This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Xi, L. Articles by Godfrey, T. E. Search for Related Content PubMed PubMed Citation Articles by Xi, L. Articles by Godfrey, T. E.

Molecular Staging of Lymph Nodes from Patients with Esophageal Adenocarcinoma

Departments of ¹ Surgery and ² Biostatistics, University of Pittsburgh; The ³ University of Pittsburgh Cancer Institute; ⁴ Redpath Integrated Pathology, Pittsburgh, Pennsylvania; and ⁵ Mount Sinai School of Medicine, New York, New York

Requests for reprints: Tony E. Godfrey, Mount Sinai School of Medicine, One Gustave L. Levy Place, Box 1668, New York, NY 10029. Phone: 212-659-9082; Fax: 212-849-2523; E-mail: tony.godfrey{at}mssm.edu.

	ABSTRACT

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Purpose: This study was designed to evaluate molecular markers for the detection of micrometastasis in esophageal adenocarcinoma, define algorithms to distinguish positive from benign lymph nodes and to validate these findings in an independent tissue set and in patients with p_N0 esophageal adenocarcinoma.

Experimental Design: Potential markers were identified through literature and database searches. All markers were analyzed by quantitative reverse transcription (QRT)-PCR on a limited set of primary tumors and benign lymph nodes. Selected markers were further evaluated on a larger tissue set and classification algorithms were generated for individual markers and combinations. Algorithms were statistically validated internally as well as externally on an independent set of lymph nodes. Selected markers were then used to identify occult disease in lymph nodes from 34 patients with p_N0 esophageal adenocarcinoma.

Results: Thirty-nine markers were evaluated, six underwent further analysis and five were analyzed in the external validation study. Two markers provided perfect classification in both the screening and validation sets, although parametric bootstrap analysis estimated 2% to 3% optimism in the observed classification accuracy. Several marker combinations also gave perfect classification in the observed data sets, and estimates of optimism were lower, implying more robust classification than with individual markers alone. Five of thirty-four patients with esophageal adenocarcinoma had positive nodes by multimarker QRT-PCR analysis and disease-free survival was significantly worse in these patients (P = 0.0023).

Conclusions: We have identified novel QRT-PCR markers for the detection of occult lymph node disease in patients with esophageal adenocarcinoma. The objective nature of QRT-PCR results, and the ability to detect occult metastases, make this an attractive alternative to routine pathology.

	INTRODUCTION

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
At the time of diagnosis, many patients with esophageal adenocarcinoma already have distant metastatic disease, and thus, surgical resection is not an option. For earlier stage patients, however, surgery offers the best hope for long-term cure. In these patients, the most important prognostic factor is the involvement of regional lymph nodes. Five-year survival for patients with any lymph node involvement averages < 25%, whereas survival in node-negative patients is 55% to 65% (1, 2). These low survival rates reflect the limitations of surgical treatment in an aggressive and frequently systemic disease for which current chemotherapy regimens are ineffective. In lymph node–negative patients, surgical resection alone should theoretically result in cure. Clearly this is not the case, suggesting that many patients may have pathologically occult lymph node involvement or hematogenous tumor spread with clinically occult distant metastases. Indeed, many recent studies have shown the presence of occult metastases in lymph nodes from patients with esophageal cancer staged as pathologically node-negative (p_N0; refs. 3–14). These studies indicate that 1% to 17% of histologically negative lymph nodes and 11% to 50% of p_N0 patients have nodal metastases that are missed by routine pathologic evaluation. Furthermore, of the studies that include clinical follow-up data (3–10, 15) , all but one (15), show a significant correlation of occult lymph node disease with poor outcome.

In the majority of reported studies, occult metastases have been detected using additional sectioning of formalin-fixed lymph nodes, followed by immunostaining with pan-cytokeratin antibodies. From the literature on several tumor types, it seems that both increased tissue sampling and the addition of immunohistochemistry contribute to increased lymph node metastasis detection rates (7, 8, 10, 16–28). Unfortunately, such detailed lymph node analysis remains restricted to diseases such as breast cancer and melanoma where the use of sentinel lymph node biopsy greatly reduces the number of nodes to be examined. In breast cancer, however, significant discordance has been reported for interpretation of sentinel lymph nodes by different pathologists (29), indicating that pathologic examination is a difficult and subjective process when small volumes of tumor are involved. In esophageal cancer, > 20 lymph nodes are typically removed, and therefore, detailed pathologic analyses of each node with immunohistochemical staining is not feasible. In a different approach, several groups have recently shown that molecular staging, using reverse transcription (RT)-PCR, can also detect occult lymph node metastasis in cancer patients. In these studies, RT-PCR typically gives positive results in 6% to 49% of lymph nodes and 50% to 100% of p_N0 patients (11, 13, 14). This high frequency of positive results in p_N0 patients, many of whom do not recur, has led to criticism of the RT-PCR technique on the basis of poor specificity. To overcome this problem, most investigators have begun to use quantitative RT (QRT)-PCR for detection of occult metastasis. In esophageal cancer patients, we recently reported that QRT-PCR allowed us to obtain high (90%) sensitivity for predicting disease recurrence in node-negative patients, although still maintaining 90% specificity. In our study, we quantified expression of carcinoembryonic antigen (CEA), a somewhat generic adenocarcinoma marker, to achieve these results. However, recent studies in breast and lung cancer have identified genes that are considerably superior to CEA for detecting lymph node metastases. Thus, the goal of this study was to identify and test potential molecular markers for detection of occult lymph node metastasis in esophageal adenocarcinoma.

	MATERIALS AND METHODS

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Identification of Potential Markers. An extensive literature and public database survey was conducted to identify any potential markers relevant to esophageal adenocarcinoma. Resources for this survey included PubMed, OMIM, UniGene (http://www.ncbi.nlm.nih.gov/), GeneCards (http://bioinfo. weizmann.ac.il/cards), and CGAP (http://cgap.nci.nih.gov). Our survey criteria were somewhat flexible but the goal was to identify genes with moderate to high expression in esophageal adenocarcinoma and low expression in normal lymph nodes. In addition, genes reported to be up-regulated in esophageal adenocarcinoma and genes with restricted tissue distribution were considered potentially useful. Finally, genes reported to be cancer-specific, such as the cancer testis antigens and hTERT, were evaluated.

Tissues and Pathologic Evaluation. Tissue specimens were obtained from tissue banks at the University of Pittsburgh Medical Center through Institutional Review Board approved protocols. One half of each lymph node, and a small piece of each primary tumor, were snap-frozen in liquid nitrogen and later embedded in optimal cutting temperature compound (oct) for frozen sectioning and RNA isolation. The remaining tissue was used for routine diagnostic pathology evaluation. Twenty 5-µm sections were cut from each frozen tissue piece for RNA isolation. In addition, sections were cut and placed on slides for H&E and immunohistochemical analysis at the beginning, middle (between the 10th and 11th sections for RNA), and end of the sections for RNA isolation. All three H&E slides from each specimen underwent pathological review to confirm presence of tumor, percentage of tumor, and to identify the presence of any contaminating tissues (such as esophagus or lung tissue on a lymph node). All of the unstained slides were stored at –20°C. Immunohistochemistry evaluation was done using the AE1/AE3 antibody cocktail (DAKO, Carpinteria, CA), and Vector Elite ABC kit and Vector AEC Chromagen (Vector Laboratories, Burlingame, CA). Immunohistochemistry was used as needed to confirm the H&E histology. Only tissues on which diagnostic pathology and frozen section pathology that were concordant were included in the screening tissue sets.

Patients with Esophageal Adenocarcinoma. A total of 314 lymph nodes were analyzed from thirty four consecutive, consenting patients (median 8.5 nodes per patient) with esophageal adenocarcinoma. All patients signed informed consent for this study using protocols and consent forms approved by the University of Pittsburgh Institutional Review Board. The final pathology staging, recurrence status, and other clinical information on these patients is shown in Table 1. Recurrence information was obtained by review of clinical charts and electronic medical records by three of the authors (J.D. Luketich, V.R. Litle, and M.C. Coello).

Marker Validation. To externally validate the classification accuracy of markers tested in the secondary screen, an independent, validation, set of lymph nodes was analyzed. The validation set consisted of 32 benign lymph nodes from patients without cancer and 24 pathology-positive lymph nodes obtained from patients undergoing surgery for esophageal adenocarcinoma. Diagnostic pathology and frozen section pathology were concordant on all nodes in the validation set. In addition, we also analyzed 14 lymph nodes from esophageal cancer patients which were determined to be positive on routine diagnostic pathology but were negative by our own pathology (H&E and immunohistochemistry) on the frozen tissue piece (discordant lymph node set). As with the screening set, all lymph nodes were from independent patients.

RNA Isolation and cDNA Synthesis. RNA was isolated using the RNeasy minikit (Qiagen, Valencia, CA) essentially as described by the manufacturer. The only modification was that we doubled the volume of lysis reagent and loaded the column in two steps. This was found to provide better RNA yield and purity. All RNAs were DNase-treated using the DNA-free kit from (Ambion, Austin, TX) and reverse transcription was done in 100 µL reaction volumes with random hexamer priming and Superscript II (Invitrogen, Carlsbad, CA) reverse transcriptase (30). For the primary screen, two reverse transcription reactions were done, each with 1,000 ng of RNA. The cDNAs were combined and quantitative PCR (QPCR) was done using the equivalent of 20 ng RNA per reaction. For the secondary screen, the RNA input for primary tumors and positive nodes was 1,200 ng per RT reaction and 20 ng per QPCR reaction, but this was increased to 2,000 ng per RT reaction and 80 ng per QPCR reaction for the benign nodes in order to improve sensitivity for detection of low background expression.

Quantitative PCR. All QPCR reactions were done on the ABI Prism 7700 Sequence Detection Instrument (Applied Biosystems, Foster City, CA). Relative expression of the marker genes was calculated using the -C_T methods previously described (30, 31) and with ß-glucuronidase as the endogenous control gene. All assays were designed for use with 5' nuclease hybridization probes although the primary screening was done using SYBR Green quantification in order to save on costs. Assays were designed using the ABI Primer Express Version 2.0 software, and where possible, amplicons spanned exon junctions in order to provide cDNA specificity. In addition, all primer pairs were tested empirically for amplification from 100 ng of genomic DNA and primers were redesigned when necessary. With the exception of cytokeratin 19 (CK19), cDNA-specific primer sets were successfully identified for all genes in this study. Finally, optimal annealing temperature was determined by testing primers for generation of a single band on gels using cDNA templates and annealing temperatures of 60°C, 62°C, and 64°C. Further details describing our methods for primer design and testing have recently been published (32). PCR efficiency was estimated using SYBR green quantification prior to use in the primary screen. Further optimization and more precise estimates of efficiency were done with 5' nuclease probes for all assays used in the secondary screen.

A mixture of the Universal Human Reference RNA (Stratagene, La Jolla, CA) and RNAs from human placenta, thyroid, heart, colon, PCI13 cell line, and SKBR3 cell line served as a universal positive expression control for all the genes in the marker screening process. Tissue RNAs were batched for reverse transcription and the positive control RNA was included in each batch. Batches of cDNAs were then analyzed by QPCR one gene at a time on 96-well plates. Thus, the positive control RNA acted as a control for both reverse transcription and QPCR of each batch. The positive control was not used for calculation of relative expression.

Quantification with SYBR Green (Primary Screen). For SYBR Green I-based QPCR, each 50 µL reaction contained 1x TaqMan buffer A (Applied Biosystems), 300 nmol/L each deoxynucleotide triphosphate, 3.5 mmol MgCl₂, 0.06 units/µL Amplitaq Gold (Applied Biosystems), 0.25x SYBR Green I (Molecular Probes, Eugene, OR), and 200 nmol/L each primer. The amplification program comprised two stages with an initial 95°C Taq activation stage for 12 minutes followed by 40 cycles of 95°C denaturation for 15 seconds, 60°C, 62°C, or 64°C anneal/extend for 60 seconds and a 10-second data collection step at a temperature 2°C to 4°C below the T_m of the specific PCR product being amplified ref. (33); more information on this technique is provided in the Supplementary Methods available online at www.mssm.edu/labs/godfrt01/index.html). After amplification, a melting curve analysis was done by collecting fluorescence data while increasing the temperature from 60°C to 95°C over 20 minutes.

Quantification with 5' Nuclease Probes (Secondary Screen). Probe-based QPCR was done as described previously (12, 31). Briefly, reactions were done with a probe concentration of 200 nmol/L and a 60-second anneal/extend phase at 60°C for ß-GUS, CK19, and CK20, 62°C for CK7 and Villin 1, and 64°C for CEA and TACSTD1. The sequences of primers and probes (purchased from IDT, Coralville, IA) for genes evaluated in the secondary screen are available in Supplemental Table 1 at www.mssm.edu/labs/godfrt01/index.html. The primer sequences for markers used in the primary screen will be provided upon request.

Quantitative PCR Analysis. In the primary screen, data from the melt curve was analyzed using the ABI Prism 7700 Dissociation Curve Analysis 1.0 software (Applied Biosystems). The first derivative of the melting curve was used to determine the product T_m as well as to establish the presence of the specific product in each sample. In general, samples were analyzed in duplicate PCR reactions and the average C_t value was used in the expression analysis. However, in the secondary screen, triplicate reactions were done each individual benign node and the lowest C_t value was used to calculate relative expression and thus obtain the highest (most conservative) estimate of background expression for the sample.

Statistical Analysis
Generation of Prediction Rules. Six markers that passed the secondary screen were evaluated individually and in combination with other markers. The characteristics used to evaluate markers were sensitivity, specificity, classification accuracy, and the area under the ROC curve. For individual markers, a cutoff value was determined that maximized the classification accuracy (proportion of lymph nodes correctly classified). In cases where classification accuracy was 100%, the cutoff was set at the midpoint between the highest expressing benign node and the lowest expressing, histologically positive node. Markers were also combined into pairs and for lymph node classification and a linear prediction rule was generated for each pair. The rule was equivalent to the linear predictor that equalized the fitted probabilities above and below the linear boundary. That is, points on the boundary line had a predicted probability midway between the numerical scores assigned to positive and negative nodes. For example, if positive nodes were assigned a score of 2 and benign nodes a score of 1, predicted scores > 1.5 were classified as positive.

Internal Validation of Prediction Rules. Properties of single and paired marker prediction rules were investigated by examining the distributional properties of the observed marker expression levels and by applying parametric bootstrap validation. Data were simulated from the lognormal and bivariate log-normal distributions using moment estimators for mean, variance, and correlation between marker pairs. One thousand parametric samples of the original data were obtained and the prediction for each bootstrap sample was applied to the original data. The difference between the observed classification accuracy and the average bootstrap classification accuracy was used to estimate the optimism in the resubstitution prediction rules (34).

External Validation of Prediction Rules. External validation of the prediction rules generated in the secondary screen was done on an independent set of benign lymph nodes (n = 32) and positive lymph nodes (n = 24) from patients with esophageal cancer. The prediction rules for individual markers and marker combinations were applied to data from the validation set and classification characteristics were calculated.

	RESULTS

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Primary Screen. Our literature and database surveys identified a total of 39 genes for evaluation in the primary screen. All of these genes were analyzed for expression in six primary esophageal tumors and 10 benign lymph nodes. Median relative expression (relative to the endogenous control gene) in the primary tumors and in benign nodes was calculated for each gene in the primary screen and is reported in Table 2. In addition, we also calculated the ratio of relative expression between the lowest expressing tumor and the highest expressing benign node and between the median expression in tumors and the highest expressing benign node. Some genes, such as CK20, had no detectable expression in benign nodes and therefore ratios could not be calculated. With the exception of CK20, however, all genes with undetectable expression in benign nodes also have very low median expression in the primary tumors and as a result, these genes are unlikely to be sensitive markers for detection of occult disease.

Secondary Screen. Histologic evaluation of the 18 primary adenocarcinoma specimens used in the secondary screen revealed a median tumor percentage of 70% with a range of 5% to 100%. Similarly, in the 20 histologically positive nodes, the median tumor percentage was 50% with a range of 2% to 100%. The relative expression profiles of the markers selected for the secondary screen are shown in Fig. 1. The data shows that these markers are expressed in positive lymph nodes as well as in primary tumors indicating that metastatic tumor cells continue to express these markers. Interestingly, however, analysis of the percentage of tumor cellularity compared with marker expression level revealed only weak associations in all cases except TACSTD1. Figure 1 also indicates the relative expression cutoff values that provide the most accurate classification of histologically positive and negative nodes. Of the six markers chosen from our original screen, all but CK7 was able to classify positive and negative lymph nodes with > 95% accuracy in the secondary screen (Table 3). Furthermore, two markers, CK19 and TACSTD1, provided perfect classification. Of these two, however, the distribution of expression levels in positive and negative nodes is less variable for TACSTD1 than for CK19 and the difference between the distributions is greater for TACSTD1. Thus, one may expect that on a larger sample set, TACSTD1 would prove superior to CK19. In an attempt to address this issue, we did an internal marker validation using parametric bootstrap analysis to estimate the optimism in our observed classification accuracy. Essentially, the expression distribution observed in positive and negative lymph nodes was statistically modeled (as illustrated in Fig. 1B and C) and 41 new data points (20 positive and 21 negative) were artificially generated. The most accurate cutoff value calculated from the artificial data points was then applied to the original 41 data points and the classification accuracy was determined. This was repeated 1,000 times and the average values for classification characteristics on the original 41 lymph nodes were calculated (Table 3, bootstrap estimates). Because the parametric method takes into account the distribution in the observed data, markers with tighter and more separated distributions should fare better in this analysis. Overall, the estimates of optimism were quite low (2-5%), indicating that robust classification of lymph nodes should be possible with these markers. Individually, CK19 and TACSTD1 were again the best markers with estimated classification accuracies of 97% and 98%, respectively.

View larger version (37K): [in this window] [in a new window] [Download PPT slide]   Fig. 1 A, secondary screen data showing the expression profiles of selected markers in primary esophageal tumors (T), histologically positive lymph nodes (PN), and benign lymph nodes (BN). Open, light blue diamonds, samples where expression of CK20 was undetectable at 40 PCR cycles. The expression levels for these six samples were artificially calculated using a CK20, Ct value of 40 cycles. B and C, examples illustrating the approach used for analysis of data from the secondary marker screen. B, data obtained from Villin 1 and CEA; C, data from CK19 and TACSTD1. In each example, the top left graph shows the observed expression data for the selected gene on the 41 lymph nodes in the secondary screen, along with the most accurate cutoff value for lymph node classification. The bottom left graphs illustrate how lymph node classification was analyzed using combinations of marker pairs. In this case, the prediction rule used to classify nodes was calculated such that it split the region into two zones of equal prediction probability as described in MATERIALS AND METHODS. For the top and bottom right graphs, the distribution characteristics of expression in the positive and negative nodes were statistically modeled for each marker, or marker combination, based on the observed expression data. Two hundred simulated positive and negative nodes were then randomly drawn from populations with these modeled distributions and plotted. These plots illustrate how the cutoff values and prediction rules calculated from the observed data may be expected to fare when a larger number of nodes is examined and more extreme values of expression levels are observed. This distribution modeling and resampling approach was also used in the calculation of parametric bootstrap estimates shown in Table 3. B, although CEA and Villin 1 were imperfect discriminators by themselves, combining their information improves prediction in both the observed and simulated data. The larger dispersion seen in the simulated positive nodes reflects the lower correlation, and greater variance, found with these two markers and in positive lymph nodes in general. Among the individual markers, TACSTD1 and CK19, yielded the best empirical discrimination and this also held up well in the simulated data. In addition to yielding the best single marker prediction properties, CK19 and TACSTD1 also benefit from being highly correlated with one another. Higher correlation tends to limit variability in the bivariate distribution.

External Marker Validation. To perform a more robust evaluation of the cutoff values and prediction rules generated in the secondary screen, five markers (CK7 was omitted due to poor classification characteristics in the secondary screen) were further evaluated in an independent, validation set of positive and benign lymph nodes. Pathology analysis of the 24 positive nodes in this set determined that the median tumor percentage was 80% with a range of 2% to 100%. In general, application of the cutoff values and prediction rules to the validation set resulted in classification characteristics of individual markers that were very similar to those estimated by the parametric bootstrap analysis used for internal validation in the secondary screen (Supplemental Table 2 and Fig. 2). The two clear exceptions to this were Villin 1 and CK20, where the specificity was clearly worse than in the secondary screen and as a result, overall classification accuracy dropped by 7% to 10% compared with the bootstrap estimate data from the secondary screen. TACSTD1 and CK19 once again stood out as the best individual markers with 100% accuracy in this set. All combinations of marker pairs resulted in classification accuracy > 92%, even those including Villin 1, and 7 of the 10 combinations gave perfect, 100% classification in the validation set.

Analysis of Nodes with Discordant Pathology. As part of our ongoing studies on esophageal cancer, we have now analyzed over 600 lymph nodes using the markers described in this study. Of these nodes, we have found 14 where the half of the lymph node sent for diagnostic pathology was positive, but our own pathology (including immunohistochemistry) on the frozen lymph node half was inconclusive. One pathologist called our sections from all 14 nodes negative, whereas a second pathologist thought that 8 of the nodes were positive. Two additional pathologists were asked to review these cases and come to a consensus opinion. Their examination concluded that 3 of the 14 nodes were positive and these 3 were among the 8 identified by the second pathologist. Two of these three nodes were positive based on QRT-PCR for multiple markers (Fig. 2), whereas the third node was positive only by QRT-PCR for CK20.

View larger version (22K): [in this window] [in a new window] [Download PPT slide]   Fig. 2 Analysis of the 56 lymph nodes in the validation set and the 14 lymph nodes with discordant pathology. Top, data on five individual gene expressions observed in benign, discordant, and positive nodes. The horizontal line indicates the most accurate cutoff value calculated from the secondary screen and applied to this independent data set. Bottom, validation set data for the same two marker pair combinations illustrated in Fig. 1B and C (CK19/TACSTD1 and CEA/Villin 1). As with the individual markers, the diagonal black line indicates the decision rule generated from the secondary screen data, applied to the validation set. Classification characteristics of the individual markers and all marker combinations are reported in Supplemental Table 2.

View larger version (15K): [in this window] [in a new window] [Download PPT slide]   Fig. 3 Kaplan-Meier disease-free survival curves for 34 patients with pN0 esophageal adenocarcinoma. Patients were stratified into occult disease-positive and occult disease-negative groups based on QRT-PCR data for CK19 and TACSTD1 using the cutoff algorithms developed in the marker screening phase. QRT-PCR-positive patients had a significantly worse disease-free survival than QRT-PCR-negative patients (P = 0.0023, log-rank test). Note that patient 25 was censored at 17.9 months and patient 4 recurred at 18.0 months. Hence, the survival curve drops to zero despite one QRT-PCR-positive patient who did not suffer recurrence in this study.

	DISCUSSION

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Marker Identification. CEA (CEACAM5) is a cell adhesion molecule that is frequently found in adenocarcinomas of endodermally derived digestive system epithelia. Historically, CEA has been the marker of choice for identifying adenocarcinoma of the gastrointestinal tract and was among the first markers to be used for RT-PCR analysis of lymph nodes. TACSTD1, also known as EPCAM, EGP-2, KS1/4, GA-733-2, and MIC-18 is a human cell surface antigen that is defined by the monoclonal antibody AUAI. TACSTD1 is also the antigen for the antibody Ber-Ep4 and, as such, has been used in several studies to detect lymph node micrometastases in a variety of tumor types (23, 35, 36). A recent study has also shown that TACSTD1 is a good RT-PCR-based marker for identification of lymph node micrometastases in non–small cell lung cancer (37).

CK19 is an acidic keratin that is expressed at high levels in many epithelial tissues and their derived tumors. CK19 is one of the keratins targeted by the AE1/AE3 antibody cocktail and, as such, is frequently used by pathologists for immunohistochemical detection or confirmation of lymph node metastases in many cancer types. CK19 has also been used in many RT-PCR-based studies to detect occult lymph node disease in breast cancer (35, 38, 39) . Unfortunately, this gene has at least two pseudogenes that make it difficult to develop a cDNA-specific assay. As a result, the utility of this marker may be limited in clinical settings given the potential for false-positive results. Cytokeratin 7 (CK7, KRT7) is a type II keratin of simple, nonkeratinizing epithelia and is typically found in tissues such as the epidermis, bronchus, and mesothelium. As such, CK7 would not typically be considered as a marker for esophageal adenocarcinoma. The last two markers, CK20 and Villin 1, are generally associated with colonic epithelium. Cytokeratin 20 (CK20, KRT20) is an integral, intermediate filament component and is a major cytoskeletal keratin of the intestinal epithelium. CK20 has been used frequently in RT-PCR studies to detect occult metastases in colon cancer patients but has not been used previously in esophageal cancer. Villin 1 is a major structural component of the brush border cytoskeleton. It is a calcium-regulated, actin-binding protein that is specifically expressed in simple epithelia of some tissues of the gastrointestinal and urogenital tracts. Because esophageal adenocarcinoma is thought to arise from an intestinal type, glandular metaplasia of the normal squamous epithelium, the finding that CK20 and Villin 1 are potentially useful markers for this disease is novel but not surprising.

Lymph Node Classification. Data generated in the screening and validation process indicate that TACSTD1 and CK19 are the best individual markers for classification of positive and benign lymph nodes in esophageal adenocarcinoma. Both markers showed 100% classification accuracy in the secondary screen, and internal validation resulted in estimated accuracies of 98% and 97%, respectively. Furthermore, when the cutoff values from the secondary screen were applied to an independent validation set of lymph nodes, these markers again showed 100% classification. Indeed, the classification accuracy observed for all markers in the validation study was very similar to that seen in the secondary screen, indicating that robust and accurate classification of lymph nodes by QRT-PCR is feasible. In our study, however, the majority of positive lymph nodes examined contained a large percentage of tumors (only seven nodes contained < 5% tumor) and were easily identified by routine pathology. Thus, the extremely high sensitivity and specificity observed in our data may not be reflected upon analysis of node-negative patients for the presence of micrometastatic disease. With this in mind, and because it is highly unlikely that classification accuracy with any individual marker will remain 100% when a larger number of nodes are examined, we also analyzed classification characteristics using combinations of marker pairs. By combining individual markers we hoped to increase the separation between positive and negative lymph node distributions and therefore obtain more reliable and sensitive lymph node classifications. When combining markers with less than perfect classification characteristics, we found that the combination frequently improved the observed classification accuracy beyond that of either individual marker alone. When comparing bootstrap estimates of accuracy between individual markers and marker combinations, the combinations were better than either individual marker in all cases (average 2.5% improvement). Furthermore, the bootstrap estimates of optimism in the individual markers ranged from 2% to 5%, whereas on marker pairs, the optimism was only 0% to 2%. Finally, analysis of the independent, validation lymph node set resulted in perfect classification for 7 of the 10 marker combinations tested. These data indicate that combining the classification characteristics of two markers is likely to result in more robust lymph node classification, in particular when examining nodes for micrometastatic disease.

Statistical Approach. In this study, the decision rules for both single and double markers have been rigorously validated. Internal validation consisted of parametric resampling of lymph node expression data to simulate sampling variability. The decision rules were further tested in an independent test set of lymph nodes from patients at our institution. Thus, we are confident of the reproducibility of classification at our institution and in our laboratory. This process should be considered only the first step in a series of more thorough efforts to generalize RT-PCR-based methods of lymph node analysis. A continuum of increasingly diverse settings needs to be introduced for full evaluation of the generalizability of our ability to discriminate and predict nodal status (40). These include obtaining lymph nodes from patients at other institutions and evaluating QRT-PCR in the hands of other pathologists and laboratory technicians.

Clinical Translation. The clinical utility of RT-PCR staging of lymph nodes relies on the assumption that RT-PCR will eventually prove superior to routine pathology, either by detecting occult disease in p_N0 patients or by providing an objective analysis to help overcome the subjectivity and discordance observed in current pathologic examination. Our work lays the foundation for these studies because we have identified appropriate markers and defined the classification criteria to be applied to lymph nodes from patients with esophageal adenocarcinoma. The detection of occult lymph node disease, and subsequent improved patient staging, could have significant consequences for the treatment of esophageal cancer. For example, although chemotherapy has failed to show a proven benefit in any large, randomized, well-controlled study (41, 42), there is now increasing interest in nonfluorouracil-containing regimens that seem to be better tolerated by patients (43). Specifically, an ongoing Eastern Cooperative Oncology Group trial is investigating platinum-based regimens in combination with either irinotecan or paclitaxel. These therapies are most likely to benefit patients with minimal residual disease, such as patients with micrometastatic involvement of regional lymph nodes. Furthermore, a major benefit of more accurate staging may actually be in our ability to identify the low risk, truly node-negative patients and spare them the potential morbidity of unnecessary chemotherapy.

Surgical approach and extent of lymph node dissection may also be determined by the lymph node status, with some surgeons advocating extensive lymph node dissection in node-positive patients. Furthermore, there is reasonably strong evidence that patients with extensive lymph node involvement have a worse prognosis (1, 7, 44) and that curative surgery may not be possible in this group. These findings have led to a push for the revision of the tumor-node-metastasis staging system in esophageal cancer to include an N₂ stage for patients with extensive lymph node disease. Along these lines, identification of occult lymph node metastases in patients determined to be N₀ by routine histology could help distinguish them from truly node-negative patients. The former subgroup might possibly benefit from more extensive nodal dissection, whereas the latter group may achieve curative resection with less aggressive surgical resection. On the other hand, among node-positive patients, the presence of occult nodal disease could identify those patients with more extensive nodal involvement (the proposed N₂ patients) who might derive a better quality of life from less invasive surgery and/or palliative or systemic treatments rather than from a relatively morbid, and ultimately unsuccessful, radical surgery. Thus, the detection of occult lymph node metastasis could play an important role in future staging and treatment options for pathologically node-positive as well as node-negative patients.

	ACKNOWLEDGMENTS

 
We thank Drs. Alyssa Krasinskas, Kenneth McCarty, Jr., and Susan Silver for reviewing slides on discordant tissues.

	FOOTNOTES

 
Grant support: NIH/National Cancer Institute CA90665-01 research grant, the Veterans Administration Pittsburgh Health Care System, and a cooperative research and development agreement with Cepheid (Sunnyvale, CA).

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.

Note: Supplementary data for this are available at Clinical Cancer Research Online (www.mssm.edu/labs/godfrt01/index.html).

Received 4/27/04; revised 10/20/04; accepted 11/ 3/04.

	REFERENCES

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 

1. Korst RJ, Rusch VW, Venkatraman E, et al. Proposed revision of the staging classification for esophageal cancer. J Thorac Cardiovasc Surg 1998;115:660–9.[Abstract/Free Full Text]
2. Ando N, Ozawa S, Kitagawa Y, Shinozawa Y, Kitajima M. Improvement in the results of surgical treatment of advanced squamous esophageal carcinoma during 15 consecutive years. Ann Surg 2000;232:225–32.[CrossRef][Medline]
3. Izbicki JR, Hosch SB, Pichlmeier U, et al. Prognostic value of immunohistochemically identifiable tumor cells in lymph nodes of patients with completely resected esophageal cancer. N Engl J Med 1997;337:1188–94.[Abstract/Free Full Text]
4. Mueller JD, Stein HJ, Oyang T, et al. Frequency and clinical impact of lymph node micrometastasis and tumor cell microinvolvement in patients with adenocarcinoma of the esophagogastric junction. Cancer 2000;89:1874–82.[CrossRef][Medline]
5. Natsugoe S, Mueller J, Stein HJ, Feith M, Hofler H, Siewert JR. Micrometastasis and tumor cell microinvolvement of lymph nodes from esophageal squamous cell carcinoma: frequency, associated tumor characteristics, and impact on prognosis. Cancer 1998;83:858–66.[CrossRef][Medline]
6. Matsumoto M, Natsugoe S, Nakashima S, et al. Clinical significance of lymph node micrometastasis of p_N0 esophageal squamous cell carcinoma. Cancer Lett 2000;153:189–97.[CrossRef][Medline]
7. Bonavina L, Ferrero S, Midolo V, Buffa R, Cesana B, Peracchia A. Lymph node micrometastases in patients with adenocarcinoma of the esophagogastric junction. J Gastrointest Surg 1999;3:468–76.[CrossRef][Medline]
8. Nakamura T, Ide H, Eguchi R, Hayashi K, Ota M, Takasaki K. Clinical implications of lymph node micrometastasis in patients with histologically node-negative (p_N0) esophageal carcinoma. J Surg Oncol 2002;79:224–9.[CrossRef][Medline]
9. Komukai S, Nishimaki T, Watanabe H, Ajioka Y, Suzuki T, Hatakeyama K. Significance of immunohistochemically demonstrated micrometastases to lymph nodes in esophageal cancer with histologically negative nodes. Surgery 2000;127:40–6.[CrossRef][Medline]
10. Vazquez-Sequeiros E, Wang L, Burgart L, et al. Occult lymph node metastases as a predictor of tumor relapse in patients with node-negative esophageal carcinoma. Gastroenterology 2002;122:1815–21.[CrossRef][Medline]
11. Kijima F, Natsugoe S, Takao S, et al. Detection and clinical significance of lymph node micrometastasis determined by reverse transcription-polymerase chain reaction in patients with esophageal carcinoma. Oncology 2000;58:38–44.[CrossRef][Medline]
12. Godfrey TE, Raja S, Finkelstein SD, Gooding WE, Kelly LA, Luketich JD. Prognostic value of quantitative reverse transcription-polymerase chain reaction in lymph node-negative esophageal cancer patients. Clin Cancer Res 2001;7:4041–8.[Abstract/Free Full Text]
13. Kassis ES, Nguyen N, Shriver SP, Siegfried JM, Schauer PR, Luketich JD. Detection of occult lymph node metastases in esophageal cancer by minimally invasive staging combined with molecular diagnostic techniques. JSLS 1998;2:331–6.[Medline]
14. Luketich JD, Kassis ES, Shriver SP, et al. Detection of micrometastases in histologically negative lymph nodes in esophageal cancer. Ann Thorac Surg 1998;66:1715–8.[Abstract/Free Full Text]
15. Glickman JN, Torres C, Wang HH, et al. The prognostic significance of lymph node micrometastasis in patients with esophageal carcinoma. Cancer 1999;85:769–78.[CrossRef][Medline]
16. Cote RJ, Peterson HF, Chaiwun B, et al. Role of immunohistochemical detection of lymph-node metastases in management of breast cancer. International Breast Cancer Study Group. Lancet 1999;354:896–900.[CrossRef][Medline]
17. Fisher ER, Swamidoss S, Lee CH, Rockette H, Redmond C, Fisher B. Detection and significance of occult axillary node metastases in patients with invasive breast cancer. Cancer 1978;42:2025–31.[CrossRef][Medline]
18. Jannink I, Fan M, Nagy S, Rayudu G, Dowlatshahi K. Serial sectioning of sentinel nodes in patients with breast cancer: a pilot study. Ann Surg Oncol 1998;5:310–4.[Abstract]
19. Isozaki H, Okajima K, Fujii K. Histological evaluation of lymph node metastasis on serial sectioning in gastric cancer with radical lymphadenectomy. Hepatogastroenterology 1997;44:1133–6.[Medline]
20. Chen ZL, Perez S, Holmes EC, et al. Frequency and distribution of occult micrometastases in lymph nodes of patients with non-small cell lung carcinoma. J Natl Cancer Inst 1993;85:493–8.[Abstract/Free Full Text]
21. Dobashi K, Sugio K, Osaki T, Oka T, Yasumoto K. Micrometastatic P53-positive cells in the lymph nodes of non-small cell lung cancer: prognostic significance. J Thorac Cardiovasc Surg 1997;114:339–46.[Abstract/Free Full Text]
22. Maruyama R, Sugio K, Mitsudomi T, Saitoh G, Ishida T, Sugimachi K. Relationship between early recurrence and micrometastases in the lymph nodes of patients with stage I non-small-cell lung cancer. J Thorac Cardiovasc Surg 1997;114:535–43.[Abstract/Free Full Text]
23. Kubuschok B, Passlick B, Izbicki JR, Thetter O, Pantel K. Disseminated tumor cells in lymph nodes as a determinant for survival in surgically resected non-small-cell lung cancer. J Clin Oncol 1999;17:19–24.[Abstract/Free Full Text]
24. Osaki T, Oyama T, Gu CD, et al. Prognostic impact of micrometastatic tumor cells in the lymph nodes and bone marrow of patients with completely resected stage I non-small-cell lung cancer. J Clin Oncol 2002;20:2930–6.[Abstract/Free Full Text]
25. Gu CD, Osaki T, Oyama T, et al. Detection of micrometastatic tumor cells in p_N0 lymph nodes of patients with completely resected nonsmall cell lung cancer: impact on recurrence and Survival. Ann Surg 2002;235:133–9.[CrossRef][Medline]
26. Motomura K, Komoike Y, Inaji H, et al. Multiple sectioning and immunohistochemical staining of sentinel nodes in patients with breast cancer. Br J Surg 2002;89:1032–4.[CrossRef][Medline]
27. Wong SL, Chao C, Edwards MJ, Simpson D, McMasters KM. The use of cytokeratin staining in sentinel lymph node biopsy for breast cancer. Am J Surg 2001;182:330–4.[CrossRef][Medline]
28. Veronesi U, Zurrida S, Mazzarol G, Viale G. Extensive frozen section examination of axillary sentinel nodes to determine selective axillary dissection. World J Surg 2001;25:806–8.[CrossRef][Medline]
29. Roberts CA, Beitsch PD, Litz CE, et al. Interpretive disparity among pathologists in breast sentinel lymph node evaluation. Am J Surg 2003;186:324–9.[CrossRef][Medline]
30. Godfrey TE, Kim S-H, Chavira M, et al. Quantitative mRNA expression analysis from formalin-fixed, paraffin-embedded tissues using 5' nuclease quantitative RT-PCR. J Mol Diagn 2000;2:84–91.[Abstract/Free Full Text]
31. Tassone F, Hagerman RJ, Taylor AK, Gane LW, Godfrey TE, Hagerman PJ. Elevated Levels of FMR1 mRNA in Carrier Males: a new mechanism of involvement in the fragile-X syndrome. Am J Hum Genet 2000;66:6–15.[CrossRef][Medline]
32. Godfrey TE, Kely LA. Development of quantitative RT-PCR assays for measuring gene expression. In: Phouthone K, Grant SG, editors. Molecular Toxicology Protocols. Humana Press;2004.
33. Morrison TB, Weis JJ, Wittwer CT. Quantification of low-copy transcripts by continuous SYBR Green I monitoring during amplification. BioTechniques 1998;24:954–8,960,962.[Medline]
34. Efron B, Tibshirani RJ. An introduction to the bootstrap. Boca Raton: Chapman & Hall/CRC;1994.
35. Schroder CP, Ruiters MH, de Jong S, et al. Detection of micrometastatic breast cancer by means of real time quantitative RT-PCR and immunostaining in perioperative blood samples and sentinel nodes. Int J Cancer 2003;106:611–8.[CrossRef][Medline]
36. Takes RP, Baatenburg de Jong RJ, Wijffels K, et al. Expression of genetic markers in lymph node metastases compared with their primary tumours in head and neck cancer. J Pathol 2001;194:298–302.[CrossRef][Medline]
37. Mitas M, Cole DJ, Hoover L, et al. Real-time reverse transcription-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]
38. 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]
39. Aerts J, Wynendaele W, Paridaens R, et al. A real-time quantitative reverse transcriptase polymerase chain reaction (RT-PCR) to detect breast carcinoma cells in peripheral blood. Ann Oncol 2001;12:39–46.[Abstract/Free Full Text]
40. Justice AC, Covinsky KE, Berlin JA. Assessing the generalizability of prognostic information. Ann Intern Med 1999;130:515–24.[Medline]
41. Blot WJ, Devesa SS, Kneller RW, Fraumeni JF. Rising incidence of adenocarcinoma of the esophagus and gastric cardia. JAMA 1991;265:1287–9.[Abstract/Free Full Text]
42. Blot WJ, McLaughlin JK. The changing epidemiology of esophageal cancer. Semin Oncol 1999;26:2–8.[Medline]
43. Anderson SE, Minsky BD, Bains M, Kelsen DP, Ilson DH. Combined modality therapy in esophageal cancer: the memorial experience. Semin Surg Oncol 2003;21:228–32.[CrossRef][Medline]
44. Lerut T, Coosemans W, Decker G, et al. Extracapsular lymph node involvement is a negative prognostic factor in T3 adenocarcinoma of the distal esophagus and gastroesophageal junction. J Thorac Cardiovasc Surg 2003;126:1121–8.[Abstract/Free Full Text]

This article has been cited by other articles:

				C. A. Klein and N. H. Stoecklein Lessons from an Aggressive Cancer: Evolutionary Dynamics in Esophageal Carcinoma Cancer Res., July 1, 2009; 69(13): 5285 - 5288. [Abstract] [Full Text] [PDF]

				S. Benlloch, J. M. Galbis-Caravajal, C. Alenda, F. M. Peiro, M. Sanchez-Ronco, J. M. Rodriguez-Paniagua, B. Baschwitz, E. Rojas, and B. Massuti Expression of molecular markers in mediastinal nodes from resected stage I non-small-cell lung cancer (NSCLC): prognostic impact and potential role as markers of occult micrometastases Ann. Onc., January 1, 2009; 20(1): 91 - 97. [Abstract] [Full Text] [PDF]

				S. M. Lagarde, F. J.W. ten Kate, J. B. Reitsma, O. R.C. Busch, and J. J. B. van Lanschot Prognostic Factors in Adenocarcinoma of the Esophagus or Gastroesophageal Junction J. Clin. Oncol., September 10, 2006; 24(26): 4347 - 4355. [Abstract] [Full Text] [PDF]

				L. Xi, M. C. Coello, V. R. Litle, S. Raja, W. E. Gooding, S. A. Yousem, T. El-Hefnawy, R. J. Landreneau, J. D. Luketich, and T. E. Godfrey A Combination of Molecular Markers Accurately Detects Lymph Node Metastasis in Non-Small Cell Lung Cancer Patients Clin. Cancer Res., April 15, 2006; 12(8): 2484 - 2491. [Abstract] [Full Text] [PDF]

				L. Xi, W. Gooding, K. McCarty, T. E. Godfrey, and S. J. Hughes Identification of mRNA markers for molecular staging of lymph nodes in colorectal cancer. Clin. Chem., March 1, 2006; 52(3): 520 - 523. [Abstract] [Full Text] [PDF]

				S. Swanson Invited commentary Ann. Thorac. Surg., August 1, 2005; 80(2): 433 - 433. [Full Text] [PDF]

				S. Raja, J. Ching, L. Xi, S. J. Hughes, R. Chang, W. Wong, W. McMillan, W. E. Gooding, K. S. McCarty Jr, M. Chestney, et al. Technology for Automated, Rapid, and Quantitative PCR or Reverse Transcription-PCR Clinical Testing Clin. Chem., May 1, 2005; 51(5): 882 - 890. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Xi, L. Articles by Godfrey, T. E. Search for Related Content PubMed PubMed Citation Articles by Xi, L. Articles by Godfrey, T. E.