Cyclooxygenase Allosterism, Fatty Acid-mediated Cross-talk between Monomers of Cyclooxygenase Homodimers*

Prostaglandin endoperoxide H synthases (PGHSs) 1 and 2, also known as cyclooxygenases (COXs), catalyze the oxygenation of arachidonic acid (AA) in the committed step in prostaglandin (PG) biosynthesis. PGHSs are homodimers that display half of sites COX activity with AA; thus, PGHSs function as conformational heterodimers. Here we show that, during catalysis, fatty acids (FAs) are bound at both COX sites of a PGHS-2 dimer. Initially, an FA binds with high affinity to one COX site of an unoccupied homodimer. This monomer becomes an allosteric monomer, and it causes the partner monomer to become the catalytic monomer that oxygenates AA. A variety of FAs can bind with high affinity to the COX site of the monomer that becomes the allosteric monomer. Importantly, the efficiency of AA oxygenation is determined by the nature of the FA bound to the allosteric monomer. When tested with low concentrations of saturated and monounsaturated FAs (e.g. oleic acid), the rates of AA oxygenation are typically 1.5-2 times higher with PGHS-2 than with PGHS-1. These different kinetic behaviors of PGHSs may account for the ability of PGHS-2 but not PGHS-1 to efficiently oxygenate AA in intact cells when AA is a small fraction of the FA pool such as during “late phase” PG synthesis.

Prostaglandin endoperoxide H synthases (PGHSs) 2 also known generically as cyclooxygenases (COXs) convert arachidonic acid (AA), two O 2 molecules, and two electrons from one or more unknown reductants to prostaglandin H 2 (PGH 2 ) (1)(2)(3)(4)(5). The conversion involves two steps. First, the cyclooxygenase activity of PGHSs catalyzes the introduction of two O 2 molecules into the AA backbone forming prostaglandin G 2 . Prostaglandin G 2 can be reduced to PGH 2 by the peroxidase (POX) activity of the enzyme. There is a functional interplay between the COX and POX sites in that the heme group at the POX site needs to be oxidized by a peroxide to oxidize Tyr-385 in the active COX site to a tyrosyl radical. The Tyr-385 radical abstracts the 13-proS hydrogen from AA to initiate the COX reaction in the rate-determining step.
PGHSs are homodimers with monomer molecular masses of ϳ72 kDa (1). Previous studies have indicated that there is crosstalk between the COX activities of the partner monomers and that the COX function exhibits half of sites activity with AA (6). That is, only one monomer of a dimer catalyzes AA oxygenation at any given time. Furthermore, the binding of inhibitors of the 2-phenylpropionic acid class such as flurbiprofen (FBP) or ibuprofen to one COX site is sufficient to inhibit all COX activity of a dimer (6,7). The existence of half of sites activity establishes that PGHSs are conformational heterodimers when the rate-determining step in catalysis occurs.
Despite exhibiting half of sites activity, there has been little evidence for cooperativity between the two COX sites during substrate turnover, and no physiological importance has been attached to the half of sites activity. However, we recently found in testing PGHS-2 with AA and another COX substrate eicosapentaenoic acid (EPA) that the enzyme showed a marked preference for AA (8). This was not predicted from the kinetic parameters determined for AA and EPA individually with PGHS-2. Also surprising was the observation that PGHS-1 differed from PGHS-2 in that PGHS-1 exhibits no apparent selectivity toward AA in the presence of EPA.
To investigate the mechanistic basis for half of sites COX activity and the basis for the selectivity of PGHS-2 for AA, we sought to identify segments of the interface between the monomers that become altered during ligand binding. We approached this by inserting cysteine residues at positions that appeared to be across from one another in the crystal structures. We determined what substitutions led to cross-linking and then determined which FAs and inhibitors would affect the cross-linking between cysteine residues located on partner monomers of dimers. In the process, we found that fatty acids (FAs) can bind to both COX sites of a dimer at the same time and that binding of FAs to one COX site can alter the activity of the enzyme at the COX site of the partner monomer. These biochemical findings led us to speculate that the cross-talk between monomers serves as a way for PGHS-2 to exhibit selectivity toward AA even when AA is a minor component of the available FA pool. In contrast, PGHS-1 does not exhibit selectivity toward AA, so during metabolic conditions in which AA is a small fraction of available FAs, PGHS-1 is effectively latent.

EXPERIMENTAL PROCEDURES
Materials-All fatty acids and their derivatives except 20:19 were purchased from Cayman Chemical Co. Diclofenac, meclofenamic acid, 1,10-phenanthroline, FLAG peptide, FLAG-affinity resin, hemin, 20:19, and FBP were from Sigma. Co 3ϩ -protoporphyrin IX was from Frontier Scientific. Horseradish peroxidase-conjugated secondary antibodies (goat antirabbit IgG and goat anti-mouse IgG) were from Bio-Rad. The detergent C 10 E 6 was from Anatrace. BCA protein reagent was from Pierce. [1-14 C]AA was from American Radiolabeled Chemicals. Hexanes, isopropanol, and acetic acid were HPLC grade from Fisher. Complete protease inhibitor was from Roche Applied Science. Restriction enzymes were from New England Biolabs, Inc. Nickel-nitrilotriacetic acid was from Qiagen. All other materials were analytical grade from Sigma.
Synthesis of 5Z,11Z,14Z-and 5Z,8Z,14Z-Eicosatrienoic Acids-In the first step of the synthesis AA was partially reduced with hydrazine 8.0 g of methyl ester of AA (25 mmol) in 200 ml of ethyl alcohol was mixed with 22 ml of hydrazine monohydrate (0.44 mole, 17.5 eq), and 30 ml of 50% hydrogen peroxide (0.44 mole, 17.5 eq) was added. The reaction mixture was stirred, and its temperature was kept at ϩ15 to ϩ20°C. After 75 min, 300 ml of distilled water was added to the reaction mixture, which was then acidified to pH ϳ 2 with 2 N HCl and extracted with ethyl acetate (2 ϫ 400 ml). The combined organic layers were washed sequentially with distilled water (3 ϫ 300 ml), saturated NaCl (2 ϫ 200 ml) and dried over 50 g of anhydrous Na 2 SO 4 with stirring for 30 min. The resulting extract was evaporated under vacuum, and the dry residue was purified by column chromatography on 100 g of silica gel (40 -63 m) eluting with 5% ethyl acetate in 95% hexane (v/v). Preparative reversed-phase HPLC was used to separate the fatty acid methyl esters (7.34 g) present in the resulting colorless oil mixture of fatty acid methyl esters; the HPLC conditions were as follows: column Phenomenex Luna-C 18 (2) (22 ϫ 250 mm), 5 m with a mobile phase of acetonitrile-water-acetic acid (65:35:0.1, v/v) and a flow rate of 30 ml/min. The methyl esters were then saponified with KOH in ethanol-water to generate the free fatty acid as described recently (9). The elution order of the product fatty acids, determined by analytical HPLC on a C 18 reversed-phase column (Phenomenex Luna-C 18 (2), 4.6 ϫ 250 mm, 5 m), with a mobile phase of acetonitrile-water-acetic acid (75:25:0.1, v/v) and flow rate of 1.5 ml/min, was dihomo-␥-linolenic acid, 5Z,11Z,14Z-eicosatrienoic (kЈ ϭ 6.3 min), 5Z,8Z,14Z-isomer (kЈ ϭ 6.6 min), and then Mead acid.
The structure assignments of the isolated fatty acids were based on the following observations. Neither isomer exhibited any UV-absorption above 210 nm indicating that the double bonds are not conjugated and in their Fourier transform-IR spectra (film) no absorption band for a trans-double bond was detected (800 -1000 cm Ϫ1 ). 1 H and 13 C NMR analyses did not permit us to distinguish between the isomers. Both the isomers produced only individual single peaks with different retention times when examined by capillary gas-liquid chromatography analysis on three different stationary phases with various polarities: Restec Rxi-1 (90 m, 0.25 mm inner diameter, 0.25 m); Agilent HP-Innowax (30 m, 0.32 mm inner diameter, 0.25 m), and Restec RT-2560 (100 m, 0.25 mm inner diameter, 0.2 m). The positions of the double bonds in the isomers were located by gas chromatography/mass spectrometry analysis of the pyrrolidides derived from them according to the procedure described previously (10). Thus, in the mass spectrum of the pyrrolidide of 5Z,11Z,14Z-eicosatrienoic acid, we observed a relatively strong signal for the molecular ion at the predicted m/z of 359 (5%) and a set of signals of diagnostic fragmentations at m/z (I%) of 181 (4%), 195 (3%), and 208 (3%); these data establish that there is no double bond at the ⌬8 position of this isomer. In contrast, we observed in the mass spectrum of pyrrolidide of 5Z,8Z,14Z-eicosatrienoic acid a set of signals of diagnostic fragmentations at m/z (I%) of 180 (4%), 192 (2%), and 206 (5%) that clearly demonstrate the absence of a double bond at the ⌬11 position of this isomer (presence of the molecular ion at the predicted m/z of 359 (5%) proves the structure assigned).
Synthesis of 14Z-Eicosenoic Acid-14Z-Eicosenoic acid was prepared by Wittig coupling between the methyl ester of 14-oxotetradecanoic acid and commercially available n-hexyltriphenylphosphonium bromide using sodium hexamethyldisilazide as a base. The methyl ester of 14-oxotetradecanoic acid was synthesized from commercially available 12-bromo-1-dodecanol by C2-elongation using the technique reported earlier (11) and subsequent oxidation of the resulting methyl ester of 14-hydroxytetradecanoic acid to the target compound with pyridinium chlorochromate in dichloromethane (12).
14Z-Eicosenoic acid was synthesized from 14-oxotetradecanoic acid as follows: to a suspension of 9.2 g (21 mmol, 1.1 eq) of n-hexyltriphenylphosphonium bromide in 200 ml of dry tetrahydrofuran at Ϫ30°C (acetone-dry ice bath), 12.2 ml of a 2 M solution of sodium hexamethyldisilazide (24 mmol, 1.25 eq) in tetrahydrofuran was added with stirring under dry nitrogen. The resulting orange solution was stirred for 30 min, and after that the flask was cooled to Ϫ78°C and 5.0 g of the methyl ester of 14-oxotetradecanoic acid (19.5 mmol) in 25 ml of tetrahydrafuran was added dropwise. The reaction mixture was stirred at Ϫ78°C for 45 min after which the bath was removed and the reaction mixture was stirred for additional 45 min, evaporated under vacuum and the residue was extracted with hexane (4 ϫ 100 ml). The combined hexane layers were evaporated under vacuum and purified by column chromatography on 50 g of silica gel (40 -63 m) eluting with 3% ethyl acetate in 95% hexane (v/v). The fraction containing methyl ester of the target 14Z-eicosenoic acid were combined and evaporated under vacuum to yield 3.32 g (52%) of a mixture of methyl 14Z-eicosenoate (94% by gas-liquid chromatography) and methyl 14E-eicosenoate (5% by gas-liquid chromatography). The methyl esters were saponified with KOH in ethanol-water as described previously for polyunsaturated fatty acids (9) and separated by preparative HPLC on a C 18 -reverse phase column (22 ϫ 250 mm, 5 m) with a mobile phase of acetonitrile-water-acetic acid (85:15:0.1, v/v) and a flow rate 30 ml/min to yield 98ϩ% of the target fatty acid as a colorless wax.
Construction of pFastBac PGHS-2 Mutations or pFastBac Dual for Heterodimer Expression-All human (hu) PGHS-2 mutations were performed using a modified pFastBac vector (Invitrogen). A QuikChange site-directed mutagenesis kit (Stratagene) was used to create mutations, which were con-firmed by sequencing by the University of Michigan DNA Sequencing Core. The construction of expression vectors encoding PGHS-2 heterodimers was performed in two steps (6). In the first step the desired DNA fragments with hexahistidine (His 6 ) tags were harvested from pFastBac using StuI-KpnI and subcloned into the Pp 10 site of pFastBacDual, which had been digested with SmaI-KpnI. In the second step, the other DNA fragment with a FLAG epitope tag also derived from pFastBac by treatment with EcoRI-HindIII was subcloned into the Pp H site of pFastBacDual, which had been cleaved with the same restriction enzymes. The correct constructs were confirmed by restriction digestion.
PGHS-2 Expression and Purification-Protein expression from pFastBac or pFastBacDual expression vectors was performed according to the instructions of the manufacturer for the Invitrogen Bac-to-Bac expression system. Sf21 cells at a density of 1.2-1.5 ϫ 10 6 cells/ml were infected with recombinant virus at a multiplicity of infection of ϳ0.1. After expression for 3-4 days, the cells were harvested, washed with phosphatebuffered saline, and stored at Ϫ80°C.
For purification of heterodimers, solubilized supernatants were prepared as described above for the purification of homodimers. Subsequently, a two-step process was used to purify the heterodimers. First, the solubilized supernatant was purified using chromatography on fast-flow nickel-nitrilotriacetic acid resin. Second, the nickel-nitrilotriacetic acid elution fraction, which contained doubly His 6 -tagged homodimers and heterodimers that were both FLAG-and His 6 -tagged, was subjected to buffer-exchange with Buffer F (50 mM Tris-HCl, pH 7.5, containing 150 mM NaCl) and further purified using FLAG affinity resin following the protocol of the manufacturer. PGHS-2 heterodimers containing both FLAG and His 6 tags were eluted with FLAG peptide (100 g/ml) in Buffer F. All protein eluants were subjected to buffer-exchange with buffer E and kept at 4°C. Protein purity was estimated visually by SDS-PAGE and staining with Coomassie Brilliant Blue G-250. Pro-tein concentrations were determined using BCA reagent with bovine serum albumin as a standard.
Disulfide Cross-linking between Monomers of PGHS-2-Before cross-linking all purified PGHS-2 variants containing cysteine substitutions (1 M of monomer) were reconstituted with a 1.5 molar excess of Co 3ϩ -protoporphyrin IX. Cross-linking to form disulfide bonds was initiated by making the solution to 250 M Cu(II)/(1,10-phenathroline) 3 (i.e. 250 M CuSO 4 plus 625 M 1,10-phenathroline) at room temperature for various times. To test the effects of various ligands (FAs or inhibitors) on cross-linking, the proteins were preincubated with different ligands for 10 min. The reactions were stopped by adding SDSloading buffer (Invitrogen) containing 20 mM N-ethylmaleimide and 20 mM EDTA (and no reductants). The reaction mixtures were subjected to SDS-PAGE using 7% or 10% gels, and immunoblotting was performed with a rabbit polyclonal antibody reactive with against huPGHS-2 (13).
Determination of Critical Micelle Concentrations-CMC values for AA and several other FAs were determined using a fluorescent dye and/or isothermal titration calorimetry. Using a fluorescent dye, a series of concentrations of AA were prepared in 100 mM sodium phosphate, pH 7.6, containing 5 M 1,6diphenyl-1,3,5-hexatriene. The samples were incubated in the dark at room temperature for 30 min before fluorescence measurements were performed. The excitation wavelength for 1,6diphenyl-1,3,5-hexatriene was 358 nm, and the intensities at 430 nm were recorded. CMC values were estimated as described by Chattopadhyay and London (14). Microcalorimetry for determination of CMC values was performed on a VP-Isothermal Titration Calorimeter (MicroCal, Amherst, MA) with a computer-controlled stirrer syringe. Typically, 8-l aliquots of a FA suspension in 100 mM sodium phosphate, pH 7.6, were injected into a 1.42-ml reaction chamber containing the same buffer at 37°C. The dissociation of FAs from the injected micelles was monitored as the heat of de-micellization. At a concentration unique for each FA, the dissociation into individual FAs from the injected micelles no longer occurs and CMC values can be calculated.
COX and Peroxidase Activity Measurements-Oxygen electrode assays for COX activity were performed as detailed in previous reports (6,15,16). COX assays of purified enzymes utilizing radio-TLC assays of PGH 2 formation were performed using [1-14 C]AA as described previously (16). POX activity was measured spectrophotometrically using 0.1 mM H 2 O 2 and 4 mM guaiacol as substrates (6). Tables 1 and 2 compare the effects of various FAs, some COX substrates and some not, on the COX activities of purified ovPGHS-1 and huPGHS-2 with AA. In Table 1, a combination of 100 M FA and 20 M AA was assayed for COX activity polarographically with an O 2 electrode and, in some cases, by radio-TLC by measuring the conversion of [1-14 C]AA to oxygenated, radioactive products. The data in Table 2 are from O 2 electrode measurements only and were performed with lower concentrations of FAs (0.5-1.0 M) and AA (2 M). The polarographic assay is accurate but does discriminate between oxygenation of AA versus the comparator FA, if the comparator FA is also a COX substrate. Although providing less precise rate data, the radio-TLC assay directly measures AA oxygenation.

Effects of FAs on COX Activities of PGHSs-
The rate of AA oxygenation by PGHS-2 was generally equal to or greater than that for PGHS-1 when AA oxygenation was measured in the presence of FAs that are not COX substrates. The one anomaly was with 100 M stearic acid and 20 M AA (Table 1); however, with 1 M stearic acid and 2 M AA (Table  2), the rate was greater with PGHS-2 than PGHS-1. The measurements performed with 1 M FA and 2 M AA are likely to be more physiologically representative, because the AA concentration that is estimated to be available to PGHSs in intact cells is 50 nM to 2 M (17, 18). 3 For FAs that can serve as COX substrates, the radio-TLC assays of Table 1 provide the only valid comparisons of rates of AA oxygenation. Consistent with our earlier report (8), EPA inhibits AA oxygenation by PGHS-1 to a greater extent than PGHS-2. This was also the trend with linoleic acid and docosahexaenoic acid. Large differences between the values obtained with O 2 electrode and radio-TLC assays reflect preferential use of the comparator FA versus AA. With physiologically relevant FAs, this phenomenon is most obvious with dihomo-␥-linolenic acid and adrenic acid both of which are 6 FAs. Finally, ⌬ 5,8,14 -eicosatrienoic acid, a newly synthesized FA, was a particularly good inhibitor of both ovPGHS-1 and huPGHS-2. It should be noted that the major product formed from [1-14 C]AA alone was [1-14 C]PGH 2 , and that none of the FAs at the concentrations used for the experiments presented in Tables 1 or 2  TABLE 1

Effects of high concentrations of various FAs on AA oxygenation by PGHS-1 versus PGHS-2
Results are shown as percentages of rates for O 2 electrode assays of COX activities of PGHSs with FAs alone (100 M) and FAs (100 M) in combination with AA (20 M) except in the case of ⌬ 5,8,14 -5,8,14 -20:3, which was tested at a concentration of 40 M. Results from radio-TLC assays are also shown for certain FAs (100 M) in combination with ͓1-14 C͔AA (20 M). Details of the experiments are provided under "Experimental Procedures." The average specific activities of preparations used in the experiments of purified ovPGHS-1 and huPGHS-2 with AA averaged 40 and 42 nmol of O 2 consumed per min per mg of protein, respectively, under standard assay conditions. Values are from at least triplicate assays, except for Mead acid and ⌬ 5,11,14 -20:3, which was assayed in duplicate, using different preparations of purified proteins. An unpaired Student's t-test, and standard errors were calculated using Microsoft Excel. ND, not determined. FAs that are not cyclooxygenase substrates Palmitic acid (16:    changed this profile of radioactive products. We emphasize this because it has been reported (19) that very high (i.e. mM) concentrations of non-substrate FAs in combination with AA cause a relative increase in the formation of hydroxyeicosatetraenoic acids from AA via PGHS-1.
Overall, the data in Tables 1 and 2 suggest that PGHS-2 oxygenates AA more efficiently than does PGHS-1 when common FAs other than adrenic acid are present. ovPGHS-1 was compared with huPGHS-2 because of the ease of handling of these particular enzyme forms. The FA substrate specificity of microsomal huPGHS-1 when tested with AA, dihomo-␥-linolenic acid, EPA, linoleic acid, and ␣-linolenic acid exhibits relative V max values (20) that are very similar to those found here for ovPGHS-1 (Table 1).
To learn more about the effects of FAs on AA oxygenation by PGHS-2 we tested the effects of 20:19. The CMC for 20:19 was determined to be between 1 and 10 M (data not shown), and values in this range would also be anticipated for palmitic acid and oleic acid. Both 1 M (e.g. Table 2) and 10 M 20:19 ( Fig.  1) had similar effects on the oxygenation of AA by PGHS-2 consistently causing a 2-to 3-fold decrease in the K m . The K m for AA with and without 10 M 20:19 calculated from the data in Fig. 1 was 3.6 M versus 9.2 M, respectively. The stimulatory effect of 20:19 on PGHS-2 was overcome as the concentration of AA was increased. A smaller (Ͻ20%) and non-significant decrease in the K m for AA was observed when 20:1 9 was tested with PGHS-1 (e.g. Tables 1 and 2).
PGHSs undergo a substrate turnover-dependent, reactionbased suicide inactivation (15). In principle, performing a reaction with AA in the presence of another FA like 20:19 could diminish the rate of suicide inactivation and have the effect of increasing the initial rate of AA oxygenation. However, there was no significant difference between the first order rates of inactivation of huPGHS-2 with 20 M AA in the presence or absence of 100 M 20:19: average t1 ⁄ 2 values of 17 s and 19 s, respectively.
Previous studies of PGHS dimers have indicated that the enzymes exhibit half of sites catalytic activity (6,7) and that the monomers are tightly associated (21). Thus, PGHSs must be conformational heterodimers during catalysis and inhibition and cross-talk must occur between partner monomers. Viewed in this context, our kinetic data suggest that the effects of FAs on PGHSs result from the binding of these FAs to one COX active site of a PGHS dimer to affect the catalytic efficiency of the partner monomer with AA. This concept is supported by the results of experiments described below in which we characterized cross-talk between PGHS-2 monomers.
Effects of Ligands on the Cross-linking of PGHS-2 Monomers-To delineate the location(s) at the dimer interface where crosstalk occurs, we first prepared a C313S C540S PGHS-2 mutant (⌬C-huPGHS-2) that lacks free cysteine residues, then substituted pairs of cysteines at five different positions along the interface between monomers and finally examined disulfide bond formation between monomers promoted by an oxidant Cu 2ϩ /o-phenanthroline (CuP) ( Table 3). The five locations where cysteines were introduced were chosen on the basis of PGHS crystal structures (22,23) where it appeared that crosslinking might occur. The cysteine-less mutant ⌬C-huPGHS-2 expressed well and had the same specific COX activity as native

TABLE 3 Cross-linking between PGHS-2 monomers having cysteine residues substituted into the dimer interface
The indicated mutant proteins were expressed in baculovirus, purified, and assayed for COX specific activities as described under "Experimental Procedures." Crosslinking (i.e. disulfide bond formation) between monomers was monitored by the appearance of higher molecular mass species following treatment with the oxidant Cu 2ϩ /o-phenanthroline and SDS-PAGE in the absence of a reducing agent as detailed under "Experimental Procedures." Values for specific activities are averages of two determinations that varied by Ͻ10%. Mutants for which cross-linking was not observed or cross-linking was not inhibited by FBP were examined only one time. ϩ, cross-linking occurred; Ϫ, cross-linking did not occur.

huPGHS-2 variant
Starting COX specific activity Cross-linking Inhibited by FBP enzyme. Mutants designed to explore cross-linking between two of the five regions, including T60C S546C ⌬C-huPGHS-2 and Q370C Q372C ⌬C-huPGHS-2 did not express well in a baculovirus expression system and were not examined further.
An S138C L334C ⌬C-huPGHS-2 was expressed, but did not undergo cross-linking in response to CuP. Mutants such as V50C E322C ⌬C-PGHS-2 exhibited ϳ60% of native PGHS-2 activity and underwent CuP-induced cross-linking, but the cross-linking was not affected by FBP. Importantly, substituting cysteine residues at positions 126 or 127 and positions 541 or 543 of each monomer of ⌬C-huPGHS-2 led to the formation of dimers, and cross-linking was attenuated by FBP (Table 3 and Fig.  2). As shown in Fig. 2 for P127C S541C ⌬C-huPGHS-2, dimers were formed in a time-dependent manner upon treatment with CuP. P127C S541C ⌬C-huPGHS-2 had Ͻ60% of the activity of native enzyme (Table 3), but the specific activity of P127C S541C ⌬C-huPGHS-2 was not affected by CuPinduced cross-linking. Three immunoreactive bands are observed with P127C S541C ⌬C-huPGHS-2 even before treatment with CuP (Fig. 2). This is because some spontaneous cross-linking occurs between monomers during expression and isolation of the enzyme. The three immunoreactive bands observed in the Western blot panels in Fig. 2 correspond, in order of greatest mobility, to the PGHS-2 monomer and singly and doubly cross-linked PGHS-2 partner monomers. The experimental evidence for this is presented in Fig. 3 and is as follows. The most rapidly migrating band has an apparent molecular mass of 72 kDa corresponding to that expected for a PGHS-2 monomer. The two slower migrating bands found with S126C A543C ⌬C-huPGHS-2 collapse into a single 72-kDa band when the protein is treated with 100 M dithiothreitol or 200 M 2-mercaptoethanol, but not upon treatment with 6 M urea (Fig. 3A). This indicates that the two more slowly migrating, upper bands in the panels in Figs. 2 and 3A are crosslinked via disulfide bond(s). The more rapidly migrating crosslinked species appears to have a single disulfide bond, because it has the same mobility as a cross-linked ⌬C-huPGHS-2 heterodimer prepared such that one monomer contains only the P127C mutation, and the partner has only the S541C mutation (designated P127C ⌬C-/S541C ⌬C-huPGHS-2 in Fig. 3B). Moreover, following extended treatment with CuP virtually all the monomers of the P127C S541C ⌬C-huPGHS-2 homodimer are converted to doubly cross-linked dimers (Figs. 2 and 3B).
In the case of the P127C S541C ⌬C-PGHS-2, the activity was ϳ55% of native enzyme, and treatment with CuP did not reduce activity; with S126C A543C ⌬C-PGHS-2, the activity was 65% that of native enzyme and cross-linking caused a 60% decrease in activity. Importantly, AA oxygenation by P127C S541C ⌬C-PGHS-2 was stimulated in a similar though less robust  manner as native huPGHS-2 by 20:19 (Fig. 4). The K m values for P127C S541C ⌬C-huPGHS-2 with AA were 9.1 M and 14 M with and without 20:19, respectively; we suspect that the lesser kinetic effect of 20:19 on the P127C S541C ⌬C-huPGHS-2 mutant is because a fraction of this mutant (ϳ30 -50%) has already become cross-linked during its isolation. Importantly, a P127C S541C ⌬C-/G533A P127C S541C ⌬C-huPGHS-2 heterodimer that has only one COX-active subunit, had Ͼ80% of the COX activity as the P127C S541C ⌬C-huPGHS-2 homodimer (Fig. 5). This indicates that P127C S541C ⌬C-huPGHS-2, like native huPGHS-2 exhibits half of sites COX activity with AA (6). Because of its kinetic similarity to native huPGHS-2, P127C S541C ⌬C-huPGHS-2 was used for most of our subsequent cross-linking studies.
Certain COX-2 substrates, including AA and EPA, attenuated cross-linking at concentrations at or near their K m values; however, linoleic acid and docosahexaenoic acid were ineffective (Fig. 6A). Oleic acid, which is not a COX substrate, also was without any effect on cross-linking. Aspirin (2 mM) had relatively little effect on cross-linking, but all other common nonsteroidal anti-inflammatory drugs (NSAIDs), which are competitive cyclooxygenase inhibitors, did inhibit CuP-induced cross-linking at concentrations of Յ5 M (Fig. 6B); included in this group of compounds are indomethacin, diclofenac, ibuprofen, FBP, and meclofenamate. Two COX-2-specific inhibitors, NS398 and SC58125, also inhibited cross-linking (data shown below). The experiments depicted in Fig. 6 (A and B) were performed with apo-enzyme. However, cross-linking reactions of P127C S541C ⌬C-huPGHS-2, and the patterns of inhibition of    cross-linking by NSAIDs and substrates were the same whether the cross-linking reactions were performed in the presence or absence of Co 3ϩ -heme; Co 3ϩ -heme can substitute for native heme in binding to the POX site of PGHSs, but the enzyme is inert because Co 3ϩ -heme cannot undergo the oxidation reaction necessary to activate the enzyme (24). Several FAs that are not COX substrates, including 20:19, failed to inhibit cross-linking when tested alone but interfered with the ability of AA to prevent cross-linking ( Fig. 7 and Table  4). Under conditions in which 20:19 interferes with the ability of AA to prevent cross-linking (Fig. 7), 20:19 stimulates AA oxygenation ( Figs. 1 and 4). These findings imply that AA is able to prevent cross-linking only if bound to both COX sites of PGHS-2 and that during catalysis AA is bound at both COX sites, although it is only being oxygenated at one site at any given time; moreover, when 20:19 and AA are incubated together with PGHS-2 at the appropriate concentrations, AA is oxygenated at the COX site of one PGHS-2 monomer while 20:19 is bound to the COX site of the partner monomer.
A number of free FAs as well as esters and amides of AA were tested both alone and in combination with AA for their abilities to block CuP-induced cross-linking of P127C S541C ⌬C-PGHS-2 in experiments like the one depicted in Fig. 7. The data are summarized in Table 4. Most straight-chain FAs having three or more double bonds and 18 -22 carbons inhibited cross-linking when tested alone. Conversely, FAs having only one or two double bonds did not by themselves inhibit crosslinking. However, most of these latter FAs, including palmitic acid and oleic acid, did prevent AA from inhibiting cross-linking. Esters, ethers, and amides of AA by themselves failed to inhibit cross-linking at concentrations as high as 100 M. The lack of protection from cross-linking observed with 2-AG suggests that there are differences in the ways that 2-AG and huPGHS-2 and AA and huPGHS-2 interact. Collectively, the cross-linking data in Table 4 suggest that at concentrations of 50 M all straight-chain FAs having 16 -22 carbons can bind at least one, and in some cases both, COX sites of a PGHS-2 homodimer.

DISCUSSION
There is a single high affinity binding site on the PGHS-2 dimer for NSAIDs of the 2-phenylpropionate class, including FBP and ibuprofen (6), and most NSAIDs inhibit COX activity upon binding one COX site of a PGHS-1 homodimer (7). Based on these observations and because PGHS-2 exhibits half of sites COX activity with AA, we had anticipated that during catalysis a substrate molecule would be bound to one COX site but that no FA would be bound at the COX site of the partner monomer (6). However, our present results indicate that both COX sites of a native PGHS-2 dimer are occupied by FAs during AA oxygenation (Fig. 8). Additionally, one monomer binds a FA with relatively high affinity and acts as an "allosteric" monomer by modulating the catalytic efficiency of its partner "catalytic" monomer.
Our most well developed evidence for the model depicted in Fig. 8   ⌬C-huPGHS-2 homodimer are also most simply interpreted by having both COX sites occupied by FAs. In the absence of 20:19, AA, at concentrations near its K m value with PGHS-2, must bind both COX sites of a dimer simultaneously to prevent cross-linking. Common FAs that have not heretofore been associated with eicosanoid metabolism, including palmitic acid, stearic acid, and oleic acid, are among the FAs that bind to the allosteric monomer and stimulate AA oxygenation in the catalytic monomer of PGHS-1 and/or PGHS-2 (Tables 1 and 2). The effects of palmitic and oleic acids on the kinetics of AA oxygenation and on cross-linking of P127C S541C ⌬C-huPGHS-2 homodimer are similar to those observed with 20:19 and are readily explained by the model depicted in Fig. 8 We have not resolved whether (a) FAs that have no apparent effect on AA oxygenation (e.g. EPA) or (b) 6 FAs that are substrates even in the presence of AA (Table 1) compete effectively with AA for binding to the allosteric site. It is also unclear how certain FAs inhibit PGHS activity. They may inhibit AA oxygenation by binding to the allosteric site or by competing with AA for the catalytic site or both.
The affinity of AA for the allosteric site must be at least somewhat lower than for the catalytic site in order for PGHS-2 to display what appears to be a conventional hyperbolic kinetic curve with AA. Kinetic analyses with 1 and 10 M 20:19 indicate that this FA has a higher affinity than AA for the allosteric site but that once bound to the allosteric site, 20:19 does not compete effectively with AA for the catalytic site of PGHS-2.
Although we have uncovered what appears to be a general phenomenon regarding the regulation of PGHS-2 by FAs, we have no clear idea about what structural features of FAs underlie their functions as allosteric regulators nor which protein residues mediate the interplay between the monomers. All PGHS x-ray structures reported to date have been performed with crystals grown at high enzyme (ϳ75 M) and excess ligand (ϳ500 M) concentrations (22)(23)(24)(25)(26)(27)(28). Ligands are seen to occupy both monomers, and there are no obvious structural differences between the monomers. It is not known if this is because the resolutions are inadequate and the differences between monomers subtle or because the individual monomers represent average structures of two different monomers. There is also no high resolution structure of a ligand-free PGHS useful for comparison with ligand-occupied PGHSs. Thus, we can only speculate about the structural changes that occur upon binding of a FA to the allosteric monomer. It is clear that Pro-127 is located in a loop emanating from a helix involving Arg-120, a residue known to interact with FA substrates and many inhibitors of PGHSs. One can imagine a ligand-induced change in the position of Arg-120 that could, in turn, alter the position of Pro-127 and neighboring residues. This structural change is presumably transmitted to the partner, catalytic monomer so that Tyr-385 and the 13-proS-hydrogen of AA become optimally juxtaposed for efficient hydrogen abstraction.
Unlike the situation with PGHS-2, examining cross-talk between monomers of PGHS-1 is technically challenging; for example, a cysteine-less version of this protein that in principle could be employed for cross-linking studies lacks activity (29). We have observed that in the presence of most other common and naturally occurring FAs (i.e. 16:0, 18:0, 18:19, 18:29, and 20:53) PGHS-1 has less activity with AA than does PGHS-2. Kinetic studies with 20:19 and PGHS-1 indicate that 20:19 stimulates PGHS-1 COX activity. Accordingly, we speculate that FAs do bind to an "allosteric" monomer of PGHS-1 but that this binding does not foster the use of 20-and 22-carbon 6 FA substrates.
We have not established whether allosteric regulation of PGHS-1 and/or PGHS-2 by FAs is physiologically significant. An intriguing possibility is that the differences in the efficiency with which PGHS-1 and PGHS-2 oxygenate AA when AA is present in mixtures of FAs having different compositions may be of physiological importance. For example, PGHS-1 can oxygenate AA efficiently in so-called "early phase" cellular PG synthesis but not in "late phase" synthesis (30 -33). In early phase synthesis, only PGHS-1 is present in cells. A bolus of AA is mobilized by cytosolic phospholipase A 2 , and because of the FA specificity of cytosolic phospholipase A 2 , AA probably represents a large fraction of the free FAs available to PGHS-1 in this early phase. However, in late phase synthesis both PGHS-1 and PGHS-2 are present and AA is mobilized in part through the action(s) of nonspecific, secretory phospholipase A 2 (s). Under these conditions AA presumably comprises a much smaller part of the FA pool (32, 34 -40). PGHS-1 may operate efficiently on AA in the early phase but be inhibited by other, competing FAs in the late phase. In contrast, PGHS-2 could catalyze AA oxygenation in the late phase, because PGHS-2 can oxygenate AA even in the presence of relatively high concentrations of other FAs (5). Studies with cultured cells under conditions of early and late phase PG will be important in testing these concepts.