Non-Human Primate Blood–Brain Barrier and In Vitro Brain Endothelium: From Transcriptome to the Establishment of a New Model

The non-human primate (NHP)-brain endothelium constitutes an essential alternative to human in the prediction of molecule trafficking across the blood–brain barrier (BBB). This study presents a comparison between the NHP transcriptome of freshly isolated brain microcapillaries and in vitro-selected brain endothelial cells (BECs), focusing on important BBB features, namely tight junctions, receptors mediating transcytosis (RMT), ABC and SLC transporters, given its relevance as an alternative model for the molecule trafficking prediction across the BBB and identification of new brain-specific transport mechanisms. In vitro BECs conserved most of the BBB key elements for barrier integrity and control of molecular trafficking. The function of RMT via the transferrin receptor (TFRC) was characterized in this NHP-BBB model, where both human transferrin and anti-hTFRC antibody showed increased apical-to-basolateral passage in comparison to control molecules. In parallel, eventual BBB-related regional differences were Investig.igated in seven-day in vitro-selected BECs from five brain structures: brainstem, cerebellum, cortex, hippocampus, and striatum. Our analysis retrieved few differences in the brain endothelium across brain regions, suggesting a rather homogeneous BBB function across the brain parenchyma. The presently established NHP-derived BBB model closely mimics the physiological BBB, thus representing a ready-to-use tool for assessment of the penetration of biotherapeutics into the human CNS.


Introduction
The blood-brain barrier (BBB) constitutes the main interface of molecular exchange between the periphery and the central nervous system (CNS) [1]. Given its extremely restrictive properties, the BBB plays a key role over the bi-directional passage of endogenous and exogenous compounds, and ultimately constitutes a formidable challenge for the entry of therapeutics into the brain [2]. In order to find new brain-specific mechanisms of transport and efficient CNS-active therapeutic alternatives, it is imperative to develop strategies to deepen the understanding of the BBB. Most in vitro BBB models derive from isolated brain capillary endothelial cells, from primary or immortalized cell line source, and rely on brain tissue from rat, mouse, porcine, bovine, and humans [3]. However, the underlying interspecies differences query the reliability and prediction capacity of most in vitro BBB models for the passage of therapeutic molecules across the human BBB and into the CNS [4,5]. (A) Diagram of the procedure conducted for the isolation of NHP-derived brain microvessels and in vitro selection of brain endothelial cells (BECs). (B1) Principal Component Analysis (PCA) of brain cortical microvascular samples clustered in 3 groups (P0D0 n = 8, P0D7 n = 8, P1D7 n = 6). (B2) PCA of brain endothelial samples clustered in 5 groups (cortex n = 8, brainstem n = 6, hippocampus n = 7, striatum n = 7, cerebellum n = 2). (C1) Comparison of the expression levels of endothelial cell and nonendothelial cell markers in the brain microvascular P0D0 and P0D7 fractions. An expression threshold of one FPKM in at least half of the independent replicates (P0D0, n = 8; P0D7, n = 8; P1D7, n = 6; FPKM ≥ 1 in n ≥ 4 samples) was applied in order to consider only the differentially expressed gene having a significant expression level. Gene levels of brain endothelial cell markers are enriched in P0D0 and P0D7 samples, and even though RNA levels of markers for microglia (or immune cells), oligodendrocytes, pericytes, astrocytes and neurons can be detected in P0D0 samples, their expression is off in P0D7. Results represent average FPKM and are expressed and mean ± SEM. (C2) Hierarchical clustering of BECs from the selected brain structures showing a lack of clustering per brain structure. An expression threshold of one FPKM in at least 3 samples was applied to ECs derived from different regions (cortex n = 8, brainstem n = 6, hippocampus n = 7, striatum n = 7, cerebellum n = 2) for the analysis of differentially expressed genes. (A) Diagram of the procedure conducted for the isolation of NHP-derived brain microvessels and in vitro selection of brain endothelial cells (BECs). (B1) Principal Component Analysis (PCA) of brain cortical microvascular samples clustered in 3 groups (P0D0 n = 8, P0D7 n = 8, P1D7 n = 6). (B2) PCA of brain endothelial samples clustered in 5 groups (cortex n = 8, brainstem n = 6, hippocampus n = 7, striatum n = 7, cerebellum n = 2). (C1) Comparison of the expression levels of endothelial cell and non-endothelial cell markers in the brain microvascular P0D0 and P0D7 fractions. An expression threshold of one FPKM in at least half of the independent replicates (P0D0, n = 8; P0D7, n = 8; P1D7, n = 6; FPKM ≥ 1 in n ≥ 4 samples) was applied in order to consider only the differentially expressed gene having a significant expression level. Gene levels of brain endothelial cell markers are enriched in P0D0 and P0D7 samples, and even though RNA levels of markers for microglia (or immune cells), oligodendrocytes, pericytes, astrocytes and neurons can be detected in P0D0 samples, their expression is off in P0D7. Results represent average FPKM and are expressed and mean ± SEM. (C2) Hierarchical clustering of BECs from the selected brain structures showing a lack of clustering per brain structure. An expression threshold of one FPKM in at least 3 samples was applied to ECs derived from different regions (cortex n = 8, brainstem n = 6, hippocampus n = 7, striatum n = 7, cerebellum n = 2) for the analysis of differentially expressed genes.
RNA-seq data analysis was performed using ArrayStudio (Qiagen, Courtaboeuf, France). Briefly, raw data QC is performed then a filtering step is applied to remove reads corresponding to rRNAs as well as reads having low quality score or shorter than 25 nt. Reads were further mapped to the Cyno.WashU 2013 genome, based on the contigs assembled from a WGS project submitted by Washington University (WashU) in 2013, using OSA4 [12] (Omicsoft Sequence Aligner, version 4, Qiagen) and quantified using Ensembl.R94 model of transcriptome, paired reads were counted at gene level. Differential analysis of gene expression was performed at gene level using DESeq2 [13]. The variable multiplicity was taken into account and false discovery rate (FDR) adjusted p-values calculated with the Benjamini-Hochberg (BH) correction [14]. Functional analysis of differentially expressed genes was performed using IPA (Qiagen) [15]. Gene Set Enrichment analysis was performed using Ingenuity and GSEA software [16,17]. High-throughput sequence data are available on the Gene Expression Omnibus (GEO) under the GSE154901 accession number. Read depth from RNAseq was counted as fragments per kilobase per million mapped fragments (FPKM). FPKM read counts were obtained for each BEC fraction (P0D0, P0D7, P1D7). A minimum of 1 FKPM in at least half of the observations per group, and in at least one of the groups, was required for a gene to be considered as expressed and included in the analysis.

Resistance Measurements
Trans-endothelial electrical resistance (TEER) was measured every 24 h following the sub-culture of BMECs onto transwell filters. Resistance was recorded using an EVOM2 epithelial voltohmmeter coupled to a cell culture cup chamber (ENDOHM-12G) (World Precision Instruments, Hitchin, Hertfordshire, United Kingdom). TEER values were presented as Ω × cm 2 following the subtraction of an un-seeded Transwell and multiplication by 1.12 cm 2 to account for the surface area. TEER measurements were taken three independent times on each sample and at least in triplicate filters for each experimental condition.

In Vitro Transcytosis and Permeability Measurements
Test (2 µg/mL human holo-Transferrin or 1 µg/mL of internally developed anti-human/cynomolgus TFRC antibody) and control molecules (10 µg/mL of FITC-coupled 70 kDa Dextran or 1 µg/mL mouse IgG, clone MG1-45, BioLegend) were added onto the upper chamber on day 4 (96 h following NHP-derived BMEC seeding on transwell). Fresh EC medium with none of these compounds was added onto the bottom chamber. Final aliquots from both chambers were taken 240 min following incubation at 37 • C, 5% CO 2 . Compound levels in mother solutions (T = 0 min), upper and lower compartments (T = 240 min) were determined by electrochemiluminescent immunoassay (MESOQuickPlex SQ120, MesoScale Discovery, Rockville, MD, USA) and by fluorescence measurement (485 nm excitation and 530 nm emission, using a multiwell-plate reader). Permeability coefficients were calculated based on the cleared volume of Dextran or holo-Transferrin from the top chamber to the bottom chamber using the following formula: where V = volume of cell medium in the bottom chamber (mL), A = surface area of the insert (cm 2 ), C luminal = compound concentration loaded in the upper chamber (µM), C abluminal = compound concentration measured in the bottom chamber (µM); t = time of the assay (min).

Western Blot
NHP-derived BECs were lysed using ice-cold RIPA buffer (CST) containing protease inhibitor cocktail (ThermoFisher). Lysates were sonicated for 5 min, centrifuged at 15,000× g for 15 min, and supernatant fractions were collected. These were loaded into 4-12% Tris-Glycine SDS-page gels (Invitrogen, Illkirch-Graffenstaden, France), and let to migrate for 1h at 180V. Samples were then transferred onto PVDF membranes using an iBlot™ 2 Dry Blotting System (Invitrogen) on the P0 program (20 V for 1 min, 23 V for 4 min, 25 V for 2 min). PVDF membranes were then rinsed with Tris-buffered saline with 0.1% Tween 20 (TBST) and blocked for 1 h in 5% non-fat dry milk in TBST (blocking buffer). Membranes were first probed overnight at 4 • C with primary antibodies in blocking buffer, and then probed with secondary antibodies diluted in TBST for 1 h at room temperature (1:10,000 diluted HRP-coupled goat anti-mouse IgG or goat anti-rabbit IgG, GE Healthcare). Following secondary antibody incubation, membranes were rinsed thoroughly with TBST, imaged using a LICOR Odyssey Imager and bands quantified using Multi-Gauge v3.0.

Statistical Analysis
The statistical significance of differences between groups was analyzed using GraphPad Prism v7.0 software (GraphPad Software, San Diego, CA, USA) t-test (unpaired, parametric, equal standard deviation, two-tailed), Mann-Whitney t-test (unpaired, nonparametric, two-tailed) and Friedman test (matched data, nonparametric) with post hoc Dunn's multiple comparison test. The application of these statistical methods to specific experiments is noted in the figure legends.

Transcriptional Profiling of Brain Microvasculature
For the detection of differentially expressed genes, we performed statistical analysis with DESeq2 [13]. Transcriptional profiles of the different cortical EC populations were compared to one another, while ECs from different brain regions were compared to those obtained from brain cortex, always using a threshold of 1.5-fold difference in the expression and a BH-adjusted p-value (p adj ≤ 0.05).
In the principal components analysis (PCA), a clustering of independent replicates is seen for each fraction, supporting a low variability between independent samples (Figure 1(B1)). The PCA of the different endothelium transcriptomes also evidences that brain ECs set in culture (P0D7, P1D7) have molecular signatures that are relatively distinct from the freshly obtained BEC-enriched fraction P0D0 (Figure 1(B1)). Such difference of P0D7 and P1D7 towards the P0D0 fraction is partially due to the presence of RNA from cell types other than endothelial cells in the P0D0 fraction. In fact, the enzymatic digestion and gradient centrifugation procedure used for obtention of brain microvessels (P0D0) yields a highly enriched fraction yet presents RNA of pericyte, glial, neuronal, and circulating blood cell origin. This also explains the important set of genes seeing their expression downregulated in P0D7 vs. P0D0 (5089 out of a 13,485 gene population, see Figure S1). On the other hand, an up-regulation of 3884 out of 13,485 genes is seen from P0D0 to P0D7 (≥1.5-fold, p ≤ 0.05), most probably representing endothelial-specific genes, and thus reflecting the in vitro selection of the endothelial cell population. The degree of purity of P0D0 and in vitro-selected P0D7 and P1D7 samples was assessed through probing the transcriptome data for the expression of 5 acknowledged cell type specific genes [18][19][20][21]  All BEC samples from five brain regions (brainstem, cerebellum, cortex, hippocampus, striatum) qualified according to the raw data quality control analysis, with the exception of 2 cerebellum with low RNA levels (consequently excluded from the subsequent analyses). The PCA analysis and hierarchical clustering suggest low variability between independent samples (Figure 1(B2)), and similar molecular signatures of endothelium transcriptomes from different brain structures as no distinct clustering of samples belonging to the same structure is formed (Figure 1(C2)). Accordingly, only 173 genes were found to be differentially expressed (fold-change > ±1.5, p < 0.05) in BECs from other brain structures vs. those from brain cortex ( Figure S2).

Endothelial Junction Mediators: Expression Levels and Transcriptional Changes
The present RNAseq analysis evidences CLDN5 as the major claudin transcript present in the cynomolgus brain vasculature, but CLDN1, CLDN12, and CLDN15 expression at the 3 different in vitro stages is also detected, at considerably lower levels ( Figure 2A). CLDN3 and CLDN6, previously reported at the mouse BBB [22,23] were not detected. Other key participants in the formation of tight junctions at the BBB, in particular JAMs (F11R, JAM3), TJP1 (ZO-1), and cadherins CDH5, CDH2, and ESAM seem to maintain high levels of expression in isolated BECs, even after sub-culturing in vitro (Figure 2A). JAM2 and to a lesser extent OCLN are progressively downregulated from P0D0 to P1D7, but present nonetheless relevant levels of expression. The expression of CTNNB1 (β-catenin), regulator of CNS-specific angiogenesis and of the expression of tight junctions at the BBB [23][24][25][26], is dominant over the three endothelial fractions, along with other genes associated to tight junction correct assembly and maintenance, such as USP53, PDCD6IP, ARHGEF2, MPDZ, and AMOTL1-2 ( Figure 2A), further sustaining the robustness of the NHP-derived BBB model. Lipolysis stimulated receptor (LSR), although downregulated (3.5-fold, p < 0.001, P0D7 vs. P0D0), together with stable levels of CGNL1 MARVELD2 found in vitro, which are known to participate in the formation of epithelial and endothelial junctions, and enriched in mouse BECs versus periphery [27,28] were also identified in our analysis.
Pharmaceutics 2020, 12, x FOR PEER REVIEW 8 of 26 correct assembly and maintenance, such as USP53, PDCD6IP, ARHGEF2, MPDZ, and AMOTL1-2 ( Figure 2A), further sustaining the robustness of the NHP-derived BBB model. Lipolysis stimulated receptor (LSR), although downregulated (3.5-fold, p < 0.001, P0D7 vs. P0D0), together with stable levels of CGNL1 MARVELD2 found in vitro, which are known to participate in the formation of epithelial and endothelial junctions, and enriched in mouse BECs versus periphery [27,28] were also identified in our analysis. Most major participants in the regulation and formation of cellular junctions, such as CDH5, TJP1, or OCLN, did not see their expression levels differ between endothelial cells obtained from the brain microvasculature of the five analyzed regions, except for increased transcript levels of CLDN5 and CLDN1 found in BECs from the brainstem and cortex vs. hippocampus (p < 0.05).  Most major participants in the regulation and formation of cellular junctions, such as CDH5, TJP1, or OCLN, did not see their expression levels differ between endothelial cells obtained from the brain microvasculature of the five analyzed regions, except for increased transcript levels of CLDN5 and CLDN1 found in BECs from the brainstem and cortex vs. hippocampus (p < 0.05).

Receptor-Mediated Transcytosis
RMT constitutes a reference strategy to overcome the BBB and promote efficient and specific delivery into the CNS of therapeutic agents. TFRC, highly expressed and enriched in the brain microvasculature [29,30], and the only validated RMT-mechanism evidencing an enhanced brain exposure of biotherapeutics [31][32][33][34][35], sees its expression levels being downregulated from P0D0 to P1D7, but retaining substantial expression levels in vitro ( Figure 3A). Insulin receptor (INSR) [36][37][38][39], LRP1 [37,[40][41][42], IGF1R [43], LDLR [44], basigin (BSG), and SLC2A1/GLUT1 [37] have been targeted using antibodies and/or peptides as attempts to improve brain uptake, but such studies did not generate much encouraging results. Shuttles targeting novel receptors have also been described, such as SLC3A2/CD98hc [37] and TMEM30 [45,46], yet poorly characterized. A sustained relevant expression from P0D0 to P1D7 is seen for the transcript levels of most of these receptors, except for LRP1, which is considerably suppressed in vitro (Figure 3(A1)). The absent expression of LRP1 is coherent with previous findings revealing strong LRP1 expression levels found in pericytes [47,48] but not in selected brain endothelial cells [47,49]. TMEM30A and IGF2R levels see a relative increase in vitro, likely due to the endothelial cell selection in culture. Expression levels of TFRC did not differ between endothelial cells from the different analyzed brain structures ( Figure 3B). Increased IGF2R and TMEM30A transcript levels were found in the brainstem-derived endothelium vs. cortex-derived endothelium (p < 0.05), while INSR showed to be significantly inferior in hippocampus vs. brainstem (≥1.5-fold change, p < 0.05, vs. striatum and cortex (≥1.5-fold change, p > 0.05).

Carrier Transporters
Comparison of Cortical Brain Endothelial P0D0, P0D7 and P1D7 Fractions The presence of cell membrane transporters at the BBB allows a perfect control of the flux of molecules between the blood and the brain, either by efflux ATP-binding cassette (ABC) transporters (e.g., ABCB1/MDR1/P-glycoprotein) [57], or highly specific solute carriers (SLC) (e.g., SLC2A1/GLUT1) [58]. Expression trends of such transporters were therefore investigated across the three studied cell fractions. Based on the overall samples, we detected a total of 398 transporters, 249 of these detectable at ≥1 FPKM in at least half of the samples from at least one of the analyzed groups.
ABC transporters: Known major drug efflux transporters, like ABCG2 (BCRP) and ABCB1, presented the highest expression levels among the 33 ABC transporters found in the three studied endothelial cell populations. ABCG2 and ABCB1 transporters were downregulated in vitro, but nonetheless remained the most expressed ABC transporters in the P0D7 and P1D7 sampling groups ( Figure 4A). ABCC4 expression levels showed to be third highest among ABC transporters in the P0D0 fraction, but its expression is considerably depleted in vitro. Similarly, ABCC6 and ABCC9 were detected in freshly isolated brain microvessels, followed by downregulation in vitro. ABCC9, but not ABCC6, have previously been reported to be present at the BBB microvasculature [59,60], but it likely originates from pericytes [61]. On the other hand, expression of ABCC1 is considerably upregulated from P0D0 to P0D7 and P1D7 culture conditions (despite controversial prior reports [62][63][64]), and similar trends are also observed for ABCB9 (reported in rat brain microvessels [60], ABCC3 and the lipid transporter ABCG1 (reported in porcine brain endothelium [65]). The upregulation of these 5 ABC transporters versus P0D0 is most likely due to the positive selection of endothelial cells in

Carrier Transporters
Comparison of Cortical Brain Endothelial P0D0, P0D7 and P1D7 Fractions The presence of cell membrane transporters at the BBB allows a perfect control of the flux of molecules between the blood and the brain, either by efflux ATP-binding cassette (ABC) transporters (e.g., ABCB1/MDR1/P-glycoprotein) [57], or highly specific solute carriers (SLC) (e.g., SLC2A1/GLUT1) [58]. Expression trends of such transporters were therefore Investig.igated across the three studied cell fractions. Based on the overall samples, we detected a total of 398 transporters, 249 of these detectable at ≥1 FPKM in at least half of the samples from at least one of the analyzed groups.
ABC transporters: Known major drug efflux transporters, like ABCG2 (BCRP) and ABCB1, presented the highest expression levels among the 33 ABC transporters found in the three studied endothelial cell populations. ABCG2 and ABCB1 transporters were downregulated in vitro, but nonetheless remained the most expressed ABC transporters in the P0D7 and P1D7 sampling groups ( Figure 4A). ABCC4 expression levels showed to be third highest among ABC transporters in the P0D0 fraction, but its expression is considerably depleted in vitro. Similarly, ABCC6 and ABCC9 were detected in freshly isolated brain microvessels, followed by downregulation in vitro. ABCC9, but not ABCC6, have previously been reported to be present at the BBB microvasculature [59,60], but it likely originates from pericytes [61]. On the other hand, expression of ABCC1 is considerably upregulated from P0D0 to P0D7 and P1D7 culture conditions (despite controversial prior reports [62][63][64]), and similar trends are also observed for ABCB9 (reported in rat brain microvessels [60], ABCC3 and the lipid transporter ABCG1 (reported in porcine brain endothelium [65]). The upregulation of these 5 ABC transporters versus P0D0 is most likely due to the positive selection of endothelial cells in culture, and not a cell derivation in vitro, given that these very same five ABC transporters showed to be positively correlated with the trend of PECAM1 expression across the three endothelial cell fractions (Pearson correlation, p < 0.001, Figure S5). Other transporters which showed considerable expression at P0D0 and for which their expression was relatively maintained in vitro are the cholesterol efflux transporters ABCA1, ABCA2 (reported in human brain microvessels [66,67]) and ABCA3, the multidrug transporter ABCC5, but also those of ABCD3 [60], ABCE1 (expression associated with cell growth and apoptosis inhibition) and ABCF1 (Figure 4A), for which the eventual played role at the BBB is absolutely unknown. A global heatmap plot representation of a hierarchical clustering of the FPKM variability of ABC transporters expressed genes across fractions is available in Figure S4.
SLC transporters: The SLC superfamily comprises known transporters of energy metabolites (e.g., SLC2A, SLC16A families), amino acids and neurotransmitters (e.g., SLC1A, SLC7A families), ions (e.g., folate/thiamine transporter family SLC19A, zinc transporter family SLC39A), organic anions (SLCO family), organic cations (SLC22A family), and others. Transporters such as SLC2A1 (GLUT1) [68,69], SLC7A5 (LAT1) [70,71], SLC7A1 (CAT1) [72], or SLCO2B1 (OATP2B1) [73,74] are known to participate in mediated-transport at the BBB. The analysis of the transcriptomic data indicates that the glucose transporter SLC2A1, the amino acid transporters SLC7A5, SLC7A1 and SLC3A2 (the latter known to form heterodimers with SLC7 family members, e.g., with SLC7A5 to form the neutral large amino acid transporter LAT-1), SLCO2B1, but also SLC38A5 [75,76] (SNAT5, carrier for glycine, alanine, serine, glutamine, asparagine, and histidine) are the most abundantly expressed SLC transporters in P0D0 ( Figure 4B). Their expression was downregulated in P0D7 and P1D7 endothelial fraction yet maintaining relevant levels of expression as they remained the most expressed SLCs. SLC7A5 and SLC38A5 although known to be prominently expressed at the BBB [76][77][78] represented important exceptions as their expression dramatically dropped from an average 339 or 169 FPKM, respectively, at P0D0 to approximately 6 and 2 FPKM, respectively, at P1D7. As SLC7A5 heterodimerizes with SLC3A2 to function as a sodium-independent, high-affinity transporter mediating uptake of large neutral amino acids such as phenylalanine, tyrosine, L-DOPA, leucine, histidine, methionine and tryptophan, its modest expression might limit the formation of such complexes and partially compromise LAT-1 activity. Expression of SLC7A11 (xCT, cystine/glutamate antiporter, also forming dimers with SLC3A2) is considerably upregulated in vitro, finding that has already been described [75], and which is associated with increased oxidative stress conditions [79]. Given the similar transcript levels of SLC3A2 and SLC7A11 found at P0D7-P1D7, formation of this heterodimer in vitro should be sustained. Some SLC38A transport family members (Na + -dependent transport of neutral amino acids), namely SLC38A1 (transport of glutamine), SLC38A2, and SLC38A10 (proton antiporter, orphan transporter) are mainly upregulated in vitro, and account for high levels of expression among SLCs, notably SLC38A2 ( Figure 4B). Other simple sugar transporters from the SLC2A family were found to be significantly expressed in the cynomolgus brain microvascular enriched fraction, such as SLC2A5 (corroborated by [80]) and SLC2A13-14. Although the expression of SLC2A5 was also depleted in vitro, an important expression of both SLC2A10 and SLC2A14 was found at both P0D7 and P1D7. The expression levels of some SLC12A (cation-coupled chloride transporters)-SLC12A2, SLC12A4, SLC12A6-and SLC16A (proton-coupled monocarboxylate uptake, such as lactate and pyruvate, but also drugs, including statins) members-SLC16A1, SLC16A3, SLC16A14-were also among the most relevant SLC transporter expressed at the cynomolgus BBB, coherent with previous reports [81,82], and whose levels of expression were maintained in vitro. On the other hand, even though a broad expression level was found at P0D0 for SLC16A2, known to carry cellular import of thyroid hormones, its expression was substantially downregulated across both P0D7 and P1D7 fractions. High levels of expression of SLC25A transporter family members (mitochondrial carriers, e.g., exchange of ADP/ATP, citrate, phosphate), namely SLC25A1, SLC25A3, SLC25A4, SLC25A5, were also observed from P0D0 to P0D7/P1D7 conditions ( Figure 4B). They account for nine out of 25 SLC transporter expression positively correlated with the expression of PECAM1 across the three endothelial cell populations ( Figure S5).
Other SLC transporters found to be present across the three endothelial cell fractions are the zinc efflux family SLC30A (SLC30A1, SLC30A9) and the zinc uptake family SLC39A (SLC39A1, SLC39A7, SLC39A9, and SLC39A13), suggesting the importance of the fine tuning of the zinc intracellular concentration levels for the cellular homeostasis and proper functioning. Our present findings allow to better establish a set of BBB highly expressed transporters which may provide novel insights into its role for the maintenance of the CNS homeostasis and nutrient requirements, but also for identifying novel molecules to target to enhance drug delivery into the CNS. A global heatmap plot representation of a hierarchical clustering of the FPKM variability of SLC transporters expressed genes across the 3 fractions can be consulted in the ( Figure S4).
across the three endothelial cell fractions are the zinc efflux family SLC30A (SLC30A1, SLC30A9) and the zinc uptake family SLC39A (SLC39A1, SLC39A7, SLC39A9, and SLC39A13), suggesting the importance of the fine tuning of the zinc intracellular concentration levels for the cellular homeostasis and proper functioning. Our present findings allow to better establish a set of BBB highly expressed transporters which may provide novel insights into its role for the maintenance of the CNS homeostasis and nutrient requirements, but also for identifying novel molecules to target to enhance drug delivery into the CNS. A global heatmap plot representation of a hierarchical clustering of the FPKM variability of SLC transporters expressed genes across the 3 fractions can be consulted in the supplementary data ( figure S4). Comparison of Brain Microvascular Endothelial Cells from 5 Brain Regions at P0D7

Comparison of Brain Microvascular Endothelial Cells from 5 Brain Regions at P0D7
No ABC transporter genes were found to be differentially expressed in the brain endothelium from either brainstem, cerebellum, hippocampus, or striatum in comparison to that from brain cortex. As in the cortical-derived brain endothelium, ABCG2, ABCB1, and ABCE1 were the most expressed ABC transporters at the BBB from the analyzed four brain regions ( Figure 5A). ABCG2 levels displayed a tendency to be higher in the brainstem and striatum when opposed to those from cortex and hippocampus. ABCA6 expression levels also showed trendily increased in hippocampus-derived endothelial cells versus brain cortex, although with no statistical difference (≥1.5-fold change, p > 0.05). (C) Representation of the 6 SLC transporters for which there is a differential expression (p < 0.05) in BECs cultivated from a given brain structure in comparison to another brain region.

Function of Transferrin Receptor-Mediated Transcytosis
Following puromycin purification in vitro during the P0 selection, NHP-derived BECs at P1 evidenced the formation of a tight monolayer, continuous and with growth inhibition at confluence. transporters for which there is a differential expression (p < 0.05) in BECs cultivated from a given brain structure in comparison to another brain region. The 16 SLC transporters found to be differentially expressed across the endothelium from the 5 brain structures are represented in Figure 5C. Among the SLC transporters with the highest expression levels across the five studied structures are SLC2A1, SLCO2B1, SLC38A2, SLC3A2, SLC2A14, and SLC20A1 ( Figure 5B). Endothelial cells derived from the hippocampal BBB were shown to be richer in SLC12A2 and SLC25A46 in comparison to cortex, but poorer in SLC5A3 (p < 0.05). Levels of expression of SLC7A1 and SLC16A1 were found to be lower in BECs from both brainstem and striatum in comparison to brain cortex, while levels of SLC11A2 were shown to be higher in brainstem vs. brain cortex endothelial cells (p < 0.05).
Cerebellum-derived brain endothelial cell transcript levels of SLC3A2, SLC1A5, SLC7A1, SLC7A5, SLC6A9, and SLCO4A1, although with no statistical value given its size sample (n = 2), showed to be increased than those found in the brainstem, striatum, and hippocampal BBB. The expression of other major SLC transporters in BECs, such as SLC7A11, SLC30A1, or SLC12A4, did not seem to differ between the five studies brain structures.

Function of Transferrin Receptor-Mediated Transcytosis
Following puromycin purification in vitro during the P0 selection, NHP-derived BECs at P1 evidenced the formation of a tight monolayer, continuous and with growth inhibition at confluence. NHP-derived BECs at P1 in monoculture or co-cultured with rat astrocytes were set in culture for four (P1D4) or seven (P1D7) days for further molecular and functional characterization. Paracellular permeability of the endothelial layer was also monitored through TEER measurement, where monolayers in monoculture presented a maximal 140 Ω·cm 2 , while brain endothelial monolayers in co-culture with astrocytes evidenced a maximal TEER of 245 Ω·cm 2 ( Figure 6A). The expression of claudin5, ZO-1 and occludin proteins were found at the cell-cell junctions ( Figure 6B). Time in culture or co-culture with astrocytes represented no changes in the protein expression in P1D7 vs. P1D4 regarding claudin5, ZO-1, and occludin ( Figure 6C), except for VE-cadherin (1.4-fold decrease found by WB, p < 0.001, Figure 6C). TFRC-mediated transcytosis activity was Investig.igated by measuring the transport of human holo-transferrin from the apical to the basolateral compartment ( Figure 7A), which showed to be significantly higher vs. Dextran 70 kDa, used as marker of paracellular passage (3.2-fold, p < 0.001, Figure 7B). TFRC-mediated transcytosis activity was statistically no different when NHP-BECs were co-cultured with or without astrocytes for seven days (P1D7). Similarly, the RMT transport of an internally developed antibody against the human and cynomolgus TFRC, was evaluated in the present model (co-culture with rat astrocytes, P1D4) vs. a control mouse IgG. A 5.7-fold higher apparent permeability was found for the anti-TFRC antibody vs. control mouse IgG ( Figure 7C, p < 0.001). Taken together, these results indicate TFRC actively participates in RMT in the BEC-derived model. Relative protein levels were obtained in the NHP-derived BBB model at P1 following either 4 or 7 days in vitro (P1D4 and P1D7, respectively) in the absence or presence of primary rat astrocytes (control and +Astrocytes, respectively) and were assessed by western blotting (levels normalized to α-tubulin). Results are expressed and means of each group ± SD (n = 3 independent experiments); pvalues were obtained by One-way ANOVA with Dunnett's multiple comparison test vs. Ctrl P1D4 groups, NS = p > 0.05, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. 50 µm Figure 6. Profiling of essential proteins for maintenance of a BBB phenotype and transport properties in NHP-derived BECs. (A) TEER values registered across endothelial cells following several days in monoculture (−A) or in co-culture with astrocytes (+A). Data are presented as mean values from 3 independent experiments, each containing 4-8 filter inserts. (B) Immunofluorescence staining of several cell-cell junction proteins in BECs. Protein staining was obtained in the NHP-derived BBB model at P1 following either 4 or 7 days in vitro (P1D4 and P1D7, respectively) in the absence of primary rat astrocytes. (C) Relative protein expression of cell-cell junction and TFRC proteins in BECs. Relative protein levels were obtained in the NHP-derived BBB model at P1 following either 4 or 7 days in vitro (P1D4 and P1D7, respectively) in the absence or presence of primary rat astrocytes (control and +Astrocytes, respectively) and were assessed by western blotting (levels normalized to α-tubulin). Results are expressed and means of each group ± SD (n = 3 independent experiments); p-values were obtained by One-way ANOVA with Dunnett's multiple comparison test vs. Ctrl P1D4 groups, NS = p > 0.05, * = p < 0.05, ** = p < 0.01, *** = p < 0.001.

Discussion
Decades Investig.ed in the search of CNS-active drugs have proven the formidable challenge of the BBB in the fight against diverse neurologic disorders. Primate studies may bring insights in the human BBB-focused research, as rhesus monkeys share greater transcriptomic similarities with human brains than mouse [83,84]. The NHP brain endothelium transcriptome is currently underrepresented, hampering a clearer elucidation of the primate BBB [10,85]. The establishment of the NHP-derived BBB model hereby presented constitutes a serious asset in comparison to cell line-based (hCMEC/D3, bEnd.3) and primary BBB models: (1) the starting material obtained from one animal is considerably greater and access to fresh post mortem tissue is much easier than for human, (2) BECs are obtained from freshly isolated post-mortem material, and therefore the BBB molecular features are well-preserved, (3) there is an evidently low variability between cultures from the same starting material, and (4) considerable retention of endothelial characteristics after several days in culture and cellular passaging. Furthermore, previous studies comparing isolated brain microvessels from rodents, primates and humans have reported consistent similarities between NHPs and humans from a molecular and pharmacokinetic standpoint, but significant differences have been retrieved between rodents and humans [10,66,86], favorizing the use of NHP or human-derived over rodent-derived BBB models as a tool in the prediction of CNS exposure of drugs and biologics in NHP and humans, as well as for target validation of novel drug targets at the primate level. On the other hand, commercially available human brain endothelial cells, such as hCMEC/D3 or human primary BECs, even though expressing some typical endothelial cell markers, they are reported to (1) have very low levels of expression of tight junctions, which are translated into low TEER values and relative high permeability towards small tracer molecules indicating paracellular leakiness, and (2) present major differences in the SLC and ABC transporter families and in diverse cell surface receptors, in comparison to freshly isolated brain microvessels, which have direct consequences when assessing transport mechanisms in these models [87]. As an alternatrive to the present model, PharmaCo-Cell Company has commercialized a BBB model prepared from primary cultures of monkey (Macaca irus) brain capillary endothelial cells, co-cultured with rat brain pericytes and astrocytes. However, its full characterization has not yet been published, with little information on how cells are obtained, and the model established. This newly generated dataset may therefore constitute a powerful tool for better understanding the BBB biology.
BECs constituting the BBB are in close contact with pericytes, encapsulated by the basement membrane, and surrounded by astrocytic endfeet processes and other cell types like microglia and neurons [88]. Consequently, the obtention of pure BECs, and thus of an accurate gene expression data on the BBB constitutes a real challenge. In the present study, brain microvessel isolation from a cynomolgus monkey brain was obtained through enzymatic digestion and gradient centrifugation, which yielded a highly enriched fraction of microvascular endothelial cells, but yet contaminated with pericyte, glial and circulating blood cell material. To eliminate any source of contamination, the isolated fraction was set in vitro and BECs selected through further exposure to puromycin [89].
Previously published comparisons of BBB gene expression profiles from freshly isolated rodent BECs versus primary cultures revealed evidence of dedifferentiation and/or down-regulation of BEC qualities (i.e., reduced levels of tight junction proteins, transporters and receptors) [75,90,91]. This is not surprising as cells cultivated in vitro are often at different stages of proliferation and differentiation. Nevertheless, we were willing to accept such limitations, as using an in vitro BBB model may yield some important insights into BEC response to external stimuli or stress, and for the prediction of passage of biotherapeutics. Nonetheless, the presented model was not thoroughly characterized at a molecular and functional level over P2, and therefore remains to be proven as an alternative model for high-throughput testing. Some authors have claimed the maintenance of barrier function properties, along with other phenotypic and cell purity properties in primary BECs following multiple passages [92,93], as inferred from the analysis of junction proteins (ZO-1 and β-catenin), endothelial cell markers (GLUT-1, vWF), and TEER measurement. However, it is very likely that more than one passage will be detrimental to the quality of the model, as documented in primary mouse models [94].
A key BBB feature is the presence of tight junctions between adjacent endothelial cells [95], where claudins (CLDN family), occludin (OCLN), and junctional adhesion molecules (JAMs) interact with cytoplasmic adaptors, including zona occludens and cingulin (CGNL1), establishing a linkage to the cytoskeleton and adherens junctions. In BECs, CLDN5 has shown to be the dominant CLDN family member and critical for the barrier formation and function [96,97]. The transcriptomic analysis hereby presented evidences CLDN5 as the highest expressed tight junction element within the cynomolgus brain vasculature, and likewise maintaining considerable high levels in vitro. A substantial expression following sub-culturing in vitro is also seen for other tight junctions, such as TJP1, OCLN, JAMs, and cadherins. The confirmation of protein expression of major tight junctions in our NHP-derived BECs, and notably their morphological distribution at the cell-to-cell interaction further hallmarks the tightness of the cell monolayer formed. Such results support that NHP-derived BECs maintain a relevant level of differentiation, representative of the brain microvasculature in its physiological state. The co-culture of BECs with astrocytes resulted in an increase on the average TEER measurements, but not in the global expression of tight junction and transporter proteins, findings corroborated by previous studies [98].
One of the most defining features of the BBB is the polarized expression of different transporters at the brain microvascular endothelium. The two major ABC transporters known at the BBB, ABCG2 and ABCB1, confirmed the highest expression levels among the ABC transporter superfamily in the NHP brain endothelium, a feature maintained in vitro, regardless of the downregulation observed for both transporters. These findings are consistent with reported transcriptomic and proteomic analyses [10,57,78], and evidence that ABCG2 greater expression over ABCB1 prevails among the primate BBB. Our analysis also confirmed ABCC4 as the third major efflux transporter present at the NHP BBB, corroborating previous studies [10,66,78]. The expression of several ABC transporters has been demonstrated to be species-dependent, where similar protein levels are found among rodents' species, and among primates, but not from rodents to primates. The human [66], cynomolgus monkey [10], and marmoset [86] BBB's are estimated to present 3-4-fold lower levels of ABCB1 than the rodent BBB. Likewise, consistent results regarding the expression levels of ABCC4 at the human and cynomolgus monkey BBB were also found [10,66], while the rodent BBB presents 5-8-fold higher abcc4 levels than those at the human and NHP BBB. Consequently, extrapolation of the BBB permeability obtained from rodent studies tends to underestimate the brain distribution of ABCB1 and ABCC4 substrates in human. In the present genomic analysis, we also identified the most abundant SLC transporters present at the primate BBB-SLC2A1, SLC7A5, SLC7A1, and SLC3A2, SLCO2B1, and SLC38A5, confirming previous findings [75][76][77][78]. These transporters remain the most expressed SLCs in vitro, except for SLC7A5 and SLC38A5 whose expressions considerably dropped in vitro. Therefore, SLC7A5 modest expression might compromise LAT-1 heterodimer activity in this model.
Our present study was also dedicated to exploring eventual differences of the vascular tree across regions of the brain, which is yet poorly addressed. In fact, regional specificities of the BBB have been reported in disease, injury [99][100][101][102], and aging [103], but to the best of our knowledge no report explores regional specificities of the BBB in the healthy brain. Given the circumventricular organs of the brain (e.g., posterior pituitary gland, pineal gland, median eminence of the hypothalamus) display a fenestrated endothelium with intercellular gaps [104], these were not included in the analyses. The present data suggest a considerable overlap of the molecular signatures of the brain endothelium obtained from the brainstem, cerebellum, cortex, hippocampus, and striatum, suggesting a rather homogeneous structure of the BBB, and therefore of its role across these brain regions. Nonetheless, these are important in the consolidation of our understanding regarding the heterogeneity of the BBB and how it may regulate brain function.
We also evaluated this NHP-derived BBB model in terms of RMT. Up to this date, TFRC remains the best validated RMT-mechanism allowing the enhancement brain exposure of biotherapeutics, widely reported by independent research groups [31][32][33][34][35]. Our study identified a TFRC-mediated transcytosis activity, as both transferrin and the anti-human/anti-cynomolgus TFRC were successfully transported from the apical to the basolateral compartment. Despite the downregulation of TFRC mRNA expression levels from isolated brain capillaries to P1, substantial expression levels are kept, which was also confirmed at protein level by WB. The present model presents improved tightness of the cell monolayer in comparison to cell line models such as hCMEC/D3, which displays low TEER values (~40 Ω·cm 2 , [3]) and for which an evaluation of the transcytosis of macromolecules across the model may only be carried using the so-called pulse-chase assay [105]. On the other hand, iPSC-derived BBB models have reported high TEER values (up to 2500 Ω·cm 2 , [106]), close to the ones reported in a physiological human BBB (~5000 Ω·cm 2 , [107]), however their differentiation conditions need to be carefully controlled to avoid an epithelial rather than endothelial phenotype [108,109]. In addition, no report can be found demonstrating transcytosis of anti-TfR antibodies in iPSC-derived BBB models, and comparable permeabilities for transferrin vs. human IgG have been shown [110].
The expression of other receptors participating in RMT was also confirmed by RNAseq in the present NHP-derived BBB model. Nonetheless, most RMT mechanisms reported to-date are ubiquitously expressed raising potential specificity and safety concerns. The identification of a novel RMT mechanism that is specifically and highly expressed at the BBB would represent a considerable asset in such competitive field, and the present transcriptomic analysis constitutes an important asset in the pursuit of such achievement. In fact, this NHP model probably represents one of the most adequate tools for the prediction of the human blood-to-brain permeability of macromolecules utilizing RMT mechanisms.