CELLULAR AND MOLECULAR MECHANISMS OF CHEMORESISTANCE IN ACUTE MYELOID LEUKEMIA

Acute Myeloid Leukemia (AML) remains an incurable disease, with a 5-year survival rate below 30%. While the introduction of new therapeutic approaches, such as targeted therapies and immunotherapy, has had a modest impact on AML prognosis, the standard treatment of AML is still based on chemotherapy regimens (like "7+3") that have been in use for over 50 years. Approximately 20% of patients do not respond to standard chemotherapy, and even among those who initially respond, around 50% eventually relapse due to the development of chemotherapy resistance. Chemoresistance is the cause of death for over 90% of AML patients. The mechanisms underlying chemoresistance in AML are not yet fully understood, and targeted therapies aimed at preventing chemoresistance are not currently available. Chemoresistance in AML is primarily driven by non-genetic mechanisms. While leukemic stem cells (LSCs) have been implicated, recent studies suggest that AML blasts that survive chemotherapy, known as Chemotherapy Persistent Blasts, enter a senescence-like and diapause state. Though they can contribute to disease recurrence, these Persistent Blasts are not inherently therapy-resistant as they revert to a drug-sensitive state upon drug withdrawal. However, a lack of robust in vivo models of clinical chemoresistance in human AML hinders our understanding of these mechanisms. To address this, I aimed to generate in vivo models of truly chemoresistant AML to investigate non-genetic mechanisms of chemotherapy resistance. I utilized mouse models of human AML, specifically Patient-Derived Xenografts (PDXs), to establish treatment protocols mimicking the standard "7+3" chemotherapy regimen used clinically. One model (AML9) mirrored the behavior of most patients treated with standard chemotherapy, exhibiting clinical remission followed by relapse. In contrast, the other model (AML20) showed a modest response to chemotherapy with rapid leukemia re-expansion. Notably, relapsed leukemias in both models were completely resistant to retreatment with the same chemotherapy regimen, and this resistant phenotype persisted even after re-transplantation. Extensive genomic profiling revealed that chemoresistance in relapsed leukemias was not associated with the emergence of de novo DNA mutations. AML9 and AML20 represent the first two available AML models of true clinical chemoresistance, mirroring primary and secondary resistance, respectively. 9 To investigate mechanisms of chemoresistance in these models, I characterized the cellular and molecular properties of Chemotherapy Persistent Blasts isolated at the end of the chemotherapy regimen. To monitor and isolate quiescent blasts during leukemia growth and after chemotherapy treatment, I incorporated a label-retaining (LR) assay into the PDX models, enabling cell division tracking through inducible expression of H2B-GFP. I found that: i) in vitro, Persistent Blasts from both models exhibited significantly higher IC50 values compared to non-treated blasts, suggesting the induction or selection of a stable and cell- autonomous chemoresistant phenotype; ii) Quiescent blasts were slightly enriched among Persistent Blasts, but a portion of leukemia cells continued to proliferate during treatment, generating a population of Persistent Blasts of both proliferating and quiescent cells; iii) Persistent Blasts retained regenerative potential in vivo, primarily attributed to persistent quiescent blasts, suggesting that quiescence, while not a specific trait of chemoresistance, is integral to the relapsing properties of these cells; iv) the AML clonal structure remained unchanged after chemotherapy treatment, as determined by lineage tracing of Persistent Blasts and relapsed leukemias, suggesting that chemotherapy does not select specific cellular lineages; v) Persistent Blasts exhibited distinct transcriptional features compared to untreated leukemias, characterized by stem-cell signatures and activation of the Interferon (IFN) signaling pathway. Notably, IFN signaling activation was a distinguishing feature of both quiescent and proliferating blasts and was predictive of clinical response to standard chemotherapy in AML patients. To investigate the contribution of the activated IFN signaling pathway to chemoresistance, I examined the effect of silencing IFI6, the top upregulated gene in the IFN signature. IFI6 silencing prolonged latency and reduced engraftment frequency in AMLs, due to its detrimental effect on leukemia-initiating cells (LICs). However, IFI6 silencing did not impact leukemia outgrowth kinetics. Strikingly, IFI6 silencing completely reversed the chemoresistant phenotype in both AML models in vivo and increased the chemosensitivity of established AML cell lines in vitro. To further explore mechanism of IFI6-mediated chemo-sensitization, I investigated the interaction between IFI6 and STING. I found that IFI6 is upregulated upon chemotherapy treatment and relocates to the ERGIC compartment in proximity to STING, where STING activation by chemotherapy-induced DNA damage initiates a cascade leading to IFN signaling and cell death. Based on these findings, I initiated experiments to determine 10 whether IFI6-mediated chemoresistance is due to its ability to downregulate STING activation, thereby preventing the execution of IFN-induced cell death.