High Cyclooxygenase-2 Expression Following Neoadjuvant Radiochemotherapy Is Associated with Minor Histopathologic Response and Poor Prognosis in Esophageal Cancer

Esophageal carcinoma is the seventh leading cancer-related death in males in the United States (1). Over the last 20 years, esophageal adenocarcinoma has been the most rapidly increasing cancer in incidence in North America and Western Europe (2, 3).

Patients with locally advanced esophageal cancer have a poor prognosis when treated exclusively by surgical resection. Therefore, neoadjuvant treatment strategies were applied in an effort to improve survival (4). Results from phase III randomized trials are encouraging; however, they revealed that only patients with major histopathologic response will finally benefit from treatment (5–7). In addition, these therapies are expensive and associated with increased therapy-associated complication rates (8). For that reason, molecular markers indicating response or nonresponse to neoadjuvant treatment would be extremely helpful in selecting patients for future treatment protocols.

Cyclooxygenase (COX) is the rate-limiting enzyme for prostaglandin synthesis and encompasses two distinct enzymatic functions: a cyclooxygenase activity that converts arachidonic acid to prostaglandin G₂ and a peroxidase activity that transforms prostaglandin G₂ to prostaglandin H₂. Prostaglandin H₂ is converted to biologically active prostaglandins, such as prostaglandin E₂, by tissue-specific isomerases (9). There are at least two distinct isoenzymes termed COX-1 and COX-2. COX-1 is constitutively expressed in many normal tissues and responsible for various physiologic functions. On the other hand, COX-2 is an inducible gene, originally found to be up-regulated by inflammation or other stimuli such as mitogens, cytokines, various growth factors, and tumor promoters (10–12). COX-2 was shown to be activated in inflammatory cells, vascular endothelial cells, and fibroblasts (13), as well as macrophages (14). High levels of COX-2 have also been reported in a variety of cancer types, including esophageal carcinoma (15).

A recent study showed that high COX-2 expression was associated with increased intratumoral microvessel density and suppression of tumor cell apoptosis in human esophageal squamous cell carcinomas (16). It has been known that cells overexpressing COX-2 tend to be resistant to undergo apoptosis (13). Therefore, esophageal cancer with high COX-2 expression may be less sensitive to radiochemotherapy because the induction of apoptosis is an important mechanism for various anticancer agents as well as radiation therapy (17–19).

The purpose of this study is to investigate the potential association of COX-2 mRNA and protein expression with histomorphologic response to neoadjuvant radiochemotherapy and survival in patients with esophageal carcinoma.

Fifty-two patients with locally advanced, resectable esophageal cancer (cT2-4, Nx, and M₀) were recruited from an ongoing prospective observation trial on neoadjuvant radiochemotherapy for esophageal cancer followed by surgery during 1996 to 2003. None of the patients received prior radiotherapy and/or chemotherapy. All patients [median age, 57.7 years; range, 29.5-72.8 years; gender, 42 (80.8%) men and 10 (19.2%) women] received standardized neoadjuvant radiochemotherapy. Cisplatin (20 mg/m²/d) was administered as short-term infusion on days 1 to 5 and 5-fluorouracil (1,000 mg/m²/d) as continuous infusion over 24 hours on days 1 to 5. Radiation was delivered in daily fractions of 1.8 Gy (days 1-5, 8-12, 15-19, and 22-26) to a total dose of 36 Gy using a multiple field technique. Standardized transthoracic en bloc esophagectomy with two-field lymphadenectomy was done 4 to 5 weeks following completion of radiochemotherapy in all patients (20). Clinical data of the patients are summarized in Table 1. Informed consent was obtained from each patient, and the scientific protocol was approved by the local ethic committee.

Histomorphologic grading of tumor regression. Because clinical response evaluation after neoadjuvant therapy for esophageal cancer is known to be highly inaccurate (21–24), objective histopathologic response analysis was applied using morphologic criteria described by Junker et al. (25). Sections of resected specimens were stained with H&E. These sections were used for both histopathologic staging according to the tumor-node-metastasis classification system (Union Internationale Contra Cancrum, 5th edition, 1997) and histomorphologic evaluation of the effect of radiochemotherapy.

The degree of histomorphologic regression was classified into four categories: grade 1, >50% vital residual tumor cells; grade 2, 10% to 50% vital residual tumor cells; grade 3, near complete regression with <10% VRCT; grade 4, complete regression (pathologic complete remission and ypT₀).

This analysis was done by two independent staff pathologists who were blinded for all other clinical data (S.E.B. and H.P.D.). Regression grades 3 and 4 were considered as major histomorphologic response compared with grades 1 and 2 constituting minor histopathologic response.

Tissue samples. Tumor tissue samples and corresponding normal epithelial tissues for COX-2 mRNA analysis were collected by endoscopic biopsy from 41 patients with esophageal cancer before neoadjuvant treatment. Samples were snap-frozen in liquid nitrogen and stored at −80°C. For immunohistochemical staining, formalin-fixed, paraffin-embedded tumor tissues from endoscopic biopsies before neoadjuvant therapy and from the post-therapeutic resected specimens of 52 patients were available for analysis.

RNA extraction and reverse transcription. Total cellular RNA was extracted using Trizol Reagent (Invitrogen GmbH, Karlsruhe, Germany) in a single-step method. RNA quantity was estimated by spectrophotometric analysis (Smart Spec, Bio-Rad, Hercules, CA). cDNA was generated using 0.5 μg total RNA with oligo (dT)₁₈ primers and Moloney murine leukemia virus reverse transcriptase (BD Biosciences PharMingen, San Diego, CA) according to the manufacture's recommendation. One-microgram human placental RNA (BD Biosciences PharMingen) was used as positive control for reverse transcription and the derived cDNA was applied as standard for the quantitative PCR.

Quantitative real-time reverse transcription-PCR. An amount of 25 ng cDNA was taken for real-time quantitative reverse transcriptase-PCR using the Taqman ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Darmstadt, Germany). β-Actin was amplified as an internal reference gene. Primers used for PCR amplification were chosen to encompass intron between exon sequences to avoid amplification of genomic DNA. The β-actin probe was labeled with a reporter dye (VIC) to the 5′-end of the probe and minor groove binder/nonfluorescent quencher at the 3′-end of the probe (Applied Biosystems). The primers and probe for COX-2 were designed as follows: forward, 5′-CCTTCCTCCTGTGCCTGATG-3′; reverse, 5′-ACAATCTCATTTGAATCAGGAAGCT-3′; probe (FAM labeled), 5′-TGCCCGACTCCCTTGGGTGTCA-3′ (Applied Biosystems). The COX-2 probe was labeled with the reporter dye 6-FAM at the 5′-end and with the quencher TAMRA at the 3′-end. The PCR conditions were 50°C for 2 minutes and 95°C for 10 minutes followed by 40 cycles at 95°C for 15 seconds and 60°C for 1 minute. All assays were run in triplicates.

The quantity of gene expression was calculated based on the standard curves generated by serial dilutions of the standard cDNA (1:1, 1:2, 1:8, 1:32, 1:128, 1:256, and 1:512). Expression of COX-2 mRNA was normalized for β-actin as an internal reference. Absolute expression levels were calculated as COX-2/β-actin in tumor and normal tissues, respectively. Relative gene expression in tumor tissue compared with normal tissue was calculated by dividing the normalized expression in tumor tissue by the expression in normal tissue (COX-2/β-actin in tumor)/(COX-2/β-actin in normal).

Immunohistochemistry. Immunohistochemical staining of COX-2 was done by using the Catalyzed Signal Amplification System (DakoCytomation, Hamburg, Germany), which is based on an avidin-biotin and peroxidase method. Paraffin-embedded tumor tissues from pretherapeutic biopsies and resected specimens were cut into 2-μm sections and mounted onto Superfrost Plus slides (Menzel-Glaeser, Braunschweig, Germany). The specificity of the assay was controlled by simultaneously evaluating negative and positive controls. For negative control staining, 3% normal goat serum was used as a substitute for the primary antibody. As internal positive controls, the staining intensity of inflammatory cells like lymphocytes as well as endothelial cells and smooth muscle cells of the muscularis propria was routinely checked in all slides. These cells consistently expressed COX-2 protein.

Tissue sections were deparaffinized in xylene and rehydrated in graded ethanol. For antigen retrieval, sections were covered with 1% SDS in TBS [100 mmol/L Tris (pH 7.4), 138 mmol/L NaCl, and 27 mmol/L KCl] for 5 minutes at room temperature. Sections were covered with 3% hydrogen peroxide for 5 minutes to deplete endogenous peroxidase. Nonspecific binding sites were blocked by 3% normal goat serum in PBS. Sections were incubated with the mouse monoclonal anti-human COX-2 antibody (BD Biosciences PharMingen) at a dilution of 1:100 (final concentration, 2.5 μg/mL) at 4°C overnight. The sections were then treated with biotinylated rabbit anti-mouse immunoglobulins and freshly prepared streptavidin/biotin/peroxidase complex, respectively. The amplification reagent (biotinyl tyramide and hydrogen peroxide in PBS) was applied before sections were incubated with streptavidin/peroxidase. The reaction product was visualized by amino ethylcarbazole substrate chromogen [amino ethylcarbazole in N, N-dimethylformamide and acetate buffer (pH 5.0); DakoCytomation]. Counter staining was done with hematoxylin.

All slides were analyzed by a staff pathologist (S.E.B.) who was blinded for all clinical data. The results were graded on a scale of 0 to 3 based on the percentage of specific tumor cell staining: grade 0, no specific COX-2 staining or <5% of the tumor cells; grade 1, ≥5% to <35% of the tumor cells; grade 2, ≥35% to <65% of the tumor cells; grade 3, ≥65% tumor cells. Grades 0 and 1 were considered low COX-2 protein expression and grades 2 and 3 as high COX-2 protein expression.

Statistical analysis. Gene expression levels (mRNA) were described using the median as point estimator and the range of values. Cutoff values for discrimination of mRNA expression levels and histopathologic response were derived from receiver operating curve data (area under the curve and the 95% confidence interval) according to Metz et al. (26).

Associations between dichotomized mRNA and protein expression levels and clinicopathologic variables were evaluated using χ² analysis and Fisher's exact test for significance.

Kaplan-Meier plots were used to describe the survival distribution by important clinical variables (27). The log-rank test was used to evaluate for survival differences (28). The level of significance was set to P < 0.05. Unless otherwise specified, Ps were given for two-sided testing.

All statistical tests were done using the Software Package SPSS for Windows, version 11.0 (Chicago, IL).

Histomorphologic regression grading and survival. Distribution of histomorphologic regression grades is displayed in Table 1. There were 32 (61.5%) tumors with minor histopathologic response and 20 (38.5%) with major histopathologic response. Kaplan-Meier survival curves based on histopathologic regression grading (minor histopathologic response and major histomorphologic response) are displayed in Fig. 1 and show a significant difference (P < 0.01) by log-rank testing. At a median follow-up of 14.6 months (range, 2.3-71.3 months) for surviving patients, 5-year survival probabilities are 26.3 ± 8.2% for minor histopathologic response and 58.6 ± 12.9% for major histopathologic response.

Cyclooxygenase-2 mRNA expression. Evaluation for COX-2 mRNA expression was possible in 41 paired patient samples. COX-2 mRNA expression levels did not show a statistically significant difference (P = 0.68) in three different measurements from three different tumor biopsies by Kendall's analysis. COX-2 mRNA expression was detected in all tumors (n = 41) and in 39 of 41 (95.1%) matched normal tissues. The median absolute COX-2 mRNA expression level (COX2/β-actin) was 2.2 (range, 0.07-126.3) in pretherapeutic tumor biopsy tissues and 0.16 (range, 0-5.6) in normal epithelial biopsy tissues, respectively. There is a highly significant difference of COX-2 mRNA expression between paired tumor and normal tissues (P < 0.0001, Wilcoxon rank test). The median relative COX-2 mRNA expression level [(COX-2/β-actin in tumor)/(COX-2/β-actin in normal)] was 6.8 (range, 0.42-963.1). No cutoff values for quantitative absolute and relative COX-2 mRNA expression levels with significant association to dichotomized histomorphologic regression grades (minor and major) were identified.

Cyclooxygenase-2 protein expression. The paired pretreatment biopsies and posttherapeutic resection specimens of 52 patients were available for immunohistochemical analysis. COX-2 immunostaining was observed predominantly in the cytoplasm of tumor cells. Inflammatory mononuclear cells, smooth muscle cells, vascular endothelial cells, and fibroblasts also showed different intensities of COX-2 immunostaining (Fig. 2). In the pretherapeutic biopsy group, 24 (46.2%) cases were grade 0, 16 (30.8%) were grade 1, six (11.5%) were grade 2, and six (11.5%) were grade 3. Therefore, 40 (77%) cases were scored as low COX-2 protein expression and 12 (23%) were scored as high COX-2 protein expression. In the group of posttherapeutic resection specimens, 31 (59.6%) were grade 0, seven (13.5%) were grade 1, four (7.7%) were grade 2, and four (7.7%) were grade 3. There were six (11.5%) cases showing pathologic complete remission, which could therefore not be evaluated in posttherapeutic resection specimens. Cases were categorized into two groups: low COX-2 protein expression (grade 0 or 1) and high protein expression (grade 2 or 3). Thirty-eight (73%) cases were scored as low COX-2 protein expression and eight (15.4%) were scored as high COX-2 protein expression.

Regulation of cyclooxygenase-2 protein expression during neoadjuvant therapy.Table 2 shows the comparison of COX-2 immunostaining between pretherapeutic biopsy tissues and posttherapeutic resection tissues. There was a statistically significant difference (P < 0.006) between the two groups. Besides the six cases with pathologic complete remission, there were 31 of 46 tumors that did not show regulation of COX-2 immunostaining during neoadjuvant therapy. Compared with pretherapeutic biopsies, 15 of 46 tumors presented regulation of COX-2 immunostaining, with 12 tumors showing down-regulation of COX-2 protein expression in resected specimens and three tumors showing up-regulation of COX-2 expression following neoadjuvant radiochemotherapy.

Pretherapeutic cyclooxygenase-2 protein expression and histopathologic regression. There were no significant associations between dichotomized COX-2 protein expression levels in pretherapeutic biopsies and clinical variables (ypT- and ypN categories and grading) or histopathologic response to neoadjuvant therapy and survival.

Posttherapeutic cyclooxygenase-2 protein expression and histopathologic regression. A pathologic complete remission was observed in 6 of 52 tumors, which could therefore not be evaluated for posttherapeutic COX-2 protein expression, because no viable tumor cells were present by definition. Forty-six of 52 samples were evaluable for posttherapeutic COX-2 protein expression. High COX-2 protein expression was only observed in ypT3 tumors (17.4%, P < 0.03). There was a statistically significant association of posttherapeutic COX-2 protein expression and histopathologic tumor regression (P < 0.04), and data are shown in Table 3. In the group of high COX-2 protein expression following neoadjuvant radiochemotherapy, there were no cases with a major histomorphologic regression. All patients (n = 8) with high posttherapeutic COX-2 expression died from tumor progression during follow-up (median survival, 16 months; minimum, 2.2 months; maximum, 35.3 months). The respective Kaplan-Meier curves are shown in Fig. 3. The individual clinicopathologic data of these eight patients are summarized in Table 4.

We analyzed the association of pretherapeutic COX-2 mRNA and pretherapeutic and posttherapeutic COX-2 protein expression with histomorphologic regression and survival following neoadjuvant radiochemotherapy and surgical resection in patients with locally advanced esophageal cancers.

We applied histomorphologic criteria and showed that histopathologic tumor regression grades were significantly associated with survival (P < 0.01). Therefore, an objective response evaluation system was available for the analysis of potential associations with COX-2 mRNA or protein expression levels.

Overexpression of COX-2 has been reported in several human malignancies including esophageal cancer (15, 29–32). In this study, a quantitative real-time reverse transcriptase-PCR method was used to analyze COX-2 mRNA expression in biopsy tissues of patients with esophageal carcinoma. Consistent with previous reports, our results show that COX-2 mRNA expression was increased significantly in tumor tissues before therapy compared with matched normal tissues (P < 0.0001). Because we did not do laser microdissection, which is difficult to achieve without degradation of the RNA, we cannot rule out that differences between samples were influenced by COX-2 expressing inflammatory or endothelial cells that were consistently present as shown by immunohistochemistry.

To date, no reports were published concerning the relationship between COX-2 mRNA expression and prognosis in esophageal cancer following neoadjuvant radiochemotherapy and surgical resection. Our data show that both absolute and relative COX-2 mRNA expression levels in pre-therapeutic tumor biopsies were not associated with histomorphologic regression to neoadjuvant therapy or survival.

There was a statistically significant difference (P < 0.006) of COX-2 protein expression between pretherapeutic biopsy tissues and posttherapeutic resection tissues. Pretherapeutic COX-2 protein expression did not show a correlation with histomorphologic tumor regression or survival. The same results were reported for patients with cervical cancer, where COX-2 protein expression was not associated with histopathologic regression to neoadjuvant chemotherapy or radiochemotherapy (33).

The most interesting finding of our study was that high COX-2 protein expression in posttherapeutic resected specimens is significantly associated with the degree of histomorphologic tumor regression induced by neoadjuvant radiochemotherapy and poor prognosis. All patients with high COX-2 protein expression levels in posttherapeutic resected specimens showed exclusively minor histomorphologic regression and died from tumor progression during follow-up (median survival, 16 months).

One previous report showed that elevated expression of COX-2 protein was associated with significantly reduced survival in patients with esophageal adenocarcinomas; however, these patients did not receive neoadjuvant treatment (31). With respect to neoadjuvant therapy, one recent report also proved that overexpression of COX-2 correlated with a poor prognosis in patients with squamous cell carcinoma of the uterine cervix treated with radiation and concurrent chemotherapy (34).

Accumulated evidence indicates that selective COX-2 inhibitors show antitumor activity and synergistic effects with chemotherapy and radiotherapy. It is well known that selective COX-2 inhibitors reduce proliferation and increase apoptosis in esophageal adenocarcinoma and squamous cell carcinoma cell lines (15, 35).

Hida et al. (36) showed that a selective COX-2 inhibitor enhanced the cytotoxicity of several chemotherapeutic agents in vitro, in part by induction of apoptosis. Recently, Hashitani et al. (37) reported that the selective COX-2 inhibitor celecoxib inhibits cell proliferation, induces apoptosis, and enhances sensitivity to anticancer drugs in human squamous cell carcinoma and adenocarcinoma cell lines.

In tumor cell cultures and animal models, selective COX-2 inhibitors could enhance tumor regression to radiation both in vitro and in vivo (38, 39). Thus, selective COX-2 inhibitors may have a potential as radiosensitizers in the treatment of human cancers (40). The radiosensitization of tumors by selective COX-2 inhibitors may result from an increase in intrinsic cell radiosensitivity and inhibition of tumor angiogenesis (40, 41).

In conclusion, pretherapeutic overexpression of COX-2 mRNA was not significantly associated with either histomorphologic regression or survival. Comparison of COX-2 protein expression in pretherapeutic tumor biopsies and post-therapeutic resected specimens showed a significant difference in COX-2 expression during neoadjuvant radiochemotherapy. High COX-2 protein expression in posttherapeutic resected specimens was significantly associated with minor histomorphologic response to neoadjuvant therapy and very poor survival probabilities. Because selective COX-2 inhibitors have been shown to enhance the sensitivity to radiotherapy or chemotherapy in various in vitro and in vivo models, it might be useful to evaluate if selective COX-2 inhibitors could either prevent up-regulation of COX-2 expression or down-regulate intrinsically high COX-2 expression to improve response and survival in patients treated with neoadjuvant radiochemotherapy for esophageal cancer. Given the preliminary nature of these data, a larger prospective study seems justified to strengthen the data before clinical consequences should be drawn.