A novel gene signature for predicting response to chemoradiotherapy in locally advanced esophageal adenocarcinoma.

Abstract

4571 Background: While neoadjuvant chemoradiotherapy (CRT) has emerged as an important treatment modality in patients with locally advanced esophageal adenocarcinoma (EAC), ~60%-70% of patients do not respond to such treatments; but are exposed to their toxicity nonetheless. This highlights the clinical need for the development of biomarkers that can robustly predict response to CRT and spare others from the toxicity and expense associated with these treatments. Herein, we systematically and comprehensively identified a biomarker signature that predicts response to neoadjuvant therapy in EAC patients. Methods: A cohort of 31 EAC patients treated with chemotherapy or chemoradiotherapy was assembled, with a majority of patients receiving carboplatin, paclitaxel and concurrent radiotherapy. Specifically, we performed a capture based targeted sequencing in paired biopsy specimens obtained at baseline and 3-6 weeks post-treatment. In addition, we also analyzed the predictive potential of a panel of immune-related genes (TIM3, LAG3, IDO1 and CXCL9) in these matched pre- and post-treatment tissues. Results: In our cohort, based upon pathologic response to neoadjuvant CRT, 8 EAC patients were categorized as non-responders, while 23 were deemed as responders to CRT. Among responders, the most frequently mutated genes were MKI67, SYNE1, PCLO, RECQL4, MSH3, NOTCH2, ILR7, CIITA, LRRK2 and EML4, and the overall tumor mutation burden (TMB) was significantly reduced for these genes in post-treatment samples ( P=5.73E-03). Similarly, in non-responders NLRP1, MAP3K1, ASMTL, and ALK harbored frequent mutations, and the TMB was significantly reduced for these genes in post-treatment samples ( P=5.57E-03). We compared responders and non-responders from the pre-treatment samples and identified differentially mutated genes including EPHA5, ZNF217, RELN, PALB2 and MYO18A. Similarly, responders had all four immune-related genes significantly up-regulated in post-treatment samples than pre-treatment samples. We constructed a risk-stratification model that comprised of mutational score from 5 differentially mutated genes, together with 4 immune-related genes, which achieved an AUC of 0.96 in predicting response to CRT in EAC patients ( P=1.03E-04). Conclusions: Using a systematic biomarker discovery approach, we have developed a novel biomarker signature that robustly predicts response to CRT in EAC patients and has a significant potential for personalized management of locally advanced EAC patients.