Prediction of Tissue Origin of Adenocarcinomas in the Esophagogastric Junction by DNA Methylation

Background Prediction of tissue origin of esophagogastric junction (EGJ) adenocarcinomas can be important for therapeutic decision, but no molecular marker is available. Here, we aimed to develop such a marker taking advantage of tissue-specific profiles of DNA methylation. Methods DNA methylation profiles of gastric adenocarcinomas (GACs) were obtained by an Infinium HumanMethylation450 BeadChip array, and those of esophageal adenocarcinoma (EACs) were obtained from the TCGA database. DNA from formalin-fixed paraffin-embedded (FFPE) samples was analyzed by bisulfite pyrosequencing. Results In the screening set, 51 of 145,841 CpG sites in CpG islands were methylated at significantly higher levels in 30 GACs compared to those in 30 EACs. Among them, SLC46A3 and cg09177106 were unmethylated in all the 30 EACs. Predictive powers of these two markers were successfully confirmed in an independent validation set (18 GACs and 18 EACs) ( SLC46A3 , sensitivity = 77.8%, specificity = 100%; cg09177106, sensitivity = 83.3%, specificity = 94.4%), and could be applied to FFPE samples (37 GACs and 18 EACs) ( SLC46A3 , P = 0.0001; cg09177106, P = 0.0028). On the other hand, EAC-specific markers informative in the FFPE samples could not be isolated. Using these GAC-specific markers, nine of 46 (19.6%) TCGA EGJ adenocarcinomas were predicted to be GACs. Conclusions Two GAC-specific markers, SLC46A3 and cg09177106, had a high specificity for identifying the tissue origin of EGJ adenocarcinoma.