SPECIFICITY OF DNA METHYLATION
For quite a long time cytosine methylation in DNA was believed to occur mainly, if not exclusively, at CpG dinucleotides (Doskocil & Sorm, 1962; Grippo et al., 1968; Gruenbaum, Stein, Cedar, & Razin, 1981). Even before the methods of DNA sequencing were elaborated, we analyzed 5mC content in pyrimidine sequences isolated from DNA of various animals and plants by chemical hydrolysis and found that most 5mCs in animal DNAs are localized in the monopyrimidine fraction (Pu-5mC-Pu), whereas in plant DNAs significant 5mC quantities are present in sequences such as Pu-5mC-Pu, Pu-5mC-T-Pu, Pu-5mC-C-Pu, and Pu-5mC-5mC-Pu (Kirnos et al., 1981). Thus, animal DNAs seemed to be methylated mostly at CpG dinucleotides, whereas methylation of plant DNAs occurred both at CpG and CpHpG sites. According to our data, in plants up to ~30% 5mCs were localized in non-CpG sequences. This estimation was in a good accordance with the data obtained at the same time by a nearest neighbor method
(Gruenbaum, Naveh-Many, et al., 1981). The presence of 5mC in non-CpG sites in animal DNA was a matter of confusion for quite a long time. In a nearest neighbor analysis of DNA methylation in human spleen DNA, more than 50% total 5mC was found in non-CpG dinucleotides (Woodcock, Crowther, & Diver, 1987). At the time, these data were ascribed to inherent artifacts of the nearest neighbor method used. The matter was resolved when 5mC detection by genomic sequencing techniques arrived. It was shown that in animal cells DNA could be methylated at CpHpG sequences, both de novo and by maintenance type activity, although non-CpG methylation still seemed to be a rather rare phenomenon compared with CpG methylation (Clark et al., 1995; Toth et al., 1990). With the advent of genome-wide approaches to DNA methylation analysis, the non-CpG methylation was shown to be most prevalent in ESCs compared with somatic tissue cells (Ramsahoye et al., 2000). The dominant form of non-CpG methyla- tion in the pluripotent cell types is 5mCpA, whereas in somatic cells CpA, CpT, and CpC are very rare and about equally methylated (Ziller et al., 2011). The enzymes responsible for non-CpG methylation are Dnmt3a and Dnmt3b: KO of their encoding genes leads to a global reduction in non-CpG methylation. The methylation levels of CpG and non-CpG sites are quite distinct. CpG sites are either not methylated at all or fully methylated. In contrast, non-CpG sites have intermediate methylation levels with a median between 30% and 50% (Ziller et al., 2011). Surprisingly, considerable levels of non-CpG methylation were found in brains of mice (Xie et al., 2012) and humans (Varley et al., 2013). In mouse frontal cortex non-CpG and CpG methylations have different genomic distributions; hence, non-CpG methylation could not be a trivial byproduct of CpG methylation. CpHpG and CpHpH methylations negatively correlate with gene expression in the mouse frontal cortex, irrespective of their location in the promoter regions or gene bodies (Xie et al., 2012). Unlike ESCs, CpHpHs are more likely to be methylated than CpHpGs in the mouse and human brains (Varley et al., 2013; Xie et al., 2012). CpH methylation accumulates during early postnatal development to maximal levels of 1.3—1.5% at the end of adolescence before diminishing slightly during aging (Lister et al., 2013). The most rapid increase in DNA methylation occurs during primary phase of synaptogenesis (from 2 to 4 weeks in mouse and at the first 2 years in humans), followed by slower accumulation during later adolescence. In mice the accumulation of 5mCpH from 1 to 4 weeks coincides with a transient increase in expression of the Dnmt3a gene. In genomic DNA purified from a relatively homogeneous population of granule neurons of the adult mouse dentate gyrus, ~25% of all 5mCs were found in mCpH sites, ~4% in mCpHpGs, and ~21% in mCpHpHs (Guo et al., 2014). A majority (~83%) of the human CpH-methylated genes had orthologs that were also CpH methylated in the mouse brain. Thus, the brain CpH methylation marks the conserved sets of genes in both mice and humans. The motif analysis identified a prominent CpApC preference for CpH methylation in neurons, predicting asymmetric methylation patterns on two DNA strands. A periodicity of ~180 bp for neuronal mCpHs was observed, suggesting a relationship to nucleosome positioning. Notably, mCpHs preferentially reside in regions of low CpG density. Both neuronal CpG and CpH methylation are inversely correlated with gene expression throughout the 5' upstream, gene-body, and 3' downstream regions. Thus, CpH methylation is a new layer of epigenetic modulation of the neuronal genome. This concept is reviewed in further detail by Mukamel and Lister in Chapter 4.