Expression of Cyclooxygenase-2 in Renal Cell Carcinoma: Correlation with Tumor Cell Proliferation, Apoptosis, Angiogenesis, Expression of Matrix Metalloproteinase-2, and Survival¹

COX³ is involved in the production of prostaglandins from arachidonic acid (1). Prostaglandins, such as prostaglandin H2, can be converted to several eicosanoids, including thromboxanes and malondialdehyde, or into other prostaglandins by specific isomerases. Several prostaglandins, such as prostaglandin E₂ and prostaglandin I₂, are known to promote carcinogenesis by increasing DNA synthesis and cell proliferation (2). Two isoforms of COX have been identified: COX-1 and COX-2. The former is constitutively expressed in most tissues (3), whereas COX-2 is not readily detectable in normal tissues under unstimulated conditions. However, COX-2 is up-regulated by a variety of stimuli including cytokines (4), growth factors (5), and oncogenes (6). In the human kidney, COX-2 is detected under certain conditions, such as aging and physiological stress, in both the cortex and medulla (7). Recent studies showed that COX-2 expression may correlate with tumor cell growth in canine renal cell carcinoma (8); however, there is no report on COX-2 expression in human renal cell carcinoma tissues.

Several epidemiological and animal studies have reported that nonsteroidal anti-inflammatory drugs (which are COX-2 inhibitors) can reduce the risk of colorectal cancer (9, 10, 11, 12, 13). Other studies reported that inactivation of COX-2 and the use of selective COX-2 inhibitors can result in 60% suppression of the incidence of intestine polyps in adenomatous polyposis coli mutant mice, a model of human familial adenomatous polyposis (14). These results provide evidence that COX-2 plays an important role in carcinogenesis. In fact, COX-2 overexpression has been detected in a variety of malignancies (15, 16, 17, 18, 19), and has been reported to be associated with tumor growth, resistance to apoptosis, angiogenesis, and tumor invasiveness (20, 21, 22, 23, 24). Furthermore, recent studies demonstrated the involvement of MMP-2 expression in tumor invasion and metastasis in renal cell cancer (25, 26, 27). On the basis of these studies, it is conceivable that COX-2 expression may correlate with the prognosis and survival of patients with renal cell carcinoma. However, the relationships between COX-2 expression and tumor growth and survival have not been examined previously.

In the present study, we investigated the relationships between COX-2 expression with several clinicopathological factors: cell proliferation, angiogenesis, MMP-2 expression, and survival in patients with renal cell carcinoma.

Clinicopathological features were reviewed in 131 patients examined between January 1988 and December 2001 at Nagasaki University Hospital. All of the tissue sections were obtained by surgical resection. Patients who received neoadjuvant therapy, e.g., renal artery embolization and immunotherapy, were excluded from this study. All of the patients were evaluated by chest X-ray, ultrasonography, computed tomography, magnetic resonance imaging, and bone scanning for tumor staging. Renal cell carcinomas were staged according to 1997 Tumor-Node-Metastasis staging system (28). Nuclear grading was based on the criteria of Fuhrman et al. (29) and divided for statistical evaluation into three groups: G1, G2, and G3 + 4. Pathological examination of the tumor specimens was performed by a single pathologist (T. H.). The study protocol was approved by the Human Ethics Review Committee of Nagasaki University School of Medicine.

Immunohistochemical staining was performed on sections from formalin-fixed and paraffin-embedded tumor specimens. Tissue sections (5-μm thick) were deparaffinized with three changes of xylene and rehydrated in graded ethanol solutions. After deparaffinization, antigen retrieval treatment was performed for anti-Ki-67 antibody and anti-CD31 antibody, or anti-COX-2 antibody and anti-MMP-2 antibody, respectively, at 121°C for 15 min or 95°C for 40 min in 0.01 m sodium citrate buffer (pH 6.0). Blocking of endogenous peroxidase was performed using 2% hydrogen peroxide for 30 min. The primary antibodies and staining procedure are summarized in Table 1. Nonspecific binding was blocked by incubation with PBS containing 5% skin milk, 2% BSA, and normal goat serum for 60 min. After incubation with primary antibody, the sections were washed extensively, and incubated with biotinylated antimouse IgG for anti-Ki-67 antibody, anti-CD31 antibody, and anti-MMP-2 antibody, or antirabbit IgG for anti-COX-2 antibody, followed by an incubation with horseradish peroxidase-conjugated avidin. Peroxidase was visualized by use of the liquid 3,3′-diaminobenzidine substrate kit (Zymed Laboratories, Inc., San Francisco, CA), and sections were counterstained with hematoxylin. As negative controls for all of the antibodies, sections were incubated with PBS instead of the primary antibody. Control sections of COX-2 expression were also incubated with antisera in the presence of 100-fold excess of the human recombinant COX-2 protein (Cayman Chemical, Ann Arbor, MI).

After deparaffinization and rehydration, tissue sections were incubated in 20 μg/ml proteinase K (Roche Diagnostics, Mannheim, Germany) for 15 min at room temperature. The sections were then washed in PBS and immersed for 5 min in a solution of 3% H₂O₂ to inactivate endogenous peroxidase. Detection of apoptosis in situ was determined using the Apop Tag In Situ Apoptosis Detection kit (Intergen Company, Purchase, NY), based on TUNEL, using the instructions provided by the manufacturer. Slides were counterstained with 2% methyl green. Negative controls consisted of consecutive sections in which the terminal deoxynucleotidyl transferase enzyme was omitted.

All of the analyses of immunohistochemical staining and TUNEL method were assessed by light microscopy within the tumor area. Necrotic areas, such as those with prominent hyalinization and hemorrhagic region, were excluded. COX-2 expression was evaluated according to previous reports (30, 31). COX-2 expression was considered to be positive when >5% of cancer cells showed clear staining. MMP-2 expression was evaluated according to Kallakury et al. (32), and the staining intensity was graded semiquantitatively into weak, moderate, or intense staining. The extent of positive staining in the tumor area was graded as focal (≤10%), regional (11–50%), or diffuse (>50%). The staining pattern of moderate diffuse, intense regional, and intense diffuse was considered to be positive. Ki-67 LI represented the percentage of positive cells (≥1000 tumor cells in three to five different fields per section were calculated under 400-fold magnification). The AI represented the number of TUNEL-positive cells among ≥1000 tumor cells. The MVD was estimated by counting the number of CD31-positive vessels in the tumor area representative of the highest MVD at ×200 magnification. Tumors with Ki-67 LI, AI, and MVD values above the median value were considered as high-index group, and those equal to or less than median value were considered as low-index group for statistical evaluation. Slides were evaluated twice at different time intervals by two investigators who were blinded to the clinical characteristics and survival.

Normality was evaluated by normal distribution and histograms for each variable. Because the data generated from immunohistochemical staining were not normally distributed, all of the data were expressed as medians with interquartile ranges, and the Mann-Whitney t test was used for analysis of continuous variables. The χ² test was used for the categorical comparison of the data. The crude and adjusted effects on nuclear grade and immunohistochemical staining, as well as other risk factors, were estimated by logistic regression analysis, and were described as ORs with 95% CIs, together with the Ps. The 5-year survival rates were compared with Kaplan-Meier analysis and the log rank test. Variables that achieved statistical significance (P < 0.05) in the univariate analysis were subsequently entered into a multivariate analysis using a Cox proportional hazard analysis. All of the statistical tests were two-sided, and significance was defined as P < 0.05. All of the statistical analysis was performed on a personal computer with the statistical package Stat View for Windows (Version 5.0).

In this study, we examined tumors from 91 men and 40 women, ranging from 34 to 88 years of age (median age, 62 years). Among the 131 tumors examined, 110 tumors (84.0%) were conventional (clear cell carcinoma), 19 tumors (14.5%) were papillary renal carcinoma, 5 tumors (3.8%) were chromophobe renal carcinoma, and 6 tumors were unclassified renal cell carcinoma. There was no tumor identified as a collecting duct carcinoma, and only 2 tumors showed sarcomatoid change. With regard to treatment, 118 patients underwent radical nephrectomy, whereas 12 underwent partial resection including 2 that underwent bilateral partial resection. The median follow-up period was 43 months (range, 2–98; interquartile range, 32–65 months).

COX-2 expression was noted in some parts of the normal kidney tissue, particularly in tubules (Fig. 1,A). However, in the specimens positive for COX-2 expression, almost all showed no or very weak staining in the tubules in the normal tissues. Glomeruli and Bowman’s capsules were not stained for COX-2. Of the 131 sections, 70 (53.4%) were positive for COX-2, which showed diffuse staining of tumor cell cytoplasm (Fig. 1,B). Importantly, COX-2 staining was reduced by competition with recombinant human COX-2 protein (Fig. 1,C). In low-stage carcinomas, only a part of tumor cells were stained (Fig. 1,D), and such staining was not detected in negative control slides (Fig. 1,E). Papillary and chromophobe carcinomas also showed diffuse staining of tumor cell cytoplasm (Fig. 1, F and G, respectively).

As shown in Table 2, COX-2 expression was associated significantly with tumor status, including T, N, and M stage, and tumor grade. The majority of COX-2-negative cases (90.1%) were patients with T₁ disease. In contrast, 23 of 78 patients (32.9%) with T₁ disease were positive for COX-2 expression, whereas 41.4% of patients with T₃ disease were positive for COX-2 expression, representing the highest percentage of those that were COX-2 positive. Patients with tumors negative for COX-2 had no lymph node metastases. Furthermore, none of the tumors negative for COX-2 were from patients with tumor grade 3 or 4. In contrast, of 70 patients positive for COX-2 expression, 11 (15.7%) and 16 patients (22.9%) had lymph node metastasis and tumor grade 3 or 4, respectively.

Fig. 2 shows representative tissue sections stained immunohistochemically for Ki-67 (Fig. 2,A), CD31 (Fig. 2,B), MMP-2 expression (Fig. 2,C), and TUNEL (Fig. 2,D). The immunohistochemical findings including Ki-67 LI, AI, and MVD are shown in Table 3. The Ki-67 LI and MVD in tumor tissues that were positive for COX-2 expression were higher than in those negative for COX-2 expression (P < 0.01) In contrast, AI did not correlate with COX-2 expression in renal cell carcinomas (P = 0.054). There was a significant correlation between COX-2 expression and MMP-2 expression (Table 4; P < 0.01).

To assess the risk factors, we used univariate logistic regression analysis for tumor size (>7 cm in diameter; Table 5, model A) and presence of invasion or metastasis (Table 5, model B). All of the factors, with the exception of AI, were identified as significant predictors by univariate analysis in both models A and B. Furthermore, among these six factors, nuclear grade 3/4, COX-2 expression, and MVD were identified as independent risk factors for large tumor size in multivariate logistic regression model (Table 5, model A). The COX-2 expression-positive group was almost four times higher compared with COX-2 expression-negative group. In addition, when we performed similar analysis for the presence of invasion or metastasis, MMP-2 expression, high KI-67 LI, and MVD were identified as independent risk factors in the multivariate logistic regression model, whereas positive COX-2 expression was not an independent factor (Table 5, model B).

The results of the log rank test of clinical features, and immunohistochemical, and TUNEL findings are shown in Table 6. All of the markers except AI correlated with cause-specific survival. The 5-year survival rates in patients that were either negative or positive for COX-2 expression were 91.1% and 65.9%, respectively. However, Cox proportional hazard analysis demonstrated that T and M classification were independent and significant influencing factors for cause-specific survival (OR, 5.19; 95% CI, 1.02–26.54; P = 0.048 and OR, 9.41; 95% CI, 2.16–41.11; P < 0.01, respectively; Table 7). In contrast, COX-2 expression-positive was not an independent factor for cause-specific survival.

Several parameters have been reported to be useful markers for staging and predicting the prognosis of patients with renal cell carcinoma. Ki-67 LI is one of the useful markers used for predicting the stage and prognosis. Jochum et al. (33) reported that the MIB-1 index in 87 patients with renal cell carcinoma ranged from 0.6 to 30.4%, with a median value of 4.7%, and this marker had provided significant prognostic information. Our results showed a similar trend with a median of 4.7% and a range of 0.7–42%. In addition, our results also showed that Ki-67 LI was a significant and useful marker for predicting the tumor stage and prognosis in univariate analysis. The Ki-67 LI has been used to evaluate cell proliferation activity. Our results demonstrated that renal cell carcinomas tissue expressing COX-2 has a high proliferation activity. Furthermore, in multivariate analysis, we found that positive COX-2 expression was an independent factor for large tumor size, whereas Ki-67 LI was not an independent factor. We speculate that COX-2 expression is associated with tumor growth in renal cell carcinoma.

Using in vitro assays, several investigators have reported that COX-2 can influence angiogenesis and that COX-2 inhibitor can reduce angiogenesis (34, 35). Williams et al. (36) reported that the vascular density in tumors grafted in COX-2^−/− mice was 30% lower than in wild-type mice, and that MVD correlated positively with the extent of COX-2 immunostaining (r = 0.41; P = 0.02). Our results also demonstrated that COX-2 expression played an important role in regulating neovascularity. Neovascularization is necessary for tumor growth and metastasis of malignant diseases (37, 38). Numerous reports have demonstrated that high MVD correlated with tumor growth and metastasis in several malignant tumors (39, 40, 41). However, there has been discussion of the pros and cons of the prognostic value of MVD for disease stage and survival in renal cell carcinoma (42, 43, 44). MVD count is influenced by several factors, such as the size of necrotic area within the tumor and the method used for measurement of MVD. Our results demonstrated that MVD was an independent predictor of large tumor size and metastasis. These results add support to the importance of neovascularity in tumor growth and that COX-2 expression is an important regulator of neovascularity in renal cell carcinoma.

In the present study, expression of COX-2 correlated with high T, N, and M stage, and high tumor grade. Several in vivo studies using a variety of human carcinoma tissues reported that COX-2 expression was associated with tumor invasion and lymph node metastasis (30, 45, 46, 47). However, there is evidence for and against the notion that COX-2 expression is associated with distant metastasis. Several investigators have reported that COX-2 expression is not associated with distant metastasis (30, 45, 46). In our study using the χ² test and univariate logistic regression analysis, we showed that positive COX-2 expression correlated significantly with metastasis. However, multivariate logistic regression analysis showed that COX-2 expression was not an independent factor of metastasis. On the other hand, multivariate analysis also showed that high Ki-67 LI, MVD, and positive MMP-2 expression were independent factors for tumor invasive or metastasis. Thus, it appears that COX-2 expression does not directly influence metastasis. Tsujii et al. (48) reported that constitutive expression of COX-2 was associated with phenotype changes including the activation of MMP-2. In the present study, COX-2 expression correlated positively with MMP-2 expression. MMP-2 is a member of a large family of endogenous proteases that degrade various components of the extracellular matrix. The proteolytic breakdown of extracellular matrix is thought to be a critical step for the tissue invasion of cancer (49, 50). Several studies reported that MMP-2 is associated with tumor aggressiveness including metastatic potential in various malignancies, such as renal cell carcinoma (26, 51). There are few reports that have examined the relationship between MMP-2 expression and metastasis in human renal cell carcinoma tissue. In addition to MMP-2, MMP-9 and tissue inhibitor of metalloproteinases were found recently to play important roles in tumor progression and metastasis (32, 52). We speculate that COX-2 expression may be associated with metastasis via regulation of MMP-2 expression. However, the MMP system, composed of both proteinases and their cognate inhibitors, has an independent activity for regulating metastatic potential that is distinct from COX-2.

A large body of evidence suggest that apoptosis correlates with the rate of tumor growth. Enhanced apoptosis has been reported to be associated with slow tumor growth in colorectal carcinoma (53). Conversely, suppression of apoptosis has been reported to correlate with tumor progression (54, 55). Several reports indicated that COX-2-specific nonsteroidal anti-inflammatory drugs can induce apoptosis in a variety of cancer cells (56, 57). However, there are no studies on the relationship between COX-2 expression and apoptosis in human renal cell carcinoma tissue. Our results showed that COX-2 expression was not associated with a lower AI. We speculate that COX-2 expression has little or no effect on the regulation of apoptosis in human renal cell carcinoma.

When the relative contribution of various risk factors for cause-specific survival was assessed, we found that T and M stages were independent predictive factors. COX-2 expression correlated closely with tumor cell proliferation, angiogenesis, and MMP-2 expression. However, multivariate analysis indicated that COX-2 expression was not an independent and significant risk factor. This finding is in agreement with those of several reports that have demonstrated that disease stage is the most significant determinant of patient survival (58).

In conclusion, we demonstrated in the present study that COX-2 expression was associated with tumor status including tumor size, metastasis, and tumor grade; and correlated with increased cell proliferation, angiogenesis, and positive MMP-2 expression. Multivariate analysis identified COX-2 expression as an independent predictive factor for large tumor size, but not for invasion and metastasis. We speculate that COX-2 expression plays an important role in the biological process of human renal cell carcinoma.

We thank Etsuji Taguchi, Takumi Shimogama, and Miki Yoshimoto for excellent assistance.