Cheminformatic identification of small molecules targeting acute myeloid leukemia

Abstract Acute myeloid leukemia (AML) is an aggressive hematological malignancy with poor prognosis and high relapse rates when treated with cytotoxic chemotherapeutics. Previously, we identified a family of small molecules that modulate mitochondrial function, referred to as PS127-family compounds. These drugs were selectively toxic to AML and were characterized by two predicted functions: apoptotic agonism and thioredoxin/glutathione reductase inhibition. Here, we uncovered a third critical predicted function, autophagic induction. Using a cheminformatic screen of ∼4.2 million compounds for molecules with high predicted probability for all three functions, we found and validated hits that selectively killed AML cells, activated apoptosis, were dependent upon autophagic activation, and compromised glutathione metabolism by interfering with glutathione reductase, all of which are consistent with predictions. Compound treatment increased pools of cytosolic and mitochondrial ROS, decreased oxygen consumption, and reduced ATP synthesis. Structurally-unrelated compounds caused the same phenotypes, validating our approach of screening for predicted function. Finally, we also observed strong synergy between these compounds and midostaurin and venetoclax, underscoring their therapeutic potential. Key phenotypes, including the compounds’ impact on glutathione metabolism and synergy with doxorubicin and midostaurin, were confirmed in AML-patient-derived primary cells, validating the potential of these compounds for the development of future AML treatments.