AML-Fated Clones Arise in Stem and Progenitor Cells in Myelofibrosis Patients Several Years Prior to AML Diagnosis

In myeloproliferative neoplasms (MPN), somatic mutations in recurrently mutated myeloid leukemia genes, apart from MPN driver genes, are common and can increase transformation risk to acute myeloid leukemia (AML). However, the temporal acquisition of these mutations, the properties of the cells bearing these mutations, and how each mutation cooperates to promote leukemic transformation remains largely unknown. We therefore examined blood and bone marrow samples collected serially from patients with myelofibrosis (MF) who developed AML to further elucidate the genetic basis of leukemic transformation. Specific aims include: 1) determining the timeline of acquisition of mutations that contribute to AML in chronic phase and 2) defining the cell type that harbours the clone fated to initiate AML.

Ten patients who transformed to AML following a diagnosis of MF (6 PMF, 3 post-PV MF, 1 post-ET MF) were studied. The time interval from the first MPN collection to AML diagnosis ranged between 1.5 to 6.6 years; and in 4 patients additional MPN samples collected between these time points were available for study. Somatic mutations were identified from whole genome sequencing (WGS) data of AML blasts (CD45^dim cells), the MPN clone (CD45^highCD33⁺ and CD34⁺ cells), and germline references (T-cells or Buccal DNA); and from targeted sequencing results of 54 recurrently mutated myeloid leukemia genes performed on AML samples as part of another study. Stem and progenitor populations (HSC, MPP, LMPP, CMP, MEP, GMP), as well as, mature cell populations (myeloid, erythroid, T-lymphoid, B-lymphoid and NK) were sorted from MPN time points and whole genome amplified using the REPLI-g Single Cell Kit (Qiagen). Somatic mutations were tracked by droplet digital PCR (ddPCR) and/or targeted sequencing.

We identified an average of 5 (range 1-8) recurrently mutated myeloid leukemia genes that were somatically mutated in each patient at AML diagnosis, with an average of 5.5 (range 1-11) different mutations affecting these genes. Many of these mutations were present at ≥5% variant allele frequency (VAF) in the MPN clone at the earliest chronic phase time point. In 8 patients, the MPN driver mutation was detected in the MPN clone and AML blasts (MPN^driver-concordant), while in 2 other patients there was discordance as JAK2^V617F was present in the MPN clone, but not the AML blasts. Myeloid leukemia genes that were mutated in 2 or more patients included SRSF2 (n=5), ASXL1 (n=4), TET2 (n=4), IDH1/2 (n=4), RUNX1 (n=4), NRAS (n=4), KRAS (n=2), U2AF1 (n=2), PHF6 (n=2), and STAG2 (n=2). We tracked mutations by ddPCR in populations sorted from MPN time points and demonstrated that 10 out of 13 mutations not identified by WGS were in fact present at low VAF in one or more cell populations. Moreover, in 5 patients with MPN^driver-concordant AML, the AML-fated clone was detectable 1.5 to 3 years before AML was diagnosed; and in 4 of these patients the AML-fated clone resided among cells with a HSC or MPP immunophenotype. In order to better establish the reservoir harbouring AML-fated clones, clonal tracking within the stem and progenitor populations by targeted sequencing of a greater number of AML-specific variants is underway. Among the two patients with JAK2^V617F-negative AML, we observed in one patient the emergence of the AML-fated clone within one year of AML diagnosis.

In summary, we show that clones fated to intiate AML in patients with MF often reside among cells with a HSC or MPP immunophenotype and can appear several years prior to the diagnosis of AML. This strongly suggests that mechanisms beyond the acquisition of somatic driver mutations (including but not limited to epigenetic alterations, acquisition of non-coding mutations, inflammation) are necessary to effectively promote full leukemic transformation. Interestingly, in one patient with JAK2^V617F-negative AML (following JAK2^V617F-positive MF), the latency period between the appearance of the AML-fated clone and AML diagnosis was much shorter, suggesting that the mechanisms underlying transformation are likely different for these 2 clonal evolutionary pathways.