Intermediate DNA methylation is a conserved signature of genome regulation

Abstract

The role of intermediate methylation states in DNA is unclear. Here, to comprehensively identify regions of intermediate methylation and their quantitative relationship with gene activity, we apply integrative and comparative epigenomics to 25 human primary cell and tissue samples. We report 18,452 intermediate methylation regions located near 36% of genes and enriched at enhancers, exons and DNase I hypersensitivity sites. Intermediate methylation regions average 57% methylation, are predominantly allele-independent and are conserved across individuals and between mouse and human, suggesting a conserved function. These regions have an intermediate level of active chromatin marks and their associated genes have intermediate transcriptional activity. Exonic intermediate methylation correlates with exon inclusion at a level between that of fully methylated and unmethylated exons, highlighting gene context-dependent functions. We conclude that intermediate DNA methylation is a conserved signature of gene regulation and exon usage.

Figure 1. IM is predominantly tissue specific.

(a) Top panel, comparison of WGBS methylation levels at CpGs carrying only MRE-seq or MeDIP-seq reads, and CpGs within IM regions. Bottom panel, comparison of 450k Infinium array methylation levels at CpGs in IM regions and outside of IM regions (66% of all IM regions overlap one or more methylation array probes). A value of 0 is unmethylated, a value of 1 is fully methylated. (b) Comparison of the number of IM regions specific to one or more of the four tissues studied. (c) Hierarchical clustering of cell type similarity based on the presence or absence of IM status. Distance metric is Jaccard; clustering method is average. (d) The known imprinted locus in the body of the Rb gene was detected as IM in all tissues except ES cells using MeDIP-seq/MRE-seq. (e) A breast-specific IM region. (d,e) Height for all tracks shows a signal range from 0 to 50 reads.

Figure 2. IM is associated with intermediate levels of epigenomic modifications and gene expression.

(a) Distribution of IM CpGs over Refseq genome feature annotations. (b) Distance from each IM region to the nearest DHS (P<0.001, χ²; odds ratio=2.53). DHSs were compiled from 41 different cell types, covering approximately 8% of the genome. (c) Comparison of H3K4me3 and H3K4me1 signals between methylated (Meth), unmethylated (Unmeth) and IM regions in five different cell types using a generalized additive model (GAM; grey outlines indicate 95% confidence interval (CI); *donor for histone ChIP-seq does not match donor used in IM analysis). (d) Comparison of DHS signals between Meth, Unmeth and IM regions (grey outlines indicate 95% CI; *donor for DNase-seq does not match donor used in IM analysis). (e) Top panel, average whole-transcript expression of genes associated with Meth, Unmeth and IM regions. Bottom panel, average exon expression relative to its gene expression for exons within 1 kb of Meth, Unmeth and IM regions (error bars represent s.e.m.; **P<0.005, Wilcoxon; ***P<0.0001, Wilcoxon). Expression analysis was based on mRNA-seq data from breast myoepithelial cells. Total transcript-associated regions: IM=6,776; Meth=3,270; Unmeth=5,605. Total exon-associated regions: IM=14,336; Meth=6,642; Unmeth=9,331.

Figure 3. Characterization of ASM and AIM regions.

(a) A novel ASM region in exon 1 of PTCHD3 validated by clonal bisulfite sequencing. Height for all tracks shows a signal range from 0 to 50 reads. (b) Scatter plots showing separation of ASM and AIM SNPs in imprinting control regions (ICRs) and in all IM regions based on allelic preference in MeDIP-seq and MRE-seq reads. Bar graphs showing relative proportions of ASM and AIM SNPs, and proportions of whole ASM and AIM IM regions based on the presence of >1 ASM or AIM SNP per region. (c) Comparison of methylation array levels between CpGs in ICRs, ASM and AIM. A value of 0 is unmethylated, and a value of 1 is fully methylated. (d) Comparison of allelic preference between histone modifications, unmethylated DNA (MRE-seq, top panel) and methylated DNA (MeDIP-seq, bottom panel) at ASM SNPs from fetal brain, in which the heterozygous genotype was verified by whole-genome sequencing. A positive correlation indicates that signals are on the same allele, whereas negative correlation indicates that signals are on opposite alleles.

Figure 4. The IM state is conserved in syntenic loci in mouse.

(a) A novel, tissue-specific human IM region in an internal exon of DCHS1 shows conserved IM state at the orthologous exon in mouse. Height for all tracks shows a signal range of 0–50 reads. (b) The pie chart indicates the distance to the nearest human IM region from each aligned mouse IM region. The bar graph shows the fold-enrichment of overlap between human and mouse IM regions at the CpG level using the complete set of human IM regions and a set restricted to cell types analogous to those in the mouse IM analysis. (c) Average phastCons conservation scores over all IM regions and regions that do not overlap coding exons. Scores are based on alignment of 46 vertebrate species.