Evolutionary expansion of DNA hypomethylation in the mammalian germline genome

Abstract

DNA methylation in the germline is among the most important factors influencing the evolution of mammalian genomes. Yet little is known about its evolutionary rate or the fraction of the methylome that has undergone change. We compared whole-genome, single-CpG DNA methylation profiles in sperm of seven species-human, chimpanzee, gorilla, rhesus macaque, mouse, rat, and dog-to investigate epigenomic evolution. We developed a phylo-epigenetic model for DNA methylation that accommodates the correlation of states at neighboring sites and allows for inference of ancestral states. Applying this model to the sperm methylomes, we uncovered an overall evolutionary expansion of the hypomethylated fraction of the genome, driven both by the birth of new hypomethylated regions and by extensive widening of hypomethylated intervals in ancestral species. This expansion shows strong lineage-specific aspects, most notably that hypomethylated intervals around transcription start sites have evolved to be considerably wider in primates and dog than in rodents, whereas rodents show evidence of a greater trend toward birth of new hypomethylated regions. Lineage-specific hypomethylated regions are enriched near sets of genes with common developmental functions and significant overlap across lineages. Rodent-specific and primate-specific hypomethylated regions are enriched for binding sites of similar transcription factors, suggesting that the plasticity accommodated by certain regulatory factors is conserved, despite substantial change in the specific sites of regulation. Overall our results reveal substantial global epigenomic change in mammalian sperm methylomes and point to a divergence in trans-epigenetic mechanisms that govern the organization of epigenetic states at gene promoters.

Figure 1.

Mammalian sperm methylome characteristics. (A) The number and average size of HMRs in native assemblies. (B) Hierarchical clustering of aligned seven-way orthologous methylomes of multiple species and cell types. (C) The number of promoter HMRs and nonpromoter HMRs in seven-way orthologous sperm methylomes. Dashed lines indicate the average number of conserved HMRs across species. (D) Sperm DNA methylation of seven species in an example orthologous region. Methylome alignment is shown along with conservation tracks from MULTIZ alignment of 100 vertebrates and human repeat elements by RepeatMasker (Smit et al. 2013–2015). Regions in solid boxes show divergent methylation states at well-conserved elements. See zoomed-in browser image for dashed boxes in Supplemental Figure S2A. (E) The median size of hypomethylated regions upstream of and downstream from TSS in somatic (orange) and sperm (blue) methylomes of different species. HMR sizes are measured in their respective native genomes. Whiskers mark the 25th and 75th percentiles of HMR sizes upstream of or downstream from TSS. Wilcoxon rank-sum tests for all pairs of species between rodent species and nonrodent species showed significantly narrower HMRs in rodents.

Figure 2.

Methylation loss exceeds methylation gain during evolution. (A) Phylogenetic tree representing consensus divergence time (Hedges et al. 2015) genome evolution and sperm methylome evolution. Unit branch length represents 1 million yr, one substitution/site, and one methylation state change/CpG site, respectively. (B) Schematic segmentation of the orthologous genome according to the history of methylation evolution and annotation of methylation mutation events. (C) Fraction of orthologous genome hypomethylated in extant and ancestral species, estimated by the interdependent-site phylo-epigenetic model. (D) Total size of different types of methylation evolution events on individual branches. (E) The distribution of distances from methylome evolution events to closest TSS in the seven-way orthologous genome. The color legend is as shown in D.

Figure 3.

Older and more conserved sperm HMRs are wider. (A) Mean sizes of promoter and nonpromoter HMRs in the seven-way orthologous genome of human and mouse sperm, grouped by estimated HMR emergence time in mammalian evolution. Error bars indicate standard errors of the mean. Significant Wilcoxon rank-sum test P-values are shown. The Boreoeutheria group contains HMRs conserved in all seven species, representing HMRs formed before and during the speciation of Boreoeutheria. Other groups represent HMRs formed during the speciation of respective lineages. (B) Distribution of HMR sizes in native genome assemblies, grouped by genomic context and conservation level.

Figure 4.

Sequence signatures driven by sperm methylome evolution. (A) CpG enrichment (observed/expected ratio of CpG occurrences) and regional average DNA methylation level in human HMRs in seven-way orthologous genome grouped by estimated hypomethylation age. (B) Two opposite scenarios explaining the evolution path to different promoter HMR sizes in two extant species and the expected CpG enrichment schematic profile in the extant species as a result of methylation-induced CpG decay. (C) Human and mouse CpG enrichment profile at hypomethylated orthologous gene promoters, where human sperm HMRs are wider than mouse. Profiles show CpG enrichment in 2-bp windows by distance to indicated HMR boundaries. Smoothed lines were generated with local polynomial regression fitting (LOESS).

Figure 5.

Sperm DNA methylation divergence is accompanied by sperm histone modification and DNA sequence divergence. (A) Mouse- and human-specific sperm HMRs and HMR extensions show species-specific H3K4me3 and H3K27me3 enrichment. (B) Lineage-specific HMRs are associated with relatively more sequence substitutions. For each pair of parallel lineages, expected distribution of RSSR is shown as a gray histogram, and fit is shown with normal distribution (black curve). The yellow area marks 90% confidence interval for mean RSSR. Upper (lower) tail of the distribution indicates significantly more substitutions in the numerator (denominator) lineage. Each lollipop marks the observed RSSR in HMRs specific to the indicated lineage, with one-sided P-value shown on top.

Figure 6.

HMRs gained on parallel lineages are associated with similar transcription factors and developmental genes. (A) Enrichment of transcription factor binding sites in HMR birth and extension regions on the human lineage and mouse lineage. (*) Factors enriched in both lineages. (B) Lineage-specific HMRs are associated with gene sets with significant overlap and that are enriched in developmental processes.