Systematic Identification of Gene-Drug Interactions Using an Advanced CRISPR Screening Platform to Predict Therapy Response across Cancer Types

Cancer entities and the presence or absence of genetic driver mutations often determine therapeutic trajectories for patients. However, the efficacy of chemotherapeutics and targeted therapies varies significantly and rarely induces long-term remissions due to the rapid emergence of therapy resistance. Consequently, there is an urgent need to identify response-predictive biomarkers that enable effective precision medicine. We hypothesized that, similar to genetic drivers that qualify patient subgroups for specific targeted therapeutics, genetic aberrations exist within cancer cells that confer intrinsic sensitivity to chemotherapeutics and small-molecule inhibitors lacking response-predictive biomarkers. To test this, we developed an advanced CRISPR/Cas9-based high-throughput screening platform to systematically interrogate gene-drug interactions of 18 clinically established and emerging therapeutics used for treating hematopoietic malignancies. These included antimetabolites, anthracyclines, hypomethylating agents, and various selective small-molecule inhibitors. Utilizing optimized sgRNA prediction algorithms and a dual-sgRNA design, we constructed low-complexity genome-wide sgRNA libraries that enhance the technical robustness and throughput of drug-modifier screens in culture and enable in vivo screens. We uncovered 180 genes including various established and previously unknown factors that upon knockout result in selective synergy or resistance to one of the 18 therapeutics. We functionally validated various gene-drug interactions in multiple leukemia and solid cancer cell lines, suggesting that the loss of these genes confers predictable drug response to specific compounds independent of the tissue context. As the mechanistic relationship of many gene-drug interactions was uncharted, we assessed their impact on drug metabolism using targeted liquid chromatography-tandem mass spectrometry (LC-MS/MS). While several genes modified cellular drug influx/efflux or metabolism, many others altered drug sensitivity by distinct means that remain to be explored. However, integrating publicly available whole-exome sequencing data pinpointed patients across diverse cancer entities harboring homozygous deletions within selected genes that modify drug response, suggesting these patients could display predictable drug sensitivity. To test this in vivo, we probed three drug modifier genes (SAMHD1, XRCC4, and USP48) that are recurrently deleted in cancer patients using colorectal cancer cell line-derived xenotransplants and examined their response to Cytarabine, Doxorubicin, and Decitabine treatment, respectively. Strikingly, while wildtype colorectal cancer transplants were unresponsive, tumors with deletions in these genes were amenable to treatment, resulting in significant tumor control. Together, our data show that our scalable screening pipeline can systematically interrogate gene-drug interactions for a broad panel of drugs lacking response-predictive biomarkers. They suggest that integrating drug modifier genes in targeted exome sequencing panels of cancer patients may pave the way for discovering unexplored treatment opportunities.