DNMT3A-Dependent DNA Methylation May Act As a Tumor Suppressor-Not a Tumor Promoter-during AML Progression

Altered DNA methylation is a well-known feature of acute myeloid leukemia (AML) genomes, but the mechanisms underlying these changes and their relevance for AML pathogenesis are unclear. We previously showed that DNMT3A is the predominant de novo methyltransferase expressed in AML cells, and that the DNMT3A^R882H mutation in AML creates a dominant negative protein that reduces in vitro DNA methylation activity by ~80%. Since DNMT3A provides themajority of the methylation activity in AML cells, we hypothesized that AML samples with and without DNMT3A^R882H could reveal novel insights about the role of this enzyme in AML initiation and progression.

We performed whole-genome bisulfite sequencing (WGBS) of 38 primary human AML samples and 17 normal human hematopoietic cell samples, as well as a remission sample from a patient with a persistent DNMT3A^R882H mutation, and blood samples from a non-leukemic patient with a constitutional DNMT3A^R882H mutation. We first identified 3,848 differentially methylated regions ('DMRs') between DNMT3A^R882H and DNMT3A^WT AMLs, virtually all of which were hypomethylated in the DNMT3A^R882H AMLs. Further, 28% (1,087/3,848) of these DMRs were also hypomethylated when compared to CD34 cells, implying that these regions are truly hypomethylated in the AML cells with the R882H mutation. In contrast, 72% (2,759/3,848) of the DMRs were unmethylated in bothDNMT3A^R882H AMLs and CD34 cells, but were hypermethylated in the DNMT3A^WT AML samples. These loci were associated with CpG dense regions, suggesting that they represent abnormal CpG island hypermethylation that occurs only in AML samples with wild-type DNMT3A. Analysis of 21 additional primary AML samples with wild-type DNMT3A identified 4,912 hypermethylated regions compared to CD34 cells, of which 4,544 (92%) were significantly less methylated in DNMT3A^R882H AMLs, implying that functional DNMT3A mediates abnormal CpG island hypermethylation in AML.

WGBS analysis of two non-leukemic hematopoietic samples with DNMT3A^R882H mutations was also performed to understand the direct effects of DNMT3A^R882H in non-transformed myeloid cells. These samples included peripheral blood (PB) neutrophils and monocytes from a newly identified 9-year old patient with an overgrowth syndrome and developmental delay (Tatton-Brown et. al., Nature Genetics 2014), who was found to have a heterozygous DNMT3A^R882H mutation in all skin and peripheral blood cells. His CBC was normal, and he had no evidence of clonal hematopoiesis by exome sequencing. We identified 2,051 DMRs in his PB myeloid cells, all of which were hypomethylated compared to control PB myeloid cells from his healthy 13-year old brother (and also normal CD34 cells), demonstrating that DNMT3A^R882H directly causes focal methylation loss. We also performed WGBS on cells expanded from single stem/progenitor cells from an AML patient with a persistent DNMT3A^R882H mutation during remission. Expanded cells with DNMT3A^R882H were hypomethylated relative to wild-type DNMT3A cells expanded from the same sample. The majority of the hypomethylated regions were also present in the patient's AML cells, implying that DNMT3A^R882H-associated hypomethylation in pre-leukemic cells is maintained during AML progression.

These findings demonstrate that DNMT3A^R882H-associated hypomethylation precedes leukemia development, and may therefore represent an important initiating phenotype for AML. Our data also suggest that the abnormal hypermethylation of CpG islands in AML cells is DNMT3A-dependent, and must occur during disease progression. This hypermethylation is absent in AMLs with DNMT3A^R882H, revealing that it is not required for leukemia progression. We therefore propose a model where DNMT3A-dependent DNA methylation in AML cells acts as a 'brake' that prevents abnormal self-renewal; the abnormal CpG island hypermethylation in DNMT3A^WT AMLs may be an adaptive response that is ultimately overcome during leukemia progression. The absence of this 'braking' activity in AMLs with DNMT3A^R882H may contribute directly to leukemia initiation. The restoration of DNMT3A activity in AML cells with the DNMT3A^R882H mutation is therefore a therapeutic goal.