Induction of human cytochrome P450 (P450) enzymes in vivo can result in significant clinical consequences primarily via reduced exposure leading to suboptimal efficacy of coadministered drugs. CYP3A4 has been recognized as the target for the most clinically significant induction-mediated drug-drug interactions. For example, the induction of CYP3A4 by some antibiotics leads to reduced ethinylestradiol levels from oral contraceptives, resulting in breakthrough pregnancies, and reduced cyclosporine levels can lead to organ rejection in transplant patients (Sinz et al., 2008). Therefore, the need to evaluate and minimize the potential for P450 induction-mediated drug-drug interactions is widely accepted within the pharmaceutical industry.

Induction of P450 by drugs and other xenobiotics occurs typically via activation of receptors that regulate transcription. The primary mechanism of CYP3A4 induction is via activation of the nuclear pregnane X receptor (PXR), a transcription factor that regulates the transcription of several P450 isoforms, UDP-glucuronosyl-transferases, and other drug-metabolizing enzymes and transporters (Hewitt et al., 2007b).

Because of distinct species differences in both the extent and pattern of ligand response (Dickins, 2004), in vitro human assays are used to predict clinical effects, and primary human hepatocytes in culture are the current system of choice (Hewitt et al., 2007a,b). However, limited availability of high-quality human hepatocytes and the interindividual variability observed in the response to inducers has encouraged the search for alternative assay systems for use in drug discovery. The use of large batches of precharacterized inducible cryopreserved human hepatocytes somewhat mitigates these issues. In addition, reporter gene assays for PXR (Wang and LeCluyse, 2003) lend themselves well to enhanced throughput screening for CYP3A4 induction liability, yet have several potential drawbacks, including that they offer a pseudoendpoint, only assay for one mechanism of CYP3A4 induction, and the reporter gene construct is typically expressed in cell lines with limited or no metabolic capability (Castell et al., 2006).

Several hepatic cell lines, including HepG2, BC2, and HepaRG (Vermeir et al., 2005; Castell et al., 2006; Kanebratt and Andersson, 2008b), and the immortalized Fa2N-4 cell line (Mills et al., 2004; Ripp et al., 2006; Youdim et al., 2007; Hariparsad et al., 2008; Kenny et al., 2008) have recently undergone preliminary evaluation as alternatives to primary human hepatocytes for CYP3A4 induction studies. Studies have demonstrated Fa2N-4 cells to be responsive to prototypical inducers of CYP1A2 and CYP3A4 (Mills et al., 2004; Kenny et al., 2008), and the potency (EC₅₀) and magnitude of effect (E_max) of CYP3A4 inducers in Fa2N-4 cells could be used to predict the in vivo induction response (Ripp et al., 2006). More recently, the expression of PXR and arylhydrocarbon receptor (AhR) were determined to be similar between Fa2N-4 cells and human hepatocytes; however, both constitutive androstane receptor (CAR) and several hepatic uptake transporters, including the organic anion-transporting polypeptides, were significantly lower in this cell line compared with primary cells (Hariparsad et al., 2008). This laboratory has recently confirmed that Fa2N-4 cells are a good surrogate for primary human hepatocytes when assessing AhR- and PXR-mediated CYP1A2 and CYP3A4 induction, respectively, but not CAR-mediated CYP2B6 induction (Kenny et al., 2008). HepaRG are newly developed human hepatoma cells that express several P450s (including CYP1A2, 2B6, 2C9, 2E1, and 3A4), PXR, AhR, and CAR at levels comparable with cultured primary human hepatocytes, unlike HepG2 cells (Aninat et al., 2006). In addition, HepaRG cells express conjugating drug-metabolizing enzymes and membrane transporters. Thus, they are a promising alternative to primary human hepatocytes not only for P450 induction studies but as an alternative model for liver metabolism and hepatic toxicity of xenobiotics (Aninat et al., 2006; Guillouzo et al., 2007; Kanebratt and Andersson, 2008a,b).

As previously reported by this laboratory, it is beneficial to determine both mRNA and activity in the assessment of induction because processes such as time-dependent P450 inhibition (TDI) can confound the interpretation of data based on activity alone (McGinnity et al., 2006; Kenny et al., 2008). To relate message to activity endpoints in this work, determination of P450 TDI was carried out for all compounds.

The present study describes a head-to-head evaluation of several promising in vitro approaches for studying CYP3A4 induction. Prototypic CYP3A4 inducers were tested in the PXR reporter gene assay, Fa2N-4 and HepaRG cells, and primary human hepatocytes, the current “gold standard” in vitro system, together with negative controls, with the aim of comparing data and assessing the applicability of the alternate in vitro test systems for use in drug discovery.

Materials and Methods

Chemicals and Reagents. Carbamazepine, 6-(4-chlorophenyl)imidazo[2,1-b][1,3]thiazole-5-carbaldehyde-O-(3,4-dichlorobenzyl)oxime (CITCO), clotrimazole, dexamethasone, mifepristone, β-naphthoflavone, nifedipine, omeprazole, phenobarbital, phenytoin, quinidine, rifampicin, rifapentine, reserpine, troglitazone, troleandomycin, verapamil, and NADPH, were purchased from Sigma-Aldrich (St. Louis, MO) and were of the highest grade available. Efavirenz, hyperforin, midazolam, 1′-hydroxymidazolam, paclitaxel, pioglitazone, rifabutin, ritonavir, and sulfinpyrazone were purchased from Sequoia Research Products Limited (Pangbourne, UK). Lansoprazole was synthesized at AstraZeneca. Collagen type I-coated 24-well plates, tissue culture-treated flasks and trypsin-EDTA, BD Hepatocyte culture medium, epidermal growth factor, 6β-hydroxytestosterone, and 6β-hydroxytestosterone-[D7] were obtained from BD Biosciences Discovery Labware (Bedford, MA). Glutamine, Fungizone, and TRIzol were purchased from Invitrogen (Carlsbad, CA). Acetonitrile, dimethyl sulfoxide (DMSO), ethanol, and formic acid were purchased from Thermo Fisher Scientific (Waltham, MA) and were of the highest grade available. Cryopreserved Fa2N-4 cells, Multi-Function Enhancing (MFE) plating media, and MFE support media with supplement A were obtained from XenoTech, LLC (Lenexa, KS). HepaRG cells were purchased as preseeded 24-well plates from Biopredic International (Rennes, France). DNase 1 RNase free kit and RNase free water were from Invitrogen (Paisley, UK). TaqMan One-Step RT-PCR Master Reaction Mix, MultiScribe Reverse Transcriptase, primers, and probes were obtained from Applied Biosystems (Foster City, CA). All other chemicals were of high-purity grade and purchased from either Sigma-Aldrich or Mallinckrodt Baker, Inc. (Phillipsburg, NJ).

Human Hepatocytes. Primary cultured fresh human hepatocytes (donors HH205 and HH215) (Table 1) were obtained from BD Biosciences Discovery Labware. Human hepatocytes plated in collagen I-coated 24-well plates were maintained in culture at least 48 h before the treatment in hepatocyte culture medium supplemented with 10 μg/l epidermal growth factor, 50 μg/ml gentamicin, 2 mM l-glutamine, and 0.75 μg/ml Fungizone. Cells were treated in triplicate with 0.08% DMSO vehicle or test compounds at concentrations as described in Table 2 for 72 h with medium change and replenishment every 24 h. Regular visual inspection of the hepatocyte cultures revealed a confluent monolayer with good cell morphology.

The CYP3A4 induction assay was performed essentially as described by Zhang et al. (2006). In brief, the induction was determined by measuring the 6β-hydroxytestosterone metabolite formation from model substrate testosterone in situ and measuring mRNA expression using TaqMan real-time RT-PCR as described below.

Determination of CYP3A4 TDI in Human Liver Microsomes. CYP3A4 TDI was assessed for all compounds in human liver microsomes (HLMs) using an adapted version of the single-time point, single concentration, automated method described by Atkinson et al. (2005). Each compound was tested at the top concentration incubated in the induction studies, with the exception of phenobarbital, which was tested at a final concentration of 500 μM to ensure solubility. In brief, a 30-min preincubation was performed, containing 1 mg/ml HLMs in 0.1 M phosphate-buffered saline, pH 7.4, 1 mM NADPH, and test compound [in 1% (v/v) DMSO]. A 20-fold dilution was performed in the presence of NADPH (1 mM) and midazolam (10 μM), and the second reaction was allowed to proceed for 15 min. Aliquots (50 μl) were quenched in ice-cold methanol (100 μl) and prepared for LC/MS/MS as described below. Projected IC₅₀ values were calculated as described previously (Atkinson et al., 2005).

Quenched samples were centrifuged (2000g; 15 min, 4°C), and supernatants (10 μl) were injected from a 96-well plate onto a Symmetry C18 (3.5 μm) column (2.1 × 100 mm; Waters, Milford, MA) maintained at 40°C. High-performance liquid chromatography separation was performed using Jasco PU-2085plus semi-micropumps coupled to a Jasco DG-2080-54 degasser, MX-2080-32 dynamic mixer, and a CO-2067plus column oven (Jasco, Tokyo, Japan). The mobile phase consisted of solvents A (0.1% formic acid in acetonitrile) and B (0.1% formic acid in water), using a linear gradient of 10% A (0–0.1 min), 54% A (0.1–3.0 min), 100% A (3.0–3.5 min), and 10% A (3.51 min) with a run time of 5.5 min and a flow rate of 0.5 ml/min. This was coupled to a triple quadruple Platinum Ultima (Waters) operating in electrospray ionization+ mode, with MassLynx version 4.1 running in multiple reaction monitoring mode for 1′-hydroxymidazolam (342.35 > 203.23, 31 V, 24 eV), which was quantified using an authentic standard.

PXR Reporter Gene Assay. Induction of CYP3A4 was assessed for all compounds in a PXR reporter gene assay as described by Persson et al. (2006). In brief, transiently transfected HepG2 cells were thawed and seeded at 2.33 million cells/ml in a 96-well plate. Test compounds were serial diluted [DMSO vehicle at 0.025% final (v/v)] and added to the cells to generate a dose response for an EC₅₀ determination. Cells were then left to incubate at 37°C with compound for 72 h. After 72 h, LucLite-lyophilized substrate was added to all 96 wells of the plate. Cells were lysed, and a signal was generated during a 10-min room temperature incubation. The 96-well plates were placed in a luminescence plate reader, and counts were measured. -Fold induction was calculated by comparing magnitude of response in treated wells relative to that observed in control (DMSO only) wells.

Fa2N-4 Cells. Induction of CYP3A4 was assessed for all compounds in Fa2N-4 cells, as described by Kenny et al. (2008). In brief, Fa2N-4 cells were grown on tissue culture-treated flasks (75 cm²), maintained in MFE support media with supplement A (1 μl/ml) at 37°C under an atmosphere of CO₂/O₂ (5:95%) at 95% relative humidity. Support media were changed every 2 days, and cells were passaged weekly as they reached confluence. For passaging, cells were trypsinized (5 ml/flask) for 15 min, harvested in MFE plating media (40 ml), centrifuged (120g for 5 min), and resuspended in fresh MFE plating media. Cells were then diluted to 0.2 × 10⁶ cells/ml and either reseeded in culture-treated flasks (∼2 × 10⁶ cells/flask) or seeded on collagen-coated 24-well plates (0.1 × 10⁶ cells/well; 0.5 ml/well). After 24 h, MFE plating media were replaced with MFE support media containing supplement A. All Fa2N-4 cells used in the experiments were between passages 2 and 8. Fa2N-4 cells were precultured for 2 days on 24-well plates in MFE support media containing supplement A. A range of six concentrations was prepared for each test compound, selected to allow full-induction dose-response curves to be generated. Test compounds in media (0.5 ml) were added to appropriate wells of a 24-well plate containing precultured Fa2N-4 cells, and this was replaced with freshly prepared solutions of media and test compound every 24 h, for up to 72 h.

HepaRG Cells. HepaRG cells were obtained as preseeded 24-well plates. Induction of CYP3A4 was assessed for all compounds in HepaRG cells as described by Kanebratt and Andersson (2008b). In brief, after a 48- to 72-h shipment, cells were allowed to recover at 37°C for 48 h in after-shipment medium (low percentage DMSO + serum) followed by an overnight recovery in differentiation media. After the cells had recovered in differentiation medium, they were maintained for 48 h in basal medium (no DMSO or serum) with daily changes of media. After 48 h of maintenance in basal medium, cells were treated daily with test compounds at varying concentrations for 48 h. The final concentration of DMSO in culture medium was 0.1% (v/v), and the vehicle control group received 0.1% DMSO. For the determination of CYP3A activities, HepaRG cells were incubated in situ with midazolam and analyzed by LC/MS/MS as described below. Quantification of CYP3A4 mRNA was carried out by one-step quantitative RT-PCR, and all mRNA were normalized to total RNA in sample as described below. The -fold induction of each P450 was calculated as described for the PXR reporter gene assay.

Determination of CYP3A Activity in Human Hepatocytes and Fa2N-4 and HepaRG Cells. CYP3A4 activity was assessed in human hepatocytes using testosterone 6β-hydroxylase activity using by LC/MS/MS analysis. After treatment, hepatocytes in culture were washed once with medium and then incubated with 200 μM testosterone at 37°C for 15 min. The reaction was stopped by removing 300 μl of incubation medium, followed by mixing with 150 μl of acetonitrile containing 0.1% formic acid and stable isotope-labeled 6β-hydroxytestosterone internal standard. The 6β-hydroxy metabolite formation was measured using a PE Sciex API4000 or API4000 Qtrap LC/MS/MS system equipped with PerkinElmer pumps and a CTC autosampler (PerkinElmer Life and Analytical Sciences, Waltham, MA).

Midazolam 1′-hydroxylation was used to assess CYP3A in Fa2N-4 and HepaRG cells as described by Kenny et al. (2008). In brief, at the end of the induction period, test compound and media were aspirated from the cells and replaced with prewarmed media (0.5 ml) containing midazolam at a final substrate concentration of 25 μM. Cells were then incubated on an orbital shaker for gentle agitation (∼30 rpm; 37°C). Aliquots (100 μl) of media were removed at 20 min, quenched in ice-cold methanol (100 μl), and stored at –30°C until analysis for 1′-hydroxymidazolam by LC/MS/MS as described above. Finally, residual media containing substrate was aspirated, cells were lysed in RLT buffer (600 μl; QIAGEN, Dorking, Surrey, UK) containing β-mercaptoethanol (10 μl/ml) and stored at –80°C until required for mRNA analysis.

CYP3A4 mRNA Analysis. The remaining human hepatocytes in each well after enzyme reaction were washed once with phosphate-buffered saline. A 1-ml aliquot of TRIzol was added to each well, and total RNA was extracted in accordance with the manufacturer's instruction. The content of total RNA was determined at 260 nm, and the purity of RNA was estimated using a ratio of 260/280 nm. The RT-PCR assay was performed with the Applied Biosystems Gene Expression kit and in a 25-μl aliquot of TaqMan One-Step RT-PCR Master Reaction Mix containing CYP3A4 and housekeeping gene β-actin forward and reverse primers, TaqMan probe, MultiScribe Reverse Transcriptase, and 50 ng of total RNA. The PCR amplification was performed, and the transcription was determined using an ABI 7300 Real-Time PCR System.

The methodology used for CYP3A4 mRNA analysis of Fa2N-4 and HepaRG cells was as described by Kenny et al. (2008). In brief, Fa2N-4 and HepaRG cell lysates (stored in RLT buffer on a culture plate) were defrosted on ice and mRNA isolated using the RNeasy 96 kit (QIAGEN) according to the manufacturer's instructions for isolation of total RNA from animal cells using vacuum technology, including an on-column DNase 1 digestion to minimize genomic DNA contamination (RNase-Free DNase Set; Invitrogen). Total RNA was quantified fluorometrically using a Quant-iT RiboGreen RNA reagent and assay kit (Invitrogen) according to the manufacturer's instructions. Purified mRNA samples (5 μl) were analyzed by one-step quantitative RT-PCR performed on an Mx3005P real-time cycler with MxPro software (Stratagene, La Jolla, CA) using QuantiTect Multiplex RT-PCR No ROX Master Mix Kit (QIAGEN) according to the manufacturer's instructions using optimized primer and probe concentrations. Relative quantification of gene expression level was determined by interpolation of threshold cycle values to a standard curve. This standard curve was generated from a dilution series of human hepatocyte total mRNA. Standard mRNA was pooled from human hepatocytes induced with 10 μM rifampicin, 50 μM omeprazole, and 2 mM phenobarbital for 72 h. Data were normalized to total RNA content, as determined in the RiboGreen assay.

Data Transformation, Statistical Analysis, and Curve Fitting. Unless otherwise specified, all results were expressed as the mean of triplicate determinations. The parameters E_max and EC₅₀ were determined from dose-response data, where E_max was the maximum observed induction at an optimum concentration, and EC₅₀ was the effective concentration that supports half-maximal induction. To estimate EC₅₀ values, dose-response data were fitted to a simple E_max model with Hill function according to the following equation: y = (E_max · x^γ)/(EC₅₀^γ + x^γ), using Origin 6.0 (OriginLab Corp., Northampton, MA). Similar to the approach taken by Ripp et al. (2006), data points from wells with some evident toxicity, based on visual inspection of cell morphology, were excluded from the fits, and the E_max was set at the observed E_max to prevent extrapolation of the curve fit beyond measured data. All CYP3A4 E_max data for CYP3A4 were normalized to the rifampicin-positive control as follows: E_max (percentage maximal rifampicin response) = (E_max test/E_max rifampicin) · 100.

Calculations. The catalytic activity for CYP3A4 was calculated using standard curves and reported as picomoles per minute per 10⁶ cells. The -fold induction was calculated as follows: (enzyme activity of test drug-treated cells)/(mean of enzyme activity of negative control). Percentage of positive control was calculated as follows: (mean of enzyme activity of test drug-treated cells – mean of enzyme activity of negative control)/(mean of enzyme activity of rifampicin-treated cells – mean of enzyme activity of negative control) × 100%. CYP3A4 mRNA -fold induction was calculated by measuring the change in the treated sample divided by the control. -Fold induction was calculated by comparing magnitude of response in treated wells relative to that observed in control (DMSO only) wells. A measure of the induction efficiency of compounds was defined as the efficacy ratio and calculated as follows: efficacy ratio = E_max/EC₅₀.

Results

Identification of CYP3A4 Time-Dependent Inhibitors in Human Liver Microsomes.Table 3 shows 14 compounds identified as time-dependent inhibitors of CYP3A4 using HLMs with the mean CYP3A4 IC₅₀ values determined as described under Materials and Methods. The CYP3A4 TDI assay correctly classified all the well established mechanism-based inhibitors of CYP3A4, such as troleandomycin (IC₅₀ 0.5 μM), mifepristone (3 μM), ritonavir (4 μM), and verapamil (43 μM), but also highlights compounds that, to our knowledge, have not been reported previously as time-dependent inhibitors of CYP3A4, such as nifedipine (projected IC₅₀ value of 300 μM).

Comparison of Activity and mRNA Endpoints of Induction. As can be seen from Fig. 1, the mRNA/activity ratio was between 1 and 10 in both donors for the majority of compounds. Very high mRNA/activity ratios (>10) observed in these experiments were consistent with the previously confirmed CYP3A4 TDI for compounds such as troleandomycin, ritonavir, and verapamil. It is interesting that quinidine also demonstrated a high mRNA/activity ratio, despite no evidence of CYP3A4 TDI.

Comparison ofE_max Values and EC₅₀ Estimates between HH205 and HH215 Using mRNA. The parameters E_max and EC₅₀ were determined from the dose-response data from both human hepatocyte donors (Table 4), where E_max was the maximum observed induction at an optimum concentration, and EC₅₀ was the effective concentration that supports half-maximal induction. To estimate EC₅₀ values, dose-response data from all the prototypic CYP3A4 inducers were fitted to a simple E_max model with Hill function as described under Materials and Methods. Rifampicin was an inducer of CYP3A4, with a mean EC₅₀ of 0.3 μM and an E_max of 55-fold using mRNA data from both donors (Table 4). A negative control for CYP3A induction β-naphthoflavone demonstrated, as expected, no inductive response in both donors. Figure 2A shows an excellent correlation between mRNA EC₅₀ estimates determined in both human hepatocyte donors. Figure 2B demonstrates a significant correlation between the maximum -fold induction between both donors, although a clear offset can be observed whereby the HH205 response was on average ∼2-fold higher than HH215. This interdonor variability in induction response can be normalized relative to the response for rifampicin (Fig. 2C) and was used throughout the rest of this analysis. Figure 2D shows the correlation of a measure of induction efficiency defined as the efficacy ratio (E_max/EC₅₀) between both donors.

Comparison ofE_max Values and EC₅₀ Estimates between PXR Reporter Gene Assay and Human Hepatocytes. The parameters E_max and EC₅₀ were determined from a dose response using eight concentrations in the PXR reporter gene assay. Table 5 summarizes the EC₅₀ and E_max data for the compound test set. Rifampicin was a potent activator of PXR, with an E_max of 19-fold and EC₅₀ of 0.9 μM. β-Naphthoflavone, a negative control, showed no evidence of PXR activation. EC₅₀ values ranged from the most potent PXR activator, rifabutin (EC₅₀, 0.1 μM), to the least, phenobarbital (EC₅₀, 100 μM). Several compounds demonstrated similar maximal response to rifampicin, including 6-(4-chlorophenyl)imidazo[2,1-b][1,3]thiazole-5-carbaldehyde O-3,4-dichlorobenzyl)oxime, omeprazole, rifapentine, and ritonavir. The estimates of E_max and EC₅₀ generated from the PXR reporter gene assay were compared with those generated in human hepatocytes. In contrast to E_max, where no significant correlation was observed, the EC₅₀ values from the reporter gene assay seem to be in reasonable agreement with those generated in human hepatocytes (Fig. 3).

Comparison ofE_max Values and EC₅₀ Estimates between Fa2N-4 Cells and Human Hepatocytes. The parameters E_max and EC₅₀ were determined from a dose response of six concentrations using Fa2N-4 cells. Rifampicin was shown to be an inducer of CYP3A4 in Fa2N-4 cells, with an mRNA EC₅₀ of 4.2 ± 1.5 μM and an E_max of 33 ± 11-fold. β-Naphthoflavone, a negative control for CYP3A4 induction, showed no evidence of CYP3A4 induction in Fa2N-4 cells. A bell-shaped dose-response curve was noted for rifapentine, and although rifapentine was clearly a potent CYP3A4 inducer in Fa2N-4 cells (EC₅₀, 3 μM; E_max, 27-fold), a decrease in induction was observed at the highest concentration of 20 μM. Bell-shaped dose-response curves were similarly observed for several other compounds (dexamethasone, paclitaxel, phenobarbital, pioglitazone, reserpine, ritonavir, sulfinpyrazone, and verapamil). The estimates of E_max and EC₅₀ generated from the Fa2N-4 cells seem to be in reasonable agreement with that generated in human hepatocytes (Fig. 4A); however, several of the compounds seemed to have ∼3-fold lower EC₅₀ in Fa2N-4 cells than determined in human hepatocytes. A notable exception was rifampicin, where the EC₅₀ was markedly higher in Fa2N-4 cells compared with human hepatocytes (4 versus 0.3 μM). Figure 4B demonstrates that Fa2N-4 cells demonstrate a weak but significant correlation with the mean E_max values from the human hepatocyte donors used in this study. Figure 4C shows a good correlation of the efficacy ratio (E_max/EC₅₀) between Fa2N-4 cells and human hepatocytes, albeit rifampicin was a notable outlier.

Comparison ofE_max Values and EC₅₀ Estimates between HepaRG Cells and Human Hepatocytes. The parameters E_max and EC₅₀ were determined from a dose response of six concentrations using HepaRG cells. Rifampicin was shown to be an inducer of CYP3A4 in HepaRG cells, with an mRNA mean EC₅₀ of 0.8 μM and an E_max of 83-fold from duplicate experiments. β-Naphthoflavone, a negative control, showed no evidence of CYP3A4 induction. The estimates of E_max and EC₅₀ generated from the HepaRG cells seemed to be in excellent agreement with those generated in human hepatocytes (Fig. 5, A and B), which translated to an excellent concordance between efficacy ratio (E_max/EC₅₀) determined using HepaRG cells and hepatocytes (Fig. 5C). Figure 6 compares CYP3A4 induction using PXR reporter gene assay, Fa2N-4 cell line, HepaRG cell line, and primary human hepatocytes in culture after exposure to rifampicin.

Discussion

Human hepatocytes are commonly used to evaluate P450 induction via an enzyme activity endpoint; however, TDI of P450 can confound data interpretation (McGinnity et al., 2006). In this work, over half of the compounds tested (14 of 24) were identified as time-dependent inhibitors of CYP3A4, as determined by the HLM assay (Table 1), including well characterized drugs such as nifedipine. The clinical consequences of such weak TDI are likely to be insignificant, yet this property may compromise the activity readout in cultured cells. Weak TDI of CYP3A4 may be relatively common, and compounds should be assessed for TDI before committing to resource-intensive P450 induction assays. Moreover, these data show the utility of mRNA as an endpoint for P450 induction; it affords a larger induction window than activity (Fig. 1) and should be a measure of the intrinsic induction potential of a compound in the absence of the confounding factor of TDI.

Because of the finite number of hepatocytes available, estimates of EC₅₀ and E_max for both human hepatocyte donors were carried out from dose-response curves with a limited number of concentrations (three). However, it is clear from the excellent correlation of EC₅₀ estimates as determined from HH205 and HH215 that such estimates are reasonably consistent. Moreover, comparison between EC₅₀ estimates from the two donors in this work and other laboratories shows an overall good concordance. For example, rifampicin (EC₅₀, 0.1 and 0.4 μM for HH205 and HH215, respectively, versus a literature range of 0.2–0.8 μM), troglitazone [0.8 μM (HH215) versus 0.3–5 μM], pioglitazone (9 and 7 μM versus 2–13 μM), phenytoin (18 and 12 μM versus 5–24 μM), carbamazepine [42 μM (HH205) versus 14–60 μM], dexamethasone (43 and 39 μM versus 40–50 μM), and phenobarbital [142 μM (HH205) versus 86–159 μM] (Ramachandran et al., 1999; LeCluyse et al., 2000; Sahi et al., 2000; Faucette et al., 2004). It should be noted that all EC₅₀ values quoted were apparent because no correction for nonspecific binding was determined in the media, and the proprietary nature of some media contents used in this analysis prevented any estimation of potential differences between all assay systems.

Figure 2B demonstrates a significant correlation between the maximum -fold induction between both donors, but it was demonstrably poorer versus the interdonor EC₅₀ determinations, and a clear offset can be observed, whereby the HH205 response was on average ∼2-fold higher than HH215. Interdonor variability can be reduced for some compounds, to some extent, by normalizing against the response to prototypic inducer rifampicin (Fig. 2C). Together, these data suggest that variability between donors comes predominantly from E_max and not EC₅₀ estimates, which is consistent with previous literature reports (Silva et al., 1998; Meunier et al., 2000; LeCluyse, 2001; Madan et al., 2003). Moreover, significant interindividual differences may be observed in the basal levels of P450 activity (and mRNA expression), which may lead to significant changes in the -fold induction observed between different donors because there seems to be a ceiling on the maximum inducible activity (LeCluyse et al., 2000). This phenomenon was also exemplified in the data from HH205 and HH215 (data not shown).

Because human hepatocytes are such a precious and finite resource, limiting data points in a dose-response curve can be a pragmatic compromise, but this, in turn, may obfuscate generating accurate estimates of E_max and EC₅₀ and so represents a further limitation of using primary cells. Because of this and the variable interdonor response to inducers, alternatives to primary human hepatocytes are required for use in drug discovery.

As expected, rifampicin was a potent activator of PXR in the reporter gene assay, with an E_max of 19-fold and an EC₅₀ of 0.9 μM, values similar to that previously reported (E_max, 32-fold; EC₅₀, 1.2 μM) (Luo et al., 2002; Sinz et al., 2006). It is encouraging that for the compounds tested, EC₅₀ values determined via the PXR assay appeared to correlate reasonably with those determined from human hepatocytes (Fig. 3A), but it was evident that the E_max values do not correlate with E_max from the human hepatocytes donors studied in this work (Fig. 3B). Luo et al. (2002) observed a relatively poor correlation between E_max values in PXR and human hepatocytes. This may be because of the involvement of more than one nuclear hormone receptor in CYP3A4 induction, the presence of other positive and negative feedback mechanisms, and the metabolic capability of primary cells. That notwithstanding, the PXR assay is a simple, rapid, robust, and enhanced throughput screen that may offer value in assessing PXR-mediated CYP3A4 induction within drug discovery programs, primarily to establish quantitative structure-activity relationships, assist in the design of compounds with reduced affinity to PXR, and select appropriate compounds for further testing in human hepatocytes or an appropriate cell line.

It has been established previously by this laboratory and others that Fa2N-4 cells are a reasonable surrogate for primary human hepatocytes when assessing AhR- and PXR-mediated CYP1A2 and CYP3A4 induction, respectively, but not CAR-mediated CYP2B6 induction (Mills et al., 2004; Ripp et al., 2006; Youdim et al., 2007; Hariparsad et al., 2008; Kenny et al., 2008). The CYP3A4 EC₅₀ of rifampicin in Fa2N-4 cells was 4 μM, similar to the values reported previously (Ripp et al., 2006; Hariparsad et al., 2008) and approximately 10× greater than that observed in primary human hepatocytes (0.3 μM). Therefore, the disparity for rifampicin may be due to specific differences in metabolism or drug transporter activity between the immortalized and primary cells (Hariparsad et al., 2008). However, this work demonstrates that for a range of compounds, CYP3A4 EC₅₀ values seem to rank reasonably well between Fa2N-4 and human hepatocytes. The rifampicin data seem anomalous compared with the majority of compounds, where EC₅₀ values determined from Fa2N-4 cells demonstrate somewhat lower EC₅₀ values (up to 10-fold) compared with those observed from human hepatocytes (Fig. 4A). Significant correlations between E_max and efficacy ratio between Fa2N-4 cells and human hepatocytes were also observed (Fig. 4, B and C). Moreover, Ripp et al. (2006) have demonstrated that relative induction scores for compounds, determined from E_max and EC₅₀ values from Fa2N-4 cells and unbound drug concentrations from clinical studies, were well correlated with in vivo decreases in AUC for coadministered CYP3A4 substrates. Overall, the Fa2N-4 data suggest this model could be used for routine screening during discovery for PXR-mediated CYP3A4 induction. The absence of functional hCAR also may allow these cells to be used for mechanistic experiments to investigate the PXR-CAR cross-talk observed for many CYP3A4 and 2B6 inducers (Faucette et al., 2006).

The observation of bell-shaped curves within the Fa2N-4 data were consistent with previous reports (Hariparsad et al., 2008). Approximately 40% of all compounds demonstrated some decrease in induction at the highest concentrations tested, perhaps suggesting Fa2N-4 cells were more sensitive to chemical insult relative to primary cells, where no bell-shaped curves were observed. It should be noted that bell-shaped curves were also observed in HepaRG (30% frequency) and HepG2 (PXR) (36%) cells. It was typically the same compounds that cause this phenomenon in more than one cell line.

HepaRG cells are derived from a hepatocellular carcinoma, which exhibits a differentiated hepatocyte-like morphology capable of expressing both phase I and II drug-metabolizing enzymes and membrane transporters normally found in the liver (Aninat et al., 2006; Guillouzo et al., 2007). In addition to rifampicin induction of CYP3A4, HepaRG cells also have been demonstrated to be responsive to prototypical CYP1A and 2B6 inducers (Aninat et al., 2006; Jossé et al., 2008; Kanebratt and Andersson, 2008b). In this work, we have shown, for several prototypic CYP3A4 inducers, E_max and EC₅₀ generated from the HepaRG cells seem to be in excellent agreement with those values generated in primary cells (Fig. 5); therefore, HepaRG seems to be a very promising surrogate for determining the CYP3A4 induction potential of test compounds and candidate drugs. It also has been suggested that HepaRG cells express CAR (Guillouzo et al., 2007), but further investigation is required to determine whether the CAR-mediated induction is present in this cell line; if demonstrated, HepaRG cells offer a clear advantage over Fa2N-4 for use in drug discovery.

Gold standard P450 induction data can be generated using primary human hepatocytes but an erratic, restricted supply, and interdonor variability issues may restrict the routine application of this assay system within early drug discovery. The PXR reporter gene assay can be considered to be an enhanced throughput assay for screening away from CYP3A4 induction liability, for known PXR-specific activators. Fa2N-4 has the advantage versus the PXR reporter gene assay of being able to assess CYP1A in addition to 3A4 induction. However, importantly, Fa2N-4 cells have been demonstrated to give no CYP2B6 response via CAR. The HepaRG cell line seems to be an excellent surrogate for CYP3A4 induction compared with primary cells and a valuable recent addition to the armory of in vitro tools for CYP3A4 induction screening in drug discovery, and this cell line warrants further investment.