Abstract
Malignant mesothelioma (MM) is a fatal tumor of increasing incidence, which is resistant to current therapy. Cyclooxygenase-2 (COX-2) plays an important role in solid tumor growth, invasiveness, and angiogenesis, in part through the synthesis of prostaglandins such as prostaglandin E2 (PGE2). In a prospective study, we evaluated COX-2 expression in snap-frozen, surgically resected MM tissue specimens using immunohistochemistry and semiquantitative Western blotting. PGE2 was assessed by enzyme immunoassay. Thirty epithelioid, 10 biphasic, and 8 sarcomatoid tumors were evaluated. Immunohistochemistry demonstrated strong cytoplasmic tumor cell and variable stromal staining in all of the cases. COX-2 protein levels were correlated with clinicopathological prognostic factors using Kaplan-Meier and Cox proportional hazards models. High COX-2 band densitometry values correlated with poor survival (P = 0.008). In multivariate analysis, high COX-2 expression (P = 0.0005), nonepithelioid cell type (P = 0.002), and chest pain (P = 0.04) were independent predictors of poor prognosis. Furthermore, COX-2 expression contributed in multivariate analysis to both European Organization for Research and Treatment of Cancer (P = 0.001) and Cancer and Leukemia Group B (P = 0.003) prognostic scoring systems. The presence of PGE2 was demonstrated in all of the samples. These results suggest that COX-2 expression is a prognostic factor in MM. COX-2 is a potential therapeutic target in MM, and trials are required of COX-2 inhibitors alone or in combination with existing treatment modalities.
INTRODUCTION
MM3 is a fatal cancer of increasing incidence associated with asbestos exposure (1) . MM responds poorly to surgery, chemotherapy, and radiotherapy (2) , and has an appalling prognosis. The median survival is approximately 6–12 months (3, 4, 5) . TNM staging (6) is of limited value in the majority of patients. In contrast, evaluation of clinicopathological features of MM has led to the formation of prognostic scoring systems. These have been derived from multivariate analyses of patients in chemotherapy trials conducted by the EORTC (4) and CALGB (5) . We have validated previously these systems in a surgical series (3) . Furthermore, investigation of biological markers of prognosis has attracted interest in other solid tumors and may provide information independent from TNM stage (7) .
Cyclooxygenases catalyze the initial, rate-limiting steps of prostaglandin synthesis from arachidonic acid (8) . Compared with its isoform COX-1, COX-2 leads preferentially to the formation of prostaglandins such as PGE2 (9) . COX-2 has been implicated in carcinogenesis through the down-regulation of cell-mediated immunity, promotion of angiogenesis, and the formation of carcinogenic metabolites such as malondialdehyde (10 , 11) . COX-2-expressing cancer cell lines are associated with increased proliferative and invasive potential (12) . COX-2 overexpression has been noted in many solid tumors, including colorectal (13) , breast (14) , gastric (15) , esophageal (16) , lung (17) , and brain (18) tumors. Selective inhibition of COX-2 is a novel therapy under investigation in both the chemoprevention and treatment of solid tumors (19 , 20) .
This study evaluated the expression of COX-2 in MM by both immunohistochemistry and Western blotting, and its product, PGE2, by enzyme immunoassay. The contribution of COX-2 expression to clinicopathologic prognostic factors, and the CALGB and EORTC prognostic scoring systems was assessed.
PATIENTS AND METHODS
Patients.
In a prospective study, patients referred to the regional Department of Thoracic Surgery for surgical biopsy, management of pleural effusion or empyema, or for radical surgery for MM were identified before their operative procedure. Case notes were reviewed and patients interviewed to retrieve relevant demographic, clinical, and pathological data. Prediagnostic variables, such as performance status and hematological indices, were taken immediately before surgery. After surgery, the detailed histopathology report was examined in each case and the pathologic IMIG stage (6) derived, where possible. Clinicopathologic prognostic factors, including CALGB (5) and EORTC (4) prognostic groups, were derived in all of the cases, as described previously (3) . In regard to the CALGB prognostic system, groups 1 and 2, 3 and 4, and 5 and 6 were combined, respectively, for statistical analyses, because of the low numbers of patients in the even-numbered groups. Causes of death and patterns of recurrence were noted, and cancer-specific survival was calculated from the date of the diagnostic biopsy.
Samples.
MM specimens were collected at video-assisted thoracoscopic biopsy or at thoracotomy, inspected macroscopically by a histopathologist, and representative blocks of tumor were snap-frozen and stored in liquid nitrogen. Histopathology slides were reviewed to assess the presence of suitable tumor for immunohistochemistry within each block.
Immunohistochemistry.
Sections 7-μm thick were cut from snap-frozen tissue blocks with a cryostat, maintaining a chamber temperature of −27°C, before mounting on silane-treated slides and drying overnight at −20°C. Slides were fixed with acetone, washed, and permeabilized in 0.1% Triton X-100 (Sigma, Gillingham, United Kingdom). Nonspecific binding was blocked with 10% normal rabbit serum. Sections were incubated overnight at 4°C with a monoclonal COX-2 primary antibody (SC-1745; Santa Cruz Biotechnology/Autogen Bioclear, Calne, United Kingdom) at a dilution of 1:80. An FITC-labeled secondary antibody (F2016; Sigma) was used at a dilution of 1:50. COX-2 immunostaining was visualized with laser scanning confocal microscopy (Leica TCS4D) at low and high power. Omission of the primary antibody was used for negative controls.
Western Blotting.
Snap-frozen samples were homogenized mechanically in a buffer [150 mM sodium chloride, 0.1 M Tris (pH 8), 1% Tween-20, 50 mM diethyldithiocarbamic acid, 1 mM EDTA pH 8 (Sigma)] containing protease inhibitors, before sonication and centrifugation at 4°C for 3 min. A Bradford assay (Bio-Rad, Hemel Hempstead, United Kingdom) was used to determine the protein concentration of each supernatant. Samples were loaded into a 10% SDS-PAGE, to give 150 μg protein/well. Once through the stacking gel, proteins were resolved at 150 V for 4 h. Transfer to a nitrocellulose membrane (Hypobond-ECL; Amersham Pharmacia Biotech, Amersham, United Kingdom) was performed by semi-dry electroblotting (Bio-Rad). Complete protein transfer was confirmed by staining both gel and membrane with Ponceau S (81462; Sigma). The blotted nitrocellulose membrane was blocked in 10% milk/Tris-buffered saline wash buffer [0.05 M Tris (pH 7.5), 0.15 M sodium chloride, 0.1% Tween-20 (Sigma)] overnight at 4°C. The membrane was probed with a monoclonal COX-2 primary antibody (SC-1745; Santa Cruz Biotechnology/Autogen Bioclear) for 90 min at room temperature. After washing, the membrane was incubated with a horseradish peroxidase-conjugated secondary antibody (SC-2020; Santa Cruz Biotechnology) for 1 h at room temperature. The membrane was washed and developed using ECL (Amersham Pharmacia Biotech) following the manufacturer’s protocol. The membrane was exposed to Hyperfilm ECL (Amersham Pharmacia Biotech) for between 2 and 20 min. Developed films were scanned and band densitometry calculated on an densitometer (Kodak Digital Science Image Station 440CF camera and ID software; PerkinElmer Life Sciences, Cambridge, United Kingdom). Background activity, the mean value of the perimeter of each template cell, was subtracted from each sample. Four samples were chosen in addition to the COX-2 standard (NP04; Oxford Biomedical Research/Biogenesis, Poole, United Kingdom) for use as internal positive controls and run on each gel. All of the gels were run under identical conditions with the same batch of reagents and densitometry results standardized between gels. COX-2 densitometry results were also normalized to membranes stripped and reprobed for α-tubulin (T-9026 primary antibody (Sigma) and P0260 secondary antibodies (Dako, Ely, United Kingdom). The semiquantitative nature of this protocol was validated by densitometric analysis of serial dilutions of the two samples with the strongest COX-2 bands. The specificity of the antibody was confirmed by preabsorbing the COX-2 primary antibody with a COX-2 blocking peptide (SC-1745P; Santa Cruz Biotechnology) overnight at 4°C.
Bicyclo-PGE2 Enzyme Immunoassay.
Because of its short half-life, PGE2 and its physiological metabolites were measured by derivatization to their stable moiety, Bicyclo-PGE2, and subsequent enzyme immunoassay, according to the manufacturer’s protocol (Cayman Chemical/Alexis Biochemicals, Nottingham, United Kingdom). Briefly, homogenized tumor supernatants were derivatized to bicyclo-PGE2 overnight in a 1 M sodium bicarbonate solution. Bicyclo-PGE2 was then measured using an enzyme immunoassay.
Statistical Analysis.
Clinicopathologic factors were assessed with Kaplan-Meier and log rank analyses. Prognostic variables on univariate analysis (P < 0.1) were entered into Cox proportional hazards models to examine HRs and perform multivariate analysis, as described previously (3) . Patients dying within 30 days were excluded from survival analysis to avoid bias from postoperative deaths. The impact of COX-2 expression, in terms of Western blot densitometry and immunohistochemical staining, was incorporated into these models both as a continuous and a categorical variable.
RESULTS
Surgical Procedures and Outcome.
Radical surgery [with intent of complete microscopic (R0) resection] was performed in 23 and palliative surgery or diagnostic surgical biopsy in 25 patients. Two patients died within 30 days: 1 from respiratory failure on day 10 and 1 from cardiac arrest after acute intrathoracic hemorrhage on day 11. Overall survival was median 8.1 (range, 0.3 to 27.0) months. Six- and 12-month survival rates were 66% and 39%, respectively. Patients who received radical surgery had a longer survival than those who had palliative or diagnostic surgery (P = 0.02; log rank). However, undergoing radical surgery was not an independent predictor of good outcome in subsequent multivariate Cox analyses. In univariate analyses, preoperative clinicopathologic prognostic factors significant with the log rank test were weight loss of >5% (P = 0.007), nonepithelioid cell type (P = 0.002), and hemoglobin <14 g/dl (P = 0.02). There was a trend toward poor survival with the presence of chest pain (P = 0.06) in this cohort of patients. Male gender, Eastern Cooperative Oncology Group performance status >0, WBC count >8.3 × 109/liter, thrombocytosis, and IMIG TNM stage were not predictors of survival. Both the CALGB (P = 0.001) and EORTC (P = 0.004) prognostic scoring systems stratified survival appropriately according to their risk groups with Kaplan-Meier analysis.
Immunohistochemistry and Western Blot Analysis.
Eighteen cases were evaluated with immunohistochemistry. There was strong cytoplasmic tumor cell staining in all of the cases (Fig. 1)⇓ . Surrounding stroma stained with variable immunointensity. With Western blotting, immunoreactive bands were seen, corresponding with the COX-2 standard, at Mr 72,000 and/or Mr 74,000, indicating different glycosylation states (Fig. 2)⇓ . Immunoreactivity was lost when the primary antibody was blocked with a blocking peptide. The mean COX-2 band densitometry value was 66,000 (SD 83,500) units. Serial dilutional studies confirmed that the relationship between COX-2 load and band densitometry was linear within the range encountered. Correcting the results between the individual gels did not affect the rank order or categories of samples. Similarly, there was a strong correlation between the COX-2 densitometry values alone and those corrected for α-tubulin (r = 0.92; P < 0.001). There were significant correlations between weight loss >5%, and both COX-2 expression (P = 0.004) and the ratio to α-tubulin (P = 0.01). There was no correlation between either COX-2 or COX-2:α-tubulin ratio and IMIG TNM stage in the 32 cases in which the latter could be derived accurately.
COX-2 immunofluorescence. There is strong immunostaining in epithelial MM tumor islets (large arrow) but weaker staining in surrounding stroma (small arrow; bar, 40 μm).
Western blotting. The top gel shows the variations in COX-2 band density for the constant protein load (PC, positive control). In the middle gel, the effects of preabsorbing the COX-2 primary antibody with a specific COX-2 blocking peptide are seen; there is minimal immunoreactivity for the positive control, but the COX-2 bands in all of the samples are blocked. The bottom gel demonstrates the bands obtained after probing for α-tubulin.
Prognostic Impact of COX-2.
In univariate analysis, high COX-2 band densitometry values correlated with poor survival (Fig. 3)⇓ when assessed as a categorical variable [greater or less than median value; P = 0.008; HR, 2.9 (95% CI, 1.3–6.3); Cox regression; Fig. 4⇓] or as a continuous variable (P = 0.003). When normalized to the corresponding α-tubulin band, the mean COX-2:α-tubulin ratio was 0.22 (SD 0.25). A COX-2:α-tubulin ratio greater than the median value was associated with poor prognosis [P = 0.02; HR, 2.7 (95% CI, 1.2–5.7)].
Scatterplot showing the inverse relationship of COX-2 Western blot band densitometry and survival.
Kaplan-Meier survival curves at a cut point of the median densitometry units (P = 0.0057; log rank). One patient in each group died before 30 postoperative days had passed, and these have been excluded from analysis.
In multivariate Cox analysis with the clinicopathologic factors (identified by univariate analysis), COX-2 expression, cell type, and the presence of chest pain at diagnosis were the only independent prognostic factors (Table 1)⇓ . The presence of weight loss, low hemoglobin, and patients who did not receive radical surgery did not contribute as poor prognostic factors to the multivariate model (Table 1)⇓ . In regard to the CALGB and EORTC prognostic scoring systems, COX-2 contributed independently to both (Table 2)⇓ . In all of the preceding multivariate analyses, COX-2 was an independent prognostic factor when entered as a continuous variable. Furthermore, COX-2 was also significant in multivariate analyses with clinicopathologic factors or prognostic scoring systems when analyzed as the ratio to α-tubulin, both as a categorical and continuous variable.
Multivariate Cox proportional hazards model with clinicopathologic prognostic factors
COX-2 values greater than the median were a poor prognostic factor independent of cell type. The presence of preoperative weight loss, hemoglobin <14.0 g/dl, or not undergoing radical surgery were not significant factors in the multivariate model.
Multivariate Cox proportional hazards models
COX-2 values greater than the median were a poor prognostic factor independent of both the CALGB (a) and EORTC (b) prognostic scoring systems. The six CALGB prognostic groups were combined, because there were insufficient numbers in the even-numbered groups for analysis.
Bicyclo-PGE2 Enzyme/Immunoassay.
Presence of Bicyclo-PGE2 was demonstrated in all of the samples with median 1.8 (range, 0.34–5.6) pg/mg protein. No correlations with either COX-2 protein levels or survival were seen.
DISCUSSION
This study demonstrates that COX-2 is expressed in MM and can be detected using immunohistochemistry and Western blotting techniques. Furthermore, PGE2 was also detected in all of the samples. An initial report of COX-2 expression in MM failed to demonstrate COX-2 protein in either tumor samples or cell lines by Western blotting (21) . Although we found little variability in MM tumor cell immunofluorescence using immunohistochemistry, high COX-2 levels on Western blotting correlated with a worse prognosis in univariate and multivariate analyses. No correlation was seen between COX-2 Western blot densitometry values and the pathological IMIG TNM stage. This is in keeping with the lack of impact of the IMIG staging system on prognosis in this and other patient series (22) , an observation suggesting that the biological behavior of the tumor is more important than disease extent in determining patient outcome.
The prognostic significance of COX-2 tumor cell expression has been examined in other cancers. With immunohistochemistry, COX-2 overexpression has been correlated with colorectal tumor growth (23) , lymph node metastasis (13) , and recurrence (24) , although results were not consistent between studies. COX-2 expression has been proposed as a significant poor prognostic factor in colorectal and gastric cancer but only with univariate analysis (25 , 26) . In non-small cell lung cancer, COX-2 immunostaining has been found to be associated with a significantly worse prognosis in stage I adenocarcinomas (27) .
Clinicopathologic correlations with COX-2 protein levels, as assessed by Western blotting and semiquantitative densitometry, have been less well characterized. COX-2 overexpression, according to this method, has been described in 15 gastric tumor samples, compared with paired normal mucosa. Samples with a COX-2 densitometry ratio between tumor and normal mucosa >2 were associated with lymphatic invasion, lymph node metastasis, and TNM stage (15) . Similarly, COX-2 Western blot densitometry values are greater in Barrett’s esophagus and esophageal adenocarcinoma than in normal tissues (28) . Our study is the first to demonstrate by Western blotting that increasing COX-2 protein levels in tumor tissue correlate with poor outcome. Furthermore, COX-2 protein levels were an independent prognostic factor in multivariate analyses and contributed to both the CALGB and EORTC prognostic scoring systems, suggesting that COX-2 is an important factor in MM.
COX-2 inhibitors have a number of beneficial effects in solid tumors. COX-2 inhibition has been shown to increase tumor cell apoptosis (29 , 30) , and reduce proliferation (31) , invasion (12) , and angiogenesis (32, 33, 34) . COX-2 inhibitors are an effective means of chemoprevention in a number of carcinogenic models (35, 36, 37, 38, 39, 40) .
A number of previous experimental studies support a role for COX activity in the pathogenesis of MM. Asbestos causes release of PGE2 from alveolar macrophages and inhibition of mesothelial cell-mediated cytotoxicity (41 , 42) . Indomethacin, an inhibitor of both COX-1 and COX-2 activity, has been demonstrated to restore the depressed lymphokine-activated killer cell activity seen in patients with MM in an ex vivo model (43) . Therefore, COX-2 may be implicated in the T-cell anergy seen in MM (44) . The suppression of antitumor immune responses may be reversed by administration of COX-2 inhibitors (45) .
We have observed previously increased expression and activation of the EGFR in pleural mesothelial cells after exposure to carcinogenic asbestos fibers (46) . EGFR autophosphorylation results in activation of the transcription factor nuclear factor κB (47) . A nuclear factor κB binding site is present in the promoter region of the COX-2 gene (48) . In keeping with this EGFR activation has been demonstrated to be associated with up-regulation of COX-2 expression (49 , 50) . A number of other growth factors, which play important roles in the pathogenesis of MM, have been shown to induce COX-2 expression. These include hepatocyte growth factor/scatter factor (51 , 52) , transforming growth factor β (53 , 54) , and platelet-derived growth factor (55 , 56) . Collectively these data suggest an important role for COX-2 in MM carcinogenesis.
In conclusion, increasing levels of COX-2 protein, as assessed by Western blot analysis, are a poor prognostic factor in MM, which contributes independently to the CALGB and EORTC prognostic scoring systems. These data support an important role for COX-2 in pathogenesis of this malignancy. The cell signaling pathways involved in the regulation of this important immunomodulatory and tumor-promoting factor require additional investigation. The inhibition of COX-2 is a potential novel therapeutic target, alone or in combination with cytotoxic chemotherapy, for the management of MM. Furthermore, COX-2 inhibitors may have a role to play in the chemoprevention of the disease.
Acknowledgments
We thank the Institute of Cancer Studies and the Institute for Lung Health, Leicester, United Kingdom.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 Supported by grants from Glenfield Hospital Research and Development, the June Hancock Memorial Fund, and the Sir Samuel Scott of Yews Trust. J. G. E. is supported by a Leicester Royal Infirmary Research Fellowship.
↵2 To whom requests for reprints should be addressed, at University Department of Oncology, Osborne Building, Leicester Royal Infirmary, Leicester, LE1 5WW, United Kingdom. Phone: 6-258-7602; Fax: 6-258-7599; E-mail: kobyrne{at}uhl.trent.nhs.uk
↵3 The abbreviations used are: MM, malignant mesothelioma; CALGB, Cancer and Leukemia Group B; CI, confidence interval; COX, cyclooxygenase; EORTC, European Organization for Research and Treatment of Cancer; HR, hazard ratio; EGFR, epidermal growth factor receptor; IMIG, International Mesothelioma Interest Group; ECL, enhanced chemiluminescence; TNM, Tumor-Node-Metastasis; PGE2, prostaglandin E2.
- Received August 23, 2001.
- Revision received February 22, 2002.
- Accepted February 25, 2002.
References
Volume 8, Issue 6
