Application of an in Vitro Blood–Brain Barrier Model in the … · 2019-11-14 · Application of an in Vitro Blood−Brain Barrier Model in the Selection of Experimental Drug Candidates

Application of an in Vitro Blood−Brain Barrier Model in theSelection of Experimental Drug Candidates for the Treatment ofHuntington’s DiseaseAnnalise Di Marco,*,† Odalys Gonzalez Paz,† Ivan Fini,† Domenico Vignone,† Antonella Cellucci,†

Maria Rosaria Battista,† Giulio Auciello,† Laura Orsatti,‡ Matteo Zini,‡ Edith Monteagudo,‡

Vinod Khetarpal,§ Mark Rose,§ Celia Dominguez,§ Todd Herbst,§ Leticia Toledo-Sherman,§

Vincenzo Summa,∥ and Ignacio Munoz-Sanjuan*,§

†In Vitro Pharmacology Unit, ‡Preclinical Research Unit, and ∥Department Chemistry, IRBM SpA, Pomezia, Rome 00071, Italy§CHDI Management, CHDI Foundation, Center Drive Los Angeles 6080, California, United States

ABSTRACT: Huntington’s disease (HD) is a neurodegenerative disease caused by polyglutamine expansion in the huntingtinprotein. For drug candidates targeting HD, the ability to cross the blood−brain barrier (BBB) and reach the site of action in thecentral nervous system (CNS) is crucial for achieving pharmacological activity. To assess the permeability of selectedcompounds across the BBB, we utilized a two-dimensional model composed of primary porcine brain endothelial cells and ratastrocytes. Our objective was to use this in vitro model to rank and prioritize compounds for in vivo pharmacokinetic and brainpenetration studies. The model was first characterized using a set of validation markers chosen based on their functionalimportance at the BBB. It was shown to fulfill the major BBB characteristics, including functional tight junctions, hightransendothelial electrical resistance, expression, and activity of influx and efflux transporters. The in vitro permeability of 54structurally diverse known compounds was determined and shown to have a good correlation with the in situ brain perfusiondata in rodents. We used this model to investigate the BBB permeability of a series of new HD compounds from differentchemical classes, and we found a good correlation with in vivo brain permeation, demonstrating the usefulness of the in vitromodel for optimizing CNS drug properties and for guiding the selection of lead compounds in a drug discovery setting.

KEYWORDS: blood−brain barrier, CNS, transport, Huntington’s disease, permeability, efflux transporters, brain penetration

■ INTRODUCTION

Huntington’s disease (HD) is a monogenic disorderencompassing a variable phenotype with progressive cognitive,psychiatric, and movement disorders, caused by an expandedCAG trinucleotide repeat (of variable length) in huntingtin(HTT), the gene that encodes the protein HTT. The disease isassociated with high morbidity and mortality rates, and thereare a number of pharmacological interventions that are beingevaluated. Currently, HD therapeutics are limited tosymptomatic treatments, and there are no treatment optionswith proven safety and efficacy to slow down diseaseprogression or enhance survival rate. Because of the lack ofapproved drugs, the medical need of HD is still very high.1

One of the major challenges in developing therapeuticagents for neurological diseases is the difficulty in deliveringthem to the brain. The complexity of the brain and the

requirement of the drug to cross the blood−brain barrier(BBB) are the main reasons for high attrition rates in thedevelopment of central nervous system (CNS)-targetingmedicines.Early assessment of the physiochemical properties required

for drugs to cross the BBB is extremely important to reduce thetime spent on unproductive or redundant developmentactivities. In addition, understanding the nature of thepermeability and transport mechanisms at the BBB is a centralpart of CNS drug development. Because of the high complexityof the brain, in vivo screening and data analysis are challenging;

Received: January 10, 2019Revised: March 6, 2019Accepted: March 27, 2019Published: March 27, 2019

Article

pubs.acs.org/molecularpharmaceuticsCite This: Mol. Pharmaceutics 2019, 16, 2069−2082

© 2019 American Chemical Society 2069 DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

Dow

nloa

ded

via

CN

R o

n M

ay 6

, 201

9 at

16:

20:3

0 (U

TC

).

See

http

s://p

ubs.

acs.

org/

shar

ingg

uide

lines

for

opt

ions

on

how

to le

gitim

atel

y sh

are

publ

ishe

d ar

ticle

s.

thus, initial BBB-related studies are frequently performed usingin vitro models. In vitro BBB systems provide a less resource-intensive alternative compared to dosing animals withsubsequent measurement of the compound brain concen-trations. In addition, when a decision is made to perform an invivo pharmacokinetic study, in vitro BBB experiments canoften help to interpret animal brain exposure at the cellular andmolecular levels.Several in vitro model systems to assess BBB permeation

have been developed using primary brain endothelial cells fromdifferent species.2,3 The best in vitro BBB models are based onthe co-culture of endothelial cells and glial cells, astrocytes,and/or brain pericytes, which participate in the acquisition ofBBB characteristics.4,5 Co-cultured models using porcine brainendothelial cells (PBECs) have been shown to be particularlyuseful for characterizing CNS drugs as this cell modelpossesses a restrictive paracellular pathway, functionalexpression of transporters, and ease of culture to supportdrug screening.6

The two-dimensional BBB model described in this work wasformed by primary PBECs and rat astrocytes. This model wasevaluated for a number of characteristics including endothelialcell organization and purity, transendothelial electricalresistance (TEER), paracellular marker permeability, qualita-tive and quantitative expression of tight junction proteins, andfunctionality of efflux transporters such as P-glycoprotein. Ourobjective was to establish and validate an in vitro BBB modelto be used for permeability evaluation of investigationalcompounds and to help prioritize compounds for in vivostudies. We assessed the in vitro permeability of therapeuticcompounds from specific HD screening programs andcompared our findings with the in vivo brain partition data.

■ EXPERIMENTAL SECTIONMaterials and Reagents. Minimum essential medium

(MEM), Dulbecco’s modified Eagle’s medium (DMEM),Medium 199 (M199), Ham’s F-12, L-15 Leibovitz media,and Lucifer yellow were purchased from Life Technologies(Thermo Fisher Scientific, Monza Italy); endothelial growthbasal medium EBM-2 with EGM-2 SingleQuot Kit Suppl. andgrowth factors were purchased from Lonza (Basel, Switzer-land). Sterile plasma-derived bovine serum (PDBS) waspurchased from First Link (Birmingham, UK); collagenasetype II from Worthington (Worthington Biochemical Corp.,NJ); fetal bovine serum (FBS), penicillin/streptomycin,glutamine, HEPES, fungizone, deoxyribonuclease I, andHoechst 33342 were from Invitrogen (Thermo FisherScientific, Monza IT); bovine serum albumin (BSA) fractionV was from Roche (Basel, Switzerland); and nylon meshes (60

and 150 μm pore size) were from Plastok Associates, UK. Rattail collagen type I, human fibronectin, cell strainer (70 μmnylon), tissue culture plastic flasks T75, and 150 mm TC-treated cell culture dishes were all purchased from Corning(Sigma-Aldrich, Milan Italy). [14C]taurocholic acid, [14C]-phenytoin, [14C]dopamine, [14C]phenylalanine, [3H]naloxone,[14C]sucrose, [3H]propranolol, [14C]caffeine, [14C]-tryptophan, [14C]leucine, [14C]mannitol, [3H]kynurenine,[3H]digoxin, [14C]arginine, [3H]verapamil, [3H]vinblastine,[3H]daunomycin, [3H]testosterone, [14C]inulin, [3H]-enkephalin (2-D-penicillamine, 5-D-penicillamine), [3H]-methyl-glucose, and Microscint 20 scintillation liquid werepurchased from PerkinElmer (Milan, IT). [14C]Theophylline,[3H]quinidine, [3H]methotrexate, [3H]gabapentin, [3H]-metoprolol, [3H]amprenavir, [3H]ritonavir, [14C]taxol, [3H]-estrone-3-sulfate, [3H]atenolol, [3H]L-DOPA (L-3,4-dihydrox-yphenylalanine), and [14C]-docosahexaenoic acid lysophos-phatidylcholine (LPC[14C]-DHA) were purchased fromBIOTREND Chemikalien GmbH (Koln, Germany); and[3H]prazosin and imatinib from Amersham, Life Science(UK).Fibronectin solution from human fibroblasts, puromycin

dihydrochloride, heparin sodium salt from porcine intestinalmucosa, hydrocortisone, fluorescein isothiocyanate (FITC)-labeled dextrans of various polymerization sizes (FITC-dextran4, FITC-dextran 40, and FITC-dextran 70), dimethyl sulfoxide(DMSO), endothelial cell growth supplement from bovineneural tissue, trypsin−ethylenediaminetetraacetic acid (EDTA)solution 1×, poly-D-lysine hydrobromide, GF120918, KO-143,MK-571, altanserin hydrochloride hydrate, biotin, rhodamine123, 3-hydroxy-kynurenine, anthralinic acid, kynurenic acid,pantothenic acid, parthenolide, triton X100, and WheatonDounce tissue grinder were purchased from Sigma-Aldrich(Milan Italy). Alvespimycin (17-DMAG), masitinib, nilotinib,and NVP-HSP990 were purchased from SelleckChem(Munich, Germany). Bafetinib was obtained from CaymanChemical (Michigan, USA); dasatinib from AstaTech (Boston,USA); laquinimod from Akos (Steinen, Germany); and 8%paraformaldehyde aqueous solution was from ElectronMicroscopy Sciences (Hatfield, PA). The rat tail collagenused for insert coating was prepared according to Strom andMichalopoulos.7 CHDI compounds, denoted cp, weresynthesized as previously described.8

Tissue culture-treated multiwell plates and transwell filterinserts (1.1 cm2 growth area, 0.4 μm pore size, transparentpolyester) were obtained from Corning Costar (Sigma-Aldrich,Milan Italy); transwell 24-well plate (0.7 cm2 growth area, 0.4μm pore size, polycarbonate membrane) was obtained fromMillipore (Merck S.p.A., Milan Italy). Optiplate-96 and white

Table 1. Antibodies Used in This Study for Immunofluorescence

target epitope species type conjugation RRIDa vendor Cat nb dilution

anti-occludin rabbit polyclonal AB_2533977 Thermo Fisher 71−1500 1:25anti-claudin-5 rabbit polyclonal AB_2533157 Thermo Fisher 34−1600 1:25anti-ZO-1 tight junction protein mouse monoclonal AB_2533147 Thermo Fisher 33−9100 1:100anti-Von Willebrand factor rabbit polyclonal AB_305689 Abcam ab6994 1:400anti-P glycoprotein [4E3] mouse monoclonal AB_297069 Abcam ab10333 1:10anti-Mfsd2A rabbit polyclonal AB_10711504 Abcam ab105399 1:50anti-rabbit IgG (H + L) goat polyclonal Alexa Fluor 488 AB_2576217 Thermo Fisher A-11034 1:3000anti-rabbit IgG (H + L) cross-adsorbed goat polyclonal Alexa Fluor 568 AB_143157 Thermo Fisher A-11011 1:3000anti-mouse IgG (H + L) cross-adsorbed rabbit polyclonal Alexa Fluor 594 AB_2534109 Thermo Fisher A-11062 1:3000

aResearch resource identifier.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2070

Opaque 96-well microplate were purchased from PerkinElmer(Milan Italy). Corning 96-well polystyrene plates wereobtained from Corning Costar (Sigma-Aldrich, Milan Italy).The antibodies used for immunochemistry studies are listed inTable 1.Isolation of Brain Microvascular Endothelial Cells.

Brain microvascular endothelial cells were isolated from freshporcine brains following the methods of Patabendige et al.3

and Skinner et al.9 with some modifications. Briefly, the cellswere isolated from porcine brains obtained from localslaughterhouses and transported on ice in an L15 mediumcontaining penicillin (100 U/mL) and streptomycin (100 μg/mL). The meninges and white matter were removed, and thegray matter was collected in an MEM containing 10% FBS,penicillin (100 U/mL), streptomycin (100 μg/mL), 2 mMglutamine, and 25 mM HEPES (MEM-H). Then, the tissuesuspension was homogenized in a glass−glass WheatonDounce tissue grinder and sequentially filtered, first througha 150 μm nylon mesh and then through a 60 μm nylon mesh.The microvessel fragments were digested with collagenase typeII, for 1 h at 37 °C, in an M199 medium containing 10% FBS.After centrifugation, for 5 min at 240g at 4 °C, the mixture wasresuspended in 20% BSA in MEM-H and centrifuged at 1000gat 4 °C to remove the supernatant containing myelin. Thepellet was resuspended in MEM-H, and the suspension wasfiltered through a 70 μm nylon cell strainer. Aftercentrifugation at 240g at 4 °C for 5 min, the pellet wasresuspended in a complete endothelial culture medium(ECM), composed of EBM-2 supplemented with 10%PDBS, 2 mM glutamine, 100 U/mL penicillin, 100 μg/mLstreptomycin, 0.25 μg/mL fungizone, 4 μg/mL puromycin,125 μg/mL heparin, and the SingleQuots kit (except heparinand serum). The capillary fragments were plated in T75 flaskscoated with 100 μg/mL rat tail collagen and 7.5 μg/mLfibronectin. We routinely obtain sufficient material for two T75

flasks from one brain. The cultures were maintained for 4 daysin ECM at 37 °C and 5% CO2 with medium change every day.

Culture on Transwell Inserts and TEER Measurement.The endothelial cells were detached using trypsin−EDTA andresuspended in an endothelial differentiation medium (EDM)consisting of DMEM/F12 1:1 containing 10% PDBS, 125 μg/mL heparin, 100 U/mL penicillin, 100 μg/mL streptomycin, 2mM glutamine, and 100 μg/mL of endothelial cell growthsupplement. The cells were plated at 200 000/cm2 on rat tailcollagen (100 μg/mL)/fibronectin (7.5 μg/mL)-coated trans-well filters. In parallel, rat astrocytes isolated from the cerebralcortex of newborn pups (1−2 days old, Sprague-Dawley (SD-CD) rats from Charles River), as described previously byAbbott et al.,10 were plated on poly-D-lysine-coated 12- or 24-well plates at 150 000 cells/cm2 in DMEM (high glucose)supplemented with 10% FBS, 2 mM glutamine, 100 U/mLpenicillin, and 100 μg/mL streptomycin. Three days later, theendothelial cells were put in co-culture with rat astrocytes inEDM containing 550 nM of hydrocortisone. The co-cultureswere monitored by measuring the TEER every day using STX2chopstick electrodes connected to an EVOM2 voltmeter(World Precision Instruments, USA). The TEER of the cellmonolayer was obtained by subtracting the resistance (Ω) ofthe coated blank filter insert (without cells) from the resistancemeasured across the insert with the cell monolayer. Theresulting value was multiplied by the surface area of the filterinsert to express the result as Ω cm2. Gaillard and De Boer11

recommended considering 200 Ω cm2 as a threshold, but asTEER is a relative measurement, we set a quality control (QC)benchmark at 400 Ω cm2 for permeability assays.

Immunocytochemistry. The cells in transwell inserts(polyester membrane Transwell-Clear) were washed withphosphate-buffered saline (PBS) and fixed in 4% paraformal-dehyde for 20 min at room temperature. The cells werepermeabilized by washing five times (2 min each) with PBS/

Table 2. Genes Evaluated in This Study and the TaqMan Assays Used for mRNA Quantification

symbol alias gene name ID (NCBI) assay ID (Thermo Fisher)

ABCB1 P-gp, MDR1 ABC, subfamily B, member 1 396910 Ss03373434_m1ABCC1 MRP1 for human ABC, subfamily C, member 1 733619 Ss03376986_u1ABCC2 MRP2 ABC, subfamily C, member 2 397535 customizeda

ABCG2 BCRP ABC, subfamily G, member 2 397073 Ss03393456_u1AGER RAGE advanced glycosylation end-product specific receptor 396591 Ss03390850_g1BSG CD147 basigin 100141312 Ss03387611_u1FABP5 E-FABP fatty acid binding protein 5 574074 Ss03392151_m1FCGRT FcRn Fc fragment of IgG receptor and transporter alpha 397399 Ss03383313_u1GAPDH GAPD GAPDH 396823 Hs02758991_g1IGF2R M6P/IGF2R insulin-like growth factor receptor II 397214 Ss03373426_m1INSR insulin receptor 396755 Ss03375405_u1LDLR LDL receptor 396801 Ss03373254_u1SCARB1 scavenger receptor class B, member 1 397018 Ss03391101_m1SLC16A1 MCT1 SLC family 16, member 1 100127159 Ss03374095_m1SLC22A5 OCTN2 organic cation/carnitine transporter 2 100520422 Ss03377456_u1SLC2A1 GLUT1 SLC family 2, member 1 102164419 Ss03374747_s1SLC7A5 LAT1, CD98 SLC family 7, member 5 8140 Hs01001189_m1SLC7A8 LAT2 SLC family 7, member 8 23428 Hs00794796_m1SLCO1A2 OATP1A2 SLC organic anion transporter family, member 1A2 397534 Ss03375623_u1TFRC TrFr, CD71 transferrin receptor 397062 Ss03391237_m1TNFRSF1A TNFR1 tumor necrosis factor receptor superfamily, member 1A 397020 Ss03391126_g1TNFRSF1B TNFR2 tumor necrosis factor receptor superfamily, member 1B 100037306 Ss03385518_u1

aForward primer: CACTGTGGGCTTTGTTCTGT; reverse primer: TTTCTGACGTCATCCTCACC; 5′Fam-3′MGB-labeled probe:CGCACTCAATATCACACAAACCCTGA.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2071

0.1% Triton X-100 and blocked in PBS containing 1% BSA(blocking buffer) for 2 h at room temperature. Primaryantibodies, diluted as reported in Table 1, were added in theblocking buffer for 2 h at room temperature. The cells werethen washed four times and incubated in the blocking bufferwith the secondary antibody, diluted as recommended by themanufacturer, and the nuclear stain Hoechst 33342 (2 μM) for1 h at room temperature. Cell images were acquired with anINCell Analyzer 2000 (GE Healthcare).mRNA Extraction and Quantification by Real-Time

qRT-PCR (TaqMan). Total RNA was extracted after cell lysisin a 96-well format from duplicate samples using an RNAextraction kit (Macherey-Nagel, Duren, Germany) on aMicrolab STARlet (Hamilton, RENO, NV), according to themanufacturer’s protocol. Quantitative analysis of specificmRNA expression was performed by real-time quantitativereverse transcription polymerase chain reaction (qRT-PCR),by subjecting the resulting cDNA to PCR amplification using384-well optical reaction plates in the QuantStudio 12k FlexReal-Time PCR System (Applied Biosystem). The thermocy-cling conditions were initiated at 48 °C for 15 min with anenzyme activation step of 95 °C for 10 min, followed by 40PCR cycles of denaturation at 95 °C for 15 s and anneal/extension at 60 °C for 1 min. Single-gene real-time PCRprimers and probes were purchased from Applied Biosystem as“TaqMan gene expression assay” for the porcine genes listed inTable 2. The housekeeping gene glyceraldehyde-3-phosphatedehydrogenase (GAPDH) was used for normalization. Therelative differences in mRNA expression were determined bythe QuantStudio 12k Flex software (version v.1.2.3) based onthe threshold cycle (Ct) data of target gene versus endogenouscontrol gene for each reaction.Permeability Assay. Permeability studies were performed

using cell monolayers with the TEER value greater than 400 Ωcm2. All permeability studies were performed on day 2−3when the highest TEER values were achieved. The transportassays were usually performed during a short time frame (1 hor less). In addition, kinetic experiments with selectedcompounds with a known transport mechanism (influx, efflux)were also performed. The permeability values were maintainedover time, indicating the functionality of the transportersduring the assay.Prior to the experiment, carried out at pH 7.4, 37 °C, and

5% CO2, the culture medium was removed from the insertsand replaced by prewarmed Hanks’ balanced salt solution(HBSS) containing 20 mM HEPES, pH 7.4, and 0.1% BSA.The inserts were removed from the co-culture plates andplaced in receiver plates, without astrocytes, containing thesame buffer. The test compound was added together with theintegrity marker (radiolabeled sucrose or FITC-dextran 40)used as a second parameter to ascertain the quality of the cellmonolayer. For each compound, inserts with endothelial cellswere used in triplicates and cell-free, collagen/fibronectin-coated inserts were used in duplicates. The plates were placedin a cell incubator for 60 min, and the inserts were then movedto another plate containing prewarmed incubation buffer tominimize the back diffusion of the compound to the uppercompartment. At the end of each experiment, the samples werecollected, and the amounts of radiotracers and fluorescenttracers were determined by liquid scintillation (Top Count-NXT, Microplate Scintillation and Luminescence counter fromPerkinElmer) and fluorescence spectrophotometry (SAFIRETECAN, Microplate Fluorescence reader), respectively, where-

as for all other compounds the concentration was determinedby liquid chromatography−tandem mass spectrometry (LC-MS/MS). Mass balance was determined considering theamount of compounds recovered in the donor and receiverchambers at the end of the assay relative to the amount addedto the donor chamber at time 0. In efflux transport assays,before the addition of the compounds, the filters werepreincubated for 30 min at 37 °C with or without 2 μMGF120918, 2 μM Ko143, or 10 μM MK571 which are theinhibitors of Pgp, breast cancer resistance proteins (BCRPs),or various multidrug resistance-associated proteins (MRPs),respectively.The permeability coefficient (P) was calculated according to

Pardridge et al.;12 PS (permeability−surface area product, incm3/s) was divided by the filter surface area A (in cm2).Therefore, the permeability coefficient P is expressed in cm/s,following eq 1

PV M

A M tcm/s d r

d[ ] =

× Δ× × Δ (1)

where Vd is the volume in the donor compartment in cm3;ΔMr is the total amount of compound in the receivercompartment after t seconds; Md is the donor amount(added at time 0); Δt is the incubation time measured inseconds; and A is the filter area in cm2. The contribution ofboth filter and substrate is subtracted from the totalpermeability using eq 2

P P P1 1 1

e t f= −

(2)

where Pt is the permeability coefficient of the total system; Pf isthe permeability coefficient of the cell-free coated filter; and Peis the permeability coefficient of the endothelial cell layer.In efflux transport assays, the efflux ratio was calculated

dividing the permeability value in the basal to apical (B−A)direction by the permeability value in the apical to basal (A−B) direction.

Bioanalysis. For LC−MS/MS analysis, aliquots from thereceiver and donor chambers were diluted with an equalvolume of quenching solution for protein precipitation [0.1%formic acid (FA) in acetonitrile (ACN) containing an internalstandard (IS)]. Routinely used ISs include ketoconazole anddextromethorphan depending on the analytical conditions, orother ISs when appropriate. The samples were then vortexedand centrifuged at 4000g for 10 min at 4 °C. A volume of 100μL of the supernatant was evaporated under nitrogen at 30 °Cand reconstituted with an appropriate solution according tothe developed LC method. In the reconstitution solvent, 10ng/mL warfarin and 50 ng/mL labetalol were included asvolumetric IS for injection evaluation. In parallel, a standardcurve for each test compound was generated, and finally, thesamples were injected into the LC−MS/MS system foranalysis.Single 12-point calibration curve for each test compound

was prepared in HBSS−HEPES−BSA 0.1% buffer. Themaximum percentage of final DMSO in each calibrationstandard (CS) was 0.1%. A 100 μL of CS solution was addedwith an equal volume of quenching solution and processed asthe samples. One set of each CS was included at the front endof the run. An analyte response for the transition with thehighest signal-to-noise ratio (S/N) was normalized to the ISpeak area, and calibration was achieved by weighted linear least

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2072

Table 3. Summary of the Permeability Values in the In Vitro BBB Model (Pin vitro) of the 54 Compounds with DifferentTransport Mechanisms and Their Relative Calculated In Vivo Permeability Coefficients (Pin vivo)

compounds BBB transport mechanism Pin vitro (×10−6 cm/s) Pin vivo (×10

−6 cm/s)

altanserina passive diffusion/active efflux (Pgp) 33.0 ± 6.2alvespimycina passive diffusion 3.1 ± 1.2amprenavira passive diffusion/active efflux (Pgp) 22.3 ± 4.3anthranilic acid passive diffusion 19.3 ± 1.9 8.014

arginine active influx (y + L) 5.2 ± 3.8 23.315

atenolol passive diffusion 1.3 ± 0.6 0.216

bafetiniba passive diffusion/active efflux (Pgp) 16.8 ± 3.2biotin active influx (SMVT) 3.6 ± 1.0 0.717

caffeine passive diffusion/active influx 40.8 ± 11.2 66.716

dasatiniba passive diffusion/active efflux (Pgp/BCRP) 7.0 ± 1.8daunomycin passive diffusion/active efflux (Pgp/MRP) 4.1 ± 2.4 13.318

digoxin passive diffusion/active efflux (Pgp) 3.7 ± 2.2 0.318

dopamine passive diffusion/active uptake 29.8 ± 17.2 53.018

bis pen enkephalin multiple mechanisms 2.7 ± 0.8 0.319

estrone-3-sulfatea passive diffusion/active efflux (MRP/BCRP) 1.7 ± 0.6FITC-dextran 4a passive diffusion 2.0 ± 1.3FITC-dextran 40a passive diffusion 0.2 ± 0.2FITC-dextran 70a passive diffusion 0.1 ± 0.1gabapentin active influx (LAT) 8.4 ± 2.4 18.720

glucose active influx (GLUT-1) 7.0 ± 3.0 13.321

3-hydroxykynurenine active influx (LAT) 1.2 ± 1.02 0.214

imatinib passive diffusion/active efflux (Pgp/BCRP) 20.6 ± 1.0 9.322

inulin passive diffusion 1.9 ± 1.1 0.0323

kynurenine active influx (LAT) 3.0 ± 2.7 0.514

kynurenic acid passive diffusion 2.0 ± 1.1 0.714

laquinimoda passive diffusion 38.8 ± 4.2L-DOPAa passive diffusion/active influx (LAT) 5.9 ± 1.8leucine active influx (LAT) 7.0 ± 3.0 1.324

LPC-DHAa MFSD2A-mediated transport 4.3 ± 0.7lucifer yellowa passive diffusion 1.7 ± 0.8mannitol passive diffusion 1.2 ± 0.8 0.125

masitiniba passive diffusion/active efflux (Pgp) 20.4 ± 10.6memantinea passive diffusion/OCT-mediated transport (influx/efflux) 50.5 ± 8.9methotrexate passive diffusion/active efflux (MRP) 5.7 ± 1.8 0.818

metoprolol passive diffusion 23.6 ± 6.6 20.16

naloxone passive diffusion/active efflux (MRP) 39.9 ± 9.2 33.326

nilotiniba passive diffusion/active efflux (Pgp) 7.4 ± 0.9NVP-HSP990a passive diffusion 33.9 ± 6.1pantothenic acid active influx (SMVT) 4.2 ± 1.3 1.227

parthenolidea passive diffusion 39.0 ± 5.0phenylalanine active influx (LAT) 13.0 ± 4.0 1.228

phenytoin passive diffusion/active efflux (MRP) 42.3 ± 8.3 42.118

prazosin passive diffusion/active efflux (BCRP) 13.4 ± 1.5 18.829

propranolol passive diffusion 41.1 ± 17.3 42118

quinidine passive diffusion/active efflux (Pgp) 16.7 ± 1.0 14.630

ritonavir passive diffusion/active efflux (Pgp) 28.8 ± 11.5 17.131

rhodaminea passive diffusion/active efflux (Pgp) 0.8 ± 0.3sucrose passive diffusion 1.6 ± 0.9 0.0325

taurocholic acid active uptake/efflux 6.4 ± 3.7 0.518

taxol passive diffusion/active efflux (Pgp) 6.7 ± 2.8 0.232

testosterone passive diffusion 72.0 ± 24.1 53018

theophylline passive diffusion 5.1 ± 1.1 3.333

tryptophan active influx (LAT) 5.4 ± 2.0 13.334

verapamil passive diffusion/active efflux (Pgp) 35.7 ± 19.3 37.335

vinblastine passive diffusion/active efflux (Pgp/MRP) 1.9 ± 1.7 2.136

aNot considered for in vitro−in vivo correlation. Fukui et al. 1991;14 Doan et al. 2000;15 Avdeef 2011;16 Park and Sinko 2005;17 Liu et al. 2004;18

Egleton and Davis 1999;19 Summerfield et al. 2007;20 Pardridge 1983;21 Bihorel et al. 2006;22 Abbruscato et al. 1997;23 Chikhale et al. 1995;24

Gratton et al. 1997;25 Zhang et al. 2006;6 Suzuki et al. 2010;26 Spector et al. 1986;27 Momma et al. 1987;28 Cisternino et al. 2004;29 Chen 2002;30

Degenais et al. 2009;31 Rice et al. 2005;32 Youdim et al. 2004;33 Diksic et al. 2000;34 Zhao et al. 2009;35 Murakami et al. 2000.36

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2073

square regression (weight 1/x or 1/x2). CSs differing by morethan ±30% of the nominal value were excluded. Carryover wasassessed by injecting blank samples after samples and CS at theupper limit of quantification. The limit of quantitation was thelowest CS included in the linear range having S/N ≥ 5.The analysis was carried out on an Acquity I-Class UPLC

System (Waters, Milford) coupled to an API-4000 or API-6500 (AB Sciex, Framingham, USA) mass spectrometer. Theproper reverse-phase or normal-phase chromatographiccolumn, mobile phases, and gradient conditions were selectedto obtain the best retention and separation of the testcompound. In general, the total run time varied from 1.50 to3.00 min for 2.1 × 30 mm or 50 mm columns. The bestconditions for sample reconstitution varied from 0.1% FA inH2O or aqueous buffer/ACN or MeOH 90/10 to 0.1% FA in100% ACN, depending on the test compound solubility andMS response.The detection of the test compound and IS was performed

by multiple reaction monitoring (MRM). MRM transitionswere optimized by direct infusion of pure standard solutions.The turbo ion spray ESI source parameters were optimized foreach compound tested and set as follows: source temperature200−700 °C, curtain gas 10−50 psi, gas1 10−45 psi, gas2 11−45 psi, ion spray voltage 4000−5500 V in positive ionizationmode, declustering potential 20−100 V, entrance potential 5−15 V, collision energy 10−70 eV, and cell exit potential 5−15V. The data analysis was performed with Analyst 1.6.2.Correlation with in Vivo Data. For in vivo correlation,

data from published values were used. The publishedcerebrovascular permeability−surface area products (PS)were converted into in vivo brain permeability values (P invivo) by dividing PS by an estimated value of the surface area(S) of perfused capillaries equal to 150 cm2/g of brain13 andare reported in Table 3.In addition, information on in vivo brain penetration was

obtained for the compounds investigated as part of thediscovery programs targeted for HD at CHDI Foundation.Following a single intravenous dosing of each compound tomice, the plasma-to-brain ratio was determined at steady state.These series of compounds showed a bicompartmental plasmapharmacokinetics, with fast uptake into the brain (within 0.5h). Because of this, the brain-to-plasma concentration ratio wascalculated at each time point (six time points from 1 to 6 h)and an average value was used for the evaluation of correlationwith the in vitro permeability data obtained in the PBECmodel. Equivalent brain/plasma ratio values were also obtainedfrom AUCbrain 1−6h divided by AUCplasma 1−6h. In addition to thetotal brain-to-plasma concentration ratio (Kp), when plasmaand brain homogenate free fractions were available, unboundbrain-to-plasma concentration ratio (Kp,uu) was also calculatedand used to further assess the correlation with in vitropermeability.

■ RESULTSBarrier Characteristic, Transcriptional Profiling, and

Protein Expression. Brain endothelial cells were obtainedfrom digested porcine brain capillary fragments after 4 days inculture (Figure 1) and the generated monolayer wascharacterized by the expression of the endothelial markerVon Willebrand factor with the characteristic staining ofWeibel−Palade bodies (Figure 2).The endothelial cells were then grown on membrane inserts

and put in co-culture with astrocytes to favor blood−brain

barrier differentiation. To create a barrier, the cells are stitchedtogether through tight junctions. The immunofluorescence(Figure 3) showed the membrane localization of claudin-5,occludin, and zonula occludens-1 (ZO-1), demonstrating theformation and the proper localization of the tight junctions.The endothelial barrier was also characterized by the TEER

value that increased from about 200 Ω cm2, measured beforeco-culture (day 3 on the transwell filter, D0 co-culture), totypically 600−1000 Ω cm2 2 days after co-culture (D2 co-culture) with rat astrocytes (Figure 4A). The TEER values areindicative of the barrier tightness, and thus permeability assayswere usually performed on day 2. The TEER and permeabilityvalues were compared, and it appeared that above 400 Ω cm2,the permeability coefficient of FITC-dextran 40 becameindependent of TEER, remaining low (Figure 4B). Belowthis threshold, the lower the TEER, the higher will be thepermeability, as observed in other studies;3,11 this supported usfor using TEER values as the first QC. In fact, for permeabilitymeasurements, only transwell filters with the TEER value equalor greater than 400 Ω cm2 were utilized.The TEER and permeability values of FITC-dextran 40 were

compared, and the relationship observed in our model (Figure4B) is similar to that reported in other studies3,11 using otherparacellular markers (mannitol, sodium fluorescein, and FITC-dextran 4). The threshold of 200 Ω cm2 is generally good forpassive permeability, but high TEER also ensures goodtransporter expression and polarity required for transcellulartransport.3 To increase reproducibility, and because the cutoffof 400 Ω cm2 was experimentally observed, we define it as theQC limit.37

The co-culture of brain endothelial cells with astrocytesexhibited a good capacity in discriminating, among paracellularmarkers of different sizes, between large and small molecules(Table 4). Actually, permeability inversely correlated with bothmolecular weight and hydrodynamic radius, with mannitol, thesmallest molecule, being more permeable (1.2 ± 0.8 × 10−6

cm/s) than FITC-dextran 70 (0.1 ± 0.1 × 10−6 cm/s), thelargest marker.The BBB is not only characterized by its tightness, but also

by the expression of several transporter systems responsibleand essential for the import of nutrients, proteins, and smallmolecules. The expression of transporters is subjected tovariation upon in vitro culture of primary cells; therefore, weanalyzed the expression levels at the same time in culture as thetransport assays were performed. The mRNA levels of themajor transporters and receptors were quantified by TaqManto explore the expression profile of our PBEC/As model. Theexpression pattern was acquired at 2 days of co-culture, but

Figure 1. Flowchart of the cell culture phases before the transportexperiments in the BBB model. Capillaries from porcine brains areisolated and put in culture for four days. At subconfluence, theendothelial cells are plated onto coated transwell filters. Three dayslater, they are put in co-culture with rat astrocytes to inducedifferentiation. Permeability assays are performed on day 9 and/or 10,corresponding to D2 and/or D3 of co-culture.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2074

similar results were obtained within a 2 day variation (data notshown). The BBB membrane transporters include ATP-binding cassette (ABC) transporters and solute carrier(SLC) transporters. These two classes of transporters areresponsible for the uptake or efflux of nutrients and toxicagents. Among the ABC transporters, ABCB1 (Pgp) andABCG2 (BCRP) were the most expressed (Figure 5). ABCG2was more expressed than ABCB1, similar to that reported inhumans.38

The expression of the ABCB1 transporter was confirmed byimmunofluorescence both in the capillary fragments and inendothelial cells after 2 days in co-culture with rat astrocyteswhere it showed predominant membrane localization (Figure6). ABCC1, formerly MRP1, which is expressed in theabluminal surface of the brain capillaries39 was more expressed(Figure 5) than ABCC2 (MRP2) which localizes to theluminal side of the capillary. Among the SLC proteinsexamined, the major expression was found for SLC2A1(GLUT1), which facilitates glucose uptake by the cells.Other transporters, for organic ions (SCL22A5 for organiccations, carnitine; SLCO1A2 for organic anions), formonocarboxylates such as lactate and pyruvate (SCL16A1),or for amino acids (SLC7A5, SLC7A8), are less expressed thanthe Na+-independent glucose transporter that helps D-glucoseto cross the BBB to supply energy to the brain. Finally, theselected surface receptors were examined, in particular the

transferrin receptor and the Fc fragment of IgG receptors,whose capacity to transport their natural ligands by trans-cytosis are studied to target therapeutics to the brain, thuscrossing the BBB.40 FCGRT was more expressed than TFRC,but both were present in our model allowing transcytosisstudies. Among the receptors important for lipid brainhomeostasis, FABP5 was the most expressed in our system,with respect to the scavenger receptor class B type 1 (SRB1)and the low-density lipoprotein receptor (LDLR), which is alsostudied to cargo drugs to the CNS. We also found a strongexpression of CD147 (or basigin) which has been shown to

Figure 2. Phase contrast micrographs of digested capillary fragments 2 h after plating (A) and of endothelial monolayer 4 days after plating (B).The monolayer was formed of tightly packed, longitudinally aligned cells, typical for differentiated capillary endothelial cells. (C) Positive stainingwith the endothelial marker Von Willebrand factor of endothelial cells (4-day monolayer) before culture in transwell inserts. The granular greensignal is characteristic of the Weibel−Palade body staining of endothelial cells, and the nuclei are stained with Hoechst.

Figure 3. Immunofluorescence micrographs for claudin-5 (A),occludin (B), and ZO-1 (C) showed intense and delineated stainingof all cell−cell contacts, indicating well-organized and fully developedtight junctions (day 2 of co-culture) of PBECs. Intracellularimmunostaining (D) of the endothelial marker Von Willebrandfactor. The nuclei were stained with Hoechst (scale bar = 20 μm).

Figure 4. Barrier function of PBEC/As model. (A) TEER ofendothelial cells on transwell increases during the first days of co-culture with astrocytes and declines after 4 days. Values are mean ±SEM from at least six different experiments, each consisting in at leastone 24-well plate. The value at seeding (D-3) was set at 0. (B)Correlation between the permeability of FITC-dextran 40 and TEER,measured before the permeability experiment performed at D2 or D3.Each point represents one transwell insert (n = 1834).

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2075

interact with SLC2A1 and SLC16A1,41 both expressed in ourmodel at different levels. Another important fatty acid receptor,Mfsd2a,42 was highly expressed in capillary fragmentscontaining cells that form the BBB as well as in the co-cultured endothelial cells as reported in Figure 6.Functional Transport and In Vivo Correlation. Once

the expression of the principal efflux pumps was confirmed, weinvestigated the functionality of the most important trans-porters in the BBB model. One of the main advantages of theculture insert system is the possibility to measure the flux ofmolecules in the A−B direction, representing the luminal toabluminal passage, that is, blood to brain, and in the B−Adirection, representing the brain to blood transit. The polarizeddistribution of the transporters is an element indicative of adifferentiated cellular model, similar to what was shown invivo. The luminal expression of Pgp and other drug-resistant

proteins is responsible for the very low penetration of manyanticancer drugs to brain tumors. We measured thepermeability of some chemotherapeutics and other drugs inboth A−B and B−A directions.As reported in Table 5, taxol, daunomycin, and vinblastine

all displayed a significant higher permeability in B−A than A−B direction, which indicated directional transport. Moreover,the efflux ratios superior to 2 confirmed that these drugs wereeffluxed. Digoxin and quinidine (cardiovascular drugs) areknown to interact with Pgp, and they displayed an unbalancedtransport in our PBEC/As model as well, even if the ratio fordigoxin was at the lower limit. The functionality of Pgp wasfurther confirmed by the increased permeability, in the A−Bdirection, of vinblastine in the presence of GF120918, the Pgpinhibitor elacridar43 (Figure 7A). This permeability increasewas accompanied by the reduction of the efflux ratio,indicating the major role of Pgp in the countertransport ofvinblastine. In a similar way, estrone-3-sulfate and prazosin, theBCRP (ABCG2) substrates, displayed vectorial transport(Table 5). In the presence of KO143,44 the inhibitor ofBCRP, the A−B permeability was increased, resulting in theabolition of efflux (Figure 7B). This indicated that BCRP isalso functional in the PBEC/As model.We further characterized the model using a panel of 54

compounds with different mechanisms of brain penetration:passive diffusion, active influx (via transporters), active efflux(principally through Pgp and BCRP and also MRP), and a mixof these, as reported in Table 3. Our PBEC modeldemonstrated a large dynamic range with the P values rangingfrom 0.1 (FITC-dextran 70) to 72 × 10−6 cm/s (testosterone),as shown in Figure 8. The resulting broad dynamic range of360 (testosterone/FITC-dextran 70) allows discriminationbetween drugs that are not permeant from the compoundswith medium or high permeability, pointing out theamenability of the BBB model for screening. This consid-eration for the compounds that freely diffuse across themonolayer is also valid for the transported substances. Forexample, glucose, which is actively transported across the BBBby the GLUT-1 transporter, had a fourfold higher permeation

Table 4. Permeability (P) in the BBB Model of ParacellularMarkers with Different Molecular Weights (MW) andHydrodynamic Radii (HR)a

paracellular marker P (×10−6 cm/s) MW (kDa) HR (nm)

mannitol 1.2 ± 0.8 0.18 0.4sucrose 1.6 ± 0.9 0.34 0.5lucifer yellow 1.7 ± 0.8 0.52 0.5inulin 1.9 ± 1.1 6 1.4FITC-dextran 4 kDa 2.0 ± 1.3 4 1.4FITC-dextran 40 kDa 0.2 ± 0.2 40 4.5FITC-dextran 70 kDa 0.1 ± 0.1 70 6.0

aResults are mean ± SD with n > 6 from at least two separatedexperiments.

Figure 5. Transporter and receptor gene expression profile inendothelial cells co-cultured for 2 days with astrocytes. mRNAs wereanalyzed by TaqMan. The levels are relative to Glut1 (SLC2A1).Results (from a representative experiment) are mean ± SD fromtriplicate determination of biological duplicates. Genes are reportedwith their symbols (see Table 2 for full-length names).

Figure 6. Immunofluorescence micrographs for Pgp (A,C) andMfsd2a receptor (B,D) in capillary fragments (A: 48 h after plating; B:24 h after plating) with endothelial cells spreading out and in PBECs(C,D) after 2 days in co-culture. The nuclei were stained withHoechst (scale bar = 50 μm).

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2076

compared to sucrose, one of the benchmarks for the passive,paracellular permeability of in vitro BBB models. Similarly,compounds that are known substrates of L-type amino acidtransporter 1 (LAT-1) such as phenylalanine, gabapentin,leucine, and tryptophan had about a 5- or 40-fold higherpermeability compared to sucrose or FITC-dextran 70,respectively, and to less extent others substrates likekynurenine or 3-hydroxykynurenine, indicating a functionaltransport system.The model was further characterized for its ability to

transport ligands by receptor-mediated mechanisms. The

Mfsd2a receptor transports DHA in the chemical form ofLPC.42 The permeability of LPC-DHA was measured to be 3.5± 0.7 × 10−6 cm/s, indicating the presence of an activetranscytotic pathway.We further wanted to check how the permeability values

obtained in our model correlated with the in vivo finding. As insitu brain perfusion is considered the best comparative in vivomodel,45 we searched in the literature for the rodent values tobe compared. From our list of 54 compounds with differenttransport mechanisms and structures (Table 3), we couldobtain the published in situ brain perfusion values for 36molecules tested in rodents (rats or mice). To compare thevalues of the same entity, all the reported values weretransformed to express permeability in the same unit. The fulltest set correlated with the coefficient of determination (r2)and correlation (r) of 0.73 and 0.86, respectively (Figure 9).This is indicative of a high positive correlation as defined byHinkle46 and confirms the PBEC/As model as a reliable invitro alternative to in situ brain perfusion for identifying lowand high permeant compounds. The correlation fits not onlythe molecules crossing the monolayer by passive diffusion (e.g.,sucrose, atenolol, theophylline, metoprolol, and propranolol),but also those actively transported into the brain (glucose,amino acids) or actively effluxed (such as taxol, imatinib,ritonavir, or verapamil).Human brain concentration data are limited; however,

imaging tools like positron emission tomography (PET)ligands allow some calculations. The CNS-positive drugmemantine, approved for Alzheimer’s disease, showed an invitro value of 50.5 × 10−6 cm/s in our model that is very closeto the value of 55 × 10−6 cm/s measured in vivo in human.47

These data further support the use of this model as aninvestigational tool for studying the transport of therapeuticsthrough the BBB.

Permeability Evaluation of Potential Therapeutics forHD. HD is a neurodegenerative disease for which there is stillno cure. In a discovery program for CNS disease, theimplementation of an in vitro assay evaluating the mechanismsof brain penetration is of importance. In Table 6, the structuresand permeability values of four compounds originating from akynurenine mono-oxygenase (KMO) inhibitor program arereported.8 KMO is the enzyme which catalyzes the synthesis of3-hydroxykynurenine from kynurenine, the levels of the formerhaving been reported to be increased in the brains of HDpatients as well as in rodent animal models of HD.48 Thepermeability was low for three of them, in agreement with theirhydrophilic nature. We considered in our system thatcompounds displaying a permeability above 3 × 10−6 cm/shave medium/high permeability, as we found all the values of

Table 5. Permeability (P) of Compounds in A−B) and B−A Directions and the Relative Efflux Ratioa

P (×10−6 cm/s) unpaired t test

compound major efflux transporters A−B B−A P value efflux ratio

digoxin Pgp 3.7 ± 2.2 6.9 ± 3.8 <0.001 1.8quinidine Pgp 16.7 ± 1.0 53.5 ± 15.9 <0.001 3.2taxol Pgp 6.7 ± 2.8 27.8 ± 11.2 <0.001 4.2daunomycin Pgp, MRP1 4.1 ± 2.4 14.5 ± 5.9 <0.001 3.6vinblastine Pgp, MRP2 1.9 ± 1.7 6.3 ± 3.7 <0.001 3.3estrone-3-sulfate BCRP, MRPs 1.7 ± 0.6 3.3 ± 1.6 <0.05 2.0prazosin BCRP 13.4 ± 1.6 22.2 ± 11.6 <0.05 1.7mannitol 1.2 ± 0.8 0.7 ± 0.3 ns (>0.05) 0.6

aResults are mean ± SD with n > 3.

Figure 7. Functionality of efflux pumps: permeability of 0.1 μMvinblastine (A) and 1 μM prazosin (B) across PBECs seeded ontranswells in A−B and B−A sides, in the absence and in the presenceof 2 μM inhibitors: GF120918 in (A) and KO143 in (B). Efflux ratiosof B−A to A−B are reported in boxes. The results are mean ± SD of arepresentative experiment. Statistically relevant data are linked withbars reporting significance (p values obtained with unpaired t test).

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2077

the paracellular markers tested below this threshold. Incontrast, the amino acid cp3 that contains the same aromaticketone functionality was significantly more permeable as the Pvalue increased from 1.2 × 10−6 to 8.8 × 10−6 cm/s, despite itsnegative log D value. The possibility that cp3 might be a LAT-1 substrate by its amino acid nature warrants furtherexperiments.To determine the predictive power of the in vitro model for

brain availability through endothelial barrier crossing, wecompared the in vitro permeability values with the in vivobrain/plasma partition (log BB), which is the classicalpharmacokinetic value used to evaluate brain penetration.When the permeability values of 46 compounds from differentHD CNS-target discovery programs were correlated with theirbrain/plasma ratios obtained after intravenous injection inmice, two main clusters appeared: one with low in vitropermeability and low Kp and another with medium/high invitro permeability and high Kp, as shown in Figure 10. Thepresence of the two clusters prompted us to use a two-classclassification for brain partitioning and for permeability. Thelimit for low in vitro permeability was set at 3 × 10−6 cm/s (logPin vitro = −5.52). This cutoff corresponds to the highestpermeability value obtained for a paracellular marker, which,

by definition, is not permeable, plus one standard deviation(FITC-dextran 4 kDa: 2.0 ± 0.2 × 10−6 cm/s; inulin: 1.9 ± 1.1× 10−6 cm/s, Table 4). On the other hand, the boundarybetween the low and high brain/plasma ratio was set at 0.3(log brain/plasma ratio = −0.52), as it has been reported49

that the compound with a brain-to-plasma ratio superior to 0.3would have sufficient access to CNS tissue.50

In so doing, quadrants for true positive (TP), true negative(TN), false negative (FN), and false positive (FP) are definedand permit the analysis of the binary test. The classificationperformance was determined using some common measuressuch as accuracy, sensitivity, and specificity. It appears that ahigh sensitivity (96%), calculated as the TP rate defined byTP/(TP + FN), compensates a poor specificity (33%) definedas TN/(TN + FP). Among the set of 46 compounds, whentheir brain/plasma ratios were corrected for the free fraction,Kp,uu, nine compounds moved upward to the TP quadrant, andonly two were incorrectly shifted downward. As a result, the

Figure 8. Permeability of known compounds in the PBECs/As model. Values are mean ± SD (n > 3).

Figure 9. Correlation between in vitro log P from PBECs and in vivolog P determined by in situ brain perfusion in rodents (data collectedfrom the literature) for 36 compounds. The solid line is the linearregression with an R2 value of 0.73.

Table 6. Permeability and Structure of KMO InhibitorSeries in the BBB Modela

aEach compound was tested in three biological replicates. The log Dvalues were extracted by ChemAxon.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2078

specificity increased to 50%, as well as the predictivity of the invitro assay, because the positive predictive value (TP/(TP +FP)) raised from 69 to 85%; meanwhile, the negativepredictive value remained at 86% (TN/(TN + FN)). Thetwo-class classification accuracy of the model was determinedto be 72% (prediction of Kp, Figure 10A) and 87% (predictionof Kp,uu, Figure 10B). This indicates that compounds judged tohave medium−high permeability in the PBEC assay willprobably have medium−high brain exposure in vivo.

■ DISCUSSIONDespite the recent progress in understanding the BBB, itsmechanisms and functions remain to be fully elucidated. ForCNS programs, the ability of drug candidates to cross the BBBis an important property that should be addressed as soon aspossible. Because of the high attrition rate of therapeutic drugsfor CNS diseases, the most promising compounds of the leadseries should be thoroughly characterized using in vitro modelsand in vivo studies to understand the pharmacological andpharmacokinetic properties. Although the most accurate

method to determine the rate of penetration across the BBBis the in situ brain perfusion technique, this method forpractical reasons could be applied later on the drug discoveryprocesses and only to a few selected preclinical candidates.Nowadays, most research on the BBB still involves

laboratory animals as human studies are restricted topostmortem investigations or imaging techniques such asmagnetic resonance imaging and PET. Given the fundamentalrole of BBB, there is a constantly increasing need for new andadvanced model systems and techniques. The availability of invitro BBB models that accurately reflect the BBB properties invivo is a valuable alternative to the use of large numbers ofanimals in drug screening and could help to generate aworkable understanding of the neuropharmacokinetic of acompound, accelerating the CNS-oriented drug discovery anddevelopment processes.HD is characterized by neurodegeneration for which there is

no current cure except alleviation of symptoms such as chorea.Thus, like for other neurodegenerative diseases of the CNS,there is a strong medical need. The aim of the present studywas to develop an in vitro assay for the selection of brain-penetrating compounds for HD disease. For this purpose, weestablished a BBB model using endothelial cells from porcinebrains, and we used this assay for the evaluation of thetransport behavior of HD therapeutics.Because of the amenable source of material, the chosen

model was made of primary cells because cell lines are quitedeficient in achieving good tightness and receptor expression.38

BBB models based on primary endothelial cells of bovine orporcine origin have been shown to be suitable for screeningpurpose because of the high junctional tightness andexpression of functional BBB markers, either in co-culturewith astrocytes and/or pericytes.3 We adopted the co-culturesystem with only rat astrocytes as it was reported that neitherthe TEER nor the permeability characteristics gained benefitfrom the presence of pericytes.51

The BBB model has been set up with differentpharmacological classes of molecules, of varying sizes andmore importantly of various mechanisms of brain entry. Themolecules were equally distributed among passive diffusion,active uptake, and efflux, and a good in vitro−in vivocorrelation was obtained. The present in vitro−in vivocorrelation results indicated that the in vitro BBB model wasable to classify drugs into categories and to reproduce thefunctionality corresponding to the in vivo situation. To obtainthe same result (R2 equal to 0.73 vs 0.72) in comparing thepermeability coefficients using a similar number of molecules(27 vs 36 compounds, respectively), Heymans and col-leagues52 had to upgrade the cellular system with glial cellscultured in the bottom of the plates to mimic tissue absorptionand to somehow correct the values for highly boundcompounds, thus improving the predictive power of theassay. In the standard in vitro measurement, we used 0.1% BSAduring the transport assay to avoid aspecific binding and helpsolubilization. We found that some compounds, characterizedby high plasma protein binding, tested in vitro with increasingserum albumin concentrations, showed a parallel concen-tration-dependent decrease of permeability (data not shown).These results suggested that the model might overestimate thepermeability of highly protein-bound compounds. However, asreported by Hammarlund-Udenaes et al.53 the plasma proteinbinding is not in itself a limitation for CNS pharmacologicalactivity. In vitro BBB permeability is a measurement of the rate

Figure 10. Correlation between in vitro permeability and brain/plasma ratio for 46 proprietary compounds. Kp was considered in (A),whereas the free fraction (Kp,uu) was used in (B). Open squaresymbols denote compounds moving to the upper correct quadrantafter correction for unbound fraction, whereas open circle symbolsdenote compounds moving downward the incorrect quadrant.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2079

of drug delivery to the brain, which should be evaluatedtogether with the extent (e.g., the ratio of drug concentrationin brain and plasma) of the distribution to fully understandbrain drug delivery. In our model, when the permeability ofcompounds from the HD CNS-target discovery programs wascorrelated with the brain/plasma ratio obtained after intra-venous injection in mice, the correction for protein bindingslightly improves the predictive value of brain/plasmapartitioning based on in vitro permeability. In other words,the in vitro PBEC/As model, within the limited data set, mayunderscore the extent of brain partition, demonstrating itsutility in a drug discovery project as a tool for prioritization ofcompounds for pharmacokinetic studies.The dynamic range of the model indicated that it can

discriminate between drugs with low, medium, and high Pvalues, and it could be used as a permeability screen toinvestigate the BBB permeation of new investigationalcompounds, displaying various physicochemical properties,including those that are substrates for transporters, whether ornot the particular transporters involved have been identified.The assessment of the apical basal polarity of the modelbecause of the presence of vectorial transport such as thatmediated by Pgp provides valuable information about themechanisms involved. Pgp has been considered as being one ofthe most important transporters playing a role in pharmaco-kinetics in human,54 and therefore it was used as a model effluxtransporter in this work. The demonstration of thefunctionality of Pgp by using bidirectional transport assayindicated that our model can be used for the prediction of Pgp-mediated efflux.Our model has been validated not only for the transport of

small-weight compounds but also for larger molecules such asLPC-DHA, transported by the major facilitator superfamilydomain containing protein 2a (Mfsd2a) highly expressed in theBBB endothelium.42,55 Mfsd2a transports DHA in the chemicalform of LPC in a sodium-dependent manner and plasma LPCscarrying long-chain fatty acids such as LPC oleate and LPCpalmitate. The expression of transporters like LDLR, TFRC,Mfsd2a, and BSG in the BBB model opens the possibility tostudy the transport of compounds engineered to use theirnatural ligand as Trojan horse or to directly monitor thetransport of macromolecules, primarily otherwise retained inthe blood.The BBB model through its medium−low throughput has

been inserted in the funnel for the drug development of HD-targeting compounds in a lead optimization phase. An earlierplacement will be time-consuming with poor information gain.In the later phase of development, the assay might be used formechanistic study such as the characterization of potentialtransporters. The assay described here has several character-istics making it useful for screening in different developmentalphases such as lead optimization: (i) amenable to a 24-wellplate format, (ii) medium throughput (about 20 test samples/endothelial cell preparations from six brains), (iii) presence ofan IS for monolayer integrity, and (iv) the possibility to testdirectional transport using inhibitors and B−A permeability.The setting of bioanalytical conditions remains unfortunatelytime-consuming as it needs to be developed for each testedcompound. The use of a cocktail of compounds in theincubation step would increase the model efficiency; however,Berezowki et al.56 showed in their publication that no morethan one Pgp substrate could be added in the transport assay,

to avoid substantial effect on the permeability of theparacellular marker sucrose.Our BBB model was used for selecting and ranking

compounds with specific transport features and not as areplacement of in vivo evaluation of the CNS target exposure.Our study relies on the endothelial cells from porcine originand shows a correlation with rodent in vivo data used forpreclinical studies. To improve the efficiency of CNS drugdiscovery, there is the need to combine in vitro and in vivopenetration assessments into an integrated preclinical work-flow. Although the porcine model used appears to correlatewell with the findings obtained in rodents, the translatability tohumans may be affected by the differences among species inthe expression of transporters and receptors. For example, PETstudies showed significant differences among species in thebrain pharmacokinetics of Pgp substrates,57 and quantitativeproteomics demonstrated the predominance of ABCG2(BCRP) efflux transporter over Pgp in the BBB of humansversus rodents.38 A recent review by O’Brown et al.58

highlights the similarities and differences between species, inparticular human brains display a higher astrocyte-to-neuronratio and increased complexity. However, the precise func-tional properties of the human barrier remain elusive and thetranslation of rodent data to human is far from straightforward.Thus, it is very important to complement data obtained inpreclinical species with the results obtained from in vitrohuman BBB models. The advent of human induced pluripotentstem cell (iPSC)-derived BBB models aims at bridging this gapin species-to-species differences, as well as provides oppor-tunity to generate the BBB model from patient-derived iPSCshaving the same genotype. Indeed, in vitro BBB models havebeen generated from HD patient-derived iPSCs, and suchmodels may be the most appropriate for these studies59,60

The alterations of the BBB functionality reported forneurodegenerative diseases, such as Alzheimer’s and/orParkinson’s diseases, suggest that human in vitro BBB modelhas an important role in drug discovery not only because it canbe used for screening purpose but also to understand thebiology of the disease involving the BBB as well as to elucidatethe specific changes during the disease progression. Theknowledge of the existence of these changes could help thedevelopment of the treatment strategies. The stem cell-derivedendothelial cells expressed the relevant BBB markers, and thetightness of the system is improved compared to other brainendothelial cell-based models.The tightness of the BBB appeared higher in healthy people

than in affected HD patients. Drouin-Ouellet reported BBBleakage in the striatum of R6/2 mice and confirmed similarBBB changes in HD patients.61 More recently, the laboratoryof L. Thompson reported drastically lower TEER values formonolayers of iBMECs derived from HD patients with longCAG repeats (71Q and 109Q) compared to shorter repeats(66Q) and healthy controls.60 The role of mHTT at the BBBremains to be elucidated, but the transport study of a panel ofknown molecules would help to better understand how leakymight be the BBB of HD patients and its relation with diseaseprogression and severity.

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: [email protected]. Phone: +1 310703 2973.*E-mail: [email protected]. Phone: +39 06 91093 221.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2080

ORCIDAnnalise Di Marco: 0000-0001-5211-6230Author ContributionsThe manuscript was written through contributions of allauthors. All authors have given approval to the final version ofthe manuscript.FundingThis work was financially supported by CHDI Foundation, Inc.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSWe would like to acknowledge Isabelle Gloaguen for hersupport in discussion and critical reading of the manuscript.We would also like to thank Sergio Altamura for his constantpresence and support. Thanks to Ralph Laufer for his help instarting this project and to Alberto Bresciani and MauroCerretani for assisting with the INCell imaging experiments.Our thanks to the public slaughterhouse in Rome, Servizioveterinario, Dip. di Prevenzione Az. USL Roma 2, andVitantonio Perrone for supplying the fresh porcine brainsamples.

■ ABBREVIATIONSHD, Huntington’s disease; CNS, central nervous system;TEER, trans-endothelial electrical resistance; PBEC, porcinebrain endothelial cell; As, astrocyte; LC−MS/MS, liquidchromatography−tandem mass spectrometry; Pgp, P-glyco-protein; BCRP, breast cancer resistance protein; MRP,multidrug resistance-associated protein

■ REFERENCES(1) McColgan, P.; Tabrizi, S. J. Huntington’s Disease: A ClinicalReview. Eur. J. Neurol. 2018, 25, 24−34.(2) Franke, H.; Galla, H.-J.; Beuckmann, C. T. Primary Cultures ofBrain Microvessel Endothelial Cells: A Valid and Flexible Model toStudy Drug Transport through the Blood-Brain Barrier in Vitro. BrainRes. Protoc. 2000, 5, 248−256.(3) Patabendige, A.; Skinner, R. A.; Morgan, L.; Joan Abbott, N. ADetailed Method for Preparation of a Functional and Flexible Blood-Brain Barrier Model Using Porcine Brain Endothelial Cells. Brain Res.2013, 1521, 16−30.(4) Toth, A.; Veszelka, S.; Nakagawa, S.; Niwa, M.; Deli, M. A.Patented in Vitro Blood-Brain Barrier Models in CNS DrugDiscovery. Recent Pat. CNS Drug Discovery 2011, 6, 107−118.(5) Deli, M. A.; Abraham, C. S.; Kataoka, Y.; Niwa, M. PermeabilityStudies on in Vitro Blood-Brain Barrier Models: Physiology,Pathology, and Pharmacology. Cell. Mol. Neurobiol. 2005, 25, 59−127.(6) Zhang, Y.; Li, C. S. W.; Ye, Y.; Johnson, K.; Poe, J.; Johnson, S.;Bobrowski, W.; Garrido, R.; Madhu, C. Porcine Brain MicrovesselEndothelial Cells as an in Vitro Model to Predict in Vivo Blood-BrainBarrier Permeability. Drug Metab. Dispos. 2006, 34, 1935−1943.(7) Strom, S. C.; Michalopoulos, G. [29] Collagen as a Substrate forCell Growth and Differentiation. Methods Enzymol. 1982, 82, 544−555.(8) Toledo-Sherman, L. M.; Prime, M. E.; Mrzljak, L.; Beconi, M.G.; Beresford, A.; Brookfield, F. A.; Brown, C. J.; Cardaun, I.;Courtney, S. M.; Dijkman, U.; et al. Development of a Series of ArylPyrimidine Kynurenine Monooxygenase Inhibitors as PotentialTherapeutic Agents for the Treatment of Huntingtons Disease. J.Med. Chem. 2015, 58, 1159−1183.(9) Skinner, R. A.; Gibson, R.; Rothwell, N.; Pinteaux, E.; Penny, J.Transport of Interleukin-1 across Cerebromicrovascular EndothelialCells. Br. J. Pharmacol. 2009, 156, 1115−1123.

(10) Abbott, N. J.; Dolman, D. E. M.; Drndarski, S.; Fredriksson, S.M. An Improved In Vitro Blood-Brain Barrier Model: Rat BrainEndothelial Cells Co-cultured with Astrocytes. Astrocytes 2012, 814,415−430.(11) Gaillard, P. J.; De Boer, A. G. Relationship betweenPermeability Status of the Blood-Brain Barrier and in VitroPermeability Coefficient of a Drug. Eur. J. Pharm. Sci. 2000, 12,95−102.(12) Pardridge, W. M.; Triguero, D.; Yang, J.; Cancilla, P. A.Comparison of in Vitro and in Vivo Models of Drug Transcytosisthrough the Blood-Brain Barrier. J. Pharmacol. Exp. Ther. 1990, 253,884−891.(13) Fenstermacher, J.; Gross, P.; Sposito, N.; Acuff, V.; Pettersen,S.; Gruber, K. Structural and Functional Variations in CapillarySystems within the Brain. Ann. N.Y. Acad. Sci. 1988, 529, 21−30.(14) Fukui, S.; Schwarcz, R.; Rapoport, S. I.; Takada, Y.; Smith, Q.R. Blood−Brain Barrier Transport of Kynurenines: Implications forBrain Synthesis and Metabolism. J. Neurochem. 1991, 56, 2007−2017.(15) Doan, K. M. M.; Lakhman, S. S.; Boje, K. M. K. Blood-BrainBarrier Transport Studies of Organic Guanidino Cations Using an inSitu Brain Perfusion Technique. Brain Res. 2000, 876, 141−147.(16) Avdeef, A. How Well Can in Vitro Brain MicrocapillaryEndothelial Cell Models Predict Rodent in Vivo Blood-Brain BarrierPermeability? Eur. J. Pharm. Sci. 2011, 43, 109−124.(17) Park, S.; Sinko, P. J. The Blood-Brain Barrier Sodium-Dependent Multivitamin Transporter: A Molecular Functional inVitro-in Situ Correlation. Drug Metab. Dispos. 2005, 33, 1547−1554.(18) Liu, X.; Tu, M.; Kelly, R. S.; Chen, C.; Smith, B. J.Development of a Computational Approach to Predict Blood-BrainBarrier Permeability. Drug Metab. Dispos. 2004, 32, 132−139.(19) Egleton, R. D.; Davis, T. P. Transport of the δ-Opioid ReceptorAgonist [D-Penicillamine2,5] Enkephalin across the Blood−brainBarrier Involves Transcytosis1. J. Pharm. Sci. 1999, 88, 392−397.(20) Summerfield, S. G.; Read, K.; Begley, D. J.; Obradovic, T.;Hidalgo, I. J.; Coggon, S.; Lewis, A. V.; Porter, R. A.; Jeffrey, P.Central Nervous System Drug Disposition: The Relationship betweenin Situ Brain Permeability and Brain Free Fraction. J. Pharmacol. Exp.Ther. 2007, 322, 205−213.(21) Pardridge, W. M. Brain Metabolism: A Perspective from theBlood-Brain Barrier. Physiol. Rev. 1983, 63, 1481−1535.(22) Bihorel, S.; Camenisch, G.; Gross, G.; Lemaire, M.;Scherrmann, J.-M. Influence of Hydroxyurea on Imatinib Mesylate(Gleevec) Transport at the Mouse Blood-Brain Barrier. Drug Metab.Dispos. 2006, 34, 1945−1949.(23) Abbruscato, T. J.; Thomas, S. A.; Hruby, V. J.; Davis, T. P.Blood-Brain Barrier Permeability and Bioavailability of a HighlyPotent and Mu-Selective Opioid Receptor Antagonist, CTAP:Comparison with Morphine. J. Pharmacol. Exp. Ther. 1997, 280,402−409.(24) Chikhale, E. G.; Chikhale, P. J.; Borchardt, R. T. Carrier-Mediated Transport of the Antitumor Agent Acivicin across theBlood-Brain Barrier. Biochem. Pharmacol. 1995, 49, 941−945.(25) Gratton, J. A.; Abraham, M. H.; Bradbury, M. W.; Chadha, H.S. Molecular Factors Influencing Drug Transfer across the Blood-Brain Barrier. J. Pharm. Pharmacol. 1997, 49, 1211−1216.(26) Suzuki, T.; Ohmuro, A.; Miyata, M.; Furuishi, T.; Hidaka, S.;Kugawa, F.; Fukami, T.; Tomono, K. Involvement of an InfluxTransporter in the Blood-Brain Barrier Transport of Naloxone.Biopharm. Drug Dispos. 2010, 31, 243−252.(27) Spector, R.; Sivesind, C.; Kinzenbaw, D. Pantothenic AcidTransport Through the Blood-Brain Barrier. J. Neurochem. 1986, 47,966−971.(28) Momma, S.; Aoyagi, M.; Rapoport, S. I.; Smith, Q. R.Phenylalanine Transport Across the Blood-Brain Barrier as Studiedwith the In Situ Brain Perfusion Technique. J. Neurochem. 1987, 48,1291−1300.(29) Cisternino, S.; Mercier, C.; Bourasset, F.; Roux, F.;Scherrmann, J.-M. Expression, Up-Regulation, and Transport Activity

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2081

of the Multidrug-Resistance Protein Abcg2 at the Mouse Blood-BrainBarrier. Cancer Res. 2004, 64, 3296−3301.(30) Chen, W. Evaluation of the Permeation Characteristics of aModel Opioid Peptide, H-Tyr-D-Ala-Gly-Phe-D-Leu-OH (DADLE),and Its Cyclic Prodrugs across the Blood-Brain Barrier Using an InSitu Perfused Rat Brain Model. J. Pharmacol. Exp. Ther. 2002, 303,849−857.(31) Dagenais, C.; Avdeef, A.; Tsinman, O.; Dudley, A.; Beliveau, R.P-Glycoprotein Deficient Mouse in Situ Blood-Brain BarrierPermeability and Its Prediction Using an in Combo PAMPAModel. Eur. J. Pharm. Sci. 2009, 38, 121−137.(32) Rice, A.; Liu, Y.; Michaelis, M. L.; Himes, R. H.; Georg, G. I.;Audus, K. L. Chemical Modification of Paclitaxel (Taxol) Reduces P-Glycoprotein Interactions and Increases Permeation across the Blood-Brain Barrier in Vitro and in Situ. J. Med. Chem. 2005, 48, 832−838.(33) Youdim, K. A.; Qaiser, M. Z.; Begley, D. J.; Rice-Evans, C. A.;Abbott, N. J. Flavonoid Permeability across an in Situ Model of theBlood-Brain Barrier. Free Radical Biol. Med. 2004, 36, 592−604.(34) Diksic, M.; Tohyama, Y.; Takada, A. Brain Net UnidirectionalUptake of α-[14C]Methyl-L-Tryptophan “α-MTrp” and Its Correla-tion with Regional Serotonin Synthesis, Tryptophan Incorporationinto Proteins, and Permeability Surface Area Products of Tryptophanand α-MTrp. Neurochem. Res. 2000, 25, 1537−1546.(35) Zhao, R.; Kalvass, J. C.; Pollack, G. M. Assessment of Blood−Brain Barrier Permeability Using the In Situ Mouse Brain PerfusionTechnique. Pharm. Res. 2009, 26, 1657−1664.(36) Murakami, H.; Takanaga, H.; Matsuo, H.; Ohtani, H.; Sawada,Y. Comparison of Blood-Brain Barrier Permeability in Mice and RatsUsing in Situ Brain Perfusion Technique. Am. J. Physiol.: Heart Circ.Physiol. 2000, 279, H1022−H1028.(37) Helms, H. C.; Hersom, M.; Kuhlmann, L. B.; Badolo, L.;Nielsen, C. U.; Brodin, B. An Electrically Tight In Vitro Blood−BrainBarrier Model Displays Net Brain-to-Blood Efflux of Substrates for theABC Transporters, P-Gp, Bcrp and Mrp-1. AAPS J. 2014, 16, 1046−1055.(38) Stanimirovic, D. B.; Bani-Yaghoub, M.; Perkins, M.; Haqqani,A. S. Blood−brain Barrier Models: In Vitro to in Vivo Translation inPreclinical Development of CNS-Targeting Biotherapeutics. ExpertOpin. Drug Discovery 2015, 10, 141−155.(39) Kilic, E.; Spudich, A.; Kilic, U.; Rentsch, K. M.; Vig, R.; Matter,C. M.; Wunderli-Allenspach, H.; Fritschy, J.-M.; Bassetti, C. L.;Hermann, D. M. ABCC1: A Gateway for PharmacologicalCompounds to the Ischaemic Brain. Brain 2008, 131, 2679−2689.(40) Jones, A. R.; Shusta, E. V. Blood-Brain Barrier Transport ofTherapeutics via Receptor-Mediation. Pharm Res. 2007, 24, 1759−1771.(41) Muramatsu, T. Basigin (CD147), a Multifunctional Trans-membrane Glycoprotein with Various Binding Partners. J. Biochem.2016, 159, 481−490.(42) Nguyen, L. N.; Ma, D.; Shui, G.; Wong, P.; Cazenave-Gassiot,A.; Zhang, X.; Wenk, M. R.; Goh, E. L. K.; Silver, D. L. Mfsd2a Is aTransporter for the Essential Omega-3 Fatty Acid DocosahexaenoicAcid. Nature 2014, 509, 503−506.(43) Hyafil, F.; Vergely, C.; Vignaud, P. Du; Grand-Perret, T. InVitro and in Vivo Reversal of Multidrug Resistance by GF120918, anAcridonecarboxamide Derivative. Cancer Res. 1993, 53, 4595−4602.(44) Li, Y.; Hayman, E.; Plesescu, M.; Prakash, S. R. Synthesis ofPotent BCRP Inhibitor-Ko143. Tetrahedron Lett. 2008, 49, 1480−1483.(45) Zhao, R.; Kalvass, J. C.; Pollack, G. M. Assessment of Blood-Brain Barrier Permeability Using the in Situ Mouse Brain PerfusionTechnique. Pharm. Res. 2009, 26, 1657−1664.(46) Hinkle, D. E.; Wiersma, W.; Jurs, S. G. Applied Statistics for theBehavioral Sciences, 5th ed.; Houghton Mifflin Harcourt, 2003.(47) Higuchi, K.; Kitamura, A.; Okura, T.; Deguchi, Y. MemantineTransport by a Proton-Coupled Organic Cation Antiporter inHCMEC/D3 Cells, an in Vitro Human Blood-Brain Barrier Model.Drug Metab. Pharmacokinet. 2015, 30, 182−187.

(48) Beaumont, V.; Mrzljak, L.; Dijkman, U.; Freije, R.; Heins, M.;Rassoulpour, A.; Tombaugh, G.; Gelman, S.; Bradaia, A.; Steidl, E.;et al. The Novel KMO Inhibitor CHDI-340246 Leads to aRestoration of Electrophysiological Alterations in Mouse Models ofHuntington’s Disease. Exp. Neurol. 2016, 282, 99−118.(49) Kikuchi, R.; De Morais, S. M.; Kalvass, J. C. In Vitro P-Glycoprotein Efflux Ratio Can Predict the in Vivo Brain PenetrationRegardless of Biopharmaceutics Drug Disposition ClassificationSystem Class. Drug Metab. Dispos. 2013, 41, 2012.(50) Bhardwaj, R.; Chandrasena, G. Optimal ADME Properties forClinical Candidate and Investigational New Drug (IND) Package.ADME-Enabling Technologies in Drug Design and Development; JohnWiley & Sons, Inc., 2012; pp 15−28.(51) Thomsen, L. B.; Burkhart, A.; Moos, T. A Triple Culture Modelof the Blood-Brain Barrier Using Porcine Brain Endothelial Cells,Astrocytes and Pericytes. PLoS One 2015, 10, No. e0134765.(52) Heymans, M.; Sevin, E.; Gosselet, F.; Lundquist, S.; Culot, M.Mimicking Brain Tissue Binding in an in Vitro Model of the Blood-Brain Barrier Illustrates Differences between in Vitro and in VivoMethods for Assessing the Rate of Brain Penetration. Eur. J. Pharm.Biopharm. 2018, 127, 453−461.(53) Hammarlund-Udenaes, M.; Friden, M.; Syvanen, S.; Gupta, A.On the Rate and Extent of Drug Delivery to the Brain. Pharm. Res.2008, 25, 1737−1750.(54) Bauer, M.; Romermann, K.; Karch, R.; Wulkersdorfer, B.;Stanek, J.; Philippe, C.; Maier-Salamon, A.; Haslacher, H.; Jungbauer,C.; Wadsak, W.; et al. Pilot PET Study to Assess the FunctionalInterplay Between ABCB1 and ABCG2 at the Human Blood−BrainBarrier. Clin. Pharmacol. Ther. 2016, 100, 131−141.(55) Ben-Zvi, A.; Lacoste, B.; Kur, E.; Andreone, B. J.; Mayshar, Y.;Yan, H.; Gu, C. Mfsd2a Is Critical for the Formation and Function ofthe Blood-Brain Barrier. Nature 2014, 509, 507−511.(56) Berezowski, V.; Landry, C.; Lundquist, S.; Dehouck, L.;Cecchelli, R.; Dehouck, M. P.; Fenart, L. Transport Screening of DrugCocktails through an in Vitro Blood-Brain Barrier: Is It a GoodStrategy for Increasing the Throughput of the Discovery Pipeline?Pharm. Res. 2004, 21, 756−760.(57) Aday, S.; Cecchelli, R.; Hallier-vanuxeem, D.; Dehouck, M. P.;Ferreira, L. Stem Cell-Based Human Blood − Brain Barrier Modelsfor Drug Discovery and Delivery. Trends Biotechnol. 2016, 34, 382−393.(58) O’Brown, N. M.; Pfau, S. J.; Gu, C. Bridging Barriers: AComparative Look at the Blood-Brain Barrier across Organisms. GenesDev. 2018, 32, 466−478.(59) Lippmann, E. S.; Azarin, S. M.; Kay, J. E.; Nessler, R. A.;Wilson, H. K.; Al-Ahmad, A.; Palecek, S. P.; Shusta, E. V. Derivationof Blood-Brain Barrier Endothelial Cells from Human PluripotentStem Cells. Nat. Biotechnol. 2012, 30, 783−791.(60) Lim, R. G.; Quan, C.; Reyes-Ortiz, A. M.; Lutz, S. E.; Kedaigle,A. J.; Gipson, T. A.; Wu, J.; Vatine, G. D.; Stocksdale, J.; Casale, M. S.;et al. Huntington’s Disease IPSC-Derived Brain MicrovascularEndothelial Cells Reveal WNT-Mediated Angiogenic and Blood-Brain Barrier Deficits. Cell Rep. 2017, 19, 1365−1377.(61) Drouin-Ouellet, J.; Sawiak, S. J.; Cisbani, G.; Lagace, M.; Kuan,W.-L.; Saint-Pierre, M.; Dury, R. J.; Alata, W.; St-Amour, I.; Mason, S.L.; et al. Cerebrovascular and Blood-Brain Barrier Impairments inHuntington’s Disease: Potential Implications for Its Pathophysiology.Ann. Neurol. 2015, 78. DOI: 10.1002/ana.24406.

Molecular Pharmaceutics Article

DOI: 10.1021/acs.molpharmaceut.9b00042Mol. Pharmaceutics 2019, 16, 2069−2082

2082