A Histone Deacetylase Inhibitor Induces Acetyl-CoA Depletion Leading to Lethal Metabolic Stress in RAS-Pathway Activated Cells

SIMPLE SUMMARY: Epigenetic therapies have been difficult to translate to solid tumors, in part because a lack of a full mechanistic understanding has not allowed the selection of tumors most likely to benefit from the therapies. Here we identified a RAS-phenotype that can be targeted by the histone deacetylase inhibitor (HDACi) romidepsin. We showed that the hyperacetylation induced by romidepsin depletes acetyl-CoA, the cell donor substrate for acetylation, and it leads to metabolic stress and death in KRAS-activated cells. Transcriptomic analysis confirmed that perturbation of two acetyl-CoA generation pathways (fatty acid metabolism and branched-chain amino acid (BCAA) metabolism) were correlated with HDACi sensitivity in a 608-cell line panel and in patients treated with belinostat. Our analysis highlights the potential utility of an acetyl-CoA phenotype to sharpen treatment choices for RAS-activated tumors, and it suggests that acetyl-CoA depletion could be a key effect underlying the myriad cellular responses that follow HDAC inhibition. ABSTRACT: Background: The mechanism of action of romidepsin and other histone deacetylase inhibitors is still not fully explained. Our goal was to gain a mechanistic understanding of the RAS-linked phenotype associated with romidepsin sensitivity. Methods: The NCI60 dataset was screened for molecular clues to romidepsin sensitivity. Histone acetylation, DNA damage, ROS production, metabolic state (real-time measurement and metabolomics), and gene expression alterations (transcriptomics) were determined in KRAS-WT versus KRAS-mutant cell groups. The search for biomarkers in response to HDACi was implemented by supervised machine learning analysis on a 608-cell transcriptomic dataset and validated in a clinical dataset. Results: Romidepsin treatment induced depletion in acetyl-CoA in all tested cell lines, which led to oxidative stress, metabolic stress, and increased death—particularly in KRAS-mutant cell lines. Romidepsin-induced stresses and death were rescued by acetyl-CoA replenishment. Two acetyl-CoA gene expression signatures associated with HDACi sensitivity were derived from machine learning analysis in the CCLE (Cancer Cell Line Encyclopedia) cell panel. Signatures were then validated in the training cohort for seven HDACi, and in an independent 13-patient cohort treated with belinostat. Conclusions: Our study reveals the importance of acetyl-CoA metabolism in HDAC sensitivity, and it highlights acetyl-CoA generation pathways as potential targets to combine with HDACi.