Abstract
The increase in cytochrome P450 (P450) enzyme activity noted upon exposure to therapeutics can elicit marked drug-drug interactions (DDIs) that may ultimately result in poor clinical outcome or adverse drug effects. As such, in vitro model systems that can rapidly and accurately determine whether potential therapeutics activate the human pregnane X receptor (PXR) and thus induce CYP3A P450 levels are highly sought after tools for drug discovery. To that end, we assessed whether DPX2 cells, a HepG2-derived cell line stably integrated with a PXR expression vector plus a luciferase reporter, could detect agents that not only cause PXR activation/CYP3A induction but also elicit clinical DDIs. All 20 clinical inducers and 9 of 15 clinical noninducers examined activated PXR in DPX2 cells (Emax > 8-fold), although activation parameters obtained with the noninducers were not predictive of DDI. The relative induction score, calculated by combining PXR activation parameters (EC50 and Emax) in DPX2 cells for seven inducers plus four noninducers with their efficacious total plasma concentrations, strongly correlated (R2 = 0.90) with the magnitude of induction of midazolam clearance. Thus, the DPX cell-based PXR activation system is not only capable of distinguishing potential inducers in a high-throughput manner but can also differentiate among compounds in predicting the magnitude of induction-mediated DDIs, providing a means for structure-activity relationship screening during discovery and development.
Introduction
It is well known that administration of specific therapeutics to patients can perturb the metabolism of other drugs and xenobiotics. Such alterations in oxidative metabolism are one principal cause of drug-drug interaction (DDI), a phenomenon that can lead to poor therapeutic outcome or toxicity. The stimulation of drug metabolism observed upon treatment with clinical agents stems primarily from induced hepatic levels of CYP3A4, the human cytochrome P450 (P450) enzyme of greatest prevalence with the most wide-ranging substrate specificity. Much progress has been made toward understanding the mechanisms that regulate expression of CYP3A4, other P450 enzymes, and drug transporters, which are all important determinants of drug clearance. Indeed, the orphan nuclear receptors human pregnane X receptor (hPXR) and human constitutive androstane receptor (hCAR) are now recognized as key regulators of not only CYP3A, CYP2B, and CYP2C gene expression but also phase II enzymes, efflux ATP-binding cassette (ABC) transporters, and uptake solute carrier (SLC) transporters (Faucette et al., 2006; Konno et al., 2008).
The induction by prospective therapeutics of CYP3A P450s, namely CYP3A4, is intensely studied during drug discovery and development. CYP3A enzyme induction by a new molecular entity (NME) can be measured using several in vitro model systems, including human hepatocytes and modified cell lines. Primary cultures of human hepatocytes have been recommended as the preferred system for assessing induction of P450 expression by therapeutics undergoing development (U.S. Food and Drug Administration Draft Guidance for Industry: Drug Interaction Studies—Study Design, Data Analysis, and Implications for Dosing and Labeling, http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM292362.pdf, 2012). In addition, human hepatocytes have been used to predict whether concurrent administration of NMEs with a known CYP3A4 substrate might lead to altered drug disposition, efficacy, or toxicity (Sinz et al., 2006; Fahmi et al., 2010). Promising alternatives for assessing CYP3A4 induction by potential drugs are the cell-based transactivation assays, which employ HepG2 cells transfected either transiently or stably with a pregnane X receptor (PXR) expression vector plus a luciferase reporter gene containing CYP3A4 response elements (proximal promoter and distal xenobiotic-responsive enhancer module). DPX2 cells, which are stable hPXR transfectants, show CYP3A enzyme induction (measured as increased luciferase activity) by therapeutics that is comparable to that reported with human hepatocytes (Krueger et al., 2002; Trubetskoy et al., 2005; Raucy and Lasker, 2010). Despite lacking certain phenotypic properties inherent to human hepatocytes, such as constitutive expression of P450 and nuclear receptor proteins, DPX2 cells nevertheless respond to typical CYP3A inducers [e.g., rifampicin (RIF)] with a marked increase in CYP3A metabolic activity (Trubetskoy et al., 2005), eliminate screening dependence on an uninterrupted supply of human hepatocytes, and exhibit much lower interexperimental variability. Despite that human hepatocytes are generally considered the “preferred system” for predicting clinical inducers, their use in high-throughput screening and structure-activity relationship (SAR) studies during early discovery is limited because of the associated costs and the variability between hepatocyte lots. DPX2 cells can readily surmount such issues (Shukla et al., 2011).
The goals of this study were to further characterize the molecular properties of DPX2 cells and to assess their utility as an in vitro paradigm for examining CYP3A induction using established clinical inducers and noninducers in a high-throughput fashion. These therapeutic agents were also used in DPX2 cells to derive relative inductions scores (RISs), an approach used previously for predicting the magnitude of clinical DDIs elicited by CYP3A induction in hepatocytes (Fahmi et al., 2008a).
Materials and Methods
Materials.
DPX2 cells that had undergone 15 to 17 passages in culture were used. The cells are available from Puracyp, Inc., as part of a hPXR activation assay kit (www.puracyp.com). The therapeutics shown in Table 1 were obtained from Sigma-Aldrich (St. Louis, MO), Sequoia Research Products Ltd. (Pangbourne, UK), and Pfizer Global Research and Development (Groton, CT). RNeasy and plasmid purification kits were purchased from QIAGEN (Valencia, CA), RETROscript was from Ambion (Austin, TX), and High Capacity cDNA Reverse Transcription Kit plus custom TaqMan low-density arrays were obtained from Applied Biosystems (Foster City, CA). CellTiter-Fluor cell viability assay, ONE-Glo luciferase assay reagent, and P450-Glo CYP3A4 Screening System with luciferin-isopropyl acetal (IPA) were from Promega (Madison, WI). All other reagents were of the highest quality available.
Analysis of Drug-Response Gene Expression in DPX2 Cells.
DPX2 cells and parental HepG2 cells were grown in 100-mm Petri dishes containing Dulbecco's modified Eagle's medium supplemented with 10% serum in a humidified 37°C/5% CO2 incubator. The cells were treated for 24 h with 10 μM RIF, 150 μM carbamazepine (CBZ), or vehicle [(0.1% dimethyl sulfoxide (DMSO)] alone. Total RNA was extracted from harvested cells using QIAGEN RNeasy Mini Kits (QIAGEN), and quantification of the specific drug-response mRNAs shown in Fig. 1 was performed using the TaqMan two-step reverse transcriptase-polymerase chain reaction (PCR) method (Fahmi et al., 2008b; Fahmi et al., 2010) using primer/probe sets (Supplemental Table S1) with TaqMan low-density arrays (Applied Biosystems) (Richert et al., 2009). Relative quantities of specific target genes versus those of the reference “housekeeping” control gene (glyceraldehyde-3-phosphate dehydrogenase/glucose 6-phosphate dehydrogenase) were determined with the ΔΔCT method.
Source of Clinical Data.
Thirty-four therapeutic agents were chosen for this study on the basis of the available data from 60 clinical studies (Table 1). Clinical DDI data, including changes in areas under the curve (AUCs) of object drugs and precipitant drug plasma concentrations, were derived from the University of Washington Metabolism and Transport Drug Interaction database (www.druginteractioninfo.org). The 20 clinical inducers and 15 clinical noninducers used are listed in Tables 2 and 3, respectively. [For the purposes of this study, modafinil was classified as both a clinical inducer and a noninducer because of its reported effects on triazolam and 17α-ethinyl estradiol (EE) pharmacokinetics, respectively (Robertson et al, 2002).]
PXR Activation in DPX2 Cells.
Construction of stable DPX2 transformants from the parental HepG2 line has been described previously (Raucy et al., 2002). DPX2 cells were grown in a humidified 37°C/5% CO2 incubator in Dulbecco's modified Eagle's medium containing 10% serum. For the hPXR transactivation assay, DPX2 cells were seeded into 96-well clear-bottom plates at a density of 2 × 104 cells/well and were maintained for 24 h at the same conditions to allow for cell recovery. After plate removal from the incubator, fresh medium containing the various therapeutic agents shown in Table 1 was added. Six concentrations of each agent, prepared in DMSO, were included, with the following exceptions (vehicle given in parenthesis): 17α-ethinyl estradiol (ethanol); gatifloxacin (0.1 N HCl/DMSO, 1:1); hyperforin (ethanol/DMSO, 1:1); leflunomide and terbinafine-HCl (methanol); pioglitazone (dimethylformamide); and nafcillin and ranitidine (water). The final concentration of vehicle ranged from 0.1 to 0.5%. RIF served as the exemplary positive control for PXR activation. Upon agent addition, plates were maintained in a humidified 37°C/5% CO2 incubator for another 24 h. Cell viability was determined with CellTiter-Fluor using fluorescent excitation and emission wavelengths of 400 and 510 nm, respectively, collected with a BioTek Synergy 2 multimode microplate reader (BioTek Instruments, Winooski, VT). The viable cell number per well was used to normalize results [average relative luminescence unit (RLU) divided by average fluorescence light unit (FLU)] obtained in the PXR activation assay upon determination of receptor activation by measuring luciferase activity with ONE-Glo reagent (Promega). PXR activation was expressed by dividing the average RLU of triplicate determinations at each of the six to eight compound doses by the average RLU of triplicate determinations obtained with the corresponding vehicle control. As expressed, PXR activation of the CYP3A4 promoter is directly proportional to luciferase activity and, thus, is a measure of CYP3A4 transcriptional activation. Emax and EC50 denote the maximum-fold PXR activation and ligand concentration associated with half-maximal PXR activation, respectively. EC50 and Emax values were derived only for those therapeutics that clearly elicited PXR activation in DPX2 cells by fitting dose-response data to a sigmoid three-parameter function (Sigma Plot; Systat Software, Inc., San Jose, CA) according to the equation f = a/{1 + exp[−(x − x0)/b]}, where a, b, and x0 denote Emax, slope, and EC50, respectively. RIS was calculated using the following equation: where I denotes the published efficacious plasma concentration of an activator/inducer achieved after a standard therapeutic dose (Hewitt et al., 2007; Fahmi et al., 2008a).
CYP3A-Mediated Metabolism of Luciferin-IPA in DPX2 Cells.
Metabolism of luciferin-IPA (Meisenheimer et al., 2011) was determined in DPX2 cells as described previously (Trubetskoy et al., 2005) except that a multiplexed assay format was used. After exposure for 48 h to RIF, CBZ, hyperforin, pioglitazone, leflunomide, or omeprazole, media was replaced with 50 μl of 3 μM luciferin-IPA. DPX2 cells were then maintained in a humidified 37°C/5% CO2 incubator for another 60 min. The contents of each well was removed to a replica plate to which P450-Glo luciferin detection reagent (50 μl/well) was added, and luminescence derived from the reaction product, d-luciferin, was measured. Cell viability and PXR activation were then determined with the original plate as described under PXR Activation in DPX2 Cells, with the viable cell number per well used to normalize luciferin-IPA metabolism results. EC50 and Emax values for the test agents were derived using nonlinear regression of typical log dose-response curves with Sigma Plot software (Systat Software, Inc.).
Results
Expression of Drug-Response Genes in DPX2 Cells.
DPX2 cells were treated with the established inducers RIF and CBZ, after which the content of mRNAs encoding specific drug-response genes (P450s, drug transporters, and nuclear receptors) was measured and compared with native HepG2 cells. As shown in Fig. 1, treatment with RIF or the analogous inducer CBZ markedly enhanced CYP3A4, CYP3A5, and CYP3A7 mRNA expression and caused a more modest increase in CYP2B6 expression. In contrast, RIF and CBZ had negligible effects on CYP2C8, CYP2C9, CYP2C19, CYP2D6, and CYP1A2 mRNA levels. Treatment of DPX2 cells with β-naphthoflavone (20 μM for 24 h) enhanced CYP1A2 transcript levels nearly 30-fold, indicating that the P450 induction process remained specific in DPX2 cells (data not shown).
Of the uptake SLC transporter genes examined in DPX2 cells, organic anion transporters (OATs) OAT2 and OATP2B1 were expressed at appreciable levels, whereas OATP1B1, OATP1B3, breast cancer resistance protein, and bile salt export pump mRNA expression was much lower, regardless of treatment. The efflux ABC transporters multidrug resistance 1 (MDR1), multidrug resistance-associated proteins (MRPs) MRP2, MRP3, and MRP4 were also found in DPX2 cells, although levels of these transporters remained unaffected by treatment with RIF or CBZ (Fig. 1). Unlike PXR, the constitutive expression of mRNAs encoding for Na/taurocholate cotransporting polypeptide, organic cation transporter 1 (OCT1), and in particular, constitutive androstane receptor (CAR) was low in DPX2 cells, and the levels of these three gene transcripts were largely uninfluenced by either of the drug treatments.
Activation of PXR in DPX2 Cells: Utility for DDI Predictions.
The in vivo DDI data from the University of Washington Metabolism and Transport Drug Interaction database used herein was derived using several CYP3A drug substrates, including MDZ, simvastatin, nifedipine, triazolam, and EE. The reported changes in these object drug AUCs (i.e., observed DDI) upon coadministration of 34 precipitant drugs are described in Table 1. Of these precipitant drugs, 20 were classified as CYP3A-inducing agents because of the considerable decrease (>20–97%) elicited in object drug AUCs. A decrease in AUC greater than 20% upon P450-inducer administration was considered clinically significant.
We initially verified the reproducibility of the DPX cell-based reporter gene assay by assessing PXR activation by 10 μM RIF on different days over a 12-month period. The magnitude of PXR activation in DPX2 cells by this RIF dose was 12.4 ± 1.2-fold (mean ± S.D., n = 24), with minimal and maximal activation of 10.7-fold and 15.3-fold, respectively (data not shown). Such results indicate that the cell-based PXR transactivation assay is subject to only modest interday variation, and that transfection of DPX2 cells with the human PXR expression vector is stable over numerous cell passages.
We then examined the ability of DPX2 cells to accurately identify clinically used CYP3A-inducing agents using the known clinical inducers and noninducers listed in Table 1. Each of the 20 agents with known in vivo inducing properties gave significant PXR activation, with estimated Emax values ranging from 3.5- to 13.3-fold (Table 2). Expressing PXR activation as a percentage relative to that observed with the positive control RIF (10 μM) demonstrated that using ≥15% as a cutoff, as recommended previously (Sinz et al., 2006), would give false negatives with 5 of the 20 clinical CYP3A inducers. Of the 15 clinical noninducers examined, 6 gave ≥2-fold activation in DPX2 cells, whereas only nifedipine elicited more than 15% PXR activation relative to RIF (Table 3). All 20 inducers exhibited dose-response effects, allowing estimates of EC50 and Emax values. Among the noninducers, dose-response effects for PXR activation were noted with nifedipine, nitrendipine, roxithromycin, leflunomide, omeprazole, and rosiglitazone.
PXR activation by clinical inducers in DPX2 cells is characterized by two kinetic parameters associated with receptor-ligand binding, namely Emax and EC50. These parameters allow for calculation of RIS, which gauges the ability of DPX2 cells to accurately predict clinical DDI. RIS values for 28 of the therapeutics examined here are shown in Tables 2 and 3, together with the parameters used to derive induction scores. The correlation between RIS and DDI (Fig. 2) was assessed by plotting RIS values versus the percentage decrease in AUC of coadministered MDZ, a widely used CYP3A4 substrate, calculated using published data from clinical DDI trials with eight of the clinical inducers (avasimibe, CBZ, phenytoin, pioglitazone, rifampin, troglitazone, pleconaril, and terbinafine) and three of the clinical noninducers (nifedipine, roxithromycin, and nitrendipine). A strong correlation (R2 = 0.90) was found between RIS and % decrease in MDZ AUC, indicating that assessment of PXR activation in DPX2 cells can serve as an accurate predictor of clinical DDI. The RIS value for each precipitant drug was plotted versus the observed DDI (percentage change in AUC), and the data were then fitted to a three-parameter Hill function using the formula y = (Emax · xγ)/(RIS50γ + xγ), where x = total Cmax in vivo, Emax = 95.42, RIS50 = 1.73, and γ = 5.15. Because DPX2 cells are stably transfected with human PXR, and because PXR activation by RIF remains consistent over numerous cell passages, the magnitude of PXR activation by NMEs can be predicted by deriving the RIS from the associated kinetic parameters determined in vitro, and then using the calibration curve to calculate the predicted percentage change in midazolam (MDZ) AUC.
Expansion of the RIS correlation plot with MDZ to other coadministered object drugs, including EE, triazolam, simvastatin, and nifedipine, gave results with varying degrees of accuracy (Tables 2 and 3). The magnitude of DDI predicted by DPX2 cells was an overestimate in the case of bosentan, rifabutin, sulfinpyrazole, and efavirenz when based on RIS established with MDZ (Table 2). No false negatives were observed, and RIS values obtained with DPX2 cells were not predictive of substantial DDI for the 15 clinical noninducers.
Analysis of CYP3A Metabolic Activity in DPX2 Cells.
The drug-metabolizing capacity of DPX2 cells was assessed by measuring the conversion of luciferin-IPA to d-luciferin, a bioluminescent reaction specific for CYP3A P450s (Meisenheimer et al., 2011). DPX2 cells treated with vehicle (0.1% DMSO) for 48 h exhibited modest luciferin-IPA metabolism (1.38 × 103 RLU/60 min), whereas cells treated with 20 μM RIF for the same time period displayed a 26-fold increase in luciferin-IPA metabolism rates [34.98 × 103 RLU/60 min (n = 3); data not shown]. In fact, the magnitude of increase in luciferin-IPA metabolism (Emax = 28.6-fold; EC50 = 1.73 μM) elicited by RIF exceeded the magnitude of PXR activation (Emax 14.4-fold; EC50 = 1.88 μM) obtained with this nuclear receptor ligand (Fig. 3A). With CBZ, the observed increase in luciferin-IPA metabolism paralleled that of PXR activation, exhibiting an Emax of 7.4-fold and an EC50 of 83.5 μM CBZ versus an Emax of 7.4-fold and an EC50 of 137 μM CBZ, respectively (Fig. 3B). Such RIF- and CBZ-mediated increases in luciferin-IPA metabolism likely stem from the enhanced expression of CYP3A4 enzyme protein in DPX2 cells, although catalysis of the reaction by CYP3A5 and/or CYP3A7 cannot be ruled out (Meisenheimer et al., 2011). Indeed, although the substrate specificity of CYP3A5 and CYP3A7 is not as broad as that of CYP3A4, the former P450s are still capable of metabolizing numerous therapeutics (Williams et al., 2002; Neunzig et al., 2011). In the cases of leflunomide (Fig. 3C) and omeprazole (Fig. 3D), neither compound gave substantial induction of luciferin-IPA metabolism (Emax of 2.0- and 4.6-fold, respectively) despite eliciting marked PXR activation (Emax = 15.0- and 25.5-fold, respectively). The latter phenomenon likely stems from the ability of these two clinical agents to serve not only as CYP3A inducers but also as inhibitors of these enzymes (Karam et al., 1996; Hosomi et al., 2011).
Discussion
The potential of NMEs to elicit DDI is attracting more attention during development because these interactions can profoundly influence patient safety, drug coadministration, dosing regimens, and overall therapeutic marketability. Curtailing DDI requires knowledge of the clinical implications associated with alterations in drug-metabolizing enzyme and ABC transporter expression. The capacity of P450 enzyme induction to elicit consequential DDI makes it essential to identify and predict, at an early stage, whether a new drug candidate can promote this phenomenon. Human hepatocyte cultures represent the most common approach for assessing P450 enzyme induction by NMEs (Hewitt et al., 2007; Fahmi and Ripp, 2010), although newer methodologies with higher throughput and less variability have become available. One such system is the DPX2 cell-based reporter gene assay incorporating hPXR (Raucy et al., 2002). The capacity of DPX2 cells to identify agents that elicit CYP3A enzyme induction in vivo upon PXR binding and activation has already been established (Raucy and Lasker, 2010). The aim of this work was to explore whether hPXR activation data obtained with DPX2 cells could serve to not only distinguish hPXR ligands but also to predict clinically relevant DDI, thereby permitting SAR analyses during early discovery. Indeed, as shown herein, the DPX2 cell-based transactivation assay can rapidly identify potential therapeutics capable of eliciting such DDI.
Cellular P450 enzyme levels are controlled primarily at the transcriptional level, with PXR serving as the predominant regulator of CYP3A gene expression (Gibson et al., 2002) and, together with CAR or hepatocyte nuclear factor 4α, CYP2B6 and CYP2C gene expression as well (Rana et al., 2010). DPX2 cell characterization revealed that except for CYP2B6, these HepG2 derivatives lacked constitutive expression of drug-metabolizing P450 mRNAs (Fig. 1A) (Yoshitomi et al., 2001; Wang et al., 2003). Treatment of DPX2 cells with the potent inducer RIF resulted in 20-fold to 150-fold increases in CYP3A4, CYP3A5, and CYP3A7 transcripts, modest enhancements of CYP2B6 mRNA, and negligible effects on CYP2C, CYP2D6, or CYP1A2 transcripts (Fig. 1A). CBZ, an antiepileptic with inducing properties in patients (Oscarson et al., 2006), gave a pattern of P450 induction in DPX2 cells that is highly similar to that with RIF. This RIF- and CBZ-mediated enhancement of CYP3A mRNA content likely stems from the overexpressed hPXR levels found in DPX2 cells (Raucy et al., 2002). PXR expression remained constant in the cells, regardless of drug treatment (Fig. 1B), whereas CAR was found at very low levels. The inability of RIF and CBZ to elevate CYP2C gene expression in DPX2 cells may stem from the lack of required transcription factors besides PXR (Rana et al., 2010). DPX2 cells also expressed appreciable levels of key efflux ABC transporters (e.g., MDR1) and uptake SLC transporters (e.g., OCT1) (Fig. 1B), although bile salt export pump, Na/taurocholate cotransporting polypeptide, OATP1B1, OATP1B3, and breast cancer-resistance protein are not detected. Except for OAT2-encoding mRNA, drug treatment had little effect on transporter expression (Fig. 1B). Although RIF treatment elevates MDR1 levels in human hepatocytes and HepaRG cells (Maglich et al., 2002; Madan et al., 2003; Mills et al., 2004), this is not the case in cell types (HepG2, lymphocytes, and LS174T cells) that lack CAR (Cerveny et al., 2007; Manceau et al., 2010).
We used the DPX2 cell-based transactivation assay to demonstrate PXR involvement in CYP3A induction by known clinical inducers such as CBZ, phenytoin, troglitazone, phenobarbital, and RIF (Table 2). The results obtained with DPX2 cells were mostly analogous to those obtained with human hepatocytes (Fahmi et al., 2010). Although 9 of 15, or 60%, of the clinical noninducers examined also activated PXR/CYP3A in DPX2 cells (Table 3), these same agents give induction of CYP3A4 mRNA in hepatocytes and other model systems (Fahmi et al., 2010). Among the noninducers, rosiglitazone promoted PXR activation in vitro only at concentrations exceeding 25 μM, which far surpasses the Cmax of 1.4 μM in vivo (Fahmi et al., 2008a), whereas omeprazole activated PXR at levels (>10 μM) that exceeded the Cmax of 0.68 μM (Soons et al., 1992). Despite eliciting substantial PXR activation (Emax >9-fold), neither leflunomide nor omeprazole increased CYP3A-mediated luciferin-IPA metabolism in DPX2 cells (Fig. 3, C and D). The capacity of these and other CYP3A-inducing therapeutics to also inhibit CYP3A catalysis (Karam et al., 1996; Ko et al., 1997; Hosomi et al., 2011) may explain why such agents fail to cause significant DDI. One compound studied here, namely modafinil, could be classified as either a noninducer when given together with EE (Table 3) or an inducer when coadministered with triazolam (Table 2). The fact that modafinil functions as a clinical inducer in vivo despite weak hPXR activation could stem from this agent's capacity to induce CYP3A enzymes in vivo via activation of CAR or its failure to enter DPX2 cells because of a possible lack of the appropriate uptake transporters.
Sinz et al. (2006) compared percentage of PXR activation by NME at their therapeutic concentrations (i.e., Cmax) with that of 10 μM RIF and defined NME that caused >40%, 15 to 40%, and <15% PXR activation at Cmax as “likely to elicit,” “may elicit,” and “not anticipated to elicit” DDI, respectively. However, we found here that such cutoffs may not be suitable for DDI predictions. For example, phenytoin exhibited an EC50 of 53 μM for PXR activation but gave only 13% of the RIF response at its Cmax of 24 μM. Likewise, pleconaril, pioglitazone, and phenobarbital also exhibited <15% PXR activation relative to RIF at their Cmax. Although the extent of PXR activation/CYP3A induction by a given NME is often expressed relative to a positive control (e.g., RIF), we now suggest a different approach. That approach uses the estimated Emax or EC50 for a NME, generated in DPX2 cells, together with the expected or observed Cmax for the same NME to predict DDI via the RIS model (Fahmi and Ripp, 2010; Fahmi et al., 2010).
Figure 2 illustrates how in vitro data may be used to predict inducer drug effects on object drug AUC. Three parameters are required, namely EC50, Emax, and [I], the latter of which denotes inducer concentration in vivo. Although estimates of [I] are controversial because of the inability to measure them directly, such estimates can be derived from either the total systemic or the unbound systemic steady-state Cmax concentration. Other obstacles to predicting DDI from in vitro studies include the lack of accurate data linking in vitro drug concentrations to those found at the hepatic site of action, as well as the contribution of protein binding or nonspecific binding. Use of total systemic drug concentrations in constructing the RIS plots gave significant improvements in DDI correlations (R2 = 0.90) obtained with DPX2 cells compared with using the corresponding unbound concentrations (R2 = 0.44; data not shown). Such improvements may be explained by the 10% serum protein used in DPX2 cell-based assay media. Although this approach fails to account for protein binding effects on drug disposition, it represents a conservative estimate using high (total) drug concentrations after the first-pass effect.
A strong correlation (R2 = 0.90) was found between RIS and the percentage decrease in MDZ AUC (Fig. 2). When combined, the parameters obtained for the 20 clinical inducers predict that these therapeutics would elicit CYP3A-mediated DDI, whereas the 7 clinical noninducers would not, a prediction borne out by clinical observations. Overall, NMEs with associated RIS values greater than 1 are likely to cause clinically relevant DDI. This approach is valuable in that once the validation has been established, there is no need to reestablish RIS parameters because the source remains constant. Expansion of the RIS correlation model to nifedipine, simvastatin, triazolam, and EE gave a less robust correlation (R2 = 0.31; data not shown) as well as DDI predictions with variable accuracies (Tables 2 and 3). DDI magnitude was significantly overpredicted with bosentan, rifabutin, sulfinpyrazole, and efavirenz and was inconsistent with their associated clinical DDI values of 31 to 41%. However, these latter DDI values were derived from studies using object drugs [EE, (R)-warfarin, and atorvastatin] metabolized by P450 enzymes in addition to CYP3A4, thereby complicating DDI prediction. Atorvastatin disposition may also entail key uptake transporters, further obscuring DDI predictions based on CYP3A induction. Most importantly, no false negatives were obtained among the 34 therapeutics analyzed in the DPX2 cell-based transactivation assay, and PXR activation data obtained with the 15 clinical noninducing agents failed to yield DDI predictions of consequence.
In conclusion, our results indicate that the DPX2 cell-based reporter gene assay represents a sensitive and accurate tool for identifying clinical inducers of CYP3A enzymes. This assay system not only is capable of distinguishing potential inducers in a high-throughput manner but also can differentiate among highly related compounds in predicting the magnitude of DDI that stems from induction, thereby providing a means for SAR screening during discovery and initial development. Inclusion of the DPX2 model together with other static or dynamic approaches of DDI prediction (e.g., hepatocytes) used during later stage development can further enhance our overall capacity to derive safer and more efficacious therapeutics.
Authorship Contributions
Participated in research design: Fahmi, Raucy, and Lasker.
Conducted experiments: Ponce and Hassanali.
Contributed new reagents or analytic tools: Fahmi.
Performed data analysis: Fahmi, Raucy, and Lasker.
Wrote or contributed to the writing of the manuscript: Fahmi, Raucy, and Lasker.
Footnotes
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
↵ The online version of this article (available at http://dmd.aspetjournals.org) contains supplemental material.
ABBREVIATIONS:
- DDI
- drug-drug interaction
- P450
- cytochrome P450
- hPXR
- human pregnane X receptor
- hCAR
- human constitutive androstane receptor
- ABC
- ATP-binding cassette
- SLC
- solute carrier
- NME
- new molecular entity
- PXR
- pregnane X receptor (NR1I2)
- RIF
- rifampicin
- SAR
- structure-activity relationship
- RIS
- relative induction score
- IPA
- isopropyl acetal
- CBZ
- carbamazepine
- DMSO
- dimethyl sulfoxide
- PCR
- polymerase chain reaction
- AUC
- area under the curve
- EE
- 17α-ethinyl estradiol
- RLU
- relative luminescence units
- OAT
- organic anion transporter
- MDR1
- multidrug resistance 1
- MRP
- multidrug resistance-associated protein
- OCT1
- organic cation transporter 1
- CAR
- constitutive androstane receptor (NR1I3)
- MDZ
- midazolam.
- Received June 22, 2012.
- Accepted August 21, 2012.
- Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics