Esophageal adenocarcinoma (EAC) develops from Barrett’s esophagus (BE), wherein normal squamous epithelia is replaced by specialized intestinal metaplasia in response to chronic gastroesophageal acid reflux. BE can progress to low- and high-grade dysplasia, intramucosal, and invasive carcinoma. Both BE and EAC are characterized by loss of heterozygosity, aneuploidy, specific genetic mutations, and clonal diversity. Given the limitations of histopathology, genomic and epigenomic analyses may improve the precision of risk stratification. Assays to detect molecular alterations associated with neoplastic progression could be used to improve the pathologic assessment of BE/EAC and to select high-risk patients for more intensive surveillance.
Key points
- •
Genetic and epigenetic alterations play a central role in the formation of Barrett’s esophagus (BE) and esophageal adenocarcinoma (EAC).
- •
Global epigenetic alterations occur early in the BE to EAC sequence.
- •
Genomic analysis of EAC and BE has revealed a set of commonly altered genes that are likely drivers of cancer formation in the esophagus.
- •
There is considerable genetic and epigenetic heterogeneity in BE and EAC.
Introduction
Esophageal cancer can be separated into 2 major histotypes, esophageal adenocarcinoma (EAC) and esophageal squamous cell carcinoma, and is the eighth most common cancer worldwide. The incidence of EAC has been rising more rapidly than any other type of solid cancer in the United States for the past several decades, possibly secondary to the increasing prevalence of risk factors, such as obesity. EAC is a particularly lethal cancer, with 5-year survival rates of less than 20%.
EAC develops from Barrett’s esophagus (BE), intestinal metaplasia of the lower esophagus, which can then progress through low-grade dysplasia and high-grade dysplasia (HGD) to intramucosal carcinoma and then invasive carcinoma. Several concurrent histologic and molecular changes have been described for BE and EAC. The molecular changes observed include structural genomic alterations (amplifications and deletions, translocations), DNA sequence alterations (eg, missense mutations), and epigenetic modifications, primarily in the form of DNA hypermethylation and hypomethylation of CpG dinucleotides.
In light of the increased risk of EAC in those with BE, individuals diagnosed with BE are advised to undergo periodic endoscopic surveillance with biopsies of the affected segment to detect early histologic changes (ie, the presence of dysplasia) thought to confer risk for EAC development. However, because the overall risk of progression to EAC is minimal, a challenge when managing individuals with BE is to balance the risks and costs of endoscopic surveillance with the potential benefit of early identification or prevention of cancer. Assays for molecular alterations in BE samples might ultimately complement histologic, demographic, and/or endoscopic data and provide a more accurate prediction of an individual’s risk for dysplasia or cancer. This article summarizes the current understanding of genetic and epigenetic alterations that underpin the development of BE, dysplastic BE, and EAC, with an emphasis on global alterations observed in BE and EAC.
Introduction
Esophageal cancer can be separated into 2 major histotypes, esophageal adenocarcinoma (EAC) and esophageal squamous cell carcinoma, and is the eighth most common cancer worldwide. The incidence of EAC has been rising more rapidly than any other type of solid cancer in the United States for the past several decades, possibly secondary to the increasing prevalence of risk factors, such as obesity. EAC is a particularly lethal cancer, with 5-year survival rates of less than 20%.
EAC develops from Barrett’s esophagus (BE), intestinal metaplasia of the lower esophagus, which can then progress through low-grade dysplasia and high-grade dysplasia (HGD) to intramucosal carcinoma and then invasive carcinoma. Several concurrent histologic and molecular changes have been described for BE and EAC. The molecular changes observed include structural genomic alterations (amplifications and deletions, translocations), DNA sequence alterations (eg, missense mutations), and epigenetic modifications, primarily in the form of DNA hypermethylation and hypomethylation of CpG dinucleotides.
In light of the increased risk of EAC in those with BE, individuals diagnosed with BE are advised to undergo periodic endoscopic surveillance with biopsies of the affected segment to detect early histologic changes (ie, the presence of dysplasia) thought to confer risk for EAC development. However, because the overall risk of progression to EAC is minimal, a challenge when managing individuals with BE is to balance the risks and costs of endoscopic surveillance with the potential benefit of early identification or prevention of cancer. Assays for molecular alterations in BE samples might ultimately complement histologic, demographic, and/or endoscopic data and provide a more accurate prediction of an individual’s risk for dysplasia or cancer. This article summarizes the current understanding of genetic and epigenetic alterations that underpin the development of BE, dysplastic BE, and EAC, with an emphasis on global alterations observed in BE and EAC.
Genetic alterations in Barrett’s esophagus, Barrett’s esophagus with dysplasia, and esophageal adenocarcinoma
Somatic Genomic Alterations in Barrett’s Esophagus
The progression of BE to EAC provides a unique system to characterize the process by which a carcinoma emerges from its precursor state. Genomic studies of BE have revealed that it is not simply a metaplastic tissue; it also harbors frequent somatic alterations. The analysis of the process of BE progression has been greatly enhanced by dramatic improvements in genomic technologies, including tools to examine genetic mutations as well as larger structural alterations in cancer (and precancer) genomes.
Early studies of BE identified frequent loss of heterozygosity (LOH) at 17p, 5q, 9p, and 13q. The 17p and 9p harbor the tumor suppressors TP53 and CDKN2A , respectively, and studies have revealed frequent LOH through mutation ( TP53 and CKN2A ) or promoter methylation ( CDKN2A ). Galipeau and colleagues analyzed a series of esophageal biopsies from patients with BE and HGD without invasive EAC, finding patients commonly develop 9p LOH before the onset of 17p LOH. A 17p LOH was associated with genomic doubling to a 4N state, consistent with the impact of p53 loss upon genomic instability. When multiple biopsies from a single patient and time point were analyzed, 9p LOH was identified frequently in a greater percentage of the overall area of BE. These data contributed to the development of a popular model, where CDKN2A loss is thought to be an initiating event in BE progression, whereas TP53 alterations are later events, associated with neoplastic progression and aneuploidy.
Beyond aneuploidy, BE progression has been associated with increasing clonal diversity. Indeed, the presence of genomically distinct clones in the field of BE has been proposed by some researchers, with data suggesting the potential for certain clones to become dominant over time, that is, a ‘clonal sweep.’
One limitation of studies of populations with BE is that the vast majority of patients with BE do not progress to cancer, making the contribution of specific genomic alterations to the process of carcinogenesis less certain. More recent prospectively established collections have permitted researchers to study differences in structural genomic profiles in BE patients who did or did not progress to cancer. Li and colleagues studied serial BE biopsies using high-density single nucleotide polymorphism arrays. They identified chromosomal instability, genome doubling, and an increase in genetic diversity in BE samples taken within 48 months of EAC diagnosis compared with BE samples from nonprogressors. Interestingly, whereas the genomes in nonprogressors were relatively stable with fewer copy number changes, 9p ( CDKN2A ) loss was still identified. These results were consistent with the model of aneuploidy being associated with neoplastic progression, but were novel in demonstrating that aneuploidy was acquired just before the diagnosis of cancer.
The advent of next-generation massively parallel sequencing technologies has enabled systematic studies of the coding mutations in BE. Agrawal and colleagues performed whole exome sequencing on a set of EAC samples, including 2 cases of EAC with adjacent BE. They were able to identify the majority of mutations found in EAC in the paired BE tissue, confirming that EAC emerges from BE and showing that many coding mutations are already present in BE, including mutations in the tumor suppressor TP53 . Through sequencing of multiple biopsy samples of BE and EAC from the same patient, Streppel and colleagues identified loss of the tumor suppressor ARID1A in both BE and EAC. In a larger cohort of patients, this group identified loss of ARID1A in 4.9%, 14.3%, 16.0%, and 12.2% of BE, BE with low-grade dysplasia, BE with HGD, and EAC, respectively. In addition, by immunohistochemical staining, they identified abnormal nuclear accumulation of P53 in 34.1% of nondysplastic BE samples.
The most comprehensive, large-scale sequencing study in BE samples to date analyzed 26 genes (selected because they are commonly mutated in EAC) in a collection of nondysplastic BE, BE with HGD, and EAC samples. A striking result of this study was that, with the exception of TP53 and SMAD4 , the other genes did not show differential mutation rates between BE and EAC, even for bona fide tumor suppressors such as CDKN2A and ARID1A . Although only 2.5% of nondysplastic BE contained a mutation in TP53 , 70% of cases of HGD and EAC were TP53 mutant. Nondysplastic BE samples were chosen because they showed no signs of progression; thus, it is notable that they contained tumor suppressor inactivation. Because most of these patients likely never progress to cancer, it will be important to determine whether mutations in genes such as ARID1A in BE are markers for increased progression risk.
Somatic Genomic Alterations in Esophageal Adenocarcinoma
Modern genomics tools are being applied widely to the study of cancers, including EAC. The earliest efforts used genome-wide array platforms for copy number analysis and found a wide range of copy number disruptions in EAC. Nancarrow and colleagues identified frequent copy number alterations including homozygous deletions at putative fragile sites in the genome at genes such as FHIT and WWOX . Goh and colleagues used comparative genomic hybridization to confine regions of amplification to targets that included known oncogenes, such as MYC and EGFR . Their results also suggested that patients with highly aneuploid tumors have a poorer prognosis. However, other studies have not validated the relationship between aneuploidy and survival.
The resolution of array platforms has recently improved, as have the statistical tools with which to analyze copy number data to identify significantly recurrent alterations. Comparisons across tumor types have also shown that copy number patterns in EAC are strikingly similar to those in gastric cancers. The copy number study with the largest number of samples to date evaluated 186 EACs in conjunction with a large set of gastric and colorectal cancers. A key finding from this group was that the predilection for recurrent genomic amplifications was an important feature distinguishing EAC (and gastric cancer) from lower intestinal tumors. Rates of genomic deletion, by contrast, were not highly divergent between upper and lower gastrointestinal cancers. Statistical analysis demonstrated that amplifications were highly recurrent at the loci of a number of established oncogenes involved with cell signaling ( EGFR , ERBB2 , KRAS , MET , FGFR2 ), the cell cycle ( CCND1 , CDK6 and CCNE1 ), and transcription factors ( MYC , GATA4 and GATA6 ). Alterations in some of these oncogenes, including CDK6 and GATA6 , were validated in other studies. Many recurrent deletions were at loci of putative fragile site genes; thus, their pathologic significance is unclear. As in studies of BE, these data were consistent with a model where aneuploidy and oncogene activation seem to be important precursors for progression to cancer. Similarly, other analyses of these data established that whole-genome doubling is a prominent feature of EAC. Newer sequencing technologies that have now been used to characterize EAC have demonstrated relatively high somatic mutation rates compared with most other epithelial cancers. Agrawal and colleagues were the first to publish data focusing on exome sequencing of esophageal cancer. They demonstrated the occurrence of common TP53 mutations and the absence of Notch family mutations in esophageal squamous cell cancers. Dulak and colleagues performed whole exome sequencing on 149 EACs, along with whole genome sequencing of 15 of these EACs ( Fig. 1 ). Canonical oncogene mutations in genes such as KRAS and PIK3CA were uncommon, whereas evidence of oncogene amplification was frequent. By contrast, there were widespread mutations affecting tumor suppressor genes, including TP53 and CDKN2A and chromatin-modifying enzymes including ARID1A , SMARCA4 , and PBRM1 . Novel recurrent mutations, including those involving TLR4 and ELMO1 , were also noted, but their pathologic significance remains unclear.
Utilizing this large-scale sequencing, Dulak and colleagues were also able to evaluate mutation patterns, and found a predilection for A to C transversions at AA dinucleotides. The etiology of these mutations is unknown, but it has been hypothesized to be linked to bile acid exposure and the induction of oxidative DNA damage. This novel mutation signature was also observed in whole genome sequencing of esophageal cancers by other groups. The Nones group also performed additional structural analysis of whole genome data, finding that EACs commonly emerge after catastrophic genomic disruptive events termed chromotripsis. These recent genomic studies are consistent with the earlier BE studies in suggesting the significant role of acquisition of aneuploidy in the transition to EAC.
Alterations in MicroRNA Expression in Barrett’s Esophagus and Esophageal Adenocarcinoma
MicroRNAs (miRNAs) are small noncoding RNA molecules that can interact with other RNA molecules, resulting in posttranscriptional regulation of gene expression and gene silencing. Although most of the data regarding the role of miRNAs in esophageal cancer pertains to squamous cell carcinoma, there is evidence that miR-21 and miR-375 play a functional role in BE and EAC. Several studies have demonstrated that miR-21 is upregulated in BE and EAC compared with the normal esophagus. Feber and colleagues showed that miRNA expression profiles distinguished normal esophagus from EAC, and that miR-21 expression was 3- to 5-fold greater in EAC compared with normal epithelia. Meanwhile, another study that utilized microarray-based technology found 34 differentially expressed miRNAs between normal squamous epithelium and BE/EACs, although the miRNA profile did not reliably distinguish BE from EAC. In a validation cohort, the 5 miRNAs chosen for validation with quantitative reverse transcriptase polymerase chain reaction, including miR-21 , were successfully able to discriminate normal esophagus from BE/EAC.
There is also evidence that differential expression of miRNAs is associated with the progression of BE to EAC. Revilla-Nuin and colleagues recently identified 23 miRNAs involved in BE progression using miRNA sequencing analysis, finding 4 miRNAs ( miR-192 , miR-194 , miR-196a , and miR-196b ) had higher expression in BE patients who progressed to cancer compared with those who did not progress.
Epigenetic alterations in Barrett’s esophagus, Barrett’s esophagus with dysplasia, and esophageal adenocarcinoma
Epigenetics broadly refers to heritable and stable alterations in gene expression that are not mediated by changes in the DNA sequence. Since the discovery of DNA hypomethylation in colorectal cancer in 1982, epigenetic research has revealed an epigenetic landscape consisting of a complex array of epigenetic regulatory mechanisms that control gene expression in both cancer and normal tissue, where it plays a crucial role in embryonic development, imprinting, and tissue differentiation. The epigenetic landscape largely impacts the condensation state of the chromatin, determining whether the DNA is accessible to transcription factors and other proteins that control gene transcription. The epigenetic mechanisms currently believed to play a role in cancer include (1) DNA methylation of cytosine bases in CG-rich sequences, called CpG islands, (2) posttranslational modifications of histones, proteins that form the nucleosomes, which regulate packaging of DNA in chromatin, (3) miRNAs and noncoding RNAs, and (4) nucleosome positioning. In this review, we focus on aberrant DNA methylation because it is the most extensively studied epigenetic mechanism in BE and EAC. A number of excellent publications focusing on other classes of epigenetic alterations, such as histone modifications, have been written recently, and the interested reader is directed to those reviews.
DNA Methylation: An Overview
DNA methylation refers to the enzymatic addition of a methyl group to the 5-carbon position of the nucleotide cytosine by DNA methyltransferases (DNMT1, DNMT3a, or DNMT3b) to produce 5-methylcytosine, a normal base in DNA. Generally, the favored substrate for the DNMTs is the CG dinucleotide sequence, which has been termed CpG. The majority of CpGs are methylated in mammalian cells with unmethylated CpGs being typically present only in regions of DNA called CpG islands, genomic regions 200 to 500 bases in length with greater than 50% GC (guanine-cytosine) content and a ratio of observed-to-expected CpGs of greater than 0.6. CpG islands overlap the promoter region of 60% to 70% of genes and tend to be protected from methylation; however, they can become aberrantly methylated in cancer. CpG methylation can lead to transcriptional inactivation via multiple mechanisms, including directly inhibiting cis-binding elements, including the following transcription factors: AP-2, CREB, E2F, CBF, and NF-KB. Although this aberrant methylation is correlated traditionally with silencing of gene expression, it seems that decreased gene expression is characteristic of only a subset of methylated genes in most cancers. The methylation that occurs in CpG sites outside of promoter regions, termed gene body methylation, paradoxically has been correlated with transcriptional activation. Moreover, DNA hypomethylation seems to be a prominent epigenetic alteration in BE and EAC and has been associated with increased gene expression.
DNA methylation is a normal mechanism in the mammalian genome by which cells regulate gene expression, and gene methylation patterns that are established during embryonic development are maintained in the adult to regulate gene expression. A prominent mechanism by which DNA methylation is thought to regulate gene expression is through cooperative interactions with enzymes that regulate the chromatin structure, which can induce a compacted chromatin environment that represses gene expression. The interaction between DNA methylation, histone modification, and chromatin structure is complex, with abundant cross-talk. DNA methylation can impact chromatin structure, but the converse is also true. Because of the epigenetic cross-talk between DNA methylation and histone modification, aberrant DNA methylation can alter chromatin structure and gene expression, and dysregulation of histones and their modifying proteins may cause aberrant DNA methylation. There is a close association between methylated CpG islands and histones containing repressive posttranslational modifications.
Feinberg and colleagues have recently enhanced our understanding of global alterations of DNA methylation in cancer. They have proposed that in addition to CpG islands there are “CpG island shores,” areas of less dense CpG dinucleotides within 2 kilobases upstream of a CpG island, that can also show abnormal methylation in cancer. Methylation of CpG island shores is also associated with transcriptional inactivation and splicing alterations, tends to be tissue specific, and has been shown to be altered in colorectal cancer. Feinberg and colleagues observed that two-thirds of cancer-associated alterations in DNA methylation can be found in large domains, termed ‘large organized chromatin lysine modifications’ (LOCKs), as well as in smaller regions immediately adjacent to hypermethylated DNA. Their findings suggest a close cooperation between the chromatin state and DNA methylation changes in cancer.
Epigenetic Alterations in Barrett’s Esophagus and Esophageal Adenocarcinoma
Global alterations in DNA methylation in Barrett’s esophagus and esophageal adenocarcinoma
Microarray-based technologies have been used to interrogate global patterns of DNA methylation in BE and EAC, and to uncover candidate epigenetic drivers of BE progression. One study used Illumina HumanMethylation27 BeadChips to interrogate more than 27,000 CpG dinucleotides. The authors noted that both BE (n = 77) and EAC (n = 117) samples were highly methylated compared with normal esophagus (n = 94), indicating that epigenetic alterations occurs early in the BE to EAC progression sequence. They also found numerous previously undescribed hypermethylated genes in BE and EAC tissues, including genes encoding ADAM (A Disintegrin And Metalloproteinase) peptidase proteins, cadherins and protocadherins, and potassium voltage-gated channels. Alvi and colleagues also used the HumanMethylation27 BeadChips to compare methylation patterns, focusing on imprinted and X chromosome genes, from 24 BE and 22 EAC samples and validated their findings in retrospective and prospective cohorts to assess the ability of methylated genes to classify individuals as having prevalent BE, dysplastic BE, or EAC. They found 4 genes ( SLC22A18 , PIGR , GJA12 , and RIN2 ) had the greatest area under curve (0.988) to distinguish between BE and dysplasia/EAC in their retrospective cohort. In the prospective cohort, this methylated gene panel was able to stratify patients into low, intermediate, or high risk groups based on the number of genes that were methylated.
Kaz and colleagues utilized GoldenGate methylation microarrays (1505 CpGs in 807 genes) to compare methylation of normal squamous (n = 30), BE (n = 29), BE plus HGD (n = 8), and EAC (n = 30) cases. Distinct global methylation signatures were seen among the different tissue types, as well as specific genes demonstrating differential methylation between these groups. Within the BE and EAC cases, there were subgroups with distinct methylation signatures (high and low methylation epigenotypes), suggesting that there may be a CpG island methylator phenotype (CIMP) molecular class of BE and EAC ( Figs. 2 and 3 ). Further studies are needed to confirm this observation.
