Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Gastroenteropathy is the most common side effect among patients taking nonsteroidal anti-inflammatory drugs (NSAIDs) for inflammatory disorders, especially rheumatoid arthritis. The pathogenesis of NSAID-induced gastric lesions is generally considered to involve the depletion of endogenous prostaglandins (PGs) caused by inhibition of cyclooxygenase (COX) activity (Vane, 1971; Whittle, 1981). There are two isozymes of COX; COX-1 is a constitutive enzyme and is expressed in various tissues, including the stomach, whereas COX-2 is an inducible enzyme, expressed in few tissues under normal conditions, but rapidly up-regulated in response to various cytokines and growth factors (Feng et al., 1993; Kargman et al., 1993; O'Neill and Fold-Hutchinson, 1993). It is, therefore, assumed that COX-1 is a housekeeping enzyme that maintains gastric mucosal integrity under physiological conditions, whereas COX-2 is responsible for inflammation (Xie et al., 1992; Seibert et al., 1994; Langenbach et al., 1995). Recently, several selective COX-2 inhibitors have been developed as new gastric sparing anti-inflammatory drugs (Futaki et al., 1993; Boyce et al., 1994; Hawkey, 1999; Langman et al., 1999; Goldstein et al., 2000; Silverstein et al., 2000).

We previously reported that the gastric lesions caused by conventional NSAIDs such as indomethacin and aspirin were markedly aggravated in rats with experimentally induced arthritis and suggested an increase in the susceptibility of the gastric mucosa to NSAIDs in arthritic animals (Kato et al., 1999, 2001a,b). The gastric ulcerogenic effect of selective COX-2 inhibitors in arthritic animals remains unknown, however.

In this study, therefore, we investigated the influence of selective COX-2 inhibitors, such as rofecoxib and celecoxib, on the integrity of gastric mucosa in rats with adjuvant-induced arthritis and compared the effects with those of indomethacin, a conventional NSAID.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Animals. Male dark Agouti rats (140~160 g; SLC, Shizuoka, Japan) were used. The animals were fed standard rat chow and tap water ad libitum. All experimental procedures described were approved by the Experimental Animal Research Committee of Kyoto Pharmaceutical University.

Induction of Arthritis. Arthritis was induced by injection of 50 µl of Freund's complete adjuvant (FCA; 10 mg/ml of heat-killed Mycobacterium tuberculosis H37Ra suspended in paraffin oil) into the plantar region of the right hindfoot. Normal rats were housed in the same manner for the same period of time so that aged and batch-matched normal and arthritic rats were used in all of the experiments. The severity of arthritis was assessed by measuring the paw volume (edema) by plethysmometry. Since the paw edema in the left (uninjected) hindfoot was observed from 10 days and reached a maximum 14 days after the injection of FCA, we used the animals of 14 days after the injection in all ulcerogenic experiments as arthritic rats. The animals with or without arthritis were deprived of food but allowed free access to tap water for 18 h before the experiments. In some case, we examined the anti-inflammatory effects of rofecoxib and indomethacin in adjuvant-induced arthritic rats. Chronic and systemic inflammation was evaluated by the development of left (uninjected) paw edema 10 days after FCA injection. Rofecoxib (3-30 mg/kg) and indomethacin (3 mg/kg) were administered orally once daily for 3 days, starting from day 7 following FCA injection.

Evaluation of Gastric Mucosal Lesions. The animals were orally administered indomethacin (3, 10, and 30 mg/kg), rofecoxib (3, 10, 30, and 100 mg/kg), or celecoxib (100 mg/kg) and were killed under deep ether anesthesia 4 h later. Then, the stomachs were removed, inflated by injecting 7 ml of 2% formalin, immersed in 2% formalin for 10 min to fix the gastric wall, and opened along the greater curvature. The area (square millimeters) of each lesion that had developed in the granular mucosa was measured under a dissecting microscope with a square grid (10×), summed per stomach, and used as a lesion score. The person measuring the lesion did not know the treatment given to the animal.

Measurement of PGE₂ in the Gastric Mucosa. The animals were killed by deep ether anesthesia 2 h after oral administration of indomethacin (10 mg/kg) or rofecoxib (3 and 10 mg/kg) on day 14 following FCA injection. The corpus mucosa was isolated, weighed, and put in a tube containing 100% ethanol plus 100 µM indomethacin to prevent further synthesis of PGs; this concentration of indomethacin did not interfere with the measurement of PGE₂ by enzyme immunoassay (Futaki et al., 1994). Then, samples were homogenized with a polytron homogenizer (IKA, Tokyo, Japan) and centrifuged at 12,000 rpm for 10 min at 4°C. After the supernatant of each sample had been evaporated with nitrogen gas, the residue was resolved in assay buffer solution and used for determination of PGE₂. The concentration of PGE₂ was measured using an enzyme immunoassay (Amersham Biosciences, Little Chalfont, Buckinghamshire, UK).

Analyses of COX-1 and COX-2 mRNA by Reverse Transcription-PCR. The animals were killed by deep ether anesthesia on day 14 following FCA injection, and the gastric mucosa was removed, frozen in liquid nitrogen, and stored at 80°C until use. The total RNA of each sample was extracted using Sepharose RNA-I (Nacalai Tesque, Kyoto, Japan). Total RNA primed by random hexadeoxy ribonucleotide was reverse-transcribed with ReverTra Ace-alpha (TOYOBO, Osaka, Japan). The sequences of sense and antisense primers for the rat COX-1 were 5'-AACCGTGTGTGTGACTTGCTGAA-3' and 5'-AGAAAGAGCCCCTCAG-AGCTCAGTG-3', respectively, giving an 887-bp PCR product (Feng et al., 1993). For the rat COX-2, the sequences of sense and antisense primers were 5'-TGATGA-CTGCCCAACTCCCATG-3' and 5'-AATGTTGAAGGTGTCCGGCAGC-3', respectively, giving a 702-bp PCR product (Feng et al., 1993). For the rat glyceraldehyde-3-phosphate dehydrogenase (GAPDH), a constitutively expressed gene, the sequences were 5'-GAACGGGAAGCTCACTGGCATGGC-3' for the sense primer and 5'-TGAGGTCCA CCACCCTGTTGCTG-3' for the antisense primer, giving a 310-bp PCR product (Iso et al., 1995). An aliquot of the reverse transcription reaction product served as a template for PCR, The reaction profile for COX-1 was 35 cycles of denaturation for 30 s at 96°C and annealing for 2 s at 65°C; for COX-2, 35 cycles of denaturation for 30 s at 94°C and annealing for 1 s at 62.5°C; for GAPDH, 25 cycles of denaturation for 30 s at 94°C, annealing for 1 s at 65°C, and extension for 30 s at 74°C on a thermal cycler (TP-240; TAKARA, Shiga, Japan) using KOD dash polymerase (TOYOBO). A portion of the PCR mixture was electrophoresed in 2% agarose gel in Tris-acetic acid-EDTA buffer (40 mM Tris buffer, 2 mM EDTA, and 20 mM acetic acid; pH 8.1), and the gel was stained with ethidium bromide and photographed.

Immunostaining of COX-1 and COX-2. Immunostaining of COX isozymes in the gastric mucosa was performed 14 days after FCA injection. The stomach was removed and washed in phosphate- buffered saline (PBS). Stomachs were fixed in 4% paraformaldehyde for 6 h and washed in PBS containing 10, 15, and then 20% sucrose, in that order. The specimens were embedded in octamer transcription factor compound (Miles, Elkhart, IN) and rapidly frozen in a mixture of dry ice and acetone. Cryostat sections cut serially at a thickness of 6 µm were mounted on silanized slides (DAKO Japan, Kyoto, Japan) and were stained with mouse monoclonal antibody against rat monocytes/macrophages (mouse anti-rat ED-1; Serotec, Oxford, England) and goat polyclonal antibodies against human COX-1 and rat COX-2 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in PBS. Immunohistochemical staining was performed with a streptavidin-biotin peroxidase method according to the manufacturers' instructions using a LSAB2 kit (DAKO Japan) for monocytes/macrophages and goat ImmunoCruz Staining System (Santa Cruz Biotechnology) for COX-1 and COX-2. The sections were finally treated with 0.03% 3,3'-diaminobenzidine (Dojin, Kumamoto, Japan) containing 0.005% hydrogen peroxidase. For double immunostaining, sections were first treated with goat polyclonal anti-rat COX-2 antibody and the goat ImmunoCruz staining system and then with mouse monoclonal anti-rat monocytes/macrophages antibody and the LSAB2 kit. Diaminobenzidine was used for the initial detection, and tetramethylbenzidine was used for the second round of detection. Counterstaining was performed with methyl green (DAKO Japan). After monocytes/macrophages were stained as described above, the infiltrating of the number of monocytes/macrophages were counted in at least five randomly chosen areas of fundic mucosa. The width inspected was 0.5 mm, and the depth of inspection depended on the height (from the base to the top) of the mucosa in the area of observation. The 200× objective of a light microscope was used, and regions to be inspected were measured with reference to the eyepiece of the microscope. Results are expressed as the number of cells per millimeter squared. Counting was done without knowledge of the treatment groups to which specimens belonged.

Determination of Microvascular Peameability. The microvascular permeability was evaluated by measuring the extravasated amount of dye (Evans blue) in the gastric mucosa according to a previous article (Takeuchi et al., 1987). The animals were given p.o. indomethacin (10 mg/kg) or rofecoxib (10 mg/kg) and killed 4 h later. In each case, 1 ml of 1% Evans blue (w/w) was injected i.v. 30 min before the killing. Under deep ether anesthesia, the animals were bled from the descending aorta, the stomach was removed, and the amount of dye that had accumulated in the corpus mucosa and the gastric juice for 30 min was measured. The extraction of dye was performed according to the method described by Katayama et al. (1978). The absorbance of each sample was measured at 620 nm on a Hitachi spectrophotometer (U-2001; Hitachi, Ibaraki, Japan), and the total amount of dye recovered was expressed as micrograms per 100 g of wet tissue.

Preparation of Drugs. Drugs used were heat-killed Mycobacterium tuberculosis (H37Ra: Difco, Detroit, MI), paraffin oil (Wako, Osaka, Japan), indomethacin, Evans blue (Sigma-Aldrich, St. Louis, MO), rofecoxib, and celecoxib (synthesized in our department). All COX inhibitors were suspended in hydroxy-propyl-cellulose (NIPPON SODA, Tokyo, Japan) solution. Evans blue was dissolved in saline. All drugs were prepared immediately before use and administered p.o. in a volume of 0.5 ml/100 g b.wt. or i.v. in a volume of 0.1 ml/100 g b.wt.

Statistical Analysis. Data are presented as the mean ± S.E. for four to six rats per group. Statistical analyses were performed using a two-tailed unpaired t test and Dunnett's multiple comparison test, and values of P < 0.05 were considered to be significant.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Gastric Ulcerogenic Response to Indomethacin, Rofecoxib, and Celecoxib. Oral administration of indomethacin (3, 10, and 30 mg/kg) caused hemorrhagic lesions in the gastric mucosa of normal rats in a dose-dependent manner, the lesion score at the doses of 10 and 30 mg/kg being 2.2 ± 0.9 and 9.8 ± 1.6 mm², respectively (Fig. 1). The gastric ulcerogenic response to indomethacin was markedly aggravated in arthritic rats, and severe lesions were observed even at the dose of 3 mg/kg (Fig. 2). On the other hand, neither rofecoxib (3, 10, 30, and 100 mg/kg) nor celecoxib (100 mg/kg) induced any damage in the stomach of normal rats even at the highest dose (100 mg/kg). In arthritic rats, however, these drugs also provoked gastric lesions at the dose of 10 mg/kg or greater, similar to indomethacin, and for rofecoxib the severity of the lesions increased dose dependently, the lesion score being 1.0 ± 0.71, 13.0 ± 4.0, 25.0 ± 5.1, and 51.8 ± 9.4 mm², respectively.

View larger version (12K): [in this window] [in a new window]   Fig. 1.   Gastric ulcerogenic effect of indomethacin, rofecoxib, and celecoxib in normal rats. Indomethacin (3-30 mg/kg), rofecoxib (3-100 mg/kg), and celecoxib (100 mg/kg) were administered orally, and the animals were killed 4 h later. Data are presented as the mean ± S.E. for four to six rats per group.

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Gastric ulcerogenic effect of indomethacin, rofecoxib and celecoxib in rats with adjuvant-induced arthritis. Arthritis was induced by injection of FCA into the plantar region of the right hindfoot, and the experiments were performed 14 days after the injection. Indomethacin (3-30 mg/kg), rofecoxib (3-100 mg/kg), and celecoxib (100 mg/kg) were administered orally, and the animals were killed 4 h later. Data are presented as the mean ± S.E. for four to six rats per group.

Effects of Rofecoxib and Indomethacin on Gastric Mucosal PGE₂ Content. The mucosal PGE₂ content in normal rat stomachs was 57.9 ± 7.0 pg/g tissue. In arthritic rats, the PGE₂ content was 128.6 ± 11.1 pg/g tissue, significantly higher than in normal rats (Fig. 3). Oral administration of indomethacin (10 mg/kg) almost totally reduced the mucosal PGE₂ content in both normal and arthritic rats, the inhibition being 99.9 and 99.7%, respectively. In contrast, rofecoxib (3 and 10 mg/kg) did not affect the PGE₂ content in normal rats (49.2 ± 5.5 and 63.6 ± 12.7 pg/g tissue, respectively) but significantly decreased the PGE₂ content in arthritic rats toward the level observed in normal rats (80.2 ± 12.2 and 58.5 ± 9.3 pg/g tissue, respectively).

View larger version (35K): [in this window] [in a new window]   Fig. 3.   Effect of indomethacin and rofecoxib on gastric mucosal PGE2 content in normal and arthritic rats. Indomethacin (3 mg/kg) and rofecoxib (10 mg/kg) were administered orally, and the animals were killed 2 h later. The PGE2 content was determined by enzyme immunoassay. Data are presented as the mean ± S.E. for five to six rats. Statistically significant difference at P < 0.05; *, from control (vehicle alone); #, from the corresponding group of normal rats.

Expression of COX-1 and COX-2 in the Stomach of Normal and Arthritic Rats. The expression of COX-2 mRNA was evident in the stomach of arthritic rats on day 14 following the FCA injection, although the gene expression of COX-2 was not detected in the gastric mucosa of normal rats; the relative density of COX-2/GAPDH in normal and arthritic rats were 0.02 ± 0.01 and 0.53 ± 0.08, respectively, and there were significant differences between these rats (Fig. 4). On the other hand, the gene expression of GAPDH and COX-1 was observed in the stomach of both normal and arthritic rats; the relative density of COX-1/GAPDH in normal and arthritic rats were 0.94 ± 0.06 and 0.89 ± 0.12, respectively, and there was no significant difference between these rats.

View larger version (90K): [in this window] [in a new window]   Fig. 4.   Expression of COX-1 and COX-2 mRNAs in the gastric mucosa of normal and arthritic rats. Note that the expression of COX-2 mRNA was not detected in the normal rat but clearly observed in the arthritic rat on day 14 after the FCA injection, whereas COX-1 mRNA was observed in the stomach of both normal and arthritic rats. Lane 1, marker (174/HaeIII digest); lane 2 to 4: normal rats; lane 5 to 7: arthritic rats.

View larger version (160K): [in this window] [in a new window]   Fig. 5.   Immunohistochemical study for COX-1, COX-2, and monocytes/macrophages in the stomach of normal and arthritic rats. A, COX-1 in a normal rat; B, COX-1 in an arthritic rat; C, COX-2 in a normal rat; D, COX-2 in an arthritic rat; E, monocytes/macrophages in a normal rat; F, monocytes/macrophages in an arthritic rat; G, double staining for COX-2 and monocytes/macrophages in an arthritic rat (400×).

Changes in Microvascular Permeability in the Stomach. In normal rats, the amount of extravasated dye in the gastric mucosa after 30 min was 48.3 ± 4.3 µg/g tissue (Fig. 6). This value increased significantly to 70.3 ± 8.9 µg/g tissue after oral administration of indomethacin (10 mg/kg), whereas rofecoxib (10 mg/kg) did not have any effect on the vascular permeability in normal rats (54.2 ± 1.7 µg/g tissue). On the other hand, the vascular permeability was found to be markedly increased to 93.2 ± 2.7 µg/g tissue in arthritic rats. The increased permeability observed in arthritic rats was further significantly enhanced by not only indomethacin but also rofecoxib, the values being 148.7 ± 12.9 and 121.6 ± 5.4 µg/g tissue, respectively.

View larger version (35K): [in this window] [in a new window]   Fig. 6.   Effects of indomethacin and rofecoxib on microvascular permeability in normal and arthritic rats. Indomethacin (3 mg/kg) and rofecoxib (10 mg/kg) were administered orally, and the animals were killed 4 h later. Evans blue (1% v/v) was administered intravenously 30 min before sacrifice. Data are presented as the mean ± S.E. for five to six rats. Statistically significant difference at P < 0.05; *, from control (vehicle alone); #, from the corresponding group of normal rats.

Anti-inflammatory Effects of Rofecoxib and Indomethacin. The injection of FCA into the planter region of the right hindfoot caused severe paw edema in the left (uninjected) hindfoot 10 days later, the increase in paw volume being 0.86 ± 0.07 ml (Table 1). Daily oral administration of rofecoxib (3, 10, and 30 mg/kg) for 3 days, starting from day 7 following FCA injection, significantly prevented the development of paw edema in a dose-dependent manner, the inhibition being 52.0, 72.1, and 87.4%, respectively. Likewise, indomethacin (3 mg/kg) almost totally prevented the paw edema observed after FCA injection (89.8%). On the other hand, the gain of body weight was markedly attenuated by induction of arthritis, the body weight in normal and arthritic rats on day 10 following FCA injection being 160.0 ± 3.8 and 128.2 ± 4.5 g, respectively. Daily administration of rofecoxib (3-30 mg/kg) for 3 days significantly prevented the loss of body weight, the value being 137.0 ± 2.4, 147.0 ± 1.7, and 152.0 ± 2.9 g, respectively. Likewise, indomethacin also prevented the decrease in body weight, the value being 149.8 ± 4.7 g.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Selective COX-2 inhibitors have been used safely in patients without causing gastric damage (Hawkey, 1999; Langman et al., 1999; Goldstein et al., 2000; Silverstein et al., 2000). In the present study, we observed that COX-2 inhibitors, such as rofecoxib and celecoxib, induced gastric mucosal lesions in rats with adjuvant-induced arthritis, similar to conventional NSAIDs such as indomethacin, although they did not cause any damage in normal rat stomachs, unlike indomethacin. We also found that in arthritic rat stomachs, the generation of PGs was significantly enhanced with a marked expression of COX-2. These findings suggest that PGs produced by COX-2 play a part in maintaining the integrity of the gastric mucosa in arthritic rats.

Conventional NSAIDs that inhibit nonselectively both COX-1 and COX-2 activities cause gastrointestinal damage in experimental animals and humans (Vane, 1971; Katayama et al., 1978; Whittle, 1981). COX-1 is constitutively expressed in the gastrointestinal tract and maintains mucosal integrity through the continuous generation of PGs, whereas COX-2 is induced predominantly in certain inflammatory cells by various cytokines, endotoxins, tumor promoters, and growth factors, as well as in response to tissue injury (Feng et al., 1993; Kargman et al., 1993; O'Neill and Fold-Hutchinson, 1993). It is believed that inhibition of COX-1 activity is critical in the ulcerogenic response to NSAIDs, whereas COX-2 mediates the synthesis of PGs responsible for the inflammatory response (Xie et al., 1992; Seibert et al., 1994; Langenbach et al., 1995). It has been demonstrated that new anti-inflammatory drugs with a several hundred-fold higher selectivity for COX-2 have minimal gastric toxicity in animals and humans (Futaki et al., 1993; Boyce et al., 1994; Hawkey, 1999; Langman et al., 1999; Warner et al., 1999; Goldstein et al., 2000; Silverstein et al., 2000). Thus, selective COX-2 inhibitors are believed to be clinically useful as safe new anti-inflammatory drugs.

We recently reported that gastric mucosal lesions induced by NSAIDs, such as aspirin and indomethacin, were significantly aggravated in experimentally induced arthritic animals when compared with normal animals (Kato et al., 1999, 2001a,b). The increased ulcerogenic response to indomethacin depended on the degree of arthritic change, with maximal aggravation observed 14 days after the injection of FCA. In addition, we also reported that healing of chronic gastric ulcers was significantly impaired in arthritic animals (Kato et al., 2001a,b). These findings together suggest that systemic chronic inflammation during adjuvant-induced arthritis has certain deleterious influences on the ulcerogenic and healing responses in the stomach. Although mere speculation, arthritic conditions may modify the gastric ulcerogenic response to selective COX-2 inhibitors. In the present study, we thus investigated whether COX-2 inhibitors, such as rofecoxib and celecoxib, have any ulcerogenic effect in the stomachs of arthritic rats. Consistent with the findings of others, these COX-2 inhibitors did not induce any gastric damage in normal rats. We found, however, that they produced severe hemorrhagic lesions in the stomach of arthritic rats, similar to indomethacin, a conventional NSAID. These results suggest that PGs produced by COX-2 plays a role in maintaining the integrity of the gastric mucosa in arthritic rats.

In the present study, we also observed that the PGE₂ content of the stomach was significantly higher in arthritic rats than normal rats. Certainly, indomethacin reduced the mucosal PGE₂ content in both normal and arthritic rats. In contrast, rofecoxib, a COX-2 inhibitor, did not affect gastric PGE₂ content in normal rats but significantly decreased the PGE₂ content observed in arthritic rats to the level of normal rats. Furthermore, we confirmed the apparent expression of COX-2 in the stomach of arthritic rats but not in normal rats. These results strongly suggest that the increase in PG production in arthritic rat stomachs is brought about by COX-2 activity. Indeed, immunohistochemical analysis showed that COX-1 was expressed in all gastric mucosa of both normal and arthritic rats, with especially strong staining being observed in mucous neck cells and epithelial cells, whereas COX-2 immuno-positive cells were seen in the gastric mucosa of arthritic rats but not normal rats. Interestingly, we also observed that the number of monocytes/macrophages was increased in arthritic rats. When performing double immunostaining for COX-2 and monocytes/macrophages, we found that COX-2-positive cells mostly consisted of monocytes/macrophages. Therefore, it is assumed that COX-2 is expressed mainly in the macrophages in the gastric mucosa of arthritic rats.

It remains unclear, however, how the macrophages accumulate in the gastric mucosa of arthritic rats and how COX-2 is up-regulated in these cells. In the present study, we observed that the microvascular permeability in the rat stomach was increased in arthritic rats compared with normal rats when determined as the amount of extravasated dye. Because it is known that the amount of extravasated dye increases in tissue with inflammation or damage including vascular injury (Menkin and Menkin, 1930; Szabo et al., 1985), it is possible to speculate that the stomach of arthritic rats is in such a condition. If so, it would be understandable in arthritic rats for the macrophages to migrate into the gastric mucosa and COX-2 to be expressed in these cells. Indeed, we observed in the immunohistochemical study that the surface of the gastric mucosal epithelium was rough in arthritic rats compared with normal rats. It is likely, therefore, that some injury might be existed in the stomach of the arthritic rats. Adjuvant arthritis is often used for animal models of rheumatoid arthritis, and these arthritic animals are known to suffer from chronic systemic inflammation and severe pain. It is possible that the increases in COX-2 expression and macrophage number in the stomach occur in association with inflammation or stress caused by pain and that the higher amount of PGE₂ produced by COX-2 plays a role in maintaining the mucosal integrity or inhibiting the progression of mild injury in arthritic rat stomachs. Certainly, further study is required to verify such a concept.

The most interesting finding of this study is that selective COX-2 inhibitors produced gastric mucosal lesions in rats with arthritis. Takahashi et al. (2000) reported that COX-2 inhibitors caused gastric lesions in Helicobacter pylori-infected animals in the stomachs, of which COX-2 had been expressed. We also reported that NS-398, a selective COX-2 selective inhibitor, attenuated the adaptive gastric protection induced by a mild irritant, especially when COX-2 expression was observed in the stomach (Yamamoto et al., 1999). Based on these findings, it can be speculated that selective inhibitors of COX-2 have deleterious influences on the stomach when a significant overexpression of COX-2 occurs under various conditions, including H. pylori infection, severe arthritis and stressful stimulation.

The mechanism by which NSAIDs cause gastric damage still remains unresolved. In general, there is some discrepancy between PGE₂ levels and ulcerogenicity induced by NSAIDs. Indeed, we have previously reported that indomethacin at the dose of 5 mg/kg almost totally inhibited PGE₂ production in the stomach but induced very few gastric damage in rats (Takeuchi et al., 1986). It is assumed that other mechanisms, in addition to PG deficiency, are involved in the gastric ulcerogenicity of NSAIDs. In the present study, rofecoxib at 10 mg/kg totally prevented the increase in PGE₂ contents and produced gross lesions in arthritic rat stomachs while this agent at 3 mg/kg induced only slight damage in the stomach, despite significantly attenuating the increase in PGE₂ contents. Thus, it is likely that the gastric PGE₂ level is not necessarily correlated with the ulcerogenicity of COX-2 inhibitors in arthritic rats, suggesting the involvement of factors other than PG deficiency in this phenomenon. Certainly, we confirmed in this study the anti-inflammatory effects of rofecoxib and indomethacin on adjuvant-induced arthritis in rats. Daily administration of rofecoxib for 3 days dose dependently prevented the development of paw edema at 10 days after FCA injection, and a significant effect was observed even at 3 mg/kg. These results suggest that the doses of rofecoxib used for evaluation of the ulcerogenic effect are within the effective dose range for inflammation in this animal model. It is still unclear at present, however, whether or not COX-2 inhibitors really cause gastric damage in patients with rheumatoid arthritis. Further study should certainly be done on this point in future.

In summary, COX-2 inhibitors, such as rofecoxib and celecoxib, cause gastric mucosal lesions in arthritic rats, similar to conventional NSAIDs such as indomethacin that inhibit nonselectively both COX-1 and COX-2. Thus, it is assumed that PGs derived from COX-2 play an important role in maintaining the integrity of the gastric mucosa in arthritic rats. Finally, when using selective COX-2 inhibitors to treat the patients with rheumatoid arthritis, one should pay attention to the severity and pathogenesis of the disease to prevent side effects.