Coexpression of Inducible Nitric Oxide Synthase and COX-2 in Hepatocellular Carcinoma and Surrounding Liver: Possible Involvement of COX-2 in the Angiogenesis of Hepatitis C Virus-positive Cases

NO, a small potent lipophilic gas with divergent biological activities, seems to play important roles in modulating tissue injury and carcinogenesis (1). Three distinct forms of NOS² catalyze the formation of NO. Endothelial NOS and neuronal NOS are constitutively expressed in different tissues, whereas iNOS is related to a high-output pathway and is responsible for various pathological processes (2). Although increased iNOS expression has been demonstrated in colon (3), prostate (4), and bladder cancer (5), its exact function in the tumor biology is not yet clear.

COX, constitutively expressed COX-1, and inducible COX-2, are the rate-limiting enzymes for production of prostanoids (prostaglandins and thromboxanes) from arachidonic acid (6). COX-2 has recently been categorized as an immediate-early gene associated with inflammation (7), cellular growth (8), differentiation (9), prevention of apoptosis (9), and tumorigenesis (3). It has also been reported that COX-2 induces angiogenesis, which is essential for tumor growth (10). Increased prostaglandin production and enhanced release of angiogenic growth factor by COX-2 may induce neovascularization (3, 8).

Evidence suggests that NO produced by iNOS enhances the activity of COX-2 (11). Expression of both iNOS and COX-2 has been reported to be increased in Barrett’s esophagus and esophageal adenocarcinomas and also in Helicobacter pylori-induced gastritis (12, 13). NO and COX-2 have carcinogenic effects achieved either directly or by producing mediators that regulate cellular growth (10, 14). These effects may be mediated by the formation of peroxinitrite and prostanoids and involve “cross-talk” between the two enzyme systems (15). To determine the roles of iNOS and COX-2 in HCC, we compared their expressions by the immunohistochemical method.

One hundred patients (77 males, 23 females) with HCC undergoing curative hepatectomy (segmental or lobar resection) between 1989 and 1997 were included in this study. The tissue samples were fixed in 10% neutral formalin and subjected to histopathological examination. None of the patients received pre- or postoperative adjuvant chemo- or radiotherapy. The age of the patients ranged from 36 to 78 years (62.8 ± 8.7 years, mean ± SD). Tumor size varied from 1.2 to 18.5 cm in diameter. Of the 100 patients, 55 were positive for antibody to HCV (anti-HCV). Preoperative ultrasonography, computed tomography scan, and angiography were performed routinely for all patients. Our criteria for definition of curability were described previously (16). All patients received follow-up at regular time intervals. Recurrence after surgery was diagnosed by serum AFP level, ultrasonography, computed tomography scan, and angiography. The criteria for histological study were described previously (17).

Tumor samples free of necrosis or hemorrhage on macroscopic inspection were selected for histology. Samples were routinely fixed in 10% formalin and subsequently embedded in paraffin. Serial sections 4-μm thick were prepared from paraffin blocks. Sections were deparaffinized and hydrated by sequential immersion in xylene and graded alcohol solutions. The sections were then incubated in 3% H₂O₂ for 30 min to block the endogenous peroxidase activity. For iNOS staining, slides were incubated with 2% skim milk for 10 min. Slides were treated with normal serum obtained from the same species in which the secondary antibody was developed for 30 min to block nonspecific staining. Subsequently, slides were incubated with primary antibodies [anti-iNOS (WAKO Pharmaceuticals Ltd., Osaka, Japan) at 1:500 dilution for 60 min at room temperature; anti-COX-2 (Cayman Chemical, Ann Arbor, MI) at 1:300 dilution overnight at 4°C; and anti-CD34 (Dako, Glostrup, Denmark) at 1:50 dilution for 1 h at room temperature], as appropriate. Immunostaining was performed according to the avidin-biotin peroxidase complex method by a commercially available kit [Histofine, SAB-PO(R); Nichirei Corporation, Tokyo, Japan]. Slides were treated with a biotin-conjugated secondary antibody for 30 min followed by incubation with peroxidase-conjugated streptavidin for 30 min at room temperature. All steps were followed by washing in PBS. Peroxidase activity was detected by incubating the samples with 3-amino-9-ethylcarbazol (Histofine, Tokyo, Japan). Finally, the sections were counterstained with hematoxylin. For CD34 staining, counter staining was not done.

The degree of staining was categorized by the extent and intensity of the staining. Two independent observers screened all sections as a semiquantitative evaluation of iNOS and COX-2 immunostaining. The immunoreactive score was determined by the sum of extension and intensity as reported previously (18) The intensity of staining was scored on a scale of 0 to 3, in which 0 = negative staining, 1 = weakly positive staining, 2 = moderately positive staining, and 3 = strongly positive staining. The extent of positivity (“extent of distribution” of positive cells) was estimated on a scale of 0 to 4, in which 0 = negative, 1 = positive staining in 1–25% of cells, 2 = positive staining in 26–50%; 3 = positive staining in 51–75%; and 4 = positive staining in 76–100%. The combined staining score (extension + intensity) ≥3 was considered as positive staining.

Because of the technical difficulty in counting linearly immunostained sinusoids, we used an image analysis system to assess the MVD as a percentage of the endothelial area. Because the immunoreactivity of CD34 showed slight heterogeneity within the same tumor, the five most highly vascularized areas (hot spots) were selected in ×200 magnification fields. The images were scanned with a Olympus CCD camera and stored on a computer for subsequent analysis. On the computer screen image, the background color was subtracted and the 3,3′-diaminobenzidine-stained area was evaluated using the NIH (Bethesda, MD) image analysis system (19). A MVD value ≥3 (median value of the MVD) was considered high MVD.

Correlation between factors was evaluated using the Spearman rank correlation coefficient test. Survival rates were obtained by the Kaplan-Meier method, and the differences were compared by the log-rank and Wilcoxon tests. Significance differences between categorical variables were compared by the χ² test or Fisher’s exact probability test when appropriate. Continuous variables were compared by the Mann-Whitney U test. Multivariate analyses was performed by the Cox proportional hazards regression analysis using StatView 4.11 (Abacus Concepts, Inc., Berkeley, CA) software on a Macintosh computer. P < 0.05 was considered significant.

iNOS immunoreactivity was observed mainly in the hepatocytes, showing a predominant cytoplasmic staining, with the positive liver cells distributed in both the tumor tissue and surrounding liver (Fig. 1, A and B). iNOS expression was not observed in liver-infiltrating mononuclear cells, vascular endothelium, or sinusoidal lining cells. iNOS immunoreactive score was significantly higher in the surrounding liver (P = 0.011) in HCV-positive HCC cases (Table 1). We found no correlation between tumor iNOS expression and clinicopathological factors in all cases (n = 100; data not shown).

In immunohistochemical analysis, cytoplasmic staining for COX-2 was observed in both tumor cells and nontumorous connective tissue cells, including those of sinusoidal endothelial cells (Fig. 1, C and D). We found no correlation between tumor COX-2 expression and clinicopathological factors when evaluated in all HCC cases (data not shown). In HCV-positive HCC cases, the COX-2 immunoreactive score was significantly higher in tumors ≥4 cm in diameter (P = 0.042) and also in tumors with positive iNOS staining scores (P = 0.034; Table 2). With the Spearman rank correlation coefficient test, there was a significant (P = 0.046) positive correlation between iNOS and COX-2 expression.

The staining of sinusoid-like vessels inside and around the tumors was evident by anti-CD34 antibody (Fig. 1,E). In the surrounding liver tissue, anti-CD34 staining was confined to vessels of the portal triad with weak staining in a few of sinusoids at the periportal area. We found no relationship between tumor CD34 expression and clinicopathological factors except for capsule formation. Significantly higher numbers of tumors with capsule formation had a higher MVD than did tumors without any capsule (data not shown). In HCV-positive cases, a significant positive correlation was found between MVD and COX-2 expression (P = 0.010; Fig. 2).

Neither iNOS nor COX-2 expression had any impact on patient survival when evaluated in all cases (data not shown). However, in HCV-positive HCC cases, patients with combined negative expression of iNOS and COX-2 had significant improvement in survival when compared with rest of the patients with variable expressions of iNOS and/or COX-2 (P = 0.041 and 0.018, log-rank test for overall and recurrence-free survival rates, respectively; Fig. 3). A total of five factors in the univariate analysis had significant effects on overall survival: age (<60 versus ≥60 years), combined iNOS and COX-2 expression (positive group versus negative group), liver cirrhosis, serum α-fetoprotein (<100 ng/ml versus ≥100 ng/ml), and differentiation (well, moderate versus others). Four factors also had prognostic impact on recurrence-free survival: age (<60 versus ≥60 years), combined iNOS and COX-2 expression (positive group versus negative group), liver cirrhosis, and α-fetoprotein (<100 ng/ml versus ≥100 ng/ml). Combined iNOS and COX-2 expression in HCV-positive HCC cases remained an independent prognostic factor both in the recurrence-free and overall survival analyses (Table 3).

Angiogenesis is a prerequisite for development and growth of different human tumors. For HCC, there are only few studies describing the mechanisms of microvessel formation. Although there are some studies addressing the role of COX-2 in angiogenesis (20), there is no report describing the relationship of angiogenesis and COX-2 expression in HCC. In both the experimental setting and clinical studies, it has been shown that COX-2 induces angiogenesis, which in turn aids tumor growth, invasion, and metastasis (10). Uefuji et al. (21) reported that COX-2 overexpression in gastric cancer samples was associated with enhanced angiogenesis. Recently, we reported that COX-2 expression correlated with tumor neovascularization in human colorectal carcinoma (22). Very recently, it was reported that iNOS and COX-2 were positively correlated with vascular endothelial growth factor and that expression of both of these factors induced angiogenesis in non-small cell lung cancers (23). In this study, we found that COX-2 overexpression had a strong correlation with MVD and tumor size in HCV-positive HCC cases. Recent evidence indicates that COX-2 modulates angiogenesis either by augmenting the release of angiogenic peptides (vascular endothelial growth factor, basic fibroblast growth factor, and NO) by tumor cells or by directly increasing the production of prostaglandins (10, 24). In this study, COX-2 expression was higher in tumors ≥4 cm in diameter in HCV-positive HCCs. COX-2 expression is also significantly higher in larger tumors in colorectal cancer (25). Prostanoid levels in adenomas from patients with familial adenomatous polyposis are significantly elevated with increased tumor size (26). Perhaps a similar mechanism is responsible for higher COX-2 expression in larger HCCs.

We have found positive correlations between COX-2 and iNOS expression and MVD in HCV-positive HCCs. This suggests that both iNOS and COX-2 might be important factors in the pathogenesis of HCV-positive HCCs. Higher expressions of the iNOS and COX-2 genes in HCV-positive HCCs might be caused by a secondary effect of cytokines produced in response to HCV infection (27, 28) or by the direct activation of the HCV core protein (29). In particular, both genes have nuclear factor κB binding sites in their promoter regions (29), and HCV is known to stimulate the activation of nuclear factor κβ (30). Evidence suggests that iNOS is capable of inducing COX-2 expression, and there is a possibility of cross-talk between the iNOS and COX-2 systems, producing a synergistic effects in tumorigenesis (15).

It has been reported that HCV infection causes elevated iNOS transcription (31). Kane et al. (32) showed a significant up-regulation of iNOS in HCV-positive hepatitis and suggested that enhanced iNOS expression might be responsible for carcinogenesis in HCV-positive chronic hepatitis. We also found higher positive iNOS expression in HCV-positive HCC cases. In our study, iNOS expression was significantly higher in the surrounding liver in cirrhotic patients. In addition, it has been reported that iNOS expression becomes higher in cirrhotic liver in animal models (33). It has been suggested that in cirrhosis, the endotoxemia and/or the increased circulating levels of cytokines may induce iNOS expression. Aberrant iNOS expression may be one of the phenotypical changes associated with the carcinogenic process in the cirrhotic liver. It might also be possible that the increased iNOS expression in the surrounding cirrhotic liver was induced by some factors released by the tumor itself. Another interesting finding of this study was a negative correlation, although not statistically significant, between iNOS expression and the serum AFP level. It was reported that the transgenic mice that expressed and produced human AFP had significantly reduced production of IFN-γ in their livers and sera. (34). IFN-γ induces iNOS; it therefore might be possible that higher AFP-secreting tumors down-regulate iNOS production through IFN-γ. In addition, reduced plasma nitrite/nitrate levels were reported in AFP-positive patients compared with AFP-negative patients (35).

The biological significance of high iNOS and COX-2 expression remains controversial and poorly understood in cancers. Most of the studies indicate that high concentrations of NO and prostanoids can be either mutagenic or tumorigenic and that these effects depend on their concentrations (36, 37). iNOS and COX-2 expression is up-regulated in different types of premalignant and malignant lesions (3, 4, 5, 8). Both iNOS and COX-2 are abundantly expressed in premalignant Barrett’s esophagus (12) and in H. pylori-related gastritis (13). We could not find any impact of iNOS and COX-2 expression on disease-free survival in all patients, but in HCV-positive cases, patients with negative expression of both iNOS and COX-2 had a significant survival advantage over the rest of the patients with variable expression of iNOS and/or COX-2. Recently, Kondo et al. (38) reported that overexpression of COX-2 in the surrounding liver was associated with a shorter disease-free survival in patients with HCV-positive HCCs.

HCC is the leading cancer in men in Taiwan and is one of the most common causes of malignancy-related death in Africa and Asia (39). Unfortunately, to date, there is no effective preventive measure for this highly malignant disease. Many reports have indicated that COX-2 is up-regulated in most human tumors (20). The clinical data of this study indicate that COX-2 expression may play a role in tumor angiogenesis because a direct correlation between the COX-2 expression score and MVD was observed in HCV-positive HCC cases. Recently, we have shown that COX-2 inhibitors induce growth arrest and apoptosis in several HCC-derived cell lines (40). All of these findings indicate that COX-2 inhibitors might be effective in prevention of both cancer development and disease progression of HCC. Further research in this direction may elucidate the roles of iNOS and COX-2 in the development of HCC, which may form the basis for future novel therapeutic and/or preventive strategies to inhibit these genes.