Introduction

For central nervous system (CNS) drugs, the step from in vitro activity to in vivo efficacy depends critically on the ability of the drug to cross the blood-brain barrier (BBB). The BBB is not merely a physical barrier but rather is an exquisitely modulated, functional gate1. Sensitive to biochemical signals from its local environment, it controls the influx and efflux of a wide variety of substances presumably required for optimal brain function2. The development of neuroactive substances in the pharmaceutical industry is an extremely costly endeavour plagued by a very high risk due to the high attrition rate in various stages of development. The success rate from first-in-human to registration for neuroactive drugs targeting the CNS is significantly lower than for other indications, such as cardiovascular, infectious, inflammatory, and metabolic diseases3,4. The development of drugs targeting the CNS requires precise knowledge of their penetration into the brain2, and ideally, this information should be obtained as early as possible to avoid failure in late clinical development, when upward of $100 million is typically invested in a drug candidate. Therefore, it is important to develop in vitro reliable screening tools that will allow the optimization of neuroactive molecules for brain penetration. Our group conceived and first developed static rodent and human BBB models using primary cells5,6,7,8. Due to the difficulties to get access to human brain tissue, several groups have attempted to develop in vitro human BBB models using induced pluripotent stem cells (iPSCs). Several in vitro models have been proposed9,10. These 2D/3D models are based on the use of iPSCs in static or microfluidic conditions11,12,13,14,15,16,17,18,19,20. Whatever their ressemblance to the in vivo BBB, in vitro BBB models must be carefully assessed for their ability to predict accurately the passage of drugs into the CNS in vivo. Nowadays, all models, whatether their characteristics are validated using the parameters of drug BBB passage measured directly in laboratory animals. Due to the poor predictive value of animal models for human responses to drugs, the cross-correlations between the in vitro human BBB model and the pharmacokinetic data in the in vivo human brain is mandatory to demonstrate the predictive value of the in vitro screening tools.

Here we report the development of a human BBB model using two different iPSC lines based on the optimization of the protocol previously reported15. Using these iPSC lines, we addressed their differentiation into brain endothelial cells (BECs) and their capacity to generate a tight monolayer in co-culture with glial cells. We also compared the permeability of 8 compounds using this in vitro iPSC-hBBB model with quantitative clinical data regarding their BBB permation obtained using brain PET imaging. We found a good correlation between the in vitro and in vivo drug brain permeability (R2 = 0.83; P = 0.008). In addition, we noted a lack of relationship between in vitro apparent permeability for 18 CNS and non-CNS compounds using the in vitro iPSC-hBBB model and the compounds’ physicochemical properties. Even if the BBB permeabilities showed the similar ranking of the tested compounds between both in vitro iPSC-hBBB and primary rat models, we noticed species differences considering their interactions with ABC transporters.

Results

iPSCs generation and human brain endothelial cells differentiation

The ED-iPSC line has been provided through the Harvard Stem Cell Institute iPS Core Facility. For the SP-iPSC line, Placenta from fibroblasts explants (Serena Pavoni, PhD, CEA Fontenay aux Roses/SupBiotech, France) was reprogrammed using the Sendai virus method until colonies began to adopt an iPSC-like morphology 15 days after transduction with OCT4, KLF4, SOX2 and C-myc (data not shown). After 6 to 10 passages on mouse embryonic fibroblasts (MEFs), iPSC lines were adapted to feeder-free conditions (Matrigel) and regularly verified for their pluripotency state. iPSC lines expressed TRA-1–60 and SSEA4 markers (Fig. 1A). We showed the expression of the endogenous pluripotency marker genes OCT4, SOX2, REX1 and NANOG (Fig. 1B) while no expression was shown for those genes in the primary cells. In addition, the iPSC lines were able to generate in vivo the three embryonic germ layers (ectoderm, endoderm and mesoderm; Fig. 1C).

Figure 1
figure 1

Establishment and characterization of iPSCs. (A) Representative analysis of TRA1-60 and SSEA-4 expression for each iPSC cell line by flow cytometry. TRA1-60 is compared with unlabeled cells and SSE-A4 is compared with the appropriate mouse IgG3 control. Control: green line; test: purple line. Samples of somatic control cells (Caco-2 epithelial cells for TRA1-60 and hBMEC for SSEA-4) were tested in parallel (respectively). Flow cytometry analysis of the iPSCs lines were performed in exponential growth phase. (B) Gene expression analysis of pluripotency markers in both iPSC and in fibroblast control by semi-quantitative PCR. (C) Ability of the iPSC-SP cell line to generate all three germ cell layer components within teratomas (endoderm – line 1, mesoderm – line 2 and ectoderm – line 3) was evaluated. Whole sections of teratoma were performed after 5 to 10 weeks of growth and stained with hematoxylin and eosin. Scale bare = 50 µm.

Full size image

iPSC-derived human brain endothelial cells (iPSC-hBECs)

Representative morphology changes of iPSC during differentiation are shown in Fig. 2A. The two step differentiation protocol adapted from Lipmann et al.15 with introduction of a selection step with puromycin yielded cells which exhibited brain endothelial cell morphology (Fig. 2A). Endothelial cells emerged at day 2 after addition of endothelial medium. The iPSC-hBECs showed morphology similar to that of primary cultures of brain endothelial cells with monolayers of tightly packed cells expressing ZO-1 and claudin-5 (Fig. 2B). In contrast to CD31 and ABCB1 mRNA, the mRNA expression level of transporters like ABCC1, ABCG2, TfRc and INSR in iPSC-hBECs was comparable to that observed in our primary brain endothelial cells (Fig. 2C).

Figure 2
figure 2

Differentiation of iPSC-hBECs. (A) Representative phase contrast images of iPSCs in cell culture at day 5 and after differentiation. By day 8, cells exhibited brain endothelial cell morphology in the presence of puromycin. Black scale bare, 500 µm. White scale bare, 50 µm. (B) ZO-1 and claudin-5 fluorescence intensity demonstrated that the differentiated cells exhibited brain endothelial characteristics. Scale bare, 50 µm. (C) Typical profile of transporter/receptor gene expression. The level of the expression of mRNA of junctional proteins like claudin-5 and ZO-1, transporters like ABCC1, ABCG2 and LRP1 in iPSC-Human BECs was comparable to that observed in our two primary hBECs samples.

Full size image

Functional and clinical validation of iPSC-hBBB models

To further demonstrate that the in vitro iPSC-hBBB monolayer displayed well-developed tight junctions, we investigated the permeability of a paracellular marker, i.e. sucrose, across the brain endothelial cell monolayer. The mean permeability value of sucrose across the iPSC-hBBB monolayer generated with the two iPSC lines was 2.9 ± 1.6 × 10−6 cm.s−1 (Fig. 3B) in comparison with propranolol, a highly permeable BBB marker [21.7 ± 3.1 × 10−6 cm.s−1 (Fig. 3D)], indicating an overall restrictive paracellular permeation as confirmed by the average of TEER values (458 ± 225 Ω.cm2).

Figure 3
figure 3

Low paracellular transport of sucrose and high permeability of propranolol in the in vitro iPSC-based human BBB model. (A,C) Each column represents the average of the apparent permeability (mean ± SD) from apical to basal compartment obtained on independent iPSC-BBB in vitro models (7 to 8 experiments). (B,D) each column represents the average of the apparent permeability from apical to basal compartment obtained in the two cell lines (mean ± SEM).

Full size image

A known P-gp substrate was used to assess the asymmetry index of the iPSC- hBBB model. In this context, the apparent permeability of vinblastine was measured bi-directionally: from apical to basal (PappAB) and from basal to apical compartments (PappBA) in the presence or absence of the P-gp inhibitor cyclosporine A. The PappBA/PappAB efflux ratio (ER) was calculated. The ER generated for each single batch of iPSCs is plotted in Fig. 4A; the mean value for the two iPSC- hBBB models from 13 independent experiments was 1.55 ± 0.40 [(CV: 25.80%); Fig. 4B]. P-gp inhibition using Cyclosporine A (CsA) resulted in a 50% decrease of vinblastine ER indicating the presence of functional P-gp efflux in the iPSC-hBBB model (Fig. 4C).

Figure 4
figure 4

Assessment of the bidirectional transport of the P-gp transporter substrate vinblastine in the iPSC-hBBB models. (A) ER of iPSC-hBBB models from different iPSC batches. (B) ER mean ratio in iPSC-hBBB models from at least 6 identical iPSC showing the ER intra-variability (mean ± SEM). (C) Effect of CsA on the vinblastine ER (mean ± SD).

Full size image

To further assess the relevance of the in vitro iPSC-hBBB model, we used compounds with known passage across the human BBB in vivo [rate constant for transfer from arterial plasma into brain obtained from compartment modeling of PET data, K1, mL.cm−3.min−1 (S1 Table 1)]. These parameters were compared with the apparent permeability obtained using the in vitro iPSC-hBBB model (PappAB). Permeability determined with the in vitro iPSC-hBBB model showed a highly significant correlation with the in vivo K1 (R2 = 0.83; P = 0.008; Fig. 5). While fluoro-A85380 and raclopride showed a moderate cerebral uptake, befloxatone, buprenorphine as well as flumazenil exhibited a high cerebral uptake in human brain and also a high permeability across the in vitro system. In contrast, loperamide and verapamil exhibited low cerebral uptake with limited K1 in vivo as well as a low apparent permeability in vitro (Fig. 5). A discrepancy was observed with erlotinib which exhibited a low cerebral uptake in human brain but high permeability in vitro. This could be explain by the fact that the concentration of erlotinib used in vitro induced inhibition of P-gp/BCRP expressed in the in vitro iPSC-hBBB model. Although this study needs to be extended to a larger set of radioligands with known PK parameters in human brain, our data suggest that the in vitro iPSC-hBBB model will be useful in drug screening strategies. In this context, we evaluated the permeability of a larger panel of marketed drugs across the in vitro iPSC-hBBB model. A robust and validated LC-MS/MS method was used to evaluate the ability of the in vitro iPSC-hBBB model to discriminate between CNS and non CNS drugs (Table 1). The mean Papp value for the marketed non-CNS compounds from the apical to basal compartment was about 2.1 ± 0.6 × 10−6 cm.s−1 (range: 1.3 × 10−6 cm.s−1 to 2.9 × 10−6 cm.s−1) which is consistent with a limited passage across the cell monolayer. With the exception of erlotinib, drugs known to be substrate of efflux transporter such as digoxin, loperamide, fluoro-A85380, verapamil and vinblastine are likely to be also ABC transporter substrates in the iPSC-hBBB model. The mean Papp value for CNS compounds such as dextromethorphan, caffeine, flumazenil, raclopride, buprenorphine, befloxatone, and propranolol, was about 30.1 ± 4 × 10−6 cm.s−1 (range: 21.7 × 10−6 cm.s−1 to 38.4 × 10−6 cm.s−1) consistent with high permeability (Fig. 5). Consistent with previous observations, no relationship between drug physicochemical properties and in vitro apparent permeability was evidenced despite the fact that compounds cLogP < 5, MW < 450 Da and PSA < 120 Å displayed high in vitro apparent permeability (Fig. 6).

Figure 5
figure 5

Clinical validation of iPSC- hBBB Blood-Brain Barrier model. Linear regression analysis shows a positive correlation between apparent permeability in human brain (K1) and in vitro apparent permeability in the iPSC-hBBB model. Spearman tests were performed and *P < 0.05 indicates y significant correlation.

Full size image
Table 1 Summary of Physicochemical Properties of the Set of Selected Compounds.
Full size table
Figure 6
figure 6

iPSC- hBBB Blood-Brain Barrier model permeability characterization. Apparent permeability values of marketed CNS and non-CNS drugs from the in vitro iPSC-hBBB model and relationship with their molecular weight (MW), cLogP, and surface polar area (PSA). Data are the average of two or three cell monolayers for each tested compound.

Full size image

Species differences study

We previously described an in vitro primary cell-based rat BBB model with relevant features of the in vivo rat BBB6. We therefore analyzed correlatively the permeability of a set of compounds obtained using the in vitro iPSC-hBBB model and the in vitro cell based rat BBB model. A good correlation was observed between the two models (R2 = 0.85; P < 0.0001). We noticed that the permeability value for compounds such as dextromethorphan, caffeine and loperamide was higher using the in vitro cell based rat BBB model than with the iPSC-hBBB model (Fig. 7A), while buprenorphine showed high permeability in only the iPSC-hBBB model. Fluoro-A85380, verapamil, vinblastine and digoxin were likely to be ABC-transporter substrates (ER ≥ 1.25) in both in vitro BBB models (Fig. 7B). However, loperamide apparently interacted with ABC transporters in the iPSC-hBBB but not in the in vitro cell based rat BBB model. Erlotinib was apparently an ABC transporter substrate in the primary rat model but not in the in vitro iPSC-hBBB model (Fig. 7B). These observations confirmed interspecies differences as has been pointed previously6,21,22,23.

Figure 7
figure 7

Species differences in blood-brain barrier transport. (A) Species correlation of apparent permeability in blood-brain barrier transport. (B) Species correlation of efflux ratio between in vitro primary rat cell based BBB model and in vitro iPSC-hBBB model. The dotted lines represent limits of ER.

Full size image

Discussions and Conclusions

We provide evidence of the physiological relevance of the iPSC-hBBB model in predicting human brain permeability to drugs by means of evaluation of the BBB permeability of radioligands studied correlatively in vivo with quantitative PET imaging in humans and in vitro. In addition, we report a lack of relationship between drug physicochemical properties and in vitro apparent permeability using the iPSC-hBBB model. Finally, we demonstrate interspecies differences in the prediction of drug permeability between the rodent BBB model and the iPSC-hBBB model in the prediction of drug BBB permeabilities. This is to our knowledge the first proof-of-concept study to compare transport results obtained using the iPSC-hBBB model and clinical data obtained using in vivo PET imaging.

The in vivo measurement of BBB permeation is expensive and time-consuming. Moreover, a reduction in animal experiments is desirable on ethical grounds. In addition, species differences in BBB permeability between rodent and human brain justify the development of in vitro human BBB models to avoid failure in late clinical development, when upward of $100 million is typically invested in a drug candidate. However, the integration of these human BBB models in research screening flow schemes requires knowledge of their in vivo BBB predictivity in brain permeation of drug candidates. Many in vitro human BBB models generated from iPSCs have been described recently11,12,13,14,15,16,17,18,19,20. The major hurdle associated with these models is a lack of a cross correlation between in vitro and in vivo pharmacokinetic data in demonstrating the predictive power of drug screening tools.

In an effort to achieve an in vivo like BBB phenotype, Lipmann’s protocol for the differentiation of iPSCs into BECs was modified by introducing a puromycin selection step. By doing so we produced well differentiated BECs as shown by localized expression of Claudin-5 and Zo-1, the integral membrane components of tight junctions. This overall expression of tight junction proteins in iPSC-BECs was associated with restricted permeability to the paracellular permeability marker sucrose in the iPSC-hBBB model and confirmed by the average (n = 13 independent experiments) of TEER value (458 ± 225 Ω.cm2). The mean Papp value of sucrose in the in vitro iPSC-hBBB model from eight different batches of cells in at least two replicates was 2.9 ± 1.6 × 10−6cm.s−1. The TEER values obtained in our experimental system without retinoic acid cell activation are in the same range as the current TEER standards16,19. In addition, our model displayed a differential permeability for diverse CNS and non CNS drugs. However, no relationship was evidenced between drug permeability and physicochemical properties of the compound such as PSA (R2 = −0.65), MW (R2 = −0.52), and lipophilicity (R2 = 0.20). The in vitro iPSC-hBBB model displayed a dynamic range, with Papp values ranging from 1.3 to 38.4 × 10−6 cm.s−1 for the CNS and non CNS drugs. Our data are in agreement with our previous observation using in vitro primary human model6.

Since the predictivity of human brain permeability of drug candidates by the recently developed in vitro iPSC-hBBB model has not yet been demonstrated, we compared previously published in vivo brain pharmacokinetic data of selected radioligands with their in vitro permeability using our iPSC-hBBB model. In vivo pharmacokinetic data were obtained with PET imaging, which is a powerful tool to assess BBB penetration of drugs in vivo in humans24. Brain kinetics can be analyzed by compartmental modelling, which allows calculation of the relevant rate constants that describe BBB passage. Although our study needs to be extended to a large set of radioligands with known BBB permeation in human brain, Fig. 5 reveals good agreement between the in vitro and in vivo human brain pharmacokinetic data (R2 = 0.83, P = 0.008), suggesting the predictive value of the in vitro iPSC-human BBB model in the evaluation of the brain penetration of drug candidates. While raclopride and fluoro-A85380 exhibited a moderate brain distribution, Fig. 5 reveals a low in vivo brain distribution in humans for loperamide and verapamil in comparison with befloxatone, buprenorphine and flumazenil. The transfer rate constant from plasma into brain (K1) and the distribution volume (VT) of these compounds was shown to be dependent on their interaction with efflux transporters, such as P-gp, as previously reported7,25,26,27. Discrepancies could be observed between the brain permeation of 11C-erlotinib in humans and the permeation of erlotinib across the iPSC-hBBB model. Saturation of the carrier-mediated efflux of erlotinib at the BBB has been reported in vivo in nonhuman primates using high doses of erlotinib28. It may be hypothesized that the tested concentration of erlotinib might inhibit its own transport by P-gp and BCRP across iPSC-hBBB in our experimental setting. The functional expression of P-gp was demonstrated by the use of the P-gp substrate vinblastine. The ER for this drug from 13 independent experiments was about 1.55 ± 0.40. This efflux was inhibited by cyclosporine-A suggesting the presence of functional P-gp efflux in the iPSC-hBBB model. We confirmed that the level of mRNA expression of ABCG2 and ABCC1 in iPSC-BECs was higher than for ABCB1. These results are consistent with previous observations showing that the abundance of BCRP is higher in human brain microvessels than that of P-gp29,30,31. In addition, Fig. 2C reveals mRNA expression of insulin and transferrin receptors in the same range as for primary human BECs, while the mRNA expression of caveoline 1 and caveoline 2 is 10 times lower than in BECs.

To summarize, the iPSC-hBBB model as described in this report mimics in vivo BBB integrity and compound permeability. Collectively, our findings suggest that the in vitro iPSC-hBBB model could be integrated in a flow chart screening of promising compounds for the treatment of brain diseases. In this view, miniaturization strategies will be developed for set up the models and for high throughput screening.

Materials and Methods

Drugs selection

Eight compounds with known pharmacokinetic parameters in human brain were selected: befloxatone (Synthélabo), flumazenil (Sigma), raclopride (tartrate salt, Sigma), erlotinib (Fluorochem), verapamil (Sigma Aldrich), buprenorphine (Sigma Aldrich), 2F-A85380-tartrate (Eras Labo) and loperamide (Sigma). In addition, eleven other compounds with diverse physicochemical properties were selected. We tested six non-radiolabeled compounds: dextromethorphan (hydrobromide monohydrate, Sigma Aldrich), levofloxacin (Sigma), sulfasalazine (Sigma), caffeine (Sigma Aldrich), taurocholate (sodium salt, Sigma), Lucifer Yellow (CH dilithium salt, Sigma); and five radiolabeled compounds: [U−14C]-sucrose (molar activity: 601 mCi.mmol−1), [3H]propranolol (molar activity: 24 Ci.mmol−1), [3H]verapamil hydrochloride (molar activity: 85 Ci.mmol−1), [3H]vinblastine sulfate (molar activity: 10.3 Ci.mmol−1), [3H]digoxin (molar activity: 40.0 Ci.mmol−1) were purchased from Perkin Elmer (Massachusetts, SA).

Animal use and assurances

Animal experiments were conducted in full compliance with local, national, ethical, and regulatory principles and local licensing regulations, per the spirit of the Association for Assessment and Accreditation of Laboratory Animal Care. The animal studies were carried out in accordance with decree 87–848, issued on October 19, 1987, concerning animal experimentation in France, in an approved facility (accreditation No. A91505) and by workers with Ministry of Agriculture authorization, as specified in the decree. The experimental protocol was approved by the Ethic committee for animal experiments, Life Sciences Division, Fundamental Research Department, CEA Fontenay aux Roses, France).

Human brain endothelial cells and derivation of patient-fibroblasts

Human brain endothelial cells used in this study were obtained from adult brain tissue after agreement of local ethics committee (Biomedical Agency-PFS09). All procedures were carried out in accordance with relevant guidelines and regulations. The informed consent was obtained from the legal guardian. BECs were isolated as described previously6.

iPSCs-SP derived from ifbroblast cells were a generous gift of SEPIA’s Laboratory (CEA Fontenay aux Roses).

Fibroblast cells were isolated from normal placenta according to the protocol study approved by the local ethics committee (Advisory Committee for the Protection of Persons in Biomedical Research Cochin Hospital, Paris, n°18-05). All procedures were carried out in accordance with relevant guidelines and regulations. The human samples were obtained with a written and informed consent, which was supervised and anonymized.

After 1–2 weeks, fibroblasts outgrowths from the explants were passaged and maintained as described previously32 in Dulbecco’s Modified Eagle Medium (DMEM, Life Technologies, Carlsbad, CA) supplemented with 10% inactivated fetal bovine serum (FBS, Life Technologies), 1% penicillin/streptomycin/neomycin (PSN, Life Technologies), 1% sodium pyruvate (Sigma-Aldrich, St. Louis, MO), and 1% L-glutamine (Life Technologies).

iPSC-Generation and Characterization

Reprogramming

Fibroblasts were reprogrammed using the Sendai virus reprogramming method as recommended by the manufacturer (Life Technologies). Briefly, fibroblasts were infected using SeV vectors encoding OCT3/4, SOX2, KLF4, and c-MYC on day 0. Two days later, cells were trypsinized and picked up onto two 10-cm gelatin-coated culture dishes that had been seeded with irradiated mouse embryonic fibroblasts (irrMEFs, GlobalStem). The cultures were maintained in human embryonic stem cell (hESC) medium containing DMEM/F12 (Gibco), 20% KnockOut Serum Replacement (Gibco), 1% minimum essential medium (MEM) nonessential amino acid (Gibco), 1 mM l-glutamine (Gibco), 0.1 mM β-mercaptoethanol (Sigma-Aldrich), and 10 ng/mL of basic fibroblast growth factor (β-FGF, Stem Cell technologies). Clones were picked starting on 20 days post infection, and expanded on irrMEFs before being adapted to feeder-free conditions.

hiPSC culturing

All iPSCs cell lines were maintained on hESC-qualified Matrigel (Corning®, NY, USA) in mTeSRTM1 medium (STEM CELL Technologies, Vancouver, BC, Canada) and passaged using mechanical dissociation at the beginning of the establishment of the cell line. After 7 to 10 passages, cells where passaged using ReLeSRTM (STEM CELL Technologies) and maintained until the passage 40 for the experiments.

Teratoma formation

To examine the developmental potential of reprogrammed clones in vivo, iPSCs grown on feeder free system, were collected by dispase treatment, and injected into hind limb muscle of 8-week-old immunodeficient NOD-scid IL2rγnull (NSG) mice (approximately 1 × 6-well plate at 70% confluence for each injection). After five to ten weeks, teratomas were dissected and fixed in 4% paraformaldehyde. Samples were embedded in paraffin and processed with hematoxylin and eosin staining. All studies were done with compliance with animal welfare regulations.

Assessment of pluripotency

The pluripotency characteristics of iPSC was demonstrated (i) by assessing the presence of totipotence gene expression marker by RT-PCR and (ii) by measuring their expression by flow cytometry.

  1. (i).

    Gene expression assessment. Total RNA was extracted from iPSCs using Nucleospin RNA II Kit (Macherey-Nagel) according to the manufacturer’s instructions. Reverse transcription (RT) was carried out using Iscript cDNA synthesis kit (Biorad) following the instructions provided by the manufacturer. A total of 1 µg of RNA was used in RT reactions. RT-PCR for the 5′ coding region was performed with primers specific for OCT4, SOX2, REX-1, NANOG and RPLO (housekeeping gene) (S2 Table 2).

  2. (ii).

    Flow Cytometry analysis. hiPSCs were harvested by accutase treatment for 5 to 10 min until the cells detach. Cells were then fixed in either paraformaldehyde 4% for 8 min. or in ice-cold methanol for 15 min.After two PBS1X wash, cells were blocked with 40% of goat serum in PBS1X for 20 min at room temperature. After 2 wash in 10% of goat serum in PBS1X (10%PBSG), cells were incubated for 1 h at room temperature or overnight at 4 °C with primary antibodies diluted in 10%PBSG. IgG controls were used at the same concentration. Cells were then washed twice in 1.5 ml of PBS1X-5%SVF and resuspended in 500 µL of FACS Flow for analysis on a FACS calibur flow cytometer. References and working dilutions of antibodies are plotted in S3 Table 3.

Derivation of human brain endothelial cells from iPSCs (hiPSC-BECs)

Human iPSCs were differentiated as described previously15 with substantial modifications. Briefly, cells were passaged using ReLeSRTM onto Matrigel and maintained in mTeSR1 medium for 2–3 days. After expansion, cells were switched on unconditioned medium (DMEM/Ham’s F-12, GlutaMAXTM supplement (Life Technologies) containing 20% KnockOut Serum Replacement (Life Technologies), 1% MEM Non-Essential Amino-Acid Solution (Life Technologies) and 0.055 mM β-mercapto-ethanol (Life Technologies)) for 6 days. Cells were cultivated for two additional days in endothelial medium (Human Serum Free Medium (hSFM, Gibco) supplemented with 20 ng/mL β-FGF (Life Technologies) and 1% of platelet-poor plasma derived bovine serum (AlfaAesar, Ward Hill, USA)). To improve the selection of endothelial cells, 1 µg/mL of puromycin (Sigma) was added during the first day of endothelial medium step. At day 8, cells were then gently dissociated with Accumax (Sigma) for 5–8 min at 37 °C and plated onto 24-well tissues culture polystyrenes plates or 1.12 cm2 Transwell® permeable inserts (0.4 µm pore size) coated firstly for 2 hours with collagen IV 100 µg/mL (Sigma) and then for 1 hour with fibronectin 10 µg/ml (Sigma). Endothelial cells were then grown in endothelial medium for co-culture experiments or immunohistochemistry.

Transcription profiles of tight junctions, efflux transporters and other receptors

To evaluate quality of differentiation of hiPSCs-BECs, the mRNA expression profile of specific endothelial markers was assessed. RNA was isolated from hiPSC-BECs or iPSCs using the RNeasy Mini kit (Qiagen), according to the manufacturer’s instructions. The concentration and purity of RNA samples were checked spectrophotometrically at 260 and 280 nm using the NanoDrop ND-1000 instrument (NanoDrop Technologies, Wilmington, DE, USA); the A260/280 ratio ranged from 1.8 to 2. A sample of 0.5 μg of total RNA was converted to cDNA with random primers in a volume of 20 μL using an RT2 HT first stand kit (Qiagen) according to the manufacturer’s protocol. cDNA was diluted with water to a final volume of 200 μL. Quantitative expression of markers was determined using 2 µL of diluted cDNA for each primer (S4 Table 4) set at 10 µM using iQ SYBR Green Supermix (Biorad) in a final volume of 12 µL. Thermocycling was carried out in CFX96 RT-PCR detection system (Bio-Rad) using SYBR green fluorescence detection. The specific amplification conditions were 10 min at 95 °C, followed by 40 amplification cycles at 95 °C for 15 s, 60 °C for 60 s and 72 °C for 30 s to reinitialize the cycle again. The specificity of each reaction was also assessed by melting curve analysis to ensure the presence of only one product. Relative gene expression values were calculated as 2−ΔCT, where ΔCT is the difference between the cycle threshold (CT) values for genes of interest and housekeeping gene (GAPDH).

iPSC-derived human BBB model

On day before the co-culture establishment, rat glial cells (2 × 104 cells) isolated as described previously6 were plated on 12 wells plate (Costar) in glial specific medium. After 24 hours, hiPSC-BECs (6 × 104cells) were plated on the upper side of a collagen-coated polyester Transwell membrane (Costar, pore size 0.4 µm; diameter 12 mm; insert growth area 1 cm2) in 0.5 mL of endothelial medium. Under those conditions, the hiPSC-BECs monolayer were confluent after 4 to 7 days of co-culture. The integrity of the cell monolayer was demonstrated (i) by assessing the Transendothelial Electrical Resistance (TEER), (ii) the assessment of presence of tight junctions between the endothelial cells by immunocytochemistry, (iii) by measuring the flux of the paracellular reference marker, [14C]sucrose, through the monolayer and (iv) by assessment of hiPSC-BECs polarity by measuring the efflux of P-gp substrates, i.e. vinblastine.

For TEER, cell culture inserts were transferred to an Endohm-6 chamber (Word Precision Instruments, Sarasota, FL) containing endothelial medium pre-warmed at 37 °C and resistance of the monolayer was measured using an EVOM2 Epithelial Volt Meter (World Precision Instruments). iPSC-hBBB models were considered mature when TEER measurements were stable and reached at least a mean of 170 ohm/cm2.

For immunocytochemistry, hiPSC-BECs monolayers were washed once in PBS1X and fixed in 4% paraformaldehyde for 8 min. After washing with PBS1X, cells were then soaked in a blocking solution containing 5% of BSA. Primary antibodies and isotype controls were diluted in the same solution and incubated on the monolayer for 1 hour at room temperature or overnight at 4 °C. After three washes with PBS1X, the cells were incubated for 30 min in the presence of secondary antibodies: goat anti-rabbit Alexa Fluor 594 (1:500; Invitrogen) or goat anti-mouse Alexa Fluor 388 (1:500; Invitrogen) diluted in blocking solution. References and working dilutions of antibodies are summarized in S2 Table 2. Cell nuclei labelling was performed at 1400 nM 4′,6-Diamidino-2-pheny-lindoldihydrochloride (DAPI) for 10 min and then washed 3 times in PBS1X. The preparations were observed with a fluorescence microscope and pictures were processed on ImageJ (NIH).

The integrity of iPSC-derived human BBB model was determined by assessment of radiolabelled sucrose permeability as a paracellular marker. The transwells with hiPSC-BECs monolayers were transferred to new 12 well plates. Transfer buffer (150 mM NaCl, 5.2 mM KCl, 2.2 mM CaCl2, 0.2 mM MgCl2, 6 mM NaHCO3, 2.8 mM glucose and 5 mM Hepes) pre-warmed at 37 °C was added: 1.5 mL to the basolateral compartment (B) and 0.5 mL to the apical compartment (A) containing 0.1µCi/mL (166-300 nM) of [14C]sucrose (Perkin Elmer). After 1 hour of incubation at 37 °C with gentle rotation, 100 µL of transfer buffer from both A and B compartments were collected and counted by scintillation. The Papp value was calculated as followed:

$${{\rm{P}}}_{{\rm{app}}}={\rm{dQ}}/{\rm{dT}}\,{\rm{X}}\,{\rm{A}}\,{\rm{X}}\,{{\rm{C}}}_{{\rm{0}}}$$

where dQ/dT is the amount of compound transported per time point; A is the membrane surface area of the insert; and Co is the donor concentration at the time point. Data were presented as the average ± SD from three monolayers. Monolayers were validated for sucrose permeability from A to B when the Papp was below 8 × 10−6 cm.s−1. The propranolol permeability was also tested using the same protocol with an apical tested dose about 1 µCi/mL (48 nM) of [3H]propranolol in transfer buffer. Monolayers were validated for propranolol permeability from A to B when the Papp was under 16 × 10−6 cm.s−1. For the activity of efflux pumps and the asymmetry of transporters distribution of the endothelial cells, we measured the efflux of [3H]vinblastine or [3H]verapamil across the cell monolayer. 0.1 µCi.mL−1 of [3H]vinbastine (4.8-9.7 nM) or [3H]verapamil (1.25 nM) were applied in the apical or basal compartment and the Papp for the passage of compartment A to B (Papp A>B) and B to A (Papp B>A) were calculated. The efflux ratio (ER) was calculated as follows: ER = Papp B>A/Papp A>B. The specificity of this efflux was confirmed by inhibition of vinblastine efflux with the P-gp inhibitor cyclosporine-A at low concentration (1.3 µM).

Permeability studies

After checking the cell monolayer integrity, the monolayers were transferred to new plates. The buffer (150 mM NaCl, 5.2 mM KCl, 2.2 mM CaCl2, 0.2 mM MgCl2, 6 mM NaHCO3, 2.8 mM glucose and 5 mM Hepes) was added: 1.5 mL to the basal chamber (B) and 0.5 mL to the apical chamber (A). Experiments were performed in triplicate for each compound. The tested compounds (5–10 μM) were introduced in the donor chamber (either the apical or the basal compartment). After 60 min, aliquots were removed from the acceptor and basal chambers for drug-concentration determination by LC-MS/MS.

For LC-MS/MS a sample aliquot (transfer buffer) was diluted by half with acetonitrile and then, 10 μL of this solution was injected into the analytical equipment made up a Liquid Chromatography system coupled to a triple quadrupole mass spectrometer Thermofisher TS Quantum Discovery or Waters Quattro Premier XE equipped with a electrospray ionization source. System control and data processing were carried out using Xcalibur or Masslynx software. The chromatographic separation was performed on a Waters ACQUITY UPLC BEH C18 column (2.1 mm × 50 mm, 1.7 μm) at 60 °C. A gradient of two solvent mixtures was delivered at 0.3 mL/min or 0.5 mL/min according to the analytical system. Solvent A comprised 0.1% formic acid in ammonium acetate (5 mM) and solvent B, 0.1% formic acid in acetonitrile. Molecules were ionized in positive mode. Multiple reaction monitoring mode was used for compounds quantification. For all tested compounds, a linear gradient of mobile phase A and B was optimized. Linearity of the response was verified in the range 0.05 to 5 µM through 3 experiments to be able to measure low passage. Quantification was carried out using standard calibration curve and quality control samples prepared with blank matrix (transfer buffer/acetonitrile; 1/1; v/v) spiked with compounds.

Statistical analysis

Statistical analysis was performed using the GraphPad Prism 5.04 program (GraphPad Software, Inc., San Diego, CA). Regression lines were calculated and correlation was estimated a nonparametric Spearman test. Statistical significance was set at p < 0.05.