Abstract 1199: NSCLC KRAS G12C and G12V mutations drive different pathways and display specific drug sensitivity patterns.

Abstract

Abstract Background: Although a recent body of literature has described the molecular environment associated with KRAS G12D mutations (6%) in lung adenocarcinoma, the global mutational and gene expression environments of the most frequent types of KRAS mutations : G12C (45%) and G12V (25%) remain largely unknown. This study aimed: (i) to describe the differences in the mutational and gene expression environment in KRAS G12C and G12V genotypes; (ii) to correlate molecular models predictive of KRAS G12C and G12V genotypes with in vitro drug sensitivity in a large drug panel screen of cell lines (Cancer Cell Line Encyclopedia, CCLE). Materials and Methods: Data sets were extracted from public repositories (TCGA LUAD n= 249, BATTLE n= 124, CCLE Lung n= 82) or those pending release upon publication (CHEMORES n= 123). We used gene set enrichment analysis (GSEA) to identify pathways associated with mutational and gene expression changes, and penalized logistic regression (ElasticNet) to build models of KRAS G12C and G12V mutations. Correlations were computed between these models and drug sensitivity represented by the IC50 of 21 drugs on the 82 lung cancer cell lines of the CCLE. Results: The KRAS G12C phenotype, defined either by the mutational environment or by the associated gene expression signature, was enriched in pathways related to direct KRAS binding proteins such as GRB2, SOS1, SHC, FRS2 pathways (adjusted P values <.01). Conversely, the KRAS G12V phenotype was consistently enriched in pathways related to cell-cycle checkpoint gene signaling including P53, Rb, ATM and PML (adjusted P values <.05). Separate G12C and G12V gene expression models trained to predict the respective mutations in TCGA were robust by independent validation in BATTLE and CCLE with AUC = 0.78 & 0.61 and AUC= 0.60 & 0.76 respectively. In CCLE we computed cell-line specific scores for our G12C and G12V models and correlated that with IC50 for the 21 available drugs. Interestingly, G12C model scores positively correlated with sensitivity to AEW541 (IGF1R inhibitor) (P=.001), while G12V model scores were positively correlated with sensitivity to PD-0332991 (cyclin D kinase 4/6 inhibitor) (P=.02) and with resistance to TKI258 (multi-target VEGFR/PDGFR kinase inhibitor) (P=.02). Conclusion: Our analyses reveal singular differences between KRAS G12C and G12V mutations in lung cancer that are consistent when viewed either in terms of associated mutations, gene expression profiles, or sensitivity to drugs. The G12C phenotype exhibits a strong association with direct KRAS interactors and correlates with sensitivity to IGF1R inhibitors. The G12V phenotype demonstrates a strong association with cell-cycle checkpoint pathways and predicts sensitivity to the Cyclin D kinase inhibitor, PD-0332991. Citation Format: Charles Ferte, Justin Guinney, Jonathan Derry, erich H. huang, Benjamin Besse, Jean-Charles Soria. NSCLC KRAS G12C and G12V mutations drive different pathways and display specific drug sensitivity patterns. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 1199. doi:10.1158/1538-7445.AM2013-1199