Implications of A Life-Long Dynamic Epigenome

Classic definition of epigenetics

Epigenetics evolved from a theory proposed by Waddington into a field of research whose main goal was explaining the mystery of how one genome encodes the multitude of cellular phenotypes represented in a multicellular organism [1,2]. Thus, epigenetics was proposed to explain changes in gene function that do not derive from a change in sequence of the DNA, but from a change in the way the gene was programmed. The classic epigenetic definition was developed to explain two widely accepted basic principles of ‘development‘. First is the principle of unidirectional irreversible differentiation, which was termed terminal differentiation. It has been generally believed that, although during development cells and tissues undergo several steps of phenotypic transformations, they reach a final terminal phenotype that is then irreversible. The second principle was the stability and heritability of the differentiated cellular phenotype. If indeed differentiation is terminal and heritable, then epigenetic processes that drive cellular development should be terminal and heritable as well. Physiological mechanisms and responses, on the other hand, are reversible and balanced. Thus, epigenetic processes were not anticipated to act as responsive lifelong physiological signals after the completion of terminal differentiation. Errors in the epigenetic markings must occur during gestation when the epigenetic markings are normally laid down. Aberrant methylation events could result in phenotypic changes, and these should again be stable and heritable within the affected cell type. It was generally believed that most of the epigenetic information is erased during early gestation, and if this erasure were complete, then errors in epigenetic markings would not be transferred across generations, but would be heritable just within the daughter cells of the somatic cell types that carried this epigenetic aberration. If an epigenetic mistake during gestation involves a region of DNA that escapes the erasure during germline maturation, then it could potentially be heritable through the germline and across generations.

A classic example of the impact of stochastic differential methylation in utero is the Agouti mouse, whose coat color is defined by the state of methylation of a retrotransposon inserted upstream to the Agouti locus [3]. Changes in maternal diet during gestation, especially those affecting the abundance of methyl donors, would then result in differences in DNA methylation that would affect the coat color phenotype or epigenotype. Epigenetic alterations according to the classic model bear some of the characteristics of genetic polymorphisms – they appear stochastically, but are then heritable. The main fundamental difference between ‘epigenetics‘ and ‘genetics‘ is that genetic polymorphisms are fixed and could not be changed by pharmacological manipulation, while ‘epigenetic alterations‘ do not involve a change in the DNA sequence itself, and are thus reversible by either pharmacological or behavioral therapy.

Epigenetic mechanisms involve histone modification and remodeling enzymes, noncoding RNAs such as microRNAs, and the covalent modification of the cytosine rings in the CG dinucleotide by methylation. DNA methylation is the only component of the epigenome that is part of the covalent structure of DNA. Tissue-specific DNA methylation patterns emerge during development, and are thus natural candidates to bear the heritability and stability of epigenetic programming. Several properties of the DNA methylation machinery were proposed to support the concept of a heritable ‘terminal‘ and stable epigenome. First, the DNA methylation reaction has been considered to be irreversible, methylation of cytosine rings is still generally believed to be a unidirectional reaction. Thus, once a methyl residue is attached to cytosine, it could not be removed by any of the known candidate enzymes including the repair enzymes. Second, maintenance DNA methyltransferases (DNMTs) were proposed to replicate the DNA methylation pattern in a semiconservative manner similar to the replication of the genetic information. This would ensure the stability and integrity of the epigenome. Indeed, the first maintenance DNMT in vertebrates that was studied in detail, DNMT1, showed preference to hemimethylated DNA in vitro[4]. This property of the enzyme predicts that it will only methylate nascent CG sites that are methylated on the parental strand, and thus accurately copy the DNA methylation pattern. Third, de novo methyltransferase enzymes that add new methyl groups to DNA were believed to be active only during early development, and to be absent in differentiated cells. The combination of these rules – first, a DNMT specific for hemimethylated DNA, second, the irreversibility of the DNA methylation pattern, and third the lack of de novo methylation in differentiated cells – ensure that DNA methylation patterns, once formed, would be maintained for the lifetime of the individual. Similarly, errors in DNA methylation that are introduced into germline precursor cells during development would be stably inherited only if they occur in regions that escape erasure early in development.

Towards a more dynamic understanding of epigenetic states

Several advances in our understanding of development and DNA methylation question these basic assumptions, and point to an alternative understanding of a more dynamic epigenome that remains so during the entire life-course of the organism. Somatic cloning [5] and derivation of pluripotent cells from adult somatic cells [6] has shaken our belief that differentiation is terminal. Both histone methylation and DNA methylation are not considered irreversible any more. Histone demethylases were discovered [7], and different mechanisms for active DNA demethylation were proposed [8–10]. Moreover, recent data showed that DNA methylation patterns are reversible in adult nondividing neurons in response to behavioral and pharmacological triggers [11,12].

The alternative hypothesis proposed here is that the DNA methylation pattern is maintained by a balance of enzymatic reactions throughout the life-course of the organism. Signaling pathways that are responsive to the different extracellular environments regulate the epigenetic enzymes. If DNA methylation patterns are maintained in a life-long equilibrium, then responses of the DNA methylation pattern to environmental cues could occur at different points in life. Since changes in DNA methylation might occur in postmitotic tissue such as neurons, liver cells or heart muscle, heritability, a classic component of the definition of epigenetics, might not be involved. However, DNA methylation changes in neurons might last for the lifetime of the individual, resembling a ‘heritable‘ alteration at least in timescale. In addition, heritability of the DNA methylation trait across generations might not involve a germline transmission of an altered DNA methylation state, but could be transmitted to the next generation by the social environment. Thus, we propose a new biological role for the commonly described epigenetic mechanisms as an adaptive process that interfaces between the environment and the genome, and adapts the function of the genome to either the real or the anticipated social physical and biological environments. Both proposed roles of epigenetics in cellular differentiation and environmental adaptation address the same challenge – deriving alternative phenotypes from a single genotype.

Epigenetics: long-term gene-expression programming

It is important to differentiate between the immediate responses of gene expression to physiological and environmental cues, and long-term programming of gene expression by epigenetic mechanisms that last long after the initial trigger is gone. Thus, it is proposed here that epigenetic mechanisms such as DNA and histone methylation and acetylation serve as a long-term memory of past exposures. This long-term gene-expression programming by epigenetic mechanisms has a lasting impact on the phenotype, and it is almost indistinguishable from a change in gene function caused by a permanent genetic mutation. The best example is perhaps the silencing of tumor suppressor genes by epigenetic mechanisms in tumors [13]. Tumor suppressor genes were originally discovered as recessive mutations driving certain familial cancers. Epigenetic silencing of tumor suppressor genes has a similar phenotypic consequence to a genetic mutation.

Similarly to sequence polymorphisms, DNA methylation alterations could serve as diagnostic tools. However, there is a cardinal difference between genetic and epigenetic silencing. While a genetic polymorphism is terminal, DNA methylation and histone modification are reversible by drugs. If the DNA methylation state is an equilibrium that could be responsive to signaling pathways in the cell, then other nonpharmacological interventions might reverse epigenetic markings as well. Therefore, understanding epigenetic aberration in disease has promising therapeutic implications in addition to its diagnostic potential.

Epigenetic variation: is it stochastic, or is there a coordinated epigenomic response to the environment?

It is quite clear that the DNA methylation and histone modification changes that occur during development are highly programmed. However, one critical question that needs to be answered is whether the DNA methylation and chromatin alterations that occur in response to different environmental cues during gestation or later in life are stochastic or programmed. It is hypothesized here that the epigenetic response to environmental cues is a well-defined adaptive response. The early-life period is an especially sensitive period for epigenetic responses to shape the life-long phenotype. It is proposed that during this period of time the epigenome is reset to handle the anticipated life-long environment, as predicted from the environmental cues that the newborn is exposed to. This early-life epigenomic adaptation could become maladaptive later in life if there is inconsistency between the anticipated and real environment, resulting in a long-term adverse impact on health and behavior. If epigenetic mechanisms are involved in such adaptive responses, they must involve programmed responses rather than stochastic drifts.

Implications of a dynamic epigenome

Expanding DNA methylation and other epigenetic mechanisms from their traditionally accepted roles in development and ‘terminal differentiation‘ to the role of a life-long interface between the environment and the genome has several important implications on diverse fields, from the social sciences to medicine. Epigenetics might provide a mechanism for biological embedding of social exposures, and the profound impact that these exposures have on the phenotype. We are already seeing the beginnings of population epigenetics. The coming years will be critical for testing whether social and physical environments have long-lasting impact on the epigenome, and whether these epigenomic responses involve related gene-expression circuitries. If the answer to these questions is positive, we will have to address the question of what are the mechanisms mediating the impact of the environment on DNA methylation and chromatin structure, and how these changes relate to the phenotype. There is data supporting the hypothesis that cellular signaling pathways elicited by neurotransmitters and other triggers might mediate the effect of the environment on the epigenome, but this needs further elucidation [14]. Understanding of these mechanisms must await a better delineation of the relationship between early and late environments, and the epigenome.

The possibility that the well-established impact of the early-life social and physical environment on health later in life is mediated by epigenetic mechanisms has obvious extremely important diagnostic and therapeutic implications. It is clear that epigenetic signatures of early-life adversity could be very important diagnostic tools. Epigenetic markings are potentially reversed by pharmacological interventions. Moreover, if social environments could impact epigenetic programs in the first place, they might also be reversed by behavioral intervention. The possibility that behavioral interventions could have biochemical consequences is extremely provocative, and might be a paradigm shift in the social sciences field. More importantly, understanding the pathways involved in these early-life responses and relating them to human disease later in life might set the stage for a new understanding of disease as a maladaptive response, rather than a series of accidental stochastic drifts. This is bound to have a profound impact on how we understand, diagnose, treat and prevent human pathology. However, many questions remain, and the main task at this stage is to test the plausibility of this hypothesis by mapping and delineating the epigenetic responses to the environment.

Financial & competing interest disclosure

Research in Moshe Szyf‘s laboratory is supported by the Canadian Institutes of Health Research and the National Cancer Institute of Canada. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Additional information

Funding

Research in Moshe Szyf‘s laboratory is supported by the Canadian Institutes of Health Research and the National Cancer Institute of Canada. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. No writing assistance was utilized in the production of this manuscript.