Variation in DNA Methylation Is Not Consistently Reflected by Sociality in Hymenoptera

Abstract

Changes in gene regulation that underlie phenotypic evolution can be encoded directly in the DNA sequence or mediated by chromatin modifications such as DNA methylation. It has been hypothesized that the evolution of eusocial division of labor is associated with enhanced gene regulatory potential, which may include expansions in DNA methylation in the genomes of Hymenoptera (bees, ants, wasps, and sawflies). Recently, this hypothesis garnered support from analyses of a commonly used metric to estimate DNA methylation in silico, CpG content. Here, we test this hypothesis using direct, nucleotide-level measures of DNA methylation across nine species of Hymenoptera. In doing so, we generated new DNA methylomes for three species of interest, including one solitary and one facultatively eusocial halictid bee and a sawfly. We demonstrate that the strength of correlation between CpG content and DNA methylation varies widely among hymenopteran taxa, highlighting shortcomings in the utility of CpG content as a proxy for DNA methylation in comparative studies of taxa with sparse DNA methylomes. We observed strikingly high levels of DNA methylation in the sawfly relative to other investigated hymenopterans. Analyses of molecular evolution suggest the relatively distinct sawfly DNA methylome may be associated with positive selection on functional DNMT3 domains. Sawflies are an outgroup to all ants, bees, and wasps, and no sawfly species are eusocial. We find no evidence that either global expansions or variation within individual ortholog groups in DNA methylation are consistently associated with the evolution of social behavior.

Keywords: CpG o/e; Neodiprion lecontei; epigenetics; eusociality; social insects; whole genome bisulfite sequencing.

Fig. 1.

—Variation in the utility of normalized CpG depletion (CpG o/e) to detect DNA methylation. Density plots of coding sequence CpG o/e measures are shown according to WGBS DNA methylation status (n = 2400 ortholog groups with data from all species). Dashed lines represent mean values of CpG o/e and Spearman’s rank correlation coefficients (rho) between CpG o/e and mCG/CG are shown, all significant at P < 10⁻¹⁵. For some species, CpG depletion is tightly correlated with DNA methylation, and there is strong separation between methylated and unmethylated genes by CpG o/e values. However, in other species, there is substantial overlap in the CpG o/e values for methylated and unmethylated genes, resulting in a weaker correlation between this metric and DNA methylation.

Fig. 2.

—Targets and levels of DNA methylation. (A) Cladogram demonstrating species relationships and independent gains of social behavior (social terminology sensuWilson 1971; Kocher and Paxton 2014). (B) Percentage of CpG sites with significant DNA methylation (FDR-corrected binomial test P < 0.05) by genomic element classification. (C) Boxplots of coding sequence methylation levels (whiskers show observations within 1.5 * IQR of the lower and upper quartiles; n = 2400 ortholog groups with data from all species). (D) Metaplots of mean CpG methylation level over all genes with five or more exons, by position are overlaid for each species.

Fig. 3.

—Taxonomic specificity of DNA methylation. (A) The number of taxa with DNA methylation in each ortholog group (n = 2400). (B) Taxonomic specificity index of DNA methylation level in CDS grouped by the number of taxa with a methylated copy (n = 1968 ortholog groups with DNA methylation in at least one taxon; Kruskal–Wallis Rank Sum Test P < 10⁻¹⁵). (C) Tissue specificity index of ortholog gene expression among ten tissues in adult D. melanogaster, grouped by the number of hymenopteran taxa with a methylated copy (n = 1676 D. melanogaster orthologs; Kruskal–Wallis Rank Sum Test P < 10⁻¹⁵).