Clinical Cancer Research CTRC-AACR San Antonio Breast Cancer Symposium Infection and Cancer: Biology, Therapeutics, and Prevention
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 Cell Growth & Differentiation
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Mitas, M.
Right arrow Articles by Hoffman, B. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Mitas, M.
Right arrow Articles by Hoffman, B. J.
Clinical Cancer Research Vol. 11, 2205-2214, March 2005
© 2005 American Association for Cancer Research

Imaging, Diagnosis, Prognosis

Accurate Discrimination of Barrett's Esophagus and Esophageal Adenocarcinoma Using a Quantitative Three-Tiered Algorithm and Multimarker Real-time Reverse Transcription-PCR

Michael Mitas1, Jonas S. Almeida2, Kaidi Mikhitarian1, William E. Gillanders1, David N. Lewin3, Demetri D. Spyropoulos3, Loretta Hoover1, Amanda Graham1, Tammy Glenn4, Peter King4, David J. Cole1, Robert Hawes4, Carolyn E. Reed1 and Brenda J. Hoffman4

Departments of 1 Surgery, 2 Biometry and Epidemiology, 3 Pathology and Laboratory Medicine, and 4 Digestive Disease Center, Medical University of South Carolina, Charleston, South Carolina

Requests for reprints: Michael Mitas, Department of Surgery, Medical University of South Carolina, Suite 420K, 96 Jonathan Lucas Street, Charleston, SC 29425. Phone: 843-792-1451; Fax: 843-792-3315; E-mail: mitasm{at}musc.edu.


   ABSTRACT
Top
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Esophageal adenocarcinoma (EA) is increasing faster than any other cancer in the U.S. In this report, we first show that EA can be distinguished from normal esophagus (NE) and esophageal squamous cell carcinoma by plotting expression values for EpCam, TFF1, and SBEM in three-dimensional Euclidean space. For monitoring progression of Barrett's esophagus (BE) to EA, we developed a highly sensitive assay for limited quantities of tissue whereby 50 ng of RNA are first converted to cDNA using 16 gene-specific primers. Using a set of training tissues, we developed a novel quantitative three-tiered algorithm that allows for accurate (overall accuracy = 61/63, 97%) discrimination of BE versus EA tissues using only three genes. The gene used in the first tier of the algorithm is TSPAN: samples not diagnosed as BE or EA by TSPAN in the first tier are then subjected to a second-tier analysis using ECGF1, followed by a third-tier analysis using SPARC. Addition of TFF1 and SBEM to the first tier (i.e., a five-gene marker panel) increases the overall accuracy of the assay to 98% (62/63) and results in mean molecular diagnostic scores (± SD) that are significantly different between EA and BE samples (3.19 ± 1.07 versus –2.74 ± 1.73, respectively). Our results suggest that relatively few genes can be used to monitor progression of BE to EA.

Key Words: Gene overexpression • ROC curve analysis • diagnostic accuracy • Epithelial cell adhesion molecule (EpCam) • Trefoil factor 1 (TFF1)


   INTRODUCTION
Top
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
From 1973 to 1994, the incidence of esophageal adenocarcinoma (EA) increased by more than 100% among Caucasian men and by about 50% among Caucasian women (1). The reason for this increase is not known. At least 80% of EA originate in the lower third of the esophagus and frequently occur in association with Barrett's esophagus (BE), a condition wherein the squamous epithelium of the distal esophagus is replaced by columnar epithelium. Furthermore, this replacement process seems to preferentially occur in Caucasian males >40 years of age. The development of EA typically follows the sequence: BE -> low-grade dysplasia -> high-grade dysplasia -> EA (2). Compared to the normal population, BE is associated with a 30- to 52-fold increase in the occurrence of EA (3). This malignancy is normally diagnosed at symptomatic and advanced stages, and is associated with poor survival (4).

Pathologic evaluation remains the foundation of clinical decision making in the evaluation of risk of progression in BE (5). The current standard of care for patients who are diagnosed with BE and no dysplasia is screening by endoscopic biopsy every 2 to 3 years, whereas those with low-grade dysplasia or who have an indefinite diagnosis for dysplasia are screened every 6 months to a year. Patients with high-grade dysplasia within BE are at the highest risk for development of EA. For example, patients with an initial diagnosis of low-grade dysplasia have a median progression-free survival of 60 months, whereas the period for those with high-grade dysplasia is reduced to 8 months (6). Given the high likelihood of metastatic disease and poor prognosis associated with invasive cancer, detection of high-grade dysplasia within BE is considered by many as the final end point requiring definitive therapy in the form of surgical resection (7). However, because esophageal resection done on patients with high-grade dysplasia reveals invasive cancer in only 30% to 50% of the specimens (8), controversy still remains as to whether all high-grade dysplasia patients should undergo resection.

In contrast to EA, the incidence of esophageal squamous cell carcinoma (ESCC) is four to five times higher in African-Americans than Caucasians (1). Furthermore, men are more commonly affected by ESCC than women at a ratio of 3 or 4:1 (9). In the U.S., the two cities with the highest rates of ESCC are Washington, DC and Charleston, SC (10), a fact that has prompted our interest in developing screening programs for the early detection of all esophageal malignancies. The identification of molecular markers with high diagnostic or prognostic values with respect to esophageal malignancies would be of great clinical value. Toward this ultimate goal, we measured, using real-time reverse transcription-PCR, the expression levels of cancer-associated genes in various esophageal tissues, including BE, EA, and ESCC. Among the genes studied, we were able to identify several that may be useful for early detection of esophageal cancer.


   MATERIALS AND METHODS
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Human Subjects. Informed consent approved by the Medical University of South Carolina Institutional Review Board was obtained from all patients enrolled in this study. Each study subject was treated in accordance with the guidelines put forth in the assurances filed with and approved by the Department of Health and Human Services, which establish policies and procedures to protect human subjects in research programs. The study population consisted of 53 cancer patients (35 EA, 18 ESCC), 34 patients with BE, and 22 normal controls aged 18 years or older.

Cancer Patients. For the majority of cases, biopsy tissues were obtained from patients who were diagnosed with esophageal cancer at a local institution and referred to the Digestive Disease Center at the Medical University of South Carolina for additional esophageal cancer staging. Mucosal biopsies of the cancers were obtained for molecular analysis from visible tumor with forceps through the working channel of an upper endoscope (GIF 160, Olympus America, Melville, NY). All biopsies were adjacent to those used for diagnosis by H&E staining.

Patients with Barrett's Esophagus. Biopsy tissues were obtained from patients undergoing surveillance endoscopy at the Digestive Disease Center, Medical University of South Carolina. Portions of each sample were examined by H&E staining and confirmed as Barrett's tissue. The remaining portions of the samples were either snap-frozen in liquid nitrogen by the Hollings Cancer Center Tissue Repository, or stored in paraffin.

Control Patients. Normal control samples were obtained from patients being seen at the Digestive Disease Center for evaluation of non–cancer related issues. Mucosal biopsies were obtained as described above at a location ~2 cm proximal to the squamocolumnar junction and/or 2 cm distal to the upper esophageal sphincter. All cancer and normal (control) mucosal biopsies were placed in test tubes containing ~1 mL saline and then packed in ice. RNA was typically extracted from the samples within 2 hours of initial procurement.

RNA Isolation from Endoscopic Biopsies and cDNA Synthesis. Total cellular RNA was isolated from pelleted biopsy material using a guanidinum thiocyanate-phenol-chloroform solution (RNA Stat-60; Tel-Test, Friendswood, TX). Briefly, pelleted biopsy specimens were resuspended in 1 mL of RNA Stat-60. Pellets larger than ~2 mm in diameter were homogenized using a model 395 type 5 polytron (Dremel, Racine, WI). Total RNA was isolated according to the manufacturer's instructions with the exception that 1 µL of a 50 mg/mL solution of glycogen (Sigma, St. Louis, MO) was added to the aqueous phase prior to addition of isopropanol. The final RNA pellet was dissolved in 50 µL of 1x RNA secure buffer (Ambion, Austin, TX). RNA was quantified by UV absorbance at 260 nm. As described in the text, cDNA was made from 200 units of Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI) and either (a) 5 µg of total RNA and oligo d(T)12-16, or (b) 50 ng of RNA and 32 µmol/L each of 16 gene-specific primers (i.e., reverse primers used for PCR: ß2m + ßActin + TSPAN + TFF1 + PDEF + SHh + EpCam + SPARC + ECFG1 + SBEM + PITX1 + PTCH + SMO + Muc4 + Gli1 + POTE).

Real-time Reverse Transcription-PCR. Real-time reverse transcription-PCR analyses were done on a PE Biosystems Gene Amp 5700 Sequence Detection System (Foster City, CA). The standard reaction volume was 10 µL and contained 1x QuantiTect SYBR Green PCR Master Mix (Qiagen, Valencia CA), 0.1 units AmpErase UNG enzyme (PE Biosystems), 0.7 µL cDNA template, and 0.25 µmol/L of both forward and reverse primer (Table 1). The initial step of PCR was 2 minutes at 50°C for AmpErase UNG activation, followed by a 15 minute hold at 95°C. Cycles (n = 40) consisted of a 15-second denaturation step at 95°C, followed by a 1 minute annealing/extension step at 60°C. The final step was a 60°C incubation for 1 minute. All reactions were done in triplicate and a negative control lacking cDNA was included.


View this table:
[in this window]
[in a new window]
  		Table 1 Primers for real-time PCR

 
Sequence Verification of Amplified Fragments. Sequences of muc1, CK19, erbB2, PDEF, and CEA gene fragments have been previously verified (11–13). For the remaining gene fragments analyzed in this study, real-time reverse transcription-PCR products were obtained using cDNA derived from various esophageal tissues. Reaction conditions were as described above. To purify amplified fragments from primers, products (100 µL) were applied to a silica gel membrane (MinElute DNA Cleanup System, Qiagen), eluted, and sequenced using an ABI 373 instrument (Foster City, CA). Each sequencing reaction (10 µL) contained 2 µL of half-term dye terminator (Sigma), 2 µL terminator ready reaction mix (ABI), 0.2 to 2 µL purified fragment, and 0.32 pmol of the forward primer listed in Table 1 was used to generate the original real-time reverse transcription-PCR product. Sequencing reactions consisted of 30 cycles at 94°C for 15 seconds, 55°C for 20 seconds, and 62°C for 4 minutes. Sequences of all fragments matched those reported for the respective Genbank accession number listed in Table 1 (data not shown).

Receiver Operator Characteristic Curve Analysis. Areas under the ROC curves (AUC) with 95% confidence intervals (CI) were obtained using MedCalc Software (Mariakerke, Belgium).

Selection of Genes for Discrimination of Normal Esophagus versus Barrett's Esophagus versus Esophageal Adenocarcinoma Tissues. Discrimination of esophageal tissues was quantified using single linkage (the shortest distance between members of distinct groups) analysis of all potential three-marker combinations. Distances were determined from Euclidean spatial coordinates obtained using {Delta}Ct values of a given gene. The implementation of the selection procedure was done in Matlab 7 R12 (MathWorks, Inc., Natick, MA) in a dual processor machine (2.6 GHz, 6 Gb RAM) under the Linux operating system. The web-based application described in the text uses this same infrastructure and the same Matlab library to produce predictions through Matlab's web server toolbox. The system makes use of a proprietary common gateway interface application for new values of selected genes.

Boundaries for Molecular Classification. Molecular classification analysis was done using a Matlab 6.5 (R13) programming environment (MathWorks). The boundaries for normal esophagus (NE) versus EA versus ESCC, or NE versus BE versus EA were set corresponding to equidistant surfaces such that all points in the surface are equidistant to the closest group member and closest non–group member. The procedure relies on computationally efficient matrix operations as described in refs (14, 15), and enables a boundary surface to be calculated and displayed every time a new position is submitted to the web-based tool. The equidistant boundary is displayed mostly for the purpose of visualization. The actual decision on group membership is based on the proportion of group members closer to the position being considered (position in this context refers to the coordinates of the three-gene marker set). This proportion is reported for n–1 members of each group, where the closest member is deducted to conservatively account for the loss of 1 df. The off-line code is also available for estimation of probabilities for each proportion, through the use of assumption-free neural network classifiers which infer density distributions directly from the experimental data (14, 15).


   RESULTS
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
To determine whether a molecular marker approach might be useful for analysis of esophageal malignancies and/or precursor lesions, we first measured the expression levels of two panels of genes. The first panel consisted of two genes (TFF1 and TFF3) encoding secretory peptides expressed in BE samples (16) and thought to play roles in epithelial repair of the gastric mucosa (17); a gene encoding a cysteine-rich protein [secreted protein acidic and rich in cysteine (SPARC)], which is up-regulated in ESCC (18) and BE/EA tissues (19); two members of the mucin gene family (muc1 and muc4); a gene [Sonic hedgehog (SHh)] involved in digestive tract morphogenesis/tumorigenesis (20, 21), and oral squamous cell carcinoma (22); and CK19, a gene overexpressed in a variety of cancers (11). The second panel consisted of two genes encoding pancarcinoma surface antigens [EpCam (23, 24) and CEA (25)]; a gene encoding an Ets transcription factor (PDEF; ref. 26) of largely unknown function that is overexpressed in breast (12, 27) and prostate (26) cancer; a gene (POTE) derived from a 10 paralog-family that is overexpressed in prostate, ovarian, testicular, and placental cancer (28, 29), a gene (HoxC6) involved in mammary and prostate development, one gene overexpressed in EA (ErbB2; ref. 30); and one gene (SBEM; ref. 31) overexpressed in breast cancer.

To define the ability of real-time reverse transcription-PCR to detect esophageal cancer, RNA was isolated from biopsies obtained during upper endoscopy and analyzed for expression of the genes described above. Gene expression was quantified by determining {Delta}Ct values. The {Delta}Ct value is the difference between the Ct value for a cancer-associated gene and the ß2-microglobin internal reference control gene. Because the amplification efficiencies of the majority of the genes used in this study are close to 100% (data not shown), {Delta}Ct values essentially correspond to a log-2 scale. For example, {Delta}Ct values that differ by 3 correspond to a 23 or 8-fold change in gene expression levels. High {Delta}Ct values are correlated with low levels of gene expression whereas low {Delta}Ct values are correlated with high levels of gene expression. In addition to use for normalization purposes, ß2-microglobin levels were also used to determine whether a given sample contained sufficient mRNA to be included in the study; those samples for which the Ct value for ß2-microglobin was ≥22.0 were excluded from analysis (~5% of total).

Samples analyzed by the first gene panel included NE (n = 18), EA (n = 20), and ESCC (n = 14) tissues. Samples analyzed by the second gene panel were largely the same as the first with some minor differences: NE (n = 17), EA (n = 21), and ESCC (n = 15). In NE tissue, the four genes expressed at the lowest levels were SHh, TFF3 (Fig. 1), POTE, and PDEF (Fig. 2). SHh and TFF3 expression were undetectable in 17 of 18 and 15 of 18 NE tissues, whereas POTE and PDEF expression were undetectable in 11 of 17 and 6 of 17 NE tissues, respectively. For all other genes, significant expression was observed in at least some normal tissue and was highest for the CEA gene.



View larger version (21K):
[in this window]
[in a new window]
  		Fig. 1 Real-time reverse transcription-PCR analysis of EA and ESCC: gene panel #1. Real-time PCR analyses of 18 normal control esophageal biopsies [left side of each matched data set; ({circ}) lower/mid-esophagus; ({bullet}), upper esophagus], 20 EAs ({triangleup}) and 14 squamous cell carcinomas ({blacklozenge}) was done as described in Materials and Methods using primer pairs for the indicated genes. Ct values for each gene were determined from triplicate reactions. {Delta}Ct values were obtained by subtracting the mean Ct value of ß2-microglobin from the mean Ct value for each respective gene.

 


View larger version (22K):
[in this window]
[in a new window]
  		Fig. 2 Real-time reverse transcription-PCR analysis of EA and ESCC: gene panel #2. Real-time PCR analyses of 17 normal control esophageal biopsies [left side of each matched data set; ({circ}), lower/mid-esophagus; ({bullet}), upper esophagus], 21 EAs ({triangleup}) and 15 squamous cell carcinomas ({blacklozenge}) was done as described in Materials and Methods using primer pairs for the indicated genes. Ct values for each gene were determined from triplicate reactions. {Delta}Ct values were obtained by subtracting the mean Ct value of ß2-microglobin from the mean Ct value for each respective gene.

 
With respect to normal tissue, expression levels of TFF1, SHh, POTE, PDEF, SBEM, and EpCam were significantly elevated in EA tissues (Figs. 1 and 2, {triangleup}), providing evidence that these genes had potential diagnostic value for detection of this malignant disease. To more precisely define this value, we did ROC curve analysis for the TFF1, SHh, POTE, PDEF, SBEM, and EpCam genes. ROC curve analysis is the most commonly used method for assessing the accuracy of diagnostic tests (32), and is based on a plot of sensitivity as a function of 1–specificity. The AUC is a measure of diagnostic accuracy such that values between 0.5 and 0.7 indicate low accuracy, values between 0.7 and 0.9 indicate moderate accuracy, and values >0.9 indicate high accuracy (33). With respect to diagnosis of EA (versus NE), EpCam exhibited the highest level of accuracy (AUC, 0.995; 95% CI, 0.906-1.000), whereas SBEM exhibited the lowest (0.799; 95% CI, 0.645-0.907; Table 2). These results provide evidence that EpCam is a highly accurate marker for diagnosis of EA.


View this table:
[in this window]
[in a new window]
  		Table 2 Diagnostic accuracy (with 95% CI) of selected genes associated with EA and esophageal squamous cell carcinoma; oligo d(T) priming method

 
In 8 of 15 (53%) ESCC tissues, expression levels of EpCam were comparable to NE tissues, providing evidence that although this gene was useful for the diagnosis of EA, it was not useful for diagnosis of ESCC. In contrast, expression levels of SBEM in all ESCC tissues were above all NE tissues. In at least some tissues, expression levels of EpCam, POTE, and PDEF were also above those measured in NE tissues (Fig. 2). To precisely define the value of SMEM, EpCam, POTE, and PDEF for diagnosis of ESCC, ROC curve analysis was done. SBEM exhibited the highest level of accuracy (AUC, 1.00; 95% CI, 0.890-1.000), whereas EpCam exhibited the lowest (AUC, 0.743; 95% CI, 0.558-0.880; Table 2).

With respect to normal tissue obtained from the lower esophagus, expression levels of SPARC were elevated in EA, a result consistent with earlier studies (19). However, expression levels of this gene were similar in EA versus ESCC tissues, suggesting that this gene could not discriminate between these two tissues. In contrast, we observed that with respect to ESCC samples, 20 of 21 EA tissues expressed the TFF1 gene at higher levels, suggesting that this gene was useful for discriminating between EA and ESCC. In fact, the AUC value of TFF1 for detection of EA (versus ESCC) was 0.977 (95% CI, 0.862-0.995; Table 2). This result suggests that the TFF1 gene can accurately distinguish EA from ESCC tissue.

A Three-Gene Marker Panel Consisting of EpCam, TFF1, and SBEM Can Discriminate among Normal Esophagus, Esophageal Adenocarcinoma, and Esophageal Squamous Cell Carcinoma Tissues. Using objective AUC criteria, three genes (EpCam, SBEM, and TFF1) were identified that could accurately discriminate between NE and EA, NE and ESCC, or EA and ESCC. Each discrimination was based on a one-gene-one-dimensional analysis. To investigate whether a combination of these genes could be used for molecular classification of esophageal malignancies, we plotted expression levels of TFF1, EpCam, and SBEM in three-dimensional Euclidean space (Fig. 3A). Expression values were derived from tissue samples for which all gene expression values were available (n = 46; 14 NE, 18 AE, 14 ESCC). We then created boundaries between the three tissue types such that the planes were placed equidistant between any two opposing pathologies. We observed that nonoverlapping, continuous, and generally nondistorted planes could be placed between the three groups of tissue pathologies [NE (blue), EA (yellow), and ESCC (red); Fig. 3A]. This result provides evidence that molecular classification of esophageal malignancies can be achieved using a relatively small number of molecular markers.



View larger version (34K):
[in this window]
[in a new window]
  		Fig. 3 Molecular classification of EA and ESCC tissues based on Euclidean spatial coordinates of EpCam, SBEM, and TFF1 mRNA expression values. Expression values of three genes (EpCam, SBEM, and TFF1) in a 46-sample esophageal tissue training set [NE (blue, n = 14); EA (yellow, n = 18); ESCC (red, n = 14)] were determined as described in Materials and Methods, and plotted in three-dimensional space. Decision boundaries separating the various tissue types were determined as described in Materials and Methods. A, training set only; B, spatial positions of the two test samples for which the pathologic and molecular diagnoses disagree are depicted by green spheres. Solid green lines designate ESCC case misdiagnosed as EA, whereas dotted green lines designate EA case misdiagnosed as NE.

 
To determine the accuracy of the three-gene assay system described above, we measured the expression levels of EpCam, SBEM, and TFF1 in 12 esophageal biopsies of unknown pathology. Expression values were entered into a software program (available on-line at http://bioinformatics.musc.edu/~jonas/mitas.html) that first plots the marker coordinates in three-dimensional Euclidean space, followed by a computer-generated molecular diagnosis. In 10 of the 12 patients, the molecular and pathologic diagnoses concurred with one another, indicating that the accuracy of molecular analysis was at least 83%. Of note, the patient that was pathologically diagnosed as ESCC but was molecularly diagnosed as EA (Fig. 3B) had a prior history of prostate adenocarcinoma that was never surgically resected.

To determine whether the genes described above were useful for early detection of EA, we measured the expression levels of EpCam, TFF1, and SBEM in limited quantities (~4 mm3) of frozen BE tissues (n = 5) obtained during surveillance endoscopy. In addition to these genes, we also determined the expression levels of POTE, PDEF, Muc4, SHh, and other genes in the Sonic activation pathway [smoothened, SMO; patched, PTCH; Gli1; see refs. (21, 34–36) ], as well as three genes [endothelial cell growth factor 1, ECGF1 (also known as thymidine phosphorylase, TP); tetraspan1, TSPAN; paired-like homeodomain transcription factor 1, PITX1 or BFT] recently reported by Brabender et al. to be able to discriminate between BE and EA (37). In addition to these test genes, we also determined expression levels of an additional internal reference control gene (ß-actin). Due to the limited quantities of BE tissue and poor amplification of the internal reference control genes in an initial set of BE tissue (n = 5; data not shown), we found it necessary to use a panel of gene-specific primers for reverse transcription. The use of gene-specific primers significantly enhances detection of small amounts of RNA (38) and can also be used for amplification of genes recovered from paraffin-embedded sections. For these reasons, we decided to use gene-specific primers for studies of BE versus EA tissues described below.

A Three-Gene Marker Panel Consisting of TSPAN, ECGF1, and SPARC Can Discriminate Between Barrett's Esophagus and Esophageal Adenocarcinoma Tissues. In the analysis of NE versus EA tissues described above where oligo d(T) primers were used (Figs. 1 and 2), we observed that the diagnostic accuracies of EpCam, POTE, TFF1, SHh, and PDEF were >0.90 (Table 2). When gene-specific primers were used for NE (n = 8) versus EA (n = 26) tissues (Fig. 4), we observed that with the exception of POTE, the diagnostic accuracies of the five genes were within 8% of the values obtained using oligo d(T) priming (see Tables 2 and 3). This result provides evidence that quantitative measurements derived from gene-specific priming are comparable to those derived from oligo-d(T) priming. Regarding the POTE gene, the diagnostic accuracy using gene-specific reverse transcription was 0.718, a 26% decrease compared with oligo d(T) priming. We suspect that the discrepancy for oligo d(T) versus gene-specific priming for the POTE gene is due to the fact that the POTE gene family contains 10 paralogs. Thus, specificity of PCR amplification using gene-specific-primed cDNA may be higher compared with oligo d(T).



View larger version (36K):
[in this window]
[in a new window]
  		Fig. 4 Real-time reverse transcription-PCR analysis of BE and EA using gene-specific primers for reverse transcription. Real-time PCR analyses of 8 normal control esophageal biopsies (left side of each matched data set), 25 BE tissues ({triangleup}) and 26 EA tissues ({blacklozenge}) was done as described in Materials and Methods using primer pairs for the indicated genes. Ct values for each gene were determined from triplicate reactions. {Delta}Ct values were obtained by subtracting the mean Ct value of ß-actin from the mean Ct value for each respective gene.

 

View this table:
[in this window]
[in a new window]
  		Table 3 Diagnostic accuracy (with 95% CI) of selected genes associated with BE and EA; gene-specific priming method

 
With respect to NE, we observed that TSPAN, TFF1, PDEF, SHh, EpCam, PTCH, and Gli1 were up-regulated in BE tissues (Fig. 4; Table 3). Of note, SHh expression was detected in 25 of 25 (100%) BE samples but in only 2 of 8 (25%) NE samples (P < 0.01, {chi}2 test), a result consistent with activation of the SHh pathway in the early phases of esophageal cancer. With respect to BE, the genes exhibiting a diagnostic accuracy >0.80 for discrimination of EA were TSPAN, ECGF1, TFF1, and PDEF (Table 3). All four genes were down-regulated in EA with respect to BE. For discrimination of NE versus EA tissues, six genes were identified with a diagnostic accuracy of >0.90 (Table 3).

In the results described above, three genes were identified that had a diagnostic accuracy of 1.00 for NE versus BE (TSPAN, TFF1, and PDEF), one gene was identified that had a diagnostic accuracy of 1.00 for NE versus EA (EpCam), and no genes were identified that had a diagnostic accuracy ≥0.90 for BE versus EA. To investigate whether a combination of genes might allow for complete discrimination of BE versus EA tissues, we first attempted single linkage analysis of all potential three-marker combinations using a Matlab 7 R12 programming environment. Our goal was to identify the combination of genes that yielded the shortest distance between members of distinct pathologic groups. However, because this analysis failed to yield satisfactory results (data not shown), we decided to determine whether a hierarchical sorting algorithm could accurately classify BE versus EA tissues.

For our sorting algorithm, we started with the gene demonstrating the highest diagnostic accuracy for discrimination of BE versus EA tissue (TSPAN; Table 3). In the first tier of this algorithm, samples were classified at a specificity level of 100% as BE if TSPAN < –0.7 {Delta}Ct units (n = 7), or as EA if TSPAN > 3.95 {Delta}Ct units (n = 13; Table 4) These threshold values corresponded to 2.02 and 1.77 SD beyond the mean expression TSPAN values measured in EA and BE tissues, respectively (Table 4). We then removed the 20 samples from the database that were diagnosed as BE or EA and subjected the remaining 31 samples to ROC curve analysis. Because ECGF1 showed a very high accuracy (0.966; Table 5), we decided to use this gene in the second tier. Samples were classified at a specificity of 100% as BE if ECGF1 > 5.20 {Delta}Ct units (n = 13) or EA if ECGF1 < 4.00 {Delta}Ct units (n = 7). Interestingly, ROC analysis of the 31 samples indicated a reduction in AUC values for PDEF and TFF1 compared with the complete set (from 0.817 and 0.816 to 0.637 and 0.650, respectively), suggesting that PDEF and TFF1 were diagnostically redundant to TSPAN.


View this table:
[in this window]
[in a new window]
  		Table 4 Parameters of three-tiered molecular diagnosis

 

View this table:
[in this window]
[in a new window]
  		Table 5 Diagnostic accuracy values for second- and third-tier analysis

 
The remaining 11 unclassified samples were again subjected to ROC curve analysis. Both SPARC and PITX1 correctly discriminated 10 of 11 samples. Thus, our results suggested that a three-tiered gene discrimination system of either TSPAN-ECGF1-SPARC or TSPAN-ECGF1-PITX1 allowed for correct classification of 50 of 51 tissue samples (98%). In terms of SD units from the mean of the opposing tissue pathology, the two nearest EA/BE neighbors to the PITX1 decision boundary were within 1.35 (± 0.34) SD units (data not shown), whereas those closest to the SPARC decision boundary were within 1.71 (± 0.04) SD units (separate values are shown in Table 4). This result indicated that the discrimination capacity of SPARC was modestly higher compared with PITX1. Thus, we decided to use SPARC as our third-tier discriminatory gene.

To quantify results from the diagnostic assay, values were assigned corresponding to the SD from which a given diagnostic {Delta}Ct value was above (or below) the mean of the opposing tissue pathology. Positive SD values were arbitrarily assigned for diagnoses of EA cases, whereas negative values were arbitrarily assigned for diagnoses of BE cases. Mean (± SD) score values for samples used in the training set are shown in Fig. 5.



View larger version (12K):
[in this window]
[in a new window]
  		Fig. 5 Molecular scores of BE and EA tissues using three- or five-gene marker panels.

 
To assess the accuracy of the TSPAN-ECGF1-SPARC assay, we determined the expression levels of these genes in 12 test samples, 11 of which were embedded in paraffin. The molecular diagnosis of 11 of 12 (92%) samples agreed with the pathologic diagnoses (Table 6), indicating that the accuracy of the molecular test was reasonably high (P = 0.004; {chi}2 test). Seven of seven (100%) samples pathologically classified as EA were diagnosed by molecular criteria as EA, although 4 of 5 (80%) samples pathologically classified as BE were diagnosed by molecular criteria as BE. Mean molecular scores of the test samples were not significantly different from the training set (Fig. 5).


View this table:
[in this window]
[in a new window]
  		Table 6 Validation of TSPAN-ECGF1-SPARC assay

 
For the single misdiagnosed test case, the TSPAN gene was used. To determine whether it was possible to increase the specificity of an EA diagnosis using the TSPAN gene, we first attempted to add an "AND" decision to the first tier of the algorithm involving EA diagnosis. Based on AUC values for discrimination of BE versus EA samples (Table 3), the most reliable genes available for AND decisions are TFF1, PDEF, and SBEM. Using TSPAN, TFF1, and SBEM, we determined that an algorithm could be written that (a) maintained the accuracy of the training set, and (b) increased the accuracy of the test set to 100%. The algorithm is as follows: Sample = EA if TSPAN > 3.95 AND TFF1 > 0.00, OR TSPAN > 3.95 AND SBEM < 16. Based on mean and SD values of the training set, the AND statements would increase the specificity of the decision by adding ~1.16 or 0.30 SD values, respectively. Accordingly, we calculated a three-tiered, five-gene molecular score by (a) recalculating mean and SD values of the entire data base, and (b) added conservative SD values contributed by the TFF1 and SBEM genes (see Table 6).


   DISCUSSION
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
BE arises when the normal esophageal squamous epithelium is replaced by small intestine-like columnar epithelium after damage by chronic gastroesophageal reflux (39). Growth of this epithelium is characterized by successive genomic and transcriptional changes, which may ultimately follow the progression of BE -> low-grade dysplasia -> high-grade dysplasia -> EA. Molecular markers that are able to accurately distinguish BE from EA might be useful for early detection of EA.

In this article, we evaluated molecular markers in two phases, the first of which was focused on discrimination of NE, EA, and ESCC tissue. For this analysis, we identified three genes (EpCam, SBEM, and TFF1) that were capable of accurately classifying the three tissue types. EpCam was useful for discriminating NE from EA tissue, a finding that is consistent with immunohistochemistry results demonstrating the presence of EpCam protein in ~100% of EA tissue samples (40–42), but not in normal esophageal mucosa (42). We observed that TFF1 was useful for discriminating EA from ESCC; TFF1 was expressed at relatively high levels in EA but at relatively low levels in ESCC (Fig. 1). In colon, rectal, and prostate tumors, TFF1 protein levels are significantly elevated in mucosa samples surrounding and in tumor samples (43, 44). In the present study, we observed that SBEM was useful for discriminating ESCC from NE. To our knowledge, this is the first report demonstrating that this gene serves as a marker for any non–breast cancer. SBEM was first identified as a breast cancer-specific gene using the public expressed sequence tag and serial analysis of gene expression databases. It is predicted to code for a putative low molecular weight, secreted sialoglycoprotein (31).

In the second phase of this study, we focused on discrimination between BE and EA tissues. Eleven genes were selected from the first phase of our study, whereas the three remaining genes (ECGF1, TSPAN, and PITX1) were selected based on a recent report by Brabender et al. (37) that these genes were able to discriminate between various esophageal tissues. Using a training set of 51 tissues, we developed a quantitative three-tiered hierarchical algorithm that allowed for accurate (50/51, 98%) discrimination of BE versus EA tissues using only three genes (TSPAN, ECGF1, and SPARC). The algorithm was based on three serial diagnostic evaluations that are listed in Table 4. For a given diagnostic decision, a score was assigned that corresponded to the SD value from which the test sample was from the mean of the opposing tissue type. Based on an analysis of 12 test cases, the three-gene marker panel was reasonably accurate (11/12, 92%). The accuracy of the test was improved to 12 of 12 (100%) in a retrospective manner by including TFF1 and SBEM in the first tier with TSPAN (i.e., a five-gene marker panel).

Of the six diagnostic decisions used in the three-tiered algorithm (Table 4), the one made with the lowest reliability is diagnosis of BE using ECGF1. The threshold value for this particular diagnosis is only 1.44 SD from the mean of EA samples. However, in spite of the apparent low specificity of the threshold setting, 5 of 5 (100%) BE test cases were diagnosed correctly as BE using the ECGF1 gene in the setting of the five-gene marker panel. To minimize the number of future samples inaccurately diagnosed as BE using ECGF1, it will likely be necessary to add an AND statement similar to that used for TSPAN.

In our analysis of BE, we used a panel of gene-specific primers for reverse transcription. Gene-specific priming has been shown to be an effective and reliable method for amplification of small amounts of RNA (38, 45), and may be especially well-suited for analysis of small biopsy samples such as those obtained during upper endoscopy. The primers we used for reverse transcription were the same as the reverse primers used for real-time reverse transcription-PCR. Previous studies have shown that the use of multiple gene-specific primers in a single reaction has been problematic due to the presumed formation of primer-dimers that interfere with the reverse transcription and/or subsequent PCR (46). In the present study, primer dimers were rarely observed. We suspect that the appearance of primer-dimers in previous studies was largely due to primer concentration issues. Unused primers from a reverse transcription reaction are inevitably carried into the subsequent PCR. When a small number of primers are used for reverse transcription, the amount of unused primers is relatively high, resulting in an increase in primer concentration in the subsequent PCR. When a large number of primers are used, the amount of unused primers carried into the PCR is relatively small.

In a previous study, Brabender et al. (37) used linear discriminant analysis of real-time data and identified a combination of genes (TSPAN, ECGFI, and PITX1) that were able to distinguish NE versus BE versus EA tissues. Of the 23 genes examined by Brabender et al., only four overlapped with the present study (TSPAN, ECGF1, PITX1, and SPARC). With respect to (a) relative levels of expression between the three tissue types, and (b) high discrimination capacity between BE and EA tissues, the results we obtained using TSPAN seemed comparable to those of Brabender et al. However, significant differences were observed for the ECGF1 and SPARC genes. For example, whereas the ECGF1 gene was not found to be useful for discrimination of BE versus EA tissue by Brabender et al., our studies indicated that the diagnostic accuracy of this gene was reasonably high. However, in spite of the apparent differences between the two studies, a three-tiered algorithm can be written for the data set of Brabender et al., that is (a) highly accurate (39/39, 100%) for discrimination of BE and EA tissues, and (b) remarkably similar to the one described in this report. The algorithm consists of a combination of TSPAN and RAR{gamma} (or GSTP or RAR{alpha}) in the first tier, followed by COX2 and PITX1 in the second and third tiers, respectively. Similar to the results described in this report, the addition of a second gene in the first tier reduces the number of BE cases misdiagnosed as EA by the TSPAN gene.

In this study, we identified six genes with a diagnostic accuracy >0.90 for discrimination of NE versus BE tissue (Table 3). One of these genes was SHh, a secreted protein ligand that has recently been shown to mediate early events of pancreatic cancer (36) by virtue of binding to a membrane-bound receptor called patched. In the normal stomach, SHh is expressed in gastric parietal cells, whereas its receptor Ptc is mainly present on gastric chief cells (47). In this report, we detected SHh expression in 2 of 22 NE samples and 30 of 30 BE samples (P < 0.001; {chi}2 test) NE tissues (Fig. 1), thus providing strong evidence that this gene is inactive in normal adult esophageal tissue but expressed at early stages of EA tumorigenesis. This finding may warrant consideration of SHh antagonists for the treatment of esophageal malignancies.


   ACKNOWLEDGMENTS
 
We thank Brenda Ferguson (Digestive Disease Center), for coordinating procurement of mucosal biopsy specimens and Margaret Romano of the Hollings Cancer Tissue Bank for obtaining paraffin-embedded tissue sections; Dr. Jennifer G. Schnellmann for critical review of this manuscript, Joy Chafin for DNA sequence analysis, and Drs. Jan Brabender and Paul Marjoram for access to their expression data on BE and EA tissues.


   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 6/ 3/04; revised 11/10/04; accepted 11/15/04.


   REFERENCES
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Blot WJ, Devesa SS, Kneller RW, Fraumeni JF Jr. Rising incidence of adenocarcinoma of the esophagus and gastric cardia. JAMA 1991;265:1287–9.[Abstract]
  2. Levine DS. Management of dysplasia in the columnar-lined esophagus. Gastroenterol Clin North Am 1997;26:613–34.[CrossRef][Medline]
  3. Overholt BF. Results of photodynamic therapy in Barrett's esophagus: a review. Can J Gastroenterol 1999;13:393–6.[Medline]
  4. Falk GW. Barrett's esophagus. Gastroenterology 2002;122:1569–91.[CrossRef][Medline]
  5. Montgomery E, Bronner MP, Goldblum JR, et al. Reproducibility of the diagnosis of dysplasia in Barrett esophagus: a reaffirmation. Hum Pathol 2001;32:368–78.[CrossRef][Medline]
  6. Montgomery E, Goldblum JR, Greenson JK, et al. Dysplasia as a predictive marker for invasive carcinoma in Barrett esophagus: a follow-up study based on 138 cases from a diagnostic variability study. Hum Pathol 2001;32:379–88.[CrossRef][Medline]
  7. Sharma P. Controversies in Barrett's esophagus: management of high grade dysplasia. Semin Gastrointest Dis 2001;12:26–32.[Medline]
  8. Romagnoli R, Collard JM, Gutschow C, Yamusah N, Salizzoni M. Outcomes of dysplasia arising in Barrett's esophagus: a dynamic view. J Am Coll Surg 2003;197:365–71.[CrossRef][Medline]
  9. Chalasani N, Wo JM, Waring JP. Racial differences in the histology, location, and risk factors of esophageal cancer. J Clin Gastroenterol 1998;26:11–3.[CrossRef][Medline]
  10. Comission SCDoHE. Columbia, 2001.
  11. Mitas M, Mikhitarian K, Walters C, et al. Quantitative real-time RT-PCR detection of breast cancer micrometastasis using a multigene marker panel. Int J Cancer 2001;93:162–71.[CrossRef][Medline]
  12. Mitas M, Mikhitarian K, Hoover L, et al. Prostate-specific ets (PSE) factor: a novel marker for the detection of metastatic breast cancer in axillary lymph nodes. Br J Cancer 2002;86:899–905.[CrossRef][Medline]
  13. Mitas M, Cole DJ, Hoover L, et al. Real-time RT-PCR Detects KS1/4 mRNA in mediastinal lymph nodes from patients with non-small cell lung cancer. Clin Chem 2003;49:312–5.[Free Full Text]
  14. Almeida JS. Predictive non-linear modeling of complex data by artificial neural networks. Curr Opin Biotechnol 2002;13:72–6.[CrossRef][Medline]
  15. Mueller M, Wagner CL, Annibale DJ, et al. Web-based prediction of extubation outcome in premature infants on mechanical ventilation using an artificial neural network. Proc AMIA Symp 2003:945.
  16. Van De Bovenkamp JH, Korteland-Van Male AM, Warson C, et al. Gastric-type mucin and TFF-peptide expression in Barrett's oesophagus is disturbed during increased expression of MUC2. Histopathology 2003;42:555–65.[CrossRef][Medline]
  17. Nie SN, Qian XM, Wu XH, et al. Role of TFF in healing of stress-induced gastric lesions. World J Gastroenterol 2003;9:1772–6.[Medline]
  18. Luo A, Kong J, Hu G, et al. Discovery of Ca2+-relevant and differentiation-associated genes downregulated in esophageal squamous cell carcinoma using cDNA microarray. Oncogene 2004;23:1291–9.[CrossRef][Medline]
  19. Brabender J, Lord RV, Metzger R, et al. Differential SPARC mRNA expression in Barrett's oesophagus. Br J Cancer 2003;89:1508–12.[CrossRef][Medline]
  20. Dimmler A, Brabletz T, Hlubek F, et al. Transcription of sonic hedgehog, a potential factor for gastric morphogenesis and gastric mucosa maintenance, is up-regulated in acidic conditions. Lab Invest 2003;83:1829–37.[CrossRef][Medline]
  21. Berman DM, Karhadkar SS, Maitra A, et al. Widespread requirement for Hedgehog ligand stimulation in growth of digestive tract tumours. Nature 2003;425:846–51.[CrossRef][Medline]
  22. Nishimaki H, Kasai K, Kozaki K, et al. A role of activated Sonic hedgehog signaling for the cellular proliferation of oral squamous cell carcinoma cell line. Biochem Biophys Res Commun 2004;314:313–20.[CrossRef][Medline]
  23. Latza U, Niedobitek G, Schwarting R, Nekarda H, Stein H. Ber-EP4: new monoclonal antibody which distinguishes epithelia from mesothelial. J Clin Pathol 1990;43:213–9.[Abstract/Free Full Text]
  24. De Leij L, Helrich W, Stein R, Mattes MJ. SCLC-cluster-2 antibodies detect the pancarcinoma/epithelial glycoprotein EGP-2. Int J Cancer Suppl 1994;8:60–3.[Medline]
  25. Maxwell P. Carcinoembryonic antigen: cell adhesion molecule and useful diagnostic marker. Br J Biomed Sci 1999;56:209–14.[Medline]
  26. Oettgen P, Finger E, Sun Z, et al. PDEF, a novel prostate epithelium-specific ets transcription factor, interacts with the androgen receptor and activates prostate-specific antigen gene expression. J Biol Chem 2000;275:1216–25.[Abstract/Free Full Text]
  27. Ghadersohi A, Sood AK. Prostate epithelium-derived Ets transcription factor mRNA is overexpressed in human breast tumors and is a candidate breast tumor marker and a breast tumor antigen. Clin Cancer Res 2001;7:2731–8.[Abstract/Free Full Text]
  28. Bera TK, Zimonjic DB, Popescu NC, et al. POTE, a highly homologous gene family located on numerous chromosomes and expressed in prostate, ovary, testis, placenta, and prostate cancer. Proc Natl Acad Sci U S A 2002;99:16975–80.[Abstract/Free Full Text]
  29. Zehentner BK, Dillon DC, Jiang Y, et al. Application of a multigene reverse transcription-PCR assay for detection of mammaglobin and complementary transcribed genes in breast cancer lymph nodes. Clin Chem 2002;48:1225–31.[Abstract/Free Full Text]
  30. Flejou JF, Paraf F, Muzeau F, et al. Expression of c-erbB-2 oncogene product in Barrett's adenocarcinoma: pathological and prognostic correlations. J Clin Pathol 1994;47:23–6.[Abstract/Free Full Text]
  31. Miksicek RJ, Myal Y, Watson PH, et al. Identification of a novel breast- and salivary gland-specific, mucin-like gene strongly expressed in normal and tumor human mammary epithelium. Cancer Res 2002;62:2736–40.[Abstract/Free Full Text]
  32. Henderson AR. Assessing test accuracy and its clinical consequences: a primer for receiver operating characteristic curve analysis. Ann Clin Biochem 1993;30:521–39.
  33. Swets JA. Measuring the accuracy of diagnostic systems. Science 1988;240:1285–93.[Abstract/Free Full Text]
  34. Sanchez P, Hernandez AM, Stecca B, et al. Inhibition of prostate cancer proliferation by interference with SONIC HEDGEHOG-GLI1 signaling. Proc Natl Acad Sci U S A 2004.
  35. Qualtrough D, Buda A, Gaffield W, Williams AC, Paraskeva C. Hedgehog signalling in colorectal tumour cells: induction of apoptosis with cyclopamine treatment. Int J Cancer 2004;110:831–7.[CrossRef][Medline]
  36. Thayer SP, di Magliano MP, Heiser PW, et al. Hedgehog is an early and late mediator of pancreatic cancer tumorigenesis. Nature 2003;425:851–6.[CrossRef][Medline]
  37. Brabender J, Marjoram P, Salonga D, et al. A multigene expression panel for the molecular diagnosis of Barrett's esophagus and Barrett's adenocarcinoma of the esophagus. Oncogene 2004;23:4780–8.[CrossRef][Medline]
  38. Mikhitarian K, Reott S, Hoover L, et al. Enhanced detection of RNA from paraffin-embedded tissue using a panel of truncated gene-specific primers for reverse transcription. Biotechniques 2004;36:474–8.[Medline]
  39. MacLennan AJ, Orringer MBBeer DG. Identification of intestinal-type Barrett's metaplasia by using the intestine-specific protein villin and esophageal brush cytology. Mol Carcinog 1999;24:137–43.[CrossRef][Medline]
  40. Went PT, Lugli A, Meier S, et al. Frequent EpCam protein expression in human carcinomas. Hum Pathol 2004;35:122–8.[CrossRef][Medline]
  41. Kumble S, Omary MB, Fajardo LF. Triadafilopoulos G. Multifocal heterogeneity in villin and Ep-CAM expression in Barrett's esophagus. Int J Cancer 1996;66:48–54.[CrossRef][Medline]
  42. Martin IG, Cutts SG, Birbeck K, Gray S, Quirke P. Expression of the 17-1A antigen in gastric and gastro-oesophageal junction adenocarcinomas: a potential immunotherapeutic target? J Clin Pathol 1999;52:701–4.[Abstract]
  43. Vizoso FJ, Fagilde MC, Corte MD, et al. Cytosolic levels of an estrogen-induced breast cancer-associated peptide (TFF1/pS2) in colorectal cancer: clinical significance and relationship with steroid receptors. Int J Biol Markers 2003;18:301–10.[Medline]
  44. Bonkhoff H, Stein U, Welter C, Remberger K. Differential expression of the pS2 protein in the human prostate and prostate cancer: association with premalignant changes and neuroendocrine differentiation. Hum Pathol 1995;26:824–8.[CrossRef][Medline]
  45. Cronin M, Pho M, Dutta D, et al. Measurement of gene expression in archival paraffin-embedded tissues: development and performance of a 92-gene reverse transcriptase-polymerase chain reaction assay. Am J Pathol 2004;164:35–42.[Abstract/Free Full Text]
  46. Dennis P, Edwards EA, Liss SN, Fulthorpe R. Monitoring gene expression in mixed microbial communities by using DNA microarrays. Appl Environ Microbiol 2003;69:769–78.[Abstract/Free Full Text]
  47. van den Brink GR, Hardwick JC, Tytgat GN, et al. Sonic hedgehog regulates gastric gland morphogenesis in man and mouse. Gastroenterology 2001;121:317–28.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 Postgrad. Med. J.Home page
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]

Home page
 GutHome page
R C Fitzgerald
Molecular basis of Barrett's oesophagus and oesophageal adenocarcinoma.
Gut, December 1, 2006; 55(12): 1810 - 1820.
[Full Text] [PDF]

Home page
 J. Mol. Diagn.Home page
F. Hube, Y. Myal, E. Leygue, J. Rollin, Y. Gruel, Y. Yatabe, and on behalf of the authors
Expression of Two Breast-Specific Molecules in the Lung
J. Mol. Diagn., July 1, 2006; 8(3): 390 - 393.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Mitas, M.
Right arrow Articles by Hoffman, B. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Mitas, M.
Right arrow Articles by Hoffman, B. J.

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 Cell Growth & Differentiation