text
string | predicted_class
string | confidence
float16 |
|---|---|---|
Consistent with the predominant excitatory nature of cortico-cortical feedback connections, reversible inactivation of area V2 (cat: Alonso et al., 1993b; Martinez-Conde et al., 1999; several species of monkeys: Sandell and Schiller, 1982; Bullier et al., 1996; Hupé et al., 2001a) results in reduction in spike-responses to optimized visual stimuli of a substantial proportion of V1 neurons. | study | 100.0 |
In our study, inactivation of the entire ipsilateral PTV cortex resulted in the reduction (albeit rarely significant) of background spike-activities of simple and complex area 19 cells. Also during inactivation, in a quarter (significant) to half of the sample, the magnitude of spike-responses was reduced. Interestingly, inactivation of higher-order pattern/form-processing areas 21a and 19, reciprocally and strongly interconnected with area 17, results in reduction in the magnitude of spike-responses to optimal visual stimuli in a great majority of affected area 17 neurons (Wang et al., 2000). The excitatory nature of feedback cortico-cortical connections is not restricted to pattern/form processing stream. In particular, reversible inactivation of cat’s motion area—posterior middle suprasylvian cortex (Galuske et al., 2002) caused reduction in the signal strength in orientation and direction maps in the parastriate cortex (area 18, V2). Similarly, cooling motion-processing medio-temporal (MT) cortex of macaque monkeys results in reductions of magnitude of spike-responses of neurons in the ipsilateral areas V1, V2 and V3 (Hupé et al., 2001b). | study | 100.0 |
The question arises if the visual signals arriving via feedback cortico-cortical connections can per se drive the spike-activity of cortical neurons. There is a clear evidence that in the cat, transient reversible inactivation of small volume of layer A of LGNd results in virtual silencing of simple and complex cells in area 17 whose RFs are in the precise retinotopic alignment with the center of GABA injections into the LGNd (Martinez and Alonso, 2001). On the other hand, when direct input from layer A of LGNd is reversibly blocked, the visuotopically corresponding part of cat’s area 18 is able to provide effective excitatory feedback drive to the supragranular layers of area 17 (Mignard and Malpeli, 1991). Overall, relatively selective removal of the feedback input from the higher-order cortical areas, without removing the other channels through which the visual input can reach a cortical neuron, might render the neuron unresponsive to visual stimuli. Thus, some of our area 19 neurons ceased to respond to visual stimuli when almost entire ipsilateral PTV cortex has been selectively inactivated, and their responsiveness to visual stimuli was restored when PTV was rewarmed. Similarly, when the entire ipsilateral area 18 (V2), rather than only the part corresponding visuotopically to the regions in V1 recorded from, was inactivated, some V1 neurons ceased to respond to visual stimuli (Sandell and Schiller, 1982). Their responsiveness to visual stimuli was restored when V2 was rewarmed (Sandell and Schiller, 1982). | study | 100.0 |
Despite the excitatory nature of feedback projections: (1) reversible inactivation of circumscribed regions of superficial layers 2/3 of cat’s area 18 results in an increase in the magnitude of spike-responses of 33%–55% of neurons in layers 2/3 in the visuotopically corresponding part of area 17 (Martinez-Conde et al., 1999); and (2) in almost a third of the present area 19 sample, inactivation resulted in increases in the magnitude of spike-responses, and in ~10% of the sample those increases were significant. Most likely, the suppressive effects are conveyed by inhibitory local interneurons, intrinsic to the area recorded from. In rats, although 98% of feedback projections from area LM (presumptive V2) form excitatory synapses on excitatory neurons in area 17, ~2% make their connections with GABAergic interneurons (Johnson and Burkhalter, 1996). Although in the cat only a small proportion of the intracortical connections appear to be inhibitory (Gabbott and Somogyi, 1986; Shao and Burkhalter, 1996; Shevelev et al., 1998), almost every neuron in the cat primary visual cortex receives some inputs from local inhibitory interneurons (for review, see Eysel, 2002). In all mammals studied so far, 20%–30% of neocortical neurons are interneurons and most of them are inhibitory (DeFelipe, 1993; Markram et al., 2004). Significant increments in peak responses of area 19 neurons during inactivation are putatively due to the removal of excitatory input from PTV on GABAergic interneurons in area 19. | review | 69.56 |
One has to be cautious in determining the overall proportions of area 19 cells receiving excitatory input vs. those receiving inhibitory input from PTV as they are not necessarily mutually exclusive. Thus, a number of area 19 cells appeared to receive an excitatory feedback input from PTV when their responses to optimally oriented stimuli were examined but appeared to receive suppressive feedback from PTV when the responses to sub-optimally oriented stimuli were examined (e.g., Figure 3Aii). Furthermore, the strength of suppressive feedback effects was substantially greater when the stimuli were moving faster (e.g., Figure 4; see also Figure 5 for area 17 cells). | study | 100.0 |
Overall, it appears that feedback projections from PTV can exert excitatory and suppressive effects on spike-responses of area 19 neurons recorded from layers 3/4. It is possible that at least some of these effects are exerted via cortico-thalamic-thalamo-cortical loops (see Sherman, 2005; Sherman and Guillery, 2011). In the case of area 17 such loops would involve: PTV cortex—lateral posterior/pulvinar thalamic complex—area 17; in the case of area 19 it would involve: PTV cortex—lateral posterior/pulvinar thalamic complex—area 19 (Rosenquist, 1985; Dreher, 1986; Casanova et al., 1997; Piché et al., 2015). There is an important caveat to our conclusions due to the fact that we do not have information about the effects of inactivation of PTV on area 19 neurons in the superficial and deep layers. | study | 100.0 |
Consistent with numerous earlier reports (Hubel and Wiesel, 1965; Duysens et al., 1982a, b; Rapaport et al., 1982; Dreher, 1986; Pettigrew and Dreher, 1987; Tanaka et al., 1987; Guillemot et al., 1993; Tardif et al., 1997), area 19 neurons exhibit clear orientation selectivity (see however Saito et al., 1988). Orientation selectivity of V1 neurons relies on multiple mechanisms, including mechanisms that are intrinsic to the area as well as the orientation biased inputs from the geniculate (for reviews, see Vidyasagar et al., 1996; Vidyasagar and Eysel, 2015) and feedback signals from layer 5 (Alonso et al., 1993b) or layers 2/3 (Martinez-Conde et al., 1999) of visuotopically corresponding parts of area 18 and/or higher-order visual cortices (e.g., Wang et al., 2000, 2010; Huang et al., 2007). It is likely, that in the case of area 19 neurons, orientation selectivity is to a large extent “inherited” via feedforward projections from the orientation selective neurons in the primary visual cortex. Nevertheless, in a small proportion of area 19 neurons, PTV inactivation caused clear (>10°) shifts in optimal orientation and/or substantial changes in the orientation-tuning widths. It is worth pointing out in this context, that in area 17 of anesthetized cats, shifts in preferred orientation not exceeding 15° do occur “spontaneously” (e.g., Henry et al., 1973). | review | 97.8 |
It is commonly assumed that velocity preferences of visual neurons closely reflect the velocity preferences of geniculate neurons which provide their “driving” (Sherman and Guillery, 1998) excitatory input (e.g., Dreher et al., 1980; Orban et al., 1981; Duysens et al., 1982a; for review, see Orban, 1984). However, a substantial proportion of area 19 cells appear to receive Y type afferents from the LGNd (presumably responding to fast moving stimuli) do not respond to fast moving stimuli (Dreher et al., 1980). Similarly, some V1 neurons which receive geniculate afferents reliably activated by fast moving stimuli, presumably due to intracortical inhibition, do not generate action potentials when fast moving stimuli are presented (Goodwin and Henry, 1978). It is also worth noting in this context that: (1) the cut-off and preferred velocities of area 18 neurons tend to be much higher than those of most area 17 neurons (e.g., Dreher et al., 1980; Orban et al., 1981; for review, see Orban, 1984); and (2) reversible inactivation of circumscribed part of layer 5 of area 18 (V2), results in reversible upward shift in cut-off velocities and/or preferred velocities of a large (40%) proportion of layer 5 neurons in the visuotopically corresponding region of area 17 (Alonso et al., 1993a). | review | 99.7 |
In this study, inactivation of PTV resulted in an upward shift in preferred velocities of some area 17 cells and in changes in their velocity preferences, with a tendency of a shift to higher velocities of over a third of area 19 neurons. Thus, it appears PTV cells providing the feedback projections to area 19 prefer higher velocities than the preferred velocities of area 19 cells based on the velocities preferred by the integrated input to area 19. Due to the fact that the highest velocity tested by us was 66°/s, and it was lower than the cut-off velocities of most area 19 neurons when PTV cortex was intact, we were unable to assess if inactivation of PTV resulted also in an upward shifts in the cut-off velocities of a substantial proportion of area 19 neurons. Nevertheless, in a couple of area 19 cells, cooling resulted in reversible upward shifts of their high cut-off velocities. Thus, the feedback from PTV might normally suppress the responses of area 19 cells to fast moving stimuli. | study | 100.0 |
The direction selectivities of cells in primary visual cortex of the cat are believed to be determined by a number of thalamo-cortical and intracortical mechanisms (for reviews, see Orban, 1984; Eysel, 2002; Humphrey and Saul, 2002). There is also substantial evidence indicating that in both carnivores and primates, feedback from the higher-order cortical areas affects the direction selectivity of neurons in the primary visual cortices. Thus, in the New-World squirrel monkey, during reversible inactivation of area 18 (V2), some direction selective cells in the ipsilateral area 17, lose most of their direction selectivity (Sandell and Schiller, 1982). In the cat, reversible inactivation of pattern/form processing area such as area 21a and a part of area 19 results in large (>20%) changes in DSIs of ~20% of neurons in the visuotopically corresponding part of area 17 (Wang et al., 2000). The increases in DSI of neurons in areas 19 and 17 (see also Huang et al., 2007) accompanying upward changes in velocity preferences during cooling of PTV, may be related to the fact that in the cat’s primary visual cortices the DSIs vary with velocities of stimulus motion (Orban et al., 1981; Orban, 1984; Humphrey and Saul, 2002). | review | 75.2 |
The DSIs of neurons in the higher-order pattern processing areas might depend on some other factors. While normally neurons in area V4 (presumed homolog of area 21a of the cat) of behaving macaque monkeys exhibit very little direction selectivity, after prolonged exposure and adaptation to unidirectional motion of coherent random dot pattern, a third of the sample became strongly direction selective (Tolias et al., 2005). | study | 100.0 |
A large body of data indicates that dynamic changes in the magnitude of responses and RF properties of neurons in the lower cortical areas, including the primary visual cortex, are affected by the recurrent feedback signals from the higher-order cortical visual areas (see “Discussion” Section and review by Gilbert and Li, 2013). Hochstein and Ahissar (2002) proposed that high-acuity-yet analytical “vision with scrutiny” is based on the information derived from feedback-modified neuronal discharge activities of cells in the lower-order visual areas (“Reverse Hierarchy Theory”). Although there is a substantial degree of cross-talk between feedback pathways in both parallel information processing streams, the effects of inactivation of higher-order areas in the pattern/form processing stream vs. those in the motion processing stream are quite different. For example, in behaving cats, reversible inactivation of the ventral-posterior suprasylvian cortex, a region corresponding closely to PTV, results in severe impairment of retention of highly-familiar simple pattern discrimination when components of the patterns are stationary or in motion (Lomber et al., 1996a,b). By contrast, inactivation of motion processing areas lining the middle suprasylvian sulcus (for review, see Spear, 1991), affects discrimination of highly familiar simple patterns only when all components of the patterns are moving (Lomber et al., 1996b). | review | 99.9 |
In anesthetized cat, silencing the feedback from motion processing areas in the middle suprasylvian region (for review, see Spear, 1991) tends to fairly selectively reduce the responsiveness of those area 18 neurons which exhibit a high degree of direction selectivity (Galuske et al., 2002). Furthermore, inactivation of feedback from motion processing middle suprasylvian region modulates direction selectivity but not orientation preferences of area 17 neurons (Shen et al., 2006). By contrast, inactivation of feedback from the pattern/form processing PTV, tends to shift upward preferred velocities and increase DSIs of areas 19 and 17 neurons (current study). Further the effects of inactivation of a particular higher-order area in a given processing stream on responses of neurons in a particular lower-order area in the same stream, are also quite specific. Thus, inactivation of area 21a affects strongly orientation (Wang et al., 2000; Tong et al., 2011) and spatial frequency tuning (Huang et al., 2004; Tong et al., 2011) but not direction selectivity tuning (Tong et al., 2011) of neurons in the ipsilateral area 17. It could be argued that feedback signals from the higher-order pattern processing areas, increase the “sparseness of coding” (Olshausen and Field, 2004) by neurons in the lower-order areas. Thus, feedback from area 21a enhances orientation selectivity of neurons in area 17 (Wang et al., 2007), while feedback from PTV suppresses responses of many area 19 and some area 17 neurons to faster moving stimuli (current study). This might reduce dynamic “cluttering” of visual scene during fast exploratory saccadic eye movements (Crommelinck and Roucoux, 1976; Blakemore and Donaghy, 1980) and allow the higher-order pattern processing areas to “concentrate” on processing visual information during small exploratory microsaccades, as well as slow drifts (Hebbard and Marg, 1960; Pritchard and Heron, 1960), plus slow disjunctive eye movements (Stryker and Blakemore, 1972) occurring during fixation of objects of interest. It appears that postulated (Pettigrew and Dreher, 1987; Guillemot et al., 1993) strong involvement of area 19 in stereoscopic analysis of the visual scene is restricted to analysis of static, rather than moving, interocular positional disparities (Mimeault et al., 2002). | review | 95.4 |
It appears that irrespective of the information processing stream in which any given areas are “embedded”, feedback from a particular higher-order cortical area modulates those properties of neurons in the lower area, which play important roles in the information processing in the higher-order areas. Thus, cortical areas in the middle suprasylvian region want to “hear” about the direction of stimulus movement and their feedback signals enhance the magnitude of responses of strongly direction selective cells in the lower areas. By contrast, the pattern/form processing PTV cortex wants to “hear” about the specific properties of stationary/slowly moving stimuli and by feedback signals suppresses the responses of cells in the lower areas such as area 19 or area 17 to fast moving stimuli. | study | 99.94 |
In about a third of the sample of area 19 neurons, inactivation of higher-order pattern-processing PTV cortex resulted in significant changes in the magnitude of their peak responses to optimally oriented elongated moving bars. Only occasionally, silencing feedback signals from PTV resulted in an abolishment of the spike-responses of area 19 neurons. In a small proportion of area 19 cells, feedback from PTV appears to play a modulatory role in relation to such RF properties as optimal orientation and/or width of orientation-tuning curves. In a large proportion of area 19 cells and a proportion of area 17 cells, feedback from PTV plays a modulatory role in relation to stimulus velocity preferences and/or direction selectivity, that is, the properties which are usually believed to be determined by the inputs from the dorsal thalamus and/or feedforward inputs from the primary visual cortices. Thus, apparent specialization of area 19 for processing information about stationary/slowly moving visual stimuli is at least partially determined, by the feedback from the higher-order pattern-processing visual area. | study | 100.0 |
Among developed countries, average maternal age at first birth has been rising steadily in recent decades . In countries such as the United Kingdom (UK), Germany, and Korea, average maternal age at first birth is around 30. In England and Wales, the fertility rate from 1981 to 2012 decreased by about 30% among women aged between 20 and 24, but has doubled among women above 35 . In Japan, average maternal age at first birth increased from 24.4 in 1950 to 30.3 in 2013 . | other | 99.9 |
From an obstetric perspective, older maternal age is associated with increased risks of adverse pregnancy and birth outcomes. For example, risks of gestational diabetes and pre-eclampsia both increase with maternal age, as well as fetal risks of low birthweight, preterm births, miscarriage, and congenital anomalies [4–7]. However, studies on the potential risks of advanced maternal age on children’s later health outcomes have remained scarce. | review | 99.75 |
To our knowledge, only a few population-based studies have explored the relationship between maternal age and children’s health. In a UK-based study, Sutcliffe et al. examined the associations of maternal age with child health and development at 9 months, 3 years, and 5 years of age . They found that older maternal age was generally associated with better health (i.e., unintentional injuries and hospital admissions) after controlling for various factors, including income, maternal education, and paternal age. | study | 99.94 |
Another population-based study was conducted in Canada using the National Longitudinal Survey of Children and Youth . Researchers found that children’s general health seemed to improve with older maternal age. However, differences in child health across maternal age groups were not statistically significant once socioeconomic factors were considered. | study | 99.94 |
The findings from the UK and Canada have been suggestive, but no similar studies have yet been undertaken with Asian populations. The rise in average maternal age has also been observed in East Asian countries such as Japan and Korea, but social and cultural context differs greatly from Western countries [3, 10]. Therefore, we sought to examine the association between maternal age and children’s general health outcomes in Japan using the nationwide Longitudinal Survey of Babies in 21st Century. Based on the UK study findings, we hypothesized that older maternal age is associated with better child health. | study | 99.94 |
The Longitudinal Survey of Babies in 21st Century (LSB21) is an ongoing study in Japan comprising two birth cohorts, in which the first started in 2001 and the second in 2010. The main purpose of the LSB21 is for the Ministry of Health, Labour and Welfare (MHLW) to develop strategies against the decline of fertility among Japanese families . Details of the LSB21 have been described elsewhere, but the 2010 cohort has not been analyzed [12–14]. | study | 99.2 |
Families from all over Japan were considered eligible to participate in the survey if their child was born between the 10th and 17th of January or July in 2001 or between the 10th and 24th of May in 2010. Baseline questionnaires were sent to all families when infants reached 6 months of age (cf. Wave 1). For both cohorts, the response rates were 88%. Follow-up questionnaires were sent to participating families every 12 months, at 18 months (Wave 2), 30 months (Wave 3), and so on. Birth record information included birthweight, gestational age, sex of the child, and parental age. We obtained information on Waves 1–12 (the 2001 cohort) and Waves 1–3 (the 2010 cohort) from the MHLW. Participant numbers and loss to follow-up are summarized in Figs 1 and 2. This study was approved by the Japanese National Center for Child Health and Development Institutional Review Board (No. 1007). | study | 100.0 |
For both cohorts, we obtained parental age at delivery from their children’s birth records. We classified maternal age into five categories: 1) younger than 25.0 years; 2) 25.0–29.9 years; 3) 30.0–34.9 years; 4) 35.0–39.9 years, and 5) 40.0 years and older. The Japan Society of Obstetrics and Gynecology considers 35 years or older as advanced maternal age for first-time births . | study | 99.94 |
In both cohorts, children’s history of unintentional injuries was ascertained from parents at Wave 2, or at the child’s age of 18 months, with the question: “Did your child experience unintentional injuries during the past 12 months?”. Respondents who confirmed their child’s experience of unintentional injuries were then asked to select the type(s) of injury from the following 10 options: 1) a fall from bed, stairs, etc; 2) trapped an arm or a leg between doors or windows; 3) cut with a sharp object such as a knife; 4) bitten by an animal or stung by a bee; 5) drowned or near-drowned; 6) swallowed a small object, such as a coin or cigarette; 7) inserted a small object into the eye, ear, or nose; 8) burnt by a hot object such as a hot iron; 9) had a traffic accident, and 10) other injury. Children’s history of hospital admissions was also ascertained from the parents at Wave 2 with the question: “Was your child admitted to hospital during the past 12 months?”. We assessed the injury outcome at Wave 2 and hospital admission outcomes at Waves 2 and 6 in the 2001 cohort (representing 18- and 66-month longitudinal follow-ups). Similarly, in the 2010 cohort, we assessed injury and hospital admission outcomes at Wave 2. The wording of ascertainment on unintentional injuries and hospital admissions remained exactly the same from the 2001-cohort survey to the 2010-cohort survey. | study | 100.0 |
For both cohorts, we considered the same set of socioeconomic and biological factors as potential confounders: maternal and paternal education, maternal smoking status, maternal employment status, household income, sex of the child, maternal parity, and paternal age. We first did not include preterm births and birthweight in our statistical model because these factors are likely to be on the causal pathway between maternal age and child health (i.e. they may be considered intermediary variables) . In the follow-up analyses, however, we additionally included preterm births and birthweight in our analyses to account for their potentially mediating effects. We also examined whether maternal parity (i.e., the experience of having more children) influences the relationship between maternal age and child health via increased knowledge of infant care. | study | 100.0 |
We obtained data on paternal age, sex of the child, and parity from birth records, maternal smoking status and household income from the first survey, and maternal and paternal education from the second survey. Paternal age was categorized in the same way as maternal age categories. Smoking status was dichotomized as smokers versus non-smokers. We aggregated educational attainment into three levels: completed high school or less, completed two years of college or vocational school, and completed four years of college or more. At Wave 1, questions about maternal employment status one year prior to delivery included seven options ranging from unemployed to full-time employment. We dichotomized these responses into “employed full-time” versus “not employed full-time”. Household income is the combined amount of annual earnings from the father, mother, and other sources, for which JPY ¥10,000 is equivalent to approximately USD $100. Based on gestational age, we created three groups following the WHO classification : term births (>37 weeks), moderately preterm births (32–36 weeks), and very preterm births (<32 weeks). We treated birthweight as continuous. | study | 100.0 |
We conducted logistic regression analyses to evaluate the relationship between maternal age and child health. We first estimated the crude odds ratios (ORs) and 95% confidence intervals (95% CIs) linking the maternal age groups with each health outcome [Unadjusted Model]. We re-calculated the ORs and 95% CIs controlling for potential confounders [Adjusted Model]. In the follow-up analyses, we added preterm births and birthweight to the list of factors included in the adjusted Model [Mediator-adjusted Model]. We also tested whether the associations between maternal age and child health outcomes were linear. In the follow-up analyses, we divided the sample into parity 1 and parity 2 and analyzed the sub-samples using the same models we used in the main analyses. The youngest maternal age group (i.e., <25) served as the reference group, and we conducted our analyses with complete cases. All statistical analyses were performed with Stata (Version 13/SE, STATA Corp., Texas, USA). | study | 100.0 |
Average maternal age was 29.9 years in the 2001 cohort (n = 47,015) and 31.4 years in the 2010 cohort (n = 38,554), which reflects the secular trend toward later childbirth. In Table 1 and S1 Table, we present the distributions of background information in the 2001 and 2010 cohorts by maternal age groups: 25.0–29.9, 30.0–34.9, 35.0–39.9, and >40.0 years. As expected, preterm births and low birthweight were more common among the advanced maternal age groups in both cohorts. The proportions of preterm birth and low birthweight were slightly higher in the <25.0 years group compared to the 25.0–29.9 group in both cohorts, which is likely because of disadvantaged socioeconomic conditions . However, government programs such as national health care coverage to support pregnant women may have been protective against adverse birth outcomes among the youngest mothers in Japan. We also observed that educational attainment and family income tended to be higher among older age groups, but full-time employment was more common among younger mothers. This reflects the pattern of labor force participation among Japanese women, who typically leave work once they begin childbearing . As shown in Figs 1 and 2, we observed some drop-out between the first and subsequent surveys. Those who were lost to follow-up tended to be younger in maternal age with low parental education, maternal smoking habits, and lower household income. | study | 100.0 |
a 14 missing cases for birthweight, 36 missing cases for preterm birth, 298 missing cases for maternal educational attainment, 792 missing cases for paternal educational attainment, 442 missing cases for maternal employment status, 278 missing cases for maternal smoking status, 612 missing cases for paternal age, and 3194 missing cases for family income. | study | 98.9 |
In Table 2, we present unadjusted and adjusted ORs with 95% CIs for child health outcomes in the 2001 and 2010 cohorts. When children were at the age of 18 months, we observed a linear relationship between older maternal age and lower risks of both unintentional injuries and hospital admission. Even after controlling for potential confounders such as educational attainment of parents, the favorable outcomes remained statistically significant among older maternal age groups compared to the youngest maternal age group. For example, compared to mothers <25 years, ORs of hospital admission were 0.76 [95% CI: 0.65, 0.90] for mothers aged 35.0–39.9 and 0.71 [0.51, 0.98] for >40.0 years, respectively. We statistically confirmed a linear trend by maternal age: ptrend<0.01 for both unintentional injuries and hospital admission. Lower risks among older mothers and the linear relationship between maternal age and child health remained after adjusting for preterm births and birthweight [the Mediator-adjusted Model in the table]. When children were at the age of 66 months, the risks of hospital admission were lower among older maternal age groups. However, the associations attenuated toward null once we controlled for potential confounders. | study | 99.94 |
In the 2010 cohort, also presented in Table 2, we observed the same trend of better child health at the age of 18 months according to older maternal age as we observed in the 2001 cohort: ptrend<0.01 for unintentional injuries and ptrend = 0.05 for hospital admission. Although the lower risk for hospital admission among older maternal age groups was not statistically significant, the odds ratios were highly consistent between the 2001 and 2010 cohorts. We did not observe much change in the odds ratios after additionally accounting for preterm births and birthweight. In the follow-up analyses, we stratified the sample by maternal parity. Examining the odds ratios obtained with the sub-samples of parity 1 and parity 2, we did not see much deviation from the original analyses in the magnitude and direction of associations. We present the results in S2 Table. | study | 100.0 |
In Japan, gendered division of labor has been the norm. Therefore, it is predominantly the mothers who take care of their infants. We confirmed this trend in our survey with the question: “who is the child’s main caregiver during the day?”. In the first survey when the child was aged 6 months, we confirmed that 91% of the 2001 cohort and 95% of the 2010 cohort identified ‘mother’ as the main caregiver. We reanalyzed the data with the restricted sample, but we did not see differences from the original analyses. | study | 99.94 |
This is the first population-based study to examine the effect of maternal age on early child health in Asian countries, where the rise in average maternal age at birth has become a serious concern . We found decreased risks of unintentional injuries and hospital admission at 18 months’ follow-up according to older maternal age in both the 2001 and 2010 cohorts. These trends seemed to be robust regarding the adjustment of detailed socioeconomic circumstances and potentially mediating factors such as preterm births and birthweight. We did not observe marked differences in hospital admission at 66 months once we controlled for biological and socioeconomic factors in the 2001 cohort. A key strength of our study is the use of a nationally representative sample of Japanese children from two large birth cohorts. Results were consistent between the two birth cohorts despite the 9-year gap, strengthening the validity of our findings. | study | 99.94 |
Previous studies have revealed that advanced maternal age is a major risk factor for various adverse obstetric and perinatal outcomes [4–6, 21, 22]. Indeed, we observed higher rates of preterm births and low birthweight among older mothers in LSB21. Our previous study, as well as other studies, has shown that preterm birth is a risk factor for unfavorable health outcomes during subsequent years [12, 23]. However, from a socioeconomic perspective, advanced maternal age may also exert a positive impact on children’s health. Japanese women are no exception to the worldwide pattern in which delayed childbearing is associated with higher educational attainment and career investment . In LSB21, women of advanced maternal age tended to have higher educational attainment and higher family income. Evidence suggests that higher socioeconomic status (e.g., higher income and educational attainment) is beneficial for child health [25, 26]. Consequently, social advantage associated with older maternal age may compensate for biological disadvantage . Conversely, when socioeconomic disadvantage threatens maternal “fitness”, there may be an advantage to commencing childbearing at an earlier age—the so-called “weathering hypothesis” proposed by Geronimus . | study | 99.94 |
Motherhood readiness linked with age may explain our findings regarding unintentional injuries . Older mothers may be more proactive concerning their children’s exposure to environmental risks of injury. They may also have higher health literacy in managing child illness, thereby avoiding hospitalization compared to younger mothers. | study | 99.5 |
Our results aligned with those from the UK study by Sutcliffe et al., which showed that advanced maternal age was associated with fewer hospital admissions at 9 months and 3 years and unintentional injuries at 9 months, 3 years, and 5 years . No statistical differences in the risk of hospital admission at 66 months among maternal age groups in our study also corroborated the null association at 5 years in the UK study. Null findings at age 5 in the present study and the UK study may suggest that the influence of other factors such as socioeconomic circumstances intensify with age. The Canadian study showed a similar trend, though the results were not statistically significant . However, the sample size was much smaller (n = 3,382) compared to our study and that of Sutcliffe et al. Although social context varies greatly among Japan, the UK, and Canada, results from all three population-based studies seemed to be in agreement. | study | 99.94 |
This study has several limitations. Firstly, inadequate statistical adjustment (i.e., over- or under-adjustment) is always possible. Bias resulting from unmeasured or residual confounding may be present, especially given the complex set of social and biological factors that are correlated with maternal childbearing age. Loss to follow-up occurred in the later waves, particularly among women with a low household income, no job, smoking habits, and women of young maternal age. As these characteristics are often considered as risk factors for child health, the observed prevalence of outcomes within younger maternal age categories may have been underestimated. Recall bias from parental reporting of unintentional injuries and hospitalization admission during the past 12 months seems unlikely because the assessment is rather objective in nature. | study | 99.94 |
Our results were comparable with those from UK and Canadian populations that suggested associations of older maternal age with improved child health outcomes in early childhood. Although delayed childbearing is an ongoing trend in East Asian countries and globally, its long-term consequences remain poorly understood. Therefore, our study adds to this scarce body of knowledge. However, more studies are needed to validate our findings in different cultural contexts and also explore the effect of older maternal and paternal age on other aspects, such as children’s cognitive and social development. | study | 99.94 |
Non-small cell lung cancer (NSCLC), a leading neoplastic disease with the highest incidence rate and cancer-related mortality worldwide, has a 5-year survival rate under 10% due to high metastasis rate and poor prognosis. Resection has been the main therapy by far . | review | 95.7 |
Targeted therapy for oncogene has been an emerging therapeutic method in recent years, whereas patient survival rate has not been improved due to late detection and therapy. Thus it was essential to find cancer marker guidance of early discovery, diagnosis and treatment, which led to biomarker becoming one of the most valuable area in cancer research. NSCLC is hardly characterized by one marker for heterogeneous disorder with a variety of biomarker , making it hard to find effective NSCLC marker. | review | 99.9 |
With rapid development of high-throughput biotechnology, biological data have been obtained from protein complex and gene expression profile, which helped to clarifying gene function and resource acquisition. Some scholars observed differentially expressed genes with microarray between Lung adenocarcinoma and adjacent normal tissue and established protein-protein interaction (PPI) network with gene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) for enhancement of lung adenocarcinoma knowledge . PPI search tool of network construction and interaction gene STRING database provided mutual information for experiment and prediction. To get gene-related information, studying topological features of disease gene product in PPI network has become a new way of gene predicting. | review | 99.06 |
Closely related core genes could be found by means of gene interaction network starting with known NSCLC gene. PARP1 had a high connectivity, namely as a basic core, and directly attracted with typical oncogene EGFR and ALK; mutating EGFR in NSCLC constitutively activated encoding protein and caused the occurrence of abnormal function and constant signal transduction out of control, which was in correction with NSCLC development ; as a representative molecularly targeted agent, ALK inhibitor could activate recombination through mutation, gene amplification and chromosome abnormality to increase carcinogenic driver expression . | study | 100.0 |
Furthermore, TCGA indicated that PARP1 was highly expressed in NSCLC, suggesting that PARP1 had a great influence on NSCLC. Coded protein in eukaryotic cells of PARP1 catalyzes poly-ADP-ribosylation and is involved in the process of DNA single-strand damage repair. The main achievement of PARP1 in cancer is stimulation for proliferation of colorectal cancerous cells, whereas there is no research on PARP1 functional mechanism in NSCLC. | study | 100.0 |
The study firstly surveyed PARP1's effect on NSCLC cells and discussed its functionary mechanism, which found that PARP1 was expressed strongly in metastatic NSCLC; and high PARP1 expression was relevant to prognosis and facilitated NSCLC cells migration and invasion, which provided not only corresponding foundation of experiment medicine for later targeted treatment studies but theory support of PARP1 expression within NSCLC and possible mechanism research. | study | 99.94 |
OMIM (Online Mendelian Inheritance in Man) is a comprehensive database involved in human gene and genetic disease . There were 3358 annotation of genes related to mutation phenotype of Mendelian inherited disease by March, 16th, 2015. Seven genes were identified from the search with the key word “non-small cell lung cancer”. | other | 96.6 |
The database COSMIC (Catalogue of Somatic Mutations in Cancer) embodied cancer-related gene extracted in high-throughput experimental data which was from the reported papers and cancer genome plan of Sanger laboratory. 19 genes were identified from the search with the key word “non-small cell lung cancer” in this database. | study | 91.6 |
GAD (The Genetic Association Database) is a comprehensive database collecting complex disease of human. The reported literature and pathogenic gene of complex disease was annotated in GWAS experimental data. 12 genes were identified from the search with the key word “non-small cell lung cancer” in this database. | study | 99.94 |
The database DisGeNET annotate pathogenic gene by integrating common database and gene-disease relationship in reported literature. Currently 381,056 gene-disease relationships included in this database (including 16,666 genes and 13,172 kinds of diseases). There are 39 genes were identified from the search with the key word “non-small cell lung cancer” in this database. In summary, 62 genes were identified through searching database in total (no repetition). The specific gene name and database source were included in the Supplementary Table 1. | study | 100.0 |
Though many databases were developed to collect pathogenic gene, there were some differences in the data with disease phenotype covered and various orientation of genotype due to the data types of different database. Meanwhile, the collection of information from above database might be incomplete considering the lag characteristic in the maintaining process of database. | other | 99.8 |
Thus, there was a need for pathogenic gene through artificial screening to further analyze in terms of specific disease phenotype. Detection of the relationship between gene and non-small cell lung cancer conducted in Pubmed, gene name and non-small cell lung cancer as the searched key words, that is (“gene symbol” or “gene”) and “non-small cell lung cancer”. The literature would be recorded if the two key words appeared both in title and abstract of the same literature, as the evidence of detection of the association between gene and non-small cell lung cancer. Finally, 81 pathogenic genes were gained from the literature, specific gene name included in the Supplementary Table 1. | study | 99.94 |
A non-directional network chart with the PPI data gained from STRING was established, and the connection of their shortest path through Dijkstra's algorithm was calculated (Figure 1). Then the Betweenness value of internal node in these paths was counted. 385 genes whose Betweenness value over 0 were gained, and these genes were listed out in the Supplementary Table 2. Permutation test was employed to check and calculate their permutation FDR in terms of further selection of these genes, similarly the results showed in the Supplementary Table 2. | study | 100.0 |
Among them, permutation FDR of 103 genes were less than 0.05. There was a strong correlation between these genes and NSCLC, which could be used in the subsequent analysis. A gene PARP1 with high betweenness value except for the famous oncogene with high connected degree was found, which indicated it played at a relatively core role in NSCLC gene network, yet it was unknown of the role of this gene in lung cancer. | study | 100.0 |
There was direct gene connections between PARP1 and the famous NSCLC genes, EGFR and ALK. Thus it was possible that PARP1 involved in the occurrence and development of NSCLC. Therefore, functional experiment was further adopted to detect the effect of PARP1 on NSCLC. | study | 100.0 |
We explored the expression of PARP1 in blood samples and tissue samples of NSCLC which showed a high consistency in its expression of blood and tissue samples (Figure 4A). It was indicated that the expression of PARP1 in blood could instruct the expression in tissue. | study | 100.0 |
Q-PCR was used to detect the PARP1 expression of tissue samples of 75 NSCLC patients. (A) The PARP1 expression of Pearson correlation scatter diagram in the matching tissue of 75 NSCLC patients and serum samples. A high consistency was found in its expression of blood and tissue samples. (B) The expression of PARP1 in metastatic NSCLC was higher than that in the non-metastatic NSCLC tissue. The results were analyzed using t-test. *p < 0.05; **p < 0.01; ***p < 0.001. Red, metastatic group; blue, non-metastatic group. | study | 100.0 |
A549, H1975 and PC9 were chosen to represent the typical NSCLC cell line in this study. Function analysis was conducted in vitro cell line to study the effects of PARP1 on NSCLC invasion by adopting overexpression and inhibition of PARP1 expression. The overexpression and silence cell line of PARP1 gene with NSCLC cell line A549, H1975 and PC9 was established, then the PARP1 expression was detected in the 3 cell lines. | study | 100.0 |
Higher expression of PARP1 was showed in NSCLC patients and metastatic NSCLC patients in the above study, which indicated PARP1 involved in the metastatic process of NSCLC. Thus, PARP1 was necessary for the cell proliferation, migration and invasion during the metastatic process. | study | 100.0 |
Thus, firstly the effects of PARP1 on the cell proliferation of A549, H1975 and PC9 were tested through MTT assays. The result indicated that PARP1 gene silence could significantly inhibit the cell growth of A549, H1975 and PC9, while PARP1 overexpression promoted the growth of A549 cell significantly (Figure 6). Secondly, cell scratch test would be conducted to identify whether PARP1 could promote the metastatic capacity of NSCLC. The result showed that PARP1 silence inhibited cell migration significantly shown in cell scratch test, while PARP1 overexpression promoted migration of NSCLC cell significantly (Figure 7). Finally, trans-well assays (Figure 8) was adopted to confirm these results, with the same conclusion drawn. | study | 100.0 |
2 × 104/ml cell suspension fluid was inoculated in 96-well plates, and cultured for 96 h, respectively. Following that, the plates were added with MTT solution, and incubated for 4 h, then added with DMSO to fully resolve MTT pyrolysis products. Optical density (OD) was measured by immunoassay analyzer at the wavelength of 570 nm. (A, B) Effect of PARP1 on the growth of A549; (C, D) effect of PARP1 on the growth of H1975; (E, F) effect of PARP1 on the growth of PC9. | study | 100.0 |
For the scratch test, cells were plated at 2×105 cells/well in a 6-well plate and grown overnight under standard conditions. After 12 h, the confluent monolayer was scratched manually with a plastic 200 μl pipette tip and after washing with PBS. Then the culture was continued in 2% serum at 37°C for 24 h. The wound area were imaged using an inverted microscope at 100× magnification. The distance was measured using ImageJ2x. (A) Images of wounds of A549 cells transfected with the indicated plasmids taken 0, 12 and 24 h afer the wounds were inflicted (× 100). (B) Overexpressing PARP1 promoted the wound-healing process in A549 cells, whereas PARP1 silencing did not contribute to wound healing. Wound-healing results for A549 cells transfected with the indicated plasmids (analyzed by t-tests). *p < 0.05; **p < 0.01;***p < 0.001. | study | 100.0 |
1 × 105 cells were plated onto 24-well cell chambers (pore size: 8 μm). The cells were plated in medium without serum, and medium with 20% serum was used as a chemoattractant. The cells were incubated at 37°C for 16 h. Cells in 10 random fields of view at 100× magnification were counted. Images were acquired using an inverted phase-contrast microscope at 100× magnification. (A) Images from the Trans-well migration assays of A549, H1975 and PC9 cells transfected with indicated plasmids (× 100). (B) Statistical results for the trans-well migration assays of A549, H1975 and PC9 cells transfected with indicated plasmids (analyzed by t-tests); %migration = [mean number of cells invading the membrane/mean number of cells migrating through the control insert membrane] × 100. *p < 0.05; **p < 0.01; ***p < 0.001. | study | 100.0 |
The survival relationship between PARP1 and patients of NSCLC was analyzed with the data in GEO database. Kaplan-meier survival curve was made using the online survival analysis tool, PROGgene. It was found that the expression of PARP1 was related to the prognosis or recurrence of patients through Kaplan-meier survival curve: The overall survival (OS) in PARP1 high expression group was significantly lower than that in PARP1 low expression group (Figure 9). | study | 100.0 |
Complex diseases were generally affected by the interaction of multiple genes . Generally, the common influence of interaction in interacting genes could not only decrease the complex degree of biological network, but also benefit to explore meaningful biological information for the researchers, which further provide scientific basis for therapy and study of disease. Previous studies indicated that many interacting genes had similar or same effects, or involved in the same biochemical pathways. Therefore, the known key gene in an established interaction network of genes would be found easily, and the functions of interacting genes can be preliminary predicted and annotated based on this gene. In this study, NSCLC-related gene was searched from database OMIM, COSMIC, GAD and DisGeNET, then a network chart was established with the adoption of PPI data from STRING, followed by the shortest path analysis and Permutation test. There is a total of 385 genes whose Betweenness value was greater than 0 gained through Dijkstra's algorithm. Permutation test showed 103 genes whose Permutation FDR was less than 0.05. There was a strong correlation between these genes and NSCLC, which could be used in the subsequent analysis. | study | 99.94 |
Almost all the NSCLC patients would recurred in different stages, and finally suffered from disease metastasis and death . Even though there was a new development of immunotherapy and targeted therapy of tumors, once the systematic metastasis of cancer cell occurred, the five-year survival rate would decrease at lessby 10% . For that it was still a challenge to expand the survival rate of NSCLC patients by inhibiting the cell migration of cancer cells. Therefore, it was of significance for the study of molecular markers and potential therapeutic small molecules in clinical stage. PARP1 played an important role in DNA damage repair, especially in base excision repair (BER) pathway . It was further verified that PARP1 played a key role in DNA repair [15–17], which made PARP1 a valuable therapeutic target in the research of disease drug. Recently, it was found that PARP1 presented high expression in a series of cancers, which provided a high-value target for the study of tumor detection, stage and biological characteristics [18–23]. The overexpression of PARP1 results from much DNA damage in cancer cells with unstable genes, instead of the induction of the activated specific carcinogenic pathway . Byers, et al. demonstrated that PARP1 could be a new target molecule for NSCLC by combined transcriptome and proteome analysis . Meanwhile, a new study claimed that PARP1 could promote the metastasis and recurrence of lung adenocarcinoma towards brain and bones by regulating several steps in the metastasis process, which was not related to DNA repair. There were several ways of PARP1 enhancing the metastasis of lung adenocarcinoma towards brain: promoting the invasion of lung adenocarcinoma cells, anoikis resistance, exosmosis, and self-updating, as well as regulating cerebral microenvironment . | review | 99.8 |
In the current study, the expression of PARP1 in the blood samples showed a high consistency with that in the tissue samples, which indicated that the expression of PARP1 in blood could instruct the expression in tissue. Because of that, the related information about the development of tumor could be obtained through the detection of PARP1 expression in blood at the early stage of NSCLC tumor. The results of MTT assays indicated that lockdown of PARP1 gene significantly inhibited the cell growth of A549, H1975 and PC9, while PARP1 overexpression promoted the growth of A431 cell significantly. It indicated that PARP1 gene promoted the proliferation of NSCLC, whose role was consistent in others malignant tumor [27, 28]. There were many extensive researches on the gene which influence NSCLC cell migration, mainly including MALAT-1 , MTA1 , BRAF , EGFR and so on. However, the role of PARP1 in NSCLC migration remains still unknown. Our experiments revealed that: expression of PARP1 was upregulated in NSCLC and showed higher expression in metastatic NSCLC; PARP1 silence inhibited cell migration significantly, while PARP1 overexpression promoted NSCLC cell migration significantly . It has demonstrated that knockout or inhibition of PARP1 could decrease the migration of lung adenocarcinoma significantly, which indicated PARP1 was involved in the migration of NSCLC. In addition, there was a direct association between PARP1 and the key cancer gene of NSCLC: EGFR, ALK in the interaction network of genes. Previous studies have suggested that EGFR takes part in DNA repair through PI3K signaling pathway [32–33]. In the cells with ALK fusion, it has illustrated that dimers can be activated abnormally by aberrant receptors, which can activate the downstream signaling pathway PI3K-AKT and drive aberrant proliferation of tumor cells [34–35]. Based on these findings, it can be see that ALK and EGFR after requiring functional mutation could drive the growth of tumor cells via PI3K-AKT pathway. So we hypothesized that the effects of PARP1 on NSCLC tumor migration may be associated with the PI3K-AKT pathway. However, the mechanism needs further exploration. | study | 100.0 |
It is of directive significance to understand the effect of prognostic factors on the prediction, diagnosis, and treatment of diseases. According to the study, PARP1 may be an independent prognostic factor of multiple malignant tumors. And PARP1 Val762Ala polymorphism may be an independent prognostic factor for cervical cancer, which might play a role in the prediction of clinical results . Collectively, the expression of PARP1 might be considered as an independent prognostic factor for the patients of breast cancer . While in the limited stage of SCLC, the expression of PARP1 was related to the progression free survival of tumor cells . The COX regression analysis conducted by KJ Xie, et al. indicated that PARP1 was an independent adverse prognostic factor of NSCLC patients. According to Kaplan-meier survival curve, the higher expression of PARP1, the lower survival rate of patients, which preliminary suggested that PARP1 might be an independent adverse prognostic factor of NSCLC. Consistently with the above result we found that PARP1 presented high expression in NSCLC, and the survival rate of patients was decreased accompanied by increased PARP1 expression. In summary, based on gene interaction network, the shortest paths method was adopted in NSCLC gene network to find a number of candidate oncogenes of NSCLC. We conducted the functional verification on PARP1, which provided convenience for the following discussion and study of the mechanisms related to the occurrence and development of NSCLC. We found the role of PARP1 in the migration and prognosis of NSCLC, laying foundations for the following clinical experiments, and providing theoretical basis for revealing the effects of PARP1 on NSCLC. | study | 99.94 |
Data collection: taking “non-small cell lung cancer” as key word to retrieve and gain NSCLC-related gene from database of OMIM (Online Mendelian Inheritance in Man) (Rashbass, 1995), COSMIC (Catalogue of Somatic Mutations in Cancer) (Forbes et al., 2011) and GAD (The Genetic Association Database) (Becker, Barnes, Bright, & Wang, 2004). | study | 61.38 |
We could make sure the shortest path of each pair of correlatively known gene of lung cancer and genes on the path by Dijkstra's algorithm: calculating optimal path of one node to all the other nodes, moreover, arraying these genes by Betweenness value that stood for internal nodes of certain gene contained all optimal paths found. | study | 99.9 |
No matter what kind of gene was included in our gene network and how the real node or gene was distributed, it was always easy to find the genes with high Betweenness value for some Betweenness value might sway by network basically structural models, yet these universal genes was not our wanted genes found from special genetic network. So we adopted permutation test to further screen NSCLC genes with optimal path in specific net. | study | 100.0 |
From gene network, we randomly selected simulated genes involved with a number of NSCLC-associated genes and then calculated the shortest path within genes (a total of 500 simulations); after we counted once, the Betweenness value in stimulation was bigger than that of actual one. After 500 randomized trials, each gene would get a frequency that was defined as permutation FDR of shortest-path gene of which small value (FDR < 0.05) meant a significantly NSCLC-involved gene of the shortest path. | study | 100.0 |
Total RNA from 8 cell lines, blood samples, 75 cancer tissues of NSCLC and the adjacent normal tissues were extracted in TRIZOL RNA extraction reagent (Invitrogen) by manual. The experiment was performed as follows: the tissues were added with 500 ul Trizol and 200 ul chloroform in order, after vibrating and standing, then were placed into centrifugal machine for 10 min, mixing with isopropanol and ethanol to discard supernatant by centrifugation; following that, the tissues were added with 20-50 ul DEPC-ddH2O to dissolving dissolve RNA using 20-50ul DEPC-ddH2O and preserved saving with absolute ethanol in refrigerator at −70°C. | study | 99.94 |
SuperScript III First-Strand Synthesis System kit (Invitrogen) was used to synthesize cDNA, with the detailed processes shown in brochure. The fluorescent primers were designed using primer premier 5.0 according to the sequence of genes. Real-time PCR measurements were performed on CFX96 (Bio-Rad). Each PCR reaction contained SYBR® Premix Ex Taq (2x) 12.5 μl, 10 μmol/L upstream and downstream primers each 1 ul, cDNA template 2 μl, dH2O 10μl. The temperature process was 95°C for 30 sec followed by 40 cycles of amplification (95°C for 5 sec, 60°C for 30 sec, and 72°C for 30 sec). 2−ΔΔt was adopted for date processing. | other | 99.9 |
NSCLC cell lines A549, H1975 and PC9, purchased from The Shanghai Institutes for Biological Sciences of the Chinese Academy of Sciences, were taken out and inoculated to RPMI-1640 culture solution complementary with 10% fetal bovine serum, and cultured at under 37°C with a humidity of 5%CO2 saturation, with liquid exchanging and passaging every 3-4 days. | study | 99.9 |
Completed PARP1 cDNA would be acquired with primers (Table 1), and then we cloned completed PARP1 cDNA into pEGFP-N2vector to construct gene overexpression plasmid. As for the PARP1 gene silencing, shRNA and negative-control shRNA (shRNA-NC) of PRAP1 were designed according the sequence of PARP1 (NM_007415) from GenBank (Table 1). After annealing, shRNA of PRAP1 were ligated into pGPU6/GFP/Neo vector to build gene silencing recombinant plasmid. And then recombinant vectors were transfected into cell lines by liposome Lipofectamine 3000. Meanwhile, the blank vector transfection group was taken as the control group. After transfection for 2 days, there were part of green fluorescence cells which were converted to use 400 μg/ml G418 DMEM medium, screened for a month to attain gene silencing and overexpression stable transgenic lines, culture and cryopreserved in DMEM + 10% FBS + P.S. medium. | study | 100.0 |
2 × 104/ml cell suspension fluid was inoculated in 96-well plates, with each well having 3 parallel wells and about 100 μl fluid, and cultured in an incubator with 5% CO2 at 37°C to 24, 48, 72, 96 h, respectively. Following that, the plates were taken out, added with 50 μl MTT solution each well that was configured into 1 g/L with normal saline, and incubated at 37°C for 4 h to discard the supernatant fluid, then added with dimethylformamide (DMSO) (100 μl per well), shaken for 10 min to fully resolve MTT pyrolysis products. Optical density (OD) was measured by immunoassay analyzer at the wavelength of 570 nm, drawing cell growth curve. | study | 99.94 |
For the scratch test, cells were plated at 2×105 cells per well in a 6-well plate and grown overnight under standard conditions. After 12 h, the confluent monolayer was scratched manually with a plastic 200 μl pipette tip and after washing with PBS. Then the culture was continued in 2% serum at 37°C for 24 h. Cells that had migrated into the wound area were imaged using an inverted microscope at 100× magnification (Olympus, CKX41). The distance was measured using ImageJ2x. | study | 99.94 |
Trans-well experiment was carried out manually. In brief, 1 × 105 cells were plated onto 24-well cell chambers with a non-coated membrane (pore size: 8 μm, BD Biosciences). The cells were plated in medium without serum, and medium with 20% serum was used as a chemoattractant in the lower chamber. The cells were incubated at 37°C for 16 h, and cells that did not migrate through the pores were removed with a cotton swab. Cells on the lower surface of the membrane were fixed with 100% methanol for 5 min and stained with 0.1% crystal violet for 2 min (Sigma). Cells in 10 random fields of view at 100× magnification were counted and expressed as the average number of cells per field of view. Images were acquired using an inverted phase-contrast microscope at 100× magnification (Olympus, CKX41). | study | 99.94 |
All statistical analyses were performed using SPSS 21.0 statistical software. Measurement data were presented by mean ± standard deviation (AV ± SD). Unpaired Student's t-test for parametric data or Mann-Whitney rank sum test for nonparametric data were applied to analyze mean value between two groups. Chi-square tests were applied to determine count data expressed in n (%). In all cases, p < 0.05 was considered statistically significant. | other | 98.75 |
Epidemiology is the study of patterns of health-related states or events in populations. The discipline has been dominated by statistics, which has resulted in a wide and flexible range of statistical methods developed in response to the particular challenges and types of data. For example, statistical epidemiological methods have been designed to cope with discrete data, control for (but not ignore) differences between individuals and groupings of individuals and to consider time to, or between, discrete outcomes (and dependence on previous state). The statistical methods developed can usefully be applied to the analysis of behavioral data. Here, we demonstrate this by applying two popular epidemiological models, survival analysis and multistate modeling (MSM), to data on guide dog behavior. | study | 99.94 |
In the study of health, many outcomes of interest are discrete. For example, does a patient have a disease or not? do they die from a disease or not? is the patient susceptible, infected, or recovered from an infectious disease? Similarly, categorical data are common in the study of behavior. For example, does the subject learn a trained task? Does a subject fight its competitor or not? Does the subject choose option a, b, or c in a choice test? A further commonality between behavior and health outcomes is that time is often an important aspect of the outcome. An epidemiologist might study whether or not a patient died and how long this took, whereas an ethologist may study whether and animal learned a task and how many trials this took. A final similarity is a lack of heterogeneity between subjects of interest. People from the same family or household may have a shared susceptibility to a disease event and animals’ from the same litter, or held in the same pen, may have a shared susceptibility to a behavioral event. Such similarities between types of data have resulted in some exploitation of similar statistical methods in both disciplines: random effects models, which account for unobserved heterogeneity, are common in both disciplines (1–4). However, statistical methods designed to study health-based events have been underutilized in behavior. | review | 99.9 |
One common epidemiological statistical method is survival analysis. Survival analysis, as the name suggests, was developed to examine mortality data and risks associated with time until death. Death is a discrete (binary) outcome. Using survival analysis, it is possible to model: a time variable, which is the time spent in a given state or the time between two events; a survival function, which is the probability of an individual surviving beyond time t; and a hazard function, which represents the probability that an individual “alive” at time t experiences “death” in the next period, t + 1. Frailty models can be used to account for a lack of heterogeneity between subjects. Survival models also allow for censoring where the outcome is not recorded at a point in time for a particular subject. Survival models may be particularly useful for studying behavior and have to date been applied to a limited range of behavioral data including: cognitive and judgment bias tasks (5); choice/preference tests (6); latency to perform behavioral response (7, 8); and time taken for pets to be rehomed (9). | review | 99.44 |
Survival analysis, typically has a binary outcome, but MSM can be extended to incorporate more than two discrete events or states. The archetypical use of MSM is to understand transitions, and the timing of transitions, between disease states. For example, when studying cancer, epidemiologists have considered the rate at which patients move between states of treatment for cancer, remission, and death (10). The multistate model accounts for dependencies of timings of subsequent events and dependencies of timings of competing events. Like survival analysis, censoring and frailty models can be used. MSMs provide a range of possible options for modeling the dependence of the transition rates on time. Models can be: time homogeneous models, which consider transitions to be constant and independent of time, Markov models where transitions depend on current state; and semi-Markov models where transitions depend on the current state but also on the entry time into the state. To date, MSM have rarely been applied to the behavioral data, with one exception being their application to understand transitions between mobility scores in livestock (11). | review | 99.8 |
To illustrate the application of epidemiological techniques to the analysis of animal behavior data, we use three previously published datasets from Guide Dogs. Guide Dogs UK are the largest breeder and trainer of dogs in the UK and have been keeping detailed records on behavior and health of dogs they breed and train for many decades. Retrospective analysis of these data has proved useful in studying patterns of health (12) and behavior (13). In addition, in recent years, Guide Dogs has been prospectively collecting more detailed behavioral data on dogs during training to better understand how to monitor and record behavior (14–16). The majority of potential guide dogs are bred internally, but a minority is purchased from breeders. All are placed with a volunteer puppy walker for the phase of Puppy Walking (obedience, socialization, and habituation), when they are approximately 6–8 weeks of age. When dogs reach approximately 12–14 months of age, and they are deemed ready, they will begin Guide Dog training during which time they are trained to guide a blind or partially sighted individual, typically in two stages by two trainers. In the last few weeks of this training, dogs are matched with a guide dog owner and the dog and owner are trained together. Dogs, which are successful in completing training are described as qualified, but those deemed unsuitable can be withdrawn from training at any stage. Here, as an example of how epidemiological methods can be used to provide insights into animal behavior research, we apply survival analysis and MSM to Guide Dogs data, considering the factors, which are associated with a dog qualifying or not, movements between different stages of training and the timing of both of these. | study | 99.94 |
The three datasets considered in this study have all been previously analyzed using alternative (more standard) statistical methods and the results published elsewhere (13, 15, 17). In brief, previously, we presented controlled observations of 6- to 8-week-old puppy behavior, the Puppy Profiling Assessment (PPA), where three (stroking, fake prey, ramp) of seven stimuli presented (also including following, retrieve, gentle restraint, noise, tunnel) could be usefully combined and associated with whether dogs completed guide dogs training to become a working guide dog (success) (17). Reactions to these stimuli could be applied to identify a small number of dogs that were later withdrawn from guide dog training with high specificity but low sensitivity, and a subsequent study suggested responses to stimuli were heritable (18). Using a questionnaire approach, completed by Guide Dog staff for trainee guide dogs of 5, 8, and 12 months of age, seven behavior traits were identified, adaptability; body sensitivity; distractibility; excitability; general anxiety; trainability and stair anxiety, which all showed associations with later success in guide dog training (15). Together, the traits could be applied to early identify a small proportion of dogs with high specificity that were later withdrawn and which were successful. A final relevant publication considered dogs withdrawn from service as a guide dog for behavior reasons based on retrospective data (13). We found that there was no peak in working life in which dogs were withdrawn for behavior, and sex and breed were important factors in age of behavioral withdrawal. In all cases, the statistical methods used to explore the association with later success in guide dog training or withdrawal from training or working were based on variants of logistic regression models. Here, we extend analysis from these three studies to include the entire span of a guide dog from training to retirement for the retrospective dataset [hence overlapping but not identical data between this paper and (13)]. The aim is to illustrate the utility of survival analysis and multistate modeling and to explore the additional insights these methods provide beyond more traditional methods of analysis. | study | 99.94 |
We used three datasets for this analysis (see Table 1): (1) retrospective data spanning 10 years (dogs born 2000–2011), (2) data collected on behavior of 6- to 8-week-old puppies in a controlled behavior test named the PPA (dogs born between March 2006 and February 2007), and (3) data collected using a validated questionnaire on dog behavior during puppy walking (dogs born between 6/12/2011 and 1/1/2013). | study | 100.0 |
In the controlled behavior test, puppies were presented with a series of controlled stimuli and their reactions were scored by a member of Guide Dogs staff on a seven point scale according to behavioral descriptors. These tests took place from May to October 2012. The later outcome of dogs (withdrawal or qualification and movements between stages) was accessed from Guide Dogs records in October 2014, but these decisions were made independently of scores on the PPA. Scores of the PPA have previously been found to be associated with qualification as a guide dog, supporting the predictive criterion validity of the test (17). Some scores on the PPA are also heritable (18) and Guide Dogs use this test routinely to record behavior of puppies. In the PPA, puppies are scored according to their response or recovery to stimuli and the human assessor. Full details can be found in Asher et al. (17), but in brief, the stimuli the puppies encounter are: “Following” where they are encouraged to follow a person, a toy, which they are encouraged to “Retrieve,” a brief and gentle “Restraint,” playback of aircraft “Noise,” “Stroking” by a person, a small furry object moved fast on a string to mimic a “Squirrel”-like prey, a “Tunnel,” which they are encouraged to move through, and a “Ramp,” which they are encouraged to walk over. | study | 100.0 |
Five of these stimuli are scored according to a dog’s immediate reaction only (following, restraint, noise, tunnel, ramp), and three are scored according to a dog’s immediate reaction and subsequent recovery reaction (retrieve, stroking, squirrel). This results in a total of 11 scores, each scored from 1 to 7. The dogs tested came from 11 breeds or crossbreeds of Golden Retrievers, Labradors, German Shepherd dogs, and Flat coat retrievers. The most common breeds were: Labradors (260) and Golden retriever × Labradors (205). The majority of dogs were bred by Guide Dogs (791/801). | study | 99.94 |
In the questionnaires, Guide Dogs staff responsible for supervising a dog’s progress during the Puppy Walking stage of training (Puppy Training Supervisors) were asked to complete a series of questions about the dog’s behavior scored on a visual analog scale (15). Data used here were collected when the dogs were 5 months of age (plus or minus 1 week) for all dogs in Guide Dogs Puppy Walking scheme from 6/12/2011 to 1/1/2013 (n = 1,401). Information on the later outcome (withdrawal or qualification and movements between stages) was retrieved from Guide Dogs records in October 2014. Outcome and movement decisions were made independently of questionnaire scores, which were not provided to Guide Dogs staff. The questions were grouped into traits and the means of these questions formed a score for each trait. This provided trait scores with the following names: distractibility, excitability, trainability, general anxiety, adaptability, body sensitivity and stair anxiety. Trialed alongside the questionnaire were postural images of dogs appearing “Neutral,” scored in five different situations. Puppy Training Supervisors were asked to rate how often the dog appeared to show body language illustrated in the neutral image using a visual analog scale from Never to Almost Always (see Figure 1) for each of the five different situations (when…it encounters another dog, …stranger(s) advance directly toward to dog, …a child/children advance directly toward the dog, …unfamiliar visitors approach the car or its kennel, …approached in a confined space) and a mean was calculated across these. | study | 100.0 |
The questionnaire was developed using a combination of etic and emic approaches. It was tested to confirm internal reliability construct validity, concurrent criterion validity (using behavior observed in a separate controlled behavior test), rank-order consistency over time (to confirm that trait scores represented temporally stable traits), and was associated with qualification as a guide dog, supporting predictive criterion validity (15). Dogs were from ten breeds and crossbreeds of Golden Retrievers, Labradors, German Shepherd dogs, and Flat coat retrievers. The most common breeds were: Golden Retriever × Labrador (474), Labrador (407), Golden Retriever (131), Labrador × Golden Retriever (132). The majority were bred by Guide Dogs (1,362/1,402). | study | 100.0 |
Survival analysis was completed using a Cox proportional hazard model implemented in the R package “survival.” A frailty term for Dam included in all models and a frailty term of assessor (ID of the person who completed the questionnaire) was included in the behavior questionnaire models. For these retrospective data, we used the age at which dogs were withdrawn permanently from training or working for behavior reasons as the outcome variable. Dogs that qualified could reach retirement and be retired at approximately 8–10 years of age or be withdrawn prior to this for behavior, health, or guide dog owner reasons. For the purposes of this, dataset dogs withdrawn for health or guide dog owner reasons were excluded from analysis. The survival curve of dogs withdrawn for behavioral reasons was plotted against dogs that retired or were censored (included in time series until the point when data were not available). The effects of breed, sex, and whether or not the dog was bred by Guide Dogs (see Table 1) were then considered on this outcome variable in univariate Cox proportional hazards models. Finally dogs’ time of movement from training to qualification was considered in comparison to dogs withdrawn for behavioral reasons using survival curves, for illustrative purposes. For data from the controlled behavior test and the behavioral questionnaire, we considered the association between the behavioral measures taken using each of these methods and subsequent withdrawal from training for behavior reasons (see Table 1). A covariate of age was applied if needed based on the assumption of proportional hazards, which was tested using the function cox.zph. For each survival analysis completed, significant results are presented with hazard ratios (HRs), 95% confidence intervals (CIs), p-values (p), and median survival times for the time taken to withdrawal. | study | 100.0 |
Multistate models were used to analyze transitions between states of training and working assuming next the state was dependent on time in current state (Semi-Markov model). The models were implemented in the “Epi” and “splines” packages in R and included frailty terms for the Dam. For the retrospective data effects of sex, breed, and Guide Dogs bred on transitions between dogs in training, qualified, and withdrawn for behavior reasons were modeled. Qualified dogs were those working with a guide dog owner. Dogs, which were withdrawn for health or guide dog owner reasons were censored prior to the withdrawals. For the controlled behavior, PPA test and the behavior questionnaire data, transitions between Puppy Walking, Training, and Withdrawal were considered. The effects of the 11 PPA scores, and seven trait scores and postural image scores of the neutral posture score from the questionnaire were considered in two separate MSMs. The number of transitions between states and total dog years spent in each state are presented. Significant effects of predictors are presented with odds ratios, CIs, and p-values presented for effects of the probability of movement between states and medians, IQR, and p-values presented for the time taken to move between states. | study | 100.0 |
Using retrospective data, there were 4,429 withdrawals for behavior during training or working. Of these withdrawals, 2,085 were females, 2,344 were males; 3,695 were bred by Guide Dogs and 734 were bred externally; and 1,055 were Golden retriever × Labradors, 1,591 were Labradors and 235 were (sire × dam) Labrador × Golden retrievers. For dogs withdrawn for behavioral reasons from training or working life, the most common time period for this withdrawal was during training (the first 2 years of life, see Figure 2). Following an initially steep survival curve during the first 2 years of life, a steady slope was found with dogs withdrawn for behavior at approximately an equal rate from 2+ until the age of 8.5 years of age, when many dogs were retired from working life. Focusing on the training period, most dogs withdrawn for behavioral reasons were withdrawn between 1 and 1.5 years of age (Figure 3), whereas dogs, which moved out of training for other reasons, such as qualifying as a guide dog did so at 1.5–2 years of age. The time prior to withdrawal was influenced by sex, breed, and whether dogs were internally bred. Labradors were withdrawn faster (by a median of 2 weeks, HR: 1.09 CI: 1.01–1.18, p = 0.026) and Labrador × Golden Retrievers slower (by a median of 2.5 weeks) than Golden Retriever × Labradors (by a median of 1.5 weeks, HR: 0.83, CI: 0.72–0.97, p = 0.014). The R squared of this model was 0.05 (where R squared indicates the proportion of the variance in the dependent variable that is attributable to the variables in the model). Dogs not bred by Guide Dogs were withdrawn faster than those bred by Guide Dogs (3.1 weeks, HR: 1.13 CI: 1.04–1.23, p = 0.002). The R squared was 0.02. Males were withdrawn faster than females by 4.2 weeks (HR: 1.26 CI: 1.19–1.34, p < 0.001). The R squared was 0.013. | study | 100.0 |
From the controlled behavior test set, 289 dogs were withdrawn from training. One of the 11 scores of behavior from this test of puppies at 6–8 weeks of age was associated with probability of withdrawal from training for behavioral reasons. Puppies that were scored higher on as recovering after exposure to the Squirrel stimuli were less likely to be withdrawn (HR: 0.81, CI: 0.69–0.94, p = 0.0047). However, dogs that scored higher on this element of the test were withdrawn a median of 1.24 weeks later. The frailty term of Dam was important in this model (chi-squared = 543, df = 76.9, p < 0.001). In this case, the inclusion or not of the frailty term had little influence, either on the estimate or the significance of the fixed effects. | study | 100.0 |
From the behavior questionnaire dataset, 548 dogs were withdrawn from training. Dogs with higher scores of excitability were withdrawn faster by a median of 2.15 weeks and were more likely to be withdrawn (HR = 1.71, CI: 1.05–2.78, p = 0.03). Dogs scored as neutral in body posture more often were less likely to withdrawn (HR = 0.58, CI: 0.38–0.88, p = 0.009). Assessor was not needed as a frailty term and was, therefore, removed from the model. The frailty term of Dam was important in the model (chi-squared = 2,555, df = 174.9, p < 0.001). In this case, the inclusion of the frailty term changed the model such that excitability and neutral would not have been significant without the frailty term. | study | 100.0 |
Using retrospective data, it appeared that the majority of dogs (6,529/10,968) transitioned from training to qualification and most of these were never withdrawn from working for behavioral reasons (see Figure 4). Dogs were a median of 1.68 years when they moved from training to being qualified. Transitions between training and withdrawal for behavior reasons were associated with sex, breed, and bred by Guide Dogs. Dogs were withdrawn faster than bitches (by 5 weeks, HR: 1.16 CI: 1.10–1.23, p < 0.001). Golden retriever × Labradors were withdrawn at a median age of 1.75, which did not differ from Labrador × Golden retrievers (median 1.62). All other breeds took less time before moving to withdrawn for behavior (Labradors, median 1.44, HR: 1.45 CI: 1.34–1.56, z = 9.25, p < 0.001; Golden Retrievers median = 1.42, HR: 1.53 CI: 1.38–1.70, p < 0.001, and other breeds median 1.41 HR: 1.58, CI: 1.44–1.71, p < 0.001). Dogs bred by Guide Dogs spent longer in training before movements to withdrawal for behavior (by a median of 10 weeks, HR: 1.68, CI: 1.56–1.82, p < 0.001). Breed and sex were associated with transitions from training to qualified: bitches took longer to move from training to qualified (median for bitches 1.78 years and dogs 1.69 years, HR: 1.49, CI: 1.30–1.70, p < 0.001). The Other breed group (median age 1.79, HR: 0.63, CI: 0.52–0.76, p < 0.001) and Labrador × Golden retrievers (median age 1.79, HR: 0.63, CI: 0.50–0.85, p < 0.002), took longer to qualify than Golden retriever × Labrador (median age: 1.70). Transitions between qualified and withdrawn for behavioral reasons were associated with breed but not sex and whether dogs were bred by Guide Dogs or not. The Other breed groups were qualified for less time (median of 3.87 years) than the reference breed of Golden retriever × Labrador (median of 4.26 years, HR: 1.43, CI: 1.18–1.74). | study | 99.94 |
Multistate model of transitions between states of guide dog training and working from retrospective data. The arrows indicate the direction of movements between states and the numbers on the arrows the number of dogs moving between states. The total dog years spent in each state across the dataset is written in each state box, with the mean years per dog in brackets. | other | 99.9 |
From the controlled behavior test, most dogs transitioned from puppy walking to training (646), with 66 withdrawn before entry to training and 191 withdrawn after entry to training. Dogs, which retrieved a toy more spent less time training (HR: 1.33, CI: 1.13–1.56, p < 0.001 by a median of 5 weeks) before being withdrawn and dogs, which responded less to a person after encountering a Squirrel-like moving object were withdrawn from puppy walking at a faster rate (HR: 0.77, CI: 0.61–0.97, p = 0.025, by a median of 2 weeks from highest to lowest scores). Dogs that reacted more to the Squirrel like object (HR: 0.90, CI: 0.82–0.99, p = 0.047, by a median of 1 week from lowest to highest scores) and dogs, which retrieved a toy less (HR: 0.91, CI: 0.83–0.99, p = 0.036, by a median of 2 weeks from highest to lowest scores) were slower to move into training. | study | 100.0 |
From the behavior questionnaire, the most common transition was between puppy walking and training, with fewer dogs moving to the withdrawn state (Figure 5). Dogs spent twice as long in puppy walking than they did in training. Most withdrawals for behavior reasons occurred during training (Figure 6). Dogs scored higher on the trait Excitability transitioned to training from puppy walking faster (HR = 3.29, 1.62–4.45, p < 0.001, by a median of 0.7 weeks from high to low scores). General anxiety was associated with movements between puppy walking and withdrawal for behavioral reasons, with more anxious dogs moving at a faster rate between these states (HR: 5.78, CI: 1.28–25.62, p = 0.022, by up to 7 weeks). Dogs that were more anxious (on the General Anxiety scale) stayed longer in puppy walking (HR: 0.53, CI: 0.32–0.85, p = 0.009, by 1.4 weeks) than dogs that scored lower on these traits. | study | 100.0 |
Multistate model of transitions between states of guide dog training, from behavioral questionnaire data. The arrows indicate the direction of movements between states and the numbers on the arrows the number of dogs moving between states. The total dog years spent in each state across the dataset is written in each state box, with the mean years per dog in brackets. | other | 99.8 |
Age and date of dogs in guide dog training, with stage and exits indicated from the behavioral questionnaire sample. Each line represents one dog. Red lines show dogs in puppy walking, green lines show dogs in training, black dots show dogs exiting training for behavioral reasons, with the absence of black dots indicating continuation within Guide Dogs. | other | 99.9 |