Sex differences in prenatal epigenetic programing of stress pathways

Abstract

Maternal stress experience is associated with neurodevelopmental disorders including schizophrenia and autism. Recent studies have examined mechanisms by which changes in the maternal milieu may be transmitted to the developing embryo and potentially translated into programing of the epigenome. Animal models of prenatal stress have identified important sex- and temporal-specific effects on offspring stress responsivity. As dysregulation of stress pathways is a common feature in most neuropsychiatric diseases, molecular and epigenetic analyses at the maternal–embryo interface, especially in the placenta, may provide unique insight into identifying much-needed predictive biomarkers. In addition, as most neurodevelopmental disorders present with a sex bias, examination of sex differences in the inheritance of phenotypic outcomes may pinpoint gene targets and specific windows of vulnerability in neurodevelopment, which have been disrupted. This review discusses the association and possible contributing mechanisms of prenatal stress in programing offspring stress pathway dysregulation and the importance of sex.

Keywords::

• Brain
• masculinization
• neurodevelopmental disorders
• placenta
• prenatal stress
• sex differences

Clinical research and epidemiological studies have linked exposure to stressful factors during pregnancy with an increased offspring disease risk. Animal models examining possible mechanisms through which such fetal antecedents contribute to disease susceptibility have identified sex and temporal specificity in the programing of offspring stress pathways (Bale et al. 2010; Dunn et al. 2010). The complex interactions between the maternal environment and the developing embryo make the placenta an intriguing candidate tissue in which potential contributing factors can be examined. Further, as most neurodevelopmental diseases exhibit a sex bias in presentation, treatment, or severity, examination of sex-specific embryonic and placental responses following maternal stress has become a recent focus in animal models. In this review, we will discuss sex differences and potential epigenetic mechanisms generated in the prenatal environment via maternal stress, with an emphasis on the novel and important contribution of the placenta as a critical sex-specific endocrine tissue serving as the gatekeeper between the maternal and embryonic compartments.

Central programing effects of prenatal stress

Stress pathway dysregulation is the most pervasive symptom in neuropsychiatric disease, making elucidation of the developmental programing and maturation of this system and the sensitive periods during which perturbations may be disruptive of critical importance. Studies have linked maternal stressors including natural disasters, death of a family member, and reported levels of maternal anxiety or depression with an increased incidence of neurodevelopmental disorders including depression, anxiety, schizophrenia, and autism (Myhrman et al. 1996; van Os and Selten 1998; Buka et al. 2000; Goldstein et al. 2000; Beversdorf et al. 2005; Khashan et al. 2008, 2009; Kinney et al. 2008; Li et al. 2009), as well as impacting brain development and cognitive abilities (King and Laplante 2005; King et al. 2009) in the offspring. In addition, the temporal specificity of maternal stress experience across pregnancy has been shown to be a factor in offspring disease risk (Huttunen and Niskanen 1978).

Detailed examination of animal models of maternal stress has provided unique insight into the programing of long-term offspring outcomes relative to timing in the development in which the exposure occurred. Further, as in humans, offspring outcomes also varied depending upon the stressor utilized (Clarke and Schneider 1993; Kapoor and Matthews 2005; Cottrell and Seckl 2009; Brunton and Russell 2010). Across species including mice, rats, guinea pigs, and nonhuman primates, prenatal stress has been shown to increase offspring hypothalamic–pituitary–adrenal (HPA) stress axis sensitivity, anxiety, and depressive-like behaviors, and cognitive deficits; and all endophenotypes that have been associated with neuropsychiatric diseases (Lemaire et al. 2000; Weinstock 2001; Schneider et al. 2002; Kapoor and Matthews 2005; Mueller and Bale 2007, 2008; Darnaudery and Maccari 2008; Kapoor et al. 2008, 2009).

Studies examining long-term consequences of prenatal stress have recently become more focused on the determination of a developmental window of vulnerability. Although prenatal stress is associated with an increased risk for neurodevelopmental disorders, the brain is developing over the course of gestation and therefore likely has points of greater sensitivity. To compare maternal stress experience across early, mid, and late stages of pregnancy on programing of offspring stress regulation, our laboratory utilized a chronic variable stress paradigm in mice (Mueller and Bale 2007, 2008). We found that stress early in pregnancy increased physiological and behavioral stress responsivity specifically in male offspring as adults (Mueller and Bale 2008). These changes in behavior were further associated with an increased expression of amygdala corticotropin-releasing factor (CRF) and reduced hippocampal glucocorticoid receptors (GR) in these mice. DNA methylation patterns for these genes were also altered by this early period of prenatal stress only in male offspring (Mueller and Bale 2008). Interestingly, similar outcomes have been previously reported in the effects of late gestational glucocorticoid exposure and following early postnatal stress on GR expression (Liu et al. 1997) and methylation (Weaver et al. 2004a) as has been elegantly reviewed (Nyirenda and Seckl 1998; Weaver et al. 2004b; Cottrell and Seckl 2009), suggesting potential common mechanisms and points of vulnerability. A recent study reported similar changes in hippocampal GR expression and methylation in human suicide victims who had suffered early life child abuse stress (McGowan et al. 2009).

Stress experience has been shown to directly impact brain development, especially the hippocampus and hypothalamus, an effect associated with increases in adult stress responsivity (Takahashi et al. 1990; Takahashi and Kalin 1991; Weinstock et al. 1992; Koehl et al. 1999; Charil et al. 2010). The hippocampus is a brain region critical in the regulation of stress pathways, and has been a particularly interesting target in the examination of prenatal and early life stress effects on hippocampal volume, neurogenesis, and epigenetic gene regulation in rodents and nonhuman primates (Uno et al. 1989; Coe et al. 2003; Korosi and Baram 2010; Korosi et al. 2010). As evidence of an impact on hippocampal function, prenatal stress also produces sex- and time-specific effects on offspring performance in hippocampal-dependent spatial learning and memory tasks including the Morris Water and Barnes mazes (Mueller and Bale 2007; Yaka et al. 2007; Kapoor et al. 2009). Although less is currently known as to how maternal stress produces such long-term changes in offspring stress pathways, studies have detected increases in both placental and hippocampal CRF production (Mairesse et al. 2007; Seckl and Holmes 2007; Charil et al. 2010), which may be activating brain CRF receptors and sensitizing them well into adulthood, altering dendritic arborization of this neuronal population (Ivy et al. 2010). The specificity of prenatal stress effects on the development of these critical brain areas has been recently elegantly reviewed (Charil et al. 2010), as it has the importance of the orchestration of these brain regions in regulating stress pathway function (Joels and Baram 2009).

Sex as a factor in disease susceptibility

In addition to the temporal specificity, offspring sex is also a critical factor in the determination of maternal stress long-term impact. Epidemiological studies linking fetal antecedents with long-term disease risk have established that gender is an important determinant in disease severity and onset. As an example, studies have reported that pregnant mothers exposed during the second trimester to the stress of the 1940 invasion of the Netherlands had male but not female offspring with an increased risk of schizophrenia as adults (van Os and Selten 1998). Further, many neurodevelopmental disorders have a sex bias, such as autism spectrum disorders (ASD) with an overall sex ratio of 4.3:1 for boys-to-girls (as reviewed in Newschaffer et al. (2007)). However, when separated for cognitive impairment, this bias is even further increased for ASD without mental retardation to 5.5:1, suggesting that distinct underlying mechanisms or predisposing factors may be involved. As exposure to maternal stress prior to 32 weeks gestation has been suggested as a potential contributing factor to ASD (Beversdorf et al. 2005), understanding the role that sex plays in the specificity of response to stress during the development may provide unique insight as to disease etiology. A recent report detected a significant effect of maternal depression during pregnancy on offspring postnatal anxiety development, particularly in males (Gerardin et al. 2010). These studies support both sex and temporal specificity in the association between maternal stress and offspring disease. Studies in animal models of prenatal stress have begun providing interesting clues as to the timing and mechanisms involved in how perturbations in the maternal milieu may have sex specificity in their effects on the developing brain. A more exhaustive discussion covering sex differences in neuropsychiatric disease has been recently reviewed elsewhere (Bale et al. 2010).

The sexually dimorphic brain and stress

The disruption of sex-dependent processes is a common theme in the prenatal stress literature. In the 1970s, the Ward laboratory produced an impressive body of work showing dysregulation of male typical behaviors following prenatal stress in rats in defining a “prenatal stress syndrome” (Ward 1972; Meisel et al. 1979). Sex differences in neurodevelopment are the result of both genetic sex and circulating gonadal hormones. Late gestational restraint stress has also been shown to disrupt the organizational perinatal testosterone surge, reduce adult testis size, shorten anogenital distance, and block the differentiation of sexually dimorphic brain nuclei in males (Ward 1972; Meisel et al. 1979; Weisz et al. 1982; Reznikov et al. 1999). In addition to the dysmasculinized stress responsivity and cognition previously reported for early prenatally stressed male mice, we found that these males also have reduced testis size and testosterone levels and a shortened anogenital distance, supporting a disruption of the normal testosterone signaling during the critical window in development. Results from current studies support an intriguing potential common target whereby prenatal stress may interfere with gonadal hormone programing, thereby disrupting normal sex differences (Ward and Ward 2009). There is evidence in humans linking prenatal testosterone levels with endophenotypes of ASD including impaired social behavior and reduced cognitive abilities (Jacklin et al. 1988; Knickmeyer et al. 2005, 2006) as well as fear reactivity in boys (Bergman et al. 2010). Further, as amniotic fluid cortisol and testosterone are positively correlated (Gitau et al. 2005; Sarkar et al. 2007a,b), there may be an important connection between maternal stress and regulation of offspring androgen production. One proposed mechanism suggested that this could occur through an impact of maternal stress on embryonic leydig cell 3β-hydroxysteroid dehydrogenase activity, thus altering the testes testosterone production and shifting the critical window of sexually dimorphic brain programing (Orth et al. 1983). Similar effects have been reported with chronic stress reductions in circulating testosterone in adult male rats (Gray et al. 1978).

Results from animal models and clinical epidemiological reports on male-biased neuropsychiatric disorders have pointed to a potential underlying “dysmasculinization” of the male brain, suggesting a development window in which normal sexual differentiation does not occur due to disruption of early sex-determining factors. Normal development of the sexually dimorphic brain occurs during the late gestational and early postnatal testosterone surge produced by the male testes (Phoenix et al. 1959), and as reviewed in Arnold (2009). In rodents, it has been well established that testosterone exposure via aromatization to estradiol in males is critical for the development of the sexually dimorphic brain, influencing masculinization of reproductive and stress behavioral neurocircuitry (as reviewed in McCarthy (1994) and Becker et al. (2007)). Recent evidence points to an additional important role of reduced testosterone to 5α-dihydrotestosterone (DHT) and binding of the androgen receptor in this process as well (Sato et al. 2004; Weinstock 2007; Bodo and Rissman 2008; Zuloaga et al. 2008). Although we have a clear appreciation for the importance of gonadal hormones in the organization of the sexually dimorphic brain, especially in animal studies focusing on reproductive behaviors, the specifics by which this hormonal milieu programs the developing stress neurocircuitry is less clear.

In rodent models, studies have taken advantage of the ability to manipulate the early postnatal critical window to masculinize females and examine organizational effects on stress responsivity (Bingaman et al. 1994; Young et al. 2001). The importance of neonatal masculinization in programing of the male HPA stress axis has been related to adult androgen receptor levels (Bingham and Viau 2008). Further evidence of this important window in stress pathway development is in the ability of neonatal female rat androgen exposure to blunt adult HPA axis response to stress (Seale et al. 2005). We recently examined physiological and behavioral responses to stress and the possible influence of these hormones on long-term changes in gene expression in adult female mice masculinized at birth (Goel and Bale 2008). Similar to outcomes reported with prenatal stress exposure, hippocampal GR expression was altered in female offspring masculinized at birth, supporting the importance of the male testosterone surge in normal wiring of these pathways. Early postnatal life is a sensitive period during which hormone exposure can have organizational effects that alter serotonin system maturation (Dominguez et al. 2003a,b), potentially leading to long-term changes in stress sensitivity related to active coping strategies. In addition, the rise in testosterone beginning in puberty exerts modulatory actions on neurotransmitter systems critical in regulation of stress physiology and coping including serotonin and γ-aminobutyric acid signaling (Bitran et al. 1993). Therefore, the coordinated impact of masculinization at both time periods may interact with the components of sex chromosomes to orchestrate a complete “normal” male phenotype. Disease predisposition may then be the result of fetal antecedents that would disrupt normal testis development, testosterone production, or downstream activation or responses to gonadal hormones resulting in a “dysmasculinized” male. However, as early pregnancy stress produces such effects but occurs prior to sexual differentiation, brain development, or the presence of a gonad, we are still searching for targets that may be important during early gestation.

Sex chromosomes: Y it matters

Expression of the sex-determining region Y (SRY) gene in males is critical for the initiation of male sex determination, testis development, and the production of testosterone (as reviewed in Sekido and Lovell-Badge (2009)). Recent studies have focused on elucidating the mechanism of SRY action and determination of the regulation of this gene during early development. In mice, the ability of SRY to induce full testis development was recently shown to occur over a very narrow time window of 6 h (11–11.25 dpc) (Hiramatsu et al. 2009), supporting a very tight regulation of the activity of SRY. SRY gene expression is also restricted to 10.5–12.5 dpc in mice. This regulation is controlled by DNA methylation of the SRY promoter in which prior to 10.5 dpc the gene is hypermethylated preventing its expression, then hypomethylated at 10.5 dpc when expression is turned on, and again hypermethylated at 12.5 dpc (Nishino et al. 2004). The downstream cascade following SRY expression includes an important transcriptional activation of SOX9 in males. In females, the absence of SRY during the critical period results in the accumulation of β-catenin and the repression of SOX9 and granulosa cell differentiation (Sekido and Lovell-Badge 2008, 2009). This tightly controlled regulation of SRY expression by epigenetic mechanisms and its timing in sexual differentiation support a vulnerable point in the development when fetal antecedents such as maternal stress could alter the genes involved in this process, thus influencing the long-term patterning of the sexually dimorphic brain. How prenatal stress could impact or reprogram this cascade is not known. However, as SRY expression is tightly regulated by methylation of its promoter, disruption of epigenetic regulation required for full SRY expression may be an interesting target.

Surprisingly, for all we know about SRY regulation and requirement during sexual differentiation, we know little as to the “dose response” requirement of SRY levels during this window—whether slightly more or less SRY would predict a change in testis size and testosterone production. In a C57BL/6J mouse carrying a Y chromosome from Mus musculus domesticus, the XY progeny develops ovaries or ovotestes during fetal life (Lee and Taketo 1994). In examining the interaction of SRY with autosomal components in these mice, it was discovered that the prolonged expression of SRY outside the normally very limited window was involved in the phenotype, supporting the importance of both the activation and inactivation of this gene (Lee and Taketo 1994). DNA-binding domains for SRY have been characterized in the promoters of the sex-specific genes encoding P450 aromatase and Mullerian-inhibiting substance (MIS) (Haqq et al. 1993). P450 aromatase catalyzes the conversion of testosterone to estradiol. Conversely, MIS is expressed in the male embryo to induce testicular differentiation and regression of female reproductive ducts. Therefore, SRY becomes a potential gene target of dysregulation by fetal antecedants that could alter male offspring masculinization—resulting in “dysmasculinization”. Interestingly, if such effects were programed in the epigenome, the phenotype could be passed on to future male generations. Studies are underway in our laboratory and others to test some of the hypotheses proposed here for SRY involvement in the prenatal stress-mediated dysmasculinization of male offspring. However, currently little has been examined.

Prenatal stress and the placenta: Where sex differences begin?

Earlier gestational periods have become a focus in epidemiological reports and animal models examining epigenetic programing associated with maternal stress (Khashan et al. 2008; Mueller and Bale 2008; Bale et al. 2010). The developing placenta serves as an intriguing candidate tissue for mechanistic examination as it is rapidly developing during this period and is the critical messenger between the maternal environment and the developing embryo. We have just begun to determine the sex-dependent mechanisms by which maternal stress could be influencing offspring outcome at an early stage of development through actions on the placenta (Rossant and Cross 2001; Hemberger 2007; O'Donnell et al. 2009). The intrauterine environment is constantly responding to changes in the maternal milieu via the placenta, and thus susceptible to insults such as maternal stress or infection. The placenta is dynamically developing during early pregnancy (early first trimester in humans), making this period of gestation highly sensitive. In humans, the onset of maternal arterial circulation as early as the end of the third week post-conception leads to a three fold rise in intra-placental oxygen (Rodesch et al. 1992). Placental tissue is largely sex specific as the maternal contribution is substantially less than that from the developing embryo, and as such has XX or XY genotype (Rossant and Cross 2001). In most placental mammals, X inactivation is random. However, in rodents, extraembryonic X inactivation is thought to be paternally imprinted (Graves 2010). Interestingly, recent studies have shown that at least 150 loci escape inactivation can be expressed from both X chromosomes (Carrel and Willard 2005). Therefore, studies in mouse and rat models demonstrating sex differences in placental responses to changes in the maternal environment may indicate an involvement of these escaped genes, in which female placentas would express twice the level of male's of a given gene. Examples of paternally expressed (maternally imprinted) genes in the mouse placenta, including Peg10 which is predominantly expressed in the placenta in mouse and human, also support exceptions to the complete maternal gene control (Rawn and Cross 2008). Alternatively, the few genes expressed on the Y chromosome may be involved. Several expressed genes on the Y chromosome code for proteins that are male specific, and may cause an immune reaction in females at the maternal–placental interface (Graves 2010), especially for placenta-specific genes (Rawn and Cross 2008). The placenta, as the endocrine tissue at which the maternal hormonal milieu can directly impact the developing embryo, is poised to be influenced by perturbations occurring early in pregnancy that could affect fetal development throughout pregnancy by altering (1) nutrient and oxygen transport, (2) inflammatory responses, and (3) epigenetic programing machinery (Mairesse et al. 2007; Seckl and Holmes 2007; Mueller and Bale 2008; Hirst et al. 2009; Lucassen et al. 2009; Bale et al. 2010). Certainly, sex differences in these placental responses could present a bias in long-term offspring outcome.

In recent studies focusing on the effects of early maternal stress in mice, we found sex differences in placental gene expression utilizing a focused PCR array for growth factors and nutrient transport genes. Interestingly, only male placentas exhibited significant increases in the expression of several genes important in growth and development including peroxisome proliferator-activated receptors α (PPARα), insulin-like growth factor-binding protein 1 (IGFBP-1), GLUT4, and HIF3α (Mueller and Bale 2008). As glucocorticoids increase PPARα expression (Lemberger et al. 1994), and PPARα regulates expression of IGFBP-1, maternal stress may directly alter these genes (Degenhardt et al. 2006). However, it remains unclear as to why this effect was specific to male placentas within the same uterus as female placentas did not respond (Mueller and Bale 2008). An elevation in placental IGFBP-1 could potentially decrease the available growth factors during critical developmental periods, and plays a role in fetal programing and brain development which were specific to males (Myatt 2006). Certainly, decreased growth-factor levels have been linked to affective and neurodevelopmental disorders, and IGFBP-1 is known to down-regulate genes involved in embryonic growth (Kajimura et al. 2005; Malberg et al. 2007).

Mechanistically, what influence maternal stress may have on programing of the placenta or embryo related to neurodevelopment is intriguing. Long-term changes in transcriptional regulation by epigenetics, including DNA methylation, histone acetylation, and noncoding RNAs, are likely involved (Weaver et al. 2004a,b; Drake et al. 2005). Tight control of placental and embryonic epigenetic machinery is critical during the early phase of gestation when a wave of genome demethylation prior to de novo re-methylation by DNA methyltransferases (DNMT), establishment of imprints, and sex determination occurs, identifying this period of early development as highly vulnerable to maternal disturbances that could result in embryonic reprogramming (Weaver 2009a,b). For instance, inflammatory cytokines can directly affect levels and activity of DNMT (Hodge et al. 2001), as well as regulate placental receptors and transporters for folate (Yasuda et al. 2008; Xia et al. 2009), an important methyl donor shown to regulate levels of methylation of nonimprinted genes during pregnancy (Waterland 2003; Waterland and Jirtle 2003; Lillycrop et al. 2005; Burdge et al. 2009). In our early prenatal stress studies, we reported a sex-specific effect of stress on placental epigenetic machinery in which maternal stress increased expression of male DNMT1 and the methyl-binding protein, MeCP2 (Mueller and Bale 2008). Previous studies in mice have reported that regulation of placental methylation patterns is predictive of similar embryonic changes critical in neurodevelopment (Rivera et al. 2008). Recent studies examining the transgenerational effects of early life stress support an involvement of stress regulation on these genes in which similar changes were reported for MeCP2 expression and methylation in the brains of first and second generation stressed mice (Franklin et al. 2010).

In search of predictive biomarkers

Despite the profound effect that neurodevelopmental diseases have on society, we have yet to identify predictive biomarkers and lack insight into underlying causes (Insel 2010). Although reports have illustrated changes in gene expression and epigenetic marks in the brains of animal models and postmortem human disease that correlate with behavioral phenotypes, the ability to obtain human neural tissue prior to disease onset or death is unlikely. Therefore, it is important to determine the potential value of tissues, such as the placenta or amniotic fluid, that are easily accessible and will produce the least disruption and risk to the developing fetus. Both amniotic fluid and placental tissue have been utilized in prenatal screening for genetic and developmental diseases for decades. Epigenetic, proteomic, and gene array analyses of these tissues in animal models may provide our best hope yet at establishing unique markers across pregnancy that could be examined in a translational setting. Further, identification of sex differences in these screens could provide insight into sex-dependent developmental points of vulnerability during gestation; how can we know when something has gone wrong if we do not know what normal looks like? Epidemiological studies examining risk factors for ASD and schizophrenia have reported many potential markers associated with disease predisposition including increased proinflammatory cytokines from maternal infection, stress hormones, gonadal hormones, and environmental neurotoxins (Newschaffer et al. 2007). However, much less is known as to how or when the developing embryo responds to these environmental exposures or the factors that may underlie individual susceptibility. Animal studies using maternal high-fat diet as a determining factor in offspring disease susceptibility recently reported that placental markers for oxidative stress were elevated and pretreatment with an antioxidant reversed these outcomes (Liang et al. 2010). Although these markers have not yet been replicated in human gestational diabetes outcomes, identification of such reproducible biological changes in relevant animal models puts us one step closer. Examination of potential placental biomarkers in predicting pre-eclampsia has developed an array of vascular transcription factors and inflammatory responses that may provide insight as to which mothers are likely to present (Aris et al. 2009; Redman and Sargent 2009). Unfortunately, the evidence we have for predictive biomarkers for neurodevelopmental disorders such as ASD or schizophrenia is not there yet. As recent reviews have defined in the search for genetic markers in schizophrenia, we are likely looking for “multiple variations, in multiple genes, in combination with environmental stressors”, the complexity of which adds to the dearth of candidates we currently hold (Schwab and Wildenauer 2009). Certainly, caution should be exercised in the potential for translation of extra-embryonic tissue biomarkers with limitations of pregnancy access and risk adding a level of difficulty for obtaining such endpoints. The exhausted post partum placenta may be less likely to reveal insight into environmental exposures or changes in the maternal milieu that occurred during key points in gestation.

Summary

Now that the field has come to a consensus as to the importance of the environment and the need to include epigenetics, as well as genetics, in understanding these diseases, what is next? Although epidemiological studies have produced strong evidence linking maternal stress experience with an increased offspring disease risk, we still lack sufficient understanding of the contributing mechanisms involved. How do we get beyond the “association” and begin to really identify the target genes involved in the programing of disease? Further, as new studies have begun to highlight the involvement of the epigenome in programing of transgenerational outcomes, the search for biomarker identification will likely become even more difficult as we begin to grasp the continued environmental involvement in each generation (McCarthy et al. 2009). What we do know is that neurodevelopmental diseases have significant sex biases in presentation, severity, and treatment. Therefore, studies examining the development and programing of the sexually dimorphic brain and sex differences at multiple levels and time points in development in response to maternal and early life stress remain of critical importance.

Acknowledgments

The author would like to thank G. Dunn and C. Morgan for assistance with the manuscript. This work was supported by funds from NIH MH073030, MH087597, and MH091258.

Declaration of interest: The author reports no conflicts of interest. The author alone is responsible for the content and writing of the paper.