• Fn14;
  • TWEAK;
  • esophageal adenocarcinoma;
  • microarray


  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Given the poor survival rate and efficacy of current therapy for esophageal adenocarcinoma (EAC), there is a need to identify and develop new therapeutic targets for treatment. Microarray analysis (Affymetrix U133A GeneChips, Robust Multi-Chip Analysis) was used to expression profile 11 normal squamous and 18 Barrett's esophagus biopsies, 7 surgically resected EACs and 3 EAC cell lines. Two hundred transcripts representing potential therapeutic targets were identified using the following criteria: significant overexpression in EAC by analysis of variance (p = 0.05, Benjamini Hochberg false discovery rate); 3-fold increase in EAC relative to normal and Barrett's esophagus and expression in at least 2 of the 3 EAC cell lines. From the list of potential targets we selected TNFRSF12A/Fn14/TWEAK receptor, a tumor necrosis factor super-family receptor, for further validation based on its reported role in tumor cell survival and potential as a target for therapy. Fn14 protein expression was confirmed in SEG-1 and BIC-1 cell lines, but Fn14 was not found to affect tumor cell survival after exposure to chemotherapeutics as expected. Instead, a novel role in EAC was discovered in transwell assays, in which modulating Fn14 expression affected tumor cell invasion. Fn14's potential as a therapeutic target was further supported by immunohistochemistry on a tissue microarray of patient samples that showed that Fn14 protein expression increased with disease progression in EAC. © 2007 Wiley-Liss, Inc.

Gastroesophageal reflux disease is a condition, in which gastric juice refluxes into the esophagus. The contents of gastric juice, including acid, pepsin and bile, can produce heartburn as a symptom and cause tissue injury to the smooth squamous epithelium lining the esophagus. In ∼ 4–10% of patients with gastroesophageal reflux symptoms the reflux-damaged squamous epithelium is replaced with an intestinal-type mucosa, leading to a condition called Barrett's esophagus (BE).1, 2 The precise origin of the metaplastic cells of BE remains unknown, although there is evidence supporting a stem cell origin first suggested by Gillen et al.3, 4 The esophageal metaplasia that occurs in BE is a premalignant lesion that can progress to dysplasia and esophageal adenocarcinoma (EAC). Estimates of the prevalence of BE in the general population and risk of progression to EAC vary: the prevalence of BE was recently reported to be 1.6%,5 while the estimated annual incidence of progression to EAC in BE patients has been estimated to be 0.4–1.8%.6 Although the risk of progression to EAC is relatively low, the incidence of EAC increased 350% in the United States from the early 1970s to the early 1990s,7 far outstripping increases in other cancers.8

The dramatic increase in EAC incidence is coupled with a poor prognosis. The main symptom of EAC is dysphagia, which often does not occur until there is significant blockage of the esophagus. As a result, the majority of patients that develop EAC present with advanced metastatic disease with an estimated 5 year survival rate of 13%.9 Surgical resection of the affected tissue is the primary treatment for EAC and improvements in surgical techniques have led to increased postoperative survival as well as 5-year survival.10 Chemotherapy with cisplatin and 5-flurouracil along with radiotherapy has been shown to improve survival11, 12 and new regimens incorporating agents, such as camptothecin, gemcitabine and paclitaxel have been tested.13, 14, 15, 16 Although survival has been improved by adding chemo- and radiotherapy to surgical resection, it appears that only a subset of patients benefit significantly; overall 5-year survival remains below 50% even in the best of circumstances.17

One way to improve the treatment and survival of EAC would be to develop novel therapies based on specific molecular targets. In an effort to identify and develop such novel therapeutic targets for treating EAC, we investigated the differences in gene expression between normal squamous tissue, BE and EAC. One gene with specific overexpression in EAC, TNFRSF12A/fibroblast growth factor-inducible 14 (Fn14)/TWEAK receptor, was selected for further confirmation and validation as a potential therapeutic target. Fn14 is a member of the tumor necrosis factor (TNF) super-family of structurally-related receptors18 that binds TNF-like weak inducer of apoptosis (TWEAK). Overexpression of Fn14, or stimulation by TWEAK has been reported to affect immune and inflammatory processes,19 branching morphology20 and resistance to chemotherapy.21 In the latter case, Fn14-mediatedresistance to chemotherapy was shown to be mediated by the nuclear factor-κB signaling pathway.22

Recently, it was reported that Fn14 mRNA expression was increased in EAC as well as tumor-associated BE relative to normal squamous tissue and nontumor associated BE; however, no functional role was demonstrated.23 In the present study, Fn14 mRNA expression is shown to correlate with protein expression in EAC tumor cells both in vivo and in vitro by immunohistochemistry and western blot and a novel functional role for Fn14 in EAC tumor cell invasion is demonstrated.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Tissue collection and cell culture

Normal and BE biopsy samples were obtained during routine endoscopy of previously diagnosed patients at the Veteran's Administration medical center in Tucson, AZ. Samples were immediately placed on dry ice, and RNA isolated within 2 hr. Samples of resected EAC were obtained from the tissue bank maintained by the gastrointestinal specialized program of research excellence at the Arizona Cancer Center. To protect patient privacy, patients were identified by a random 4 digit number. Patient information is available in Supplemental Table I at: All samples were obtained under protocols and procedures approved by the University of Arizona human subjects committee.

The BIC-1, SEG-1 and TE-7 EAC cell lines were maintained in Dulbecco's modified eagle medium and 5% fetal bovine serum supplemented with 50 μg/ml penicillin/streptozotocin (Invitrogen, Carlsbad, CA).

Microarray analysis

Samples were pulverized using microtube pestles (VWR, West Chester, PA) and total RNA isolated, using the Qiagen RNeasy minikit (Valencia, CA) according to manufacturer's instructions. The isolated total RNAs were used to produce labeled target, hybridized to Affymetrix U133A GeneChips, and read using the Agilent/Affymetrix 2500A scanner according to manufacturer's protocols. Raw data (CEL files) were summarized to transcript-level signal and normalized using the GC-RMA algorithm24 as implemented in GeneSpring v7.0 (Silicon Genetics, Redwood City, CA). Samples were labeled according to the tissue class, from which they were derived (N, normal; BE; EAC) and each transcript was normalized to its median. MIAME compliant files and raw data are available at the Gene Expression Omnibus website: Supplemental Tables of genelists and patient information are available as supplemental data at

Statistical analysis

Analysis of variance was performed using GeneSpring v7.0 with variances not assumed equal between sample groups (Welch's ANOVA). The global error model was used based on replicates of the 3 tissue types analyzed (normal, BE, EAC). Multiple testing was corrected using Benjamini and Hochberg False Discovery Rate set to 0.05—i.e., 5% of genes identified as significant were false positives. The Tukey post hoc test was used to evaluate where significant differences lay between the sample classes.

Power analysis

Power analysis was performed as described in Dobbin and Simon25 for single–label microarrays. Within class variance was calculated using all data from the normal samples. The median variance was 0.09, the 75th percentile was 0.21, and the 90th percentile was 0.46. The 90th percentile was used to provide a sample size estimate valid for 90% of the transcripts measured. Alpha was set at 0.05; power set to 0.90. Using the 90th percentile variance of 0.46, the number of independent replicate samples per class necessary to detect 3-fold differences between classes was ∼ 10. Since all classes of tissue analyzed contained 10 or more samples, a 3-fold cut-off for fold change was used to filter the data before clustering.

Visualization of gene expression profiles by K-means clustering

Transcripts expression profiles were clustered using K-means in GeneSpring 7.0. The parameters used: number of clusters 4, Number of iterations 100, Similarity Measure Pearson Correlation. An additional 25 random clusters were tested. Clustering converged after 14 iterations. Transcripts with increased signal in EAC relative to normal and BE samples were apparent in 1 of the 4 resulting clusters and were selected for further filtering.

Filtering for expression in cell lines

A transcript was considered to be expressed in the EAC cell lines if the raw signal after GC-RMA normalization was greater than 70 in 2 of the 3 cell lines. This signal strength cut-off was determined from a histogram of the raw signal data for all experiments, which showed that the 75th percentile of transcript signal across all measurements was 70.

siRNA transfection

siRNA oligonucleotides specific for GL2 Luciferase (5′ AAC GTA CGC GGA ATA CTT CGA TT) as a control, and Fn14 mRNA (region 475–495; 5′-AAG GAG ATT GGT GCT GTA AAA)26 and (region 2–22; 5′ GAG GGA GAA TTT ATT AAT AAA) were designed according to Elbashir et al.27 and purchased from Qiagen (Valencia, CA). Transient transfection of siRNA was carried out using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) as previously described.28

Adenovirus infection

Adenoviruses expressing Fn14 wild-type receptor or LacZ as a control for viral infection were infected at a multiplicity of infection of 5–20 as previously described.22

Western blot analysis

Western blotting was performed as described previously.22 Briefly, monolayer of cells were washed in phosphate-buffered saline containing 1 mM phenylmethylsulfonyl fluoride and then lysed in 2× sodium dodecyl sulfate sample buffer. Thirty micrograms of total cellular proteins were separated by 12% SDS-PAGE, and then transferred to nitrocellulose by electroblotting at 4C. The nitrocellulose membranes were blocked with 5% nonfat dry milk in Tris-buffered saline, pH 8.0, with 0.1% Tween 20 for 1 hr prior to addition of primary antibodies, followed by secondary horseradish peroxidase-conjugated anti-rabbit or -mouse IgG (Promega, Madison, WI). Alpha-tubulin primary antibody was obtained from Upstate Biotechnology (Lake Placid, NY) and Fn14 polyclonal sera was obtained from J. Winkles (University of Maryland).18 Protein bands were identified by chemiluminescence and exposed on X-Omat AR film (Eastman Kodak, Rochester, NY).

Invasion assay

Cell invasion was performed using modified Boyden chambers consisting of transwells with precoated Matrigel™ membrane filters inserted in 24-well tissue culture plates (BD Biosciences, Bedfold, MA). A monolayer of cells (75% confluence) was transfected with siRNA directed against Fn14 mRNA; controls were infected with siRNA directed against GL2 luciferase. In experiments on BIC-1, cells were infected with adenoviruses expressing wild-type human Fn14 or the LacZ gene as a control. Cells (2 × 105) were suspended in 100 μl of Dulbecco's modified Eagle's medium containing 1 mg/ml bovine serum albumin and 0.5% serum. After 16 hr, noninvading cells were removed by wiping the upper side of the membrane, and invading cells were fixed with methanol and stained with crystal violet. The number of invading cells was quantified by counting 3 random fields (total magnification, 200×) per filter. Three membrane filters were used for each condition within one experiment. Values represent the mean and 95% confidence interval (1.96 times standard error,29 of 3 independent experiments).


Tissue microarrays containing 99 samples of normal squamous, BE, dysplasia and EAC, as well as lymph and liver metastases were prepared as previously described.30 The tissue microarray was processed, using a Discovery XT and the Discovery DAB Map Detection kit (Ventana Medical Systems, Tucson, AZ) according to manufacturer's instructions. Antigen retrieval was performed with borate buffer at pH 8 followed by hydrogen peroxide and additional blocking steps according to manufacturer's instructions. Fn14 primary antibody (gift of Biogen IDEC, Cambridge, MA) was applied at 1:200 dilution for 60 min. Immunohistochemical images were captured using a Dmetrix automated optical slide scanner (Dmetrix, Tucson, AZ). To prevent bias in image processing image brightness and contrast were simultaneously adjusted automatically on all representative images shown.


  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Identification of tissue-specific gene expression

Twenty-nine biopsies, representing 11 normal and 18 Barrett's samples, along with 10 samples of EAC (seven surgical resections and 3 cell lines: SEG-1, BIC-1 and TE-7; Supplemental Table I —supplemental data is available at: were analyzed using the Affymetrix U133A GeneChip representing 22,283 human transcripts. ANOVA analysis was used to identify 892 transcripts with significant differences (p = 0.05 Benjamini Hochburg False Discovery Rate) in expression between normal and EAC, as well as BE and EAC. Power analysis using a variance estimate at the 90th percentile for all transcripts measured indicated that a limit of 3-fold change could be detected reliably in the study. Filtering the 892 significant transcripts found by ANOVA analysis for at least a 3-fold difference between EAC and the other 2 tissues reduced the list to 864 transcripts. To analyze the expression profiles of the 864 transcripts, they were classified using a combination of hierarchical and K-means clustering. Hierarchical clustering of the 864 transcripts indicated there were four major patterns of gene expression across the three tissues (Fig. 1a, the list of clustered transcripts is in Supplemental Table II). Clusters of genes appeared to be tissue specific in their expression profile; inspection of the clusters revealed transcripts previously reported as tissue or cell type-specific in the normal and BE tissue. Transcripts unique to the normal squamous samples included keratinocyte-specific genes: TGM1, a marker for keratinocyte terminal differentiation,31 Keratins 5 and 14 and MUC4, a tracehobronchial mucin not found in the EAC or Barrett's tissues.32 Transcripts with previously reported BE-specific expression, including: carbonic anhydrase,33 mucin 1334 and trefoil factor 3.35

thumbnail image

Figure 1. (a) Visualization of gene expression profiles. Hierarchical clustering of 864 transcripts identified major patterns in the expression profiles across the 3 tissues and was used to determine the number of clusters to use in K-means clustering. The 4 major sets of expression profiles shown in the subsequent K-means clustering (b) are identified numerically as 1–4. Clusters of the dendrogram that indicate tissue-specific expression are identified: Normal squamous (N), Barrett's esophagus (BE) and EAC with examples of tissue or cell type-specific transcripts mentioned in the text listed at right. The 864 transcripts clustered have an expression level 3-fold different in EAC relative to N and BE and are statistically significant. Coloring is based on relative expression with black representing no change; an increased relative expression is shown by red; and a decreased relative expression is shown by green. Each row denotes a single gene and each column represents a tissue as labeled. The dendrograms represent the correlation of the tissues and transcripts at top and left, respectively. The length of joining segments increases with decreasing Pearson correlation. (b) K-means clustering of 864 transcripts to visualize patterns in the expression profiles across the 3 tissue classes: normal squamous (N), Barrett's esophagus (BE) and EAC based on Pearson correlation. Each transcript is represented by 1 line and graphed to show relative expression on a log scale.

Download figure to PowerPoint

Based on the 4 major clusters of expression found by hierarchical clustering, the 864 transcripts were clustered using K-means into four groups to visualize gene expression patterns across the three tissues (Fig. 1b). The K-means clustering isolated four distinct patterns of gene expression across the three tissues: (i) transcripts down-regulated in EAC and BE relative to normal (Cluster 1, Fig. 1b); (ii) transcripts up-regulated in EAC relative to both normal and BE (Cluster 2, Fig. 1b); (iii) transcripts down-regulated in EAC relative to normal and BE (Cluster 3, Fig. 1b) and (iv) transcripts down-regulated in normal and EAC relative to BE (Cluster 4, Fig. 1b).

Identification of potential therapeutic targets

To identify potential therapeutic targets for the treatment of EAC, we focused on the 380 transcripts in Cluster 2 of Figure 1b that are up-regulated specifically in EAC relative to both normal and BE. Because 7 of the 10 EAC samples were surgically resected esophagus containing cell types from the entire thickness of the esophagus while the N and BE samples were biopsies, it seemed probable that the EAC samples contained additional cell types that would result in false positives during identification of EAC-specific transcripts. To overcome the mismatched sample composition between the biopsies and the resected surgical samples, the EAC-specific transcripts from Cluster 2 were further filtered for expression in at least 2 of the 3 EAC cell lines analyzed. The resulting list of 200 transcripts contained potential therapeutic targets, including MIG2,36 TCF8,37 IL8,38 PLAB39, 40 and Fn14 with previously reported involvement in tumor growth, invasion and drug resistance in other tumors (Supplemental Table III).

Fn14 protein expression and function in EAC cells

The EAC-specific transcripts identified in supplemental Table III are of interest for their potential to be developed as therapeutic targets. Several genes of particular interest were identified in the EAC-specific transcripts because of their secreted or cell-surface expression or prior association with progression in other gastrointestinal tumors. One such transcript encodes the protein Fn14 (TWEAK receptor) that appears to be multifunctional in a tissue-dependent manner with previous reports showing Fn14 involvement in angiogenesis,41, 42 migration43 and chemotherapy resistance.22 We investigated Fn14 expression and function in the EAC cell lines SEG-1 and BIC-1 as well as clinical samples. A summary of Fn14 expression in the microarray samples using a boxplot showed that Fn14 expression was significantly over-expressed in EAC relative to both normal squamous tissue and BE (Fig. 2a), while a western blot on control treated cells showed that the Fn14 protein was expressed in both cell lines as predicted by the microarray results (Fig. 2b). Additionally, Fn14 protein levels could be reduced by transfection with siRNAs directed against Fn14 mRNA; thereby establishing a model for investigating Fn14 function in EAC cells (Fig. 2b). Two independent siRNAs targeting different portions of Fn14 mRNA were tested in SEG-1 cells to ensure that the results were Fn14-specific.

thumbnail image

Figure 2. Fn14 mRNA and protein expression in microarray samples and EAC cell lines. (a) Boxplot of Fn14 (TNFRSF12a) RNA expression as measured by microarray analysis in the 3 tissues: N, normal squamous; BE, Barrett's esophagus; EAC, esophageal adenocarcinoma. The p-value shown is from the analysis of variance of normal squamous and Barrett's esophagus tissues compared to EAC (variances not assumed equal; Benjamini Hochburg false discovery rate used for multiple testing correction). (b) Western blot for Fn14 protein expression in SEG-1 and BIC-1 esophageal adenocarcinoma cell lines. Treatment lanes: ctrl, transfection with a siRNA against GL2 luciferase; Fn14 siRNA-475 and Fn14 siRNA-2, transfection with independent siRNAs against Fn14 that target different regions of the transript; top panels show Fn14 protein detection, while the lower panels show alpha-tubulin as a loading control.

Download figure to PowerPoint

Functional assays were performed to determine if Fn14 plays a role in tumor cell survival and invasion using the SEG-1 and BIC-1 cell lines transfected with siRNA directed against Fn14 mRNA or infected with adenovirus expressing Fn14. The first experiments tested whether cell survival after exposure to carboplatin and camptothecin was affected by a siRNA-mediated decrease in Fn14 expression. Results showed no change in cell survival in SEG-1 or BIC-1 cells treated with siRNA-475 against Fn14 relative to cells transfected with a control siRNA (data not shown). We next tested Fn14 function in tumor cell invasion using Matrigel™-coated transwells. Untreated SEG-1 cells were invasive, with an average of 716 cells per well invading through the Matrigel. In contrast, siRNAs targeting Fn14 mRNA reduced SEG-1 cell invasiveness significantly (p < 0.001, paired t-test, variances not assumed equal) to an average of 269 and 212 invading cells per well using siRNA-475 and siRNA-2, respectively (Fig. 3a). In contrast to SEG-1, untreated BIC-1 cells exhibited less invasiveness, with an average of only 30 cells invading the Matrigel. Nonetheless, siRNA targeting Fn14 mRNA reduced the number of invasive cells significantly to 20 (p < 0.01, paired t-test, variances not assumed equal). Because the number of invasive BIC-1 cells was low, we tested the effect of adenoviral overexpression of Fn14 in BIC-1 cells. Fn14 overexpression in BIC-1 cells significantly (p < 0.001, paired t-test, variances not assumed equal) increased the number of invasive cells from an average of 14 in the control infected with adenovirus expressing LacZ to an average of 44 invasive cells after infection with adenovirus expressing Fn14 (Fig. 3b).

thumbnail image

Figure 3. Fn14 plays a role in EAC tumor cell invasion. (a) Invasion assay with SEG-1 cells exhibiting significantly reduced invasion through Matrigel after transfection with 2 independent siRNAs directed against Fn14 mRNA (Fn14 siRNA-475 and Fn14 siRNA-2) or a control siRNA targeting GL2 luciferase. Error bars show 95% confidence intervals based on 3 independent experiments; the asterisk indicates significance at p < 0.001 in a paired t-test with variances assumed unequal. (b) Matrigel invasion assay with BIC-1 cells treated with control siRNA against GL2 luciferase (ctrl siRNA) or Fn14 (Fn14 siRNA-475). Error bars show 95% confidence intervals based on 3 independent experiments; the asterisk indicates p < 0.01 for a paired t-test between ctrl siRNA treated cells and Fn14 siRNA-475 cells. The 2 conditions at right show Matrigel invasion of BIC-1 cells infected with adenovirus overexpressing LacZ as a control (LacZ-Ad) or wild type Fn14 (wt Fn14-Ad); the double asterisk indicates p < 0.001 for a paired t-test between LacZ-Ad treated cells and wt Fn14-Ad treated cells.

Download figure to PowerPoint

Fn14 protein expression correlates with disease progression in patient samples

To confirm Fn14 protein expression in clinical samples immunohistochemistry was performed on a tissue microarray containing 99 samples of esophagus representing normal squamous, BE, low-grade and advanced dysplasia, EAC, lymph and liver metastases. Figure 4 shows representative images from the tissue microarray stained with antibody against Fn14, in which paired samples from 3 patients are shown. Fn14 protein expression in BE is negligible compared to EAC from the same patient (Figs. 4a and 4b); in another patient normal squamous tissue is shown to be Fn14 negative compared to matched EAC (Figs. 4c and 4d). The last 4 panels show lower Fn14 expression in 2 patients' dysplasia (Figs. 4e and 4g) as compared to their EAC (Figs. 4f and 4h).

thumbnail image

Figure 4. Paired representative images showing Fn14 protein expression increasing with disease progression in clinical samples. Immunohistochemical results from the tissue microarray scored in Figure 4 are shown at ×10 magnification. (a) Barrett's esophagus with negligible apparent Fn14 protein expression, (b) Fn14 esophageal adenocarcinoma from the same patient as Panel a. (c) normal squamous tissue with negligible Fn14 protein expression, (d) esophageal adenocarcinoma from the same patient as Panel c. (e) example of dysplasia with expression of Fn14 protein, (f) esophageal adenocarcinoma from the same patient in Panel e showing higher Fn14 expression than in the dysplastic tissue. (g) example of dysplasia with expression of Fn14 protein, (h) esophageal adenocarcinoma from the same patient as Panel g showing increased Fn14 expression.

Download figure to PowerPoint

Results from Figure 4 were semiquantitatively scored independently by 2 pathologists for strength of Fn14 expression and percent of cells of a given cell-type expressing Fn14. As predicted by the expression microarray results, Fn14 protein expression was low to negligible in normal squamous and BE samples and progressively increased in both the amount expressed and the percentage of positive cells in dysplastic tissue and EAC (Figs. 5a and 5b). Twelve percent of normal squamous cells and 21% of BE cells were positive for Fn14 expression with an average stain strength of 0.34 and 0.55, respectively. In contrast, 88% of EAC cells were positive for Fn14 expression with average stain strength of 2.2. Both scoring pathologists noted that the strongest Fn14 expression was in the invasive tumor cells of advanced EAC, consistent with the functional role for Fn14 in tumor cell invasion suggested by the invasion assays with SEG-1 and BIC-1. A comparison of the Fn14 stain strength between the tumor or dysplastic cells, and both the normal squamous and glandular BE cells, found a significant difference (p < 0.0001; 2-sided t-test with variances not assumed equal, Fig. 5a). Similarly, a significant difference (p < 0.0001; 2-sided t-test with variances not assumed equal) was found for the number of cells that stained positive for Fn14 across cell types (Fig. 5b). There was no significant difference in Fn14 stain strength or percent Fn14 positive cells when normal squamous was compared to glandular BE cells.

thumbnail image

Figure 5. Fn14 protein expression increases with disease progression in clinical samples from EAC patients. (a) Average strength of Fn14 protein stain by cell type as determined by 2 independent pathologists. Normal squa., normal squamous epithelial cell; BE, tumor-associated columnar cells of Barrett's esophagus; dysplasia, dysplastic cells arising from Barrett's esophagus; EAC, esophageal adenocarcinoma cells. Stain strength was assessed on a scale of 0–3; error bars represent 95% confidence interval; asterisks indicate p < 0.001 for the comparison of dysplastic or tumor cells to both normal and Barrett's esophagus in a 2-sided t-test with variances not assumed equal (b) Percent of cells positive for Fn14 expression by cell type as in Panel a.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Using microarrays we have determined gene expression profiles for normal esophagus, BE and EAC. Analysis of the genes differentiating between the 3 tissue types was complicated by heterogeneity between the samples, some of which were biopsies (the normal and BE samples) and some of which were surgical resections (the EAC samples). The additional cell types (e.g., smooth muscle) present in the resected EAC samples was expected since the resection contained all cell types present in the entire thickness of the resected portion of the esophagus. To identify EAC-specific transcripts, the comparison of resected EAC samples to the much shallower biopsies of normal and metaplastic tissue required an additional filter: expression in 2 of the 3 EAC cell lines analyzed. In addition to expression in EAC cell lines, the selected genes were statistically increased in EAC relative to normal and BE tissue at a level supported by the power of the experiment.

In several cases the transcripts identified as induced in EAC have been reported as potential biomarkers in other gastrointestinal cancers or to play a role in tumor progression. These genes of interest include MIG2, TCF8, IL8, PLAB and TNFRSF12a, which codes for Fn14 (Supplemental Table III). All of these genes show significant increases in expression in EAC when compared to both normal and Barrett's tissue and are expressed in all EAC cell lines analyzed. Previously identified as a serum induced gene, MIG2 is a component of cell-extracellular matrix adhesions and is involved in cell spreading.36 TCF8, also known as deltaEF1 or ZEB1, is a transcriptional regulator of growth and differentiation.37 IL8 cooperates with VEGF to induce tumor angiogenesis and increased levels of IL8 in drainage veins was associated with shorter survival times in a study of gastric cancer times.38 PLAB, also known as macrophage inhibiting-factor, is a member of the transforming growth factor-β super-family involved in tissue differentiation and maintenance. An association between serum levels of PLAB and neoplastic progression of colon cancer has been reported and a combination of serum levels of PLAB and CA19-9 showed diagnostic accuracy in pancreatic cancer.39, 40

Fn14 has been reported to play a role in both tumor invasion and resistance to chemotherapy. While Fn14 did not modulate sensitivity to chemotherapy in SEG-1 and BIC-1, reducing Fn14 expression with siRNA constructs significantly reduced invasion in both cell lines. Our transwell assay results agree with a previous report in glioma cells, which showed that Fn14 altered glioma tumor cell invasion and went further to show demonstrate that the effects were mediated by Fn14 signaling via nuclear factor-κB.26 In addition to reducing invasiveness with siRNAs targeting Fn14 mRNA, adenoviral overexpression of Fn14 in the poorly invasive BIC-1 cell line significantly increased the number of invasive cells over control cells infected with LacZ. Although the overexpression of Fn14 in BIC-1 increased the number of invasive cells, the total number of invasive cells remained considerably lower than in SEG-1. The difference in invasiveness between BIC-1 and SEG-1 suggests they differ in expression of proteins that modulate the invasive phenotype in addition to Fn14. Because BIC-1 was poorly invasive to begin with, the reduction in the number of invasive cells after treatment with siRNA targeting Fn14 was small, but significant nonetheless. In addition, Fn14 protein overexpression induced a significant increase in BIC-1 invasion. Taken together with the immunohistochemical results on patient samples which showed that the highest levels of Fn14 were in the most invasive tumor cells, these results suggest that Fn14 may play a role in EAC tumor invasion.

Another gene of particular interest is CYR61 that appears to be an EAC-specific gene expressed in the EAC resected samples and SEG-1, but not BIC-1. CYR61 is an integrin ligand whose forced expression in a gastric cancer cell line has been recently been shown to increase invasiveness; in addition Cyr61 expression was recently reported to be induced by low pH, which is associated with the gastric reflux thought to lead to EAC.44, 45 It is possible that the difference in expression of CYR61, and perhaps other proteins involved in invasion, between SEG-1 and BIC-1 may explain the difference in overall invasiveness reported here.

A recent publication by Wang et al. reported a comparison of normal, BE and EAC tissues.23 Of the 67 transcripts identified as overexpressed in EAC and Barrett's relative to normal tissue we were able to identify 66 matching probesets in our data. Of the 66 probesets, 29 overlapped with the 864 we identified as statistically significant and 3-fold changed in EAC relative to the other two tissues. There were no discrepancies in the induction or suppression of all 29 probesets that overlapped in the two studies. Six of the 29 overlapping probesets were increased in EAC relative to normal tissue and represented 4 distinct genes: COL4A2, SERPINH1, TCEAL1 and KIAA0657; the remaining 23 probesets were suppressed. COL4A2 and SERPINH1 passed the remaining filters used in our study to identify potential therapeutic targets and are found in the set of 200 genes on supplemental Table III. The report by Wang et al. also compared Barrett's tissue associated with progression to cancer to Barrett's tissue from patients that had not progressed. Of the resulting 13 genes reported to predict progression only Fn14 was identified by our filters as overexpressed in EAC. The remaining 12 markers of progression failed to pass our filters because they were suppressed in EAC, not 3-fold changed in EAC relative to normal and Barrett's tissue, or most importantly, did not have expression in the EAC cell lines. When expression in the cell line models of EAC was taken into account, only 3 of the 67 genes (COL4A2, SERPINH1 and TCEAL1) reported by Wang et al. met the criteria used to identify the genes listed in supplemental Table III. The lack of overlap between the 2 studies likely resulted from the studies' different objectives and thus different criteria for selection of differentially expressed genes.

Analysis of Fn14 expression in the present study found little to no expression in BE tissue in contrast to the tumor associated Barrett's tissue studied by Wang et al. A likely reason for this discrepancy was that the present study was not designed to identify markers for progression to EAC. Thus many, if not all, of the BE samples that were analyzed were likely not associated with progression to EAC. In the immunohistochemical analysis of Fn14 protein expression we did observe a slight increase in the percent of cells that scored positive for Fn14 expression in BE tissue from EAC patients (Fig. 5b); however, the confidence intervals overlapped with the normal squamous tissue.


  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We acknowledge the kind gift of the SEG-1 and BIC-1 EAC cell lines from Dr. David Beers, University of Michigan and the TE-7 cell line from Dr. Xiao-Chun Xu, University of Texas. This work was supported by the Arizona Cancer Center's GI SPORE (Gastroinstestinal Specialized Program of Research Excellence) and EAC tissues were obtained through the GI SPORE Tissue Core; both supported by NIH grant CA95060. Immunohistochemistry was performed by the Arizona Cancer Center's Tissue Acquisition Cellular/Molecular Analysis Shared Service supported by NCI grant CA023074-26. Microarray data was generated by the Arizona Cancer Center Genomics Shared Service, supported by NCI grant CA023074-26 and NIEHS grant ES06694. Construction of the tissue microarray was supported by funding from The Roy L. Jeannotte Memorial Foundation.


  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information
  • 1
    Veldhuyzen van Zanten SJ, Thomson AB, Barkun AN, Armstrong D, Chiba N, White RJ, Escobedo S, Sinclair P. The prevalence of Barrett's oesophagus in a cohort of 1040 Canadian primary care patients with uninvestigated dyspepsia undergoing prompt endoscopy. Aliment Pharmacol Ther 2006; 23: 5959.
  • 2
    Malfertheiner P, Lind T, Willich S, Vieth M, Jaspersen D, Labenz J, Meyer-Sabellek W, Junghard O, Stolte M. Prognostic influence of Barrett's oesophagus and Helicobacter pylori infection on healing of erosive gastro-oesophageal reflux disease (GORD) and symptom resolution in non-erosive GORD: report from the ProGORD study. Gut 2005; 54: 74651.
  • 3
    Gillen P, Keeling P, Byrne PJ, West AB, Hennessy TP. Experimental columnar metaplasia in the canine oesophagus. Br J Surg 1988; 75: 11315.
  • 4
    Seery JP. Stem cells of the oesophageal epithelium. J Cell Sci 2002; 115(Part 9): 17839.
  • 5
    Ronkainen J, Aro P, Storskrubb T, Johansson SE, Lind T, Bolling-Sternevald E, Vieth M, Stolte M, Talley NJ, Agreus L. Prevalence of Barrett's esophagus in the general population: an endoscopic study. Gastroenterology 2005; 129: 182531.
  • 6
    Shaheen NJ, Crosby MA, Bozymski EM, Sandler RS. Is there publication bias in the reporting of cancer risk in Barrett's esophagus? Gastroenterology 2000; 119: 3338.
  • 7
    Devesa SS, Blot WJ, Fraumeni JF,Jr. Changing patterns in the incidence of esophageal and gastric carcinoma in the United States. Cancer 1998; 83: 204953.
  • 8
    Pera M, Manterola C, Vidal O, Grande L. Epidemiology of esophageal adenocarcinoma. J Surg Oncol 2005; 92: 1519.
  • 9
    Eloubeidi MA, Mason AC, Desmond RA, El-Serag HB. Temporal trends (1973–1997) in survival of patients with esophageal adenocarcinoma in the United States: a glimmer of hope? Am J Gastroenterol 2003; 98: 162733.
  • 10
    Lerut T, Coosemans W, De Leyn P, Van Raemdonck D, Deneffe G, Decker G. Treatment of esophageal carcinoma. Chest 1999; 116 ( 6 Suppl): 463S465S.
  • 11
    Herskovic A, Martz K, al-Sarraf M, Leichman L, Brindle J, Vaitkevicius V, Cooper J, Byhardt R, Davis L, Emami B. Combined chemotherapy and radiotherapy compared with radiotherapy alone in patients with cancer of the esophagus. N Engl J Med 1992; 326: 15938.
  • 12
    Forastiere AA, Orringer MB, Perez-Tamayo C, Urba SG, Zahurak M. Preoperative chemoradiation followed by transhiatal esophagectomy for carcinoma of the esophagus: final report. J Clin Oncol 1993; 11: 111823.
  • 13
    Ilson DH, Saltz L, Enzinger P, Huang Y, Kornblith A, Gollub M, O'Reilly E, Schwartz G, DeGroff J, Gonzalez G, Kelsen DP. Phase II trial of weekly irinotecan plus cisplatin in advanced esophageal cancer. J Clin Oncol 1999; 17: 32705.
  • 14
    Sandler AB, Kindler HL, Einhorn LH, Mitchell E, Masters G, Kraut M, Nicol S, Raghavan D. Phase II trial of gemcitabine in patients with previously untreated metastatic cancer of the esophagus or gastroesophageal junction. Ann Oncol 2000; 11: 11614.
  • 15
    Wright CD, Wain JC, Lynch TJ, Choi NC, Grossbard ML, Carey RW, Moncure AC, Grillo HC, Mathisen DJ. Induction therapy for esophageal cancer with paclitaxel and hyperfractionated radiotherapy: a phase I and II study. J Thorac Cardiovasc Surg 1997; 114: 81115; discussion 816.
  • 16
    Williamson SK, McCoy SA, Gandara DR, Dakhil SR, Yost KJ, Paradelo JC, Atkins JN, Blanke CD, Abbruzzese JL. Phase II trial of gemcitabine plus irinotecan in patients with esophageal cancer: a Southwest Oncology Group (SWOG) trial. Am J Clin Oncol 2006; 29: 11622.
  • 17
    Mooney MM. Neoadjuvant and adjuvant chemotherapy for esophageal adenocarcinoma. J Surg Oncol 2005; 92: 2308.
  • 18
    Meighan-Mantha RL, Hsu DK, Guo Y, Brown SA, Feng SL, Peifley KA, Alberts GF, Copeland NG, Gilbert DJ, Jenkins NA, Richards CM, Winkles JA. The mitogen-inducible Fn14 gene encodes a type I transmembrane protein that modulates fibroblast adhesion and migration. J Biol Chem 1999; 274: 3316676.
  • 19
    Maecker H, Varfolomeev E, Kischkel F, Lawrence D, LeBlanc H, Lee W, Hurst S, Danilenko D, Li J, Filvaroff E, Yang B, Daniel D, et al. TWEAK attenuates the transition from innate to adaptive immunity. Cell 2005; 123: 93144.
  • 20
    Michaelson JS, Cho S, Browning B, Zheng TS, Lincecum JM, Wang MZ, Hsu YM, Burkly LC. Tweak induces mammary epithelial branching morphogenesis. Oncogene 2005; 24: 261324.
  • 21
    Winkles JA, Tran NL, Berens ME. TWEAK and Fn14: new molecular targets for cancer therapy? Cancer Lett 2006; 235: 1117.
  • 22
    Tran NL, McDonough WS, Savitch BA, Sawyer TF, Winkles JA, Berens ME. The tumor necrosis factor-like weak inducer of apoptosis (TWEAK)-fibroblast growth factor-inducible 14 (Fn14) signaling system regulates glioma cell survival via NFκB pathway activation and BCL-XL/BCL-W expression. J Biol Chem 2005; 280: 348392.
  • 23
    Wang S, Zhan M, Yin J, Abraham JM, Mori Y, Sato F, Xu Y, Olaru A, Berki AT, Li H, Schulmann K, Kan T, et al. Transcriptional profiling suggests that Barrett's metaplasia is an early intermediate stage in esophageal adenocarcinogenesis. Oncogene 2006; 25: 334656.
  • 24
    Wu Z, Irizarry RA. Preprocessing of oligonucleotide array data. Nat Biotechnol 2004; 22: 65658; author reply 658.
  • 25
    Dobbin K, Simon R. Sample size determination in microarray experiments for class comparison and prognostic classification. Biostatistics 2005; 6: 2738.
  • 26
    Tran NL, McDonough WS, Savitch BA, Fortin SP, Winkles JA, Symons M, Nakada M, Cunliffe HE, Hostetter G, Hoelzinger DB, Rennert JL, Michaelson JS, et al. Increased fibroblast growth factor-inducible 14 expression levels promote glioma cell invasion via Rac1 and nuclear factor-{κ}B and correlate with poor patient outcome. Cancer Res 2006; 66: 953542.
  • 27
    Elbashir SM, Harborth J, Lendeckel W, Yalcin A, Weber K, Tuschl T. Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 2001; 411: 4948.
  • 28
    Chuang YY, Tran NL, Rusk N, Nakada M, Berens ME, Symons M. Role of synaptojanin 2 in glioma cell migration and invasion. Cancer Res 2004; 64: 82715.
  • 29
    Streiner DL. Maintaining standards: differences between the standard deviation and standard error, and when to use each. Can J Psychiatry 1996; 41: 498502.
  • 30
    Hansel DE, Dhara S, Huang RC, Ashfaq R, Deasel M, Shimada Y, Bernstein HS, Harmon J, Brock M, Forastiere A, Washington MK, Maitra A, et al. CDC2/CDK1 expression in esophageal adenocarcinoma and precursor lesions serves as a diagnostic and cancer progression marker and potential novel drug target. Am J Surg Pathol 2005; 29: 3909.
  • 31
    Yamada K, Matsuki M, Morishima Y, Ueda E, Tabata K, Yasuno H, Suzuki M, Yamanishi K. Activation of the human transglutaminase 1 promoter in transgenic mice: terminal differentiation-specific expression of the TGM1-lacZ transgene in keratinized stratified squamous epithelia. Hum Mol Genet 1997; 6: 222331.
  • 32
    Guillem P, Billeret V, Buisine MP, Flejou JF, Lecomte-Houcke M, Degand P, Aubert JP, Triboulet JP, Porchet N. Mucin gene expression and cell differentiation in human normal, premalignant and malignant esophagus. Int J Cancer 2000; 88: 85661.
  • 33
    Pohler E, Craig AL, Cotton J, Lawrie L, Dillon JF, Ross P, Kernohan N, Hupp TR. The Barrett's antigen anterior gradient-2 silences the p53 transcriptional response to DNA damage. Mol Cell Proteomics 2004; 3: 53447.
  • 34
    Williams SJ, Wreschner DH, Tran M, Eyre HJ, Sutherland GR, McGuckin MA. Muc13, a novel human cell surface mucin expressed by epithelial and hemopoietic cells. J Biol Chem 2001; 276: 1832736.
  • 35
    Van De Bovenkamp JH, Korteland-Van Male AM, Warson C, Buller HA, Einerhand AW, Ectors NL, Dekker J. Gastric-type mucin and TFF-peptide expression in Barrett's oesophagus is disturbed during increased expression of MUC2. Histopathology 2003; 42: 55565.
  • 36
    Tu Y, Wu S, Shi X, Chen K, Wu C. Migfilin and Mig-2 link focal adhesions to filamin and the actin cytoskeleton and function in cell shape modulation. Cell 2003; 113: 3747.
  • 37
    Postigo AA, Depp JL, Taylor JJ, Kroll KL. Regulation of Smad signaling through a differential recruitment of coactivators and corepressors by ZEB proteins. EMBO J 2003; 22: 245362.
  • 38
    Konno H, Ohta M, Baba M, Suzuki S, Nakamura S. The role of circulating IL-8 and VEGF protein in the progression of gastric cancer. Cancer Sci 2003; 94: 73540.
  • 39
    Brown DA, Ward RL, Buckhaults P, Liu T, Romans KE, Hawkins NJ, Bauskin AR, Kinzler KW, Vogelstein B, Breit SN. MIC-1 serum level and genotype: associations with progress and prognosis of colorectal carcinoma. Clin Cancer Res 2003; 9: 264250.
  • 40
    Koopmann J, Buckhaults P, Brown DA, Zahurak ML, Sato N, Fukushima N, Sokoll LJ, Chan DW, Yeo CJ, Hruban RH, Breit SN, Kinzler KW, et al. Serum macrophage inhibitory cytokine 1 as a marker of pancreatic and other periampullary cancers. Clin Cancer Res 2004; 10: 238692.
  • 41
    Lynch CN, Wang YC, Lund JK, Chen YW, Leal JA, Wiley SR. TWEAK induces angiogenesis and proliferation of endothelial cells. J Biol Chem 1999; 274: 84559.
  • 42
    Ho DH, Vu H, Brown SA, Donohue PJ, Hanscom HN, Winkles JA. Soluble tumor necrosis factor-like weak inducer of apoptosis overexpression in HEK293 cells promotes tumor growth and angiogenesis in athymic nude mice. Cancer Res 2004; 64: 896872.
  • 43
    Tran NL, McDonough WS, Donohue PJ, Winkles JA, Berens TJ, Ross KR, Hoelzinger DB, Beaudry C, Coons SW, Berens ME. The human Fn14 receptor gene is up-regulated in migrating glioma cells in vitro and overexpressed in advanced glial tumors. Am J Pathol 2003; 162: 131321.
  • 44
    Lin MT, Zuon CY, Chang CC, Chen ST, Chen CP, Lin BR, Wang MY, Jeng YM, Chang KJ, Lee PH, Chen WJ, Kuo ML. Cyr61 induces gastric cancer cell motility/invasion via activation of the integrin/nuclear factor-κB/cyclooxygenase-2 signaling pathway. Clin Cancer Res 2005; 11: 580920.
  • 45
    Duggan SP, Gallagher WM, Fox EJ, Abdel-Latif MM, Reynolds JV, Kelleher D. Low pH induces co-ordinate regulation of gene expression in oesophageal cells. Carcinogenesis 2006; 27: 31927.

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

This article contains supplementary material available via the Internet at .

Suppl_Data_-_Table_S1_-_patient_information.doc61KSupporting Information file Suppl_Data_-_Table_S1_-_patient_information.doc
Suppl_Data_-_Table_S2_-_864_genes.doc1006KSupporting Information file Suppl_Data_-_Table_S2_-_864_genes.doc
Suppl_Data_-_Table_S3_-_200_genes.doc250KSupporting Information file Suppl_Data_-_Table_S3_-_200_genes.doc

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.