Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Rofecoxib is one of the first selective cyclooxygenase (COX)-2 inhibitors approved for the treatment of pain and inflammation in osteoarthritis. In this context, it has been described to provide efficacy advantages over acetaminophen and the selective COX-2 inhibitor celecoxib (Geba et al., 2002). It has also been found to be effective in rheumatoid arthritis and as an analgesic in the treatment of acute inflammatory pain after third molar extraction (Morrison et al., 2000) or orthopedic surgery (Reuben and Connelly, 2000). Because rofecoxib does not inhibit COX-1 activity at doses up to 500 mg/day (10 times the highest recommended dose) (Matheson and Figgitt, 2001) and thus does not affect physiological prostaglandin synthesis in the gastrointestinal tract it causes considerably less gastrointestinal toxicity than nonselective COX inhibitors such as naproxen or diclofenac (Bombardier et al., 2000; Hawkey et al., 2000; Gretzer et al., 2001). However, it was noted that rofecoxib causes a relatively strong sodium and water retention (Kammerl et al., 2001; Whelton et al., 2001) and reduction of the glomerular filtration rate (Swan et al., 2000) compared with other selective and nonselective nonsteroidal anti-inflammatory drugs [celecoxib (COX-2-specific) and diclofenac, ibuprofen, indomethacin (nonspecific), respectively] (Zhao et al., 2001). In a recent study based on spontaneous reports of adverse drug reactions in the World Health Organization/Uppsala Monitoring Centre safety database, rofecoxib-treated patients experienced a significantly higher incidence of peripheral (low-extremity) edema than patients treated with celecoxib or unselective NSAIDs (Swan et al., 2000; Zhao et al., 2001). Another group found no difference in the incidence of peripheral edema when half-life-associated dosing patterns of rofecoxib (25 mg once daily), celecoxib (200 mg twice daily), and naproxen (500 mg twice daily) (Schwartz et al., 2001) have been compared. In addition, the VIGOR study has revealed that rofecoxib treatment is associated with an increase of the systolic and diastolic blood pressure, which is more pronounced than that observed with the nonselective COX inhibitor naproxen (Mukherjee et al., 2001). Furthermore, in some studies there were some hints that rofecoxib may increase the risk of cardiovascular diseases, including myocardial infarction compared with naproxen treatment (Bombardier et al., 2000; Rainsford, 2001). Thus, it is discussed that rofecoxib causes an imbalance of prothrombotic and antithrombotic arachidonic acid metabolites, which does not occur with unselective COX inhibitors because nonselective NSAIDs simultaneously inhibit prothrombotic thromboxane synthesis in platelets (through COX-1) and antithrombotic prostacyclin synthesis in endothelial cells (through COX-2), whereas COX-2-selective drugs inhibit only the latter (for review, see Hinz and Brune, 2002). Hence, the COX-2 selectivity of rofecoxib might at least partly contribute to the increase of the cardiovascular risk. Another feature with rofecoxib is the lack of a clear dose-effect relationship concerning its clinical analgesic and also its anti-inflammatory efficacy (Day et al., 2000; Truitt et al., 2001). Thus, there is no reliable information about which dose will probably work for an individual patient. In some studies, there were no differences in efficacy between doses of 12.5, 50, and 125 mg of rofecoxib (Ehrich et al., 1999; Schnitzer et al., 1999).

We have recently observed that the COX-2-mediated anti-inflammatory activity of the other clinically available COX-2 inhibitor, celecoxib, is negatively affected at high doses because at high concentrations celecoxib activates the transcription factor NF-B, thereby causing an increase of NF-B-regulated proinflammatory genes such as COX-2 and TNF (Niederberger et al., 2001). The clinical data of rofecoxib together with the previous results obtained with celecoxib prompted us to hypothesize that the lack of a dose dependency with rofecoxib and its renal unwanted effects might be mediated through alterations of certain transcription factors. We therefore evaluated effects of rofecoxib on the transcription factors NF-B and activating protein-1 (AP-1) and some of their target genes and assessed the anti-inflammatory effects of rofecoxib at low, medium, and very high doses of the drug to find a possible explanation for the special clinical features of rofecoxib outlined above.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Materials. Antibodies for rat COX-2, NF-B, and I-B were obtained from Santa Cruz Biotechnology, Inc. (Heidelberg, Germany). The antibody and the cDNA probe for rat iNOS and the cDNA probe for COX-2 were a kind gift from Prof. J. Pfeilschifter (University of Frankfurt, Frankfurt, Germany) (Xie and Nathan, 1993) and Prof. D. DeWitt (Michigan State University, East Lansing, MI) (DeWitt and Meade, 1993), respectively. Rofecoxib was a gift from Prof. W.J. Wechter (Loma Linda University, Loma Linda, CA). Etoricoxib, used as an analytical internal standard, has been synthesized by Witega (Berlin, Germany). Identity and purity were checked by mass spectrometry and ^1HNMR and was greater than 98.5%. For the animal experiments the commercially available suspension of rofecoxib (Vioxx) (5 mg/ml) was used.

Zymosan-Evoked Inflammation. Male Sprague-Dawley rats (Charles River, Sulzfeld, Germany), weighing 250 to 300 g, were used. They were housed in groups of five in standard cages and maintained in climate- and light-controlled rooms (22 ± 0.5°C, 12-h dark/light cycle). Unilateral hind paw inflammation was induced by subcutaneous injection of 1.25 mg of zymosan (Sigma-Aldrich, Steinheim, Germany) suspended in 100 µl of phosphate buffer into the midplantar region of the right hind paw (Meller and Gebhart, 1997). The paw volume was measured before zymosan injection (time 0) and at 0.25, 0.5, 1, 2, 4, 6, 8, 24, 30, and 96 h using a plethysmometer (Ugo Basile, Varese, Italy) according to the manufacturer's instructions. At each time point, four measurements of the paw volume were taken and the median was used to calculate the percentage of increase of the paw volume compared with the value before zymosan injection (PW). At completion of the experiments rats were deeply anesthetized and killed by cardiac puncture. The spinal cord was rapidly excised and tissue samples from lumbar spinal cord were snap frozen in liquid nitrogen and kept at 80°C until further analysis. In all experiments, the ethics guidelines for investigations in conscious animals were obeyed and the procedures were approved by the local ethics committee for animal research.

Drug Treatment and Data Analysis. Twenty-four hours before drug administration, animals were deprived of food with free access to tap water. Rofecoxib suspension (Vioxx) (2.5 or 5 mg/ml) was administered by gastric gavage at doses of 1, 10, and 50 mg/kg; 50 mg/kg was the highest dose that could be administered (2.5-3 ml of the suspension) to the animals without causing regurgitation or serious diarrhea. Controls received the appropriate volume of tylose suspension. Six to 10 rats were used in each group. The drugs were administered 15 min before the intraplantar injection of zymosan.

Rofecoxib Plasma Concentrations. For determination of the plasma concentrations, blood was taken at time points 1, 2, 3, 5, 8, and 25 h after drug treatment. Rat plasma samples were extracted using a liquid-liquid extraction method. An aliquot of 100 µl of plasma was mixed with the internal standard etoricoxib and 500 µl of 0.1 M sodium carbonate buffer (pH 9.8) and extracted for 15 min with 2 ml of t-butylmethylether. The organic layer was separated and evaporated under a gentle stream of nitrogen. The residue was reconstituted in 100 µl of mobile phase.

Cell Culture. RAW 264.7 mouse macrophages (courtesy of Prof. J. Pfeilschifter, University of Frankfurt) were cultured and incubated in RPMI 1640 medium containing 10% fetal calf serum and 1% penicillin/streptomycin. At this serum concentration, about 10% of the administered rofecoxib concentration was sequestered by protein binding as determined by measuring free and total rofecoxib concentrations in culture medium with HPLC.

Analysis of prostaglandin E₂ (PGE₂), Nitrite/Nitrate, and TNF. RAW-cells were stimulated with 10 µg/ml lipopolysaccharide (LPS) for 24 h in the absence or presence of various concentrations of rofecoxib. PGE₂ concentrations in culture supernatants were assessed using a commercially available enzyme immunoassay (Biotrend Chemicals, Köln, Germany) according to the manufacturer's protocol. The reliable limit of quantification was 36 pg/ml, and the mean percentage deviation over the calibration range of 36 to 5000 pg/ml was less than 15%.

Preparation of Crude Protein Extracts. RAW 264.7 cells were seeded in 10-cm dishes at a density of 5·× 10⁵ cells/dish. At 80% confluence, cells were stimulated with LPS for 24 h in the presence or absence of rofecoxib. Unstimulated cells were used as controls. At the end of the incubation period, cells were washed with PBS and then scraped with a rubber policeman and collected in 1.5-ml tubes. After short centrifugation, the pellet was resuspended in lysis buffer (10 mM Tris-HCl buffer, pH 7.4, containing 20 mM CHAPS, 0.5 mM EDTA, 1 mM DTT, 0.5 mM PMSF, and 1 µM Pefabloc; Alexis, Grünberg, Germany) and kept on ice for 30 min. After sonication for 10 s the suspension was centrifuged at 13,000 rpm in an Eppendorf centrifuge and the supernatant was stored at 80°C until further analysis.

Preparation of Nuclear and Cytosolic Cellular Fractions. Cells were incubated for 30 min with rofecoxib and were then stimulated for 30 min with 10 µg/ml LPS. Cell pellets were resuspended in 1 ml of lysis buffer I (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 1 mM PMSF, and 2 mM DTT) and incubated for 10 min on ice. After addition of Nonidet P-40 (final concentration 0.5%), the solution was vortexed and centrifuged at 400g for 5 min. The supernatant was kept as the cytosolic fraction. The nuclear pellet was washed with lysis buffer I. Pellets were then resuspended in 2 volumes of lysis buffer II (20 mM HEPES-KOH, pH 7.4, 600 mM KCl, 0.2 mM EDTA, 1 mM PMSF, and 2 mM DTT) and incubated for 30 min on ice. After centrifugation (10,000g for 10 min), the supernatant was diluted by the addition of 1 volume of lysis buffer III (20 mM HEPES-KOH, pH 7.4, 0.2 mM EDTA, 0.5 mM PMSF, and 2 mM DTT). Glycerol was added to obtain a final concentration of 20% and aliquots were stored at 80°C.

Western Blot Analysis. Proteins of cell lysates (50 µg), nuclear extracts (20 µg), or spinal cord homogenates (30 µg) were separated electrophoretically by 10 or 12% SDS-PAGE and then transferred onto nitrocellulose membranes by semidry blotting. The membrane was incubated overnight at 4°C in blocking buffer (5% skimmed milk in PBS/0.3% Tween 20). It was then incubated with the primary antibodies diluted 1:100 in blocking buffer for 90 min at room temperature. After washing in PBS/0.3% Tween 20, it was incubated for 60 min with a secondary antibody conjugated with peroxidase. Protein-antibody complexes were detected with the enhanced chemiluminescence system (Amersham Biosciences). Densitometric analysis of the blots were performed with Quantity One software (Bio-Rad, Munich, Germany).

Electrophoretic Mobility Shift Assay (EMSA). Nuclear extracts (5 µg) were incubated in 10% glycerol, 10 mM HEPES-KOH, pH 7.9, 50 mM KCl, 4 mM MgCl₂, 4 mM Tris-HCl, 0.5 mM DTT, 0.5 mM EDTA, 1 µg of bovine serum albumin, and 1 µg of poly(dI-dC) together with 25 fmol of [³²P]ATP-labeled oligonucleotide in a final volume of 20 µl for 30 min at room temperature. The oligonucleotide sequence corresponds to the NF-B binding site (5'-AGTTGAGGGGACTTTCCCAGGC-3') and the AP-1 binding site (5'-CGCTTGATGACTCAGCCGGAA-3'), respectively. For competition experiments, a 100-fold molar excess of unlabeled probe was added to the reaction 15 min before addition of the radiolabeled probe. The nucleotide-protein complex was separated on a 6% native polyacrylamide gel in 0.25× Tris borate-EDTA buffer (0.5×: 45 mM Tris-borate 45, 1 mM EDTA) at 100 V at room temperature. The gel was dried and radioactive bands were detected by autoradiography.

Northern Blot Analysis. Total RNA was isolated from the cells by the method of Chomczynski (1993). Total RNA (20 µg) was dissolved in 10 µl of H₂O, mixed with 10 µl of denaturing solution (500 µl of formamide, 162 µl of 37% formaldehyde, and 100 µl of 0.2 M MOPS), and incubated at 60°C for 15 min. RNA loading dye (4 µl) was added. Total RNA was separated on a 1% agarose gel containing 1% formaldehyde (80 V, 4 h). The RNA was then transferred to a nylon membrane overnight with 10× SSC (1.5 M sodium chloride, 0.15 M sodium citrate) and immobilized with a UV transilluminator (254 nm, 150 mJ).

Real-Time PCR. RNA was isolated as described above. Two micrograms of total RNA has been used for the reverse transcription, which was performed with a reverse transcription kit (QIAGEN GmbH, Hilden, Germany).

(Ct)

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Effects of Rofecoxib on Transcription Factor Regulation of NF-B and AP-1 in RAW 264.7 Cells. The transcription factor NF-B consists of two dimers (p50/p65), which are bound to the inhibitor I-B and thus sequestered in the cytoplasm in unstimulated cells. Upon stimulation, I-B is phosphorylated by I-B kinases, subsequently ubiquitinylated, and then degraded by a proteasome complex. Degradation of I-B allows NF-B to translocate into the nucleus where it binds to the promoter region of various genes such as COX-2, TNF, or iNOS and activates their transcription (for reviews, see Baeuerle, 1998; Pahl, 1999). In unstimulated control cells, DNA binding of NF-B was minimal, whereas treatment with LPS considerably increased its DNA binding activity. The specificity of the NF-B DNA binding was confirmed by the reversal of the binding by a 100-fold molar excess of unlabeled probe (data not shown). Rofecoxib dose dependently (1, 10, and 100 µM) reduced the LPS stimulated DNA binding activity of NF-B. At 100 µM, effects of rofecoxib were indistinguishable from those of the positive control Bay 11-7085, which is an inhibitor of I-B kinase. Figure 1A shows the results of a representative experiment.

View larger version (32K): [in this window] [in a new window]   Fig. 1.   A, EMSA. RAW 264.7 cells were preincubated for 30 min with rofecoxib at the indicated concentrations and then stimulated with 10 µg/ml LPS for further 30 min. The nuclear fractions were extracted and 5 µg was subjected to EMSA using a 32P-labeled NF-B consensus oligonucleotide. B, Western blot analysis of the p65 subunit of NF-B in nuclear extracts and I-B in cytosolic extracts. Cells were treated for 30 min with 1, 10, and 100 µM rofecoxib followed by the stimulation with 10 µg/ml LPS for further 30 min. C, EMSA with RAW cells preincubated for 30 min with rofecoxib at the indicated concentrations and then stimulated with 10 µg/ml LPS for further 30 min. Nuclear fractions were subjected to EMSA using a 32P-labeled AP-1 consensus oligonucleotide. D, Western blot analysis of c-Fos and c-Jun in nuclear extracts of RAW 264.7 macrophages. Cells were treated as described above. Representative results of four experiments are shown.

Expression of NF-B- and AP-1-Regulated Genes in RAW 264.7. To evaluate whether the observed inhibition of NF-B and the activation of AP-1 affected the expression of proinflammatory genes, we assessed the expression of COX-2 and iNOS. Northern blot analyses as well as real-time PCR (Fig. 2A) revealed that iNOS and COX-2 mRNA, which were only slightly detectable in unstimulated RAW 264.7 cells, considerably increased after stimulation with LPS. This was associated with a significant increase of NO₂/NO₃ and PGE₂ release (Fig. 2, B and C). When cells were treated with the positive control dexamethasone, both COX-2 and iNOS mRNA expression was reduced. Incubation with rofecoxib also resulted in a reduction of iNOS and COX-2 mRNA. However, this inhibitory effect showed no clear dose dependency.

View larger version (28K): [in this window] [in a new window]   Fig. 2.   A, expression of iNOS and COX-2 mRNA in RAW 264.7 cells stimulated with LPS for 24 h in the absence or presence of rofecoxib or dexamethasone at the indicated concentrations. Top, Northern blot analysis of iNOS and COX-2; 20 µg of RNA was separated in a 1% agarose gel, blotted on a nylon membrane, and then hybridized with the radioactively labeled probe. The blots show representative results of four experiments. 18S RNA was assessed as a loading control. Bottom, real-time reverse transcription-PCR. One hundred nanograms of RNA-equivalent was subjected to real-time PCR in an SDS 7700 with Sybr Green staining. Ct values were calculated with 18S RNA as internal standard. The diagram shows the relative amount of mRNA compared with the LPS stimulated control, which was set as 1. B, release of nitrite/nitrate in RAW 264.7 stimulated for 24 h with 10 µg/ml LPS in the absence or presence of rofecoxib or 0.5 µM dexamethasone (Dex). Nitrite/nitrate concentrations (mean ± S.E.M.) were measured by the Griess method (*, statistical significance mean difference, p = 0.05). C, effects of rofecoxib on the release of PGE2 (mean ± S.E.M.) in RAW264.7. Cells were treated as described above. PGE2 concentrations were measured by an enzyme immunoassay (*, statistical significance mean difference, p = 0.05).

Effects of Rofecoxib on Zymosan-Evoked Inflammation in Rats. Because, in contrast to celecoxib, rofecoxib has not activated NF-B but rather inhibited its DNA binding activity, we hypothesized that its anti-inflammatory activity should not be abolished at high doses. We tested this hypothesis in the zymosan-induced paw inflammation model in rats. In vehicle-treated rats, intraplantar injection of 1.25 mg of zymosan led to a maximum increase of the paw volume of 110.3 ± 5.3% (mean ± S.E.M.). As hypothesized, rofecoxib inhibited paw inflammation at doses of 1, 10, and 50 mg/kg (Fig. 3A). Statistical comparison of the area under the paw volume increase versus time curves (AUC_PW from 0-24 h) revealed statistically significant differences in treatment means: F(3,25) = 13.54, p < 0.001. Results of the post hoc analysis are shown in Fig. 3B. Interestingly, 50 mg/kg produced significantly stronger anti-inflammatory effects than 1 and 10 mg/kg, whereas the anti-inflammatory effect of 10 mg/kg rofecoxib was not stronger than that with 1 mg/kg.

View larger version (12K): [in this window] [in a new window]   Fig. 3.   A, time course of the anti-inflammatory effects of rofecoxib in zymosan-induced hind paw inflammation in rats (mean ± S.E.M.). Six to 10 rats were used in each group. Rofecoxib was administered orally 15 min before an intraplantar injection of 1.25 mg of zymosan. Control animals received the appropriate volume of tylose slime. B, for statistical comparison of drug effects, the areas under the paw volume increase versus time curves (AUCPW from 0-24 h, mean ± S.E.M.) were calculated using the linear trapezoidal rule and subjected to univariate analysis of variance with subsequent Bonferroni post hoc tests. The box represents the interquartile range, the line within the box the median, the ends of the "whiskers" show the 5th and 95th percentile. Open dots show individual values. (* and ***, statistically significant mean difference with p < 0.05 and p < 0.001, respectively).

View larger version (13K): [in this window] [in a new window]   Fig. 4.   Mean ± S.E.M. plasma concentration-time curves of rofecoxib in rats after the oral administration of 1 mg/kg (), 10 mg/kg (), and 50 mg/kg ().

View larger version (19K): [in this window] [in a new window]   Fig. 5.   Expression of COX-2, iNOS, and TNF in rat lumbar spinal cord as assessed by Western blot analysis (COX-2, iNOS) or immunoassay (TNF). The spinal cord was excised 96 h after induction of a peripheral inflammation by injection of 1.25 mg of zymosan into one hind paw. Animals (n = 3-4 for each dose) were treated with a single oral dose of 1, 10, and 50 mg/kg rofecoxib 15 min before the zymosan injection. A, top, Western blot analysis of iNOS and COX-2. The blots show one representative experiment of three. Bottom, densitometric analysis of all investigated spinal cord samples (three per value; mean ± S.E.M.). Zymosan-treated control rats were set as 1. B, TNF expression (* and **, statistically significant mean difference with p < 0.05 and p < 0.01, respectively.).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Rofecoxib is a selective COX-2 inhibitor with anti-inflammatory and analgesic efficacy and a low gastrointestinal toxicity compared with conventional NSAIDs. Although rofecoxib does not inhibit COX-1 at therapeutically relevant doses, its use is associated with a relatively high incidence of renal side effects. This may be partly, although not exactly defined, due to the constitutive expression of COX-2 in the kidney. Depending on the study design, rofecoxib treatment is controversially discussed to be more frequently associated with salt and water retention, reduction of the glomerular filtration rate, reports of acute renal failure, and increase in blood pressure compared with other selective (celecoxib) or unselective NSAIDs (Swan et al., 2000; Kammerl et al., 2001; Schwartz et al., 2001; Whelton et al., 2001; Zhao et al., 2001). Another feature with rofecoxib is that its effects are somewhat arbitrarily associated with the concentration or dose, i.e., it is not clear whether a high dose will provide higher efficacy than a lower one. In the present study, we assessed the influence of rofecoxib on transcription factor regulation as a potential explanation for the above-mentioned controversial discussion.

The major finding of the present study is that rofecoxib inhibited NF-B but increased AP-1 DNA binding activity. Because NF-B and AP-1 regulate similar genes, but not necessarily in the same direction, the simultaneous action on both transcription factors is probably the cause for the here-observed, at the first view, confusing pattern of up- and down-regulation of proinflammatory genes, including COX-2, TNF, and iNOS. For example, the simultaneous inhibition of NF-B and activation of AP-1 might have caused the significant decrease of iNOS protein expression in the spinal cord at the highest rofecoxib dose because iNOS transcription is stimulated by NF-B but inhibited by AP-1 (Kleinert et al., 1998). Thus, the observed inhibition of iNOS expression is probably the sum of both effects and might have contributed to the strong anti-inflammatory effects of the highest rofecoxib dose (50 mg/kg), particularly in the late phase of the zymosan-induced paw edema. Nevertheless, it cannot be excluded that the stronger effect of rofecoxib at 50 mg/kg compared with 1 and 10 mg/kg might be caused by higher plasma concentrations and therefore longer inhibition of COX-2 activity rather than by additional effects on transcription factors.

The effects of rofecoxib on NF-B are directly opposed to those observed with the other less potent "coxib", celecoxib in a previous study. In that study, celecoxib was shown to activate NF-B at high concentrations. This resulted in a complete loss of its anti-inflammatory efficacy at high doses (100-200 mg/kg) (Niederberger et al., 2001). Because rofecoxib has no NF-B activating effects, it is not surprising that rofecoxib did not lose its anti-inflammatory efficacy even at very high doses. For rofecoxib, the 50-mg/kg dose was the maximum that could be safely administered to rats and considering its about 10 times higher potency in terms of COX-2 inhibition, this dose is even more potent to inhibit COX-2 than the celecoxib doses of 100 to 200 mg/kg at which anti-inflammatory effects of celecoxib were found to be abolished. These data demonstrate that there are considerable differences among these two COX-2 inhibitors. In terms of NF-B inhibition, the effects of rofecoxib are more similar to those of acetylsalicylic acid, salicylic acid, or flurbiprofen, which have been previously found to inhibit NF-B activation (Muller et al., 2001; Tegeder et al., 2001). In contrast to these drugs, however, rofecoxib did not inhibit the nuclear translocation of NF-B but prevented its DNA binding activity, suggesting that although the resultant inhibition of NF-B-dependent gene transcription is similar, the step at which NF-B activation is inhibited is obviously different. Rofecoxib's stimulating effects on AP-1 also clearly differ from those of other NSAIDs, because both acetylsalicylic acid (Huang et al., 1997) and flurbiprofen (Tegeder et al., 2001) were found to inhibit AP-1 DNA binding. This may explain why flurbiprofen had no effect on iNOS protein expression, whereas rofecoxib reduced its zymosan-induced up-regulation at high doses.

COX-1 inhibition and thereby reduction of thromboxane synthesis and platelet aggregation is the major mechanism underlying the beneficial cardiovascular effects of aspirin. Recently, it has been suggested, that the treatment of patients with COX-2-selective drugs may be associated with a higher risk of cardiovascular events (Boers, 2001; Hennan et al., 2001; Mukherjee et al., 2001), probably because these drugs may increase prothrombotic activity by decreasing the vasodilatory and antiaggregatory prostacyclin production without a simultaneous inhibition of platelet aggregation. This feature, however, is shared by all COX-2 inhibitors (Zhao et al., 2001). In addition, some studies showed that the incidence of peripheral edemas under treatment with rofecoxib was about twice that observed with celecoxib or other nonselective NSAIDs (Zhao et al., 2001).

Considering that AP-1 regulates the transcription of renal sodium channels (Otulakowski et al., 1999), it may be hypothesized that the rofecoxib-induced activation of AP-1 may be involved in the increase of salt and water retention that occurs during treatment with this drug. The action of rofecoxib on the expression of certain sodium transporters in the kidney remains to be evaluated.

In summary, we show in the present study that the pattern of transcription factor regulation caused by rofecoxib is specific for this coxib and may explain the inconsistent pattern of up- and down-regulation of proinflammatory genes such as COX-2, iNOS, and TNF. This may be the reason for the occurrence of side effects such as salt and water retention and the lack of a linear dose dependence of the analgesic and anti-inflammatory effects of rofecoxib (Day et al., 2000; Truitt et al., 2001).