DGAT1 Inhibition Enhances Metformin-Induced Ferroptosis in AML By Targeting Lipid Metabolism

Metabolic rewiring is a well-established hallmark of cancer, crucial for sustaining leukemogenesis. In acute myeloid leukemia (AML) patients, high dependency in oxidative phosphorylation (OXPHOS) is often associated with poor outcomes, and inhibition of mitochondrial respiration has shown significant efficacy in AML therapy. Here, we used metformin, an FDA-approved drug known for impairing mitochondrial respiration, to assess its efficacy in a genetically diverse panel of AML patients. Given the heterogeneous response to metformin, our goal was to identify metabolic profiles that could predict susceptibility to the drug and to further elucidate its mechanism of action in AML. Using label-free quantitative proteome analysis (11272 proteins) combined with metabolomic screening (172 metabolites) on sorted CD34⁺/CD117⁺ AML blasts (n=26), we performed single-sample Gene Set Enrichment Analysis (ssGSEA) focused on metabolic terms and correlated enrichment scores with metformin sensitivity. Increased response to treatment was associated with terms related to lipid metabolism and ferroptosis, with FLT3-ITD and IDH1/2 mutant samples showing the highest enrichment scores. To elucidate the mechanisms underlying our findings, we used the isogenic TF1 cell line harboring an IDH2^R140Q mutation which exhibited disrupted lipid homeostasis and increased sensitivity to OXPHOS inhibition compared to the wild-type (TF1^wt). Extracellular flux analysis confirmed that IDH2^R140Q cells are more reliant on OXPHOS compared to its counterpart, and metformin treatment completely abolished mitochondrial respiration in these cells. While TF1^wt cells were able to rewire their metabolism towards glycolysis upon OXPHOS inhibition, supporting its viability, this was not observed to the same extent in IDH2^R140Q cells. Measurements of glucose consumption and lactate production post-treatment further confirmed a stronger upregulation of glycolysis in TF1^wt cells compared to the mutant setting. Since our in silico analysis showed an association between ferroptosis transcriptional signature and metformin, we assessed lipid peroxidation levels and induction of reactive oxygen species (ROS) after treatment. We confirmed an increase in lipid peroxidation and ROS levels in both cell lines and ex vivo-treated AML samples, together with a strong reduction in viability, an indicative of ferroptosis. Metabolome analysis showed an imbalance in the lipid pool, with a higher relative contribution of poly-unsaturated fatty acids in FLT3-ITD and IDH2 mutant samples, which increases their susceptibility to lipid peroxidation at baseline. In line, we also observed increased expression of CD36, a fatty acid (FA) transporter, in this subset of patients. Co-treatment with palmitate, a saturated FA, increased metformin sensitivity. Contrarily, CD36 knockdown rendered IDH2^R140Q cells more resistant to treatment, suggesting that disturbed lipid homeostasis is a key factor in determining metformin sensitivity in AML. Given the increased lipid import in IDH1/2 and FLT3-mutant cells, we investigated whether the machinery involved in lipid droplet formation was also affected. Transcriptome analysis of metformin-treated IDH2^R140Q cells revealed increased expression of DGAT1, a key enzyme in converting FA-CoA into triacylglycerides for lipid droplet formation. ssGSEA also identified an enrichment of terms related to lipid droplet formation and organization in IDH2^R140Q cells which could be validated experimentally. Combining metformin with a DGAT1 inhibitor, T863, resulted in an additive-to-synergistic effect in our cell lines and a panel of primary AML samples. In conclusion, we demonstrated that OXPHOS inhibition by metformin is effective across a wide range of genetically diverse AMLs, particularly in those with disturbed lipid metabolism. We showed that defective metabolic rewiring towards glycolysis, along with increased lipid peroxidation, ROS production, and lipid droplet formation, are mechanisms associated with enhanced metformin sensitivity in AML. Our results highlight the potential of leveraging metabolic vulnerabilities in AML for the development of more effective and personalized therapeutic strategies.