Conversion of Prostaglandin G/H Synthase-1 into an Enzyme Sensitive to PGHS-2-selective Inhibitors by a Double His 513 3 Arg and Ile 523 3 Val Mutation*

Modeling of the active site of prostaglandin G/H syn- thase-2 (PGHS-2) onto PGHS-1 utilizing the known crystal structure of PGHS-1 shows that the only residues impinging directly on the active site that were not conserved in the two enzymes are His 513 and Ile 523 of PGHS-1 (Arg 499 and Val 509 of PGHS-2). These residues of human PGHS-1 were each mutated to the corresponding PGHS-2 residues (His 513 3 Arg and Ile 523 3 Val) and a double mutant (His 513 3 Arg,Ile 523 3 Val) containing both residues was also constructed. The mutant enzyme forms were expressed in COS-7 cells, and their properties were compared with those of the normal isoforms using microsomal membranes. The mutated enzyme forms all had apparent K m values within 1.4-fold that of the wild type enzyme, and the specific activity of the mutants were within 2-fold of that of PGHS-1. DuP697, NS-398, DFU, and SC-58125 are selective PGHS-2 inhibitors that act as time-dependent inhibitors of PGHS-2 and rapidly reversible competitive inhibitors of PGHS-1. The single Ile 523 3 Val mutation increased the sensitivity to each of these selective inhibitors with most of the effect detected using instantaneous inhibition assays, except for DuP697, whose potency was fur- ther increased by preincubation with the enzyme. The double PGHS-1 His 513 3 Arg,Ile during the mutagenesis procedure. The cDNA of interest was subcloned into the pcDNA3 vector (Invitrogen) for transient expression in COS-7 (ATCC) cells. Expression of PGHS-1 Recombinant Proteins— PGHS-1 cDNA constructs were transfected into COS-7 cells using a calcium phosphate transfection kit (Life Technologies, Inc.). The cells were washed with fresh medium 24 h after transfection and harvested 72 h post-transfec- tion. The harvested cells were washed with phosphate-buffered saline (Life Technologies, Inc.) and resuspended in 100 m M Tris-HCl, pH 7.4, 10 m M EDTA, 2 (cid:109) g/ml aprotinin, 2 (cid:109) g/ml leupeptin, 2 (cid:109) g/ml soybean trypsin inhibitor, 0.5 m M phenylmethylsulfonyl fluoride. The cells were disrupted with a Kontes Micro Ultrasonic Cell Disruptor for 3 (cid:51) 10 s. The crude homogenate was centrifuged at 1000 (cid:51) g for 10 min at 4 °C. The supernatant fraction was then subjected to a 100,000 (cid:51) g spin for 1 h at 4 °C. The resulting microsomal pellet fraction for each mutant and the wild type PGHS-1 was resuspended in 100 m M Tris-HCl, pH 7.4, 10 m M EDTA. The protein concentration was quantitated by using the Pierce Coomassie protein assay as described by the manufacturer.

Prostaglandins are derived from arachidonic acid and act as mediators of pain, fever, and other inflammatory responses (1).
Prostaglandin G/H synthase (PGHS) 1 converts arachidonic acid into prostaglandin G 2 by the addition of molecular oxygen (a cyclooxygenase step) and then catalyzes the conversion of prostaglandin G 2 to prostaglandin H 2 by a peroxidase reaction (2,3). Prostaglandin H 2 is the precursor to the formation of all prostaglandins, thromboxane, and prostacyclin. Nonsteroidal anti-inflammatory drugs (NSAIDs), such as aspirin and indomethacin, abrogate prostaglandin synthesis through inhibition of the cyclooxygenase reaction of PGHS (4).
A second isoform of PGHS has been discovered (PGHS-2) that is induced in inflammatory situations in response to cytokines or growth factors (5)(6)(7)(8)(9)(10). This has lead to the development of selective PGHS-2 inhibitors, which have demonstrated that inhibition of PGHS-2 alone is sufficient to obtain an antiinflammatory effect while eliminating the gastric ulceration seen in animal models with NSAIDs, which inhibit both PGHS-1 and -2 without a large degree of selectivity (9,11,12). The observation with selective PGHS-2 inhibitors and the difference in the regulation of the expression of the two isoforms has led to the suggestion that PGHS-1 is responsible for normal physiological PG synthesis, whereas PGHS-2 is the main isoform responsible for elevated PG production during inflammatory responses (9). Recent examples of selective PGHS-2 inhibitors are DuP697 (13), NS-398 (14), SC-58125 (9), and DFU (5,5-dimethyl-3-(3-fluorophenyl)-4-(4-methylsulfonyl)phenyl-2(5H)-furanone). 2 These compounds have similar mechanisms of action, and they inhibit PGHS-2 by a time-dependent mechanism, whereas inhibition of PGHS-1 is through a time-independent and competitive mechanism (13,16). The molecular basis for this time-dependent inhibition has not been elucidated.
The cDNA and corresponding amino acid sequences of both human PGHS isoforms have been published, and the two enzyme forms have 63% sequence identity (17). The x-ray crystal structure of sheep seminal vesicle PGHS-1 has demonstrated that the cyclooxygenase active site is comprised of a long hydrophobic channel that is also the site for binding of NSAIDs (18). Several distinctive features of the active site are: 1) an Arg 120 residue at the mouth of the channel that has been demonstrated to be important for binding of arachidonic acid and NSAIDs containing a carboxylic acid residue (19,20); 2) a Tyr 385 residue at the upper portion of the active site that is involved in the formation of a radical at the C-13 position of arachidonic acid (21); and 3) a Ser 530 residue (just below Tyr 385 ) is the residue that is the site of acetylation by aspirin (22). The aspirin acetylated PGHS-2 results in an enzyme form that oxidizes arachidonic acid to  with altered sensitivity to inhibition by certain NSAIDs (24). Molecular modeling of the active site of PGHS-2 using the coordinates of PGHS-1 and site-directed mutagenesis have identified Val 509 of PGHS-2 as essential for inhibition by PGHS-2 selective inhibitors (25,26). We have utilized molecular modeling of the active site of PGHS-2 to determine residues of PGHS-1 that could be mutated to recover inhibition of PGHS-2 selective inhibitors on a modified form of PGHS-1.

EXPERIMENTAL PROCEDURES
Materials-Diclofenac was obtained from Sigma and indomethacin and arachidonic acid were purchased from Cayman Chemical Co. Sulindac sulfide, DFU, DuP697, NS-398, SC-58125, and L-745,296 (compound 23) (27) were synthesized by the Medicinal Chemistry Department of the Merck Frosst Center for Therapeutic Research. Hematin and glutathione were purchased from Sigma; phenol was obtained from Life Technologies, Inc.
Molecular Modeling-Human PGHS-2 was homology built based The labeled tyrosine residue in both isoforms is located below the heme group of the enzyme, and the arginine residue is present on the putative mouth to the channel into which arachidonic acid enters (note: several regions of the enzyme have been removed for clarification). The blue dots represent regions of solvent accessible surface within the two active sites. Ile 523 of PGHS-1 acts as a key residue to constrict one area of the active site of PGHS-1, whereas the Val 509 and Arg 499 residues of PGHS-2 results in a larger solvent accessible surface, suggesting a larger active site and inhibitor binding site.
upon an amino acid sequence alignment with residues 33-586 of sheep PGHS-1 using GCG and the published crystal structure of sheep PGHS-1. For amino acid changes between the two isoforms, the sheep PGHS-1 side chains were replaced with those corresponding to human PGHS-2. Following manual correction of the worst steric clashes of the introduced side chains, an all atom model was built and progressively minimized using AMBER 4.1 (28,29). Restraints to the sheep PGHS-1 peptide backbone were maintained at all times. The protein was heated to 300 K over 10 ps and equilibrated over 45 ps in vacuo using a distance-dependent dielectric constant. Snapshots of the trajectory at 0, 15, 30, and 45 ps were cooled and minimized; comparison of the resulting structures showed little structural variation. In the active site where flurbiprofen was shown to bind in sheep PGHS-1, there was almost no structural variation between the human PGHS-2 minimized structures. A solvent accessible surface area was generated (water probe radius, 1.4 Å) around the area corresponding to the flurbiprofen binding site in sheep PGHS-1.
Mutagenesis of PGHS-1-The three mutants of PGHS-1 (His 513 3 Arg; Ile 523 3 Val; and His 513 3 Arg,Ile 523 3 Val) were constructed by using the Sculptor in vitro mutagenesis system (Amersham Corp.). The coding region of PGHS-1 was subcloned into the multiple cloning site of Bluescript SK (Stratagene), and single stranded DNA was obtained using M13 K07 helper phage (Life Technologies, Inc.) as described by the manufacturer's instructions. Oligonucleotides were obtained from Research Genetics (Huntsville, AL) with a one-or two-nucleotide mismatch. The oligonucleotides were hybridized to the single stranded DNA, and mutants were generated according to the manufacturer's instructions. The mutants were confirmed by sequencing using a Prism DyeDeoxy Terminator Kit (Applied Biosystems) and an ABI 373 DNA sequencer. The correct mutants were completely sequenced on both cDNA strands to confirm no misincorporation of nucleotides during the mutagenesis procedure. The cDNA of interest was subcloned into the pcDNA3 vector (Invitrogen) for transient expression in COS-7 (ATCC) cells.
Expression of PGHS-1 Recombinant Proteins-PGHS-1 cDNA constructs were transfected into COS-7 cells using a calcium phosphate transfection kit (Life Technologies, Inc.). The cells were washed with fresh medium 24 h after transfection and harvested 72 h post-transfection. The harvested cells were washed with phosphate-buffered saline (Life Technologies, Inc.) and resuspended in 100 mM Tris-HCl, pH 7.4, 10 mM EDTA, 2 g/ml aprotinin, 2 g/ml leupeptin, 2 g/ml soybean trypsin inhibitor, 0.5 mM phenylmethylsulfonyl fluoride. The cells were disrupted with a Kontes Micro Ultrasonic Cell Disruptor for 3 ϫ 10 s. The crude homogenate was centrifuged at 1000 ϫ g for 10 min at 4°C. The supernatant fraction was then subjected to a 100,000 ϫ g spin for 1 h at 4°C. The resulting microsomal pellet fraction for each mutant and the wild type PGHS-1 was resuspended in 100 mM Tris-HCl, pH 7.4, 10 mM EDTA. The protein concentration was quantitated by using the Pierce Coomassie protein assay as described by the manufacturer. Protein expression was quantitated by immunoblot analysis. Protein samples were subjected to electrophoresis on 10% SDS-polyacrylamide gels and transferred electrophoretically to nitrocellulose membranes. The PGHS-1 protein was detected with a PGHS-1 polyclonal antiserum (30), and a secondary antibody developed with enhanced chemiluminesce (DuPont NEN). The immunoblot was exposed to Kodak Biomax MR film. The developed film was analyzed for expression levels by laser densitometric scanning using a Molecular Dynamics Laser Densitometer.
Enzyme Assays-PGHS activity was determined based on the conversion of arachidonic acid to PGE 2 by radioimmunoassay. Microsomal membrane protein preparations (final concentration, 15 g/ml) of PGHS-1 were incubated in the absence or the presence of inhibitor for various times in 100 mM Tris-HCl, pH 7.4, 10 mM EDTA, 1 mM glutathione, 0.5 mM phenol, 100 M hematin. The reaction was initiated with either 2 or 10 M arachidonic acid and was terminated at 1 or 3 min by acidification with 0.1 N HCl (final concentration). This mixture was neutralized with an equivalent amount of NaOH and analyzed for PGE 2 formation using the methyl oximated PGE 2 RIA kit (Amersham Corp.).
A time course of PGE 2 formation for all of the microsomal preparations was obtained and found to result in a rapid product formation that was linear for the first minute and reached a maximum after 3-5 min. The K m was determined using a reaction time of 1 min over a broad range of arachidonic acid concentrations (0.1-10 M) and was calculated using a hyperbolic regression program created by Dr. J. S. Easterby (University of Liverpool, UK). The IC 50 calculations of the inhibitors tested were determined Ϯ standard error from at least four values. Microsomal PGHS-2 was prepared from COS-7 cells infected with a vaccinia virus PGHS-2 construct as described previously (12).

Molecular Modeling of the Active Site of PGHS-2-
The development of selective PGHS-2 inhibitors has led to attempts to identify residues of PGHS-2 involved in the binding of these inhibitors. Molecular modeling of the three-dimensional structures of the human isoforms was performed utilizing the published crystal structure of sheep PGHS-1 (18) to identify residues of PGHS-1 that can be converted to PGHS-2 to mimic the inhibitor binding site of PGHS-2. The predicted structure of PGHS-2 demonstrates that the majority of the changes between the two isoforms occur at the N-terminal region forming the amphipathic helices and at the C-terminal tail. The cyclooxygenase active site of PGHS-1 and 2 are very similar. The important features of the active site are depicted in Fig. 1 with Tyr 385 (PGHS-1 residues, Fig. 1A) representing the upper part of the active site located just below the heme moiety and presumably responsible for the abstraction of the hydrogen atom at the C-13 position of arachidonic acid. The Ser 530 residue is the site of aspirin acetylation and is equivalent to Ser 516 are two residues that impinge on the active site and that are not conserved between the two isoforms. The solvent accessible space in the active site of these isoforms (depicted as blue dots in Fig. 1) shows that the His 513 and Ile 523 of PGHS-1 allow for  only a small available space in the vicinity of these residues for inhibitor interaction. Specifically, Ile 523 appears to severely restrict access to this space. In contrast, the Val 509 and Arg 499 of PGHS-2 result in the creation of a larger active site of PGHS-2 accessible to substrates and inhibitors. We have replaced the residues His 513 and Ile 523 of PGHS-1 with the corresponding residues of PGHS-2 to mimic the inhibitor binding site of PGHS-2.
Expression and Activity of Mutants of PGHS-1-The double mutant of PGHS-1 (PGHS-1 His 513 3 Arg,Ile 523 3 Val (HR-IV)) was constructed to completely mimic the lower end of the active site of PGHS-2 and the single mutants (PGHS-1 His 513 3 Arg (HR) and PGHS-1 Ile 523 3 Val (IV)) were constructed to establish the contribution of the single residue on the inhibitor selectivity. Levels of PGHS protein expression of the various mutants in microsomal membranes from transfected COS-7 cells were quantitated by immunoblot analysis using a specific anti-PGHS-1 antiserum (Fig. 2). Laser densitometric scanning indicates that the level of expression of each of the various proteins was very similar with relative levels of 0.9-, 0.6-, and 0.9-fold that of PGHS-1 for HR, IV, and HR-IV mutants, respectively. The PGHS activity of the various mutants, as assessed by conversion of arachidonic acid to PGE 2 ranged from 0.6 to 2-fold that of the wild type PGHS-1 (Table I). Membranes from mock transfected COS-7 cells contained less than 3% of the PGHS activity measured for any of the recombinant PGHS proteins. The time course of PGE 2 production was similar for all recombinant PGHS preparations with a rapid substrate conversion over the first minute of the reaction and a plateau of product formation after a 3-5-min reaction as previously reported (19). The apparent K m values for arachidonic acid were determined and found to be very similar for PGHS-2, PGHS-1, and the mutants of PGHS-1 (0.5-0.9 M) ( Table I).
Altered Sensitivity of Mutated Enzyme Forms to PGHS-1 and PGHS-2 Selective Inhibitors-The effect of the various mutations introduced into PGHS-1 on inhibitor sensitivity was evaluated using selective PGHS-2 inhibitors (DuP697, NS-398, SC-58125, and DFU), nonselective NSAIDs (sulindac sulfide, indomethacin, and diclofenac), and a selective PGHS-1 inhibitor (L-745,296). The results are summarized in Table II, and titration curves for the selective PGHS-2 inhibitors are shown in Fig. 3. The HR mutant resulted in an enzyme form that was inhibited by NSAIDs with similar potencies as observed for wild type PGHS-1, and there was no significant change in sensitivity to PGHS-2 selective inhibitors. The most significant change for the HR mutant was a 5-fold increase in the IC 50 for the selective PGHS-1 inhibitor, L-745,296. In contrast to the modest effect of the HR mutation on inhibitor sensitivity, the IV mutant was found to have an increased sensitivity to all of the PGHS-2 inhibitors. For example, DuP697 was about 300fold and SC-58125 at least 15-fold more potent at inhibiting the IV mutant than the wild type PGHS-1. The IV mutant was also more sensitive to inhibition by NS-398 or DFU, and the IC 50 for these selective inhibitors was at least 4-fold lower than that observed for PGHS-1. The double mutant HR-IV was more sensitive to inhibition by the selective PGHS-2 compounds, NS-398, SC-58125, and DFU, than either of the single point mutants. DFU and SC-58125, which both have IC 50 s of Ͼ50 M for the inhibition of PGHS-1, inhibited HR-IV with IC 50 values of 1.6 and 1.8 M, respectively. The HR-IV mutant remained sensitive to inhibition by the PGHS-1 selective inhibitor L-745,296, although the potency of this inhibitor was decreased by more than 10-fold as compared with native PGHS-1.
Mechanism of Inhibition of Mutant Forms of PGHS-1-The inhibition of PGHS-2 by selective compounds such as DuP697 is time-dependent for PGHS-2 and rapidly reversible for PGHS-1 (13,16). This difference is readily observed with microsomal preparations where only the inhibition of PGHS-2 was found to depend on the time of preincubation with enzyme as shown in Fig. 4 (A and B). We examined this time dependence for inhibition of the IV mutant, and as shown in Fig. 4C, inhibition of this mutant by DuP697 follows a similar pattern as that seen for inhibition of PGHS-2. DuP697 was also found to be a time-dependent inhibitor of HR-IV (data not shown). SC-58125 was a time-independent inhibitor of both the IV and HR-IV mutant (data not shown).
Both instantaneous inhibition assays in which the inhibitor was added at the time of initiation of the reaction and time-dependent inhibition have been used to evaluate the effect of NSAIDs and PGHS-2 inhibitors on enzyme activity (31). IC 50 values were determined for both the IV and HR-IV mutants using assays with either no preincubation or a 15-min preincubation of the inhibitor prior to initiation of the enzyme reaction. These results are summarized in Table III for DuP697, DFU, and NS-398. The single IV mutation caused only a slight increase in inhibitor potency as determined for instantaneous inhibition, which was not further modified by increasing the preincubation time in the case of DFU and NS-398. In contrast, DuP697, for which an IC 50 value of 0.1 M for PGHS-1 is observed independently of the time of preincubation, was a more potent inhibitor of the IV mutant after a 15-min preincubation (IC 50 ϭ 0.3 nM), consistent with the time-dependent inhibition observed in Fig. 4C. The introduction of the second HR mutation did not further change the parameters for the inhibition by DuP697. For DFU and NS-398, the second HR mutation caused a further decrease in IC 50 values for instantaneous inhibition to 5.2 and 2.6 M, respectively. In addition, the IC 50 values for these two inhibitors were about 3-fold lower using the 15-min preincubation assay. All of these selective PGHS-2 inhibitors are essentially inactive when tested for instantaneous inhibition versus PGHS-2 (IC 50 Ͼ 50 M). Altogether, the results of Table III and Fig. 4 demonstrate that the IV mutant is inhibitable in a time-dependent fashion by DuP697, and the HR-IV mutant is inhibitable in a time-dependent fashion by DuP697, NS-398, and DFU.

DISCUSSION
The discovery of a second isoform of PGHS several years ago has led to the identification of several selective inhibitors of PGHS-2 that have anti-inflammatory properties and a significantly decreased capacity for gastric ulceration. The mechanism of action of these selective PGHS-2 inhibitors is through a two-step mechanism. The first step involves the formation of an enzyme inhibitor complex (EI) followed by the slower formation of a tightly bound EI* complex as shown in the equation below.
The method of inhibition for PGHS-1 by PGHS-2 selective inhibitors is via a competitive rapidly reversible mechanism (13,16). Molecular modeling of the active site of PGHS-2 based on the known structure of PGHS-1 was performed to predict the major differences in the active sites of the two enzymes. The model suggests that in PGHS-2, the solvent accessible surface area is increased by the residues Val 509 and Arg 499 of PGHS-2 ( Fig. 1), resulting in the ability of PGHS-2 to utilize a variety of lipid substrates (16) and a larger active site for inhibitor binding. Because the only primary structure of the differences in the active site were localized to these latter residues and the equivalent PGHS-1 residues are His 513 and Ile 523 , we constructed mutants of PGHS-1 in attempts to convert the cyclooxygenase active site of PGHS-1 into that of PGHS-2. Because the apparent K m values of the mutants are within 1.4-fold and the specific activity is within 2-fold of that of PGHS-1 and the mutants retained sensitivity to inhibition by all NSAIDs tested, the mutations have not significantly altered the cyclooxygenase active site of prostaglandin synthase.
The single IV mutation in PGHS-1 results in an enzyme form that is more sensitive to inhibition by all four of the PGHS-2 selective inhibitors, although NS-398 and DFU are much weaker inhibitors of this mutant compared with either the double HR-IV mutant or PGHS-2. The single IV mutant was also more sensitive to inhibition by nonselective NSAIDs and the selective PGHS-1 inhibitor L-745,296, although the effect was most pronounced with the PGHS-2 selective inhibitors. The increase in potency for the IV mutant as compared with the wild type PGHS-1 is observable at instantaneous inhibition for DFU, SC-58125, and NS-398 with a further increase in the potency of DuP697 by preincubation of the inhibitor with enzyme before the measurement of activity. These results suggest that the IV mutation has increased the affinity of PGHS-1 for all selective PGHS-2 inhibitors but without a change in the mechanism of inhibition (rapidly reversible) for DFU, SC-58125, and NS-398. However, this single mutation was sufficient to change the rapidly reversible mechanism of inhibition of PGHS-1 by DuP697 to the time-dependent mechanism characteristic of PGHS-2 inhibition. Reconstitution of time-dependent inhibition would be consistent with a two-step mechanism as shown above for PGHS-2, although alternate mechanisms cannot be ruled out. The introduction of the second HR mutation to generate the HR-IV mutant resulted in an enzyme form very sensitive to inhibition by all four of the PGHS-2 selective inhibitors with DFU and NS-398 now showing timedependent inhibition in addition to DuP697. It is of interest to note that the changes of the HR mutation on inhibitor potency were observed on the IV mutant of PGHS-1 rather than on PGHS-1, demonstrating the advantage of combining mutations to show the importance of the various residues. Another difference in the active site of PGHS-2 as compared with PGHS-1 is the increased capacity of PGHS-2 to synthesize 15-R-HETE from arachidonic acid when acetylated by aspirin (23,24). Both the single IV and the double mutant HR-IV did not synthesize any appreciable 15-R-HETE as compared with PGHS-2 when treated with 100 M aspirin (data not shown). This suggests that the major determinants for the binding of selective PGHS-2 inhibitors by the PGHS-1 mutants are not sufficient to restore all of the properties of the active site of PGHS-2.
Two recent reports have shown that mutation of Val 509 of PGHS-2 to the corresponding Ile 523 in PGHS-1 results in the loss of sensitivity to PGHS-2 inhibition (25,26). Our results demonstrate that the converse in PGHS-1 is sufficient to gain partial sensitivity to PGHS-2 inhibitors, but the double HR-IV mutation results in an enzyme form that is more sensitive to a greater number of PGHS-2 selective inhibitors from three diverse structural classes, with most having a similar mechanism of inhibition. These data suggest that the molecular modeling of PGHS-2 has accurately predicted residues that are important for PGHS-2 inhibitor selectivity. During preparation of this manuscript, the crystal structure of PGHS-2 was published and confirms the major points depicted by the molecular modeling (15). The crystal structure data demonstrate the importance of Val 509 as the residue that provides access to a larger active site for PGHS-2 as compared with PGHS-1.
In conclusion, we have demonstrated that the combination of molecular modeling, mutagenesis, and inhibitor characterization have delineated both residues Arg 499 and Val 509 to be essential determinants in the differentiation between interaction of nonselective NSAIDs and selective PGHS-2 inhibitors and their mechanism of inhibition.