Leveraging single-cell multiomics for a generalizable approach to differentiation therapy

Among the different types of systemic therapies used in cancer treatment, differentiation therapy, perhaps the least frequently utilized but typically most tolerable form of treatment, aims to set cancer cells back on the developmental path from which they strayed, leading to cell death through multiple mechanisms. Acute promyelocytic leukemia (APL) is the exemplar for differentiation therapy wherein leukemic blasts arrested in a progenitor-like state are induced to continue their myeloid differentiation after treatment with all-trans retinoic acid (ATRA) and arsenic trioxide, resulting in >95% cure rates. In contrast to APL, other types of acute myeloid leukemia (AML) have only a 5-year survival of ~33%. However, recent advances in targeted therapies with novel, differentiation-inducing agents for AML with alterations in IDH1/2 (isocitrate dehydrogenase 1 and 2) , NPM1 (nucleophosmin 1) , and KMT2A ( lysine methyltransferase 2A ) have demonstrated efficacy in patients who are older, frail, or who have relapsed or refractory disease. However, the underlying mechanism of differentiation induction remains incompletely understood, limiting adoption for other AML subtypes or non-hematologic diseases. Establishing a rational means of discovering regulators of differentiation therapy could accelerate development of this class of drugs. Here, we present a platform for discovering the central regulators of differentiation as a means of identifying targets for downstream therapeutic development. Using single-cell multiomic (RNA- and ATAC-seq - scMultiome) data from clinical samples of IDH1/2 -mutated (IDHm) AML as a test case, we utilized a software package called CellOracle to dissect the gene-regulatory networks (GRNs) underpinning the differentiation response throughout the course of IDH inhibition. With these analyses, we recapitulated a known pattern of “BRCAness” in IDHm AML with in silico knockout of BRCA1 (breast cancer gene 1) appearing to support the differentiation effect of IDH inhibition . We also discovered putative as-yet unidentified regulators of the differentiation phenotype including several Ikaros transcription factors. With this new platform, we also sought to identify targets for differentiation therapy in a solid tumor type, wherein examples of differentiation therapy are almost entirely lacking, with the noteworthy exception of ATRA for neuroblastoma. We utilized scMultiome data from H3K27M diffuse midline gliomas of the pons, which are known to have an epigenetically-based inhibition of terminal differentiation. These aggressive and devastating brain cancers affect exclusively school-aged children and have an approximately 95% 2-year mortality. Using the same pipeline we have identified several putative regulators of a malignant-to-normal transition, including several genes known to be important to normal glial development and maintenance of the cancer state.