Establishing an AML Ex Vivo Platform to Generate Physiologically Relevant Models

Acute myelogenous leukemia (AML) comprises a highly polyclonal and heterogeneous disease containing leukemic stem cells, progenitors, and differentiated blasts. Despite advances in treatment, the long-term relapse-free survival for AML remains poor. Due to complex mutational profiles and highly divergent subtypes in AML, the use of physiologically relevant, patient-derived models is needed. Traditional cell line models of AML almost exclusively represent differentiated blasts, are homogenous in clonality, and often do not encompass major genetic subtypes. To address these gaps, we have established a culture system that supports the growth of higher-order AML samples ex vivo.

By optimizing culture conditions, we successfully cultivated 70% (10/14) of AML disseminated PDX models ex vivo and demonstrated they continue to proliferate after long term cultures. We established a 16-color flow panel to characterize their cell maturity state and show they maintain cell surface heterogeneity. Importantly, these models have not ‘drifted’ transcriptomically after multiple passages. Our platform includes 2 NPM1c mutant and 3 IDH1/2 mutant ex vivo lines, in addition to other subtypes underrepresented in traditional culture methods.

Furthermore, we treated AML PDXs in vivo with a clinically-relevant dose of venetoclax + azacytidine (Ven/AZA) for 2 cycles (14 days) and then evaluated post-VEN/AZA cells in our ex vivo assay. Growing these models ex vivo, we show using our flow panel that Ven/AZA treatment shifts the differentiation status from a stem-like state to a more mature-like cell state which is consistent with literature evidence of treatment resistance. Additionally, a second PDX model treated with three cycles of Ven/AZA in vivo and harvested in ex vivo culture conditions maintained both Ven and AZA resistance. Interestingly, compared with the vehicle treatment, Ven/AZA treatment caused resistance to other cell death agents including selective MCL1 and dual Bcl2/xL inhibitors. Lastly, we performed a 7-day cell compound panel screen using common standard of care agents for the treatment of AML. Many of the targeted agent activity correlated with expected genetic subtypes, such as FLT3 inhibitors in FLT3-ITD mutants and venetoclax in WT TP53 models. This established ex vivo platform provides physiologically relevant models to explore novel target biology, better understand compound mechanism of action (MoA), and generate drug resistant models through in vivo driven clinically relevant dose scheduling using standard of care agents.