DNMT3a Mutations Define a Pre-Leukemic Stem Cell Reservoir In Human Acute Myeloid Leukemia

Leukemia in humans arises from the multistep accumulation of mutations. However, the identity of the cell of origin, the nature of the first genetic lesion and the order of subsequent mutations remain poorly understood, as most cases of de novo acute myeloid leukemia (AML) are diagnosed without prior observation of a pre-leukemic clonal expansion. As part of studies to examine intra-tumoral genetic heterogeneity in AML, we carried out deep targeted sequencing (read depth 250×) of 101 commonly mutated leukemia genes on samples from 12 patients at diagnosis. Normal T-cells from each sample were expanded in vitro to provide a non-leukemic hematopoietic tissue for comparison. In 3 of 4 patients, we unexpectedly identified DNMT3a mutation not only in AML cells but also in T-cells at a low allele frequency (1-20%). Other genetic alterations such as NPM1 mutation (mutNPM1) were found only in AML cells and not in T-cells, ruling out contaminating AML cells as the source of the DNMT3a signal in cultured T-cells. To investigate the prevalence of T cell involvement, an additional 71 samples from AML patients at diagnosis were screened by Sanger sequencing for DNMT3a mutations. 17 of 71 AML samples (24%) carried R882 codon mutations (mutDNMT3a), and 15 of 17 (88%) also carried mutNPM1. Mutant allele frequency in freshly isolated T-cells was measured by droplet digital PCR (ddPCR). mutDNMT3a with no evidence of mutNPM1 was detected in T-cells of 12 of 17 patients (70.5%), suggesting that DNMT3a mutation occurs before NPM1 mutation in an ancestral stem/progenitor cell that gives rise to both T-cells and the dominant AML clone present at diagnosis. To directly determine whether phenotypic stem/progenitor cells that carried the mutDNMT3a allele could be identified within the non-leukemic hematopoietic compartment of AML blood and bone marrow samples, we undertook genetic analysis of highly-resolved phenotypically-defined normal stem, progenitor and mature lymphoid cell fractions from 10 patient samples. mutDNMT3a without detectable mutNPM1was present in stem cells and all downstream progenitors, with mean allele frequency among multipotent progenitor (MPP), multilymphoid progenitor (MLP) and common myeloid progenitor (CMP) of 31.7%. mutDNMT3a and mutNPM1 were found together only in granulocyte monocyte progenitor (GMP) and CD33+ blasts. Importantly, even for patients in whom mutDNMT3a was not detected in mature lymphoid populations, mutDNMT3a without mutNPM1 was found in MPP, MLP, CMP, providing strong evidence that mutDNMT3a precedes mutNPM1 during leukemogenesis. Analysis of diagnostic and remission samples revealed similar or higher proportion of cells with mutDNMT3a alone at remission compared to diagnosis. Xenotransplantation of cells from the diagnostic samples of 2 patients with mutDNMT3a and mutNPM1generated predominantly non-leukemic multilineage grafts (18 of 19 mice) with predominance of cells bearing mutDNMT3a without mutNPM1 (mean allele frequency 57%), confirming that mutDNMT3a was present in HSC. Kinetic analysis at 8 and 16 weeks revealed increasing mutDNMT3a allele frequency in multilineage xenografts over time, suggesting that mutDNMT3a confers a competitive growth advantage over non-mutated HSC. Collectively, our results are consistent with the clonal expansion in AML patients of mutDNMT3a HSC that survive chemotherapy. These cells may therefore represent a reservoir for further genomic progression leading to relapse. Our findings now offer the possibility of therapeutic intervention during remission to eliminate these surviving pre-leukemic clones and prevent relapse in a large proportion of AML patients carrying mutDNMT3a. As well, our work provides a framework for the identification of other early events in leukemogenesis and examination of how these changes disrupt normal HSC function and lead to leukemia.