| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Human Cancer Biology |
Authors' Affiliation: Gastrointestinal Tumor Program, H. Lee Moffitt Cancer Center and Research Institute and Departments of Oncology, Medicine, Epidemiology, Pathology, and Surgery, University of South Florida, Tampa, Florida
Requests for reprints: James Helm, Gastrointestinal Tumor Program, H. Lee Moffitt Cancer Center and Research Institute, University of South Florida, 12902 Magnolia Drive, Tampa, FL 33612. Phone: 813-979-7257; Fax: 813-979-7229; E-mail: helmjf{at}moffitt.usf.edu.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Experimental Design: Microarray analysis was used to obtain individual gene expression profiles from endoscopic biopsies of nine esophageal adenocarcinomas and the Barrett's epithelia from which three of the cancers had arisen. Pooled samples from the Barrett's epithelia of six patients without cancer or dysplasia served as a reference.
Results: Barrett's epithelia from which cancer had arisen differed from the reference Barrett's epithelia primarily by underexpression of genes, many of which function in governing cell differentiation. These changes in gene expression were found even in those specimens of Barrett's epithelia from which cancer had arisen that lacked dysplasia. Each cancer differed from the Barrett's epithelium from which it had arisen primarily by an overexpression of genes, many of which were associated with tissue remodeling and invasiveness. Cancers without identifiable Barrett's epithelium differed from cancers that had arisen from a Barrett's epithelium by having an even greater number of these overexpressed genes.
Conclusions: Histologic progression from Barrett's epithelium to cancer is associated with a gradient of increasing changes in gene expression characterized by an early loss of gene function governing differentiation that begins before histologic change; gain in function of genes related to remodeling and invasiveness follows later. This correlation of histologic progression with increasing changes in gene expression suggests that gene expression changes in biopsies taken from Barrett's epithelium potentially could serve as a marker for neoplastic progression that could be used to predict risk for developing cancer.
Key Words: Esophageal adenocarcinoma • Barrett's esophagus • Tumor markers • Invasiveness • Differentiation • Gene expression profiling
Sequential progression from Barrett's epithelial metaplasia to cancer occurs over years, affording clinicians the opportunity to detect high-grade dysplasia or curable early-stage adenocarcinoma by endoscopic surveillance. However, use of dysplasia as a marker of malignant potential in Barrett's esophagus has shortcomings, including disagreement in histologic interpretation among pathologists and sampling error in biopsies (5, 6). For this reason, investigators have studied a large number of biological markers in search of an alternative to dysplasia as a marker of malignant potential (4, 7). Abnormalities in oncogenes, growth regulatory factors, and tumor suppressor genes have been described at different stages in the neoplastic progression of Barrett's esophagus (8). The complexity of genetic abnormalities may explain why no single genetic marker has been found to replace dysplasia as a marker of malignant potential.
The aim of this exploratory study was to use genome-wide expression profiling to characterize the broad changes in gene expression that underlie progression to esophageal adenocarcinoma from a Barrett's epithelium and assess the potential for using these gene expression changes as a marker predictive of malignant progression in Barrett's epithelium. We used microarray-based expression profiling to determine the changes in gene expression that occurred in esophageal adenocarcinoma and the Barrett's epithelium from which it arises. Genome-wide expression profiling is well suited for exploring the complexity of genetic abnormalities associated with the neoplastic progression of Barrett's metaplasia to cancer because of its ability to evaluate thousands of genes from multiple tissues simultaneously.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Biopsy specimens were examined by frozen section technique and microdissected to eliminate areas containing normal adjacent tissue, intervening stroma, and necrotic debris. Microdissection was done with a scalpel tip by a single pathologist (D.C.). Although it was not possible to eliminate all stroma and debris, microdissection of tumors by this method has been judged to yield tissue containing >90% tumor cells (9). For specimens from each patient, total RNA was prepared by Trizol (Invitrogen, Carlsbad, CA) extraction and the yield was measured. Integrity of extracted RNA was verified by both gel electrophoresis and the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). We found that 10 endoscopic biopsy specimens obtained from an esophageal adenocarcinoma and processed in this way typically would yield 50 to 100 µg total RNA, a quantity that far exceeded the minimum 5 µg needed for the subsequent microarray analysis. Samples were either pooled in equimolar amounts or used independently for GeneChip (Affymetrix, Santa Clara, CA) hybridization.
All experiments used the Human Hu U133A GeneChip, which contains 22,214 probe sets that survey expression level of >15,000 named genes based on the National Center for Biotechnology Information UniGene Build 133 (Affymetrix). Each chip was hybridized using targets synthesized from 5 µg of starting material (total RNA). Pooled samples were made by combining equal amounts of total RNA from each component tissue to obtain 5 µg total RNA for study (9). Target synthesis, hybridization, and posthybridization staining were done using standard Affymetrix protocols (10). Stained chips were scanned on an Affymetrix GeneArray Scanner. Scanned output files were visually inspected for hybridization artifacts and then analyzed by using Affymetrix Microarray suite software version 5.0. This software uses a statistical algorithm to assess increases or decreases in mRNA abundance in a direct comparison between two samples (11).
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Of >15,000 genes surveyed by microarray analysis, 1,097 different genes exhibited expression changes in cancers and the Barrett's epithelia from which cancer had arisen. To make this list, a gene expression change was required to have been found in each of the three Barrett's epithelia from which cancer had arisen, each of the three cancers arising from the Barrett's epithelia, or at least five of the six cancers without identifiable adjacent Barrett's epithelium. In nearly all cases, a gene expression change found in the Barrett's epithelia was also found in the cancers that had arisen from them. A complete list of differentially expressed genes may be obtained online (http://mriweb.moffitt.usf.edu/mpv/) or from the authors on request. In general, oncogenes, such as the cyclins A, D1, D2, D3, and E, were not overexpressed in cancer and the Barrett's epithelia from which cancer had arisen, and tumor suppressor genes, such as p53, and the cyclin-dependent kinase inhibitors p16, p21, and p27, were not underexpressed. The great majority of differentially expressed genes identified in this analysis do not exhibit functions that are usually considered to play a role in tumorigenesis.
We found few changes in expression of genes involved in regulation of the cell cycle whether in cancers or in the Barrett's epithelia from which cancer had arisen. Specifically, expression was not substantially or consistently different for genes governing cyclins and cyclin-dependent kinases, their inhibitors, or important substrates experimentally shown to have a major impact on control of cell growth. In the few instances in which a change was observed, the change was in a direction opposite to that expected. For example, expression of the gene cyclin D2, which promotes cell cycle progression, was observed to decrease an average of 1.8-fold in seven of nine cancers, whereas expression of genes governing the possible tumor suppressors ras-induced senescence 1 (RIS1) and hypermethylated in cancer 2 (HIC2) was increased 
4-fold. The one example of differential gene expression for a cell cycle regulator that might promote unregulated cell growth is CDC25B, which is a dual specificity phosphatase that activates the mitosis-promoting kinase CDC2 by dephosphorylation (12). Gene expression for CDC25B was increased in eight of nine cancers by 2-fold. Yet, this change may have been counterbalanced by an equal 2-fold increase in WEE1, which catalyzes the inhibitory phosphorylation of CDC2.
Other genes that might contribute to tumorigenesis exhibited similarly low-level changes in expression. Programmed cell death 4 and 10 (PDCD4 and PDCD10), genes induced during apoptosis, were down-regulated in all cancers, as were transcripts of caspase-6 and caspase-7, whereas baculoviral inhibitors of apoptosis repeat-containing 5 (BIRC5), also known as survivin, and SERPINB9, which can inhibit caspases, were increased. Changes in expression of genes involved in repair of DNA damage were found in cancers and the Barrett's epithelium from which cancer had arisen, but only XRCC4 was down-regulated. Loss of this protein, which is necessary for nonhomologous end joining and repair of double-strand breaks in DNA, can lead to genetic instability and further genetic changes (13). Expression of several other DNA repair genes was increased in cancers.
Although >1,000 genes were differentially expressed in cancers or the Barrett's epithelia from which cancer had arisen, the increase or decrease in any single transcript was usually relatively small, with the average change being <4-fold for most genes. The low magnitude of these changes led us to use hierarchical clustering analysis to look for patterns in the gene expression profiles that could be used to classify specimens into four groups: cancer without identifiable Barrett's epithelium, cancer arising from a Barrett's epithelium, Barrett's epithelium from which cancer had arisen, and the reference Barrett's epithelia without cancer or dysplasia. Clustering analysis using data from all probe sets on each microarray failed to clearly delineate these groups. Restriction of the analysis to only the 1,097 genes that displayed differential expression was more successful but still did not classify specimens into entirely distinct groups (Fig. 1). Cancers arising from a Barrett's epithelium tended to cluster together with the Barrett's epithelia from which they had arisen, whereas cancers without identifiable Barrett's epithelium tended to cluster away.
|
|
|
Transcripts for several proteins with a putative role in cellular defense against oxidative damage were reduced in Barrett's epithelia from which cancer had arisen. These genes included glutathione peroxidase (GPX2), selenoprotein P (SEPP1), and paraoxonase 2 and 3 (PON2 and PON3). Reduced expression of peroxiredoxin 2 (PRDX2) and thioredoxin (TXN) in some cancers could further impair cellular resistance to oxidative damage. Yet, some cancers also showed increased expression of superoxide dismutase 2 (SOD2), which could be protective.
Expression of genes involved in detoxification of xenobiotics was decreased in Barrett's epithelia from which cancer had arisen. Transcripts for several cytochrome P450s and glycosyltransferases were down-regulated. Expression of flavin-containing mono-oxygenase 5 (FMO5), arylacetamide deacetylase (AADAC), some ATP-binding cassette transporters associated with the export of xenobiotics, and several members of the aldo-keto reductase family was reduced as well. Decrease in these normal epithelial protective functions could increase cancer risk.
We unexpectedly found changes in expression of several autosomal genes ordinarily considered male specific. Several genes located on the Y chromosome were greatly decreased in both cancers and the Barrett's epithelia from which cancer had arisen: deleted in azoospermia 2, 3, and 4 (DAZ2, DAZ3, and DAZ4) and testes-specific protein, Y-linked (TSPY). Only four such genes were slightly up-regulated in the cancers: SOX9 (involved in early development of males), guanine nucleotide binding protein ß5 (GNB5), sperm-associated antigen 9 (SPAG9), and sperm adhesion molecule 1 (SPAM1).
We developed a list of candidate gene predictors of progression to cancer (Table 3). These genes were chosen from those that exhibited a change in expression in all three Barrett's epithelia from which cancer had arisen. Down-regulated genes were selected because of a sizeable change in expression that indicated virtually complete loss of function in most cases. Up-regulated genes were the ones most consistently and highly expressed. Furthermore, these genes exhibited the same expression changes in cancers as they did in the Barrett's epithelia.
|
Gene expression in cancer without identifiable Barrett's epithelium. The gene expression changes found in cancers without identifiable Barrett's epithelium were the same as those found in cancers arising from a Barrett's epithelium, with only a relatively small number of exceptions. The number of genes overexpressed in cancers without a Barrett's epithelium, however, was almost twice that found in cancers that had arisen from a Barrett's mucosa (Table 1). As was the case for cancers arising from a Barrett's epithelium, many of these overexpressed genes were associated with tissue remodeling and invasiveness. Expression of a few such genes, however, seemed to be genuinely increased in one group of cancers and decreased in the other. For example, genes for collagen type III 
1 (COL3A1) and matrix metalloproteinase 1 (MMP1) both exhibited decreased expression in cancers arising from a Barrett's epithelium and increased expression in cancers without evident Barrett's epithelium. Conversely, matrix metalloproteinase 7 (MMP7), -1-antichymotrypsin (SERPINA3), and fibroblast growth factor receptor 2 (FGFR2) were up-regulated in cancers arising from a Barrett's epithelium and down-regulated in the cancers without evident Barrett's epithelium. Because these genes govern extracellular tissue components, differences in their expression between groups of cancers may perhaps reflect a slightly different interaction of the cancers with the underlying tissue.
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
We did not find clear indication that differential expression of genes involved in cell cycle regulation contributes to cell proliferation in esophageal adenocarcinoma or the Barrett's epithelium from which it had arisen. One possible explanation for this observation may be that any changes in growth control had already occurred in the reference Barrett's epithelia from patients who had neither cancer nor dysplasia. Thus, no additional cell cycle dysregulation would be needed for progression from a benign Barrett's metaplasia to invasive adenocarcinoma. Indeed, differential expression of a large number of genes involved in G2-M functions and cell cycle check points has been reported in spontaneously arising 4N (G2-tetraploid) cell populations present in Barrett's epithelial cell cultures (17). These tetraploid cell populations have been found in several human neoplasms and are associated with malignant progression. Another possibility is that reduced cell death rate, rather than an increased birth rate, could allow cells to accumulate. Down-regulation of genes induced during apoptosis in cancers suggests a reduced rate of apoptosis, although a definitive answer will require further evaluation of the cell life cycle in cancers. One further possibility is that individual cancers developed unique pathways toward unregulated growth. Unique changes in individual cancers would not be identified by this analysis, which sought consensus changes.
Although a change in expression of 1,097 genes was found in cancers and the Barrett's epithelia from which cancers had arisen, most of these genes are not among those expected to be expressed differentially as a frankly invasive tumor evolves. Many of the changes in gene expression that we observed reflect a change in differentiation and complexity of the tissue as the tumor phenotype emerges. As an example of how such changes in expression might come about, consider that the development of any tissue and maintenance of its integrity are dependent on communications between two or more cell types (18–21). In epithelia, these communications can occur through receptors or ligands present on each cell type, through paracrine stimulation, or across gap junctions. As a neoplasm develops, it becomes less reliant on these forms of cell-to-cell communication for survival. Removing one partner from a communicating group of cell types leads to phenotypic changes in remaining cell types. Thus, loss of intercellular communication could explain some of the changes in gene expression that we observed.
Columnar Barrett's epithelium exhibits enterocytic differentiation as evidenced by its expression of proteins associated with intestinal mucosal function, such as sucrase-isomaltase, cytokeratins, and mucins (22–25). We found that in Barrett's epithelia from which cancer had arisen the predominant change in gene expression was down-regulation of differentiation-specific genes involved in gastrointestinal mucosal functions, such as transport, metabolism, and mucous secretion. On average, the decrease in expression of differentiation-specific genes was substantial; however, individual variability in expression was equally striking and of uncertain origin. The degree to which expression of enterocytic differentiation is lost may represent simple variation among individuals, but it also may reflect the extent of neoplastic progression in a given individual specimen, as expression of differentiation-specific genes tended to be less in cancers than in the Barrett's epithelia from which cancers had arisen. The various possibilities cannot be addressed by this study because of the limited sample size. A further issue to be determined is the extent to which changes in expression of differentiation-specific genes may be secondary consequences rather than directly attributable to neoplastic transformation itself.
Barrett's epithelia from which cancer had arisen exhibited a decrease in expression of genes encoding proteins with putative roles in cellular defense against oxidative damage. Transcripts for proteins involved in the detoxification of xenobiotics were reduced in Barrett's epithelia from which cancer had arisen as well, indicating a decreased tolerance for various forms of chemical insult. Reduced protection against outside insults would place the epithelium at risk for further genetic damage and therefore may represent an early event in the progression to cancer. In addition to the possibility of impaired defenses against DNA damage, there is some evidence for a diminished capacity to repair DNA damage in both cancers and the Barrett's epithelium from which the cancer has arisen.
Inflammation is commonly present in esophageal adenocarcinoma as well as in Barrett's epithelium. Accordingly, changes in expression of genes associated with immune function and inflammatory response are to be expected in both cancer and the Barrett's epithelium from which it arises. Transcripts for proteins that promote or accompany inflammatory response were increased in both tissues. Decreased transcript for the IgG-binding protein Fc
BP suggests increased susceptibility to infection, but whether this finding is of any significance in the neoplastic transformation of Barrett's epithelium is unknown.
Surprisingly, expression of several genes located on the Y chromosome was evident in the reference metaplastic Barrett's epithelia but substantially decreased or virtually undetectable in both cancers and the Barrett's epithelia from which cancer had arisen. Very little is known about these genes; however, deletion of any one of the genes results in azoospermia (26). This is not the first example of Y-linked gene expression in a tissue that has no role in sexual reproduction (27) or the first indication that progression of tumors of the esophagus might be influenced by changes on the Y chromosome, which is often entirely lost in cancer (28, 29).
The observation that gene underexpression was the predominant change in Barrett's epithelia from which cancer had arisen suggests importance of a loss of function, rather than gain, in initiating the neoplastic progression from metaplasia to cancer. Many of the underexpressed genes govern differentiation, which is compatible with the model for development of cancer by regression to a dedifferentiated state. In contrast, gene expression changes with progression to invasive cancer reflect a distinct gain in function. That is, gene overexpression accounted for most changes in cancers individually compared with the Barrett's epithelium from which each had arisen. Many overexpressed genes were associated with tissue remodeling and invasiveness, which correlates with the more aggressive phenotype of invasive cancer.
Adenocarcinoma arises from an endoscopically identifiable Barrett's epithelium in 30% to 85% of patients (30, 31). The remaining cancers may have overgrown the Barrett's epithelium from which they had arisen; however, it could be argued that adenocarcinoma may have arisen de novo without a precursor Barrett's epithelium. We first thought that if these two groups of cancers had different origins, then gene expression might differ as well. Indeed, gene expression in cancers without an identifiable Barrett's epithelium differed from those arising from a Barrett's mucosa by having nearly twice the number of overexpressed genes, many of which were involved in governing extracellular tissue components related to tissue remodeling. It is difficult to know, however, if this difference in gene expression between the two groups of cancers was the result of the tumors having distinctly different origins or whether it reflects only differences in growth properties of the two groups. Greater overexpression of genes associated with tissue remodeling and invasiveness in cancers without identifiable Barrett's epithelium could have led to a more aggressive cancer that overgrew the Barrett's mucosa.
It is remarkable that gene expression changes were already present in two of the three Barrett's epithelia from which cancer had arisen, although they were histologically indistinguishable from the reference Barrett's epithelia of patients who had neither cancer nor dysplasia. The high-grade dysplasia of the third Barrett's epithelium exhibited a greater number of gene expression changes that more closely resembled cancer than the two epithelia that were histologically identical to the reference Barrett's epithelia. Although the number of patients is small, these findings are compatible with the hypothesis that the histologic progression from Barrett's metaplasia to cancer through grades of dysplasia is associated with an underlying gradient of increasing changes in gene expression that begin before the development of dysplasia (Fig. 2). This correlation of histologic progression with increasing changes in gene expression suggests that gene expression changes in biopsies taken from Barrett's epithelium could serve as a marker for neoplastic progression that could be used to predict which individuals are at greatest risk for developing cancer. To gain an understanding of the findings that might be expected in a larger study population, we developed a list of candidate gene predictors. Not surprisingly, this list is dominated by genes associated with epithelial differentiation, with a smaller number of genes linked to tissue remodeling and growth.
|
|
Acknowledgments |
|---|
|
Footnotes |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received 1/ 7/04; revised 1/ 1/05; accepted 1/ 6/05.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Feber, L. Xi, J. D. Luketich, A. Pennathur, R. J. Landreneau, M. Wu, S. J. Swanson, T. E. Godfrey, and V. R. Litle MicroRNA expression profiles of esophageal cancer. J. Thorac. Cardiovasc. Surg., February 1, 2008; 135(2): 255 - 260. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Zagorowicz and J. Jankowski Molecular changes in the progression of Barrett's oesophagus Postgrad. Med. J., August 1, 2007; 83(982): 529 - 535. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Garg, D. Chaurasiya, R. Rana, N. Jagadish, D. Kanojia, N. Dudha, N. Kamran, S. Salhan, A. Bhatnagar, S. Suri, et al. Sperm-Associated Antigen 9, a Novel Cancer Testis Antigen, Is a Potential Target for Immunotherapy in Epithelial Ovarian Cancer Clin. Cancer Res., March 1, 2007; 13(5): 1421 - 1428. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R C Fitzgerald Molecular basis of Barrett's oesophagus and oesophageal adenocarcinoma. Gut, December 1, 2006; 55(12): 1810 - 1820. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
