Adenocarcinoma of the stomach and esophagogastric junction with low DNA methylation show poor prognoses.

Abstract

Gastric cancer (GC) is characterized by unique DNA methylation epigenotypes (MEs). However, MEs including adenocarcinomas of the esophagogastric junction (AEG) and background non-neoplastic columnar mucosae (NM) remain to be clarified. We analyzed the genome-wide DNA MEs of AEG, GC, and background NM using the Infinium 450k beadarray, followed by quantitative pyrosequencing validation. Large-scale data from The Cancer Genome Atlas (TCGA) were also reviewed. Unsupervised two-way hierarchical clustering using Infinium data of 21 AEG, 30 GC, and 11 NM revealed four DNA MEs: extremely high-ME (E-HME), high-ME (HME), low-ME (LME), and extremely low-ME (E-LME). Promoter methylation levels were validated by pyrosequencing in 146 samples. Non-inflammatory normal mucosae were clustered into E-LME, whereas gastric or esophagogastric junction mucosae with chronic inflammatory changes caused by either Helicobacter pylori infection or reflux esophagitis were clustered together into LME, suggesting that inflammation status determined DNA MEs regardless of the cause. Three cases of Barrett's-related adenocarcinoma were clustered into HME. Among 94 patients whose tumors could be clustered into one of four MEs, 11 patients with E-LME cancers showed significantly shorter overall survival than that in the other MEs, even with the multivariate Cox regression estimate. TCGA data also showed enrichment of AEG in HME and a poorer prognosis in E-LME. E-LME cases, newly confirmed in this study, form a unique subtype with poor prognosis that is not associated with inflammation-associated elevation of DNA methylation levels. LME could be acquired via chronic inflammation, regardless of the cause, and AEG might preferentially show HME.