Effects of cyclooxygenase inhibition on cardiovascular function in a hypercholesterolemic swine model of chronic ischemia

Abstract

The cardiovascular effects of cyclooxygenase (COX) inhibition remain controversial, especially in the setting of cardiovascular comorbidities. We examined the effects of nonselective and selective COX inhibition on cardiovascular function in a hypercholesterolemic swine model of chronic ischemia. Twenty-four intact male Yorkshire swine underwent left circumflex ameroid constrictor placement and were subsequently given either no drug (HCC; n = 8), a nonselective COX inhibitor (440 mg/day naproxen; HCNS; n = 8), or a selective COX-2 inhibitor (200 mg/day celecoxib; HCCX; n = 8). After 7 wk, myocardial functional was measured and myocardium from the nonischemic ventricle and ischemic area-at-risk (AAR) were analyzed. Regional function as measured by segmental shortening was improved in the AAR of HCCX compared with HCC. There was no significant difference in perfusion to the nonischemic ventricle between groups, but myocardial perfusion in the AAR was significantly improved in the HCCX group compared with controls at rest and during pacing. Endothelium-dependent microvessel relaxation was diminished by ischemia in HCC animals, but both naproxen and celecoxib improved vessel relaxation in the AAR compared with controls, and also decreased the vasoconstrictive response to serotonin. Thromboxane levels in the AAR were decreased in both HCNS and HCCX compared with HCC, whereas prostacyclin levels were decreased only in HCNS, corresponding to a decrease in prostacyclin synthase expression. Chronic ischemia increased apoptosis in Troponin T negative cells and intramyocardial fibrosis, both of which were reduced by celecoxib administration in the AAR. Capillary density was decreased in both the HCNS and HCCX groups. Protein oxidative stress was decreased in both HCNS and HCCX, whereas lipid oxidative stress was decreased only in the HCCX group. Thus nonselective and especially selective COX inhibition may have beneficial myocardial effects in the setting of hypercholesterolemia and chronic ischemia. Whether these effects modulate cardiovascular risk in patients taking these drugs remains to be seen, but evidence to date suggests that they do not.

nonsteroidal anti-inflammatory drugs (NSAIDs) function by inhibiting the conversion of arachadonic acid into prostaglandin H₂ by cyclooxygenase (COX), inhibiting the production of a family of prostanoids that includes prostacyclin and thromboxane. Unfortunately, the gastrointestinal side effects of NSAIDs somewhat limit their use. The constitutively expressed COX isoform, COX-1, is involved in maintenance of gastrointestinal mucosa and platelet aggregation, whereas the inducible isoform, COX-2, mediates inflammation and pain at sites of cellular injury (24). Based on these findings, selective COX-2 inhibitors were developed to specifically target the undesirable effects of COX-2 while sparing the physiological effects of COX-1. However, in 2000 the Vioxx gastrointestinal outcomes research (VIGOR) trial revealed that, although the COX-2 inhibitor rofecoxib (Vioxx) did decrease the incidence of gastrointestinal ulcers compared with naproxen, it was also associated with a higher incidence of myocardial infarction (3). Since that time, some studies have also associated nonselective NSAIDs with increased cardiac risk (13), whereas others have reported no increased risk of cardiovascular events with either nonselective or selective NSAIDs (6). There have been numerous studies examining the safety of celecoxib, with the majority showing no increased cardiovascular risk. However, the Adenoma Prevention with Celecoxib (APC) trial found that, although twice-daily celecoxib had a significant antitumor effect, there was also a significantly increased number of adverse cardiovascular events compared with placebo, with a significant interaction between history of atherosclerotic heart disease and adverse events (2). In a similar trial in which patients were given celecoxib only once daily, there was no significant difference in cardiovascular risk between celecoxib- and placebo-treated patients (1).

The cardiovascular effects of COX are thought to derive from the balance between two downstream products, prostacyclin and thromboxane A₂. Prostacyclin, primarily produced by endothelial cells, is a vasodilator and inhibits platelet aggregation, whereas thromboxane A₂, produced primarily by platelets, is a vasoconstrictor and promotes platelet aggregation. FitzGerald (11) hypothesized that selective blockade of prostacyclin synthesis by COX-2 inhibitors tips prostanoid balance in favor of thromboxane A₂, leading to thrombogenesis, vasoconstriction, and cardiovascular complications. However, the variation in clinical outcomes of these drugs demonstrates that their cardiovascular effects may not be so clear cut. Notably, studies on patients with comorbidities such as hypercholesterolemia and chronic ischemia, which affect a significant proportion of patients with cardiovascular disease, are lacking.

Chronic hypercholesterolemia, by altering levels of vasoactive substances including nitric oxide and endothelin and increasing oxidative stress, decreases the ability of myocardial microvessels to relax (16), termed endothelial dysfunction. In this study, we used a swine model of hypercholesterolemia and chronic ischemia to examine the effects of nonselective COX and selective COX-2 inhibition on cardiovascular function. We hypothesized that the endothelial dysfunction and increased oxidative stress resulting from hypercholesterolemia might sensitize the heart and vessels to the deleterious effects of nonselective COX and selective COX-2 inhibition.

MATERIALS AND METHODS

Swine.

Yorkshire (Parsons Farm, Amherst, MA) were fed a high cholesterol diet (Sinclair Research, Columbia, MO) starting at 4 wk of age and continuing for the duration of the experiment. At 8 wk of age, swine were made chronically ischemic by left circumflex coronary artery (LCx) ameroid placement (Research Instruments SW, Escondido, CA) and divided into three groups. One group received no drug (HCC; n = 8), one group received 440 mg orally per day of the nonselective COX inhibitor naproxen (HCNS; n = 8), and the third group received 200 mg orally per day of the selective COX-2 inhibitor celecoxib (HCCX; n = 8). The mean body weights of swine in the three groups at the time of death were 33.2 ± 0.9 kg in HCC, 31.07 ± 0.9 in HCNS, and 31.05 ± 0.8 in HCCX (P = 0.17).

Surgical procedures.

Anesthesia was induced with intramuscular telazol (4.4 mg/kg) and maintained with 3.0% isoflurane. Animals were intubated and mechanically ventilated. Titanium ameroid constrictors (1.75–2.25 mm internal diameter, depending on vessel size) were placed around the proximal LCx via left thoracotomy. After 7 wk of drug treatment, swine were again anesthetized, and whole blood samples were obtained. Naproxen and celecoxib were discontinued 1 day before the final surgery to avoid confounding results with acute drug effects. After midline sternotomy, measurements were performed, followed by cardiac harvest. Thick transverse slices (1 cm) were taken through the left ventricle (LV), and the resulting rings were divided into eight sections each. Myocardial samples were rapidly frozen in liquid nitrogen (histologic and molecular studies), placed in 4°C Krebs solution (microvessel studies), or dried at 60 degrees (microsphere analysis).

All experiments were approved by the Beth Israel Deaconess Medical Center and Rhode Island Hospital Institutional Animal Care and Use Committees. Animals were cared for in accordance with the Principles of Laboratory Animal Care formulated by the National Society for Medical Research and the Guide for the Care and Use of Laboratory Animals (NIH publication no. 85-23, revised 1996).

Measurement of global and regional myocardial function.

Heart rate (HR), mean arterial pressure (MAP), developed left ventricular pressure (DLVP), first derivative of LV pressure (+dP/dt), and regional LV function in the ischemic area-at-risk (AAR) were recorded before cardiac harvest using single-sensor pressure catheters inserted into the femoral artery and apex of the heart (Millar Instruments, Houston, TX) and the Sonometrics system (Sonometrics London, ON, Canada) as described previously (25). Rate pressure product (RPP), defined as HR multiplied by MAP, was calculated for each animal.

Coronary angiography.

X-ray coronary angiography with iohexol (Omnipaque; GE Healthcare, Princeton, NJ) was carried out via femoral artery approach to verify LCx occlusion at the time of death. A 5-French AR-1 catheter (Cordis, Bridgewater, NJ) was advanced into the right and left coronary artery ostia, and 4 ml of contrast was injected per side to visualize coronary vessels.

Myocardial perfusion analysis.

Myocardial perfusion was measured with isotope-labeled microspheres (BioPAL, Worcester, MA). Briefly, 1.5 × 10⁷ gold-labeled microspheres were injected during temporary LCx occlusion to demarcate the nonischemic ventricle (NV) and the AAR. Lutetium (resting HR) and Europium (pacing to 150 beats/min)-labeled microspheres were injected at the final procedure while simultaneously withdrawing arterial blood from a femoral artery catheter. LV samples from the NV and AAR were dried in a 60°C oven for at least 48 h, then exposed to neutron beams and microsphere densities measured using a γ-counter (BioPAL). Myocardial blood flow was determined using the following equation: blood flow = (withdrawal rate/tissue weight) × (tissue microsphere count/blood microsphere count).

Microvessel studies.

Coronary arterioles (80–180 μm diameter) were isolated and placed in a microvessel chamber as described previously (5). Vessels were preconstricted with thromboxane A₂ analog U46619 (0.1–1.0 μM), then treated with endothelium-independent vasodilator sodium nitroprusside (SNP; 10⁻⁹ to 10⁻⁴ mol/l) and endothelium-dependent vasodilator ADP (ADP; 10⁻⁹ to 10⁻⁴ mol/l). Responses were defined as percent relaxation of the preconstricted diameter. Other vessels were treated with the vasoactive agent serotonin (10⁻⁹ to 10⁻⁵ mol/l), and percent constriction was measured. All reagents were obtained from Sigma-Aldrich (St. Louis, MO).

Tissue prostaglandin assays.

Tissue levels of the stable breakdown products of thromboxane A₂ and prostacyclin, 11-dehydrothromboxane-B₂ (11-d-TXB₂) and 6-keto prostaglandin-F_1α (6-k-PGF_1α), respectively, were measured using ELISA kits according to the manufacturer's specifications (Neogen, Lexington, KY). Briefly, tissue lysates underwent liquid-liquid exchange, extraction, and concentration and were then loaded into 96-well plates containing antibody to either 11-d-TXB₂ or 6-k-PGF_1α. Plates were washed, and colorimetric substrate was added. Absorbance was read at 650 nm, and results were plotted against a standard curve.

Immunoblotting.

Sixty micrograms of total protein from myocardial homogenates were fractionated by SDS-PAGE (Invitrogen, San Diego, CA) and transferred to PVDF membranes (Millipore, Bedford, MA). Membranes were incubated with antibodies against thromboxane synthase (Abcam, San Francisco, CA), prostacyclin synthase (Cayman Chemical, Ann Arbor, MI), proangiogenic and vasodilatory proteins VEGF and endothelial nitric oxide synthase (eNOS), cell survival proteins Erk 1/2 and phosphorylated-Erk 1/2, anti-apoptotic protein Bcl-2, apoptosis marker poly-ADP ribose polymerase (PARP), caspase-9 (C9), cleaved caspase-3 (Asp175; Cell Signaling Technology, Danvers, MA), and RIP (H-207) (Santa Cruz Biotechnology, Santa Cruz, CA) at dilutions recommended by the manufacturer, followed by the appropriate HRP-linked secondary antibodies (Jackson ImmunoResearch, West Grove, PA). Immune complexes were detected with chemiluminescence (Amersham, Piscataway, NJ) and digitalized using G:Box (Syngene, Cambridge, England). Densitometry was performed using ImageJ software (National Institutes of Health, Bethesda, MD). Bands were normalized to GAPDH expression (Cell Signaling).

Oxidative stress.

To assess protein oxidation, dinitrophenylhydrazine-derivatized tissue homogenates were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were incubated with primary antibody specific to dinitrophenylhydrazine, followed by incubation with HRP-linked secondary antibody per manufacturer's recommendations (Millipore, Billerica, MA). Immune complexes were visualized with chemiluminescence. Densitometry was performed on entire lanes using ImageJ software to measure total protein oxidation. Lipid oxidation of issue homogenates was measured using the Bioxytech MDA-586 spectrophotometric assay (OxisResearch, Burlingame, CA). Briefly, samples were incubated with probucol, hydrochloric acid, and N-methyl-2-phenylindole, which reacts with malondialdehyde (MDA), a major product of lipid oxidation, to produce a stable dye. Absorbance at 586 nm was read using a SynergyMx plate reader (Biotek, Winooski, VT) and plotted against a standard curve. Results are expressed as [MDA] in micrometers.

Immunofluorescence staining for vascular density.

Frozen sections of myocardium (12 μm thick) were formalin fixed, then incubated with antibodies against porcine CD-31 (R&D Systems, Minneapolis, MN) and smooth muscle actin (SMA; Sigma-Aldrich), followed by the appropriate Alexa fluor conjugated secondary antibodies (Jackson ImmunoResearch). Sections were mounted in Vectashield (Vector Laboratories, Burlingame, CA). Photomicrographs were taken with a Zeiss Axiolab microscope (Carl Zeiss, Thornwood, NY) under 20× magnification. Capillaries, defined as CD31-positive structures between 5 and 25 μm² in cross-sectional area, and arterioles, defined as structures costaining for both CD31 and SMA, were counted using ImageJ software (National Institute of Health, Bethesda, MD). Data is presented as vessels per squared millimeters.

TUNEL staining for apoptosis.

Frozen sections of myocardium (12 μm thick) were fixed in formalin. Apoptotic cells were identified using the ApopTag Red kit (Chemicon International, Temecula, CA) to detect DNA breaks according to the manufacturer's specifications (Chemicon International). Three 20× images of each section were taken, and apoptotic nuclei counted using ImageJ software. The results of the three images were averaged to obtain the number of apoptotic nuclei for each section. Results are expressed as nuclei per squared millimeter. In double labeling with TUNEL, a Troponin T mouse monoclonal antibody was used at 1:300 dilution (United States Biological, Swampscott, MA).

Trichrome staining for fibrosis.

Frozen sections of myocardium (12 μm thick) were fixed in formalin, then stained with Masson's trichrome stain. Digital images of the stained slices were captured using Aperio slide scanning software (Aperio Technologies, Vista, CA). The amount of blue-stained collagen was quantified in a blinded fashion for three randomly selected 10× fields per section using ImageJ software (NIH, Bethesda, MD) and expressed as a percentage of the total section area. Measurements from the three fields were averaged to obtain representative percent fibrosis for each section.

Statistical analysis.

All results are presented as means ± SE. Microvessel responses are expressed as percent relaxation of the preconstricted diameter and were analyzed using two-way, repeated-measures ANOVA with a post hoc Bonferroni test. All other comparisons were carried out using one-way ANOVA with a post hoc Bonferroni test using GraphPad Prism 5.0 Software (GraphPad Software, San Diego, CA).

RESULTS

Hypercholesterolemic swine model.

All groups achieved mean serum cholesterol levels consistent with hypercholesterolemia (HCC, 207 ± 33 mg/dL; HCNS, 254 ± 58 mg/dL; HCCX, 244 ± 22 mg/dL; P = 0.78). All values were at least 2.5-fold higher than reported serum cholesterol levels for normocholesterolemic swine (86 ± 7 mg/dL; n = 6) (4). Coronary angiography at the terminal procedure demonstrated complete LCx occlusion by the ameroid constrictor in all animals.

Global and regional ventricular function.

There was no significant difference in global LV function (MAP, HR, DLVP, or +dP/dt) between groups, although the HCNS and HCCX groups tended to have lower MAP, HR, DLVP, and RPP compared with controls (Fig. 1, A–E).

Fig. 1.

Hemodynamic parameters. Shown are mean arterial blood pressure (MAP, A), heart rate (HR, B), developed LV pressure (DLVP, C), LV contractility (+dP/dt, D), and rate pressure product (RPP, E) measurements for all three groups. There was no significant difference between groups in any parameter.

Segmental shortening along the vertical axis in the AAR was not significantly different between groups (Fig. 2A), but there was a strong trend toward improved horizontal segmental shortening in the HCNS group and a significant improvement in horizontal segmental shortening in the HCCX group compared with control (P < 0.05; Fig. 2B).

Fig. 2.

Regional function. Shown are % segmental shortening in ischemic myocardium along the vertical axis (VSS, A) and horizontal axis (HSS, B). Horizontal segmental shortening tended to be increased in both drug-treated groups compared with control, but was significantly increased only in the HCCX group. *P < 0.05.

Myocardial perfusion.

Myocardial perfusion in the AAR under resting conditions trended toward improvement in the HCNS group compared with control and was significantly increased in the HCCX group compared with control (P < 0.01; Fig. 3A). Under paced conditions, blood flow in the AAR was similar between the HCC and HCNS groups, but still significantly increased in the HCCX group compared with the other groups (P < 0.05; Fig. 3B).

Fig. 3.

Myocardial perfusion. Perfusion in the ischemic and nonischemic myocardium was measured using radiolabeled microspheres at rest and under ventricular pacing. Perfusion was increased in ischemic territory of the HCCX group compared with the HCC group both at rest (A) and under ventricular pacing (B), while there was no significant increase in perfusion in the HCNS group. There was no difference in perfusion to the nonischemic myocardium between groups (C,D) *P < 0.05, †P < 0.01.

There was no difference in perfusion in the NV between groups either at rest or under conditions of ventricular pacing (Fig. 3, C and D).

Microvascular reactivity.

There were no significant differences in baseline microvessel diameter or percent preconstriction between groups. Vessel relaxation response to the endothelium-dependent vasodilator ADP was decreased in the AAR of HCC animals compared with the NV. However, both the HCNS and HCCX groups demonstrated significantly improved relaxation responses in the AAR compared with HCC (P < 0.05, P < 0.01, respectively; Fig. 4A). The relaxation response to the endothelium-independent vasodilator SNP was similar between all groups (Fig. 4B). The contractile response to serotonin was similar between HCC-NV and HCC-AAR, but diminished in the AAR of both HCNS (P < 0.01) and HCCX (P < 0.06) groups (Fig. 4C).

Fig. 4.

Microvessel function. Microvessel relaxation response to the endothelium-dependent vasodilator adenosine diphosphate (ADP) and the endothelium-independent vasodilator sodium nitroprusside (SNP) were measured and expressed as % relaxation from preconstricted diameters. Microvessel contractile response to serotonin was expressed as % contraction from baseline diameter. Ischemia resulted in worsened relaxation response to ADP in HCC animals, but this effect was abrogated by both HCNS and HCCX treatment (A). There was no difference in relaxation response to SNP between groups (B). The contractile response to serotonin was not significantly different between the AAR and NV of HCC animals (C). Only the HCNS group demonstrated significantly reduced contractile response to serotonin compared with HCC, though the contractile response in HCCX tended to be less than HCC as well (P = 0.06). *P < 0.05, †P < 0.01.

Tissue prostaglandin levels.

Levels of 11-d-TXB₂ were increased in the AAR of HCC animals compared with the NV, but both naproxen and celecoxib administration significantly decreased tissue thromboxane in the AAR (P < 0.001; Fig. 5A). Levels of 6-keto-PGF_1α were increased by ischemia in the HCC group as well, but were significantly decreased in the AAR of HCNS animals alone (P < 0.001; Fig. 5B). Prostacyclin levels in the AAR were unaffected by celecoxib treatment.

Fig. 5.

Tissue prostanoid levels. Shown are tissue levels of 11-d-thromboxane B₂(11-d-TXB₂), the stable metabolite of thromboxane A₂, and 6keto-prostaglandin F_1α (6k-PGF_1α), the stable metabolite of prostacyclin. 11-d-TXB₂was increased in the AAR of HCC animals, but HCNS and HCCX treatment significantly inhibited 11-d-TXB₂ levels (A). 6k-PGF_1αwas increased in the AAR of HCC animals as well, but only naproxen administration decreased levels of 6k-PGF_1α, while celecoxib administration had no significant effect. *P < 0.05, ‡P < 0.001.

Immunoblotting.

Prostacyclin synthase expression in the AAR was significantly lower in the HCNS group compared with both the HCC (P < 0.05) and HCCX (P < 0.01) groups (Fig. 6A). There was no significant difference in the expression of thromboxane synthase, eNOS, VEGF, phospho-Erk 1/2, Bcl-2, or cleaved PARP between groups (Fig. 6, B-G). Immunoblotting detection of key proteins involved in cell death showed similar levels of RIP and procaspase-9 were similar among groups. Although cleaved caspase-9 was undetectable in any of the three groups, cleaved caspase-3 was undetectable only in the HCNS group. Loading control was obtained with α-tubulin (Fig. 6H).

Fig. 6.

Immunoblotting results. Immunoblotting of prostacyclin synthase (PGIS), thromboxane synthase (TS), endothelial nitric oxide synthase (eNOS), vascular endothelial growth factor (VEGF), phosphorylated-Erk 1/2 (p-Erk), B-cell lymphoma 2 (Bcl-2), and cleaved poly-ADP ribose polymerase (PARP) expression levels were performed, and expressed in terms of optical density (OD). Selected representative bands are shown (actual n = 8 per group). Additional blots were performed for RIP, caspase-9 and cleaved caspase-3 (H). Bands were normalized to either GAPDH (A–G) or α-tubulin loading controls (H). Prostacyclin synthase expression was significantly different between groups [(A) *P < 0.05] and cleaved caspase-3 was not detected in the naproxen-treated group (H). The same loading controls were used in figures 6A, F, and G and in figures 6B, C, D, and E because they came from the same membranes that were stripped and re-probed with different antibodies.

Oxidative stress.

Protein oxidative stress was significantly increased in the AAR of HCC animals compared with the NV (P < 0.001). However, administration of naproxen and celecoxib significantly reduced protein oxidation in the AAR (P < 0.01). The celecoxib group also demonstrated less protein oxidation than the naproxen group (P < 0.01; Fig. 7A).

Fig. 7.

Oxidative stress. Protein oxidation was measured using the Oxyblot kit for detection of dinitrophenylhydrazine-derived proteins. Densitometry was performed on entire lanes to measure total protein oxidation. Representative lanes are shown (A, total n = 8 per group). Protein oxidation was increased by ischemia, but reduced by naproxen treatment, and even more so by celecoxib treatment in the AAR. Lipid oxidation was measured by ELISA assay, and was significantly reduced by celecoxib treatment alone (B). *P < 0.05, †P < 0.01, ‡P < 0.001.

Lipid oxidation was significantly lower in the HCCX-AAR compared with all other groups (P < 0.05; Fig. 7B).

Vascular density.

Capillary density in the AAR was significantly increased in the AAR of HCC animals compared with the NV (P < 0.01), but decreased in the AAR of HCNS and HCCX animals compared with controls (P < 0.05 and P < 0.01, respectively; Fig. 8A). There was no significant difference in arteriolar density between groups (Fig. 8B).

Fig. 8.

Vessel density. Myocardial sections were stained for CD31 and SMA to measure capillary and arteriolar density/mm². Capillary density was increased by ischemia in HCC animals, but reduced in the AAR by both naproxen and celecoxib treatment (A). There was a strong trend towards increased arteriolar density in the AAR of HCC animals compared with other groups as well, but this did not reach statistical significance (B). *P < 0.05, †P < 0.01.

Apoptosis and intramyocardial fibrosis.

Apoptosis, as assessed by TUNEL staining, was increased in the AAR of HCC and HCNS hearts compared with HCC-NV (P < 0.05). However, apoptosis in the AAR of celecoxib-treated animals was not significantly different than in HCC-NV, and in fact was significantly lower than in the AAR of naproxen-treated animals (P < 0.01; Fig. 9A). A similar trend was seen with respect to intramyocardial fibrosis between the groups. Fibrosis was increased in both HCC-AAR and HCNS-AAR compared with the NV of HCC animals, but fibrosis in HCCX-AAR was similar to HCC-NV and significantly lower than HCNS-AAR (P < 0.05; Fig. 9B).

Fig. 9.

Apoptosis and intramyocardial fibrosis. Apoptosis was measured using TUNEL staining, and intramyocardial fibrosis measured using trichrome staining. Both apoptosis and fibrosis were increased by myocardial ischemia compared with nonischemic tissue, and were similar in the the AAR of HCNS animals and HCC animals. However, the AAR of HCCX animals demonstrated reduction in both apoptosis (A), and fibrosis (B) compared with both HCC and HCNS animals. *P < 0.05, †P < 0.01.

TUNEL assay staining frozen sections of the ischemic left ventricular myocardium revealed that the majority of TUNEL-positive cells are cardiac troponin T (cTnT) negative and sparsely distributed through the heart irrespectively of the group analyzed (Fig. 10).

Fig. 10.

TUNEL assay staining of ischemic left ventricle frozen sections. The majority of TUNEL positive cells are sparsely distributed and cardiac troponin T (cTnT) negative. TUNEL positive (red), cTnT positive (green), and nuclei counterstained with DAPI (blue). Arrows point to TUNEL positive cells.

DISCUSSION

In this study, we tested the hypothesis that COX-2 inhibition exacerbates cardiovascular dysfunction in a hypercholesterolemic model of chronic ischemia, but we found quite the opposite. Hypercholesterolemic swine treated with the COX-2 inhibitor celecoxib demonstrated improved regional function, myocardial perfusion, microvessel relaxation, and decreased oxidative stress, apoptosis, and intramyocardial fibrosis in the AAR compared with untreated animals.

We chose to use a model of chronic ischemia in this study to simulate an important comorbidity affecting a large number of chronic NSAID users. The dominant pathology in the ameroid model is myocardial hibernation, but the degree of myocardial dysfunction can be quite variable. Although the ameroid is closing, myocardial hibernation likely predominates, since there is a persistent perfusion deficit. As collateralization and angiogenesis increase, the perfusion deficit decreases and may even completely disappear, with subsequent resolution in cardiac dysfunction (23). However, the angiogenic response has been shown to be impaired in the setting of hypercholesterolemia, which may lead to more permanent flow and functional deficits (4).

Hypercholesterolemia also significantly worsens relaxation responses to both endothelium-dependent ADP and endothelium-independent SNP (4). In the current study, both naproxen and celecoxib improved endothelium-dependent vasorelaxation compared with controls, suggesting that COX inhibition, in the setting of hypercholesterolemia, in part counteracts endothelial dysfunction and improves microvascular relaxation to the nitric oxide-releasing agonist ADP. Furthermore, COX inhibition decreased the contractile response of microvessels to serotonin. We demonstrated two factors that likely contribute to this improved microvessel function.

The first is vasoactive prostanoid balance. Although both the nonselective and the selective COX-2 inhibitor improved vasodilation in this study, the COX-2 inhibitor improved myocardial perfusion to a significantly greater degree. COX regulates the production of thromboxane A₂ and prostacyclin, two vasoactive prostanoids with quite opposite functions. It was initially thought that selective COX-2 inhibition preferentially inhibits endothelial prostacyclin production, leading to unopposed vasoconstriction by thromboxane A₂ and increased cardiovascular risk. In contrast, we found that although nonselective COX inhibition decreased myocardial levels of both thromboxane and prostacyclin in the AAR, selective COX-2 inhibition decreased myocardial levels of thromboxane A₂ only. In a previous study, we showed that myocardial ischemia increased the contractile response of coronary microvessels to serotonin, but that both nonselective COX inhibition with indomethacin and selective COX-2 inhibition with NS398 markedly reduced vasoconstriction (17). This response was associated with a significant increase in gene and protein expression of COX-2, with no change in expression of COX-1. Furthermore, the contractile response was inhibited in the presence of a thromboxane synthase inhibitor (17), suggesting that the response to serotonin is in large part attributable to the actions of thromboxane, mediated by COX-2. A similar effect of COX-2 inhibition is evident in the present study. Selective COX-2 inhibition likely primarily inhibits thromboxane release, shifting the prostanoid balance in ischemic myocardium to favor prostacyclin, resulting in improved microvessel relaxation, decreased contraction, and increased perfusion. Recent studies suggest that thromboxane A₂is produced not only by platelets but also in endothelium, where COX-2 inhibition may also regulate it (10). Additional studies have reported vasodilatory effects of COX-2 inhibitors in other vascular beds as well (8), which would be unlikely if these drugs solely inhibited prostacyclin production. It is apparent that the balance between prostacyclin and thromboxane may not be as simple as previously thought.

The second is oxidative stress. Hypercholesterolemia increases oxidative stress by a variety of mechanisms. In the setting of hypercholesterolemia, eNOS has been shown to generate reactive oxygen species (ROS) instead of nitric oxide (NO) (19). Hypercholesterolemia also activates the renin-angiotensin system, leading to hypertension and the production of ROS. ROS, in turn, react with and inactivate NO (20). However, Jerez et al. (12) showed that the NSAID indomethacin improved endothelium-dependent aortic ring relaxation, while blocking angiotensin II-reactivity in hypercholesterolemic rabbits. Inhibition of the renin-angiotensin pathway by NSAID treatment likely contributes to the reduction in oxidative stress and improvement in microvessel relaxation that we observed in this study. Furthermore, both nonselective and selective COX inhibitors have been shown to decrease lipid peroxidation in mice(14), and celecoxib was found to reduce plasma lipid peroxidation in obese rats(9). Interestingly, our study found that only celecoxib decreased myocardial lipid peroxidation, and that celecoxib decreased protein oxidation to a greater degree than did naproxen. To date, no studies have directly shown increased antioxidant activity of selective COX-2 inhibitors compared with nonselective COX inhibitors; it could also be that the improvements in oxidative stress in the celecoxib group are a result of improved perfusion rather than a cause.

Zhu et al. showed in 2006 that diet-induced hypercholesterolemia increased intramyocardial fibrosis and myocyte apoptosis, ultimately leading to myocardial dysfunction in swine (28). These effects were attenuated with antioxidant supplementation. Additionally, di Lisa et al. showed that oxidation of myocardial proteins such as tropomyosin in the hearts of patients with heart failure contributes to contractile dysfunction(7). These experiments clearly establish a link between oxidative stress, myocardial fibrosis, apoptosis, and myocardial dysfunction. Thus, it is not surprising that we found decreased intramyocardial fibrosis, apoptosis, and increased regional contractility in the AAR of celecoxib-treated animals, where both protein and lipid oxidation were reduced compared with controls. However, we did not see any difference in the expression of proteins involved in apoptosis like Bcl-2, PARP cleavage, and cleaved caspase 9 as well as RIP involved in necrosis. Interestingly, caspase-3 seems to be consistently inactive only in the naproxen group. It may be that the majority of molecular mechanisms leading up to myocyte apoptosis occurred some time prior to the end of our experiment, seven wk after the initiation of ischemia. Cleavage (activation) of caspase-3 has also been associated with functions distinct from apoptosis. Regardless, the beneficial effects of celecoxib on ischemic myocardium that we demonstrate in this study are unique and significant.

COX-inhibition and angiogenesis.

We measured capillary and arteriolar density in the AAR to determine whether angiogenesis contributed to the improved myocardial perfusion, and surprisingly found significantly fewer capillaries in both naproxen- and celecoxib-treated groups. Nonselective NSAIDs have been shown to inhibit Erk 1/2 activation, an important step involved in cell proliferation and angiogenesis(18), while COX-2 inhibitors are known to have antiangiogenic effects in tumor models(15). In our study, neither naproxen nor celecoxib affected myocardial expression of pro-angiogenic proteins Erk 1/2, VEGF, or eNOS, yet capillary density was decreased in both drug-treated groups. Furthermore, arteriolar density remained similar between groups, ruling out enhanced arteriolar maturation as an explanation for increased flow. It may simply be an improvement in the function of arterioles, rather than an increase in their number, that improves myocardial perfusion to the ischemic territory and decreases the impetus for angiogenesis. However, we currently do not have a good explanation for these findings other than that the products of COX are involved with the development of new vessels.

Effects of hypercholesterolemia and ischemia.

Similar experiments by our lab on nonischemic myocardium and normo-cholesterolemic animals did not demonstrate such pronounced improvements in microvascular function with COX inhibition(21, 22). The combination of hypercholesterolemia and chronic ischemia seems to create an environment that is particularly susceptible to the beneficial effects of COX-2 inhibition. Under conditions of inflammation, prostacyclin production decreases, and thromboxane production predominates (26). Thus, COX inhibition in the inflammatory setting of hypercholesterolemia and chronic ischemia may preferentially inhibit thromboxane production, allowing vasodilatory mechanisms to prevail. This theory is supported by a recent study of a rat model of cirrhosis, where cirrhotic rats demonstrated mesenteric vessel vasodilation in response to both nonselective and selective COX-2 inhibition while noncirrhotic rats did not (27).

Limitations.

Though the differences in MAP, HR, and RPP between groups were not statistically significant, there was a trend towards decreases in all three of these parameters in the drug treated groups in this study, and one should keep in mind that measures of contractility are sensitive to differences in afterload. It was also somewhat unexpected that there was not a more significant increase in myocardial perfusion under ventricular pacing in both the NV and AAR. It may be that ventricular pacing did not adequately increase metabolic demand, or decreased perfusion pressure below the threshold at which autoregulation can take place; unfortunately we did not take hemodynamic or functional measurements during pacing. Additionally, though it is known that COX-2 is induced in inflammatory states including ischemia and hypercholesterolemia, we were unable to quantify COX-1 and COX-2 expression to prove that this is the case in swine as well. Finally, we examined the effects of COX-inhibition on the vasoactive components of the COX pathway, but did not investigate the thrombotic properties of the pathway. Increased platelet aggregation and thrombosis afforded by selective COX-2 inhibition may still contribute to adverse cardiac events, regardless of improvements in cardiac function.

Conclusions

We investigated the effects of COX inhibition in a model of hypercholesterolemia that is highly relevant to patients with cardiovascular disease, and found that both nonselective and selective COX-2 inhibition improved microvascular reactivity and reduced oxidative stress in the AAR. Selective COX-2 inhibition significantly improved myocardial perfusion and regional function while reducing apoptosis and intramyocardial fibrosis. These findings should not be taken to mean that COX-2 inhibition is absolutely protective against adverse cardiac events, as it may still increase platelet aggregation and thrombus formation, and may have as yet undiscovered effects on cardiovascular function. Nonetheless, the selective COX-2 inhibitor celecoxib clearly has a significant effect in reducing endothelial dysfunction, which may be beneficial in patients with hypercholesterolemia and chronic ischemia.

GRANTS

This project was funded by grants from the National Heart, Lung, and Blood Institute (RO1HL46716, RO1HL69024, and RO1HL85647, Dr. Sellke), NIH training grant 5T32HL094300 (Dr. Chu), the Irving Bard Memorial Fellowship (Drs. Chu and Robich), and NIH training grant 5T32HL007734 (Dr. Robich).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s). Dr. Frank W. Sellke was a consultant Medicines, Regado Pharma, and CSL Behring. Dr. Sellke was a consultant for the law firms representing Pfizer (Princeton, NJ) in the Bextra/Celebrex litigation, but no funding was received for this study, and there was no consultation or notification regarding this study.

AUTHOR CONTRIBUTIONS

Author contributions: L.M.C., M.P.R., and F.W.S. conception and design of research; L.M.C., M.P.R., C.B., J.F., Y.L., S.-H.X., and T.A.B. performed experiments; L.M.C., M.P.R., C.B., J.F., Y.L., S.-H.X., T.A.B., and F.W.S. analyzed data; L.M.C., M.P.R., C.B., J.F., Y.L., S.-H.X., T.A.B., and F.W.S. interpreted results of experiments; L.M.C., C.B., J.F., Y.L., S.-H.X., and T.A.B. prepared figures; L.M.C., Y.L., and S.-H.X. drafted manuscript; L.M.C., M.P.R., C.B., J.F., T.A.B., and F.W.S. edited and revised manuscript; L.M.C., M.P.R., C.B., J.F., Y.L., S.-H.X., T.A.B., and F.W.S. approved final version of manuscript.