Abstract
Six proton pump inhibitors (PPIs), omeprazole, lansoprazole, esomeprazole, dexlansoprazole, pantoprazole, and rabeprazole, were shown to be weak inhibitors of cytochromes P450 (CYP3A4, -2B6, -2D6, -2C9, -2C8, and -1A2) in human liver microsomes. In most cases, IC50 values were greater than 40 μM, except for dexlansoprazole and lansoprazole with CYP1A2 (IC50 = ∼8 μM) and esomeprazole with CYP2C8 (IC50 = 31 μM). With the exception of CYP2C19 inhibition by omeprazole and esomeprazole (IC50 ratio, 2.5 to 5.9), there was no evidence for a marked time-dependent shift in IC50 (IC50 ratio, ≤2) after a 30-min preincubation with NADPH. In the absence of preincubation, lansoprazole (IC50 = 0.73 μM) and esomeprazole (IC50 = 3.7 μM) were the most potent CYP2C19 inhibitors, followed by dexlansoprazole and omeprazole (IC50 = ∼7.0 μM). Rabeprazole and pantoprazole (IC50 = ≥25 μM) were the weakest. A similar ranking was obtained with recombinant CYP2C19. Despite the IC50 ranking, after consideration of plasma levels (static and dynamic), protein binding, and metabolism-dependent inhibition, it is concluded that omeprazole and esomeprazole are the most potent CYP2C19 inhibitors. This was confirmed after the incubation of the individual PPIs with human primary hepatocytes (in the presence of human serum) and by monitoring their impact on diazepam N-demethylase activity at a low concentration of diazepam (2 μM). Data described herein are consistent with reports that PPIs are mostly weak inhibitors of cytochromes P450 in vivo. However, two members of the PPI class (esomeprazole and omeprazole) are more likely to serve as clinically relevant inhibitors of CYP2C19.
Introduction
Proton pump inhibitors (PPIs) inhibit the gastric (parietal cell) H+/K+ATPase that is involved in the final step of hydrochloric acid secretion. Consequently, such agents are used to treat acid-related conditions such as peptic ulcers and their complications (e.g., bleeding, gastroesophageal reflux disease, nonsteroidal anti-inflammatory drug-induced gastrointestinal lesions, Zollinger-Ellison syndrome, and dyspepsia). In combination with antibiotics, PPIs are also used to treat Helicobacter pylori infections (Blume et al., 2006; Shi and Klotz, 2008).
Since the introduction of omeprazole (5-methoxy-2-[[(4-methoxy-3,5-dimethyl-2-pyridinyl)methyl]sulfinyl]-1H-benzimidazole) in 1989, the market has expanded to encompass newer PPIs, such as lansoprazole (2-[[3-methyl-4-(2,2,2-trifluoroethoxy)pyridin-2-yl]methylsulfinyl]-1H-benzimidazole), pantoprazole (5-(difluoromethoxy)-2-[(3,4-dimethoxypyridin-2-yl)methylsulfinyl]benzimidazol-1-ide), and rabeprazole (2-[[4-(3-methoxypropoxy)-3-methylpyridin-2-yl]methylsulfinyl]benzimidazol-1-ide), and now also includes the (S)-enantiomer of omeprazole (esomeprazole; (S)-5-methoxy-2-[(4-methoxy-3,5-dimethylpyridin-2-yl)methylsulfinyl]-3H-benzimidazole) and the (R)-enantiomer of lansoprazole (dexlansoprazole; (R)-2-[[3-methyl-4-(2,2,2-trifluoroethoxy)pyridin-2-yl]methylsulfinyl]-1H-benzimidazole) (Andersson et al., 2001; Shi and Klotz, 2008; Vakily et al., 2009). As a drug class, the PPIs are well characterized in terms of their pharmacokinetics, absorption, distribution, metabolism, and excretion properties. For example, it is known that PPIs undergo extensive metabolism by cytochromes P450 (P450s), and that CYP2C19 phenotype substantially influences pharmacokinetics, pharmacodynamics, and clinical outcomes (e.g., speed and degree of gastric acid suppression) (Li et al., 2004; Baldwin et al., 2008; Hunfeld et al., 2008; Shi and Klotz, 2008). In a crowded market, therefore, a considerable effort has been made to differentiate the various PPI class members on the basis of their pharmacokinetics, efficacy, and drug-drug interaction profile (Andersson et al., 2001; Blume et al., 2006; Shi and Klotz, 2008; Vakily et al., 2009; Ogawa and Echizen, 2010).
Recently, PPI-associated drug interactions have garnered the attention of various regulatory agencies and researchers. The growing interest has been fueled by reports of drug interactions in subjects who received the combination of a PPI with clopidogrel, despite the acceptance by many that such an interaction only leads to a slight increase in cardiovascular risk (Rassen et al., 2009; Zhang et al., 2009; Furuta et al., 2010; Liu and Jackevicius, 2010; Oyetayo and Talbert, 2010; Ray et al., 2010; Shmulevich et al., 2011; Ohbuchi et al., 2012; Shah et al., 2012). Clopidogrel, a P2Y12 adenosine diphosphate receptor antagonist, undergoes extensive metabolism to both inactive and active (thiol) metabolites. Although a number of P450s have been shown to catalyze the formation of the active thiol (H4), CYP2C19 has received the most attention; CYP2C19-catalyzed metabolism of clopidogrel is a low Km process in vitro (Hagihara et al., 2009; Kazui et al., 2010), and CYP2C19 genotype is associated with antiplatelet activity and circulating levels of H4 (Shuldiner et al., 2009; Furuta et al., 2010; Boulenc et al., 2012). Consequently, the prescribing information for clopidogrel includes a boxed warning regarding the effectiveness of clopidogrel in CYP2C19 poor metabolizers (PMs). It is also recommended to avoid concomitant use of clopidogrel with omeprazole and esomeprazole.
To date, no attempt has been made to systematically evaluate six marketed PPIs as inhibitors of multiple human P450s under the same experimental conditions. Moreover, data for P450s such as CYP2C8, CYP1A2, and CYP2B6 are limited, and there are no reports of various PPIs as time-dependent (metabolism-dependent) inhibitors of P450s (VandenBranden et al., 1996; Ko et al., 1997; Li et al., 2004; Liu et al., 2005; Walsky et al., 2005, 2006). In fact, there are only two reports describing time-dependent inhibition, with a focus on CYP2C19 (multiple PPIs) or multiple P450s (omeprazole only) (Ogilvie et al., 2011; Boulenc et al., 2012).
Evaluation of six PPIs is important because it affords ranking of a single compound across different P450s, as well as the different compounds against a single P450, and because time-dependent inhibition has been reported for numerous P450s beyond CYP3A4 (Venkatakrishnan and Obach, 2007). Therefore, omeprazole, lansoprazole, esomeprazole [(S)-isomer of omeprazole], dexlansoprazole [(R)-isomer of lansoprazole], pantoprazole, and rabeprazole were assessed as reversible and time-dependent inhibitors of P450 activities in human liver microsomes (HLM) (CYP1A2, -2B6, -2C8, -2C9, -2C19, -2D6, and CYP3A4). During the course of the study, it was evident that some of the PPIs behaved as relatively potent inhibitors of CYP2C19 (versus other P450s) and two of them (omeprazole and esomeprazole) also behaved as time-dependent inhibitors. The latter result supports the findings of Ogilvie et al. (2011) and Boulenc et al. (2012), and the more recent findings of Ohbuchi et al. (2012) using 2-oxo-clopidogrel as substrate. Additional studies were conducted with recombinant CYP2C19 (rCYP2C19) using three different substrates [3-cyano-7-ethoxy-coumarin (CEC), (S)-mephenytoin, and diazepam] and with serum-coincubated human primary hepatocytes using diazepam as substrate. Where appropriate, the determined IC50 was corrected for fu,inc to generate IC50(u) and support a comparison of HLM with rCYP2C19. Finally, the in vitro CYP2C19 inhibition parameters (Ki,u, kinact, and KI,u) were used to predict % inhibition in vivo (% inhibitionpredicted) based on static (Cmax, Cmax,u, Cmax,portal, and Cmax,portal,u) and dynamic (time-dependent) concentrations of each PPI. The approach enabled assessment of CYP2C19 inhibition for various substrates (irrespective of fm,2C19) and the comparison of % inhibitionpredicted versus % inhibition observed in vivo (% inhibitionin vivo).
Materials and Methods
Materials.
Omeprazole, esomeprazole [(S)-isomer of omeprazole], lansoprazole, phenacetin, N-desmethyl-diazepam, diazepam, and [2H5]N-desmethyl-diazepam were obtained from Sigma-Aldrich (St. Louis, MO). Dexlansoprazole [(R)-isomer of lansoprazole] and pantoprazole were obtained from Synfine Research Inc. (Ontario, Canada). Rabeprazole, bupropion, [2H6]hydroxy-bupropion, [2H5]desethyl-amodiaquine, [2H4]acetaminophen, and (S)-mephenytoin were purchased from Toronto Research Chemicals Inc. (North York, Ontario, Canada). Pooled HLM (150 different organ donors), and Supersomes containing rCYP2C19 (coexpressed with P450 reductase), were purchased from BD Biosciences (Woburn, MA). CEC, [13C6]4′-hydroxy-diclofenac, [2H3]4′-hydroxy-mephenytoin, [2H3]dextrorphan, and [13C3]1′-hydroxy-midazolam were also obtained from BD Biosciences (Woburn, MA). AlgiMatrix firming buffer and 24-well plates (AlgiMatrix 3D culture system) were ordered from Invitrogen (Carlsbad, CA). Two preparations of cryopreserved human primary hepatocytes were obtained from Celsis In Vitro Technologies (Baltimore, MD). Both preparations had greater than 80% postthaw viability as determined by trypan blue exclusion. The first represented a single organ donor with relatively high CYP2C19 activity [(S)-mephenytoin 4′-hydroxylase = 95 pmol/min per 106 cells]. The second was a pool of 20 different organ donors with medium CYP2C19 activity [(S)-mephenytoin 4′-hydroxylase = 15 pmol/min per 106 cells]. Human serum was obtained from Bioreclamation LLC (Westbury, NY); purchased frozen, stored at −80°C, and thawed only once before each experiment. Hepatocyte culture medium [Hepatozyme-SFM, l-glutamine (200 mM), and penicillin/streptomycin (10,000 IU/ml, 10,000 μg/ml)] was ordered from Invitrogen. Before cell culture, the medium was diluted and adjusted to a final concentration of 2 mM (glutamine), 50 IU/ml (penicillin), and 50 μg/ml (streptomycin). All other reagents and chemicals were of analytical grade and of the highest quality available commercially.
Inhibition Studies with HLM.
The HLM panel consisted of assays for seven different P450s (CYP1A2, -2B6, -2C8, -2C9, -2C19, -2D6, and -3A4). The assays used pooled HLM and well established substrates (final substrate concentration ∼Km) that produce P450 isoform-selective metabolites (Walsky and Obach, 2004). Each assay was performed in a time-dependent format to assess both reversible IC50 and time (metabolism)-dependent shifts in IC50: test compounds were preincubated at 37°C with HLM in the presence of NADPH (1 mM) for 0 and 30 min. Assays were performed in 384-well microplates in a total volume of 30 μl. Automated liquid handling equipment was used in the various steps of the assay process: Genesis 150 (Tecan Group Ltd., Mannedorf, Switzerland), BenchCel System (Velocity 11 Inc., Menlo Park, CA), ECHO 550 (LabCyte Inc., Sunnyvale, CA), and Multidrop Combi (Thermo Electron Corporation, Vantaa, Finland). Each drug substance was tested as a single point at each of 10 concentrations ranging from 2 nM to 40 μM, final dimethyl sulfoxide (DMSO) concentration in the reaction mixture was 0.2%. The IC50 value for each compound was determined using a four-parameter logistic regression model (see Data Analysis). See supplemental data for Sample Preparation, Assay, RapidFire-Mass Spectrometry (RF-MS/MS), LC-MS/MS Analysis of CYP1A2 (HLM-Phenacetin) Reaction Samples, Inhibition Studies with rCYP2C19; CEC O-deethylase, (S)-Mephenytoin 4′-Hydroxylase, Diazepam N-Demethylase, and Determination of fu,inc.
Incubations with Human Primary Hepatocytes in the Presence of Human Serum.
Cryopreserved human primary hepatocytes were thawed rapidly at 37°C. The cells were washed with fresh culture medium (50 ml) and then centrifuged (120g) for 3 min. After removal of the supernatant, the cells were resuspended in hepatocyte culture medium containing 10% (v/v) AlgiMatrix firming buffer to yield a final density of 4.2 × 106 cells/ml. An aliquot (300 μl) of the suspension was transferred into the middle of each well of a 24-well AlgiMatrix 3D culture plate. After gentle horizontal shaking, the plate was placed in an incubator (37°C; humidified atmosphere of 5% CO2) for 10 min to allow the hepatocytes to seed on the AlgiMatrix sponge substratum. Finally, human serum (700 μl) was added to each well so that the final volume was 1 ml and the cell density was 1.25 × 106 cells/ml. For each of the six PPIs, a stock solution was prepared in culture medium containing 40% DMSO, and 5-μl aliquots were added to the specific wells in the assay plate. The final PPI concentration was based on the calculated Cmax.portal (see legend to Table 3) and was 2.5 μM (omeprazole), 18.7 μM (esomeprazole), 2.9 μM (lansoprazole and dexlansoprazole), 6.7 μM (pantoprazole), and 1.4 μM (rabeprazole). DMSO alone (0.2%) served as a control. Diazepam was prepared as a stock solution in 100% DMSO (at 1 or 2 mM final concentration) and was added (1 μl) to each well. Therefore, the final concentrations of DMSO and serum in the reaction mixtures were 0.3 and 70% (v/v), respectively. The final concentration of diazepam (1–2 μM) approximated the calculated Cmax,portal after an intravenous (IV) dose (data not shown) and is below the Km reported for CYP2C19 (Yasumori et al., 1994). At the specific time points of the incubation time course (0, 2, 4, 7, 24, and 30 h), an aliquot (50 μl) of the incubate (serum layer) was removed and added to acetonitrile (500 μl) containing 0.3 μM [2H5]N-desmethyl-diazepam (internal standard). The sample was vortexed for 3 min, centrifuged, and the supernatant (10 μl) subjected to liquid chromatography/tandem mass spectrometry analysis (see supplemental data). For the purposes of quantitation, a calibration curve of N-desmethyl-diazepam was prepared in human serum matrix at final concentrations of 0, 1, 2, 4, 8, 16, 32, 64, and 128 nM. The rate of reaction was calculated using the concentration of N-desmethyl-diazepam at each time point (see Data Analysis).
Data Analysis.
Determination of IC50.
The endpoint of the RapidFire-mass spectrometry readout for the assays using HLM was the signal intensity of the metabolite, which was then normalized to the signal of internal standard in the same sample. Therefore, the sample signal intensity was expressed as a signal ratio. The endpoint readout for the CEC O-deethylase assay was the fluorescence intensity of the metabolite. For both the liquid chromatography/mass spectrometry-based and fluorescence assays, the sample readout (signal ratio or fluorescence intensity) was then normalized to the signal ratio or fluorescence intensity of the reactions performed in the absence of the test substance (total signal, 0% inhibition), and the reactions performed in the presence of the inhibitor cocktail (background signal, 100% inhibition). These normalized results were expressed as percentage of inhibition calculated as shown in eq. 1: where S = sample, T = average total, B = average background.
The results were then imported into custom curve fitting software, which uses MathIQ package (ID Business Solutions, Ltd., Guilford, England), to determine the IC50 values for each test compound. The IC50 was defined as the concentration corresponding to 50% inhibition derived from the fitted 10-point curve using a four-parameter logistic regression model (eq. 2): where Y = response at a given concentration of inhibitor (X), A = minimum response, B = maximal response, D = Hill coefficient (slope), and C = X at which Y = A + [(B − A)/2].
Determination of KI and kinact.
Inhibition parameters (kinact and KI) were determined by nonlinear fitting of the kobs at each concentration of PPI versus PPI concentration ([I]) (eq. 3) (GraFit version 6; Erithacus Software Ltd., Surrey, UK).
CYP2C19 inhibition in the presence of human primary hepatocytes.
For the human hepatocyte studies, the rate of N-desmethyl-diazepam formation was determined in the presence of DMSO alone or in the presence of each individual PPI at its respective predicted Cmax,portal. Formation of N-desmethyl-diazepam was linear up to 7 h, so it was possible to determine the rate of reaction. In this instance, % inhibition was calculated as follows (eq. 4): where VDMSO and VPPI is the rate of N-desmethyl-diazepam formation (pmol/min per 106 cells) in the presence of DMSO alone and PPI, respectively.
Prediction of CYP2C19 % Inhibition In Vivo Based on In Vitro Data (Using Static Concentration of Each PPI).
The in vitro-derived parameters (HLM and rCYP2C19) were used to predict the % inhibition in vivo as follows (eq. 5): In this instance, FRpredicted = δ × γ
For reversible inhibition (γ), Ki,u was obtained by simply dividing IC50(u) by 2 (assuming competitive inhibition; substrate concentration ∼Km for each assay). When time-dependent (mechanism-based) inhibition was evident (δ), the experimentally derived estimate of KI was corrected for fu,inc, and kdeg was set at 0.0005 min−1 (half-life of CYP2C19 is ∼24 h) (Nishiya et al., 2009). For the determination of both γ and δ, [I] is the inhibitor concentration and is based on Cmax, Cmax,u, Cmax,portal, and Cmax,portal,u (see below). Some of the PPIs did not exhibit time-dependent inhibition, so δ = 1.
Where possible, the Cmax,portal for each PPI was calculated as shown in eq. 6 (Vuppugalla et al., 2010; references therein): where Q is the hepatic blood flow (1500 ml/min) and fa approaches unity for each PPI. The Cmax and Cmax,portal were corrected for plasma fu to generate Cmax,u and Cmax,portal,u, respectively. For each PPI, the estimated ka (deconvoluted oral data based on published IV data) was obtained from the literature as follows: 0.037 (omeprazole), 0.178 (esomeprazole), 0.013 (lansoprazole), 0.011 (pantoprazole), and 0.005 (rabeprazole) min−1 (Landahl et al., 1992; Pue et al., 1993; Gerloff et al., 1996; Andersson et al., 2001; Setoyama et al., 2005). Unfortunately, in the absence of published IV data, it was not possible to obtain estimates of Cmax,portal for dexlansoprazole. For omeprazole, esomeprazole, lansoprazole, dexlansoprazole, pantoprazole, and rabeprazole, plasma fu was 0.05, 0.03, 0.03, 0.02, 0.02, and 0.04, respectively. Unless otherwise indicated, plasma Cmax (Cmax,u) values for each PPI were based on those reported for a specific diazepam drug interaction; 0.3 (0.02) μM after 20 mg of oral omeprazole, 5.2 (0.16) μM after 30 mg of oral esomeprazole, 3.4 (0.11) μM after 60 mg of oral lansoprazole, 4.0 (0.08) μM after 90 mg of oral dexlansoprazole, 57 (1.1) μM after 240 mg of IV pantoprazole, and 0.45 (0.02) μM after 20 mg of oral rabeprazole (see Supplemental Table S2) (Andersson et al., 1990, 2001; Lefebvre et al., 1992; Ishizaki et al., 1995; Gugler et al., 1996; Vakily et al., 2009). Equation 6 yielded the following estimates for Cmax,portal: 1.7 μM (omeprazole), 15.5 μM (esomeprazole), 4.8 μM (lansoprazole), 57 μM (pantoprazole), and 0.8 μM (rabeprazole). Corresponding Cmax,portal,u values were as follows: 0.1 μM (omeprazole), 0.47 μM (esomeprazole), 0.14 μM (lansoprazole), 1.1 μM (pantoprazole), and 0.03 μM (rabeprazole).
Prediction of CYP2C19 % Inhibition In Vivo Based on In Vitro Data (Using Time-Dependent Concentrations of Each PPI).
Simulations were performed to project the interaction of each PPI with diazepam using a semimechanistic compartment model (Supplemental Fig. S1). In brief, the observed plasma concentration-time data of each PPI, digitized from literature reports (Supplemental Table S2), were fitted using a one- or two-compartment model (vide infra). Likewise, the observed concentration-time profiles of diazepam without PPI, digitized from literature reports (Supplemental Table S2), were fitted using a two-compartment model (vide infra). The elimination of diazepam was assumed to occur via two pathways, with CYP2C19-mediated metabolism fixed at fm,2C19 = 0.57 (vide infra; eq. 16). Subsequently, the interaction of each PPI with diazepam was simulated by incorporating in vitro inhibition parameters. The simulated profiles of diazepam after PPI administration were compared with the observed data obtained from the literature. All PPI-diazepam interaction modeling was performed using WinNonlin (Enterprise version 5.0; Pharsight, Mountain View, CA) and Berkeley Madonna (Berkeley Madonna Inc., University of California, Berkeley, CA).
Modeling the pharmacokinetics of each PPI.
In brief, the plasma concentration-time profile of each PPI after single-dose oral administration was fitted to a one- or two-compartment model (Supplemental Fig. S2), wherein ka and kel are the first-order absorption and elimination rate constants and kcp and kpc are the distribution rate constants from the central and the peripheral compartments. In the case of pantoprazole, which was administered intravenously, the absorption compartment was not used and ka was set to zero. The differential equations are shown below (eqs. 7, 8, and 9): wherein “GI,” “PPI,” and “PeripheralPPI” are the amounts of inhibitor (PPI) in gastrointestinal tract, plasma, and peripheral compartment, respectively. The single-dose pharmacokinetics of omeprazole, esomeprazole, lansoprazole, pantoprazole, and rabeprazole were obtained from the literature (Andersson et al., 1990, 2001; Lefebvre at al., 1992; Pue et al., 1993; Setoyama et al., 2005). In the case of rabeprazole, the absorption was modeled using a Weibull function, because a first-order process could not adequately capture the time course of absorption. For dexlansoprazole, two first-order absorption rate constants (ka1 and ka2) were used to model the absorption kinetics after administration of the modified release tablet (Vakily et al., 2009). Values for kcp and kpc were fixed to zero in all cases except for esomeprazole, where they were estimated due to the bi-exponential nature of the profile.
Modeling the pharmacokinetics of diazepam.
The pharmacokinetics of IV diazepam in the absence of PPI was modeled (fitted) using a two-compartment model with the elimination occurring from the central compartment as shown in eqs. 10 to 12 (Supplemental Fig. S1). Because diazepam was administered orally (5 mg) in the drug-drug interaction study with dexlansoprazole (Vakily et al., 2009), an absorption compartment with a first-order absorption rate constant of kad (3.0 h−1) was used to model diazepam kinetics. In addition, because of the low clearance of diazepam relative to hepatic blood flow (<10%), the oral bioavailability was assumed to be complete (Klotz et al., 1975). The elimination was assumed to occur via two pathways, with CYP2C19-mediated metabolism fixed at fm,2C19 = 0.57 (see below; eq. 16). The clearance of diazepam via the CYP2C19-mediated component was modeled using a well stirred model equation (eq. 12): Wherein “Diazepam” is the plasma concentration of diazepam, “PeripheralDZ” is the amount of diazepam in the peripheral compartment, and CL1 and CL2 are the two clearance pathways for diazepam; with CL1 assumed to be CYP2C19-mediated. Vc refers to the volume of the central compartment (300 ml/kg), and k12 (0.3 h−1) and k21 (0.13 h−1) are the distribution rate constants from the central and peripheral compartments. Q is the hepatic blood flow in man (20.5 ml/min per kg), and CLint is the intrinsic clearance of diazepam in man via the CYP2C19-mediated pathway. Based on data digitized from literature studies (Andersson et al., 1990, 2001), the pharmacokinetic profile of diazepam in the placebo leg of the study was fitted to the above equations to estimate CLint, CL2, k12, k21, and Vc (above).
Simulating the effect of each PPI on the pharmacokinetics of diazepam.
Simulations were performed to predict the effect of each PPI on the pharmacokinetics of diazepam. The results of these simulations were compared with diazepam concentration-time profiles and area under the curve (AUC) ratios obtained from drug-drug interaction studies reported in the literature. The dose and dosing regimen used in these simulations were similar to literature reports (see Supplemental Fig. S1 and Table S2). In brief, using the semimechanistic compartment model, concentration-time profiles were simulated after “administration” of diazepam as a short intravenous infusion (0.1 mg/kg) or a single oral dose (5 mg) after PPI dosing (Supplemental Fig. S1 and Table S2). The intrinsic clearance of diazepam in the presence of PPI (CLint′) was assumed to be inhibited as a function of [I] as follows (eq. 13): wherein kdeg, kinact (determined in vitro), KI,u (determined in vitro), and Ki,u (determined in vitro) have been defined previously (above) and were obtained using rCYP2C19 with diazepam as substrate. For PPIs that inhibit CYP2C19 only by a reversible mechanism (lansoprazole, pantoprazole, dexlansoprazole, and rabeprazole), kdeg, kinact, and KI,u were fixed at zero. Finally, to evaluate the influence of protein binding, drug-drug interaction predictions were performed under two scenarios using time-dependent free plasma PPI concentrations and total PPI plasma concentrations as surrogates for [I]. Data were reported as the predicted ratio of diazepam AUC (PPI versus placebo), and it was possible to determine % inhibitionpredicted for each PPI (eq. 14): In eq. 14, FRpredicted was determined using eq. 15 (fm,2C19 = 0.57; see below):
Calculation of fm,2C19 for Diazepam.
The AUC of diazepam is increased 1.8- to 2.3-fold in CYP2C19 PM versus extensive metabolizer (EM) subjects, so the drug can serve as a CYP2C19 probe (Andersson et al., 1990; Ishizaki et al., 1995). Moreover, diazepam is a low clearance compound (0.29–0.46 ml/min per kg), with good oral bioavailability, and undergoes extensive metabolism (>99% of the dose) (Klotz et al., 1975). Therefore, the fraction of the diazepam dose metabolized by CYP2C19 can be calculated based on the clinically determined AUCPM/AUCEM ratio (eq. 16) (Ito et al., 2005). Based on the AUCPM/AUCEM ratio of 2.3, fm,2C19 for diazepam is calculated to be 0.57.
Determination of CYP2C19 % Inhibition In Vivo For Each PPI.
In vivo inhibition of CYP2C19 was assessed based on the impact of each PPI on the pharmacokinetics of diazepam; diazepam is one of the few well characterized probes that has been studied with all six PPIs described. For each PPI with diazepam, the % inhibitionin vivo was then determined as follows (eq. 17): In eq. 17, FRin vivo was determined using eq. 18 (fm,2C19 = 0.57);
The diazepam AUCi/AUCc ratio (observed) with each PPI (AUC in the presence of PPI versus placebo) has been reported [90% confidence interval (CI)]; 1.28 (1.19, 1.39; Ishizaki et al., 1995) and 1.36 (1.19, 1.53; Andersson et al., 1990) for omeprazole; 1.12 (0.99, 1.23) for lansoprazole (Lefebvre et al., 1992); 1.81 (1.42, 2.31) for esomeprazole (Andersson et al., 2001); 1.06 (1.01, 1.12) for dexlansoprazole (Vakily et al., 2009); 0.99 (0.87, 1.13) for pantoprazole (Gugler et al., 1996); and 0.91 (0.81, 1.02) for rabeprazole (Ishizaki et al., 1995). The CI (90%) was reported or estimated from the mean ± S.D. data (ratio variance approximated by the Taylor expansion).
Results
Reversible Inhibition of P450s in the Absence of Preincubation with NADPH.
HLM.
Data describing the reversible inhibition of P450 activities in HLM by omeprazole, lansoprazole, esomeprazole, dexlansoprazole, pantoprazole, and rabeprazole are presented in Table 1. The PPI concentration range tested (2 nM–40 μM) afforded coverage of known clinically relevant plasma concentrations and differentiation of the inhibitory potency across the various P450s. Overall, relatively minimal inhibition (IC50 > 40 μM) of CYP3A4, CYP2B6, CYP2D6, CYP2C9, CYP2C8, and CYP1A2 activity was observed. However, it was possible to obtain an IC50 (∼31 μM) for esomeprazole with CYP2C8 and an IC50 (∼8 μM) for both lansoprazole and dexlansoprazole with CYP1A2. Of the P450s tested, the most potent inhibition was observed with CYP2C19-catalyzed (S)-mephenytoin 4′-hydroxylase activity (Table 2). In this instance, lansoprazole (IC50 = 0.73 μM) was found to be the most potent inhibitor, followed by esomeprazole (IC50 = 3.7 μM), dexlansoprazole (IC50 = 6.0 μM), and omeprazole (IC50 = 7.4 μM). Both rabeprazole and pantoprazole were relatively weak inhibitors of CYP2C19 (IC50 ≥ 25 μM).
rCYP2C19.
In agreement with HLM data, lansoprazole and esomeprazole were the most potent inhibitors of rCYP2C19-catalyzed CEC O-deethylase activity (IC50 = 0.4 μM), followed by omeprazole and dexlansoprazole (IC50 = 1.2 and 2.2 μM, respectively), and then by pantoprazole and rabeprazole (IC50 ∼ 4.0 μM). Relatively potent inhibition with lansoprazole and esomeprazole was observed also with rCYP2C19-catalyzed diazepam (IC50 = 2.4 and 4.6 μM, respectively) and (S)-mephenytoin (IC50 = 1.1 and 1.3 μM, respectively) metabolism (Table 2). This meant that the difference in CYP2C19 inhibitory potency observed in HLM for omeprazole and its (S)-isomer, as well as lansoprazole and its (R)-isomer, was also evident with rCYP2C19 using three different substrates; omeprazole was less potent than esomeprazole (∼2-fold with HLM and 1.7 to 9.6-fold with rCYP2C19), whereas lansoprazole was more potent than dexlansoprazole (∼8-fold with HLM and ∼5-fold with rCYP2C19).
Time-Dependent Inhibition of P450s in HLM after Preincubation with NADPH.
The six PPIs were also assessed as time-dependent inhibitors of various P450 activities in HLM (Tables 1 and 2), involving a 30-min preincubation of the PPI with NADPH-fortified HLM, and the IC50(t) was compared with IC50 obtained without preincubation with NADPH. In most cases, a relatively minimal time-dependent shift in IC50 (IC50/IC50(t) ratio <2.0) was observed in comparison to positive controls such as tienilic acid (CYP2C9), paroxetine (CYP2D6), verapamil (CYP3A4), ticlopidine (CYP2C19, CYP2B6), and furafylline (CYP1A2) (see legend to Supplemental Table S1). However, a time-dependent shift in IC50 was evident with omeprazole and CYP1A2 activity (IC50 > 40 μM, IC50(t) = 20.6 μM); esomeprazole and CYP2D6 (IC50 > 40 μM, IC50(t) = 20.9 μM); and with rabeprazole and CYP1A2 (IC50 > 40 μM, IC50(t) = 18.1 μM), CYP2C8 (IC50 > 40 μM, IC50(t) = 13.9 μM), and CYP2D6 (IC50 > 40 μM, IC50(t) = 28.5 μM) (Table 1). Only omeprazole (IC50/IC50(t) ratio = 2.5) and esomeprazole (IC50/IC50(t) ratio = 4.9) exhibited time-dependent inhibition of CYP2C19-catalyzed (S)-mephenytoin 4′-hydroxylation (Table 2). The higher IC50/IC50(t) ratio for esomeprazole (∼9.5) versus omeprazole (∼2.9) was confirmed with rCYP2C19 using diazepam as substrate (data not shown).
Determination of kinact and KI for CYP2C19 (HLM and rCYP2C19).
The time-dependent shift in IC50 with (S)-mephenytoin described above is consistent with the observations of Ogilvie et al. (2011) and Boulenc et al. (2012). In fact, both groups have reported omeprazole (up to 100 μM) as a mechanism-based inhibitor of CYP2C19 in HLM and reported KI values ranging from 1.7 to 9.1 μM and kinact values ranging from 0.016 to 0.046 min−1. Likewise, a ∼2-fold shift in IC50 has also been reported by Ohbuchi et al. (2012) using 2-oxo-clopidogrel as substrate of rCYP2C19. Therefore, we also sought to determine KI and kinact with HLM using (S)-mephenytoin as substrate, but we wanted to confirm the parameters for both omeprazole and esomeprazole with rCYP2C19 also.
After a 30-min preincubation at a higher protein concentration and a 10-fold dilution of incubate with the assay buffer containing substrate (∼10-fold higher than Km), it was possible to verify that the kinact/KI ratio is higher for esomeprazole (kinact/KI ratio = 0.056 min−1μM−1) than for omeprazole (kinact/KI ratio = 0.027 min−1μM−1) in HLM (Fig. 1). Ticlopidine was tested as a positive control under the same assay conditions, and similar parameters to those of Nishiya et al. (2009) were obtained (kinact/KI ratio = 0.068 min−1μM−1; KI = 1.53 ± 0.03 μM; kinact = 0.105 ± 0.011 min−1) (data not shown). Likewise, rCYP2C19 with (S)-mephenytoin rendered a higher kinact/KI ratio for esomeprazole (kinact/KI ratio = 0.085 min−1μM−1) compared with omeprazole (kinact/KI ratio = 0.018 min−1μM−1). The results obtained with (S)-mephenytoin were confirmed with diazepam as rCYP2C19 substrate; esomeprazole kinact/KI ratio = 0.042 min−1μM−1 and omeprazole kinact/KI ratio = 0.01 min−1μM−1 (Fig. 1). No attempt was made to determine kinact and KI using HLM with diazepam as substrate, because at higher diazepam concentrations (>CYP2C19 Km), additional P450s are involved in metabolism and the data would be difficult to interpret. In contrast, (S)-mephenytoin is relatively more CYP2C19-selective over a wide concentration range. Overall, the kinact (0.030–0.048 min−1) and KI (1.1–3.8 μM) values for omeprazole described herein fell within the range reported by Ogilvie et al. (2011). However, both Ogilvie et al. (2011) and Boulenc et al. (2012) reported higher KI values using HLM (∼9 μM) when higher concentrations (>30 μM) of omeprazole were used (e.g., 40–100 μM). Although 30 μM was the highest concentration of omeprazole tested in the present study, the use of seven to 10 different omeprazole concentrations enabled robust fitting of the data and parameter determination (Fig. 1).
Various PPIs as Inhibitors of Diazepam N-Demethylation Catalyzed by Human Primary Hepatocytes Coincubated with Human Serum.
As shown in Table 3, the six PPIs were evaluated as inhibitors of diazepam N-demethylase activity in the presence of human primary hepatocytes using the AlgiMarix 3D culture system. Two preparations of cells were used, both coincubated with human serum (70% v/v). Under the incubation conditions described, the formation of N-desmethyl diazepam was linear with time and the scaled diazepam clearance (∼1.0 ml/min per kg) was similar to that observed clinically after an IV dose (0.29–0.46 ml/min per kg) (Klotz et al., 1975). Moreover, use of low diazepam concentrations (∼2 μM) ensured that N-demethylase activity was largely reflective of CYP2C19 (Yasumori et al., 1994).
Each individual PPI was added at a single concentration based on its estimated Cmax,portal (see Materials and Methods), and the serum was added to account for the differences in fu, in addition to hepatocyte uptake and binding. Consistent with the estimates of % inhibition in vivo, the greatest inhibition was observed with esomeprazole and omeprazole. In comparison, relatively minimal inhibition (≤13.5%) was evident with lansoprazole, dexlansoprazole, rabeprazole, and pantoprazole (Table 3).
Prediction of CYP2C19 Inhibition In Vivo Based on In Vitro-Derived Inhibition Data and a Static Concentration of PPI (Cmax, Cmax,u, Cmax,portal, and Cmax,portal,u).
HLM [(S)-mephenytoin] and rCYP2C19 (diazepam, (S)-mephenytoin, and CEC as substrate) IC50(u) data were used to derive Ki,u values (Tables 4 and 5). In turn, the Ki,u values were used to estimate the degree of CYP2C19 inhibition in vivo based on published data for each compound (e.g., Cmax, dose, fu, fa). For omeprazole and esomeprazole, the experimentally derived parameters for metabolism-dependent inhibition (KI,u and kinact) were also considered. Where possible, the ka for each individual PPI was calculated by leveraging published human oral and IV pharmacokinetic data (Landahl et al., 1992; Pue et al., 1993; Gerloff et al., 1996; Andersson et al., 2001; Setoyama et al., 2005). With the exception of dexlansoprazole, it was possible to calculate Cmax,portal and Cmax,portal,u (see Materials and Methods).
As shown in Table 4, for lansoprazole, dexlansoprazole, and rabeprazole, it was possible to rationalize the inhibition in vivo by considering either fu-corrected Cmax or fu-corrected Cmax,portal. Moreover, it was possible to show that inhibition of CYP2C19 with rabeprazole was negligible and inhibition with lansoprazole (8–34%) was greater than that observed with dexlansoprazole (1–7%). With the exception of rabeprazole, the % inhibition in vivo was greatly overestimated by not correcting for plasma protein binding. Based on the plasma Cmax (57 μM) after a 240-mg IV dose of pantoprazole (Gugler et al., 1996), the % inhibition of CYP2C19 was overestimated even with fu correction. Only HLM-derived (S)-mephenytoin Ki data (>20 μM) rendered a % inhibitionpredicted value close to that in vivo (<5%).
In the case of omeprazole, fu correction was needed to avoid overestimation of % inhibition (Table 5). Furthermore, it was also evident that one had to consider metabolism-dependent inhibition. In this regard, rCYP2C19-derived values for % inhibitionpredicted using (S)-mephenytoin (38–77%) and diazepam (26–66%), and fu-corrected Cmax, were closest to the % inhibitionin vivo range reported for diazepam. Unfortunately, data for the inhibition of (S)-mephenytoin 4′-hydroxylase in vivo are not available, so an attempt was made to determine % inhibitionin vivo with two additional CYP2C19 substrates (moclobemide, fm,CYP2C19 = 0.72; escitalopram, fm,CYP2C19 = 0.44). Like diazepam, the AUC of moclobemide (1.3–2.2-fold) and escitalopram (1.4–1.6-fold) is increased in CYP2C19 extensive metabolizer subjects codosed with omeprazole (Yu et al., 2001; Malling et al., 2005). Therefore, the % inhibitionin vivo with moclobemide (32–76%) is closer to that observed with diazepam (28–61%) versus escitalopram (64–89%). It is interesting to note that HLM (S)-mephenytoin 4′-hydroxylase-derived values for % inhibitionpredicted were higher (51–85%) even after incorporation of plasma fu (Table 5). Whether or not the degree of CYP2C19 inhibition by omeprazole is substrate-dependent requires further investigation; in our study, the kinact/KI ratio (rCYP2C19) with diazepam (0.01 min−1μM−1) and (S)-mephenytoin (0.018 min−1μM−1) were comparable (Fig. 1).
As presented in Table 5, esomeprazole has been shown to elicit a relatively marked effect on the AUC of diazepam, and % inhibitionin vivo ranged from 52 to 99%. Therefore, the extent of CYP2C19 inhibition with esomeprazole was greater than that observed with any other PPI. Although the predictions correctly presented esomeprazole as the most potent CYP2C19 inhibitor, % inhibition was overestimated (% inhibitionpredicted ≥ 95%) when using Cmax or Cmax,portal (Fig. 4). Furthermore, correction for plasma fu, consideration of portal concentration, use of HLM versus rCYP2C19, and substrate type did not improve the prediction. It should be noted that the rate constant for CYP2C19 degradation (kdeg) is a key parameter, and if incorrectly applied it can lead to erroneous estimates of % inhibition in vivo. Unfortunately, the kdeg for CYP2C19 has not been determined in vivo, and so the value used in the current analysis (0.0005 min−1) has been derived in vitro and used by others (Nishiya et al., 2009; Ogilvie et al., 2011).
Prediction of CYP2C19 Inhibition In Vivo Based on In Vitro-Derived Inhibition Parameters and a Semimechanistic Compartment Model Describing the Pharmacokinetics of Both PPI and Diazepam.
The analyses described above focused on the use of static PPI concentrations. Therefore, the impact of time-dependent PPI concentrations was considered also. Toward this end, a semimechanistic compartment model was developed to describe each PPI-diazepam drug interaction and leverage in vitro-derived inhibition parameters (rCYP2C19-catalyzed diazepam N-demethylation) (see Materials and Methods). The plots of the observed in vivo and predicted plasma concentration-time profiles are shown for each of the six PPIs (Supplemental Fig. S2), after oral dosing of omeprazole, esomeprazole, lansoprazole, rabeprazole, and dexlansoprazole at 20, 30, 60, 20, and 90 mg, respectively, and pantoprazole IV dose at 240 mg (Supplemental Table S2). In all cases, the observed plasma concentrations were adequately captured using a one- or two-compartment model (Supplemental Fig. S1). The subsequent plots of observed and model-simulated plasma concentration-time profiles of diazepam after PPI dosing are shown in Fig. 2 (incorporating total PPI plasma concentration as [I]) and in Fig. 3 (incorporating free PPI plasma concentration as [I]). As shown in Fig. 2, the semimechanistic compartment model incorporating total PPI concentrations was able to reasonably capture the diazepam time course of placebo and PPI-treated subjects, and % inhibitionpredicted was 50% (esomeprazole), 29% (omeprazole), 16% (lansoprazole), 16% (dexlansoprazole), 29% (pantoprazole), and <1% (rabeprazole). In contrast, incorporation of free [I] resulted in the underprediction of the diazepam AUC ratio compared with the observed geometric mean ratio (<1% inhibitionpredicted) (Fig. 3). Pantoprazole behaved as an outlier, because fu-adjusted plasma concentrations rendered a % inhibitionpredicted value (16%) that fell within the in vivo range (<1%, 20%) as defined by the 90% CI. It is interesting to note that, based on total PPI concentration in plasma, the model predicted that only a partial decrease in inhibition is possible when diazepam is dosed 12 h after omeprazole (29 versus 16%) and esomeprazole (50 versus 29%) (Supplemental Fig S3). For these two PPIs, it may not be possible to successfully mitigate CYP2C19 inhibition in a clinical setting by separating the dose of PPI from that of the victim drug.
Summary Comparison of the Six PPIs as CYP2C19 Inhibitors.
To facilitate a comparison across the seven different modeling methods used, summary data are presented for the six PPIs (Fig. 4). In all cases, esomeprazole was predicted to be the most potent CYP2C19 inhibitor, whereas rabeprazole was predicted to be the weakest inhibitor. It was possible to differentiate esomeprazole from omeprazole, especially when fu-corrected static PPI concentration or when time-dependent changes in total PPI concentration were considered. Consistent with in vivo data, both lansoprazole and dexlansoprazole were predicted to be weaker inhibitors versus esomeprazole and omeprazole. However, these two PPIs could only be differentiated from each other (lansoprazole > dexlansoprazole) only when fu-corrected Cmax was considered (8 versus 1% inhibitionpredicted). In the case of pantoprazole, only hepatocyte data correctly predicted minimal inhibition (≤5%) of diazepam metabolism (Fig. 4).
Discussion
As described herein, it was possible to evaluate six PPIs as reversible and time-dependent inhibitors of seven different human P450s under the same assay conditions in vitro. To our knowledge, this has not been reported previously. In the absence of preincubation, the IC50s obtained (Tables 1 and 2) more or less complimented the results of others who have reported IC50 data for different PPI-P450 combinations (VandenBranden et al., 1996; Ko et al., 1997; Li et al., 2004; Liu et al., 2005; Walsky et al., 2005, 2006; Ogilvie et al., 2011). For the first time, it is possible to report that the PPIs studied do not exhibit marked time-dependent inhibition of six P450s (CYP1A2, -2B6, -3A4, -2C9, -2C8, and -2D6) in HLM (IC50/IC50(t) ratio <2.0). Only with omeprazole (CYP1A2), esomeprazole (CYP2D6), and rabeprazole (CYP1A2, -2C8, and -2D6) did the IC50 shift to any measurable extent. In part, this may reflect the generation of metabolites that serve as reversible inhibitors. For example, the thioether metabolite of rabeprazole is known to be a more potent inhibitor of CYP2D6 in HLM (Li et al., 2004). Based on parent PPI pharmacokinetics, however, inhibition of these six P450s is predicted to be minimal, assuming that Cmax,u (≤5% inhibition) or Cmax,portal,u (≤12% inhibition) governs the interaction (data not shown). This is consistent with the minimal effect of various PPIs on the pharmacokinetics of probes such as (S)-warfarin (CYP2C9), theophylline (CYP1A2), and metoprolol (CYP2D6) (Andersson et al., 2001; Blume et al., 2006; Uno et al., 2008; Vakily et al., 2009; Ogawa and Echizen, 2010). In the absence of appreciable CYP3A4 inhibition in vitro, the only drug interactions that could not be rationalized were the ∼1.3-fold increase in AUCs of cisapride with esomeprazole, nifedipine with omeprazole, and tacrolimus with lansoprazole, pantoprazole, and rabeprazole (Andersson et al., 2001; Ogawa and Echizen, 2010). It should be noted that the various metabolites of each PPI were not studied as CYP3A inhibitors. In addition, the effect of CYP2C19 phenotype on the pharmacokinetics of each PPI, and its impact on the [I]/Ki ratio for CYP3A4, was not considered in each case.
With the exception of pantoprazole, the inhibition of CYP2C19 activity in HLM rendered the lowest IC50 values (Tables 1 and 2). Such a result was not unexpected, given that PPIs are known to mostly serve as low Km CYP2C19 substrates (Karam et al., 1996; Abelo et al., 2000; Li et al., 2005). With all three rCYP2C19 substrates chosen, rabeprazole and pantoprazole were weaker than lansoprazole. This observation is in accord with the findings of Li et al. (2004), Zhang et al. (2009), and Ohbuchi et al. (2012). Liu et al. (2005) have determined that lansoprazole is more potent (∼3-fold) than dexlansoprazole with (S)-mephenytoin as substrate. We report an 8-fold (HLM) and 6-fold (rCYP2C19) greater potency with the same substrate. Li et al. (2004), Liu et al. (2005), and Ohbuchi et al. (2012) have documented esomeprazole and omeprazole as more or less equipotent inhibitors of CYP2C19. In our hands, esomeprazole was more potent when incubated with HLM (∼2-fold) and recombinant CYP2C19 (1.7–10-fold). This is in contrast to Ogilvie et al. (2011), who reported a lower IC50 for omeprazole (6.9 ± 0.7 vs. 15 ± 1 μM).
In terms of the time-dependent inhibition observed with omeprazole and esomeprazole (Fig. 1), the results are in accord with other reports (Ogilvie et al., 2011; Boulenc et al., 2012; Ohbuchi et al., 2012). Ogilvie et al. (2011) have shown that the time-dependent inhibition is consistent with mechanism-based inactivation resulting in irreversible inactivation of CYP2C19 in HLM. This implies that both PPIs undergo P450-mediated oxidation to a product that covalently binds to CYP2C19. Because metabolism is known to be catalyzed by CYP2C19 (low Km) and CYP3A4 (high Km) in HLM (Abelo et al., 2000; Li et al., 2005), we also sought to assess the time-dependent inhibition with preparations of rCYP2C19 using both (S)-mephenytoin and diazepam as substrates. In our study, it was possible to confirm that the KI with CYP2C19 ranged from 1.8 to 3.8 μM (Fig. 1) and was consistent with the low Kms reported for both omeprazole and esomeprazole (Abelo et al., 2000; Li et al., 2005). The HLM-derived estimates of KI (∼1.0 μM) reported herein are largely reflective of low Km CYP2C19-dependent metabolism (Li et al., 2005) and of the PPI concentration range used in the present study (≤30 μM). Therefore, the higher HLM-derived values of KI (∼9 μM) reported by Ogilvie et al. (2011) and Boulenc et al. (2012) may reflect the use of higher PPI concentrations (>30 μM) and metabolism by high Km P450s (e.g., CYP3A4) giving rise to CYP2C19 inhibition also.
Overall, the data presented indicate that lansoprazole is the PPI that serves as the most potent reversible CYP2C19 inhibitor in vitro. As such, one would anticipate drug interactions with known CYP2C19 substrates. That is not the case, however, because lansoprazole interactions with drugs such as diazepam and phenytoin are relatively minimal (Lefebvre et al., 1992; Ogawa and Echizen, 2010). In contrast, esomeprazole and omeprazole are weaker reversible inhibitors of CYP2C19 in vitro, but their effect on the pharmacokinetics of diazepam and phenytoin is relatively greater. For example, the AUC of phenytoin is increased 25, 20, <1, 3, and <1% by omeprazole, esomeprazole, dexlansoprazole, lansoprazole, and pantoprazole, respectively. Likewise, the AUC of diazepam is increased 28, 81, 6, 12, and <1%, respectively (Andersson et al., 1990, 2001; Ishizaki et al., 1995; Ogawa and Echizen, 2010). Such clinical data are in agreement with the hepatocyte data described herein (Table 3). Given these differences, an attempt was made to rationalize clinical findings based on in vitro-derived inhibition parameters.
It is accepted that integration of in vitro P450 inhibition data with in vivo data is difficult, and one has to consider multiple factors in any modeling exercise (Vuppugalla et al., 2010). Toward this end, inhibition data were obtained using three different model systems. In addition, PPI plasma protein binding was considered when determining % inhibitionpredicted, as were PPI plasma Cmax and Cmax,portal values. Finally, the pharmacokinetic profile of each PPI was also considered (see Materials and Methods). It is noteworthy that inhibitory metabolites, possible interactions between (R)- and (S)-forms of each racemic PPI, and accumulation in hepatocytes were not considered (Jones et al., 2004; Li et al., 2004, 2005; Ogilvie et al., 2011).
As summarized in Fig. 4, use of diazepam in the modeling exercise was useful, because it is a well established CYP2C19 probe and all six PPIs have been studied clinically as perpetrators. In addition, the degree of inhibition varies across the series of compounds, ranging from 79% (esomeprazole) and 46% (omeprazole) to <1% (pantoprazole and rabeprazole). With such a range, it was possible to leverage in vitro inhibition data and compare the seven different modeling approaches in an attempt to differentiate the PPIs and predict % inhibition in vivo. All seven methods presented esomeprazole as the most potent CYP2C19 inhibitor, whereas rabeprazole was predicted to be the weakest inhibitor. Furthermore, it was possible to differentiate esomeprazole from omeprazole, especially when fu-corrected static PPI concentrations or when time-dependent changes in total PPI concentration were considered (Fig. 4). The same two methods also correctly predicted that both lansoprazole and dexlansoprazole are weaker inhibitors versus esomeprazole and omeprazole. Of the PPIs studied, pantoprazole was the most problematic. The compound is known to be a weak inhibitor of CYP2C19 in vivo (<1%), even when plasma Cmax is high (57 μM) after an IV dose of 240 mg (Gugler et al., 1996). Only the semimechanistic model, with correction for fu, rendered a % inhibitionpredicted value that fell within the 90% CI reported in vivo (Fig. 4). Typically, pantoprazole is dosed orally (40 mg), and the Cmax,portal (∼7 μM) is calculated to be well below the Cmax reported by Gugler et al. (1996). When pantoprazole is added to serum coincubated hepatocytes, at such a low concentration, the degree of inhibition (≤5%) is more consistent with in vivo data (Table 3). It should be noted that, as reported by other authors (Li et al., 2004; Ogilvie et al., 2011), pantoprazole was shown to be a relatively weak inhibitor of HLM-catalyzed (S)-mephenytoin 4′-hydroxylase activity in our study (Table 2). For pantoprazole, this implies that a rCYP2C19-derived Ki,u is problematic and renders overestimates of % inhibition in vivo.
In conclusion, the results of the present study have shown that six PPIs are not potent reversible or metabolism-dependent inhibitors of P450s such as CYP2D6, -2C8, -2C9, -1A2, -2B6, and -3A4 in vitro. Of the P450s tested, the lowest IC50s were obtained with CYP2C19. In this regard, lansoprazole was the most potent inhibitor, whereas pantoprazole and rabeprazole were the weakest. Moreover, it was confirmed that two of the six PPIs (esomeprazole and omeprazole) are time-dependent inhibitors of CYP2C19. Despite the CYP2C19 inhibition potency rank in vitro, when one considers plasma protein binding and exposure after a dose, the integrated data set predicts that pantoprazole and rabeprazole will cause relatively minimal inhibition of CYP2C19. In contrast, it is anticipated that esomeprazole will exhibit the greatest inhibition of CYP2C19, more so than omeprazole, dexlansoprazole, and lansoprazole. When it comes to the inhibition of drug-metabolizing P450s, therefore, it may be possible to differentiate the various PPIs in terms of their ability to inhibit CYP2C19 and conclude that drug interactions involving the inhibition of this particular P450 are not a class effect (Angiolillo et al., 2011; Frelinger et al., 2012; Shah et al., 2012).
Authorship Contributions
Participated in research design: Zvyaga, Chang, Chen, and Rodrigues.
Conducted experiments: Zvyaga, Chen, Thorndike, Hurley, and Wagner.
Performed data analysis: Zvyaga, Vuppugalla, Chimalakonda, Chang, Yang, and Rodrigues.
Wrote or contributed to the writing of the manuscript: Zvyaga, Yang, Chimalakonda, Vuppugalla, Chen, and Rodrigues.
Acknowledgments
We thank Dr. Fabrice Hurbin (Sanofi-Aventis, Research and Development, Montpellier, France) for input and suggestions.
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:
- PPI
- proton pump inhibitor
- P450
- cytochrome P450
- HLM
- human liver microsomes
- r
- recombinant
- CEC
- 3-cyano-7-ethoxy-coumarin
- fu,inc
- free fraction in the incubation
- IC50
- concentration of inhibitor that decreases activity by 50% (not corrected for fu,inc)
- IC50(t)
- concentration of inhibitor that decreases activity by 50% after a preincubation time (t)
- IC50(u)
- IC50 corrected for fu,inc
- Cmax
- maximal concentration in plasma
- Cmax,u
- maximal concentration in plasma corrected for free fraction in plasma
- Cmax,portal
- maximal concentration in portal vein
- Cmax,portal,u
- maximal concentration in portal vein corrected for plasma protein binding
- DMSO
- dimethyl sulfoxide
- RF-MS/MS
- RapidFire-mass spectrometry
- KI
- inhibitor concentration that supports half the maximal rate of inactivation
- kinact
- maximal rate of inactivation
- [I]
- concentration of inhibitor
- Ki,u
- Ki corrected for fu,inc
- KI,u
- KI corrected for fu,inc
- kdeg
- rate of P450 (CYP2C19) holoenzyme degradation in the absence of inhibitor
- fa
- fraction absorbed
- ka
- absorption rate constant
- kobs
- rate constant for inactivation at a given concentration of inhibitor
- IV
- intravenous
- AUC
- area under the curve
- fm,2C19
- fraction cleared via CYP2C19
- PM
- poor metabolizer
- EM
- extensive metabolizer
- AUCPM
- AUC in poor metabolizers
- AUCEM
- AUC in extensive metabolizers
- AUCi
- AUC in the presence of inhibitor
- AUCc
- AUC in the absence of inhibitor
- CI
- confidence interval
- FR
- fractional activity remaining.
- Received March 7, 2012.
- Accepted May 30, 2012.
- Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics