Abstract
The epigenome serves as an interface between the dynamic environment and the inherited static genome. The epigenome is comprised of chromatin and a covalent modification of DNA by methylation. The epigenome is sculpted during development to shape the diversity of gene expression programs in the different cell types of the organism by a highly organized process. Epigenetic aberrations have similar consequences to genetic polymorphisms resulting in variations in gene function. Recent data suggest that the epigenome is dynamic and is therefore responsive to environmental signals not only during the critical periods in development but also later in life as well. It is postulated here that not only chemicals but also exposure to social behavior, such as maternal care, could affect the epigenome. It is proposed that exposures to different environmental agents could lead to interindividual phenotypic diversity as well as differential susceptibility to disease and behavioral pathologies. Interindividual differences in the epigenetic state could also affect susceptibility to xenobiotics. Although our current understanding of how epigenetic mechanisms impact on the toxic action of xenobiotics is very limited, it is anticipated that in the future, epigenetics will be incorporated in the assessment of the safety of chemicals.
EPIGENETICS AND INTERINDIVIDUAL DIFFERENCES IN RESPONSE TO XENOBIOTICS
The genetic revolution has focused on unraveling the sequence of the genetic material of humans and other organisms. It is commonly held that the sequence holds the secret for the phenotypic diversity of humans and their susceptibility to disease. Thus, the major effort in mapping the human haplotype map. This genome-focused notion also guided our assessment of the potential health hazards of different agents. The basic supposition in the field has been that the interindividual variations in response to xenobiotics are defined by genetic differences and that the main hazard anticipated at the genomic level from xenobiotics is mutagenesis or physical damage to DNA. In accordance with this basic hypothesis, the main focus of attention in pharmacogenetics has been on identifying polymorphisms in genes encoding drug metabolizing enzymes and receptors. New xenobiotics were traditionally tested for their genotoxic effects. However, it is becoming clear that epigenetic programming plays an equally important role in generating interindividual phenotypic differences, which could affect drug response. Moreover, the emerging notion of the dynamic nature of the epigenome and its responsivity to multiple cellular signaling pathways suggests that it is potentially vulnerable to the effects of xenobiotics not only during critical period in development but later in life as well. Thus, nongenotoxic agents might affect gene function through epigenetic mechanisms in a stable and long-term fashion with consequences, which might be indistinguishable from the effects of physical damage to the DNA. Epigenetic programming has the potential to persist and even being transgenerationally transmitted (Anway et al., 2005) and this possibility creates a special challenge for toxicological assessment of safety of xenobiotics.
The combinations of mechanisms, which confer long-term programming to genes and could bring about a change in gene function without changing gene sequence are termed epigenetic. Epigenetic programming occurs during development to generate the complex patterns of gene expression characteristic of complex organisms such as humans, however, epigenetic programs in difference from the genetic sequence itself are somewhat dynamic and responsive to different environmental exposures during fetal development as well as early in life. Although it is assumed that the majority of epigenetic marks established during embryonal development remain stable through life, a fraction of these marks are potentially dynamic even later in life. Thus, many of the phenotypic variations seen in human populations might be a result of disparity in long-term programming of gene function rather than the sequence per se. Any analysis of interindividual phenotypic diversity should therefore take into account epigenetic variations in addition to genetic sequence polymorphisms (Meaney and Szyf, 2005b). For example, differences in expression between individuals in the level of activity of a p450 enzyme or a DNA repair enzyme might result from genetic polymorphism but might as well be derived from altered epigenetic programming. Moreover, it is possible that epigenetic processes might override genetic polymorphisms.
Epigenetic variations could potentially be established at distinct points in life and in specific tissues exclusively. This has implications on drug action and toxic effects of xenobiotics since cell-specific epigenetic variation could result in differential pharmacodynamics, pharmacokinetics, and toxicity of drugs in different tissues. Thus, whereas a germ-line polymorphism is a static property of an individual and might be mapped in any tissue at any point in life, epigenetic differences must be examined at different time points and at diverse cell types. It is clear that we need to develop the capacity to address this daunting task.
Some critical environmental exposure could alter the progression of epigenetic programming during development both in utero as well as postnatally. Thus, variation in environmental exposures during these critical periods could result in epigenetic and therefore phenotypic differences later in life. It stands to reason therefore that exposure to nutritional deprivation and chemical toxins would affect the epigenetic machinery during development. Recent data suggest that in addition to these physical exposures, psychosocial exposures early in life could also impact on the epigenome resulting in differential epigenetic programming and as a consequence in behavioral variations later in life (Meaney and Szyf, 2005a). Thus, certain behavioral pathologies might be a consequence of early life exposures, which altered epigenetic programming. We therefore need to consider the possibility that early life social environment might have an impact on responsivity to drugs later in life through epigenetic programming of critical genes involved in drug responsivity. Although these notions are purely speculative at this stage, such epigenetic mechanisms should be considered in future analyses of interindividual variation in drug and toxin response and are bound to expand the horizons of classic pharmacogenetics and toxicology.
It is critical to understand the mechanisms driving variations in epigenetic programming in order to identify the exposures, which modulate interindividual variations in response to drug and toxin actions. Epigenetic considerations should also be applied in drug development to identify potential toxic hazards to the epigenome as well as to discover agents, which modulate the epigenome in a therapeutically advantageous manner. Drugs, which target the epigenetic machinery, are currently tested in clinical trials in cancer (Kramer et al., 2001; Weidle and Grossmann, 2000) and psychiatry disorders (Simonini et al., 2006).
THE EPIGENOME
Chromatin
The epigenome consists of the chromatin and its modifications as well as a covalent modification by methylation of cytosine rings found at the dinucleotide sequence CG (Razin, 1998). The epigenome determines the accessibility of the transcription machinery to the genome. Inaccessible genes are silent whereas accessible genes are transcribed. We therefore distinguish between open and closed configuration of chromatin (Groudine et al., 1983; Grunstein, 1997; Marks et al., 1985; Ramain et al., 1986; Varga-Weisz and Becker, 2006). Large chromatin domains, which are tightly silenced are termed heterochromatin, these regions are located around centromeres and telomeres. Regions of chromatin permissive for gene expression are termed euchromatin. However, some silent genes are found in “euchromatic neighborhoods” and are marked by similar chromatin modifications to those, which are responsible for heterochromatin silencing. Recently, another new level of epigenetic regulation by small noncoding RNAs termed microRNA has been discovered (Bergmann and Lane, 2003). A large number of loci in the human genome encode noncoding RNAs, which are processed to short RNAs and target-specific genes for silencing. microRNAs regulate gene expression at different levels; silencing of chromatin, degradation of mRNA, and blocking translation. microRNAs were found to play an important role in cancer (Zhang et al., 2007) and could potentially play an important role in modulating drug response as well (Vo et al., 2005).
An interesting but yet unexplored possibility is that microRNAs regulate the levels of expression of genes involved in drug metabolism and response to xenobiotics. Interindividual differences in microRNA expression might in turn define subject-specific drug and toxin responsivity. On the other hand, both drug and xenobiotics exposure might alter the repertoire of microRNA expression (Saito and Jones, 2006). Moreover, microRNA expression is itself regulated by epigenetic factors such as DNA methylation and chromatin structure (Saito and Jones, 2006) and thus could mediate the impact of epigenetic reprogramming in response to environmental exposure on a panel of other genes. Since microRNAs also act by changing chromatin structure (Chuang and Jones, 2007) they could be considered as a component of chromatin modification and DNA methylation machineries.
The Histone Code
The DNA is wrapped around a protein-based structure termed chromatin. The basic building block of chromatin is the nucleosome, which is formed of an octamer of histone proteins. There are five basic forms of histone proteins termed H1, H2A, H2B H3, and H4 (Finch et al., 1977) as well as other minor variants, which are involved in specific functions such as DNA repair or gene activation (Sarma and Reinberg, 2005). The octamer structure of the nucleosome is composed of an H3–H4 tetramer flanked on either side with an H2A–H2B dimer (Finch et al., 1977). The N-terminal tails of these histones are extensively modified by methylation (Jenuwein, 2001), phosphorylation, acetylation (Wade et al., 1997), and ubiquitination (Shilatifard, 2006). The state of modification of these tails plays an important role in defining the accessibility of the DNA wrapped around the nucleosome core. Different histone variants, which replace the standard isoforms also play a regulatory role and serve to mark active genes in some instances (Henikoff et al., 2004). The specific pattern of histone modifications was proposed to form a “histone code” that delineates the parts of the genome to be expressed at a given point in time in a given cell type (Jenuwein and Allis, 2001). A change in histone modifications around a gene will change its level of expression and could convert an active gene to become silent resulting in “loss of function,” or switch a silent gene to be active leading to “gain of function.”
Histone-Modifying Enzymes
The most investigated histone-modifying enzymes are histone acetyltransferases (HATs), which acetylate histone H3 at the K9 residue as well as other residues and H4 tails at a number of residues, and histone deacetylases (HDACs), which deacetylate histone tails (Kuo and Allis, 1998). Histone acetylation is believed to be a predominant signal for an active chromatin configuration (Lee et al., 1993; Perry and Chalkley, 1982). Deacetylated histones signal inactive chromatin, chromatin associated with inactive genes. Many repressors and repressor complexes recruit HDACs to genes, thus causing their inactivation (Wolffe, 1996).
Histone tail acetylation is believed to enhance the accessibility of a gene to the transcription machinery, whereas deacetylated tails are highly charged and are believed to be tightly associated with the DNA backbone and thus limiting accessibility of genes to transcription factors (Kuo and Allis, 1998). Whole-genome chromatin immunoprecipitation (ChIP) microarrays (ChIP on chip) analyses indicate that H3 histone acetylation at the K9 residue is a hallmark of active promoters and the most consistent mark of active genes (Sinha et al., 2006). HDACs are emerging drug targets in cancer (Marks et al., 2000) and schizophrenia therapy (Simonini et al., 2006). However, it is also necessary to consider the possibility that different unsuspected environmental toxic and pharmacological agents might target this important class of chromatin modifiers and thus affect the long-term programming of the genome in diverse tissues. A classic example is the antiepileptic drug Valproic acid, which was considered for years to be a GABA receptor stimulator but was later found to be an HDAC inhibitor (Gottlicher et al., 2001), which can impact both chromatin and DNA methylation (Detich et al., 2003a; Milutinovic et al., 2007; Veldic et al., 2005). It stands to reason that future assessment of safety of different agents would test for HDAC inhibitory activity. On the other hand, valproate could serve as a paradigm of the potential for rediscovering a “new” role as chromatin modifiers for “old drugs” with novel clinical applications. Valproate is now tested for its HDACi activity in cancer therapy (Kuendgen et al., 2006; MacFarlane et al., 2005).
Histone modification by methylation is catalyzed by different histone methyltransferases (HMTs). All HMTs bear a characteristic SET domain (Bryk et al., 2002). Several classes of HMT were characterized. Some specific methylation events are associated with gene silencing and others with gene activation. For example, methylation of the K9 residue of H3-histone tails is catalyzed by the HMT SUV3–9 and is associated with silencing of the associated gene (Lachner et al., 2001). HMTs are targeted to specific genomic loci by factors, which recognize specific cis-acting elements in their genomic targets. Most notable in targeting repressive chromatin-modifying enzymes to specific genomic loci are the Polycomb group proteins (PcG), which were originally identified in Drosophila and repress gene expression at a long distance (Schwartz and Pirrotta, 2007). For example, a member of the PcG group of proteins is EZH2, which was shown to contain a SET domain and function as HMT targeting K27 and in the H3 histone tail (Cao and Zhang, 2004; Cao et al., 2002).
Particular factors recognize histone modifications and further stabilize an inactive state. For example, the heterochromatin associated protein HP-1, binds H3-histone tails methylated at the K9 residue and precipitates an inactive chromatin structure (Lachner et al., 2001). Recently described histone demethylases remove the methylation mark by oxidative demethylation causing either activation or repression of gene expression (Shi et al., 2004; Tsukada et al., 2006). Several classes of histone demethylases are currently known.
Chromatin Remodeling
Chromatin remodeling complexes, which are adenosine triphosphate (ATP) dependent, alter the position of nucleosomes around the transcription initiation site and define its accessibility to the transcription machinery (Varga-Weisz and Becker, 2006). It is becoming clear now that there is an interrelationship between chromatin modification and chromatin remodeling. For example, the presence of BRG1, the catalytic subunit of SWI/SNF-related chromatin remodeling complexes is required for histone acetylation and regulation of β-globin expression during development (Bultman et al., 2005). Mutations in SWI/SNF proteins play an important role in human disease. For example, mutations in ATRX a member of the SNF2 family of helicase/ATPases give rise to a syndrome characterized by severe mental retardation, facial dysmorphism, urogenital abnormalities, and alpha-thalassaemia (Gibbons et al., 1997; Picketts et al., 1996). Agents that block these proteins might have long-term downstream effects on chromatin programming, which might extend well beyond their immediate-early effects.
How is the Gene Specificity of Chromatin Structure Established; the Concept of Targeting
The main role of the epigenome is to orchestrate a specific program of gene expression, which is unique for each cell type. In addition, the epigenome has to respond to specific physiological, developmental, and environmental triggers. A basic principle in epigenetic regulation is targeting. Histone-modifying enzymes are generally not gene specific. Specific transcription factors and transcription repressors recruit histone-modifying enzymes to specific genes and thus define the gene-specific profile of histone modification (Jenuwein and Allis, 2001). For example, transcription factors such as octamer binding 4 and NANOG seem to be important for recruitment of PcG complexes to critical genes and to maintain them in a repressed state in stem cells (Lee et al., 2006). The targets of these complexes in embryonic stem cell were found to be some of the most conserved noncoding elements in the genome (Lee et al., 2006).
In addition to developmental regulators such as PcG and TRITHORAX complexes, the epigenome could potentially respond to physiological and environmental signals throughout life after developmental programs are established and executed (Weaver et al., 2004a, 2005). These are most probably mediated by signaling pathways in the cell. Specific transacting factors are responsive to cellular signaling pathways. Signal transduction pathways, which are activated by cell-surface receptors, could thus serve as conduits for epigenetic change linking the environmental trigger at cell-surface receptors with gene-specific chromatin alterations and reprogramming of gene activity. Signaling pathways are highly responsive to drugs and toxic agents. Thus, the possible impact of repetitive firing of signaling pathway by an environmental agent could have an epigenomic impact well beyond the recognized immediate transient effects. For example, numerous signaling pathways including those triggered by G protein–coupled cell-surface receptors in the brain alter the concentration of cyclic adenosine 3′,5′ monophosphate (cAMP) in the cell resulting in activation of protein kinase A (PKA). One of the transcription factors, which is phosphorylated and activated by PKA in response to elevated cAMP, is CREB (cAMP response element binding protein). CREB binds cAMP response elements in certain genes. Phosphorylated CREB also recruits CREB-binding protein (CBP) (Chrivia et al., 1993). CBP is a HAT, which acetylates histones (Ogryzko et al., 1996). Thus, elevation of cAMP levels in response to an extracellular signal would result in a change in the state of histone acetylation of specific genes. Obviously environmental or physiological events, which interfere at any point along the signaling pathway, might result in chromatin alterations. Thus, toxic agents which interfere with cAMP-mediated signaling pathways could potentially elicit in addition to the immediate physiological consequences of elevated cAMP, stable reprogramming of gene expression which will last long after the initial exposure is over. An example of such a pathway that leads from maternal behavior to long-term programming of gene expression in the hippocampus will be discussed in detail here (Meaney and Szyf, 2005b).
DNA Methylation
In contrast to the situation in a number of lower model organisms, DNA methylation is an important component of epigenetic regulation in plants and vertebrates. This fundamental difference between epigenetics in vertebrates where DNA methylation plays a nodal role and the popular model organisms with no DNA methylation raises methodological issues as to the validity of these organisms as models for evaluating the potential epigenetic effects of environmental exposures. DNA methylation provides a unique mechanism of epigenetic marking since in contrast with chromatin, which is associated with DNA, the DNA molecule itself is chemically modified by methyl residues at the 5′ position of the cytosine rings predominantly in the dinucleotide sequence CG in vertebrates (Razin, 1998).
What distinguishes DNA methylation in vertebrate genomes is the fact that not all CGs are methylated in any given cell type (Razin, 1998). Distinct CGs are methylated in different cell types (between 60 and 80% of all CGs in a given cell type), generating cell type–specific patterns of methylation (Razin and Szyf, 1984). Thus, the DNA methylation pattern confers upon the genome its cell type–specific chemical identity (Razin, 1998). Since DNA methylation is part of the chemical structure of the DNA itself, it is more stable than other epigenetic marks and thus it has extremely important diagnostic potential (Beck et al., 1999). Epigenetic profiling is rapidly advancing as a method for early diagnosis of cancer as well as for staging cancers. It is yet to be utilized in behavioral disorders as well as other diseases. It is still to be recognized in the toxicological sciences as a valid biomarker of potential hazardous effects of environmental agents.
The DNA methylation pattern is established during development and is then maintained faithfully through life by the maintenance DNA methyltransferase (DNMT) (Razin and Riggs, 1980). The DNA methylation reaction was believed to be irreversible, thus the common consensus was that the only manner by which methyl residues could be lost was through replication in the absence of DNMT by passive demethylation (Razin and Riggs, 1980). Recent data support the idea that similar to chromatin modification, DNA methylation is also potentially reversible (Ramchandani et al., 1999a) even in postmitotic tissues (Weaver et al., 2004a). Recent results suggest that the DNA methylation pattern is highly dynamic in neurons and plays a critical role in memory and fear conditioning (Levenson et al., 2006; Miller and Sweatt, 2007).
DNA Methylation, Chromatin, and Gene Expression
DNA methylation patterns in vertebrates are distinguished by their correlation with chromatin structure. Active open-configuration regions of the chromatin, which enable gene expression, are associated with hypomethylated DNA, whereas hypermethylated DNA is packaged in inactive chromatin (Razin, 1998; Razin and Cedar, 1977). Thus, a general inverse correlation between DNA methylation in certain regulatory regions and gene expression was proposed by Razin and Riggs (1980) more than two decades ago.
It is clear from an evolutionary perspective that epigenetic regulation could occur in the absence of any further modification of DNA. There is no DNA methylation in some of the most well-studied genetic models of development such as the yeast Saccharomyces cerevisiae, the nematode Caenorhabditis elegans and there is a very minute level of DNA methylation in Drosophila melanogaster (Lyko et al., 2000; Ramsahoye et al., 2000; Urieli-Shoval et al., 1982). It therefore stands to reason that the basic level of epigenetic regulation, at least from an evolutionary perspective, is the chromatin. This evolutionary relationship between chromatin and DNA methylation might also be at work mechanistically in higher organisms where DNA methylation does play a role. It is becoming clear that chromatin states define and delineate DNA methylation patterns in both directions of the DNA methylation equilibrium; methylation and demethylation (D'Alessio and Szyf, 2006). This link between chromatin and DNA methylation provides a conduit through which site-specific DNA methylation patterns are laid down and for the DNA methylation pattern to respond to changing environmental and cellular cues.
A recent high-resolution analysis of DNA methylation patterns and histone H3 and H4 acetylation patterns in the HOXA cluster region revealed no acetylated histones in the hypermethylated regions, demonstrating reciprocal relationship between DNA methylation and histone H3 and H4 acetylation (Hayashi et al., 2007) which was predicted by the original studies of Razin and Cedar (1977) three decades ago. Recent genome-wide analyses of DNA methylation suggest a more complex relationship between DNA methylation and gene expression as was proposed two decades ago (Razin and Szyf, 1984). At a genomic level, promoters are poorly methylated whereas repetitive sequences and heterochromatic regions are methylated supporting the general role of DNA methylation in gene silencing (Zhang et al., 2006).
Although a significant fraction of promoters in the human genome (30%) exhibit a tight inverse correlation between gene expression and DNA methylation, certain promoters are not expressed even in the unmethylated state and several genes are heavily methylated even in their expressed state (Eckhardt et al., 2006). Thus, although methylation is a long-term repressive signal, gene expression involves a combination of factors. Only once all these factors including chromatin marks and RNA polymerase II occupancy are concurrently analyzed at the genomic level, we will be able to establish the first principles of the involvement of DNA methylation in gene expression.
A plausible model is that there are multiple states of silencing of transcription, which are marked by different epigenetic marks and that DNA methylation signals only one of the most stable state of repression. Promoters, poised to expression in response to transient signals, would be unmethylated while permanently silenced promoters, will be found in a methylated state.
An additional point that must be considered when analyzing whole-genome maps is that differential methylation is in certain instances exquisitely site specific. Current genome-wide analyses methods do not measure DNA methylation at a single base resolution.
An emerging unresolved question is whether methylation in regions other than promoters is involved in transcriptional control. Recent data from Arabidopsis suggest a positive correlation between methylation in the body of genes and transcription (Zhang et al., 2006). Methylation of the body of expressed genes was noted decades ago but its role in gene expression remains to be resolved (Stein et al., 1983). Similarly, the 3′ promoter proximal region was recently implicated in gene silencing (Appanah et al., 2007).
Mechanisms of Silencing of Gene Expression by DNA Methylation
Notwithstanding the problems raised by the whole-genome level of DNA methylation analysis discussed in the previous sections, there are now overwhelming data indicting that aberrant silencing of tumor suppressor genes by DNA methylation is a common mechanism in cancer (Baylin et al., 2001). Thus, there is no question that DNA methylation is a highly effective mechanism of silencing of gene expression in vertebrates and plants. DNA methylation silences gene expression by two principal mechanisms. The first mechanism involves direct interference of a methyl residue with the binding of a transcription factor to its recognition element in DNA resulting in silencing of gene expression (Comb and Goodman, 1990; Inamdar et al., 1991). A second mechanism is indirect. A certain density of DNA methylation moieties in the region of the gene attracts the binding of methylated DNA binding proteins (MBDs) such as MeCP2 (Nan et al., 1997). MeCP2 recruits other proteins such as SIN3A and histone-modifying enzymes, which lead to formation of a “closed” chromatin configuration and silencing of gene expression (Nan et al., 1997). Several MBDs, such as MBD1, MBD2, and MBD3, suppress gene expression by a similar mechanism (Fujita et al., 1999; Hendrich and Bird, 1998; Ng et al., 1999b).
DNA Methylation and Demethylation Enzymes
The DNA methylation reaction is catalyzed by DNMT (Fig. 1) (Razin and Cedar, 1977). Methylation of DNA occurs immediately after replication by a transfer of a methyl moiety from the donor S-adenosyl-L-methionine (AdoMet) in a reaction catalyzed by DNMTs. Three distinct phylogenic DNMTs were identified in mammals. DNMT1 shows preference for hemimethylated DNA in vitro, which is consistent with its role as a maintenance DNMT, whereas DNMT3a and DNMT3b methylate unmethylated and methylated DNA at an equal rate which is consistent with a de novo DNMT role (Okano et al., 1998). Two additional DNMT homologs were found; DNMT2 whose substrate and DNA methylation activity is unclear (Vilain et al., 1998) but was shown to methylate tRNA (Goll et al., 2006; Rai et al., 2007) and DNMT3L which is essential for the establishment of maternal genomic imprints but lacks key methyltransferase motifs, and is possibly a regulator of methylation rather than an enzyme that methylates DNA (Bourc'his et al., 2001). Knockout mouse data indicate that DNMT1 is responsible for a majority of DNA methylation marks in the mouse genome (Li et al., 1992) as well as the human genome (Chen et al., 2007), whereas DNMT3a and DNMT3b are responsible for some but not all de novo methylation during development (Okano et al., 1999).
The DNA methylation equilibrium. The DNA methylation pattern is determined by a balance of methylation and demethylation reactions. The direction of the equilibrium is determined by chromatin-modifying enzymes and by the state of chromatin. The DNA methylation reaction is catalyzed by DNMTs, which transfers a methyl moiety from the methyl donor SAM to the 5′ position in the cytosine ring in the dinucleotide sequence CG in DNA releasing SAH. Demethylase(s) remove the methyl group by a yet undetermined mechanism. Genes associated with acetylated histones HAT and are unmethylated at K9 and K27 positions of H3-histone tails are preferred substrates for DNA demethylases. Genes associated with deacetylated histones, which are methylated at K9 and K27 positions of H3 histones tails, and associated with the HP-1 and the HMT EZH2, are preferred substrates for DNMTs. HMTASE, histone methyltransferase; SAH, S-adenosylhomocysteine; EZH2, Polycomb complex member HMT; dMTase, DNA demethylase; HDMase, histone demethylase; K9, K9 residue on H3 histone tail; Ac, acetyl group; CH3, methyl group; HP-1, methyl K9 H3 histone binding protein.
Razin and Riggs proposed that the DNA methylation pattern is accurately inherited during replication since maintenance DNMT could only methylate hemimethylated sites. Such as sites are generated on the nascent DNA strand during DNA replication when a methylated CG dinucleotide in the template strand is replicated. DNA methylation was therefore proposed to be truly heritable by a semiconservative mechanism similar to DNA replication (Razin and Riggs, 1980). This concept has led to the basic notion that although DNA methylation patterns are sculpted during development by demethylases and de novo methyltransferases, they are fixed thereafter and are inherited faithfully similar to the genetic sequence.
The DNA Methylation Pattern is Reversible; DNA Demethylation Enzymes
It was a long held belief that the DNA methylation pattern is solely dependent on DNMTs and that the reverse reaction cannot occur. Thus, it was believed that DNA methylation pattern could be altered only during cell division when new unmethylated DNA is synthesized and serves as a substrate for maintenance DNMT. The implication of this classic hypothesis to toxicology is that toxic agents would affect DNA methylation only in dividing somatic cells but not in nondividing tissues such as muscle, heart, brain, etc. The second implication is that the only enzyme, which might be a target for toxic agents, is DNMT. The third implication is that the only possible alteration in the DNA methylation pattern is hypomethylation by inhibition of DNMT during cell division. The semiconservative model of replication is inconsistent with significant de novo methylation that does happen in somatic cells (Szyf et al., 1989).
If DNA methylation only happens when DNMT is copying DNA methylation patterns during cell division as suggested by the classic model (Razin and Riggs, 1980), there should be no requirement for DNMTs in postmitotic neurons. Nevertheless, DNMTs are present in postmitotic neurons (Goto et al., 1994) and there are data suggesting that DNMT levels in neurons change in certain pathological conditions such as schizophrenia (Veldic et al., 2005). The presence of DNMT in neurons would make sense only if DNA methylation is dynamic in postmitotic tissues and is a balance of methylation and demethylation reactions (Szyf, 2001). Without active replication-independent demethylation there is no need for DNA methylation activity in neurons. Recent data suggest rapid demethylation and methylation of specific genes in the brain in response to contextual fear conditioning (Miller and Sweatt, 2007). These data raise the interesting possibility that the DNA methylation is highly dynamic in neurons and that it plays a role in memory generation.
We have proposed a while ago that the DNA methylation pattern is a balance of methylation and demethylation reactions (Ramchandani et al., 1999b). This equilibrium is responsive to physiological and environmental signals and thus forms a platform for gene–environment interactions as well as a target for toxic agents (Fig. 1). There is a long list of data from both cell culture and early mouse development supporting the hypothesis that active methylation occurs in both embryonal and somatic cells. A replication-independent active demethylase was assayed in nuclear extracts from Zebrafish embryo (Collas, 1998). A global active demethylation, which is independent of replication occurs in the male pronucleus of mice shortly after fertilization (Oswald et al., 2000). Active demethylation was reported for the myosin gene in differentiating myoblast cells (Lucarelli et al., 2001), for the Il2 gene upon T-cell activation (Bruniquel and Schwartz, 2003), the interferon γ gene upon antigen exposure of memory CD8 T cells (Kersh et al., 2006) and in the glucocorticoid receptor (GR) gene promoter in adult rat brains upon treatment with the HDAC inhibitor trichostatin A (TSA) (Weaver et al., 2004a). Thus, environmental agents could potentially affect the DNA methylation state in both directions. If the hypothesis proposed here is true, then all the components of the DNA methylation equilibrium are potential targets for toxic agents. This hypothesis also predicts that certain xenobiotics could affect the DNA methylation pattern in nondividing tissues such as the brain as well as in dividing tissues. Thus, it might be important to assess the potential epigenotoxic effect of xenobiotics in nondividing tissues as well.
The main challenge in the field is to identify the enzymes responsible for demethylation. This is especially important for evaluating the impact of candidate environmental toxins on DNA demethylation. The biochemical properties of the enzymes responsible for active demethylation are still controversial. One proposal has been that a G/T mismatch repair glycosylase also functions as a 5-methylcytosine DNA glycosylase, recognizes methyl cytosines, and cleaves the bond between the sugar and the base. The abasic site is then repaired and replaced with a nonmethylated cytosine resulting in demethylation (Jost, 1993). An additional protein with a similar activity was identified, the MBD4 (Zhu et al., 2000). While such mechanism can explain site-specific demethylation, global demethylation by a glycosylase would cause widespread DNA damage and would compromise genomic integrity. Another report from our laboratory has proposed that MBD2 has demethylase activity. MBD2b (a shorter isoform of MBD2) was shown to directly remove the methyl group from methylated cytosine in methylated CpGs (Bhattacharya et al., 1999). This enzyme was therefore proposed to reverse the DNA methylation reaction. However, other groups disputed this finding and reproducing the in vitro activity of this protein has been tough (Ng et al., 1999b). Very recent data suggest that active demethylation early in embryogenesis as well as in somatic cells is catalyzed by a nucleotide excision repair mechanism, whereby methylated cytosines are replaced by unmethylated cytosines, which involves the growth arrest and damage response protein Gadd45a and the DNA repair endonuclease XPG (Barreto et al., 2007). The main problem with a repair-based mechanism is that it would cause extensive damage to the DNA and it is difficult to believe that such a mechanism participates in the global demethylation in embryogenesis.
Although a number of biochemical processes were implicated in demethylation, it is unclear how and when these different enzymes participate in shaping and maintaining the overall pattern of methylation and how these activities respond to different environmental exposures in the brain. This remains one of the most important unresolved questions in the field. In vitro assays of demethylase activity are still finicky and are not amenable as of yet for high-throughput screening. We therefore developed a cellular assay for demethylase activity, which measures the demethylation of transiently transfected nonreplicating in vitro methylated Cyto megalo virus promoter-green fluorescence protein in HEK 293 cells (Cervoni and Szyf, 2001; Detich et al., 2003b). Demethylation results in expression and green fluorescence, which could be detected by a high throughput spectrofluorometer. This assay could be used to identify potential agents that modulate the DNA methylation machinery.
How are Sequence-Specific Methylation Patterns Generated and Maintained? Targeting DNA Methylation and Demethylation; Chromatin and DNA Methylation
The DNA methylation pattern is exquisitely cell type and site specific. How is this pattern generated and maintained and how does it respond in a distinct mode to specific signals later in life?
As discussed above, chromatin states are maintained by targeting of chromatin-modifying enzymes to specific loci in the genome. We propose that the chromatin configuration then gates the accessibility of genes to either DNA methylation or demethylation machineries (Cervoni and Szyf, 2001; D'Alessio and Szyf, 2006) (Fig. 2). This primacy of chromatin in defining DNA methylation patterns is supported by the evolutionary relationship between chromatin modification and DNA methylation as discussed above (section IIg). In support of this hypothesis we have previously shown that the HDACi trichostatin A, which causes histone hyperacetylation also causes active replication-independent DNA demethylation (Cervoni and Szyf, 2001). Since it is well established that cellular signaling pathways regulate chromatin modification states, this hypothesis provides a link between central cellular signaling pathways and the DNA methylation pattern.
Targeting of DNMTs to specific loci. Under normal conditions DNMT activity is limiting. Both active genes marked by histone acetylation at K9 residues of H3-histones and inactive genes marked by K9 and K27 methylation of H3 histone tails (M) and are associated with EZH2 are present in the genome. Elevation of DNMT levels as a consequence of induction of oncogenic signaling pathways results in targeting of DNMT to genes associated with EZH2 causing methylation of DNA (circled M). Other genes are not affected. This results in gene- and region-specific methylation. EZH2, Polycomb complex member HMT; K9, K9 residue on H3 histone tail; M, methyl residue; oval, nucleosome; horizontal arrow, transcription.
Histone modification enzymes interact with DNA methylating enzymes and participate in recruiting them to specific targets (Fig. 2). A growing list of histone-modifying enzymes has been shown to interact with DNMT1, and DNMT3a such as HDAC1 and HDAC2, the HMTs SUV3–9, EZH2, a member of the multiprotein Polycomb complex PRC2/3, which methylates H3 histone at the K27 residue (Fuks et al., 2000, 2003; Rountree et al., 2000; Vire et al., 2006) as well as the heterochromatin protein HP-1 which binds H3–K9 methylated histones (Smallwood et al., 2007). The MBD MeCP2 interacts with the HMT SUV3–9 (Fuks et al., 2003).
The link between EZH2, methylation of H3-Histones at K9 and K27 residues, and DNA methylation of tumor suppressor genes has been unraveled recently (Fig. 2). A survey of CG islands methylated in lung cancer revealed that they were also PcG EZH2 targets (Rauch et al., 2007). Although these CG island targets are not methylated in normal tissues, they are marked by histone methylation and are associated with EZH2 (Schlesinger et al., 2007). During tumorigenesis, DNMT1 levels are induced by activation of several oncogenic pathways (Bigey et al., 2000; MacLeod et al., 1995; Slack et al., 1999; Szyf, 2006; Szyf et al., 2000). Since DNMT1 interacts with EZH2, it is guided by EZH2 to specific loci. The increased capacity of DNMT1 in oncogenesis results in increased occupation of EZH2 targets in the genome and methylation of CG islands associated with EZH2 (Vire et al., 2006). This mechanism illustrates why an increase in DNMT1 results in specific alteration of DNA methylation patterns in distinct CG islands but not in a global change in DNA methylation (Fig. 2).
Trans-acting repressors target both histone-modifying enzymes and DNMTs to specific cis-acting signals (see model in Fig. 3). A good example is the promyelocytic leukemia Promyelocytic leukemia-retinoic acid receptor gene fusion protein that engages HDACs and DNMTs to its target binding sequences and produces de novo DNA methylation of adjacent genes (Di Croce et al., 2002).
The DNA and chromatin state equilibrium. The DNA methylation equilibrium could be altered by different environmental cues, which trigger signaling pathways in the cell, which direct either chromatin activating or silencing factors to genes facilitating either methylation or demethylation. Environmental exposures including toxins could affect DNA methylation and chromatin modification enzymes at different points resulting in tilting of the DNA methylation equilibrium. HMTASE, histone methyltransferase; TF, transcription factor; AC, acetyl; M, methyl group; circled M, methyl group in DNA; HP-1, methyl K9 H3-histone binding protein heterochromatin protein 1; K9, K9 residue on H3 histone tail; M, methyl residue; oval, nucleosome; horizontal arrow, transcription.
Transcription factors recruit HATs to specific genes. This triggers gene-specific acetylation and we propose that this could facilitate also the demethylation of the gene (Fig. 3). There are examples in the literature indicating that enhancers are required for replication-independent active demethylation. For example, the intronic kappa chain enhancer and the transcription factor nuclear factor kappaB are required for B cell-specific demethylation of the kappa immunoglobulin gene (Lichtenstein et al., 1994). Maternal care is employing this mechanism to program gene expression through recruitment of the transcription factor nerve growth factor induced A (NGFI-A) to one of the GR gene promoters in the hippocampus (Weaver et al., 2007). This is a mechanism, which could potentially mediate between external signals from the environment and demethylation of specific genes in neurons.
In summary, we propose that the DNA methylation pattern and chromatin structure are found in a dynamic balance through life. The direction of the equilibrium is maintained and defined by sequence-specific factors, which deliver histone modification and DNA modification enzymes to genes. These factors are responsive to signaling pathways in the cell (Fig. 3). The state of this equilibrium is defined during development and in the process of cellular differentiation. Physiological or environmental signals, which alter the signaling pathways in the cell, would result in tilting of this balance (Fig. 3). A good example is the HAT CBP, which is activated by cAMP mediated signaling pathways and causes gene-specific histone acetylation (Ogryzko et al., 1996). The implication of this model for toxicology is that xenobiotics, which interfere with cellular signaling pathways, might also alter critical DNA methylation patterns. It is important therefore to develop assays that measure the potential DNA methylation effects of agents, which are known to affect cellular signaling pathways.
MECHANISMS OF EPIGENETIC PROGRAMMING BY THE ENVIRONMENT AND THEIR POSSIBLE IMPLICATIONS FOR TOXICOLOGY, HUMAN HEALTH, AND HUMAN BEHAVIOR
The dynamic DNA methylation pattern and its responsiveness to trans-acting factor and cellular signaling pathways provide a path for the environment to affect and modulate DNA methylation patterns. Physiological, behavioral, pharmacological, and toxic agents could act at different levels to either activate or block signaling pathways, thus leading to alterations in chromatin structure and as a consequence in DNA methylation. Xenobiotics might also directly act on epigenetic enzymatic machineries such as histone methylases and demethylases and DNA methylases and demethylases. Modulation of epigenetic machineries by toxins might have long-term effects on human health by affecting diseases such as cancer as well as behavior.
Hypo and Hypermethylation in Cancer
A well-established example of the involvement of aberrant DNA methylation in human disease is cancer. DNA methylation aberrations are a hallmark of cancer. Two seemingly contradictory aberrations are seen in cancer, global hypomethylation (Ehrlich, 2002; Feinberg and Vogelstein, 1983a,b; Feinberg et al., 1988; Narayan et al., 1998) and DNA hypermethylation of tumor suppressor genes (Baylin et al., 1998). Regional hypermethylation causes silencing of tumor suppressor genes and hypomethylation was recently shown to be involved in activating prometastatic genes and possibly other tumor promoting genes (Pakneshan et al., 2004, 2005; Shukeir et al., 2006). Hypomethylation is a well-known aberration in cancer and could be triggered by nongenotoxic carcinogens (Bachman et al., 2006a,b; Counts and Goodman, 1995a,b). It is impossible to explain the coexistence of both regional hypermethylation and global demethylation in the same cell if the only enzyme responsible for DNA methylation is the semiconservative maintenance DNMT1. However, our current understanding is that two different enzymatic machineries define the methylation pattern; DNA methylation and demethylation and that targeting is required for either methylation or demethylation (Fig. 3) (Szyf, 2001). This hypothesis could accommodate both the hypermethylation and demethylation observed in cancer. Cancer involves deregulation of the enzymatic DNA methylation and possibly demethylation machineries. However, increase in DNMT1 by itself does not cause hypermethylation, the DNMT needs to be targeted to specific loci. One exciting candidate for targeting DNMT1 is the PcG complex (Ohm et al., 2007; Schlesinger et al., 2007; Schwartz and Pirrotta, 2007; Vire et al., 2006; Widschwendter et al., 2007) discussed above. Thus, it appears that the level of DNMT1 is the limiting factor in methylation of PcG associated genomic loci (Fig. 2). Although PcG binding and K9/27 H3 histone methylation poises these sites for methylation, under normal conditions the capacity of DNMT1 in the cell is insufficient to bind all the potential PcG complexes in the nucleus. By upregulation of the levels of DNMT1 and other DNMTs, oncogenic pathways increase the fraction of PcG associated loci which become associated with DNMT1 and undergo methylation (Fig. 2). Thus, agents which modulate DNMT activity such as cigarette smoke could cause regional methylation of specific genes (Liu et al., 2007).
We have identified a number of prometastatic genes, such as uPA and MMP2 which are demethylated in metastatic breast and prostate cancer (Pakneshan et al., 2004, 2005; Shukeir et al., 2006). We showed that these genes are demethylated by a protein, which we had previously proposed to serve as a demethylase MBD2 (Bhattacharya et al., 1999). Knock-down of MBD2 reversed demethylation and the ability of cancer cells to invade bone and metastasize. Therefore, we propose that induction of demethylase results in demethylation of prometastatic genes concurrently with the regional methylation of tumor suppressor genes. The profile of demethylated genes is determined as well by factors, which target the demethylase to these genes. These factors remain to be discovered.
Implications for Toxicology; Environmental Modulation of Epigenetic Machineries and Cancer
Both the methylation and demethylation machineries could be potentially modulated by environmental interventions such as diet, alcohol, and environmental hazards such as cigarette smoke and arsenic. We found that the methyl donor S-adenosylmethionine (SAM) inhibits demethylase activity in human cells (Detich et al., 2003b) and in vitro and demethylation of metastatic genes in breast and prostate cancer cells (Pakneshan et al., 2004; Shukeir et al., 2006). SAM levels could be modulated by methyl content of diet, by dietary intake of folic acid and vitamin B12 (Doi et al., 1989; Fenech, 2001; Selhub, 2002), as well as by alcohol (Rambaldi and Gluud, 2006). Thus, agents that modulate one carbon metabolism or directly affect levels of methionine or SAM might have an effect on epigenetic programming of critical cancer and metastasis promoting genes. The effect of potential hazardous agents on one carbon metabolism should be assessed to exclude epigenetic side effects.
Cigarette smoke was recently shown to stimulate the demethylation of metastatic genes in lung cancer cells through inhibition of expression of the DNMTB (Liu et al., 2007). This is a clear example how an environmental hazard could impact the genome through nongenotoxic mechanisms. It is possible that this effect of cigarette smoke plays a critical role in progression of cancer. It is therefore important to screen and identify xenoobiotics, which affect the state of expression of putative DNMTs and demethylases using cellular assays such as the one used by Liu et al. (2007) or Detich et al. (2003b). Another example is the nongenotoxic carcinogen phenobarbital which induces global hypomethylation and regional hypermethylation in rodents (Bachman et al., 2006b).
Prenatal Epigenetic Effects on the Offspring of Maternal Environmental Exposures
Environmental exposures have particularly long-term effects on epigenetic programming of the offspring genome when they occur prenatally. A nice model for assessing the prenatal epigenetic effects of environmental exposures is the Agouti mouse. The yellow coat color of Agouti mice is determined by the level of expression of the agouti gene, which is controlled by a transposable element. Exposures of the dams to high methyl content diet rich in folic acid, vitamin B(12), choline, and betaine alter the phenotype of their A(vy)/a offspring via increased CpG methylation at the A(vy) locus (Waterland and Jirtle, 2003). This experiment illustrates the remarkable potential impact of maternal exposure to different environmental agents which target the epigenetic machinery. The Agouti mouse model could be used as a biosensor to detect environmental agents and toxins, which alter the epigenetic machinery during pregnancy (Dolinoy et al., 2007). For example, exposure of pregnant Agouti mice to genistein, the major phytoestrogen in soy beans results in methylation of the retrotransposon upstream of the transcription start site of the Agouti gene and as a consequence a change in the Agouti coat color in the offspring (Dolinoy et al., 2006). The fact that exposure of a mother to some common environmental agents can result in the persistent chemical modification of the genome of the offspring points out to the critical urgency for screening and identifying environmental epigenetic modifiers. Such agents would have escaped detection using classic assays for genotoxic agents and environmental hazards.
Behavior as an Agent and Target of Epigenetic Modification
It is anticipated and well accepted that chemical exposure could have adverse effects on biological processes. Epigenetics adds the notion that a short exposure to a chemical could be memorized through epigenetic mechanisms long after the chemical trigger is gone. The main focus in toxicology has been on the biological response to exposure to chemicals. However, surprisingly our current understanding of epigenetics points to the possibility that exposure to the social environment might affect the epigenome as well. Thus, adverse social exposures might be acting through similar mechanisms to those, which respond to adverse chemical exposures. Moreover, epigenetic exposures might impact behavior and memory in addition to their effects on physical health. Thus, the long-term impact on behavior of chemicals acting through epigenetic mechanisms might be included in the future in toxicological assessment of the hazardous potential of environmental and chemical exposures.
There are extensive studies, which correlate the socioeconomic status early in life with health outcomes later in life. However, the mechanisms mediating the effect of the social environment on health are unclear. The best-documented case to date of epigenetic programming triggered by the social environment is the long-term impact that maternal care in rat has on expression of the GR gene in the hippocampus of the offspring. In the rat, the adult offspring of mothers that exhibit increased levels of pup licking/grooming (i.e., High LG mothers) over the first week of life show increased hippocampal GR expression, enhanced glucocorticoid feedback sensitivity, decreased hypothalamic corticotrophin releasing factor (CRF) expression, and more modest HPA stress responses compared to animals reared by Low LG mothers (Francis et al., 1999; Liu et al., 1997). Cross-fostering studies suggest direct effects of maternal care on both gene expression and stress responses (Francis et al., 1999; Liu et al., 1997). These studies supported an epigenetic mechanism rather than a genetic mechanism for the effect of maternal care since the fostering mother and not the biological mother defined the stress response of its adult offspring (Francis et al., 1999; Liu et al., 1997). We hypothesized that the maternal behavior of the caregiver triggered an epigenetic change in the brain of the offspring (Meaney and Szyf, 2005b).
This model has two main implications on our understanding of the relationship between behavior and epigenetics. First, social behavior of one subject can affect epigenetic programming in another subject. Thus, our model provides a molecular mechanism mediating the effects of nurture on nature. These socially driven epigenetic variations might impact on health as well as on the response to toxic agents. Second, epigenetic programming by exposure to chemicals and environmental pollutants can have long-term impact on behavior. Thus, the epigenetic-mediated effects of toxic agents on behavior need to be considered and evaluated.
Maternal Care Epigenetically Programs Stress Responses in the Offspring
We have previously published evidence supporting the hypothesis that epigenetic mechanisms mediate the maternal effect on stress response. Maternal care affects the chromatin, DNA methylation, and transcription factor binding to the GR exon 17 promoter illustrating the basic principles of epigenetic regulation discussed above (Weaver et al., 2004a). This epigenetic programming persists long after the period of maternal care is over. Maternal care early in life affected the expression of hundreds of genes in the adult hippocampus (Weaver et al., 2006). Thus, these data illustrate the profound effect of the social environment early in life on gene expression programming throughout life. These results have quite tantalizing implications. They imply that differences in maternal care early in life can result in gene expression changes, which remain persistent into adulthood in numerous genes. This illustrates the potential impact of behaviorally driven epigenetic processes on our genomic inheritance. An important implication of this mechanism is that exposures to different chemicals and toxins during early life might have a profound impact on behavior later in life by interfering with epigenetic programming. An additional implication is that early life experience might modulate the response to environmental exposures later in life through epigenetic programming. For example, early life adversity might result in altered epigenetic programming of genes involved in response to toxins, therapeutics, and entertainment drugs and might modulate and predispose to alcoholism and drug addiction. Indeed it is known from studies in nonhuman primates that subjects who underwent early developmental stress developed behaviour patterns similar to those predisposing to early-onset alcoholism among humans (Heinz et al., 1998). It stands to reason that this response involves epigenetic mechanisms similar to those described for the rat maternal care model. Future experiments need to address this possibility and to identify the epigenetics marks involved. If this is true, it has interesting implications on predicting susceptibility to adverse effects of environmental hazards, customizing therapy, and treatment of drug and alcohol addiction.
Epigenetic Programming by Maternal Care is Reversible in the Adult Animal
Although epigenetic programming by maternal care is limited to a critical window of time after birth, is highly stable and results in long-term changes in gene expression, it is nevertheless potentially reversible. The steady-state methylation pattern is defined by a dynamic equilibrium of methylation–demethylation reactions (Szyf, 2001). We have previously proposed that increasing histone acetylation by a HDAC inhibitor such as TSA would tilt the balance of the DNA methylation equilibrium towards demethylation (Cervoni and Szyf, 2001; Cervoni et al., unpublished data). Treating adult offspring of low Licking and grooming arched back nursing (LG-ABN) maternal care with TSA reversed the epigenetic marks on the GR exon 17 promoter; histone acetylation increased, the gene was demethylated, and there was increased occupancy of the promoter with the transcription factor NGFI-A, resulting in increased GR exon 17 promoter expression. The epigenetic reversal was accompanied with a behavioral change so that the stress response of the TSA treated adult offspring of low LG-ABN was indistinguishable from the offspring of high LG-ABN (Weaver et al., 2004b). The combination of reversibility and stability is one of the appealing aspects of epigenetics. This study was the first illustration of reversal of early life behavioral programming by pharmacological modulation of the epigenome during adulthood.
This experiment has two major implications for toxicology. First, it illustrates how epigenetic variability generated by differences in maternal behavior early in life results in differential response to drugs and toxic agents in the adult offspring. Only low LG-ABN offspring responded to TSA. The stress response was not altered by exposure to the drug in offspring of High LG mothers. Thus, in the future prediction of drug response would require understanding the epigenetic state of the subject in addition to the genetic factors. We might have to consider early life social exposure in our predictions of drug response later in life. Future toxicology research will need to focus on mapping epigenetic variations, which define differential responsivity to environmental agents and drugs.
The second implication of this study is that it is plausible that environmental agents, which affect the epigenetic machinery, might also affect behavior. It is thus critical to identify epigenetic modifiers in our repertoire of chemical exposures and to include behavioral testing as one of the endpoints included in assessment of safety of chemicals. Entertainment drugs, therapeutic drugs with other predicted mechanisms of action, or other toxic exposures, which reach the central nervous system might potentially alter the epigenetic state of critical genes and alter behavior (Fig. 4). An interesting example is the antiepileptic and mood stabilizing agent valproic acid. Although known for other mechanisms of action, this drug was found to be a HDAC inhibitor similar to TSA. We showed that valproate triggered replication-independent DNA demethylation in tissue culture (Detich et al., 2003a; Milutinovic et al., 2007) and inhibited DNA methylation in the brain in an animal model (Tremolizzo et al., 2002). Several drugs, which had well-established mechanisms of action and therapeutic targets and were not anticipated to affect the epigenome were later shown to also affect DNA methylation. For example, the antihypertensive drug hydralazine and the antiarrhythmic procainamide inhibit DNA methylation and cause broad epigenetic reprogramming in T cells (Cornacchia et al., 1988).
(a) Environmental exposure at a single point in life could result in epigenetic reprogramming which would stably alter phenotype and responsivity to toxins. (b) Exposure to toxins could result in epigenetic reprogramming which would result in phenotypic variations later in life.
One fundamental element of the hypothesis presented here is that the DNA methylation pattern is a dynamic equilibrium of methylation and demethylation throughout life. We therefore asked the question of whether it is possible also to reverse the DNA methylation equilibrium in the hippocampus in the opposite direction by increasing DNA methylation through pharmacological manipulation. The methyl donor SAM inhibits the demethylation reaction (Detich et al., 2003b). Thus, changing SAM levels would alter the DNA methylation equilibrium by either increasing the rate of the DNA methylation reaction or by inhibiting the demethylation reaction or both. Systemic injection of methionine was previously shown to increase SAM concentrations in the brain (Tremolizzo et al., 2002). Injection of methionine to the brain led to hypermethylation and reduced expression of the GR exon 17 expression in the adult hippocampus of offspring of high LG-ABN and reversal of its stress response to a pattern which was indistinguishable from offspring of low LG-ABN (Weaver et al., 2005).
In summary, we illustrated that the epigenetic state is dynamic throughout life and responsive to environmental agents. The response to chemicals is delineated by the epigenetic state of the subject and in turn that epigenetic state could be shaped by early life experiences. Epigenetic reprogramming by environmental agents might result in behavioural change.
Mechanisms Leading from Maternal Care to Epigenetic Programming
One of the mysteries of epigenetic programming by maternal behavior is identifying the conduit leading from the social behavior of the mother to a chemical modification in chromatin and DNA. Our working hypothesis is that maternal care triggers a signaling pathway in the offspring, which activates certain transcription factors delivering the epigenetic machinery (DNA and chromatin-modifying enzymes) to specific targets in the genome. Maternal LG in early life elicits a thyroid hormone-dependent increase in serotonin (5-HT) activity at 5-HT7 receptors, and the subsequent activation of cAMP and cAMP-dependent PKA (Laplante et al., 2002; Meaney et al., 1987, 2000). This is accompanied by increased hippocampal expression of NGFI-A transcription factor. The GR exon 17 promoter region contains a binding site for NGFI-A (McCormick et al., 2000). NGFI-A was previously shown to interact with the transcriptional coactivator and histone acetyl transferase CBP. Signaling pathways that result in increased cAMP also activate CBP. Recruitment of CBP to the GR exon 17 promoter in response to maternal care could explain the increased acetylation and demethylation observed in offspring of high LG-ABN (Weaver et al., 2004a). Thus, we propose that NGFI-A delivers chromatin and modifying enzymes such as the HAT CBP to the GR exon 17 promoter in response to firing of a signaling pathway by maternal care (Weaver et al., 2007). This acetylation of histones facilitates demethylation. We show that similar to acetylation induced in response to pharmacological administration of TSA, targeted acetylation by recruitment of a transcription factor leads to demethylation of DNA (Weaver et al., 2007).
The remaining question was to identify the protein(s) involved in demethylation of DNA. We previously proposed that the MBD2 was a DNA demethylase and could bring about DNA demethylation in vitro (Bhattacharya et al., 1999). Other groups (Ng et al., 1999a) hotly contested the in vitro demethylation activity of MBD2 but more recent data from our laboratory supported a demethylation role for MBD2 (Detich et al., 2002, 2003b). Our recent data indicate that MBD2 causes global demethylation in cells and causes widespread activation of genes involved in cancer metastasis. Interestingly, inhibition of MBD2 results in blocking cancer metastasis (Pakneshan et al., 2004; Shukeir et al., 2006). MBD2, which is involved so heavily in tumorigenesis and metastasis, is also found to be recruited to the GR exon 17 promoter in response to maternal care. Using a transient transfection assay we show that ectopically expressed MBD2 transcriptionally activates in vitro methylated GR exon 17 promoter-luciferase promoter, and demethylates a CG site found in the NGFI-A recognition element. Binding of NGFI-A to its response element was required for MBD2 action (Weaver et al., unpublished data). Our data are consistent with the hypothesis that NGFI-A facilitates the accessibility of the sequence to MBD2 leading to target specific demethylation.
In summary, these experiments provide a general paradigm for how behavior exposure can result in changes in DNA methylation of specific genes. This conduit could be activated by other environmental exposures. Agents targeting a step in this pathway would result in epigenetic changes leading to behavioral changes in the brain and other changes in somatic tissues or tumors (Pakneshan et al., 2004; Shukeir et al., 2006).
SUMMARY
The notion of a dynamic epigenome implies that variations in gene function and phenotype are modulated not only by DNA sequence polymorphisms but also by reversible but nevertheless very stable epigenetic changes. The epigenetic pattern is cell type specific, which distinguishes it from genomic polymorphisms and could thus create cell type–specific phenotypic variations between individuals. An important element of the hypothesis presented here is that the epigenome is found in a dynamic equilibrium, implying the possibility of change in response to shifting environmental and physiological exposures throughout life. Epigenetics provides a mechanism through which transient exposures to an environmental hazard have persistent life-long phenotypic effects. Our data suggest that the direction of the epigenetic equilibrium could be altered by pharmacological manipulations. This model has many potential implications in toxicology, as a consequence toxicological analyses will have to take the epigenetic factor into account. First, nongenotoxic agents might have an epigenotoxic effect. Second, variations in epigenetic states would alter the response to toxic agents amongst individuals. The interindividual variation might be tissue specific. Interindividual epigenetic variations could result from genetic differences as well as from exposures to different behavioral conditions, diets, and chemicals at critical periods and throughout life (Fig. 4). A novel concept introduced by our studies is that epigenetic variations could be caused not only by exposure to chemicals or nutrients but also by different behavioral exposures. Thus, social exposures such as childhood adversity might impact on drug responses later in life and should be considered in our assessment of adverse effects of potential toxic agents. It is important to incorporate these concepts in toxicology by establishing in the future a methodology to identify agents with a potential epigenotoxic effect. This effort to incorporate epigenetic considerations in toxicology might result in a new field of epi-genotoxicology.
Epigenetic mechanisms confer long-term programming to genes and bring about a change in gene function without changing the gene sequence.
Differences in epigenetic marks between individuals could result in phenotypic variation.
In contrast to genetic polymorphisms epigenetic variants are potentially reversible.
The epigenome consists of the chromatin and its modifications as well as a covalent modification by methylation of cytosine rings found at the dinucleotide sequence CG.
Small noncoding RNAs termed microRNA are involved in epigenetic regulation through chromatin modification and perhaps also DNA methylation as well as other mechanisms.
H3-histone acetylation at the K9 residue is a hallmark of active promoters and a consistent mark of active genes.
Specific transcription factors and transcription repressors recruit histone-modifying enzymes to specific genes and thus define the gene-specific profile of histone modification.
Distinct CGs are methylated in different cell types (between 60 and 80% of all CGs in a given cell type), generating cell type–specific patterns of methylation.
DNMTs catalyze the transfer of a methyl moiety from the donor AdoMet to DNA.
Recent data suggest that the DNA methylation pattern is reversible by demethylases and is thus potentially responsive to environmental exposures.
DNA methylation in critical regulatory regions silences gene expression.
DNA methylation silences gene expression by either blocking binding of transcription factors or by recruiting MBDs.
MBDs recruit chromatin-modifying enzymes to methylated regions in genes resulting in silencing of gene expression.
There is a bilateral relationship between chromatin structure and DNA methylation.
Sequence-specific factors, which deliver histone modification and DNA modification enzymes to genes, delineate the epigenetic state.
These factors are responsive to signaling pathways in the cell, thus creating a conduit between environmental exposures and epigenetic states.
Environmental exposure could alter epigenetic reprogramming, creating a phenotypic difference.
Early life experiences such as the quality of maternal care could result in persistent life-long variations in gene expression and phenotype through epigenetic programming.
Epigenetic differences could alter the response to xenobiotics.
Nongenotoxic xenobiotics could alter phenotype through epigenetic mechanisms.
It is important to screen and identify the potential epigenotoxic effects of xenobiotics and other environmental exposures.
FUNDING
These studies were supported by a grant from the Canadian Institutes for Health Research and from the National Cancer Institute of Canada to MS.
Comments