Cyclooxygenase 2-Selective and Nonselective Nonsteroidal Anti-Inflammatory Drugs Induce Oxidative Stress by Up-Regulating Vascular NADPH Oxidases

Nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit the activity of cyclooxygenase (COX). COX is the pacemaker enzyme in the synthesis of prostaglandins and thromboxane from arachidonic acid. Two isoforms of COX have been identified; COX1 is constitutively expressed in many tissues and generates prostanoids mediating normal physiological functions. COX2, on the other hand, is induced at sites of inflammation, but it is also expressed in response to physiologic stimuli (see below). Aspirin and other nonselective NSAIDs inhibit both COX isoenzymes. COX2-selective inhibitors (coxibs) were developed to provide the anti-inflammatory and analgesic effects of NSAIDs with reduced gastrointestinal toxicity (Antman et al., 2005).

However, serious cardiovascular side effects have been reported for coxibs, including rofecoxib, celecoxib, parecoxib, and valdecoxib (Bresalier et al., 2005; Nussmeier et al., 2005; Solomon et al., 2005). Compared with placebo, allocation to a coxib is associated with a 42% relative increase in the incidence of total vascular events (1.2%/year versus 0.9%/year), mainly attributable to an increased risk of myocardial infarction (0.6%/year versus 0.3%/year), with little apparent difference in other vascular outcomes: stroke (0.4%/year versus 0.4%/year) and vascular death (coxib 0.3%/year versus placebo 0.2%/year) (Kearney et al., 2006). The most widely accepted hypothesis to explain the greater numbers of cardiovascular events associated with coxibs is that these agents cause an imbalance between vascular prostacyclin and platelet thromboxane (Antman et al., 2005). Platelets are without COX2 and produce thromboxane via COX1. Endothelial cells express COX1, and the shear stress of the flowing blood stimulates COX2 expression (Inoue et al., 2002). In fact, COX2 is the dominant source of prostacyclin from endothelial cells in vivo. Thus, according to the prostanoid imbalance theory, COX2 inhibition by coxibs would markedly reduce prostacyclin production from endothelial cells without inhibiting COX1 in platelets, and thus lead to a thromboxane overproduction and promote platelet-dependent thrombosis (Antman et al., 2005).

However, there is an increasing body of evidence indicating that this paradigm may be too simplistic and cannot explain the available clinical data (Hermann et al., 2005). For example, addition of aspirin (a preferential COX1 inhibitor) does not prevent the adverse cardiovascular effects of coxibs (Nussmeier et al., 2005). Moreover, also nonselective NSAIDs, which inhibit primarily COX1, have been found to be associated with an increased rate of cardiovascular events (Hippisley-Cox and Coupland, 2005; Kearney et al., 2006; McGettigan and Henry, 2006).

Thus, prostanoid imbalance cannot explain all of the adverse cardiovascular effects of coxibs and traditional NSAIDs observed in clinical studies, and other pathogenic mechanisms have to be postulated.

Over the last years, evidence has accumulated showing that oxidative stress plays an important role in the pathogenesis of cardiovascular disease (Griendling and Fitz-Gerald, 2003; Förstermann and Münzel, 2006), and there are some hints that NSAIDs may affect the redox status of vascular tissues. For example, rofecoxib has been shown to increase the susceptibility of human low-density lipoprotein and cell membrane lipids to oxidative modification (Walter et al., 2004). Rofecoxib can also promote nonenzymatic formation of isoprostanes from biological lipids (Mason et al., 2007). In contrast, in salt-sensitive Dahl rats fed a high-sodium diet, elevated levels of 8-isoprostane and endothelial dysfunction were normalized by celecoxib and not affected by rofecoxib or diclofenac (Hermann et al., 2003).

Therefore, the current study was designed to investigate whether NSAIDs induce or reduce vascular oxidative stress, and if so, by which mechanism. Cardiovascular side effects of NSAIDs are being observed more frequently in patients with pre-existing cardiovascular disease (Bennett et al., 2005). Patients with hypertension, dyslipidemia, diabetes mellitus, congestive heart failure, or chronic renal disease seem to have a significantly higher risk associated with the use of NSAIDs compared with patients without these conditions (Huang et al., 2006). For this reason, we performed our experiments in an animal model of cardiovascular disease, the spontaneously hypertensive rat (SHR). Similar to human pathology, these animals show an enhanced production of reactive oxygen species (ROS) in their vasculature (Nabha et al., 2005).

Materials and Methods

Animals and Treatment with NSAIDs. Male SHR (6 months of age) were obtained from Charles River Laboratories (Sulzfeld, Germany) and assigned randomly to receive rofecoxib [4-(4-methylsulfonylphenyl)-3-phenyl-5H-furan-2-one, CAS number 162011-90-7, chemical formula C₁₇H₁₄O₄S, 10 mg/kg/day] (MSD Sharp & Dohme, Munich, Germany; obtained before the withdrawal in October 2004), celecoxib [4-[5-(4-methylphenyl)-3-(trifluoromethyl)pyrazol-1-yl]benzenesulfonamide, CAS number 169590-42-5, chemical formula C₁₇H₁₄F₃N₃O₂S, 30 mg/kg/day] (Pfizer, Karlsruhe, Germany), diclofenac [2-(2-(2,6-dichlorophenylamino)-phenyl)acetic acid, CAS number 15307-86-5, chemical formula C₁₄H₁₁Cl₂NO₂, 5 mg/kg/day] (Sigma-Aldrich, Deisenhofen, Germany), or naproxen [(+)-(S)-2-(6-methoxynaphthalen-2-yl)propanoic acid, CAS number 22204-53-1, chemical formula C₁₄H₁₄O₃, 50 mg/kg/day] (Sigma-Aldrich) mixed into the chow for 2 weeks. The doses of the NSAIDs were extrapolated from doses applied to humans. COX1 activity was not affected at the doses selected for rofecoxib (Höcherl et al., 2002) and celecoxib (Kitahara et al., 2002). Apocynin [1-(4-hydroxy-3-methoxyphenyl)ethanone, CAS number 498-02-2, chemical formula C₉H₁₀O₃ (Sigma-Aldrich), 70 mg/kg/day], a selective NADPH oxidase inhibitor, was given in drinking water. The use of animals in this study complies with the Institute of Laboratory Animal Resources (1996) and the German law on the protection of animals. The investigation conforms with the principles outlined in the Declaration of Helsinki.

Real-Time Reverse Transcription-Polymerase Chain Reaction for mRNA Expression of NADPH Oxidase Subunits and Endothelial Nitric-Oxide Synthase. mRNA expression of endothelial nitric-oxide synthase (eNOS), Nox isoforms of NADPH oxidase (Nox1, Nox2, and Nox4), and NADPH oxidase subunit p22phox was analyzed with quantitative real-time reverse transcription-polymerase chain reaction (RT-PCR) using an iCyclerTM iQ System (Bio-Rad Laboratories, Munich, Germany), as described previously (Li et al., 2006). Total RNA was isolated from aorta and heart of SHR using an RNeasy Fibrous Tissue Kit (QIAGEN, Hilden, Germany). Five hundred nanograms of total RNA were used for real-time RT-PCR analysis with the QuantiTect Probe RT-PCR kit (QIAGEN). Gene expression of the target genes was normalized to the endogenous control, glyceraldehyde 3-phosphate dehydrogenase mRNA, and the amount of target gene mRNA expression in each sample was expressed relative to that of control.

Measurement of Superoxide Concentrations in Heart Membrane Fractions. Rat heart tissues were homogenized (glass/glass) in buffer containing 50 mM Tris-HCl, pH 7.4, 10 mM dithiothreitol, and a protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany), and centrifuged at 2000g for 5 min at room temperature. The supernatant was removed and centrifuged at 20,000g for 20 min at 4°C. Again the supernatant was removed and centrifuged at 100,000g for 60 min at 4°C. The pellet was resuspended in 50 mM Tris-HCl, pH 7.4, containing protease inhibitors, and further diluted in phosphate-buffered saline to give a final protein concentration of 0.7 mg/ml and a final dithiothreitol concentration of 200 μM. The lucigenin (5 μM)-derived chemiluminescence of the membrane suspensions was detected in a Lumat LB 9507 (Berthold Technologies, Bad Wildbad, Germany) in the presence of 200 μM NADPH (Daiber et al., 2004).

Measurement of Aortic Superoxide Concentrations by Lucigenin Chemiluminescence. Aortas from SHR were cut into 3-mm rings and equilibrated in a Krebs-HEPES buffer for 30 min at 37°C, with or without the nitric-oxide synthase (NOS) inhibitor N^G-nitro-l-arginine methyl ester (l-NAME; 1 mM). Then, lucigenin (5 μM)-derived chemiluminescence was detected in a Lumat LB 9507 (Berthold Technologies). The photon counts were normalized for the counting time and the dry weight of aortic tissues (Oelze et al., 2006).

Measurement of Aortic ROS Content by L-012 Chemiluminescence. Aorta rings at 3-mm length were incubated for 30 min at 37°C in 96-well plates in Hanks' balanced salt solution (PAA Laboratories, Cölbe, Germany) containing 100 μM of the luminol derivative L-012 (Daiber et al., 2004; Li et al., 2006). L-012-derived chemiluminescence was measured using a Microplate Luminometer (Berthold Technologies). The photon counts were normalized for the dry weight of aortic tissue.

Measurement of ROS Content by Dihydroethidium Fluorescence. The oxidative fluorescent dye dihydroethidium (DHE) was used to evaluate the amount of superoxide in situ, as described previously (Li et al., 2006). Frozen sections (10 μm) from aorta or heart were incubated with 10 μM DHE. Photographs were taken using a fluorescent microscope at an excitation wavelength of 520 nm and an emission wavelength of 610 nm. To analyze the potential contribution of eNOS to ROS production, some sections were preincubated with l-NAME (1 mM) for 30 min (Li et al., 2006).

Plasma Total Peroxide As a Marker of Oxidative Stress. Plasma total peroxide concentrations were determined photometrically in EDTA plasma using a PerOx assay kit (Immundiagnostik, Bensheim, Germany), which is based on the reaction of horseradish peroxidase with plasma peroxides using tetramethylbenzidine as a chromogen substrate (450-nm wavelength) (Hildebrandt et al., 2002).

Western Blot Analysis. Isolated aortic tissue was frozen and homogenized in liquid nitrogen. Proteins were separated by SDS-polyacrylamide gel electrophoresis and blotted onto a nitrocellulose membrane. After blocking, immunoblotting was performed with a mouse monoclonal phosphorylated vasodilator-stimulated phospho-protein (P-VASP) phosphoserine 239 antibody (clone 16C2) (Merck, Darmstadt, Germany). α-Actinin was detected using a mouse monoclonal antibody (Sigma-Aldrich) for normalization. The bands were evaluated by densitometry (Oelze et al., 2000).

Detection of Nitrotyrosine in Aortic Sections. Frozen sections of aorta (10 μm) were incubated with a rabbit polyclonal anti-nitrotyrosine antibody (catalog no. 06-284; Millipore Bioscience Research Reagents, Temecula, CA) at a dilution of 1:100. Biotinylated goat anti-rabbit IgG (Vector Laboratories, Burlingame, CA) was used as secondary antibody. After addition of avidin/biotinylated enzyme (peroxidase) complex (Vectastain ABC kit; Vector Laboratories), the reaction was visualized with 3,3′-diaminobenzidine. Finally, the sections were additionally stained with hematoxylin.

Determination of Nitric-Oxide Synthesis As Nitrite in Rat Plasma. Nitrite, the oxidation product of nitric oxide (NO), was assayed in rat plasma as a measure of NO synthesis. Nitrite was determined by ozone chemiluminescence after chemical reduction to NO using a NOA 280 Nitric Oxide Analyzer (Sievers, Boulder, CO) (Li et al., 2006).

Measurement of Aortic cGMP Content. Aorta segments in 3-mm length were equilibrated for 60 min at 37°C in Hanks' balanced salt solution containing 200 U/ml superoxide dismutase, and 300 μM 3-isobutyl-1-methylxanthine, 100 μM l-arginine, with or without 20 μMBH₄. Then, 300 nM acetylcholine was added, and an incubation for 3 min was performed to stimulate eNOS. Aortic cGMP content was determined by radioimmunoassay.

Reporter Cell Assay for Determination of Bioactive NO in Human Endothelial Cells. Human EA.hy 926 endothelial cells, derived from human umbilical vein endothelial cells, were treated with either rofecoxib, celecoxib, or diclofenac (10 μM each) for 24 h. NO production was bioassayed with RFL-6 rat lung fibroblasts used as reporter cells (Li and Förstermann, 2000). In brief, after the 24-h pretreatment, EA.hy 926 cells were incubated for 3 min with the respective NSAID (10 μM) and 300 μM ATP. Then, the conditioned media were transferred onto the RFL-6 cells. On the RFL-6 cells, another 3-min incubation was performed. The cGMP content of the RFL-6 samples was determined by radioimmunoassay as described previously (Li and Förstermann, 2000).

Assessment of Nitroglycerin-Induced Dilatation of the Right Brachial Artery in Healthy Human Volunteers. Ten healthy volunteers (for baseline clinical characteristics see Table 1) were randomly assigned to 5-day treatments with either rofecoxib (50 mg once daily), celecoxib (200 mg twice daily), diclofenac (75 mg twice daily), or placebo. Volunteers received all three NSAIDs, but only seven of the subjects received rofecoxib due to the restricted availability of the drug. All volunteers were from the scientific laboratory personnel, and all gave informed consent to participate in the study. Nitroglycerin-induced dilatation was assessed before and in the morning of the last day of a 5-day treatment period (2 h after the last NSAID dose). Brachial arterial diameter was measured noninvasively (Warnholtz et al., 2005) immediately before and 4 min after sublingual administration of nitroglycerin (0.8 mg Nitrolingual; Pohl-Boskamp, Hohenlockstedt, Germany). Two-dimensional baseline images of brachial artery diameter were obtained with an ATL HDI5000 ultrasound system (Philips, Best, The Netherlands). Drug-free periods of 9 days were allowed between the different NSAID treatments.

Statistics. Changes of brachial artery diameter in response to nitroglycerin in control and NSAID-treated individuals were compared for statistical differences using Student's t test for paired values. Because the rofecoxib group did not show Gaussian distribution, the Wilcoxon signed-rank test was used. For all other comparisons of different means, analyses of variance followed by Fisher's protected least-significant-difference test were performed.

Results

NSAIDs Enhance the Expression of NADPH Oxidases in Aorta and Heart of SHR. Treatment of SHR with rofecoxib (10 mg/kg/day), celecoxib (30 mg/kg/day), diclofenac (5 mg/kg/day), or naproxen (50 mg/kg/day) for 2 weeks resulted in enhanced mRNA expression of NADPH oxidases in the aorta (Fig. 1A) as well as in the heart (Fig. 1B). In the aorta and in the heart, a marked increase of Nox1, Nox2, and Nox4 was seen with the nonselective NSAIDs diclofenac and naproxen. Rofecoxib and celecoxib also increased the expression of these Nox isoforms, but to a lesser extent. p22phox was significantly increased by diclofenac and naproxen, whereas rofecoxib and celecoxib had little (heart) or no effect (aorta).

NSAIDs Induce Oxidative Stress in SHR. Treatment of SHR with NSAIDs resulted in increased vascular ROS content as determined with DHE staining (Fig. 2A) and L-012 chemiluminescence (Fig. 2C). In accordance with the effects of NSAIDs on the expression of NADPH oxidases (Fig. 1), diclofenac and naproxen were more efficacious than rofecoxib or celecoxib at enhancing the superoxide content of the aorta from treated SHR. ROS content in heart was also increased by NSAIDs, as measured by DHE staining (Fig. 2B) and by lucigenin chemiluminescence (Fig. 2D). In agreement with the enhanced oxidative stress in cardiovascular tissues, plasma concentrations of total peroxides tended to be higher in NSAID-treated SHR (Fig. 2E).

Treatment with apocynin, a NADPH oxidase inhibitor, prevented diclofenac-induced ROS content, both in the aorta (Fig. 3, A and C) and the heart (Fig. 3, B and D).

NSAIDs Decrease the Level of Bioactive NO, but Up-Regulate eNOS Expression. Treatment with NSAIDs decreased the plasma levels of nitrite, the oxidation product of NO (Fig. 4A). However, eNOS expression was not reduced but was even up-regulated by diclofenac or naproxen (Fig. 4B). In the aorta, vasodilator-stimulated phosphoprotein (VASP) phosphorylation was decreased by treatment with NSAIDs (Fig. 4, C and D), indicating a reduction in the levels of bioactive NO. Also in this respect, diclofenac and naproxen showed stronger effects than rofecoxib and celecoxib.

The increased superoxide concentrations in diclofenac-treated animals were reduced by the NOS inhibitor l-NAME, as determined with L-012 or lucigenin chemiluminescence (Fig. 5, A and B). Aortas from diclofenac-treated SHR showed a reduced cGMP content, which could be reversed by the NOS cofactor BH₄ (Fig. 5C). In diclofenac-treated SHR, an enhanced accumulation of protein nitrotyrosine was observed in the aorta, which was prevented by apocynin (Fig. 5D).

NSAIDs Increase Nox2 Expression and Decrease the Level of Bioactive NO in Cultured Human Endothelial Cells. Nox2 mRNA expression was increased in human EA.hy 926 endothelial cells by rofecoxib or diclofenac but not by celecoxib (10 μM each, 24 h; Fig. 6A).

When RFL-6 reporter cells were treated with conditioned media from nonstimulated EA.hy 926 endothelial cells, a basal cGMP level of 2.2 ± 0.1 pmol/10⁶ cells was measured. The cGMP level increased to 33.7 ± 1.5 pmol/10⁶ cells, when endothelial cells were stimulated with 300 μM ATP (Fig. 6B). This increase could be totally prevented with the NOS inhibitor l-NAME (1 mM), which reduced the cGMP level to 2.3 ± 0.2 pmol/10⁶ cells (data not shown). Pretreatment of EA.hy 926 cells for 24 h with rofecoxib or diclofenac significantly reduced cGMP levels in the RFL-6 reporter cell assay; a 24-h treatment with celecoxib only showed a tendency for reduced cGMP levels (Fig. 6B).

NSAIDs Impair Nitroglycerin-Induced Dilatation of the Human Brachial Artery. Healthy volunteers received the NSAIDs rofecoxib, celecoxib, or diclofenac at the highest recommended daily dose for 5 days (Fig. 7). Before and after the 5-day treatment with one of the NSAIDs, nitroglycerin-induced (NO-mediated) dilatation of the right brachial artery was measured noninvasively using high-resolution vascular ultrasound. All three NSAIDs impaired nitroglycerin-induced dilatation significantly (Fig. 7).

Discussion

The present study provides evidence in favor of a novel theory explaining the adverse cardiovascular effects of coxibs and non-selective NSAIDs. Our data demonstrate that both COX2-selective coxibs (rofecoxib and celecoxib) and, to an even greater extent, nonselective NSAIDs (diclofenac and naproxen) induce oxidative stress in the cardiovascular system of spontaneously hypertensive rats. This is indicated by enhanced vascular superoxide content (Fig. 2, A and C) as well as elevated plasma peroxides (at least for diclofenac; Fig. 2E). The increased superoxide content is consistent with an enhanced expression of NADPH oxidases in aorta and heart (Fig. 1). Treatment with the NADPH oxidase inhibitor apocynin prevented the enhanced vascular superoxide production by diclofenac (Fig. 3). In addition, an “uncoupled,” dysfunctional eNOS (see below) can contribute to oxidative stress.

Plasma nitrite was decreased in NSAID-treated animals (Fig. 4A), pointing to a diminished synthesis of vascular NO. In principle, this could be the result of a decreased expression of eNOS. However, as demonstrated in Fig. 4B, eNOS mRNA expression was not reduced; diclofenac and naproxen even (paradoxically) up-regulated eNOS mRNA expression. The increased expression of eNOS mRNA is probably a consequence of an increased production of the reactive oxygen species H₂O₂.H₂O₂, the dismutation product of superoxide, has been shown to increase eNOS expression at the transcriptional and post-transcriptional levels (Drummond et al., 2000). In accordance with this hypothesis, the NSAIDs producing the most oxidative stress (diclofenac and naproxen; Fig. 2) were the ones that produced the greatest up-regulation of eNOS expression (Fig. 4B).

An involvement of the inducible NO synthase (iNOS) in the NSAID-induced reduction in plasma nitrite seems unlikely. It has been demonstrated that iNOS is expressed in aorta and heart of SHR (Vaziri et al., 2000). However, NSAIDs (diclofenac or celecoxib) did not modify iNOS expression or plasma nitrite/nitrate levels in a rat model (Hamilton and Warner, 1998). Thus, the changes in plasma nitrite in response to NSAIDs observed in our study probably reflect NO production from eNOS.

Our data further indicate that the decreased NO production in response to NSAIDs can be explained by a loss of eNOS enzyme activity or functionality. Enhanced oxidative stress can lead to an uncoupling of oxygen reduction from NO synthesis in eNOS (Förstermann and Münzel, 2006). An uncoupled eNOS will produce superoxide at the expense of NO. We and others have shown previously that in vascular diseases associated with oxidative stress, eNOS is also up-regulated but dysfunctional. Thus, the NSAIDs are likely to produce oxidative stress and eNOS uncoupling similar to that observed in animal models of cardiovascular disease such as hypertension, hyperlipidemia, and diabetes mellitus (Förstermann and Münzel, 2006).

Indeed, the increased superoxide content in diclofenac-treated SHR could be reduced by the NOS inhibitor l-NAME (Fig. 5, A and B), indicating that eNOS was in a dysfunctional state under these conditions. In many cardiovascular disease states associated with eNOS uncoupling, oxidative degradation of the NOS cofactor BH₄ is probably the common cause of eNOS dysfunction (Förstermann and Münzel, 2006). This seems to be true also in NSAID-induced eNOS uncoupling, because substitution with BH₄ restored eNOS functionality (Fig. 5C). It is important to note that not all ROS are able to oxidize BH₄. Particularly peroxynitrite—the direct reaction product of NO and superoxide—is able to do so (Förstermann and Münzel, 2006). Peroxynitrite also causes tyrosine nitration of proteins, and, indeed, NSAIDs led to an enhanced nitrotyrosine accumulation (Fig. 5D).

Additional evidence for eNOS uncoupling after NSAID administration came from the demonstration of a reduced VASP phosphorylation (Fig. 4, C and D) in the aorta. This indicates a functional impairment of the NO/cGMP/cGMP-dependent protein kinase (cGK) pathway (most probably caused by the diminished formation of bioactive NO). VASP is a validated substrate of cGMP-dependent protein kinase (cGK), and P-VASP can be used as an indicator of the integrity of the NO/cGMP/cGK pathway (Oelze et al., 2000). The reduction in P-VASP was more pronounced with the nonselective NSAIDs than with coxibs.

When exposing human endothelial cells to COX inhibitors, rofecoxib and diclofenac increased Nox2 expression and reduced cGMP (the second messenger generated in response to bioactive NO) significantly, whereas celecoxib did not (Fig. 6, A and B). This result is interesting in view of clinical trials demonstrating a more favorable cardiovascular risk profile of celecoxib relative to rofecoxib (McGettigan and Henry, 2006). In an animal study, celecoxib even improved endothelial function, whereas rofecoxib did not (Hermann et al., 2003). Different intrinsic physicochemical properties of celecoxib (sulfonamide) and rofecoxib (sulfone) may be responsible (in part) for the difference in cardiovascular effects of the two coxibs (Walter et al., 2004; Mason et al., 2007).

It is established that coxibs and nonselective NSAIDs can produce increases in blood pressure in rats (Höcherl et al., 2002; Hermann et al., 2003) and in humans (Mukherjee et al., 2001). However, NSAID-induced up-regulation of NADPH oxidases and the resulting oxidative stress are probably independent of changes in blood pressure, because they could also be demonstrated in cultured endothelial cells (Fig. 6, A and B).

As a last piece of evidence, the current study demonstrates in healthy volunteers that the nitroglycerin-induced, NO-mediated vasodilation was impaired by NSAIDs (Fig. 7). The reduced relaxation in response to nitroglycerin could be caused by several mechanisms: 1) enhanced oxidative stress by NSAIDs (induction of NADPH oxidase and uncoupling of eNOS) can reduce the bioavailability of nitroglycerin-derived NO (Mollnau et al., 2002); 2) NSAIDs, by inducing oxidative stress, may lower the expression and activity of guanylyl cyclase (Münzel et al., 2005; Gerassimou et al., 2007); and 3) endogenous prostacyclin can contribute to the relaxing effect of nitroglycerin (Hink et al., 2003). Thus, NSAIDs could also reduce the relaxation in response to nitroglycerin by blocking prostacyclin synthesis.

Because eNOS-derived NO is vasoprotective, whereas vascular oxidative stress facilitates vasospasm and promotes atherosclerosis (Förstermann and Münzel, 2006), the data of the current study can (at least in part) provide a potential pathomechanism for the outcome of clinical studies demonstrating that coxibs and nonselective NSAIDs increase cardiovascular events, especially in high-risk patients.

Whereas both coxibs and nonselective COX inhibitors produce oxidative stress, this was more pronounced with the nonselective NSAIDs (Fig. 2). This correlates with a clinical study reporting a similar risk of myocardial infarction for celecoxib and rofecoxib but a somewhat higher risk for diclofenac (Hippisley-Cox and Coupland, 2005). However, in other studies, rofecoxib has been found to be associated with a higher risk for cardiovascular disease than diclofenac (Kearney et al., 2006; McGettigan and Henry, 2006). Thus, oxidative stress probably represents one of the mechanisms responsible for adverse cardiovascular effects of NSAIDs, but it is not the only one. Other contributing mechanisms include effects on prostanoids regulating platelet-vessel wall interactions. For instance, the nonselective NSAIDs prevent platelet aggregation/adhesion (by inhibiting COX1-dependent thromboxane formation), whereas coxibs reduce preferentially vessel wall-derived prostacyclin (Antman et al., 2005). This could be the reason why total cardiovascular risk for nonselective NSAIDs is not higher than that for coxibs.

In conclusion, we provide evidence that COX inhibition by NSAIDs enhances the expression of vascular and cardiac NADPH oxidases and reduces vascular NO bioavailability in SHR and human endothelial cells. This seems to have functional consequences even in man in vivo, because the NO-mediated nitroglycerin-induced vasodilation was clearly attenuated. At least in SHR, coxibs produce less oxidative stress than the nonselective NSAIDs. The enhanced production of reactive oxygen species in response to COX inhibition may be responsible, at least in part, for the adverse cardiovascular effects of NSAIDs observed in clinical studies.