REVIEWS
`
`THE DEVELOPMENT OF COX2
`INHIBITORS
`
`Rod J. Flower
`
`Aspirin, arguably the world’s favourite drug, has been around since the late nineteenth century,
`but it wasn’t until the late 1970s that its ability to inhibit prostaglandin production by the
`cyclooxygenase enzyme was identified as the basis of its therapeutic action. Early hints of a
`second form of the cyclooxygenase that was differentially sensitive to other aspirin-like drugs
`ultimately ushered in an exciting era of drug discovery, culminating in the introduction of an
`entirely new generation of anti-inflammatories. This article reviews the story of this discovery and
`looks at the future of cyclooxygenase pharmacology.
`
`ISOZYME (ISOENZYME)
`One of several forms of an
`enzyme in an individual or
`population that catalyse the
`same reaction but differ from
`each other in such properties
`as substrate affinity and
`maximum rates of
`enzyme–substrate reaction.
`
`ANTIPYRETIC
`Describes the fever-suppressive
`activity of a drug.
`
`Department of Biochemical
`Pharmacology,
`The William Harvey
`Research Institute,
`Queen Mary University
`of London,
`Charterhouse Square,
`London EC1M 6BQ, UK.
`e-mail:
`r.j.flower@mds.qmul.ac.uk
`doi:10.1038/nrd1034
`
`C A S E H I S TO R Y
`
`The league table listing the ‘top 20’ drugs includes
`rofecoxib (Vioxx) and celecoxib (Celebrex), two
`inhibitors of the prostaglandin-forming cyclooxygenase
`(COX) (FIG. 1), which between them commanded sales
`in excess of US $4 billion in the year 2000 (REF. 1). This
`statistic might seem surprising — after all, the thera-
`peutic use of COX inhibitors has a venerable history
`dating back to the introduction of aspirin in 1898 (or
`even earlier if the use of salicylate-containing plant
`extracts is included) and, since then, the field has a
`record of almost continuous development. The 1940s
`saw the introduction of phenylbutazone, the fenamates
`appeared in the 1950s, indomethacin in the 1960s, the
`proprionates in the 1970s and the oxicams in the 1980s.
`With such a long record of drug discovery in the area
`and such a vast range of drugs to choose from, one
`might be forgiven for thinking that the wellspring of
`chemical innovation that nurtured the field so effi-
`ciently over the past century must have long since run
`dry — or at least begun to falter — and that the current
`market would have little room for new versions of what
`would seem to be a rather tired and well-worn formula.
`However, the discovery in the early 1990s of a second
`COX ISOZYME revitalized the field and stimulated a hunt
`for new and selective isoform inhibitors. This culmi-
`nated in the introduction of Vioxx and Celebrex in
`Europe and North America within a mere ten years, and,
`at the same time, brought a fresh perspective on the
`
`unusual therapeutic profile of several existing non-
`steroidal anti-inflammatory drugs (NSAIDs). But we are
`getting ahead of ourselves — how did this idea of the
`second isoform come about in the first place and what
`are the therapeutic advantages of these new inhibitors
`over the many older drugs which, after all, have seen
`valiant clinical service over the decades? To answer these
`questions, we must return to the 1970s when much of
`the seminal work on COX inhibition was published.
`
`Early work
`Although the NSAIDs do not reverse the course of
`systemic diseases such as arthritis, they form the main-
`stay symptomatic treatment of many inflammatory
`disorders and soft-tissue injuries. Approximately 50
`NSAID preparations are listed in Monthly Index of
`Medical Specialities and, as a class, these are among the
`most commonly prescribed drugs. In the United
`Kingdom, for example, recent data (1999) indicate that
`18.5 million prescriptions are written for NSAIDs each
`year at a cost of £170 million — and this figure doesn’t
`take into account the over-the-counter sales, which are
`considerable. Aspirin itself is still consumed in prodi-
`gious amounts around the world and new uses are con-
`tinually being found for this drug.
`NSAIDs are sometimes known as the aspirin-like
`drugs because they have an activity profile that is
`broadly similar to that of aspirin. That is, they all possess
`
`NATURE REVIEWS | DRUG DISCOVERY
`
`VOLUME 2 | MARCH 2003 | 1 7 9
`
`© 2003 Nature Publishing Group
`
`Page 1 of 13
`
`Patent Owner Ex. 2060
`Mylan v. Pozen
`IPR2017-01995
`
`

`

`R E V I E W S
`
`Mitogens, cytokines and many other
`stimuli acting through the MAPK pathway
`
`+
`
`Other esterified
`arachidonate
`
`Esterases
`
`Free
`arachidonate
`
`Phospholipases,
`especially cPLA2
`
`Phospholipid
`arachidonate
`
`PGG2
`
`PGH2
`
`Cyclooxygenase
`complex
`
`Thromboxane
`synthase
`
`Prostacyclin
`synthase
`
`Reductase
`
`S
`
`Isomerase
`
`S
`
`S
`
`Isomerase
`
`TXA2
`
`S
`
`TXB2
`
`PGI2
`
`S
`
`6-keto
`PGF1α
`
`PGF2α
`
`PGE2
`
`PGD2
`
`13,14-reductase; 15-dehydrogenase
`
`13,14-dihydro; 15-keto metabolites
`
`β-and ω-oxidation
`
`Urinary metabolites
`
`Figure 1 | An overview of prostaglandin synthesis and metabolism. In theory, free fatty acids
`such as arachidonate can be formed from several sources, although phospholipid-bound
`arachidonate is probably the most significant pool. Phospholipases, especially cytosolic
`phospholipase A2 (cPLA2), are highly-regulated enzymes (by the MAP kinase (MAPK) pathway) that
`liberate arachidonate, which is then transformed by the cyclooxygenase (COX) complex. The
`mechanism of this reaction is complex — 2 moles of molecular oxygen are introduced sequentially
`by a lipoxygenase reaction, followed by a COX reaction. This generates PGG2, a 15-hydro-peroxide
`prostaglandin that is reduced to PGH2, the corresponding hydroxy product. Both of these
`intermediates are short-lived but may have independent biological activity. However, in most cases
`it is not yet clear whether this independent activity is significant in vivo. A further battery of enzymes
`transform PGH2 into a variety of products. Some of these enzymes, such as thromboxane (TX)
`synthase and prostacyclin (PGI2)synthase, show marked tissue localization (for example, TX
`synthase in platelets and PGI2 synthase in vascular endothelium). Other enzymes, such as the
`endoperoxide reductases and isomerases, are quite widely distributed, although in some cases they
`can be induced following inflammatory stimuli. In vitro at least, the endoperoxides PGG2 and PGH2
`can also decay spontaneously (S) to PGF2α, PGE2 and PGD2. The evanescent products, TXA2 and
`PGI2, decay spontaneously to their respective inactive metabolites, TXB2 and 6-keto PGF1α. The
`biological activity of the other prostaglandins is curtailed following uptake into cells, by a series of
`metabolic enzymes that are present in some tissues (for example, the lung) at high concentrations.
`Inactive metabolites of these prostanoids undergo carbon chain shortening (especially in the liver)
`prior to secretion in the urine, in which estimates of total body prostaglandin turnover can be made
`by selectively monitoring these products. The most significant products from the point of view of this
`review are PGE2, because of its importance in inflammation, fever and pain; PGI2 because of its anti-
`aggregatory action and possible role in hyperalgesia, and TXA2 because of its important role in
`platelet aggregation. Colour code: red, precursors; orange, intermediates; yellow, ‘primary’
`prostaglandins that mediate most of the biological activity of this system; green, inactive or largely
`inactive metabolites; brown, end metabolites that are excreted primarily in the urine.
`
`analgesic, anti-inflammatory and ANTIPYRETIC properties
`to some degree, and produce characteristic side effects,
`including gastric intolerance and depression of blood
`clotting through inhibitory action on platelet function.
`As a group, the NSAIDs are structurally diverse, with
`most (but not all) being carboxylic acids (FIG. 2). The
`main question, from the pharmacologist’s point of view,
`was how these apparently disparate therapeutic and side
`effects were mechanistically linked.
`There were several early suggestions (REFS 2,3), but
`the real breakthrough came in 1971. Vane tells us4 that
`the idea that the aspirin-like drugs blocked the conver-
`sion of substrate arachidonic acid to prostaglandins
`came to him while reviewing experiments in which
`aspirin blocked the release of ‘rabbit aorta contracting
`substance’ (RCS) from guinea-pig and dog lung.
`Believing that RCS was an intermediate in prostaglandin
`synthesis, he wrote “…a logical corollary was that
`aspirin might well be blocking the synthesis of
`prostaglandins”. A quartet of papers appeared that year
`from Vane and members of his group, showing that
`aspirin itself, indomethacin and (less effectively) sodium
`salicylate blocked prostaglandin synthesis in a cell-free
`system5 and in isolated perfused spleen of dogs6. In
`humans, therapeutic doses of aspirin taken by volun-
`teers reduced prostaglandin generation by aggregating
`platelets ex vivo 7 or in seminal plasma samples collected
`during the course of the treatment 8. The overall message
`was clear — at least some NSAIDs were able to prevent
`the generation of prostaglandins by direct action on the
`COX enzyme, and did so in humans in clinical doses.
`But how was this linked to their therapeutic actions?
`The late 1960s and early 1970s had seen an explosion
`of interest in the biology of the prostaglandins. It had
`already been shown that prostaglandins are generated
`during platelet aggregation to produce fever, HYPERALGESIA
`and inflammation (reviewed by Willis9). Prostaglandins
`had also been detected in gastric mucosa and been
`shown to inhibit ulcer formation in rodents10. In other
`words, the ability of NSAIDs to block COX provided the
`much sought after link between the therapeutic and side
`effects of these drugs.
`Over the next couple of years, several other important
`findings emerged. Significantly, it was shown that the
`entire gamut of NSAIDs inhibit COX at concentrations
`well within their therapeutic plasma range and that the
`overall order of potency corresponded to their thera-
`peutic activity11. Other types of anti-inflammatories,
`such as the glucocorticoids and the so-called disease-
`modifying drugs, were inactive in these cell-free assays,
`providing further evidence for the specificity of the effect.
`Using ENANTIOMERIC pairs of NSAIDs, such as naproxen12,
`an exquisite correlation was observed between the anti-
`inflammatory and anti-COX activity, and many more
`studies confirmed the notion that this is a fundamental
`mechanism of this class of drug (reviewed in REF. 13).
`By 1974, Vane’s concept was firmly established.
`Researchers were quick to seize upon the fact that
`NSAIDs could be used to probe the functions of
`prostanoids in physiology and pathology, thereby
`opening an entirely new chapter in eicosanoid research.
`
`180 | MARCH 2003 | VOLUME 2
`
`www.nature.com/reviews/drugdisc
`
`© 2003 Nature Publishing Group
`
`Page 2 of 13
`
`Patent Owner Ex. 2060
`Mylan v. Pozen
`IPR2017-01995
`
`

`

`CO2H
`
`O
`
`O
`
`C CH3
`
`CO2H
`
`OH
`
`O
`
`N
`
`N
`
`O
`
`CH3
`
`R E V I E W S
`
`O
`
`N
`
`N
`
`O
`
`CH3
`
`Aspirin
`(Acetylsalicylic acid)
`
`Salicylic acid
`
`Phenylbutazone
`
`HO
`Oxyphenbutazone
`
`CH3
`
`CH3
`
`HN
`
`CO2H
`
`FF
`
`CF
`
`HN
`
`CO2H
`
`FF
`
`F
`C
`
`Cl
`
`HN
`
`Cl
`
`CO2H
`
`Meclofenamic acid
`
`Flufenamic acid
`
`Mefenamic acid
`
`CO2H
`
`CH3
`
`H3C
`
`CH3
`
`CH3
`CHCO2H
`
`F
`
`Ibuprofen
`
`Flurbiprofen
`
`CH3
`CHCO2H
`
`CH3
`CHCO2H
`
`O
`
`Ketoprofen
`
`CH3O
`
`Naproxen
`
`O
`
`H3C
`
`CO2H
`
`CH3
`
`N
`
`C O
`
`Cl
`Indomethacin
`
`CH2CO2H
`
`NH
`
`Cl
`
`Cl
`
`O
`
`S
`
`O
`
`N
`
`CH3
`
`CONH
`
`N
`
`OH
`
`Diclofenac
`
`Piroxicam
`
`Figure 2 | Chemical structures of NSAIDs and related compounds. Structures of some ‘classical’ NSAIDs, including
`representative salicylates, pyrazolones, fenamates, proprionates, oxicams and indomethacin. Note the general presence of
`a carboxylic-acid moiety.
`
`Parenthetically, one might add that the pharmaceutical
`industry now possessed, probably for the first time, a
`simple and robust in vitro technique to screen com-
`pounds for putative anti-inflammatory activity. This
`in itself was a significant advance, and the number of
`chemical abstracts dealing with potential inhibitors of
`the COX enzyme rose markedly, with more than 2,500
`per year recorded within a decade of these ideas
`taking hold.
`But right from the beginning of the story, several
`anomalies were noted. The most relevant concerns
`paracetamol (acetaminophen) — another hugely popu-
`lar drug that was also introduced in the 1890s (although
`it was not in common use until some 40 years later). As
`for all other aspirin-like drugs, paracetamol possessed
`antipyretic and analgesic activity but, unlike most, had
`little anti-inflammatory activity and caused virtually no
`gastric or platelet side effects. Apparently in accord with
`its therapeutic profile, paracetamol was found to have a
`different pattern of inhibitory activity, being more effec-
`tive against brain COX than enzyme prepared from
`peripheral tissues such as the spleen14. At the time, this
`was put forward as a putative explanation for the selec-
`tivity of its therapeutic action, and the idea that there are
`several forms of the enzyme was formulated. Wide
`
`variations in the inhibitory potency of indomethacin
`against COX enzymes prepared from a range of tissues
`was subsequently reported15 and the isoenzyme idea
`was further elaborated in several reviews13,16.
`
`The development of the field
`Despite the early indications for alternative forms of
`COX, little concrete evidence emerged to support this
`idea for several years. The structural features of COX,
`which has dual hydroperoxidase and cyclooxygenase
`activity17, were not well understood. With the notable
`exception of bovine or ovine seminal vesicles, COX is
`usually expressed in tissues in low abundance. As a
`dimeric membrane-bound protein, it posed many
`challenges to purification and was not sequenced until
`1988 (REFS 18,19). Other factors complicating the inter-
`pretation of potency differences included the wide
`variations in assay conditions that were used by differ-
`ent groups — which is still a problem today — as well
`as differences in the kinetics of the COX inhibitors,
`some of which produced ‘competitive reversible’ inhi-
`bition, whereas others exhibited unusual inhibitory
`effects such as the ‘competitive non-reversible’ mecha-
`nism that is observed in the case of indomethacin20.
`The discovery21,22 that, alone among the group, aspirin
`
`HYPERALGESIA
`An abnormal state of increased
`sensitivity to painful stimuli.
`
`ENANTIOMERS
`A pair of compounds whose
`molecular structures are mirror
`images of each other.
`
`NATURE REVIEWS | DRUG DISCOVERY
`
`VOLUME 2 | MARCH 2003 | 1 8 1
`
`© 2003 Nature Publishing Group
`
`Page 3 of 13
`
`Patent Owner Ex. 2060
`Mylan v. Pozen
`IPR2017-01995
`
`

`

`R E V I E W S
`
`CONTRALATERAL
`On or affecting the opposite side.
`
`HOUSEKEEPING GENE
`A gene that is usually expressed
`at a fairly constant rate in cells
`as it subserves some constant
`physiological requirement. By
`comparison, an inducible gene is
`one that normally appears at a
`very specific time and in
`response to a specific stimulus.
`
`irreversibly acetylated a serine residue (Ser530) within
`the COX active site to produce its effects, reinforced
`the opinion held by many workers at that time that the
`NSAIDs were a chemically heterogeneous group of
`drugs with widely differing modes of inhibitory action.
`In short, it seemed at least possible that some of these
`differences in inhibitor potency could be the result of
`factors other than the presence of isozymes.
`Meanwhile, other apparently unrelated investigations
`were to provide the backdrop for future dramatic revela-
`tions. Using a model of rabbit kidney inflammation
`induced by ligation of the urethra, Needleman’s group
`observed23 that the affected, but not the CONTRALATERAL,
`organ unexpectedly developed an enormous capacity to
`generate prostaglandins. In the following year24, the
`group showed that this increase was due to de novo
`synthesis of fresh enzyme, although no suggestion was
`made at this time that the new enzyme was a variant
`form. Nevertheless, this theme was pursued by the
`Needleman group over the next few years and, in 1982, a
`paper appeared that suggested the presence of two dis-
`tinct forms of COX in brain tissue with differing sensitivi-
`ties to indomethacin25. Other studies in gastrointestinal
`tissue were also supportive of the selectivity of action of
`NSAIDs in different tissues26.
`The problem was taken up again by several labora-
`tories towards the end of the 1980s. Treatment of vas-
`cular smooth muscle cells with epidermal growth factor
`(EGF)27 or fibroblasts, and monocytes with pro-
`inflammatory stimuli such as interleukin-128,29 or
`lipopolysaccharide30, induced COX mRNA and de novo
`synthesis of enzyme which, when partially sequenced,
`seemed to be identical to the bovine seminal vesicle
`COX. In one paper31, Needleman’s group wrote “Clearly,
`those putative enzyme pools may arise as different gene
`products, possibly through the expression of different
`COX genes…”. Interestingly, glucocorticoids inhibited
`the induction of this new COX protein without altering
`the amount of enzyme that was constitutively present
`in cells32. Elsewhere, further evidence emerged indi-
`cating different forms of the enzyme. For example,
`Wong and Richards, using immunological techniques,
`reported two isoforms in the rat ovary, one of which
`was regulated by hormones33.
`
`The discovery of COX2
`Paradoxically, the next significant step forward came in
`1991 from workers in an entirely different field. While
`investigating the expression of early-response genes in
`fibroblasts transformed with Rous sarcoma virus,
`Simmons and his colleagues34 identified a novel mRNA
`transcript that coded for a protein that had a high
`sequence similarity, but was not identical, to the seminal
`vesicle COX enzyme. The suggestion was that a COX
`isozyme had been discovered. Independent and simulta-
`neous confirmation of this exciting finding came from
`the laboratory of Herschman and colleagues, who dis-
`covered a novel cDNA species encoding a protein with a
`predicted structure similar to COX1 while studying
`phorbol-ester-induced genes in Swiss 3T3 cells35. The
`same laboratory subsequently showed that this gene
`
`product was indeed a novel COX36 and that its induction
`was inhibited by dexamethasone37. Similarly indicative
`findings were also reported in mouse fibroblasts38, cul-
`tured rat mesangial cells39, RAW 264.7 cells40, rat alveolar
`macrophages41,42, the ovary43,44 and other cell types45.
`But was this enzyme physiologically relevant?
`Needleman’s group conclusively identified their
`inflammation-inducible form of COX as the species
`that both Simmons and Herschman had cloned32. As
`the evidence for the relevance of the two enzymes
`mounted, they were renamed COX1, referring to the
`original enzyme isolated from seminal vesicles and sub-
`sequently found to be distributed almost ubiquitously,
`and COX2, denoting the ‘inducible’ form of the enzyme
`(although it was expressed basally in the brain46). The
`two genes had a different chromosomal organization in
`rodents47 and humans48, and COX1/COX2 mRNA was
`differentially expressed49 in human tissues. Promoter
`analysis confirmed a fundamental difference between
`the two isozymes, with the COX2 promoter containing
`elements strongly reminiscent of those genes that are
`switched on during cellular stress and switched off by
`glucocorticoids50, whereas COX1 had the appearance of
`a ‘HOUSEKEEPING’ GENE. Histological and other studies con-
`firmed this apparent division of labour between the
`two enzymes, and COX1 seemed to be the predomi-
`nant isoform in healthy gastrointestinal tissue from rat,
`dog and monkey51.
`These facts begged a key question — if COX2 was
`the predominant inflammatory species, shouldn’t this
`be the target for NSAIDs? If this was so, the corollary
`was surely that COX2 inhibition was the true thera-
`peutic modality of NSAIDs, whereas COX1 inhibition
`might be the cause of side effects such as gastric irrita-
`tion and depression of platelet aggregation. This was
`the notion put forward by more than one group52–54,
`which came to be known in a rather Orwellian way as
`the ‘COX2-bad:COX1-good’ hypothesis. If these
`propositions were true, then a selective COX2
`inhibitor would be an ideal drug, possessing the anti-
`inflammatory action but lacking the gastric and other
`side effects. But would it be possible to find or design
`such a drug?
`Some encouragement for this idea could be
`derived from publications of the day52,53. Most of the
`non-steroidal drugs available at that time actually
`inhibited both enzymes to a greater or lesser degree,
`but some selectivity of action was seen with experi-
`mental drugs such as 6-MNA and BF389. But other
`data that subsequently emerged from transgenic
`approaches were less encouraging. Cox1-null (REF. 55)
`and Cox2-null (REF. 56) mice did not behave exactly as
`expected — Cox1-deficient mice did not have sponta-
`neous stomach ulcers and, whereas the ulceration
`caused by indomethacin in these animals was less than
`in the wild types, it was still very much in evidence.
`Again contrary to theoretical predictions, homo-
`zygous mutant mice lacking Cox2 showed a normal
`(albeit acute) inflammatory response when treated
`with several agonists, apparently contradicting the
`idea that Cox2 was the predominant enzyme in
`
`182 | MARCH 2003 | VOLUME 2
`
`www.nature.com/reviews/drugdisc
`
`© 2003 Nature Publishing Group
`
`Page 4 of 13
`
`Patent Owner Ex. 2060
`Mylan v. Pozen
`IPR2017-01995
`
`

`

`R E V I E W S
`
`O
`
`O
`
`O
`
`CH3
`
`O
`
`N
`S
`O
`CGP28238
`
`SO2NH2
`
`O
`
`H3C
`
`S
`
`O
`
`Rofecoxib
`
`CH3
`
`N
`
`CH3
`
`CH3
`
`O
`
`N
`
`N
`H3C
`
`SO2CH3
`
`F
`
`O
`
`F
`
`H
`
`F
`
`DuP697
`
`F
`
`Br
`
`S
`
`NHSO2CH3
`
`O F
`
`O
`
`Flosulide
`
`Phenazone
`(antipyrine)
`
`Dipyrone
`(Methampyrone)
`
`Amidopyrine
`(aminopyrine)
`
`Figure 3 | Chemical structures of NSAIDs and related compounds. a | Structures of DuP697, NS398 and other similar
`compounds. b | Selective COX2 inhibitors that were discovered as a result of a search for selective isoform inhibitors (celecoxib and
`rofecoxib) or that were ‘revealed’ as being COX2 selective (meloxicam, etodolac and nimesulide). c | Structures of some
`compounds that are more effective inhibitors of COX3 according to Simmons124.
`
`inflammation56. From the medicinal chemist’s point
`of view, there was also another depressing fact — the
`sequence of the catalytic domains of the two isozymes
`was so similar that the prospect of finding a specific
`inhibitor seemed remote. Fortunately, this perception
`was soon to be changed in a rather dramatic way.
`
`‘Sleepers’
`DuP697 was a drug reported in 1990 by the Dupont
`Company to be an effective anti-inflammatory agent
`with reduced ulcerogenic properties57. Significantly,
`DuP697 showed only feeble activity in in vitro assays
`of COX using seminal vesicle or rat kidney prepara-
`tions (known to predominantly contain COX1), but
`was more effective against rat brain prostanoid syn-
`thesis. The authors originally attributed the gastro-
`intestinal safety of this compound to its chemical
`structure (a non-acidic thiophene) (FIG. 3a), which was
`presumed to have different pharmacokinetic proper-
`ties to other COX inhibitors, most of which were car-
`boxylic acids. Reports of other compounds with simi-
`lar properties were published later, including some
`experimental compounds such as NS398, flosulide
`and CGP28238 (REF. 58) (FIG. 3a). However, in light of
`the discovery of COX2, it was not long before alert phar-
`macologists realized that these drugs might have this
`unusual behaviour because they acted predominately on
`the COX2 isozyme.
`
`The moment of this Damascene revelation cannot
`be pinpointed precisely but, in the context of future
`drug development, two ‘prostaglandin’ meetings in
`1992 (in Keystone, Colorado, in January, and Montreal
`in July) seem to have been of particular significance. At
`the first meeting, the Dupont Group presented evi-
`dence that gastric tolerance to DuP697 might be
`accounted for by differential inhibition of COX
`enzymes. At the second meeting, workers from Taisho
`described similar data with another structurally unre-
`lated compound, NS398 (REFS 59,60). These meetings
`were attended by many industrial scientists, and it
`seems clear from published accounts that these events
`initiated — or at least greatly accelerated — many in-
`house ‘COX2’ programmes. It is also clear that the
`structures of DuP697 and NS398 were crucial starting
`points in the hunt for new selective inhibitors. Two
`companies who capitalized on these leads were Searle
`Monsanto (now Pharmacia) and Merck. The former
`focused on sulphonamide-substituted 1,5-diaryl pyra-
`zole compounds, whereas Merck scientists settled on a
`series of methylsulphonylphenyl compounds.
`While medicinal chemistry programmes were being
`developed, the field continued to produce data that were,
`on the whole, supportive to the COX1/COX2 concept.
`For example, COX2 was found in human SYNOVIAL TISSUE
`taken from patients with rheumatoid arthritis, and
`the ‘inducibility’ of this enzyme by cytokines was
`
`SYNOVIAL TISSUE
`The tissues that surround
`the joints.
`
`NATURE REVIEWS | DRUG DISCOVERY
`
`VOLUME 2 | MARCH 2003 | 1 8 3
`
`© 2003 Nature Publishing Group
`
`a
`
`b
`
`c
`
`NHSO2CH3
`O
`
`NO2
`
`NS398
`
`CH3
`
`S
`
`N
`
`OH
`
`CONH
`
`N
`
`O
`
`S
`
`O
`
`CH3
`
`Meloxicam
`
`NHCOCH3
`
`NHCOCH3
`
`OH
`
`Paracetamol
`
`OCH2H5
`Phenacetin
`
`NHSO2CH3
`O
`
`N
`
`N
`
`F 3C
`
`CH2
`CO2H
`C2H5
`C
`
`HN
`
`C2H5
`
`NO2
`
`Etodolac
`
`Nimesulide
`
`CH3
`Celecoxib
`
`Na
`
`OO
`
`S
`
`CH3
`O
`
`N
`
`CH3
`
`O
`
`N
`
`N
`H3C
`
`O
`
`N
`
`N
`H3C
`
`CH3
`
`Page 5 of 13
`
`Patent Owner Ex. 2060
`Mylan v. Pozen
`IPR2017-01995
`
`

`

`R E V I E W S
`
`CARRAGEENAN
`A sulphated cell-wall
`polysaccharide that is found in
`certain red algae, which
`contains repeating sulphated
`disaccharides of galactose, and
`sometimes anhydrogalactose,
`and is used to induce an
`inflammatory lesion when
`injected into experimental
`animals.
`
`DYSMENORRHOEA
`Painful menstruation, often
`associated with nausea,
`vomiting, headache and
`faintness. It is thought to be
`related to excessive
`prostaglandin production.
`
`confirmed61,62. Support for the COX1/COX2 concept
`also came from pharmacological studies. Masferrer et
`al.63 showed that the onset of inflammation in the rat air
`pouch correlated with the appearance of COX2 in the
`lesion, and that NS398 blocked the production of
`prostaglandins at this site without causing intestinal
`lesions or affecting gastric prostaglandin synthesis.
`Similar results were seen in various inflammatory models
`with an early Searle Monsanto selective inhibitor
`SC58125 (REFS 64,65). A number of other useful, selective
`COX2 inhibitors were soon described in the literature —
`SC558 (a celecoxib prototype) showed good efficacy in
`rodent models of inflammation, fever and pain, whereas
`the closely related SC560, a selective COX1 inhibitor, was
`ineffective66. Celecoxib itself (then known as SC58635)
`reversed CARRAGEENAN-induced hyperalgesia and local
`prostaglandin production67 in rats, and a related com-
`pound was active intrathecally68. Recombinant COX1
`and COX2 had by now been prepared, and the selectivity
`of DuP697 and NS398 had been confirmed with in vitro
`assay systems using these enzymes69,70.
`The solution of the crystal structures of COX1 in
`1994 (REF. 71) and COX2 in 1996 (REF. 72) made a sub-
`stantial impact on the field. In the former paper, atten-
`tion was drawn to the crucial roles of arginine 120
`(Arg120, which interacted with the carboxyl group of
`both substrate and inhibitors) and tyrosine 355
`(Tyr355) in determining the stereospecificity of NSAID
`binding (FIGS 4, 5). In the latter paper, the structure of
`COX2 was determined in the presence of several
`inhibitors, enabling the authors to deduce the confor-
`mational and other changes required for selective inhi-
`bition. Despite the great similarity in the sequence data,
`detailed examination of the structure of the catalytic
`sites revealed the substrate binding ‘channel’ in the two
`enzymes to be quite different. A single amino-acid
`change, from the comparatively bulky isoleucine (Ile)
`in COX1 to valine at position 523 in COX2 (equivalent
`to position 509 in COX1), and the conformational
`changes that this produced resulted in enhanced access
`to a ‘side pocket’ that allowed the binding of COX2-
`specific inhibitors by providing a docking site for the
`bulky phenylsulphonamide residue of drugs such as
`SC558 (FIG. 4). In an elegant demonstration of how
`crucial this single change was, Gierse et al.73 showed
`that mutation of this residue in COX2 back to Ile
`largely prevented selective inhibitors such as SC58125,
`SC236 and NS398 from working. Further work on the
`structural basis for inhibition of both isoforms delin-
`eated the role of Arg106 in the binding of substrate and
`certain NSAIDs74, as well as the role of Tyr355 located
`at the entrance of the active site of COX2 (REF. 75).
`These structural data also helped explain differences
`in the inhibitory kinetics of COX1 and COX2 with
`drugs such as DuP697 and NS398 (REF. 76). There are, as
`stated above, several distinct mechanisms by which
`COX1 inhibitors can inhibit the enzyme, but many are
`of the competitive reversible type. By contrast, the
`COX2 inhibitors are irreversible, time-dependent (in
`the context of enzyme kinetics) inhibitors, partly as a
`result of the binding of the sulphonamide (or related)
`
`moiety into the enzyme ‘side pocket’. In an analysis of
`the kinetic behaviour of several COX inhibitors, Gierse
`et al.77 subsequently discerned four separate modes of
`enzyme inhibition, ranging from the simple competi-
`tive inhibition of drugs such as ibuprofen, through the
`‘weak binding,
`time-dependent’ mechanism of
`naproxen and the oxicams and the ‘tight binding, time-
`dependent’ inhibition of indomethacin , to the covalent
`modification produced by aspirin.
`
`The development of the ‘coxibs’
`Encouraged by the ‘concept testing’ experiments with
`selective inhibitors, and armed with several solid leads
`and a clear idea of the nature of the binding site, devel-
`opment of this field was rapid. Celecoxib arose from the
`Searle Monsanto programme and showed marked selec-
`tivity for COX2 in vitro78. Preclinical studies with this
`compound revealed that the drug had good efficacy in
`rodent models of inflammation, fever and pain. Early
`human studies confirmed its effectiveness in the treat-
`ment of osteoarthritis, rheumatoid arthritis and post-
`surgical pain when tested in comparison with a placebo
`or with comparator NSAIDs such as naproxen, ibupro-
`fen and diclofenac. Crucially, there was also confirmation
`of the reduced incidence of platelet and gastrointestinal
`side effects (reviewed by Lefkowith79), and celecoxib was
`subsequently licensed in the United Kingdom for
`osteoarthritis and rheumatoid arthritis in 2000.
`From Merck’s methylsulphonylphenyl series came
`MK0966, later named rofecoxib. Once again, a series of
`preclinical studies confirmed the efficacy and gastro-
`intestinal safety of this compound in rodent models of
`inflammation, and early human studies established its
`clinical utility in fever and pain (dental, DYSMENORRHOEA
`and post-operative models) when compared with stan-
`dard NSAIDs such as ibuprofen or naproxen (reviewed
`by Morrison80). In osteoarthritis and rheumatoid arthri-
`tis, rofecoxib proved superior to placebo and comparable
`in efficacy to standard doses of other NSAIDs such as
`diclofenac. Rofecoxib was also found to have greater
`gastrointestinal safety and not to affect platelet aggrega-
`tion — it was licensed in the United Kingdom for
`osteoarthritis in 1999 and rheumatoid arthritis in 2001.
`But it was not just the quest for new drugs that was
`stimulated by the discovery of COX2. The enolcarbox-
`amide meloxicam (Boehringer Ingleheim), as well as the
`tetrahydropyranoindole etodolac (Wyeth/Shire), were
`drugs already under development at the time that the
`COX2 story ‘broke’, whereas a sulphonanilide drug,
`nimesulide (Helsinn), had been marketed in Europe
`since 1985 (FIG. 3b). In each case, clinical and experimen-
`tal evidence already indicated that these agents were dif-
`ferent from the other NSAIDs, especially in terms of
`their good gastrointestinal tolerance. Pharmacologists
`could now view these anomalous data with a vision
`greatly sharpened by the emerging COX2 concept, and
`these drugs turned out to be effective COX2
`inhibitors81–83. In an authoritative survey, Warner et al.84
`found, for example, that meloxicam and etodolac
`showed almost the same order of selectivity for
`COX1/COX2 as some of the newer agents.
`
`184 | MARCH 2003 | VOLUME 2
`
`www.nature.com/reviews/drugdisc
`
`© 2003 Nature Publishing Group
`
`Page 6 of 13
`
`Patent Owner Ex. 2060
`Mylan v. Pozen
`IPR2017-01995
`
`

`

`COX1
`
`COX2
`
`NSAID binding space
`
`R E V I E W S
`
`'Side pocket'
`
`Intracellular membrane
`
`F
`
`S
`
`Br
`
`SO2CH3
`
`Bulky grouping
`
`COX2 inhibitor
`DuP697
`
`F
`
`CHCO2H
`CH3
`
`COX1 inhibitor
`Flurbiprofen
`
`Figure 4 | Comparison of the NSAID binding sites of COX1 and COX2 after Browner. Schematic cartoon (modified from REF. 142),
`showing the differences in