Background: The BCL-2 inhibitor Venetoclax (VEN) with hypomethylating agents (HMA) has become the standard treatment for unfit AML patients. Recently, we reported that therapy outcome depends on leukemic stem cells (LSCs). By developing the “Mediators-of-Apoptosis-Combinatorial Score” (MAC-Score) linking the ratio of protein expression of BCL-2, BCL-xL, and MCL-1 in LSCs, we demonstrated that BCL-2 apoptotic dependency in LSCs is currently the most accurate therapy response predictor. Aims: In this study we aimed to identify the molecular mechanisms of VEN resistance and apoptotic dependency of LSCs. Methods: Our study cohort comprises samples from 87 AML patients who underwent HMA/VEN therapy including 19 diagnostic/relapse pairs. We aimed to decipher causes of VEN response in FACS-sorted GPR56+ LSCs using transcriptome and methylome analysis as well as functional xenotransplantation and apoptotic dependency assays. Results: To account for heterogeneous patterns of therapy resistance we mapped the gene expression profiles of LSCs onto a single-cell reference map of healthy hematopoiesis. 54 % of LSCs resembled lymphoid-myeloid progenitors (LMPP), 34 % megakaryocytic progenitors (MkP) and 12 % of LSCs showed transcriptomes most similar to hematopoietic stem cells (HSC). Importantly, VEN response and event-free survival (EFS) was associated with LMPP LSCs while VEN refractory patients predominantly resembled other LSC subtypes. The favorable EFS of LMPP LSCs correlated with higher BCL-2 and lower MCL1/BCL-xL expression suggesting a higher BCL-2 dependency for survival compared to other LSC subtypes. Our data further suggest that the LSC subtype is determined by cytogenetic/mutational drivers. AMLs harboring VEN therapy favorable mutations like IDH1/2, splicing factor or RUNX1 mutations were associated to LSCs with an LMPP subtype, while TP53, RAS and JAK2/CALR preferentially result in LSCs with MkP or HSC phenotypes. Further evidence for the LSC-subtype as key determinant of VEN response was found in the differentially expressed genes between responders and non-responders. Non-responders showed higher expression of cell cycle regulators also highly expressed in healthy MkP. We functionally validated these data by performing NSG xenotransplantation assays using AML cells from >40 patients. These results show that VEN refractory patients rapidly induce AML in NSG mice while responders largely fail to engraft even after 8 months, connecting LSC engraftment potential with VEN resistance. Furthermore, LSC potency was significantly associated with reduced EFS and differentiation stage of LSCs, cytogenetic/mutational drivers and apoptotic dependency. In line with these results, paired diagnosis – relapse samples showed that LSC activity is acquired at relapse, while simultaneously BCL-2 dependency decreases. Last, we identified a rare functional LSC population displaying markers of monocytic differentiation (Mono-LSCs) with predominantly KMT2A-rearrangements. These Mono-LSC were distinct from our previously described LSCs as well as from monocytic AML cells, and displayed a unique transcriptomic signature with both monocytic and stem cell features making them poor clinical responders to VEN. Summary/Conclusion: In summary, our data show that LSCs resemble several mutation driven differentiation stages, which shape apoptotic dependencies, LSC potential and ultimately determine the upfront response to HMA/VEN. At time of relapse both apoptotic dependencies and stemness are reshaped and resemble LSCs from refractory patients.Keywords: Leukemic stem cell, Venetoclax, Acute myeloid leukemia, BCL2
