate side-effects. The study design, materials and neuromuscular assessments are available for reference in protocols.io (dx.doi.org/10.17504/protocols.io.beadjaa6) Neuromuscular assessment in the resting position was performed at t 0 (baseline) prior to exercise intervention. The assessment consisted maximum voluntary contraction (MVC) of the knee extensors, interpolated with a high frequency (T MVC ) twitch (10 ms interstimuli interval), followed 3 s later by a 100 Hz doublet (T 100 ), followed 3 s later by a 20 Hz (50 ms interstimuli interval) doublet (T 20 ), and 3 s later by a potentiated single twitch (TW). The assessment procedure was executed according to <ns0:ref type='bibr'>(Millet et al., 2011)</ns0:ref> and repeated at 1 minute (t f ), as well as at 15 (t f15 ) and 30 minutes (t f30 ) after the final 9minute intervention. All neuromuscular assessments were performed on the right leg. Subjects Ten healthy subjects (6 men and 4 women; age: 21.1 1.41 years, mass: 77.8 The sample size was estimated by means of a power analysis aiming to detect large effect sizes (f = 1.2; alpha = 0.05; power = 0.80).</ns0:p></ns0:div> <ns0:div><ns0:head>Intervention</ns0:head><ns0:p>The interventions were performed on a side-alternating vibration platform (Galileo Fit, Novotec Medical, Germany) which was running at a frequency of 26 Hz <ns0:ref type='bibr'>(Rittweger, Mutschelknauss &amp; Felsenberg, 2003;</ns0:ref><ns0:ref type='bibr'>Cochrane et al., 2010)</ns0:ref> and off, respectively, for WBV and SHAM conditions. Subjects were instructed to maintain a half-squat position with their knees flexed at an angle of 60&#176; <ns0:ref type='bibr'>(Ritzmann et al., 2010)</ns0:ref> for several sets with 2minute rest between sets. Kinematics were controlled with a goniometer. The subjects stood with their feet 40 cm apart at a point where the tilting platform reached peak-topeak displacement amplitude of 5 mm <ns0:ref type='bibr'>(Ritzmann, Gollhofer &amp; Kramer, 2013)</ns0:ref>.</ns0:p><ns0:p>At the beginning of each session, subjects underwent a 6-minute warm-up routine consisting of bench stepping (20 cm high) at a frequency of 0.5 Hz, swapping the leading leg at one minute intervals.</ns0:p></ns0:div> <ns0:div><ns0:head>Testing protocols</ns0:head><ns0:p>During the neuromuscular assessment, the subjects remained seated in a custom-built isometric knee extension apparatus equipped with a force transducer (MES, Maribor, Slovenia) <ns0:ref type='bibr'>(Tomazin, Dolenec &amp; Strojnik, 2008;</ns0:ref><ns0:ref type='bibr'>Garc&#237;a-Ramos et al., 2016)</ns0:ref>. The force transducer was calibrated prior to testing sessions. Each subject was seated in an upright position, hip at 100&#176; and trunk leaning against the backrest of the testing apparatus, fixed by straps over the pelvis and a horizontal pad over the distal third of the thigh. The knee joint axis was aligned with the mechanical axis of the dynamometer.</ns0:p><ns0:p>The shin pad was placed just superior to the medial malleolus. The right knee joint was fixed at a 60&#176; angle (0&#176; = full extension) (Fig. <ns0:ref type='figure' target='#fig_5'>1C</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Femoral nerve electrical stimulation</ns0:head><ns0:p>The femoral nerve was stimulated by pressing a monopolar cathode (10-mm in diameter, Ag-AgCl, Type 0601000402, Controle Graphique Medical, Brie-Comte-Robert, France) into the femoral triangle of the iliac fossa (Fig. <ns0:ref type='figure' target='#fig_5'>1C</ns0:ref>). A larger (102mm x 52mm, Compex, SA, Ecublens, Switzerland) self-adhesive electrode placed over the gluteal fold served as the anode. Electrical impulses (single, square wave, 1-ms duration) elicited by a high voltage constant current electrical stimulator (DS7A; Digitimer, Hertfordshire, UK) were used to trigger the muscle response, which was detected as a change in torque of the knee extensors. The stimulation intensity to elicit maximum knee extensors isometric twitch was determined in each subject at the beginning of each trial and maintained for the entire trial. Starting from an intensity of 10 mA, the stimulation intensity was progressively increased by 10 mA until no further increase in torque was observed despite further increment in electrical current. The current at maximal twitch torque was additionally increased by a factor of 1.5 to obtain a supramaximal stimulus <ns0:ref type='bibr'>(Verges et al., 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Single twitch</ns0:head><ns0:p>The torque change induced by a single supramaximal femoral nerve stimulus <ns0:ref type='bibr'>(Place et al., 2007)</ns0:ref> was analysed to obtain the peak torque value (TW PT ).</ns0:p></ns0:div> <ns0:div><ns0:head>High-and low-frequency doublets</ns0:head><ns0:p>The torque change induced by the paired high-frequency (100 Hz, i.e. 10-ms interstimuli interval) and low-frequency (20 Hz, i.e. 50-ms interstimuli interval) supramaximal electrical stimuli <ns0:ref type='bibr'>(Place et al., 2007;</ns0:ref><ns0:ref type='bibr'>Verges et al., 2009)</ns0:ref> was analysed to obtain the following parameters: peak torque from 100 Hz doublet (T 100 ), peak torque from 20 Hz doublet (T 20 ). In addition, the low-to high-frequency ratio (T 20/100 ) was calculated using the following formula:</ns0:p><ns0:formula xml:id='formula_0'>&#119879; 20 100 = &#119879; 20 &#119879; 100 * 100</ns0:formula><ns0:p>This ratio was used as a surrogate of low-to high-frequency tetanic stimulation <ns0:ref type='bibr'>(Verges et al., 2009)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Maximal voluntary contraction with double twitch interpolated techniques</ns0:head><ns0:p>Subjects were asked to perform a 5 s maximal isometric voluntary knee extension <ns0:ref type='bibr'>(Verges et al., 2009)</ns0:ref>. The signal was smoothed using a 0.5 s window moving average filter and peak torque (MVC) was retained for analysis. The double twitch interpolated technique <ns0:ref type='bibr'>(Allen, Gandevia &amp; McKenzie, 1995)</ns0:ref> was performed by superimposing a 100 Hz doublet on the isometric plateau (T MVC ). A second analogous stimulation (T 100 ) on the relaxed muscle followed after 3 seconds (Fig. <ns0:ref type='figure' target='#fig_5'>1B</ns0:ref>). The ratio of the amplitude of the T MVC over T 100 was then calculated to obtain the level of voluntary activation (%VA):</ns0:p><ns0:formula xml:id='formula_1'>%&#119881;&#119860; = ( 1 -&#119879; &#119872;&#119881;&#119862; -&#119872;&#119881;&#119862; &#119879; 100</ns0:formula><ns0:p>) * 100 Statistics A two-way factorial ANOVA (Type III) was conducted in R(3.5.1) with the afex package <ns0:ref type='bibr'>(Singmann et al., 2018)</ns0:ref> to compare the main effects of time (t 0 , t f , t f15 , t f30 ) and trial (WBV 30 , WBV 60 , WBV 180 , SHAM 30 , SHAM 60 , SHAM 180 ) and the interaction effect of time trial. Generalised eta squared ( ) effect sizes were calculated for the ANOVA main</ns0:p><ns0:formula xml:id='formula_2'>&#215; &#120578; 2</ns0:formula><ns0:p>&#119866; and interaction effects. In the case of statistically significant interactions, post hoc comparisons with Sidak corrections were applied using the emmeans package <ns0:ref type='bibr'>(Lenth et al., 2018)</ns0:ref> in order to compare WBV and SHAM condition. Tukey-corrected pairwise post hoc tests were used to calculate differences to baseline within trials.</ns0:p><ns0:p>In addition to inference statistics, standardised changes in the mean of each measure were used to assess the magnitudes of effect (ES) between WBV and SHAM conditions of the same set duration (e.g. SHAM 30 -WBV 30 , SHAM 60 -WBV 60 , SHAM 180 -WBV 180 ) and were then calculated using Cohen d. </ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:50532:1:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Descriptive statistics for MVC and %VA are displayed in Table <ns0:ref type='table'>1</ns0:ref>; descriptive statistics for T 20 , T 100 and T 20/100 , TW PT are listed in Table <ns0:ref type='table' target='#tab_3'>2</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Maximum voluntary contraction</ns0:head><ns0:p>There was a statistically significant time effect (F (3, 27) = 24.40, p &lt; 0.001, = 0.02), &#120578; 2 &#119866; but no significant trial effect (F (5, 45) = 2.13, p = 0.08, = 0.01) nor trial x time &#120578; 2 &#119866; interaction effect (F (15, 135) = 0.60, p = 0.87, = 0.002) for MVC. Within-trial post hoc &#120578; 2 &#119866; tests showed differences between baseline and post-assessments (Fig. <ns0:ref type='figure' target='#fig_6'>2a</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Level of voluntary activation (%VA)</ns0:head><ns0:p>There was a statistically significant time (F (3, 27) = 3.67, p = 0.024, = 0.02) and trial &#120578; 2 &#119866; (F (5, 45) = 2.52, p = 0.042, = 0.08) effect, but no trial x time interaction (F (15, 135) = &#120578; 2 &#119866; 1,21, p = 0.26, = 0.03) for %VA. Post hoc tests did not reveal significant differences &#120578; 2 &#119866; between baseline and post-assessments (Fig. <ns0:ref type='figure' target='#fig_6'>2b</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Peripheral fatigue</ns0:head><ns0:p>There was a significant time effect (F (3, 27) = 64.43, p &lt; 0.001, = 0.25) for T 20 . Trial &#119866; revealed significant differences between baseline and post-assessments for each of the trials (Fig. <ns0:ref type='figure'>3a</ns0:ref>, Table <ns0:ref type='table'>3</ns0:ref>). There was a significant time effect (F (3, 27) = 60.33, p &lt; 0.001, = 0.15) for T 100 . Trial <ns0:ref type='table'>3</ns0:ref>). There was a significant time effect (F (3, 27) = 46.33, p &lt; 0.001, = 0.17) for T20/100. &#119866; revealed significant differences between baseline and post-assessments for each of the trials (Fig. <ns0:ref type='figure'>3c, Table 3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Single twitch</ns0:head><ns0:p>There was a significant time effect (F (3, 27) = 48.80, p &lt; 0.001, = 0.23). Trial effects &#119866; revealed significant differences between baseline and post-assessments for each of the trials (Fig. <ns0:ref type='figure' target='#fig_6'>2c</ns0:ref>, Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50532:1:0:NEW 2 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The current study aimed to investigate the magnitude and origin of neuromuscular fatigue induced by long-exposure half-squat whole-body vibration intervention (WBV) with sets of different duration and compare it with non-vibration (SHAM) conditions. Our findings revealed no superimposed effect of WBV compared to control conditions without vibrations.</ns0:p></ns0:div> <ns0:div><ns0:head>Maximal voluntary contraction</ns0:head><ns0:p>Knee extensors MVC torque dropped by 7 to 12% after each fatiguing protocol, which is in line with other WBV induced fatigue studies, where MVC torque decreased by approximately 8% <ns0:ref type='bibr'>(de Ruiter et al., 2003;</ns0:ref><ns0:ref type='bibr'>Erskine et al., 2007;</ns0:ref><ns0:ref type='bibr'>Colson et al., 2009;</ns0:ref><ns0:ref type='bibr'>Zory et al., 2013)</ns0:ref>. Only Maffiuletti et al. ( <ns0:ref type='formula'>2013</ns0:ref>) reported a more substantial decline in MVC (-23%), which is likely associated with the application of additional loads coupled with shorter inter-set rest periods compared to other studies and to our specific experimental setting. This finding is in contrast with our hypothesis that longer set duration exercises superimposed with vibration (WBV 180 ) would produce greater fatigue compared to SHAM 180 condition. However, it has been previously suggested that potentiated electrically elicited supramaximal doublets represent a more suitable indicator of peripheral fatigue and contractile impairments compared to MVC torque <ns0:ref type='bibr'>(Place et al., 2007)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Central fatigue</ns0:head><ns0:p>The level of voluntary activation (%VA) of the knee extensors was not significantly depressed by any intervention utilised in this study, which suggest that mechanisms located in the central nervous system (CNS) were not significantly involved in the decline of MVC. These findings are in line with Colson et al. ( <ns0:ref type='formula'>2009</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Peripheral fatigue</ns0:head><ns0:p>To the best of our knowledge, this is the first study where electrically elicited supramaximal low-and high-frequency doublets were used to assess the origin and magnitude of peripheral fatigue after WBV exposure. For all protocols, T 20 was more affected than T 100 leading to a decreased T 20/100 ratio (Fig. <ns0:ref type='figure'>3c</ns0:ref>). These declines suggest the occurrence of low-frequency fatigue (LFF) in all trials. It is noteworthy that the T 20/100 ratio for SHAM interventions returned to baseline values 15 minutes after the intervention, while WBV interventions remained significantly depressed up to 30 minutes after the intervention. This suggests that LFF is stronger and more long-lasting when a WBV exercise is executed with an emphasis on exposing to longer sets of vibration. The observation favouring LFF as an underlying mechanism can additionally be supported by the findings obtained from single twitch data. Similar to T 20 and T 100 , TW PT progressively decreased as the intervention continued.</ns0:p></ns0:div> <ns0:div><ns0:head>Underlying mechanisms</ns0:head><ns0:p>The lack of difference between WBV and SHAM conditions observed in this study suggest that no beneficial effects on neuromuscular function exist when using superimposed WBV. This is particularly true for MVC and the level of voluntary activation. Even though some studies reported that WBV can induce modulation in the neuronal control, which is manifested as a facilitated central drive <ns0:ref type='bibr'>(Mileva, Bowtell &amp; Kossev, 2009;</ns0:ref><ns0:ref type='bibr'>Krause et al., 2016)</ns0:ref> this does not translate into central fatigue. Furthermore, the decline in low-and high-frequency doublets, as well as single twitch torque, suggests that a mechanism underlying the decrease in force production in both WBV and SHAM treatments is an impairment in Ca 2+ handling. This is followed by a gradual recovery of the Ca 2+ depletion within the 15-30 min following WBV equal to the SHAM intervention. Underlying cellular fatiguing mechanisms explaining the results for SHAM and WBV may refer to three aspects <ns0:ref type='bibr'>(Westerblad et al., 2000;</ns0:ref><ns0:ref type='bibr'>Allen &amp; Westerblad, 2001;</ns0:ref><ns0:ref type='bibr'>Williams &amp; Ratel, 2009</ns0:ref>): a) since doublet peak torques progressively dropped at low-and high-frequencies of stimulation, there could be direct inhibition of inorganic phosphates (P i ) on Ca 2+ , thereby producing an impairment in the cross-bridge force generation <ns0:ref type='bibr'>(Millar &amp; Homsher, 1990</ns0:ref>). However, it is unlikely that this mechanism alone accounts for low-frequency fatigue <ns0:ref type='bibr'>(Allen, Lannergren &amp; Westerblad, 1995)</ns0:ref>. b) It is likely that the larger drop in T 20 compared to T 100 could indicate a precipitation in Ca 2+ -P i in the sarcoplasmic reticulum, leading to a decrease in free Ca 2+ available for release <ns0:ref type='bibr'>(Allen &amp; Westerblad, 2001)</ns0:ref>. In addition, c) reduced myofibrillar Ca 2+ sensitivity can also affect force production <ns0:ref type='bibr'>(Bruton et al., 2008)</ns0:ref>. Both mechanisms (b and c) have little impact on force production at high frequencies but a large effect on low frequencies <ns0:ref type='bibr'>(Westerblad et al., 2000)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Limitations</ns0:head><ns0:p>The study might have some limitations. An important limitation of this study (similar to the majority of other vibration studies) is the lack of WBV load normalisation, as this may have considerable side-effects on the results, as was demonstrated by Di <ns0:ref type='bibr'>Giminiani et al. (2009)</ns0:ref>. Another limiting aspect deals with different work/rest ratios between long sets (180 s work -120 s rest) compared to other shorter set durations (30 s -120 s and 60 s -120 s). There is a great diversity in scientific and practitioner protocols and therefore, future studies should consider the variability in work/rest ratios and duration sets within the experimental design. Furthermore, the experiment was executed in the right leg only. The leg dominance has thereby not been considered as a variable of influence on fatigue and fatigue mechanisms.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The outcomes of this study suggest the origin of fatigue induced by half-squat with superimposed vibrations is no different from the control conditions without vibrations. Due to a lack of significant modulation of voluntary activation, it can be assumed that the fatiguing protocols used in this study predominantly affected peripheral mechanisms rather than central ones. The primary induced peripheral fatiguing mechanism seems to find its origin in low-frequency fatigue which most probably involves Ca 2+ handling. The outcomes of this investigation seems to suggest that static squat with superimposed whole-body vibrations does not represent a larger fatiguing stimulus compared to static squat alone in recreationally active athletes. WBV differences from baseline (&#9650;&#9650;&#9650; p &lt; 0.001; &#9650;&#9650; p &lt; 0.01; &#9650; p &lt; 0.05). White circles represent statistically significant SHAM differences from baseline (&#9675;&#9675;&#9675; p &lt; 0.001; &#9675;&#9675; p &lt; 0.01; &#9675; p &lt; 0.05).</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 3</ns0:head><ns0:p>Relative changes from baseline.</ns0:p><ns0:p>(A) low-frequency doublet (T 20 ), (B) high-frequency doublet (T 100 ) and (C) low-high torque frequency ratio (T 20/100 ) for WBV (connected black triangles) and SHAM (connected white circles) for trials with different set durations (30 s, 60 s and 180 s). Values are expressed as mean and standard errors. Black triangles represent statistically significant WBV differences from baseline (&#9650;&#9650;&#9650; p &lt; 0.001; &#9650;&#9650; p &lt; 0.01; &#9650; p &lt; 0.05). White circles represent statistically significant SHAM differences from baseline (&#9675;&#9675;&#9675; p &lt; 0.001; &#9675;&#9675; p &lt; 0.01; &#9675; p &lt; 0.05). Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>45) = 1.91, p = 0.11, = 0.03) and trial x time interaction effects (F (15, &#120578; 2 &#119866; 135) = 0.90, p = 0.56, = 0.007) remained statistically insignificant. Post hoc tests &#120578; 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>45) = 2.15, p = 0.07, = 0.03) and trial x time interaction effect (F (15, 135) &#120578; 2 &#119866; = 0.43, p = 0.97, = 0.002) remained statistically insignificant. Post hoc tests revealed &#120578; 2 &#119866; significant differences between baseline and post-assessments for each of the trials (Fig. 3b, Table</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>&#120578; 2 &#119866;</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Trial effect (F (5, 45) = 1.06, p = 0.40, = 0.02) and trial x time interaction effect (F (15,&#120578; 2 &#119866; 135) = 0.97, p = 0.49, = 0.02) remained statistically insignificant. Post hoc tests &#120578; 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>45) = 0.86, p = 0.52, = 0.006) and trial x time interaction effect (F (15, 135) = &#120578; 2 &#119866; 1.05, p = 0.41, = 0.006) remained statistically insignificant for TW PT . Post hoc tests &#120578; 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>2009) and Jordan et al. (2010) but in contrast to de Ruiter et al. (2003) who reported a vibration-induced decline in knee extensors voluntary activation. Despite de Ruiter et al. (2003) reported a similar drop in %VA compared to the present study (approx. 4%), any difference in interpretation between the two studies could be biased by the lack of a control group or control condition in Ruiter et al. (2003) experiment coupled with the eligibility criteria for volunteers: in our study the population consisted of recreationally trained athletes and de Ruiter's et al. (2003) enrolled untrained students. Evidence exist for physiological differences and perquisites in motor control between sedentary and trained active subjects (Buford &amp; Manini, 2010). Being hypothesis-driven, our findings indicate that there are no evident superimposed effects of WBV on central fatiguing mechanisms compared to control conditions without WBV. This should be taken into consideration when designing exercise programs or research studies which intend to induce centralfatigue. As such, WBV superimposed exercises are unlikely to be more effective than maintaining a static squat alone.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>95% CI] Maximum voluntary contraction (MVC)</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell /><ns0:cell>t 0</ns0:cell><ns0:cell /><ns0:cell>t f</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>t f15</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>t f30</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='11'>mean (SD) Cohen d [SHAM 30 mean (SD) % &#120607; Cohen d [95% CI] mean (SD) % &#120607; Cohen d [95% CI] mean (SD) % &#120607; 206.04 (67.09) 190.31 (68.15) -7.63 -0.21 [-0.31, -0.11] 188.93 (68.26) -8.30 -0.23 [-0.37, -0.09] 191.15 (63.97) -7.23 -0.21 [-0.35, -0.06]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 30</ns0:cell><ns0:cell>219.46 (64.63)</ns0:cell><ns0:cell>191.82 (55.25)</ns0:cell><ns0:cell>-12.59</ns0:cell><ns0:cell>-0.42 [-0.66, -0.17]</ns0:cell><ns0:cell>193.06 (67.16)</ns0:cell><ns0:cell>-12.03</ns0:cell><ns0:cell>-0.36 [-0.54, -0.18]</ns0:cell><ns0:cell>192.23 (64.92)</ns0:cell><ns0:cell>-12.41</ns0:cell><ns0:cell>-0.38 [-0.54, -0.23]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 60</ns0:cell><ns0:cell>215.61 (64.31)</ns0:cell><ns0:cell>201.05 (63.94)</ns0:cell><ns0:cell>-6.75</ns0:cell><ns0:cell>-0.21 [-0.43, 0.02]</ns0:cell><ns0:cell>199.34 (61.18)</ns0:cell><ns0:cell>-7.55</ns0:cell><ns0:cell>-0.23 [-0.44, -0.03]</ns0:cell><ns0:cell>200.52 (62.06)</ns0:cell><ns0:cell>-7.00</ns0:cell><ns0:cell>-0.22 [-0.45, 0.01]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 60</ns0:cell><ns0:cell>209.75 (63.53)</ns0:cell><ns0:cell>194.58 (56.78)</ns0:cell><ns0:cell>-7.23</ns0:cell><ns0:cell>-0.23 [-0.46, 0.00]</ns0:cell><ns0:cell>201.37 (57.87)</ns0:cell><ns0:cell>-3.99</ns0:cell><ns0:cell>-0.12 [-0.34, 0.09]</ns0:cell><ns0:cell>197.64 (56.89)</ns0:cell><ns0:cell>-5.77</ns0:cell><ns0:cell>-0.18 [-0.38, 0.01]</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>SHAM 180 207.81 (62.38)</ns0:cell><ns0:cell>186.06 (51.73)</ns0:cell><ns0:cell>-10.47</ns0:cell><ns0:cell>-0.34 [-0.53, -0.16]</ns0:cell><ns0:cell>189.68 (52.72)</ns0:cell><ns0:cell>-8.72</ns0:cell><ns0:cell>-0.28 [-0.49, -0.08]</ns0:cell><ns0:cell>188.45 (57.65)</ns0:cell><ns0:cell>-9.32</ns0:cell><ns0:cell>-0.29 [-0.47, -0.11]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 180</ns0:cell><ns0:cell>200.36 (62.85)</ns0:cell><ns0:cell>175.58 (53.91)</ns0:cell><ns0:cell>-12.37</ns0:cell><ns0:cell>-0.38 [-0.68, -0.09]</ns0:cell><ns0:cell>185.85 (61.52)</ns0:cell><ns0:cell>-7.24</ns0:cell><ns0:cell>-0.21 [-0.44, 0.02]</ns0:cell><ns0:cell>177.76 (65.08)</ns0:cell><ns0:cell>-11.28</ns0:cell><ns0:cell>-0.32 [-0.66, 0.02]</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Level of voluntary activation (%VA)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SHAM 30</ns0:cell><ns0:cell>93.05 (3.00)</ns0:cell><ns0:cell>89.94 (6.14)</ns0:cell><ns0:cell>-3.34</ns0:cell><ns0:cell>-0.58 [-1.22, 0.05]</ns0:cell><ns0:cell>89.57 (7.14)</ns0:cell><ns0:cell>-3.74</ns0:cell><ns0:cell>-0.58 [-1.34, 0.19]</ns0:cell><ns0:cell>90.12 (4.47)</ns0:cell><ns0:cell>-3.15</ns0:cell><ns0:cell>-0.40 [-0.72, 0.52]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 30</ns0:cell><ns0:cell>90.74 (3.98)</ns0:cell><ns0:cell>89.47 (5.07)</ns0:cell><ns0:cell>-1.41</ns0:cell><ns0:cell>-0.25 [-0.97, 0.46]</ns0:cell><ns0:cell>89.49 (4.58)</ns0:cell><ns0:cell>-1.38</ns0:cell><ns0:cell>-0.26 [-0.87, 0.34]</ns0:cell><ns0:cell>91.85 (4.21)</ns0:cell><ns0:cell>1.22</ns0:cell><ns0:cell>0.25 [-0.39, 0.88]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 60</ns0:cell><ns0:cell>89.27 (4.78)</ns0:cell><ns0:cell>87.71 (6.00)</ns0:cell><ns0:cell>-1.75</ns0:cell><ns0:cell>-0.26 [-0.78, 0.26]</ns0:cell><ns0:cell>88.14 (5.32)</ns0:cell><ns0:cell>-1.27</ns0:cell><ns0:cell>-0.20 [-0.74, 0.34]</ns0:cell><ns0:cell>87.52 (5.39)</ns0:cell><ns0:cell>-1.96</ns0:cell><ns0:cell>-0.31 [-0.85, 0.23]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 60</ns0:cell><ns0:cell>90.20 (4.85)</ns0:cell><ns0:cell>87.41 (5.96)</ns0:cell><ns0:cell>-3.09</ns0:cell><ns0:cell>-0.31 [-0.70, 0.55]</ns0:cell><ns0:cell>91.33 (4.29)</ns0:cell><ns0:cell>1.26</ns0:cell><ns0:cell>0.17 [-0.59, 0.67]</ns0:cell><ns0:cell>89.66 (4.41)</ns0:cell><ns0:cell>-0.59</ns0:cell><ns0:cell>-0.03 [-0.64, 0.62]</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>SHAM 180 87.40 (6.82)</ns0:cell><ns0:cell>84.36 (5.47)</ns0:cell><ns0:cell>-3.47</ns0:cell><ns0:cell>-0.45 [-1.08, 0.19]</ns0:cell><ns0:cell>88.73 (5.06)</ns0:cell><ns0:cell>1.52</ns0:cell><ns0:cell>0.20 [-0.55, 0.95]</ns0:cell><ns0:cell>87.45 (6.01)</ns0:cell><ns0:cell>0.05</ns0:cell><ns0:cell>0.01 [-0.75, 0.77]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 180</ns0:cell><ns0:cell>88.54 (5.36)</ns0:cell><ns0:cell>87.48 (4.31)</ns0:cell><ns0:cell>-1.20</ns0:cell><ns0:cell>-0.17 [-0.67, 0.59]</ns0:cell><ns0:cell>86.00 (6.28)</ns0:cell><ns0:cell>-2.87</ns0:cell><ns0:cell>-0.40 [-1.03, 0.24]</ns0:cell><ns0:cell>87.43 (4.33)</ns0:cell><ns0:cell>-1.25</ns0:cell><ns0:cell>-0.21 [-0.98, 0.57]</ns0:cell></ns0:row></ns0:table><ns0:note>1PeerJ reviewing PDF | (2020:06:50532:1:0:NEW 2 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Descriptive statistics (mean and SD), within trial relative change from baseline and Cohen d effects size for T 20 , T 100 , T 20/100 and TW PT .</ns0:figDesc><ns0:table><ns0:row><ns0:cell>t 0 , baseline; t f , after intervention; t f15 , 15 minutes after intervention; t f30 , 30 minutes after</ns0:cell></ns0:row><ns0:row><ns0:cell>intervention.</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:06:50532:1:0:NEW 2 Oct 2020)Manuscript to be reviewed</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>95% CI] Low-frequency doublet (T 20 )</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>SHAM 30</ns0:cell><ns0:cell>75.26 (18.37)</ns0:cell><ns0:cell>57.14 (12.56)</ns0:cell><ns0:cell>-24.07</ns0:cell><ns0:cell>-1.04 [-1.51, -0.58]</ns0:cell><ns0:cell>56.97 (11.84)</ns0:cell><ns0:cell>-24.30</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>57.87 (10.33)</ns0:cell><ns0:cell>-23.11</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 30</ns0:cell><ns0:cell>80.47 (20.30)</ns0:cell><ns0:cell>59.70 (13.91)</ns0:cell><ns0:cell>-25.80</ns0:cell><ns0:cell>-1.08 [-1.43, -0.73]</ns0:cell><ns0:cell>59.47 (15.47)</ns0:cell><ns0:cell>-26.10</ns0:cell><ns0:cell>-1.05 [-1.39, -0.72]</ns0:cell><ns0:cell>60.67 (15.04)</ns0:cell><ns0:cell>-24.61</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 60</ns0:cell><ns0:cell>82.03 (22.79)</ns0:cell><ns0:cell>62.84 (17.27)</ns0:cell><ns0:cell>-23.39</ns0:cell><ns0:cell>-0.86 [-1.45, -0.27]</ns0:cell><ns0:cell>66.78 (19.18)</ns0:cell><ns0:cell>-18.59</ns0:cell><ns0:cell>-0.66 [-1.34, 0.03]</ns0:cell><ns0:cell>62.78 (15.23)</ns0:cell><ns0:cell>-23.46</ns0:cell><ns0:cell>-0.90 [-1.39, -0.41]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 60</ns0:cell><ns0:cell>76.11 (18.34)</ns0:cell><ns0:cell>56.88 (17.20)</ns0:cell><ns0:cell>-25.26</ns0:cell><ns0:cell>-0.98 [-1.44, -0.52]</ns0:cell><ns0:cell>57.06 (17.68)</ns0:cell><ns0:cell>-25.03</ns0:cell><ns0:cell>-0.96 [-1.34, -0.58]</ns0:cell><ns0:cell>55.79 (17.33)</ns0:cell><ns0:cell>-26.70</ns0:cell><ns0:cell>-1.03 [-1.52, -0.54]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 180</ns0:cell><ns0:cell>82.24 (17.44)</ns0:cell><ns0:cell>60.38 (12.87)</ns0:cell><ns0:cell>-26.58</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>60.35 (13.30)</ns0:cell><ns0:cell>-26.62</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>60.86 (13.15)</ns0:cell><ns0:cell>-26.00</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 180</ns0:cell><ns0:cell>79.59 (18.84)</ns0:cell><ns0:cell>53.86 (16.14)</ns0:cell><ns0:cell>-32.33</ns0:cell><ns0:cell>-1.33 [-1.69, -0.97]</ns0:cell><ns0:cell>53.62 (17.56)</ns0:cell><ns0:cell>-32.63</ns0:cell><ns0:cell>-1.29 [-1.68, -0.90]</ns0:cell><ns0:cell>53.66 (13.88)</ns0:cell><ns0:cell>-32.58</ns0:cell><ns0:cell>-1.42 [-1.82, -1.02]</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>High-frequency doublet (T 100 )</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SHAM 30</ns0:cell><ns0:cell>78.56 (20.46)</ns0:cell><ns0:cell>63.90 (15.30)</ns0:cell><ns0:cell>-18.66</ns0:cell><ns0:cell>-0.74 [-1.02, -0.45]</ns0:cell><ns0:cell>63.09 (15.68)</ns0:cell><ns0:cell>-19.69</ns0:cell><ns0:cell>-0.77 [-1.02, -0.52]</ns0:cell><ns0:cell>62.73 (13.77)</ns0:cell><ns0:cell>-20.15</ns0:cell><ns0:cell>-0.82 [-1.10, -0.54]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 30</ns0:cell><ns0:cell>82.52 (21.65)</ns0:cell><ns0:cell>67.91 (15.83)</ns0:cell><ns0:cell>-17.71</ns0:cell><ns0:cell>-0.70 [-0.94, -0.45]</ns0:cell><ns0:cell>65.79 (16.42)</ns0:cell><ns0:cell>-20.28</ns0:cell><ns0:cell>-0.79 [-1.02, -0.56]</ns0:cell><ns0:cell>66.10 (16.90)</ns0:cell><ns0:cell>-19.90</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 60</ns0:cell><ns0:cell>87.98 (22.21)</ns0:cell><ns0:cell>71.22 (18.06)</ns0:cell><ns0:cell>-19.05</ns0:cell><ns0:cell>-0.75 [-1.04, -0.46]</ns0:cell><ns0:cell>71.91 (20.34)</ns0:cell><ns0:cell>-18.27</ns0:cell><ns0:cell>-0.68 [-1.04, -0.33]</ns0:cell><ns0:cell>71.07 (18.78)</ns0:cell><ns0:cell>-19.23</ns0:cell><ns0:cell>-0.75 [-1.08, -0.41]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 60</ns0:cell><ns0:cell>81.00 (19.98)</ns0:cell><ns0:cell>64.55 (20.02)</ns0:cell><ns0:cell>-20.30</ns0:cell><ns0:cell>-0.75 [-1.32, -0.17]</ns0:cell><ns0:cell>63.64 (19.06)</ns0:cell><ns0:cell>-21.43</ns0:cell><ns0:cell>-0.81 [-1.36, -0.25]</ns0:cell><ns0:cell>61.50 (18.58)</ns0:cell><ns0:cell>-24.06</ns0:cell><ns0:cell>-0.92 [-1.57, -0.26]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 180</ns0:cell><ns0:cell>88.05 (21.36)</ns0:cell><ns0:cell>69.93 (16.90)</ns0:cell><ns0:cell>-20.58</ns0:cell><ns0:cell>-0.85 [-1.18, -0.52]</ns0:cell><ns0:cell>68.09 (16.89)</ns0:cell><ns0:cell>-22.68</ns0:cell><ns0:cell>-0.94 [-1.28, -0.60]</ns0:cell><ns0:cell>67.69 (15.60)</ns0:cell><ns0:cell>-23.12</ns0:cell><ns0:cell>-0.99 [-1.39, -0.59]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 180</ns0:cell><ns0:cell>83.10 (21.07)</ns0:cell><ns0:cell>63.93 (20.07)</ns0:cell><ns0:cell>-23.07</ns0:cell><ns0:cell>-0.84 [-1.10, -0.59]</ns0:cell><ns0:cell>62.99 (18.42)</ns0:cell><ns0:cell>-24.20</ns0:cell><ns0:cell>-0.92 [-1.20, -0.64]</ns0:cell><ns0:cell>62.06 (16.54)</ns0:cell><ns0:cell>-25.33</ns0:cell><ns0:cell>-1.01 [-1.30, -0.71]</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>Low-to high-frequency doublet ration (T 20/100 )</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SHAM 30</ns0:cell><ns0:cell>0.96 (0.06)</ns0:cell><ns0:cell>0.90 (0.07)</ns0:cell><ns0:cell>-6.56</ns0:cell><ns0:cell>-0.89 [-1.45, -0.32]</ns0:cell><ns0:cell>0.91 (0.08)</ns0:cell><ns0:cell>-5.15</ns0:cell><ns0:cell>-0.68 [-1.09, -0.27]</ns0:cell><ns0:cell>0.93 (0.08)</ns0:cell><ns0:cell>-3.27</ns0:cell><ns0:cell>-0.43 [-0.78, -0.07]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 30</ns0:cell><ns0:cell>0.97 (0.05)</ns0:cell><ns0:cell>0.88 (0.05)</ns0:cell><ns0:cell>-10.11</ns0:cell><ns0:cell>-0.60 [-0.77, 0.43]</ns0:cell><ns0:cell>0.90 (0.07)</ns0:cell><ns0:cell>-7.50</ns0:cell><ns0:cell>-0.49 [-0.74, 0.48]</ns0:cell><ns0:cell>0.92 (0.07)</ns0:cell><ns0:cell>-5.77</ns0:cell><ns0:cell>-0.44 [-0.73, 0.50]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 60</ns0:cell><ns0:cell>0.97 (0.07)</ns0:cell><ns0:cell>0.88 (0.09)</ns0:cell><ns0:cell>-9.24</ns0:cell><ns0:cell>-0.54 [-0.75, 0.46]</ns0:cell><ns0:cell>0.93 (0.08)</ns0:cell><ns0:cell>-4.27</ns0:cell><ns0:cell>-0.42 [-0.73, 0.51]</ns0:cell><ns0:cell>0.94 (0.09)</ns0:cell><ns0:cell>-3.87</ns0:cell><ns0:cell>-0.35 [-0.71, 0.53]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 60</ns0:cell><ns0:cell>0.98 (0.06)</ns0:cell><ns0:cell>0.89 (0.04)</ns0:cell><ns0:cell>-9.58</ns0:cell><ns0:cell>-1.56 [-2.23, -0.89]</ns0:cell><ns0:cell>0.90 (0.06)</ns0:cell><ns0:cell>-8.42</ns0:cell><ns0:cell>-1.22 [-1.86, -0.57]</ns0:cell><ns0:cell>0.91 (0.08)</ns0:cell><ns0:cell>-7.24</ns0:cell><ns0:cell>-0.90 [-1.78, -0.02]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 180</ns0:cell><ns0:cell>0.96 (0.08)</ns0:cell><ns0:cell>0.88 (0.08)</ns0:cell><ns0:cell>-7.64</ns0:cell><ns0:cell>-0.83 [-1.28, -0.38]</ns0:cell><ns0:cell>0.91 (0.09)</ns0:cell><ns0:cell>-4.98</ns0:cell><ns0:cell>-0.49 [-0.82, -0.15]</ns0:cell><ns0:cell>0.92 (0.08)</ns0:cell><ns0:cell>-4.15</ns0:cell><ns0:cell>-0.44 [-0.80, -0.08]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 180</ns0:cell><ns0:cell>0.98 (0.06)</ns0:cell><ns0:cell>0.87 (0.12)</ns0:cell><ns0:cell>-11.12</ns0:cell><ns0:cell>-1.00 [-1.59, -0.40]</ns0:cell><ns0:cell>0.88 (0.07)</ns0:cell><ns0:cell>-9.53</ns0:cell><ns0:cell>-1.28 [-1.80, -0.75]</ns0:cell><ns0:cell>0.88 (0.09)</ns0:cell><ns0:cell>-9.78</ns0:cell><ns0:cell>-1.10 [-1.54, -0.67]</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>Single twitch peak torque (TW PT )</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>SHAM 30</ns0:cell><ns0:cell>26.81 (7.53)</ns0:cell><ns0:cell>20.62 (6.21)</ns0:cell><ns0:cell>-23.10</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>20.04 (5.12)</ns0:cell><ns0:cell>-25.28</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>20.74 (5.39)</ns0:cell><ns0:cell>-22.65</ns0:cell><ns0:cell>-0.84 [-1.29, -0.40]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 30</ns0:cell><ns0:cell>26.62 (8.07)</ns0:cell><ns0:cell>19.50 (5.65)</ns0:cell><ns0:cell>-26.77</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>19.00 (5.53)</ns0:cell><ns0:cell>-28.64</ns0:cell><ns0:cell>-1.00 [-1.36, -0.64]</ns0:cell><ns0:cell>18.85 (5.34)</ns0:cell><ns0:cell>-29.19</ns0:cell><ns0:cell>-1.03 [-1.42, -0.64]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 60</ns0:cell><ns0:cell>27.37 (8.27)</ns0:cell><ns0:cell>19.83 (4.11)</ns0:cell><ns0:cell>-27.54</ns0:cell><ns0:cell>-1.05 [-1.53, -0.56]</ns0:cell><ns0:cell>19.06 (3.70)</ns0:cell><ns0:cell>-30.35</ns0:cell><ns0:cell>-1.18 [-1.81, -0.54]</ns0:cell><ns0:cell>19.41 (4.35)</ns0:cell><ns0:cell>-29.10</ns0:cell><ns0:cell>-1.09 [-1.58, -0.60]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 60</ns0:cell><ns0:cell>26.77 (8.11)</ns0:cell><ns0:cell>19.76 (5.50)</ns0:cell><ns0:cell>-26.17</ns0:cell><ns0:cell>-0.92 [-1.26, -0.57]</ns0:cell><ns0:cell>18.25 (5.31)</ns0:cell><ns0:cell>-31.84</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell><ns0:cell>19.71 (5.78)</ns0:cell><ns0:cell>-26.37</ns0:cell><ns0:cell>-0.91 [-1.38, -0.44]</ns0:cell></ns0:row><ns0:row><ns0:cell>SHAM 180</ns0:cell><ns0:cell>26.72 (7.68)</ns0:cell><ns0:cell>19.85 (6.47)</ns0:cell><ns0:cell>-25.72</ns0:cell><ns0:cell>-0.88 [-1.09, -0.67]</ns0:cell><ns0:cell>18.90 (5.35)</ns0:cell><ns0:cell>-29.29</ns0:cell><ns0:cell>-1.07 [-1.36, -0.78]</ns0:cell><ns0:cell>18.94 (6.39)</ns0:cell><ns0:cell>-29.15</ns0:cell><ns0:cell>-0.63 [-0.78, 0.42]</ns0:cell></ns0:row><ns0:row><ns0:cell>WBV 180</ns0:cell><ns0:cell>27.20 (8.23)</ns0:cell><ns0:cell>18.06 (7.74)</ns0:cell><ns0:cell>-33.60</ns0:cell><ns0:cell>-1.04 [-1.32, -0.75]</ns0:cell><ns0:cell>16.85 (7.37)</ns0:cell><ns0:cell>-38.06</ns0:cell><ns0:cell>-1.20 [-1.58, -0.82]</ns0:cell><ns0:cell>17.95 (7.35)</ns0:cell><ns0:cell>-34.01</ns0:cell><ns0:cell>-1.07 [-1.43, -0.72]</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:06:50532:1:0:NEW 2 Oct 2020)Manuscript to be reviewed</ns0:note> </ns0:body> "

"Article ID 50532 Response to reviewers Dear Amador García-Ramos, PhD, Academic Editor, PeerJ. We are grateful to the editors and reviewers for their time and constructive comments on our manuscript. We have implemented their comments and suggestions and wish to submit a revised version of the manuscript for further consideration in the journal. Please find attached a revised manuscript with track changes as well as a “clear” version of the manuscript with no track changes. Below, we also provide a point-by-point response explaining how we have addressed each of the editors or reviewers’ comments. We believe that the manuscript is now suitable for publication in PeerJ. Yours sincerely, On behalf of the co-authors Miloš Kalc Point-by-point response to reviewers: Reviewer 1 (Anonymous) Basic reporting The manuscript is well written overall. I only have few comments. Reply: Thank you for your comments. We would like to express a general comment since it applies to most of your comments. We discussed the issue of magnitude-based inference within the authors and decided to follow your suggestion and remove the additional statistics, so we adjusted the results section and discussion accordingly. We interpret the results using only RM ANOVA and post-hoc tests where appropriate. The interpretation of the results and the discussion is now focused on the lack of differences between WBV and SHAM condition. Background in the abstract - increasing evidence ...induce a decline of neuromuscular parameters. I could be improved with an acute impairment in neuromuscular function Reply: Thank you for your comment. Please find an improved version of the abstract background section. Introduction, line 34. Not sure about the use of reflectors here. Check english Reply: Thank you for the comment. We changed »reflectory activation« into »reflexive activation« Lines 48-51. The authors should also mention the differences in protocols and subjects with De Ruiter's findings (in their experiment they had untrained students) but in Maffiuletti's work there was external load added. I suggest the authors clarify this aspect and/or only cite literature with similar protocols and subjects to make their case Reply: Thank you for your comment. We were trying to present a broader view of the topic (level of voluntary activation), since all the cited articles in this paragraph use different protocols, it is difficult to cite research where similar fatiguing protocols or subjects were used. As you noted in Maffiuletti 's work additional load was added, in Zory's work (recreationally trained students) a longer exercise exposure (10 minutes) was used. Taking a closer look into the data presented in De Ruiter’s work (untrained students), the drop of the level of voluntary activation is really similar to those presented in other studies including the present study (approx. 3-5%). Since DeRuiter's work is the only one lacking a control condition/group to bias the results, it is probably the main limiting factor. Please find an adjusted version of the paragraph, where De Ruiter’s work has been removed. Experimental design The cross-over design is appropriate, the experiment is well described and written so it can be replicated. Reply: Thank you very much for the positive evaluation of our scientific work which is very much appreciated by the authors of the study. The statistical analysis as well, however I am not sure why the authors decided to add the magnitude based inferences approach. I think this makes the findings confusing. I suggest the authors to report the results only in the context of the two-way factorial ANOVA reported and discuss the effect sizes (eta squared) to interpret the magnitude of change rather than adding the magnitude of effects. Reply: Thank you for your suggestion. We discussed this issue within the authors and decided to remove the magnitude-based inference statistics. Validity of the findings It is clear that there is no difference between trials and trial x time interaction and therefore the discussion should be focused on that aspect. Reply: Thank you for this suggestion. We discussed this issue within the authors of this study and we do agree with you. Therefore, we revised the discussion section according to your suggestion and put an emphasis on the simple results that there is no effect disregarding which statistical approach has been used. What is the biological meaning of a likely small to possibly small fatiguing effect on single twitch? Reply: Thank you for your comment. We removed the magnitude-based inference statistics, so this sentence has been removed as well. The comparisons with the results on central fatigue with De Ruiter's work should be discussed possibly due to the difference in the subjects used (the subjects in this study were recreationally trained athletes vs students untrained in De Ruiter's work). Reply: Thank you for the comment. Please find a revised version of the paragraph, where we discussed differences between our study and de Ruiter's work. Underlying mechanisms. Improve the first sentence. Probably the authors mean that there seems to be no beneficial effects of using WBV on neuromuscular function due to the lack of difference observed? Reply: Please find an improved version of the first sentence of this paragraph. Lines 336-339 are a bit confusing, this sentence needs improving it is not clear what the authors are trying to say. Reply: Thank you for your comment. Please find an adjusted version of the sentence. Line 340. Poor English. Are the authors attributing the observed acute decrease in force production in both treatments to an impairment in Ca2+ handling? Reply: Thank you for your comment. Please find an improved version of this sentence. line 342- Following the intervention. Reply: We agree with you. We revised the sentences according to your suggestion. The following sentences are a bit too long and have few improvements needed. Also this part should be discussed in the context of the findings. The way it is currently written it is assuming that the intervention (WBV) caused an impairment in neuromuscular function, while the statistics analysis reveals no difference with the sham intervention. So the authors should clarify the last part better and make clear that there was no difference and therefore, the observed reduction over time is possibly caused by the same mechanisms of the sham intervention? Reply: Thank you for your comment. We improved this paragraph making clear that there are no differences between WBV and SHAM intervention and that the observed changes of the assessed parameters is probably caused by the same underlying mechanisms. Limitations While the authors explain well the limitations of the study. I am not sure if the work/rest ratios is an actual limitation. The study was designed for this scope and therefore it was a design choice? I am surprised as well that there was no larger evidence of fatigue with this protocol (180s work - 120s rest). Reply: Thank you for your comment. When designing this study we made the choice to change the exercise set duration maintaining the rest unmutated to 120s for all conditions. In our opinion this could be seen as a limitation since the work/ratio may change. This is an aspect which is worth for us to be addressed, Another approach could be to change the duration of the work and rest, per example use a 1:1 work/rest ratio (in example: 30 s work / 30 s rest; 60 s work / 60s rest …. ). The latter systematics is based on work/rest equity and is rather close to practice also. We found it important that – with the perspective for future evidences – also other protocols which are widely applied and approved in practice should be analyzed. With an emphasis on the training diversity, we mentioned this issue in the discussion section. We were also surprised by the lack of differences between protocols. Conclusions. As indicated above, results should be discussed in light of the statistical findings and not with the MBI. I have the impression that at all costs there is the willingness to identify a positive aspect when the data are clearly showing the opposite. I am confident the authors can address this properly as negative findings are as important as positive ones. Reply: Thank you for your comment. Since we removed the MBI statistic, we improved the conclusion section summarizing our work in the light of statistical findings, concluding there are no differences between WBV and SHAM. Also, I am am a bit puzzled of how the findings can be used by practitioners. Since there is no difference between treatments on the acute measures of neuromuscular function, what is the advantage of using vibration with the protocols and equipment used? This should go in the conclusion Reply: Thank you for your comment. The conclusion has been improved in order to emphasize the lack of differences between WBV and SHAM, My first impression is that probably the sample size was too small despite the a-priori power calculation (Jordan's study used 24 subjects) and they only found a difference in voluntary muscle activation. On what parameter was the power analysis conducted for the Jordan study? Reply: Thank you for your comment. We agree with you, the power analysis presented in the original manuscript was incorrect and misleading. We performed a power analysis aiming to detect large effect sizes and did not use data from other studies. Please find a revisited version at the end of the “Subjects” paragraph. Reviewer 2 (Anonymous) Basic reporting The study is very clear, well written. Experimental design No comments. Validity of the findings Studies show that more than 10 minutes of WBVE can cause fatigue; why do you use 9 minutes ? Thank you for your comment which raised an important aspect. There are numerous different articles which demonstrate an immediate increase of fatigue in response to WBV. To our knowledge, the threshold in terms of cumulative duration is sufficiently high to trigger fatigue throughout vibration exposure is unknown. Additionally, we expect that beside the duration, the vibration amplitude, frequency and length of vibration intervals with reference to intermitted breaks determine the effect of fatigue. To address the issue raised by you, we revised the introduction and clearly stated that fatigue has been manifested after different cumulative durations of vibration exposure ranging between 4 and 10 min. Thereby, the duration we have chosen – 9 min – is included. We also mentioned two articles which manifested fatigue and specified that fatiguing effects are attributed to a neuromuscular origin. We hope these changes bring clarity to the readership of the journal and address your suggestion. Comments for the Author Congratulations. This study is very important to our community. We need to know about WBVE and fatigue. Reply: Thank you very much for the positive evaluation of our scientific work which is very much appreciated by the authors of the study. Reviewer 3 (Grzegorz Sobota) Basic reporting The reviewed paper adheres to all PeerJ policies and within the journal scope. The manuscript is written in the proper English style and conform to professional standards of courtesy and expression. The introduction is sufficient with relevant referenced literature. The structure of the article is typical for an original research article. Reply: Thank you very much for the positive evaluation of our scientific work which is very much appreciated by the authors of the study. Experimental design The authors define the clear research question, connected with the identified gap of knowledge, presented at the broad background in the introduction. Methods and protocols are described very well, also with very detailed statistical analysis. The measurements performed with high technical and ethical standards. Reply: Thank you very much for your generous comment which is very much appreciated. Minor comments: Line: 104: „different visits with at least seven days rest in-between.” – please check it with Figure 1A, there is „separated by 6 days rest in between”. Reply: Thank you for your comment. We corrected Figure 1A to “7 days in between” Validity of the findings The data have been provided, but some of them weren't compatible (I described this at the end in detail) and the figures should be revised. The results are enough to verify hypotheses and make conclusions. The discussion section is prepared nicely. Conclusions are well stated, linked to the research question. The authors have noted some limitations of the study. Reply: Thank you very much for your comment which is very much appreciated. Minor comments: Line 270: „…statistically insignificant for TWPT”. – did you mean „…for TWPT”? {last two letters in subsricpt} Reply: Thank you for your comment. TWTP (last two letters in subscript) is the correct form. We made the correction according to your comment. Line 325 and 326 – there is no explanation of ‘LFF’. Is it low-frequency fatigue? Add abbreviation in line 319 - should be enough. Reply: Thank you for your comment. We inserted the abbreviation LFF as suggested. Description of Figure 4: you put one black triangle with one white circle together as p<0,05. Reply: Thank you for your comment which is very much appreciated. We made the correction according to your comment. I have one question for the research protocol: only right legs were measured, so are they all were right –legged persons? Did you check it? If yes, describe how; if not, please put some opinion about the lower limb lateralization and its influence on your results. Reply: We confirm that only the right leg was used for the neuromuscular assessment; the left leg has not been investigated. We have not considered the leg dominance as fatigue is postulated to have holistic effects. There is no scientific evidence about a significant effect of leg dominance on central and peripheral fatigue in fit and healthy young adults. Therefore, we have not considered the leg dominance as a major covariate in our study design. Nevertheless, we addressed the issue of leg dominance raised by you and added it to the discussion section. The other minor comment on data presentation at figures is about showing native units at graphical parts. It suggests that you present raw data or maybe raw differences for readers who didn’t read carefully the description of the figure. As you noted – the graphics show ES as d Cohen value, so that value hasn’t any unit. Please verify this for Figures 3 and 5. Please consider is it necessary to show the same both in numerical and graphical parts? What is ‘MBI’ abbreviation? Response: Thank you for your comment. We agree with you that the captions (titles of sublots) were presented with their respective units, and as you noted the graphics present Cohen d effect size, which have not any unit. We also agree that the units in the subplot titles could be misleading. However, according to major comments raised by the first reviewer concerning the Magnitude based inference (MBI - additional statistic we used), we decided to remove this statistic, since it made the interpretations of the results less clear and potentially misleading. Consequently, we removed figures 3 and 5 from the manuscript. The major comment is about data compatibility. I compared the relative changes from baseline between Figure 2 and Table 1 for MVC and %VA. The graphical representations didn’t match numerical values. For example, MVC(%) at Figure 1A-30s set and connected black triangles for WBV. I clearly see that the relative change is bigger at tf15 then tf, and the biggest at tf30 but data from Table 1 are -12.59% at tf, -12.03 at tf15, and -12.41 at tf30; that’s why I’m confused. Look also for %VA where for every set (30, 60, and 180s) I found differences. For example (Fig. 1B, 180s set) relative difference at tf for WBV is a positive value (around 2-3% based on the graph) but Table 1 showed -1.20 value. I didn't check all values at Figures 2 and 4 with connection to Tables 1 and 2 – now it’s your job to do it precisely. Replay: Thank you for your comment which is very much appreciated. You are right, there were mistakes in the calculated data presented in the figures; data in the tables and text corpus were correct in the submitted manuscript. We are very grateful for the attention you have drawn to the manuscript and revised the figures according to your suggestion. We submitted an adjusted version of Figures 2 and 3 (please note the new figure numbering, since we removed the figures representing magnitude base inference – figures 3 and 5 in the first submission). Now the data plotted in figures 2 and 3 representing the relative changes from baseline are equivalent to the data presented in the tables and text corpus. Comments for the Author Summarise my review – the paper is very good but necessary is a major revision on data compatibility between graphical and numerical representation. Other comments are minor. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Motile cryptofauna inhabiting coral reefs are complex assemblages that utilize the space available among dead coral stands and the surrounding coral gravel substrate. They comprise a group of organisms largely overlooked in biodiversity estimates because they are hard to collect and identify, and their collection causes disturbance that is unsustainable in light of the widespread reef degradation. Artificial substrate units (ASUs) provide a better sampling alternative and have the potential to enhance biodiversity estimates. The present study examines the effectiveness of ASUs made with defaunated coral gravel to estimate the diversity of motile cryptic crustaceans in the back-reef zone of the Puerto Morelos Reef National Park, Mexico. Species richness, Simpson's diversity index, Shannon-Wiener index and composition of assemblages were compared between ASUs and the surrounding coral gravel substrate. A combined total of 2,740 specimens of 178 different species, belonging to five orders of Crustacea (Amphipoda, Cumacea, Isopoda, Tanaidacea and Decapoda) were collected. Species richness was higher in the surrounding coral gravel and Shannon-Wiener and Simpson indexes were higher in ASUs.</ns0:p><ns0:p>Species composition differed between methods, with only 71 species being shared among sampling methods. Decapoda was more speciose in ASUs and Peracarids in the surrounding coral gravel. Combining the use of ASUs with surrounding gravel provided a better inventory of motile cryptic crustacean biodiversity, as 65% of the species were represented by one or two specimens.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Estimating the biodiversity of coral reefs is challenging as many invertebrate species are rare, small, and inhabit microhabitats that are difficult to access. This is especially true of cryptofauna, which are a major component of the biodiversity of coral reefs <ns0:ref type='bibr' target='#b48'>(Reaka-Kudla, 1997)</ns0:ref> with the subphylum Crustacea being one of the most abundant and speciose groups. Its representatives occupy cracks, crevices and cavities within the reef, ranging from a few millimeters to several centimeters in diameter, including coral framework, bioerosion galleries, and the interstices between large clasts in deposits of skeletal gravel <ns0:ref type='bibr' target='#b30'>(Hutchings &amp; Weate, 1977;</ns0:ref><ns0:ref type='bibr' target='#b45'>Peyrot-Clausade, 1980;</ns0:ref><ns0:ref type='bibr' target='#b48'>Reaka-Kudla, 1997)</ns0:ref>. Skeletal gravel is common on coral reefs that are impacted by tropical cyclones and is generated when storm and hurricane waves destroy live coral stands on the shallow inner shelf, and deposit the fragmented corals as a layer of coarse gravel covering the PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed shallow reef zones <ns0:ref type='bibr' target='#b2'>(Blanchon, Jones &amp; Kalbfleisch, 1997)</ns0:ref>. In Caribbean fringing reefs, coral sand and gravel produced during these events is deposited mainly over the crest and the backreef causing a retrograde accretion through time <ns0:ref type='bibr' target='#b4'>(Blanchon et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Skeletal gravel deposits are reported to be colonized by cryptic crustaceans in a little as 2 to 4 weeks <ns0:ref type='bibr' target='#b51'>(Takada, Abe &amp; Shibuno, 2007)</ns0:ref>, as they provide microhabitats, feeding areas, and protection against predation <ns0:ref type='bibr' target='#b40'>(Moran &amp; Reaka-Kudla, 1991;</ns0:ref><ns0:ref type='bibr' target='#b5'>Buhl-Mortensen et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b29'>Humphries, La Peyre &amp; DeCossas, 2011</ns0:ref>). Yet cryptic crustacean inhabiting coral gravel have been largely overlooked in biodiversity estimates because individuals are hard to collect and identify. Furthermore, their collection is commonly destructive and involves disturbance to the collection site <ns0:ref type='bibr' target='#b19'>(Enochs, 2012;</ns0:ref><ns0:ref type='bibr' target='#b20'>Enochs &amp; Manzello, 2012)</ns0:ref>, which is incompatible with coral reef health and prohibited in marine protected areas.</ns0:p><ns0:p>Artificial substrate units (ASUs) are fabricated structures that mimic the characteristics of natural habitats <ns0:ref type='bibr' target='#b57'>(Walker, Schlacher &amp; Schlacher-Hoenlinger, 2007)</ns0:ref>. Their design can provide high spatial diversity, they are easy to place, recover, and relocate, and can provide a standardized sampling effort, allowing direct comparison between different sites <ns0:ref type='bibr' target='#b8'>(Chapman, 2002;</ns0:ref><ns0:ref type='bibr' target='#b51'>Takada, Abe &amp; Shibuno, 2007;</ns0:ref><ns0:ref type='bibr' target='#b0'>Baronio, 2008;</ns0:ref><ns0:ref type='bibr' target='#b55'>Takada et al., 2016)</ns0:ref>. ASUs can also be tracked over time to study recruitment and succession processes <ns0:ref type='bibr'>(Perkol-Finkel &amp; Beneyahu, 2005)</ns0:ref>, and the response of biota to environmental gradients or short-term disturbances <ns0:ref type='bibr' target='#b57'>(Walker, Schlacher &amp; Schlacher-Hoenlinger, 2007)</ns0:ref>.</ns0:p><ns0:p>Several types of ASUs have been developed to study the biodiversity of hard bottom marine habitats <ns0:ref type='bibr' target='#b46'>(Plaisance et al., 2011;</ns0:ref><ns0:ref type='bibr'>Enoch &amp; Mazello, 2012;</ns0:ref><ns0:ref type='bibr' target='#b55'>Takada et al., 2016)</ns0:ref>. Artificial Reef Matrix Structures, for example, are ASUs made of affordable materials which are designed to mimic large head corals <ns0:ref type='bibr' target='#b61'>(Zimmerman &amp; Martin, 2004)</ns0:ref>. By contrast, ASUs designed to study PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020) motile cryptofauna diversity commonly employ mesh trays filled with defaunated coral gravel, which is reported to have the highest species richness, compared to live or recently dead coral <ns0:ref type='bibr'>(Enoch &amp; Mazello, 2012)</ns0:ref>. This type of ASU has been employed on Pacific reefs <ns0:ref type='bibr'>(Enoch et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b53'>Takada, Abe &amp; Shibuno, 2012;</ns0:ref><ns0:ref type='bibr' target='#b55'>Takada et al., 2016)</ns0:ref>, but has been used to a lesser extent in Caribbean, despite the fact that coral gravel is one of the dominant substrates <ns0:ref type='bibr' target='#b9'>(Choi &amp; Ginsburg, 1983;</ns0:ref><ns0:ref type='bibr' target='#b24'>Gischler &amp; Ginsburg, 1996)</ns0:ref>. In order to determine their efficiency, however, data derived from their employment needs to be compared with data obtained through other sampling methods.</ns0:p><ns0:p>In this study, we evaluate the efficiency of ASUs made with plastic mesh-bags filled with defaunated coral gravel as a means of obtaining the crustacean motile cryptofauna diversity and improve the species inventory in the back-reef zone of a Mexican Caribbean reef, where the diversity of cryptic crustaceans in coral gravel has been reported previously (Monroy-Vel&#225;zquez, <ns0:ref type='bibr' target='#b39'>Rodr&#237;guez-Mart&#237;nez &amp; Alvarez, 2017)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>Study site</ns0:head><ns0:p>The study was carried out in the back-reef zone of the Bonanza reef site (20&#176;57&#697;58&#698; N, 086&#176;48&#697;27&#698; W; Fig. <ns0:ref type='figure' target='#fig_4'>1A</ns0:ref>), within the Puerto Morelos Reef National Park, in the Mexican Caribbean. The site is characterized by well-developed back-reef and crest zones, and a reeffront with limited structural relief and only small (&lt; 50 cm) scattered coral colonies <ns0:ref type='bibr' target='#b33'>(Jord&#225;n-Dahlgren, 1979)</ns0:ref>. Between the reef and the shore, lies a reef lagoon (~ 2.5 km wide) colonized by seagrasses and macroalgae. The back-reef environment at Puerto Morelos is the main zone of Manuscript to be reviewed active coral growth at present and is dominated by Acropora palmata, Orbicella spp., Pseudodiploria spp., Siderastrea siderea, Agaricia agaricites, and Porites astreoides <ns0:ref type='bibr'>(Caballero-Arag&#243;n et al., 2020)</ns0:ref>, whereas the crest zone is dominated by <ns0:ref type='bibr'>A. palmata and Millepora complanata (Jord&#225;n-Dahlgren, 1979)</ns0:ref>. After tropical storms and hurricanes, a large amount of skeletal detritus from these coral species accumulates in the back-reef. Based on historical evidence, 27 hurricanes have passed within 50 km of the town of Puerto Morelos between 1852 to 2019, with Hurricanes Gilbert <ns0:ref type='bibr'>(1988)</ns0:ref> and <ns0:ref type='bibr'>Wilma (2005)</ns0:ref> being the most intense (National Hurricane Center, 2020). The site is also under the influence of trade winds, which are interrupted by mild cold fronts for periods of 3-10 days in the winter <ns0:ref type='bibr'>(Ruiz-Renter&#237;a, van Tussenbroek &amp; Jord&#225;n-Dahlgren, 1998)</ns0:ref>. The Yucatan current flows northward along the narrow shelf and, during the trade wind season, its superficial waters are transported into the reef area.</ns0:p><ns0:p>Monthly average sea surface temperature ranges from 25.1 to 29.9&#176;C <ns0:ref type='bibr' target='#b49'>(Rodr&#237;guez-Mart&#237;nez et al., 2010)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Artificial substrate unit design</ns0:head><ns0:p>The Artificial substrate unit (ASU) was designed using a plastic tray (50 cm high by 40 cm wide) within which was placed a mesh bag (with a 35 mm mesh) filled with 3 kg of coral gravel (collected from the beach behind to the study site and dried for five days to ensure that it was uncolonized; Fig. <ns0:ref type='figure' target='#fig_4'>1</ns0:ref>). The coral gravel selected was naturally porous and ranged in size from 5 to 20 cm (Fig. <ns0:ref type='figure' target='#fig_4'>1B</ns0:ref>). The crate was anchored with a concrete block to prevent its displacement by waves and currents (Fig. <ns0:ref type='figure' target='#fig_4'>1C</ns0:ref>); the block holes were open to the surface, allowing the recruitment of cryptofauna (Fig. <ns0:ref type='figure' target='#fig_4'>1D</ns0:ref>). Using scuba, two ASUs were placed on the seafloor of the back-reef zone at a depth of 3 m, in the area where coral gravel accumulates after storms and hurricanes. <ns0:ref type='table'>PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:ref> Manuscript to be reviewed These ASUs were replaced every two to three months with fresh gravel (May, August, and November of 2013, and January of 2014); this period was selected based on the studies of <ns0:ref type='bibr' target='#b51'>Takada et al. (2007)</ns0:ref> and <ns0:ref type='bibr'>Takada et al. (2008)</ns0:ref> who showed that a period of 2-4 weeks is sufficient for the establishment of cryptofauna on coral reefs. For retrieval, each ASU was placed into a plastic bag to prevent specimen loss. At the same time, 3 kg of the same-sized coral gravel was collected in-situ from the area surrounding the ASU with an area no larger than 9 m 2 (Fig. <ns0:ref type='figure' target='#fig_4'>1E</ns0:ref>). Once in the boat, both bagged samples were placed in buckets containing seawater and immediately transported to the laboratory. In total eight samples were obtained from ASUs and eight from coral gravel collected in situ throughout the study. All surveys were conducted under permit DGOPA.00008.080113.0006 granted by SAGARPA (Agriculture, Natural Resources and Fisheries Secretariat) to F. Alvarez.</ns0:p></ns0:div> <ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div> <ns0:div><ns0:head>Laboratory work</ns0:head><ns0:p>This work was done at the same time as the study on cryptic peracarid crustacean of the Puerto Morelos Reef National Park, previously reported by <ns0:ref type='bibr' target='#b39'>(Monroy-Vel&#225;zquez, Rodr&#237;guez-Mart&#237;nez &amp; Alvarez, 2017)</ns0:ref>. Laboratory methods and identification keys employed for the two studies were the same as previously described by these authors.</ns0:p></ns0:div> <ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>Species diversity obtained using the two sampling methods was assessed using Hill Numbers of the effective number of species <ns0:ref type='bibr' target='#b26'>(Hill, 1973;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chao et al., 2014)</ns0:ref>, namely species richness (q = 0), the exponential of Shannon entropy index, or Shannon diversity (q = 1), and the inverse of the Manuscript to be reviewed Simpson concentration index, or Simpson diversity (q = 2). Hill Numbers and curves, and measures of sample coverage, were obtained by means of the package iNEXT in the R environment <ns0:ref type='bibr' target='#b27'>(Hsieh et al., 2016)</ns0:ref>. Sample coverage is a measure of sample completeness, giving the proportion of the total number of individuals in a community that belong to the species represented in the sample; subtracting the sample coverage from unity gives the probability that the next individual collected belongs to a species not previously collected in the sample <ns0:ref type='bibr' target='#b27'>(Hsieh et al., 2016)</ns0:ref>.</ns0:p><ns0:p>To compare species composition between methods, non-metric multidimensional scaling (NMDS) ordination was employed, using the metaMDS function (Vegan package), with Bray-Curtis dissimilarity measure and 999 permutations. Species data were converted into a proportion of the total number of occurrences by sample, thereby accounting for differences among samples in the total number of individuals. Differences in composition among methods were tested by a permutational multivariate analysis of variance with 999 permutations, using the nonparametric ADONIS function of the Vegan package in the R environment <ns0:ref type='bibr' target='#b42'>(Oksanen et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The Importance Value Index (IVI) <ns0:ref type='bibr' target='#b15'>(Curtis &amp; McIntosh, 1951)</ns0:ref> was used as a proxy to estimate the relative importance of each taxon within each substrate. The IVI of each taxon is calculated as IVI = (RA+RF)/2, where RA is relative abundance, calculated from the number of individuals per taxon with respect to the number of individuals of all species found in the assemblage, and where RF is relative frequency, estimated as the proportion of surveys where a taxon is present, normalized to the frequency of all species in the assemblage. All analyses were done in R 3.6.3.</ns0:p><ns0:p>(R Core Team, 2016).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>A total of 2,740 specimens belonging to at least 178 species, encompassing five orders of Crustacea (Amphipoda, Cumacea, Isopoda, Tanaidacea and Decapoda) and 58 families, were identified and recorded throughout the study. Of these, 129 taxa were identified to species, 39 to genus and ten to higher taxonomic levels. The taxonomic composition of the samples taken using the two methods is summarized in Supplementary Table <ns0:ref type='table'>1</ns0:ref>. Fifty-five species (31%) were represented by a single specimen and 60 (34%) by two specimens each. Forty percent of the species were shared among methods. Decapoda was the most speciose order, with 57 species, followed by Isopoda (N = 48), Amphipoda (N = 39), Tanaidacea (N = 18) and Cumacea (N = 16). Three species of Decapoda represent new records for the Mexican Caribbean (Paguristes hernancortezi, Processa profunda, and Processa riveroi). Other specimens that were rarely observed in samples, and were not included in the data analyses, were crustaceans belonging to the class Ostracoda and to the subclass Copepoda, as well as animals belonging to Mollusca, Polychaeta, and Echinodermata.</ns0:p><ns0:p>In total, 868 specimens of crustaceans, consisting of at least 116 species, were obtained from the ASUs , and 1,872 specimens, consisting of at least 133 species, were obtained from coral gravel collected in situ (Table <ns0:ref type='table'>1</ns0:ref>). Species richness was not significantly different between methods (confidence intervals overlap; Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>) but the identity of the species differed, showing that both contribute to unique taxa; 45 were exclusive to ASUs and 62 were unique to coral gravel collected in situ. In both methods, over half of the species were represented by one or two specimens (ASUs = 57%; Coral gravel = 52%). Overall, Decapoda was more speciose in ASUs, while Isopoda, Amphipoda, Tanaidacea and Cumacea were more speciose in coral gravel collected in situ (Table <ns0:ref type='table'>1</ns0:ref>). Regarding the number of individuals, Isopoda was the most abundant Manuscript to be reviewed order in ASUs and Tanaidacea in coral gravel collected in situ; Cumacea was rare in samples obtained by both methods (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>Shannon and Simpson indexes were significantly higher in ASUs (H' = 39.4; D = 18.2) than in coral gravel collected in situ (H' = 27.3; D = 13.5) (confidence intervals don't overlap; Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>).</ns0:p><ns0:p>Rarefaction curves of species richness constructed to estimate the reliability of diversity estimates for both methods (Fig. <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>) failed to reach a plateau, indicating that sample size was insufficient to reliably estimate the total number of species and thus diversity measurements for each method are conservative (Fig. <ns0:ref type='figure' target='#fig_5'>2</ns0:ref>). Estimates of sample coverage, a measure of sampling completeness, were 0.95 for ASUs, 0.97 for coral gravel collected in situ, and 0.98 when both methods were combined (Fig <ns0:ref type='figure' target='#fig_6'>3</ns0:ref>).</ns0:p><ns0:p>The nMDS conducted to analyze the beta-diversity of cryptic Crustacea for ASUs and for coral gravel collected in situ shows significant difference between methods (ANOSIM test, Stress: 0.1465, R = 0.9790, p = 0.0073). ASU samples were more similar to each other than to samples of coral gravel collected in situ, nevertheless the two types of samples cannot be clearly separated (Fig. <ns0:ref type='figure'>4</ns0:ref>).</ns0:p><ns0:p>According to the Importance Value Index (IVI), the dominant species differed between methods.</ns0:p><ns0:p>In ASUs, the dominant species were the tanaidacean Chondrochelia dubia (IVI = 9.5%) and the isopod Cirolana parva (IVI = 6.2%), with other relatively important species being the amphipod Elasmopus rapax (IVI = 3.5%) and the tanaidacean Apseudes sp. A (IVI = 3.4%) (Fig. <ns0:ref type='figure' target='#fig_7'>5</ns0:ref>). In coral gravel collected in situ, the dominant species were the tanaidaceans Apseudes sp. A (IVI = 8.6%), Paratanais sp. A (IVI = 7.9%), Pseudoleptochelia sp. A (IVI = 6.2%) and Chondrochelia dubia (IVI = 5.2%) (Fig. <ns0:ref type='figure' target='#fig_7'>5</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed DISCUSSION Artificial sampling units (ASUs) made with fresh coral gravel and deployed in different seasons for short periods of time (2-3 months) are an effective method for improving species inventory of motile cryptic crustaceans on Caribbean coral reefs. Using this method we recorded 116 species of this group during the one-year study duration at the Bonanza site, 45 of which were not recorded in the surrounding coral gravel. Nevertheless, the ASUs failed to record 62 species that were unique to the surrounding coral gravel. However, the rarefaction curves of species richness for both methods failed to reach a plateau, indicating that more samples were needed to have a complete inventory. By combining both ASUs and surrounding gravel samples, we recorded a total of 178 species, with 65% being represented by one or two individuals, and reached a sample coverage of 98% in our sampling size. Our results therefore support the finding of other studies which suggest that sampling of coral gravel using different techniques would render a higher taxonomic richness <ns0:ref type='bibr'>(Costello et al., 2017)</ns0:ref> and a greater potential for the discovery of new species <ns0:ref type='bibr'>(Souza, Oliveira &amp; Almeida, 2012;</ns0:ref><ns0:ref type='bibr' target='#b43'>Paz-R&#237;os, Sim&#245;es &amp; Ardisson, 2013)</ns0:ref>.</ns0:p><ns0:p>ASUs were more effective in sampling decapods, with 23 out of the 57 species recorded being exclusive to this method, while the surrounding coral gravel was more effective for recording unique species of Amphipoda, Cumacea, Isopoda and Tanaidacea, even though some families of these orders were only sampled by ASUs, including the Amphilochidae and Bateidae, of the order Amphipoda, and the Munnidae, of the order Isopoda. Decapod families exclusively found in ASUs were: Hippolytidae, Paguridae, Pilumnidae, Porcellanidae, Spongicolidae and Thoridae.</ns0:p><ns0:p>Differences between sampling methods in motile cryptic crustacean species richness, diversity, and assemblage composition could also be explained by the duration of time that each substrate remained underwater, and thus differences in the composition and coverage of algal turfs.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed Peracarids like a layer of turf algae and fine sediment particles on which to feed, while decapods were more likely looking for space in which to live away from predators. Coral gravel within ASUs had low algal turf coverage, as it stayed in the water for only a few months (&#8804; 3).</ns0:p><ns0:p>Although biofilms formed by bacteria and microalgae can be formed within hours <ns0:ref type='bibr'>(Cuba &amp; Blake, 2018)</ns0:ref>, the composition and coverage of the algal assemblage can change significantly within months <ns0:ref type='bibr'>(Frike et al., 2011)</ns0:ref>, as opportunistic filamentous species are replaced by more competitive fleshy ones and Cyanobacteria <ns0:ref type='bibr' target='#b58'>(Wanders, 1977;</ns0:ref><ns0:ref type='bibr'>Frike et al., 2011)</ns0:ref>. Given that the pattern of succession of algae can shape their communities <ns0:ref type='bibr' target='#b10'>(Connell &amp; Slatyer, 1977)</ns0:ref>, the absence, or low coverage, of certain algal species could have inhibited the colonization or permanence of some of the cryptic crustacean species in the ASUs. Biofilms, for example, are known to release peptides that induce the settlement of several species of sessile invertebrate larvae <ns0:ref type='bibr' target='#b32'>(Johnson et al., 1997)</ns0:ref> and sessile assemblages on coral gravel may later affect the colonization of cryptic motile fauna <ns0:ref type='bibr' target='#b34'>(Klumpp, McKinnon &amp; Mundy, 1988;</ns0:ref><ns0:ref type='bibr' target='#b35'>Kramer, Bellwood &amp; Bellwood, 2012)</ns0:ref>.</ns0:p><ns0:p>In the coral gravel collected around the ASUs, the algal-turf cover was higher, increasing habitat heterogeneity and allowing detritus to be trapped <ns0:ref type='bibr' target='#b16'>(Danovaro &amp; Fraschetti, 2002)</ns0:ref>. This possibly favored a higher species richness of peracarids, which are typically found in early successional stages. Despite its apparent permanence in back-reef environments, coral gravel cannot be seen as a static substrate, particularly in shallow reef sites, where it can be periodically reworked by currents and large wave events during storms, hurricanes, and north winds, or disturbed by fish feeding and bioerosion, among other factors, thus becoming periodically available for colonization <ns0:ref type='bibr' target='#b51'>(Takada, Abe &amp; Shibuno, 2007)</ns0:ref>. All these factors may drive the distribution and structure of cryptic assemblages <ns0:ref type='bibr' target='#b9'>(Choi &amp; Ginsburg, 1983;</ns0:ref><ns0:ref type='bibr' target='#b38'>Meesters et al., 1991)</ns0:ref> Manuscript to be reviewed to the maintenance of high species diversity, by avoiding competitive exclusion and facilitating the colonization of less competitive species <ns0:ref type='bibr' target='#b21'>(Enochs et al., 2011)</ns0:ref>. A higher peracarid species richness would likely be obtained by increasing the number of ASUs per survey and allowing the artificial substrate to become covered by an algal matrix before deployment.</ns0:p><ns0:p>The dominant cryptic crustacean species, as obtained by the IVI, differed between sampling methods. Coral gravel substrates are dominated by tanaids (Apseudes sp., Paratanais sp., Pseudoleptochelia sp. and Chondrochelia dubia), while in ASUs, C. dubia is co-dominant with isopod Cirolana parva, and the amphipod Elasmopus rapax: these species are probably opportunistic colonizers of new habitat space. Juveniles and ovigerous females of E. rapax, C. and decapods (families Alpheidae and Mithracidae) were observed in ASUs throughout the study. The dominant species in ASUs were previously reported as abundant in coral gravel substrates on the Puerto Morelos reef <ns0:ref type='bibr' target='#b39'>(Monroy-Vel&#225;zquez, Rodr&#237;guez-Mart&#237;nez &amp; Alvarez, 2017;</ns0:ref><ns0:ref type='bibr' target='#b60'>Winfield et al., 2017)</ns0:ref>, suggesting that, despite their artificial nature, ASUs were not only colonized by some of the most abundant reef species, but also by rare ones too. More studies are needed to determine if ASUs have an effect on the abundance, life stage, and sizes of the individuals recruited.</ns0:p><ns0:p>Our findings show that when assessing the effectiveness of ASUs on coral reefs, or other ecosystems, care should be taken in comparing the experimental results with controls collected simultaneously from the same sample station. Changes in either of these variables can produce significant differences in species composition and abundance that will affect comparisons <ns0:ref type='bibr' target='#b40'>(Moran &amp; Reaka-Kudla, 1991;</ns0:ref><ns0:ref type='bibr' target='#b51'>Takada et al., 2007)</ns0:ref>. Recruitment of the most abundant species in the reef should be expected and, due to the temporary use of ASUs, mostly motile cryptofauna should be evaluated. Sessile encrusting or colony forming species are not expected to be PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:p><ns0:p>Manuscript to be reviewed common in ASUs, unless they remain in the water for several months or more <ns0:ref type='bibr' target='#b37'>(Malella, 2007;</ns0:ref><ns0:ref type='bibr' target='#b18'>Duckworth &amp; Wolff, 2011)</ns0:ref>. Once a broad survey of the species composition of the local coral gravel has been undertaken, it is then possible to evaluate the effectiveness of ASUs. Our results</ns0:p><ns0:p>show that the use of ASUs made with defaunated coral gravel is effective in detecting cryptic and rare motile crustaceans, and can help improve species inventories of this group on Caribbean coral reefs.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>Artificial sampling units (ASUs) made with defaunated coral gravel constitute a valuable tool to study the diversity of motile cryptic Crustacea in Caribbean coral reefs without using destructive methods in an already deteriorating ecosystem. Our results show that combining data from ASUs with that from surrounding coral gravel gives a more complete inventory of species, as both methods contribute unique species. ASUs are better to estimate diversity, whereas the surrounding coral gravel gives better estimates of species richness. By combining both methods we record an assemblage of motile cryptic Crustacea composed of at least 178 species encompassing five orders at a single reef site in one year. Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>and contribute PeerJ reviewing PDF | (2020:06:49584:1:1:NEW 26 Aug 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 5 Ecological</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> </ns0:body> "

"PeerJ Editor August 19 23rd 2020 Dear Editor, Please find below the answers (in blue) to the queries made for the article “How good are artificial collectors to estimate crustacean cryptofauna diversity in coral reefs?” (ID: 49584), which has been renamed as “Use of artificial substrate units to improve species inventory of cryptic Crustacea on Caribbean coral reefs”. Reviewers comments Editor 1. Concerns about whether the authors have the power to test their stated hypotheses. Concerns about replication and statistical analyses. Include both the additional analyses suggested and the appropriate caveats about potential shortcomings of the approach and future directions for improvement. Concerns about sampling effort and design that must be addressed in your revision, and I think most readers would appreciate a power analysis that shows exactly what is your ability to detect differences and account for the variability of the sampled communities. Even if the study lacks statistical power and the differences among collectors are purely observational, the second referee points out that these are extremely understudied communities (especially in the Caribbean) and the survey of species detected is a valuable scientific contribution in itself. Following your recommendation and that of referee 2, we have changed the focus of the manuscript to evaluate the effectiveness of artificial substrate units to improve species inventory. We removed all tests regarding species abundance. For diversity estimates we provide information on sample completeness, which show values of 95% or higher (Fig. 3). 2. Make clear your specific goals for the study – is it to collect the most species, or minimize variance among collectors, or create a natural habitat with real-world assemblages of cryptic organisms? If your initial goal was different – but you lack the replication and statistical power to convince referees that you have achieved your goal in this study – the study could be written up in a different manner that is far more likely to be acceptable. For example, if you lack the power to test your stated hypothesis, you could revise the focus of the paper as a preliminary evaluation of a method to sample this understudied community and future directions for refinement of the approach In the revised version the goal is to evaluate the effectiveness of artificial substrate units made with defaunated coral rubble to estimate motile cryptofauna diversity and improve species inventory. 3. English language must be improved. The new version was edited by an English-speaking researcher. Reviewer 1 1. Problems with the design of the study. We remove all statistical tests in the revised version and concentrate only on the contribution to the species inventory. l. 95. Nice that all organisms were removed by sun drying, but rubble in the sea is not sun-dried, usually, so the collectors won’t start to replicate the rubble zone until there is colonization by microbes and micro-algae, etc. Perhaps having left them in place for two months would have allowed that to happen, but I think that could explain some of the results. We modified the text to clarify that the rubble was collected from the beach that runs parallel to the reef site in order to ensure it was defaunated. In the discussion we now comment that letting rubble to become covered by algae before deployment could help increase species inventory. l. 96. I have no problem with the collectors per se, although finding a way to standardize the collector might have been better. But I do wonder about placing the cinder block on the collector in the plastic crate. That will influence access to the rubble in some way, I would expect. Cinder blocks were placed above each ASU to prevent their displacement by waves and currents. In the new version we mention this and also that the holes of the block were facing the sea surface to allow the recruitment of cryptofauna. The animals also had access through the sides of the ASU. The collectors were placed adjacent to the back reef zone, but not in the back reef zone, or for that matter, on the rubble, a part of which was collected for comparative purposes Collectors were placed in the area where coral gravel accumulates after storms and hurricanes. This has been clarified in the text. The biggest problem with this study is that there were too few collectors to accommodate the variability of colonizers. Only two replicates each of two times. Would have been much better to deploy minimum of three collectors, but probably more, up to six, each of two times, maybe three times. I recognize that the person power to handle all those crustaceans might not be present or affordable, but I think one could also narrow the scope of the groups dealt with, or look for the 10-20 most commonly occurring species. It is nice, but not necessary to identify everything that is in the collector as many species are rare, as the authors have seen. We agree with the reviewer that more replicates would be ideal for some studies, depending on their objective. As mentioned above the focus of the manuscript has now been changed to study only the effectiveness of the ASUs to improve species inventory, thus we could not only focus on the most common species, as rare species may play important ecological functions too (see: Mouillot et al. (2013) Rare Species Support Vulnerable Functions in High-Diversity Ecosystems. PLoS Biology.DOI: 10.1371/journal.pbio.1001569). We removed from the manuscript all analyses that would have required larger sample size. The second problem with the collectors is that they were placed near the back reef zone and not on or near the rubble they were intended to replicate. That most likely cut off a lot of potential colonizers. None of the peracarids have water borne larvae, and some, like the cumaceans, tanaids, etc., tend to be sedentary unless conditions get bad enough to make them move. We modified the text to make clear that ASUs were placed in the area where coral rubble accumulates after storms and hurricanes. This has been clarified in the text and we have included a photo in figure 1 that shows the area of accumulation. The data in Table 1 also suggests that the different crustacean groups respond to the relative newness of the rubble in the collectors in different ways. Cumaceans and taniads like a nice layer of microalgae and fine sediment particles on which to feed. On the other hand, the decapods are more likely looking for space in which to live away from predators. We agree with the reviewer and have included this information in the discussion. l. 155-158. There should be some discussion of the taxa that were more abundant in the collectors vs not. Why are they there? I note that tanaids and gnathiids and a few other crustaceans were more abundant in the collectors. We agree with the reviewer and have included this information in the discussion. l. 156. No confidence intervals visible in the Fig. 2 The figure now includes confidence intervals. l. 193 ff. That is probably due to the small amount of rubble material, thus limiting the habitat space. More rubble collectors would undoubtedly collected more of the species known from the area. This is now mentioned in the discussion l. 198-199. I don’t agree with this sentence at all. Peracarids are not typically found in early successional stages of habitats. Some species are, but the vast majority have very specific habitat requirements. In studies of recovery of bottom habitats after trawling, for example, it has been shown that cumaceans are the last group to arrive, long after the worms. Same with a number of the amphipod taxa. This paragraph was removed from the discussion. l. 202-205. Yes, but it is also likely that rubble gets well “glued” to the bottom as it were during periods of relative calm due to microbial and microalgal growth. Those periods can last for months or, in the absence of strong winter storms or hurricanes, maybe even years. During those periods one would expect the rubble community to reach some state of stability with a different cryptofauna than what might be expected show up in rubble baked in the sun and placed in bags. We have included in the manuscript that “Coral gravel cannot be seen as a static substrate in shallow reef sites, where it can be periodically reworked by currents, large wave energy events (e.g. storms, hurricanes, and north winds), or disturbed by fish feeding and bioerosion, among other factors, thus becoming periodically available for colonization and contribute to the maintenance of high species diversity, by avoiding competitive exclusion and facilitating the colonization of less competitive species. L 212-226. Exactly the point. We based our sampling design in the results of Takada et al. (2007, 2008) that show that a period of 2–4 weeks was appropriate for the establishment of cryptofauna in artificial sampling units. However, we now mention in the discussion that species richness could be increased by letting the artificial substrate become covered by turf algae before deployment and by increasing sample size. l. 230. Chondrochelia dubia is probably a weedy species. Early and fast colonizer of new habitat space. This is now included in the discussion. In the end, I am not sure that these collectors do what I think the authors want them to do. That is, provide an inventory of species from the rubble without destructive sampling. Perhaps a better designed, and aged collector would do a better job. We disagree. ASU proved effective in recording 45 species that were not recorded in the surrounding coral gravel, three of them being new records for the Mexican Caribbean. We agree, however, that an aged collector could help to obtain a more complete inventory of cryptic crustaceans, as many species are rare. We also recommend employing different techniques to improve this. For example, the collector could be dipped in fresh water to remove all the macrofauna, then placed in bags with 0.5 mm mesh, or smaller, and then allowed to incubate in the water for a month or two to allow the turf to develop. Then they could be placed on the bottom to see what they collect. If the objective is to get a good inventory of the cryptofauna without destructive sampling, then the cryptic habitat provided needs to be a good approximation of the real thing. We now make clear that the coral gravel employed was obtained from the beach and was defaunated. We have included the recommendation of the referee about allowing the rubble to become covered by an algal matrix before deployment (aged) Reviewer 2 (Ian Enochs) There are some gaps that should be corrected, particularly with reference to prior studies that use artificial reef substrates to assess decapod diversity. We have included information about prior studies that use artificial substrates. I have some concerns with replication and the clarity of the goals, specifically the definition of what a 'good' collector is. It's not clear to me if the aim is to collect the most species or minimize variance, or create a natural habitat with real-world assemblages of cryptic organisms. We have changed the focus of the manuscript and removed all tests regarding species abundance. In the revised version the focus is to evaluate the effectiveness of ASUs to improve the inventory of motile cryptic Crustacea. I have several concerns pertaining to sampling effort and design, as well as the results and corresponding conclusions. I hope and trust that they can be sufficiently addressed before publication. Since the focus of the manuscript was changed we removed all tests regarding species abundance. For diversity estimates we evaluated sample completeness and presented the results (Fig. 3), which show values of 95% or higher. Line 1. Suggest: “Are artificial substrates effective tools for estimating crustacean cryptofauna diversity on coral reefs?” The title was modified to “Use of artificial substrate units to improve species inventory of cryptic Crustacea on Caribbean coral reefs” Line 16. They comprise a group of organisms largely overlooked in biodiversity estimates because they are hard to collect and identify. Done Line 18. That may not be sustainable in light of the widespread degradation of coral reef habitat. Done Line 20: examines Done Line 23: the Simpson’s … and the Shannon-Wiener Done Line 24: Separate species or taxa? If not all identified to species, could say at least 178 different species, belonging to five orders of Crustacea were collected. Done Line 26. Again, “taxa richness” should be substituted with “richness” or “species richness” Done Line 28: reword Done Line 31: “and sampling of animals associated with coral rubble” Done Line 38 “Cryptofauna are” Done Line 38: The part of this sentence on fishes and reef-building organisms is not needed an arbitrary. There are numerous other taxa on reefs and therefore highlighting just these two groups is ineffective. The part of the sentence regarding fish and reef-building organisms was removed. Line 40: largest is not specific, you mean most biodiverse? Most speciose? Most abundant? Modified to “most speciose and abundant”. Line 41: replace “either in” with “including within gaps” Done Line 47-50: This sentence needs to be cleaned up, perhaps broken into two Done Line 53: Technically, I would say that Enochs et al. 2011 was more sustainable than Enochs and Manzello 2012 or Enochs 2012 (Marine Biology, not cited) in that it just used rubble in bags, that were then returned. The other studies involved sampling of framework and are therefore more destructive, though similar to this study, rubble was a large component of that... Enochs et al. 2011 was removed and Enochs 2012, and Enochs and Manzello 2012 were included. Line 58: I think it is critical that you incorporate the work of Plaisance and others that have used NOAA and Smithsonian artificial reef monitoring structures (ARMS) to monitor cryptofauna biodiversity (primarily decapods like this study). These structures are probably the most widely-used, ubiquitous structures around the world, and are definitely worth discussing in this paper. We have included the information about ARMS Line 62: Inappropriate use of the word factual. Changed to “effective” Line 63-64: This makes it seem like the references are about artificial collectors that use rubble, rather than just the rubble itself. The sentence was modified Line 62-68: It strikes me that Enochs et al. 2011 employs artificial bags of rubble, directly as discussed here, rather than references to the rubble. There are also other studies that have used rubble bags as well, both in the Pacific and to a lesser extent in the Caribbean. I would suggest highlighting them as you are introducing the topic here. Done Line 95: Are there permitting issues with this type of collection, drying, and redeployment? Numerous organisms naturally grow on this substrate and aerial exposure can be somewhat destructive (and smell bad). In the methodology section it has now been clarified that the rubble used for the artificial substrates was collected from the beach to ensure it was defaunated. Line 96: Are there shading concerns from the block? It is now mentioned in the text that the holes of the block were facing the sea surface. Line 98: From this I am interpreting that the effective sample size is two for each of these timepoints and eight in total? Please clarify here and in the text. This is a little lower than I am comfortable with given the high variability observed within this community (especially over time) and may have ramifications for statistical analysis. We removed all statistical analyses regarding species abundance. Line 101: Can you report the average depth of the rubble collected? Not water depth, but the actual depth of the layer of rubble on the sea floor. 50cm high rubble bag collectors may provide more shelter per unit area, but they also may have more restricted water exchange, and less epibenthic flora/fauna for organisms to feed on. In essence, if you sample three kilograms of rubble, how spread out of an area are you talking about vs. the 40 cm diameter of the collector? Done Line 101: Can you speak to the structural similarity, perhaps size frequency distribution or volume of the rubble used in bags vs. collected? Were they the same general sizes? I think this is important for exploring the similarity of the approaches, as well as differences in the habitat/sheltering potential. Also, a detailed description of the rubble is necessary for general understanding of the habitat sampled. What species of coral and sizes? Was it highly eroded or recently dead and intact? Was it covered with algae or sponges, or bare and uncolonized? It is now mentioned in the text that coral rubble was porous and had diameters of 5-20 cm. We also mention the dominant coral species in the reef. We did not identify the coral species that composed the rubble, as it was old, dead and eroded, and the one collected in situ was covered by turf. In the text it is mentioned that the bags were filled with 3 kg and approximately the same amount was collected in situ. Line 108: Was the rubble reused in the next deployment of the bags? If so, was there mortality of the sessile organisms or did that community mature over the subsequent sampling events throughout the year. Coral rubble was not reused. It is now explained in the text that every time collectors were placed they were filled with new rubble (collected from the beach). Line 119: Is there a citation for Hill numbers? I am not familiar with them. Two references were included Hill 1973; Chao et al., 2014 Line 123: Why extreme transformation? This analysis was removed. Line 126: This statement isn’t clear to me? All time points were pooled or they were pooled only within time points? This analysis was removed. Line 127: P values equal to or less than This text was removed. Line 141: wording Done Line 144: The taxonomic composition of the samples taken using the two methodologies Done Line 152: , as well as animals belonging to Mollusca, Polychaeta, and Echinodermata Done Line 155 and Line 163: Are you confident in the statistical power and sample sizes used to assess this or do you think the lack of significance is reflective of replication This analysis was removed. Line 157. nMDS or other multivariate ordination would be an effective method to assess variation in community composition, showing overlap and separation between the two methodologies. nMDS is now included. Lines 160: This underscores my question of habitat depth and the similarities in substrate porosity/ shelter potential The rubble employed on ASUs was from the same coral species as that collected scattered around ASUs and thus most likely had similar porosity, although this last was covered by turf. The depth was the same for both methods. This has been clarified in the text. Line 185: reword Done Line 189: I don’t agree with this. Just because they are not common in the samples does not mean that they were transient or opportunistic. Maybe they are rare, or their preferred food source wasn’t on the rubble. Maybe sampling effort wasn’t high enough to collect patchily distributed communities. There are lots of potential reasons this could be the case Agreed. This paragraph was removed Line 208: et al. Done Line 218, not just algae. Other sessile organisms may be important food sources as well. This is mentioned in the discussion: “Biofilms, for example, are known to release peptides that induce the settlement of several species of sessile invertebrate larvae (Johnson et al., 1997) and sessile assemblages on coral gravel may later affect the colonization of cryptic motile fauna (Klumpp, McKinnon & Mundy, 1988; Kramer, Bellwood & Bellwood, 2012). Line 224: Do you have data on this? We don’t have data on this but we now mention that coral gravel collected in situ was covered by turf coverage. This was clear from observation. Line 228: I think just tanaids is preferable as their classification as shrimps is colloquial and factually incorrect. Done Line 233: Yes abundant species are recruiting to them, but also rare. I also think the last part of the sentence is not necessarily correct. There could still be an effect in their colonization as their abundances, sizes, etc. could be different. We agree, and have included a sentence indicating that “More studies are needed to determine if ASUs have an effect in the abundance, life stage, and sizes of the individuals recruited.” Line 237: reword Done Line 241: How do you know that great replication or longer deployment time would have yielded the same results? Do you have enough data to state that the substrate/collector is the difference, not the amount of material or time of community development/succession? As suggested by the editor, the focus of the manuscript has been changed to determine the effectiveness of ASUs to improve species inventory. We have commented in the discussion that increasing replication and allowing the gravel to become covered by turf algae could help to increase the inventory. Line 241: I think you need to be more clear on what the actual goal is. Are you trying to collect the maximum number of decapod species possible or are you trying to determine if the 2-month collector deployment reflects the same species composition, diversity, and abundances? These are different goals and the conclusions as to the effectiveness of each are different. It also strikes me that part of the “effectiveness” of the collectors is their community variance and sensitivity to environmental differences/change. We have now made clear the goal is to evaluate ASUs effectiveness to improve species inventory. We conclude that their use combined with collecting coral rubble in situ improves biodiversity estimates. Are successive collections similar enough that they can be used to detect the influence of environmental gradients? That would be very valuable or “effective” and wouldn’t necessarily result in the highest number of species collected. I think community similarity should be investigated further, and I think ordination would be helpful. We have included an nMDS analyses Line 254: What about flow, porosity, and benthic cover as investigated in Enochs et al. 2011? We didn’t record this information as it was not part of the objectives. Line 266: substitute access for sample Done "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Motile cryptofauna inhabiting coral reefs are complex assemblages that utilize the space available among dead coral stands and the surrounding coral rubble substrate. They comprise a group of organisms largely overlooked in biodiversity estimates because they are hard to collect and identify, and their collection causes disturbance that is unsustainable in light of widespread reef degradation. Artificial substrate units (ASUs) provide a better sampling alternative and have the potential to enhance biodiversity estimates. The present study examines the effectiveness of ASUs made with defaunated coral rubble to estimate the diversity of motile cryptic crustaceans in the back-reef zone of the Puerto Morelos Reef National Park, Mexico. Species richness, Simpson's diversity index, Shannon-Wiener index and the composition of assemblages were compared between ASUs and samples from the surrounding coral rubble substrate. A combined total of 2,740 specimens of 178 different species, belonging to five orders of Crustacea (Amphipoda, Cumacea, Isopoda, Tanaidacea and Decapoda) were collected. Species richness was higher in the surrounding coral rubble and Shannon-Wiener and Simpson indexes were higher in ASUs. Species composition differed between methods, with only 71 species being shared among sampling methods. Decapoda was more speciose in ASUs and Peracarids in the surrounding coral rubble. Combining the use of ASUs with surrounding rubble provided a better inventory of motile cryptic crustacean biodiversity, as 65% of the species were represented by one or two specimens.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>INTRODUCTION</ns0:head><ns0:p>Estimating the biodiversity of coral reefs is challenging as many invertebrate species are rare, small, and inhabit microhabitats that are difficult to access. This is especially true of cryptofauna, which are a major component of the biodiversity of coral reefs that are hard to estimate <ns0:ref type='bibr' target='#b55'>(Reaka-Kudla, 1997;</ns0:ref><ns0:ref type='bibr' target='#b58'>Small et al., 1998)</ns0:ref>, with the subphylum Crustacea being one of the most abundant and speciose groups. Its representatives occupy cracks, crevices and cavities within the reef, ranging from a few millimeters to several centimeters in diameter, including coral framework, bioerosion galleries, and the interstices between large clasts in deposits of skeletal rubble <ns0:ref type='bibr' target='#b31'>(Hutchings &amp; Weate, 1977;</ns0:ref><ns0:ref type='bibr' target='#b53'>Peyrot-Clausade, 1980;</ns0:ref><ns0:ref type='bibr' target='#b55'>Reaka-Kudla, 1997)</ns0:ref>. Skeletal rubble is common on coral reefs that are impacted by tropical cyclones and is generated when storm and hurricane waves destroy live coral stands on the shallow inner shelf, and deposit the fragmented Manuscript to be reviewed corals as a layer of coarse rubble covering the shallow reef zones <ns0:ref type='bibr' target='#b2'>(Blanchon, Jones &amp; Kalbfleisch, 1997)</ns0:ref>. In Caribbean fringing reefs, coral sand and rubble produced during these events is deposited mainly over the crest and the back-reef causing a retrograde accretion through time <ns0:ref type='bibr' target='#b4'>(Blanchon et al., 2017)</ns0:ref>.</ns0:p><ns0:p>Skeletal rubble deposits are reported to be colonized by cryptic crustaceans in as little as two to four weeks <ns0:ref type='bibr' target='#b60'>(Takada, Abe &amp; Shibuno, 2007)</ns0:ref>, as they provide microhabitats, feeding areas, and protection against predation <ns0:ref type='bibr' target='#b47'>(Moran &amp; Reaka-Kudla, 1991;</ns0:ref><ns0:ref type='bibr' target='#b5'>Buhl-Mortensen et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b30'>Humphries, La Peyre &amp; DeCossas, 2011</ns0:ref>). Yet cryptic crustaceans inhabiting coral rubble have been largely overlooked in biodiversity estimates because individuals are hard to collect and identify. Furthermore, their collection is commonly destructive and involves disturbance to the collection site, which is incompatible with coral reef health and prohibited in marine protected areas.</ns0:p><ns0:p>Artificial substrate units (ASUs) are fabricated structures that mimic the characteristics of natural habitats <ns0:ref type='bibr' target='#b68'>(Walker, Schlacher &amp; Schlacher-Hoenlinger, 2007)</ns0:ref>. Their design can provide high spatial diversity, they are easy to place, recover, and relocate, and can provide a standardized sampling effort, allowing direct comparison between different sites <ns0:ref type='bibr' target='#b8'>(Chapman, 2002;</ns0:ref><ns0:ref type='bibr' target='#b60'>Takada, Abe &amp; Shibuno, 2007;</ns0:ref><ns0:ref type='bibr' target='#b0'>Baronio, 2008;</ns0:ref><ns0:ref type='bibr' target='#b64'>Takada et al., 2016)</ns0:ref>. ASUs can also be tracked over time to study recruitment and succession processes <ns0:ref type='bibr'>(Perkol-Finkel &amp; Beneyahu, 2005)</ns0:ref>, and the response of biota to environmental gradients or short-term disturbances <ns0:ref type='bibr' target='#b68'>(Walker, Schlacher &amp; Schlacher-Hoenlinger, 2007)</ns0:ref>.</ns0:p><ns0:p>Several types of ASUs have been developed to study the biodiversity of hard bottom marine habitats <ns0:ref type='bibr' target='#b54'>(Plaisance et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b22'>Enochs et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b64'>Takada et al., 2016)</ns0:ref> Manuscript to be reviewed large head corals <ns0:ref type='bibr' target='#b73'>(Zimmerman &amp; Martin, 2004)</ns0:ref>. By contrast, ASUs designed to study motile cryptofauna diversity commonly employ mesh trays filled with defaunated coral rubble, which is reported to have the highest species richness, compared to live or recently dead coral <ns0:ref type='bibr'>(Enoch &amp; Mazello, 2012)</ns0:ref>. This type of ASU has been employed on Pacific reefs <ns0:ref type='bibr'>(Enoch et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b62'>Takada, Abe &amp; Shibuno, 2012;</ns0:ref><ns0:ref type='bibr' target='#b64'>Takada et al., 2016)</ns0:ref>, but has been used to a lesser extent in the Caribbean, despite the fact that coral rubble is an abundant substrate and plays an important role in harboring diverse cryptofaunal communities, including fish <ns0:ref type='bibr' target='#b9'>(Choi &amp; Ginsburg, 1983;</ns0:ref><ns0:ref type='bibr' target='#b25'>Gischler &amp; Ginsburg, 1996;</ns0:ref><ns0:ref type='bibr' target='#b67'>Valles et al., 2006)</ns0:ref>. In order to determine their efficiency, however, data derived from their employment needs to be compared with data obtained through other sampling methods.</ns0:p><ns0:p>In this study, we evaluate the efficiency of ASUs made with plastic mesh-bags filled with defaunated coral rubble as a means of obtaining the crustacean motile cryptofauna diversity and improve the species inventory in the back-reef zone of a Mexican Caribbean reef, where the diversity of cryptic crustaceans in coral rubble has been reported previously <ns0:ref type='bibr' target='#b46'>(Monroy-Vel&#225;zquez, Rodr&#237;guez-Mart&#237;nez &amp; Alvarez, 2017)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>MATERIALS AND METHODS</ns0:head></ns0:div> <ns0:div><ns0:head>Study site</ns0:head><ns0:p>The study was carried out in the back-reef zone of the Bonanza reef site (20&#176;57&#697;58&#698; N, 086&#176;48&#697;27&#698; W; Fig. <ns0:ref type='figure' target='#fig_8'>1A</ns0:ref>), within the Puerto Morelos Reef National Park, in the Mexican Caribbean. The site is characterized by well-developed back-reef and crest zones, and a reeffront with limited structural relief and only small (&lt; 50 cm) scattered coral colonies (Jord&#225;n- Manuscript to be reviewed <ns0:ref type='bibr' target='#b33'>Dahlgren, 1979)</ns0:ref>. Between the reef and the shore, lies a reef lagoon (~ 2.5 km wide) colonized by seagrasses and macroalgae. The back-reef environment at Puerto Morelos is the main zone of active coral growth at present and is dominated by Acropora palmata, Orbicella spp., Pseudodiploria spp., Siderastrea siderea, Agaricia agaricites, and Porites astreoides <ns0:ref type='bibr'>(Caballero-Arag&#243;n et al., 2020)</ns0:ref>, whereas the crest zone is dominated by <ns0:ref type='bibr'>A. palmata and Millepora complanata (Jord&#225;n-Dahlgren, 1979)</ns0:ref>. After tropical storms and hurricanes, a large amount of skeletal detritus from these coral species accumulates in the back-reef. Based on historical evidence, 27 hurricanes have passed within 50 km of the town of Puerto Morelos between 1852 to <ns0:ref type='bibr'>2019</ns0:ref><ns0:ref type='bibr'>, with Hurricanes Gilbert (1988)</ns0:ref> and <ns0:ref type='bibr'>Wilma (2005)</ns0:ref> being the most intense (National Hurricane Center, 2020). The site is also under the influence of trade winds, which are interrupted by mild cold fronts for periods of 3-10 days in the winter <ns0:ref type='bibr'>(Ruiz-Renter&#237;a, van Tussenbroek &amp; Jord&#225;n-Dahlgren, 1998)</ns0:ref>. The Yucatan current flows northward along the narrow shelf and, during the trade wind season, its superficial waters are transported into the reef area.</ns0:p><ns0:p>Monthly average sea surface temperature ranges from 25.1 to 29.9&#176;C <ns0:ref type='bibr' target='#b56'>(Rodr&#237;guez-Mart&#237;nez et al., 2010)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Artificial substrate unit design</ns0:head><ns0:p>The artificial substrate unit (ASU) was designed using a plastic tray (50 cm high by 40 cm wide) within which was placed a mesh bag (with a 35 mm mesh) filled with 3 kg of coral rubble (collected from the beach behind to the study site and dried for five days to ensure that it was uncolonized; Fig. <ns0:ref type='figure' target='#fig_8'>1</ns0:ref>). The coral rubble selected was naturally porous and ranged in diameter from recruitment of cryptofauna (Fig. <ns0:ref type='figure' target='#fig_8'>1D</ns0:ref>). Using scuba, two ASUs were placed on the seafloor of the back-reef zone at a depth of 3 m, in the area where coral rubble accumulates after storms and hurricanes. These ASUs were replaced every two to three months with fresh rubble (May, August, and November of 2013, and January of 2014); this period was selected based on the studies of <ns0:ref type='bibr' target='#b60'>Takada et al. (2007)</ns0:ref> and <ns0:ref type='bibr'>Takada et al. (2008)</ns0:ref> who showed that a period of 2-4 weeks is sufficient for the establishment of cryptofauna on coral reefs. For retrieval, each ASU was placed into a plastic bag to prevent specimen loss. At the same time, 3 kg of the same-sized coral rubble was collected in-situ from the area surrounding the ASU with an area no larger than 9 m 2 (Fig. <ns0:ref type='figure' target='#fig_8'>1E</ns0:ref>). Once in the boat, both bagged samples were placed in buckets containing seawater and immediately transported to the laboratory. In total eight samples were obtained from ASUs and eight from coral rubble collected in situ throughout the study. All surveys were conducted under permit DGOPA.00008.080113.0006 granted by SAGARPA (Agriculture, Natural Resources and Fisheries Secretariat) to F. Alvarez.</ns0:p></ns0:div> <ns0:div><ns0:head>Laboratory work</ns0:head><ns0:p>In the laboratory, the coral rubble obtained from the ASUs and in situ was placed in separate buckets filled with fresh water to provoke osmotic shock and force organisms out of their cavities. The residue material was sieved through a 0.5 mm mesh. Organisms were preserved in 70% ethanol and later identified to the lowest possible taxonomic level and counted.</ns0:p><ns0:p>Identifications followed <ns0:ref type='bibr' target='#b59'>Su&#225;rez-Morales et al. (2004)</ns0:ref> for Tanaidacea, <ns0:ref type='bibr' target='#b26'>Heard, Roccatagliata &amp; Petrescu (2007)</ns0:ref> for Cumacea, <ns0:ref type='bibr' target='#b34'>Kensley &amp; Schotte (1989)</ns0:ref> for Isopoda, <ns0:ref type='bibr' target='#b66'>Thomas (1993)</ns0:ref> and <ns0:ref type='bibr' target='#b40'>LeCroy (2000</ns0:ref><ns0:ref type='bibr' target='#b41'>LeCroy ( , 2002</ns0:ref><ns0:ref type='bibr' target='#b42'>LeCroy ( , 2004</ns0:ref><ns0:ref type='bibr' target='#b43'>LeCroy ( , and 2007) )</ns0:ref> for Amphipoda, and <ns0:ref type='bibr' target='#b71'>Williams (1984)</ns0:ref> for Decapoda. <ns0:ref type='table'>PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:ref> Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>PeerJ reviewing</ns0:head></ns0:div> <ns0:div><ns0:head>Data analysis</ns0:head><ns0:p>Species diversity obtained using the two sampling methods was assessed using Hill Numbers of the effective number of species <ns0:ref type='bibr' target='#b27'>(Hill, 1973;</ns0:ref><ns0:ref type='bibr' target='#b7'>Chao et al., 2014)</ns0:ref>, namely species richness (q = 0), the exponential of Shannon entropy index, or Shannon diversity (q = 1), and the inverse of the Simpson concentration index, or Simpson diversity (q = 2). Hill Numbers and curves, and measures of sample coverage, were obtained by means of the package iNEXT in the R environment <ns0:ref type='bibr' target='#b28'>(Hsieh et al., 2016)</ns0:ref>. Sample coverage is a measure of sample completeness that gives the proportion of the total number of individuals in a community that belong to the species represented in the sample <ns0:ref type='bibr' target='#b28'>(Hsieh et al., 2016)</ns0:ref>. Subtracting the sample coverage from unity gives the probability that the next individual collected belongs to a species not previously collected in the sample <ns0:ref type='bibr' target='#b28'>(Hsieh et al., 2016)</ns0:ref>.</ns0:p><ns0:p>To compare species composition between methods, non-metric multidimensional scaling (NMDS) ordination was employed, using the metaMDS function (Vegan package), with Bray-Curtis dissimilarity measure and 999 permutations. Assemblage compositions were computed based on presence/absence of species. Differences in composition among methods were tested by a permutational multivariate analysis of variance with 9999 permutations, using the nonparametric ADONIS function of the Vegan package in the R environment <ns0:ref type='bibr' target='#b49'>(Oksanen et al., 2013)</ns0:ref>.</ns0:p><ns0:p>The Importance Value Index (IVI) <ns0:ref type='bibr' target='#b15'>(Curtis &amp; McIntosh, 1951)</ns0:ref> was used as a proxy to estimate the relative importance of each taxon within each substrate. The IVI of each taxon is calculated as IVI = (RA+RF)/2, where RA is relative abundance, calculated from the number of individuals per taxon with respect to the number of individuals of all species found in the assemblage, and where RF is relative frequency, estimated as the proportion of surveys where a taxon is present, Manuscript to be reviewed normalized to the frequency of all species in the assemblage. All analyses were done in R 3.6.3.</ns0:p><ns0:p>(R Core Team, 2016).</ns0:p></ns0:div> <ns0:div><ns0:head>RESULTS</ns0:head><ns0:p>A total of 2,740 specimens belonging to at least 178 species, encompassing five orders of Crustacea (Amphipoda, Cumacea, Isopoda, Tanaidacea and Decapoda) and 58 families were identified and recorded throughout the study. Of these, 129 taxa were identified to species, 39 to genus and ten to higher taxonomic levels. The taxonomic composition of the samples taken using the two methods is summarized in Supplementary Table <ns0:ref type='table'>1</ns0:ref>. Fifty-five species (31%) were represented by a single specimen and 60 (34%) by two specimens each. Forty percent of the species were shared among methods. Decapoda was the most speciose order, with 57 species, followed by Isopoda (N = 48), Amphipoda (N = 39), Tanaidacea (N = 18) and Cumacea (N = 16). Three species of Decapoda represent new records for the Mexican Caribbean (Paguristes hernancortezi, Processa profunda, and Processa riveroi). Other specimens that were rarely observed in samples, and were not included in the data analyses, were crustaceans belonging to the class Ostracoda and to the subclass Copepoda, as well as animals belonging to Mollusca, Polychaeta, and Echinodermata.</ns0:p><ns0:p>In total, 868 specimens of crustaceans, consisting of at least 116 species, were obtained from the ASUs, and 1,872 specimens, consisting of at least 133 species, were obtained from coral rubble collected in situ (Table <ns0:ref type='table'>1</ns0:ref>). Species richness was not significantly different between methods (confidence intervals overlap; Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>) but the identity of the species differed, showing that both contribute to unique taxa; 45 were exclusive to ASUs and 62 were unique to coral rubble collected in situ. In both methods, over half of the species were represented by one or two Manuscript to be reviewed specimens (ASUs = 57%; Coral rubble = 52%). Overall, Decapoda was more speciose in ASUs, while Isopoda, Amphipoda, Tanaidacea and Cumacea were more speciose in coral rubble collected in situ (Table <ns0:ref type='table'>1</ns0:ref>). Regarding the number of individuals, Isopoda was the most abundant order in ASUs and Tanaidacea in coral rubble collected in situ; species of the order Cumacea were rare in samples obtained by both methods (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>Shannon and Simpson indexes were significantly higher in ASUs (H' = 39.4; D = 18.2) than in coral rubble collected in situ (H' = 27.3; D = 13.5) (confidence intervals don't overlap; Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>).</ns0:p><ns0:p>Rarefaction curves of species richness constructed to estimate the reliability of diversity estimates for both methods (Fig. <ns0:ref type='figure' target='#fig_10'>3</ns0:ref>) failed to reach a plateau, indicating that sample size was insufficient to reliably estimate the total number of species and thus diversity measurements for each method are conservative (Fig. <ns0:ref type='figure' target='#fig_9'>2</ns0:ref>). Estimates of sample coverage, a measure of sampling completeness, were 0.95 for ASUs, 0.97 for coral rubble collected in situ, and 0.98 when both methods were combined (Fig <ns0:ref type='figure' target='#fig_10'>3</ns0:ref>).</ns0:p><ns0:p>The nMDS plot, based on presence/absence data in Figure <ns0:ref type='figure'>4</ns0:ref>, showed no distinct separation of the cryptic crustacean assemblages in the two methods, as confirmed by the high stress value (0.1704). Assemblages obtained from coral rubble collected from in situ samples at different periods were more similar than samples of rubble in ASUs, nevertheless the samples from both methods overlap for some sampling periods; ASUs samples from the first and last surveys were more similar to coral rubble samples collected in situ than to ASUs samples collected in the second and third surveys (Fig. <ns0:ref type='figure'>4</ns0:ref>). ADONIS test indicated that the method had a small effect, although it was significant (R 2 = 0.1142, p = 0.0027).</ns0:p><ns0:p>According to the Importance Value Index (IVI), the dominant species differed between methods.</ns0:p><ns0:p>In ASUs, the dominant species were the tanaidacean Chondrochelia dubia (IVI = 9.5%) and the Manuscript to be reviewed isopod Cirolana parva (IVI = 6.2%), with other relatively important species being the amphipod Elasmopus rapax (IVI = 3.5%) and the tanaidacean Apseudes sp. A (IVI = 3.4%) (Fig. <ns0:ref type='figure' target='#fig_11'>5</ns0:ref>). In coral rubble collected in situ, the dominant species were the tanaidaceans Apseudes sp. A (IVI = 8.6%), Paratanais sp. A (IVI = 7.9%), Pseudoleptochelia sp. A (IVI = 6.2%) and Chondrochelia dubia (IVI = 5.2%) (Fig. <ns0:ref type='figure' target='#fig_11'>5</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>DISCUSSION</ns0:head><ns0:p>Artificial sampling units (ASUs) made with fresh coral rubble and deployed in different seasons for short periods of time (2-3 months) are an effective method for improving species inventory of motile cryptic crustaceans on Caribbean coral reefs. Using this method we recorded 116 species of this group during the one-year study duration at the Bonanza site, 45 of which were not recorded in the surrounding coral rubble. Nevertheless, the ASUs failed to record 62 species that were unique to the surrounding coral rubble. However, the rarefaction curves of species richness for both methods failed to reach a plateau, indicating that more samples were needed to have a complete inventory. By combining both ASUs and surrounding rubble samples, we recorded a total of 178 species, with 65% being represented by one or two individuals, and reached a sample coverage of 98% in our sampling size. The nMDS analyses showed no distinct separation of the cryptic crustacean assemblages obtained by the two methods, as samples obtained from ASUs in the first and last surveys overlapped with those obtained from coral rubble collected in situ.</ns0:p><ns0:p>However, ASUs samples from intermediate surveys were dissimilar to all others suggesting that it would take more ASUs to provide estimates of the community structure recorded in coral rubble samples, or that ASUs need to be left in place for longer periods. Further studies are Manuscript to be reviewed needed to determine at which point in time, or after how many samples, the two methods would yield similar results.</ns0:p><ns0:p>Our results support the finding of other studies which suggest that sampling of coral rubble using different techniques would render a higher taxonomic richness <ns0:ref type='bibr'>(Costello et al., 2017)</ns0:ref> and a greater potential for the discovery of new species <ns0:ref type='bibr'>(Souza, Oliveira &amp; Almeida, 2012;</ns0:ref><ns0:ref type='bibr' target='#b50'>Paz-R&#237;os, Sim&#245;es &amp; Ardisson, 2013)</ns0:ref>. Our ASUs were more effective in sampling decapods, with 23 out of the 57 species recorded being exclusive to this method, while the surrounding coral rubble was more effective for recording unique species of Amphipoda, Cumacea, Isopoda and Tanaidacea, even though some families of these orders were only sampled by ASUs, including the Amphilochidae and Bateidae, of the order Amphipoda, and the Munnidae, of the order Isopoda.</ns0:p><ns0:p>Decapod families exclusively found in ASUs were: Hippolytidae, Paguridae, Pilumnidae, Porcellanidae, Spongicolidae and Thoridae.</ns0:p><ns0:p>Differences between sampling methods in motile cryptic crustacean species richness, diversity, and assemblage composition could also be explained by the duration of time that each substrate remained underwater, and thus differences in the composition and coverage of algal turfs.</ns0:p><ns0:p>Peracarids like a layer of turf algae and fine sediment particles on which to feed, while decapods were more likely recruiting to ASUs for shelter from predators or could be actively foraging within the ASUs. Coral rubble within ASUs had low algal turf coverage, as it stayed in the water for only a few months (&#8804; 3). Although biofilms formed by bacteria and microalgae can be formed within hours <ns0:ref type='bibr'>(Cuba &amp; Blake, 2018)</ns0:ref>, the composition and coverage of the algal assemblage can change significantly within months <ns0:ref type='bibr'>(Frike et al., 2011)</ns0:ref>, as opportunistic filamentous species are replaced by more competitive fleshy ones and Cyanobacteria <ns0:ref type='bibr' target='#b69'>(Wanders, 1977;</ns0:ref><ns0:ref type='bibr'>Frike et al., 2011)</ns0:ref>. Given that the pattern of succession of algae can shape their Manuscript to be reviewed communities <ns0:ref type='bibr' target='#b10'>(Connell &amp; Slatyer, 1977)</ns0:ref>, the absence, or low coverage, of certain algal species could have inhibited the colonization or permanence of some of the cryptic crustacean species in the ASUs. Biofilms, for example, are known to release peptides that induce the settlement of several species of sessile invertebrate larvae <ns0:ref type='bibr' target='#b32'>(Johnson et al., 1997)</ns0:ref> and sessile assemblages on coral rubble may later affect the colonization of cryptic motile fauna <ns0:ref type='bibr' target='#b35'>(Klumpp, McKinnon &amp; Mundy, 1988;</ns0:ref><ns0:ref type='bibr' target='#b36'>Kramer, Bellwood &amp; Bellwood, 2012)</ns0:ref>. In the coral rubble collected around the ASUs, the algal-turf cover was higher, increasing habitat heterogeneity and allowing detritus to be trapped <ns0:ref type='bibr' target='#b16'>(Danovaro &amp; Fraschetti, 2002)</ns0:ref>. This possibly favored a higher species richness of peracarids, in particular of tanaidaceans, which are typically found in early successional stages <ns0:ref type='bibr' target='#b38'>(Larsen &amp; Shimomura, 2008)</ns0:ref>.</ns0:p><ns0:p>Despite its apparent permanence in back-reef environments, coral rubble cannot be seen as a static substrate, particularly in shallow reef sites, where it can be periodically reworked by currents and large wave events during storms, hurricanes, and north winds, or disturbed by fish feeding and bioerosion, among other factors, thus becoming periodically available for colonization <ns0:ref type='bibr' target='#b60'>(Takada, Abe &amp; Shibuno, 2007)</ns0:ref>. All these factors may drive the distribution and structure of cryptic assemblages <ns0:ref type='bibr' target='#b9'>(Choi &amp; Ginsburg, 1983;</ns0:ref><ns0:ref type='bibr' target='#b45'>Meesters et al., 1991)</ns0:ref> and contribute to the maintenance of high species diversity, by avoiding competitive exclusion and facilitating the colonization of less competitive species <ns0:ref type='bibr' target='#b22'>(Enochs et al., 2011)</ns0:ref>. A higher peracarid species richness would likely be obtained by increasing the number of ASUs per survey and allowing the artificial substrate to become covered by an algal matrix before deployment. Manuscript to be reviewed with isopod Cirolana parva, and the amphipod Elasmopus rapax: these species were probably opportunistic colonizers of new habitat space. Juveniles and ovigerous females of E. rapax, C. and decapods (families Alpheidae and Mithracidae) were observed in ASUs throughout the study. The dominant species in ASUs were previously reported as abundant in coral rubble substrates on the Puerto Morelos reef <ns0:ref type='bibr' target='#b46'>(Monroy-Vel&#225;zquez, Rodr&#237;guez-Mart&#237;nez &amp; Alvarez, 2017;</ns0:ref><ns0:ref type='bibr' target='#b72'>Winfield et al., 2017)</ns0:ref>, suggesting that, despite their artificial nature, ASUs were not only colonized by some of the most abundant reef species, but also by rare ones too. More studies are needed to determine if ASUs have an effect on the abundance, life stage, and sizes of the individuals recruited.</ns0:p><ns0:p>Our findings show that when assessing the effectiveness of ASUs on coral reefs, or other ecosystems, care should be taken in comparing the experimental results with controls collected simultaneously from the same sample station. Changes in either of these variables can produce significant differences in species composition and abundance that will affect comparisons <ns0:ref type='bibr' target='#b47'>(Moran &amp; Reaka-Kudla, 1991;</ns0:ref><ns0:ref type='bibr' target='#b60'>Takada et al., 2007)</ns0:ref>. Sessile encrusting or colony forming species are not expected to be common in ASUs, unless they remain in the water for several months or more <ns0:ref type='bibr' target='#b44'>(Malella, 2007;</ns0:ref><ns0:ref type='bibr' target='#b18'>Duckworth &amp; Wolff, 2011)</ns0:ref>. Once a broad survey of the species composition of the local coral rubble has been undertaken, it is then possible to evaluate the effectiveness of ASUs. Our results show that the use of ASUs made with defaunated coral rubble is effective in detecting cryptic and rare motile crustaceans, and can help improve species inventories of this group on Caribbean coral reefs.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Artificial sampling units (ASUs) made with defaunated coral rubble constitute a valuable tool to study the diversity of motile cryptic crustaceans in Caribbean coral reefs. Our results show that combining data from ASUs with that from surrounding coral rubble gives a more complete inventory of species, as both methods contribute unique species. ASUs gave a better estimate of diversity, whereas the surrounding coral rubble gave a better estimate of species richness. By combining both methods we recorded an assemblage of motile cryptic crustaceans composed of at least 178 species encompassing five orders at a single reef site in one year. </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>The dominant cryptic crustacean species, as obtained by the IVI, differed between sampling methods. Coral rubble substrates were dominated by tanaids (Apseudes sp., Paratanais sp., Pseudoleptochelia sp. and Chondrochelia dubia), while in ASUs, C. dubia was co-dominant PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Figure 5 Ecological</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='25,42.52,280.87,525.00,315.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>. Artificial Reef Matrix Structures, for example, are ASUs made of affordable materials which are designed to mimic</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:note></ns0:figure> <ns0:note place='foot' n='5'>to 20 cm (Fig.1B). The crate was anchored with a concrete block to prevent its displacement by waves and currents (Fig.1C); the block holes were open to the surface, allowing the PeerJ reviewing PDF | (2020:06:49584:2:0:NEW 19 Oct 2020)</ns0:note> </ns0:body> "

" PeerJ Editor October 19th 2020 Dear Dr. Robert Toonen, Please find below the answers (in blue) to the queries made for the article “The use of artificial substrate units to improve inventories of cryptic crustacean species on Caribbean coral reefs” (#2020:06:49584:1:1:REVIEW). Reviewer 1 (Anonymous) Basic reporting This second version is much more modest in its goals. The paper is well written with only a few sentences that need checking for English. There were no page or line numbers but an example is 'Cumacea was...'. The document was reviewed to correct mistakes. Figure legends are missing from revised draft. We don’t know why the reviewer could not see the figure legends nor the line numbers, as these appear fine in the PDF file that was created when the document was uploaded. Experimental design The authors have reduced their goals due to the very limited sampling. The design obviously cannot be changed at this time, so how to make the best of the data they have. I think they have done what can be done but need to add a few more thoughts to the Discussion. We added more comments to the discussion regarding the reviewers concerns. Validity of the findings The MDS tells us something the authors could point out. They do say that the ASU and the CG provide mostly different fauna lists. They admit that could be due to their limited sampling. But they should also talk about sampling issues for the cryptofauna in general. Why would the two sampling methods produce such differences in faunal composition. I disagree with the authors that there is much overlap of the two faunal lists, but that's not as important as understanding the differences. There is some discussion of that but that discussion avoids a question of how many samples would it take to get all the fauna. Figure 3 suggests that it would not take many in order to account for most of the species. The rarefaction curve, however, is an estimate of species accumulation. The MDS was run on species abundance data, not presence-absence data, and it shows that the ASUs don't help to estimate community structure of the cryptofauna, i.e., there is very little overlap of the two data sets. That is a difference that I think should be pointed out. Might also be worth running the MDS on presence data. I suspect there would be a greater overlap. The raw counts for the samples are not provided in the supplemental data so that can't be checked. In any case, it does seem like it would take a lot of ASUs to provide estimates of the community structure as seen in the coral gravel samples. In the new version, we followed the recommendation of the reviewer and ran the nMDS using presence/absence data, which in fact showed greater overlap. This is now discussed in the document. We also mention that the differences between methods could be due to the time the ASUs were in the water and to the amount of algal turf developed during that period of time, in contrast to in situ coral rubble which had spent months, or even years, on the site. We pointed out that 40% of the species recorded were shared by the two methods, and we recognize that either more ASUs will be needed to provide estimates of the community structure recorded in coral rubble samples, or that ASUs need to be left in place for longer periods. We also comment that further studies are needed to determine at which point in time, or after how many samples, the two methods would yield similar results. All data used in the analyses have been available since submission of the document in the Github folder mentioned in the methodology section. Comments for the Author This is much better and more modest. I think the idea of using ASUs to sample this habitat is an interesting idea, but you need to strengthen your case that it will be better than sampling the coral gravel itself, which is the conservation goal (I am very interested in conservation but I am not so sure that much is lost by sampling the coral gravel... you could take 10 samples of 250 cc and not make much of an impact on that habitat, although that might not be legal, as you point out). To replace that sampling, many more ASUs will need to be used. The results of our study show that diversity estimates of cryptic crustaceans would be improved by employing both methods; we never mention that ASUs would be a better method. All that said, I am pleased that someone is paying attention to the cryptofauna. Please keep working on this topic. Reviewer 2 (Ian Enochs) Basic reporting I think this manuscript is greatly improved, minor comments are provided directly below. Experimental design The shift in direction and design of the current study are entirely appropriate and I commend the authors for making the difficult decision to do so, as well as the work involved with revising the paper. I think it has been a worthwhile endeavor. Validity of the findings See comments below. Findings are sound. Comments for the Author Suggest: The use of artificial substrate units to the improve species inventories of … -or- improve inventories of cryptic crustacean species on … The title was modified as suggested by the reviewer Line 23: can cause disturbance that is unsustainable in light of widespread Done Line 25: coral rubble Done Line 27: the composition Done Line 31: rubble not gravel, here and throughout. Done Line 34: with sampling of the surrounding… Done Line 44: comma after citation, before “with” Done Line 44: Suggest checking out the following reference: Small AM, Adey WH, Spoon D (1998) Are current estimates of coral reef biodiversity too low? The view through the window of a microcosm. Atoll Res Bull 458:1-20. The citation is now included Line 55: write out numbers less than 10… here and throughout. Done Line 58: crustaceans Corrected Line 61: Not entirely fair to single these authors out as the only destructive samplers. Numerous people have sampled rubble (including Monroy-Velazques et al. 2017) and live corals as they did. We meant to say that the comment was mentioned by these authors, not that they are destructive samplers! The references were removed to avoid confusion. Line 72: Wrong paper. Enochs et al. 2011 used artificial rubble bags. Corrected Line 78: Suggest checking out: Valles H, Kramer DL, Hunte W (2006) A standard unit for monitoring recruitment of fishes to coral reef rubble. J Exp Mar Biol Ecol 336:171-183. The citation is now included Line 79: Suggest different word than “dominant.” Not clear what you mean. I’m guessing most abundant substrate, though the references cited indicate “play an important role in harboring diverse cryptofaunal communities. Modified as suggested by reviewer Line 116: Length? Width? Circumference? Corrected to diameter Line 136: Need to give brief description here so the reader can figure out what you did without reading the other paper. Okay to provide less details and point the reader to the other paper but still need a synopsis. Modified as suggested by reviewer Line 146. Suggest new sentence Done Line 149: I would suggest sqrt transformation to decrease the potential for extremely abundant species to strongly skew your assemblage comparison. This is pretty common practice. It may also help to deal with the very high stress in your nmds plot. In the new version we conducted the nMDS using presence/absence data as recommended by the other reviewer. Line 188: italicize. Done Lines 194: and 205: Odd spacing. Corrected Line 235: Looking for space is conversational. Suggest “recruiting to ASUs for shelter from predators” …though previous studies have found abundant cryptic predator populations. I also think that it is entirely possible that herbivorous, detritivorous, and carnivorous species were actively foraging within the ASU’s. Trophic consideration of the decapods considered may be helpful in making this case. The sentence was modified. Line 251: reference? A reference was included Line 278: Not sure if I follow. What if the most abundant species are dependent on a food source that takes time to recruit to the ASU’s? The sentence was removed as we agree with the reviewer that it was unclear. We are very thankful for the comments of the editor and the two reviewers, as they were very helpful to improve the manuscript. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>A variety of predictors are available for ovarian stimulation cycles in assisted reproductive technology (ART) forecasting ovarian response and reproductive outcome in women including biomarkers such as anti-Mullerian hormone (AMH). The aim of our present study was to compare the relationship between AMH levels and pregnancy outcomes in patients undergoing intra-cytoplasmic sperm injection (ICSI). Overall, fifty patients (n=50), aged 20-45 years were recruited for the present prospective study. Three AMH levels were presented with high often poly cystic ovarian syndrome (PCOS) amongst 52.4% patients, 40.5% in normal and 7.1% in low to normal, correspondingly. There was statistically significant relationship between AMH and day of embryo transfer (p&lt; 0.05). The Pearson analysis between AMH, age, E2 and FSH displayed no statistically significant relationship between E2 and AMH (p &lt; 0.05) and negative correlation between FSH and age (p &gt; 0.05).</ns0:p><ns0:p>The area under the receiver operating characteristic curve for E 2 was 0.725 and for AMH levels as predictors of CPR was 0.497 indicating E 2 as better predictor than AMH. The number of oocytes, mature oocytes and fertilized oocytes all presented a weak positive relationship to AMH. Our results confirm the clinical significance of AMH to accurately predict ovarian reserve as a marker and its limitations to use as predictor for a positive pregnancy outcome. Additional prospective studies should be conducted to validate the predictive capability of AMH levels for the outcome of clinical pregnancy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Couples in modern societies postpone childbearing amidst busy schedules and career positionings;</ns0:p><ns0:p>trying to conceive at a more advanced age contributing to a rise in the occurrence of infertility <ns0:ref type='bibr' target='#b2'>(Caroppo et al., 2006)</ns0:ref>. Most women are unaware that fertility starts to decline after the early thirties in some individuals. With increasing female age, fecundity in natural and stimulated ovarian cycles declines, as observed in population-based studies <ns0:ref type='bibr' target='#b8'>(Grynnerup et al., 2012)</ns0:ref> as well as in IVF studies <ns0:ref type='bibr' target='#b19'>(Scheffer et al., 2018)</ns0:ref>. For this reason, there is a growing number of women of advanced age seeking treatment for infertility <ns0:ref type='bibr' target='#b14'>(Oskayli et al., 2019)</ns0:ref>.</ns0:p><ns0:p>The hormonal control of ovarian function by the gonadotropins plays a vital part in the physiological process of follicular growth <ns0:ref type='bibr' target='#b17'>(Richards, 2018)</ns0:ref>. Over the last decade, Roberto <ns0:ref type='bibr' target='#b15'>(Palermo, 2007)</ns0:ref> in 2007, correlated the contribution of follicle stimulating hormone (FSH) and luteinizing hormone (LH) to follicular development by clinical data obtained from assisted reproductive technique (ART) cycles performed using gonadotropin-releasing hormone (GnRH) agonist protocols. In all patients, the serum hormone levels of these hormones can be used to evaluate the endocrine environment of the follicles.</ns0:p><ns0:p>In assisted reproduction, serum levels for several hormones are used to assess the ovarian reserve and to monitor the development of the follicles that have been stimulated by gonadotrophins <ns0:ref type='bibr' target='#b0'>(Alson et al., 2018)</ns0:ref>. Traditional techniques used to predict ovarian stimulation have incorporated serum levels of hormones such as FSH, LH and estrogen (E 2 ) along with ultrasonographic guides such as ovarian volume and number of early antral follicles as a reliable predictor of the outcome of in vitro fertilization (IVF) <ns0:ref type='bibr' target='#b11'>(Kunt et al., 2011)</ns0:ref>. Over the last years, Anti-Mullerian hormone <ns0:ref type='bibr'>(AMH)</ns0:ref> has been projected as a novel marker for predicting ovarian response to gonadotrophin stimulation <ns0:ref type='bibr' target='#b0'>(Alson et al., 2018</ns0:ref><ns0:ref type='bibr'>, Zargar et al., 2018)</ns0:ref>. AMH is a dimeric glycoprotein strongly produced by the granulosa cells of the pre-antral (primary and secondary) and small antral follicles (AF's) in the ovary and shown to be age dependent <ns0:ref type='bibr' target='#b18'>(Sahmay et al., 2014)</ns0:ref>. Measurement of anti-M&#252;llerian hormone in serum is much more precise measure of the ovarian reserve than the other hormones that have previously been available to us <ns0:ref type='bibr' target='#b1'>(Anderson et al., 2012)</ns0:ref>.</ns0:p><ns0:p>The aim of this prospective study was to investigate the relationship between Anti-M&#252;llerian hormone levels and pregnancy outcomes in patients undergoing in-vitro fertilization or intracytoplasmic sperm injection (ICSI). </ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Patients</ns0:head></ns0:div> <ns0:div><ns0:head>GnRH antagonist Protocol:</ns0:head><ns0:p>A gonadotrophin-releasing hormone (GnRH) antagonist protocol with recombinant FSH (GONAL-f, Merck Serono, Darmstadt, Germany) was used as downregulatory <ns0:ref type='bibr' target='#b16'>(Park et al., 2015)</ns0:ref>. The second approach was followed by administering 0.25 mg/day Cetrotide (Merck Serono). When at least 3 or more follicles reach a diameter equal or above 17 -18 mm, the endometrial thickness reached at least 7 mm by ultrasound and E2 levels were about 1500-1800 pmol/L then Human chorionic gonadotropin (hCG) was administered. All patients received 5000 -10 000 IU hCG (Ovitrelle&#174;, Merck Serono). Oocyte retrieval was performed 36 hours after the administration of the hCG. Conventional IVF or ICSI was performed according to previously described protocols.</ns0:p><ns0:p>Sample collection: Blood samples were collected every 3-4 days on commencement of the treatment. The blood samples were centrifuged at 3000 rpm for 10 minutes using a Biofuge centrifuge (Biofuge Primo -Heraeus) to obtain the blood serum. AMH and FSH levels were recorded, upon the first visit. Estrogen(E 2 ) and LH levels were monitored throughout the program until a peak E 2 and LH level were reached.</ns0:p><ns0:p>Hormone Assays: Ultra-sensitive Beckman Coulter second generation AMH assay was used to estimate AMH levels from the blood serum <ns0:ref type='bibr' target='#b10'>(Kumar et al., 2010)</ns0:ref>. Similarly, Enzyme-linked immunosorbent assays (ELISA) (Beckman Coulter) was used to analyse hormone levels (FSH, E 2 and LH) from the blood serum. Insemination and intra cytoplasmic sperm injection (ICSI), oocyte retrieval, culture, fertilization, embryo culture, and transfer were carried out as previously described by Gardener et al. <ns0:ref type='bibr' target='#b5'>(Gardner et al., 2001)</ns0:ref>. Inclusion Criteria: The population of the study included female patients ranging between the ages of 20-45 Exclusion Criteria: Patients undergoing cancer therapy and patients on immune suppressant drugs were excluded from study.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis:</ns0:head><ns0:p>The data were analysed using IBM SPSS software (Chicago, IL, USA).</ns0:p><ns0:p>A non-parametric Pearson's correlation was used to determine the direction, strength, and significance of the correlation between X and Y variables between the different semen parameters.</ns0:p><ns0:p>A parametric multiple linear regression analysis was used to evaluate the relationship between AMH and other available endocrine markers. ROC curves were used to assess predictive value for E 2 and AMH and evaluating cut off values to optimise sensitivity and specificity. A p value of &lt; 0.05 was considered statistically significant. </ns0:p></ns0:div> <ns0:div><ns0:head>Institutional Review</ns0:head></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The prospective study included fifty patients who met the inclusion criteria. From the initial sample size of fifty, forty-two presented with data that could be analysed whilst 8 patients had oocytes that where abnormal and did not result in transfer. The data from these 8 patients were not included in the study due to poor embryo development (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Table 1. AMH distribution in blood samples</ns0:head><ns0:p>Amongst the 42 patients analysed, 4.76% were between 20-24 years, 9.52% were between 25-29 years, 40.47% were between 30-34 years, 35.7% were between 35-39 years and 9.52% were between 40-44 years, respectively. As demonstrated by this study the clinical pregnancy rate for patients 20 -24 years was 100%, 25 -29 years was 50%, 30 -34 years was 17.6%, 35 -39 years was 26.6% and 40 -44 years was 25% (Fig. <ns0:ref type='figure'>1</ns0:ref>).</ns0:p><ns0:p>Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref> shows number of oocytes retrieved, number of oocytes matured, and number of oocytes fertilized into respective categories. Not all eggs obtained were at the metaphase 2 stages and had to be matured in the incubator overnight and injected the following day. The results shown were to some extent anticipated as AMH has been used an indicator of oocyte reserve in previous studies <ns0:ref type='bibr'>(Yarde et al., 2013</ns0:ref><ns0:ref type='bibr' target='#b24'>, Yao et al., 2015)</ns0:ref> whereas the resulting fertilized or transferred embryo's may be due to a chance process based on many various factors such as quality of oocyte and sperm.</ns0:p><ns0:p>The Chi-square test for Independence was performed to check whether there was an association between the number of oocytes fertilized and the AMH category (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). A Chi-squared value of 18.5, degrees of freedom = 12, with a p = 0.10 was found. There was no statistically significant relationship between numbers of oocytes fertilized versus AMH category (p &gt; 0.05).</ns0:p><ns0:p>Out of 22 patients, 43 embryos were transferred. Embryos were transferred depending on embryo development and the number of embryos obtained. Most patients in the high and normal categories resulted in a day 5 transfer, the Chi-squared test for independence of AMH and number of embryos transferred gave a Chi-squared value of 6.384 with df = 4 and p-value = 0.172(Table <ns0:ref type='table'>3</ns0:ref>), thus statistically no significant association between AMH and number of embryos transferred was observed. Whilst, Chi-square test for independence between the variables AMH and day of embryo transfer (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>) gave a Chi-square value of 14.117, 6 degrees of freedom and p = 0.028 indicating statistically significant relationship between AMH and day of embryo transfer (p&lt; 0.05).</ns0:p><ns0:p>Pregnancy outcome and AMH category are as shown in Table <ns0:ref type='table' target='#tab_4'>5</ns0:ref>. Out of twenty two cases in high category, 6 resulted in a positive pregnancy; 6 resulted in a positive outcome (6/12 = 50.0%) (Normal); while out of the 3 cases where the AMH was 'Low to Normal', there were no pregnancies reported. The Chi-squared test for independence of AMH category and pregnancy outcome gave a Chi-Squared value of 0.502, 2 degrees of freedom and p = 0.778. There was no statistically significant association between the pregnancy outcome and the AMH category (p &gt; 0.05).</ns0:p></ns0:div> <ns0:div><ns0:head>Pearson correlation</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Pearson correlation coefficients were calculated to determine if any statistical significance exists between AMH on a quantitative scale and age, E 2 and FSH (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>). The Pearson Correlation coefficient of 0.151 indicates that a very weak positive relationship existence between E 2 and AMH, which is not statistically significant (p = 0.341). Furthermore, Pearson correlation coefficient between the AMH and age had a coefficient of -0.028 thus showing a weak, negative correlation p = 0.859 (p &gt; 0.05). The Pearson Correlation between AMH and FSH produced a coefficient of -0.185 thus showing a weak, negative correlation p = 0.240 (p &gt; 0.05). Pearson correlation coefficient showed no significant relationship between AMH and number of oocytes (p = 0.191), number of mature oocytes (p = 0.300) and number of oocytes fertilized (p = 0.146). The number of oocytes, mature oocytes and oocytes fertilized all presented a no statistically significant correlation with AMH (0.206, 0.164, and 0.228, respectively).</ns0:p></ns0:div> <ns0:div><ns0:head>Logistic regression analysis</ns0:head><ns0:p>A logistic regression model was used to determine the possible predictor variables for the pregnancy outcome. The model was fitted to the data with the result of the pregnancy namely, 'Positive' or 'Negative' as the binary dependent variable and age, E 2 LH, Basal FSH, Basal AMH and number of oocytes fertilized as independent variables (Table <ns0:ref type='table'>7</ns0:ref>). As shown in Table <ns0:ref type='table'>7</ns0:ref>, LH has p = 0.042 (p &lt; 0.05) and E 2 has p = 0.065, which is statistically not significant at a 5% level. The SPSS output for the model is given in Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>(Supplementary data S1) signifying that overall, 73.8 % of the cases were correctly classified, while 5/12 =0.417 or 41.7 % of the positives were correctly classified, and 86.7% of the negative cases were correctly classified.</ns0:p></ns0:div> <ns0:div><ns0:head>Area under the curve</ns0:head><ns0:p>The ROC curves of the serum AMH concentrations and E 2 for the prediction of the clinical pregnancy are depicted in Fig. <ns0:ref type='figure'>2</ns0:ref>. The areas under the curves (AUC) for E 2 were 0.725 and for AMH (AUC = 0.497). E 2 is therefore a better single predictor of pregnancy outcome when compared to AMH. It has been shown that E 2 can better predict the number of oocytes obtained.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In the current study, we investigated the relationship between AMH levels and pregnancy outcomes in patients undergoing intra-cytoplasmic sperm injection.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between basal AMH and E 2</ns0:head><ns0:p>Pearson analysis between E 2 and AMH presented a Pearson Correlation co-efficient of 0.151 with p = 0.341 (p &lt; 0.05) which indicates that a weak significant relationship exists between E 2 and AMH. Most previous studies (Ramalho <ns0:ref type='bibr' target='#b16'>de Carvalho et al., 2012</ns0:ref><ns0:ref type='bibr' target='#b20'>, Ubaldi et al., 2005)</ns0:ref> have shown a relationship between a raised basal E 2 level and a reduced ovarian response using different values to express elevated estrogen levels which replicated the findings in this study therefore showing that a low AMH can result in low estrogen levels. Also, it can be concluded that a poor AMH value results in a poor ovarian reserve indicating follicles produced will not be correlated to a raised estrogen level, therefore indicating poor follicle growth, thus reducing the number of oocytes produced. However, it was determined that meagre response to stimulus in IVF, indicative of a lower ovarian reserve, is associated with declined baseline serum AMH concentrations <ns0:ref type='bibr' target='#b23'>(Van Rooij et al., 2004)</ns0:ref>. Consequently, when women have regular ovarian reserve and decent retort, disappointment of IVF must look for additional infertility reasons, e.g., male specific issue i.e. Y chromosome microdeletion. Furthermore, this conclusion is reinforced by the data of woman undergoing IVF which indicated that male factor infertility resulted in an unsuccessful cycle.</ns0:p><ns0:p>Although E2 levels in these cases were above those of controls, they are still within the range of 25-100 pg/ml <ns0:ref type='bibr' target='#b18'>(Sahmay et al., 2014)</ns0:ref>, suggesting that E2 single-handedly is not capable of predicting the female reproductive potential.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between basal AMH and Age</ns0:head><ns0:p>Pearson correlation between AMH and age (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) presented a co-efficient of -0.028 thus displaying a weak, negative correlation and with a p = 0.859 (p&gt; 0.05). A stronger relationship between these two variables was expected as it is known that as age increases, AMH should decrease. This contrary association corresponds as reported by <ns0:ref type='bibr' target='#b23'>Van Rooij et al., (2004)</ns0:ref>, where serum AMH levels decline with age in normal women with proven fertility. Also, it is suggested that serum AMH is identified as the improved endocrine marker to measure the reproductive capability in advanced age.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between Basal AMH and FSH</ns0:head><ns0:p>Basal FSH is one of the primary endocrine markers presented into ART program. The Pearson correlation amid AMH and FSH (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) had a coefficient of -0.185 thus displaying a weak, negative correlation and with a p= 0.240 (p&gt; 0.05). This study specifies a negative correlation, i.e., the higher the FSH the higher the chances the patient can present with a poor ovarian reserve and early menopause. It is routine practice to frequently measure the basal FSH level and to start IVF treatment only when the FSH level is lower than threshold value in a cycle. This was strategic on the awareness that these women will retort well to ovarian stimulation while the basal FSH level is lesser at the beginning of the cycle. The outcomes of this study revealed that woman who were poor respondents or had a reduced ovarian reserve had a poor outcome and frequent testing is of no worth. Women who had a history of high FSH level must undergo treatment without further delay. By postponing treatment for these patients can be detrimental as they get older and fast approaching menopause <ns0:ref type='bibr' target='#b21'>(Uzumcu and Zama, 2016)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between Age and FSH and Number of oocytes</ns0:head><ns0:p>The Pearson Correlation between FSH and age (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) displayed no statistical significance, p = 0.583 (p &gt; 0.05). For most of Pearson Correlation analysis, no significant relationships were found with most of the p-values, being greater than 0.05. This may be due to the small sample size used in this study of 42 patients. AMH being compared to age and number of oocytes showed a slightly negative correlation which is expected as it is shown in previous studies that AMH and number of oocytes decrease with maternal age <ns0:ref type='bibr' target='#b23'>(van Rooij et al., 2004)</ns0:ref> and <ns0:ref type='bibr'>Gobikrushant et al. (2018)</ns0:ref>. This inverse relationship is in agreement by <ns0:ref type='bibr' target='#b23'>Van Rooij et al. (2004)</ns0:ref>, who reported that serum AMH levels deteriorate with age in normal women with proven fertility. Additionally, serum AMH indicates the simplest endocrine marker to measure the age-related decline of reproductive competence. AMH levels, in our group who were high respondents were over 3.0 ng/ml, normal respondent over 1.0 ng/ml and low respondents found to be below 0.9 ng/ml. Oocytes were still recovered even with low AMH levels. Neither fertilization rate nor embryo quality can be assessed using basal AMH levels. This contrasts with the findings reported by <ns0:ref type='bibr' target='#b22'>Vaegter et al. (2017)</ns0:ref>, where embryos had superior morphology and cleavage performance in patients with AMH levels &gt; 2.7 ng/ml as compared with patients with values below this threshold.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between AMH and the number of oocytes, number of mature oocytes and number of oocytes fertilized</ns0:head><ns0:p>Our second objective of the study was to examine if AMH levels affected oocyte quality. In this study, the Pearson Correlation test (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) showed no significant relationship between AMH and number of oocytes (p = 0.191), several mature oocytes (p = 0.300) and number of oocytes fertilized (p = 0.146). The number of oocytes, mature oocytes and oocytes fertilized all showed a weak positive relationship to AMH (0.206, 0.164, and 0.228, respectively). These findings are in agreement with that reported by La Marca and Sunkara (2014); La <ns0:ref type='bibr' target='#b12'>Marca et al., (2010)</ns0:ref> and <ns0:ref type='bibr'>Dehghani-Firouzabadi et al. (2008)</ns0:ref>, where mean amount of oocytes was lower in poor responding patients than in normal patients attending IVF programs. This therefore led to the inference that ovarian response can be regarded as a reflection of the ovarian reserve. The Chi-square test for Independence was done to determine whether there is an association between the number of oocytes collected and the AMH category (Table <ns0:ref type='table'>8</ns0:ref>). A Chi-squared value of 21.246, degrees of freedom = 8, with a p = 0.007 was observed. There was a significant relationship between the numbers of oocytes collected versus AMH category (p &lt; 0.05). The Chi-square test for Independence was performed to see whether there is an association between the number of oocytes fertilized and the AMH category (Table <ns0:ref type='table'>9</ns0:ref>). A Chi-squared value of 18.5, degrees of freedom = 12, with a p = 0.10 was found. There was thus no statistically significant relationship between the numbers of embryo's fertilized versus AMH category (p &gt; 0.05).This is anticipated as AMH has been used an indicator of oocyte reserve in previous studies whereas the resulting fertilized or transferred embryo's may be due to a chance process based on many various factors such as the quality of the oocyte and sperm <ns0:ref type='bibr'>(Yarde et al., 2013)</ns0:ref>. <ns0:ref type='bibr' target='#b4'>Ebner et al. (2006)</ns0:ref>, demonstrated that AMH serum levels were related with oocyte quality in stimulated cycles. The quality of the embryos was not assessed using baseline AMH which agrees with our findings. However, the fertilization rate was not correlated with the serum AMH which varied with the results of the present study.</ns0:p></ns0:div> <ns0:div><ns0:head>AMH category and positive pregnancies</ns0:head><ns0:p>Embryo quality has been suggested to be of paramount importance to predict the occurrence of Manuscript to be reviewed</ns0:p><ns0:p>In this study, AMH value for predicting pregnancy outcomes does not exist because oocyte quality is not accounted for by ovarian reserve markers. As demonstrated in this study the clinical pregnancy rate for patients 20 -24 years was 100%, 25 -29 was 50%, 30 -34 years was 18%, 35</ns0:p><ns0:p>-39 years was 27% and 40 -44 years was 25% (Figure <ns0:ref type='figure'>1</ns0:ref>). Patients presenting with a low AMH did not vary from those women presenting with higher AMH concentrations in same age group. A positive pregnancy outcome was logged across all age groups regardless of the AMH level. These results advocate that low ovarian reserve is not correlated with low oocyte quality in patients and the prediction remains the similar despite low AMH concentrations. <ns0:ref type='bibr' target='#b9'>Kini et al. (2010)</ns0:ref> stated the role of AMH in foreseeing cumulative pregnancy outcome during IVF treatment. It was recognized that serum AMH concentration on day 6 of stimulation was suggestively higher in participants who resulted in an ongoing pregnancy in IVF compared to those who did not. Serum AMH is a suitable indicator of ovarian hyper-response. In a metanalysis study conducted by <ns0:ref type='bibr' target='#b24'>Yao et al. (2015)</ns0:ref> to evaluate role of serum AMH role in forecasting the pregnancy outcome in IVF/ICS, it was concluded that there is positive correlation between serum AMH and pregnancy. Nevertheless, association between serum AMH and non-pregnancy cannot be ruled out either.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In conclusion, the outcomes of our investigations specify that AMH has established to be a valuable marker for ovarian reserve and might benefit woman who plan for pregnancy. AMH hormone seems to be the best endocrine marker, however, the valuable role of AMH and its role in ovarian function should be looked at in relation to the other markers to assess the decline of the ovarian pool. Manuscript to be reviewed </ns0:p></ns0:div> <ns0:div><ns0:head>Pregnancy outcome and AMH category</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Pearson correlation between basal AMH and E 2, Age and FSH and oocytes</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>:</ns0:head><ns0:label /><ns0:figDesc>Fifty women(n=50), aged 20-45 years were recruited from Centre of Assisted Reproduction and Endocrinology (C.A.R.E) Clinic in Westville, Durban, South Africa who were undergoing IVF treatment. This study was approved by Ethical Committee of the Durban University of Technology (Project reference 128/16) and Research Committee, C.A.R.E. Clinic, Durban, South Africa. After approving the study by the research ethics committees, written informed consents were obtained from all the patients.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Board approval: This study was approved by Ethical Committee of the Durban University of Technology (Project reference 128/16) and Research Committee, C.A.R.E. Clinic, Durban, South Africa and was performed in accordance with the Helsinki Declaration of 1975 (as revised in 1983).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>pregnancy after IVF. In a regression model E 2 has a p =0.017 (p &lt;0.05) and LH has a p =0.035 (p &lt;0.05). Both variables are significant and age and basal AMH play a role in the pregnancy outcome and the model is thus adjusted for these two variables PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Yarde F, Voorhuis M, Dolleman M, Knauff EAH, Eijkemans MJC &amp; Broekmans FJM. 2013. Antimullerian hormone as predictor of reproductive outcome in subfertile women with elevated basal folliclestimulating hormone levels: a follow-up study. Fertility and Sterility, 100, 831-+. Zargar M, Najafian M &amp; Zamanpour Z. 2018. Relationship between follicular fluid and serum anti-Mullerian hormone levels and pregnancy rate in ART cycles. Perinatolog&#237;a y Reproducci&#243;n Humana, 32, 3-8.</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,188.32,525.00,294.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>While appropriate reference values are being generated per age category and until the consequences of having a low or high AMH for one's age are being established, AMH should only be determined in the context of clinical studies. At present, the most important clinical role of AMH at this stage is to serve as a red flag for reduced ovarian reserve in women of reproductive age who must undergo further diagnostics. As per the study conducted, we can infer that AMH can accurately predict ovarian reserve but cannot predict the oocyte quality or a positive pregnancy outcome. The more oocytes obtained, increases a patient's chance of more viable embryos and therefore, improving chances of a healthy pregnancy and ultimately a live birth. Further research</ns0:figDesc><ns0:table /><ns0:note>on the implication of varying levels of AMH within the follicular fluid may be representative as an indicator of 'quality' in addition to the number of growing follicles.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='2'>. AMH distribution in blood samples</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Category</ns0:cell><ns0:cell cols='3'>AMH Blood Level Concentration Frequency % Patients</ns0:cell></ns0:row><ns0:row><ns0:cell>High (often PCOS)</ns0:cell><ns0:cell>&gt; 3.0 ng/ml</ns0:cell><ns0:cell>22</ns0:cell><ns0:cell>52.4</ns0:cell></ns0:row><ns0:row><ns0:cell>Normal</ns0:cell><ns0:cell>&gt; 1.0 ng/ml</ns0:cell><ns0:cell>17</ns0:cell><ns0:cell>40.5</ns0:cell></ns0:row><ns0:row><ns0:cell>Low Normal Range</ns0:cell><ns0:cell>&#8804; 0.3 -0.9 ng/ml</ns0:cell><ns0:cell>3</ns0:cell><ns0:cell>7.1</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Correlation between AMH and number of oocytes collected, matured, and fertilised during stimulation</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Total no.</ns0:cell><ns0:cell>Total no.</ns0:cell><ns0:cell>Total no.</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>of oocytes collected</ns0:cell><ns0:cell>of oocytes matured</ns0:cell><ns0:cell>of oocytes fertilized</ns0:cell><ns0:cell>% collected oocytes</ns0:cell><ns0:cell>% matured oocytes</ns0:cell><ns0:cell>% oocytes fertilized</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>High</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>62</ns0:cell><ns0:cell>69</ns0:cell><ns0:cell>65.4%</ns0:cell><ns0:cell>60%</ns0:cell><ns0:cell>61.6%</ns0:cell></ns0:row><ns0:row><ns0:cell>AMH</ns0:cell><ns0:cell cols='2'>Normal 38</ns0:cell><ns0:cell>36</ns0:cell><ns0:cell>38</ns0:cell><ns0:cell>30.6%</ns0:cell><ns0:cell>35%</ns0:cell><ns0:cell>33.9%</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>category Low to</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>4.0%</ns0:cell><ns0:cell>5%</ns0:cell><ns0:cell>4.5%</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell /><ns0:cell>124</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Chi Square analysis of AMH and day of embryo transfer. . 10 cells (83.3%) have expected count less than 5. The minimum expected count is 0.05</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Value</ns0:cell><ns0:cell>Degrees of Freedom</ns0:cell><ns0:cell>Asymptotic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(df)</ns0:cell><ns0:cell>Significance (2-sided)</ns0:cell></ns0:row><ns0:row><ns0:cell>Pearson Chi Square</ns0:cell><ns0:cell>14.117 a</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0.028</ns0:cell></ns0:row><ns0:row><ns0:cell>Likelihood Ratio</ns0:cell><ns0:cell>6.432</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0.377</ns0:cell></ns0:row><ns0:row><ns0:cell>N of Valid Cases</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>aPeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 5 (on next page)</ns0:head><ns0:label>5</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Pregnancy outcome and AMH category</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Pregnancy Result</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 6 (on next page)</ns0:head><ns0:label>6</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Pearson correlation between basal AMH and E 2, Age and FSH and oocytes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>E 2</ns0:cell><ns0:cell>Age</ns0:cell><ns0:cell>FSH</ns0:cell><ns0:cell>Number</ns0:cell><ns0:cell>Number of</ns0:cell><ns0:cell>Number of oocytes</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>of</ns0:cell><ns0:cell>mature</ns0:cell><ns0:cell>fertilized</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>oocytes</ns0:cell><ns0:cell>oocytes</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Pearson</ns0:cell><ns0:cell cols='2'>0.151 -0.028</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.206</ns0:cell><ns0:cell>0.164</ns0:cell><ns0:cell>0.228</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>correlation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.185</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>coefficient</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>AMH</ns0:cell><ns0:cell>Significance</ns0:cell><ns0:cell cols='3'>0.341 0.859 0.240</ns0:cell><ns0:cell>0.191</ns0:cell><ns0:cell>0.300</ns0:cell><ns0:cell>0.146</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>value (2-tailed)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>No. in the</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>sample</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell>Pearson</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>-</ns0:cell><ns0:cell>-0.271</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Correlation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.087</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Significance</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.583</ns0:cell><ns0:cell>0.082</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>value (2-tailed)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>No. in the</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>sample</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:52172:1:1:NEW 13 Oct 2020)</ns0:note> <ns0:note place='foot' n='103'>112100% 100%</ns0:note> </ns0:body> "

"Correlation between AMH level and FSH as predictors of pregnancy outcomes in patients undergoing intra-cytoplasmic sperm injection Manuscript ID: 52172 Reviewer 1 comments Reviewers comments Basic Reporting: The objective of the study was to investigate the relationship between Anti-Müllerian hormone levels and pregnancy outcomes in patients undergoing in-vitro fertilization or intracytoplasmic sperm injection (ICSI), although, it does not match (not reflect) the title of the manuscript Experimental Design 1. The age of the patients (> 40 years) can in itself negatively influence the assisted reproduction outcomes, ends up becoming a study bias. Authors Comments We appreciate reviewer comments for bringing this to our attention. In revised manuscript, we have modified title to meet the criteria and to attract great International readers of PeerJ as “The relationship between Anti-Mullerian Hormone (AMH) levels and pregnancy outcomes in patients undergoing Assisted Reproductive Techniques (ART)” 1. Women aged between 20 to 45 years old without history of other diseases and coming from different ethnic backgrounds were recruited for the present study. We therefore looked at females in all age groups within reproductive age for the study. Furthermore, fecundability deteriorates significantly since the early 30s (Faddy et al., 1992) and prevalence of infertility increases suggestively after the age of 35 years. Besides, 99% of patients are expected to be infertile by the age of 45 (Menken et al., 1986). 2. What was the total dose of gonadotropins used? Did all patients use the same amount, regardless of age? Outcomes should not be compared if different protocols or different amounts of gonadotropins were used. 2. All patients received different dose of gonadotrophins and same protocol was adopted for all patients in present study. Females are born with a fixed number of primordial follicles, resting in a dormant stage of meiosis II until puberty. The quality as well as the quantity of the primordial follicle constitutes the ovarian reserve (Broer et al., 2012). Moreover, the primary focus of the study was to determine the relationship between AMH and egg reserves, embryo quality and pregnancy outcomes (Note: Manuscript title is also changed accordingly. 3. What were the infertility causes of the patients? Ovulatory factors, endometriosis and severe male factor also negatively influence reproduction outcomes, influence AMH serum values and therefore are biases in the study. Did all patients have both ovaries? 3. Causes of infertility were unknown for patients in the present study. Serum AMH levels have been measured at different times in the menstrual cycle, with extremely subtle or non-existent fluctuation. Minimal fluctuations in serum AMH levels may be consistent with continuous noncyclic growth of small follicles. Hence AMH is relatively convenient to determine especially as it seems to be stable throughout the cycle making it determinant of ovarian activity (Marca et al., 2004 and Grujters et al., 2003). Yes, all patients had both ovaries. 4. The number of patients is another very important issue. The analysis of 50 or 42 patients is insufficient to analyze whether a serum measurement can be used as a biomarker 4. AMH is considered to be a marker that can estimate the quantity and activity of retrievable follicles in early stages of maturation, thus being more reliable for the prediction of ovarian response and reproductive potential which is supported by many other studies (Franchin et al., 2003, Visser et al., 2005, Rooij et al., 2002 and Muttukrishna et al., 2005). Overall, the number of patients to be recruited for present study was approved by qualified statistian (G Mathews, HoD, Durban University of Technology) to meet the criteria of study and to meet the regulations of the 1 University which is approved by Institutional Ethics committee in local setting. This is ongoing project and Prof Matthews went through previous articles before settling on a sample size and given the fact that we were in a small clinical setting which needed to be taken into consideration as primary limitation in South Africa. We take this feedback extremely valuable suggestion and increase the patient’s number in all our future studies. Validity of the findings 1. It is very difficult to assess whether the results are valid in the face of so many biases found at the study. 2. You need to adjust the number of patients and especially the patient inclusion criteria. 1. AMH was the only marker of ovarian reserve showing a decline over time both in woman younger (<35 years) and in woman older than 40 years of age. With respect to other known markers, AMH seemed to better reflect the continuous decline of the oocyte/follicle pool with age. The decrease in AMH with advancing age indicates that serum AMH levels may be the best marker for ovarian ageing and menopausal transition (Van Rooij et al., 2002). 2. The revised manuscript has now been supplemented with patient’s inclusion criteria (Line 113-114) Reviewer 2 comments Basic reporting 1. Clarity is needed on lines 61-67. Is the intent in this paragraph to indicate that GnRH agonist induced gonadotropin levels correlate with antral follicle counts? Is this referring purely to endogenous gonadotropin in circulation? 2. 1. Please include the trade names and catalogue numbers for the immunoassays for the purpose of comparisons to other similar studies in the literature. 2. Yes, it refers to endogenous gonadotropin circulation. As woman age, they have fewer oocytes (primordial follicles) remaining and they have fewer antral follicles. Antral follicle counts are a good predictor of mature follicles that can be stimulated in the ovary when stimulation medication is administered. Counting antral follicles using an ultrasound can be a subjective process. An ideal antral follicle count depends on the age of the woman; older women are not expected to have the same antral follicle count as reported by Anon. (2016). The trade names of the compounds/drugs procured are listed in manuscript in Materials and Methods. However, for clarity the catalogue numbers are as follows: • B84493 Access Sensitive Estradiol Reagent, 100 Determinations, 2 x 50 tests. • B13127 Access AMH (Anti-Mullerian Hormone) Reagent, 100 Determinations, 2 x 50 Tests. • 33520 Access hFSH Reagent, 100 Determinations, 2 x 50 tests 2 1. Overall, the number of patients to be recruited for present study is approved by Statistian to meet the criteria of study and to meet the regulations of the University which is approved by Institutional Ethics committee in local setting. This is ongoing project; we take this feedback seriously and increase the patient’s number in all future studies. 2. Yes, we agree with reviewer and amended the manuscript likewise to Pearson correlation to avoid the confusion (L 116) with the readers. 1. Yes, we agree with reviewers’ comment and completed necessary amendments as “not statistically significant” in the text (L 178)-red font 2. Yes, we noted the error and made amendments in the text as now it reads as “statistically not significant at 5%” (Line 186-187). A p-value <0.05 is regarded as significant if the level of significance is taken as 5%; however, a p-value of 0.065 would be <0.10 (i.e. at a 10% level. Experimental design le size of 42 analysable patients is very small for a study with correlative analysis and may be wered. arify what is meant by “non-parametric Pearson correlation”. Typically, Pearson correlation is c and Spearman-rank correlation is non-parametric. Validity of the findings 1. Lines 167-174 if the p-value of the correlative analysis is greater than 0.05, then it indicates that there is no evidence of an association and that the slope of the line has arisen through random variation within the data. It is not appropriate to refer to these results as “weak correlations” or “weak negative correlations”. 2. On lines 120-121 the level of significance is indicated to be p < 0.05 but on line 181 it refers to a p-value of 0.065 as significant at the 10% level. This lack of consistency should be addressed. Following previous studies support our views strongly to our present study. References 1. Anon. 2016. Antral Follicle Counts, Resting follicle and Ovarian Reserve. Available: http://www.advancedfertility.com/antralfollicles.htm (access 16/09/16) 2. Broer S.L, Mol B, Dollerman M, Frauser B.C and Broekmans F.J.M. 2012. “the role of anti – Mullerian hormone assessment in assisted reproductive technology outcomes. Current Opinion in Obstetrics and Gynaecology, 22:3 193-201 3. Faddy MJ, Gosden RG, Gougeon A, Richardson SJ and Nelson JF. 1992. Acceleration disappearance of ovarian follicles in mid-life: implications for forecasting menopause. Human Reproduction, 7:1342-1346 4. Fanchin, R., Schonäuer, L.M., Righini, C., Frydman, N., Frydman, R. and Taieb, J., 2003. Serum anti‐Müllerian hormone dynamics during controlled ovarian hyperstimulation.Human reproduction, 18(2), pp.328-332. 5. Fanchin, R., Schonäuer, L.M., Righini, C., Guibourdenche, J., Frydman, R. and Taieb, J., 2003. Serum anti‐Müllerian hormone is more strongly related to ovarian follicular status than serum inhibin B, estradiol, FSH and LH on day 3. Human Reproduction, 18(2), pp.323-327. 3 6. Gruijters, M.J., Visser, J.A., Durlinger, A.L. and Themmen, A.P., 2003. Anti-Müllerian hormone and its role in ovarian function.Molecular and cellular endocrinology, 211(1), pp.85-90. 7. La Marca, A., Malmusi, S., Giulini, S., Tamaro, L.F., Orvieto, R., Levratti, P. and Volpe, A., 2004. Anti-Müllerian hormone plasma levels in spontaneous menstrual cycle and during treatment with FSH to induce ovulation. Human Reproduction, 19(12), pp.2738-2741. 8. Menken J, Trussell J and Larsen U. 1986.Age and Inferility.Science, 233: 1389-1394 9. Muttukrishna, S., McGarrigle, H., Wakim, R., Khadum, I., Ranieri, D.M. and Serhal, P., 2005. Antral follicle count, anti‐mullerian hormone and inhibin B: predictors of ovarian response in assisted reproductive technology?.BJOG: An International Journal of Obstetrics & Gynaecology, 112(10), pp.1384-1390. 10. Van Rooij IAJ, Broekmans FJM, Te Velder ER, Frauser BCJM, Bancsi LFJMM, De Jong FH et al. 2002. Serum Anti-Müllerian Hormone level: a novel measure of ovarian reserve. Human Reproduction, 17:3065-3071. 11. Visser, J.A., Durlinger, A.L., Peters, I.J., van den Heuvel, E.R., Rose, U.M., Kramer, P., de Jong, F.H. and Themmen, A.P., 2007.Increased oocyte degeneration and follicular atresia during the estrous cycle in Anti-Müllerian Hormone null mice. Endocrinology, 148(5), pp.2301-2308. -------------------------------------------------XXXX------------------------------------------------------- 4 "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>A variety of predictors are available for ovarian stimulation cycles in assisted reproductive technology (ART) forecasting ovarian response and reproductive outcome in women including biomarkers such as anti-Mullerian hormone (AMH). The aim of our present study was to compare the relationship between AMH levels and pregnancy outcomes in patients undergoing intra-cytoplasmic sperm injection (ICSI). Overall, fifty patients (n=50), aged 20-45 years were recruited for the present prospective study. Three AMH levels were presented with high often poly cystic ovarian syndrome (PCOS) amongst 52.4% patients, 40.5% in normal and 7.1% in low to normal, correspondingly. There was statistically significant relationship between AMH and day of embryo transfer (p&lt; 0.05). The Pearson analysis between AMH, age, E2 and FSH displayed no statistically significant relationship between E2 and AMH (p &lt; 0.05) and negative correlation between FSH and age (p &gt; 0.05).</ns0:p><ns0:p>The area under the receiver operating characteristic curve for E 2 was 0.725 and for AMH levels as predictors of CPR was 0.497 indicating E 2 as better predictor than AMH. The number of oocytes, mature oocytes and fertilized oocytes all presented a weak positive relationship to AMH. Our results confirm the clinical significance of AMH to accurately predict ovarian reserve as a marker and its limitations to use as predictor for a positive pregnancy outcome. Additional prospective studies should be conducted to validate the predictive capability of AMH levels for the outcome of clinical pregnancy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Couples in modern societies postpone childbearing amidst busy schedules and career positionings;</ns0:p><ns0:p>trying to conceive at a more advanced age contributing to a rise in the occurrence of infertility <ns0:ref type='bibr' target='#b3'>(Caroppo et al., 2006)</ns0:ref>. Most women are unaware that fertility starts to decline after the early thirties in some individuals. With increasing female age, fecundity in natural and stimulated ovarian cycles declines, as observed in population-based studies <ns0:ref type='bibr' target='#b10'>(Grynnerup et al., 2012)</ns0:ref> as well as in IVF studies <ns0:ref type='bibr' target='#b22'>(Scheffer et al., 2018)</ns0:ref>. For this reason, there is a growing number of women of advanced age seeking treatment for infertility <ns0:ref type='bibr' target='#b16'>(Oskayli et al., 2019)</ns0:ref>.</ns0:p><ns0:p>The hormonal control of ovarian function is influenced by administering exogenous follicle stimulating hormone (FSH) <ns0:ref type='bibr' target='#b19'>(Richards, 2018)</ns0:ref>. Prediction of ovarian responses prior to stimulation is not only useful for patient counselling, but also important in tailoring the optimal dosage of gonadotrophin for individual patients. The recruitment and development of multiple follicles in response to gonadotrophin stimulation are essential for the successful treatment of infertility by assisted reproductive techniques (ART) <ns0:ref type='bibr' target='#b5'>(Dewailly et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b28'>Yang, Wu, and Zhang 2020)</ns0:ref>.</ns0:p><ns0:p>Besides, poor ovarian response has been suggested to be associated with high cycle cancellation rates <ns0:ref type='bibr' target='#b20'>(Saldeen, K&#228;llen, and Sundstr&#246;m 2007)</ns0:ref>. <ns0:ref type='bibr' target='#b2'>Chang et al (1998)</ns0:ref> found that patients with antral follicle number &#8804; 3 had a significantly higher rate of cycle cancellation and higher human menopausal gonadotropin (HMG) dosage as compared with those patients with antral follicle number 4-10 or &#8805; 10. Nevertheless, the AFC is presently believed to be the finest specific predictor of ovarian response to stimulation in ART, and it can be used in clinical practice for pretreatment counselling targets.</ns0:p><ns0:p>In assisted reproduction, serum levels for several hormones are used to assess the ovarian reserve and to monitor the development of the follicles that have been stimulated by gonadotrophins <ns0:ref type='bibr' target='#b0'>(Alson et al., 2018)</ns0:ref>. Traditional techniques used to predict ovarian stimulation have incorporated serum levels of hormones such as FSH, LH and estrogen (E 2 ) along with ultrasonographic guides such as ovarian volume and number of early antral follicles as a reliable predictor of the outcome of in vitro fertilization (IVF) <ns0:ref type='bibr' target='#b13'>(Kunt et al., 2011)</ns0:ref>. Over the last years, Anti-Mullerian hormone (AMH) has been projected as a novel marker for predicting ovarian response to gonadotrophin stimulation <ns0:ref type='bibr' target='#b0'>(Alson et al., 2018</ns0:ref><ns0:ref type='bibr' target='#b30'>, Zargar et al., 2018)</ns0:ref>. AMH is a dimeric glycoprotein strongly produced by the granulosa cells of the pre-antral (primary and secondary) and small antral follicles (AF's) in the ovary and shown to be age dependent <ns0:ref type='bibr' target='#b21'>(Sahmay et al., 2014)</ns0:ref>. Measurement of anti-M&#252;llerian hormone in serum is much more precise measure of the ovarian reserve than the other hormones that have previously been available to us <ns0:ref type='bibr' target='#b1'>(Anderson et al., 2012)</ns0:ref>.</ns0:p><ns0:p>The aim of this prospective study was to investigate the relationship between Anti-M&#252;llerian hormone levels and pregnancy outcomes in patients undergoing in-vitro fertilization or intracytoplasmic sperm injection (ICSI). </ns0:p></ns0:div> <ns0:div><ns0:head>Materials and Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Patients</ns0:head></ns0:div> <ns0:div><ns0:head>GnRH antagonist Protocol:</ns0:head><ns0:p>A gonadotrophin-releasing hormone (GnRH) antagonist protocol with recombinant FSH (GONAL-f, Merck Serono, Darmstadt, Germany) was used as downregulatory <ns0:ref type='bibr' target='#b18'>(Park et al., 2015)</ns0:ref>. The second approach was followed by administering 0.25 mg/day Cetrotide (Merck Serono). When at least 3 or more follicles reach a diameter equal or above 17 -18 mm, the endometrial thickness reached at least 7 mm by ultrasound and E2 levels were about 1500-1800 pmol/L then Human chorionic gonadotropin (hCG) was administered. All patients received 5000 -10 000 IU hCG (Ovitrelle&#174;, Merck Serono). Oocyte retrieval was performed 36 hours after the administration of the hCG. Conventional IVF or ICSI was performed according to previously described protocols.</ns0:p><ns0:p>Sample collection: Blood samples were collected every 3-4 days on commencement of the treatment. The blood samples were centrifuged at 3000 rpm for 10 minutes using a Biofuge centrifuge (Biofuge Primo -Heraeus) to obtain the blood serum. AMH and FSH levels were recorded, upon the first visit. Estrogen(E 2 ) and LH levels were monitored throughout the program until a peak E 2 and LH level were reached.</ns0:p><ns0:p>Hormone Assays: Gen II ELISA (Beckman Coulter Inc., USA, catalog number A79765/A79766, unmodified version). (Beckman Coulter, USA) kit was used to estimate hormone levels (FSH-Cat.</ns0:p><ns0:p>No. 33520 Access hFSH reagent, 100 determinations, 2 x 50 tests); E2 (Cat. No. B84493 Access Sensitive Estradiol Reagent, 100 determinations) and AMH (Cat. No. B13127 Access AMH Reagent, 100 determinations, 2 x 50 tests) from the blood serum according to manufacturer's instructions. Insemination and intra cytoplasmic sperm injection (ICSI), oocyte retrieval, culture, fertilization, embryo culture, and transfer were carried out as previously described by Gardener et al. <ns0:ref type='bibr' target='#b8'>(Gardner et al., 2001)</ns0:ref>. Inclusion Criteria: The population of the study included female patients ranging between the ages of 20-45 Exclusion Criteria: Patients undergoing cancer therapy and patients on immune suppressant drugs were excluded from study.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis:</ns0:head><ns0:p>The data were analysed using IBM SPSS software (Chicago, IL, USA).</ns0:p><ns0:p>Pearson's correlation was used to determine the direction, strength, and significance of the correlation between X and Y variables between the different semen parameters. A parametric multiple linear regression analysis was used to evaluate the relationship between AMH and other available endocrine markers. ROC curves were used to assess predictive value for E 2 and AMH and evaluating cut off values to optimise sensitivity and specificity. A p value of &lt; 0.05 was considered statistically significant. </ns0:p></ns0:div> <ns0:div><ns0:head>Institutional Review</ns0:head></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>The prospective study included fifty patients who met the inclusion criteria. From the initial sample size of fifty, forty-two presented with data that could be analysed whilst 8 patients had oocytes that where abnormal and did not result in transfer. The data from these 8 patients were not included in the study due to poor embryo development (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Table 1. AMH distribution in blood samples</ns0:head><ns0:p>Amongst the 42 patients analysed, 4.76% were between 20-24 years, 9.52% were between 25-29 years, 40.47% were between 30-34 years, 35.7% were between 35-39 years and 9.52% were between 40-44 years, respectively. As demonstrated by this study the clinical pregnancy rate for patients 20 -24 years was 100%, 25 -29 years was 50%, 30 -34 years was 17.6%, 35 -39 years was 26.6% and 40 -44 years was 25% (Fig. <ns0:ref type='figure'>1</ns0:ref>).</ns0:p><ns0:p>Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref> shows number of oocytes retrieved, number of oocytes matured, and number of oocytes fertilized into respective categories. Not all eggs obtained were at the metaphase 2 stages and had to be matured in the incubator overnight and injected the following day. The results shown were to some extent anticipated as AMH has been used an indicator of oocyte reserve in previous studies <ns0:ref type='bibr' target='#b29'>(Yarde et al., 2013</ns0:ref><ns0:ref type='bibr' target='#b27'>, Yao et al., 2015)</ns0:ref> whereas the resulting fertilized or transferred embryo's may be due to a chance process based on many various factors such as quality of oocyte and sperm.</ns0:p><ns0:p>The Chi-square test for Independence was performed to check whether there was an association between the number of oocytes fertilized and the AMH category (Table <ns0:ref type='table' target='#tab_2'>2</ns0:ref>). A Chi-squared value of 18.5, degrees of freedom = 12, with a p = 0.10 was found. There was no statistically significant relationship between numbers of oocytes fertilized versus AMH category (p &gt; 0.05).</ns0:p><ns0:p>Out of 22 patients, 43 embryos were transferred. Embryos were transferred depending on embryo development and the number of embryos obtained. Most patients in the high and normal categories resulted in a day 5 transfer, the Chi-squared test for independence of AMH and number of embryos transferred gave a Chi-squared value of 6.384 with df = 4 and p-value = 0.172(Table <ns0:ref type='table'>3</ns0:ref>), thus statistically no significant association between AMH and number of embryos transferred was observed. Whilst, Chi-square test for independence between the variables AMH and day of embryo transfer (Table <ns0:ref type='table' target='#tab_3'>4</ns0:ref>) gave a Chi-square value of 14.117, 6 degrees of freedom and p = 0.028 indicating statistically significant relationship between AMH and day of embryo transfer (p&lt; 0.05). Manuscript to be reviewed (Normal); while out of the 3 cases where the AMH was 'Low to Normal', there were no pregnancies reported. The Chi-squared test for independence of AMH category and pregnancy outcome gave a Chi-Squared value of 0.502, 2 degrees of freedom and p = 0.778. There was no statistically significant relationship between the pregnancy outcome and the AMH category (p &gt; 0.05).</ns0:p></ns0:div> <ns0:div><ns0:head>Pregnancy outcome and AMH category are as shown in</ns0:head></ns0:div> <ns0:div><ns0:head>Pearson correlation</ns0:head><ns0:p>Pearson correlation coefficients were calculated to determine if any statistical significance exists between AMH on a quantitative scale and age, E 2 and FSH (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>). The Pearson Correlation coefficient of 0.151 indicates that a very weak positive relationship existence between E 2 and AMH, which is not statistically significant (p = 0.341). Furthermore, Pearson correlation coefficient between the AMH and age had a coefficient of -0.028 thus showing no statistical significance p = 0.859 (p &gt; 0.05). The Pearson Correlation between AMH and FSH produced a coefficient of -0.185 thus indicating no statistical significance p = 0.240 (p &gt; 0.05). Pearson correlation coefficient showed no significant association between AMH and number of oocytes (p = 0.191), number of mature oocytes (p = 0.300) and number of oocytes fertilized (p = 0.146). The number of oocytes, mature oocytes and oocytes fertilized all presented a no statistically significant correlation with AMH (0.206, 0.164, and 0.228, respectively).</ns0:p></ns0:div> <ns0:div><ns0:head>Logistic regression analysis</ns0:head><ns0:p>A logistic regression model was used to determine the possible predictor variables for the pregnancy outcome. The model was fitted to the data with the result of the pregnancy namely, 'Positive' or 'Negative' as the binary dependent variable and age, E 2 LH, Basal FSH, Basal AMH and number of oocytes fertilized as independent variables (Table <ns0:ref type='table'>7</ns0:ref>). As shown in Table <ns0:ref type='table'>7</ns0:ref>, LH has p = 0.042 (p &lt; 0.05) and E 2 has p = 0.065, which is statistically not significant at a 5% level. The SPSS output for the model is given in Table <ns0:ref type='table'>1</ns0:ref>(Supplementary data S1) signifying that overall, 73.8 % of the cases were correctly classified, while 5/12 =0.417 or 41.7 % of the positives were correctly classified, and 86.7% of the negative cases were correctly classified.</ns0:p></ns0:div> <ns0:div><ns0:head>Area under the curve</ns0:head><ns0:p>The ROC curves of the serum AMH concentrations and E 2 for the prediction of the clinical pregnancy are depicted in Fig. <ns0:ref type='figure'>2</ns0:ref>. The areas under the curves (AUC) for E 2 were 0.725 and for AMH (AUC = 0.497). E 2 is therefore a better single predictor of pregnancy outcome when compared to AMH. It has been shown that E 2 can better predict the number of oocytes obtained.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In the current study, we investigated the relationship between AMH levels and pregnancy outcomes in patients undergoing intra-cytoplasmic sperm injection.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between basal AMH and E 2</ns0:head><ns0:p>Pearson analysis between E 2 and AMH presented a Pearson Correlation co-efficient of 0.151 with p = 0.341 (p &lt; 0.05) which indicates that a weak significant relationship exists between E 2 and AMH. Most previous studies (Ramalho <ns0:ref type='bibr' target='#b18'>de Carvalho et al., 2012</ns0:ref><ns0:ref type='bibr' target='#b23'>, Ubaldi et al., 2005)</ns0:ref> have shown a relationship between a raised basal E 2 level and a reduced ovarian response using different values to express elevated estrogen levels which replicated the findings in this study therefore showing that a low AMH can result in low estrogen levels. Also, it can be concluded that a poor AMH value results in a poor ovarian reserve indicating follicles produced will not be correlated to a raised estrogen level, therefore indicating poor follicle growth, thus reducing the number of oocytes produced. However, it was determined that meagre response to stimulus in IVF, indicative of a lower ovarian reserve, is associated with declined baseline serum AMH concentrations <ns0:ref type='bibr' target='#b26'>(Van Rooij et al., 2004)</ns0:ref>. Consequently, when women have regular ovarian reserve and decent retort, disappointment of IVF must look for additional infertility reasons, e.g., male specific issue i.e. Y chromosome microdeletion. Furthermore, this conclusion is reinforced by the data of woman undergoing IVF which indicated that male factor infertility resulted in an unsuccessful cycle.</ns0:p><ns0:p>Although E2 levels in these cases were above those of controls, they are still within the range of 25-100 pg/ml <ns0:ref type='bibr' target='#b21'>(Sahmay et al., 2014)</ns0:ref>, suggesting that E2 single-handedly is not capable of predicting the female reproductive potential.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between basal AMH and Age</ns0:head><ns0:p>Pearson correlation between AMH and age (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) presented a co-efficient of -0.028 thus displaying a weak, negative association with a p = 0.859 (p&gt; 0.05). A stronger relationship between these two variables was expected as it is known that as age increases, AMH should decrease. This contrary association corresponds as reported by <ns0:ref type='bibr' target='#b26'>Van Rooij et al., (2004)</ns0:ref>, where serum AMH levels decline with age in normal women with proven fertility. Also, it is suggested that serum AMH is identified as the improved endocrine marker to measure the reproductive capability in advanced age.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between Basal AMH and FSH</ns0:head><ns0:p>Basal FSH is one of the primary endocrine markers presented into ART program. The Pearson correlation amid AMH and FSH (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) had a coefficient of -0.185 thus displaying a weak, negative relationship and with a p= 0.240 (p&gt; 0.05). This study specifies a negative correlation, i.e., the higher the FSH the higher the chances the patient can present with a poor ovarian reserve and early menopause. It is routine practice to frequently measure the basal FSH level and to start IVF treatment only when the FSH level is lower than threshold value in a cycle. This was strategic on the awareness that these women will retort well to ovarian stimulation while the basal FSH level is lesser at the beginning of the cycle. The outcomes of this study revealed that woman who were poor respondents or had a reduced ovarian reserve had a poor outcome and frequent testing is of no worth. Women who had a history of high FSH level must undergo treatment without further delay. By postponing treatment for these patients can be detrimental as they get older and fast approaching menopause <ns0:ref type='bibr' target='#b24'>(Uzumcu and Zama, 2016)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between Age and FSH and Number of oocytes</ns0:head><ns0:p>The Pearson Correlation between FSH and age (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) displayed no statistical significance, p = 0.583 (p &gt; 0.05). For most of Pearson Correlation analysis, no significant relationships were found with most of the p-values, being greater than 0.05. This may be due to the small sample size used in this study of 42 patients. AMH being compared to age and number of oocytes showed a slightly negative correlation which is expected as it is shown in previous studies that AMH and number of oocytes decrease with maternal age <ns0:ref type='bibr' target='#b26'>(van Rooij et al., 2004)</ns0:ref> and <ns0:ref type='bibr'>Gobikrushant et al. (2018)</ns0:ref>. This inverse relationship is in agreement by <ns0:ref type='bibr' target='#b26'>Van Rooij et al. (2004)</ns0:ref>, who reported that serum AMH levels deteriorate with age in normal women with proven fertility. Additionally, serum AMH indicates the simplest endocrine marker to measure the age-related decline of reproductive competence. AMH levels, in our group who were high respondents were over 3.0 ng/ml, normal respondent over 1.0 ng/ml and low respondents found to be below 0.9 ng/ml. Oocytes were still recovered even with low AMH levels. Neither fertilization rate nor embryo quality can be assessed using basal AMH levels. This contrasts with the findings reported by <ns0:ref type='bibr' target='#b25'>Vaegter et al. (2017)</ns0:ref>, where</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed embryos had superior morphology and cleavage performance in patients with AMH levels &gt; 2.7 ng/ml as compared with patients with values below this threshold.</ns0:p></ns0:div> <ns0:div><ns0:head>Correlation between AMH and the number of oocytes, number of mature oocytes and number of oocytes fertilized</ns0:head><ns0:p>Our second objective of the study was to examine if AMH levels affected oocyte quality. In this study, the Pearson Correlation test (Table <ns0:ref type='table' target='#tab_6'>6</ns0:ref>) showed no significant relationship between AMH and number of oocytes (p = 0.191), several mature oocytes (p = 0.300) and number of oocytes fertilized (p = 0.146). The number of oocytes, mature oocytes and oocytes fertilized all showed a weak positive relationship to AMH (0.206, 0.164, and 0.228, respectively). These findings are in agreement with that reported by La Marca and Sunkara (2014); La <ns0:ref type='bibr' target='#b14'>Marca et al., (2010)</ns0:ref> and <ns0:ref type='bibr'>Dehghani-Firouzabadi et al. (2008)</ns0:ref>, where mean amount of oocytes was lower in poor responding patients than in normal patients attending IVF programs. This therefore led to the inference that ovarian response can be regarded as a reflection of the ovarian reserve. The Chi-square test for Independence was done to determine whether there is an association between the number of oocytes collected and the AMH category (Table <ns0:ref type='table'>8</ns0:ref>). A Chi-squared value of 21.246, degrees of freedom = 8, with a p = 0.007 was observed. There was a significant relationship between the numbers of oocytes collected versus AMH category (p &lt; 0.05). The Chi-square test for Independence was performed to see whether there is an association between the number of oocytes fertilized and the AMH category (Table <ns0:ref type='table'>9</ns0:ref>). A Chi-squared value of 18.5, degrees of freedom = 12, with a p = 0.10 was found. There was thus no statistically significant relationship between the numbers of embryo's fertilized versus AMH category (p &gt; 0.05).This is anticipated as AMH has been used an indicator of oocyte reserve in previous studies whereas the resulting fertilized or transferred embryo's may be due to a chance process based on many various factors such as the quality of the oocyte and sperm <ns0:ref type='bibr' target='#b29'>(Yarde et al., 2013)</ns0:ref>. <ns0:ref type='bibr' target='#b7'>Ebner et al. (2006)</ns0:ref>, demonstrated that AMH serum levels were related with oocyte quality in stimulated cycles. The quality of the embryos was not assessed using baseline AMH which agrees with our findings. However, the fertilization rate was not correlated with the serum AMH which varied with the results of the present study.</ns0:p></ns0:div> <ns0:div><ns0:head>AMH category and positive pregnancies</ns0:head><ns0:p>Embryo quality has been suggested to be of paramount importance to predict the occurrence of pregnancy after IVF. In a regression model E 2 has a p =0.017 (p &lt;0.05) and LH has a p =0.035</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed (p &lt;0.05). Both variables are significant and age and basal AMH play a role in the pregnancy outcome and the model is thus adjusted for these two variables</ns0:p><ns0:p>In this study, AMH value for predicting pregnancy outcomes does not exist because oocyte quality is not accounted for by ovarian reserve markers. As demonstrated in this study the clinical pregnancy rate for patients 20 -24 years was 100%, 25 -29 was 50%, 30 -34 years was 18%, 35</ns0:p><ns0:p>-39 years was 27% and 40 -44 years was 25% (Figure <ns0:ref type='figure'>1</ns0:ref>). Patients presenting with a low AMH did not vary from those women presenting with higher AMH concentrations in same age group. A positive pregnancy outcome was logged across all age groups regardless of the AMH level. These results advocate that low ovarian reserve is not correlated with low oocyte quality in patients and the prediction remains the similar despite low AMH concentrations. <ns0:ref type='bibr' target='#b11'>Kini et al. (2010)</ns0:ref> stated the role of AMH in foreseeing cumulative pregnancy outcome during IVF treatment. It was recognized that serum AMH concentration on day 6 of stimulation was suggestively higher in participants who resulted in an ongoing pregnancy in IVF compared to those who did not. Serum AMH is a suitable indicator of ovarian hyper-response. In a metanalysis study conducted by <ns0:ref type='bibr' target='#b27'>Yao et al. (2015)</ns0:ref> to evaluate role of serum AMH role in forecasting the pregnancy outcome in IVF/ICS, it was concluded that there is positive correlation between serum AMH and pregnancy. Nevertheless, association between serum AMH and non-pregnancy cannot be ruled out either.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In conclusion, the outcomes of our investigations specify that AMH has established to be a valuable marker for ovarian reserve and might benefit woman who plan for pregnancy. AMH hormone seems to be the best endocrine marker, however, the valuable role of AMH and its role in ovarian function should be looked at in relation to the other markers to assess the decline of the ovarian pool. on the implication of varying levels of AMH within the follicular fluid may be representative as an indicator of 'quality' in addition to the number of growing follicles.</ns0:p></ns0:div> <ns0:div><ns0:head>Study Limitation</ns0:head><ns0:p>A noteworthy restraint of the current study was the lack of antral follicle count (AFC) at time of oocyte collection.</ns0:p></ns0:div> <ns0:div><ns0:head>Conflict of Interest</ns0:head><ns0:p>Authors declare no conflict of interest</ns0:p></ns0:div> <ns0:div><ns0:head>Authors Contribution</ns0:head><ns0:p>[SU] and Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>:</ns0:head><ns0:label /><ns0:figDesc>Fifty women(n=50), aged 20-45 years were recruited from Centre of Assisted Reproduction and Endocrinology (C.A.R.E) Clinic in Westville, Durban, South Africa who were undergoing IVF treatment. This study was approved by Ethical Committee of the Durban University of Technology (Project reference 128/16) and Research Committee, C.A.R.E. Clinic, Durban, South Africa. After approving the study by the research ethics committees, written informed consents were obtained from all the patients.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Board approval: This study was approved by Ethical Committee of the Durban University of Technology (Project reference 128/16) and Research Committee, C.A.R.E. Clinic, Durban, South Africa and was performed in accordance with the Helsinki Declaration of 1975 (as revised in 1983).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>[JKA] contributed to the study conception and design. [SU] performed all experiments, [SU], [KSBN] and [JKA] performed material preparation, data collection and analysis. The first draft of the manuscript was written by [KSBN] and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='18,42.52,188.32,525.00,294.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Out of twenty two cases in high category, 6 resulted in a positive pregnancy; 6 resulted in a positive outcome (6/12 = 50.0%)</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>While appropriate reference values are being generated per age category and until the consequences of having a low or high AMH for one's age are being established, AMH should only be determined in the context of clinical studies. At present, the most important clinical role of AMH at this stage is to serve as a red flag for reduced ovarian reserve in women of reproductive age who must undergo further diagnostics. As per the study conducted, we can infer that AMH can accurately predict ovarian reserve but cannot predict the oocyte quality or a positive pregnancy</ns0:figDesc><ns0:table /><ns0:note>outcome. The more oocytes obtained, increases a patient's chance of more viable embryos and therefore, improving chances of a healthy pregnancy and ultimately a live birth. Further research</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Correlation between AMH and number of oocytes collected, matured, and fertilised during stimulation</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>Total no.</ns0:cell><ns0:cell>Total no.</ns0:cell><ns0:cell>Total no.</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>of oocytes collected</ns0:cell><ns0:cell>of oocytes matured</ns0:cell><ns0:cell>of oocytes fertilized</ns0:cell><ns0:cell>% collected oocytes</ns0:cell><ns0:cell>% matured oocytes</ns0:cell><ns0:cell>% oocytes fertilized</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>High</ns0:cell><ns0:cell>81</ns0:cell><ns0:cell>62</ns0:cell><ns0:cell>69</ns0:cell><ns0:cell>65.4%</ns0:cell><ns0:cell>60%</ns0:cell><ns0:cell>61.6%</ns0:cell></ns0:row><ns0:row><ns0:cell>AMH</ns0:cell><ns0:cell cols='2'>Normal 38</ns0:cell><ns0:cell>36</ns0:cell><ns0:cell>38</ns0:cell><ns0:cell>30.6%</ns0:cell><ns0:cell>35%</ns0:cell><ns0:cell>33.9%</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>category Low to</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>5</ns0:cell><ns0:cell>4.0%</ns0:cell><ns0:cell>5%</ns0:cell><ns0:cell>4.5%</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Normal</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Total</ns0:cell><ns0:cell /><ns0:cell>124</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 4 .</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Chi Square analysis of AMH and day of embryo transfer. . 10 cells (83.3%) have expected count less than 5. The minimum expected count is 0.05</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell>Value</ns0:cell><ns0:cell>Degrees of Freedom</ns0:cell><ns0:cell>Asymptotic</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>(df)</ns0:cell><ns0:cell>Significance (2-sided)</ns0:cell></ns0:row><ns0:row><ns0:cell>Pearson Chi Square</ns0:cell><ns0:cell>14.117 a</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0.028</ns0:cell></ns0:row><ns0:row><ns0:cell>Likelihood Ratio</ns0:cell><ns0:cell>6.432</ns0:cell><ns0:cell>6</ns0:cell><ns0:cell>0.377</ns0:cell></ns0:row><ns0:row><ns0:cell>N of Valid Cases</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>aPeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 5 (on next page)</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Pregnancy outcome and AMH category</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 5 .</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Pregnancy outcome and AMH category</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Pregnancy Result</ns0:cell></ns0:row></ns0:table><ns0:note>PeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head>Table 6 (on next page)</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Pearson correlation between basal AMH and E 2, Age and FSH and oocytes</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:08:52172:2:0:NEW 22 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Pearson correlation between basal AMH and E 2, Age and FSH and oocytes</ns0:figDesc><ns0:table><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>E 2</ns0:cell><ns0:cell>Age</ns0:cell><ns0:cell>FSH</ns0:cell><ns0:cell>Number</ns0:cell><ns0:cell>Number of</ns0:cell><ns0:cell>Number of oocytes</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>of</ns0:cell><ns0:cell>mature</ns0:cell><ns0:cell>fertilized</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>oocytes</ns0:cell><ns0:cell>oocytes</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>Pearson</ns0:cell><ns0:cell cols='2'>0.151 -0.028</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>0.206</ns0:cell><ns0:cell>0.164</ns0:cell><ns0:cell>0.228</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>correlation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.185</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>coefficient</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>AMH</ns0:cell><ns0:cell>Significance</ns0:cell><ns0:cell cols='3'>0.341 0.859 0.240</ns0:cell><ns0:cell>0.191</ns0:cell><ns0:cell>0.300</ns0:cell><ns0:cell>0.146</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>value (2-tailed)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>No. in the</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>sample</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Age</ns0:cell><ns0:cell>Pearson</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>-</ns0:cell><ns0:cell>-0.271</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Correlation</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.087</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>Significance</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.583</ns0:cell><ns0:cell>0.082</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>value (2-tailed)</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>No. in the</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>42</ns0:cell><ns0:cell>42</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell>sample</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='103'>112100% 100%</ns0:note> </ns0:body> "

" 22-10-2020 To Wei Cui Academic Editor PeerJ Re: Revised manuscript ID- 52172:0:2:REVIEW Dear Editor We appreciate the positive valuable comments made by reviewers in strengthening our manuscript and attended all points raised by reviewers (attached rebuttal letter). We also wish to inform you that we are open to necessary amendments in support of our manuscript. Highly appreciate your support in this regard. Regards Dr Suresh Naidu Corresponding author Durban University of Technology Durban, South Africa The Relationship Between Anti-Mullerian Hormone (AMH) Levels and Pregnancy Outcomes in Patients Undergoing Assisted Reproductive Techniques (ART) Article No: 52172 Basic reporting 1.Reviewer’s response: Clarity is still needed in this paragraph, as no changes have been made. The intent of the authors is not clear and it needs to be redrafted. The authors response mentions antral and mature follicle counts which do not feature in the paragraph. It is not clear whether the authors are referring to endogenous gonadotropin or gonadotropin administered during stimulation cycles Authors response: Towards this we merged the latest literature available from lines 64-76 in text The hormonal control of ovarian function is influenced by administering exogenous follicle stimulating hormone (FSH) (Richards, 2018). Prediction of ovarian responses prior to stimulation is not only useful for patient counselling, but also important in tailoring the optimal dosage of gonadotrophin for individual patients. The recruitment and development of multiple follicles in response to gonadotrophin stimulation are essential for the successful treatment of infertility by assisted reproductive techniques (ART) (Dewailly et al. 2014; Yang, Wu, and Zhang 2020). Besides, poor ovarian response has been suggested to be associated with high cycle cancellation rates (Saldeen, Källen, and Sundström 2007). Chang et al (1998) (Chang 1998) found that patients with antral follicle number ≤ 3 had a significantly higher rate of cycle cancellation and higher human menopausal gonadotropin (HMG) dosage as compared with those patients with antral follicle number 4-10 or ≥ 10. Nevertheless, the AFC is presently believed to be the finest specific predictor of ovarian response to stimulation in ART, and it can be used in clinical practice for pretreatment counselling targets. The following references are added to the reference section accordingly. Chang, M.Y., Chiang, C.H., T’sang-T’ang Hsieh, M.D., Soong, Y.K. and Hsu, K.H., . 1998. 'Use of the antral follicle count to predict the outcome of assisted reproductive technologies. ', Fertility and Sterility 69: 505-10. Dewailly, Didier, Claus Yding Andersen, Adam Balen, Frank Broekmans, Nafi Dilaver, Renato Fanchin, Georg Griesinger, Tom W Kelsey, Antonio La Marca, and Cornelius Lambalk. 2014. 'The physiology and clinical utility of anti-Müllerian hormone in women', Human Reproduction Update, 20: 370-85. Saldeen, Pia, Karin Källen, and Per Sundström. 2007. 'The probability of successful IVF outcome after poor ovarian response', Acta Obstetricia Et Gynecologica Scandinavica, 86: 457-61. Yang, Peiwen, Ruxing Wu, and Hanwang Zhang. 2020. 'The effect of growth hormone supplementation in poor ovarian responders undergoing IVF or ICSI: a meta-analysis of randomized controlled trials', Reproductive Biology and Endocrinology, 18: 76. 2.Reviewer’s response: These reagent details listed above have not been included in the methods section of the manuscript. The readers of the final article will need to know where the reagents came from if they want to replicate this study or compare it with others. The methods section of the manuscript refers to a manual 96-well plate format AMH Gen II ELISA (referencing Kumar et al. 2010) whereas the regent listed above is for an automated immunoassay. These two immunoassays have different properties. Which one was used? (state this in the manuscript) Authors response: We have amended the methods section (lines 115-120) Hormone assays: Gen II ELISA (Beckman Coulter Inc., USA, catalog number A79765/A79766, unmodified version). (Beckman Coulter, USA) kit was used to estimate hormone levels (FSH-Cat. No. 33520 Access hFSH reagent, 100 determinations, 2 x 50 tests); E2 (Cat. No. B84493 Access Sensitive Estradiol Reagent, 100 determinations) and AMH (Cat. No. B13127 Access AMH Reagent, 100 determinations, 2 x 50 tests) from the blood serum according to manufacturer’s instructions. Experimental design 1.Reviewer’s response: The manuscript methods section still reads “non-parametric Pearson correlation”. Pearson correlation is parametric. The authors have not made the change they claim above. Authors response: amended the correction Line 128 in text Validity of the findings 1.Reviewer’s response: If the p-value is >0.05 then there is no evidence of correlation. The manuscript still refers to these results as “correlations” in the results section. The requested changes have not been made, nor addressed with a rebuttal Authors response: We made necessary corrections in the text by deleting “correlations” (Lines 181-184) and replaced with association respectively as suggested by reviewer. ------------------------------------------------XXX-------------------------------------------------------- "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Secondary fungal metabolites are important sources for new drugs against infectious diseases and cancers. Methods. To obtain a library with enough diversity, we collected about 2395 soil samples and 2324 plant samples from 36 regions in Africa, Asia, and North America. The collection areas covered various climate zones in the world. We examined the usability of the global fungal extract library (GFEL) against parasitic malaria transmission, Gram-positive and negative bacterial pathogens, and leukemia cells.</ns0:p></ns0:div> <ns0:div><ns0:head>Results.</ns0:head><ns0:p>Nearly ten thousand fungal strains were isolated. Sequences of nuclear ribosomal internal transcribed spacer (ITS) from 40 randomly selected strains showed that over 80% were unique. Screening GFEL, we found that the fungal extract from Penicillium thomii was able to block P. falciparum transmission to Anopheles gambiae, and the fungal extract from Tolypocladium album was able to kill myelogenous leukemia cell line K562.</ns0:p><ns0:p>We also identified a set of candidate fungal extracts against bacterial pathogens.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Natural products, produced by living organisms in nature, have been used as medicine for thousands of years <ns0:ref type='bibr'>(Dias, Urban et al. 2012</ns0:ref><ns0:ref type='bibr' target='#b3'>, Buyel 2018)</ns0:ref>. For instance, the treatment of malaria was recorded in China with the Qinghao plant (Artemisia annua) as early as the second century. The active ingredient, qinghaosu (artemisinin), was isolated from the plant by Youyou</ns0:p><ns0:p>Tu and her colleagues in 1971 <ns0:ref type='bibr' target='#b18'>(Luo and</ns0:ref><ns0:ref type='bibr'>Shen 1987, Hsu 2006)</ns0:ref>. The establishment of microbiology in the early modern era led to drug discoveries from microbes. The first antibiotics, penicillin, was discovered from a fungus by Alexander Fleming <ns0:ref type='bibr'>(A 1929)</ns0:ref>. Fungi have initially been and still are used to produce medicines to treat infectious diseases <ns0:ref type='bibr' target='#b5'>(Elder 1944)</ns0:ref>.</ns0:p><ns0:p>Furthermore, people use natural products to treat tumors. Indeed, more than half of anti-tumor drugs or leads in current clinical trials are from natural products <ns0:ref type='bibr' target='#b32'>(Wolfender and Queiroz 2012)</ns0:ref>. Microbial metabolites are essential resources for drug discovery <ns0:ref type='bibr'>(Lenzi, Costa et al. 2018)</ns0:ref>. Compared with other natural product resources, fungi have the following advantages.</ns0:p><ns0:p>First, there are enormous fungal species: about 120,000 fungal species have been described <ns0:ref type='bibr' target='#b10'>(Hawksworth and Lucking 2017)</ns0:ref> and 5.1 million fungal species are estimated <ns0:ref type='bibr' target='#b2'>(Blackwell 2011)</ns0:ref>.</ns0:p><ns0:p>Second, fungi produce broad and diverse secondary metabolites with a vast difference in chemical structures <ns0:ref type='bibr'>(Pham, Yilma et al. 2019)</ns0:ref>. Third, large-scale fermentation can generate a large amount of fungal secondary metabolites, which was exampled by the production of alcohol and lactic acid. However, yield of many target fungal secondary metabolites is restricted by fungal growth and differentiation <ns0:ref type='bibr' target='#b20'>(Nielsen and Nielsen 2017</ns0:ref><ns0:ref type='bibr' target='#b14'>, Keller 2019</ns0:ref><ns0:ref type='bibr'>, Pham, Yilma et al. 2019)</ns0:ref>, which is resolved by new technologies that enable us to engineer a fungus to produce a specific compound in high yield by modifying its metabolic pathways (van Dijk and Wang 2016).</ns0:p><ns0:p>Also, the recent development of genomic sequencing technology and the identification of more biosynthetic gene clusters accelerate the discovery and application of new compounds from fungi <ns0:ref type='bibr'>(Hussain, Al-Sadi et al. 2017</ns0:ref><ns0:ref type='bibr' target='#b14'>, Keller 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Isolation of fungi from soil and plants:</ns0:head><ns0:p>For each soil sample, we transferred 50 mg soil to a 1.5 mL plastic tube, and 1 mL autoclaved distilled water was added. The sample was vortexed for 30 secs and centrifuged at 500 g for 2 minutes (min) to get rid of the soil particles. For plant samples, the first step was to sterilize the plant surface by rinsing the sample with distilled H 2 O, then soaking in 70% ethanol for 10 seconds (sec) and rinsing again with distilled H 2 O. After sterilization, the plants were cut into 0.5 cm x 0.5 cm pieces and transferred into a sterile mortar.</ns0:p><ns0:p>Then, 2 mL of distilled H 2 O was added, followed by grinding with a pestle for 1-3 min and the slurry was transferred to a 1.5 mL plastic tube and then centrifuged at 500 g for 2 min to get rid of the particles. About 100 &#956;L of upper supernatant from treated soil or plants was evenly spread onto a 100 x 15 mm Petri Dish plate containing 14 mL Malt Extract Agar (MEA), made of 10 g malt extract, 1 g yeast extract, 15 g agar, and 0.05 g chloramphenicol (Sigma-Aldrich, St.</ns0:p><ns0:p>Louis, MO) in 1 L distilled H 2 O and autoclaved at 121 &#186;C for 20 min. The plates were sealed with parafilm, and the fungi were allowed to grow for 7-14 days at room temperature (RT) with cycles of 12 hours (hr) of darkness and 12 hr of light.</ns0:p><ns0:p>The fungal colonies on the MEA medium plates were picked with a toothpick and inoculated in a new MEA plate by streaking. If the colonies were mixtures of two or more species, we kept inoculating and streaking until a single colony appeared. Finally, a piece of fungal agar containing mycelium or spores was cut and transferred to a 1.5 mL Eppendorf tube containing 500 &#956;L of sterile 20% glycerol in distilled H 2 O. We stored the cells in a -80 &#176;C freezer for long-term storage.</ns0:p></ns0:div> <ns0:div><ns0:head>Metabolite production and extraction:</ns0:head><ns0:p>Cereal based medium was used to grow fungi to produce secondary metabolites <ns0:ref type='bibr'>(Niu, Wang et al. 2015)</ns0:ref>. Briefly, six pieces of Cheerios Breakfast cereals (General Mills, Minneapolis, MN) were placed to a glass test tube, capped with a plastic lid and autoclaved for 20 min, and then 2 mL sterile sucrose water (3 g of sucrose and 50 mg chloramphenicol in 1 L distilled H 2 O) was added into the tube. Later, the fungal colony grown on the MEA plate was inoculated into the cereal medium and incubated at RT for PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed one month to produce sufficient metabolites. A month late, 2 mL ethyl acetate was added into a tube to extract the fungal metabolites, mixed with a glass stirring rod, and placed overnight in a chemical hood with gentle shaking. The next day, 1 mL upper layer of supernatant was transferred to a 1.5 mL tube. After centrifugation (2, 000 g for 2 min), around 950 &#956;L clear supernatant was transferred to a pre-weighed 1.5 mL plastic tube and dried with SpeedVac concentrator (Thermo Fisher Scientific, Waltham, MA). Finally, the dry extracts were weighed and dissolved in an appropriate amount of dimethyl sulfoxide (DMSO) to prepare 10 mg/mL stock solution and stored in a -20 &#176;C freezer for future screening assays.</ns0:p><ns0:p>Determination of fungal species: Forty fungal isolates were randomly picked to evaluate the fungal library's diversity. The fungi were cultured with liquid malt extract medium at RT for one week, and mycelium was collected for DNA extraction using DNAzol (Thermo Fisher). Genomic DNA applied as PCR templates were isolated using DNAzol Reagent following the manual (Thermo Fisher Scientific). To identify the fungal species, nuclear ribosomal ITS regions were amplified with by PCR with specific primers <ns0:ref type='bibr'>(Schoch, Seifert et al. 2012</ns0:ref>) (Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>) using the following approach: 94 &#176;C 2 min; 94 &#176;C 30 sec, 55 &#176;C 30 sec, 72 &#176;C 1 min, 35 cycles; 72 &#176;C 5 min. The amplified products were sequenced and blasted against the NCBI database to identify fungal species <ns0:ref type='bibr'>(Raja, Miller et al. 2017)</ns0:ref>.</ns0:p><ns0:p>Screening the fungal extract library to identify malaria transmission-blocking candidates with ELISA assays: As described previously <ns0:ref type='bibr'>(Niu, Wang et al. 2015)</ns0:ref> Manuscript to be reviewed washed three times with RPMI-1640 at 300 &#215; g for 4 m. The cell pellets were re-suspended in PBST (PBS containing 0.2% Tween-20) and homogenized by ultra-sonication with six cycles of 10 sec pulse and 50 sec resting on ice for each period. The lysates were centrifuged at 8,000 g for 2 min to remove insoluble materials and cellular debris. With the iRBC lysate and insect cellexpressed recombinant FREP1, the ELISA assay was used to screen the fungal extract library to block FREP1-parasite interaction <ns0:ref type='bibr'>(Niu, Wang et al. 2015)</ns0:ref>. A 96-well ELISA plate was coated with 50 &#956;L iRBC lysate (2 mg/mL protein) overnight at 4&#176;C. After coating, the plate was blocked with 100 &#956;L of PBS plus 0.2% bovine serum albumin (BSA) per well for 1.5 hr at RT. After removal of the blocking solution, FREP1 (10 &#956;g/mL) in blocking buffer (PBS plus 0.2% BSA)</ns0:p><ns0:p>was added to each well, and 1&#956;L fungal extract was taken from a 96-well plate containing 2 mg/mL crude extract dissolved in DMSO in each well with a multiple-channel pipette and transferred to the ELISA plate, then incubated for 1 hr at RT with gentle shaking. After washing three times with PBST, 50 &#956;L rabbit anti-FREP1 polyclonal antibody <ns0:ref type='bibr' target='#b23'>(Niu et al., 2015)</ns0:ref> (diluted 1: 5,000 in blocking buffer, 1 &#956;g/mL) was added to each well and incubated for 1 hr at RT. About 50 &#956;L alkaline phosphatase-conjugated anti-rabbit IgG (diluted 1: 20,000 in blocking buffer) was added to each well and incubated for 45 min at RT. The wells were washed three times with PBST between incubations. After washing, each well was developed with 50 &#956;L pNPP substrate (Sigma-Aldrich) until the colors were visible, and absorbance at 405 nm was measured. The functional FREP1 supplemented with 1 &#956;L solvent (DMSO) was used as non-inhibition control, and the heat-inactivated FREP1 (65 &#176;C for 15 min) was used as a 100% inhibition control.</ns0:p></ns0:div> <ns0:div><ns0:head>Determination of the transmission-blocking activity of the fungal extracts in mosquitoes:</ns0:head><ns0:p>Following the previous protocol <ns0:ref type='bibr'>(Zhang, Niu et al. 2015)</ns0:ref>, the 15-to 17-day old cultured P.</ns0:p><ns0:p>falciparum containing 2-3% stage V gametocytes were collected and diluted with new O+ type human blood to get 0.2% stage V gametocytes in the blood. Then, the 150 &#61549;L blood was mixed with the same volume of heat-inactivated AB+ human serum. Then, 3 &#61549;L candidate fungal</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed extract in DMSO (10 mg/mL or 2 mg/mL) was mixed with 297 &#61549;L infected blood, the final fungal extract concentration in blood was 100 or 20 &#181;g/mL, respectively. SMFA was performed to feed about 100 3-5 days old A. gambiae G3 female mosquitoes for 15 min, and the engorged mosquitoes were maintained with 8% sugar in a BSL-2 insectary (28 &#176;C, 12-h light/dark cycle, 80% humidity). The midguts were dissected seven days post-infection and stained with 0.1% mercury dibromofluorescein disodium salt in PBS for 16 m. The oocysts in midguts were counted under a light microscope. This standard membrane feeding assays were conducted at least twice to confirm the results. were seeded in 96-well microplates and incubated at 37 &#176;C with 5% CO 2 . The next day, one &#181;L fungal extract in DMSO was added (final concentration of the fungal extract was 20 &#181;g/mL), and the cells were incubated at 37 &#176;C with 5% CO 2 for another 24 hr. Next, the microplate was centrifuged at 500 g for 10 min to pellet the cells, the medium was carefully removed as much as possible, and 100 &#181;L of fresh medium was then added. About 10 &#181;L of the 12 mM MTT stock solution was added, mixed, and incubated for 4 hr at 37 &#176;C. The microplate was centrifuged again at 500 g for 10 m. After removing 75 &#181;L of the medium from the wells with 25 &#181;L medium with cells left, 50 &#181;L of DMSO was added to each well, mixed, and incubated at 37 &#176;C for 10 min to dissolve formazan crystal for measurement. The same amount of DMSO without drugs was applied as a control. The optical density was measured at an absorbance wavelength of 540 nm. Cell growth inhibition rate (%) = (A 540 of control -A 540 of treatment/ A 540 of control) &#215; 100%.</ns0:p></ns0:div> <ns0:div><ns0:head>Screening the fungal extract library to identify antibiotics</ns0:head><ns0:p>Triplicates were conducted for each sample.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis:</ns0:head><ns0:p>All the experiments were independently repeated at least twice and analyzed with the Wilcoxon-Mann-Whitney test using GraphPad Prism (GraphPad Software, CA, USA).</ns0:p><ns0:p>Sequence availability: All fungal ITS sequences obtained in this project have been deposited</ns0:p><ns0:p>into GenBank at NCBI (https://www.ncbi.nlm.nih.gov/genbank/sequenceids/).</ns0:p><ns0:p>Construction of phylogenetic tree: ITS sequences were aligned each other using online clustalW2 tool (https://www.ebi.ac.uk/Tools/msa/clustalw2/). A phylogenetic tree construction</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed tool (https://www.ebi.ac.uk/Tools/phylogeny/simple_phylogeny/) took the multiple sequence alignment to build the phylogenetic tree.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head><ns0:p>Extensive fungal library with nearly ten thousand isolates A large and diverse fungal library is powerful in discovering new drugs. To maximize the diversity and quantity in our fungal library, we selected fungal colonies based on their location, color, and morphology on culture plates. Therefore, the fungal isolates in the library look strikingly different (Fig <ns0:ref type='figure' target='#fig_6'>1C</ns0:ref>). This visible criterion facilitates the fungal library construction. However, we also discarded many fungi that are different species with similar morphology. We examined the species diversity of our fungal library at the molecular level by randomly picking 40 fungal isolates. Their genomic DNA was isolated, the ITS regions were amplified, and PCR products were sequenced. These sequences have been deposited into GenBank at https://www.ncbi.nlm.nih.gov/genbank/sequenceids/. Their Accession numbers are MT594355-MT594393 and MT584204. We searched these sequences against NCBI DNA databases using blast. Results show that about 12.5% of total fungal isolates have identical ITS sequences to others in the library (Table <ns0:ref type='table'>3</ns0:ref>). For instance, 126-G10, 117-B9, and 45-F10 have identical ITS to Fusarium solani, and 2 of 3 might be duplicates. Stains 99-H5 and 78-D10 have identical ITS to Penicillium sclerotiorum. More than 80% of fungal isolates belong to different species or strains, indicating the fungal library is highly diverse. A small portion (&lt;12.5%) are duplications of the other. Based on these ITS sequences, a phylogenetic tree was constructed, displaying the fungal diversity in samples (Supplemental File 1). Three genera (Trichoderma, Fusarium and Penicillium) present in three big branches, which is consistent to our sampling sources, e.g., soil and plants. Trichoderma and Fusarium are the most prevalent soil fungi and many are associated with plants <ns0:ref type='bibr'>(Harman, Howell et al. 2004)</ns0:ref>. Penicillium is ubiquitous genus with more than 350 species already identified <ns0:ref type='bibr'>(Visagie, Houbraken et al. 2014)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Endophytic fungi from Chinese medicinal plants</ns0:head><ns0:p>Our fungal library includes many endophytic fungi. Since endophytic fungi produce many plant metabolites with medical functions, we collected plant samples from the three most <ns0:ref type='table'>4</ns0:ref>). We separated different parts from each plant, e.g., leaf, bark, stem, fruits, and seeds, and sterilized their surfaces with 75% ethanol. Following grinding masses and culturing on Malt Extract Agar (MEA) plates, more than 50 endophytic fungi were isolated on MEA plates. The </ns0:p></ns0:div> <ns0:div><ns0:head>Construction of a fungal extract library</ns0:head><ns0:p>To generate a fungal metabolite library, we used the cereal-based medium to produce the secondary metabolites as reported <ns0:ref type='bibr'>(Niu, Wang et al. 2015)</ns0:ref>. Each fungus was cultured in a testing tube with six small pieces of cereals. After culturing for one month, we used ethyl acetate to extract the secondary metabolites. We obtained 9,053 crude extracts in total, each of which corresponds to a specific fungal isolate. As anticipated, different fungal strains produced different amounts of secondary metabolites from 1 mg to 20 mg per gram culture, and have various physical features such as stickiness, odors, and solubility. More than 90% of the extracts have colors, including green, orange, red, yellow, purple, and others. The crude extracts were dissolved in DMSO to generate 2 mg/mL solution. For future reference, we name this library 'Global Fungal Extract Library' or GFEL in brief.</ns0:p></ns0:div> <ns0:div><ns0:head>Screen the fungal extract library against P. falciparum transmission to mosquitoes</ns0:head><ns0:p>Malaria remains a devastating disease, and Anopheles midgut protein fibrinogen-related protein 1 (FREP1) mediates Plasmodium transmission <ns0:ref type='bibr'>(Li, Wang et al. 2013, Zhang, Niu et al.</ns0:ref> PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed 2015, <ns0:ref type='bibr'>Niu, Franca et al. 2017</ns0:ref>). FREP1 mediates Plasmodium invasion in mosquitoes by binding to P. falciparum gametocytes or ookinetes <ns0:ref type='bibr'>(Zhang, Niu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>To examine the usability of the newly constructed fungal library, we screened 460 fungal extracts obtained in later 2016 and early 2017 for their inhibition activity against FREP1-P. Penicillium thomii and Penicillium pancosmium, respectively (Table <ns0:ref type='table'>5</ns0:ref>). Notably, this is the first report about P. thomii and P. pancosmium that produce secondary metabolites with antimalarial activities. An independent project in our lab identified Asperaculane B as an active compound from this GFEL that inhibited malaria transmission to mosquitoes <ns0:ref type='bibr'>(Niu, Hao et al. 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Examine the usability of fungal extract library in finding antibiotic leads</ns0:head><ns0:p>Antibiotic-resistant bacteria threaten public health <ns0:ref type='bibr' target='#b29'>(Todd 2017)</ns0:ref>. We screened potential antibiotics from this newly established fungal library against antibiotic-resistant bacterial</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed pathogens. We randomly picked 574 fungal extracts and examined their activity in inhibiting Gram-positive (methicillin-resistant S. aureus MRSA) and Gram-negative (S. flexneri) bacterial pathogens. These fungi were isolated from samples collected in late 2017. Among them, 47 inhibited the growth of S. aureus MRSA (hit rate of 10.8%), and one inhibited the growth of S.</ns0:p><ns0:p>flexneri (hit rate of 0.17%, Table <ns0:ref type='table'>6</ns0:ref>). The hit rate of antibiotics against Gram-positive bacteria was higher than that of the Gram-negative bacteria, which consistent to the well-known challenges in antibiotic discovery against gram-negative pathogens. Gram-negative pathogens have unique outer membrane and efflux pumps <ns0:ref type='bibr' target='#b6'>(Fair and Tor 2014)</ns0:ref>.</ns0:p><ns0:p>We also examined the effect of 288 fungal extracts on non-methicillin-resistant S.</ns0:p><ns0:p>aureus. The results showed that 22 prevented the growth of non-methicillin-resistant S. aureus (hit rate of 7.6%). Besides, we analyzed another two featured bacteria, Mycobacterium smegmatis, which is gram-positive and acid-fast dye staining cell wall, and E. coli-AS17tolc, which has gram-negative cell wall, but more permeable than the wild type E. coli. We obtained eight candidates against Mycobacterium smegmatis and three candidates against E. coli-AS17tolc. The results showed 10.8% and 1% hit rates to Mycobacterium smegmatis and E. coli-AS17tolc, respectively (Table <ns0:ref type='table'>6</ns0:ref> and Supplemental File 3). The results show that the secondary metabolites produced by different types of fungi in the GFEL contained compounds biologically against multi-resistant, various pathogenic strains of bacteria.</ns0:p><ns0:p>Examine the usability of fungal extract library in finding anti-chronic myeloid leukemia candidates Finally, we studied the possibility of obtaining drug candidate leads against chronic myelocytic leukemia (CML), a malignant tumor of the blood system <ns0:ref type='bibr'>(Kaleem, Shahab et al. 2015)</ns0:ref>. A small subset of endophytic fungi from Chinese medicinal plants was used for this purpose. Fifty extracts were examined for their inhibition against the human immortalized myelogenous leukemia cell line K562 using the MTT method. The final concentrations of crude extracts were 20 &#956;g/mL. The results from triplicates showed that 4 extracts <ns0:ref type='bibr'>(#17, #22, #29, #35)</ns0:ref> PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed and alkaloids, and has good antibacterial activity, immunomodulatory effect, and anti-tumor effect. We PCR-amplified the ITS region of the fungal candidate (#29) and sequenced the region. Based on the ITS region sequence (Accession # at GenBank: MT594489), the candidate fungus was Tolypocladium album. Notably, a tetrameric acid from Tolypocladium album has been reported to inhibit the tumor's growth <ns0:ref type='bibr'>(Fukuda, Sudoh et al. 2015)</ns0:ref>. Further studies will be conducted to isolate and identify the bioactive compounds from this fungus.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>We established a comprehensive fungal library that is diverse and useful for communities. We also demonstrated the usability of GFEL and identified a set of fungal strains that produce secondary metabolites to inhibit Plasmodium falciparum's transmission, chronic pneumonia development, and bacteria proliferation. Further studies will identify the active compounds for drug development. The fungal isolates and the corresponding Chinese medicinal host plants.</ns0:p><ns0:p>The images in rows 1, 3, and 5 were the fugal isolates grown on MEA plates, and the photos below in row 2, 4, and 6 are the corresponding host plants where they were isolated.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>The transmission-blocking activity of the extracts candidates by screening the fungal library with the in vitro FREP1-parasite interaction-based ELISA assays. Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4</ns0:note><ns0:p>Anti-chronic myeloid leukemia screening results.</ns0:p><ns0:p>About 50 extracts of endophytic fungi isolated from the Chinese medicinal plant-fungal metabolite library were examined against K562 cells using the MTT method. The results show Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>: A subset of fungal extracts, randomly selected, were tested for the antibacterial activity to inhibit the growth of Shigella flexneri (ATCC 9199), Staphylococcus aureus (ATCC 14775), methicillin-resistant Staphylococcus aureus (MRSA, ATCC BAA-44) and E.coli (AS17tolc). The drug-screening was performed in a 384-well microplate format using the following procedures. Bacteria were cultured in 50 mL Mueller Hinton broth (MHB) in a 150-mL flask overnight at 37&#186;C. The next day, the cells were first OD 600 adjusted to 0.1 and then further diluted 1:100 in MHB, and a volume of 50 &#181;L (~10 5 CFU) is added to each well of the 384-well sterile microplates (Thermo Fisher Scientific). Fungal extracts (0.5 &#181;L in DMSO) were then added to each test well at a final concentration of 40 &#181;g/mL. The plates were incubated for 18-20 hr at 37 &#186;C. At the end of this incubation, resazurin (Sigma-Aldrich) was added to the wells to determine the growth of bacteria. The final concentration was 0.02%, and the plates were further incubated for 4-6 hr at 37 &#186;C. In the presence of viable cells, resazurin was reduced to resorufin (pink) along with an increase in fluorescence(O'Brien, Wilson et al. 2000). Extracts showing antibiotic activity (hits) were scored as those that prevented the color change and also reduced the fluorescence (Ex 540, Em 590nm) by 90% when compared to the control wells containing no inhibitor.Ciprofloxacin was used as a positive control for bacterial growth inhibition. Three wells werePeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020) Manuscript to be reviewed used for each sample. For the positive candidate extracts, we repeated the experiments at least once to confirm the results. Screening the fungal extract library to identify drugs leads against chronic myeloid leukemia with MTT assays: Cell proliferation was analyzed by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) using Vybrant&#174; MTT Cell Proliferation Assay (Thermo Fisher Scientific) with the human immortalized myelogenous leukemia cell line K562. Around 2x10 4 cell in 100 &#181;L culture medium (RPMI 1640 + 2mM glutamine + 10% fetal bovine serum)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020) Manuscript to be reviewed collection in location, weather, climate, and altitude (Fig 1B) promises the diversity of fungal species and their genetic background.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>extensive tropical botanic gardens in China, including South China, Hainan Xinglong, and PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020) Manuscript to be reviewed Yunnan Tropical. There are many diverse plants in these gardens. We collected 27 well-known Chinese medicinal plants, such as Chinese black olive (Canarium pimela), Chinese croton (Excoecaria cochinchinensis), Lemon-scented gum (Eucalyptus citriodora Hook.f), Sweet osmanthus (Osmanthus fragrans), and Yellow cow wood (Cratoxylum cochinchinense) (Table</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>colors and morphology of these endophytic fungi and their corresponding plants look strikingly different (Fig 2). The ITS of three fungi were PCR-amplified and sequenced (Accession # in GenBank are MT994711, MT994712, and MT594489). They were identified as Stephanonectria keithii (Fig 2, C1), Aspergillus sp. (Fig 2, E1), and Tolypocladium album (Fig 2, H3), respectively.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>falciparum</ns0:head><ns0:label /><ns0:figDesc>interaction and found 4 extracts that prevented FREP1 from binding to P. falciparum lysates by over 90%. The fungal colonies of these four fungal candidates on MEA plates showed different colors and shapes (Fig 3A). Then, we determined the activities of the four candidates against P. falciparum to An. gambiae in vivo using the standard membrane feeding assays (SMFA). Results show that the fungal extracts of 37C6 and 22E8 could block malaria transmission at a concentration of 100 &#181;g/mL, and 100D3 and 45F10 did not (Fig 3B, Supplemental File 2). The oocyst numbers of 37C6 or 22E8 extract-treated mosquitoes were nearly zero, while the oocyst numbers in the 100D3 or 45F10 extract-treated mosquitoes were not significantly different from that of the DMSO control (Fig. 3B). After further dilution of the two positive candidate fungal extracts (37C6 and 22E8) to 20 &#181;g/mL, we found that the extract of 22E8 still significantly inhibited the activity in P. falciparum transmission to mosquitoes (Fig 3C, Supplemental File 2). We sequenced the ITS sequences of the four candidate fungi and their accession number are MT594486-MT594488 and MT613342 at GenBank at https://www.ncbi.nlm.nih.gov/genbank/sequenceids/. According to ITS sequences of 22E8 (Acc #: MT613342) and 37C6 (Acc#: MT594487), the fungal species of 22E8 and 37C6 were</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>significantly inhibited the proliferation of K562 cells (p&lt;0.05, Fig 4, Supplemental File 4). The hit rate of the plant-fungal metabolite library was 8%. Notably, the survivorship of K562 cells with extract #29 was about 24.8% compared with the control, e.g., the inhibition of extract #29 on K562 proliferation was as high as 75.2%. The specimen of #29 was from the fungus (Fig 2, H3) isolated from the plant Litsea glutinosa (Fig 2, H4) collected from the Medical Botanical Garden of South China Botanical Garden in Guangzhou. Litsea glutinosa has been used to treat diarrhea, traumatic injuries, mumps, and rheumatism. It contains abundant flavonoids, terpenes,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>The morphology of fungal isolates 37C6 and 22E8T on the MEA agar plate. (B): The final concentrations of the fungal extracts were 100 &#956;g/mL and the results show that the fungal extracts (37C6 and 22E8) significantly reduced the oocyst number compared with the DMSO control while the oocyst number of the other two (100D3 and 45F10) was not significantly different with the DMSO control, respectively. (C): Further, the fungal extract of 37C5 continued to show a significant reduction of the oocyst number in midgut while the 22E8 fungal extract did not have significant effects on P. falciparum infection in mosquitoes when the concentration of the fungal extracts was decreased to 20 &#956;g/mL. N: the number of mosquitoes for each treatment; mean: the average number of oocysts per midgut; PR: infection prevalence in mosquitoes. p: the p-value was calculated by the Mann-Whitney-Wilcoxon test. The experiments were repeated three times. PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>that 4 extracts(#17, #22, #29, #35) significantly inhibited the growth of K562 cells (p &lt;0.05). Notably, the survivorship of K562 cells with extract #29 was about 24.3% compared with the control, or the inhibition of extract #29 on K562 was 75.7%. The DMSO was applied as the control. The data shows the means and standard deviations of triplicates.PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,70.87,305.48,672.95' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head /><ns0:label /><ns0:figDesc>To achieve this goal, we collected samples globally. Current collection includes samples from Kenya, Myanmar, USA, and China. We received about 2395 soil samples and 2324 plant samples. We collected the whole plant or separated plant parts such as roots, stems, leaves, flowers, fruits, or various combinations of components. The samples were from 36 regions, including Nairobi in Kenya, 10</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Kenya. Approximate 4% were from tropical/subtropical highland climate areas such as Nairobi</ns0:cell></ns0:row><ns0:row><ns0:cell>in Kenya, the Lijiang National Park, and the Potatso National Park in China. A small portion</ns0:cell></ns0:row><ns0:row><ns0:cell>(2.1%) was from the cold areas, such as Jiuzhaigou National Park in China, and Alaska in the</ns0:cell></ns0:row><ns0:row><ns0:cell>USA (Fig 1A). We collected the samples from different landforms, including hills, mountains,</ns0:cell></ns0:row><ns0:row><ns0:cell>plateaus, canyons, valleys, and bays. Notably, some fungi were isolated from samples</ns0:cell></ns0:row></ns0:table><ns0:note>regions in the United States, Yangon in Myanmar, and 24 districts in China (Table2). The collection places cover various climate zones.From these samples, 9,053 fungal isolates in total were cultured. Among them, 2,356 were from the plant samples, and 6,688 were from soil samples. About one fungal strain per plant-part sample and 2.8 fungal isolates per soil sample were obtained by average. Nearly 69.4% of fungal isolates were from the subtropical climate in China (Shanghai, Guangzhou, and Chongqing) and the USA (e.g., Dallas, New Orleans, and Oklahoma City). About 8% of fungal isolates were from tropical climates such as Miami in the USA, Yangon in Myanmar, Kisumu in collected from the mountains with an altitude over 3,000 meters, such as the Cang Mountains and Meili Snow Mountains in the Yunnan Province, China. The vast difference of sample</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>PCR Primers for fungal ITS regions</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:note></ns0:figure> <ns0:note place='foot'>Manuscript to be reviewed</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49300:1:1:NEW 3 Oct 2020)</ns0:note> </ns0:body> "

"September 9, 2020 Dear Dr. Gomez, We thank the reviewers for their time and comments on the manuscript. We have edited the manuscript to address their concerns. In particular, 1) Supplemental File 1 was added to display a phylogenetic tree of 40 fungi from the fungal library; 2) Figure 2 was modified to add indexes, plant species and some fungal species; 3) A citation (Molecules, 2020. 25:13) was added to demonstrate a bioactive compound identified from this newly established fungal extract library; 4) Figure 4 was modified as following: Figure 4A was re-done to improve the quality. Panel B was deleted. The plant and fungus photos were referred in Figure 2B. 5) Dr. Stephen Munga from Kenya Medical Research Institute was added as a co-author for his contribution in Kenya; 6) ITS primer sequences were added as a new table (Table 1); 7) Taxonomic names in Table 3 (Table 4 now) were corrected; and 8) More references were added as requested. In summary, all issues raised by reviewers were carefully evaluated and addressed in the revision. Sincerely, Dr. Jun Li Associate Professor Tel: 305-348-7618 Email: [email protected] On behalf of all authors DEPARTMENT OF BIOLOGICAL SCIENCES 11200 S.W. 8th St. Modesto Maidique Campus, OE-167. Miami, FL 33199. Tel: 305-348-2201. Reviewer 1 (Anonymous) Basic reporting With regard to the manuscript titled 'A diverse global fungal library for drug discovery,' I believe the work can be published. Building a fungal library with such a wide diversity and at a global level is very useful for high throughput screening of fungal secondary metabolites for bioactive compounds. I am concerned about the taxonomic names of the fungi, please see my comment below. I think the reference citations can be improved. For some, reason I felt the authors have few references cited. Figure 2. The plates are all streaked like bacteria, but in Mycology that is not usually how the fungal transfers are performed - just a suggestion. Figure 4b. Really cannot see anything clearly. Also, check the spelling of DMSO Answer: These concerns were addressed one by one in the following paragraphs. Comments for the Author Chemistry and Pharmacology testing Comment #1: Citing references is needed for the information stated at lines 46-47, lines 47-48, and line 52. Response: Citations (8-15) were added. Comment #2: Time units such as seconds, minutes, and hours can be abbreviated as sec, min, and h, respectively. Response: Agree. Sec, min and hr were used as abbreviations of seconds, minutes, and hours. Comment #3: The picture used in figure 2, it is better to add a label on the left side of rows 1, 3, and 5 as fungal isolates and rows 2, 4, and 6 as host plants. Response: Done as suggested. In addition, we added the plant host name to label plant. We also added fungal species names to three identified isolates. Comment #4: In figure 4, authors need to specify whether the anti-chronic myeloid leukemia screening was performed in triplicate, and what the error bars actually represent? Response: Done as suggested. In the results (line 341) and Figure 4 legends, we clarified that the results showed the means and standard deviations from triplicates. Comment #5: It is nice to see that crude extracts were obtained for the total 9,053 fungal strains. It would be good to study the chemical diversity of these extract samples and develop a dereplication method for the purpose of discovering novel bioactive leads. Response: Indeed, one active compound that inhibits malaria transmission was identified from this newly established fungal extract library, and the related work was recently published in Molecules, 2020. 25:13. The citation is added (line 312, Citation 32). Comment #6: Out of the 9, 053 crude fungal extracts, some were randomly selected for antimalarial, antimicrobial, and cytotoxic activities. The authors need to identify the method followed to randomly select these samples for screening. Were these samples representative for all different locations and climate conditions. Were there any extracts that were tested against more than one screening assay? Response: The fungal extracts were screened against malaria, bacteria and myeloid leukemia by three different labs (Jun Li’s lab, Yukching Tse-Dinh’s lab, and Guomin Niu’s lab respectively). Each lab took fungal extracts randomly without describable patterns (lines 292-293, 318, 338-339). The screened extracts against malaria were from fungi collected earlier (line 292-293). The screened extracts against bacteria were from fungi DEPARTMENT OF BIOLOGICAL SCIENCES 11200 S.W. 8th St. Modesto Maidique Campus, OE-167. Miami, FL 33199. Tel: 305-348-2201. collected later (line 318). About 50 extracts from endophytic fungi collected from South China Botanical Garden were examined against myeloid leukemia (line 338). Therefore, three sets were different without overlap. Comment # 7: The authors might in the future make a fraction library from their extracts and then perform bioassays. Each extract can undergo flash chromatography to make the fractions, which can then be tested against specific bioassays. Thus, when activity is found it would help zoom into the compound peak that is responsible for the activity. Response: We agree. We will construct a fraction library to facilitate the drug discovery. Mycology Comment #1 Reference for ITS primers is not given. Response: The primers sequences were added as a new table (Table 1), and the related reference (24) was added. Comment #2 The identification to species level with the ITS region might not necessarily be accurate. Especially for genera like Penicillium and Fusarium. Thus, please only use the genus name for identification. All Table 3 species names might not be accurate unless they are identical to the type sequence for ITS in GenBank. Thus the authors need to thoroughly check their identification to species-level with the ITS region just via BLAST search in NCBI GenBank with no consultation with type databases Response: We agree. Species in Table 3 were modified as suggested, e.g., only use the genus name for identification unless they are identical to the type sequence for ITS in GenBank. Comment #3 Did the authors follow the Nagoya protocol for collections in Africa? I don't see any collaborator from an African University/institution? Response: We followed the Nagoya protocol. All available sequence data have been shared with the public through GenBank. As stated in acknowledgement, many persons provided assistance for this work. The samples from Kenya were cobtained through the collaboration between Florida International University and Kenya Medical Research Institute. Dr. Stephen Munga from KEMRI was added as a co-author. DEPARTMENT OF BIOLOGICAL SCIENCES 11200 S.W. 8th St. Modesto Maidique Campus, OE-167. Miami, FL 33199. Tel: 305-348-2201. Reviewer 2 (Anonymous) Basic reporting Review of: A diverse global fungal library for drug discovery The manuscript entitled “A diverse global fungal library for drug discovery” describes the construction of a fungal extract library containing extracts from 9053 fungal isolates obtained from approximately 5000 (soil and plant samples harvested in 36 regions in Africa, Asia, and North America. The biological activity of randomly selected extracts was evaluated in several assays including P. falciparum transmission to mosquitoes, antibacterial, and anti-chronic myeloid leukemia. Generally, this manuscript is well written, however, some references are missing in the introduction section. Additionally, reference 9 is not appropriate for the text, it does talk about actinomycetes instead of ascomycetes. Response: More citations were added. The reference 9 (new citation 16) has been corrected. If well, the isolation of 9053 fungal microorganisms is a titanic task, the information provided about the diversity in the chemistry and taxonomy of the isolates is not well presented in the submitted version; a phylogenetic tree must be ideal to represent the taxonomic distribution of the isolates. Response: The phylogenetic tree of 40 randomly selected fungi was constructed based on their ITS sequences (Supplemental File 1). The phylogenetic tree construction was added in method (line 211-214), the related result and discussion were added at line 254-259. More importantly, the manuscript does not present any evidence of the chemistry of the isolates, possibly, an untargeted metabolomics study will provide such information. Ideally, the isolation and structure elucidation of some bioactive compounds is recommended. Response: An active compound against malaria has been isolated from this GFEL by our lab. Its structure was determined to be Asperaculane B. The results and related work have just been recently published in Molecules 2020 25(13):3018 (line 311-312, citation 32). Comments for the Author Review of: A diverse global fungal library for drug discovery Generally, this manuscript is well written, however, some references are missing in the introduction section. Additionally, reference 9 is not appropriate for the text, it does talk about actinomycetes instead of ascomycetes. If well, the isolation of 9053 fungal microorganisms is a titanic task, the information provided about the diversity in the chemistry and taxonomy of the isolates is not well presented in the submitted version; a phylogenetic tree must be ideal to represent the taxonomic distribution of the isolates. More importantly, the manuscript does not present any evidence of the chemistry of the isolates, possibly, an untargeted metabolomics study will provide such information. Ideally, the isolation and structure elucidation of some bioactive compounds is recommended. Response: The issues have been addressed one by one as described in sections above. Taking into account the above concerns and the suggestions marked in the attached pdf file, the manuscript must be accepted after major revisions. Response: All comments in marked pdf were addressed completely (shown in tracking manuscript). DEPARTMENT OF BIOLOGICAL SCIENCES 11200 S.W. 8th St. Modesto Maidique Campus, OE-167. Miami, FL 33199. Tel: 305-348-2201. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Secondary fungal metabolites are important sources for new drugs against infectious diseases and cancers. Methods. To obtain a library with enough diversity, we collected about 2395 soil samples and 2324 plant samples from 36 regions in Africa, Asia, and North America. The collection areas covered various climate zones in the world. We examined the usability of the global fungal extract library (GFEL) against parasitic malaria transmission, Gram-positive and negative bacterial pathogens, and leukemia cells.</ns0:p></ns0:div> <ns0:div><ns0:head>Results.</ns0:head><ns0:p>Nearly ten thousand fungal strains were isolated. Sequences of nuclear ribosomal internal transcribed spacer (ITS) from 40 randomly selected strains showed that over 80% were unique. Screening GFEL, we found that the fungal extract from Penicillium thomii was able to block P. falciparum transmission to Anopheles gambiae, and the fungal extract from Tolypocladium album was able to kill myelogenous leukemia cell line K562.</ns0:p><ns0:p>We also identified a set of candidate fungal extracts against bacterial pathogens.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Natural products, produced by living organisms in nature, have been used as medicine for thousands of years <ns0:ref type='bibr' target='#b2'>(Dias, Urban et al. 2012</ns0:ref><ns0:ref type='bibr' target='#b1'>, Buyel 2018)</ns0:ref>. For instance, the treatment of malaria was recorded in China with the Qinghao plant (Artemisia annua) as early as the second century. The active ingredient, qinghaosu (artemisinin), was isolated from the plant by Youyou</ns0:p><ns0:p>Tu and her colleagues in 1971 <ns0:ref type='bibr' target='#b21'>(Luo and</ns0:ref><ns0:ref type='bibr'>Shen 1987, Hsu 2006)</ns0:ref>. The establishment of microbiology in the early modern era led to drug discoveries from microbes. The first antibiotics, penicillin, was discovered from a fungus by Alexander Fleming <ns0:ref type='bibr' target='#b5'>(Fleming 1929)</ns0:ref>. Fungi have initially been and still are used to produce medicines to treat infectious diseases <ns0:ref type='bibr' target='#b3'>(Elder 1944)</ns0:ref>.</ns0:p><ns0:p>Furthermore, people use natural products to treat tumors. Indeed, more than half of anti-tumor drugs or leads in current clinical trials are from natural products <ns0:ref type='bibr' target='#b36'>(Wolfender and Queiroz 2012)</ns0:ref>. Microbial metabolites are essential resources for drug discovery <ns0:ref type='bibr' target='#b18'>(Lenzi, Costa et al. 2018)</ns0:ref>. Compared with other natural product resources, fungi have the following advantages.</ns0:p><ns0:p>First, there are enormous fungal species: about 120,000 fungal species have been described <ns0:ref type='bibr' target='#b11'>(Hawksworth and Lucking 2017)</ns0:ref> and 5.1 million fungal species are estimated <ns0:ref type='bibr' target='#b0'>(Blackwell 2011)</ns0:ref>.</ns0:p><ns0:p>Second, fungi produce broad and diverse secondary metabolites with a vast difference in chemical structures <ns0:ref type='bibr' target='#b29'>(Pham, Yilma et al. 2019)</ns0:ref>. Third, large-scale fermentation can generate a large amount of fungal secondary metabolites, which was exampled by the production of alcohol and lactic acid. However, yield of many target fungal secondary metabolites is restricted by fungal growth and differentiation <ns0:ref type='bibr' target='#b23'>(Nielsen and Nielsen 2017</ns0:ref><ns0:ref type='bibr' target='#b17'>, Keller 2019</ns0:ref><ns0:ref type='bibr' target='#b29'>, Pham, Yilma et al. 2019)</ns0:ref>, which is resolved by new technologies that enable us to engineer a fungus to produce a specific compound in high yield by modifying its metabolic pathways (van Dijk and Wang 2016).</ns0:p><ns0:p>Also, the recent development of genomic sequencing technology and the identification of more biosynthetic gene clusters accelerate the discovery and application of new compounds from fungi <ns0:ref type='bibr' target='#b14'>(Hussain, Al-Sadi et al. 2017</ns0:ref><ns0:ref type='bibr' target='#b17'>, Keller 2019)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Isolation of fungi from soil and plants:</ns0:head><ns0:p>For each soil sample, we transferred 50 mg soil to a 1.5 mL plastic tube, and 1 mL autoclaved distilled water was added. The sample was vortexed for 30 secs and centrifuged at 500 g for 2 minutes (min) to get rid of the soil particles. For plant samples, the first step was to sterilize the plant surface by rinsing the sample with distilled H 2 O, then soaking in 70% ethanol for 10 seconds (sec) and rinsing again with distilled H 2 O. After sterilization, the plants were cut into 0.5 cm x 0.5 cm pieces and transferred into a sterile mortar.</ns0:p><ns0:p>Then, 2 mL of distilled H 2 O was added, followed by grinding with a pestle for 1-3 min and the slurry was transferred to a 1.5 mL plastic tube and then centrifuged at 500 g for 2 min to get rid of the particles. About 100 &#956;L of upper supernatant from treated soil or plants was evenly spread onto a 100 x 15 mm Petri Dish plate containing 14 mL Malt Extract Agar (MEA), made of 10 g malt extract, 1 g yeast extract, 15 g agar, and 0.05 g chloramphenicol (Sigma-Aldrich, St.</ns0:p><ns0:p>Louis, MO) in 1 L distilled H 2 O and autoclaved at 121 &#186;C for 20 min. The plates were sealed with parafilm, and the fungi were allowed to grow for 7-14 days at room temperature (RT) with cycles of 12 hours (hr) of darkness and 12 hr of light.</ns0:p><ns0:p>The fungal colonies on the MEA medium plates were picked with a toothpick and inoculated in a new MEA plate by streaking. If the colonies were mixtures of two or more species, we kept inoculating and streaking until a single colony appeared. Finally, a piece of fungal agar containing mycelium or spores was cut and transferred to a 1.5 mL Eppendorf tube containing 500 &#956;L of sterile 20% glycerol in distilled H 2 O. We stored the cells in a -80 &#176;C freezer for long-term storage.</ns0:p></ns0:div> <ns0:div><ns0:head>Metabolite production and extraction:</ns0:head><ns0:p>Cereal based medium was used to grow fungi to produce secondary metabolites <ns0:ref type='bibr' target='#b26'>(Niu, Wang et al. 2015)</ns0:ref>. Briefly, six pieces of Cheerios Breakfast cereals (General Mills, Minneapolis, MN) were placed to a glass test tube, capped with a plastic lid and autoclaved for 20 min, and then 2 mL sterile sucrose water (3 g of sucrose and 50 mg chloramphenicol in 1 L distilled H 2 O) was added into the tube. Later, the fungal colony grown on the MEA plate was inoculated into the cereal medium and incubated at RT for PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed one month to produce sufficient metabolites. A month late, 2 mL ethyl acetate was added into a tube to extract the fungal metabolites, mixed with a glass stirring rod, and placed overnight in a chemical hood with gentle shaking. The next day, 1 mL upper layer of supernatant was transferred to a 1.5 mL tube. After centrifugation (2, 000 g for 2 min), around 950 &#956;L clear supernatant was transferred to a pre-weighed 1.5 mL plastic tube and dried with SpeedVac concentrator (Thermo Fisher Scientific, Waltham, MA). Finally, the dry extracts were weighed and dissolved in an appropriate amount of dimethyl sulfoxide (DMSO) to prepare 10 mg/mL stock solution and stored in a -20 &#176;C freezer for future screening assays.</ns0:p><ns0:p>Determination of fungal species: Forty fungal isolates were randomly picked to evaluate the fungal library's diversity. The fungi were cultured with liquid malt extract medium at RT for one week, and mycelium was collected for DNA extraction using DNAzol (Thermo Fisher). Genomic DNA applied as PCR templates were isolated using DNAzol Reagent following the manual (Thermo Fisher Scientific). To identify the fungal species, nuclear ribosomal ITS regions were amplified with by PCR with specific primers ( min. The amplified products were sequenced and blasted against the NCBI database to identify fungal species <ns0:ref type='bibr' target='#b30'>(Raja, Miller et al. 2017)</ns0:ref>.</ns0:p><ns0:p>Screening the fungal extract library to identify malaria transmission-blocking candidates with ELISA assays: As described previously <ns0:ref type='bibr' target='#b26'>(Niu, Wang et al. 2015)</ns0:ref> Manuscript to be reviewed washed three times with RPMI-1640 at 300 &#215; g for 4 m. The cell pellets were re-suspended in PBST (PBS containing 0.2% Tween-20) and homogenized by ultra-sonication with six cycles of 10 sec pulse and 50 sec resting on ice for each period. The lysates were centrifuged at 8,000 g for 2 min to remove insoluble materials and cellular debris. With the iRBC lysate and insect cellexpressed recombinant FREP1, the ELISA assay was used to screen the fungal extract library to block FREP1-parasite interaction <ns0:ref type='bibr' target='#b26'>(Niu, Wang et al. 2015)</ns0:ref>. A 96-well ELISA plate was coated with 50 &#956;L iRBC lysate (2 mg/mL protein) overnight at 4&#176;C. After coating, the plate was blocked with 100 &#956;L of PBS plus 0.2% bovine serum albumin (BSA) per well for 1.5 hr at RT. After removal of the blocking solution, FREP1 (10 &#956;g/mL) in blocking buffer (PBS plus 0.2% BSA)</ns0:p><ns0:p>was added to each well, and 1&#956;L fungal extract was taken from a 96-well plate containing 2 mg/mL crude extract dissolved in DMSO in each well with a multiple-channel pipette and transferred to the ELISA plate, then incubated for 1 hr at RT with gentle shaking. After washing three times with PBST, 50 &#956;L rabbit anti-FREP1 polyclonal antibody <ns0:ref type='bibr' target='#b26'>(Niu et al., 2015)</ns0:ref> (diluted 1: 5,000 in blocking buffer, 1 &#956;g/mL) was added to each well and incubated for 1 hr at RT. About 50 &#956;L alkaline phosphatase-conjugated anti-rabbit IgG (diluted 1: 20,000 in blocking buffer) was added to each well and incubated for 45 min at RT. The wells were washed three times with PBST between incubations. After washing, each well was developed with 50 &#956;L pNPP substrate (Sigma-Aldrich) until the colors were visible, and absorbance at 405 nm was measured. The functional FREP1 supplemented with 1 &#956;L solvent (DMSO) was used as non-inhibition control, and the heat-inactivated FREP1 (65 &#176;C for 15 min) was used as a 100% inhibition control.</ns0:p></ns0:div> <ns0:div><ns0:head>Determination of the transmission-blocking activity of the fungal extracts in mosquitoes:</ns0:head><ns0:p>Following the previous protocol <ns0:ref type='bibr' target='#b37'>(Zhang, Niu et al. 2015)</ns0:ref>, the 15-to 17-day old cultured P.</ns0:p><ns0:p>falciparum containing 2-3% stage V gametocytes were collected and diluted with new O+ type human blood to get 0.2% stage V gametocytes in the blood. Then, the 150 &#61549;L blood was mixed with the same volume of heat-inactivated AB+ human serum. Then, 3 &#61549;L candidate fungal</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed extract in DMSO (10 mg/mL or 2 mg/mL) was mixed with 297 &#61549;L infected blood, the final fungal extract concentration in blood was 100 or 20 &#181;g/mL, respectively. SMFA was performed to feed about 100 3-5 days old A. gambiae G3 female mosquitoes for 15 min, and the engorged mosquitoes were maintained with 8% sugar in a BSL-2 insectary (28 &#176;C, 12-h light/dark cycle, 80% humidity). The midguts were dissected seven days post-infection and stained with 0.1% mercury dibromofluorescein disodium salt in PBS for 16 m. The oocysts in midguts were counted under a light microscope. This standard membrane feeding assays were conducted at least twice to confirm the results. were seeded in 96-well microplates and incubated at 37 &#176;C with 5% CO 2 . The next day, one &#181;L fungal extract in DMSO was added (final concentration of the fungal extract was 20 &#181;g/mL), and the cells were incubated at 37 &#176;C with 5% CO 2 for another 24 hr. Next, the microplate was centrifuged at 500 g for 10 min to pellet the cells, the medium was carefully removed as much as possible, and 100 &#181;L of fresh medium was then added. About 10 &#181;L of the 12 mM MTT stock solution was added, mixed, and incubated for 4 hr at 37 &#176;C. The microplate was centrifuged again at 500 g for 10 m. After removing 75 &#181;L of the medium from the wells with 25 &#181;L medium with cells left, 50 &#181;L of DMSO was added to each well, mixed, and incubated at 37 &#176;C for 10 min to dissolve formazan crystal for measurement. The same amount of DMSO without drugs was applied as a control. The optical density was measured at an absorbance wavelength of 540 nm. Cell growth inhibition rate (%) = (A 540 of control -A 540 of treatment/ A 540 of control) &#215; 100%.</ns0:p></ns0:div> <ns0:div><ns0:head>Screening the fungal extract library to identify antibiotics</ns0:head><ns0:p>Triplicates were conducted for each sample.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis:</ns0:head><ns0:p>All the experiments were independently repeated at least twice and analyzed with the Wilcoxon-Mann-Whitney test using GraphPad Prism (GraphPad Software, CA, USA).</ns0:p><ns0:p>Sequence availability: All fungal ITS sequences obtained in this project have been deposited</ns0:p><ns0:p>into GenBank at NCBI (https://www.ncbi.nlm.nih.gov/genbank/sequenceids/).</ns0:p></ns0:div> <ns0:div><ns0:head>Construction of phylogenetic tree:</ns0:head><ns0:p>Forty fungal ITS sequences in FASTA format were input into an online multiple sequence alignment tool (https://www.ebi.ac.uk/Tools/msa/clustalo/)</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed using Clustal Omega algorithm <ns0:ref type='bibr' target='#b12'>(Higgins and Sharp 1988)</ns0:ref>. The parameter 'DNA' was selected for input sequence and 'ClustalW' was selected as output format. All other parameters were kept as defaults. Multiple sequence alignment was conducted by clicking on 'Submit'. After alignment was completed, the tab of 'Phylogenetic Tree' was clicked and the checkbox 'Real' was selected for 'Branch length' to visualize the phylogenetic tree. The phylogenetic tree was saved as a pdf file through 'print' under 'File'.</ns0:p></ns0:div> <ns0:div><ns0:head>Results and Discussion</ns0:head><ns0:p>Extensive fungal library with nearly ten thousand isolates A large and diverse fungal library is powerful in discovering new drugs. To achieve this goal, we collected samples globally. Current collection includes samples from Kenya, Myanmar, USA, and China. We received about 2395 soil samples and 2324 plant samples. We collected the whole plant or separated plant parts such as roots, stems, leaves, flowers, fruits, or various combinations of components. The samples were from 36 regions, including Nairobi in Kenya, 10 regions in the United States, Yangon in Myanmar, and 24 districts in China ( To maximize the diversity and quantity in our fungal library, we selected fungal colonies based on their location, color, and morphology on culture plates. Therefore, the fungal isolates in the library look strikingly different (Fig <ns0:ref type='figure' target='#fig_6'>1C</ns0:ref>). This visible criterion facilitates the fungal library construction. However, we also discarded many fungi that are different species with similar morphology. We examined the species diversity of our fungal library at the molecular level by randomly picking 40 fungal isolates. Their genomic DNA was isolated, the ITS regions were amplified, and PCR products were sequenced. These sequences have been deposited into GenBank at https://www.ncbi.nlm.nih.gov/genbank/sequenceids/. Their Accession numbers are MT594355-MT594393 and MT584204. We searched these sequences against NCBI DNA databases using blast. Results show that about 12.5% of total fungal isolates have identical ITS sequences to others in the library (Table <ns0:ref type='table'>3</ns0:ref>). For instance, 126-G10, 117-B9, and 45-F10 have identical ITS to Fusarium solani, and 2 of 3 might be duplicates. Stains 99-H5 and 78-D10 have identical ITS to Penicillium sclerotiorum. More than 80% of fungal isolates belong to different species or strains, indicating the fungal library is highly diverse. A small portion (&lt;12.5%) are duplications of the other. Based on these ITS sequences, a phylogenetic tree was constructed, displaying the fungal diversity in samples (Supplemental File 1). Three genera (Trichoderma, Fusarium and Penicillium) present in three big branches, which is consistent to our sampling sources, e.g., soil and plants. Trichoderma and Fusarium are the most prevalent soil fungi and many are associated with plants <ns0:ref type='bibr' target='#b10'>(Harman, Howell et al. 2004)</ns0:ref>. Penicillium is ubiquitous genus with more than 350 species already identified <ns0:ref type='bibr' target='#b35'>(Visagie, Houbraken et al. 2014)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='table'>4</ns0:ref>). We separated different parts from each plant, e.g., leaf, bark, stem, fruits, and seeds, and sterilized their surfaces with 75% ethanol. Following grinding masses and culturing on Malt Extract Agar (MEA) plates, more than 50 endophytic fungi were isolated on MEA plates. The </ns0:p></ns0:div> <ns0:div><ns0:head>Construction of a fungal extract library</ns0:head><ns0:p>To generate a fungal metabolite library, we used the cereal-based medium to produce the secondary metabolites as reported <ns0:ref type='bibr' target='#b26'>(Niu, Wang et al. 2015)</ns0:ref>. Each fungus was cultured in a testing tube with six small pieces of cereals. After culturing for one month, we used ethyl acetate to extract the secondary metabolites. We obtained 9,053 crude extracts in total, each of which corresponds to a specific fungal isolate. As anticipated, different fungal strains produced different amounts of secondary metabolites from 1 mg to 20 mg per gram culture, and have various physical features such as stickiness, odors, and solubility. More than 90% of the extracts have colors, including green, orange, red, yellow, purple, and others. The crude extracts were dissolved in DMSO to generate 2 mg/mL solution. For future reference, we name this library 'Global Fungal Extract Library' or GFEL in brief.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Screen the fungal extract library against P. falciparum transmission to mosquitoes</ns0:head><ns0:p>Malaria remains a devastating disease, and Anopheles midgut protein fibrinogen-related protein 1 (FREP1) mediates Plasmodium transmission <ns0:ref type='bibr' target='#b20'>(Li, Wang et al. 2013</ns0:ref><ns0:ref type='bibr'>, Zhang, Niu et al. 2015</ns0:ref><ns0:ref type='bibr' target='#b24'>, Niu, Franca et al. 2017)</ns0:ref>. FREP1 mediates Plasmodium invasion in mosquitoes by binding to P. falciparum gametocytes or ookinetes <ns0:ref type='bibr' target='#b37'>(Zhang, Niu et al. 2015)</ns0:ref>.</ns0:p><ns0:p>To examine the usability of the newly constructed fungal library, we screened 460 fungal extracts obtained in later 2016 and early 2017 for their inhibition activity against FREP1-P.</ns0:p><ns0:p>falciparum interaction and found 4 extracts that prevented FREP1 from binding to P. falciparum lysates by over 90%. The fungal colonies of these four fungal candidates on MEA plates showed different colors and shapes (Fig <ns0:ref type='figure'>3A-D</ns0:ref>). Then, we determined the activities of the four candidates against P. falciparum to An. gambiae in vivo using the standard membrane feeding Penicillium thomii and Penicillium pancosmium, respectively (Table <ns0:ref type='table'>5</ns0:ref>). Notably, this is the first report about P. thomii and P. pancosmium that produce secondary metabolites with antimalarial activities. An independent project in our lab identified Asperaculane B as an active compound from this GFEL that inhibited malaria transmission to mosquitoes <ns0:ref type='bibr' target='#b25'>(Niu, Hao et al. 2020)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>Examine the usability of fungal extract library in finding antibiotic leads Antibiotic-resistant bacteria threaten public health <ns0:ref type='bibr' target='#b32'>(Todd 2017)</ns0:ref>. We screened potential antibiotics from this newly established fungal library against antibiotic-resistant bacterial pathogens. We randomly picked 574 fungal extracts and examined their activity in inhibiting Gram-positive (methicillin-resistant S. aureus MRSA) and Gram-negative (S. flexneri) bacterial pathogens. These fungi were isolated from samples collected in late 2017. Among them, 47 inhibited the growth of S. aureus MRSA (hit rate of 10.8%), and one inhibited the growth of S.</ns0:p><ns0:p>flexneri (hit rate of 0.17%, Table <ns0:ref type='table'>6</ns0:ref>). The hit rate of antibiotics against Gram-positive bacteria was higher than that of the Gram-negative bacteria, which consistent to the well-known challenges in antibiotic discovery against gram-negative pathogens. Gram-negative pathogens have unique outer membrane and efflux pumps <ns0:ref type='bibr' target='#b4'>(Fair and Tor 2014)</ns0:ref>.</ns0:p><ns0:p>We also examined the effect of 288 fungal extracts on non-methicillin-resistant S.</ns0:p><ns0:p>aureus. The results showed that 22 prevented the growth of non-methicillin-resistant S. aureus (hit rate of 7.6%). Besides, we analyzed another two featured bacteria, Mycobacterium smegmatis, which is gram-positive and acid-fast dye staining cell wall, and E. coli-AS17tolc, which has gram-negative cell wall, but more permeable than the wild type E. coli. We obtained eight candidates against Mycobacterium smegmatis and three candidates against E. coli-AS17tolc. The results showed 10.8% and 1% hit rates to Mycobacterium smegmatis and E. coli-AS17tolc, respectively (Table <ns0:ref type='table'>6</ns0:ref> and Supplemental File 3). The results show that the secondary metabolites produced by different types of fungi in the GFEL contained compounds biologically against multi-resistant, various pathogenic strains of bacteria.</ns0:p><ns0:p>Examine the usability of fungal extract library in finding anti-chronic myeloid leukemia candidates Finally, we studied the possibility of obtaining drug candidate leads against chronic myelocytic leukemia (CML), a malignant tumor of the blood system <ns0:ref type='bibr' target='#b16'>(Kaleem, Shahab et al. 2015)</ns0:ref>. A small subset of endophytic fungi from Chinese medicinal plants was used for this</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed been reported to inhibit the tumor's growth <ns0:ref type='bibr'>(Fukuda, Sudoh et al. 2015)</ns0:ref>. Further studies will be conducted to isolate and identify the bioactive compounds from this fungus.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>We established a comprehensive fungal library that is diverse and useful for communities. We also demonstrated the usability of GFEL and identified a set of fungal strains that produce secondary metabolites to inhibit Plasmodium falciparum's transmission, chronic pneumonia development, and bacteria proliferation. Further studies will identify the active compounds for drug development. The fungal isolates and the corresponding Chinese medicinal host plants.</ns0:p><ns0:p>The images in rows 1, 3, and 5 were the fugal isolates grown on MEA plates, and the photos below in row 2, 4, and 6 are the corresponding host plants where they were isolated.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:p>The transmission-blocking activity of the extracts candidates by screening the fungal library with the in vitro FREP1-parasite interaction-based ELISA assays. Manuscript to be reviewed Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>: A subset of fungal extracts, randomly selected, were tested for the antibacterial activity to inhibit the growth of Shigella flexneri (ATCC 9199), Staphylococcus aureus (ATCC 14775), methicillin-resistant Staphylococcus aureus (MRSA, ATCC BAA-44) and E.coli (AS17tolc). The drug-screening was performed in a 384-well microplate format using the following procedures. Bacteria were cultured in 50 mL Mueller Hinton broth (MHB) in a 150-mL flask overnight at 37&#186;C. The next day, the cells were first OD 600 adjusted to 0.1 and then further diluted 1:100 in MHB, and a volume of 50 &#181;L (~10 5 CFU) is added to each well of the 384-well sterile microplates (Thermo Fisher Scientific). Fungal extracts (0.5 &#181;L in DMSO) were then added to each test well at a final concentration of 40 &#181;g/mL. The plates were incubated for 18-20 hr at 37 &#186;C. At the end of this incubation, resazurin (Sigma-Aldrich) was added to the wells to determine the growth of bacteria. The final concentration was 0.02%, and the plates were further incubated for 4-6 hr at 37 &#186;C. In the presence of viable cells, resazurin was reduced to resorufin (pink) along with an increase in fluorescence<ns0:ref type='bibr' target='#b27'>(O'Brien, Wilson et al. 2000)</ns0:ref>. Extracts showing antibiotic activity (hits) were scored as those that prevented the color change and also reduced the fluorescence (Ex 540, Em 590nm) by 90% when compared to the control wells containing no inhibitor.Ciprofloxacin was used as a positive control for bacterial growth inhibition. Three wells were PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)Manuscript to be reviewed used for each sample. For the positive candidate extracts, we repeated the experiments at least once to confirm the results.Screening the fungal extract library to identify drugs leads against chronic myeloid leukemia with MTT assays: Cell proliferation was analyzed by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) using Vybrant&#174; MTT Cell Proliferation Assay (Thermo Fisher Scientific) with the human immortalized myelogenous leukemia cell line K562. Around 2x10 4 cell in 100 &#181;L culture medium (RPMI 1640 + 2mM glutamine + 10% fetal bovine serum)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>We collected the samples from different landforms, including hills, mountains, plateaus, canyons, valleys, and bays. Notably, some fungi were isolated from samples collected from the mountains with an altitude over 3,000 meters, such as the Cang Mountains and Meili Snow Mountains in the Yunnan Province, China. The vast difference of sample collection in location, weather, climate, and altitude (Fig 1B) promises the diversity of fungal species and their genetic background.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Endophytic fungi from Chinese medicinal plantsOur fungal library includes many endophytic fungi. Since endophytic fungi produce many plant metabolites with medical functions, we collected plant samples from the three mostextensive tropical botanic gardens in China, including South China, Hainan Xinglong, and Yunnan Tropical. There are many diverse plants in these gardens. We collected 27 well-known Chinese medicinal plants, such as Chinese black olive (Canarium pimela), Chinese croton (Excoecaria cochinchinensis), Lemon-scented gum (Eucalyptus citriodora Hook), Sweet osmanthus (Osmanthus fragrans), and Yellow cow wood (Cratoxylum cochinchinense) (Table</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>colors and morphology of these endophytic fungi and their corresponding plants look strikingly different (Fig 2). The ITS of three fungi were PCR-amplified and sequenced (Accession # in GenBank are MT994711, MT994712, and MT594489). They were identified as Stephanonectria keithii (Fig 2C), Aspergillus sp. (Fig 2E), and Tolypocladium album (Fig 2Z), respectively.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>assays (SMFA). Results show that the fungal extracts of 37C6 and 22E8 could block malaria transmission at a concentration of 100 &#181;g/mL, and 100D3 and 45F10 did not (Fig 3E, Supplemental File 2). The oocyst numbers of 37C6 or 22E8 extract-treated mosquitoes were nearly zero, while the oocyst numbers in the 100D3 or 45F10 extract-treated mosquitoes were not significantly different from that of the DMSO control (Fig. 3E). After further dilution of the two positive candidate fungal extracts (37C6 and 22E8) to 20 &#181;g/mL, we found that the extract of 22E8 still significantly inhibited the activity in P. falciparum transmission to mosquitoes (Fig 3F, Supplemental File 2). We sequenced the ITS sequences of the four candidate fungi and their accession number are MT594486-MT594488 and MT613342 at GenBank at https://www.ncbi.nlm.nih.gov/genbank/sequenceids/. According to ITS sequences of 22E8 (Acc #: MT613342) and 37C6 (Acc#: MT594487), the fungal species of 22E8 and 37C6 were</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>purpose. Fifty extracts were examined for their inhibition against the human immortalized myelogenous leukemia cell line K562 using the MTT method. The final concentrations of crude extracts were 20 &#956;g/mL. The results from triplicates showed that 4 extracts(#17, #22, #29, #35) significantly inhibited the proliferation of K562 cells (p&lt;0.05, Fig 4, Supplemental File 4). The hit rate of the plant-fungal metabolite library was 8%. Notably, the survivorship of K562 cells with extract #29 was about 24.8% compared with the control, e.g., the inhibition of extract #29 on K562 proliferation was as high as 75.2%. The specimen of #29 was from the fungus (Fig2Z)isolated from the plant Litsea glutinosa (Fig 2HH) collected from the Medical Botanical Garden of South China Botanical Garden in Guangzhou. Litsea glutinosa has been used to treat diarrhea, traumatic injuries, mumps, and rheumatism. It contains abundant flavonoids, terpenes, and alkaloids, and has good antibacterial activity, immunomodulatory effect, and anti-tumor effect. We PCR-amplified the ITS region of the fungal candidate (#29) and sequenced the region. Based on the ITS region sequence (Accession # at GenBank: MT594489), the candidate fungus was Tolypocladium album. Notably, a tetrameric acid from Tolypocladium album has</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>The morphology of fungal isolates 37C6, 100D3, 22E8T, and 45F10 on the MEA agar plate. (E): The final concentrations of the fungal extracts were 100 &#956;g/mL and the results show that the fungal extracts (37C6 and 22E8) significantly reduced the oocyst number compared with the DMSO control while the oocyst number of the other two (100D3 and 45F10) was not significantly different with the DMSO control, respectively. (F): Further, the fungal extract of 37C5 continued to show a significant reduction of the oocyst number in midgut while the 22E8 fungal extract did not have significant effects on P. falciparum infection in mosquitoes when the concentration of the fungal extracts was decreased to 20 &#956;g/mL. N: the number of mosquitoes for each treatment; mean: the average number of oocysts per midgut; PR: infection prevalence in mosquitoes. p: the p-value was calculated by the Mann-Whitney-Wilcoxon test. The experiments were repeated three times.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 4 Anti</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,70.87,305.48,672.95' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>) (Op De Beeck, Lievens et al. 2014) using</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>). The</ns0:cell></ns0:row></ns0:table><ns0:note>in Kenya, the Lijiang National Park, and the Potatso National Park in China. A small portion (2.1%) was from the cold areas, such as Jiuzhaigou National Park in China, and Alaska in the</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 (on next page)</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>PCR Primers for fungal ITS regions</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:note></ns0:figure> <ns0:note place='foot'>Manuscript to be reviewed</ns0:note> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:05:49300:2:1:NEW 24 Oct 2020)</ns0:note> </ns0:body> "

"October 23, 2020 Dear Dr. Gomez, We thank you and the reviewer for your time and comments. We have edited the manuscript to address the minor issues. In particular, the font of “sp” in fungal species were corrected to normal. The original articles where published those ITS primer sequences was correctly cited. The method to construct the phylogenetic tree was described in more detail and the phylogenetic tree was reconstructed. All issues raised by reviewers have been addressed in this revision. Sincerely, Dr. Jun Li Associate Professor Tel: 305-348-7618 Email: [email protected] On behalf of all authors DEPARTMENT OF BIOLOGICAL SCIENCES 11200 S.W. 8th St. Modesto Maidique Campus, OE-167. Miami, FL 33199. Tel: 305-348-2201. Reviewer 1 (Anonymous) Comments for the Author The authors have addressed most of my comments, but some very minor issues remain. 1. 'sp.' should not be italicized throughout. Response: We have changed all italicized “sp” normal “sp”, including in manuscript text (line 279), table 3, and Supplemental Figure 1. 2. Please use the original citations for the ITS primers utilized. Response: We replaced the citation of Schoch Seifert et al 2012 with the correct citation of Op De Beeck, et al 2014 that listed the actual primers (line 125). 3. The authors have oversimplified the making of the phylogenetic tree. Please elaborate a bit. Response: We added the websites, citation, and step by step procedure with parameters in detail (line 212-219). DEPARTMENT OF BIOLOGICAL SCIENCES 11200 S.W. 8th St. Modesto Maidique Campus, OE-167. Miami, FL 33199. Tel: 305-348-2201. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Thiolases are important enzymes involved in lipid metabolism in both prokaryotes and eukaryotes, and are essential for a range of metabolic pathways, while, little is known for this important family in insects. To shed light on the evolutionary models and functional diversities of the thiolase family, 137 thiolase genes were identified in 20 representative insect genomes. They were mainly classified into five classes, namely cytosolic thiolase (CT-thiolase), T1-thiolase, T2-thiolase, trifunctional enzyme thiolase (TFE-thiolase), and sterol carrier protein 2 thiolase (SCP2-thiolase). The intron number and exon/intron structures of the thiolase genes reserve large diversification. Subcellular localization prediction indicated that all the thiolase proteins were mitochondrial, cytosolic, or peroxisomal enzymes. Four highly conserved sequence fingerprints were found in the insect thiolase proteins, including CxS-, NEAF-, GHP-, and CxGGGxG-motifs. Homology modeling indicated that insect thiolases share similar 3D structures with mammals, fishes, and microorganisms. In Bombyx mori, microarray data and reverse transcriptionpolymerase chain reaction (RT-PCR) analysis suggested that some thiolases were involved in steroid metabolism, juvenile hormone (JH), and sex pheromone biosynthesis pathways.</ns0:p><ns0:p>In general, sequence and structural characteristics were relatively conserved among insects, bacteria and vertebrates, while different classes of thiolases might have differentiation in specific functions and physiological processes. These results will provide an important foundation for future functional validation of insect thiolases.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Thiolases are ubiquitous enzymes that play important roles in lipid-metabolizing pathways <ns0:ref type='bibr' target='#b20'>(Thompson et al. 1989;</ns0:ref><ns0:ref type='bibr'>Igual et al. 1992;</ns0:ref><ns0:ref type='bibr' target='#b13'>Pereto, Lopez-Garcia &amp; Moreira 2005)</ns0:ref>. There are two major kinds of thiolases based on the direction of the catalytic reaction <ns0:ref type='bibr' target='#b7'>(Masamune et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b11'>Modis &amp; Wierenga 2000)</ns0:ref>. One is degradative thiolase I (3-ketoacyl-CoA thiolase, E.C. 2.3.1.16), which catalyzes the thiolytic cleavage of medium-to long-chain unbranched 3-oxoacyl-CoAs (from 4 to 22 carbons) into acetyl-CoA and a fatty acyl-CoA (Fig. <ns0:ref type='figure' target='#fig_9'>S1A</ns0:ref>) <ns0:ref type='bibr'>(Clinkenbeard et al. 1973</ns0:ref><ns0:ref type='bibr' target='#b14'>, Schiedl et al. 2004</ns0:ref><ns0:ref type='bibr'>, Houten &amp; Wanders 2010)</ns0:ref>. It is mainly involved in fatty acid &#946;-oxidation and preferentially catalyzes the last step. The other is biosynthetic thiolase II (acetoacetyl-CoA thiolase, EC2.3.1.9), which is capable of catalyzing the Claisen condensation reaction of two molecules of acetyl-CoA to acetoacetyl-CoA (Fig. <ns0:ref type='figure' target='#fig_9'>S1B</ns0:ref>). Thiolase II might be involved in poly beta-hydroxybutyric acid synthesis, steroid biogenesis, etc. <ns0:ref type='bibr'>(Clinkenbeard et al. 1973)</ns0:ref>.</ns0:p><ns0:p>Based on the function, oligomeric state, substrate specificity, and subcellular localization, six different classes of thiolases (CT, AB, SCP2, T2, T1, and TFE) have been identified in humans <ns0:ref type='bibr'>(Fukao 2002;</ns0:ref><ns0:ref type='bibr' target='#b9'>Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. In trypanosomatid and bacterial kingdoms, another four classes were identified, including thiolase-like protein (TLP), SCP2thiolase-like protein (SLP), unclassified thiolase (UCT), and TFE-like thiolase (TFEL) <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. The CT-thiolase, located in the cytosol, has a key role in catalyzing the condensation of two molecules of acetyl-CoA to acetoacetyl-CoA, which is the first reaction of the metabolic pathway leading to the synthesis of cholesterol <ns0:ref type='bibr' target='#b6'>(Kursula et al. 2005)</ns0:ref>.</ns0:p><ns0:p>AB-thiolase and SCP2-thiolase as degradative thiolases occur in peroxisomes <ns0:ref type='bibr'>(Antonenkov et al. 1997;</ns0:ref><ns0:ref type='bibr'>Antonenkov et al. 1999)</ns0:ref>. T1-, T2-, and TFE-thiolases are mitochondrial degradative enzymes. Except for the degradation of acetoacetyl-CoA and 2-methyl-acetoacetyl-CoA, T2thiolase has a biosynthetic function in the synthesis of acetoacetyl-CoA in ketone body metabolism Manuscript to be reviewed <ns0:ref type='bibr'>(Fukao et al. 1997)</ns0:ref>. In general, AB-, SCP2-, T1-, and TFE-thiolases belong to degradative thiolase I, and CT class is biosynthetic thiolase II. Importantly, T2-thiolase is a bi-functional enzyme with synthesis and degradation activity.</ns0:p><ns0:p>As enzymes responsible for broad pathways, thiolases have been performed the phylogenetic analysis in mycobacteria and functional studies in humans <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. In insects, several thiolase genes have been characterized <ns0:ref type='bibr'>(Fujii et al. 2010)</ns0:ref>. In Helicoverpa armigera, an acetoacetyl-CoA thiolase was cloned and performed functional analysis <ns0:ref type='bibr'>(Zhang et al. 2017)</ns0:ref>. It was indicated that the thiolase was involved in the early step of the juvenile hormone pathway, i.e. mevalonate biosynthesis. One acetoacetyl-CoA thiolase was purified to apparent homogeneity by column chromatography in Bombus terrestris, suggesting that it might be the first enzyme in the biosynthesis of terpenic sex pheromone <ns0:ref type='bibr'>(Brabcova et al. 2015)</ns0:ref>. Besides, acetyl-CoA can be used as the precursor for de novo biosynthesis of sex pheromones in female moths, and four 3-ketoacyl-CoA thiolase genes were identified in sex pheromone gland transcriptome of the noctuid moth Heliothis virescens <ns0:ref type='bibr' target='#b21'>(Vogel et al. 2010)</ns0:ref>. In total, there are few and scattered studies on insect thiolase, lacking systematic identification and comparative studies.</ns0:p><ns0:p>In this study, we selected 20 representative species from 7 insect orders to perform genomewide identification of the thiolase family proteins. Gene structure, chromosome location, and three-dimensional (3D) structure and motif characteristics of proteins were compared. In addition, Bombyx mori is an important model species for studying juvenile hormone and sex pheromone biosynthesis <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010;</ns0:ref><ns0:ref type='bibr' target='#b26'>Xia, Li &amp; Feng 2014)</ns0:ref>. Expression profiles of the thiolase genes </ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data resources</ns0:head><ns0:p>In this study, 20 representative species were selected from Lepidoptera, Hymenoptera, Hemiptera, Diptera, Coleoptera, Phthiraptera, and Isoptera (Table <ns0:ref type='table'>1</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Identification of insect thiolase genes</ns0:head><ns0:p>The known thiolase sequences of Homo sapiens and Mycobacterium tuberculosis were retrieved from GenBank (Table <ns0:ref type='table'>S1</ns0:ref>; <ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref> and used as queries to perform BLASTP search (E-value &lt; 0.01) against the protein database of predicted genes in each species. Hidden Markov Model (HMM) files of Thionlase_N (PF00108) and Thionlase_C (PF02803) domains were downloaded from Pfam database (http://pfam.xfam.org/), which were used to screen the protein database of each species with hmmsearch in HMMER 3.0 (E-value &lt; 0.01). Based on BLASTP and hmmsearch analyses, the candidate thiolase genes were identified and subsequently checked by conserved domain search (CD-Search) in NCBI and hmmscan against Pfam database (E-value &lt; 1e-5). The candidate sequences have Thionlase_N and/or Thionlase_C domains were recogized as thiolases. Those identified thiolases were used as new queries to perform BLASTP search against the protein database of each species until no more novel loci can be found. All the validated thiolase genes were used for further analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Phylogenetic analysis</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The protein sequences of thiolases from 20 insects, H. sapiens and M. tuberculosis were aligned using MUSCLE <ns0:ref type='bibr'>(Edgar 2004)</ns0:ref>. Positions that had a high percentage of gaps (&gt;70%) were trimmed.</ns0:p><ns0:p>The handled alignment of protein sequences was used for checking the most suitable model of </ns0:p></ns0:div> <ns0:div><ns0:head>Molecular modeling of protein structure</ns0:head><ns0:p>The three-dimensional (3D) structure prediction of insect thiolases was conducted using the homology modeling method. Structures of T1-, T2-, CT-, AB-, TFE-, and SCP2-thiolases were predicted on-line at the SWISS-MODEL Interactive Workspace <ns0:ref type='bibr'>(Arnold et al. 2006)</ns0:ref>. The known protein that has the highest sequence similarity to the thiolase to be analyzed is used for homology modeling. The predicted models of monomer and multimer were visualized in Swiss-PdbViewer 4.1.0 (Guex &amp; Peitsch 1997). To understand the 3D structural similarities among the insect thiolases, all the other structures were compared with BmorT2 using magic fit algorithm in Swiss-PdbViewer, respectively. The root-mean-square distance (RMSD) values were calculated to express the structural similarity. The lower value of RMSD means higher similarity between two structures <ns0:ref type='bibr' target='#b12'>(Carugo &amp; Pongor 2001)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Reverse transcription-polymerase chain reaction (RT-PCR)</ns0:head><ns0:p>The various tissues on day 3 of fifth-instar larvae were dissected in the silkworm. The sex pheromone glands (PGs) from 5 individuals were used as one sample at each developmental stage.</ns0:p><ns0:p>All the samples were preserved in RNAlater (Ambion, 98 Austin, USA) and stored at -80 &#176;C for RNA isolation. Total RNA was extracted using Trizol reagent (Invitrogen, USA). The first strand of cDNA was synthesized by M-MLV reverse transcriptase following the manufacturer's instructions (Promega, USA). RT-PCR primers were listed in Table <ns0:ref type='table'>S2</ns0:ref>. The silkworm RpL3 gene was used as an internal control for relative quantitative analysis of RT-PCR. PCRs were performed with the following cycling parameters: 95 &#176;C for 3 minutes (min), followed by 25 cycles of 30 seconds (s) at 95 &#176;C, 30 s annealing (temperatures listed in Table <ns0:ref type='table'>S2</ns0:ref>) and 30 s extension (72 &#176;C), and a final extension at 72 &#176;C for 10 min. The amplification products were monitored on 1.5% agarose gels.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Genome-wide identification and phylogeny of insect thiolase proteins</ns0:head><ns0:p>To identify thiolases in insects, human and M. tuberculosis thiolase protein sequences were used as queries to perform homologous searches in whole genomes. In total, 137 thiolase genes were identified in 20 insects from 7 orders, and the gene numbers of the species were ranged from 4 to 15 (Table <ns0:ref type='table'>1</ns0:ref>; Table <ns0:ref type='table'>S1</ns0:ref>). The thiolase protein sequences were used to reconstruct the maximumlikelihood phylogenetic tree (Fig. <ns0:ref type='figure' target='#fig_9'>1</ns0:ref>). Based on the nomenclature rules in humans and M.</ns0:p><ns0:p>tuberculosis <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014)</ns0:ref>, each thiolase was named in insects. It was indicated that insect thiolases were grouped 7 classes, namely CT, T1, T2, TFE, SCP2 (type-1), TFEL (type-2), and AB (Table <ns0:ref type='table'>1</ns0:ref>; Fig. <ns0:ref type='figure' target='#fig_9'>1</ns0:ref>). Relatively, the classification of insect thiolases was more similar to that of the human than M. tuberculosis. Unlike humans, out of 20 insect species, only P. xuthus has gene members in AB class. Interestingly, 5 out of 6 Lepidopteran insects have no CT-thiolase.</ns0:p><ns0:p>Furthermore, TFEL (type-2) class was only detected in C. lectularius (Hemipter) and bacterium</ns0:p></ns0:div> <ns0:div><ns0:head>M. tuberculosis.</ns0:head><ns0:p>To understand the evolutionary mode, gene gain and loss of thiolases were analyzed. It was indicated that most of the gain and loss events were occurred in a certain species (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>Especially, P. xylostella (Lepidoptera), A. pisum (Hemipter), and P. xuthus (Lepidoptera) showed more duplications after or during the formation of the species, resulting in a total number of 15, 12, and 10 genes, respectively. Except for the gene duplication of a single species lineage, the common ancestor of D. alloeum, A. mellifera and B. impatiens showed 2 duplications (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>This phenomenon was also noted in the clade of Hemipteran H. halys, C. lectularius, and A. pisum.</ns0:p><ns0:p>For those recent duplication genes, they were often phylogenetically closely related to its ancestral genes (Fig. <ns0:ref type='figure' target='#fig_9'>1</ns0:ref>). Conversely, A. mellifera (Hymenoptera), H. halys (Hemipter), and Z. nevadensis (Isoptera) presented 3, 3, and 2 gene losses during speciation, resulting in fewer genes in these species. Generally, gene gain and loss rates are important for understanding the role of natural selection and adaptation in shaping gene family sizes. For the species with more gene expansion, whether these duplicated genes play roles in adapting to special habitats deserves further study.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene structures of insect thiolase genes</ns0:head><ns0:p>A comparative analysis of exon-intron structures was conducted for the 137 insect thiolase genes (Fig. <ns0:ref type='figure' target='#fig_11'>3A</ns0:ref> and 3B; Fig. <ns0:ref type='figure'>S2A</ns0:ref>). The insect thiolase genes have a different number of introns ranging from 0 to 22. It was indicated that only 12 genes have no intron, and 17 genes have only one intron, accounting for 21.17% in total. The intronless genes were distributed in T2 (5), SCP2 (type-1) (2), CT (2), AB (2), and TFEL (1). Previous studies revealed that introns can delay regulatory response and are selected against in genes whose transcripts need to be adjusted quickly to meet environmental challenges <ns0:ref type='bibr' target='#b2'>(Jeffares, Penkett &amp; Bahler 2008)</ns0:ref>. The intronless thiolase genes and the genes contained fewer introns might play important roles in survival for environmental changes.</ns0:p><ns0:p>In addition, the intron number and exon/intron structures of thiolase genes are very different, even the orthologous genes of different species in the same class have a large differentiation. It was suggested that the differentiation of the intron number may result in the diversification of thiolase gene structures in insects.</ns0:p></ns0:div> <ns0:div><ns0:head>Chromosome distribution and gene synteny</ns0:head><ns0:p>In order to explore the chromosomal distribution of thiolase genes, five representative species were analyzed. It was indicated that most of the thiolase genes were randomly distributed on different chromosomes (Fig. <ns0:ref type='figure' target='#fig_11'>S3A-E</ns0:ref>), for example, 6 thiolase genes were scattered on 4 chromosomes in D.</ns0:p><ns0:p>melanogaster (Fig. <ns0:ref type='figure' target='#fig_11'>S3A</ns0:ref>), which is similar to thiolase genes in the human <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014</ns0:ref>).</ns0:p><ns0:p>However, different members of a certain class are often tandem distribution, such as DmelSCP2 (type-1)-1 and DmelSCP2 (type-1)-2 in D. melanogaster (Fig. <ns0:ref type='figure' target='#fig_11'>S3A</ns0:ref>) and BmorT1-1, BmorT1-2, and BmorT1-3 in B. mori (Fig. <ns0:ref type='figure' target='#fig_11'>S3C</ns0:ref>). Meanwhile, we also detected the distribution of the 10 T1thiolase genes in P. xylostella and 7 CT-thiolase genes in A. pisum, which were distributed on several small unassembled scaffolds. Whether they are distributed in tandem, we still need to wait Manuscript to be reviewed for the scaffold sequences to be integrated into the corresponding chromosome in future. In general, tandem duplication might be the main mechanism for enlarging thiolase family in insects.</ns0:p><ns0:p>The syntenic relationships of thiolase genes were investigated among B. mori, H. melpomene, D. melanogaster, T. castaneum, and A. mellifera, which their genome sequences have been assembled into chromosome levels. The results indicated that only four genes exhibited the syntenic relationships between B. mori and H. melpomene, that is, BmorT2 and HmelT2, BmorTFE</ns0:p><ns0:p>and HmelTFE (Fig. <ns0:ref type='figure' target='#fig_11'>S3F</ns0:ref>). Interestingly, except for the tandemly duplicated genes, amount of thiolase genes often present orthologous relationships among insect, human, and M. tuberculosis (Fig. <ns0:ref type='figure' target='#fig_9'>1</ns0:ref>). It was suggested that thiolase family is an ancient gene family. Even in insects, the age of thiolase gene differentiation is relatively long. Thus, the discovery of fewer syntenic genes implies that thiolases might mainly locate in some non-conserved genomic blocks (The Heliconius Genome Consortium 2012).</ns0:p></ns0:div> <ns0:div><ns0:head>Subcellular localization of thiolase proteins</ns0:head><ns0:p>Subcellular localization refers to the specific location of a certain protein or the expression product and cytosolic localization was related to the biosynthesis of acetoacetyl-CoA <ns0:ref type='bibr' target='#b6'>(Kursula et al. 2005)</ns0:ref>.</ns0:p><ns0:p>However, in a certain class of thiolase, there are always a few exceptions to the cellular location in some species (Fig. <ns0:ref type='figure' target='#fig_11'>3C</ns0:ref>; Fig. <ns0:ref type='figure'>S2B</ns0:ref>), which suggested that its function might have diverged during evolution.</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved domain characteristics and catalytic residues</ns0:head><ns0:p>To identify the potential domains of insect thiolase proteins (Table <ns0:ref type='table'>S1</ns0:ref>), it was performed hmmscan analysis in Pfam database. The results indicated that all the thiolases contained Thiolase_N and Thiolase_C domains (Fig. <ns0:ref type='figure'>4A</ns0:ref>). In addition to the thiolase domains, SCP2-thioloase (type-1) has a</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed typical sterol carrier protein 2 (SCP2) domain at C terminal. Unexceptly, some of the members in TFE, CT, and T2 classes contained a ketoacyl-synt (beta-ketoacyl synthase) domain within Thiolase_N (Table <ns0:ref type='table'>S1</ns0:ref>). We carefully checked the alignments of hmmscan search. It was found that the E value was around the threshold 1e-5, and only about 50 amino acids can be aligned, which are much shorter than 250 amino acids of the ketoacyl-synt domain (Pfam ID, PF00109).</ns0:p><ns0:p>Thus, thiolases may not contain the real ketoacyl-synt domain, and just show certain similarities with it <ns0:ref type='bibr'>(Huang et al. 1998)</ns0:ref>. Therefore, based on the domain characteristics, all the insect thiolase encoding genes were classified as 2 groups (Fig. <ns0:ref type='figure'>4A</ns0:ref>).</ns0:p><ns0:p>The conserved sequence blocks of representative 4 insects, humans, and M. tuberculosis were analyzed (Fig. <ns0:ref type='figure'>4B</ns0:ref>). The CxS-motif is the most important sequence fingerprint in the N-terminal domain, which provides the nucleophilic cysteine <ns0:ref type='bibr'>(Zeng &amp; Li 2004;</ns0:ref><ns0:ref type='bibr' target='#b9'>Mazet et al. 2011)</ns0:ref>. Except for some incomplete sequences, all the thiolases contained the cysteine residue (Fig. <ns0:ref type='figure'>4B</ns0:ref>). The histidine of the GHP-motif contributes to the oxyanion hole of the thioester oxygen <ns0:ref type='bibr' target='#b10'>(Merilainen et al. 2009)</ns0:ref>.</ns0:p><ns0:p>It was indicated that only PxylT1-1 and PxylT1-2 lose this motif. The cysteine of CxGGGxGmotif provides the catalytic residue of the active sites. Only a few duplicated genes of T1 class were missing the catalytic cysteine in P. xylostella and B. mori. In addition, the asparagine side chain of the NEAF-motif interacts with important catalytic water <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011</ns0:ref>). However, NEAF-motif was replaced by HDCF-motif in the SCP2-thiolases of all the insect species (Fig. <ns0:ref type='figure'>4B</ns0:ref>). Based on the comparison of the sequence fingerprints, it was indicated that the catalytic mechanisms of the insect thiolases might be similar to that of thiolases from mammals and bacteria.</ns0:p></ns0:div> <ns0:div><ns0:head>Molecular modeling of insect thiolases</ns0:head><ns0:p>In recent years, the crystal structures of some thiolases have been gradually resolved in bacteria, fish, and mammals <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b4'>Kim et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019</ns0:ref>). The higher sequence similarities (&gt;60%) may help to build more accurate 3D structures for the insect thiolases <ns0:ref type='bibr'>(Arnold et al. 2006)</ns0:ref>. Based on homology modeling using SWISS-MODEL Interactive Workspace, we found that thiolase sequences within a class were very conserved among different organisms. For instance, BmorTFE-thiolase and BmorSCP2-thiolase (type-1) shared 67.58% and 61.63 % identities with its corresponding modeling templates from human (PDB ID: 6dv2.1.A) and zebrafish (6hrv.2.A), respectively. In this study, the modeling structures of some representative thiolases were presented (Fig. <ns0:ref type='figure'>5A-F</ns0:ref>). The structural similarities of the monomeric forms were Manuscript to be reviewed RMSD values of BmorT1-1, DmelCT, PxutAB-1, BmorTFE, and BmorSCP2 (type-1) were 0.93 &#197;, 0.91 &#197;, 0.95 &#197;, 1.19 &#197;, and 1.39 &#197;, respectively. It was indicated that T1, T2, CT, and AB classes share more similar 3D structures than TFE-thiolase and SCP2-thiolase (type-1) (Fig. <ns0:ref type='figure'>5A-F</ns0:ref>). This phenomenon is widespread in both humans and M. tuberculosis. Generally, T1-, T2-, and CT-thiolases were homo-tetramers (dimer of dimers) (Fig. <ns0:ref type='figure'>5G</ns0:ref>), while, SCP2 (type-1) and ABthiolases were homo-dimers (Fig. <ns0:ref type='figure'>5H</ns0:ref>). Especially, TFE-thiolase can also make a tightly bound homo-dimer. It can be used as the &#946;-subunits of mitochondrial trifunctional enzyme complex, which is composed of an &#945;2&#946;2 heterotetramer as a biological unit <ns0:ref type='bibr' target='#b24'>(Xia et al. 2019)</ns0:ref>. The results of 3D structural modeling showed that different classes of thiolase genes still present some extent divergence in tertiary or quaternary structures.</ns0:p><ns0:p>SCP2-thiolase (type-1) was widely distributed in insects, mammals, and bacteria (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>One single structural gene referred to as the SCPx encodes a full-length protein comprised of 3oxoacyl-CoA thiolase (known as SCP2-thiolase) and sterol carrier protein 2 <ns0:ref type='bibr' target='#b16'>(Seedorf et al. 1994;</ns0:ref><ns0:ref type='bibr'>Gallegos et al. 2001)</ns0:ref>. The C-terminal SCP2-domain containing the peroxisomal targeting signal is needed for the targeting of full-length SCPx into the peroxisomes. The SCP2-thiolase and SCP2</ns0:p><ns0:p>protein are produced from SCPx via proteolytic cleavage by peroxisomal proteases <ns0:ref type='bibr' target='#b16'>(Seedorf et al. 1994)</ns0:ref>. Based on the homology modeling, the tertiary and quaternary structures of mature SCP2thiolase (type-1) protein were presented in Fig. <ns0:ref type='figure'>5F</ns0:ref> and 5H, respectively. For insect SCP2-thiolases, the canonical CxGGGxG-motif is also absent, and the NEAF-motif has been replaced by HDCFmotif (Fig. <ns0:ref type='figure'>4B</ns0:ref>). The previous studies indicated that HDCF-motif might provide the catalytic cysteine in bacteria, mammals, and fish <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b3'>Kiema et al. 2019)</ns0:ref>. Based on the structural modeling, the cysteine of HDCF-motif is very close to the other two catalytic sites in protein spatial conformation (Fig. <ns0:ref type='figure'>5F</ns0:ref>). Therefore, the catalytic cysteine of the insect SCP2thiolases might be not provided by CxGGGxG-motif but HDCF-motif.</ns0:p></ns0:div> <ns0:div><ns0:head>Expression profile and potential functional diversity</ns0:head><ns0:p>To understand the potential functional diversity of the insect thiolases, the silkworm, B. mori, was used as a model organism to perform expression profile analysis in the various tissues and sex pheromone glands (PGs) at different developmental stages. In the silkworm, genome-wide microarray with 22,987 oligonucleotides was designed and surveyed the gene expression profiles in multiple tissues on day 3 of the fifth-instar larvae <ns0:ref type='bibr' target='#b25'>(Xia et al. 2007)</ns0:ref>. 5 out of 6 thiolase genes were found its corresponding probes (Fig. <ns0:ref type='figure' target='#fig_14'>6A</ns0:ref>). The microarray data indicated that BmorSCP2</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Relatively, BmorT2 and BmorT1-1 showed ubiquitous expressions. Meanwhile, the expression profiles of the four genes were similar between females and males, respectively.</ns0:p><ns0:p>To validate the expression profiles of the silkworm thiolase genes, the mixed male and female tissues were used to perform RT-PCR validation on day 3 of the fifth-instar larvae (Fig. <ns0:ref type='figure' target='#fig_14'>6B</ns0:ref>). In total, 5 out of the 6 thiolase genes presented expression evidence. Relatively, BmorT1-2</ns0:p><ns0:p>and BmorTFE showed predominant expressions in head and hemocyte, respectively (Fig. <ns0:ref type='figure' target='#fig_14'>6B</ns0:ref>).</ns0:p><ns0:p>While, BmorT1-1 was widely expressed in various tissues. In addition, sex pheromone glands of different developmental stages were used to detect the expressions of thiolase genes (Fig. <ns0:ref type='figure' target='#fig_14'>6C</ns0:ref>).</ns0:p><ns0:p>Four thiolase genes presented expression signals in the silkworm PGs. Relatively, BmorT1-1</ns0:p><ns0:p>showed the highest expression on day 8 of pupae. BmorTFE, BmorT2, and BmorSCP2 (type-1)</ns0:p><ns0:p>presented expressions at all the developmental stages. Interestingly, the expression levels of all three genes were declined in the mated female PGs (Fig. <ns0:ref type='figure' target='#fig_14'>6C</ns0:ref>). These expression analyses might help us understand the functional divergence of the thiolase genes in the silkworm.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Thiolases are widely distributed in all organisms and are essential for a range of metabolic pathways. With the development of sequencing technology, it provides the possibility for us to identify and compare insect thiolase at the whole genome level. In this study, 137 thiolase genes were identified in the 20 representative species from 7 insect orders (Table <ns0:ref type='table'>1</ns0:ref>; Table <ns0:ref type='table'>S1</ns0:ref>). The insect thiolases were mainly classified into five classes, including CT-thiolase, T1-thiolase, T2-thiolase, TFE-thiolase, and SCP2-thiolase. It was indicated that P. xylostella, A. pisum, and P. xuthus showed more duplications, resulting in a total number of 15, 12, and 10 genes, respectively. Z.</ns0:p><ns0:p>nevadensis and H. melpomene have the least number of genes (Table <ns0:ref type='table'>1</ns0:ref>). In addition to a certain differentiation in the number of genes, Thiolase_N or Thiolase_C domains of 9 thiolase genes were missing (Table <ns0:ref type='table'>S1</ns0:ref>). It is worth noting that the quality of the genome may have a certain impact on the number of genes and the integrity of gene structures. Whether the incomplete thiolase genes were pseudogenes or not (Table <ns0:ref type='table'>S1</ns0:ref>) needs further verification by the high-quality genome in the future. <ns0:ref type='bibr'>et al. 2006)</ns0:ref>. It presented expressions in the midgut, fat body, and head on day 3 of the fifth-instar larvae in the silkworm (Fig. <ns0:ref type='figure' target='#fig_14'>6B</ns0:ref>), which was suggested that the SCP2 protein might have a similar function with that of S. litura <ns0:ref type='bibr'>(Guo et al. 2009)</ns0:ref>. More important, SCP2-thiolase (type-1) is also encoded by a part of a fusion gene encoding SCPx. In vertebrates, SCP2-thiolase (type-1) plays a crucial role in the oxidation of the branched side chain of cholesterol to form bile acids <ns0:ref type='bibr'>(Ferdinandusse et al. 2000)</ns0:ref>, while the physiological role of SCP2-thiolase (type-1) has not been characterized in insects. Fortunately, the expression of the SCP2-thiolase (type-1) has also been detected in the prothoracic glands of Spodopera littoralis, which are the main tissue producing the insect molting hormone <ns0:ref type='bibr' target='#b18'>(Takeuchi et al. 2004</ns0:ref>). Thus, whether SCP2-thiolase (type-1) of the silkworm and other insects play role in the oxidation of cholesterol and participates in ecdysone synthesis needs further study.</ns0:p><ns0:p>In insects, several thiolases have been cloned and suggested to be related to juvenile hormone (JH) <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhu et al. 2016;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. Juvenile hormone is an important regulator for insect growth and development <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007)</ns0:ref>. Acetoacetyl-CoA thiolase catalyzes two molecules of acetyl-CoA to form acetoacetyl-CoA, which is the first enzyme in JH biosynthesis <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. The candidate acetoacetyl-CoA thiolases related to JH biosynthesis were cloned in B. mori and Helicoverpa armigera <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. In this study, those two acetoacetyl-CoA thiolase genes were classified as T2thiolases (BmorT2 and HarmT2), and they shared higher sequence identities with the other T2thiolases (Table <ns0:ref type='table'>S3</ns0:ref>). For example, BmorT2-thiolase shared 82.71% sequence identity with</ns0:p><ns0:p>HarmT2. In H. armigera, temporal expressions of HarmT2-thiolase keep pace with JH fluctuations, and its expression can be inhibited by a juvenile hormone analog <ns0:ref type='bibr'>(Zhang et al. 2017)</ns0:ref>.</ns0:p><ns0:p>The expression of BmorT2-thiolase was relatively abundant in the head where the JH synthetic Manuscript to be reviewed gland, corpora allata (CA), is located (Fig. <ns0:ref type='figure' target='#fig_14'>6A and 6B</ns0:ref>). Thus, insect T2-thiolases may be involved in the JH synthesis pathway. Interestingly, we found BmorTFE-and BmorT1-1-thiolase also showed high expressions in the larval head (Fig. <ns0:ref type='figure' target='#fig_14'>6A</ns0:ref>). In humans, TFE-and T1-thiolases catalyze thiolytic cleavage of 3-ketoacyl-CoA into acetyl-CoA and acyl-CoA <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. However, T1-thiolase has been found synthetic and degradative activities in Ostrinia scapulalis (Lepidoptera: Crambidae). Therefore, except for T2-thiolase, whether TFE-and T1thiolases were also involved in JH biosynthesis is still worthy of validation.</ns0:p><ns0:p>Acetyl-CoA is often used as the initial precursor for sex pheromone biosynthesis in insects <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010)</ns0:ref>. Degradative thiolases may supplement with sufficient acetyl-CoA for sex pheromone synthesis <ns0:ref type='bibr'>(Brabcova et al. 2015)</ns0:ref>. In this study, expression profiles of the thiolase genes were detected in the sex pheromone glands at different developmental stages in the silkworm (Fig. <ns0:ref type='figure' target='#fig_14'>6C</ns0:ref>). Relatively, BmorSCP2 (type-1) maintains a high level of expression in the PGs on day 4 of pupae to 24-h-old virgin female moth. However, its expression level was sharply declined in the mated female PGs (Fig. <ns0:ref type='figure' target='#fig_14'>6C</ns0:ref>). The previous study suggested that an over 6-h mating duration can terminate the sex pheromone production in the silkworm <ns0:ref type='bibr' target='#b1'>(Ando et al. 1996)</ns0:ref>. The expression pattern of BmorSCP2 (type-1) was consistent with sex pheromone production <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010)</ns0:ref>, which suggested that it might be involved in sex pheromone biosynthesis. Simultaneously, BmorTFE and BmorT2 showed similar expression patterns as BmorSCP2 (type-1). Generally, it is tempting to assume that a thiolase expressed in a specific tissue might obtain a specific role. Thus, the functional diversification and physiological roles of insect thiolases need yet further experimental validation.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In the present study, genome-wide identification of the thiolase gene family was conducted for the first time in multiple insect genomes. A total of 137 thiolase genes were identified in 20 insects from 7 orders. About 80% of the thiolase genes have 2 or more introns, and its exon/intron structures reserve diversification. Based on the prediction, all the thiolase proteins are located in the mitochondria, cytosol, or peroxisome, and thiolases of the same class often have similar cellular localization. Four highly conserved sequence fingerprints were found in the insect thiolase proteins, including CxS-, NEAF-, GHP-, and CxGGGxG-motifs. Homology modeling analysis indicated that 3D structures of the insect thiolases share similar to mammals, fishes, and microorganisms. Expression pattern analysis suggests some thiolase genes were involved in Manuscript to be reviewed steroid metabolism, JH, and sex pheromone biosynthesis pathways in B. mori. These results might provide valuable information for the functional exploration of thiolase proteins in insects. The species tree was obtained from timetree database ( http://www.timetree.org/ ). Gain and loss analysis was conducted by Notung-2.9 software with default parameters. The orange and green vertical bars on branch presented gene gain and loss, respectively. The number in each node is gene count. Gene number of each species was presented in brackets.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed (H) Homo-dimer of BmorSCP2 (type-1). </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 6</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>were detected in various tissues and developmental sex pheromone gland of B. mori. Combining structural characteristics and expression patterns, the potential functions and involved PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020) physiological processes were hypothesized. The present study can help us understand the functional differentiation of thiolase genes in insects.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>of a certain gene in the cell. Protein subcellular localization is closely related to protein functions<ns0:ref type='bibr' target='#b13'>(Pereto, Lopez-Garcia &amp; Moreira 2005;</ns0:ref><ns0:ref type='bibr' target='#b23'>Wang et al. 2014)</ns0:ref>. Only when the protein is positioned correctly can it perform normal biological functions. In this study, subcellular localization of //www.genscript.com/tools/psort), which were cytosolic, mitochondrial, or peroxisomal enzymes (Fig.3C; Fig.S2B). Generally, most of the TFE-and T2-thiolase proteins were located in the mitochondrion, T1-and CT-thiolases were cytosolic, and SCP2-thiolases were peroxisomal proteins. Previous studies suggested that the mitochondrial and peroxisomal thiolase proteins were mainly involved in the fatty acid &#946;-oxidation pathway<ns0:ref type='bibr' target='#b13'>(Pereto, Lopez-Garcia &amp; Moreira 2005),</ns0:ref> </ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>detected with magic fit in Swiss-PdbViewer(Guex &amp; Peitsch 1997). Compared with BmorT2, the PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>type-1), BmorT2, BmorTFE, and BmorT1-1 have expression signals at least one of the 9 tissues.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>Two groups of thiolases were identified in animals: 3-oxoacyl-CoA thiolase and acetoacetyl-CoA thiolase, which participates in different catabolic (fatty acid oxidation and bile acid formation) and anabolic (cholesterogenesis, ketone body synthesis, fatty acid elongation)PeerJ reviewingPDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)Manuscript to be reviewed processes(Antonenkov, Van Veldhoven &amp; Mannaerts 1999). It is well known that cholesterol is a precursor of molting hormone, 20-hydroxyecdysone (20E), and is a structural component of cell membranes(Gilbert, Rybczynski &amp; Warren 2002). Due to the lack of squalene monooxygenase and lanosterol synthase for the synthesis of cholesterol, insects can not autonomously synthesize the 20E precursor(Guo et al. 2009). Alternatively, insects can obtain cholesterol or other sterols from their diet to meet the needs of growth and development. In Spodoptera litura, SCPx showed predominant expression in the midgut, and its coding SCP2 protein was involved in the absorption and transport of cholesterol(Guo et al. 2009). In the silkworm, SCPx gene has been cloned (Gong</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:1:1:REVIEW 29 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Fig. 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig. 1. Phylogenetic tree of insect thiolases using the maximum-likelihood (ML) method.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 2 Fig. 2 .</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Fig. 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig. 3. Exon/intron structure and subcelluar localization analyses.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>Figure 4 Fig. 4 .</ns0:head><ns0:label>44</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head>Figure 5 Fig. 5 .</ns0:head><ns0:label>55</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head>Fig. 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Fig. 6. Expression profiles of the thiolase genes in the silkworm.</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>). The annotated genes and genomes of B. mori were retrieved from SilkDB v3.0 (https://silkdb.bioinfotoolkits.net). The sequence information of Danaus plexippus and Heliconius melpomene were downloaded from http://metazoa.ensembl.org/.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Manduca</ns0:cell><ns0:cell>sexta</ns0:cell><ns0:cell>was</ns0:cell><ns0:cell>from</ns0:cell></ns0:row></ns0:table><ns0:note>ftp://ftp.bioinformatics.ksu.edu/pub/Manduca/OGS2/. The other sequences were retrieved from GenBank (https://www.ncbi.nlm.nih.gov/), including Papilio xuthus, Plutella xylostella, Culex quinquefasciatus, Anopheles gambiae, Drosophila melanogaster, Anoplophora glabripennis, Nicrophorus vespilloides, Tribolium castaneum, Halyomorpha halys, Acyrthosiphon pisum, Cimex lectularius, Diachasma alloeum, Apis mellifera, Bombus impatiens, Pediculus humanus, and Zootermopsis nevadensis.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Chromosome distribution, gene structure, and syntenic analysis</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='7'>evolution by ProtTest 3.2 (Darriba et al. 2011). Maximum-likelihood (ML) trees were</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>reconstructed using RAxML version 8.2.12 (Stamatakis 2014) with the most suitable model</ns0:cell></ns0:row><ns0:row><ns0:cell>(PROTGAMMAVTF)</ns0:cell><ns0:cell>and</ns0:cell><ns0:cell>500</ns0:cell><ns0:cell>bootstrap</ns0:cell><ns0:cell>replicates.</ns0:cell><ns0:cell>FigTree</ns0:cell><ns0:cell>v1.4.3</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>(http://tree.bio.ed.ac.uk/software/figtree/) was used for plotting the final phylogenetic tree. The</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>clustering and classification of the thiolase sequences in the ML tree were done using known</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>functional properties of H. sapiens and M. tuberculosis (Anbazhagan et al. 2014).</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='7'>To localize the thiolase genes on chromosomes, B. mori, H. melpomene, D. melanogaster, T.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>castaneum, and A. mellifera were selected, which their genome sequences have been assembled</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>into chromosomes. Based on the GFF (General Feature Format) file of each species, every thiolase</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>genes was mapped to the corresponding chromosomes. Using protein sequences of the thiolases,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>the precise exon/intron structures were generated through BLAT search against the genome</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>sequences with Scipio server (https://www.webscipio.org/). The synteny events between two</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>species were detected by Multiple Collinearity Scan toolkit (MCScanX) with the default</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>parameters (Wang et al. 2012). The syntenic map of B. mori and H. melpomene was constructed</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>with family_circle_plotter.java in MCScanX software.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> </ns0:body> "

"College of Life Science China West Normal University Nanchong 637009 Sichuan China September 22th, 2020 Dear editors, First, thank you for your timely processing of my manuscript. Next, I thank the reviewers for their generous comments on the manuscript. Based on the comments and suggestions, I have carefully revised the manuscript and responded one-by one. I believe that the manuscript is now suitable for publication in PeerJ. Dr. Shou-Min Fang Associate Professor of Genomics Response Editor comments (Kenta Nakai) MAJOR REVISION Your manuscript has been reviewed by three experts in the field. As you can see from their comments below, all of them admit the value of your work; one even recommends its acceptance as it is. However, I think that manuscript should be revised following their comments by the remaining two reviewers (mainly Reviewer 2). Please read their comments carefully and revise the manuscript accordingly. Looking forward to your revised manuscript. [# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #] Response: Thank you for your timely processing of my manuscript. I have carefully revised the manuscript, and answered to all the comments one by one. Reviewer 1 Basic reporting no comment Experimental design No comment Validity of the findings No comment Comments for the author This research focused on thiolase family in insects. Thiolases plays important roles in lipid metabolism but it is poorly understood in insects. This study identifies the genes and provides key information for the future function analysis. The manuscript is generally written well. I have some comments as following: 1、Line 164, it is better to mention somewhere genome quality potentially influences the results, like the number of genes identified and the deduced evolutionary process. Response: Agreed. Genome quality may affect the gene number and evolutionary process. It was added to the last of the first paragraph of Results and Discussion section. The revision is as follows: “It is worth noting that the quality of the genome may have a certain impact on the number of genes and integrity of gene structures. Whether the incomplete thiolase genes were pseudogenes or not (Table S1) needs further verification by the high-quality genome in the future. ” 2、Line 178, the author can specify which group have no intron. Response: It was added to Line 189-190. 3、Line 181, this part reads like introns can help genes respond to enviromental challenges, which is contradicting to the first part of the sentence. Response: Thanks for your remind. I checked the reference. The sentence should be “Previous studies revealed that introns can delay regulatory response and are selected against in genes whose transcripts need to be adjusted quickly to meet environmental challenges”. Thus, the next sentence was also revised. 4、Line 182-183, the findings/results cannot support this discussion/opinion yet. Response: It was revised to “The intronless thiolase genes and the genes contained fewer introns might play imortant roles in survival for environmental changes.”. 5、Line 204, reference. Response: The related reference (The Heliconius Genome Consortium 2012) was added. 6、Line 243, the author should list protein identifies between target protein and human template for all proteins. But the author only give the information for TFE and SCP2. Response: This information was presented in Fig. 5 legend: The monomer structures of some representative genes were shown, including BmorT2 (A), DmelCT (B), PxutAB-1 (C), BmorT1-1 (D), BmorTFE (E), and BmorSCP2 (type-1) (F). The corresponding templates for homology modeling were PDB ID 6bjb.1.A, 4wyr.1.A, 1afw.1.A, 4wyr.1.A, 6dv2.1.A, and 6hrv.2.A, respectively. Reviewer 2 Basic reporting The author used clear and unambiguous English, but there are several simple grammatical and typographical errors to be corrected before publication, which are listed below. Lines 21, 311: “. While, ” => “, while ” Line 31: Italicize species name. Line 44: Fig S1A => Fig. S1A Line 62: belonged => belong Lines 124–125: Use singular form after “every “ Line 247: We => we Line 267: contained => containing Line 273: “Therefore ” => “Therefore, ” Line 338: delined => declined Response: Agreed. According to the suggestions, all the simple grammatical and typographical errors were revised. There are also several texts and words which should be improved for clarity and better understanding. Line 90: “gene models” may be “genes”? Lines 154–155: The sentence, “In total, …”, should be improved into a clearly understandable text. Line 222: “Except for” here may confuse readers. It may be better to use “In addition to” instead. Line 224: “seem to be contained” => “contained” Line 227: largely => much Line 269: “SCPx by proteolytic cleavage via peroxisomal” => “SCPx via proteolytic cleavage by peroxisomal” Line 350: and => or Response: According to the suggestions, all the mentioned texts and words were were revised. Some figures and table seem to be in insufficient quality: Fig. 1: The visibility of dots on the nodes is poor. Response: The dots were presented by red. It might be more visible. Figs. 3B and S2A: Scale bars? cannot be read and look strange. Response: Yes, they are scale bars. The font sizes of the scale bars were increased. The gene structures was produced with Scipio server (https://www.webscipio.org/). The different genes have different scales, including exons and introns. Thus, many scale bars were presented. Fig. 5: The resolution of some figures seems to be low. Numbers with dotted surfaces are shown, but hard to recognize. Response: We revised this. The catalytic amino acids and its positions were presented more clear in Fig. 5. Please check it. Fig. 6: (A) The change of the color for expression <400 into white is recommended for clarity. (B) and (C) The legends seem to be wrong texts. Response: The color for expression <400 has been changed into white. (B) Expression patterns in the various tissues of the fifth-instar larvae. (C) Expression profiles in the sex pheromone glands at different developmental stages of females. 0-h and 24-h vergin: 0-h and 24-h old vergin adults after eclosion; 3-h, 6-h and 9-h mated: female moths mated 3 hours, 6 hours and 9 hours. Table 1: Double-byte spaces appear in Iso_Z. nevadensis line and should be removed. Response: The double-byte spaces in Iso_Z. nevadensis line were removed. Titles of supplemental figures are missing and should be placed. In addition, Figs S2 and S3 may also need legends for better understanding. Response: The titles and legends have been prepared. It might be not uploaded before. Fig. S1. The catalytic processes of 3-oxoacyl-CoA thiolase and acetoacetyl-CoA thiolase. It was retrieved from https://www.uniprot.org/. Fig. S2. Analyses of exon/intron organization and subcelluar localization for the thiolase genes from the other 13 insects. (A) Exon/intron structure analysis. (B) Subcelluar localization analysis. Mit: mitochondrion; Cyt: cytosol; Pox: peroxisome. Fig. S3. Chromosomal distribution and synteny analyses of the thiolase genes. (A) to (E) Chromosomal distribution of the thiolase genes was presented for D. melanogaster, T. castaneum, B. mori, A. mellifera, H. melpomene, respectively. Chr: chromosome. (F) Synteny events of the thiolase genes between H. melpomene and B. mori. Gray lines indicate all synteny blocks for the 4 chromosomes and red lines indicate synteny relationships of the thiolase genes. Experimental design There are several concerns about details of the methods listed below. Line 91: The URL of SilkDB (http://silkworm.genomics.org.cn) cannot be accessed. Response: It was replaced by https://silkdb.bioinfotoolkits.net. Line 101, 153: The detail of the known sequences should be mentioned, probably by using Table S2. Response: The known sequences of the human and M. tuberculosis were listed in the Table S1. Line 102: The condition of BLASTP search should be written. Response: The threshold value (E-value < 0.01) was added. Lines 108–109: The detail of how did the author validate “The validated thiolases” should be clearly written. Response: One sentence was supplemented. It is as follows: The candidate sequences have Thionlase_N and/or Thionlase_C domains were recogized as thiolases. Line 123: The author should specify the species used as “three model species” here or in the result texts and explain why selected them. Response: During the revision, five species were used for analysis of chromosome distribution. Because these species’ genome sequences have been assembled into corresponding chromosomes. In the manuscript, it was revised to “B. mori, H. melpomene, D. melanogaster, T. castaneum, and A. mellifera were selected, which their genome sequences have been assembled into chromosomes.”. Line 135: The author should specify the kind of “similarity”. Is it sequence similarity? Response: It is sequence similarity. It was revised. Lines 137–138: “Using BmorT2 as the reference…”. I cannot understand where this was used. Response: This sentence was revised. It is as follows: To understand the 3D structural similarities among the insect thiolases, all the other structures were compared with BmorT2 using magic fit algorithm in Swiss-PdbViewer, respectively. The root-mean-square distance (RMSD) values were calculated to express the structural similarity. The lower value of RMSD means higher similarity between two structures (Carugo & Pongor 2001). Lines 209–210: The method for “In this study, subcellular localization of all the 137 insect thiolase proteins was predicted” was not written. Response: It was predicted by PSORT II server (https://www.genscript.com/tools/psort). This sentence was revised to “In this study, subcellular localization of all the 137 insect thiolase proteins was predicted by PSORT II server (https://www.genscript.com/tools/psort),...” Validity of the findings Several sentences in the conclusion seem to be exaggerated or over generalized and should be improved. Line 349: “randomly and unevenly distributed” has weak evidence derived from small number of samples. Response: The sentence “The thiolase genes are randomly and unevenly distributed on chromosomes.” was deleted. Lines 350–351, 354–355: The results are based on prediction. Response: “Based on the prediction” was added to the corresponding sentence. Lines 353–354: “3D structures were highly conserved between insects and other vertebrates” was not clearly shown in general. Response: It was revised to “Homology modeling analysis indicated that 3D structures of the insect thiolases share similar with mammals, fishes, and microorganisms. Comments for the author The author identified thiolase genes in 20 insects and analyzed their features about phylogeny, genes, 3D-structures, expressions of proteins, etc. Major concerns: * The analyses seem to be well done, but some of them were not comprehensively done: for example, chromosome distribution and conserved domain analysis. Response: At present, most of the insect genomes have not been assembled into chromosome levels. In this study, the genomes of B. mori, H. melpomene, D. melanogaster, T. castaneum, and A. mellifera have been assembled into chromosomes. In the previous version of the manuscript, only B. mori, T. castaneum, and D. melanogaster were presented in Fig. S3. In this revision, A. mellifera and H. melpomene were also presented in Fig. S3D and E, respectively. In this study, the conserved Thiolase-N and Thiolase-C domains were predicted for all the insect thiolases (Table S1). The conserved signature motifs in the domains were analyzed in the different species (Fig. 4). Furthermore, the alignement logos of the motifs were presented in Fig. 4. Thus, no other analysis has been done. * The manuscript needs more discussion “based on the results”, especially for the expression profile parts and should be largely improved. Response: Based on the suggestion, more discussion has been done. Please check it in the manuscript. * Some results are discussed only in insects thiolases (e.g., Lines 190–194, 254–255). The author should evaluate the results by comparing them with the known thiolases of the other organisms. Whether the results are insects-specific or not is important for better understanding the thiolases of insects. Lines 190–194: Human thiolases were also randomly distributed on different chromosomes. Thus, we supplemented a comparitive information “…., which is similar to thiolase genes in human (Anbazhagan et al. 2014).” Lines 254–255: The 3D structures of the thiolases in the humans and Mycobacterium tuberculosis was compared, respectively. It was also indicated that structures of TFE-thiolases and SCP2thiolases showed more greater differentiation. Thus, a sentence was supplemented. “This phenomenon is widespread in both humans and Mycobacterium tuberculosis.”. * Fragment sequences like only N-terminal parts are treated same as whole genes. Those sequences may be pseudogenes and should be separately treated. In addition, as the author also mentioned in the text, some of the full-length genes lack the catalytic residues important for thiolases and may lose their functions as thiolases. Those incomplete genes should be distinguished with likely genuine genes (e.g., in Table 1). Response: The incomplete genes were marked in the Table S1. Based on the present genome sequences, I can not determine whether the incomplete thiolase genes were pseudogenes or not. Thus, some sentences was added to the Discussion section. Line 414-417: It is worth noting that the quality of the genome may have a certain impact on the number of genes and integrity of gene structures. Whether the incomplete thiolase genes were pseudogenes or not (Table S1) needs further verification by the high-quality genome in the future. Minor concerns: Lines 185–187: I couldn’t understand why the sentence, “It seems that ...”, was derived. Maybe, more explanation is needed here. Response: The sentence was revised as “It was suggested that the differentiation of intron number may result in the diversification of thiolase gene structures in insects.”. Line 251: The values of sequence similarity are better to be shown. This sentence may be better to appear in line 246. Response: Agreed. The sentence as move to the previous line 246 position. It was revised as “The higher sequence similarities (>60%) may help to build more accurate 3D structures for the insect thiolases (Arnold et al. 2006).” Line 255: The values of 3D structural similarity are better to be shown. Response: The root-mean-square distance (RMSD) values were calculated to express the structural similarity. The lower value of RMSD means higher similarity between two structures (Carugo & Pongor 2001). This method was supplemented in Materials & Methods section: “To understand the 3D structural similarities among the insect thiolases, all the other structures were compared with BmorT2 using magic fit algorithm in Swiss-PdbViewer, respectively. The root-mean-square distance (RMSD) values were calculated to express the structural similarity. The lower value of RMSD means higher similarity between two structures (Carugo & Pongor 2001).” And then, one sentence was supplemented in the previous Line 255: “Compared with BmorT2, the RMSD values of BmorT1-1, DmelCT, PxutAB-1, BmorTFE, and BmorSCP2 (type-1) were 0.93 Å, 0.91 Å, 0.95 Å, 1.19 Å, and 1.39 Å, respectively.” Lines 258–260: “However, …” The meaning of this sentence is hard to understand for me. Maybe more explanation is needed. Response: It was revised as “It can be used as the β-subunits of mitochondrial trifunctional enzyme complex, ...”. Line 264: How did the author judge “this type of thiolase is the most divergent”? Response: It is a comprehensive result from gene and protein structures. The description without the following results is very difficult to understand. Thus, the sentence “It is worth noting that this type of thiolase is the most divergent member of the thiolase family.” was deleted. Lines 271–276: The order of sentences may be better to be reordered: known knowledge as first. Response: Agreed. The revison was as follows: The previous studies indicated that HDCF-motif might provide the catalytic cysteine in bacteria, mammals, and fish (Harijan et al. 2013, Kiema et al. 2019). Based on the structural modeling, the cysteine of HDCF-motif is very close to the other two catalytic sites in protein spatial conformation (Fig. 5F). Therefore, the catalytic cysteine of the insect SCP2-thiolases might be not provided by CxGGGxG-motif but HDCF-motif. Lines 324: “relatively high enriched” seems to be not appropriate phrase for this. Response: The sentence was revised to “It was indicated that the expression of BmorT2-thiolase was relatively abundant…”. Lines 325–326: “It further supports that insect T2-thiolase genes may be involved in the JH synthesis pathway.” Generalization as “insect T2-thiolase genes“ seems to be unreasonable. Response: The “insect T2-thiolase genes” was revised to BmorT2-thiolase. Reviewer 3 Basic reporting In this manuscript, the author (Shou-Min Fang) identified thiolase genes in 20 representative insect genomes using bioinformatics analysis. Insect thiolases were mainly classified into five classes (CTthiolase, T1-thiolase, T2-thiolase, TFE-thiolase, and SCP2-thiolase). Their intron numbers and exon/intron structures, and predicted subcellular localization are variable. Homology modeling indicated that insect thiolases share similar 3D structures with mammals and other vertebrates. Transcription analysis suggested that, in B. mori, some thiolases may be involved in steroid metabolism, juvenile hormone and sex pheromone biosynthesis. In general, the manuscript is well written and organized. Experimental design The manuscript is pretty well organized and the experiment is carefully designed. Identification of thiolase genes in sequenced insect genomes provide a useful resource for further functional studies. Validity of the findings no comment Response: Thanks for your affirmation and omments. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Thiolases are important enzymes involved in lipid metabolism in both prokaryotes and eukaryotes, and are essential for a range of metabolic pathways, while, little is known for this important family in insects. To shed light on the evolutionary models and functional diversities of the thiolase family, 137 thiolase genes were identified in 20 representative insect genomes. They were mainly classified into five classes, namely cytosolic thiolase (CT-thiolase), T1-thiolase, T2-thiolase, trifunctional enzyme thiolase (TFE-thiolase), and sterol carrier protein 2 thiolase (SCP2-thiolase). The intron number and exon/intron structures of the thiolase genes reserve large diversification. Subcellular localization prediction indicated that all the thiolase proteins were mitochondrial, cytosolic, or peroxisomal enzymes. Four highly conserved sequence fingerprints were found in the insect thiolase proteins, including CxS-, NEAF-, GHP-, and CxGGGxG-motifs. Homology modeling indicated that insect thiolases share similar 3D structures with mammals, fishes, and microorganisms. In Bombyx mori, microarray data and reverse transcriptionpolymerase chain reaction (RT-PCR) analysis suggested that some thiolases might be involved in steroid metabolism, juvenile hormone (JH), and sex pheromone biosynthesis pathways. In general, sequence and structural characteristics were relatively conserved among insects, bacteria and vertebrates, while different classes of thiolases might have differentiation in specific functions and physiological processes. These results will provide an important foundation for future functional validation of insect thiolases.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Thiolases are ubiquitous enzymes that play important roles in lipid-metabolizing pathways <ns0:ref type='bibr' target='#b20'>(Thompson et al. 1989;</ns0:ref><ns0:ref type='bibr'>Igual et al. 1992;</ns0:ref><ns0:ref type='bibr' target='#b13'>Pereto, Lopez-Garcia &amp; Moreira 2005)</ns0:ref>. There are two major kinds of thiolases based on the direction of the catalytic reaction <ns0:ref type='bibr' target='#b7'>(Masamune et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b11'>Modis &amp; Wierenga 2000)</ns0:ref>. One is degradative thiolase I (3-ketoacyl-CoA thiolase, E.C. 2.3.1.16), which catalyzes the thiolytic cleavage of medium-to long-chain unbranched 3-oxoacyl-CoAs (from 4 to 22 carbons) into acetyl-CoA and a fatty acyl-CoA (Fig. <ns0:ref type='figure' target='#fig_6'>S1A</ns0:ref>) <ns0:ref type='bibr'>(Clinkenbeard et al. 1973</ns0:ref><ns0:ref type='bibr' target='#b14'>, Schiedl et al. 2004</ns0:ref><ns0:ref type='bibr'>, Houten &amp; Wanders 2010)</ns0:ref>. It is mainly involved in fatty acid &#946;-oxidation and preferentially catalyzes the last step. The other is biosynthetic thiolase II (acetoacetyl-CoA thiolase, EC2.3.1.9), which is capable of catalyzing the Claisen condensation reaction of two molecules of acetyl-CoA to acetoacetyl-CoA (Fig. <ns0:ref type='figure' target='#fig_6'>S1B</ns0:ref>). Thiolase II might be involved in poly beta-hydroxybutyric acid synthesis, steroid biogenesis, etc. <ns0:ref type='bibr'>(Clinkenbeard et al. 1973)</ns0:ref>.</ns0:p><ns0:p>Based on the function, oligomeric state, substrate specificity, and subcellular localization, six different classes of thiolases (CT, AB, SCP2, T2, T1, and TFE) have been identified in humans <ns0:ref type='bibr'>(Fukao 2002;</ns0:ref><ns0:ref type='bibr' target='#b9'>Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. In trypanosomatid and bacterial kingdoms, another four classes were identified, including thiolase-like protein (TLP), SCP2thiolase-like protein (SLP), unclassified thiolase (UCT), and TFE-like thiolase (TFEL) <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. The CT-thiolase, located in the cytosol, has a key role in catalyzing the condensation of two molecules of acetyl-CoA to acetoacetyl-CoA, which is the first reaction of the metabolic pathway leading to the synthesis of cholesterol <ns0:ref type='bibr' target='#b6'>(Kursula et al. 2005)</ns0:ref>.</ns0:p><ns0:p>AB-thiolase and SCP2-thiolase as degradative thiolases occur in peroxisomes <ns0:ref type='bibr'>(Antonenkov et al. 1997;</ns0:ref><ns0:ref type='bibr'>Antonenkov et al. 1999)</ns0:ref>. T1-, T2-, and TFE-thiolases are mitochondrial degradative enzymes. Except for the degradation of acetoacetyl-CoA and 2-methyl-acetoacetyl-CoA, T2thiolase has a biosynthetic function in the synthesis of acetoacetyl-CoA in ketone body metabolism PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>(Fukao et al. 1997)</ns0:ref>. In general, AB-, SCP2-, T1-, and TFE-thiolases belong to degradative thiolase I, and CT class is biosynthetic thiolase II. Importantly, T2-thiolase is a bi-functional enzyme with synthesis and degradation activity.</ns0:p><ns0:p>As enzymes responsible for broad pathways, thiolases have been performed the phylogenetic analysis in mycobacteria and functional studies in humans <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. In insects, several thiolase genes have been characterized <ns0:ref type='bibr'>(Fujii et al. 2010)</ns0:ref>. In Helicoverpa armigera, an acetoacetyl-CoA thiolase was cloned and performed functional analysis <ns0:ref type='bibr'>(Zhang et al. 2017)</ns0:ref>. It was indicated that the thiolase was involved in the early step of the juvenile hormone pathway, i.e. mevalonate biosynthesis. One acetoacetyl-CoA thiolase was purified to apparent homogeneity by column chromatography in Bombus terrestris, suggesting that it might be the first enzyme in the biosynthesis of terpenic sex pheromone <ns0:ref type='bibr'>(Brabcova et al. 2015)</ns0:ref>. Besides, acetyl-CoA can be used as the precursor for de novo biosynthesis of sex pheromones in female moths, and four 3-ketoacyl-CoA thiolase genes were identified in sex pheromone gland transcriptome of the noctuid moth Heliothis virescens <ns0:ref type='bibr' target='#b21'>(Vogel et al. 2010)</ns0:ref>. In total, there are few and scattered studies on insect thiolase, lacking systematic identification and comparative studies.</ns0:p><ns0:p>In this study, we selected 20 representative species from 7 insect orders to perform genomewide identification of the thiolase family proteins. Gene structure, chromosome location, and three-dimensional (3D) structure and motif characteristics of proteins were compared. In addition, Bombyx mori is an important model species for studying juvenile hormone and sex pheromone biosynthesis <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010;</ns0:ref><ns0:ref type='bibr' target='#b26'>Xia, Li &amp; Feng 2014)</ns0:ref>. Expression profiles of the thiolase genes </ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data resources</ns0:head><ns0:p>In this study, 20 representative species were selected from Lepidoptera, Hymenoptera, Hemiptera, Diptera, Coleoptera, Phthiraptera, and Isoptera (Table <ns0:ref type='table'>1</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Identification of insect thiolase genes</ns0:head><ns0:p>The known thiolase sequences of Homo sapiens and Mycobacterium tuberculosis were retrieved from GenBank (Table <ns0:ref type='table'>S1</ns0:ref>; <ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref> and used as queries to perform BLASTP search (E-value &lt; 0.01) against the protein database of predicted genes in each species. Hidden Markov Model (HMM) files of Thionlase_N (PF00108) and Thionlase_C (PF02803) domains were downloaded from Pfam database (http://pfam.xfam.org/), which were used to screen the protein database of each species with hmmsearch in HMMER 3.0 (E-value &lt; 0.01). Based on BLASTP and hmmsearch analyses, the candidate thiolase genes were identified and subsequently checked by conserved domain search (CD-Search) in NCBI and hmmscan against Pfam database (E-value &lt; 1e-5). The candidate sequences that have Thiolase_N and/or Thiolase_C domains were recognized as thiolases. Those identified thiolases were used as new queries to perform BLASTP search against the protein database of each species until no more novel loci can be found. All the validated thiolase genes were used for further analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Phylogenetic analysis</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The protein sequences of thiolases from 20 insects, H. sapiens and M. tuberculosis were aligned using MUSCLE <ns0:ref type='bibr'>(Edgar 2004)</ns0:ref>. Positions that had a high percentage of gaps (&gt;70%) were trimmed.</ns0:p><ns0:p>The handled alignment of protein sequences was used for checking the most suitable model of </ns0:p></ns0:div> <ns0:div><ns0:head>Molecular modeling of protein structure</ns0:head><ns0:p>The three-dimensional (3D) structure prediction of insect thiolases was conducted using the homology modeling method. Structures of T1-, T2-, CT-, AB-, TFE-, and SCP2-thiolases were predicted on-line at the SWISS-MODEL Interactive Workspace <ns0:ref type='bibr'>(Arnold et al. 2006)</ns0:ref>. The known protein that has the highest sequence similarity to the thiolase to be analyzed is used for homology modeling. The predicted models of monomer and multimer were visualized in Swiss-PdbViewer 4.1.0 (Guex &amp; Peitsch 1997). To understand the 3D structural similarities among the insect thiolases, all the other structures were compared with BmorT2 using magic fit algorithm in Swiss-PdbViewer, respectively. The root-mean-square distance (RMSD) values were calculated to express the structural similarity. The lower value of RMSD means higher similarity between two structures <ns0:ref type='bibr' target='#b12'>(Carugo &amp; Pongor 2001)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Reverse transcription-polymerase chain reaction (RT-PCR)</ns0:head><ns0:p>The various tissues on day 3 of fifth-instar larvae were dissected in the silkworm. The sex pheromone glands (PGs) from 5 individuals were used as one sample at each developmental stage.</ns0:p><ns0:p>All the samples were preserved in RNAlater (Ambion, 98 Austin, USA) and stored at -80 &#176;C for RNA isolation. Total RNA was extracted using Trizol reagent (Invitrogen, USA). The first strand of cDNA was synthesized by M-MLV reverse transcriptase following the manufacturer's instructions (Promega, USA). RT-PCR primers were listed in Table <ns0:ref type='table'>S2</ns0:ref>. The silkworm RpL3 gene was used as an internal control for relative quantitative analysis of RT-PCR. PCRs were performed with the following cycling parameters: 95 &#176;C for 3 minutes (min), followed by 25 cycles of 30 seconds (s) at 95 &#176;C, 30 s annealing (temperatures listed in Table <ns0:ref type='table'>S2</ns0:ref>) and 30 s extension (72 &#176;C), and a final extension at 72 &#176;C for 10 min. The amplification products were monitored on 1.5% agarose gels.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Genome-wide identification and phylogeny of insect thiolase proteins</ns0:head><ns0:p>To identify thiolases in insects, human and M. tuberculosis thiolase protein sequences were used as queries to perform homologous searches in whole genomes. In total, 137 thiolase genes were identified in 20 insects from 7 orders, and the gene numbers of the species were ranged from 4 to 15 (Table <ns0:ref type='table'>1</ns0:ref>; Table <ns0:ref type='table'>S1</ns0:ref>). The thiolase protein sequences were used to reconstruct the maximumlikelihood phylogenetic tree (Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>). Based on the nomenclature rules in humans and M.</ns0:p><ns0:p>tuberculosis <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014)</ns0:ref>, each thiolase was named in insects. It was indicated that insect thiolases were grouped 7 classes, namely CT, T1, T2, TFE, SCP2 (type-1), TFEL (type-2), and AB (Table <ns0:ref type='table'>1</ns0:ref>; Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>). Relatively, the classification of insect thiolases was more similar to that of the human than M. tuberculosis. Unlike humans, out of 20 insect species, only P. xuthus has gene members in AB class. Interestingly, 5 out of 6 Lepidopteran insects have no CT-thiolase.</ns0:p><ns0:p>Furthermore, TFEL (type-2) class was only detected in C. lectularius (Hemipter) and bacterium</ns0:p></ns0:div> <ns0:div><ns0:head>M. tuberculosis.</ns0:head><ns0:p>To understand the evolutionary mode, gene gain and loss of thiolases were analyzed. It was indicated that most of the gain and loss events were occurred in a certain species (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>Especially, P. xylostella (Lepidoptera), A. pisum (Hemipter), and P. xuthus (Lepidoptera) showed more duplications after or during the formation of the species, resulting in a total number of 15, 12, and 10 genes, respectively. Except for the gene duplication of a single species lineage, the common ancestor of D. alloeum, A. mellifera and B. impatiens showed 2 duplications (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>This phenomenon was also noted in the clade of Hemipteran H. halys, C. lectularius, and A. pisum.</ns0:p><ns0:p>For those recent duplication genes, they were often phylogenetically closely related to its ancestral genes (Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>). Conversely, A. mellifera (Hymenoptera), H. halys (Hemipter), and Z. nevadensis (Isoptera) presented 3, 3, and 2 gene losses during speciation, resulting in fewer genes in these species. Generally, gene gain and loss rates are important for understanding the role of natural selection and adaptation in shaping gene family sizes. For the species with more gene expansion, whether these duplicated genes play roles in adapting to special habitats deserves further study.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene structures of insect thiolase genes</ns0:head><ns0:p>A comparative analysis of exon-intron structures was conducted for the 137 insect thiolase genes (Fig. <ns0:ref type='figure' target='#fig_8'>3A</ns0:ref> and 3B; Fig. <ns0:ref type='figure'>S2A</ns0:ref>). The insect thiolase genes have a different number of introns ranging from 0 to 22. It was indicated that only 12 genes have no intron, and 17 genes have only one intron, accounting for 21.17% in total. The intronless genes were distributed in T2 (5), SCP2 (type-1) (2), CT (2), AB (2), and TFEL (1). Previous studies revealed that introns can delay regulatory response and are selected against in genes whose transcripts need to be adjusted quickly to meet environmental challenges <ns0:ref type='bibr' target='#b2'>(Jeffares, Penkett &amp; Bahler 2008)</ns0:ref>. The intronless thiolase genes and the genes contained fewer introns might play important roles in survival for environmental changes.</ns0:p><ns0:p>In addition, the intron number and exon/intron structures of thiolase genes are very different, even the orthologous genes of different species in the same class have a large differentiation. It was suggested that the differentiation of the intron number may result in the diversification of thiolase gene structures in insects.</ns0:p></ns0:div> <ns0:div><ns0:head>Chromosome distribution and gene synteny</ns0:head><ns0:p>In order to explore the chromosomal distribution of thiolase genes, five representative species were analyzed. It was indicated that most of the thiolase genes were randomly distributed on different chromosomes (Fig. <ns0:ref type='figure' target='#fig_8'>S3A-E</ns0:ref>), for example, 6 thiolase genes were scattered on 4 chromosomes in D.</ns0:p><ns0:p>melanogaster (Fig. <ns0:ref type='figure' target='#fig_8'>S3A</ns0:ref>), which is similar to thiolase genes in the human <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014</ns0:ref>).</ns0:p><ns0:p>However, different members of a certain class are often tandem distribution, such as DmelSCP2 (type-1)-1 and DmelSCP2 (type-1)-2 in D. melanogaster (Fig. <ns0:ref type='figure' target='#fig_8'>S3A</ns0:ref>) and BmorT1-1, BmorT1-2, and BmorT1-3 in B. mori (Fig. <ns0:ref type='figure' target='#fig_8'>S3C</ns0:ref>). Meanwhile, we also detected the distribution of the 10 T1thiolase genes in P. xylostella and 7 CT-thiolase genes in A. pisum, which were distributed on several small unassembled scaffolds. Whether they are distributed in tandem, we still need to wait</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed for the scaffold sequences to be integrated into the corresponding chromosome in future. In general, tandem duplication might be the main mechanism for enlarging thiolase family in insects.</ns0:p><ns0:p>The syntenic relationships of thiolase genes were investigated among B. mori, H. melpomene, D. melanogaster, T. castaneum, and A. mellifera because their genome sequences have been assembled into chromosome levels. The results indicated that only four genes exhibited the syntenic relationships between B. mori and H. melpomene, that is, BmorT2 and HmelT2, BmorTFE</ns0:p><ns0:p>and HmelTFE (Fig. <ns0:ref type='figure' target='#fig_8'>S3F</ns0:ref>). Interestingly, except for the tandemly duplicated genes, amount of thiolase genes often present orthologous relationships among insect, human, and M. tuberculosis (Fig. <ns0:ref type='figure' target='#fig_6'>1</ns0:ref>). It was suggested that thiolase family is an ancient gene family. Even in insects, the age of thiolase gene differentiation is relatively long. Thus, the discovery of fewer syntenic genes implies that thiolases might mainly locate in some non-conserved genomic blocks (The Heliconius Genome Consortium 2012).</ns0:p></ns0:div> <ns0:div><ns0:head>Subcellular localization of thiolase proteins</ns0:head><ns0:p>Subcellular localization refers to the specific location of a certain protein or the expression product and cytosolic localization was related to the biosynthesis of acetoacetyl-CoA <ns0:ref type='bibr' target='#b6'>(Kursula et al. 2005)</ns0:ref>.</ns0:p><ns0:p>However, in a certain class of thiolase, there are always a few exceptions to the cellular location in some species (Fig. <ns0:ref type='figure' target='#fig_8'>3C</ns0:ref>; Fig. <ns0:ref type='figure'>S2B</ns0:ref>), which suggested that its function might have diverged during evolution.</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved domain characteristics and catalytic residues</ns0:head><ns0:p>To identify the potential domains of insect thiolase proteins (Table <ns0:ref type='table'>S1</ns0:ref>), it was performed hmmscan analysis in Pfam database. The results indicated that all the thiolases contained Thiolase_N and Thiolase_C domains (Fig. <ns0:ref type='figure'>4A</ns0:ref>). In addition to the thiolase domains, SCP2-thioloase (type-1) has a</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed typical sterol carrier protein 2 (SCP2) domain at C terminal. Unexpectedly, some of the members in TFE, CT, and T2 classes contained a ketoacyl-synt (beta-ketoacyl synthase) domain within Thiolase_N (Table <ns0:ref type='table'>S1</ns0:ref>). We carefully checked the alignments of hmmscan search. It was found that the E value was around the threshold 1e-5, and only about 50 amino acids can be aligned, which are much shorter than 250 amino acids of the ketoacyl-synt domain (Pfam ID, PF00109).</ns0:p><ns0:p>Thus, thiolases may not contain the real ketoacyl-synt domain, and just show certain similarities with it <ns0:ref type='bibr'>(Huang et al. 1998)</ns0:ref>. Therefore, based on the domain characteristics, all the insect thiolase encoding genes were classified as 2 groups (Fig. <ns0:ref type='figure'>4A</ns0:ref>).</ns0:p><ns0:p>The conserved sequence blocks of the 20 insects, humans, and M. tuberculosis were analyzed (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). The CxS-motif is the most important sequence fingerprint in the N-terminal domain, which provides the nucleophilic cysteine <ns0:ref type='bibr'>(Zeng &amp; Li 2004;</ns0:ref><ns0:ref type='bibr' target='#b9'>Mazet et al. 2011)</ns0:ref>. Except for some incomplete sequences, almost all the thiolases contained the cysteine residue (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). The histidine of the GHP-motif contributes to the oxyanion hole of the thioester oxygen <ns0:ref type='bibr' target='#b10'>(Merilainen et al. 2009)</ns0:ref>. It was indicated that GHP-motif was highly conserved in in insects, humans, and M. tuberculosis (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). The cysteine of CxGGGxG-motif provides the catalytic residue of the active sites. Except for SCP2-thiolases, the catalytic cysteine was retained in almost all the other thiolases (Fig. <ns0:ref type='figure'>S4</ns0:ref>). In addition, the asparagine side chain of the NEAF-motif interacts with important catalytic water <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011</ns0:ref>). However, NEAF-motif was replaced by HDCF-motif in all of the SCP2-thiolases (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). Based on the comparison of the sequence fingerprints, it was indicated that the catalytic mechanisms of the insect thiolases might be similar to that of thiolases from mammals and bacteria.</ns0:p></ns0:div> <ns0:div><ns0:head>Molecular modeling of insect thiolases</ns0:head><ns0:p>In recent years, the crystal structures of some thiolases have been gradually resolved in bacteria, fish, and mammals <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b4'>Kim et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. The high sequence similarities (&gt;60%) may help to build more accurate 3D structures for the insect thiolases <ns0:ref type='bibr'>(Arnold et al. 2006)</ns0:ref>. Based on homology modeling using SWISS-MODEL Interactive Workspace, we found that thiolase sequences within a class were very conserved among different organisms. For instance, BmorTFE-thiolase and BmorSCP2-thiolase (type-1) shared 67.58% and 61.63 % identities with its corresponding modeling templates from human (PDB ID: 6dv2.1.A) and zebrafish (6hrv.2.A), respectively. In this study, the modeling structures of some representative thiolases were presented (Fig. <ns0:ref type='figure'>5A-F</ns0:ref>). The structural similarities of the monomeric forms were BmorTFE and BmorSCP2 (type-1) shared from 1.12 &#197; to 2.01 &#197; with the others (Table <ns0:ref type='table'>S3</ns0:ref>). It was indicated that T1, T2, CT, and AB classes share more similar 3D structures than TFE-thiolase and SCP2-thiolase (type-1) (Fig. <ns0:ref type='figure'>5A-F</ns0:ref>). This phenomenon is widespread in both humans and M.</ns0:p><ns0:p>tuberculosis <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. For the quaternary structrue, different thiolases also have certain differences. For example, BmorT1-1 and BmorSCP2 (type-1) were homo-tetramer and homo-dimer, respectively (Fig. <ns0:ref type='figure'>5G and 5F</ns0:ref>). The results of 3D structural modeling showed that different classes of thiolase genes still present some extent divergence in tertiary or quaternary structures.</ns0:p><ns0:p>SCP2-thiolase (type-1) was widely distributed in insects, mammals, and bacteria (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>One single structural gene referred to as the sterol carrier protein x (SCPx) gene encodes a fulllength protein comprised of 3-oxoacyl-CoA thiolase (known as SCP2-thiolase) and sterol carrier protein 2 <ns0:ref type='bibr' target='#b16'>(Seedorf et al. 1994;</ns0:ref><ns0:ref type='bibr'>Gallegos et al. 2001)</ns0:ref>. The C-terminal SCP2-domain containing the peroxisomal targeting signal is needed for the targeting of full-length SCPx into the peroxisomes.</ns0:p><ns0:p>The SCP2-thiolase and SCP2 protein are produced from SCPx via proteolytic cleavage by peroxisomal proteases <ns0:ref type='bibr' target='#b16'>(Seedorf et al. 1994)</ns0:ref>. Based on the homology modeling, the tertiary and quaternary structures of mature SCP2-thiolase (type-1) protein were presented in Fig. <ns0:ref type='figure'>5F and 5H</ns0:ref>, respectively. For insect SCP2-thiolases, the canonical CxGGGxG-motif is also absent, and the NEAF-motif has been replaced by HDCF-motif (Fig. <ns0:ref type='figure'>4B</ns0:ref>). The previous studies indicated that HDCF-motif might provide the catalytic cysteine in bacteria, mammals, and fish <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b3'>Kiema et al. 2019)</ns0:ref>. Based on the structural modeling, the cysteine of HDCF-motif is very close to the other two catalytic sites in protein spatial conformation (Fig. <ns0:ref type='figure'>5F</ns0:ref>). Therefore, the catalytic cysteine of the insect SCP2-thiolases might be not provided by CxGGGxG-motif but HDCF-motif.</ns0:p></ns0:div> <ns0:div><ns0:head>Expression profile and potential functional diversity</ns0:head><ns0:p>To understand the potential functional diversity of the insect thiolases, the silkworm, B. mori, was used as a model organism to perform expression profile analysis in the various tissues and sex pheromone glands (PGs) at different developmental stages. In the silkworm, genome-wide microarray with 22,987 oligonucleotides was designed and surveyed the gene expression profiles in multiple tissues on day 3 of the fifth-instar larvae <ns0:ref type='bibr' target='#b25'>(Xia et al. 2007)</ns0:ref>. 5 out of 6 thiolase genes Manuscript to be reviewed were found its corresponding probes (Fig. <ns0:ref type='figure' target='#fig_11'>6A</ns0:ref>). The microarray data indicated that BmorSCP2</ns0:p><ns0:p>(type-1), BmorT2, BmorTFE, and BmorT1-1 have expression signals at least one of the 9 tissues.</ns0:p><ns0:p>Relatively, BmorT2 and BmorT1-1 showed ubiquitous expressions. Meanwhile, the expression profiles of the four genes were similar between females and males, respectively.</ns0:p><ns0:p>To validate the expression profiles of the silkworm thiolase genes, the mixed male and female tissues were used to perform RT-PCR validation on day 3 of the fifth-instar larvae (Fig. <ns0:ref type='figure' target='#fig_11'>6B</ns0:ref>). In total, 5 out of the 6 thiolase genes presented expression evidence. Relatively, BmorT1-2</ns0:p><ns0:p>and BmorTFE showed predominant expressions in hemocyte and head, respectively (Fig. <ns0:ref type='figure' target='#fig_11'>6B</ns0:ref>), while BmorT1-1 was widely expressed in various tissues. In addition, sex pheromone glands of different developmental stages were used to detect the expressions of thiolase genes (Fig. <ns0:ref type='figure' target='#fig_11'>6C</ns0:ref>).</ns0:p><ns0:p>Four thiolase genes presented expression signals in the silkworm PGs. Relatively, BmorT1-1</ns0:p><ns0:p>showed the highest expression on day 8 of pupae. BmorTFE, BmorT2, and BmorSCP2 (type-1)</ns0:p><ns0:p>presented expressions at all the developmental stages. Interestingly, the expression levels of all three genes were declined in the mated female PGs (Fig. <ns0:ref type='figure' target='#fig_11'>6C</ns0:ref>). These expression analyses might help us understand the functional divergence of the thiolase genes in the silkworm.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Thiolases are widely distributed in all organisms and are essential for a range of metabolic pathways. With the development of sequencing technology, it provides the possibility for us to identify and compare insect thiolase at the whole genome level. In this study, 137 thiolase genes were identified in the 20 representative species from 7 insect orders (Table <ns0:ref type='table'>1</ns0:ref>; Table <ns0:ref type='table'>S1</ns0:ref>). The insect thiolases were mainly classified into five classes, including CT-thiolase, T1-thiolase, T2-thiolase, TFE-thiolase, and SCP2-thiolase. It was indicated that P. xylostella, A. pisum, and P. xuthus showed more duplications, resulting in a total number of 15, 12, and 10 genes, respectively. Z.</ns0:p><ns0:p>nevadensis and H. melpomene have the least number of genes (Table <ns0:ref type='table'>1</ns0:ref>). In addition to a certain differentiation in the number of genes, Thiolase_N or Thiolase_C domains of 9 thiolase genes were missing (Table <ns0:ref type='table'>S1</ns0:ref>). It is worth noting that the quality of the genome may have a certain impact on the number of genes and the integrity of gene structures. Whether the incomplete thiolase genes were pseudogenes or not (Table <ns0:ref type='table'>S1</ns0:ref>) needs further verification by the high-quality genome in the future.</ns0:p><ns0:p>Two groups of thiolases were identified in animals: 3-oxoacyl-CoA thiolase and acetoacetyl-CoA thiolase, which participates in different catabolic (fatty acid oxidation and bile acid while the physiological role has not been characterized in insects. Fortunately, the expression of the SCP2-thiolase (type-1) has also been detected in the prothoracic glands of Spodoptera littoralis, which are the main tissue producing the insect molting hormone <ns0:ref type='bibr' target='#b18'>(Takeuchi et al. 2004</ns0:ref>).</ns0:p><ns0:p>Thus, whether SCP2-thiolase (type-1) of the silkworm and other insects play role in the oxidation of cholesterol and participates in ecdysone synthesis needs further study.</ns0:p><ns0:p>In insects, juvenile hormone (JH) is an important regulator for growth and development <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007</ns0:ref>) and several thiolases have been cloned and suggested to be related to JH biosynthesis <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhu et al. 2016;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. Acetoacetyl-CoA thiolase catalyzes two molecules of acetyl-CoA to form acetoacetyl-CoA, which is the first enzyme in JH biosynthesis <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. The candidate acetoacetyl-CoA thiolases related to JH biosynthesis were cloned in B. mori and Helicoverpa armigera <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. In this study, those two acetoacetyl-CoA thiolase genes were classified as T2thiolases (BmorT2 and HarmT2), and they shared high sequence identities with the other T2thiolases (Table <ns0:ref type='table'>S4</ns0:ref>). For example, BmorT2-thiolase shared 82.71% sequence identity with</ns0:p><ns0:p>HarmT2. In H. armigera, temporal expressions of HarmT2-thiolase keep pace with JH fluctuations, and its expression can be inhibited by a juvenile hormone analog <ns0:ref type='bibr'>(Zhang et al. 2017)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The expression of BmorT2-thiolase was relatively abundant in the head where the JH synthetic gland, corpora allata (CA), is located (Fig. <ns0:ref type='figure' target='#fig_11'>6A and 6B</ns0:ref>). Interestingly, we found BmorTFE-and BmorT1-1-thiolase also showed high expressions in the larval head (Fig. <ns0:ref type='figure' target='#fig_11'>6A</ns0:ref>). In humans, TFEand T1-thiolases catalyze thiolytic cleavage of 3-ketoacyl-CoA into acetyl-CoA and acyl-CoA <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. However, T1-thiolase has been found synthetic and degradative activities in Ostrinia scapulalis (Lepidoptera: Crambidae). Therefore, whether T2-, TFE-and T1-thiolases were involved in JH biosynthesis is still worthy of experimental validation.</ns0:p><ns0:p>Acetyl-CoA is often used as the initial precursor for sex pheromone biosynthesis in insects <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010)</ns0:ref>. Degradative thiolases may supplement with sufficient acetyl-CoA for sex pheromone synthesis <ns0:ref type='bibr'>(Brabcova et al. 2015)</ns0:ref>. In this study, expression profiles of the thiolase genes were detected in the sex pheromone glands at different developmental stages in the silkworm (Fig. <ns0:ref type='figure' target='#fig_11'>6C</ns0:ref>). Relatively, BmorSCP2 (type-1) maintains a high level of expression in the PGs on day 4 of pupae to 24-h-old virgin female moth. However, its expression level was sharply declined in the mated female PGs (Fig. <ns0:ref type='figure' target='#fig_11'>6C</ns0:ref>). The previous study suggested that an over 6-h mating duration can terminate the sex pheromone production in the silkworm <ns0:ref type='bibr' target='#b1'>(Ando et al. 1996)</ns0:ref>. The expression pattern of BmorSCP2 (type-1) was consistent with sex pheromone production <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010)</ns0:ref>, which suggested that it might be involved in sex pheromone biosynthesis. Generally, it is tempting to assume that a thiolase expressed in a specific tissue might obtain a specific role. Thus, the functional diversification and physiological roles of insect thiolases need yet further experimental validation.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In the present study, genome-wide identification of the thiolase gene family was conducted for the first time in multiple insect genomes. A total of 137 thiolase genes were identified in 20 insects from 7 orders. About 80% of the thiolase genes have 2 or more introns, and its exon/intron structures reserve diversification. Based on the prediction, all the thiolase proteins are located in the mitochondria, cytosol, or peroxisome, and thiolases of the same class often have similar cellular localization. Four highly conserved sequence fingerprints were found in the insect thiolase proteins, including CxS-, NEAF-, GHP-, and CxGGGxG-motifs. Homology modeling analysis indicated that 3D structures of the insect thiolases share similar to mammals, fishes, and microorganisms. Expression pattern analysis suggested some thiolase genes may be involved in Manuscript to be reviewed steroid metabolism, JH, and sex pheromone biosynthesis pathways in B. mori. These results might provide valuable information for the functional exploration of thiolase proteins in insects. The species tree was obtained from timetree database ( http://www.timetree.org/ ). Gain and loss analysis was conducted by Notung-2.9 software with default parameters. The orange and green vertical bars on branch presented gene gain and loss, respectively. The number in each node is gene count. Gene number of each species was presented in brackets.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed (H) Homo-dimer of BmorSCP2 (type-1). </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 6</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>were detected in various tissues and developmental sex pheromone gland of B. mori. Combining structural characteristics and expression patterns, the potential functions and involved PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020) physiological processes were hypothesized. The present study can help us understand the functional differentiation of thiolase genes in insects.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>of a certain gene in the cell. Protein subcellular localization is closely related to protein functions<ns0:ref type='bibr' target='#b13'>(Pereto, Lopez-Garcia &amp; Moreira 2005;</ns0:ref><ns0:ref type='bibr' target='#b23'>Wang et al. 2014)</ns0:ref>. Only when the protein is positioned correctly can it perform normal biological functions. In this study, subcellular localization of //www.genscript.com/tools/psort), which were cytosolic, mitochondrial, or peroxisomal enzymes (Fig.3C; Fig.S2B). Generally, most of the TFE-and T2-thiolase proteins were located in the mitochondrion, T1-and CT-thiolases were cytosolic, and SCP2-thiolases were peroxisomal proteins. Previous studies suggested that the mitochondrial and peroxisomal thiolase proteins were mainly involved in the fatty acid &#946;-oxidation pathway<ns0:ref type='bibr' target='#b13'>(Pereto, Lopez-Garcia &amp; Moreira 2005),</ns0:ref> </ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)Manuscript to be reviewed detected with magic fit in Swiss-PdbViewer(Guex &amp; Peitsch 1997). The RMSD values were ranged from 0.25 &#197; to 0.91 &#197; among BmorT2, BmorT1-1, DmelCT, and PxutAB-1, while</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)Manuscript to be reviewed formation) and anabolic (cholesterogenesis, ketone body synthesis, fatty acid elongation) processes(Antonenkov, Van Veldhoven &amp; Mannaerts 1999). It is well known that cholesterol is a precursor of molting hormone, 20-hydroxyecdysone (20E), and is a structural component of cell membranes(Gilbert, Rybczynski &amp; Warren 2002). Due to the lack of squalene monooxygenase and lanosterol synthase for the synthesis of cholesterol, insects can not autonomously synthesize the 20E precursor(Guo et al. 2009). Alternatively, insects can obtain cholesterol or other sterols from their diet to meet the needs of growth and development. In Spodoptera litura, one sterol carrier protein x (SCPx) gene encoding a sterol carrier protein 2 and a 3-oxoacyl-CoA thiolase known as SCP2-thiolase (type-1) showed predominant expression in the midgut, and its coding SCP2 was involved in the absorption and transport of cholesterol(Guo et al. 2009). In the silkworm, SCPx gene has been cloned(Gong et al. 2006). It presented expressions in the midgut, fat body, and head on day 3 of the fifth-instar larvae in the silkworm (Fig.6B), which suggested that the SCP2 protein might have a similar function with that of S. litura(Guo et al. 2009). More important, the SCP2-thiolase (type-1) encoded by SCPx plays a crucial role in the oxidation of the branched side chain of cholesterol to form bile acids in vertebrates(Ferdinandusse et al. 2000),</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Fig. 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig. 1. Phylogenetic tree of insect thiolases using the maximum-likelihood (ML) method.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Figure 2 Fig. 2 .</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Fig. 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig. 3. Exon/intron structure and subcelluar localization analyses.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 4 Fig. 4 .</ns0:head><ns0:label>44</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 5 Fig. 5 .</ns0:head><ns0:label>55</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Fig. 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Fig. 6. Expression profiles of the thiolase genes in the silkworm.</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>). The annotated genes and genomes of B. mori were retrieved from SilkDB v3.0 (https://silkdb.bioinfotoolkits.net). The sequence information of Danaus plexippus and Heliconius melpomene were downloaded from http://metazoa.ensembl.org/.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Manduca</ns0:cell><ns0:cell>sexta</ns0:cell><ns0:cell>was</ns0:cell><ns0:cell>from</ns0:cell></ns0:row></ns0:table><ns0:note>ftp://ftp.bioinformatics.ksu.edu/pub/Manduca/OGS2/. The other sequences were retrieved from GenBank (https://www.ncbi.nlm.nih.gov/), including Papilio xuthus, Plutella xylostella, Culex quinquefasciatus, Anopheles gambiae, Drosophila melanogaster, Anoplophora glabripennis, Nicrophorus vespilloides, Tribolium castaneum, Halyomorpha halys, Acyrthosiphon pisum, Cimex lectularius, Diachasma alloeum, Apis mellifera, Bombus impatiens, Pediculus humanus, and Zootermopsis nevadensis.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Chromosome distribution, gene structure, and syntenic analysis</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='7'>evolution by ProtTest 3.2 (Darriba et al. 2011). Maximum-likelihood (ML) trees were</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>reconstructed using RAxML version 8.2.12 (Stamatakis 2014) with the most suitable model</ns0:cell></ns0:row><ns0:row><ns0:cell>(PROTGAMMAVTF)</ns0:cell><ns0:cell>and</ns0:cell><ns0:cell>500</ns0:cell><ns0:cell>bootstrap</ns0:cell><ns0:cell>replicates.</ns0:cell><ns0:cell>FigTree</ns0:cell><ns0:cell>v1.4.3</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>(http://tree.bio.ed.ac.uk/software/figtree/) was used for plotting the final phylogenetic tree. The</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>clustering and classification of the thiolase sequences in the ML tree were done using known</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>functional properties of H. sapiens and M. tuberculosis (Anbazhagan et al. 2014).</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='7'>To localize the thiolase genes on chromosomes, B. mori, H. melpomene, D. melanogaster, T.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>castaneum, and A. mellifera were selected because their genome sequences have been assembled</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>into chromosomes. Based on the GFF (General Feature Format) file of each species, every thiolase</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>gene was mapped to the corresponding chromosomes. Using protein sequences of the thiolases,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>the precise exon/intron structures were generated through BLAT search against the genome</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>sequences with Scipio server (https://www.webscipio.org/). The synteny events between two</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>species were detected by Multiple Collinearity Scan toolkit (MCScanX) with the default</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>parameters (Wang et al. 2012). The syntenic map of B. mori and H. melpomene was constructed</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>with family_circle_plotter.java in MCScanX software.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:52282:2:0:NEW 11 Oct 2020)</ns0:note> </ns0:body> "

"College of Life Science China West Normal University Nanchong 637009 Sichuan China October 11, 2020 Dear editors, First, thank you for your timely processing of my manuscript. I also thank the reviewers for rigorous review. Some grammatical and typographical errors have been discovered. This will greatly help improve the quality of the manuscript. Based on the comments and suggestions, I have carefully revised the manuscript and responded one-by-one. I hope that the manuscript is now suitable for publication in PeerJ. Dr. Shou-Min Fang Associate Professor of Genomics Editor comments (Kenta Nakai) MINOR REVISIONS Your revised manuscript has been reviewed by the same two reviewers, whose comments are shown below. As you will see in their comments, one of them now recommends its acceptance as is while the other still raises several points. Please read these comments carefully and re-revise the manuscript, unless you disagree with them. Particularly, you should clarify why you have used only four out of 20 insects data. [# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.  It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter.  Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #] Response: Thank you for your timely processing of my manuscript. I have carefully revised the manuscript, and answered to all the comments. Reviewer 1 (Anonymous) Basic reporting No comment. Experimental design No comment. Validity of the findings No comment. Comments for the Author The author has addressed my concerns well, and I have no further comment. Response: Thanks for your review and affirmation. Reviewer 2 (Anonymous) Basic reporting The author has corrected most of grammatical and typographical errors, and problems of figures and tables, I previously commented, but there are still some are left to be corrected and new ones appeared: Line 108: have => “that have” or “having”; Thionlase => Thiolase; recogized => recognized Line 124: “, which” => because Line 126: genes => gene Line 236: “Unexceptly” may be “Unexpectedly”? Line 258: higher => high Line 304: “head and hemocyte” => “hemocyte and head” Lines 304–305: “. While, ” => “, while ” Line 338: “was suggested” => “suggested” Line 344: “Spodopera littoralis” => “Spodoptera littoralis” Line 355: higher => high Line 390: suggests => suggested The title of Table S2 in the word file is Table S1. Response: Agreed. According to the suggestions, all the mentioned texts and words were revised. Experimental design The author has adequately addressed all my comments. Validity of the findings The author has adequately addressed all my comments. Comments for the Author The author has addressed most of my comments. However, there are still some concerns that need to be resolved. Major concerns: * The author should use all the insects thiolases in the conserved sequence block analysis. In the present manuscript, only 4 of 20 insects were used to derive the conclusion for insects thiolases, which is unreasonable. Response: Based on the suggestion, the highly conserved sequence blocks of all the 20 insects, human and M. tuberculosis were analyzed and presented in Figure S4. The corresponding result section was appropriately revised. * The author seems to regard BmorT2-thiolase as involving JH synthesis pathway only based on weak estimation from gene expression and it still needs further experimental validations similar to those for HarmT2-thiolase by (Zhang et al. 2017). ** Lines 360–361: “Thus, insect T2-thiolases may be involved in the JH synthesis pathway.”: I think it is unreasonable to generalize the JH synthesis about insects from only an example in H. armigera and an estimation in B. mori. ** Lines 365–366: “except for T2-thiolase” is not appropriate, I think. Response: “Thus, insect T2-thiolases may be involved in the JH synthesis pathway.” and “except for T2-thiolase” were deleted. “Therefore, except for T2-thiolase, whether TFE- and T1-thiolases were also involved in JH biosynthesis is still worthy of validation.” was revised to “Therefore, whether T2-, TFE- and T1-thiolases were involved in JH biosynthesis is still worthy of experimental validation.” Minor concerns: Lines 266–270: Why did the author use only BmorT2 as the reference to calculate the RMSD values here? It is not enough result to support the description: “It was indicated that T1, T2, CT, and AB classes share more similar 3D structures than TFE-thiolase and SCP2-thiolase (type-1)”. Response: In this revision, RMSD values were estimated by Swiss-PdbViewer among all the 6 thiolases. The RMSD matrix was presented by Table S3. These results should support “It was indicated that T1, T2, CT, and AB classes share more similar 3D structures than TFE-thiolase and SCP2-thiolase (type-1)”. The corresponding results were also revised. Please see the lines 267-269. Line 270: “This phenomenon is widespread in both humans and M. tuberculosis” needs references. Response: Two references (Harijan et al. 2013; Anbazhagan et al. 2014) were cited. Lines 270–274: The sentences, “Generally, T1-, T2-, ... biological unit (Xia et al. 2019)”, seem to mostly describe the general fact and are not appropriate in the Results section. The author should better describe what was derived from the results of Figs. 5G and 5H, or remove this part. Response: Some of the sentences were deleted. Based on the Figs. 5G and 5H, a result was added. Lines 303, 307: “In total, 5 out of the 6 thiolase genes presented expression evidence.”, “Four thiolase genes presented expression signals in the silkworm PGs.”: Figs. 6B and 6C do not show the results of all the 6 thiolase genes. The author should show all the results in the figure. Response: Those genes have no signals in the various tissues on day 3 of the fifth-instar larvae or PGs were omitted in the Figs. 6B and 6C. Because no amplified target bands could be detected. Lines 337–339: “It” in the sentence, “It presented expressions ...”, indicates SCPx gene, but Fig.6B shows the expression result of SCP2 thiolase gene. It is better to fill the gap between SCPx gene, SCP2 thiolase gene/protein, and SCP2 protein just before here for easy understanding. Response: Agreed. Based on the suggestion, information “…one sterol carrier protein x (SCPx) gene encoding a sterol carrier protein 2 and a 3-oxoacyl-CoA thiolase known as SCP2-thiolase (type-1)…” was added to point out the relationship among SCPx gene, SCP2 thiolase gene/protein, and SCP2 protein. Please check the lines 335-337. Lines 348–350: The two sentences may be better to be combined into one sentence. Response: Agreed. It was revised as follows: In insects, juvenile hormone (JH) is an important regulator for growth and development (Kinjoh et al. 2007) and several thiolases have been cloned and suggested to be related to JH biosynthesis (Kinjoh et al. 2007; Zhu et al. 2016; Zhang et al. 2017). Lines 376–377: The author wrote that “Simultaneously, BmorTFE and BmorT2 showed similar expression patterns as BmorSCP2 (type-1).”, but BmorTFE and BmorT2 seem to show different expression patterns from BmorSCP2 (type-1) in Fig. 6C for me. Response: Indeed, expression level of BmorTFE and BmorT2 was relatively low. To avoid disagreement, “Simultaneously, BmorTFE and BmorT2 showed similar expression patterns as BmorSCP2 (type-1).” was deleted. Lines 31–33, 390–391: “may be involved” or “might be involved” is appropriate instead of “were involved” because of the weak evidences. Response: Agreed. both of them were revised. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Thiolases are important enzymes involved in lipid metabolism in both prokaryotes and eukaryotes, and are essential for a range of metabolic pathways, while, little is known for this important family in insects. To shed light on the evolutionary models and functional diversities of the thiolase family, 137 thiolase genes were identified in 20 representative insect genomes. They were mainly classified into five classes, namely cytosolic thiolase (CT-thiolase), T1-thiolase, T2-thiolase, trifunctional enzyme thiolase (TFE-thiolase), and sterol carrier protein 2 thiolase (SCP2-thiolase). The intron number and exon/intron structures of the thiolase genes reserve large diversification. Subcellular localization prediction indicated that all the thiolase proteins were mitochondrial, cytosolic, or peroxisomal enzymes. Four highly conserved sequence fingerprints were found in the insect thiolase proteins, including CxS-, NEAF-, GHP-, and CxGGGxG-motifs. Homology modeling indicated that insect thiolases share similar 3D structures with mammals, fishes, and microorganisms. In Bombyx mori, microarray data and reverse transcriptionpolymerase chain reaction (RT-PCR) analysis suggested that some thiolases might be involved in steroid metabolism, juvenile hormone (JH), and sex pheromone biosynthesis pathways. In general, sequence and structural characteristics were relatively conserved among insects, bacteria and vertebrates, while different classes of thiolases might have differentiation in specific functions and physiological processes. These results will provide an important foundation for future functional validation of insect thiolases.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>Thiolases are ubiquitous enzymes that play important roles in lipid-metabolizing pathways <ns0:ref type='bibr' target='#b20'>(Thompson et al. 1989;</ns0:ref><ns0:ref type='bibr'>Igual et al. 1992;</ns0:ref><ns0:ref type='bibr' target='#b13'>Pereto, Lopez-Garcia &amp; Moreira 2005)</ns0:ref>. There are two major kinds of thiolases based on the direction of the catalytic reaction <ns0:ref type='bibr' target='#b7'>(Masamune et al. 1989;</ns0:ref><ns0:ref type='bibr' target='#b11'>Modis &amp; Wierenga 2000)</ns0:ref>. One is degradative thiolase I (3-ketoacyl-CoA thiolase, E.C. 2.3.1.16), which catalyzes the thiolytic cleavage of medium-to long-chain unbranched 3-oxoacyl-CoAs (from 4 to 22 carbons) into acetyl-CoA and a fatty acyl-CoA (Fig. <ns0:ref type='figure' target='#fig_5'>S1A</ns0:ref>) <ns0:ref type='bibr'>(Clinkenbeard et al. 1973</ns0:ref><ns0:ref type='bibr' target='#b14'>, Schiedl et al. 2004</ns0:ref><ns0:ref type='bibr'>, Houten &amp; Wanders 2010)</ns0:ref>. It is mainly involved in fatty acid &#946;-oxidation and preferentially catalyzes the last step. The other is biosynthetic thiolase II (acetoacetyl-CoA thiolase, EC2.3.1.9), which is capable of catalyzing the Claisen condensation reaction of two molecules of acetyl-CoA to acetoacetyl-CoA (Fig. <ns0:ref type='figure' target='#fig_5'>S1B</ns0:ref>). Thiolase II might be involved in poly beta-hydroxybutyric acid synthesis, steroid biogenesis, etc. <ns0:ref type='bibr'>(Clinkenbeard et al. 1973)</ns0:ref>.</ns0:p><ns0:p>Based on the function, oligomeric state, substrate specificity, and subcellular localization, six different classes of thiolases (CT, AB, SCP2, T2, T1, and TFE) have been identified in humans <ns0:ref type='bibr'>(Fukao 2002;</ns0:ref><ns0:ref type='bibr' target='#b9'>Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. In trypanosomatid and bacterial kingdoms, another four classes were identified, including thiolase-like protein (TLP), SCP2thiolase-like protein (SLP), unclassified thiolase (UCT), and TFE-like thiolase (TFEL) <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. The CT-thiolase, located in the cytosol, has a key role in catalyzing the condensation of two molecules of acetyl-CoA to acetoacetyl-CoA, which is the first reaction of the metabolic pathway leading to the synthesis of cholesterol <ns0:ref type='bibr' target='#b6'>(Kursula et al. 2005)</ns0:ref>.</ns0:p><ns0:p>AB-thiolase and SCP2-thiolase as degradative thiolases occur in peroxisomes <ns0:ref type='bibr'>(Antonenkov et al. 1997;</ns0:ref><ns0:ref type='bibr'>Antonenkov et al. 1999)</ns0:ref>. T1-, T2-, and TFE-thiolases are mitochondrial degradative enzymes. Except for the degradation of acetoacetyl-CoA and 2-methyl-acetoacetyl-CoA, T2thiolase has a biosynthetic function in the synthesis of acetoacetyl-CoA in ketone body metabolism PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed <ns0:ref type='bibr'>(Fukao et al. 1997)</ns0:ref>. In general, AB-, SCP2-, T1-, and TFE-thiolases belong to degradative thiolase I, and CT class is biosynthetic thiolase II. Importantly, T2-thiolase is a bi-functional enzyme with synthesis and degradation activity.</ns0:p><ns0:p>As enzymes responsible for broad pathways, thiolases have been performed the phylogenetic analysis in mycobacteria and functional studies in humans <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. In insects, several thiolase genes have been characterized <ns0:ref type='bibr'>(Fujii et al. 2010)</ns0:ref>. In Helicoverpa armigera, an acetoacetyl-CoA thiolase was cloned and performed functional analysis <ns0:ref type='bibr'>(Zhang et al. 2017)</ns0:ref>. It was indicated that the thiolase was involved in the early step of the juvenile hormone pathway, i.e. mevalonate biosynthesis. One acetoacetyl-CoA thiolase was purified to apparent homogeneity by column chromatography in Bombus terrestris, suggesting that it might be the first enzyme in the biosynthesis of terpenic sex pheromone <ns0:ref type='bibr'>(Brabcova et al. 2015)</ns0:ref>. Besides, acetyl-CoA can be used as the precursor for de novo biosynthesis of sex pheromones in female moths, and four 3-ketoacyl-CoA thiolase genes were identified in sex pheromone gland transcriptome of the noctuid moth Heliothis virescens <ns0:ref type='bibr' target='#b21'>(Vogel et al. 2010)</ns0:ref>. In total, there are few and scattered studies on insect thiolase, lacking systematic identification and comparative studies.</ns0:p><ns0:p>In this study, we selected 20 representative species from 7 insect orders to perform genomewide identification of the thiolase family proteins. Gene structure, chromosome location, and three-dimensional (3D) structure and motif characteristics of proteins were compared. In addition, Bombyx mori is an important model species for studying juvenile hormone and sex pheromone biosynthesis <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010;</ns0:ref><ns0:ref type='bibr' target='#b26'>Xia, Li &amp; Feng 2014)</ns0:ref>. Expression profiles of the thiolase genes </ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Data resources</ns0:head><ns0:p>In this study, 20 representative species were selected from Lepidoptera, Hymenoptera, Hemiptera, Diptera, Coleoptera, Phthiraptera, and Isoptera (Table <ns0:ref type='table'>1</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Identification of insect thiolase genes</ns0:head><ns0:p>The known thiolase sequences of Homo sapiens and Mycobacterium tuberculosis were retrieved from GenBank (Table <ns0:ref type='table'>S1</ns0:ref>; <ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref> and used as queries to perform BLASTP search (E-value &lt; 0.01) against the protein database of predicted genes in each species. Hidden Markov Model (HMM) files of Thionlase_N (PF00108) and Thionlase_C (PF02803) domains were downloaded from Pfam database (http://pfam.xfam.org/), which were used to screen the protein database of each species with hmmsearch in HMMER 3.0 (E-value &lt; 0.01). Based on BLASTP and hmmsearch analyses, the candidate thiolase genes were identified and subsequently checked by conserved domain search (CD-Search) in NCBI and hmmscan against Pfam database (E-value &lt; 1e-5). The candidate sequences that have Thiolase_N and/or Thiolase_C domains were recognized as thiolases. Those identified thiolases were used as new queries to perform BLASTP search against the protein database of each species until no more novel loci can be found. All the validated thiolase genes were used for further analysis.</ns0:p></ns0:div> <ns0:div><ns0:head>Phylogenetic analysis</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The protein sequences of thiolases from 20 insects, H. sapiens and M. tuberculosis were aligned using MUSCLE <ns0:ref type='bibr'>(Edgar 2004)</ns0:ref>. Positions that had a high percentage of gaps (&gt;70%) were trimmed.</ns0:p><ns0:p>The handled alignment of protein sequences was used for checking the most suitable model of </ns0:p></ns0:div> <ns0:div><ns0:head>Molecular modeling of protein structure</ns0:head><ns0:p>The three-dimensional (3D) structure prediction of insect thiolases was conducted using the homology modeling method. Structures of T1-, T2-, CT-, AB-, TFE-, and SCP2-thiolases were predicted on-line at the SWISS-MODEL Interactive Workspace <ns0:ref type='bibr'>(Arnold et al. 2006)</ns0:ref>. The known protein that has the highest sequence similarity to the thiolase to be analyzed is used for homology modeling. The predicted models of monomer and multimer were visualized in Swiss-PdbViewer 4.1.0 (Guex &amp; Peitsch 1997). To understand the 3D structural similarities among the insect thiolases, all the other structures were compared with BmorT2 using magic fit algorithm in Swiss-PdbViewer, respectively. The root-mean-square distance (RMSD) values were calculated to express the structural similarity. The lower value of RMSD means higher similarity between two structures <ns0:ref type='bibr' target='#b12'>(Carugo &amp; Pongor 2001)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Reverse transcription-polymerase chain reaction (RT-PCR)</ns0:head><ns0:p>The various tissues on day 3 of fifth-instar larvae were dissected in the silkworm. The sex pheromone glands (PGs) from 5 individuals were used as one sample at each developmental stage.</ns0:p><ns0:p>All the samples were preserved in RNAlater (Ambion, 98 Austin, USA) and stored at -80 &#176;C for RNA isolation. Total RNA was extracted using Trizol reagent (Invitrogen, USA). The first strand of cDNA was synthesized by M-MLV reverse transcriptase following the manufacturer's instructions (Promega, USA). RT-PCR primers were listed in Table <ns0:ref type='table'>S2</ns0:ref>. The silkworm RpL3 gene was used as an internal control for relative quantitative analysis of RT-PCR. PCRs were performed with the following cycling parameters: 95 &#176;C for 3 minutes (min), followed by 25 cycles of 30 seconds (s) at 95 &#176;C, 30 s annealing (temperatures listed in Table <ns0:ref type='table'>S2</ns0:ref>) and 30 s extension (72 &#176;C), and a final extension at 72 &#176;C for 10 min. The amplification products were monitored on 1.5% agarose gels.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Genome-wide identification and phylogeny of insect thiolase proteins</ns0:head><ns0:p>To identify thiolases in insects, human and M. tuberculosis thiolase protein sequences were used as queries to perform homologous searches in whole genomes. In total, 137 thiolase genes were identified in 20 insects from 7 orders, and the gene numbers of the species were ranged from 4 to 15 (Table <ns0:ref type='table'>1</ns0:ref>; Table <ns0:ref type='table'>S1</ns0:ref>). The thiolase protein sequences were used to reconstruct the maximumlikelihood phylogenetic tree (Fig. <ns0:ref type='figure' target='#fig_5'>1</ns0:ref>). Based on the nomenclature rules in humans and M.</ns0:p><ns0:p>tuberculosis <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014)</ns0:ref>, each thiolase was named in insects. It was indicated that insect thiolases were grouped 7 classes, namely CT, T1, T2, TFE, SCP2 (type-1), TFEL (type-2), and AB (Table <ns0:ref type='table'>1</ns0:ref>; Fig. <ns0:ref type='figure' target='#fig_5'>1</ns0:ref>). Relatively, the classification of insect thiolases was more similar to that of the human than M. tuberculosis. Unlike humans, out of 20 insect species, only P. xuthus has gene members in AB class. Interestingly, 5 out of 6 Lepidopteran insects have no CT-thiolase.</ns0:p><ns0:p>Furthermore, TFEL (type-2) class was only detected in C. lectularius (Hemipter) and bacterium</ns0:p></ns0:div> <ns0:div><ns0:head>M. tuberculosis.</ns0:head><ns0:p>To understand the evolutionary mode, gene gain and loss of thiolases were analyzed. It was indicated that most of the gain and loss events were occurred in a certain species (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>Especially, P. xylostella (Lepidoptera), A. pisum (Hemipter), and P. xuthus (Lepidoptera) showed more duplications after or during the formation of the species, resulting in a total number of 15, 12, and 10 genes, respectively. Except for the gene duplication of a single species lineage, the common ancestor of D. alloeum, A. mellifera and B. impatiens showed 2 duplications (Fig. <ns0:ref type='figure'>2</ns0:ref>).</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>This phenomenon was also noted in the clade of Hemipteran H. halys, C. lectularius, and A. pisum.</ns0:p><ns0:p>For those recent duplication genes, they were often phylogenetically closely related to its ancestral genes (Fig. <ns0:ref type='figure' target='#fig_5'>1</ns0:ref>). Conversely, A. mellifera (Hymenoptera), H. halys (Hemipter), and Z. nevadensis (Isoptera) presented 3, 3, and 2 gene losses during speciation, resulting in fewer genes in these species. Generally, gene gain and loss rates are important for understanding the role of natural selection and adaptation in shaping gene family sizes. For the species with more gene expansion, whether these duplicated genes play roles in adapting to special habitats deserves further study.</ns0:p></ns0:div> <ns0:div><ns0:head>Gene structures of insect thiolase genes</ns0:head><ns0:p>A comparative analysis of exon-intron structures was conducted for the 137 insect thiolase genes (Fig. <ns0:ref type='figure' target='#fig_7'>3A</ns0:ref> and 3B; Fig. <ns0:ref type='figure'>S2A</ns0:ref>). The insect thiolase genes have a different number of introns ranging from 0 to 22. It was indicated that only 12 genes have no intron, and 17 genes have only one intron, accounting for 21.17% in total. The intronless genes were distributed in T2 (5), SCP2 (type-1) (2), CT (2), AB (2), and TFEL (1). Previous studies revealed that introns can delay regulatory response and are selected against in genes whose transcripts need to be adjusted quickly to meet environmental challenges <ns0:ref type='bibr' target='#b2'>(Jeffares, Penkett &amp; Bahler 2008)</ns0:ref>. The intronless thiolase genes and the genes contained fewer introns might play important roles in survival for environmental changes.</ns0:p><ns0:p>In addition, the intron number and exon/intron structures of thiolase genes are very different, even the orthologous genes of different species in the same class have a large differentiation. It was suggested that the differentiation of the intron number may result in the diversification of thiolase gene structures in insects.</ns0:p></ns0:div> <ns0:div><ns0:head>Chromosome distribution and gene synteny</ns0:head><ns0:p>In order to explore the chromosomal distribution of thiolase genes, five representative species were analyzed. It was indicated that most of the thiolase genes were randomly distributed on different chromosomes (Fig. <ns0:ref type='figure' target='#fig_7'>S3A-E</ns0:ref>), for example, 6 thiolase genes were scattered on 4 chromosomes in D.</ns0:p><ns0:p>melanogaster (Fig. <ns0:ref type='figure' target='#fig_7'>S3A</ns0:ref>), which is similar to thiolase genes in the human <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014</ns0:ref>).</ns0:p><ns0:p>However, different members of a certain class are often tandem distribution, such as DmelSCP2 (type-1)-1 and DmelSCP2 (type-1)-2 in D. melanogaster (Fig. <ns0:ref type='figure' target='#fig_7'>S3A</ns0:ref>) and BmorT1-1, BmorT1-2, and BmorT1-3 in B. mori (Fig. <ns0:ref type='figure' target='#fig_7'>S3C</ns0:ref>). Meanwhile, we also detected the distribution of the 10 T1thiolase genes in P. xylostella and 7 CT-thiolase genes in A. pisum, which were distributed on several small unassembled scaffolds. Whether they are distributed in tandem, we still need to wait</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed for the scaffold sequences to be integrated into the corresponding chromosome in future. In general, tandem duplication might be the main mechanism for enlarging thiolase family in insects.</ns0:p><ns0:p>The syntenic relationships of thiolase genes were investigated among B. mori, H. melpomene, D. melanogaster, T. castaneum, and A. mellifera because their genome sequences have been assembled into chromosome levels. The results indicated that only four genes exhibited the syntenic relationships between B. mori and H. melpomene, that is, BmorT2 and HmelT2, BmorTFE</ns0:p><ns0:p>and HmelTFE (Fig. <ns0:ref type='figure' target='#fig_7'>S3F</ns0:ref>). Interestingly, except for the tandemly duplicated genes, amount of thiolase genes often present orthologous relationships among insect, human, and M. tuberculosis (Fig. <ns0:ref type='figure' target='#fig_5'>1</ns0:ref>). It was suggested that thiolase family is an ancient gene family. Even in insects, the age of thiolase gene differentiation is relatively long. Thus, the discovery of fewer syntenic genes implies that thiolases might mainly locate in some non-conserved genomic blocks (The Heliconius Genome Consortium 2012).</ns0:p></ns0:div> <ns0:div><ns0:head>Subcellular localization of thiolase proteins</ns0:head><ns0:p>Subcellular localization refers to the specific location of a certain protein or the expression product and cytosolic localization was related to the biosynthesis of acetoacetyl-CoA <ns0:ref type='bibr' target='#b6'>(Kursula et al. 2005)</ns0:ref>.</ns0:p><ns0:p>However, in a certain class of thiolase, there are always a few exceptions to the cellular location in some species (Fig. <ns0:ref type='figure' target='#fig_7'>3C</ns0:ref>; Fig. <ns0:ref type='figure'>S2B</ns0:ref>), which suggested that its function might have diverged during evolution.</ns0:p></ns0:div> <ns0:div><ns0:head>Conserved domain characteristics and catalytic residues</ns0:head><ns0:p>To identify the potential domains of insect thiolase proteins (Table <ns0:ref type='table'>S1</ns0:ref>), it was performed hmmscan analysis in Pfam database. The results indicated that all the thiolases contained Thiolase_N and Thiolase_C domains (Fig. <ns0:ref type='figure'>4A</ns0:ref>). In addition to the thiolase domains, SCP2-thioloase (type-1) has a</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed typical sterol carrier protein 2 (SCP2) domain at C terminal. Unexpectedly, some of the members in TFE, CT, and T2 classes contained a ketoacyl-synt (beta-ketoacyl synthase) domain within Thiolase_N (Table <ns0:ref type='table'>S1</ns0:ref>). We carefully checked the alignments of hmmscan search. It was found that the E value was around the threshold 1e-5, and only about 50 amino acids can be aligned, which are much shorter than 250 amino acids of the ketoacyl-synt domain (Pfam ID, PF00109).</ns0:p><ns0:p>Thus, thiolases may not contain the real ketoacyl-synt domain, and just show certain similarities with it <ns0:ref type='bibr'>(Huang et al. 1998)</ns0:ref>. Therefore, based on the domain characteristics, all the insect thiolase encoding genes were classified as 2 groups (Fig. <ns0:ref type='figure'>4A</ns0:ref>).</ns0:p><ns0:p>The conserved sequence blocks of the 20 insects, humans, and M. tuberculosis were analyzed (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). The CxS-motif is the most important sequence fingerprint in the N-terminal domain, which provides the nucleophilic cysteine <ns0:ref type='bibr'>(Zeng &amp; Li 2004;</ns0:ref><ns0:ref type='bibr' target='#b9'>Mazet et al. 2011)</ns0:ref>. Except for some incomplete sequences, almost all the thiolases contained the cysteine residue (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). The histidine of the GHP-motif contributes to the oxyanion hole of the thioester oxygen <ns0:ref type='bibr' target='#b10'>(Merilainen et al. 2009)</ns0:ref>. It was indicated that GHP-motif was highly conserved in insects, humans, and M. tuberculosis (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). The cysteine of CxGGGxG-motif provides the catalytic residue of the active sites. Except for SCP2-thiolases, the catalytic cysteine was retained in almost all the other thiolases (Fig. <ns0:ref type='figure'>S4</ns0:ref>). In addition, the asparagine side chain of the NEAF-motif interacts with important catalytic water <ns0:ref type='bibr' target='#b9'>(Mazet et al. 2011</ns0:ref>). However, NEAF-motif was replaced by HDCFmotif in all of the SCP2-thiolases (Fig. <ns0:ref type='figure'>S4</ns0:ref>; Fig. <ns0:ref type='figure'>4B</ns0:ref>). Based on the comparison of the sequence fingerprints, it was indicated that the catalytic mechanisms of the insect thiolases might be similar to that of thiolases from mammals and bacteria.</ns0:p></ns0:div> <ns0:div><ns0:head>Molecular modeling of insect thiolases</ns0:head><ns0:p>In recent years, the crystal structures of some thiolases have been gradually resolved in bacteria, fish, and mammals <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b4'>Kim et al. 2015;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. The high sequence similarities (&gt;60%) may help to build more accurate 3D structures for the insect thiolases <ns0:ref type='bibr'>(Arnold et al. 2006)</ns0:ref>. Based on homology modeling using SWISS-MODEL Interactive Workspace, we found that thiolase sequences within a class were very conserved among different organisms. For instance, BmorTFE-thiolase and BmorSCP2-thiolase (type-1) shared 67.58% and 61.63 % identities with its corresponding modeling templates from human (PDB ID: 6dv2.1.A) and zebrafish (6hrv.2.A), respectively. In this study, the modeling structures of some representative thiolases were presented (Fig. <ns0:ref type='figure'>5A-F</ns0:ref>). The structural similarities of the monomeric forms were BmorTFE and BmorSCP2 (type-1) shared from 1.12 &#197; to 2.01 &#197; with the others (Table <ns0:ref type='table'>S3</ns0:ref>). It was indicated that T1, T2, CT, and AB classes share more similar 3D structures than TFE-thiolase and SCP2-thiolase (type-1) (Fig. <ns0:ref type='figure'>5A-F</ns0:ref>). This phenomenon is widespread in both humans and M.</ns0:p><ns0:p>tuberculosis <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b0'>Anbazhagan et al. 2014)</ns0:ref>. For the quaternary structure, different thiolases also have certain differences. For example, BmorT1-1 and BmorSCP2 (type-1) were homo-tetramer and homo-dimer, respectively (Fig. <ns0:ref type='figure'>5G and 5F</ns0:ref>). The results of 3D structural modeling showed that different classes of thiolase genes still present some extent divergence in tertiary or quaternary structures.</ns0:p><ns0:p>SCP2-thiolase (type-1) was widely distributed in insects, mammals, and bacteria (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p><ns0:p>One single structural gene referred to as the sterol carrier protein x (SCPx) gene encodes a fulllength protein comprised of 3-oxoacyl-CoA thiolase (known as SCP2-thiolase) and sterol carrier protein 2 <ns0:ref type='bibr' target='#b16'>(Seedorf et al. 1994;</ns0:ref><ns0:ref type='bibr'>Gallegos et al. 2001)</ns0:ref>. The C-terminal SCP2-domain containing the peroxisomal targeting signal is needed for the targeting of full-length SCPx into the peroxisomes.</ns0:p><ns0:p>The SCP2-thiolase and SCP2 protein are produced from SCPx via proteolytic cleavage by peroxisomal proteases <ns0:ref type='bibr' target='#b16'>(Seedorf et al. 1994)</ns0:ref>. Based on the homology modeling, the tertiary and quaternary structures of mature SCP2-thiolase (type-1) protein were presented in Fig. <ns0:ref type='figure'>5F and 5H</ns0:ref>, respectively. For insect SCP2-thiolases, the canonical CxGGGxG-motif is also absent, and the NEAF-motif has been replaced by HDCF-motif (Fig. <ns0:ref type='figure'>4B</ns0:ref>). The previous studies indicated that HDCF-motif might provide the catalytic cysteine in bacteria, mammals, and fish <ns0:ref type='bibr'>(Harijan et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b3'>Kiema et al. 2019)</ns0:ref>. Based on the structural modeling, the cysteine of HDCF-motif is very close to the other two catalytic sites in protein spatial conformation (Fig. <ns0:ref type='figure'>5F</ns0:ref>). Therefore, the catalytic cysteine of the insect SCP2-thiolases might be not provided by CxGGGxG-motif but HDCF-motif.</ns0:p></ns0:div> <ns0:div><ns0:head>Expression profile and potential functional diversity</ns0:head><ns0:p>To understand the potential functional diversity of the insect thiolases, the silkworm, B. mori, was used as a model organism to perform expression profile analysis in the various tissues and sex pheromone glands (PGs) at different developmental stages. In the silkworm, genome-wide microarray with 22,987 oligonucleotides was designed and surveyed the gene expression profiles in multiple tissues on day 3 of the fifth-instar larvae <ns0:ref type='bibr' target='#b25'>(Xia et al. 2007)</ns0:ref>. 5 out of 6 thiolase genes Manuscript to be reviewed were found its corresponding probes (Fig. <ns0:ref type='figure' target='#fig_10'>6A</ns0:ref>). The microarray data indicated that BmorSCP2</ns0:p><ns0:p>(type-1), BmorT2, BmorTFE, and BmorT1-1 have expression signals at least one of the 9 tissues.</ns0:p><ns0:p>Relatively, BmorT2 and BmorT1-1 showed ubiquitous expressions. Meanwhile, the expression profiles of the four genes were similar between females and males, respectively.</ns0:p><ns0:p>To validate the expression profiles of the silkworm thiolase genes, the mixed male and female tissues were used to perform RT-PCR validation on day 3 of the fifth-instar larvae (Fig. <ns0:ref type='figure' target='#fig_10'>6B</ns0:ref>). In total, 5 out of the 6 thiolase genes presented expression evidence. Relatively, BmorT1-2</ns0:p><ns0:p>and BmorTFE showed predominant expressions in hemocyte and head, respectively (Fig. <ns0:ref type='figure' target='#fig_10'>6B</ns0:ref>), while BmorT1-1 was widely expressed in various tissues. In addition, sex pheromone glands of different developmental stages were used to detect the expressions of thiolase genes (Fig. <ns0:ref type='figure' target='#fig_10'>6C</ns0:ref>).</ns0:p><ns0:p>Four thiolase genes presented expression signals in the silkworm PGs. Relatively, BmorT1-1</ns0:p><ns0:p>showed the highest expression on day 8 of pupae. BmorTFE, BmorT2, and BmorSCP2 (type-1)</ns0:p><ns0:p>presented expressions at all the developmental stages. Interestingly, the expression levels of all three genes were declined in the mated female PGs (Fig. <ns0:ref type='figure' target='#fig_10'>6C</ns0:ref>). These expression analyses might help us understand the functional divergence of the thiolase genes in the silkworm.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Thiolases are widely distributed in all organisms and are essential for a range of metabolic pathways. With the development of sequencing technology, it provides the possibility for us to identify and compare insect thiolase at the whole genome level. In this study, 137 thiolase genes were identified in the 20 representative species from 7 insect orders (Table <ns0:ref type='table'>1</ns0:ref>; Table <ns0:ref type='table'>S1</ns0:ref>). The insect thiolases were mainly classified into five classes, including CT-thiolase, T1-thiolase, T2-thiolase, TFE-thiolase, and SCP2-thiolase. It was indicated that P. xylostella, A. pisum, and P. xuthus showed more duplications, resulting in a total number of 15, 12, and 10 genes, respectively. Z.</ns0:p><ns0:p>nevadensis and H. melpomene have the least number of genes (Table <ns0:ref type='table'>1</ns0:ref>). In addition to a certain differentiation in the number of genes, Thiolase_N or Thiolase_C domains of 9 thiolase genes were missing (Table <ns0:ref type='table'>S1</ns0:ref>). It is worth noting that the quality of the genome may have a certain impact on the number of genes and the integrity of gene structures. Whether the incomplete thiolase genes were pseudogenes or not (Table <ns0:ref type='table'>S1</ns0:ref>) needs further verification by the high-quality genome in the future.</ns0:p><ns0:p>Two groups of thiolases were identified in animals: 3-oxoacyl-CoA thiolase and acetoacetyl-CoA thiolase, which participates in different catabolic (fatty acid oxidation and bile acid</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed formation) and anabolic (cholesterogenesis, ketone body synthesis, fatty acid elongation)</ns0:p><ns0:p>processes <ns0:ref type='bibr'>(Antonenkov, Van Veldhoven &amp; Mannaerts 1999)</ns0:ref>. It is well known that cholesterol is a precursor of molting hormone, 20-hydroxyecdysone (20E), and is a structural component of cell membranes <ns0:ref type='bibr'>(Gilbert, Rybczynski &amp; Warren 2002)</ns0:ref>. Due to the lack of squalene monooxygenase and lanosterol synthase for the synthesis of cholesterol, insects can not autonomously synthesize the 20E precursor <ns0:ref type='bibr'>(Guo et al. 2009</ns0:ref>). Alternatively, insects can obtain cholesterol or other sterols from their diet to meet the needs of growth and development. In Spodoptera litura, one sterol carrier protein x (SCPx) gene encoding a sterol carrier protein 2 and a 3-oxoacyl-CoA thiolase known as SCP2-thiolase (type-1) showed predominant expression in the midgut, and its coding SCP2 was involved in the absorption and transport of cholesterol <ns0:ref type='bibr'>(Guo et al. 2009</ns0:ref>). In the silkworm, SCPx gene has been cloned <ns0:ref type='bibr'>(Gong et al. 2006)</ns0:ref>. It presented expressions in the midgut, fat body, and head on day 3 of the fifth-instar larvae in the silkworm (Fig. <ns0:ref type='figure' target='#fig_10'>6B</ns0:ref>), which suggested that the SCP2 protein might have a similar function with that of S. litura <ns0:ref type='bibr'>(Guo et al. 2009)</ns0:ref>. More important, the SCP2-thiolase (type-1) encoded by SCPx plays a crucial role in the oxidation of the branched side chain of cholesterol to form bile acids in vertebrates <ns0:ref type='bibr'>(Ferdinandusse et al. 2000)</ns0:ref>,</ns0:p><ns0:p>while the physiological role has not been characterized in insects. Fortunately, the expression of the SCP2-thiolase (type-1) has also been detected in the prothoracic glands of Spodoptera littoralis, which are the main tissue producing the insect molting hormone <ns0:ref type='bibr' target='#b18'>(Takeuchi et al. 2004</ns0:ref>).</ns0:p><ns0:p>Thus, whether SCP2-thiolase (type-1) of the silkworm and other insects play role in the oxidation of cholesterol and participates in ecdysone synthesis needs further study.</ns0:p><ns0:p>In insects, juvenile hormone (JH) is an important regulator for growth and development <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007</ns0:ref>) and several thiolases have been cloned and suggested to be related to JH biosynthesis <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhu et al. 2016;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. Acetoacetyl-CoA thiolase catalyzes two molecules of acetyl-CoA to form acetoacetyl-CoA, which is the first enzyme in JH biosynthesis <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. The candidate acetoacetyl-CoA thiolases related to JH biosynthesis were cloned in B. mori and Helicoverpa armigera <ns0:ref type='bibr' target='#b5'>(Kinjoh et al. 2007;</ns0:ref><ns0:ref type='bibr'>Zhang et al. 2017)</ns0:ref>. In this study, those two acetoacetyl-CoA thiolase genes were classified as T2thiolases (BmorT2 and HarmT2), and they shared high sequence identities with the other T2thiolases (Table <ns0:ref type='table'>S4</ns0:ref>). For example, BmorT2-thiolase shared 82.71% sequence identity with</ns0:p><ns0:p>HarmT2. In H. armigera, temporal expressions of HarmT2-thiolase keep pace with JH fluctuations, and its expression can be inhibited by a juvenile hormone analog <ns0:ref type='bibr'>(Zhang et al. 2017)</ns0:ref>.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>The expression of BmorT2-thiolase was relatively abundant in the head where the JH synthetic gland, corpora allata (CA), is located (Fig. <ns0:ref type='figure' target='#fig_10'>6A and 6B</ns0:ref>). Interestingly, we found BmorTFE-and BmorT1-1-thiolase also showed high expressions in the larval head (Fig. <ns0:ref type='figure' target='#fig_10'>6A</ns0:ref>). In humans, TFEand T1-thiolases catalyze thiolytic cleavage of 3-ketoacyl-CoA into acetyl-CoA and acyl-CoA <ns0:ref type='bibr' target='#b0'>(Anbazhagan et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b24'>Xia et al. 2019)</ns0:ref>. However, T1-thiolase has been found synthetic and degradative activities in Ostrinia scapulalis (Lepidoptera: Crambidae). Therefore, whether T2-, TFE-and T1-thiolases were involved in JH biosynthesis is still worthy of experimental validation.</ns0:p><ns0:p>Acetyl-CoA is often used as the initial precursor for sex pheromone biosynthesis in insects <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010)</ns0:ref>. Degradative thiolases may supplement with sufficient acetyl-CoA for sex pheromone synthesis <ns0:ref type='bibr'>(Brabcova et al. 2015)</ns0:ref>. In this study, expression profiles of the thiolase genes were detected in the sex pheromone glands at different developmental stages in the silkworm (Fig. <ns0:ref type='figure' target='#fig_10'>6C</ns0:ref>). Relatively, BmorSCP2 (type-1) maintains a high level of expression in the PGs on day 4 of pupae to 24-h-old virgin female moth. However, its expression level was sharply declined in the mated female PGs (Fig. <ns0:ref type='figure' target='#fig_10'>6C</ns0:ref>). The previous study suggested that an over 6-h mating duration can terminate the sex pheromone production in the silkworm <ns0:ref type='bibr' target='#b1'>(Ando et al. 1996)</ns0:ref>. The expression pattern of BmorSCP2 (type-1) was consistent with sex pheromone production <ns0:ref type='bibr' target='#b8'>(Matsumoto 2010)</ns0:ref>, which suggested that it might be involved in sex pheromone biosynthesis. Generally, it is tempting to assume that a thiolase expressed in a specific tissue might obtain a specific role. Thus, the functional diversification and physiological roles of insect thiolases need yet further experimental validation.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In the present study, genome-wide identification of the thiolase gene family was conducted for the first time in multiple insect genomes. A total of 137 thiolase genes were identified in 20 insects from 7 orders. About 80% of the thiolase genes have 2 or more introns, and its exon/intron structures reserve diversification. Based on the prediction, all the thiolase proteins are located in the mitochondria, cytosol, or peroxisome, and thiolases of the same class often have similar cellular localization. Four highly conserved sequence fingerprints were found in the insect thiolase proteins, including CxS-, NEAF-, GHP-, and CxGGGxG-motifs. Homology modeling analysis indicated that 3D structures of the insect thiolases share similar to mammals, fishes, and microorganisms. Expression pattern analysis suggested some thiolase genes may be involved in Manuscript to be reviewed steroid metabolism, JH, and sex pheromone biosynthesis pathways in B. mori. These results might provide valuable information for the functional exploration of thiolase proteins in insects. The species tree was obtained from timetree database ( http://www.timetree.org/ ). Gain and loss analysis was conducted by Notung-2.9 software with default parameters. The orange and green vertical bars on branch presented gene gain and loss, respectively. The number in each node is gene count. Gene number of each species was presented in brackets.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed (H) Homo-dimer of BmorSCP2 (type-1). </ns0:p><ns0:note type='other'>Figure 3</ns0:note><ns0:note type='other'>Figure 6</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>were detected in various tissues and developmental sex pheromone gland of B. mori. Combining structural characteristics and expression patterns, the potential functions and involved PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020) physiological processes were hypothesized. The present study can help us understand the functional differentiation of thiolase genes in insects.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>of a certain gene in the cell. Protein subcellular localization is closely related to protein functions<ns0:ref type='bibr' target='#b13'>(Pereto, Lopez-Garcia &amp; Moreira 2005;</ns0:ref><ns0:ref type='bibr' target='#b23'>Wang et al. 2014)</ns0:ref>. Only when the protein is positioned correctly can it perform normal biological functions. In this study, subcellular localization of //www.genscript.com/tools/psort), which were cytosolic, mitochondrial, or peroxisomal enzymes (Fig.3C; Fig.S2B). Generally, most of the TFE-and T2-thiolase proteins were located in the mitochondrion, T1-and CT-thiolases were cytosolic, and SCP2-thiolases were peroxisomal proteins. Previous studies suggested that the mitochondrial and peroxisomal thiolase proteins were mainly involved in the fatty acid &#946;-oxidation pathway<ns0:ref type='bibr' target='#b13'>(Pereto, Lopez-Garcia &amp; Moreira 2005),</ns0:ref> </ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)Manuscript to be reviewed detected with magic fit in Swiss-PdbViewer(Guex &amp; Peitsch 1997). The RMSD values were ranged from 0.25 &#197; to 0.91 &#197; among BmorT2, BmorT1-1, DmelCT, and PxutAB-1, while</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Fig. 1 .</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig. 1. Phylogenetic tree of insect thiolases using the maximum-likelihood (ML) method.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 2 Fig. 2 .</ns0:head><ns0:label>22</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Fig. 3 .</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig. 3. Exon/intron structure and subcelluar localization analyses.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 4 Fig. 4 .</ns0:head><ns0:label>44</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 5 Fig. 5 .</ns0:head><ns0:label>55</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Fig. 6 .</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Fig. 6. Expression profiles of the thiolase genes in the silkworm.</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>). The annotated genes and genomes of B. mori were retrieved from SilkDB v3.0 (https://silkdb.bioinfotoolkits.net). The sequence information of Danaus plexippus and Heliconius melpomene were downloaded from http://metazoa.ensembl.org/.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Manduca</ns0:cell><ns0:cell>sexta</ns0:cell><ns0:cell>was</ns0:cell><ns0:cell>from</ns0:cell></ns0:row></ns0:table><ns0:note>ftp://ftp.bioinformatics.ksu.edu/pub/Manduca/OGS2/. The other sequences were retrieved from GenBank (https://www.ncbi.nlm.nih.gov/), including Papilio xuthus, Plutella xylostella, Culex quinquefasciatus, Anopheles gambiae, Drosophila melanogaster, Anoplophora glabripennis, Nicrophorus vespilloides, Tribolium castaneum, Halyomorpha halys, Acyrthosiphon pisum, Cimex lectularius, Diachasma alloeum, Apis mellifera, Bombus impatiens, Pediculus humanus, and Zootermopsis nevadensis.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Chromosome distribution, gene structure, and syntenic analysis</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='7'>evolution by ProtTest 3.2 (Darriba et al. 2011). Maximum-likelihood (ML) trees were</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>reconstructed using RAxML version 8.2.12 (Stamatakis 2014) with the most suitable model</ns0:cell></ns0:row><ns0:row><ns0:cell>(PROTGAMMAVTF)</ns0:cell><ns0:cell>and</ns0:cell><ns0:cell>500</ns0:cell><ns0:cell>bootstrap</ns0:cell><ns0:cell>replicates.</ns0:cell><ns0:cell>FigTree</ns0:cell><ns0:cell>v1.4.3</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>(http://tree.bio.ed.ac.uk/software/figtree/) was used for plotting the final phylogenetic tree. The</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>clustering and classification of the thiolase sequences in the ML tree were done using known</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>functional properties of H. sapiens and M. tuberculosis (Anbazhagan et al. 2014).</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='7'>To localize the thiolase genes on chromosomes, B. mori, H. melpomene, D. melanogaster, T.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>castaneum, and A. mellifera were selected because their genome sequences have been assembled</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>into chromosomes. Based on the GFF (General Feature Format) file of each species, every thiolase</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>gene was mapped to the corresponding chromosomes. Using protein sequences of the thiolases,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>the precise exon/intron structures were generated through BLAT search against the genome</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>sequences with Scipio server (https://www.webscipio.org/). The synteny events between two</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>species were detected by Multiple Collinearity Scan toolkit (MCScanX) with the default</ns0:cell></ns0:row><ns0:row><ns0:cell cols='7'>parameters (Wang et al. 2012). The syntenic map of B. mori and H. melpomene was constructed</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>with family_circle_plotter.java in MCScanX software.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:52282:3:0:NEW 23 Oct 2020)</ns0:note> </ns0:body> "

"College of Life Science China West Normal University Nanchong 637009 Sichuan China October 23, 2020 Dear editors, First, thank you for your timely processing of my manuscript. I also thank the reviewers for rigorous review. Based on the comments, I have carefully revised the manuscript and responded one-by-one. I hope that the manuscript is now suitable for publication in PeerJ. Dr. Shou-Min Fang Associate Professor of Genomics Editor comments (Kenta Nakai) MINOR REVISIONS Your revised manuscript has been reviewed by the same reviewer who made some comments in its previous version. This time, the reviewer gives only a few minor comments. I believe that I can accept the manuscript in the next revision without another round of review. Response: Thank you for your timely processing of my manuscript. I have carefully revised the manuscript, and answered to all the comments. Reviewer 2 (Anonymous) Basic reporting There are two new errors appeared in this revision: Line 249: “in in”=> “in” Line 272: structrue => structure Response: Agreed. The two errors were revised. Experimental design No comment. Validity of the findings No comment. Comments for the Author The author has addressed all my comments. Below is just a recommendation for easy understanding. The omission of data with no signals in Figs. 6B and 6C is better to be explicitly written in the main manuscript and/or the figure legend. Response: Thanks for your suggestion. The omission of data with no signals in Figs. 6B and 6C was explicitly written in the figure legend. It is as follows: (B) Expression patterns of the thiolase genes in the various tissues of the fifth-instar larvae. Expression signal of BmorT1-3 was not detected. It was not presented in the figure. (C) Expression profiles in the sex pheromone glands at different developmental stages of females. Expressions of BmorT1-2 and BmorT1-3 were not detectable in the PGs. 0-h and 24-h vergin: 0-h and 24-h old vergin adults after eclosion ; 3-h, 6-h and 9-h mated: female moths mated 3 hours, 6 hours and 9 hours. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Immunoglobulin A nephropathy (IgAN) is immune-mediated primary glomerulonephritis, which is the most common reason leading to renal failure worldwide, the exact pathogenesis of IgAN is not well defined. Accumulating evidence indicates that circular RNAs (circRNAs) play crucial roles in the immune disease by involving in the competing endogenous RNA (ceRNA) network mechanism. At present, the studies of the circRNA profiles and circRNA-associated ceRNA networks in the IgAN are still scarce. This study aimed to elucidate the potential roles of circRNA-associated ceRNA networks of peripheral blood mononuclear cells (PBMCs) in IgAN patients.Method. CircRNA sequencing was used to identify the differential expressed circRNAs (DEcircRNAs) of PBMCs in IgAN and healthy controls; limma packages from data sets GSE25590 and GSE73953 in the Gene Expression Omnibus (GEO) database, were used to identify differentially expressed micro RNAs (miRNAs) and message RNAs (mRNAs). A circRNA-miRNA-mRNA ceRNA network was constructed to further investigate the mechanisms of IgAN. Then, GO analysis and KEGG enrichment analyses were used to annotate the genes involved in the circRNAassociated ceRNA network. Further, Protein-protein interaction (PPI) networks were established to screen potential hub genes, by using Search Tool for the Retrieval of Interacting Genes/Proteins (STRING). Last, a quantitative real-time polymerase chain reaction (qRT-PCR) was applied to verify the hub genes in the ceRNA network.Result. A total of 145 circRNAs, 22 miRNAs, and 1117 mRNAs were differentially expressed in IgAN compared with controls (P&lt;0.05). A ceRNA network was constructed which contained 16 DEcircRNAs, 72 differential expressed mRNAs (DEmRNAs) and 11 differential expressed miRNAs (DEmiRNAs). KEGG pathway enrichment analysis illustrated the underlying biological functions of the ceRNA-associated genes, such as Nitrogen compound metabolic process, COPII-coated ER to Golgi transport vesicle, CAMP response element protein binding process (P&lt;0.01), meanwhile, Hepatitis B, GnRH signaling, and Prion disease were</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>IgA nephropathy (IgAN) has a global incidence exceeding 1.5 per 100000 persons, which is the most common type of glomerulonephritis worldwide <ns0:ref type='bibr' target='#b43'>(Wyatt &amp; Julian. 2013)</ns0:ref>. The clinical manifestation often presents as macroscopic hematuria following an upper respiratory or gastrointestinal infections. Diagnosed of IgAN is exclusively identified by renal biopsy with characteristics of predominant IgA deposition in the glomerular mesangium <ns0:ref type='bibr' target='#b43'>(Wyatt &amp; Julian. 2013)</ns0:ref>, and the 10-year risk of end-stage renal disease varies widely from 5% to 60% <ns0:ref type='bibr' target='#b25'>(Reich et al., 2007)</ns0:ref>. In the exploring of pathogenesis in IgAN, many mechanisms including the four-hit hypothesis had been proposed in the academic: the unknown upstream inducing the production of galactose-deficient IgA1(Gd-IgA1), which were identified as auto-antigen and targeted by autoantibody, formatted as immune complex in the circulation, deposited in kidney thus resulting in following renal injuries <ns0:ref type='bibr' target='#b35'>(Suzuki et al., 2011)</ns0:ref>. Noticeably, recurrence of IgAN is 50% over 5 years after kidney transplantation <ns0:ref type='bibr' target='#b22'>(Mulay et al., 2009)</ns0:ref>, while it's reported that several IgAN cases experienced a recovery after bone marrow transplantation <ns0:ref type='bibr' target='#b24'>(Park et al., 2008)</ns0:ref>, suggesting IgAN appears to be a systemic immune disease in which the kidneys are damaged as innocent bystanders <ns0:ref type='bibr' target='#b43'>(Wyatt &amp; Julian. 2013)</ns0:ref>. Recently, circular RNAs (circRNAs) arouse attention as one of the factors correlated with immune response, they are special types of single-stranded non-coding RNA which has a closed feature and without 3'poly (A) and 5'-cap structure <ns0:ref type='bibr' target='#b1'>(Chen. 2020;</ns0:ref><ns0:ref type='bibr' target='#b44'>Yan., 2020)</ns0:ref>. CircRNA is formed by the back-splicing of pre-mRNAs processing, in which a downstream 5&#8242; splice site is joined to an upstream 3&#8242; splice site in reverse order across an exon or exon. Investigator recently found that after B cell infected with Kaposi's sarcoma herpesvirus, hundreds of differentially expressed human circRNAs were identified, which suggests those newly expressed circRNAs may work as an antiviral mechanism by suppressing crucial viral genes <ns0:ref type='bibr' target='#b36'>(Tagawa et al., 2018)</ns0:ref>. The expression of circRNAs is usually stable, for it was resistant to exonucleases, and largely exported to the cytoplasm <ns0:ref type='bibr' target='#b1'>(Chen. 2020</ns0:ref>). The harboring miRNA recognition elements (MREs) of circRNAs could competitively bind to certain miRNAs, and regulate miRNA-mediated downstream target gene silencing at the post-transcriptional level, thus participate in the manipulation process of the target genes and been described involved in the competing endogenous RNA (ceRNA) hypothesis <ns0:ref type='bibr' target='#b37'>(Thomson &amp; Dinger. 2016</ns0:ref> ). The ceRNA hypothesis has been demonstrated involved in kidney disease, for example, novel_circ-0004153/rno-miR-1443p/Gpnmb ceRNA relationship has been reported involved in acute kidney injury of rat model <ns0:ref type='bibr' target='#b4'>(Cheng et al., 2019)</ns0:ref>; circHLA-C could function as a sponge to decoying miRNA-150, then promoted renal fibrosis by regulating fibrosis-associated genes in lupus nephritis <ns0:ref type='bibr' target='#b17'>(Luan et al., 2018)</ns0:ref>. However, the studies of the circRNA profiles and circRNAassociated ceRNA networks in the IgAN are still scarce, therefore, exploring the expression of non-coding RNAs including circRNAs and miRNAs may bring potential opportunities in the understanding mechanism of IgAN. PBMCs are blood cells with round nuclei that encompass a heterogeneous cell population (with 70-90% T cells, B cells, and NK cells) <ns0:ref type='bibr' target='#b28'>( Sallustio et al., 2019)</ns0:ref>, which originate from hematopoietic stem cells that reside in the bone marrow. In this study, PBMCs from IgAN patients and healthy controls were collected and sequenced with next-generation technology, and combined analysis with the genes expression profiles in the GEO database, a circRNA associated ceRNA network of IgAN was established (Fig. <ns0:ref type='figure'>1</ns0:ref>), aiming to comprehensively investigate the potential circRNA related molecular mechanism in IgAN.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>PBMCs collection and RNA preparation</ns0:head><ns0:p>This research was approved by the Ethics Committee of the Second Xiangya Hospital (IRB2018-S095). Participants were from The Second Xiangya Hospital, and all of them had provided signed informed consent. The diagnosis of IgAN was based on renal biopsy with the deposition of IgA in the glomerular mesangium <ns0:ref type='bibr' target='#b43'>(Wyatt &amp; Julian. 2013)</ns0:ref>, patients with any other systemic diseases or secondary IgAN were excluded. Peripheral blood from 3 IgAN patients and 3 healthy controls were centrifuged at 2000 rpm for 10 min, PBS was added to diluted the remaining blood, by Ficoll-Hypaque (GE Healthcare) gradient (2000 rpm for 30 min) to isolate the PBMCs by density separation. The isolated PBMCs were used for RNA isolation. Then, total RNA was extracted using Trizol reagent (Invitrogen, CA, USA) following the manufacturer's procedure. The total RNA quantity and purity were analyzed using Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent, CA, USA) with RNA integrity number &gt;7.0. Approximately 5 ug of total RNA were used to deplete ribosomal RNA according to the Ribo-Zero&#8482; rRNA Removal Kit (Illumina, San Diego, USA). After removing ribosomal RNAs, the remaining RNAs were fragmented into small pieces using divalent cations under high temperatures. Then the cleaved RNA fragments were reverse-transcribed to create the cDNA, which was next used to synthesized U-labeled second-stranded DNAs with E. coli DNA polymerase I, RNase H and dUTP. An A-base is then added to the blunt ends of each strand, preparing them for ligation to the indexed adapters. Each adapter contains a T-base overhang for ligating the adapter to the Atailed fragmented DNA. Single-or dual-index adapters are ligated to the fragments, and size selection was performed with AMPureXP beads. After the heat-labile UDG enzyme treatment of the U-labeled second-stranded DNAs, the ligated products were amplified with PCR with the following conditions: initial denaturation at 95&#8451; for 3 min; 8 cycles of denaturation at 98&#8451; for 15s, annealing at 60&#8451; for 15s, and extension at 72&#8451; for 30s; and then final extension at 72&#8451; for 5 min. The average insert size for the final cDNA library was 300 bp (&#177;50 bp).</ns0:p></ns0:div> <ns0:div><ns0:head>CircRNA sequencing and microarray data download</ns0:head><ns0:p>First, we performed the paired-end sequencing (PE150) for the library on Illumina Hiseq 4000 (LC Science Co., LTD., Hangzhou, China) following the vendor's recommended protocol. Cutadapt <ns0:ref type='bibr' target='#b21'>(Martin et al., 2011)</ns0:ref> was used to remove the reads that contained adaptor contamination, low quality bases and undetermined bases. Then sequence quality was verified using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). We used Bowtie2 <ns0:ref type='bibr' target='#b13'>(Langmead et al., 2012)</ns0:ref> and Hisat2 <ns0:ref type='bibr' target='#b9'>(Kim et al., 2015)</ns0:ref> to map reads to the genome of Homo sapiens (GRCh38). Remaining reads (unmapped reads) were still mapped to genome using tophat-fusion <ns0:ref type='bibr' target='#b10'>(Kim et al., 2011)</ns0:ref>. CIRCExplorer2 <ns0:ref type='bibr' target='#b49'>(Zhang et al., 2014 and</ns0:ref><ns0:ref type='bibr' target='#b47'>Zhang et al., 2016)</ns0:ref> was used to de novo assemble the mapped reads to circRNAs at first. Then, back splicing reads were identified in unmapped reads by tophat-fusion. All samples generated unique circular RNAs ( step1: the two ends of splice sites must be GU/AG; step 2: mismatch&#8804;2; step 3: back spliced junctions read&#8805;1; step 4: the distance between two splice sites on the genome is not more than 100kb). We next performed an electronic search in GEO (http://www.ncbi.nlm.nih.gov/geo/) database using the keywords 'IgA nephropathy, IgA nephritis, or Berger's disease' to identify studies involving samples with PBMCs from IgAN patients. Studies were included if they met the following criteria: (1) patients were diagnosed with biopsy-proven IgAN; (2) case-control studies and the number of cases and controls in each dataset must be &#8805;2; (3) all datasets were genome-wide; (4) complete microarray raw data were available. We excluded any animal or duplicated studies. The whole-genome raw expression data of included studies were downloaded from the GEO dataset. Therefore, the GSE73953 (mRNA) and GSE25590 (miRNA) microarray datasets that contain IgAN's PBMC related information were selected. GSE73953 including 15 IgAN patients, 2 healthy controls <ns0:ref type='bibr' target='#b23'>(Nagasawa et al., 2016)</ns0:ref>, which was based on the GPL4133 platform (Agilent-014850 Whole Human Genome Microarray 4x44K G4112F). GSE25590 contained 7 IgAN patients and 7 matched healthy samples <ns0:ref type='bibr' target='#b31'>(Serino et al., 2012)</ns0:ref> and was based on the GPL7731 platform (Agilent-019118 Human miRNA Microarray 2.0 G4470B).</ns0:p></ns0:div> <ns0:div><ns0:head>Differential expression analysis of circRNAs, miRNAs, and mRNAs</ns0:head><ns0:p>The unit of measurement for circRNA is Fragment Per Kilobase of exon per Million fragments mapped (FPKM). The differentially expressed circRNAs were selected with log2 (fold change) &gt;1 or log 2 (fold change) &lt;-1 (|log 2 (FC)|&gt;1) and with statistical significance (P-value &lt; 0.05) by R package-edgeR <ns0:ref type='bibr' target='#b27'>(Robinson et al., 2010)</ns0:ref>.</ns0:p><ns0:p>The expression status of GSE73953 and GSE25590 were obtained by R-package agimicrorna processing, normalizeBetweenArrays in the Limma package was used to normalize the expression values in the quartile <ns0:ref type='bibr' target='#b26'>(Ritchie et al., 2015)</ns0:ref>. Finally, log2 transformation was used to obtain standardized expression values. Significance of differential miRNAs and mRNAs expression were defined by |log 2 (FC)| &gt;1, q-value &lt;0.05 and P-value &lt; 0.05. The heat maps were made by Pheatmap software, and the volcano plot was made by Ggplot2 software.</ns0:p></ns0:div> <ns0:div><ns0:head>Function enrichment analysis of the differential expression genes.</ns0:head><ns0:p>To investigate the potential function of the differential expression genes (DEGs) in the PBMCs of IgAN, GO (http://www.geneontology.org/ ) and KEGG (http://www.kegg.jp/) pathway enrichment analysis of the identified DEGs were performed. The identified DEGs were classified in terms of the biological process (BP), molecular function (MF), and cellular component (CC) categories <ns0:ref type='bibr' target='#b0'>(Ashburner et al., 2000)</ns0:ref>. KEGG analysis was utilized to interpret the potential functions and pathways of the DEGs <ns0:ref type='bibr' target='#b8'>(Kanehisa et al., 2010)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>CeRNA network construction and functional analysis of the genes in the network</ns0:head><ns0:p>We analyzed the circRNAs and mRNAs which are expressed in significantly different levels between the IgAN patients and the control group. The sequences of circRNAs, miRNAs, and mRNAs were screened to search the potential MREs. We used 4 databases including miRanda (http://www.microrna.org/microrna/home.do), PITA (http://genie.weizmann.ac.il/pubs/mir07/mir07_dyn_data.html), RNA22 (http://cm.jefferson.edu/rna22/Interactive/) and TarPmiR (http://hulab.ucf.edu/research/projects/miRNA/TarPmiR/) to find miRNAs-circRNAs relationship, only overlapping genes were selected as candidate. Next, we used 5 kinds of databases including MiRDB (http://mirdb.org/), TargetScan (http://www.targetscan.org/), miRanda, miRMap (https://mirmap.ezlab.org/) and miTarBase (http://mirtarbase.mbc.nctu.edu.tw/) to screen mRNAs corresponding targets of miRNAs, and retained the interaction genes which are identified at least 4 of the databases. The circRNA-miRNA-mRNA regulatory network was constructed using a combination of circRNA-miRNA and miRNA-mRNA interaction. Finally, the network was visualized and mapped using Cytoscape v3.7.0 software. All of the DEGs coming from co-expression and prediction ceRNA network were united for genes functional annotation and enrichment analysis, to investigate the potential molecular mechanism of the genes in IgAN.</ns0:p></ns0:div> <ns0:div><ns0:head>Protein-protein Interaction network analysis and qRT-PCR of the hub genes.</ns0:head><ns0:p>To assess the interactions between DEGs in the ceRNA network, we used the Search Tool for the Retrieval of Interacting Genes (STRING, https://stringdb.org/) online tool, which can provide comprehensive interactions among proteins and genes, and established a Protein-Protein Interaction (PPI) network of our ceRNA network. A Required Confidence (combined score) &gt;0.7 was used to visualize the established PPI network. Then we used Cytoscape to analyze the topology of genes in the PPI network, the central proteins in the network are found by using the scale-free nature of the interaction PPI network. Also, 4 IgAN patients and 4 healthy controls in the Second Xiangya Hospitals that signed informed consent, were recruited for qRT-PCR validation of the expression of hub genes. The way of PBMCs collection and RNA extraction were mentioned previously, the qRT-PCR reaction was performed using the UltraSYBR Mixture (Cwbiotech, China), in the Pikoreal PCR Detection System (Thermo, USA) with the following conditions: 95&#8451; for 10 mins, then followed by 40 cycles of 95&#8451; for 15s and 60&#8451; for the 30s. The quantitative primers of ankyrin repeat and SOCS box containing 16 (ASB16) , major histocompatibility complex, class I, B (HLA-B), tripartite motif containing 21(TRIM21), SEC24 homolog C, COPII coat complex component (SEC24C) , were designed and synthesized by Sangon Biotech (Sangon Biotech, China) and are listed in Table <ns0:ref type='table'>S1</ns0:ref>, &#946;-actin was used as the house-keeping gene for normalization, qRT-PCR relative fold change results were calculated using the 2-&#9651;&#9651;Ct method.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Overview of circRNA-Seq and identification of DEcircRNAs</ns0:head><ns0:p>After the removal of low-quality reads, adapters, poly-N&gt;5%, and other contaminant-containing reads from the raw data, clean reads of circRNA-sequencing (circRNA-seq) were obtained.</ns0:p><ns0:p>In the present study, a total of 16,940 circRNA transcripts were identified, among them, 145 circRNAs were differentially expressed in the IgAN patients compared to the healthy controls, including 112 up-regulated and 33 down-regulated circRNAs. The basic characteristic of the top 10 DEcircRNAs are listed in Table <ns0:ref type='table'>1</ns0:ref>, the most up-regulated circRNA was hsa_circ_0038725 with 6.03 log 2 (FC), and the most down-regulated circRNA was circRNA11137 (gene symbol: EMB) with -6.01 log 2 (FC), those DEcircRNAs were used for the subsequent analysis (Table <ns0:ref type='table' target='#tab_2'>S2</ns0:ref>).</ns0:p><ns0:p>A heat-map of DEcircRNAs was illustrated in Fig. <ns0:ref type='figure' target='#fig_1'>2A</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Identification of DEmiRNAs and DEmRNAs</ns0:head><ns0:p>A total of 22 DEmiRNAs were screened from the GSE25590 dataset, including 18 up-regulated and 4 down-regulated miRNAs in IgAN patients (Table <ns0:ref type='table'>S3</ns0:ref>). In the GSE73953 dataset, a total of 1117 DEmRNAs were identified, including 522 up-regulated and 595 down-regulated mRNAs in IgAN patients (Table <ns0:ref type='table'>S4</ns0:ref>). The DEmiRNAs and DEmRNAs between IgAN patients and controls are illustrated as heat-map in Fig. <ns0:ref type='figure' target='#fig_1'>2BC</ns0:ref>. The statistical analysis of the differential expressed circRNAs, miRNAs and mRNAs were summarized in Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>. A volcano plot was utilized to visualize the statistical significance of DEGs between the IgAN patients and the controls (Fig. <ns0:ref type='figure'>3A</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Pathway enrichment analysis of the DEGs</ns0:head><ns0:p>KEGG analysis illustrated those down-regulated DEGs were mainly enrichment in the NOD-like receptor signaling pathway (hsa04621, 19 genes were enriched with P&lt;10 -6 ), Fc gamma Receptor-mediated phagocytosis (hsa04666, 13 genes were enriched with P&lt;10 -6 ), Measles pathway (hsa05162, 15 genes were enriched with P&lt;10 -5 ), etc (Fig. <ns0:ref type='figure'>3B</ns0:ref>). Those upregulated DEGs were mainly enriched in the Mineral absorption pathway (hsa04978, 8 genes were enriched with P&lt;10 -5 ), Maturity onset diabetes of the young pathway (hsa04950, 4 genes were enriched with P&lt;10 -4 ), Linoleic acid metabolism pathway (hsa00591, 4 genes were enriched with P&lt;10 -4 ), etc (Fig. <ns0:ref type='figure'>3C</ns0:ref>). The results of GO analysis shows that the down-regulated DEGs were enriched in functions associated with Leukocyte activation involved in immune response (GO:0002366, 55 genes were enriched with P&lt;10 -15 ), Cytoplasm (GO:0005737, 382 genes were enriched with P&lt;10 -16 ), Identical protein binding (GO:0042802, 80 genes were enriched with P&lt;10 -5 ). The upregulated DEGs were enriched in function associated with Cellular response to copper ion (GO:0071280, 10 genes were enriched with P&lt;10 -11 ), Integral component of postsynaptic specialization membrane (GO:0099060, 8 genes were enriched with P&lt;10 -5 ), Hormone activity (GO:0005179, 8 genes were enriched with P&lt;10 -4 ) (Table <ns0:ref type='table'>S5</ns0:ref> and Fig. <ns0:ref type='figure'>S1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Construction of the ceRNA network and functional annotation of genes in ceRNA network</ns0:head><ns0:p>According to the ceRNA hypothesis, the members of ceRNA (circRNAs, miRNAs, and mRNAs) compete for the same miRNA response elements (MREs) to regulate each other. To establish a circRNA-miRNA-mRNA ceRNA network, we found 32 DEmiRNAs-DEcircRNAs pairs according to the circRNAs-miRNAs corresponding relationship, which overlapped in 4 databases including miRanda, PITA, RNA22, and TarPmiR. (Fig. <ns0:ref type='figure'>4A</ns0:ref>) Then, we used 5 kinds of databases including MiRDB, TargetScan, miRanda, miRMap and miTarBase to identify mRNAs corresponding targets of miRNAs, and 12 DEmRNAs-DEmiRNAs relationship pairs which are overlapping at least 4 of the databases. Overlapping datasets were visualized using Venn diagrams (Fig. <ns0:ref type='figure'>4B</ns0:ref>).</ns0:p><ns0:p>Finally, we constructed a ceRNA network based on 16 circRNA nodes, 11 miRNA nodes, and 72 mRNA nodes (Fig. <ns0:ref type='figure'>4C</ns0:ref>). These RNA interactions may serve as a novel perspective for exploring the underlying mechanism of IgAN. More details are listed in Tables S6.</ns0:p><ns0:p>All of the related genes coming from the established ceRNA network were united for genes functional annotation enrichment analysis to investigate the potential roles of these differentially expressed circRNAs and mRNAs (Table <ns0:ref type='table'>S7</ns0:ref>). The KEGG analysis showed that the pathway of genes in the network were mostly related with Hepatitis B (hsa05161, 4 genes were enriched with P=0.003), GnRH signaling pathway (hsa04912, 3 genes were enriched with P=0.005), and Prion disease (hsa05020, 2 genes were enriched with P=0.007) (Fig. <ns0:ref type='figure'>4D</ns0:ref>). The most significant biological process is the regulation of Nitrogen compound metabolic process (GO:0051171, 35 genes were enriched with P&lt;10 -4 ), Vesicle targeting (GO:0006903, 4 genes were enriched with P&lt;10 -4 ), Vesicle budding from the membrane (GO:0006900, 4 genes were enriched with P&lt;10 -4 ) (Fig. <ns0:ref type='figure'>4E</ns0:ref>). The most relevant cellular component are COPII-coated ER to Golgi transport vesicle (GO:0030134, 5 genes were enriched with P&lt;10 -4 ), Coated vesicle (GO:0030135, 7 genes were enriched with P&lt;10 -4 ), and Heterochromatin (GO:0000792, 4 genes were enriched with P&lt;10 -3 ) (Fig. <ns0:ref type='figure'>4F</ns0:ref>). CAMP response element binding (GO:0008140, 2 genes were enriched with P&lt;10 -3 ), Non-membrane spanning protein tyrosine kinase activity (GO:0004715, 3 genes were enriched with P&lt;10 -5 ), and Calcium-dependent phospholipid binding (GO:0005544, 3 genes were enriched with P=0.001) were the most enrichment molecular function (Fig . <ns0:ref type='figure'>4G</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Protein-protein Interaction network analysis and validation of the hub genes</ns0:head><ns0:p>Furthermore, the genes were selected which has potential functional associated (in the STRING database) with other genes in the ceRNA network, and then established PPI networks. The PPI network involving 7 nodes and 4 edges (Fig. <ns0:ref type='figure'>5A</ns0:ref>). Among then, ASB16, HLA-B, TRIM21, SEC24C were relatively highly connected in the PPI network (associated with more than 3 molecules) and were considered to be the potential hub genes <ns0:ref type='bibr' target='#b7'>(He &amp; Zhang. 2006</ns0:ref>). qRT-PCR was utilized to validates those hub genes in the PPI networks (Fig. <ns0:ref type='figure'>5B</ns0:ref>/C/D/E), as a result, ASB16 was confirmed significant up-regulated (P=0.015) (Fig. <ns0:ref type='figure'>5B</ns0:ref>) and SEC24C was significantly downregulated in IgAN patients (P=0.021) (Fig. <ns0:ref type='figure'>5D</ns0:ref>). This indicates the up-regulated ASB16 and downregulated SEC24C may play important roles in the pathogenesis and pathological process in IgAN.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>In the presenting study, we successfully identified DEcircRNAs in PBMCs of IgAN (P&lt;0.05), combined with the identified DEmiRNAs and DEmRNAs (P&lt;0.05) in the RNA profiles in the GEO database, a ceRNA network was constructed with 16 DEcircRNAs, 11 DEmiRNAs, and 72 DEmRNAs according to the ceRNA hypothesis. The function analysis revealed the genes in the ceRNA network mainly involved in the Nitrogen compound metabolic process, COPII-coated ER to Golgi transport vesicle, CAMP response element protein binding process in the GO term annotation, meanwhile, it's showed that the Hepatitis B, GnRH signaling pathway, Prion disease and Human T-cell leukemia virus 1 infection were the most enriched KEGG pathway, and interestingly, 3 of them were closely related to the virus infection. The protein interaction analysis revealed that ASB16, SEC24C were considered as hub genes in the ceRNA networks. Therefore, our results have firstly revealed the potential mechanism of circRNAs-associated ceRNA networks in IgAN.</ns0:p><ns0:p>Since the 1990s, many studies have found that competing endogenous circRNAs could work as endogenous miRNA 'sponges', which take part in regulating miRNA related downstream gene expression <ns0:ref type='bibr' target='#b1'>(Chen. 2020</ns0:ref>). Ongoing investigation illustrated that significant expression of circRNAs was observed in platelets, hematopoietic progenitor cell differentiation into lymphoid and myeloid cells, and been observed they have functionally involved in neuronal function, cell proliferation, and innate immunity. But the pathological function of circRNA is still largely unknown, for instances, rheumatoid arthritis is a chronic and systemic autoimmune disease with unknown etiology <ns0:ref type='bibr' target='#b5'>(Coutant and Miossec. 2020)</ns0:ref>, multiple groups have independently identified several differentially expressed circRNAs in PBMCs between rheumatoid arthritis patients and healthy controls <ns0:ref type='bibr' target='#b42'>(Wen JT et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b45'>Yang X et al., 2019)</ns0:ref>. Similarly, systemic lupus erythematosus (SLE) in one of the autoimmune disease and could also be manifested as glomerulonephritis, several circRNAs including hsa_circ_0044235 have been found to be significantly decreased in PBMCs, and are proposed as potential negative biomarkers for SLE diagnosis <ns0:ref type='bibr' target='#b19'>(Luo Q et al., 2019)</ns0:ref>.</ns0:p><ns0:p>In the field of IgAN, several studies have investigated the non-coding RNA in PBMCs. For examples, UDP-N-acetyl-&#945;-d-galactosamine: polypeptide N-acetylgalactosaminyltransferase 2 (GALNT2) could catalysis the attachment of N-acetylgalactosamine (GalNAc) to the serine/threonine of the hinge region of the molecule, which functions in the early step in Olinked glycosylation in IgA1, the significantly down-regulated GALNT2 in the PBMCs thus correlated with the Gd-IgA1 level in IgAN, Serino et al had reported that miRNA let-7b not only involving influence the expression of GALNT2, it could also play roles as a biomarker for detecting primary IgAN <ns0:ref type='bibr' target='#b30'>(Serino et al., 2016)</ns0:ref>. In the presenting study, we have first time identified 112 up-regulated and 33 down-regulated circRNAs in IgAN, the interaction regulatory mechanism of the related circRNA have been carefully predicted. Consistent with the literature, the GALNT2 were significantly down-regulated in the PBMCs of IgAN in our study, while the hsa_circ_0070562&#12289;hsa_circ_0066719&#12289;hsa_circ_0073237 and circRNA3302 were identified as ceRNAs of miR765, therefore could participate in the manipulation the expression of target GALNT2. However, unexpectedly all the 3 circRNAs were up-regulated in the study, therefore didn't follow the classic circRNA (down in IgAN)-miRNA (up in IgAN)-mRNA (down in IgAN) interaction mechanism, therefore, it remains unknown whether the regulation of GLANT2 in IgAN is influenced by the circRNAs in PBMCs. It's should be mentioned that Thomson &amp; Dinge used to discuss in 2016 <ns0:ref type='bibr' target='#b37'>(Thomson &amp; Dinger. 2016</ns0:ref> ) , the underlying argument opinion which against the ceRNA hypothesis is that, the change in expression of an individual noncoding RNA may only impact relatively a small fraction of the target mRNA abundance. Therefore, the real mechanism of ceRNA interaction needs further molecular biology experiments, to carefully verify this non-coding factors on the molecules in the disease, such as the previous literature, using knock-out or overexpression method to illustrated a TGF-&#946;/Smad3interacting-lncRNA inhibits renal fibrogenesis in IgAN <ns0:ref type='bibr' target='#b17'>(Wang et al., 2018)</ns0:ref>.</ns0:p><ns0:p>Next, we identified the hub genes of the ceRNA network. Normally most proteins interact with only a few other proteins, while a small number of proteins that have many interaction partners (the genes encoding them, namely hub genes) <ns0:ref type='bibr' target='#b7'>(He &amp; Zhang. 2006)</ns0:ref>, and usually play important roles in the protein network, exploring of the hub genes may bring further insights into the presented ceRNA network. In our study, ASB16 and SEC24C were confirmed by qRT-PCR to have the same gene expression direction as detected by microarray, thus were identified as potentially the hub genes of the ceRNA networks. ASB16 is a member of the ankyrin repeat and SOCS box-containing (ASB) family of proteins, which may be a substrate-recognition component of E3 ubiquitin-protein ligase complex that mediates the ubiquitination and subsequent proteasomal degradation of target proteins <ns0:ref type='bibr' target='#b15'>(Liu P et al., 2019)</ns0:ref>; SEC24C is a subunit of COPII complex, which is the coat protein complex responsible for vesicle budding from the endoplasmic reticulum (ER), and plays a role in shaping the vesicle <ns0:ref type='bibr' target='#b32'>(Subramanian et al., 2019)</ns0:ref>, as well as in cargo selection and concentration, It's been reported SEC24C were significantly upregulated during B cell differentiation into a plasma cell and was assumed associated with the antibody-producing related preprocessing <ns0:ref type='bibr' target='#b11'>(Kirk et al., 2010)</ns0:ref>.</ns0:p><ns0:p>In our study, the functional analysis of the DEGs in the PBMCs according to the genes profiles GSE73953 had shown, the NOD-like receptor signaling pathway, Fc gamma R-mediated phagocytosis, Mineral_absorption pathway were the most enriched pathway. In another study that had comprehensively analyzed 3 microarray data of PBMC in IgAN <ns0:ref type='bibr' target='#b39'>(Liu D et al., 2019)</ns0:ref>, the KEGG enrichment pathway analysis was confirmed with enriched in human T-cell leukemia virus 1 infection, but also including proteoglycans in cancer and intestinal immune network for IgA production. However, noticeably, our study had suggested that the genes in the circRNAassociated ceRNA network were relatively more enriched in virus infection-related pathway. The KEGG analysis indicated that the Hepatitis B, Prion disease, T-cell leukemia virus 1 infection, and cytomegalovirus infection were significantly enriched in the genes of the ceRNA network. It's well known that viruses could manipulate host gene expression, and among the various mechanism of this pathological process, the most efficient way is a non-coding RNAs strategy <ns0:ref type='bibr' target='#b38'>(Tycowski et al., 2015)</ns0:ref>. Many kinds of research have revealed that viruses produce non-coding RNAs could serve as miRNA sponges, play as an adaptation to the ceRNA, and increased ceRNA activity <ns0:ref type='bibr' target='#b38'>(Tycowski et al., 2015)</ns0:ref>. But the mechanism investigation of virus-related ceRNA in IgAN was less explored up till now. In the related literature, the relationship between IgAN and HBV has been often noticed, it's reported 17-18% of overall IgAN patients in China were HBV carriers <ns0:ref type='bibr' target='#b50'>(Zhao., 2020)</ns0:ref>. Presently it's unknown that, whether the humoral immune response inducing by HBAg-HBAb immune complex causing the dysregulation of IgA producing or the mesangial cell injuries directly causing by HBV in situ were most involved in the pathogenesis of IgAN. While in another way, considering the prevalence of HBV infection could reach 5-7.99% of the population in China, therefore, whether IgAN was secondary to HBV infection, or presenting as a geographical coincidence, both of them should be further carefully demonstrated. Besides, previous research of IgAN has revealed that in vitro EBVtransformed peripheral-blood cells from healthy individuals produce almost exclusively IgA1 subclass. In a recent study of Zachova et al <ns0:ref type='bibr' target='#b46'>(Zachova et al., 2020)</ns0:ref>, which indicates EBV infection may be involved in the pathogenesis of IgAN. They found that B cells and their IgA + subpopulation in peripheral blood of IgAN patients displayed a significantly higher frequency of EBV infection compared to the controls and displayed increased expression of homing receptors for targeting the upper respiratory tract. Similarly, in the animal model of IgAN, by oral immunization with the Sendai virus, a parainfluenza virus likewise to human respiratory tract viruses can induce IgAN in mice <ns0:ref type='bibr' target='#b34'>(Suzuki &amp; Suzuki. 2018)</ns0:ref>. These in vivo and in vitro phenotypes had a raised speculation, that different types of viruses infection have shared some similar molecular biological processes, which aroused response of circRNAs expression and involved in the disease progression of IgAN.</ns0:p><ns0:p>Moreover, the GO enrichment analysis has shown that the circRNAs-associated genes' most related cellular component were intracellular membrane-bounded organelle and cytoplasm, especially the process associated with vesicle transport between the ER to Golgi. Although the molecule function of intracellular vesicle transport is relatively less discuss in IgAN, however, it's naturally closely related to the immunoglobulin's processing <ns0:ref type='bibr' target='#b11'>(Kirk et al., 2010)</ns0:ref>. For instance, the C1GALT1, which function as a crucial enzyme of process IgA1 galactosylation, normally C1GALT1 is synthesized in ER under the help of molecular chaperone Cosmc, and it's necessary to be packaged in vesicles and transport to Golgi then finish the galactosylation of IgA1 in Golgi <ns0:ref type='bibr' target='#b6'>(Cummings. 2019)</ns0:ref>. In IgAN, the significant deficiency of C1GALT1 (expression or/and activity) have resulted in the elevated of Gd-IgA1, and directly correlated with serum levels of Gd-IgA1 in IgAN patients which have been demonstrated by many works of literature <ns0:ref type='bibr' target='#b12'>(Lai et al., 2016)</ns0:ref>. In our previous study, Wang et al <ns0:ref type='bibr' target='#b39'>(Wang et al., 2019)</ns0:ref> have illustrated the Golgi matrix protein 130 plays the role of docking C1GALT1 vesicles in Golgi, the dysregulation with this molecular process had been demonstrated associated with Gd-IgA1 in IgAN, and firstly aroused the attention about the molecular process of ER-Golgi transportation in IgAN. Combined with our findings according to the ceRNA network, exploring the virus infectionrelated intracellular vesicle transport mechanism might be an interesting direction of IgAN in the future.</ns0:p><ns0:p>Lastly, our study was designed as combined with the patient's data from Italian, Japanese and Chinese populations, for better avoiding the bias brought by single-center sample, but in another way, the different race' sample sources bring the heterogeneity challenges to the comprehensive analysis of genes profile and validation of hub genes. For example, the molecular in this ceRNA network was relatively less than other ceRNA comprehensively studies, and we couldn't find ceRNA interaction pair with certain order (up-down-up or down-up-down) to validates the potential associated relationship corresponding with GALNT, ASB16 or SEC24C. The small sample size of the hub gene verification is also the one of limitations of this study, validation of a larger sample size will be needed, and exploring the potential capability of circRNAs as biomarkers of IgAN would be a research in the future. Future research should further identify the hub molecules with therapeutic potential in IgAN and conducted the overexpression or inhibition experiments of related ceRNAs, and observing the reciprocal relationship of target genes, for a better understanding of the molecular pathogenesis of IgAN.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>In summary, we successfully identified IgAN associated circRNAs using RNA-seq analysis, and elucidated the circRNA-associated ceRNA networks of IgAN through the integrated analysis of RNA expression profile. To our knowledge, this is the first report examining the expression of circRNAs in IgAN. These findings had expanded our understanding of circRNAs-associated ceRNA networks in IgAN, a future exploring the related molecular regulatory mechanism will be needed for a better understanding IgAN.</ns0:p></ns0:div> <ns0:div><ns0:head>Figure 1(on next page)</ns0:head><ns0:p>Main steps of the construction of the circRNA-associated ceRNA network in IgAN.</ns0:p><ns0:p>Step 1: The circRNA expression profiles of PBMCs in IgAN patients and healthy controls were explored by high-throughput sequencing, expression profiles of miRNAs and mRNAs were obtained from the data sets in the GEO database. Step 2: Significantly differential circRNAs, miRNAs and mRNAs were identified, the GO and KEGG enrichment analysis were conducted to reveal the functions of differential expressed mRNAs.</ns0:p><ns0:p>Step 3: The ceRNA interaction relationships were predicted using online tools, then constructed the ceRNA network. Steps Manuscript to be reviewed Statistical analysis of deferentially expressed circRNAs, miRNAs and mRNAs.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50257:1:1:NEW 15 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>4: Functional analysis of the related genes in the ceRNA network. Identification hub genes of the ceRNA network and validate the expression with qRT-PCR. IgAN: IgA nephropathy.DEcircRNA: differentially expressed circularRNA. DEmiRNA: differentially expressed microRNA. DEmRNA: differentially expressed mRNA. PBMCs: peripheral blood mononuclear cells. CeRNA: competing endogenous RNA. GO: Gene Ontology. KEGG: Kyoto Encyclopedia of Genes and Genomes.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='21,42.52,334.42,525.00,408.00' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='22,42.52,323.18,525.00,391.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='24,42.52,70.87,525.00,396.00' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Gene symbol Chromosome Strand Regulation log 2 (FC) P-value</ns0:head><ns0:label /><ns0:figDesc /><ns0:table><ns0:row><ns0:cell cols='2'>circRNA9482 hsa_circ_0038725</ns0:cell><ns0:cell>IL4R</ns0:cell><ns0:cell>Chr16</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>6.03134 0.01027</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA11137</ns0:cell><ns0:cell>novel_circ</ns0:cell><ns0:cell>EMB</ns0:cell><ns0:cell>Chr5</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>6.01153 0.00447</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA12492</ns0:cell><ns0:cell>novel_circ</ns0:cell><ns0:cell>SP140L</ns0:cell><ns0:cell>Chr2</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>5.98330 0.00015</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA4713</ns0:cell><ns0:cell>novel_circ</ns0:cell><ns0:cell>TLN1</ns0:cell><ns0:cell>Chr9</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>Down</ns0:cell><ns0:cell>5.69939 0.02184</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA10024</ns0:cell><ns0:cell>novel_circ</ns0:cell><ns0:cell>RPS6KA5</ns0:cell><ns0:cell>Chr14</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>5.59830 0.00903</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>circRNA8550 hsa_circ_0028670</ns0:cell><ns0:cell>TAOK3</ns0:cell><ns0:cell>Chr12</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>5.46949 0.00979</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>circRNA10745 hsa_circ_0007612</ns0:cell><ns0:cell>ORC5</ns0:cell><ns0:cell>Chr7</ns0:cell><ns0:cell>-</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>5.16406 0.03555</ns0:cell></ns0:row><ns0:row><ns0:cell>circRNA11518</ns0:cell><ns0:cell>novel_circ</ns0:cell><ns0:cell>HERC3</ns0:cell><ns0:cell>Chr4</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>5.13206 0.00143</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>circRNA4236 hsa_circ_0004893</ns0:cell><ns0:cell>PTP4A2</ns0:cell><ns0:cell>Chr1</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>Down</ns0:cell><ns0:cell>4.74473 0.00931</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>circRNA11045 hsa_circ_0009096</ns0:cell><ns0:cell>UTRN</ns0:cell><ns0:cell>Chr6</ns0:cell><ns0:cell>+</ns0:cell><ns0:cell>Up</ns0:cell><ns0:cell>4.49705 0.03553</ns0:cell></ns0:row></ns0:table><ns0:note>2PeerJ reviewing PDF | (2020:06:50257:1:1:NEW 15 Oct 2020)</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 . Statistical analysis of differentially expressed circRNAs, miRNAs and mRNAs.</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /></ns0:figure> <ns0:note place='foot' n='2'>2PeerJ reviewing PDF | (2020:06:50257:1:1:NEW 15 Oct 2020)</ns0:note> </ns0:body> "

" The Second Xiangya Hospital, Central South University No.139 Renming Road Changsha 410011 [email protected] Oct 15th, 2020 Dear Editors We really thank the reviewers for their helpful and valuable comments on the manuscript, we have carefully edited the manuscript to address their concerns. We hope that the manuscript is now suitable for publication in PeerJ. If you have any questions, please don’t hesitate to contact us. Dr. Hong Liu On behalf of all authors. Reviewer 1 ( Raquel Echavarria) Basic reporting The manuscript ''Comprehensive analysis of circRNA expression profiles and circRNA associated competing endogenous RNA networks in IgA nephropathy'' is interesting, well-structured (conforms to PeerJ standards), and professional English language is used throughout. However, the whole manuscript could use some editing to make it more clear and unambiguous. Some specific changes are suggested below: Line 85 suggest -> suggesting Thank you! We have changed the “suggest” to “suggesting” (revised line 87). Line 98 have providing evidence -> provide evidence Agreed, and then we deleted this sentence “Those studies have providing evidences that these miRNA may have universal role in…” according to the rest of the other comments. Thank you! Line 102 procession -> processing We apologize for the error. We have changed the “procession” to “processing” (revised line 93). Thank you! Line 104 remove: therefore the circRNAs been produced Agreed, we have removed it (revised line 94), thank you! Line 104 are -> is Thank you for pointing this out! We have revised the text (revised line 98). Line 107 regulating -> regulate Thank you, we have changed the word to “regulate” (revised line 100). Line 108 please check: thus been discrible involed Sorry for the spelling mistake, we have changed to “described involved” (revised line 102), thank you! Line 109 have -> has Agreed, we should use the third person singular for “hypothesis”, we have changed to “has” (revised line 104), thank you! Line 110 remove: ''many'' Agreed, it’ will be more clear and we have removed the“many”(revised line 104). Thank you! Line 112/116 please check: ''; novel_circ-113 0004153/rno-miR-1443p/Gpnmb ceRNA relationship have been reported involved in acute kidney injury of rat model (Cheng et al., 2019); in another way, human antiviral circRNAs are activated for responding to Kaposi's sarcoma herpesvirus infection, which been proved work as an antiviral mechanism by suppressing crucial viral genes'' Thank you very much for pointing out these sentences! We have split the sentence and revised the text, to address your concerns, and hope that it is now clearer. Please see the revised line 104-111 for introducing ceRNA hypothesis, “The ceRNA hypothesis has been demonstrated involved in kidney disease, for example, novel_circ-0004153/rno-miR-1443p/Gpnmb ceRNA relationship has been reported involved in acute kidney injury of rat model (Cheng et al., 2019); circHLA-C could function as a sponge to decoying miRNA-150, then promoted renal fibrosis by regulating fibrosis-associated genes in lupus nephritis (Luan et al., 2018).” And the revised line 92-97 have introduced the circRNA related mechanism, “CircRNA is formed by the back-splicing of pre-mRNAs processing, in which a downstream 5′ splice site is joined to an upstream 3′ splice site in reverse order across an exon or exon. Investigator recently found that after B cell infected with Kaposi's sarcoma herpesvirus, hundreds of differentially expressed human circRNAs were identified, which suggests those newly expressed circRNAs may work as an antiviral mechanism by suppressing crucial viral genes (Tagawa et al., 2018).”. Thank you! Line 122 the change of -> changes in Thank you for pointing this out! We have modified this expression according to your next comment “Line 123 please check” Line 123 please check: ''Because PBMCs are full of lymphocytes (with 70-90% T cells, B cells, and NK cells) could partially represent the change of the immune system, it had been broadly investigated in IgAN'' Thank you! We have deleted “could partially represent the change of the immune system, it had been broadly investigated in IgAN” , which may be unclear and unnecessary, and changed the sentence to “PBMCs are blood cells with round nuclei that encompass a heterogeneous cell population (with 70-90% T cells, B cells, and NK cells)( Sallustio et al., 2019), which originate from hematopoietic stem cells that reside in the bone marrow. In this study, PBMCs from IgAN patients and healthy controls were collected….” (revised line 112). Thank you for point it out! Line 127 remove: ''underlying'' We have removed the “underlying” and it looks more concise (revised line 117), thank you! Line 135/136 patients were excluded from any other systemic diseases or secondary IgAN -> patients with any other systemic diseases or secondary IgAN were excluded Agreed, we have modified accordingly (revised line 126), thank you for helping us make the sentence more elegant! Line 141 of -> using This phrase was modified accordingly (revised line 131), thank you! Line 143 ugs -> g Sorry for this mistake. We have corrected the typo (revised line 132), thank you! Line 144 left -> remaining We have modified this expression (revised line 134), thank you! Line 146 synthesis -> synthesized We have changed the “synthesis” to“synthesized”(revised line 137)，thank you! Line 164 denovo -> de novo Thank you! This phrase was modified accordingly (revised line 154), thank you! Line 165 ''; '' -> ''. '' We have changed the “;” to “.” (revised line 155), thank you! Line 166 remove: ''were'' We have removed the “were” (revised line 156), thank you! Line 187 are -> were We have modified this expression (revised line 177), thank you! Line 188/189 please check: ''Normalize Between Arrays in the Limma package was used to normalize'' We have modified it to “The expression status of GSE73953 and GSE25590 were obtained by R-package agimicrorna processing, normalizeBetweenArrays in the Limma package was used” (revised line 178). Thank you for pointing this out! Line 190 Significantly differential -> Significance of differential This sentence was rephrased accordingly (revised line 180), thank you! Line 201 which expressed -> which are expressed Agreed, we have added the “are” (revised line 191), thank you! Line 207 found -> find We have modified “found” to “find” (revised line 197), thank you! Line 211 screened -> screen We have modified “screened” to “screen” (revised line 201), thank you! Line 226 have -> that Agreed, we have changed to “that” (revised line 216). Thank you! Line 227 validation the expression -> validation of the expression We confused verbs with nouns here, thank you for pointing this out! We have modified this expression (revised line 217). Line 229 with the following: -> with the following conditions Thank you for the suggestion! We have added the “conditions” (revised line 220). Line 232 and listed -> and are listed Thank you very much! We have added the missing “are” (revised line 224). Line 239 differential -> differentially We have modified this expression (revised line 233), thank you! Line 241 is listed -> are listed We have modified “is listed” to “are listed” (revised line 235), thank you! Line 250 remove: ''briefly'' Agreed, we have removed the word “briefly” (revised line 244). Line 256 Fc gamma R- -> Fc gamma Receptor- Thank you! We have added the “Receptor” (revised line 250). Line 258 upregulate -> upregulated We have changed to “upregulated” (revised line 252), thank you! Line 262 function -> functions We have modified “function” to “functions” (revised line 257), thank you! Line 265 ''; '' -> ''. '' Agreed. We have changed to “.” here (revised line 260). Thank you! Line 265 upregulate -> upregulated We have modified “upregulate” to “upregulated” (revised line 260), thank you! Line 274 remove: ''have'' Thank you! The perfect past tense shouldn’t be used here, we have removed “have” (revised line 268). Line 275 overlapping -> overlapped Agreed, we have changed to “overlapped” (revised line 269). Line 278 identified -> identify We have modified “identified” to “identify” (revised line 272), thank you! Line 285 and 286 remove: '', '' We have removed these “,” (revised line 279, 280), thank you! Line 287 most -> mostly We have modified “most” to “mostly” (revised line 281), thank you! Line 296 remove: ''And'' Thank you, we have removed the “And” (revised line 290). Line 299 was -> were We have modified “was” to “were” (revised line 293), thank you! Line 368 transformer -> differentiation? Agreed, we have changed the “transformer” to “differentiation” (revised line 369 ). Thank you for pointing this out! Line 369 prepossessing -> preprocessing Sorry for this mistake. We have corrected the typo (revised line 370). Line 381 remove: ''using'' Agreed, we have removed the “using” (revised line 382). Line 382 has -> have We have modified the “has” to “have” (revised line 383), thank you! Line 383 please check: ''adaptions'' Thank you for point it out! We have changed the “adaptions” to “adaptation”( revised line 384). Line 401 These Vivo and Vitro phenotypes -> These in vivo and in vitro phenotypes Agreed, we have modified this expression (revised line 402), thank you! Line 432 remove: ''of'' We have removed the “of” (revised line 435), thank you! The authors explain the context of their study and support their data with well referenced and relevant literature. The figures are relevant and high quality. However, there are some errors in some of the titles: Table 1- Title: Most -> most Thank you! We have changed the “Most” to “most” (please see the revised line 584, Table 1). Table 1- Column 2: please check ''CricBase ID'', is it CircBase? Yes, we quote the ID from CircBase (if they have ID) for readers, we have changed the “CircBase ID” to “CircBase”, thank you! (please see the revised line 584, Table 1) Table 1- Columns are too close to each other. A suggestion is to reduce the unnecessary decimals in the last two columns (log2(FC) and p-value). Thank you! This is a very good suggestion and we have modified according to the comment. (please see the revised line 584, Table 1) Table 2- Title: deferentially -> differentially Thank you! We have corrected the typo. (please see the revised line 586, Table 2) Table 2- Column 4- Title: Different -> Differential Thank you for pointing this out, we apologize for our error. (please see the revised line 586, Table 2) Additionally, there is a lack of cohesion throughout the manuscript when referencing Figures. Sometimes the full word appears. Other times they are named as ''Fig.''. I suggest homogenizing. Thank you! According to your suggestion, I rechecked all the Figures, and homogenized their names as 'Fig. x'. The authors supplied raw data for circRNA (available at GEO accession number GSE154046) and RT-PCR (Excel file). Experimental design The original primary research of this manuscript is within the scope of the journal. IgA nephropathy remains a significant cause of renal failure that results from the formation of IgA1-containing immune complexes. Our current understanding of its pathogenesis is mostly based on the abnormal increase of circulating O-galactosylated IgA1 and the production of O-glycan-specific antibodies. However, it has become evident that other genetic, epigenetic, and immunological mechanisms are responsible for mesangial IgA1 deposition and renal injury. Hence, the authors´ research question is relevant and meaningful in furthering our understanding of IgA nephropathy's pathogenesis. The authors state that their research fills an identified knowledge gap (understanding the molecular pathogenesis of IgA nephropathy) that could lead to improved targeted therapies to ameliorate disease progression. Rigorous investigation is performed to a high technical and ethical standard. Overall, the methods are described with sufficient detail and information to replicate. Some minor comments: Methods section- A description of RT-PCR data normalization and analysis is missing. Looking at the Excel file containing the raw data it is evident that they used actin as the house-keeping gene for normalization. However, this is not stated in the manuscript and the primers used for actin amplification are also missing in Table S1. Also, the full names of ASB16, HLA-B, TRIM21, and SEC24C are not provided. Thank you very much for pointing out those deficiencies! Please see the revised line 224, we have added “β-actin was used as the house-keeping gene for normalization, qRT-PCR relative fold change results were calculated using the 2-△△Ct method. ”into the Methods section. The primer used for β-actin: ' F:ACCCTGAAGTACCCCATCGAG. R:AGCACAGCCTGGATAGCAAC' have added to the Table S1 (please see the revised 588, Table S1). The 'ankyrin repeat and SOCS box containing 16 (ASB16) , major histocompatibility complex, class I, B (HLA-B), tripartite motif containing 21(TRIM21), SEC24 homolog C, COPII coat complex component (SEC24C), ' were added into the manuscript according to the comment (revised line 221). Validity of the findings All underlying data have been provided, they are robust and statistically sound. I found the results to be supported by the data presented. Conclusions are well stated. However, I was left wondering how their findings compare to other studies of non-coding RNAs in PMBCs from patients diagnosed with glomerular diseases other than IgA nephropathy. We are grateful for the suggestion! This section was revised according to your comments, please see the revised line 327 “Similarly, systemic lupus erythematosus (SLE) in one of the autoimmune disease and could also be manifested as glomerulonephritis, several circRNAs including hsa_circ_0044235 have been found to be significantly decreased in PBMCs, and are proposed as potential negative biomarkers for SLE diagnosis (Luo Q et al., 2019).”. Could any of their circRNAs, miRNAs, and mRNAs have the potential to be used as diagnostic or prognostic markers? This is a great question! This presenting study is mainly focused on a comprehensive ceRNA network and its functions enrichment analysis. As your suggesting, many non-coding RNA are used as biomarkers of diseases, especially circRNAs, because the expression is usually stable. Therefore, we plan to take whether circRNA can be used as a biomarker of IgAN (even its relationship with disease progression and clinical manifestation) as an important content of our next research project, and stated in the discussion: “validation of larger sample size will be needed, and exploring the potential capability of circRNAs as biomarkers of IgAN would be a research in the future.” (revised line 431), thank you very much for the suggestion! Additionally, I found the number of patients in which they performed the validation of their findings through RT-PCR rather small (IgA nephropathy n=4 and healthy controls n=4). I suggest elaborating on this issue as a limitation of their study. Thank you for pointing this out this deficiency! Accordingly, we stated of this limitation in the discussion (revised line 430). ' The small sample size of the hub gene verification is also the one of limitations of this study, validation of a larger sample size will be needed, and exploring the potential capability of circRNAs as biomarkers of IgAN would be a research in the future. Future research should further identify the hub molecules with therapeutic potential in IgAN and conducted the overexpression or inhibition experiments of related ceRNAs, and observing the reciprocal relationship of target genes.' Comments for the author The GEO accession number of the circRNA data is not written in the manuscript. Is this intentional? Thank you for let us know your question! I noticed the PeerJ usually putting GSE numbers in the “ADDITIONAL INFORMATION AND DECLARATIONS” part, like “Data Availability: The following information was supplied regarding data availability: The raw data is available at GEO: GSE135111.” We would add any information needed by PeerJ, thank you! Reviewer 2 Basic reporting 1. The title of this manuscript is “Comprehensive analysis of circRNA expression profiles and circRNA associated competing endogenous RNA networks in IgA nephropathy”. It should be better using “circRNA-associated” instead of “circRNA associated”. Thank you for the suggestion! Modified throughout the text according to the comment. 2. The English language should be improved to ensure that an international audience can clearly understand your text. Some examples where the language and grammar could be corrected or improved include lines 48, 75, 85, 90, 107, 108, 122, 358 and so on – the current phrasing makes comprehension difficult. Thank you for underlining this deficiency! We have modified this expression throughout the text according to the comment. Lines 48, we have changed the original text “This study aimed to elucidate the potential roles of circRNA, microRNA (miRNA), and messenger RNA (mRNA) ceRNA network in peripheral blood mononuclear cells (PBMCs) of IgAN.”To “This study aimed to elucidate the potential roles of circRNA-associated ceRNA networks of peripheral blood mononuclear cells (PBMCs) in IgAN patients.”(revised line 46) Lines 75, we have changed the original text“The clinical manifestation often presenting as macroscopic hematuria following an upper respiratory or gastrointestinal infections.” To “The clinical manifestation often presents as macroscopic hematuria following an upper respiratory or gastrointestinal infections.” (revised line 76) Lines 85, we have changed the original text“while it's reported that several IgAN cases experienced a recovery after bone marrow transplantation (Park et al., 2008), suggest IgAN appears to be a systemic immune disease in which the kidneys are damaged as innocent bystanders (Wyatt & Julian. 2013).” to “while it's reported that several IgAN cases experienced a recovery after bone marrow transplantation (Park et al., 2008), suggesting IgAN appears to be a systemic immune disease in which the kidneys are damaged as innocent bystanders (Wyatt & Julian. 2013).” (revised line 86) Lines 90, this section was revised and modified according to the comment No.3 “line 88 – line 99 elucidating that miRNA has universal role in the pathophysiology of IgAN could be redundant.”, we have more focused on the background of circRNA-mediated ceRNA networks in IgA nephropathy in the Introduction section. Line 107, we have changed the “regulating”to “regulate”. (revised line 100) Line 108, sorry for the spelling mistake, we have changed to ‘described involved’. (revised line 102) Line 122, we have deleted “could partially represent the change of the immune system, it had been broadly investigated in IgAN” , which may be unclear and unnecessary, and changed the sentence “Because PBMCs are full of lymphocytes (with 70-90% T cells, B cells, and NK cells) could partially represent the change of the immune system, it had been broadly investigated in IgAN (Sallustio et al., 2019; Katerina et al., 2020). In this study, PBMCs from IgAN and healthy controls were collected…” to “PBMCs are blood cells with round nuclei that encompass a heterogeneous cell population (with 70-90% T cells, B cells, and NK cells)( Sallustio et al., 2019), which originate from hematopoietic stem cells that reside in the bone marrow. In this study, PBMCs from IgAN patients and healthy controls were collected….” (revised line 112). Thank you for point it out! Line 358, we have change the original text “Normally most proteins interact with only a few other proteins, while a small number of proteins (the genes encoding them, namely hub genes) have many interaction partners (He & Zhang. 2006), and usually been demonstrated play important roles in the protein network, an exploring of hub genes may bring further insights of the presented ceRNA network.”, to “Normally most proteins interact with only a few other proteins, while a small number of proteins that have many interaction partners (the genes encoding them, namely hub genes) (He & Zhang. 2006), and usually play important roles in the protein network, exploring of the hub genes may bring further insights into the presented ceRNA network.” (revised line 356) We are grateful for the suggestion! 3. The introduction needs to be more concise, and highlights the background of circRNA-mediated ceRNA networks in IgA nephropathy. Therefore, line 88 – line 99 elucidating that miRNA has universal role in the pathophysiology of IgAN could be redundant. Thank you for pointing this out! This section was revised and modified according to the comment, we have more focused on the background of circRNA-mediated ceRNA networks in IgA nephropathy in the introduction, shorten and left the part of miRNA to the discussion part. 4. Line 124, the last sentence of the introduction described that “PBMCs from IgAN and healthy controls were collected and sequenced with next-generation technology”. It’s necessary to indicate clearly PBMCs were collected from IgA nephropathy patients instead of animal model. Thank you for the suggestion! We have added the information required as explained above. Please see Line.. “PBMCs from IgAN patients and healthy controls were collected and sequenced with next-generation technology ” (revised line 114) 5. Figure 2A: The spelling of “cricRNA” was wrong. Please correct it. Sorry for the spelling mistake, we 've corrected the typo. Thank you for point out! Please see Figure 2A (revised). Experimental design no comment. Validity of the findings 6. The author didn’t use qRT-PCR to confirm the differential expressed circRNAs identified in the circRNA-sequencing data. It will improve the accuracy of this manuscript if the qRT-PCR confirmation experiment could be supplemented. Thank you for your suggestion! In this study, we want to firstly focus on the hub genes in the ceRNA network, and then next look for corresponding circRNAs to gain potential regulatory opportunities. But, the expression relationship between circRNAs and the hub genes with a certain order (up-down-up or down-up-down) has not been found in the ceRNA network, therefore the expression of any specific circRNA may not the topic in this study, and we put it in the discussion (please see line 427 ), as one of the limitations of the study. “For example, the molecular in this ceRNA network was relatively less than other ceRNA comprehensively studies, and we couldn't find ceRNA interaction pair with certain order (up-down-up or down-up-down) to validates the potential associated relationship corresponding with GALNT, ASB16 or SEC24C. The small sample size of the hub gene verification is also the one of limitations of this study, validation of a larger sample size will be needed, and exploring the potential capability of circRNAs as biomarkers of IgAN would be a research in the future.” We plan to put the validation of the circRNAs in our next project, for exploring the capacity of circRNAs as biomarker of IgAN, thank you very much! 7. In Conclusions (line 439), the manuscript elucidated that “4 circRNAs competed with miR765 to regulate the expression of GALNT2 in IgAN”. However, the experiment design of this article didn’t validate this point. Therefore, the author couldn’t draw this definite conclusion. Suggest the author could modify the content of “Conclusions”. Thank you for the suggestion! We have revised the text to address your concerns, the original text “Differentially expressed genes and their potential functions in the ceRNAs network have revealed, including 4 circRNAs competes with miR765 to regulate the expression of GALNT2 in IgAN. Besides, ASB16, SEC24C were the hub genes in the network.” were deleted, and changed the conclusion to “In summary, we successfully identified IgAN associated circRNAs using RNA-seq analysis, and elucidated the circRNA-associated ceRNA networks of IgAN through the integrated analysis of RNA expression profile. To our knowledge, this is the first report examining the expression of circRNAs in IgAN. These findings had expanded our understanding of circRNAs-associated ceRNA networks in IgAN, a future exploring the related molecular regulatory mechanism will be needed for a better understanding IgAN.”, and hope that it is now clearer. Comments for the author 8. In the Discussion, line 338 – line 343 described that “the hsa_circ_007056、hsa_circ_006671、hsa_circ_0073237 and circRNA3302 were significantly upregulated in our study, and all of them may play roles as ceRNAs of miR765, therefore participating in the manipulation the expression of target GALNT2. However, all the 3 circRNAs were up-regulated, therefore didn’t follow the classic circRNA(down in IgAN)-miRNA(up in IgAN)-mRNA(down in IgAN) interaction regulation”. This finding was interesting, but the explanation of this part was not sufficient. Hope the author could discuss deeply about this finding. Thank you for underlining this deficiency! This section was revised and modified according to the information showed in the work suggested by the reviewer: “Consistent with the literature, the GALNT2 were significantly down-regulated in the PBMCs of IgAN in our study, while the hsa_circ_0070562、hsa_circ_0066719、hsa_circ_0073237 and circRNA3302 were identified as ceRNAs of miR765, therefore could participate in the manipulation the expression of target GALNT2. However, unexpectedly all the 3 circRNAs were up-regulated in the study, therefore didn’t follow the classic circRNA (down in IgAN)-miRNA (up in IgAN)-mRNA (down in IgAN) interaction mechanism, therefore, it remains unknown whether the regulation of GLANT2 in IgAN is influenced by the circRNAs in PBMCs. It's should be mentioned that Thomson & Dinge used to discuss in 2016 (Thomson & Dinger. 2016 ) , the underlying argument opinion which against the ceRNA hypothesis is that, the change in expression of an individual non-coding RNA may only impact relatively a small fraction of the target mRNA abundance. Therefore, the real mechanism of ceRNA interaction needs further molecular biology experiments, to carefully verify this non-coding factors on the molecules in the disease, such as the previous literature, using knock-out or overexpression method to illustrated a TGF-β/Smad3-interacting-lncRNA inhibits renal fibrogenesis in IgAN (Wang et al., 2018).” (revised line 341) We would like to thank the referee again for taking the time to review our manuscript! Yours Sincerely, Hong Liu "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>In bio-medicine, inferring causal relation from experimental intervention or perturbation is believed to be a more reliable approach than inferring causation from cross-sectional correlation. However, we point out here that even in interventional inference there are logical traps. In homeostatic systems, causality in a steady state can be qualitatively different from that in a perturbed state. On a broader scale there is a need to differentiate driver causality from navigator causality. A driver is essential for reaching a destination but may not have any role in deciding the destination. A navigator on the other hand has a role in deciding the destination and the path but may not be able to drive the system to the destination. The failure to differentiate between types of causalities is likely to have resulted into many misinterpretations in physiology and bio-medicine. Methods:</ns0:p><ns0:p>We illustrate this by critically re-examining a specific case of the causal role of insulin in glucose homeostasis using five different approaches (1) Systematic review of tissue specific insulin receptor knock-outs, (2) Systematic review of insulin suppression and insulin enhancement experiments, (3) Differentiating steady state and post-meal state glucose levels in streptozotocin treated rats in primary experiments, (4) Mathematical and theoretical considerations and (5) Glucose insulin relationship in human epidemiological data. Results: All the approaches converge on the inference that although insulin action hastens the return to a steady state after a glucose load, there is no evidence that insulin action determines the steady state level of glucose. Insulin, unlike the popular belief in medicine, appears to be a driver but not a navigator for steady state glucose level. It is quite likely therefore that the current line of clinical action in the field of type 2 diabetes has limited success largely because it is based on a misinterpretation of glucose-insulin relationship. The insulin-glucose example suggests that we may have to carefully reexamine causal inferences from perturbation experiments and set up revised norms for experimental design for causal inference.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head></ns0:div> <ns0:div><ns0:head n='1.'>Causal relationships in Biology:</ns0:head><ns0:p>Inferring causal relationships in biology is a more complex, philosophical and methodological issue than what is generally perceived. Much of the debate has centered around making causal inferences from observed associations or correlations <ns0:ref type='bibr' target='#b18'>(Boudon, 1965;</ns0:ref><ns0:ref type='bibr' target='#b25'>Chawla et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b48'>Grace, 2002;</ns0:ref><ns0:ref type='bibr' target='#b54'>A. B. Hill, 1965)</ns0:ref>.Methods of analysis specifically addressing the question of causal inferences have been developed for various contexts and at various levels of reasoning including Hill criteria (A. B. <ns0:ref type='bibr' target='#b54'>Hill, 1965)</ns0:ref>, path analysis (C. C. <ns0:ref type='bibr' target='#b74'>Li, 1956;</ns0:ref><ns0:ref type='bibr' target='#b88'>Niles, 1923;</ns0:ref><ns0:ref type='bibr' target='#b136'>Wright, 1960)</ns0:ref>, Granger causality <ns0:ref type='bibr' target='#b49'>(Granger, 1969)</ns0:ref>, steady state causality <ns0:ref type='bibr' target='#b25'>(Chawla et al., 2018)</ns0:ref>, genomic and network causality <ns0:ref type='bibr' target='#b69'>(Kulkarni et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b81'>Meinshausen et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b125'>Triantafillou et al., 2017)</ns0:ref>.</ns0:p><ns0:p>While on the one hand, complex statistical and computational tools are being developed for determining causality under different contexts, certain simple problems about causality are not sufficiently appreciated and addressed. We focus on one such apparently simple problem here, the lack of appreciation of which has led to serious misinterpretations of experimental results.</ns0:p><ns0:p>In classical experimental physiology, interventions or perturbations are believed to be reliable indicators of causation and there is little debate about it. If the experimenter perturbs A and finds a significant effect on B after following all fundamental principles of experimental design, the change in A is inferred to be causal to the change in B. However, there are many subtleties in drawing a causal inference from experimental interventions that have not yet attracted sufficient philosophical as well as methodological attention among experimental biologists. One such thinking trap is that in homeostatic systems the nature of causality in a perturbed state can be qualitatively different from that in equilibrium or steady state and the failure to distinguish between the two may have substantially misled biomedical research.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.'>The role of growth rates in Lotka-Volterra competition models</ns0:head><ns0:p>A well worked out theoretical model that can be used to distinguish clearly between perturbed and steady state causation is the Lotka-Volterra (LV) competition model. This model describes the dynamics for interspecific competition <ns0:ref type='bibr' target='#b47'>(Gotelli, 2008)</ns0:ref>. The growth of two interacting populations is modelled using logistic equations for both the populations. The changes in populations depend on the individual growth rates and the carrying capacities of the two populations (equations 1 and 2). Additional parameters are the competition coefficients &#945; and &#946; which represent the effects of the two species on each other. Thus the carrying capacities K 1 , K 2 and the competition coefficients &#945; and &#946; determine the equilibrium population (equations 3 and 4) <ns0:ref type='bibr' target='#b47'>(Gotelli, 2008)</ns0:ref>. Equation <ns0:ref type='formula'>1</ns0:ref>&#119889;&#119873; 1 &#119889;&#119905; = &#119903; 1 . &#119873; 1 (</ns0:p><ns0:formula xml:id='formula_0'>&#119870; 1 -&#119873; 1 -&#120572;.&#119873; 2 &#119870; 1 ) Equation 2 &#119889;&#119873; 2 &#119889;&#119905; = &#119903; 2 . &#119873; 2 ( &#119870; 2 -&#119873; 2 -&#120573;.&#119873; 1 &#119870; 2 ) Equation 3 &#119873; 1 = &#119870; 1 -&#120572;.&#119870; 2 1 -&#120572;.&#120573; Equation 4 &#119873; 2 = &#119870; 2 -&#120573;.&#119870; 1 1 -&#120572;.&#120573;</ns0:formula><ns0:p>where, N 1 and N 2 are the population sizes of the two competing species respectively, and &#119873; 1 &#119873; 2 representing steady state populations.</ns0:p><ns0:p>r 1 and r 2 are the growth rates K 1 and K 2 are the carrying capacities &#945; is the competition coefficient which shows the effect of population 2 on population 1</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed &#946; is the competition coefficient which shows the effect of population 1 on population 2</ns0:p><ns0:p>Figure <ns0:ref type='figure'>1</ns0:ref>:Simulated population dynamics of two competing species A (black lines) and B (grey lines) at different growth rates. Solid lines represent high growth rates reaching the equilibrium faster whereas dotted lines represent slow growth rates reaching the equilibrium more slowly. In both cases species A has slower growth rate but greater carrying capacity than species B. At time T1 the populations are in proportion to their growth rates but at T2, approaching equilibrium, growth rates become increasingly irrelevant in determining population sizes. Thus, growth rates are important determinants of population size in a perturbed state but not at a steady state.</ns0:p><ns0:p>If, in an experiment, we start at a non-equilibrium state, and observe at time T1 (figure <ns0:ref type='figure'>1</ns0:ref>), the standing populations would be inferred as a function of intrinsic growth rates and . But if &#119903; 1 &#119903; 2 observed at time T2, the inference would be different. The intrinsic growth rates of both the competing populations do not determine the equilibrium populations of the two species. The magnitude of r determines the time taken by the population to reach the equilibrium or steady state (figure <ns0:ref type='figure'>1</ns0:ref>). The role of growth rates in population dynamics is well recognized and is demonstrable in a perturbed state but it needs to be realized that it has no role in determining the steady state populations. Nevertheless, existence of non-zero positive growth rates is essential for attaining the equilibrium or returning to it if perturbed. If either or both the growth rates are made zero, the system will never attain back a stable equilibrium coexistence. Thus, the two growth rates are causal for attaining equilibrium, but they have no causal role in deciding the position of the equilibrium point. Thus, we need to distinguish between the driver cause and the navigator cause. Driver causality is a process that takes a homoeostatic system to an equilibrium point but may not have any role in deciding the attributes of the equilibrium. Navigator causality refers to the processes that determine the location of the equilibrium point and lead the driver there, but in the absence of the driver, may not be able to take the system to the steady state.</ns0:p><ns0:p>For a homeostatic system, the distinction between perturbed state and steady state causality is practically equivalent to driver and navigator causality. However, the driver-navigator distinction can be applied, in principle, to non-homeostatic systems as well and therefore is a broader concept.</ns0:p><ns0:p>This has relevance to experimental physiology. If knocking out a certain gene, protein or function disables homeostatic control, it does not provide us any clue as to whether it has a driver or navigator function. The experiment does not necessarily demonstrate that the gene, protein or function determines the steady state levels of the controlled variable. Since distinction between driver and navigator causality has not been explicitly made in experimental physiology, currently there are no norms or methods to resolve between the two types of causations. We use this distinction below to re-examine the role of insulin in glucose homeostasis and show that the failure to distinguish between driver and navigator causality has led to a fundamentally flawed understanding of glucose homeostasis and type 2 diabetes in particular.</ns0:p></ns0:div> <ns0:div><ns0:head n='3.'>Why is insulin believed to regulate fasting blood sugar: A burden of history?</ns0:head><ns0:p>After the classical demonstration by Claud Bernard that damage to medulla oblongata causes hyperglycaemia <ns0:ref type='bibr' target='#b12'>(Bernard, 1879)</ns0:ref>, the second major breakthrough was the demonstration by von</ns0:p><ns0:p>Mering and Minkowski that pancreatectomy resulted in hyperglycaemia <ns0:ref type='bibr' target='#b84'>(Mering &amp; Minkowski, 1890)</ns0:ref> and further that pancreatic extracts resulted in lowering of plasma glucose. The active principle eventually purified became known as insulin <ns0:ref type='bibr' target='#b66'>(Karamitsos, 2011)</ns0:ref>. The discovery and success of insulin in treating diabetes was so overwhelming that insulin became the key molecule in glucose homeostasis and the role of brain and other mechanisms were practically forgotten. It should be noted that the prevalent type of diabetes then was what we would call type 1 diabetes (T1D) today in which there is almost complete destruction of pancreatic &#946;-cells. The distinction between type 1 and 2 developed gradually over the next five decades along with the realization that insulin levels may be normal or raised in type 2 diabetes (T2D) and that a substantial population of &#946;-cells survives lifelong <ns0:ref type='bibr' target='#b22'>(Butler et al., 2003;</ns0:ref><ns0:ref type='bibr' target='#b26'>Clark et al., 1990;</ns0:ref><ns0:ref type='bibr' target='#b100'>Porte &amp; Kahn, 2001)</ns0:ref>. However, by now the thinking about glucose homeostasis was so insulin-centered, that the inability of normal or raised levels of insulin to keep plasma glucose normal was labelled as 'insulin resistance' <ns0:ref type='bibr' target='#b102'>(Reaven, 1988)</ns0:ref> without adequately examining and eliminating alternative possibilities and the concept got wide uncritical acceptance. Although insulin receptor and downstream functions are known to be highly variable at the cellular level, the question whether altered insulin signalling is solely or mainly responsible for fasting hyperglycaemia of T2D, or other insulin independent mechanisms play a significant role is not clearly answered.</ns0:p><ns0:p>There are multiple reasons to doubt and re-examine the role of insulin in glucose regulation in relation to T2D <ns0:ref type='bibr' target='#b30'>(Corkey, 2012;</ns0:ref><ns0:ref type='bibr' target='#b98'>Pories &amp; Dohm, 2012;</ns0:ref><ns0:ref type='bibr' target='#b131'>M. Watve, 2013)</ns0:ref>. Exogenous insulin and other insulin-centered lines of treatment have largely failed to reduce diabetic complications and mortality in T2D although short term glucose lowering may be achieved <ns0:ref type='bibr' target='#b0'>(ACCORD, 2008;</ns0:ref><ns0:ref type='bibr' target='#b67'>King et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b80'>Meinert et al., 1970;</ns0:ref><ns0:ref type='bibr'>UK Prospective Diabetes Study Group, 1998a</ns0:ref><ns0:ref type='bibr' target='#b130'>, 1998c</ns0:ref><ns0:ref type='bibr' target='#b128'>, 1998b)</ns0:ref>.</ns0:p><ns0:p>In the long run even the glucose normalization goal is not achieved in majority of cases (UK Prospective Diabetes Study <ns0:ref type='bibr'>Group, 1998b</ns0:ref><ns0:ref type='bibr'>Group, , 1998c))</ns0:ref>. A number of mechanisms are known to influence glucose dynamics, partially or completely independent of insulin signalling, including autonomic signals <ns0:ref type='bibr' target='#b89'>(Nonogaki, 2000;</ns0:ref><ns0:ref type='bibr' target='#b108'>Schwartz, 2005)</ns0:ref>, glucocorticoids (Di <ns0:ref type='bibr' target='#b33'>Dalmazi et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b43'>Gathercole &amp; Stewart, 2010;</ns0:ref><ns0:ref type='bibr' target='#b44'>Goldstein et al., 1993;</ns0:ref><ns0:ref type='bibr' target='#b70'>Kuo et al., 2015)</ns0:ref>, insulin independent glucose transporters <ns0:ref type='bibr' target='#b23'>(Carruthers et al., 2009)</ns0:ref> and certain other hormones and growth factors <ns0:ref type='bibr' target='#b27'>(Clemmons, 2004;</ns0:ref><ns0:ref type='bibr' target='#b62'>Jansen et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b85'>Messmer-Blust et al., 2012;</ns0:ref><ns0:ref type='bibr'>Suh et al., 2014)</ns0:ref>. Analysis of multi-organ signalling network models have also raised doubts about the central role of insulin and insulin resistance in T2D <ns0:ref type='bibr' target='#b69'>(Kulkarni et al., 2017)</ns0:ref>.</ns0:p><ns0:p>The definitions as well as clinical measures of insulin resistance are such that the effects of all other mechanisms are accounted for under the name of 'insulin resistance'. For example, the HOMA-IR index is calculated as a product of fasting glucose and fasting insulin <ns0:ref type='bibr' target='#b78'>(Matthews et al., 1985;</ns0:ref><ns0:ref type='bibr' target='#b126'>Turner et al., 1979)</ns0:ref>. The belief that this product reflects insulin resistance is necessarily based on the assumption that insulin signalling alone quantitatively determines glucose level in a fasting steady state. The assumption has seldom been critically examined. If any other mechanisms are contributing to impaired fasting glucose, they will be included in the HOMA-IR index going by the way it is calculated and would be labelled as insulin resistance.</ns0:p><ns0:p>This amounts to a circular logic. Insulin resistance is hypothesized to be responsible for the failure of insulin to control fasting glucose, and insulin resistance is measured as the inability of insulin to control fasting glucose. This makes the insulin resistance concept unfalsifiable from clinical data.</ns0:p><ns0:p>We have previously showed using mathematical and statistical tools of causal analysis <ns0:ref type='bibr' target='#b25'>(Chawla et al., 2018)</ns0:ref> that the classical pathway of obesity induced insulin resistance leading to a hyperinsulinemic normoglycemic prediabetic state and the faithfulness of HOMA indices in measuring insulin resistance cannot be simultaneously true. Either the HOMA indices do not represent insulin resistance faithfully or the classically believed pathway of compensatory insulin response leading to hyperinsulinemic normoglycemic state is wrong according to this analysis <ns0:ref type='bibr' target='#b25'>(Chawla et al., 2018)</ns0:ref>. All the mounting inconsistencies and paradoxes in the insulin resistance</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed concept along with the limited success of type 2 diabetes treatment warrant a re-examination of the foundational concepts.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.'>Approach used in this paper</ns0:head><ns0:p>We examine here the long-held belief that altered insulin signalling is responsible for fasting as well as post prandial hyperglycemia in T2D using five different approaches</ns0:p><ns0:p>(1) Systematic review of experiments involving tissue specific insulin receptor knock-outs (IRKOs)</ns0:p><ns0:p>(2) Systematic review of experiments to chronically raise or lower insulin levels</ns0:p><ns0:p>(3) Primary experiments on streptozotocin (STZ) induced hyperglycaemia in rats, that differentiate between steady (fasting) and perturbed (post-feeding) state</ns0:p><ns0:p>(4) Examining the insulin resistance hypothesis for being mathematically possible and theoretically sound</ns0:p><ns0:p>(5) Analysis of insulin-glucose relationship in steady state versus post-meal perturbed state in human epidemiological data for testing the predictions of mathematical models.</ns0:p><ns0:p>The first three approaches have the advantage of using specific molecular interventions where the target is precisely known. For the analyses we chose mechanisms of insulin level/action modification which have been used extensively and have been reproduced by multiple labs world over. The possible disadvantage is that they are mostly animal experiments and doubts are expressed about whether the results are directly relevant to humans <ns0:ref type='bibr' target='#b3'>(Akhtar, 2015;</ns0:ref><ns0:ref type='bibr' target='#b5'>Ali et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b19'>Bracken, 2009)</ns0:ref>. However, some of the experiments reported are human and they converge with the inferences of the animal experiments. In the last two approaches, human epidemiological data are used in which the experimental molecular precision is not expected, but we test certain specific predictions of the insulin resistance hypotheses using novel analytical approaches and examine whether they converge on similar inferences. The convergence of human and animal data is important to reach robust conclusions.We will now describe each approach in detail here.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.1'>Systematic review of experiments involving tissue specific insulin receptor knock-outs</ns0:head><ns0:p>The first step in insulin signaling is the binding of insulin to insulin receptor <ns0:ref type='bibr'>(Bevan, 2001)</ns0:ref>. The downstream actions of this event finally lead to insulin-dependent glucose uptake in insulin dependent tissues of the body. Experimentally, disruption of insulin signaling is achieved by knocking out or inhibiting various players in the signaling cascade. We chose to look at the effects of knocking out the tissue specific insulin receptor on fasting and post-meal/feeding or post glucose load levels in rodent models. Studying tissue specific insulin receptor knockouts enables us to differentiate between the roles of insulin signaling in different tissues. A classical belief is that the post-meal glucose curve is mainly influenced by the rate of glucose uptake by tissues, mainly muscle, whereas the fasting glucose levels are mainly determined by the rate of liver glucose production <ns0:ref type='bibr' target='#b16'>(Bock et al., 2007)</ns0:ref>. If this belief is true one expects that muscle specific knockout would mainly affect the GTT curve but may not affect fasting glucose level, whereas liver specific knockout would mainly affect the fasting glucose level. We searched the published literature for experiments in which the fasting and post-feeding levels of glucose are measured in tissue specific insulin receptor knock-outs.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.2'>Systematic review of experiments to chronically raise or lower insulin levels</ns0:head><ns0:p>The insulin receptor knockout experiments assume that the main action of insulin is through the specific receptors. for experiments where a stable and sustained increase or decrease in insulin was achieved and then the effect on fasting glucose and GTT studied.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.3'>Fasting versus post feeding glucose in STZ rats</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Streptozotocin (STZ) induced diabetes is a popular model of rodent diabetes <ns0:ref type='bibr' target='#b1'>(Akbarzadeh et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b40'>Gajdos&#237;k et al., 1999)</ns0:ref>. STZ acts by specifically destroying the insulin producing &#946; cells of the pancreatic islets <ns0:ref type='bibr' target='#b116'>(Szkudelski, 2001)</ns0:ref>. A low dose of STZ that destroys a substantial population of &#946; cells but does not lead to total destruction of their population is often perceived as a good model for T2D, whereas a high dose of STZ that destroys the &#946; cell population almost entirely is perceived as a model of T1D <ns0:ref type='bibr' target='#b1'>(Akbarzadeh et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b7'>Asrafuzzaman et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b39'>Freitas et al., 2015;</ns0:ref><ns0:ref type='bibr' target='#b40'>Gajdos&#237;k et al., 1999;</ns0:ref><ns0:ref type='bibr' target='#b65'>Kagami et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b68'>Kovacs et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b116'>Szkudelski, 2001;</ns0:ref><ns0:ref type='bibr' target='#b140'>Zhang et al., 2008)</ns0:ref>. We searched literature to look for studies that carefully differentiated between steady state glucose from post load glucose in STZ models but did not find any studies that make this distinction clear. Therefore, we designed and conducted experiments to differentially study the steady state and perturbed state glucose levels in rats treated with STZ.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.4'>Theoretical and mathematical considerations</ns0:head><ns0:p>In this approach we elaborate on the theoretical underpinnings of insulin-glucose relationship.</ns0:p><ns0:p>We also explore possible explanations for the unexpectedly consistent failure of experimental insulin signal impairment to alter steady state glucose level. Simultaneously we make differential predictions from alternative homeostasis models that can be tested in human epidemiological data. The two models which we use here answer the critical question in glucose homeostasis:</ns0:p><ns0:p>whether the fasting/steady state glucose level is a consequential balance between glucose production and glucose utilization rates (consequential steady state CSS) or whether there is a target glucose level that is maintained by sensing and correcting any changes in it (targeted steady state TSS).</ns0:p></ns0:div> <ns0:div><ns0:head n='4.5'>Analysis of insulin glucose relationship in steady and perturbed state in human data:</ns0:head></ns0:div> <ns0:div><ns0:head>Epidemiological inquiry</ns0:head><ns0:p>Here we use human epidemiological data to test the correlational predictions made by the CSS versus TSS models of glucose homeostasis.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head><ns0:p>We will outline the methods used in each of the 5 approaches described in the last section of the introduction.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Systematic Review Registration</ns0:head><ns0:p>1. The four meta-analyses were registered in PROSPERO.</ns0:p><ns0:p>2. We did not start the project with the intention of doing these systematic reviews. Our primary goal was to test the alternative evolutionary hypotheses for the origins of type 2 diabetes (M. <ns0:ref type='bibr' target='#b132'>Watve &amp; Diwekar-Joshi, 2016)</ns0:ref>. While pursuing this work we noticed certain basic anomalies in the classical theories of diabetes. At the stage when we thought of the question addresses in this paper, we had already scanned much of the literature used here.</ns0:p><ns0:p>3. To ensure that our review is unbiased and rigorous enough, we repeated the literature search following PRISMA and PROSPERO guidelines. Therefore, much of the search for the Meta-analyses was done before the actual PROSPERO registration. The secondary screening, data extraction and the data analysis was done during the time required after the submission to the PROSPERO and the actual approval of the registration by them.</ns0:p><ns0:p>4. But we were glad to find after following proper procedures that our prior work met the standards of unbiased and rigorous review.</ns0:p><ns0:p>5. Details about the search strategies, quality assessment and data analysis for the four metaanalyses are included in Supplementary information 1.</ns0:p><ns0:p>6. The flowcharts for the search strategies and screening procedures followed are given in the form of PRISMA flowcharts and checklists and are given as Supplementary Information files.</ns0:p></ns0:div> <ns0:div><ns0:head>Systematic Review/Meta-analyses: Search and screening</ns0:head><ns0:p>1. Manawa Diwekar-Joshi and Milind Watve decided the search strategy and the actual searches and screening based on the strategy was done by Manawa Diwekar-Joshi.</ns0:p><ns0:p>2. Data analysis was done by both the authors.</ns0:p><ns0:p>3. Disagreements were dissolved by discussion and input from other lab members.</ns0:p><ns0:p>4. The referees for the manuscript were other post-doc and graduate students of the Watve Lab.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical approach (used for all the meta-analyses)</ns0:head><ns0:p>Although we short-listed papers that used similar methods, small differences in protocols can make considerable differences in the results. There is substantial variation in results across studies. Therefore, we use non-parametric methods for analyzing the pooled data. We first look at in how many of the experiments the treatment means are greater than the control means and in how many they are smaller. If this difference is significant, we conclude that there is enough qualitative consistency across experiments to reach a reliable inference. If there is a consistent direction of difference, we look at how many are individually significant. As a conservative approach we avoid pooling data quantitatively since across studies there are differences in age or weight of animals, number of days after treatment, number of hours of fasting and other variables. This approach is maintained for the analysis of all the meta-analyses.</ns0:p></ns0:div> <ns0:div><ns0:head>1.Systematic review of experiments involving tissue specific insulin receptor knock-outs</ns0:head><ns0:p>The details of the method of the systematic literature review involving the tissue specific insulin receptor knock-outs are given in table 1. The systematic review was registered in PROSPERO (ID: CRD42019132379).The details of the experiments of the shortlisted studies can be seen in table 1 of the supplementary information 1 which shows that similar methods have been utilized</ns0:p><ns0:p>to create the knockouts and therefore a comparative analysis is justified. </ns0:p></ns0:div> <ns0:div><ns0:head n='2.'>Systematic review of experiments to chronically raise or lower insulin levels</ns0:head></ns0:div> <ns0:div><ns0:head n='2.1'>Increase in insulin</ns0:head><ns0:p>A model for sustained increase in insulin levels is a knock-out or inhibition of the insulin degrading enzyme (IDE). An interplay between insulin secretion and insulin degradation maintains the level of insulin in plasma <ns0:ref type='bibr' target='#b9'>(Authier et al., 1996;</ns0:ref><ns0:ref type='bibr' target='#b36'>Duckworth et al., 1998;</ns0:ref><ns0:ref type='bibr'>Hulse, Ralat, &amp; Wei-Jen, 2009;</ns0:ref><ns0:ref type='bibr' target='#b110'>Shen et al., 2006)</ns0:ref>. Plasma insulin has a half-life of 4 to 9 minutes <ns0:ref type='bibr'>(Hulse, Ralat, &amp; Wei-Jen, 2009;</ns0:ref><ns0:ref type='bibr' target='#b123'>Tomasi et al., 1967)</ns0:ref> and it is degraded predominantly by the insulin degrading enzyme (IDE) <ns0:ref type='bibr'>(Hulse, Ralat, &amp; Wei-Jen, 2009;</ns0:ref><ns0:ref type='bibr' target='#b110'>Shen et al., 2006)</ns0:ref>. Inhibition of IDE has been considered as a therapeutic option for type 2 diabetes with limited and questionable success <ns0:ref type='bibr' target='#b31'>(Costes &amp; Butler, 2014;</ns0:ref><ns0:ref type='bibr' target='#b76'>Maianti et al., 2014)</ns0:ref>. We performed a systematic literature review to find out experiments in which IDE was inhibited to obtain a sustained high plasma insulin level and, in such animals, GTT was performed (table 2) (PROSPERO registration ID: CRD42019140619). </ns0:p></ns0:div> <ns0:div><ns0:head n='2.2'>Decrease in insulin</ns0:head><ns0:p>We performed a systematic literature review for experiments in which there was sustained suppression of insulin production. Two insulin suppressing agents have been repeatedly used to lower insulin production in rodent models as well as in humans.</ns0:p><ns0:p>(i). Diazoxide (DZX): Diazoxide is a potassium channel activator which causes reduction in insulin secretion by the &#946;-cells by keeping the cells in a hyperpolarized state by opening the channel <ns0:ref type='bibr' target='#b94'>(Panten et al., 1989)</ns0:ref>. It has been used as a drug to modulate insulin secretion for research and therapeutic purposes <ns0:ref type='bibr' target='#b35'>(Doyle, 2003)</ns0:ref>.</ns0:p><ns0:p>(ii). Octreotide (OCT): Octreotide is a somatostatin analogue which inhibits insulin and growth hormone. It has been used to reduce insulin secretion in vitro and in vivo <ns0:ref type='bibr' target='#b71'>(Lamberts et al., 1996)</ns0:ref>.</ns0:p><ns0:p>We searched the literature systematically for studies where the insulin levels have been altered</ns0:p><ns0:p>using either DZX or OCT and glucose tolerance has been examined using a GTT after DZX/OCT treatment (table <ns0:ref type='table' target='#tab_3'>3</ns0:ref>) (PROSPERO registrationID for DZX: CRD42020141688;</ns0:p><ns0:p>PROSPERO registration ID for OCT: CRD42020141464). It should be noted that this literature includes a significant proportion of human trials. We also searched literature for studies in which insulin was suppressed by other methods. </ns0:p></ns0:div> <ns0:div><ns0:head>3.Steady state versus perturbed state glucose in STZ rats</ns0:head></ns0:div> <ns0:div><ns0:head n='3.1'>Animal model and conditions</ns0:head></ns0:div> <ns0:div><ns0:head>Ethics Approval</ns0:head><ns0:p>The experiments performed on Sprague Dawley (SD) rats had been approved by the Institutional</ns0:p><ns0:p>Animal Ethics Committee at IISER, Pune (Protocol Number IISER/IAEC/2016-02/006) constituted by the Committee for the Purpose of Control and Supervision of Experiments on Animals (CPCSEA), Govt. of India.</ns0:p></ns0:div> <ns0:div><ns0:head>Housing of the animals</ns0:head><ns0:p>All the rats were housed in a facility with a temperature of 23&#177;2&#176;C and a 12-hour light/dark cycle with standard rat chow (Altromin rat/mice maintenance diet) and water available ad libitum. The bedding of the cages was changed every three days and every day after injection of STZ. There were no extra measures taken for the enrichment of the animals.</ns0:p></ns0:div> <ns0:div><ns0:head>Euthanasia</ns0:head><ns0:p>The animals were not euthanized before the end of the experiments. At the end of the 12 days of glucose and insulin readings, the animals were euthanized with an intraperitoneal injection of thiopentone (100-120mg/kg body weight).</ns0:p></ns0:div> <ns0:div><ns0:head n='3.2'>STZ treatment for insulin suppression</ns0:head><ns0:p>Male, SD rats weighing 180-200 g were injected with STZ at 50 mg/kg body weight. The STZ was dissolved in Citrate Buffer (Citric Acid: 0.1M and Sodium Citrate: 0.1M). Injection of citrate buffer alone was used as control.</ns0:p></ns0:div> <ns0:div><ns0:head n='3.3'>Fasting and post-feeding glucose in 12 day follow up</ns0:head><ns0:p>Three days after the STZ injection, the rats were fasted for 16 hours and glucose was measured using the hand held Accu-Chek Glucometer. The rats were then given 40 grams of Standard Chow for 8 hours. Food was weighed and post-meal glucose was measured after three hours. The protocol was repeated for 12 days and body weight, food weight and glucose readings were taken daily. 12 animals per group were used for this experiment.</ns0:p></ns0:div> <ns0:div><ns0:head n='3.4'>Duration of fasting</ns0:head><ns0:p>An experiment was also performed to see how much time was required to reach a steady state of glucose after removal of food. The food was removed from the STZ and Control animals after ad libitum availability and glucose readings were taken after 3 hours, 6 hours, 9 hours, 12 hours and 16 hours. After a recovery of three days, glucose levels were measured only at 16 hours after removing the food. 9 STZ treated animals and 10 Control animals (injected with citrate buffer)</ns0:p><ns0:p>were used for this experiment.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head n='4.'>Theoretical and mathematical considerations 4.1 Choice of models for glucose homeostasis</ns0:head><ns0:p>The fasting state has been generally accepted to be a steady state for glucose concentration for several reasons. In a given healthy individual the fasting glucose levels are stable in time <ns0:ref type='bibr' target='#b53'>(Halter et al., 1985;</ns0:ref><ns0:ref type='bibr' target='#b73'>Lerner &amp; Porte, 1972)</ns0:ref>. The post-meal peak of glucose and insulin returns to the fasting level within a few hours and remains stable over a long time. The fasting state is considered and modelled as a steady state by the widely used HOMA model <ns0:ref type='bibr' target='#b78'>(Matthews et al., 1985;</ns0:ref><ns0:ref type='bibr' target='#b126'>Turner et al., 1979)</ns0:ref>. Classically the negative feedback loops are assumed to work through insulin and insulin is taken as a determinant of steady state glucose level. Most popular models of glucose homeostasis work on this assumption although non-steady state models of insulin resistance exist <ns0:ref type='bibr' target='#b92'>(Palumbo et al., 2013)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.2'>Models used for this analysis</ns0:head><ns0:p>A critical question in glucose homeostasis is whether the fasting steady state glucose level is a consequential balance between glucose production and glucose utilization rates (consequential steady state CSS) or whether there is a target glucose level that is maintained by sensing and correcting any changes in it (targeted steady state TSS). The difference in the two can be visualized by the tank water level analogy (figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>). If a tank has an input tap releasing water in it at a constant rate and has an outlet at the bottom through which water escapes proportionate to the pressure of the water column, a steady state is invariably reached (figure <ns0:ref type='figure' target='#fig_12'>2A</ns0:ref>). The steady state level is decided by the rate of intake and the size of the outlet. This is a CSS which will change with any change in the size/capacity of the input or the outlet tap. In contrast to CSS, in a TSS there is a desired water level and sensors are placed above and below the desired level such that when the level goes below the lower sensor the input is switched on or its rate increased and/or output switched off or its rate decreased (figure <ns0:ref type='figure' target='#fig_12'>2B</ns0:ref>). Refer to supplementary information 2 for the details of the CSS model. Manuscript to be reviewed reach a steady state but will not change the steady state level due to the presence of sensors which regulate the inlet and outlet taps</ns0:p></ns0:div> <ns0:div><ns0:head>5.Analysis of insulin glucose relationship in steady and perturbed state in human data:</ns0:head><ns0:p>Epidemiological inquiry</ns0:p></ns0:div> <ns0:div><ns0:head n='5.1'>Epidemiological data</ns0:head><ns0:p>The three data sets used here come from two different studies: (i) Coronary Risk of Insulin Sensitivity in Indian Subjects (CRISIS) study, Pune, India <ns0:ref type='bibr'>(Yajnik et al., 2007)</ns0:ref> and (ii)</ns0:p><ns0:p>Newcastle Heart Project (NHP), UK <ns0:ref type='bibr' target='#b14'>(Bhopal et al., 1999)</ns0:ref>. Data from the latter is divided into two groups as the subjects belong to different ethnicities namely European white and south Asian and we will prefer to analyze the two groups separately since certain ethnic differences are likely to be present in the tendency to develop metabolic syndrome <ns0:ref type='bibr' target='#b13'>(Bhopal, 2013;</ns0:ref><ns0:ref type='bibr' target='#b51'>Gujral et al., 2013)</ns0:ref>.</ns0:p><ns0:p>Hence all the comparison of predictions with the data has been done independently for the three data sets. All the studies are population surveys that include non-diabetic (fasting glucose values less than 110mg/dl) and diabetic individuals (fasting glucose values above 110 mg/dl) and the clinical history, morphometric parameters, glucose and insulin during fasting and oral glucose tolerance test (OGTT) of the subjects were recorded. In the analysis, we included only the nondiabetic groups in which the homeostatic mechanism can be assumed to be intact and therefore any hypothesis about it can be tested. Most of the individuals in the diabetic group would be under different drug regime affecting glucose-insulin dynamics in different ways and therefore we exclude that group for the analysis.</ns0:p></ns0:div> <ns0:div><ns0:head n='5.2'>Statistics</ns0:head><ns0:p>Linear regression and correlation were used to compare the glucose-insulin relationship in steady state (fasting) versus perturbed state (post glucose load) in the three data sets along with the relationships between HOMA-IR and HOMA-&#946; derived from the fasting data.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head n='1.'>Systematic review of experiments involving tissue specific insulin receptor knock-outs</ns0:head></ns0:div> <ns0:div><ns0:head>Pooled results of IRKO experiments</ns0:head><ns0:p>We shortlisted 16 papers describing 46 independent experiments in which glucose tolerance curves of insulin receptor knockouts and controls were compared (table 1 of supplementary information 1). The experiments could be segregated in four different tissue specific knockouts for the analysis: fat/adipose insulin receptor knockout or FIRKO (figure <ns0:ref type='figure' target='#fig_1'>3</ns0:ref>), Muscle insulin receptor knockout or MIRKO (figure <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>), liver insulin receptor knockout or LIRKO (figure <ns0:ref type='figure' target='#fig_3'>5</ns0:ref>) and &#946;-cell insulin receptor knockout or &#946;IRKO (figure <ns0:ref type='figure' target='#fig_4'>6</ns0:ref>). A generalized trend in the total picture summed up over all four IRKOs was that along the GTT curve, significantly higher glucose levels are seen in the knockouts as compared to the controls, particularly and consistently at 30, 60 and 120 minutes. However, the fasting glucose level was not significantly different. In some studies, fasting glucose was significantly higher in the knockouts than the controls, however in some other studies it was significantly lower as well. In 29 out of 46 experiments there was no significant difference (table 4) in the fasting glucose levels of knockouts and controls. This trend was consistently seen in MIRKO (figure <ns0:ref type='figure' target='#fig_2'>4</ns0:ref>), LIRKO (figure <ns0:ref type='figure' target='#fig_3'>5</ns0:ref>) and &#946;IRKO (figure <ns0:ref type='figure' target='#fig_4'>6</ns0:ref>). Only in FIRKO (figure <ns0:ref type='figure' target='#fig_1'>3</ns0:ref>) there were greater number of studies showing fasting glucose significantly higher in the knockouts than in the controls <ns0:ref type='bibr' target='#b50'>(Guerra et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b104'>Sakaguchi et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b111'>Softic et al., 2016)</ns0:ref>, but in the non-parametric analysis the collective trend was not significant table 4 .</ns0:p><ns0:p>These inconsistencies in the FIRKO fasting glucose levels compared to the MIRKO LIRKO could be explained with the help of the duration of fasting used for the glucose tolerance tests.</ns0:p><ns0:p>Although the glucose levels in the WT are higher than the FIRKO/BATIRKO in the fasting conditions, there could be possible reasons for that. For example, in the study <ns0:ref type='bibr' target='#b104'>(Sakaguchi et al., 2017)</ns0:ref>, the fasting for the glucose tolerance test was carried out only for 6 hours as compared to 16 hours/ON in other studies. Secondly, one of the knockouts used in this study is a double knockout of insulin receptor as well as the insulin-like growth factor 1 receptor which could be a possible reason for the higher glucose levels in the fasting condition. In case of the <ns0:ref type='bibr' target='#b50'>(Guerra et al., 2001;</ns0:ref><ns0:ref type='bibr' target='#b111'>Softic et al., 2016)</ns0:ref>, the animals have been fasted ON for the glucose tolerance test and the tests have been performed on FIRKO and WT of different ages. The impairment of glucose tolerance increases with age, through this also is not seen consistently across all the studies. In the case of <ns0:ref type='bibr' target='#b15'>(Bl&#252;her et al., 2002)</ns0:ref>, the fasting duration for the GTT is 16 hours, highest in all the studies and in this case the treated glucose levels in the fasting condition are equal to or lower than that of the controls. It is important to note that in FRIKO, although there are maximum number of cases where in the treated is higher than the control in the fasting condition, this trend is not significant in the non-parametric analysis (table <ns0:ref type='table' target='#tab_4'>4</ns0:ref>). Also, only in FIRKO, the 30, 60-and 120-minute glucose was not significantly different in the knockouts than the controls (figure <ns0:ref type='figure' target='#fig_1'>3</ns0:ref>).</ns0:p><ns0:p>It is notable that in none of LIRKO experiments the fasting sugar was significantly higher than the controls. This contradicts the classical belief that liver insulin resistance is mainly responsible for fasting hyperglycemia in T2D <ns0:ref type='bibr' target='#b16'>(Bock et al., 2007;</ns0:ref><ns0:ref type='bibr' target='#b64'>Johnson et al., 1972)</ns0:ref>.</ns0:p><ns0:p>A possible problem in comparing fasting glucose across different studies was that different fasting intervals have been used ranging from 4 to 16 hours. No study clearly reported how much time is required to reach a steady state in a knockout. In 10 of the experiments in which fasting time was reported as 16 hours, none had fasting sugar significantly different from controls. In the 13 experiments in which it was high, the fasting duration was between 4 to 12 hours or not precisely reported. Therefore, it is likely that in at least some of the experiments, glucose steady state was not yet achieved at the time point defined as fasting. This bias increases the probability that higher fasting glucose is reported for the knockouts. However, since we do not see a significant difference in the collective analysis, the inference that IRKO does not alter fasting glucose is unlikely to be a result of the bias. In fact, any possible correction to the bias might further reduce the apparent residual difference. Therefore, in spite of some inconsistency across studies, a robust generalization is that IRKOs have significantly increased plasma glucose over controls at 30 to 120 minutes post glucose load but they do not appear to affect steady state fasting glucose. The time required to reach the steady state is evidently increased. shows, out of the total number of experiments used for the analysis, in how many the mean of the knockouts (T) was greater than the control means (C) and in how many the trend was reverse. This relative position of the means across studies is compared non-parametrically to see whether the trend across studies was non-random, significant ones being indicated by asterisk. The table also gives in how many studies T was significantly greater than C and vice versa. For fasting glucose, the difference is not significant in majority of studies and where there is statistical significance there is lack of consistency across studies. However, at 30, 60 and 120 minutes the knockouts have consistently elevated levels of glucose as compared to the corresponding controls. </ns0:p></ns0:div> <ns0:div><ns0:head n='2.'>Systematic review of experiments to chronically raise or lower insulin levels</ns0:head></ns0:div> <ns0:div><ns0:head n='2.1'>Increase in insulin by suppression of IDE</ns0:head><ns0:p>We found 6 publications that described 18 experiments that allowed comparison of GTT between raised insulin groups and control group (table 2 of supplementary information 1).</ns0:p><ns0:p>Analysis revealed no significant difference in the fasting glucose. In only one out of 18 experiments the treatment group had lower fasting glucose than the control. During the GTT curve, at 90 and 120 minutes the difference between treatment and control were significant but in the opposite direction of the expectation (table 5, figure <ns0:ref type='figure' target='#fig_5'>7</ns0:ref>). While rise in insulin level should reduce plasma glucose, it increased in 15 out of 18 studies, two of which were individually significant and the difference was significant in non-parametric collective analysis (table 5, figure <ns0:ref type='figure' target='#fig_5'>7</ns0:ref>). Across all time points along the GTT, the plasma glucose in the treated group was greater than the control group in majority of the experiments. Thus, in this class of experiments increasing insulin failed to reduce glucose at the steady state as well as post glucose load.</ns0:p><ns0:p>Table <ns0:ref type='table'>5</ns0:ref>: Analysis of the fasting and post-feeding glucose levels in the control and IDE-inhibition: The table shows, out of the total number of experiments used for the analysis, in how many the mean of the IDE inhibition models (T) was greater than the control means (C) and in how many the trend was reverse. This relative position of the means across studies is compared non-parametrically to see whether the trend across studies was non-random, significant ones being indicated by asterisk. The table also gives in how many studies T was significantly greater than C and vice versa. There is no significant difference in the control at treated fasting glucose level. At 90 and 120 minutes the difference between the control and treated was significant, but in the opposite direction. significantly from the control. At 90 and 120 minutes the trend is higher mean glucose than control which is contrary to the expectation in an experiment with sustainable rise in insulin.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.2'>Decrease in insulin: Suppression by diazoxide or octreotide</ns0:head><ns0:p>We found 8 papers describing 14 experiments for diazoxide treatment and 10 papers with 15 experiments for octreotide treatment (tables 3 and 4 from supplementary information 1 respectively). It can be seen from table <ns0:ref type='table'>6</ns0:ref> and figure 8 that for both insulin suppressing agents, suppression of insulin did not result into increased fasting glucose. Further at 120 minutes post glucose load there was a marginally significant rise in glucose in the insulin suppressed group as compared to control group. This demonstrates that pharmacological suppression of insulin was</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed unable to raise plasma glucose level in a fasting steady state. There was somewhat inconsistent but significant rise post glucose load. Diazoxide is the only agent in which both rodent and human data are available. The results of the two are very similar and separating them does not alter the significance.</ns0:p><ns0:p>Table <ns0:ref type='table'>6</ns0:ref>:Analysis of the fasting and post-feeding glucose levels in the control and treatment with Diazoxide or Octreotide: The table shows, out of the total number of experiments used for the analysis, in how many the mean of the treated (T) was greater than the control means (C) and in how many the trend was reverse. This relative position of the means across studies is compared non-parametrically to see whether the trend across studies was non-random, significant ones being indicated by asterisk. The table also gives in how many studies T was significantly greater than C and vice versa. We found more means of insulin suppression in which GTT after suppression was reported, but there were not many published replications of the experiments coming independently from different research groups. Therefore, any systematic review was not warranted. We briefly review their results here.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.3'>Suppression by Protein restriction:</ns0:head><ns0:p>Dietary protein deprivation is another method of insulin suppression. This also led to a decrease in plasma insulin levels; however fasting glucose levels did not increase <ns0:ref type='bibr' target='#b106'>(Schteingart et al., 1979)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.4'>Suppression by insulin siRNA:</ns0:head><ns0:p>Transgenic mice for insulin-siRNA along with IDE overexpression, showed decreased levels of insulin. Again the fasting glucose levels remained normal while there was a change in glucose tolerance curve (figure <ns0:ref type='figure' target='#fig_19'>9</ns0:ref>) <ns0:ref type='bibr' target='#b61'>(Hwang et al., 2007)</ns0:ref>. The where, in individuals with impaired insulin signalling the glucose peak is higher which returns to steady state much later than the controls, but the fasting steady state level is not different.</ns0:p><ns0:p>Figure <ns0:ref type='figure' target='#fig_19'>9</ns0:ref>: Intra peritoneal glucose tolerance test of control and siRNA treated mice. ). Fasting glucose levels in both the siRNA treated (represented by solid lines) and untreated group (represented by dotted lines) remain unaltered in male (represented by black lines) and female (represented by grey lines) mice. 15 minutes after the glucose injection, the treated mice show higher glucose levels relative to the untreated mice in both the male and female groups; and this effect is seen throughout till 120 minutes. Figure redrawn from data by <ns0:ref type='bibr' target='#b61'>(Hwang et al., 2007)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head n='2.5'>Suppression of insulin by partial gene ablation:</ns0:head><ns0:p>In rodents, there are two insulin genes Ins1 and Ins2 <ns0:ref type='bibr' target='#b37'>(Duvillie et al., 1997)</ns0:ref>. A double knockout of both the genes results in death, but ablation of either of the genes does not alter the glucose tolerance significantly suggesting redundancy <ns0:ref type='bibr' target='#b79'>(Mehran et al., 2012)</ns0:ref>. There are studies in which one gene is completely knocked out and the other one is a heterozygote <ns0:ref type='bibr' target='#b34'>(Dionne et al., 2016;</ns0:ref><ns0:ref type='bibr' target='#b79'>Mehran et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b90'>Page et al., 2019;</ns0:ref><ns0:ref type='bibr' target='#b117'>Templeman et al., 2015</ns0:ref><ns0:ref type='bibr' target='#b119'>Templeman et al., , 2017))</ns0:ref>. Reduced insulin gene dosage did not consistently result into fasting hyperglycemia in these studies although it offered protection against some of the effects of hyperinsulinemia.</ns0:p></ns0:div> <ns0:div><ns0:head>3.Steady state versus perturbed state glucose in STZ rats</ns0:head><ns0:p>Among the STZ treated rats, all the animals showed significantly higher post load glucose than the control group on all the 12 days sampled. However, in 10 out of the 12 days the 16-hour fasting glucose was not significantly different from the control although the variance was substantially greater than that of the control (figure <ns0:ref type='figure' target='#fig_8'>10</ns0:ref>).</ns0:p><ns0:p>A close look at the time course of fasting in the two groups revealed that in 4 out of 9 STZ animals the glucose levels reached the normal range but with substantial delay as compared to control animals. In two more animals the levels did not reach the normal range till 16 hours but a monotonic decrease continued throughout the period, indicating that their blood glucose may not have reached a steady state in 16 hours. Only in 3 animals the 16-hour glucose was higher than the control range with some indications of stabilizing at a higher level. In the time course experiment, the animals are handled frequently leading to some unavoidable stress, which may have influenced the glucose levels. In the 12 day follow up experiment, plasma glucose is estimated only after 16 hours and here there is no significant difference in the control and STZ animals on 10 out of 12 days. Furthermore, the individuals that showed higher 16 hour fasting glucose did not do so consistently. In the 12 days follow up, the distribution of 16-hour fasting glucose was typically skewed with one or two outliers having high glucose levels. Interestingly the outliers were not the same animals every day. There was considerable day to day variation in individuals and averaged over the 12 days, none of the STZ animals showed significantly higher fasting glucose than the controls although they consistently showed higher post feeding glucose.</ns0:p><ns0:p>Thus, these experiments show on the one hand that STZ treatment failed to increase steady state glucose levels significantly and consistently. showed a monotonic decrease in glucose levels. In three animals the glucose levels reduced at or below the control levels and in two others they showed a continued monotonic decrease but did not reach the normal level in 16 hours.</ns0:p><ns0:p>Filled triangles with dotted lines represent the individual time courses of the three STZ treated rats which showed some indications of stabilizing at a steady state above the normal.</ns0:p></ns0:div> <ns0:div><ns0:head n='4.'>Theoretical and mathematical considerations</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>In glucose homeostasis, in a fasting state, liver glucose production is analogous to the inlet tap and tissue glucose uptake analogous to the size of the outlet (figure <ns0:ref type='figure' target='#fig_0'>2</ns0:ref>), both being a function of insulin signalling. Most models of glucose regulation assume CSS <ns0:ref type='bibr' target='#b10'>(Bergman, 1989;</ns0:ref><ns0:ref type='bibr' target='#b11'>Bergman, 2005;</ns0:ref><ns0:ref type='bibr' target='#b77'>Makroglou et al., 2006;</ns0:ref><ns0:ref type='bibr' target='#b78'>Matthews et al., 1985;</ns0:ref><ns0:ref type='bibr' target='#b123'>Tomasi et al., 1967;</ns0:ref><ns0:ref type='bibr' target='#b126'>Turner et al., 1979;</ns0:ref><ns0:ref type='bibr' target='#b92'>Palumbo et al., 2013)</ns0:ref>. According to the CSS, the steady state/fasting level of glucose (level of water in the water tank analogy) is decided by the difference between the inlet tap (liver glucose production) and outlet (tissue glucose uptake) (figure <ns0:ref type='figure' target='#fig_12'>2A</ns0:ref>). Both liver glucose production and tissue glucose uptake depend on insulin signalling, hence a change in insulin signalling (change in inlet and outlet flow rates) would result in an altered level of glucose (steady-statewater level would change).</ns0:p><ns0:p>It </ns0:p></ns0:div> <ns0:div><ns0:head>Testable predictions from the models:</ns0:head><ns0:p>It is possible to make other testable predictions of TSS and CSS models. In the normal healthy individual, increased glucose utilization is expected to decrease fasting glucose levels by the CSS PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed model but not by the TSS model. Human experiments have shown that sustained exercise does not reduce plasma glucose, in fact it might increase <ns0:ref type='bibr' target='#b29'>(Coggan, 1991)</ns0:ref>. In order to match with experimental data, CSS based models of glucose dynamics during exercise need to include additional terms which involve neuronal mechanisms such as direct stimulation of liver glucose production in response to exercise through sympathetic route <ns0:ref type='bibr' target='#b103'>(Roy &amp; Parker, 2006)</ns0:ref>. This brings the model close to a TSS model. If TSS model describes glucose homeostasis more appropriately, reduced insulin signalling is not expected to change steady state glucose but only alter the time course to reach a steady state.</ns0:p><ns0:p>The mechanism of attaining a hyperinsulinemic normoglycemic prediabetic state is different by the CSS and TSS models. By the classical CSS based pathway, obesity induced insulin resistance is believed to be primary. The insulin resistance reduces glucose uptake and the excess glucose triggers a compensatory insulin response. The resultant hyperinsulinemia compensates for insulin resistance keeping the fasting glucose levels normal. Detailed analysis of the model and matching its prediction with empirical data has refuted this model <ns0:ref type='bibr' target='#b25'>(Chawla et al., 2018)</ns0:ref>. One of the intuitively appealing reasons for this refutation is that after the heightened insulin levels normalize glucose, there is no reason why insulin levels remain high. Therefore, a steady state with hyperinsulinemia and normoglycemia is impossible by the CSS model but it exists in a prediabetic state. If a 'compensatory' insulin response is mediated by glucose, one would expect a positive correlation between fasting glucose (FG) and fasting insulin (FI) and no correlation between insulin resistance and &#946; cell responsiveness.</ns0:p><ns0:p>By the TSS model, on the other hand, compensatory response is possible in either way. Primary insulin resistance may increase the glucose levels transiently, but when glucose sensing mechanisms detect the change a compensatory response can be operational. By this mechanism a hyperinsulinemic normoglycemic state is possible. Alternatively, primary hyperinsulinemia <ns0:ref type='bibr' target='#b30'>(Corkey, 2012;</ns0:ref><ns0:ref type='bibr' target='#b42'>Garvey et al., 1986;</ns0:ref><ns0:ref type='bibr' target='#b109'>Shanik et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b135'>Weyer et al., 2000)</ns0:ref> can also be compensated by increased insulin resistance by hitting the lower level of sensing which would trigger compensatory insulin resistance. Even in this case a hyperinsulinemic normoglycemic state is possible. Both glucose sensing neurons and neuronal regulation of insulin release and liver glucose production are well known. In the compensatory response mediated by TSS PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed pathways there need not be a correlation between fasting insulin and fasting glucose, but insulin resistance and &#946; cell response would be correlated.</ns0:p><ns0:p>Also using a simple CSS model (see supplementary information 2 for details), simulations show that the correlation coefficient and regression slope in the insulin-glucose relationship would remain the same in the fasting as well as post-meal state although the range of glucose and insulin levels will be different. On the other hand, in a TSS model the post-meal glucose and insulin levels are expected to be correlated but the steady state levels may not. We test these predictions by the two alternative models using human epidemiological data below.</ns0:p><ns0:p>We argued above that since on impairment of insulin signalling, the time required to reach a steady state can be substantially longer, overnight fasting may not ensure a steady state in all individuals. Fasting hyperglycaemia in T2D can have two alternative (but not mutually exclusive) causes. Either it represents the failure to reach a steady state in the specified fasting period, or it is because of mechanisms other than reduced insulin action. The TSS model can make differential predictions from the two alternative causes since it predicts a positive correlation between plasma glucose and plasma insulin in the post-meal state but loss of this correlation on reaching a steady state. In population data, if some individuals have reached a steady state but a few others haven't we would expect a correlation significantly weaker than the post-meal correlation. These predictions can be tested in epidemiological data.</ns0:p></ns0:div> <ns0:div><ns0:head n='5.'>Analysis of insulin glucose relationship in steady and perturbed state in human data:</ns0:head></ns0:div> <ns0:div><ns0:head>Epidemiological inquiry</ns0:head><ns0:p>In all the three data sets there was weak (R 2 range 0.017 to 0.057) but significant correlation between fasting glucose (FG) and fasting insulin (FI) and strong correlation between HOMA-IR and HOMA-&#946; (R 2 range 0.20 to 0.83) (figure <ns0:ref type='figure' target='#fig_9'>11</ns0:ref>). It was seen in all three data sets that the correlation coefficients for glucose and insulin were an order of magnitude higher in the post-meal cross sectional data than in the fasting state (Table <ns0:ref type='table' target='#tab_7'>7</ns0:ref> and fig <ns0:ref type='figure' target='#fig_9'>11</ns0:ref>). Also, the regression slopes in the post-meal data were substantially different from fasting data unlike what is expected by the CSS model (Supplementary information 2). By both the sets of predictions the CSS model predictions are rejected. The HOMA-IR HOMA-&#946; correlation, as well as the difference between the regression correlation parameters between fasting and post-meal data are compatible with predictions of the TSS model. However, although weak, there is significant correlation between FG and FI unlike what may be expected by a steady state TSS model. This incompatibility is not sufficient to falsify the TSS model since the failure of a small proportion of individuals to have reached a steady state at overnight fasting is sufficient to explain the weak correlation. It is also likely that the assumption of fasting may not be true for the entire sample. Even if a small number of individuals do not comply with the overnight fasting instructions, a positive correlation can result and this possibility is extremely difficult to exclude in human data.</ns0:p><ns0:p>The support of TSS model over CSS model is important because it accounts for the failure of impairment of insulin signalling to alter fasting glucose but increase only post load glucose. </ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The five approaches examined above fail to support the classical belief about glucose insulin relationship. The insulin receptor knock-out experiments and insulin suppression or enhancement experiments converge to show that alteration in insulin levels or insulin sensitivity does not change the steady state glucose levels. Evidence that it changes the shape of the glucose curve after food intake or glucose loading is more convincing in spite of some inconsistency across different experiments. Typically return to the steady state is delayed by impaired insulin</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed signalling but the steady state glucose level remains unchanged. Convergence of experiments using other means of causing specific alterations in insulin action strengthens the inference.</ns0:p><ns0:p>A number of mathematical models attempt to capture the dynamics of glucose homeostasis. A good model should be able to explain all the empirical results summed up here namely the inability of insulin receptor knockouts, insulin suppression and insulin enhancement experiments to alter steady state glucose levels; the difference in the regression correlation parameters between insulin and glucose in the steady versus perturbed state; the extremely weak correlation between fasting glucose and fasting insulin, but very strong correlation between HOMA-IR and HOMA-&#946;; the hyperinsulinemic-normoglycemic prediabetic state and the phenomenon of impaired glucose tolerance but normal fasting glucose. Reviewing models of glucose homeostasis is beyond the scope of this paper, but we outline here what a good model of glucose homeostasis needs to explain. In our observation, all existing models explain only some of the empirical findings. We suggest here that this inability is because of a questionable common baseline assumption of all models that insulin signalling determines the glucose level in the fasting as well as post feeding conditions. It should be possible to construct a model which is compatible with all experimental and epidemiological findings, if we realize that insulin affects glucose only in the post feeding but not in fasting conditions.</ns0:p><ns0:p>It is difficult to defend the classical assumptions about glucose-insulin relationship against the multiple convergent lines of evidence. Although results of these experiments have been there in the published literature for about two decades, these results were mostly explained away giving different excuses for different sets of experiments. The possible lines of defense would include difference between homeostatic mechanisms in rodents and humans or the possibility of nonlinear nature of glucose-insulin relationship. The evidence reviewed here comes from rodents as well as humans and the glucose insulin scatters do not show any clear indication of non-linearity. Manuscript to be reviewed other explanations will have to be supported by giving evidence for the assumptions made in those explanations.</ns0:p><ns0:p>The failure of experimental alteration in insulin signalling to alter steady state glucose raises two distinct possibilities about fasting hyperglycaemia in T2D. One is that fasting hyperglycaemia in T2D is a result of processes independent of insulin signalling such as autonomic signalling or other insulin independent mechanisms. The sympathetic tone is known to be altered in metabolic syndrome <ns0:ref type='bibr' target='#b121'>(Thorp &amp; Schlaich, 2015)</ns0:ref> and increased sensitivity of liver to sympathetic signal is likely to be mainly responsible to fasting hyperglycaemia <ns0:ref type='bibr' target='#b21'>(Bruce et al., 1992)</ns0:ref>. The other possibility is that with impaired insulin signalling overnight fasting is not sufficient to reach a steady state, therefore fasting hyperglycaemia in T2D is a non-steady state phenomenon in type 2 diabetes. The considerably weaker but still significant correlation between glucose and insulin in fasting as compared to post glucose load data suggests that both the factors are likely to be operational differentially in different individuals.</ns0:p><ns0:p>In either case certain fundamental concepts in our understanding of T2D need to be revised. First of all, the definition and measurement of insulin resistance using steady state glucose and insulin levels needs to be questioned. Most commonly used indices of insulin resistance are based on the assumption that insulin signalling decides the fasting steady state glucose levels, although nonequilibrium methods of assessing insulin resistance have been described <ns0:ref type='bibr' target='#b95'>(Patarr&#227;o et al., 2012)</ns0:ref>.</ns0:p><ns0:p>In the classical view other mechanisms of glucose regulation are assumed to be absent or nonsignificant. If increased sympathetic signalling increases liver glucose production, HOMA-IR will still account it as 'insulin resistance'. The same is true about insulin resistance measured by hyperinsulinemic euglycemic clamp. The way insulin resistance is measured at the clinical level eliminates the chance of separately accounting for other mechanisms of glucose regulation. Even when experiments show that certain agents affect glucose dynamics independent of insulin action, they are typically labelled as 'insulin sensitizing' agents <ns0:ref type='bibr' target='#b56'>(Hossain et al., 2018)</ns0:ref>. As a result, the belief that insulin is the only mechanism of glucose regulation relevant to T2D is artificially strengthened. There is a subtle circularity in the working definition of insulin resistance. Insulin resistance is blamed for the failure of normal or elevated levels of insulin to regulate glucose. In order to test this hypothesis, we should have an independent definition and The question can be turned upside down to examine whether steady state glucose level determines steady state insulin. If glucose is infused with a constant rate over a long time, insulin levels will come back to the baseline levels if glucose is not a determinant of fasting insulin. If it is, then insulin levels will stabilize at a new heightened steady state level. <ns0:ref type='bibr' target='#b63'>Jetton et al. (2008)</ns0:ref> infused intra venous glucose (20% glucose w/v) continuously for 4 days in rats. Both glucose and insulin levels increased significantly after the infusion. However, later both glucose and insulin levels came back to normal even as the infusion continued. Increase in the concentration of the infused glucose (up to 35%) also yielded similar results <ns0:ref type='bibr' target='#b112'>(Steil et al., 2001)</ns0:ref>. Thus, immediately on perturbation, glucose affected insulin levels, however after allowing sufficient time to regain steady state, the continued infusion of glucose had no significant effect on insulin levels. This demonstrates that even glucose does not hold a causal relationship with insulin in a steady state whereas glucose level perturbation is certainly known to stimulate insulin response.</ns0:p><ns0:p>The evidence reviewed above indicates that insulin is a driver but not a navigator of glucose homeostasis. A non-zero level of insulin is required for reaching a homeostatic steady state. In type 1 diabetes, the almost complete absence of insulin prevents glucose homeostasis. In type 2 diabetes there are non-zero insulin levels and therefore, a steady state is possible, but insulin itself plays little role in deciding the steady state glucose level. It is more likely that neuronal and other hormonal-metabolic factors affect the steady state glucose in T2D.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>From the insulin-glucose lesson we can learn certain general principles and norms for making causal inferences from experimental results. There are certainly more examples and implications of the driver navigator distinction than the ones captured by the insulin example. We see these phenomena in simple physico-chemical systems as well as in complex biological systems. An azeotropic mixture or a constant boiling solution has a steady state proportion of two or more liquids. If one starts from a non-equilibrium mixture, the rate of heating affects the time required to reach the azeotropic equilibrium but does not affect the equilibrium composition. This is because the molecular dynamics that decides the equilibrium has more complex elements than simple evaporation(D. <ns0:ref type='bibr' target='#b75'>Li et al., 2020)</ns0:ref> It is possible that in a given system some causal factors affect both steady and perturbed states where as others affect only one of them. For example, stirring increases the rate of dissolution of a solute, but does not affect the saturation concentration. Heating, on the other hand, can affect both.</ns0:p><ns0:p>Some of the well studies examples in complex systems include the regulation of plasmid numbers in bacteria. The rate of replication of plasmids does not decide the stable plasmid copy number since there is a complex set of mechanisms that decide the steady state independent of the rate of DNA replication (M. M. <ns0:ref type='bibr' target='#b134'>Watve et al., 2010)</ns0:ref>.</ns0:p><ns0:p>The differences between driver and navigator causes may also be specific to the context. For example, the rate of lexicon building at an early age may or may not be determining the adult vocabulary and language skills depending upon what is being tested (F. <ns0:ref type='bibr' target='#b55'>Hill et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b60'>Hurtado et al., 2008;</ns0:ref><ns0:ref type='bibr' target='#b96'>Peter et al., 2019)</ns0:ref> In warm blooded animals, in a healthy state, the body temperature is maintained constant by homeostatic mechanisms. If the environmental temperature changes suddenly, heat transfer between the body and the environment is inevitable and as a result the body temperature will change transiently. However, the homeostatic loops will be operative and bring the body temperature back to normal soon. If an experimenter measures the body temperature immediately after changing the environmental temperature, a causal relationship might be apparent. If the body temperature is measured after reaching the steady state, one may infer that environmental temperature has no causal relationship with body temperature. If the experimenter is measuring heat transfer as the mechanism of causation, then it will be demonstrable at any time. So, a causal mechanism exists all the time but demonstration of a causal mechanism or pathway is not sufficient to establish a causal role. A sub-normal driver will delay the time to destination but will not change the destination. On the other hand, changing the navigator may or may not alter the time, but will alter the position of the destination. In the history of insulin research, early experiments such as total pancreatectomy demonstrated the necessary role of insulin in glucose homeostasis but the distinction between driver or navigator causality was not even conceptually perceived. So, it was assumed that insulin does both the roles. Although the absence of correlation between fasting glucose and insulin but good correlation after perturbation was noted as early as 1969 <ns0:ref type='bibr' target='#b45'>(Goodner et al., 1969)</ns0:ref> in the absence of conceptual differentiation between steady state and perturbed state causality, a clear interpretation did not emerge. Now in the presence of multiple experiments</ns0:p><ns0:p>showing the precise role of insulin, we need to revive our concepts of causality. Although the experiments reviewed above have been there in the literature, the inability of insulin to alter fasting glucose was not appreciated before, only because the distinction between driver and navigator causality was not there in the philosophical foundation of biological causality.</ns0:p><ns0:p>For any homoeostatic system in which there can be one or more feedback loops working as homeostatic mechanisms and one or many possible perturbations, all the perturbations can have a demonstrable causal role in a non-equilibrium state but only some of the components of the feedback loops may have steady state causal relationship with the controlled parameter. If there are more than one homeostatic mechanisms operating on a controlled variable, it is possible that components of any one loop do not affect the steady state level of the variable. This distinction is important in building the philosophy of experimental physiology and medicine. Most of the experiments in physiology are perturbation experiments and often we try to understand the steady state in a homeostatic system using such experiments. This situation warrants care in making inferences in the absence of which our understanding of physiology can remain flawed.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Lotka-Volterra model of two competing species at different growth rates Simulated population dynamics of two competing species A (black lines) and B (grey lines) at different growth rates. Solid lines represent high growth rates reaching the equilibrium faster whereas dotted lines represent slow growth rates reaching the equilibrium more slowly. In both cases species A has slower growth rate but greater carrying capacity than species B. At time T1 the populations are in proportion to their growth rates but at T2, approaching equilibrium, growth rates become increasingly irrelevant in determining population sizes.</ns0:p><ns0:p>Thus, growth rates are important determinants of population size in a perturbed state but not at a steady state. Manuscript to be reviewed List of publications used in the final analysis: references numbers <ns0:ref type='bibr' target='#b15'>Bl&#252;her et al., 2002;</ns0:ref><ns0:ref type='bibr'>Br&#252;ning et al., 1998;</ns0:ref><ns0:ref type='bibr'>Cohen et al., 2004;</ns0:ref><ns0:ref type='bibr'>Dodson Michael et al., 2000;</ns0:ref><ns0:ref type='bibr'>Ealey et al., 2008;</ns0:ref><ns0:ref type='bibr'>Escribano et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b50'>Guerra et al., 2001;</ns0:ref><ns0:ref type='bibr'>Haas et al., 2012;</ns0:ref><ns0:ref type='bibr'>Kawamori et al., 2009;</ns0:ref><ns0:ref type='bibr'>Lauro et al., 1998;</ns0:ref><ns0:ref type='bibr'>Mauvais-Jarvis et al., 2000;</ns0:ref><ns0:ref type='bibr'>Okada et al., 2007;</ns0:ref><ns0:ref type='bibr'>Otani, 2003;</ns0:ref><ns0:ref type='bibr' target='#b104'>Sakaguchi et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b111'>Softic et al., 2016;</ns0:ref><ns0:ref type='bibr'>Wojtaszewski et al., 1999)</ns0:ref> </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2: The consequential steady state (CSS) (A) and targeted steady state (TSS) (B) models of homeostasis illustrated with a tank water level analogy. In CSS a change in the size of inlet or outlet tap, analogous to insulin sensitivity can change the steady state level. In a TSS model, a change in the tap size will alter the time required to</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3: Glucose levels for control (black squares) and FIRKO (red squares) at steady state and perturbed state. The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4: Glucose levels for control (black squares) and MIRKO (red squares) at steady state and perturbed state. The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the shortlisted papers in which control and MIRKO were compared using an OGTT.MIRKO glucose levels are normalized to that of the control and the difference is expressed with &#177; 95% CI. Glucose levels of control are expressed as 0 &#177; 95% CI. Steady state is represented by the fasting glucose (A) and the perturbed state is represented by different time points post glucose load when the readings are taken: (B) 15 minutes (C) 30 minutes (D) 60 minutes (E) 90 minutes and (F) 120 minutes.Note that fasting glucose does not differ from the control in any of the experiments.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 5 :</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure5: Glucose levels for control (black squares) and LIRKO (red squares) at steady state and perturbed state. The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the shortlisted papers in which control and LIRKO were compared using an OGTT.LIRKO glucose levels are normalized to that of the control and the difference is expressed with &#177; 95% CI. Glucose levels of control are expressed as 0 &#177; 95% CI. Steady state is represented by the fasting glucose (A) and the perturbed state is represented by different time points post glucose load when the readings are taken: (B) 15 minutes (C) 30 minutes (D) 60 minutes (E) 90 minutes and (F) 120 minutes.Note that inconsistent with classical belief, liver specific insulin receptor knockout does not show significant effect on fasting glucose in any of the experiments. On the other hand, post load glucose is consistently higher. Figure6: Glucose levels for control (black squares) and &#946;IRKO (red squares) at steady state and perturbed state.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 6 :</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure5: Glucose levels for control (black squares) and LIRKO (red squares) at steady state and perturbed state. The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the shortlisted papers in which control and LIRKO were compared using an OGTT.LIRKO glucose levels are normalized to that of the control and the difference is expressed with &#177; 95% CI. Glucose levels of control are expressed as 0 &#177; 95% CI. Steady state is represented by the fasting glucose (A) and the perturbed state is represented by different time points post glucose load when the readings are taken: (B) 15 minutes (C) 30 minutes (D) 60 minutes (E) 90 minutes and (F) 120 minutes.Note that inconsistent with classical belief, liver specific insulin receptor knockout does not show significant effect on fasting glucose in any of the experiments. On the other hand, post load glucose is consistently higher. Figure6: Glucose levels for control (black squares) and &#946;IRKO (red squares) at steady state and perturbed state.The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the shortlisted papers in which control and &#946;IRKO were compared using an OGTT.&#946;IRKO glucose levels are normalized to that of the control and the difference is expressed with &#177; 95% CI. Glucose levels of control are expressed as 0 &#177; 95% CI. Steady state is represented by the fasting glucose (A) and the perturbed state is represented by different time points post glucose load when the readings are taken: (B) 15 minutes (C) 30 minutes (D) 60 minutes (E) 90 minutes and (F) 120 minutes.Note that fasting glucose does not differ from the control in any of the experiments.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Figure 7 :</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Glucose levels for control (black squares) and IDE-inhibition (red squares) models at steady and perturbed state. The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the shortlisted papers in which control and IDE-inhibition were compared using an OGTT. IDEinhibition glucose levels are normalized to that of the control and the difference is expressed with &#177; 95% CI.Glucose levels of control are expressed as 0 &#177; 95% CI. Steady state is represented by the fasting glucose (A) and the perturbed state is represented by different time points (B to F) post glucose load when the readings are taken: (B) 15 minutes (C) 30 minutes (D) 60 minutes (E) 90 minutes and (F) 120 minutes.Note that fasting glucose does not differ</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Figure 8 :</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Glucose levels for control (black squares) and treated (red squares) models at steady and perturbed state.The X-axis represents the glucose levels and the Y-axis represents experiments from different studies from the shortlisted papers in which control and treated were compared using an OGTT. Glucose levels after treatment with Diazoxide/Octreotide are normalized to that of the control and the difference is expressed with &#177; 95% CI. Glucose levels of control are expressed as 0 &#177; 95% CI. For Diazoxide treatment, steady state is represented by the fasting glucose (A) and the perturbed state is represented by (B) at 120 minutes post glucose load. For Octreotide treatment, steady state is represented by the fasting glucose (C) and the perturbed state is represented by (D) at 120 minutes post glucose load.Note the inconsistencies across studies.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020) Manuscript to be reviewed curves in figure 9 are typical of insulin receptor knockout or insulin suppression experiments</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 10 :</ns0:head><ns0:label>10</ns0:label><ns0:figDesc>Figure 10: Treatment of SD rats with STZ. (A): The 16-hour fasting and post-meal glucose values of treated (STZ 50mg/kg) and control rats (Citrate buffer CB) over 12 days. N= 12 for each group. Note that on 10 out of 12 days the mean fasting glucose of the treated group (STZ-F) was not significantly different from the control mean (CB-F). Post feeding the treated group (STZ-P) has substantially greater mean than the control (CB-P). (B) Time course of glucose during 16 hours fasting. X axis represents the time after removal of food when the glucose readings are taken and Y axis represent the glucose levels. The grey band represents the upper and lower bounds of 95% CI of the control group with the mean glucose values represented by filled circles. Filled squares represent individuals that</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 11 :</ns0:head><ns0:label>11</ns0:label><ns0:figDesc>Figure 11: The Fasting Glucose-Fasting Insulin, Post-meal Glucose-post-meal Insulin and HOMA-IR HOMA-&#946; scatter plots in non-diabetic populations in the three data sets. The FG-FI correlation is weak as compared to post-</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Further</ns0:head><ns0:label /><ns0:figDesc>it would be prudent to avoid making inferences based on dietary or other complex interventions since they can have multiple mechanisms of action. Specific genetic or molecular interventions are more revealing with respect to the underlying mechanisms since we can be more confident about their specificity of action. Therefore our inference that insulin action does not influence fasting glucose levels is the most straightforward and parsimonious inference. Any PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)Manuscript to be reviewed measure of insulin resistance. Only then we can test whether and to what extent insulin resistance can alter glucose dynamics. However, clinically insulin resistance is measured by the inability of insulin to regulate glucose. Such a measure cannot be used to test the hypothesis that insulin resistance leads to the failure of insulin to regulate glucose. The unfalsifiability of the insulin resistance hypothesis arising out of this circularity has halted any attempts towards realistic assessments of the true causes of fasting hyperglycaemia in type 2 diabetes. In the molecular approach to induce insulin resistance, we have an independent definition and causality for insulin resistance and therefore such experiments are free from circularity of definition. The results of such experiments reviewed here are therefore more revealing and reliable. Since all of them converge to show that altering insulin signalling does not alter steady state glucose levels, the insulin resistance and inadequate compensation hypothesis for steady statehyperglycaemia stands clearly rejected.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_15'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_16'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_17'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_18'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_19'><ns0:head>Figure 9</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_20'><ns0:head /><ns0:label /><ns0:figDesc>Figure redrawn from data by<ns0:ref type='bibr' target='#b61'>(Hwang et al., 2007)</ns0:ref> .</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_21'><ns0:head>Figure 10 Steady</ns0:head><ns0:label>10</ns0:label><ns0:figDesc>Figure 10</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_22'><ns0:head /><ns0:label /><ns0:figDesc>secondary screening Study showing similar methods of making the insulin receptor knockout; had fasting and post glucose bolus readings of the control and knockout Number of papers shortlisted based on screening the full-text and back referencing (data extracted from these papers) 16</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='42,42.52,382.87,525.00,363.00' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Systematic literature review for studies on tissue specific insulin receptor knockouts.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 2 :</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Systematic literature review for studies on insulin degrading enzyme inhibition/knockout.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 3 :</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Systematic literature review for studies on insulin suppression with diazoxide and octreotide</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 4 :</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Collective analysis of the fasting and post-feeding glucose levels in the control and IRKOs: The table</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_6'><ns0:head /><ns0:label /><ns0:figDesc>The TSS model requires mechanisms of sensing any departure from the targeted steady state. Such mechanisms are not known in peripheral systems but glucose sensing neurons are certainly known to be present in the brain. Therefore, if TSS is a more appropriate model, the CNS mechanisms are likely to be central to glucose homeostasis, particularly in determining the steady state levels; whereas insulin signalling would play a role in determining the rate at which a steady state is reached after perturbation.</ns0:figDesc><ns0:table /><ns0:note>has not been critically examined whether CSS or TSS describes glucose homeostasis more appropriately. This is important because if TSS model is appropriate, insulin resistance and relative insulin deficiency will not result into altered steady state glucose levels although the time required for reaching a steady state after perturbation might change. Comparing the TSS with the water tank analogy, if insulin resistance affects the liver glucose production (inlet tap) or the tissue glucose uptake (outlet tap), because of the presence of upper and lower set points, insulin resistance and the relative insulin insufficiency might change the time required to attain the steady state, but not the steady state itself (figure2B). If CSS model is appropriate, insulin resistance or altered insulin levels are bound to change fasting glucose levels. The failure of insulin receptor knockouts and insulin suppression experiments to alter the fasting steady state, along with the delay in reaching the steady state indicates that TSS model is likely to describe glucose homeostasis more appropriately.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_7'><ns0:head>Table 7 :</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Correlation and regression parameters of glucose-insulin relationship at steady and perturbed states.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:figure type='table' xml:id='tab_8'><ns0:head /><ns0:label /><ns0:figDesc>These set of examples demonstrate that causation may work differently in steady versus perturbed states and the need to differentiate between drive and navigator causality is widespread in different fields of science. There is a need for a detailed philosophical account on this issue as well as a set of working norms for practicing science. We suggest a few norms specifically for the field of experimental physiology.Certain kinds of experimental interventions are unable to distinguish between driver versus navigator causality. Knocking out a driver or a navigator will disable the journey to the destination. Therefore, complete knockout of a cause may not distinguish between driver and navigator causality. On the other hand, experiments quantitatively altering the level of the causal</ns0:figDesc><ns0:table /><ns0:note>factor while keeping it non-zero has a different set of implications. If a perturbation is momentary or transient, the results obtained would certainly reflect perturbed state causality, but may not reflect steady state causality. On the other hand, sustained perturbations held constant for sufficiently long to allow the system to regain a steady state are necessary to establish steady state causality. If upon sustainably altering a causal factor the effect variable returns to the same steady state, it reflects only perturbed state and not steady state causality. If, on the other hand, sustained alteration in the causal factor results into an altered steady state, it indicates steady state causality.</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_9'><ns0:head>Table 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>IRKO meta analysisSystematic literature review for studies on tissue specific insulin receptor knockouts.</ns0:figDesc><ns0:table /></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:07:50680:1:1:NEW 26 Oct 2020)</ns0:note> </ns0:body> "

"Dear Editors, We thank the reviewers for the constructive critical review. Please find below a point wise response to the comments and a guide to the changes in the manuscript. Reviewer 1 Basic reporting The study by Joshi and Watve highlights the possible misinterpretations of experimental data in biomedicine due to lack of the understanding of driver versus navigator causation. As a case study, they analyzed the glucose-insulin relationship to provide a potential proof of concept. Authors have systematically analyzed the published data/literature and performed the relevant experiments to revisit the glucose-insulin relationship and highlighted the previously misinterpreted and exaggerated roles of insulin in regulating blood glucose levels in humans. Experimental design The study is novel, rigorously performed and systematically presented. The methodology and search criteria used are relevant and the limitations are also clearly mentioned which makes the study more reliable. Sufficient evidences from the literature and the experimental data supports the claim of the study. The manuscript is written in a lucid and comprehensive way. We thank the reviewer for the appreciation. Validity of the findings However, some minor concerns must be addressed before publication. 1. References are missing at some places throughout the manuscript (for example in paragraph 246- 255, lines 471 and 473) The relevant references have been added in the lines 254-268, 486-499 as well as some other places (lines 153, 245) which we found after a thorough check again. 2. Figure No 8 and 9 are not sufficiently described and labeled. The figure legend should be written more comprehensively. It is difficult to understand what is shown in figure from the legend. More detail should be added in figure legends for figure no. 3-7. The legends for figures 3-8 have been expanded to now explain them in more detail. Additionally, figure 9 has been changed since we found that a different figure had got inserted rather than in intended one. Although both figures indicate the same pattern, the substituted figure is the appropriate one as cited. The legend also has been expanded. 3. Authors claim that fasting glucose in FIRKO knockouts remain unchanged, however some studies does not support this observation. Authors should carefully review these studies in detail to find the possible reason of inconsistency towards the conclusion of their original result. We looked carefully again at the FIRKO data to discern the causes for the inconsistencies. We now discuss the nature and causes of inconsistencies in the FIRKO data in the revised manuscript. (lines 488-504). 4. To generalize the concept of driver and navigator causation in biology and/or in a broader prospective, some more cases with similar observation from the literature should be added (with references) in the discussion., 5. Authors cite one hypothetical case of body temperature in the conclusion section, but this not supported relevant references. We have added a few more relevant examples to illustrate the generality of driver navigator causation. For the examples that represent classical textbook science, we do not cite specific references, but for all specific biological examples we have appropriate references in the revision. (lines 922-950). 6. Figure is not cited in result 2.5 (line no. 576) Here we have not done any new analysis, we are simply citing what the authors found. So, no new figure was generated other than what they have already published. Reviewer: ChaoyiJin Basic reporting The author is trying to say there are logical traps in interventional inferences because steady state and perturbed state are different. They concluded that it is necessary to differentiate between types of causalities. To demonstrates this, they used the insulin and glucose as an example and reached the conclusion that researchers may have to carefully re-examine causal inferences from perturbation experiments and set up revised norms for experimental design for causal inference. The paper is well written and reads well, the background information is adequate. The result is clear and informative. The figures are clear, and the subtitles are detailed. The reasonings in the paper are also clear. We thank the reviewer for the appreciation. However, there are some minor parts that needs to be improved. 1. Figure 2 is a great analogy, the author explained the analogy between 2A and insulin sensitivity, it would be great if they also explain the analogy between 2B and the glucose homeostasis model. We have expanded the explanation for the analogy between figure 2B and glucose homeostasis (lines 421-434 and 705-722). Experimental design The experiment design is good. The question is well defined and then the author used 5 different approaches to prove the concepts. We thank the reviewer for the appreciation. Validity of the findings The conclusion are well-stated. The reasoning is also reasonable and clear. We thank the reviewer for the appreciation. Kindly find uploaded the revised MS in the track changes mode as well as the clean copy. We hope to have incorporated all the suggestions and addressed all the concerns of the reviewers. Thanking you once again. -Manawa Diwekar-Joshi and Prof. Milind Watve "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Yes associated protein 1 (YAP1), which is a standout amongst the most essential effectors of the Hippo pathway, assumes a vital part in a few kinds of cancer. However, whether YAP1 is an oncogene in CRC (colorectal cancer) remains controversial, and the association between the subcellular localization of YAP1 and clinical implications in CRC remains unknown. Patients and methods: In this study, we investigated the subcellular localization of YAP1 in CRC cells by immunohistochemistry and then associate these findings with clinical information in a large CRC cohort with 922 CRC patients. Results:</ns0:p><ns0:p>The results show that CRC tissues has a significant higher expression of cytoplasmic YAP1 compared to adjacent normal tissues (all P&lt;0.001). Cytoplasmic YAP1 expression was significantly associated with the number of lymph nodes removed and differentiation grade (all P &lt; 0.001). Furthermore, after correcting confounding variables, for example, TNM stage and differentiation grade, the multivariate Cox analysis confirmed cytoplasmic YAP1-high subgroup had a significant shorter DFS(HR=3.255; 95% CI=2.290-4.627; P&lt;0.001) and DSS(HR=4.049; 95% CI=2.400-6.830; P&lt;0.001) than cytoplasmic YAP1-low subgroup. As compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy. Conclusion: Cytoplasmic YAP1 could be could be utilized as a prognosis factor in CRC patients, and may be an indicator of whether certain patients population could benefit from postoperative chemotherapy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The prevalence of CRC ranks third among all cancers in male and second in female <ns0:ref type='bibr' target='#b26'>(Torre et al. 2015)</ns0:ref>, and CRC has the third highest mortality rate among all cancers <ns0:ref type='bibr' target='#b37'>(Zeng et al. 2014)</ns0:ref>.</ns0:p><ns0:p>Surgical resection combined with chemotherapy remains the mainstay of treatment for CRC, in any case, numerous patients will progress to metastatic CRC and develop resistance to chemotherapeutic drugs <ns0:ref type='bibr' target='#b5'>(Fisher et al. 2015)</ns0:ref>, because signs or symptoms diagnose CRC usually appear in advanced phases <ns0:ref type='bibr' target='#b1'>(Binefa et al. 2014)</ns0:ref>. Even if some patients are diagnosed with CRC and undergo surgery at an early stage, 20%-30% of these patients will relapse within five years. <ns0:ref type='bibr' target='#b8'>(Hardingham et al. 2015)</ns0:ref>. Current CRC treatment regimen is heterogenous for patients, even for patients with the same TNM stage <ns0:ref type='bibr' target='#b21'>(Nagtegaal et al. 2012)</ns0:ref>, in any case, the indication for treatment should be assessed on an individual basis by considering the risk factors of relapse <ns0:ref type='bibr' target='#b17'>(Marin et al. 2012)</ns0:ref>. Currently, the only effective marker for the CRC prognosis and appropriate chemotherapy selection is microsatellite instability (MSI) <ns0:ref type='bibr' target='#b10'>(Hemminki et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b24'>Popat et al. 2005)</ns0:ref>, however, MSI as a CRC marker has not been applied clinically. Therefore, there is an urgent need for new biomarkers to assess the prognosis of CRC patients before and after treatment.</ns0:p><ns0:p>YAP1 is a standout amongst the most essential effectors of the Hippo pathway, which is a critical pathway regulating cell proliferation, apoptosis, and organ growth <ns0:ref type='bibr' target='#b11'>(Justice et al. 1995)</ns0:ref>.</ns0:p><ns0:p>Several studies have shown that YAP1 is an oncogene highly express in numerous cancer types including bladder cancer <ns0:ref type='bibr' target='#b15'>(Liu et al. 2013)</ns0:ref>, breast cancer <ns0:ref type='bibr' target='#b13'>(Kim et al. 2014)</ns0:ref>, gastric cancer <ns0:ref type='bibr' target='#b12'>(Kang et al. 2011)</ns0:ref>, hepatocellular cancer <ns0:ref type='bibr' target='#b33'>(Xu et al. 2009</ns0:ref>), nonsmall-cell lung cancer <ns0:ref type='bibr' target='#b31'>(Wang et al. 2010)</ns0:ref>, and CRC <ns0:ref type='bibr' target='#b30'>(Wang et al. 2013a;</ns0:ref><ns0:ref type='bibr' target='#b33'>Xu et al. 2009</ns0:ref>) that associate with tumor progression and poor prognosis. On the contrary, abundant literature suggested that YAP1 is a tumor suppressor gene and nuclear expression is reduced in different cancers, such as breast cancer <ns0:ref type='bibr' target='#b18'>(Matallanas et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b35'>Yu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b36'>Yuan et al. 2008)</ns0:ref>, head and neck cancers <ns0:ref type='bibr' target='#b4'>(Ehsanian et al. 2010)</ns0:ref>, hematological cancers <ns0:ref type='bibr' target='#b2'>(Cottini et al. 2014)</ns0:ref>, and CRC <ns0:ref type='bibr' target='#b14'>(Levy et al. 2007</ns0:ref>). These paradoxical reports remind us that the role of YAP1 in cancer is controversial, and it is crucial to make it clear the relationship between YAP1 expression and its clinical relevance in CRC. In addition, the nuclear overexpression of YAP1 is associate with poor survival in gastric cancer <ns0:ref type='bibr' target='#b12'>(Kang et al. 2011)</ns0:ref>, actually, previous researches suggest that subcellular localization of proteins is associated with functions associated of tumorigenesis and tumor progression <ns0:ref type='bibr' target='#b6'>(Garcia et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b16'>Lobo et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b28'>Vaquero et al. 2017)</ns0:ref>, a few studies suggested that YAP1 overexpression is associate with poor survival in CRC <ns0:ref type='bibr' target='#b30'>(Wang et al. 2013a;</ns0:ref><ns0:ref type='bibr' target='#b32'>Wang et al. 2013b;</ns0:ref><ns0:ref type='bibr' target='#b34'>Yang et al. 2018)</ns0:ref>, however, the association between subcellular localization of YAP1 and clinical significance in CRC has been previously <ns0:ref type='bibr' target='#b22'>(Pan et al. 2015)</ns0:ref>. Disease-free survival (DFS) was defined as the number of months Manuscript to be reviewed </ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Differences in YAP1 expression between CRC tissues and adjacent normal tissues.</ns0:head><ns0:p>To analyze the expression pattern of YAP1 in CRC tissues, we firstly utilized the datasets from public database, the results showed that in one TCGA dataset and three GEO datasets, YAP1 mRNA expression level was consistently significantly elevated in CRC tissues compared with the adjacent normal tissue (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_9'>1A-D</ns0:ref>), the other five GEO datasets also show the same results (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_9'>S1A</ns0:ref>). We subsequently investigated the expression pattern of YAP1 by IHC method in 997 CRC and 70 adjacent normal tissue samples which derive from patients who underwent surgery at Yunnan Cancer Hospital. The positive immunostaining results from YAP1 predominantly occurred in the cytoplasm and nucleus of colorectal epithelial cells (Fig. <ns0:ref type='figure' target='#fig_9'>1H-M</ns0:ref>), whereas the staining was negative or weak in mesenchymal cells (Fig. <ns0:ref type='figure' target='#fig_9'>1H-M</ns0:ref>). We calculated the H-score of cytoplasmic YAP1 and nuclear YAP1 independently, then the YAP1 NCR (nuclear/cytoplasmic ratio) was calculated, and there is a very weak positive correlation between cytoplasmic H-score and nuclear YAP1 H-score (Fig. <ns0:ref type='figure' target='#fig_9'>S1B</ns0:ref>). the results show that cytoplasmic YAP1 expression was significantly elevated in CRC tissues compared with the adjacent normal tissues (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_9'>1E</ns0:ref>), and nuclear YAP1 expression was significantly elevated in primary cancer tissues compared with the adjacent normal tissues (P&lt;0.0001; Fig. <ns0:ref type='figure' target='#fig_9'>1F</ns0:ref>), but the expression of nuclear YAP1 in adenomas and metastasis CRC tissues have no significant differences with the adjacent normal tissues (Fig. <ns0:ref type='figure' target='#fig_9'>1F</ns0:ref>), we also found YAP1 NCR (Nuclear/Cytoplasmic Ratio) gradually decrease in adjacent normal tissues, adenomas, primary cancers, and metastatic CRC (all P&lt;0.05; Fig. <ns0:ref type='figure' target='#fig_9'>1G</ns0:ref>). The results above indicated that the increased cytoplasmic YAP1 expression may be associated with the progression of CRC.</ns0:p></ns0:div> <ns0:div><ns0:head>Associations between YAP1 expression and CRC patients' clinicopathological characteristics</ns0:head><ns0:p>To obtain further information, we analyzed the association between cytoplasmic YAP1 expression levels or YAP1 NCR and CRC patients' clinicopathological characteristics. We found that the expression of cytoplasmic YAP1 protein was significantly higher in poorly+moderate grades than that in the well grade (P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_9'>S1C</ns0:ref>), and YAP1 NCR was significantly lower in poorly+moderate grades than that in the well grade (P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_9'>S1E</ns0:ref>), but there is no significant differences between poorly+moderate grades and well grade in the expression of nuclear YAP1 protein (Fig. <ns0:ref type='figure' target='#fig_9'>S1D</ns0:ref>), the clinicopathological features for the patients at poor+moderate grade or well grade were described in Table <ns0:ref type='table' target='#tab_2'>S1</ns0:ref>. Next, we classified the 919 patients (patients lost follow-up information were excluded) into cytoplasmic YAP1-low and cytoplasmic YAP1-high subgroups by the optimal cut-off value (H-score= 202.5) determined by the maxstat R package, the results showed there were significant differences between cytoplasmic YAP1-low and cytoplasmic YAP1-high subgroups with respect to the number of resected lymph nodes and differentiation grade. (all P &lt; 0.001; Table <ns0:ref type='table' target='#tab_2'>1</ns0:ref>). We also classified the 919 patients into YAP1 NCR-low and YAP1 NCR-high subgroups by the optimal cut-off value (NCR=0.0482) determined by the maxstat R package, the results showed a significant difference between the YAP1 NCR-low and YAP1 NCR-high subgroups in the TNM stage (P = 0.02; Table <ns0:ref type='table' target='#tab_4'>S2</ns0:ref>). The above results revealed that high cytoplasmic YAP1 expression may be involved in the aggressiveness of CRC.</ns0:p></ns0:div> <ns0:div><ns0:head>High cytoplasmic YAP1 expression is associated with a worse survival in CRC patients</ns0:head><ns0:p>A univariate and multivariate Cox regression analyses (Further evaluation of meaningful prognostic factors in univariate analysis in multivariate analysis) was applied to determined the independence of the prognostic value of YAP1 on the basis of the DFS and DSS of CRC patients, the results showed that high cytoplasmic YAP1 expression was an independent risk factor of DFS (HR=3.255; 95%CI=2.290-4.627; P&lt;0.001) and DSS (HR=4.049; 95%CI=2.400-6.830; P&lt;0.001) for CRC patients ( for CRC patients (Table <ns0:ref type='table' target='#tab_1'>2</ns0:ref>). Kaplan-Meier analyses with log-rank tests showed that DFS and DSS in the cytoplasmic YAP1-high subgroup were significantly shorter than the cytoplasmic YAP1-low subgroup (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_10'>2A, B</ns0:ref>), moreover, cytoplasmic YAP1-high subgroups were consistently had shorter DFS and DSS than cytoplasmic YAP1-low subgroups in stage &#8544;, &#8545;, or &#8546; CRC patients respectively (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_10'>2B-D, F-H</ns0:ref>). We also found DFS and DSS were significantly lower in YAP1 NCR-low subgroup than YAP1 NCR-high subgroup (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_10'>S2A</ns0:ref>). However, there is no significant differences between nuclear YAP1-high subgroup and nuclear YAP1-low subgroup in Kaplan-Meier analyses (all P&gt;0.05; Fig. <ns0:ref type='figure' target='#fig_10'>S2B</ns0:ref>).</ns0:p><ns0:p>Stage III CRC patients with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy To evaluate whether cytoplasmic YAP1 expression level could be an indicator of whether certain patients population could benefit from adjuvant chemotherapy, the stage III patients were divided into two groups respectively (all stage III patients received adjuvant chemotherapy), either did or did not receive adjuvant chemotherapy (Table <ns0:ref type='table'>S3</ns0:ref>), for stage III patients who received adjuvant chemotherapy, Kaplan-Meier analyses with log-rank tests showed that DFS and DSS in the cytoplasmic YAP1-high subgroup were significantly shorter than the cytoplasmic YAP1-low subgroup (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_11'>3A-B</ns0:ref>), but there were no significant differences between YAP1-high subgroup and YAP1-low subgroup in DFS and DSS for stage III patients without adjuvant chemotherapy (all P&gt;0.05; Fig. <ns0:ref type='figure' target='#fig_11'>3C-D</ns0:ref>) .</ns0:p><ns0:p>Besides, for stage III patients who received adjuvant chemotherapy, Kaplan-Meier analyses also showed that DFS and DSS in the low YAP1 NCR subgroup were significantly shorter than the high YAP1 NCR subgroup (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_11'>S3A-B</ns0:ref>), but there were no significant differences between high YAP1 NCR subgroup and low YAP1 NCR subgroup in DFS and DSS for stage III patients without adjuvant chemotherapy (all P&gt;0.05; Fig. <ns0:ref type='figure' target='#fig_11'>S3C-D</ns0:ref>). Therefore, as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy,in fact, the DFS and DSS shorten in cytoplasmic-high YAP1 subgroups who received adjuvant chemotherapy.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Primarily, the results in this study showed that in one TCGA dataset and eight GEO datasets, the mRNA expression of YAP1 in CRC tissues was consistently higher in CRC tissues Manuscript to be reviewed compared with the adjacent normal tissue. Further, the IHC examination of YAP1 confirmed that epithelial cytoplasmic YAP1 protein expression were significantly elevated in CRC tissues compared with the adjacent normal tissue in the Yunnan Cancer Hospital, and YAP1 NCR gradually decrease in adjacent normal tissues, adenomas, primary cancers, and metastatic CRC.</ns0:p><ns0:p>Prior studies had illustrated the expression of YAP1 in a various cancers including CRC <ns0:ref type='bibr' target='#b2'>(Cottini et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b4'>Ehsanian et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b12'>Kang et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kim et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b14'>Levy et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b15'>Liu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b18'>Matallanas et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b30'>Wang et al. 2013a;</ns0:ref><ns0:ref type='bibr' target='#b31'>Wang et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b33'>Xu et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Yu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b36'>Yuan et al. 2008)</ns0:ref>, but the association between subcellular localization of YAP1 and aggressiveness of CRC has been neglected. In this study, the expression pattern of YAP1 in the Yunnan Cancer Hospital cohort reveal that the increased cytoplasmic YAP1 expression may be associated with the progression of CRC.</ns0:p><ns0:p>The analysis of association between YAP1 expression and CRC patients' clinicopathological features showed that cytoplasmic YAP1 expression was related to differentiation grade and YAP1 NCR was related to TNM stage. Further, CRC patients were divided into cytoplasmic-high YAP1 and cytoplasmic-low YAP1 subgroups by the optimal cutoff value (H-score=202.5), meanwhile, classify CRC patients into YAP1 NCR-low and YAP1</ns0:p><ns0:p>NCR-high subgroups according to the optimal cut-off value (NCR=0.0482). We found that DFS and DSS in the cytoplasmic YAP1-high subgroup were significantly shorter than the cytoplasmic There is some evidence to suggest that YAP1 is retained in the cytoplasm by AKT phosphorylation <ns0:ref type='bibr' target='#b0'>(Basu et al. 2003)</ns0:ref> or through binding LATS1 <ns0:ref type='bibr' target='#b18'>(Matallanas et al. 2007</ns0:ref>), and YAP1 functions as an oncogene which can promote CRC progression by activating the ERK/PI3K-AKT signaling pathway <ns0:ref type='bibr' target='#b29'>(Wang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b38'>Zhang et al. 2016)</ns0:ref>. And LATS1/2 has been reported have a suppress role in cancer immunity <ns0:ref type='bibr' target='#b20'>(Moroishi et al. 2016)</ns0:ref>, and this phenomenon may be a reason why YAP1 cytoplasmic localization is associate with the progression and poor prognosis of CRC. In another way, YAP1 acts as a tumor suppressor gene interacting with p73 to cause transcription of proapoptotic gene puma <ns0:ref type='bibr' target='#b18'>(Matallanas et al. 2007</ns0:ref>), but the apoptosis can be suppressed by enhancing the retention of YAP1 in cytoplasm. This may be the reason why high cytoplasmic YAP1 expression and low YAP1 NCR is associated with the progression and poor prognosis of CRC. Recent research has suggested that upregulation of EGFR by YAP1 has contributed to confer chemoresistance to esophageal cancer cells <ns0:ref type='bibr' target='#b25'>(Song et al. 2015)</ns0:ref> , another study suggested that YAP1 confers Colon cancer cells chemoresistance to 5FU chemotherapy <ns0:ref type='bibr' target='#b27'>(Touil et al. 2014)</ns0:ref>, Therefore, YAP1 may promote CRC progression, as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy. However, the suggestions above are speculative, further mechanistic studies are required to explain the results.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In this study, we provided important evidence that increased cytoplasmic YAP1 correlated with the malignant phenotype in CRC. Importantly, the results show that increased cytoplasmic YAP1 was significantly associated with poor prognosis in CRC patients. More importantly, as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy. Our study has revealed that Cytoplasmic YAP1 could be utilized as prognostic factors in CRC patients and may be indicators of whether a certain patient population could benefit from postoperative chemotherapy, however, the molecular mechanisms behind it remain unknown and need to be further investigated.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure legends</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>from the first treatment to the first relapse. Disease-specific survival (DSS) as the number of months from the first treatment to the date of death due to CRC. The patients were divided into two subgroups (cytoplasmic YAP1 high vs. cytoplasmic YAP1 low, nuclear YAP1 high vs. nuclear YAP1 low, or YAP1 NCR high vs. YAP1 NCR low) by the optimal cut-off values for maximum discrimination in survival difference, the cut-off values were determined by the maxstat R package PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>YAP1-low subgroup, and DFS and DSS were significantly lower in YAP1 NCR-low subgroup than YAP1 NCR-high subgroup. High cytoplasmic YAP1 expression and low YAP1 NCR were found to be independent risk factors for CRC prognosis in multivariate Cox analysis (after correcting confounding variables), above results indicated that cytoplasmic YAP1 may be used as a indicator for staging of tumor. This is the first study to show the potential association between subcellular localization of YAP1 and CRC patients' clinicopathological characteristics.Adjuvant chemotherapy (FOLFOX/CapeOX regimen) is currently the most effective cytotoxic regimen for the treatment of CRC, FOLFOX adjuvant therapy can significant improve the survival of CRC patients<ns0:ref type='bibr' target='#b7'>(Gustavsson et al. 2015)</ns0:ref>. However, adjuvant chemotherapy also has some side effects, such as myelotoxicity, neurotoxicity or gastrointestinal toxicity which can befatal and cause complications (Mohelnikova-Duchonova et al. 2014), therefore, biomarkers predicting the benefit of chemotherapy are urgently needed. Our study clearly demonstrated that as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy, in fact, adjuvant chemotherapy did not contribute to improve the survival of the cytoplasmic-high YAP1 subgroup and the YAP1 NCR-low subgroup patients. Currently, microsatellite instability (MSI) is the only effective indicator for prognosis and suitable chemotherapy regime for colorectal cancer patients (Hemminki et al. 2000; Popat et al. 2005), therefore, a new biomarker is urgently needed to instruct us to determine if a population is suitable for adjuvant chemotherapy. Therefore, cytoplasmic YAP1 may have crucial clinical implications and deserve further study.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Fig. 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig. 1 Differences in YAP1 expression between CRC tissues and adjacent normal tissues. (A-D) Bioinformatics analyses of YAP1 mRNA expression between cancer and cancer related specimens in one TCGA dataset and three GEO datasets. (E) Comparison of YAP1 expression level among different colorectal pathological tissues by cytoplasmic YAP1 H-score, (F) nuclear YAP1 H-score, or (G) YAP1 NCR H-score. (H-M) Representative YAP1 staining in normal tissues and cancer tissues, the blue staining represents the nuclear staining and the brown staining represents the YAP1 positive staining, cancer tissue have the higher cytoplasmic YAP1 H-score, higher nuclear YAP1 H-score and lower NCR than normal tissue, scale bars: 100&#956;m. *P&lt;0.05; **P&lt;0.01; ***P&lt;0.001; ****P&lt;0.0001; ns, no significance.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Fig. 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Fig. 2 High cytoplasmic YAP1 expression is associated with a worse survival in CRC patients. (A-D) Associations between cytoplasmic YAP1 expression and DFS in the patient subgroups with different stage. (E-H) Associations between cytoplasmic YAP1 expression and DSS in the patient subgroups with different stage. Patients with stages &#8544;-&#8546;, stage &#8544;, stage &#8545;, or stage &#8546; were dichotomized into the cytoplasmic YAP1-high subgroups and cytoplasmic YAP1-low subgroups according to optimal cut-off value. Kaplan-Meier survival curves reveal DFS and DSS in patients with each TNM stage CRC. P-values are from Kaplan-Meier analysis with logrank test.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Fig. 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig. 3 Adjuvant chemotherapy has a differential effect on patients with different cytoplasmic YAP1 expression levels. Associations between cytoplasmic YAP1 expression and DFS (A) or DSS (B) in the stage &#8546; patients with chemotherapy. Associations between cytoplasmic YAP1 expression and DFS (C) or DSS (D) in the stage &#8546; patients without chemotherapy. P-values are from Kaplan-Meier analysis with log-rank test.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Fig</ns0:head><ns0:label /><ns0:figDesc>Fig. S1 Supplementary bioinformatics analyses and differences in YAP1 expression between different differentiation grades. (A) Bioinformatics analyses of YAP1 mRNA expression levels in cancer and cancer-related specimens in five GEO datasets. (B) Correlation scatter plot of Cytoplasmic vs Nuclear YAP1 H-score. (C) Comparison of YAP1 expression levels between different differentiation grades by cytoplasmic YAP1 H-score (D) nuclear YAP1 H-score, or (E) YAP1 NCR H-score. *P&lt;0.05; **P&lt;0.01; ***P&lt;0.001; ****P&lt;0.0001; ns, no significance.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Fig</ns0:head><ns0:label /><ns0:figDesc>Fig. S2 Low YAP1 NCR is associated with a worse survival in CRC patients. (A) Associations between nuclear YAP1 expression and DFS or DSS in patients with stages &#8544;-</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Fig</ns0:head><ns0:label /><ns0:figDesc>Fig. S3 Adjuvant chemotherapy has a differential effect on patients with different YAP1 NCR. Associations between YAP1 NCR and DFS (A) or DSS (B) in the stage &#8546; patients with</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,42.52,70.87,525.00,411.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>), likewise, low YAP1 NCR was an independent risk factor ofDFS (HR=2.295; P=0.024) and DSS (HR=2.873; 95%CI=1.045-7.902; P=0.041) for CRC patients (Table2), but the univariate Cox regression analysis showed that nuclear YAP1 expression level was not a meaningful prognostic factor either for DFS PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)Manuscript to be reviewed (HR=0.684; 95%CI=0.453-1.031; P=0.07) or DSS(HR=0.860; P=0.688) </ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head>Table 1 . Associations of cytoplasmic YAP1 expression with demographic and clinical variables of 919 CRC patients</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>** Mann-Whitney U test (non-parametric). Missing values are excluded for all statistic tests.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Characteristics</ns0:cell><ns0:cell>Total (n=919)</ns0:cell><ns0:cell cols='2'>Cytoplasmic YAP1 expression level Low(n=457) High(n=462)</ns0:cell><ns0:cell>P Value *</ns0:cell></ns0:row><ns0:row><ns0:cell>Mean age&#177;SD(year)</ns0:cell><ns0:cell>60.1&#177;12.4</ns0:cell><ns0:cell>61.4&#177;12.3</ns0:cell><ns0:cell>60.7&#177;12.5</ns0:cell><ns0:cell>0.389 **</ns0:cell></ns0:row><ns0:row><ns0:cell>Sex (n (%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.686</ns0:cell></ns0:row><ns0:row><ns0:cell>Men</ns0:cell><ns0:cell>549(59.7)</ns0:cell><ns0:cell>270(59.1)</ns0:cell><ns0:cell>279(60.4)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Women</ns0:cell><ns0:cell>370(40.3)</ns0:cell><ns0:cell>187(40.9)</ns0:cell><ns0:cell>183(39.6)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Disease location (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.632</ns0:cell></ns0:row><ns0:row><ns0:cell>Rectum</ns0:cell><ns0:cell>512(55.7)</ns0:cell><ns0:cell>251(54.9)</ns0:cell><ns0:cell>261(56.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Colon</ns0:cell><ns0:cell>407(44.3)</ns0:cell><ns0:cell>206(45.1)</ns0:cell><ns0:cell>201(43.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Differentiation grade (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>&lt;0.001 ***</ns0:cell></ns0:row><ns0:row><ns0:cell>Well</ns0:cell><ns0:cell>95(10.3)</ns0:cell><ns0:cell>67(14.7)</ns0:cell><ns0:cell>28(6.1)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Moderately</ns0:cell><ns0:cell>779(84.8)</ns0:cell><ns0:cell>369(80.7)</ns0:cell><ns0:cell>410(88.7)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Poorly</ns0:cell><ns0:cell>35(3.8)</ns0:cell><ns0:cell>14(3.1)</ns0:cell><ns0:cell>21(4.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Missing</ns0:cell><ns0:cell>10(1.1)</ns0:cell><ns0:cell>7(1.5)</ns0:cell><ns0:cell>3(0.6)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Resected lymph nodes (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>&lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;12</ns0:cell><ns0:cell>201(21.9)</ns0:cell><ns0:cell>140(30.6)</ns0:cell><ns0:cell>61(13.2)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8805;12</ns0:cell><ns0:cell>718(78.1)</ns0:cell><ns0:cell>317(69.4)</ns0:cell><ns0:cell>401(86.8)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>TNM stage (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.362 ***</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8544;</ns0:cell><ns0:cell>140(15.2)</ns0:cell><ns0:cell>65(14.2)</ns0:cell><ns0:cell>75(16.2)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8545;</ns0:cell><ns0:cell>459(49.9)</ns0:cell><ns0:cell>245(53.6)</ns0:cell><ns0:cell>214(46.3)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8546;</ns0:cell><ns0:cell>320(34.8)</ns0:cell><ns0:cell>147(32.2)</ns0:cell><ns0:cell>173(37.4)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Chemotherapy (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.730</ns0:cell></ns0:row><ns0:row><ns0:cell>Yes</ns0:cell><ns0:cell>671(73.0)</ns0:cell><ns0:cell>336(73.5)</ns0:cell><ns0:cell>335(72.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>No</ns0:cell><ns0:cell>248(27.0)</ns0:cell><ns0:cell>121(26.5)</ns0:cell><ns0:cell>127(27.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serum CEA (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.451</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;5ng/ml</ns0:cell><ns0:cell>568(61.8)</ns0:cell><ns0:cell>288(63)</ns0:cell><ns0:cell>280(60.6)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8805;5ng/ml</ns0:cell><ns0:cell>351(38.2)</ns0:cell><ns0:cell>169(37)</ns0:cell><ns0:cell>182(39.4)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serum CA19-9 (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.686</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;37U/ml</ns0:cell><ns0:cell>788(85.7)</ns0:cell><ns0:cell>394(86.2)</ns0:cell><ns0:cell>394(85.3)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8805;37U/ml</ns0:cell><ns0:cell>131(14.3)</ns0:cell><ns0:cell>63(13.8)</ns0:cell><ns0:cell>68(14.7)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Notes: * &#967;2 test.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='5'>Abbreviations: YAP1, Yes associated protein 1; TNM, tumor-node-metastasis; CEA, carcinoembryonic</ns0:cell></ns0:row><ns0:row><ns0:cell cols='2'>antigen; CA19-9, carbohydrate antigen 19-9.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>** Student t-test. *</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Cox regression analysis of immunohistochemistry YAP1 expression and clinicopathological covariates in patients with CRCAbbreviations: HR, hazard ratio; CI, confidence interval; YAP1, Yes associated protein 1; TNM, tumor-node-metastasis; CEA, carcinoembryonic antigen; CA19-9, carbohydrate antigen 19-9; NCR, Nuclear/Cytoplasmic Ratio.</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 2 . Cox regression analysis of immunohistochemistry YAP1 expression and clinicopathological covariates in patients with 2 CRC Disease-free Survival Disease-specific Survival</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Characteristics</ns0:cell><ns0:cell>Univariate</ns0:cell><ns0:cell /><ns0:cell>Multivariate</ns0:cell><ns0:cell /><ns0:cell>Univariate</ns0:cell><ns0:cell /><ns0:cell>Multivariate</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>P</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>YAP1-high vs. YAP1-low(cytoplasmic) 3.891 (2.758-5.490) &lt;0.001</ns0:cell><ns0:cell cols='2'>3.255 (2.290-4.627) &lt;0.001</ns0:cell><ns0:cell cols='2'>4.291 (2.545-7.236) &lt;0.001</ns0:cell><ns0:cell cols='2'>4.049 (2.400-6.830) &lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>YAP1-low vs. YAP1-high(NCR)</ns0:cell><ns0:cell>2.709(1.331-5.511)</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell>2.295(1.118-4.711)</ns0:cell><ns0:cell>0.024</ns0:cell><ns0:cell>3.346(1.219-9.183)</ns0:cell><ns0:cell>0.019</ns0:cell><ns0:cell>2.873(1.045-7.902)</ns0:cell><ns0:cell>0.041</ns0:cell></ns0:row><ns0:row><ns0:cell>YAP1-high vs. YAP1-low(nuclear)</ns0:cell><ns0:cell>0.684(0.453-1.031)</ns0:cell><ns0:cell>0.070</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.860(0.412-1.975)</ns0:cell><ns0:cell>0.688</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Age (&gt;=60 vs. &lt;60)</ns0:cell><ns0:cell>0.897 (0.667-1.207)</ns0:cell><ns0:cell>0.474</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.891 (0.568-1.398)</ns0:cell><ns0:cell>0.617</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Sex (female vs. male)</ns0:cell><ns0:cell>0.867 (0.638-1.177)</ns0:cell><ns0:cell>0.360</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.817 (0.512-1.302)</ns0:cell><ns0:cell>0.395</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Location (colon vs. rectum)</ns0:cell><ns0:cell>1.068 (0.793-1.440)</ns0:cell><ns0:cell>0.665</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>1.159 (0.737-1.823)</ns0:cell><ns0:cell>0.522</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>TNM (per increase in stage)</ns0:cell><ns0:cell cols='2'>1.874 (1.474-2.381) &lt;0.001</ns0:cell><ns0:cell cols='2'>1.863 (1.471-2.360) &lt;0.001</ns0:cell><ns0:cell>1.256 (0.889-1.775)</ns0:cell><ns0:cell>0.196</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Grade (per increase in grade)</ns0:cell><ns0:cell cols='2'>3.001 (1.948-4.625) &lt;0.001</ns0:cell><ns0:cell cols='2'>3.435 (2.127-5.548) &lt;0.001</ns0:cell><ns0:cell>2.992 (1.575-5.685)</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell>2.732 (1.383-5.394)</ns0:cell><ns0:cell>0.004</ns0:cell></ns0:row><ns0:row><ns0:cell>Chemotherapy (yes vs. no)</ns0:cell><ns0:cell>1.705 (1.156-2.515)</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell>1.029 (0.647-1.637)</ns0:cell><ns0:cell>0.902</ns0:cell><ns0:cell>1.125 (0.662-1.912)</ns0:cell><ns0:cell>0.663</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Resected lymph nodes (&#8805;12 vs. &lt;12)</ns0:cell><ns0:cell cols='2'>2.675 (1.689-4.236) &lt;0.001</ns0:cell><ns0:cell>1.780 (1.111-2.853)</ns0:cell><ns0:cell>0.017</ns0:cell><ns0:cell>2.610 (1.375-4.954)</ns0:cell><ns0:cell>0.003</ns0:cell><ns0:cell>1.685 (0.874-3.251)</ns0:cell><ns0:cell>0.120</ns0:cell></ns0:row><ns0:row><ns0:cell>Serum CEA (&#8805;5 vs. &lt;5ng/ml)</ns0:cell><ns0:cell>1.646 (1.223-2.215)</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell>1.513 (1.120-2.043)</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell>1.378 (0.876-2.170)</ns0:cell><ns0:cell>0.166</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serum CA19-9 (&#8805;37 vs. &lt;37U/ml)</ns0:cell><ns0:cell>1.766 (1.224-2.549)</ns0:cell><ns0:cell>0.002</ns0:cell><ns0:cell>1.350 (0.914-1.995)</ns0:cell><ns0:cell>0.132</ns0:cell><ns0:cell>1.619 (0.906-2.894)</ns0:cell><ns0:cell>0.104</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='3'>Abbreviations: HR, hazard ratio; CI, confidence interval; YAP1, Yes associated protein 1; TNM, tumor-node-metastasis; CEA, carcinoembryonic antigen; 4 CA19-9, carbohydrate antigen 19-9; NCR, Nuclear/Cytoplasmic Ratio.PeerJ reviewing PDF | (2020:05:49163:1:1:NEW 6 Sep 2020)</ns0:note> </ns0:body> "

"Response letter Dear editor, On behalf of my myself and the co-authors, thank you for the constructive feedback received on 07,10, 2020, and for providing us with an opportunity to revise our manuscript (manuscript ID: 49163). We appreciate the editors’ and reviewers’ positive and constructive comments and suggestions on our manuscript entitled “High cytoplasmic YAP1 expression predicts a poor prognosis in patients with colorectal cancer”. We have carefully reviewed the reviewers’ comments and we have made the relevant revisions, which have been highlighted in the revised manuscript. We hope that the revised manuscript meets all the requirements to be accepted for publication in PeerJ. Responses to reviewers’ comments: Reviewer 1 Basic reporting The manuscript broadly meets all the criteria and expected standards. However, I recommend authors to provide a English-translated version of the written consent form (currently in Chinese), as this has been presented as a Supplemental evidence. Response: As suggested, we have added the English-translated version of the written consent form to the “Supplemental material” section in the revised manuscript. Experimental design No Comments Response: Thank you. Validity of the findings I have a few comments on the results presented and conclusions drawn by the authors. 1. There is an instance of minor image duplication in Figure 1E. In the upper panel comprising of Normal (Upper) Cytoplasmic (Left) and Nuclear YAP1 (middle), authors use slightly shifted field-of-views of the same image to extract the inset crop. If possible, provide an image of another field for the Nuclear YAP1 data. If you intend to use the same image, please provide the exact same field of view for both data set and mention the same in the figure legend. Response: As suggested, an image of another field for the Nuclear YAP1 data is provided in the revised manuscript. 2. Authors report overall similar trends in their results when scoring for either ‘High cytoplasmic YAP1’ or ‘Low YAP1 NCR’. While these two are intuitively linked, the ratio could be biased by differences in Nuclear Levels of YAP1. Authors have provided independent quantification of the same in Fig 1 B-D. I feel this result could be bolstered by adding a correlation scatter plot of Cytoplasmic v/s Nuclear YAP1 H-score, to rule there are any non-trivial relations between them, which might bias their results. Response: As suggested, we have added a correlation scatter plot of Cytoplasmic v/s Nuclear YAP1 H-score (Fig. S1B) to the ‘Differences in YAP1 expression between CRC tissues and adjacent normal tissues’ subsection of the results in the revised manuscript. Comments for the author The differential effects of chemotherapy on high cytosolic YAP1 cells constitute the key finding of this manuscript. Please highlight this in the abstract more clearly. Response: Thank you, we have carefully reviewed and revised the abstract to highlight the finding as follows ‘In particular, the Kaplan-Meier analysis showed that high cytoplasmic YAP1 expression is associated with potential harm from adjuvant chemotherapy in stage III CRC patients.’ Reviewer 2 Basic reporting 1. Clear legend should be provided for Figure 1E. It is not clear what the colors represents, and it is not clear what are the differences between the images presented and what is the message each image is delivering to the readers. Response: Thank you, we have revised the fig.1E legend as suggested. 2. Comparing the H-scores in Figure 1E with that in Figure 1B-D, it seems that images presented in Figure 1E are examples of extreme H-scores. Those images are not representative staining. Response: We have changed the images as suggested. 3. In Line 62 of the Introduction, the authors mentioned the need of new biomarkers for CRC prognosis. I think the authors should introduce if there are other biomarkers for CRC prognosis and discuss about how YAP1 is different from or better than existing biomarkers. Response: Thank you for your suggestion. We have added the relevant information to the “Introduction” section as follows: ‘Currently, the only effective marker for the CRC prognosis and appropriate chemotherapy selection is microsatellite instability (MSI) [8, 9], however, MSI as a CRC marker has not been applied clinically. Therefore, new biomarkers to assess the prognosis of CRC patients before and after treatment are urgently needed.’ 4. In Lines 66-74. The authors first introduced the paradoxical reports of YAP1 either as an oncogene or a tumor suppressor gene, then proposed the need of establishing a clear relationship between YAP1 expression and its clinical relevance. I think the authors should discuss why they think the relationship they have established is clearer than those in previous reports instead of simply one more example being added to the existing paradoxical reports. Response: Thank you for your suggestion. In view of the contradictions of existing reports, we firstly showed the potential associations between subcellular localization of YAP1 and CRC patients’ clinicopathological characteristics, utilizing the large cohort (919 CRC patients with prognostic information) and detailed clinicopathological characteristics. We tried to elucidate the relationship between YAP1 expression pattern and CRC prognosis more clearly. 5. Raw data. A lot of information, including TNM stage, differentiation grade, manual grading of H-score before averaging, etc., are not provided in the raw data used for Figures 2-3 and Tables 1-2. Raw data for Figure 1B-E are not provided. Response: Thanks for your suggestions. Manual grading of H-scores before averaging and the whole field of the representative IHC staining figures have added to the raw data. Due to the patients’ raw clinicopathological characteristics involve the original confidential information of the patients, and the data is complex and irregular, therefore, it is reasonable to show the existing raw data about patients’ clinicopathological characteristics. 6. The font size in Figure 2 is too small. Response：We have revised it as suggested. Experimental design 1. In Lines 88-95 bioinformatics analysis, the authors used publicly available gene datasets. The authors should describe how and why they select those specific datasets for their analysis. Are those all the related datasets available from the database, or are they randomly selected from the database? In Line 90, the authors should provide more details on how the data are normalized. Response: Thank you for your suggestion. We have changed the relevant expression to the “Bioinformatics analysis” section as follows: All available related mRNA expression profiles of YAP1 in CRC tissues were downloaded from the TCGA database and normalized by EDASeq package which can take gene length and GC-content into account. The microarray expression profiles of eight datasets (GSE8671, GSE37364, GSE41258, GSE23878, GSE22598, GSE9348, GSE81582, and GSE77955) associated with CRC tissues were randomly downloaded from the Gene Expression Omnibus (GEO) database, and then the mRNA expression profiles of YAP1 in CRC tissues with different pathological features were extracted from the microarray expression profiles and compared by independent sample t-tests or paired sample t-tests. Validity of the findings 1. In Lines 193-194, the authors concluded that high cytoplasmic YAP1 expression may contribute to CRC aggressiveness. This conclusion is not justified as the data only suggest for a correlation but not for a causation. It is equally likely that YAP1 expression is modified as CRC progresses to a more advanced stage, i.e. CRC progression contributes to cytoplasmic YAP1 expression. Response: We have revised the expression from “The above results revealed that high cytoplasmic YAP1 expression may contribute to the aggressiveness of CRC” to “The above results revealed that high cytoplasmic YAP1 expression may be involved in the aggressiveness of CRC”. 2. The conclusion (Lines 215-216, Lines 228-229, and Lines 264-265) generated from Figure 3 is not justified. Firstly, the sample size for the groups without chemotherapy is too small, it is premature to draw any conclusions based on the current data (i.e. Fig. 3C-D, Fig. S3C-D). Secondly and most importantly, to evaluate whether chemotherapy has an effect on cytoplasmic YAP1-low group or cytoplasmic YAP1-high group, and whether the effect of chemotherapy is different due to the expression level of cytoplasmic YAP1 level, the authors should compare the group with chemotherapy (either high cytoplasmic YAP1 or low cytoplasmic YAP1) to the corresponding group without chemotherapy first, and then check, for example, if chemotherapy is effective for the cytoplasmic YAP1-low patients but not effective for the cytoplasmic YAP1-high patients. The current comparison in Figure 3 only answers the question whether chemotherapy influences the prognostic role of cytoplasmic YAP1 level, but not the question whether cytoplasmic YAP1 level influences the chemotherapy effect. Response: Thank you for your constructive suggestion. We have compared the group with chemotherapy (either high cytoplasmic YAP1 or low cytoplasmic YAP1) to the corresponding group without chemotherapy, but there are no significant difference in survival between two groups, so we have changed subtitle from ‘Adjuvant chemotherapy has a differential effect on patients with different cytoplasmic YAP1 expression levels.’ to ‘Stage III CRC patients with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy.’ And changed the expression from ‘Therefore, our results indicated that adjuvant chemotherapy differentially affected the survival of patients (stage III CRC) with different cytoplasmic YAP1 expression levels.’ to ‘Therefore, as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy.’ And the relevant expressions in other parts of the manuscript have also been changed accordingly. Comments for the author In this manuscript, Dong and colleagues investigated the relationship between YAP1 expression and colorectal cancer (CRC) prognosis. They found a correlation between CRC progression with cytoplasmic but not nuclear YAP1 expression, with high cytoplasmic YAP1 corresponding to poor prognosis. This is an interesting finding as YAP is a transcriptional co-activator which regulates gene expression when transported to the nucleus. However, some of the conclusions are not justified. Please see specific comments for details. Response: Thank you! Reviewer 3 Basic reporting The authors have investigated the subcellular loacalization of YAP-1 protein in CRC samples and examined its potential correlation to CRC prognosis and Disease specific/free survival. They calculated YAP-1 staining intensity h scores and studies TMA. The language is simple, clear and non ambiguous. Authors have provided proper references. A little descriptive background about why YAP-1 plays a role in CRC, its dual role in CRC as an oncogene and a tumor suppressor, and its association to hippo pathway might be insightful. Response: Thank you! Experimental design The overall observational analysis suggesting that higher nuclear and cytoplasmic YAP-1 expression is associated with aggravated CRC. 1. The authors suggest that YAP1 mRNA expression was elevated in 3 GEO and 1 TCGA dataset upon analysis following which they mention that other 5 GEO datasets also exhibit similar results. Is there a significance to segregating the first 3 and other 5 GEO datasets while analysis and result interpretation? Response: Thank you for your suggestion. Due to the limited space in the manuscript, we exhibit the other five GEO datasets which also show the same results in the supplementary figures. 2. Suppression of YAP-1 phosphorylation leads to its higher nuclear accumulation and has been previously reported to be correlated to CRC prognosis. However, the authors have shown that CRC samples had both high nuclear and cytoplasmic YAP-1 expression. What is the significance of higher cytoplasmic YAP-1? YAP-1 regulates and drives multiple pathways and it would add to the study if authors look into the pathways that drive its high cytoplasmic expression in CRC. Response: Thank you for your suggestion. As you said, YAP-1 regulates and regulated by multiple pathways, as we mentioned in the discussion part ‘YAP1 has also been reported to act as a tumor suppressor gene that interacts with p73 to cause transcription of the proapoptotic gene puma [17], but apoptosis can be suppressed by enhancing the retention of YAP1 in cytoplasm, which may be why high cytoplasmic YAP1 expression and a low YAP1 NCR are associated with the progression and poor prognosis of CRC.’ And LATS1/2 has been reported have a suppress role in cancer immunity (Moroishi et al., 2016, Cell), and this phenomenon may be a reason why YAP1 cytoplasmic localization is associate with the progression and poor prognosis of CRC. We have added the sentence above to the discussion part. 3. Authors observe that subjects that underwent adjuvant chemotherapy and had high cytoplasmic YAP-1 had a shorter DFS and DSS. Authors can extrapolate this observation to in vitro studies to understand the actual mechanism of how cytoplasmic YAP-1 expression can determine the efficacy of adjuvant chemotherapy. If authors are able to verify this observation through mechanistic evaluation, it is the most important and interesting finding of this study. Response: Thank you for your constructive suggestion. However, the actual mechanism of how cytoplasmic YAP-1 expression can determine the efficacy of adjuvant chemotherapy is very unfathomable, and the mechanism of this action has been under investigation in our laboratory, due to the impact of the COVID-19, unfortunately, results are unavailable at this point. We think that the phenomenon was be found in the current study is very meaningful, and we are eager to share it with peers in order to promote mechanistic evaluation. 4. Did authors see any age dependent or gender dependent cytoplasmic/ nuclear YAP-1 variability? Response: As shown in Table 1, there is no age dependent or gender dependent cytoplasmic YAP-1 variability, and we did not see any age dependent or gender dependent nuclear YAP-1 variability too. 5. In the discussion section, authors have reported previous literature on how the upregulation of EGFR by YAP1 confers chemoresistance to esophageal cancer cells. It will be interesting to see if their observation on effect of high cytoplasmic YAP-1 on adjuvant chemotherapy is also due to EGF-R upregulation or other factors. Response: Thank you for your constructive suggestion. We understand that the experiment (the reviewer suggested) may better reveal the relationship between YAP-1 and chemotherapy in CRC. However, in the present study, we mainly focused on the prognosis value of YAP-1 in CRC, which shed light on the role of YAP-1 in CRC chemotherapy. We think that, while the experiments in this study may not be optimal, they should be sufficient to draw the conclusion that high cytoplasmic YAP1 expression is associated with potential harm from adjuvant chemotherapy in stage III CRC patients. 6. If I understand correctly, the cytoplasmic YAP-1 expression was high in all malignant CRC irrespective of stage. What is the prognostic value of YAP-1 in differentiating the stage based on its expression? Response: Thank you for your constructive suggestion. After analyzed the association between cytoplasmic YAP-1 H-score and TNM stage of CRC patients, we did not see any significant difference between different stages respect to cytoplasmic expression level. The results indicated that the cytoplasmic YAP-1 expression was elevated in all malignant CRC stages, therefore, CRC staging based on YAP-1 expression need further investigation. Validity of the findings The study is interesting in many facets owing to its clinical significance however it is mostly observational and bioinformatic. To assess and confirm these observations, elucidating the underlying cause of YAP-1 cytoplasmic up-regulation in CRC will be highly useful. The English needs minor revision and language editing. Overall the paper is easy to understand and direct. Fine tuning the discussion can help interpret the finding of the study more significantly. Authors can add the most recent work on correlation between YAP-1 and CRC. Response: Thank you! We have carefully reviewed and revised the manuscript to reduce English errors. We have revised and added the relevant information to discussion section in revised manuscript as follows: ‘and LATS1/2 has been reported have a suppress role in cancer immunity [35], and this phenomenon may be a reason why YAP1 cytoplasmic localization is associate with the progression and poor prognosis of CRC.’ Thank you for your constructive suggestions to help improve the quality and scientific integrity of our manuscript. Sincerely, Shangyong Zheng "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Yes associated protein 1 (YAP1), which is a standout amongst the most essential effectors of the Hippo pathway, assumes a vital part in a few kinds of cancer. However, whether YAP1 is an oncogene in CRC (colorectal cancer) remains controversial, and the association between the subcellular localization of YAP1 and clinical implications in CRC remains unknown. Patients and methods: In this study, we investigated the subcellular localization of YAP1 in CRC cells by immunohistochemistry and then associate these findings with clinical information in a large CRC cohort with 922 CRC patients. Results:</ns0:p><ns0:p>The results show that CRC tissues has a significant higher expression of cytoplasmic YAP1 compared to adjacent normal tissues (all P&lt;0.001). Cytoplasmic YAP1 expression was significantly associated with the number of lymph nodes removed and differentiation grade (all P &lt; 0.001). Furthermore, after correcting confounding variables, for example, TNM stage and differentiation grade, the multivariate Cox analysis confirmed cytoplasmic YAP1-high subgroup had a significant shorter DFS(HR=3.255; 95% CI=2.290-4.627; P&lt;0.001) and DSS(HR=4.049; 95% CI=2.400-6.830; P&lt;0.001) than cytoplasmic YAP1-low subgroup. High cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy. Conclusion: Cytoplasmic YAP1 could be could be utilized as a prognosis factor in CRC patients, and may be an indicator of whether certain patients population could benefit from postoperative chemotherapy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>The prevalence of CRC ranks third among all cancers in male and second in female <ns0:ref type='bibr' target='#b25'>(Torre et al. 2015)</ns0:ref>, and CRC has the third highest mortality rate among all cancers <ns0:ref type='bibr' target='#b36'>(Zeng et al. 2014)</ns0:ref>.</ns0:p><ns0:p>Surgical resection combined with chemotherapy remains the mainstay of treatment for CRC, in any case, numerous patients will progress to metastatic CRC and develop resistance to chemotherapeutic drugs <ns0:ref type='bibr' target='#b5'>(Fisher et al. 2015)</ns0:ref>, because signs or symptoms diagnose CRC usually appear in advanced phases <ns0:ref type='bibr' target='#b1'>(Binefa et al. 2014)</ns0:ref>. Even if some patients are diagnosed with CRC and undergo surgery at an early stage, 20%-30% of these patients will relapse within five years. <ns0:ref type='bibr' target='#b8'>(Hardingham et al. 2015)</ns0:ref>. Current CRC treatment regimen is heterogenous for patients, even for patients with the same TNM stage <ns0:ref type='bibr' target='#b21'>(Nagtegaal et al. 2012)</ns0:ref>, in any case, the indication for treatment should be assessed on an individual basis by considering the risk factors of relapse <ns0:ref type='bibr' target='#b17'>(Marin et al. 2012)</ns0:ref>. Currently, the only effective marker for the CRC prognosis and appropriate chemotherapy selection is microsatellite instability (MSI) <ns0:ref type='bibr' target='#b9'>(Hemminki et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b23'>Popat et al. 2005)</ns0:ref>, however, MSI as a CRC marker has not been applied clinically. Therefore, there is an urgent need for new biomarkers to assess the prognosis of CRC patients before and after treatment.</ns0:p><ns0:p>YAP1 is a standout amongst the most essential effectors of the Hippo pathway, which is a critical pathway regulating cell proliferation, apoptosis, and organ growth <ns0:ref type='bibr' target='#b10'>(Justice et al. 1995)</ns0:ref>.</ns0:p><ns0:p>Several studies have shown that YAP1 is an oncogene highly express in numerous cancer types including bladder cancer <ns0:ref type='bibr' target='#b15'>(Liu et al. 2013)</ns0:ref>, breast cancer <ns0:ref type='bibr' target='#b13'>(Kim et al. 2014)</ns0:ref>, gastric cancer <ns0:ref type='bibr'>(Kang et al. 2011)</ns0:ref>, hepatocellular cancer <ns0:ref type='bibr' target='#b32'>(Xu et al. 2009</ns0:ref>), nonsmall-cell lung cancer <ns0:ref type='bibr' target='#b30'>(Wang et al. 2010)</ns0:ref>, and CRC <ns0:ref type='bibr' target='#b29'>(Wang et al. 2013a;</ns0:ref><ns0:ref type='bibr' target='#b32'>Xu et al. 2009</ns0:ref>) that associate with tumor progression and poor prognosis. On the contrary, abundant literature suggested that YAP1 is a tumor suppressor gene and nuclear expression is reduced in different cancers, such as breast cancer <ns0:ref type='bibr' target='#b18'>(Matallanas et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b34'>Yu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b35'>Yuan et al. 2008)</ns0:ref>, head and neck cancers <ns0:ref type='bibr' target='#b4'>(Ehsanian et al. 2010)</ns0:ref>, hematological cancers <ns0:ref type='bibr' target='#b2'>(Cottini et al. 2014)</ns0:ref>, and CRC <ns0:ref type='bibr' target='#b14'>(Levy et al. 2007</ns0:ref>). These paradoxical reports remind us that the role of YAP1 in cancer is controversial, and it is crucial to make it clear the relationship between YAP1 expression and its clinical relevance in CRC. In addition, the nuclear overexpression of YAP1 is associate with poor survival in gastric cancer <ns0:ref type='bibr'>(Kang et al. 2011)</ns0:ref>, actually, previous researches suggest that subcellular localization of proteins is associated with functions associated of tumorigenesis and tumor progression <ns0:ref type='bibr' target='#b6'>(Garcia et al. 2005;</ns0:ref><ns0:ref type='bibr' target='#b16'>Lobo et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b27'>Vaquero et al. 2017)</ns0:ref>, a few studies suggested that YAP1 overexpression is associate with poor survival in CRC <ns0:ref type='bibr' target='#b29'>(Wang et al. 2013a;</ns0:ref><ns0:ref type='bibr' target='#b31'>Wang et al. 2013b;</ns0:ref><ns0:ref type='bibr' target='#b33'>Yang et al. 2018)</ns0:ref>, however, the association between subcellular localization of YAP1 and clinical significance in CRC has been patients signed a written informed consent for using their tissues for research purpose.</ns0:p></ns0:div> <ns0:div><ns0:head>Immunohistochemistry (IHC)</ns0:head><ns0:p>IHC is performed on 4 &#956;m thick array slides. Specifically, the array slides were primarily immersed into the citrate solution (pH 6.0) and boil for 5 minutes for antigen retrieval, then </ns0:p></ns0:div> <ns0:div><ns0:head>Quantitative evaluation of immunostaining</ns0:head><ns0:p>Aperio ScanScope (Aperio Technologies, Vista, CA, USA) was used to digitally scan the stained TMA slides, then the scan image can be used for Quantitative evaluation of immunostaining, the YAP1 protein expression level was quantified by H-score method as reported previously <ns0:ref type='bibr' target='#b3'>(Detre et al. 1995)</ns0:ref>. Specifically, the staining intensity in the epithelial cell was scored as 0, 1, 2, or 3 corresponding to the presence of negative, weak, intermediate, and strong brown staining, respectively, then the number of cells stained at each intensity was counted. The H-score is the multiplication of the proportion of positive cells and the corresponding staining intensity score (0, 1, 2 or 3), thus an H-score between 0 and 300 was obtained. The quantitative evaluation of immunostaining was performed separately by two coauthors who were blinded to the clinicopathological information, and the scores were averaged.</ns0:p></ns0:div> <ns0:div><ns0:head>Follow-up and patients</ns0:head><ns0:p>919 CRC patients' follow-up information was collected using a standard methods reported previously <ns0:ref type='bibr' target='#b22'>(Pan et al. 2015)</ns0:ref>. Disease-free survival (DFS) was defined as the number of months Manuscript to be reviewed in R 3.2.0. </ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Differences in YAP1 expression between CRC tissues and adjacent normal tissues.</ns0:head><ns0:p>To analyze the expression pattern of YAP1 in CRC tissues, we firstly utilized the datasets from public database, the results showed that in one TCGA dataset and three GEO datasets, YAP1 mRNA expression level was consistently significantly elevated in CRC tissues compared with the adjacent normal tissue (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_8'>1A-D</ns0:ref>), the other five GEO datasets also show the same results (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_8'>S1A</ns0:ref>). We subsequently investigated the expression pattern of YAP1 by IHC method in 997 CRC and 70 adjacent normal tissue samples which derive from patients who underwent surgery at Yunnan Cancer Hospital. The positive immunostaining results from YAP1 predominantly occurred in the cytoplasm and nucleus of colorectal epithelial cells (Fig. <ns0:ref type='figure' target='#fig_8'>1H-M</ns0:ref>), whereas the staining was negative or weak in mesenchymal cells (Fig. <ns0:ref type='figure' target='#fig_8'>1H-M</ns0:ref>). We calculated the H-score of cytoplasmic YAP1 and nuclear YAP1 independently, then the YAP1 NCR (nuclear/cytoplasmic ratio) was calculated, and there is a very weak positive correlation between cytoplasmic H-score and nuclear YAP1 H-score (Fig. <ns0:ref type='figure' target='#fig_8'>S1B</ns0:ref>). the results show that cytoplasmic YAP1 expression was significantly elevated in CRC tissues compared with the adjacent normal tissues (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_8'>1E</ns0:ref>), and nuclear YAP1 expression was significantly elevated in primary cancer tissues compared with the adjacent normal tissues (P&lt;0.0001; Fig. <ns0:ref type='figure' target='#fig_8'>1F</ns0:ref>), but the expression of nuclear YAP1 in adenomas and metastasis CRC tissues have no significant differences with the adjacent normal tissues (Fig. <ns0:ref type='figure' target='#fig_8'>1F</ns0:ref>), we also found YAP1 NCR (Nuclear/Cytoplasmic Ratio) gradually decrease in adjacent normal tissues, adenomas, primary cancers, and metastatic CRC (all P&lt;0.05; Fig. <ns0:ref type='figure' target='#fig_8'>1G</ns0:ref>). The results above indicated that the increased cytoplasmic YAP1 expression may be associated with the progression of CRC.</ns0:p></ns0:div> <ns0:div><ns0:head>Associations between YAP1 expression and CRC patients' clinicopathological characteristics</ns0:head><ns0:p>To obtain further information, we analyzed the association between cytoplasmic YAP1 expression levels or YAP1 NCR and CRC patients' clinicopathological characteristics. We found that the expression of cytoplasmic YAP1 protein was significantly higher in poorly+moderate grades than that in the well grade (P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_8'>S1C</ns0:ref>), and YAP1 NCR was significantly lower in poorly+moderate grades than that in the well grade (P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_8'>S1E</ns0:ref>), but there is no significant differences between poorly+moderate grades and well grade in the expression of nuclear YAP1 protein (Fig. <ns0:ref type='figure' target='#fig_8'>S1D</ns0:ref>), the clinicopathological features for the patients at poor+moderate grade or well grade were described in Table <ns0:ref type='table' target='#tab_3'>S1</ns0:ref>. Next, we classified the 919 patients (patients lost follow-up information were excluded) into cytoplasmic YAP1-low and cytoplasmic YAP1-high subgroups by the optimal cut-off value (H-score= 202.5) determined by the maxstat R package, the results showed there were significant differences between cytoplasmic YAP1-low and cytoplasmic YAP1-high subgroups with respect to the number of resected lymph nodes and differentiation grade. (all P &lt; 0.001; Table <ns0:ref type='table' target='#tab_3'>1</ns0:ref>). We also classified the 919 patients into YAP1 NCR-low and YAP1 NCR-high subgroups by the optimal cut-off value (NCR=0.0482) determined by the maxstat R package, the results showed a significant difference between the YAP1 NCR-low and YAP1 NCR-high subgroups in the TNM stage (P = 0.02; Table <ns0:ref type='table' target='#tab_5'>S2</ns0:ref>). The above results revealed that high cytoplasmic YAP1 expression may be involved in the aggressiveness of CRC. for CRC patients (Table <ns0:ref type='table' target='#tab_4'>2</ns0:ref>). Kaplan-Meier analyses with log-rank tests showed that DFS and DSS in the cytoplasmic YAP1-high subgroup were significantly shorter than the cytoplasmic YAP1-low subgroup (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_9'>2A, B</ns0:ref>), moreover, cytoplasmic YAP1-high subgroups were consistently had shorter DFS and DSS than cytoplasmic YAP1-low subgroups in stage &#8544;, &#8545;, or &#8546; CRC patients respectively (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_9'>2B-D, F-H</ns0:ref>). We also found DFS and DSS were significantly lower in YAP1 NCR-low subgroup than YAP1 NCR-high subgroup (all P&lt;0.01; Fig. <ns0:ref type='figure' target='#fig_9'>S2A</ns0:ref>). However, there is no significant differences between nuclear YAP1-high subgroup and nuclear YAP1-low subgroup in Kaplan-Meier analyses (all P&gt;0.05; Fig. <ns0:ref type='figure' target='#fig_9'>S2B</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>High cytoplasmic YAP1 expression is associated with a worse survival in CRC patients</ns0:head></ns0:div> <ns0:div><ns0:head>High cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy</ns0:head><ns0:p>To evaluate whether cytoplasmic YAP1 expression level could be an indicator of whether certain patients population could benefit from adjuvant chemotherapy, the stage III patients were divided into two groups respectively (all stage III patients received adjuvant chemotherapy), either did or did not receive adjuvant chemotherapy (Table <ns0:ref type='table'>S3</ns0:ref>), for stage III patients who received adjuvant chemotherapy, Kaplan-Meier analyses with log-rank tests showed that DFS and DSS in the cytoplasmic YAP1-high subgroup were significantly shorter than the cytoplasmic YAP1-low subgroup (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_10'>3A-B</ns0:ref>), but there were no significant differences between YAP1-high subgroup and YAP1-low subgroup in DFS and DSS for stage III patients without adjuvant chemotherapy (all P&gt;0.05; Fig. <ns0:ref type='figure' target='#fig_10'>3C-D</ns0:ref>) . Besides, for stage III patients who received adjuvant chemotherapy, Kaplan-Meier analyses also showed that DFS and DSS in the low YAP1 NCR subgroup were significantly shorter than the high YAP1 NCR subgroup (all P&lt;0.001; Fig. <ns0:ref type='figure' target='#fig_10'>S3A-B</ns0:ref>), but there were no significant differences between high YAP1 NCR subgroup and low YAP1 NCR subgroup in DFS and DSS for stage III patients without adjuvant chemotherapy (all P&gt;0.05; Fig. <ns0:ref type='figure' target='#fig_10'>S3C-D</ns0:ref>). Therefore, high cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Primarily, the results in this study showed that in one TCGA dataset and eight GEO datasets, the mRNA expression of YAP1 in CRC tissues was consistently higher in CRC tissues compared with the adjacent normal tissue. Further, the IHC examination of YAP1 confirmed that epithelial cytoplasmic YAP1 protein expression were significantly elevated in CRC tissues compared with the adjacent normal tissue in the Yunnan Cancer Hospital, and YAP1 NCR gradually decrease in adjacent normal tissues, adenomas, primary cancers, and metastatic CRC.</ns0:p><ns0:p>Prior studies had illustrated the expression of YAP1 in a various cancers including CRC <ns0:ref type='bibr' target='#b2'>(Cottini et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b4'>Ehsanian et al. 2010;</ns0:ref><ns0:ref type='bibr'>Kang et al. 2011;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kim et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b14'>Levy et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b15'>Liu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b18'>Matallanas et al. 2007;</ns0:ref><ns0:ref type='bibr' target='#b29'>Wang et al. 2013a;</ns0:ref><ns0:ref type='bibr' target='#b30'>Wang et al. 2010;</ns0:ref><ns0:ref type='bibr' target='#b32'>Xu et al. 2009;</ns0:ref><ns0:ref type='bibr' target='#b34'>Yu et al. 2013;</ns0:ref><ns0:ref type='bibr' target='#b35'>Yuan et al. 2008)</ns0:ref>, but the association between subcellular localization of YAP1 and aggressiveness of CRC has been neglected. In this study, the expression pattern of YAP1 in the Yunnan Cancer Hospital cohort reveal that the increased cytoplasmic YAP1 expression may be associated with the progression of CRC. The analysis of association between YAP1 expression and CRC patients' clinicopathological features showed that cytoplasmic YAP1 expression was related to differentiation grade and YAP1 NCR was related to TNM stage. Further, CRC patients were divided into cytoplasmic-high YAP1 and cytoplasmic-low YAP1 subgroups by the optimal cutoff value (H-score=202.5), meanwhile, classify CRC patients into YAP1 NCR-low and YAP1</ns0:p><ns0:p>NCR-high subgroups according to the optimal cut-off value (NCR=0.0482). We found that DFS and DSS in the cytoplasmic YAP1-high subgroup were significantly shorter than the cytoplasmic Manuscript to be reviewed</ns0:p><ns0:p>There is some evidence to suggest that YAP1 is retained in the cytoplasm by AKT phosphorylation <ns0:ref type='bibr' target='#b0'>(Basu et al. 2003)</ns0:ref> or through binding LATS1 <ns0:ref type='bibr' target='#b18'>(Matallanas et al. 2007</ns0:ref>), and YAP1 functions as an oncogene which can promote CRC progression by activating the ERK/PI3K-AKT signaling pathway <ns0:ref type='bibr' target='#b28'>(Wang et al. 2017;</ns0:ref><ns0:ref type='bibr' target='#b37'>Zhang et al. 2016)</ns0:ref>. And LATS1/2 has been reported have a suppress role in cancer immunity <ns0:ref type='bibr' target='#b20'>(Moroishi et al. 2016)</ns0:ref>, and this phenomenon may be a reason why YAP1 cytoplasmic localization is associate with the progression and poor prognosis of CRC. In another way, YAP1 acts as a tumor suppressor gene interacting with p73 to cause transcription of proapoptotic gene puma <ns0:ref type='bibr' target='#b18'>(Matallanas et al. 2007</ns0:ref>), but the apoptosis can be suppressed by enhancing the retention of YAP1 in cytoplasm. This may be the reason why high cytoplasmic YAP1 expression and low YAP1 NCR is associated with the progression and poor prognosis of CRC. Recent research has suggested that upregulation of EGFR by YAP1 has contributed to confer chemoresistance to esophageal cancer cells <ns0:ref type='bibr' target='#b24'>(Song et al. 2015)</ns0:ref> , another study suggested that YAP1 confers Colon cancer cells chemoresistance to 5FU chemotherapy <ns0:ref type='bibr' target='#b26'>(Touil et al. 2014)</ns0:ref>, Therefore, YAP1 may promote CRC progression, high cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy. However, the suggestions above are speculative, further mechanistic studies are required to explain the results.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>In this study, we provided important evidence that increased cytoplasmic YAP1 correlated with the malignant phenotype in CRC. Importantly, the results show that increased cytoplasmic YAP1 was significantly associated with poor prognosis in CRC patients. More importantly, high cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy. Our study has revealed that Cytoplasmic YAP1 could be utilized as prognostic factors in CRC patients and may be indicators of whether a certain patient population could benefit from postoperative chemotherapy, however, the molecular mechanisms behind it remain unknown and need to be further investigated.</ns0:p><ns0:p>promotes the malignant potential of colon cancer cells by activating the YAP-ERK/PI3K-AKT signaling pathway. Oncol Rep 36:3619-3626. Manuscript to be reviewed Manuscript to be reviewed </ns0:p><ns0:note type='other'>Figure legends</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>from the first treatment to the first relapse. Disease-specific survival (DSS) as the number of months from the first treatment to the date of death due to CRC. The patients were divided into two subgroups (cytoplasmic YAP1 high vs. cytoplasmic YAP1 low, nuclear YAP1 high vs. nuclear YAP1 low, or YAP1 NCR high vs. YAP1 NCR low) by the optimal cut-off values for maximum discrimination in survival difference, the cut-off values were determined by the maxstat R package PeerJ reviewing PDF | (2020:05:49163:2:0:NEW 25 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>YAP1-low subgroup, and DFS and DSS were significantly lower in YAP1 NCR-low subgroup than YAP1 NCR-high subgroup. High cytoplasmic YAP1 expression and low YAP1 NCR were found to be independent risk factors for CRC prognosis in multivariate Cox analysis (after correcting confounding variables), above results indicated that cytoplasmic YAP1 may be used as a indicator for staging of tumor. This is the first study to show the potential association between subcellular localization of YAP1 and CRC patients' clinicopathological characteristics.Adjuvant chemotherapy (FOLFOX/CapeOX regimen) is currently the most effective cytotoxic regimen for the treatment of CRC, FOLFOX adjuvant therapy can significant improve the survival of CRC patients<ns0:ref type='bibr' target='#b7'>(Gustavsson et al. 2015)</ns0:ref>. However, adjuvant chemotherapy also has some side effects, such as myelotoxicity, neurotoxicity or gastrointestinal toxicity which can be fatal and cause complications (Mohelnikova-Duchonova et al. 2014), therefore, biomarkers predicting the benefit of chemotherapy are urgently needed. Our study clearly demonstrated that high cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy. Currently, microsatellite instability (MSI) is the only effective indicator for prognosis and suitable chemotherapy regime for colorectal cancer patients<ns0:ref type='bibr' target='#b9'>(Hemminki et al. 2000;</ns0:ref><ns0:ref type='bibr' target='#b23'>Popat et al. 2005)</ns0:ref>, therefore, a new biomarker is urgently needed to instruct us to determine if a population is suitable for adjuvant chemotherapy. Therefore, cytoplasmic YAP1 may have crucial clinical implications and deserve further study.PeerJ reviewing PDF | (2020:05:49163:2:0:NEW 25 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Fig. 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Fig. 1 Differences in YAP1 expression between CRC tissues and adjacent normal tissues. (A-D) Bioinformatics analyses of YAP1 mRNA expression between cancer and cancer related specimens in one TCGA dataset and three GEO datasets. (E) Comparison of YAP1 expression level among different colorectal pathological tissues by cytoplasmic YAP1 H-score, (F) nuclear YAP1 H-score, or (G) YAP1 NCR H-score. (H-M) Representative YAP1 staining in normal tissues and cancer tissues, the blue staining represents the nuclear staining and the brown staining represents the YAP1 positive staining, cancer tissue have the higher cytoplasmic YAP1 H-score, higher nuclear YAP1 H-score and lower NCR than normal tissue, scale bars: 100&#956;m. *P&lt;0.05; **P&lt;0.01; ***P&lt;0.001; ****P&lt;0.0001; ns, no significance.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Fig. 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Fig. 2 High cytoplasmic YAP1 expression is associated with a worse survival in CRC patients. (A-D) Associations between cytoplasmic YAP1 expression and DFS in the patient subgroups with different stage. (E-H) Associations between cytoplasmic YAP1 expression and DSS in the patient subgroups with different stage. Patients with stages &#8544;-&#8546;, stage &#8544;, stage &#8545;, or stage &#8546; were dichotomized into the cytoplasmic YAP1-high subgroups and cytoplasmic YAP1-low subgroups according to optimal cut-off value. Kaplan-Meier survival curves reveal DFS and DSS in patients with each TNM stage CRC. P-values are from Kaplan-Meier analysis with logrank test.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Fig. 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Fig. 3 Adjuvant chemotherapy has a differential effect on patients with different cytoplasmic YAP1 expression levels. Associations between cytoplasmic YAP1 expression and DFS (A) or DSS (B) in the stage &#8546; patients with chemotherapy. Associations between cytoplasmic YAP1 expression and DFS (C) or DSS (D) in the stage &#8546; patients without chemotherapy. P-values are from Kaplan-Meier analysis with log-rank test.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Fig</ns0:head><ns0:label /><ns0:figDesc>Fig. S1 Supplementary bioinformatics analyses and differences in YAP1 expression between different differentiation grades. (A) Bioinformatics analyses of YAP1 mRNA expression levels in cancer and cancer-related specimens in five GEO datasets. (B) Correlation scatter plot of Cytoplasmic vs Nuclear YAP1 H-score. (C) Comparison of YAP1 expression levels between different differentiation grades by cytoplasmic YAP1 H-score (D) nuclear YAP1 H-score, or (E) YAP1 NCR H-score. *P&lt;0.05; **P&lt;0.01; ***P&lt;0.001; ****P&lt;0.0001; ns, no significance.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head>Fig</ns0:head><ns0:label /><ns0:figDesc>Fig. S2 Low YAP1 NCR is associated with a worse survival in CRC patients. (A) Associations between nuclear YAP1 expression and DFS or DSS in patients with stages &#8544;-</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head>Fig</ns0:head><ns0:label /><ns0:figDesc>Fig. S3 Adjuvant chemotherapy has a differential effect on patients with different YAP1 NCR. Associations between YAP1 NCR and DFS (A) or DSS (B) in the stage &#8546; patients with</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:05:49163:2:0:NEW 25 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='19,42.52,70.87,525.00,411.75' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_2'><ns0:head /><ns0:label /><ns0:figDesc>A univariate and multivariate Cox regression analyses (Further evaluation of meaningful prognostic factors in univariate analysis in multivariate analysis) was applied to determined the independence of the prognostic value of YAP1 on the basis of the DFS and DSS of CRC</ns0:figDesc><ns0:table><ns0:row><ns0:cell>patients, the results showed that high cytoplasmic YAP1 expression was an independent risk</ns0:cell></ns0:row><ns0:row><ns0:cell>factor of DFS (HR=3.255; 95%CI=2.290-4.627; P&lt;0.001) and DSS (HR=4.049; 95%CI=2.400-</ns0:cell></ns0:row><ns0:row><ns0:cell>6.830; P&lt;0.001) for CRC patients (Table 2), likewise, low YAP1 NCR was an independent risk</ns0:cell></ns0:row></ns0:table><ns0:note>factor ofDFS (HR=2.295; P=0.024) and DSS (HR=2.873; 95%CI=1.045-7.902; P=0.041) for CRC patients (Table2), but the univariate Cox regression analysis showed that nuclear YAP1 expression level was not a meaningful prognostic factor either for DFSPeerJ reviewing PDF | (2020:05:49163:2:0:NEW 25 Oct 2020)Manuscript to be reviewed (HR=0.684; 95%CI=0.453-1.031; P=0.07) or DSS(HR=0.860; P=0.688) </ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_3'><ns0:head>Table 1 . Associations of cytoplasmic YAP1 expression with demographic and clinical variables of 919 CRC patients</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>YAP1, Yes associated protein 1; TNM, tumor-node-metastasis; CEA, carcinoembryonic antigen; CA19-9, carbohydrate antigen 19-9.</ns0:figDesc><ns0:table><ns0:row><ns0:cell>Characteristics</ns0:cell><ns0:cell>Total (n=919)</ns0:cell><ns0:cell cols='2'>Cytoplasmic YAP1 expression level Low(n=457) High(n=462)</ns0:cell><ns0:cell>P Value *</ns0:cell></ns0:row><ns0:row><ns0:cell>Mean age&#177;SD(year)</ns0:cell><ns0:cell>60.1&#177;12.4</ns0:cell><ns0:cell>61.4&#177;12.3</ns0:cell><ns0:cell>60.7&#177;12.5</ns0:cell><ns0:cell>0.389 **</ns0:cell></ns0:row><ns0:row><ns0:cell>Sex (n (%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.686</ns0:cell></ns0:row><ns0:row><ns0:cell>Men</ns0:cell><ns0:cell>549(59.7)</ns0:cell><ns0:cell>270(59.1)</ns0:cell><ns0:cell>279(60.4)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Women</ns0:cell><ns0:cell>370(40.3)</ns0:cell><ns0:cell>187(40.9)</ns0:cell><ns0:cell>183(39.6)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Disease location (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.632</ns0:cell></ns0:row><ns0:row><ns0:cell>Rectum</ns0:cell><ns0:cell>512(55.7)</ns0:cell><ns0:cell>251(54.9)</ns0:cell><ns0:cell>261(56.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Colon</ns0:cell><ns0:cell>407(44.3)</ns0:cell><ns0:cell>206(45.1)</ns0:cell><ns0:cell>201(43.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Differentiation grade (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>&lt;0.001 ***</ns0:cell></ns0:row><ns0:row><ns0:cell>Well</ns0:cell><ns0:cell>95(10.3)</ns0:cell><ns0:cell>67(14.7)</ns0:cell><ns0:cell>28(6.1)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Moderately</ns0:cell><ns0:cell>779(84.8)</ns0:cell><ns0:cell>369(80.7)</ns0:cell><ns0:cell>410(88.7)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Poorly</ns0:cell><ns0:cell>35(3.8)</ns0:cell><ns0:cell>14(3.1)</ns0:cell><ns0:cell>21(4.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Missing</ns0:cell><ns0:cell>10(1.1)</ns0:cell><ns0:cell>7(1.5)</ns0:cell><ns0:cell>3(0.6)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Resected lymph nodes (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>&lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;12</ns0:cell><ns0:cell>201(21.9)</ns0:cell><ns0:cell>140(30.6)</ns0:cell><ns0:cell>61(13.2)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8805;12</ns0:cell><ns0:cell>718(78.1)</ns0:cell><ns0:cell>317(69.4)</ns0:cell><ns0:cell>401(86.8)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>TNM stage (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.362 ***</ns0:cell></ns0:row><ns0:row><ns0:cell>&#8544;</ns0:cell><ns0:cell>140(15.2)</ns0:cell><ns0:cell>65(14.2)</ns0:cell><ns0:cell>75(16.2)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8545;</ns0:cell><ns0:cell>459(49.9)</ns0:cell><ns0:cell>245(53.6)</ns0:cell><ns0:cell>214(46.3)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8546;</ns0:cell><ns0:cell>320(34.8)</ns0:cell><ns0:cell>147(32.2)</ns0:cell><ns0:cell>173(37.4)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Chemotherapy (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.730</ns0:cell></ns0:row><ns0:row><ns0:cell>Yes</ns0:cell><ns0:cell>671(73.0)</ns0:cell><ns0:cell>336(73.5)</ns0:cell><ns0:cell>335(72.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>No</ns0:cell><ns0:cell>248(27.0)</ns0:cell><ns0:cell>121(26.5)</ns0:cell><ns0:cell>127(27.5)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serum CEA (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.451</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;5ng/ml</ns0:cell><ns0:cell>568(61.8)</ns0:cell><ns0:cell>288(63)</ns0:cell><ns0:cell>280(60.6)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8805;5ng/ml</ns0:cell><ns0:cell>351(38.2)</ns0:cell><ns0:cell>169(37)</ns0:cell><ns0:cell>182(39.4)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serum CA19-9 (n(%))</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell>0.686</ns0:cell></ns0:row><ns0:row><ns0:cell>&lt;37U/ml</ns0:cell><ns0:cell>788(85.7)</ns0:cell><ns0:cell>394(86.2)</ns0:cell><ns0:cell>394(85.3)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>&#8805;37U/ml</ns0:cell><ns0:cell>131(14.3)</ns0:cell><ns0:cell>63(13.8)</ns0:cell><ns0:cell>68(14.7)</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell>Notes: * &#967;2 test.</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table><ns0:note>** Student t-test. *** Mann-Whitney U test (non-parametric). Missing values are excluded for all statistic tests. Abbreviations:</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_4'><ns0:head>Table 2 (on next page)</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Cox regression analysis of immunohistochemistry YAP1 expression and clinicopathological covariates in patients with CRCAbbreviations: HR, hazard ratio; CI, confidence interval; YAP1, Yes associated protein 1; TNM, tumor-node-metastasis; CEA, carcinoembryonic antigen; CA19-9, carbohydrate antigen 19-9; NCR, Nuclear/Cytoplasmic Ratio.</ns0:figDesc><ns0:table /><ns0:note>PeerJ reviewing PDF | (2020:05:49163:2:0:NEW 25 Oct 2020)Manuscript to be reviewed 1</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_5'><ns0:head>Table 2 . Cox regression analysis of immunohistochemistry YAP1 expression and clinicopathological covariates in patients with 2 CRC Disease-free Survival Disease-specific Survival</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>Characteristics</ns0:cell><ns0:cell>Univariate</ns0:cell><ns0:cell /><ns0:cell>Multivariate</ns0:cell><ns0:cell /><ns0:cell>Univariate</ns0:cell><ns0:cell /><ns0:cell>Multivariate</ns0:cell><ns0:cell /></ns0:row><ns0:row><ns0:cell /><ns0:cell /><ns0:cell>P</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell><ns0:cell /><ns0:cell>P</ns0:cell></ns0:row><ns0:row><ns0:cell /><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell><ns0:cell>HR (95%CI)</ns0:cell><ns0:cell>Value</ns0:cell></ns0:row><ns0:row><ns0:cell cols='3'>YAP1-high vs. YAP1-low(cytoplasmic) 3.891 (2.758-5.490) &lt;0.001</ns0:cell><ns0:cell cols='2'>3.255 (2.290-4.627) &lt;0.001</ns0:cell><ns0:cell cols='2'>4.291 (2.545-7.236) &lt;0.001</ns0:cell><ns0:cell cols='2'>4.049 (2.400-6.830) &lt;0.001</ns0:cell></ns0:row><ns0:row><ns0:cell>YAP1-low vs. YAP1-high(NCR)</ns0:cell><ns0:cell>2.709(1.331-5.511)</ns0:cell><ns0:cell>0.006</ns0:cell><ns0:cell>2.295(1.118-4.711)</ns0:cell><ns0:cell>0.024</ns0:cell><ns0:cell>3.346(1.219-9.183)</ns0:cell><ns0:cell>0.019</ns0:cell><ns0:cell>2.873(1.045-7.902)</ns0:cell><ns0:cell>0.041</ns0:cell></ns0:row><ns0:row><ns0:cell>YAP1-high vs. YAP1-low(nuclear)</ns0:cell><ns0:cell>0.684(0.453-1.031)</ns0:cell><ns0:cell>0.070</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.860(0.412-1.975)</ns0:cell><ns0:cell>0.688</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Age (&gt;=60 vs. &lt;60)</ns0:cell><ns0:cell>0.897 (0.667-1.207)</ns0:cell><ns0:cell>0.474</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.891 (0.568-1.398)</ns0:cell><ns0:cell>0.617</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Sex (female vs. male)</ns0:cell><ns0:cell>0.867 (0.638-1.177)</ns0:cell><ns0:cell>0.360</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>0.817 (0.512-1.302)</ns0:cell><ns0:cell>0.395</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Location (colon vs. rectum)</ns0:cell><ns0:cell>1.068 (0.793-1.440)</ns0:cell><ns0:cell>0.665</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell>1.159 (0.737-1.823)</ns0:cell><ns0:cell>0.522</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>TNM (per increase in stage)</ns0:cell><ns0:cell cols='2'>1.874 (1.474-2.381) &lt;0.001</ns0:cell><ns0:cell cols='2'>1.863 (1.471-2.360) &lt;0.001</ns0:cell><ns0:cell>1.256 (0.889-1.775)</ns0:cell><ns0:cell>0.196</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Grade (per increase in grade)</ns0:cell><ns0:cell cols='2'>3.001 (1.948-4.625) &lt;0.001</ns0:cell><ns0:cell cols='2'>3.435 (2.127-5.548) &lt;0.001</ns0:cell><ns0:cell>2.992 (1.575-5.685)</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell>2.732 (1.383-5.394)</ns0:cell><ns0:cell>0.004</ns0:cell></ns0:row><ns0:row><ns0:cell>Chemotherapy (yes vs. no)</ns0:cell><ns0:cell>1.705 (1.156-2.515)</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell>1.029 (0.647-1.637)</ns0:cell><ns0:cell>0.902</ns0:cell><ns0:cell>1.125 (0.662-1.912)</ns0:cell><ns0:cell>0.663</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Resected lymph nodes (&#8805;12 vs. &lt;12)</ns0:cell><ns0:cell cols='2'>2.675 (1.689-4.236) &lt;0.001</ns0:cell><ns0:cell>1.780 (1.111-2.853)</ns0:cell><ns0:cell>0.017</ns0:cell><ns0:cell>2.610 (1.375-4.954)</ns0:cell><ns0:cell>0.003</ns0:cell><ns0:cell>1.685 (0.874-3.251)</ns0:cell><ns0:cell>0.120</ns0:cell></ns0:row><ns0:row><ns0:cell>Serum CEA (&#8805;5 vs. &lt;5ng/ml)</ns0:cell><ns0:cell>1.646 (1.223-2.215)</ns0:cell><ns0:cell>0.001</ns0:cell><ns0:cell>1.513 (1.120-2.043)</ns0:cell><ns0:cell>0.007</ns0:cell><ns0:cell>1.378 (0.876-2.170)</ns0:cell><ns0:cell>0.166</ns0:cell><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Serum CA19-9 (&#8805;37 vs. &lt;37U/ml)</ns0:cell><ns0:cell>1.766 (1.224-2.549)</ns0:cell><ns0:cell>0.002</ns0:cell><ns0:cell>1.350 (0.914-1.995)</ns0:cell><ns0:cell>0.132</ns0:cell><ns0:cell>1.619 (0.906-2.894)</ns0:cell><ns0:cell>0.104</ns0:cell><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot' n='3'>Abbreviations: HR, hazard ratio; CI, confidence interval; YAP1, Yes associated protein 1; TNM, tumor-node-metastasis; CEA, carcinoembryonic antigen; 4 CA19-9, carbohydrate antigen 19-9; NCR, Nuclear/Cytoplasmic Ratio.PeerJ reviewing PDF | (2020:05:49163:2:0:NEW 25 Oct 2020)</ns0:note> </ns0:body> "

"Response letter Dear editor, On behalf of myself and the co-authors, thank you for the constructive feedback received on Sep. 17, 2020, and for providing us with an opportunity to revise our manuscript (manuscript ID: 49163). We appreciate the editor’s and reviewer2’s constructive comments on our manuscript entitled ‘High cytoplasmic YAP1 expression predicts a poor prognosis in patients with colorectal cancer’. We have carefully reviewed the reviewer2’s comments and we have made the relevant revisions, which have been highlighted in the revised manuscript. We hope that the revised manuscript meets all the requirements to be accepted for publication in PeerJ. Responses to reviewer2’s comments: Basic reporting Raw data. The authors declined to provide raw data such as TNM stages, differentiation grade for each patient due to the concern for patient confidential information, although in principle information such as TNM stage should not be any different from the already provided DFS, DSS, Chemotherapy raw information. The authors provided the manual grading of H-scores before averaging. However, the nuclear H-score appear weird. All nuclear H-scores are in the increment of 5, which should be the case if it is calculated in the same way as the cytoplasmic H-score. Response: Thank you. Due to the patients’ raw clinicopathological characteristics involve the original confidential information of the patients, we decided to attach the raw clinicopathological characteristics to this letter only for review by editors and reviewers. The percentage of each staining intensity cannot be very precise because H-score is a manual scoring, we calculated the percentage of each staining intensity of the nucleus and part of the cytoplasm in increment of 5, so H-scores are in the increment of 5. Experimental design No comment Validity of the findings 1. The authors said they had compared the group with chemotherapy (either high cytoplasmic YAP1 or low cytoplasmic YAP1) to the corresponding group without chemotherapy as suggested, and revised their conclusion based on this comparison to “Stage III CRC patients with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy”. However, they did not make any update to Figure 3 to reflect this comparison. Response: Thank you and sorry for the misunderstanding caused by language description, in the previous response letter we wrote that ‘We have compared the group with chemotherapy (either high cytoplasmic YAP1 or low cytoplasmic YAP1) to the corresponding group without chemotherapy, but there are no significant difference in survival between two groups’, we have conducted this comparison according to your suggestion, but there are no significant differences in survival among these two groups, and according to your recent comment, in further revised manuscript, we have changed our conclusion drawn by figure3, therefore we did not change the content of figure3. 2. The authors rewrote their final conclusion of Figure 3 to “therefore, as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy”. This conclusion is exciting, but it is not supported by their data at all and is misleading. It sounds like for patient with lower cytoplasmic YAP1 expression, they did receive significant survival benefit from adjuvant chemotherapy; and for patient with high cytoplasmic expression, they did not. Related to the point above, the author did not show this comparison in their Figure 3. And based on the data presented in Figure 3, chemotherapy did not seem to benefit patient with lower cytoplasmic YAP1. Instead, chemotherapy even appeared to have an adverse effect on the survival of patients with low cytoplasmic YAP1, although this may largely be due to the small sample size of the group with low cytoplasmic YAP1 and without chemotherapy. Either way, the data does not support the conclusion drawn by the authors. Response: Thank you for your constructive suggestion. According to the existing figure3 data, we have changed subtitle from ‘Stage III CRC patients with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy’ to ‘High cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy’ And changed the expression from ‘Therefore, as compared to the patients in stage III with lower cytoplasmic YAP1 expression, those with high cytoplasmic YAP1 expression did not receive significant survival benefit from adjuvant chemotherapy, in fact, the DFS and DSS shorten in cytoplasmic-high YAP1 subgroups who received adjuvant chemotherapy.’ to ‘Therefore, High cytoplasmic YAP1 expression is associated with a worse survival in stage III CRC patients who received chemotherapy.’ And the relevant expressions in other parts of the manuscript have also been changed accordingly. The current expression plainly describes the existing figure3 data, and this finding is clinically significant because there were no significant differences between YAP1-high subgroup and YAP1-low subgroup in DFS and DSS for stage III patients without adjuvant chemotherapy. Comments for the author In this revised manuscript, Dong and colleagues addressed some of my questions. However, they did not address my concern on the adjuvant chemotherapy result, which is a particularly important part of this study. Their conclusion is not supported by the data presented. Please see specific comments for details. Response: Thank you, we have changed the way we describe the figure3 data, and this finding is clinically significant. Thank you for your constructive suggestions to help improve the quality and scientific integrity of our manuscript. Sincerely, Shangyong Zheng patinet# 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 DFS RECURE 1 1 2 2 4 5 5 5 5 5 6 6 7 7 7 7 7 7 7 7 8 8 8 10 10 10 10 10 10 10 10 10 10 10 11 11 11 11 11 12 12 12 12 12 12 12 12 12 13 13 13 14 14 14 14 14 14 15 15 15 15 15 15 15 15 16 16 16 16 17 17 17 17 17 17 17 17 17 18 18 18 18 18 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 20 20 20 20 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 23 23 23 23 23 23 23 23 23 23 23 23 23 23 23 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 26 26 26 26 26 26 26 DSS 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 1 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 0 1 1 0 0 0 1 0 0 1 0 0 1 1 0 1 1 1 0 1 0 0 0 0 0 0 1 1 1 0 0 1 1 0 1 1 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 1 0 0 1 1 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 0 1 0 0 0 1 0 0 1 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 DEATH 5 5 2 23 12 20 5 5 38 5 29 21 7 12 7 63 10 11 32 7 25 25 8 10 37 33 38 35 10 29 10 15 52 10 19 11 26 14 20 61 12 21 27 60 13 15 12 14 13 13 59 15 18 24 24 20 14 26 15 54 37 29 15 24 26 16 16 17 23 42 32 17 18 18 23 24 20 29 32 27 20 20 18 34 54 24 20 20 29 24 27 19 32 19 19 20 20 33 20 20 27 20 20 53 20 20 21 53 24 21 54 21 21 21 21 21 21 55 24 51 21 21 23 67 22 60 24 32 22 22 22 22 22 25 22 22 22 22 22 22 29 22 23 22 43 23 23 23 25 27 60 26 23 23 23 35 23 23 23 24 24 24 24 24 24 24 24 24 24 65 24 25 24 24 24 24 40 24 24 24 24 24 24 24 24 57 25 25 25 62 31 60 25 25 25 25 25 25 25 25 25 25 26 26 26 26 26 26 26 CHEMOTHERAPAY 1 0 0 0 0 1 1 0 1 0 1 0 0 0 0 0 1 1 1 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 1 1 1 0 0 0 0 0 0 0 1 0 0 1 1 0 0 0 1 0 0 1 0 0 1 0 1 1 1 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 1 1 1 1 0 1 1 1 1 1 0 0 1 0 1 1 1 1 0 1 1 1 1 0 1 0 0 0 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 1 1 0 1 0 1 0 0 0 1 1 0 1 1 1 1 1 1 1 1 1 0 1 1 0 1 0 0 1 1 0 0 0 1 1 0 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 1 0 1 1 1 1 1 1 1 1 0 1 0 0 1 1 1 1 0 1 1 1 1 1 1 1 1 0 1 0 1 1 0 1 0 1 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 0 1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 1 1 0 1 0 1 1 1 1 CYTOPLASMIC H-SCORE 270 300 300 264 240 282 240 234 300 300 270 276 240 300 264 300 252 300 300 288 264 282 252 246 252 234 252 228 258 252 300 264 288 264 270 288 270 204 276 258 264 288 240 240 240 240 300 276 300 240 264 276 252 246 300 246 216 228 258 300 300 264 300 300 210 282 300 252 300 270 228 228 252 252 276 300 264 264 252 240 300 276 276 276 252 258 300 258 270 230 288 288 288 242 270 253 253 288 300 242 276 270 242 247 242 300 230 230 259 236 230 230 230 299 265 300 300 242 299 247 276 247 293 265 300 288 230 299 224 247 230 270 276 161 288 276 219 242 300 300 253 230 201 247 299 253 288 247 219 300 293 276 259 265 288 288 265 207 242 230 300 253 230 276 300 253 288 270 288 300 253 230 293 219 259 300 299 253 253 259 224 300 300 247 276 270 253 173 230 300 213 253 253 236 230 224 265 300 300 237 231 300 237 237 300 300 292 220 253 NUCLEAR H-SCORE 175 140 150 135 100 255 140 50 150 150 10 200 170 155 150 50 60 160 225 170 80 240 10 125 20 40 60 170 150 160 250 20 190 100 140 150 20 30 110 40 80 250 120 40 200 200 50 140 225 10 20 225 125 195 285 95 20 90 200 120 50 80 170 270 30 95 170 170 225 160 100 150 110 170 160 280 200 10 170 185 125 260 125 190 60 50 200 50 80 180 10 125 230 50 240 170 100 250 270 120 10 200 50 50 140 240 155 125 90 90 10 110 60 200 225 270 130 125 40 50 190 90 220 60 100 80 125 20 195 260 280 250 80 150 190 125 170 170 230 265 170 90 5 250 150 130 125 125 120 100 200 220 155 50 175 80 170 5 50 280 150 80 30 215 200 30 50 150 240 210 170 110 230 195 20 200 60 60 125 155 10 170 250 170 180 150 150 30 0 10 20 50 235 170 200 95 205 280 60 125 180 225 35 150 150 240 200 40 155 NCR 0.648148 0.466667 0.500000 0.511364 0.416667 0.904255 0.583333 0.213675 0.500000 0.500000 0.037037 0.724638 0.708333 0.516667 0.568182 0.166667 0.238095 0.533333 0.750000 0.590278 0.303030 0.851064 0.039683 0.508130 0.079365 0.170940 0.238095 0.745614 0.581395 0.634921 0.833333 0.075758 0.659722 0.378788 0.518519 0.520833 0.074074 0.147059 0.398551 0.155039 0.303030 0.868056 0.500000 0.166667 0.833333 0.833333 0.166667 0.507246 0.750000 0.041667 0.075758 0.815217 0.496032 0.792683 0.950000 0.386179 0.092593 0.394737 0.775194 0.400000 0.166667 0.303030 0.566667 0.900000 0.142857 0.336879 0.566667 0.674603 0.750000 0.592593 0.438596 0.657895 0.436508 0.674603 0.579710 0.933333 0.757576 0.037879 0.674603 0.770833 0.416667 0.942029 0.452899 0.688406 0.238095 0.193798 0.666667 0.193798 0.296296 0.782609 0.034783 0.434783 0.800000 0.207039 0.888067 0.671937 0.395257 0.869565 0.900000 0.496894 0.036232 0.740056 0.207039 0.202224 0.579710 0.800000 0.673913 0.543478 0.347826 0.381760 0.043478 0.478261 0.260870 0.668896 0.850662 0.900000 0.433333 0.517598 0.133779 0.202224 0.688406 0.364004 0.750213 0.226843 0.333333 0.278261 0.543478 0.066890 0.869565 1.051567 1.217391 0.925069 0.289855 0.931677 0.660870 0.452899 0.778032 0.703934 0.766667 0.883333 0.671937 0.391304 0.024845 1.011122 0.501672 0.513834 0.434783 0.505561 0.549199 0.333333 0.682012 0.797101 0.599034 0.189036 0.608696 0.278261 0.642722 0.024155 0.207039 1.217391 0.500000 0.316206 0.130435 0.778986 0.666667 0.118577 0.173913 0.555042 0.834783 0.700000 0.671937 0.478261 0.784314 0.892449 0.077295 0.666667 0.200669 0.237154 0.494071 0.599034 0.044593 0.566667 0.833333 0.687563 0.652174 0.555042 0.592885 0.173913 0.000000 0.033333 0.094007 0.197628 0.928854 0.721103 0.869565 0.423634 0.775047 0.933333 0.200000 0.528541 0.779221 0.750000 0.147992 0.634249 0.500000 0.800000 0.686106 0.181818 0.612648 CYTOPLASMIC H-SCORE1 257 300 300 265 235 286 242 230 300 300 283 264 236 300 277 300 271 300 300 278 277 281 262 242 256 244 256 238 262 251 300 280 283 260 260 290 252 223 258 285 280 277 249 254 248 239 300 258 300 247 270 257 267 238 300 236 215 230 255 300 300 262 300 300 202 275 300 251 300 281 227 243 250 258 282 300 276 275 252 250 300 262 266 255 251 272 300 255 285 221 281 283 278 255 281 283 271 278 300 235 252 254 241 247 247 300 212 228 247 240 244 243 236 299 249 300 300 251 300 240 260 237 292 280 300 285 223 300 249 238 225 253 268 149 287 268 205 252 300 300 268 249 209 254 300 283 279 249 197 300 295 260 277 254 277 285 272 208 239 220 300 281 234 262 300 265 286 254 286 300 255 231 292 201 261 300 300 261 248 261 217 300 300 236 262 263 257 187 247 300 217 274 265 244 224 225 257 300 300 241 226 300 213 230 300 300 292 217 264 CYTOPLASMIC H-SCORE2 283 300 300 263 245 278 238 238 300 300 257 288 244 300 251 300 233 300 300 298 251 283 242 250 248 224 248 218 254 253 300 248 293 268 280 286 288 185 294 231 248 299 231 226 232 241 300 294 300 233 258 295 237 254 300 256 217 226 261 300 300 266 300 300 218 289 300 253 300 259 229 213 254 246 270 300 252 253 252 230 300 290 286 297 253 244 300 261 255 239 294 292 297 228 260 223 235 297 300 248 300 287 242 248 236 300 248 232 271 232 216 217 224 299 280 300 300 232 298 255 292 258 295 249 300 290 237 298 200 257 235 288 284 173 288 284 232 231 300 300 238 211 194 241 298 223 296 246 240 300 292 292 241 275 298 290 257 206 244 240 300 225 226 290 300 241 289 287 289 300 251 229 295 236 257 300 298 245 258 257 232 300 300 259 290 278 249 158 213 300 209 232 241 228 236 224 272 300 300 232 236 300 260 243 300 300 291 223 242 NUCLEAR H-SCORE1 175 145 165 125 105 245 140 45 150 145 5 195 155 165 145 60 60 175 220 155 95 250 5 125 15 35 55 165 155 160 260 25 185 100 150 165 15 30 120 40 95 255 115 50 215 210 60 145 230 10 15 230 120 185 285 100 25 90 220 105 40 80 175 265 30 95 165 160 215 160 110 165 115 170 170 275 205 10 160 185 140 255 130 185 65 60 220 45 85 170 10 140 235 60 245 155 100 255 260 110 10 195 50 40 145 230 165 130 100 100 5 105 65 200 230 255 130 135 40 40 200 80 215 70 95 80 120 25 185 280 275 250 90 150 185 135 170 175 240 255 170 85 0 260 160 130 135 135 105 105 225 220 155 45 170 80 165 10 60 280 155 80 35 225 200 40 45 145 245 210 165 105 225 200 15 200 55 55 125 160 5 160 245 175 175 165 150 30 0 10 25 55 250 175 200 95 210 275 65 130 175 215 30 145 145 230 225 45 165 NUCLEAR H-SCORE2 175 135 135 145 95 265 140 55 150 155 15 205 185 145 155 40 60 145 230 185 65 230 15 125 25 45 65 175 145 160 240 15 195 100 130 135 25 30 100 40 65 245 125 30 185 190 40 135 220 10 25 220 130 205 285 90 15 90 180 135 60 80 165 275 30 95 175 180 235 160 90 135 105 170 150 285 195 10 180 185 110 265 120 195 55 40 180 55 75 190 10 110 225 40 235 185 100 245 280 130 10 205 50 60 135 250 145 120 80 80 15 115 55 200 220 285 130 115 40 60 180 100 225 50 105 80 130 15 205 240 285 250 70 150 195 115 170 165 220 275 170 95 10 240 140 130 115 115 135 95 175 220 155 55 180 80 175 0 40 280 145 80 25 205 200 20 55 155 235 210 175 115 235 190 25 200 65 65 125 150 15 180 255 165 185 135 150 30 0 10 15 45 220 165 200 95 200 285 55 120 185 235 40 155 155 250 175 35 145 Patient_age Patient_sex 1male 2female 83 52 44 59 76 75 60 74 59 62 76 68 45 55 64 45 81 61 49 55 50 71 79 71 50 77 47 43 55 74 74 55 60 37 52 64 73 88 42 73 74 63 54 55 63 49 51 63 43 37 64 53 32 34 43 85 59 59 66 70 44 51 57 68 52 70 59 56 65 67 73 76 66 67 56 72 72 53 73 45 61 46 68 59 62 87 67 77 77 68 59 76 64 74 81 81 84 75 30 68 84 35 72 39 46 77 41 39 71 49 56 69 54 57 60 70 40 69 70 54 38 62 72 73 55 65 44 39 40 64 75 49 65 59 59 58 46 87 81 71 57 60 67 71 60 54 60 54 78 45 66 75 49 44 51 58 57 78 79 83 55 53 39 47 53 67 71 71 74 40 69 80 76 36 81 78 60 59 26 71 79 51 60 44 66 66 74 65 59 50 57 78 66 53 81 45 74 76 75 59 57 56 71 45 67 57 40 72 77 Location1-rectum, 2-colon 1 1 2 1 1 2 2 1 1 1 1 1 1 1 1 1 2 1 2 1 2 2 2 1 2 1 1 2 2 2 2 1 1 1 1 2 1 2 2 1 1 1 2 1 2 1 1 2 1 1 2 1 1 1 1 2 2 1 1 2 2 1 1 1 2 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 1 1 2 2 1 2 1 1 2 1 1 1 2 2 1 2 2 2 1 2 1 1 1 2 2 2 1 1 2 2 1 1 1 1 2 1 1 2 2 2 2 2 2 1 2 2 1 2 1 2 1 1 2 1 2 1 1 2 1 1 1 1 1 1 2 1 2 1 1 1 1 1 2 1 1 1 1 2 1 1 1 2 2 1 1 1 1 1 1 1 2 1 2 2 1 1 1 2 1 2 2 1 1 1 1 1 2 1 2 1 1 1 1 1 1 2 2 1 2 1 1 TNM 2 1 1 1 2 2 1 2 1 1 1 1 1 2 1 2 2 2 1 1 1 2 1 2 2 1 1 1 2 2 2 1 1 1 1 2 2 1 2 1 2 1 2 1 2 2 1 1 2 1 2 1 2 2 2 2 2 1 1 1 1 2 2 1 1 2 1 1 1 1 1 2 2 2 1 1 2 2 2 2 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 2 2 1 1 1 2 2 2 2 1 1 1 2 1 1 1 2 2 1 1 1 1 2 2 1 2 1 2 2 1 2 1 2 1 1 1 1 1 2 1 2 1 2 2 1 2 2 1 1 2 1 2 1 2 2 1 1 1 2 1 2 2 1 2 2 2 2 2 2 2 2 1 1 2 2 2 2 2 1 1 1 1 2 2 2 1 2 1 2 2 2 1 1 2 1 1 2 1 1 2 1 1 2 1 1 1 2 1 2 2 2 2 2 2 3 3 3 3 3 2 2 3 3 3 2 2 3 3 3 2 2 3 1 3 3 2 3 2 3 2 2 2 1 2 3 1 1 3 1 2 3 3 2 3 3 3 2 3 3 3 2 3 2 3 3 2 2 3 3 2 3 3 1 3 2 3 3 2 3 3 2 2 1 3 1 2 2 2 2 2 3 3 3 3 3 3 3 3 3 2 2 2 3 2 1 1 3 2 2 1 3 2 3 2 3 2 2 3 3 2 3 2 1 3 3 3 3 3 3 2 1 2 3 2 2 2 3 3 1 2 1 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 3 1 2 3 2 3 3 3 3 2 1 3 1 2 2 3 2 2 2 2 2 2 3 2 1 3 2 3 2 3 2 2 2 2 2 2 2 2 3 2 2 1 3 2 1 3 2 3 1 3 2 2 2 2 2 3 2 1 3 2 1 2 1 3 3 2 3 Grade1poor2medium3well 2 2 1 2 2 2 2 2 2 2 2 2 1 1 1 1 2 1 2 2 2 2 1 2 2 2 2 3 2 1 2 2 2 2 2 1 2 2 2 2 2 2 2 3 3 2 2 2 2 2 2 1 1 3 2 2 2 2 2 2 2 1 2 2 2 2 2 3 2 2 1 2 1 1 3 2 2 2 3 1 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 #NULL! 2 #NULL! 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 3 CEA-1 CA199-1 2 2 2 2 1 2 1 2 2 1 2 2 1 1 2 1 2 2 1 2 2 2 2 2 1 1 1 2 1 2 2 2 2 1 1 1 2 2 1 1 1 1 2 2 2 2 2 2 2 2 2 2 1 1 2 2 1 2 2 1 1 2 1 1 2 2 1 1 1 2 1 2 2 1 2 2 2 2 1 2 2 2 1 1 1 1 1 2 1 2 2 1 1 1 1 1 2 2 1 2 1 2 2 2 2 2 1 1 1 2 2 1 2 1 2 1 1 2 2 1 1 1 1 2 1 2 1 2 2 1 1 2 2 2 1 1 2 1 1 1 1 1 2 1 1 1 1 2 2 1 2 1 2 1 1 1 1 N/A 2 1 2 1 1 2 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 2 1 2 1 1 2 1 1 1 1 1 2 1 1 1 2 1 1 2 1 1 1 2 2 1 1 2 1 2 1 1 1 2 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 2 2 1 1 1 2 1 1 1 1 2 1 1 1 1 1 2 2 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 1 2 1 1 2 1 1 2 1 2 1 1 1 1 2 1 N/A 2 2 1 2 1 1 2 1 2 2 2 2 1 1 1 1 1 1 2 2 1 2 1 2 1 2 2 2 1 1 1 1 1 1 1 1 2 1 1 2 2 1 1 1 1 2 1 2 2 2 1 2 1 1 2 1 1 1 1 1 2 2 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 2 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 2 1 2 1 1 1 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 26 26 26 26 26 26 26 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 27 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 30 30 30 30 30 30 31 31 31 31 31 31 32 32 32 32 32 32 32 32 33 33 33 33 34 34 34 35 35 35 35 35 35 36 36 36 36 37 37 37 37 38 38 38 39 39 39 40 40 41 41 41 41 41 41 41 41 41 41 41 41 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 44 44 44 44 44 44 44 44 44 44 44 44 44 44 45 45 45 45 45 45 45 45 45 45 45 45 45 45 45 45 45 45 0 0 0 0 1 1 1 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 0 0 1 1 0 0 1 1 1 0 0 1 0 0 0 1 0 1 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 26 26 26 26 31 53 59 27 27 27 27 27 27 27 27 27 27 28 27 27 27 27 27 27 27 27 27 35 27 27 27 27 27 27 27 27 28 60 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 28 29 62 62 29 29 29 29 29 29 29 29 29 29 29 29 29 29 38 29 58 30 30 30 30 32 31 31 31 31 31 46 34 32 32 32 32 32 32 32 33 33 42 49 34 34 34 47 41 53 35 35 57 45 36 36 43 39 53 37 37 48 38 38 39 50 39 88 40 48 54 47 41 41 41 41 41 41 41 41 41 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 43 43 43 43 43 43 43 53 43 43 43 43 43 43 43 43 43 43 50 43 43 43 43 44 44 44 44 44 44 44 44 44 44 44 44 44 44 45 45 45 45 45 45 58 45 45 45 45 45 45 45 45 45 45 45 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0 1 0 0 1 1 1 0 0 1 0 0 1 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 0 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 0 0 1 0 1 1 1 1 1 0 0 0 1 1 1 0 1 0 0 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 0 1 0 0 0 0 1 1 0 1 1 1 0 1 1 1 1 1 1 1 0 1 1 1 1 0 0 1 0 1 0 1 1 1 1 1 0 0 0 0 0 1 0 0 0 0 1 1 1 1 1 1 0 1 1 1 1 1 0 1 0 1 1 1 0 1 1 1 1 1 0 1 0 1 1 1 1 1 1 0 0 1 0 1 1 0 1 1 0 1 1 0 1 1 1 0 1 0 0 0 1 1 300 231 248 300 226 242 286 215 198 300 226 237 275 209 220 300 231 198 193 292 286 300 215 215 264 242 253 237 300 231 231 176 242 286 242 220 242 198 198 275 204 253 198 300 253 220 286 220 231 187 209 242 275 220 220 275 193 300 300 242 248 220 209 215 209 220 220 209 220 275 231 200 210 273 221 263 194 299 210 263 221 210 247 300 263 210 284 263 231 221 168 210 194 210 205 210 221 210 210 221 210 210 231 210 252 184 242 210 210 210 284 242 210 300 231 210 210 210 210 210 210 210 221 215 210 194 236 210 231 189 210 236 247 184 210 205 168 221 273 226 189 210 231 205 210 205 226 231 231 221 245 226 242 231 200 158 116 236 200 215 242 210 105 221 210 231 210 278 231 215 226 221 231 210 147 184 231 210 200 126 189 257 189 210 221 215 263 231 247 210 189 231 210 179 299 168 210 231 220 190 200 210 205 195 250 220 180 200 250 260 175 235 20 250 30 20 50 110 20 20 110 50 150 20 75 230 200 100 235 200 195 255 185 185 160 80 215 80 280 50 60 95 125 160 20 90 50 20 30 10 60 50 120 170 120 155 225 30 120 90 130 210 170 170 220 40 180 275 250 100 20 45 30 30 125 140 110 155 125 210 155 155 140 250 200 50 80 175 80 140 155 200 125 100 190 170 125 150 120 90 20 170 95 185 80 80 165 140 80 210 80 125 80 125 200 50 60 30 125 100 10 140 180 230 190 100 125 200 0 170 50 50 185 220 20 125 210 80 125 100 90 90 125 70 80 90 100 150 125 125 60 90 205 155 185 80 210 125 235 235 225 225 150 280 100 155 5 170 170 175 220 155 270 80 200 80 80 200 260 170 170 155 280 80 10 30 230 70 50 80 170 100 90 80 30 100 125 80 100 200 20 120 200 150 295 90 90 120 80 10 100 260 140 125 290 80 100 50 175 260 0.583333 1.017316 0.080808 0.833333 0.133038 0.082645 0.174825 0.512821 0.101010 0.066667 0.487805 0.211416 0.545455 0.095694 0.340909 0.766667 0.865801 0.505051 1.220779 0.686106 0.681818 0.850000 0.862471 0.862471 0.606061 0.330579 0.849802 0.338266 0.933333 0.216450 0.259740 0.539773 0.516529 0.559441 0.082645 0.409091 0.206612 0.101010 0.151515 0.036364 0.294840 0.197628 0.606061 0.566667 0.474308 0.704545 0.786713 0.136364 0.519481 0.481283 0.622010 0.867769 0.618182 0.772727 1.000000 0.145455 0.935065 0.916667 0.833333 0.413223 0.080808 0.204545 0.143541 0.139860 0.598086 0.636364 0.500000 0.741627 0.568182 0.763636 0.670996 0.776942 0.666667 0.915751 0.907029 0.190476 0.411840 0.584795 0.380952 0.533333 0.702948 0.952381 0.506586 0.333333 0.723810 0.809524 0.440917 0.571429 0.519481 0.408163 0.119048 0.809524 0.489060 0.880952 0.390720 0.380952 0.748299 0.666667 0.380952 0.952381 0.380952 0.595238 0.346320 0.595238 0.793651 0.272109 0.248447 0.142857 0.595238 0.476190 0.035273 0.579710 0.857143 0.766667 0.822511 0.476190 0.595238 0.952381 0.000000 0.809524 0.238095 0.238095 0.839002 1.022067 0.095238 0.643501 0.888889 0.380952 0.541126 0.529101 0.428571 0.380952 0.506586 0.380952 0.380952 0.439560 0.595238 0.680272 0.457875 0.553710 0.317460 0.428571 0.887446 0.757021 0.880952 0.390720 0.930233 0.541126 1.017316 1.065760 0.918367 0.996678 0.621118 1.212121 0.501253 0.984127 0.043290 0.719577 0.852130 0.813008 0.910973 0.738095 2.571429 0.362812 0.952381 0.346320 0.380952 0.718778 1.125541 0.789779 0.753045 0.702948 1.212121 0.380952 0.068027 0.163265 0.995671 0.333333 0.250627 0.634921 0.899471 0.388727 0.476190 0.380952 0.136054 0.464576 0.476190 0.346320 0.405268 0.952381 0.105820 0.519481 0.952381 0.840336 0.985798 0.535714 0.428571 0.519481 0.363636 0.052632 0.500000 1.238095 0.682927 0.641026 1.160000 0.363636 0.555556 0.250000 0.700000 1.000000 300 219 238 300 211 254 282 218 170 300 248 228 252 218 250 300 246 188 176 286 284 300 200 211 251 243 270 221 300 217 227 203 255 285 251 250 250 195 174 278 209 278 182 300 266 225 290 211 228 180 203 254 276 239 218 254 198 300 300 252 249 241 220 204 198 250 242 219 224 271 230 225 217 269 249 268 183 299 210 256 232 197 254 300 271 214 281 267 237 225 169 202 192 201 207 213 217 210 221 218 223 196 213 213 278 193 242 206 197 216 278 255 210 300 221 200 204 201 217 221 218 215 219 217 196 202 224 224 244 205 212 230 254 183 203 201 175 232 273 245 170 204 228 213 197 201 211 240 236 226 235 248 252 246 225 158 109 220 224 223 239 214 106 238 223 250 200 282 218 209 228 237 220 205 148 184 226 214 206 133 187 273 181 199 242 211 263 239 252 204 179 232 216 187 299 149 202 247 228 168 204 205 206 184 274 242 195 195 247 275 300 243 257 300 240 230 290 211 226 300 203 245 298 200 190 300 216 208 209 297 288 300 229 218 277 241 236 252 300 245 235 149 229 287 233 190 234 201 222 272 198 228 214 300 240 215 282 229 234 194 215 230 274 201 222 296 187 300 300 232 246 199 198 225 220 190 198 199 216 279 232 174 203 277 192 257 206 300 210 269 209 223 240 300 254 206 286 258 225 216 167 218 197 219 203 207 224 210 199 223 197 224 249 207 226 175 241 214 223 204 289 228 210 300 241 220 216 219 203 199 202 205 222 214 224 187 249 196 218 173 208 243 240 185 217 209 161 209 273 207 208 216 234 197 223 209 241 222 226 215 255 204 231 216 174 157 122 253 175 208 244 206 104 203 197 212 220 275 244 222 224 204 242 215 146 184 236 206 193 119 191 242 197 221 199 220 262 223 242 216 199 230 204 170 300 187 218 215 212 212 196 215 204 206 226 198 165 205 253 245 185 240 20 250 30 25 50 105 15 20 105 40 145 25 75 230 195 110 240 210 195 240 195 185 160 95 205 85 275 45 60 110 125 165 20 100 40 20 25 5 60 60 110 170 130 160 220 35 110 100 125 210 175 170 225 40 170 265 240 95 15 40 35 30 120 145 115 155 120 205 160 145 140 255 210 55 80 175 75 140 155 200 120 110 200 155 130 155 120 80 15 165 95 200 90 95 155 155 80 220 95 120 75 130 215 55 70 25 125 100 5 135 175 230 185 110 130 200 0 165 45 40 180 215 25 125 220 80 135 105 90 90 130 75 75 90 100 150 120 130 70 95 220 145 185 85 205 130 250 245 225 230 155 285 95 145 10 175 170 170 230 155 280 95 195 95 90 205 270 175 175 160 275 80 10 40 225 70 45 75 160 110 85 90 25 95 130 85 110 215 15 105 210 155 295 95 100 125 90 5 95 275 155 120 295 85 105 40 175 265 165 230 20 250 30 15 50 115 25 20 115 60 155 15 75 230 205 90 230 190 195 270 175 185 160 65 225 75 285 55 60 80 125 155 20 80 60 20 35 15 60 40 130 170 110 150 230 25 130 80 135 210 165 170 215 40 190 285 260 105 25 50 25 30 130 135 105 155 130 215 150 165 140 245 190 45 80 175 85 140 155 200 130 90 180 185 120 145 120 100 25 175 95 170 70 65 175 125 80 200 65 130 85 120 185 45 50 35 125 100 15 145 185 230 195 90 120 200 0 175 55 60 190 225 15 125 200 80 115 95 90 90 120 65 85 90 100 150 130 120 50 85 190 165 185 75 215 120 220 225 225 220 145 275 105 165 0 165 170 180 210 155 260 65 205 65 70 195 250 165 165 150 285 80 10 20 235 70 55 85 180 90 95 70 35 105 120 75 90 185 25 135 190 145 295 85 80 115 70 15 105 245 125 130 285 75 95 60 175 255 62 46 81 67 57 77 47 55 73 64 73 86 40 56 61 54 46 68 41 43 55 60 59 60 75 64 67 54 68 82 60 50 65 74 46 87 51 53 59 54 65 88 54 58 62 54 61 57 67 75 77 73 47 75 79 71 71 30 65 57 50 73 37 64 72 50 76 41 47 59 74 65 51 68 62 76 60 52 67 53 76 45 55 56 51 65 31 58 73 77 58 55 52 59 71 83 67 77 73 73 55 41 64 65 44 64 49 67 84 52 71 59 78 77 57 39 71 76 71 82 78 71 60 59 61 77 74 45 60 61 41 21 71 82 78 70 53 48 78 64 55 48 45 58 69 38 63 75 60 44 56 72 39 35 64 62 71 76 63 62 84 73 54 48 50 75 76 41 73 63 57 55 54 79 66 53 54 46 66 57 74 54 60 58 66 75 73 61 81 67 61 30 60 61 45 72 60 44 58 67 56 61 36 74 69 83 55 64 54 69 1 1 1 1 2 1 2 2 1 2 2 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 2 2 1 2 2 1 2 2 1 1 1 1 1 1 2 1 1 1 1 2 2 2 2 2 2 2 1 2 2 1 1 2 1 1 1 1 2 1 1 2 1 1 1 2 1 2 1 2 1 1 2 1 1 1 2 1 1 2 2 2 2 1 2 1 1 2 1 1 1 1 1 2 2 1 1 1 2 2 1 1 1 2 1 2 2 1 2 2 1 1 1 1 2 2 2 2 2 1 2 1 1 2 2 1 2 1 1 1 2 1 2 1 1 2 2 1 1 2 1 2 1 2 2 2 1 2 1 2 2 1 1 2 1 1 1 1 2 1 1 1 1 1 2 2 2 1 1 2 1 2 2 1 1 1 2 1 2 2 2 1 2 1 2 2 1 1 1 2 1 2 1 2 1 2 1 1 1 1 1 2 1 2 1 1 2 2 1 2 1 2 1 2 1 2 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 1 2 1 1 1 2 1 2 1 2 1 2 1 1 2 2 1 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 1 2 2 1 1 1 2 2 1 1 1 1 2 1 2 1 1 1 2 2 2 2 1 2 2 1 1 2 2 1 1 2 1 2 2 1 1 2 2 1 1 2 2 2 1 2 2 1 1 2 2 1 1 1 2 1 2 1 2 1 1 1 1 2 2 1 1 2 1 2 1 1 1 1 1 1 1 1 2 2 1 1 1 2 1 2 1 1 2 1 1 1 1 1 1 2 1 1 2 1 1 2 2 1 1 1 1 1 2 2 2 2 1 1 1 1 2 1 1 1 2 1 2 1 1 1 1 1 1 1 2 1 1 1 1 2 3 2 2 2 3 2 3 3 2 1 1 3 2 2 2 2 3 3 3 3 3 3 3 3 3 3 2 2 2 2 2 3 3 2 2 3 2 2 3 1 2 3 1 2 3 2 3 2 2 2 2 2 3 3 1 2 2 1 3 2 3 2 2 2 2 2 2 2 3 3 2 2 3 2 3 3 3 2 2 2 3 3 3 2 2 2 3 3 2 1 2 2 3 3 3 3 2 2 2 3 2 3 3 1 2 3 2 2 2 1 1 2 1 3 3 2 3 3 2 2 2 2 2 2 3 2 3 1 3 2 2 3 1 1 3 3 3 3 3 3 2 2 2 1 1 1 1 1 2 1 2 2 1 3 2 2 2 2 3 3 3 3 3 3 3 1 3 2 3 2 3 1 2 2 2 2 2 2 2 2 2 2 3 3 3 2 1 2 2 1 2 2 1 2 2 1 2 2 1 2 2 2 2 2 1 1 3 2 2 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 3 #NULL! 3 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 3 2 3 #NULL! 2 2 2 2 2 2 3 2 3 2 3 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 #NULL! 2 2 2 2 2 2 2 2 1 3 2 3 3 3 3 3 3 2 2 2 3 3 1 2 2 2 3 2 2 1 2 2 2 2 2 2 2 3 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 1 2 1 2 1 2 1 1 2 2 1 2 1 2 2 1 2 1 2 1 2 1 1 2 1 2 1 1 2 1 1 1 2 2 1 1 2 1 2 1 1 1 1 2 2 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 2 2 1 2 1 1 1 2 2 2 2 1 1 2 2 1 1 1 1 2 1 1 2 1 1 1 2 2 1 2 2 2 1 1 1 2 1 2 1 1 1 2 2 1 1 1 2 2 1 2 2 2 1 1 1 2 1 2 1 1 2 1 1 2 2 2 2 1 1 2 1 2 1 1 2 1 1 2 1 1 1 1 1 1 1 1 1 1 2 1 1 2 1 2 1 2 1 2 2 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 2 1 2 1 1 1 1 2 1 1 1 1 1 2 1 1 1 1 1 2 1 2 1 1 2 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 2 2 1 1 1 2 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 2 1 1 2 2 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 2 2 2 1 2 1 1 2 1 2 1 1 2 2 2 1 2 1 2 1 1 1 1 1 2 2 1 2 1 1 2 1 2 2 2 1 2 2 1 1 1 2 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 46 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 47 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 51 51 51 51 51 51 51 51 51 51 51 51 51 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 54 54 54 54 54 54 54 54 54 54 54 54 54 54 54 54 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 46 46 46 46 46 46 46 90 46 46 46 46 46 46 46 46 46 46 46 46 46 46 58 47 47 47 47 47 47 47 47 47 47 47 47 57 47 47 47 47 47 47 47 47 47 60 47 47 47 47 47 47 47 47 47 47 47 47 47 47 48 48 48 48 48 48 48 48 48 48 48 48 90 48 48 48 48 48 48 48 48 48 48 48 48 48 60 48 48 48 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 50 50 50 50 50 50 50 50 57 71 60 50 50 50 50 50 50 51 50 50 51 51 51 51 51 51 51 51 51 51 51 51 51 52 52 52 52 52 52 56 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 54 54 54 54 54 66 54 54 54 54 54 54 54 54 54 54 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 1 1 1 1 1 0 1 0 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 0 0 0 1 0 0 0 0 0 1 0 1 0 1 1 1 1 0 0 0 0 0 1 1 1 0 1 0 1 1 0 0 1 1 1 0 0 1 0 1 1 0 1 0 0 1 1 1 0 1 1 1 1 0 0 1 1 0 1 0 1 0 1 1 1 0 1 0 1 0 1 0 1 0 1 1 0 0 0 1 0 1 1 1 1 1 0 1 0 0 0 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0 1 0 1 1 1 1 1 0 0 1 0 0 1 0 0 1 0 0 1 0 1 1 0 0 1 1 0 1 1 0 1 1 1 1 1 1 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 185 215 185 240 200 215 200 200 200 225 110 270 215 220 250 200 185 220 140 200 200 275 270 185 220 140 210 200 275 185 180 180 150 195 150 200 195 205 235 205 140 180 290 255 160 190 160 205 185 185 175 200 140 215 245 150 195 180 235 195 205 165 225 210 200 240 160 190 260 225 175 160 240 185 180 205 185 285 245 280 200 225 195 265 215 185 230 215 165 225 255 185 235 235 195 225 195 235 185 255 195 230 220 220 105 220 200 190 190 220 200 200 200 250 257 181 209 214 176 152 209 214 257 219 181 162 181 181 219 209 190 105 228 200 152 114 124 152 190 171 238 190 190 171 219 133 190 176 190 171 238 114 143 162 200 190 200 181 86 171 190 181 190 181 190 171 209 214 162 238 204 209 181 190 209 176 219 209 190 162 190 190 190 190 200 190 209 190 257 162 152 200 190 171 190 209 171 219 86 95 133 171 190 143 171 171 180 189 189 171 125 155 155 260 80 60 60 90 90 50 0 260 130 230 220 5 40 170 20 50 40 155 125 30 175 0 50 30 250 30 140 60 50 155 90 20 80 80 170 225 10 80 200 280 200 170 150 125 50 25 65 200 60 120 180 100 180 125 250 155 170 170 80 155 170 145 110 140 110 240 145 140 110 80 30 300 230 250 205 225 65 95 290 285 215 30 120 140 115 110 220 5 230 30 110 175 230 230 60 260 180 255 190 270 20 125 20 100 60 140 180 170 70 190 280 10 80 250 60 80 90 100 240 170 170 120 70 90 80 120 20 10 40 210 30 5 150 30 200 160 185 160 80 185 210 50 120 130 140 170 160 40 110 20 200 50 150 200 50 155 250 60 190 80 90 20 60 180 30 150 100 100 100 100 160 10 230 145 170 125 80 160 200 185 40 170 60 140 145 80 40 130 120 90 100 110 70 255 10 10 100 50 200 80 80 75 190 50 80 60 0.675676 0.720930 0.837838 1.083333 0.400000 0.279070 0.300000 0.450000 0.450000 0.222222 0.000000 0.962963 0.604651 1.045455 0.880000 0.025000 0.216216 0.772727 0.142857 0.250000 0.200000 0.563636 0.462963 0.162162 0.795455 0.000000 0.238095 0.150000 0.909091 0.162162 0.777778 0.333333 0.333333 0.794872 0.600000 0.100000 0.410256 0.390244 0.723404 1.097561 0.071429 0.444444 0.689655 1.098039 1.250000 0.894737 0.937500 0.609756 0.270270 0.135135 0.371429 1.000000 0.428571 0.558140 0.734694 0.666667 0.923077 0.694444 1.063830 0.794872 0.829268 1.030303 0.355556 0.738095 0.850000 0.604167 0.687500 0.736842 0.423077 1.066667 0.828571 0.875000 0.458333 0.432432 0.166667 1.463415 1.243243 0.877193 0.836735 0.803571 0.325000 0.422222 1.487179 1.075472 1.000000 0.162162 0.521739 0.651163 0.696970 0.488889 0.862745 0.027027 0.978723 0.127660 0.564103 0.777778 1.179487 0.978723 0.324324 1.019608 0.923077 1.108696 0.863636 1.227273 0.190476 0.568182 0.100000 0.526316 0.315789 0.636364 0.900000 0.850000 0.350000 0.760000 1.091618 0.055402 0.382775 1.169591 0.341394 0.526316 0.430622 0.467836 0.935673 0.778032 0.941828 0.743034 0.387812 0.498615 0.366133 0.574163 0.105263 0.095694 0.175439 1.052632 0.197368 0.043860 1.214575 0.197368 1.052632 0.935673 0.778947 0.842105 0.421053 1.081871 0.961098 0.375940 0.631579 0.739687 0.736842 0.994152 0.673684 0.350877 0.771930 0.123839 1.002506 0.263158 0.751880 1.108033 0.584795 0.906433 1.315789 0.332410 1.000000 0.443213 0.473684 0.116959 0.287081 0.842105 0.185759 0.631579 0.489596 0.478469 0.554017 0.526316 0.765550 0.056899 1.052632 0.693780 0.894737 0.773994 0.421053 0.842105 1.052632 0.973684 0.200501 0.894737 0.287081 0.736842 0.565302 0.495356 0.263158 0.651629 0.631579 0.526316 0.526316 0.526316 0.409357 1.167048 0.116959 0.105263 0.751880 0.292398 1.052632 0.561404 0.467836 0.438596 1.055556 0.264550 0.423280 0.350877 178 211 175 240 216 217 220 224 206 225 106 281 208 213 240 202 183 237 153 221 201 278 278 191 220 135 220 215 281 195 175 181 145 197 159 196 168 222 240 224 140 181 295 263 167 173 168 215 190 195 171 206 138 220 241 160 166 194 230 178 217 146 211 196 195 251 152 202 247 246 177 151 240 199 198 225 172 282 255 277 207 238 199 270 214 184 241 223 151 228 266 202 220 239 181 215 202 216 179 264 184 246 220 216 114 216 216 202 205 242 204 222 195 240 274 170 215 209 193 147 204 220 274 224 199 174 205 169 225 221 172 100 235 211 152 107 117 149 172 187 215 191 181 201 211 134 181 166 184 195 214 121 135 145 206 174 221 195 81 180 181 196 171 167 169 197 203 211 172 247 210 203 186 197 206 187 209 195 190 169 171 178 191 195 215 203 219 201 282 155 154 221 189 192 175 196 169 215 91 93 125 180 186 155 197 166 183 165 166 178 192 219 195 240 184 213 180 176 194 225 114 259 222 227 260 198 187 203 127 179 199 272 262 179 220 145 200 185 269 175 185 179 155 193 141 204 222 188 230 186 140 179 285 247 153 207 152 195 180 175 179 194 142 210 249 140 224 166 240 212 193 184 239 224 205 229 168 178 273 204 173 169 240 171 162 185 198 288 235 283 193 212 191 260 216 186 219 207 179 222 244 168 250 231 209 235 188 254 191 246 206 214 220 224 96 224 184 178 175 198 196 178 205 260 239 191 203 219 159 157 214 208 239 213 162 149 156 192 212 197 208 109 221 188 152 121 130 155 208 155 260 189 199 141 226 132 199 186 196 147 261 107 150 178 193 206 178 166 90 162 199 165 209 194 211 145 215 217 151 228 199 215 175 183 212 165 228 223 190 154 209 202 189 185 184 177 199 179 231 168 150 178 191 150 205 222 173 222 80 97 141 162 194 130 145 176 177 213 212 164 130 145 150 260 75 55 65 85 80 40 0 275 140 235 225 0 55 155 20 40 35 155 120 35 175 0 60 35 250 25 135 70 50 165 100 25 95 85 170 215 5 90 195 280 225 170 155 130 40 35 60 195 70 125 185 110 175 125 250 145 160 175 85 165 165 150 125 155 125 230 135 145 125 85 40 300 235 250 205 215 65 105 295 285 225 25 105 145 125 110 215 10 230 30 125 175 235 240 60 260 180 240 190 255 25 130 15 105 65 150 180 165 65 180 290 10 80 245 70 85 95 110 235 170 160 120 85 95 85 125 20 10 40 210 35 10 155 25 210 155 195 165 95 200 220 45 105 130 150 170 165 50 130 15 220 60 160 210 45 160 245 65 190 90 95 20 65 170 30 165 100 95 95 100 165 5 225 140 175 130 90 165 205 180 50 170 65 155 140 80 35 140 120 100 110 125 65 245 10 5 90 55 210 95 75 65 185 45 85 60 120 165 160 260 85 65 55 95 100 60 0 245 120 225 215 10 25 185 20 60 45 155 130 25 175 0 40 25 250 35 145 50 50 145 80 15 65 75 170 235 15 70 205 280 175 170 145 120 60 15 70 205 50 115 175 90 185 125 250 165 180 165 75 145 175 140 95 125 95 250 155 135 95 75 20 300 225 250 205 235 65 85 285 285 205 35 135 135 105 110 225 0 230 30 95 175 225 220 60 260 180 270 190 285 15 120 25 95 55 130 180 175 75 200 270 10 80 255 50 75 85 90 245 170 180 120 55 85 75 115 20 10 40 210 25 0 145 35 190 165 175 155 65 170 200 55 135 130 130 170 155 30 90 25 180 40 140 190 55 150 255 55 190 70 85 20 55 190 30 135 100 105 105 100 155 15 235 150 165 120 70 155 195 190 30 170 55 125 150 80 45 120 120 80 90 95 75 265 10 15 110 45 190 65 85 85 195 55 75 60 54 47 63 53 37 59 59 51 71 70 72 71 69 44 56 35 58 74 70 54 65 40 45 35 69 79 39 69 76 32 76 58 62 74 67 63 68 56 86 70 46 44 66 48 46 75 43 71 84 53 64 71 54 72 80 66 49 68 58 70 59 82 58 67 74 68 68 54 60 65 65 40 69 39 67 61 65 55 68 64 61 39 80 71 52 54 45 51 47 65 84 63 58 54 61 67 75 54 71 71 58 57 68 79 52 70 30 75 70 36 53 76 61 52 64 57 60 57 62 36 58 54 62 71 40 56 76 65 61 68 49 66 52 48 62 42 38 63 42 67 44 54 67 46 56 67 70 62 60 69 33 58 73 58 69 66 43 63 54 67 55 60 41 59 45 71 63 55 70 78 77 37 72 59 64 71 75 85 66 57 77 48 39 32 67 49 55 60 47 34 50 60 54 54 58 64 76 54 77 68 67 55 77 48 52 76 51 66 65 69 1 2 2 1 1 2 1 2 1 2 1 2 2 1 2 2 2 1 1 1 2 1 1 2 1 2 2 2 1 1 1 2 2 1 1 1 1 1 1 2 1 2 1 2 1 1 1 1 1 1 1 1 2 1 2 2 1 1 1 2 2 2 2 1 1 2 1 1 1 1 2 2 2 2 1 1 2 1 1 1 2 2 1 1 1 2 1 1 2 2 1 2 2 1 2 1 2 1 1 1 1 2 1 1 1 1 2 1 1 1 1 2 2 1 1 1 2 1 2 1 1 1 2 1 2 2 1 1 1 1 1 1 1 2 2 2 1 1 2 1 2 2 2 1 1 1 1 1 2 2 2 1 1 1 2 2 1 2 2 1 2 1 1 1 1 2 2 2 1 2 2 1 1 2 2 2 1 1 1 2 2 1 1 2 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 1 2 1 1 2 2 1 1 1 1 1 2 1 1 2 1 1 2 1 1 1 1 1 1 2 2 1 1 2 2 2 1 1 1 2 1 2 1 1 2 2 1 1 1 1 1 2 1 1 2 2 1 1 1 1 1 2 2 2 2 1 1 1 2 1 1 2 1 1 1 1 1 1 1 1 1 2 1 1 2 2 1 1 1 2 1 1 1 1 2 2 2 2 1 1 1 1 2 1 1 2 2 1 2 2 2 1 2 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 2 1 1 2 1 2 1 1 1 1 1 1 1 1 2 1 1 2 2 1 1 1 1 1 1 1 2 1 1 2 1 1 2 1 1 1 1 2 1 1 2 1 1 2 2 1 1 1 1 2 2 1 2 1 1 1 1 1 1 3 2 1 3 1 3 2 3 2 3 2 3 1 2 2 3 3 2 2 2 3 3 2 3 1 1 3 2 2 2 1 1 2 2 1 1 1 1 2 2 1 3 1 3 2 3 2 1 1 1 1 1 2 2 3 1 2 1 3 3 1 1 3 2 2 1 1 2 1 3 3 2 2 1 2 2 3 3 2 3 3 3 2 1 1 3 2 1 2 3 2 1 2 2 3 1 3 1 2 2 3 1 3 2 2 3 1 1 1 3 1 2 2 2 3 3 1 3 1 1 1 3 3 2 2 1 3 2 3 3 2 2 3 3 3 3 1 2 2 2 3 3 3 2 3 1 2 3 2 3 2 3 2 1 2 1 3 2 2 3 2 1 1 2 2 1 3 1 1 3 1 1 3 1 2 3 2 2 2 2 2 2 2 2 3 3 2 3 2 3 1 3 3 3 2 2 2 1 1 2 2 2 2 3 2 2 2 2 3 3 2 2 2 2 2 2 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 2 2 2 1 2 1 1 2 1 1 1 1 2 1 2 2 1 1 1 1 2 2 1 1 2 2 2 1 1 1 1 2 1 1 1 1 2 1 1 2 1 1 1 2 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 2 2 1 2 1 1 1 1 2 1 1 1 1 1 1 2 1 2 2 1 2 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 1 2 1 1 1 2 1 1 1 1 2 1 2 2 1 2 1 1 2 2 1 1 1 1 1 1 1 1 N/A 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 1 1 1 1 1 2 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 N/A 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 N/A 2 2 1 2 1 2 2 1 1 2 2 2 1 1 1 2 1 2 1 1 1 1 1 2 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 1 1 2 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 2 1 1 2 1 2 2 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 2 1 1 1 1 1 1 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 55 55 55 55 55 55 55 55 55 55 55 55 55 56 56 56 56 56 56 56 56 56 56 56 56 56 56 57 57 57 57 57 57 57 57 57 57 57 57 57 58 58 58 58 58 58 58 58 58 58 58 58 58 58 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 62 62 62 62 62 62 62 62 62 62 63 63 63 63 63 64 64 64 64 65 65 65 65 65 65 65 65 66 66 66 66 66 67 67 67 67 67 67 68 68 68 68 68 69 69 69 69 70 70 70 70 70 70 70 71 71 71 71 71 71 72 72 72 72 72 73 73 73 73 73 73 73 73 73 73 73 74 74 74 74 74 74 74 74 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 55 55 55 55 55 55 55 55 55 55 59 55 55 56 56 56 56 56 56 56 56 56 56 56 56 56 56 57 57 57 57 57 57 57 57 57 57 57 57 57 58 58 58 58 58 58 58 58 58 58 58 58 58 58 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 60 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 65 61 61 61 61 61 61 61 62 62 62 62 62 62 62 62 62 62 63 63 63 65 63 64 64 64 64 65 65 65 65 65 65 68 65 66 66 66 66 66 67 67 67 67 67 67 68 68 68 68 68 69 69 69 69 70 70 70 70 70 70 70 71 71 71 71 71 71 72 72 72 72 72 73 73 73 73 73 73 73 73 73 73 73 74 74 74 74 74 74 74 74 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 1 0 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 0 1 1 0 0 1 1 0 1 0 1 1 1 0 1 1 1 1 1 1 0 1 1 0 1 0 1 0 1 1 1 1 1 1 0 1 0 1 0 0 1 1 1 1 0 0 0 1 0 1 0 1 1 1 0 1 0 0 1 1 1 1 1 1 1 0 0 1 0 1 1 1 1 1 1 1 0 0 1 0 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 0 1 0 1 0 0 1 1 1 1 0 1 1 1 1 0 1 1 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1 1 0 1 0 1 1 1 1 0 0 0 0 1 1 0 0 1 1 1 0 0 1 1 1 1 1 1 180 153 198 180 180 108 135 171 207 180 207 180 108 207 198 189 180 162 108 180 162 180 180 162 176 162 180 162 144 135 189 180 180 198 99 198 252 162 180 153 176 185 72 171 189 189 252 171 243 180 207 252 189 234 184 216 180 225 225 203 180 252 252 198 189 216 171 171 180 180 216 225 198 198 189 189 180 243 207 180 207 225 180 252 180 180 234 80 152 144 232 208 80 104 176 160 200 156 176 224 216 172 168 168 120 136 80 160 216 200 160 208 160 160 160 152 144 180 136 160 148 160 160 200 160 184 200 152 160 80 112 144 136 144 160 152 168 176 80 120 200 88 144 168 128 120 200 96 160 184 104 96 160 120 104 80 88 168 152 152 192 160 208 208 184 144 152 152 144 216 184 168 232 208 144 168 168 120 112 224 72 208 136 216 160 120 152 72 164 120 208 224 208 88 160 144 144 96 88 96 160 88 120 88 88 120 200 84 80 136 130 20 125 150 190 125 100 100 150 80 100 90 10 125 125 50 170 140 60 50 170 70 75 20 170 100 200 60 190 125 135 10 150 100 10 50 250 5 50 10 40 10 0 10 50 65 125 50 20 30 200 80 170 65 50 125 20 50 125 75 10 280 110 220 50 160 170 125 50 125 125 150 150 100 30 100 10 150 50 150 50 100 150 20 20 30 50 30 175 125 170 100 10 20 150 50 50 150 50 50 40 60 10 200 10 50 10 150 200 125 20 230 100 50 50 130 10 20 130 125 145 155 150 250 10 100 100 65 220 10 140 170 190 140 200 105 100 20 60 125 100 70 195 170 85 170 200 70 190 230 40 50 185 90 10 10 60 180 50 50 80 140 220 190 70 80 30 20 110 110 60 140 70 110 10 170 110 125 40 210 20 150 20 130 150 145 20 40 80 80 50 40 30 70 90 10 10 70 40 130 180 100 40 10 150 60 130 80 50 130 0.722222 0.130719 0.631313 0.833333 1.055556 1.157407 0.740741 0.584795 0.724638 0.444444 0.483092 0.500000 0.092593 0.603865 0.631313 0.264550 0.944444 0.864198 0.555556 0.277778 1.049383 0.388889 0.416667 0.123457 0.968661 0.617284 1.111111 0.370370 1.319444 0.925926 0.714286 0.055556 0.833333 0.505051 0.101010 0.252525 0.992063 0.030864 0.277778 0.065359 0.227920 0.054201 0.000000 0.058480 0.264550 0.343915 0.496032 0.292398 0.082305 0.166667 0.966184 0.317460 0.899471 0.277778 0.271739 0.578704 0.111111 0.222222 0.555556 0.370370 0.055556 1.111111 0.436508 1.111111 0.264550 0.740741 0.994152 0.730994 0.277778 0.694444 0.578704 0.666667 0.757576 0.505051 0.158730 0.529101 0.055556 0.617284 0.241546 0.833333 0.241546 0.444444 0.833333 0.079365 0.111111 0.166667 0.213675 0.375000 1.151316 0.868056 0.732759 0.480769 0.125000 0.192308 0.852273 0.312500 0.250000 0.961538 0.284091 0.223214 0.185185 0.348837 0.059524 1.190476 0.083333 0.367647 0.125000 0.937500 0.925926 0.625000 0.125000 1.105769 0.625000 0.312500 0.312500 0.855263 0.069444 0.111111 0.955882 0.781250 0.979730 0.968750 0.937500 1.250000 0.062500 0.543478 0.500000 0.427632 1.375000 0.125000 1.250000 1.180556 1.397059 0.972222 1.250000 0.690789 0.595238 0.113636 0.750000 1.041667 0.500000 0.795455 1.354167 1.011905 0.664063 1.416667 1.000000 0.729167 1.187500 1.250000 0.384615 0.520833 1.156250 0.750000 0.096154 0.125000 0.681818 1.071429 0.328947 0.328947 0.416667 0.875000 1.057692 0.913462 0.380435 0.555556 0.197368 0.131579 0.763889 0.509259 0.326087 0.833333 0.301724 0.528846 0.069444 1.011905 0.654762 1.041667 0.357143 0.937500 0.277778 0.721154 0.147059 0.601852 0.937500 1.208333 0.131579 0.555556 0.487805 0.666667 0.240385 0.178571 0.144231 0.795455 0.562500 0.069444 0.069444 0.729167 0.454545 1.354167 1.125000 1.136364 0.333333 0.113636 1.704545 0.500000 0.650000 0.952381 0.625000 0.955882 171 154 200 200 172 113 134 204 218 174 209 190 99 203 193 198 185 147 112 196 162 181 187 163 199 147 190 154 150 129 196 195 188 172 103 180 271 156 202 151 174 196 65 180 171 197 260 182 253 167 195 261 168 230 187 220 167 244 226 201 203 278 264 189 184 204 201 178 205 185 218 245 196 175 193 178 202 235 206 167 222 232 203 270 167 177 225 88 155 140 225 214 75 104 205 162 210 154 178 240 213 185 152 148 133 129 75 161 214 200 153 195 158 156 171 145 152 202 132 159 141 163 168 214 156 190 199 157 150 85 108 145 123 142 173 146 146 174 78 122 208 86 138 170 121 120 217 102 169 196 108 95 159 132 112 88 88 157 171 159 184 165 224 210 170 155 149 161 151 208 200 175 246 216 144 165 172 118 99 228 82 222 115 197 161 116 155 77 168 122 206 239 210 84 145 140 151 89 71 99 145 91 133 85 87 123 201 73 95 115 189 152 196 160 188 103 136 138 196 186 205 170 117 211 203 180 175 177 104 164 162 179 173 161 152 177 170 170 138 141 182 165 172 224 95 216 233 168 158 155 177 173 79 162 207 181 244 160 233 193 219 243 210 238 181 212 193 206 224 204 157 226 240 207 194 228 141 164 155 175 214 205 200 221 185 200 158 251 208 193 192 218 157 234 193 183 243 72 149 148 239 202 85 104 147 158 190 158 174 208 219 159 184 188 107 143 85 159 218 200 167 221 162 164 149 159 136 158 140 161 155 157 152 186 164 178 201 147 170 75 116 143 149 146 147 158 190 178 82 118 192 90 150 166 135 120 183 90 151 172 100 97 161 108 96 72 88 179 133 145 200 155 192 206 198 133 155 143 137 224 168 161 218 200 144 171 164 122 125 220 62 194 157 235 159 124 149 67 160 118 210 209 206 92 175 148 137 103 105 93 175 85 107 91 89 117 199 95 65 157 140 15 130 165 190 125 100 95 145 90 100 85 10 130 125 55 165 145 65 55 155 75 80 25 175 105 220 55 190 130 135 5 160 95 10 60 245 10 60 5 45 10 0 5 50 70 130 60 15 40 210 90 165 75 45 120 15 50 130 80 10 285 105 225 60 165 155 120 50 130 125 145 165 110 30 95 10 145 40 150 60 110 165 20 20 40 55 35 185 135 175 105 10 20 150 40 60 155 45 40 40 55 5 205 10 55 5 160 220 125 25 225 95 50 60 120 10 15 120 125 135 155 145 245 5 105 100 55 215 10 145 170 180 155 205 105 95 20 65 135 95 70 190 165 75 160 225 80 195 235 35 60 185 95 10 5 70 190 55 50 85 155 215 195 75 75 25 15 105 130 60 155 75 115 5 160 110 125 40 210 15 160 25 120 145 140 25 45 75 80 45 45 25 75 80 10 5 75 50 140 185 95 45 5 145 55 130 75 60 120 120 25 120 135 190 125 100 105 155 70 100 95 10 120 125 45 175 135 55 45 185 65 70 15 165 95 180 65 190 120 135 15 140 105 10 40 255 0 40 15 35 10 0 15 50 60 120 40 25 20 190 70 175 55 55 130 25 50 120 70 10 275 115 215 40 155 185 130 50 120 125 155 135 90 30 105 10 155 60 150 40 90 135 20 20 20 45 25 165 115 165 95 10 20 150 60 40 145 55 60 40 65 15 195 10 45 15 140 180 125 15 235 105 50 40 140 10 25 140 125 155 155 155 255 15 95 100 75 225 10 135 170 200 125 195 105 105 20 55 115 105 70 200 175 95 180 175 60 185 225 45 40 185 85 10 15 50 170 45 50 75 125 225 185 65 85 35 25 115 90 60 125 65 105 15 180 110 125 40 210 25 140 15 140 155 150 15 35 85 80 55 35 35 65 100 10 15 65 30 120 175 105 35 15 155 65 130 85 40 140 64 68 63 58 49 60 66 70 72 70 58 52 54 53 50 48 58 55 53 77 55 77 62 46 75 81 60 60 56 40 74 89 57 69 54 74 77 78 73 55 62 52 76 38 55 62 46 76 73 77 56 74 55 63 62 58 69 70 58 66 79 57 57 46 59 53 59 83 50 56 74 55 60 60 47 70 62 54 43 59 65 70 58 47 66 80 82 61 44 77 73 74 58 77 34 74 60 71 48 55 58 48 58 59 62 44 72 66 37 66 62 60 58 62 62 51 40 67 72 55 50 59 72 66 69 62 55 60 57 66 62 52 51 65 72 67 61 44 75 73 46 56 70 33 49 74 61 57 69 55 72 45 52 66 78 58 71 84 57 48 76 49 55 62 65 71 80 50 67 56 57 64 56 45 72 37 67 63 61 63 58 70 58 54 80 47 73 65 77 59 59 48 83 59 62 47 49 69 57 86 57 34 52 75 63 72 43 80 75 39 1 1 1 1 1 1 2 1 2 1 2 1 1 1 1 1 1 2 1 1 1 2 1 2 2 1 1 1 2 2 2 1 2 1 1 1 1 2 2 2 1 2 1 1 1 1 1 1 1 2 1 1 2 1 1 2 1 1 1 1 2 2 1 1 2 2 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 2 1 2 1 2 2 2 1 1 1 1 2 1 1 1 1 2 1 1 1 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 2 2 2 2 1 1 1 1 1 2 1 1 2 1 2 1 1 1 2 1 2 2 1 1 2 2 1 1 1 2 2 1 1 2 2 1 2 2 1 2 2 1 1 2 2 2 1 2 1 2 1 2 1 1 2 2 1 2 1 1 2 2 2 1 2 1 1 1 1 2 1 1 1 1 2 1 1 1 2 2 1 2 2 2 1 2 1 1 2 1 1 2 2 1 1 1 1 1 2 2 1 2 1 2 1 1 1 1 1 2 1 1 2 2 1 2 2 2 2 1 1 1 2 2 1 1 1 1 2 1 1 2 1 1 1 1 2 2 1 1 1 2 1 2 1 2 1 1 1 2 1 1 2 1 2 1 1 1 2 1 2 2 1 1 1 1 1 2 1 2 1 1 1 1 2 1 1 2 1 2 2 2 2 2 1 2 2 2 1 2 2 1 1 2 1 2 1 2 1 2 1 2 1 1 1 1 1 1 1 1 1 1 2 2 2 1 2 1 1 2 1 1 1 2 1 2 2 1 1 2 1 1 2 2 1 2 1 2 1 2 1 2 2 2 2 1 2 2 2 1 1 1 1 2 2 2 2 1 1 2 2 1 2 2 1 2 2 1 2 2 1 1 2 2 2 2 2 1 2 2 2 2 1 2 2 1 2 1 2 1 2 2 2 1 2 2 1 2 1 1 1 1 1 2 2 2 2 2 2 2 2 3 2 3 2 3 1 3 2 1 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 1 3 2 2 3 2 2 3 3 2 1 3 2 3 2 2 2 3 2 1 2 3 2 1 2 1 2 3 2 2 3 3 3 2 2 2 2 2 2 3 3 1 3 1 2 3 2 2 2 2 1 3 1 2 1 1 2 2 3 2 2 2 2 3 1 3 2 2 2 2 1 3 2 3 2 3 2 3 3 2 3 1 2 2 2 2 3 2 2 2 2 3 1 2 3 2 3 3 2 2 2 3 2 3 3 2 1 2 3 2 2 2 2 3 2 3 2 2 2 2 3 3 1 2 2 2 2 1 2 3 2 2 2 3 2 2 3 2 2 2 2 3 1 2 2 2 3 3 3 1 2 2 2 2 3 1 2 2 2 2 1 1 1 2 2 2 2 2 2 2 3 1 2 2 2 3 2 3 2 2 2 2 2 1 3 2 2 2 2 2 2 3 2 2 2 2 2 2 2 3 3 2 3 3 2 2 2 3 3 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 3 2 2 2 2 3 2 2 3 2 3 2 3 3 2 2 3 2 2 #NULL! 1 2 2 2 2 1 2 3 2 3 2 2 3 2 2 3 2 2 2 2 1 2 2 2 3 3 2 2 2 3 3 2 2 2 2 2 #NULL! 2 2 1 2 3 2 2 2 2 2 3 3 2 2 2 2 3 2 2 3 1 1 1 2 2 1 1 2 1 2 1 1 2 1 2 2 1 1 1 2 1 2 2 1 2 1 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 1 2 1 2 2 1 1 1 1 1 1 2 2 1 1 1 1 2 2 1 1 1 1 1 1 1 2 2 2 1 2 1 1 1 1 2 2 2 1 2 2 1 2 1 1 1 2 1 2 2 1 1 1 2 1 1 1 1 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 1 1 1 1 1 1 1 1 1 1 1 2 2 2 1 1 1 1 1 1 2 2 1 1 1 2 1 1 2 2 2 1 1 1 2 2 1 1 1 2 2 2 1 1 2 1 2 2 1 1 1 1 1 1 1 2 1 1 1 1 1 2 2 2 1 2 1 2 2 1 1 1 1 1 1 1 1 1 2 2 2 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 2 1 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1 2 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 2 1 2 1 1 1 1 1 2 2 1 1 2 1 1 1 1 1 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 74 74 75 75 75 75 75 76 76 76 76 76 77 77 77 77 77 77 78 79 80 80 80 81 81 82 82 82 82 82 83 83 83 85 85 85 85 85 86 86 86 86 87 88 88 88 88 89 89 89 89 91 91 91 91 92 92 92 93 94 95 101 101 102 102 102 102 103 103 103 103 104 107 107 111 113 113 115 115 117 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 74 74 75 75 75 75 75 81 76 76 76 76 77 82 77 77 77 77 78 79 80 80 80 81 81 82 82 82 82 82 83 83 83 85 85 85 85 85 86 86 86 86 87 88 88 88 88 96 89 89 89 91 91 91 91 92 92 92 93 94 95 101 101 102 102 102 102 103 103 103 103 104 107 107 111 113 113 115 115 117 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1 0 1 1 1 1 1 1 1 1 1 0 1 1 0 1 0 1 1 1 1 1 1 1 1 1 1 1 1 136 200 128 88 80 92 160 96 144 232 144 164 160 204 160 64 136 160 40 76 120 136 160 144 120 152 152 168 104 120 160 128 160 96 144 216 208 176 160 160 164 80 152 176 160 80 160 128 160 176 128 216 104 112 160 144 136 72 144 80 112 184 88 152 104 184 96 160 160 96 176 112 144 152 136 160 168 128 160 120 80 110 140 10 55 45 130 175 20 200 10 65 120 20 40 100 100 125 10 170 90 40 200 30 80 280 180 60 40 190 200 190 150 80 70 160 120 90 80 120 45 20 30 180 60 30 70 80 200 50 100 235 100 10 250 200 70 70 110 40 190 30 140 40 140 40 120 200 10 10 210 20 20 70 50 30 20 120 10 100 0.588235 0.550000 1.093750 0.113636 0.687500 0.489130 0.812500 1.822917 0.138889 0.862069 0.069444 0.396341 0.750000 0.098039 0.250000 1.562500 0.735294 0.781250 0.250000 2.236842 0.750000 0.294118 1.250000 0.208333 0.666667 1.842105 1.184211 0.357143 0.384615 1.583333 1.250000 1.484375 0.937500 0.833333 0.486111 0.740741 0.576923 0.511364 0.500000 0.750000 0.274390 0.250000 0.197368 1.022727 0.375000 0.375000 0.437500 0.625000 1.250000 0.284091 0.781250 1.087963 0.961538 0.089286 1.562500 1.388889 0.514706 0.972222 0.763889 0.500000 1.696429 0.163043 1.590909 0.263158 1.346154 0.217391 1.250000 1.250000 0.062500 0.104167 1.193182 0.178571 0.138889 0.460526 0.367647 0.187500 0.119048 0.937500 0.062500 0.833333 119 222 138 87 97 95 159 97 136 236 151 170 172 216 167 75 134 169 48 69 130 128 167 141 127 155 172 175 99 134 168 121 158 100 155 221 202 178 168 167 168 97 164 189 153 72 166 133 172 182 123 213 103 106 172 149 132 76 143 88 104 186 88 156 101 193 92 150 145 93 181 111 136 164 137 147 153 117 174 132 153 178 118 89 63 89 161 95 152 228 137 158 148 192 153 53 138 151 32 83 110 144 153 147 113 149 132 161 109 106 152 135 162 92 133 211 214 174 152 153 160 63 140 163 167 88 154 123 148 170 133 219 105 118 148 139 140 68 145 72 120 182 88 148 107 175 100 170 175 99 171 113 152 140 135 173 183 139 146 108 90 125 150 10 60 45 130 170 20 195 5 60 130 20 40 100 90 125 10 155 95 55 210 25 80 275 185 65 40 200 200 185 145 75 85 175 115 100 80 115 55 20 25 190 65 30 80 90 200 55 95 240 110 10 260 210 75 85 130 45 185 40 150 35 155 50 110 200 10 10 210 25 25 75 55 30 15 110 10 110 70 95 130 10 50 45 130 180 20 205 15 70 110 20 40 100 110 125 10 185 85 25 190 35 80 285 175 55 40 180 200 195 155 85 55 145 125 80 80 125 35 20 35 170 55 30 60 70 200 45 105 230 90 10 240 190 65 55 90 35 195 20 130 45 125 30 130 200 10 10 210 15 15 65 45 30 25 130 10 90 66 58 42 53 64 56 44 71 56 80 66 56 42 73 56 73 60 51 38 80 62 69 20 74 74 72 51 57 62 52 77 60 68 65 51 46 76 72 73 67 49 73 69 46 68 55 72 50 71 68 74 56 76 54 75 73 50 78 34 50 53 63 73 76 71 39 50 70 67 62 70 54 76 50 41 55 77 62 69 61 1 1 2 1 1 1 2 2 1 2 1 2 2 1 2 1 1 1 1 2 1 1 2 2 2 1 1 2 2 1 1 2 2 1 2 2 2 2 2 1 1 1 1 1 1 2 1 1 1 2 2 2 2 2 1 2 1 2 1 1 2 2 2 2 2 1 1 2 2 2 1 2 1 1 1 2 1 1 2 2 2 2 2 1 1 1 2 2 1 2 2 1 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 1 1 1 2 2 2 2 1 2 2 2 2 2 2 2 2 2 1 2 2 1 2 2 2 1 2 2 2 2 1 2 2 2 2 1 2 1 1 2 1 2 1 2 2 1 2 1 2 2 2 1 1 2 2 2 2 2 3 2 3 2 2 2 2 3 2 3 3 3 3 3 3 2 3 3 3 2 3 2 3 2 2 2 2 2 3 2 3 2 3 3 2 2 2 3 2 2 3 3 3 3 2 2 2 2 2 2 3 3 3 3 2 3 3 2 2 3 2 2 2 2 2 3 3 2 2 2 2 2 3 2 3 3 #NULL! #NULL! 3 2 2 2 2 3 2 3 2 3 2 3 3 3 2 2 2 3 3 2 3 2 3 2 2 2 2 2 2 2 2 #NULL! 3 2 3 3 2 2 2 2 2 2 3 2 3 3 2 2 2 2 2 2 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 1 2 2 1 1 2 2 2 1 1 2 2 2 2 2 2 2 2 1 2 2 2 2 1 2 1 1 1 2 1 1 1 2 1 1 1 1 1 1 2 1 2 2 1 2 2 2 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 2 1 1 1 1 2 1 1 1 2 1 1 1 1 2 2 2 2 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 2 1 1 1 2 1 1 1 1 2 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 1 1 1 2 1 1 1 1 1 1 1 1 "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Social communication difficulties are a diagnostic feature in autism. These difficulties are sometimes attributed, at least in part, to impaired ability in making inferences about what other people mean. In this registered report, we test a competing hypothesis that the communication profile of adults on the autism spectrum can be more strongly characterised by reduced confidence in making inferences in the face of uncertain information. We will test this hypothesis by comparing the performance of 100 autistic and 100 non-autistic adults on a test of implied meaning, using a test of grammaticality judgements as a control task. We hypothesise that autistic adults will report substantially lower confidence, allowing for differences in accuracy, than non-autistic adults on the test of implied meaning compared to the grammaticality test. In addition, we hypothesise that reduced confidence in drawing inferences will relate to the cognitive trait Intolerance of Uncertainty and self-reported social communication challenges. Finally, we will conduct exploratory analysis to assess the specificity of the communication profile of the autistic adults by comparing their performance to that of dyslexic adults, who might also be expected to experience challenges with language and communication.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Persistent challenges with social communication are a defining feature for the diagnosis of autism <ns0:ref type='bibr'>(American Psychiatric Association, 2013)</ns0:ref>. The underlying nature of these challenges remains unclear, although they are sometimes attributed to a core impairment in pragmatics (e.g. <ns0:ref type='bibr' target='#b4'>Baron-Cohen, 1988;</ns0:ref><ns0:ref type='bibr' target='#b34'>Rapin &amp; Dunn, 2003)</ns0:ref>. Pragmatics refers to the role of context in communication, including the ability to 'read between the lines' to infer intended meaning beyond what is explicitly stated <ns0:ref type='bibr' target='#b3'>(Baird &amp; Norbury, 2016)</ns0:ref>. However, empirical research suggests that pragmatic difficulties may be rather subtle in autistic people, and mostly attributable to language ability <ns0:ref type='bibr' target='#b25'>(Kalandadze, Norbury, Naerland &amp; Naess, 2018;</ns0:ref><ns0:ref type='bibr' target='#b27'>Loukusa &amp; Moilanen, 2009</ns0:ref>).</ns0:p><ns0:p>An alternative suggestion is that social communication difficulties are less the result of an impairment in pragmatics, but more impacted by cognitive preferences that differ between autistic and non-autistic people. We propose that a preference for certainty and explicit communication commonly occurs in autistic people, and that this trait may be a critical factor in the communication difficulties experienced by autistic people, as communicative situations often involve ambiguity, uncertainty and implied meanings.</ns0:p></ns0:div> <ns0:div><ns0:head>Intolerance of Uncertainty</ns0:head><ns0:p>A 'preference for certainty and explicit communication' may link to the widelyresearched cognitive trait, Intolerance of Uncertainty (IU). IU has been defined as a tendency to negatively evaluate uncertain situations and information <ns0:ref type='bibr'>(Shihata, McEvoy, Mullan &amp; Carleton,</ns0:ref> PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed 2016). We use the term 'Intolerance of Uncertainty' in line with previous research and intend to convey a value-neutral meaning in using it, as we recognise that high levels of IU may be an understandable, even adaptive, response where individuals have experienced mishaps in confusing situations. IU has mostly been investigated as a transdiagnostic construct that plays a central role in emotional disorders across the general population (see <ns0:ref type='bibr' target='#b38'>Shihata et al., 2016</ns0:ref> for a review), but it also seems especially relevant to autism, with autistic children and adults showing significantly elevated levels of the trait compared to the general population (e.g. <ns0:ref type='bibr'>Hwang et al., 2020;</ns0:ref><ns0:ref type='bibr'>Vasa et al., 2018)</ns0:ref>. IU has been closely linked to anxiety in autistic people <ns0:ref type='bibr' target='#b24'>(Jenkinson et al., 2020)</ns0:ref>, and also relates to core features of autism, including social difficulties, sensory sensitivities, insistence on sameness and repetitive behaviours (e.g. <ns0:ref type='bibr'>Hwang et al., 2020;</ns0:ref><ns0:ref type='bibr'>Vasa et al., 2018;</ns0:ref><ns0:ref type='bibr'>Wigham et al., 2015)</ns0:ref>.</ns0:p><ns0:p>A possible link between IU and communication in autistic people remains largely unexplored. However, there are reasons to believe that a link is plausible. First, inferential models of communication, such as Relevance Theory <ns0:ref type='bibr' target='#b43'>(Sperber &amp; Wilson, 1986)</ns0:ref>, propose that communication inherently involves uncertainty. Under Relevance Theory, language comprehension is not simply a process of 'understanding what the words mean', as there are often indeterminacies and ambiguities in uses of language; instead, the words are used as evidence by the listener in supporting a hypothesis about what the speaker probably means in the context, i.e. inferring intended meaning under uncertainty. Relevance Theory suggests that there is a gradient of uncertainty in communication. Sometimes, the listener can rely mostly on the explicit content of the utterance to compute the intended meaning, but in other situations there is a greater reliance on inferential processing to understand the speaker's probable meaning by integrating the utterance with contextual cues and world knowledge. Compare for instance the PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed utterances 'No, let's stay inside' and 'It's quite cold today' as responses to a suggestion to go outside. In the second example, the speaker communicates implicitly, leaving the listener to process the implicature that they would probably prefer to stay inside. In a previous study, we provided evidence for cognitive differences between autistic and non-autistic people in processing implicatures <ns0:ref type='bibr' target='#b51'>(Wilson &amp; Bishop, 2020b)</ns0:ref>. Crucially, it seemed that a cognitive preference for certainty and explicit communication, and not simply reduced ability, may account for some of the differences. Participants completed the Implicature Comprehension Test, which required individuals to listen to short conversational interchanges that are followed by a comprehension question to assess whether an implied meaning has been processed; test-takers responded with 'yes', 'no' or 'don't know'. Controlling for grammar/vocabulary ability, we found that autistic adults (N = 66) were 6.19, 95% <ns0:ref type='bibr'>CI [3.63,</ns0:ref><ns0:ref type='bibr'>10.39]</ns0:ref>, times more likely to select the 'don't know' rather than the correct response compared to non-autistic people (N = 118), and also 2.56, 95% <ns0:ref type='bibr'>CI [1.76,</ns0:ref><ns0:ref type='bibr'>3.77]</ns0:ref>, times more likely to choose the 'incorrect' response <ns0:ref type='bibr' target='#b51'>(Wilson &amp; Bishop, 2020b</ns0:ref>). Group differences were large, and performance on the test gave 76% sensitivity and specificity for differentiating between autistic and non-autistic groups. On the face of it, these results suggest that autistic people have difficulties inferring the gist of a speaker's meaning, as predicted by the 'central coherence' theory, which proposes that autistic people may show less tendency than non-autistic people to process information at a global level <ns0:ref type='bibr' target='#b15'>(Frith, 1989)</ns0:ref>.</ns0:p><ns0:p>However, in an alternative version of the test without a 'don't know' response, autistic individuals showed high accuracy for items for which they had selected 'don't know' first time round. This marked tendency to select 'don't know' when given a chance, but to process the inference as intended when constrained by the task, suggested reduced confidence in the face of Manuscript to be reviewed uncertain information and a preference for explicit communication. This could be due to possible difficulties around metacognition in autistic people, who may experience a mismatch between performance and confidence in their performance due to differences in self-monitoring. There is evidence that autistic people may show such a mismatch <ns0:ref type='bibr' target='#b17'>(Grainger et al., 2016;</ns0:ref><ns0:ref type='bibr'>Nicolson et al., 2019)</ns0:ref> although there is some concern about the replicability of these results <ns0:ref type='bibr' target='#b28'>(Maras et al., 2020)</ns0:ref>.</ns0:p><ns0:p>An alternative view would be that it is less an issue of metacognitive 'ability', and more about differences in personality/cognitive preference, with the well-replicated elevated levels of IU in autistic people (e.g. <ns0:ref type='bibr'>Hwang et al., 2020;</ns0:ref><ns0:ref type='bibr'>Vasa et al., 2018)</ns0:ref> accounting for this apparent preference for explicit communication observed in our previous study. In the present study, we aim to replicate this finding with more refined methods. In an adapted version of the Implicature Comprehension Test, individuals will respond using a 4-point scale of 'yes', 'maybe yes', 'maybe no', and 'no', allowing us to capture accuracy and confidence in the same measure. We hypothesise that confidence is likely to be affected specifically in a pragmatic language task (i.e.</ns0:p><ns0:p>where the individual needs to make flexible context-dependent inference about uncertain implied meanings), and not on tasks focused on more structural, codified aspects of language such as grammatical competence. As such, we present the Grammaticality Decision Test as a control task with a similar response format to the pragmatic task to test the specificity of any differences.</ns0:p><ns0:p>We propose that reduced confidence on the Implicature Comprehension Test will be a marker of IU in autistic people, and may be a more influential factor in the communication difficulties diagnostic of autism, as opposed to a 'deficit' in understanding social meanings. If this claim is borne out, it would have a couple of implications for psychological practice. <ns0:ref type='table'>2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:ref> Manuscript to be reviewed instruction in social skills, and reviews suggest modest effectiveness although there are questions about the extent to which skills transfer to daily life <ns0:ref type='bibr' target='#b16'>(Gates, Kang, &amp; Lerner, 2017;</ns0:ref><ns0:ref type='bibr' target='#b41'>Spain &amp; Blainey, 2015)</ns0:ref>. A focus on IU may be a useful alternative target. Existing cognitive interventions involve integrating psychoeducation and cognitive challenge techniques to target a client's beliefs about (un)certainty, and these have shown some effectiveness for treating mental health difficulties and particularly anxiety in the general population <ns0:ref type='bibr' target='#b38'>(Shihata et al., 2016)</ns0:ref>. It remains to be seen whether such interventions could be adapted to support autistic individuals with distressing communication experiences, although this is a promising possibility given that early studies suggest that such interventions may be feasible and acceptable for autistic groups <ns0:ref type='bibr' target='#b35'>(Rodgers et al., 2018;</ns0:ref><ns0:ref type='bibr'>Rodgers et al., 2016)</ns0:ref>. Second, if a cognitive preference for certainty is especially significant as an explanation for social difficulties, then it supports an autism-positive approach to intervention which focuses on awareness of cognitive differences across communities. In addition, if performance on the Implicature Comprehension Test is a sensitive marker of IU, that in itself might have clinical and research utility, since measurement of IU is currently limited to self-and informant-report questionnaires.</ns0:p><ns0:p>A remaining question is whether any differences observed on our tasks are specific to autism or might also be relevant to other neurodevelopmental diagnoses. This is certainly plausible in the light of dimensional models of neurodiversity, where features of autism, developmental language disorder, dyslexia, ADHD, etc., might show some overlap and exist as a continuum in the general population <ns0:ref type='bibr' target='#b45'>(Thapar, Cooper, &amp; Rutter, 2017)</ns0:ref>. To test the specificity of any cognitive differences observed on our tests, we will compare performance by autistic people to both a dyslexic and a general population sample. As neurodevelopmental conditions are often co-occurring, we view these three groups as defined less by a specific diagnostic label but rather Manuscript to be reviewed as varying along a communication continuum. As such, one group is defined by social communication differences potentially alongside co-occurring language/literacy impairments (the autism group), a second group by language/literacy impairments but no diagnosed social communication difficulties (the dyslexia group), and a final group without any communication, language or literacy related diagnosis. It is possible that dyslexic adults may show some difficulty on our pragmatic task (i.e. the Implicature Comprehension Test), as previous research has documented some limited evidence for pragmatic difficulties in dyslexic individuals (e.g. <ns0:ref type='bibr' target='#b8'>Cappelli et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b11'>Cardillo et al., 2018;</ns0:ref><ns0:ref type='bibr' target='#b21'>Griffiths, 2007)</ns0:ref>. An alternative possibility is that adults with dyslexia will show greater difficulty with tasks focused more on structural language skills compared to pragmatics. For instance, a meta-analysis has found that dyslexic adults perform less well on language measures, such as vocabulary, speeded naming, verbal memory and phonological processing, than people without a diagnosis of dyslexia, with moderate to large effect sizes <ns0:ref type='bibr' target='#b44'>(Swanson &amp; Hsieh, 2009)</ns0:ref>. Given that there is no clear reason to support one of these possibilities over the other, we will take a more exploratory approach with the dyslexic group to examine how they compare with autistic adults.</ns0:p><ns0:p>In summary, we propose the following hypotheses:</ns0:p><ns0:p>(1) Autistic adults will score lower on a pragmatic language task when responses are coded purely in terms of confidence (number of yes and no responses, regardless of polarity) than when responses are coded in terms of accuracy (with yes and maybe yes, and maybe no and no responses, combined according to polarity), compared to adults without any neurodevelopmental diagnosis, but will not show this same disparity between accuracy and confidence on a core language task.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p><ns0:p>(2)</ns0:p><ns0:p>The number of less confident responses (maybe responses) on the pragmatic language task, the score on the Intolerance of Uncertainty Scale, and selfreported social communication difficulties will significantly intercorrelate across the full sample.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods</ns0:head><ns0:p>Ethical approval for this project was granted on 30/03/2020 by the Medical Science Interdivisional Research Ethics Committee at Oxford University (Ref: R68518/RE001). The script for the power analysis and example materials are available on the Open Science</ns0:p><ns0:p>Framework here: https://osf.io/wk97s/. We are very happy to share full materials for this study, but to protect the validity of the items for future uses, we ask that researchers contact us to request a link to the full assessments. See further information on requesting access by following the link above.</ns0:p></ns0:div> <ns0:div><ns0:head>Power calculation</ns0:head><ns0:p>We determined power to detect the three-way interaction described in Hypothesis 1 using simulations. We used data reported in <ns0:ref type='bibr' target='#b51'>Wilson and Bishop (2020b)</ns0:ref> to estimate the likely size of fixed and random effects in the mixed model described in Data Analysis below. Using R package simr <ns0:ref type='bibr' target='#b19'>(Green &amp; MacLeod, 2016)</ns0:ref> we ran 1000 simulations with a sample size of 200 people (100 autistic, 100 non-autistic) and a significant three-way interaction was found in 9830 simulations, indicating that power was over 98% to detect our effect of interest at an alpha level of .05.</ns0:p><ns0:p>Effectively, this allows us to detect a significant difference where approximate Cohen's d values Manuscript to be reviewed in favour of the non-autistic group are 0.70 and 1.10 for the implicature accuracy and confidence variables and 0.20 for the grammar variables, as suggested by our previous data. Allowing for exclusion of up to 10% of participants during the outlier exclusion phase described in Data Analysis, power remains very high (98% in a sample of 180). For hypothesis 2, a sample of 200 is powered at over 99% to detect a correlation of .3.</ns0:p></ns0:div> <ns0:div><ns0:head>Participants</ns0:head><ns0:p>We will recruit individuals with autism, dyslexia, and no neurodevelopmental diagnosis.</ns0:p><ns0:p>Based on the power calculation, we will recruit 100 autistic adults and 100 adults without a neurodevelopmental diagnosis in order to run the confirmatory analysis. In addition, we will aim to recruit 50 dyslexic adults as a clinical control group for exploratory analysis. All participants will meet the following eligibility criteria: (i) age of 18 years or over, (ii) native-level fluency in English, (iii) no history of acquired brain injury, (iv) no significant uncorrected sensory impairment, and (v) access to a computer with internet and audio.</ns0:p><ns0:p>Individuals will be recruited into three groups defined by communication and language/literacy problems. One group will be recruited on the basis of a clinical diagnosis of Manuscript to be reviewed individuals must score below the clinical threshold of 6 on the Autism Spectrum Quotient (AQ)</ns0:p><ns0:p>and at 6 or above on the Reading Scale of the Adult Reading Questionnaire (ARQ); this latter score translates to over 1.5 SDs above the mean in individuals not self-reporting dyslexia in the original validation study <ns0:ref type='bibr' target='#b40'>(Snowling, Dawes, Nash, &amp; Hulme, 2012</ns0:ref>). Individuals will be recruited through charitable organisations such as the British Dyslexia Association and social media.</ns0:p><ns0:p>Other neurodevelopmental diagnoses will be noted but will not be grounds for exclusion from these groups. A third group will have no neurodevelopmental diagnosis, and will be recruited via the online participant platform, Prolific (https://prolific.co). Individuals will be excluded from this third group if they score above threshold on either the AQ or ARQ (i.e. above 6 on either) and if they have ever been diagnosed with: a global or specific learning disability, attention deficit hyperactivity disorder, dyspraxia/developmental coordination disorder, a genetic variation (such as Down's syndrome or Fragile-X) or a neurological condition (such as epilepsy).</ns0:p></ns0:div> <ns0:div><ns0:head>Procedure</ns0:head><ns0:p>The study will be presented online using Gorilla, the online platform for behavioural experiments and surveys (https://gorilla.sc/). Individuals complete an online set of tasks and questionnaires in one sitting at a time and place of their choosing. After providing informed written consent to participate, individuals will complete a Study Questionnaire (please see the OSF link) on which they will be asked to report on demographics and any neurodevelopmental diagnoses. Then they will complete questionnaires/tasks in two sections. The first section will include the experimental tasks required for the hypothesis-testing, and the second will include several brief measures for characterising the sample. The two experimental tasks will be randomized between participants, and all other measures will be administered in the order set out below.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Measures Section 1: Experimental tasks assessing ability and confidence with pragmatics and core language</ns0:head><ns0:p>Implicature Comprehension Test-2 (ICT-2). In this test of pragmatic language comprehension, participants complete an adapted version of the Implicature Comprehension Test <ns0:ref type='bibr' target='#b47'>(Wilson &amp; Bishop, 2019)</ns0:ref>. There is a sequence of 56 videos, each approximately 8 s long, consisting of a conversational adjacency pair between two characters: the first character asks a closed question (eliciting a 'yes' or 'no' answer) and the second character produces a short answer but does not say yes or no. Each utterance is between 5 and 10 words in length, grammatically simple, and age of acquisition of the words does not exceed middle primary school level. Following the dialogue, the participant hears a comprehension question directly based on the structure of the first character's question. The participant answers the question on a 4-point scale (yes, maybe yes, maybe no, no) by clicking buttons arranged horizontally on the screen. This is a timed task, with a time limit of 10 s for a response from the offset of the question. There are two item types: implicature and explicit-response. Utterance length and psycholinguistic variables (word frequency, word age-of-acquisition and word concreteness) are controlled across the two item types.</ns0:p><ns0:p>For 40 videos, the second character's answer is indirect, and the participant needs to</ns0:p><ns0:p>process implicature to answer the comprehension question appropriately. Example: Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Answer: No</ns0:head><ns0:p>Half of the comprehension questions are correctly answered by 'yes' and half by 'no'.</ns0:p><ns0:p>There are two measured variables: total accuracy (collapsing yes and maybe yes, and maybe no and no responses, according to polarity) out of 40 and total confidence (number of yes and no responses, regardless of polarity) out of 40.</ns0:p><ns0:p>Alongside the implicature items, there are 16 explicit-response items where the second character's answer is more explicit. In these items, the speaker intends to convey uncertainty explicitly, whereas in the implicature items, the uncertainty is in the mind of the listener.</ns0:p><ns0:p>Example:</ns0:p><ns0:p>Character 1: Will we get there by seven?</ns0:p><ns0:p>Character 2: Mmm, yes maybe, I think we're near.</ns0:p><ns0:p>Comprehension Question: Will they get there by seven?</ns0:p><ns0:p>Answer: Maybe yes For these items, the comprehension questions will encourage the participant to use the full scale, with four questions each correctly answered by 'yes', 'maybe yes', 'maybe no' and 'no'.</ns0:p><ns0:p>There is one measured variable: total accuracy out of 16. Manuscript to be reviewed grammatical, indicating 'yes', 'maybe yes', 'maybe no' and 'no' as their answer by clicking buttons arranged horizontally on the screen, as in the ICT-2. After offset of the sentence, participants have 10 seconds to give their response. This is a norm-referenced questionnaire measuring self-reported communication challenges.</ns0:p></ns0:div> <ns0:div><ns0:head>Grammaticality Decision</ns0:head><ns0:p>Participants will be presented with the pragmatic language scale (22 items). For each item, participants identify how frequently certain communication behaviours apply to them on a 4point scale from 'less than once a week (or never)' to 'several times a day (or all the time)'. An example item is 'People tell me that I ask the same question over and over'.</ns0:p><ns0:p>Adult Reading Questionnaire (ARQ) Reading Scale (Snowling, Dawes, Nashe, &amp; Hulme, 2012). Self-reported reading difficulties will be measured using this 5-item questionnaire. In the original validation study, it showed good construct validity (correlating with observed literacy ability at -.67) and, along with self-reported dyslexia status, discriminated with 88% accuracy in identifying those with weaker literacy skills.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Young &amp; Keith, 2020).</ns0:p><ns0:p>Synonyms Test <ns0:ref type='bibr' target='#b47'>(Wilson, &amp; Bishop, 2019)</ns0:ref>. General verbal ability will be measured using this 25-item test of vocabulary knowledge. Participants select which of five written words is synonymous with a target word, under a 12-second time limit. The original version of the GDT and this task showed a moderate correlation in both autistic and non-autistic samples, suggesting they are overlapping measures of core language ability <ns0:ref type='bibr' target='#b47'>(Wilson &amp; Bishop, 2019</ns0:ref><ns0:ref type='bibr' target='#b50'>, 2020a)</ns0:ref>. statements about uncertainty, ambiguous situations, and the future. They rate how closely each statement relates to them on a 5-point scale from 'not at all characteristic of me' to 'entirely characteristic of me'. An example item is: 'When I am uncertain, I can't function very well.'</ns0:p></ns0:div> <ns0:div><ns0:head>Intolerance of Uncertainty</ns0:head></ns0:div> <ns0:div><ns0:head>Data Analysis</ns0:head><ns0:p>Individuals will be excluded from the dataset if they have an outlying score for either (a)</ns0:p><ns0:p>accuracy on the GDT or the positive control items of the ICT-2 (b) total number of timeouts across the ICT-2 and GDT. Outliers will be defined according to the method of <ns0:ref type='bibr'>Hoaglin and Inglewicz (1987)</ns0:ref>: more than 2.2 times the interquartile range below the first quartile. In previous Manuscript to be reviewed work, these criteria led to exclusion of approximately 5% of participants, and captured individuals scoring below approximately 50% on the GDT and 70% on the positive control items of the original version of the ICT <ns0:ref type='bibr' target='#b50'>(Wilson &amp; Bishop, 2020a</ns0:ref>).</ns0:p><ns0:p>Data will be analysed in R (R Core Team, 2019). After exclusions, total scores on the two experimental tasks for the groups with autism and no neurodevelopmental diagnosis will be turned into long format, and each participant's total will be coded for task (ICT-2 or GDT), group (autistic or no neurodevelopmental diagnosis), response (accuracy or confidence) and participant. We will run a mixed effects linear regression using the lme4 R package <ns0:ref type='bibr' target='#b5'>(Bates, Maechler, Bolker, &amp; Walker, 2015)</ns0:ref>. The model will include three fixed effects (task, group and response) and the interactions between these, as well as a random effect (participant). The significance level of the three-way interaction will offer a test of Hypothesis 1. We will also compute correlations between confidence on the ICT-2, self-reported communication challenges on the CC-SR and total score on IUS-12 across the full sample; this will test Hypothesis 2 and represents a dimensional analysis of the relationship between communication difficulties and sensitivity to uncertainty. Table <ns0:ref type='table'>1</ns0:ref> shows a summary of our planned analyses, linking research questions, hypotheses, tests and power calculations.</ns0:p><ns0:p>In exploratory analysis, we will examine how the dyslexic group compares to the autistic group and the group without a neurodevelopmental diagnosis on the ICT-2 and GDT in terms of accuracy and confidence.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>First, it would suggest that interventions targeting IU may be useful for autistic people wanting support with communication challenges. Current interventions for communication focus on explicit PeerJ reviewing PDF | (</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head /><ns0:label /><ns0:figDesc>autism; participants will need to declare where, by whom and what label was used for their diagnosis on the Study Questionnaire. For inclusion, the diagnosis must have been made in a clinical service by appropriately trained individuals, such as clinical psychologists, psychiatrists or developmental paediatricians. We will recruit autistic individuals through Autistica, the research network for families and individuals with autism, as well as support groups arranged privately and by the National Autistic Society, and through social media. A second group will include individuals reporting dyslexia or specific reading difficulties. For inclusion in this group, PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head>Character 1 :</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Did you hear what the police said? Character 2: There were lots of trains going past. Comprehension Question: Did he hear what the police said? PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>Test (GDT; based on<ns0:ref type='bibr' target='#b47'>Wilson &amp; Bishop, 2019)</ns0:ref>. In this test of core language ability, participants listen to a sequence of 50 sentences and decide if the sentence is grammatical and well-formed or not. Half the sentences are grammatical. Grammatical violations represent mistakes that native speakers would not tend to make, such as using an incorrect verb form (e.g. I went out after I have eaten dinner) or atypical placing of adverbs (e.g.If you can't find it, I can send again the letter). Participants are asked whether the sentences are PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>Autistic traits will be measured using this 10-item version of the Autism Spectrum Quotient (AQ). In the original validation study, the measure had 85% correct discrimination between almost 450 autistic adults and over 800 control adults. The National Institute for Health and Care Excellence (NICE, 2012) recommend use of the questionnaire for identifying individuals for comprehensive autism assessment. A clinical cut-off of 6 or more is taken as indicating possible autism.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>Scale (IUS-12; Carleton, Norton, &amp; Asmundson, 2007). In this self-report measure of intolerance of uncertainty, participants are presented with 12</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52215:1:1:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>International Cognitive Ability Resource (ICAR) Sample Test (Condon &amp; Revelle, 2014).</ns0:head><ns0:label /><ns0:figDesc>This is an open-access test of general cognitive ability, which requires participants to complete 16 items across four item types: matrix reasoning, verbal reasoning, three-dimensional</ns0:figDesc><ns0:table><ns0:row><ns0:cell>rotation, and letter-number sequences. The ICAR sample test has good internal consistency</ns0:cell></ns0:row><ns0:row><ns0:cell>(alpha = .81), and good convergent validity (correlating at approximately .8 with commercial IQ</ns0:cell></ns0:row><ns0:row><ns0:cell>measures when correcting for reliability and restriction of range; Condon &amp; Revelle, 2014;</ns0:cell></ns0:row></ns0:table></ns0:figure> </ns0:body> "

"Dear Dr Borghi, Thank you for sending through comments on our paper. We are very appreciative of the positive and constructive feedback from yourself and your reviewer, and we provide point-by-point responses to all your comments below. Editor’s Comments I have now received one independent review on your report, which you submitted to PeerJ, I had the possibility to see three reviews of a previous submission of this paper and have carefully read it myself. As you will see, the reviewer is positively oriented toward your report, and I share his/her opinion. The reviewer simply asks to make the materials available, and I also think that you could refer more extensively to literature on metacognition, directly relevant to the issue of confidence and tolerance for uncertainty. We agree that it would be sensible to refer to the literature on metacognition, so have provided some brief discussion on p4 lines 24-25 and p5 lines 1-3. We are also in agreement about the importance of sharing materials in the interests of open and reproducible science. However, we are cautious about making the materials publicly available as it may reduce their validity as research and possible clinical measures if individuals have seen them prior to assessment. As a compromise, we propose having them available in a closed-access repository to which researchers can request a link, which we mention on p8 lines 7-10. For now, we have uploaded the materials to private components on the Open Science Framework (OSF). We share links to these private components below for yourself and the reviewer to see the materials. I was also wondering – but this is a personal curiosity, feel free to avoid referring to it in the revised version - whether this difficulty with uncertainty might be related to the difficulty of autistic people with abstractness. Finally, I think that the impactful question of how to intervene to improve the level of acceptance of uncertainty should be addressed. We share your interest in whether challenges with abstractness may contribute to inferencing difficulties, and see links between this suggestion and the ‘central coherence’ hypothesis of the cognitive differences in autism. There is some discussion of this hypothesis on p4 lines 15-18. We have added to our discussion of interventions on p6 lines 1-9 However, we do need to be cautious about the question of intervention for two reasons. First, while there is much to suggest that ‘intolerance of uncertainty’ is an important maintaining factor in mental distress, there is relatively little intervention-based research targeting this construct. In addition, there is very little research into the effectiveness of any cognitive therapies in autism. All the same, we agree that there is some worth in exploring possible impacts, as you suggest. Reviewer’s Comments The article is very well written and easy to follow. Appropriate literature is cited. Hypotheses are clearly stated and follow logically from the introduction. Link to OSF given. My only major suggestion is to share the full set of items so that it is easier to understand what is being tested. We sympathise with this suggestion, and would ideally share all items publicly. However, as explained in our response to the Editor above, we are concerned that sharing items in this way may impact on test validity. Therefore, we hope you will agree that it would be an appropriate compromise to have the items available by request on OSF, as described above. We do invite yourselves to take a look at the full tests using the links above. It might also help to expand on what is proposed to be involved in processing an inference/ implicature by giving some theoretical background. Reference to relevant theoretical frameworks would be helpful as there is some debate about when inferences are and are not necessary and what they involve in different cases (e.g., Kintsch, 1998, Comprehension). Lack of a theoretical statement on this and lack of access to materials made me wonder what was in fact being tested. Our approach to implicature/inference is based on Relevance Theory of Sperber & Wilson (1986), and we provide some of this theoretical background on p3 lines 13-25. If a value-neutral term is intended would 'preference for certainty' be better or is 'intolerance of uncertainty' preferred as a conventional expression?  As ‘Intolerance of uncertainty’ is the term used in the literature, we feel it is best to stick with it so that it is clear how our research relates to previous work. However, we have included a sentence about intending a value-neutral meaning when we use the term. The section in the intro on the inclusion of a dyslexic group - and the selection of this group rather than others - was a little vague at first but became clearer in the method around line 172. Perhaps this justification could be given earlier. Thank you for this suggestion. We have moved the justification to p6 lines 22-25 and p7 lines 1-3. The proposed research is original primary research within the aims and scope of the journal. With the caveat above, the research question is well defined, the means of testing it are justified and the results are likely to advance knowledge. Methodological and analytic standards are high. Detail is sufficient to replicate with the exception of provision of key inference test materials. Unless there is a good reason not too, these materials should presumably be shared and this would be my only major request for revision. This is all the more important given how difficult it is to define inference, to come up with good tests of it and to be clear why they are a test of inference (perhaps even this is a question of individual differences to some extent, with some arriving at a given 'inference' more through spreading activation and others by making a more proactive effort after meaning). Of course, this is work in progress and some of this has to be left underspecified for now. The purpose of sharing materials would not be to scrutinise every item before the test is run, but to help everyone to make sense of what is being tested and what is found. Thank you for your positive feedback about the experimental design. Please see our responses above regarding the sharing of materials. Having said this, I had specific notes about some test items. For example, in the video about free seats, there are free seats available in the background of the picture, so the answer is also available visually. We agree that the visual cue makes this item less ideal in a test of inferencing so have redone the visuals. We have checked other items but did not spot any similar issues. I also wondered whether there were specific reasons why people were likely to struggle with different items - for example, reasons relating to real world knowledge and ability to generate world models that would make the speech observed plausible. It appears this could relate to uncertainty in several ways. For example, someone could be uncertain because they struggled to generate plausible real world contexts that would justify the answer or because they could generate several scenarios and weren't sure which was more likely to hold (e.g., 'I heard the police say there were a lot of trains' or 'there were so many trains, I couldn't hear what the police said').  This is a very good point, and it will be interesting to consider this when we discuss results. It will likely be difficult to distinguish between these possibilities, although our results will indicate whether participants are uncertain because they are unsure what an appropriate inference would be (in which case we would expect low accuracy and low confidence) or whether they tend to form a normative inference but are less confident in that (where we would expect higher accuracy). However, we appreciate this is not quite the point you are making, and identifying possible causes of uncertainty is likely to be difficult to disentangle using our methods. Missing word on line 179 > the online platform Thank you for pointing out this error. This has been corrected. Perhaps clarify that for the literal items, the speaker intends to convey uncertainty. For the implicature items, the uncertainty is in the mind of the listener. This is a lovely way of conceptualising the difference between the item types, and we have adopted this description in our paper on p12 lines 10-11. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Evidence was brought forward in England and the USA that Black, Asian, Latino and Minority Ethnic people exhibit higher mortality risk from COVID-19 than White people.</ns0:p><ns0:p>While socioeconomic factors were suggested to contribute to this trend, they arguably do not explain the range of the differences observed, allowing for possible genetic implications. Almost concurrently, the analysis of a cohort in Chinese COVID-19 patients proposed an association between the severity of the disease and the presence of the minor allele of rs12252 of the Interferon-induced transmembrane protein 3 (IFITM3) gene. This SNP, together with rs34481144, are the two most studied polymorphisms of IFITM3 and have been associated in the past with increased severity in Influenza, Dengue, Ebola, and HIV viruses. IFITM3 is an immune effector protein that is pivotal for the restriction of viral replication, but also for the regulation of cytokine production. Following up on these two developments in the ongoing SARS-CoV-2 pandemic, the present study investigates a possible association between the differences in mortality of ethnic groups in England and the combined haplotypes of rs12252 and rs34481144. The respective allele frequencies were collected for 26 populations from 1000 Genomes Project and subgroups were pooled wherever possible to create correspondences with ethnic groups in England. A significant correlation (r=0.9687, p= 0.0003) and a striking agreement was observed between the reported Standardized Mortality Ratios and the frequency of the combined haplotype of both reference alleles, suggesting that the combination of the reference alleles of the specific SNPs may be implicated in more severe outcomes of COVID-19. This study calls for further focus on the role of IFITM3 variants in the mechanism of cellular invasion of SARS-CoV-2, their impact in COVID-19 severity and their possible implications in vaccination efficacy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Emerging scientific evidence from international (Kirby, 2020) and UK ( <ns0:ref type='bibr'>Aldridge et al., 2020)</ns0:ref> COVID-19 patient reports and death records, indicate a disproportionate effect of the novel coronavirus on ethnic minorities. According to CDC (CDC, 2020), Black, Asian and Minority Ethnic (BAME) people are at higher risk of death from COVID-19. Importantly, an Indirect Standardization of NHS mortality data in England ( <ns0:ref type='bibr'>Aldridge et al., 2020)</ns0:ref>, revealed that the adjusted for age and region Standardized Mortality Ratios (SMRs), were highest in Black African, Black Caribbean, Pakistani, Bangladeshi, and Indian minority ethnic groups. On the contrary, White Irish and White British ethnic groups exhibited a significantly lower risk of death. Similarly, in the USA <ns0:ref type='bibr' target='#b17'>(Garg et al., 2020)</ns0:ref>, preliminary data compiled from hospitals in 14 US states, confirmed the UK study outcomes, showing that African Americans are also disproportionately affected by COVID-19. Specifically, African Americans represented 33% of COVID-19 hospitalizations, despite only making up 18% of the total population studied. In a subsequent analysis, among COVID-19 deaths in New York City, for which race and ethnicity data were available, death rates from COVID-19 among black or African Americans and Hispanic or Latinos were substantially higher than that of white or Asian people <ns0:ref type='bibr' target='#b17'>(Garg et al., 2020)</ns0:ref>.</ns0:p><ns0:p>Several reasons have been proposed to explain these ethnic discrepancies in COVID-19 mortality risk arising from these preliminary studies. Chronic pre-existing conditions, such as Cardio Vascular Diseases (CVD), diabetes, hypertension, obesity, etc. are more common in minorities compared to Caucasian populations and have all been associated with adverse outcomes in COVID-19 (Centers for Disease Control and Prevention, 2020; <ns0:ref type='bibr' target='#b27'>Kirby, 2020)</ns0:ref>. However, race disparities in those diseases are not large enough to fully explain the COVID-19 death disparity <ns0:ref type='bibr'>(Aldridge et al., 2020)</ns0:ref>. Factors such as housing and living conditions, use of public transportation, lack of regular access to primary health, and occupation-related differences that prohibit the work from home, or require more frequent and/or close social contact, may have all played an important role in producing disproportionate death rates among BAME groups <ns0:ref type='bibr'>(Aldridge et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b27'>Kirby, 2020;</ns0:ref><ns0:ref type='bibr' target='#b33'>Niedzwiedz et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b26'>Khunti et al., 2020)</ns0:ref>. Nevertheless, it is suggested that inequalities in socioeconomic status parameters do not seem to adequately explain the range of differences, and in some instances, the extreme variations observed among ethnic minorities in mortality rates from COVID-19 infection <ns0:ref type='bibr' target='#b27'>(Kirby, 2020)</ns0:ref>.</ns0:p><ns0:p>As the importance of genetic polymorphisms (SNPs) in the modulation of individual susceptibility to, and severity of, infectious diseases has been well established <ns0:ref type='bibr' target='#b10'>(Chapman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b51'>Zhao et al, 2018)</ns0:ref>, we turned our focus to two very highly studied polymorphisms of the interferon-induced transmembrane protein 3 (IFITM3) gene: rs12252 and rs34481144. IFITM3 encodes an immune effector protein that is pivotal for restriction of viral replication <ns0:ref type='bibr'>(Brass et al., 2009)</ns0:ref> of many enveloped RNA viruses including HIV-1, influenza A virus (IAV), Ebola and Dengue virus ( <ns0:ref type='bibr'>Brass et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b16'>Feeley et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b22'>Huang et al., 2011;</ns0:ref><ns0:ref type='bibr'>Everitt et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b11'>Compton et al., 2014)</ns0:ref>. IFITM3 has also been demonstrated to affect severity of infection and improve the host cellular defenses against viruses ( <ns0:ref type='bibr'>Brass et al., 2009;</ns0:ref><ns0:ref type='bibr'>Everitt et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b11'>Compton et al., 2014)</ns0:ref>. Interestingly, IFITM3 has also been shown to act as a regulator of antiviral immunity that controls cytokine production to restrict viral pathogenesis, in CMV <ns0:ref type='bibr' target='#b39'>(Stacey et al., 2017)</ns0:ref> and Sendai virus <ns0:ref type='bibr' target='#b25'>(Jiang et al., 2017)</ns0:ref>. This finding is particularly important since cytokine storm in influenza can lead to a rapid progression of the infection in humans <ns0:ref type='bibr' target='#b42'>(Wang et al., 2014)</ns0:ref> and the same observation is also apparent in COVID-19 severe and deadly cases <ns0:ref type='bibr' target='#b18'>(Giamarellos-Bourboulis et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b4'>Blanco-Melo et al., 2020)</ns0:ref>. Moreover, IFITM3 was found to be explicitly upregulated in SARS-CoV-2 infected cells <ns0:ref type='bibr' target='#b4'>(Blanco-Melo et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b19'>Hachim et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b20'>He et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The minor allele of rs12252 (C in minus, or G in plus strand orientation) has been associated with rapid progression of acute HIV infection <ns0:ref type='bibr' target='#b47'>(Zhang et al., 2015)</ns0:ref>, with the severity of influenza <ns0:ref type='bibr' target='#b46'>(Zhang et al., 2013)</ns0:ref> and recently with COVID-19 severity <ns0:ref type='bibr' target='#b48'>(Zhang et al, 2020)</ns0:ref>. The minor allele of rs34481144 (A in minus, or T in plus strand orientation) was previously found to be correlated with increased severity of IAV infection <ns0:ref type='bibr' target='#b3'>(Allen et al., 2017)</ns0:ref>. Moreover, the minor allele of rs34481144 is also associated with enhanced methylation on the IFITM3 promoter of CD8+ T cells, and general transcriptional repression of the broader locus surrounding IFITM3, which includes several genes known to be involved in host responses to viral infection <ns0:ref type='bibr' target='#b43'>(Wellington et al., 2019)</ns0:ref>. SARS-CoV-2 uses primarily the ACE2 receptor as main point of entry and the host cell serine protease TMPRSS2 for viral spike (S) protein priming <ns0:ref type='bibr' target='#b21'>(Hoffmann et al., 2020)</ns0:ref>. Severe acute respiratory syndrome coronavirus (SARS-CoV), which also uses ACE2 as a receptor, has been shown to be restricted more efficiently by IFITM1 than by IFITM3, presenting a different restriction pattern than IAV <ns0:ref type='bibr' target='#b22'>(Huang et al., 2011)</ns0:ref>. Interestingly, it was recently shown that TMPRSS2 is specifically allowing evasion of IFITM3 restriction for bat SARS-Like WIV1 coronavirus <ns0:ref type='bibr' target='#b52'>(Zheng et al., 2020)</ns0:ref>, opening the possibility for a similar mechanism in the case of SARS-CoV-2. Further potential involvement of IFITM3 in COVID-19 outcome was revealed in the context of syncytial pneumocytes in severe cases with extensive lung damage, where it was suggested that the cellular location of IFITMs 1-3 could be playing a role in syncytia formation <ns0:ref type='bibr'>(Buchrieser et al., 2020)</ns0:ref>. Indeed, the accumulation of many direct and indirect layers of evidence linking IFITM3 with COVID-19 severity, has also led to explicit calls for further investigation of the role of this highly relevant first-line of cellular defense protein <ns0:ref type='bibr' target='#b49'>(Zhao, 2020)</ns0:ref>. Following up to the analysis of COVID-19 NHS mortality data in BAME groups ( <ns0:ref type='bibr'>Aldridge et al., 2020)</ns0:ref>, the purpose of the present study was to investigate a possible association between the stand-alone and combined frequencies of the alleles of the IFITM3 gene variants rs12252 and rs34481144, with COVID-19 standardized mortality ratio of ethnic groups in England.</ns0:p><ns0:p>The sole incentive of the current research is to help improve the existing and future treatment protocols for severe COVID-19 patients, and in no case to provide DNA-based arguments that may be used to mask existing social inequalities or racism.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods</ns0:head><ns0:p>The Standardized Mortality Ratios (SMR) of ethnic groups in England, adjusted for age and region, were adopted from the study by <ns0:ref type='bibr'>Aldridge et al. in</ns0:ref> the exact form in which they were presented <ns0:ref type='bibr'>(Aldridge et al., 2020)</ns0:ref>. Specific dataset details, including age, region and ethnicity information, are available at the repository address of the respective publication (https://discovery.ucl.ac.uk/ id/eprint/10096589). The rs12252 (A&gt;G) and rs34481144 (C&gt;T) allele and haplotype frequencies were collected for all available 1000 Genomes Project ancestral populations, from LDlink (LDhap tool : https://ldlink.nci.nih.gov/?tab=ldhap) , specifically 5 major groups, i.e. African (AFR), Ad Mixed American (AMR), European (EUR), East Asian (EAS) and South Asian (SAS), comprising 26 subgroups in total <ns0:ref type='bibr' target='#b31'>(Machiela et al., 2015)</ns0:ref> (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>The plus orientation for the reference and minor alleles was retained throughout this analysis, for better data handling and in compliance with dbSNP. Combined rs12252_rs34481144 haplotypes were defined as A_C (H1), G_C (H2), A_T (H3), while G_T haplotype was not represented at all in the data Rankings were examined by sorting all populations by individual reference allele (rs12252:A, rs34481144:C) and by combined haplotype frequency ratios (rs12252_rs34481144: h1_ratio=A_C/(A_T+G_C), h2_ratio=G_C/(A_C+A_T), h3_ratio=A_T/(G_C+A_C)) (Figure <ns0:ref type='figure' target='#fig_1'>1a</ns0:ref>), and subsequently compared visually to the reported Standardized Mortality Ratios (SMR) of ethnic groups in England (Figures <ns0:ref type='figure' target='#fig_1'>1b and 2</ns0:ref>). Additionally, an attempt was made to correlate directly the two rankings, i.e. SMR and IFITM3 haplotype frequencies by specific reported ethnic subgroup. UK demographics sources were therefore consulted (Office for National Statistics, UK, 2011; <ns0:ref type='bibr' target='#b9'>Chanda &amp; Ghosh, 2012)</ns0:ref> in order to pool, wherever possible, the ancestral subgroups to the reported ethnic groups in England. With all reservations tied to the inevitable discrepancies of this type of simplified socio-genetic correspondences, the following pools were formed: [AFR-YRI, AFR-LWK, AFR-GWD, AFR-MSL, AFR-ESN]&gt;'African', [SAS-STU, SAS-GIH, SAS-PJL]&gt;'Indian', [EAS-CDX, EAS-CHS, EAS-CHB]&gt;'Chinese', [EUR-CEU, EUR-IBS, EUR-TSI, EUR-FIN]&gt;'White Other' (see Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref> for full subgroup descriptions and Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref> for pooled subgroups and correspondences to ethnic minorities). A pool for the reported Pakistani group failed to form from ancestral populations, as the Punjabi (SAS-PJL), being the only related subgroup, account roughly for just 45% of Pakistan's demographics, while in London the community includes comparable numbers of Punjabis, Pathans and Kashmiris, with small communities of Sindhis and Balochis (Department for Communities and Local Government, UK, 2009). Moreover, the Punjabi form also a considerable part of Indians' pool (at least 40% of Delhi's total population), therefore a single-ended direct correspondence between Punjabi and British Pakistani was not warranted in this case. Indian Telugu (SAS-ITU) were not included in Indians' pool, as no demographic report was suggestive of comparable numbers to the other 3 included subgroups, for people of Indian origin in England. The same rationale applied for the non-inclusion of AFR-ASW (Americans of African Ancestry in SW USA) in the African pool. The haplotype frequencies were simply averaged within pooled groups, and both the ratios and SMR were normalized to the White British result (represented uniquely by EUR-GBR subgroup) (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>We extracted rs12252 and rs34481144 allele frequencies of various ethnic groups from the 1000 Genomes Project, in order to examine whether the distribution of any one of the combined haplotypes is directly correlated with the reported SARS-CoV-2 related SMR in England. Two levels of SMR vs haplotype comparisons were applied: first, we compared at the level of major groups (i.e., un-pooled comparison: EAS vs SAS vs EUR vs AFR, see Figure <ns0:ref type='figure'>2</ns0:ref>), second, at the level of ethnic subgroups, wherever possible (pooled comparison, see Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>). The ranking that visually appeared in line with the reported SMR (Figure <ns0:ref type='figure'>2</ns0:ref>) was produced by the A_C/(A_T+G_C) ratio (h1). For the evaluation of rank correlation we considered two scenarios; in both cases the correlation proved significant. First, we considered all 10 groups that are available with an SMR (Figure <ns0:ref type='figure'>2</ns0:ref>) and we observed an almost perfect alignment with h1 ratios, which corresponds to a permutation event of 10 items, with 1 / 3,628,800 chance to occur randomly, corresponding to p = 3*10 -7 (5&#963;). Secondly, we considered only 5 aligned items, specifically the sequence of (a) African groups, followed by (b) South Asian, followed by (c) White Non-British, (d) Chinese, and finally (e) White British, a permutation event with 1 / 120 chance to occur randomly, or p = 0.008. It is noteworthy, that EUR-IBS (Iberian Population in Spain) and EUR-TSI (Toscani in Italia), representative of two countries that suffered higher death rates than other European countries, share the highest h1 ratio between all European subgroups. Subsequently, in order to assess the potential strength of the theorized correlation, the rankings of pooled h1 ratios and SMR, per group, were traced one versus the other with Pearson correlation r=0.9687, p=3*10 -4 (&gt;3.5&#963;) (Figure <ns0:ref type='figure'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The calculated level of correlation appears to be remarkable, considering the possible discrepancies in the pooling of the available ancestral groups, but also the expected multiparametric causes of the observed COVID-19 SMR in England's ethnic groups (as previously described, potentially involving prior health status, income level, household density, behavioral biases, questionable attribution of death to COVID-19, etc.). On one hand, the alignment of ethnic group rankings between SMR and un-pooled h1 ratio, is less than 1/120 probable to occur randomly at the worst case, and less than 1 in 3.6 million probable to occur randomly at the best case. In other words, if one hypothesizes that the reported SMR rankings are solely due to socioeconomic factors, then one would conclude that socioeconomic factors would be in perfect alignment with h1 ratios. The possibility of the above to occur seems highly unlikely, thus pointing to the fact that differential allele frequencies play a potentially important role in the reported SMR.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52069:1:1:NEW 19 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>On the other hand, the level of correlation (&gt;3.5&#963;) between SMR and pooled h1 ratio, confirms the previous alignment and appears strong enough to suggest a possible causal link, albeit in this case, the pooling process may have introduced some discrepancies. The potential introduction of pooling discrepancies is expected and is impossible to quantify given the available data. However, the purpose of the pooled analysis was mainly to reinforce the primary association between IFITM3 and COVID-19 severity rather than to (indirectly) infer causality. Considering the parametric uncertainty of the pooled analysis it is probably inadequate to suggest a causal link, based on the strength of the observed correlation alone. However, taken together, this set of results constitutes a clear and valid starting point for designing further investigations regarding the role of IFITM3 in COVID-19 severity and appears as one more piece of evidence towards this direction. Therefore, the main point of this analysis is that the two examined SNPs should preferably henceforth be studied under a combined haplotype and not separately, as was performed so far for SARS-CoV-2, and a great variety of other viruses.</ns0:p><ns0:p>Before proceeding to the discussion of the implications of the above conclusion, it is interesting to also view the present observations in the context of data from the USA. A preliminary analysis of death rates from COVID-19 in New York City shows 92.3 deaths per 100,000 population among black or African American people, followed by Hispanic or Latino people (74.3), then by white (45.2) or Asian (34.5) people <ns0:ref type='bibr' target='#b27'>(Kirby, 2020)</ns0:ref>. The same trend was clearly displayed in the initial ranking by h1 ratio (Figure <ns0:ref type='figure' target='#fig_1'>1a</ns0:ref>), with American populations (AMR) occupying the middle of the chart, between African and European / Asian populations.</ns0:p><ns0:p>Interestingly, the reported lower death rate of Asian people compared to white people in the New York data, which represents an inversion of the respective numbers in England (Figure <ns0:ref type='figure' target='#fig_1'>1b</ns0:ref>), could be justified by the fact that White British likely constitute a smaller proportion of the reported 'white phenotype' in the USA (1.5M in USA &amp; Canada) and likewise, Japanese people possibly constitute a bigger proportion of of the reported 'Asian phenotype' in the USA (&gt;1.5M). Both White British (EUR-GBR) and Japanese (EAS-JPT) ethnic subgroups have among the lowest h1 ratio between all subgroups.</ns0:p><ns0:p>The fact that the proposed risk haplotype (A_C) involves the reference alleles of both studied SNPs, appears as counterintuitive. Especially so, since it has been suggested after analysis of a Chinese cohort, that it is the minor allele rs12252:G that is linked to increased COVID-19 severity <ns0:ref type='bibr' target='#b48'>(Zhang et al., 2020)</ns0:ref>. In fact the minor allele rs12252:G was linked to worse outcome in almost every related study, such as increased influenza severity <ns0:ref type='bibr' target='#b46'>(Zhang et al., 2013)</ns0:ref>, or more rapid HIV progression <ns0:ref type='bibr' target='#b47'>(Zhang et al., 2015)</ns0:ref>, although always observed in Chinese patients and not in European or American cohorts. This is noteworthy, as minor allele rs12252:G is found frequently in Chinese populations (roughly 50%), but is on the contrary rare in European populations (1-8%), or infrequent in South Asian (10-18%), or African groups (21%-33%). Interestingly, an inverted trend is observed in the other half of the discussed A_C haplotype, with namely rs34481144:T being rare in Chinese populations (1-2%), rare or infrequent in African groups (2-14%), but fairly frequent in European groups (38-56%). Rs34481144:T was found to correlate strongly with increased influenza severity in 3 independent cohorts <ns0:ref type='bibr' target='#b3'>(Allen et al., 2017)</ns0:ref>. These 3 independent cohorts, however, did not confirm the link between rs12252:G and increased influenza severity, as was suggested in Chinese cohorts. To add to the controversy of the possible antiviral effects linked to rs12252, a detailed study on 293T cells of the putative truncated variant &#916;1-21 that is theorized to result from the rs12252:G mutant, showed increased potential to restrict HIV replication and therefore an advantage compared to the complete IFITM3 protein carrying the reference allele <ns0:ref type='bibr' target='#b12'>(Compton et al., 2016)</ns0:ref>. However, this truncated version was not observed later in the blood of IAV or HIV patients <ns0:ref type='bibr' target='#b3'>(Randolph et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b32'>Makvandi-Nejad et al., 2018)</ns0:ref>, while rs12252:G was, on the contrary, found to enhance HIV-1 infection in Chinese patients <ns0:ref type='bibr' target='#b47'>(Zhang et al., 2015)</ns0:ref>. The reports for the functional role and consequences of minor rs12252:G allele are therefore conflicting and thus inconclusive. Although it is shown that &#916;1-21 variant redistributes the protein to the plasma membrane, by prohibiting the phosphorylation of residue Y20 that produces a signal for endocytosis <ns0:ref type='bibr'>(Jia et al., 2012)</ns0:ref>, a functional link between &#916;1-21 and rs12252:G has yet to be established. In the other examined SNP, the minor rs34481144:T allele is currently believed to favor the binding of transcriptional repressor CTCF, also known as CCCTC-binding factor, at IFITM3 promoter, seemingly leading to an inactive IFITM3 profile <ns0:ref type='bibr' target='#b3'>(Allen et al., 2017)</ns0:ref>. However, the exact functional effect of rs34481144:T is still not well understood. Manuscript to be reviewed repertoire, in conjunction with selective post translational modifications, and therefore should not be considered de facto as risk factors but rather as trade-offs in antiviral specificity <ns0:ref type='bibr' target='#b12'>(Compton et al., 2016)</ns0:ref>. This is further supported by the pronounced variability of rs12252:A and rs34481144:C frequencies, which is seen between the major groups in 1000 Genome Project populations, but not as much between subgroups of the same major group. Indeed, the observed spectrum of h1 haplotype prevalence across ancestral populations seems consistent with an evolutionary adaptation to specific immunological challenges and local factors of environmental pressure. In the case of SARS-CoV-2, the observed strong correlation of reference haplotype H1 (A_C) with increased morbidity in ethnic groups in England, could be pointing at a specific antiviral advantage conferred by the presence of each minor allele. However, since both minor alleles are not observed simultaneously (haplotype G_T is not represented), it is harder to conceive an independently equivalent beneficial effect by each distinct minor allele in the mixed reference/minor haplotypes H2 (G_C, here minor allele&gt; rs12252:G) and H3 (A_T, here minor allele&gt; rs34481144:T). Instead, it is more plausible to consider an effective hijacking of IFITM3 by SARS-CoV-2 in order to infect the cell, or to replicate, or to spread, or involving more than one of these phases. Indeed, there are known examples of similar hijacking, for example by the coronavirus that causes the common cold, HCoV-OC43 <ns0:ref type='bibr' target='#b50'>(Zhao et al., 2014)</ns0:ref>, or by human cytomegalovirus (HCMV) <ns0:ref type='bibr' target='#b45'>(Xie et al., 2015)</ns0:ref>.</ns0:p><ns0:p>More specifically for HCoV-OC43, it was shown that all three types of interferons, IFN-&#945;, IFN-&#947;, and IFN-&#955;, actually enhance HCoV-OC43 infection, while IFITM3 possibly promotes the low-pH-activated membrane fusion between the viral envelope and endosomal membranes. In contrast, human cytomegalovirus hijacks BST-2/tetherin to promote its entry into host cells and co-opts viperin to facilitate its replication, with IFITM3 facilitating the formation of the virion assembly compartment, but the virus is otherwise less sensitive to IFNs. In the case of SARS-CoV-2, it is therefore not inconceivable that if there is in fact a pro-infection role of IFITM3, that the virus could have evolved to exploit the most abundant haplotype A_C (59% abundance across all populations). The different effect between the H1 haplotype and H2/H3 haplotypes most probably involves the cellular distribution of IFITM3, which is mainly controlled by post-translational modifications, which in turn may be influenced by key polymorphisms such as the two examined here. Of great relevance, in this context, is the finding that plasma membrane localization of IFITM3 enhances SARS-CoV-2 infection, while endocytosis of IFITM3 effectively restricts the virus <ns0:ref type='bibr' target='#b20'>(Shi et al., 2020)</ns0:ref>. The same study confirms an even greater enhancement of SARS-CoV-2 in PeerJ reviewing PDF | (2020:08:52069:1:1:NEW 19 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed bypassing IFITM3 defense via TMPRSS2 activation of plasma membrane fusion, and reports compatibility with HCoV-OC43 mode of enhanced infection. Moreover, in another study by <ns0:ref type='bibr'>Bozzo et al.</ns0:ref>, IFITM3, together with IFITM2, were shown to boost SARS-CoV-2 infection, rather than restrict it, both in the absence and presence of interferon, which is consistent with our current suggestion of viral hijacking of IFITM3 <ns0:ref type='bibr'>(Bozzo et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The recent suggestion that rs12252:G is the risk allele in a n=80 COVID-19 cohort with Chinese patients <ns0:ref type='bibr' target='#b48'>(Zhang et al., 2020)</ns0:ref>, appears to challenge our conclusions, claiming the inverse effect. Considering that the cohort took place at Beijing You'an Hospital, if it is safe to assume that patients belonged to EAS-CHB group (Han Chinese in Beijing), the subgroup with the highest frequency in rs12252:G (54%), then an alternative interpretation of the result may be possible:. With 28/80 patients hospitalized with pneumonia being homozygotes rs12252:GG, 37/80 being heterozygotes rs12252:AG and 15/80 being homozygotes rs12252:AA, this results to 58% (93/160) abundance for the G allele (minor) and 42% (67/160) abundance for the A allele (reference). As the prior probability for the G allele was as high as 54%, the above result appears inconclusive (i.e. 58% observed vs 54% expected for rs12252:G) and therefore the suggestion that rs12252:G alone is a COVID-19 severity risk allele seems unfounded in this case. The same conclusion is reached, with whichever possible mix of the 3 available Chinese subgroups from 1000 Genomes Projects (EAS-CHB, EAS-CHS, EAS-CDX), as they all show high rs12252:G frequencies (0.47-0.54), surpassed only by the Japanese subgroup (EAS -JPT, 0.64). Interestingly, in an independent study that compared worldwide COVID-19 mortality statistics with rs12252 genotypic information per country (extracted from the PubMed database), rs12252:G was found to be negatively correlated with the SARS-CoV-2 mortality rate (p=0.0008), in agreement with our current outcomes <ns0:ref type='bibr' target='#b40'>(Pati et al., 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>The role of the examined IFITM3 variants in the severity of COVID-19 should be elucidated, by in vitro investigation of the effect of H1 (A_C), vs H2 (G_C), vs H3 (A_T) rs12252_rs34481144 haplotypes in SARS-CoV-2 infectivity and viral spread. Related in silico investigations should include the examined combined haplotype in their analysis and consider non-additive interactions. It is notable how various ongoing GWAS studies (e.g. The COVID-19 Host Genetics Initiative, 2020; 23andMe, 2020). did not confirm in their meta-analysis some other independently established genetic variant effects, which are otherwise broadly accepted by the scientific community as influencing COVID-19 severity, such as the ABO blood group <ns0:ref type='bibr' target='#b30'>(Li et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b44'>Wu et al., 2020)</ns0:ref>, or APOE e4 genotype <ns0:ref type='bibr'>(Kuo et al., 2020)</ns0:ref>. If functional differences between the examined IFITM3 haplotypes are shown to produce distinct profiles of COVID-19 progression in severe patients, then an improved understanding of the underlying mechanisms may allow more adequate or personalized treatment protocols. Last but not least important, is the need for investigating the examined variant implications in raising an effective immune response after future vaccination against SARS-CoV-2, as it was recently demonstrated that homozygote rs12252:GG, actually reduces the level of antibody response after influenza vaccination <ns0:ref type='bibr' target='#b29'>(Lei et al., 2020)</ns0:ref>. It could be important to verify whether this also stands for the upcoming SARS-CoV-2 vaccines, and whether a reduced antibody response could be instead elicited in this case by the major allele (rs12252:A) , as was suggested throughout our analysis regarding COVID-19 severity.</ns0:p><ns0:p>In conclusion, this study a) presented one more piece of evidence associating IFTM3 variants with the severity of COVID-19, b) suggested that the two most highly studied IFITM3 polymorphisms should be considered as a combined haplotype, and c) is calling for further research focus on this important first line of cellular antiviral defense. East Asian subgroups (in green) present the highest ranking among all 1000 Genomes Project populations in h2 ratios and rs34481144:C frequencies (5th column from left and second to last column), while European subgroups (in blue) present the highest ranking in h3 ratios and rs12252:A frequencies (7th &amp; 9th column from left). H1 haplotype ratio ranking (3rd-4th column from left) presents a similar alignment with the reported SMR of major groups in England (1st column from left). Note: ranked items are color-tagged by their major group, i.e., continent of origin: AFR, SAS, EAS, EUR -no subgroup pooling is shown here (see </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>As part of the IFITM family of proteins, one of the evolutionary ancient first lines of antiviral cellular defenses, the localization in endosomal or lysosomal membrane, or at the surface, e.g. of CD4 + T cells, and the exact antiviral mechanism of IFITM3, is regulated by many different post translational modifications, mainly palmitoylation, ubiquitination and phosphorylation. It is shown that genotypic variants of IFITM3 play a role in diversifying a host's potential antiviral PeerJ reviewing PDF | (2020:08:52069:1:1:NEW 19 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='20,42.52,205.87,525.00,525.00' type='bitmap' /></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 for</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>pooling details).</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 . Detailed allele and haplotype frequencies per ethnic subgroup derived from 1000 Genomes Project, for rs12252 and rs34481144.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>T A_T</ns0:cell><ns0:cell>A_C</ns0:cell><ns0:cell>G_C</ns0:cell></ns0:row></ns0:table></ns0:figure> </ns0:body> "

"Dr. Joseph J Gillespie Academic Editor for PeerJ Dear Dr. Gillespie, Thank you for your decision on our PeerJ submission entitled “The frequency of combined IFITM3 haplotype involving the reference alleles of both rs12252 and rs34481144 is in line with COVID-19 standardized mortality ratio of ethnic groups in England” (#2020:08:52069:1:0:NEW). We also thank the reviewers for their diligence in reading our manuscript and providing us with their valued constructive criticism. Please find detailed below major revisions as suggested or replies to the reviewers’ comments. 1) Dr. Dannielle Wellington suggested that we should reorganize our set of figures to avoid repetitiveness or ambiguity, as well as to find a better way to represent the visual comparison of SMR and h1 ratio rankings. In response, we made the following changes: a) We have completely removed figures 2 &3 from the previous version. b) We have created a two-panel figure, labeled now as “Figure 1”, showing haplotype ratios on panel A and SMR ratios on panel B. c) We have relabeled Figure 4 from previous version to Figure 3 at the revised version. d) We have created an additional figure (Figure 2) that shows a colored stacking of haplotype ratio and SMR rankings, highlighting the striking similarities between SMR and h1 ratio. Overall, the new version contains 3 figures and 2 tables. 2) We are thankful to Dr. Wellington for acknowledging that the literature review is well written, and for her insightful suggestion to include previously described data showing IFITM protein restriction of SARS-CoV, by Huang et al., Plos Path., 2011. We have now added a comment in lines 83-86, including the suggested reference and highlighting the fact that IFITM3 was shown to be less efficient than IFITM1 in SARS-CoV restriction. 3) We added two leading sentences to the results section explaining what we have done briefly, as suggested by Dr. Wellington (lines 141-146 of the revised manuscript). 4) We also made careful proof-reading which led to the correction of very few more grammatical errors, including the two that were spotted by Dr. Wellington. 5) We added a short comment at the end of the Introduction, lines 98-100 of the revised manuscript, explicitly stating the incentive of our research and clarifying any possible confusion regarding any connections with social inequality and racism. Though it may be unusual, we feel that in this case it would be beneficial to put a very clear note for the readers regarding the strict scientific ethics of the study. 6) In lines 146-154 and 165-171 of the revised manuscript, we restructured our rationale regarding the comparison of un-pooled rankings between SMR in England and h1 ratio. The argument of rankings alignment is now stronger and more clear. 7) We added a short comment connecting the increased variability of rs12252 and rs34481144 allele frequencies in populations across the world, with local, ancestral, evolutionary pressure (lines 228-233 of the revised manuscript). 8) Dr. Ronghua Jin mentioned that there is no precise basic information to support this analysis, for example, age, gender, health status, medical conditions. This in silico analysis is based, as far as the Standardized Mortality Ratios are concerned, on the results and conclusions of the peer reviewed study made by Aldridge et al., 2020. Therefore, we built our analysis on the exact form that was chosen for the representation of their results, namely column 8, in Table 1 of the respective reference (Aldridge et al., 2020). We acknowledge that the above fact may not have been stated explicitly enough in our previous manuscript, and therefore we added two clarifying sentences at the beginning of our Methods section (lines 103-107). Furthermore, we have added the repository link of the dataset of the related manuscript to the data links of our study (UCL DISCOVERY, Dataset: Black, Asian and Minority Ethnic groups in England are at increased risk of death from COVID-19; https://discovery.ucl.ac.uk/id/eprint/10096589) 9) Dr. Jin commented that although the research question was well defined, the idea of creating correspondences with ethnic groups and the reported standardized mortality ratios was unreliable. In response to the reviewer, we would like to point out that in our original manuscript, we clearly acknowledge the limitations of ethnic pools and groupings (paragraph between lines 161-182 in the revised manuscript) and we state that, at the worst case, the pooled comparison greatly reinforces the ranking comparison at the level of major groups (i.e. un-pooled comparison, EAS vs EUR vs SAS vs AFR, see lines 143-147). Hence, the decision of pooling was a necessary step that validated our initial assumption of h1 ratio and SMR alignment in English population. Additionally, it is unclear in which ways our pooling methodology was deemed unreliable. It would be greatly appreciated, if the reviewer is still not convinced, to kindly specify. 10) In relation to the previous comment and in support to our manuscript, we would also like to bring to your attention that the observations of our study (in preprint form) were recently cited in an accepted manuscript in The Journal of Infectious Diseases (Pati et al., 2020; https://doi.org/10.1093/infdis/jiaa630), which to our opinion supports and further corroborates the importance of our findings and the reliability of our method. We added a related sentence in lines 274-277 of the revised manuscript. 11) Dr. Wellington commented directly to us that our manuscript is well written and presents an interesting observation in regards to the link between COVID-19 mortality and IFITM3, while Dr. Jin commented that the research question is very important and that further analysis with more precise data is necessary to support the findings. We thank both for their very useful views and opinion and wish to emphasize that our study aims to add one more piece to the continuously increasing volume of evidence linking IFITM3 to COVID-19 severity. In the current setting of the ongoing pandemic, adding pieces together from this unprecedented flow of worldwide scientific observations and results, has been so far the most fruitful approach towards the goals of understanding and containing both SARS-CoV-2 and COVID-19. Our team made the observation that the major alleles of the two most well studied polymorphisms of IFTM3 could be favoring the hijacking of the restriction factor by SARS-CoV-2, as early as 15th of May 2020 (see our 1st related preprint version, https://www.preprints.org/ manuscript/202005.0273/v1). Since then, numerous independent studies implicated IFITM3 specifically in COVID-19 severity, although none so far succeeded in correlating the two most studied polymorphisms as a combined haplotype. Perhaps the most prominent of all new pieces of evidence is yet another recent study proposing explicitly that “SARS-CoV-2 appears to hijack IFITMs” (Bozzo et al, 2020; preprint, here: https://www.biorxiv.org/content/10.1101/ 2020.08.18.255935v1, covered by Nature, here: https://www.nature.com/articles/s41577-020-00448-0, covered by medical news site, here: https://www.news-medical.net/news/20200820/SARS-CoV-2-hijacks-antiviral-factors-to-promote-infection-in-human-lung-cells.aspx). A related sentence was added in the revised manuscript, in lines 257-260. The importance of IFITM3 in SARS-CoV-2 Infection is further illustrated by the NIH funding available for this research; see: “Deciphering the Double-Edged Role of IFITM3 During SARS-CoV-2 Infection”, A.Compton, awardee of NIH’s Office of Intramural Research, ITAC Awards (https://oir.nih.gov/about/intramural-targeted-anti-covid-19-itac-awards). It seems there is now an established consensus regarding a potentially crucial role of IFITM3 in COVID-19. We are very confident on the validity of our results and strongly believe that our data will shed more light as to the contribution of host genetics on SARS-CoV-2 cellular invasion and COVID-19 severity. Your sincerely, Dr. D.Nikoloudis On behalf of Dimitris Nikoloudis, Dimitrios Kountouras and Asimina Hiona. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Evidence was brought forward in England and the USA that Black, Asian, Latino and Minority Ethnic people exhibit higher mortality risk from COVID-19 than White people.</ns0:p><ns0:p>While socioeconomic factors were suggested to contribute to this trend, they arguably do not explain the range of the differences observed, allowing for possible genetic implications. Almost concurrently, the analysis of a cohort in Chinese COVID-19 patients proposed an association between the severity of the disease and the presence of the minor allele of rs12252 of the Interferon-induced transmembrane protein 3 (IFITM3) gene. This SNP, together with rs34481144, are the two most studied polymorphisms of IFITM3 and have been associated in the past with increased severity in Influenza, Dengue, Ebola, and HIV viruses. IFITM3 is an immune effector protein that is pivotal for the restriction of viral replication, but also for the regulation of cytokine production. Following up on these two developments in the ongoing SARS-CoV-2 pandemic, the present study investigates a possible association between the differences in mortality of ethnic groups in England and the combined haplotypes of rs12252 and rs34481144. The respective allele frequencies were collected for 26 populations from 1000 Genomes Project and subgroups were pooled wherever possible to create correspondences with ethnic groups in England. A significant correlation (r=0.9687, p= 0.0003) and a striking agreement was observed between the reported Standardized Mortality Ratios and the frequency of the combined haplotype of both reference alleles, suggesting that the combination of the reference alleles of the specific SNPs may be implicated in more severe outcomes of COVID-19. This study calls for further focus on the role of IFITM3 variants in the mechanism of cellular invasion of SARS-CoV-2, their impact in COVID-19 severity and their possible implications in vaccination efficacy.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Emerging scientific evidence from international (Kirby, 2020) and UK ( <ns0:ref type='bibr'>Aldridge et al., 2020)</ns0:ref> COVID-19 patient reports and death records, indicate a disproportionate effect of the novel coronavirus on ethnic minorities. According to CDC (CDC, 2020), Black, Asian and Minority Ethnic (BAME) people are at higher risk of death from COVID-19. Importantly, an Indirect Standardization of NHS mortality data in England ( <ns0:ref type='bibr'>Aldridge et al., 2020)</ns0:ref>, revealed that the adjusted for age and region Standardized Mortality Ratios (SMRs), were highest in Black African, Black Caribbean, Pakistani, Bangladeshi, and Indian minority ethnic groups. On the contrary, White Irish and White British ethnic groups exhibited a significantly lower risk of death. Similarly, in the USA <ns0:ref type='bibr' target='#b16'>(Garg et al., 2020)</ns0:ref>, preliminary data compiled from hospitals in 14 US states, confirmed the UK study outcomes, showing that African Americans are also disproportionately affected by COVID-19. Specifically, African Americans represented 33% of COVID-19 hospitalizations, despite only making up 18% of the total population studied. In a subsequent analysis, among COVID-19 deaths in New York City, for which race and ethnicity data were available, death rates from COVID-19 among black or African Americans and Hispanic or Latinos were substantially higher than that of white or Asian people <ns0:ref type='bibr' target='#b16'>(Garg et al., 2020)</ns0:ref>.</ns0:p><ns0:p>Several reasons have been proposed to explain these ethnic discrepancies in COVID-19 mortality risk arising from these preliminary studies. Chronic pre-existing conditions, such as Cardio Vascular Diseases (CVD), diabetes, hypertension, obesity, etc. are more common in minorities compared to Caucasian populations and have all been associated with adverse outcomes in COVID-19 (Centers for Disease Control and Prevention, 2020; <ns0:ref type='bibr' target='#b26'>Kirby, 2020)</ns0:ref>. However, race disparities in those diseases are not large enough to fully explain the COVID-19 death disparity <ns0:ref type='bibr'>(Aldridge et al., 2020)</ns0:ref>. Factors such as housing and living conditions, use of public transportation, lack of regular access to primary health, and occupation-related differences that prohibit the work from home, or require more frequent and/or close social contact, may have all played an important role in producing disproportionate death rates among BAME groups <ns0:ref type='bibr'>(Aldridge et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b26'>Kirby, 2020;</ns0:ref><ns0:ref type='bibr'>Niedzwiedz et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b25'>Khunti et al., 2020)</ns0:ref>. Nevertheless, it is suggested that inequalities in socioeconomic status parameters do not seem to adequately explain the range of differences, and in some instances, the extreme variations observed among ethnic minorities in mortality rates from COVID-19 infection <ns0:ref type='bibr' target='#b26'>(Kirby, 2020)</ns0:ref>.</ns0:p><ns0:p>As the importance of genetic polymorphisms (SNPs) in the modulation of individual susceptibility to, and severity of, infectious diseases has been well established <ns0:ref type='bibr' target='#b10'>(Chapman et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b52'>Zhao et al, 2018)</ns0:ref>, we turned our focus to two very highly studied polymorphisms of the interferon-induced transmembrane protein 3 (IFITM3) gene: rs12252 and rs34481144. IFITM3 encodes an immune effector protein that is pivotal for restriction of viral replication <ns0:ref type='bibr'>(Brass et al., 2009)</ns0:ref> of many enveloped RNA viruses including HIV-1, influenza A virus (IAV), Ebola and Dengue virus ( <ns0:ref type='bibr'>Brass et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b15'>Feeley et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b21'>Huang et al., 2011;</ns0:ref><ns0:ref type='bibr' target='#b14'>Everitt et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b11'>Compton et al., 2014)</ns0:ref>. IFITM3 has also been demonstrated to affect severity of infection and improve the host cellular defenses against viruses ( <ns0:ref type='bibr'>Brass et al., 2009;</ns0:ref><ns0:ref type='bibr' target='#b14'>Everitt et al., 2012;</ns0:ref><ns0:ref type='bibr' target='#b11'>Compton et al., 2014)</ns0:ref>. Interestingly, IFITM3 has also been shown to act as a regulator of antiviral immunity that controls cytokine production to restrict viral pathogenesis, in CMV <ns0:ref type='bibr' target='#b39'>(Stacey et al., 2017)</ns0:ref> and Sendai virus <ns0:ref type='bibr' target='#b24'>(Jiang et al., 2017)</ns0:ref>. This finding is particularly important since cytokine storm in influenza can lead to a rapid progression of the infection in humans <ns0:ref type='bibr' target='#b42'>(Wang et al., 2014)</ns0:ref> and the same observation is also apparent in COVID-19 severe and deadly cases <ns0:ref type='bibr' target='#b17'>(Giamarellos-Bourboulis et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b4'>Blanco-Melo et al., 2020)</ns0:ref>. Moreover, IFITM3 was found to be explicitly upregulated in SARS-CoV-2 infected cells <ns0:ref type='bibr' target='#b4'>(Blanco-Melo et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b18'>Hachim et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b19'>He et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The minor allele of rs12252 (C in minus, or G in plus strand orientation) has been associated with rapid progression of acute HIV infection <ns0:ref type='bibr' target='#b48'>(Zhang et al., 2015)</ns0:ref>, with the severity of influenza <ns0:ref type='bibr' target='#b47'>(Zhang et al., 2013)</ns0:ref> and recently with COVID-19 severity <ns0:ref type='bibr' target='#b49'>(Zhang et al, 2020)</ns0:ref>. The minor allele of rs34481144 (A in minus, or T in plus strand orientation) was previously found to be correlated with increased severity of IAV infection <ns0:ref type='bibr' target='#b3'>(Allen et al., 2017)</ns0:ref>. Moreover, the minor allele of rs34481144 is also associated with enhanced methylation on the IFITM3 promoter of CD8+ T cells, and general transcriptional repression of the broader locus surrounding IFITM3, which includes several genes known to be involved in host responses to viral infection <ns0:ref type='bibr' target='#b43'>(Wellington et al., 2019)</ns0:ref>. SARS-CoV-2 uses primarily the ACE2 receptor as main point of entry and the host cell serine protease TMPRSS2 for viral spike (S) protein priming <ns0:ref type='bibr' target='#b20'>(Hoffmann et al., 2020)</ns0:ref>. Severe acute respiratory syndrome coronavirus (SARS-CoV), which also uses ACE2 as a receptor, has been shown to be restricted more efficiently by IFITM1 than by IFITM3, presenting a different restriction pattern than IAV <ns0:ref type='bibr' target='#b21'>(Huang et al., 2011)</ns0:ref>. Interestingly, it was recently shown that TMPRSS2 is specifically allowing evasion of IFITM3 restriction for bat SARS-Like WIV1 coronavirus <ns0:ref type='bibr' target='#b53'>(Zheng et al., 2020)</ns0:ref>, opening the possibility for a similar mechanism in the case of SARS-CoV-2. Further potential involvement of IFITM3 in COVID-19 outcome was revealed in the context of syncytial pneumocytes in severe cases with extensive lung damage, where it was suggested that the cellular location of IFITMs 1-3 could be playing a role in syncytia formation <ns0:ref type='bibr'>(Buchrieser et al., 2020)</ns0:ref>. Indeed, the accumulation of many direct and indirect layers of evidence linking IFITM3 with COVID-19 severity, has also led to explicit calls for further investigation of the role of this highly relevant first-line of cellular defense protein <ns0:ref type='bibr' target='#b50'>(Zhao, 2020)</ns0:ref>. Following up to the analysis of COVID-19 NHS mortality data in BAME groups ( <ns0:ref type='bibr'>Aldridge et al., 2020)</ns0:ref>, the purpose of the present study was to investigate a possible association between the stand-alone and combined frequencies of the alleles of the IFITM3 gene variants rs12252 and rs34481144, with COVID-19 standardized mortality ratio of ethnic groups in England.</ns0:p><ns0:p>The sole incentive of the current research is to help improve the existing and future treatment protocols for severe COVID-19 patients, and in no case to provide DNA-based arguments that may be used to mask existing social inequalities or racism.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods</ns0:head><ns0:p>The Standardized Mortality Ratios (SMR) of ethnic groups in England, adjusted for age and region, were adopted from the study by <ns0:ref type='bibr'>Aldridge et al. in</ns0:ref> the exact form in which they were presented <ns0:ref type='bibr'>(Aldridge et al., 2020)</ns0:ref>. Specific dataset details, including age, region and ethnicity information, are available at the repository address of the respective publication (https://discovery.ucl.ac.uk/ id/eprint/10096589). The rs12252 (A&gt;G) and rs34481144 (C&gt;T) allele and haplotype frequencies were collected for all available 1000 Genomes Project ancestral populations, from LDlink (LDhap tool : https://ldlink.nci.nih.gov/?tab=ldhap) , specifically 5 major groups, i.e. African (AFR), Ad Mixed American (AMR), European (EUR), East Asian (EAS) and South Asian (SAS), comprising 26 subgroups in total <ns0:ref type='bibr' target='#b30'>(Machiela et al., 2015)</ns0:ref> (Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref>).</ns0:p><ns0:p>The plus orientation for the reference and minor alleles was retained throughout this analysis, for better data handling and in compliance with dbSNP. Combined rs12252_rs34481144 haplotypes were defined as A_C (H1), G_C (H2), A_T (H3), while G_T haplotype was not represented at all in the data Rankings were examined by sorting all populations by individual reference allele (rs12252:A, rs34481144:C) and by combined haplotype frequency ratios (rs12252_rs34481144: h1_ratio=A_C/(A_T+G_C), h2_ratio=G_C/(A_C+A_T), h3_ratio=A_T/(G_C+A_C)) (Figure <ns0:ref type='figure' target='#fig_0'>1A</ns0:ref>), and subsequently compared visually to the reported Standardized Mortality Ratios (SMR) of ethnic groups in England (Figures <ns0:ref type='figure' target='#fig_1'>1B and 2</ns0:ref>). Additionally, an attempt was made to correlate directly the two rankings, i.e. SMR and IFITM3 haplotype frequencies by specific reported ethnic subgroup. UK demographics sources were therefore consulted (Office for National Statistics, UK, 2011; <ns0:ref type='bibr' target='#b9'>Chanda &amp; Ghosh, 2012)</ns0:ref> in order to pool, wherever possible, the ancestral subgroups to the reported ethnic groups in England. With all reservations tied to the inevitable discrepancies of this type of simplified socio-genetic correspondences, the following pools were formed: [AFR-YRI, AFR-LWK, AFR-GWD, AFR-MSL, AFR-ESN]&gt;'African', [SAS-STU, SAS-GIH, SAS-PJL]&gt;'Indian', [EAS-CDX, EAS-CHS, EAS-CHB]&gt;'Chinese', [EUR-CEU, EUR-IBS, EUR-TSI, EUR-FIN]&gt;'White Other' (see Table <ns0:ref type='table' target='#tab_1'>1</ns0:ref> for full subgroup descriptions and Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref> for pooled subgroups and correspondences to ethnic minorities). A pool for the reported Pakistani group failed to form from ancestral populations, as the Punjabi (SAS-PJL), being the only related subgroup, account roughly for just 45% of Pakistan's demographics, while in London the community includes comparable numbers of Punjabis, Pathans and Kashmiris, with small communities of Sindhis and Balochis (Department for Communities and Local Government, UK, 2009). Moreover, the Punjabi form also a considerable part of Indians' pool (at least 40% of Delhi's total population), therefore a single-ended direct correspondence between Punjabi and British Pakistani was not warranted in this case. Indian Telugu (SAS-ITU) were not included in Indians' pool, as no demographic report was suggestive of comparable numbers to the other 3 included subgroups, for people of Indian origin in England. The same rationale applied for the non-inclusion of AFR-ASW (Americans of African Ancestry in SW USA) in the African pool. The haplotype frequencies were simply averaged within pooled groups, and both the ratios and SMR were normalized to the White British result (represented uniquely by EUR-GBR subgroup) (Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>We extracted rs12252 and rs34481144 allele frequencies of various ethnic groups from the 1000 Genomes Project, in order to examine whether the distribution of any one of the combined haplotypes is directly correlated with the reported SARS-CoV-2 related SMR in England. At first we compared the trend lines of reference allele frequencies with those of combined haplotype ratios (Figure <ns0:ref type='figure' target='#fig_0'>1A</ns0:ref>). The fluctuation of frequencies of major alleles rs12252:A and rs34481144:C follows a trend similar to h3 and h2 ratios, respectively, while h1 ratio shows a unique trend. The ranking that visually appeared in line with the reported SMR, adjusted for age and NHS region (Figure <ns0:ref type='figure' target='#fig_0'>1B</ns0:ref>), was produced by the h1 ratio. Two levels of SMR vs haplotype comparisons were applied: first, we compared at the level of major groups (i.e., un-pooled comparison: EAS vs SAS vs EUR vs AFR, see Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>), second, at the level of ethnic subgroups, wherever possible (pooled comparison, see Table <ns0:ref type='table' target='#tab_0'>2</ns0:ref>). For the evaluation of un-pooled rank correlation we considered two scenarios. First, we considered all 10 ethnic groups with an available SMR (Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>) and we observed an almost perfect alignment with h1 ratios, which corresponds to a permutation event of 10 items, with 1 / 3,628,800 chance to occur randomly, corresponding to p = 3*10 -7 (5&#963;). Secondly, we considered only 5 aligned items, specifically the sequence of (a) African groups, followed by (b) South Asian, followed by (c) White Non-British, (d) Chinese, and finally (e) White British, a permutation event with 1 / 120 chance to occur randomly, or p = 0.008. In both cases the correlations proved highly significant. It is noteworthy, that subgroups EUR-IBS (Iberian Population in Spain) and EUR-TSI (Toscani in Italia), representative of two countries that suffered higher death rates than other European countries, share the highest h1 ratio between all European subgroups. Subsequently, in order to assess the potential strength of the theorized association, the rankings of pooled h1 ratios and SMR, per group, were linearly and significantly correlated, with Pearsonr=0.9687, p=3*10 -4 (&gt;3.5&#963;) (Figure <ns0:ref type='figure'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The calculated level of correlation appears to be remarkable, considering the possible discrepancies in the pooling of the available ancestral groups, but also the expected multiparametric causes of the observed COVID-19 SMR in England's ethnic groups (as previously described, potentially involving prior health status, income level, household density, behavioral biases, questionable attribution of death to COVID-19, etc.). On one hand, the alignment of ethnic group rankings between SMR and un-pooled h1 ratio, is less than 1/120 probable to occur randomly at the worst case, and less than 1 in 3.6 million probable to occur randomly at the best case. In other words, if one hypothesizes that the reported SMR rankings are solely due to socioeconomic factors, then one would conclude that socioeconomic factors would be in perfect alignment with h1 ratios. The possibility of the above to occur seems highly unlikely, thus pointing to the fact that differential allele frequencies play a potentially important role in the reported SMR.</ns0:p><ns0:p>On the other hand, the level of correlation (&gt;3.5&#963;) between SMR and pooled h1 ratio, confirms the previous alignment and appears strong enough to suggest a possible causal link, albeit in this case, the pooling process may have introduced some discrepancies. The potential introduction of pooling discrepancies is expected and is impossible to quantify given the available data. However, the purpose of the pooled analysis was mainly to reinforce the primary association between IFITM3 and COVID-19 severity rather than to (indirectly) infer causality. Considering the parametric uncertainty of the pooled analysis it is probably inadequate to suggest a causal link, based on the strength of the observed correlation alone. However, taken together, this set of results constitutes a clear and valid starting point for designing further investigations regarding the role of IFITM3 in COVID-19 severity and appears as one more piece of evidence towards this direction. Therefore, the main point of this analysis is that the two examined SNPs should preferably henceforth be studied under a combined haplotype and not separately, as was performed so far for SARS-CoV-2, and a great variety of other viruses.</ns0:p><ns0:p>Before proceeding to the discussion of the implications of the above conclusion, it is interesting to also view the present observations in the context of data from the USA. A preliminary analysis of death rates from COVID-19 in New York City shows 92.3 deaths per 100,000 population among black or African American people, followed by Hispanic or Latino people (74.3), then by white (45.2) or Asian (34.5) people <ns0:ref type='bibr' target='#b26'>(Kirby, 2020)</ns0:ref>. The same trend was clearly displayed in the initial ranking by h1 ratio (Figure <ns0:ref type='figure' target='#fig_0'>1A</ns0:ref>), with American populations (AMR) occupying the middle of the chart, between African and European / Asian populations.</ns0:p><ns0:p>Interestingly, the reported lower death rate of Asian people compared to white people in the New York data, which represents an inversion of the respective numbers in England (Figure <ns0:ref type='figure' target='#fig_0'>1B</ns0:ref>), could be justified by the fact that White British likely constitute a smaller proportion of the reported 'white phenotype' in the USA (1.5M in USA &amp; Canada) and likewise, Japanese people possibly constitute a bigger proportion of of the reported 'Asian phenotype' in the USA (&gt;1.5M). Both White British (EUR-GBR) and Japanese (EAS-JPT) ethnic subgroups have among the lowest h1 ratio between all subgroups.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:08:52069:2:0:NEW 27 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>It shThe fact that the proposed risk haplotype (A_C) involves the reference alleles of both studied SNPs, appears as counterintuitive. Especially so, since it has been suggested after analysis of a Chinese cohort, that it is the minor allele rs12252:G that is linked to increased COVID-19 severity <ns0:ref type='bibr' target='#b49'>(Zhang et al., 2020)</ns0:ref>. In fact the minor allele rs12252:G was linked to worse outcome in almost every related study, such as increased influenza severity <ns0:ref type='bibr' target='#b47'>(Zhang et al., 2013)</ns0:ref>, or more rapid HIV progression <ns0:ref type='bibr' target='#b48'>(Zhang et al., 2015)</ns0:ref>, although always observed in Chinese patients and not in European or American cohorts. This is noteworthy, as minor allele rs12252:G is found frequently in Chinese populations (roughly 50%), but is on the contrary rare in European populations (1-8%), or infrequent in South Asian (10-18%), or African groups (21%-33%). Interestingly, an inverted trend is observed in the other half of the discussed A_C haplotype, with namely rs34481144:T being rare in Chinese populations (1-2%), rare or infrequent in African groups (2-14%), but fairly frequent in European groups (38-56%). Rs34481144:T was found to correlate strongly with increased influenza severity in 3 independent cohorts <ns0:ref type='bibr' target='#b3'>(Allen et al., 2017)</ns0:ref>. These 3 independent cohorts, however, did not confirm the link between rs12252:G and increased influenza severity, as was suggested in Chinese cohorts. To add to the controversy of the possible antiviral effects linked to rs12252, a detailed study on 293T cells of the putative truncated variant &#916;1-21 that is theorized to result from the rs12252:G mutant, showed increased potential to restrict HIV replication and therefore an advantage compared to the complete IFITM3 protein carrying the reference allele <ns0:ref type='bibr' target='#b12'>(Compton et al., 2016)</ns0:ref>. However, this truncated version was not observed later in the blood of IAV or HIV patients <ns0:ref type='bibr' target='#b3'>(Randolph et al., 2017;</ns0:ref><ns0:ref type='bibr' target='#b31'>Makvandi-Nejad et al., 2018)</ns0:ref>, while rs12252:G was, on the contrary, found to enhance HIV-1 infection in Chinese patients <ns0:ref type='bibr' target='#b48'>(Zhang et al., 2015)</ns0:ref>. The reports for the functional role and consequences of minor rs12252:G allele are therefore conflicting and thus inconclusive. Although it is shown that &#916;1-21 variant redistributes the protein to the plasma membrane, by prohibiting the phosphorylation of residue Y20 that produces a signal for endocytosis <ns0:ref type='bibr'>(Jia et al., 2012)</ns0:ref>, a functional link between &#916;1-21 and rs12252:G has yet to be established. In the other examined SNP, the minor rs34481144:T allele is currently believed to favor the binding of transcriptional repressor CTCF, also known as CCCTC-binding factor, at IFITM3 promoter, seemingly leading to an inactive IFITM3 profile <ns0:ref type='bibr' target='#b3'>(Allen et al., 2017)</ns0:ref>. However, the exact functional effect of rs34481144:T is still not well understood.</ns0:p><ns0:p>As part of the IFITM family of proteins, one of the evolutionary ancient first lines of antiviral cellular defenses, the localization in endosomal or lysosomal membrane, or at the surface, PeerJ reviewing PDF | (2020:08:52069:2:0:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed e.g. of CD4 + T cells, and the exact antiviral mechanism of IFITM3, is regulated by many different post translational modifications, mainly palmitoylation, ubiquitination and phosphorylation. It is shown that genotypic variants of IFITM3 play a role in diversifying a host's potential antiviral repertoire, in conjunction with selective post translational modifications, and therefore should not be considered de facto as risk factors but rather as trade-offs in antiviral specificity <ns0:ref type='bibr' target='#b12'>(Compton et al., 2016)</ns0:ref>. This is further supported by the pronounced variability of rs12252:A and rs34481144:C frequencies, which is seen between the major groups in 1000 Genome Project populations, but not as much between subgroups of the same major group. Indeed, the observed spectrum of h1 haplotype prevalence across ancestral populations seems consistent with an evolutionary adaptation to specific immunological challenges and local factors of environmental pressure. In the case of SARS-CoV-2, the observed strong correlation of reference haplotype H1 (A_C) with increased morbidity in ethnic groups in England, could be pointing at a specific antiviral advantage conferred by the presence of each minor allele. However, since both minor alleles are not observed simultaneously (haplotype G_T is not represented), it is harder to conceive an independently equivalent beneficial effect by each distinct minor allele in the mixed reference/minor haplotypes H2 (G_C, here minor allele&gt; rs12252:G) and H3 (A_T, here minor allele&gt; rs34481144:T). Instead, it is more plausible to consider an effective hijacking of IFITM3 by SARS-CoV-2 in order to infect the cell, or to replicate, or to spread, or involving more than one of these phases. Indeed, there are known examples of similar hijacking, for example by the coronavirus that causes the common cold, HCoV-OC43 <ns0:ref type='bibr' target='#b51'>(Zhao et al., 2014)</ns0:ref>, or by human cytomegalovirus (HCMV) <ns0:ref type='bibr' target='#b45'>(Xie et al., 2015)</ns0:ref>.</ns0:p><ns0:p>More specifically for HCoV-OC43, it was shown that all three types of interferons, IFN-&#945;, IFN-&#947;, and IFN-&#955;, actually enhance HCoV-OC43 infection, while IFITM3 possibly promotes the low-pH-activated membrane fusion between the viral envelope and endosomal membranes. In contrast, human cytomegalovirus hijacks BST-2/tetherin to promote its entry into host cells and co-opts viperin to facilitate its replication, with IFITM3 facilitating the formation of the virion assembly compartment, but the virus is otherwise less sensitive to IFNs. In the case of SARS-CoV-2, it is therefore not inconceivable that if there is in fact a pro-infection role of IFITM3, that the virus could have evolved to exploit the most abundant haplotype A_C (59% abundance across all populations). The different effect between the H1 haplotype and H2/H3 haplotypes most probably involves the cellular distribution of IFITM3, which is mainly controlled by post-translational modifications, which in turn may be influenced by key polymorphisms such as the two examined here. Of great relevance, in this context, is the finding that plasma membrane localization of IFITM3 enhances SARS-CoV-2 infection, while endocytosis of IFITM3 effectively restricts the virus <ns0:ref type='bibr' target='#b19'>(Shi et al., 2020)</ns0:ref>. The same study confirms an even greater enhancement of SARS-CoV-2 in bypassing IFITM3 defense via TMPRSS2 activation of plasma membrane fusion, and reports compatibility with HCoV-OC43 mode of enhanced infection. Moreover, in another study by <ns0:ref type='bibr'>Bozzo et al.,</ns0:ref><ns0:ref type='bibr'>IFITM3,</ns0:ref><ns0:ref type='bibr'>together with IFITM2,</ns0:ref><ns0:ref type='bibr'>were</ns0:ref> shown to boost SARS-CoV-2 infection, rather than restrict it, both in the absence and presence of interferon, which is consistent with our current suggestion of viral hijacking of IFITM3 <ns0:ref type='bibr'>(Bozzo et al., 2020)</ns0:ref>.</ns0:p><ns0:p>The recent suggestion that rs12252:G is the risk allele in a n=80 COVID-19 cohort with Chinese patients <ns0:ref type='bibr' target='#b49'>(Zhang et al., 2020)</ns0:ref>, appears to challenge our conclusions, claiming the inverse effect. Considering that the cohort took place at Beijing You'an Hospital, if it is safe to assume that patients belonged to EAS-CHB group (Han Chinese in Beijing), the subgroup with the highest frequency in rs12252:G (54%), then an alternative interpretation of the result may be possible.</ns0:p><ns0:p>With 28/80 patients hospitalized with pneumonia being homozygotes rs12252:GG, 37/80 being heterozygotes rs12252:AG and 15/80 being homozygotes rs12252:AA, this results to 58% (93/160) abundance for the G allele (minor) and 42% (67/160) abundance for the A allele (reference). As the prior probability for the G allele was as high as 54%, the above result appears inconclusive (i.e. 58% observed vs 54% expected for rs12252:G) and therefore the suggestion that rs12252:G alone is a COVID-19 severity risk allele seems unfounded in this case. The same conclusion is reached, with whichever possible mix of the 3 available Chinese subgroups from 1000 Genomes Projects (EAS-CHB, EAS-CHS, EAS-CDX), as they all show high rs12252:G frequencies (0.47-0.54), surpassed only by the Japanese subgroup (EAS -JPT, 0.64). In the case where heterozygote rs12252:AG (37/80 or 46% abundance) is expected to behave similarly to homozygote rs12252:AA (19%), so that homozygote rs12252:GG (35%) appears as the risk genotype, then a chi-square statistic would report a non-significant p = 0.36, when q (frequency of rs12252:G) = 0.54 and q 2 = 0.29 according to Hardy-Weinberg equilibrium, or borderline significant p=0.04, when q (frequency of rs12252:G) = 0.47, with q 2 = 0.22. Therefore, it is equally unsafe to associate homozygote rs12252:GG with COVID-19 severity.</ns0:p><ns0:p>It is acknowledged that the conclusions of the present investigation take into consideration only allele frequencies and not direct genotyped data. Nevertheless, allele frequencies from large PeerJ reviewing PDF | (2020:08:52069:2:0:NEW 27 Oct 2020) Manuscript to be reviewed sequencing studies, such as the 1000 Genomes project, are deemed important to be taken into account when available, in order to provide further insight and direction to studies based exclusively on genotypic analysis. In further support to the above, an independent study that compared worldwide COVID-19 mortality statistics with rs12252 allele frequency information per country (extracted from the PubMed database), found that rs12252:G was negatively correlated with the SARS-CoV-2 mortality rate (p=0.0008), in agreement with our current outcomes <ns0:ref type='bibr' target='#b40'>(Pati et al., 2020)</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusion</ns0:head><ns0:p>The role of the examined IFITM3 variants in the severity of COVID-19 should be elucidated, by in vitro investigation of the effect of H1 (A_C), vs H2 (G_C), vs H3 (A_T) rs12252_rs34481144 haplotypes in SARS-CoV-2 infectivity and viral spread. Related in silico investigations should include the examined combined haplotype in their analysis and consider nonadditive interactions. It is notable how various ongoing GWAS studies (e.g. The COVID-19 Host Genetics Initiative, 2020; 23andMe, 2020). did not confirm in their meta-analysis some other independently established genetic variant effects, which are otherwise broadly accepted by the scientific community as influencing COVID-19 severity, such as the ABO blood group <ns0:ref type='bibr' target='#b29'>(Li et al., 2020;</ns0:ref><ns0:ref type='bibr' target='#b44'>Wu et al., 2020)</ns0:ref>, or APOE e4 genotype <ns0:ref type='bibr'>(Kuo et al., 2020)</ns0:ref>. If functional differences between the examined IFITM3 haplotypes are shown to produce distinct profiles of COVID-19 progression in severe patients, then an improved understanding of the underlying mechanisms may allow more adequate or personalized treatment protocols. Last but not least important, is the need for investigating the examined variant implications in raising an effective immune response after future vaccination against SARS-CoV-2, as it was recently demonstrated that homozygote rs12252:GG, actually reduces the level of antibody response after influenza vaccination <ns0:ref type='bibr' target='#b28'>(Lei et al., 2020)</ns0:ref>. It could be important to verify whether this also stands for the upcoming SARS-CoV-2 vaccines, and whether a reduced antibody response could be instead elicited in this case by the major allele (rs12252:A), as was suggested throughout our analysis regarding COVID-19 severity.</ns0:p><ns0:p>In conclusion, this study a) presented one more piece of evidence associating IFTM3 variants with the severity of COVID-19, b) suggested that the two most highly studied IFITM3 polymorphisms should be considered as a combined haplotype, and c) is calling for further research focus on this important first line of cellular antiviral defense. East Asian subgroups (in green) present the highest ranking among all 1000 Genomes Project populations in h2 ratios and rs34481144:C frequencies (5th column from left and second to last column), while European subgroups (in blue) present the highest ranking in h3 ratios and rs12252:A frequencies (7th &amp; 9th column from left). H1 haplotype ratio ranking (3rd column from left) presents an almost identical alignment with the reported SMR of major groups in England (1st column from left). Note: ranked items are color-tagged by their major group, i.e., continent of origin: AFR, SAS, EAS, EUR -no subgroup pooling is shown here (see </ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:52069:2:0:NEW 27 Oct 2020)Manuscript to be reviewed</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head>Table 2 for</ns0:head><ns0:label>2</ns0:label><ns0:figDesc /><ns0:table /><ns0:note>pooling details).</ns0:note></ns0:figure> <ns0:figure type='table' xml:id='tab_1'><ns0:head>Table 1 . Detailed allele and haplotype frequencies per ethnic subgroup derived from 1000 Genomes Project, for rs12252 and rs34481144.</ns0:head><ns0:label>1</ns0:label><ns0:figDesc /><ns0:table><ns0:row><ns0:cell>T A_T</ns0:cell><ns0:cell>A_C</ns0:cell><ns0:cell>G_C</ns0:cell></ns0:row></ns0:table></ns0:figure> <ns0:note place='foot'>PeerJ reviewing PDF | (2020:08:52069:2:0:NEW 27 Oct 2020)</ns0:note> </ns0:body> "

"Dr. Joseph J Gillespie Academic Editor for PeerJ Dear Dr. Gillespie, Thank you very much for your decision on our PeerJ submission entitled “The frequency of combined IFITM3 haplotype involving the reference alleles of both rs12252 and rs34481144 is in line with COVID-19 standardized mortality ratio of ethnic groups in England” (#2020:08:52069:2:0:NEW). We would also like to thank Dr. Wellington for her very helpful and constructive comments and suggestions, in order to significantly improve our manuscript. Please find detailed below minor revisions as suggested, or replies to the reviewer’s comments. 1) We reworked the titles and legends of most Figures and Tables. Also, we added three sentences between lines 143-147 (Results section), which describe more clearly the results that derive from figure 1A (and removed the respective comments from Figure 1A legend). Regarding the legend of Figure 2, we feel that the description of ratio/frequency ranking “leaders” serves the acquaintance of the reader with the Figure, rather than presenting results. The actual result linked to Figure 2, specifically that SMR is aligned with h1 ratios, is presented in both the Results section (lines 146-147) and the legend. However, if it is judged that this description is not necessary in the Figure legend, please feel free to remove it, as the Figure 2 title is self-explanatory. 2) We further elaborated our rationale regarding the inconclusiveness of n=80 COVID-19 cohort with Chinese patients (Zhang et al., 2020), with respects to the expectation that rs12252 heterozygosity would behave similarly to the wild type homozygous genotype (lines 276-281). We also followed the reviewer’s suggestion to add an acknowledgement in the Discussion (lines 282-285) of the fact that our study takes into consideration allele frequencies and not genotyped data. Your sincerely, Dr. D.Nikoloudis On behalf of Dimitris Nikoloudis, Dimitrios Kountouras and Asimina Hiona. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Medical imaging is an important approach for the diagnosis of hepatocellular carcinoma (HCC), a common life threaten disease, however, the diagnostic efficiency is still not optimal. Developing a novel method to improve diagnosis is necessary. The aim of this project was to formulate a material that can combine with GPC3 of HCC for targeted enhanced ultrasound. Methods. A material of sulfur hexafluoride (SF 6 ) filled liposome microbubbles and conjugated with synthesized peptide (LSPMbs) was prepared and assessed in vitro and vivo. Liposome microbubbles were made of DPPC, DPPG, DSPE-PEG2000,and SF 6 , using thin film method to form shell, followed filling SF 6 , and conjugating peptide. A carbodiimide method was used for covalent conjugation of peptide to LSMbs. Results. The prepared LSPMbs appeared round shaped, with size of 380.9&#177;176.5 nm, and Zeta potential of -51.4&#177;10.4mV. LSPMbs showed high affinity to Huh-7 cells in vitro, presented good enhanced ultrasound effects, did not show cytotoxicity, and did not exhibit targeted fluorescence and enhanced ultrasound in animal xenograft tumors. Conclusion. Extravascular contrast-enhanced ultrasound targeted GPC3 on HCC may not be realized, and the reason may be that targeted contrast agents of microbubbles are hard to access and accumulate in the tumor stroma and matrix.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Hepatocellular carcinoma (HCC) is a common primary malignant neoplasm derived from hepatocytes, especially in some Asia-Pacific regions, where the underline diseases of hepatitis B virus infection and relevant diseases are in high prevalence <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>. At present, the diagnosis of small and atypical HCC is still challenging <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. Healthcare professionals and scientists have been searching new methods to improve diagnosis efficiency. Targeted imaging has been a heat topic and interest of researchers in recent decades, and is expected to be an ideal non-invasive imaging method <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref><ns0:ref type='bibr' target='#b3'>[4]</ns0:ref><ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>. Although the lack of a basement membrane and smooth muscle and the expansion of the intercellular space in cancer vasculature result in a maximum pore size of approximately 380-780 nm, which exhibits leaky and/or defective blood vessels, microbubbles (Mbs) with diameter more than 1000 nm cannot migrate from the tumor vasculature to the cellular target site to exert the desired diagnostic effect <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. Therefore, the development of nanoscale targeted ultrasound contrast agents (UCAs), which may permeate through the tumor vasculature gap and bind to tumor cells, with extravascular imaging function, is required. On HCC cellular membrane, there is a high expression of Glypican-3 (GPC3) protein, which can be used for a target for molecular imaging <ns0:ref type='bibr' target='#b3'>[4,</ns0:ref><ns0:ref type='bibr' target='#b4'>5]</ns0:ref>. However, because of the antigenicity and larger size of GPC3, if it is used for the ligand of a targeted material, the material may not produce desired effect in the vivo. On this condition, if a small size peptide without antigenicity but possessing similar targeting ability, it may be used for the fabrication of a targeted contrast material. We hypothesized that a new material may be fabricated, with function of targeted contrast-enhanced ultrasound (CEU) imaging. Based on previous studies, we established a sort of liposome microbubbles and conjugated with a synthesized peptide targeting GPC3 of the HCC, with liposome as shell and sulfur hexafluoride gas (SF 6 ) as the core <ns0:ref type='bibr' target='#b3'>[4,</ns0:ref><ns0:ref type='bibr' target='#b4'>5]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Ethics statement</ns0:head><ns0:p>All the experimental procedures in this study were in compliance with the National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee of Hainan Medical University (2018-02-27).</ns0:p><ns0:p>&#61623; &#61623; &#61623; preparation. The above processed lipid dispersion was put into a 50 mL Falcon tube, the air above the aqueous dispersion in the tube was replaced with SF 6 gas, and the tube was sealed with parafilm. The temperature of lipid dispersion was heated to 30&#8451;, the homogenizer was operated to create high shear mixing (15000 rpm, 5 min) to form microbubbles. The mixture was centrifuged at 12000 rpm at 4&#8451;for 5 min, washed with deionized water, three times. 15% (w/v) sucrose solution was added to the mixture in a 5:1 volume ratio (mixture: sucrose), small glass vial of 4 mL volume was used for the loading, each with 2 mL mixture. The air in the vials was replaced with SF 6 gas before lyophilized in a -85&#8451; lyophilizer (SP Scientific, VirTis, USA).</ns0:p><ns0:p>After 24 h completely freeze-drying, vials were refilled with SF 6 gas and sealed, stored at 4&#8451;. Pure liposomes (LS) were prepared using the above protocols without filling SF 6 gas.</ns0:p></ns0:div> <ns0:div><ns0:head>Preparation of GPC3 Targeted Liposome Microbubbles</ns0:head><ns0:p>Firstly, after liposome microbubbles (LSMbs) were prepared using the protocols above, but did not add sucrose solution and lyophilize. Next, a carbodiimide method was used for covalent conjugation of the synthesized peptide to the free carboxyl groups on the surface of LSMbs. The prepared LSMbs were resuspended in MES buffer (0.1 mol/L, pH 6.0), and an adequate amount The mixture was centrifuged at 12000 rpm at 4&#8451;for 5 min, washed by deionized water, three times. Next, 15% (w/v) sucrose solution was added to the mixture in a 5:1 volume ratio (mixture: sucrose), small glass vial of 4 mL volume was used for the loading, each with 2 mL mixture. The air in the vials was replaced with SF 6 gas before lyophilized in a -85&#8451; lyophilizer (SP Scientific, VirTis, USA). After 24 h completely freeze-drying, vials were refilled with SF 6 gas and sealed, stored at 4&#8451;. Liposome microbubbles (LSMbs), and GPC3 antibody (BM1846; Wuhan Boster Biological Technology, Ltd., Wuhan, China) conjugated liposome microbubbles (LSGMbs) were prepared in the similar protocol above.</ns0:p></ns0:div> <ns0:div><ns0:head>Characterization of LSPMbs Transmission Electron Microscopy Evaluation</ns0:head><ns0:p>The LSPMbs were observed using transmission electron microscope (TEM) for particle size and shape assessment. LSPMbs were suspended with deionized water (1:50), and one drop of the suspension was dropped onto a carbon-coated copper grid. After drying and adhesion in 25&#8451;, samples were negatively stained by sodium phosphotungstate solution (2%, w/w) and analyzed with a 120-kV TEM (TEM; JEM 2100, JEOL, Tokyo, Japan). Suspensions of samples with different concentration were dropped on coverslips and observed under light microscopy.</ns0:p></ns0:div> <ns0:div><ns0:head>LSPMbs Size and Zeta Potential Measurements</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>LSPMbs size and Zeta potential of each sample was measured using a Zetasizer Nano S90 (Malvern Instruments Ltd., Malvern, UK) by Laser Doppler Anemometry (LDA) using electropheoretic light scattering at 25&#8451;. An adequate amount of LSPMbs was suspended and diluted to enable the microbubbles concentration was maintained to ensure that multiple scattering and microbubble-microbubble interactions were negligible. The microbubble size and zeta potential of each sample were measured three times, and the mean value was taken as the final microbubble size and zeta potential. LSGMbs were assessed in the same methods.</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Biocompatibility of LSPMbs</ns0:head></ns0:div> <ns0:div><ns0:head>MTT Assay for Cytotoxicity</ns0:head><ns0:p>RAW 264.7 cells [Cell bank of the Chinese academy of sciences (Shanghai, China)] in exponential phase of growth were taken, transferred 200 &#61549;L to each well of a 96-well plate, and adjust the cell density to 5000/well. The cells were cultured with DMEM (Wuhan Boster Biological Technology, Ltd., Wuhan, China), 5% fetal bovine serum (FBS) (Gibco, Australia), and 1% penicillin-streptomycin at 37 &#8451; in 5%CO 2 atmosphere for 24 h, when the well were fully covered, added different concentration gradients LSPMbs (2mg/mL, 5mg/mL, 10mg/mL, 15mg/mL, and 20mg/mL) prepared using with DMEM and 5% FBS to different wells, 200 &#61549;L per well, then continuing incubated for 24h. LSPMbs were burst using ultrasound at the experiment. Next, 10 &#61549;L 0.5% 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) solution was added to each well, and continuing incubated for 24h, 48h and 72h in the dark place, respectively. The incubation was ended at 24h, 48h and 72h, respectively, and the solution in the wells were carefully absorbed out and discarded. Next, 200&#61549;L dimethyl sulfoxide was added to each well, and oscillated at a frequency of 20 times per minute for 10min to fully dissolve the crystal. Optical density (OD) is measured at the 570 nm wavelength by using a spectrophotometric microplate reader (Bio-Tek ELX-800, Winooski, VT, USA).</ns0:p><ns0:p>Controls were established and processed using the similar protocols.</ns0:p></ns0:div> <ns0:div><ns0:head>Evaluation of Enhanced Ultrasound Imaging of LSPMbs in Vitro</ns0:head><ns0:p>To assess the contrast-enhanced effect of LSPMbs, LSPMbs were suspended with deionized water and diluted into various concentrations (1.6, 0.8, and 0.4mg/mL) and placed in different plastic tubes for ultrasound evaluation. LSPMbs suspension filled tubes were fixed in a water container, and their CEU effect was assessed using GE Logiq E9 ultrasound system (GE Healthcare, Milwaukee, WI, USA), using a ML6-15-D linear transducer with a frequency of 4-15 MHz. During the ultrasound performance, the frequency of the transducer was set to 12 MHz, the depth and focus were adjusted to optimize imaging, and the model was shifted to contrast imaging, using the default parameter (MI 0.1). Before the ultrasound scanning, the tubes were agitated slightly. Controls of commercial ultrasound contrast agent SonoVue (Shanghai Bracco Sine Pharmaceutical Corp. Ltd., Shanghai, China), degas deionized water, and air were established, and these were assessed using the same protocols above. Manuscript to be reviewed Photoshop, activated menus of 'Window, Information, and Histogram' consecutively, selected 'rectangle, and statistics display ' tools, set the same region of interest to the tube, parameters yielded automatically, measured three times in each image, and adopted the mean value of scales (arbitrary units) as the result of a single image.</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Affinity of LSPMbs to the Liver Cancer Cells</ns0:head><ns0:p>Fluorescence experiment was used for the assessment of affinity of LSPMbs to the liver cancer cells. A tiny amount of 1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate (DiI) (Yeasen Biotech Co., Ltd., Shanghai, China) was added into 5mL of <ns0:ref type='bibr' target='#b17'>16</ns0:ref> Manuscript to be reviewed (10&#181;g/mL) was applied before adding DiI labelled LSPMbs as a blocking control, and the other protocols were the same as the above.</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of LSPMbs in Vivo</ns0:head><ns0:p>All animal experimental procedures were approved by the Our University Association for Accreditation of Laboratory Animal Care. Animals of female BALB/C mice were used for the evaluation of targeting ability and contrast-enhanced effect. Twenty Huh-7 cell xenograft tumor models of female BALB/C mice were established, and the Huh-7 cell line was acquired from the cell bank of the Chinese academy of sciences (Shanghai, China). All animals expect four mice were sacrificed by euthanasia using isoflurane after fluorescent imaging were sacrificed using carbon dioxide in a closed box at the end of the animal experiments. The criteria of animal death are that the mice were in collapse state, losing muscular tension, no breath, no heart beating, and the skin color became gray.</ns0:p></ns0:div> <ns0:div><ns0:head>Enhanced Ultrasound Imaging Assessment of LSPMbs</ns0:head><ns0:p>Fifteen of the mice were used for the targeted CEU experiments in five groups, with each group of three mice. The CEU effect was assessed using GE Logiq E9 ultrasound system (as has addressed in the previous section). During the ultrasound scanning, the frequency of the transducer was set to 12 MHz, the depth and focus were adjusted to optimize imaging, and the model was shifted to contrast imaging, using the default parameter (MI 0.1). The shape of the xenograft tumors was ovoid, and the longitudinal diameter of the tumors of 24 mice was 10.4&#177;0.53 mm in the fifth week after cells seeded. Control experiments were conducted in four Manuscript to be reviewed groups using LSGMbs, LSPMbs (GPC3 antibody or synthesized peptide blocked previously), SonoVue (a commercial ultrasound contrast-enhanced agent), and LS, respectively. LSPMbs suspension was prepared using 14mg LSPMbs and 4 mL 0.9% sodium chloride solution, agitated slightly before injection. The animals with Huh-7 xenograft tumor in five groups were intravenously injected suspension of LSPMbs, LSGMbs, LSPMbs (injected 0.1mL synthesized peptide or 0.1mL dilated GPC3 antibody for blocking in 3 minutes ahead), SonoVue (15mg SonoVue in 4mL 0.9% sodium chloride solution), and LS (14mg LS and 4 mL 0.9% sodium chloride solution, agitated slightly before injection), respectively, each with the volume of 0.2mL.</ns0:p><ns0:p>The contrast imaging time was counted since the bolus intravenous injection of 0.2mL LSPMbs suspension via the mouse tail vein. The images were saved in the ultrasound system and exported for study late.</ns0:p><ns0:p>Enhancement ultrasound imaging effects of LSPMbs, LSGMbs after synthesized peptide or GPC3 antibody blocking, SonoVue, LS were determined using Photoshop software, and the methods had been addressed in the previous section.</ns0:p></ns0:div> <ns0:div><ns0:head>Fluorescent Imaging Assessment of LSPMbs</ns0:head><ns0:p>Ten mice with xenograft tumors of Huh-7 cells allotted to two groups of five each were used to Manuscript to be reviewed Jinan, China). 0.2mL Cy5.5 labelled LSPMbs suspension was intravenously injected into six mice with Huh-7 xenograft tumors via the tail veins, images were acquired at one minute, six hours, and 24 hours. Cy5.5 labelled LSPMbs suspension was prepared using 14mg Cy5.5 labelled LSPMbs and 4 mL 0.9% sodium chloride solution, agitated slightly before injection.</ns0:p><ns0:p>The control experiments were conducted in five mice with Huh-7 xenograft tumors, with the same methods after injection of 0.1mL synthesized peptide (10&#181;g/mL) for blocking in three minutes. Four mice were sacrificed by euthanasia using 3m L isoflurane in a closed small box.</ns0:p><ns0:p>The tumor, liver, heart, lung, kidney, and spleen of the mice were isolated for fluorescent imaging assessment at 24hours.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Quantitative data are presented as mean &#177; SD (standard deviation) , and qualitative data are presented as percentile. Statistical significance of differences between groups of quantitative variables were analyzed using paired-sample t tests or univariate analysis of variance, and qualitative variables were analyzed using Chi-square test. All statistical analyses were performed using SPSS software (Version 20; IBM, Armonk, NY, USA). P &lt; 0.05 was considered significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Characterization of LSPMbs</ns0:head><ns0:p>The LSPMbs appeared round shaped on a sectional view, with different size, without aggregation, identified by transmission electron microscopy (Figure <ns0:ref type='figure' target='#fig_16'>1</ns0:ref>). Manuscript to be reviewed Size and Zeta potential of LSPMbs were 380.9&#177;176.5 nm and -51.4&#177;10.4mV, respectively. The determination results showed 'Good' (Figures <ns0:ref type='figure' target='#fig_20'>2 and 3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Biocompatibility of LSPMbs</ns0:head></ns0:div> <ns0:div><ns0:head>MTT Assay for Cytotoxicity</ns0:head><ns0:p>The cell index had no significant differences among different LSPMbs concentrations at different time (all P&gt;0. 05), indicating that LSPMbs did not cause significant toxic effect on RAW 264.7 cells. As shown on Figure <ns0:ref type='figure' target='#fig_21'>4</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Evaluation of Enhanced Ultrasound Imaging of LSPMbs in Vitro</ns0:head><ns0:p>LSPMbs and SonoVue suspension with different concentrations presented different enhancement effects, they presented almost the same enhancement effect at the same concentration (Figures <ns0:ref type='figure' target='#fig_10'>5</ns0:ref> A and D were obtained from 1.6mg/mL, figures 5B and E was obtained from 0.8mg/mL, and figures 5C and F was obtained from 0.4mg/mL). At higher concentration (1.6mg/mL), they all presented homogeneous hyperechogenicity with marked attenuation (Figures <ns0:ref type='figure' target='#fig_23'>5A and D</ns0:ref>), and the echogenicities became weaker when the concentrations decreased (Figures <ns0:ref type='figure' target='#fig_23'>5B and E</ns0:ref>, and figures 5C and F), and they all much stronger than control of degas deionized water, which presented homogeneous anechogenicity (Figure <ns0:ref type='figure' target='#fig_23'>5G</ns0:ref>). Control of air presented strong echogenicity at the interface of the tube, and the distal field presented marked attenuation (Figure <ns0:ref type='figure' target='#fig_23'>5H</ns0:ref>), which was substantial different from those obtained from LSPMbs and SonoVue suspension. </ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Affinity of LSPMbs to the Liver Cancer Cells</ns0:head><ns0:p>The fluorescence on the cellular membrane of Huh-7 cells was intensive (Figure <ns0:ref type='figure' target='#fig_24'>6</ns0:ref>), indicating that there was high GPC3 expression. Cells and DiI labled LSPMbs and controls obtained from light microscope (Figure <ns0:ref type='figure' target='#fig_25'>7</ns0:ref> A, E and I). Cell nucleus of Huh-7 cells presented blue after DAPI staining and being incubated with DiI labled LSPMbs (Figure <ns0:ref type='figure' target='#fig_25'>7</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of LSPMbs in Vivo Enhanced Ultrasound Imaging Assessment of LSPMbs</ns0:head><ns0:p>All xenograft tumors in the mice of the five groups presented a complex of isoechogenicity, hypoechogenicity and anechogenicity ( <ns0:ref type='figure'>Figures A, E, I, M and Q</ns0:ref> These indicate that LSPMbs has good capability in CEUS imaging, but the experiments of it did not show targeted imaging in vivo.</ns0:p></ns0:div> <ns0:div><ns0:head>Fluorescent Imaging Assessment of LSPMbs</ns0:head><ns0:p>The fluorescent signal could be visualized all over the body soon after the administration of Manuscript to be reviewed could not be visualized in the mice 24 hours after injection (Figure <ns0:ref type='figure' target='#fig_31'>9C</ns0:ref>). Experiment carried out in the mice blocked previously by injection of GPC3 antibody or synthesized peptide presented the same fluorescent imaging characteristics as those in the mice with Huh-7 xenograft tumors (Figures <ns0:ref type='figure' target='#fig_31'>9D, E and F</ns0:ref>). 24 hours after intravenous injection of Cy5.5 labelled LSPMbs, four mice with Huh-7 cell xenograft tumors of two in each groups were sacrificed, the tumors and visceral organs were assessed, there were fluorescence in the lungs and liver, and no fluorescence in the tumor and the heart, spleen and kidneys (Figure <ns0:ref type='figure' target='#fig_32'>10</ns0:ref>). These indicate that the Cy5.5 labelled LSPMbs did not selective accumulated in the xenograft tumor, and the LSPMbs did not present detectable targeting ability to the tumor. The reason that the fluorescent signal intensity in the liver and spleen was higher than other areas is believed that the liver and spleen have abundant capillaries and macrophage cells, the macrophage cells can engulf the liposomes, so liposomes in these regions are richer than other regions. The more Cy5.5 labelled LSPMbs aggregated, the stronger the fluorescent signal intensity.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>LSPMbs presented similar enhancement effect at the same concentration as that the SonoVue performed in vitro and vivo, indicating that the LSPMbs has good capability of enhancement imaging. Optical imaging in vivo using fluorescence and bioluminescence has high sensitivity and resolution <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. The near infrared dye Cy5.5 allows a fluorescent imaging of deep tissue in rodent animals with low background, providing a possibility of evaluation of the molecular imaging agents. In this study, LSPMbs were labelled with Cy5.5 for targeting GPC3 imaging Manuscript to be reviewed evaluation, the results showed that they had not presented aggregated fluorescence imaging, indicating that LSPMbs were not targeting retained in the tumor. LSPMbs did not present targeted imaging both at ultrasound imaging and optical imaging.</ns0:p><ns0:p>Unlike iodinated and gadolinium contrast agents for x-ray, CT and MRI that can enter extravascular tissue, common UCAs are confined to the blood pool when administered intravenously, which are consist of microbubbles in suspension which strongly interplay with the ultrasound beam and are readily detectable by ultrasound imaging systems <ns0:ref type='bibr' target='#b7'>[7]</ns0:ref>. Molecularly targeted UCAs are created by conjugating the microbubble shell with a peptide, antibody, or other ligand designed to target an endothelial biomarker associated with tumor angiogenesis or inflammation. These microbubbles then accumulate in the microvasculature at target sites where they can be imaged <ns0:ref type='bibr' target='#b7'>[7]</ns0:ref>.</ns0:p><ns0:p>Our findings were significantly different from previous reports that vascular endothelial growth factor receptor 2 (VEGFR2) based targeted UCAs <ns0:ref type='bibr' target='#b8'>[8]</ns0:ref>. These targeted UCAs do not need to extravasate the blood vessels. The VEGFR2 based targeted UCAs can contact and combine with the VEGFR2 on the blood vessels when they enter and flow through the tumor's vessels, forming focal contrast agent accumulation, and can display focal enhanced imaging at ultrasound scanning. However, our preparation of LSPMbs targeting cellular membranous receptor of the HCC confronts substantial challenge for targeted imaging. Sizes of LSPMbs are 380.9&#177;176.5 nm, which can extravasate the fissure of blood vessel wall of tumor if there are no other impeding factors. But the experimental results did not gain the desired goal. The reasons may be the Manuscript to be reviewed following factors. The previous study showed that the gap between the epithelia of cancer blood vessel is large enough (380-780 nm) to allow nanoscale materials passing, but the blood vessels contact the cells of tumor and interstitial closely, elevated interstitial fluid pressure in the tumor could restrict convective flow and antibody extravasation, except in large necrosis and hypoxia areas <ns0:ref type='bibr' target='#b9'>[9,</ns0:ref><ns0:ref type='bibr' target='#b10'>10]</ns0:ref>. Similarly, the targeted LSPMbs needs overcome interface pressure gradient and get enough space to access and bind to the cancer cells of tumor. How the LSPMbs penetrate the blood vessels of tumor, distribute in the tumor and uptake by the cells are difficult to understand or predict <ns0:ref type='bibr' target='#b11'>[11]</ns0:ref>. A study by Opie showed that the osmotic parameters of tumors (hepatoma, etc) are much lower than that of normal tissues <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>. In this circumstance, the suspension of LSPMbs is harder to be absorbed into tissue by osmotic force of tumor. At tumor sites, the disorganized tumor vascular network, extensively distributed stromal cells (e.g., tumor-associated macrophage, cancer-associated fibroblasts, etc.) and the dense physical barriers of extracellular matrix comprise of the abominable obstacles hampering nanoparticles transport in a tumor. <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref> The electron microscopy results on a study have confirmed that the opening in tumor extracellular matrix barriers surrounding the cancer cells is generally less than 40 nm. <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> On this condition, LSPMbs of 380.9&#177;176.5 nm are impossible to pass the extracellular matrix opening to access to the cancer cells. A recent study revealed that only 0.7% of systemically administered nanoparticles can reach the tumor sites and less than 14 out of 1 million (0.0014% injected dose) of them are accessed by cancer cells, and that only 2 out of 100 cancer cells interacted with the nanoparticles. <ns0:ref type='bibr' target='#b16'>[15]</ns0:ref> Therefore, if the number and volume of the targeted LSPMbs entering the tumor extravascular part are not enough, it will be impossible to yield visible enhanced imaging Manuscript to be reviewed effect.</ns0:p><ns0:p>Many researches on extravascular targeted contrast-enhanced ultrasound have been reported in literature, but only a few of them validate in vivo of animals. Mai et al reported that a chitosan-vitamin C lipid system had been fabricated and had achieved tumor-selective enhanced ultrasound imaging in a mouse tumor model, but they confirmed only that the fluorescence accumulated highly at the tumor site, other than the nanobubbles <ns0:ref type='bibr' target='#b17'>[16]</ns0:ref>. Another extravascular targeted nanobubbles fabricated by Gao et al remains to be further verification because of the preliminary results and substantial limitations <ns0:ref type='bibr' target='#b18'>[17]</ns0:ref>. Theoretically, if microbubbles enter the tumor interstitials, some liquid solution also enters. There must be enough extravascular space in the tumor to contain and distribute them, only in this condition can the microbubbles in the oscillations of ultrasonic compression and expansion wave generate stronger backscattered acoustic signal and second harmonics for enhanced ultrasound imaging. If many microbubbles are compacted together, their size will be big, and which will generate little backscattered acoustic signal and second harmonics <ns0:ref type='bibr' target='#b19'>[18]</ns0:ref>.</ns0:p><ns0:p>Our experimental results, together with earlier published reports by others <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref><ns0:ref type='bibr' target='#b14'>[14]</ns0:ref><ns0:ref type='bibr' target='#b16'>[15]</ns0:ref>, strongly suggested that to develop a targeted material for extravascular contrast-enhanced ultrasound imaging, cutting-edge precisive experiments should be conducted firstly to ascertain that the material can penetrate the blood vessel and wade through the cellular matrix and stroma to access and bind to the target cells, and the accommodation space for the materials is adequate. In future, the development of materials for extravascular contrast-enhanced ultrasound imaging Manuscript to be reviewed may be emphasis that using specific materials such as cell-penetrating peptides, a disulfidebridged cyclic RGD peptide, named iRGD (internalizing RGD, c(CRGDK/RGPD/EC)), which is a tumor-homing peptide that can bind to avb3 integrin with high affinity and specificity to construct the targeted material. A material integrated iRGD peptide may increase penetration of the blood vessels and matrix, and facilitate accumulation and increase the probability of enhanced imaging <ns0:ref type='bibr' target='#b20'>[19,</ns0:ref><ns0:ref type='bibr' target='#b22'>20]</ns0:ref>. Augmentation of enhanced permeability and retention effect of targeted material through using NO-releasing agent such as nitroglycerin or angiotensinconverting enzyme inhibitors, and albumin-protein interactions using S -nitrosated human serum albumin dimer, etc, may increase targeted material accumulation and the probability of enhanced imaging <ns0:ref type='bibr' target='#b23'>[21,</ns0:ref><ns0:ref type='bibr' target='#b25'>22]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Collectively, a new material of LSPMbs has been prepared, which has good effect of enhanced ultrasound imaging, but it did not exhibit targeted imaging effect in vivo of animal experiments. The causes may be that the volume of LSPMbs pass the tumor blood vessels and enter the tumor parenchyma was very limited, and the LSPMbs cannot pass the fissures of extracellular stroma and matrix surrounding the cancer cells to access and bind to the cancer cells. Therefore, a potential target of GPC3 on hepatocellular carcinoma for extravascular targeted imaging may not be realized in contrast-enhanced ultrasound. The future research should focus on that whether a candidate targeted material for extravascular contrast-enhanced ultrasound imaging can penetrate the blood vessel and wade through the cellular matrix and Manuscript to be reviewed stroma to access and bind to the target cells, and whether the accommodation space for the materials is adequate.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Image of the LSPMbs obtained by transmission electron microscope. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 2</ns0:note><ns0:note type='other'>Figure 4 Figure 4</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:note type='other'>Figure 6 Figure 6</ns0:note><ns0:note type='other'>Figure 7</ns0:note><ns0:note type='other'>Figure 9</ns0:note><ns0:note type='other'>Figure 10</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>of EDC/NHS [1-ethyl-3(3-dimethylaminopropyl) carbodiimide (EDC): N-hydroxy succinimide (NHS) = 1:4, Sigma-Aldrich Chemical Co., Inc, USA] were added into the suspension. The mixture suspension was oscillated and incubated for 2 h at room temperature (25&#8451;). The remaining EDC/NHS was removed by three-time centrifugation at 12000 rpm using MES (pH 6.0), 5 min each time. The precipitate was dispersed into MES buffer (0.1 mol/L, pH 8.0), and an adequate amount of synthesized peptide was added and incubated with stirring for 2 h at room PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020) Manuscript to be reviewed temperature. The peptide was compounded by GL Biochem (Shanghai) Ltd. (Shanghai, China) in accordance to a 12-mer peptide with the sequence of DHLASLWWGTEL reported from previous study that it can target GPC3 of HepG2 [4] [ DOI: 10.1021/acs.bioconjchem.6b00030].</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Enhancement ultrasound imaging effects of LSPMbs, LSGMbs, SonoVue, LS, degas deionized water, and air were determined using Photoshop software (Adobe Photoshop CS3, Adobe Systems Inc, CA, USA). To analyze ultrasound images, observers opened the image in PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>mg LSPMbs suspension to form DiI labelled LSPMbs suspension. Huh-7 cells [Cell bank of the Chinese academy of sciences (Shanghai, China)] were seeded in a six-well plate with one glass slip placed in each well at a concentration of 3&#215;10 4 cells/well. Next day, the Huh-7 cells were fixed by 90% cold ethanol for 20 min and blocked by 10% bovine serum albumin (BSA) at 37 &#176;C for 1 h and subsequently incubated with 4 drops of DiI labelled LSPMbs suspension for 3 h in dark place. Next, the wells and slips were washed with PBS three times, and the cells on the slips were mounted with 4&#8242;, 6-diamidino-2-phenylindole (DAPI, Wuhan Boster Biological Technology, Ltd., China) for nuclei visualization and detected using a laser confocal microscope (Fluoview FV 10001000, Olympus, Japan). Images of bright, DAPI staining, DiI staining, and merged were obtained. LSGMbs were assessed in the same protocols for control. To confirm the specificity of binding of DiI labelled LSPMbs to GPC3 in the Huh-7 cells, 0.1mL synthesized peptide PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>conduct</ns0:head><ns0:label /><ns0:figDesc>Cy 5.5 fluorescence experiment to test the specificity and affinity of the peptide in LSPMbs to GPC3 of the liver cancer. An IVIS Lumina image system (Xenogen) (IVIS &#174; Lumina XR) (Caliper life sciences) was used for the evaluation. During the fluorescence imaging, mice were under gas anesthesia with oxygen and isoflurane (Jinan Shengqi pharmaceutical Co, Ltd., PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>figure C vs. F, P=0.428; and figure A vs. G, A vs. H, B vs. H, G vs. H, A vs. B, A vs. C, B vs. C,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>B, F and J). LSPMbs with DiI staining combined with the membrane of Huh-7 cells presented red color fluorescence on image (Figure 7 C and D); LSGMbs presented a very similar appearance (Figure 7 G and H); and LSPMbs with DiI staining after synthesized peptide blocking previously did not show red color fluorescence on the membrane of Huh-7 cells, indicating that LSPMbs had not combined with the membrane of Huh-7 cells (Figure 7 K and L).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>). Of the study group, after administration of LSPMbs suspension, Huh-7 xenograft tumor presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis) at two seconds (Figure 8 B); the tumor enhancement lasted over 20 seconds with little change; at 60 seconds, the tumor still presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis) (Figure 8 C); at 10 minutes, the tumor presented iso-enhancement with central hypo-enhancement (necrosis) (Figure PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)Manuscript to be reviewed 8 D). Of the four control groups, the mice injected with LSGMbs, LSPMbs (blocked with GPC3 antibody or synthesized peptide previously), and SonoVue, respectively, presented similar enhancement patterns and sustain time as those in Huh-7 xenograft tumors (Figures8 F,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>Figures 8 J, K and L, and Figures 8 N, O and P), and there was no significant difference; the</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Cy5. 5 labelled</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>LSPMbs suspension. Images acquired at one minute (Figure 9A), six hours (Figure 9B) after the initial fluorescent imaging, the fluorescent signal intensity in the tumor area has no significant difference from other areas of the body expect the liver and spleen. The fluorescent signal intensity in the liver and spleen area of the mice was marked stronger than other areas, and the fluorescent signal intensity was similar in three times. The fluorescent signal PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_15'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:1:1:NEW 8 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_16'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1 Image of the LSPMbs obtained by transmission electron microscope. The LSPMbs present round shaped on a sectional view, with different size, without aggregation.</ns0:figDesc><ns0:graphic coords='26,42.52,229.87,525.00,349.50' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_17'><ns0:head /><ns0:label /><ns0:figDesc>Size and distribution of LSPMbs</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_18'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2 Size of LSPMbs are 380.9&#177;176.5 nm, and the determination results show 'Good'.</ns0:figDesc><ns0:graphic coords='27,42.52,204.37,525.00,357.00' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_19'><ns0:head>Figure 3 Zea</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_20'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3 Zea potential of LSPMbs was -51.4&#177;10.4mV, and the determination results show 'Good'.</ns0:figDesc><ns0:graphic coords='28,42.52,229.87,525.00,339.00' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_21'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4 MTT assay for cytotoxicity showed the cell index had no significant differences among different LSPMbs concentrations at different time (all P&gt;0. 05).</ns0:figDesc><ns0:graphic coords='29,42.52,229.87,525.00,217.50' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_22'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_23'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5 Ultrasound images of the LSPMbs, SonoVue, degas deionized water, and air in the tubes. Figures A and D were obtained from LSPMbs and SonoVue of 1.6mg/mL, B and E were obtained from LSPMbs and SonoVue of 0.8mg/mL, and C and F were obtained from LSPMbs and SonoVue of 0.4mg/mL. Figure G was obtained from degas deionized water, and H was obtained from air. The echogenic intensity decreased with decreasing concentrations of LSPMbs, and degas deionized water; the echogenic intensity was strong at the interface between the air in the tube and the outside water, and the echogenicity in the tube was attenuated.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_24'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6 Image of GPC3 expression of Huh-7 cells obtained by confocal laser scanning microscope. The fluorescence on the cellular membrane appears intensive.</ns0:figDesc><ns0:graphic coords='32,42.52,229.87,525.00,492.75' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_25'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_26'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7 Images of Huh-7 cells incubated with DiI lablled LSPMbs and controls obtained by light microscope and confocal laser scanning microscope. (Images of A, E and I) Cells and DiI lablled LSPMbs and controls obtained from light microscope. (Images of B, F and J) Cell nucleus of Huh-7 cells presented blue after DAPI staining and being incubated with DiI lablled LSPMbs. (Images of C and D) LSPMbs with DiI staining combined with the membrane of Huh-7 cells presented red color fluorescence on image. (Images of G and H) LSGMbs presented a very similar appearance. (Images of K and L) LSPMbs with DiI staining after blocked by synthesized peptide did not show red color fluorescence on the membrane of Huh-7 cells, indicating that LSPMbs had not combined with the membrane of Huh-7 cells.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_27'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_28'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_29'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8 Row 1 is a study group, and rows 2-5 are control groups. (Images of A, E, I, M and Q) Images of all xenograft tumors presented a complex of isoechogenicity, hypoechogenicity and anechogenicity obtained by convention ultrasound. (Image B of LSGMbs) After administration of LSPMbs suspension, tumors presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis) at two second. (Image C of LSGMbs) At 60 seconds, the tumor still presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis). (Image D of LSGMbs) At 10 minutes, the tumor presented isoenhancement with central hypo-enhancement (necrosis). (Images F, G and H of LSGMbs) and (Images J, K and L of LSPMbs, after GPC3 blocking) and they presented similar enhancement patterns and sustain time as those in Huh-7 xenograft tumors, and there were no appreciable difference. (Images N, O and P of SonoVue) The xenograft tumors presented similar enhancement patterns and sustaining time as those using LSPMbs after administration of SonoVue suspension. (Images R, S and T of LS) The xenograft tumors did not present enhancement at 2, 20, and 60 seconds, and 10 minutes after administration of LS suspension.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_30'><ns0:head>Figure 9</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_31'><ns0:head>Figure 9</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9 Image A acquired at one minute, and image B acquired six hours after the initial fluorescent imaging, the fluorescent signal intensity in the tumor area has no significant difference from other areas of the body expect the liver and spleen. Image C acquired 24 hours after injection, the fluorescent signal could not be visualized in the mice. Images D, E and F acquired from experiments of mice with Huh-7 xenograft tumors blocked using GPC3 antibody presented the same fluorescent imaging characteristics as those in the mice with Huh-7 xenograft tumors.</ns0:figDesc><ns0:graphic coords='37,42.52,357.37,525.00,351.75' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_32'><ns0:head>Fluorescence in the organsFigure 10</ns0:head><ns0:label>10</ns0:label><ns0:figDesc>Figure 10 There were fluorescence in the lungs and liver, and no fluorescence in the tumor, heart, spleen, and kidneys 24 hours after intravenous injection of Cy5.5 labelled LSPMbs, after animal sacrificed.</ns0:figDesc><ns0:graphic coords='38,42.52,255.37,525.00,268.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='34,42.52,70.87,525.00,344.25' type='bitmap' /></ns0:figure> </ns0:body> "

"Response Letter Dear Reviewers and Editors, We read your comments with due respect and really appreciated your views. According to your comments and suggestion, the manuscript has been revised, and the response was given point to point. However, due to our limited academic level, there will be still drawbacks and mistakes in the revised manuscript. We are looking forward to hearing good news. Sincerely, Size Wu On behalf of all authors Responses to Reviewer's Comments Reviewer 1 Comments: 2. Materials and Methods, Line 70. In this study, there is only one image intensifier material, and there is no description of the selection from multiple candidates during the development process. Response: We agree with you. A complete paper should include the hypothesis, candidates seeking, screening experiments, optimization, validation, etc. However, according to the present journal requirements of methods and limitation of manuscript words, we had not included other content into the manuscript. We believed that absence of omitted part may not affect the general quality of such article. In fact, we had designed and selected materials and prescriptions with reference to other literature, and selected the present prescription according to the optimizing microbubbles’ enhancement effects, microbubble size, stability, etc. Comments: 3. Materials and Methods, Line 71. There is no description about the reason for selecting the gas (SF6) to be enclosed in the microbubbles. Response: Thank you for your comments. SF6 (sulfur hexafluoride) is a sort of inflammable inert gas, without toxicity in a certain volume, it can be eliminated through breath completely. SF6 has been used for the preparation of commercial contrast agent (SonoVue; Bracco Ltd, Italy) for many years. SF6 is familiar to our peers (sonographers, sonologists), so we had not addressed it in the manuscript. Comments: 4. Materials and Methods, Line 72. There is no detailed description of GPC3 antibody and peptides that bind to the microbubbles. Response: We agree with you. Thank you for your comments. The description of GPC3 antibody and peptides were added in the “Introduction” part, highlighted. Comments: 5. Materials and Methods Line 130. There is no description of the experiment on the particle size of liposome microbubbles filled with SF6, considering only the change in the solution concentration. Response: Thank you for your comments. The sizes of LSPMbs were 380.9±176.5 nm, which was listed at the “Results” part. Validity of the findings The following issues need to be fixed. Comments: 1. Figure 4. The concentration in the right column is unknown. After 72 hours, there is a slight decrease in the Cell Index. In the safety evaluation, it is necessary to extend the observation period. Response: Thank you for your comments. We agree with you of the losing of the right column, and we get it emerging now. The RAW 264.7 cells grew fast, so it is not convenient to extend the observation period. We performed several time experiments, the results were similar, and it did not shew toxicity. In other literature, 72 hours observation is also a frequently used time. The general project and experiments have ended, and we are not convenient to make up an experiment for it. Comments: 2. Figure 5-8. Please insert a size bar on each image. Response: We agree with you. A size bar was inserted to Figures 6 and 7, Figure 5 and 8 are not suitable to add a size bar, for the images in different plots are not a miniature of the original images, and they are not in the same proportion. Comments: 3. Figure 6. The explanation of “The fluorescence on the cellular membrane appears stronger than that in cellular plasma. ”. These findings cannot be recognized in the image. Response: We agree with you. We replaced it with “The fluorescence on the cellular membrane appears intensive”. Because the fluorescence was intensive, and we had not been able to adjust it to an adequate level, so the fluorescence on cellular membrane and cellular plasm were mixed together, and hard to be discerned. Comments: 4. Figures 7. Fluorescence images cannot be recognized in photographs. It is better to replace images with Supplement Figure images. Response: We agree with you. According to the requirements of journal, original figures were processed into different style, as a result, the color and resolution changed, becoming blurring. The original figures were the same as those of the Supplement Figure images, and they had been spoiled in the computer. Comments: 5. Figure 8, There is no explanation for 3rd row. And the initial elapsed time is missing in this figure. Response: Thank you for your comments. The explanation had been listed in the text. As are===Of the four control groups, the mice injected with LSGMbs, LSPMbs (blocked with GPC3 antibody or synthesized peptide previously), and SonoVue, respectively, presented similar enhancement patterns and sustain time as those in Huh-7 xenograft tumors (Figures 8b-d in 2nd row, Figures 8b-d in 3rd row, and Figures 8b-d in 4th row), and there was no significant difference;== Comments: 6. Figure 10. In the safety evaluation of systemic administration (intravenous administration), evaluations of the pulmonary embolus are necessary. A histopathological image of the organs including the lungs should be presented. Response: Thank you for your comments. The aim of Figure 10 was to present whether the LSPMbs had target imaging effect or not. This study is of primary study, the evaluations of the pulmonary embolus are overlooked; if further validation was conducted, evaluations of the pulmonary embolus are necessary. Comments: 7. Supplement Figure. Image of xenograft tumor of the mouse obtained by duplex ultrasound with a scanner of frequency of 12 MHz Response: Xenograft tumors of the mouse locate at superficial position, and their images were obtained by duplex ultrasound with a scanner of frequency of 12 MHz. Supplement Figures were not to be published. Comments: 8. HCC tumors are rich in blood flow. However, the Doppler echo image of the xenograft tumor in the Supplement Figure does not show a blood flow signal. There are problems in selecting the xenograft tumor. Response: Thank you for your comments. Good question. However, not all HCC with rich blood vessels, color Doppler flow imaging cannot detect low speed blood flow and render signal, so some figures did not show a blood flow signal. Comments: 1. It has little advantage over existing intravenous image intensifier materials for ultrasonography. Response: We agree with you. This article is a report of negative final result. The existing intravenous image intensifier materials for extravascular ultrasonography usually showed “good” results, however, they are not reproducible. Comments: 2. Since this is negative final results, it is necessary to describe details above-mentioned problems for the future developments. Response: We agree with you. Relevant address was added, as highlighted in the text. Comments: 3. As a result of the peer review, a major revision is required. Response: Thank you for your comments. A major revision was done. Reviewer 2 Comments: As the authors noted, the material described in this manuscript appears not to be utilized in clinical setting. Furthermore, the authors did not clearly show the reason of an unsuccessful attempt. In this point, the authors discussed it based on previous literature that suggest small amount of LSPMBs of 380 +/- 176.5 nm cannot access to HCC cells. Response: Thank you for your comments. We did not clearly show the reason of an unsuccessful attempt, we had just delivered a report of our experimental results, and it’s hard to show the reason (need other cutting-edge precisive experiments). We cited other research results to support and explain the phenomena of our negative result. Comments: This reviewer fully respects author’s efforts and the results of this experiment, though still wonders the fundamental meaning of this paper. Response: Thank you for your comments. The fundamental meaning of this paper is that the nanoscale targeting material for contrast-enhanced ultrasound can be fabricated, but it cannot display desired effect of extravascular contrast-enhanced ultrasound imaging. This paper presented a negative result, it may be used for a reference for researchers who are going to engage in such research, and let them have a second think before any project or find an alternative method to develop a novel contrast-enhanced material for extravascular contrast-enhanced ultrasound imaging. In literature, there is little negative report, and some researchers may think that there are chance to conduct such research. Responses to Editor's Comments Comments 1: Tracked Changes Manuscript Source File Please could you upload the manuscript with computer-generated tracked changes, rather than highlighting, to the Revision Response Files section. The reviewers and Academic Editor will want to see all of the changes documented and will normally request it if some changes appear to be missing. Please use the Compare Function in Microsoft Word to track all changes made to the manuscript since the last submission. Our response: I changed the revision. Comments 2: Authors We need proof that this email: [email protected] belongs to the Corresponding Author: Size Wu, at their stated institution. Please do one of the following: • Enter an institutional email for this author and have them confirm authorship using it. • Ask a co-author to email [email protected] with the code: 51803 from their institutional email. • Supply a link to an institutional webpage displaying this author's name and [email protected] email: Enter a note to staff. • Upload a photo/scan of an Institutional ID card that includes the institution name, author's name, and author photo: Upload a Supplemental File. • Upload a photo/scan of a signed letter from a department head (or equivalent, and they cannot be a coauthor) on institution-headed paper that includes: (A) the author’s name, (B) validates the non-institutional email address in use (EMAIL), AND (C) the printed name and title of the department head: Upload a Supplemental File. Our response: Our institution has more than 2500 staff, only senior experts present their profile on website, youngers staff do not to present their profile on the institutional website. Our institute and each department do not have Institutional ID card  and public sharing email address. I provided the snap images and other materials to certify that I’m a faculty of the first affiliated hospital of Hainan medical university. Supplemental materials will be uploaded. The coauthor is a new faculty graduated from our university two years ago, and she has no available information on institutional website. Comments 3: Title Your title does not mention that this work had a negative finding. It will improve the discoverability and readability of your article if you edit your title to state the main finding of your work: • in your manuscript and re-upload it, • and in the metadata. • Our response: We agree with you. Thank you very much. The title has been changed as: Negative outcome in developing nanoscale materials for extravascular ultrasound targeting hepatocellular carcinoma Comments 4: Vertebrate Animal Study Thank you for including the name of the Institutional Animal Care and Use Committee in the Methods. Please also add the approval reference number (2018-02-27) in the Methods section of your manuscript and re-upload your manuscript here. Our response: Thank you very much. It was added. Comments 5: Figures Figures 7, 8, and 9 have multiple parts but each of the individual parts have multiple parts as well and this has resulted in multiple cited parts that correspond to A, B, C, D, etc. Please correct this by relabeling the existing figures so they do not have apparently duplicate labels. In this case: • the 12 parts of Figure 7 should be labeled A-L. • the 20 parts of Figure 8 should be labeled A-T. • the 6 parts of Figure 9 should be labeled A-F. Please provide replacement figures measuring minimum 900 pixels and maximum 3000 pixels on all sides, saved as PNG, EPS, or PDF (vector images only) file format without excess white space around the images. Our response: Thank you very much. The figures were changed according to your requirements. Comments 6: Figure Accessibility For better visibility, we recommend that you use white backgrounds for your figures. Please provide a replacement Figure 4 measuring minimum 900 pixels and maximum 3000 pixels on all sides, saved as PNG, EPS, or PDF (vector images) file format without excess white space around the image. Our response: Thank you very much. The figures’ backgrounds are not able to change, for their original patterns are the styles we had presented. Figure 4 was replaced. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>Background. Medical imaging is an important approach for the diagnosis of hepatocellular carcinoma (HCC), a common life threaten disease, however, the diagnostic efficiency is still not optimal. Developing a novel method to improve diagnosis is necessary. The aim of this project was to formulate a material that can combine with GPC3 of HCC for targeted enhanced ultrasound. Methods. A material of sulfur hexafluoride (SF 6 ) filled liposome microbubbles and conjugated with synthesized peptide (LSPMbs) was prepared and assessed in vitro and vivo. Liposome microbubbles were made of DPPC, DPPG, DSPE-PEG2000,and SF 6 , using thin film method to form shell, followed filling SF 6 , and conjugating peptide. A carbodiimide method was used for covalent conjugation of peptide to LSMbs. Results. The prepared LSPMbs appeared round shaped, with size of 380.9&#177;176.5 nm, and Zeta potential of -51.4&#177;10.4mV. LSPMbs showed high affinity to Huh-7 cells in vitro, presented good enhanced ultrasound effects, did not show cytotoxicity, and did not exhibit targeted fluorescence and enhanced ultrasound in animal xenograft tumors. Conclusion. Extravascular contrast-enhanced ultrasound targeted GPC3 on HCC may not be realized, and the reason may be that targeted contrast agents of microbubbles are hard to access and accumulate in the tumor stroma and matrix.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Hepatocellular carcinoma (HCC) is a common primary malignant neoplasm derived from hepatocytes, especially in some Asia-Pacific regions, where the underline diseases of hepatitis B virus infection and relevant diseases are in high prevalence <ns0:ref type='bibr' target='#b0'>[1]</ns0:ref>. At present, the diagnosis of small and atypical HCC is still challenging <ns0:ref type='bibr' target='#b1'>[2]</ns0:ref>. Healthcare professionals and scientists have been searching new methods to improve diagnosis efficiency. Targeted imaging has been a heat topic and interest of researchers in recent decades, and is expected to be an ideal non-invasive imaging method <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref><ns0:ref type='bibr' target='#b3'>[4]</ns0:ref><ns0:ref type='bibr' target='#b4'>[5]</ns0:ref>. Although the lack of a basement membrane and smooth muscle and the expansion of the intercellular space in cancer vasculature result in a maximum pore size of approximately 380-780 nm, which exhibits leaky and/or defective blood vessels, microbubbles (Mbs) with diameter more than 1000 nm cannot migrate from the tumor vasculature to the cellular target site to exert the desired diagnostic effect <ns0:ref type='bibr' target='#b2'>[3]</ns0:ref>. Therefore, the development of nanoscale targeted ultrasound contrast agents (UCAs), which may permeate through the tumor vasculature gap and bind to tumor cells, with extravascular imaging function, is required. On HCC cellular membrane, there is a high expression of Glypican-3 (GPC3) protein, which can be used for a target for molecular imaging <ns0:ref type='bibr' target='#b3'>[4,</ns0:ref><ns0:ref type='bibr' target='#b4'>5]</ns0:ref>. However, because of the antigenicity and larger size of GPC3, if it is used for the ligand of a targeted material, the material may not produce desired effect in the vivo. On this condition, if a small size peptide without antigenicity but possessing similar targeting ability, it may be used for the fabrication of a targeted contrast material. We hypothesized that a new material may be fabricated, with function of targeted contrast-enhanced ultrasound (CEU) imaging. Based on previous studies, we established a sort of liposome microbubbles and conjugated with a synthesized peptide targeting GPC3 of the HCC, with liposome as shell and sulfur hexafluoride gas (SF 6 ) as the core <ns0:ref type='bibr' target='#b3'>[4,</ns0:ref><ns0:ref type='bibr' target='#b4'>5]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Materials &amp; Methods</ns0:head></ns0:div> <ns0:div><ns0:head>Ethics statement</ns0:head><ns0:p>All the experimental procedures in this study were in compliance with the National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee of Hainan Medical University (2018-02-27).</ns0:p><ns0:p>&#61623; &#61623; &#61623; preparation. The above processed lipid dispersion was put into a 50 mL Falcon tube, the air above the aqueous dispersion in the tube was replaced with SF 6 gas, and the tube was sealed with parafilm. The temperature of lipid dispersion was heated to 30&#8451;, the homogenizer was operated to create high shear mixing (15000 rpm, 5 min) to form microbubbles. The mixture was centrifuged at 12000 rpm at 4&#8451;for 5 min, washed with deionized water, three times. 15% (w/v) sucrose solution was added to the mixture in a 5:1 volume ratio (mixture: sucrose), small glass vial of 4 mL volume was used for the loading, each with 2 mL mixture. The air in the vials was replaced with SF 6 gas before lyophilized in a -85&#8451; lyophilizer (SP Scientific, VirTis, USA).</ns0:p><ns0:p>After 24 h completely freeze-drying, vials were refilled with SF 6 gas and sealed, stored at 4&#8451;. Pure liposomes (LS) were prepared using the above protocols without filling SF 6 gas.</ns0:p></ns0:div> <ns0:div><ns0:head>Preparation of GPC3 Targeted Liposome Microbubbles</ns0:head><ns0:p>Firstly, after liposome microbubbles (LSMbs) were prepared using the protocols above, but did not add sucrose solution and lyophilize. Next, a carbodiimide method was used for covalent conjugation of the synthesized peptide to the free carboxyl groups on the surface of LSMbs. The prepared LSMbs were resuspended in MES buffer (0.1 mol/L, pH 6.0), and an adequate amount The mixture was centrifuged at 12000 rpm at 4&#8451;for 5 min, washed by deionized water, three times. Next, 15% (w/v) sucrose solution was added to the mixture in a 5:1 volume ratio (mixture: sucrose), small glass vial of 4 mL volume was used for the loading, each with 2 mL mixture. The air in the vials was replaced with SF 6 gas before lyophilized in a -85&#8451; lyophilizer (SP Scientific, VirTis, USA). After 24 h completely freeze-drying, vials were refilled with SF 6 gas and sealed, stored at 4&#8451;. Liposome microbubbles (LSMbs), and GPC3 antibody (BM1846; Wuhan Boster Biological Technology, Ltd., Wuhan, China) conjugated liposome microbubbles (LSGMbs) were prepared in the similar protocol above.</ns0:p></ns0:div> <ns0:div><ns0:head>Characterization of LSPMbs Transmission Electron Microscopy Evaluation</ns0:head><ns0:p>The LSPMbs were observed using transmission electron microscope (TEM) for particle size and shape assessment. LSPMbs were suspended with deionized water (1:50), and one drop of the suspension was dropped onto a carbon-coated copper grid. After drying and adhesion in 25&#8451;, samples were negatively stained by sodium phosphotungstate solution (2%, w/w) and analyzed with a 120-kV TEM (TEM; JEM 2100, JEOL, Tokyo, Japan). Suspensions of samples with different concentration were dropped on coverslips and observed under light microscopy.</ns0:p></ns0:div> <ns0:div><ns0:head>LSPMbs Size and Zeta Potential Measurements</ns0:head><ns0:p>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:p></ns0:div> <ns0:div><ns0:head>Manuscript to be reviewed</ns0:head><ns0:p>LSPMbs size and Zeta potential of each sample was measured using a Zetasizer Nano S90 (Malvern Instruments Ltd., Malvern, UK) by Laser Doppler Anemometry (LDA) using electropheoretic light scattering at 25&#8451;. An adequate amount of LSPMbs was suspended and diluted to enable the microbubbles concentration was maintained to ensure that multiple scattering and microbubble-microbubble interactions were negligible. The microbubble size and zeta potential of each sample were measured three times, and the mean value was taken as the final microbubble size and zeta potential. LSGMbs were assessed in the same methods.</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Biocompatibility of LSPMbs</ns0:head></ns0:div> <ns0:div><ns0:head>MTT Assay for Cytotoxicity</ns0:head><ns0:p>RAW 264.7 cells [Cell bank of the Chinese academy of sciences (Shanghai, China)] in exponential phase of growth were taken, transferred 200 &#61549;L to each well of a 96-well plate, and adjust the cell density to 5000/well. The cells were cultured with DMEM (Wuhan Boster Biological Technology, Ltd., Wuhan, China), 5% fetal bovine serum (FBS) (Gibco, Australia), and 1% penicillin-streptomycin at 37 &#8451; in 5%CO 2 atmosphere for 24 h, when the well were fully covered, added different concentration gradients LSPMbs (2mg/mL, 5mg/mL, 10mg/mL, 15mg/mL, and 20mg/mL) prepared using with DMEM and 5% FBS to different wells, 200 &#61549;L per well, then continuing incubated for 24h. LSPMbs were burst using ultrasound at the experiment. Next, 10 &#61549;L 0.5% 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) solution was added to each well, and continuing incubated for 24h, 48h and 72h in the dark place, respectively. The incubation was ended at 24h, 48h and 72h, respectively, and the solution in the wells were carefully absorbed out and discarded. Next, 200&#61549;L dimethyl sulfoxide was added to each well, and oscillated at a frequency of 20 times per minute for 10min to fully dissolve the crystal. Optical density (OD) is measured at the 570 nm wavelength by using a spectrophotometric microplate reader (Bio-Tek ELX-800, Winooski, VT, USA).</ns0:p><ns0:p>Controls were established and processed using the similar protocols.</ns0:p></ns0:div> <ns0:div><ns0:head>Evaluation of Enhanced Ultrasound Imaging of LSPMbs in Vitro</ns0:head><ns0:p>To assess the contrast-enhanced effect of LSPMbs, LSPMbs were suspended with deionized water and diluted into various concentrations (1.6, 0.8, and 0.4mg/mL) and placed in different plastic tubes for ultrasound evaluation. LSPMbs suspension filled tubes were fixed in a water container, and their CEU effect was assessed using GE Logiq E9 ultrasound system (GE Healthcare, Milwaukee, WI, USA), using a ML6-15-D linear transducer with a frequency of 4-15 MHz. During the ultrasound performance, the frequency of the transducer was set to 12 MHz, the depth and focus were adjusted to optimize imaging, and the model was shifted to contrast imaging, using the default parameter (MI 0.1). Before the ultrasound scanning, the tubes were agitated slightly. Controls of commercial ultrasound contrast agent SonoVue (Shanghai Bracco Sine Pharmaceutical Corp. Ltd., Shanghai, China), degas deionized water, and air were established, and these were assessed using the same protocols above. Manuscript to be reviewed Photoshop, activated menus of 'Window, Information, and Histogram' consecutively, selected 'rectangle, and statistics display ' tools, set the same region of interest to the tube, parameters yielded automatically, measured three times in each image, and adopted the mean value of scales (arbitrary units) as the result of a single image.</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Affinity of LSPMbs to the Liver Cancer Cells</ns0:head><ns0:p>Fluorescence experiment was used for the assessment of affinity of LSPMbs to the liver cancer cells. A tiny amount of 1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate (DiI) (Yeasen Biotech Co., Ltd., Shanghai, China) was added into 5mL of <ns0:ref type='bibr' target='#b17'>16</ns0:ref> Manuscript to be reviewed (10&#181;g/mL) was applied before adding DiI labelled LSPMbs as a blocking control, and the other protocols were the same as the above.</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of LSPMbs in Vivo</ns0:head><ns0:p>All animal experimental procedures were approved by the Our University Association for Accreditation of Laboratory Animal Care. Animals of female BALB/C mice were used for the evaluation of targeting ability and contrast-enhanced effect. Twenty Huh-7 cell xenograft tumor models of female BALB/C mice were established, and the Huh-7 cell line was acquired from the cell bank of the Chinese academy of sciences (Shanghai, China). All animals expect four mice were sacrificed by euthanasia using isoflurane after fluorescent imaging were sacrificed using carbon dioxide in a closed box at the end of the animal experiments. The criteria of animal death are that the mice were in collapse state, losing muscular tension, no breath, no heart beating, and the skin color became gray.</ns0:p></ns0:div> <ns0:div><ns0:head>Enhanced Ultrasound Imaging Assessment of LSPMbs</ns0:head><ns0:p>Fifteen of the mice were used for the targeted CEU experiments in five groups, with each group of three mice. The CEU effect was assessed using GE Logiq E9 ultrasound system (as has addressed in the previous section). During the ultrasound scanning, the frequency of the transducer was set to 12 MHz, the depth and focus were adjusted to optimize imaging, and the model was shifted to contrast imaging, using the default parameter (MI 0.1). The shape of the xenograft tumors was ovoid, and the longitudinal diameter of the tumors of 24 mice was 10.4&#177;0.53 mm in the fifth week after cells seeded. Control experiments were conducted in four Manuscript to be reviewed groups using LSGMbs, LSPMbs (GPC3 antibody or synthesized peptide blocked previously), SonoVue (a commercial ultrasound contrast-enhanced agent), and LS, respectively. LSPMbs suspension was prepared using 14mg LSPMbs and 4 mL 0.9% sodium chloride solution, agitated slightly before injection. The animals with Huh-7 xenograft tumor in five groups were intravenously injected suspension of LSPMbs, LSGMbs, LSPMbs (injected 0.1mL synthesized peptide or 0.1mL dilated GPC3 antibody for blocking in 3 minutes ahead), SonoVue (15mg SonoVue in 4mL 0.9% sodium chloride solution), and LS (14mg LS and 4 mL 0.9% sodium chloride solution, agitated slightly before injection), respectively, each with the volume of 0.2mL. The contrast imaging time was counted since the bolus intravenous injection of 0.2mL LSPMbs suspension via the mouse tail vein. The images were saved in the ultrasound system and exported for study late.</ns0:p><ns0:p>Enhancement ultrasound imaging effects of LSPMbs, LSGMbs after synthesized peptide or GPC3 antibody blocking, SonoVue, LS were determined using Photoshop software, and the methods had been addressed in the previous section.</ns0:p></ns0:div> <ns0:div><ns0:head>Fluorescent Imaging Assessment of LSPMbs</ns0:head><ns0:p>Ten mice with xenograft tumors of Huh-7 cells allotted to two groups of five each were used to Manuscript to be reviewed Jinan, China). 0.2mL Cy5.5 labelled LSPMbs suspension was intravenously injected into six mice with Huh-7 xenograft tumors via the tail veins, images were acquired at one minute, six hours, and 24 hours. Cy5.5 labelled LSPMbs suspension was prepared using 14mg Cy5.5 labelled LSPMbs and 4 mL 0.9% sodium chloride solution, agitated slightly before injection.</ns0:p><ns0:p>The control experiments were conducted in five mice with Huh-7 xenograft tumors, with the same methods after injection of 0.1mL synthesized peptide (10&#181;g/mL) for blocking in three minutes. Four mice were sacrificed by euthanasia using 3m L isoflurane in a closed small box.</ns0:p><ns0:p>The tumor, liver, heart, lung, kidney, and spleen of the mice were isolated for fluorescent imaging assessment at 24hours.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analysis</ns0:head><ns0:p>Quantitative data are presented as mean &#177; SD (standard deviation) , and qualitative data are presented as percentile. Statistical significance of differences between groups of quantitative variables were analyzed using paired-sample t tests or univariate analysis of variance, and qualitative variables were analyzed using Chi-square test. All statistical analyses were performed using SPSS software (Version 20; IBM, Armonk, NY, USA). P &lt; 0.05 was considered significant.</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Characterization of LSPMbs</ns0:head><ns0:p>The LSPMbs appeared round shaped on a sectional view, with different size, without aggregation, identified by transmission electron microscopy (Figure <ns0:ref type='figure' target='#fig_15'>1</ns0:ref>). Manuscript to be reviewed Size and Zeta potential of LSPMbs were 380.9&#177;176.5 nm and -51.4&#177;10.4mV, respectively. The determination results showed 'Good' (Figures <ns0:ref type='figure' target='#fig_19'>2 and 3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Biocompatibility of LSPMbs</ns0:head></ns0:div> <ns0:div><ns0:head>MTT Assay for Cytotoxicity</ns0:head><ns0:p>The cell index had no significant differences among different LSPMbs concentrations at different time (all P&gt;0. 05), indicating that LSPMbs did not cause significant toxic effect on RAW 264.7 cells. As shown on Figure <ns0:ref type='figure' target='#fig_20'>4</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Evaluation of Enhanced Ultrasound Imaging of LSPMbs in Vitro</ns0:head><ns0:p>LSPMbs and SonoVue suspension with different concentrations presented different enhancement effects, they presented almost the same enhancement effect at the same concentration (Figures <ns0:ref type='figure' target='#fig_10'>5</ns0:ref> A and D were obtained from 1.6mg/mL, figures 5B and E was obtained from 0.8mg/mL, and figures 5C and F was obtained from 0.4mg/mL). At higher concentration (1.6mg/mL), they all presented homogeneous hyperechogenicity with marked attenuation (Figures <ns0:ref type='figure' target='#fig_22'>5A and D</ns0:ref>), and the echogenicities became weaker when the concentrations decreased (Figures <ns0:ref type='figure' target='#fig_22'>5B and E</ns0:ref>, and figures 5C and F), and they all much stronger than control of degas deionized water, which presented homogeneous anechogenicity (Figure <ns0:ref type='figure' target='#fig_22'>5G</ns0:ref>). Control of air presented strong echogenicity at the interface of the tube, and the distal field presented marked attenuation (Figure <ns0:ref type='figure' target='#fig_22'>5H</ns0:ref>), which was substantial different from those obtained from LSPMbs and SonoVue suspension. </ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of Affinity of LSPMbs to the Liver Cancer Cells</ns0:head><ns0:p>The fluorescence on the cellular membrane of Huh-7 cells was intensive (Figure <ns0:ref type='figure' target='#fig_23'>6</ns0:ref>), indicating that there was high GPC3 expression. Cells and DiI labled LSPMbs and controls obtained from light microscope (Figure <ns0:ref type='figure' target='#fig_24'>7</ns0:ref> A, E and I). Cell nucleus of Huh-7 cells presented blue after DAPI staining and being incubated with DiI labled LSPMbs (Figure <ns0:ref type='figure' target='#fig_24'>7</ns0:ref> </ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of LSPMbs in Vivo Enhanced Ultrasound Imaging Assessment of LSPMbs</ns0:head><ns0:p>All xenograft tumors in the mice of the five groups presented a complex of isoechogenicity, hypoechogenicity and anechogenicity ( <ns0:ref type='figure'>Figures A, E, I, M and Q</ns0:ref> These indicate that LSPMbs has good capability in CEUS imaging, but the experiments of it did not show targeted imaging in vivo.</ns0:p></ns0:div> <ns0:div><ns0:head>Fluorescent Imaging Assessment of LSPMbs</ns0:head><ns0:p>The fluorescent signal could be visualized all over the body soon after the administration of Manuscript to be reviewed could not be visualized in the mice 24 hours after injection (Figure <ns0:ref type='figure' target='#fig_30'>9C</ns0:ref>). Experiment carried out in the mice blocked previously by injection of GPC3 antibody or synthesized peptide presented the same fluorescent imaging characteristics as those in the mice with Huh-7 xenograft tumors (Figures <ns0:ref type='figure' target='#fig_30'>9D, E and F</ns0:ref>). 24 hours after intravenous injection of Cy5.5 labelled LSPMbs, four mice with Huh-7 cell xenograft tumors of two in each groups were sacrificed, the tumors and visceral organs were assessed, there were fluorescence in the lungs and liver, and no fluorescence in the tumor and the heart, spleen and kidneys (Figure <ns0:ref type='figure' target='#fig_31'>10</ns0:ref>). These indicate that the Cy5.5 labelled LSPMbs did not selective accumulated in the xenograft tumor, and the LSPMbs did not present detectable targeting ability to the tumor. The reason that the fluorescent signal intensity in the liver and spleen was higher than other areas is believed that the liver and spleen have abundant capillaries and macrophage cells, the macrophage cells can engulf the liposomes, so liposomes in these regions are richer than other regions. The more Cy5.5 labelled LSPMbs aggregated, the stronger the fluorescent signal intensity.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>LSPMbs presented similar enhancement effect at the same concentration as that the SonoVue performed in vitro and vivo, indicating that the LSPMbs has good capability of enhancement imaging. Optical imaging in vivo using fluorescence and bioluminescence has high sensitivity and resolution <ns0:ref type='bibr' target='#b5'>[6]</ns0:ref>. The near infrared dye Cy5.5 allows a fluorescent imaging of deep tissue in rodent animals with low background, providing a possibility of evaluation of the molecular imaging agents. In this study, LSPMbs were labelled with Cy5.5 for targeting GPC3 imaging Manuscript to be reviewed evaluation, the results showed that they had not presented aggregated fluorescence imaging, indicating that LSPMbs were not targeting retained in the tumor. LSPMbs did not present targeted imaging both at ultrasound imaging and optical imaging.</ns0:p><ns0:p>Unlike iodinated and gadolinium contrast agents for x-ray, CT and MRI that can enter extravascular tissue, common UCAs are confined to the blood pool when administered intravenously, which are consist of microbubbles in suspension which strongly interplay with the ultrasound beam and are readily detectable by ultrasound imaging systems <ns0:ref type='bibr' target='#b7'>[7]</ns0:ref>. Molecularly targeted UCAs are created by conjugating the microbubble shell with a peptide, antibody, or other ligand designed to target an endothelial biomarker associated with tumor angiogenesis or inflammation. These microbubbles then accumulate in the microvasculature at target sites where they can be imaged <ns0:ref type='bibr' target='#b7'>[7]</ns0:ref>.</ns0:p><ns0:p>Our findings were significantly different from previous reports that vascular endothelial growth factor receptor 2 (VEGFR2) based targeted UCAs <ns0:ref type='bibr' target='#b8'>[8]</ns0:ref>. These targeted UCAs do not need to extravasate the blood vessels. The VEGFR2 based targeted UCAs can contact and combine with the VEGFR2 on the blood vessels when they enter and flow through the tumor's vessels, forming focal contrast agent accumulation, and can display focal enhanced imaging at ultrasound scanning. However, our preparation of LSPMbs targeting cellular membranous receptor of the HCC confronts substantial challenge for targeted imaging. Sizes of LSPMbs are 380.9&#177;176.5 nm, which can extravasate the fissure of blood vessel wall of tumor if there are no other impeding factors. But the experimental results did not gain the desired goal. The reasons may be PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed the following factors. The previous study showed that the gap between the epithelia of cancer blood vessel is large enough (380-780 nm) to allow nanoscale materials passing, but the blood vessels contact the cells of tumor and interstitial closely, elevated interstitial fluid pressure in the tumor could restrict convective flow and antibody extravasation, except in large necrosis and hypoxia areas <ns0:ref type='bibr' target='#b9'>[9,</ns0:ref><ns0:ref type='bibr' target='#b10'>10]</ns0:ref>. Similarly, the targeted LSPMbs needs overcome interface pressure gradient and get enough space to access and bind to the cancer cells of tumor. How the LSPMbs penetrate the blood vessels of tumor, distribute in the tumor and uptake by the cells are difficult to understand or predict <ns0:ref type='bibr' target='#b11'>[11]</ns0:ref>. A study by Opie showed that the osmotic parameters of tumors (hepatoma, etc) are much lower than that of normal tissues <ns0:ref type='bibr' target='#b12'>[12]</ns0:ref>. In this circumstance, the suspension of LSPMbs is harder to be absorbed into tissue by osmotic force of tumor. At tumor sites, the disorganized tumor vascular network, extensively distributed stromal cells (e.g., tumorassociated macrophage, cancer-associated fibroblasts, etc.) and the dense physical barriers of extracellular matrix comprise of the abominable obstacles hampering nanoparticles transport in a tumor. <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref> The electron microscopy results on a study have confirmed that the opening in tumor extracellular matrix barriers surrounding the cancer cells is generally less than 40 nm. <ns0:ref type='bibr' target='#b14'>[14]</ns0:ref> On this condition, LSPMbs of 380.9&#177;176.5 nm are impossible to pass the extracellular matrix opening to access to the cancer cells. A recent study revealed that only 0.7% of systemically administered nanoparticles can reach the tumor sites and less than 14 out of 1 million (0.0014% injected dose) of them are accessed by cancer cells, and that only 2 out of 100 cancer cells interacted with the nanoparticles. <ns0:ref type='bibr' target='#b16'>[15]</ns0:ref> Therefore, if the number and volume of the targeted LSPMbs entering the tumor extravascular part are not enough, it will be impossible to yield Manuscript to be reviewed visible enhanced imaging effect.</ns0:p><ns0:p>Many researches on extravascular targeted contrast-enhanced ultrasound have been reported in literature, but only a few of them validate in vivo of animals. Mai et al reported that a chitosan-vitamin C lipid system had been fabricated and had achieved tumor-selective enhanced ultrasound imaging in a mouse tumor model, but they confirmed only that the fluorescence accumulated highly at the tumor site, other than the nanobubbles <ns0:ref type='bibr' target='#b17'>[16]</ns0:ref>. Another extravascular targeted nanobubbles fabricated by Gao et al remains to be further verification because of the preliminary results and substantial limitations <ns0:ref type='bibr' target='#b18'>[17]</ns0:ref>. Theoretically, if microbubbles enter the tumor interstitials, some liquid solution also enters. There must be enough extravascular space in the tumor to contain and distribute them, only in this condition can the microbubbles in the oscillations of ultrasonic compression and expansion wave generate stronger backscattered acoustic signal and second harmonics for enhanced ultrasound imaging. If many microbubbles are compacted together, their size will be big, and which will generate little backscattered acoustic signal and second harmonics <ns0:ref type='bibr' target='#b19'>[18]</ns0:ref>.</ns0:p><ns0:p>Our experimental results, together with earlier published reports by others <ns0:ref type='bibr' target='#b13'>[13]</ns0:ref><ns0:ref type='bibr' target='#b14'>[14]</ns0:ref><ns0:ref type='bibr' target='#b16'>[15]</ns0:ref>, strongly suggested that to develop a targeted material for extravascular contrast-enhanced ultrasound imaging, cutting-edge precisive experiments should be conducted firstly to ascertain that the material can penetrate the blood vessel and wade through the cellular matrix and stroma to access and bind to the target cells, and the accommodation space for the materials is adequate. In future, the development of materials for extravascular contrast-enhanced ultrasound imaging Manuscript to be reviewed may be emphasis that using specific materials such as cell-penetrating peptides, a disulfidebridged cyclic RGD peptide, named iRGD (internalizing RGD, c(CRGDK/RGPD/EC)), which is a tumor-homing peptide that can bind to avb3 integrin with high affinity and specificity to construct the targeted material. A material integrated iRGD peptide may increase penetration of the blood vessels and matrix, and facilitate accumulation and increase the probability of enhanced imaging <ns0:ref type='bibr' target='#b20'>[19,</ns0:ref><ns0:ref type='bibr' target='#b22'>20]</ns0:ref>. Augmentation of enhanced permeability and retention effect of targeted material through using NO-releasing agent such as nitroglycerin or angiotensinconverting enzyme inhibitors, and albumin-protein interactions using S -nitrosated human serum albumin dimer, etc, may increase targeted material accumulation and the probability of enhanced imaging <ns0:ref type='bibr' target='#b23'>[21,</ns0:ref><ns0:ref type='bibr' target='#b25'>22]</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Collectively, a new material of LSPMbs has been prepared, which has good effect of enhanced ultrasound imaging, but it did not exhibit targeted imaging effect in vivo of animal experiments. The causes may be that the volume of LSPMbs pass the tumor blood vessels and enter the tumor parenchyma was very limited, and the LSPMbs cannot pass the fissures of extracellular stroma and matrix surrounding the cancer cells to access and bind to the cancer cells. Therefore, a potential target of GPC3 on hepatocellular carcinoma for extravascular targeted imaging may not be realized in contrast-enhanced ultrasound. The future research should focus on that whether a candidate targeted material for extravascular contrast-enhanced ultrasound imaging can penetrate the blood vessel and wade through the cellular matrix and Manuscript to be reviewed stroma to access and bind to the target cells, and whether the accommodation space for the materials is adequate.</ns0:p><ns0:note type='other'>Figure 1</ns0:note><ns0:p>Image of the LSPMbs obtained by transmission electron microscope. Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed Manuscript to be reviewed</ns0:p><ns0:note type='other'>Figure 4 Figure 4</ns0:note><ns0:note type='other'>Figure 5</ns0:note><ns0:note type='other'>Figure 6 Figure 6</ns0:note><ns0:note type='other'>Figure 7</ns0:note><ns0:note type='other'>Figure 9</ns0:note><ns0:note type='other'>Figure 10</ns0:note></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>of EDC/NHS [1-ethyl-3(3-dimethylaminopropyl) carbodiimide (EDC): N-hydroxy succinimide (NHS) = 1:4, Sigma-Aldrich Chemical Co., Inc, USA] were added into the suspension. The mixture suspension was oscillated and incubated for 2 h at room temperature (25&#8451;). The remaining EDC/NHS was removed by three-time centrifugation at 12000 rpm using MES (pH 6.0), 5 min each time. The precipitate was dispersed into MES buffer (0.1 mol/L, pH 8.0), and an adequate amount of synthesized peptide was added and incubated with stirring for 2 h at room PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020) Manuscript to be reviewed temperature. The peptide was compounded by GL Biochem (Shanghai) Ltd. (Shanghai, China) in accordance to a 12-mer peptide with the sequence of DHLASLWWGTEL reported from previous study that it can target GPC3 of HepG2 [4] [ DOI: 10.1021/acs.bioconjchem.6b00030].</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head /><ns0:label /><ns0:figDesc>Enhancement ultrasound imaging effects of LSPMbs, LSGMbs, SonoVue, LS, degas deionized water, and air were determined using Photoshop software (Adobe Photoshop CS3, Adobe Systems Inc, CA, USA). To analyze ultrasound images, observers opened the image in PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head /><ns0:label /><ns0:figDesc>mg LSPMbs suspension to form DiI labelled LSPMbs suspension. Huh-7 cells [Cell bank of the Chinese academy of sciences (Shanghai, China)] were seeded in a six-well plate with one glass slip placed in each well at a concentration of 3&#215;10 4 cells/well. Next day, the Huh-7 cells were fixed by 90% cold ethanol for 20 min and blocked by 10% bovine serum albumin (BSA) at 37 &#176;C for 1 h and subsequently incubated with 4 drops of DiI labelled LSPMbs suspension for 3 h in dark place. Next, the wells and slips were washed with PBS three times, and the cells on the slips were mounted with 4&#8242;, 6-diamidino-2-phenylindole (DAPI, Wuhan Boster Biological Technology, Ltd., China) for nuclei visualization and detected using a laser confocal microscope (Fluoview FV 10001000, Olympus, Japan). Images of bright, DAPI staining, DiI staining, and merged were obtained. LSGMbs were assessed in the same protocols for control. To confirm the specificity of binding of DiI labelled LSPMbs to GPC3 in the Huh-7 cells, 0.1mL synthesized peptide PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>conduct</ns0:head><ns0:label /><ns0:figDesc>Cy 5.5 fluorescence experiment to test the specificity and affinity of the peptide in LSPMbs to GPC3 of the liver cancer. An IVIS Lumina image system (Xenogen) (IVIS &#174; Lumina XR) (Caliper life sciences) was used for the evaluation. During the fluorescence imaging, mice were under gas anesthesia with oxygen and isoflurane (Jinan Shengqi pharmaceutical Co, Ltd., PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_5'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_6'><ns0:head /><ns0:label /><ns0:figDesc>figure C vs. F, P=0.428; and figure A vs. G, A vs. H, B vs. H, G vs. H, A vs. B, A vs. C, B vs. C,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_7'><ns0:head /><ns0:label /><ns0:figDesc>B, F and J). LSPMbs with DiI staining combined with the membrane of Huh-7 cells presented red color fluorescence on image (Figure 7 C and D); LSGMbs presented a very similar appearance (Figure 7 G and H); and LSPMbs with DiI staining after synthesized peptide blocking previously did not show red color fluorescence on the membrane of Huh-7 cells, indicating that LSPMbs had not combined with the membrane of Huh-7 cells (Figure 7 K and L).</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_8'><ns0:head /><ns0:label /><ns0:figDesc>). Of the study group, after administration of LSPMbs suspension, Huh-7 xenograft tumor presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis) at two seconds (Figure 8 B); the tumor enhancement lasted over 20 seconds with little change; at 60 seconds, the tumor still presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis) (Figure 8 C); at 10 minutes, the tumor presented iso-enhancement with central hypo-enhancement (necrosis) (Figure PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020) Manuscript to be reviewed 8 D). Of the four control groups, the mice injected with LSGMbs, LSPMbs (blocked with GPC3 antibody or synthesized peptide previously), and SonoVue, respectively, presented similar enhancement patterns and sustain time as those in Huh-7 xenograft tumors (Figures 8 F,</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_9'><ns0:head /><ns0:label /><ns0:figDesc>figures N, O, and P of SonoVue) all had no significant difference (all P&gt;0.05), and there were</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_10'><ns0:head>Cy5. 5 labelled</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>LSPMbs suspension. Images acquired at one minute (Figure 9A), six hours (Figure 9B) after the initial fluorescent imaging, the fluorescent signal intensity in the tumor area has no significant difference from other areas of the body expect the liver and spleen. The fluorescent signal intensity in the liver and spleen area of the mice was marked stronger than other areas, and the fluorescent signal intensity was similar in three times. The fluorescent signal PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_11'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_12'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_13'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_14'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:08:51803:2:0:NEW 27 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_15'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1 Image of the LSPMbs obtained by transmission electron microscope. The LSPMbs present round shaped on a sectional view, with different size, without aggregation.</ns0:figDesc><ns0:graphic coords='26,42.52,229.87,525.00,349.50' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_16'><ns0:head>Figure 2 Size</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_17'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2 Size of LSPMbs are 380.9&#177;176.5 nm, and the determination results show 'Good'.</ns0:figDesc><ns0:graphic coords='27,42.52,204.37,525.00,357.00' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_18'><ns0:head>Figure 3 Zea</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_19'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3 Zea potential of LSPMbs was -51.4&#177;10.4mV, and the determination results show 'Good'.</ns0:figDesc><ns0:graphic coords='28,42.52,229.87,525.00,339.00' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_20'><ns0:head>Figure 4</ns0:head><ns0:label>4</ns0:label><ns0:figDesc>Figure 4 MTT assay for cytotoxicity showed the cell index had no significant differences among different LSPMbs concentrations at different time (all P&gt;0. 05).</ns0:figDesc><ns0:graphic coords='29,42.52,229.87,525.00,217.50' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_21'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_22'><ns0:head>Figure 5</ns0:head><ns0:label>5</ns0:label><ns0:figDesc>Figure 5 Ultrasound images of the LSPMbs, SonoVue, degas deionized water, and air in the tubes. Figures A and D were obtained from LSPMbs and SonoVue of 1.6mg/mL, B and E were obtained from LSPMbs and SonoVue of 0.8mg/mL, and C and F were obtained from LSPMbs and SonoVue of 0.4mg/mL. Figure G was obtained from degas deionized water, and H was obtained from air. The echogenic intensity decreased with decreasing concentrations of LSPMbs, and degas deionized water; the echogenic intensity was strong at the interface between the air in the tube and the outside water, and the echogenicity in the tube was attenuated.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_23'><ns0:head>Figure 6</ns0:head><ns0:label>6</ns0:label><ns0:figDesc>Figure 6 Image of GPC3 expression of Huh-7 cells obtained by confocal laser scanning microscope. The fluorescence on the cellular membrane appears intensive.</ns0:figDesc><ns0:graphic coords='32,42.52,229.87,525.00,492.75' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_24'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_25'><ns0:head>Figure 7</ns0:head><ns0:label>7</ns0:label><ns0:figDesc>Figure 7 Images of Huh-7 cells incubated with DiI lablled LSPMbs and controls obtained by light microscope and confocal laser scanning microscope. (Images of A, E and I) Cells and DiI lablled LSPMbs and controls obtained from light microscope. (Images of B, F and J) Cell nucleus of Huh-7 cells presented blue after DAPI staining and being incubated with DiI lablled LSPMbs. (Images of C and D) LSPMbs with DiI staining combined with the membrane of Huh-7 cells presented red color fluorescence on image. (Images of G and H) LSGMbs presented a very similar appearance. (Images of K and L) LSPMbs with DiI staining after blocked by synthesized peptide did not show red color fluorescence on the membrane of Huh-7 cells, indicating that LSPMbs had not combined with the membrane of Huh-7 cells.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_26'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_27'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_28'><ns0:head>Figure 8</ns0:head><ns0:label>8</ns0:label><ns0:figDesc>Figure 8 Row 1 is a study group, and rows 2-5 are control groups. (Images of A, E, I, M and Q) Images of all xenograft tumors presented a complex of isoechogenicity, hypoechogenicity and anechogenicity obtained by convention ultrasound. (Image B of LSGMbs) After administration of LSPMbs suspension, tumors presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis) at two second. (Image C of LSGMbs) At 60 seconds, the tumor still presented hyper-enhancement in periphery and hypo-enhancement in center (necrosis). (Image D of LSGMbs) At 10 minutes, the tumor presented isoenhancement with central hypo-enhancement (necrosis). (Images F, G and H of LSGMbs) and (Images J, K and L of LSPMbs, after GPC3 blocking) and they presented similar enhancement patterns and sustain time as those in Huh-7 xenograft tumors, and there were no appreciable difference. (Images N, O and P of SonoVue) The xenograft tumors presented similar enhancement patterns and sustaining time as those using LSPMbs after administration of SonoVue suspension. (Images R, S and T of LS) The xenograft tumors did not present enhancement at 2, 20, and 60 seconds, and 10 minutes after administration of LS suspension.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_29'><ns0:head>Figure 9</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_30'><ns0:head>Figure 9</ns0:head><ns0:label>9</ns0:label><ns0:figDesc>Figure 9 Image A acquired at one minute, and image B acquired six hours after the initial fluorescent imaging, the fluorescent signal intensity in the tumor area has no significant difference from other areas of the body expect the liver and spleen. Image C acquired 24 hours after injection, the fluorescent signal could not be visualized in the mice. Images D, E and F acquired from experiments of mice with Huh-7 xenograft tumors blocked using GPC3 antibody presented the same fluorescent imaging characteristics as those in the mice with Huh-7 xenograft tumors.</ns0:figDesc><ns0:graphic coords='37,42.52,357.37,525.00,351.75' type='bitmap' /></ns0:figure> <ns0:figure xml:id='fig_31'><ns0:head>Fluorescence in the organsFigure 10</ns0:head><ns0:label>10</ns0:label><ns0:figDesc>Figure 10 There were fluorescence in the lungs and liver, and no fluorescence in the tumor, heart, spleen, and kidneys 24 hours after intravenous injection of Cy5.5 labelled LSPMbs, after animal sacrificed.</ns0:figDesc><ns0:graphic coords='38,42.52,255.37,525.00,268.50' type='bitmap' /></ns0:figure> <ns0:figure><ns0:head /><ns0:label /><ns0:figDesc /><ns0:graphic coords='34,42.52,70.87,525.00,344.25' type='bitmap' /></ns0:figure> </ns0:body> "

"Response Letter Dear Reviewers and Editors, We read your comments with due respect and really appreciated your work. According to your comments and suggestion, the manuscript has been revised, and the response was given point to point. However, due to our limited academic level, there will be still drawbacks and mistakes in the revised manuscript. We are looking forward to hearing good news. Sincerely, Size Wu On behalf of all authors Responses to Editor’s and Reviewer's Comments Editor comments (Yoshinori Marunaka) I have received the review result from a reviewer who has recommended that your manuscript would be acceptable. However, the reviewer is also concerned about the title of your manuscript including the term of 'negative outcome', which would reduce the value of the manuscript. Therefore, I would suggest your revising the title to convey a more neutral impression. I would suggest 'Trials in developing a nanoscale material for extravascular contrast-enhanced ultrasound targeting hepatocellular carcinoma' Our response: We agree with you. Thank you for your comments and suggestion. The title was changed as you suggested. Reviewer 1's Comments Basic reporting Comment 1：This article is to develop an image enhanced material for the ultrasound examination targeting specific tumor such as HCC. Our response: We agree with you. Comment 2：This Article is re-reviewed at the standpoint of an oncology clinician and the experimental pathology of malignant tumors. Our response: We agree with you. Experimental design Comment 1：This experimental study was executed in an adequate animal evaluation, data processing including statistical analyses scientifically and ethically. Our response: We agree with you. You had understood our article sufficiently. Comment 2：The response for each reviewers’ comments seems to be rewritten adequately and precisely. Our response: Thank you. The previous response had been written seriously. Validity of the findings Comment 1：The author carefully revised Figures according to reviewer’s comments. Our response: We agree with you. Comments for the Author Comment 1：Reviewers have already pointed out the poor experimental results. And article authors are also aware of this issue. Our response: Thank you for your understanding. We agree with you. Comment 2：Although final results is negative, this article details the process of development materials for extravascular ultrasound targeting. Our response: We agree with you. Thank you for your understanding. Comment 3：The title 'negative outcome' might lose the value of the paper, so it is better to rewrite the title in a different expression. Our response: We agree with you. The title was rephrased as “Trials in developing a nanoscale material for extravascular contrast-enhanced ultrasound targeting hepatocellular carcinoma” Comment 4：A result of the peer review, acceptable except the title expression. Our response: Thank you for your understanding. The title was changed as “Trials in developing a nanoscale material for extravascular contrast-enhanced ultrasound targeting hepatocellular carcinoma” "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:p>We report here five new localities across the distribution of the lizard Sphaerodactylus samanensis, extending its current geographic range to the west, in the Cordillera Central of Hispaniola. We also report phenotypic variation in the color pattern and scutellation on throat and pelvic regions of males from both eastern and western populations, which is described below. Furthermore, based on these new data, we confirm that the species is not fitting in its current IUCN category, and in consequence propose updating its conservation status.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Lizards of the genus Sphaerodactylus (107 recognized species, <ns0:ref type='bibr' target='#b42'>Uetz et al. 2020)</ns0:ref>, have diversified remarkably on Caribbean islands, and occur in Central and Northern South America and in the Pacific Island of Cocos <ns0:ref type='bibr' target='#b9'>(Hass 1991;</ns0:ref><ns0:ref type='bibr' target='#b17'>Henderson and Powell 2009;</ns0:ref><ns0:ref type='bibr' target='#b15'>Hedges et al. 2019;</ns0:ref><ns0:ref type='bibr' target='#b13'>Hedges 2020)</ns0:ref>. This is a clade of small geckos (geckolet) containing also one of the smallest amniote vertebrates in the world with a maximum snout-vent length of 18 mm <ns0:ref type='bibr' target='#b16'>(Hedges and Thomas 2001)</ns0:ref>. Likewise, the largest species of this genus reaches up to a maximum of 39 mm <ns0:ref type='bibr'>(Barbour 1914;</ns0:ref><ns0:ref type='bibr' target='#b36'>Schwartz and Garrido 1985;</ns0:ref><ns0:ref type='bibr' target='#b5'>Fong and Diaz 2004;</ns0:ref><ns0:ref type='bibr' target='#b6'>Griffin et al. 2018)</ns0:ref>.</ns0:p><ns0:p>These geckos are one of the most dominant herpetofauna in the Antilles <ns0:ref type='bibr' target='#b33'>(Scantlebury et al. 2011)</ns0:ref>, reaching densities greater than 60,000 ind/ha <ns0:ref type='bibr' target='#b32'>(Rodda et al. 2001)</ns0:ref>. Nonetheless, nearly 20% of the species are known only for the type locality <ns0:ref type='bibr' target='#b28'>(Meiri et al. 2017)</ns0:ref> and several others for a small number of localities <ns0:ref type='bibr' target='#b10'>(Hedges 1996;</ns0:ref><ns0:ref type='bibr' target='#b30'>Powell and Inchaustegui 2009;</ns0:ref><ns0:ref type='bibr' target='#b35'>Schwartz 1970;</ns0:ref><ns0:ref type='bibr' target='#b37'>Schwartz and Henderson 1991)</ns0:ref>. Among them, Sphaerodactylus samanensis is a species previously reported at a few places near to the type locality, along the southern side of the Samana Bay, Dominican Republic with an elevation range from 0 to 181 m.a.s.l. <ns0:ref type='bibr' target='#b37'>(Schwartz and Henderson 1991;</ns0:ref><ns0:ref type='bibr' target='#b41'>Thomas and Hedges 1993;</ns0:ref><ns0:ref type='bibr' target='#b23'>Landestoy et al. 2016)</ns0:ref>. Because its restricted distribution range and small extent of occurrence (100 km 2 ), S. samanensis is currently classified as a Critically Endangered species by both the IUCN Red List (2020), and the Dominican Republic's Red List of threatened species <ns0:ref type='bibr'>(MIMARENA, 2019)</ns0:ref>. According to the records, this species inhabits the northeastern edge of the island (Figure <ns0:ref type='figure'>1</ns0:ref>) alongside Cordillera Oriental, a low mountain chain with Miocene Karst terrain <ns0:ref type='bibr' target='#b0'>(Bowin 1966</ns0:ref><ns0:ref type='bibr' target='#b1'>(Bowin , 1975))</ns0:ref>.</ns0:p><ns0:p>The recent discovery of an individual of Sphaerodactylus samanensis in the surroundings of Pueblo Viejo Mine (PVM) by one of the authors (JU) encouraged us to perform new field Morphological revision. We euthanized the specimens in the field with Lidocaine 10%, then preserved and stored in alcohol 96%. We used a digital calliper to measure snout-vent length (SVL) of individuals to the nearest tenth of a millimeter. Our scale counts follow <ns0:ref type='bibr' target='#b39'>Thomas and Schwartz (1966)</ns0:ref> and <ns0:ref type='bibr' target='#b40'>Thomas et al. (1992)</ns0:ref> and consists in: 1) escutcheon length, we considered the maximum number of scales (anterior to posterior); 2) escutcheon width, we considered the maximum number of scales transversally across the patch (including extensions onto thighs); and 3) escutcheon total scales, we considered all scales on the pelvic scutcheon. In order to support our observations, we added two more scale counts: 1) number of gular scales in contact with the first infralabial, here we considered all adjacent scales (including postmentals) to the first infralabial scale; and 2) number of scales per dorsal band, we considered the maximum number of pigmented scale rows covered by a dorsal band in a longitudinal count. Specimens were sexed by examining the sexually dimorphic color pattern and the gonads to confirm the presence of hemipenes. We used photographs taken in the field by ML to describe the coloration in life of the specimens. Also, we followed Kohler (2012) to name the colors in our description. In addition, we follow taxonomy stablished by Poe (2012) which unlike <ns0:ref type='bibr' target='#b21'>K&#246;hler et al. 2019</ns0:ref>, regards Anolis as a valid genus for Dactyloid lizards from La Hispaniola. Data Analysis. We estimated the occurrence of this species based on our field measurements of the extension of Karst (where we observed Sphaerodactylus samanensis), additionally supported by the estimation of the area of Karst in contact with them, through the data previously reported by Servicio Geologico Nacional (2010). Geographic data and map designing were drawn in ArcGIS version 10.3. Additionally, we follow IUCN (2001) defining: 1) Extent of occurrence (EOO) and 2) Area of occupancy (AOO).</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head><ns0:p>We observed this species mostly under Karst rocks and out of the surroundings of the type locality for the first time (specimens collected per locality are detailed in Table <ns0:ref type='table'>S1</ns0:ref>).</ns0:p><ns0:p>Subsequently, we are confirming its occurrence in the Cordillera Central and adding five localities to its currently known occurrence (Figure <ns0:ref type='figure'>1</ns0:ref>). This extends its geographic range by 82.2 km to the northwest. All individuals were observed by day, under rocks in habitat mixed between karst-rock clusters and tropical forest, with bushes and trees approaching 30 m tall, ground covered in leaf-litter and rocks covered with moss, lichens, ferns and other epiphytes. <ns0:ref type='table'>2020:07:51336:1:1:CHECK 30 Sep 2020)</ns0:ref> Manuscript to be reviewed All individuals of S. samanensis agree with the original description <ns0:ref type='bibr' target='#b2'>(Cochran 1932</ns0:ref>) bearing a combination of the following characters: a moderately short snout, a large rostral scale with a median groove, a medium-sized superciliar spine, a large third supralabial exceeding the center of the eye, imbricate-keeled dorsal scales and an orange head in males. Nevertheless, we noted some phenotypic variation between S. samanensis individuals from the surroundings of the type locality (Ca&#241;o Hondo) and nearby eastern places (Cueva Casa Grande and Batey Piedra), and the western populations (Chacuey Abajo, Cueva de Sanabe, and PVM) (See Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>). The eastern individuals have 2.5-5.5 (average=4.1, SD=0.8) gular scales in contact with first infralabial instead of 4.5-7 (average=5.1, SD=0.6) in western individuals (p &lt; 0.001) (See Figure <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>), and a lower total number of pelvic scutcheon scales ranging from 25-32 scales (average=28.4, SD=2.5) instead of 30-39 scales (average=35.7, SD=2.9) in western specimens (p &lt; 0.001). Eastern populations also differ in coloration by bearing dorsal bands and scapular ocelli in females and most males, which are absent in males of western samples (Figure <ns0:ref type='figure' target='#fig_1'>2</ns0:ref>). Eastern females have 3-4 dorsal bands vs 4-5 in western females (p &lt; 0.001), and wider dorsal bands covering 3-7 dorsal scales (average=5, SD=1.2) instead of the thin dorsal bands of western females covering only 3 dorsal scales (average=3, SD=0; p &lt; 0.001). Further details on measurements, coloration and scutellation are provided in Table <ns0:ref type='table'>S2</ns0:ref>.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>Our results update the distribution of Sphaerodactylus samanensis which now range from the region of the type locality (Boca del Infierno) in the Samana Bay <ns0:ref type='bibr' target='#b2'>(Cochran 1932</ns0:ref>) and surrounding areas <ns0:ref type='bibr'>(Thomas and</ns0:ref><ns0:ref type='bibr'>Hedges 1993, Landestoy et al. 2016)</ns0:ref> to the Central Cordillera (Figure <ns0:ref type='figure'>1</ns0:ref>), an east-west airline distance of 82.2 km. Therefore, the distribution of this gecko is now only exceeded by those of S. copei, S. darlingtoni, S. difficilis, and S. elegans <ns0:ref type='bibr' target='#b37'>(Schwartz and Henderson 1991;</ns0:ref><ns0:ref type='bibr' target='#b13'>Hedges 2020)</ns0:ref>, species previously recognized as widely spread on Hispaniola <ns0:ref type='bibr' target='#b9'>(Hass 1991;</ns0:ref><ns0:ref type='bibr' target='#b37'>Schwartz and Henderson 1991)</ns0:ref>. We also report the maximum altitude so far recorded for this species: 257 m. a. s. l. exceeding by 200 meters former records reported by <ns0:ref type='bibr' target='#b2'>Cochran (1932)</ns0:ref> and <ns0:ref type='bibr' target='#b23'>Landestoy et al. (2016)</ns0:ref>. These novel geographic data exceed those formerly known for this species confirming that it is not a short-ranged species but rather a widely distributed lineage that could be distributed even further.</ns0:p><ns0:p>Since large geographic ranges are scarcely recorded in Sphaerodactylus lizards, phenotypic variation has been barely noted and subsequently poorly studied <ns0:ref type='bibr' target='#b34'>(Schwartz 1966;</ns0:ref><ns0:ref type='bibr' target='#b4'>Dood Jr and Ortiz 1984</ns0:ref>). Here we provide for the first time evidence of differences between eastern (including type locality) populations (n=24) and western populations (n=28), mainly in color PeerJ reviewing PDF | (2020:07:51336:1:1:CHECK 30 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed pattern and scutellation (Table <ns0:ref type='table'>S2</ns0:ref>). Measurements did not differ. In spite of scutellation mostly overlapping between eastern and western populations, gular scales are longer in eastern individuals, better noted in the proximal rows of the throat (including postmentals) which have contact with the first infralabial and are clearly smaller in western individuals (Table <ns0:ref type='table'>S2</ns0:ref>, Figure <ns0:ref type='figure' target='#fig_3'>3</ns0:ref>). Likewise, the escutcheon plate in western males tends to contain more scales than those from eastern individuals. The differences between scutcheon width and scutcheon length are not significant (p = 0.7 and p = 0.1 respectively). This is because the difference does not depend on the width or length of the rows, but rather the number of additional escutcheon scales surrounding the proximal edge of the escutcheon (Table <ns0:ref type='table'>S2</ns0:ref>). Concerning coloration, eastern individuals have 3-4 wide dorsal bands (each covering 3-7 dorsal scales) which are present in all females and some males (especially in males from the type locality); contrasting with western individuals which have 4-5 thin dorsal bands (each covering three dorsal scales) only present in females.</ns0:p><ns0:p>The geological history of the island of Hispaniola is influenced mainly by water incursions and plate movements occurring since the late Mesozoic and into the Cenozoic <ns0:ref type='bibr' target='#b27'>(Mann et al. 1991;</ns0:ref><ns0:ref type='bibr' target='#b25'>MacPhee and Iturralde-Vinent 1994;</ns0:ref><ns0:ref type='bibr' target='#b10'>Hedges 1996;</ns0:ref><ns0:ref type='bibr'>Iturralde-Vinent and McPhee 1999;</ns0:ref><ns0:ref type='bibr'>Ricklefs and Bermingham 2008;</ns0:ref><ns0:ref type='bibr' target='#b3'>Daza et al. 2019)</ns0:ref>. This likely originated the vicariance phenomenon in the Proto Antilles as well as the overwater dispersion and later (approximately during Mid-Tertiary sensu Hedges 1996) divergence of lineages in vertebrate fauna on this island <ns0:ref type='bibr' target='#b27'>(Mann et al. 1991</ns0:ref><ns0:ref type='bibr'>, Hedges 1992</ns0:ref><ns0:ref type='bibr' target='#b10'>, Hedges 1996</ns0:ref><ns0:ref type='bibr' target='#b3'>, Daza et al. 2019</ns0:ref>). These events could cause isolation <ns0:ref type='bibr' target='#b10'>(Hedges 1996</ns0:ref><ns0:ref type='bibr'>, Daza et al. 1994</ns0:ref>) and the subsequent geographic restriction of emergent taxa to small areas, explaining why very few Sphaerodactylus species had been able to spread widely on Hispaniola. Those geologic events could have influenced dispersion and also the evolution of phenotypic features of Sphaerodactylus samanensis. Certainly, the distribution of this species seems to follow a geologic pattern overlapping two ancient karst formations (Figure <ns0:ref type='figure'>1</ns0:ref>): Los Haitises karst to the east and El Hatillo karst to the west, both structures raised in the Late Tertiary (Servicio Geol&#243;gico Nacional 2010). This would agree with the phenotypic variation reported here, which follows an east-west geographic pattern. Future research should target molecular analysis and the revision of new specimens to determine patterns in the phenotypic variation in S. samanensis.</ns0:p><ns0:p>Because of its restricted range of distribution and threats to its habitat, both the Dominican Republic and IUCN Red-Lists currently list Sphaerodactylus samanensis as a Critically Endangered species (IUCN 2020). Nevertheless, our findings demonstrate that the occurrence PeerJ reviewing PDF | (2020:07:51336:1:1:CHECK 30 Sep 2020)</ns0:p><ns0:p>Manuscript to be reviewed of Sphaerodactylus samanensis is wider than previously reported, with an estimated EOO of 500 km 2 . We observed that S. samanensis inhabits karst rocks, in contrast to sympatric congeners such as S. darlingtoni and S. difficilis which are more often recorded in leaf litter usually on soil, reducing therefore its AOO within this range. We also suggest that loss of karst formations, in particular loss of tree cover within karst areas, could threat some populations.</ns0:p><ns0:p>Nonetheless, given its widened extent of occurrence, including its presence in protected areas (Los Haitises National Park to the east and Aniana Vargas National Park to the west), as well as the number of locations and mature individuals observed during fieldwork, we propose that the species be reclassified by the IUCN. Certainly, based on new information it would appear unlikely that the species would become extinct barring catastrophic climate events, however, continued destruction of karst habitat could become a future problem for the species, therefore we propose the category Near Threatened for S. samanensis.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>Our report allows us to confirm that Sphaerodactylus samanensis is a widely distributed species that inhabits both eastern forests and Central Cordillera of Hispaniola Island, also confirms its presence in five previously unreported localities. Likewise, we observed a phenotypic variation between eastern and western populations, however this is not consistent enough to consider them different taxon but rather this is evidence of an inter-population variation of S. samanensis. Moreover, our findings extend the area of occupancy of the species and lead us to suggest that its current IUCN category (CR) is not fitting with these novel data, instead we propose Near Threatened as a proper IUCN category. Manuscript to be reviewed A) Eastern male (MNHNSD 23.3716) from Ca&#241;o Hondo, B) western male <ns0:ref type='bibr'>(MNHNSD 23.3734)</ns0:ref> from Chacuey Abajo. Photographs by Miguel A. Landestoy T.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>Additionally, we recorded two other geckolets: Sphaerodactylus darlingtoni and S. difficilis in sympatry with S. samanensis. Other sympatric lizards recorded during surveys were Celestus sepsoides, C. stenurus, Anolis cybotes, and A. distichus PeerJ reviewing PDF | (</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 2 .</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2. Color pattern variation in Sphaerodactylus samanensis between Eastern and Western specimens.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> </ns0:body> "

"Dear PeerJ Editor We want to thank you for the value comments to our manuscript entitled “New distributional records of the Samana Least Gecko (Sphaerodactylus samanensis, Cochran, 1932) with comments on its morphological variation and conservation status”, which we have improved following all of your recommendations. We also are attaching a new version of the Figure 3 as reviewer 1 requested and additional files to support or statements answering the questions from an anonymous reviewer concerning on the taxonomic identity of the furthest populations reported in our manuscript. We also have accepted all typographic corrections and made changes on the manuscript following your suggestions which are detailed in the next pages. The details of the changes following major suggestion are below. We believe now our manuscript is ready to be published in PeerJ. Joaquín A. Ugarte-Núñez Project Manager Knight Piésold Consulting REVIEWER 1. Suggestions: Suggestion 1: The botanical assessment of the area is very vague, and basically would apply to many areas in the Caribbean. I recommend including additional microhabitat information about this species, which would be probably very useful (e.g. species of trees on the area, is the leaf litter moist or dry, is the leaf litter dense or sparce, etc) Reply: We have added all information required including genus of the trees and some vegetation found in the area. This could be noted in the lines Suggestion 2: You mention these geckos were found under rocks, would you classify the as rupiculous, or a combination of leaf litter dwellers using occasionally karst outcrops? Reply: Despite we have observed the specimens mostly under rocks, the lack of sampling points out of Karst rocks prevents us to confirm if Sphaerodactylus samanensis is exclusively a rocky dweller or may inhabit another microhabitat. Otherwise, our analysis is targeting to report the range extension and not its ecological habits. However, we have added a phrase mentioning these rocky shelters in the first line of Results item. Suggestion 3: Is unclear from the photographs, but aren't here more than 2 postmental scales? (Referring to Figure 3A) Reply: No, there are only two postmentals, the irregular speckling on one of them could have resulted in a visual illusion of that. We are uploading additional files to support this, specifically close up photographs of the throat of three specimens (one of them is the one in the Figure 3A) from the same locality pointing the two postmentals present in each one. Comments: Introduction, Line 44: update this number as new species have been described. Reply: We have updated the current number of species of Sphaerodactylus. Introduction, Line 46: and in the Pacific island of Cocos Reply: We have added this locality to the distribution of the genus. Introduction, Line 51: Griffing et al has reports higher than this: Developmental Osteology of the Parafrontal Bones of the Sphaerodactylidae Reply: We have added this information and the correspond citation to the manuscript. Introduction, Line 52: I certainly disagree with this, I think both sphaeros and anoles are very similar in importance. Was Scantlebury mention referring to population densities? Reply: Scantlebury et al. (2011) stated literally that Sphaerodactylus is one of the most dominant groups among Antillean reptiles and amphibians. We have included that cite because we believe is important to highlight the importance of the genus. Likewise, we do not think this statement is underestimating other groups (which are not necessary to mention in a manuscript focused on Sphaerodactylus lizards), is just focusing on the main topic of our manuscript. Methods, line 96: what euthanasia protocol you found? Reply: we have added a line explaining our euthanasia method. Methods, line 102: this number probably changes between males and females, did you find such differences? Reply: We did not find any sexual difference on this character. (See table S2). Methods, line 108: scutcheon scales is usually sufficient to determine the sex in Sphaeros. Reply: That is right, we had the chance to make a small dissection to confirm it, and that is what we done. Results, line 132: maybe talk about the combination, all of these features are general to sphaeros, except the orange head in males. Reply: We have changed the sentence following this suggestion. Results, line 139: maybe a chart showing the differences in numbers across localities would show this better. Reply: we have already showing the variation in the Figure 2, with the most highlighting changes in the bands. To include, including other individuals would be repetitive and would not show the differences. Discussion, line 168: Why is this surprising? (referring to the word initiating the sentence) Reply: we have changed the sentence to avoid any confusion or overrate on the sense of the phrase. REVIEWER 2: Suggestions Suggestion 1. First, a recent molecular phylogeny, that included many Hispaniolan species of Sphaerodactylus (Daza et al. 2019. The sprightly little sphaerodactyl: Systematics and biogeography of the Puerto Rican dwarf geckos Sphaerodactylus (Sphaerodactylidae, Gekkota). Zootaxa 4712:151–201) found S. ladae, S. difficilis, & S. darlingtoni were actually species complexes composed of more than one valid taxon. Thus, multiple widespread species of Hispaniolan Sphaerodactylus are in need of taxonomic revision and that could also include the now widespread S. samanensis. Analysis of molecular data would be helpful in this case to determine whether the eastern and western S. samanensis populations are reproductively isolated. Reply: We have performed a genetic analysis of all populations reported here already. Nevertheless, this analysis is being part of another project which is led by another researcher and it will be part of another scientific article. Despite this fact will prevent us to publish those data, we are uploading a phylogenetic tree (resulted from that research) as additional file to support this statement. In there you will find, that Sphaerodactylus samanensis’s lineages are clustered all together with some divergent branches. However, these branches are not separated enough from type locality samples. Therefore, the monophyly of those populations is well supported and do not belong to a different taxon. Likewise, this fact supports the hypothesis that the Samana’s least gecko is a widely distributed species in La Hispaniola. Suggestion 2. Second, the authors detail significant phenotypic variation between the eastern and western S. samanensis populations. These morphological differences, when considered together, are sufficient to diagnose each population and perhaps even call them separate species. Reply: As we said earlier in the response to the first suggestion, our genetic analysis support the monophyly of those populations and we are quite confident regarding all of them as Sphaerodactylus samanensis. Suggestion 3. Third, the authors describe a biogeographic scenario (lines 174-190) that that could have resulted in allopatric speciation between eastern and western S. samanensis populations. Thus, while S. samanensis clearly has a larger distribution than previously known, it may not be as big as stated in this manuscript and the authors in all likelihood have discovered an undescribed species, closely related to S. samanensis sensu stricto. Thus, reconsidering the conservation status of S. samanensis, while clearly warranted, must be contingent upon a thorough taxonomic evaluation first. Reply: Indeed, this was one of our major issues to solve. Nonetheless, after the genetic analysis was entirely cleared. Please check uploaded additional file (a phylogenetic tree) to corroborate this, as we explained earlier, we are not able to publish yet but we have decided to upload it to support our manuscript considering all populations reported here as S. samanensis. Comments. Introduction, Lines 52-54 - Should also mention that 21 Sphaerodactylus species are known only from the type locality (Meiri et al. 2017. Extinct, obscure or imaginary: The lizard species with the smallest ranges. Diversity & Distributions 24:262-273). Reply: We have added a sentence mentioning this and also add the citation in the text and bibliography. Methods, Lines 109-110 - Perhaps cite (Poe. 2012. 1986 Redux: New genera of anoles (Squamata: Dactyloidae) are unwarranted. Zootaxa 3626:295–299) as justification for using Anolis. Reply: We have added the citation in text and Bibliography. Line 158 - Should be Dodd Jr. & Ortiz (see also line 225). Reply: Actually, should be Dodd Jr and Ortiz, we have corrected it. Line 168 - '(intruders)' - not sure what this refers to. Reply: It refers to the additional scales mentioned in the previous word into the sentence. In order to understand correctly the sentence, we have deleted the word “intruders”. "

Here is a paper. Please give your review comments after reading it.

"<ns0:abstract xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Background:</ns0:head><ns0:p>In idiopathic sudden sensorineural hearing loss (ISSNHL), the relationship between the administration duration of vitamin B12 (vit B12) with adenosine triphosphate (ATP) and their therapeutic effect is not fully understood.</ns0:p><ns0:p>Objective: To investigate the therapeutic effect of long-term (&#8805;16 weeks) administration of vit B12 with ATP on the prognosis of ISSNHL patients and compare it with those of short-term (&lt;8 weeks) and middleterm (&#8805;8 weeks, &lt;16 weeks) administration.</ns0:p></ns0:div> <ns0:div><ns0:head>Methods:</ns0:head><ns0:p>We retrospectively reviewed the medical records of 117 patients with ISSNHL treated between 2015 and 2018.</ns0:p></ns0:div> <ns0:div><ns0:head>Results:</ns0:head><ns0:p>The overall recovery rate was 32.5%. Initial higher hearing threshold and initial higher grade of hearing loss (HL) were associated with a poor prognosis. However, the administration duration of vit B12 and ATP did not influence the overall hearing improvement. With regard to the time course of hearing recovery, there was no significant difference in hearing recovery among the long-, middle-, and shortterm administration groups until 16 weeks after treatment. However, at 16-24 weeks after initial treatment, the short-term administration group exhibited significantly lower hearing recovery than did the long-term administration groups.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions:</ns0:head><ns0:p>The administration duration of vit B12 and ATP did not influence the overall hearing prognosis in ISSNHL, but long-term administration of vit B12 and ATP helped prevent the progression of HL after ISSNHL. Our results suggest that long-term administration of vit B12 and ATP is not necessarily required to treat ISSNHL patients, except for slowly progressing HL in the affected ears.</ns0:p></ns0:div> </ns0:abstract> <ns0:body xmlns:ns0='http://www.tei-c.org/ns/1.0'> <ns0:div><ns0:head>Introduction</ns0:head><ns0:p>Idiopathic sudden sensorineural hearing loss (ISSNHL) usually occurs as an acute unilateral hearing loss (HL) <ns0:ref type='bibr' target='#b23'>(Staecker et al. 2019)</ns0:ref>. The exact cause of ISSNHL has not been identified, but several possible etiologies have been proposed, such as vascular dysfunction, neurological disorder, autoimmune disease, or viral infection <ns0:ref type='bibr' target='#b26'>(Watanabe et al. 2019)</ns0:ref>. Accordingly, the optimal treatment modality for ISSNHL remains controversial, but various comprehensive approaches have been widely adopted, including topical application of steroids, vasodilator, and vitamin supplementation, with systemic steroid therapy being the mainstream treatment <ns0:ref type='bibr' target='#b0'>(Ahmadzai et al. 2019</ns0:ref>). However, despite systemic treatment, the spontaneous recovery rate only ranges from 30% to 60% <ns0:ref type='bibr' target='#b20'>(Qiang et al. 2017)</ns0:ref>. Further, although steroid treatment has profound therapeutic benefits, it does not confer therapeutic benefit if the treatment is started more than 2 weeks after onset <ns0:ref type='bibr' target='#b3'>(Amarillo et al. 2019)</ns0:ref>. After completion of the initial steroid therapy, patients are prescribed various medications such as vitamin supplementation and vasodilator drugs continuously. Particularly, a combination of vitamin B12 (vit B12) and adenosine triphosphate (ATP) is prescribed for ISSNHL patients to improve the auditory neural function and cochlear blood circulation <ns0:ref type='bibr' target='#b0'>(Ahmadzai et al. 2019)</ns0:ref>.</ns0:p><ns0:p>The duration of vit B12 and ATP treatment depends on the period during which the treatment effect of these drugs is sustained. A patient would be willing to take vit B12 and ATP orally for the long term if these medications are likely to provide benefits for a longer period of time. However, the influence of the duration of vit B12 treatment with ATP on hearing prognosis is unclear. Thus, the present study aimed to investigate the influence of treatment duration of vit B12 with ATP on the hearing prognosis of ISSNHL. Towards</ns0:p></ns0:div> <ns0:div><ns0:head>Assessment of hearing function</ns0:head><ns0:p>Pure-tone audiometry was performed using a conventional device (AA-78; Rion, Tokyo, Japan) in a soundproof room. First, the hearing thresholds were obtained through air conduction (AC) and bone conduction (BC) in frequencies of 0.25, 0.5, 1, 2, and 4 kHz for both ears. To prevent cross-hearing phenomenon causing an erroneous measurement, masking noise was used to occupy the non-test ear while the other ear was tested, as necessary. Briefly, the necessity of masking during the measurement of the AC thresholds was identified based on the minimum interaural attenuation level of 40 dB for retesting through AC. Meanwhile, in the measurement of BC thresholds, a masking process was applied using ABC methods <ns0:ref type='bibr' target='#b14'>(Kurioka et al. 2020c)</ns0:ref>. Thresholds were obtained across all the frequency octaves from 0.25 kHz to 4 kHz, and the arithmetic average AC and BC thresholds were calculated from the thresholds at 0.25, 0.5, 1, 2, and 4 kHz. The grade of HL, defined according to the Japanese Ministry of Health and Welfare guidelines <ns0:ref type='bibr' target='#b17'>(Nakashima et al. 2014)</ns0:ref>, was then determined using the initial audiogram data. Hearing recovery was calculated as the difference between average hearing thresholds at different time points, including initial day of treatment and 2, 4, and 6 months after initial treatment. Siegel's criteria were employed to assess treatment results 6 months after treatment (Table <ns0:ref type='table'>1</ns0:ref>), and the patients were accordingly classified into two groups as the recovery group (i.e., complete and partial recovery) and the no recovery group (i.e., slight recovery and no improvement). These two groups were treated with same drug regimen, except for the duration of vit B12 and ATP administration.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Treatment All patients were treated with a 10-day course of systemic corticosteroids (betamethasone 8 mg via intramuscular injection for the first day followed by betamethasone 4 mg via oral administration for the first 3 days, tapered to 2 mg for the second 3 days, and finally to 1 mg for the last 3 days). Oral administration of vit B12 (1.5 mg daily) and ATP (300 mg daily) was started from initial day of treatment, concurrent with the corticosteroids. To investigate the therapeutic effects of long-term administration of vit B12 and ATP on hearing outcomes in ISSNHL, the patients were divided into three groups according to the administration duration as the short-term (&lt;8 weeks), middle-term (&#8805;8 weeks, &lt;16 weeks) and long-term (&#8805;16 weeks) administration groups.</ns0:p></ns0:div> <ns0:div><ns0:head>Statistical analyses</ns0:head><ns0:p>The Chi-squared test was used to evaluate clinical characteristics and possible prognostic factors. Student's t-tests and non-parametric Mann-Whitney U tests were used to investigate continuous prognostic factors. For comparisons between more than two groups, a one-way analysis of variance was used followed by Dunn's multiple comparisons for the post hoc test. The parameters that were statistically significant in the univariate analysis were entered into a binary logistic regression model for multivariate analysis. All statistical analyses were performed using GraphPad Prism 8 (GraphPad Software Inc., La Jolla, CA) or JMP 14.2 (SAS Institute Japan Inc., Tokyo, Japan). A p value of &lt;0.05 was considered statistically significant.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed</ns0:p></ns0:div> <ns0:div><ns0:head>Results</ns0:head></ns0:div> <ns0:div><ns0:head>Patient characteristics and treatment outcome</ns0:head><ns0:p>The mean age was 58.2 &#177; 16.6 years, and there were 66 and 51 male and female patients, respectively. The patients' clinicodemographic characteristics are presented in Table <ns0:ref type='table'>2</ns0:ref>. The mean interval between symptom onset and initial treatment was 4.2 &#177; 3.8 days. As accompanying symptoms and complications, 56 patients (47.9%) had vertigo and 32 patients (27.4%) had diabetes. The mean hearing threshold at the initial examination was 80.7 &#177; 24.0 dB. With respect to initial HL grade, 8 patients had grade 1; 18 patients, grade 2; 45 patients, grade 3; and 46 patients, grade 4. The average duration of vit B12 and ATP administration was 4.2 &#177; 2.4 months. With regard to final recovery according to Siegel's criteria, the overall recovery rate (complete + partial recovery) was 32.5%. Specifically, 19.7%, 12.8%, 39.3%, and 28.2% of the patients achieved complete recovery, partial recovery, slight recovery, and no improvement, respectively (Table <ns0:ref type='table'>1</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Prognostic factors</ns0:head><ns0:p>In univariate analysis, the initial hearing threshold and initial grade of HL was significantly lower in the recovery group than that in the no recovery group (p &lt; 0.0001). With regard to the associated symptoms, there was a significantly higher incidence of vertigo in the no recovery group than in the recovery group (p = 0.04). The other variables, including age, sex, days to initiation of treatment, and diabetes were not significantly different between the two groups. The duration of vit B12 and ATP administration was slightly PeerJ reviewing <ns0:ref type='table'>PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:ref> Manuscript to be reviewed longer in the no recovery group than that in the recovery group, but the difference was not significant. The significant variables in the univariate analysis were included in the multivariate analysis. The results showed that higher initial hearing threshold and higher initial grade of HL were associated with a poor prognosis in ISSNHL patients (Table <ns0:ref type='table'>3</ns0:ref>).</ns0:p></ns0:div> <ns0:div><ns0:head>Administration duration of vit B12 and ATP</ns0:head><ns0:p>The short-term, middle-term, and long-term administration groups involved 15, 41, and 38 patients, respectively. The patients with complete recovery within 24 weeks after onset (n = 23) were excluded because the vit B12 and ATP administration was terminated following the confirmation of complete recovery. Specifically, the duration of drug administration was shortened due to complete recovery, but the shortened administration did not contribute to complete recovery. Therefore, it was not appropriate to include the patients with complete recovery within 24 weeks after onset for the analysis of the effect of drug administration duration. In this study, the duration of administration was determined by the patients after consultation with their physician, even though they did not achieve complete recovery. The clinical features of each group are presented in Table <ns0:ref type='table'>4</ns0:ref>. The mean duration of drug administration in the short, middle, and long-term groups was 1.9, 3.3, and 6.3 months, respectively. In univariate analysis, the duration of drug administration was significantly different among the groups (p &lt; 0.0001), but the clinical characteristics and hearing results were comparable (p &gt; 0.05). This indicated that the duration of drug administration was not influenced by the patient's clinical characteristics and did not affect the overall hearing outcome. Ultimately, these results show that long-PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed term administration of vit B12 and ATP had no significant impact on the hearing prognosis within our observation period.</ns0:p><ns0:p>We also investigated the time course of hearing recovery in each group. As shown in Figure <ns0:ref type='figure' target='#fig_1'>1A</ns0:ref>, the thresholds (dB) of total hearing recovery within 24 months after symptom onset in the short-, middle-, and long-term administration groups were 18.9 &#177; 5.8, 22.4 &#177; 2.3, and 22.7 &#177; 3.0, respectively, with no significant differences (p = 0.76). The time course of hearing recovery in the overall population was 18.9 &#177; 1.7, 3.1 &#177; 0.7, and 0.04 &#177; 0.4 within 0-2, 2-4, and 4-6 months after symptom onset, respectively (p &lt; 0.0001; Fig. <ns0:ref type='figure' target='#fig_1'>1B</ns0:ref>). Hearing recovery within the initial 2 months after symptom onset was significantly higher than that within 2-4 months (p &lt; 0.0001) and 4-6 months (p &lt; 0.0001). Moreover, hearing recovery within 2-4 months after symptom onset was significantly higher than that within 4-6 months (p = 0.04). This result indicated that hearing recovery in the initial 2 months after onset is higher and that the probability of recovery decreases as the duration after symptom onset increases. The time course of hearing recovery of both the affected and contralateral ears in the three groups are shown in Figure <ns0:ref type='figure' target='#fig_3'>2</ns0:ref>. There was no significant difference in hearing recovery among the three groups until 16 weeks after onset. However, hearing recovery of the affected ears was significantly lower in the shortterm administration group than that in the long-term administration group at 16-24 weeks after symptom onset (p = 0.03). Meanwhile, there was no significant difference in hearing recovery of the contralateral ears among the three groups throughout the study period.</ns0:p><ns0:p>This indicated a slow progression of HL in the affected ear and that long-term administration of vit B12 and ATP might prevent progressive HL in the affected ear in ISSNHL. We performed an additional analysis of hearing recovery at each frequency. As PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed shown in Figure <ns0:ref type='figure' target='#fig_4'>3</ns0:ref>, hearing recovery appeared to be greater at 0.5 and 1.0 kHz than that at 2.0 and 4.0 kHz until 8 weeks after onset. In addition, significant differences between long-term and short-term drug administration were observed at only 2.0 kHz during the 4-6 months after the onset of ISSNHL.</ns0:p></ns0:div> <ns0:div><ns0:head>Discussion</ns0:head><ns0:p>The influence of the duration of vit B12 and ATP treatment on hearing prognosis has been unclear. In our study, the administration periods of vit B12 and ATP were not prognostic factors. However, this result does not necessarily indicate that long-term administration of vit B12 and ATP has no therapeutic effect on the overall prognosis of ISSNHL as hearing recovery was significantly lower in the short-term administration group than that in the middle and long-term administration groups within 4-6 months after onset. This indicates that vit B12 and ATP treatment might have gradual therapeutic effects on preventing the deterioration in hearing rather than promoting the hearing recovery.</ns0:p><ns0:p>Hearing recovery within the initial 2 months was also significantly higher than that in the later phase, indicating that early initial treatment within 2 months after symptom onset is extremely important.</ns0:p><ns0:p>The optimal treatment strategy for ISSNHL has not been established to date because the cause and etiology of ISSNHL remain unclear. However, circulation and neurological disorders are thought to be one of the major causes for ISSNHL <ns0:ref type='bibr' target='#b5'>(Hsu et al. 2016)</ns0:ref>.</ns0:p><ns0:p>Because the cochlea is supplied by the labyrinthine artery, which has no collateral circulation, the obstruction of blood supply by thrombosis or hemorrhage may cause PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed cochlear damages that results in ISSNHL. Diabetes, hyperlipidemia, and aging are also well-known factors of microvascular disease that results in blood circulation disorders in the inner ear <ns0:ref type='bibr' target='#b1'>(Akinpelu et al. 2014;</ns0:ref><ns0:ref type='bibr' target='#b18'>Orita et al. 2007)</ns0:ref>. Various empirical treatments have been applied to improve blood circulation and to increase oxygen supply to the inner ear, including vasodilators, steroids, plasma expanders, and anticoagulant agents <ns0:ref type='bibr' target='#b27'>(Xie et al. 2018)</ns0:ref>. Particularly, systemic steroids have become the most commonly used treatment for ISSNHL patients, with the dose tapered within 2 weeks <ns0:ref type='bibr' target='#b22'>(Slattery et al. 2005)</ns0:ref>. The use of intra-tympanic steroid injection for the treatment of ISSNHL has also recently increased owing to its comparable effects to systemic steroids and the additional benefit of salvage treatment <ns0:ref type='bibr' target='#b10'>(Kordis &amp; Battelino 2017)</ns0:ref>. After the initial steroid treatment, which usually lasts for 1-2 weeks, patients are generally prescribed vit B12 and ATP supplementation for further management.</ns0:p><ns0:p>Vitamin B12 plays an important role in the maintenance of normal neural function <ns0:ref type='bibr' target='#b2'>(Altun &amp; Kurutas 2016)</ns0:ref>. Accordingly, vit B12 deficiency leads to anemia, demyelination, axonal degeneration, and, ultimately, neuronal loss. In the auditory system, vit B12 deficiency is known to have negative effects on hearing by affecting myelinization of the auditory neurons at the retro-cochlear region. Additionally, auditory neural degenerations, such as disruption of cochlear synapses, demyelination, and shrinkage of auditory nerves, gradually progress after decreased auditory inputs <ns0:ref type='bibr' target='#b12'>(Kurioka et al. 2020a;</ns0:ref><ns0:ref type='bibr' target='#b13'>Kurioka et al. 2020b</ns0:ref>). Vitamin B12 supplementation has been shown to increase the number of Schwann cells and myelinated nerve fibers and the diameter of the axons, thereby promoting the regeneration of myelinated nerve fibers and the proliferation of Schwann cells <ns0:ref type='bibr'>(Lopatina et al. 2011)</ns0:ref>. Therefore, supplementary vit B12 treatment is considered in PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed auditory diseases to prevent neural disorders and to enhance neurovascular endothelial function <ns0:ref type='bibr' target='#b21'>(Singh et al. 2016)</ns0:ref>. Vitamin B12 supplementation also reduces homocysteine synthesis, which is a vascular and thrombotic risk factor and causes vascular injury by reducing the amount of nitric oxide (NO). The reduced homocysteine synthesis leads to vasodilatation as a result of an increase in the amount of NO <ns0:ref type='bibr' target='#b25'>(Toda &amp; Okamura 2016)</ns0:ref>.</ns0:p><ns0:p>Thus, vit B12 can cause an increase in vascular perfusion and cellular metabolism in the cochlea.</ns0:p><ns0:p>ATPs are often used as vasodilators to increase cochlear blood flow in auditory diseases. The vasodilator effect of ATP is mediated by the endothelium and follows the release of NO. A previous study reported that treatment with ATP resulted in a large increase in cochlear blood flow and a decrease in blood pressure owing to the vasodilator actions of ATP <ns0:ref type='bibr' target='#b16'>(Munoz et al. 1999)</ns0:ref>. Another study reported that a 300 mg dose of ATP has therapeutic effects for inner ear pathologies, such as Meniere's disease <ns0:ref type='bibr'>(Mizukoshi et al. 1983</ns0:ref>). Thus, ATP and vit B12 are widely used for the treatment of ISSNHL owing to their various biological therapeutic effects.</ns0:p><ns0:p>The prognosis for recovery from ISSNHL can be influenced by various factors, including age, accompanying symptoms such as vertigo, degree of HL, audiometric configuration, and time to treatment initiation <ns0:ref type='bibr' target='#b9'>(Kim et al. 2018;</ns0:ref><ns0:ref type='bibr' target='#b11'>Kuhn et al. 2011;</ns0:ref><ns0:ref type='bibr'>Lin et al. 2016)</ns0:ref>. The findings of the present study indicated that the initial hearing threshold and initial grade of HL were useful prognostic factors, consistent with the findings of a previous study <ns0:ref type='bibr' target='#b11'>(Kuhn et al. 2011)</ns0:ref>. With respect to the treatment period after ISSNHL onset, it has been reported that hearing remained relatively stable after a period of 2-3 months from symptom onset <ns0:ref type='bibr' target='#b6'>(Kallinen et al. 2001;</ns0:ref><ns0:ref type='bibr' target='#b7'>Kanzaki et al. 1988;</ns0:ref><ns0:ref type='bibr' target='#b19'>Psifidis et al. 2006)</ns0:ref>. It is possible Manuscript to be reviewed that the period of 2 months could be consistent with the natural history of the disease, regardless of which therapeutic strategy is applied. Thus, any additional treatment after 2 months should not affect the outcome of the hearing, consistent with our results.</ns0:p><ns0:p>However, a long-term follow-up study reported that 9.9% patients showed an improvement over a period of 3 months after the onset of ISSNHL <ns0:ref type='bibr' target='#b28'>(Yeo et al. 2007</ns0:ref>).</ns0:p><ns0:p>Nevertheless, one study found that 25% of patients had long-term hearing deterioration after the onset of ISSNHL <ns0:ref type='bibr' target='#b4'>(Furuhashi et al. 2002)</ns0:ref>. This is consistent with our results that ISSNHL affected the involved ear more gradually than the contralateral ears. This hearing deterioration appeared to have no relationship with age-related HL because the deterioration was observed only in the affected ears in this study. Although the overall hearing recovery did not differ according to the duration of administration, we found some benefit for the prevention of hearing deterioration with long-term administration. However, these results generally do not indicate that long-term vit B12 and ATP should be adopted for ISSNHL patients. Furthermore, it is generally believed that hearing recovery would differ between acute low-tone sensorineural HL and ISSNHL. Therefore, the initial type of HL in ISSNHL might be associated with delayed changes in hearing after the onset.</ns0:p><ns0:p>Additional studies with a longer observation period are needed to elucidate the accurate therapeutic effects of vit B12 and ATP in ISSNHL.</ns0:p><ns0:p>This study has some limitations. First, this was a retrospective study conducted in a single hospital, and the sample size was relatively small. Second, the duration of vit B12 and ATP treatment in this study was influenced by various factors, such as patient preferences and drug therapeutic effects. However, despite these limitations, we believe that our study is valuable because to the best of our knowledge, this is the first to focus on the PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed relationship between the duration of vit B12 and ATP administration and prognosis in ISSNHL. Our findings have important clinical implications in the understanding and management of patients with ISSNHL. Although the underlying therapeutic mechanisms of vit B12 and ATP in ISSNHL are not well known, its effects need to be considered in the management of ISSNHL in the clinic. Further prospective studies with larger populations are needed to validate our findings.</ns0:p></ns0:div> <ns0:div><ns0:head>Conclusions</ns0:head><ns0:p>The administration duration of vit B12 and ATP did not influence the hearing prognosis in ISSNHL. Higher initial hearing threshold and higher initial grade of HL were associated with a poor prognosis. Compared to the contralateral ear, HL progressed gradually in the affected ear. Long-term administration of vit B12 and ATP might help prevent progressive HL in the affected ear.</ns0:p><ns0:p>PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:p><ns0:p>Manuscript to be reviewed Hearing recovery in the affected ears was comparable among the three groups within 0-2 and 2-4 months after onset. Hearing recovery differed significantly between the short-term (black) and long-term administration groups (red) at 4-6 months after onset only at 2.0 kHz (C). P values are indicated as *p &lt;0.05.</ns0:p></ns0:div><ns0:figure xml:id='fig_0'><ns0:head /><ns0:label /><ns0:figDesc>PeerJ reviewing PDF | (2020:06:50084:1:1:NEW 30 Oct 2020)</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_1'><ns0:head>Figure 1</ns0:head><ns0:label>1</ns0:label><ns0:figDesc>Figure 1</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_2'><ns0:head>(</ns0:head><ns0:label /><ns0:figDesc>A) The total hearing threshold recovery from symptom onset to 6 months was comparable among the three groups. (B) Hearing recovery in the initial 2 months was significantly higher than that in the later periods. Hearing recovery within 2-4 months after symptom onset was also significantly higher than that at 4-6 months after onset. p values are indicated as **p &lt; 0.01, ****p &lt; 0.0001.</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_3'><ns0:head>Figure 2</ns0:head><ns0:label>2</ns0:label><ns0:figDesc>Figure 2</ns0:figDesc></ns0:figure> <ns0:figure xml:id='fig_4'><ns0:head>Figure 3</ns0:head><ns0:label>3</ns0:label><ns0:figDesc>Figure 3</ns0:figDesc></ns0:figure> <ns0:figure type='table' xml:id='tab_0'><ns0:head /><ns0:label /><ns0:figDesc>Watanabe Y, Watanabe I, Okubo J, Matsunaga T, Matsunaga T, Takayasu S, Kato I, and Tanaka T. 1983. Subjective and objective evaluation of medical</ns0:figDesc><ns0:table><ns0:row><ns0:cell cols='9'>Lin CF, Lee KJ, Yu SS, and Lin YS. 2016. Effect of comorbid diabetes and</ns0:cell></ns0:row><ns0:row><ns0:cell cols='9'>hypercholesterolemia on the prognosis of idiopathic sudden sensorineural hearing</ns0:cell></ns0:row><ns0:row><ns0:cell cols='6'>loss. Laryngoscope 126:142-149. 10.1002/lary.25333</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell cols='9'>Lopatina T, Kalinina N, Karagyaur M, Stambolsky D, Rubina K, Revischin A, Pavlova G,</ns0:cell></ns0:row><ns0:row><ns0:cell cols='9'>Parfyonova Y, and Tkachuk V. 2011. Adipose-derived stem cells stimulate</ns0:cell></ns0:row><ns0:row><ns0:cell cols='9'>regeneration of peripheral nerves: BDNF secreted by these cells promotes nerve</ns0:cell></ns0:row><ns0:row><ns0:cell>healing</ns0:cell><ns0:cell>and</ns0:cell><ns0:cell>axon</ns0:cell><ns0:cell>growth</ns0:cell><ns0:cell>de</ns0:cell><ns0:cell>novo.</ns0:cell><ns0:cell>PLoS</ns0:cell><ns0:cell>One</ns0:cell><ns0:cell>6:e17899.</ns0:cell></ns0:row><ns0:row><ns0:cell cols='4'>10.1371/journal.pone.0017899</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row><ns0:row><ns0:cell>Mizukoshi K,</ns0:cell><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /><ns0:cell /></ns0:row></ns0:table></ns0:figure> </ns0:body> "

"Article ID: 50084 Title: Long-term administration of vitamin B12 and adenosine triphosphate for idiopathic sudden sensorineural hearing loss: a retrospective study Point-by-Point Response In this document, we have included our point-by-point responses to the Reviewers’ comments, as well as descriptions of the changes made to the manuscript. The Reviewer comments are in black font, while our responses are in red font. Throughout the text, we have incorporated minor editorial changes to improve readability and deleted some redundant phrases or words. Changes in the manuscript text are indicated by red font. Please note that the changes made do not influence the content, conclusions, or framework of the paper. Editor comments I would also like to see some more background on the use of ATP as a treatment ... my understanding is that ATP at this dose would not be significantly bio-available, and while the paper cited on guinea pig cochlear blood flow and ATP is certainly relevant preclinical evidence, is there any evidence in people that might speak to how ATP might affect recovery from/progression of hearing loss ? Response: Thank you for your comments. Although a 300 mg dose of ATP is widely used in otology clinics (Sano et al., 2010) (Kitoh et al., 2017), the precise effects and bioavailability of this ATP dose is not clear. It has been reported that a 300 mg dose of ATP has therapeutic effects for inner ear pathology, such as Meniere’s disease (Mizukoshi et al., 1983). We have discussed this issue and added the relevant references to the revised manuscript. Reviewer 1 1, My understanding is that hearing level moves longer in low-tone frequencies that in high-tone frequencies after the onset of ISSNHL. What was the frequency area of the slowly progressing hearing loss during 4 – 6 months after the onset of ISSNHL? Response: Thank you for your critical comments. We have performed an additional analysis of hearing recovery at each frequency. As shown in Fig. 3 of the revised manuscript, significant differences between long-term and short-term drug administration were only observed at 2.0 kHz during 4–6 months after the onset of ISSNHL. Furthermore, hearing recovery was greater at 0.5 and 1.0 kHz than that at 2.0 and 4.0 kHz. We have added these results to the “Results” section. 2, It is generally considered that hearing recovery can be expected longer in acute low-tone SNHL than in ordinary ISSNHL. I expect some comments about the initial hearing type of ISSNHL in association with the delayed hearing change after the onset of ISSNHL. Response: We agree with the reviewer’s comments. The prognosis of acute low-tone SNHL would be better and hearing would recover more slowly. Therefore, the initial type of hearing loss in ISSNHL might be associated with delayed changes in hearing after onset. However, as shown in Fig. 3 of the revised manuscript, hearing recovery was significantly lower in the short-term administration group than in the long-term administration group at 16–24 weeks after symptom onset only at a frequency of 2.0 kHz, which is not a low frequency. This might indicate that the patients showing progressive HL in this study have ISSNHL but not acute low-tone SNHL. We have added these comments to the “Results” and “Discussion” sections. 3, Furuhashi’s paper (Clinical Otolaryngol, 2002) described recurrence of ISSNHL for a very long period like ten years. I would like to recommend a paper investigating for more than three months after the onset of ISSNHL for one of the references. “Yeo, SW, Lee DH et al.: Hearing outcome of sudden sensorineural hearing loss: Long-term follow-up. Otolaryngol Head Neck Surg 136;221-224. 2007” Response: Thank you for your suggestion. We have added the above reference. "

Here is a paper. Please give your review comments after reading it.