Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B

ENS-06138; No of Pages 9 Evolution and Human Behavior xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Evolution and Human Behavior journal homepage: www.ehbonline.org

Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B Gabriel L. Schlomer ⁎, Hyun-Jin Cho Division of Educational Psychology and Methodology, University at Albany, State University of New York, United States

a r t i c l e

i n f o

Article history: Initial receipt 19 December 2016 20 June 2017 Final revision received 30 June 2017 Available online xxxx Keywords: Father absence Menarche LIN28B Life history theory Gene-environment interaction

a b s t r a c t Substantial research and theory over a number of years have linked father absence to earlier age at menarche (AAM). More recent work has centered on explaining the relative genetic and environmental contributions to this correlation. The purpose of the current study was to evaluate the combined effects of father absence and variation in the LIN28B gene on AAM. A sample of 300 women (age 18–25) successfully genotyped for two LIN28B single nucleotide polymorphisms (SNPs; rs364663 and rs314273) were used to test gene-environment interaction models. Results for both SNPs were consistent with the hypothesis that father absence would attenuate later AAM associated with LIN28B. Genetic index analysis of combined LIN28B SNPs showed that girls with at least one copy of the T/T genotype had later AAM if they were father present. Study strengths and the implications of GxE research for life history models are discussed. © 2017 Published by Elsevier Inc.

1. Introduction Age at menarche (AAM) marks a significant transition point in pubertal development as girls process through a pre-reproductive state into reproductive maturity. This transition may also come with significant negative health and psychosocial consequences however, especially among girls whose AAM is relatively early. Several studies have shown early AAM is related to greater risk for reproductive cancers (Gong, Wu, Vogtmann, Lin, & Wang, 2013), obesity, and cardiovascular disease (Prentice & Viner, 2013); early maturing girls are also at greater risk for psychosocial problems including internalizing problems, substance use/delinquency, earlier onset of sexual behaviors, and greater subjection to sexual harassment (Ellis et al., 2003; Lichty & Campbell, 2012; Mendle, Turkheimer, & Emery, 2007; Mrug et al., 2013; Skoog & Ozdemir, 2016). Discerning the etiology of AAM variation, and potential strategies for intervention, extends beyond adolescent development and into adulthood and has implications for the physical and psychological well-being of women. Because early menarche has been linked to an array of biopsychosocial issues, intrapersonal (e.g., race, height, weight) and contextual (e.g., socioeconomic status, family structure) factors that predict AAM have been the focus of considerable research. One contextual factor that has been reliably linked to AAM is biological father absence (i.e., separation or divorce of the birth parents followed by absence of the birth father from the home, Ellis, 2004). Although this extant research has demonstrated a replicable empirical phenomena, the ⁎ Corresponding author. E-mail address: [email protected] (G.L. Schlomer).

causal mechanism of this association continues to be debated. On one hand are models that presume father absence is an environmental pressure that calibrates reproductive timing. On the other hand are models that suggest the association can be attributed to a third variable confound, such as genetic inheritance. Although these areas of inquiry have provided useful insights into possible causes of AAM variation, hypotheses that position genetic and environmental influences in opposition often overlook important gene-environment transactions. In the current study we conceptualize the interplay between genes and father absence on AAM as a gene-environment interaction (GxE) using a molecular genetic approach. We begin by outlining prior theory and research on the father absence-AAM association, discuss genetically informed research aimed at teasing apart genetic and environmental contributions, and review recent evidence from genome-wide association studies (GWAS) that implicate genetic variation in the LIN28B gene as material in regulating developmental timing, and develop hypotheses about the combined and conditional effects of father absence and LIN28B that have yet to be studied in the father absence literature. 1.1. Evolutionary perspectives on father absence and AAM Early theorizing regarding why father absence is associated with AAM (and puberty more generally) stemmed from life history theory (see Ellis, Figueredo, Brumbach, & Schlomer, 2009). The central tenet of life history theory is that organisms make (non-conscious) strategic allocations of time and energy into fundamental components of growth and reproduction. Applied to life course development, for example, organisms make trade-offs between investment in somatic effort, which includes growth, development, and physical maintenance, and

http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002 1090-5138/© 2017 Published by Elsevier Inc.

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

2

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

investment in reproductive effort, such as sexual maturation and finding and/or retaining mates (see Ellis et al., 2009 for an extensive review). From this perspective, developmental experiences serve to entrain developmental trajectories toward life history phenotypes that will ultimately match expected adult reproductive ecologies. In one of the earliest examples of life history theory applied in human developmental research, Draper and Harpending (1982) theorized that experiencing father absence during development is a critical cue for daughter's expected adult reproductive ecology. Father absence specifically was hypothesized to signify that male participation in child rearing is not critical to reproductive success. In such an ecology, faster life history strategies, characterized by earlier maturation and reproduction, would be evolutionarily favorable. From this perspective, early pubertal timing associated with father absence represents an early facet of an overall developmentally calibrated faster life history strategy. Alternatively, father presence is hypothesized to cue that male parental investment is important for reproduction, which favors slower life history strategies, including later developmental timing. Building from Draper and Harpending (1982), Belsky, Steinberg, and Draper (1991) further hypothesized that early psychosocial stressors in and around the family also contribute to faster life history strategies (Belsky et al., 1991). This theory expanded on Draper and Harpending (1982) by positing that early experience (the first 5 to 7 years) is a sensitive period for calibrating perceptions of resource predictability, the trustworthiness of others, and the utility of interpersonal relationships. Children's experiences that inform these perceptions – including father absence – result in developmental and behavioral patterns that evolved to optimize reproductive success under these varying conditions. As it has come to be called, psychosocial acceleration theory has been generally well supported by empirical research. For example, a recent metaanalysis of research that examined the relation between father absence and AAM showed a small to moderate and statistically significant correlation across a set of 33 studies (Webster, Graber, Gesselman, Crosier, & Orozco-Schember, 2014). This theoretical foundation and empirical support suggests father absence and associated stressors function as evolutionarily important environmental cues for entraining life history development. 1.2. Father absence and age at menarche may be genetically confounded Although there is empirical support for the association between father absence and AAM, the causal status of father absence continues to be unclear, which stems, in part, from possible genetic confounding (Barbaro, Boutwell, Barnes, & Shackelford, 2017; Mendle et al., 2006). More specifically, psychosocial acceleration theory assumes father absence and associated stressors exogenously act upon AAM resulting in an evolutionarily coordinated phenotype (Belsky et al., 1991). An alternative explanation is that genetic factors are related to both father absence and AAM, which are passed down from father to daughter. Behavioral genetic models show that AAM is highly heritable – approximately 50% – leading to the conclusion that the association between father absence and AAM may be due to common genetic liability (Rowe, 2002). Evidence that AAM is heritable does not, however, constitute evidence that the association between father absence and AAM is due to a genetic confound since father absence effects could also operate through (non)shared environmental pathways (see Mendle et al., 2006). Surprisingly, there is relatively little research aimed at directly testing these alternative hypotheses and what research does exist is conflicting. Mendle et al. (2006) found that the association between daughters' stepfather exposure (which indicates biological father absence) and AAM was confounded by a shared genetic or environmental factor, but the data did not permit distinguishing between these two sources of variation. In another study, Tither and Ellis (2008) found that father absence was related to earlier age at menarche after controlling for genetic and family-wide confounds (see also Ellis, Schlomer,

Butler, & Tilley, 2012). Using a candidate gene approach, Comings, Muhleman, Johnson, and MacMurray (2002) found the X-linked androgen receptor gene (AR) was correlated with parental divorce, father absence, and AAM in girls, lending support for the genetic confounding hypothesis. The association between AR and father absence was not replicated, however, in a larger sample of Australian women (Jorm, Christensen, Rodgers, Jacomb, & Easteal, 2004). Taken together, these studies are suggestive that while at least a portion of the association between father absence and AAM may be attributed to genetic loading (i.e., heritability), father absence may additionally influence AAM through environmental means. Genetically informed studies that examine the relative contributions of genes and environments have moved research on life history theory forward by highlighting that genetic effects must be considered to get an accurate picture of putative environmental influences (Barbaro et al., 2017). This approach assumes, however, that genes and environments have independent main effects on phenotypic development. From a biological point-of-view, the genes plus (or versus) environment approach is problematic because genes cannot operate independently from environments (Meaney, 2010). Instead, gene-environment interactions (GxE) are key determinants for understanding how genes and environments co-act in developmental processes that produce complex phenotypes. GxE research is lacking, however, in the study of father absence and AAM. A notable exception, although not centered on father absence, are primary and replication candidate GxE studies (cGxE) that found the effect of early family relationships on AAM was moderated by variation in the estrogen receptor-α gene (ESR1; Hartman, Widaman, & Belsky, 2015; Manuck, Craig, Flory, Halder, & Ferrell, 2011). However, father absence was not directly addressed in either of these studies. 1.3. LIN28B, age at menarche, and father absence Incorporating molecular genetic information into research on father absence and AAM has the potential to move the field further by directly testing the complex interplay between genes and environments. One of the foremost challenges of cGxE research, however, is deciding which gene should be studied. Recent progress has been made on this issue and several papers have offered recommendations for gene-choosing strategies (Dick et al., 2015; Schlomer, Cleveland, Vandenbergh, Fosco, & Feinberg, 2015). Collectively, these researchers argue for a more comprehensive approach than has been previously used in cGxE research, which includes considering results from highly powered GWAS. Markers identified by GWAS are particularly good candidates since they have a high a priori probability of producing meaningful associations and interactions (Dick et al., 2015). Relevant to the current study, results from five independent GWAS have identified a robust and replicable association between variants linked to the LIN28B gene and AAM. He et al. (2009) identified six single nucleotide polymorphisms (SNPs; rs314277, rs314263, rs369065, rs314280, rs4946651, and rs314262) in or upstream from LIN28B that reached genome wide significance in a sample of 17,438 women; in a meta-analysis of eight different GWAS (N = 17,510), Perry et al. (2009) found an association between rs7759938 (near LIN28B) and AAM; Ong et al. (2009) showed one SNP in intron 2 of LIN28B (rs314276) reached genome-wide significance (N = 4,714); Sulem et al. (2009) in a study of 15,297 Icelandic women with replication sets from Iceland, Denmark, and the Netherlands found rs314280 was reliably associated with AAM, replicating He et al. (2009); in another meta-analysis of 32 GWAS (N = 87,802), Elks et al. (2010) also found rs7759938 was strongly related to AAM. This SNP (rs7759938) has also been linked to AAM in non-European ancestry populations including Korean (Hong et al., 2013), Japanese (Tanikawa et al., 2013), and Filipino (Croteau-Chonka, Lange, Lee, Adair, & Mohlke, 2013) samples (see also Witchel, 2016). Importantly, these SNPs are part of a large linkage-disequilibrium (LD) block associated with LIN28B and can be identified by two tag-SNPs, which are also

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

in moderate LD. Fig. 1 shows that two SNPs – rs364663 and rs314273 – together tag a large portion of the genomic region associated with LIN28B, including SNPs identified by these GWAS. In addition to considering GWAS when choosing genes to study, the biological role of the gene should also be evaluated (Schlomer et al., 2015). Although genes identified by GWAS do not necessarily have a known genomic function, it turns out that LIN28B plays a central role in developmental timing, especially in girls (Cousminer, Widen, & Palmert, 2016). The LIN28B gene is located at 6q21 and produces a protein of the Lin28 family that suppresses the formation of the let-7 microRNA (miRNA). The let-7 miRNA regulates gene expression via messenger RNA interference following DNA transcription and was initially discovered as a key regulator of developmental timing in C. elegans (Reinhart et al., 2000). Overexpression of Lin28 proteins in mice is associated with delayed growth cessation, later first estrus, later first litter, and delayed vaginal opening (a puberty marker in mice; Zhu et al., 2010). In addition, Zhu et al. (2011) found transgenic mice with human LIN28B were remarkably resistant to weight gain when fed a high fat diet. Notably, these phenotypic characteristics associated with Lin28 proteins resemble a slower life history strategy. The proposed mechanism for the relation between the Lin28/let-7 axis and pubertal development centers on let-7-dependent metabolic regulation. Animals with high Lin28 protein expression – and therefore less let-7 availability – are more insulin sensitive and more resistant to diabetes and obesity (Zhu et al., 2010). This work suggests the link between LIN28B and AAM may be mediated by metabolic processes related to later growth and development. Human research on LIN28B, let-7 suppression, and AAM is lacking, however. Nonetheless, the Lin28/let7 axis is conserved across evolutionarily disparate species suggesting an ancestral mechanism for regulating growth and development (Thornton & Gregory, 2012) and is a prime candidate for studying life history strategy regulation.

3

related to AAM. In addition, GWAS and functional genomic findings link LIN28B variation with later AAM. Given these lines of evidence, the purpose of this study was to test the combined effects of father absence and LIN28B on AAM. We expected earlier AAM among father absent girls and that LIN28B variation (rs364663 and rs314273) would be associated with later AAM, consistent with prior research. In addition, we expected the association between LIN28B and AAM would be moderated by father absence such that later AAM linked to LIN28B would be attenuated among father absent girls. We conceptualize father absence as moderating LIN28B, however a significant interaction in this study is mathematically equivalent to LIN28B moderating father absence. In the interest of completeness, we present significant findings from both conceptualizations below. Follow-up analyses were conducted to test the robustness of the results. 2. Material and methods 2.1. Participants Data used in this study came from a larger safer sex intervention study with women drawn from a moderately sized mid-western university. All data in the current report were collected prior to the intervention. In total, 321 women were recruited over 17 data collection sessions (M = 18.88 per session). The current analytic sample included data from 300 participants with complete genotype and phenotype data. The mean age of participants within the analytic sample was 20.46 years (SD = 1.31, range = 18 to 25). Sample participants primarily identified as Caucasian (n = 258; 86.3%) and included African American (n = 21; 7.0%), Asian (n = 19; 6.4%), and American Indian/Alaskan (n = 1; 0.3%). One participant did not report their ethnicity. The data used in this study have been made available in an online appendix following recommendations by the Center for Open Science's Transparency and Openness Promotion (TOP) Guidelines (Nosek et al., 2015).

1.4. The current study and hypotheses 2.2. Procedure There is an abundance of theory and empirical research that places father absence and associated stressors as important contextual factors

Participants were recruited primarily through advertising in university courses and listservs. Preceding enrollment, participants were informed that they would complete surveys that include questions about their sexual behavior and attitudes, participate in a safer sex or stress reduction intervention, and that their DNA would be collected via oral swab. Potential participants were provided a weblink where they could sign-up to participate in the study. On the day of data collection sessions, participants received email reminders to help ensure attendance in their scheduled session. Data collection sessions took place in the evening in reserved classrooms. Upon checking-in, participants were compared against an enrollment list (to prevent participating twice) and read an information script that provided more information about both the study and the data collection process. All identifying information was confidential. Participants were provided with informed consent, including consent to provide DNA, following a brief question/answer session. Participants were given a pre-test survey marked with a random ID number and completed surveys were deposited into a sealed container to help protect participant confidentiality. Following pre-test survey completion, DNA was collected from each participant via buccal swab. Participants received remuneration for participating. 2.3. Measures

Fig. 1. Linkage disequilibrium map of LIN28B tagged by rs364663 and rs314273. Darker shading indicates higher LD. The boxes to the right of the rs numbers indicate minor allele frequency, the more fully shaded the closer MAF is to 50%. Dots indicate tag-SNPs. Lighter colored rs numbers fall within the 85 kb flanking region. (http://snpinfo.niehs. nih.gov/snpinfo/snptag.htm). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2.3.1. Age at menarche Participants were asked the following question: How old were you when you first began to menstruate (got your period)? Retrospective reports of age at menarche have demonstrated strong reliability among young women (e.g., Koprowski, Coates, & Bernstein, 2001), such as those in the current study. Mean age at menarche was 12.63 years (SD

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

4

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

= 1.44; range = 8 to 17), which is similar to other ethnically comparable community and national samples (Hendrick, Cance, & Maslowsky, 2016; Jorm et al., 2004; Krieger, Kiang, Kosheleva, Waterman, Chen, & Beckfield, 2015). 2.3.2. Father absence Father absence was operationalized in this study as separation or divorce of the participants' birth parents and was assessed via biological parents' current marital status. Response options were 1 = married, 2 = separated/divorced, and 3 = my parents never married. Participants primarily reported their parents were married (n = 216; 72.0%) with 23% (n = 69) and 4.9% (n = 15) reporting their parents as separated/divorced or never married, respectively. Participants whose parents were married were coded 0 = Father present (n = 216, 72.0%); participants whose parents never married or were separated/divorced were coded 1 = Father absent (n = 84, 28.0%). Mean age at father absence was 10.51 (SD = 6.15; range = 1 to 22, n = 66); 40% (n = 26) were father absent before age 7. The operationalization of father absence used in this study leaves open the possibility that some portion of women coded as father absent did not experience absence of their biological father from the home, such as cases of non-married cohabitation or divorce followed by absence of the biological mother. Data on cohabitation suggest that although cohabiting couples are historically on the rise, cohabiting couples' relationships tend to be relatively unstable (Bumpass & Lu, 2000) and children born in this family structure are more likely to experience the dissolution of their parents' union compared to married couples (Manning, 2004). In addition, children most often reside in the custody of their biological mother in dissolved unions, whether previously married or cohabitating (e.g., Seltzer, 1991). Although the current measure primarily reflects biological father absence, it likely also captures variance attributed to more general family instability. Importantly, this more general conceptualization of father absence (i.e., father absence and associated stressors) is consistent with thinking from psychosocial acceleration theory, which stipulates a role for familial instability in addition to father absence (Belsky et al., 1991). Links between father absence and AAM have been found in research using similar measures (e.g., Maestripieri, Roney, DeBias, Durante, & Spaepen, 2004). 2.3.3. LIN28B SNPs used in the current analysis (rs364663 and rs314273) were identified as tag SNPs using the National Institute of Environmental Health Sciences LD Tag SNP Selection (TagSNP) online tool (http:// snpinfo.niehs.nih.gov/snpinfo/snptag.htm), which uses an algorithm developed by Xu, Kaplan, and Taylor (2007) for tag SNP identification. Using this method, rs364663 (A/T) was identified as tagging 13 other SNPs in or near LIN28B (see Fig. 1) with a minimum LD threshold of R2 = 0.80; rs314273 (G/T) was identified as tagging 11 other SNPs in the same region. In total, these two SNPs reflect variation in 26 polymorphisms. Both SNPs are intron variants and are in moderate LD with each other (R2 = 0.67; SNAP Pairwise LD). DNA for this study was collected via buccal swab. SNPs were genotyped using the OpenArray system from Life Technologies Inc. (now part of Thermo Fisher, Inc.), which utilizes TaqMan genotyping assays applied to an array. Collectively, rs364663 and rs314273 had a 96.6% genotyping success rate. Regenoypting 20% of the sample showed a 97.0% and 95.5% genotyping accuracy rate for rs364663 and rs314273, respectively. Discordant cases (n = 5) were dropped from the analysis. Table 1 Genotype frequencies of LIN28B SNPs. SNP

Genotype, N(%)

rs364663 rs314273

AA, 94(31.3) GG, 138(46)

Note: MAF = minor allele frequency.

MAF AT, 152(50.7) GT, 127(42.3)

TT, 54(18) TT, 35(11.7)

0.43 0.33

Both SNPs were in Hardy-Weinberg Equilibrium (both p's N 0.50; see Table 1 for genotype frequencies). Results of GWAS (reviewed above) indicate the T allele for both SNPs as related to later pubertal development. SNPs were dummy coded such that the reference group for each SNP was no copies of the T allele (A/A and G/G for rs364663 and rs314273, respectively; see Aliev, Latendresse, Bacanu, Neale, & Dick, 2014). 2.3.4. Covariates Two control variables were used in these analyses to help account for possible confounding. First was ethnicity, which was coded 0 = Caucasian (n = 258), 1 = Non-Caucasian (n = 41; one participant did not report their ethnicity). Ethnicity has a robust association with AAM (Wu, Mendola, & Buck, 2002) and ethnicity controls can help account for possible genetic stratification due to allele frequency differences across ancestral populations (Cardon & Palmer, 2003). Second, was financial hardship, which prior research indicates is correlated with AAM (e.g., Boynton-Jarrett & Harville, 2012). Financial hardship was measured with 4 items that shared the stem, “How strongly do you agree or disagree with the following statements about your family's financial situation when you were growing up?” Items assessed if participants felt their family had enough money to afford the kind of home, clothing, food, and medical care they needed (1 = Strongly Disagree, 5 = Strongly Agree; higher values equal less hardship). These items were adapted from a family economic hardship scale (Conger & Donnellan, 2007; Conger, Ge, Elder, Lorenz, & Simons, 1994). The 4 items showed good internal consistency (α = 0.96) and were averaged (M = 4.58, SD = 0.76). 3. Results 3.1. Preliminary analyses Correlations between variables used in this study showed women classified as father absent had earlier AAM (r = − 0.15), were more likely to be non-Caucasian (r = − 0.14), and reported more financial hardships growing up (r = −0.22, all p's b 0.05). Although not part of the current study, father absence was also correlated with girls' earlier age at first vaginal intercourse (r = −0.14), more past year sexual partners (r = 0.13), and more lifetime sexual partners (r = 0.18, all p's b 0.05). This pattern of correlations is consistent with other studies that use prospective measures of father absence (Ellis, 2004; Ellis et al., 2003). In addition, negligible correlations were found between LIN28B SNPs and father absence (both SNPs r = −0.01, ns), which suggest interactions between these factors are unlikely due to gene-environment correlation (Manuck & McCaffery, 2014). Allele frequencies for both LIN28B SNPs did not systematically vary between Caucasian and nonCaucasians (χ2(2) b 1.68, ns, both SNPs) which suggests genomic stratification likely has a trivial impact on these results. Both SNPs were unrelated to financial hardship (r's b 0.05, ns). 3.2. Model 1: Father absence and rs364663 Hypotheses were tested using hierarchical regression models that included father absence and dummy coded SNP variables and two product terms between father absence and each dummy variable (see Table 2). Results for rs364663 showed that girls who came from father absent households had earlier age at menarche (b = −0.489, p b 0.05; partial r2 = 0.023). No main effects were found for rs364633 (A/A vs. A/T: b = 0.146, ns; A/A vs. T/T: b = 0.306, ns; partial r2's b 0.001). Adding the two product terms showed a significant interaction between father absence and A/A vs. T/T (b = −1.120, p b 0.05; partial r2 = 0.016). In Fig. 2, father present girls showed significantly later AAM if they had the T/T genotype compared to A/A (b = 0.675, p b 0.05; partial r2 = 0.017); father absent girls, however, did not show this genotype difference (b = −0.444, ns; partial r2 = 0.003). Notably, the 95% confidence intervals

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

3.3. Model 2: Father absence and rs314273

Table 2 Results of interaction models for father absence and LIN28B SNPs. b(SE) Variables Main effects Father absence rs364663 (AA vs. AT) rs364663 (AA vs. TT) rs314273 (GG vs. GT) rs314273 (GG vs. TT) Two-way interactions Father absence × rs364663 (AA vs. AT) Father absence × rs364663 (AA vs. TT) Father absence × rs314273 (GG vs. GT) Father absence × rs314273 (GG vs. TT)

5

Model 1

Model 2

−0.489(0.185)⁎ 0.146(0.188) 0.306(0.244)

−0.449(0.184)⁎ 0.405(0.204)⁎ 0.178(0.336)

−0.111(0.415) −1.120(0.517)⁎ −0.072(0.400) −0.847(0.561)

Residual degrees of freedom in both models = 296. ⁎ p b 0.05.

for father present A/A vs. T/T girls overlap somewhat in Fig. 2 despite significantly different means (see Schenker & Gentleman, 2012). Further, girls with the T/T genotype showed significantly later AAM if they were father present compared to father absent (b = − 1.33, p b 0.05; partial r2 = 0.035). To additionally test for differences between girls of A/T vs. T/T genotype, a second regression was conducted except rs364663 was dummy coded such that A/T was the reference group. Results showed a marginally significant difference in AAM between A/T vs. T/T girls who were father present (b = −0.481, p = 0.08; partial r2 = 0.011) and no significant effect among father absent girls (b = 0.528, ns; partial r2 = 0.006). The interaction between A/T vs T/T and father absence was statistically significant (b = − 1.009, p b 0.05; partial r2 = 0.014), suggesting that girls with two copies of the rs364663 T allele experienced later AAM compared to girls with one copy, a difference that was not found in father absent households.

Model 2 consisted of a hierarchical regression similar to Model 1. Like Model 1, father absence was again significantly associated with earlier AAM (b = −0.449, p b 0.05; partial r2 = 0.200). In addition, there was a significant main effect of LIN28B between G/G and G/T genotypes (b = 0.405, p b 0.05; partial r2 = 0.018) but not between G/G and T/T (b = 0.178, ns; partial r2 = 0.001). The interaction between father absence and G/G vs. T/T did not reach statistical significance (b = −0.847, p = 0.13; partial r2 = 0.008) but showed a similar pattern as was found for rs364663. The father absence by G/G vs. G/T interaction was not significant (b = −0.072, ns; partial r2 b 0.001). 3.4. Model 3: Father absence and LIN28B index Because rs364663 and rs314273 are in moderate LD and showed a strong correlation in these data, the two SNPs were combined to create a LIN28B genetic index. The SNPs were combined such that 0 = no copies of the T/T genotype (n = 246) and 1 = at least one copy of the T/T genotype (n = 54). Results of a 2 (father absence) × 2 (LIN28B T/T genotype) ANOVA showed a significant difference between father present (M = 12.94, SE = 0.13) and father absent (M = 12.13, SE = 0.19) girls' AAM (F(1, 296) = 12.65, p b 0.05; partial η2 = 0.041) and no main effect of LIN28B (F(1, 296) b 1.0, ns; partial η2 b 0.001). The interaction between father absence and LIN28B was statistically significant (F(1, 296) = 5.17, p b 0.05; partial η2 = 0.017; see Fig. 3). Follow-up simple effects tests showed AAM was significantly later among father present girls with at least one copy of the T/T genotype compared to girls with no copies (13.22 vs 12.67, respectively; t(214) = 2.22, p b 0.05; r2 = 0.023). This difference was not significant among father absent girls (T/T = 11.89 vs no T/T = 12.37; t(82) = 1.18, ns; r2 = 0.016), consistent with prior models. In addition, father present girls with at least one copy of the T/T genotype showed significantly later AAM compared to T/T genotype girls who were father absent (t(52) = 2.87, p b 0.05; r2 = 0.137). Last, adding participant ethnicity and financial hardship as covariates reduced the p-value for the interaction to 0.051 (F(1, 292) = 3.83). Of note, the effect size for the interaction in this model (partial η2 = 0.013) was similar to the effect size for ethnicity (partial η2 = 0.014), suggesting the interaction explained a similar amount of additional variance in AAM as non-European ethnicity. 3.5. Follow-up analyses A limitation of the current research is our retrospective measure of father absence, which may be less precise than prospective measures based on family composition metrics. For example, it is possible that among our n = 15 participants that reported their parents were never married, some may have cohabited. To help test the sensitivity of the

Fig. 2. Interactions between father absence and rs364663 (top) and rs314273 (bottom).

Fig. 3. Interaction between combined LIN28B SNPs and father absence.

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

6

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

results to this issue, Model 3 was conducted excluding these cases. Results showed the interaction remained statistically significant (F(1, 281) = 4.48, p b 0.05; partial η2 = 0.016). In addition, n = 23 (27.4%) father absent women reported that their biological parents divorced/ separated after menarche (2 women did not report how old they were when their biological parents divorced/separated). Model 3 was conducted again dropping these 23 cases. Results showed that excluding these cases did not affect the significance of the interaction term (F(1, 273) = 5.25, p b 0.05) and a slightly stronger effect was detected (partial η2 = 0.019 vs. 0.017). In addition, given that father absence is correlated with ethnicity and financial hardship, an alternative explanation of the present findings is that father absence serves as a proxy for these other factors. To test these alternative hypotheses, Model 3 was conducted two additional times using ethnicity and financial hardship, respectively, in place of father absence. Results showed ethnicity did not interact with LIN28B to predict AAM (b = 0.250, ns; partial r2 = 0.003) nor did financial hardship (F(1, 295) b 1.00, ns; partial η2 = 0.003). Because life history models of father absence emphasize early experience as critical for calibrating developmental trajectories (i.e. first 5 to 7 years; Belsky et al., 1991), we sought to further test father absence timing effects in conjunction with LIN28B variation. Overall, there was a modest and non-significant correlation between age at father absence and AAM (r = 0.19, ns) among the n = 65 participants with age at father absence data. To make direct comparisons between early and late father absence girls, a three-level family status variable coded 0 = father present (n = 216), 1 = father absent at age 7 and under (n = 26), 2 = father absent at age 8 and up (n = 39) was created and entered into a univariate ANOVA with AAM as the dependent variable. Results showed an overall main effect of family status (F(2278) = 3.18, p b 0.05; partial η2 = 0.022). A Helmert planned contrast (level 1 vs. later, level 2 vs. 3) showed a significant difference between father present and father absent girls' AAM (t = 2.23, p b 0.05; r2 = 0.017). Age at menarche did not significantly differ between father absent girls prior to age 7 (M = 12.04) and those after age 7 (M = 12.59; t = 1.55, ns; r2 = 0.008), which suggests father absence effects on AAM did not appreciably differ by age in these data. Including LIN28B in this model showed a marginally significant interaction with family status (F(2, 275) = 2.58, p = 0.08; partial η2 = 0.018). Consistent with prior models, AAM was significantly later among father present girls with at least one copy of the T/T genotype (t(214) = 2.215, p b 0.05; r2 = 0.023) and no difference was observed for girls who were father absent below or above age 7 (t(24) = 0.401, ns, r2 = 0.006; t(37) = 1.18, ns, r2 = 0.036 respectively). Last, we examined the hypothesis that number of years without a biological father prior to AAM would predict AAM above and beyond father absence status. Girls who reported being father present were assigned a value of zero for years without a biological father prior to AAM. This analysis permits us to examine father absence timing effects without the need to artificially categorize continuous variables, such as father absence above or below 7 years (MacCallum, Zhang, Preacher, & Rucker, 2002). Results of this analysis showed no significant association between AAM and years without a biological father prior to AAM (b = −0.032, ns; partial r2 = 0.002). Taken together, these results suggest the effects of father absence, LIN28B, and their interaction are statistically consistent across these different timings of father absence. Finally, many molecular GxE studies have been cast in the differential susceptibility theoretical framework (DST; Belsky & Pluess, 2009; Ellis, Boyce, Belsky, Bakermans-Kranenburg, & van IJzendoorn, M.H., 2011) based on the premise that some genetic variants will contribute to a neurobiologically sensitive phenotype. Briefly, DST posits that intrapersonal characteristics, such as genes, that are liabilities for negative outcomes in adverse environments may permit exceptional advantages in more favorable contexts. There is, however, no a priori reason to think that variation in LIN28B, which is functionally related to metabolic processes, would have contributed to an evolved system of neural plasticity. Thus, we expected the interaction between father absence and

LIN28B would be ordinal, consistent with diathesis stress, rather than disordinal, which would suggest DST. To formally test this hypothesis, we implemented the Widaman, Helm, Castro-Schilo, Pluess, and Belsky (2012) reparametrized regression approach. The reparametrized model includes the interaction crossover point as a model parameter, which permits confidence intervals to be estimated for the crossover. Inferences about the likely population interaction form – ordinal vs. disordinal – can be made based on the crossover point estimate and its confidence intervals. Results of this analysis using SAS 9.4 (2013) showed the estimated crossover point was −0.28 and had a 95% confidence interval between − 0.82 and 0.26. Because the crossover point falls outside the range of the predictor (LIN28B coded 0/1), the interaction form is consistent with an ordinal interaction. However, because the upper 95% confidence limit falls within the predictor range, a disordinal interaction in the population cannot be ruled out (Widaman et al., 2012). 4. Discussion Understanding the etiology of AAM can potentially inform prevention efforts that can have lifetime effects on the health and well-being of women. Research that has looked to family factors for casual predictors of AAM have generally focused on father absence. Although this research has demonstrated a replicable empirical phenomena, the causal status of father absence on AAM remains elusive due to genetic confounds. Research on genetic and environmental contributions to AAM have generally been designed to either tease apart these influences or estimate the influence of one while controlling for the other. Rather than pose genes and environments as alternative explanations, the current research sought to examine the interplay, by virtue of GxE, between father absence and variation in the LIN28B gene (rs364663 and rs314273). LIN28B has been identified in several GWAS of AAM and is functionally linked to later pubertal development, likely through metabolic processes, and is a key developmental regulator conserved across evolutionarily disparate species. Drawing on psychosocial acceleration theory we hypothesized earlier menarche among women who grew up in father absent compared to father present homes. In addition, we hypothesized that LIN28B genetic variation would be linked to later pubertal timing and that this association would be attenuated by father absence. The results of this study were generally consistent with hypotheses. In all models father absence was related to earlier AAM. The hypothesized association between LIN28B and later AAM was less clear, however, given the significant difference between G/G and G/T genotypes for rs314273, but not between G/G and T/T or G/T and T/T genotypes. Since the mean ages at menarche were similar for G/T and T/T girls, the null finding for G/G versus T/T was likely due to smaller N of the T/T group for this SNP. Consistent with our hypothesis was a significant interaction between father absence and rs364663 whereby girls with the T/T genotype showed significantly later AAM compared to A/A girls, but only if they were from a father present household. A similar but non-significant pattern of results was detected for rs314273. These interactions suggests the main effect of LIN28B on later AAM is stronger among father present girls compared to father absent. Results from a 2 × 2 ANOVA of father absence and the combined LIN28B genetic index showed that father present girls with at least on copy of the T/T genotype for either rs364663 or rs317273 evidenced significantly later AAM relative to girls with no T/T genotype, which was not detected among father absent girls. These results were robust to financial hardship and ethnicity controls. Follow-up analyses centered on testing the consistency of the findings to sampling, measurement, and father absence timing effects. The initial results were robust in these analyses. 4.1. Genetics, life history theory, and father absence Significant advances can be made in future life history research on father absence and AAM by more explicitly integrating genetic

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

7

components (e.g., Barbaro et al., 2017). We contend that a GxE approach in particular can significantly enhance this research. If humans have evolved to calibrate development based on early cues, then conditional adaptations should be conceptualized as GxE. For example, the caterpillar Nemoria arizonaria evolved mechanisms to phenotypically respond to their early diet and adopt one of two morphs that confer an adaptive advantage (i.e., appropriate camouflage; Greene, 1989). These adaptive mechanisms must be underlain by selected genes that conditionally express. These caterpillars have an evolved genetic architecture that contingently respond a specific environmental input and adjusts their developmental trajectory in evolutionarily advantageous ways. In effect, conditional adaptations are GxE. This is the first study to use a molecular genetic approach to study the relation between father absence and AAM as a GxE. As such, these findings should be considered preliminary and await replication and extension in larger more diverse samples. In addition, conceptualizing variation in AAM as an emergent, nonadditive property of father absence and genetic disposition is more consistent with biological views of how genes and environments produce developmental phenotypes. Statistical models that parse genetic and environmental sources of variance provide a limited view of how these factors actually operate at a genetic level given no genes are actually measured. Nonetheless, research that examines father absence as either a genetic or environmental effect continues to be important and will have implications for GxE research. For example, it remains unclear if the E in the current GxE truly is an environmental effect or if it reflects genetic loading, and thereby a GxG. Given the negligible correlation between LIN28B variation and father absence, any G that might be included in father absence is independent of LIN28B. Last, it is important to point out that we did not find evidence that timing of father absence was related to AAM. Psychosocial acceleration theory places early father absence – during the first 5 to 7 years – as a key period for life history strategy development (Belsky et al., 1991; Draper & Harpending, 1982) and other empirical research has found evidence for a timing effect, whereby early father absence is more strongly linked with AAM (e.g., Culpin, Heron, Araya, Melotti, Lewis, & Johnson, 2014; Quinlin, 2003). The null findings reported in this paper are most likely due to the reduced number of father absent girls required for this analysis (n = 65). In fact, the pattern of results, both the correlation and the ANOVA, are in line with an early father absence effect. We view the null findings as more likely the result of reduced power to statistically detect this difference rather than an actual absence of the association.

moderating LIN28B, in part, due to this reasoning. It may be the case, as suggested by these data, that LIN28B effect sizes found in GWAS studies are attenuated by not considering father presence/absence subgroups. Another critique of cGxE involves how researchers choose what gene to study and concern has been expressed for overreliance on “usual suspect” markers (e.g., 5-HTTLPR, DRD4, MAOA). Although there is nothing inherently problematic with studying these makers, gene-choosing strategies should take advantage of multiple sources of information that implicate the gene(s) as relevant for the outcome. In this paper, we relied on results from several independent GWAS, research on LIN28B genomic function, and LD patterns in LIN28B that lead us to choose rs364663 and rs314273 for studying AAM. As a result, the prior probability that these SNPs and their interactions with father absence are meaningful is presumed to be high (Dick et al., 2015; Ioannidis, 2005). Greater prior probability for meaningful associations lend additional confidence that these findings can be replicated in future studies. Related to the “usual suspect” critique is the notion that single candidate gene studies do not reflect the state of the gene-finding science (e.g., Dick et al., 2015), which has moved beyond single marker approaches. In this study, however, rs364663 and rs314273 identify variation in 26 SNPs associated with LIN28B. Choosing tag-SNPs such as these can be an attractive alternative for developmental scientists since larger genomic regions can be evaluated with fewer markers. Additionally, cGxE in developmental research often does not fall under the category of gene-finding research, characteristic of GWAS, and should be considered a confirmatory approach, especially when motivated by theory and hypotheses. Last, statistically detecting interactions can be highly dependent on measurement scale and in many psychological studies choice of scale is arbitrary. Age at menarche does not suffer this limitation since it is measured on a non-arbitrary ratio scale (years) that has a true zero. In light of these strengths, limitations should also be acknowledged. For instance, the sample size used is relatively small for a genomic study. Additional studies with larger samples are needed to further test the hypothesized associations that were detected as well as those that were not. In addition, because the sample is drawn from a university population, additional research is needed to determine if these findings can be replicated in a more representative sample. Last, stronger tests of moderation may be possible using longitudinal data and prospective measurement of father absence and AAM. Despite these limitations, the strengths of this study described above provide confidence that these results are not unique to this sample.

4.2. Critiques of cGxE research and the current findings

Acknowledgements

It is important to note that cGxE research has been the subject of much debate and some have argued that all cGxE studies should be treated with skepticism (e.g., Duncan, Pollastri, & Smoller, 2014). In light of these concerns, this paper addresses several of these critiques by drawing on recommendations for best practices in cGxE research (e.g., Cleveland, Schlomer, Vandenbergh, & Wiebe, 2016; Dick et al., 2015; Schlomer et al., 2015). Perhaps the primary criticism of cGxE research is a suspicion that researchers may be leveraging chance results from multiple testing, which stems from the ambiguous replication history of cGxE studies and small effect sizes typically detected in GWAS. One approach to assuage these concerns is for researchers to rely heavily on theory for guiding choices about what environments and outcomes to examine (Schlomer et al., 2015). In the current study, we drew on life history models of pubertal development that conceptualize father absence as a contextual influence and is explicit about the role of fathers on daughters' AAM. Regarding small effect sizes, it is possible that one reason GWAS find small main effects may be due to the fact that genetic associations vary across environments. If an association is minuscule in one population and larger in another, the overall main effect may be small. We chose to conceptualize father absence as

The authors are indebted to the Jacobs Foundation who provided support to collect and analyze these data. The authors would like to thank Amanda M. Griffin and Brianna Yates for their assistance with data collection and David J. Vandenbergh and Erin R. Baker who provided comments on earlier versions of this manuscript. We would also like to thank the Penn State Genomics Core Facility for genotyping. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.evolhumbehav.2017.06.002. References Aliev, F., Latendresse, S. J., Bacanu, S. A., Neale, M. C., & Dick, D. M. (2014). Testing for measured gene-environment interaction: Problems with the use of cross-product terms and regression model reparameterization solution. Behavior Genetics, 44, 165–181. http://dx.doi.org/10.1007/s10519-014-9642-1. Barbaro, N., Boutwell, B. B., Barnes, J. C., & Shackelford, T. K. (2017). Genetic confounding of the relationship between father absence and age at menarche. Evolution and Human Behavior, 38, 357–365. http://dx.doi.org/10.1016/jevolhumbehav.2016.11. 007.

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

8

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx

Belsky, J., & Pluess, M. (2009). Beyond diathesis stress: Differential susceptibility to environmental influences. Psychological Bulletin, 135, 885–908. http://dx.doi.org/10.1037/ a0017376. Belsky, J., Steinberg, L., & Draper, P. (1991). Childhood experience, interpersonal development, and reproductive strategy: An evolutionary theory of socialization. Child Development, 62, 647–670. Boynton-Jarrett, R., & Harville, E. W. (2012). A prospective study of childhood social hardships and age at menarche. Annals of Epidemiology, 22, 731–737. http://dx.doi.org/10. 1016/j.annepidem.2012.08.005. Bumpass, L., & Lu, H. (2000). Trends in cohabitation and implications for children's family contexts in the United States. Population Studies, 54, 29–41. Cardon, L. R., & Palmer, L. J. (2003). Population stratification and spurious allelic association. The Lancet, 361, 598–604. http://dx.doi.org/10.1016/S0140-6736(03)12520-2. Cleveland, H. H., Schlomer, G. L., Vandenbergh, D. J., & Wiebe, R. P. (2016). Gene x intervention designs. A promising step toward understanding etiology and building better preventive interventions. Criminology & Public Policy, 15(3), 1–10. http://dx.doi.org/ 10.1111/1745-9133.12221. Comings, D. E., Muhleman, D., Johnson, J. P., & MacMurray, J. P. (2002). Parent-daughter transmission of the androgen receptor gene as an explanation of the effect of father absence on age of menarche. Child Development, 73, 1046–1051. Conger, R. D., & Donnellan, M. B. (2007). An interactionist perspective on the socioeconomic context of human development. Annual Review of Psychology, 58, 175–199. Conger, R. D., Ge, X., Elder, G. H., Lorenz, F. O., & Simons, R. L. (1994). Economic stress, coercive family process, and developmental problems of adolescence. Child Development, 65, 541–561. Cousminer, D. L., Widen, E., & Palmert, M. R. (2016). The genetics of pubertal timing in the general population: Recent advances and evidence for sex specificity. Current Opinion in Endrocrinology, Diabetes, & Obesity, 23, 57–65. http://dx.doi.org/10.1097/MED. 0000000000000213. Croteau-Chonka, D. C., Lange, L. A., Lee, N. R., Adair, L. S., & Mohlke, K. L. (2013). Replication of LIN28B SNP association with age of menarche in young Filipino women. National Institutes of Health, 8, e50–e53. http://dx.doi.org/10.1111/j.2047-6310.2013. 00178. Culpin, I., Heron, J., Araya, R., Melotti, R., Lewis, G., & Johnson, C. (2014). Father absence and timing of menarche in adolescent girls from a UK cohort: The mediating role of maternal depression and major financial problems. Journal of Adolescence, 37, 291–301. http://dx.doi.org/10.1016/j.adolesence.2014.02.003. Dick, D. M., Agrawal, A., Keller, M. C., Adkins, A., Aliev, F., Monroe, S., ... Sher, K. J. (2015). Candidate gene-environment interaction research: Reflections and recommendations. Perspectives on Psychological Science, 10, 37–59. http://dx.doi.org/10.1177/ 17456914556682. Draper, P., & Harpending, H. (1982). Father absence and reproductive strategy: An evolutionary perspective. Journal of Anthropological Research, 38, 255–273. Duncan, L. E., Pollastri, A. R., & Smoller, J. W. (2014). Mind the gap: Why many geneticists and psychological scientists have discrepant views about gene-environment interaction (GxE) research. American Psychologist, 69, 249–268. http://dx.doi.org/10.1037/ a0036320. Elks, C. E., Perry, J. R. B., Sulem, P., Chasman, D. I., Franceschini, N., He, C., ... Murray, A. (2010). Thirty new loci for age at menarche identified by a meta-analysis of genome-wide association studies. Nature Genetics, 42, 1077–1085. http://dx.doi.org/ 10.1038/ng.714. Ellis, B. J. (2004). Timing of pubertal maturation in girls: An integrated life history approach. Psychological Bulletin, 130, 920–958. http://dx.doi.org/10.1037/0033-2909. 130.6.920. Ellis, B. J., Bates, J. E., Dodge, K. A., Fergusson, D. M., Horwood, L. J., Pettit, G. S., & Woodward, L. (2003). Does father absence place daughters at special risk for early sexual activity and teenage pregnancy? Child Development, 74, 801–821. Ellis, B. J., Boyce, W. T., Belsky, J., Bakermans-Kranenburg, M. J., & van IJzendoorn, M.H. (2011). Differential susceptibility to the environment: An evolutionaryneurodevelopmental theory. Development and Psychopathology, 23, 7–28. Ellis, B. J., Figueredo, A. J., Brumbach, B. H., & Schlomer, G. L. (2009). Fundamental dimensions of environmental risk: The impact of harsh versus unpredictable environments on the evolution and development of life history strategies. Human Nature, 20. http:// dx.doi.org/10.1007/s12110-009-9063-7. Ellis, B. J., Schlomer, G. L. Butler, & Tilley, E. A. (2012). The impact of fathers on risky sexual behavior in daughters: A genetically and environmentally controlled sibling study. Development and Psychopathology, 24, 317–332. http://dx.doi.org/10.1017/ s095457941100085x. Gong, T. T., Wu, Q. T., Vogtmann, E., Lin, B., & Wang, Y. L. (2013). Age at menarche and risk of ovarian cancer: A meta-analysis of epidemiological studies. International Journal of Cancer, 132, 2894–2900. http://dx.doi.org/10.1002/ijc.27952. Greene, E. (1989). A diet-induced developmental polymorphism in a caterpillar. Science, 243, 643–646. Hartman, S., Widaman, K. F., & Belsky, J. (2015). Genetic moderation of effects of maternal sensitivity on girl's age of menarche: Replication of the Manuck et al. study. Development and Psychopathology, 27, 747–756. http://dx.doi.org/10.1017/ s0954579474000856. He, C., Kraft, P., Chen, C., Buring, J. E., Pare, G., Hankinson, S. E., ... Chasman, D. I. (2009). Genome-wide association studies identify loci associated with age at menarche and age at natural menopause. Nature Genetics, 41, 724–728. http://dx.doi.org/10.1038/ng.385. Hendrick, C. E., Cance, J. D., & Maslowsky, J. (2016). Peer and individual risk factors in adolescence explaining the relationship between girls' pubertal timing and teenage childbearing. Journal of Youth and Adolescence, 45, 916–927. http://dx.doi.org/10. 1007/s10964-016-0413-6. Hong, K., Kim, C., Min, H., Park, S., Park, J. K., Ahn, Y., ... Kim, Y. (2013). Replication of genomewide association studies on age at menarche in the Korean

population. Genes & Genomics, 35, 69–75. http://dx.doi.org/10.1007/s13258013-0060-1. Ioannidis, J. P. A. (2005). Why most published research findings are false. PLoS Medicine, 8, 696-071. Jorm, A. F., Christensen, H., Rodgers, B., Jacomb, P. A., & Easteal, S. (2004). Association of adverse childhood experiences, age of menarche, and adult reproductive behavior: Does the androgen receptor gene play a role? American Journal of Medical Genetics: Part B Neuropsychiatric Genetics, 125, 105–111. http://dx.doi.org/10.1002/ajmg.b. 20114. Koprowski, C., Coates, R. J., & Bernstein, L. (2001). Ability of young women to recall past body size and age at menarche. Obesity Research, 9, 478–485. Krieger, N., Kiang, M. V., Kosheleva, A., Waterman, P. D., Chen, J. T., & Beckfield, J. (2015). Age at menarche: 50-year socioeconomic trends among US-born black and white women. American Journal of Public Health, 105, 388–397. Lichty, L. F., & Campbell, R. (2012). Targets and witnesses: Middle school students' sexual harassment experiences. Journal of Early Adolescence, 32, 414–430. http://dx.doi.org/ 10.1177/0272431610396090. MacCallum, R. C., Zhang, S., Preacher, K. J., & Rucker, D. D. (2002). On the practice of dichotomization of quantitative variables. Psychological Methods, 7, 19–40. Maestripieri, D., Roney, J. R., DeBias, N., Durante, K. M., & Spaepen, G. M. (2004). Father absence, menarche and interest in infants among adolescent girls. Developmental Science, 560–566. Manning, W. D. (2004). Children and the stability of cohabitating couples. Journal of Marriage and Family, 66, 674–689. http://dx.doi.org/10.1111/j.0022-2445.2004. 00046.x. Manuck, S. B., Craig, A. E., Flory, J. D., Halder, I., & Ferrell, R. E. (2011). Reported early family environment covaries with menarcheal age as a function of polymorphic variation in estrogen receptor-α. Development and Psychopathology, 23, 69–83. http://dx.doi.org/ 10.1017/s0954579410000659. Manuck, S. B., & McCaffery, J. M. (2014). Gene-environment interaction. Annual Review of Psychology, 65, 41–70. http://dx.doi.org/10.1146/annurev-psych-010213-115100. Meaney, M. J. (2010). Epigenetics and the biological definition of gene × environment interactions. Child Development, 81, 41–79. Mendle, J., Turkhiemer, E., D'Onofrio, B. M., Lynch, S. K., Emery, R. E., Slutske, W. S., & Martin, N. G. (2006). Family structure and age at menarche: A children-of-twins approach. Development and Psychopathology, 42, 533–542. http://dx.doi.org/10.1037/ 0012-1649.42.3.533. Mendle, J., Turkheimer, E., & Emery, R. E. (2007). Detrimental psychological outcomes associated with early pubertal timing in adolescent girls. Developmental Review, 27, 151–171. http://dx.doi.org/10.1016/j.dr.2006.11.001. Mrug, S., Elliott, M. N., Davies, S., Tortolero, S. R., Cuccaro, P., & Schuster, M. A. (2013). Early puberty, negative peer influence, and problem behaviors in adolescent girls. Pediatrics, 133, 1–8. http://dx.doi.org/10.1542/peds.2013-0628. Nosek, B. A., Alter, G., Banks, G. C., Borsboom, D., Bowman, S. D., Breckler, S. J., ... Yarkoni, T. (2015). Promoting an open research culture. Science, 348, 1422–1425. http://dx.doi. org/10.1126/science.aab2374. Ong, K. K., Elks, C. E., Li, S., Zhao, J. H., Luan, J., Anderson, L. B., ... Aareham, N. J. (2009). Genetic variation in LIN28B is associated with the timing of puberty. Nature Genetics, 41, 729–733. http://dx.doi.org/10.1038/ng.382. Perry, J. R. B., Stolk, L., Franceschini, N., Lunetta, K. L., Zhai, G., McArdle, P. F., ... Murabito, J. M. (2009). Meta-analysis of genome-wide association data identifies two loci influencing age at menarche. Nature Genetics, 41, 648–650. http://dx.doi.org/10. 1038/ng.386. Prentice, P., & Viner, R. M. (2013). Pubertal timing and adult obesity and cardiometabolic risk in women and men: A systematic review and meta-analysis. International Journal of Obesity, 37(8), 1036–1043. Quinlin, R. J. (2003). Father absence, parental care, and female reproductive development. Evolution and Human Behavior, 24, 376–390. http://dx.doi.org/10.1016/s10905138(03)00039-4. Reinhart, B. J., Slack, F. J., Basson, M., Pasquinelli, A. E., Bettinger, J. C., Rougvie, A. E., ... Ruvkun, G. (2000). The 21-nucleotide let-7 RNA regulates developmental timing in Caenorhabditis elegans. Nature, 403, 901–906. Rowe, D. C. (2002). On genetic variation in menarche and age at first sexual intercourse. A critique of the Belsky-Draper hypothesis. Evolution and Human Behavior, 23, 365–372. http://dx.doi.org/10.1016/s1090-5138(02)00102-2. SAS 9.4 (2013). SAS Institute Inc., Cary, NC, USA. Schenker, N., & Gentleman, J. F. (2012). On judging the significance of differences by examining the overlap between confidence intervals. The American Statistician, 55, 182–186. http://dx.doi.org/10.1198/000313001317097960. Schlomer, G. L., Cleveland, H. H., Vandenbergh, D. J., Fosco, G. M., & Feinberg, M. E. (2015). Looking forward in candidate gene research: Concerns and suggestions. Journal of Marriage and Family, 77, 351–354. http://dx.doi.org/10.111/jomf.12165. Seltzer, J. A. (1991). Relationships between fathers and children who live apart: The father's role after separation. Journal of Marriage and Family, 53, 79–101. Skoog, T., & Ozdemir, S. B. (2016). Explaining why early maturing girls are more exposed to sexual harassment in early adolescence. The Journal of Early Adolescence, 36, 490–509. http://dx.doi.org/10.1177/0272431614568198. Sulem, P., Gudbjartsson, D. F., Rafnar, T., Holm, H., Olafsdottir, E. J., Olafsdottir, G. H., ... Stefansson, K. (2009). Genome-wide association study identifies sequence variants on 6q21 near LIN28B associated with age at menarche. Nature Genetics, 41, 734–738. http://dx.doi.org/10.1038/ng.383. Tanikawa, C., Okada, U., Takahashi, A., Oda, K., Kamatani, N., Kubo, M., ... Matsuda, K. (2013). Genome wide association study of age at menarche in the Japanese population. PLoS ONE, 8, e63821. Thornton, J. E., & Gregory, R. I. (2012). How does lin28 let-7 control development and disease? Trends in Cell Biology, 22, 474–482. http://dx.doi.org/10.1016/j.tcb.2012.06.001.

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002

G.L. Schlomer, H.-J. Cho / Evolution and Human Behavior xxx (2017) xxx–xxx Tither, J. M., & Ellis, B. J. (2008). Impact of fathers on daughters' age at menarche: A genetically and environmentally controlled sibling study. Developmental Psychology, 44, 1409–1420. Webster, G. D., Graber, J. A., Gesselman, A. N., Crosier, B. S., & Orozco-Schember, T. (2014). A life history theory of father absence and menarche: A meta-analysis. Evolutionary Psychology, 12, 273–294. Widaman, K. F., Helm, J. L., Castro-Schilo, L., Pluess, M., & Belsky, J. (2012). Distinguishing ordinal and disordinal interactions. Psychological Methods, 17, 615–622. http://dx.doi. org/10.1037/a0030003. Witchel, S. F. (2016). Genetics, genome-wide association studies, and menarche. Seminars in Reproductive Medicine, 34, 205–214. http://dx.doi.org/10.1055/s-0036-1585405. Wu, T., Mendola, P., & Buck, G. M. (2002). Ethnic differences in the presence of secondary sex characteristics and menarche among US girls: The third national health and nutrition examination survey. Pediatrics, 110, 752–757.

9

Xu, Z., Kaplan, N. L., & Taylor, J. A. (2007). Tag SNP selection for candidate gene association studies using HapMap and gene resequencing data. European Journal of Human Genetics, 15, 1063–1070. Zhu, H., Shah, S., Shyh-Chang, N., Shinoda, G., Einhorn, W. S., Viswanathan, S. R., ... Daely, G. Q. (2010). Lin28a transgenic mice manifest size and puberty phenotypes identified in human genetic association studies. Nature Genetics, 42, 626–630. http://dx.doi.org/ 10.1038/ng.593. Zhu, H., Shyh-Chang, N., Segre, A. V., Shinoda, G., Shah, S. P., Einhorn, W. S., ... Daley, G. Q. (2011). The Lin28/let-7 axis regulates glucose metabolism. Cell, 147, 81–94. http://dx. doi.org/10.1016/j.cell.2011.08.033.

Please cite this article as: Schlomer, G.L., & Cho, H.-J., Genetic and environmental contributions to age at menarche: Interactive effects of father absence and LIN28B, Evolution and Human Behavior (2017), http://dx.doi.org/10.1016/j.evolhumbehav.2017.06.002