A detailed method for preparation of a functional and flexible blood–brain barrier model using porcine brain endothelial cells^☆

2.1. Isolation of brain capillary endothelial cells

A homogeneous and smaller diameter vessel fraction was collected from the finer filters (60 µm mesh) than from the coarser filters (150 µm mesh). Furthermore, TEER of PBEC monolayers cultured from the 60 µm fraction was higher, consistent with the 60 µm fraction being derived from purer capillaries (60s: 625±21 Ω cm², n=6, cf. 150s: 237±10 Ω cm², n=6).

2.2. Characterisation

Characterisation of the brain endothelial cell monolayers produced by this method (Patabendige et al., this issue) and the co-culture variant (Skinner et al., 2009) are published elsewhere. By a range of morphological, immunocytochemical and functional criteria, the cells reproduce well in vivo endothelial and BBB features, from expression of endothelial markers, to organisation of tight junction proteins, and exp-ression of typical BBB enzymes and transport systems. They have been used for a number of studies on the cellular and molecular function of the BBB (in preparation).

2.3. Morphology, assessment of barrier integrity and reproducibility

TEER is one of the best measures of the barrier function of an in vitro BBB model, and has been used throughout the optimi-sation of this method and applications of the resulting model variants.

3.1. Isolation of microvessel rather than large vessel endothelium

The careful removal of meninges, including its invaginating folds into sulci, was designed to remove the large surface vessels, including many of the penetrating arterioles which run perpendicularly into the brain cortex (Dacey and Duling, 1982). This will not only remove most of the potential contamination by leptomeningeal cells with fibroblast-like properties, but also by arterial and arteriolar smooth muscle cells, which tend to grow more rapidly than endothelial cells in culture. The two-stage filtration is designed to retain vessel fragments, allowing isolated cells including most glial cells to pass through. Examination of the material collected from the coarser and finer filters (150 µm and 60 µm mesh respectively) shows that the 150 µm filters retain a less pure (and generally larger diameter) vessel fraction than the 60 µm filters; the latter generate a more homogeneous and higher TEER monolayer consistent with it being derived from relatively pure capillary endothelium. Isolation of predominantly capillary rather than arteriolar or venular microvessels is important as there are several phenotypic and functional differences between the endothelial cells of these different segments of the microvasculature. In particular, compared with arteriolar or venular endothelium, cerebral capillary endothelium has more a more complex and complete pattern of tight junction strands in freeze-fracture images (Nagy et al., 1984) consistent with tighter tight junctions (Wolburg and Lippoldt, 2002), high expression of solute transporters including efflux transporters (Ge et al., 2005; Macdonald et al., 2010; Saubamea et al., 2012), and of certain receptors involved in transcytosis such as transferrin receptor (Ge et al., 2005). Arteriolar endothelium shows higher expression of certain enzymes including 5′-nucleotidase, Mg²⁺-ATPase and Na⁺-K⁺-ATPase than capillary or venular endothelium (Vorbrodt et al., 1982, 1988), and significant absence of P-glycoprotein (Saubamea et al., 2012); bidirectional transcytosis of horseradish peroxidase (creating a local ‘leak’) has been reported in certain brain arterioles but not in capillaries or venules (Westergaard and Brightman, 1973; van Deurs, 1977). The post-capillary venule segment is specialised as a site regulating adhe-sion and traffic of leucocytes into the perivascular space (Bechmann et al., 2007; Owens et al., 2008; Muldoon et al., 2013), shows higher expression of genes involved in inflam-mation-related tasks (Macdonald et al., 2010), and is more affected in inflammatory conditions than capillary endothelium (Paul et al., 2013). Given the much greater area of the cerebral microvascular surface contributed by capillary endothelium compared with arteriolar or venular endothelium (Abbott et al., 2006), preparation of cultures from relatively pure capillary fragments should give the tightest monolayers reflecting most closely the transporting endothelium of the BBB.

3.2. Pericyte contamination

In cultures of rat brain endothelial cells, contaminating pericytes frequently grow in the same plane as the endothelial cells, and are typically surrounded by a cell-free zone leading to holes in the endothelial monolayer (Abbott et al., 1992; Parkinson and Hacking, 2005). By contrast, in the porcine model the pericytes generally grow below the endothelial layer, close to or directly on top of the extracellular matrix (see Fig. 2) (Abbott et al., 1997). Hence high TEER can be achieved even in the presence of a small percentage of pericyte contaminants, since they do not necessarily cause holes in the PBEC monolayer. However, PBECs growing on top of pericytes show a slightly altered morphology, with broader cells and irregular cell borders, compared to the elongated spindle-shaped morphology of PBECs without pericyte growth underneath (Fig. 2). In our experience, treatments to remove pericytes as thoroughly as possible gave the tightest monolayers. Puromycin, substrate of the brain drug efflux transporter P-glycoprotein (P-gp) was used to reduce pericytes contamination. Brain endothelial cells have stronger expression of P-gp than pericytes, so can restrict cellular uptake of the cytotoxic puromycin, while pericytes are more vulnerable, tend to be killed by puromycin treatment (Perrière et al., 2005). Proliferating endothelial cells release platelet-derived growth factor (PDGF) that attracts pericytes, and can lead to vessel (tube) formation and release of vascular endo-thelial growth factor (VEGF) through interactions between endothelial cells and pericytes (von Tell et al., 2006). VEGF increases the permeability of the BBB (Dobrogowska et al., 1998). Therefore, reducing the number of pericytes in the culture favours monolayers rather than vessel formation and leads to uniform monolayers of contact-inhibited endothelial cells with low permeability.

3.3. Supplementation to increase barrier integrity

Supplementation with treatments to elevate cAMP_i was based on a successful protocol for bovine brain endothelial cells (Rubin et al., 1991), and was consistently found to give tighter monolayers in the PBEC model. The treatment of choice now also includes supplementation with hydrocortisone, found to sustain tighter layers in many brain endothelial models (Förster et al., 2008; Hoheisel et al., 1998). In a porcine brain endothelial model developed by Galla and co-workers (Franke et al., 1999, 2000), the presence of ox serum in the medium was found to reduce TEER (Nitz et al., 2003), attributed to the presence of permeabilising factors including lysophosphatidic acid (LPA) and VEGF. Growth fac-tors such as PDGF and VEGF can increase BBB permeability by disrupting tight junctions and stimulating angiogenesis (Dobrogowska et al., 1998; Harhaj et al., 2002; Wang et al., 1996, 2001). To induce better barrier properties, some plasma-derived sera are treated with charcoal to reduce the concentrations of these growth factors. However the charcoal-stripping of serum can lead to removal/reduction of other biologically important factors such as hormones, vitamins, enzymes and electrolytes (Cao et al., 2009). In the present model, we chose to use BPDS, which being derived from adult bovine plasma, is collected with generally less stress to the donor, and contains lower concentrations of growth factors (e.g. PDGF, VEGF) and other vasoactive/proliferative factors than foetal or neonatal calf serum (Abbott et al., 1992). BPDS increased the TEER of the brain endothelial cells compared with serum-free medium, consistent with observations that serum proteins stabilise capillary endothelial permeability, by cross-linking the glycocalyx and possibly also the exposed proteins of the outer zones of the junctional complexes (Curry and Michel, 1980). Where experiments need to be done under serum-free conditions, the monolayers withstand serum removal for 24 h before experiments.

3.4. Comparison of variants

Both mono-culture (Patabendige et al., this issue) and co-culture (Skinner et al., 2009) of the PBEC model variants are capable of giving monolayers of TEER >400 Ω cm². For many applications examining the BBB flux of drug-like molecules and other small solutes, this is sufficient to give good resolution between transcellular and paracellular flux (Gaillard and de Boer, 2000). The relationship between P_app mannitol and TEER observed in our model (Fig. 10) is similar to that reported by Gaillard and de Boer (2000) using two other paracellular permeability markers, sodium fluorescein and 4 kDa FITC-dextran; in our model, P_app was relatively independent of TEER when TEER was >200 Ω cm². As TEER is inversely related to the small ion conductance (and hence permeability) of the monolayer, TEER recorded at the start of an experiment is a good measure of the ‘basal’ paracellular permeability of the cells, as reference for studies e.g. with drugs which may themselves alter permeability. For leakier monolayers, the TEER can be used to derive a corrected permeability coefficient for a drug from the measured P_app (Gaillard and de Boer, 2000); however, when TEER is high enough for P_app to be relatively independent of TEER, the measured P_app is sufficient without correction, and suitable for comparisons between laboratories.

There is an extensive literature showing that exposure to astrocytes or astrocyte-conditioned medium increases the expression of several BBB features in brain endothelial monolayers (Dehouck et al., 1990; Pottiez et al., 2011) so using the astrocyte co-culture or ACM variants of the method may be required for some applications (Gaillard et al., 2001) including those where vesicular-mediated transcytosis of large molecules is involved (Candela et al., 2008; Demeule et al., 2002; Skinner et al., 2009). However, our experience has been that the state of differentiation of the endothelium also plays a large part in maintaining BBB features, and supplementation with hydrocortisone plus elevation of cAMP_i, combined with growth on extracellular matrix mimicking the native brain endothelial/astrocyte basement membrane, without addition of astrocyte-derived influence, may be sufficient for many applications.

3.5. Comparison with PBEC models in the literature

Several promising PBEC models have been introduced over the last decade (Cohen-Kashi Malina et al., 2009; Franke et al., 2000; Smith et al., 2007; Zhang et al., 2006). However, several things drove our development of an alternative to published methods. At the start of this process (early 1990s) there was no reliable published method for generating porcine brain endothelial cells. Since then, several methods have been described, but intra-batch and batch-to-batch variation was still a problem with many of them (Franke et al., 2000; Zhang et al., 2006). There was some variability in the effects of adding serum, reported to either increase or decrease permeability (Nitz et al., 2003), and it was not always clear whether astrocytic influence was necessary. While astrocytes were not required to generate a high TEER in the PBEC model described by Franke et al. (2000), others have reported that astrocytic influence is necessary to produce a practical model (Cohen-Kashi Malina et al., 2009; Smith et al., 2007). In general, where a brain endothelial cell culture model achieves a high TEER without astrocytic influence (Franke et al., 2000; Lohmann et al., 2002; Patabendige et al., this issue; Zhang et al., 2006), functional expression of small solute transporters (SLCs) and efflux transporters is found to be sufficient to allow use of the monocultures for drug permeability assay. For leakier models, co-culture with astrocytes (Cohen-Kashi Malina et al., 2009; 2012; Kido et al., 2002) or C6 glioma cells, or exposure to glial-conditioned medium (Smith et al., 2007) may be necessary to tighten the barrier and improve expression of other BBB properties such as enzymes and transporters, to produce functional assay systems. For certain specialised features of the brain endothelium such as receptor-mediated transcytosis, astrocyte co-culture may be necessary even with tighter monolayers (Skinner et al., 2009).

A detailed comparison of the methods and barrier characteristics of the main PBECs models in comparison to our model is given in Patabendige et al. (this issue). The strengths of the present method are that it is relatively simple, involving fewer preparative steps, and that it gives a high yield. With this method (Patabendige et al., this issue), reliably tight brain endothelial cell monolayers can be grown on inserts without astrocyte influence, and if needed, serum can be removed for the last 24 h to provide suitable starting conditions for experiments, without significantly compromising tightness. The alternative co-culture variant of this method described provides considerable flexibility for experimental design, depending on the application.

The ultimate goal for most BBB researchers is to be able to study the human BBB. However, the difficulties associated with developing robust and realistic in vitro human BBB models have led to the use of animal models (Patabendige, 2012). A porcine BBB model is a good alternative as the biology of the pig is closer than that of other laboratory animals to the biology of the human (Walters et al., 2011). The PBEC model presented in this paper is one of the best BBB models giving high TEER. However, as with all BBB models, there are some limitations. Strict adherence to the experimental procedure is required to produce high yields of pure PBEC cultures and to minimise variation between batches. Only limited in vivo data is available for porcine models compared to rodent models; however, with the increased use of transgenic and miniature pigs this will improve in future. Availability of good porcine primers and antibodies is currently an issue, but this also will improve with the recent publication of a high-quality draft pig genome sequence (Groenen et al., 2012). Further examination of expression and function of transporters and receptors on the PBEC model is currently under way.

In summary, this method combines simplicity and reproducibility with optimum cell yield and purity, making the resulting PBEC model robust, reliable and flexible, with good preservation of BBB features, suitable for a range of appli-cations.

4.1. Time required

• 8 h isolation of brain capillaries and freezing (from 6 pig brains)

• 3 days cell culture

• 1 h passaging of cells onto Transwells filter inserts

• 2–3 days cell culture to confluence

• 15 min medium exchange (‘switch medium’)

• 24 h later: ready for experiments

• The total time required from cell preparation to having suitable cell cultures for mechanistic and transport studies is 8 days, but the cells stay in well-differentiated state for up to 4 days after this.

4.2. Materials

Culture medium L-15 Leibovitz (L-15); medium 199 (M199); DMEM; Penicillin (10,000 U/mL)/Streptomycin (10 mg/mL) (P/S); Glutamine (2 mM stock soln); Heparin; Puromycin; cell permeant cAMP analogue, CPT-cAMP; Hydrocortisone; Trypsin-EDTA for endothelial cells; Hanks’ balanced salt solution (HBSS) without (w/o) Ca²⁺,Mg²⁺; FCS; poly-D-lysine; human fibronectin; dimethyl sulfoxide (DMSO); all from Sigma. Type IV phosphodiesterase inhibitor, RO 20-1724 from Calbiochem/Merck. Enzymes from Lorne Laboratories Limited, UK. Collagenase, Trypsin, DNase I. Minimal essential medium (MEM+HEPES) from MP Biomedicals. Phosphate buffered saline (PBS) with Ca²⁺ and Mg²⁺ from Cambrex Bio Science. BPDS from First Link UK. Nylon meshes (60 µm and 150 µm pore size) from Plastok Associates, UK. Rat tail collagen type I from Becton Dickinson. Tissue culture plastics (flasks, plates, Petri dishes). Filter inserts: Costar ‘Transwell Clear’ 12-well tissue-culture-treated sterile polyester membrane, 0.4 µm pore, 12 mm membrane, pre-loaded on cluster plates. [¹⁴C]sucrose (0.15 µCi/mL final concentration, specific activity 643 mCi/mmol) and [¹⁴C]mannitol (0.20 µCi/mL final concentration, specific activity 56 mCi/mmol) from GE Healthcare. Anti-smooth muscle-specific actin (monoclonal-mouse) from Dako Ltd.; monoclonal antibody against p100/120 from Transduction Laboratories (now BD Biosciences). Anti-mouse secondary antibodies were from Jackson Immunoresearch Laboratories Inc. and nuclear stain Hoechst 33342 was from Sigma. FITC-labelled IB4 was from Gibco, Paisley, UK and ProLong Mounting Medium containing Dapi was from Invitrogen, UK. Lab-made rat-tail collagen (Strom and Michalopoulos, 1982). All other chemicals not quoted specifically were obtained from commercial sources at the highest quality available.

4.3. Laboratory equipment

• •

  Refrigerated centrifuge

• •

  Dounce glass homogeniser (40 mL, with loose and tight pestle)

• •

  Dissecting instruments set: fine forceps (for fine dissection, removal of meninges), curved forceps (for separation of white matter from grey matter), coarse forceps (‘rat-toothed’), scalpel

• •

  Nalgene reusable filter holder unit for 47 mm diameter membranes (500 mL receiver) from Fisher Scientific

• •

  Three clean glass beakers

• •

  Sterile plastics: 1 L containers, 15 cm Petri dishes, 50 mL centrifuge tubes, 50 mL syringe, T75 and T175 flasks, sin-gle-wrapped tissue culture pipettes (10 mL, 25 mL)

• •

  200 mL glass bottle for preparation of ‘Digest Mix’ stock

• •

  60 µm and 150 µm nylon mesh, cut slightly larger than the size of filter unit; gauze and paper towels

• •

  Measurement of TEER: EVOM2 voltohmeter with ENDOHM-12 electrode chamber, and/or STX2 chopstick electrodes, both from World Precision Instruments, USA

• •

  Automatic pipettes+tips (e.g. Gilson)

4.4. Composition of solutions and media

• •

  Transport solution for transferring brains to laboratory. L15 medium with added penicillin (100 U/mL), streptomycin (100 µg/mL) (Pen/Strep).

• •

  Washing and dissection solution: HEPES buffered MEM containing 10% FCS and 1% P/S (% by volume).

• •

  Digest Mix: Enzymes in M199 medium with FCS and P/S. Weigh out 188 mg collagenase (223 U/mg), 86 mg trypsin (211 U/mg), 10.9 mg DNase I (2108 U/mg). Add 178 mL of M199 and filter sterilise into an autoclaved 200 mL glass bottle. Then add 20 mL FCS and 2 mL P/S. Aliquot into centrifuge tubes and freeze at −20 °C.

• •

  Freezing mix: Resuspend vessel fragments in 10% DMSO in FCS and aliquot to cryovials (1 mL each), bring slowly to −80 °C (use a cryo freezing container to achieve −1 °C/min cooling rate; place vials in the container and place it in −80 °C freezer for 24 h), then store vials in liquid nitrogen.

• •

  Basic growth medium: DMEM, 10% BPDS, 1% P/S, 1% Glutamine and 125 μg/mL heparin. Pass through 0.22 µm filter before use.

• •

  Switch (differentiation) medium: Once cells are growing well on inserts, change growth medium to serum-free switch medium containing DMEM, 1% P/S, 1% Glutamine, 125 μg/mL heparin and 550 nM hydrocortisone. Then treat cells with 250 µm CPT-cAMP and 17.5 µm RO 20-1724 (see below for timings).

4.5. Animals and yield

The culture of each batch of cells starts with six pig brains (from abattoir), and generates 12 cryovials each of ‘60s’ and ‘150s’, indicating the filter mesh size used for their isolation. One vial is sufficient for two T75 flasks and cells from two T75 flasks are enough for 18–24 Transwell 12 mm diameter inserts (1×10⁵ cells/insert). Hence six brains yield ∼24×20=480 Transwell inserts with confluent cells.

4.6. Preparations (day before isolation)

• •

  Sterilise dissecting instruments, glass beakers, homogeniser, filter unit, six circles each of 60 µm and 150 µm nylon mesh, gauze and sterile 1 L containers

• •

  Prepare a solution of L-15 medium with 1% P/S (L-15+)

• •

  Prepare a solution of PBS with Ca²⁺ and Mg²⁺ with 1% P/S (PBS+)

• •

  Prepare a solution of MEM/HEPES medium with 10% FCS and 1% P/S (Mem/HEPES+)

• •

  Sterile disposables: Scalpel, cell scrapers, 50 mL syringes, single-wrapped tissue culture pipettes, Petri dishes, centrifuge tubes, labelled cryovials, T75 and T175 flasks

4.7. Detailed experimental procedure

4.8. Co-culture variant

4.9. Assessing barrier integrity and reproducibility: Measuring TEER and mannitol or sucrose permeability

4.10. Immunocytochemistry and cell staining

4.11. Statistical analysis

Data are expressed as mean (±standard error of the mean, SEM) and analysed and presented using GraphPad Prism. Groups of two were analysed using Student's t-test, groups of three or more were analysed using either one-way analysis of variance (ANOVA) with a Dunnets post-hoc test or, if multiple variables were involved, two-way ANOVA with Bonferroni post-hoc test was applied. Values were considered to be significantly different when p<0.05.