Prostacyclin synthase active sites. Identification by molecular modeling-guided site-directed mutagenesis.

Prostacyclin synthase (PGIS), a cytochrome P450 enzyme, catalyzes the biosynthesis of a physiologically important molecule, prostacyclin. In this study we have used a molecular modeling-guided site-directed mutagenesis to predict the active sites in substrate binding pocket and heme environment of PGIS. A three-dimensional model of PGIS was constructed using P450BM-3 crystal structure as the template. Our results indicate that residues Ile67, Val76, Leu384, Pro355, Glu360, and Asp364, which were suggested to be located at one side of lining of the substrate binding pocket, are essential for catalytic activity. This region containing β1-1, β1-2, β1-3, and β1-4 strands is predicted well by the model. At the heme region, Cys441 was confirmed to be the proximal axial ligand of heme iron. The conserved Phe and Arg in P450BM-3 were substituted by Leu112 and Asp439, respectively in PGIS. Alteration of Leu112 to Phe retained the activity, indicating that Leu112 is a functional substitution for Phe. In contrast, mutant Asp439 → Ala exhibited a slight increase in activity. This result implies a difference in the heme region between P450BM-3 and PGIS. Our results also indicate that stability of PGIS expression is not affected by heme site or active site mutations.

Prostacyclin (PGI 2 ) 1 is a potent inhibitor of platelet activation, vasoconstriction, and leukocyte interaction with endothelium (1,2). It is considered to be an important vasoprotective molecule. Its biosynthesis in vascular endothelial cells is catalyzed by a series of enzymes of which PGI 2 synthase (PGIS) catalyzes the final conversion of prostaglandin H 2 (PGH 2 ) to PGI 2 . PGIS was purified from bovine aortic tissues to homogeneity and the purified enzyme was characterized as a cytochrome P450 (3)(4)(5). Bovine and human PGIS cDNA have been isolated recently (6 -8). The bovine vascular PGIS cDNA is highly homologous to the human vascular enzyme. The cDNA from both species contains an open reading frame of 1500 nucleotides coding for a 500-amino acid peptide with a calculated molecular mass of about 57 kDa. Sequence comparison with known P450s shows less than 30% sequence identity between PGIS and other cytochrome P450s. Interestingly, de-spite similarities in enzymatic reactions, PGIS has only about a 16% sequence identity to thromboxane A 2 synthase (TXAS). Thus, PGIS has been classified as a member of a new family of P450, CYP8. The active site structure of PGIS has not been delineated. On sequence comparison with P450s, we observed the conservation of several P450 structural elements in PGIS, including a putative membrane-anchoring segment, a helix I which forms an ␣-helix backbone through the center of the enzyme, and a heme binding pocket. This raised the possibility of mapping the active site amino acid residues by molecular modeling-guided site-directed mutagenesis. Cytochrome P450 BM-3 from Bacillus megaterium catalyzes the monooxygenation of fatty acids, including arachidonic acid (AA). Its catalytic function and primary structure resemble mammalian microsomal cytochrome P450s. The heme domain of P450  shows about 25-30% sequence identity to several microsomal P450 enzymes (9). Since AA is a functional substrate of P450 BM-3 , and the substrate is closely related to the substrate of TXAS and PGIS, we reasoned that the crystal structure of the hemoprotein domain could serve as a useful template for constructing a three-dimensional model of these two enzymes.
The three-dimensional model of TXAS has been reported by our group (10) and proves to be useful in identifying TXAS active site residues (11). In the present study, we constructed a PGIS three-dimensional model based on the crystallographic structure of P450  . From this model, we identified several amino residues potentially involved in substrate access and binding and in heme binding. To test the functional role of these residues in catalysis, we have altered these residues by site-directed mutagenesis. Results from these experiments have allowed for identifying amino acid residues in the substrate channel that are important in substrate access and binding and in interaction with heme.

EXPERIMENTAL PROCEDURES
Materials-COS-1 cells (ATCC CRL-1650) were obtained from the American Type Culture Collection. pBluescript (SKϩ) plasmid DNA and eukaryotic expression vector pSG5 were from Stratagene. Human lung cDNA was from CloneTech. Cell culture media and antibiotics were from either Life Technologies, Inc. or HyClone Laboratories. DEAE-dextran and pGEX-2T plasmid DNA were from Pharmacia Biotech Inc. Isopropyl-␤-D-thiogalactopyranoside, dimethyl sulfoxide, and chloroquine were from Sigma. [ 14 C]Arachidonic acid ([ 14 C]AA) was from Amersham Corp. PGH 2 was from Cayman Chemical (Ann Arbor, MI). Oligonucleotides were synthesized by Genosys (The Woodlands, TX). Linear-K preadsorbent TLC plates with silica gel (60 Å) and 250-m thickness were from Whatman (LK6D).
Molecular Modeling-The strategy used for constructing the PGIS three-dimensional model followed that developed for TXAS (10) and other mammalian P450s (12). The strategy took into consideration six procedures: (i) sequence alignment, (ii) framework construction, (iii) loop structural determination, (iv) site-chain placement, (v) molecular docking, and (vi) energy minimization, which are briefly described below. A sequence similarity alignment was made for PGIS and the hemoprotein domain of the P450 BM-3 (Fig. 1), and the main-chain conformation of PGIS was built with the Quanta-Charmm protein model-ing package by transferring the crystal coordinates of P450 BM-3 to the aligned components of PGIS. Thus, the conserved helices and strand framework in P450 BM-3 provided the backbone section coordinates of PGIS. The backbone segments were linked to each other using a fragment searching approach (13)(14)(15) and a data base containing 58-protein three-dimensional structures, developed by Ruan et al. (10). Several three-dimensional structural candidates were obtained from the data base and one was chosen based on the best similarity of primary structure and the best fit of C␣ distance, with a cutoff value of 0.5 Å root mean square deviation (10,12,16). The three-dimensional structural coordinates of the heme in PGIS was adopted directly from the x-ray structure of P450 BM-3 and then fixed into the backbone structure of PGIS. The substrate, PGH 2 , structure was constructed, energy-minimized, and subjected to conformation search. One of the 200 conformations of the PGH 2 three-dimensional structure with the best score was docked into the proposed PGIS substrate binding pocket of the constructed three-dimensional model of PGIS, which corresponded to the P450 BM-3 substrate binding cavity. An energy minimization with 500 steps of steepest descent was performed for the PGIS three-dimensional structural model containing heme and PGH 2 structures. After the energy minimization, some steric clashes between atoms of the model were removed, and a reasonable protein folding and the substrate in coordinate position were obtained.
Expression Vector of PGIS cDNA-The full-length cDNA of PGIS was amplified by a two-step polymerase chain reaction (PCR) with two pairs of primers (I34 (CAGCCCCGCGATGGCTTG) and I37 (TGTGCACACA-GAAAGCTG), outer primer set; and I35 (CATGGATCCGCGATGGCTT GGGCC) and I36 (CGAGCACGTGGATCCATC), inner primer set). Outer primer set was used for first PCR in a 60-l reaction mixture under the cycle schedule of 95°C for 35 s, 53°C for 1 min, and 72°C for 1 min 10 s with 2-s extension for each additional cycle for a total of 30 cycles on a Perkin-Elmer DNA thermal cycler. 0.4 g of human lung cDNA, 2.5 units of Taq DNA polymerase (Promega or Perkin-Elmer), 0.1 unit of Pfu DNA polymerase (Stratagene), 100 nM of each primer, and buffer for Pfu DNA polymerase were used. 1 l of the first PCR product (using outer primers) was used as template for second PCR amplification using inner primers in a 60-l reaction mixture under the cycle schedule of 95°C for 35 s, 55°C for 1 min, and 72°C for 1 min 10 s with 2-s extension for each additional cycle for a total of 30 cycles. The final PCR products, which contained the full-length PGIS cDNA, were gel-purified and subcloned into the BamHI site of pBluescript. The subcloned PGIS cDNA was verified by DNA sequencing.
Site-directed Mutagenesis-The PGIS cDNA was subcloned into pSG5 plasmid at either BamHI or BamHI-BglII sites for transient expression in COS-1 cells. Orientations of insert were verified by both AccI and EcoRI restriction enzyme digestions. The final construct containing the full-length PGIS cDNA (pSG5-PGIS, wild-type) was verified by DNA sequencing. The following single mutants of PGIS, Ile 67 3 Lys, Arg 72 3 Glu, Arg 72 3 Gln, Val 76 3 Asp, Thr 206 3 Val, Leu 210 3 Asp, Leu 213 3 Glu, Pro 355 3 Val, Glu 360 3 Gly, Asp 364 3 Val, Leu 384 3 Asp, Phe 476 3 Val, Cys 231 3 Ser and Cys 441 3 Ser, Leu 112 3 Asp, Leu 112 3 Gly, Leu 112 3 Phe, Pro 113 3 Ala, and Asn 439 3 Ala, were produced according to the method described previously (17) with some modifications. In brief, a primer bearing the mutated sequence and its paired primer were used in PCR to amplify a 400 -500-base pair product using pSG5-PGIS DNA as the template. The PCR products were purified with Wizard PCR prep (Promega) to get rid of primers. Purified DNA of the first PCR product was used as a mega-primer for the second PCR with a primer on the other side of the mutation. The PCR product (about 5 ng), which contained the full-length cDNA and a portion of the vector, was used as the template for the second PCR. The final products were gel-purified, restriction enzyme-digested, and subcloned into pSG5-PGIS to replace the target segment. DNA sequencing was used to verify the sequence at mutation and its surrounding region. The plasmids for transfection were prepared using Qiagen plasmid purification kits (Qiagen).
Antibody Development-To generate the expression vector for glutathione S-transferase (GST) and PGIS (GST-PGIS) fusion protein, PGIS cDNA (wild-type) was subcloned into pGEX-2T vector (pGEX-PGIS). Transformed JM105 cells containing pGEX-PGIS DNA were cultured at 37°C in 2 ϫ YT medium to A 600 ϭ 0.8. Isopropyl-␤-D-thiogalactopyranoside (0.2 mM) was then added to induce protein expression, and cells were further incubated at 37°C for 3 h. Harvested cells were sonicated for 35 s and centrifuged at 1000 ϫ g for 10 min. The pellet which contained the fused GST-PGIS protein was washed with lysis buffer (100 mM NaCl, 50 mM Tris-Cl, pH 8.0, 0.5% Triton X-100, 10 mM EDTA) and centrifuged before being applied on SDS-polyacrylamide gel electrophoresis gels. The ϳ68-kDa GST-PGIS protein was dissected out from the SDS-polyacrylamide gel electrophoresis gel and electroeluted. The antibody against purified GST-PGIS was prepared in rabbits by H.T.I. Bio-products (Ramona, CA).
COS-1 Cell Transfection-COS-1 cells were grown to near-confluence in 100-mm tissue culture dishes in the presence of Dulbecco's modified Eagle's medium containing 2% fetal calf serum and 8% bovine calf serum at 37°C in a humidified 5% CO 2 atmosphere. Transfection procedure was performed as described by Wang et al. (11) with some modifications. In brief, cells were activated in fresh medium for 2 h before transfection. After washing with phosphate-buffered saline (PBS), COS-1 cells were incubated in 1 ml of PBS containing DEAEdextran (0.5 mg/ml) and DNA (5 g/ml) for 30 min in a lamina hood. 5 ml of Dulbecco's modified Eagle's medium containing 2% fetal calf serum and chloroquine (60 g/ml) was then added, and the cells were incubated for additional 3 h at 37°C in 5% CO 2 atmosphere. After incubation, the cells were shocked with dimethyl sulfoxide (10%) for 2 min, and the medium was replaced with 12 ml of complete growth medium. The transfected cells were then incubated for 42-60 h. To harvest cells, COS-1 cells were scraped from the plate, pelleted by centrifugation, washed with PBS, and resuspended in 350 l of PBS. The protein concentration of each sample was determined using BCA protein assay reagent kit (Pierce).
Determination of PGIS Protein Level by Western Blot-Immunoblotting was performed by a procedure described previously (18). 15-20 g of whole cells were boiled in an electrophoresis sample buffer (19) for 8 -10 min before being applied to a 10% polyacrylamide minigel for electrophoresis. The resolved proteins were electrotransferred at 300 -450 mA for 1 h to nitrocellulose membrane. Subsequently, the membrane was blocked with 3% non-fat milk in PBS and probed with a 1 to 1000 dilution of rabbit serum containing antibody against GST-PGIS fusion protein at 25°C for 1 h or 4°C overnight. A second antibody of goat anti-rabbit IgG conjugated with horseradish peroxidase was used as recommended. The protein bands were visualized by incubation with either the peroxidase substrate 4-chloro-1-naphthol (Bio-Rad) or Supersignal CL-HRP substrate system (Pierce). The levels of wild-type and mutant proteins on the Western blots were scanned by densitometer as described below. Organic products generated were extracted with 300 l of diethyl ether twice. The organic extract which contained 92% Ϯ 2% (n ϭ 10) of the radioactivity was concentrated under nitrogen to less than 60 l and applied to a TLC plate. The TLC plate was chilled on ice before placing in a developing tank (on ice) in the organic phase of ethyl acetate/2,2,4-trimethylpentane/acetic acid/water (110:50:20:100, v/v/ v/v; upper phase) (21). After development, the radioactive signals on the TLC plate were detected by autoradiography. PGI 2 was hydrolyzed to 6-keto-PGF 1␣ during extraction, therefore the amount of [ 14 C]6-keto-PGF 1␣ produced reflects the amount of PGI 2 converted from PGH 2 . The position of 6-keto-PGF 1␣ on TLC plate was determined by co-migration of enzymatic products of a wild-type sample and nonlabeled 6-keto-PGF 1␣ which was visualized by iodine vapor. The radioactive signal of [ 14 C]6-keto-PGF 1␣ on the autoradiography was detected by using a densitometer (UMAX Vista-T630 or Epson ES-800C scanners with the Adobe Photoshop program), and the intensity of each signal was analyzed on a Macintosh computer using the public domain NIH Image program.
Radioimmunoassay was also used to determine the catalytic activity of PGIS. Whole cell homogenate (60 g of protein) was incubated with PGH 2 (5 M) at room temperature for 3 min, the reaction was terminated, and the organic products were extracted by the procedure described above. The organic extract was dried under nitrogen and resuspened in methanol. The extract was applied to reverse-phase HPLC as described previously (22). The fractions containing 6-keto-PGF 1␣ were collected, dried, and resuspended in methanol. 6-Keto-PGF 1␣ contents in these extracts were measured by radioimmunoassay as described previously (23).

RESULTS
Substrate Binding Pocket-A three-dimensional model based on the crystal structure of the hemoprotein domain of P450 BM-3 was constructed. Details of the PGIS model will be published elsewhere. The backbone structure of the PGIS model matched well with that of P450 BM-3 , especially the helical regions (data not shown). The crystal structure of P450 BM-3 reveals a long binding pocket lined with mostly nonaromatic hydrophobic res-idues (9). We compared the substrate access channel of the PGIS model with that of P450 BM-3 by superimposing the substrate binding pocket structure of PGIS model on that of P450 BM-3 crystal structure. Residues which were predicted from the three-dimensional model to be important in substrate binding pocket and heme site are shown in Fig. 2. Along the substrate access channel of PGIS were hydrophobic residues, Ile 67 , Val 76 , Thr 206 , Leu 210 , Leu 213 , and Phe 476 , which correspond to Phe 42 , Tyr 51 , Leu 181 , Met 185 , Leu 188 , and Leu 437 of P450 BM-3 channel, respectively. To test the hypothesis that these hydrophobic residues were important in PGIS catalytic activity, the cDNA sequences at these six positions were altered individually to code for recombinant PGIS mutants: Ile 67 3 Lys (on ␤1-1), 2 Val 76 3 Asp (on ␤1-2), Thr 206 3 Val, Leu 210 3 Asp, Leu 213 3 Glu (on ␣F), and Phe 476 3 Val (between ␤4 -1 and ␤4 -2). Transient expression of each of these mutants in COS-1 cells generated a level of PGIS protein comparable with that of the wild-type (Fig. 3, A and B). 6-Keto-PGF 1␣ levels produced by recombinant mutants Ile 67 3 Lys, Val 76 3 Asp, and Leu 210 3 Asp were markedly reduced to the level of the mock-transfected control as analyzed by TLC (Fig. 4). The TLC results were confirmed by radioimmunoassay of 6-keto-PGF 1␣ fractions separated by reverse-phase HPLC (data not shown). The catalytic activities of PGIS mutants from multiple experiments expressed as percent of the wild-type PGIS activity are summarized in Table I. The activities of Ile 67 3 Lys, Val 76 3 Asp, and Leu 210 3 Asp were 5.1, 8.3, and 6.8% of the wild-type enzyme, respectively. By contrast, catalytic activity of mutants Thr 206 3 Val, Leu 213 3 Glu, or Phe 476 3 Val was not significantly different from that of the wild-type (100, 97, and 83%, respectively; Table I and Fig. 4). These results indicate that, as predicted by the three-dimensional model, Ile 67 , Val 76 , and Leu 210 of PGIS are important in substrate access to the active site, while contrary to prediction from the model, Thr 206 , Leu 213 , and Phe 476 are not critically involved in substrate access. 2 The designation for ␤ strands of PGIS was according to that for P450 BM-3 crystal structure (9) and is depicted in Fig. 1.

FIG. 2. PGIS residues predicted from the three-dimensional model to be important in substrate binding pocket and heme site.
Green denotes PGH 2 . Pink denotes the heme moiety. Blue denotes the charged residues. Red and yellow denote the hydrophobic residues.
Arg 47 , corresponding to Arg 72 in PGIS model, is considered to be a "gate-keeper" for P450 BM-3 enzyme. To determine whether this charged residue plays a similar role in PGIS, we altered it to Glu or Gln by site-directed mutagenesis. These two mutants expressed a similar quantity of proteins on Western blots as the wild-type (Fig. 3A) and had comparable catalytic activities as the wild-type enzyme (Table I and Fig. 4).
Ravichandran et al. (9) predicted from the crystal structure of P450 BM-3 that ␤1-4 strand was part of the substrate binding sites and residues Ala 328 , Ala 330 , and Met 354 (on ␤1-3) were involved in the substrate binding. In this region, according to sequence alignment and PGIS modeling, three corresponding amino acid residues, Pro 355 , Glu 360 , and Leu 384 , respectively, were suggested to be involved in substrate binding (Fig. 2). A nearby residue in TXAS, Arg 413 (on ␤1-4) corresponding to Asp 364 in PGIS, was suggested to have interaction with substrate (11). Each of these four amino acid residues in the full-length PGIS cDNA was individually altered to Pro 355 3 Val, Glu 360 3 Gly, Asp 364 3 Val, and Leu 384 3 Asp. The level of each mutant protein expressed in transient transfected COS-1 cells was comparable with that of the wild-type (Fig. 3,  A and B). PGIS activity was significantly diminished in Asp 364 3 Val and Leu 384 3 Asp mutants (6.8 and 6.9%, respectively; Table I and Fig. 4), whereas Pro 355 3 Val and Glu 360 3 Gly mutants retained a fraction of the wild-type enzymatic activity (34 and 44%, respectively). These results indicate that Pro 355 , Glu 360 , Leu 384 , and Asp 364 are important in substrate binding as predicted by molecular modeling.
Heme Environment-A cysteine residue which serves as the proximal axial ligand for the heme iron through a thiolate bond is conserved among all P450s. There are only two Cys residues in PGIS, Cys 231 (on ␣G) and Cys 441 , of which Cys 441 corresponds to the consensus P450 cysteine (Figs. 1 and 2). Mutation of Cys 441 to Ser by site-directed mutagenesis resulted in a diminished enzyme activity (13% ; Table I and Fig. 4), without alteration in the expressed protein level (Fig. 3B). Mutation of Cys 231 to Ser did not alter the catalytic activity (98%) ( Table I and Fig. 4).
In the heme binding region of cytochrome P450, an Arg or a His residue is conserved which forms a hydrogen bond with the D-ring propionate group of the heme moiety. Alignment of amino acid sequence of the three-dimensional model of PGIS with that of P450 BM-3 did not reveal a corresponding Arg or His. Instead, it suggested that Asn 439 of PGIS was the corresponding residue (Figs. 1 and 2). This raised the possibility that the heme binding environment of PGIS may differ from that of other P450s and Asn 439 may be functionally important in heme binding. To test this hypothesis, Asn 439 was altered to Ala. The protein level of mutant Asn 439 3 Ala in transient transfected COS-1 cells is similar to that of the wild-type. Contrary to the hypothesis, mutation of Asn 439 to Ala resulted in a slight increase in the PGIS activity.
Crystallographic structure of P450 BM-3 suggests that Phe 87 of P450 BM-3 is involved in heme interaction (9). However, neither sequence alignment nor molecular modeling disclosed a corresponding Phe residue around this region in PGIS. The corresponding residue derived from molecular modeling was Leu 112 (Fig. 2). To test if this Leu residue can provide hydrophobic interactions with heme and substrate as the Phe residue in other P450s does, Leu 112 was mutated to Asp, Phe, or Gly by site-directed mutagenesis. The catalytic activity of mutant Leu 112 3 Phe was only slightly lower (89%) than that of the wild-type, while the activity of mutants Leu 112 3 Asp and Leu 112 3 Gly was reduced to 11 and 15% of the wild-type enzyme, respectively (Table I).
Pro 113 was mutated as a control. Although situated next to Leu 112 , this residue was not located at the vicinity of heme or substrate pocket on the PGIS model. Pro 113 3 Ala mutant as predicted from the model retained most of the catalytic activity (Table I). DISCUSSION Molecular modeling coupled with site-directed mutagenesis is a powerful tool in studying the structure-activity relationship of cytochrome P450 enzymes. PGIS is a new family of P450 which shares several enzymatic and spectral characteristics with TXAS. Based on our previous observations that the crystallographic structure of P450 BM-3 serves as a suitable template for constructing the three-dimensional model of TXAS (10), we have taken a similar approach in constructing the three-dimensional model of PGIS, and the results indicate that the generated model is valuable for identifying the amino acid residues involved in substrate binding and heme environment.
Substrate Binding Pocket-The PGIS model suggests that the substrate access channel of PGIS is similar to that of P450 BM-3 in that it is very long and lined with hydrophobic residues. This result is not surprising since the substrate (PGH 2 ) for PGIS is a 20-carbon metabolite of arachidonate, a substrate for P450 BM-3 . Our site-directed mutagenesis results indicate that Ile 67 , Val 76 , and Leu 384 , which are situated along the lower portion of the channel according to the PGIS model (Fig. 2), are critically important in PGIS catalytic activity. Mutations of these hydrophobic residues to charged residues reduced the catalytic activity to the background value. Hence, this portion of the binding pocket in PGIS is comparable with that of P450 BM-3 . Two residues (Pro 355 and Glu 360 ) are located inside the lower portion of the binding pocket adjacent to heme (Fig. 2). Mutations of these two residues, Glu 360 3 Gly and Pro 355 3 Val retained 44 and 34% activity, respectively, suggesting that these two charged residues are involved in catalytic activity. We have shown previously that TXAS Arg 413 , which is not predicted to be located at the binding pocket based on P450 BM-3 structure, is important in TXAS catalytic activity (11). This residue corresponds to PGIS Asp 364 . Mutation of Asp 364 to Val reduced the catalytic activity to the background value, indicating that these two corresponding residues are similarly important in PGIS and TXAS activities. According to the PGIS model, this charged residue was located near the lining of the substrate binding pocket but not in the vicinity of PGH 2 or heme (Fig. 2). Taken together, these results indicate that, with minor exceptions, the lower part of the substrate binding pocket containing ␤1-1, ␤1-2, ␤1-3, and ␤1-4 strands in PGIS model (Fig. 2) is comparable with that of P450 BM-3 .
By contrast, the upper portion of the substrate binding pocket of PGIS was not as well predicted from the P450 BM-3 structure. Only one of the four predicted residues is important in substrate binding and catalytic activity: Leu 210 3 Asp lost the activity, whereas Thr 206 3 Val, Leu 213 3 Glu, and Phe 476 3 Val retained the activity completely. It has been indicated that Met 185 and Leu 437 in P450 BM-3 form strong van der Waals interactions (9). These two residues correspond to Leu 210 and Phe 476 in PGIS by modeling. As mentioned above, the mutant Leu 210 3 Asp completely lost the catalytic activity as predicted from the P450 BM-3 structure, whereas mutation of Phe 476 to Val surprisingly had no effect on the activity. The discrepancy between the predicted and experimental data by site-directed mutagenesis indicates that the molecular model around the upper portion of the channel of PGIS (Fig. 2) containing most of ␣F helix (Thr 206 , Leu 210 , and Leu 213 ) is different from that of P450 BM-3 . Since the polarity and size of AA are different form PGH 2 , it is expected to find disagreement between the structure of P450 BM-3 (using AA as substrate) and PGIS (using PGH 2 as substrate) around the substrate access channel.
Arg 72 in PGIS model is corresponding to Arg 47 in P450 BM-3 structure. This P450 BM-3 Arg residue is located at the mouth of its access channel in strand ␤1-2 close to the molecular surface and is not well defined in the crystal structure (9). Mutation of Arg to Glu in P450 BM-3 blocks the enzymatic reaction, and this residue was suggested to be important in substrate recognition and binding (9,24). Surprisingly, mutation of the conserved Arg in PGIS to charged residue, Glu or Gln, of comparable size did not significantly change the catalytic activity (101 and 80%, respectively). These results imply that the charge group of this Arg is not important in substrate access and binding in PGIS. Unlike the P450 BM-3 , which is a soluble enzyme, the PGIS has a trans-membrane domain at its N-terminal, which is close to the entrance of the substrate binding pocket. We speculated that the entrance of substrate channel of PGIS differs from that of P450 BM-3 because of the influence of membrane topology of PGIS on substrate channel orientation.
Heme Environment-The three-dimensional model of PGIS predicts Cys 441 to be the proximal axial ligand for heme iron. This prediction was confirmed by site-directed mutagenesis. Hatae et al. (25) have recently obtained a similar result. This cysteine provides the thiol moiety to coordinate the heme. It has been shown that P450s exhibit a conserved sequence motif (F-G/S-X-G-X-R/H-X-C-hy-G, where hy denotes any hydrophobic residue) at the cysteine region near the C terminus of P450  proteins (26). This motif in PGIS, WGAGHNHCLG, differs from the conserved motif by two residues: substitutions of Trp (W) for Phe (F) and Asn (N) for Arg/His (R/H). Since Phe to Trp substitution also occurs in nitric oxide synthase, we altered the second substitution from Asn 439 to Ala. This mutant surprisingly exhibited a slight increase in the catalytic activity. Asn 439 is unlikely to be the corresponding conserved Arg or His. This result implies that structure around the heme environment of PGIS is different from that of other P450s.
The crystal structure of P450 BM-3 predicted that Phe 87 forms close van der Waals interactions with the heme on the distal side and the corresponding residue in P450 cam provides hydrophobic interactions with the substrate camphor (27). The corresponding residue in PGIS is Leu 112 . It is interesting to note that alteration of Leu 112 to Phe retained the catalytic activity whereas change of it to a charged residue (Asp) or a smaller residue (Gly) reduced the activity markedly. This result indicates that Leu can substitute Phe in this region to provide hydrophobic interaction with heme and probably also with PGH 2 .
Despite a marked loss of catalytic activity, mutants, Cys 441 3 Ser and Leu 112 3 Asp or Leu 112 3 Gly, in which the heme environment is severely perturbed still expressed intact proteins as detected by Western blots. By contrast, TXAS in which the heme ligation or environment is perturbed by site-directed mutation expresses a very low level of proteins in cells (11). The stability of TXAS protein expression requires heme in a correct orientation. On the other hand, stability of PGIS protein appears to be less dependent on heme. These results epitomize major structural differences between PGIS and TXAS despite a similar substrate binding pocket.
Conclusion-Guided by the three-dimensional model, we have identified nine amino acid residues which are near or line the substrate binding pocket and are important in catalytic activity, especially the residues in ␤1-1, ␤1-2, ␤1-3, and ␤1-4 strands region. We have also confirmed Cys 441 as the proximal axial ligand of heme iron. However, the model is imperfect and fails to predict amino acid residues in other regions of the active site pocket especially in the ␣F helix region, heme environment and the substrate entrance gate-keeper. These inconsistencies between the model and the experimental data offer an interesting opportunity for further experiments to elucidate the different and potential new structural characteristics of PGIS. One approach is to refine the model based on the site-directed mutagenesis data from which additional functionally important residues can be identified. It is hoped that, through these experiments, the PGIS active site pocket and heme site can be more precisely mapped. Eventually, these structures will have to be confirmed by three-dimensional structure derived from x-ray crystallography and/or NMR spectroscopy.