Synthetic lethality-based predictive biomarker identification of splicing modulators in lung cancer

Identification of predictive biomarkers for targeted drugs is an active research area in precision oncology. Usually, a genetic alteration in the drug target is used to identify responders, which can limit the use of the drug in other responsive cancers. Alternatively, cell line-based drug sensitivity data are used to build machine-learning models to identify predictive biomarkers. These methods require large amounts of data and the black-box models developed are often hard to interpret. Synthetic lethality provides an alternate approach for predictive biomarker identification of targeted drugs. In synthetic lethal (SL) interactions, a defect in one gene leads to dependency on a second gene. Neither defect by itself is essential for survival, but together they lead to cell death. Thus, the inhibition of the drug target in the presence of a cancer-specific genetic alteration would result in cancer cell death. The genetic alterations can be used as biomarkers for the drug. We developed a novel computational pipeline for predictive biomarker identification based on our tool, Mining Synthetic Lethals (MiSL), that mines SL interactions from pan-cancer primary tumor data (such as TCGA). Since MiSL utilizes primary human tumor data, it enables the identification of SL interactions in the native context of human tumors and is more likely to find relationships relevant to in vivo tumor biology than shRNA/CRISPR screens. To discover biomarkers of a targeted drug, we used MiSL to identify genetic alterations that are SL with targets of the drug in a specific cancer type. As proof of principle, we applied our method to identify biomarkers of sudemycin-D6 (SD6), a potent splicing modulator, in non-small cell lung cancer (NSCLC). First, we identified putative drug targets of SD6. Since SD6 is known to inhibit SF3B1, any gene belonging to the same protein complex as SF3B1 was termed a SD6 drug target. Using these SD6 drug targets, our MiSL-based pipeline identified KRAS mutations, present in ~30% of NSCLC patients, as a SD6 biomarker. To confirm our predictions, we treated a panel of KRAS-mutant-and-dependent NSCLC lines (H358, H23 and H441) and a KRAS-WT NSCLC line (H2228) with SD6 and measured cell viability. All KRAS-mutant lines tested were significantly more sensitive to SD6 (> 5-30X lower IC50) than the KRAS-WT line. Furthermore, the KRAS-mutant lines were also more sensitive (5-20X lower IC50), compared to the KRAS-WT line, to a splicing modulator of a different class – a kinase inhibitor that phosphorylates components of the spliceosome. In vivo studies are underway to further validate our in vitro findings. This study identifies KRAS mutation as a novel predictive biomarker for splicing modulators in NSCLC, which could lead to new therapeutic options for KRAS-mutated NSCLC. Furthermore, our work demonstrates the feasibility of a synthetic lethality-based pipeline to predict biomarkers for targeted drugs. Citation Format: Claire E. Repellin, Puja Patel, Yihui Shi, Helena Gong, Thomas R. Webb, Lidia Sambucetti, Subarna Sinha. Synthetic lethality-based predictive biomarker identification of splicing modulators in lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 4015.