A human DNA methylation atlas reveals principles of cell type-specific methylation and identifies thousands of cell type-specific regulatory elements

Abstract

DNA methylation is a fundamental epigenetic mark that governs chromatin organization, cell identity, and gene expression. Here we describe a human methylome atlas, based on deep whole-genome bisulfite sequencing allowing fragment-level analysis across thousands of unique markers for 39 cell types sorted from 207 healthy tissue samples. Replicates of the same cell-type are >99.5% identical, demonstrating robustness of cell identity programs to genetic variation and environmental perturbation. Unsupervised clustering of the atlas recapitulates key elements of tissue ontogeny, and identifies methylation patterns retained since gastrulation. Loci uniquely unmethylated in an individual cell type often reside in transcriptional enhancers and contain DNA binding sites for tissue-specific transcriptional regulators. Uniquely hyper-methylated loci are rare and are enriched for CpG islands, polycomb targets, and CTCF binding sites, suggesting a novel role in shaping cell type-specific chromatin looping. The atlas provides an essential resource for interpretation of disease-associated genetic variants, and a wealth of potential tissue-specific biomarkers for use in liquid biopsies.

Cell-free, methylated DNA in blood samples reveals tissue-specific, cellular damage from radiation treatment

“…We segmented the genome into blocks of homogenous methylation as previously described in Loyfer et al 2022 using wgbstools (with parameters segment --max_bp 5000) 22, 39 . In brief, a multi-channel Dynamic Programming segmentation algorithm was used to divide the genome into continuous genomic regions (blocks) showing homogenous methylation levels across multiple CpGs, for each sample.…”

“…This resulted in a variable number of cell-type specific blocks available for each tissue and cell type. Each DNA fragment was characterized as U (mostly unmethylated), M (mostly methylated) or X (mixed) based on the fraction of methylated CpG sites as previously described 22 . We used thresholds of :S33% methylated CpGs for U reads and 2’66% methylated CpGs for M. We then calculated a methylation score for each identified cell-type specific block based on the proportion of U/X/M reads among all reads.…”

“…We segmented the genome into blocks of homogenous methylation as previously described in Loyfer et al 2022 using wgbstools (with parameters segment --max_bp 5000) 22,39 CpG sites and coverage across 90% of samples were selected. We computed the average methylation for each block and sample using wgbstools (--beta_to_table).…”

“…This resulted in a variable number of cell-type specific blocks available for each tissue and cell type. Each DNA fragment was characterized as U (mostly unmethylated), M (mostly methylated) or X (mixed) based on the fraction of methylated CpG sites as previously described 22 .…”

“…The short half-life of cfDNA (15 mins - 2 hours) is ideal for monitoring real-time changes in tissue homeostasis due to therapeutic interventions 27,28 . Also, few cfDNA analyses have taken advantage of CpG pattern analysis to increase sensitivity and specificity of cell type proportion estimates 22,23,29–31 . Since each cfDNA molecule has independent origins, pattern analysis of sequence reads allows for individual classification of each fragment as opposed to traditional methods that average the methylation status across a population of fragments aligned at single CpG sites 27,28 .…”