1521-0103/358/2/294–305$25.00 http://dx.doi.org/10.1124/jpet.116.232447THE JOURNAL OF PHARMACOLOGY AND EXPERIMENTAL THERAPEUTICS J Pharmacol Exp Ther 358:294–305, August 2016Copyright ª 2016 by The American Society for Pharmacology and Experimental Therapeutics

Examining the Uptake of Central Nervous System Drugs andCandidates across the Blood-Brain Barrier s

Scott G. Summerfield, Yanyan Zhang, and Houfu LiuGlaxoSmithKline R&D, Platform Technology and Science, Park Road, Ware (S.G.S); and GlaxoSmithKline R&D, PlatformTechnology and Science, Zhangjiang Hi-Tech Park, Pudong, Shanghai, China (Y.Z., H.L.)

Received February 7, 2016; accepted May 17, 2016

ABSTRACTAssessing the equilibration of the unbound drug concentrationsacross the blood-brain barrier (Kp,uu) has progressively replacedthe partition coefficient based on the ratio of the total concen-tration in brain tissue to blood (Kp). Here, in vivo brain distributionstudies were performed on a set of central nervous system(CNS)–targeted compounds in both rats and P-glycoprotein(P-gp) genetic knockout mice. Several CNS drugs are charac-terized by Kp,uu values greater than unity, inferring facilitateduptake across the rodent blood-brain barrier (BBB). Examplesare shown in which Kp,uu also increases above unity on knockout

of P-gp, highlighting the composite nature of this parameter withrespect to facilitated BBB uptake, efflux, and passive diffusion.Several molecules with high Kp,uu values share common struc-tural elements, whereas uptake across the BBB appears moreprevalent in the CNS-targeted drug set than the chemicaltemplates being generated within the current lead optimizationparadigm. Challenges for identifying high Kp,uu compounds arediscussed in the context of acute versus steady-state data andcross-species differences. Evidently, there is a need for betterpredictive models of human brain Kp,uu.

IntroductionDrug distribution across the blood-brain barrier (BBB)

forms a central element in linking pharmacokinetics to thepharmacological action of xenobiotics within the centralnervous system (CNS) (Reichel, 2015). Unbound drug thatcrosses the BBB into brain extracellular fluid (ECF) is avail-able to distribute to the biophase or bind nonspecificallywithin the lipid-rich components of brain tissue (Jeffrey andSummerfield, 2007). Recent examples linking unbound drugbrain concentrations with potency, target engagement oreffect (Large et al., 2009; Watson et al., 2009; Liu and Chen,2015) have helped to transition the thinking of CNS drugdiscovery scientists away from total concentration measure-ments and more toward unbound drug concentrations. Simi-larly, interpreting brain distribution has shifted away fromthe brain-to-blood partition ratio based on total concentra-tions (Kp) toward the unbound ratio (Kp,uu) comprising brainECF and unbound blood concentrations (Liu and Chen, 2015).Kp,uu is a steady-state distribution term denoting the

unbound concentration gradient across the BBB (Guptaet al., 2006; Hammarlund-Udenaes et al., 2008). If Kp,uu issubstantially lower than unity, then drug passage across theBBB is restricted by some factor, such as poor permeability

(e.g., atenolol; Summerfield et al., 2007) or carrier-mediatedefflux (Loryan et al., 2014). The most widely recognized BBBefflux transporter for reducing brain Kp,uu is P-glycoprotein(P-gp), which exerts its effects on many chemical templates(Raub, 2006). Much of the founding in vivo work characteriz-ing P-gp efflux relied on Kp measurements; however, thisparameter in itself does not indicate whether a drug moleculeis the subject of active transport; indeed, additional measure-ments are required, such as determining Kp in an in vivotransporter knockout model (chemical or genetic) or deter-mining the P-gp efflux ratios in an in vitro system.A Kp,uu value approaching unity suggests that passive

diffusion is an important distribution property of the drugmolecule, or that multiple carrier-mediated processes arebalanced (e.g., influx and efflux). Although Kp,uu alone cannotdistinguish these two scenarios, concentration dependency inthis value or the permeability-surface product might suggesttransporter involvement (Cisternino et al., 2013). Finally, aKp,uu value in excess of unity would suggest facilitated trans-fer of unbound drug across a cellular barrier and the estab-lishment of a concentration gradient. An example of this isgabapentin, where the uptake into brain intracellular fluid(Fridén et al., 2007) supported the involvement of LAT1carrier-mediated transport (Su et al., 1995).Fewer literature examples are available demonstrating

uptake into the CNS relative to drug efflux; presumably, thisis because drug efflux has a far more negative impact on drug

dx.doi.org/10.1124/jpet.116.232447.s This article has supplemental material available at jpet.aspetjournals.org.

ABBREVIATIONS: aCSF, artificial cerebrospinal fluid; BBB, blood-brain barrier; CNS, central nervous system; ECF, extracellular fluid; ER, efflux ratio;GF 120918A, XXX 9,10-dihydro-5-methoxy-9-oxo-N-[4-[2-(1,2,3,4-tetrahydro-6,7-dimethoxy-2-isoquinolinyl) ethyl]phenyl]-4-acridine-carboxamide; GSK,GlaxoSmithKline; HPLC, high-performance liquid chromatography; Kp, total concentration of gradient across the BBB; Kp,uu, unbound concentration ofgradient across the BBB; MS/MS, tandem mass spectrometry; PBS, phosphate-buffered saline; P-gp, P-glycoprotein; pKa, acidic and basic ionizationconstants; SD, Sprague-Dawley; TEER, transepithelial electrical resistance.

294

http://jpet.aspetjournals.org/content/suppl/2016/05/18/jpet.116.232447.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

efficacy measurements during lead optimization and earlydrug development and is therefore identified more readily infavorable chemical templates. However, Kp,uu provides apowerful tool for probing the prevalence of facilitated uptakein CNS drugs or lead optimization compounds and whetherthere are any key structural features or physicochemicalproperties that promote the passage of unbound drug acrossthe BBB. In this study, we evaluated Kp,uu across approxi-mately 50 CNS-targeted compounds. These Kp,uu values werederived from steady-state intravenous CNS penetration stud-ies in rodents [rat, mdr1a/b(2/2) mice, and mdr1a/b(1/1)mice] to yield Kp, values which were then corrected fornonspecific binding to blood and brain. Intracerebral micro-dialysis was performed on two examples in which Kp,uu wassubstantially greater than unity (bupropion and clozapine)with paracetamol included as a control (e.g., Kp,uu � 1). Anassessment of Kp,uu across the lead optimization space withinGlaxoSmithKline (GSK) is also presented as a guide to thelikelihood of other structural hotspots promoting increasedbrain distribution of unbound drug. In this case, braindistribution information was collated from both steady-stateand acute CNS penetration studies.

Materials and MethodsCompound Selection. Fifty-three molecules were examined in

theCNS compound set (Supplemental Table 1 for structures) andwereselected on the basis of their spread in physicochemical properties(Summerfield et al., 2006). The following compounds were obtainedfrom commercial sources: risperidone, chlorpromazine, bupropion,pergolide mesylate, amitriptyline, diazepam, meprobamate, and tema-zepam (SigmaChemicalCompany, Poole, Dorset, UK); and venlafaxine,citalopram, pemoline, andmidazolam (Sequoia Research Products Ltd.,Oxford,UK). All other compoundswere available from in-house sources.Compounds used as retrodialysers for microdialysis were purchased asfollows: fluperlapine (Biomol International L.P., Plymouth Meeting,PA) and [13C2,

15N]acetaminophen (Cambridge Isotope Laboratories,Andover, MA). [2H6]Bupropion was synthesized at GlaxoSmithKline.All other chemicals were of analytical grade, and all solvents were ofhigh-performance liquid chromatography (HPLC) grade.

Animals. Adult male Sprague-Dawley (SD) rats (body weight274–376 g; Charles River UK Ltd., Tranent Edinburgh, Scotland)were used. Male FVB mdr1a/b(2/2) knockout mice and geneticallymatched mdr1a/b(1/1) wild-type expressing P-gp were purchasedfrom Taconic Farms, Inc. (Germantown, NY). Mouse unbound brainand blood fraction determinations by means of equilibrium dialysiswere conducted in male CD-1 mice (Charles River Laboratories,Margate, UK); this mouse strain was selected based on availabilitytogether with the fact that internal and external work has showedlittle impact in nonspecific brain binding across species (Summerfieldet al., 2008) and within strain (Kalvass and Maurer, 2002). All animalstudies were ethically reviewed and carried out in accordance withAnimals (Scientific Procedures) Act 1986 and the GSK Policy on theCare, Welfare and Treatment of Animals.

Preparation for Mouse Steady-State Intravenous BrainDistribution Studies. Each mouse (at least 6 weeks old at the timeof surgery) was cannulated via the jugular vein (for drug administra-tion) while under isoflurane anesthesia, with the cannula exteriorizedat the back of the neck. Animals were placed in a jacket and tethercounterbalance system and allowed to recover for at least 3 days priorto dosing. Three male control mdr1a/b (1/1) and three male mdr1a/b(2/2) knockout mice were dosed with each compound.

The dose solution for each test compound was prepared on the dayof dosing using an appropriate formulation for intravenous admin-istration. The infusion rate and infusion time required to reach

steady-state blood concentrations (taken as at least 3.5 half-lives)were determined separately for each compound, and were based onprior studies to establish the intravenous pharmacokinetics. Bloodsamples (25 ml) were taken from the tail vein at 0.5-hour intervalsduring the last 2 hours of the infusion to confirm steady-state bloodconcentrations had been achieved. At the end of the infusion period,the mice were decapitated and the brain removed and homogenizedwith an equal weight of purified water (ELGA maxima; ELGA Lab-Water, HighWycombe, UK). Triplicate weighed aliquots (50ml) of brainhomogenate were stored at ca. 280°C prior to HPLC/tandem massspectrometry (MS/MS) analysis. Samples were extracted by a methodbased on protein precipitation using acetonitrile (250 ml containing asuitable internal standard). Brain concentrations were corrected forresidual blood volume using 15ml/g of brain tissue as the vascular space(Brown et al., 1986).

Preparation for Rat Steady-State Intravenous Brain Distri-bution Studies. Adult male SD rats (body weight 274–376 g;Charles River UK Ltd.) were used. Each rat was cannulated via thejugular vein (for drug administration) and femoral vein (for bloodsampling) while under isoflurane anesthesia, the cannulae beingexteriorized at the back of the neck. Animals were placed in a jacketand tether and allowed to recover for at least 2 days prior to dosing.The dose solution for each test compound was prepared on the day ofdosing using an appropriate formulation for intravenous administra-tion. Three rats were dosed for each compound. The infusion rate andinfusion time were determined separately for each compound. Bloodsamples were taken at 0.5-hour intervals during the last 2 hours of theinfusion to confirm steady-state blood concentrations had beenachieved. At the end of the infusion period, the rats were exsangui-nated and decapitated, and the brain was removed. Blood samples (ca.80 ml) were collected into tubes containing K2EDTA as anticoagulantand aliquoted immediately after collection. An aliquot (50 ml) of eachblood sample was diluted with an equal volume of purified water andstored at ca.280°C to await analysis. Each brainwashomogenizedwithan equal weight of purified water (ELGA maxima; ELGA LabWater,High Wycombe, UK). Weighed aliquots (50 ml) of brain homogenatewere taken and stored at ca. 280°C prior to HPLC/MS/MS analysis.Samples were extracted by a method based on protein precipitationusing acetonitrile (250 ml containing a suitable internal standard).Brain concentrations were corrected for residual blood volume using15 ml/g as the vascular space (Brown et al., 1986).

Preparation for Steady-State Intravenous In Vivo BrainMicrodialysis Studies. Adult male SD rats (body weight 274–376 g;Charles River UK Ltd.) were used. Prior to surgery, rats receivedantibiotic (ampicillin or Oxytetracycline) and analgesic (carprofen).Under isoflurane anesthesia, rats were cannulated via a femoral veininto the vena cava (for drug administration) and jugular vein (for bloodsampling) (Griffiths et al., 1996). Following venous cannulation,animals were placed in a stereotaxic frame (David Kopf Instruments,Tujunga, CA), and a 20-mm midline incision was made to expose theskull surface. The periosteum was infiltrated with local anesthetic(0.15 ml of mepivacaine hydrochloride) and reflected to expose thebregma. Measurements were taken from the bregma using appropri-ate anterior and lateral stereotaxic coordinates (anterior, 13.2 mm;lateral, 11.2 mm), and a hole was drilled for the guide probe. Fourfurther holeswere drilled in the skull surface to enable fixing screws tobe inserted. The guide probe was lowered (ventral, –2.0 mm; allcoordinates according to Pellegrino et al., 1979) and cemented intoposition, and the was skin sutured around the headcap as previouslydescribed (Criado et al., 2003). Jugular and femoral cannulae wereexteriorized at the back of the neck. Animals were placed in jacketsand collars, housed individually, and allowed to recover for 72 hoursprior to dosing.

In vivo Studies to Determine the Concentration of TestCompounds in Blood, Brain, and Cerebrospinal Fluid Follow-ing Intravenous Administration to Steady State. Microdialysisprobes (CMA11, 2-mm probe length; CMA Microdialysis, Kista,Sweden) perfused with artificial cerebrospinal fluid (aCSF; 145 mM

Uptake of CNS Drugs and Candidates across the BBB 295

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

NaCl, 2.7 mMKCl, 1.2 mM CaCl2.2H20, 1.0 mMMgCl2.6H20, 2.0 mMNa2HPO4, pH 7.4) at a flow rate of 1.5 ml/min were inserted throughthe previously implanted guide cannula into the cortex. Rats wereplaced in Perspex bottom Raturn bowls (BASi, West Lafayette, IN),were freely moving, and had access to food and water ad libitumthroughout the experiment. Probes were dialyzed with aCSF for atleast 1 hour prior to dosing to allow tissue equilibration (Chaurasiaet al., 2007).

Dose solutions of test compounds and reference solutions forretrodialysers were prepared on the day of dose administration andfiltered before administration. Aliquots of dose solutions and referencesolutions were collected prefilter, postfilter, and postdose and storedat approximately 280°C prior to analysis by HPLC-MS/MS. Dosesolutions of test compounds were prepared in an appropriate formu-lation for intravenous administration. Test compounds were admin-istered as constant rate intravenous infusions over 6 or 8 hours via thefemoral vein cannula. The infusion rate and infusion time weredetermined for each test compound from pharmacokinetic datagenerated previously “in house” (data not shown) to achieve targetedsteady-state blood concentrations (Css, range 0.1–1 mg/ml) designed tooptimize quantitative measurement of test compounds. The actualdose rates used (mg/kg/h) and the subsequent blood clearances(ml/min/kg) were as follows: acetaminophen, 2887 6 12 mg/kg/h and45 6 2 ml/min/kg; clozapine, 1474 6 19 mg/kg/h and 115 613 ml/min/kg; and bupropion, 1826 6 106 mg/kg/h and 127 623 ml/min/kg. Reference solutions were prepared at a concentrationof 250 ng/ml retrodialyser in aCSF containing 0.1% (v/v) dimethylsulf-oxide. Reference solutions were administered via the probe prior toand throughout the i.v. infusion period.

Serial blood samples were collected via the jugular cannula atregular intervals throughout the infusion to characterize the bloodconcentration-time profile and confirm the steady-state blood concen-trations had been achieved (Cblood). Blood samples (ca. 90 ml) werecollected into tubes containing K2EDTA as anticoagulant, a 50-mlaliquot diluted with an equal volume of purified water and stored atapproximately 280°C prior to analysis by HPLC-MS/MS. For thedetermination of brain ECF concentrations (Cubrain,ECF) and proberecovery, serial aCSF samples were collected at regular intervalsthroughout the infusion period into a refrigerated fraction collectormaintained at 4°C.

Following the end of the infusion, rats were exsanguinated underanesthesia (Euthatal [Merial Animal Health, Woking, UK] adminis-tered via the jugular vein cannula), CSF samples were collected fromthe cisterna magna for determination of CSF concentration (CCSF), andthe brains were removed. A 50-ml aliquot of exsanguinated bloodsamples was diluted with an equal volume of deionized water andstored at approximately 280°C prior to analysis by HPLC-MS/MS fordetermination of drug blood concentration. Another aliquot of exsan-guinated blood samples was dilutedwith an equal volume of phosphate-buffered saline (PBS) for determination of the unbound drug fraction(fublood) by equilibrium dialysis and stored at 4°C. Brains were rinsed inchilled saline, and superficial blood vessels were removed, weighed, andhomogenized with an equal volume of PBS. Brain homogenate wasstored as either weighed aliquots (50 ml) at280°C for determination oftotal brain concentration (Cbrain) by HPLC-MS/MS, or as bulk at 4°C fordetermination of the unbound drug fraction (fubrain) by equilibriumdialysis. Total brain concentrations were corrected for residual bloodvolume using a value of 15 ml/g as the vascular space (Brown et al.,1986).

Probe Calibration. Probes were calibrated by retrodialysisin vivo with probe recoveries being calculated as follows:

Lossin vivo 5Cin– Cout

Cin(1)

where Cout is the concentration in the dialysate, Cin is the concentra-tion in the perfusate, and percentage of loss is assumed to equalpercentage of recovery. Similarly, probes were calibrated in vitro by

loss of retrodialyser and recovery of test compound following thein vivo phase:

Lossin vitro 5Cin– Cout

Cin(2)

where Cout is the concentration in the dialysate and Cin is theconcentration in the perfusate, and

Recoveryin vitro 5Cout

Cin(3)

where Cout is the concentration in the dialysate and Cin is theconcentration in the bathing solution. Recovery values used toestimate Cubrain,ECF were selected based on the most appropriatemethod for each test compound.

Bioanalysis of Blood, CSF, ECF, and Brain HomogenateSamples. Aliquots of blood, brain, and CSF (50 ml) andmicrodialyste(22.5 ml) were extracted by means of protein precipitation (250 ml)containing a suitable internal standard. The performance of internalstandard chosen was considered to be acceptable by the analyzers inall cases. Calibration curve ranges and analytical conditions aredetailed in Supplemental Table 2. Quality control samples (n 5 3 perconcentration) accompanied the calibration curves at the assay limit ofquantification (LLQ; 4 � LLQ), the geometric mean of the calibrationrange, and the upper limit of quantification. All samples were analyzedby means of LC-MS/MS on a PE-Sciex API-4000 tandem quadrupolemass spectrometer (Applied Biosystems, Ontario, Canada) using aTurbo V Ionspray operated at a source temperature of 700°C (80 psi ofnitrogen). Samples (3–10 ml) were injected using a CTC Analytics HTSPal autosampler (Presearch, Hitchin, UK).

Equilibrium Dialysis Experiments. A 96-well equilibrium di-alysis apparatus was used to determine the free fraction in the bloodand brain for each drug (HT Dialysis LLC, Gales Ferry, CT).Membranes (3-kDA cutoff) were conditioned in deionized water for40 minutes, followed by conditioning in 80:20 deionized water:ethanolfor 20 minutes, and then rinsed in deionized water before use. Ratbrains were obtained fresh on the day of the experiment and werehomogenized with PBS to a final composition of 1:2 brain:PBS bymeans of ultrasonication (Tomtec Autogiser; Receptor Technologies,Adderbury, Oxon, UK) in an ice bath. Diluted brain homogenatewas spiked with the test compound (1000 ng/g), and 100-ml aliquots(n 5 6 replicate determinations) were loaded into the 96-wellequilibrium dialysis plate. Dialysis versus PBS (100 ml) was carriedout for 5 hours in a temperature-controlled incubator at 37°C (StuartScientific, Watford, UK) using an orbital microplate shaker at 125 rpm(Stuart Scientific). At the end of the incubation period, aliquots ofbrain homogenate or PBS were transferred to Matrix ScreenMatetubes (Matrix Technologies, Hudson,NH), and the composition in eachtube was balanced with control fluid such that the volume of PBS tobrain was the same. Sample extraction was performed by the additionof 200ml of acetonitrile containing an internal standard. Sampleswereallowed to mix for 15 minutes and then centrifuged at 2465 � g in96-well blocks for 20 minutes (Eppendorf 5810R; VWR International,Poole, Dorset, UK). The unbound fraction was determined as the ratioof the peak area in buffer to that in brain, with correction for dilutionfactor according to following equation (Kalvass and Maurer, 2002):

fu5ð1=DÞ��

1fuapparent

– 1�1 1

D

� (4)

where D 5 dilution factor in brain homogenate, and fuapparent is themeasured free fraction of diluted brain tissue.

MDR1-MDCKII Cell Passive Membrane Permeability andP-gp Transport Measurements. Cell culture media, bovine serumalbumin, medium supplements (L-glutamine, 20 mM sodium bicar-bonate, nonessential amino acids, sodium pyruvate, colchicine), andHanks’ balanced salt solution was purchased from Invitrogen

296 Summerfield et al.

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

(Carlsbad, CA). Transwells (12-well, 1.13-cm2 area, 0.4-mm pores)were purchased from Costar (Cambridge, MA). Krebs-Ringer bicar-bonate buffer was obtained from Sigma-Aldrich (St. Louis, MO), anddrugs were supplied by GlaxoSmithKline (Stevenage, UK).

Multidrug resistance Madin-Darby canine kidney (MDR1-MDCK)type II cells were obtained from the National Institutes of Health(Bethesda, MD). Cells were maintained in minimum essential Eagle’smedium containing 2 mM L-glutamine, 20 mM sodium bicarbonate,0.1 mM nonessential amino acids, 1 mM sodium pyruvate, and 10%bovine serum albumin supplemented with 0.2 mM colchicine tomaintain P-gp expression. For permeability experiments, cells withpassage numbers 24–33were seeded at a density of 60,000 cells/cm2 onrat type I collagen-coated polycarbonate membranes in 12-well trans-well plates. The experiments were performed on the eighth day afterseeding. Prior to the permeation assay, the transepithelial electricalresistance (TEER) was measured on each cell monolayer using anEndohm Voltameter (World Precision Instruments, Sarasota, FL).The MDR1-MDCK cell systems used in transport experiments had ahigh TEER value ranging from 1800 to 2000V cm2, which differs fromlower TEER MDR1-MDCK models often used in CNS screens (Braunet al., 2000). The permeability assay buffer was Hanks’ balanced saltsolution containing 10 mMHEPES and 15 mM glucose at pH 7.4. Thecells were dosed on the apical side and basolateral side, and incubatedat 37°C with 5% CO2 and 90% relative humidity under shakingconditions (150 rpm) to reduce the influence of the unstirred waterlayer. The test compounds were prepared to a final concentration of3 mM. In the P-gp inhibition assays, the cells were incubated with2 mM GF 120918A (9,10-dihydro-5-methoxy-9-oxo-N-[4-[2-(1,2,3,4-tetrahydro-6,7-dimethoxy-2-isoquinolinyl) ethyl]phenyl]-4-acridine-carboxamide), a known P-gp inhibitor. Drug permeation was testedin two directions, apical-to-basolateral (A-to-B) and basolateral-to-apical (B-to-A), in triplicate. Sampling was done 60 minutes afterdosing. The apparent permeability coefficient (Papp) was calculatedas follows:

Papp 5dCrdt

�

VrA×C0

�(5)

where dCr/dt is the slope of the cumulative concentration in thereceiver compartment versus time in mM/s, Vr is the volume oftransport assay buffer in the receiver compartment, A is the area ofthe cell monolayer, and C0 is the drug concentration in the dosingsolution. The recovery of most compounds exceeded 80%. The bi-directional efflux ratio was determined as follows:

Efflux ratio5Papp ðB-AÞPapp ðA-BÞ

(6)

P-gp effluxwas confirmed by determining the bidirectional efflux ratioin the presence of elacridar (2 mM) and observing an efflux ratio ofunity.

Data Analysis. Total blood concentrations were measured inserial blood samples. Total Cblood values were determined using themean measured blood concentrations during the final hours of the i.v.infusion once steady state had been achieved. Total Cbrain values weremeasured in terminal brain homogenate samples, and CCSF wasmeasured in terminal CSF samples. Cbrain was used with Cblood todetermine the CNS penetration of test compounds.

The unbound brain-to-blood concentration, Kp,uu, was determinedas follows, where Cu represents the unbound concentration in therelevant compartment:

Kp;uu 5Cubrain;ECF

Cublood(7)

Molecular and Physicochemical Descriptors. Acidic and ba-sic ionization constants (pKa)were calculated fromChemAxon version5.4.1.1 (ChemAxon Kft, Budapest, Hungary). cLogP and cLogD7.4were calculated using the ChemAxon .NET 5.11.4.30. Topological

polar surface area was calculated by the method of P. Ertl, and only Nand O are considered as polar atoms (Ertl et al., 2000). Hydrogen bondacceptor and hydrogen bond donor were calculated with any N, O, or Satoms with a lone pair, excluding basic amines and anilines, etc.

ResultsTable 1 shows rat and mouse brain Kp,uu values and in vitro

P-gp efflux ratios (ERs) for a series of CNS-targetedmolecules.Kp,uu covers a 320-fold range from 0.022 (ergotamine) to 7.1(perphenazine). Atenolol is included as a marker of theextravascular brain volume (Kp,uu 5 0.012; Summerfieldet al., 2007). Individual Kp and unbound fraction values aresupplied in Supplemental Table 3.In Fig. 1, Kp,uu is plotted versus increasing in situ BBB

permeability (data reported in Summerfield et al., 2008) andshows no strong correlation between rate of drug passageacross the BBB and the extent of distribution at steady state(Hammarlund-Udenaes et al., 2008; Summerfield and Dong,2013). Several compounds that are characterized have lowKp,uu values (,0.5) and are identified as substrates forhuman P-gp in vitro (ER . 2): sumatriptan (Kp,uu 5 0.034,ER 5 2.9), primidone (Kp,uu 5 0.38, ER 5 3.6), amoxapine(Kp,uu 5 0.35, ER 5 4.1), rizatriptan (Kp,uu 5 0.043, ER 58.4), metoclopramide (Kp,uu 5 0.40, ER 5 13), risperidone(Kp,uu 5 0.20, ER 5 21), thiothixene (Kp,uu 5 0.13, ER 5 48),ergotamine (Kp,uu5 0.022,ER5 50), andmesoridazine (Kp,uu50.10, ER5 87). Gabapentin is also associated with a relativelylow rat brain Kp,uu value of 0.44, although it is subject tofacilitated transport (Su et al., 1995).Several of the CNS-targeted molecules possess tricyclic

groups (Fig. 2) and were also characterized by rat brain Kp,uu

values greater than 2; this includes the phenothiazinemoiety of chlorpromazine, perphenazine, thioridazine, andtrifluorperazine. Related tricyclic ring systems lead to Kp,uu .2 for clozapine, olanzapine, doxepin, imipramine, loxepin,mianserin, and mirtazapine. For the tricyclic CNS drugswhere Kp,uu is close to or less than unity, BBB efflux appearsto occur. From steady-state CNS penetration studies usingwild-type and genetic P-gp knockout mice, maprotiline isidentified as a moderate P-gp substrate (Table 1). Indeed, themouse brain Kp,uu of maprotiline increases from 1.42 to3.06 between the wild-type and mdr1a/b(2/2) knockout mice,which suggests the presence of competing efflux and uptakeprocesses. Similarly, amitriptyline has been reported as aP-gp substrate based on in vivo knockout studies (Uhr et al.,2000, 2007; Abaut et al., 2009). Sun et al. (2006) reportedmultiple transporters involved in the passage of carbamaze-pine across cultured rat brain microvascular endothelial cells.For fluphenazine, the rat brain Kp,uu is 1.58 but increases toover 6 in both mdr1a/b(1/1) and mdr1a/b(2/2) mice (Table 1).Amoxapine and thiothixene are identified as P-gp substratesin vitro (Table 1) but are not entirely excluded from the ratbrain since their respective Kp,uu values are 0.35 and 0.13. Thebrain Kp,uu for tacrine was determined as 1.61 in rats(Table 1), increasing to over 2 in mice (Table 1). Striatalextracellular fluid concentrations of tacrine were shown todecrease with increasing dose (Sung et al., 2005), and trans-porter studies in human embryonic kidney 293 cells suggestedthe involvement of multiple cationic transporter systems inrats. Quetiapine and midazolam are the only tricyclic exam-ples in this compound set where the rat brain Kp,uu values are

Uptake of CNS Drugs and Candidates across the BBB 297

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

close to unity but show no propensity for efflux by P-gp in vitroor in vivo (Table 1). Currently, no further information isreported in the literature to suggest whether quetiapine issubject to efflux via another transport system.Eight additional compounds were characterized by Kp,uu

values greater than 2 (Fig. 2): amantadine, atomoxetine,bupropion, fluoxetine, lamotrigine, memantine, pergolide,and sertraline. Of these, amantadine and memantine share

a common tricyclodecanamine moiety, and a concentrationdependency on the BBB permeability of memantine has beenreported that arises from facilitated transport via a cationicproton antiporter (Mehta et al., 2013).The rat brain Kp,uu of lamotrigine was determined as 4.86 in

this study, which accords with it being a substrate for OCT1 inhuman brain endothelial cells (Dickens et al., 2012). It isnoteworthy that themouse brain Kp,uu for lamotrigine ismuch

TABLE 1Brain Kp,uu values for rat and mouse [mdr1a/b(+/+) and mdr1a/b(2/2)] quoted as mean 6 standarddeviationThree rats and mice (n = 3) were used for each drug. Kp values derived from oral dosing studies.

CNS Drug In Vitro P-gpEfflux Ratio

Kp,uu Kp,uu Kp,uu

Rat mdr1a/b(+/+) Mouse mdr1a/b(2/2) Mouse

Mean S.D. Mean S.D. Mean S.D.

Atenolol — 0.012 0.003 — — — —Amantadine 0.8 3.01 0.85 — — — —Amitriptyline 1.9 1.33 0.40 — — — —

Amoxapine 4.1 0.35 0.17 0.23 0.06 0.55 0.10Atomoxetine 3.4 2.13 0.85 2.35 0.17 4.10 0.35Bupropion 1.4 4.97 0.91 6.14 2.54 5.41 0.52Carbamazepine 0.8 0.76 0.20 — — — —Chlorpromazine 0.8 3.41 0.81 2.79 0.42 3.19 0.45Citalopram 20.1 1.30 0.33 1.47 0.17 4.58 0.48Clozapine 1.3 3.83 0.80 2.51 0.14 2.84 0.32Diazepam 1.0 0.96 0.16 — — — —

Donepezil 2.3 1.77 0.23 1.19 0.09 2.64 0.29Doxepin 1.9 3.11 0.59 — — — —Ergotamine 49.6 0.02 0.01 ,0.01 — 0.30 0.12Ethosuximide 0.8 1.50 0.75 — — — —Fluoxetine 1.2 5.23 1.87 — — — —

Fluphenazine 2.3 1.58 0.57 6.12 1.31 7.77 1.25Gabapentin 0.8 0.44 0.11 — — — —Haloperidol 1.3 1.77 0.71 1.54 0.21 1.89 0.49Imipraminea — 3.52 0.32 — — — —Isocarboxazid 1.2 0.18 0.04 — — — —Lamotrigine 1.6 4.86 0.72 1.00 0.11 1.09 0.08Loxapine 1.2 2.95 0.89 — — — —Maprotiline 4.0 0.98 0.47 1.42 0.44 3.06 0.45Meprobamate 3.3 0.77 0.10 0.61 0.10 0.80 0.20Memantinea — 6.21 1.56 — — — —Mesoridazine 87.1 0.10 0.03 0.19 0.02 3.30 0.73Metoclopramide 13.2 0.40 0.08 0.90 0.15 6.23 0.83Mianserina — 3.22 0.43 — — — —Midazolam 2.0 1.08 0.10 — — — —

Mirtazapine 1.3 3.85 0.50 — — — —Olanzapine 4.9 2.13 0.60 6.86 1.52 9.53 1.38Pemoline 1.2 0.55 0.15 — — — —Pergolide 1.8 3.19 0.59 — — — —Perphenazine 4.7 7.12 2.17 5.99 1.53 6.17 1.58Phenelzine 1.4 1.48 0.34 — — — —Phenytoin 2.8 0.66 0.11 3.05 0.42 2.79 0.37Primidone 3.6 0.38 0.10 0.68 0.10 1.25 0.22Quetiapine 1.5 1.26 0.14 1.15 0.20 1.27 0.34Risperidone 20.8 0.21 0.04 0.18 0.07 2.29 0.42Rizatriptan 8.4 0.04 0.01 0.40 0.27 0.53 0.10Selegiline 0.8 1.08 0.17 — — — —Sertraline 2.6 2.67 0.51 3.87 1.03 4.22 1.52Sumatriptan 2.9 0.04 0.01 0.02 0.01 0.08 0.02Tacrine 1.1 1.61 0.56 2.25 0.83 2.68 0.97Temazepam 1.6 0.80 0.12 — — — —

Thioridazine 33.7 3.50 1.67 — — — —Thiothixene 47.7 0.13 0.04 0.34 0.06 3.67 0.47Tiagabine 25.8 0.62 0.21 0.77 0.14 1.17 0.20Trazodone 1.1 1.54 0.28 — — — —Trifluoperazine 7.3 3.35 1.10 2.18 0.47 3.35 0.70Venlafaxine 4.7 1.91 0.30 3.49 0.71 5.43 0.98Zaleplon 5.6 0.59 0.10 0.77 0.13 1.38 0.23Ziprasidone 1.7 0.75 0.22 4.10 0.91 7.17 1.85Digoxin — — — 0.12 0.03 3.41 0.42

—, not determined.aKp values derived from oral dosing studies.

298 Summerfield et al.

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

closer to unity (Table 1), but there is no indication of efflux byP-gp substrate (Table 1). The unbound blood fractions in bothrat and mouse are comparable for lamotrigine (0.28 and 0.29,respectively), as are the unbound brain fractions (0.27 and0.20, respectively). Taken together, these data may support aspecies difference in the BBB transport of lamotrigine be-tween rat and mouse.The rat brain Kp,uu of fluoxetine was determined as 5.23 in

this study. In contrast, results from Doran et al. (2005)suggested that the mouse brain Kp,uu was close to unity inboth mdr1a/b(1/1) or mdr1a/b(2/2) mice. Fluoxetine is sub-ject to an extremely high degree of nonspecific brain tissuebinding, and fubrain was measured as 0.00094 in mice (Doranet al., 2005) but was around 4-fold higher in this study at0.0040. Since greater variability is likely to be seen at thelowest extremity of the fubrain assay, the high Kp,uu could be anexperimental artifact in this study, although comparable datain a single species are not available for fluoxetine. It isnoteworthy that the structural analog atomoxetine wascharacterized by a rat brain Kp,uu of 2.13 and increased from2.35 to 4.10 in the P-gp knockout mouse model (Table 1).Conversely, in vivo microdialysis determined the Kp,uu ofatomoxetine as 0.7 (Kielbasa et al., 2009). The unboundblood/plasma fractions are comparable between these studies,although Kp differs almost 2-fold between this study (Kp 5 19)and the following microdialysis (Kp 5 11).Pergolide is an ergot derivative characterized by a rat brain

Kp,uu value of 3.19 and was not identified as a P-gp substrate,in agreement with another reported study (Vautier et al.,2008). Although few reports are available on the transport ofergotamine itself in vivo, an ergot-containing analog, lysergicacid, was seen to accumulate in frog choroid plexus cells by anNa/K ATPase–dependent process with the characteristics of

active transport (Wright, 1972). Mulac et al. (2012) alsoreported ergotamine as a breast cancer resistance proteinsubstrate based on in vitro observations, and hence, bidirec-tional transport may be occurring. Last, sertraline wasmeasured with a rat brain Kp,uu of 2.67 and a mouse brainKp,uu around 4. In contrast, Doran et al. (2005) reported themouse brain Kp,uu of sertraline as 1.5 following subcutaneousadministration, rather than i.v. infusion to steady-state usedin this study.Mdr1a/b Mouse Model. Brain penetration data collected

in mice highlight several CNS-targeted molecules where thebrain Kp,uu increases markedly above unity following P-gpknockout (Table 1); these drugs are citalopram (Kp,uu 5 4.58),risperidone (Kp,uu 5 2.29), ziprasidone (Kp,uu 5 7.17), digoxin(Kp,uu5 3.41), mesoridazine (Kp,uu5 2.79), thiothixene (Kp,uu53.67), metoclopramide (Kp,uu 5 6.23), and venlafaxine (Kp,uu 57.59). Mesoridazine and thiothixene both possess the pheno-thiazine moiety but are also strong P-gp substrates. Forcitalopram, Bundgaard et al. (2012) reported an increase inthe mouse brain Kp,uu to 4.8 in P-gp–deficient mice. Althoughtheprimary focus of thisworkwas onaP-gp species difference, theputative involvement of an uptake mechanism has been recog-nized (Rochat et al., 1999).For risperidone, the mouse brain Kp increased from 0.78 to

8.02 in the study reported by Doran et al. (2005), whichequates a change in Kp,uu from 0.26 to 2.63, respectively, whencorrecting for the fuplasma and fubrain data cited in the samestudy. In contrast, Bundgaard et al. (2012) reported Kp,uu

values around unity in P-gp knockout rats and mice followinginfusion to steady state. A structural analog, ziprasidone, wascharacterized by a mouse brain Kp,uu value of 7.17 in ourpresent study, which may point to a potential mechanism forfacilitated uptake for both structurally related drugs. Digoxin

Fig. 1. Bar graph showing steady-state Kp,uu values for CNS drugs, ordered along the x-axis in terms of increasing in situ BBB permeability (referencedata reported in Summerfield et al., 2008).

Uptake of CNS Drugs and Candidates across the BBB 299

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

overdose is known to induce CNS symptoms (Kelly and Smith,1992), and despite P-gp playing a key role in maintaining lowbrain concentrations of digoxin, high affinity for a specificorganic anion transporter may help to increase Kp,uu aboveunity in the mdr1a/b(2/2) model (Noé et al., 1997; Taub et al.,2011). For metoclopramide, the rat brain Kp,uu (0.40) in thisstudy agrees with another report (0.37) (Doran et al., 2012),but no other comparative work in P-gp–deficient species isavailable, nor could any literature reports be found thatsuggest the involvement of carrier-mediated transport formetoclopramide. Work in transfected human embryonic kid-ney 293 cells has shown metoclopramide is a substrate forOATP1A2 along with several triptans, although this trans-porter does not have rodent homologs (Cheng et al., 2012). Forvenlafaxine, the Kp,uu was determined to be 1.91 in the ratbrain, 4.87 in mdr1a/b(1/1) mice, and 7.59 in mdr1a/b(2/2)mice. Recently, its structural isomer tramadol was character-ized by a brain Kp,uu of 2.3 following microdialysis, and

subsequent mechanistic studies suggested carrier-mediatedtransport by means of a proton-coupled organic cation anti-porter (Kitamura et al., 2014).In Vivo Microdialysis. To cross-reference the in vitro

determinations of Cubrain (e.g., Cbrain � fubrain,homogenate), twohigh brain Kp,uu drugs were selected for rat brain micro-dialysis (clozapine and bupropion). Acetaminophen was alsoincluded as a control because Kp,uu was measured as beingclose to unity. Clozapine and bupropion were also selected, asthey differ greatly in terms of their unbound blood fractions(0.066 vs. 0.391) and unbound brain fractions (0.011 vs.0.171).The blood and brain ECF unbound concentration-time profilesfor bupropion are displayed in Fig. 3. The correspondingconcentration-time profiles for acetaminophen and clozapinewere reported previously (Summerfield and Dong, 2013).For acetaminophen, the unbound concentrations were

comparable between blood (4.97 6 1.00 mM), brain ECF (3.52 60.52mM), and cisternalCSF (4.9560.27mM) (Table 2). Following

Fig. 2. Structures for CNS drugs con-taining tricyclic moieties. Tricyclic CNS-targeted compounds where rat brainKp,uu $ 2 (A), tricyclic CNS-targetedcompounds where rat brain Kp,uu , 2 (B),and other CNS-targeted compounds whererat brain Kp,uu $ 2 (C). *Data availableto suggest BBB efflux and/or multiplecarrier-mediated transport processes.

300 Summerfield et al.

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

microdialysis, the rat brain Kp,uu of acetaminophen was de-termined as 0.7360.16. Equilibriumdialysis of ex vivo blood andbrain homogenates from the rats infused with acetaminophenrealized a mean unbound blood fraction of 0.741 6 0.128and a mean unbound brain fraction of 0.724 6 0.050. Thesevalues were used to correct the total concentrations in bloodand brain to unbound concentrations, which yields anex vivo rat Kp,uu of 0.613 6 0.136.In the case of bupropion, the rat brain Kp,uu was determined

as 3.3 6 1.7 following microdialysis (range 2.35–5.77). Theparticular rat associated with the highest Kp,uu value was alsocharacterized by an approximately 2-fold larger brain Kp

(18 vs. 10) and 2-fold higher brain ECF concentrations (2.64vs. 1.28 mM). The unbound concentrations of bupropion werehigher in brain ECF (1.626 0.69mM) relative to both the blood(0.504 6 0.066 mM) and cisternal CSF (0.250 6 0. 053 mM).Equilibrium dialysis of ex vivo blood and brain homogenatesfrom the rats infusedwith bupropion realized amean unboundblood fraction of 0.537 6 0.056 and a mean unbound brainfraction of 0.149 6 0.008. Using the total concentrations inbrain and correcting for the ex vivo brain binding gave acomparable ex vivo rat brain Kp,uu of 3.08 6 0.71.

For clozapine, the rat brain Kp,uu was determined as 2.2 60.2 followingmicrodialysis, consistent with the data presentedin Table 1. The initial brain ECF measurements for clozapinewere below the quantification limit of the analytical method,but concentrations were measurable at steady state (albeitwith more variability than bupropion and acetaminophen).The unbound concentration of clozapine was highest incisternal CSF (0.156 6 0.005 mM), followed by brain ECF(0.077 6 0.008 mM), and then lowest in blood (0.036 6 0.002mM). Equilibrium dialysis of ex vivo blood and brain homog-enates from the rats infused with clozapine realized a meanunbound blood fraction of 0.059 6 0.012 and a mean unboundbrain fraction of 0.0109 6 0.0015. Using the total concentra-tions in brain and correcting for the ex vivo brain binding gavea comparable ex vivo rat brain Kp,uu of 3.18 6 0.50.Kp,uu Assessments of CNS Drug Candidates from

Lead Optimization. Rat brain Kp,uu data were available for771 proprietary compounds synthesized against CNS targets,of which Kp,uu was $2 for 41 compounds (5.5%) and $3 for19 compounds. Of these, CNS penetration measurementswere taken at steady state for four compounds, whereas theremainder were from acute studies. Table 3 outlines the

Fig. 3. Unbound drug concentrations of bupropion in blood and brain extracellular fluid following in vivo microdialysis in male Sprague-Dawley rats.Unbound blood concentrations (d) are reported as the mean 6 standard deviation (n = 4); concentrations at individual brain extracellular fluid arereported for rat #1 (X), rat #2 (s), rat #3 (¤) and rat #4 (◻).

TABLE 2Summary of pharmacokinetic and brain distribution values for acetaminophen, bupropion, and clozapinein rats

Acetaminophen Bupropion Clozapine

fu(blood) 0.741 6 0.128 0.537 6 0.056 0.0590 6 0.0120fu(brain) 0.724 6 0.050 0.149 6 0.008 0.0109 6 0.0015Cublood (mM) 4.97 6 1.00 0.504 6 0.066 0.036 6 0.002Cubrain ECF (mM) 3.52 6 0.52 1.62 6 0.69 0.077 6 0.008Cisternal CSF (mM) 4.95 6 0.27 0.250 6 0.053 0.156 6 0.005Kp,uu ECF 0.73 6 0.16 3.30 6 1.70 2.16 6 0.23Cubrain estimated (mM) 3.13 6 0.21 1.64 6 0.08 0.124 6 0.018Ex vivo Kp,uu 0.613 6 0.136 3.10 6 0.70 3.18 6 0.50

Uptake of CNS Drugs and Candidates across the BBB 301

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

physicochemical properties of these high Kp,uu compounds.Generally, the number of hydrogen bond donors and acceptorswas low (median values of 1 and 2, respectively), and manywere characterized by reasonably acidic pKa values. Of theseCNS discovery candidates, a wide range of structures werenoted that were mostly distinct from the CNS compound set,presumably because of the novel CNS targets being prose-cuted in lead optimization. Four of these proprietary mole-cules were characterized by fused ring systems, but in general,the chemotypes were not comparable to those detailed in theCNS compound set.In comparison, the CNS-targeted compound set was char-

acterized by 19 out of 53 (36%) with Kp,uu $ 2, and 15 (28%)with Kp,uu $ 3. Analysis of physicochemical propertiesassociated with rat Kp,uu values revealed that drugs withKp,uu . 2 had reduced topological polar surface area ascompared with drugs with Kp,uu , 0.5 or 0.5 # Kp,uu # 2(Supplemental Fig. 1; Supplemental Table 1). However, thecompound set is not large enough to form a definitive conclu-sion. A more detailed analysis in relationships between com-pound physicochemical properties and rat Kp,uu values in alarger data set was published elsewhere (Zhang et al., 2016).Selection of the compound set was originally conceived torepresent the span of physiochemical properties of moleculestargeting the CNS, e.g., rather than maximizing the span oftargets themselves. The compound set comprises agonists,antagonists, inhibitors, and reuptake inhibitors of variousimportant CNS receptors, such as dopamine (e.g., clozapineand perphenazine), serotonin (e.g., risperidone and citalopram),gamma aminobutyric acid (e.g., midazolam and meprobamate),noradrenaline (e.g., amoxapine and venlafaxine), and mono-amine oxidase (e.g., selegiline and isocarboxazid).Many of theseCNS drugs were also approved in an era when pharmacologyscreening played a more prominent role in progressing earlyleads, whereas today’s paradigm relies far more heavily onbatteries of in vitro Drug Metabolism and Pharmacokinetics(DMPK) screens to triage drug candidates. Evidently, duringthat period, the concept of Kp,uu had not been conceived, andhence serendipity played a large part in selecting drugcandidates subject to uptake across the BBB.

DiscussionThe introduction of the Kp,uu parameter has provided new

insights into the processes that occur during drug passage acrossthe BBB. This work focused on molecules in which there isevidence to support facilitated passage across theBBB—namely,where Kp,uu is markedly greater than unity. However, thepresent study also highlights examples inwhich there are clearlyat least two transport processes acting in opposing directions,such that the overall Kp,uu is closer to unity. Examples of this arecitalopram, metoclopramide, and thiothixene.

Donepezil poses an interesting example of potential cross-species differences in Kp,uu. Being a modest P-substrate in themouse, the donepezil Kp,uu increases from 1.19 and 2.64(Table 1), whereas Kim et al. (2010) noted concentration-dependent transport of donepezil and proposed this to bemediated by organic cation transporters, such as the cholinetransport system. Recently, we reported the minipig Kp ofdonepezil as 14.9, identified as compound “L” following a 1-houri.v. infusion at a dose of 1mg/kg (Summerfield andDong, 2013).The unbound fractions in plasma and brain were 0.2606 0.027and 0.162 6 0.017, realizing a minipig brain Kp,uu of 7.2. Inhumans, the brain Positron Emission Tomography (PET) bio-distribution of 11C-donepezil was approximately 20 in healthyelderly subjects (Okamura et al., 2008), which translates to aKp,uu estimate of 10, after correcting for plasma protein bind-ing (Nagy et al., 2004) and assuming the brain tissue bindingis comparable across species (Summerfield et al., 2008).Given the interplay between passive transfer across the

BBB and the potential for competing efflux, experimentalconditions are likely to play an important part in whether adrug candidate is identified as a substrate for drug uptake.Absolute Kp,uu values may be dependent on the circulatingunbound concentration bathing the BBB (e.g., the doseadministered) and whether any of these transport processesare subject to saturation. At the limit, brain Kp,uu wouldapproach unity if the transport processes are saturated andpassive diffusion becomes the dominant process of drugpassage across the BBB. Interestingly, a review by theInternational Transporter Consortium noted that, other thancarrier-mediated transfer via P-gp, decreases or increases inbrain Kp were generally less than a few-fold in many trans-porter knockout models (Kalvass et al., 2013; Tweedie et al.,2013).It is important to recognize not only that Kp,uu is a steady-

state parameter, but also that it is most often generatedindirectly, e.g., by correcting the total brain concentration forthe unbound brain fraction. Hence, it is imperative thatsteady-state conditions are achieved to accurately reflect thebrain Kp,uu, which is exemplified in Fig. 4. The BBB initiativethat led to many of the publications reported by GSKperformed a large battery of in vivo and in vitro tests on thisset of CNS-targeted molecules, including both acute (oral) andsteady-state (i.v.) brain distribution studies. Acute studieswere included because many of the early estimates of braindistribution during lead optimizationwere derived from short-term pharmacology models, whereas steady-state brain dis-tribution studies were supplemented for promising molecules.Comparison of the acute and steady-state brain Kp datahighlighted a strong correlation between the two data sets;however, the slope of correlation was approximately 0.5rather than unity, the steady-state brain Kp values generallybeing higher than the acute Kp values. Figure 4 shows this

TABLE 3Proprietary CNS drug candidates with Kp,uu $ 2 in rat and their physicochemical properties

Rat Kp,uu MW cLogP cLogD7.4 tPSA HBA HBD pKa (Most Acidic) pKa (Most Basic)

Max 11.6 484 7.5 7.5 85.8 5 2 14 9.9Min 2.01 221 1 0.3 3.2 0 0 5.9 0Median 2.91 403 3.7 2.8 47.5 2 1 14 7.6

HBA, hydrogen bond acceptor; HBD, hydrogen bond donor; MW, Molecular Weight; tPSA, topological polar surfacearea.

302 Summerfield et al.

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

correlation in terms of Kp,uu, where the difference is mostmarked for the higher Kp,uu compounds. Similarly, the mousedata set originally reported by Doran et al. (2005) has been avaluable resource for understanding brain distribution interms of unbound concentrations, although only one com-pound was described by a brain Kp,uu greater than 2 in themdr1a/b(1/1) mouse (e.g., methylphenidate; Liu et al., 2008).In the correspondingmdr1a/b(2/2) strain, eight compoundsarecharacterized by Kp,uu $ 2, e.g., cyclobenzaprine, fluvoxamine,hydrocodone, methylphenidate, metoclopramide, nortriptyline,propoxyphene, and risperidone. These Kp,uu values were de-termined from the area under the curve measurements follow-ing subcutaneous dosing over 5 hours (concentration samplingfrom four time points: 0.5, 1, 2.5, or 5 hours). The acute studiesshown inFig. 4 were performed following oral dosingwith bloodand brain samples collected up to 8 hours.Many CNS drugs possessing tricyclic functions, such as

chlorpromazine and olanzapine, appear to undergo facilitatedtransfer across the rodent BBB. An uptake mechanism forclozapine across HL-60 cells has been proposed previously(Henning et al., 2002), but further work is required tocharacterize whether there is a common carrier-mediatedsystem or whether this structural element would be useful inother CNS-targeted drug candidates. Since the pH values ofblood and brain ECF are comparable, these increases in brainKp,uu for tricyclic molecules are unlikely to be the result of iontrapping, and in general, their acidities and basicities varywidely (see Supplemental Data File 1; Supplemental Table S2).Not all of the examples highlighted in this study (where

Kp,uu is significantly greater than unity) can be supportedby literature data, but perhaps this is unsurprising. Drugefflux, particularly P-gp, has blighted drug discovery becausethe inherent reduction in brain penetration leads to a corre-sponding reduction in efficacy. However this is not the case foruptake across the BBB; higher unbound brain concentrations

would still lead to an efficacious effect in vivo, albeit at lowerdoses. The traditional Kp measurement would have providedno indication of uptake, and discovery pharmacokinetic/pharmacodynamic models are rarely refined enough to high-light efficacy at an unexpectedly low dose. Indeed, prior to theinception of Kp,uu, carrier-mediated transport may have onlybeen speculated when the extent of brain penetrationappeared at odds with the physicochemical properties (e.g.,gabapentin and lamotrigine). In the case of lipophilic mole-cules, which readily traverse the BBB, a difference of a few-fold in brain penetration arising from facilitated uptakemechanism would not be discernible without the determina-tion of Kp,uu.Predicting difference in Kp,uu across species remains a

significant challenge, especially when only rodent data oracute brain distribution studies are performed. So how doesthe industry obtain better estimates for the human Kp,uu?Take an example such as donepezil, where literature datasuggest Kp,uu is far higher in humans than would be antici-pated from the rodent data. The early estimates of the humandose are highly sensitive to the value of brain Kp,uu; a 3-foldchange in Kp,uu would attenuate the corresponding humandose prediction 3-fold (assuming unbound brain concentra-tions are driving efficacy). In vitro BBB models for humanKp,uu have been reported based on stem cells (Cecchelli et al.,2014), and the promise is immense for helping progress thenext generation of CNS drugs. Although the number ofreported in vitro and in vivo human brain Kp,uu compoundsis currently small, it is noteworthy that both bupropion andlamotrigine showed human brain Kp,uu estimates close tounity. However, the estimates for human Kp,uu were madeusing CSF concentrations, whereas this may not be a reliablesurrogate for a variety of reasons (de Lange, 2013). A questionalso remains concerning the junction tightness in human cell-based systems. Low TEER values could lead to increased

Fig. 4. Scatter plot showing rat brain Kp,uu values derived from brain penetration studies in male Sprague-Dawley rats; the x-axis denotes steady-statedata, and the y-axis shows acute data following oral dosing. The dashed line represents the line of unity, and the solid line is a best fit to the data by linearregression. AUC, area under the curve.

Uptake of CNS Drugs and Candidates across the BBB 303

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

directional flux of drug across the in vitro BBB system; inparticular, Kp,uu values tend to be only a few-fold greater thanunity.In summary, the introduction of Kp,uu has enabled a deeper

insight into the factors affecting the unbound concentrationgradient across the BBB. In vivo microdialysis studies onclozapine and bupropion confirmed the corresponding in vitroKp,uu estimates (Table 2). Clearly, examples such as citalo-pram and maprotiline demonstrate that a measured brainKp,uu around unity need not necessarily be the result of apurely passive transfer process (Table 1). Differences in braindistribution across species, concentration dependency of braindistribution (in vitro or in vivo), or evidence of uptake fromother pharmacokinetic information may highlight the poten-tial for uptake across the BBB.

Acknowledgments

The authors thank the following people for assisting in theexperimental work: Lee Abberley, Leanne Cutler, Kim Matthews,Ann Metcalf, Maria Osuna, Kathryn Patterson, Alexander Stevens,and Sac-Pham Tang. The authors also acknowledge the advice andsupport of David Begley, Phil Jeffrey, Rod Porter, and Jasminder Sahithroughout the blood-brain barrier initiative efforts within GSK.

Authorship Contributions

Participated in research design: Summerfield.Conducted experiments: Summerfield.Performed data analysis: Summerfield, Zhang, Liu.Wrote or contributed to the writing of the manuscript: Summerfield,

Zhang, Liu.

References

Abaut AY, Chevanne F, and Le Corre P (2009) Influence of efflux transporters onliver, bile and brain disposition of amitriptyline in mice. Int J Pharm 378:80–85.

Braun A, Hämmerle S, Suda K, Rothen-Rutishauser B, Günthert M, Krämer SD,and Wunderli-Allenspach H (2000) Cell cultures as tools in biopharmacy. Eur JPharm Sci 11 (Suppl 2):S51–S60.

Brown EA, Griffiths R, Harvey CA, and Owen DA (1986) Pharmacological studieswith SK&F 93944 (temelastine), a novel histamine H1-receptor antagonist withnegligible ability to penetrate the central nervous system. Br J Pharmacol 87:569–578.

Bundgaard C, Jensen CJ, and Garmer M (2012) Species comparison of in vivoP-glycoprotein-mediated brain efflux using mdr1a-deficient rats and mice. DrugMetab Dispos 40:461–466.

Cecchelli R, Aday S, Sevin E, Almeida C, Culot M, Dehouck L, Coisne C, EngelhardtB, Dehouck MP, and Ferreira L (2014) A stable and reproducible human blood-brain barrier model derived from hematopoietic stem cells. PLoS One 9:e99733.

Chaurasia CS, Müller M, Bashaw ED, Benfeldt E, Bolinder J, Bullock R, Bungay PM,DeLange EC, Derendorf H, and Elmquist WF, et al. (2007) AAPS-FDA workshopwhite paper: microdialysis principles, application and regulatory perspectives.Pharm Res 24:1014–1025.

Cheng Z, Liu H, Yu N, Wang F, An G, Xu Y, Liu Q, Guan CB, and Ayrton A (2012)Hydrophilic anti-migraine triptans are substrates for OATP1A2, a transporterexpressed at human blood-brain barrier. Xenobiotica 42:880–890.

Cisternino S, Chapy H, André P, Smirnova M, Debray M, and Scherrmann JM (2013)Coexistence of passive and proton antiporter-mediated processes in nicotinetransport at the mouse blood-brain barrier. AAPS J 15:299–307.

Criado A, Barford J, Parker F, Bate S, and Whelan G (2003) Use of cyanoacrylate gelas a substitute for dental cement in intracerebroventricular cannulations in rats.Contemp Top Lab Anim Sci 42:13–16.

de Lange EC (2013) The mastermind approach to CNS drug therapy: translationalprediction of human brain distribution, target site kinetics, and therapeutic effects.Fluids Barriers CNS 10:12 10.1186/20458118-10-12.

Dickens D, Owen A, Alfirevic A, Giannoudis A, Davies A, Weksler B, Romero IA,Couraud PO, and Pirmohamed M (2012) Lamotrigine is a substrate for OCT1 inbrain endothelial cells. Biochem Pharmacol 83:805–814.

Doran A, Obach RS, Smith BJ, Hosea NA, Becker S, Callegari E, Chen C, Chen X,Choo E, and Cianfrogna J, et al. (2005) The impact of P-glycoprotein on the dis-position of drugs targeted for indications of the central nervous system: evaluationusing the MDR1A/1B knockout mouse model. Drug Metab Dispos 33:165–174.

Doran AC, Osgood SM, Mancuso JY, and Shaffer CL (2012) An evaluation of usingrat-derived single-dose neuropharmacokinetic parameters to project accuratelylarge animal unbound brain drug concentrations. Drug Metab Dispos 40:2162–2173.

Ertl P, Rohde B, and Selzer P (2000) Fast calculation of molecular polar surface areaas a sum of fragment-based contributions and its application to the prediction ofdrug transport properties. J Med Chem 43:3714–3717.

Fridén M, Gupta A, Antonsson M, Bredberg U, and Hammarlund-Udenaes M (2007)In vitro methods for estimating unbound drug concentrations in the brain in-terstitial and intracellular fluids. Drug Metab Dispos 35:1711–1719.

Griffiths R, Lewis A, and Jeffrey P (1996) Models of drug absorption in situ and inconscious animals. Pharm Biotechnol 8:67–84.

Gupta A, Chatelain P, Massingham R, Jonsson EN, and Hammarlund-Udenaes M(2006) Brain distribution of cetirizine enantiomers: comparison of three differenttissue-to-plasma partition coefficients: K(p), K(p,u), and K(p,uu). Drug MetabDispos 34:318–323.

Hammarlund-Udenaes M, Fridén M, Syvänen S, and Gupta A (2008) On the rate andextent of drug delivery to the brain. Pharm Res 25:1737–1750.

Henning U, Löffler S, Krieger K, and Klimke A (2002) Uptake of clozapine into HL-60promyelocytic leukaemia cells. Pharmacopsychiatry 35:90–95.

Jeffrey P and Summerfield SG (2007) Challenges for blood-brain barrier (BBB)screening. Xenobiotica 37:1135–1151.

Kalvass JC and Maurer TS (2002) Influence of nonspecific brain and plasma bindingon CNS exposure: implications for rational drug discovery. Biopharm Drug Dispos23:327–338.

Kalvass JC, Polli JW, Bourdet DL, Feng B, Huang SM, Liu X, Smith QR, ZhangLK, and Zamek-Gliszczynski MJ; International Transporter Consortium(2013) Why clinical modulation of efflux transport at the human blood-brainbarrier is unlikely: the ITC evidence-based position. Clin Pharmacol Ther 94:80–94.

Kelly RA and Smith TW (1992) Recognition and management of digitalis toxicity. AmJ Cardiol 69:108G–118G, 118G–119G.

Kielbasa W, Kalvass JC, and Stratford R (2009) Microdialysis evaluation of atom-oxetine brain penetration and central nervous system pharmacokinetics in rats.Drug Metab Dispos 37:137–142.

Kim MH, Maeng HJ, Yu KH, Lee KR, Tsuruo T, Kim DD, Shim CK, and Chung SJ(2010) Evidence of carrier-mediated transport in the penetration of donepezil intothe rat brain. J Pharm Sci 99:1548–1566.

Kitamura A, Higuchi K, Okura T, and Deguchi Y (2014) Transport characteristics oftramadol in the blood-brain barrier. J Pharm Sci 103:3335–3341.

Large CH, Kalinichev M, Lucas A, Carignani C, Bradford A, Garbati N, Sartori I,Austin NE, Ruffo A, and Jones DN, et al. (2009) The relationship between sodiumchannel inhibition and anticonvulsant activity in a model of generalised seizure inthe rat. Epilepsy Res 85:96–106.

Liu X and Chen C (2015) Free Drug Hypothesis for CNS Drug Candidates, in Blood-Brain Barrier in Drug Discovery: Optimizing Brain Exposure of CNS Drugs andMinimizing Brain Side Effects for Peripheral Drugs (Di L and Kerns EH eds) pp42–65, John Wiley & Sons, Inc, Hoboken, NJ.

Liu X, Chen C, and Smith BJ (2008) Progress in brain penetration evaluation in drugdiscovery and development. J Pharmacol Exp Ther 325:349–356.

Loryan I, Sinha V, Mackie C, Van Peer A, Drinkenburg W, Vermeulen A, Morrison D,Monshouwer M, Heald D, and Hammarlund-Udenaes M (2014) Mechanistic un-derstanding of brain drug disposition to optimize the selection of potential neu-rotherapeutics in drug discovery. Pharm Res 31:2203–2219.

Mehta DC, Short JL, and Nicolazzo JA (2013) Memantine transport across the mouseblood-brain barrier is mediated by a cationic influx H1 antiporter. Mol Pharm 10:4491–4498.

Mulac D, Hüwel S, Galla HJ, and Humpf HU (2012) Permeability of ergot alkaloidsacross the blood-brain barrier in vitro and influence on the barrier integrity. MolNutr Food Res 56:475–485.

Nagy CF, Kumar D, Cullen EI, Bolton WK, Marbury TC, Gutierrez MJ, Hutman HW,and Pratt RD (2004) Steady-state pharmacokinetics and safety of donepezil HCl insubjects with moderately impaired renal function. Br J Clin Pharmacol 58 (Suppl1):18–24.

Noé B, Hagenbuch B, Stieger B, and Meier PJ (1997) Isolation of a multispecificorganic anion and cardiac glycoside transporter from rat brain. Proc Natl Acad SciUSA 94:10346–10350.

Okamura N, Funaki Y, Tashiro M, Kato M, Ishikawa Y, Maruyama M, Ishikawa H,Meguro K, Iwata R, and Yanai K (2008) In vivo visualization of donepezil bindingin the brain of patients with Alzheimer’s disease. Br J Clin Pharmacol 65:472–479.

Pellegrino LJ, Pellegrino AS, and Cushman AJ (1979) A Stereotaxic Atlas of the RatBrain, 2nd ed, Plenum Press, New York.

Raub TJ (2006) P-glycoprotein recognition of substrates and circumvention throughrational drug design. Mol Pharm 3:3–25.

Reichel A (2015) Pharmacokinetics of CNS Penetration, in Blood-Brain Barrier inDrug Discovery: Optimizing Brain Exposure of CNS Drugs and Minimizing BrainSide Effects for Peripheral Drugs (Di L and Kerns EH eds) pp 7–41, John Wiley &Sons, Inc, Hoboken, NJ.

Rochat B, Baumann P, and Audus KL (1999) Transport mechanisms for the anti-depressant citalopram in brain microvessel endothelium. Brain Res 831:229–236.

Su TZ, Lunney E, Campbell G, and Oxender DL (1995) Transport of gabapentin, agamma-amino acid drug, by system l alpha-amino acid transporters: a comparativestudy in astrocytes, synaptosomes, and CHO cells. J Neurochem 64:2125–2131.

Summerfield SG and Dong KC (2013) In vitro, in vivo and in silico models of drugdistribution into the brain. J Pharmacokinet Pharmacodyn 40:301–314.

Summerfield SG, Lucas AJ, Porter RA, Jeffrey P, Gunn RN, Read KR, Stevens AJ,Metcalf AC, Osuna MC, and Kilford PJ, et al. (2008) Toward an improved pre-diction of human in vivo brain penetration. Xenobiotica 38:1518–1535.

Summerfield SG, Read K, Begley DJ, Obradovic T, Hidalgo IJ, Coggon S, Lewis AV,Porter RA, and Jeffrey P (2007) Central nervous system drug disposition: the re-lationship between in situ brain permeability and brain free fraction. J PharmacolExp Ther 322:205–213.

Summerfield SG, Stevens AJ, Cutler L, del Carmen Osuna M, Hammond B, Tang SP,Hersey A, Spalding DJ, and Jeffrey P (2006) Improving the in vitro prediction ofin vivo central nervous system penetration: integrating permeability,P-glycoprotein efflux, and free fractions in blood and brain. J Pharmacol ExpTher 316:1282–1290.

304 Summerfield et al.

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from

Sun JJ, Xie L, and Liu XD (2006) Transport of carbamazepine and drug interactionsat blood-brain barrier. Acta Pharmacol Sin 27:249–253.

Sung JH, Yu KH, Park JS, Tsuruo T, Kim DD, Shim CK, and Chung SJ (2005)Saturable distribution of tacrine into the striatal extracellular fluid of the rat:evidence of involvement of multiple organic cation transporters in the transport.Drug Metab Dispos 33:440–448.

Taub ME, Mease K, Sane RS, Watson CA, Chen L, Ellens H, Hirakawa B, Reyner EL,Jani M, and Lee CA (2011) Digoxin is not a substrate for organic anion-transporting polypeptide transporters OATP1A2, OATP1B1, OATP1B3, andOATP2B1 but is a substrate for a sodium-dependent transporter expressed inHEK293 cells. Drug Metab Dispos 39:2093–2102.

Tweedie D, Polli JW, Berglund EG, Huang SM, Zhang L, Poirier A, Chu X, and FengB; International Transporter Consortium (2013) Transporter studies in drug de-velopment: experience to date and follow-up on decision trees from the In-ternational Transporter Consortium. Clin Pharmacol Ther 94:113–125.

Uhr M, Grauer MT, Yassouridis A, and Ebinger M (2007) Blood-brain barrier pen-etration and pharmacokinetics of amitriptyline and its metabolites inp-glycoprotein (abcb1ab) knock-out mice and controls. J Psychiatr Res 41:179–188.

Uhr M, Steckler T, Yassouridis A, and Holsboer F (2000) Penetration of amitripty-line, but not of fluoxetine, into brain is enhanced in mice with blood-brain barrier

deficiency due to mdr1a P-glycoprotein gene disruption.Neuropsychopharmacology22:380–387.

Vautier S, Milane A, Fernandez C, Buyse M, Chacun H, and Farinotti R (2008)Interactions between antiparkinsonian drugs and ABCB1/P-glycoprotein atthe blood-brain barrier in a rat brain endothelial cell model. Neurosci Lett 442:19–23.

Watson J, Wright S, Lucas A, Clarke KL, Viggers J, Cheetham S, Jeffrey P, Porter R,and Read KD (2009) Receptor occupancy and brain free fraction. Drug MetabDispos 37:753–760.

Wright EM (1972) Active transport of lysergic acid diethylamide. Nature 240:53–54.Zhang Y, Liu H, Summerfield SG, Luscombe CN, and Sahi J (2016) Integratingin silico and in vitro approaches to predict drug accessibility to the central nervoussystem. Mol Pharm 13:1540–1550. DOI: 10.1021/acs.molpharmaceut.6b00031.

Address correspondence to: Dr. Scott G. Summerfield, Drug Metabolismand Pharmacokinetics Department, GlaxoSmithKline, Park Road, Ware,Hertfordshire, United Kingdom, SG12 0DP. E-mail: Scott.G.Summerfield@GSK.com

Uptake of CNS Drugs and Candidates across the BBB 305

at ASPE

T Journals on D

ecember 16, 2020

jpet.aspetjournals.orgD

ownloaded from