NON-CG METHYLATION AND BRAIN CELL DIVERSITY
Maintaining a diverse population of specialized cell types is important in most tissues, and it is especially critical for the function of brain circuits. Cognitive processes such as perception, memory, and motor control rely on the balanced interaction, via coordinated gene expression, protein synthesis, and synaptic signaling, of a myriad of excitatory and inhibitory neuron types. Distinct neuronal cell types express unique DNA-binding transcription factors, have different histone modification profiles, and also differ substantially in their landscape of CG and non-CG DNA methylation. Among these cell type-specific characteristics, the pattern of non-CG methylation is one of the most distinctive molecular-genetic fingerprints of cell type-specific identity. In a comparison of three neuronal cell types (excitatory pyramidal cells, and inhibitory cells expressing parvalbumin or vasoactive intestinal peptide), almost half of all protein-coding genes were found to harbor neuron type-specific non-CG methylation (Mo et al., 2015). The correlation of gene body mCH levels between neuronal cell types (r = 0.83-0.86) was lower than the corresponding correlation of transcriptional levels by RNA-seq (r = 0.95-0.96), again suggesting that mCH captures essential aspects of neuronal epigenetic diversity. By contrast, both CG and non-CG methylation are precisely conserved in replicate experiments using tissue samples from different individuals, and non-CG methylation is conserved at homologous sequences between human and mouse brain neurons (Lister et al., 2013).
In cell types that harbor non-CG methylation, the mark appears across almost all genomic compartments, including exons, introns, and outside of gene bodies. Active regulatory elements located outside of gene promoters, as defined by a suite of epigenetic and chromatin marks, are depleted of non-CG methylation (Lister et al., 2009; Mo et al., 2015). DMRs, which are classically defined based on statistically significant differences in the level of CG methylation, are also marked by differential non-CG methylation as well as cell type-specific active histone marks (Lister et al., 2013; Mo et al., 2015). These active regulatory regions are also marked by open chromatin as assayed by in vitro transposition of native chromatin by Tn5 transposase (ATAC-seq) (Buenrostro, Giresi, Zaba, Chang, & Greenleaf, 2013). By computationally identifying ATAC-seq footprints lying over transcription factor sequence motifs, putative cell type-specific transcription factor binding sites could be profiled. Cell type-specific footprints corresponding to more than 100 transcription factors were depleted for both CG and non-CG methylation. Only two factors, CTCF and Zfp410, seemed to lack this correlation between transcription factor binding and methylation level (Mo et al., 2015), which suggests a tight relationship between methylation level and regulatory activity. However, this association remains correlational and the causal role of dynamic methylation at these sites remains to be determined (Schubeler, 2015).
In addition to gene expression and transcription factor binding, DNA methylation in both the CG and non-CG contexts can be modulated by the pattern of nucleosome occupancy (Chodavarapu et al., 2010; Lister et al., 2009). In neurons, autocorrelation analysis of non-CG methylation shows a weak modulation with periods of ~10 bp, corresponding to the DNA helix coil length, and ~180 bp, corresponding to nucleosome spacing (Lister et al., 2013). This phased modulation of DNA methylation is similar in plants and mammals (Chodavarapu et al., 2010), suggesting it is a universal feature of DNA methyltransferase activity. Integrating MethylC-seq data with nucleosome positions estimated from ATAC-seq chromatin accessibility data showed that both CG and non-CG methylation are depleted at the center of nucleosomes and enriched in the space between nucleosomes (Mo et al., 2015).