A Revised Mechanism for Human Cyclooxygenase-2*

The mechanism of ω-6 polyunsaturated fatty acid oxidation by wild-type cyclooxygenase 2 and the Y334F variant, lacking a conserved hydrogen bond to the catalytic tyrosyl radical/tyrosine, was examined for the first time under physiologically relevant conditions. The enzymes show apparent bimolecular rate constants and deuterium kinetic isotope effects that increase in proportion to co-substrate concentrations before converging to limiting values. The trends exclude multiple dioxygenase mechanisms as well as the proposal that initial hydrogen atom abstraction from the fatty acid is the first irreversible step in catalysis. Temperature dependent kinetic studies reinforce the novel finding that hydrogen transfer from the reduced catalytic tyrosine to a terminal peroxyl radical is the first irreversible step that controls regio- and stereospecific product formation.

steps with different temperature dependences instead of the reportedly lower thermal activation barrier to deuterium transfer than protium transfer (17).
In the present work, new data collected under physiologic conditions argue that a second hydrogen transfer, downstream of AA ⅐ or LA ⅐ formation and O 2 trapping of the substrate radical, is the first irreversible step. In this reaction, the terminal peroxyl radical is reduced to the hydroperoxide product, and Tyr-371 is reoxidized to Tyr-371 ⅐ . In the proposed mechanism, kinetic factors dictate regio-and stereospecific product formation during dioxygenase catalysis, a finding with implications for designing mechanism-based COX-2-specific inhibitors (18).
Protein Preparation and Characterization-N-terminal His 6tagged proteins were expressed recombinantly in baculovirustransfected Sf9 insect cells following published protocols (19). Wild-type (WT), Y334F, and Y371F constructs (20) were sequenced and deposited at the Baylor College of Medicine Baculovirus/Monoclonal Antibody facility where proteins were expressed following amplification, titration, and Western blotting characterization (5).
The following procedure was used to purify COX-2 proteins with all manipulations conducted at 4°C unless noted. Cell pellets were suspended (1 g/4 ml) in a pH 7.4 buffer consisting of 25 mM NaH 2 PO 4 , 100 mM NaCl, 20 mM imidazole, and 1 mM PMSF prior to lysis. Next, sonication was performed using 50% amplitude and cycles of 10 s on and 60 s off. The lysate was isolated and centrifuged for 1 h at 43,000 ϫ g. The supernatant was discarded, and the membrane fraction resuspended in the above lysis buffer supplemented with a 0.1% (v/v) protease inhibitor mixture (Sigma) and Tween 20. The suspension was homogenized (Dounce), stirred on ice for 1.5 h, and then centrifuged for 1 h at 36,000 ϫ g. The cell pellet was discarded, and the supernatant was incubated with nickel-nitrilotriacetic acid resin pretreated with the lysis buffer for 90 min before loading onto a nickel-nitrilotriacetic acid affinity column (2-cm diameter, 10-ml bed volume). Chromatography utilized a pH 7.4 buffer containing 25 mM NaH 2 PO 4 , 20 mM imidazole, 0.1% (w/v) Tween 20, and 100 mM NaCl at a flow rate of approximately 0.65 ml/min. The ionic strength () was raised using 300 mM NaCl prior to eluting the protein with 25 mM NaH 2 PO 4 , 100 mM NaCl, 0.1% (w/v) Tween 20, and 200 mM imidazole. Aliquots were removed and assayed for peroxidase activity. This procedure involved adding hematin, H 2 O 2 , and 2,2Ј-azinobis(3-ethylbenzothiazoline-6-sulfonic acid) diammonium salt to samples followed by colorimetric analysis of COX-2 peroxidase activity. Fractions with peroxidase activity were pooled, loaded onto a PD-10 desalting column (Bio-Rad), and eluted with 100 mM NaH 2 PO 4 at pH 7.0 or 16 mM Na 2 H 2 P 2 O 7 at pH 8.0 in the presence of 10% (v/v) glycerol and 0.1% (w/v) Tween 20, allowed isolation of imidazole-free solutions of COX-2 apoproteins lacking the Fe(III)(Por). Alternatively, apoproteins were purified by dialysis for Ͼ12 h against the pH 8.0 buffer above.
COX-2 apoproteins were reconstituted at 20°C. Two equivalents of hematin or Mn(III)(Por)Cl were incubated with solutions of homodimeric forms of COX-2 for 15-20 min. Hydrated DE53 resin was then added, and after 5 min, the mixture was loaded onto a second PD-10 column to remove any unbound metalloporphyrin. The holoprotein was eluted with 100 mM NaH 2 PO 4 buffered to pH 7.0 or 16 mM Na 2 H 2 P 2 O 7 buffered to pH 8.0, each containing 10% (v/v) glycerol and 0.1% (w/v) Tween 20. Solutions of holoprotein were concen-FIGURE 1. Overlaid structures of human COX-2 with bound AA (yellow) and LA (cyan). The Fe(III)(Por) was replaced by Co(III)(Por) in both proteins (7,8).
The Dioxygenase Mechanism of Human COX-2 trated to Յ5 ml using an Amicon stirred cell equipped with a 30-kDa ultrafiltration membrane (Millipore), apportioned into 50 -100-l aliquots, and stored at Ϫ80 or Ϫ30°C inside the freezer of the N 2 -filled glove box described above.
Assessing Active Enzyme Concentration-The concentration of active COX-2 was analyzed using extinction coefficients calculated as outlined above and a standard assay conducted at 30 (22).
Ascertaining the concentration of active enzyme is critical to the investigation of rate constants and deuterium KIEs at variable temperatures. In such experiments, changes in protein structure might result in loss of activity. Therefore, enzyme concentrations were checked by preincubating aliquots of stock solution at reaction temperatures between 5.0 and 50.0°C for 3 min before performing standard assays. Under these conditions, no decrease in enzyme activity was detectable, and no correction for enzyme concentration was applied.
Steady-state Kinetics-Initial rates were measured at 30.0 Ϯ 0.2°C unless noted using the O 2 electrode (above). All experiments were conducted at pH 8.0 in 0.016 mM sodium pyrophosphate ( ϭ 0.1 M) using known concentrations of AA or LA and O 2 along with optimized concentrations of phenol, hydroperoxide initiator, and 1 M hematin or Mn(III)(Por)Cl.
Kinetic measurements were initiated by injecting 1-10 l of enzyme stock solution into 1.0 -1.5 ml of stirring reaction solution contained inside a chamber with no headspace. Standard assays were used to check for changes in enzyme concentration over the course of experiments. Diminution in dioxygenase activity is clearly detected as COX-2 undergoes turnover. However, initial rates extracted from the most linear portion of the reaction progress curve after allowing ϳ5 s for mixing varied in proportion to the enzyme concentration. This behavior is consistent with no significant loss of enzyme activity on the experimental time scale. In some cases, correction to the O 2 uptake rate for background drift of the YSI electrode was required. In no instance was this correction Ͼ30% of the measured rate.
Kinetic parameters were obtained by fitting data to the hyperbolic Michaelis-Menten equation: ) using Kaleidagraph 4.0 (Synergy Software). The concentra-tion of fatty acid or O 2 , [S], was varied, whereas the co-substrate concentration was held fixed at kinetic saturation; i.e. Ͼ6 K m . The apparent maximal rate (V max ) divided by the active enzyme concentration, [E], plotted versus [S] gave a hyperbolic trend that was fitted to obtain the unimolecular rate constant k cat and bimolecular rate constants k cat /K m (LA) or k cat /K m (AA) and k cat /K m (O 2 ). Bimolecular rate constants were also extracted by linear fitting of data at concentrations below 1 ⁄ 2K m (S).
Studies of temperature-dependent kinetics at high and low [O 2 ] were performed using protiated and deuterated substrates. The resulting substrate deuterium KIEs upon k cat and k cat / K m (O 2 ) were determined over a 5.0 or 10.0 to 50.0°C range. Data were analyzed by linear regression of natural log versus inverse temperature plots assuming the phenomenological Arrhenius expression: Viscosity Effects-Experiments were carried out to probe for diffusion-limited contributions to k cat /K m (O 2 ) in accord with the Stokes-Einstein relation (23). Initial rates were examined at varying [O 2 ] while [AA] or [LA] was held constant at 50 -100 M (Ͼ6 K m ) in the presence of sucrose or polyethylene glycol, which served as micro-and macroviscosogens, respectively. An Ostwald viscometer allowed the relative viscosity, / 0 , to be measured and correlated to k cat /K m (O 2 ) as a means of probing for diffusion-controlled steps (24,25). The effect of added micro-or macroviscosogen was examined up to / 0 ϭ 4 to test whether O 2 -trapping of AA ⅐ or LA ⅐ and hydroperoxide product release steps contribute to k cat /K m (O 2 ).
Product Analysis-Products of AA and LA oxidation were quantified in relation to the O 2 consumed to confirm the reaction stoichiometries of Equations 1 and 2. Experiments were carried out by adding enzymes to 3-10-ml reaction solutions. After monitoring O 2 uptake, 1 ml of acetic acid was added to quench the reaction. Acidification also neutralized the fatty acid oxidation products, allowing for extraction into CH 2 Cl 2 . Although COX-2 possesses peroxidase activity, 5% (v/v) trimethyl phosphite was added to ensure reduction of hydroperoxide products; i.e. PGG 2 was converted to PGH 2 , HPETEs were converted to HETEs, and HPODEs were converted to HODEs (26). Solutions of these compounds were evaporated to a residue under a stream of N 2 and then redissolved in a known volume of methanol. Samples were analyzed by ultra-high performance liquid chromatography (UPLC) on a Waters Acquity/ Xevo-G2 system equipped with a reverse phase C 18 column (HSS T3 1.8 m, 2.1 ϫ 100 mm) and quadrupole time-of-flight mass spectrometer operating in negative ion mode. The mobile phase (60% acetonitrile and 40% water), flowing at a rate of 0.5 ml/min, allowed separation of products, which were identified based on comparisons with authentic samples.
Competitive Deuterium KIEs-Apparent deuterium KIE upon LA oxidation, ap D k cat /K m (LA), was measured competitively in 100 M solutions of perprotiated (h 31 -LA) and perdeuterated (d 31 -LA) fatty acid combined in a 1:3 ratio. Experiments were performed with WT and Y334F COX-2 over a range of O 2 concentrations by analyzing the ratio of deuterated to protiated products at low conversion (Յ10%) relative to 100% conversion. The latter measurements used excess COX-2 or soybean lipoxygenase (26). Following the protocol above, UPLC-MS exposed two baseline resolved signals corresponding to the monodeprotonated isomers of h 31 -HODE and d 31 -HODE. This ratio was corroborated by analysis of electronic absorption at 235 nm, the wavelength of maximum absorption associated with the conjugated diene in the product.
Solvent Isotope Exchange-Experiments were undertaken to test whether the reactive hydrogen of the substrate undergoes exchange with the solvent in the presence and absence of Tyr-371 ⅐ -containing COX-2. WT, Y334F, and Y371F proteins were incubated anaerobically in 3-10 ml of N 2 -saturated pH 8 buffered H 2 O containing 50 M d 31 -LA. HPODEs in this substrate were present at trace levels sufficient to ensure full enzyme activation. 2 mM phenol was also added to protect enzymes from overoxidation by HPODEs. Samples were incubated at 4°C under N 2 for 12 h before reisolation of unreacted d 31 -LA. The procedure used was analogous to that described in under "Product Analysis." Acid-quenched samples were extracted with CH 2 Cl 2 , evaporated to a residue, redissolved in methanol, and then analyzed by UPLC-MS. A sample containing all components but no protein was analyzed as a second type of control for detecting spurious isotope exchange.

Results
Recently, Danish et al. (5) demonstrated that the dioxygenase activity of COX-2 could be analyzed independently of the peroxidase activity of the enzyme at sufficiently high concentrations of phenol. Such investigations are not possible with COX-1 where rate acceleration is followed by inhibition upon adding phenol (6). However, rates of dioxygenase catalysis were found to saturate hyperbolically upon increasing the phenol concentration to 1.0 -3.0 mM with Fe(III)(Por)-containing WT COX-2 and Y334F COX-2. With these enzymes, trace hydroperoxide impurities present in AA or LA were sufficient to observe optimal rates of turnover. The lower peroxidase activity of Mn(III)(Por)-WT COX-2 eliminates the requirement of added phenol. Even in the presence of 3-10 M HPODE or HPETE needed to optimize initial rates, no phenol dependence was detected. Furthermore, the inhibition of dioxygenase activity by phenol was undetected in all forms of COX-2 examined in this study.
LA Oxidation Kinetics-Steady-state rate constants for LA oxidation were determined for WT COX-2 and the Y334F variant, which lacks a conserved hydrogen bond to Tyr-371 ⅐ or Tyr-371 (7,8,11,12). In Figs. 2 and 3, the apparent rate at kinetically saturating co-substrate concentration (Ͼ6 K m ) is given as V/[E] and plotted versus varied substrate to determine k cat /K m and k cat as outlined under "Experimental Procedures." Rate constants summarized in Table 1 show that WT and Y334F COX-2 exhibit similar k cat, k cat /K m (AA), and k cat / K m (LA). Comparisons of AA with LA require correction by a factor of 2 for the difference in O 2 equivalents consumed in Equations 1 and 2. Quantification of dioxygenase products in tandem with measurements of O 2 uptake confirmed the stoichiometry for formation of PGH 2 as the major product of AA oxidation and HODEs as the products of LA oxidation. The k cat /K m (O 2 ) was, however, an order of magnitude greater for AA than LA due to a reduction in K m (O 2 ).
Deuterium KIEs on LA Oxidation-The apparent rate constants for consumption of h 31 -LA versus d 31 -LA afford deute-rium KIEs that increase with [O 2 ] until a limiting D k cat /K m (LA) is attained. Most measurements were made competitively by mixing the substrate isotopologues and analyzing the h 31 -HODE to d 31 -HODE product ratios after reduction of h 31 -and d 31 -HPODEs following a published procedure (26). Apparent competitive deuterium KIEs upon LA oxidation, ap D k cat / K m (LA), were found to increase from subsaturating to saturating [O 2 ] in Fig. 4. This behavior is in line with the apparent deuterium KIEs determined non-competitively by measuring O 2 uptake rates with h 31 -LA or d 31 -LA (see Table 1).
Remarkably, the D k cat /K m (LA) values for WT and Y334F COX-2 are equivalent to the deuterium KIEs upon bimolecular rate constants for O 2 uptake, D k cat /K m (O 2 ), determined under non-competitive conditions. This unusual deuterium KIE upon O 2 consumption arises from retention of the hydrogen abstracted from LA at Tyr-371; solvent exchange with this residue is slow on the time scale of enzyme turnover (see below).
The observed equivalence of D k cat /K m (LA) to D k cat /K m (O 2 ) is consistent with a common irreversible hydrogen transfer in WT and Y334F COX-2 after LA and O 2 enter the catalytic cycle. Table 2 consists of data obtained with two forms of COX-2 containing a Fe(III)(Por) or Mn(III)(Por) prosthetic group. The WT COX-2 exhibits a D k cat of ϳ21, which is 2-3 times larger than D k cat /K m (LA) and D k cat /K m (O 2 ). These results, together with reported temperature dependences of D k cat and D k cat / K m (O 2 ), expose a change in the irreversible hydrogen transfer step as the [O 2 ] is raised (5). In contrast, Y334F COX-2 exhibits a D k cat of ϳ30 that is indistinguishable from D k cat /K m (LA) and D k cat /K m (O 2 ) within the error limits.
The large [O 2 ]-and [LA]-independent deuterium KIEs implicate a common irreversible and rate-controlling step for Y334F COX-2 turnover at all substrate concentrations. Temperature studies of D k cat and D k cat /K m (O 2 ) described below also support a common irreversible step. In addition, competitive oxygen-18 KIEs upon k cat /K m (O 2 ), reflecting steps that begin with O 2 encounter and lead up to and include the first irreversible step, are indistinguishable for WT and Y334F COX-2 catalysis with LA (1.0133-1.0156) and AA (1.0194 -1.0205). These results, which correlate the magnitude of 18 Table 2) in sodium pyrophosphate (0.016 M) at 30°C, pH 8.0, and ϭ 0.1 M as detailed under "Experimental Procedures."

TABLE 1 Rate constants for oxidations of AA and LA at 30°C
Limiting rate constants are quoted with Ϯ1 S.E. in parentheses.

FIGURE 5. Isotope exchange between d 31 -LA and H 2 O is mediated by WT and Y334F COX-2 enzymes (spectra on right) but not the Y371F mutant or in O 2 -free solutions (spectra on left).
Solutions were incubated for equivalent times (ϳ12 h) at 4°C, pH 8.0, and ϭ 0.1 M as described under "Experimental Procedures." In the experiments labeled controls 1 and 2, the catalytic Tyr-371 ⅐ required to abstract hydrogen from the substrate is absent. AA and LA using WT and Y334F COX-2. Data published for WT COX-2 (5) are compared with those determined for the Y334F variant in Fig. 7. These two enzymes exhibit similar competitive 18 O KIEs as described above with each fatty acid substrate, suggesting that the transition state is the same for the step that limits k cat /K m (O 2 ) (28,35). This result is consistent with a common step giving rise to D k cat /K m (O 2 ) when the hydrogen initially abstracted from the substrate is retained at Tyr-371 and followed by Tyr-371 ⅐ formation via irreversible hydrogen transfer after O 2 enters the catalytic cycle. Previous findings on LA oxidation by WT COX-2 indicated a change in the first irreversible step upon increasing [O 2 ] from below K m (O 2 ) to Ͼ6K m (O 2 ) (5). The activation energy (E a ) associated with k cat /K m (O 2 ) and k cat was determined for h 31 -LA and d 31 -LA reacting with Y334F COX-2. In contrast to WT COX-2, the variant shows indistinguishable deuterium KIEs upon k cat , k cat /K m (LA), and k cat /K m (O 2 ) with each approaching 30 at 30°C. This indication of kinetically simple behavior where the rate constant for dioxygenase catalysis is controlled by a common irreversible step is also consistent with the negligible ⌬E a associated with D k cat ϭ 2.7 Ϯ 0.7 kcal mol Ϫ1 and D k cat / K m (O 2 ) ϭ 2.2 Ϯ 0.6 kcal mol Ϫ1 derived from data at saturating and subsaturating [O 2 ], respectively, where E a (d 31 -LA) ϭ 9.4 Ϯ 0.7 or 7.9 Ϯ 0.3 and E a (h 31 -LA) ϭ 6.7 Ϯ 0.7 or 5.7 Ϯ 0.5. At low [O 2 ], where results reveal irreversible hydrogen transfer from Tyr-371 to an O 2 -derived peroxyl radical, Y334F COX-2 exhibits a ⌬E a that is ϳ2 times smaller than that reported for the WT enzyme where ⌬E a ϭ 5.0 Ϯ 0.8 kcal mol Ϫ1 and A H /A D is much more inverse (5). These results must be reconciled with the proposal of a common irreversible step in WT and Y334F COX-2 under conditions where the enzymes exhibit indistinguishable competitive 18 (15) in which deuterium KIEs were found to vary inversely with temperature at pH 7.0, resulting in E a (h 39 -AA) Ͼ E a (13,13-d 2 -AA). Importantly, the same source of 13,13-d 2 -AA used earlier (13,15) was used in the present study. In the earlier study, no attention was devoted to optimizing the [phenol] required to observe maximal rates at different temperatures. This oversight might cause kinetic complexity that results in the aberrant ⌬E a .
Data reported previously (5) and in this study are compiled in Table 3. Arrhenius analyses suggest a higher thermal barrier to the oxidation of AA than the oxidation of LA. As discussed below, these observations could reflect a greater energy barrier due to reorganization of AA in the active site of COX-2 as well as a mechanism that involves additional pre-equilibrium steps before the first irreversible hydrogen transfer step that forms PGG 2 and regenerates Tyr-371 ⅐ .

Discussion
The similar distributions of products formed by COX-1 and COX-2 suggest that the enzymes react by common mechanisms. There are fundamental differences in their substrate specificity profiles, however, because of the larger active site in COX-2 (1). In this study, dioxygenase catalysis with LA to produce HPODEs was examined as a model for HPETE formation from AA, which likely results from the same initial steps as PGG 2 .
Analyses of Steady-state Kinetic Parameters-In the works cited above (5, 6), COX-1 and COX-2 exhibit apparent second order rate constants that depend upon co-substrate concentrations in a hyperbolic manner. To the best of our knowledge, the  present studies are the first to focus upon how substrate deuterium KIEs in COX respond to changes in [O 2 ]. Although COX-1 reacts with high specificity for AA over LA, COX-2 exhibits nearly identical k cat , k cat /K m (AA), and k cat /K m (LA) when a correction is applied for the O 2 consumed. In contrast, k cat /K m (O 2 ) is an order of magnitude greater for AA than LA. This difference is due to a diminution in K m (O 2 ), which could indicate a more favorable O 2 uptake. The mechanism for LA is less complex because a single equivalent of O 2 traps the LA ⅐ before irreversible production of 9R-and 13S-HPODEs. Fewer steps contributing to k cat /K m (O 2 ) could also explain why E a determined with LA is 5-6 kcal mol Ϫ1 smaller than E a determined with AA as discussed below.
Deuterium KIEs upon LA Oxidation-The ap D k cat /K m (LA) and ap D k cat /K m (O 2 ) exhibit hyperbolic dependences upon cosubstrate concentrations analogous to that seen for apparent rate constants (5,6). Each increases to a limiting deuterium KIE at sufficiently high co-substrate concentration. The apparent competitive deuterium KIEs upon LA consumption were analyzed at varying [O 2 ] to test the kinetic mechanism and the relationship of D k cat /K m (LA) to D k cat /K m (O 2 ) and D k cat (27,35).
The ap D k cat /K m (LA) increases upon raising the [O 2 ] until an [O 2 ]-independent D k cat /K m (LA) is reached in Fig. 4. The D k cat / K m (LA) is indistinguishable from D k cat /K m (O 2 ) in WT and Y334F COX-2, indicating that the same irreversible hydrogen transfer contributes to the deuterium KIE. Unlike WT COX-2 where a change in the first irreversible step occurs as the [O 2 ] is raised, Y334F COX-2 exhibits similar deuterium KIEs at all substrate concentrations; i.e. D k cat , D k cat /K m (LA), and D k cat / K m (O 2 ) all approach 30. These results indicate a common irreversible and rate-limiting step at all substrate concentrations. On these grounds, the mechanism in Fig. 9 is proposed to involve reversible LA ⅐ formation and O 2 trapping followed by irreversible peroxyl radical reduction by Tyr-371.
Certain mechanisms can be excluded based on the [O 2 ] dependence of ap D k cat /K m (LA) and the [LA] dependence of ap D k cat / K m (O 2 ). For instance, a "ping-pong" type mechanism where the fatty acid is oxidized and O 2 is reduced in kinetically independent steps is inconsistent with the co-substrate dependence of the apparent deuterium KIEs (36). Ordered sequential mechanisms where (i) initial LA or AA binding is followed by O 2 uptake and (ii) initial O 2 binding is followed by LA or AA uptake can also be ruled out based on the increase in ap D k cat /K m (LA) upon raising the [O 2 ]. Published derivations (36,37) indicate that in (i) the ap D k cat /K m (LA) should decrease upon raising [O 2 ] and in (ii) the ap D k cat /K m (LA) should be O 2 -independent.
Findings with WT and Y334F COX-2 expose an increase in ap D k cat /K m (LA) upon raising the [O 2 ] until a limiting D k cat / K m (LA) equal to D k cat /K m (O 2 ) is reached. This behavior is explained by the random kinetic mechanism in Fig. 9 under the simplifying assumptions that reversible LA binding couples to reversible hydrogen transfer and that the reactions exhibit negligible equilibrium isotope effects. These assumptions allow use of the previously derived Equation 3 as a starting point (37 A ternary complex is hypothesized to result from O 2 binding to LA ⅐ in Fig. 9. This intermediate, consisting of the reduced catalytic Tyr-371 and LA-derived peroxyl radical, is expected to react by irreversible hydrogen transfer to form either 9R-or 13S-HPODE. In competition, the ternary complex could release O 2 from the LA-derived peroxyl radical via k Ϫ2 . Alternatively, O 2 could dissociate from the peroxyl radial and partition into an active site pocket in COX-2, and reverse hydrogen transfer from Tyr-371 to LA ⅐ could release LA via k Ϫ4 . When k Ϫ4 Ͼ Ͼ k Ϫ2 , raising [O 2 ] results in an increase in the ap D k cat / K m (LA) to the intrinsic KIE defined by H k 5 / D k 5 . This scenario reproduces the trends in Fig. 4 in accord with Equations 3-5.
The mechanism also allows for observations that D k cat / K m (LA) is equivalent to D k cat /K m (O 2 ). The random kinetic mechanism in Fig. 9 predicts that the deuterium KIE approaches a minimum as the [O 2 ] is lowered. At the lowest [O 2 ], deviation from unity is expected when the equilibrium isotope effect upon initial hydrogen transfer is significant. A negligible effect is anticipated in COX-2 because of the similar vibrational  The Large Deuterium KIE on AA Oxidation-The present results reveal a large D k cat /K m (O 2 ) when oxidation of h 39 -AA is compared with 13,13-d 2 -AA. As with LA, this result requires the initially abstracted hydrogen to be retained at Tyr-371 to explain the presence of a deuterium KIE upon the rate constant for O 2 consumption. During catalysis with AA, Tyr-371 is reoxidized by PGG 2 ⅐ in k 8 Ј of Fig. 9. Evidence that initial hydrogen transfer from AA to Tyr-371 ⅐ might be reversible comes from earlier work in which the AA ⅐ was detected at vanishingly low [O 2 ] (12).
A large D k cat of ϳ21 and smaller D k cat /K m (O 2 ) of ϳ7 are observed during LA oxidation by WT COX-2. This behavior has been attributed to a change in the irreversible hydrogen transfer step upon varying [O 2 ]. A similar change in the first irreversible step might also occur during AA oxidation where D k cat is ϳ3 and D k cat /K m (O 2 ) is ϳ18. Alternatively, an ordered sequential mechanism where AA binding and formation of AA ⅐ occur before O 2 is consumed might account for the difference in deuterium KIEs. Assignment of the mechanism requires examining apparent competitive KIEs at varying [O 2 ].
The results highlighted in this study pertain to physiologically relevant conditions for the oxidation of LA and AA by COX-2. Previously, D k cat /K m (AA) was estimated to be ϳ2 in non-competitive experiments at an [O 2 ] of ϳ250 M; however, the use of non-optimal [phenol] could cause kinetic complexity under these conditions (5). Inhibition of dioxygenase activity by peroxidase turnover could also obscure D k cat /K m (AA) so that it is smaller than the intrinsic H k 8 Ј/ D k 8 Ј in Fig. 9.
The D k cat /K m (O 2 ) for AA is proposed to reflect Tyr-371 oxidation by the terminal peroxyl radical. It is unclear, however, if the deuterium KIE arises from pre-equilibrium hydrogen transfer (from AA to Tyr-371 ⅐ ) coupled to reversible 5-exocyclization steps in Fig. 9. Although D k cat /K m (O 2 ) exposes irreversible hydrogen transfer from Tyr-371 to PGG 2 ⅐ , the possibility that O 2 equivalents are consumed independently cannot be rigorously excluded. Such a reaction seems unlikely in view of the greater E a and larger 18 (27,35), the size of the 18 O KIE should increase in response to polarization of the hydrogen transfer to a proton/ deuteron (H ϩ /D ϩ )-coupled electron transfer-like transition state (38). Such reactivity could explain the moderate to large D k cat /K m (O 2 ) and competitive 18 O KIEs measured with AA and LA as well as the absence of viscosity effects upon k cat /K m (O 2 ), which reveals thermally activated hydrogen transfer rather than diffusion control.
Temperature Studies of Dioxygenase Catalysis-The temperature-dependent rate constants for WT and Y334F COX-2 summarized in Table 3 contradict findings recently reported by Wu et al. (15) that E a (h 39 -AA) significantly exceeds E a (13,13d 2 -AA). Such results conflict with expectations that in the normal thermodynamic range tunneling is more probable for protium than deuterium (39) and requires less reorganization to achieve the reactive configuration. These properties make E a smaller for the lighter isotope of hydrogen. Although there are instances where inversely temperature-dependent deuterium KIEs have been attributed to hydrogen tunneling, this generally requires a highly exergonic "inverted region" in which the reorganization energy is offset by very favorable Gibbs free energy ( Յ Ϫ⌬G 0 ) (17). This situation is unlikely in COX-2 where oxidation of Tyr-371 by a peroxyl radical is estimated to be thermoneutral, ⌬G 0 ϳ 0, because of the similar O-H bond strengths in Tyr-371 and the hydroperoxide product (40).
It is unsurprising that hydrogen transfer from Tyr-371 to the terminal peroxyl radical derived from LA is associated with a smaller E a than the analogous reaction of AA. Following antarafacial O 2 trapping of LA ⅐ or AA ⅐ , rearrangement of the terminal peroxyl radical must occur to remove hydrogen from Tyr-371. The hydrogen transfer that limits k cat /K m (O 2 ) is expected to require less thermal activation to reorganize the smaller 9R or 13S peroxyl radical derived from LA than the larger PGG 2 ⅐ derived from AA, although multiple explanations are possible.
The reduction of PGG 2 ⅐ might take place over a shorter distance than the 9R or 13S peroxyl radical, explaining the larger magnitude of the Arrhenius prefactor (A). Favorable pre-equilibria could also elevate A to a value approaching 10 16 M Ϫ1 s Ϫ1 in the case of AA (see Table 3). Although the temperature dependence of k cat also indicates a large A for AA compared with LA, the value of 10 11 -10 12 s Ϫ1 falls within the upper limit defined by transition state theory where A is associated with k B T/h value of ϳ10 13 s Ϫ1 where is a probability factor related to hydrogen tunneling, k B is the Boltzmann constant, T is temperature, and h is the Planck constant.
Catalysis by WT and Y334F COX-2-Oxidation of Tyr-371 by the oxidized prosthetic group is slower in Y334F COX-2 than in the WT enzyme (22); however, there are no major differences in the steady-state rate constants of the enzymes (see Table 1). This behavior is consistent with the observed deuterium and 18 O KIEs but raises questions concerning how their magnitudes might depend upon polarization of the hydrogen transfer in transition states with differently sized AA-and LAderived peroxyl radicals (27,35).
Mutating Tyr-334 to Phe reduces complexity of the enzyme kinetics and unmasks a single irreversible step controlling LA oxidation. The evidence for this change is inflation of D k cat , D k cat /K m (LA), and D k cat /K m (O 2 ) to values approaching 30. Additional support comes from temperature studies that reveal similar a E a for k cat and k cat /K m (O 2 ). This behavior is consistent with irreversible hydrogen transfer from Tyr-371 to the LA-derived peroxyl radical.
In contrast to Y334F COX-2, the WT enzyme exhibits a D k cat of ϳ21 along with D k cat /K m (LA) and D  Although the magnitude of k cat /K m (O 2 ) is essentially the same for WT and Y334F COX-2 oxidizing LA, a 4-fold increase in D k cat /K m (O 2 ) is observed. This is possibly the result of expanding the hydrogen transfer distance upon removal of the conserved hydrogen bond to Tyr-371. The smaller E a observed for Y334F COX-2 relative to WT COX-2 could be associated with greater protein flexibility accommodating reorganization of enzyme-bound intermediates required for hydrogen transfer from Tyr-371 to the terminal peroxyl radical.
Conclusions-Conditions have been identified where dioxygenase and peroxidase catalysis can be examined independently in recombinant human COX-2. This feature allows for the analyses of apparent rate constants and deuterium KIEs at variable co-substrate concentrations to address the kinetic mechanism of the enzyme. Studies of dioxygenase catalysis by Y334F COX-2, which lacks a conserved hydrogen bond to the catalytic Tyr-371 and Tyr-371 ⅐ , reveal kinetically uncomplicated behavior. Using the Y334F variant together with the simpler reacting LA substrate exposes a random sequential mechanism and hydrogen transfer from Tyr-371 to a terminal peroxyl radical in the first irreversible step.
Temperature-dependent deuterium KIEs at subsaturating [O 2 ] revealed differences in the behaviors of WT and Y334F COX-2. In both enzymes, the deuterium KIE upon k cat /K m (O 2 ) arises from retention of the hydrogen abstracted from the substrate at Tyr-371 in the absence of solvent isotope exchange. In the first irreversible step, the terminal peroxyl radical is proposed to accept the hydrogen retained at Tyr-371 via a polarized hydrogen transfer or H ϩ /D ϩ -coupled electron transferlike transition state. The less temperature-dependent D k cat / K m (O 2 ) observed for Y334F COX-2 relative to the WT enzyme suggests that removal of the conserved hydrogen bond creates greater protein flexibility or lower reorganization energy within the active site.
In this study, the reactivity of WT COX-2 with AA was also examined at physiologically relevant [O 2 ] for the first time, and a large deuterium KIE was observed upon k cat /K m (O 2 ). This result again indicates irreversible hydrogen transfer step after O 2 enters the catalytic cycle at physiologically relevant concentrations. Smaller deuterium KIEs at saturating [AA] and [O 2 ], i.e. D k cat , could arise from a change in hydrogen tunneling distance or in the kinetic mechanism. Comparing the temperature dependences of D k cat and D k cat /K m (O 2 ) suggests a higher activation barrier required to attain the distance required for hydrogen tunneling from Tyr-371 to the enzyme-bound PGG 2 ⅐ derived from AA than from Tyr-371 to the smaller acyclic peroxyl radical derived from LA. Future studies of deuterium KIEs upon AA oxidation will address the origin of discrepancies in the temperature-dependent deuterium KIEs and mechanistic inconsistencies in the literature to date.
Author Contributions-Y. L. was responsible for the acquisition of all data in this paper. J. P. R. was responsible for the preparation of the manuscript. Both authors participated in the analysis of data described herein.