Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Cytochrome P450 enzymes, a superfamily of heme-thiolate proteins, are found in all living organisms and are involved in the biotransformation of a diverse range of xenobiotics and endobiotics. Human P450 isoforms, which are mainly expressed in the liver, play a central role in drug metabolism. Variations in individual metabolism often result in unexpected toxicities because drug clearance is affected by a range of factors, including genetic variation, enzyme induction (activation), and inhibition of drug metabolism. Therefore, characterization of the P450 enzyme family has been of unceasing interest for the prediction and identification of drug metabolism and drug-drug interactions for discovery, development, and clinical therapy (Gonzalez and Nebert, 1990; Daly, 1995; Nebert, 1997; Kleyn and Vesell, 1998; Evans and Relling, 1999).

CYP2D6 is one of the first of the well characterized phase I polymorphic drug-metabolizing enzymes and is involved in the oxidation of numerous drugs including antiarrhythmics, antihypertensives, B-blockers, opioids, antipsychotics, and tricyclic antidepressants (Idle and Smith, 1979; Gonzalez et al., 1988; Nebert, 1997). CYP2D6 also metabolizes certain neurotoxins, including 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine, a potent inducer of Parkinson's disease (Fonne-Pfister et al., 1987; Gilham et al., 1997).

The molecular basis of the CYP2D6 polymorphism has been studied intensely in recent years, and it is now believed that this gene exhibits more than 70 allelic variations (Marez et al., 1997). The extrapolation of in vitro data measured for the resulting proteins and decreased drug-metabolizing capacity has garnered a significant amount of interest recently (Johansson et al., 1994; Oscarson et al., 1997; Fukuda et al., 2000; Shimada et al., 2001; Tsuzuki et al., 2001; Zanger et al., 2001). Some individuals carry "null" copies of the CYP2D6 gene in which one or more nucleotide substitutions encode for a faulty message or truncated protein, unable to bind heme and therefore unable to produce recognizable P450 enzyme activity. Other 2D6 alleles contain point mutations resulting in one or more amino acid changes in the proteins compared with CYP2D6.1. This variety results in a range of phenotypes, from poor metabolizers with no 2D6 enzyme activity, as seen in up to 10% of Caucasians, to ultra rapid metabolizers and ranges in between (Daly, 1995; Raimundo et al., 2000; Zanger et al., 2001).

A correlation between the presence of 2D6 allelic isoforms *2, *10, and *17 and decreased in vivo capacity for marker reactions is well established (Sachse et al., 1997; Panserat et al., 1999; Zanger et al., 2001). Assignment of 2D6*2 activity is often confounded by the multiplicity of allelic variants containing the same mutations, promoter polymorphism, and gene duplication (Raimundo et al., 2000; Zanger et al., 2001). Modestly to severely decreased catalytic activities are also observed with most allelic isoforms using cDNA-transfected mammalian and yeast cell membranes (Johansson et al., 1994; Oscarson et al., 1997; Fukuda et al., 2000) and genotyped and phenotyped human liver microsomes (Shimada et al., 2001; Zanger et al., 2001). In some cases, specific ethnic populations appear to correspond with specific metabolizer ranges. For example, the CYP2D6*10 allele is present at a high frequency (0.408 to 0.495) in Asian populations and is associated with substantially decreased turnover of CYP2D6 substrates (Droll et al., 1998; Tateishi et al., 1999; Garcia-Barcelo et al., 2000; Teh et al., 2001; Zanger et al., 2001; this study). CYP2D6*17, common among African-Americans and black Africans at frequencies of 0.150 to 0.34, also correlates with a marked decrease in activity toward probe substrates (Aklillu et al., 1996; Masimirembwa et al., 1996; Leathart et al., 1998; Wennerholm et al., 1999; Wan et al., 2001). The CYP2D6*2 allele is present in Caucasian populations at frequencies ranging from 0.271 to 0.324 compared with 0.347 to 0.364 for CYP2D6*1.

Previously, we have shown that the purification of CYP2D6 from baculovirus-infected Trichoplusia ni is feasible and results in high-activity enzyme in a reconstituted system. What is lacking to date is a uniform biochemical examination of the intrinsic properties of each allelic variant. We intend to fill this gap via the expression of each individual isoform in an identical environment followed by a high degree of purification and quality control. In this manner, we can strictly control the ratios of cofactors and lipids used in reconstitution procedures and avoid many of the pitfalls associated with the use of human microsomes or in vitro expression systems. The primary goal of the present work then is the in vitro expression and purification of CYP2D6.1, 2D6.2, 2D6.10, and 2D6.17 allelic isoforms of human CYP2D6 followed by the determination of Michaelis-Menten kinetic parameters of each toward codeine, fluoxetine, and dextromethorphan. Our intent is to test the hypothesis that the functional consequences of allelic variation found in human CYP2D6 are substrate-dependent, indicating that the extrapolation of data from in vivo phenotyping and genotyping experiments with a single marker substrate cannot reliably be used to predict the effects toward other important xenobiotics.

Toward this end, we generated CYP2D6*2, CYP2D6*10, and CYP2D6*17 amino acid coding sequences from CYP2D6*1 cDNA and adapted a baculovirus-mediated insect cell system for the high-level expression of each. We purified these allelic isoforms and determined their molecular weights by mass spectrometry. Reaction kinetics were then followed in a reconstituted enzyme/lipid system for the formation of: dextrorphan (DXO) and 3-methoxymorphinan (MEM) from dextromethorphan (by O-demethylation and N-demethylation, respectively) for the formation of morphine by codeine O-demethylation and for the formation of norfluoxetine from fluoxetine. The large number of substrates affected by the 2D6 polymorphism and the possibility of substrate-dependent effects warrants the systematic study of the decreased catalytic activities of the proteins encoded by these high-frequency alleles toward each individual substrate. In this study, we compare the functional differences among CYP2D6.1, CYP2D6.2, CYP2D6.10, and CYP2D6.17 protein products toward three clinically important substrates and confirm the hypothesis that enzymatically active allelic variants of CYP2D6 indeed show substrate-dependent alterations in their metabolic capacities.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Chemicals. Dextromethorphan, dextrorphan, 3-methoxymorphinan, 3-hydroxymorphinan, fluoxetine, and norfluoxetine were purchased from Sigma/RBI (Natick, MA). Codeine, norcodeine, morphine, reduced NADPH, L--dilauroylphosphatidylcholine (DLPC), dithiothreitol, phenylmethylsulfonyl fluoride, Octyl-Sepharose CL-4B, trifluoroacetic acid (TFA), and 60% perchloric acid were purchased from Sigma-Aldrich (St. Louis, MO). DEAE-Sepharose Fast-Flow was from Pharmacia (Peapack, NJ), and ceramic hydroxyapatite was from Bio-Rad Laboratories (Hercules, CA). Emulgen 911 was a gift from Kao-Atlas (Tokyo, Japan). HPLC solvents and other chemicals were of the highest grade commercially available and were used as received.

Molecular Biology. Restriction enzymes were purchased from Roche Diagnostics (Indianapolis, IN), Invitrogen (Carlsbad, CA), or New England Biolabs (Beverly, MA) and were used in buffer systems provided by the manufacturers. High-five T. ni cells were obtained from Invitrogen. HyQCCM-SFX medium and fetal bovine serum were from Hyclone Laboratories (Logan, UT). General molecular biology methods were performed by standard procedures (Sambrook et al., 1989), and routine insect cell culture methods were followed as described by O'Reilly et al. (1994).

Protein Expression and Purification. In all cases, spectral P450 was determined by the method of Omura and Sato (1964) and was used as a means to establish optimal conditions for P450 expression. T. ni cells were maintained routinely on 100-mm culture dishes (Sarstedt, Newton NC) at 27 ± 1°C and subcultured every 2 to 3 days using HyQCCM-SFX medium (Hyclone Laboratories) supplemented with 8 to 10% fetal bovine serum. Penicillin-G (100 µg/ml), streptomycin sulfate (61 µg/ml), and amphotericin-B (0.6 µg/ml) were routinely added to the medium to reduce contamination by bacteria or fungi. Baculovirus stocks were amplified by subsequent "passages" (up to three as necessary to achieve suitably high P450 expression) of viral supernatant onto a new 100-mm dish of T. ni with excess (30 ml) medium followed by incubation for 7 to 8 days. Suspension cultures were inoculated directly from culture dishes and were grown in batches of 250 ml in 2-liter Erlenmeyer flasks using a spin bar on a magnetic stir platform. Vigorous stirring was used to reduce cell clumping and to ensure adequate oxygenation of the medium. Infection was carried out at a cell density of 0.8 to 1.6 × 10⁶ cells/ml using 1 to 10 ml of amplified viral supernatant. Freshly prepared hematin solution (1-5 mg/ml dissolved in 10 mM NH₄OH) was added 2 days postinfection at a 1:1000 dilution (final heme concentration = 1-5 µg/ml). Cells were pelleted 3 to 4 days postinfection, resuspended and washed once in glycerol-containing buffer (100 mM potassium phosphate, 20% glycerol, 0.33 mM dithiothreitol, 1 mM EDTA, pH 7.4), repelleted, and stored at 80°C until further use.

Determination of 2D6 Protein Molecular Weights. ESI/liquid chromatography mass spectrometry analyses (Koenigs et al., 1999) were performed on a Micromass Quattro II tandem quadrupole mass spectrometer (Micromass, Ltd., Manchester, UK) coupled to an HPLC (Shimadzu LC-10AD with SPD-10AV UV-vis variable detector; Shimadzu Scientific Instruments, Inc., Columbia, MD). The instrument was controlled by a computer running Windows NT based Micromass MassLynxNT 3.2 software. The source temperature was 150°C, with the cone voltage set to 55 kV. The mobile phase consisted of buffer B (0.05% TFA in water) and buffer C [0.05% TFA in acetonitrile and water mixture (950/50, v/v)]. A linear gradient elution of 35 to 100% buffer C from 0 to 12 min was used to separate the proteins. Solvent flow through the POROS R2 perfusion column (2.1 × 150 mm) from Perspective Biosystems (Cambridge, MA) was 0.2 ml/min with 100% of the flow (50 pmol of protein injected) being diverted to the mass spectrometer. CYP2D6 allelic isozymes eluted at approximately 9.70 min. Acquisition was carried out from m/z 500 to 2000 Da in the CONTINUUM scanning mode. ESI mass spectra were collected, the individual scans across the HPLC peak were combined, and each spectrum was deconvoluted using the MaxEnt program (Micromass).

Kinetic Studies of Purified Recombinant CYP2D6 Isoforms. Incubation reactions were carried out in 100 mM potassium phosphate buffer, pH 7.4, containing 0.1 µM CYP2D6, 0.2 µM P450 reductase, 10 µg (80 µM) DLPC, 1 mM NADPH, and substrate in a final volume of 200 µl. When fluoxetine was used as the substrate, the amount of each enzyme (P450 and reductase) was doubled. CYP2D6 and reductase were added together first and left to incubate at room temperature for 15 min. DLPC was then added for a further 15-min incubation period before the addition of buffer and substrate. Reactions were initiated by the addition of NADPH and terminated by the addition of 10 µl of 60% perchloric acid. The mixtures were subjected to centrifugation at 14,000g for 5 min before HPLC injection and analysis. Dextromethorphan concentrations ranged from 0 to 500 µM for the O-demethylation reaction and from 0 to 8000 µM for the N-demethylation reaction (according to a biphasic-kinetics model; Yu et al., 2001). The mixtures were incubated at 37°C for 5 min with CYP2D6.1 and 2, whereas a 15-min of incubation was required with CYP2D6.10 and 17. Codeine concentrations ranged from 0 to 3000 µM, and all incubations were carried out for 15 min at 37°C. Fluoxetine concentrations ranged from 0 to 100 µM, and all reaction mixtures were incubated at 37°C for 15 min. In an attempt to model the heterozygotic condition, two purified CYP2D6 allelic isoforms were mixed together (1:1, pmol/pmol) at room temperature before the addition of reductase. DXM concentrations used ranged from 0 to 2000 µM. All reactions were performed in duplicate.

Metabolite Analyses. HPLC analyses were carried out on a Waters Alliance system (Milford, MA) consisting of the 2690 separation module, the 2487 dual  absorbance detector, and the 474 scanning fluorescence detector. The Alliance HPLC system was controlled with Millennium32 software. A 250 × 4.6-mm i.d. hi-chrome phenyl column (Regis Technologies, Inc., Morton Grove, IL) was used to separate the metabolites. The flow rate through the column at ambient temperature was 1 ml/min. Analyses of dextromethorphan and its metabolites were performed as described previously (Yu and Haining, 2001). Detection limits for the O- and N-demethylated metabolites, DXO and MEM, respectively, were 5 pmol. Separations of fluoxetine and its metabolites were achieved with a mobile phase containing 60% buffer A (10 mM potassium phosphate in water, pH 3.5 adjusted with orthophosphoric acid) and 40% acetonitrile. The excitation and emission wavelengths of the fluorescence detector were set at 235 and 310 nm, respectively. Fluoxetine and its N-demethylated metabolite norfluoxetine eluted at 14.1 and 11.9 min, respectively, under these conditions. The detection limit for norfluoxetine was 10 pmol. A mobile phase consisting of 65% water (with 0.1% TFA) and 35% acetonitrile and water (400:600, v/v) was used to separate codeine and its metabolites norcodeine and morphine, which eluted at 15.8, 9.2, and 8.2 min, respectively, under these conditions. The excitation and emission wavelengths of the fluorescence detector were set at 280 and 335 nm, respectively, for the analysis of codeine and its metabolites. The detection limit for morphine was 5 pmol.

Data Analysis. Enzyme Michaelis-Menten parameters, K_m and V_max, and error estimates thereof were generated by nonlinear regression analysis (GraphPad Prizm 3.02; GraphPad Software, San Diego, CA). Initial estimates for nonlinear regression were generated graphically using Eadie-Hofstee plots (Vo versus Vo/[S]). Linear regression analyses were conducted using Microsoft Excel 2000 (Microsoft, Redmond, WA).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

cDNA Expression and Purification of CYP2D6 Allelic Variants. The creation of high titer baculovirus stocks, growth, and infection of T. ni, and heme addition proceeded in a manner analogous to that used for CYP2C9 (Haining et al., 1996). Cultures exhibiting greater than 50 nmol/l equivalents of P450 enzymes were used for purification. Holoprotein yields were estimated at each stage by measuring carbon monoxide difference spectra (Omura and Sato, 1964). Each CYP2D6 allelic variant behaved in a chromatographically similar manner throughout the purification procedure. Contaminants were removed by passage through Octyl-Sepharose and DEAE-Sepharose. Highly purified, detergent-free CYP2D6 was collected after dialysis of fractions from the hydroxyapatite column (Fig. 1). Typically, 100% estimated yield is achieved at the first step using cholate to extract P450 enzymes from cell membrane. The total final yields for CYP2D6.1, CYP2D6.2, CYP2D6.10, and CYP2D6.17 enzymes are shown in Table 1. The carbon monoxide difference spectra for CYP2D6.1, CYP2D6.2, and CYP2D6.10 isoforms exhibited Soret maxima at 450 nm, with no evidence of cytochrome P420 formation (Fig. 2). CYP2D6.17, however, showed strong absorbance at 420 nm after passing through the hydroxyapatite column, indicating the denaturation of some active enzyme, perhaps due to improper pH of the buffer used in this step or altered detergent susceptibility of this isoform.

View larger version (115K): [in this window] [in a new window]   Fig. 1.   Analysis of stages in the protein purification of CYP2D6 variants by SDS-polyacrylamide gel electrophoresis. Ten picomoles of cytochrome P450 was loaded into each lane. Proteins were separated on a resolving gel polymerized from 9% acrylamide and visualized by staining with Coomassie blue. Molecular weight markers in lanes 1 and 9 and (marked with arrowheads in lane 1) correspond in descending order to phosphorylase b (97.4 kDa), bovine serum albumin (68 kDa), ovalbumin (43 kDa), carbonic anhydrase (29 kDa), and lysozyme (14.3 kDa). Lane 2, membrane preparation from T. ni culture expressing CYP2D6.2. Lane 3, pooled fractions following an Octyl-Sepharose chromatography step during purification of CYP2D6.2. Lane 4, pooled fractions from a DEAE-Sepharose chromatography step (CYP2D6.2). Lane 5, final purified product following hydroxyapatite chromatography and dialysis (CYP2D6.2). Lane 6, final purified CYP2D6.1. Lane 7, final purified CYP2D6.17. Lane 8, final purified CYP2D6.10.

View larger version (23K): [in this window] [in a new window]   Fig. 2.   Carbon monoxide difference spectroscopy of purified CYP2D6.1, 2D6.2, 2D6.10, and 2D6.17 allelic variants. Spectra were determined by the method of Omura and Sato (1964).

Analysis of Purified CYP2D6 Allelic Isoforms. Samples throughout the purification procedure were analyzed using SDS-polyacrylamide gel electrophoresis (Fig. 1) and Western blot (not shown), with a selective anti-2D6 antibody. As expected, full-sized CYP2D6 protein products were detected in each sample. Equal amounts of each allelic variant were loaded per lane (10 pmol in Fig. 1) based on CO-difference spectroscopy. In each case, the band intensities are practically identical, indicating that the ratio of active enzyme to total protein for each variant is also identical. The isolated CYP2D6 protein isoforms were submitted to mass spectral analysis to check for gross abnormalities and compare the mass differences caused by the encoded amino acid changes. For example, phosphorylation would be expected to add some 80 mass units to a protein, a change easily within our detection limits and approximately 4-fold greater than the largest difference between predicted versus measured masses of our proteins (Table 2). The experimentally determined mol.wt. was calculated using the observed charge state distribution or ion envelope. The experimental mol.wt. for CYP2D6.1, CYP2D6.2, CYP2D6.10, and CYP2D6.17 compared with the mol.wt. values, predicted based on the amino acid sequences, are shown in Table 2.

Dextromethorphan O-Demethylation by CYP2D6 Allelic Isoforms. Dextromethorphan, an over-the-counter antitussive agent, is a widely used probe drug for polymorphic CYP2D6 activity both in vivo (Sachse et al., 1997) and in vitro (Yu et al., 2001; Yu and Haining, 2001). Kinetic analyses of DXM O-demethylation activities were carried out with purified CYP2D6 allelic isozymes (Fig. 3A). Eadie-Hofstee transformations of these data are shown in Fig. 4A; calculated K_m and V_max values are given in Table 3. Apparently, the intrinsic clearance values, as estimated by the V_max/K_m ratio, of DXM by CYP2D6.2, 0.10, and 0.17 decreased approximately 5-, 100-, and 10-fold, respectively, compared with CYP2D6.1.

View larger version (18K): [in this window] [in a new window]   Fig. 3.   Kinetic plots of product formation versus substrate concentration for metabolic turnover of dextromethorphan, codeine, and fluoxetine catalyzed by highly purified recombinant CYP2D6 isoforms in vitro. Equal amounts of purified CYP2D6.1, CYP2D6.2, CYP2D6.10, and CYP2D6.17 enzyme (estimated by the method of Omura and Sato, 1964) were individually reconstituted in a lipid environment with purified rat cytochrome P450 reductase and substrate. Reactions were initiated by the addition of NADPH. Shown are Michaelis-Menten curves for the formation of dextrorphan from dextromethorphan (A), morphine from codeine (B), and nor-fluoxetine from fluoxetine (C). All assays were conducted in duplicate. , CYP2D6.1; , CYP2D6.2; , CYP2D6.10; , CYP2D6.17.

View larger version (18K): [in this window] [in a new window]   Fig. 4.   Eadie-Hofstee plots for the O-demethylation of dextromethorphan to dextrorphan by purified CYP2D6 allelic isozymes in vitro. A, shows the transformation of data presented in Fig. 3A for individual enzymatic variants (, CYP2D6.1; , CYP2D6.2; , CYP2D6.10; , CYP2D6.17). B, shows the result of simple mixing of isoforms in 1:1 ratio [, CYP2D6(.1/.2); , CYP2D6(.1/.10); ×, CYP2D6(.2/.10); *, CYP2D6(.1/.17)]. Note the clear deviation from linearity in the latter. Kinetic constants derived from either a single-site enzyme model or two-site hyperbolic models (curve fit shown) are given in Table 4.

Dextromethorphan N-Demethylation by CYP2D6 Allelic Variants. Dextromethorphan is metabolized to 3-methoxymorphinan through N-demethylation as well as to DXO by CYP2D6.1 (Yu et al., 2001; Yu and Haining, 2001); thus both metabolites were monitored in reactions catalyzed by the CYP2D6.2, 10, and 17 proteins. Calculated K_m values for MEM formation were in the millimolar range for each isozyme, and V_max values were relatively high (Table 3). N-Demethylation activity catalyzed by CYP2D6.2, 10, and 17 isoforms continued to increase at higher substrate concentrations, far past the point where O-demethylation activity reached saturation (not shown) similar to the result obtained for the CYP2D6.1 enzyme (Yu et al., 2001) and very much resembling the situation presented by Korzekwa et al. (1998) in which two substrates (in this case both the same substrate) are thought to occupy the active site of some P450s simultaneously. Therefore, kinetic measurements of this reaction are a rough measure at best and should be used for comparison purposes only.

Codeine O-Demethylation and Fluoxetine N-Demethylation. Codeine and fluoxetine were also used as substrates to compare catalytic efficiencies of CYP2D6.1, 2, 10, and 17 allelic isoforms. Michaelis-Menten plots for each substrate are shown in Fig. 3, B and C, and estimated K_m values for codeine O-demethylation and fluoxetine N-demethylation by CYP2D6.1, 2, and 17 are shown in Table 3. The CYP2D6.10 isoform did not produce detectable O-demethylated metabolite from codeine (morphine) under the experimental conditions used. Thus, wild-type CYP2D6.1 enzyme was found to have the highest affinity and catalytic efficiency for codeine O-demethylation, followed by CYP2D6.2 and CYP2D6.17 isoforms. Apparent K_m and V_max values for fluoxetine N-demethylation by CYP2D6.1, 2, 10, and 17 allelic isoforms are also shown in Table 2. Enzyme efficiency decreased in the order of CYP2D6.1 > CYP2D6.2 > CYP2D6.17 > CYP2D6.10 consistent in rank order with the results obtained from DXM O-demethylation and codeine N-demethylation.

Modeled "Heterozygotes" of CYP2D6 Alleles. In a crude attempt to estimate the drug-metabolizing capacity of heterozygotes in vitro, a simple model was used in which two different CYP2D6 allelic isoforms were mixed in equal proportions (based on CO-difference spectra and hence presumably functional protein). Of course, the exact ratio of CYP2D6 allelic isoforms expressed in heterozygous subjects is unknown and undoubtedly highly variable. In this study, equal amounts of two allelic isoforms were combined together as rough in vitro models for CYP2D6*1/*2, *1/10, *1/17, and *2/*10 heterozygotes. As expected, biphasic kinetics are evident for DXM O-demethylation in all cases (Fig. 4B). If the data are fit to a one-enzyme Michaelis-Menten equation, apparent K_m values largely reflect the apparent affinity of CYP2D6.1 or CYP2D6.2 protein individually (Tables 3 and 4). Only in the case of 2D6.1/2D6.10 mixtures did the K_m of 2D6.1 not predominate over the lower activity (2D6.10) enzyme. Two-enzyme kinetic parameters were also derived from the data (Table 4). The V_max values for the in vitro-modeled heterozygous subjects were decreased more than expected based on simple arithmetic. Notably, the estimated intrinsic clearance values of these in vitro modeled heterozygous subjects were markedly decreased when compared with CYP2D6*1/*1 (Tables 3 and 4). Also note that 2D6.2/2D6.10 mixtures appear to retain more activity than 2D6.1/2D6.10 mixtures (Fig. 4b), a situation not predicted on the basis of individual enzyme activity (Table 3).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

In the present study, active human CYP2D6.1, 2, 10, and 17 proteins were over-expressed in insect (T. ni) cells followed by individual purification to a high degree of homogeneity. The authenticity of expressed mutant proteins is crucial to our final structure/function interpretations. Toward this end, manipulated DNA sequences were confirmed in their entirety before use, and as a final confirmation, we determined whole-protein molecular weights of our expressed variants by ESI/liquid chromatography mass spectrometry (Koenigs et al., 1999). This method is capable of revealing any gross abnormalities, such as truncated protein or the loss of heme, or any major post-translational modifications, such as phosphorylation or glycosylation.

Previous studies with CYP2D6 allelic variants (Johansson et al., 1994; Oscarson et al., 1997; Fukuda et al., 2000; Tsuzuki et al., 2001; Zanger et al., 2001; Marcucci et al., 2002) have generally used unpurified enzyme in membrane homogenates from cDNA-transfected mammalian (typically COS-1), yeast, bacterial, or insect cells. The use of purified enzyme, however, is superior in many ways to the use of such cell preparations. For example, consistency in expression level is difficult to achieve, and batch-to-batch variability is often large. Purification overcomes this variable and also allows us to eliminate concerns over the participation of anything other than the desired isoforms (such as P450 endogenous to the host cell) in the NADPH-dependent oxidations. Although the process involves the removal of protein from a membrane environment, subsequent reconstitution with a single lipid (dilauryl-phosphatidylcholine) eliminates the largely unknown effects of lipid composition in the membrane of a host organism. Also, the concentration of cofactors required for electron transfer can be stringently controlled, eliminating another large source of inter-laboratory variability.

Literature values for the kinetics of dextromethorphan O-demethylation provide an independent check of our methods. For example, using the extract of a yeast expression system containing CYP2D6.1 cDNA, Ching et al. (1995) measured a K_m of 5.4 µM and V_max of 0.47 nmol/nmol/min for this reaction, similar to the K_m value of 8.5 µM and V_max of 0.7 nmol/nmol/min reported by Krynetski et al. (1995). Our initial measurement for these parameters with purified 2D6 was a K_m of 6.2 µM and V_max of 14 nmol/nmol/min; we interpret this to mean that our product had the same approximate binding affinity but a much higher proportion of active enzyme to inactive protein. Subsequent refinement of our methodologies and kinetic measurements resulted in a steady decline in the apparent K_m and leveling off of the V_max such that a K_m value of 1.9 µM and V_max of 8.5 nmol/nmol/min were eventually reported using a relatively short (2 min) assay time (Yu et al., 2001). In the current study, to better compare less efficient isoforms, either a 5- or 15-min assay was used, resulting in the higher apparent K_m (DXM) of 3.0 ± 0.64 µM and V_max of 9.0 ± 0.5 nmol/nmol/min.

Most recently, Marcucci et al. (2002) compared the kinetics of metabolism by 2D6.1, 2D6.2, and 2D6.17 when expressed in two different heterologous systems, using insect and COS-7 cell lines. Interestingly, they did not find any substantial changes in kinetic parameters between 2D6.1 and 2D6.2 using either system for dextromethorphan, bufuralol, or debrisoquine. This lies in stark contrast to our findings in which the intrinsic clearance (V/K) of 2D6.2 toward dextromethorphan, fluoxetine, and codeine decreased to levels between 20 and 35% that of 2D6.1. In their hands, only with the addition of the T107I substitution does the V/K drop to some 18 to 22% of that of 2D6.1. In our hands, the addition of the T107I mutation of 2D6.17 further decreased the V/K for each to between 6 and 15% of that measured for 2D6.1. The reason for the discrepancy between laboratories is unknown at this time. Variability in the reported kinetics of dextromethorphan O-demethylation may be explained in part by the low enzyme expression levels achieved in most in vitro systems and the resulting longer reaction times required to measure metabolite formation. Another common pitfall lies in the use of artificially high substrate concentrations, at which point the unusual biphasic kinetics exhibited for this reaction as catalyzed by CYP2D6 (Yu et al., 2001; Korzekwa et al., 1998) and exhibited by allelic variants (not shown) can only aggravate interpretation.

Fluoxetine has a different structure than the fused heterocyclic ring structures of either dextromethorphan or codeine; therefore, it is an attractive candidate to test the hypothesis that changes in kinetics caused by allelic variation will be substrate-dependent. Presumably the secondary amine portion of this substrate cannot interact with an active site aspartic acid and be simultaneously oxidized at the N-methyl moiety. In addition, the highly electronegative character of the trifluoromethyl group may serve as a Lewis acid donor for D301, thereby orienting the electron-dense nitrogen atom toward the heme for oxidation by heme oxy-iron. Previously, Margolis et al. (2000) examined fluoxetine metabolism by recombinant CYP2D6.1. Using racemic fluoxetine, they estimate a K_m of 2.2 µM, a value very similar to ours (Table 3). Their V_max of 6.32 pmol/pmol of P450/min is very nearly 10-fold higher than ours. The reason for this discrepancy is not known. Overall, our estimated intrinsic clearances for fluoxetine N-demethylation decreased approximately 4-fold when comparing 2D6.2 versus 2D6.1 and another 2-fold upon the addition of the T107I substitution (2D6.17; Table 3). Thus, the substrate-dependent effect of this substitution appears to be confirmed.

Other reports shed further light on the effect of the T107I substitution, encoded only in the 2D6*17 allele (Masimirembwa et al., 1996; Oscarson et al., 1997). Expression of variant and wild-type cDNAs in vitro revealed that the CYP2D6.17 enzyme had only 20% of the wild-type activity toward bufuralol, but the Thr107Ile substitution on its own had no significant effect on bufuralol metabolism (Oscarson et al., 1997). Both T107I and R296C were required to raise the K_m for this substrate, but the S486T change appeared unimportant alone or in combination. When codeine was used as a substrate, however, these authors found that the T107I mutation alone was sufficient to increase the K_m. In the current study, we wished to see if the observations by Oscarson et al. (1997) regarding codeine hold true for the similarly structured dextromethorphan. From Table 3, however, in comparing CYP2D6.2 with CYP2D6.17, it appears that the addition of the T107I substitution reduces the estimated intrinsic clearance (V/K) of the enzyme 6-fold when codeine is used as substrate, whereas the V/K for the O-demethylation of dextromethorphan decreased only 2-fold.

The in vitro levels of expression and the ease in which heme is incorporated into apoprotein to create active enzyme, as well as the purification process itself, may provide important clues to the inherent stability of 2D6 isoforms in vivo. In general, the final purification yields of each isoform mirror the expression levels seen in cell culture (Table 1). Notably, 2D6.17 was unusually sensitive to inactivation during exposure to the hydroxyapatite matrix during purification, resulting in a loss of P450 spectral binding and a concomitant increase in the peak at 420 nM (Omura and Sato, 1964). This peak is often attributed to inactive protein with heme bound improperly, probably lacking the thiolate ligand. It is consistent with observations made in a cell culture in which protein expression levels (as indicated by immunoblot and 420-nm absorbance) may be quite high while P450 content as estimated by CO-difference spectroscopy may be quite low.

Notable among our findings is that 2D6.2 protein does not appear to possess any unusual stability or drastic instability, confirming that the increase in 2D6 enzyme activity associated with certain alleles encoding the same two amino acid changes is not due to inherent protein stability. The Arg296Cys change is remarkable considering the distinct alteration in bulk and charge. According to existing models, this residue is probably located early in the I-helix, which forms the backbone of all P450s, and passes directly over the heme pocket to become one of the most determinative of the substrate recognition sequences (Gotoh, 1992). Although position 296 resides in the N-terminal end of the I-helix, it is near enough (approximately two turns of an a-helix) to Asp301 such that it is possible that Arg296 normally serves to neutralize the charge of this carboxylic acid in the absence of cationic substrate. Conversion of the imidazole side chain to a thiol raises the possibility of disrupting any such interaction and/or instead introducing an intra- or interchain disulfide bond. Interestingly, mutation of this residue to a serine, rather than cysteine, resulted in a nonfunctional protein in preliminary experiments (R. Haining, unpublished observation).

Another important finding from the current work is that CYP2D6.10, although severely impaired in the overall turnover of all substrates examined (V/K reduced 50-fold or more; Table 3), was nonetheless relatively stable during purification (Table 1) and catalyzed the formation of 3-methoxymorphinan from dextromethorphan with kinetics similar to 2D6.1. Why might the DXM to dextrorphan pathway be decreased 100-fold, whereas the DXM to 3-methoxymorphinan pathway is apparently decreased only 2-fold? Since 2D6.2 and 2D6.17 both also contain the Ser486Thr change found in 2D6.10 yet do not exhibit the same level of discrimination against the O-demethylation pathway of DXM metabolism (Table 2), the Pro34Ser change found in CYP2D6.10 must instead be implicated. The location of this known helix-breaking residue suggests that it may act as a hinge between the predicted membrane-spanning N-terminal -helix and the soluble portion of the molecule comprising the bulk of the protein, a structure widely accepted for mammalian P450s. One article (Yamazaki et al., 1993) examined the importance of conserved proline residues in this region for the proper folding and membrane-insertion of rat CYP2C11. They found that the replacement of Pro30, which corresponds to Pro34 of CYP2D6.1, to alanine resulted in a membrane-associated protein with complete lack of observable heme binding. Based on our expression and purification data, the analogous substitution of 2D6.10 appears to impair proper folding and heme binding but not prevent it.

Although CYP2D6.10 is capable of carrying out enzymatic oxidations, its low expression in T. ni and humans may indicate a difficulty in membrane association. The replacement of this residue by Ser may hinder the formation of a kink and thus retard the proper folding of residues downstream. Indeed, the content of CYP2D6 in human liver microsomes is about 3-fold lower in CYP2D6*1/*10 heterozygotes and CYP2D6*10/*10 homozygotes than CYP2D6*1/*1 homozygotes (Shimada et al., 2001). The CYP2D6.10 allelic isozyme is also poorly expressed in COS-1 cells compared with CYP2D6.1 (Johansson et al., 1994). Published kinetic analyses using bufuralol, codeine, or venlafaxine as a substrate indicate that CYP2D6.10 (Fukuda et al., 2000; Shimada et al., 2001) exhibits a higher K_m value, yielding a lower V_max/K_m ratio than the CYP2D6.1 isoform. One recent study following (+)-bufuralol 1'-hydroxylation, however, showed that CYP2D6.10 exhibits only slight K_m changes, but marked V_max changes, compared with the CYP2D6.1 isoform (Zanger et al., 2001).

Another consideration regarding 2D6.10 and the P34S change lies in substrate access. It seems likely, based on their hydrophobic nature, that P450 substrates might access the active-site oxy-ferrous species directly from the lipid aliphatic side-chain portion of the membrane environment, thus providing an explanation for the specific lipid requirement for P450 enzyme activity. In this model, Pro34 may form part of a direct membrane-to-protein access channel that is disrupted in the 2D6.10 protein; the lowered activity of this variant could reflect a disruption in this substrate access channel. The loss of peptide-bond restraint imposed by the proline and/or the introduction of a hydroxyl-containing amino acid such as serine may discourage hydrophobic substrates from reaching the oxy-heme moiety. An interesting phenomenon observed with the allelic variants is that the DXM N-demethylated product continues to be formed at higher substrate concentrations (not shown) far beyond the point of saturation for the O-demethylation reaction (Fig. 3) similar to the phenomenon observed with the CYP2D6.1 isoform (Yu et al., 2001). This observation raises the possibility that this reaction is achieved through an alternate binding orientation or active site. In the membrane access model when the membrane access route is full or simply at high enough concentrations, substrate could be forced in through the normal product egress pathway, presumably directly from the aqueous phase.

Finally, a crude in vitro representation of the phenotype resulting from heterozygotic carriers of functional 2D6 enzymes was created by simply mixing highly purified allelic isoforms in a 1:1 ratio (based on CO-difference spectroscopy) during the lipid reconstitution step. The content of wild-type CYP2D6.1 isoforms in these in vitro modeled heterozygotes is half than that in modeled CYP2D6*1/*1 homozygotes. Based on simple arithmetic, one might expect to see at least 50% of the homozygotic activity in heterozygotes. Modeled 2D6*1/2D6*10 heterozygotes, however, appeared to have only 25% or less of the activity seen in modeled CYP2D6*1/*1 homozygote, as reflected in the decreased apparent V_max for each reaction mixture and increased K_m (Tables 2 and 3; Fig. 4). If instead, by an alternate explanation, the enzyme is active in a dimeric form rather than a monomeric form, it is tempting to speculate that the mixing of two isoforms, one of which that is partially denatured, causes a decrease in the turnover of its otherwise normal partner. This possibility will require further investigation.

In summary, the overall pattern of effect on the enzymatic efficiency of 2D6 isoforms examined in the present study toward three substrates was the same. Estimated intrinsic clearance values for the 2D6 isozymes examined in the present study decreased in the order CYP2D6.1 > CYP2D6.2 > CYP2D6.17 > CYP2D6.10. Thus, individuals containing one or more variant alleles are predicted to eliminate dextromethorphan by favoring the N-demethylation pathway (as catalyzed by CYP3A4; Yu and Haining, 2001) and have lessened analgesic effect from codeine. The effect on the efficacy of fluoxetine therapy is less straightforward due to the active nature of its major metabolite, norfluoxetine, and requires further study. The substrate dependence of changes in enzyme activity is confirmed, and thus, phenotypic markers for impairment in 2D6 function should be used with caution.