Role of Hepatic Organic Anion Transporter 2 in the ... - ACS Publications

Feb 12, 2018 - of R- and S‑Warfarin: In Vitro Studies and Mechanistic Evaluation ... To further evaluate the potential role of hepatic uptake in R- ...
2 downloads 0 Views 3MB Size
Article Cite This: Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Role of Hepatic Organic Anion Transporter 2 in the Pharmacokinetics of R- and S‑Warfarin: In Vitro Studies and Mechanistic Evaluation Yi-an Bi, Jian Lin, Sumathy Mathialagan, Laurie Tylaska, Ernesto Callegari, A. David Rodrigues, and Manthena V. S. Varma* Pharmacokinetics, Dynamics, and Metabolism, Pfizer Global Research and Development, MS 8220-2451, Groton, Connecticut 06340, United States S Supporting Information *

ABSTRACT: Interindividual variability in warfarin dose requirement demands personalized medicine approaches to balance its therapeutic benefits (anticoagulation) and bleeding risk. Cytochrome P450 2C9 (CYP2C9) genotype-guided warfarin dosing is recommended in the clinic, given the more potent S-warfarin is primarily metabolized by CYP2C9. However, only about 20− 30% of interpatient variability in S-warfarin clearance is associated with CYP2C9 genotype. We evaluated the role of hepatic uptake in the clearance of R- and S-warfarin. Using stably transfected HEK293 cells, both enantiomers were found to be substrates of organic anion transporter (OAT)2 with a Michaelis−Menten constant (Km) of ∼7−12 μM but did not show substrate affinity for other major hepatic uptake transporters. Uptake of both enantiomers by primary human hepatocytes was saturable (Km ≈ 7−10 μM) and inhibitable by OAT2 inhibitors (e.g., ketoprofen) but not by OATP1B1/1B3 inhibitors (e.g., cyclosporine). To further evaluate the potential role of hepatic uptake in R- and S-warfarin pharmacokinetics, mechanistic modeling and simulations were conducted. A “bottom-up” PBPK model, developed assuming that OAT2−CYPs interplay, well recovered clinical pharmacokinetics, drug−drug interactions, and CYP2C9 pharmacogenomics of R- and S-warfarin. Clinical data were not available to directly verify the impact of OAT2 modulation on warfarin pharmacokinetics; however, the bottom-up PBPK model simulations suggested a proportional change in clearance of both warfarin enantiomers with inhibition of OAT2 activity. These results suggest that variable hepatic OAT2 function, in conjunction with CYP2C, may contribute to the high population variability in warfarin pharmacokinetics and possibly anticoagulation end points and thus warrant further clinical investigation. KEYWORDS: warfarin, organic anion transporter, CYP2C9, physiological based modeling, pharmacokinetics



INTRODUCTION Warfarin is an oral anticoagulant widely prescribed for the management of deep vein thrombosis, pulmonary embolism, atrial fibrillation, as well as stroke in high-risk patients.1,2 However, the frequency of bleeding events (7.6−16.5 per 100 patients per year3) and the risk of serious hemorrhage (1.3−4.2 per 100 patient exposure years4) complicate warfarin therapy. Interpatient variability in the response to warfarin makes its effective daily dose ranging from 0.5 to 60 mg.5,6 Clinical implementation of such a variable and narrow therapeutic index drug is therefore challenging, despite efforts that lead to international guidance on monitoring. Notably, warfarin dose requirements are individualized on the basis of the presence of © XXXX American Chemical Society

genetic variants in cytochrome P-450 2C9 (CYP2C9, related to pharmacokinetics) and vitamin K epoxide reductase (VKORC1, related to clot formation).7−9 Warfarin is a racemic mixture of R- and S-enantiomers, where R-warfarin is metabolized by CYP2C19 and other CYP enzymes while S-warfarin is predominantly metabolized by CYP2C9.10,11 Apparently, stereoselective pharmacological differences also exist with S-warfarin, suggested to be 3−5 Received: Revised: Accepted: Published: A

December 8, 2017 January 19, 2018 February 12, 2018 February 12, 2018 DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics times more potent than those for R-warfarin.12,13 Due to the genetic polymorphism in CYP2C9, associated with considerable variability in catalytic activity across different variants,14,15 and likely higher potency of S-warfarin, the pharmacodynamic variability and major hemorrhage risk are linked to highly variable S-warfarin pharmacokinetics.3,4,16,17 Clearly, studies have demonstrated that patients with lower dose requirements have a greater prevalence of low catalytic CYP2C9*2 and *3 allele variants.3,7,8,17,18 Additionally, carriers of homozygous CYP2C9*2/*2 and CYP2C9*3/*3 present a significantly reduced S-warfarin oral clearance compared to wild-type (CYP2C9*1/*1).17,19,20 However, interestingly, multiple studies also indicated that only about 20−30% of interpatient variability in S-warfarin clearance is associated with CYP2C9 genotype and that there is a significant correlation between Rwarfarin (not metabolized by CYP2C9) and S-warfarin clearance in all CYP2C9 genotype subgroups including CYP2C9*1/*1.17,19 On the basis of these clinical observations, we hypothesized that non-CYP pathways likely play an important role in the clearance of R- and S-warfarin. Hence, we evaluated the potential role of transporter-mediated hepatic uptake in the disposition and clearance of both warfarin enantiomers. Na +-taurocholate cotransporting polypeptide (NTCP, SLC10A1), organic anion transporter 2 (OAT2, SLC22A7), organic anion-transporting polypeptide 1B1 (OATP1B1, SLCO1B1), OATP1B3 (SLCO1B3), OATP2B1 (SLCO2B1), and organic cation transporter 1 (OCT1, SLC22A1) are major uptake transporters on the sinusoidal membrane of hepatocytes.21,22 OATPs/NTCP-mediated hepatic uptake plays an important role in the pharmacokinetics of several acidic and zwitterionic drugs.22−25 OCT1 was implicated in the import of cationic drugs like metformin.26 However, information on the role of hepatic OAT2 in drug disposition and clearance is limited.27 In the process of evaluating the uptake transporter’s role in the hepatic clearance of R- and S-warfarin, we conducted (i) in vitro transporter activity studies using transporter-transfected cells and primary human hepatocytes (PHHs), (ii) in vitro transporter phenotyping studies with a range of inhibitors to elucidate the molecular mechanisms driving uptake into human hepatocytes, and (iii) in vitro data-informed “bottom-up’ physiologically based pharmacokinetic (PBPK) modeling and simulations (assuming transporter−enzyme interplay in hepatic clearance) to quantitatively assess the significance of hepatic transport and/or metabolic activity on the clinical pharmacokinetics of R- and S-warfarin.

ketoprofen were purchased from Sigma-Aldrich (St. Louis, MO). Rosuvastatin was purchased from Sequoia Research Products Ltd. (Oxford, U.K.). [3H]-taurocholate was purchased from PerkinElmer Life Sciences (Boston, MA). [3H]-cGMP was purchased from American Radiolabeled Chemicals, Inc. (St. Louis, MO). In Vitro Transport Studies. Cell culture conditions and uptake experiments using HEK293 cells singly transfected with NTCP, OATP1B1, OATP1B3, OATP2B1, OCT1, or OAT2(tv-1) were similar to that previously reported.28,29 For uptake studies, the cells were washed three times with warm uptake buffer (HBSS with 20 mM HEPES, pH 7.4) and then incubated (0.5, 1, 2 min) with uptake buffer containing warfarin (S or R forms), [3H]-cGMP (1 μM), [14C]-metformin (10 μM), [3H]taurocholic acid (0.4 μM), or rosuvastatin (5 μM), on a heated plate shaker set to 37 °C and 150 rpm. Rosuvastatin, 3H-cGMP, 14 C-metformin, and 3H-taurocholate were used as positive controls for OATP1B1/1B3/2B1, OAT2, OCT1 and NTCP, respectively. Incubation time points were selected based on preliminary studies, where the accumulation in these cells is typically linear. Cellular uptake was terminated by washing the cells three times with ice-cold transport buffer and then lysed with 0.2 mL of 1% NP-40 in water or methanol containing internal standard. Uptake was determined either by mixing the cell lysate with scintillation fluid and analyzed by a liquid scintillation counter for radiolabeled compounds or by LC-MS/ MS system. The total cellular protein content was determined by using a Pierce BCA protein assay kit according to the manufacturer’s specifications. Plateable cryopreserved human hepatocytes were thawed and plated as described previously.30 Briefly, hepatocytes were thawed and seeded into 24-well collagen I coated plates with 0.5 mL at a density of ∼7 × 105 cell/mL using InVitro-HT and InVitro-CP media. Plated cells were cultured for 6 h before initiation of uptake experiments. Uptake rates were measured following preincubation with inhibitors or buffer for 10 min. The reactions were initiated by addition of R-/S-warfarin (0.2 μM) and probe substrates alone or with inhibitors. Rosuvastatin (5 μM), 3H-cGMP (1 μM), 14C-metformin (10 μM), and 3Htaurocholate (0.4 μM) were used as positive controls for OATP1B1/1B3/2B1, OAT2, OCT1, and NTCP, respectively. At predetermined time points (0.5, 1, 2, 5 min), reactions were terminated by washing the cells three times with ice-cold HBSS. These time points were selected based on the linearity of warfarin uptake in the preliminary studies. The cells were lysed with 100% methanol containing internal standard, and the samples were analyzed by LC-MS/MS. Analyses of R- and S-warfarin and rosuvastatin were performed by liquid chromatography/tandem mass spectrometry (LC-MS/MS). The HPLC consisted of Shimadzu LC20AD pumps and ADDA autosampler. Mobile phases were 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B). The gradient was maintained at 5% B for 0.2 min, followed by a linear increase to 95% B in 0.5 min and kept at 95% B for 0.3 min followed by a linear decrease to 5% B in 0.02 min. The column was equilibrated at 5% B for 0.48 min. The total run time for each injection was 1.5 min. The chromatographic separation was carried out on a Phenomenex Kinetex C18 100 Å 30 × 2.1 mm column with a flow rate of 0.8 mL/min (injection volume of 10 μL). Mass spectrometric detection was performed using an AB mass spectrometer, model Triple Quad 5500 (Applied Biosystems/Sciex, Thornhill, Ontario, Canada). The TurboIon-



MATERIALS AND METHODS Materials. Cryopreserved human hepatocyte lots, HH1027 (female, Caucasian, 59 year old), Hu8246 (female, Caucasian, 37 year old), and Hu4165 (female, Caucasian, 57 year old) were purchased from In Vitro ADMET Laboratories, LLC (Columbia, Maryland), Thermo Fisher Scientific (Carlsbad, CA), and Invitrogen (Carlsbad, CA), respectively. Human Embryonic Kidney (HEK) 293 cells stably transfected with human OATP1B1, OATP1B3, or OATP2B1 were generated at Pfizer Inc. (Sandwich, UK). HEK293 cells stably transfected with human NTCP, OCT1, and OAT2 were generated by the laboratories of Per Artursson (Uppsala University, Sweden), Kathleen Giacomini (University of California, CA), and Ryan Pelis (Dalhousie University, Canada), respectively. R-warfarin, S-warfarin, cyclosporine A, rifampicin, rifamycin SV, and B

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics Table 1. Summary of Input Parameters for R- and S-Warfarin Physiologically Based Pharmacokinetic Modelsa parameters Physicochemical Properties molecular weight (g/mol) log P compound type pKa fraction unbound ( f u,p) blood/plasma ratio (Rb) Absorption absorption type fraction absorbed MDCK-LE permeability (×10−6 cm/s) permeability calibrator - propranolol (×10−6 cm/s) unbound fraction in gut (fugut) Distribution distribution model Kp scalar Elimination CYP2C9 CLint,met (μL/min/mg-microsomal protein) CYP2C19 CLint,met (μL/min/mg-microsomal protein) CYP1A2 CLint,met (μL/min/mg-microsomal protein) renal elimination (%) Hepatobiliary transport liver unbound fraction intracellular extracellular PSpd (μL/min/106 cells) OAT2 Jmax (pmol/min/106 cells) OAT2 Km (μM)

R-warfarin inputs

S-warfarin inputs

308.3 2.7 monoprotic acid 5.1 0.013 0.59

308.3 2.7 monoprotic acid 5.1 0.013 0.59

ADAM 1.0 24 23 1

ADAM 1.0 24 23 1

refs 9, 36 measuredd measured

full PBPK 2

full PBPK 2

Rodgers et al. estimated based on oral data, assuming F ≈ 1b

source

measuredc calculated (MoKa 2.5.4)c measuredc measured

0.4

estimated from in vitro measurementse

0.14 0.05 0.003

0.05 0.003

in vivo data

0.69 0.025 1.1 335 10.4

0.69 0.025 0.87 169.6 7.3

model predicted model predicted measured (Figure 2) measured (Figure 2) measured (Figure 2)

a

Variables: ADAM, Advanced dissolution, absorption and metabolism model; P, partition coefficient; pKa, acid dissociation constant; Kp, tissue partition constant; fm, fraction metabolized; PSpd, intrinsic passive uptake clearance. bEstimated by fitting to oral plasma concentration−time profile data. See the Materials and Methods section. cLog P measured by the shake-flask method. MoKa software from Molecular Discovery, http://www. moldiscovery.com/software/moka. The unbound fraction was measured by the equilibrium dialysis method. dThe permeability of the racemic mixture was measured using MDCK-low efflux cells.34 eMeasured in pooled human hepatocytes by the Relay method37 and corrected for intracellular concentration using Kpuu measured by Izumi et al.’s method.38 Hepatocyte intrinsic clearance (μL/min/106 cells) was converted to microsomal intrinsic clearance (μL/min/mg-microsomal protein) using a physiological scalar: 106 cells is equivalent to 0.33 mg-microsomal protein (this corresponds to 120 × 106 cells per g liver and 40 mg-microsomal protein per g liver). fmCYP2C9 was assumed to be 90% for S-warfarin.39 No defined fmCYP2C19 value was available for R-warfarin in the literature; therefore, it was arbitrarily set at 75% considering reported semiquantitative data.11

clearance, and C is the incubation concentration. In the case of transport kinetics in HEK293 cells, the uptake rate in wildtype cells was subtracted from the uptake rate in OAT2trasfected cells at each concentration; therefore, PSpassive in eq 1 was assumed to be zero. Intrinsic clearance values generated using human hepatocytes and liver microsomes are reported in units of μL/min/106 cell and μL/min/mg microsomal protein, respectively. The in vitro intrinsic values were scaled using physiological scalars: 118 × 106 hepatocytes per g liver, 39.8 mg microsomal protein per g liver, 24.5 g liver per kg body weight (mean values employed in the healthy volunteer population file of Simcyp). PBPK Modeling and Simulations of Clinical DDIs. Whole-body PBPK modeling and simulations of clinical pharmacokinetics and DDIs of R- and S-warfarin were performed using a population-based ADME simulator, Simcyp (version 15.1, Certara, Sheffield, U.K.). Methodology adopted in model building for R- and S-warfarin is similar to that applied for OATP substrates.31−33 Physicochemical properties and in vitro data used along with the source are provided in Table 1. The advanced dissolution, absorption and metabolism (ADAM) model, informed with in vitro permeability data

Spray interface was operated in the negative ion mode at 5500 V and 500 °C. Quadrupoles Q1 and Q3 were set on unit resolution. Multiple-reaction-monitoring mode using specific precursor/product ion transitions was used for quantification. Detection of the ions was performed by monitoring the transitions of mass/charge ratio (m/z) with a collision energy of 30 eV as follows: warfarin (307 → 117) and tolbutamide (internal standard, IS) (269 → 170). Stock solutions of each warfarin enantiomer were prepared in DMSO and quantitated from standard curves ranging from 0.1 to 500 nM. Linear regression was fitted to data of warfarin standards using 1/x2 weighting. Data processing was performed using Analyst software (version 1.6.2, Sierra Analytics LLC). Data Analysis. An initial uptake rate based on the time course (typically 0.5−2 min) was used to obtain uptake clearance.30 Kinetic parameters of hepatic uptake in human hepatocytes were estimated using the following equation Uptake rate = PSpd ·C +

Vmax ·C Km + C

(1)

where Km and Vmax are the active transport affinity and maximum uptake rate, respectively. PSpd is the passive C

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Figure 1. Substrate activity of R- and S-warfarin toward hepatic uptake transporters in transporter-transfected HEK293 cells. Uptake of R-warfarin (A), S-warfarin (B), and positive controls (C) by NTCP, OAT2, OATP1B1/1B3/2B1, and OCT1 was measured at 0.5 μM concentration. The effect of various transporter inhibitors on the OAT2-mediated transport of R-warfarin (D), S-warfarin (E), and cGMP (F) was also evaluated. Data in bar graphs represent the uptake ratio in transporter-transfected cells versus wild-type cells (mean+s.d., n = 3). Horizontal lines represent an uptake ratio of unity. Positive controls used for OATP1B1/1B3/2B1, OAT2, OCT1, and NTCP are rosuvastatin (5 μM), 3H-cGMP (1 μM), 14C-metformin (10 μM), and 3H-taurocholate (0.4 μM), respectively. Insets show concentration-dependent OAT2-specific transport in HEK293 cells. CSA, cyclosporine; Rif, rifampicin; RIF SV, rifamycin SV; Keto, ketoprofen.

measured using MDCK-low efflux cells,34 was adopted to capture intestinal absorption and predict oral pharmacokinetics of R- and S-warfarin. Both enantiomers were shown to be substrates to BCRP;35 however, given that the observed bioavailability (F) of warfarin is complete in humans,9,36 the intestinal efflux mechanism was not considered in the model. For the formulation component, solution with no precipitation was assumed, supported by its complete F. Full-PBPK model using Rodgers et al. method considering rapid equilibrium between blood and tissues (except liver) was adopted to obtain the R-/S-warfarin distribution into all organs except the liver. The volume of distribution (Vdss/F) was underpredicted by the Rodgers et al. method, and therefore, a partitioning scalar of 2 was assigned to recover the observed pharmacokinetics in humans following oral dosing, assuming F ≈ 1. A permeabilitylimited model, assuming transporter−enzyme interplay, was considered to model hepatic disposition, wherein sinusoidal active uptake kinetics (Jmax and Km) and passive diffusion obtained from the current plated hepatocyte studies were incorporated. Intrinsic metabolic clearance was measured in pooled human hepatocytes by the relay incubation method37

and corrected for intracellular concentration with the unbound cell-to-media concentration (Kpuu) measured using Izumi et al.’s method.38 FmCYP2C9 was assumed to be 90% for S-warfarin, as reported previously.39 No defined fmCYP2C19 value was available for R-warfarin in the literature; therefore, it was arbitrarily set at 75% considering reported semiquantitative data.11 The effect of CYP2C9 reduced-function variants on the pharmacokinetics of S-warfarin was simulated assuming the catalytic CYP2C9 activity for *1/*3, *2/*3 and *3/*3 variants to be 60, 30, and 12% of wild-type (*1/*1), respectively.17,19,39 The effect of the CYP2C19 genotype on the pharmacokinetics of R-warfarin was not evaluated due to lack of definitive information on the genotype-dependent activity for CYP2C19. For DDI assessment, inhibitor models of fluconazole and cimetidine were directly adopted from the default compound library of the simulator (V15.1). These models were previously verified for their predictability of plasma concentration−time profiles and DDIs with probe substrates.40−42 PBPK models for amiodarone and its major circulating metabolite, monodesethyl-amiodarone, were reproduced from Chen et al.43 Minimal PBPK models for tienilic acid and miconazole were developed D

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Figure 2. Uptake transport characteristics of R- and S-warfarin in the PHHs. Effect of various transporter inhibitors on the uptake of 0.2 μM R- and S-warfarin (A,B) and probe substrates of OAT2 (C), NTCP (D), and OATPs+NTCP (E). Kinetics of active uptake of R- and S-warfarin in PHHs (F,G). Active uptake of R- and S-warfarin by three different lots of PHHs (H). Positive controls used for OATP1B1/1B3/2B1, OAT2, OCT1, and NTCP are rosuvastatin (5 μM), 3H-cGMP (1 μM), 14C-metformin (10 μM), and 3H-taurocholate (0.4 μM), respectively. Unless mentioned, all studies were carried using hepatocyte lot Hu8246. Data in the bar graphs represent mean ± s.d. (n = 3).

duration of R- and S-warfarin and inhibitor drugs were identical to those reported in the original clinical studies. The predictive performance of models was assessed by the “Rpredicted/observed value” [ = (mean predicted parameter)/(mean observed parameter)], with predefined acceptance criteria of 0.80− 1.25.46

using physicochemical properties, in vitro data, and estimates of Vdss and CL were obtained by fitting plasma concentration− time profiles.44,45 The inhibition potency of these drugs against CYP2C9, CYP2C19, and CYP1A2 were obtained from literature reports. The inhibition potency of OAT2 generated in the current study was also considered in the PBPK modeling, where applicable. Model input parameters and the source for values of the inhibitor drugs are provided in Supporting Information Table 1. Verification of tienilic acid and miconazole models in recovering the pharmacokinetics is presented in Supporting Information Table 2. The virtual population (10 × 10 trials) of healthy subjects with a body weight of ∼80 kg and age ranging from 18 to 65 years included both sexes. Dose, dosing interval, and dosing



RESULTS In Vitro Transport Activity Using Transfected Cells. Both R- and S-enantiomers of warfarin were evaluated in HEK293 cells singly transfected with six major hepatic uptake transporters (NTCP, OAT2, OATP1B1/1B3/2B1, and OCT1). Both enantiomers exhibited a high uptake ratio (i.e., ratio of uptake by transporter-transfected to wild-type cells) E

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Figure 3. PBPK model-based prediction of R- and S-warfarin clinical pharmacokinetics assuming OAT2-CYPs interplay in hepatic disposition. (A,B) Prediction of R- and S-warfarin pharmacokinetics following racemic warfarin oral dose. Data points represent dose-normalized (to 10 mg) mean observed values from several independent studies dosed between 7.5 and 30 mg. Solid and dashed lines represent mean and individual trials of model-predicted plasma concentration−time profiles. Insets show early time profiles up to 12 h. (C,D) Effect of multiple dose fluconazole (400 mg daily) on the oral pharmacokinetics of R- and S-warfarin. (E) S-warfarin pharmacokinetics in carriers of CYP2C9 genetic variants.20

respectively. Collectively, studies with transporter-transfected cells provide definitive evidence for OAT2-mediated transport of both R- and S-warfarin. R- and S-Warfarin Transport Characteristics in Primary Human Hepatocytes. Uptake by PHHs was evaluated for R- and S-warfarin and probe substrates of OAT2, OATPs, and NTCP (Figure 2). Consistent with findings from transfected cells studies, warfarin enantiomers showed uptake by PHH that was inhibitable by rifamycin SV (a pan-SLC inhibitor) and ketoprofen (an OAT2 inhibitor).30,47 However, no inhibition was evident with cyclosporine and rifampicin at concentrations well above those typically required to inhibit OATPs.23 On the contrary, cyclosporine and rifampicin markedly reduced uptake of taurocholate and rosuvastatin by human hepatocytes. Overall, the inhibition patternin the presence of several transporter inhibitors showed that the hepatic uptake of warfarin enantiomers is similar to that of the OAT2-selective substrate, cGMP, and

with OAT2; however, they did not register the substrate affinity for the other five solute carriers (SLCs) tested (Figure 1). We further studied the effect of various transporter inhibitors on the OAT2-mediated uptake in HEK293 cells (Figure 1D−F). Interestingly, only rifamycin SV (1000 μM), which is a panSLC inhibitor,30 and ketoprofen, an OAT2 inhibitor (IC50 ≈ 21 μM),47 inhibited the OAT2-specific uptake of both enantiomers. Cyclosporine (10 μM) and rifampicin (20 μM), which are known to inhibit OATPs at the concentrations tested,23 did not show any impact on the uptake of warfarin enantiomers. The inhibition pattern of both enantiomers with multiple transporter inhibitors was found to be similar to that of cyclic guanosine monophosphate (cGMP, an OAT2 -selective substrate27), although complete inhibition by rifamycin SV and ketoprofen was not achieved for R-warfarin (Figure 1D− F). For both enantiomers, concentration-dependent OAT2mediated transport conformed to single-Km Michaelis kinetics: Km of 12.5 ± 2.4 and 7.1 ± 1.5 μM for R- and S-warfarin, F

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Table 2. Summary of Bottom-up PBPK (OAT2−CYP2C9 Interplay) Model Predictions of Pharmacokinetics and the Impact of CYP2C9 Pharmacogenomics on Oral Exposure of R-/S-Warfarin warfarin pharmacokinetic parameters

predicteda

Oral Pharmacokinetics R-Warfarin oral AUC (μg·h/mL) oral Cmax (μg/mL) S-Warfarin oral AUC (μg·h/mL) oral Cmax (μg/mL) Impact of CYP2C9 Genotype S-Warfarin CYP2C9*1/*1 oral AUC (μg·h/mL) CYP2C9*1/*3 oral AUC (μg·h/mL) CYP2C9*2/*3 oral AUC (μg·h/mL) CYP2C9*3/*3 oral AUC (μg·h/mL)

observedb

Rpredicted/observed value

refs for observed 20 and 57−60

27.6 0.43

27.8 0.52

0.99 0.83

17.8 0.42

15.7 0.52

1.13 0.81 20

19.3 26.5 40.2 56.9

18.2 28.5 52.7 56.5

1.06 0.93 0.77 1.01

a Mean of 10 × 10 population simulations. bThe mean of mean values is represented when data from multiple studies are available. Data presented after dose-normalized to 10 mg of racemic warfarin, where necessary.

Table 3. Effect of Drugs Showing Clinically Significant or No DDIs with R- and S-Warfarin on the OAT2-Mediated Transport In Vitro and PBPK Model-Based Prediction of DDIs in vitro inhibition (%) (at 10 μM concentration) or IC50 (μM) inhibitor drug

cGMP HEK-OAT2 uptake

R-warfarin hepatocyte uptake

S-warfarin hepatocyte uptake

amiodarone cimetidine fluconazole

−5 −3 22

−1 17 0

7 0 0

miconazole ranitidine rifampicin (single dose) sulfinpyrazone tienilic acid

33 6 −5

33 36 12

14 31 10

IC50 = 66 μM IC50 = 5.7 μM

12 43

10 60

PBPK model predicted mean AUC ratio

observed mean AUC ratio

R-warfarin

S-warfarin

R-warfarin

S-warfarin

refs

1.02 1.02 1.88 (1.76−2.06) 2.11 a a

1.02 1.02 2.40 (2.26−2.54) 3.62 a a

1.22, 1.62 1.26 1.86

1.27, 2.10 1.03 2.54

61,62 63 64

1.66 1.05 1.02

4.74 1.01 0.90

65 63 57

0.98 1.08

1.70 2.92

66 67

b 1.00

b 2.87

a Ranitidine and Rifampicin show no significant inhibition of CYP1A2, CYP2C9, and CYP2C19 metabolism and OAT2-mediated uptake in vitro, and therefore, PBPK modeling was not conducted. bUnbound Cmax of sulfinpyrazone (∼1.5 μM)68 is well below the IC50 of OAT2 and CYP2C9 (∼85 μM),69 and cannot explain the observed AUC ratio of 1.7 for S-warfarin, and therefore, PBPK modeling was not conducted.

incorporated), models considerably underpredicted oral clearance of both enantiomers (Supporting InformationFigure 1) A set of inhibitor drugs reported to cause significant or no DDIs with R- and S-warfarin were examined for their effect on hepatic uptake of both enantiomers (Table 3). After a thorough literature search, the significant interactions (AUC ratio > 1.25) for warfarin reported are with amiodarone, fluconazole, miconazole, tienilic acid sulfinpyrazone, and cimetidine. We included all of these in our in vitro and modeling assessments and additionally included ranitidine and rifampicin, which showed lack of DDI in clinic as negative controls for their ability to inhibit OAT2 in vitro. Of the eight inhibitor drugs tested, only tienilic acid (IC50 ≈ 5.7 μM) and sulfinpyrazone (IC50 ≈ 66 μM) showed appreciable inhibition of OAT2mediated uptake in vitro. Model-based DDI predictions with fluconazole (CYP2C9 and CYP2C19 inhibitor) were able to recover the changes in plasma R- and S-warfarin concentration (Figure 3C,D). Additionally, we developed inhibitor models for tienilic acid (time-dependent inhibitor of CYP2C9 and OAT2) and miconazole (inhibit CYP2C9 and CYP2C19) and simulated their effect on R- and S-warfarin exposure. Here, the magnitude of the interaction with tienilic acid and miconazole is in reasonable agreement with the observed area

distinct from the NTCP and OATPs/NTCP probe substrates, taurocholate and rosuvastatin, respectively (Figure 2A−E). Kinetic studies revealed saturable active uptake with Km values of 10.4 ± 1.4 and 7.3 ± 1.4 μM and Vmax values of 335.1 ± 18.4 and 169.6 ± 11.3 pmol/min/mg-protein for R- and S-warfarin, respectively (Figure 2F,G). Finally, total uptake clearance is within 20% of the mean across three separate lots of PHHs (Figure 2H). Model-Based Prediction of Warfarin Pharmacokinetics, DDIs and CYP2C9 Pharmacogenomics. A “bottom-up” full PBPK model considering OAT2−CYPs interplay in hepatic disposition was developed for R- and S-enantiomers individually. Models were informed with in vitro transporter kinetic data obtained using PHHs (Figure 2F,G) and metabolic intrinsic clearance estimated from incubations of pooled human hepatocytes after correcting for free cell-to-media concentrations (Kpuu) (Table 1). These bottom-up models recovered the plasma concentration−time profiles of each enantiomer reasonably well, with the estimated pharmacokinetic parameters in good agreement with the observed clinical oral pharmacokinetic data (Figure 3, Table 2). In contrast, when only CYPmediated metabolism was assumed (no hepatic uptake kinetics G

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Figure 4. Model-based sensitivity analysis of the impact of intrinsic OAT2-mediated active uptake clearance and intrinsic CYP2C19 or CYP2C9 clearance on the plasma clearance of (A) R-warfarin and (B) S-warfarin. Insets are enlargement of initial profiles.

and clinical efficacy/safety of warfarin and warrant further clinical investigation. Our initial in vitro studies showed significant active uptake of R- and S-warfarin by PHHs, wherein rifamycin SV, at a concentration (1000 μM) high enough to inhibit six major hepatic uptake transporters (NTCP, OAT2, OATP1B1/1B3/ 2B1, and OCT1),30 reduced the uptake of both enantiomers to about 10−18% of the control (Figure 2). In an effort to identify the key membrane transporters involved, singly transfected expression (HEK293 cells) systems revealed OAT2-selective transport that is inhibited by certain inhibitors (rifamycin SV 1000 μM and ketoprofen) but not by OATP/NTCP inhibitors (cyclosporine and rifampicin). The inhibition pattern for Rand S-warfarin with human hepatocytes resembled that obtained with the HEK293-OAT2 cell, implying that OAT2 is the major uptake transporter. Additionally, the transport characteristics of R- and S-warfarin in human hepatocytes are similar to that of an OAT2 probe substrate (cGMP) but markedly distinct from the probe substrates of NTCP (taurocholate) and OATPs (rosuvastatin). Finally, the transport affinity (Km ≈ 10 μM for R-warfarin and ∼7 μM for Swarfarin) appears to be similar between the two systems. While OAT2 Km values are well above the unbound Cmax of R- and Swarfarin, it should be noted that their Km for 7-hydroxylation in human liver microsomes (a major pathway via CYP2C) is also in the range of ∼5 μM or above.48 This implies that substrate interaction with OAT2 and CYP2C is somewhat similar at clinically relevant concentrations. Overall, these comprehensive in vitro data provide definitive evidence in support of OAT2mediated hepatic uptake for both enantiomers of warfarin. On the basis of the well-documented evidence, warfarin dose requirements can be individualized considering the presence of genetic variants in CYP2C9 and vitamin K epoxide reductase (VKORC1 related to clot formation).7,8 S-warfarin is primarily metabolized by CYP2C9, while R-warfarin is metabolized by CYP2C19 and CYP1A2.10,11 Clearance of S-warfarin is therefore associated with CYP2C9 genetic variants, with clearance significantly dropping particularly in subjects homozygous for *2, and *3 alleles.17,19,20 However, results from several clinical studies have implied that (i) a considerable portion of the patient population (about 30%) requiring low maintenance doses and having low warfarin clearance are homozygous for *1/*1 (wild-type),17,19 (ii) patients with a wild-type genotype show a wide range (almost 12-fold) of clearance values, which is overlapping that of most patients carrying mutant alleles,17,19 (iii) nongenotype factors like age contribute more to the variability in warfarin dose requirement

under the plasma concentration−time curve (AUC) ratios (Table 3). However, DDI with amiodarone was notably underpredicted (although observed data is variable across separate clinical studies), which may be due to complexity in capturing the interactions associated with amiodarone and its major metabolite pair.43 A probe OATP inhibitor, single-dose rifampicin (600 mg), did not alter the oral clearance of warfarin enantiomers due to the lack of OAT2 and CYPs inhibition. On the basis of the in vitro data, OAT2 inhibition did not contribute to the observed interaction for all eight inhibitor drugs evaluated. The effect of CYP2C9 reduced-function variants on the pharmacokinetics of S-warfarin, following a single oral dose of racemic warfarin, was also reasonably predicted by the mechanistic model, assuming the catalytic CYP2C9 activity for *1/*3, *2/*3, and *3/*3 variants to be 60, 30, and 12% of wild-type (*1/*1), respectively.17,19,39 Overall, the observed clinical pharmacokinetics, DDIs, and CYP2C9 pharmacogenomics can be quantitatively described primarily considering on OAT2−CYP2C9 interplay for the S-enantiomer and OAT2− CYP2C19 interplay for the R-enantiomer. Clinical data were not available to directly evaluate the impact of OAT2 inhibition on warfarin pharmacokinetics; however, the verified models were further used to simulate the effect of change in intrinsic clearance mediated by CYP2C9/CYP2C19 and OAT2 on oral clearance of both enantiomers (Figure 4). It was determined that a change in OAT2 function has a linear impact on the oral clearance for R- and S-warfarin. On the other hand, inhibition of CYPs shows a linear reduction in oral clearance, while induction of CYPs causes a less than proportional change in clearance. Finally, simultaneous changes in the function of both OAT2 and CYP2C, with similar magnitude, have a marked effect on clearance of R- and S-warfarin.



DISCUSSION High interindividual variability in the pharmacokinetics of warfarin is a major complication in managing its therapy with minimal risk, and thus, a thorough understanding of the mechanisms involved in its clearance is critical in support of a personalized dose strategy. Collective findings from our in vitro studies, along with the mechanistic PBPK modeling and simulations, suggest a potential role for OAT2-mediated hepatic uptake in determining the pharmacokinetics of both R- and S-enantiomers of warfarin. This previously unrecognized clearance mechanism, in conjunction with CYP2C9/2C19, may contribute to the interpatient variability in pharmacokinetics H

DOI: 10.1021/acs.molpharmaceut.7b01108 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics and plasma exposure of both R- and S-enantiomers than the CYP2C9 genotype,18 and (iv) more importantly, there is good correlation between clearance of both enantiomers following dosing of racemic warfarin within the population carrying CYP2C9 wild-type (*1/*1) as well as reduced-function allelic variants. For example, Scordo et al. reported a linear correlation between R- and S-warfarin free plasma clearance with a coefficient of determination (R2) of 0.75 in CYP2C9 wild-type and 0.39 in subjects carrying mutated allele.17 Herman et al. showed similar correlations in CYP2C9*1/*1 (R2 ≈ 0.46), CYP2C9*1/*X (∼0.58), and CYP2C9*X/*X (∼0.62) subjects.19 From these collective clinical observations and prior knowledge of metabolic pathways, we hypothesized that (i) disposition mechanisms other than CYP metabolism contribute significantly to the clearance of both R- and S-enantiomers and (ii) the major clearance mechanism of the two enantiomers is the same or closely related. Our findings of the OAT2-mediated active hepatic uptake mechanism can explain the above-noted gaps in current understanding of warfarin clinical pharmacokinetics. We employed bottom-up PBPK modeling to describe pharmacokinetics of R- and S-warfarin considering permeability-limited (OAT2−CYPs interplay) hepatic disposition, which in static terms can be expressed by the extendedclearance concept.23,24 Assuming no active basolateral and canalicular efflux, hepatic intrinsic clearance (CLh,int) can be derived using uptake (CLuptake,int), passive (CLpassive), and metabolic (CLmet,int) intrinsic clearances. CL h,int =

CLuptake,int CLpassive + CLmet,int

with a larger data set of OAT2 substrate drugs are warranted to verify the reasonable in vitro−in vivo extrapolation noted here. Admittedly, clinical data were not available to directly verify the quantitative role of OAT2 in R- and S-warfarin pharmacokinetics. However, on the basis of the OAT2− CYP2C interplay, the bottom-up PBPK models were able to quantitatively describe the pharmacokinetics of R- and Swarfarin (Figure 3A,B, Table 2). In contrast, when only CYPmediated metabolism was assumed (no hepatic uptake kinetics incorporated), bottom-up models considerably underpredicted oral clearance of both enantiomers (Supporting Information Figure 1). Overall, this provides further evidence for the clinical significance of OAT2 in hepatic clearance of warfarin enantiomers. In vitro studies suggest that the uptake by PHHs is 30% higher for R-warfarin, but S-warfarin intrinsic metabolic clearance is ∼2.4-fold higher than that of the R-enantiomer. This difference in metabolic intrinsic clearance could explain the relatively higher systemic clearance of S-warfarin. The PBPK model also well predicted the DDIs with fluconazole, miconazole, and tienilic acid, which show a significant increase in warfarin AUCs. With the lack of meaningful in vivo OAT2 inhibition based on in vitro IC50 data, these DDIs are explained primarily by CYP2C9/2C19 inhibition (Table 3). Inhibitor drugs with no or weak inhibition of OAT2 and CYP2C in vitro did not alter clinical pharmacokinetics of R- and S-warfarin. Additionally, the effect of CYP2C9 reduced-function variants on S-warfarin clearance is well described by the model. Although hepatic uptake plays a significant role in the clearance of Swarfarin, the mechanistic model supports the current guidance on using S-warfarin as a sensitive and selective clinical probe for assessing CYP2C9 activity.50,51 However, caution should be applied in the conduct and interpretation of such clinical studies as we note that simultaneous inhibition of OAT2 and CYPs could result in a higher change in S-warfarin exposure in the clinic (Figure 4). We recently reported similar findings for tolbutamide, which is also a recommended clinical probe for CYP2C9 activity.52 Knowledge regarding the clinical relevance of OAT2 (SLC22A7) in drug disposition is very limited, although recent evidence points to its involvement in the handling of physiologically important endogenous compounds, like creatinine and cGMP.27 To the best of our knowledge, this is the first study demonstrating the potential role of hepatic OAT2 in the pharmacokinetics of a drug. Due to the lack of clinically relevant OAT2 inhibitors and limited knowledge regarding the functional polymorphic variant(s) of SLC22A7 (gene encoding OAT2), it will be currently challenging to clinically assess the contribution of OAT2 to the pharmacokinetic variability of Rand S-warfarin. For example, the expression of OAT2 in human liver samples is known to vary ∼10-fold but is not associated with SLC22A7 genotype.53 From the standpoint of DDIs, none of the inhibitor drugs studied in combination with warfarin in the clinic present as significant OAT2 inhibitors (Table 3). A review of the literature suggests that indomethacin can at least partially (∼30%) inhibit OAT2 at clinically relevant concentrations.29 Although there are no clinical DDI data available, indomethacin has been reported to greatly prolong the prothrombin time of warfarin. 54 On the other hand, interferon-α2b is known to potentiate warfarin activity and has been shown to down-regulate OAT2 mRNA and protein expression in human hepatocytes in vitro.55,56 While there is a need for new tools (e.g., clinical probe inhibitors) to further

·CLmet,int

For warfarin enantiomers, CLpassive (∼105 μL/min/g liver; unit conversion of values in Table 1) is much higher than CLmet,int (