CpG-level DNA methylation analysis identifies molecular events that are definitive of progressive lineage commitment in hematopoiesis

Abstract

Abstract Epigenetic mechanisms such as DNA methylation play critical roles in orchestrating hematopoietic differentiation, yet the fine-scale dynamics of methylation programming during lineage commitment remain poorly understood. Traditional analyses focus on differentially methylated regions (DMRs), which may fail to account for heterogeneous behavior of individual CpGs within these intervals, and hence potentially obscure the temporal and functional organization of the regulatory events underlying hematopoietic cell fate decisions. We aimed to resolve the temporal and lineage-specific architecture of DNA methylation programming in murine hematopoiesis at single-CpG resolution and sought to identify epigenetic features that mark lineage commitment and early fate priming. We profiled 25 immunophenotypically defined murine hematopoietic populations by whole-genome bisulfite sequencing. These included hematopoietic stem cells (HSCs), multipotent progenitors (MPP1 - MPP5), and downstream erythroid, myeloid, dendritic, and lymphoid cells. Differentially methylated regions (DMRs) were identified and deconvoluted into differentially methylated CpGs (dmCpGs), which were clustered based on shared programming dynamics. To link epigenetic states with transcriptional output, we performed combined single-cell bisulfite and RNA sequencing (scMT-seq) on over 550 hematopoietic stem and progenitor cells (HSPCs) and mature blood cells. We identified over 122,000 DMRs that formed 28 lineage or cell type-specific clusters. CpG-level analysis of the over 593,000 dmCpGs revealed that DNA methylation changes within DMRs are modulated by systematically ordered programming events occurring at distinct differentiation stages, which indicated pseudotemporal progression of DNA methylation changes within DMRs. Based on this finding, we established a DMR “expansion state” framework that starts with the focal loss of methylation at “seed” CpGs, followed by sequential demethylation of “primary” and “tuning” CpGs along the differentiation trajectories of all major hematopoietic lineages. Using this integrated map of DMR and dmCpG programming of hematopoiesis, we were able to robustly classify the epigenetic states of individual cells from scMT-seq data despite the method-inherent sparse coverage. This identified 15 methylation-defined HSPC states that recapitulated progression through a shared early trajectory, characterized by multi-lineage priming within multipotent progenitor cells, followed by bifurcating erythroid vs. lympho-myeloid fate decisions. Applying the DMR “expansion state” framework to the scMT-seq data, we confirmed at the single cell-level that CpGs within DMRs are demethylated in a fixed sequence (seed → primary → tuning). Early dmCpG-specific seeding events occurring in progenitors were retained in mature cells from multiple lineages. For example, B cells showed demethylation at seed dmCpGs in T cell-specific DMRs, while full demethylation of these DMRs was restricted to T cells. This suggested that mature cells retain an epigenetic memory of early multi-lineage priming. Integration of the methylome and transcriptome layers of the scMT-seq data revealed that the initial multi-lineage priming observed within early progenitor cells was associated with decreasing expression of stemness marker genes, but was not immediately associated with measurable upregulation of lineage-specific transcription programs. Progression along the epigenetically defined erythroid and lympho-myeloid trajectories was correlated with the activation of lineage-defining gene expression programs. In summary, this study provides the first high-resolution, CpG-resolved map of DNA methylation programming across murine hematopoiesis, revealing a highly ordered, pseudotemporal sequence of epigenetic changes that precede and define lineage commitment. Our findings redefine how DMRs function as regulatory units, showing that individual CpGs within DMRs possess distinct temporal programming logic. Our CpG-resolved framework enables refined trajectory inference, uncovers hidden heterogeneity within progenitor populations, and offers a powerful resource for investigating the epigenome of normal hematopoiesis and its dysregulation in hematopoietic malignancies. Furthermore, our work provides a model for how progressive and modular CpG-level control of DMR regulatory states may operate in other differentiating systems.