Reconstruction of Human AML Reveals Stem Cell Origin and Therapeutic Targets for Treatment Resistant CD34^-/Lo MLL-Rearranged Leukemia

Identification of the origins of acute myeloid leukemia (AML) stem cells has been a holy grail for a better understanding of the developmental biology of the disease and the design of effective treatments. Studies using primary AML samples on patient-derived xenograft models identified AML stem cells in multiple different CD34/CD38 cellular fractions, which shared similar immunophenotypes of hematopoietic stem/progenitor cells including hematopoietic stem cells (HSCs), lymphoid-primed multipotent progenitors (LMPPs) and granulocyte-macrophage progenitors (GMPs). However inference of cells-of-origin based on the retrospective approach has major limitations as the reported HSC/LMPP/GMP-like AML stem cells are phenotypically different from their normal counterparts; and AML stem cells identified in late developmental stages do not necessarily maintain the same immunophenotypes of the disease initiating (pre-leukemic) cells, which can retain a relatively normal differentiation potential, and only their descendants acquire additional events becoming AML stem cells. While prospective disease modelling using mouse cells has provided unique insights into the potential origins of AML stem cells, human and mouse cells have different transformation requirements, distinct telomere biology, and a significant degree divergence of transcriptional regulation, chromatin state and gene regulatory networks that can profoundly affect their transformation potential and associated cancer biology. Therefore we reason that deconstruction of AML stem cell hierarchy from primary human samples followed by reconstruction of the corresponding human disease using candidate cell populations will give novel insights into this issue.

Given that AML is a highly heterogeneous disease, our study initially focused on genetically well-defined MLL-rearranged AML frequently found in both infant and adult leukemia. By analyzing primary human AML patient samples, the current study unexpectedly revealed that AML stem cells driven by MLL fusions almost exclusively resided in immunophenotypically mature CD34^-/loCD38⁺ compartments, as demonstrated by in vitro long-term culture initiating cells and in vivo limiting dilution xeno-transplantation assays into immunodeficient NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice. Using highly purified populations of hematopoietic stem/progenitor cells from normal human umbilical cord blood (UCB) to reconstruct of the human disease, we found that only human HSCs and common myeloid progenitors (CMPs), but not developmentally branched LMPPs and GMPs, could be disease initiating cells, transformed by MLL-fusions and induced leukemia in vivo, suggesting a key difference between human and mouse leukemia. As revealed by RNA-sequencing, LMPPs and GMPs failed to activate stem cell transcriptional programs and genes essential for MLL-leukemia including MEIS1 and HMGA2 in the early phase of transformation. MLL-fusions transformed HSCs and CMPs were immunophenotypically indistinguishable to leukemic cells of human MLL-AML patients, and also had an enrichment of AML stem cells in CD34^-/loCD38⁺ compartment. Using machine learning, a specific gene signature could stratify patients into HSCs- and CMPs-derived AML, in which HSCs-derived AML showed a significantly poorer prognosis. To further investigate the biology of HSCs/CMPs-derived MLL-AML related to treatment responses, we subjected HSC/CMP MLL-AML leukemia with chemotherapeutic drugs and inhibitor of Bromodomain and Extra-Terminal motif (BET) currently used in AML trial. Interestingly, CMP-derived MLL-AML was treatment sensitive and PDX models transplanted with CMP-like AML samples could be largely cured by chemotherapy or BET inhibitor targeted therapy. In contrast, human HSCs-derived AML was highly resistant to the treatments. Strikingly, shRNA-mediated knockdown or pharmacological inhibition by fidaxomicin targeting ATP-binding cassette (ABC) transporters, ABCC3 that is highly expressed in HSCs-derived MLL-AML could re-sensitize the cells to the current chemotherapy. Together, the current study not only for the first time functionally identifies the origin of human MLL-AML stem cells, but also provides a new actionable venue for overcoming stem cells-associated treatment resistance by repositioning an anti-diarrhea drug, fidaxomicin currently available in the clinics.