Deconvolution of Hematopoietic Commitment Decisions By Genome-Wide Analysis of Progressive DNA Methylation Changes

Recent advances in single cell transcriptome analyses have resulted in the derivation of new models to describe the hierarchical organization of the mammalian hematopoietic system. While such an approach appears to be effective at discerning the trajectory of differentiation from hematopoietic stem cells (HSCs) to a given mature lineage, it remains a challenge to identify definitive points where specific lineage fates become restricted. The characterization of molecular events that correspond to such commitment decisions is critical to our interrogation of the existence and nature of serial bifurcation steps that are hypothesized to underlie the process of hematopoiesis.

5-Methylcytosine is a stable epigenetic modification, whose remodeling at specific CpG residues appears to be integral to the process of enforcing lineage-restricted gene expression programs. We have previously observed that the remodeling of the DNA methylome appears to be both progressive and irreversible during the process of hematopoietic differentiation, suggesting that this modification could be used to unambiguously identify molecular marks of lineage commitment. In order to pursue this concept further, we used tagmentation-based whole-genome bisulfite sequencing to generate a genome-wide DNA methylation map of murine hematopoiesis. This map encompasses 26 different FACS-purified populations, ranging from LT-HSCs through to terminally differentiated blood cell lineages. Across all cell populations studied, we identified 147,232 differentially methylated regions (DMRs). In line with our previous data, hierarchical clustering of these DMRs revealed coordinately regulated regions that show progressive and unidirectional lineage-specific DNA methylation dynamics during hematopoietic differentiation that would be indicative of a molecular mechanism of cell fate restriction. Single cell DNA methylome analysis indicated that methylation programming may be exclusive for a specific lineage within each cell analyzed, supporting the use of this data to establish the discreet points at which lineage commitment occurs. Along these lines, lineage-specific DMRs could already be identified within the early hematopoietic stem and multipotent progenitor compartments, supporting the concept that lineage restriction occurs early during differentiation and providing a potential molecular basis for so called lineage-priming/bias. Indeed, a phylogenetic tree of the entire hematopoietic system could be constructed purely based on methylation remodeling events that took place in the Lin^-, Sca-1⁺, c-Kit⁺ compartment. To gain further insight into how the DNA methylation programming relates to regulation of gene expression, we generated a comprehensive single cell transcriptome map encompassing the entire hematopoietic component of the bone marrow. Integration of DNA methylation dynamics with single cell gene expression dynamics provided evidence for anti-correlation between the transcriptional program and DNA methylation. However, loss of DNA methylation was not invariably associated with an increase in gene expression, suggesting that DNA methylation has more a permissive rather than an instructive role in regulating gene expression programs.

We next applied our data set to the exploration of the phenomenon of myeloid lineage bias that has been described in aged mice, by investigating the methylomes of young and aged HSCs. Compared to young HSCs, we identified 3,275 DMRs in aged HSCs, which were predominantly associated with loss of DNA methylation and affected genes involved in HSC adhesion and migration, such as Vwf and ITGB3. Remarkably, 46 % of aging DMRs overlapped with the methylome programs identified in hematopoietic differentiation and were enriched for genes involved in integrin signaling, platelet activation and aggregation. This data suggests that HSC aging results in remodeling of the DNA methylome in vivo, in a manner that is associated with a differentiation bias towards the megakaryocytic lineage. Together, our work provides a rich resource to investigate DNA methylation changes in normal or diseased hematopoiesis, across a broad range of conditions such as aging.