Position-dependent correlations between DNA methylation and the evolutionary rates of mammalian coding exons

Abstract

DNA cytosine methylation is a central epigenetic marker that is usually mutagenic and may increase the level of sequence divergence. However, methylated genes have been reported to evolve more slowly than unmethylated genes. Hence, there is a controversy on whether DNA methylation is correlated with increased or decreased protein evolutionary rates. We hypothesize that this controversy has resulted from the differential correlations between DNA methylation and the evolutionary rates of coding exons in different genic positions. To test this hypothesis, we compare human-mouse and human-macaque exonic evolutionary rates against experimentally determined single-base resolution DNA methylation data derived from multiple human cell types. We show that DNA methylation is significantly related to within-gene variations in evolutionary rates. First, DNA methylation level is more strongly correlated with C-to-T mutations at CpG dinucleotides in the first coding exons than in the internal and last exons, although it is positively correlated with the synonymous substitution rate in all exon positions. Second, for the first exons, DNA methylation level is negatively correlated with exonic expression level, but positively correlated with both nonsynonymous substitution rate and the sample specificity of DNA methylation level. For the internal and last exons, however, we observe the opposite correlations. Our results imply that DNA methylation level is differentially correlated with the biological (and evolutionary) features of coding exons in different genic positions. The first exons appear more prone to the mutagenic effects, whereas the other exons are more influenced by the regulatory effects of DNA methylation.

Fig. 1.

(A) Pearson’s r between CpG_O/E and the mCG density for the six analyzed cell lines. (B) Pearson’s r between CpG_O/E and the mCG density for the first, last, and internal exons in the six cell lines. (C) Comparison of the median evolutionary measurements (d_N/d_S, d_N, and d_S) in the first, last, and internal exons. Statistical significance was estimated using Pearson’s r (A and B) and a two-tailed Wilcoxon rank-sum test (C): *P < 0.05, **P < 0.01, and ***P < 0.001. NS, not significant. The dashed curves in C indicate the statistical significance in comparisons between the different exon groups.

Fig. 2.

Spearman’s rank coefficient of correlation (ρ) between mCG density and (A) d_N/d_S; (B) d_N; and (C) d_S. The left and right columns under each heading illustrate the correlations in human–macaque and human–mouse comparisons before and after controlling for the four potential confounding factors (CpG density, G + C content, exon length, and the CSE/ASE exon type). *P < 0.05, **P < 0.01, and ***P < 0.001.

Fig. 3.

The correlations among mCG density, evolutionary rates, sample specificity of mCG Density (τ_m), and exonic expression level in (A) the first exons and (B) the last or internal exons. Green and red lines represent positive and negative correlations, respectively. NS, not significant.