Tight Regulation of H3K79 Methylation Levels in KMT2A-Rearranged AML

Rationale: KMT2A-rearrangements (KMT2A-r) in acute myeloid leukemia (AML), encompassing both KMT2a-fusions (KMT2A-F) and KMT2A-partial tandem duplications (KMT2A-PTD), represent a subgroup of AML with a particularly poor prognosis. Both KMT2A-F and KMT2A-PTD share a dependency on the H3K79 methyltransferase DOT1L for proper histone 3 lysine 79 dimethylation (H3K79me2) on target genes such as HOXA9 and MEIS1. Pharmacologic inhibition of DOT1L results in downregulation of KMT2A fusion/PTD target genes. Albeit rare, complete responses observed in patients with relapsed/refractory KMT2A-r leukemia treated with a DOT1L inhibitor underscore the clinical relevance of this pathway. The second most commonly co-occurring mutation with KTM2A-PTD are mutations in isocitrate dehydrogenase 1 and 2 (mIDH1/2). The canonical oncogenic function of mIDH1/2 involves aberrant production of the oncometabolite 2HG and inhibition of TET2. However, 2HG also induces histone hypermethylation, including H3K79 hypermethylation as reported by several groups. Based on this, we sought to study the effect of increasing H3K79 methylation in KMT2A-r AML either via direct overexpression of DOT1L or introduction of mIDH1/2.

Results: As increased H3K79 methylation has only been observed upon 2HG exposure or mIDH1/2 expression in experimental systems, we analyzed H3K79 methylation in primary AML patient samples. We found increased H3K79 methylation in IDH1/2 mutant AML at diagnosis. Given the high percentage of IDH1/2 mutations in KMT2A-PTD leukemias, we sought to understand what effect increasing H3K79me2 would have on KMT2A-r leukemia. To this end, we overexpressed DOT1L in KMT2A-MLLT3 murine AML cells. Surprisingly, we found H3K79me2 was counterselected, and DOT1L overexpressing cells grew slower compared to controls. Furthermore, ChIP-Seq revealed that DOT1L overexpression increased H3K79me2 peak height and width outside KMT2A-fusion target genes, effectively blunting the difference between fusion targets and other genomic loci. This was associated with a decreased expression of target genes. This counterintuitive finding led us to analyze global (rather than locus specific) H3K79 methylation in AML patient samples. We found that H3K79me2 is globally low in KMT2A-r AML. This constellation is reminiscent of CALM-AF10 AML, which typically loses the second AF10 allele, resulting in globally low H3K79me2/3, preserved/increased methylation on fusion target loci, and DOT1L dependence. We next tested the effect of IDH mutations on KMT2A-r AML. Co-expression of mIDH1 and MLL-MLLT3 in murine lin^-sca1⁺ckit⁺ bone marrow cells resulted in an aggressive AML that was dependent on DOT1L, but not mIDH1. As we found when DOT1L was overexpressed, H3K79me2 was counterselected in the mouse. This finding was confirmed by introducing KMT2A-MLLT3 into mIDH2 knock-in bone marrow recapitulating the order of mutations in AML patients, as well as by overexpressing mIDH1 in a KMT2A-PTD cell line. In two patient samples with KMT2A-PTDs and IDH1/2 mutations, we found no difference in H3K79 methylation compared to non-IDH1/2 mutant KMT2A-PTD samples.

Summary: While two groups have reported cooperation of IDH1/2 mutations with retroviral overexpression of KMT2A targets (HOXA9/Meis1), we find that there is no cooperation between KMT2A-rearrangements themselves and mIDH1/2. This may in part be due to H3K79 hypermethylation, which is not tolerated in KMT2A-r disease, while HOXA9/Meis1 retroviral models are insensitive to H3K79 modulation. Interestingly, H3K79me2 is globally low in KMT2A-F patient samples. Our findings suggest that KMT2A-F leukemias maintain tight control over H3K79me2, and neither increases nor decreases are tolerated. This may have implications for the clinical/biomarker development of DOT1L directed therapies.