Novel Predictive and Prognostic Gene Signatures for Chemoradiotherapy in Locally Advanced Esophageal Adenocarcinoma

Abstract

Esophageal adenocarcinoma is the sixth leading cause of cancer-related deaths in the United States. For locally advanced patients, neoadjuvant chemoradiotherapy followed by surgery is the standard of care. Risk stratification relies heavily on clinicopathologic features, particularly pathologic response, which is inadequate. This highlights the need for the development of biomarkers that can help improve outcomes. Herein, we have developed 2 distinct biomarker signatures to either predict response to neoadjuvant therapy or stratify patients into risk categories for survival.We assembled a cohort of 34 patients with locally advanced esophageal adenocarcinoma, with the majority (21) receiving concurrent chemoradiotherapy with carboplatin/paclitaxel. A capture based targeted sequencing was performed in paired biopsy specimens obtained at baseline, 3- and 6-weeks post-treatment. Differentially mutated gene analysis between responders and non-responders of treatment was performed to determine predictors of response. A univariate Cox proportional hazard regression was used to examine associations between gene mutation status and overall survival. A risk score was then calculated for each patient using a linear combination of univariate coefficient and mutation status of gene.15 patients (71%) receiving the CROSS regimen were classified as responders. Among responders to chemotherapy, the most frequently mutated genes were MKI67, SYNE1, PCLO, MSH3, RECQL4, NOTCH2, ILR7, CIITA, LRRK2 and EML4. Tumor mutation count was significantly reduced for these genes in post-treatment samples (P = 5.89E-03). We have identified a 3 gene signature based on mutations in EPHA5, BCL6, and ERBB2 that robustly predicts response to carboplatin/paclitaxel based neoadjuvant therapy. For this model, sensitivity is 84.6% and specificity is 100%. In the overall cohort, a 9 gene signature using APC, MAP3K6, ETS1, CSF3R, PDGFRB, GATA2, ARID1A, and FGF6, was also created that significantly stratifies patients into risk categories, prognosticating for improved relapse-free (P = 0.005) and overall survival (P = 3.325E-06). Sensitivity for this model is 73.33% and specificity is 94.74%.We have identified a 3-gene signature (EPHA5, BCL6 and ERBB2) that is predictive of response to neoadjuvant therapy and a separate 9 gene classifier which prognosticates for survival outcomes. These provide significant potential for personalized management of locally advanced esophageal cancer.