Increased Cyclooxygenase-2 Expression in Duodenal Compared with Colonic Tissues in Familial Adenomatous Polyposis and Relationship to the −765G → C COX-2 Polymorphism

Familial adenomatous polyposis (FAP) is caused by germ-line mutations in the adenomatous polyposis coli (APC) gene, which leads to the development of innumerable adenomatous polyps throughout the colorectum. Without colectomy, colorectal carcinoma is almost inevitable usually by the fifth decade of life. In recent decades, colorectal cancer screening and prophylactic surgery have significantly improved the survival of patients with FAP. However, the life expectancy of patients with FAP is still below that of the general population, largely due to the risk of developing upper gastrointestinal tract malignancy.

The duodenum is the second commonest site of adenoma development in patients with FAP, and ∼5% of patients with FAP will develop duodenal cancer during their lifetime (1–3). Currently, the main management options for patients with duodenal adenomatosis are endoscopic surveillance and selective surgical resection. Duodenal surgery, either a pancreas-preserving duodenectomy or pancreaticoduodenectomy, is indicated for patients with either severe duodenal polyposis or duodenal carcinoma. However, these therapeutic options do not adequately manage the duodenal neoplasia that arises in the setting of FAP. Colorectal adenomas occurring in patients with FAP have been shown to regress with sulindac, a nonsteroidal anti-inflammatory drug, and with cyclooxygenase 2 (COX-2) inhibitors (4, 5). These drugs have been targeted towards treating duodenal adenomas but results have been modest and conflicting. Some investigators have shown that sulindac and COX-2 inhibitors can reduce small duodenal adenomas (6, 7), but not large adenomas (8), whereas others have not (9, 10).

The rationale behind using nonsteroidal anti-inflammatory drugs as chemopreventive agents is their inhibitory effect on the COX enzymes. Two isoforms of COX exist, COX-1 and COX-2, which are involved in the conversion of arachidonic acid into prostaglandins. Prostaglandins, in turn, regulate cellular functions such as angiogenesis and cell proliferation and have been associated with progression of tumor development in a size-dependent manner (11). In the colorectum, COX-2 is increasingly up-regulated with progression from adenoma to carcinoma (12–16). The importance of COX-2 expression in the pathogenesis of colorectal neoplasia is dramatically illustrated by the marked reduction of polyps seen in APC^MIN mice crossed with COX-2 knockout mice (17). The mechanisms by which COX-2 overexpression may contribute to tumorigenesis have been extensively studied and include antiapoptotic, proangiogenic, and proinvasive effects (18). Other gastrointestinal tract neoplasms are associated with increases in COX-2 expression including esophageal and gastric neoplasia (19, 20). However, duodenal neoplasms in the setting of FAP have not, to our knowledge, been investigated for COX-2 expression.

Therefore, to provide information concerning the potential to prevent the progression of duodenal neoplasms in patients with FAP using COX-2 inhibitors, we evaluated the expression of COX-2 in duodenal adenomas of patients with FAP compared with that in duodenal tissues from patients with sporadic duodenal adenomas and duodenal carcinomas as well as with that in colonic mucosa from patients with FAP. In addition, we examined the association between intestinal COX-2 expression and the presence of a common COX-2 promoter polymorphism (−765C→G; refs. 21, 22).

Patients. Patients were selected by searching the Johns Hopkins Surgical Pathology archives for patients with FAP and colonic and duodenal polyps and for patients with sporadic duodenal adenoma and/or duodenal carcinoma. Thirty-six patients with FAP had colon polyps (19 male, mean age: 34.5 ± 18.1 (SD) years; 17 female, mean age 33.8 ± 13.9 years) and 22 patients with FAP had duodenal polyps (7 male, mean age 48.4 ± 11.5 years; 15 female, mean age 44.8 ± 11.7 years). Two patients with FAP had colorectal adenocarcinoma and two had duodenal adenocarcinoma in association with colonic and duodenal polyposis, respectively. In addition, 22 patients with sporadic duodenal adenomas (11 male, mean age: 66.1 ± 12.5 years; 11 female, mean age: 61.2 ± 17.1 years) and 17 patients with sporadic duodenal carcinomas (11 male, mean age: 66.2 ± 9.7 years; 7 female, mean age: 54.1 ± 14.9 years) were included in the analysis (Table 1). To investigate the correlation between the −765G/C COX-2 promoter genotype and COX-2 expression, normal duodenal mucosa was obtained from 93 patients who underwent pancreaticoduodenectomy for pancreatic adenocarcinoma. To determine the prevalence of the COX-2 promoter polymorphism by ethnicity, 219 unselected Caucasian patients and 50 African American patients were genotyped (206 with pancreatic cancer, 46 with familial adenomatous polyposis, 17 with benign gallbladder disease, and 5 with chronic pancreatitis). The study was approved by the Institutional Review Board of the Johns Hopkins University.

Tissue microarrays. Formalin-fixed paraffin embedded tissues were collected to generate 13 tissue microarrays (Table 1). For tissue microarray construction, representative areas containing morphologically defined normal mucosa, adenoma, or carcinoma were identified on an H&E-stained reference slide by an experienced pathologist (C.A.I.D.) and encircled on the paraffin block. Tissue microarrays were constructed using a manual Tissue Puncher/Arrayer (Beecher Instruments, Silver Spring, MD). The diameter of the punched core was 1.4 mm. Serial sections were cut from these tissue microarrays, one of which was stained with H&E as a reference.

Immunohistochemistry. Immunohistochemistry for COX-2 was done on unstained 4 μm sections as previously described (23). An anti–COX-2 monoclonal antibody (Cayman Chemical, Ann Arbor, MI) was used at a dilution of 1:100. Two independent observers (C.A.I.D. and L.A.A.B.) scored the intensity of epithelial COX-2 staining in a semiquantitative manner on a five-grade scale: absent, weak, moderate, strong, or very strong COX-2 labeling (24, 25). Stromal staining was scored separately. In assigning scores, the observers assessed all of the tissue cores on the tissue microarrays. In most cases, multiple cores from multiple polyps and multiple cores from normal mucosa or carcinoma were present. The highest score was used in the statistical analysis.

DNA isolation. Genomic DNA was obtained from deparaffinized formalin-fixed paraffin-embedded tissue of 46 patients with FAP using TK buffer [200 μg/mL of proteinase K and 0.5% Tween 20, 50 mmol/L Tris (pH 9), 1 mmol/L NaCl, 2 mmol/L EDTA]. After overnight incubation in 50 μL TK buffer at 56°C, tubes were incubated at 95°C for 10 minutes to inactivate the proteinase K (26). In addition, DNA was isolated using Qiagen Tissue Kits (Qiagen, Valencia, CA) from fresh-frozen normal duodenum or normal pancreas tissue of 206 patients with pancreatic adenocarcinoma who had undergone Whipple resection, as well as formalin-fixed paraffin-embedded gall bladder tissue from 17 patients with benign gall bladder disease who had undergone cholecystectomy and 5 patients with chronic pancreatitis who had undergone Whipple resection.

Real-time PCR. The −765G → C promoter polymorphism was detected in the SmartCycler1 (Cepheid, Sunnyvale, CA) using the following primers: COX2RealTimeFor: 5′-cattaactatttacagggtaactgcttagg-3′ and COX2RealTimeRev: 5′-ccccctccttgtttcttgga-3′. Fluorescent MGB probes (Applied Biosystems), which were used to detect the G allele (probe 1, 765G: 6-FAM-5′-ctttcccgcctctct-3′) and the C allele (probe 2, 765C: TET-5′-ctttcccccctctct-3′; ref. 27). Samples were assayed in a 26 μL reaction mixture containing 12.5 μL Quantitect Buffer (Qiagen), 1.25 μL of each primer (final concentration 0.5 μmol/L), 0.25 μL of each probe (final concentration 0.1 μmol/L), 9.5 μL diethyl pyrocarbonate–treated H₂O, and 50 ng of sample genomic DNA. PCR reactions were done starting with 94°C for 15 minutes to activate HotStarTaq DNA polymerase, followed by 45 cycles of 94°C for 15 seconds and 60°C for 30 seconds. Two samples with known genotype and a water control were simultaneously assayed in each run.

Sequencing. Fourteen samples were sequenced to validate the single nucleotide polymorphism real-time PCR assay. After initial amplification of the promoter region containing the single nucleotide polymorphism of interest (primers, COX2For: 5′-gcatacgttttggacatttag; COX2Rev: 5′-ctaccttcagtgtacatagc), the PCR product was purified using the QIAquick PCR Purification Kit (Qiagen). Subsequently, samples were sequenced using an internal forward primer (COX2IntFor: 5′-gttttggacatttagcgtcc) and the Applied Biosystems 3730 DNA Analyzer.

Statistics. Nonparametric χ² tests were used to assess differences between groups and to assess the correlation between COX-2 expression and COX-2 genotype. In addition, χ² tests were used to compare the observed genotype prevalence with the expected prevalence of each genotype for a population in Hardy-Weinberg equilibrium.

Cyclooxygenase-2 expression in the colon of patients with familial adenomatous polyposis. COX-2 was expressed in the normal-appearing and adenomatous colonic epithelium of all patients. The level of COX-2 expression in normal colonic mucosa was similar to levels found in colorectal adenomas from patients with FAP. The colon carcinomas from patients with FAP had significantly higher COX-2 expression than normal-appearing (χ² = 14.9, P = 0.0019) and adenomatous mucosa (χ² = 17.9, P = 0.00045; Table 2; Fig. 1A, C, and E).

Cyclooxygenase-2 expression in the duodenum of patients with familial adenomatous polyposis. COX-2 was similarly expressed in the normal duodenal mucosa, duodenal adenomas, and duodenal carcinomas (Table 2; Fig. 1B, D, and F). Normal-appearing duodenal mucosa from FAP patients showed significantly higher levels of COX-2 than normal-appearing FAP colonic mucosa (Fig. 1B versus A; χ² = 12.5, P = 0.014). Similarly, duodenal adenomas from patients with FAP showed higher levels of COX-2 than colonic adenomas from patients with FAP (χ² = 13.3, P = 0.01; Fig. 1C and D).

Cyclooxygenase-2 expression in sporadic and familial adenomatous polyposis duodenal adenomas. Normal duodenal mucosa was available from 15 of 22 patients that had sporadic duodenal adenomas. The majority (80%) showed moderate COX-2 expression in normal duodenal mucosa whereas COX-2 was expressed in all duodenal adenomas. Furthermore, sporadic duodenal adenomas had significantly higher COX-2 expression than normal duodenal mucosa (χ² = 11.4, P = 0.0224) with most adenomas exhibiting strong COX-2 expression (Table 2). Moreover, normal duodenal mucosa from FAP patients showed statistically significantly higher levels of COX-2 than normal duodenal mucosa from patients with sporadic duodenal adenomas (χ² = 10.2, P = 0.037). This observation is further supported by the finding that 69.2% of matched normal and adenomatous duodenal mucosa of FAP patients showed the same COX-2 intensity, whereas only 33.3% of sporadic duodenal adenomas showed the same level of COX-2 expression in their normal duodenal mucosa (χ² = 8.4, P = 0.015; Table 3). These results could not be explained by differences in the grade of adenoma or how the adenoma was obtained (by resection, polypectomy, or biopsy) between patients with FAP and those with sporadic duodenal adenomas.

Cyclooxygenase-2 expression in sporadic duodenal carcinomas. All 17 patients with sporadic duodenal carcinoma expressed COX-2 in normal duodenal mucosa, mostly with moderate labeling intensity. Most of the 17 sporadic duodenal carcinomas exhibited strong or very strong COX-2 immunoreactivity. In addition, all nine adenomas associated with duodenal carcinoma displayed COX-2 immunostaining (Table 2). Duodenal carcinomas had statistically significantly higher levels of COX-2 expression than normal duodenal mucosa (χ² = 7.5, P = 0.024).

Stromal cyclooxygenase-2 immunoreactivity. Stromal COX-2 labeling was focal. It was observed mainly in macrophages underlying the epithelium (Fig. 1A, arrow). COX-2 labeling was strong in eroded areas and seen in macrophages underlying the erosion.

Distribution of the −765G → C polymorphism. In patients with FAP, the GG, GC, and CC genotype frequencies were 65.2%, 28.3%, and 6.5%, respectively (Table 4). The GG, GC, and CC genotypes occurred at 63.2%, 32.4%, and 4.4% in the disease control group, respectively. African Americans were more likely to carry a 765C polymorphism than Caucasians (χ² = 10.01, P = 0.0067; Table 4). All genotypic distributions are in Hardy-Weinberg equilibrium (P ≥ 0.05).

Correlation between −765 genotype and cyclooxygenase-2 expression. Carriers of the −765GG genotype with FAP had higher COX-2 expression in their normal (χ² = 9.4, P = 0.009) colonic and duodenal mucosa than the −765GC/CC carriers (Table 5). In contrast, there was no significant difference in COX-2 expression in the adenomatous mucosa of GG carriers (χ² = 1.4. P = 0.495). There was also no correlation between the level of COX-2 expression in the normal duodenal mucosa of 93 patients with pancreatic cancer and their −765G/C COX-2 genotype (data not shown).

Recent studies indicate that COX-2 activity may be influenced in certain tissues by estrogen (28). However, we found no evidence of any difference in COX-2 expression in normal-appearing mucosa or adenomas by gender.

In this study, we find that COX-2 levels increase with increasing stage of duodenal neoplasia among patients with sporadic disease. This pattern of COX-2 expression has been described for colorectal and other neoplasms and is consistent with a similar adenoma-carcinoma progression sequence for duodenal neoplasia as has been observed for colorectal and other neoplasms (29).

Second, we also find greater COX-2 expression in the normal duodenal mucosa and duodenal adenomas of patients with FAP than in patients with sporadic duodenal neoplasias. Other investigators have noted that normal duodenal mucosa from patients with FAP exhibits increased cell proliferation and ultrastructural changes in cell adhesion function compared with non-FAP controls (30, 31). In addition, several genes, including COX-2, are up-regulated in the normal-appearing colon mucosa of APC^MIN mice and in patients with colorectal cancer (32). Increased COX-2 expression in histologically normal mucosa from patients with FAP may result from impaired Wnt signaling through a possible transcription factor 4 binding element in the COX-2 promoter (33). In addition, gastrin expression is increased by APC inactivation, and in certain tissues gastrin expression can increase COX-2 expression (34, 35).

In addition, we also find that among patients with FAP, COX-2 expression is higher in normal duodenal mucosa than in normal colonic mucosa. Whereas previous studies have compared COX-2 expression in small intestinal cancers and colorectal cancers, where expression patterns were found to be similar (36), COX-2 expression has not been studied in the duodenal tissues of patients with FAP. The higher expression of COX-2 in duodenal mucosa than in colonic mucosa could explain the lower response of duodenal adenomas compared with colonic adenomas to chemoprevention with nonsteroidal anti-inflammatory drugs and COX-2 inhibitors. Other factors that could contribute to differences between colonic and duodenal polyp responses to COX-2 inhibitors, such as greater resistance of duodenal adenomas to apoptosis or differences in the bioavailability of COX-2 inhibitor drugs in the colon versus the duodenum, require investigation. Interestingly, although it is plausible that treatment responses to standard doses of COX-2 inhibitors would be influenced by the amount of COX-2 protein in target tissues, such a relationship has not been clearly shown. Nonsteroidal anti-inflammatory drugs have been shown to reduce COX-2 expression in vitro (37) and in Apc^MIN/+ mouse (38). Indeed, in a previous study examining molecular correlation of adenoma responses and resistance to sulindac, COX-2 expression was lower (although still present) in sulindac-resistant colonic adenomas than in pretreatment adenomas that subsequently regressed with sulindac (23). Our results raise the possibility that higher dosages of COX-2 inhibitors could be more effective against duodenal adenomas as has been suggested for sulindac-resistant adenomas. Future studies should consider measuring duodenal COX-2 expression in patients with FAP undergoing treatment to determine if expression levels predict response to COX-2 inhibitors.

The mechanism(s) responsible for higher COX-2 expression in duodenal mucosa is not known. Several studies have suggested a role for bile acids, such as the unconjugated bile acid chenodeoxycholate, in the development of duodenal neoplasia in Apc^MIN/+ mice (39). Also, a correlation has been observed between the site of a patients' duodenal adenoma development and the site of exposure of their mucosa to bile (40). Ex vivo experiments have shown that COX-2 expression increases in response to exposure to pulses of bile acids and stomach acid (19, 41). Other studies have indicated a chemopreventive effect for ursodeoxycholic acid (42) and combined sulindac and ursodeoxycholic acid in mouse and rat intestine (43).

Finally, we investigated whether a recently reported single nucleotide polymorphism in the COX-2 promoter affected the level of COX-2 expression in intestinal mucosa. Papafili et al. (22) showed that the −765C allele had lower COX-2 promoter activity than the −765G allele. In addition, the −765GC and −765CC genotypes correlate with decreased risk of myocardial infarct and stroke and decreased COX-2 expression in atherosclerotic plaques compared with −765GG (21). We found higher COX-2 expression in the normal, but not in the adenomatous, mucosa of patients with FAP who carried −765GG alleles than in those with the −765GC or −765CC genotype. Because the −765 COX-2 polymorphism influences COX-2 expression in the normal-appearing gastrointestinal mucosa of patients with FAP, it is possible that this polymorphism could influence the number of adenomas that develop in these patients, similar to the effect observed when COX-2 is knocked out in animal models of FAP (17). However, our results indicate that once polyps have formed, COX-2 genotype does not influence COX-2 expression, suggesting that COX-2 genotype may not influence the progression of these neoplasias. Interestingly, we also found that African Americans are significantly more likely to be carriers of −765C alleles, raising the possibility that African Americans could be more prone to the beneficial and adverse effects (such as toxicity from nonsteroidal anti-inflammatory drugs) of having lower COX-2 expression. Further study is needed to assess whether this polymorphism acts as a modifier of FAP phenotype or affects COX-2 expression elsewhere in the gastrointestinal tract, and whether genetic differences in the level of COX-2 expression influence patients' response to COX-2 inhibitors.

There is a need to improve the chemopreventive strategies for duodenal neoplasia occurring in the setting of FAP. In addition to COX-2, several other molecular targets merit consideration in chemoprevention studies including peroxisome proliferator-activated receptors δ and γ (44), epidermal growth factor receptor (45), ornithine decarboxylase (46), and nuclear factor κB pathway (47).

In conclusion, we have found that COX-2 is more highly expressed in the duodenal adenomas and normal duodenal mucosa of patients with FAP and this increased COX-2 expression may contribute to the poorer response of these neoplasms to chemoprevention with COX-2 inhibitors. Further investigation is needed to determine the role of the −765G/C COX-2 promoter polymorphism on COX-2 gene expression in the gastrointestinal tract and its effect on the response to COX-2 inhibitors.

We thank Kieran Brune, Mike Mullendore, and Katharine Romans for technical support.