Molecular profiling of adenocarcinoma of gastroesophageal junction compared to esophageal and gastric adenocarcinoma

Abstract

Introduction: Biologic understanding of adenocarcinoma of gastroesophageal junction (AGEJ) and similarity to gastric or esophageal adenocarcinoma has been long standing controversial issue. The purpose of our study is to evaluate molecular characteristics of AGEJ compared to esophageal (EAC) or gastric adenocarcinoma using next generation sequencing NGS data of the Cancer Genome Atlas (TCGA) and Seoul National University (SNU) cohorts. Methods: We retrieved NGS data of esophageal adenocarcinoma (EAC, n=78), adenocarcinoma of gastroesophageal junction or cardia (GEJ/cardia, n=48) and gastric adenocarcinoma located at fundus or body of the stomach (GCFB, n=102) from TCGA cohort. For SNU cohort, whole exome and transcriptome sequencing were carried out for each pair of tumor and corresponding normal gastric mucosae of AGEJ II (n=16 pairs), AGEJ III (n=16 pairs) and upper third gastric adenocarcinoma (UT, n=14 pairs). Class prediction model was developed using Bayesian compound covariate predictor (BCCP) with Leave-one-out cross validation between EAC and GCFB of TCGA cohort, and tested for GEJ/cardia tumors from TCGA and all tumors from SNU cohort. Results: The class prediction model using 400 differentially expressed classifier genes (90.2% of sensitivity and 89.7% of specificity) showed a spectral transition of clusters between EAC-like and GCFB-like group without any entirely distinguishable cluster. Using 0.4535 of BCCP score as a cut-off value, 68.8% of GEJ/Cardia of TCGA cohort and AGEJ II of SNU cohort were identified as GCFB-like group. AGEJ III of SNU cohort consisted of 93.7% of GCFB-like adenocarcinoma, and there was no significant relationship between involvement of GEJ and molecular classification of AGEJ III. EAC-like group was significantly related to differentiated and intestinal type, and showed significantly amplified copy number of ERBB2 compared to GCFB group. Reverse phase protein array and tissue microarray revealed significant overexpression of EGFR and ERBB2 in EAC-like than GCFB-like group. Drug response analysis of lapatinib from Cancer Cell Line Encyclopedia database demonstrated significantly lower half maximal inhibitory concentration for EAC-like than GCFB-like. Conclusions: Molecular classification of AGEJ using BCCP with 400 classifier genes demonstrated that GEJ/cardia in TCGA cohort and AGEJ II in SNU cohort were a combination of 31.2% of EAC-like group and 68.8% of GCFB-like group. EAC-like group was significantly related to differentiated, intestinal type and shows significant copy number amplification of ERBB2 and overexpression of ERBB2 and EGFR. EAC-like group can be a promising target for EGFR and ERBB2 tyrosine kinase inhibitor.