Engineering Mouse P450coh to a Novel Corticosterone 15a-Hydroxylase and Modeling Steroid-binding Orientation in the Substrate Pocket*

The F209L mutation alters specificity of P450coh from coumarin 7-hydroxylation to 15a-hydroxylation of 1 1-deoxysteroids such as testosterone and 1 l-deox- ycorticosterone. Neither the wild-type nor F209L exhibits activity toward 11B-hydroxysteroids including corticosterone. Mutation of Phe-209 to Asn, however, confers on mutant F209N a high corticosterone 15a-hydroxylase activity. F209V also exhibits low corti- costerone 15a-hydroxylase activity; K , and V,,, are 10-fold higher and lower, respectively, than for F209N. The results are consistent with the hypothesis that direct interaction of Asn-209 with l l O H is responsible for high corticosterone 15a-hydroxylase ac- tivity. To support this hypothesis, a possible steroid-binding orientation is modeled in the substrate pocket of P45Ocam. Our weighted homology and constrained alignments map residue 209 of P450coh to Met-184 and Met- 191 of P450cam. Energy minimization of cor- ticosterone in the substrate pocket results ia the 110H of the steaoid directed toward Met-184 (7 A) and Met- 191 (16 A), and in C15 initially minimized (using the united atom force field) only resi- dues 86-101, 173-177, 244-250, and 395-397 by fixing (no force on their atoms) the heme, steroid, and remaining parts of the P450cam.

The F209L mutation alters specificity of P450coh from coumarin 7-hydroxylation to 15a-hydroxylation of 1 1-deoxysteroids such as testosterone and 1 l-deoxycorticosterone. Neither the wild-type nor F209L exhibits activity toward 11B-hydroxysteroids including corticosterone. Mutation of Phe-209 to Asn, however, confers on mutant F209N a high corticosterone 15ahydroxylase activity. F209V also exhibits low corticosterone 15a-hydroxylase activity; K , and V,,, are 10-fold higher and lower, respectively, than for F209N. The results are consistent with the hypothesis that direct interaction of Asn-209 with l l O H is responsible for high corticosterone 15a-hydroxylase activity. To support this hypothesis, a possible steroidbinding orientation is modeled in the substrate pocket of P45Ocam. Our weighted homology and constrained alignments map residue 209 of P450coh to Met-184 and Met-191 of P450cam. Energy minimization of corticosterone in the substrate pocket results ia the 110H of the steaoid directed toward Met-184 (7 A) and Met-191 (16 A), and in C15 located near the sixth axial position of the heme. The steroid-binding model suggests that the P450cam's substrate pocket may be conserved in the mammalian P450 and can accommodate a steroid molecule, and that residue 209 appears to be located at the critical site that determines the steroidsubstrate specificity of a P450 depending on the type of group at the 1 1-position of steroid molecule.
As a family of structurally related enzymes, P450s exhibit extreme diversity in hydroxylase activities. T o understand the structural basis for the divergent activities, the substratebinding orientation in the heme-pocket must first be deline-* 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. ated, and this question is currently of major interest in P450 research.
Recent site-directed mutagenesis studies indicate that the specificity of the P450s can be altered by single amino acid substitutions. Mutation of Phe-209 to Leu, for example, converts the substrate specificity of P450coh (2A5)' from COUmarin to steroid hydroxylation (1). The type of residue 478 of P4502B1 determines the regioselectivity of steroid metabolites (2), and Arg-346 is responsible for delineating the 17ahydroxylase activity in P45Ol7, from the lyase activity (3). Other reports have also described how amino acid substitutions affect the hydroxylase activities of P450s (4-7). Moreover, homology alignment and computer modeling based on the bacterial P450cam have provided the structural basis for some, but not all, of the observed specificities (8).
To study the steroid-binding orientation in the substrate pocket, we have focused on residue 209 of mouse P450coh, because our previous work suggests that this residue is close to the sixth ligand of the heme and plays a key role in altering the P450 activity (9,10). To this end, we measured steroid hydroxylase activity of the mutants of P450coh using different steroids as substrates, and subsequently engineered a novel P450 which catalyzes corticosterone l5a-hydroxylase activity. We then organized the biochemical information to define the steroid-binding orientation in the substrate pocket of the P450 using the three-dimensional structure of P450cam (P450 101) as the model. We propose a steroid-binding orientation in the substrate pocket of the P450.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis and Expression in Yeasts-Construction of P450coh mutants used in this study was as described in our previous papers (1,9,10). The mutated cDNAs were ligated to yeast expression vector pAAH5 and transformed to Sacchromyces cerevisiae AH22 cells as previously described by Oeda et al. (11).
Purification of P450 and Steroid Hydroxylase Activity-We prepared microsomes from recombinant yeast and purified the P450s using previously published methods (9), except that an additional hydroxylapatite column was used to remove endogenous yeast P450. Hydroxylapatite column was equilibrated with 10 mM potassium phosphate buffer, pH 7.25, containing 20% glycerol, 1 mM dithiothreitol, and 0.2% sodium cholate. The P450 from the previous octylamino-Sepharose 4B was bound on the column, washed by 100 mM potassium phosphate buffer, pH 7.25, and then eluted by increasing the buffer concentration to 400 mM.
Steroid hydroxylase activity was reconstituted as described previously (12). The reconstitution system consisted of purified P450 (10 pmol), rat NADPH-cytochrome P450 reductase (30 pmol), NADPH and steroid (100 p~) in 0.5 ml of 50 mM Tris-HC1 buffer, pH 7.5. The steroid metabolites were extracted with methylene chloride, separated by thin layer chromatography with toluene/acetone (1:l) as the developing solvent, and exposed to x-ray film. Finally, spots containing the metabolites were scraped from plates and counted using liquid scintillation counting.
Proton NMR Identification of the Steroid Metabolites-Approximately 350 pmol of each P450 (F209L or F209N) and 1.4 +mol of Coumarin 7-hydroxylase P450coh is a member of the mouse 2A subfamily. P450 2A5 and Cyp 2A5 are the given standard nomenclatures for the P450 and its gene. Accordingly, the standard name of P450cam is P450 101. 759 steroid (deoxycorticosterone or corticosterone) were incubated at 37 "C for 1 h under the above described reconstitution conditions, except that the reaction volume was 4 ml. The steroid metabolites were extracted with chloroform/methanol (1:l) and applied to a silica gel column (5 mm X 3.5 cm). The column was washed with toluene and the metabolite eluted with toluene/methanol(82), then subjected to HPLC chromatography on a TSK gel ODs-80TM (4.6 mm X 25 cm) using methanol/water (9:l). Approximately 0.2 mg of the metabolite was obtained, dried, and dissolved in 0.4 ml of "100%" chloroform-dl (Cambridge Isotope Laboratories). The samples were then transferred to 5-mm (outer diameter) sample tubes to obtain the 'H NMR spectrum at 500 MHz using a GN 500 spectrometer (GE Instruments) equipped with a Nicolet 1280 data system and 2933 pulse programmer.
Homoiogy Alignment and Graphic Analysis-The GCG program BESTFIT was used with default penalties for gaps and insertions to align P450coh to P450cam. To obtain a constrained alignment .Of P450coh against P450cam, we first defined a sphere (radius of 15 A) around the sixth axial position in the P450cam pocket by taking the distance between the axial position and the a-carbon position of residue 209. The standard Needleman-Wunsch algorithm (13) was then modified to find the best alignment for which residue 209 mapped to some bacterial residue in the sphere. To model a steroid into the P450cam pocket, we used a Silicon Graphics workstation and the program MULTI (14) to dock corticosterone (15) into the pocket, then energy-minimized with the program AMBER 3.0a (16). Residues Phe-87, Tyr-96, Phe-98, Thr-101, Thr-185, Leu-244, Val-247, Gly-248 Val-295, Ile-395, and Val-396 all had atoms that fit within a 6.5-A sphere centered on camphor bound to P450cam. Thus we initially minimized (using the united atom force field) only residues 86-101, 173-177, 244-250, and 395-397 by fixing (no force on their atoms) the heme, steroid, and remaining parts of the P450cam. In this alignment, C15 of the steroid molecule is 2.9 8, from the activating oxygen in the sixth axial position and no atom of the P450cam is closer than 2.3 A to the steroid. Finally, the structure from this constrained minimization was subjected to a full minimization of all atoms in the P450cam protein and heme while freezing the position of the steroid.

RESULTS AND DISCUSSION
The metabolites of deoxycorticosterone and corticosterone were determined by the analysis and complete assignment of their 'H NMR spectra. Two-dimensional, double quantumfiltered homonuclear shift correlation spectroscopy (DQCOSY)' (17) was used to determine the scalar J-connectivity between geminal and vicinal pairs of protons. In comparison with the 'H NMR spectra of the precursors, the most significant feature in the spectra of the metabolites is the appearance of a new resonance, corresponding to a single proton in the CHOH region a t approximately 4.2 ppm. According to the DQCOSY spectra, this multiplet is assigned to H15 by virtue of its connectivity via strong vicinal 3Jcouplings to the 14a and 16p protons. To determine the stereospecific assignment of the H E , we used one-dimensional selective cross-relaxation or NOE spectroscopy (18). As a result, the peak assigned to H15 in the deoxycorticosterone metabolite cross-relaxes with the C18 methyl protons (Fig. 1). This indicates a p orientation for H15 and implies an a-substituted OH. Also consistent with an a-hydroxylated product is the similarity of chemical shifts for the HlGa,@ and H17 protons of the metabolite and for those of the closely related compound, 15a-hydoxyprogesterone (19). Similar results and conclusions concerning 15a-hydroxylation were obtained for the corticosterone metabolite. We conclude, therefore, that the metabolites formed by the mutants P450coh are 15a-hydroxydeoxycorticosterone and 15a-hydroxycorticosterone.
P450coh (2A5) catalyzes coumarin 7-hydroxylase but shows little steroid hydroxylase activity. The substrate specificity of The abbreviations used are: DQCOSY, double quantum-filtered homonuclear shift correlation spectroscopy; NOE, nuclear Overhauser effect. the P450, however, is altered from coumarin 7-hydroxylase to testosterone 15a-hydroxylase activity by a mutation of Phe-209 to Leu (1). This mutant F209L also displays high deoxycorticosterone 15a-hydroxylase activity in addition to testosterone hydroxylase activity (Fig. 2). The Phe + Leu mutation, therefore, provides P450coh with high 11-deoxysteroid 15ahydroxylase activity. Because our previous work showed that P45015a (2A4) specifically catalyzes the l5a-hydroxylation of many *A 3-ketone steroids but not corticosterone (1, 12), we examined whether F209L catalyzes the hydroxylation of corticosterone. However, this mutant exhibits no hydroxylase activity toward corticosterone (Fig. 2).  Table I). Besides corticosterone, hydrocortisone is also hydroxylated by mutant F209N but not by either the wild-type or mutant F209L.3 Moreover, the L209N mutation in P45015a also leads to high corticosterone l5a-hydroxylase activity (20,21). The llphydroxysteroid hydroxylase activity, therefore, depends on the presence of Asn a t position 209 but not the type of the P450s. As a result, the substrate specificity of P450coh can be altered from coumarin to 11-deoxysteroids, and then to Ilp-hydroxysteroids by mutations of residue 209 from Phe to Leu, and then to Asn. In addition to the wild-type P450coh, mutants F209L and F209N, seven other mutants (F209A, F209V, F209G, F209S, F209M, F209D, and F209K) were examined for their activities to catalyze the l5a-hydroxylation of corticosterone and deoxycorticosterone (Table I) ity, although its V,,, value is at least 10-fold lower than the mutant F209N (Table I). Moreover, the K,,, of F209V for the corticosterone 15a-hydroxylase activity is maximally 10-fold higher than that of the mutant F209N (Table I). These results suggest, therefore, that a direct interaction of Asn-209 with the lip-hydroxyl of the steroid molecule plays the key role in conferring high corticosterone 15a-hydroxylase activity. Deoxycorticosterone, on the other hand, is hydroxylated by many, but not all, mutants to various degrees, although mutants F209V and F209L catalyze the hydroxylation most efficiently ( Table I).
Because of the absence of a three-dimensional structure for mammalian P450, we used the P450cam as the model for our computational and graphic analysis of the steroid-binding orientation in the substrate-heme pocket. First, the amino acid sequence of P450coh is aligned with the P450cam by two simple procedures: which map the Phe-209 to Met-191 (by the weighted homology alignment) and Met-184 (by the constrained alignment), respectively, in P450cam. Our alignments, therefore, are somewhat similar to that reported recently by Gotoh

Steroid 15a-hydroxylase activity of residue 209 mutants P450coh
Steroid 15a-hydroxylase activity was reconstituted as described under "Experimental Procedures." In A, corticosterone l5a-hydroxylase and deoxycorticosterone 15a-hydroxylase activities were measured using the steroid concentration of 100 PM. The values reported were obtained by averaging two separate results. ND means that the activity was not detectable. In B, for the K , and Vmax values of mutants P450 for corticosterone 15a-hydroxylase activity, activity was measured at various concentrations of 2,10,20,40,100, and  of P450cam by our constrained alignment. The crystal structure of P450cam (8, 23) shows a binding pocket large enough to accommodate a camphor molecules and a narrow entrance channel. It is of interest to determine whether the binding pocket of this enzyme is also large enough and flexible enough to accept a steroid molecule, assuming that channel adjustments will permit entrance. To investigate this question, we docked corticosterone into the pocket using graphics, then systematically used molecular mechanics to minimize the energy of the system. The most successful starting position occurred when corticosterone was placed approximately perpendicular to the heme plane in the substrate pocket of P450cam, so that the steroid l l O H was directed toward Met-184 and Met-191. Energy minimization verified that the P450cam pocket can indeed accommodate the steroid molecule; the nearest hydrogen bonding distance between the protein and steroid is 2.7 A (210H to the NH1 of Arg-186 in P450cam). The energy-minimized structure is shown in Fig. 3 with key di$ances given. The oxygen of the l l O H is approximately 7 A from the &carbon of Met-184 toward which it is direc!ed. Moreover, the carbon-15 of the steroid molecule is 2.9 A from the activating oxygen in the sixth axial position. The 3-carbonyl oxygen (03) of thq steroid molecule is near the side-chain oxygen of Tyr-96 (4.0 A). This phenolic oxygen (Tyr-96 in P450cam) is known to form a hydrogen bond with the carbonyl oxygen of camphor (23). The overall energy-minimized structure (1215 backbonea atoms) has an average root mean square deviation of 0.66 A to the crystal structure. We overlaid the backbone atoms of the following residues to the crystal structure by calculating the distance between the a-carbons (minimized uersus crystal structure) and found relatively small diGplacements for residues with atoms within the original $5-A sphere that defines the P450~am pocket: Phe-;7 (0.  1.40 A). The results indicate that our binding model maintains the backbone of the crystal structure very well. Moreover, the corticosterone molecule ends up aligned in the substrate pocket, so that its llOH is directed toward Phe-209 and C15 resides near the activating oxygen in the sixth axial position of the heme.
Graham-Lorence et al. (7) have also modeled androstenedione by placing the C19-methyl group over the C5 of camphor, thereby placing the steroid at a 45" angle to the heme plane. Although this model provides useful information for understanding the structural function behavior of aromatase P450, the energy minimization was sufficiently constrained so that it is not possible to infer that the P450cam pocket can accommodate androstenedione.
A sequence alignment proposed by Nelson and Strobel (24) assigns Phe-209 of P450coh in the region corresponding to helix E in P450cam. The importance of residue 209 has been viewed as a mystery (8), because this helix E does not directly interact with either the heme or the substrate in P450cam. Our alignment and modeling, however, indicate that residue 209 of P450coh may be located either at the end of helix F or in the F-G loop. The helix and loop constitute part of the substrate pocket and/or substrate-access channel in P450cam (8, 23). If residue 209 is located in the F-G loop, its mutation may result in the structural alteration of the substrate-access channel. Alternatively, the mutation may refold the loop structure toward the heme and bring residue 209 of P450coh in closer contact with the 1lOH of the steroid molecule. We favor the refolding hypothesis, because our studies with mutants P450coh suggest a direct interaction between the Asn-209 and the llOH. If residue 209 resides at the end of the F helix, on the other hand, a side-chain alteration alone would be enough to cause the direct interaction between the Asn-209 and the llOH. Further modeling will require details of the three-dimensional structures of more closely related P450s.
In conclusion, we have engineered a novel P450, which has a high corticosterone 15a-hydroxylase activity, by mutating Leu-209 to Asn of P450coh. Using the biochemical findings, we have then modeled a possible binding orientation in the substrate pocket of P450coh (2A5) based on the three-dimensional structure of P450cam. Our model proposes a steroidbinding orientation in the substrate pocket, for which the C15 position of the steroid may near the sixth axial position of the heme and the l l O H of the steroid appears to interact with residue 209 of P450coh. As a result, a structural variation of helix F and/or the F-G loop plays a key role in selecting the steroid-substrate specificity based on the group identity at the 11 position of steroid molecule.