Background

Early life stress (ELS) encompasses childhood abuse, neglect, poverty and parental illness, alongside a multitude of other stressors. Some forms of ELS affect 30–40 % of the Western population and have been implicated in approximately half of all childhood and a third of adulthood psychiatric disorders [1, 2]. Exposure to ELS results in enhanced susceptibility to neuropsychiatric disorders such as major depressive disorder, generalised anxiety disorder, schizophrenia and autism spectrum disorders [1, 3, 4]. Chronic health conditions such as obesity, type two diabetes mellitus, respiratory disorders and cardiovascular diseases are also increased in individuals with a history of ELS [5, 6].

The impact of early adversity on the susceptibility to psychiatric disorders in later life is influenced by a number of factors. Environmental factors include nature of stressors [7], time of exposure in development [8, 9] and severity and cumulative exposure effects [10–12]. Biological factors include gender, age of assessment [13] and predisposing genetic polymorphisms in genes associated with mood regulation, stress response and inflammatory processes. These include genes such as serotonin transporter (5-HTT), brain-derived neurotrophic factor (BDNF) and FK506 binding protein (FKBP5) [14–16]. A dysfunctional hypothalamic-pituitary-adrenal (HPA) stress axis and impaired immune responses such as increased cytokines have also been implicated in the increased vulnerability to ELS [17, 18]. In spite of the increasing knowledge, the molecular mechanisms underlying ELS-mediated long-term vulnerability to later-life stressors are unclear.

Gene-environment interactions, such as those occurring when exposed to ELS, often encompass epigenetic changes. Epigenetics processes occur at the level above the genome, which includes DNA methylation, posttranslational histone modifications and gene regulation by micro-RNA (miRNA). Collectively, these epigenetic changes can stably mark the genome in response to environment, potentially altering gene expression across lifespan [19, 20] and across generations [21]. As such, epigenetic alterations may represent one of the key mechanisms underlying the long-lasting nature of ELS-induced changes in neurobiology, behaviour and disease susceptibility [22, 23].

We first present a brief overview of the stress response pathways followed by a detailed review of evidence demonstrating epigenetic alterations following ELS in animal models and humans. Changes in the epigenetic signature of candidate genes and alterations in genome-wide methylation will be reported. Finally, we aim to understand how these changes mediate long-term effects such as their role in risk to developing psychiatric disorders in adulthood.

The central role of hypothalamus-pituitary-adrenal system

Adversities during early postnatal life are able to shape the experience-dependent maturation of stress-regulating pathways, such as the HPA system. This can lead to persistent alterations in stress responsivity during adulthood—a phenomenon often referred to as “early-life programming”. Tight regulation of the HPA axis is therefore core to the long-term control of systems governing stress responsivity. The HPA axis involves the release, following a stressor, of the neuropeptide corticotrophin-releasing hormone (CRH) and arginine vasopressin (AVP) from the paraventricular nucleus (PVN) of the hypothalamus. These bind to their specific receptors (the CRHR1 and V1b receptors) in the anterior pituitary that stimulate the release of adrenocorticotrophic hormone (ACTH) which stimulates the adrenal cortex to release glucocorticoid (GC) hormones, cortisol in human and corticosterone in rodents. These GCs in turn mobilise glucose from energy stores and increase cardiovascular tone, among further widespread effects. Feedback loops, primarily mediated through glucocorticoid receptors (GRs) in the PVN and pituitary, regulate responsiveness of the HPA axis ensuring a return to a homeostatic balance when it is no longer challenged (Fig. 1).

Fig. 1
figure 1

The hypothalamic-pituitary-adrenal (HPA) axis and its response to stress stimuli: the signalling events (green, solid lines) in the HPA axis in response to stress stimuli and how glucocorticoids (GCs) produced by the adrenal gland can have a negative feedback role in maintaining GC levels in the blood. The negative feedback in the hypothalamus and pituitary (red, dotted lines) are both mostly regulated by glucocorticoid receptors (GRs), and dysfunctional negative feedback system is often seen associated with chronic exposure to stress stimuli

Full size image

A loss of this negative feedback control, particularly following periods of chronic stress, may influence the development of affective disorders. Indeed, dysregulated HPA activity is one of the most commonly observed neuroendocrine symptoms in major depressive disorder (for review see [24]). Childhood stress has also been shown to be a strong predictor of impaired inhibitory feedback regulation of the HPA axis with evidence linking to a role of CRH and/or AVP systems. For example, postmortem brain tissue of depressed individuals revealed elevated CRH and AVP in the hypothalamus [25, 26]. Studies in rodent models further support the concept that exposure to a chronic stressor can lead to long-term changes in HPA regulation and behaviour stemming from changes in neuropeptide regulation (for review see [27]).

The role of candidate genes outside of the HPA axis

Candidate genes such as the serotonin transporter SLC6A4 and BDNF have been highly implicated in stress response and in increased risk for psychiatric disorders [28–32]. BDNF is the most prevalent growth factor in the central nervous system (CNS) and important in neuronal development and plasticity [33]. Serotonin transporter is involved in the reuptake of serotonin from the brain synapses regulating serotonin signalling and is the target for many antidepressants [34]. SLC6A4 or 5-HTT carries a genetic polymorphism in the promoter region resulting in a short “s” and a long “l” allele version of the promoter [35]. The “s” allele is associated with poor transcriptional efficiency of SLC6A4 compared to “l” allele [35]. The BDNF gene carries a Val66Met polymorphism which impacts an activity-dependent expression of BDNF and the intracellular trafficking [36]. In combination with exposure to ELS events, both SLC6A4 and BDNF polymorphisms have been attributed to increased risk for depression in later life [28, 30, 37]. Further, steroid hormone estrogen and its receptors have been shown to influence brain function and psychiatric disorders (for review see [38]). Animal models analysing maternal care in rats identified estrogen receptor α expression was altered with the type of maternal care, and this was passed across generations [39]. A detailed analysis of the role of these and other candidate genes implicated in ELS and later-life psychopathology is reviewed in the following sections.

ELS-induced epigenetic modifications in animal models

A variety of animal models are currently being used to model ELS paralleling childhood adversity in humans (Table 1). Each paradigm facilitates investigation into ELS-induced alterations in the developing animal and centres on the importance of mother for normal nervous, immune and endocrine system development [40–42]. A majority of the literature on animal models discussed in this review will therefore be on variations in maternal care.

Table 1 Commonly used models of early adversity in animal studies
Full size table

ELS-induced epigenetic changes in HPA axis genes

Given the central role of the HPA axis in stress responsivity and adaptation to ELS, the genes involved in regulating this system have been of focus in ELS-induced epigenetic studies (see Table 2 for summary of studies).

Table 2 Early stress-induced epigenetic changes in stress-regulatory genes in animal studies
Full size table

Glucocorticoid receptor gene

Pioneering studies on epigenetic alterations in GR promoter in response to variations in maternal care were first shown by Weaver and colleagues [20]. They reported increased methylation of the 5′ exon17 GR promoter and decreased H3K9 acetylation both associated with reduction in GR messenger RNA (mRNA) expression in the hippocampus of pups raised by low licking grooming arched-back nursing (LG-ABN) dams [20]. Extended studies showed that increased 5′ cytosine phosphate guanine (CpG) site methylation in the low LG-ABN pups reduced binding of transcription factor nerve growth factor inducible protein A (NGFI-A) to GR exon 17 promoter and reduced recruitment of CREB binding protein (CBP), subsequently reducing the levels of GR mRNA in hippocampus [20, 43]. These changes were observed both at postnatal day (PND) 6 (early) and PND90 (adulthood) suggesting the long-lasting nature of the epigenetic mark. In contrast, Daniels and colleagues reported no differences in the methylation status of exon 17 GR promoter in maternally separated (MS) compared to control rats on PND21 [44]. The conflicting results could be due to differences in the early stress model (maternal care vs MS) and strain (Long-Evans vs Sprague Dawley) which may exert different effects on the epigenetic signature of the glucocorticoid receptor. Other studies also report epigenetic alterations in the GR promoter using different ELS models of rat/mouse strains [45, 46] (Table 2).

Crh

Chen and colleagues reported hypomethylation of the Crh promoter in the PVN of maternally deprived (MD) Sprague Dawley rats on PND61 [47]. This was associated with increased phosphoCREB binding to the Crh cAMP response element (CRE), critical in the regulation of transcription of Crh. Similarly, Wang and colleagues reported increased H3 acetylation and hypomethylation of the Crh promoter region in the hippocampal cornu ammonis 1 (CA1) region of rats with postnatal MS [48]. Franklin and colleagues reported hypomethylation of the CRH receptor 2 (Crhr2) in maternally separated male C57/BL6 mice at 3–8 months of age and demonstrated a transgenerational effect [21]. The hypomethylation was associated with decreased Crhr2 gene expression which in this case was assessed in vitro using zebularine, a DNA methylation inhibitor [21]. This is in contrast to conventional understanding that DNA methylation is repressive [49, 50]; however, results correlate with the expected ELS-induced changes in HPA axis regulation and decreased Crhr2 expression.

Avp

Murgatroyd and colleagues reported hypomethylation of the AVP enhancer sequence in the parvocellular neurons of the PVN [23]. Hypomethylation of the AVP enhancer was associated with increased Avp expression following ELS (maternal separation) from 6 weeks of age and was still evident at 1 year suggesting the long-lasting nature of this epigenetic mark. Mechanistic analysis using mouse hypothalamic-like cells revealed that the hypomethylated CpG sites bound MeCP2 during postnatal life that in turn recruited DNA methyl transferase (DNMTs) and histone deacetylase (HDACs) to regulate expression of Avp [23] and that MeCP2 occupancy further depended on polycomb binding earlier in hypothalamic development [51]. Contrasting results were reported in a more recent study showing hypermethylation of the Avp enhancer CpG site in the hippocampus of maternally deprived C57BL/6 and DBA/2 males [45]. This could be due to the type of tissue analysed and the fact that the CpG sites analysed by these two studies did not overlap.

ELS-induced epigenetic changes in genes outside of the HPA axis

Brain-derived neurotrophic factor

ELS has been shown to decrease Bdnf expression whilst enhancing anxiety and depression-like behaviours [41, 52]. Roth and colleagues exposed Long-Evans rat pups to an abusive mother during the first week of life [53]. They reported hypermethylation of the Bdnf exons IV and IX in the prefrontal cortex on PND8 with associated reduction of Bdnf mRNA which persisted into adulthood. Bai and colleagues reported significantly decreased Bdnf mRNA and protein and increased miR-16 expression in the hippocampus of MD rats compared to those exposed to chronic unpredictable stress in adulthood (CUPS) and control rats [54] (see Table 3). The results suggested significant association of depression induced by MD to Bdnf and miR-16 levels but not the late-life stressors such as the CUPS thus emphasising the role of ELS-induced epigenetic alterations.

Table 3 Early stress-induced epigenetic modifications in other candidate genes
Full size table

Serotonin transporter (5-HTT or Slc6a4) gene

Reduction in 5-HTT expression has been observed in response to ELS in non-human primate and rodent models [55, 56]. Kinnaly and colleagues investigated the role of ELS-induced epigenetic modifications in the 5-HTT gene using non-human primate models [57, 58] (see Table 3). These studies assessed the relationship between 5-HTT expression and 5-HTT gene methylation following ELS in infant rhesus macaques [57] and adolescent bonnet macaques [58]. No significant effect of rearing was observed on the 5-HTT methylation status in either study. However, carriers of the short allele of the 5-HTT polymorphism had higher mean 5-HTT CpG methylation, and this was significantly associated to the levels of peripheral 5-HTT expression in the infant rhesus macaques [57]. In adolescent and adult Bonnet macaques, genome-wide methylation levels were associated with 5-HTT expression in those exposed to early stress as infants compared to controls [58]. The above studies suggest that 5-HTT gene methylation is not susceptible to variations in maternal care; however, polymorphisms in the 5-HTT gene and methylation at other cis or Trans regulating sites as may be conferring increased stress reactivity.

Estrogen receptor-α gene

Increased expression of estrogen receptor-α (ERα) in response to variations in maternal care was first reported by Champagne and colleagues using the LG-ABN rat model [39]. The study reported elevated ERα mRNA expression in the medial preoptic area of high LG-ABN dams compared to low LG-ABN, and this effect was transmitted to the female offspring of the high LG-ABN dams [39, 59]. Champagne and colleagues then demonstrated decreased ERα1b promoter methylation in high vs low LG-ABN offspring on PND6 [59] (Table 3). It was characterised that this hypomethylation of the promoter region enhanced binding of the STAT5 transcription factor to the ERα1b promoter and a corresponding increase in ERα mRNA expression in response to increased maternal care.

Glutamate decarboxylase-1 and Reelin genes

Glutamate decarboxylase 1 (GAD1) is a key enzyme in the synthesis of gamma amino-butyric acid (GABA), and Reelin (Reln) is important in migration of new neurons in the central nervous system (CNS). Postmortem studies of schizophrenic patient brains have shown decreased forebrain expression of GAD1 [60, 61] and increased methylation of the GAD1 promoter [62, 63] and decreased RELN expression [64, 65]. Zhang and colleagues assessed the methylation status of the Gad1 promoter in the offspring of the high vs low LG-ABN dams [66] (See Table 3). Hypomethylation, increased histone H3K9 acetylation and increased NGFI-A association with the Gad1 promoter were observed in those raised by high LG-ABN dams leading to increased Gad1 mRNA expression [66]. Qin and colleagues reported hypermethylation of the Reln gene and subsequent down-regulation of Reln mRNA in the hippocampus of Wistar rats exposed to MS from PND2–15 compared to controls [67].

Early stress-induced epigenetic modifications in humans

Epigenetic alterations in candidate genes

In human studies, most focus to date has been on the role of GR due to its negative feedback control in stress responsivity and serotonin transporter (5-HTT) due to its polymorphisms and role in mediating ELS and later-life stress effects on adult depression status [14, 68] (see Table 4 for summary of studies).

Table 4 Early stress-induced epigenetic modifications of candidate genes in humans
Full size table

GR gene

McGowan and colleagues were the first to demonstrate hypermethylation of the GR promoter exon 1F and decreased GR mRNA in the hippocampus of adults who were exposed to childhood abuse [19]. When compared to suicide victims or controls with no childhood abuse the GR promoter of suicide victims with a history of childhood abuse showed a significant increase (p < 0.05; d = 1.07) in methylation. This study was thus able to translate the results previously described in LG-ABN rat model [43]. McGowan and colleagues extended research by analysing a wider locus containing the GR gene on chromosome 5 and reported differential methylation in promoters of the protocadherin (PCDH) gene family which are implicated in synaptic plasticity [69]. Whilst McGowan and colleagues analysed the epigenetic changes in hippocampal GR gene, other human studies have reported similar hypermethylation GR gene promoter in peripheral blood leukocytes of adults exposed to early childhood stress [70, 71]. Murgatroyd and colleagues have recently demonstrated maternal stroking to modify CpG methylation within this GR region further translating the rat LG results [72]. These studies have shown that hypermethylation of specific CpG sites were associated to type of early stress or the severity of early maltreatment (see Table 4).

5-HTT gene

Philibert and colleagues were the first to report hypermethylation of the 5-HTT promoter and subsequent reduction in 5-HTT mRNA; however, this study did not assess the role of early maltreatment [73]. Further work from Beach and colleagues reported association of increased methylation of the 5-HTT promoter to childhood abuse. In their 2010 study, Beach and colleagues report significant association of overall methylation of the 5-HTT promoter region (CpGs analysed = 71, CpGs used in analyses = 26) to childhood abuse (p < 0.0004, d = 0.73) [74]. In particular, increased methylation of the sites CpG1 and CpG3 in the 5-HTT promoter was significantly associated in abused females. In their 2011 study, the same group showed significant association of childhood abuse to 5-HTT promoter methylation, 5-HTT mRNA levels and a correlation to 5-HTT genotype and adult anti-social personality disorder [75]. The study reported that 9 % of the variation in the personality disorder was attributed to increased 5-HTT promoter methylation observed only in those that were sexually abused (see Table 4 for details). Kang and colleagues also reported similar hypermethylation of the 5-HTT promoter (total of seven CpG sites) in abused compared to non-abused people with current MDD [76]. The average methylation percentage of the seven CpG sites for any adversity was significantly higher in those who were abused in childhood (p < 0.001, d = 1.1).

Genome-wide methylation

A number of studies have assessed the effect of early stress on genome-wide methylation patterns (See Table 5 for summary of studies). Naumova and colleagues demonstrated hypermethylation of 28 genes involved in brain development and function including those regulating the arginine vasopressin 1A receptor (AVPR1A), GABA A receptor (GABRA5), glutamate receptor (GRM5), among others, in 7- to 10-year-old children following institutionalisation [77]. Hypermethylation of candidate genes has also been shown in adulthood, with Labonte and colleagues reporting hypermethylation of 248 gene promoters and associated decreases in mRNA expression of these promoters in the hippocampi of adult men who completed suicide [78]. Genes responsible for cellular and neuronal plasticity were the most differentially methylated, including the alsin (ALS2) gene promoter. Many other studies have also shown significant global methylation differences between those exposed to early adversity [79–81]. These studies suggest that alterations in the early environment have the ability to cause changes in the methylation status of numerous genes across the genome, including those involved in control of nervous and immune system development and function.

Table 5 Effects of early stress on genome-wide methylation in humans
Full size table

Discussion

As evidenced in the studies reviewed above, changes in DNA methylation in response to the early environment remain the best characterised to date. This could be due to DNA methylation being a robust epigenetic mark and available to study across lifespan and generations [21, 82]. More recent studies however have also analysed the role of microRNAs [54], histone modifications [83] and DNA hydroxymethylation [84] in mediating early stress effects.

Results from the animal models reviewed suggest that ELS exerts significant methylation changes in the stress-regulatory genes. Enhanced Crh and Avp expression in the hypothalamus may lead to a hypersensitive HPA axis [23, 47], whilst reduced Crhr2 and GR expression (in the hypothalamus and hippocampus respectively) may affect negative feedback of the HPA response to stress [20, 21]. Epigenetic signatures in the GR gene promoter in humans exposed to a variety of early adversity translate the results observed in animal models and confirm the central role of GR methylation in the dysfunctional negative feedback of HPA axis [19, 70, 71]. Analysis of a broader GR locus (~7 Mpb centred with Nr3c1) in both rats and humans exposed to early adversity revealed conserved methylated sequences in the GR locus including the GR promoter methylations thus emphasising an evolutionary role of the GR locus in stress responsivity [46, 69] (See Tables 2 and 4 for details).

Studies of animal models have centred on other candidate genes implicated in the pathogenesis of neuropsychiatric disorders. Altered 5-HTT expression early in life may moderate the impact of early environment on emotion development [28] as also shown by pharmacological blocking studies [85, 29]. The findings on 5-HTT promoter methylation therefore support the role of reduced 5-HTT activity in the onset of adult depression-like behaviours [57, 58]. Reduced BDNF expression has been implicated in anxiety and depression-like behaviours [41, 52]; therefore, hypermethylation of the Bdnf gene alongside reduced Bdnf mRNA expression in the prefrontal cortex (PFC) of rats exposed to abuse (67) suggests early stress may play a role in the development of anxiety and depression disorders. In humans exposed to early adversity, 5-HTT hypermethylation has been associated to adult depression status depending on the polymorphic status of the 5-HTT gene, emphasising the role of genetic polymorphisms on epigenetic effects [14, 76]. Interestingly, studies to date have only analysed the effect of prenatal stress on the methylation status of the BDNF gene in humans [86, 87] and not variations in postnatal stress. BDNF is known to regulate neuronal development, serotonergic functions and signalling [88]. For example, BDNF promotes the development and function of serotonergic neurons where high affinity receptors for BDNF, Trkb, are also found [89]. Therefore, ELS-induced reduced BDNF could also potentially lead to decreased function of the serotonergic system thus leading to mood and affective behavioural dysfunctions as reported in the above studies. In addition, BDNF has been shown to be regulated by estrogen via ERα within the hippocampus of rat brains [90]. In this context, the increased ERα expression in the brains of the high LG-ABN rats [59] suggests that estrogen is important in regulating normal brain development and possibly involves maturation of neural systems via BDNF regulation. The down-regulation of Reelin post ELS [67] may negatively impact hippocampal function, potentially altering stress axis regulation and predisposing to anxiety and depression across the lifespan. A neurodevelopmental origin for schizophrenia is shown from the study suggesting increased methylation and reduced Gad1 expression in rats [66] which fits with the observations in postmortem brains of schizophrenic patients [62, 63].

One of the major questions in psychiatric epigenetics is whether epigenetic changes observed in the periphery reflect the changes in the brain. For example, hypermethylation of GR gene promoter and subsequent increase in GR mRNA was reported in both peripheral leukocytes [70, 71] and in the neurons of the hippocampus [19]. Analysis of methylome of cells in the PFC and T cells in mother vs surrogate-reared rhesus macaques revealed similarities in the methylation of cells in PFC and T cells, specifically the GR receptor promoter region alongside genes in the immune response, transcription and response to stimulus [91]. These findings suggest that peripheral GR methylation or cell specific (example T cells) methylation changes could be a potential biomarker for assessing early stress-induced HPA dysfunction. Genome-wide methylation studies show that childhood maltreatment leaves a systemic, genome-wide mark by altering the methylation status of key genes in regulatory pathways such as intra and extra cellular signalling [77, 79, 80, 92, 93]. Although most work to date has focused on the stress response, it may be interesting to consider whether prolonged immune dysfunction that often persists after early stress is secondary to epigenetic changes in immunoregulatory genes as reported by Smith and colleagues in PTSD [81] or if it results indirectly from epigenetic changes in stress-regulatory genes.

Conclusions

Research to date highlights the remarkable susceptibility of the genome, and particularly of genes involved in stress and emotion regulation, to environmental alterations early in the lifespan. Many of these changes persisted into adulthood, and epigenetic mechanisms have been shown to contribute to the long-lasting nature of ELS and its contribution to an individual’s disease risk and susceptibility to neuropsychiatric conditions across the lifespan.