Gene expression can be regulated by epigenetic processes. Two examples include coordinated epigenetic modifications of chromatin by DNA methylation and post-translational covalent modifications of histone proteins [4,11,12] and micro-RNA-induced suppression of gene expression during development [13]. DNA methylation is the most extensively investigated of the epigenetic mechanisms and involves the addition of a methyl group at the carbon-5 position of cytosine in CpG dinucleotides. While CpG dinucleotides are underrepresented in mammalian genomes overall, and usually exist in a methylated state, proximal gene promoter regions often overlap with CpG rich regions known as “CpG islands” that are typically unmethylated. In these regions, cytosine methylation serves a regulatory function [14]. By extending into the major groove of DNA, the methyl group of 5-methylcytosine (5-mC) interferes with transcription binding proteins, inhibiting transcription, and effectively silencing the gene [7]. More importantly, DNA methylation acts as a docking site for methyl-DNA binding proteins that recruit other chromatin remodeling proteins. The importance of methylation changes at non CpG islands is now being recognized [15]. Both hypermethylation and hypomethylation of DNA can result from exposure to exogenous chemicals. For some genes, even a small change in the level of DNA methylation at a few CpG sites might subtly alter gene expression and increase disease risk [16,17].

The epigenome is susceptible to dysregulation throughout life; however, it is thought to be most vulnerable to environmental factors during embryogenesis, which is a period of rapid cell division and epigenetic remodeling [16,18]. Following a complex choreography, following fertilization, DNA methylation patterns are largely erased and established early in mammalian development (reviewed in [7,19]). Fig. 1 illustrates the normal timetable for reprogramming of methylation of non-imprinted and imprinted genes during early development, beginning with the primordial germ cells (PGCs) of each of the parents (F₀) through gametogenesis, fertilization, the embryonic period of the offspring (F₁), followed by the maintenance of methylation in somatic cells and the development of germ cells that will become F₂ [7,20,21]. These dynamic stages represent windows of potential vulnerability to epigenetic dysregulation [7]. While the maintenance of imprinted genes throughout the preimplantation period is essential for normal embryonic development, demethylation of other genes is needed to make the genome broadly available to the developing embryo. Thus, after fertilization and prior to implantation, the embryo undergoes genome-wide demethylation, with the exception of imprinted genes (which retain the methylation profile of the parent-of-origin) and some retrotransposable elements [16]. Beginning when the embryo is in the blastocyst stage (starting day 5 post fertilization for humans) and before implantation into the uterine wall (about 7 days post fertilization), methylation patterns in non-imprinted genes are reestablished de novo by the DNA methyltransferases DNMT3a and DNMT3b and their cofactor DNMT3L [20,22]. DNA methylation patterns are maintained by DNMT1, which restores full methylation to hemi-methylated CpG sites following DNA replication; this maintenance is critical for normal development [20,23].

As noted, imprinted genes do not undergo genome-wide demethylation before implantation but maintain their methylation patterns throughout this period of reprogramming, allowing for the inheritance of parental-specific monoallelic expression in somatic tissues throughout adulthood [20]. Primordial germ cells (PGC) (the precursor cells that develop into spermatogonia and oogonia) have differential methylation by parent-of-origin at imprinted genes until they enter the genital ridge, when their DNA undergoes global demethylation of both imprinted and non-imprinted genes [24,21]. Remethylation of imprinted genes occurs in a sex-specific manner during gametogenesis (the division of gametocytes into haploid sperm and oocytes by meiosis) [19,20]. Imprints are established perinatally in the male germ line and are maintained throughout the mitotic divisions of the spermatogonial stem cells [20,21,48]. In the female germline, imprints are established during oocyte growth while they are arrested during the meiotic prophase I and are erased soon thereafter in the primordial germ cells of the next generation [25,26]. It can be seen from this brief summary that, prior to complete cell differentiation and the persistence of a stable epigenetic pattern, there is an opportunity for prenatal endogenous and exogenous exposures to alter the elaborate DNA methylation patterning required for normal tissue development [18]. Imprinted genes may be a particularly susceptible target for environmentally induced epigenetic effects [27]. The early developmental period is thought to be the most susceptible to epigenetic insults because the DNA synthesis rate is high and the elaborate DNA methylation patterning and chromatin structure required for normal tissue development is established at that time [28]. However, after birth, somatic cell methylation patterns continue to adjust in response to developmental and environmental factors [2,3,29].

In 1992, Barker and colleagues laid the groundwork for the “fetal basis of adult disease” (FEBAD) hypothesis, postulating that, because organs undergo developmental programming in utero that predetermines subsequent physiologic and metabolic adaptation during adult life, prenatal insults such as nutritional deprivation or environmental exposures that disturbed developmental programming could lead to a higher risk of disease in adulthood. They showed that abdominal fatness in adult men, an indicator of increased risk of cardiovascular disease and diabetes independent of body mass, was associated with retarded fetal growth, suggesting a persisting response to adverse conditions in fetal life [30]. Since 1992, the evidence has grown that developmental plasticity allows the fetus to make anticipatory responses to the external environment by altering the course of cellular and organ differentiation in utero in order to gain adaptive advantage for later life challenges [31,32]. However, a mismatch between the prenatal and the postnatal environment or synthetic environmental agents that mimic internal or natural cues can result in disease. The FEBAD hypothesis has been supported by evidence that fetal nutrient availability, other intrauterine factors, and external environmental factors can cause serious consequences in later life by permanently reprogramming the functional capacity of organs. Classical examples include the association of low or lower birth weight with increased risk of adult onset cardiovascular disease [31], type 2 diabetes mellitus, osteoporosis [33], depressive disorders [34] and certain cancers [35].

There is compelling evidence that epigenetic dysregulation underlies the observed associations between adult disease and adverse environmental/nutritional conditions early in development. For example, Heijmans and colleagues reported that individuals who were periconceptionally exposed to famine during the Dutch Hunger Winter in 1944–1945 had, six decades later, significantly less DNA methylation of the imprinted insulin-like growth factor II (IGF2) gene compared to their unexposed same-sex siblings [36]. IGF2 is a key factor in human growth and development and is maternally imprinted. Diseases that have been associated with early gestational exposure to famine include schizophrenia and coronary heart disease – diseases in which IGF2 may play a role.

In addition, a series of elegant studies in mice has shown that prenatal exposure to dietary methyl-donor supplementation with folic acid, Vitamin B12, choline, and betaine not only increased DNA methylation at specific CpG sites but also altered subsequent phenotypes such as coat color and obesity in the A^vy mouse model (reviewed in [7]). The fact that CpG sites were altered in tissues derived from the ectodermal, endodermal, and mesodermal lineages indicates that methylation profiles were changed early in embryonic development [7,18,37].

Another often-cited illustration of the importance of methylation changes attributable to environmental factors, albeit one involving neonatal exposure, is the work of Weaver et al. [38,39] showing that maternal stress and subsequent nurturing behaviors alter the epigenotype in rodent offspring, affecting their glucocorticoid receptor (GR) expression and behavior. The epigenetic changes could be reversed in adulthood by administering methionine or histone deacetylatase (HDAC) inhibitor. These epigenetic effects are not germline inherited but are passed on to the offspring directly from the mother through her behavior during the first week of postnatal life [40,29]. In a related study, newborns of mothers who had symptoms of depression during pregnancy had increased methylation of the glucocorticoid receptor gene in umbilical cord blood cells and the infants had elevated salivary cortisol concentrations at three month of age [41].

Bagot and Meaney conclude that epigenetic remodeling can occur both during early and later stages of development in response to environmental events that regulate development and function, with increased risk for psychopathology [42]. Most studies have focused on the influence of the maternal environment and maternal-infant interactions. However, recent evidence suggests that paternal factors (nutritional, toxicological, age, and phenotypic variation) can affect offspring and in some cases grandoffspring [43].

With respect to the lifecourse, a well studied example of an exogenous in utero exposure affecting adult disease is diethylstilbesterol (DES), the estrogenic pharmaceutical agent. This non-genotoxic, epigenetic carcinogen induced reproductive disorders and cancers in daughters exposed in utero and even in their granddaughters (reviewed in [44]). DES has been shown to alter gene methylation in mice exposed in utero suggesting that epigenetic mechanisms are involved [45].

With respect to prenatal exposures, there is an increasing body of evidence that diverse pollutants alter epigenetic programming and disease risk in the F₁ and even F₂ and F₃ generations. These include arsenic, tobacco smoke, air pollutants, and endocrine disrupting chemicals.

PAHs have been shown to be neurotoxic and affect gene expression in humans. Laboratory studies exposing experimental animals to PAHs during the prenatal and neonatal periods have reported neurodevelopmental and behavioral effects including depression-like symptoms and memory impairment in the absence of other overt toxicological effects [106–111]; others have shown that exposure affects neurotransmitter levels and gene expression patterns in the brain [80,112–114]. For example, prenatal treatment of rats with BaP impaired memory and ability to learn, consistent with alterations in the expression profile of glutamate receptor (GluR) subunits, which are key genes involved in long-term potentiation (LTP), considered the cellular correlate of learning and memory [81,111].

Many studies have shown that PAHs such as BaP are endocrine disruptors, affecting gene expression in hormonal regulatory pathways important in early brain development. Gene targets include the aryl hydrocarbon receptor (AhR) [115–117], CYP1A1 and CYP1B1 and CYP19A1; these genes are expressed in the fetal brain and peripheral lymphocytes [118–123]. In experimental studies, BaP caused alterations in levels of noradrenaline, dopamine, and serotonin and/or their metabolites in discrete brain regions [112–114]. In gestationally exposed rats, BaP caused significant reductions in expression of the N-methyl-D-aspartic acid (NMDA) glutamate receptor subunit NMDAR2B [81,111,124], and reduced LTP across the perforant path granular cells synapses in the hippocampus [125]. These data indicate that BaP and other PAHs disrupt the glutamate pathway and the dopaminergic and serotonergic systems, consistent with observed disturbances of learning and emotional behavior [113,125].

In addition, PAHs are immunotoxic contaminants known to affect expression of pro-inflammatory cytokines including interleukin-1beta, tumor necrosis factor-alpha (TNFα) IFN-γ and the chemokine CCL1 and [126,127]. Cytokines are also produced by CNS tissue and peripheral leukocytes [128]. These cytokines are among those most often implicated in neurodevelopment [129–131].

Finally, exposure to BaP upregulated COX-2, a key enzyme involved in inflammation and associated with reactive oxygen species (ROS) production, in rat astrocytes [132,133] in human cells and rats exposed to BaP or its main metabolite BPDE [133–136]. Residents of cities with severe air pollution had significantly higher expression of COX-2 in the frontal cortex and hippocampus compared to controls at autopsy [137].

Fewer studies have evaluated methylation changes due to PAHs than have looked at gene regulation or expression. However, in the CCCEH cohort study described above, a number of genes in addition to ACSL3 were found to be differentially methylated in UCWBC of newborns with high vs. low prenatal PAH exposure [88]. Several are known to be expressed in leukocytes and brain cells and have functions related to inflammatory and or immune pathways. They include CCL17 (a chemokine also known as TARC) that selectively induces migration of Th2 lymphocytes [138], which is expressed constitutively in thymus and in phytohemagglutinin-stimulated peripheral blood mononuclear cells, and has been found to be overexpressed in autistic brains [139].

The experimental studies described above have drawn an inferential link between PAH-related epigenetic alterations and neurodevelopmental effects, suggesting that alterations in methylation/gene expression mediate the neurodevelopmental effects of PAHs. For example, the observed neurodevelopmental effects of PAHs on learning and memory in humans [83] and the observation that prenatal treatment of rats with BaP impaired their ability to learn [111] are consistent with the observed disruptions of the glutamate pathway [111].

BPA is another endocrine disrupting chemical capable of exerting developmental effects. A recent epidemiological study has linked prenatal exposure to BPA with subtle, gender-specific alterations in behavior of 2-year olds [140]. Experimental evidence indicates that gestational exposure to environmentally relevant doses of BPA abrogates sexual dimorphism in brain structure and behavior and disrupts cognition, social behaviors, and other aspects of brain function [141–143]. Perinatal exposure to BPA altered sex differences in anxiety and depression-like responses in rodents [144–148]. In addition, male mouse offspring of dams treated from mating through weaning with low dose BPA exhibited impairment in memory [149]. Hyperactivity in male mice has also been demonstrated in response to perinatal exposure to BPA (females were not examined) [150]. Importantly, these behavioral changes are induced through low dose exposures to BPA in these experimental models. BPA treatment of pregnant female mice led to disruption in neocortical patterning in offspring during adulthood, possibly by accelerating neuronal differentiation and migration [151], and caused changes in gene expression in the fetal forebrain [152].

As is the case for PAHs, most mechanistic research on BPA, an estrogen-mimicking chemical, has focused on gene expression, rather than DNA methylation. BPA has been shown to interact with estrogen signaling pathways through binding to the estrogen receptors ERα and ERβ [153–155], and is also believed to interfere with non-classical estrogen signaling pathways at very low concentrations [156–158]. Studies in mice have shown that perinatal exposure to BPA can disrupt estrogen signaling in the offspring, with changes in gene expression at low doses [159]. Gestational exposure to BPA caused permanent upregulation of ERβ mRNA in the preoptic area of the hypothalamus in male rat offspring [160] and increased ERα and ERβ levels in the dorsal raphe nucleus of male mouse offspring of mice [161].

Prenatal BPA exposure has also been shown to interfere with other endocrine pathways. Prenatal BPA exposure dramatically increased expression levels of AhR in mouse embryonic cerebra, cerebella, and gonads, showing a U-shaped dose–response curve with extremely high response at the lowest dose tested [162]. Prenatal exposure to BPA also interfered with thyroid hormone, required for normal fetal and neonatal brain development [163–165], and increased expression of RC3/neurogranin in the dentate gyrus brain region, suggesting that BPA differentially affects the beta-thyroid receptor (TRβ) vs. the alpha-TR (TRα) [166] and changes the temporal expression patterns of TRα and TRβ [152]. Cell-based assays suggest that BPA blocks expression of TRβ [167,168].

Prenatal exposure to BPA affects immune cells and the expression of genes involved in inflammation such as IL-4 and INFγ [169,170]. BPA also affects methylation of genes involved in several immune pathways. For example, gestational exposure of rats to BPA caused differential methylation of Cebpα [171], which is involved in macrophage differentiation, cytokine signaling, microglial activation, and neuronal signaling [172–175]. As noted above, studies by Ho et al. found that PDE4D (involved in macrophage differentiation, neutrophil recruitment in inflammation, responses to oxidative stress, and possibly in regulation of dopaminergic neurotransmission [176–179]) was hypomethylated in the prostates of rats treated neonatally with BPA and also showed increased expression [35]. All of the genes mentioned above are known or believed to be expressed in both the brain and the blood cells.

A number of studies have attempted to link BPA-related changes in gene expression to neurodevelopmental outcomes. In utero treatment of mice with BPA eliminated sex differences in the size of the anteroventral periventricular preoptic (AVPV) area, significantly reduced the number of TH (tyrosine hydroxylase) positive neurons in the female AVPV, and abrogated sex differences in the number of TH-positive neurons [143]. In the same study, BPA reduced sexual dimorphism in anxiety-related behaviors in the open-field test. Finally, BPA treatment from mating through weaning both increased levels of neurotransmitter-producing choline acetyltransferase in the hippocampus of male offspring and affected performance on the step-through test, indicating memory impairment as a consequence of BPA exposure [149].

There is growing evidence that environmental variations experienced by both fathers and mothers may lead to phenotypic variation in the development and behavior of offspring resulting from transmission through the germline [182]. Transmission can result either from altered programming within germ cells of the epigenome of the retrotransposons and imprinted genes or through altered expression of RNA within gametes. “Transgenerational epigenetic inheritance” refers to the transmission of a biological trait to subsequent generations via epigenetic modifications in the germline [7]. As elaborated by Jirtle and Skinner [7], in order to consider transgenerational effects on the epigenome to be a plausible mechanism for a disease phenotype, the epigenetic changes and the disease phenotype must be observed in the F₃ generation. This is because gestational exposure of an F₀ female directly exposes both the F₁ embryo and the F₂ germline. Therefore, phenotypes in the F₁ and F₂ generations may be due to their direct exposure to the environmental factor rather then germline transmission (see Fig. 2).

With respect to genetic susceptibility, as noted above, Breton and colleagues observed a significant interaction between the GSTM null genotype of the child and the prenatal exposure to tobacco smoke on global methylation in the child [71]. Foley et al. [16] have reviewed other genetic factors that directly affect DNA methylation. These include the C-to-T substitution at nucleotide 677 of the methylene tetrahydro-folate reductase (MTHFR) gene: TT homozygous individuals have lower levels of DNA methylation than CC homozygous individuals. In addition, variants of the DNMT gene family discussed earlier have been associated with diseases including cancer [190]. As another example, a specific variant (C > T) in the O-6-methylquanine-DNA transferase (MGMT) tumor suppressor gene has been associated with O-6-methylguanine-DNA methyltransferase promoter methylation and gene silencing. Wright [191] has reviewed the evidence that psychosocial stressors and physical environmental toxicants play a joint (synergistic) role in disrupting immune and endocrine pathways involved in respiratory and cognitive development and function. She notes that oxidative stress pathways that may influence health are disrupted by both psychological stressors and environmental pollutants such as tobacco smoke and air pollution, all of which generate reactive oxygen species. Further, by causing dysregulatory behavioral states such as depression and anxiety, psychosocial stressors may produce long-lasting effects on shared physiologic processes and thereby increase risk from pollutant exposures [192]. Wright suggests that both factors may be acting through early life reprogramming of the hypothalamic-pituitary-adrenal (HPA) axis and the autonomic nervous system which are particularly susceptible to both stress and physical environmental toxicants.