Identification of thromboxane A2 synthase active site residues by molecular modeling-guided site-directed mutagenesis.

Human thromboxane A2 synthase (TXAS) exhibits spectral characteristics of cytochrome P450 but lacks monooxygenase activity. Its distinctive amino acid sequence makes TXAS the sole member of family 5 in the P450 superfamily. To better understand the structure-function relationship of this unusual P450, we have recently constructed a three-dimensional model for TXAS using P450BM-3 as the template (Ruan, K.-H., Milfeld, K., Kulmacz, R. J., and Wu, K. K. (1994) Protein Eng. 7, 1345-1551) and have identified a potential active site region. The catalytic roles of several putative active site residues were evaluated using selectively mutated recombinant TXAS expressed in COS-1 cells. Mutation of Ala-408 to Glu or Arg-413 to Gly led to a complete loss of enzyme activity despite expression of mutant protein levels equivalent to that of the wild-type TXAS. Mutation of Ala-408 to Gly or Leu retained the enzyme activity at levels of 30 or 40%, respectively. This suggests that Ala-408 provides a hydrophobic environment for substrate binding. Mutation of Arg-413 to Lys or Gln completely abolished the enzyme activity, indicating that this residue is essential to catalytic activity and supports its identification as an active site residue. Mutation of Arg-410 to Gly or Glu-433 to Ala resulted in >50% reduction in the enzyme activity without appreciably altering mutant protein expression, consistent with a more subtle effect of these residues on TXAS catalytic efficiency. Mutation of residues predicted to be involved in binding the heme prosthetic group, including the heme thiolate ligand Cys-480, Arg-478, Phe-127, and Asn-110, each markedly reduced the expressed protein level and abolished enzyme activity. This suggests that proper heme binding is important to synthesis or stability of recombinant TXAS. Mutation of Ile-346, which corresponds to P450cam-Thr-252, an essential amino acid involved in dioxygen bond scission, to Thr increased the enzymatic activity by 40%, suggesting that oxygen bond cleavage is not a rate-limiting step in thromboxane A2 biosynthesis. The present results from site-directed mutagenesis support the overall structure of the TXAS active site predicted by homology modeling and have allowed refinement of the position of bound substrate.

Thromboxane A 2 (TXA 2 ) 1 is a potent mediator of platelet aggregation, vasoconstriction, and bronchoconstriction and plays an important role in major human diseases, including myocardial infarction, stroke, septic shock, and asthma (1,2). TXA 2 biosynthesis is catalyzed in succession by phospholipase A 2 , which liberates arachidonic acid from membrane phospholipids, prostaglandin H synthase, which converts arachidonate to prostaglandin H 2 (PGH 2 ), and TXA 2 synthase (TXAS, EC 5.3.99.5), which converts PGH 2 to TXA 2 as well as to 12Lhydroxy-5,8,10-heptadecatrienoic acid (3). Because PGH 2 can also be converted to several other prostanoids, TXAS plays a pivotal role in determining the amount of TXA 2 synthesized. In mammalian tissues, TXAS is a membrane-bound hemoprotein exhibiting spectral characteristics of a cytochrome P450, but it lacks monooxygenase activity and does not require reductase to initiate the reaction (4). TXAS cDNA cloned from human lungs and platelets encodes a 534-amino acid protein with a molecular weight of 60,684 (5)(6)(7). Amino acid sequence comparison between TXAS and other cytochrome P450s revealed less than 40% identity with the most closely related P450s, those in family 3 (7), making TXAS the sole member of family 5 in P450s (8). TXAS appears to be bound to the endoplasmic reticulum membrane and its NH 2 -terminal topology in the membrane has recently been shown to fit one model proposed for mammalian cytochrome P450s (9 -11).
Little is known about the active site structure of TXAS. In an attempt to define the heme environment and the substrate binding pocket better, we have recently constructed a threedimensional model of TXAS using a homology-modeling approach based on the crystallographic structure of the hemoprotein domain of P450 BM-3 (12,13). P450 BM-3 is a bacterial fatty acid monooxygenase that shares a high degree of sequence similarity with TXAS. The P450 BM-3 substrates, including arachidonic acid, are structurally related to the TXAS substrate PGH 2 . The structural model of TXAS allowed identification of several residues in the vicinity of the expected positions of the heme prosthetic group and the substrate. To test the importance of several of these predicted active site residues to TXAS catalytic activity, we have altered these residues by site-directed mutagenesis and analyzed the effects on enzyme activity. Results from this study indicate that Ala-408, Arg-410, Glu-433, and Arg-413 influence TXAS activity, supporting their proposed interactions with substrate or heme. These residues are conserved in human, rat, and pig TXAS (6,7,14,15), and all but Arg-410 are conserved in mouse TXAS (16). Alter-ations of several amino acid residues expected to be in the heme environment, including the conserved thiolate ligand Cys-480, caused a marked reduction in the expressed protein levels, suggesting that proper heme binding is required for normal TXAS synthesis or stability.
Molecular Modeling of Three-dimensional Structure-A detailed working model of the TXAS three-dimensional structure has been constructed using a homology-modeling technique (12). Briefly, a sequence similarity alignment was made for TXAS and the hemoprotein domain of P450 BM-3 by incorporating secondary structural predictions and experimental information, including conserved residues and heme ligand identity. The final alignment shows an overall 26.4% residue identity and 48.4% residue similarity between these two proteins. The main chain conformation of TXAS was then built with a Quanta 3.3 proteinmodeling package by transferring the crystal coordinates of P450 BM-3 to the aligned components of TXAS. The main chain backbone sections were linked to one another using a fragment-searching approach (17)(18)(19). The PGH 2 structure was built by using Quanta 3.3 with an energy minimization and conformation search. PGH 2 was initially docked into the proposed TXAS substrate binding site, which corresponds to the P450 BM-3 substrate binding site. Finally, an energy minimization routine with 500 steps of steepest descent was performed for the TXAS model containing heme and PGH 2 structures to remove steric clashes between atoms and to produce reasonable protein folding. The same approaches were used to modify the three-dimensional structure of the TXAS-PGH 2 complex based on the site-directed mutagenesis study in this article.
Site-directed Mutagenesis-A vector coding for TXAS was constructed by inserting the TXAS cDNA at the EcoRI site of the eukaryotic expression vector pSG5, which uses the SV40 early promoter to facilitate in vivo expression (7), to obtain pSG5-TXAS. The construct was validated to have a correct orientation with respect to the SV40 promoter in pSG5 by appropriate restriction enzyme digestion. Uracilcontaining single-stranded pSG5-TXAS was used to introduce specific base changes following Kunkel's method (20). Briefly, a phosphorylated oligonucleotide containing the base mismatch(es) was incubated with the single-stranded cDNA at 70°C, and then the mixture was allowed to cool to 30°C over 30 min. The oligonucleotide was extended and ligated by Sequenase and T4 DNA ligase. The resultant duplex DNA was used to transform E. coli XL-1 Blue cells. Plasmids with the desired cDNA mutations were confirmed by double-stranded sequencing using a Sequenase kit (U. S. Biochemical Corp.) or a double-stranded cycle sequencing kit (Life Technologies, Inc.). The oligonucleotides used for the mutations were (in 5Ј to 3Ј direction with mutated bases under- To generate mutant Ala-408 3 Gly, polymerase chain reaction-mediated site-directed mutagenesis was used. The primers used for mutation were (in 5Ј to 3Ј direction with mutated bases underlined): sense strand, ACCCGCCAGGTTTCAGATTC; and antisense strand, ATCT-GAAACCTGGCGGGTAC. Polymerase chain reaction was carried out to generate a 0.4-kb fragment by amplifying the wild-type TXAS cDNA with the sense primer and an antisense primer corresponding to nucleotides 1601-1621 downstream from the translational start codon. In a separate tube, a 0.15-kb fragment was generated by polymerase chain reaction using the antisense primer and a sense primer corresponding to nucleotides 1062-1081 downstream from the translation start codon. Both 0.4-and 0.15-kb fragments were combined, diluted 200-fold, and used as templates for polymerase chain reaction in a buffer containing 100 nM each of the primers 1601-1621 and 1062-1081 plus 1 nM each of the sense and antisense primers. The resultant 0.55-kb fragment was digested with HindIII and KpnI to obtain a 0.43-kb DNA. The 0.43-kb fragment was gel purified and subcloned into the unique HindIII-KpnI sites of pSG5-TXAS. Mutants Ala-408 3 Leu, Arg-413 3 Lys and Arg-413 3 Gln were generated identically, except that the primers used for mutations were: Ala-408 3 Leu, ACCCGCCACTTTTCAGATTC and ATCTGAAAAGTGGCGGGTAC; Arg-413 3 Lys, AGATTCACAAAG-GAGGCAGC and CTGCCTCCTTTGTGAATCTG; and Arg-413 3 Gln, AGATTCACACAGGAGGCAGC and CTGCCTCCTGTGTGAATCTG. The mutants were confirmed by DNA sequencing. The plasmids for transfection were prepared using Qiagen plamid purification kits (Qiagen, Inc., Studio City, CA).
Transfection of COS-1 Cells-COS-1 cells were grown to near confluency in 100-mm tissue culture dishes at 37°C in a humidified 5% CO 2 atmosphere with Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Transfections were performed as described by Lopata et al. (21). Briefly, the cells were incubated with 3 ml of DEAE-dextran (250 g/ml) containing 7.5 g of plasmid DNA. After a 1-h incubation at 37°C, 7 ml of complete medium containing chloroquine (52 g/ml) was added, and the cells were incubated for additional 5 h at 37°C in a 5% CO 2 atmosphere. After the incubation period, the medium was removed and replaced with 10 ml of complete growth medium, and the transfected cells were further incubated for 40 h until harvest.
Preparation of COS-1 Cell Microsomes-Cells were scraped from the plates, pelleted by centrifugation at 900 ϫ g for 10 min, washed with phosphate-buffered saline, and sonicated by a Sonifier Cell Disruption, model W185 (five 10-s pulses) in 5 ml of phosphate-buffered saline containing 1 g/ml aprotinin, 1 mM EDTA, 10 g/ml pepstatin, 10 g/ml leupeptin, 0.5 mM Pefabloc SC, and 0.1% Triton X-100. After removing cell debris by centrifugation at 10,000 ϫ g for 10 min, a microsomal fraction was isolated by centrifugation at 105,000 ϫ g for 1 h, and then resuspended in 400 l of phosphate-buffered saline. Protein concentrations were determined by bicinchoninic acid assay using bovine serum albumin as a standard (22).
Determination of TXAS Protein Level by Western Blot-Immunoblotting was performed according to the procedure previously described (9,23). In brief, microsomal samples were boiled in electrophoresis sample buffer (24) for 4 min before applying to a 10% polyacrylamide minigel for electrophoresis. The resolved proteins were transferred electrophoretically at 200 mA for 1 h to a nitrocellulose membrane. Subsequently, the membrane was blocked with 1% nonfat milk in phosphatebuffered saline and probed with a 1:500 dilution of rabbit antibody against TXAS conjugated to glutathione S-transferase (9) at 4°C overnight, followed by incubation with 10 g/ml goat anti-rabbit IgG conjugated to horseradish peroxidase, and finally visualized by incubation with the peroxidase substrate 4-chloro-1-naphthol (Sigma).
Assay of TXAS Activity-TXAS activity was assayed by incubating 100 g of microsomal protein prepared from transfected COS-1 cells with 5 M PGH 2 in 1 ml of 60 mM Tris-HCl, pH 7.4, at 37°C for 5 min. The reaction was terminated by acidification with 50 l of 2 N HCl. The mixture was subsequently neutralized by adding 100 l of 1.5 M Tris, pH 7.4. TXB 2 , the chemically stable hydration product of TXA 2 , was measured by radioimmunoassay as described previously (25). TXAS activity values are reported as the amount of TXB 2 produced per g of total protein per min.
To determine K m and V max values, microsomal proteins (50 g) from COS-1 cells transfected with wild-type or mutant TXAS cDNA were incubated with 0 -10 M PGH 2 for 2 min at 37°C in the absence or presence of 150 M U46619, a TXAS inhibitor. The reaction was stopped by acidification, and TXB 2 was measured by radioimmunoassay. Estimations of the kinetic parameters were obtained from nonlinear least squares fits to the Michaelis-Menten equation, using the Marquardt-Levenberg algorithm.

RESULTS AND DISCUSSION
Predicted Substrate Binding Region-Based on the sequence alignment and crystal structure of P450 cam , Gotoh and Fujii-Kuriyama (26) predicted six substrate binding domains in P450s. One of these domains is located between helix K and the conserved heme binding Cys region (Fig. 1). In this region, a ␤ sheet configuration was identified, and two pairs of antiparallel ␤ strands (Fig. 1, ␤5-␤8 and ␤6 -␤7) were conserved in the crystal structures of P450 cam , P450 BM-3 , and P450 terp . The ␤5-␤8 strand was considered part of the substrate binding sites for these P450s (13, 27, 28). Ravichandran et al. (13) predicted that, in this region of P450 BM-3 , three amino acid residues (Ala-328, Ala-330, and Met-354) were involved in the substrate binding. Our three-dimensional model of TXAS (12) also suggested that this ␤ sheet region located between Pro-407 and Val-436 was involved in substrate binding, with the corresponding amino acid residues being Ala-408, Arg-410 and Glu-433 ( Fig. 2A). The presence of the two charged residues in the TXAS pocket (Arg-410 and Glu-433) was suspected to reflect the difference in substrate specificity between TXAS and P450 BM-3 and implicated these residues in TXAS substrate binding and catalytic activity. To test this hypothesis, the cDNA sequences for these two residues, as well as a conserved residue, Ala-408, and an additional nearby charged residue, Arg-413, were altered to code for recombinant TXAS with Ala-408 3 Glu, Arg-410 3 Gly, Arg-413 3 Gly, or Glu-433 3 Ala mutations. Transient expression of each of the four mutants in COS-1 cells produced a full-length recombinant TXAS protein of M r 60,000 at levels equivalent to that observed for wild-type TXAS, as determined by Western blot analysis (Fig. 3A). In the platelet microsomes, a second band at M r 28,000 corresponding to endogenous glutathione S-transferase was also detected by the antibodies, which were raised against TXAS conjugated to glutathione S-transferase (9). The enzymatic activity was significantly diminished by each of the above mutations, with the Ala-408 3 Glu and Arg-413 3 Gly mutants being indistinguishable from the mock-transfected control (Table I). In contrast, the Arg-410 3 Gly and Glu-433 3 Ala mutants retained a significant fraction of the wild-type enzymatic activity (49 and 36%, respectively). The low amounts of TXAS synthesized in the COS-1 system preclude purification of TXAS and quantification of the enzyme-bound heme to determine the heme/ TXAS ratios. However, the wild-type recombinant TXAS from the COS-1 microsome had a specific activity similar to native TXAS in platelet microsomes, as judged by Western blot analysis and activity assay (data not shown). This result indicated that the recombinant wild-type TXAS maintained its native conformation for heme binding and enzymatic activity, validat-ing the use of the COS-1 expression system.
Ala-328 of P450 BM-3 constitutes part of the hydrophobic pocket and is located in the vicinity of the heme moiety in the crystal structure (13). In the TXAS model (12), Ala-408 is superimposed on Ala-328 of P450 BM-3 . The loss of activity in the Ala-408 3 Glu mutant suggested that Ala-408 is important for TXAS enzymatic activity (Table I and Fig. 3A). In the TXAS structural model ( Fig. 2A), alteration of this alanine residue to glutamic acid can be envisioned as disrupting the substrate binding pocket required for interaction with hydrophobic region of PGH 2 . To test this hypothesis, we made additional, more conservative, mutations, changing the alanine residue to either glycine or leucine, which would be likely to maintain the hydrophobic environment of the substrate binding pocket. The Ala-408 3 Gly and Ala-408 3 Leu mutant constructs were expressed at levels similar to that of wild-type TXAS in the COS-1 cells (Fig. 3C). Additionally, the Ala-408 3 Gly and Ala-408 3 Leu mutants retained 30 and 44% of the wild-type enzymatic activities, respectively ( Table I). The retention of TXAS activity in these conservative Ala-408 mutants supports our hypothesis that the loss of activity in the Ala-408 3 Glu mutant was due to disruption of the hydrophobic environment rather than some general structural perturbation.
In the original structural model ( Fig. 2A), Arg-413 did not appear to have any direct interactions with either the heme or bound PGH 2 . The complete loss of activity in the Arg-413 3 Gly mutant (Table I) prompted us to test alternative binding orientations for PGH 2 . For this, the substrate was reoriented without changing the TXAS protein backbone structure, and a new round of energy minimization was performed. One resulting modified model, shown in Fig. 2B, suggests that Arg-413 could participate in substrate binding by forming a charge interaction with the carboxylate group of PGH 2 . This provides a plausible explanation for the loss of activity observed in the Arg-413 3 Gly mutant. To test this model further, we altered the positively charged arginine residue at this position to either glutamine or lysine. As expected, mutation of Arg-413 to a neutral glutamine completely abolished enzymatic activity even though the recombinant protein was expressed at comparable levels ( Fig. 3C and Table I). Interestingly, the positively charged Arg-413 3 Lys mutant also lacked enzymatic activity, suggesting that the precise positioning of the positive charge is important.
Surprisingly, alteration of Arg-410 or Glu-433 to hydrophobic residues, changes that were anticipated from the TXAS model to disrupt substrate binding, produced only partial reduction in enzyme activity (Table I). To characterize further the effects of these mutations on TXAS catalysis, kinetic analyses of these mutant proteins were carried out in the presence or absence of U46619, a nonmetabolized PGH 2 analog with moderate affinity for TXAS (29). The apparent K m and V max values of the wild-type and mutant proteins are summarized in Table II. The kinetic analysis indicated that the Arg-410 3 Gly mutation did not greatly affect the apparent K m . For wild-type TXAS, U46619 had no significant effect on apparent V max but increased the apparent K m value by 2-fold, confirming that U46619, as previously reported (29), is a competitive inhibitor. U46619 did not appreciably change the apparent K m value for the Arg-410 3 Gly mutant. Glu-433 3 Ala had an apparent K m that was approximately twice that of the wild type. The presence of U46619 decreased the apparent K m value for Glu-433 3 Ala. These results suggest that the Arg-410 3 Gly and Glu-433 3 Ala mutants have altered affinity for PGH 2 and/or its analog, consistent with the proximity of both residues to the TXAS active site in the structural model.
As a control, one amino acid residue that is predicted by the  (Fig. 2) to be located away from the substrate binding pocket, Thr-350, was also mutated. As expected, alteration of Thr-350 to Ala did not exert any influence on the TXAS protein level (Fig. 3A) or on the enzyme activity (Table I).
To address the possibility of altered product specificity, reversed-phase high performance liquid chromatographic analysis was carried out for products formed from [1-14 C]PGH 2 by recombinant TXAS and several of the mutants, particularly those mutants that failed to synthesize detectable TXB 2 . The wild-type recombinant enzyme produced both TXB 2 and 12Lhydroxy-5,8,10-heptadecatrienoic acid, as previously reported (23). No appreciable amounts of alternative products were found with the Ala-408 3 Glu and Arg-413 3 Gly mutants (data not shown), making it unlikely that the low apparent activity in these mutants was due to a shift in product profile.
Another important feature of the crystal structures of P450 BM-3 , P450 cam , and P450 terp is that helix I forms a hydrophobic backbone passing through the center of the molecule and creates a unusual groove adjacent to the heme region (13,28). The helix I residue corresponding to Thr-252 in P450 cam is conserved as either a Thr or Ser across all monooxygenase P450 sequences, and mutagenesis studies in P450 cam showed that the hydroxylation reaction required a Thr or Ser at position 252 (30,31). A mechanism was proposed in which the Ser-Thr hydroxy group is involved in dioxygen bond scission, one of the rate-limiting steps of the P450 cam monooxygenation reaction (32). Our TXAS model (12) predicts that Ile-346 corresponds to Thr-252 of P450 cam . The atypical Ile substitution in TXAS prompted us to examine the effect of "reverting" Ile-346 to a Thr. This Ile-346 3 Thr mutant was found to be produced in COS-1 cells at levels comparable with those of the wild type (Fig. 3A), and its catalytic activity was actually somewhat higher than the activity of the wild-type enzyme (Table I). It is noteworthy that TXAS, unlike other P450s, catalyzes scission and isomerization of an endoperoxide rather than using molecular oxygen for a monooxygenation reaction. Epidioxy cleavage requires less energy for a peroxide than for a dioxygen molecule (33) and is thus less likely to be the rate-limiting step in TXA 2 biosynthesis. This may explain why threonine or serine is not    Heme Environment-Two amino acid residues in the heme binding region are conserved among all P450s: a cysteine residue, which serves as the proximal axial ligand for the heme iron through a thiolate bond, and an arginine or a histidine residue, which forms a hydrogen bond with the D-ring propionate group of the heme moiety. In TXAS, these residues correspond to Cys-480 and Arg-478. Our molecular modeling, as shown in Fig. 2B, also predicted that Asn-110 of TXAS is linked to the heme propionate group by hydrogen bonds (12). From the molecular model, Phe-127 was also predicted to be in the vicinity of TXAS heme. Mutation of Cys-480 to Ser, Arg-478 to Ala, Phe-127 to Val, and Asn-110 to Ile by site-directed mutagenesis each resulted in a complete loss of the enzyme activity (Table I), attributable to a very low level of TXAS protein expression (Fig.  3, B and C). All four mutant proteins did have molecular weights of 60,000, indistinguishable from wild-type recombinant TXAS and human platelet microsomes. The low amounts of the Cys-480 3 Ser, Arg-478 3 Ala, Phe-127 3 Val, and Asn-110 3 Ile proteins expressed in COS-1 cells precluded determination of their heme content.
Recombinant TXAS with a mutation of Cys-480 has been found to be inactive, even when expressed at high levels in insect cells (34), supporting the assignment of Cys-480 as the heme thiolate ligand. The present observation of dramatically decreased expression of the Cys-480 3 Ser mutant in COS-1 cells (Fig. 2B) suggests that the loss of the proximal heme ligand somehow affects the stability of recombinant TXAS in the mammalian cells. It is interesting to note that greatly decreased expression was also found for three other mutations of residues predicted to participate in binding the TXAS heme: Asn-110 3 Ile, Phe-127 3 Val, and Arg-478 3 Ala (Fig. 3B). Misfolded proteins are known to be susceptible to accelerated degradation in the endoplasmic reticulum compartment of mammalian cells (35). It may be that any disruption of proper heme binding in TXAS alters the overall protein folding and targets the misfolded protein for degradation by endogenous proteases.
Conclusion-Our revised three-dimensional structure of the TXAS model (Fig. 2B) predicts that Arg-413 can act as a bridge, interacting with Glu-433 and the carboxylate group of PGH 2 . The revised orientation of PGH 2 also puts Arg-410 closer to the center of the PGH 2 molecule, where this arginine residue might participate in the catalytic reaction of TXA 2 formation. Ala-408 is also located near PGH 2 in the new orientation and could plausibly contribute to the hydrophobic environment to stabilize substrate binding. As in the original model, Cys-480, Arg-478, Phe-127, and Asn-110 are in the immediate vicinity of the heme moiety. Cys-480, the thiolate ligand of the TXAS heme iron, is located on the heme face away from the substrate binding pocket. Arg-478 is predicted to play an important role in maintaining the heme in its proper environment through a charge interaction with the propionate. In the revised model, Phe-127 is not only adjacent to the heme propionate ring but also close to the bound substrate. Furthermore, the C-9 endoperoxide oxygen atom of PGH 2 is oriented toward the heme iron, in accordance with the spectroscopic data indicating an interaction between this oxygen and the heme iron (36).
The results of the present mutagenesis study support the general features of the active site in the TXAS structural model and have led to refinement of the substrate binding site. The revised TXAS model should be useful for further characterization of the active site structure and function of this unusual cytochrome P450.