Kinetic and molecular analysis of 5-epiaristolochene 1,3-dihydroxylase, a cytochrome P450 enzyme catalyzing successive hydroxylations of sesquiterpenes.

The final step of capsidiol biosynthesis is catalyzed by 5-epiaristolochene dihydroxylase (EAH), a cytochrome P450 enzyme that catalyzes the regio- and stereospecific insertion of two hydroxyl moieties into the bicyclic sesquiterpene 5-epiaristolochene (EA). Detailed kinetic studies using EA and the two possible monohydroxylated intermediates demonstrated the release of 1beta-hydroxy-EA ((OH)EA) at high EA concentrations and a 10-fold catalytic preference for 1beta(OH)EA versus 3alpha(OH)EA, indicative of a preferred reaction order of hydroxylation at C-1, followed by that at C-3. Sequence alignments and homology modeling identified active-site residues tested for their contribution to substrate specificity and overall enzymatic activity. Mutants EAH-S368C and EAH-S368V exhibited wild-type catalytic efficiencies for 1beta(OH)EA biosynthesis, but were devoid of the successive hydroxylation activity for capsidiol biosynthesis. In contrast to EAH-S368C, EAH-S368V catalyzed the relative equal biosynthesis of 1beta(OH)EA, 2beta(OH)EA, and 3beta(OH)EA from EA with wild-type efficiency. Moreover, EAH-S368V converted approximately 1.5% of these monohydroxylated products to their respective ketone forms. Alanine and threonine mutations at position 368 were significantly compromised in their conversion rates of EA to capsidiol and correlated with 3.6- and 5.7-fold increases in their Km values for the 1beta(OH)EA intermediate, respectively. A role for Ile486 in the successive hydroxylations of EA was also suggested by the EAH-I468A mutant, which produced significant amounts 1beta(OH)EA, but negligible amounts of capsidiol from EA. The altered product profile of the EAH-I486A mutant correlated with a 3.6-fold higher Km for EA and a 4.4-fold slower turnover rate (kcat) for 1beta(OH)EA. These kinetic and mutational studies were correlated with substrate docking predictions to suggest how Ser368 and Ile486 might contribute to active-site topology, substrate binding, and substrate presentation to the oxo-Fe-heme reaction center.

The mevalonate and methylerythritol phosphate pathways are responsible for the biosynthesis of isoprenoids, a diverse group of organic natural products found in animals, plants, fungi, insects, and bacteria. Isoprenoids are further divided into classes of primary and secondary metabolites. Isoprenoids that are primary metabolites include sterols, carotenoids, hormones, and long chain hydrocarbons used to tether particular enzymes to membrane systems, compounds essential for viability. Isoprenoids classified as secondary metabolites include monoterpenes, sesquiterpenes, diterpenes, and triterpenes, and many of these mediate interactions between organisms and their environments (1)(2)(3)(4)(5).
Capsidiol is a bicyclic dihydroxylated sesquiterpene produced by several solanaceous plants in response to pathogen or elicitor challenge (6 -10) and is considered an important plant defense response because it can prevent the germination and growth of several fungal species (11). Capsidiol biosynthesis is regulated by the expression of two key enzymes (see Scheme 1). 5-Epiaristolochene synthase catalyzes the cyclization of farnesyl diphosphate to the bicyclic intermediate 5-epiaristolochene (EA) 1,2 and is responsible for the diversion of farnesyl diphosphate from the mevalonate pathway toward antimicrobial compound biosynthesis (12,13). Recently, two highly homologous cDNAs for EA dihydroxylase (EAH) were isolated from Nicotiana tabacum and functionally expressed in yeast, and the encoded proteins were characterized as cytochrome P450 enzymes catalyzing the stereo-and regiospecific hydroxylation of EA at C-1 and C-3 (see Scheme 1) (14).
A preferred order for the successive hydroxylation of EA has yet to be determined (see Scheme 1). Of the two possible monohydroxylated intermediates of EA, only 3␣-hydroxy-EA (3␣(OH)EA) has been reported as a minor metabolite accumulating in pepper fruits challenged with fungal spores (34) Ralston et al. (14) documented the ready conversion of 3␣(OH)EA to capsidiol by microsomes from yeast overexpressing the EAH gene, they also reported the biosynthesis of capsidiol without the accumulation of either monohydroxylated intermediate, 3␣(OH)EA or 1␤(OH)EA, in reactions incubated with low concentrations (Ͻ1 M) of the EA substrate. Although 1␤(OH)EA has not been reported as a reaction product of EAH or observed in plants exposed to exogenous stimuli, preference for P450-mediated hydroxylations of mono-and diterpenes occurring allylic to carbon-carbon double bonds (16,20,21) suggests that the initial hydroxylation catalyzed by EAH could occur at C-1, followed by that at C-3.
This work describes investigations of the successive regioand stereospecific hydroxylation of EA by EAH. The ready availability of EA, 3␣(OH)EA, and 1␤(OH)EA greatly facilitated our ability to perform detailed kinetic studies to address the preferred order of hydroxylation. Sequence comparisons with related P450 enzymes and modeling of EAH based upon the mammalian 2C5 steroid hydroxylase structure (36) enabled us to identify amino acid residues potentially providing catalytic specificity to these reactions. These structural hypotheses were then tested by site-directed mutagenesis and kinetic analysis. Together, the results demonstrate a preferred reaction order and provide evidence that Ser 368 and Ile 486 play a significant role in orchestrating the successive hydroxylation of EA to capsidiol.
Yeast Expression and Microsome Preparations-Expression vectors harboring wild-type or mutant EAH were introduced into the WAT11 yeast strain, a strain previously engineered with an Arabidopsis NADPH-cytochrome P450 reductase gene (40) and kindly provided by Dr. P. Urban (Centre de Génétique Moléculaire, CNRS, Gif-sur-Yvette, France). Transformants were grown from single colonies; expression of the EAH cDNAs was induced by galactose addition to the growth medium; and microsomes were prepared as described previously (41).
CO Difference Spectra-CO difference spectra were determined to estimate the amount of properly folded wild-type and mutant EAH proteins in microsome preparations. The amount of functional P450 was calculated using an extinction coefficient of 91 mM Ϫ1 cm Ϫ1 at 450 nm (42). The wild-type and mutant enzymes showed a clear absorption maximum at 450 nm (Supplemental Fig. 1), consistent with properly folded enzymes having the heme cofactor oriented in the correct electron spin state.
EAH Assays-Standard assays were performed in 1-ml glass vials with a final reaction volume of 200 l containing 4% Me 2 SO, 2.4 mM NADPH, 100 mM Tris-HCl (pH 7.5), and microsomes at a final concentration of 90 or 180 pmol of EAH protein eq (determined by CO difference spectroscopy)/ml. Substrate concentrations were varied from 2 to 100 M for EA, 3 to 30 M for 3␣(OH)EA, and 0.5 to 30 M for 1␤(OH)EA. After preincubation of the reaction mixtures at 30°C for 5 min, the reactions were initiated by the addition of 2.4 mM NADPH and allowed to proceed for 2 min (EA), 1 min (1␤(OH)EA), or 5 min (3␣(OH)EA), which allowed for 5-20% conversion of substrate to reaction product(s). The reactions were terminated by the rapid addition of 400 l of ethyl acetate and mixed by vortexing, followed by removal of the ethyl acetate extract. A second ethyl acetate extract was combined with the first, and the combined organic extracts were carefully concentrated on ice under N 2 gas, redissolved in 20 l of ethyl acetate containing 10 ng/l valencene as an internal standard, and subjected to gas chromatography (GC) and GC/mass spectrometry (MS) analyses. Assays were performed in triplicate for each substrate concentration, and kinetic constants were calculated by a nonlinear regression fit to the Michaelis-Menten equation using EnzymeKinetics Version 1.5 software (Trinity Software). Kinetic constants calculated from Lineweaver-Burk, Eadie-Hofstee, and Hanes-Woolf plots were not significantly different from those derived from nonlinear regression. Optimized reactions were linear with respect to reaction product generation over time and corresponded to a 5-20% conversion of substrate to product, sufficient to generate 1 ng or more of reaction product(s) readily quantified and identifiable by GC/MS.
Reaction Product Analyses-Quantification of the reaction products as well as standards of EA, 1␤(OH)EA, 3␣(OH)EA, and capsidiol was routinely performed using (ϩ)-valencene (Fluka, Buchs, Switzerland) as an internal standard. Accuracy of the valencene standard was periodically verified relative to a (Ϫ)-␣-cedrene standard (Fluka). Reaction products (1-l aliquots) were quantified with an HP-5890 gas chromatograph equipped with an HP-5 capillary column (30 m ϫ 0.25 mm, 0.25-m phase thickness) and a flame ionization detector as described previously (38). Splitless injections were performed at an injection port temperature of 250°C with an initial column temperature of 140°C maintained for 0.5 min. The column temperature was then increased to 230°C with a 4°C/min gradient. In addition to comigration and retention time comparisons with authentic standards, reaction products were identified by MS using a Thermo Finnigan DSQ GC/MS system equipped with a Restec Rtx-5 capillary column (30 m ϫ 0.32 mm, 0.25-m phase thickness). Samples were injected in the splitless mode at 250°C with an initial oven temperature of 70°C for 1 min, followed by an 8°C/min gradient to 230°C. Mass spectra were recorded at 70 eV, scanning from 35 to 300 atomic mass units, and compared with authentic standards for verification.
Purification of the EAH-S368V Reaction Products-To isolate and identify the EAH-S368V reaction products, ethyl acetate extracts from the in vitro assays were carefully concentrated under an N 2 stream and resuspended with hexane such that the final ethyl acetate concentration was Ͻ2%. The sample was then applied to a 230 -400 mesh silica column (0.5 ϫ 1 cm), and the hydrocarbon fraction was collected with a 2-ml hexane wash. Hydroxylated products were selectively eluted with 2 ml of a 1:19 (v/v) ethyl acetate/hexane solution, followed by 2 ml of a 1:4 (v/v) ethyl acetate/hexane solution. The fractions were carefully reconcentrated under an N 2 stream before GC/MS analysis.
Computational Studies-Homology modeling was performed with Modeler Version 6.2 (43-45) by threading the EAH amino acid sequence (GenBank TM /EBI accession number AF368376) onto the structural coordinates for the mammalian P450 2C5 protein (Protein Data Bank code 1DT6) (36). Ligands EA, 1␤(OH)EA, and 3␣(OH)EA were created using Chemdraw Ultra Version 7.0.1, energy-minimized with MOPAC in Chem3D Ultra Version 7.0 (CambridgeSoft Corp., Cambridge, MA), and subsequently used in docking simulation experiments. Docking calculations were conducted with GOLD Version 1.2 software (Genetic Optimization for Ligand Docking, Cambridge Crystallographic Data Centre, Cambridge, United Kingdom) (46 -49).

RESULTS
Optimization of the EAH Activity-Numerous preliminary experiments were performed to optimize EAH activity for the conversion of EA to capsidiol. For example, the ability of a miscible organic co-solvent and various detergents to improve EA solubility was assessed. Maximum conversion of EA to capsidiol activity was observed at final concentrations of 2-5% (v/v) Me 2 SO, but was inhibited at concentrations Ͼ10%. Detergents were also evaluated at concentrations ranging from 0.01 to 0.1%. Triton X-100 (v/v), deoxycholate (w/v), and 1-O-octyl ␤-D-glucopyranoside (w/v) all inhibited EAH activity, whereas CHAPS (w/v) and Tween 20 (v/v) had no effect. Standard assays were therefore performed using EA stock solutions dissolved in Me 2 SO and assayed at a final Me 2 SO concentration of 4% (v/v) in all reaction assays. The pH optimum was determined using 100 mM Tris-HCl (pH 7.2-9.2) and phosphate (pH 6.0 -7.5) buffering systems. Maximum activity was observed at pH 7.5 with little preference for either the Tris-HCl or phosphate buffer. The optimum NADPH concentration for EAH activity was also determined. The maximum conversion rate   Kinetic Analysis of EAH Indicates a Specific Reaction Order-To initially address the hydroxylation specificity of EAH (Scheme 1), we tested the enzyme for substrate preference for either of the putative reaction intermediates, 3␣(OH)EA and 1␤(OH)EA (Fig. 1). Microsomes from WAT11 yeast overexpressing the EAH gene were used as the enzyme source, and the absolute amount of active EAH enzyme was calculated from the carbon monoxide difference spectra. Both 3␣(OH)EA and 1␤(OH)EA were converted to capsidiol by EAH in NADPHdependent reactions, but with different efficiencies. The K m for 1␤(OH)EA was ϳ4 times lower and the k cat was ϳ3 times higher than those for 3␣(OH)EA, resulting in a catalytic efficiency (k cat /K m ) for 1␤(OH)EA ϳ10 times greater than that for 3␣(OH)EA (Table I). The kinetic constants for the hydrocarbon FIG. 3. GC and MS profiles for the reaction products generated by EAH incubated with EA. Shown are total ion chromatograms for reaction products generated by EAH in the presence and absence of NADPH (A) and mass spectrum matches for peaks 1 and 2 to authentic standards of 1␤(OH)EA and capsidiol (B).

5-Epiaristolochene 1,3-Dihydroxylase
substrate EA were also determined ( Fig. 2 and Table I). Although the K m for the hydrocarbon substrate was considerably higher than that for either of the monohydroxylated intermediates, the turnover rates for EA and 1␤(OH)EA were similar and significantly greater than that for 3␣(OH)EA.
The kinetic constants reported in Table I are consistent with a catalytic barrier to the successive hydroxylation of the 3␣(OH)EA intermediate and suggested that 1␤(OH)EA would be the preferred monohydroxylated intermediate. In an effort to capture the release of an initial reaction intermediate, assays were performed with increasing concentrations of EA (2-100 M), and the reaction product profiles were determined (Fig. 2). Reaction products were observed only after incubations with NADPH and were identified by GC/MS comparisons with authentic standards (Fig. 3). Capsidiol production reached a maximum at ϳ40 M EA and gradually decreased after that. Although no monohydroxylated intermediates were detectable at EA concentrations below 16 M, a proportionate increase in the biosynthesis of 1␤(OH)EA was observed at EA concentrations above 20 M. In contrast, no release of 3␣(OH)EA was observed over the entire concentration range investigated.
Molecular Analysis of the Dihydroxylase Activities-A combination of computational methods was used to identify structural elements possibly contributing to the successive hydroxylation activity of EAH (Fig. 4). Sequence alignment of EAH with the known substrate recognition sites (SRSs) of closely related P450 enzymes of the CYP71D subfamily (50) revealed several potentially important sites. SRS-5 and SRS-6 were of particular interest given their proximity to the active site in several P450 structures (51) and because specific amino acid positions within these regions have previously been correlated with regio-and stereospecific reaction mechanisms (Fig. 4A) (52)(53)(54)(55)(56)(57)(58)(59)(60)(61). A homology model of EAH derived from comparison with the mammalian CYP2C5 structure was constructed (Fig.  4B), and the resultant model suggested that Ser 368 and Ile 486 project into the active-site cavity oriented toward the heme catalytic site (Fig. 4C).
To investigate the contributions, if any, of Ser 368 and Ile 486 to either the first or second hydroxylation reaction as predicted by computational studies, site-specific mutations were generated at each position. The mutant genes were expressed in WAT11 yeast lines, and the absolute amount of newly synthesized microsomal enzyme was calculated from the carbon monoxide difference spectra (Supplemental Fig. 1). Mutant enzymes were then examined for substrate preferences and product specificity by steady-state kinetic analyses (Fig. 5 and Table II). Mutant enzymes harboring smaller amino acid substitutions such as Ala, Thr, Cys, and Val demonstrated reasonable CO binding spectra and were catalytically active. EAH-S368A and EAH-S368T retained their ability to fully convert EA to capsidiol, but  (15), CYP71D16 for a diterpene hydroxylase (72), and CYP71D20 for EAH (14). A homology model of EAH was generated using the mammalian P450 2C5 protein (Protein Data Bank code 1DT6) (36) as described under "Experimental Procedures" and is shown here as a ribbon diagram (B). A close-up is shown of the active-site model of EAH showing the spatial location of Ser 368 and Ile 486 relative to the heme reaction center (C).

5-Epiaristolochene 1,3-Dihydroxylase
the turnover rates for capsidiol formation were 3-13 times lower compared with that of the wild-type enzyme. EAH-S368A and EAH-S368T were also able to convert 1␤(OH)EA to capsidiol with k cat values comparable with that of wild-type EAH, but with K m values for 1␤(OH)EA 4 -6 times greater compared with that of wild-type EAH. Although mutant enzymes containing bulky amino acid substitutions such as Ile and Phe expressed well in yeast and were capable of binding CO, indicative of proper folding, these mutant enzymes were devoid of any hydroxylase activity for EA or 1␤(OH)EA.
EAH-S368A and EAH-S368T possess a K m for the EA substrate that is similar to that of the wild-type enzyme, yet have a significantly higher K m for 1␤(OH)EA, suggesting that Ser 368 may be more instrumental for the second hydroxylation step rather than the first. Additional mutations at Ser 368 provided support for this notion. Both EAH-S368C and EAH-S368V readily catalyzed the biosynthesis of 1␤(OH)EA without the generation of any capsidiol when incubated over a wide range of EA concentrations (Fig. 5). Interestingly, the EAH-S368C mutant produced 1␤(OH)EA with a catalytic efficiency ϳ30% that of the wild-type enzyme and only a very negligible amount of EA-1-one (Table II). In contrast, the EAH-S368V mutant produced significant and approximately equal amounts of 1␤(OH)EA, 2␤(OH)EA, and 3␤(OH)EA; a small amount of 3␣(OH)EA; another unidentified monohydroxylated compound accounting for 7% of the total products (presumed to be 2␣(OH)EA based on retention time and MS); and EA-3-one and EA-1-one at 1.5% (Fig. 6 and Scheme 2). Control microsomes (microsomes from yeast harboring an expression vector not engineered with any EAH gene constructs) were not able to metabolize 1␤(OH)EA, but did exhibit a very minor activity (4.6% of that associated with EAH-S368V) for conversion of 3␣(OH)EA to the corresponding ketone (data not shown). Overall, the catalytic efficiency of EAH-S368V for the turnover of EA to 1␤(OH)EA was slightly superior to the ability of the wild-type enzyme to convert EA to capsidiol (Table II). However, when the biosynthetic rates for all the monohydroxylated and ketone reaction products are taken into account, EAH-S368V exhibited a turnover rate for EA twice that of the wildtype enzyme.
Ile 486 within SRS-6 was also chosen for mutagenesis because this position exhibits significant variability between other members of the CYP71D subfamily, and the spatial location of its R group is oriented into the active site based on molecular modeling (Fig. 4). Mutation of Ile 486 to Ala in EAH also caused a significant reduction in the successive hydroxylation of EA to capsidiol ( Fig. 7 and Table II). The EAH-I486A mutant enzyme turned over EA to the monohydroxylated intermediate 1␤(OH)EA at approximately one-half the rate of the wild-type enzyme, but produced only negligible amounts of capsidiol from EA. Because an accurate determination of the EA-to-capsidiol conversion was not possible, no kinetic constants for this reaction were calculated. However, the turnover rate for 1␤(OH)EA to capsidiol by EAH-I486A had a 12-fold higher K m and a 4-fold lower turnover rate compared to that of the wild-type enzyme (Table II). DISCUSSION The aim of this work was to determine whether successive hydroxylations of EA to capsidiol catalyzed by EAH occur with a preferred reaction order, and if so, if there are structural features of the EAH active site that can be associated with this specificity. Earlier work had demonstrated that EA is readily converted to capsidiol by EAH without the apparent release of any reaction intermediate(s) (14,41). This in turn prompted speculation that the successive hydroxylations might occur sequentially within a single catalytic cascade terminating with the release of only the dihydroxylated capsidiol product. The relatively large cavities described for the active site of several P450 enzymes (62,63) lent support to this notion. We therefore suggested that the initial monohydroxylated intermediate might not be released from EAH, but instead might be flipped and/or rotated within the active site in a manner to reposition the intermediate for the second hydroxylation event. The second hydroxylation event would then provide a trigger for product release. The present results clearly refute such speculation and provide strong kinetic evidence for independent and stepwise hydroxylation of EA. EAH readily accepts and catalyzes the conversion of both monohydroxylated intermediates to capsidiol. However, although the turnover rate for 3␣(OH)EA to capsidiol is comparable with that for the native hydrocarbon substrate EA, the relative efficiency for 1␤(OH)EA conversion is 10-fold greater ( Table I). The latter is largely accounted for by a much lower K m for 1␤(OH)EA. The accumulation of 1␤(OH)EA in reactions initiated with concentrations of EA exceeding 20 M also clearly demonstrates a specific reaction pathway. When EA concentrations are below 20 M, EA must be converted first to 1␤(OH)EA, which is released but rapidly recaptured and converted to the dihydroxylated product. At concentrations exceeding 20 M (the approximate K m for EA), FIG. 5. Substrate-dependent activities of the EAH Ser 368 mutants. Site-specific mutations corresponding to Ser 368 were introduced into the EAH gene as described under "Experimental Procedures"; the mutant genes were expressed in yeast; and yeast microsomes were prepared as the source for each mutant P450 enzyme. CO difference spectroscopy was used to qualify and normalize the amount of properly folded P450 enzyme used in the activity assays. Microsome preparations were incubated at the indicated concentrations of EA or 1␤(OH)EA before profiling the reaction products by GC/MS. Except for the EAH-S368V mutant, which generated multiple reaction products, the only NADPH-dependent reaction products observed were capsidiol (‚) and 1␤(OH)EA (E).
The successive hydroxylation activity of EAH exhibits similarities and differences relative to other well characterized multifunctional P450 enzymes. For example, CYP27A1 catalyzes the successive hydroxylation of vitamin D 3 (64) and cho-lesterol (65). Based on the time-dependent appearance of multiple reaction products during in vitro reactions, Sawada et al. (64) concluded that CYP27A1 initially catalyzes hydroxylation at C-25 of vitamin D 3 , followed by a second hydroxylation at C-24, C-1, or C-26, yielding a complex mixture of a single monohydroxylated product and three or more dihydroxylated ones. No C-1 or C-24 monohydroxylation products were ob-  6. Reaction product analysis of EAH-S368V incubated with EA. EAH-S368V was incubated with 40 M EA, and the total reaction products were fractionated by silica gel chromatography. Total ion chromatograms of the reaction products eluted from the silica gel with 5% ethyl acetate in hexane (A, left panel) are compared with those of authentic standards for EA-3-one and EA-1-one (A, right panel). Mass spectra for peaks 1 and 2 are compared directly with those for the respective ketones (B). Total ion chromatograms of reaction products sequentially eluted from the silica gel with 20% ethyl acetate in hexane (C, left panel) are compared with those of authentic standards for 1␤(OH)EA, 3␣(OH)EA, and 3␤(OH)EA (C, right panel). The mass spectrum for peak 3 matches that for 1␤(OH)EA (D), shown previously in Fig. 3. The mass spectra for peaks 5 and 6 are compared directly with those for 3␤(OH)EA and 3␣(OH)EA (D). However, peak 5 resolves into two nearly equal sized peaks by chiral GC analysis (Supplemental Fig. 2), with peak 5-1 corresponding to 3␤(OH)EA and peak 5-2 to 2␤(OH)EA (as determined by NMR analysis). RT, retention time.

5-Epiaristolochene 1,3-Dihydroxylase
served, nor products with more than two inserted hydroxyls. Consistent with this complex reaction product profile, the K m for the C-25 monohydroxylated intermediate is equal to that for the original substrate, but the apparent reaction velocity for the successive hydroxylation is only one-tenth that of the initial hydroxylation event. P450 11␤ (66) and P450 17␣ lyase (67) are likewise P450 enzymes catalyzing successive hydroxylation reactions in aldosterone and androgen metabolism, respectively, that generate mono-and dihydroxylated reaction mixtures. Using rapid quench methods, rate constants for the initial hydroxylation step were shown to be ϳ10-fold greater than the rate constants for the second hydroxylation step, and the dissociation rates of the monohydroxylated intermediates were also competitive with the rate constants for the second hydroxylation step, altogether consistent with the release and accumulation of monohydroxylated intermediates over time. In light of these reports and the current results, our initial observation of complete conversion of EA to a single dihydroxylated product without release of monohydroxylated intermediate (14) can now be attributed to the greater affinity and faster turnover rate of EAH for the monohydroxylated intermediate than for the initial EA substrate.
Previous site-directed mutagenesis studies and the threedimensional structure of mammalian P450 2C5 (36) provided important frameworks for targeting active-site residues that might play a role in the successive hydroxylation specificity of EAH (68,69). Using only those sequence regions that map to the active-site cavity of 2C5 and correspond to SRS-5 and SRS-6 (50), sequence alignment of select CYP71D family members revealed that the positions corresponding to amino acids 368 and 486 in EAH are particularly variable (Fig. 4). In fact, previous studies had already documented the functional significance of this variability. Mutagenesis of the position equiva-lent to amino acid 368 in several members of the CYP2B subfamily was shown to affect the regiospecific hydroxylation of steroids (52,53). The same position was also reported to play a role in substrate recognition for CYP2A5 (54) and CYP3A4 (55). Swapping this region between two highly homologous plant P450 enzymes, limonene 6-hydroxylase and limonene 3-hydroxylase, followed by reciprocal site-directed mutagenesis, demonstrated that the corresponding position also contributes to the regiospecificity of monoterpene hydroxylation. Exchange of Phe 363 with Ile changes the regiospecificity of limonene 6-hydroxylase (CYP71D18) from hydroxylating limonene at C-6 to C-3, but the reciprocal mutation, I363F, completely inactivates limonene 3-hydroxylase activity (56). A similar modification of the corresponding site (position 371) in cinnamate hydroxylase (CYP73A1), another plant-specific P450 family, from Ile to Phe leads to a dramatic decrease in substrate binding and enzymatic activity, but without a major perturbation of protein folding (57).
The position corresponding to amino acid 486 in SRS-6 of EAH has also been extensively evaluated in mammalian hydroxylases. For instance, exchange of Gly 478 of CYP2B1 with larger hydrophobic amino acids (Ala, Val, Ile, and Leu) alters the stereospecific hydroxylation of androstenedione (58,59). In contrast, substitution of Phe for Ile at position 479 of CYP3A4 alters the regiospecific hydroxylation of 7-hexylcoumarin (60). Similarly, substitution of smaller hydrophobic amino acids (Ala, Val, and Leu) for Phe 494 in CYP94A2 also alters the regiospecificity of lauric acid hydroxylation (61). Finally, docking putative substrates into molecular models of four phenyl propanoid metabolism P450 enzymes from Arabidopsis thaliana (CYP84A1, CYP75B1, CYP98A3, and CYP73A5) was used to map substrate contact points to positions equivalent to SCHEME 2. Chemical rationalization for the reaction products generated by the EAH-S368V mutant. Solid arrows refer to reaction products identified by comparison with authentic standards or directly by NMR, whereas the dashed arrow refers to a reaction product structure inferred from GC retention times and MS profiles. The percentages denote the amount of a particular reaction product generated in a typical reaction assay. unk, unknown FIG. 7. Substrate-dependent activities of the EAH-I486A mutant. The EAH-I486A mutation was created as described under "Experimental Procedures"; the mutant gene was expressed in yeast; and yeast microsomes were prepared as the source of the EAH-I486A enzyme. CO difference spectroscopy was used to qualify and normalize the amount of properly folded P450 enzyme used in the activity assays. Assays were incubated at the indicated concentrations of EA for 5 min (upper panel) or 1␤(OH)EA for 1 min (lower panel) before profiling the reaction products by GC/MS. The only NADPH-dependent reaction products observed were capsidiol (‚) and 1␤(OH)EA (E). Capsidiol levels were near the detection limits for reactions incubated with EA.
This work, summarized in Table III, demonstrates that positions 368 and 486 in EAH are both critical for regulating successive hydroxylation reactions as well as dictating regioand stereospecificity. Substitution of Cys for Ser at position 368 completely abolished the capacity of EAH to generate capsidiol from either EA or 1␤(OH)EA, yet the initial hydroxylation reaction of EA to 1␤(OH)EA was preserved. Position 486 likewise is critical for the successive hydroxylation reaction. Although the EAH-I486A mutant readily catalyzed the biosynthesis of 1␤(OH)EA from EA, it was severely compromised for capsidiol biosynthesis. The latter limitation was in fact correlated with a significant reduction in the binding affinity for the monohydroxylated intermediate.
The diversity of hydroxylated products generated by the EAH-S368V mutant underscores its importance in controlling stereo-and regiospecific hydroxylation. Although, like EAH-I486A, this mutant is severely compromised in its ability to convert 1␤(OH)EA to capsidiol, it has a total activity 2-fold greater than that of the wild-type enzyme and generates at least five novel hydroxylation products (Fig. 6). These new products represent unique stereospecific products (3␤(OH)EA), regiospecific isomers (2␣(OH)EA and 2␤(OH)EA), and novel successive hydroxylation products (EA-1-one and EA-3-one) that must arise from multiple hydroxylation reactions at C-1 and C-3, respectively (Scheme 2).
Substrate docking experiments using the EAH homology model provide a basis to rationalize the catalytic specificities of EAH and the various Ser 368 and Ile 486 mutations (Fig. 8).
Multiple docking solutions for EA as ligand cluster around two distinct binding modes in which the plane of the bicyclic ring is rotated by ϳ180°. In binding mode A (Fig. 8A), the ␤-face of EA is directed toward the heme center, placing C-1 in position for hydroxylation to produce 1␤(OH)EA. In binding mode B (Fig.  8B), C-3 is now poised for hydroxylation on the ␣-face leading to 3␣(OH)EA. It is apparent from these binding models that the ␣-face is more hindered, given the methyl group on C-4. Therefore, one might expect that hydroxylation on the ␤-face may proceed faster. Docking solutions using 1␤(OH)EA as ligand cluster around a single binding mode, placing the ␤-hydroxy in the vicinity of Ser 368 for a potential hydrogen bonding interaction (Fig. 8C). Docking solutions for the 3␣(OH)EA ligand produced a single nonproductive binding mode that directs the 3␣-hydroxy group toward the heme center and C-1 away (data not shown). A compelling explanation for the greater catalytic efficiency of EAH for 1␤(OH)EA than for 3␣(OH)EA is that the hydroxyl group of Ser 368 anchors the 1␤(OH)EA intermediate via a hydrogen bonding network (70) in a proper orientation for the second hydroxylation event. The EAH-S368A mutant may retain successive hydroxylation activity because it accommodates an additional water molecule in the active-site pocket along with the monohydroxylated intermediate, and this water molecule complements the missing hydroxyl function of the Ser or Thr side chain. Confirmation of a hydrogen bonding network or the presence of an additional water molecule in the active site of EAH-S368A will be difficult without direct structural information on the EAH enzyme complexed with substrate analogs (62,63,70,71).
In summary, this work has established a preferred, kinetically discriminated reaction order for the successive hydroxylation of EA by EAH and demonstrated that successive hydroxylation shares important parallels with current models for how regioand stereospecificity arise in cytochrome P450 enzymes. Sequence alignments of P450 enzymes, especially those representing the SRS regions (50), active-site residues (69), and those alignments between highly homologous but functionally distinct enzymes (56), have proven extremely useful in identifying structural features and amino acids contributing to such specificities of these enzymes. Elucidation of such specificities in the successive hydroxylation reactions of P450 enzymes such as EAH should also benefit from this comparative type of approach.
FIG. 8. Docking of EA and 1␤(OH)EA into a modeled EAH active site. A homology model of EAH was created using the mammalian P450 2C5 (Protein Data Bank code 1DT6) as template and was used for docking experiments with EA and 1␤(OH)EA ligands as described under "Experimental Procedures." Two distinct binding modes of EA that position the ␤-face of EA toward the heme center for oxidation at C-1 (A) or that place the ␣-face toward the heme center for C-3 oxidation (B) were predicted from docking experiments. A single binding mode for 1␤(OH)EA (C) that poises C-3 near the heme center for oxidation was predicted. The ␤-hydroxy is directed toward the hydroxyl group of Ser 368 for possible hydrogen bonding interactions. Relevant carbon atoms are numbered, and important active-site residues are displayed and labeled.