Background
Saccharomyces cerevisiae represses respiration in the presence of adequate glucose, mimicking the Warburg effect, termed aerobic glycolysis. We conducted yeast phenomic experiments to characterize differential doxorubicin-gene interaction, in the context of respiration vs. glycolysis. The resulting systems level biology about doxorubicin cytotoxicity, including the influence of the Warburg effect, was integrated with cancer pharmacogenomics data to identify potentially causal correlations between differential gene expression and anti-cancer efficacy. Methods
Quantitative high-throughput cell array phenotyping (Q-HTCP) was used to measure cell proliferation phenotypes (CPPs) of the yeast gene knockout/knockdown library, treated with escalating doxorubicin concentrations in fermentable and non-fermentable media. Doxorubicin-gene interaction was quantified by departure of the observed and expected phenotypes for the doxorubicin-treated mutant strain, with respect to phenotypes for the untreated mutant strain and both the treated and untreated reference strain. Recursive expectation-maximization clustering (REMc) and Gene Ontology-based analyses of interactions were used to identify functional biological modules that buffer doxorubicin cytotoxicity, and to characterize their Warburg-dependence. Yeast phenomic data was applied to cancer cell line pharmacogenomics data to predict differential gene expression that causally influences the anti-tumor efficacy, and potentially the anthracycline-associated host toxicity, of doxorubicin. Results
Doxorubicin cytotoxicity was greater with respiration, suggesting the Warburg effect can influence therapeutic efficacy. Accordingly, doxorubicin drug-gene interaction was more extensive with respiration, including increased buffering by cellular processes related to chromatin organization, protein folding and modification, translation reinitiation, spermine metabolism, and fatty acid beta-oxidation. Pathway enrichment was less notable for glycolysis-specific buffering. Cellular processes exerting influence relatively independently, with respect to Warburg status, included homologous recombination, sphingolipid homeostasis, telomere tethering at nuclear periphery, and actin cortical patch localization. Causality for differential gene expression associated with doxorubicin cytotoxicity in tumor cells was predicted within the biological context of the phenomic model. Conclusions
Warburg status influences the genetic requirements to buffer doxorubicin toxicity. Yeast phenomics provides an experimental platform to model the complexity of gene interaction networks that influence human disease phenotypes, as in this example of chemotherapy response. High-resolution, systems level yeast phenotyping is useful to predict the biological influence of functional variation on disease, offering the potential to fundamentally advance precision medicine.
