DNMT3A is a de novo DNA methyltransferase that is crucial to the maintenance of normal hematopoiesis. DNMT3A is frequently mutated in hematopoietic diseases, with mutations reported in over 20% of adult acute myeloid leukemia (AML), and 10-15% of myelodysplastic syndrome (MDS) patients. Our lab has shown that loss of Dnmt3a in mice leads to enhanced self-renewal of the HSC compartment and a block in differentiation, contributing to its role in clonal hematopoiesis (Challen et al. 2011). We have also shown that loss of DNMT3A leads to a spectrum of hematologic malignancies; however, changes in DNA methylation alone are insufficient to explain the leukemic transformation, suggesting that DNMT3A could also regulate stem cells independent of methylation. The mechanisms through which DNMT3A exerts its specific effects are unclear, in part due to our relatively poor understanding of its interactome.

To address this gap in knowledge, we generated a doxycycline-inducible biotinylated-DNMT3A transgene in mouse ES cells. Using co-immunoprecipitation and mass spectrometry analysis of biotinylated-DNMT3A in murine ESCs, we identified several candidate interacting proteins. Along with known partners such as DNMT3L and DNMTB, factors associated with chromatin regulation (HDACs, NuRD complex), splicing factors (SF proteins, Hnrnps, and DDXs), and ribosomes (RPSs and RPLs) were pulled down. We validated these interactions by over-expressing myc-tagged DNMT3A in HEK293T cells. Since epigenetic regulators and splicing factors are recurrently co-mutated in hematologic malignancies, we validated its interaction with splicing factors such as DDX39B, SF3B3, and SF3B1 (among others) through co-IP of DNMT3A in CD34+ HSPCs. Furthermore, we found that these interactions are RNA and DNA-independent. N-terminal deletions of DNMT3A domains revealed that the catalytic domain is crucial for its interaction with splicing factors, and AML mutant, Arg882His, disrupts this interaction. These data suggest that through its interaction with splicing factors, DNMT3A may have a crucial role in RNA splicing efficacy.

To determine how the putative interaction between DNMT3A and splicing factors affects hematopoiesis, we examined whether loss of Dnmt3a impacted RNA splicing in purified HSCs after competitive transplant using our Mx1-cre-inducible Dnmt3a knockout mouse model. Using intron retention as a measure of splicing efficacy, we performed deep RNA sequencing of poly-A-enriched mRNA from HSCs (LSK CD150+ CD48-). When comparing exon-intron junctions in Dnmt3a KO versus WT HSCs, Dnmt3a KO HSCs exhibited a greater degree of intron retention. Aberrant alternative splicing was observed in genes involved in HSC differentiation and self-renewal, splicing, ribosomal biogenesis, and cell cycle regulation, suggesting that its role in RNA splicing may be an important mechanism through which DNMT3A influences hematopoiesis. In addition, we delineate a key relationship between DNMT3A, intron retention, and HSC differentiation. Inducing differentiation of WT and Dnmt3a KO HSCs with 5-Fluorouracil (5-FU) revealed key differences in alternative splicing. As WT HSCs differentiate, intron retention decreases; however, Dnmt3a KO HSCs are unable to decrease intron retention in key hematopoietic differentiation genes, leading to a block in differentiation. This was also demonstrated after inducing differentiation in WT and Dnmt3a KO mouse ESCs with Retinoic Acid (RA).

To determine if loss of DNMT3A also impacts splicing in human HSPCs, we used CRISPR/Cas9 to knock out DNMT3A in human HSPCs and transplanted CD34+ cells into NSG mice. RNA sequencing of CD34+ cells 3 months after transplantation revealed aberrant alternative splicing, specifically, greater intron retention in DNMT3A KO than control cells. Together, these data suggest that DNMT3A plays a crucial role in maintaining proper splicing in HSCs.

In summary, this study reveals a novel role for DNMT3A in hematopoiesis. We show that DNMT3A interacts with splicing factors and impacts global alternative splicing patterns in HSCs. In addition, we demonstrate that decreasing intron retention has a specific role in HSC differentiation. The aberrant splicing patterns seen in Dnmt3a KO HSCs may be instrumental in decrypting the crosstalk between DNA methylation and alternative splicing, and their role in HSC differentiation.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.