Materials and Methods

Chemicals.

RPR 106541 (Batch TP2969 and Lot JYW 417/3), RPR 104065 (methyl 16α,17α-butylidene bis(oxy)−6α,9α-difluoro-11β,21-dihydroxy-3-oxoandrosta-1,4-diene-17β-thiocarboxylate; Batch NCP80), RPR 112020 (S-diastereomer sulfoxide of RPR 106541; Batch JYW 473P), RPR 112903 (R-diastereomer sulfoxide of RPR 106541; Batch JYW 490/2), RPR 112022 (sulfone metabolite; Batch JYW 47P), and RPR 112023 (3-hydroxy derivative of RPR 106541; Batch JYW 47P) were all obtained from RPR, Dagenham Research Center (Dagenham, England). CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonic acid}, glucose 6-phosphate, NADP⁺, β-NADPH, glucose 6-phosphate dehydrogenase, and troleandomycin (TAO) were purchased from Sigma Chemical Co. (St. Louis, MO). Methyl-t-butyl ether and isopropyl alcohol were purchased from EM Science (Gibbstown, NJ). Acetonitrile was purchased from J.T. Baker (Philipsburg, NJ). All other chemicals were obtained from standard vendors.

Biological Reagents.

Human liver samples were obtained through organ procurement agencies (Anatomic Gift Foundation, Woodbine, GA; International Institute for the Advancement of Medicine/In Vitro Technologies, Exton, PA; National Disease Research Interchange, Philadelphia, PA; and the Association for Human Tissue Users, Tucson, AZ) in accordance with proper ethical procedures for consent. Rat (Sprague-Dawley), dog (beagle), and human liver microsomes were prepared according to the procedure of Wang et al. (1983). Anti-CYP3A1 IgG was purchased from Human Biologics (Phoenix, AZ), and IgG stock solutions were diluted with PBS. Microsomes prepared from human lymphoblastoid cells or baculovirus-infected insect cells transfected with individual CYP forms were purchased from Gentest Corp. (Woburn, MA). Mutagenesis of CYP3A4 residues and enzyme expression were performed as described previously (Domanski et al., 1998).

Enzyme Assays.

The following enzyme assays were used to monitor specific CYP forms as described previously (Heyn et al., 1996): 100 μM phenacetin O-deethylation for CYP1A2, 50 μM coumarin 7-hydroxylation for CYP2A6, 200 μM S-mephenytoinN-demethylation for CYP2B6, 200 μMS-mephenytoin 4′-hydroxylation for CYP2C19, 0.5 mM chlorzoxazone 6-hydroxylation for CYP2E1, and 50 μM nifedipine oxidation for CYP3A4. CYP2D6 activity was measured by bufuralol 1′-hydroxylation (Kronbach et al., 1987) at a substrate concentration of 100 μM, and diclofenac 4′-hydroxylation (20 μM diclofenac) was used as a CYP2C9 marker activity (Leemann et al., 1993).

Incubations of RPR 106541 with liver microsomes were conducted under the following conditions: 0.65 mg/ml microsomal protein, 50 mM Tris/HCl (pH 7.4), 1 mM NADP⁺, 1 U/ml glucose 6-phosphate dehydrogenase, and 100 μM RPR 106541 in a total volume of 1 ml. For the antibody inhibition experiments with RPR 106541, the microsomal protein concentration was reduced to 0.12 mg/ml and the incubation time was increased to 20 min to conserve antibody. Antibody, microsomal protein, and all other incubation components except substrate and glucose 6-phosphate were incubated with gentle shaking at room temperature for 30 min before the addition of the substrate. Incubations using CYP forms expressed in baculovirus-infected insect cells (Gentest Corp.) used 50 pmol CYP in a total volume of 0.2 ml, 60 min incubation time, and direct precipitation of protein followed by HPLC analysis as described below. Samples were preincubated for 3 min at 37°C in a shaking water bath, and the reaction was started by the addition of glucose 6-phosphate (10 mM final concentration). Reactions were terminated by the addition of 4 ml of methyl-t-butyl ether and the samples were extracted by mechanical shaking for 15 min followed by centrifugation for 10 min at 4000 rpm. For quantitative studies, 3 ml of the organic phase were removed and transferred to a clean tube. Samples were then evaporated with nitrogen and reconstituted in mobile phase (50% A/50% B, see below) before injection.

Samples were analyzed by HPLC using a Spherisorb ODS-1, 250 × 4.6 mm, 5-μm analytical column (Phase Separations, Franklin, MA) with a Zorbax ODS 4 mm × 1.25 cm guard column (MacMod Analytical, Chadds Ford, PA). A gradient elution was used at 1 ml/min with mobile phase A (30% acetonitrile/65% water/5% isopropyl alcohol) and B (55% acetonitrile/40% water/5% isopropyl alcohol): initial conditions were 50% solvent B for the first 5 min increasing to 70% B from 5 to 19 min, then increasing to 100% B from 22 to 26 min, and holding at 100% B until the end of the run (34 min). A Spherisorb ODS-1, 250 × 10 mm, 5-μm semipreparative column (Phase Separations) in combination with a 60 × 10 mm guard column packed with the identical material was used for larger scale separations. The flow rate of 4.73 ml/min represents a linear velocity equivalent to that of the analytical column. Kinetic analysis of the human liver microsome experiments was performed using a one-site model (GraphPad Software, Inc., San Diego, CA).

Mutagenesis and Initial Characterization of CYP 3A4 Residues 373, 479, and 480.

Plasmid pS3A4His (Domanski et al., 1998) was used as the template for construction of L373A, L479F, and G480Q. The Expand PCR kit (Boehringer Mannheim, Indianapolis, IN) was used with a common reverse primer that overlapped 21 bases with 479F and 480Q primers, which contained the desired mutations plus a silent mutation that removed a HindIII site (Fig.2). After amplification, the reactions were incubated with DpnI, an enzyme that will only digest methylated DNA. This process removed any circular, unmutated template plasmid. The DNA was transformed into DH5α cells, and plasmid DNA from the resulting colonies was isolated and digested withHindIII to screen for the desired mutations, because these constructs lacked one HindIII site. The entire 3A4 coding region was confirmed by sequencing (University of Arizona Sequencing Facility, Tucson, AZ). Primers L373A and L373 reverse were used to introduce the L373A mutation into the 3A4 cDNA (Fig. 2). After amplification of the desired region, the MunI toKpnI fragment was subcloned into the pS3A4His plasmid, replacing the wild-type fragment. The mutated region was sequenced to verify the presence of the desired mutation. The construction of the remaining mutants presented in Table 3 has been described previously (Harlow and Halpert, 1997, 1998; He et al., 1997; Domanski et al., 1998). The Expand PCR kit and DpnI were purchased from Boehringer Mannheim. All other restriction enzymes and Taqpolymerase were purchased from Life Technologies (Grand Island, NY).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Primers used to construct mutations as CYP 3A4 residues 373, 479, and 480. The underlined residue denotes the silent mutation removing a HindIII restriction site. The boldface bases denote mutations to make residue conversions.

Initial characterization of L373A, L479F, and G480Q was performed using protein preparations that were expressed in Escherichia coliand purified from solubilized membrane preparations using the Talon Affinity system (Clontech, Palo alto, CA) as described previously (Domanski et al., 1998). Testosterone and progesterone hydroxylation assays were also performed as described previously (Domanski et al., 1998).

Mutant Enzyme Assays.

CYP3A4 mutants were provided as purified preparations (Harlow and Halpert, 1997; He et al., 1997;Domanski et al., 1998). Preparations were reconstituted before incubation with RPR 106541 by the addition of the following reagents (in order of addition): CYP, 4-morpholinepropanesulfonic acid (MOPS) buffer [100 mM MOPS (pH 7.3), 10% glycerol, 0.2 mM dithiothreitol, 1 mM EDTA], CHAPS (10% stock), dioleoylphosphatidylcholine (DOPC; 1 mg/ml stock), reductase, and cytochromeb₅. The CYP/reductase/b₅ ratio was chosen as 1:4:2 based on previous data showing this ratio was optimal for testosterone and progesterone hydroxylase activities (Domanski et al., 1998). The CHAPS concentration was 0.25–0.4%, and the DOPC concentration was 0.1 mg/ml. The MOPS buffer was added only to dilute all the CYP samples to an equal concentration before adding the remaining reagents. The reconstitution was carried out at room temperature for 10–15 min. The reaction volume ranged from 666 μl to 1.040 ml with the following conditions: 50 pmol CYP, 100 μM RPR 106541, 50 mM HEPES buffer (pH 7.6), 15 mM MgCl₂, 0.1 mM EDTA, 0.04% CHAPS, and 0.1 mg/ml DOPC, and the reactions were started by the addition of 1 mM NADPH (final concentration). The incubation time was 60 min, and the reactions were terminated and analyzed as described above.

Computer Modeling.

The model of CYP 3A4 developed previously (Szklarz and Halpert, 1997) was used to dock substrate RPR 106541 into the active site in the proper orientation for the formation of theR-sulfoxide metabolite to understand the changes in specificity resulting from certain amino acid substitutions. The substrate was placed into a reactive binding orientation, with the oxidation site fixed at 3.24–3.5 from the heme oxygen as described previously (Szklarz and Halpert, 1997). This leads to sulfoxidation and formation of the R-sulfoxide metabolite. Conformational analysis of RPR 106541 was performed with the SEARCH COMPARE module of INSIGHT II (Molecular Simulations, Inc, San Diego, CA), and the total energy of the docked substrate and protein were analyzed using the DOCKING module (Szklarz et al., 1995). The docking interaction between the enzyme and the substrate was optimized using energy minimizations within INSIGHT II (Szklarz et al., 1994).

Structural Analysis.

Liquid chromatography/mass spectrometry (LC/MS) analyses were performed on a Sciex API III spectrometer (Sciex, Foster City, CA) interfaced to a Hewlett-Packard model 1050 chromatography system (Palo Alto, CA). The HPLC separations used a combined gradient on a microbore column (Spherisorb ODS-1, 1 × 150 mm, 5 μm, Phase Separations). The mobile phase comprised A = 20 mM ammonium acetate/isopropyl alcohol, 90:10, and B = 20 mM ammonium acetate/acetonitrile/isopropyl alcohol, 45:45:10. The gradient used was 50% B for the first 5 min, then increased to 70% B over 20 min and held for 5 min, and finally increased to 100% B over 5 min. Using MS/MS experiments, structural information was derived from molecular fragment ions produced either by collision-induced dissociation using IonSpray, or at higher energies with heated nebulizer and corona discharge. Electron impact mass spectra were acquired at 70 eV using a Finnegan (San Jose, CA) model 4500 spectrometer with a direct insertion probe. Proton NMR spectra were obtained at 500 MHz with a Varian UnityPlus (Varian, Palo Alto, CA) spectrometer atmospheric pressure chemical ionization with the analyte in CDCl₃ and tetramethylsilane as internal reference. Infrared spectra were recorded at 4 cm⁻¹ resolution on a Nicolet model 740 FTIR with an IRPLAN microscope (Nicolet, Madison, WI).

X-Ray Crystallographic Structure Determination of RPR 112020.

A single crystal of RPR 112020 at a temperature of 153 K was used for the x-ray crystallographic measurements on an Enraf-Nonius CAD-4 diffractometer (Enraf-Nonius, the Netherlands), with graphite crystal monochromatized Mo radiation. An empirical ψ-scan absorption correction was applied with the data from nine reflections. The intensities of six standard reflections, monitored every hour of crystal X-ray exposure time, showed no change in the average intensity over the measurement period. Calculations were performed using TEXSAN (Molecular Structure Corp., Molecular Structure, The Woodlands, TX) and SHELXTL (Siemens Industrial Automation, Inc.) software programs (Siemens, Munich, Germany). The structure was solved with direct methods and refinement was by full-matrix least-squares with anisotropic temperature factors for the C, F, O, and S atoms and isotropic terms for the H atoms. The C-linked H atoms were positioned from the heavy atom framework; the hydroxyl H atoms were located in a difference electron density map. There were two crystallographically independent molecules per unit cell. The absolute configuration was established by refinement of the Flack xparameter (0.038 for this determination; Flack, 1989).

Results

Identification of RPR 106541 Sulfoxide Metabolites.

Preliminary metabolite profiling studies consisted of the incubation of rat, dog, and human liver microsomes with RPR 106541 followed by HPLC analysis. These incubations produced two major metabolites with retention times of 13.1 (M1) and 13.9 min (M2;data not shown). Preliminary LC/MS analysis of the organic extract of incubations of RPR 106541 with dog liver microsomes showed two closely eluting peaks that each produced a molecular ion of 473 (M + H⁺). Because of the highly efficient conversion of RPR 106541 to M2 by female rat liver microsomes, larger scale metabolite generation was undertaken using this enzyme source. Specifically, 30 identical 60-min incubations of 3 mg/ml female rat liver microsomal protein and 300 μM RPR 106541 were conducted. Following extraction, the samples were pooled and injected onto a semipreparative (250 × 10 mm) Spherisorb ODS-1 column.M2 was collected from multiple injections, evaporated under nitrogen, and dried in vacuo. The isolated sample was then analyzed by infrared, NMR (proton), and mass (electron impact, LC MS/MS) spectrometry.

Figure 3 shows the electron impact spectra of M2 from dog liver microsome incubations. The fragment ions revealed the loss of CH₃SOH (product ion m/z 409) and C₄H₇OH (product ionm/z 337), consistent with the structure of a sulfoxide of RPR 106541. The fragmentation patterns for M2 isolated from rat liver microsome incubations and from the dog liver microsome incubations were identical with each other and to the fragmentation pattern for M1. The proton NMR spectra for M2produced characteristic shifts in the positions of the hydrogen atoms adjacent to the sulfur atom. Specifically, the hydrogens of the thiomethyl group (H13) of RPR 106541 had a chemical shift of 2.134 ppm; in contrast M2 exhibited a H13 chemical shift of 2.694 ppm. The deshielded position (relative to RPR 106541) of theS-methyl resonance (hydrogen position H13) for M2is indicative of a sulfoxide. In addition, the appearance of this resonance as a singlet at high resolution indicates that M2consists of a single sulfoxide enantiomer. Shifts were also observed for hydrogens at positions 3 (t, 5.616 ppm), 5 (d, 5.273, 5.262), and 23 (s, 1.106).M2 was later shown to cochromatograph with a synthetic standard of the S diastereomer RPR 112020 (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Electron impact spectra of M2 isolated from incubations of RPR 106541 with rat liver microsomes. The fragmentation ions show the consecutive losses of the CH₃SOH [m/z 409, (M + H)⁺ ] and C₄H₇OH [m/z 337, (M + H)⁺] groups, consistent with sulfoxide formation. In combination with NMR and X-ray crystallography data, this compound was identified as the S-sulfoxide metabolite (RPR 112020).

Finally, a single crystal X-ray structure determination of RPR 112020 was performed to unambiguously determine the stereochemistry of the sulfoxide position. Based on the refined Flack x parameter of 0.038, the absolute configuration of the sulfoxide was assigned asS. By default, RPR 112903 (M1) was assigned theR configuration.

Identification of the Human CYP Forms Responsible for RPR 106541 Sulfoxidation.

Based on preliminary kinetic experiments examining the effect of protein concentration and incubation time on the linearity of R- and S-sulfoxide formation, a protein concentration of 0.65 mg/ml and a 10-min incubation time were used. The kinetics of RPR 106541 sulfoxidation were then determined for three human liver microsome samples (HL-04, HL-05, and HL-06). The respective K_m andV_max values for all three samples are given in Table 1. ForR-sulfoxide formation, theK_m values ranged from 55 to 73 μM and the V_max values from 0.70 to 2.01 nmol/min/mg protein. For S-sulfoxide formation, theK_m values ranged from 21 to 59 μM and the V_max values from 0.84 to 2.46 nmol/min/mg protein. The three K_mvalues for each metabolite were compared by a two-sample ttest and determined not to be statistically different (p ≥ .05). A concentration of 100 μM RPR 106541 was thus chosen as saturating for the determination of the rates of sulfoxidation for a panel of human liver microsome samples. The results for 15 human liver microsome samples and the correlation with CYP3A4-marker nifedipine oxidase activity are shown in Fig.4A and B. For R-sulfoxide formation, the enzyme activity ranged 5.8-fold from 0.31 to 1.81 nmol/min/mg protein, and for S-sulfoxide formation, enzyme activity ranged 4.7-fold from 0.44 to 2.06 nmol/min/mg protein. These activities were then correlated with CYP marker activities previously determined for the same panel of human liver microsome samples. As shown in Table 2, of the eight marker activities included in the analysis, CYP3A4-catalyzed nifedipine oxidase activity produced the best correlation with the formation of both sulfoxides. Specifically, the correlation coefficients (r²) were 0.84 and 0.91 forS- and R-sulfoxide formation, respectively. The relatively strong correlation of sulfoxide formation with CYP2D6-marker bufuralol 1-hydroxylase activity was not supported by subsequent studies examining the effect of the CYP2D6 inhibitor quinidine or of expressed CYP2D6 on RPR 106541 sulfoxidation (see below).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Rates of formation of the R- (A) andS-sulfoxides (B) for 15 human liver microsomes samples correlated with the corresponding nifedipine oxidase activity. Ther² values were calculated from linear regression analysis.

The ability of CYP form-selective inhibitors to alter the sulfoxidation of RPR 106541 was also examined. As shown in Fig.5, RPR 106541 sulfoxidation was not affected by inhibitors of CYP 1A2 (furafylline), 2C9 (sulfaphenazole), 2D6 (quinidine), or 2E1 (diethyldithiocarbamate). Coumarin, a competitive inhibitor of CYP2A6, produced minimal inhibition of sulfoxide formation (∼15% for each diastereomer). The CYP3A4 substrate cyclosporin A produced 44% inhibition ofR-sulfoxide formation and 38% inhibition ofS-sulfoxide formation. Troleandomycin produced ∼50% inhibition of each reaction at a concentration of 50 μM, thus implicating CYP3A forms in RPR 106541 sulfoxidation. Antibody inhibition results for RPR 106541 sulfoxidation paralleled the chemical inhibition data. As shown in Fig. 6, 5 mg anti-CYP3A1 IgG/mg microsomal protein produced 64% and 70% inhibition of S- and R-sulfoxide formation, respectively. The same ratio of antibody inhibited 82% of CYP3A4-marker nifedipine oxidase activity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Inhibition of human liver microsomal RPR 106541 sulfoxidation by CYP-form selective inhibitors. The open bars representR-sulfoxide formation, and the striped bars representS-sulfoxide formation. Abbreviations used are Sulf (sulfaphenazole), CsA (cyclosporin A), TAO, Quin (quinidine), and DDC (diethyldithiocarbamate). Furafylline, sulfaphenazole, TAO, and DDC were preincubated with human liver microsomes in the presence of NADPH before the addition of substrate. The percentage of activity remaining is calculated based on the respective control values.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Inhibition of RPR 106541 sulfoxidation by anti-CYP3A1 antibody. The percentage of control represents enzyme activity in the presence of anti-CYP3A1 antibody relative to enzyme activity in the presence of the same ratio of preimmune (rabbit) sera to microsomal protein. The rate of R-sulfoxide (▪) andS-sulfoxide (▴) formation is shown in comparison to inhibition of CYP3A4 marker nifedipine oxidase activity (○). Details are provided in Materials and Methods.

Finally, the sulfoxidation of RPR 106541 by a panel of microsome samples of expressed CYP forms was measured. CYP1A2, -2B6, -2D6, -2E1, and -2A6 did not metabolize RPR 106541. CYP2C19 and -3A4 catalyzed the formation of the R-sulfoxide at rates of 0.91 and 1.23 nmol/min/nmol CYP, respectively. For S-sulfoxide formation, the respective rates in nmol/min/nmol were as follows: CYP3A4, 1.33; CYP3A5, 1.21; CYP2C9, 2.27; and CYP2C19, 3.71. Because other investigators have shown a requirement for cytochromeb₅ with CYP3A-dependent reactions (Gillam et al., 1993; Yamazaki et al., 1996), a separate experiment examined the effect of b₅ addition on RPR 106541 sulfoxidation by CYP3A and -2C forms. The addition of 1 or 2 nmol b₅/nmol CYP did not affect the conversion of RPR 106541 to the R-sulfoxide, except that CYP3A5 now displayed a rate of R-sulfoxide formation of 0.5 nmol/min/nmol CYP. Addition of b₅ did not affect the rate of S-sulfoxide formation by CYP3A4, but did produce low levels (<50%) of stimulation ofS-sulfoxide formation by CYP3A5, -2C9, and -2C19 (data not shown).

Although CYP2C forms displayed the highest rates ofS-diastereomer formation, the involvement of CYP2C forms in the metabolism of RPR 106541 was not supported by other phenotyping approaches. Specifically, RPR 106541 sulfoxidation did not correlate with human liver microsomal CYP2C9-catalyzed tolbutamide 4-hydroxylase activity (r² = 0.01 and 0.00 forR- and S-sulfoxide formation, respectively). Also, 100 μM sulfaphenazole produced only 4% inhibition ofR-sulfoxide formation and 13% of S-sulfoxide formation. This finding is in contrast to the >80% inhibition of tolbutamide hydroxylation reported by Newton et al. (1995) at the same concentration of sulfaphenazole.

RPR 106541 Sulfoxidation by CYP3A4 Mutants.

The effect of site-directed mutagenesis of various CYP3A4 SRS residues on RPR 106541 sulfoxidase activity was studied using several purified preparations. These mutants were engineered based on a proposed CYP3A4 three-dimensional model and the localization of putative 3A4 SRSs based on alignment of 3A4 with nonmammalian CYP sequences (Szklarz and Halpert, 1997) as originally introduced by Gotoh (1992) for the CYP2 family. Table 3 shows the rates of formation of each enantiomer and the ratio of R:S-sulfoxide formation for each mutant and wild-type 3A4. The wild type produced an enantiomer ratio of 0.89, similar to that of commercially available CYP3A4 expressed in insect cells (0.92). Within SRS-2, the Leu210 → Ala mutation resulted in a sharp decrease in the sulfoxide ratio (0.24). The effect was decreased (0.55) for L211A. Substitution of Phe for Leu at 211, however, did not alter the ratio of sulfoxide formation compared with the wild type.

SRS-4 is located within the highly conserved I helix and has been shown to influence both the rate and stereoselectivity of endogenous steroid hydroxylation (Domanski et al., 1998). With RPR 106541, SRS-4 substitutions were clearly more detrimental to formation of theR-diastereomer in comparison to theS-diastereomer. For I301A, the decrease in the ratio to 0.42 was entirely dependent on the drop in R-sulfoxide formation. With F304A the ratio decreased to 0.25 due to a 3-fold decrease inR-sulfoxide formation and a slight increase in theS formation. S-sulfoxide formation was reduced by the change from the α-methyl at 305 to an α-hydrogen, resulting in a lower overall ratio. The substitution of valine at 305 produced both the lowest enantiomer ratio and the lowest rate ofR-sulfoxide formation of any mutant tested (0.20).

Residues 369, 370, and 373 have been shown to dictate testosterone and progesterone metabolite profiles (He et al., 1997). Although the corresponding 6β and 16α sites of metabolism are not available for hydroxylation with RPR 106541, altered sulfoxide production could provide evidence that these residues are involved in substrate binding within the enzyme active site. RPR 106541 sulfoxidation was not altered by I369V relative to the wild type, and L373A showed a diastereomer ratio close to the wild type, although with uniformly decreased rates of R- and S-sulfoxide formation. However, A370V showed a 50% decrease in R-sulfoxide formation whereasS-sulfoxide formation was increased ∼2-fold. This mutation has been postulated to stabilize the 16α binding orientation of progesterone (He et al., 1997) and may have a similar effect on the positioning of RPR 106541 for sulfoxidation in the Sconfiguration.

The effect of amino acid substitutions in the proposed SRS-6 of CYP3A4 on enzyme activity has not been reported. R-sulfoxide formation was selectively affected by a L479 → Phe alteration as evidenced by the 0.42 ratio. This change in activity for L479F is in contrast to the lack of effect of this mutation on the 6β-hydroxy to 16α-hydroxy ratio for the metabolism of progesterone or testosterone. (All of these mutants maintained 6β-OH to 16α-OH ratios similar to CYP 3A4 wild type, although G480Q had an overall decrease in activity of ∼8-fold and L373A displayed a 2-fold decrease.)

Interpretation of Mutagenesis Data Using Molecular Modeling.

Previously, the CYP-3A4 three-dimensional model has been used successfully to interpret mutagenesis data (Harlow and Halpert, 1997,Domanski et al., 1998). Therefore, this methodology was employed to interpret the changes in R-sulfoxide formation displayed by several mutants studied including L210A, I301A, F304A, A305V, A370V, and L479F. Figure 7 clearly illustrates and supports the results obtained in several of these instances. For example, the increased size of the valine side chain in A305V and A370V causes increased van der Waals overlaps, inhibiting formation of theR-sulfoxide. The same decrease in R-sulfoxide formation was observed for mutant L479F, where a leucine is replaced with a very bulky phenylalanine. I301A and F304A, however, have large side chains replaced by the much smaller alanine. The positioning of these residues suggests that the lack of the bulkier side chains may allow for excess movement of the substrate in the binding pocket, limiting residence time in a productive orientation. The side chain of mutant L210A appears to be too far from the docked substrate to play a significant role in substrate specificity; in fact, it is located more than 5 Å from the docked substrate. However, this residue may be located in the channel of entry, a possibility that cannot be addressed with current modeling methodology.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Substrate RPR 106541 docked into a CYP3A4 model in the proper orientation for the formation of theR-sulfoxide metabolite. The substrate is shown in gray, as are the side chains of the mutated residues A210, A213, A301, A304, V305, V370, and F479.

Discussion

The importance of identifying the major routes of metabolism for a new chemical entity and the enzymes involved is well recognized. The primary strategy for generating this information first typically involves coupling in vitro metabolism systems with structural analysis (LC/MS, NMR, etc.) to identify the major metabolites. If the reaction is then determined to be CYP-catalyzed, a variety of approaches (correlation with marker activities in a bank of human liver microsomes, chemical/antibody inhibition, expressed enzymes) are then used to determine the responsible human CYP form(s) (Wrighton et al., 1993; Guengerich, 1995). The experimental resources required for this work are significant, and thus limits the number of new chemical entities that can be evaluated as human pharmaceuticals. The development of more empirical and predictive methods could expedite this process and allow a description of the enzymology of a reaction to be incorporated in the new chemical entity selection process at an earlier stage. Specifically, methods using photoaffinity probes, molecular modeling, chimeric enzymes, and site-directed mutagenesis have been used to identify critical binding residues of mammalian CYP forms or to predict the site(s) of CYP metabolism (von Wachenfeldt and Johnson, 1995). In addition, homology models based on the crystal structure of bacterial CYP101 that could be applied to computer aided drug design have been generated for several human CYP forms, including CYP2C9 (Korzekwa and Jones, 1993), CYP2D6 (Koymans et al., 1993), and CYP3A4 (Ferenczy and Morris, 1989; Szklarz and Halpert, 1997).

In our present study, the sulfoxidation of RPR 106541 by human liver microsomes has been characterized by using a combination of reaction phenotyping techniques followed by an analysis of changes in metabolism by CYP3A4 SRS mutations. The R- and S-sulfoxide diastereomers are the major oxidative metabolites formed by human liver microsomes. The correlation of RPR 106541 sulfoxidase activities with human liver microsomal CYP3A4 marker nifedipine oxidase activity and the metabolism of RPR 106541 by expressed CYP3A4 and CYP3A5 strongly suggest that CYP3A forms are the primary catalysts of RPR 106541 sulfoxidation in human liver microsomes. The involvement of expressed CYP2C9 and CYP2C19 in RPR 106541 sulfoxidation was not supported by correlation analysis or chemical inhibition data. Yamazaki and Shimada (1997) have shown that CYP3A4, CYP2C9, and CYP2C19 are all involved in the 21-hydroxylation of progesterone, a site similar in spatial configuration to the S-methyl substituent of the D ring of RPR 106541. The contribution of CYP2C forms to RPR 106541 metabolism could explain why chemical inhibition studies showed that the CYP3A-selective inhibitor TAO produced only moderate inhibition of RPR 106541 sulfoxidation (∼50%). However, there is precedence for the finding of incomplete inhibition by TAO of reactions shown to be CYP3A-dependent by other reaction phenotyping approaches (Guengerich, 1990; Fleming et al., 1992; Kumar et al., 1994). Finally, antibody inhibition studies show that anti-CYP3A1 antibody inhibits the conversion of RPR 106541 to both the R- (70% of control) and S-sulfoxides (64% of control).

The evaluation of 14 CYP3A4 mutants for RPR 106541 sulfoxidase activity represents a comprehensive evaluation of CYP3A4 SRS sites 2, 4, 5, and 6 and a logical extension of the phenotyping studies to CYP3A4 structure-function analysis. Within SRS-2, the Leu210 → Ala mutation decreased the enantiomer ratio from 0.89 (wild type) to 0.25, predominantly due to a decrease in R-sulfoxide formation. This substitution has been shown previously to decrease the rates of testosterone 2β- and 15β-hydroxylation and to alter the response to stimulation by α-naphthoflavone, suggesting that effector and binding sites of CYP3A4 may overlap (Harlow and Halpert, 1997). Also, Leu210 and Leu211 of CYP3A4 are highly conserved between species; however, CYP3A5 has Phe at residue 210. The Leu210 → Phe substitution did not alter the enantiomer ratio (1.06). This finding suggests that the difference in the ratio of sulfoxide enantiomer formation between CYP3A4 and 3A5 cannot be accounted for by residue 210, a conclusion that is consistent with the 3A4 modeling data.

The residues that encompass SRS-4 are highly conserved between species and have been directly implicated in substrate specificity and metabolism (Szklarz and Halpert, 1997;Fukuda et al., 1993;Raag and Poulos, 1989). Of the four SRS-4 mutants tested, only T309A showed sulfoxidase activity comparable with that of the wild type. F304A and I301A showed decreases only in R-sulfoxide formation. Interestingly, F304A has previously shown greater progesterone hydroxylase activity in comparison to the wild type, suggesting that the mutation does not effect the inherent function of the enzyme, but more likely the binding of the particular substrate. This interpretation is supported by the prediction of a distance of only 4 Å between a docked progesterone molecule and the 304 residue (Szklarz and Halpert, 1997). Residue T309 also has been mapped to close proximity to CYP3A4 substrates and has also been implicated in the CYP catalytic cycle, a finding supported by the change in enzyme activity with the mutation of the aligned residue of rabbit 2E1 and rat 2A1 (T303; Fukuda et al., 1993). However, steroid 6β-hydroxylation (B ring of the steroid) and now RPR 106541 sulfoxidation (D ring substituent) have been shown to be unaltered by the threonine to alanine change at residue 309, thus highlighting the importance of model validation with metabolism data.

Steroid 6β-hydroxylation by CYP3A4 has also been shown to be altered by the mutation of residues I369, A370, and L373 (He et al., 1997). With RPR 106541 sulfoxidase activity, however, only the A370V mutation resulted in a dramatic change in the rate of metabolism and the enantiomeric ratio (0.23). The A370V substitution has been implicated in the stabilization of progesterone in the 16α binding orientation (D ring as opposed to the 6β or B ring orientation), and may explain why the A370V mutant had the highest rate of S-sulfoxide formation of any mutant (Szklarz and Halpert, 1997). The effect of alterations of SRS-6 residues on steroid hydroxylase activity for CYP2B forms has been shown to be dramatic. Substrate docking studies have shown that residues 478–480 map to SRS-6 and potentially interact with the substrates erythromycin and progesterone (Szklarz and Halpert, 1997). Consistent with the docked substrate model, substitution of a Phe for L479 selectively decreased R-sulfoxide formation. The fact that the ratio of 6β-OH and 16α-OH metabolites of progesterone and testosterone were not effected by this mutation illustrates the increased sensitivity of RPR 106541 to SRS-6 mutations.

In conclusion, we have shown that the use of homology modeling in combination with site-directed mutagenesis can add important information on CYP-450-substrate interactions to CYP reaction phenotyping studies. There are three reasons why CYP3A4 is an excellent candidate for studies that could potentially advance the descriptive phenotyping data of individual compounds to more predictive models of CYP3A4-substrate interactions. First, CYP3A forms are responsible for the majority of xenobiotic oxidations in humans, with CYP3A4 representing the largest percentage of CYP forms in liver and a significant barrier to bioavailability in the intestine. Second, CYP3A4 displays a relatively high homology (88%) to CYP3A5, a polymorphically expressed form whose levels can equal or exceed those of CYP3A4 in certain individuals. This amino acid sequence similarity could be exploited to narrow the search for differences in CYP3A4/3A5 substrate interactions and provide critical information on interindividual variability in the metabolism of compounds with overlapping CYP3A substrate specificity. Finally, there is a large literature base on the stimulation of CYP3A4-catalyzed reactions in liver microsomes and, more recently, in human hepatocytes (Maenpaa et al., 1998). The combination of homology modeling with site-directed mutagenesis may also help to differentiate a number of hypotheses proposed on the mechanism of CYP3A stimulation (Schwab et al., 1988; Shou et al., 1994; Ueng et al., 1997;Harlow and Halpert, 1998). The historical anomalies associated with many CYP3A4-catalyzed reactions represent a growing research area that can now be addressed by a variety of substrates and genetic engineering techniques.