Background
The incidence of esophageal adenocarcinoma (EAC) is increasing rapidly and patient outcomes remain poor. Known risk factors for EAC include obesity, gastro-esophageal reflux, and the presence of Barrett's esophagus (BE). Repeated injury from gastro-duodenal reflux is thought to result in the replacement of the esophageal squamous mucosa with a metaplastic columnar lined epithelium. The presence of goblet cells within columnar epithelium is diagnostic for BE. Approximately 0.5 – 1% of patients with BE will develop EAC each year and patients with BE have a 50- to 100-fold increased risk of EAC compared to the general population [
1,
2].
The progression from BE to EAC is generally accepted to proceed via the histological stages of low-grade and high-grade dysplasia. Progress along this pathway appears to mirror the accumulation of genetic abnormalities, with a number of reports suggesting a stepwise progression of genetic changes. Abnormalities in CDKN2A, seen in BE metaplastic tissue, followed by altered TP53 expression, generally reported in dysplastic tissue, have been associated with the transition from BE to EAC [
3,
4]. Alterations in the expression of many other genes have also been described at different stages of the progression to cancer [
5,
6].
Genes can be aberrantly down-regulated as a result of genomic alterations such as mutation, deletion, or DNA methylation. In humans, methylation of cytosines generally occurs in the context of a CpG dinucleotide. Regions of relatively high CpG content, termed CpG islands, are found in the promoter region of many genes. Methylation within these regions is associated with suppression of the expression of certain genes, for example the switching off of key developmental genes in adult tissues. However, aberrant methylation of normally unmethylated CpG islands is a common cause of altered gene expression in cancer. As esophageal biopsies are easily obtained at endoscopy, it is possible to determine at what stage methylation occurs during the progression to cancer. In this study we compared the methylation of nine genes, frequently methylated in other cancers, between squamous, BE and EAC tissues.
Discussion
Methylation is an important mechanism for the silencing of genes in the development of cancer. We compared the frequency and extent of methylation of APC, CDKN2A, ID4, MGMT, RBP1, RUNX3, SFRP1, TIMP3, and TMEFF2 in esophageal adenocarcinoma (EAC), high grade dysplastic Barrett's from patients with adenocarcinoma (D-EAC), metaplastic Barrett's from patients without dysplasia or adenocarcinoma (BE), and histologically normal esophageal squamous epithelium. All nine genes showed an increase in frequency of methylation in EAC compared to squamous epithelium. We found that for seven of these genes there was no difference in the frequency of methylation between the BE and EAC. For only CDKN2A and RUNX3 was there a significant increase in frequency of methylation in EAC compared to BE. The extent of methylation in EAC was significantly greater than in BE for six of the nine genes (CDKN2A, ID4, RBP1, RUNX3, SFRP1 and TMEFF2) but was not different to D-EAC for any gene. This suggests that methylation of these genes can occur early in the development of Barrett's metaplasia, and for only some of these genes does the extent of methylation increases during the progression to cancer.
In the technique we used to analyse methylation, bisulphite modified DNA is amplified using primers and conditions that do not discriminate between methylated and unmethylated sequences. Amplification is therefore independent of the presence or extent of methylation within the region being assessed. Unlike two common methods for measuring methylation, Methylation-Specific PCR [
7] or MethyLight [
8], the method we used will detect small amounts of methylation anywhere along the target sequence. Thus the methylation frequency we measure may differ from that reported by others. The methylation extent reflects a combination of the percentage of the alleles that are methylated, and the density of methylation on each allele. To determine if there was a relationship between the methylation measured in our assays and transcriptional silencing we measured gene expression in esophageal cell lines treated with the demethylating drug aza-dC. There was an increase in expression of ID4, MGMT, RBP1, RUNX3, SFRP1 and TIMP3, suggesting an association between methylation and silencing of expression for these genes. A relationship could not be determined for APC or CDKN2A, which were unmethylated, or TMEFF2, for which we could not measure any transcript in either cell line. For these three genes we have shown using cell lines established from other cancers that methylation of the regions we amplified are associated with gene silencing (unpublished observations). In our EAC tissues only methylation of MGMT was associated with a significant reduction in mRNA expression. The lack of a significant difference in gene expression for the other genes may be due to methylation of only one allele, or the presence of unmethylated cells in the tissue, such as clonal variants of the tumor, or stromal or infiltrating cells. Methylation in the cancer tissues was always greater than in the squamous epithelium of patients without cancer, so even if methylation did not result in gene silencing, it was still associated with cancer.
The processes involved in the transition from squamous epithelium to Barrett's metaplasia are unknown, although chronic gastro-esophageal reflux is widely believed to be the main trigger, with approximately 10% of gastro-esophageal reflux patients having BE [
9]. It is accepted that BE is a pre-malignant condition, which over time, as a result of the accumulation of multiple genetic and epigenetic modifications, can progress through dysplasia to EAC in a percentage of patients. A large number of such molecular changes have been reported in BE and EAC, although the sequence of these changes appear not to be as predictable as in some other cancers such as in the colon [
6].
It is not clear from the literature when aberrant DNA methylation occurs during the progression to EAC. There may be differences in molecular profile of BE tissue obtained from cancer resection specimens, compared to BE tissue obtained from patients with metaplasia only. Our samples of BE were obtained from patients with no detectable dysplasia or cancer, and were compared to cancer tissue from different patients. High levels of methylation in such BE specimens have been reported by others, such as methylation of SFRP1 being as frequent in BE (81 – 100% of samples) as it is in EAC (93 – 100%) [
10,
11]. In several studies high levels of methylation have been found in BE for many genes, but a few have reported low methylation in BE and then high in EAC tissues. Eads et al examined a single biopsy of metaplastic tissue from six patients with BE without associated EAC and reported methylation of APC, MGMT and TIMP3 in BE and also in cancers from other patients, but they did not compare the amounts of methylation in each [
12]. They also reported four genes, CDKN2A, ESR1 and MYOD1 and CALCA as unmethylated in these BE patients, and methylated in the cancers. Schulmann et al, using MethyLight, measured methylation in 93 biopsies of BE from an undisclosed number of patients [
13]. Methylation was as frequent in samples of EAC (n = 77) as in BE for APC, MGMT, RBP1 and TIMP3, but was significantly greater in EAC compared to BE for CDKN2A, RUNX3 and TMEFF2. Furthermore, the methylation level, their measure of the number of extensively methylated or hypermethylated molecules in the specimen, increased significantly from BE to dysplasia for these three genes. Clement reported that APC, TIMP3 and TERT were significantly more methylated in BE of patients who progressed to EAC compared to patients with BE which had not progressed during a follow-up of between four to ten years [
10]. Of the nine genes which we studied, only methylation of CDKN2A and RUNX3 appeared to increase in frequency in the progression from BE to cancer, and only six of the nine genes increased in the extent of methylation.
The underlying mechanisms that cause aberrant DNA methylation in Barrett's metaplasia and cancer are unknown. Our study confirms that aberrant methylation of multiple genes is an early event, and mostly occurs independently of dysplasia or EAC. Thus, we speculate that methylation of some genes occurs at or shortly after the transition from squamous to columnar metaplasia. To date there are no longitudinal studies investigating methylation changes during metaplasia.
This is the first report of ID4 methylation in BE or EAC. The other genes in this study have previously been reported to be methylated in esophageal disease. ID4 is a member of the inhibitor of DNA binding (ID) family of proteins that inhibit the binding of basic helix-loop helix transcription factors to DNA. It regulates the transcription of genes important in development and differentiation, and is a candidate tumor suppressor gene [
14,
15]. Methylation of ID4 has been reported in and associated with the silencing of gene transcription and loss of protein expression in lymphoma [
16], and gastric [
17], colorectal [
18] and breast carcinoma [
19,
20]. Methylation of ID4 correlated with increased risk of lymph node metastasis in T1 stage breast cancer [
20], and histopathological tumor grade and poorer prognosis in colorectal carcinoma [
18]. Our findings that ID4 is frequently methylated in BE and EAC but not in the normal squamous mucosa and that demethylation of cancer cell lines significantly increases expression, suggests loss of ID4 expression is important in the neoplastic progression of BE, supporting its role as a tumor suppressor.
DNA methylation is not the only abnormality reported in the genome of metaplastic Barrett's epithelium of patients without dysplasia or EAC. Using gene expression microarrays, Wang et al reported that the gene expression profile of BE more closely resembled EAC than esophageal squamous epithelium. Rather than BE being a benign tissue, they concluded that it was biologically closer to cancer than to normal squamous epithelium [
21]. Genomic loss, chromosomal gains and amplifications, mutations, and aneuploidy are observed in BE [
22,
23], and loss of heterozygosity of CDKN2A is reported in 47 – 75% of patients with BE, in the absence of dysplasia or EAC [
24‐
26]. In metaplastic epithelium from patients with dysplasia or EAC, 5q (APC), 13q (RB1), 17p (TP53), and 18q (DCC) are commonly lost, whilst 8q, 8p, and 6p are frequently gained [
27‐
29]. In contrast, mutations in BE are relatively uncommon. Mutations in CDKN2A have been reported in up to 7% of patients with metaplasia [
24,
30], but most reports suggest that mutations are absent in BE from patients without dysplasia or EAC [
31].
In this study of methylation in esophageal disease we have measured little methylation in any esophageal squamous epithelium, but in BE tissues there was significant methylation which for all but two of the nine genes examined did not increase in frequency in the progression from BE to EAC. We have also reported methylation of ID4 for the first time in BE or EAC. Together these findings confirm that BE is a precancerous tissue, and that aberrant promoter methylation occurs early in metaplasia before histological evidence of progression towards cancer, and that metaplastic BE is nearly as abnormal epigenetically as EAC.
Competing interests
The authors declare that they have no competing interests.