Abstract P38: Expanding the repertoire of physiologically relevant AML models to support new target drug discovery, compound screening, and predicting patient responses in clinical trials

Introduction: Traditional cell line models of Acute Myeloid Leukemia (AML) almost exclusively represent differentiated blasts, are homogenous in clonality, and often do not encompass major genetic subtypes. The application of physiologically relevant, patient-derived models (PDX) is essential to capture the complexity of mutational profiles and highly divergent subtypes that are a feature of this disease,. To address these gaps, we established a culture system that supports the long-term growth of higher-order AML samples, created a 16-color immunophenotyping flow panel, and validated the use of these samples using common standard of care or investigational agents to better predict in vivo and patient responses. Methods:Ex vivo culture conditions were optimized using physiologically relevant stroma-free cytokine-rich media that supports primary AML cells with minimal phenotypic shift, allowing for high throughput functional drug sensitivity testing. We developed a 16-color high-dimensional immune profiling panel that captures AML heterogeneity by delineating primitive and mature AML blasts along with normal hematopoietic lineages. Results: 70% (10/14) of disseminated AML PDX models tested were successfully expanded ex vivo and continued to proliferate after long term (> 30 days) culture. Our platform included examples of subtypes (NPM1c, DNMT3A, and mut IDH1 & 2) allowing us to expand on the existing repertoire of underrepresented cell lines. Our 16-color flow panel established baseline characteristics of cell heterogeneity and state; critically, cultures maintained cell surface expression over time and transcriptomic drift was minimal. Our compound panel established baseline drug sensitivity across a range of cell death, cell cycle, epigenetic, and other targeted agents. There was the expected concordance and response to FLT3i (Quizartinib) and sensitivity in FLT3-ITD, Menini (Revumenib/Ziftomenib) and sensitivity to both NPM1 and MLL mutants, and cell death (Bcl2 and/or xL) sensitivity in WT TP53 lines. AML PDXs were also tested in vivo with a clinically relevant dose of venetoclax + azacytidine (VEN/AZA) for 2 cycles (14 days) and sensitivity evaluated in VEN/AZA exposed cell culture. It was notable that VEN/AZA treatment enforced an immunophenotypic shift from a stem-like state to a more mature-like cell state (decreases in CD123+, TIM3+, increases in cKIT+, CD14+ and CD34+) consistent with literature and evidence of treatment resistance. Of note, VEN/AZA treatment coincided with the onset of resistance of other cell death agents including selective MCL1 (AZD5911) and dual Bcl2/xL inhibitor (AZD0466). Conclusions: Our established platform provides novel physiologically relevant models to explore unique target biology, better understand compound mechanism of action, and ability to generate and deeply characterize drug resistant models. Applying single and combination compound screening to these models will allow more accurate prioritization of assets moving toward the clinic.

Citation Format: Omid Tavana. Expanding the repertoire of physiologically relevant AML models to support new target drug discovery, compound screening, and predicting patient responses in clinical trials [abstract]. In: Proceedings of the Blood Cancer Discovery Symposium; 2024 Mar 4-6; Boston, MA. Philadelphia (PA): AACR; Blood Cancer Discov 2024;5(2_Suppl):Abstract nr P38.