Cyclo-oxygenase isoenzymes: physiological and pharmacological role


P. C. A.Kam Senior Staff Anasthetist


Prostaglandins play important roles in inflammation and the maintenance of normal physiological function of several organ systems. Prostaglandin production requires the conversion of arachidonic acid to the intermediate prostaglandin H2 catalysed by the cyclo-oxygenase (COX) enzyme. There are two isoforms of the COX enzyme, COX-1 and COX-2. These isoforms vary in their distribution and expression but are similar in size, substrate specificity and kinetics. Normal physiological functions are mediated by ‘constitutive’ COX-1, while the inflammatory response is mediated by ‘inducible’ COX-2. Current nonsteroidal anti-inflammatory drugs inhibit both enzymes to varying degrees and can cause adverse effects in the gastrointestinal tract, kidney, respiratory system and platelets. Newer, selective COX-2 inhibitors offer real hope for safer anti-inflammatory drugs although their long-term safety and efficacy need to be studied as questions remain unanswered about possible physiological functions of COX-2.

Arachidonic acid products such as prostaglandins (PGs), thromboxane and 15-hydroxy-eicosatetraenoic acids (HETE) are collectively referred to as eicosanoids. The regulation of eicosanoid production has become an area of intensive research because these fatty acid metabolites are known to affect multiple signalling pathways that modulate physiological processes such as inflammation, ovulation, modulation of immune responses and mitogenesis. There are several sites in the biosynthesis of PGs that may be regulated. The first step is the liberation of arachidonic acid from membrane phospholipid by phospholipase. The free arachidonic acid is converted to PGG2 by cyclo-oxygenase (COX, or PGH synthase). Prostaglandin G2 is then converted to PGH2 by a peroxidase reaction that is also catalysed by the COX enzyme. This COX enzyme present normally in different tissues is known as COX-1 and was purified by Miyamoto et al. in 1976 [ 1]. Prostaglandin H2 is then converted to other PGs, prostacyclins and thromboxanes by tissue-specific isomerases [ 2]. In platelets, nearly all the PGH2 is converted to thromboxane A2 by thromboxane synthase. In smooth muscle cells, PGI2 is produced by PGI synthase and, in vascular tissues, the endothelial cells produce PGE2 and PGI2. In 1989, Simmons et al. [ 3] identified a second inducible form of COX, known as COX-2 or PG endoperoxide synthase-2.

Cyclo-oxygenase-2 expression is inducible by a variety of extracellular and intracellular stimuli, including lipopolysaccharide (LPS), foskolin, interleukin (IL)-1, tumour necrosis factor (TNF), epidermal growth factor (EGF)-α, platelet-activating factor, interferon-γ (IFγ) and endothelin. Formation of COX-2 protein is also associated with the increase in prostanoid production that results from mitogenic stimulation in a wide variety of cell types such as intestinal epithelium. Although COX-1 and COX-2 produce similar products, it is likely that these two isoforms have different roles because they are located in different subcellular compartments.

Flowers & Vane [ 4] demonstrated that COX preparations from the brain were sensitive to paracetamol compared to those from spleen. Vane & Botting [ 5] proposed that there may be a COX-3 isoform on which paracetamol has a preferential action. Paracetamol has analgesic and antipyretic effects but weak anti-inflammatory activity.

Cyclo-oxygenase isoforms

Both COX isoforms are haemoproteins that catalyse the oxygenation of arachidonic acid to form PGG2, which is reduced to form PGH2. A proposed model for the structure of the active site of the COX isoenzymes contains a single haem group near the binding site for arachidonic acid. There are separate substrate and peroxidase binding sites. Arachidonic acid, derived from phospholipids on the cytoplasmic face of endoplasmic reticulum and the nuclear envelope, enters a four-sided pore formed by 4α-helical regions of the membrane-binding domain of the isoenzyme to bind to the COX active site. The genes [ 6] for the two enzymes are found on different chromosomes, chromosome 9 for COX-1 and chromosome 1 for COX-2.


Cyclo-oxygenase-1 is a membrane-bound haemoglycoprotein with a molecular weight of 71 kDa found in the endoplasmic reticulum of PG-producing cells. The enzyme has two functions: it first cyclises arachidonic acid (converts it to a closed-ring structure) and adds a 15-hydroperoxy group to form PGG2, and then the hydroperoxy moiety of PGG2 is reduced to the hydroxy form of PGH2 by a peroxidase enzyme moiety in the same COX enzyme protein. The COX–1 integrates into only a single leaflet of the lipid bilayer and this is described as a ‘monotopic’ arrangement. Pitot et al. [ 7] determined the three-dimensional structure of COX-1. The enzyme has three independent folding units: an EGF-like domain, a membrane-binding domain and an enzymatic domain. The α-helices of the membrane binding domains, forming a channel entrance to the active site, are inserted into the membrane, thereby allowing arachidonic acid to gain access into the interior of the lipid bilayer. The sites for COX and peroxidase activity are spatially distinct but are adjacent to each other. The COX active site is a long hydrophobic channel with tyrosine 385 and serine 530 at its apex. Nonsteroidal anti-inflammatory drugs (NSAIDs) block COX-1 at about halfway down the channel by hydrogen bonding to the polar arginine at position 120. Aspirin acetylates serine 530, irreversibly inhibiting COX-1, preventing access for arachidonic acid into the COX-1 active site.


In the 1990s, studies demonstrated increased COX activity in a variety of cells after exposure to LPS and cytokines, indicating the presence of an inducible form of COX that was labelled COX-2. This enzyme was originally discovered as a transcriptional protein that was upregulated during inflammation and decreased by glucocorticoids. No increase in COX-1 messenger ribonucleic acid (mRNA) was observed by reverse transcription and polymerase chain reaction analysis. It was suggested that COX-1 was responsible for physiological or ‘housekeeping’ functions while COX-2 was an inducible enzyme involved in inflammation, mitogenesis and specialised signal transduction. This led to the notion that the anti-inflammatory, analgesic and antipyretic effects of NSAIDs were due to their inhibitory action on COX-2 whilst the antithrombotic and adverse effects might be due to inhibition of COX-1. Cyclo-oxygenase-2 has a molecular weight of 70 kDa with similar active sites for the attachment of arachidonic acid to the NSAIDs. The three-dimensional structure of COX-2 closely resembles that of COX-1. However, its active site has a slightly larger volume because it has a larger central channel (by about 17%) with a wider entrance and a secondary internal pocket, and can accommodate larger drugs. A single amino acid difference at position 523 is critical for selectivity of the NSAIDs: an isoleucine molecule in COX-1 and a valine in COX-2. As the valine at 523 in COX-2 is smaller (by one methyl group), a gap in the wall of the channel is present, giving access to a side pocket which is the binding site of the COX-2 selective inhibitors. The larger isoleucine at position 523 in COX-1 blocks access to the side pocket. When blocked by NSAIDs, neither enzyme is capable of producing PGH2. When blocked by aspirin or NSAIDs, COX-1 is inactivated completely but COX-2 converts arachidonic acid to 15-R-hydroeicosatetraenoic acid (15-R-HETE) [ 8].

Physiological actions of prostaglandins

Cyclo-oxygenase-1 is constitutively expressed in most tissues. Under physiological conditions, COX-1 activity predominates and serves to produce PGs that regulate rapid physiological responses such as regulating vascular homeostasis, gastric function, platelet activity and renal function. The concentration of the enzyme is low but may increase two- to four-fold in response to stimulation by hormones or growth factors.

Although it was previously thought that COX-2 is only present in cells that are induced by inflammatory factors, low basal (normal) concentrations of the enzyme can be detected in the brain, kidney and the gravid uterus. Cyclo-oxygenase-2 is regarded as an inducible enzyme and its expression by monocytes, synovial cells and fibroblasts may be increased 10- to 80-fold when stimulated by growth factors, cytokines, bacterial LPS or phorbol esters as part of an inflammatory response. During inflammation, there is increased expression of COX-2 mRNA followed by increased COX-2 production and tissue PGE2 concentrations.

Gastrointestinal tract

The cytoprotective actions of PG preventing gastric ulceration are mediated by endogenous prostacyclin and PGE2, which reduce gastric acid production, stimulate gastric fluid secretion, increase secretion of viscous mucus and exert a direct vasodilator action on gastric mucosa. In addition, secretion of bicarbonate by the duodenum is also enhanced [ 9].

Epidemiological evidence suggests a reduced relative risk of colon cancer in individuals taking NSAIDs. There is increased expression of COX-2 in human colorectal adenomas and carcinoma, with a marked increase in COX-2 mRNA and consequently increased PGE2. Cyclo-oxygenases catalyse the formation of PGs that promote tumour growth. Blocking the activity of COX-1 and COX-2 may be important for preventing cancer [ 10].

Renal function

Prostaglandins are involved in three areas of normal renal function: control of renin release, regulation of vascular tone and control of tubular function [ 11]. The normal renal cortex synthesises both PGE2 and PGI2, whereas the renal medulla produces predominantly PGE2. Production of these PGs is mediated predominantly by COX-1. Prostaglandin I2 stimulates renin release directly from the juxtaglomerular cells. Prostaglandin E2 increases renin release indirectly and requires an intact macula densa. Prostaglandins also modulate vascular tone as PDF2 and thromboxane A2 are vasoconstrictors, and PGE2 and PGI2 vasodilate the afferent arteriole, maintaining renal function in the presence of increased levels of angiotensin II (a vasoconstrictor). Prostaglandins are important for renal water and salt transport as they exert their effects locally at or near the site of synthesis, acting as autocoids. Prostaglandin E2 inhibits sodium and chloride reabsorption from the ascending loop of Henle. The PGs also attenuate the antidiuretic actions of vasopressin on the collecting ducts, increasing urine flow. Cyclo-oxygenase-1 is predominantly found in the cells of the medullary collecting ducts with small amounts in the cells of the cortical collecting tubules; it is absent in the macula densa and the cells of the thick ascending limb of Henle's loop of the cortical nephrons.

Cyclo-oxygenase-2 is present primarily in the macula densa of the juxtaglomerular apparatus and adjacent cells of the cortical thick ascending limb of the loop of Henle in normal rats, rabbits and dogs. Volume depletion leads to upregulation of COX-2 in the macula densa of these animal models [ 12]. It is suggested that the macula densa mediates an increase in renin release through PGs derived from COX-2 in these animals [ 13]. However, COX-2 is not found in the macula densa of normal human kidneys and it is not known whether volume depletion leads to upregulation of this enzyme in humans. The introduction of very selective COX-2 inhibitors should lead to a better understanding of the role of COX-2 in human renal physiology.

Mice that lack the gene for COX-1 production do not show significant signs of renal disease, whereas in those that lack the COX-2 gene, the kidneys fail to develop normally resulting in death [ 14].

Central nervous system

The brain contains high concentrations of PGD2 and PGE2. Prostaglandin D2 in the pre-optic area induces sleep whereas intracerebral injections of PGE2 decrease sleep [ 15]. Brains of neonates have higher concentrations of PGs, synthesised mainly by ‘constitutive’ COX-2, which are responsible for regulation of blood flow in the newborn [ 16].

Cyclo-oxygenase-1 is distributed throughout the brain but is most abundant in the forebrain. The PGs in the forebrain may be involved in the control of the autonomic nervous system and sensory processing [ 17]. High basal levels of COX-2 are found in the brain, especially the granule cell and pyramidal cell layers of the hippocampus. The pyriform cortex, layers I and II of the neocortex and the amygdala complex have relatively high concentrations of COX-2. The thalamus, hypothalamus and the caudate–putamen complex contain low levels of COX-2. Increased neural activity such as seizures or N-methyl d-aspartate (NMDA)-dependent synaptic activity will increase expression of COX-2 in the hippocampus, whereas acute stress raises the concentrations in the cerebral cortex [ 18].

Fever is thought to be the effect of PGE2 released by induced COX-2 on the endothelium of cerebral blood vessels acting on the pre-optic area [ 19]. Increased COX-2 expression has been demonstrated in the basilar artery in the acute and chronic phases of experimental subarachnoid haemorrhage, suggesting that elevated inflammatory cytokines may be involved in the induction of COX-2, which may produce eicosanoids that cause vasospasm [ 20].

Cyclo-oxygenase-2 expression is increased in the spinal cord of normal animals owing to processing of nociceptive stimuli [ 21].

Reproductive function

Prostaglandin E2 and PGF are synthesised by the decidua and myometrium of the uterus, fetal membranes and umbilical cord. The capacity of these tissues to produce PGs is increased during pregnancy via induction of COX-2. The concentrations of PGs are increased in blood and amniotic fluid during labour [ 22]. During early pregnancy, COX-1 and COX-2 are expressed and may be important for implantation of the ovum and angiogenesis, which is important for development of the placenta.

Constitutive COX-1 in the amnion may be important for the maintenance of a healthy pregnancy. Production of COX-1 in the amnion is increased by human chorionic gonadotropin. Prostaglandins derived from COX-2 activity may be important for initiating uterine contractions during labour. Nonsteroidal anti-inflammatory drugs such as indomethacin delay premature labour by inhibiting uterine production of PGs. Cyclo-oxygenase-2 mRNA increases markedly immediately before and after the start of labour. Glucocorticoids, EGF, IL-1β and IL-4 stimulate COX-2 production in human amnion cells.

During the luteal and menstrual phases of the ovulatory cycle, PG production is stimulated by increased COX-2 and this is associated with the occurrence of period pains during the ovulatory cycle [ 23]. Mice lacking COX-2 are infertile because of reduced ovulation [ 24].

Monocytes and macrophages

Monocytes and macrophages contain both COX-1 and COX-2 enzymes. Induction of COX-2 by LPS has been shown in human alveolar macrophages [ 25]. Cytokines such as IFγ and TNF have a synergistic effect on induction of COX-2 whilst IL-10 down-regulates COX-2 expression.

Endothelial cells

Cyclo-oxygenase-2 has been demonstrated in human umbilical vein endothelial cells after induction by IL-1α and phorbol ester [ 26]. During inflammation, the increased permeability of the vascular endothelium is caused by retraction of endothelial cells leading to exudation and migration of phagocytic cells [ 27].

Synovial tissue

Both COX isoforms are present in synovial tissues in patients with rheumatoid arthritis and osteoarthritis; upregulation of COX-2 by IL-1 has been reported in human chrondrocytes and osteoblasts [ 28].


Cyclo-oxygenase-2 in the lung may be induced either locally in pulmonary structures (airway epithelium, airway smooth muscle, lung macrophages and activated leucocytes) after airway damage or as part of a systemic response to cytokines. Endogenous PGE2 may have a bronchoprotective function such as modulation of airway and vascular tone, inflammatory cell activity and recruitment, cytokine release, mucus secretion, and cholinergic and sensory nerve function [ 29, 30].

Vascular smooth muscle

Induction of COX-2 has been demonstrated in vitro in arterial smooth muscle cells treated with platelet-derived growth factor, EGF, thrombin and also after vessel injury, indicating a pathophysiological role of COX-2 in the modulation of vascular responses to injury [ 31].

Pharmacological aspects

The currently available NSAIDs inhibit both COX-1 and COX-2 to varying degrees. Inhibition of inducible COX-2 is the principal anti-inflammatory mechanism of the NSAIDs. Different NSAIDs have relatively equivalent efficacy but individual responses to each drug may vary.

Long-term use of NSAIDs is associated with significant adverse effects as they all inhibit production of COX-1. Inhibition of this enzyme also causes suppression of local and systemic production of PGE2 leading to reduced pain and inflammation. However, reduced PGE2 production in the gastrointestinal tract leads to gastritis and ulceration. Epidemiological studies have shown that 20–30% of hospitalisations and deaths in patients above the age of 65 years were related to NSAID use [ 32]. Risk factors for gastrointestinal complications include age > 65 years, history of peptic ulcer disease, concomitant use of corticosteroids or anticoagulants, smoking and alcohol consumption. Nonsteroidal anti-inflammatory drugs may also cause reversible renal failure and the risk factors for this complication include advanced age, hypertension, congestive cardiac failure, and the concomitant use of diuretics, gentamicin or angiotensin converting enzyme inhibitors. Combinations of NSAIDs (with the exception of aspirin at doses under 325−1 as prophylaxis against cardiovascular events) should not be used because of a greater risk of adverse effects and lack of evidence that efficacy is increased [ 33]. The NSAIDs can also cause reversible platelet dysfunction, liver dysfunction and bronchospasm. In the lungs, inflammatory cells express COX-2 and 5-lipoxygenase resulting in the release of prostanoids and leukotrienes. In the presence of aspirin, COX-2 is modified enzymatically to produce 15-HETE instead of PGs. The 15-HETE is further metabolised by 5-lipooxygenase present in leucocytes to form leukotrienes such as 15-epilpoxin A2, which mediatesmooth muscle constriction and mucus secretion. It is suggested that the diversion of arachidonic acid to 5-lipooxygenase products by inhibition of COX-2 leads to the production of leukotrienes, which mediate bronchospasm [ 34].

Overviews of clinical data indicate that the blood pressure of patients with controlled hypertension can be increased by 3–6 mmHg during concurrent treatment with NSAIDs, increasing the risk of stroke, congestive cardiac failure or renal failure [ 35]. The use of NSAIDs, particularly indomethacin, to prevent premature labour and in the treament of polyhydramnios has been associated with constriction of the fetal ductus arteriosus, with serious adverse consequences to the fetus [ 36].

Animal studies have indicated that COX-2 mediates lamellar bone formation induced by mechanical strain. The selective COX-2 inhibitor NS-398 completely blocked bone formation with depression of mineral deposition [ 37] and this may delay bone healing in orthopaedic patients.

Differential inhibition of cyclo-oxygenase isoforms

As the pharmacological profile of COX-1 and COX-2 are different, it is possible to develop new anti-inflammatory drugs that specifically inhibit COX-2 and reduce the adverse effects of NSAIDs. Cultured bovine aortic endothelial cells serve as a source of COX-1, and J774.2 macrophages, which contain little or no COX-1 stimulated by bacterial LPS, serve as a means of testing for inhibition of COX-2. In both test systems, the release of PGs is measured. The selectivity of various COX inhibitors is evaluated by determining the ratio of IC50 (mean concentration of drug that reduces PG synthesis by 50%) for COX-2 relative to COX-1. A ratio < 1 indicates a greater inhibitory effect on COX-2 whilst a ratio > 1 indicates preferential inhibition of COX-2. The absolute IC50 values as well as the IC50 ratio of COX-2 vs. COX-1 inhibition vary greatly depending on the test tissue (model) used. However, the rank order of selectivity for COX-2 vs. COX-1 enzymes within a range of compounds appears to be reproducible from one test tissue to another [ 38, 39].

The NSAIDs have been tested for their relative inhibition of COX-1 and COX-2 isoforms. Piroxicam causes 250 times greater inhibition of COX-1 than of COX-2; aspirin is 166 times more active and indomethacin is 60 times more active. Diclofenac and naproxen are equipotent in inhibiting COX-1 and COX-2. These drugs are well known for their propensity to cause gastrointestinal ulceration and bleeding.

As NSAIDs are used by a large number of patients with osteoarthritis, rheumatologists have re-evaluated the use of NSAIDs as primary treatment because of the high incidence of gastrointestinal side-effects. Although paracetamol is effective against mild to moderate pain, it is only effective for a short time. Other more potent NSAIDs are as effective as paracetamol for the treatment of mild to moderate pain in osteoarthritis but also allow faster recovery from joint pain after activities.

Selective COX-2 inhibitors

A number of compounds that preferentially inhibit COX-2 relative to COX-1 should have less side-effects and less irritant action on the stomach.

The pharmaceutical company Monsanto-Searle has synthesised inhibitors 1000 times more potent against COX-2 than against COX-1 [ 40]. SC5815 has an IC50 of 0.07 μm for COX-2 and 100 μm for COX-1, making it 1400 times more selective for COX-2. Even at greater than anti-inflammatory doses, SC58125 did not cause gastric ulceration and no changes in the production of renal PGs were observed. The anti-inflammatory effects of a prototype of this compound, celecoxib (SC58635), have been investigated in clinical trials. In a standard dental pain model, celecoxib appeared to be similar to aspirin in its analgesic effects. Sixty per cent of patients who received placebo required rescue medication (paracetamol) within 1 h and 90% within 3 h. In contrast, most patients taking aspirin or celecoxib did not require rescue medication within 1 h and 40% taking aspirin and 50% of celecoxib-treated patients did not require any rescue medication within 4 h [ 41]. Two short-term studies indicate that celecoxib is effective in the treatment of pain in osteoarthritis and rheumatoid arthritis without causing side-effects related to COX-1 inhibition such as gastrointestinal irritation, platelet dysfunction, renal insufficiency or liver toxicity [ 42, 43].

L-475.337, another selective COX-2 inhibitor from Merck-Frosst, is 1000-fold more selective for COX-2 in vitro with good anti-inflammatory activity in vivo. This agent reduced carragenin-induced rat paw oedema and LPS-induced pyrexia in rats [ 44].

Flosulide (CGP 28238) is a highly selective inhibitor of COX-2 with a COX-2 : COX-1 ratio of 0.0002. There is remarkably reduced gastric damage with the use of flosulide [ 45].

Meloxicam (Boehringer Ingelheim) has potent anti-inflammatory properties with a COX-1 : COX-2 ratio of 0.8, and an estimated three- to 77-fold selectivity for COX-2 [ 46]. The mode of action of meloxicam is unknown, although it has been suggested that it may work at the apex of the channel. It has an elimination half-life of 20 h. Clinical trials suggest that it is efficacious in the dose range 7.5–22.5−1. In a double-blind study, meloxicam 7.5 mg caused no more gastric mucosal injury than placebo over 23 days, but there was an increase in gastric erosions at a dose of 15−1. In a large trial, meloxicam at a dose of 7.5−1 caused less gastrointestinal side-effects than diclofenac 100−1, but was slightly less efficacious [ 47]. The safety profile of higher doses of meloxicam is not known. Recent data [ 48] suggest that meloxicam inhibits platelet thromboxane synthesis at a dose of 15−1 although it spared COX-1 activity at a lower dose of 7.5−1.

Rofecoxib (Merck and Co.) is a methylsulphonylphenyl derivative with an 800-fold selectivity for COX-2. Phase II trials suggest that a dose range of 12.5 or 25−1 was as effective as diclofenac 150−1 [ 49]. In doses of 250 mg daily (well above the recommended dose), rofecoxib was well tolerated, with no change in bleeding time or increase in gastric erosions or ulcers [ 50]. The incidence of ulcers in patients taking rofecoxib for periods up to 6 months is not yet available.


The discovery of at least two forms of COX has given a new impetus to the search for less toxic NSAIDs. The good relationship found between preferential inhibition of COX-2 relative to COX-1 in vitro and the improved pharmacological profile in vivo for various NSAIDs indicates that preferential inhibition of COX-2 may maintain or enhance anti-inflammatory activity whilst minimising the gastric and renal side-effects of these drugs. A high degree of COX-2 selectivity may not ensure freedom from adverse effects, because COX inhibition in the different tissues depends on a combination of COX selectivity, COX potency and drug pharmacokinetics. Ideally, NSAIDs should be prescribed in doses that are effective for their anti-inflammatory, antipyretic and analgesic actions, without producing serum and tissue concentrations that reduce COX-1 activity. Furthermore, the physiological importance of constitutive expression of COX-2 in various tissues such as the gastric mucosa and the kidneys needs to be investigated more thoroughly. As such, the selective COX-2 inhibitors should be used cautiously until these issues are resolved. Studies on COX-2 expression in various pathological conditions suggest new potential therapeutic indications for COX-2 inhibitors such as neuronal injury, pain, inflammation and the management of preterm delivery, as well as adjunctive therapy in the management of large bowel cancer or adenoma.