Pluripotent Reprogramming of Human AML Resets Leukemic Behavior and Models Therapeutic Targeting of Subclones

Understanding the contribution of abnormal genetic and epigenetic programs to acute myeloid leukemia (AML) is necessary for the integrated design of genetic and epigenetic targeted therapies. While epigenetic therapies have been used to treat AML, so far efficacy has been limited. To this point, the reversibility of epigenetic modifications in AML is poorly understood, as are the relative contributions of the leukemic genetic and epigenetic programs to disease pathogenesis. To investigate these questions, we sought to reprogram primary AML leukemic blasts into induced pluripotent stem cells (iPSCs) and assess the effect of epigenetic reprogramming on leukemic behavior.

We generated iPSCs from several AML patients with 11q23/MLL translocations using classic iPSC reprogramming factors (Sox-2, Klf4, Oct4, and Myc). Reprogrammed AML cells generated iPSCs as indicated by classic pluripotent features including embryonic stem cell morphology, expression of pluripotent markers, and formation of teratomas in vivo. In addition, AML iPSCs retained all the karyotypic and genetic abnormalities of the original AML patient. Despite presence of these mutations, AML iPSCs were able to differentiate into multiple cell types with normal function. Thus, human AML cells can be epigenetically reprogrammed into a pluripotent state despite the presence of leukemic driver mutations.

Surprisingly, when AML iPSCs were differentiated in CD43+45+ hematopoietic cells, a leukemia phenotype re-emerged. Differentiated hematopoietic cells from AML iPSCs exhibited exclusively granulocytic-monocytic differentiation in colony forming assays, and serial replating potential in contrast to control iPSCs. When transplanted into immunodeficient mice, hematopoietic cells from AML iPSCs formed aggressive myeloid leukemias as evidenced by peripheral blood and bone marrow CD33+ myeloid engraftment, enlarged spleens, secondary transplant potential, and death from fulminant leukemia. These data indicate that epigenetic reprogramming alone was insufficient to eliminate leukemic behavior.

To further investigate these epigenetic changes during reprogramming, we performed DNA methylation and gene expression analysis using 450K BeadChip arrays and RNAseq, respectively. Unsupervised hierarchical clustering demonstrated clustering of undifferentiated AML iPSCs with control iPSCs. In contrast, hematopoietic cells differentiated from AML iPSCs clustered exclusively with primary AML cells. Through gene ontology enrichment analysis, hematopoietic differentiated AML iPSCs were enriched for hypomethylation and gene activation of hematopoietic and leukemogenesis gene sets, including MLL gene targets, as compared to undifferentiated AML iPSCs. These data demonstrate that DNA methylation and gene expression profiles are reset upon epigenetic reprogramming of AML cells, but re-emerge upon hematopoietic differentiation, coinciding with re-emergence of the leukemic phenotype.

Lastly, we demonstrated that our AML iPSC model can be used to physically separate and functionally profile genetic subclones within an AML patient. In one patient, we identified distinct AML iPSC genetic subclones (KRAS wildtype and G13D mutant) that could be prospectively separated and demonstrated differential growth properties and therapeutic susceptibilities. To determine the clinical relevance of these subclones, we sequenced this patient at relapse. Strikingly, the KRAS mutant clone, which was dominant at diagnosis, was absent at disease relapse implying relative chemotherapy resistance and competitive outgrowth of the KRAS wildtype subclone. Indeed, in our AML iPSC model, the KRAS wildtype clone demonstrated increased resistance to cytarabine in colony assays. We then investigated in a cohort of AML patients whether cytarabine sensitivity differed between KRAS wildtype and mutant patients. Strikingly, KRAS AML patients were more resistant to cytarabine in vitro compared to KRAS G13D patients. Thus, our AML iPSC model predicted clonal relapse in this patient and identified differential subclonal sensitivity to chemotherapy as a mechanism for relapse.

In summary, pluripotent reprogramming of AML resets leukemic methylation/gene expression patterns, facilitates clonal targeting, and predicts subclonal relapse.