Resolving the Distribution−Metabolism Interplay of Eight OATPSubstrates in the Standard Clearance Assay with Suspended HumanCryopreserved HepatocytesPar Nordell, Susanne Winiwarter, and Constanze Hilgendorf*

Drug Safety and Metabolism, AstraZeneca R&D Molndal, Pepparedsleden 1, SE-431 83 Molndal, Sweden

*S Supporting Information

ABSTRACT: Uptake transporters may act to elevate theintrahepatic exposure of drugs, impacting the route and rate ofelimination, as well as the drug−drug interaction potential. Wehave here extended the assessment of metabolic drug stabilityin a standard human hepatocyte incubation to allow forelucidation of the distribution−metabolism interplay estab-lished for substrates of drug transporters. Cellular concen-tration−time profiles were obtained from incubations of eightknown OATP substrates at 1 μM, each for two different 10-donor batches of suspended cryopreserved human hepatocytes. Thekinetic data sets were analyzed using a mechanistic mathematical model that allowed for separate estimation of active uptake,bidirectional diffusion, metabolism and nonspecific extracellular and intracellular binding. The range of intrinsic clearancesattributed to active uptake, diffusion and metabolism of the test set spanned more than 2 orders of magnitude each, with medianvalues of 18, 5.3, and 0.5 μL/min/106 cells, respectively. This is to be compared with the values for the apparent clearance fromthe incubations, which only spanned 1 order of magnitude with a median of 2.6 μL/min/106 cells. The parameter estimates ofthe two pooled 10-donor hepatocyte batches investigated displayed only small differences in contrast to the variability associatedwith use of cells from individual donors reported in the literature. The active contribution to the total cellular uptake ranged from55% (glyburide) to 96% (rosuvastatin), with an unbound intra-to-extracellular concentration ratio at steady state of 2.1 and 17,respectively. Principal component analysis showed that the parameter estimates of the investigated compounds were largelyinfluenced by lipophilicity. Active cellular uptake in hepatocytes was furthermore correlated to pure OATP1B1-mediated uptakeas measured in a transfected cell system. The presented approach enables the assessment of the key pathways regulating hepaticdisposition of transporter and enzyme substrates from one single, reproducible and generally accessible human in vitro system.

KEYWORDS: drug transporters, modeling, statins, kinetics, uptake, hepatocytes

■ INTRODUCTION

Isolated human hepatocytes represent a physiologically relevantexperimental platform for the evaluation of liver-relatedmetabolism, toxicity and drug−drug interactions, of wide usein the preclinical phase of drug discovery programs.1 In nearlyall new chemical entities’ characterization the in vitro measureof the metabolic stability obtained from hepatocyte incubationsis a key component in the prediction of the pharmacokineticprofile.2−6 For the majority of oral drugs, passive membranepermeability is sufficient to quickly equilibrate the unboundextra- and intracellular concentrations of drug. As a result theapparent clearance from an in vitro incubation will approximatethe intrinsic metabolic clearance driven by the unboundconcentration of the substrate at the site of the enzymaticprocess.7 However, it is now widely recognized that druguptake mediated by membrane-associated transporter proteinscan have clinically relevant effects on drug disposition in vivo aswell as at the cellular level in vitro.8−10 Members of the organicanion transporter polypeptide (OATP) family have beenidentified as important uptake transporters expressed in theliver, with a substrate profile that covers a broad range of

amphiphilic xenobiotics.11,12 The OATP-mediated transportcontributes for instance to the hepatoselectivity of HMG-CoAreductase inhibitors (“statins”).13−16 A dependency on activeuptake is, however, also associated with a certain susceptibilityto drug−drug interactions and variability caused by geneticpolymorphisms, with the particular attention of regulatoryauthorities.17 With optimization strategies to improve metabolicstability becoming more successful, distribution mediated bytransport processes will continue to have an increasing impacton elimination, emphasizing the need to assess transportcharacteristics early on in drug discovery.18 Ideally, suchcharacterization would comprise the drug clearance processesattributed to transporters and metabolic enzymes in the same invitro study, and additionally allow for a reproducible use, thusfacilitating optimization during the discovery phase.

Received: April 26, 2013Revised: August 26, 2013Accepted: October 8, 2013Published: October 8, 2013

Article

pubs.acs.org/molecularpharmaceutics

© 2013 American Chemical Society 4443 dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−4451

The principal manner in which active transport affects theclearance from a conventional hepatocyte incubation is wellillustrated by decomposing the apparent clearance viewed fromthe unbound medium concentration CLmed into intrinsicprocesses at steady-state conditions.19−21

= ×+

+ +CL CL

CL CL

CL CL CLmed int ,metint ,up int ,diff

int ,diff int ,eff int ,met

(1)

where CLint,up, CLint,diff, CLint,eff and CLint,met are the intrinsicclearances of active uptake, passive diffusion, active efflux andmetabolism, respectively. While biliary excretion is one of theprimary mechanisms of elimination in vivo, it is not clear towhat extent the function of efflux transporters is retained afterisolation.22−24 In Figure 1 the apparent clearance from the

medium is calculated from eq 1 (neglecting active efflux). Fromsuch an illustration it is obvious how the same observedclearance may have a diverse mechanistic origin, resulting fromthe interplay of active uptake, diffusion and metabolism.To be able to define an in vitro kinetic profile of actively

transported drugs according to eq 1 the approach typicallycomprises (1) experiments in which the cellular and mediafractions are separated and (2) analysis based on physiologicallyinspired mechanistic models that decompose flux between theextra- and intracellular compartment into a unidirectionaltransporter-driven and a bidirectional passive part.25−28 Theconcepts were effectively applied by Paine et al.,29 who studiedthe overall kinetics of atorvastatin, cerivastatin and indometha-cin in rat hepatocyte suspensions and described theextrapolation of data to the in vivo pharmacokinetics in therat. With a larger set of 16 known transporter substrates Yabe etal. developed a saturable uptake model to investigate the activeuptake phase in suspended rat hepatocytes.30 Reported fromthe same lab, a comprehensive evaluation of the uptake kineticsfor 7 OATP substrates with plated rat31 and plated humancryopreserved hepatocytes32 gave a mechanistic background tointerspecies discrepancy observed on the transporter level.Importantly, the latter study highlighted how humaninterdonor variability in uptake transporter activity maycompromise the predictive value of the in vitro estimates: forrosuvastatin, relying extensively on carrier mediation, active

uptake clearance varied more than 6-fold between the donorsinvestigated.In the present study, pooled cryopreserved hepatocytes were

selected, as to gain a general understanding independent frominterindividual differences. This paradigm made pooledhepatocytes the in vitro system most widely applied throughoutdrug discovery programs that allow for reproducibility in theassessment of metabolic properties of new chemical entities. Weanalyzed the apparent depletion of eight OATP substrates withdifferent degrees of metabolism in two batches of pooledhuman cryopreserved hepatocytes in suspension. A mechanistickinetic model, similar to that applied by Paine and co-workers,29 enabled the extraction of the intrinsic processespassive diffusion, active uptake and metabolism. This modelallowed for detailed understanding of underlying mechanismscritical to the identification of the key pathways regulatinghepatic disposition of transporter and enzyme substrates fromone single, generally accessible human in vitro system.

■ EXPERIMENTAL SECTIONChemicals. Valsartan, pravastatin, bosentan, atorvastatin

and glyburide were obtained as 10 mM DMSO stock solutionsfrom the AstraZeneca Compound Management Team inMolndal, Sweden, while rosuvastatin, fexofenadine andpitavastatin were obtained in solid form from Chemtronica,Sweden. The hepatocyte suspension medium (HSM) wasprepared by supplementing Williams Medium E (Sigma-AldrichResearch, St. Louis, MO, USA) with 25 mM HEPES and 2 mML-glutamine (pH 7.40). Mineral and silicon oil (Sigma-AldrichResearch, St. Louis, MO, USA) were mixed to give a finaldensity of 1.015 g/mL for use in the oil-spin experiments.

Assessment of Hepatocyte Concentration−Time Pro-files Using an Oil-Spin Procedure. The two batches (UMJand IRK) of pooled human cryopreserved hepatocytes wereobtained from Celsis In Vitro Technologies (Brussels,Belgium). Beckman 0.5 mL microtubes (Fisher ScientificGTF AB, Sweden) were prepared ahead of incubation startby addition of 15 μL of 4% cesium chloride (CsCl) and 140 μLof the oil mixture followed by spinning at 4000g for 2 min.Compound DMSO stocks were on the day of each experimentdiluted in HSM to 2 μM (0.1% DMSO). Cryopreservedhepatocytes were thawed and resuspended in HSM accordingto the manufacturer’s recommended procedures. Cell densityand viability (consistently >82%) were determined using aCASY cell counter (Innovatis AG, Germany). The cellsuspension was diluted to 3.2 million cells/mL, and experi-ments were started by mixing equal volumes of prewarmedcompound and cell solutions in a glass vial to yield 1 μM drugand 1.6 × 106 cells/mL in the final incubation. The temperaturewas controlled using a water bath set to 50 rpm (linear shaking)at 37 °C. At selected time points (typically 15, 30, 45 s, 1, 2, 3,5, 15, 30, 60, 90, and 120 min) a 100 μL aliquot was removedfrom the incubation, dispensed into a microtube prepared withoil and immediately centrifuged at 7000g for 15 s to separatecells from medium using a benchtop Eppendorf MiniSpincentrifuge equipped with the appropriate rotor/adapters.Control incubations at 4 °C were performed for eachcompound to account for the passive processes, diffusion andnonspecific binding. Medium samples were immediately takenfrom the supernatant above the oil layer, whereafter the tubescontaining the cell pellets were directly put on dry ice. Cellsamples were prepared for analysis by cutting frozen tube tipscontaining the cell pellets below the oil/CsCl interface into a

Figure 1. CLmed surfaces calculated from eq 1, varying CLint,up andCLint,diff from 0.1 to 100 μL/min/106 cells (CLint,eff = 0) at threeCLint,met levels (sheets from top to bottom represent CLint,met = 10, 1,and 0.1 μL/min/106 cells).

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514444

96-well plate. 200 μL of methanol containing warfarin as avolume marker was added to each well, and the plate was mixedat room temperature using a plate shaker for 1 h. After 100 μLof water had been added to each well, the plate was centrifugedat 4000g at 4 °C for 20 min. The supernatants were, togetherwith samples from the medium fraction, diluted to attain thesame methanol concentration, and finally analyzed by a LC/MS/MS system consisting of an HTS PAL injector (CTCAnalytics, Zwingen, Switzerland), HP 1100 LC binary pump(Agilent Technology, Germany) equipped with a reversed-phase C18 Atlantis T3 column (Waters Corp., USA) and triplequadrupole mass spectrometer API4000 (Applied Biosystems/MDS Sciex, Canada). For each combination of drug andhepatocyte batch (UMJ and IRK), data was collected fromduplicate incubations at 37 °C and 4 °C run in parallel at thesame test occasion.Mechanistic Mathematical Model To Elucidate the

Underlying Processes of Drug Distribution, Binding andMetabolism. The sets of kinetic data from the two batches ofcryopreserved hepatocytes were evaluated in terms of theunsaturable mechanistic model as outlined in Figure 2. In the

model, the incubation volume is divided into three compart-ments, a medium, an intracellular compartment and an outermembrane compartment, similar to the model used by Paineand co-workers.29 The drug quantified from the cellular fractionwas considered to include drug in the intracellular and themembrane compartments.Extracellular binding was assumed to be dominated by drug

associating with the outer surface of the cellular membrane,described as a rapid equilibrium established between unbounddrug in the medium (Dmed,u) and drug in the membranecompartment (Dmem):

=×

K[D]

[D] [M]memmem

med,u (2)

where [M] denotes the concentration of membrane withrespect to the medium volume of unit million cells/mL andKmem the membrane association constant. If fucell is the fractionof unbound drug inside the hepatocytes, the unboundintracellular concentration [D]cell,u is related to the totalintracellular concentration [D]cell,tot by eq 3:

= ×[D] [D] fucell,u cell,tot cell (3)

Equations 4 and 5 define the change of the unbound drugconcentration in the cell (of volume Vcell) and the medium (ofvolume Vmed) compartments, respectively, with time:

= × − + ×t V V

d[D]

dCL

[D](CL CL )

[D]cell,uin

med,u

cellout int ,met

cell,u

cell

(4)

= − × + ×t V V

d[D]

dCL

[D]CL

[D]med,uin

med,u

medout

cell,u

med (5)

where CLin and CLout represent the sum of clearancesassociated with drug transport from the medium into thecells and from the cells to the medium, respectively, andCLint,met is the intrinsic clearance of drug due to metabolism inthe cellular compartment (Figure 2). At steady-state conditions,when [D]cell,u/dt = 0 (eq 4), the unbound cell-to-mediumconcentration ratio [D]cell,u/[D]med,u, described by the partitioncoefficient Kp,uu, can be calculated from the clearanceestimates:19,20

= =+

K[D]

[D]CL

CL CLtp,uu

cell,u

med,u

in

ou int ,met (6)

In a general description of transport in the system, CLin andCLout would both comprise an active and a passive component.However, since the localization of canalicular drug transportproteins may not be retained in the suspended hepatocytemodel, active efflux from the cellular compartment to themedium is assumed to be limited.24 If CLint,up and CLint,diffdescribe intrinsic clearance due to active uptake and bidirec-tional passive diffusion, respectively, CLin and CLout are thengiven by

= +CL CL CLin int ,diff int ,up (7)

=CL CLout int ,diff (8)

Taking into account extracellular (eq 2) and intracellular (eq3) binding, numerical integration of eqs 4 and 5 gives theconcentration profiles of the cellular and the mediumcompartments. Using the nonlinear least-squares solver of thecommercial software package Matlab 7.12 (MathWorks Inc.,Natick, MA, 2011), profiles were simultaneously fitted to datameasured at 37 and 4 °C data from both the UMJ and the IRKbatches of cells. At 4 °C processes associated with transporteror enzymatic activity, CLint,up and CLint,met, were consideredinactivated. The fitting procedure allowed the values of thesetwo adjustable parameters, CLint,up and CLint,met, to be variedindividually with batch and temperature, while parametersdescribing passive diffusion and binding (CLint,diff, Kmem andfucell) were restricted to one common estimate throughout allfour sets of data. It was observed that medium-loss data showedlarger variability than data from the cellular fraction, inparticular for the low turnover drugs. Medium data wastherefore excluded from the model fitting procedure and onlyused as an additional control of the mass balance. Also, only the4 °C data obtained at approximate steady-state conditions(here generally regarded at >60 min after incubation start) wereused in the parameter fit.The structural identifiability of model parameters was

confirmed using the Matlab-based toolbox GenSSI (GeneratingSeries Approach for Testing Structural Identifiability).33

Figure 2. Schematic representation of the proposed mechanisticmodel comprising a medium, a cellular compartment and an outer cellmembrane compartment. Transport of drug from the medium into thecell and out from the cell to the medium is described by CLin andCLout, respectively. Kmem and fucell describe binding to the outer cellmembrane and intracellular binding, respectively. The cellular fractionscollected experimentally include the cellular and the membrane modelcompartments (shaded area).

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514445

Concentration profiles were calculated based on theassumption of a cellular volume Vcell of 4.0 μL/106 cells.34

The variances of parameter estimates were obtained from theJacobian evaluated at the point estimates of the parameters.35

The sum of simulated concentration−time profiles in themedium, cell and membrane compartments at 1 million cells/mL represents a total depletion curve typically obtained inconventional hepatocyte incubations, from which an estimate ofthe clearance from the incubation (CLinc) can be calculated bylinear regression.

Assessment of Drug Uptake into OATP1B1-Express-ing HEK-293 Cells. Stably transfected OATP1B1-expressingand empty vector (Mock) transfected HEK-293 cells wereobtained from the AZ-cell bank. The cell line has beengenerated in-house and validated as published previously.36

Cells were cultured in Dulbecco’s modified Eagle medium(DMEM, Sigma-Aldrich Research, St. Louis, MO, USA),supplemented with 10% fetal bovine serum, 4 mM L-glutamine,and 1 mg/mL Geneticin. Transport studies were performed oncell monolayers in Hank’s balanced salt solution (HBSS, Gibco,Life Technologies Europe) set to pH = 7.40 in poly-D-lysine

Figure 3. (Top) Cellular amount (intracellular + membrane bound) of transporter substrates rosuvastatin, pitavastatin and glyburide in the 100 μLsamples collected from 1 μM incubations at 1.6 million cells/mL. Black symbols indicate experimentally observed values at 37 °C for the UMJ (△)and the IRK (▽) batches, for which the model best-fit simulated profiles are represented by the solid and the dashed line, respectively. Gray symbolsrepresent observed values at 4 °C, with the fitted steady-state amount given by the dotted line. Data for each batch and temperature were obtainedfrom two parallel incubations. (Bottom) Simulated total (solid) and unbound (dashed) intracellular concentration at 1 million cells/mL plottedtogether with the medium concentration profile (dotted) for the UMJ batch.

Table 1. Best-Fit Parameter Estimates for the Set of Transporter Substrates, Ordered from Low to High Lipophilicity (SE inParentheses)a

drug LogD7.4b

CLint,up (μL/min/106 cells)

CLint,diff (μL/min/106 cells)

CLint,met (μL/min/106 cells) fucell

Kmem (mL/106

cells) Kp/Kp,uu

CLinc(μL/min/106

cells) f up.active (%)

OATP1B1c

(pmol/min/mg protein)

valsartan −2.1 4.6 (0.16) 0.42 (0.055) 0.15 (0.031) 0.45 7.0 × 10−3 19/9.0 1.9 92 12 (1.4)4.9 (0.17) 0.10 (0.022) (0.039) (0.43 × 10−3) 23/10 1.4 92

pravastatin −0.77 3.1 (0.21) 0.70 (0.13) 0.37 (0.056) 0.87 5.1 × 10−3 4.1/3.5 2.0 81 2.9 (0.55)4.0 (0.26) 0.78 (0.074) (0.057) (0.3 × 10−3) 3.7/3.2 3.9 85

rosuvastatin −0.44 12 (0.49) 0.53 (0.10) 0.11 (0.022) 0.53 4.8 × 10−3 38/20 2.8 96 4.7 (0.40)7.8 (0.42) 0.086 (0.023) (0.098) (0.95 × 10−3) 26/14 1.6 94

fexofenadine 0.47 0.75 (0.067) 0.24 (0.029) 0d 0.46 7.0 × 10−3 8.8/4.2 0 76 0.20 (0.092)0.65 (0.086) 0d (0.046) (0.40 × 10−3) 7.8/3.7 0 73

pitavastatin 0.91 150 (11) 13 (2.1) 1.1 (0.16) 0.058 0.056 190/11 8.8 92 64 (2.0)120 (9.2) 1.2 (0.17) (0.0074) (0.0098) 150/8.9 8.5 90

bosentan 1.0 32 (3.9) 9.9 (2.1) 1.0 (0.15) 0.041 0.096 93/3.8 3.7 76 14 (1.0)24 (3.9) 0.85 (0.20) (0.0038) (0.0075) 77/3.2 2.6 71

atorvastatin 1.1 61 (6.8) 13 (2.5) 0.62 (0.15) 0.037 0.063 140/5.4 2.7 82 30 (1.6)72 (7.8) 0.29 (0.12) (0.0045) (0.011) 170/6.3 1.3 84

glyburide 2.2 130 (17) 100 (12) 8.9 (0.69) 0.024 0.084 86/2.1 18 56 15 (1.0)120 (17) 5.6 (0.43) (0.0011) (0.012) 87/2.1 11 55

aFor parameters allowed to be individually varied (CLint,up and CLint,met) obtained values are given for both the UMJ (top) and the IRK (bottom)batches of cryopreserved hepatocytes. bExperimental values for logD7.4 were taken from the AstraZeneca internal database. cRate of OATP1B1-mediated uptake determined 2 min after incubation start. dCLint,diff set to 0 during fitting procedure.

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514446

coated 24-well plates (Becton Dickinson labware, U.K.) 72 hafter seeding. In brief, after the culture medium was removed,cells were washed twice followed by 10 min preincubation inHBSS. Experiments were started by addition of substratesolutions diluted from 10 mM DMSO stocks in HBSS to give afinal concentration of 1 μM. At selected time points (0.5, 1, 2,5, and 10 min) the solution was removed completely and thecells were lysed using a 50% acetonitrile solution. After mixingand centrifugation sample supernatants were analyzed by theLC/MS/MS system described above and quantified from astandard curve. Each sample was run in triplicate. Concen-trations were normalized against total protein determined usingPierce BCA Protein Assay Kit (Thermo Fisher Scientific Inc.,USA). The OATP1B1-mediated uptake rate was at each timepoint determined from (amount (OATP1B1) − amount(Mock))/time.Analysis of Known Molecular Properties, Model

Parameters and OATP1B1-Mediated Uptake of Inves-tigated Drugs by Principal Component Analysis (PCA). Aprincipal component analysis of the resulting parameters wasperformed using SIMCA-P+ 12.0.1 (Umetrics AB, Umea,Sweden). All CL and OATP1B1 transporter values werelogarithmized. Values for logD7.4 and conventional CLint inhuman liver microsomes (HLM) were taken from theAstraZeneca internal database. Regression analysis wasperformed to assess the linear relation between logD7.4 andparameters of interest.

■ RESULTSAssessment of Hepatocyte Drug Concentration

Profiles Using an “Oil-Spin” Procedure. Eight transportersubstrates were incubated with two separate batches ofcryopreserved human hepatocytes UMJ and IRK at 1 μM,and the intracellular and medium concentration−time profilesfrom two incubations were collected. Cellular fraction dataobtained for three representative drugs, rosuvastatin, pitavasta-tin and glyburide, are shown in the top panel of Figure 3 (datafor all eight compounds are given as Supporting Information,Figure S1).Mechanistic Mathematical Model To Elucidate the

Underlying Processes of Drug Distribution, Binding andMetabolism. The concentration data from the UMJ and IRKbatches overlapped in general well which allowed both data setsto be fitted simultaneously. The procedure restricted estimatesof bidirectional diffusion (CLint,diff), membrane binding (Kmem)and intracellular binding (fucell) to be shared between the twobatches and gave individual (batch-specific) estimates of themodel descriptors of active uptake (CLint,up) and metabolism(CLint,met). Model best-fit estimates are summarized in Table 1,with model simulations for the three representative drugsincluded in Figure 3 (simulations for all eight compounds aregiven as Supporting Information Figure S1, with residuals inSupporting Information Figure S2). The ranges of CLint,up,CLint,diff, and CLint,met for the investigated drugs span more than2 orders of magnitude each as shown in Figure 4. Medianvalues for CLint,up, CLint,diff and CLint,met are 18, 5.3, and 0.5 μL/min/106 cells, respectively, indicating the trend that CLint,upvalues were in general higher than both CLint,met and CLint,diff forthese compounds. Fexofenadine, a rather polar compound, wasconsidered metabolically stable (CLint,met fixed to 0 duringfitting procedure), since no compound depletion wasmeasurable during the time of the incubation. Figure 4 alsoshows the batch-specific differences for the estimates of

individually set parameters describing transporter and meta-bolic activity: CLint,up estimates for the test compounds arewithin a factor of 1.3, whereas CLint,met estimates are slightlymore variable (within a factor of 2.1). Accordingly, the uptakeand metabolic descriptors agree in general well between thetwo batches, and a batch-related difference was not indicated bya paired Student’s t test at significance level 0.05. The fractionof the total uptake that is attributed to CLint,up is >90% forvalsartan, rosuvastatin and pitavastatin (Table 1), illustratingthe importance of this route for the internalization. For themore lipophilic glyburide, associated with higher passivepermeability, just about 50% of the total uptake is due to anactive process, even though the active uptake CLint,up ofglyburide is one of the highest in the data set.

Assessment of Cell-to-Medium Concentration Ratiosand the Apparent Metabolic Clearance from theIncubation. Log-transformed medium and intracellularconcentration simulated for rosuvastatin, pitavastatin andglyburide over time at 1 million cells/mL are displayed inFigure 3 (bottom). The cell-to-medium concentration ratios atsteady state, described by the distribution parameters Kp andKp,uu in Table 1, can be calculated from eq 6. Kp given by themean total (bound and unbound) concentration ratio for theUMJ and the IRK batch, varies between approximately 4(pravastatin) and 170 (pitavastatin) (Table 1). The meanunbound concentration ratio Kp,uu, which in contrast to Kp isindependent of the binding properties, varies instead betweenapproximately 2 (glyburide) and 17 (rosuvastatin).In conventional assessment of metabolic clearance the cells

and medium are homogeneously sampled over time at a typicalhepatocyte concentration of 1 million cells/mL. An estimate ofthe apparent clearance of drug obtained from such a standardassay (CLinc) can be obtained by analyzing the simulateddepletion of drug from the whole incubation (cells, mediumand membrane compartments) at 1 million cells/mL. Since alldrugs in the set exhibit Kp,uu ratios above 1, estimated CLincvalues are generally larger than CLint,met values: simulated CLincdisplayed a median value of 2.6 μL/min/106 cells, compared tothe CLint,met median value of 0.5 μL/min/106 cells (Figure 4

Figure 4. Distribution of fitted model clearance parameters CLint,up,CLint,diff and CLint,met as well as the estimated apparent clearanceparameter CLinc. Dotted lines indicate the median value for eachparameter set. Coding (increasing logD7.4): 1 = valsartan, 2 =pravastatin, 3 = rosuvastatin, 4 = fexofenadine, 5 = pitavastatin, 6 =bosentan, 7 = atorvastatin, 8 = glyburide for the UMJ (open symbols)and the IRK batch (filled symbols). For CLint,diff, the two batches sharea common estimate (represented by open symbols).

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514447

and 5). The CLinc values span roughly 1 order of magnitude,which is considerably smaller than the range of the CLint,met

values.

Assessment of Drug Uptake into OATP1B1-Express-ing HEK-293 Cells. The specific uptake transport viaOATP1B1 was evaluated from time-course data collectedusing HEK 293-cells, stably transfected with OATP1B1 orempty vector (Mock). Rates were measured 2 min afterincubation start within the approximately linear phase of uptake(1 μM of substrate). The rate of OATP1B1-specific uptake wascalculated as difference in uptake between OATP1B1 andempty vector transfected cells. In Figure 6, the rate ofOATP1B1-specific uptake is compared to the mean CLint,upestimate from incubations with the two hepatocyte batches.The active uptake observed in the two systems correlates forthe presented data set, with an R2 of 0.79.

Analysis of Known Molecular Properties, ModelParameters and OATP1B1-Mediated Uptake of Inves-tigated Drugs by Principal Component Analysis (PCA).The obtained model estimates, experimental data for logD7.4and human liver microsome (HLM) CLint from theAstraZeneca database were subjected to a principal componentanalysis (PCA). Figure 7 shows the resulting score (A) and

loadings plots (B) for the first two components, which explainabout 80% of the variability of the data. There is a high degreeof colinearity of the data, as more than 60% of the variability isexplained by the first component alone. The first componentshows mainly the influence of lipophilicity, where the morelipophilic compounds like glyburide and atorvastatin are foundon the right-hand side of the score plot, whereas hydrophiliccompounds like fexofenadine and valsartan are situated more tothe left. In the loadings plot it can be seen that all CL values aswell as the binding parameter Kmem are positively correlated tologD7.4. fucell, on the other hand, is found on the opposite sideand thus negatively influenced by lipophilicity. Pairwise linearregression strengthens these findings; CLint,diff, CLint,met andKmem values show a positive correlation with logD7.4,demonstrating R2 values between 0.67 and 0.9, whereas fucellshows a negative correlation with R2 = 0.79 (plots andequations shown in Supporting Information, Figure S3). ThePCA detects furthermore colinearity of transporter relatedparameters in the second component: both uptake parameters(CLint,up and OATP1B1 uptake) and Kp,uu are found in theupper part of the loadings plot.

Figure 5. Comparison of estimated intrinsic metabolic clearance fromthe mechanistic model to the estimated apparent clearance from theincubation (given values represent a mean of the estimates obtainedfor the UMJ and IRK data). CLinc = CLint,met, expected for highlypermeable drugs, indicated by dashed line. Coding (increasinglogD7.4): 1 = valsartan, 2 = pravastatin, 3 = rosuvastatin, 5 =pitavastatin, 6 = bosentan, 7 = atorvastatin, 8 = glyburide.Metabolically stable fexofenadine not included.

Figure 6. Comparison of active hepatocellular to OATP1B1-specificuptake. Initial rate of OATP1B1-mediated uptake calculated from thedifference in amount internalized with OATP1B1 and empty vectortransfected HEK-293 cells after 2 min of incubation. Data pointsrepresent a mean obtained from three 1 μM incubations. CLint,up is themean of the estimates obtained for the UMJ and IRK batch. Dashedline taken from linear regression (log(HEK-OATP1B1 rate) = 0.81 ×log(CLint,up) − 0.067, R2 = 0.79). Coding (increasing logD7.4): 1 =valsartan, 2 = pravastatin, 3 = rosuvastatin, 4 = fexofenadine, 5 =pitavastatin, 6 = bosentan, 7 = atorvastatin, 8 = glyburide.

Figure 7. Score (A) and loadings (B) plots describing the first twocomponents obtained from PCA. Score plot coding (increasinglogD7.4): 1 = valsartan, 2 = pravastatin, 3 = rosuvastatin, 4 =fexofenadine, 5 = pitavastatin, 6 = bosentan, 7 = atorvastatin, 8 =glyburide.

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514448

■ DISCUSSION

The search for new chemical entities with high metabolicstability has led to compounds with a higher dependency onactive transport mechanisms for both uptake and elimination.For such compounds the conventional hepatocyte clearanceassays, where homogeneous samples of cells and medium arecollected, may be less appropriate for determining the intrinsicmetabolic clearance since the unbound intracellular concen-tration then can deviate strongly from that of the medium(Figure 1). While an in vitro assay that utilizes the loss of parentdrug from the incubation medium into hepatocytes (“medium-loss” assay) can allow for more effective prediction of the in vivosituation,37 a further definition of the underlying mechanisms islikely to rely on the use of physiologically inspiredmathematical models. Recent reports have presented hepato-cellular models for description of kinetics in different in vitrosettings.26,27,29−32 In the current study we aimed at elucidatingprocesses of drug distribution, binding and metabolism fromincubations replicating the conditions at which hepatocytemetabolic clearance routinely is assessed in drug discovery:incubations at one single drug concentration using pooledbatches of human cryopreserved hepatocytes, with samplestypically being collected over a longer period of time duringone to three hours.The obtained kinetic data sets were analyzed using a

mechanistic mathematical model including five adjustableparameters: active uptake clearance (CLint,up), bidirectionalpassive diffusion (CLint,diff), intracellular metabolism (CLint,met),intracellular unbound fraction of the drug (fucell) and theunspecific binding term Kmem describing the rapid equilibriumbinding of the drug to the outer cell membrane (Figure 2). Theobserved and simulated intracellular drug concentrations matchin general well over the time of the incubations (Figure 3 andSupporting Information Figures S1 and S2). The model fitparameters (Table 1) span a wide range of values (Figure 4),illustrating the applicability of the approach for drugs withdiverse properties.In the current parametrization, efflux transporter function-

ality, or any bidirectionality of uptake transporters, isconsidered limited (CLout = CLint,diff). Even if this assumptionfinds support in the literature,24 there are reports of effluxactivity also in short-term hepatocyte cultures.22,23 While thecurrent experimental approach is not appropriate for identifyingactive efflux back into the medium (CLint,eff), it is of interest tocomment on the implications of falsely ignoring such activity inthe suspended hepatocyte model. First, since a generalparametrization would attribute CLout to CLint,eff + CLint,diff,active efflux is inevitably smaller than or equal to CLint,diff inTable 1. Knowing that the studied drugs at least to some extentdo permeate the cellular membranes passively, CLint,eff ishowever likely to be considerably smaller. Furthermore, sinceCLin = CLint,diff + CLint,up, CLint,eff > 0 would result in a higherCLint,up. However, considering the trend of CLint,up being largerthan CLint,diff (Figure 4), the effect on CLint,up would in generalbe relatively small. Therefore, the CLint,diff and CLint,up estimatespresented in Table 1 can be considered as an upper limit to thepassive permeation and a lower limit to the active uptake,respectively. Estimates of binding (Kmem and fucell) anddistribution (Kp,uu) are independent of the specific interpreta-tion of CLin and CLout.Critical to this strategy outlined is the control measurement

at which the activity of active transport and metabolism is low.

Use of 4 °C incubations for this purpose is debated, since thedecrease of membrane fluidity at low temperature also mayaffect the passive permeability.27 However, by limiting the useof control data to steady-state conditions, collected between 60and 120 min after incubation start, the kinetics of the 4 °Cexperiment has no influence on CLint,diff in our model. Thesteady-state conditions, equivalent to unbound intra- andextracellular drug concentrations being equal, allows instead forestimation of fucell when a limited difference in intracellularbinding between 4 and 37 °C is assumed.The in vitro unbound cell-to-medium concentration ratio

(Kp,uu) provides a key measure for evaluating organ exposure ofdrugs distributing into the liver cells. Rosuvastatin andglyburide show clear differences, as illustrated by simulationsof their unbound and total cell and medium concentrationsover time (Figure 3, bottom). The relatively lipophilic glyburideis effectively internalized by uptake transporters. However, onlya moderate Kp,uu ratio is attained (approximately 2), primarilyas a consequence of the high passive permeability effectivelyequilibrating large unbound cell-to-medium concentrationratios. The mean (UMJ and IRK batch) active uptake clearanceof the less permeable rosuvastatin is about 8% of that estimatedfor glyburide, but results, attributed to the larger contributionto the total uptake, in a mean Kp,uu ratio of 17.Figure 5 illustrates the impact of differences in cellular

unbound exposure (Kp,uu) on the rate of metabolic clearanceobtained directly from the incubation. The intrinsic metabolicclearance (CLint,met) for bosentan and rosuvastatin differs by afactor of 10, showing the greater susceptibility for bosentan toundergo metabolic transformation. However, due to the higherKp,uu ratio for rosuvastatin, the apparent clearance (CLinc) ofbosentan is actually only a factor of 1.4 higher than that ofrosuvastatin. In a conventional clearance assay the twocompounds would thus appear as displaying comparablemetabolic stability. Likewise, the higher apparent clearancepredicted for pitavastatin compared to bosentan is not primarilyan effect of a lower stability, but rather is caused by the effectiveraise of the pitavastatin intracellular concentration due totransporter-mediated uptake. The considerably smaller varia-tion of the CLinc values, compared to the separate intrinsicclearances, within the data set (Figure 4) illustrates in analogywith Figure 1 the limited mechanistic information suchmeasures contain.The interbatch difference of active uptake clearance was in

general small compared to the difference observed in betweentested compounds, indicative of a similar overall transporteractivity profile. Menochet et al. reported on average 2.8-folddifference in uptake clearance between two individual donors.32

For comparison, the UMJ batch gave on average 11% higheruptake clearance than the IRK batch. A review of compoundsubstrate profiles did not indicate that the differences stillidentified with the pooled batches are related to the directfunction of specific uptake transporters. Comparison to uptakedata from monolayer cultures showed good agreement with ourdata despite the differences in experimental setting and theanalytical approach: the mean (UMJ and IRK batch) uptakeclearance estimates are 1.6 (valsartan), 1.3 (pravastatin), 1.1(rosuvastatin), 3.2 (pitavastatin), 1.6 (bosentan) times higherthan those reported for the donor exhibiting the highesttransporter activity by Menochet and co-workers.32 Higherrates with isolated hepatocytes may to some extent be directlyattributed to the culture format, such as the larger mediumexposed membrane area. However, the comparison shows that

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514449

preparation of the pooled batches, involving one additionalfreezing cycle, does not necessarily compromise the drugtransporter function extensively. The finding agrees with recentstudies specifically addressing the utility of cryopreservedhuman hepatocyte suspensions for uptake studies.38,39

Various transporter proteins may be involved in thehepatocellular uptake of the drugs investigated. However,OATP1B1 is generally considered a main contributor, withparticular impact to the ADME properties of, e.g., statins.OATP1B1 specific kinetics were obtained from time-coursedata with OATP1B1-transfected HEK-293 cells, using vector-transfected cells as baseline. OATP1B1 uptake was wellcorrelated to CLint,up as obtained from the hepatocytes,especially for the statin subset (Figure 6). Two of thecompounds, fexofenadine and glyburide, showed somewhathigher uptake in hepatocytes than in the OATP1B1 experi-ment. This could fit with their uptake being largely influencedby OATP1B340 and OATP2B141 and not solely dependent onOATP1B1. However, also atorvastatin is likely influenced byother transporters42,43 and does not diverge from the generalcorrelation.PCA was used to further investigate intercorrelations within

the data set including OATP1B1-uptake data and internalAstraZeneca data on logD7.4 and metabolic stability in humanliver microsomes (HLM). Lipophilicity is known to be animportant molecular property that influences passive perme-ation, membrane association and substrate binding to, e.g.,metabolizing enzymes and thereby metabolic stability.44−46

Parameter estimates are indeed found to be highly correlated tologD7.4, the main denominator of the first component of theloadings plot (Figure 7, bottom). The CL parameters areprimarily distinguished in the second component. It can benoted that the closeness of CLint,met and HLM CLint in theloadings plot indicates that it was possible to extract the puremetabolism descriptor in the present analysis of the hepatocytedata.The detoxification machinery of the hepatocyte relies on the

interplay of uptake, biotransformation and efflux transport.While the apparent clearance obtained from the conventionalhepatocyte incubation directly reflects metabolic stability whendistribution mainly is governed by passive permeation,interpretation becomes less straightforward when activetransport contributes to the uptake. We have found theproposed three-compartmental model applicable for resolvingsingle concentration, time-course data obtained from incuba-tions with eight drugs of diverse pharmacokinetic profiles. Wehave furthermore found pooled batches of cryopreservedhuman hepatocytes, allowing for an extended, reproducibleuse, a relevant alternative to preparations from individualdonors for screening purposes. This rationalized procedure isforeseen to have good potential for application in the drugdiscovery setting allowing larger numbers of compounds to beexplored with a reduced experimental effort compared to fullkinetic characterization with multiple test concentrations.

■ ASSOCIATED CONTENT

*S Supporting InformationCellular fraction experimental and simulated profiles fortransporter for all eight substrates (with corresponding residualplots) and linear relation of model parameters to logD7.4. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*E-mail: constanze.hilgendorf@astrazeneca.com. Tel: +46 (0)317065349. Fax: +46 (0)317763743.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

The authors gratefully acknowledge Drs. Ken Grime and PatrikLundquist for careful review and suggestions on the manu-script.

■ ABBREVIATIONS USED

CLinc, apparent metabolic clearance with respect to the totalincubation concentration; CLmed, metabolic clearance withrespect to the medium concentration; CLin, sum of clearancesassociated with transport from the medium into the cells;CLint,diff, intrinsic diffusion clearance; CLint,eff, intrinsic activeefflux clearance; CLint,met, intrinsic metabolic clearance; CLint,up,intrinsic active uptake clearance; CLout, sum of clearancesassociated with transport from the cells to the medium; fucell,intracellular fraction unbound; HLM, human liver microsomes;Kp/Kp,uu, total/unbound cell-to-media concentration ratio atsteady-state; Kmem, outer membrane binding constant; OATP,organic anion transporter polypeptide

■ REFERENCES(1) Hewitt, N. J.; Lechon, M. J.; Houston, J. B.; Hallifax, D.; Brown,H. S.; Maurel, P.; Kenna, J. G.; Gustavsson, L.; Lohmann, C.;Skonberg, C.; Guillouzo, A.; Tuschl, G.; Li, A. P.; LeCluyse, E.;Groothuis, G. M.; Hengstler, J. G. Primary hepatocytes: currentunderstanding of the regulation of metabolic enzymes and transporterproteins, and pharmaceutical practice for the use of hepatocytes inmetabolism, enzyme induction, transporter, clearance, and hepatotox-icity studies. Drug Metab. Rev. 2007, 39, 159−234.(2) Houston, J. B. Utility of in vitro drug metabolism data inpredicting in vivo metabolic clearance. Biochem. Pharmacol. 1994, 47,1469−1479.(3) Ito, K.; Houston, J. B. Prediction of human drug clearance fromin vitro and preclinical data using physiologically based and empiricalapproaches. Pharm. Res. 2005, 22, 103−112.(4) Riley, R. J.; McGinnity, D. F.; Austin, R. P. A unified model forpredicting human hepatic, metabolic clearance from in vitro intrinsicclearance data in hepatocytes and microsomes. Drug Metab. Dispos.2005, 33, 1304−1311.(5) Grime, K.; Riley, R. J. The impact of in vitro binding on in vitro -In vivo extrapolations, projections of metabolic clearance and clinicaldrug-drug interactions. Curr. Drug Metab. 2006, 7, 251−264.(6) Sohlenius-Sternbeck, A.-.; Jones, C.; Ferguson, D.; Middleton, B.J.; Projean, D.; Floby, E.; Bylund, J.; Afzelius, L. Practical use of theregression offset approach for the prediction of in vivo intrinsicclearance from hepatocytes. Xenobiotica 2012, 42, 841−853.(7) Soars, M. G.; Webborn, P. J.; Riley, R. J. Impact of hepatic uptaketransporters on pharmacokinetics and drug-drug interactions: use ofassays and models for decision making in the pharmaceutical industry.Mol. Pharmaceutics 2009, 6, 1662−1677.(8) Chandra, P.; Brouwer, K. L. R. The complexities of hepatic drugtransport: Current knowledge and emerging concepts. Pharm. Res.2004, 21, 719−735.(9) International Transporter Consortium; Giacomini, K. M.; Huang,S. M.; Tweedie, D. J.; Benet, L. Z.; Brouwer, K. L.; Chu, X.; Dahlin, A.;Evers, R.; Fischer, V.; Hillgren, K. M.; Hoffmaster, K. A.; Ishikawa, T.;Keppler, D.; Kim, R. B.; Lee, C. A.; Niemi, M.; Polli, J. W.; Sugiyama,Y.; Swaan, P. W.; Ware, J. A.; Wright, S. H.; Yee, S. W.; Zamek-

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514450

Gliszczynski, M. J.; Zhang, L. Membrane transporters in drugdevelopment. Nat. Rev. Drug Discovery 2010, 9, 215−236.(10) Klaassen, C. D.; Aleksunes, L. M. Xenobiotic, bile acid, andcholesterol transporters: Function and regulation. Pharmacol. Rev.2010, 62, 1−96.(11) Hilgendorf, C.; Ahlin, G.; Seithel, A.; Artursson, P.; Ungell, A.-L.; Karlsson, J. Expression of thirty-six drug transporter genes inhuman intestine, liver, kidney, and organotypic cell lines. Drug Metab.Dispos. 2007, 35, 1333−1340.(12) Ohtsuki, S.; Schaefer, O.; Kawakami, H.; Inoue, T.; Liehner, S.;Saito, A.; Ishiguro, N.; Kishimoto, W.; Ludwig-Schwellinger, E.; Ebner,T.; Terasaki, T. Simultaneous absolute protein quantification oftransporters, cytochromes P450, and UDP-glucuronosyltransferases asa novel approach for the characterization of individual human liver:Comparison with mRNA levels and activities. Drug Metab. Dispos.2012, 40, 83−92.(13) Elsby, R.; Hilgendorf, C.; Fenner, K. Understanding the criticaldisposition pathways of statins to assess drugdrug interaction riskduring drug development: It’s not just about OATP1B1. Clin.Pharmacol. Ther. 2012, 92, 584−598.(14) Kalliokoski, A.; Niemi, M. Impact of OATP transporters onpharmacokinetics. Br. J. Pharmacol. 2009, 158, 693−705.(15) Garcia, M. J.; Reinoso, R. F.; Sanchez Navarro, A.; Prous, J. R.Clinical pharmacokinetics of statins. Methods Find. Exp. Clin.Pharmacol. 2003, 25, 457−481.(16) Hamelin, B. A.; Turgeon, J. Hydrophilicity/Lipophilicity:Relevance for the pharmacology and clinical effects of HMG-CoAreductase inhibitors. Trends Pharmacol. Sci. 1998, 19, 26−37.(17) Shitara, Y.; Horie, T.; Sugiyama, Y. Transporters as adeterminant of drug clearance and tissue distribution. Eur. J. Pharm.Sci. 2006, 27, 425−446.(18) Funk, C. The role of hepatic transporters in drug elimination.Expert Opin. Drug Metab. Toxicol. 2008, 4, 363−379.(19) Iwatsubo, T.; Suzuki, H.; Sugiyama, Y. Determination of therate-limiting step in the hepatic elimination of YM796 by isolated rathepatocytes. Pharm. Res. 1999, 16, 110−116.(20) Shitara, Y.; Sato, H.; Sugiyama, Y. Evaluation of drug-druginteraction in the hepatobiliary and renal transport of drugs. Annu. Rev.Pharmacol. Toxicol. 2005, 45, 689−723.(21) Webborn, P. J. H.; Parker, A. J.; Denton, R. L.; Riley, R. J. Invitro-in vivo extrapolation of hepatic clearance involving active uptake:Theoretical and experimental aspects. Xenobiotica 2007, 37, 1090−1109.(22) Lam, J. L.; Benet, L. Z. Hepatic microsome studies areinsufficient to characterize in vivo hepatic metabolic clearance andmetabolic drug-drug interactions: Studies of digoxin metabolism inprimary rat hepatocytes versus microsomes. Drug Metab. Dispos. 2004,32, 1311−1316.(23) Li, M.; Yuan, H.; Li, N.; Song, G.; Zheng, Y.; Baratta, M.; Hua,F.; Thurston, A.; Wang, J.; Lai, Y. Identification of interspeciesdifference in efflux transporters of hepatocytes from dog, rat, monkeyand human. Eur. J. Pharm. Sci. 2008, 35, 114−126.(24) Bow, D. A. J.; Perry, J. L.; Miller, D. S.; Pritchard, J. B.; Brouwer,K. L. R. Localization of P-gp (Abcb1) and Mrp2 (Abcc2) in freshlyisolated rat hepatocytes. Drug Metab. Dispos. 2008, 36, 198−202.(25) Gardiner, P.; Paine, S. W. The impact of hepatic uptake on thepharmacokinetics of organic anions. Drug Metab. Dispos. 2011, 39,1930−1938.(26) Watanabe, T.; Kusuhara, H.; Maeda, K.; Shitara, Y.; Sugiyama,Y. Physiologically based pharmacokinetic modeling to predicttransporter-mediated clearance and distribution of pravastatin inhumans. J. Pharmacol. Exp. Ther. 2009, 328, 652−662.(27) Poirier, A.; Lave, T.; Portmann, R.; Brun, M. E.; Senner, F.;Kansy, M.; Grimm, H. P.; Funk, C. Design, data analysis, andsimulation of in vitro drug transport kinetic experiments using amechanistic in vitro model. Drug Metab. Dispos. 2008, 36, 2434−2444.(28) Nordell, P.; Svanberg, P.; Bird, J.; Grime, K. PredictingMetabolic Clearance for Drugs That Are Actively Transported intoHepatocytes: Incubational Binding as a Consequence of in Vitro

Hepatocyte Concentration Is a Key Factor. Drug Metab. Dispos. 2013,41, 836−843.(29) Paine, S. W.; Parker, A. J.; Gardiner, P.; Webborn, P. J.; Riley, R.J. Prediction of the pharmacokinetics of atorvastatin, cerivastatin, andindomethacin using kinetic models applied to isolated rat hepatocytes.Drug Metab. Dispos. 2008, 36, 1365−1374.(30) Yabe, Y.; Galetin, A.; Houston, J. B. Kinetic characterization ofrat hepatic uptake of 16 actively transported drugs. Drug Metab. Dispos.2011, 39, 1808−1814.(31) Menochet, K.; Kenworthy, K. E.; Houston, J. B.; Galetin, A.Simultaneous assessment of uptake and metabolism in rat hepatocytes:A comprehensive mechanistic model. J. Pharmacol. Exp. Ther. 2012,341, 2−15.(32) Menochet, K.; Kenworthy, K. E.; Houston, J. B.; Galetin, A. Useof mechanistic modeling to assess interindividual variability andinterspecies differences in active uptake in human and rat hepatocytes.Drug Metab. Dispos. 2012, 40, 1744−1756.(33) Chis, O.; Banga, J. R.; Balsa-Canto, E. GenSSI: A softwaretoolbox for structural identifiability analysis of biological models.Bioinformatics 2011, 27, 2610−2611.(34) Reinoso, R. F.; Telfer, B. A.; Brennan, B. S.; Rowland, M.Uptake of teicoplanin by isolated rat hepatocytes: Comparison with invivo hepatic distribution. Drug Metab. Dispos. 2001, 29, 453−459.(35) Bonate, P. L. Pharmacokinetic-Pharmacodynamic Modeling andSimulation; Springer-Verlag GmbH: Heidelberg, Germany, 2005.(36) Sharma, P.; Holmes, V. E.; Elsby, R.; Lambert, C.; Surry, D.Validation of cell-based OATP1B1 assays to assess drug transport andthe potential for drugdrug interaction to support regulatorysubmissions. Xenobiotica 2010, 40, 24−37.(37) Soars, M. G.; Grime, K.; Sproston, J. L.; Webborn, P. J. H.;Riley, R. J. Use of hepatocytes to assess the contribution of hepaticuptake to clearance in vivo. Drug Metab. Dispos. 2007, 35, 859−865.(38) De Bruyn, T.; Ye, Z.-W.; Peeters, A.; Sahi, J.; Baes, M.;Augustijns, P. F.; Annaert, P. P. Determination of OATP-, NTCP- andOCT-mediated substrate uptake activities in individual and pooledbatches of cryopreserved human hepatocytes. Eur. J. Pharm. Sci. 2011,43, 297−307.(39) Badolo, L.; Trancart, M. M.; Gustavsson, L.; Chesne, C. Effectof cryopreservation on the activity of OATP1B1/3 and OCT1 inisolated human hepatocytes. Chem.-Biol. Interact. 2011, 190, 165−170.(40) Shimizu, M.; Fuse, K.; Okudaira, K.; Nishigaki, R.; Maeda, K.;Kusuhara, H.; Sugiyama, Y. Contribution of OATP (organic anion-transporting polypeptide) family transporters to the hepatic uptake offexofenadine in humans. Drug Metab. Dispos. 2005, 33, 1477−1481.(41) Satoh, H.; Yamashita, F.; Tsujimoto, M.; Murakami, H.; Koyabu,N.; Ohtani, H.; Sawada, Y. Citrus juices inhibit the function of humanorganic anion-transporting polypeptide OATP-B. Drug Metab. Dispos.2005, 33, 518−523.(42) Karlgren, M.; Vildhede, A.; Norinder, U.; Wisniewski, J. R.;Kimoto, E.; Lai, Y.; Haglund, U.; Artursson, P. Classification ofinhibitors of hepatic organic anion transporting polypeptides(OATPs): Influence of protein expression on drug-drug interactions.J. Med. Chem. 2012, 55, 4740−4763.(43) Shitara, Y.; Maeda, K.; Ikejiri, K.; Yoshida, K.; Horie, T.;Sugiyama, Y. Clinical significance of organic anion transportingpolypeptides (OATPs) in drug disposition: Their roles in hepaticclearance and intestinal absorption. Biopharm. Drug Dispos. 2013, 34,45−78.(44) Arnott, J. A.; Planey, S. L. The influence of lipophilicity in drugdiscovery and design. Expert Opin. Drug Discovery 2012, 7, 863−875.(45) Winiwarter, S.; Ridderstrom, M.; Ungell, A.-L.; Andersson, T.B.; Zamora, I. Use of Molecular Descriptors for ADME Predictions.Comprehensive Medicinal Chemistry II. In ADME-Tox Approaches;Triggle, D. J., Taylor, J. B., Eds.; Elsevier: 2007; Vol. 5, p 531.(46) Testa, B.; Crivori, P.; Reist, M.; Carrupt, P.-A. The influence oflipophilicity on the pharmacokinetic behavior of drugs: Concepts andexamples. Perspect. Drug Discovery Des. 2000, 19, 179−211.

Molecular Pharmaceutics Article

dx.doi.org/10.1021/mp400253f | Mol. Pharmaceutics 2013, 10, 4443−44514451