Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Early-life social experience affects offspring DNA methylation and later life stress phenotype

Abstract

Studies in rodents and captive primates suggest that the early-life social environment affects future phenotype, potentially through alterations to DNA methylation. Little is known of these associations in wild animals. In a wild population of spotted hyenas, we test the hypothesis that maternal care during the first year of life and social connectedness during two periods of early development leads to differences in DNA methylation and fecal glucocorticoid metabolites (fGCMs) later in life. Here we report that although maternal care and social connectedness during the den-dependent life stage are not associated with fGCMs, greater social connectedness during the subadult den-independent life stage is associated with lower adult fGCMs. Additionally, more maternal care and social connectedness after den independence correspond with higher global (%CCGG) DNA methylation. We also note differential DNA methylation near 5 genes involved in inflammation, immune response, and aging that may link maternal care with stress phenotype.

Introduction

Early social experiences shape many aspects of an organism’s future phenotype. Over 60 years ago, experiments with laboratory rodents and non-human primates revealed that maternal care and interactions with peers have lasting effects on an organism’s stress physiology and behavior1,2. In rodents, lower rates of maternal licking and grooming during the offspring’s first 10 days of life caused elevated adult plasma corticosterone in response to external stressors3 and exhibition of fearful behaviors4,5. Beyond the beneficial effect of maternal care, interactions with group members were protective against an elevated corticosterone response to a standardized stressor in Sprague–Dawley rats6. Similarly, the physiological toll of maternal separation during early life later manifested as a flat cortisol trajectory in rhesus macaques7. The importance of early social experiences extends to our own species. For example, children who lived in orphanages had altered hypothalamic-pituitary-adrenal (HPA) activity, including elevated basal cortisol levels that remained evident even after adoption8. This body of literature emphasizes the importance of maternal care and early social interactions in the ontogenetic development of stress phenotypes. Yet, the underlying biological mechanisms remain unclear.

DNA methylation is a mitotically stable epigenetic mark that is responsive to environmental cues and is involved in the regulation of gene expression9. It has been championed as a potential biological mechanism linking early social experiences to later stress phenotype. A landmark cross-fostering study in rodents showed that lower rates of maternal licking and grooming corresponded with higher DNA methylation of the promoter region of the hippocampal glucocorticoid receptor (GR; NR3C1) gene, lower GR RNA expression, and elevated plasma corticosterone among adult offspring10. Subsequent work in rodents demonstrated that maternal separation is correlated with widespread differences in brain tissue DNA methylation, not only at specific promoter regions but also in the form of lower global DNA methylation11. In rhesus macaques, maternal vs. peer-led rearing corresponded with genome-wide changes in DNA methylation in brain tissue and T cells during adulthood12. Finally, human epidemiological studies show that lower quality of maternal care corresponds with higher blood leukocyte DNA methylation of brain-derived neurotrophic factor (BDNF) and oxytocin receptor (OXTR) in adulthood13.

Although there is a large body of literature on the topic of early social experience as a determinant of an organism’s future stress phenotype and DNA methylation as a potential mediator, we identify three major lacunae. First, there is a lack of studies that measure natural variation in the quantity or quality of early social experience, as much of this research includes experimental studies involving maternal separation and peer isolation, which are routine procedures in studies of captive primates and rodents14,15. While informative, experiences of extreme deprivation do not capture the range of normative social interactions that are relevant to development3,16,17. Second, although studies have examined the relationships among social experiences, DNA methylation, and stress phenotype in a piecemeal fashion, few have measured all three components in the same population10,18,19,20. Doing so is necessary to explicitly test the hypothesis that DNA methylation represents a mechanistic link between early social experience and future stress phenotype. Third, few studies explicitly consider that variation in the type and/or timing of social experiences may affect DNA methylation21 and stress phenotypes22, particularly in wild animal populations where development is subject to environmental fluctuation and pressures of natural selection.

We address these three gaps in the literature by examining the relationships among early social experience, DNA methylation, and stress phenotype in a long-term field study of wild spotted hyenas (Crocuta crocuta). Based on earlier work in primates and rodents, our primary hypothesis is that unfavorable early social experiences correspond with adverse stress physiology later in life, as indicated by higher fecal glucocorticoid concentrations. Second, we hypothesize that these early social experiences influence patterns of DNA methylation measured in subadult and adult hyenas. Third, we hypothesize that differential DNA methylation is on the causal pathway, and thus a potential mechanism linking early social experience to future stress physiology.

Here, we show that maternal care and social connectedness during the earlier den-dependent life stage are not associated with fecal glucocorticoid metabolites (fGCMs; an indicator of stress phenotype). However, greater social connectedness during the later subadult den-independent life stage is associated with lower adult fGCMs. In addition, more maternal care and social connectedness after den independence corresponds with higher global (%CCGG) DNA methylation, a marker of genomic stability and overall health. We also note differential DNA methylation near five genes involved in inflammation, immune response, and aging that may link maternal care with stress phenotype. Our findings suggest that both maternal care during the first year of life and social connections after leaving the den influence DNA methylation and contribute to a developmentally plastic stress response.

Results

Study population characteristics and results from bivariate analyses are provided in the Supplementary Material.

Part 1: associations among early social environment and adult fGCMs

Among hyena cubs, none of the maternal care or social connectivity measures during the communal den (CD) period were associated with adult fGCMs levels in the adjusted models (Table 1). However, during the later den-independent (DI) period of development, higher network degree, and strength were each associated with lower adult fGCMs after adjusting for a hyena’s sex, level of human disturbance in its birth year, age, reproductive state, and the time of day of sample collection (−11% [95% confidence interval (CI): −20, −2] and −13% [95% CI: −23, −3] for network degree and strength, respectively; Table 1).

Table 1 Association of maternal care and early-life social network metrics with adult fGCMs.

Part 2: associations between early social environment and global DNA methylation

We observed multiple positive associations between early social experiences and %CCGG methylation that were robust to covariate adjustment. In our demographic covariates model, which adjusted for offspring sex and age, we found that each 1-SD greater proportion of time spent in close proximity to the mother corresponded with 1.36 (95% CI: 0.72, 2.03) greater %CCGG methylation in cub and subadult offspring. The association was robust to adjustment in the social experience covariates model (1.28 [95% CI: 0.34, 2.20] %CCGG), and the ecological covariates model (1.07 [95% CI: 0.34, 1.83] %CCGG). In the model that included all covariates, there was a slight attenuation of the estimate, and the confidence interval narrowly crossed the null (0.90 [95% CI: −0.04, 1.86] %CCGG; Fig. 1). We also found that more time spent nursing was associated with higher %CCGG methylation in models that accounted for the demographic covariates (0.86 [95% CI: 0.14, 1.60] %CCGG) and social experience (1.12 [95% CI: 0.10, 2.08] %CCGG) covariates. However, adjustment for ecological covariates (human disturbance and migration status in the birth year) attenuated the effect of nursing time on %CCGG methylation. There was no effect of proportion of time spent grooming on %CCGG methylation (Fig. 1), and maternal care and maternal rank interactions were not significant.

Fig. 1: Associations of maternal care behaviors with cub and subadult offspring %CCGG methylation among N = 55 hyenas.
figure1

All models included a random intercept for maternal ID to account for relatedness and the offspring’s age in months at the time of blood sample collection for DNA methylation quantification. Models were further adjusted according to four types of covariates (demographic covariates [dark blue], early social experience covariates [dark green], ecological covariates [light blue], and a fully adjusted all covariates model [light green]). 95% CIs were based on percentile bootstrapping (2000 iterations). Source data are provided as a Source Data file.

Next, we assessed relationships between social network metrics derived during the hyenas’ CD and DI periods of development and %CCGG methylation. During the CD period, strength (0.75 [95% CI: 0.01, 1.48]) and betweenness (0.83 [95% CI: 0.10, 1.58]) were each positively associated with %CCGG methylation in demographic covariates models (Fig. 2). These relationships were attenuated after adjusting for early social experience and ecological covariates. Network degree was not associated with %CCGG methylation in any of our models during the CD period and none of the interaction terms involving social network metrics and maternal social rank were significant.

Fig. 2: Associations of hyena CD social network metrics with cub and subadult offspring %CCGG methylation among N = 81 hyenas.
figure2

All models included a random intercept for maternal ID to account for relatedness, and the offspring’s age in months at the time of blood sample collection for DNA methylation quantification. Models were further adjusted according to four types of covariates (demographic covariates [dark blue], early social experience covariates [dark green], ecological covariates [light blue], and a fully adjusted all covariates model [light green]). 95% CIs were based on percentile bootstrapping (2000 iterations). Source data are provided as a Source Data file.

During the DI period, network degree was positively associated with %CCGG. Each 1-SD increment in degree corresponded with 0.69 (95% CI: 0.03, 1.38) higher %CCGG methylation in the demographic covariates model. This association was strengthened after accounting for ecological covariates (1.04 [95% CI: 0.25, 1.86] %CCGG methylation), and also in the fully adjusted model that included all covariates (1.19 [95% CI: 0.24, 2.13] %CCGG methylation). Estimates for network degree from the social experience model included the null (0.79 [95% CI: −0.09, 1.69] %CCGG methylation, Fig. 3), although the direction and magnitude of estimates were similar to that of other models. Greater network strength was associated with 0.71 (95% CI: 0.03, 1.43) higher %CCGG methylation in the demographic covariates model. This estimate was materially unchanged after accounting for ecological covariates (0.68 [95% CI: −0.05, 1.40] %CCGG methylation) and adjustment for all covariates simultaneously (0.83 [95% CI: −0.03, 1.72] %CCGG methylation), although we note that the 95% CIs were slightly wider in these models and narrowly crossed the null (Fig. 3). Adjustment for social experience covariates attenuated the estimate for network strength towards the null (0.49 [95% CI: −0.36, 1.33] %CCGG methylation). Betweenness was not associated with %CCGG methylation, and there were no significant interactions between DI network metrics and maternal rank.

Fig. 3: Associations of hyena DI social network metrics with cub and subadult offspring %CCGG methylation among N = 81 hyenas.
figure3

All models included a random intercept for maternal ID to account for relatedness, and the offspring’s age in months at the time of blood sample collection for DNA methylation quantification. Models were further adjusted according to four types of covariates (demographic covariates [dark blue], early social experience covariates [dark green], ecological covariates [light blue], and a fully adjusted all covariates model [light green]). 95% CIs were based on percentile bootstrapping (2000 iterations). Source data are provided as a Source Data file.

Finally, offspring %CCGG methylation was not associated with adult fGCMs.

Part 3: formal mediation analysis

In the subset of hyenas for which we had data on maternal care (n = 30 total; n = 17 females and n = 13 males) or social network metrics (n = 52 total; n = 32 females and n = 20 males), as well as %CCGG DNA methylation and adult fGCMs, we conducted a three-step mediation analyses. We found no association between our explanatory variables (maternal care, CD social connectedness, DI social connectedness) and the outcome, fGCMs (Supplementary Table 7). In the second step of the mediation analyses, we found no evidence that social experiences (the explanatory variables) were associated with %CCGG methylation (the mediator), or that %CCGG methylation was associated with fGCMs (the outcome variable, Supplementary Table 7). Given the lack of associations in the first two steps, we did not perform the final step of the mediation analysis.

Part 4: genome-wide (multiplexed enhanced reduced representation bisulfite sequencing (mERRBS)) DNA methylation analyses

Our epigenome-wide association study (EWAS) resulted in 18 differentially methylated sites (DMSs) that were associated with fGCMs after false discovery rate (FDR) correction (Table 2 and Supplementary Fig. 10). While we did not observe systemic bias in the enrichment of positive or negative associations from our EWAS, within particular genomic regions, including CpG islands, promoters and introns there appeared to be fewer Beta estimates near zero than compared to all CpG sites across the whole genome (Supplementary Fig. 11). Based on the CpG density annotation, we observed that 62% of CpG sites from our EWAS occurred in CpG islands and CpG shores, while among both positive and negative DMSs, 72% of sites were in CpG islands and shores (Supplementary Fig. 12a, c). According to our genic annotation, 38% of CpG sites from our EWAS occurred in either promoters, exons, or introns, and 39% of DMSs were found in the same genic annotations (Supplementary Fig. 12b, d). There were equal total numbers of positive and negative DMSs, but in both annotations more positive DMSs were found near CpG islands and gene bodies than in intergenic regions.

Table 2 Epigenome-wide association results showing DMSs associated with fGCMs BLUPs among N = 25 hyenas.

Among the 18 CpG sites identified in the EWAS, we examined the relationships between maternal care metrics and maternal rank with DNA methylation and identified six CpG sites that were associated with maternal care. Four of these CpG sites exhibited positive Beta estimates from fGCMs models and corresponding negative Beta estimates for maternal grooming, while DNA methylation at one CpG site was negatively associated with fGCMs and positively associated with maternal grooming. At a CpG site located at the human teashirt zinc finger homeobox 2 (TSHZ2) gene23,24, the proportion of time that cubs were groomed by their mothers was positively associated with DNA methylation (P = 0.074; Table 3), and fGCMs were negatively associated with DNA methylation (P = 0.031; Table 2). We observed a negative association of time spent grooming (P < 0.001; Table 3) and a positive association of fGCMs (P = 0.026; Table 2) with respect to DNA methylation at a CpG that aligned at the tsukushi, small leucine-rich proteoglycan (TSKU) gene in humans23,24. We also observed a negative association for time spent grooming (P = 0.097; Table 3) and a positive association of fGCMs (P = 0.045; Table 2) with DNA methylation at a CpG that aligned at the RAB43, member RAS oncogene family (RAB43) and the ISY1 splicing factor homolog (ISY1) gene complex in humans23,24. Maternal grooming was negatively associated with DNA methylation (P = 0.048; Table 3) and fGCMs were positively associated with DNA methylation (P = 0.007; Table 2) at a CpG site at the human adenosine deaminase (ADA) gene23,24. Similarly, we observed a negative association for time spent grooming (P = 0.018; Table 3) and a positive association for fGCMs (P = 0.026; Table 2) with DNA methylation at a CpG site at zinc finger and SCAN domain containing 2 (ZSCAN2) gene in humans23,24.

Table 3 Associations of DMSs by maternal rank and maternal care variables among the 18 DMSs identified in the stress phenotype EWAS for female spotted hyenas that have (maternal rank data (n = 23) and maternal care data (n = 9).

In addition, one CpG site near the human DENN domain containing 3 (DENND3) gene23,24 was associated with the proportion of time spent in close proximity to the mother hyena (P = 0.042; Table 3) and fGCMs (P = 0.032; Table 2), but the Beta estimates were in the same direction, and thus did not align with our criteria for a plausible mechanism or biomarker of a mechanism linking early social environment to future stress phenotype. Neither maternal rank during the year in which the offspring was born nor the proportion of time spent nursing was associated with the stress-related DMSs identified in this sample of female hyenas (Table 3).

Candidate gene DNA methylation: putative GR promoter

Among 78 hyenas that passed the pyrosequencing quality checks (Pyromark Q96 MD instrument), preliminary data revealed invariant and near-zero percent methylation at six CpG sites, including the putative NGFI-A transcription factor binding site (Supplementary Fig. 13). Due to lack of variation in DNA methylation at these CpG sites, we did not conduct any additional analyses with respect to this region

Discussion

Leveraging ecological and behavioral data from a long-term prospective study of wild hyenas, we investigated the extent to which social experiences during two early developmental windows are associated with DNA methylation and adult stress phenotype. We found beneficial effects of social connectedness during the DI period of development, when young hyenas no longer shelter in dens, on adult stress phenotype. We also found that higher-quality maternal care in the first year of life, which roughly corresponds with the CD period of development, and greater social connectedness during the DI period were each associated with higher global DNA methylation. Although we found no evidence that the relationship between social experience and stress phenotype is mediated by global DNA methylation, our genome-wide DNA methylation analysis identified multiple DMSs associated with both receipt of maternal care and adult fGCMs, and thus might serve as mechanisms or biomarkers of mechanisms for future investigation.

The early social environment and adult stress phenotype

Higher social connectedness is associated with lower fGCMs

Greater social connectedness—specifically, higher degree centrality and strength—during the DI period, when hyenas have left the CD, corresponded with lower adult fGCMs. This supports the hypothesis that social connections promote a healthy stress response and suggests that connections with group members are particularly important among subadult animals as they expand their social network beyond a limited number of associations with group members at the CD.

Our findings corroborate previous work in humans, rodents, and baboons. For example, in young adults, Kornienko et al. showed that more gregarious individuals have lower salivary cortisol levels25,26. Similarly, Ponzi et al. reported that a children’s perception of their social connections (i.e., higher density of friendships) is inversely related to salivary cortisol levels, as well as salivary alpha-amylase reactivity, an indicator of reactivity to psychological stressors27. In Sprague–Dawley rats, postweaning rearing in social isolation corresponded with higher corticosterone concentrations than in rats reared with other group members6. The importance of social networks has also been observed in wild baboons, for which membership in more close-knit and focused grooming networks are associated with lower fGCMs28. However, it is worth noting that social connections are not universally associated with stress phenotypes. For instance, a study on wild yellow-bellied marmots found no association of social network metrics with fGCMs measured in a mixed-age, mixed-sex population29, although it should be noted that this species is not obligately social like those in other above-mentioned studies.

Maternal care is not associated with fGCMs

Contrary to our predictions and to findings from rodents, non-human primates, and humans3,4,7,30,31, we found no evidence that maternal care affects future stress phenotype. There are multiple possible explanations for our null findings. First, in contrast to controlled rodent experiments, wild hyenas are subject to a multitude of stressors over development, which may hamper our ability to isolate the specific effect of maternal care on the stress phenotype. Second, compared to maternal separation contrived in many experimental studies, natural variation in maternal care is far more subtle and may require a more sensitive measure of physiological stress than average fGCMs, which is a summary baseline stress indictor32 subject to unmeasured environmental factors and the animal’s condition when the sample is collected33. Considering not only the discrepancies in how early-life experience and stress outcomes are measured but also differences between controlled laboratory settings vs. studies of wild animals, we urge caution when comparing our results to previous studies involving captive primates and rodents.

The early social environment and global DNA methylation

Maternal care and social connectivity are associated with global DNA methylation

Greater maternal care (specifically, close maternal proximity during the first year of life) and higher social connectivity (namely, degree during the DI life stage) were both associated with higher global DNA methylation later in life. These results corroborate those of previous studies that report associations of early social experiences with differential patterns of DNA methylation in rodents, rhesus macaques, and humans11,12,13. Specifically, offspring that spent more time in close proximity to their mothers during the first year of life had higher global DNA methylation later in life, even after adjustment for key demographic, social, and ecological covariates. We also found that more time spent nursing corresponded to higher global DNA methylation after adjusting for demographic and social experience covariates, although the effect of nursing was attenuated when we included the early-life ecological covariates in the model. This suggests that ecological context may account for some variation in the relationship between time spent nursing and global DNA methylation, which is plausible given previously reported associations in hyenas between maternal care behavior as well global DNA methylation with human disturbance and prey availability34,35. One potential explanation for the robust effect of close proximity, but not the other metrics of maternal care (nursing and grooming), on global DNA methylation may reflect the fact that close proximity includes the time mother hyenas spent nursing and grooming. Thus, close proximity is a composite metric of maternal investment in offspring that may capture more variation in offspring global DNA methylation.

We also found that higher social connectedness during the DI, but not CD, period of development was positively associated with global DNA methylation. Specifically, the degree of a hyena’s social connectedness, which reflects the number of group members with which a hyena associates, corresponded with higher global DNA methylation even after adjustment for demographic and ecological covariates. Controlling for early-life social factors, including the number of littermates, group size, and maternal rank slightly attenuated the effect of social connections towards the null, which is not surprising given that these variables are known determinants of a young hyena’s social network36 as well as of global DNA methylation34.

Our findings that less maternal care during the first year of life and less social connectedness during the DI period of development are associated with lower global DNA methylation suggest that inadequate early social experiences may be risk factors for suboptimal health later in life. Previous human research has shown that lower global DNA methylation corresponds with adverse health outcomes related to cardiovascular disease37, pulmonary disease38, shorter telomere length (a marker of accelerated aging) in both adolescents39 and adults40, and cancer41. It is worth noting that higher global DNA methylation has also been associated with some cardiovascular diseases42. Thus, deviations in either direction, both lower and higher DNA methylation, may be a marker of aberrant physiology and should be interpreted with caution. Finally, the effect sizes for %CCGG methylation we observed are modest but biologically relevant, as they are comparable in magnitude to the effect sizes of dietary43 and toxicant44 exposures reported in the existing literature.

Mediation by global DNA methylation

No evidence of mediation by global DNA methylation

We found no evidence of mediation by %CCGG methylation. This could be due to low power given the relatively small sample sizes, and/or related to the fact that %CCGG methylation averaged over the entire genome may be too broad a metric to isolate important regulatory pathways involved in the stress response. For example, in humans, we previously found that maternal prenatal socioeconomic status was associated with differential methylation at specific regions of the genome, but not with global DNA methylation in humans45. Therefore, we complemented our global DNA methylation analysis in hyenas with mERRBS46, a high-throughput sequencing technology that assays single-nucleotide DNA methylation resolution at a genome-wide scale.

Biological interpretation of CpG sites identified in genome-wide DNA methylation assessment

Using mERRBS, we identified 18 DMSs that were associated with fGCMs after FDR correction, six of which were also associated with early-life maternal care metrics. Five of these DMSs followed our prediction given that greater maternal care is hypothesized to be protective against an adverse stress phenotype. More specifically, for a given DMS, a positive association between DNA methylation and stress hormone levels should correspond to a negative association between maternal care and DNA methylation at that DMS, or vice versa. We discuss the direction of the associations of these five DMSs with respect to fGCMs and early-life maternal care metrics below, as well as their potential biological function.

DNA methylation of a CpG site that mapped near the human TSHZ2 gene was positively associated with receipt of maternal grooming and inversely associated with fGCMs. This transcription factor is involved in regulating gene expression and plays a role in development in multicellular organisms47,48,49,50. TSHZ2 is expressed during early development of the nervous system in diverse vertebrate taxa51,52. The TSHZ2 gene is also associated with a number of pathophysiological states, including inflammatory disease53, dysregulated sleep patterns54, and tumorigenesis55,56 via hypomethylation of TSHZ2 promoter in cancerous epithelia cell lines57. Second, we identified a CpG site that mapped near the human TSKU gene which was negatively associated with receipt of maternal grooming and positively associated with fGCMs. TSKU is part of gene ontogeny pathways involved in gene expression and development as well as differentiation of the ectoderm germ layer, wound healing, Wnt signaling, and cholesterol processing47,48,49,50. Expression of TSKU has been implicated in renal disease through fibrosis and inflammation58. The third CpG site, which was negatively associated with grooming and positively associated with fGCMs, mapped near the RAB43 and ISY1 genes. RAB43 plays a role in phagosome maturation, response to interferon-gamma, Golgi organization, and intracellular protein transport, while ISY1 contributes to messenger RNA (mRNA) processing and splicing as well as transcription-coupled nucleotide-excision repair47,48,49,50. As a regulator of cellular trafficking and signaling, there is evidence that RAB43 helps orchestrate the immune response to infection59,60. More recent work also implicates overexpression of this gene in the progression of multiple cancers61,62. The fourth CpG site exhibited a negative association with grooming and a positive association with fGCMs, and it mapped to the ADA gene, which is involved in T and B cell regulation, inflammation, aging, embryonic development, and organic molecule metabolism47,48,49,50. Research in humans and rodent models of human health suggests that insufficient expression of ADA gene leads to immunodeficiency63,64,65 and ADA activity has been identified as an indicator of aging66,67,68,69. Of particular interest to our study, corticosterone has been shown to inhibit ADA activity in mice70, suggesting that regulation of this gene may link early-life experience with stress phenotype and aging. Finally, DNA methylation of a CpG site near the ZSCAN2 gene was negatively associated with receipt of grooming and positively associated with fGCMs. ZSCAN2 is involved in the regulation of transcription and cell differentiation, particularly during development47,48,49,50, but we found no reported pathologies associated with this gene. Despite having only correlative evidence from our EWAS, the above molecular biomarkers in hyenas share sequence similarities to human genes. Furthermore, these genes have been implicated in basic biological processes related to a range of chronic diseases including inflammation, immune function, tumorigenesis, and aging, suggesting that these pathways may be particularly important for understanding the relationship between early-life maternal care and later-life stress phenotypes.

No evidence of DNA methylation variation at the putative hyena GR

Although some previous work in rodents and humans has shown that inadequate maternal care and greater early adversity corresponded with higher methylation of CpG sites in the GR promoter region10,71, this relationship between maternal care and GR DNA methylation has not been replicated in all studies72. In the present study, we measured DNA methylation at CpG sites in the putative hyena GR gene promoter and found invariant and near-zero percent methylation, suggesting a potential lack of plasticity in DNA methylation in this promoter region, at least when using peripheral leukocytes. Future studies may better replicate results from laboratory rodents by using brain tissue, wherein expression of the GR gene is more tightly coupled with HPA regulation, or by exploring DNA methylation of additional CpGs with more extensive coverage of the hyena GR gene.

Limitations of this study

Our study has several limitations. First, we used a large database of existing behavioral, demographic, and biological variables to test our hypotheses. As a result, we lacked complete data overlap between different variables of interest resulting in somewhat differing sample sizes across analyses (thereby hindering our ability to make direct comparisons across analytical subsamples). Further, the observational nature of the data limits causal inference due to the potential for unmeasured confounding. Next, our genome-wide analysis was limited by budgetary constraints, so sample selection focused on animals of similar age and sex for the sake of statistical power. Consequently, results of this particular analysis may not be generalizable to both sexes and preclude between sex comparisons. We acknowledge the tradeoff between improved statistical power and internal validity at the expense of external validity of results given the opportunistic nature of this analysis. Second, we used the average of multiple measures of fGCMs as an indicator of the animals’ stress phenotype. However, a more nuanced measure of stress, like reaction and recovery to a standardized stressor or the ratio of cortisol to dehydroepiandrosterone, may provide a more direct assessment of acute HPA function73,74 over time. Third, our mediation and mERRBS analyses comprised small sample sizes and therefore may have low external validity—particularly for the mERRBS analysis, which included only females. Fourth, we used archived DNA from whole blood that lacked information on cell type composition, and therefore we were not able to control for cell-type heterogeneity as a source of variation in DNA methylation. However, cell-type composition may be influenced by factors like social stress75, and therefore might be on the causal pathway between the explanatory variables of interest and DNA methylation. Future analyses that aim to test hypotheses about functional genomic pathways involving a larger sample of genome-wide DNA methylation data are warranted. With larger sample sizes and additional annotation of the hyena genome, future studies could home in on specific genomic contexts containing differentially methylated CpG sites and identify pathways enriched in response to early-life social experience. Finally, we recommend future studies incorporate low coverage genome-wide genetic sequence data as a more accurate alternative to a pedigree to control for genetic correlations in DNA methylation.

In conclusion, in a wild population of spotted hyenas, maternal care and social connections early in life were key determinants of global DNA methylation and fGCMs assessed in adulthood, even after accounting for key demographic, biological, and social factors. These findings contribute to the literature in multiple ways. First, few studies have been able to assess early social experience, DNA methylation, and future stress phenotype together, a study design that allows for explicit assessment of DNA methylation as a potential mediator. Second, existing literature has focused predominantly on singular social experiences during specific developmental windows, which is not only unrealistic given the multitude of experiences across early life that may contribute to stress phenotypes but also precludes the ability to assess the relative importance of the type and timing of experiences. In the present study, we were able to examine prospective associations of multiple aspects of social experience during two sensitive windows of development.

Our work on wild hyenas joins a growing body of evidence, predominantly from rodents and primates, showing the critical effects of early experience on molecular biomarkers and future stress phenotype. Given that such studies are seldom possible in free-living populations of long-lived social animals, this study provides insight into the type, timing, and mechanisms that link early-life experiences with the developmental plasticity of stress phenotypes and highlights the extent to which natural variation in maternal care and social interactions have persistent effects on developing offspring.

Methods

Study population

We used behavioral data and biological samples collected between June 1988 and July 2016 by the Mara Hyena Project, an ongoing field study of wild spotted hyenas (C. crocuta) in the Masai Mara National Reserve, Kenya. All protocols used for the present study have been approved by Michigan State University (MSU) Institutional Animal Care and Use Committee (IACUC) and the Kenyan Wildlife Service (KWS). For each hyena in our study population, we have information on demographic, social, and ecological conditions throughout their lifetimes. Using blood samples from opportunistically immobilized hyenas paired with archived behavioral and physiological data, we constructed three primary datasets for analyses: global DNA methylation (n = 186 total; n = 99 females and n = 87 males), genome-wide DNA methylation (n = 29 total; n = 29 females and n = 0 males), and candidate gene DNA methylation (n = 78 total; n = 43 females and n = 35 males).

Spotted hyenas are an appropriate species in which to test our hypotheses because they exhibit a wide range of social behaviors, live in large fission–fusion clans that can contain more than 100 individuals, and show a protracted period of maternal care76. Female hyenas reach reproductive maturity around 2 years of age and give birth to one or two offspring every 14–17 months starting in their third year of life77. Following a 110-day gestation period, hyena cubs spend the first few weeks of life interacting exclusively with their mothers at a natal den78. During development, young hyenas rely on their mothers for sustenance and social support until they are at least 2 years of age78. Young hyenas also socialize with other members of their clan during an earlier life stage when they reside at their clan’s CD—known as the CD period, followed by a later, DI period when hyenas venture out into their clan’s territory and develop relationships with other group members36,79. In order to balance our sampling design, we matched the duration of the DI period with that of the CD period (mean length of both periods was 6.98 ± 1.74 months)80 for each hyena.

Early social experiences

We derived the maternal care variables from focal animal survey (FAS) data collected during observation sessions in which: (1) mother–offspring pairs were present together for a minimum of 5 min and offspring were <13 months old and (2) mothers were lactating since our intention was to focus on maternal care received early in life while offspring were dependent on nursing for sustenance. We quantified durations of maternal care behaviors from FAS data81 based on counts of behaviors occurring during each minute of observation in which both the mother and offspring were present together. Behavioral data were collected daily between roughly 0530–0900 and 1700–2000 hours. We focused on three maternal care behaviors: minutes the mother and cub spent in close proximity (≤1 m apart), minutes offspring spent latched to their mother’s nipple (nursing), and minutes during which mothers were observed grooming (i.e., licking) their offspring. Information on additional behavioral data processing appears in the Supplementary Material.

We measured social connectivity by generating association networks among hyenas based on co-occurrences between each hyena and its group members36 as described in the Supplementary Material. For each hyena, we constructed separate association networks during the CD and DI periods (two life stages during which social interactions were previously identified as key determinants of fitness80).

From each hyena’s association networks during its CD and DI periods, we extracted three metrics that quantify how connected a hyena is with its group members: degree centrality, strength, and betweenness centrality. We focused on these metrics, rather than the overall network structure, because they reflect an individual hyena’s connectedness within its network. A description of social network methods has been published in ref. 36 and is summarized in the Supplementary Material.

DNA methylation: global (%CCGG), genome-wide, and candidate gene

We quantified global (%CCGG) DNA methylation derived from whole blood using the LUminometric Methylation Assay (LUMA)82. Descriptions of the LUMA assay, laboratory procedures, and data cleaning protocol are available in ref. 34, from which the current global DNA methylation samples were drawn. The majority of CpG sites assessed via LUMA occur in intergenic regions of the genome, where they may repress repetitive elements83 and enhance chromosome stability84, as well as introns, where they may function in transcription regulation and alternative splicing85 (Supplementary Fig. 8). Coupled with evidence that global DNA methylation is responsive to early-life environmental factors86,87,88, and the fact that lower global DNA methylation is associated with a range of adverse health outcomes in general-risk human populations (e.g., shorter telomere length as a metric of accelerated aging39,40; hypertension37; and chronic obstructive pulmonary disease38), we interpret average %CCGG methylation as a relevant biomarker of developmental plasticity where lower global DNA methylation is disadvantageous to health.

We measured genome-wide DNA methylation at a single-nucleotide resolution in whole blood collected from hyenas 11–27 months old. To prepare the mERRBS sample library, we followed the protocol of Garrett-Bakelman et al.89 using 100 ng of high-quality genomic DNA. Each sample was spiked with 1 ng of non-methylated lambda guide DNA to estimate bisulfite conversion efficiency. Five libraries were pooled per lane of an Illumina HiSeq4000® platform for single-end sequencing with a 50-nucleotide read length. After DNA sequencing, we ran a standard bioinformatic pipeline to clean, align, and call DNA methylation reads using the draft spotted hyena genome90. We filtered read counts to a minimum of 10× coverage. Details on DNA library preparation and the bioinformatics pipeline are given in the Supplementary Material.

Given the extensive literature on the relevance of the GR gene to both early social experiences and stress phenotypes, we also assessed CpG methylation in the putative GR promoter region of DNA from hyenas. Candidate gene bioinformatics and laboratory methods are described in the Supplementary Material, and a list of all primers used in the candidate gene analysis is provided in Supplementary Table 2. We identified CpG sites in hyena DNA that aligned with those in DNA from humans and rats (Supplementary Figs. 25; refs. 91,92). We calculated CpG site-specific DNA methylation in the putative hyena GR promoter region using pyrosequencing.

Fecal glucocorticoid metabolites

Since January 1993, we have opportunistically collected fecal samples any time an individually identifiable hyena was seen defecating. Fecal samples were mixed and transferred to 2 mL cryovials before flash freezing in liquid nitrogen within 12 h of collection. The frozen samples were then transported from our field site to the United States. Here, we focus on the hormone corticosterone, as indicated by the concentration of fGCMs measured via a validated hormone extraction process and radioimmunoassay developed for our study population93,94.

Demographic, social experience, and ecological covariates

We considered three categories of potential confounding variables (i.e., those that are associated with the explanatory variable but not caused by the explanatory variable and potential determinants of the dependent variable) and included them in multiple variable models to improve causal inference. Sex was a demographic confounder; maternal rank, litter size, parity, and clan size were social experience confounders; and human disturbance and local prey abundance were ecological confounders. Prey abundance was defined as either high or low, respectively, based on the annual mass migration of wildebeest and zebran the study area. The Supplementary Material provides details on data collection germane to these variables.

Statistical modeling framework

We analyzed data in four distinct parts in accordance with our goals and using a methodical approach to make causal inferences from observational data95. First, we characterized the relationship between early social experience according to maternal care and social network metrics (explanatory variables) and adult stress phenotype as indicated by adult fGCMs (dependent variable). Second, we examined associations between early social experience and global DNA methylation, and between global DNA methylation and fGCMs. Third, given that global DNA methylation may be a potential mechanism (in addition to an outcome), we conducted a formal mediation analysis following Baron and Kenny’s method (1986) on a subset of hyenas (n = 30 maternal care and n = 52 social networks) for which we had data on early social experience, global DNA methylation, and adult fGCMs (Fig. 7). Fourth, to complement the third objective, we used genome-wide DNA methylation data to identify potential functional biomarkers that might link early-life experiences with future stress phenotypes and that might be formally assessed as mediators in future analyses. In all models, we considered α = 0.05 as the threshold for statistical significance, unless otherwise indicated.

Part 1: associations among early social environment, DNA methylation, and adult fGCMs

First, we examined each maternal care variable (proportion of time in close proximity, nursing, and grooming) and social network metric (degree centrality, strength, and betweenness centrality during the CD and DI periods) as explanatory variables in relation to fGCMs as the dependent variable using mixed-effects linear regression with a random intercept for hyena ID to account for correlations between repeated measures of fGCMs. In these models, we included covariates that are associated with the explanatory variable and are potential determinants of the dependent variable (confounders), as well as factors that account for variability in the exposure only or the outcome only (precision covariates). Confounders included offspring sex and level of human disturbance during the year in which offspring were born (low, medium, and high). Precision covariates included hyena’s age, reproductive state for females (nulliparous, pregnant, lactating, or other), and time of day (a.m. or p.m.) when the fecal sample was collected (Fig. 4).

Fig. 4: Directed acyclic graph (DAG) for Part 1 of the analysis showing the hypothesized relationships between early-life social experience and adult fecal glucocorticoid metabolites (fGCMs).
figure4

Covariates were included based on a priori biological knowledge and bivariate analysis (see Supplementary Tables 36). Precision covariates (gray text) were assessed when the fecal sample was collected. Variable positions over the timeline roughly corresponded with the timing of the assessment.

Part 2: associations between early social environment and global DNA methylation, and between global DNA methylation and fGCMs

First, we modeled the total effect each maternal behavior and social network metric had as separate explanatory variables with global DNA methylation (%CCGG methylation) as a continuous dependent variable, again using mixed-effects linear regression models with a random intercept for maternal ID to account for correlations among siblings, and included the covariates depicted in Fig. 5. In multiple variable analysis, we controlled for offspring age in months as a precision covariate in all models because age is a known determinant of DNA methylation in this and other species34. In models where a maternal care behavior was the explanatory variable, we further accounted for three categories of covariates using separate models: demographic covariates (offspring sex), social experience covariates (number of littermates, mother’s parity, and mother’s social rank), ecological covariates (human disturbance and prey abundance on the date offspring were born), and a fully adjusted model that included all covariates. When CD or DI social network metrics were the explanatory variables of interest, we controlled for the same confounders, except we included clan size rather than the mother’s parity in the social experience covariates model. Given the importance of maternal rank to offspring DNA methylation34, and the potential for maternal rank to mitigate or exacerbate effects of mother–offspring interactions35,96, we tested for statistical interactions between the explanatory variables of interest (maternal care and social connectedness) and maternal rank in models where %CCGG methylation was the outcome. We considered α = 0.10 for the interaction term as evidence of effect modification by maternal rank.

Fig. 5: Directed acyclic graph (DAG) for Part 2 of the analysis showing the hypothesized relationships between early-life social experiences and %CCGG methylation.
figure5

Potential confounding variables were modeled in three different context groups, which included demographic covariates (dark blue), early social experience covariates (dark green), and ecological covariates (light blue). *Maternal rank was considered a potential effect modifier of the associations of interest. Age at darting was included as a precision covariate. Variable positions over the timeline roughly corresponded with the timing of the assessment.

Next, we assessed the relationship between global DNA methylation and adult fGCMs (Fig. 6). We used mixed-effects linear regression with a random effect for individual ID to account for repeated fGCM measurements and controlled for sex and human disturbance during the year offspring were born. Precision covariates included the hyena’s age, reproductive state, and time of day when the fecal sample was collected.

Fig. 6: Directed acyclic graph (DAG) for Part 2 of the analysis showing the hypothesized relationships between %CCGG methylation with adult fecal Glucocorticoid Metabolites (fGCMs).
figure6

Covariates are included based on a priori biological knowledge and bivariate analysis (see Supplementary Tables 36). Precision covariates (gray text) are assessed when the fecal sample was collected. Variable positions over the timeline roughly correspond with the timing of the assessment.

Of note, one of our aims was to examine associations with DNA methylation of the GR promoter region among 78 hyenas with adequate high-quality DNA. However, our assays revealed invariant and near-zero percent methylation at six CpG sites, including the putative NGFI-A transcription factor binding site, so no additional analyses were performed with respect to this region.

Part 3: formal mediation analysis

The third part of our analysis assessed the extent to which maternal care and social network metrics during CD and subadult DI life stages are associated with adult fGCMs before and after adjustment for global DNA methylation (%CCGG) as a mediator. Here, we included the subsample of hyenas with complete data on early social experience, DNA methylation, and stress phenotype. Using the complete data subsample, we followed Baron and Kenny’s mediation procedure (1986) and re-implemented the Parts 1 and 2 from above, which correspond to Steps 1 and 2 in a mediation analysis (Fig. 7). If associations from Steps 1 and 2 were significant, then we conducted Step 3 where we included %CCGG as an additional covariate in models in which each early-life social variable was the explanatory variable and fGCMs was the outcome (Step 3; Fig. 7) and compared estimates of association for each early-life social variable with vs. without the inclusion of the DNA methylation mediator97. We considered evidence for mediation by DNA methylation if the inclusion of %CCGG methylation attenuated the estimate for a given early-life social metric by >10%.

Fig. 7: A conceptual diagram outlining the analytical strategy for the three-step mediation analysis.
figure7

X represents the explanatory variables of interest, M represents the potential mediator variable, and Y represents the outcome variable.

Part 4: genome-wide (mERRBS) DNA methylation analyses

Finally, as a complement to Part 3, we sought to identify differentially methylated CpG sites that were associated with later-life corticosterone levels (fGCMs) as well as maternal care and/or maternal social rank during a hyena’s first year of life as a way to identify biomarkers of this relationship following the “meet-in-the-middle” approach for high-dimensional data98. For this exploratory analysis, we selected 30 female hyenas from a single social group, born between 2011 and 2014 (except for one hyena born in 2010 and another born in 2007), and ranging in age from 11 to 27 months old at the time of darting in order to reduce variability in background characteristics. Of the selected 30 hyenas, we had information on fGCMs during the subadult or adult life stages (≥13 months) for 25 hyenas; these individuals comprised the study sample for the EWAS (Supplementary Fig. 7).

Prior to formal analysis, we filtered the DNA methylation data by removing 5% of CpG sites with the lowest interindividual variation in DNA methylation and removing CpG sites for which the average DNA methylation was <10 or >90%99. Next, to prepare the fGCMs data for the EWAS, we consolidated the repeated measures (range 1–18) of fGCMs such that each individual had one value for their stress phenotype using a mixed-effects linear regression model where repeated measures of fGCMs were the outcomes and a random intercept for hyena ID was included to output the best linear unbiased predictors (BLUPs) from this model (see Supplementary Material). These BLUPs represent the deviation of each individual’s average fGCMs concentration (i.e., the mean across all available measurements) from the population average. We then analyzed associations of the BLUPs with genome-wide DNA methylation using binomial regression models in MACAU100. In these models, the BLUPs were the explanatory variable and DNA methylation of each sequenced CpG site (count of methylated divided by total CpG sites) was the outcome (Supplementary Fig. 7a). We controlled for offspring age in months at the time of darting as a precision covariate, and we accounted for genetic correlations by including a relatedness covariance matrix based on a pedigree of our study population. After running the EWAS, we corrected estimates of association and P values for inflation and bias due to unmeasured confounding using the Bioconductor package, “bacon” in order to reduce false-positive associations101. We accounted for multiple comparisons using a Benjamini–Hochberg FDR of 5% on the inflation and bias-corrected P values102. Finally, among the 25 individuals for which we assayed mERRBS, we annotated CpG sites from the EWAS in genomic regions of the draft hyena genome as described in the Supplementary Material. We then assessed patterns of positive and negative associations of DNA methylation with fGCMs and compared the distribution of Beta estimates within genomic regions vs. across the entire genome to assess for bias in the direction of associations from the EWAS.

Among the DMSs identified in the EWAS, we focused on those that were also associated with maternal care and/or maternal rank. By focusing on CpG sites that are associated with later-life fGCMs and with maternal care and/or maternal rank, we identified potential biomarkers of stress-related genomic pathways responsive to early social experience (Supplementary Fig. 7b, c).

To accomplish this, we ran separate generalized linear regression models where the explanatory variable was a given maternal care metric or maternal rank, and the outcome was counts of methylated DNA sequence reads at each stress-related DMS identified from the EWAS. We treated the total number of DNA sequence read counts (methylated + unmethylated) at each DMS as an offset in our models. We considered a CpG site to be of interest based on two criteria. First, the Beta coefficient for the relationship between a given early social characteristic and DNA methylation for the CpG site of interest needed to reach statistical significance at a nominal P value cut-off of <0.1. Second, given our overarching hypothesis that better quality or quantity of maternal care is protective against an adverse stress phenotype in offspring, we focused on CpG sites that exhibited the opposite direction of associations for the early social metrics vs. fGCMs. That is, our hypothesis proposes that a more favorable early social environment is associated with a better stress phenotype later in life. If this is the case, then a CpG site that is positively associated with maternal care (i.e., better quantity or quality of maternal care correlates with higher DNA methylation of that CpG site) should be inversely related to cortisol (i.e., higher DNA methylation of the CpG site should be associated with lower fGCMs). Thus, CpG sites of interest should theoretically be associated with maternal care and fGCMs in opposite directions in order to be potential markers of this relationship.

To interpret DMSs of interest, we copied an ~20 kb sequence of spotted hyena DNA centered on the CpG site of interest and saved the sequences as Fasta files. We first mapped each DNA sequence containing the CpG site of interest to the domestic cat (Felis catus Nov. 2017 [felCat9] Assembly) using the UCSC Genome Browser tool, the BLAST-like alignment tool (BLAT)23 and identified nearest genes from multiple species alignment. Next, we mapped each hyena DNA sequence to the human genome (UCSC Human Dec. 2010 [GRCh38/hg38] Assembly) to again identify the nearest gene(s) and cross-checked the alignments with those from the more closely related cat genome. We used default BLAT parameters (see Supplementary Material) and retained the highest-scoring DNA sequence matches in cases where there was concordance between the multiple species alignment when querying hyena DNA against the cat and human reference genomes. We do not report the nearest human gene(s) when there was discordance between cat and human queries and when the BLAT score and % identical sequence matches were low (<500 BLAT score, <85% match).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The full data used in this paper are available at https://doi.org/10.5281/zenodo.4967924103. Source data are provided with this paper. Sequence data for the spotted hyena putative glucocorticoid receptor promoter region are available at GenBank at https://www.ncbi.nlm.nih.gov/nuccore/MZ399371, accession number BankIt2471386 Crocuta MZ399371. Short read sequence data from the epigenome wide association study are available at the Sequence Read Archive at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA740026, accession number PRJNA740026.

Code availability

Source code is provided with this paper at https://doi.org/10.5281/zenodo.4967924103.

References

  1. 1.

    Harlow, H. F., Dodsworth, R. O. & Harlow, M. K. Total social isolation in monkeys. Proc. Natl Acad. Sci. USA 54, 90–97 (1965).

    ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 

  2. 2.

    Levine, S. Infantile experience and resistance to physiological stress. Science 126, 405 (1957).

    ADS  CAS  PubMed  Article  Google Scholar 

  3. 3.

    Liu, D. et al. Maternal care, hippocampal glucocorticoid receptors, and hypothalamic-pituitary-adrenal responses to stress. Science 277, 1659–1662 (1997).

    CAS  PubMed  Article  Google Scholar 

  4. 4.

    Francis, D., Diorio, J., Liu, D. & Meaney, M. J. Nongenomic transmission across generations of maternal behavior and stress responses in the rat. Science 286, 1155–1158 (1999).

    CAS  PubMed  Article  Google Scholar 

  5. 5.

    Caldji, C. et al. Maternal care during infancy regulates the development of neural systems mediating the expression of fearfulness in the rat. Proc. Natl Acad. Sci. USA 95, 5335–5340 (1998).

    ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 

  6. 6.

    Vargas, J., Junco, M., Gomez, C. & Lajud, N. Early life stress increases metabolic risk, HPA axis reactivity, and depressive-like behavior when combined with postweaning social isolation in rats. PLoS ONE 11, 1–21 (2016).

    CAS  Google Scholar 

  7. 7.

    Sánchez, M. M. et al. Alterations in diurnal cortisol rhythm and acoustic startle response in nonhuman primates with adverse rearing. Biol. Psychiatry 57, 373–381 (2005).

    PubMed  Article  CAS  Google Scholar 

  8. 8.

    Fries, A. B. W., Shirtcliff, E. A. & Pollak, S. D. Neuroendocrine dysregulation following early social deprivation in children. Dev. Psychobiol. 50, 588–599 (2008).

    CAS  PubMed  Article  Google Scholar 

  9. 9.

    Li, E. & Bird, A. In Epigenetics (eds Allis, C. D., Jenuwein, T., Reinberg, D. & Caparros, M.-L.), 343–356 (Cold Spring Harbor Laboratory Press, 2007).

  10. 10.

    Weaver, I. C. G. et al. Epigenetic programming by maternal behavior. Nat. Neurosci. 7, 847–854 (2004).

    CAS  PubMed  Article  Google Scholar 

  11. 11.

    Anier, K. et al. Maternal separation is associated with DNA methylation and behavioural changes in adult rats. Eur. Neuropsychopharmacol. 24, 459–468 (2014).

    CAS  PubMed  Article  Google Scholar 

  12. 12.

    Provencal, N. et al. The signature of maternal rearing in the methylome in rhesus macaque prefrontal cortex and T cells. J. Neurosci. 32, 15626–15642 (2012).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  13. 13.

    Unternaehrer, E. et al. Childhood maternal care is associated with DNA methylation of the genes for brain-derived neurotrophic factor (BDNF) and oxytocin receptor (OXTR) in peripheral blood cells in adult men and women. Stress 18, 451–461 (2015).

    PubMed  Article  CAS  Google Scholar 

  14. 14.

    Sánchez, M. M., Ladd, C. O. & Plotsky, P. M. Early adverse experience as a developmental risk factor for later psychopathology: evidence from rodent and primate models. Dev. Psychopathol. 13, 419–449 (2001).

    PubMed  Article  Google Scholar 

  15. 15.

    Van Bodegom, M., Homberg, J. R. & Henckens, M. J. A. G. Modulation of the hypothalamic-pituitary-adrenal axis by early life stress exposure. Front. Cell. Neurosci. 11, 1–33 (2017).

    Google Scholar 

  16. 16.

    Moore, S. R. et al. Epigenetic correlates of neonatal contact in humans. Dev. Psychopathol. 29, 1517–1538 (2017).

    PubMed  Article  Google Scholar 

  17. 17.

    Sanchez, M. M. The impact of early adverse care on HPA axis development: nonhuman primate models. Horm. Behav. 50, 623–631 (2006).

    PubMed  Article  Google Scholar 

  18. 18.

    Houtepen, L. C. et al. Genome-wide DNA methylation levels and altered cortisol stress reactivity following childhood trauma in humans. Nat. Commun. 7, 10967 (2016).

  19. 19.

    Coley, E. J. L. et al. Cross-generational transmission of early life stress effects on HPA regulators and bdnf are mediated by sex, lineage, and upbringing. Front. Behav. Neurosci. 13, 1–17 (2019).

    Article  CAS  Google Scholar 

  20. 20.

    Kember, R. L. et al. Maternal separation is associated with strain-specific responses to stress and epigenetic alterations to Nr3c1, Avp, and Nr4a1 in mouse. Brain Behav. 2, 455–467 (2012).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  21. 21.

    Dunn, E. C. et al. Sensitive periods for the effect of childhood adversity on DNA methylation: results from a prospective, longitudinal study. Biol. Psychiatry 85, 838–849 (2019).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  22. 22.

    Hennessy, M. B., Hornschuh, G., Kaiser, S. & Sachser, N. Cortisol responses and social buffering: a study throughout the life span. Horm. Behav. 49, 383–390 (2006).

    CAS  PubMed  Article  Google Scholar 

  23. 23.

    Kent, W. J. et al. The Human Genome Browser at UCSC. Genome Res. 12, 996–1006 (2002).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  24. 24.

    Kent, W. J. BLAT—the BLAST-like alignment tool. Genome Res. 12, 656–664 (2002).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  25. 25.

    Kornienko, O., Clemans, K. H., Out, D. & Granger, D. A. Hormones, behavior, and social network analysis: exploring associations between cortisol, testosterone, and network structure. Horm. Behav. 66, 534–544 (2014).

    CAS  PubMed  Article  Google Scholar 

  26. 26.

    Kornienko, O., Clemans, K. H., Out, D. & Granger, D. A. Friendship network position and salivary cortisol levels. Soc. Neurosci. 8, 385–396 (2013).

    PubMed  Article  Google Scholar 

  27. 27.

    Ponzi, D., Muehlenbein, M. P., Geary, D. C. & Flinn, M. V. Cortisol, salivary alpha-amylase and children’s perceptions of their social networks. Soc. Neurosci. 11, 164–174 (2016).

    PubMed  Article  Google Scholar 

  28. 28.

    Wittig, R. M. et al. Focused grooming networks and stress alleviation in wild female baboons. Horm. Behav. 54, 170–177 (2008).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  29. 29.

    Wey, T. W. & Blumstein, D. T. Social attributes and associated performance measures in marmots: bigger male bullies and weakly affiliating females have higher annual reproductive success. Behav. Ecol. Sociobiol. 66, 1075–1085 (2012).

    Article  Google Scholar 

  30. 30.

    Priebe, K. et al. Maternal influences on adult stress and anxiety-like behavior in C57BL/6J and BALB/CJ mice: a cross-fostering study. Dev. Psychobiol. 47, 398–407 (2005).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  31. 31.

    McLaughlin, K. A. et al. Causal effects of the early caregiving environment on development of stress response systems in children. Proc. Natl Acad. Sci. USA 112, 5637–5642 (2015).

    ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 

  32. 32.

    Sheriff, M. J., Dantzer, B., Delehanty, B., Palme, R. & Boonstra, R. Measuring stress in wildlife: techniques for quantifying glucocorticoids. Oecologia 166, 869–887 (2011).

    ADS  PubMed  Article  PubMed Central  Google Scholar 

  33. 33.

    Goymann, W. On the use of non-invasive hormone research in uncontrolled, natural environments: the problem with sex, diet, metabolic rate and the individual. Methods Ecol. Evol. 3, 757–765 (2012).

    Article  Google Scholar 

  34. 34.

    Laubach, Z. M. et al. Early life social and ecological determinants of global DNA methylation in wild spotted hyenas. Mol. Ecol. 28, 3799–3812 (2019).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  35. 35.

    Greenberg, J. R. Developmental Flexibility in Spotted Hyneas (Crocuta crocuta): The Role of Maternal and Anthropogenic Effects (Michigan State University, 2017).

  36. 36.

    Turner, J. W., Bills, P. S. & Holekamp, K. E. Ontogenetic change in determinants of social network position in the spotted hyena. Behav. Ecol. Sociobiol. 72, 1–5 (2018).

  37. 37.

    Smolarek, I. et al. Global DNA methylation changes in blood of patients with essential hypertension. Med. Sci. Monit. 16, 149–155 (2010).

    Google Scholar 

  38. 38.

    Zinellu, A. et al. Blood global DNA methylation is decreased in non-severe chronic obstructive pulmonary disease (COPD) patients. Pulm. Pharmacol. Ther. 46, 11–15 (2017).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  39. 39.

    Dong, Y. et al. Associations between global DNA methylation and telomere length in healthy adolescents. Sci. Rep. 7, 1–6 (2017).

    ADS  Article  CAS  Google Scholar 

  40. 40.

    Wong, J. Y. Y. et al. The association between global DNA methylation and telomere length in a longitudinal study of boilermakers. Genet. Epidemiol. 38, 254–264 (2014).

    PubMed  PubMed Central  Article  Google Scholar 

  41. 41.

    Woo, H. D. & Kim, J. Global DNA hypomethylation in peripheral blood leukocytes as a biomarker for cancer risk: A meta-analysis. PLoS ONE 7, e34615 (2012).

    ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 

  42. 42.

    Sharma, P. et al. Detection of altered global DNA methylation in coronary artery disease patients. DNA Cell Biol. 27, 357–365 (2008).

    CAS  PubMed  Article  Google Scholar 

  43. 43.

    Ono, H. et al. Association of dietary and genetic factors related to one-carbon metabolism with global methylation level of leukocyte DNA. Cancer Sci. 103, 2159–2164 (2012).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  44. 44.

    Basu, N. et al. Effects of methylmercury on epigenetic markers in three model species: mink, chicken and yellow perch Niladri. Comp. Biochem. Physiol. C 157, 322–327 (2013).

    CAS  Google Scholar 

  45. 45.

    Laubach, Z. M. et al. Socioeconomic status and DNA methylation from birth through mid-childhood: a prospective study in Project Viva. Epigenomics https://doi.org/10.2217/epi-2019-0040 (2019).

  46. 46.

    Meissner, A. et al. Reduced representation bisulfite sequencing for comparative high-resolution DNA methylation analysis. Nucleic Acids Res. 33, 5868–5877 (2005).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  47. 47.

    Brown, G. R. et al. Gene: a gene-centered information resource at NCBI. Nucleic Acids Res. 43, D36–D42 (2015).

    CAS  PubMed  Article  Google Scholar 

  48. 48.

    National Library of Medicine (US), National Center for Biotechnology Information. Gene. https://www.ncbi.nlm.nih.gov/gene/ (2004).

  49. 49.

    Binns, D. et al. QuickGO: a web-based tool for Gene Ontology searching. Bioinformatics 25, 3045–3046 (2009).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  50. 50.

    Huntley, R. P. et al. The GOA database: Gene Ontology annotation updates for 2015. Nucleic Acids Res. 43, D1057–D1063 (2015).

    CAS  PubMed  Article  Google Scholar 

  51. 51.

    Chang, I. & Parrilla, M. Expression patterns of homeobox genes in the mouse vomeronasal organ at postnatal stages. Gene Expr. Patterns 21, 69–80 (2016).

    CAS  PubMed  Article  Google Scholar 

  52. 52.

    Santos, J. S., Fonseca, N. A., Vieira, C. P., Vieira, J. & Casares, F. Phylogeny of the teashirt-related zinc finger (tshz) gene family and analysis of the developmental expression of tshz2 and tshz3b in the zebrafish. Dev. Dyn. 239, 1010–1018 (2010).

    CAS  PubMed  Article  Google Scholar 

  53. 53.

    Zhou, T. et al. Peripheral blood gene expression as a novel genomic biomarker in complicated sarcoidosis. PLoS ONE 7, 1–13 (2012).

    CAS  Google Scholar 

  54. 54.

    Scheinfeldt, L. B. et al. Using the Coriell Personalized Medicine Collaborative Data to conduct a genome-wide association study of sleep duration. Am. J. Med. Genet. B 168, 697–705 (2015).

    Article  Google Scholar 

  55. 55.

    Riku, M. et al. Down-regulation of the zinc-finger homeobox protein TSHZ2 releases GLI1 from the nuclear repressor complex to restore its transcriptional activity during mammary tumorigenesis. Oncotarget 7, 5690–5701 (2016).

    PubMed  Article  Google Scholar 

  56. 56.

    Tapia-Carrillo, D., Tovar, H., Velazquez-Caldelas, T. E. & Hernandez-Lemus, E. Master regulators of signaling pathways: an application to the analysis of gene regulation in breast cancer. Front. Genet. 10, 1–11 (2019).

    Article  CAS  Google Scholar 

  57. 57.

    Yamamoto, M., Cid, E., Bru, S. & Yamamoto, F. Rare and frequent promoter methylation, respectively, of TSHZ2 and 3 genes that are both downregulated in expression in breast and prostate cancers. PLoS ONE 6, 1–10 (2011).

    Article  Google Scholar 

  58. 58.

    Zhou, S. et al. Proteomic landscape of TGF-β1-induced fibrogenesis in renal fibroblasts. Sci. Rep. 10, 1–17 (2020).

    Article  CAS  Google Scholar 

  59. 59.

    Seto, S., Tsujimura, K. & Koide, Y. Rab GTPases regulating phagosome maturation are differentially recruited to mycobacterial phagosomes. Traffic 12, 407–420 (2011).

    CAS  PubMed  Article  Google Scholar 

  60. 60.

    Kretzer, N. M. et al. RAB43 facilitates cross-presentation of cell-associated antigens by CD8α+ dendritic cells. J. Exp. Med. 213, 2871–2883 (2016).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  61. 61.

    Huang, Z., Liang, H. & Chen, L. Rab43 promotes gastric cancer cell proliferation and metastasis via regulating the pi3k/akt signaling pathway. OncoTargets Ther. 13, 2193–2202 (2020).

    CAS  Article  Google Scholar 

  62. 62.

    Han, M. Z. et al. High expression of RAB43 predicts poor prognosis and is associated with epithelial-mesenchymal transition in gliomas. Oncol. Rep. 37, 903–912 (2017).

    CAS  PubMed  Article  Google Scholar 

  63. 63.

    Blackburn, M. R., Datta, S. K., Wakamiya, M., Vartabedian, B. S. & Kellems, R. E. Metabolic and immunologic consequences of limited adenosine deaminase expression in mice. J. Biol. Chem. 271, 15203–15210 (1996).

    CAS  PubMed  Article  Google Scholar 

  64. 64.

    Bradford, K. L., Moretti, F. A., Carbonaro-Sarracino, D. A., Gaspar, H. B. & Kohn, D. B. Adenosine deaminase (ADA)-deficient severe combined immune deficiency (SCID): molecular pathogenesis and clinical manifestations. J. Clin. Immunol. 37, 626–637 (2017).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  65. 65.

    Parish, S. T. et al. Adenosine deaminase modulation of telomerase activity and replicative senescence in human CD8 T lymphocytes. J. Immunol. 184, 2847–2854 (2010).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  66. 66.

    Sánchez-Melgar, A., Albasanz, J. L., Pallàs, M. & Martín, M. Adenosine metabolism in the cerebral cortex from several mice models during aging. Int. J. Mol. Sci. 21, 1–20 (2020).

    Article  CAS  Google Scholar 

  67. 67.

    Geiger, J. D. & Nagy, J. I. Ontogenesis of adenosine deaminase activity in rat brain. J. Neurochem. 48, 147–153 (1987).

    CAS  PubMed  Article  Google Scholar 

  68. 68.

    Vasudha, K. C., Nirmal Kumar, A. & Venkatesh, T. Studies on the age dependent changes in serum adenosine deaminase activity and its changes in hepatitis. Indian J. Clin. Biochem. 21, 116–120 (2006).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  69. 69.

    Sims, B., Powers, R. E., Sabina, R. L. & Theibert, A. B. Elevated adenosine monophosphate deaminase activity in Alzheimer’s disease brain. Neurobiol. Aging 19, 385–391 (1998).

    CAS  PubMed  Article  Google Scholar 

  70. 70.

    Singh, L. S. & Sharma, R. Developmental expression and corticosterone inhibition of adenosine deaminase activity in different tissues of mice. Mech. Ageing Dev. 80, 85–92 (1995).

    CAS  PubMed  Article  Google Scholar 

  71. 71.

    McGowan, P. O. et al. Epigenetic regulation of the glucocorticoid receptor in human brain associates with childhood abuse. Nat. Neurosci. 12, 342–348 (2009).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  72. 72.

    Pan, P., Fleming, A. S., Lawson, D., Jenkins, J. M. & McGowan, P. O. Within- and between-litter maternal care alter behavior and gene regulation in female offspring. Behav. Neurosci. 128, 736–748 (2014).

    CAS  PubMed  Article  Google Scholar 

  73. 73.

    Romero, L. M., Dickens, M. J. & Cyr, N. E. The Reactive Scope Model - a new model integrating homeostasis, allostasis, and stress. Horm. Behav. 55, 375–389 (2009).

    PubMed  Article  Google Scholar 

  74. 74.

    Kamin, H. S. & Kertes, D. A. Cortisol and DHEA in development and psychopathology. Horm. Behav. 89, 69–85 (2017).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  75. 75.

    Engler, H., Bailey, M. T., Engler, A. & Sheridan, J. F. Effects of repeated social stress on leukocyte distribution in bone marrow, peripheral blood and spleen. J. Neuroimmunol. 148, 106–115 (2004).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  76. 76.

    Kruuk, H. The Spotted Hyena: A Study of Predation and Social Behavior (University of Chicago Press, 1972).

  77. 77.

    Holekamp, K., Smale, L. & Szykman, M. Rank and reproduction in the female spotted hyaena. J. Reprod. Fertil. 108, 229–237 (1996).

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  78. 78.

    Holekamp, K. E. & Smale, L. Behavioral development in the spotted hyena. Bioscience 48, 997–1005 (1998).

    Article  Google Scholar 

  79. 79.

    Holekamp, K. E. et al. Patterns of association among female spotted hyenas (Crocuta crocuta). J. Mammal. 78, 55–64 (1997).

    Article  Google Scholar 

  80. 80.

    Turner, J. W., Robitaille, A. L., Bills, P. S. & Holekamp, K. E. Early-life relationships matter: social position during early life predicts fitness among female spotted hyenas. J. Anim. Ecol. 90, 183–196 (2021).

    PubMed  Article  Google Scholar 

  81. 81.

    Altmann, J. Observational study of behavior: sampling methods. Behaviour 49, 227–267 (1974).

    CAS  PubMed  Article  Google Scholar 

  82. 82.

    Karimi, M., Johansson, S. & Ekström, T. J. Using LUMA. A luminometric-based assay for global DNA methylation. Epigenetics 1, 45–48 (2006).

    PubMed  Google Scholar 

  83. 83.

    Coluccio, A. et al. Individual retrotransposon integrants are differentially controlled by KZFP/KAP1-dependent histone methylation, DNA methylation and TET-mediated hydroxymethylation in naïve embryonic stem cells. Epigenet. Chromatin 11, 1–18 (2018).

    Article  CAS  Google Scholar 

  84. 84.

    Eden, A., Gaudet, F., Waghmare, A. & Jaenisch, R. Chromosomal instability and tumors promoted by DNA hypomethylation. Science 300, 455 (2003).

    CAS  PubMed  Article  Google Scholar 

  85. 85.

    Lev Maor, G., Yearim, A. & Ast, G. The alternative role of DNA methylation in splicing regulation. Trends Genet. 31, 274–280 (2015).

    CAS  PubMed  Article  Google Scholar 

  86. 86.

    Doherty, T. S., Forster, A. & Roth, T. L. Global and gene-specific DNA methylation alterations in the adolescent amygdala and hippocampus in an animal model of caregiver maltreatment. Behav. Brain Res. 298, 55–61 (2016).

    CAS  PubMed  Article  Google Scholar 

  87. 87.

    Noguera, J. C. & Velando, A. Bird embryos perceive vibratory cues of predation risk from clutch mates. Nat. Ecol. Evol. 3, 1225–1232 (2019).

    PubMed  Article  Google Scholar 

  88. 88.

    Crudo, A. et al. Prenatal synthetic glucocorticoid treatment changes DNA methylation states in male organ systems: multigenerational effects. Endocrinology 153, 3269–3283 (2012).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  89. 89.

    Garrett-Bakelman, F. E. et al. Enhanced reduced representation bisulfite sequencing for assessment of DNA nethylation at base pair resolution. J. Vis. Exp. https://doi.org/10.3791/52246, 1–15 (2015).

  90. 90.

    Yang, C. et al. A draft genome assembly of spotted hyena, Crocuta crocuta. Sci. Data 7, 1–10 (2020).

    CAS  Article  Google Scholar 

  91. 91.

    Mccormick, J. A. et al. 5’-Heterogeneity of glucocorticoid receptor messenger RNA is tissue specific: differential regulation of variant transcripts by early-life events. Mol. Endocrinol. 14, 506–517 (2000).

    CAS  PubMed  Google Scholar 

  92. 92.

    Szyf, M., Weaver, I. C. G., Champagne, F. A., Diorio, J. & Meaney, M. J. Maternal programming of steroid receptor expression and phenotype through DNA methylation in the rat. Front. Neuroendocrinol. 26, 139–162 (2005).

    CAS  PubMed  Article  Google Scholar 

  93. 93.

    Van Meter, P. E. et al. Fecal glucocorticoids reflect socio-ecological and anthropogenic stressors in the lives of wild spotted hyenas. Horm. Behav. 55, 329–337 (2009).

    PubMed  Article  CAS  Google Scholar 

  94. 94.

    Dloniak, S. M. et al. Non-invasive monitoring of fecal androgens in spotted hyenas (Crocuta crocuta). Gen. Comp. Endocrinol. 135, 51–61 (2004).

    CAS  PubMed  Article  Google Scholar 

  95. 95.

    Laubach, Z. M., Murray, E. J., Hoke, K. L., Safran, R. J. & Perng, W. A biologist’s guide to model selection and causal inference. Proc. R. Soc. Ser. B https://doi.org/10.1098/rspb.2020.2815 (2021).

  96. 96.

    Engh, A. L., Esch, K., Smale, L. & Holekamp, K. E. Mechanisms of maternal rank ‘inheritance’ in the spotted hyaena, Crocuta crocuta. Anim. Behav. 60, 323–332 (2000).

    CAS  PubMed  Article  Google Scholar 

  97. 97.

    Baron, R. M. & Kenny, D. A. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J. Pers. Soc. Psychol. 51, 1173–1182 (1986).

    CAS  PubMed  Article  Google Scholar 

  98. 98.

    Chadeau-Hyam, M. et al. Meeting-in-the-middle using metabolic profiling-a strategy for the identification of intermediate biomarkers in cohort studies. Biomarkers 16, 83–88 (2011).

    CAS  PubMed  Article  Google Scholar 

  99. 99.

    Lea, A. J., Altmann, J., Alberts, S. C. & Tung, J. Resource base influences genome-wide DNA methylation levels in wild baboons (Papio cynocephalus). Mol. Ecol. 25, 1681–1696 (2016).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  100. 100.

    Lea, A. J., Tung, J. & Zhou, X. A flexible, efficient binomial mixed model for identifying differential DNA methylation in bisulfite sequencing data. PLoS Genet. 11, 1–31 (2015).

    Article  CAS  Google Scholar 

  101. 101.

    van Iterson, M. et al. Controlling bias and inflation in epigenome- and transcriptome-wide association studies using the empirical null distribution. Genome Biol. 18, 1–13 (2017).

    Article  CAS  Google Scholar 

  102. 102.

    Hochberg, Y. & Benjamini, Y. More powerful procedures for multiple statistical significance testing. Stat. Med. 9, 811–818 (1990).

    CAS  PubMed  Article  Google Scholar 

  103. 103.

    Laubach, Z. M. et al. Early-life social experience affects offspring DNA methylation and later life stress phenotype. https://doi.org/10.5281/zenodo.4967924 (2021).

Download references

Acknowledgements

We thank the Kenyan National Commission for Science, Technology, and Innovation, the Naboisho Conservancy, the Narok County Government, the Kenya Wildlife Service (KWS) and Brian Heath for permission to conduct research in the Mara ecosystem. We also thank the Mara Hyena Project field crew and the residents of the Mara ecosystem for their direct and indirect support of long-term field research in their backyard. We are indebted to all those who have contributed to long-term data and sample collection on the Mara Hyena Project, including undergraduate researchers, Karee Lesko, Shannon Carvey, and Kaycee Morra. We are also grateful to Dr. Elise Zipkin for suggestions and feedback regarding the statistical modeling, as well as Dr. Lu Tang and Patrick Bills for help processing the behavioral data. This work was supported by National Science Foundation Grants DEB1353110, OISE1556407, and IOS1755089 to KEH, and Doctoral Dissertation Improvement Grant from NSF (DDIG 1701384) to ZML. This work was also supported in part by funds from NSF Grant OIA 0939454 to the BEACON Center for the Study of Evolution in Action as well as Michigan Lifestage Environmental Exposures and Disease (M-LEEaD), NIEHS Core Center (P30 ES017885), as well as the UM NIEHS Institutional Training Grant T32 ES007062 to DCD. Dr. Perng is supported by the Center for Clinical and Translational Sciences Institute via KL2-TR002534.

Author information

Affiliations

Authors

Contributions

Conceptualization, Z.M.L.; data generation and curation, Z.M.L., J.R.G., J.W.T., T.M.M., M.A.S., L.S., R.G.C., K.R.P., C.L. M.O.P.; formal analysis, Z.M.L., R.G.C., W.P.; funding acquisition, Z.M.L., C.D.F., D.C.D., K.E.H.; investigation, Z.M.L.; methodology, Z.M.L., R.G.C., K.R.P., C.L., B.v.H., C.D.F., K.E.H., W.P.; resources, Z.M.L., C.D.F., D.C.D., K.E.H.; supervision, B.v.H., C.D.F., D.C.D., K.E.H., W.P.; visualization, Z.M.L. R.G.C.; writing – original draft, Z.M.L.; writing – review & editing, Z.M.L., J.R.G., J.W.T., T.M.M., M.A.S., M.O.P., L.S., R.G.C., K.R.P., C.L., B.v.H., C.D.F., D.C.D., K.E.H., W.P.

Corresponding author

Correspondence to Zachary M. Laubach.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Communications thanks Richard Hunter, Kelly Lambert and the other, anonymous, reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Laubach, Z.M., Greenberg, J.R., Turner, J.W. et al. Early-life social experience affects offspring DNA methylation and later life stress phenotype. Nat Commun 12, 4398 (2021). https://doi.org/10.1038/s41467-021-24583-x

Download citation

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing