DNA methylation is associated with an increased level of conservation at nondegenerate nucleotides in mammals

Abstract

DNA methylation at CpG dinucleotides can significantly increase the rate of cytosine-to-thymine mutations and the level of sequence divergence. Although the correlations between DNA methylation and genomic sequence evolution have been widely studied, an unaddressed yet fundamental question is how DNA methylation is associated with the conservation of individual nucleotides in different sequence contexts. Here, we demonstrate that in mammalian exons, the correlations between DNA methylation and the conservation of individual nucleotides are dependent on the type of exonic sequence (coding or untranslated), the degeneracy of coding nucleotides, background selection pressure, and the relative position (first or nonfirst exon in the transcript) where the nucleotides are located. For untranslated and nonzero-fold degenerate nucleotides, methylated sites are less conserved than unmethylated sites regardless of background selection pressure and the relative position of the exon. For zero-fold degenerate (or nondegenerate) nucleotides, however, the reverse trend is observed in nonfirst coding exons and first coding exons that are under stringent background selection pressure. Furthermore, cytosine-to-thymine mutations at methylated zero-fold degenerate nucleotides are predicted to be more detrimental than those that occur at unmethylated nucleotides. As zero-fold and nonzero-fold degenerate nucleotides are very close to each other, our results suggest that the "functional resolution" of DNA methylation may be finer than previously recognized. In addition, the positive correlation between CpG methylation and the level of conservation at zero-fold degenerate nucleotides implies that CpG methylation may serve as an "indicator" of functional importance of these nucleotides.

Keywords: DNA methylation; degeneracy of nucleotide; genomics; methylation-associated mutation; single-nucleotide evolution.

Fig. 1.

Comparisons of evolutionary properties among different types of exonic regions for the six cell types examined. (A) Examples of six groups of exonic regions categorized: G1, 5U-F; G2, 5U-nonF; G3, CDS-F; G4, CDS-nonF; G5, 3U-F; and G6, 3U-nonF. (B) Pearson’s coefficient of correlation (r) between CpG_O/E and mCG density. (C) Average conservation scores (phastCons scores) of exons for the six exonic groups.

Fig. 2.

Comparisons of the conservation scores (the median PhyloP scores) at methylated and unmethylated sites in six groups of exonic regions across six cell types examined (S1–S6) (A) before and (B) after dividing each of the exonic region group into lowly conserved (left column) and highly conserved regions (right column). The statistical significance was evaluated by using the two-tailed Wilcoxon rank-sum test.

Fig. 3.

Comparisons of the median PhyloP scores of zero-fold (i = 0), two-/three-fold (i = 2 or 3), and four-fold (i = 4) degenerate nucleotides at methylated and unmethylated sites in the coding exonic regions (including CDS-F and CDS-nonF regions) (A) before and (B) after dividing each of the exonic region group into lowly conserved (left column) and highly conserved regions (right column). The statistical significance was evaluated by using the two-tailed Wilcoxon rank-sum test.

Fig. 4.

The percentages of damaging C-to-T mutations at methylated and unmethylated zero-fold-degenerate sites (measured by the SIFT and PolyPhen-2 predictions) and the odds ratios of such substitutions occurring at methylated sites being predicted to be damaging over those occuring at unmethylated sites in the lowly (left column) and highly (right column) conserved regions of the (A) CDS-F and (B) CDS-nonF regions. The statistical significance was evaluated by using the two-tailed Fisher’s exact test: *P < 0.05, **P < 0.01, and ***P < 0.001. NS, not significant.