The social and physical environment influences human development after birth and during different life cycle stations. For example, social adversity early in life has a profound impact on life-long physical health and behavior.12–14
Thus, differentiation of the phenotype occurs in response to external signals from the social environment. Similar to cellular differentiation it involves diversification of the phenotype without altering the genotype. An attractive hypothesis is that DNA methylation might play a role in diversification of the phenotypic potential of a single genome in response to external signals during post partum development as much as it plays a role in diversification of genome function in response to innate signals of differentiation ().
Figure 1 Adaptive response of the genome in early life. Signals triggered by early life environments turn on signaling pathways in brain as well as peripheral tissues that target chromatin and DNA methylation/demethylation enzymes to specific loci in the genome (more ...)
The first evidence that early environmental exposures could alter the phenotype through altering DNA methylation patterns came from the Jirtle lab that demonstrated an effect of maternal diet on the agouti color phenotype in agouti mice which was mediated through methylation of a transposable element.15
The impact of methyl-rich diets during gestation or the impact of other chemicals during gestation could be explained just as a stochastic chemical interference in the enzymatic DNA methylation reactions that are actively laying down the DNA methylation pattern during embryogenesis. However, the responsivity of the DNA methylation pattern to social-adversity signals after birth and completion of embryogenesis as discussed below could not be explained just as a stochastic change in DNA methylation reaction kinetics.
If DNA methylation acts as a responsive biological signal even in postmitotic tissues such as neurons, the DNA methylation reaction has to be reversible;16
both demethylation and de novo methylation should be possible in nondividing tissues. It was long understood that during gestation changes in DNA methylation that sculpt the DNA methylation pattern in a tissue specific manner do occur and that these changes must be catalyzed by enzymatic processes that add and remove DNA methylation.1
For decades, DNA methylation enzymology focused on DNA methyl transferases. Several DNA methyl transferases (DNMTs) were characterized. DNMT1 is a hemimethylated DNA methyltransferase that is believed to be responsible for replicating the DNA methylation pattern during cell division and to maintain the fidelity of DNA methylation during cell division.9
DNMT3a and DNMT3b were shown to act as de novo methyltransferases.17
De novo methylation was originally believed to be limited to the early stages of development. This is essential if the DNA methylation pattern is to remain rigid after completion of DNA replication. De novo methylation in a differentiated cell would alter the DNA methylation pattern. It is clear however that DNMT3A is present in adult neurons18
supporting the possibility of change in DNA methylation in postmitotic neurons.
CG is a palindrome sequence and therefore a methylated CG in the parental strand lends itself to template dependent copying during cell division. The discovery of a large number of non-CG methylation in the genome4
raises questions on the mechanisms involved in maintaining these patterns of methylation.19
Although these non-CG methylation sites were discovered mainly in stem cells,4
it is still possible that non-CG methylation is present to a certain extent in mature cells as well.20
Methylation of nonpalindromic sequences cannot be guided by the state of methylation of the template and each round of methylation following DNA replication is essentially de novo methylation. Is there a mechanism that guides the de novo enzymes to specific sites? If there are such mechanisms and indeed there is evidence for targeting of DNMTs to specific sites, then it implies that replication of the DNA methylation pattern is not exclusively an automatic copying process. This is more consistent with a dynamic DNA methylation state.
DNA demethylation is critical for a dynamic DNA methylation pattern. DNA demethylation could occur as a passive process during cell division when DNA methyltransferases are blocked by specific factors.1
Since cell division is obviously abundant during embryogenesis, passive demethylation could theoretically explain DNA demethylation during gestation and cellular and tissue differentiation. This could explain why DNA demethylases didn't attract much attention in the past. The absence of a DNA demethylase is consistent with a rigid DNA methylation pattern post cellular differentiation. A situation where maintenance DNA methyltransferases exclusively maintain the fidelity of DNA methylation in mitotic differentiating cells in the absence of de novo methyltransferases and demethylases would prevent a drift in the DNA methylation pattern and is critical for a rigid “terminal” DNA methylation pattern that guards “terminal” differentiation. However, data that has consistently pointed to replication-independent DNA demethylation has forced us to revisit this issue and a wealth of enzymatic processes that could remove DNA demethylation in the absence of cell division have been defined.21–25
It has been shown that brain extracts are capable of demethylating “naked” DNA substrate in vitro.20,26,27
The strongest evidence for dynamic methylation-demethylation comes from several studies showing active demethylation in postmitotic neurons.18,28–30
Conditional knock out of DNMT1 in postmitotic neurons results in DNA demethylation suggesting the presence of demethylation activity in nondividing neurons which is critical for a dynamic methylation pattern in the brain.31
The main issue in the field remains however whether DNA methylation is truly a reversible reaction that involves removal of the methyl moiety and its release16,32
or whether DNA demethylation requires excision of the methylated base and its replacement by an unmethylated cytosine through a process of DNA repair. The vast majority of the data to date points to a repairbased demethylation process. First, the methylated cytosine could be removed by a glycosylase activity and the abasic site that was created is then repaired and replaced with an unmethylated cytosine.33,34
Second, DNMTs were proposed to deaminate the methyl cytosine to thymidine creating a C/T mismatch, which is then corrected by a mismatch-repair mechanism.35
DNMTs were previously shown to deaminate 5-methylcytosines36,37
under conditions of low SAM. Third, growth arrest and DNA-damage-inducible, a (GADD45A), a DNA repair protein was proposed to participate in catalysis of active DNA demethylation by an unknown DNA repair based mechanism.38
However, this was disputed.39
Other studies have suggested involvement of GADD45B in demethylation in the brain.40
Fourth, a complex sequence of coupled enzymatic reactions of deamination and mismatch repair were shown to be involved in demethylation in zebrafish: activation-induced cytidine deaminase (AID, which converts 5-meC to thymine), a G:T mismatch-specific thymine glycosylase methyl-CpG binding domain protein 4 (MBD4) and repair promoted by GADD45A.41
AID has been implicated in the global demethylation in mouse primordial germ cells as well.42
An open question is the role of the newly discovered modification 5-hydroxymethylcytosine as a potential intermediate in the DNA demethylation reaction.5
5-hydroxymethylcytosine was proposed to serve as a modification of 5-methylcytosine that marks it for base excision repair and demethylation. Recent data suggest that TET1 the enzyme that catalyzes the hydroxylation of 5-methylcytosine is present and required for stem cell maintenance of inner cell mass specification43
and for activity driven demethylation in neurons. 5-hydroxymethylation catalyzed by TET1 is followed by deamination of the 5-hydroxymethylated base by AID (activation-induced deaminase)/APOBEC (apolipoprotein B mRNA-editing enzyme complex) family of cytidine deaminases and base excision repair enzymes replace the deaminated base with an unmethylated cytosine (BER).44
The main challenge in accepting such a complex multienzyme repair-based mechanisms that involves a sequence of modifications to 5-methylcytosine as a life-long physiological process is that it invokes constant mutagenic stress and damage to the integrity of DNA; constant modification by deamination, breaking and fixing of the DNA seems to be an extremely dangerous way to maintain the DNA methylation equilibrium in both developing embryo and postmitotic neurons. Nevertheless, there is evidence for BER activity during reprogramming of mouse germ cells, a time-point in development that involves extensive DNA demethylation.45
It is unclear however why is there a need for modification of the 5-methylcytosine base first by hydroxylation and then by deamination to target it for base excision especially since glycosylases that could recognize methylated cytosines such as MBD4 and 5-methylcytosine DNA glycosylase do exist.46,47
In contrast to these complicated repair based mechanisms we have previously proposed that demethylation is truly a reversible reaction that involves removal of the methyl moiety rather than modifying and breaking the DNA and then fixing it with an unmethylated cytosine.16
We proposed that the methylated DNA binding protein MBD2 was a bona fide demethylase that removed methyl groups from DNA and truly reversed the DNA methylation reaction. This is to date the only proposed bona fide demethylase. MBD2 has been implicated in the activation of both methylated and unmethylated genes.48,49
have contested the demethylase and transcriptional activating properties of MBD2. Studies by Detich et al. have demonstrated however MBD2 demethylase activity in vitro.52
Hamm et al. have proposed an oxidative mechanism of 5-methylcytosine DNA demethylation by MBD2.53
According to this mechanism, oxidation of the methyl moiety generates 5-hydroxymethylcytosine, which is followed by release of the methyl residue as formaldehyde. Although this mechanism implicates 5-hydroxymethylcytosine in demethylation it suggests however that it is an intermediary in the demethylation enzymatic reaction rather than a modification that targets 5-hydroxymethylcytosine for excision repair.
There is no evidence that the recently described complex enzymatic reactions could catalyze demethylation of methylated DNA in vitro. Most of the evidence is based on depletion of the different predicted components of the complex in cells. However, there is evidence from several groups for enzymatic demethylation activity that doesn't require DNA repair in brain cell extracts.26,27
In a very interesting study Fuso et al. showed DNA demethylase activity in brain extracts and that depletion of MBD2 by an antibody blocks this activity supporting a role for MBD2 as a demethylase in the brain.20
Moreover the authors elegantly show that this activity is modulated in vivo by modulating one carbon metabolism. These data suggest that it is possible that DNA demethylation is a true enzymatic reaction that reveres the DNA methylation state rather than a complex of repair and modification activities. Future studies are critical for resolving this central question in DNA methylation.