S419 Discovery of Methylated DNA Biomarkers for the Potential Non-Endoscopic Detection of Barrett's Esophagus (BE)

Introduction: Barrett’s esophagus (BE) is a known precursor of esophageal adenocarcinoma (EAC), the type of esophageal cancer that accounts for over 80% of all esophageal cancer cases in the United States. The current gold standard for diagnosis of BE is endoscopic biopsy with histologic examination. Unfortunately, most individuals in the United States do not undergo endoscopy and thus BE is not diagnosed, resulting in progression toward fatal EAC in most undiagnosed BE cases. For this reason, minimally invasive sponge or balloon devices have been studied in conjunction with various types of biomarkers. However, most markers used in current BE diagnostic biomarker panels are outdated, lacking in sufficient specificity and sensitivity, and warrant substantial improvement. Methods: Using a novel approach, we accessed and integrated 6 Infinium HumanMethylation450 BeadChip datasets from various research groups within the Gene Expression Omnibus (GEO) database, from which we selected probes that were highly methylated in Barrett’s (beta ≥ 0.30) and mostly unmethylated in normal esophageal and gastric tissues (beta ≤ 0.05) yielding 30 candidate BE-specific markers. All 30 were identified from research groups who used microdissection and/or careful histopathologic review of each tissue biopsy. We then analyzed 248 BE, 184 normal esophageal, and 101 normal gastric tissue samples from our archives. After designing qMSP primers and probes, and further testing, we assayed 14 candidate markers in 21 matched normal-BE tissue pairs, 8 matched normal-BE-tumor tissue triplets, and 17 matched normal-tumor tissue pairs. Results: All 14 biomarkers tested exhibited significantly higher methylation levels in BE DNAs vs. matched normal DNAs (p < 0.01 by Wilcoxon rank-sum test). 4 of the 14 markers showed significantly higher methylation levels in tumor DNAs vs. matched normal DNAs (P < 0.01). 3 markers were statistically significantly different across matched normal, BE, and tumor tissue triplets (Kruskal-Wallis test, p < 0.05), and these 3 markers were used to develop a diagnostic panel for future testing on minimally invasively obtained cytologic DNAs from sponge-capsule samples. Conclusion: This discriminatory biomarker panel shows potential for BE diagnosis using an inexpensive, minimally invasive sampling technique and thus merits further study in case-control sponge studies. Due to our systematic and rigid method of selecting these markers, these genes are expected to be extremely important for the diagnosis of BE.