Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Recent clinical experiences with several compounds have highlighted the importance of assessing the potential for inhibition (Honig et al., 1993; Ahmad and Wolfe, 1995) or induction of drug metabolism (Parkinson, 1996 and references therein) and has promoted heavy investment in the early prediction of the extent and route of candidate drug metabolism and pharmacokinetic parameters in humans. In addition, with the increasingly large number of compounds generated as a result of the emergence of combinatorial chemistry and high throughput screening techniques, it is becoming important to develop rapid and relatively high throughput in vitro drug metabolism models with which to assess and eradicate such unfavorable properties from a series of potential drug candidates.

Currently, a combined in vitro approach using both heterologous expression systems (eventually for all xenobiotic-metabolizing enzymes) and human hepatic tissue fractions is recommended as the most thorough route to developing safe drugs with appropriate metabolic properties (Parkinson, 1996). Identifying the specific cytochrome P-450 (CYP)¹ enzyme involved in the metabolism of drug candidates is a labor- and time-intensive process involving correlation analyses within a bank of human liver microsomes (HLM), chemical and antibody isoform-selective inhibition, and confirmation with expressed enzymes.

As confidence in the application of recombinant enzymes to such studies has increased, their use as first line approaches in human hepatic CYP inhibition screens for drug candidates (Remmel and Burchell, 1993; Crespi and Penman, 1997; Crespi et al., 1997; Moody et al., 1999) and for the prediction of HLM kinetics (Becquemont et al., 1998) has been proposed.

Many CYP isoforms have now been expressed in various heterologous cell lines including bacteria, yeast, cultured insect, and mammalian cells (Waterman et al., 1995; Gonzalez and Korzekwa, 1995). E. coli expression systems offer many advantages over the major commercially available materials including generally stable and easy to manipulate plasmid constructs; the expressed enzyme is easily replenished at a high yield and at relatively low cost. Furthermore, although a high level of expression may require some modification of the N-terminal amino acid sequence of the CYP protein (Larson et al., 1991; Waterman, 1993), recent data have indicated that functional expression can still be achieved with the major drug-metabolizing CYPs with properties similar to the human hepatic microsomal enzymes (Blake et al., 1996; Iwata et al., 1998). The availability of an in-house panel of characterized human hepatic CYP enzymes would obviate the reliance on expensive commercial supplies and provide local control of expression levels, supply, interbatch variability, and quality.

With any expressed enzyme system, it is extremely important to have confidence that the recombinant CYPs are surrogates for the human microsomal enzymes by comparative analysis of the kinetic properties of these enzymes. This characterization and validation could include an assessment of substrate and inhibitor selectivity together with comparative kinetic analyses, which represents a substantial undertaking. This process is facilitated markedly through the application of automation technology, which enables a comprehensive in-house characterization of enzyme activities and eliminates awkward interlaboratory comparisons (Moody et al., 1999).

This laboratory has recently developed fully automated inhibition screens for the major human hepatic CYPs (Moody et al., 1999). As part of an extensive collaboration between academia and major pharmaceutical companies, this paper describes the validation of the major drug-metabolizing human CYP isoforms CYP1A2, 2C9, 2C19, 2D6, and 3A4 expressed in E. coli. These enzymes have been characterized with respect to their kinetic properties and comparisons have been made with their counterparts in HLM as well as the most widely used commercially available recombinant CYP preparations and to the wider literature.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Chemicals. All chemicals and reagents used were of the highest available commercial grade. Ampicillin, diazepam, erythromycin, dextromethorphan, naproxen, ethoxyresorufin, resorufin, -aminolevulinic acid, glycerol, and -nicotinamide adenine dinucleotide phosphate, reduced form (-NADPH) were purchased from Sigma Chemical Co. (Poole, UK). Isopropylthio--D-galactoside was purchased from Gibco BRL (Grand Island, NY). Bactotryptone, yeast extract, and bactopeptone were purchased from Difco (Detroit, MI).

Construction and Expression of Plasmids in E. coli. The LINK consortium, a collaboration between UK-based academia and industry, provided stocks of transformed cells with human CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 individually coexpressed with human NADPH-P-450 reductase in E. coli. The strategy for expressing CYP1A2 (in strain JM109) and CYP2C9 (in strain DH5) was similar to that described for CYP2D6 (Pritchard et al., 1998) and involved coexpressing each CYP with NADPH-P-450 reductase from separate compatible plasmids using ampicillin and chloramphenicol as the respective selection markers. For optimal expression of CYP1A2 and CYP2C9 the first eight codons of each CYP cDNA were replaced with eight codons of the bovine pregnenolone and progesterone 17-hydroxylase sequence (hereafter referred to as 17) to generate the constructs 17-CYP1A2 and 17-CYP2C9 (Pritchard et al., 1998).

Commercial CYP Sources. Microsomes prepared from human B-lymphoblastoid cells coexpressing recombinant human NADPH-P-450 reductase and human CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 and microsomes from insect cells transfected with a baculovirus containing the cDNA for human CYP2C19 together with cytochrome b₅ (Supersomes) were purchased from Gentest Corp. (Woburn, MA). Similarly, microsomes prepared from insect cells transfected with a baculovirus expressing CYP2C19 and rabbit NADPH-P-450 reductase were purchased from PanVera Corp. (Madison, WI). Pooled HLM (batch numbers 217, 219, and 220) were supplied by IIAM (Leics, UK).

Manual Enzyme Assays. Ethoxyresorufin O-deethylation (EROD; Riley et al., 1995), naproxen O-demethylation (Rodrigues et al., 1996), dextromethorphan O-demethylation (Rodrigues et al., 1994) and erythromycin N-demethylation (Riley and Howbrook, 1998) assays were used as probe reactions for CYP1A2, CYP2C9, CYP2D6, and CYP3A4, respectively, and based on methods described previously. [¹⁴C]-diazepam N-demethylation has been developed as an assay for CYP2C19 activity in this laboratory (Moody et al., 1999).

Substrate Selectivity. The isoform selectivity of these probe substrate reactions was investigated in incubations with E. coli membranes expressing CYP1A2, CYP2C9, CYP2D6, and CYP3A4 (all at 50 pmol P-450/ml) in 0.1 M potassium phosphate buffer pH 7.4 in a final volume of 200 µl. Each radiometric assay included 0.05 to 0.1 µCi of [¹⁴C]naproxen, [¹⁴C]dextromethorphan, [¹⁴C]diazepam, or [¹⁴C]erythromycin. An appropriate amount of cold substrate was added to give final substrate concentrations of dextromethorphan (30 and 3 µM), diazepam (100 and 20 µM), naproxen (2 mM and 300 µM), and erythromycin (300 and 40 µM). Substrate concentrations were chosen to reflect K_m and V_max conditions.

Automated K_m and V_max Determination. K_m and V_max determination assays were performed by a robotic sample processor (RSP; Genesis RSP 150; Tecan, Reading, UK). All of the assays performed by the RSP were programmed by the user and are not default program-supplied with the hardware.

SPE. Fully automated positive pressure SPE was performed and samples counted using the liquid scintillation analyzer as described elsewhere (Moody et al., 1999).

View larger version (16K): [in this window] [in a new window]   Fig. 1.   Hanes plots of the five probe assays using different enzyme sources. The kinetic determination was carried out as described in Materials and Methods. The activity of the isoforms at different substrate concentrations for probe reactions. a, CYP1A2-EROD. b, CYP2C9-naproxen O-demethylation. c, CYP2C19-diazepam N-demethylation. d, CYP2D6-dextromethorphan O-demethylation. e, CYP3A4-erythromycin N-demethylation. Full kinetic plots were determined using E. coli membranes expressing each individual CYP (), B-lymphoblasts expressing CYP1A2, CYP2C9, CYP2D6, and CYP3A4 (), baculosomes expressing CYP2C19 (), and for the three specific probe reactions (CYP1A2, CYP2D6, and CYP3A4) HLM (). Data shown are from one experiment typical of at least three, performed in duplicate.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

E. coli Expression of CYP and NADPH P-450-Reductase. Expression levels of holo-CYP determined by CO-reduced difference spectroscopy were similar to those recorded by Pritchard et al. (1997, 1998). Typical levels of spectrally active CYP (coexpressed with NADPH-reductase) recovered in E. coli membrane fractions were approximately 60 pmol P-450/mg protein for CYP2C19 and between 300 and 500 pmol P-450/mg protein for CYP1A2, CYP2C9, CYP2D6, and CYP3A4. NADPH-P-450 reductase activities were typically between 100 and 600 nmol cytochrome c reduced · min¹ · mg¹ protein.

Catalytic Activities of Recombinant CYPs. The catalytic activities of the CYPs were determined with membrane fractions of E. coli transformed with plasmids carrying the individual CYP and NADPH P-450-reductase. Time and protein concentration linearity profiles of all five probe assays with the different enzyme sources were determined (data not shown). Based on these data, Table 1 summarizes the experimental conditions that were adopted routinely for all assays and enzyme sources. Under these restrictions, the uppermost CYP content and incubation time (up to 15 min) were utilized to increase assay sensitivity. However, because of dissimilar CYP expression levels between enzyme sources, the overall protein concentration was variable (see Table 1).

Substrate Selectivity. Table 2 demonstrates the importance of examining substrate selectivity at several substrate concentrations. EROD and erythromycin N-demethylation were found to be specific reactions for CYP1A2 and CYP3A4, respectively, with minor contributions from other isoforms. Similarly, at a substrate concentration at or near the K_m for CYP2D6 (3 µM), dextromethorphan O-demethylation was metabolized almost exclusively by CYP2D6 (CYP2C19 activity was some 12% of that catalyzed by CYP2D6), whereas oxidation by low-affinity CYP2C isoforms was more dominant at a substrate concentration of 30 µM. The O-demethylation of naproxen was catalyzed principally by CYP2C9 but also by CYP1A2 and CYP2C19 (at 2 mM), whereas at a substrate concentration at or near the K_m for CYP2C9 (300 µM) catalysis by CYP1A2 was predominant. Diazepam (100 µM) was N-demethylated by CYP2C19 and CYP3A4 under the conditions used with negligible catalysis from the other CYPs. At 20 µM diazepam, the relative rate of metabolism by CYP3A4 was much less (approximately 30% that of CYP2C19).

Kinetics of HLM and Recombinant Human CYPs. V_max, TN, and K_m values for the probe substrate reactions were calculated and intrinsic clearance (Cl_int) values estimated from V_max/K_m.

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

This study describes the characterization and validation of the five major drug-metabolizing human hepatic CYPs functionally coexpressed with human NADPH P-450-reductase in E. coli. This laboratory has recently developed fully automated inhibition screens for these enzymes using both HLM and recombinant human CYPs and has demonstrated that their inhibition characteristics are remarkably similar (Moody et al., 1999). This report now details an efficient investigation into the kinetic properties of CYP1A2, 2C9, 2C19, 2D6, and 3A4 expressed in E. coli using this automated technology and comparison of the data with results obtained for HLM, recombinant CYPs from the major commercial sources, and the wider literature.

For the last 20 to 30 years the predominant in vitro tool for drug metabolism studies has been HLM, which have supplied qualitative (e.g., identifying which CYP isoform metabolizes the compound of interest) and quantitative (e.g., predicted Cl_int) information. Recent improvements in transplant technology suggest there will no longer be an inexhaustible supply of good quality human liver for the increasing amount of drug metabolism studies required by the pharmaceutical industry. As the biochemistry of the CYP family of enzymes is more fully understood, increased emphasis on cloning, expression, and the subsequent use of these enzymes is commonplace. Although initial use of expressed enzymes in vitro was as confirmation of HLM data (Aoyama et al., 1990; Tassaneeyakul et al., 1992; Rodrigues et al., 1994), sufficient confidence has recently been gained for recombinant CYP isoforms to be proposed as a first line approach (Becquemont et al., 1998; Moody et al., 1999). A major advantage of using recombinant enzymes is the potential to have a relatively low cost, inexhaustible supply of enzymes. However, assurance is required that such recombinant enzymes are faithful surrogates for HLM CYPs.

In general, the apparent K_m values, V_max, and Cl_int estimates agreed very well for the different sources of recombinant enzymes. For isoform-specific substrates such as ethoxyresorufin, erythromycin, and at low concentrations, dextromethorphan, the apparent K_m values were also equivalent to HLM and available literature data (Tables 3 and 4), demonstrating that nonspecific binding appears to be consistent for the substrates studied. The specificity of each probe reaction at high and low substrate concentrations was also determined for the five major drug-metabolizing human CYPs expressed in E. coli. At a substrate concentration at or near the K_m for CYP1A2, EROD catalysis by the other major human hepatic CYP isoforms studied is negligible and this reaction is specific for CYP1A2. Indeed, under V_max conditions for CYP1A2, of the isoforms studied, only CYP2C19 catalyzed this reaction to a minor extent (approximately 10% of CYP1A2).

In HLM, naproxen O-demethylation was characterized by a two-enzyme model, although this may be an oversimplification. Indeed, Korzekwa et al. (1998) have defined the atypical kinetics observed with naproxen metabolism by CYP2C9 in much more detail and invoked simultaneous binding of more than one substrate molecule to the enzyme, which was not apparent from our more limited data. Nevertheless, using recombinant CYP2C9 in E. coli membranes, the apparent K_m and TN obtained compared well with the high-affinity HLM component and other expression systems (Table 3). Detailed kinetic analysis suggested that CYP1A2 may exhibit a lower K_m for naproxen O-demethylation than CYP2C9 but also a lower V_max, a result in agreement with Tracy et al. (1997). Thus naproxen O-demethylation is suitable as a probe reaction for analysis of recombinant CYP2C9 but, even at low concentrations, not as a specific marker for CYP2C9 activity in HLM. The kinetics of the specific CYP2C9 probe, tolbutamide 4-hydroxylation, were also shown to be comparable in E. coli membranes (TN 2.5 min¹ and K_m 179 µM) and B-lymphoblasts (TN 2.8 min¹ and K_m 75 µM) expressing CYP2C9 and agreed well with literature values (data not shown).

Interestingly, initial data obtained with the Gentest materials indicated that diazepam N-demethylation might be highly dependent on the presence of cytochrome b₅. Moreover, the apparent K_m and TN for the E. coli-expressed CYP2C19 originating from the one plasmid construct compared well with the commercial baculovirus expression system coexpressing rabbit reductase, which suggested a pivotal role of the substantially greater NADPH P-450-reductase levels in the former system. Indeed, in vitro activities of the CYP2C family have been reported previously to be highly dependent on the supply of electrons via reductase and b₅ (Richardson et al., 1995; Crespi and Miller, 1997; Chang et al., 1997; Venkatakrishnan et al., 1998). Diazepam N-demethylation, under V_max conditions for the high-affinity CYP2C19, is catalyzed significantly by CYP3A4. At a substrate concentration nearer the pharmacological levels (20 µM) metabolism by CYP3A4 was still apparent. As the relative abundance of CYP3A may be up to 20-fold higher than CYP2C19 in an average human liver (Shimada et al., 1994; Jung et al., 1997; Inoue et al., 1997), the contribution of CYP3A4 to diazepam N-demethylation in HLM may be significant depending on the relative levels of these two enzymes. In HLM, two apparent K_m and V_max values for this reaction were determined by fitting the data to a simple two-enzyme model confirming the complex, multienzyme catalysis of diazepam N-demethylation in HLM (Yasumori et al., 1993; Andersson et al., 1994; Zomorodi and Houston, 1996).

Dextromethorphan was O-demethylated by CYP2D6, CYP2C19, and CYP2C9 under V_max conditions for CYP2D6. However, at a substrate concentration at or near the K_m for CYP2D6 (3 µM) the rate of dextromethorphan O-demethylation by CYP2C19 was approximately one-tenth of CYP2D6 activity whereas the catalysis by CYP2C9 is negligible. Thus at low dextromethorphan concentrations, this reaction would be almost exclusively CYP2D6-dependent in HLM. However, the low-affinity CYP2C19 (Moody et al., 1999) and CYP2C9 may represent the residual dextromethorphan O-demethylation pathway in poor metabolizers (Kroemer and Eichelbaum, 1995; Mahgoub et al., 1977).

This report confirms that erythromycin N-demethylation is catalyzed almost exclusively by the CYP3A family (Cook et al., 1993). Indeed, at a substrate concentration of 40 µM (~K_m for CYP3A4), catalysis by the other human hepatic CYPs studied was negligible. Interestingly, erythromycin N-demethylase activity in E. coli membranes expressing CYP3A4 was reduced in TSE buffer (0.32 ± 0.04 min¹) compared with 0.1 M phosphate pH 7.4 (2.5 ± 0.1 min¹). Other reports have also determined CYP3A4 activities to be dependent on the buffer composition (Maenpaa et al., 1998) and yet this phenomenon appears to be substrate (Shet et al., 1993; Yamazaki et al., 1995) and expression system-specific (see Table 3). The selective effect on TN may indicate that the buffer composition alters the local membrane environment and electron flow via the interaction with NADPH P-450-reductase rather than effecting a conformational change of the CYP3A4 active site (Ingelman-Sundberg et al., 1996; Table 3). Indeed, the critical nature of this interaction was suggested further with data from a one-plasmid CYP3A4 system, which exhibited much higher levels of human NADPH P-450-reductase (1600 nmol · min¹ · mg¹). This recombinant system showed nonlinear product formation with respect to time for several substrates, perhaps as a result of excess reactive oxygen species production (data not shown).

In conclusion, this paper demonstrates the characterization and validation of the five major drug-metabolizing human hepatic CYPs coexpressed with human NADPH P-450-reductase in E. coli. These expression systems have been validated with respect to five probe assays using predominantly marketed drug substrates and expedited by automation. The kinetic parameters of these recombinant enzymes were similar to their HLM counterparts, where directly comparable, and thus they would appear to be faithful surrogates. In addition, the activities of the optimized E. coli-expressed CYPs are equal to or even greater than the most widely used commercial sources. The data also suggest a pivotal role for the levels of coexpressed human NADPH P-450-reductase expression, which may be isoform- and substrate-dependent. These data make available this panel of enzymes for more detailed biochemical and pharmacological investigations and enhanced throughput screening. This laboratory has already advocated these materials as a first line screen for early CYP inhibition analysis, and work to expand their applications in the prediction of the routes and rates of human oxidative metabolism is ongoing.