Cross-tissue patterns of DNA hypomethylation reveal genetically distinct histories of cell development

Abstract

Establishment of DNA methylation (DNAme) patterns is essential for balanced multi-lineage cellular differentiation, but exactly how these patterns drive cellular phenotypes is unclear. While >80% of CpG sites are stably methylated, tens of thousands of discrete CpG loci form hypomethylated regions (HMRs). Because they lack DNAme, HMRs are considered transcriptionally permissive, but not all HMRs actively regulate genes. Unlike promoter HMRs, a subset of non-coding HMRs is cell-type specific and enriched for tissue specific gene regulatory functions. Our data further argues not only that HMR establishment is an important step in enforcing cell identity, but also that complex HMR patterns are functionally instructive to gene regulation. To understand the significance of non-coding HMRs, we systematically dissected HMR patterns across diverse human cell types and developmental timepoints, including embryonic, fetal, and adult tissues. Unsupervised clustering of 102,390 distinct HMRs revealed that levels of HMR specificity reflects a developmental hierarchy supported by enrichment of stage-specific transcription factors and gene ontologies. Using a pseudo-time course of development from embryonic stem cells to adult stem and mature hematopoietic cells, we find that most HMRs observed in differentiated cells (~70-75%) are established at early developmental stages and accumulate as development progresses. HMRs that arise during differentiation frequently (~35%) establish near existing HMRs ([≤] 6kb away), leading to the formation of HMR clusters associated with stronger enhancer activity. Using SNP-based partitioned heritability from GWAS summary statistics across diverse traits and clinical lab values, we discovered that genetic contribution to trait heritability is enriched within HMRs. Moreover, the contribution of heritability to cell-relevant traits increases with both increasing developmental specificity and HMR clustering, supporting the role of distinct HMR subsets in regulating normal cell function. Altogether, our findings reveal that HMRs can predict cellular phenotypes by providing genetically distinct historical records of a cells journey through development. AUTHOR SUMMARYStudies aiming to understand the relationship between DNA methylation patterns and phenotypic outcomes have focused largely on individual differentially methylated regions without consideration of combinatorial changes that drive phenotypes. In non-disease contexts, most of the human genome is stably methylated, except for thousands of discrete DNA hypomethylated regions (HMRs) coinciding with gene regulatory elements. Here, we comprehensively characterize HMR relationships both within and between developmentally diverse cell types to understand the functional significance of complex HMR patterns. We show that levels of HMR specificity across cell-types captures time-point specific branchpoints of development. Our analysis further reveals that HMRs form clusters in proximity to cell identity genes and are associated with stronger gene enhancer activity. This is a wide-spread phenomenon and only a very small subset of HMR clusters is explained by overlapping super-enhancer annotations. Partitioned heritability revealed the functional significance of different HMR patterns linked to specific phenotypic outcomes and indicates a quantitative relationship between HMR patterns and complex trait heritability. Altogether, our findings reveal that HMRs can predict cellular phenotypes by providing genetically distinct historical records of a cells journey through development, ultimately providing novel insights into how DNA hypo-methylation mediates genome function.