Genome-wide signatures of early life stress: influence of sex

Abstract

Both history of early life stress (ELS) and female sex are associated with increased risk for depression. The complexity of how ELS interacts with brain development and sex to impart risk for multifaceted neuropsychiatric disorders is also unlikely to be understood by examining changes in single genes. Here, we review an emerging literature on genome-wide transcriptional and epigenetic signatures of ELS and the potential moderating influence of sex. We discuss evidence that there are both latent sex differences revealed by ELS, and ELS itself produces latent transcriptomic changes revealed by adult stress. In instances where there are broad similarities in global signatures of ELS among females and males, genes that contribute to these patterns are largely distinct based on sex. As this area of investigation grows, effort should be made to better understand sex-specific impact of ELS within the human brain, specific contributions of chromosomal vs hormonal sex, how ELS alters the time course of normal transcriptional development, and the cell-type specificity of transcriptomic and epigenomic changes in the brain. A better understanding of how ELS interacts with sex to alter transcriptomic and epigenomic signatures in the brain will inform individualized therapeutic strategies to prevent or ameliorate or prevent depression and other psychiatric disorders this vulnerable population.

Early life stress (ELS) is strongly linked to lifetime risk of depression and other mood, anxiety, and substance use disorders (1–3). Childhood adversity can include social and environmental insults such as abuse, neglect, family separation, forced migration, and poverty. The impact of ELS is uneven across communities, and is exacerbated by current global issues such as climate change (4) and the COVID-19 pandemic (5). In humans, non-human primates, and rodent models, ELS has been found to alter development of the brain, detectable even at the molecular level. Developmental changes in gene expression and epigenetic regulation of transcription can impact brain circuit formation, maturation, and function, and mediate susceptibility or resilience to neuropsychiatric disorders throughout life.

There are well-documented contributions of sex (including genetic, hormonal, and gender aspects) to the experience of ELS and mental health outcomes following ELS. Female and transgender children (not mutually-exclusive) have higher rates of adverse childhood experiences (ACEs) and abuse (1,6–8), which in turn may influence rates of depression across the lifespan. While in childhood the incidence of depression is equal among self-identified girls and boys, after puberty prevalence of major depressive disorder is nearly twice as high among self-identified women as men. Sex hormones likely play a mediating role for at least some of this disparity, as higher incidence of depression is observed throughout the reproductive period and the risk for females peaks during periods associated with large hormonal changes including during puberty, peripartum periods, and perimenopause (9). Historically, however, studies of ELS and/or depression either included only male subjects or did not examine/report interactions of ELS experience with sex.

The complexity of how ELS interacts with brain development and sex to impart risk is not adequately captured in candidate gene studies. Studies of polymorphisms or changes in transcription or epigenetic regulation of a handful of candidate genes [including SlC6A4 (serotonin transporter), NR3C1 (glucocorticoid receptor), FKBP5 (FK506-binding protein 1; modulates glucocorticoid receptor activity), and BDNF (brain-derived neurotrophic factor)] have been fruitful, but there is a growing appreciation that multifaceted mental health disorders are more likely the result of many small transcriptional changes than large changes in expression or function of any one single gene. Advances in the last 15 years in massively parallel, deep sequencing technology are transforming neuropsychiatric research by enabling researchers to examine molecular correlates of disease on a genome-wide scale. Evolutions in technology now enable complete genome sequencing of a single individual at base resolution in the timescale of hours to days. Sequencing itself is only part of the battle, and genome alignment and analyses have also become ultra-fast, aided by high performance computing clusters, parallelization of processes, and open-source sharing of analysis pipelines. Genome-wide sequencing of gene expression and of epigenetic regulators in humans and model organisms now enable discovery of how ELS and sex alter entire gene networks and broader patterns of transcription.

Here we review the literature to-date on genome-wide signatures of ELS and the potential moderating influence of sex. We focus on studies of early adversity in children or neonatal stress in animal models. Although adolescent stress is often considered distinct from ELS, we include discussion on two studies of sex differences in transcriptomic response to adolescent stress. For further discussion of genome-wide sex differences in adult depression, we refer readers to current reviews and other articles in this special issue (10). For a detailed discussion of sex differences in gene expression in the brain more broadly, we direct readers to a recent review by Gegenhuber and Tollkuhn (11).

Definitions of sex and strategies to study sex differences in transcriptomics

“Sex” refers to biological and physiological characteristics, including X/Y chromosome complement, gonadal anatomy, hormone levels and function, or a combination thereof. While sex is usually categorized as male or female, there can be variation in each of these characteristics. In the studies reviewed here we presume that sex is self-identified in human subjects and is assigned according to external genitalia in animal models. Gender, on the other hand, is a constructed perception of one’s sex as a function of social and cultural expectations. Human studies that use the term “gender” often do so erroneously to mean sex, and model organisms do not have gender. This review specifically discusses transcriptomic differences related to sex, not gender.

It is useful to consider different categories of sex differences (12,13) which may be reflected within transcriptomic or epigenomic patterns (Figure 1). In most stress research, studies measure the same behavioral or transcriptional endpoints, which exist on a continuum, and quantitative sex differences are observed if female and male averages are different. For example, there are well-described sex differences in steroid hormone receptor expression in the brain (11). A trait may be sexually dimorphic if the endpoint has multiple forms that are largely distinct between females and males, such as with mating behaviors. In transcriptional studies sexual dimorphism may occur if females and males engage distinct networks of genes (as shown in 14). It is worth noting that sex differences in expression or sexually dimorphic gene networks may serve to promote sex differences in behavior, or distinct transcriptional trajectories may converge on the same functional cellular or behavioral endpoint, an example of compensation (15) or latent sex differences (16). There may also be latent sex differences in which expression is similar at baseline, but undetected sex differences elsewhere in the system (for example in epigenetic regulation) cause expression to diverge in response to a challenge such as ELS. Alternatively, baseline sex differences may converge or reverse after stress or other challenges. For example, adolescent social isolation stress eliminates sex differences in some defensive behaviors and leads to feminization, masculinization, or sometimes a reversal of sexually dimorphic gene expression in different groups of genes (Figure 1; 14).

Figure 1:

Example types of sex differences in transcriptomic patterns and potential response to early life stress (ELS).

Sex differences in neuropsychiatric and behavioral response to early-life stress

The moderating effect of sex on consequences of ELS in humans and animals has been comprehensively reviewed in a recent paper (17). Clinical research suggests that outcomes are dependent upon the type (dimensions of threat versus deprivation) and timing of ELS and the timing of data collection. Studies of Romanian children orphaned at birth found cognitive delays and behavioral and emotional problems in toddlers, but no sex differences (18). Behavioral problems emerged more predominately in boys by about 5 years, and increased sensitivity to internalizing disorders emerged among females by adolescence (19). Despite sex-linked differences in depression and other neuropsychiatric disorders, only about half of properly powered clinical studies found sex differences in outcome associated with ELS (17). Effects of ELS and sex in rodent studies are also mixed and may likewise depend on type and timing of ELS. Different studies using a limited bedding and nesting ELS paradigm have reported anxiety-like and depression-like behaviors and impaired cognitive behavior among both female and male rodents, though some studies directly examining sex differences report female-specific effects (20–23). The impact of ELS on rodent behavior may also emerge earlier in females, but last longer across the lifespan in males. Maternal separation stress in rodents is associated with increased anxiety-like behavior more consistently among male rodents (24,25), although other studies found similar effects on anxiety-like behavior in females interacting with hormonal state (26). Our research uses a combination of maternal separation and reduced nesting material in mice and finds that ELS by itself minimally impacts depression-like behavior, but increases susceptibility to a second hit of stress in adulthood similarly among females and males (27–29).

GENOME-WIDE SIGNATURES OF SEX OR ELS

Sex differences in transcriptomics prior to stress

Despite similarities in neuropsychiatric and behavioral outcomes, there are significant sex differences in transcriptional control of neural mechanisms serving these behaviors even at baseline without stress. Some sex differences in transcription are hormonally linked and can be observed at different stages of the female menstrual cycle (women) or estrous cycle (rodents and other animals). The rodent estrous cycle is typically divided into four stages that can be determined by vaginal cytology and which correspond to differing levels of circulating hormones: proestrus, estrus, metestrus, and diestrus. Estradiol levels are highest during proestrus in rodents (13). Hormonally linked differences in transcription were identified in the medial prefrontal cortex (PFC) of male mice and female mice in diestrus or proestrus. In fact, there were 20-fold more differentially expressed genes (DEGs) in PFC between proestrus and diestrus females than between females in either cycle state and males (30). Cycle-state dependent gene expression was related to neuronal signaling, cellular metabolism, and synaptic function among other pathways. Four-Core-Genotypes mice can further be used to specifically distinguish between chromosomal and gonadal hormone effects of sex (12). In genetically modified Four-Core-Genotypes mice, the testis-determining gene Sry has been moved to an autosome, which enables expression of Sry (and differentiation of the bipotential gonad into testes) independent from the Y chromosome, and thus separates genetic (XX or XY) from gonadal (Sry+ with testes or Sry+ with ovaries) sex. While Four-Core-Genotypes mice have not been used to study transcriptomic effects of ELS as of yet, baseline differences in transcription were recently described between chromosomally XX females and XY males in three key brain regions including the basolateral amygdala (BLA), PFC, and nucleus accumbens (NAc) (31). Genetic sex was found to drive approximately 25% of sex differences in gene expression, and gonadal sex (presence or absence of Sry and gonad-dependent developmental hormone exposure) between 23–31%, particularly genes involved in immune function (31). Activity-dependent immediate early genes were also driven by gonadal hormones and more highly expressed among mice with ovaries than testes. Moreover, chromatin organization is dynamically regulated by estrous cycle state in gonadal females (32). Together, these studies indicate widespread basal influence of sex chromosomes and hormones on the transcriptome even prior to stress experience. These basal differences need to be carefully considered when assessing the additional effect and interaction of ELS on the transcriptome.

Genome-wide signatures of human child abuse and maltreatment

Several important human studies have examined the transcriptomic and epigenomic consequences of ELS and risk for depression within both brain and peripheral tissue. These studies have included female and male subjects, but statistically controlled for sex rather than examined sex as a mediating factor. A history of child abuse is associated with altered DNA methylation at hundreds of CpG sites across all 23 chromosomes within postmortem hippocampus (33), postmortem anterior cingulate cortex (34), saliva of children (35,36), and blood of adults (37). Both hypo- and hyper-methylation were detected in all tissue types, indicating specificity of changes rather than global gain or loss. Neuron-specific differential methylation in hippocampus occurred at a cluster of genes involved in plasticity (33). In anterior cingulate, child abuse prior to age 14 most significantly altered DNA methylation and transcription of oligodendrocyte and myelin-related genes associated with myelin thickness in this brain region (34). Differential methylation sampled from peripheral tissues predicted depression and altered functional brain connectivity in maltreated children (36), and altered cortisol stress reactivity in adults with prior history of childhood trauma (37), advocating for the utility of peripheral methylome signatures as biomarkers. While it is important that these studies included both female and male subjects, future studies should be powered sufficiently to detect sex differences rather than controlling for sex as a covariate.

PUTTING IT TOGETHER: GENOME-WIDE SIGNATURES OF ELS AND SEX

Sex differences in the DNA methylome in response to ELS

Sex differences in peripheral T-cell 5mC DNA methylation patterns were examined in a non-human primate model of maternal deprivation (38). T-cell samples were taken across juvenile development in order to understand the development and persistence of peripheral DNA methylation changes. Sex-specific developmental changes in DNA methylation increased with age and were predicted to be downstream of the steroid hormone estradiol, although the majority of age-related DNA methylation changes were common between females and males. Maternal deprivation altered DNA methylation in opposite directions among females and males (latent sex differences) with the most pronounced sex-specific differences early in development (day 14), fading with age, and regulated by inflammatory pathways (38).

Sex differences in the transcriptomic response to ELS

Few studies to date have compared transcriptomic response to ELS in females and males, although more have included either males or females in different contexts (39–42). An important question for ELS studies is the time-course through which molecular and behavioral changes emerge and either persist or are returned to control levels. One time-course study examined hypothalamic gene expression changes among female and male chickens after ELS (brief daily social isolation of chicks from postnatal days 4–26), at both juvenile and adult timepoints using a microarray. Changes in transcription, hormones, and exploratory and anxiety-like behavior persisted into adulthood in males, but not females (43). Gene ontology analysis of differentially expressed genes in male chickens revealed changes predominately in neuronal differentiation, morphogenesis, and axon guidance, suggesting that ELS interacts directly with developmental rather than steady-state processes in the hypothalamus.

We examined the long-lasting transcriptomic response to ELS (a combination of maternal separation and limited nesting material from P10–17) in adult female and male mice, across three different brain regions [ventral tegmental area (VTA), nucleus accumbens (NAc) and PFC (27,29)]. Our study sought to understand how ELS might alter lasting transcriptomic patterns both on its own and in response to a second stressor in adulthood. However, it is important to note that direct comparisons between female and male mice were not made due to differences in cohort and type of adult stress between female and male mice. In this study, there was not a significant relationship between female cycle state and behavior, and tissue was taken when most female mice were in diestrus (29).

Several broad patterns emerged from this transcriptomic analysis. First, new threshold-free rank-rank hypergeometric overlap (RRHO) analyses of these datasets (GEO accession GSE89692) reveal opposite genome-wide signatures of response to ELS among females and males in VTA and NAc, though similar responses to ELS in PFC (Figure 2). This is consistent with the opposing transcriptional signatures of depression previously identified in postmortem brain tissue from women and men (44,45) and in response to adolescent stress in rats (46), confirming divergent responses to stress across the lifespan. This RRHO analysis does not, however, reflect whether there are baseline sex differences in expression or whether different networks of genes are used to reach the same cellular endpoint. Second, when RRHO analyses were performed within each sex separately, ELS alone (prior to depression-like behavior) still programmed depression-like gene expression patterns similar to those from adult stressed mice that did show behavioral changes (29). This depression-like pattern was observed in each brain region similarly among females and males. Together, these studies consistently show that stress across the life span (ELS, adolescent stress, adult stress and depression) induces latent, opposing transcriptomic signatures in females and males, but that within a sex, response to stress is similar at each age. Finally, exposure to a second stress in adulthood did not simply exaggerate expression changes induced by ELS or adult stress alone, but induced a modified transcriptional response. In other words, in both males and females, ELS induces latent transcriptomic changes – possibly through changes in the epigenome – that are later revealed by adult stress.

Figure 2: Opposite transcriptomic signatures of ELS in male and female mouse brain.

New rank-rank hypergeometric overlap (RRHO) analyses of data from (29) indicate opposite signatures of ELS in male and female NAc and to a lesser extent VTA, while signatures of ELS in PFC are similar among males and females. Analysis methods were identical to those previously reported (29). Pixels represent the overlap between the transcriptome of each comparison as noted, with the significance of overlap (−log10(p-value) of a hypergeometric test) color coded. Maximum significance of overlap was allowed to be different in each brain region and is noted in parentheses at the top right of each image. A two-sided version of RRHO was used to test for coincident and opposite enrichment such that the lower left quadrants include co-upregulated genes, upper right quadrants include co-downregulated genes, and upper left and lower right quadrants include oppositely regulated genes (up-down and down-up, respectively). Genes along each axis are sorted from most to least significantly regulated from the middle to outer corners.

A final pattern to emerge was that the combination of ELS and adult stress together enriches for gene expression changes in the PFC associated with critical period plasticity. Plasticity-associated genes were also enriched among differentially methylated genes in human hippocampus following child abuse (33). In mice, transcriptional signatures of plasticity were enriched among female and depleted among male mice after ELS. This suggests that females and males arrive at similar behavioral stress sensitivity through divergent transcriptional/plasticity mechanisms. Given timing of ELS during a sensitive window (27), it is possible that ELS interferes with natural developmental plasticity to either inappropriately preserve plasticity beyond the normal window and render mice hyper-reactive to future stimuli, or prematurely or otherwise inappropriately reduce plasticity such that mice cannot actively cope with future stress.

Although some broad patterns (such as induction of a depression-like transcriptomic state and latent response to adult stress) were similar among female and male mice, the genes contributing to and regulating these transcriptional states were largely sex-specific. These differences may reflect different baseline gene expression levels as demonstrated in other studies (31,47), or sex-specific gene network correlations. Nevertheless, a male-specific predicted upstream regulator in VTA from this pre-clinical mouse model of ELS (the transcription factor OTX2) was ultimately useful to predict clinical outcomes following ELS in children. Genome-wide analysis of salivary DNA methylation from female and male children aged 8–15 years found that child maltreatment and OTX2 methylation predicted depression in the children, and increased OTX2 methylation was associated with altered functional connectivity within the brain (36). This demonstrates the translational potential of pre-clinical animal models to understanding molecular correlates of child maltreatment in humans.

Sex differences in transcriptomic response to adolescent stress

Adolescence is marked by a surge of gonadal steroid hormones, heightened plasticity in some brain regions, increasing independence, and changes in motivated behavior (48). While adolescence precedes adulthood, it is considered separately here from animal models of ELS earlier in the neonatal/juvenile period (although human studies typically lump stress occurring prior to age 18).

Similar to ELS and adult depression, adolescent stress from P38–49 (mixed social defeat and restraint) in rats led to opposing transcriptional signatures between females and males in the hippocampus. Adolescent stress likewise modified the transcriptional response to adult stress, similar to ELS in mice, indicating latent regulatory mechanisms (46). Adolescent social isolation from P22–42 also diminished baseline sex differences in transcription in the medial amygdala (a sexually dimorphic brain region) such that a large portion of sexually dimorphic genes were feminized, masculinized, or reversed (14). Unfortunately, similar analyses focusing on sexually dimorphic gene expression response to stress have not yet been applied to datasets of neonatal ELS or adult stress, so it is not yet known whether stress at different ages can similarly disrupt basal sex differences or whether this effect is specific to stress occurring during the onset of puberty.

Common to both ELS and adolescent stress-induced transcriptomic changes is a regulatory role for hormone-responsive transcription factors. The top predicted upstream regulator of adolescent stress response in female and male hippocampus was estrogen receptor alpha (ESR1) (46), which was also the only predicted upstream regulator common to females and males after ELS in NAc (29). In the amygdala, a key driver of sex-specific transcriptomic response to adolescent stress is crystalin mu (Crym), a thyroid hormone binding protein (14). Both ESR1 and CRYM are transcription factors simultaneously responsive to hormonal environment and capable of directly regulating transcription of thousands of genes. Finally, these studies show that manipulation of sex-specific predicted transcriptional regulators leads to sex-specific functional impact on behavior. Overexpression of Crym in amygdala led to distinct transcriptional alterations in females vs males, and abolished sex differences in exploratory and anxiety-like behavior (14). This demonstrates the predictive validity of bioinformatic analyses of complex genome-wide data.

CURRENT LIMITATIONS AND OPPORTUNITIES FOR FUTURE DIRECTIONS

Our use of sequencing technology and bioinformatic analyses to understand the impact of ELS on the brain, and how sex may mediate these effects, in still in its infancy. We see four main areas for future research to expand: 1) human studies of ELS that directly examine sex differences rather than control for sex; 2) understanding the contributions of hormone levels; 3) understanding how ELS sex-specifically alters transcriptional dynamics across development; and 4) expanding sequencing methods.

Need for additional human studies

Human studies have not yet directly examined sex differences in global response to ELS, much less attempted to understand the role of gonadal hormones or sex as a nonbinary variable with intersex and transgender individuals. Some existing data may be analyzed again with the express purpose of identifying sex differences (though existing datasets may be underpowered), but there is an opportunity to generate new datasets to a priori determine how ELS and sex interact to alter transcriptional development. Of course, studying the interaction of ELS and sex in the human brain is difficult for a number of reasons including access to brain tissue, complete history including information on childhood adversity, and variability in human genetics and life experience. Building off of existing genome-wide PsychENCODE datasets or formation of a consortium specific for ELS may facilitate future research (49).

Need for understanding hormonal contributions

Including sex as a biological variable and even accounting for sex differences does not necessarily mean studying hormonal contributions, as it is not always the case that variability is accounted for by hormonal state (12,50). However, steroid hormone levels directly impact gene expression through binding of their cognate receptors and initiating activity as transcription factors. Additionally, chromatin organization in the female hippocampus dynamically fluctuates across the estrous cycle, and thus hormones are likely to play a role in the sex differences observed in genome-wide response to ELS (32). Hormonal contributions can be examined – particularly in animal models – by directly measuring hormone levels, tracking female cycle state and measuring outcome variables at specific stages, artificially after gonadectomy and/or hormone clamping, or with the Four-Core-Genotypes mouse model.

Need for longitudinal assessment

There are demonstrated acute and long-lasting effects of ELS on transcription, but longitudinal studies examining transcription across postnatal development are needed to understand how ELS interacts with and potentially disrupts the time course of transcriptional development itself. For example, expression sampled at only a single time point may erroneously appear to be suppressed when in fact development of that gene program is delayed, or there may appear to be no effect of ELS when in fact maturation was accelerated (27,51,52). A transcriptomic time course of normal and ELS-impacted neurodevelopment will also reveal sensitive periods of heightened plasticity and potential vulnerability, and whether there are sex differences in the timing thereof.

Opportunities to harness new technology

The majority of studies to date on the genome-wide impact of ELS and sex have used bulk-tissue RNA-sequencing. There is a current lack of understanding of how chromatin dynamics mediate enduring consequences of ELS on a genome-wide scale. Indeed, chromatin dynamics likely underlie the altered transcriptomic response to adult stress observed after ELS (27,46). There is also enormous potential to apply recent advances in single-cell sequencing technologies in future studies to examine cell-type specific effects of ELS in the brain. Several genome-wide studies of depression in adults have identified a role for oligodendrocytes and microglia (34,45). Single-cell sequencing is quickly becoming easier and less expensive and may provide important cellular resolution to understand sex-specific effects of ELS.

CONCLUSIONS

Sex and ELS interact to impact transcription across the genome. There are both latent sex differences revealed by ELS, and that ELS itself produces latent transcriptomic changes revealed by adult stress. In instances where there are broad similarities in global signatures of ELS among females and males, the individual genes differentially regulated by ELS, and the predicted upstream regulators of these genes, are largely non-overlapping. These important transcriptional and epigenetic patterns – above and beyond individual gene changes – could only be revealed through sequencing, highlighting the unique power of such genome-wide approaches.

Sex differences in response to ELS have major implications for response to neuropsychiatric disorder treatments. Post-traumatic stress disorder treatment response is predicted by and interaction between history of child abuse and DNA methylation profiles specifically in female patients (53). Antidepressant treatment response or non-response is also influenced by gonadal hormones (54). Rich genome-wide datasets of sex-specific response to ELS may be mapped to drug signatures available in different cell lines and tissues to identify novel and potentially sex-specific treatments for this vulnerable population (55). Additional human studies, a focus on hormonal contributions to sex differences, and increased cellular specificity of transcriptomic and epigenomic changes across development will lead to a better understanding of how ELS interacts with sex to increase risk for depression and other neuropsychiatric disorders, as well as facilitate development of individualized treatment.