MTHFD2 is a selective metabolic vulnerability in AML that supports mitochondrial redox homeostasis and venetoclax resistance

Abstract Dysregulation of mitochondrial biology is a common occurrence in acute myeloid leukemia (AML) and is often associated with chemotherapy resistance. However, the molecular components that support mitochondrial dysfunction in AML remain largely undetermined. The mitochondrial enzyme, MTHFD2, is a key component of the one-carbon (1C) metabolic network, which supports numerous anabolic processes, including de novo purine synthesis. Recent evidence suggests that MTHFD2 may support AML; however, the molecular role of MTHFD2 in AML biology and chemotherapy resistance remains largely unknown. To explore the role of MTHFD2 in AML, we employed the CRISPR-Cas9 gene editing system and found that, compared to non-targeting controls, MTHFD2 deletion significantly impaired the expansion of several human AML cell lines, including MOLM14, NOMO1, OCIAML2, and OCIAML3. Human AML cells lacking MTHFD2 displayed increased Annexin V staining and reduced BrdU incorporation, indicating that these cells rely on MTHFD2 to support cell proliferation and survival. We also generated a mouse model to study how deletion of Mthfd2 impacts AML driven by the MLL-AF9 leukemogenic allele in vivo. Using this model, we found that deletion of Mthfd2 led to a significant delay in the onset of disease in this model (p = 0.0023). Importantly, deletion of Mthfd2 did not significantly alter the distribution of various mature or hematopoietic stem and progenitor cells (HSPCs) in healthy mice, nor did it significantly impact the ability of healthy HSPCs to reconstitute hematopoiesis in a competitive transplant model. The central role of MTHFD2 in 1C-metabolism is to convert 5,10-dimethylene (5,10me)-tetrahydrofolate (THF) into 10-formyl-THF, which is subsequently used to synthesize purines. To assess the impact of MTHFD2 deletion on 1C-metabolism, we carried out isotopologue tracing assays using a non-radioactive isotopic form of serine carrying three deuterium atoms (2,3,3-2H-serine). Consistent with previous reports in solid cancers, we found that while MTHFD2 deletion reduced the contribution of mitochondria 1C-metabolism to purine and deoxy-thymidine triphosphate (dTTP) synthesis, this could be compensated for by the cytosolic arm of 1C-metabolism. During this generation of 10-formyl-THF, MTHFD2 also reduces NADP/NAD+ into NADPH/NADH. Notably, MTHFD2 deletion resulted in a significant decrease of both mitochondrial NADPH and NADH, both of which play key roles in regulating mitochondrial redox homeostasis and metabolism. We next performed RNA-seq analysis in both control and MTHFD2-deficient mouse (MLL-AF9) and human (MOLM14) AML cells. Consistent with our metabolic tracing assays, gene set enrichment analysis of these datasets showed that MTHFD2 deletion disrupts de novo purine and pyrimidine synthesis and mitochondrial redox homeostasis, and leads to an induction of inflammatory, apoptotic, and oxidative stress gene signatures. We also found that deletion of MTHFD2 increased mitochondrial superoxide production and the overexpression of Catalase (CAT) and Superoxide dismutase (SOD), both of which promote mitochondrial superoxide detoxification, partially restored cell viability in sgMTHFD2 cells compared to the sgNT expressing cells. Collectively, these results suggest MTHFD2 plays a key role in regulating mitochondrial redox homeostasis. To assess the therapeutic potential of targeting MTHFD2, we employed a small molecule inhibitor of MTHFD2 called DS18561882 (DS). Similar to MTHFD2 deletion, DS disrupted mitochondrial redox homeostasis and decreased mouse and human AML cell growth and survival. Next, we examined the effects of this compound on two healthy donor (HD) samples and 50 genetically-diverse patient-derived (PD) AML samples. While HD cell numbers were unaffected, DS-treatment displayed varying but dose-dependent anti-leukemic effects on PD-AML samples. We also evaluated DS in combination with current FDA-approved AML therapies and found that DS selectively cooperated with Venetoclax (and not cytarabine or anthracyclines) in killing PD-AML cells. Importantly, the combination of DS and Venetoclax often outperformed the combination of Venetoclax plus azacytidine. Lastly, we generated Venetoclax-resistant human AML cell lines and found that DS re-sensitized these cells to Venetoclax treatment. From these data, our study establishes MTHFD2 as a stand-alone target in AML as well as a potential adjunct to Venetoclax-based therapies.