Cyclooxygenase-2 Expression: A Significant Prognostic Indicator for Patients With Colorectal Cancer

Colorectal cancer is the third most common cancer in the world and the second most common cause of cancer-related death (1). During 2000, in the United States, 130,200 new cases of colon cancer and rectal cancer were reported (2). In Japan, ∼36,000 patients dead of this cancer in 2001 (3). Despite numerous attempts to detect cancer at an early stage, the overall long-term outcome of patients curatively resected has not significantly changed in the last decade, the 5-year survival rate being ∼60% (4). Assessment of molecular prognostic factors associated with a distinct prognostic outcome would, therefore, be of great help for identification of patients who are likely to benefit from adjuvant therapies, leading to an improvement in prognosis.

Recent studies have suggested that an increase in the expression of cyclooxygenase (Cox)-2, a key-inducible enzyme involved in the production of prostaglandins and other eicosanoids, may play a significant role in carcinogenesis in addition to its well-known role in inflammatory reactions (5, 6, 7, 8, 9, 10).

Interestingly, Cox-2 has been reported to be significantly increased in up to 85% of human sporadic colorectal carcinoma, predominantly within neoplastic epithelial cells (5, 6, 11, 12, 13, 14, 15, 16, 17, 18), in which Cox-2 may induce resistance to apoptosis, alter extracellular matrix adhesion, modulate tumor angiogenesis, and increase metastatic potential (7, 9, 19). In addition, a specific cause-effect relationship between the overexpression of Cox-2 and colon cancer has been determined in studies on intestinal carcinogenesis in the adenomatous polyposis coli (APC)^Δ716 (8, 20).

Moreover, Kawamori et al. (21) have reported that celecoxib, a specific Cox-2 inhibitor suppressed both incidence and multiplicity of azoxymethane-induced colon cancer in rats by ∼93 and 97%, respectively. In a previous clinical trial, we also showed that rofecoxib, another specific Cox-2 inhibitor, significantly decreased the number and size of rectal polyps in familial adenomatous polyposis patients (22).

However, it is controversial whether Cox-2 expression in itself is a prognostic factor for local recurrence and/or survival of patients with colorectal cancer (23).

Therefore, to clarify additionally the prognostic role of Cox-2 in colorectal carcinoma and to evaluate the possible contribution of isoenzyme Cox-1, we immunohistochemically determined the expression of both proteins in a series of patients with primary tumors, who were treated at our institution. In addition, sections of adjacent normal mucosa, lymph node metastases, and hepatic metastases samples were also evaluated.

A total of 288 consecutive patients who had undergone surgical resections for primary sporadic colorectal carcinoma at the Department of Surgical Oncology, Tokyo Medical and Dental University (Tokyo, Japan), between January 1986 and December 1996 was included in this study. There were 174 colon cancers and 114 rectal cancers; 183 males and 105 females; mean age, 61 ± 10 years. None of them had a history of hereditary colon cancer syndromes and regular use of aspirin-like drugs. There was no preoperative chemotherapy or radiotherapy; however, after surgery, for both colon and rectal cancer, patients with stage III tumor received oral 5-fluorouracil, and those with stage IV tumor had 5-fluorouracil–based systemic chemotherapy without any radiation. Patients were followed up until death or December 31, 2002, with a median postoperative follow-up duration of 63 ± 33 months.

In all cases, archival H&E slides of the primary and secondary tumors were retrieved and reviewed to confirm pathological features and to select suitable tissue blocks for immunohistochemical analysis.

The samples included 288 primary tumors, 25 available lymph node metastases, 16 available hepatic metastases. In all 25 metastatic lymph nodes and 4 of 16 hepatic metastases, concurrent primary tumors were examined. Additionally, for comparison, sections of 10 adjacent normal mucosa, 10 normal lymph node, 10 normal hepatic parenchyma, were also included in this study.

The clinicopathologic findings were determined according to the International Union Against Cancer Tumor-Node-Metastasis (TNM) classification of malignant tumors (24).

A universal immunoenzyme polymer method was used for immunostaining. Three-μm thick sections were cut from formalin-fixed, paraffin-embedded tissue blocks and mounted on poly-lysine coated slides, dewaxed in xylene, and rehydrated through a graded series of ethanol. After deparaffinization, antigen retrieval treatment was performed at 121°C (autoclave) for 5 minutes in 10 mmol/L sodium citrate buffer (pH 6.0), then treated with 3% hydrogen peroxide in methanol solution for 20 minutes to quench endogenous peroxidase activity. Nonspecific bindings were blocked by treating slides with 10% normal goat serum for 10 minutes. To block intrinsic biotin-binding capabilities, hepatic tissue slides were treated with avidin-biotin blocking kit reagents (Vector Laboratories, Inc., Burlingame, CA) for 15 minutes. Thereafter, the slides were incubated with mouse monoclonal antibodies against human Cox-2 (dilution 1:250; Cayman Chemical Co., Ann Arbor, MI) for 2 hours at room temperature for Cox-2 and rabbit polyclonal antibodies against human Cox-1 (dilution 1:200; Cayman Chemical Co.) for 2 hours at room temperature for Cox-1. Next, slides were incubated with labeled polymer (N-Histofine Simple Stain MAX PO MULTI; Nichirei Co., Tokyo, Japan) for 30 minutes at room temperature. Color development was done with 0.02% 3,3′-diaminobenzidine tetrahydrochloride (Sigma, Poole, United Kingdom) and 0.06% hydrogen peroxide in 50 mmol/L Tris-HCl (pH 7.6) for 5 minutes. Finally, the slides were counterstained with 1% Meyer’s hematoxylin.

As a negative control for Cox-2 and Cox-1 staining, tissue sections were treated with normal serum instead of each primary antibody.

All sections were scored blind by two investigators (L.T.S. and Y.T.) under a light microscope; and, in cases of occasional scoring discrepancy, consensus was always achieved after discussion of findings. For Cox-2 and Cox-1 assessment, the entire tissue section was scanned to assign the scores. The staining intensity was scored as 0 (negative), 1 (weak), 2 (medium), and 3 (strong). Extent of staining was scored as 0 (0%), 1 (1 to 25%), 2 (26 to 50%), 3 (51 to 75%), and 4 (76 to 100%), according to the percentages of the positive staining areas in relation to the whole carcinoma area or entire section for the normal samples. The sum of the intensity and extent score was used as the final staining score (0 to 7) for Cox-2 or Cox-1. This relatively simple, reproducible scoring method that gives highly concordant results between independent evaluators has been used in past studies (14, 17).

For the purpose of statistical evaluation, tumors having a final staining score of ≥3 were considered to be positive.

All statistical analyses were carried out with the StatView Software (version 5.0). The correlations between expression of Cox-2 or Cox-1 and several clinicopathologic parameters were assessed with the χ² test or the Spearman rank test as indicated. The Kaplan-Meier method was used to estimate survival as a function of time, and survival differences were analyzed with the log-rank test. The Cox proportional hazards model was used for multivariate analysis of prognostic factors. P < 0.05 was considered to be significant.

Both cyclooxygenase proteins were mainly detected in the cytoplasm of the epithelial cells, vascular endothelial cells, and some stromal cells probably fibroblast and mononuclear cells. The allocation of staining score by the two investigators was concordant in >94% of Cox-2 samples and >96% of Cox-1 samples. In addition, as we stated in Materials and Methods, the discrepant cases were reevaluated and scored by consensus. The distribution of primary tumors according to the extent and intensity of Cox-2 and Cox-1 immunoreactivity are presented in Tables 1 and 2.

The expression of Cox-2 was positive (grade 3 to 7) in 70.8% of primary tumors, 92.0% of metastatic lymph nodes, and all of hepatic metastases (Table 3). The normal mucosa adjacent to the Cox-2–positive tumor and the normal samples showed a little or no stained for Cox-2. In contrast, the positivity rate for Cox-1 was 42.7, 84.0, 37.5, and, 80.0% in the primary tumors, metastatic lymph nodes, hepatic metastases, and normal samples, respectively (Table 3). Moreover, we found an elevated expression of both Cox-1 and Cox-2 in the normal liver tissue and biliary epithelium adjacent to the metastatic lesion, even after endogenous avidin-biotin blocking treatment. Representatives of the staining patterns are shown in Fig. 1.

When compared primary and secondary tumors versus normal samples, a significant difference was observed for Cox-2 expression: primary tumor versus normal mucosa (P < 0.0001), metastatic lymph node versus normal lymph node (P = 0.0001), and hepatic metastases versus normal hepatic parenchyma (P = 0.0016); and for Cox-1 expression, primary tumor versus normal mucosa (P = 0.0195), metastatic lymph node versus normal lymph node (P = 0.6468), and hepatic metastases versus normal hepatic parenchyma (P = 0.1069).

In Table 4, the association between Cox-2 and Cox-1 expression and the clinicopathologic parameters is shown. Cox-2 expression was significantly correlated with primary tumor size (Rho = 0.411, P < 0.0001), depth of invasion (P = 0.0004), lymph node metastasis (P = 0.0312), lymphatic invasion (P = 0.0253), venous invasion (P = 0.0186), TNM stage (P = 0.0013), and tumor recurrence (P = 0.0097). There was no significant correlation between Cox-2 expression and age, gender, tumor site, and histologic type. Cox-1 was significantly correlated with only the primary tumor size (Rho = 0.208, P = 0.0004).

Within a median postoperative follow-up duration of 63 ± 33 months, 57 cancer-related deaths occurred: 10 in patients with Cox-2–negative tumors and 47 in the positive group (P = 0.0311) and 30 in patients with Cox-1–negative tumors and 27 in the positive group (P = 0.4271).

In the entire cohort, the overall survival rates of patients with Cox-2–negative tumors were significantly higher when compared with those of the Cox-2–positive group (P = 0.0006, log-rank test; Fig. 2,A). The same significant difference in survival between the two groups of patients was also observed in combined stage I and stage II (P = 0.0271; Fig. 2,B) and in stage III cases (P = 0.0081; Fig. 2 C). The 5-year survival rate of Cox-2–negative and Cox-2–positive cases in stage I and stage II was 97 and 82%, respectively, and that in stage III was 88 and 67%, respectively. Stage IV was not considered alone because of the limit number of cases.

The Kaplan-Meier curves demonstrated no association between Cox-1 expression status and patients survival (P = 0.4016; log-rank test).

Furthermore, in the multivariate analysis, potential prognosis factors such as age and gender of the patient, tumor site, size, depth of invasion, lymphatic invasion, venous invasion, degree of differentiation, lymph node metastasis, TNM stage, and Cox-1 and Cox-2 expression were included in the Cox proportional hazards model. The results indicated that, in addition to lymph node metastasis and TNM stage, Cox-2 expression status was an independent prognostic factor in our cohort (P = 0.0103; Table 5).

In this study of human colorectal carcinoma, the expression of Cox-2 protein showed a markedly increased from normal samples to primary tumors and metastatic samples. We found that elevated expression of Cox-2 was significantly associated with prognostically worse clinicopathologic variables, including tumor size, depth of invasion, lymph node metastases, and vessels invasion. These findings are consistent with our previous reports that tumors with higher Cox-2 gene expression grow larger and in a more invasive manner in colorectal (25) and gastric carcinoma (26). Contrary to the results from some recent studies (14, 17), we could not prove a statistically significant relationship between Cox-2 expression with the degree of tumor cells differentiation.

Interestingly, we showed a significant correlation between high-grade Cox-2 expression with a shortened survival of patients and Cox-2 status as well as lymph node metastasis and TNM stage was an independent prognostic factor.

Actually, the negative prognostic value of Cox-2 overexpression has been reported in several studies on various carcinomas such as those of colon and rectum (12, 14, 17, 27, 28), stomach (29), esophagus (30), breast (31), lung (32, 33), and ovaries (34).

However, there is a disagreement over the independent prognostic capability of Cox-2 as compared with other known prognostic factors, including tumor features and molecular characteristics (14, 27). Also, some authors that used immunohistochemical analysis method claimed that Cox-2 may be involved in tumorigenesis but lack a prognostic significance for colorectal cancer patients (18, 23, 35, 36, 37). At this point, it is considered that tumor invasiveness, frequent metastasis, and expression in larger tumors are responsible for the worse prognosis for patients bearing Cox-2–positive tumors (14). We believe that variability in population size and characteristics, antibodies, lack of consensus on and standardization of methods, and differences in the choice of cutoff levels may explain the discrepancies between the different studies. In fact, our findings were obtained by analysis of a relatively large number of consecutively resected cases in a single institution and, thus, should be of considerable interest. Moreover, the monoclonal antibodies we used have been shown in several excellent studies (13, 14, 30, 31, 38) and is considered to give the most sensitive, specific, and reproducible immunoreactivity of Cox-2 protein. Nevertheless, as proposed by Buecher et al. (28), it will be of critical importance to standardize the methods used for the evaluation of tumor Cox-2 status and to apply common interpretative criteria to allow direct comparison of results between laboratories.

To date, several molecular pathways have been postulated to elucidated the mechanism by which up-regulation of the Cox-2 gene can contribute to tumor development and progression. Among them, the hypothesis of an inhibition of apoptosis (7, 39), an increased in cell proliferation and angiogenesis (14, 15, 19, 20, 39, 40), and a local immunosuppression (16) as carcinogenic properties of Cox-2 have in the recent years received more attention. These possible pathways may all influence the prognosis of patients with colorectal cancer to worse.

One distinctive feature in the present study is that cancer cells in 23 of 25 lymph node metastases and all hepatic metastases, as well as their corresponding primary tumors, were positive for Cox-2 protein. Hull et al. (37), who investigated 54 colorectal cancer liver metastases, reported that Cox-2 protein was detected in cancer cells in 100% of cases. Unfortunately, the authors were unable to obtain data regarding Cox-2 status in the corresponding primary tumors. To our knowledge, this is the first investigation in which Cox-2 protein expression is evaluated in both primary and secondary tumors. This finding may indicate that Cox-2 expression by cancer cells probably occurred before metastasis, and only Cox-2 bearing cancer cells have more of a chance to metastasize to the lymph node and the liver. Additional studies are warranted to clarify this question and, more interestingly, to assess the utility of Cox-2 inhibitors in the treatment of metastatic tumors.

In contrast to Cox-2, the possible roles of Cox-1 expression in human tumorigenesis have received less attention. Our attempt to show the clinicopathologic significance of Cox-1 expression in colorectal cancer, as has been shown in ovarian cancer (41) and testicular tumors (42), showed only an association with tumor size.

However, in conjunction with a genetic study reporting that disruption of Cox-1 gene in APC^Min mice causes suppression of polyposis (43) and also given the fact that some nonsteroidal anti-inflammatory drugs such as aspirin that inhibit only Cox-1 can reduce colon cancer incidence and mortality (44, 45, 46), it is likely that both cyclooxygenases play critical roles in tumorigenesis at different stages. In support to this possibility, Takeda et al. (47), in studying intestinal polyp formation, suggested that, whereas Cox-2 induction is essential for polyps to expand beyond ∼1 mm, nascent polyps are likely to develop to ∼1 mm with prostaglandin E₂ supplied by Cox-1 alone. Additional studies are necessary to gain more insight into the functional role of Cox-1 in the development and progression of colorectal carcinoma.

Finally, evidence from several experimental and clinical studies has shown that selective Cox-2 inhibitors possess significant antitumor activity on their own with favorable safety profile (1, 8, 21, 22, 48) and can also enhance tumor response to radiation or cytotoxic agents (49, 50). However, recent clinical studies (51, 52) with rofecoxib in the treatment of metastatic colorectal cancer pointed out an increased toxicity and lack of efficacy. Nevertheless, although this is an interesting observation, the few patients in those studies precludes it from being presented as a definitive conclusion. More clinical trials with more patients are needed to really assess the therapeutic benefit of Cox-2 inhibitors in human colorectal cancer.

We conclude that elevated Cox-2 expression, but not that of its isoform Cox-1, is significantly associated with reduced survival in patients with colorectal adenocarcinoma. Accordingly, our results taken together with those of other prompted us to believe that assessment of Cox-2 level in colorectal cancer may be used as a marker to select not only high-risk patients group for recurrence and offer them more frequent follow-up but also patients who may likely benefit from Cox-2 inhibitors as adjuvant therapy either alone or in combination with established anti cancer drugs.