A predictive signature for oxaliplatin and 5-fluorouracil based chemotherapy in locally advanced gastric cancer

Abstract

Adjuvant chemotherapy(AC) plays a substantial role in the treatment of locally advanced gastric cancer (LAGC), but the response remains poor. We aims to improve its efficacy in LAGC. Therefore, we identified the expression of eight genes closely associated with platinum and fluorouracil metabolism (RRM1, RRM2, RRM2B, POLH, DUT, TYMS, TYMP, MKI67) in the discovery cohort (N=291). And we further validated the findings in TCGA (N=279) and GEO. Overall survival (OS) was used as an endpoint. Univariate and multivariate Cox models were applied. A multivariate Cox regression model was simulated to predict the OS. In the discovery cohort, the univariate Cox model indicated that AC was beneficial to high-RRM1, high-DUT, low-RRM2, low-RRM2B, low-POLH, low-KI67, low-TYMS or low-TYMP patients, the results were validated in the TCGA cohort. The multivariate Cox model showed consistent results. Cumulative analysis indicated that patients with low C-Score respond poorly to the AC, whereas the high and medium C-Score patients significantly benefit from AC. A risk model based on the above variables successfully predicted the OS in both cohorts (AUC=0.75 and 0.67, respectively). Further validation in a panel of gastric cancer cell (GC) lines (N=37) indicated that C-Score is significantly associated with IC50 value to fluorouracil. Mutation profiling showed that C-Score was associated with the number and types of mutations. In conclusion, we successfully simulated a predictive signature for the efficacy of AC in LAGC patients and further explored the potential mechanisms. Our findings could promote precision medicine and improve the prognosis of LAGC patients.

Keywords

• Locally advanced gastric cancer
• Adjuvant chemotherapy
• Overall survival
• Prediction
• Mutation