Interactions of 2-O-arachidonylglycerol ether and ibuprofen with the allosteric and catalytic subunits of human COX-21[S]

Prostaglandin (PG) endoperoxide H synthase (PGHS)-2, also known as cyclooxygenase (COX)-2, can convert arachidonic acid (AA) to PGH2 in the committed step of PG synthesis. PGHS-2 functions as a conformational heterodimer composed of an allosteric (Eallo) and a catalytic (Ecat) monomer. Here we investigated the interplay between human (hu)PGHS-2 and an alternative COX substrate, the endocannabinoid, 2-arachidonoylglycerol (2-AG), as well as a stable analog, 2-O-arachidonylglycerol ether (2-AG ether). We also compared the inhibition of huPGHS-2-mediated oxygenation of AA, 2-AG, and 2-AG ether by the well-known COX inhibitor, ibuprofen. When tested with huPGHS-2, 2-AG and 2-AG ether exhibit very similar kinetic parameters, responses to stimulation by FAs that are not COX substrates, and modes of inhibition by ibuprofen. The 2-AG ether binds Ecat more tightly than Eallo and, thus, can be used as a stable Ecat-specific substrate to examine certain Eallo-dependent responses. Ibuprofen binding to Eallo of huPGHS-2 completely blocks 2-AG or 2-AG ether oxygenation; however, inhibition by ibuprofen of huPGHS-2-mediated oxygenation of AA engages a combination of both allosteric and competitive mechanisms.


Structural analysis of products formed upon oxygenation of 2-AG ether by huPGHS-2
A reaction was performed using a standard COX assay mixture that included 20 g of 2-AG ether as the substrate and sufficient huPGHS-2 to consume approximately 80% of the substrate during a 2 min incubation. Immediately afterwards, a volume of 100 mM SnCl 2 in methanol was added to the sample such that the final SnCl 2 concentration was 1 mM. The sample was vortexed and incubated at room temperature, then extracted with ethyl acetate, dried under N 2 , and kept in a sealed tube until analyzed. Control reactions were performed with assay buffer alone, with assay buffer that included heme and phenol, with huPGHS-2 in buffer alone, with huPGHS-2 in buffer that included heme and phenol, with 2-AG ether in buffer alone, with 2-AG ether in buffer containing heme and phenol, and with 2-AG ether plus huPGHS-2 in buffer alone.
Acetylation of the extracted reaction products was performed with a 1:1 mixture of acetic anhydride and pyridine for 20 min at 100°C. The reaction mixture was dried under nitrogen, dissolved in methanol, and then diluted with water to a final methanol concentration of less than 15%. Solid phase extraction and reversed phase chromatography were performed essentially as previously described (16). The LC effluent was directly interfaced into the electrospray ionization source of a triple quadrupole mass spectrometer (Sciex API 5500; PE-Sciex, Thornhill, ON, Canada) where mass spectrometric analyses were performed in the positive ion mode (m/z 250-800) or as MS 2 product ions using nitrogen as collision gas at a collision voltage of 25 V. TMS ether derivatives were prepared from HPLC-purified metabolites and analyzed by electron ionization by capillary GC/MS, as previously described (17), using a Finnigan DSQ GC-MS system (Thermo Finnigan, Thousand Oaks, CA) with a ZB-l column (30 m, 0.25 mm inner diameter 0.25 mm film thickness; Phenomenex). The gas chromatograph was programmed from 150 to 270°C at 30°C/min, 270 to 315°C at 10°C/min, and finally held at 315°C for 6 min. The injector was maintained at 230°C, the transfer line was maintained at 290°C, and the ion source at 200°C.

Statistical analyses
Student's t-tests were performed in Microsoft Excel. If the experiments had the same numbers of repetitions, probabilities were calculated with a Student's paired t-test, with a two-tailed distribution. If the experiments had different numbers of repetitions, probabilities were calculated with a Student's unequal variance t-test, with a two-tailed distribution.
with higher affinity to E cat than E allo of murine (mu)PGHS-2 (11); in contrast, AA binds with much higher affinity to E allo than E cat (7). Interestingly, a nonsubstrate (ns)FA, 13-methyl AA, increases the rate of oxygenation of 2-AG by muPGHS-2 by increasing the V max , but not the K m , toward 2-AG (11).
The 2-AG is unstable and readily rearranges to 1-AG and hydrolyzes to AA and glycerol (12). This instability presents experimental difficulties in studying the interactions of 2-AG with PGHSs. The 2-O-arachidonylglycerol ether (2-AG ether) is a stable analog of 2-AG. In the first part of the present study, we report the characterization of 2-AG ether as a substrate of huPGHS-2. We find that the 2-AG ether behaves very much like 2-AG with huPGHS-2. Because of its stability, 2-AG ether can serve as a surrogate for 2-AG in enzyme studies.
In related work described here, we examined the ability of the commonly used NSAID, (S)-(+)-ibuprofen (IBP), to interact with E allo and E cat to inhibit huPGHS-2. We confirm results of earlier studies that IBP is an allosteric inhibitor of 2-AG oxygenation (13) and extend this finding to 2-AG ether. We also observed that IBP binding to E allo of huPGHS-2 allosterically inhibits AA oxygenation, but does so only incompletely. Complete inhibition involves the binding of IBP to both E cat and E allo of huPGHS-2.

Expression, purification, and assay of huPGHS-2 variants
Procedures for the expression and purification of recombinant native huPGHS-2 and mutant huPGHS-2 heterodimer variants from insect cells were as described previously (7,15). The purity of the recombinant huPGHS-2 was determined by SDS-PAGE and Western blot analysis (9). In most cases, COX activity was determined using measurements of O 2 consumption with an O 2 electrode (7). One unit of COX activity is defined as 1 mol of O 2 consumed per minute at 37°C in the standard assay mixture. The average specific activity of purified huPGHS-2 with 100 M AA was 40 units per milligram protein. This specific activity is similar to that reported in earlier studies from our laboratory

Products formed from 2-AG ether by huPGHS-2
When 2-AG ether (75 nmoles) was incubated with an excess of huPGHS-2 (240 nmoles) at 37°C in a standard COX assay mixture for 2 min, 112 nmoles of O 2 were consumed. We assumed that the excess enzyme led to complete conversion of the substrate and that either one or two O 2 molecules were incorporated into the 2-AG ether substrate. These assumptions were corroborated by the mass spectrometric results described below. Accordingly we calculated that 1.49 mol O 2 were incorporated per mole of 2-AG ether, indicating that about 70% of the products were bis-oxygenated (i.e., 2-PGH 2 -glycerol ether) and 30% were mono-oxygenated [i.e., 2-(hydroxy-eicosatetraenoyl)glycerol ether(s)].
Mass spectrometric studies using LC-MS and LC-MS/ MS were performed to directly characterize the structural products obtained following the action of huPGHS-2 on 2-AG ether (Fig. 1). In order to impart favorable mass spectrometric characteristics, the products extracted from the reaction mixture were first reduced with SnCl 2 and then derivatized by acetylation. RP-HPLC was able to separate two less lipophilic products from the starting 2-AG ether that were not present in control incubations with huPGHS-2 when no phenol or heme was present (Fig. 1A). These are labeled peak A and the more lipophilic peak B. Because these were presumed to be acetylated ether diglycerides, positive ion electrospray ionization as the ammonium adduct ion (NH 4 + ) was employed. The observed [M+NH 4 ] + adduct ions were m/z 642 (peak A; Fig. 1B) and m/z 524 (peak B; Fig. 1C), while the signal for unreacted starting material was found to produce m/z 466 (peak C; Fig. 1D). Collisional activation of the starting material (peak C) yielded a major product ion at m/z 273 corresponding to cleavage of the arachidonyl chain at the ether bond (Fig. 1D).
The spectrum of least lipophilic reaction product (acetyl derivative, ammonium ion adduct, m/z 642) ( Fig. 1B) was consistent with the addition of 3-hydroxyl groups (analyzed as acetyl esters) and reduction of one double bond. This was consistent with a PGF 2 -like structure generated from a PGH 2 endoperoxide intermediate formed by the COX activity of huPGHS-2 acting on the arachidonyl ether chain of 2-AG ether. The collisional activation of m/z 642 [M+NH 4 ] + yielded a very prominent product ion at m/z 445 (Fig. 1B), consistent with three neutral losses of acetic acid (60 Da each) and ammonia (NH 3 ). The abundant ion at m/z 269 could then be understood as cleavage of the arachidonyl carbon-ether bond with positive charge retention on the 20-carbon alkyl leaving group via the mechanism suggested in supplementary Fig. 1. The corresponding ether-bond fragment ion was observed in the MS/MS spectra of peaks B and C, but each was 2 Da and 4 Da, respectively, higher in measured m/z because the alkyl carbocation generated by collisional activation (supplementary Fig. 1) has five and six rings or double bonds, respectively, compared with the arachidonyl carbocation generated from CID of 2-AG ether, which has only four for 2-AG (15); and V max of 43 units/mg and K m 10 M for AA (15). Thus, the catalytic efficiencies (V max /K m ) with huPGHS-2 are similar when each substrate is tested individually.
To estimate the relative affinities of E allo versus E cat of huPGHS-2 for 2-AG ether, we first examined the effects of rings or double bonds. Considering the presence of three hydroxyl groups and loss of a double bond, the data are consistent with the presence of a novel ether lipid having a PGF 2 structural element, 2-O-(PGF 2 )-glycerol. The scale employed in these experiments was not sufficient for NMR analysis of this metabolite.
The increase in the observed adduct molecular ion mass for peak B was 58 Da (Fig. 1C), consistent with one additional hydroxyl group along the arachidonyl carbon chain that had been converted to an acetate ester. Collisional activation of peak B yielded ions at m/z 447 [M+H-CH 3 COOH] + , m/z 387 (m/z 477-CH 3 COOH), and the most abundant product ion at m/z 271 (Fig. 1C). This latter ion corresponded to the most abundant product ion observed in the MS/MS spectrum of the starting material (that being m/z 273), but 2 Da lower, corresponding to cleavage of the ether bond in 2-AG ether and one additional double bond introduced by the presence of one acetoxy group that had been lost as acetic acid following collisional activation. The position of this hydroxyl group on the 20-carbon chain was determined by electron ionization MS as the TMS derivative (Fig. 2), and the ion at m/z 225 [CH 3 (CH 2 ) 4 CH=CH-CH=CH-CH=O + -TMS] was consistent with an introduction of a hydroxyl group at C-11 of the arachidonoyl carbon chain and migration of the  11,12 double bond to  12,13 . Thus, this metabolite was determined to be 2-O-(11-hydroxy-eicosatetraenyl)-glycerol.

Comparison of 2-AG and 2-AG ether as huPGHS-2 substrates
We determined the following V max and K m values for recombinant huPGHS-2 when comparing 2-AG ether, 2-AG, and AA as substrates: V max of 27 units/mg and K m = 4.6 M for 2-AG ether (Fig. 3); V max of 30 units/mg and K m 7 M    The results in Table 1 provide further evidence that 2-AG ether binds more tightly to E cat than E allo of huP-GHS-2. In contrast to what is observed with PA or with 5 M AA itself, 2-AG ether, at an initial concentration of 5 M, more than 25 times that of unreacted [1-14 C]AA (0.18 M), failed to displace [1-14 C]AA from E allo of huP-GHS-2. Only at concentrations higher than 7.5 M does 2-AG ether cause any significant displacement of [1-14 C] AA from E allo . These findings provide additional evidence that 2-AG ether fails to bind efficiently to E allo of PGHS-2.
We also compared the interactions of 2-AG and 2-AG ether with nsFAs (Fig. 4). PA was previously found to activate the oxygenation of 2-AG by huPGHS-2, although the magnitude of the effect was much less than that seen with AA (15). As shown in Fig. 4, nsFAs that bind to E allo cause very similar levels of activation of both 2-AG and 2-AG ether oxygenation by huPGHS-2. These data, viewed in the nsFA, PA, on the oxygenation of 2-AG ether (Fig. 3). Increasing the ratio of 2-AG ether to PA did not change the ratio of the rates with 2-AG ether alone versus 2-AG ether plus PA. This indicates that 2-AG at any of the concentrations used in the assays fails to compete with 25 M PA for E allo . PA binds to E allo of huPGHS-2 with a K d 7.5 M, but binds only very weakly to E cat (K d >50 M) (7). PA increased the V max , but did not change the K m , of huPGHS-2 for 2-AG ether. Because it increases the rate of 2-AG ether oxygenation, PA is not competing for E cat , but rather must act via E allo ; moreover, 2-AG ether did not compete with PA for E allo at the concentrations tested. This indicates that 2-AG ether binds significantly less tightly to E allo than PA, and thus, less tightly to E allo than E cat . The K d for PA binding to E allo is 7.5 M, which is below its critical micelle concentration [25 M (7,9)], where PA effectively binds only E allo . The 2-AG ether did not displace PA from E allo when the ratio of 2-AG ether/PA was 1.0, indicating that the K d for 2-AG ether binding is significantly greater than 7.5 M and, thus, greater than the K d of 2-AG ether for E cat . The K d for 2-AG ether binding to E cat is the K m of huPGHS-2 for 2-AG [5 M (Fig. 3)]. This situation is unlike what is observed with AA (7) or EPA (10) that bind E allo 30 times more tightly than E cat of huPGHS-2.  Purified huPGHS-2 was added to initiate oxygenation reactions. O 2 consumption was monitored as described in the Experimental Procedures. Results are shown for a single experiment involving triplicate determinations. The error bars indicate the average ± SD. When control values are normalized to 100% activity, significant differences from the control value (with no ibuprofen) were seen at all the ibuprofen concentrations tested, as determined by the Student's t-test (P < 0.05). IC 50 values shown in the two panels were calculated using GraphPad Prism software. For the monophasic mode shown in the two panels in this figure, IC 50 values were calculated by using nonlinear regression to fit the data to the log [ibuprofen] versus normalized response (variable slope) curve.
We recently described a recombinant huPGHS-2 heterodimer variant, denoted as Y385F/Native huPGHS-2 (15). This variant has 90% of the activity of native huP-GHS-2, indicating that 90% of E allo monomers bear the Tyr-385Phe mutation and 90% of E cat monomers are native monomers; moreover, once E allo and E cat of huPGHS-2 are formed, they do not interconvert (15). This variant can be used as a platform to determine the effect of amino acid substitutions in the COX binding site of E allo , the subunit with the Tyr385Phe substitution, as compared with the native E cat subunit (15). Figure 6 presents the results of studies of Arg-120 substitutions in the Y385F/Native huPGHS-2 platform on the effect of IBP on 2-AG ether oxygenation; supplementary Fig. 2 presents these data, showing actual rates as opposed to the relative COX activities shown in Fig.  6. Arg-120 is the residue that interacts with the carboxylate group on substrates and inhibitors that bind within the COX active site. Inhibition of oxygenation of 2-AG ether by IBP is largely impeded by having Arg-120 substitutions in both subunits (R120A/R120A huPGHS-2), but only modestly attenuated by having an Arg120Ala substitution in E allo (i.e., Y385F R120A/Native huPGHS-2). This suggests that the context of the kinetic data indicating that 2-AG and 2-AG ether have similar properties as huPGHS-2 substrates, establish that 2-AG ether can be used as a stable surrogate for 2-AG in monitoring interactions between this latter endocannabinoid and PGHS-2.

Comparison of IBP as an inhibitor of the oxygenation of 2-AG, 2-AG ether, and AA
Many COX inhibitors are time-dependent inhibitors, but IBP is a rapidly reversible inhibitor of AA oxygenation (19,20). As first reported by Marnett and coworkers, IBP is a more potent inhibitor of the oxygenation of 2-AG than of AA by muPGHS-2 (13). According to their model, IBP binds one monomer of PGHS-2 to allosterically inhibit 2-AG oxygenation occurring in the partner catalytic monomer, whereas IBP binding to both monomers is required to inhibit AA oxygenation.
We compared the effects of IBP on the oxygenation of 2-AG, 2-AG ether, and AA by huPGHS-2. As shown in Fig. 5, instantaneous inhibition of huPGHS-2-mediated oxygenation of 2-AG and 2-AG ether occurs at similar concentrations of IBP. Fig. 6. Instantaneous inhibition by IBP of 2-AG ether oxygenation by huPGHS-2 variants having an Arg-120 substitution in E allo or E cat . Assays of O 2 consumption using an O 2 electrode were performed as described in the Experimental Procedures using 50 M 2-AG ether as the substrate. huPGHS-2 heterodimer variants were described previously (15). The designations indicate the mutations in the subunits. For example, Y385F R120A/Native indicates a PGHS-2 molecule having one subunit with both Y385F and R120A mutations and the other subunit as having no mutations (i.e., a Native subunit). R120A/R120A indicates a PGHS-2 molecule with mutations in both subunits. Results are shown for a single experiment involving triplicate determinations. The error bars indicate the average ± SD. Control values for each enzyme variant are normalized to 100%. Asterisks denote a significant difference between the value with inhibitor and the control (native) value with the same level of inhibitor as determined by the Student's t-test (P < 0.05). Specific activities for the huPGHS-2 variants with 2-AG ether were as follows: for Native, 25 units/mg; for R120A/ R120A, 17 units/mg; for Y385F/Native, 19 units/mg; for Y385F R120A/Native, 24 units/mg; and for Y385F/ R120A, 22 units/mg. 2-AG ether and ibuprofen toward E allo and E cat of PGHS-2. Figure 8 compares the specificities of a number of ligands toward E allo and E cat of both PGHS-1 and PGHS-2.
Our present studies indicate that huPGHS-2 has essentially the same kinetic properties and forms a homologous set of products with 2-AG and 2-AG ether. These two substrates have indistinguishable K m and V max values, are activated to similar extents by common nsFAs, and are inhibited to similar extents and apparently via similar mechanisms by IBP The 2-AG ether exhibits a preference for binding E cat versus E allo . These results are consistent with recent studies by Marnett and coworkers that have indicated that 2-AG binds E cat with about 10 times higher affinity than E allo (11). The K m for 2-AG ether, which is equivalent to the K d for 2-AG ether binding to E cat , is 5-10 M (15). The marked preference of 2-AG and 2-AG ether for E cat of huPGHS-2 is unlike that observed with FA substrates, such as AA and EPA, that bind E allo with 20-30 times greater affinity than E cat (7,15). Presumably neither the ester nor ether groups of 2-AG or 2-AG ether, respectively, effectively interact with Arg-120 of E allo .
Because 2-AG ether is more stable than 2-AG, which readily hydrolyzes or rearranges (12,21), 2-AG ether can be used as a stable surrogate for examining 2-AG as a COX IBP functions by binding E allo in native huPGHS-2 to cause near complete allosteric inhibition of 2-AG ether oxygenation. However, IBP appears to bind only slightly less well to E cat to cause complete inhibition when IBP cannot bind to E allo (i.e., in Y385F R120A/Native huPGHS-2).
As shown in Figs. 5, 6, 50% inhibition of 2-AG ether oxygenation occurs with 5-10 M IBP with essentially complete inhibition occurring with 30-40 M IBP. As illustrated in Fig. 5 with native huPGHS-2, the inhibition curves for 2-AG and 2-AG ether involve a single phase. Figure 7 shows the effect of IBP on AA oxygenation by huPGHS-2. Unlike that seen with 2-AG and 2-AG ether, the best fit curve is biphasic for inhibition of AA oxygenation. IBP causes approximately 50% inhibition through a higher affinity IBP binding event, with an IC 50 of about 120 M. The remainder of the inhibition occurs through a second lower affinity IBP binding event, having an IC 50 of about 1 mM (i.e., 948 M). One simple explanation for these data is that the initial and apparently incomplete phase of inhibition of AA oxygenation reflects partial allosteric inhibition upon IBP occupancy of E allo , while the second phase represents competitive IBP binding to E cat . The fact that the IC 50 for the first phase is 10-20 times higher than the IC 50 for 2-AG ether oxygenation is consistent with the idea that IBP needs to displace AA bound to E allo to cause the first phase of inhibition.

DISCUSSION
PGHSs function in solution as conformational heterodimers composed of E cat and E allo monomers. In previous studies, we have identified the E allo versus E cat binding specificities of huPGHSs toward FAs from a number of functional classes and of a variety of NSAIDs (7,8,10,15). Here we have determined the binding specificities of Fig. 7. Instantaneous inhibition of huPGHS-2-mediated oxygenation of AA by IBP. The indicated concentrations of ibuprofen were present in a standard O 2 electrode assay mixture along with 25 M AA, and purified huPGHS-2 was added to initiate oxygenation. O 2 consumption was monitored as described in the Experimental Procedures. Results are shown for a single experiment involving triplicate determinations. The error bars indicate the average ± SD. Control values for each enzyme variant are normalized to 100%. Significant differences from the control value (with no inhibitor) were seen at all IBP concentrations, as determined by the Student's t-test (P < 0.05). IC 50 values were calculated using GraphPad Prism software. For the biphasic mode shown in this figure, IC 50 values were calculated by using nonlinear regression to fit the data to a biphasic dose-response. Fig. 8. Isoform specific interactions of COX substrates, nsFAs, and COX inhibitors with huPGHS-1 and huPGHS-2. Both PGHS isoforms are sequence homodimers that function as conformational heterodimers composed of an allosteric (E allo ) and a catalytic (E cat ) subunit. The individual subunits of huPGHSs differ in their affinities for ligands and in the nature and amplitude of their effects. COX substrates are shown in blue in approximate order of their catalytic efficiencies. Note that FAs that can function as COX substrates can interfere with PGH 2 formation by competing with the most common substrate, AA, for the E cat . Ligands shown in green allosterically stimulate COX activity. Ligands shown in red interfere with COX activity either allosterically by binding E allo or competitively by binding E cat . Ligands that bind E allo can also affect responses to COX inhibitors. For example, nsFAs bound to E allo of huPGHS-1 increase the rate of aspirin acetylation, whereas celecoxib (in dark red) bound to E allo of huPGHS-1 may interfere with aspirin action. DHLA, dihomo--linolenic acid; DPA, docosapentaenoic acid; nsNSAIDs, nonspecific NSAIDs. substrate. The 2-AG ether is reportedly formed in vivo (22), but its physiologic importance is unclear (23), and it is generally less biologically potent than 2-AG itself (24).
We have previously reported that nsFAs cause about a 2-fold decrease in the K m of PGHS-2 for AA (7,9); however, the difference the between K m values with and without ns-FAs are small and are based on experiments in which it is difficult to estimate the amount of AA versus nsFA bound to E allo because the AA concentration rapidly decreases during measurements of COX activities. The 2-AG ether, because it appears not to effectively compete with nsFAs for E allo , is a useful substrate for discriminating between effects on K m versus V max . As illustrated in Fig. 3, PA primarily affects the V max and not the K m for 2-AG ether. Marnett and coworkers have reported similar results for the effect of another allosteric activator, 13-methyl AA, on 2-AG oxygenation (11). Thus, in contrast to what we had previously hypothesized (7), we now suspect that the effect of nsFAs to stimulate AA oxygenation by huPGHS-2 results primarily from an effect on the V max for AA. This is also consistent with naproxen being a negative allosteric regulator of huPGHS-2 by reducing the V max without changing the K m for AA (7).
Marnett and coworkers reported that IBP causes a noncompetitive inhibition of 2-AG oxygenation that they attributed to a high affinity binding of one molecule of IBP to muPGHS-2 (13). Additionally, they indicated that IBP is a competitive inhibitor of AA oxygenation by muPGHS-2 and that inhibition of AA oxygenation involves the binding of two IBP molecules per dimer. Our results on the inhibition of 2-AG ether oxygenation by IBP parallel those with 2-AG reported by the Marnett laboratory. Studies with huPGHS-2 variants having Arg120Ala substitutions in E allo and/or E cat indicate that IBP can completely inhibit the oxygenation of 2-AG ether by binding either E allo or E cat . However, IBP is slightly better able to bind E allo than E cat in the presence of 2-AG ether. In contrast to what was reported for muPGHS-2 (13), we find that IBP inhibition of AA oxygenation by huP-GHS-2 is a mixed inhibition involving allosteric and competitive components, probably mediated by sequential binding of IBP to E allo then E cat . The basis for the species difference in IBP inhibition of AA oxygenation by muP-GHS-2 versus huPGHS-1 is unclear. The allosteric inhibitory effect of IBP on AA oxygenation may simply be much more pronounced with huPGHS-2 than with muPGHS-2.