The DNMT3A mutations (DNMT3AMT) that occur in acute myeloid leukemia (AML) and Myelodysplastic syndromes (MDS) are either hypomorphic or inactivating, consistent with bone fide tumor suppression function of DNMT3A . The most recurrent mutation involves the canonical R882, resulting in a hypomorphic protein most likely also acting in a dominant-negative manner by affecting the methyltransferase activity of the remaining wild type DNMT3A allele (DNMT3AWT).

To assess the clinical context of DNMT3A alterations, 1059 samples from patients with myeloid neoplasia (422 MDS, 240 MDS/myeloproliferative neoplasms (MPN), 163 secondary (s) AML, 234 primary (p) AML were examined. DNMT3AMT were found in 21% of patients, of which 71% were missense mutations [52% R882; 48% non-canonical mutation (NCM)] and 29% were truncating mutations or frameshift mutations. DNMT3AR882 was significantly overrepresented in pAML (P=.04) and underrepresented in MDS (P=.03). In contrast, truncating mutations were more common in MDS and sAML vs. pAML (55% vs. 29%; P<.01). DNMT3AMT were associated with ASXL1 (P=.05) and SF3B1 (P<.01) mutations in MDS but not in pAML. In contrast, DNMT3AMT associated with NPM1 (P<.01), FLT3ITD/TKD (P=.04) and IDH1 mutations in pAML but not sAML. Clonal architecture analyses showed that 49% of R882 were ancestral events with TET2, FLT3ITD/TKD and NPM1 as the most common secondary hits. In contrast, only 33% of non-R882 variants (39% truncating; 33% NCM) were an ancestral event (P=.04). Gene expression analyses of AML [n=167 (26% DNMT3AMT)] and MDS [n=175 (8% DNMT3AMT)] showed a decrease in DNMT3A mRNA levels in DNMT3AMTvs. DNMT3AWT AML (mean ± SD= 1079 ± 582 vs. 1376 ± 556), though not in MDS, probably due to the high percentage of truncating mutations and resulting nonsense mediated decay of mRNA.

Exclusively heterozygous DNMT3AMT configurations imply that biallelic DNMT3A loss may be lethal and thus that mutant cells may be amenable to synthetic lethal therapeutic approaches. DNMT3A inhibitors may however, also phenocopy DNMT3A loss in normal DNMT3AWT cells. Although such concerns are not huge, as they would likely be reversed by withdrawing the inhibitor, there is also rationale to instead overcome haploinsufficiency by increasing the throughput of the remaining DNMT3AWT allele, i.e. to revert malignant cells toward normalcy rather than attempt to kill them. We achieve this by supplementation of the universal methyl donor S-adenosyl- L -methionine (SAM), i.e. by increasing DNMT3A substrate availability.

To test this principle we quantified the differential sensitivity of DNMT3AMTvs. DNMT3AWT and primary cells to SAM (disulfate tosylate) supplementation. As in vitro models, we selected a myeloid derived cell line, OCI-AML3, which harbors R882 and shows a 9% decrease in global DNA-methylation. As controls we used a panel of myeloid cell lines with DNMT3AWT (K562, THP1, U937, MOLM13). We observed that OCI-AML3R882 exhibited higher sensitivity to 24h SAM supplementation than DNMT3AWT cell lines using an MTT assay (IC50 of ~2mM). This dose did not affect healthy human bone marrow cells as shown by no change in growth upon SAM treatment.

We then assessed DNMT3A enzymatic activity using an ELISA-like colorimetric assay and found that 2mM SAM exposures increased DNMT3A activity by 20% vs. baseline (P=.04) in OCI-AML3R882 cells, suggesting that SAM supplementation was able to increase DNMT3AWT throughput. DNMT3AWT cells did not show increases in DNMT3A activity with SAM supplementation.

We also generated a cellular model of DNMT3A insufficiency using lentiviral shRNA in K562 (55% reduction by qPCR) (K562-shRNA DNMT3A). SAM treatment (500µM, 1mM, 2mM, 3mM) induced a decrease in cell growth in K562-shRNA DNMT3A vs. not (30% vs. 60%; P<.01).

In sum, we present the first novel DNMT3AMT therapeutic strategy based on the principle of overcoming the functional deficiency of a monoallelic mutants through up-modulation of DNMT3AWT activity. SAM showed a promising increase in DNMT3AWT activity in mutant cells at safe/feasible doses, opening the possibility of combinational strategies with existing/ approved agents. SAM could be beneficial in healthy older adults with clonal hematopoiesis of indeterminate potential and DNMT3AMT. We are in the process of determining if SAM treatment will enhance DNMT3AWT in murine models of DNMT3A (Dnmt3aMT, Flt3ITD, Npm1C) resulting in a fully penetrant leukemic phenotype.


Sekeres: Celgene: Membership on an entity's Board of Directors or advisory committees.

Author notes


Asterisk with author names denotes non-ASH members.