An active site substitution, F87V, converts cytochrome P450 BM-3 into a regio- and stereoselective (14S,15R)-arachidonic acid epoxygenase.

Cytochrome P450 BM-3 catalyzes the high turnover regio- and stereoselective metabolism of arachidonic and eicosapentaenoic acids. To map structural determinants of productive active site fatty acid binding, we mutated two amino acid residues, arginine 47 and phenylalanine 87, which flank the surface and heme ends of the enzyme's substrate access channel, respectively. Replacement of arginine 47 with glutamic acid resulted in a catalytically inactive mutant. Replacement of arginine 47 with alanine yielded a protein with reduced substrate binding affinity and arachidonate sp3 carbon hydroxylation activity (72% of control wild type). On the other hand, arachidonic and eicosapentaenoic acid epoxidation was significantly enhanced (154 and 137%, of control wild type, respectively). As with wild type, the alanine 47 mutant generated (18R)-hydroxyeicosatetraenoic, (14S,15R)-epoxyeicosatrienoic, and (17S,18R)-epoxyeicosatetraenoic acids nearly enantiomerically pure. Replacement of phenylalanine 87 with valine converted cytochrome P450 BM-3 into a regio- and stereoselective arachidonic acid epoxygenase ((14S,15R)epoxyeicosatrienoic acid, 99% of total products). Conversely, metabolism of eicosapentaenoic acid by the valine 87 mutant yielded a mixture of (14S,15R)- and (17S,18R)-epoxyeicosatetraenoic acids (26 and 69% of total, 94 and 96% optical purity, respectively). Finally, replacement of phenylalanine 87 with tyrosine yielded an inactive protein. We propose that: (a) fatty acid oxidation by P450 BM-3 is incompatible with the presence of residues with negatively charged side chains at the surface opening of the substrate access channel or a polar aromatic side chain in the vicinity of the heme iron; (b) the high turnover regio- and stereoselective metabolism of arachidonic and eicosapentaenoic acids involves charge-dependent anchoring of the fatty acids at the mouth of the access channel by arginine 47, as well as steric gating of the heme-bound oxidant by phenylalanine 87; and (c) substrate binding coordinates, as opposed to oxygen chemistries, are the determining factors responsible for reaction rates, product chemistry, and, thus, catalytic outcome.

Recombinant DNA techniques have greatly facilitated the molecular characterization of highly homologous P450 1 isoforms by providing access to sufficient quantities of recombinant proteins and have expedite the task of purifying and characterizing, as single molecular entities, otherwise nearly unresolvable protein isoforms. Nevertheless, a three-dimensional structural description of a single mammalian, membrane-bound P450 is lacking, despite the availability of several highly purified eukaryotic P450 isoforms. Atomic structures are currently available for four water-soluble hemoproteins of bacterial origin (1)(2)(3)(4). Among these, the hemoprotein domain of P450 BM-3, a fatty acid hydroxylase isolated from Bacillus megaterium, has been crystallized and its three-dimensional structure determined at 2-Å resolution (2). Of all the structurally characterized P450s, the hemoprotein domain of P450 BM-3 shows the highest degree of functional and sequence homology to the mammalian enzymes and, in particular, to members of the CYP 4A gene subfamily of microsomal fatty acid hydroxylases (5)(6)(7). Native P450 BM-3 is a single polypeptide (119 kDa) containing a cytochrome P450 domain fused to a cytochrome P450 reductase domain with a 1:1:1 stoichiometric ratio between heme, FAD, and FMN (5)(6)(7). The recombinant P450 BM-3 holoenzyme has been shown to catalyze the NADPHdependent hydroxylation of medium and long chain saturated fatty acids with optimal lengths of 14 -16 carbons (8). While the regiochemistry of saturated fatty acid hydroxylation by P450 BM-3 is more or less carbon chain length-dependent, i.e. as chain length increases, regioselectivity shifts from the -1 to the -2 or -3 fatty acid carbon (7)(8)(9), the enzyme appears unable to catalyze fatty acid -oxidation, a reaction common to several mammalian P450 4A isoforms (6). Purified forms of P450 BM-3 were shown to catalyze the hydroxylation and the epoxidation of monounsaturated fatty acids such as palmitoleic acid (10). Furthermore, studies with palmitoleic acid indicated that the partition ratio between -2 hydroxylation and olefin epoxidation was pH-dependent and subject to differential inhibition by antibodies raised against purified P450 BM-3 (10,11). * This work was supported by National Institutes of Health Grants GM 37922 (to J. H. C.), 31278 (to J. R. F.), and 43479 (to J. A. P.) and a grant from the Renal Care Group, Nashville, TN (to J. H. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
** To whom correspondence should be addressed: Vanderbilt University Medical School, Medical Center North S-3223, Nashville, TN 37232.
Since, in those studies, a single protein was shown to be responsible for fatty acid sp 3 carbon hydroxylation and olefin epoxidation, it was then concluded that P450 BM-3 existed in two or more pH-dependent substrate binding conformations and that antigen-antibody interactions could alter the equilibrium between these conformations (10,11).
Microsomal P450 catalyzes the NADPH-dependent oxidation of AA to regioisomeric monohydroxylated and epoxidized metabolites (12)(13)(14)(15)(16). Depending on the P450 isoform composition of the enzyme preparation under study, arachidonic acid metabolism by P450 can proceed by one or more of the following pathways or reaction types: 1) allylic oxidation, 2) sp 3 carbon hydroxylation at C 16 through C 20 , and 3) olefin epoxidation. While the physiological significance of P450-dependent allylic oxidation remains obscure, the products of the latter two reactions display a variety of potent biological activities and have been implicated in processes such as hormonal signaling (12)(13)(14)(15)(16), the control of cell ion permeability and vascular tone (12)(13)(14)(15)(16), as well as the pathophysiology of experimental hypertension (12)(13)(14)(15)(16). The demonstration of in vivo chiral EET formation by human, rat, and rabbit organs established P450 as a member of the endogenous AA metabolic cascade and suggested new and important roles for the hemoprotein in the biosynthesis of lipid derived mediators of cell and organ function (9,13). As with most eicosanoids, the regio-and stereochemical features of the P450-derived metabolites of AA define biological activity and/or potency and thus, functional significance (9). Studies using purified and/or recombinant mammalian P450 isoforms confirmed that P450 controlled, in an isoform-specific fashion, the regio-and stereoselectivity of oxygen insertion into the AA molecular template (9,11,13).
To initiate the delineation of protein and substrate determinants involved in productive substrate binding and metabolism, we recently characterized the metabolism of AA and several AA analogs by P450 BM-3 (17). These studies demonstrated that AA was metabolized by P450 BM-3 at the fastest rates ever reported for an NADPH-dependent P450 monooxygenase (3.2 mol of product/nmol of P450/min at 30°C), and that AA metabolism by P450 BM-3 regio-and stereoselectively generated (18R)-OH-AA and (14S,15R)-EET in a 5:1 molar ratio and with 96 and 99% optical purities, respectively (17). Similarly, EPA was metabolized by P450 BM-3 to (17S,18R)-EETA (97% optical purity) as the sole reaction product (17). Based on these results and similar experiments performed with the AA analogs EPA and eicosatrienoic acid, we developed a working model of the AA-bound active site of P450 BM-3 from which one concludes that during AA metabolism: 1) a single oxidant species accounts for fatty acid sp 3 carbon hydroxylation and olefin epoxidation; 2) that catalytic outcome is critically dependent on active site spatial coordinates responsible for productive substrate binding and orientation between the heme-bound active oxygen and the acceptor carbon bond(s); and 3) the active site spatial coordinates that control AA binding restricts the degrees of freedom of substrate C-C bond rotation while, at the same time, allowing the molecule a moderate degree of lateral mobility (17). The above mentioned high catalytic rate, as well as the regio-and stereoselectivity displayed by P450 BM-3 during AA metabolism are indicative of a high degree of structural evolutionary specialization and, in particular, of those spatial coordinates that define protein substrate binding and orientation. We have utilized site-specific mutagenesis to change the properties of two amino acid residues flanking either the surface end or the heme end of the P450 BM-3 substrate access channel (2). Here we report that single amino acid replacements in the substrate access channel can either markedly alter the enzyme's regioselectivity or re-sult in inactive proteins. Furthermore, a valine for phenylalanine replacement at position 87 coverts P450 BM-3 into an active and stereoselective (14S,15R)-AA epoxygenase.

MATERIALS AND METHODS
cDNA Manipulations and Protein Purification-The original plasmid containing the cDNA coding for P450 BM-3 was a gift of Dr. Armand Fulco (Department of Biochemistry, UCLA, Los Angeles, CA). The cDNA was subcloned into the pIBI expression vector and expressed in DH5␣ Escherichia coli exactly as described (5). Site-specific mutations were introduced into the P450 domain using the M13 two-primer method of Kunkel et al. (18). Mutants were screened with the radiolabeled mutagenic oligomers and the incorporated changes confirmed by DNA sequencing. Mutated cDNAs were subcloned into the pIBI vector and expressed as above.
The recombinant P450 BM-3 proteins were purified as described in Ref. 5. Briefly, P450 BM-3 was collected from cell lysates by precipitation with ammonium sulfate (60% saturation). After centrifugation and dialysis versus 50 mM MOPS buffer (pH 7.4) containing 50 mM KCl, the sample was applied to a DEAE-cellulose column equilibrated with dialysis buffer and the proteins were eluted using a linear salt gradient from 50 to 250 mM KCl. Fractions with a 416 nm to 280 nm absorbance ratio greater than 0.9 were pooled, concentrated, applied to a Sephacryl 200 HR column and eluted with 50 mM KCl and 50 mM MOPS buffer (pH 7.4). Fractions with 416 nm to 280 nm absorbance ratios greater than 1.2 were pooled and utilized as such. P450 concentrations were determined from the difference absorbance spectrum of the CO-complex of the ferrous form of the enzyme using an extinction coefficient of 91 mM Ϫ1 cm Ϫ1 for the wavelength pair of 450 versus 490 nm (19). The substrate binding spectra used for the determination of K s , the spectral binding constant, were obtained after the addition of increasing concentration of substrate at 25°C. Difference spectra were recorded with a Hewlett Packard diode array spectrophotometer equipped with a Peltier temperature control device. Absorbance data for substrate binding were analyzed with the program Microsoft Excel.
Enzyme Activity Determinations-Oxygen and NADPH consumption were measured with a Clark-type oxygen electrode (YSI Instruments, Yellow Springs, OH) and a Hewlett Packard spectrophotometer, respectively, at room temperature and using 20 -50 nM solutions of the enzyme in 50 mM MOPS buffer (pH 7.4) containing 50 -100 M fatty acid. After a 5-min preincubation, reactions were started by the addition of NADPH. NADPH utilization was measured at 340 nm and concentrations calculated using an extinction coefficient of 6.22 mM Ϫ1 cm Ϫ1 . For product quantification and structural characterization, incubations were performed at 30°C under atmospheric air and with vigorous mixing. Reaction mixtures in 50 mM Tris-Cl buffer (pH 7.4) containing 10 mM MgCl 2 , 150 mM KCl, 8 M sodium isocitrate, isocitrate dehydrogenase (1.0 IU/ml), dilauroylphosphatidylcholine (0.05 g/ml), and P450 BM-3 (2-10 nM, final concentration) were incubated 2.5 min prior to the addition of the sodium salt of either AA or EPA (25 mM each in 0.05 mM Tris-Cl buffer (pH 8.0) to final concentrations of 50 -100 M, each. After 1 min, reactions were started by the addition of NADPH (1 mM, final concentration). At different time points, aliquots of the incubates were withdrawn and the organic soluble products extracted thrice with equal volumes of ethyl ether containing HOAc (0.05%, v/v). After solvent evaporation under an stream of N 2 , the products were resolved by RP-HPLC on a 5-m Dynamax Microsorb C 18 column (4.6 ϫ 250 mm, Rainin Instruments Co., Woburn, MA) using a linear solvent gradient from 49.95% CH 3 CN, 49.95% H 2 O, 0.1% HOAc to 99.9% CH 3 CN, 0.1% HOAc over 40 min at 1 ml/min (20). Products were quantified by on-line liquid scintillation using a Radiomatic Flo-one ␤-Detector (Radiomatic Instruments, Tampa, FL).
Structural Characterizations-The identification of 18-OH-AA, 14,15-EET, and of 17,18-EETA was done using published methodology (20 -22) and confirmed with synthetic standards. For the characterization of 14,15-EETA, the organic soluble material extracted from incubates containing [1-14 C]EPA (100 M final concentration, 0.1 Ci/mol), 5 nM P450 BM-3, and 1 mM NADPH was resolved by RP-HPLC as described above. The radioactive fraction eluting from the HPLC column with a retention time of 19.7 min was collected batchwise and submitted to catalytic hydrogenation under PtO 2 . For structural characterization, the NICI/GC/MS and chromatographic properties of the metabolite were compared to those of authentic 14,15-epoxyeicosanoic acid. PFB esters were prepared by reaction with pentafluorobenzyl bromide in triethylamine as described (21,22). RP-HPLC was done on a 5-m Dynamax Microsorb C 18 column (4.6 ϫ 250 mm, Rainin Instruments Co.) using a linear solvent gradient from 49.95% CH 3 CN, 49.95% H 2 O, 0.1% HOAc to 99.9% CH 3 CN, 0.1% HOAc over 40 min at 1 ml/min. Normal phase HPLC was performed on a 5-m Dynamax Microsorb Silica column (4.6 ϫ 250 mm, Rainin Instruments Co.) using either a solvent mixture of 99.65% n-hexane, 0.25% isopropanol, and 0.1% HOAc at 2 ml ml/min (for the eicosanoic acid) or 99.95% n-hexane, 0.05% isopropanol at 2 ml/min (for the PFB ester). The RP-HPLC retention times for the hydrogenated EPA metabolite and for synthetic 14,15-epoxyeicosanoic acid and for the corresponding PFB esters were 33.1 and 46 min, respectively, The normal phase HPLC retention time for the hydrogenated EPA metabolite and the corresponding synthetic standards were 21 and 12 min, respectively, for the free acid and PFB ester.
Gas Chromatography-Mass Spectral Analysis-Samples were dissolved in dodecane and analyzed by NICI/GC/MS on a Nermag R1010C quadrupole instrument interfaced to a Varian Vista Gas chromatograph utilizing helium and CH 4 as carrier and reagent gases, respectively. Splitless injections were made onto a 30 m SPB-1 fused silica capillary column (0.32 mm (inner diameter), 0.25 m coating thickness, Supelco Inc., Bellefonte, PA). After 1 min at 100°C, the oven temperature was raised to 300°C at 10°C/min.

RESULTS AND DISCUSSION
The three-dimensional structure of P450 BM-3 revealed a heme prosthetic group positioned at the end of a long and narrow channel (8 -10 Å wide) lined with mostly hydrophobic amino acid residues (2). It was proposed that during catalysis, this channel, "the substrate access channel," provided access to the heme active site and could participate in determining substrate selectivity and the regio-and stereoselectivity of oxygen insertion (2). An arginine residue (Arg 47 ) flanks the end of the substrate access channel facing the surface of the molecule (2). At the opposite end, the heme end of the channel, a phenylalanine residue (Phe 87 ) is located in close proximity to the distal face of the heme prosthetic group (2). Based on the unique spatial geometries of these residues, it was proposed that, on the one hand, Arg 47 controls entry to the substrate access channel and, therefore, fatty acid selectivity and, on the other hand, the aromatic side chain of Phe 87 limits and/or controls substrate C-H acceptor bond access to the enzyme's heme iron and thus, the regioselectivity of oxygen attack (2). In the absence of atomic coordinates for substrate-bound P450 BM-3, we utilized a molecular modeling approach to visualize AA binding to P450 BM-3 (17). For these studies, the fatty acid carboxylate was placed within charge coupling distance of the Arg 47 guanidinium group and the remainder of the molecule was built into the volume of the access channel defined by Arg 47 at the surface and Phe 87 above the heme (2,17). To optimize occupancy, Phe 87 was moved slightly back, thus allowing the -end of AA to occupy an available active site cavity (2,17). In this active site induced conformation, the fatty acid methyl end, i.e. the carbon, bends upward and away from the heme iron and the AA C18 pro-R hydrogen is positioned in close proximity to the enzyme's heme iron (17) (Fig. 1). Moreover, an additional small displacement of Phe 87 places the AA 14,15-olefin in optimal proximity to the P450 BM-3 heme iron (17). Based on this model, we suggested that AA-bound P450 BM-3 oscillates between two alternate conformers characterized by the relative position of Phe 87 with respect to the heme iron. These alternate conformations allow for a small degree of substrate lateral displacement, responsible for oxygen attack at C18 and the 14,15-olefin of AA (2,17). To further define the roles of Arg 47 and Phe 87 , as the outer and inner boundaries of the substrate access channel (2) (Fig. 1), we replaced them by site-specific mutagenesis. Mutated cDNAs were expressed, the corresponding recombinant proteins purified and their enzymatic properties studied as described under "Materials and Methods." A summary of the amino acid replacements introduced into the protein are shown in Table I. The positively charged side chain of Arg 47 (Fig. 1) was either eliminated by replacement with alanine or replaced with a negatively charged glutamic acid ( Table I). The aromatic ring of Phe 87 (Fig. 1) was either removed by replacement with valine or, alternatively, its polarity was increased by substitution with tyrosine (Table I).
Substrate Access Channel: Arg 47 Replacements-The poorly defined nature of Arg 47 in the two crystallographically independent P450 BM-3 molecules analyzed indicated high struc- The atomic coordinates for "molecule A" of P450 BM-3 (2) were used to model the conformation of arachidonic acid in the substrate access channel. The fatty acid was docked into the access channel using the program InsightII (Biosym Corp., San Diego, CA). Inappropriate van der Waals contacts were minimized manually. The arachidonic acid carboxylate was positioned within electrostatic interaction distance of the Arg 47 guanidino group with its -end inserted into a pre-existing cavity above the heme group. The ␤-sheet strands on either side of Arg 47 are colored in blue. The carbon atoms in Arg 47 and Phe 87 are yellow and the nitrogens blue. The heme group is in dark gray with its nitrogens blue, oxygens red and iron orange. The ␣-carbon backbone of the I-helix is green and the fatty acid carbon chain pink.
tural flexibility and suggested a role for this residue in the monooxygenation of fatty acids of different chain lengths (2). The protein crystal structure showed that the flexible, positively charged, side chain of Arg 47 protruded into the lumen of the substrate access channel (2). The catalytic significance of these Arg 47 -associated structural features, i.e. side chain length and charge, were addressed by replacing the amino acid with either glutamic acid or alanine ( Fig. 1) (Table I). Replacement of Arg 47 by a negatively charged glutamic acid residue resulted in a mutant enzyme (Table I) spectroscopically identical to the wild type, however, unable to metabolize AA or EPA (data not shown). To reduced the potential for charge repulsion between the ionizable fatty acid carboxylate and the side chain of Glu 47 in the mutant enzyme, we prepared the methyl ester derivatives of AA and EPA and incubated them with the mutant protein. Even at high P450 concentrations (Յ0.5 M), the R47E mutant was unable to metabolize the methyl esters of AA and EPA (data not shown). Furthermore, no significant NADPH utilization or NADPH-dependent oxygen reduction was observed with the R47E mutant P450 BM-3 incubated in the absence or presence of AA. Thus, catalytic turnover appears to be incompatible with the presence of a negative charge at the surface opening of the substrate access channel. Finally, the addition of AA or EPA (from 1 to 50 M) to a solution of this mutant P450 BM-3 failed to induce a typical spectral manifestation of binding to the enzyme's active site.
The replacement of Arg 47 (Fig. 1) by alanine (Table I) generated a mutant protein lacking a flexible and positively charged protruding side chain (2) and with potentially enhanced active site accessibility. Incubation of the R47A mutant P450 BM-3 with [1-14 C]AA and NADPH (2.5 nM, 100 M, and 1 mM final concentrations, respectively) resulted in the formation of two radioactive metabolites with HPLC retention times identical to those of authentic 18-OH-AA and 14,15-EET (Fig. 2). While the products formed are the same as those previously reported for wild type P450 BM-3 (17), the Arg for Ala replacement resulted in a reduced overall rate of AA oxidation (82% of wild type rates, Table II) (17). Furthermore, the amino acid replacement affected AA sp 3 carbon hydroxylation and olefin epoxidation in opposite manners, i.e. an increased AA 14,15olefin epoxygenase activity was accompanied by a concomitant reduction in AA -2 carbon hydroxylation (154 and 72% of wild type rates, respectively) (Fig. 2, Table II). An enhanced epoxygenase activity was also observed when the R47A mutant was incubated with EPA. As shown in Fig. 2 and Table III, the mutant enzyme metabolized EPA to 17,18-EETA with nearly absolute regioselectivity and at reaction rates higher than those reported for wild type P450 BM-3 (137% of wild type rates; Fig. 2, Table III) (17). Neither wild type P450 BM-3 nor the R47A mutant metabolized methyl-AA or methyl-EPA (data not shown). As with wild type enzyme, during AA metabolism by the R47A mutant, NADPH oxidation is coupled to oxygen reduction and substrate oxidation, as revealed by a 1:1 stoichiometry between NADPH and oxygen utilization (Table IV). However, in contrast with wild type P450 BM-3, the R47A mutant enzyme diverts a small but significant portion of NADPH supplied electrons to H 2 O 2 formation (Table IV). These result indicate that electron flow and oxygen chemistries are only marginally disturbed by the Arg 47 for Ala replacement (Tables III and IV).
To probe for changes in substrate binding affinity and/or spatial orientation resulting from the Arg 47 for Ala substitution, we determined the spectral dissociation constants for AA and EPA as well as the chirality of the AA and EPA metabolites generated by the R47A mutant. Compared to wild type P450 BM-3, and as estimated by the values of the corresponding spectral dissociation constants shown in Table V, the mutated enzyme showed a substantially decreased binding affinity for AA and EPA. These results, in conjunction with the above mentioned changes in reaction rates, suggest that charge interactions between the Arg 47 side chain and the fatty acid carboxylates may stabilize substrate binding in the spatial conformation that favors oxygenation at the substrate C18 C-H bond, i.e. AA sp 3 -2 carbon hydroxylation and EPA 17,18epoxidation. Similarly, the increased AA and EPA epoxygenase activity observed with the R47A mutant may reflect a decreased binding rigidity and greater freedom of substrate lateral displacement brought about by the removal of Arg 47 -asso-   ciated charge interactions with the fatty acid carboxylate (17) (Fig. 1). Chiral phase HPLC analysis of the 18-OH-AA, 14,15-EET, and 17,18-EETA isolated from incubates containing either AA or EPA, R47A P450 BM-3, and NADPH showed that the enantiofacial selectivity of the mutant enzyme was not only identical to that reported for wild type P450 BM-3 (Table VI) (17) but also that the high degree of catalytic asymmetry was preserved (Table VI). It was therefore concluded that removal of the positively charged and protruding Arg 47 side chain does not change and/or modify the active site spatial coordinates responsible for active site substrate orientation with regard to the heme-iron oxygenating locus, i.e. nearly enantiomerically pure (18R)-OH-AA, (14S,15R)-EET and (17S,18R)-EETA are the metabolites generated by mutant and wild type P450s (17) (Table VI).
Based on the above results, we concluded that: (a) while a positively charged residue at position 47 is not essential for catalysis, catalytic turnover is incompatible with the presence of a negatively charged glutamic acid residue at this position; and (b) the change of Arg 47 to alanine appears to enhance the occupancy of a second binding orientation, which facilitates 14,15-epoxidation over AA hydroxylation at carbon 18 (Table  II). This second substrate binding orientation in the R47A mutant most likely allows the AA molecule to penetrate deeper into the access channel, permitting the 14,15-olefin to approach the heme iron at a higher frequency. In summary, charge interactions between the Arg 47 side chain and the fatty acid carboxylic acid may provide an anchoring point, which leads to increased substrate binding rigidity and, thus, higher turnover and regiochemical selectivity (Fig. 1).
Substrate Heme-Iron Access: Phe 87 Replacements-A distinctive feature of AA regioselective metabolism, and particularly of oxidations occurring at the fatty acid , -1, -2, or -3 carbon atoms, is the efficient, active site-dependent insulation of reactive olefins and bis-allylic methylene carbons from the heme-bound reactive oxygen. As mentioned, in P450 BM-3 the aromatic ring of Phe 87 extends into the heme pocket and is positioned above and nearly perpendicular to the porphyrin plane ( Fig. 1) (2). It is readily apparent from the protein crystal structure (2), as well as from the active site model of AA-bound P450 BM-3 (17), that this residue could restrict substrate access to the enzyme's heme-iron and thus control the regioselectivity of oxygenation (17). Using the atomic coordinates for P450 BM-3, we constructed a model of the solvent-accessible surface at the enzyme's substrate access channel (2,17). As shown in Fig. 3A, the aromatic side chain of Phe 87 occupies a substantial portion of the access channel volume proximal to the heme face, markedly reducing substrate heme accessibility and thus increasing binding rigidity (17). Furthermore, the volume of the substrate binding region of P450 BM-3 available to solvent approximates that of a regular cone with its base facing the surface opening of the access channel and its vertex the heme end (Fig. 3A). A similarly constructed model of the F87V mutant substrate access channel is shown in Fig. 3B. Replacement of Phe 87 by Val eliminates the steric effects of the Phe aromatic side chain in the vicinity of the heme and increases substrate heme accessibility. Importantly, while the replacement of Phe 87 by Val results in a significant expansion   of the volume available for substrate binding near the heme prosthetic group, it has little or no effect at the opposite end, the surface opening end of the access channel. Thus, compared to wild type P450 BM-3, the volume of the substrate binding region in F87V is more or less cylindrical instead of conical (Fig. 3, A and B). To study the functional role of Phe 87 , we replaced it with tyrosine or valine. With the Tyr for Phe 87 replacement (Table I), we sought to increase the polarity of the phenyl ring and thus add a polarity component to its steric effects. On the other hand, in the valine mutant (F87V) ( Table  I), we sought to remove from the heme environment steric effects attributable to the aromatic ring. The addition of a hydroxyl group to the Phe 87 aromatic side chain, i.e. replacing this residue for tyrosine, generated a mutant (F87Y) protein catalytically inactive toward AA, EPA, and their corresponding methyl esters. Importantly, the Soret region absorbance spectra of the F87Y mutant were identical to those of the wild type P450 BM-3 (data not shown). Assuming that the tyrosine replacement does not introduce significant protein structural alterations, the metabolism of polyunsaturated fatty acids by P450 BM-3 appears to be incompatible with a heme environment containing the oxygen binding site in close proximity to a polar aromatic amino acid side chain. Significant restrictions in substrate heme accessibility are likely to be responsible for the aforementioned lack of catalytic turnover. In support of the latter, the addition of AA or EPA to a solution of F87Y P450 BM-3 failed to elicit typical spectral manifestations of substrate binding and heme iron spin change. While unable to metabolize AA, the F87Y mutant did support significant NADPH-dependent oxygen reduction (Table IV). Furthermore, the absence of measurable substrate oxidation and of H 2 O 2 formation, as well as the observed 1.9 -2.1 stoichiometric ratio between NADPH and oxygen utilization (Table IV), indicates that the F87Y mutant catalyzes a four-electron reduction  F87V mutant proteins (B). The atomic coordinates of "Molecule A" of P450 BM-3 (2) were utilized to construct both models. Conally surfaces (25) were created using default parameters (In-sightII, Biosym Corp.) and are shown as white dots. van der Waals contacts between amino acid side chains and carbon backbone atoms were minimized maintaining appropriate bond angles. A spacefilling model of Arg 47 , Phe 87 , and the heme group is displayed. The ␣-carbon backbone of the I-helix is shown in dark green. For clarity, additional residues and/or atoms were not included. Carbon, green; nitrogen, blue; oxygen, red. of molecular oxygen to H 2 O. Control experiments demonstrated that: (a) the F87Y mutant protein lacked significant catalase-like activity, and (b) NADPH and oxygen utilization was partially fatty acid-independent (data not shown). The four-electron reduction of oxygen to water is characteristic of cytochrome oxidase(s), the structurally complex and multifunctional hemoproteins that serve as the terminal oxidases of the respiratory chain. Because of its importance for hemoprotein biochemistry and function, this unusual activity of the F87Y mutant is currently under investigation.
The nearly absolute degree of enantiofacial selectivity with which wild type P450 BM-3 catalyzed AA sp 3 hydroxylation and 14,15-olefin epoxidation (17) indicated that (a) AA binding to the active site of P450 BM-3 severely limits the molecule's freedom of C-C bond rotation, (b) the active site binding coordinates allow for a limited degree of substrate displacement along the longitudinal axis of the access channel resulting in more than one catalytically productive active site-AA complex, and (c) Phe 87 provides a distinct steric barrier to active sitesubstrate configurations favoring oxygen delivery to the AA or EPA 14,15-olefin (17) (Figs. 2 and 3A). Incubation of [1-14 C]AA with the F87V mutant P450 BM-3 and NADPH, followed by product extraction and HPLC analysis, revealed this single amino acid substitution caused important alterations in the regioselectivity of AA oxidation by P450 BM-3. Thus, as reported (17) and shown in Fig. 4, AA metabolism by wild type P450 BM-3 generated a 5:1 mixture of 18-OH-AA and 14,15-EET. Significantly, removal of the protruding aromatic side chain associated with Phe 87 completely eliminated the catalysis of AA sp 3 carbon hydroxylation and converted the mutant protein into a regio-and stereoselective AA (14S,15R)-epoxygenase (100% of total products, 99% optical purity) (Fig. 4, Tables VII and VIII). Importantly, the enantiofacial selectivity of the mutant F87V epoxygenase is identical to that of wild type P450 BM-3 (Table VIII) ,suggesting that: (a) the amino acid replacement did not substantially change the spatial orientation of the acceptor bond(s), and (b) the wild type and mutant P450s utilize similar substrate binding coordinates. In agreement with this, the AA binding constants for wild type and the F87V mutant enzymes were also similar (Table V).
Compared to wild type, the F87V mutant enzyme metabolized AA at a significantly reduced overall rate (approximately 36% of wild type rates, Table VII). This reduced catalytic rate could be accounted for by the loss of AA sp 3 carbon hydroxylation activity since, under identical conditions, the mutant enzyme catalyzed 14,15-EET formation at a rate twice that of wild type P450 BM-3 (Table VII). It is of interest that, in addition to the above effects on reaction rates, the replacement of Phe 87 by Val resulted in the generation of small amounts of H 2 O 2 due to non-productive cycles of O 2 reduction and increased NADPH oxidase activity (Table IV). These results indicate that the aromatic side chain of Phe 87 may facilitate tight coupling between electron transfer, oxygen reduction and activation, and acceptor C-H bond insertion. Assuming that the Phe 87 for Val replacement does not introduce major protein structural alterations and/or changes in the kinetics of product release, the documented catalytic dissociation between P450 BM-3 supported hydroxylation and epoxidation reactions is in agreement with the proposal that these reactions are mediated by distinct P-450 BM-3 active site-substrate conformations (Fig. 3, A and B) (2,17). Studies of the metabolism of AA and of several AA analogs by P450 BM-3 indicated that steric factors, as opposed to the chemistry of the reactive oxygen intermediate, may play the predominant role in determining the nature of the reaction products (17). Furthermore, our current understanding of the biochemistry of oxygen activation by P450 enzymes supports the concept that sp 3 carbon hydroxylation   and olefin epoxidation are mediated by a common oxidant. Inasmuch as olefin epoxidation is thermodynamically preferred over sp 3 carbon hydroxylation, the decreased rates of AA metabolism observed with the F87V mutant enzyme may reflect changes in electron transfer kinetics brought about by the removal of the Phe 87 aromatic side chain from the environment of the protein redox active center. Alternatively, charge interactions between the side chain of Arg 47 and the substrate carboxylic acid, the proposed substrate anchoring site, may limit heme-iron access to the AA and EPA 14,15-olefins by providing a significant energy barrier to the freedom of substrate longitudinal displacement.
Incubation of [1-14 C]EPA with the F87V mutant protein resulted in the NADPH-dependent formation of two radioactive products with HPLC retention times identical to those of au-thentic 17,18-EETA and its hydration product, 17,18-DHETA (17) (Fig. 5). A third metabolite, eluting with an HPLC retention of 19.7 min was identified as 14,15-EETA based on the following criteria: (a) catalytic hydrogenation over PtO 2 generated a product with normal and RP-HPLC chromatographic properties identical to those of authentic 14,15-epoxyeicosanoic acid (21 and 33.1 min for normal and RP-HPLC, respectively), (b) catalytic hydrogenation followed by derivatization to the corresponding PFB ester yielded a product with normal and RP-HPLC retention times identical to those of the synthetic PFB ester of 14,15-epoxyeicosanoic acid (12 and 46 min for normal and RP-HPLC, respectively), and (c) catalytic hydrogenation, followed by derivatization to the corresponding PFB ester and NICI/GC/MS analysis showed the presence of major negative fragment ions at m/z 325 (base peak; M-PFB), 309 (25% abundance; loss of O and PFB) and 307 (21% abundance; loss of PFB and H 2 O). This mass spectrum was identical to that generated, under similar conditions by the PFB ester of synthetic 14,15-epoxyeicosanoic acid (data not shown). For chiral characterization of enzymatically derived 14,15-EETA, the metabolite was reduced by catalytic hydrogenation, converted to the corresponding PFB ester, and compared by chiral phase HPLC with racemic and/or enantiomerically pure synthetic standards. As shown in Fig. 6, Chiralcel OD chromatography separated the optical antipodes of synthetic 14,15-epoxyeicosanoate-PFB with base-line resolution. Co-injection of a mixture of racemic standards with enzymatically produced, 1-14 C-labeled metabolite showed coelution with (14S,15R)-epoxyeicosanoate-PFB (Fig. 6).
The reported regio-and stereoselective epoxidation of EPA to (17S,18R)-EETA by P450 BM-3 was attributed to efficient trapping of the heme-bound active oxygen by the fatty acid reactive 17,18-olefin electron cloud (17). As illustrated by the HPLC chromatograms in Fig. 5   of total products for 14,15-and 17,18-EETA ϩ 17,18-DHETA, respectively) (Fig. 5, Table IX). As with AA, the removal of the Phe 87 aromatic ring from the heme environment was associated with decreased rates of EPA oxidation (Table IX) and concomitant increases in substrate-dependent, NADPH oxidase (Table IV). Furthermore, as shown in Tables V and VIII, the Phe 87 replacement by Val did not significantly change the affinity of the protein for EPA, nor did it affect the enantiofacial selectivity of oxygenation (Table VIII). Thus, while the removal of the aromatic ring in Phe 87 allows further displacement of the EPA molecule along the longitudinal axis of the access channel (Fig. 3B), the overall EPA binding coordinates do not appear to be significantly altered by this replacement (Table VIII). It is of interest that regardless of the fatty acid, the si, re faces of the 14,15-and 17,18-olefin are always selected for epoxidation (Fig.  7), indicating that regioselectivity results predominantly from freedom of longitudinal displacement within the confines of the substrate access channel and not from oxygen chemistries and/or unique binding configurations.
The presence in P450 BM-3 of an hydrophobic substrate binding site uniquely adapted to AA binding and metabolism not only indicates high evolutionary specialization, it also raises interesting questions with regards to the mechanism and/or driving force(s) responsible for the release of the oxygenated products. As reported (17), the products of AA oxidation are efficiently recycled through the enzyme system and undergo extensive secondary oxidations, even in the presence of high AA concentrations (17). Thus, (14S,15R)-EET and (18R)-OH-AA compete effectively with AA for binding to the enzyme active site. Hence, it is likely that during catalytic turnover, reversible changes in active site configuration may participate in facilitating oxygenated product release.
All the changes in catalytic outcome and/or reaction rates reported here are the consequence of single amino acid replacements in the vicinity of the heme prosthetic group. Furthermore, these substitution-dependent changes appear to take place in the absence of major protein and/or active site structural modifications, i.e. changes in enzyme regioselectivity occurred without alterations in stereoselectivity. The pioneering studies of Negishi and collaborators and others with microsomal P450 isoforms (23,24), in conjunction with the results reported here, show that substrate selectivity and catalytic outcome can be critically dependent on the chemical nature of unique residues located in finite, more or less predictable, areas of the protein active site. Thus, instead of the chemical properties of the oxidant species, the chemistry of the reaction products may be controlled by active site binding coordinates and the resulting spatial orientation of acceptor bonds with respect to the heme-bound reactive oxygen. These as well as other published studies are beginning to provide a molecular description of active site determinants for regio-and stereoselective metabolism in different P450 isoforms (24). These studies also contribute to the delineation of the molecular basis responsible for the catalytic versatility of these functionally diverse but, otherwise, structurally homologous proteins.
Finally, while it is apparent that charge interactions associated with the side chain of Arg 47 serve as a fatty acid anchoring point and thus may regulate substrate displacement along the longitudinal axis of the access channel, the Phe 87 aromatic side chain controls heme access at the distal end of the access channel. Consequently, we propose that the high turnover regio-and stereoselective metabolism of fatty acids such as AA and EPA is controlled by substrate anchoring at the surface opening of the access channel and by steric gating of the hemebound reactive oxygen. Furthermore, the experimental data also shows that substrate-active site binding coordinates, as opposed to oxygen chemistries, are the predominant factors responsible for reaction rates and the chemistry of the reaction products.