the Departments of Hematopathology, Pathology, Experimental Therapeutics, Biostatistics and Applied Math, GI Medicine & Nutrition, Thoracic & Cardiovascular Surgery, and GI Medical Oncology, The University of Texas M.D. Anderson Cancer Center, Houston, TX

    ABSTRACT

    PURPOSE: Patients with localized esophageal carcinoma have a 5-year survival rate of less than 20%. Patients are often treated similarly (ie, with preoperative chemoradiotherapy) but the outcomes vary greatly. Chemoradiotherapy and surgery can result in significant undesirable consequences. Currently, however, there are no tools to help select optimum therapy. We hypothesized that gene expression profiling could provide clues and biomarkers for selection of therapy.

    METHODS: Pretreatment endoscopic cancer biopsies from 19 patients (16 with adenocarcinoma, two with squamous cell carcinoma, and one with adenosquamous carcinoma) enrolled onto a preoperative chemoradiotherapy protocol were profiled using oligonucleotide microarrays. Surgical specimens following therapy were assessed for the degree of pathologic response. On the basis of array data, selected genes were analyzed by polymerase chain reaction.

    RESULTS: Unsupervised hierarchical cluster analysis segregated the cancers into two molecular subtypes, each consisting 10 and nine specimens, respectively. Most cancers (five of six) that had pathologic complete response (pathCR) clustered in molecular subtype I. Subtype II, with one exception, consisted cancers that had less than pathCR (< pathCR). Using a combination marker approach, levels of PERP, S100A2, and SPRR3 allowed discrimination of pathCR from < pathCR with high sensitivity and specificity (85%). Pathway analysis identified apoptotic pathway as one of the key functions downregulated in molecular type II in comparison with type I.

    CONCLUSION: These encouraging, albeit preliminary, data suggest that expression profiling may distinguish cancers with different pathologic outcome. This is the first report to show subtypes of esophageal cancers with distinct molecular signatures. The potential of PERP, S100A2, and SPRR3 as biomarkers of pathCR warrants further validation.

    INTRODUCTION

    Esophageal cancer (ECA) is the ninth most common malignancy in the world and is estimated to be responsible for approximately 13,000 deaths and 14,000 new diagnoses in the United States in 2004.1,2 Even when localized, the 5-year survival rate of less than 20% has not changed significantly in several decades.2,3 The incidence of adenocarcinomas (ACAs) of the esophagus has risen faster than any other malignancy, especially in white males with an estimated increase in incidence by more than 70% in 20 years, thus making ACA the most common histologic type in the West.4,5 Progression of Barrett's metaplasia appears to be one of the major contributors to the observed increase in incidence of ACA.6-8

    The most common approach to treating patients with localized carcinoma of the esophagus, irrespective of the histologic type, is preoperative chemoradiotherapy. This approach provides hypothetical advantages including, higher rate of curative surgery, reduced local relapse, and early therapy of micrometastases. Because of empiric nature, current approaches lead to considerable uncertainty in patient outcome and result in administration of toxic therapies.

    Pretreatment clinical parameters such as TNM classification, primary location, sex and histologic type are unable to predict differences in the biologic behavior of these cancers in patients receiving preoperative chemoradiotherapy.9 One can, however, predict outcome after surgery by reviewing the American Joint Committee on Cancer (AJCC) stage. The most favorable survival is noted in patients who do not have any residual cancer in the resected specimen (pathologic complete response [pathCR]).10-12 The fraction of pathCR patients is approximately 25%. However, biomarkers are not available to identify patients who respond to chemoradiotherapy and thus may be spared from potentially harmful interventions, and patients who benefit from more aggressive treatments.

    Many expression profiling studies have been conducted over the last few years to understand the biology of ECA and to identify bio-markers that can be targeted.13-23 However, these studies lacked treatment and pathologic outcome data to correlate with specific transcriptional signatures. Identification of molecular signatures that predict outcome would be of value in individualizing management of these patients. With this ultimate goal, we profiled pretreatment endoscopic cancer biopsies from patients with ECA using Affymetrix U133A Chip (Santa Clara, CA) and correlated their molecular profiles with pathologic response. The expression levels of a few genes selected on the basis of array data were assessed by polymerase chain reaction (PCR) as biomarkers of pathologic response. In addition, we used Ingenuity Pathways Analysis Software (Ingenuity Systems, Mountain View, CA) to identify, from the microarray data, key biologic pathways, and functions associated with chemoradiotherapy resistance.

    METHODS

    Patient Selection and Evaluation

    All patients in this report participated in a clinical trial approved by The University of Texas M.D. Anderson Cancer Center (Houston, TX) institutional review board. Patients with localized histologically confirmed squamous cell carcinoma (SCCA) or ACA of the thoracic esophagus were considered eligible. Patients were evaluated by chest radiograph, computed tomography of the chest and abdomen, upper GI double-contrast barium radiographs, an esophagogastroduodenoscopy with endoscopic ultrasonography (EUS), ECG, SMA-12, electrolytes, CBC including platelet count, and serum baseline carcinoembryonic antigen (CEA) level. Positron emission tomography (PET) was performed when available. Patients with T2-3 with any N, patients with M1a cancer (celiac nodes associated with a gastroesophageal junction carcinoma), and patients with T1N1 carcinoma were considered eligible. All patients were evaluated before registration by a multidisciplinary team that included thoracic oncology surgeons, radiation oncologists, gastroenterologists, and medical oncologists. Eligible patients had to have cancer that was considered technically resectable and medically operable on the basis of the clinical staging and evaluation. All patients signed a written informed consent, which was approved by the institutional review board.

    Patients with T4 cancer and patients with T1N0 lesions were excluded. Patients with any evidence of metastatic cancer were also not enrolled. Patients with uncontrolled medical conditions (such as diabetes, hypertension, heart condition classified as New York Heart Association class III or IV, or psychiatric illness) were not eligible. Patients who could not comprehend the purpose of this clinical trial or comply with its requirements were not enrolled.

    Treatment

    The objective of the protocol was to determine the feasibility of the three-step approach using three chemotherapy agents before and during preoperative chemoradiotherapy. If a patient had an R0 resection, no further therapy was planned. Patients who underwent an R1 resection (microscopic carcinoma at the margin) or R2 resection (gross carcinoma after surgery) or who had M1 disease were offered palliative care.

    Step 1: Induction Chemotherapy

    All patients had a central venous line placed before starting chemotherapy. Patients received docetaxel as intravenous (IV) bolus (at 33 mg/m2), irinotecan (at 55 mg/m2) as IV bolus, and fluorouracil (at 2 g/m2) infusion over 24 hours weekly for 2 weeks, followed by 1 week off. One cycle was 6 weeks long. If there was no cancer progression after the first cycle, patients received a second cycle of the induction chemotherapy. Standard premedications were used.

    Step 2: Preoperative Chemoradiotherapy

    Patients received up to 50.4 Gy of radiotherapy in 28 fractions. Concurrently, patients received docetaxel (20 mg/m2) IV bolus weekly, irinotecan (30 mg/m2) IV bolus weekly, and fluorouracil (300 mg/m2/24 hours as continuous infusion Monday through Friday of each radiotherapy week). Standard premedications were used.

    Step 3: Surgery

    Approximately 5 to 6 weeks after the end of chemoradiotherapy, patients were restaged fully. If there was no contraindication for surgery, patients underwent an attempted surgical resection of the primary and regional nodes. Patients were then followed for 5 years or until death.

    Tissue Collection

    Patients undergoing therapeutic or diagnostic endoscopic procedure volunteered in an approved tissue collection protocol, thus allowing collection and storage of specimens (blood and cancer tissue). Up to 15 biopsy specimens (eight cancer, four junction, and three adjacent, noncancerous tissues) were collected. The size of usual biopsy was approximately 1.0 mm, and tissue specimens were snap frozen in liquid nitrogen until use. Pretreatment cancer tissues from 19 of the patients, 16 with ACA, two with SCCA, and one with adenosquamous carcinoma (ASCCA) who participated in the tissue analysis and also underwent surgery after therapy were subjected to gene expression profiling by microarray analysis.

    Histologic Evaluation

    For each specimen analyzed by microarray, an adjacent tissue biopsy was given to a pathologist for assessing the presence of cancer and its histology. Routine hematoxylin and eosin–stained slides were used to evaluate for the presence of cancer in pretreatment endoscopic biopsies and esophagectomy specimens.

    Postchemoradiation resected surgical specimens with no residual cancer were classified as achieving pathCR, whereas others with the presence of any cancer cell in the specimen were classified as less than pathCR (< pathCR).

    Synthesis of Biotin-Labeled cRNA and Hybridization

    RNAs from the tissue biopsies were isolated using RNeasy Mini kit (Qiagen, Valencia, CA) according to the manufacturer's recommendations. The quantity of the RNA was determined spectrophotometrically at 260 nm and the integrity of RNA was assessed by Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA). Only high-quality RNA with intact 18S and 28S RNA was used for the synthesis of biotin-labeled cRNA. Because the RNA yield from some of the biopsies was less than the 5 μg required for our standard protocol, an alternative small sample protocol with a second round of amplification that was established by the M.D. Anderson Affymetrix core facility was used to generate biotin-labeled cRNA. Briefly, 200 ng of total RNA from each specimen is converted to cDNA using the SuperScript choice system (Invitrogen, San Diego, CA), then to cRNA by in vitro transcription using Ambion MeGAscript T7 kit (Ambion, Austin, TX). In the second cycle, biotin-labeled cRNA is generated from the second round cDNA using the Enzo BioArray High Yield RNA Transcript Labeling Kit (Enzo Diagnostics, Farmingdale, NY). The yield of biotin labeled cRNA is determined by measuring absorbance at 260 nm. Fifteen micrograms of cRNA is then fragmented and hybridized to Affymetrix U133A Chip as per manufacturer's instructions. A total of 19 RNAs isolated from pretreatment cancer tissue from patients with ECA were subjected to microarray analysis.

    Oligonucleotide Microarray Analysis

    Each U133A microarray contains 22,215 noncontrol probe sets that correspond to more than 18,400 distinct transcripts, including 14,593 well-characterized human genes. The list of probe sets and corresponding genes is available from the Affymetrix Web site (http://www.affymetrix.com/support/technical/libraryfilesmain.affx). Hybridization of biotin-labeled cRNA to the oligonucleotide arrays and image analysis was performed in the DNA Microarray Core Facility at the M.D. Anderson Cancer Center according to protocols available on their Web site (http://www.mdanderson.org/departments/dnamicroarray).

    Microarray Suite (MAS) 5.0 software and custom tools developed by the M.D. Anderson Cancer Center Bioinformatics Department were used to analyze the data. Briefly, the microarray data were processed using the positional dependent nearest neighbor model to normalize and to extract gene expression values.24 Then, a hierarchical clustering algorithm was used to cluster genes and samples.25 The absent genes and the invariant genes were filtered out before clustering. The genes with below-median expression value were regarded as absent genes. The invariant genes were selected according to the standard deviation of expression values across all samples (). The genes that have  less than three times the average of  over all the genes on the array are regarded as invariant genes. The cluster analysis was performed using the uncentered Pearson correlation as similarity metric and average linkage algorithm to combine cluster branches.

    Differentially expressed genes were identified using standard t test. The false discovery rate of the list of the differentially expressed genes was estimated using the beta-uniform mixture (BUM) distribution model.26

    Ingenuity Pathways Analysis

    Ingenuity Pathways (INGP) Analysis software was used to identify key functions and pathways differentially regulated between the two molecular subtypes of ECAs. The INGP Analysis software is a Web-delivered application that enables biologists to discover, visualize, and explore therapeutically relevant networks significant to gene expression array data sets. The INGP allows concurrent analysis of multiple data sets across different experimental platforms based on the Ingenuity Knowledge Base (Ingenuity Systems), a database consisting of millions of individually modeled relationships between proteins, genes, cells, tissues, drugs and diseases for the identification of key functions and pathways distinguishing biologic states. A detailed description of INGP analysis is available at Ingenuity Systems' Web site (http//www.ingenuity.com).

    The average log2 expression values were used to calculate the fold change (log2 FC) between cancer subtypes I and II. The data set containing gene identifiers and their corresponding expression values (log2 FC values) were then uploaded into the INGP as a tab-delimited text file to perform the analysis. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. A fold-change cutoff of 2 was set to identify genes whose expressions were differentially regulated. These genes, called focus genes, were then used as the starting point for generating biologic networks. To start building networks, the application queries the Ingenuity Pathways Knowledge Base for interactions between focus genes and all other gene objects stored in the knowledge base, and generates a set of networks with a network size of 20 genes/proteins. INGP Analysis then computes a score for each network according to the fit of the user's set of significant genes. The score is derived from a P value and indicates the likelihood of the focus genes in a network being found together as a result of random chance. A score of 2 indicates that there is a 1-in-100 chance that the focus genes are together in a network as a result of random chance. Therefore, scores of 2 or higher have at least a 99% confidence of not being generated by random chance alone. Biologic functions are then calculated and assigned to each network.

    Biologic functions were assigned to each gene network by using the findings that have been extracted from the scientific literature and stored in the Ingenuity Pathways Knowledge Base. The biologic functions assigned to each network are ranked according to the significance of that biologic function to the network. Fisher's exact test is used to calculate a P value determining the probability that the biologic function assigned to that network is explained by chance alone.

    Real-Time Quantitative PCR

    To validate microarray data, we chose few genes based on differential expression greater than two-fold between the two molecular subtypes, and performed real-time quantitative PCR (qPCR). cDNA for real-time qPCR was generated for each sample using a kit from Invitrogen according to manufacturer's instructions. Briefly, 100 ng of total RNA from the same aliquot of RNA that was used for microarray analysis was reverse transcribed using random primers and SuperScript II reverse transcriptase in a total volume of 20 μL. Each reaction was performed in triplicate and final reaction products were pooled and stored at –20°C until further use. The TaqMan minor grove binder probe and the ABI Prism 7900 Sequence Detection system (PE Applied Biosystems, Foster City, CA) were used for detecting real-time PCR products. Primers and probes for the target and internal control genes were designed by Perkin Elmer Applied Biosystems and obtained via their Assays-on-Demand Gene Expression Products services. All gene expression assays have TaqMan (Applied Biosystems) minor groove binder probe with a corboxyfluorescein reporter dye at the 5' end and a fluorescent quencher at the 3' end of probe. Each target was amplified individually; PCR assays included 10 μL of TaqMan universal master mix No AmpErase UNG (2x), 1 μL of 20x Assays-on-Demand Gene Expression Assay Mix, and 2 μL of cDNA diluted in RNase free water, in a final volume of 20 μL. The PCR thermal cycling conditions performed for all of the samples was as follows: 10 minutes at 95°C for AmpliTaq Gold (Applied Biosystems) activation; and 40 cycles for the melting (95°C for 15 seconds) and annealing/extension (60°C for 1 minute) steps. PCR reactions for each target and control (18S RNA) genes were performed in duplicate.

    Comparative CT Method for Relative Quantification of Gene Expression

    The comparative CT method (2- CT) was used to determine relative gene expression levels for each target gene.27 Results of real time reverse transcriptase PCR (RT-PCR) data are expressed as CT values, where CT is defined as the threshold PCR cycle number at which an amplified signal above the baseline is detected. There is an inverse relationship between CT and amount of target; thus, lower target amounts correspond to higher CT and vice versa. In order to determine relative gene expression levels, first, the duplicate CT values for the control (18S RNA) and the target gene were averaged for of each sample. The relative expression levels of target genes in comparison wtih control gene were then calculated using the formula 2-CT where CT represents the difference between each target gene and the control gene (average CT for the target minus average CT for 18S RNA). The relative gene expression values were multiplied by a factor of 106 to make the values greater than 0.01, and to simplify presentation of the data.

    Discrimination Analysis

    The potential of the three genes to discriminate the two cancer subtypes was assessed by linear discrimination analysis (LDA) using S-PLUS software package (Insightful Corp, Seattle, WA). The log expression values of S100A2, PERP, and SPPR3 were used as predictors and labels of subtype I and subtype II were used as response variables.

    RESULTS

    Patient characteristics are described in Table 1. PathCR was observed in 32% of cancers (six of 19). Unsupervised hierarchical cluster analysis segregated the cancers into two major categories, each consisting of 10 and nine cancers respectively (Fig 1). Approximately 400 genes were differentially expressed between the two subtypes, with an estimated false-discovery rate of 5%. The molecular subtype I comprised seven ACAs and two SCCAs and one ASCCA, whereas subtype II contained only ACAs. Thus, ACAs segregated into two categories. It is worth noting that the segregation of ACAs into two subtypes remained same when two of the SCCAs were excluded from the clustering analysis (data not shown). Five of the cancers with pathCR (four of five ACAs and one of one SCCA) clustered together in type I. Subtype II, with one exception, consisted of cancers with < pathCR. The clustering pattern was robust against the gene filtering process and clustering algorithm used in the study. For instance, the partitioning of the subtypes remained unchanged when complete linkage algorithm was used instead of average linkage algorithm. Additionally, the partitioning of the two main sub-branches (ie, the two subtypes) and the partitioning of the pathCR samples in to the two sub-branches remained the same when the number of variant genes included in the cluster analysis changed from 50 to 800 by altering the 3 x  boundary.

    The median time to locoregional and metastatic progression has not yet been reached by either of the molecular subtypes. Nevertheless, the molecular subtype II appears to portend shorter disease-free survival (DFS) time, with a mean time to DFS of 22.42 months (95% CI, 15 to 29) compared to 28.55 months (95% CI, 21 to 36) for the molecular subtype I. At 14 months 54% of subtype II was free of disease compared with 75% of subtype I. Similarly, the median time of overall survival (OS) has not yet been reached by either of the molecular subtypes. Again molecular subtype II portends a worse OS, with a median OS time of 23 months (95% CI, 16 to 30) compared with 27.3 months (95% CI, 20 to 35) for subtype I. At 14 months, 57.4% of subtype II survived compared with 77.7% of subtype 1.

    Greater than two-fold differences in the expression levels were observed in 80 genes using the t test (P < .0001). Genes associated with apoptosis, calcium homeostasis, stress response, and proliferation were downregulated in molecular subtype II in comparison with subtype I (Table 2). They include genes encoding annexin 1, chromosome 1 open reading frame 10 (C1orf10), cystatin A and B (stefin A and B), S100 calcium binding proteins, (S100A2, S100A7-9 and S100A14), small proline-rich proteins (SPRR1A, SPRR1B, SPRR2A, SPRR2C, SPRR3), heat shock protein 27 (Hsp27), TACSTD2, and transglutaminase 3 (TGM3). Several of these proteins are Ca2+-binding or -regulating proteins and are components of the cornified cell envelope, which is a specialized structure that forms in terminally differentiated epithelial cells and provides a barrier against mechanical and chemical stress. For instance, TGM3, a Ca2+-dependent enzyme that catalyzes covalent cross-linking reactions between proteins or peptides by - glutamyl lysine isopeptide bonds is important for effective epithelial barrier formation and the assembly of the cell envelope.

    The top four functions identified by IGNP to be differentially regulated between the two molecular subtypes of ECA were embryonic development, tissue development, cell-to-cell signaling and interactions, and cell death. The network profile shown in Figure 2 generated by INGP highlights the inter-relationship between various genes and the apoptotic pathway downregulated in subtype II.

    The relative expression levels of genes PERP, S100A2, and SPRR3 evaluated by real-time qPCR are shown in Figure 3. Because of insufficient quantities of RNA, specimens 24 and 20 were not included in the real-time PCR analysis. PERP (TP53 effector related to peripheral myelin protein 22 [PMP22]) is a novel type of effector involved in p53-dependent apoptosis.28 This protein is a member of expanding family of tetraspan membrane proteins, including PMP22 and the epithelial membrane proteins 1, 2 and 3 (EMP1-3).29 Overexpression of EMP proteins has been shown to induce cell death through a mechanism that involves association with the P2X7 cation channel and the consequent induction of membrane blebbing. Because of significant sequence homology to both PMP22 and EMPs, it is postulated that PERP, too, can induce membrane blebbing that contributes to activation of the apoptotic pathway. The S100A2 gene encoding a calcium binding protein is considered as candidate tumor suppressor gene because of its underexpression in several cancers, including esophageal SCCA, in comparison with healthy epithelia.30-32 In addition, S100A2 recently has been shown to be a novel downstream mediator of Np63.33 SPRR3, a member of small proline-rich proteins, is a component of the cell envelope and is expressed in stratified squamous epithelia during differentiation. This gene has been identified as a marker of esophageal cancer progression.34-37

    The relative expression values of all the three genes were lower in tumors belonging to subtype II in comparison with tumors in type I (Fig 3), confirming our microarray data. For example, the expression values of PERP were below 75 (range, 1.4 to 75) in subtype II; they were higher than 100 (range, 100 to 394) with one exception in cancers belonging to subtype I. Levels of S100A2 ranged between 0.3 and 38 and were below 10 in subtype II cancers except for cancer 11, and ranged between 5 and 50,000 with values above 10 in subtype I except for cancers 6 and 16. The expression of SPRR3, though overall lower in type II tumors, varied similarly among tumors, ranging from 0.01 to 6 in subtype II and from 0.13 to 23,522 in subtype I. Thus, no single marker was able to segregate the two molecular subtypes without an overlap. We used a statistical method (LDA) to see if the combination of genes examined by PCR has the potential to separate subtype I and subtype II into two distinct groups. Using SPRR3 and S100A2, the separation is statistically significant (P = .014; Hotelling T squared for differences in means between subtype I and subtype II). The P value was .0006 when sample 6, which appears to be an outlier, is omitted from the analysis. Thus, combining S100A2 and SPPR3 produces a classifier that separated subtype I and subtype II samples with only one outlier (sample 6).

    We noticed that expression values of the three marker genes were substantially higher in cancers that achieved pathCR compared with cancer with < pathCR. When we used an arbitrary cutoff value of 100 for relative expression, seven of 17 cancers showed expression values > 100 in at least two of the three markers. These seven included five cancers that achieved pathCR (cancers 1, 3, 16, 56, and 23). Thus, only two of 17 cancers, 2 and 19, with < pathCR showed expression values > 100 in at least two of the three markers. The specificity (true negatives/true negatives plus false positives) and the sensitivity (true positives/true positives plus false negatives) of the combination marker approach for identifying pathCR were 85% (11 of 13) and 86% (six of seven), respectively.

    DISCUSSION

    The clinical course of patients with ECA is heterogeneous. Thus, patients with the same disease stage have variable outcomes from uniform therapy. Patients with chemoradiotherapy-resistant cancer have a high likelihood of developing metastases.38 Currently, an empiric approach is utilized for patients with locoregional esophageal cancer, because one is not able to predict the degree of chemoradiotherapy resistance before surgery. The early identification of nonresponders would allow physicians to discontinue ineffective treatment regimens and institute alternative treatments, thereby avoiding both overtreatment and undertreatment of patients. Therefore, the need for markers that predict response early during the course of therapy is widely acknowledged.

    In an attempt to identify a panel of biomarkers that allow us to predict response to chemoradiotherapy, we profiled pretreatment cancer biopsies from 19 patients enrolled in a clinical protocol. Six of these patients (32%) had a pathCR. Unsupervised cluster analysis separated the cancers into two categories. Interestingly, five (83%) of the six cancers that achieved pathCR clustered in one molecular subtype (type I). Only one cancer with pathCR fell in subtype II.

    There was no clear segregation, however, of pathCR from < pathCR in subtype I, because 30% (five of 13) of cancers with < pathCR also clustered in this subtype. Nevertheless, our PCR data point out that expression analysis of a limited set of biomarkers selected from the list of genes that were regulated differentially between the two subtypes increases the predictive power. Thus, simply using three markers, PERP, S100A2, and SPRR3, and choosing an arbitrary expression cutoff value of 100, we were able to assign cancers to pathCR and < pathCR categories in 15 of 17 cancers tested by PCR.

    Median time to locoregional and metastatic progression was not reached by either of the molecular subtypes. Similarly, the median time of OS was not yet reached by either of the molecular subtypes. However, the molecular subtype II appears to portend shorter DFS time, and a worse OS compared with subtype I.

    Many of the genes with differential expression between the two types of ECAs have previously been reported to show altered expression in esophageal cancers by other investigators, confirming that they were cancer related.19,21,30,36,39-43 It is interesting to note that Luo et al,21 using high-density cDNA microarray platform, also observed that several genes including annexin 1, SPRRS, S100A8 and A9, TGM3, CK4, CK13, and CK15, were downregulated in SCCA in comparison with healthy squamous epithelium.

    Collective down regulation of several members of apoptotic pathway such Bcl-2/EIB 19 kDa interacting protein 3 (BNIP3), PERP, epithelial membrane protein (EMP1), p63, stratifin (SFN)/14-3-3, and S100A2 in nonresponder cancer type as illustrated in network profile (Fig 2) implicates a critical role of apoptotic pathway in chem-oradiotherapy resistance in ECA. Solid tumors are poorly oxygenated as compared with healthy tissues and consist regions of hypoxia. These hypoxic regions often correlate with poor prognosis as a result of the ability of cells within these regions to become resistant to chemotherapeutic reagents and radiation therapy. Apoptosis induced by hypoxia is a mechanism for elimination of stressed cells. Similarly, ionizing radiation and chemotherapeutic agents use the process of programmed cell death to induce cancer cell death. In vitro studies have shown that several genes we noticed to be differentially regulated between the two molecular types were indeed associated with response to chemoradiotherapy. For example, BNIP3 encoded by Bnip3L, a unique member of the Bcl2 family members is downregulated in cancer cells that are resistant to fluorouracil.44 Van de Velde et al45 have shown that forced overexpression of BNIP3 induces cell death characterized by localization at the mitochondria, loss of membrane potential and reactive oxygen species production. More recent studies have demonstrated that Bnip3L is inducible by p53 under hypoxia, and its knockdown promotes tumor growth.46 Similarly, Hermeking et al47 have demonstrated that SFN is induced after DNA damage in a p53 dependent manner. It is also shown to play a crucial role in the G2 checkpoint by sequestering the mitotic initiation complex, cdc2-cyclin B1, in the cytoplasm after DNA damage.48 Further, on cisplatin induced DNA damage, SFN is shown to bind with phosphorylated Np63 isoform and mediate nuclear export of Np63 into cytoplasm.49

    To our knowledge, this is the first report showing two types of esophageal ACA with distinct molecular signatures. It is clear from published studies that the genes expressed differentially in the two molecular subtypes in our study are cancer-related genes. Because many of these genes are highly and uniformly expressed in healthy squamous epithelium, earlier profiling studies comparing tumors with healthy squamous epithelium may have clustered tumors with varying degree of loss of expression in to one category. Excluding healthy esophageal mucosa in microarray analysis in our strategy may, in fact, have accentuated the separation of the molecular subtypes based on differences in the relative expression levels among the tumors and not between tumors and healthy mucosa. Thus, it appears that it is not the loss or gain of expression of these genes in comparison with healthy squamous epithelium, but it is the relative levels in different tumors that distinguish responders from nonresponders. Because our tumor specimens were unselected with regard to percentage of stromal infiltration or inflammation, we realize that the clustering results might reflect contributions from non-neoplastic cellular elements to the expression signatures. However, both tumor and its surrounding microenvironment are important in tumor growth and response, inclusion of these components may be more beneficial than detrimental in studies such as ours that are designed to associate molecular signatures with pathologic response. Our data indicate that analysis of combination of biomarkers that are analyzed easily by quantitative assays such as PCR may be sufficient for distinguishing cancers that respond to therapy from those resistant to therapy. However, our study included only a small number of specimens; hence, vigorous validation with a larger set of samples is warranted to assess the predictive power of these potential markers.

    Authors' Disclosures of Potential Conflicts of Interest

    The authors indicated no potential conflicts of interest.

    Author Contributions

    Conception and design: Rajyalakshmi Luthra

    Provision of study materials or patients: Tsung-Teh Wu, Jeffrey H. Lee, Robert Bresalier, Asif Rashid, Stephen G. Swisher, Jaffer A. Ajani

    Collection and assembly of data: Rajyalakshmi Luthra, Madan G. Luthra, Enrique Lopez-Alvarez, Li Zhang, Jaime Bailey

    Data analysis and interpretation: Rajyalakshmi Luthra, Madan G. Luthra, Li Zhang, Asif Rashid, Jaffer A. Ajani

    Manuscript writing: Rajyalakshmi Luthra, Madan G. Luthra, Julie Izzo, Jaffer A. Ajani

    Final approval of manuscript: Rajyalakshmi Luthra, Jaffer A. Ajani

    GLOSSARY

    Apoptotic pathway:

    Also called programmed cell death, apoptosis is a signaling pathway that leads to cellular suicide in an organized manner. Several factors and receptors are specific to the apoptotic pathway. The net result is that cells shrink, develop blebs on their surface, and their DNA undergoes fragmentation.

    Barrett's metaplasia:

    A condition in which stratified squamous cells of the esophagus are replaced by specialized columnar epithelial cells similar to ones found in the intestine. Barrett's metaplasia, generally developing in patients suffering from gastroesoplageal reflux disease (GERD), is a predisposing condition that increases the risk of developing esophageal adenocarcinoma.

    Biomarker:

    A functional biochemical or molecular indicator of a biologic or disease process that has predictive, diagnostic, and/or prognostic utility.

    Cell envelope:

    A specialized structure that forms in terminally differentiated epithelial cells and provides a barrier against mechanical and chemical stress.

    Clustering:

    Organization of data consisting of many variables (multivariate data) into classes with similar patterns. Hierarchical clustering creates a dendrogram based on pairwise similarities in gene expression within a set of samples. Samples within a cluster are more similar to one another than to samples outside the clus-ter. The vertical length of branches in the tree represents the ex tent of similarity between the samples. Thus, shorter the branch length, the fewer the differences between the samples.

    Expression profiling:

    The expression of a set of genes in a biologic sample (eg, blood, tissue) using microarray technology.

    Less than pathologic complete response (< pathCR):

    The presence of any residual tumor cells in a histologic evaluation of a tumor specimen is defined as a less than complete pathologic response.

    Microarray:

    A miniature array of regularly spaced DNA or oligonucleotide sequences printed on a solid support at high density that is used in a hybridization assay. The sequences may be cDNAs or oligonucleotide sequences that are synthesized in situ to make a DNA chip.

    Molecular signature:

    With the advent of bioinformatics, molecular signatures are a new discipline that uses a variety of approaches to generate a global view of mRNA, protein patterns, and DNA alterations in various cell types. Thus, molecular signatures of disease processes may be seen as distinct from healthy cells, and therapeutic approaches may be tailored on the basis of molecular signature.

    Pathologic complete response (pathCR):

    The absence of any residual tumor cells in a histologic evaluation of a tumor specimen is defined as a complete pathologic response.

    NOTES

    Supported by a Multidisciplinary Research Program Grant from The University of Texas M.D. Anderson Cancer Center, the Cantu, Smith, Dallas, and Park families, and the Rivercreek Foundation.

    Terms in blue are defined in the glossary, found at the end of this article and online at www.jco.org.

    Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.

    REFERENCES

    Jemal A, Murray T, Samuels A, et al: Cancer statistics, 2003. CA Cancer J Clin 53:5-26, 2003

    Ilson DH: Oesophageal cancer: New developments in systemic therapy. Cancer Treat Rev 29:525-532, 2003

    Enzinger PC, Mayer RJ: Esophageal cancer. N Engl J Med 349:2241-2252, 2003

    Pera M: Recent changes in the epidemiology of esophageal cancer. Surg Oncol 10:81-90, 2001

    Bollschweiler E, Wolfgarten E, Gutschow C, et al: Demographic variations in the rising incidence of esophageal adenocarcinoma in white males. Cancer 92:549-555, 2001

    Wild CP, Hardie LJ: Reflux, Barrett's oesophagus and adenocarcinoma: Burning questions. Nat Rev Cancer 3:676-684, 2003

    Winters C Jr, Spurling TJ, Chobanian SJ, et al: Barrett's esophagus: A prevalent, occult complication of gastroesophageal reflux disease. Gastroenterology 92:118-124, 1987

    Pera M: Trends in incidence and prevalence of specialized intestinal metaplasia, Barrett's esophagus, and adenocarcinoma of the gastroesophageal junction. World J Surg 27:999-1008, 2003

    Chirieac LR, Swisher SG, Ajani JA, et al: Posttherapy pathologic stage predicts survival in patients with esophageal carcinoma receiving preoperative chemoradiation. Cancer 103:1347-1355, 2005

    Rohatgi P, Swisher SG, Correa AM, et al: Characterization of pathologic complete response after preoperative chemoradiotherapy in carcinoma of the esophagus and outcome after pathologic complete response. Cancer October 21, 2005 [epub ahead of print]

    Berger AC, Farma J, Scott WJ, et al: Complete response to neoadjuvant chemoradiotherapy in esophageal carcinoma is associated with significantly improved survival. J Clin Oncol 23:4330-4337, 2005

    Darnton SJ, Archer VR, Stocken DD, et al: Preoperative mitomycin, ifosfamide, and cisplatin followed by esophagectomy in squamous cell carcinoma of the esophagus: Pathologic complete response induced by chemotherapy leads to long-term survival. J Clin Oncol 21:4009-4015, 2003

    Lu J, Liu Z, Xiong M, et al: Gene expression profile changes in initiation and progression of squamous cell carcinoma of esophagus. Int J Cancer 91:288-294, 2001

    Xu Y, Selaru FM, Yin J, et al: Artificial neural networks and gene filtering distinguish between global gene expression profiles of Barrett's esophagus and esophageal cancer. Cancer Res 62:3493-3497, 2002

    Xu SH, Qian LJ, Mou HZ, et al: Difference of gene expression profiles between esophageal carcinoma and its pericancerous epithelium by gene chip. World J Gastroenterol 9:417-422, 2003

    Hourihan RN, O'Sullivan GC, Morgan JG: Transcriptional gene expression profiles of oesophageal adenocarcinoma and normal oesophageal tissues. Anticancer Res 23:161-165, 2003

    Zhou J, Zhao LQ, Xiong MM, et al: Gene expression profiles at different stages of human esophageal squamous cell carcinoma. World J Gastroenterol 9:9-15, 2003

    Ishibashi Y, Hanyu N, Nakada K, et al: Profiling gene expression ratios of paired cancerous and normal tissue predicts relapse of esophageal squamous cell carcinoma. Cancer Res 63:5159-5164, 2003

    Dahlberg PS, Ferrin LF, Grindle SM, et al: Gene expression profiles in esophageal adenocarcinoma. Ann Thorac Surg 77:1008-1015, 2004

    McManus DT, Olaru A, Meltzer SJ: Biomarkers of esophageal adenocarcinoma and Barrett's esophagus. Cancer Res 64:1561-1569, 2004

    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 23:1291-1299, 2004

    Kazemi-Noureini S, Colonna-Romano S, Ziaee AA, et al: Differential gene expression between squamous cell carcinoma of esophagus and its normal epithelium; altered pattern of mal, akr1c2, and rab11a expression. World J Gastroenterol 10:1716-1721, 2004

    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 23:4780-4788, 2004

    Zhang L, Miles MF, Aldape KD: A model of molecular interactions on short oligonucleotide microarrays. Nat Biotechnol 21:818-821, 2003

    Eisen MB, Spellman PT, Brown PO, et al: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95:14863-14868, 1998

    Pounds S, Morris SW: Estimating the occurrence of false positives and false negatives in microarray studies by approximating and partitioning the empirical distribution of p-values. Bioinformatics 19:1236-1242, 2003

    Livak KJ, Schmittgen TD: Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25:402-408, 2001

    Ihrie RA, Reczek E, Horner JS, et al: PERP is a mediator of p53-dependent apoptosis in diverse cell types. Curr Biol 13:1985-1990, 2003

    Jetten AM, Suter U: The peripheral myelin protein 22 and epithelial membrane protein family. Prog Nucleic Acid Res Mol Biol 64:97-129, 2000

    Ji J, Zhao L, Wang X, et al: Differential expression of S100 gene family in human esophageal squamous cell carcinoma. J Cancer Res Clin Oncol 130:480-486, 2004

    Nagy N, Brenner C, Markadieu N, et al: S100A2, a putative tumor suppressor gene, regulates in vitro squamous cell carcinoma migration. Lab Invest 81:599-612, 2001

    Hitomi J, Kimura T, Kusumi E, et al: Novel S100 proteins in human esophageal epithelial cells: CAAF1 expression is associated with cell growth arrest. Arch Histol Cytol 61:163-178, 1998

    Hibi K, Fujitake S, Takase T, et al: Identification of S100A2 as a target of the DeltaNp63 oncogenic pathway. Clin Cancer Res 9:4282-4285, 2003

    Chen BS, Wang MR, Cai Y, et al: Decreased expression of SPRR3 in Chinese human oesophageal cancer. Carcinogenesis 21:2147-2150, 2000

    Smolinski KN, Abraham JM, Souza RF, et al: Activation of the esophagin promoter during esophageal epithelial cell differentiation. Genomics 79:875-880, 2002

    Kimos MC, Wang S, Borkowski A, et al: Esophagin and proliferating cell nuclear antigen (PCNA) are biomarkers of human esophageal neoplastic progression. Int J Cancer 111:415-417, 2004

    Kimchi ET, Posner MC, Park JO, et al: Progression of Barrett's metaplasia to adenocarcinoma is associated with the suppression of the transcriptional programs of epidermal differentiation. Cancer Res 65:3146-3154, 2005

    Rohatgi P, Swisher SG, Correa AM, et al: Failure patterns correlate with the proportion of residual carcinoma after preoperative chemoradiotherapy for carcinoma of the esophagus. Cancer 104:1349-1355, 2005

    Abraham JM, Wang S, Suzuki H, et al: Esophagin cDNA cloning and characterization: A tissue-specific member of the small proline-rich protein family that is not expressed in esophageal tumors. Cell Growth Differ 7:855-860, 1996

    Soldes OS, Kuick RD, Thompson IA II, et al: Differential expression of Hsp27 in normal oesophagus, Barrett's metaplasia and oesophageal adenocarcinomas. Br J Cancer 79:595-603, 1999

    Doak SH, Jenkins GJ, Parry EM, et al: Differential expression of the MAD2, BUB1 and HSP27 genes in Barrett's oesophagus-their association with aneuploidy and neoplastic progression. Mutat Res 547:133-144, 2004

    Paweletz CP, Ornstein DK, Roth MJ, et al: Loss of annexin 1 correlates with early onset of tumorigenesis in esophageal and prostate carcinoma. Cancer Res 60:6293-6297, 2000

    Shiraishi T, Mori M, Tanaka S, et al: Identification of cystatin B in human esophageal carcinoma, using differential displays in which the gene expression is related to lymph-node metastasis. Int J Cancer 79:175-178, 1998

    de Angelis PM, Fjell B, Kravik KL, et al: Molecular characterizations of derivatives of HCT116 colorectal cancer cells that are resistant to the chemotherapeutic agent 5-fluorouracil. Int J Oncol 24:1279-1288, 2004

    Van de Velde C, Cizeau J, Dubik D, et al: BNIP3 and genetic control of necrosis-like cell death through the mitochondrial permeability transition pore. Mol Cell Biol 20:5454-5468, 2000

    Fei P, Wang W, Kim SH, et al: Bnip3L is induced by p53 under hypoxia, and its knockdown promotes tumor growth. Cancer Cell 6:597-609, 2004

    Hermeking H, Lengauer C, Polyak K, et al: 14-3-3 sigma is a p53-regulated inhibitor of G2/M progression. Mol Cell 1:3-11, 1997

    Chan TA, Hwang PM, Hermeking H, et al: Cooperative effects of genes controlling the G(2)/M checkpoint. Genes Dev 14:1584-1588, 2000

    Fomenkov A, Zangen R, Huang YP, et al: RACK1 and stratifin target DeltaNp63alpha for a proteasome degradation in head and neck squamous cell carcinoma cells upon DNA damage. Cell Cycle 3:1285-1295, 2004

查询更多Carcinomas相关信息在本站>>

《临床肿瘤学医学期刊》2006年1月第24卷第1期