By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
William F. Stenson, Division of Gastroenterology, Box 8124, Washington University School of Medicine, 660 South Euclid Ave, St Louis, MO 63110, USA. Tel.: +1 314 362 8952; fax: +1 314 362 9035. E-mail: firstname.lastname@example.org
Summary. Background: We found a high incidence of thrombotic deaths in COX-1+/−COX-2−/− mice and sought to define the mechanism of these events. The cyclooxygenase products thromboxane A2 and prostacyclin are important in the regulation of coagulation but their role in fibrinolysis is largely unexplored. PAI-1 blocks fibrinolysis by inhibiting plasminogen activator. Aim: Our objective was to explain the mechanism of increased thrombosis associated with the COX-1+/−COX-2−/− genotype. Methods: Carotid artery occlusion times were measured after photochemical injury. PAI-1 levels were measured in the plasma by ELISA. PAI-1 levels in the aorta were measured by RT-PCR and Western blotting. Urinary metabolites of Thromboxane A2 and prostacyclin were measured by ELISA. Results: The COX-1+/−COX-2−/− genotype is associated with a decreased time to occlusion in the carotid artery thrombosis model (30 ± 5 minutes vs 60 ± minutes in wild type, p<.001). The COX-1−/−COX-2+/+, COX-1+/−COX-2+/− and COX-1+/− COX-2+/+ all had occlusion times similar to wild type. COX-1+/+ COX-2−/− had a prolonged occlusion time. COX-1+/− COX-2−/− had increased PAI-1 levels in the plasma and aorta and with a prolonged euglobulin lysis time (37.4 ± 10.2 hours vs 15.6 ± 9.8 hours in wild type, p<.004). The decreased time to occlusion in the COX-1+/−COX2−/− mice was normalized by an inhibitory antibody to PAI-1 whereas the antibody had no effect on the time to occlusion in wild type mice. Conclusion: The COX-1+/−COX-2−/− genotype is associated with a shortened time to occlusion in the carotid thrombosis model and the shortened time to occlusion is mediated through increased PAI-1 levels resulting in decreased fibrinolysis.
We attempted to generate cyclooxygenase-1 and cyclooxygenase-2-deficient (COX-1−/−COX-2−/−) mice on a C57BL/6J background by crossing male COX-1+/−COX-2−/− with female COX-1+/−COX-2+/− and female COX-1−/−COX-2+/− mice. This approach did not yield viable COX-1−/−COX-2−/− mice but we noted that many of the COX-1+/− COX-2−/− mice died by 2–3 months of age. Live COX-1+/−COX-2−/− mice submitted for necropsy were found to have thrombotic events including clots in the heart, liver and lung. We sought to define the mechanism of the thrombotic events associated with this genotype. Cyclooxygenase products have been closely tied to coagulation. Thromboxane A2 (TxA2) promotes thrombosis through platelet aggregation and vasoconstriction [1,2]. In contrast, prostacyclin (PGI2) inhibits thrombosis through vasodilation and inhibition of platelet aggregation [3–5]. Selective COX-2 inhibitors and NSAIDs, which inhibit COX-1 and COX-2, are associated with an increased incidence of myocardial infarction and other thrombotic events [6,7]. Meta-analyses show an increased risk of cardiovascular events with one selective COX-2 inhibitor, rofecoxib, and perhaps with another, celecoxib [8,9]. There appears to be an increased risk of cardiovascular events with some NSAIDs, including diclofenac, but not with naprosyn . The mechanism of the increased incidence of cardiovascular events is not clear but appears to relate to decreased production of PGI2 .
Myocardial infarctions and other thrombotic events can be the result of either enhanced coagulation or impaired clot resolution. The contribution of COX-1 and COX-2 products to thrombosis is complex. TxA2, which promotes platelet aggregation, is produced in platelets through COX-1 . PGI2, which is produced in blood vessels in both endothelial cells and vascular smooth muscle cells, is a product of prostacyclin synthase (PGI synthase), which can be downstream of either COX-1 or COX-2 . At baseline COX-2 expression in vascular tissue is low and most vascular PGI2 is produced through COX-1 [14,15]. Some investigators have been unable to identify COX-2 in vascular tissue in the absence of injury ; however, others find that in response to shear forces associated with blood flow COX-2 is induced in vascular tissues and more vascular PGI2 is produced through COX-2 [16,17]. These findings led to the suggestion that the increase in myocardial infarctions and other thrombotic events seen in patients taking selective COX-2 inhibitors is the product of the inhibition of vascular PGI2 in the absence of any inhibition of platelet thromboxane production .
Although the role of cyclooxygenase products in coagulation has been extensively investigated, their role in clot resolution or fibrinolysis is largely unexplored. Incubation of stable PGI2 analogs with vascular smooth muscle cells in culture results in a decreased expression of plasminogen activator inhibitor 1 (PAI-1), a protein that decreases fibrinolysis by inhibiting both tissue-type plasminogen activator and urokinase-type plasminogen activator and thus plasmin formation [18,19]. Treatment of diabetic patients with iloprost, a stable PGI2 analog, results in decreased plasma PAI-1 activity . Mice deficient in PAI-1 have increased fibrinolysis and a prolonged time to thrombosis in a carotid artery injury model . There is experimental and epidemiological evidence that PAI-1 contributes to the development of cardiovascular disease [22,23]. PAI-1 excess causes spontaneous coronary artery occlusion in mice . ApoE−/− mice fed a high fat diet develop severe hypercholesterolemia and have elevated plasma PAI-1 levels . These mice exhibit a prothrombotic phenotype with shortened times to thrombotic occlusion in carotid artery injury models. The deletion of the PAI-1 gene in these mice reverses the prothrombotic tendency. Vitronectin (VN), an abundant plasma and matrix glycoprotein, binds PAI-1 and regulates its activity by stabilizing the active PAI-1 conformation . Deletion of either PAI-1 or VN in wild-type mice delays thrombotic occlusion following arterial injury . Complete deficiency of PAI-1 in man results in a mild bleeding disorder .
In this study we explored the relationship between PAI-1 and cyclooxygenase genotypes using euglobulin lysis times and a carotid artery occlusion model in the mouse. We found that the COX-1+/− COX-2−/− genotype was associated with increased plasma and aortic levels of PAI-1 and with both a prolonged euglobulin lysis time and a shortened time to occlusion in the carotid artery injury model. Moreover, we found that the shortened time to occlusion was normalized by a neutralizing antibody to PAI-1.
Materials and methods
Except where noted, mice are on a C57BL/6J background. Knockout mice on a C57BL/6J background were COX-1−/−COX-2+/+, COX-1+/−COX-2+/+, COX-1−/−COX-2+/−, COX-1+/+COX-2−/−, and COX-1+/−COX-2−/− where the COX-2 gene was disrupted in exon 8 . COX-1+/+COX-2+/− mice on the B6;129 background were a gift from Garret FitzGerald, University of Pennsylvania. In these mice the COX-2 gene is disrupted in exon 1 . We bred them to produce wild-type and COX-2−/− mice. Mice were maintained on a 12-h light/dark cycle in a pathogen-free barrier facility and fed standard laboratory chow ad libitum. All experiments using mice were approved by the Washington University School of Medicine Animal Studies Committee.
Clot lysis assay
A euglobulin clot lysis assay for global fibrinolysis was performed according to a protocol modified from Smith et al. . See Supporting Information Data S1 for detailed protocol.
Arterial thrombosis model
Carotid artery thrombosis was induced as described previously . Male and female wild-type and knockout C57BL/6J and B6;129 mice (6–9 weeks of age; 16–19 g body weight) were anesthetized with intraperitoneal sodium pentobarbital, secured in a supine position, and placed under a dissecting microscope. The right carotid artery was isolated through a midline cervical incision, and an ultrasonic flow probe (Model 0.5 VB; Transonic Systems, Ithaca, NY, USA) was applied. A 1.5 mW, 540 nm laser beam (Melles Griot, Carlsbad, CA, USA) was applied to the artery from a distance of 6 cm. Rose Bengal dye (3′, 4′, 5′, 6′– tetrachloro-2, 4, 5, 7- tetraiodofluorescein) (Fisher Scientific, Fair Lawn, NJ, USA) at a dose of 50 mg kg−1 body weight was then injected into the lateral tail vein, and blood flow was monitored continuously. The occlusion time was taken as the interval between the injection of Rose Bengal dye and complete and stable (> 5 min) cessation of flow.
In a subset of experiments, mice were given a single intraperitoneal injection of NS-398 (Biomol, Plymouth Meeting, PA, USA) at a dose of 1 mg kg−1 body weight 2 h before injecting Rose Bengal dye. NS-398 was reconstituted in dimethylsulfoxide (DMSO) as a concentrated stock solution at 25 mg mL−1. The working solution was made by diluting an aliquot 125-fold (8 μl mL−1) in 5% sodium bicarbonate. Control mice received vehicle. In another subset of experiments, mice were given a single tail vein injection of a mouse monoclonal inhibitory antibody, H4B3, against mouse PAI-1 (Molecular Innovations, Novi, MI, USA)  at a dose of 0.5 mg kg−1 body weight 15 min before injecting Rose Bengal dye.
Total RNA was extracted from aorta tissues by homogenization in RiboZol RNA extraction reagent (Amresco, Solon, OH, USA). qRT-PCR was used to measure gene expression, with actin expression as a control. The final results were expressed as fold differences in gene expression relative to the actin gene. The Ct for each amplicon was determined as the PCR cycle at which the fluorescence intensity crossed a user-established threshold. The following primers were synthesized by Integrated DNA Technologies (Coralville, IA, USA):
COX-1 (5′-AGGAGATGGCTGCTGAGTTGG and 5′-AATCTGACTTTCTGAGTTGCC), COX2 (5′-ACACACTCTATCACTGGCACC and 5′-TTCAGGGCGAAGCGTTTGC), PAI-1 (5′-TCAGAGCAACAAGTTCAACTACACTGAG and 5′CCCACTGTCAAGGCTCCATCACTTGCCCCA), vitronectin (5′-CATCACGTTCAATCTCGTTCTCTTT and 5′-TTGGGCTGGGAGAAAGAGATGAGG) and actin (5′-CAACGAGCGGTTCGGATG and 5′-GCCACAGGATTCCATACC). Commercially available primers for PGI synthase (PPM-05311A) were purchased from SA Biosciences (Frederick, MD, USA).
SDS-PAGE and western blot analysis
Aortas from C57BL/6J wild-type, COX-1+/+COX2−/− and COX1+/−COX2−/− mice were homogenized in ice cold proteinase inhibitor cocktail containing antipain (25 μg mL−1), aprotinin (25 μg mL−1), leupeptin (25 μg mL−1), chymostatin (25 μg mL−1), phenanthroline (50 μmol L−1), phenylmethylsufonyl fluoride (100 μmol L−1), pepstatin A (10 μg mL−1) and dithiothreitol (2 nmol L−1) in N-Tris (hydroxymethyl)-methyl-2-aminomethane sulfonic acid (20 mmol L−1), pH 7.4. Equal amounts (40 μg) of protein from all mice were separated by SDS-PAGE on a 10% polyacrylamide gel and transferred to an Immobilon membrane (Millipore, Bedford, MA, USA). Antibodies used for detecting proteins were goat anti-mouse COX-1 (SC-1754, 1:200; Santa Cruz Biotechnology, Santa Cruz, CA, USA), mouse monoclonal anti-COX-2 (610204, 1:500; BD Biosciences, San Jose, CA, USA), rabbit anti-human PGI synthase (SC-20933, 1:200; Santa Cruz) and sheep anti-human PAI-1 (SAPAI-IG, 1:500; Affinity Biologicals, Ontario, Canada). Visualizing of protein bands was carried out with HRP-linked secondary antibodies and ECL reagent (GE Healthcare Amersham, Piscataway, NJ, USA) with fluorographic detection on Blue-sensitive autoradiography film (Mid Sci, St Louis, MO, USA). Determination of protein loading was by reprobing blots with goat anti-human actin (SC-1616, 1:200; Santa Cruz).
Blood was obtained from mice by puncturing the right mandibular vein with a 5.5-mm animal lancet and collecting the blood directly into plasma separator tubes kept on ice and centrifuged at 10 000 × g for 5 min within 10 min of collection. Plasma was transferred to cryovials and snap frozen in liquid nitrogen and stored at −80 °C until the time of assay. ELISA kits from Molecular Innovations were used to measure plasma levels of active PAI-1 (MPAI KT) and total PAI-1 (MPAIKT-TOT) in wild-type, COX-1+/+COX-2−/− and COX-1+/−COX-2−/− mice.
Tail vein bleeding time
Tail vein bleeding time was determined as described by Broze et al. . Mice were anesthetized (ketamine 75mg kg−1, medetomidine 1 mg kg−1 intraperitoneally) and placed prone on a warming pad. A transverse incision was made with a #11 scalpel blade over a lateral vein at a position where the diameter of the tail was 2.25 to 2.5 mm. The tail was immersed in isotonic normal saline (37 °C) in a hand-held test tube. The time from incision to the cessation of bleeding was recorded as the bleeding time.
Analysis of urine for metabolites of PGI2 and TxA2
Urine was collected from mice by lining the bottoms of clean mouse cages with clean sheets of plastic wrap and placing a single mouse in each cage . Once a mouse urinated, the voided urine was transferred to a micro centrifuge tube and snap frozen in liquid nitrogen. Urine samples were stored at −80 °C. Urine collected from wild-type and knockout C57BL/6J mice was assayed for the prostacyclin metabolite 2,3-dinor-6-keto PGF1α and the thromboxane A2 metabolite 11-dehydro-thromboxane B2 by ELISA (Cayman Chemical, Ann Arbor, MI, USA). Urinary metabolites were normalized to urine creatinine. Urinary creatinine was measured by ELISA (Cayman Chemical).
Breeding schemes to generate COX genotypes
COX-1−/− (gene disrupted in exon 11) and COX-2−/− (gene disrupted in exon 8) mice on a mixed B6;129 genetic background were backcrossed 10 generations to wild-type mice on the C57BL/6J genetic background to yield COX-1−/− and COX-2−/− mice on a C57BL/6J genetic background. These mice were bred as in Table S1. Because COX-1−/− females have delayed parturition, COX-1−/− mice were generated from COX-1+/− females. The percentage of COX-1−/− mice generated was 33% from COX-1+/− × COX-1+/− matings and 45% from male COX-1−/− × female COX-1+/− matings. For generating COX-2−/− mice, females expressing one allele of COX-2 were used, because female COX-2−/− mice are infertile . The percentage of COX-2−/− mice generated was less than Mendelian expectations, with 6% from COX-2+/−× COX-2+/− matings and 23% from male COX-2−/− × female COX-2+/− matings. For the purpose of generating COX-1+/−COX-2+/−, COX-1−/−COX-2+/− and COX-1+/−COX-2−/− mice, female COX-1+/−COX-2+/− breeders were used. In contrast to female COX-1+/−COX-2+/− mice on the mixed B6;129 genetic background, which have delayed parturition , female COX-2+/−COX-2+/− mice on the C57B6/6J genetic background have a normal gestation period and consistently deliver live pups. Male COX-1+/+COX-2−/− × female COX-1+/−COX-2+/− matings were used to produce COX-1+/−COX-2−/− mice. Male COX-1+/−COX-2−/−× female COX-1+/−COX-2+/− matings were also used and these also generated COX-1−/−COX-2−/− mice, which died within 24 h from patent ductus arteriosus (PDA) . Male COX-1−/−COX-2+/+ × female COX-1+/−COX-2+/− matings were used to generate COX-1−/−COX-2+/− mice.
COX-1+/−COX-2−/− genotype is associated with premature death and spontaneous thrombosis
COX-1+/−COX-2−/− mice died prematurely, with 50% survival at 3 months and 10% survival at 1 year (Fig. 1A). All the other genotypes generated (COX-1−/−, COX-2−/−, COX-1+/−COX-2+/−, COX-1−/−COX-2+/−) had survival curves similar to wild-type mice. Neonate survival for the various COX-1 and COX-2 knockout genotypes on the C57BL/6J genetic background was greater than in mice on the mixed B6;129 genetic background . The occurrence of PDA on the B6;129 background has been reported to be 35% in COX-2−/−, 79% in COX-1+/−COX-2−/− and 100% in COX-1−/−COX-2−/− mice . In our mice on the C57BL/6J genetic background there was still 100% occurrence of deaths within 24 h of birth in COX-1−/−COX-2−/−, apparently due to PDA, but there was no indication from neonate survival of PDA in the COX-2−/− or in COX-1+/−COX-2−/− mice on the C57BL/6J genetic background.
At about 1–2 months of age the COX-1+/−COX-2−/− mice on the C57BL/6J genetic background began to die. Some died suddenly whereas others first manifested signs of disease, including decreased activity and hunched posture. Some of these animals were sacrificed and necropsied. The necropsies demonstrated spontaneous thromboses in a variety of anatomic sites, including the superior vena cava (Fig. 1B), tricuspid valve (Fig. 1C), pulmonary artery (Fig. 1D) and central vein in the liver (Fig. 1E). The liver thrombosis is associated with an area of hepatic infarction. The liver and pulmonary artery thromboses have considerable fibrosis, suggesting that they are at least several days old.
COX-1+/− COX-2−/− genotype is associated with a shortened time to occlusion in the Rose Bengal carotid artery injury model
To address the mechanism of the thrombotic deaths in the COX-1+/−COX-2−/− mice we assessed the time to occlusion in a carotid artery injury model. Wild-type C57BL/6J mice had a 63-min time to occlusion in the carotid injury model (Fig. 2A). COX-1−/− COX-2+/+, COX-1−/− COX-2+/− and COX-1+/− COX-2+/+ all had times to occlusion similar to that seen with wild-type mice. COX-1+/+ COX-2−/− mice had a markedly prolonged time to occlusion at 100 min. COX-1+/− COX-2−/− mice had a markedly shortened carotid occlusion time at 30 min.
The administration of a selective but not specific COX-2 inhibitor, NS-398, to wild-type C57BL/6J mice had no effect on the time to occlusion in the carotid injury model. In contrast, administration of NS-398 to the COX-1−/− COX-2+/−, COX-1+/− COX-2+/+ or COX-1+/+ COX2−/− mice resulted in a shortening of the time to occlusion.
Our studies on the carotid injury model used mice on the C57BL/6J genetic background. Other investigators had used mice on a mixed B6;129 background to study the effects of cyclooxygenase products on thrombosis . We therefore repeated some of the carotid artery injury studies using B6;129 mice. Wild-type mice on the B6;129 genetic background had a shorter time to occlusion (41 min) than did the wild-type on the C57BL/6J genetic background (63 min) (Fig. 2B and A). In contrast to the C57BL/6J mice, the administration of NS-398 to wild-type B6;129 mice resulted in a marked diminution in the time to occlusion, from 41 to 20 min. In COX-1+/+ COX-2−/− mice on the B6;129 genetic background the time to occlusion was markedly prolonged, just as it was in the COX-1+/+ COX-2−/− mice on the C57BL/6J genetic background. Similarly, administration of NS-398 resulted in a marked shortening of the time to occlusion in the COX-1+/+ COX-2−/− mice on the B6;129 genetic background, just as it had in the mice with the similar genotype on the C57BL/6J genetic background.
COX1+/−COX2−/− genotype is associated with decreased urinary prostacyclin metabolites
The cyclooxygenase products most frequently associated with thrombosis are TxA2, which promotes thrombosis through vasoconstriction and platelet aggregation, and PGI2, which impairs thrombosis through vasodilation and inhibition of platelet aggregation. Platelet TxA2 is produced only by COX-1 whereas TxA2 synthesized in other cell types can be the product of either COX-1 or COX-2. Vascular PGI2 can be produced by either COX-1 or COX-2. We next sought to correlate the COX genotypes with urinary levels of metabolites of TxA2 and PGI2 (Fig. 3). COX-1−/−, COX-1−/−COX-2+/−, COX-2−/− and COX-1+/−COX-2−/− mice all had lower levels of 2,3-dinor -6-keto PGF1α, a metabolite of PGI2, when compared with wild-type mice (Fig. 3A). Although all four of these genotypes had diminished levels of the PGI2 metabolite, the lowest levels were in the COX-1+/−COX-2−/− mice. Urinary levels of the TxA2 metabolite, 11-dehydro-thromboxane B2, were also diminished in the COX-1−/−, COX-1−/−COX-2+/− and COX-2−/− mice but not in COX-1+/−COX-2−/− mice. (Fig. 3B). Although the levels of the thromboxane metabolite were diminished in several genotypes, the lowest levels were in the mice completely deficient in COX-1. Other investigators demonstrated an association of thrombotic events with the ratio of PGI2 urinary metabolites to TxA2 urinary metabolites; the higher the ratio the less likely a thrombotic event [37,38]. COX-1−/−, COX-1−/−COX-2+/− and COX-2−/− mice all had ratios of 0.6–0.7 compared with the ratio of 1.2 seen in wild-type mice (Fig. 3C). In contrast, COX-1+/−COX-2−/− mice had a much lower ratio of 0.25. Thus the COX-1+/−COX-2−/− mice had a PGI2/TxA2 metabolite ratio associated with an increased risk of thrombosis.
COX-1+/− COX-2−/− genotype is associated with increased aortic levels of VN and PAI-1
One determinant of the time to thrombosis in the carotid artery injury model is the rate of fibrinolysis. PAI-1 blocks fibrinolysis by inhibiting plasminogen activator and VN stabilizes the active form of PAI-1 [18,19]. Deletion of PAI-1 or VN in wild-type mice delays thrombotic occlusion in arterial injury . We next sought to determine if the shortened time to occlusion in the carotid artery seen with COX-1+/−COX-2−/− mice was associated with an increase in PAI-1. Total mRNA was isolated from the aortas of wild-type, COX-1+/+ COX-2−/− and COX-1+/− COX-2−/− mice, all on the C57BL/6J genetic background. Levels of VN and PAI-1 mRNA were markedly increased in the aortas of COX-1+/+ COX-2−/− mice and were further increased in the aortas of COX-1+/− COX-2−/− mice (Fig. 4A). The levels of PGI synthase mRNA were reduced in the COX-1+/+ COX-2−/− mice and further reduced in the COX-1+/− COX-2−/− mice. We also extracted protein from the aortas of wild-type, COX-1+/+ COX-2−/− and COX-1+/− COX-2−/− mice. Western blots for PGI synthase, and PAI-1, confirmed the results of the qRT-PCR data (Fig. 4B).
There was an 8-fold increase in PAI-1 in the COX-1+/− COX-2−/− aortas compared with wild-type mice. There was also a marked decrease in the expression of PGI synthase in the COX-1+/− COX-2−/− mouse aortas. COX-1 levels were diminished in the COX-1+/− COX-2−/− mice compared with wild-type mice. As expected, COX-2 was not detectable in either the COX-1+/− COX-2−/− or COX-1+/+ COX-2−/− mice compared with wild-type mice.
Plasma levels of PAI-1 are significantly higher in COX-1+/− COX-2−/− mice compared with wild-type or COX-1+/+ COX-2−/− mice
Having found increased levels of PAI-1 in the aortas of COX-1+/−COX-2−/− mice we next sought to determine if there were increased plasma levels of PAI-1 in these mice. We measured both total PAI-1 and active PAI-1 levels in the plasma of wild-type, COX-1+/+ COX-2−/− and COX-1+/− COX-2−/− mice. There was a 5-fold increase in both total PAI-1 and active PAI-1 in the plasma of COX-1+/− COX-2−/− mice compared with the levels in the plasma of either wild-type or COX-1+/+ COX-2−/− mice (Fig. 5).
The shortened occlusion time seen in the COX-1+/− COX-2−/− mice was normalized by the administration of an inhibitory antibody to PAI-1
Wild-type and COX-1+/− COX-2−/− mice were given an inhibitory antibody to PAI-1 prior to the administration of Rose Bengal dye in the carotid artery injury model. Administration of the antibody to PAI-1 had no effect on the occlusion time in the wild-type mice but reversed the shortening of the time to occlusion seen in the COX-1+/− COX-2−/− mice (Fig. 6).
COX-1+/−COX-2−/− genotype is associated with impaired fibrinolysis
We performed euglobulin lysis times using platelet-poor plasma (Fig. 7). The euglobulin lysis time in the COX-1+/−COX-2−/− mice was significantly prolonged in comparison with wild-type mice. Addition of an inhibitory antibody to mouse PAI-1 to the wild-type plasma had no effect on the euglobulin lysis time, whereas addition of an inhibitory antibody to mouse PAI-1 to COX-1+/−COX-2−/− plasma resulted in a trend towards a shorter euglobulin lysis time although this trend did not reach statistical significance. Addition of exogenous mouse PAI-1 to the wild-type serum resulted in a marked prolongation of the lysis time such that the optical density never fell below 0.50 (data not shown).
Tail vein bleeding time in COX-1+/− COX-2−/− mice is the same as that in wild-type mice
The tail vein bleeding time in the COX-1+/− COX-2−/− mice was the same as in the wild-type mice (Fig. 8). The normal tail vein bleeding time in the COX-1+/−COX-2−/− mice suggests that there is no defect in platelet function. As expected, the tail vein bleeding time in COX-1−/− COX-2+/+ mice was markedly prolonged.
In attempting to explain the premature deaths and spontaneous thromboses we saw in COX-1+/−COX-2−/− mice we found that this genotype is associated with a shortened time to occlusion in the carotid artery injury model, with increased levels of aortic and plasma PAI-1 and with a prolonged euglobulin lysis time. Moreover the shortened time to occlusion seen in the COX-1+/− COX-2−/− genotype was normalized by an inhibitory antibody to mouse PAI-1. This suggests that the increased risk of thrombosis seen in this genotype is the result of decreased fibrinolysis mediated by an increase in PAI-1.
The association of elevated levels of PAI-1 with spontaneous thrombosis and a prolonged euglobulin lysis time is consistent with previous studies. Mice deficient in PAI-1 have increased fibrinolysis and a prolonged time to thrombosis in a carotid artery injury model  and PAI-1 excess is associated with spontaneous coronary artery occlusions. . The spontaneous thromboses seen in the PAI-1 overexpressing mice is consistent with the spontaneous thromboses seen in the COX-1+/−COX-2−/− mice. The COX-1+/− COX-2−/− genotype was also associated with an increase in VN, which stabilizes the active form of PAI-1. There is some evidence for regulation of VN expression by cyclooxygenase products .
We attempted to determine if specific cyclooxygenase products were involved in mediating the decreased time of occlusion or the increased PAI-1 expression seen in COX-1+/− COX-2−/− mice by measuring urinary metabolites of PGI2 and TxA2. The analysis of the urinary metabolites demonstrated that both PGI2 and TxA2can be produced through either COX-1 or COX-2 although COX-1 is the major pathway for TxA2 production. The urinary metabolite studies also demonstrated that the COX-1+/− COX-2−/− genotype was associated with the lowest levels of PGI2 metabolites and the lowest ratio of PGI2 metabolites to TxA2 metabolites. There are several other lines of evidence that support the suggestion that decreased PGI2 production mediates the elevated PAI-1 levels and the shortened time to occlusion associated with the COX-1+/− COX-2−/− genotype. Stable analogs of PGI2 decrease PAI-1 expression in vascular smooth muscle cells [18,19]. Moreover, the administration of iloprost, a stable analog of PGI2, to diabetic patients resulted in diminished PAI-1 activity . Administration of PGI2 to patients undergoing hemodialysis resulted in effective anticoagulation with laboratory evidence for increased fibrinolysis although PAI-1 levels were not measured . Therefore, a genotype associated with low levels of PGI2 production might be expected to be associated with increased PAI-1 and decreased fibrinolysis.
The COX-1+/− COX-2−/− genotype would be expected to result in decreased vascular PGI2 production because vascular PGI2 is produced through both COX-1 and COX-2. This is supported by the finding that, of all the genotypes tested, the COX-1+/− COX-2−/− genotype was associated with the lowest level of urinary PGI2 metabolites. Finally, the COX-1+/− COX-2−/− genotype is associated with decreased aortic PGI synthase as assessed both by qRT-PCR and Western blot.
How do these data fit with earlier studies of the role of cyclooxygenase products in thrombosis? No previous studies have specifically addressed the effects of cyclooxygenase genotype on PAI-1 levels or fibrinolysis in vivo; however, there have been studies of the role of cyclooxygenase products in the carotid injury model. Cheng et al. addressed the role of cyclooxygenases in the Rose Bengal carotid artery injury model used here. Mice deficient in the prostacyclin receptor (IP−/−) had a markedly shortened time to occlusion compared with wild-type mice. The time to occlusion in IP+/− mice was intermediate between that seen in wild-type and IP−/− mice. These data support the suggestion that decreased PGI2 could lead to increased thrombosis but do not directly address the mechanism. The same investigators found that administration of DFU, a selective COX-2 inhibitor, to wild-type mice resulted in a shortening of the time of occlusion. This study was carried out using B6;129 mice and is consistent with our finding that NS-398 shortened the time to occlusion in wild-type mice on the same genetic background. These investigators also developed a mouse (PGHS-1KD) in which COX-1 activity was knocked down by 95%. They found that this genotype is associated with a prolonged time to occlusion. Administration of DFU to these mice shortened the time to occlusion. The 95% knockdown of COX-1 would be expected to decrease platelet thromboxane production and thus lead to a prolonged time to occlusion. In contrast, a 50% decrease in COX-1 activity, as is seen in COX-1+/− mice, has no effect on platelet aggregation as demonstrated by the normal bleeding time.
One remarkable aspect of the results with the carotid artery injury model was the observation that although COX-1+/− COX-2−/− mice had a shortened time to occlusion, neither COX-1+/− COX-2+/+ nor COX-1+/+COX-2−/− mice had a shortened time to occlusion. This suggests that a shortened time to occlusion requires complete, or near complete, loss of COX-2 activity and partial loss of COX-1 activity.
Why is the time to occlusion shorter in the B6;129 wild-type mice than in the C57BL/6J wild-type mice? Different mouse strains have different baseline values in the carotid artery thrombosis assay . Potential explanations for the differences in time to occlusion between the two strains include events that affect coagulation and events that affect fibrinolysis; these events could include COX- related events.
Why is the time of occlusion prolonged in COX-1+/+ COX-2−/− mice both on the C57BL/6J genetic background and on the B6;129 genetic background? We would have predicted that COX-1+/+COX-2−/− mice would have a time to occlusion similar to or shorter than wild-type mice. We were surprised by the prolonged time to occlusion in the COX-1+/+ COX-2−/− mice on the C57BL/6J genetic background, where the COX-2 gene is disrupted in exon 8, and thought this might reflect some non-COX-2-related property in these mice. To confirm this observation we repeated the study in COX-1+/+COX-2−/− mice on the B6;129 genetic background, where the COX-2 gene is disrupted in exon 1; we obtained similar results. Thus the time to occlusion is prolonged in COX-1+/+COX-2−/− mice on different genetic backgrounds with the COX-2 genes disrupted in different exons. The prolonged time to occlusion in the COX-1+/+COX-2−/− mice is not associated with decreased PAI-1 levels and, in fact, is associated with PAI-1 levels that are increased compared with wild-type mice. Thus the prolonged time to occlusion does not appear to be due to increased fibrinolysis resulting from decreased PAI-1.
The effects of NS-398 on the time to occlusion in the different genotypes fits with the model that a shortened time to occlusion requires a complete, or almost complete, loss of COX-2 activity with partial loss of COX-1 activity. As expected administration of NS-398 resulted in a shortening of the time to occlusion in COX-1−/−COX2+/− and COX-1+/−COX-2+/+ mice. The response of wild-type mice to NS-398 varied with the background, with NS-398 having no effect on the time to occlusion in C57BL/6J mice but resulting in a shorter time to occlusion in the B6;129 mice. This may reflect the relative contribution of COX-1 and COX-2 to PGI2 synthesis in the two genetic backgrounds. Other investigators using a different selective COX-2 inhibitor in B6;129 mice also found that inhibition of COX-2 results in shortening of the time to occlusion in this model . The prolonged time to occlusion in the COX-1+/+COX2−/− mice is shortened by treatment with NS-398. This effect is seen in COX-1+/+COX-2−/− mice on both backgrounds. NS-398 is not completely specific for COX-2, and it is possible that the shortening of the occlusion time is due to the partial inhibition of COX-1 by NS-398.
Previous studies have demonstrated that cyclooxygenase products have profound effects on thrombosis[1–5]. Here we have demonstrated that cyclooxygenase genotypes influence PAI-1 levels and thus occlusion times in carotid injury. The relative importance of these effects in clinical thrombosis is not clear; however, in the carotid artery injury model the short time to occlusion associated with the COX-1+/− COX-2−/− genotype was completely reversed by an inhibitory antibody against mouse PAI-1. In contrast, the antibody to PAI-1 had no effect on the time to occlusion in wild-type mice. These findings suggest that PAI-1 levels do not influence the time to occlusion in wild-type mice but that increased PAI-1 levels were responsible for the shortened time to occlusion in the COX-1+/− COX-2−/− mice. There is, however, evidence that complete absence of PAI-1 induces a mild fibrinolytic state, as is shown by the accelerated clot lysis in PAI-1 deficient mice, .
The current study raises the possibility that the role of decreased fibrinolysis in the development of thrombosis may be increased in the face of drugs that inhibit COX-1 and COX-2, and especially drugs that block COX-2 activity completely and block COX-1 activity partially. This study has two potential implications for defining the mechanism of the increased number of myocardial infarctions and the other thrombotic events seen with selective COX-2 inhibitors. The first is that agents that completely inhibit COX-2 activity and partially inhibit COX-1 activity may be more likely to block fibrinolysis and thus result in thrombotic events than agents that completely inhibit COX-2 without affecting COX-1 activity. This suggests that the association of NSAIDs with cardiovascular events may depend on their selectivity for COX-1 and COX-2. The second potential implication is that the administration of a selective COX-2 inhibitor to an individual with diminished COX-1 activity, either on a genetic basis or from taking an NSAID, may result in decreased fibrinolysis and an increased risk of thrombotic events.
T. E. Riehl helped design the experiments, performed experiments, analyzed data and contributed to the manuscript; L. He performed carotid artery occlusion studies; L. Zheng performed experiments; S. Greco performed the necropsies, D. M. Tollefsen participated in the design of the study, interpreted data and contributed to the manuscript; W. F. Stenson designed the overall study, interpreted data and finalized the manuscript.
We thank G. Broze for help with the tail vein bleeding times and A. Morrison for reviewing the manuscript. This project supported by DK33165 (WFS), DK55753 (WFS) and HL55520 (DMT)
Disclosure of Conflict of Interests
The authors declare that they have no conflict of interest.