DNA methylation profiling in Barrett's esophagus and esophageal adenocarcinoma reveals unique methylation signatures and molecular subclasses

Abstract

Barrett's esophagus (BE) is a metaplastic process whereby the normal stratified, squamous esophageal epithelium is replaced by specialized intestinal epithelium. Barrett's is the only accepted precursor lesion for esophageal adenocarcinoma (EAC), a solid tumor that is rapidly increasing in incidence in western countries. BE evolves into EAC through intermediate steps that involve increasing degrees of dysplasia. Current histologic criteria are quite subjective and the clinical behavior of BE is highly variable and difficult to predict using these standards. It is widely believed that molecular alterations present in BE and EAC will provide more precise prognostic and predictive markers for these conditions than the current clinical and histologic features in use. In order to further define molecular alterations that can classify unique groups of BE and EAC, we utilized methylation microarrays to compare the global gene methylation status of a collection of normal squamous, BE, BE + high-grade dysplasia (HGD), and EAC cases. We found distinct global methylation signatures, as well as differential methylation of specific genes, that discriminated these histological groups. We also noted high and low methylation epigenotypes among the BE and EAC cases. Additional validation of those CpG sites that distinguished BE from BE + HGD and EAC may lead to the discovery of useful biomarkers with potential clinical applications in the diagnosis and prognosis of BE and EAC.

Figures

Figure 1

Box plot depicting distribution of β values of 706 CpGs from the GoldenGate array for each sample group. Each box contains the distribution of β values from 25–75%, with the upper and lower ‘whiskers’ representing the upper and lower quartile of β value distributions, respectively. The line within each box represents the median β value. The median β value was lower in the squamous group, increased in the BE group, and highest in the HGD and EAC groups (ρ

Figure 2

Dendrograms (‘heat maps’) generated from unsupervised hierarchical clustering analyses revealed distinct methylation profiles based on histological subtype. Each column represents a single sample, and each row an individual CpG site. The darker blue colors correspond to higher β values (i.e., a greater degree of methylation) at particular CpGs. (A) EAC vs. SQ samples. (B) BE vs. SQ samples. (C) HGD vs. SQ samples.

Figure 3

Dendrograms generated from unsupervised hierarchical clustering analyses within a single histological subtype. Each column represents a single sample, and each row an individual CpG site; the darker blue colors correspond to higher β values. When analyzing the BE cases independently (A) or the EAC cases independently (B), we noted two distinct methylation profiles within each of these tissue types: a high and low methylation epigenotype (HME and LME) subgroup.

Figure 4

Bisulfite sequencing results are shown for four genes. For each gene, one or two HME and one LME case were chosen for sequencing, and the region containing the CpG from the GoldenGate array was selected for analysis. Shown under each gene name is the gene structure along with the location of the CpG island (gray box) and the region that was selected for sequencing (black bar). Individual sequences from clones are shown beneath this; methylated CpGs are depicted by a black circle and unmethylated CpGs by a white circle. The β value of the CpG from the array can be compared with the methylation index (MI = methylated CpGs/total CpGs), a measure of overall methylation that captures the CpG of interest as well as surrounding CpG dinucleotides, which is calculated from the sequencing data.