2 juvenile and nine adult cynomolgus monkeys (Macaca fascicularis), imported from Le Tamarinier (Mauritius), or Nafovanny (Vietnam) comprising four males and seven females were included in the study. The animals were part of other studies (Table 1) and were socially housed in primate cages under controlled conditions of humidity, temperature, and light (12-h light/12-h dark cycle, light on at 8.00 a.m.), within a dedicated primate facility. Primate diet was provided daily in amounts appropriate for the size and age of the animals. Fresh fruits were also given to the animals daily, and unsweetened treats were scattered in the litter as part of the Testing Facility Environmental Enrichment Program. Tap water was available ad libitum to each animal. Animal care was supervised by veterinarians experienced in NHP husbandry. Following acceptance of the study design by the IACUC for NHP experiments (National Veterinary School of Lyon, Lyon, France), brains were collected from animals undergoing experiments carried out in accordance to the European Communities Council Directive (2010/63/EU) for the care of laboratory animals. The eleven cynomolgus monkeys and their inclusion in the different aspects of the study are described in Table 1.

Animals were euthanized under deep anaesthesia (intramuscular injection of Ketamine and Midazolam) with 4 g of intravenous pentobarbital (Dolethal, Vetoquinol). Brains were collected and quickly immerged in cold HBSS (Gibco^®, Fisher Scientific, France). Lateral ventricle choroid plexuses (LVCP) and fourth ventricle choroid plexuses (4VCP), as well as pieces of meninge-free cortex and cerebellum were collected. Samples collected for gene expression analyses were snap-frozen in liquid nitrogen and kept at −80°C. For immunohistochemistry, additional tissue fragments were snap-frozen in isopentane at −50°C, embedded in Tissue-Tek (Sakura Finetek Europe, Netherland) and stored at −80°C. Other fragments were fixed in RCL2 fixative (Alphelys, France) before paraffin embedding. In two animals, freshly collected cortical gray matter was used for immediate microvessel isolation, or frozen at −80°C for further microvessel isolation.

Brain microvessels were isolated as described previously in details for rats (Gazzin et al., 2008). Briefly, an average of 5 g of cortical tissue was used for each preparation. The tissue was carefully cleaned from meninges and superficial blood vessels prior to microvessel isolation. Minced cortical tissue was homogenized in Krebs-Ringer buffer (KR) by mechanical homogenization, and an aliquot of homogenate was sampled for enzymatic measurement. The preparation was further diluted with KR supplemented with 1% (w/v) bovine serum albumin (BSA) (5/1, v/w of tissue), homogenized, and subjected sequentially to several steps of centrifugation, to a 70 kDa dextran gradient, and to filtration on sieves of decreasing mesh sizes. Microvessels were collected on a 40 µm-mesh sieve in 0.1% BSA-supplemented KR.

Total RNA was prepared from LVCP, 4VCP, cerebral cortex, cerebellum, and brain microvessel fractions using Qiagen RNeasy Micro and Mini kits (Qiagen, Valencia, CA, United States) according to the manufacter’s instructions. Tissue was homogenized in RLT buffer using Soft tissue homogenizing tubes in the Minilys tissue grinder (both from Bertin Technologies, France) prior to the column purification step. Digestion with proteinase K (Qiagen) was performed on choroidal and microvascular samples. All tissues were treated with DNAse as recommended in the kit instructions. RNA was quantified by OD measurement at 260 nm using a NanoDrop spectrophotometer (ThermoScientific, Baltimore, MA, United States) and quality was assessed with the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, United States). Only samples with an RNA integrity number above 7.8 were used in the study.

BioChain^® total RNA from cynomolgus monkey peripheral tissues (colon, liver, lung, and kidney) were purchased from Clinisciences (Nanterre, France) to be used as positive control tissues to define PCR conditions. All RNAs were reverse-transcribed using the iScript cDNA Synthesis kit (Bio-Rad, Marnes-la-Coquette, France).

Quantitative polymerase chain reaction (qPCR) was performed using the LightCycler FastStart-DNA Master SYBR Green I kit and the LightCycler^® 2.0.5 Instrument (Roche Diagnostics GmbH, Mannheim, Germany), or the LightCycler^® 480 SYBR Green I Master kit and the LightCycler^® 480 Instrument II (Roche). Primers were designed using NCBI Primer-BLAST or chosen from the NCBI probe bank. They were selected to generate amplicons with a length of 85–200 bp (Table 2). After an initial step of DNA polymerase activation at 95°C for 8 min, the following amplification conditions were used: 45 cycles of denaturation at 95°C for 10 s, annealing for 10 s and extension at 72°C for 10 s. A touchdown protocol was applied by setting the initial annealing temperature at 68°C and decreasing this value by 0.5°C for the next 12 cycles, and running all the remaining annealing steps at 62°. Melting-curve analysis was then performed to verify the amplification of a single product with a specific melting temperature. For genes analyzed using the LightCycler FastStart-DNA Master SYBR Green I kit, MgCl₂ concentration was optimized. A negative PCR control without cDNA was included in all runs. A standard curve was generated for each gene of interest by nonlinear regression analysis of crossing points (threshold cycles, Ct) generated for at least five serial dilutions of a cDNA pool, using the LightCycler^® Software 4.1 and LightCycler^® 480 Software 1.5.1.

Ct values of unknown samples were then used to determine in each sample a relative cDNA concentration of the target gene. Possible sample-to-sample variations in reverse transcription efficiency and in qPCR processing were corrected by normalizing the data to the expression of the gene dolichyl-phosphate N-acetylglucosaminephosphotransferase 1 (DPAGT1).

In order to provide an index of abundance for the different tight junction proteins and transporter gene products, expression levels of all genes were estimated first in a reference sample, arbitrarily chosen as LVCP cynomolgus #2 as follows: D P A G T 1  efficiency   D P A G T 1 Ct /  Target efficiency   Target Ct where efficiencies of amplification were calculated from the linear part of the standard curves using the LightCycler Software 4.1. The obtained gene values were normalized to CLDN1 value set as 100 after correction for the differences in the size of the amplification products. Then for each target gene, the expression level for each sample was expressed relative to the value of the reference sample.

Hematoxylin (Biolyon 0942)–phloxin (Ral diagnostics 361470)–safran (BDH Gurr 35093) was used to stain nuclei, cytoplasm and connective tissue, respectively, according to the following sequence. Parafin sections were deparafinized in methylcyclohexane (three times for 5 min), rehydrated through a graded series of ethanol/water solutions (100%–0% of ethanol), and incubated in hematoxylin (0.5% w/v) for 5 min. After a 10-min wash in running water, sections were immersed for 15 s in hydrochloric alcool (3 drops of pure hydrochloric acid in 100 mL of absolute ethanol). After a 10-min wash in running water, sections were immersed for 3 s in phloxin (1% w/v). The slides were then immersed for 10 s in an alcoholic safran solution before being rinsed in 100% ethanol. Sections were rinsed quickly with water before being dehydrated through a graded series of ethanol/water solutions (70%–100% ethanol) followed by 3 baths of methylcyclohexane. Sections were examined with an Axioplan microscope (Zeiss, France).

For immunofluorescence labelling, frozen sections were fixed differently depending on the primary antibodies (Table 3). Fixation in 4% (w/v) paraformaldehyde in phosphate buffer was performed at room temperature for 10 min. Fixation in 1% (w/v) paraformaldehyde in phosphate buffer was performed at room temperature for 30 s. Fixation in methanol/acetone (50/50, v/v) was performed at −20°C for 5 min. Fixation in acetone was performed at −20°C for 8 min. Fixed sections were blocked for 1 h at room temperature in phosphate buffer saline (PBS, Euromedex, France) containing 0.2% BSA w/v, 10% normal goat serum v/v, 0.3% Triton X100 v/v for claudins (CLDN-1, -2, -3, -4, and -5). For all other primary antibodies, sections were blocked in a solution containing 5% BSA, 5% normal goat serum, 0.3% Triton X-100 in PBS. Sections were incubated overnight at 4°C in the blocking solution containing the primary antibody. The list of primary antibodies, with company names, product references and final concentrations is given in Table 3. Following three washes in the blocking solution, the sections were incubated with a secondary anti-rabbit, anti-rat or anti-mouse Alexa-conjugated antibody (Invitrogen, 2 μg/mL in blocking solution) for 1 h at room temperature. Nuclei were stained with 4′,6-Diamidine-2′-phenylindole dihydrochloride (Roche, 5 μg/mL in PBS) for 10 min at room temperature and sections were mounted with fluorescence mounting medium (F/TA-030-FM, Thermo Scientific). Negative controls were performed by omitting the primary antibody. Fluorescent immunolabelling was observed using an Axio Imager M2 fluorescence microscope (Zeiss).

Gamma-glutamyltransferase activity was determined in tissue homogenates and microvessel preparations using a dual-beam Cary 100 spectrophotometer (Varian) with L-γ-glutamyl-3-carboxy-4-nitroanilide as a substrate, in the presence of glycylglycine and Triton X-100, as previously described (Ghersi-Egea et al., 1994). The specific activity was calculated using an extinction coefficient of 9900 M^-1.cm^-1 for carboxy 4-nitoaniline. Glutathione-S-transferase activity was measured by kinetic spectrophotometry in the Cary 100 spectrophotometer using 1-chloro-2,4-dinitrobenzene (Sigma), a multispecific substrate of glutathione-S-transferase isoenzymes, as previously described (Ghersi-Egea et al., 2006). The specific activity was calculated using an extinction coefficient of 9600 M^-1.cm^-1 for the glutathione conjugate. Total protein content was measured by the method of Peterson (Peterson, 1977) using BSA to generate the standard curve.

Considering the number of structures (choroidal tissue, isolated microvessels, cerebral and cerebellar parenchyma), and the variable number of samples per tissue (n = 3 for isolated microvessels), the expression levels of barrier-related molecular attributes were compared among the different cerebral tissues by the non-parametric Kruskal and Wallis test corrected for multiple comparisons by the false discovery rate method of Benjamin and Hachberg (PRISM software). For clarity, only statistical differences discussed in the result section are marked on Figures 2, 4, 6. The complete set of statistic data is found in Supplementary Table S1. In all figures and in Supplementary Table S1, *, and ** indicate discoveries (statistical differences) when Q was set at 0.05, and 0.01, respectively. q Values close to significance when Q was set at 0.05 are also reported.

The brain microvessel isolation technique established for the rat (Gazzin et al., 2008) was applied to the cynomolgus monkey. It produced preparations that were highly enriched in capillaries and virtually devoid of tissue microfragments (Figure 1A). The activity of γ-glutamyltransferase, a key enzyme of the glutathione cycle used as a marker of cerebral endothelial cells (Ghersi-Egea et al., 1987), was enriched on average 40 times in microvessel homogenates compared to the initial cortex homogenates (Figure 1B). Of note, activities measured in both the microvessel fraction and cerebral cortex of cynomolgus were twice those measured in rat preparations (data not shown), indicating that glutathione-dependent metabolic pathways may be especially active in cynomolgus monkey brain. Isolated microvessels also comprise pericytes as they are embedded in a common basal membrane with endothelial cells. As a result, the expression of PDGFRβ ¹ , a marker of pericytes, was increased 3.8 ± 0.9-fold (n = 3 p < 0.01, paired t-test) in capillaries versus initial whole brain tissue. In contrast, while astrocytic end-feet ghosts can remain occasionally attached to isolated capillaries, mRNA from glial origin was low in the microvessel preparations, exemplified by GFAP mRNA level which was only 7.1 ± 4.8% of the levels measured in corresponding whole brain tissue (n = 3, p < 0.05, paired t-test). Choroid plexuses were microdissected from the lateral and fourth ventricles of cynomolgus monkey brains. Choroid plexuses from the two locations were kept separated for the analyses. Although closely related, they may not share the same level of neuroprotective capacity. Conventional histology showed a preserved organization of the choroidal villi with a well-preserved epithelial layer surrounding the vascularized stroma (Figure 1C). Total RNA from choroidal tissue, isolated microvessels, cortical and cerebellar parenchyma were isolated, reverse transcribed and subjected to qPCR. All mRNA preparations used in this study were assessed for quality (see method). As expected, the choroidal epithelial marker transthyretin was expressed at a high level in the choroid plexuses (Figure 1D), and was undetected in cortical and cerebellar parenchyma. Three microvessel fractions were incorporated in the qRT-PCR analyses as they yielded mRNA of good quality. They were obtained from two different cortical tissue pieces of cynomolgus #7 (2 preparations) and one cortical piece of cynomolgus #8 (1 preparation). The purity of these microvessels was illustrated by the expression of the endothelial marker PECAM-1 which was enriched 30 times in comparison to the initial cortical tissue used for capillary isolation (n = 2). These isolated cynomolgus monkey microvessel and choroid plexus preparations were therefore validated for further characterization of the molecular determinants responsible for the barrier phenotype of blood-brain interfaces.

Figure 2 shows the relative mRNA levels of claudins and occludin in cerebral microvessels and choroid plexuses. Data were obtained from juvenile and young adult females (3.3- to 4.6-year-old) with one adult and one older male (5- and 9-year-old, respectively, distinguished on the graphs) incorporated in the study. We combined all data from males and females in the statistical analyses. Of note, we could not evidence any correlation between expression and age when analysing juvenile and adult animals. The 9-year-old male animal yielded a mildly lower than average level of expression for four tight junction proteins in the choroid plexuses.

The molecular composition of tight junctions in cynomolgus monkey blood-brain interfaces shared the following features with that of tight junctions in rat and human barriers we described in Kratzer et al. (2012): CLDN2, 1, 19 (described in rat only), and CLDN3 (by order of abundance), which are hallmarks of the BCSFB, were specifically enriched in both LVCP and 4VCP, while CLDN5, characteristic of the BBB, was highly expressed and enriched in microvessels (Figure 2, blue and orange graphs). The junctional localization of these various proteins was confirmed by immunohistochemistry on monkey choroidal and parenchymal tissue (Figures 3A,B,F). OCLN coding for Occludin was highly expressed in both barrier fractions (Figure 2, green graph). CLDN11 was mostly expressed in parenchymal tissue (Figure 2, black graph), as expected from its involvement in myelin sheet organization, and as we already reported for adult rat brain parenchyma (Kratzer et al., 2012).

Of note, we found that CLDN7 and CLDN23 mRNAs were enriched in the cynomolgus monkey microvessel and choroid plexus preparations, respectively. Cldn-7 may play a role in tight junction repair at the BBB as inferred from the recent study performed on kidney epithelial cells (Higashi et al., 2023). The relative distribution of these two claudins between the two interfaces has not been reported previously in any species. We detected mRNAs for tight junction-associated ZO-1 and ZO-2 in all tissues tested, with no clear enrichment in either the cerebral microvessels or choroid plexuses (Supplementary Figure S1).

This study also provided evidence for a number of differences in the composition of tight junctions in cynomolgus monkey compared to the rat (Kratzer et al., 2012).

The localization of CLDN-4, selectively expressed at the BBB in rat (Kratzer et al., 2012), was confirmed in cynomolgus monkey (Figure 3E), but the mRNA analysis revealed that in the NHP, CLDN4 is also expressed in the choroidal tissue (Figure 2), as observed in human choroid plexuses (Rodriguez-Lorenzo et al., 2020). CLDN16 mRNA, selectively enriched at the BBB in rat (Kratzer et al., 2012), was further enriched in choroid plexuses of cynomolgus monkey. CLDN22 mRNA, selectively enriched at the BCSFB in rat (Kratzer et al., 2012), was also expressed at the BBB in the cynomolgus monkeys (Figure 2).

CLDN-3, selectively expressed and immunoreactive in the BCSFB in rat (Kratzer et al., 2012), was detected by immunohistochemistry at both interfaces in cynomolgus monkey (Figures 3C,D). We also observed in cynomolgus monkey additional CLDN-1 immunoreactivity in inter-endothelial junctions of cortical microvessels and larger vessels (data not shown), an observation previously reported for human (Tran et al., 2016). The expression of CLDN-3 at the BBB is debated and may vary between species (Kratzer et al., 2012; Steinemann et al., 2016; Tran et al., 2016). Given the low level of CLDN3 mRNA measured in the monkey microvessels compared to the level in choroid plexuses, and the absence of detection of CLDN1 mRNA in the microvessel fraction (Figure 2), these two claudins, if present at the cynomolgus monkey BBB, are probably only marginally expressed. Although anti-claudin antibodies are claudin-specific, a cross-reactivity between CLDN-1 or CLDN-3 and other tight junction proteins at the cerebral capillaries may occur, as reported for Cldn-3 in mice (Castro Dias et al., 2019). The parenchymal expression of CLDN3 restricted to the cerebellum remains also to be understood (Figure 2). Yet, a potential contamination of the cerebellar tissues by appended 4VCP fragments was ruled out by the complete absence of transthyretin expression in these preparations. A robust CLDN-5 immunoreactivity was observed in cynomolgus monkey, not only at the BBB and in endothelial junctions of large choroidal vessels as observed previously in rats (Kratzer et al., 2012), but also in junctions between choroidal epithelial cells (Figures 3G,H). The latter labelling has not been observed in the rat choroidal tissue, but has been observed in the choroidal epithelium during human development (Mollgard et al., 2017; Virag et al., 2017). Using RNA sequencing of adult human CPs, Rodriguez-Lorenzo and colleagues showed a substantial CLDN5 expression in this tissue (Rodriguez-Lorenzo et al., 2020), but did not assess CLDN-5 localization in their study. Collectively, these observations call for a clear understanding of the exact contribution of the numerous tightening (e.g., CLDN-5) and pore-forming (e.g., CLDN-2) claudins present at the BCSFB in setting the fence function of this interface. Differences in tight junction proteins between species may change the function of epithelial junctions and subsequently modify selectively blood to CSF permeation.

Taken together, our data indicate that the blood-brain and blood-CSF barriers in cynomolgus monkey share more analogy with the human barriers than the rodent barriers with regard to the protein composition of their tight junctions.

Besides ABC efflux transporters, various carriers of the SLC superfamily can also influence the cerebral bio-availability of drugs, toxic compounds, and biologically active metabolites. They include members of the SLCO, SLC22 and SLC47 families, which display distinct substrate specificity profiles.

The present work has focused among all SLCO carriers, on the ubiquitous members SLCO3A1 and SLCO4A1, and the BBB-specific members (within the brain) SLCO2B1 and SLCO1A2 (Bronger et al., 2005; Roth et al., 2012; Ronaldson and Davis, 2013). The OATP proteins encoded by these four genes accept various xenobiotics as substrates, such as antibiotics, endothelin A receptor antagonists, beta-blockers, antiretroviral and antineoplastic agents. Endogenous compounds such as steroids, thyroid hormones, prostaglandins are also transported by these carriers. The SLCO2A1-encoded OATP carrier more specifically transports prostaglandins (for reviews on substrate selectivity of these transporters, see Roth et al., 2012; Ronaldson and Davis, 2013). The expression profile of these transporters in brain barriers and parenchyma of cynomolgus is described in Figure 6. The robust expression of SLCO2B1 and SLCO1A2 genes in the non-human primate BBB, in line with data obtained in human tissues, stresses the probable influence of these two proteins on cerebral drug bioavailability (Morris et al., 2017; Schafer et al., 2020; Kinzi et al., 2021). Both messengers were enriched in microvessels compared to the cortex and cerebellum, corroborating the high microvessel-to-tissue ratios previously reported for mRNA levels in human brain (Gao et al., 2000; Geier et al., 2013; Suhy et al., 2017). Immunohistochemical and quantitative proteomic analyses reported by other groups consistently detected SLCO2B1-encoded protein in microvessels isolated from human brain, while SLCO1A2-encoded protein was not always detected (Uchida et al., 2011; Gao et al., 2015; Billington et al., 2019; Bao et al., 2020; Schafer et al., 2020). In cynomolgus choroid plexuses, we found high mRNA levels of SLCO1A2, but not SLCO2B1 (Figure 6). This is in contrast with human transcriptomic data showing a higher expression of SLCO2B1 as compared to SLCO1A2 in this tissue (Rodriguez-Lorenzo et al., 2020), and with a proteomic study that failed to detect SLCO1A2-encoded protein in one specimen of human choroid plexus tissue (Uchida et al., 2015). The expression of SLCO2A1 and SLCO4A1 genes was low in all cynomolgus tissues we investigated, with no apparent specificity for either barrier. In contrast, SLCO3A1 expression was higher in choroid plexuses than in microvessels isolated from monkey brain, and intermediate in parenchyma (Figure 6). In line with this pattern, immunohistochemical analysis of human brain sections has shown that the two splice variants of human OATO3A1 were associated with the membrane of choroid plexus epithelial cells, and were detected in neural cells of the frontal cortex (Huber et al., 2007). Comparing the expression of these SLCO genes between primate and rodent is strongly impeded by the low homology of sequences between these species, which is characteristic of the SLCO family (Hagenbuch and Meier, 2004). Overall, our data suggest that the expression of SLCO genes of interest in the primate BBB resembles that observed in human, and further provide a detailed pattern of expression for these genes at the BCSFB. Of note, OATP-dependent transport processes can be bidirectional, and the membrane localization of most OATP proteins is unknown in the primate blood-brain and blood-CSF barriers. Whether these transporters influence positively or negatively the cerebral penetration of their substrates remains a poorly understood field of research in primates.

The organic anion transporter SLC22A8 has been shown to play an important role as an efflux transporter for many drugs including antivirals and antibiotics, and for endogenous compounds such as steroid derivatives at both brain barriers in rats and mice (Strazielle et al., 2003; Sykes et al., 2004; Ose et al., 2009). We found a high expression of this gene in choroid plexuses isolated from monkey brain, while its expression was barely detectable in other tissues investigated including microvessels (Figure 6). Accordingly, the corresponding protein has not been detected in a quantitative proteomic analysis of microvessels isolated from cynomolgus monkey brains (Ito et al., 2011). In human brain microvessels, the protein was identified in only one study, albeit at a low level compared to ABC transporters (Al-Majdoub et al., 2019), and was not detected in several other targeted proteomic analyses (Shawahna et al., 2011; Uchida et al., 2011; Billington et al., 2019; Bao et al., 2020). SLC22A8 expression level was higher than that of ABCC1 in human CP (Rodriguez-Lorenzo et al., 2020), and the protein was present at a level similar to that of ABCC1 in one sample of human choroid plexus (Uchida et al., 2015). Collectively, these data suggest that, with respect to the relative expression of SC22A8 between the two barriers, a higher similarity of SLC22A8-dependent efflux processes exists between non-human and human primates than between primates and rodents.

Finally, we investigated carriers of the SLC47 family for which differences in expression have been reported between human and non-primate species. The protein multidrug and toxin exclusion (MATE)1 encoded by SLC47A1 was detected in substantial amounts by absolute targeted proteomic in human choroid plexuses, but was under the limit of quantification in rat tissue (Uchida et al., 2015). In cynomolgus monkey, we found that SLC47A1 and SLC47A2 levels of expression were both higher in choroid plexuses than in brain microvessels (Figure 6), a finding that is consistent with the modest expression, if any, of SLC47A1 in human brain microvessels (Shawahna et al., 2011). We also detected SLC47A1 mRNA in monkey cortex and cerebellum, while SLC47A2 expression was more specific of the choroid plexuses (Figure 6). Fairly high levels of message for both transporters were also detected in human choroid plexuses (Rodriguez-Lorenzo et al., 2020). Yet, the protein MATE2-K, the main variant isoform encoded by SLC47A2 in kidney, was under the limit of quantification in human choroid plexus (Uchida et al., 2015). The proteins encoded by SLC47 genes, MATE1 and MATE2-K are proton-coupled antiporters for small organic cations. They accept endogenous and exogenous cations as substrates, including a large range of therapeutic agents, natural products such as flavonoids, and potentially toxic molecules such as creatinine, some neurosteroids, or else methylphenylpyridinium (Motohashi and Inui, 2013; Nies et al., 2016). The substrate specificity of other SLC47A2 variant isoforms is currently unknow except for tetraethylammonium (Nies et al., 2016). More data on the protein isoforms actually present in the BCSFB, their membrane localization, and their substrate specificities are needed before their involvement in the clearance of organic cations from the CNS can be fully clarified in primate.