Identification of active regulatory regions from DNA methylation data

Abstract

We have recently shown that transcription factor binding leads to defined reduction in DNA methylation, allowing for the identification of active regulatory regions from high-resolution methylomes. Here, we present MethylSeekR, a computational tool to accurately identify such footprints from bisulfite-sequencing data. Applying our method to a large number of published human methylomes, we demonstrate its broad applicability and generalize our previous findings from a neuronal differentiation system to many cell types and tissues. MethylSeekR is available as an R package at www.bioconductor.org.

Figure 1.

Identification of regulatory regions from Bis-seq data. (a) Typical methylation pattern in mammalian methylomes (dots represent individual CpGs, methylation levels averaged over three consecutive CpGs). UMRs (blue rectangles) and LMRs (red triangles) are identified as regions with methylation levels <50% (dashed gray line). CGI: CpG islands. (b) The number of CpGs per hypomethylated region versus its median methylation. The regions separate into two classes: CpG-rich, unmethylated UMRs and CpG-poor LMRs with residual methylation. (c) Some methylomes contain regions of highly disordered methylation (PMDs, orange bar, dots represent individual CpGs), which need to be identified and masked for the identification of regulatory regions. Unsmoothed methylation levels are shown. (d) Workflow of MethylSeekR.

Figure 2.

LMRs are highly dynamic distal regulatory elements. (a) Methylation profiles for trophoblasts differentiated from H1 (h1_bmp4) and B cells (bcell) for the same locus as in Figure 1a. UMRs and LMRs are shown as blue rectangles and red triangles, respectively, and n indicates the minimal number of CpGs required to identify a region. (b) Number of UMRs and LMRs identified in selected human methylomes: H1 ESCs (h1), trophoblasts differentiated from H1 (h1_bmp4), fetal lung fibroblasts (imr90), adipocytes differentiated from adipose-derived stem cells (ads_adipose), hematopoietic stem and progenitor cells (hspc) and B cells (bcell). The regions have been grouped by the number of cell types they exist in. (c) Transcription factor motif enrichments for cell type-specific and constitutive LMRs. Only motifs with an enrichment >1.5 in at least one cell type are shown.