The development and improvement of in vitro models of the blood-CNS barriers is an ongoing effort towards both understanding the biology of these important regulatory tissues, and being able to overcome the formidable obstacle that they present to the delivery of therapeutics for the treatment of debilitating neurological diseases. Human, rat and mouse BBB cell lines have been developed but, although cheap and convenient to use, these cells produce barriers with high intrinsic paracellular permeability making them poorly-suited for applications such as drug transport screening and characterisation [5, 76–78]. Due to this fact, in vitro primary cell models of highly differentiated endothelial cells from the BBB and BSCB remain a critical tool for investigative and pharmaceutical biology, particularly in the species typically used in pre-clinical studies. Such in vitro primary models have been utilised for genomics and proteomics studies [30, 79–81], analysing endothelial transporter function [15, 16, 65, 82, 83], studying brain metastasis of cancer cells , and applied to translational pharmaceutical studies, investigating small molecule drug transport [77, 85, 86] and toxicity [67, 87].
Here, we describe a further development in techniques producing such in vitro barrier models from primary rat CNS tissue, providing the first description of the provision of both brain and spinal cord endothelial cells from the rat, a species of pre-clinical importance in pharmaceutical CNS drug development and a commonly used laboratory model organism. High yields of differentiated cells are cultured from the same donor animals, reducing cost, labour and number of animals required. Furthermore, the endothelial cells obtained by this method are able to form monolayers with excellent barrier characteristics in vitro, making them suitable for use in biological investigations and in drug transport and toxicity studies. Importantly, our methods detail the first procedure for the culture of robust and in vivo-like spinal cord endothelial cells from the rat, complementing existing descriptions from mouse , and we also provide the first description of a functional in vitro barrier phenotype for spinal cord endothelial cells from any species.
Our aim was to achieve high yields of RBECs and RSCECs to provide a large number of cells that could be used for biological studies and drug discovery. We therefore introduced steps into our protocol to facilitate enhanced recovery of endothelial cells. Firstly, to increase cell numbers, we omitted size-dependent filtration from our microvessel isolation protocol. Filtering the microvessels enriches for smaller capillaries, which are hypothesised to be more “BBB-like” , but decreases the overall yield of endothelial cells. By filtering through 40 μm or 70 μm cell strainers, we observed that many microvessels of a wide range of sizes, including small capillaries, were retained on the cell strainer. We thus plated out the whole microvessel pellet isolated by BSA density centrifugation and subjected it to puromycin purification [27, 31, 51, 61, 65]. In addition to maintaining purity, the puromycin also enforces a selective pressure on the endothelial cells; RBECs and RSCECs with high expression of P-gp, a characteristic of in vivo blood-CNS barriers, are able to survive and proliferate in vitro. Indeed our data indicates that both RBECs and RSCECs in culture express P-gp efflux transporters (Figures 5h and 7h) that retain functionality (Figure 8a, b). By culturing cells using this selective method, we obtained large numbers of primary endothelial cells, which were then passaged onto cell culture inserts or tissue culture plastic and retained a highly-differentiated barrier phenotype.
We next looked to improve reproducibility in obtaining tight barriers, as a lack of robustness is a common problem when culturing primary brain endothelial cells in vitro. Our observations from passaging brain endothelial cells led us to an approach of splitting cells on the basis of the surface areas of the receiving culture dishes, rather than by traditional dilution splitting on the basis of cell numbers. If RBECs or RSCECs were transferred to new dishes or cell culture inserts as a dilution passage (e.g. splitting 1:4–1:2) they frequently grew as islands of cells which stopped proliferating and did not form a continuous monolayer, rendering them unsuitable for barrier studies on cell culture inserts. If, however, the cells were transferred on the basis of the surface area of the dish they were passaged into, they quickly reached confluence and formed functional barriers (Figure 1). Thus, a key element of our protocol is the concept of the ~1:1 passage of endothelial cells. This transfer method facilitated excellent barrier phenotypes for both RBECS and RSCECs (Figures 3, 4 and 6), with no obvious endothelial de-differentiation as judged by immunocytochemistry which demonstrated well organised, mature tight junctions and expression of endocytic transport machinery (Figures 5 and 7). Furthermore, the technique reliably resulted in the provision of useable barriers in almost every insert. We observed very low losses of individual inserts where barriers did not form, as often happened with dilution passaging. Individual inserts where the barrier failed were usually found to be due to handling technique and mechanical damage to the monolayer.
A major finding of our study was that barriers with high TEER and low Pe to small molecules such as LY were reproducibly obtained when culturing RBECs and RSCECs in Lonza’s EBM-2 basal medium with the EGM-2 BulletKit minus VEGF. The Lonza BulletKit contains supplements, such as hydrocortisone and FGF, that are well validated to improve endothelial barrier function in vitro[22, 61]. This EBM-2/EGM-2 media combination outperformed DMEM supplemented with another commercial supplement, MVGS. The optimised conditions also included the addition of 15% plasma-derived serum and the monolayers did not display the sensitivity to serum-derived factors that has been observed in some in vitro BBB cell culture models .
Historically, the BBB has been more highly studied than the BSCB both in vivo and in vitro. An emerging consensus is that the two are broadly similar with some subtle differences, for example in their permeability and their vulnerability to certain insults and diseases [4, 89]. The BSCB has an almost identical physical structure to that of the BBB, with tight junction-containing endothelial cells surrounded by and interacting with astrocytes and pericytes . The BSCB also appears to be more permeable than the BBB in certain sub-regions but is still a tight and highly regulated barrier that protects the spinal cord parenchyma. Studies in vivo have indicated that the BSCB is more permeable than the BBB to small tracers, cytokines and neurotrophins, with lumbar regions of the spinal cord in particular being more permeable [90–94]. Some cytokines and growth factors however, such as IL-1α and granulocyte-macrophage colony-stimulating factor (GMCSF), have been shown to have similar transport across the BSCB as the BBB in vivo[95, 96]. Other evidence suggests that the BBB and BSCB have similarly low permeability to large plasma proteins such as IgGs and albumin [95, 97]. Similarities between the BBB and BSCB have also been noted in the expression and functionality of ABC transporters, which are hypothesised to play key roles in disease and drug resistance. Isolated capillaries from mouse brain and spinal cord show similar expression and functionality of P-gp, MDR2 and BRCP [82, 83]. ABC-transporters at the BBB and BSCB also share similar increased expression and functionality following exposure to dioxins, and in mouse models of amyotropic lateral sclerosis (ALS) in vivo[82, 83]. Interestingly, our models show that RSCEC monolayers are generally slightly more permeable than with RBECs, in spite of the fact that both are cultured in the presence of glial cells derived from brain tissue. Although data is lacking attesting to differences between brain and spinal cord astrocytes in their ability to induce barrier phenotypes, our observations may indicate that some aspects of the permeability properties of RSCECs are cell-intrinsic.
The only previously published in vitro study comparing BBB and BSCB cells was carried out using endothelial cells derived from mouse . In that study, culture conditions for both endothelial cell types were established and expression levels of proteins associated with barrier function were characterised, although no functional barrier data was presented. Ge and Pachter (2006) found that cultured endothelial cells from both type of CNS tissue were indistinguishable under the microscope and showed identical expression of the endothelial markers vWF and PECAM-1, as well as similar uptake of LDL . Our extensive characterisation data for endothelial markers, tight junction proteins, endocytic machinery and the P-gp efflux transporter, suggests a similar situation to be true for brain and spinal cord endothelial cells from rat (Figures 5 and 7). Additionally, Ge and Pachter provided a highly useful comparison of several genes important for barrier function in these cultured endothelial cells . Gene expression of claudin-1, claudin-5, P-gp and transferrin receptor were unchanged between both types of endothelial cell in mouse, but expression levels of ZO-1, occludin, β-catenin and VE-cadherin were lower in spinal cord endothelial cells compared to those from brain tissue . This observation is in agreement with in vivo descriptions of the BSCB being more permeable than the BBB.
Our observations further support and extend these observations on the structure and function of the BBB and the BSCB. We have shown that cells from both the rat BBB and BSCB can be cultured on cell culture inserts in vitro to form functionally restrictive cell monolayers, with the endothelial cells of the brain forming slightly tighter barriers than those of the spinal cord. In this regard our in vitro models apparently mimic the in vivo situation for the BBB and BSCB. We observed an excellent relationship between pre-experimental TEER values and Pe to LY in permeability assays for both models (R2 = 0.78 for RBECs and 0.91 for RSCECs). Importantly, this indicates that the pre-experimental TEER value is predictive of the Pe to LY, allowing consistent and reproducible experiments to be performed. Cell culture inserts with high TEER values can be selected from the outset and matched with inserts of similar barrier tightness, allowing robust experiments to be performed on primary-derived cells that have similar intrinsic permeability properties.
Our BBB and BSCB models exhibited excellent discrimination characteristics for limiting the passage of small and large molecules that cross the barrier by paracellular diffusion, such as LY and FITC-dextrans, and also for small molecules that enter the CNS poorly on the basis of their low lipophilicity. This makes our models ideally suited for in vitro permeability studies, particularly for small molecule drugs where tight in vitro barriers are critically required to minimise non-specific paracellular transport that would mask true permeability characteristics and kinetics. Small molecule permeability across an in vitro barrier has been demonstrated by several groups using different species, including human, porcine, mouse and rat, but only across endothelial monolayers derived from cells of the BBB [16, 25, 54, 65]. Our methods show that we are able generate a large number of tight in vitro barriers representing the rat BBB, but we also demonstrate for the first time an in vitro model of rat spinal cord endothelial cells that shows similar restrictive properties to small molecules. These data suggest that our models would be suitable for a broad range of CNS drug discovery studies, particularly for instances where a drug target is located within the spinal cord as well as, or instead of, in the brain. The RBEC and RSCEC barriers also show expression and functionality of the clinically important efflux transport P-gp (Figures 5h, 7h and 8a, b). These BBB and BSCB models could thus be used for in vitro studies of barrier function involving this transporter, such as determining the efflux of chemotherapeutic small molecule drugs which are often also P-gp substrates.
Since differences between the BBB and BSCB exist, it is therefore essential that in vitro models for both barriers are available for research purposes. An in vitro model for one barrier may not necessarily be an appropriate substitute for the other. This may be of particular relevance when studying diseases which affect one CNS compartment more than the other [4, 89]. Our in vitro models of both types of blood-CNS barrier are thus of great potential value for the investigation of such disease processes. Since these models are optimised for rat tissues, a species for which relevant and well-characterised in vivo models of CNS-disease exist, they have great potential utility in translational studies. Our novel in vitro RSCEC model of the rat BSCB may also contribute to the furthering of knowledge about this poorly-understood blood-CNS barrier and could be applied to genomics and proteomics studies in the future.