Valdecoxib: Assessment of Cyclooxygenase-2 Potency and Selectivity

The mechanism of action of the nonsteroidal anti-inflammatory drugs (NSAIDs), as well the side effects, are explained by inhibition of prostaglandin (PG) synthesis by cyclooxygenase (COX) (Vane, 1971). The recent finding of a second COX isoform (COX-2) provided the basis for the discovery of anti-inflammatory drugs with improved safety. COX-1 is expressed in most tissues and cells and is abundant in the GI tract, kidney, and platelets. Prostaglandins formed by this enzyme are important for normal physiological function in these tissues. The second isoform, COX-2, is prominently expressed in inflamed tissues, where it produces proinflammatory prostaglandins (Mitchell et al., 1993; Masferrer et al., 1994; Seibert et al., 1994; Crofford, 1997) and to a lesser extent constitutively expressed in brain and kidney (Seibert et al., 1997). This suggested that COX-2 could provide a well defined molecular target for rational drug development, with the hypothesis that specific inhibitors of this enzyme may achieve anti-inflammatory and analgesic efficacy without affecting production of physiological PGs (Needleman and Isakson, 1997).

Testing schemes were developed that relied on assessing the ability of compounds to selectively inhibit COX-2 over COX-1. Initially, cells were utilized that expressed the proper isoform [for example, lipopolysaccharide (LPS) stimulated fibroblasts for COX-2 or platelets for COX-1]. Purified ovine COX-1 and COX-2 (Futaki et al., 1994) were also used in assays that followed the production of PGE₂ by ELISA or radiometric methods. Later, recombinant enzymes, expressed in mammalian cells (Meade et al., 1993) or expressed in insect cells and purified (Barnett et al., 1994; Gierse et al., 1995), were utilized in screening assays. These in vitro as-says provided a powerful means for assessing COX selectivity and potency and led to the discovery and clinical development of the first rationally designed COX-2-selective inhibitors, celecoxib (Penning et al., 1997) and rofecoxib (Black et al., 1999; Chan et al., 1999).

Recombinant enzymes have been useful for defining the kinetic mechanism of COX-2-selective inhibition and for structural studies of the molecular basis underlying this phenomenon. The results of these mechanistic experiments indicated that COX-2-selective inhibitors such as celecoxib weakly inhibit COX-1 in a competitive fashion but potently inhibit COX-2 through a time-dependent, slowly reversible mechanism (Gierse et al., 1999). The potent and specific time-dependent mechanism confers selectivity for COX-2 (Copeland et al., 1994; Gierse et al., 1995; Ouellet and Percival, 1995). The specificity for COX-2 demonstrated by diaryl heterocyclic inhibitors such as celecoxib is based on their interaction with a unique side pocket in the COX-2 active site (Kurumbail et al., 1996); a single amino acid difference within the catalytic sites of the two isoforms provides a major contribution to COX-2 specificity (Gierse et al., 1996).

Despite advances in understanding the molecular and kinetic bases for specificity, methods for evaluating in vitro activity on COX isoforms vary widely, leading to considerable differences in reported selectivity (Meade et al., 1993; Laneuville et al., 1994; Smith et al., 1994, 1995, 1997; Chan et al., 1995; Gierse et al., 1995; Jouzeau et al., 1997). This suggests the need for methods to clearly evaluate COX isoform specificity in vivo. The observation by Masferrer et al. (1994) that the inflamed rat air pouch produces PGs derived from COX-2, whereas gastric PG levels mirror COX-1 activity, suggests a means for directly evaluating selectivity in vivo.

In this report, we utilize in vitro and in vivo methodology to assess COX isoform selectivity and describe the pharmacological activity of valdecoxib, a rationally designed COX-2 inhibitor (Talley et al., 2000), as well as the other marketed COX-2-selective inhibitors, celecoxib, rofecoxib (Black et al., 1999), and etoricoxib (Riendeau et al., 2001) (Fig. 1).

Materials and Methods

Materials

Arachidonic acid, supplied as the sodium salt, was obtained from NuChek Prep (Elysian, MN); PGE₂ and thromboxane B2 (TxB₂) ELISA kits were obtained from Cayman Chemical (Ann Arbor, MI); celecoxib, valdecoxib, rofecoxib, etoricoxib, meloxicam, and SC-560 [5-(4-chlorophenyl)-1-(4-methoxyphenyl)-3-trifluoromethylpyrazole] were all prepared by the Amersham Biosciences Inc. Medicinal Chemistry Department (Piscataway, NJ); indomethacin, naproxen, ibuprofen, diclofenac, etodolac, nabumetone, piroxicam, and all standard buffer reagents were obtained from Sigma-Aldrich (St. Louis, MO).

In Vitro Potency and Selectivity

Human Recombinant Enzyme Assay. Compounds were evaluated for selectivity of inhibition in vitro using baculovirus-expressed recombinant human COX-1 and COX-2 enzymes as previously described (Gierse et al., 1995). For COX-2-selective inhibitors, IC₅₀s were generated from a mean of at least four separate determinations, and NSAIDs for comparison purposes only were tested once. Inactivation rate constants were determined as previously described (Walker et al., 2001) by measuring oxygen consumption directly with a Clark-style polarographic electrode. In addition, kinetic constants were determined by measuring the cyclooxygenase activity indirectly by utilizing tetramethyl-p-phenylenediamine as a cosubstrate with arachidonic acid by the method previously described (Gierse et al., 1999).

Radioligand Binding Assay. A direct binding assay that measures radiolabeled inhibitor binding to enzyme was developed to assess binding directly to COX-1 and COX-2 without the confounding influence of enzyme kinetics. The method is described in detail by Hood et al. (2003) and is summarized here. COX-1- or COX-2-specific monoclonal antibodies at 10 μg/ml in 100 mM NaHCO₃, pH 8.2 were coated (100 μl/well) onto 96-well Immulon-2 microtiter plates (Dynex Technologies Inc., Chantilly, VA) and incubated overnight at room temperature in a humidified chamber. The coated plates were washed with Dulbecco's phosphate-buffered saline (d-PBS; Invitrogen, Carlsbad, CA), without CaCl₂ and MgCl₂, pH 7.4 and then treated with a blocking reagent consisting of 10% skim milk in d-PBS (0.2 ml) for 90 to 120 min at 37°C to decrease nonspecific binding to the plate. The coated and blocked plates were washed, COX enzyme was added at 20 to 35 μg/ml in 50 μl of binding buffer (100 mM Tris and 1 μM hemin, pH 8.0), and then was incubated at room temperature for 60 to 120 min. Finally, these antibody-captured enzyme-coated plates were washed with d-PBS and aspirated to dryness immediately prior to the binding assay. To determine competitive binding with valdecoxib, various concentrations of the indicated compound were incubated with [³H]valdecoxib and allowed to compete for the binding to COX-2 for 120 min. K_i values were determined from logit-log transformations of the binding data. For determination of dissociation rates, ³H compounds were incubated with murine COX-2 immobilized on a 96-well plate for 120 min before excess cold inhibitor (10 μM) was added to initiate the dissociation time course. At the indicated times, the incubation was halted by aspiration, and the remaining radiolabeled inhibitor was released from the enzyme and counted. Specific binding of valdecoxib was 93%, celecoxib was 73%, and rofecoxib was 50%.

COX Inhibition in Human Whole Blood. The assay of Patrignani et al. (1994) was used to assess COX inhibition in human whole blood. For COX-2-selective inhibitors, IC₅₀s were generated from a mean of at least five separate determinations, and NSAIDs for comparison purposes only were tested once. To evaluate COX-1-mediated TxB₂ production, venous blood from healthy human donors was collected in tubes without anticoagulants and allowed to clot. Blood (0.5 ml) was incubated in 96-well culture plates for 1 h at 37°C with compound suspended in dimethylsulfoxide (DMSO; 0.4% final concentration), and the mixtures were incubated for 10 min at 37°C. The reaction was stopped by cold centrifugation at 800g for 10 min at 4°C to pellet the cells. The supernatants were recovered and diluted 1:200 in ELISA buffer for quantitation of TxB₂ by ELISA. Expected levels of TxB₂ formed in this assay are 50 to 100 ng/ml serum. Twelve concentrations of compound starting at 200 μM with 3-fold dilutions were examined in duplicate.

To evaluate compounds for COX-2-mediated PGE₂ production, venous blood from healthy human donors was collected in heparinized tubes. Blood (0.5 ml) was incubated in 96-well collection plates for 24 h at 37°C with 100 μg/ml LPS (Sigma L-2630; Sigma-Aldrich) and compound dissolved in DMSO (0.4% final concentration of DMSO). The reaction was stopped by cold centrifugation at 800g for 10 min at 4°C to pellet the cells. Plasma supernatant (40 μl) was precipitated with 4 volumes of methanol (160 μl) and spun at 800g for 10 min. Supernatants were recovered and diluted 1:50 in ELISA buffer for quantitation of PGE₂ by ELISA. Expected PGE₂ levels in this assay are approximately 15 ng/ml for unstimulated and 50 to 150 ng/ml for simulated plasma. Twelve concentrations of compound starting at 200 μM with 3-fold dilutions were examined in duplicate. IC₅₀ values were generated by fitting the data with a four-parameter logistic regression fit then determining the point that intersects 50% of the difference between negative (unstimulated) and positive (stimulated with LPS) uninhibited controls (control IC₅₀).

In Vivo Potency and Selectivity

Rat Air Pouch Model of Inflammation. Air pouches were produced by subcutaneous injection of 20 ml of sterile air into the intrascapular area of the back of male Lewis rats (175-200 g) (Charles River Laboratories, Inc., Wilmington, MA), six animals per dose group. Pouches were allowed to develop for 1 day. Animals were fasted with free access to water. Compounds or vehicle were administered by gavage 2 h prior to injection of 2 ml of a 1% suspension of carrageenan (Sigma-Aldrich) dissolved in saline into the pouch. At 3 h postcarrageenan injection, the pouch fluid was collected by lavage with 1 ml of cold heparin-saline. The fluid was centrifuged at 800g for 10 min at 4°C, and the supernatants were collected for analysis of PGE₂ by ELISA. At the end of the 3-h postcarrageenan injection, rats were anesthetized with a CO₂/O₂ gas, and blood was collected by heart stick. The gastrointestinal tract was exposed and observed for lesions. Sections of stomach mucosa were dissected and immediately frozen for further prostaglandin analysis. Frozen tissues were processed by homogenization in 70% ethanol. After centrifugation, the supernatants were collected, dried under a stream of nitrogen, and resuspended in ELISA buffer for PGE₂ determination by ELISA.

Acute Carrageenan-Induced Edema and Hyperalgesia in the Rat. Paw edema was induced by injecting 0.1 ml of a 1% carrageenan saline solution into the hind paw of male Sprague-Dawley rats (180-280 g), six animals per dose group. Prior to experiments, the rats were allowed free access to food and water. Compounds or vehicle (0.5% methylcellulose in water) were administered orally 1 h before the carrageenan injection, and paw volume was measured by water displacement at intervals thereafter. Analgesic activity was assessed as inhibition of hyperalgesia produced in response to a thermal stimulus from a radiant heat source (high-intensity projector lamp bulb) positioned under a Plexiglas floor directly beneath the hind paw. The withdrawal latency of the affected paw was compared with the contralateral (i.e., normal) paw and determined to the nearest 0.1 s with an electronic clock circuit and a microcomputer. Each point represents either the change in paw volume or withdrawal.

Rat Adjuvant Arthritis Model. Dose-response curves for anti-inflammatory activity were established as described by Billingham (1983). Male Lewis rats (180-200 g), six animals per dose group, were injected with 1 mg of heat-killed mycobacterium butyricum (Difco, Detroit, MI) suspended in light mineral oil (1 mg/0.025 ml) [subcutaneously in the plantar surface of the right hind paw (day 0)]. Control animals were injected with mineral oil only. Rats were assigned numbers, and body weights were followed each week. Fourteen days after injection, the left paw volume (without mycobacterium injection) was measured by the model 7150 plethysmometer (Ugo Basile, Comerio, Italy). Rats were qualified for use in the assay by measuring the left paw volume. A paw volume of greater than 0.375 ml after subtracting the vehicle-treated left paw volume (left paw volume from injected rat-left paw volume from normal rat greater than or equal to 0.375 ml) was required for inclusion in the study. After the animals were regrouped, rats received either vehicle (0.5% methylcellulose/0.025% Tween 20, 1 ml/rat b.i.d.) or compound (0.01-10 mg/kg/day in 1 ml of vehicle p.o. b.i.d.). On days 22 (7 days after compound dosing) and 26 (11 days after compound dosing), left paw volume was measured. ED₅₀ and ED₈₀ data were calculated based on the data from day 26.

Results

Assessment of COX Inhibition in Vitro and in Vivo. Inhibition of COX isoforms in vitro by valdecoxib and several NSAIDs was assessed using recombinant human enzymes (Table 1, left). Consistent with previous reports, most NSAIDs appear to nonselectively inhibit both isoforms of COX. Valdecoxib and diclofenac are the most potent inhibitors of COX-2 in this setting, exhibiting IC₅₀ values of 0.005 and 0.01 μM, respectively. In contrast to the NSAIDs evaluated, valdecoxib demonstrates clear potency and selectivity for COX-2, consistent with a previous report (Talley et al., 2000).

In an effort to more fully characterize the potency of COX-2 enzyme inhibition, kinetic parameters for time-dependent inhibition were determined (Table 2). Time-dependent parameters of K_i and K_inact obtained with valdecoxib for COX-2 are 35 μM and 3.8 s for the oxygen uptake assay (Table 2). The efficiency of compound inhibition for COX-2 can be more conveniently expressed as a function of the K_inact/K_i (seconds per micromolar). When expressed in this manner, valdecoxib and celecoxib have a similar efficiency for COX-2 inhibition, with valdecoxib having a faster rate of inactivation (K_inact). Consistent with the standard endpoint recombinant enzyme assay, both rofecoxib and etoricoxib were less potent, primarily due to their slow rate of inactivation of COX-2.

In the two-step model of COX-2 inhibition, potency is largely due to time-dependent inhibition, where the forward rate (K_inact) is vastly greater than the off-rate (K_-2) due to tight binding. Thus, the off-rate cannot be directly determined from enzyme kinetic measurements. To assess the off-rate of inhibitors from COX-2, a binding assay was utilized that directly measures the kinetics of binding of radio-labeled compounds to COX-2. Using this method, it was found that valdecoxib had a slow dissociation rate from COX-2 (t_1/2 = 98 min versus 50 min for celecoxib and 17 min for rofecoxib). Competitive binding K_is were also determined using the binding assay; the rank order of potency for binding was similar to that found with the enzyme inhibition assay, although quantitative differences in potency are apparent (Table 3).

Isoform specificity in human cells was also assessed using the whole-blood assay of Patrignani et al. (1994) as modified by Chan et al. (1995) (Fig. 2; Table 1, center). In this analysis, some of the NSAIDs showed some selectivity for COX-2 (e.g., diclofenac and etodolac). However, there are several anomalies evident in these data; naproxen and the active metabolite of nabumetone were inactive on COX-2 despite their known anti-inflammatory activity in humans, whereas drugs that exhibited specificity for COX-2, such as diclofenac and etodolac, are known to produce GI toxicity in patients with the same incidence as other NSAIDs (Physicians' Desk Reference, 1998). Interestingly, only slight differences in potency for COX-2 inhibition were observed among the COX-2-selective inhibitors.

Since in vitro measurements of COX isoform selectivity may be unreliable predictors of in vivo activity, we directly assessed the effect of various doses of several NSAIDs and valdecoxib on PG content in vivo derived from either COX-1 (gastric mucosa) or COX-2 (inflamed air pouch). This provided a quantitative biochemical assessment of the specificity of inhibition of COX isoforms in vivo. As shown in Fig. 3, valdecoxib dose-dependently decreased inflammatory PGE₂ production, with ED₅₀ occurring at approximately 0.06 mg/kg; little inhibition of COX-1-derived gastric PG content was observed over a wide dose range. In contrast, NSAIDs showed no specificity for either COX isoform (meloxicam and nabumetone) or apparent COX-1 specificity (etodolac) in this in vivo assay. Quantitative comparisons of several NSAIDs derived from dose-response analyses are shown in Table 1, panel 3.

Activity in Acute Inflammation and Hyperalgesia. In vivo potency and efficacy of valdecoxib was evaluated in a standard model of acute inflammation and pain. The injection of carrageenan into the rat paw caused marked increases in paw volume (edema) and thermal hyperalgesia that were maximal within 1 to 3 h. Prophylactic administration of either valdecoxib or naproxen produced sustained inhibition of edema and hyper-algesia (Fig. 4). When administered prior to carrageenan injection, valdecoxib was as efficacious as naproxen in blocking the inflammatory pain response (Table 4).

Anti-Inflammatory Activity in Adjuvant-Induced Arthritis. Adjuvant-induced arthritis in Lewis rats was used as a model of chronic anti-inflammatory activity. Valdecoxib exhibited potent activity in this assay (ED₅₀ = 0.03 mg/kg) (Table 5). This was equivalent in maximal efficacy to the standard NSAID indomethacin and the steroid dexamethasone, as seen from the comparison of the respective time courses of treatment (Fig. 5).

Discussion

Evidence is presented here that valdecoxib potently and selectively inhibits COX-2 in vitro and in vivo and, in animals, possesses lower potential for inhibition of COX-1 as well as anti-inflammatory and analgesic activity comparable with NSAIDs. The evidence for potency and selectivity of valdecoxib in vitro was seen in the recombinant enzyme assay and was maintained in the human whole-blood assay.

Factors underlying the high potency of valdecoxib for COX-2 were assessed mechanistically. Consistent with previous studies, the primary kinetic factors underlying the potency and COX-2 selectivity of valdecoxib was time dependent inhibition of COX-2 and not COX-1 (Copeland et al., 1994). This was manifested by a very rapid rate of inactivation (K_inact) and observed rate of inactivation (K_observed) of COX-2 by valdecoxib. These results, coupled with the finding of a very slow off-rate of binding, provide a mechanistic basis for the potency and specificity of valdecoxib for COX-2. The in vitro potency and selectivity of valdecoxib for COX-2 was confirmed in vivo, using biochemical measures of COX activity. Thus, dose-dependent inhibition of PG formation at sites of inflammation was observed (ED₅₀ = 0.03-0.06 mg/kg) without significant effects on COX-1-dependent PG production in the stomach (ED₅₀ > 200 mg/kg). These data are significant because it is now recognized that predictions of COX-2 selectivity based solely on in vitro results are suspect (Jouzeau et al., 1997) and require confirmation in vivo (Meade et al., 1993).

As would be expected of a potent COX-2 inhibitor, valdecoxib was as effective as indomethacin in both the acute carrageenan inflammation model (ED₅₀ ∼6 mpk) and in the adjuvant-induced arthritis model (ED₅₀ ∼0.03 mg/kg). The maximal effects seen with valdecoxib in these studies are similar to those produced by NSAIDs. Reports describing knockout animals, in which the COX-2 gene for has been deleted (Dinchuk et al., 1995; Morham et al., 1995), along with increasing awareness of physiological roles for COX-2, have raised concern for adverse consequences of selective COX-2 inhibition (Richardson and Emery, 1996). Although the studies described here were not designed to evaluate adverse effects, the level of inhibition of COX-2 produced by valdecoxib in animal models is no different from that already seen with drugs used extensively in patients.

It is believed that inhibition of PG synthesis by COX-1 is pivotal to the production of gastrointestinal injury by NSAIDs (Mitchell et al., 1993; Seibert et al., 1995). In addition to the minimal effects on gastric PGs, valdecoxib produced no functional evidence of COX-1 inhibition in acute models that screen for NSAID-type GI toxicity. The absence of gastric or intestinal injury after single doses of valdecoxib in rodents (compared with indomethacin, which produced clear gastric injury, peritonitis, and fibrosis in all animals) suggests that valdecoxib possesses lower potential for GI toxicity than a nonselective inhibitor.

Assays using human whole blood, which typically measure COX-1 inhibition in platelets and inhibition of COX-2 induced in monocytes with LPS, were developed to mitigate the uncertainty of in vitro systems. However, as shown here, these assays appear to suffer the same limitations seen with in vitro assays. Thus, inconsistencies noted between these results (Table 1) and the known clinical activity of several drugs (e.g., diclofenac, naproxen, nabumetone, and meloxicam) suggest that this approach does not necessarily predict either selectivity or improved GI safety in humans.

In vivo assessments of COX-2 selectivity are ultimately more relevant because they represent pharmacological activity in whole animals. PGs measured in experimentally induced inflammatory exudate and gastric extracts provide simultaneous measures of the activities of COX-2 and COX-1, respectively. In vivo, all NSAIDs appear nonselective with little or no separation of dose-response curves for inhibition of inflammation and gastric PGs. This is consistent with the well recognized toxicity of these drugs in patients given ordinary clinical doses.

In conclusion, the exquisite anti-inflammatory potency of valdecoxib in vivo is due in large part to its ability to potently inhibit COX-2. By in vitro measurements, valdecoxib is a highly potent COX-2 inhibitor. Further enzyme kinetic analyses revealed that the potency of valdecoxib for COX-2 is due to its fast rate of inactivation of COX-2 and tightness of binding, manifested by a slow off-rate. Valdecoxib's potency is maintained in animal models of acute inflammation and pain without apparent effect on COX-1 at maximal doses administered. Valdecoxib's efficacy in a chronic rat model of arthritis is reflective of its low clinical dose. Clinical evaluations of the effectiveness and side effects of valdecoxib have confirmed these conclusions.