1521-009X/45/2/152–159$25.00 http://dx.doi.org/10.1124/dmd.116.072306DRUG METABOLISM AND DISPOSITION Drug Metab Dispos 45:152–159, February 2017Copyright ª 2017 by The American Society for Pharmacology and Experimental Therapeutics

Target-Site Investigation for the Plasma Prolactin Response:Mechanism-Based Pharmacokinetic-Pharmacodynamic Analysis of

Risperidone and Paliperidone in the Rat s

Shinji Shimizu,1 Sandra M. den Hoedt, Victor Mangas-Sanjuan, Sinziana Cristea, Jana K. Geuer,Dirk-Jan van den Berg, Robin Hartman, Francisco Bellanti, and Elizabeth C. M. de Lange

Division of Pharmacology, Leiden Academic Center for Drug Research, Leiden, The Netherlands

Received June 30, 2016; accepted November 7, 2016

ABSTRACT

To understand the drivers in the biological system response todopamine D2 receptor antagonists, a mechanistic semiphysiologi-cally based (PB) pharmacokinetic-pharmacodymanic (PKPD) modelwas developed to describe prolactin responses to risperidone (RIS)and its active metabolite paliperidone (PAL). We performed amicrodialysis study in rats to obtain detailed plasma, brain extra-cellular fluid (ECF), and cerebrospinal fluid (CSF) concentrations ofPAL and RIS. To assess the impact of P-glycoprotein (P-gp)functioning on brain distribution, we performed experiments in theabsence or presence of the P-gp inhibitor tariquidar (TQD). PK andPKPDmodelingwas performed by nonlinearmixed-effect modeling.

Plasma, brain ECF, and CSF PK values of RIS and PAL were welldescribed by a 12-compartmental semi-PBPK model, includingmetabolic conversion of RIS to PAL. P-gp efflux functionality wasidentified on brain ECF for RIS and PAL and on CSF only for PAL. Inthe PKPD analysis, the plasma drug concentrations were morerelevant than brain ECF or CSF concentrations to explain the prolactinresponse; the estimated EC50 was in accordance with reports in theliterature for both RIS and PAL. We conclude that for RIS and PAL, theplasma concentrations better explain the prolactin response than dobrain ECF or CSF concentrations. This research shows that PKPDmodeling is of high value to delineate the target site of drugs.

Introduction

It is known that prolactin is released from the pituitary upon inhibition ofthe dopamine-D2 (D2) receptor. As D2 receptors are present in brain caudateputamen, which feeds into the hypothalamus and further to the pituitary, butalso D2 on the pituitary itself, the question is whether D2 antagonists exerttheir effects on prolactin release directly at the level of pituitary or at the levelof the brain. If prolactin release is mostly under the control of brain D2receptors, it would represent a biomarker of brain dopaminergic activity,which is the target site for anti-schizophrenia drug action.In a previous publication, brain extracellular fluid (ECF) concentra-

tions of remoxipride (REM) in rats have been identified as the target-siteconcentrations by using a modified pool effect model (Stevens et al.,2012). Interestingly, this pharmacokinetic-pharmacodymanic (PKPD)model could be translated successful to the human situation to predictdata that had been obtained on REM PK and prolactine PD in plasma.

The question, however, is whether this effect model is more generallyapplicable and thus drug-independent.Risperidone (RIS) is awidely used antipsychotic drug. Paliperidone (PAL)

is a pharmacologically activemetabolite ofRIS and is nowalso prescribed forschizophrenia. These twoD2 antagonists have been investigated their PKPDproperties, including throughmathematical modeling. Kozielska et al. (2012)successfully developed a physiology- and mechanism-based model todescribe D2 receptor occupancy, as well as whole-brain concentrations ofRIS and PAL in rats. Friberg et al. (2009) examined mechanistic models(including pool model) to describe prolactin response in humans. They usedplasma concentrations of RIS and PAL as a driver to induce plasma prolactinelevation. The aim of the current study was to investigate whether the PDmodel developed for REM can also be applied to two other D2 antagonists,RIS and (its active metabolite) PAL, or that direct interaction of the drugs inplasmawithD2 receptors at the level of the pituitarywould be themechanismof action for inducing prolactin release from the pituitary into plasma.To describe the plasma PK and brain distribution, a microdialysis

study in rats was carried out to provide serial samples of plasma, brainECF, and cerebrospinal fluid (CSF). RIS or PAL was administered i.v.,with or without coadministration of tariquidar (TQD) as P-glycoprotein(P-gp) blocker, to assess potential changes in brain distribution by P-gpfunctionality. A physiologically based pharmacokinetic (PBPK)

This work was supported by Daichi Sankyo Co., Ltd., and TI Pharma PKPDModeling Platform 2.

1Current affiliation: Daiichi-Sankyo Co., Ltd., Shinagawa R&D Center, Tokyo, Japan.dx.doi.org/10.1124/dmd.116.072306.s This article has supplemental material available at dmd.aspetjournals.org.

ABBREVIATIONS: AAI, agonist-antagonist interaction; BASE, baseline prolactin concentrations; BBB, blood-brain barrier; bPF, bufferedmicrodialysis perfusion fluid; CM, cisterna magna; CP, caudate putamen; CSF, cerebrospinal fluid; D2, dopamine D2; DE, drug effect; ECF,extracellular fluid; GLU, glucose; ID, internal diameter; IS, internal standard; LV, lateral ventricle; MP, mobile phase; MQ, purified millipore water; MS,mass spectrometry; NMEM, nonlinear mixed-effect modeling; OFV, objective function value; PAL, paliperidone; PB, physiologically based; P-gp,P-glycoprotein; PKPD, pharmacokinetic-pharmacodynamic; REM, remoxipride; RIS, risperidone; TFV, third and fourth ventricle; Tmax, peak time;TQD, tariquidar; VPC, visually predictive check.

152

http://dmd.aspetjournals.org/content/suppl/2016/11/11/dmd.116.072306.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

approach was taken to describe PK profiles in plasma, brain ECF, andCSF. Nonlinear mixed-effect modeling (NMEM) was used for PKPDmodeling, in which themetabolite conversion fromRIS to PALwas alsotaken into account.

Materials and Methods

Chemicals and Reagents

For all procedures, purified Millipore water (MQ, 18.2 MV cm) from aMilli-QPF Plus system was used (Millipore B.V., Amsterdam, The Netherlands). RIS, PAL,RIS-D4, and PAL-D4 were obtained from Sigma-Aldrich (Zwijndrecht, The Nether-lands). TQD was obtained from Xenova group PLC (Cambridge, United Kingdom).Ammonium formate, ammonium bicarbonate, and ammonium hydroxide [all ultrahigh performance liquid chromatography (ULC)/mass spectrometry (MS) grade] wereobtained from Sigma-Aldrich. Acetonitrile (high-performance liquid chroma-tography S grade), methanol, isopropanol, and formic acid (all ULC/MS grade)were obtained from Biosolve B.V. (Valkenswaard, The Netherlands). All pHmeasurements were performed using a Schott CG 820 pHmeter (Schott-GeräteGMBH, Mainz, Germany).

Animals

All animal procedures were performed in accordance with Dutch laws onanimal experimentation. The study protocol was approved by the Animal EthicsCommittee of Leiden University (UDEC nr. 12049). Male Wistar WU rats (n =61, 269 6 16.6 g, 8–10 weeks old; Charles River Laboratories, Den Bosch, TheNetherlands) were housed in groups for a minimum period of 7 days understandard environmental conditions (ambient temperature 20 6 1�C, humidity40%–60%, 12-hour light/dark cycle, background noise, daily handling) with adlibitum access to food (Laboratory Chow; Hope Farms, Woerden, The Nether-lands) and acidified water. Between surgery and experiments, the animals werekept individually in Makrolon type 3 cages (Tecniplast, Buguggiate, Italy) withspecial grid for 7 days to recover from the surgical procedures.

Surgical Procedures

Anesthesia. At t = 27 days, the animals underwent surgery. Anesthesia wasinduced using amixture of oxygen and 5% isoflurane andmaintained by amixtureof oxygen and 2%–3% isoflurane. The head, neck, and left groin were shaved.Animals received 0.15 ml of antibiotics (ampicillin 20%, s.c.) and 0.09 ml of s.c.analgesic buprenorphine (Temgesic: AST Farma, Oudewater, The Netherlands)before operation. Body temperaturewasmaintained at 37�Cby an electric heating pad.

Blood Cannulas. Animals received a cannula in the left femoral artery and femoralvein for blood sampling and drug administration, respectively. All cannulas were firstdisinfected with 0.1% benzalkonium chloride. The cannula inserted into the femoralartery consisted of a 4-cm piece with an internal diameter (ID) of 0.28mm connected toa 16-cm piece with 0.58-mm ID cannula (Portex fine-bore polythene tubing; SmithsMedical, Kent, England). The cannula inserted in the femoral vein consisted of a 3.5-cmpiece ID 0.58-mm ID cannula connected to 16-cm ID 0.58-mm cannula. Both cannulaswere led s.c. to the back of the head and fixated into the neck with a rubber ring.

Microdialysis Guide and Dummy Implantation. Animals were chronicallyinstrumented with a CMA/12 microdialysis guide (Aurora Borealis Control,Schoonebeek, The Netherlands) in the caudate putamen (CP, for ECF sampling)and the cisterna magna (CM; for CSF sampling), as described earlier (Westerhoutet al., 2012). The CMA/12 guide in the CP was inserted at the coordinate’santerior-posterior 21.0 mm, lateral +3.2 mm, and ventral 23.5 mm from thebregma. The CMA/12 guide in the CM was inserted 11 degrees from the lateralaxis and 25 degrees from the dorsoventral axis (toward posterior) at coordinatesanterior-posterior 22.51 mm, lateral +2.04 mm, and ventral 28.34 mm from l.

After surgery, animals received 3 ml of saline (B. Braun Melsungen AG,Melsungen, Germany, s.c.) to prevent excessive dehydration and weight loss. Torecover from the surgical procedures, the animals were kept individually inMakrolon type 3 cages (Covestro) with special grid for 7 days.

Experimental Procedures

After the animals were placed in the experimental chamber, CMA/12microdialysis guides were replaced by activated probes; CMA Elite PAES,4-mm (CP) or 1-mm (CM) polycarbonate membrane, cutoff 20 kD (Aurora

Borealis Control Schoonebeek, The Netherlands). Before placement of theanimal, probes were flushed with buffered microdialysis perfusion fluid (bPF).Microdialysis bPF was prepared, consisting of phosphate buffer (2 mM, pH 7.4)containing 145 mM sodium, 2.7 mM potassium, 1.2 mM calcium, 1.0 mMmagnesium, 150mMchloride, and 0.2 mMascorbate, whichwas filtered and thenstored in glass vials (220�C). The bPF was sonified and degassed with heliumbefore use. After implantation of the activated probe, the animals were kept in theexperimental chamber until start of the experiments.

The in vivo recovery was determined by the reversed dialysis method “byloss.” The microdialysis probes in CP (for brain ECF) and CM (for CSF) wereperfused with a bPF containing PAL or RIS (both 50 ng/ml). The in vivo recoverywas defined (eq. 1) as the ratio of the concentration difference between thedialysate (Cdial) and perfusion fluid (Cin) and divided by the concentration in bPFfor each microdialysis probe location (Scheller and Kolb, 1991):

In    vivo    recovery ¼ Cin 2Cdial

Cinð1Þ

At t = 0 of the study, one experimental group received an i.v. dose of 2.0 mg/kgRIS in saline during a 20-minute infusion through an automated pump (HarvardApparatus 22, model 55-2222; Holliston, MA) with (n = 8) or without (n = 8) ani.v. injection of 15 mg/kg TQD in glucose (GLU, 5%) for 10 minutes starting att = 230 minutes. The second experimental group received an i.v. dose of0.5 mg/kg PAL in saline during a 10-minute infusion with (n = 8) or without(n = 13) an i.v. injection of 15 mg/kg TQD in GLU (5%) for 10 minutes starting att =230minutes. The experimental control group (animals without TQD) receivedan i.v. injection of 1.25 ml GLU (5%) instead of TQD per 250-kg rat. Start andduration of infusion were corrected for internal volume of the tubing so thatinfusion started at t = 0 minute.

Blood samples (150 ml) were obtained from the arterial cannula at t = 215,1 (PAL only)/20(RIS only), 30, 60, 90, 120, 180, 240, 300, and 360 minutes,shaken in heparin- or EDTA-coated Eppendorf cups (Sarstedt, Nümbrecht,Germany) and stored temporarily in the refrigerator. During and after theexperiments, the samples were centrifuged for 10 minutes to separate plasmafrom the whole blood. The obtained plasma samples were stored at 220�C untilanalysis.

Tubing (1.2 ml/100 mm FEP tubing; CMA/Microdialysis, Stockholm,Sweden) was connected with tubing adapters (CMA/microdialysis) to themicrodialysis probe’s inlet and outlet. Microdialysis vials were preweighed andplaced in a cooled fraction collector (Univentor 820 Microsampler; Antec, TheNetherlands) to collect microdialysate samples. Microdialysis probes werecontinuously flushed with bPF (1 ml/min, Bee-Hive; Bioanalytical SystemsInc., W. Lafayette, IN), and samples were collected at 20-minute intervalsbetween t = 22 and t = 6 hours. After sample collection, vials were weighed todetermine the true probe perfusion rate and stored at280�C until analysis. At theend of the experiments, the animals were sacrificed by receiving a lethal dosage ofNembutal (Ceva Sante Animale, Naaldwijk, The Netherlands), after whichanimals underwent transcardial perfusion with 50 mM phosphate buffer (pH 7.4).

Plasma Protein Binding

For the determination of plasma protein binding, the plasma samples werepooled by a group of 10 and 30 (PAL) or 20 and 30minutes (RIS), 60, 90, 120 and180 minutes, which resulted in aliquots of 150–500 ml of plasma. The obtainedpooled plasma (fromwhich an aliquot was taken for analysis of total PAL or RIS)was put into protein binding tubes, with a cutoff of 10 kDa (Centrifree device;Merck Millipore, Amsterdam, The Netherlands). The tubes were incubated at37�C for 30 minutes and then centrifuged for 20 minutes in a thermostattedcentrifuge (37�C, 2000g). The filtrate was analyzed with the same analysismethod as the pooled plasma. An average of protein bindings of three groups foreach compound was used to calculate unbound plasma concentrations.

Measurement of Risperidone and Paliperidone

Sample Preparation of Microdialysis Samples. The calibration standardsfor the microdialysate analysis were prepared in bPF with concentrations of 0,0.05, 0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 7.5, and 10.0 ng/ml for PAL and RIS. Analyte andcalibration samples were mixed with internal standard (IS) solution (PAL-D4 andRIS-D4 in MQ) (1:1, v/v).

PKPD Analysis of Risperidone and Paliperidone 153

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

Sample Preparation of Plasma Samples. The calibration standards for theplasma analysis were prepared inMQwith concentrations of 0, 5, 10, 20, 50, 100,200, 500, and 1000 ng/ml for PAL and RIS. Calibration samples were mixed withblank plasma and IS solution (1:1:1, v/v). For analyte samples, those samplesweremixed with MQ and IS solution (1:1:1, v/v). For analyte and calibration samples,400 ml of 0.1M ammonium bicarbonate solution (pH 10) was added to precipitatethe proteins. After brief vortexing (VortexMS2Minishaker IKA; Beun de Ronde,Breda, The Netherlands) of analyte and calibration samples, 5 ml of isopropy-lether was added to the mixtures. Resulting samples were centrifuged in glassvials at 4000 rpm for 10 minutes (Heraeus Multifuge 3s; Dijkstra VereenigdeBV, Lelystad, The Netherlands). Supernatant (4 ml) was pipetted in glassvials, and isopropylether was evaporated (Labconco vortex evaporator; Beunde Ronde). Residues were resolved in 200 ml ammonium formate (5 mM) inMQ (pH 4.1), sonicated, and vortexed. An aliquot (150 ml) was pipetted inan Eppendorf vial and centrifuged at 14,000 rpm for 10 minutes (Labofuge GLcentrifuge, Micro CL 21R; Thermo Fischer Scientific, Breda, The Netherlands).

Liquid Chromatography. An aliquot of 15 ml per sample was injected intothe liquid chromatography-tandem mass spectroscopy (LC-MS/MS) system.After each injection (10�C, Surveyor auto sampler, Thermo Fischer Scientific),the injection needle was washed with 100ml methanol:acetonitrile:isopropanol inMQ (1:1:1, v/v) and with 100 ml of methanol in MQ (1:1, v/v), to reducecarryover. The MS pump (Surveyor MS pump, Thermo Fischer Scientific) wasused to separate the microdialysis analytes on a Hyperclone 34 BOS C18 130AHPLC column (100mm� 2.00mm, 3mm; Phenomenex, Utrecht, TheNetherlands).Mobile phaseA (MPA) consisted of 10%methanol and 10mMammonium formate atpH 4.1; mobile phase B (MPB) consisted of 90% methanol and 10 mM ammoniumformate at pH 4.1. The column was preconditioned (0–1 minute) with 26% methanolusing a 20/80 ratio of MPA and MPB. A linear gradient (1–6 minutes) from 26%methanol to 74%methanol was used to elute RIS, PAL, and IS. At 8 minutes, the LCsystem was set to 26% methanol to equilibrate the column.

Mass Spectrometry. The Finnigan TSQ Quantum Ultra Mass SpectrometerSystem (Thermo Fischer Scientific) was tuned by infusing standard solutions ofPAL and RIS or IS with the aid of a T-piece in the LC eluent. Electro spray-ionization mode was chosen tomonitor all four compounds (PAL, RIS, and their IS),and nitrogen was used as a desolvation gas. After collision-induced dissociation byargon gas (0.8 psi), the total ion current was measured using the fragmented ions ofPAL (MH+ = 207), RIS (MH+ = 191), PAL-D4 (MH+ = 211), and RIS-D4 (MH+ =195). Tuning of the MS was performed for all four compounds.

Linear regression was used to determine the slope, intercept, and correlationcoefficient of the relation between peak-area ratio and the drug concentration inthe calibration standards data after analysis. Relationships were accepted whencorrelation coefficient is greater than 0.98. For the calibration curves, weightingfactors of 1, 1/Y, and 1/Y^2 were compared.

Prolactin Analysis

A fourplex bead assay (rat pituitary magnetic bead panel, Milliplex Map,Millipore) was used to measure the plasma levels of prolactin on a Luminex100/200TM platform (Luminex BV, Oosterhout, The Netherlands). The standardprotocol was adjusted by diluting plasma samples in serum matrix (1:1, v/v), andbeads were diluted with assay buffer (1:1, v/v). Each sample was measured induplicate. For calculations, the average value of the duplicates was used as a singlevalue.

PK Data Analysis

All plasma concentrations were converted to unbound plasma concentrationsby correction for plasma protein binding. All microdialysate concentrations fromCP and CM were converted into brain ECF (CECF) or CSF concentrations (CCSF)by division of the dialysate concentrations by the average in vivo recovery asdetermined for each microdialysis probe location (eq. 2);

CECF   or CCSF ¼ Cdial

in  vivo recoveryð2Þ

Pharmacometric Analysis

Model Development. Pharmacokinetic analysis was performed using NMEMimplemented in NONMEM VII (GloboMax LLC, Hanover, MD) with the

subroutine ADVAN 6 and first-order conditional. Model selection was basedon parameter estimates and their relative standard errors, residual errorvalues, goodness-of-fit plots, visual inspection, visually predictive check(VPC, 1000 replicates), and the likelihood ratio test (P, 0.05). During thoseanalyses, software Pirana (Keizer et al., 2011) was used for NONMEM runmanagement and Xpose implemented in R 3.0.0 (The R Foundation forStatistical Computing, Vienna, Austria, 2011) for plots. For example,predicted concentrations were plotted against measured concentrations, andconditional weighted residuals were calculated and plotted against time andconcentration.

The interindividual variation of a parameter was described by the exponentialvariance model in eq. 3:

ui ¼ upop × expðhiÞ; ð3Þ

where ui is the parameter in the ith animal; upop, the parameter in a typicalanimal; and hi, the interanimal variability, which is assumed to be a standarddeviation v. The parameter value for the ith animal from the typical value waspredicted from the regression model. Interindividual variation was incorporatedonly for parameters for which it significantly (P, 0.05: objective function value,OFV; a reduction of 3.84 units) improved the model.

TQD treatment was defined as a covariate to examine its effect on theparameter estimates (eq. 4):

ui ¼ upop × expðhiÞ × ucovarCOVARIATEð1 or 0Þ; ð4Þ

where ucovar describes the influence of TQD treatment on parameter estimatesupop. The exponent COVARIATEwas assigned a value of 1 when TQD is treated.

Covariate effect was included for parameters for which it significantly (P ,0.05: OFV, reduction of 3.84 units) improved the model.

Physiologically Based PK Model. Plasma, brain ECF, and CSF concentra-tions of RIS and PAL were described based on the previously published semi-PBPK model (Westerhout et al., 2013) with slight modification. In short, thephysiologic volumes of each brain compartment were fixed. Also, drug transportswere governed by brain ECF and CSF flows, as well as transfer clearancesbetween plasma and each of the brain compartments (Fig. 1). During developmentof the semi-PBPK model, a stepwise approach was carried out: 1) a plasma PKmodel was developed for both compounds, and 2) the resulting plasma PKparameters were fixed and then used for development of the brain distributionmodel. In the brain distribution model, the lateral ventricle (LV) and thirdand fourth ventricle (TFV) were lumped into one compartment (LV and TFV),and it was assumed that for each compartment, the transfer clearances wereproportional to their physiologic volumes, which gives the following equations(eq. 5, 6, and 7):

CLPL2LV&TFV ¼ CLPL2CM � VLV&TFV

VCMð5Þ

CLLV&TFV 2PL ¼ CLCM2PL � VLV&TFV

VCMð6Þ

CLP2 gp;LV&TFV ¼ CLP2 gp;CM � VLV&TFV

VCM; ð7Þ

in which CLPL-LV&TFV and CLPL-CM represent transfer clearance from plasmato CSF in the LV and TFV and CM compartment, respectively. CLLV&TFV-PL andCLCM-PL represent CSF to plasma transfer clearances from the LV and TFV andthe CM compartment, respectively. CLP-gp,LV&TFV and CLP-gp,CM represent P-gpclearance in the LV and TFV and CM compartments, respectively. VLV&TFV andVCM represent physiologic volume in LV&TFV and CM compartment,respectively.

PD Model. After fixing all of the PK parameters obtained in the semi-PBPKanalysis, so-called sequential PK/PD modeling approach was performed. Todescribe prolactin response (Fig. 2), the previously developed pool model(Stevens et al., 2012) was applied except for feedback component. Briefly,prolactin is synthesized and stored in pool compartment with a zero-order rateconstant (ks,pr) and then released into plasma with a first-order rate constant (kr,pr).Plasma prolactin profile is described by a simple indirect response model(turnover model, eq. 8 and 9) in which prolactin comes from a pool compartmentby kr,pr and then is eliminated by a first-order rate constant (kel,pr):

154 Shimizu et al.

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

dCpool;pr

dt¼ ks;pr 2Cpool;pr � kr;pr ð8Þ

dCplasma;pr

dt¼ Cpool;pr � kr;pr 2Cplasma;pr � kel;pr ð9Þ

in which Cpool,pr and Cplasma,pr represent prolactin concentrations in poolcompartment and in plasma compartment, respectively.

At baseline (t = 0) without any drug influence, the following model results (eq.10, 11, and 12):

C0plasma;pr ¼ BASE ð10Þks;pr ¼ C0pool;pr � kr;pr ð11ÞC0pool;pr ¼ BASE � kel;pr

kr;prð12Þ

in which C0plasma,pr and C0pool,pr represent prolactin concentrationsat t=0 in the plasma compartment and in the pool compartment, re-spectively. BASE represents base line prolactin concentrations in plasmacompartment.

A drug effect (DE) on the release rate constant of prolactin was tested byinvestigating the use of linear- and Emax-response relationships shown in eq. 13and 14:

DElinear ¼ c� Cdrug ð13ÞDEEmax ¼ Emax � Cdrug

EC50 þ Cdrugð14Þ

in which c is the slope of the linear relationship between DE and theunbound drug concentrations, Cdrug; Emax is the maximum stimulation ofrelease rate; and EC50 is the drug concentration producing 50% of maximumstimulation.

For the drug effect part of the model, concentrations of RIS and PAL wereadded together as the “active drug concentration” because it was reasonable to

assume that Emax and EC50 of RIS and PAL were indistinguishable based on theirin vitro Ki literature values of 1.2 ng/ml for RIS and 1.7 ng/ml for PAL (vanBeijsterveldt et al., 1994). Incorporation of the drug effect part in the pool modelstructure gave eq. 15 and 16:

Fig. 1. Proposed semi-PBPK model to describe brain distribution for RIS and PAL. Hatched compartments have observations. QECF, flow rate of brain; ECF; QCSF, flow rateof CSF.

Fig. 2. Proposed pool model to describe prolactin profile in plasma. Prolactin issynthesized and stored in a pool compartment with a zero-order rate constant (ks,pr)and then released into plasma with a first-order rate constant (kr,pr). Releasedprolactin in plasma is eliminated by a first-order rate constant (kel,pr). Drug effect isdriven on kr,pr by unbound drug concentrations.

PKPD Analysis of Risperidone and Paliperidone 155

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

dCpool;pr

dt¼ ks;pr 2Cpool;pr � kr;pr � ð1þ DEÞ ð15Þ

dCplasma;pr

dt¼ Cpool;pr � kr;pr� � ð1þ DEÞ2Cplasma;pr � kel;pr ð16Þ

For the elimination rate of prolactin, a reported value (5.72 hours21) obtainedin the pool model analysis (Stevens et al., 2012) was taken for this study.

All the equations used in the analyses and the NONMEMmodel code used inthe final model are shown in Supplemental Appendixes 1 and 2, respectively.

Results

PK Data Analysis. Plasma protein binding was 5.8% and 8.2% forRIS and PAL, respectively. In vivo recoveries for the CP and CMprobeswere 22% and 10% for RIS, 18% and 10% for PAL, respectively. Thesedata were used to calculate unbound plasma, brain ECF and CSFconcentrations.PBPKModel. Based on the literature, a two-compartment model was

selected to describe PAL concentrations in plasma (Kozielska et al.,2012), and a three-compartment model was selected for RIS as thisgave more accurate descriptions than two-compartment model. PALplasma PK was connected to RIS plasma PK via metabolism clearance(CLRIS-PAL metab) (Fig. 1). In the end, five interindividual variabilityparameters were identified (onCLRIS, PL-E,CLRIS-PAL metab,CLPAL,PL-E, andVPAL,plasma). The effect of the coadministration of TQD was significanton CLRIS, PL-E and VPAL,plasma. The obtained plasma PK parameters werefixed and then used for the development of brain distribution model.As a start, the previously developed brain distribution model structure

(Westerhout et al., 2013) was used to describe brain distribution of RISand PAL. For P-gp functioning, the following transfer clearances acrossthe blood-brain barrier (BBB) were tested (eq. 17 and 18):

CLPL-BRAIN ¼ CLPL-BRAIN;passive-CLPL-BRAIN;P-gp� ¼ CLPL-BRAIN

� ucovar_TQD;   ucovar_TQD.1⋯BBB  influx  hindrance�

ð17ÞCLBRAIN-PL ¼ CLBRAIN-PL;passive þ CLBRAIN-PL;P-gp

� ¼ CLBRAIN-PL� ucovar_TQD;   0,ucovar_TQD,1⋯BBB  efflux enhancement

�

ð18Þ

where the subscript passive denotes passive clearance, P-gp denotesP-gp transport, subscript PL-BRAINmeans the direction from plasma tobrain ECF or CSF, and BRAIN-PL means the direction from brain ECFor CSF to plasma. ucovar_TQD represents the influence of TQD treatmenton the transfer clearances. For BBB influx hindrance, the apparentclearance from plasma to each brain compartment (CLPL-BRAIN)becomes larger (1 , ucovar_TQD) and is assumed to turn to passiveclearance (CLPL-BRAIN,passive) when a P-gp inhibitor TQD is treated. Asubtractionof apparent clearance frompassiveclearance (CLPL-BRAIN,passive2CL

PL-BRAIN) is considered to give P-gp transport clearance (CLPL-BRAIN,P-gp).

For BBB efflux enhancement, the apparent clearance from each braincompartment to plasma (CLBRAIN-PL) becomes smaller (0, ucovar_TQD, 1)and is assumed to turn to passive clearance (CLBRAIN-PL,passive) whenTQD is treated. A subtraction of passive clearance from apparentclearance (CLPL-BRAIN 2 CLPL-BRAIN,passive) is considered to give P-gptransport clearance (CLPL-BRAIN,P-gp). The data were better explained bythe influx hindrance mechanism 21 points lower OFV compared withthat of the efflux enhancement mechanism). In an earlier period of themodel development, we did not include the deep-brain compartmentsince there was no observation in this compartment unlike in theprevious study (Westerhout et al., 2013); however, the addition of deepbrain compartment significantly improved the model for RIS (OFVdropped 30 points), although this attempt did not significantly improvethe model for PAL.

The PK parameters finally estimated and the physiological parametersused for this analysis are shown in Table 1 and Table 2, respectively.P-gp contributed to the clearances from plasma-to-brain ECF for RISand PAL and from plasma-to-CSF for PAL. Also, interindividualvariability was identified on CLRIS,PL-ECF, CLRIS.PL-CSF, CLPAL,PL-ECF,and CLPAL,PL-CSF. The results of VPC in plasma, brain ECF, and CSFcompartments are shown in Fig. 3. Basically, the proposed modeldescribed well the observations in all compartments for both RIS andPAL. In addition, differences in RIS and PAL PK profiles with orwithout TQD treatment were nicely captured (Fig. 3). The resultingestimated PK parameters were fixed then provided for the followingprolactin PD analysis.PD Model. Plasma prolactin responses after administration of RIS or

PAL are shown in Fig. 4. Initially, brain ECF or CSF concentrationswere used to drive a drug effect based on our previous study (Stevenset al., 2012), which, however, resulted in a termination problem of the

TABLE 1

PK parameters estimated in the NMEM analysis

Population Parameters Unit Estimated %SEa

CLRIS,PL-E Liters/min 0.482 31.5VRIS plasma Liters 20.0 17.3QRIS,PL-PER1 Liters/min 0.404 80.9VRIS,PL-PER1 Liters 9.25 55.7QRIS,PL-PER2 Liters/min 0.0604 59.1VRIS,PL-PER2 Liters 13.5 55.1CLRIS-PAL, metab Liters/min 0.239 23.0CLRIS,PL-ECF ml/min 8.46 30.0CLRIS,PL-ECF, passive ml/min 10.91 —

CLRIS,PL-ECF, P-gp ml/min 2.45 —

CLRIS,ECF-PL, ml/min 3.38 25.6CLRIS,PL-CSF ml/min 0.373 18.6CLRIS,PL-CSF, passive ml/min 0.373 —

CLRIS,PL-CSF, P-gp ml/min 0 —

QRIS,ECF-Deep ml/min 2.64 38.3CLPAL,PL-E Liters/min 0.214 10.9VPAL,plasma Liters 19.7 9.7QPAL,PL-PER1 Liters/min 0.244 41.0VPAL,PER1 Liters 7.09 27.2CLPAL,PL-ECF ml/min 2.87 19.9CLPAL,PL-ECF, passive ml/min 8.50 —

CLPAL,PL-ECF, P-gp ml/min 5.63 —

CLPAL,ECF-PL ml/min 6.56 18.8CLPAL,PL-CSF ml/min 0.092 18.7CLPAL,PL-CSF, passive ml/min 0.112 —

CLPAL,PL-CSF, P-gp ml/min 0.020 —

QPAL,ECF-Deep ml/min 0 —

Covariate of TQD onCLRIS,PL-E 0.308 63.6CLRIS,PL-ECF 1.29 26.3CLPAL,PL-CSF 1.22 28.6CLPAL,PL-ECF 2.96 22.7VPAL,plasma 1.53 13.3

Interindividual variabilityCLRIS,PL-E 0.909 27.6CLRIS-PAL, metab 0.234 27.1CLPAL,PL-E 0.164 18.0VPAL,plasma 0.199 13.1CLRIS,PL-ECF 0.204 14.5CLRIS,PL-CSF 0.559 35.3CLPAL,PL-ECF 0.475 16.8CLPAL,PL-CSF 0.741 11.6

Proportional residual errorRIS plasma 0.274 13.2RIS brainECF 0.203 11.7RIS CSF 0.328 13.2PAL plasma 0.0871 9.2PAL brain ECF 0.138 9.0PAL CSF 0.248 9.1

aPercent relative standard error as calculated by NONMEM covariance step.

156 Shimizu et al.

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

mathematical model analysis. In contrast, the PD analysis was successfulwhen plasma concentrations were used to determine the drug effect.Therefore, we decided to keep plasma concentrations as a driver toinduce the prolactin response during this PKPD analysis. (In fact,analysis was again terminated with final PKPD model structure in casebrain ECF, or CSF concentrations were used for a drug effect instead ofplasma concentrations.)To describe the plasma prolactin profile, a simple indirect response

model with linear drug-effect relationship was tested first. Subsequently,the pool model was applied, which improved the OFV by 73 points. Thepool model was further improved by changing the drug-effect relation-ship from linear to Emax (OFV dropped 29 points). While applying theEmax drug-effect relationship, we attempted to estimate the EC50 for RISand PAL, separately; however, the coefficient of variation for estimatedEC50 of PAL was greater than 100%; in addition, the OFV did not dropso much (by 1.7 points). Thus, we decided to assume that the Emax andEC50 values of RIS and PAL are the same, which is consistent with thein vitro value (1.2 ng/ml for RIS and 1.7 ng/ml for PAL) in the literature(van Beijsterveldt et al., 1994). Consequently, the estimated EC50 values(1.73 ng/ml for both RIS and PAL) were close to those reported in theliterature. The inclusion of a positive feedback component into the poolmodel was not successful, probably as a result of overparameterization.Therefore, the feedback component was not included in the finalanalysis. The parameters estimated in the final PKPD model are shownin Table 3. Interindividual variability was identified on BASE. The VPC(Fig. 5) shows that, although not all of observations fell within the range

of the 90% prediction interval, the proposed model almost completelycaptured the median of observations.

Discussion

To describe prolactin response for RIS and PAL (active metabolite ofRIS), dopamine D2 receptor antagonists, a mechanism-based semi-PBPKPD model including detailed target site distribution, has beendeveloped. To our best knowledge, this is the first study thatsuccessfully developed a 12-compartmental semi-PBPKmodel based onextensive data in the different compartments. With this model, it waspossible to describe RIS and PAL profiles in brain, CSF, and plasma,which then was further linked to PD models to investigate themechanism of the effect of RIS and PAL on the release of prolactin inplasma.We proposed a semi-PB intrabrain distributionmodel and have shown

that this model structure can be applied to a variety of drugs from 1)acetaminophen, a model compound for only passive transport into andout of the brain kinetics (Westerhout et al., 2012); 2) quinidine, a modelcompound for P-gp–mediated transport kinetics (Westerhout et al.,2013); to 3) methotrexate, a model compound for substrate of multipletransporters (Westerhout et al., 2014). In this study, we appliedessentially the same model structure to the dataset of RIS and PAL,both of which are reported to be P-gp substrates in vitro (Ejsing et al.,2005; Zhu et al., 2007 and in vivo in mice (Kirschbaum et al., 2008;Bundgaard et al., 2012) and rats (Pacchioni et al., 2010; Doran et al.,2012). As a result, PK parameters including P-gp clearances wereestimated with manageable standard errors (less than 81%). EstimatedP-gp clearances for PAL were larger than those for RIS in both brainECF and CSF compartments (CLRIS,PL-ECF, P-gp vs. CLPAL,PL-ECF, P-gp,CLRIS,PL-CSF, P-gp vs. CLPAL,PL-CSF, P-gp), which is in line with reportsthat PAL has a greater affinity for P-gp than does RIS shown in knockout

Fig. 3. VPC results of RIS (upper row) and PALafter PAL dosing (middle row), after RIS dosing(lower row), in the presence or absence of TQD forplasma (left two columns), brain ECF (middle twocolumns), and CSF (right two columns). Blackdotted, 90% confidence interval; red, median ofsimulations; light green, median of observations.

Fig. 4. Plasma prolactin profiles (mean 6 S.D.) after administration of RIS (leftpanel) or paliperidone (right panel). Open and closed symbols represent pretreatmentof vehicle or TQD, respectively.

TABLE 2

Physiologic parameters used in the NMEM analysis

Parametersa Unit Value Description

Vdeep_brain ml 1.44 Volume of deep brainVbrainECF ml 0.29 Volume of brain ECFVLV&TFV_CSF ml 0.10 Volume of LV and TFV in CSFVCM_CSF ml 0.017 Volume of CM in CSFVSAS_CSF ml 0.18 Volume of SAS in CSFQECF ml/min 0.2 Flow rate of brain ECFQCSF ml/min 2.2 Flow rate of CSF

aSource: Westerhout et al., (2014).

PKPD Analysis of Risperidone and Paliperidone 157

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

mice (Wang et al., 2004; Doran et al., 2012). Regarding the direction ofP-gp functionality between plasma and CSF, some groups reported itsdirection from plasma to CSF (Rao et al., 1999; Kassem et al., 2007). Inour analysis, however, its direction was identified from CSF to plasmafor PAL (no functioning for RIS), which is quite similar to the finding inour previous study with quinidine (Westerhout et al., 2013).Kozielska et al. (2012) reported (total concentration-based) plasma

PK parameters of RIS and PAL in their PKPD modeling work forreceptor occupancy (Kozielska et al., 2012). If we calculate unbound-based PK parameters, assuming that unbound-based clearance (orvolume) equals total-based clearance (or volume) divided by plasma-unbound fraction ratio, corresponding PK parameters would be 0.10(liters/min) for CLRIS,PL-E, 4.8 (L) for VRIS,PL, 0.0055 (liters/min) forQRIS,PL-PER1, or QRIS,PL-PER2, 0.64 (V) for VRIS,PL-PER1 or VRIS,PL-PER2,0.049 (liters/min) for CLmRIS-PAL, metab, 0.041 (liters/min) for CLPAL,PL-E,3.0 (V) for VPAL,PL, 0.010 (liters/min) for QPAL,PL-PER1, and 0.58 (V) forVPAL,PL-PER1. (Rat body weight was also assumed to be 300 g.) Not allthese PK parameters are consistent with our PK parameters; however, only2 of 12 study cohorts in their investigation were conducted by i.v. dosing(the remaindere of the cohorts were done by s.c. or i.p. dosing), andstandard errors for some of the estimated parameters were relatively high(e.g., 54% for VPAL,PL-PER1, 128% for QPAL,PL-PER1). In addition, the doserange in their study was much wider: from 0.01–40 mg/kg for RIS andfrom 0.16–10 mg/kg for PAL.Based on these semi-PBPK models, we further investigated the PD

model to describe plasma prolactin response. Regarding a model todescribe the prolactin response, several other models have been reported,

such as the agonist-antagonist interaction (AAI) model (Bagli et al.,1999) and the AAI model incorporated with circadian rhythm (Friberget al., 2008; Ma et al., 2010). The AAI model used a hypotheticaldopamine concentration to inhibit prolactin synthesis, however, whichmay lead to unstable analysis. Also, circadian rhythm in rats was not sosignificant in relation to the magnitude of prolactin response afteradministration of a D2 receptor antagonist (Stevens et al., 2012).Therefore the focus of the current study was on a relatively simple PDmodel, being a pool model without circadian rhythm. Previously, wedeveloped the pool model with another D2 receptor antagonist, REM,using a drug input on brain ECF (Stevens et al., 2012). Here we appliedthe pool model structure to the prolactin response induced by RIS andPAL. Drug input was identified not on brain ECF or CSF, however, buton plasma in this RIS and PAL study. The Tmax (peak time) of prolactinwas 10 to 30 minutes; the Tmax of brain ECF concentrations was 40–60 minutes, and that of CSF concentrations was even later, 60–180 minutes. A pool model, an extended model of indirect responsemodel, describes a delayed PD response in relation to PK; however, theTmax of brain ECF or CSF was later than the Tmax of prolactin, whichmight be a part of why it was not possible to use brain ECF or CSFconcentrations to determine the prolactin response. Prolactin is producedin the anterior pituitary, one of the brain endocrine glands, but it can bedistinct from the other brain regions owing to the lack of BBB(Schlosshauer, 1993). In fact, Eyal et al. (2010) found that a distributionof [11-C] verapamil, an established P-gp substrate, into the pituitary was5 times greater than into almost any other intrabrain regions. Also, Kapuret al. (2002) investigated regional (pituitary and striatum) difference inD2 receptor occupancy using a variety of BBB-penetrating drugs andfound that a limited BBB-penetrating drug shows high receptoroccupancy in the pituitary but low in the striatum, whereas a rapidlyBBB-penetrating drug shows similar receptor occupancy between thepituitary and striatum. This report suggests that plasma concentration ismore relevant than brain concentration to receptor occupancy in thepituitary. Additionally, Arakawa et al. (2010) found that the plasmaprolactin level is positively correlated not with D2 receptor occupancy inthe temporal cortex but with that in the pituitary. Taken together, RISand PAL would increase the plasma prolactin level not via the brain butactually via plasma exposure since the target tissue can be (at leastmainly) accessed from plasma.In our previous REM study, only the parent compound was measured

and considered an active moiety; however, Mohell et al. (1993) reportedthe production of pharmacologically active metabolites FLA797 andFLA908, of which in vitro efficacies in rats are 230 and 25 times morepotent than REM, respectively. FLA797 concentrations in the brain were2%–4% of that of REM (at 20–60 minutes) (Ögren et al., 1993).Additionally, i.p. dosing showed significantly higher efficacy than s.c.dosing (Ahlenius et al., 1997), where a greater amount of FLA797 wasproduced for i.p. dosing than for s.c. dosing; REM concentration itself

Fig. 5. Prolactin VPC results after administration of RIS and PAL. Black dotted,90% confidence interval; red, median of simulations; light green, median ofobservations.

TABLE 3

PD parameters estimated in the NMEM analysis

Parameter Unit Estimate %S.E.a Description

Pool model kel,pr /h 5.72 (fixed) Elimination rate constantkr,pr /h 0.223 17.9 Release rate constant from pool compartmentBASE ng/ml 7.1 9.7 Baseline prolactin concentration in plasma

Drug-effect model Emax,RIS 4.24 17.8 Risperidone maximum effectEC50,RIS ng/ml 1.73 51.4 Risperidone concentration inducing half the Emax,RIS

Emax,PAL 4.24 17.8 Paliperidone maximum effectEC50,PAL ng/ml 1.73 51.4 Paliperidone concentration inducing half the Emax,PAL

Interindividual variability BASE 0.141 18.5 Variability on BASEResidual error 0.314 6.1 Proportional

aPercent relative S.E. as calculated by NONMEM covariance step.

158 Shimizu et al.

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from

was lower for i.p. dosing (Ögren et al., 1993). Furthermore, a biphasicresponse (on dopamine turnover) was seen after oral dosing of REM,whereas a monophasic response was seen after i.p. dosing (Magnussonet al., 1987). The authors did not state specific reason for such a biphasicresponse, but formation of metabolites, such as FLA797 or FLA908,through first-pass hepatic metabolism may explain the appearance of thesecond peak observed after oral dosing. Given those reports, activemetabolites (or parent plus active metabolites) may significantlycontribute to its in vivo efficacy after administration of REM to rats.Since active metabolites were not considered in our previous study(Stevens et al., 2012), we have concluded that brain ECF can be thetarget site for REM drug effects on prolactin in plasma. It remains to bedetermined if there is indeed a role for active metabolites in governingthe prolactin plasma concentrations after REM administration. It isshown that (semi-)mechanistic PKPDmodeling can be used to delineatethe target site of drugs specifically by examining whether the main druginput to the effect is from plasma, brain ECF, or CSF. Our semi-PBPKmodel made it possible to examine the target site since it adequatelyprovided PK profiles in plasma, brain ECF, and CSF for both RIS andPAL.In conclusion, a semi mechanism-based PKPDmodel including target

site distribution of RIS and its active metabolite PAL has beensuccessfully developed to describe plasma prolactin response. 1) BrainECF and CSF distributions of RIS and PAL were adequately describedby applying already established semi-PBPK model to our data set with aslight modification. In addition, differences in PK between with andwithout TQD treatment were nicely captured for both RIS and PAL. 2)The plasma prolactin profile was sufficiently described by applying apool model. 3) During the PD model development, we confirmed thatplasma drug concentrations are more relevant than CSF or brain ECFconcentrations to explain the prolactin response.

Acknowledgments

The authors thank Coen van Hasselt and Yumi Emoto-Yamamoto (Division ofPharmacology, Leiden Academic Center for Drug Research, Leiden, TheNetherlands) for their support in pharmacometric analysis and the editorialassistance of Corine Visser (Division of Pharmacology, Leiden Academic Centerfor Drug Research, Leiden, The Netherlands).

Authorship ContributionsConducted experiments: den Hoedt, Mangas-Sanjuan, Cristea, K Geuer, van

den Berg, Hartman.Performed data analysis: Shimizu, Bellanti, de Lange.Wrote or contributed to the writing of the manuscript: Shimizu, de Lange.

References

Ahlenius S, Ericson E, Hillegaart V, Nilsson LB, Salmi P, and Wijkström A (1997) In vivo effectsof remoxipride and aromatic ring metabolites in the rat. J Pharmacol Exp Ther 283:1356–1366.

Arakawa R, Okumura M, Ito H, Takano A, Takahashi H, Takano H, Maeda J, Okubo Y,and Suhara T (2010) Positron emission tomography measurement of dopamine D2 receptoroccupancy in the pituitary and cerebral cortex: relation to antipsychotic-induced hyper-prolactinemia. J Clin Psychiatry 71:1131–1137.

Bagli M, Süverkrüp R, Quadflieg R, Höflich G, Kasper S, Möller H-J, Langer M, Barlage U,and Rao ML (1999) Pharmacokinetic-pharmacodynamic modeling of tolerance to the prolactin-secreting effect of chlorprothixene after different modes of drug administration. J Pharmacol ExpTher 291:547–554.

Bundgaard C, Jensen CJN, and Garmer M (2012) Species comparison of in vivo P-glycoprotein-mediated brain efflux using mdr1a-deficient rats and mice. Drug Metab Dispos 40:461–466.

Doran AC, Osgood SM, Mancuso JY, and Shaffer CL (2012) An evaluation of using rat-derivedsingle-dose neuropharmacokinetic parameters to project accurately large animal unbound braindrug concentrations. Drug Metab Dispos 40:2162–2173.

Ejsing TB, Pedersen AD, and Linnet K (2005) P-glycoprotein interaction with risperidone and9-OH-risperidone studied in vitro, in knock-out mice and in drug-drug interaction experiments.Hum Psychopharmacol 20:493–500.

Eyal S, Ke B, Muzi M, Link JM, Mankoff DA, Collier AC, and Unadkat JD (2010) RegionalP-glycoprotein activity and inhibition at the human blood-brain barrier as imaged by positronemission tomography. Clin Pharmacol Ther 87:579–585.

Friberg LE, Vermeulen AM, Petersson KJ, and Karlsson MO (2009) An agonist-antagonist in-teraction model for prolactin release following risperidone and paliperidone treatment. ClinPharmacol Ther 85:409–417.

Kapur S, Langlois X, Vinken P, Megens AA, De Coster R, and Andrews JS (2002) The differentialeffects of atypical antipsychotics on prolactin elevation are explained by their differential blood-brain disposition: a pharmacological analysis in rats. J Pharmacol Exp Ther 302:1129–1134.

Kassem NA, Deane R, Segal MB, Chen R, and Preston JE (2007) Thyroxine (T4) transfer fromCSF to choroid plexus and ventricular brain regions in rabbit: contributory role of P-glycoproteinand organic anion transporting polypeptides. Brain Res 1181:44–50.

Keizer RJ, van Benten M, Beijnen JH, Schellens JH, and Huitema AD (2011) Piraña and PCluster:a modeling environment and cluster infrastructure for NONMEM. Comput Methods ProgramsBiomed 101:72–79.

Kirschbaum KM, Henken S, Hiemke C, and Schmitt U (2008) Pharmacodynamic consequences ofP-glycoprotein-dependent pharmacokinetics of risperidone and haloperidol in mice. Behav BrainRes 188:298–303.

Kozielska M, Johnson M, Pilla Reddy V, Vermeulen A, Li C, Grimwood S, de Greef R, GroothuisGM, Danhof M, and Proost JH (2012) Pharmacokinetic-pharmacodynamic modeling of the D2

and 5-HT (2A) receptor occupancy of risperidone and paliperidone in rats. Pharm Res 29:1932–1948.

Ma G, Friberg LE, Movin-Osswald G, and Karlsson MO (2010) Comparison of the agonist-antagonist interaction model and the pool model for the effect of remoxipride on prolactin. Br JClin Pharmacol 70:815–824.

Magnusson O, Mohringe B, Thorell G, and Lake-Bakaar DM (1987) Effects of the dopamine D2selective receptor antagonist remoxipride on dopamine turnover in the rat brain after acute andrepeated administration. Pharmacol Toxicol 60:368–373.

Mohell N, Sällemark M, Rosqvist S, Malmberg A, Högberg T, and Jackson DM (1993) Bindingcharacteristics of remoxipride and its metabolites to dopamine D2 and D3 receptors. Eur JPharmacol 238:121–125.

Ögren S, Lundström J, and Nilsson L (1993) Concentrations of remoxipride and its phenolicmetabolites in rat brain and plasma: relationship to extrapyramidal side effects and atypicalantipsychotic profile. J Neural Transm 94:199–216.

Pacchioni AM, Gabriele A, Donovan JL, DeVane CL, and See RE (2010) P-glycoprotein inhibitionpotentiates the behavioural and neurochemical actions of risperidone in rats. Int J Neuro-psychopharmacol 13:1067–1077.

Rao VV, Dahlheimer JL, Bardgett ME, Snyder AZ, Finch RA, Sartorelli AC, and Piwnica-WormsD (1999) Choroid plexus epithelial expression of MDR1 P glycoprotein and multidrugresistance-associated protein contribute to the blood-cerebrospinal-fluid drug-permeability bar-rier. Proc Natl Acad Sci USA 96:3900–3905.

Scheller D and Kolb J (1991) The internal reference technique in microdialysis: a practical ap-proach to monitoring dialysis efficiency and to calculating tissue concentration from dialysatesamples. J Neurosci Methods 40:31–38.

Schlosshauer B (1993) The blood-brain barrier: morphology, molecules, and neurothelin. Bio-Essays 15:341–346.

Stevens J, Ploeger BA, Hammarlund-Udenaes M, Osswald G, van der Graaf PH, Danhof M, and deLange EC (2012) Mechanism-based PK-PD model for the prolactin biological system responsefollowing an acute dopamine inhibition challenge: quantitative extrapolation to humans. JPharmacokinet Pharmacodyn 39:463–477.

van Beijsterveldt LE, Geerts RJ, Leysen JE, Megens AA, Van den Eynde HM, Meuldermans WE,and Heykants JJ (1994) Regional brain distribution of risperidone and its active metabolite9-hydroxy-risperidone in the rat. Psychopharmacology (Berl) 114:53–62.

Wang J-S, Ruan Y, Taylor RM, Donovan JL, Markowitz JS, and DeVane CL (2004) The brainentry of risperidone and 9-hydroxyrisperidone is greatly limited by P-glycoprotein. Int J Neu-ropsychopharmacol 7:415–419.

Westerhout J, Ploeger B, Smeets J, Danhof M, and de Lange EC (2012) Physiologically basedpharmacokinetic modeling to investigate regional brain distribution kinetics in rats. AAPS J 14:543–553.

Westerhout J, Smeets J, Danhof M, and de Lange EC (2013) The impact of P-gp functionality onnon-steady state relationships between CSF and brain extracellular fluid. J PharmacokinetPharmacodyn 40:327–342.

Westerhout J, van den Berg D-J, Hartman R, Danhof M, and de Lange EC (2014) Prediction ofmethotrexate CNS distribution in different species - influence of disease conditions. Eur J PharmSci 57:11–24.

Zhu H-J, Wang J-S, Markowitz JS, Donovan JL, Gibson BB, and DeVane CL (2007) Risperidoneand paliperidone inhibit p-glycoprotein activity in vitro. Neuropsychopharmacology 32:757–764.

Address correspondence to: Shinji Shimizu, 140-8710, Daiichi-Sankyo CO.,LTD., Shinagawa R&D Center 1-2-58, Hiromachi, Shinagawa-ku, Tokyo, Japan.E-mail: shimizu.shinji.d3@daiichisankyo.co.jp. Elizabeth C. M. de Lange, SystemsPharmacology HB810b, LACDR, Leiden University, Einsteinweg 55, 2333 CC/P.O.Box 9502, 2300 RA, Leiden, The Netherlands. E-mail: ecmdelange@lacdr.leidenuniv.nl

PKPD Analysis of Risperidone and Paliperidone 159

at ASPE

T Journals on M

arch 9, 2020dm

d.aspetjournals.orgD

ownloaded from