IDH1 Mutant AML Is Susceptible to Targeting De Novo Lipid Synthesis Independent of 2-Hydroxyglutarate and Has a Distinct Metabolic Profile from IDH2 Mutant AML

Introduction: Mutations in IDH1 and IDH2 are recurrent in AML and several other cancers, resulting in the aberrant production of the onco-metabolite, R-2-hydroxyglutarate (2-HG), as well as an inability of mutant IDH1 to convert cytoplasmic alpha-ketoglutarate to isocitrate via reductive carboxylation. Currently, inhibitors of the neomorphic enzymes that abrogate the production of 2-HG, such as AG-120, are FDA-approved, but are not curative. Using a novel computational method (MiSL), we identified acetyl CoA carboxylase (ACACA) as a potential druggable target specifically in IDH1-mutated AML. ACACA regulates the de novo synthesis of lipid precursors by converting acetyl CoA to malonyl CoA building blocks. We hypothesize that IDH1 mutant AML exhibits a defect in reductive carboxylation and de novo fatty acid synthesis conferring preferential susceptibility to ACACA inhibition. Here, we investigate this hypothesis by comprehensively quantifying the metabolic landscape, including non-polar lipid metabolites, conferred by IDH1 R132H mutation compared to IDH2 mutation in isogenic cell lines and primary samples. Moreover, we investigate the in vitro and in vivo effects of targeting de novo lipid synthesis on IDH1 and IDH2 mutant AML.

Methods: Comprehensive metabolomic profiling of primary FACS-purified AML blasts was performed using an in-house protocol optimised for extraction of non-polar lipid metabolites from less than 1 million primary cells. CD33+CD45+ leukemic blasts were profiled from 17 patient samples with IDH1 mutation (n=6), IDH2 mutation (n=5), or IDH1/2 wildtype (n=6) after culturing in serum-free media. 6 independent cord-blood CD34+ cells were profiled as a negative control. For validation of IDH1-specific effects, isogenic THP-1 cells transduced with doxycycline-inducible wildtype and R132H mutant IDH1 or R140Q mutant and wildtype IDH2 were profiled. Molecules were identified according to their molecular weight and retention time using Mass Hunter software (Agilent) and the Human Metabolome Database. For in vivo studies, primary AML samples were engrafted in NSG mice that were subjected to dietary modification with low lipid diet and/or treatment with selective inhibitors of ACACA and mutant IDH1.

Results: Principle component analysis of metabolite abundance of 1400 unique compounds revealed striking differences between IDH1 and IDH2 mutant AML. Both IDH1 and IDH2 mutant samples produced high levels of 2-HG compared to wildtype AML and CD34+ cells (50 fold increase, P=4.5E-05). A major perturbation in multiple phospholipid fatty acid species was conferred by IDH1 R132H, but not by IDH2 mutation. The same pattern was observed in cell lines with 49 lipid species decreased in the presence of mutant IDH1 compared to only 2 perturbed with mutant IDH2. Direct comparison of IDH1 vs IDH2 mutant primary samples revealed 54 lipid metabolites significantly down-regulated in IDH1 mutant blasts (adjusted P value <0.05).

To investigate the effects of targeting de novo lipid synthesis on IDH1 mutant AML in vivo, we engrafted primary IDH1 mutant AML and tested growth with lipid-free compared to normal diet. At 12 weeks, IDH1 mutant AML showed reduced growth in the bone marrow of mice on lipid-free diet (SU389 11% vs 40%, n=10 mice, P=0.03 and SU372 21% vs 34%, n=10, P=0.02 Mann-Whitney U). IDH1 mutant AML was susceptible to ACACA inhibition with shRNA, CRISPR targeting, or selective nanomolar inhibitors. Knockdown of ACACA with independent shRNAs caused a defect in cell growth in the presence of IDH1 R132H, but not in its absence or with scrambled shRNA (p=0.009, shRNA #1 vs. scrambled; p=0.01, shRNA #2 vs. scrambled) in vitro and in xenografts. Primary IDH1 R132 mutated AML blasts were selectively sensitive to ACACA inhibitor treatment compared to IDH1 wildtype normal karyotype blasts (IC50 0.6 uM vs 6 uM, p=0.009). Notably, IDH1-mutant AML blasts pre-treated with 10mM AG-120 remained susceptible to ACACA inhibition, identifying a 2-HG independent vulnerability. Similar findings were observed in a solid tumor IDH1 mutant sarcoma model in vivo.

Conclusion: These results support our hypothesis that IDH1 mutant AML exhibits a defect in de novo fatty acid synthesis conferring preferential susceptibility to ACACA inhibition, and suggests that pharmacologic inhibitors of ACACA may complement IDH1 mutation-specific inhibitors in the clinic.