DNA methylation is an important epigenetic modification, although the regulation of its deposition is not well-understood. Furthermore, our current understanding is mostly limited to CpG methylation; methylation can also occur at cytosines in other contexts. Two recent studies shed light on the regulation of CpG methylation deposition during differentiation and on the patterns of non-CpG methylation.

Stadler and colleagues made base-pair-resolution maps of CpG methylation in mouse embryonic stem cells (ESCs) by performing whole-genome bisulphite sequencing. As expected, they identified methylated genomic regions and non-methylated CpG islands. They also found a class of regions that they termed low-methylated regions (LMRs) that had levels of CpG methylation of 10–50%. These LMRs contain features that are characteristic of distal regulatory elements, and they function as enhancers in experimental assays. Mapping DNA methylation in neural precursor cells showed that LMRs are specific to particular cell types. Furthermore, the formation of an LMR is linked to an increase in gene expression at adjacent genes and remethylation correlates with reduced expression. Using chromatin immunoprecipitation followed by sequencing (ChIP–seq), the authors identified cell-type-specific distribution of transcription factors that were bound to these LMRs. Knocking out the transcription factor REST or ablating binding sites for the transcriptional regulator CTCF in ESCs showed that, for these two cases, DNA binding is necessary and sufficient for reduced methylation at the LMRs at which they bind. The cell-type-specific patterns show that DNA methylation is more plastic during differentiation than was previously thought and that cell-specific distribution of transcription factors may be responsible for regulating this plasticity. Furthermore, the identification of active regulatory regions may be performed by only considering DNA methylation.

A different type of variation in DNA methylation patterns was studied by Ziller and colleagues. These authors investigated the genome-wide levels of non-CpG methylation in human pluripotent and somatic cell lines by reduced representation bisulphite sequencing (RRBS) — a technique that they showed was able to capture a small but representative fraction of the non-CpG methyl sites. Consistent with previous studies, they found that non-CpG methylation was much more widespread in pluripotent cells, including in induced pluripotent cells, than in somatic cells. However, they showed that non-CpG methylation is highly variable in pluripotent cells and that there are differences among ESC lines and among cells from the same cell line that have undergone different numbers of passages. That is, non-CpG methylation seems to be more prone to stochastic variation than CpG methylation. Knocking down the de novo methylases DNMT3A and DNMT3B in human ESCs resulted in decreased levels of non-CpG methylation. Furthermore, non-CpG methylation was found to be correlated with CpG methylation that is in close vicinity to it, suggesting that the deposition of the two is somehow linked.

These two studies both reveal new aspects to DNA methylation dynamics; it will be interesting to explore further the mechanisms that are involved in these changes in CpG and non-CpG methylation.