Expansion of Substrate Specificity of Cytochrome P450 2A6 by Random and Site-directed Mutagenesis*

The natural product indole is a substrate for cytochrome P450 2A6. Mutagenesis of P450 2A6 was done to expand its capability in the oxidization of bulky substituted indole compounds, which are not substrates for the wild-type enzyme or the double mutant L240C/N297Q, as determined in our previous work (Wu, Z.-L., Aryal, P., Lozach, O., Meijer, L., and Guengerich, F. P. (2005) Chem. Biodivers. 2, 51-65). Error-prone PCR and site-directed mutagenesis led to the identification of two critical amino acid residue changes (N297Q and I300V) that achieve the purpose. The new mutant (N297Q/I300V) was able to oxidize both 4- and 5-benzyloxy(OBzl)indoles to form colored products. Both changes were required for oxidation of these bulky substrates. The colored product derived from 5-OBzl-indole was mainly 5,5′-di-OBzl-indirubin, whereas the dominant blue dye isolated upon incubations with 4-OBzl-indole was neither an indigo nor an indirubin. Two-dimensional NMR experiments led to assignment of the structure as 4-OBzl-2-(4′-OBzl-1′,7′-dihydro-7′-oxo-6′H-indol-6′-ylidene)indolin-3-one, in which a pyrrole ring and a benzene ring are connected with a double bond instead of the pyrrole-pyrrole connection of other indigoids. Monomeric oxidation products were also isolated and characterized; three phenols (4-OBzl-1H-indol-5-ol, 4-OBzl-1H-indol-6-ol, and 4-OBzl-1H-indol-7-ol) and one quinone (4-OBzl-1H-indole-6,7-dione, the postulated immediate precursor of the final blue dye) were identified. The results are interpreted in the context of a crystal structure of a P450 2A6-coumarin complex. The I300V change opens an additional pocket to accommodate the OBzl bulk. The N2297Q change is postulated to generate a hydrogen bond between Gln and the substrate oxygen. Thus, the substrate specificity of P450 2A6 was expanded, and new products were obtained in this study.

Microsomal P450 2 enzymes (also termed "heme thiolate P450" (1)) are well known for their remarkable capabilities in the catalysis of diverse oxygenation reactions (2)(3)(4)(5)(6). These enzymes have been studied mostly as the principal catalysts involved in sterol synthesis, drug metabolism, and xenobiotic disposition. The use of P450 enzymes in the area of biocatalysis and fine chemical production is still largely unexploited (7). Random mutagenesis is one of the main approaches in terms of developing P450 enzymes with new functions and can also be used to enhance the knowledge of structure-function relationships of these enzymes and thus provide more information for drug development.
Random mutagenesis and molecular breeding approaches can be very useful when there is no crystal structure available to perform rational design, which is still the case for most P450 enzymes today. Furthermore, even rational designs based on crystal structures do have limitations because the enzymes possess both rigidity and flexibility. Predictions are even more complex in the case of oxidoreductase-catalyzed reactions when several factors are involved in the catalytic cycle. In addition to the active site (where substrate is bound) that is obvious in the crystal structure, other residues in the protein may also play important roles. With the availability of various high throughput screening methods and robotic systems, random mutagenesis has become a widely used tool in protein engineering both to improve catalytic efficiency and to investigate structure-function relationships (8). For P450 enzymes, random mutagenesis and molecular breeding are emerging areas. Bacterial P450 102A1 has been used as a starting point to develop catalysts that can hydroxylate alkanes in a regio-and enantioselective manner (9) or that have special features (10). This laboratory has developed P450 1A2 using several strategies to improve the catalytic efficiency toward several substrates (11)(12)(13).
Molecular breeding of P450 2A6 was done in this laboratory (14), and a double mutant (L240C/N297Q) was selected from libraries constructed by random mutagenesis using randomized primers, each covering four positions of one of the substrate recognition site regions (2), and a staggered extension process method (15). The screening was based on the metabolism of indole, a substrate of P450 2A6 (16,17) that can be hydroxylated to a product (indoxyl) that dimerizes to the blue compound indigo. The activity of the selected mutant was enhanced toward indole as well as some substituted indoles. This mutant has been used successfully to produce various substituted indigoids with enhanced activities as potent kinase inhibitors (18,19).
In the course of previous work on the biotransformation of 45 indole compounds catalyzed by the L240C/N297Q mutant, we found that this mutant showed limitations in using some substrates, among which were those bearing bulky groups (19). In the x-ray crystal structure of the P450 2A6-coumarin complex recently determined by Johnson and coworkers (20), the active site of P450 2A6 contains large aromatic residues that reduce the volume of the substrate-binding site. The active site is ϳ6-fold smaller than that of P450 2C8 (Protein Data Bank code 1PQ2) (21), consistent with the size of typical substrates such as coumarin (22), nicotine (23,24), and indole.
We are interested in expanding the substrate specificity of P450 2A6  using mutagenesis methods to augment the active pocket so that it can accommodate indoles with bulky substitution groups to increase the diversity of libraries of products generated for use as chemicals for screening of biological activity (19). In this work, error-prone PCR was used to construct randomized libraries, and a colorimetric colony-based method was used in the screening of mutants. Two critical amino acid residue changes that allow oxidization of both 4-and 5-benzyloxy(OBzl) indoles to form colored products were identified. The results are interpreted in the context of the x-ray crystal structure (20).

EXPERIMENTAL PROCEDURES
Chemicals-Indole, 4-chloroindole, and 4-and 5-OBzl-indoles were purchased from Aldrich or Lancaster Synthesis (Windham, NH). All other reagents and solvents were obtained from general commercial suppliers and used without further purification.
Spectroscopy-UV-visible spectra were recorded either with a Hewlett-Packard 1040A diode array detector (on-line with HPLC) or in Me 2 SO or H 2 O using a Cary 14/OLIS or Aminco DW2a/OLIS spectrophotometer for quantitative studies. Mass spectra were obtained in the Vanderbilt facility either on-line with HPLC or via direct infusion using a Finnigan TSQ 7000 triple quadrupole mass spectrometer (Thermo Electron Corp., San Jose, CA) equipped with a standard API-1 atmospheric pressure chemical ionization (APCI) source in the positive or negative ion mode. N 2 was used as a sheath gas (50 p.s.i.). The vaporizer temperature was set to 500°C, and the corona current was maintained at 5 A. The capillary was set at 220°C and 25 V (or Ϫ25 V in the negative mode). The tube lens voltage was set to 80 V (or to Ϫ96 V in the negative mode). The collision-induced dissociation was set to Ϫ30 V in tandem mass spectrometry experiments. Data acquisition and spectral analysis were conducted with Finnigan ICIS software on a Digital Equipment Alpha workstation.
All NMR experiments were carried out in the Vanderbilt facility at 298 K with Bruker Avance TM DRX 500-and 400-MHz instruments. Samples were dissolved in d 6 -Me 2 SO or CD 3 CN, and semiquantitative estimates of the amounts of indigoids were made by adding known amounts of CH 3 CO 2 H and referencing to this internal -CH 3 signal (␦ 1.91). Detailed structural information for four was obtained from four separate two-dimensional NMR experiments. Through-space 1 H-1 H proximity was observed using nuclear Overhauser effect correlation spectroscopy (NOESY) with a mixing time of 500 ms. Single C-H bond connectivity was obtained using heteronuclear multiple quantum correlation methods. Long-range C-H bond connectivity (two to three bonds) was obtained using heteronuclear multiple bond correlation (HMBC) with a 55-ms delay and 1 H-13 C CIGAR-HMBC with a delay for evolution of long-range couplings being modulated between 6 and 12 Hz and a low pass J-filter to suppress one-bond correlations. The NOESY and COSY spectra of compounds 6, 8, and 10 were recorded (mixing time of 600 ms for NOESY) to determine the hydroxylation positions. All NMR data were processed using XWIN-NMR TM software on an Octane workstation (Silicon Graphics, Mountain View, CA).
Expression Systems-The system involved expression of human P450 2A6 in Escherichia coli (25). The P450 2A6 double mutant L240C/N297Q used to construct the library had been selected from libraries generated in selected regions of the P450 2A6 sequence on the basis of their ability to produce colored indigoids (14). All of the wild-type (WT) and mutant P450 2A6 enzymes and human NADPH-cytochrome P450 reductase were expressed using a bicistronic pCWЈ vector (pCW2A6bc) (26) and a tryptophanase-negative strain (trnA Ϫ ) of E. coli as the host bacteria (19).
Construction of P450 2A6 Libraries-The open reading frame region of the P450 2A6 gene (1.5 kb) was mutated using a low fidelity PCR method and amplified in a 50-l PCR mixture containing 250 ng of pCW2A6bc-L240C/N297Q or pCW2A6bc-WT plasmid, 120 ng of forward primer 5Ј-TAGGAGGTCATATGGCTGCT-3Ј and reverse primer 5Ј-ATTTCTAGACCGGAAGGCTT-3Ј, 2.5 units of Mutazyme DNA polymerase (Stratagene, La Jolla, CA), 0.2 mM each dNTP, and 10ϫ Mutazyme reaction buffer (Stratagene). The 1.5-kb amplified PCR library fragment was purified by agarose gel electrophoresis and cloned into the pCW2A6bc vector using the NdeI and XbaI restriction sites. The ligation mixture was transformed into E. coli DH10B ultracompetent cells (Invitrogen) by electroporation. The library plasmid DNA was purified using a QIAprep miniprep kit (Qiagen Inc.).
Phenotypic Selection Based on Color Formation-The bicistronic library DNA was transformed into the E. coli trnA Ϫ strain by heat shock at 41°C, plated on large bioassay Q tray plates, and allowed to grow at 30°C for 2 days. Individual colonies with blue color were picked and grown in 3 ml of LB medium containing ampicillin (100 g/ml) as starter cultures for 15 h at 32°C (see Fig. 1). Plasmid DNA was prepared from the cultures using a QIAprep miniprep kit. Verification of the new mutants was performed by inoculating the starter cultures into 1 ml of Terrific Broth expression medium (1%, v/v) fortified with 1 mM 5-OBzl-indole on 24-well plates. The plates were then incubated at 30°C for 30 h with gyratory shaking at 250 rpm. Clones able to transform 5-OBzl-indole to form blue cultures were selected.
Nucleotide Sequence Analysis-Sequencing of the plasmid DNA in the clones that produced blue cultures (following verification) was performed in the Vanderbilt facility using an Applied Biosystems Model 3700 fluorescence sequencing unit with a Taq dye terminator kit (Applied Biosystems, Foster City, CA). The mutated sequences were verified by comparing the sequences of the sense and antisense strands (entire open reading frame).
Construction of Site-directed Mutants-Site-directed mutagenesis was conducted using the QuikChange mutagenesis kit (Stratagene) following the supplier's protocol. The primers were designed so that the mutation position was close to the middle with ϳ10 -15 bases of template complementary sequence on both sides and with a melting temperature of Ն75°C (supplemental data, Table 1S). For practical purpose, a pBluescript SK(ϩ) vector of the selected P450 2A6 mutant DNA (Bn1; ϳ4.5 kb) was first constructed and subjected to site-directed mutagenesis (see Fig. 1). The 1.5-kb mutated fragment was subcloned into the pCW2A6bc vector (ϳ8.5 kb) using the NdeI and XbaI restriction sites. The ligation mixture was transformed into E. coli, and positive clones were confirmed following purification of plasmids. The constructed site-directed mutant plasmids were verified by sequencing both strands of DNA (see above).
Whole Cell Assay for Oxidation of Indoles-Cells were incubated at 28°C for 14 h with gyratory shaking at 200 rpm. Cell pellets containing expressed mutant P450 2A6 and NADPH-cytochrome P450 reductase were harvested from 500 ml of Terrific Broth expression medium of the E. coli trnA Ϫ strain by centrifugation at 5000 ϫ g for 20 min, washed with 200 ml of M9 minimal medium (28), resuspended in 100 ml of M9 minimal medium, and used directly (after measuring the P450 concentration). The reaction mixtures consisted of 200 pmol of P450 (whole cells), 100 mM potassium phosphate buffer (pH 7.4), and varying concentrations of substrate in a total volume of 1 ml. The reactions were carried out at 37°C in 24-well plates with gyratory shaking at 280 rpm and quenched by the addition of 0.1 ml of 0.5 M aqueous NaOH after 30 min (for indole and 4-chloroindole) or after 100 min (for 4-and 5-OBzlindoles). The formation of colored products (indigo and indirubin) was measured by recording absorbance spectra in the visible region, and the estimated concentrations were calculated based on the extinction coefficient of the major product (subsequently determined). Estimates of k cat and K m were made using nonlinear regression with GraphPad Prism software.
Preparation of Membranes and Purification of Recombinant WT P450 2A6 and Mutants-Bacterial inner membranes containing mutant P450 2A6 and NADPH-cytochrome P450 reductase were isolated from 500 ml of Terrific Broth expression medium of the E. coli trnA Ϫ strain. The pellets were resuspended in a final volume of 10 ml of 50 mM Tris acetate (pH 7.4) containing 250 mM sucrose and 0.25 mM EDTA and homogenized with a Dounce homogenizer. The homogenized membrane fractions were stored on ice until analysis.
Estimation of Substrate Binding Affinity-The binding affinity of P450 2A6 (WT and mutants) for indole or 5-OBzl-indole was estimated by titration (29). Substrates were dissolved in Me 2 SO and added to 1.0 ml of 100 mM potassium phosphate buffer (pH 7.4) containing 2-4 nmol of purified enzyme, and spectra were recorded following each addition (350 -500 nm). The absorbance changes at 390 and 420 nm were used to estimate a spectral K d (K s ) using GraphPad Prism software, with fitting to a one-site binding hyperbola equation using nonlinear regression.
Coumarin 7-Hydroxylation Assay-The reaction mixtures contained 20 pmol of P450 (cell membranes), 100 mM potassium phosphate buffer (pH 7.4), and varying concentrations of coumarin (from a 100-fold stock in H 2 O) in a total volume of 0.5 ml. The reactions were carried out at 37°C and quenched with 0.1 ml of 2 M aqueous HCl after 5 min. The product was extracted as described (30) and assayed fluorometrically in a plate reader.
Large Scale Biotransformation and Product Separation-The biotransformations were performed in 2.8-liter Fernbach flasks with 500 ml of Terrific Broth expression medium. The cultures were incubated for 5 h at 30°C with gyratory shaking at 250 rpm. At that time, the substrate (4-or 5-OBzl-indole) in Me 2 SO solution (1 M stocks) was added to the medium (0.5 ml), and shaking was continued for 22 h at 30°C.
The pelleted cells and supernatants were separated by centrifugation at 5000 ϫ g for 30 min. The supernatant was extracted with ethyl acetate (3 ϫ 200 ml) and dried with Na 2 SO 4 . The pellet (from 500 ml of bacterial culture) was stirred with 500 ml of (CH 3 ) 2 CO and filtered through Celite (18,19). The ethyl acetate and (CH 3 ) 2 CO extracts were combined and concentrated in vacuo, and the resulting solids were fractionated by semipreparative HPLC. HPLC was done on a Beckman Ultra[sphere column (10 ϫ 250 mm, 5 m) with a CH 3 CN/H 2 O mobile phase at a flow rate of 3 ml/min with 40% (v/v) CH 3 CN (in H 2 O) for 10 min, a 40 -60% CH 3 CN linear gradient over 20 min, a 60 -80% CH 3 CN linear gradient over 10 min, and a 80 -100% CH 3 CN linear gradient over 5 min. To separate compounds 5 and 6, HPLC conditions of 30% (v/v) CH 3 CN (in H 2 O) for 10 min and a 30 -35% CH 3 CN linear gradient over 20 min were used with the same column. Detection was at 254 or 600 nm (single wavelength) or with the on-line HPLC/diode array detector system. Peaks were collected manually, and CH 3 CN was removed with a rotary evaporator. The remaining aqueous phase was lyophilized to dryness. In analytical systems, a Beckman Ultrasphere HPLC column (4.6 ϫ 250 mm, 5 m) was used with a CH 3 (14). This mutant served as the starting material in this study. The cDNA was subjected to error-prone PCR, and the resulting proteins were screened for their ability to convert 5-OBzl(OBn)-indole to colored products. The result was a mutant (Bn1) that was found to have a total of five mutations relative to WT P450 2A6 (I140M, L240C, N297Q, I300V, and I366V). Site-directed mutagenesis was used to produce all combinations of these substitutions and to establish which are necessary to allow P450 2A6 to oxidize the bulky 4-and 5-OBzl-indoles. The result was Bn1/125, with only two substitutions (N297Q and I300V). SRS, substrate recognition site.    (20) in silico using interactive graphics techniques in the program O (31), followed by manual docking of each of the two substrates into the binding site cavity. A rotamer of Gln capable of donating a hydrogen bond to the substrate bound in the active site was selected. Subsequent energy minimization (conjugate gradient method) was performed using a CNS software routine with no experimental energy term (32). Energy minimization resulted in molecular models of P450 2A6 with bound substrates having a root mean square deviation of 0.55 Å for C-␣ atoms compared with the original structure. Subsequently, ligand coordinates were removed, and protein coordinates were used as a rigid receptor for docking of these same substrates using the FlexX Suite module of Tripos SYBYL software (33), including the pharmacophore constraint of Gln 297 in the enzyme active site as a hydrogen bond donor.

RESULTS
Library Screening-Two randomized libraries were constructed using error-prone PCR following the manufacturer's instructions (Stratagene) to adjust the mutation rate to zero to three mutations/1000 bp using either WT P450 2A6 or the double mutant L240C/N297Q as a template (Fig. 1). The screening method was a convenient colorimetric colony-based method that allows rapid screening of thousands of colonies by simply plating them on LB expression plates fortified with the target substrate (5-OBzl-indole in this case) and picking up the blue colonies directly from the plate. LB expression medium was used instead of Terrific Broth to decrease the background color. A tryptophanase-negative strain (trnA Ϫ ) of E. coli (19) was used to avoid the contribution of indole to the products, i.e. to force the P450 2A6 mutants to use only OBzl-substituted indoles. Because of the temperature sensitivity of this strain ( prophage can be switched on at 42°C) (34), a slightly lower temperature than 42°C was used for the heat shock to achieve a sufficient number of colonies to work with. A total number of ϳ3 ϫ 10 3 colonies were screened from each of the two libraries. No blue colonies were found upon screening the WT P450 2A6-derived library. The double mutant-derived library yielded three colonies, designated Bn1 (with a clear blue color), Bn2, and Bn3 (with a brown color), among which only Bn1 yielded a distinctive blue color when incubated in 1 ml of Terrific Broth expression medium fortified with 5-OBzlindole (Fig. 1). This mutant was determined to be P450 2A6 I140M/ L240C/N297Q/I300V/I366V by sequence analysis, with three more mutation positions added to the template DNA.
Critical Residue Identification and Kinetic Parameters for WT P450 2A6 and Mutants-To determine which mutation positions were critical in gaining the ability to oxidize 5-OBzl-indole, 13 site-directed mutants were constructed from the Bn1 mutant, with one to four mutation sites recovered ( Fig. 1 and TABLE ONE, Bn1/1 to Bn1/1254). The hydroxylation of indole (the native substrate), 4-chloroindole (bearing a small substituent), and 4-and 5-OBzl-indoles (bearing bulky substituents) was measured based on the formation of colored dimers of the substrates (16). The low solubility of the 4-and 5-OBzl-indole substrates in H 2 O made vigorous shaking a necessity to increase diffusion; therefore, whole cells containing WT and mutant P450 2A6 were used instead of membranes or purified enzymes, which were more fragile under these conditions. The amounts of colored products were calculated based on their extinction coefficients, which were obtained either from the literature (indigo) (35) or from the isolation, identification, and quantification of the products as described under "Characterization of Colored Products Generated from 4-Chloroindole and 4-and 5-OBzlindoles" (see below).
For the small substrates indole and 4-chloroindole, all of these mutants (as well as WT P450 2A6) had catalytic activity; the original double mutant (L240C/N297Q) was one of the most efficient in terms of catalytic efficiencies. No dramatic enhancement of activity was found for indole or 4-chloroindole oxidation catalyzed by the new mutants, which was not surprising because the screening method used in this work focused specifically on 5-OBzl-indole biotransformation.
Convenient as the colorimetric method is, the caveat exists that those mutants that showed no color derived from 4-or 5-OBzl-indole might catalyze the oxidation of the substrates to products without absorbance in the visible range. To address this possibility, HPLC profiles (monitored at 254 nm) were analyzed with crude extracts from the biotransformation of 4-and 5-OBzl-indoles catalyzed by WT P450 2A6, the original double mutant (L240C/N297Q), N297Q/I300V (Bn1/125), and the pCWЈ-hor vector (with only human NADPH-cytochrome P450 reductase) as negative controls (Fig. 3). The expression levels were normal 11 h after induction. In the HPLC profiles, the cell extracts from WT P450 2A6 and P450 2A6 L240C/N297Q showed the same patterns as negative controls, and no products were detected.

Substrate Binding Affinities of WT P450 2A6 and Mutants-
The binding affinities of WT P450 2A6 and some mutants for indole and 5-OBzl-indole were estimated (TABLE TWO). In general, there was no correlation between the catalytic activities and the apparent K d values. WT P450 2A6 had the lowest catalytic efficiency toward indole among the enzymes tested; however, this enzyme showed the tightest binding of indole (apparent K d of 13 M). For 5-OBzl-indole, which was the substrate for the Bn1 and Bn1/1 enzymes but not for WT P450 2A6, the K d values are extremely high for Bn1 (apparent K d of 710 M) and Bn1/1 (apparent K d of 414 M), but very low for the WT enzyme (apparent K d of 8 M). These discrepancies suggest that some of the apparent binding may be nonproductive.
Coumarin 7-Hydroxylation Activity-Analysis of the rates of coumarin 7-hydroxylation with membranes prepared from WT P450 2A6 and the double mutants L240C/N297Q and N297Q/I300V (Bn1/125) indi-  DECEMBER 9, 2005 • VOLUME 280 • NUMBER 49 cated that the k cat for both mutants was increased ϳ5-fold over that for the WT enzyme, whereas the catalytic efficiency indicated by k cat /K m was not enhanced due to an increase in K m (TABLE THREE).

Characterization of Colored Products Generated from 4-Chloroindole and 4-and 5-OBzl-indoles-
The colored products from cell cultures fortified with indole, 4-chloroindole, and 4-and 5-OBzl-indoles were analyzed by HPLC at a wavelength of 600 nm. The main product derived from indole was indigo (compound 1), as established by comparison with standard material. 4-Chloroindole yielded 4,4Ј-dichloroindigo (compound 2) as a single peak (Fig. 4). 5-OBzl-indole yielded mainly 5,5Ј-di-OBzl-indirubin (compound 3) (Fig. 4), an expected dimer based on our previous work (18,19). The main product of 4-OBzl-indole was originally expected to be an indigo because the substituent group at the 4-position makes formation of an indirubin unfavorable due to its interference with the carbonyl group at the 3Ј-position (19,36). However, the dominant blue dye isolated upon incubations with 4-OBzl-indole was neither indigo nor indirubin, although it had the same molecular mass as that of the indigoids and a UV-visible spectrum similar to that of indigo. 1 H NMR clearly showed an unsymmetrical structure with one group of two doublets and one triplet (hydrogens from a 4-substituted benzene ring) plus another group of a singlet and two triplets with small coupling constants (hydrogens at the 2-and 3-positions of the pyrrole ring), suggesting that a pyrrole ring and a benzene ring are connected with a double bond instead of a pyrrole-pyrrole connection. 1 H-1 H COSY, 1 H-1 H NOESY, 1 H-13 C heteronuclear multiple quantum correlation, 1 H-13 C HMBC, and 1 H-13 C CIGAR-HMBC NMR experiments (supplemental data, Figs. 1S-5S) were used to assign the structure as 4-OBzl-2-(4Ј-OBzl-1Ј,7Ј-dihydro-7Ј-oxo-6ЈH-indol-6Ј-ylidene)indolin-3-one (compound 4) (Fig. 4).
The formation of this compound was as reproducible as that of the other indigoids. The yield of this dye from cultures (13 mol/liter of culture) is at the same level as that of 5,5Ј-di-OBzl-indirubin (9 mol/ liter of culture) by the same mutant, P450 2A6 N297Q/I300V (Bn1/ 125), and similar to that reported in our previous work on 6-nitroindole and methyl indole-5-carboxylate converted by P450 2A6 L240C/ N297Q (19). Other gain-of-activity mutants such as Bn1/52, Bn1/51, etc., also produced this dimer as the dominant colored product.
The spectroscopic data of the four colored dimers follow. Indigo Identification of Intermediates Generated in Vivo by Recombinant P450 2A6 N297Q/I300V-As shown in Fig. 3, many other products were found in the cell extracts in addition to the colored dimers that have absorbance at 254 nm. Considering the possible interference of other compounds from the E. coli system, liquid chromatography/tandem mass spectrometry (APCI positive mode) was done with the mixtures to help identify peaks that have a fragment at m/z 91 (benzyl group) as evidence of the origin from the substrate 4-or 5-OBzl-indole.   Although there were several dimers detected by liquid chromatography/tandem mass spectrometry and 1 H NMR, the amounts were insufficient for full characterization. 4-OBzl-indolin-2-one (compound 7) and 5-OBzl-indolin-2-one (compound 11) appear to be the primary oxidation products formed from each corresponding substituted indole, consistent with our previous study of indole oxidation (17).
Postulated Mechanism for 4-OBzl-indole Biotransformation-On the basis of the above results and previous work (17), we propose that the mechanism involved in the formation of compound 4 follows the course shown in Fig. 6. The substrate 4-OBzl-indole is first hydroxylated at one of the five free positions to form compounds 7-10 and an indoxyl (compound 13a), which undergoes further oxidation to form isatin (compound 5) via a tautomer (compound 13b) and an intermediate com- pound (compound 13c). Accordingly, the primary compounds 9 and 10 can be oxidized to give compound 6, possibly via radical intermediates (compounds 9a and 10a, respectively). Alternatively, oxidation of compounds 9 and 10 to catechols can be proposed, followed by 2-electron oxidation to the o-quinone (compound 6). The final blue dimer (compound 4) then could be the result of combinational reaction of compounds 13c and 6 or compounds 10a and 5.
Substrate Docking and Modeling of the P450 2A6 N297Q/I300V Mutant-Analysis of the active site in the crystal structure of the P450 2A6-coumarin complex indicates that coumarin occupies a cavity of ϳ359 Å 3 . The single mutation I300V, despite elimination of only one methylene group in Val compared with Leu, results in expansion of the active site to an adjacent hydrophobic cavity built by Val 110 , Ile 208 , Leu 241 , and Leu 244 and adds an additional 83 Å 3 to the total cavity volume, increasing the latter up to 440 Å 3 (Fig. 7A). The functionally conservative substitution N297Q does not significantly affect the cavity volume (353 Å 3 ). However, the longer side chain of glutamine presumably enables a new hydrogen bond with the oxygen atom in the substrate   Orientation of the cavity axis at ϳ60°to the heme plane (Fig. 7A) suggests a likely substrate-binding model in the active site and explains the reaction products obtained upon conversion of both 4-and 5-OBzlindoles. Alignment of the long axis of 5-OBzl-indole with the cavity axis and putative flipping of the compound along this axis would always present the pyrrole ring to the oxygen scission site (Figs. 7C and 8A). In contrast, flipping of 4-OBzl-indole along its long axis would present both pyrrole and benzene rings to the oxygen scission site (Fig. 8, B and C). The representative binding modes for both substrates resulting from automated docking are shown (Fig. 8). In about one-half (15 of 30) of the highest scoring solutions 5-OBzl-indole was bound in the conformation represented by that shown in Fig. 8A, with the pyrrole ring presented to the oxygen scission site. As expected, automated docking of 4-OBzlindole returned a larger variety of docked complexes, with the highest scoring conformers represented in Fig. 8B, where the benzene ring is exposed to the oxygen scission site. An alternative family of conformers featuring the pyrrole ring of 4-OBzl-indole as a target for oxygenation is shown in Fig. 8C.

DISCUSSION
We have reported the identification of P450 2A6 enzymes with expanded function using a process of sequential random and site-directed mutagenesis. New chemical reactions were also found for some products formed by the new mutants. These mutations provide for more diversity of product formation. The available x-ray structure of the P450 2A6-coumarin complex (20) has been used to explain the effects of the mutations.
Although exhaustive systematic screening of the WT P450 2A6 and P450 2A6 L240C/N297Q (the original double mutant) libraries to cover 95% possible changes (38) was not performed, this strategy used here was successful. Searching for gain-of-activity mutants, for which a low frequency is expected, using the colony-based method described here should be considered more of a selection than a screening process. All blue colonies definitely indicate a gain-of-activity mutant. Any incorrect identification resulted from the uncertainty in reading the color as "blue" due to the disturbance of background color and the brown color of the colonies, which made the second confirmation necessary.
From the results of site-directed mutagenesis, we know that one of the critical changes is the N297Q mutation, an existing mutation in the original double mutant enzyme. This mutant (L240C/N297Q) had been selected previously in a molecular breeding experiment (14) that was based on the use of randomized oligonucleotides (39) targeted to the substrate recognition site regions expected to comprise the active site (2). The nucleotide change is from AAC to CAA, a complex double transversion that would not likely be achieved by error-prone PCR. This could be the main reason why no gain-of-activity mutants were found in the WT P450 2A6 library.
Analysis of the N297Q and I300V substitutions in silico using the available crystal structure revealed that, although the I300V mutation adds ϳ83 Å 3 to the wild-type binding site, resulting in 440 Å 3 of total space, it is not enough to accommodate either of the OBzl-substituted indoles used in this study (which would require a volume of ϳ600 Å 3 ). Presumably substrate-induced fit adjustment to the enzyme active site is required to complement the favorable steric and electronic effects of both mutations as deduced from the substrate-docking model. Alternatively, the active site of the N297Q/I300V mutant may have changed enough from that of the WT enzyme to expand the space, although we prefer the former hypothesis regarding induced fit.
The compound 4 colored dimer appears to be novel in that no such structure has been reported. In our previous work, ϳ20 indigoids were isolated from P450 2A6-catalyzed reactions (18,19). All of them have parent rings of indigo or indirubin, in which both positions of the pyr- role ring (positions 2 and 3) are oxidized and one of the positions is connected to another monomer, leaving the benzene ring intact, consistent with the reactivity of the indole. Indole is a -excessive heteroaromatic, which means that the -electron densities on carbon atoms of the pyrrole ring are greater than those of the benzene, and the pyrrole ring should be more readily oxidized. The most important resonance structures of indole maintain an unbroken benzene ring and bear the positive charge on nitrogen and the negative charge on the C-3 atom. The 3-position of the indole ring is highly reactive toward electrophilic reagents (40). Although it has been reported that indole can be hydroxylated at the 5-and 6-positions in P450-catalyzed oxidation systems, the amounts of these products are limited, and they apparently do not undergo further oxidation to form dimers (17,41). The intermediate 4-OBzl-1H-indole-6,7-dione (compound 6) does have analogs in nature, e.g. the redox cofactors in methylamine dehydrogenase (tryptophan tryptophylquinone) (42) and amine dehydrogenase (cysteine tryp-tophylquinone) (43), which are also 6,7-diones derived from L-tryptophan and formed during post-translational modification. How the two quinone oxygens are incorporated into tryptophan tryptophylquinone is still unclear, although it is believed that other enzymes, probably monooxygenases, are required for the biogenesis instead of a self-processing mechanism (44). Tryptophan tryptophylquinone could be considered as an indole dimer as well in that the pyrrole ring (C-2) of Trp 108 is connected to the benzene ring (C-4) of Trp 57 with a single bond. Another report on the oxidation of indole derivatives includes structures involving formation of dimers with the benzene-benzene ring connected; in this case, however, the 3-position of pyrrole was protected and the benzene ring was dioxidized in advance (45).
The isolation of six intermediates upon incubation with 4-OBzl-indole, especially those derived from hydroxylation on the benzene ring, not only supports the formation of compound 4, but also clearly indicates that the substrate is accommodated inside the active site with Expansion of Substrate Specificity of P450 2A6 DECEMBER 9, 2005 • VOLUME 280 • NUMBER 49 variable orientations. On the other hand, the gain-of-activity mutants retained their normal activities (producing indigo or indirubin), like WT P450 2A6, with the smaller indole compounds tested (indole and 4-chloroindole), which are presumed to be more flexible in the active site. The lack of oxidation of the benzene ring of indole by the mutant might be explained partially by the electron-donating property of the OBzl group. The 4-OBzl substituent may activate the para-position to be hydroxylated to compound 8, which contributes to the whole process (Fig. 6). In addition, the rigidity of substrate docking of a bulky substrate compared with a substrate of smaller size may also contribute to the increased chance of hydroxylation on the benzene ring, as long as the substrate can be accommodated properly inside the active site with the benzene ring presented to the active oxygen species.
In conclusion, we have isolated new mutants of P450 2A6 with expanded substrate specificity toward bulky substituted indoles from a randomized library constructed by error-prone PCR and identified the critical residues that contribute to the effect by site-directed mutagenesis. The products (dimers and monomeric oxidation products) were characterized, and we proposed a rational mechanism for the formation of an unusual dimer derived from 4-OBzl-indole. We interpreted the results in the context of a crystal structure of a P450 2A6-coumarin complex. The same approach, together with the analysis of the crystal structure, could be adapted to generate more diversity of products from the P450 2A6-catalyzed metabolism and could be used as a tool in the development of P450 enzymes with new functions.