J Biol Chem, Vol. 274, Issue 33, 22903-22906, August 13, 1999

MINIREVIEW
Arachidonic Acid Oxygenation by COX-1 and COX-2
MECHANISMS OF CATALYSIS AND INHIBITION^*,

Lawrence J. Marnett, Scott W. Rowlinson, Douglas C. Goodwin, Amit S. Kalgutkar, and Cheryl A. Lanzo^§

From the Departments of Biochemistry and Chemistry, Center in Molecular Toxicology, and the Vanderbilt Cancer Center, Vanderbilt University School of Medicine, Nashville, Tennessee 37232

	INTRODUCTION

TOP INTRODUCTION Cyclooxygenase Catalysis Cyclooxygenase Inhibition Future Directions REFERENCES

Prostaglandins were discovered in human semen in 1930, but their low concentrations and instability precluded identification for nearly 30 years (for a brief historical review, see Ref. 1). Once they were identified, it was clear they arose from polyunsaturated fatty acids by a complex series of reactions involving oxygenation, cyclization, and the generation of five chiral centers from an achiral substrate. The mechanism of prostaglandin biosynthesis was outlined in 1967 by Hamberg and Samuelsson (2), and the basic tenets have been confirmed in subsequent studies. The key step in their proposed mechanism was the formation of bicyclic peroxides (endoperoxides) as the initial products of polyunsaturated fatty acid oxygenation (Fig. 1). The term cyclooxygenase (COX)¹² was coined to describe the enzyme that carried out this complex chemical transformation, and its role was confirmed by the isolation of prostaglandin endoperoxides in 1973 (3, 4).

View larger version (6K): [in this window] [in a new window]   Fig. 1.   Chemical steps in the conversion of arachidonic acid to PGG2. The enzyme removes the 13-pro-S-hydrogen, which generates a pentadienyl radical with maximal electron density at C-11 and C-15. Trapping of the carbon radical at C-11 with O2 produces a peroxyl radical, which adds to C-9 generating a cyclic peroxide and a carbon-centered radical at C-8. The C-8 radical adds to the double bond at C-12, generating the bicyclic peroxide and an allylic radical with maximal electron density at C-13 and C-15. Trapping of the carbon radical at C-15 with O2 generates a peroxyl radical which is reduced to PGG2.

In addition to catalyzing a fascinating metabolic transformation, COX is an enormously important pharmacological target. Vane reported in 1971 (5) that non-steroidal anti-inflammatory drugs (NSAIDs) inhibit prostaglandin formation and demonstrated that their relative inhibitory potency in vitro correlates to their anti-inflammatory activity in vivo. This not only explained the beneficial activity of NSAIDs but also their side effects such as gastrointestinal toxicity and bleeding because prostaglandins and related molecules (i.e. thromboxane) are involved in a very broad range of physiological and pathophysiological responses. The importance of these molecules as autocrine and paracrine mediators has been confirmed recently by the phenotypes of mice bearing targeted deletions in COX genes or prostaglandin receptor genes.

The discovery of a second gene (COX-2) coding for cyclooxygenase and the demonstration that its protein product is distributed differently from the originally discovered enzyme (COX-1) raised the possibility that some of the beneficial effects of NSAIDs may be separable from their side effects by development of isoform-selective inhibitors (6-9). This hypothesis has been dramatically validated by the demonstration that selective COX-2 inhibitors are anti-inflammatory and analgesic but lack the gastric toxicity associated with all currently available NSAIDs (10, 11).

	Cyclooxygenase Catalysis

TOP INTRODUCTION Cyclooxygenase Catalysis Cyclooxygenase Inhibition Future Directions REFERENCES

Substantial evidence supports the hypothesis that COX oxygenates arachidonic acid by a free radical mechanism (Fig. 1). Thus, COX appears to have co-opted the process that gives rise to isoprostanes to generate prostaglandins. The major differences between COX-catalyzed and spontaneous oxidation of arachidonic acid are the increased rate and high degree of stereochemical control of the enzymatic reaction (1 of 64 possible isomers predominates). Thus, the overall role of COX is rather simple: stereospecifically remove the 13-pro-S-hydrogen and control the stereochemistry of oxygenation. How does it do this?

Oxidizing Agent-- A protein tyrosyl radical appears to be the oxidizing agent that initiates arachidonic acid oxygenation (12). A tyrosyl radical is formed during COX turnover, and although there has been debate over the identity of the spectroscopically detected radicals, they appear capable of oxidizing arachidonic acid (13-15). The crystal structures of both COX-1 and COX-2 reveal that Tyr-385 is positioned perfectly to react with the fatty acid substrate (Fig. 2) (16-18). Indeed, the Y385F mutant is not catalytically active and does not oxidize arachidonic acid when it is treated with peroxide (19). Incubation of wild-type enzyme with arachidonate in the presence of nitric oxide quenches the EPR signal of the tyrosyl radical and leads to the formation of nitrotyrosine at position 385 in the protein (20, 21).

View larger version (66K): [in this window] [in a new window]   Fig. 2.   Subunit structure of COX-1. Each COX subunit comprises three domains, an epidermal growth factor domain (yellow), a membrane-binding domain (lavender), and a catalytic domain (blue). The catalytic domain contains the cyclooxygenase active site and the peroxidase active site separated by the heme prosthetic group (red). In the present structure, the cyclooxygenase active site is occupied by a molecule of iodosuprofen (lime). Arg-120, Tyr-355, and Glu-524 comprise a H-bonding network that introduces a constriction at the base of the cyclooxygenase active site. They are depicted in gold. The volume beneath this constriction is termed the lobby and is bordered on three sides by the membrane-binding domain. The catalytically important Tyr-385 residue is depicted in white. The peroxidase active site is at the top of the protein in this drawing and is visible as the wide opening to the heme prosthetic group.

Protein radicals require an oxidant for their formation, which in most cases is a metal-containing prosthetic group (22). COX is a homodimer of 70-kDa subunits that each contain one molecule of heme (16). The iron is ferric in the resting enzyme and is likely thermodynamically incapable of oxidizing Tyr-385 (E_1/2 = 0.9 V for Tyr^·  Tyr and E_1/2 = 0.2 to +0.2 V for Fe³⁺  Fe²⁺ for most hemes) (22, 23). Reaction of the heme of COX with peroxides generates a ferryl-oxo complex analogous to compound I of classic heme peroxidases (24). The redox potential of such higher oxidation states is typically on the order of +1 V so the compound I of COX is capable of oxidizing Tyr-385. Ruf and co-workers (25) demonstrated some time ago that oxidation of COX with organic hydroperoxides or fatty acid hydroperoxides generates a spectroscopically detectable tyrosyl radical, and they postulated that the tyrosyl radical oxidizes arachidonic acid. Support for this hypothesis is provided by the existence of significant lag phases for the induction of cyclooxygenase activity of Mn-protoporphyrin IX-reconstituted enzyme or site-directed mutants that exhibit diminished rates of reaction with hydroperoxide (26).

The identity of the hydroperoxide activator has been uncertain. Our laboratory has reported that peroxynitrite, the coupling product of nitric oxide and superoxide anion, is an excellent oxidant for the heme of COX and activates the enzyme even in the presence of concentrations of glutathione peroxidase and glutathione that inhibit activation by fatty acid hydroperoxides (27). Activation is inhibited by superoxide dismutase, which scavenges peroxynitrite or prevents its formation from NO and O₂. Lipophilic superoxide dismutase mimetic agents inhibit prostaglandin biosynthesis by cultured mouse macrophages, which is consistent with a role for peroxynitrite in cyclooxygenase activation in intact cells. These findings provide a biochemical link between NO biosynthesis and prostaglandin biosynthesis and may explain the finding that NO synthase inhibitors reduce prostaglandin biosynthesis in inflammatory lesions in vivo (Equation 1) (28). Peroxynitrite activation of cyclooxygenase may be especially important in activated macrophages because inducible NO synthase and COX-2 are immediate early genes that are dramatically expressed in response to exposure to inflammatory stimuli such as lipopolysaccharide. The identity of the cyclooxygenase activator in non-inflammatory cells remains to be determined.

View larger version (5K): [in this window] [in a new window]   Equation 1.

There has been considerable experimental debate about whether the Tyr-385 radical is regenerated at the end of each cyclooxygenase catalytic cycle or requires reoxidation by another peroxidase catalytic cycle (29, 30). Detailed kinetic investigations confirm observations extant at the onset of the debate that strongly support the regeneration of the catalytic tyrosyl radical at the end of each cycle of arachidonic acid oxidation (4, 31, 32). Thus, the final step in each round of arachidonic acid oxygenation is reduction of the peroxyl radical precursor of PGG₂ by Tyr-385, which regenerates the Tyr-385 radical for the next round of cyclooxygenase catalysis (Fig. 1). This leads to multiple turnovers per activation event and allows the accumulation of PGG₂ (4, 32).

Stereochemistry-- How does COX ensure that a single stereoisomer of PGG₂ is produced from arachidonic acid? One can predict on purely chemical grounds that the enzyme must bind arachidonate in a conformation similar to that illustrated in Fig. 1. Removal of the 13-pro-S hydrogen followed by reaction with O₂, serial cyclization, and reaction with the second O₂ could occur with minimal motion of the reaction intermediates to produce PGG₂ with all the correct stereocenters. We and others have docked arachidonate into the cyclooxygenase active site of sheep COX-1 to test whether such a conformation can be accommodated. The carboxylate was positioned adjacent to Arg-120, which plays a crucial role in binding arachidonate and arylalkanoic acid-type inhibitors (33), and the 13-pro-S hydrogen was placed near the phenolic hydroxyl of Tyr-385. The -end of the fatty acid was inserted into a channel at the top of the cyclooxygenase active site that eventually leads to the surface of the protein. This end of arachidonate straddles the -helix containing Ser-530, the site acetylated by aspirin. Acetylation by aspirin blocks arachidonate binding to COX-1 (34). The arachidonate-enzyme complex was then minimized to obtain the final conformation displayed in Fig. 3. It is clear that the fatty acid substrate can be nicely accommodated within the active site in a conformation expected to yield PGG₂. One prediction of this model is that O₂ molecules diffuse up the central channel to couple to the solvent-exposed sides of the carbon radical intermediates to form PGG₂. This prediction is consistent with the fact that the 13-pro-S hydrogen is on the opposite side of the fatty acid from the direction of both O₂ coupling reactions.

View larger version (42K): [in this window] [in a new window]   Fig. 3.   Stereo drawing of a model of arachidonic acid bound at the cyclooxygenase active site of sheep COX-1. The COX-1-flurbiprofen structure was used as a starting point for model building. The flurbiprofen was removed, and the arachidonic acid molecule was positioned as described in the text. The overall structure was then minimized. The arachidonate is highlighted with different colors for various regions of the molecule. The carboxylate group (blue and red) is positioned near Arg-120 (yellow) and Tyr-355 (lime). Carbon-13 is yellow, and the pro-S and pro-R hydrogens are white and lavender, respectively. The -end of arachidonate is positioned above the helix containing Ser-530 (lavender) and Gly-533 (red).

This model was developed using coordinates for sheep COX-1, but a similar model can be developed using the coordinates for mouse COX-2. The active site structures are quite similar for the two forms of the enzyme, with a few exceptions that are detailed below. Despite this similarity, clear differences in substrate specificities exist between the two enzymes. COX-2 appears much more accommodating than COX-1 in that it oxidizes 18 carbon polyunsaturated fatty acids with much higher efficiency than COX-1. Also, COX-2 oxidizes the hydroxyethylamide derivative of arachidonic acid (anandamide) whereas COX-1 does not (35, 36). Consistent with the latter observation, the R120Q mutant of COX-1 demonstrates a 100-fold increase in K_m for arachidonate whereas the corresponding mutant of COX-2 displays a K_m for arachidonate that is quite similar to the wild-type enzyme (33).³ The molecular basis for these differences is not understood. Another difference between the two enzymes is the ability of aspirin-acetylated COX-2 to oxygenate arachidonic acid to 15-(R)-hydroxyeicosatetraenoic acid (37, 38). Acetylated COX-1 is unable to carry out this transformation. The importance of the channel at the top of the active site for binding the -end of arachidonate was recently confirmed by mutagenesis of Gly-533. Bulky substitutions at this position inhibit the oxidation of arachidonic acid but not fatty acids with three less carbons at their -end.⁴

	Cyclooxygenase Inhibition

TOP INTRODUCTION Cyclooxygenase Catalysis Cyclooxygenase Inhibition Future Directions REFERENCES

The anticipated pot-of-gold at the end of the rainbow awaiting a selective COX-2 inhibitor triggered a world class race to develop candidate drugs. Some of the efforts have been been successful and are chronicled in several recent reviews (39-42). The most extensively represented class of inhibitors is diarylheterocycles; other classes of inhibitors include acidic sulfonamides, indomethacin analogs, zomepirac analogs, and di-t-butylphenols. All appear to be slow, tight binding inhibitors in which the selectivity is manifest in the second step (Equation 2) (43). This slow step involves the conversion of the (E·I) complex to the (E*·I) complex, in which the inhibitor is bound more tightly to the enzyme. Formation of the E*·I occurs in seconds to minutes and may reflect the induction of a subtle protein conformational change. The time-dependent change is not associated with covalent modification for most inhibitors. Aspirin is the only COX inhibitor that covalently modifies the protein (44). It acetylates Ser-530, which is juxtaposed to Arg-120, and it is more potent against COX-1 than COX-2 (34, 45). Recently, an aspirin-like molecule (acetoxyphenylheptynyl sulfide) was developed that exhibits 20-fold selectivity for COX-2 and acetylates only Ser-530 (46, 47).

View larger version (4K): [in this window] [in a new window]   Equation 2.

Crystal structures of complexes of sheep COX-1, mouse COX-2, and human COX-2 with non-selective and with selective inhibitors have been solved at 3-3.5-Å resolution (16-18). Carboxylic acid-containing NSAIDs ion pair with the guanidinium group of Arg-120, which also ion pairs with the carboxylate of arachidonate (18, 48, 49). Site-directed mutagenesis of the arginine residue in COX-1 to glutamine or glutamate renders the protein resistant to inhibition by carboxylic acid-containing NSAIDs (33, 50). Arg-120 is part of a hydrogen-bonding network with Glu-524 and Tyr-355, which stabilizes substrate/inhibitor interactions and closes off the upper part of the cyclooxygenase active site from the spacious opening at the base of the channel, which we call the lobby (Fig. 4) (16). Disruption of this hydrogen-bonding network opens the constriction and enables substrate/inhibitor binding and release to occur (17). In addition, Tyr-355 sterically hinders the mouth of the COX active site, which accounts for the preferential inhibition exhibited by S-stereoisomers of -methyl-substituted arylalkanoic inhibitors (e.g. the profens) (51). Opening and closing of the Arg-120-Glu-524-Tyr-355 constriction may contribute to the time dependence of all COX inhibitors.

View larger version (34K): [in this window] [in a new window]   Fig. 4.   A, stereo drawing of the active site of a complex of murine COX-2 with the diarylheterocycle SC558. B, stereo drawing of the active site of a complex of human COX-2 with an acylsulfonamide derivative of zomepirac. In both frames, the heme prosthetic group is shown in red, and the bound drug is shown in lavender. The highlighted residues include Arg-120 (yellow), Tyr-355 (white), Tyr-385 (orange), Arg-513 (blue), Val-523 (green), and Glu-524 (red). The zomepirac derivative is colored lavender.

Structures of COX-inhibitor complexes presumably reflect the (E*·I) complex in which the inhibitor is bound tightly to the enzyme. That COX-2-selective inhibitors, especially diarylheterocycles, bind to regions accessible in COX-2 but not COX-1 is consistent with the hypothesis that these structures reveal the molecular basis for their selectivity. Fig. 4A demonstrates that the sulfonamide moiety in SC-558 wedges into a hydrophobic "side pocket" of COX-2 bordered by Val-523 (18) and hydrogen bonds to Arg-513 and the peptide bond of Phe-518 (18). A similar hydrophobic side pocket off the main channel in COX-1 is not accessible because of the presence of an isoleucine instead of valine at position 523, which sterically hinders inhibitor approach. Other changes between COX-2 and COX-1 that contribute to rigidification of this side pocket include the substitutions R513H and V434I. The COX-2 mutant V523I is resistant to time-dependent inhibition by diarylheterocycles but not arylalkanoic acid-type NSAIDs (52-54). Conversely, the COX-1 mutant I523V is sensitive to time-dependent inhibition by diarylheterocycles (53). Movement of Val-523 and insertion of the sulfonamide or methylsulfone moiety into the side pocket may contribute to the time dependence of inhibition by diarylheterocycles.

The structural basis for selectivity by indomethacin analogs is dependent on substitutions at the top rather than the side of the cyclooxygenase active site. The structure of human COX-2 complexed with a 4-bromobenzyl indomethacin analog reveals that the 4-bromobenzyl group is in van der Waals contact with Leu-503 at the apex of the COX-2 active site.⁵ Position 503 in COX-1 is substituted with phenylalanine, which is not as easily displaced by the bromobenzyl group as leucine. Decreased flexibility at the top of the COX-1 active site may reduce the affinity of the protein for the indomethacin analog, thereby accounting for its COX-2 selectivity.

Not all structures of COX-2 inhibitor complexes provide insights into the mechanism of selectivity. The NS-398·COX-2 complex demonstrates that the sulfonamide group ion pairs to Arg-120 in a manner similar to carboxylate-containing NSAIDs rather than inserting into the Val-523 side pocket (55). However, the reason that NS-398 preferentially interacts with COX-2 is not evident from the structure. Likewise, crystallographic analysis of zomepirac-derived COX-2-selective inhibitors does not provide a rationale for their selective COX-2 inhibition. The zomepirac-acylsulfonamide inhibitor breeches the constriction at the mouth of the COX active site and projects into the sterically uncongested lobby region (Fig. 4B) (17). The sulfonamide moiety of the inhibitor hydrogen bonds to Arg-120, Glu-524, and Tyr-355 in COX-2 in a manner similar to the arylalkanoic acid inhibitors. Because these three residues are common to COX-1 and COX-2, their importance in determining the selectivity of this inhibitor for COX-2 is uncertain.

	Future Directions

TOP INTRODUCTION Cyclooxygenase Catalysis Cyclooxygenase Inhibition Future Directions REFERENCES

The cyclooxygenase mechanism represents a beautiful marriage of peroxidase chemistry and free radical chemistry. The involvement of protein radicals in arachidonate oxidation is now generally accepted, but there is a dearth of details on the electron transfer that generates them. Furthermore, there are proposals of non-peroxidatic mechanisms for tyrosyl radical generation that need to be explored (56).

A static picture of substrate-protein and inhibitor-protein interactions is emerging from crystallographic and mutagenesis studies, but the dynamics of these interactions is poorly understood. Fluorescence quenching should be especially useful for studying inhibitor-COX interactions in real time, and preliminary studies indicate diarylheterocycle binding is more complex than the simple two-step model suggests (57).

Two selective COX-2 inhibitors are now on the market and if they are as safe in the general population as they have been in clinical trials (11), they will represent a major public health advance by reducing mortality from bleeding ulcers. Equally exciting is the potential selective inhibitors and COX knockout mice have for uncovering this enzyme's involvement in a wide range of physiological and pathophysiological responses in human beings (58). In addition, it is likely that a broader range of structures of COX-2 inhibitors will emerge over the next few years as exemplified by a survey of the recent patent literature.⁶

Finally, although this review has not covered regulation, this topic remains an extremely important and exciting area of investigation. Despite intense study, the reason for the existence of distinct COX genes (sometimes expressed in the same cell type) is not understood; the source of arachidonic acid for the two enzymes remains incompletely defined, and the mechanism by which glucocorticoids inhibit COX-2 expression is not certain. Elucidating the answer to these questions not only will provide important fundamental knowledge but also may uncover new molecular targets for pharmacological intervention against a range of human diseases.

	ACKNOWLEDGEMENTS

We are grateful to R. Kurumbail and M. Browner for COX-2-inhibitor coordinates and helpful discussions; to B. McKeever and W. Smith for communication of submitted manuscripts; and P. Isakson for helpful discussions.

	FOOTNOTES

^* This minireview will be reprinted in the 1999 Minireview Compendium, which will be available in December, 1999. This is the first article of five in "A Thematic Series on Oxidation of Lipids as a Source of Messengers." Work in this laboratory is supported by National Institutes of Health Grant CA47479.

Because of space limitations, many pertinent references were not cited in the body of the text. A supplementary reference list is provided in the on-line version of this article (available at http://www.jbc.org) that contains citations relevant to the following topics: discovery and identification of prostaglandins; biological responses linked to prostaglandin biosynthesis; chemical precedents for free radical mechanisms of oxygenation; peroxidase activation of cyclooxygenase activity; structures of COX-2 inhibitors; and baculovirus expression of COX-2.

To whom correspondence should be addressed. Tel.: 615-343-7329; Fax: 615-343-7534; E-mail: marnett@toxicology.mc.vanderbilt.edu.

^§ Present address: Depts. of Pharmaceutical Chemistry and Biochemistry and Biophysics, University of California, San Francisco, CA 94143-0446.

² The term cyclooxygenase is used to describe the enzyme activity or to refer to the active site for that activity on the protein.

³ Rieke, C. J., Mulichak, A. M., Garavito, R. M., and Smith, W. L. (1999) J. Biol. Chem. 274, 17109-17114.

⁴ Rowlinson, S. W., Crews, B. C., Lanzo, C. A., and Marnett, L. J. (1999) J. Biol. Chem., in press.

⁵ B. M. McKeever, S. R. Pandya, M. D. Percival, M. Ouellet, C. Bayly, G. P. O'Neill, L. Bastien, B. P. Kennedy, M. Adam, W. Cromlish, P. Roy, W. C. Black, D. Guay, and Y. Leblanc, submitted for publication.

⁶ Information regarding compounds for which patents have been filed recently is available at the IBM patent server web site. The reference numbers for chromenes, pyranoindoles, and substituted phenylacetic acids are PCT WO9847890, US 5776967, and PCT WO9709977, respectively.

	ABBREVIATIONS

The abbreviations used are: COX, cyclooxygenase protein or the gene that codes for it; NSAIDs, non-steroidal anti-inflammatory drugs; PGG₂, prostaglandin G₂.

	REFERENCES

TOP INTRODUCTION Cyclooxygenase Catalysis Cyclooxygenase Inhibition Future Directions REFERENCES