Arachidonic acid is a constituent of the SN-2 domain of membrane phospholipids. It is released from membrane phospholipids by the actions of phospholipases, particularly cytosolic phospholipase A2, which is mobilized to the membrane by an elevation in intracellular calcium, which results from a response to very diverse stimuli.

After it is released into the cytosole, arachidonic acid is acted on by enzymes called prostaglandin GH synthase, which possess both cyclooxygenase (COX) and hydroperoxidase (HOX) activity. These enzymes are colloquially known as cyclooxygenases, and they generate intermediate products, which are then acted on by isomerases expressed in a tissue-specific manner to generate prostaglandins.

Prostaglandins are known to activate G-protein coupled receptors, and these receptors derive from an ancestral receptor for the prostaglandin E, with one exception, which is the DP-2 or CRTH-2 receptor, described in Japan, which is a member of the FMLP superfamily of receptors.

Activating these receptors, prostaglandins transduce very diverse biological activities, which are often contrasting within a given system, including the cardiovascular system.

In this system, the capacity to release arachidonic acid, either by the cytosolic phospholipase (cPLA1), or by other phospholipases A-2 (PLA2), such as secretory phospholipases (sPLA2), and the presence of cyclooxygenases, either COX-1 or COX-2, and the capacity to express COX-2, tend to be present in most cells.  However, the expression of the downstream enzymes exemplified by thromboxane synthase, prostacyclin synthase, and PGE synthase, tends to occur in a tissue-specific manner.

There is some evidence, at least in heterologous expression systems, for a functional coupling of either COX-1 or COX-2 with particular downstream isomerases. An example of that is the functional coupling of COX-1 with the cytoplasmic variant of PGE synthase and COX-2 with the membrane variant of PGE synthases.  Interestingly, co-localization of these enzymes is evident in certain model systems.

The dynamic interplay between the COXs in a variety of cardiovascular settings: 1) in thrombosis, how COX-1 derived thromboxane A-2 in platelets has an interaction with vascular derived prostacyclin, 2) in atherogenesis, the role of thromboxane A-2, either derived from COX-1 in platelets or COX-2 in other cells, such as macrophages, in dynamic interplay with prostacyclin derived from COX-2, and 3) in a model of heart failure, the interplay between thromboxane A-2 and prostacyclin, both as products of COX-2 expressed in cardiomyocytes.

There is evidence for the functional coupling of COXs and downstream isomerases in heterologous expression systems, and some evidence for the co-localization of these enzymes in model systems. FitzGerald’s lab has obtained evidence for the co-localization of immunostaining of the microsomal variant of PGE synthase in the gills and branchial arches of vascular structures of the branchial arches of zebra fish, with the co-expression of COX-2 as determined by in situ analysis.

By contrast, the differential expression of the cytosolic PGE synthase in the zebra fish embryo, where it co-localizes with a different COX enzyme, COX-1, is evident in pronephric ducts, where immunostaining of cPGES can be seen and in situ detection of COX-1. In the ear, there is detection of cPGES message and detection of COX-1.

The interplay of platelet derived COX-1 formed thromboxane A-2 and vascular derived COX-2 dependent PGI-2 formation in relation to thrombosis was discussed.

The mechanism of action of aspirin, a well-known inhibitor of platelet derived thromboxane involves the hydrophobic channel of COXs, which bores into the center of these enzymes and permit the access of the lipid substrate to the active site, which involves residue such as tyrosine385.  And arachidonic acid proceeds through this hydrophobic channel and adopts a constrained hairpin configuration that puts it in approximation with the catalytic site.

Aspirin interferes with this process by irreversibly targeting a serine residue at position 530, which is close to, but not within the catalytic site.  However, acetylation of this serine residue interpolates the residue with access of the substrate to the catalytic site, thereby inactivating the enzyme.

By contrast, traditional nonsteroidal anti-inflammatory drugs (NSAIDs) are competitive active site inhibitors.  Now the contrast between the mechanism of action of inhibition of this enzyme by aspirin, and a traditional NSAID like ibuprofen, is illustrated in this study of healthy volunteers. Low doses of aspirin were given daily to attain steady state effects on inhibition of enzyme action as measured by the detection of serum thromboxane B-2, and inhibition of consequent function, that is, platelet aggregation ex vivo.

After steady state effects were attained, the offset of action over the following 24 hours after the last tablet was observed. With aspirin there was sustained maximal inhibition of serum thromboxane, and sustained maximal inhibition of function, consistent with its irreversible acetylation of serine 530.
By contrast, multiple daily doses of the NSAID administered to steady state, after discontinuation, have a pronounced offset of action of inhibition of enzyme, and an even more pronounced offset of action of function, which relates to the nonlinear relationship between inhibition of the capacity of platelets to form thromboxane and inhibition of platelet derived thromboxane-dependent platelet aggregation.

This distinction between rapid offset of functional inhibition and sustained inhibition of function is thought to underlie the cardioprotective effects of aspirin, and the reason aspirin rather than NSAIDs would be effective in this domain.

Indirect comparisons of aspirin at various doses in an overview analysis of controlled clinical trials performed by the Oxford Antithrombin Group showed that the reduction in important CV events is at least as impressive in the case of low doses of aspirin as the reduction attained at higher doses of aspirin.  No direct comparisons between low and high doses of aspirin have been performed to address this question in adequately sized clinical trials.

Cyclooxygenase enzymes, which exist as a dimer, have a hydrophobic channel that provides access to the catalytic site.  The target serine for aspirin action, using position 529, the position in the human platelet enzyme, and arachidonate accessed the catalytic site. Aspirin can prevent this access by acetylation of the serine residue.

But they hypothesized that if a NSAID had competed with the substrate for the active site, and was occupying the catalytic site, it may also prevent access of aspirin to its serine residue, and therefore its capacity to afford sustained inhibition of platelet function.

To study this, they dosed individuals to a steady state with either a low-dose aspirin daily or multiple daily doses of ibuprofen, but altering the order in which these drugs were given.  If aspirin is given first and then ibuprofen, there is sustained inhibition of thromboxane formation and sustained inhibition of platelet aggregation, similar to when aspirin is given alone.  But, if ibuprofen precedes aspirin, it looks like just ibuprofen was given. The offset of action is amplified when looking at the offset of function. This raised the possibility that ibuprofen and aspirin may undergo a pharmacodynamic interaction.

Cyclooxygenases come in two varieties, COX-1 and COX-2. A variant of cyclooxygenase-1 has been termed COX-3, although its biological importance in human systems remains to be determined.  But COX-1 and COX-2, even at the level of tertiary structure, exhibit remarkable conservation of structure.

Despite this remarkable conservation, there is evidence that these two enzymes differ substantially in terms of their biology.  And the more readily inducible COX-2 is thought to account largely for the formation of prostaglandins in inflammatory conditions.

An example of this is a study in a human model of acute inflammation, where healthy volunteers are administered bacterial lipopolysaccharide.  In this model, they develop a transient flu-like syndrome associated with ex vivo expression of both cyclooxygenase-1 and cyclooxygenase-2.  Coincident with these systemic symptoms, there was a marked increase in prostaglandin biosynthesis, by the augmentation of excretion of a urinary metabolite of prostacyclin.

Prior administration of a selective inhibitor of COX-2, celecoxib, depresses this signal very dramatically, but not completely.  And the fact that it is not complete is shown by prior administration of a mixed inhibitor of COX-1 and COX-2, ibuprofen, which further depresses this signal.  So it seems in this model as if COX-2 accounts for roughly 80%-90% of the signal. But COX-1, which is co-expressed with COX-2 in various inflammatory tissues, including atherosclerosis, accounts for about 10% of the signal.

The rationale for the development of selective COX-2 inhibitors was configured on the assumption that COX-1 was the major source of cytoprotective prostaglandins, and that the GI side effects that are associated with NSAIDs were largely accountable for by inhibition of COX-1.  So a selective inhibitor for COX-2 would target inflammation but spare the GI tract.

The proof of principle of this hypothesis is the VIGOR study in which a selective COX-2 inhibitor, rofecoxib, was compared with a mixed inhibitor, naproxen, and important GI events, particularly bleeding from ulcers, were measured.  This study resulted in a significant difference between the groups, with the reduction in the incidence of important GI events on naproxen from about 4% to about 2% on the selective COX-2 inhibitor.

Selective COX-2 inhibitors would be expected to have a very different effect on platelets than do mixed inhibitors, which inhibit both COX-1 and COX-2 with similar potency.  The reason is that in mature human platelets, COX-2 is not detected, although there is clear expression of COX-1.

The existence of COX-2 in very immature platelets forms has been recently described, in circumstances of accelerated platelet turnover, such as patients who have had a splenectomy, they may be detectible in the circulation.  But the relevance of these observations to function must be determined.

Given the absence of COX-2 in human platelets, the pharmacodynamic interaction between selective COX-2 inhibitors and low-dose aspirin would not be anticipated to be similar to that for conventional NSAIDs like ibuprofen. And indeed, that turns out to be the case.

In individuals dosed to steady state with low-dose aspirin or a selective COX-2 inhibitor, rofecoxib, differing the order of drug administration so that aspirin precedes rofecoxib or rofecoxib precedes aspirin, the selective COX-2 inhibitor has no impact on the sustained inhibition of serum thromboxane or platelet aggregation by aspirin.

Selective COX-2 inhibitors may not inhibit platelet function, but they have other effects.  Two studies showed that even in healthy volunteers, the impact of conventional NSAIDs on biosynthesis of prostacyclin, as reflected by excretion of its urinary metabolite, appeared to be largely accounted for by COX-2.  Thus, two structurally distinct COX-2 inhibitors, celecoxib and rofecoxib, depressed this index of prostacyclin biosynthesis to a degree that was comparable with the mixed inhibitors.

Before that observation, it was assumed that the primary source of endothelial prostacyclin was indeed COX-1, because in culture under static conditions, COX-2 was not expressed in these cells.  However, other investigators showed that subjection of endothelial cells to laminar shear results in sustained up-regulation of COX-2. In these human umbilical and endothelial cells, the existence of COX-2 is easily detected by in situ hybridization.

This finding raised the hypothesis that even under physiological circumstance, there might be shear-dependent induction of COX-2 and that this accounted largely for prostacyclin biosynthesis in vivo.

A paradigm is emerging of COX-1 as being the predominant source of thromboxane A-2 formed by platelets, and this activates its receptor, the TP.  By contrast, COX-2 is the dominant, although not the sole source of prostacyclin formation, and it activates its receptor, the IP.  Conventional NSAIDs and high doses of aspirin coincidentally inhibit COX-1 and COX-2. Low doses of aspirin preferentially inhibit COX-1. Selective COX-2 inhibitors, like coxibs, inhibit COX-2 dependent prostacyclin formation without coincidental inhibition of thromboxane A-2.

This seemed a rather arcane observation until the outcome of the VIGOR study.  So the same study that established, that supported the COX-2 hypothesis in terms of GI cytoprotection had another result: When CV events were evaluated a priori, there was a 5-fold difference in the incidence of myocardial infarction (MI) in patients with rheumatoid arthritis.

Several epidemiological studies have suggested that the relative risk of MI is elevated in patients with rheumatoid arthritis, compared to patients with osteoarthritis or patients without arthritis.  This has been hypothesized to result from cytokine-induced vascular damage with secondary thrombosis.  So in this population, which has a greater risk of thrombosis, there was a rather surprising divergence in terms of MI between the two groups.

This observation may be explained by naproxen behaving quite differently to the other NSAIDs, as illustrated by ibuprofen, and was actually affording cardioprotection.  Another possibility was that there was a CV associated with rofecoxib.  Another possibility was that both mechanisms might be operative, or that indeed this all derived from chance. The likelihood that it derived from chance has been diminished recently by the observation of a similar divergence in the incidence of cardiovascular events in patients receiving another COX-2 inhibitor, atericoxib, compared to naproxen.

FitzGerald’s group attempted to determine the possible mechanism by which COX-2 inhibitors, and specifically COX-2 dependent depression of prostacyclin formation, might contribute to a CV hazard.  In the catheter-induced vascular injury model, using carotid damage in mice deficient in the prostacyclin receptor IP, or the thromboxane receptor TP, or both receptors together, it was shown that the proliferative response was augmented in the absence of the prostacyclin receptor. There was an increase in the signal in the absence of the prostacyclin receptor, whether looking at the intima:media ratio or the number of proliferative cells.

In these mice, there is also a procedure related platelet activation reflected by excretion of the major murine metabolite of thromboxane, just like the increase in thromboxane metabolites that occurs peri-procedurally in patients undergoing angioplasty. Dramatically, there was a marked augmentation in thromboxane generation in response to vascular injury in the IP knockout mouse.

The evidence that prostacyclin was acting as a counter hormone against thromboxane was the complete rescue of both the proliferative phenotype and the platelet phenotype, with the coincidental deletion of the thromboxane receptor along with the IP.  So this afforded, at least in mice, a mechanism by which COX-2 inhibitors could cause a CV hazard.

Deletion of the IP does not result in thrombosis, but rather it predisposes mice that otherwise are at risk of thrombogenic stimuli.  This mechanism would not be expected to be relevant in patients of low CV risk.  Indeed review analyses of CV events in elderly patients at low CV risk, show no suggestion of increased hazard in those persons treated with a coxib.

In the CLASS study, a second large trial focusing on GI outcomes, using celecoxib, the full data set failed to detect a significant difference between celecoxib and the two traditional nonsteroidal comparators, ibuprofen and diclofenac. There was a virtual superimposition of outcome between celecoxib and diclofenac.

Looking at selectivity of NSAIDs using ex vivo whole blood assays, it is possible to plot the IC50 for inhibition of COX-2 against the IC50 for inhibition of COX-1. Increasingly selective drugs will move in this direction.  Rofecoxib is a bit more selective than celecoxib. But, they differ in potency. The selectivity of celecoxib is very similar in this type of assay to the selectivity for COX-2 of diclofenac.  This is consistent with the outcome of the CLASS Study.

The pharmacodynamic interaction between aspirin and diclofenac, compared to the pharmacodynamic interaction between aspirin and ibuprofen, shows very different effects. In the case of aspirin and diclofenac, the sustained inhibition under chronic dosing conditions looks like aspirin alone.  This is very reminiscent of the study that failed to see an interaction with the selective COX-2 inhibitor rofecoxib.  So these pharmacodynamic data are very consistent with the other suggestion that diclofenac exhibits preferential selectivity for COX-2. And by contrast, under steady state chronic dosing conditions with multiple daily doses of ibuprofen, there is the same dramatic offset of functional effect, which is somewhat more pronounced than the offset of enzyme inhibition.

Interestingly, this pharmacodynamic interaction was studied in MI patients discharged from hospital. CV mortality was assessed over the first month of discharge. Aspirin reduced CV quite significantly. Interestingly, the people who took chronic ibuprofen coincident with aspirin had an elevated odds ratio for CV hazard. By contrast, in an epidemiological study looking at real CV events, with chronic diclofenac plus aspirin, there was no such pharmacodynamic interaction.

In relation to COXs and thrombosis, the inhibition of platelet COX-1 derived thromboxane is at least sufficient to explain cardioprotection for aspirin. However, direct comparisons of low- and high-dose aspirin have not been performed.  The observations with diclofenac and ibuprofen raise the possibility that either diclofenac or a selective COX-2 inhibitor might be preferable to ibuprofen in chronic combination with low-dose aspirin. Although more epidemiological studies are underway and hopefully a clear picture will emerge.

Two plausible mechanistic hypotheses, depression of prostacyclin or cardioprotection from naproxen, might explain the results of the VIGOR trial. These mechanisms are not mutually exclusive.  Even if the CV hazard is real, it would be expected to be confined to those who are otherwise at risk of thrombosis.

The deletion of the prostacyclin receptor in prostacyclin receptor knockout mice subjected to carotid allograft accelerated the generation of graft atherosclerosis, compared to wild type mice, in work in FitzGerald’s lab.

Another study in this lab, in conventional hypercholesterolemic atherosclerosis in LDL receptor deficient mice, showed that atherosclerotic disease progresses faster in males than in females. Interestingly, deletion of the prostacyclin receptor along with deletion of the LDL receptor had no particular impact on the development of atherosclerosis in male mice.  However, this had a dramatic gene dose-dependent impact on development of atherosclerosis in the female mice, such that they caught up with males in the rate of atherosclerosis development.

This finding raised the possibility that prostacyclin plays an important role in the atheroprotection that is associated with female gender. Other studies in this lab showed that estrogens can upregulate the expression of COX-2 dependent prostacyclin in vascular smooth muscle cells (SMC).

One obvious mechanism for this finding would be that the deficiency of prostacyclin action would augment the platelet activation that accompanies the progressive development of atherosclerosis. Indeed, at 3 months, in both male and female mice, the platelet activation reflected by excretion of the thromboxane metabolite was augmented in the absence of the prostacyclin receptor.  Interestingly, even at baseline, this was evident in the female mice.

But, the finding in relation to oxidant stress is perhaps more dramatic. At baseline and at 3 months, there was a dramatic increase in lipid peroxidation during atherogenesis in the female mice lacking the prostacyclin receptor. No significant impact was seen in the male mice.

The generation of reactive oxygen species was also studied to look at the impact of prostacyclin receptor activation on oxidant stress. Even at baseline, the absence of the prostacyclin receptor is associated with an increased signal, and stimulation with H2O2 even more dramatically augmented the reactive oxygen species signal, compared to wild type mice. In further study, hydrogen peroxide-stimulated isoprostane generation, was assessed in murine aortic SMC.  The signal is greatly augmented in vascular SMC lacking the prostacyclin receptor.  By contrast, overexpression of the prostacyclin receptor completely reverses response.

In sum, this work suggests that shear dependent upregulation of COX-2, resulting in prostacyclin formation, may contribute to atheroprotection in females. The data on estrogen-stimulated COX-2 dependent prostacyclin formation, suggests that prostacyclin formed in this way may contribute in part to the supposed antioxidant effect of estrogens.

A study in FitzGerald’s lab using a mixed inhibitor, indomethacin, showed that products of COX-1 and COX-2 had some impact in atherogenesis. Indomethacin significantly retarded atherogenesis in LDL receptor deficient mice, in contrast to a nonstatistical effect of selective COX-2 inhibition. This suggests that COX-1 derived products are dominant.

Other investigators had reported that thromboxane receptor antagonism retarded atherogenesis, and thromboxane would be a prime candidate product of COX-1 in this disease.  So, using a double knockout model of atherogenesis, FitzGerald’s lab showed that indomethacin retarded atherogenesis. However, thromboxane receptor antagonism in a dose-dependent fashion had a more dramatic impact than shutting down COX-1 and COX-2 dependent thromboxane formation in these mice.

What happens to thromboxane formation during atherogenesis?  In the case of prostacyclin and isoprostanes, platelet activation increases as reflected by excretion of the thromboxane metabolized in these atherogenic mice as they developed the disease. In the animals administered the mixed inhibitor, this increase is abrogated.   The increase is not affected with either low or high doses of the thromboxane antagonist until the very terminal phases of atherogenesis.

Prostacyclin behaves like a counter-hormone, at least in the pattern of its formation, because its generation is also increasing during atherogenesis.  Indeed in patients with severe atherosclerosis, there is augmented formation of both eicosanoids.  The inhibitor depressed thromboxane and prostacyclin. The pattern with the antagonist is the same; no real effect until the very late stages of the disease.

So, now there is a pattern in which COX-1 derived thromboxane A-2 in platelets can activate the TP and appears to be contributing to the development of atherogenesis. Other data suggests that COX-2 perhaps in macrophages may contribute to a minor degree in this formation of thromboxane.  But this raises the question as to whether pre-radical catalyzed products, such as isoprostanes, may also be contributing to TP activation. Work in this lab showed that vasoactive effects of isoprostanes may transduce through the TP, with similar effects to thromboxane ligands themselves.

In relation to COXs in atherogenesis, there is evidence that COX-derived thromboxane A-2 promotes atherogenesis. TP antagonism is even more effective than shutting down thromboxane formation. The possible reasons for this divergence of effect may be the relative sparing of prostacyclin by the antagonist and indeed the prevention of the coincidental activation of the receptor by cyclooxygenase independent ligands, such as isoprostanes.

Tissue transglutaminase, also known as Gh, has been implicated in the pro-apoptotic pathway. Transglumination stabilizes intracellular structures prior to phagocytosis. However, transglutaminate, a high molecular weight protein, also binds GTP. Transglutaminate was shown to transduce alpha-adrenoreceptor-dependent activation of the delta isozyme of phospholipase C. Transglutaminate copurifies with the thromboxane receptor, and further that it could co-immunoprecipitate the TP and Gh tissue transglutaminase from a variety of CV tissues, including platelets and vascular SMCs. Interestingly, the TP exists in two cytoplasmic variants, two carboxy terminal tail variants, the alpha and the beta isoform.  In heterologous expression systems, these isoforms do not discriminate between GQ members or G-11, G-12, or G-13 family members and their ability to transduce a biochemical signal. It was somewhat curious that only Gh of these G proteins coupled in an isoform-specific fashion with TP-alpha.

FitzGerald’s laboratory explored the in vivo biology of tissue transglutaminase with respect to its relationships with the cyclooxygenase system.  In mice overexpresseing Gh in cardiomyocytes, there was consistent evidence of activation of tissue transglutaminase activity in the transgenic lines.  So, in the presence of calcium, transglutaminase activity was increased in the two transgenic lines, and this effect was abrogated by the addition to GTP.

There was a marked increase in left ventricular mass in the Gh-overexpressing cells compared to wild type cells, with a concomitant reduction in stroke volume and increase in heart rate. There was increased expression of COX-2 and thromboxane synthase in these animals, along with varied expression of various receptors for the prostacyclin, the F-2 alpha prostaglandin and thromboxane. COX-1 expression was not altered.

The dominant product of COX-2 in cardiomyocytes was prostacyclin. In this model, they showed that prostacyclins were cardioprotective.  It had been previously shown that doxorubicin-induced cardiomyopathy is associated with upregulated expression of COX-2 and that the dominant product is prostacyclin, which was cardioprotective against doxorubicin-induced cardiotoxicity. Other investigators have shown that in various models that IP deletion exacerbates cardiac reperfusion injury.

In their work, FitzGerald’s group found increased prostacyclin biosynthesis, as reflected by excretion of its major metabolite, in the overexpressing cells in their transgenic animals.
However, there was surprising concomitant increased excretion of the thromboxane metabolite, and increased oxidant stress, as reflected by increased excretion of the isoprostane.

FitzGerald’s group then looked at the impact of COX-2 inhibition on expression of this cardiac phenotype.  They found that apoptosis was abrogated with either of the two COX-2 inhibitors. In the fibrotic phenotype, the COX-2 inhibitor abrogated the signal. Left ventricular mass, cardiac output, heart rate, and systolic blood pressure were improved with COX-2 inhibition in the transgenic animals compared to the wild-type mice in their work.

In addition to increased tissue transglutaminase activity, the transgenic animals also have increased activation of TP-dependent signaling.  Further, the thromboxane-dependent activation of ERK is mediated by a Gh, as shown by blockade of this effect in the transgenic animals with an antibody directed against Gh.

Treatment with a thromboxane receptor antagonist had an effect very similar to the effect with the COX-2 inhibitor, and improved cardiac output.

Because both COX-1 and COX-2 could form thromboxane A-2, this group then studied whether the elevated oxidant stress might have functional consequences. They compared a COX-2 inhibitor with a dosing regimen of vitamin E sufficient to completely suppress isoprostane generation in these animals.  Elevated left ventricular mass, stroke volume, heart rate, and systolic blood pressure in the transgenic animals were improved with either the COX-2 inhibitor or vitamin E.  The COX-2 inhibitor rescued the phenotype and vitamin E had a significant ameliorative effect. 

In summary, directed overexpression of Gh tissue transglutaminase in myocytes results in a cardiac phenotype that is mediated by TP activation.  TP activation by COX-2 derived thromboxane A-2 and lipid peroxidation products may also be relevant in this phenotype.

Both Gh and COX-2 are upregulated in human cardiomyocytes in patients with heart failure. The implications for human heart failure are not clear yet.  In models of heart failure, either prostacyclin or thromboxane A-2 may predominate as the major product of COX-2, depending on the model.  The impact of COX-2 inhibitors in human heart failure may be quite heterogeneous.