CRISPR-based genetic screens to interrogate therapeutic targets on tumor maintenance and heterogeneity in AML patient cells

Functional genomics directly in primary tumor cells is warranted to better understand the complexity of the diseases in a context of high clinical relevance. We optimized ex vivo and in vivo CRISPR-based genome/epigenome editing approaches in patient-derived xenograft (PDX) and primary acute myeloid leukemia (AML) cells, including those harboring DNMT3A/FIT3-ITD/NPM1c that predicts a poor clinical outcome and has no model to fully represent. We deployed CRISPR and CRISPRi “drop-out” screenings on patient cells in vivo and prioritized AML-biased gene and cis-element dependencies, including an uncharacterized MYB regulatory element. We further developed a Perturb-seq pipeline for patient cells, where we perturbed known AML therapeutic targets and delineated complex regulatory networks and changes of cell cycle stages and cellular hierarchy in response to perturbations by scRNA-seq. We uncovered an unanticipated role for SETDB1, a suppressor of interferon response, whose depletion caused the emergence of Hematopoietic stem cell (HSC) and progenitor-like population marked with high HOXA expression in a PDX model, and myeloid differentiation were re-induced by DOT1L inhibition, preceding a combinatorial therapy to co-suppress SETDB1 and the HOXA cluster. Our study proposed a strategy to annotate personalized vulnerabilities to predict therapeutic response in AML.