Unique and shared inflammatory profiles of human brain endothelia and pericytes

The experiments conducted here were approved by the Northern Regional Ethics Committee (New Zealand) for biopsy tissue and The University of Auckland Human Participants Ethics Committee for post-mortem brain tissue. All methods were carried out in accordance with the approved guidelines. Biopsy human brain tissue was obtained with informed written consent from the patient and family members (Additional file 1). Tissue used in this study was derived from paediatric epilepsy surgery, surgery for adult drug-refractive epilepsy, and normal brain regions resected from patients with deep tumours (Additional file 1: Table S1). This technique provided sufficient endothelial yields for experiments from as little as 0.25 g tissue. Cultures were also attempted from post-mortem brains (Additional file 1: Table S1), however yielded too few viable endothelia for experimentation.

The endothelial cell culture protocol was adapted to work in concert with a previously established method to generate mixed glial cultures from surgical specimens, routinely used in our laboratory [48]. Cultures of endothelial cells were attempted using autopsy specimens as well, but yielded few or no viable endothelial cells. Following collection of tissue, HBSS and meninges were removed. Tissue was then thoroughly mechanically dissociated with a sterile scalpel and enzymatically dissociated (10 mL/g tissue; 10 U/mL DNase I (Invitrogen, CA, USA), 2.5 U/mL papain (Worthington, NJ, USA) in Hibernate-A medium (Gibco, CA, USA)) for 15 min at 37 °C with agitation. The suspension was triturated, and then left to dissociate for a further 15 min at 37 °C with agitation. Digestion was quenched with the addition of one volume of media, either complete DMEM as per [48] (Dulbecco’s modified Eagle’s media with F12 supplement (DMEM:F12; Gibco) containing 10% FBS (Gibco, CA, USA), 1% penicillin/streptomycin (Gibco), and 1% GlutaMAX (Gibco)); neural stem cell medium as per [49, 50] (DMEM:F12 with 1% B27 supplement (Gibco), 1% penicillin/streptomycin, 1% GlutaMAX); or neuronal medium as per Park et al. (In submission) (DMEM:F12 with 1% vitamin A-containing B27 supplement (Gibco), 1% penicillin/streptomycin (Gibco), 1% GlutaMAX (Gibco), and BDNF, NGF, NT3, GDNF, and IGF-1 (Peprotech, NJ, USA) at 40 ng/mL). At this point, the suspension was allowed to settle and the supernatant, containing cell suspension, was passed through a 70- or 100-μm nylon strainer (Bector Dickinson, NJ, USA). The remaining debris and vessel component was triturated and passed through the strainer.

Components of the suspension that passed through the strainer contain other cerebral cell types and can be cultured to obtain pericytes [48], microglia [49], neural stem cells [50], and neurons (Park et al., In submission). Here, the single cell suspension was plated into a T75 and cultured in complete DMEM and left to generate a mixed glial culture containing pericytes, astrocytes, microglia, and endothelia and otherwise treated as per [48]. At passage 2, mixed glial cultures were used as a positive control for GFAP, PU.1, and PDGFRβ immunostaining to compare isolated endothelia in Fig. 1 (Additional file 1: Table S2). Due to the proliferation of pericytes, but not other cell types in mixed glial cultures, serial passaging of mixed glial cultures yields pericyte monocultures, as we have established previously [36]. Pericyte monocultures were used as a negative control for endothelial marker immunostaining in Fig. 2. In experiments where the inflammatory response of endothelial cells and pericytes was compared, only pericyte monocultures were used (Figs. 3, 4, and 5). These were treated as previously described and pericytes used for comparison [51, 52]. Briefly, pericytes were grown in complete DMEM in T75 flasks until confluent. Pericytes were subcultured by trypsinisation (0.25% trypsin-EDTA; Gibco) and seeded at a density of 15,000 cells/cm² for most experiments and 30,000 cells/cm² for ECIS experiments. See Additional file 2: Table S2 for details of epilepsy-derived pericyte monocultures used.

The strainers used to obtain the aforementioned single cell suspension were placed in a sterile petri dish and strainers inverted so that debris could be washed off strainers in complete DMEM. The debris was collected and spun at 170×g for 5 min. Media was aspirated, and the pellet was gently resuspended for in ScienCell Endothelial Cell Medium (ECM; ScienCell, CA, USA) plating into a Matrigel® (thin coating 2.5 μL/cm²; Corning, NY, USA)-coated flask as per the manufacturer’s instructions (See Table 1 for approximate plating guidelines). Vessels and debris were left to adhere overnight, before being washed gently to detach loosely adhered debris. The media was changed to puromycin (0.5 μg/mL; Sigma-Aldrich, MO, USA)-containing ECM. Elongated, cobblestone-shaped endothelia emerged and could be distinguished from other cell types initially present in cultures, including pericytes (large, stellate), microglia (rhomboidal, phase bright), and astrocytes (complex, thin processes). Human brain endothelial cell cultures were maintained in puromycin-containing ECM for 1–2 weeks in order to eliminate other cell types. Brain endothelia are resistant to puromycin-induced toxicity due to their ability to pump it out, and puromycin therefore selects for endothelial cells [53]. Following this, cells were grown in ECM without puromycin for the remainder of their time in culture. Subculture was achieved by detaching endothelia with StemPro® Accutase® (5 min, 37 °C; Gibco), and plating into new Matrigel®-coated plates at a density of 30,000 cells/cm², except for barrier studies where they were seeded at 60,000 cells/cm² onto collagen I-coated plates (1 μg/cm²; Gibco). In general, barrier studies were performed in passages 2–3, while secretion analysis was performed at passages 4–5. Endothelia can be maintained over multiple passages (Range: 3–8 passages) and are amenable to cryostorage in 90% FBS:10% DMSO for extended periods. In order to successfully passage endothelia multiple times, flasks should not be split at a ratio of less than 1:2.

Matrigel® (Corning) was allowed to thaw on ice, before being diluted 1:80 in DMEM:F12 without any additives (Gibco) and kept on ice. Ice cold diluted Matrigel® was added to plates at 2.5 μL Matrigel® per square centimetre of plastic and left at room temperature for at least 1 h. Plates and flasks were then washed three times for at least 30 min in additive-free DMEM:F12, before this was aspirated and surfaces allowed to air-dry overnight. Plates and flasks were then sealed and kept at 4 °C for up to 3 months.

For initial attempts to isolate endothelia, magnetic-activated cell sorting (MACS) based on CD31 expression was used (Additional file 1: Figure S1). Briefly, endothelial cells were cultured as above; however, puromycin was omitted. At the first passage, Accutase (Gibco, CA, USA) was used to dissociate cultures, before being washed in MACS buffer (1% bovine serum albumin, 1 mM EDTA in PBS) and Fc receptor blocking agent added, followed by addition of magnetic bead-conjugated mouse IgG1 anti-human CD31 antibody (Miltenyl Biotech, Cologne, Germany) for 15 min at 4 °C. The cell suspension was washed in MACS buffer and passed through a MACS separation column, as per the manufacturer’s instructions to elute the negative fraction, followed by elution of the CD31-positive endothelial fraction in MACS buffer. The positive and negative fractions were spun at 170×g for 5 min and plated at a density of 30,000 cells/cm² into 96-well plates.

Inflammatory cytokines interleukin-1β (IL-1β), tumour necrosis factor-α (TNFα), lipopolysaccaride (LPS), interferon-γ (IFN-γ), transforming growth factor-β1 (TGF-β₁), interleukin 6 (IL-6), and interleukin 4 (IL-4) or vehicle (a mixture of the vehicles for IL-1β, TNFα, LPS, IFN-γ, IL-6, and IL-4 (0.1% bovine serum albumin (BSA) in phosphate buffered saline (PBS)) and 1 mM citric acid, pH 3 with 0.1% BSA, for TGF-β₁) were added as a 100-fold dilution of a 100× stock to both pericytes and endothelia. See Table 2 for details.

Cells in 96-well plates were fixed in 4% paraformaldehyde for 15 min and washed with PBS with Triton™ X-100 (0.1%; Sigma, MO, USA) (PBS-T). Primary antibodies, diluted appropriately in immunobuffer (PBS with 0.2% Triton™ X-100, 1% goat serum (Gibco, CA, USA) and 0.04% thimerosal (Sigma, MO, USA)), were added overnight at 4 °C (See Additional file 2: Table S3 for details). The following day, secondary antibodies were added at room temperature for 3 h. Hoechst nuclear counterstain (1 μM; Hoechst 33258, Sigma, MO, USA) in Tris-NaCl-EDTA (TNE) buffer was added for 10 min at room temperature. For stains that were sensitive to detergent (zona occludens-1; ZO-1), Triton™ X-100 was excluded until after the primary antibody incubation. Wells were then washed again and counterstained with Hoechst-33258 nuclear stain. For experiments directly comparing endothelial and pericyte inflammatory response, both endothelia and pericytes were seeded and imaged on the same plate. Images were acquired on an ImageXpress Micro XLS™ automated microscope (Molecular Devices).

Proliferation was measured by 5-ethynyl-2′-deoxyuridine (10 μM; Click-iT® EdU AlexaFluor® 647 Imaging Kit; Molecular Probes, CA, USA) incorporation for 48 h at 37 °C, 5% CO₂. Fixed cells that had been immunostained, as described previously, were then washed in 3% BSA in PBS-T twice. EdU staining was visualised using the Click-iT® AlexaFluor™ 647 kit as per the manufacturer’s instructions. Nuclei positive for EdU were scored using MetaXpress™ version 5.3.04 (Molecular Devices) analysis software.

Quantification of cell populations was performed by immunocytochemistry with well-characterised lineage markers. Nuclei positive for PU.1 were designated as microglial using MetaXpress™ software, while in mixed glial cultures, CD31-positive cells lacking PU.1-positive nuclei were designated as endothelia. Remaining cell populations could not be accurately scored by automated analysis and were therefore counted manually. Glial fibrillary acidic protein (GFAP)-positive cells were determined to be astrocytes. The remaining cells were found to be platelet derived growth factor receptor-β (PDGFRβ)-positive and designated as pericytes. In endothelial cultures, the remainder of GFAP-negative, PU.1-negative, and PDGFRβ-negative cells were designated endothelia. Quantification of intercellular cell adhesion molecule-1 (ICAM1) and MCP-1 intensity was performed by setting a low threshold and measuring the integrated intensity of the thresholded pixels, relative to the cell number. Quantification of signal transducer and activator of transcription-1 (STAT1) and similar to mothers against dodecapentaplegic-2/3 (SMAD2/3) was performed by setting a high threshold and scoring the percentage of Hoechst-positive nuclei above the threshold value. Quantification of nuclear factor kappa-light-chain-enhancer of activated B cells p65 (NF-κB) translocation was performed using the translocation assay built in to MetaXpress™ software.

Total RNA was extracted from endothelia or pericytes grown in a 12-well plate using the Ambion RNA Micro-scale extraction kit (Ambion, CA, USA) following the manufacturer’s instructions. For complementary DNA (cDNA) synthesis, 500 ng total cDNA was made per sample. RNA was added to each reaction mix accordingly and diluted to give the same volume, before deoxyribonuclease I (DNase I) from the RQ1 RNase-free DNase kit (Promega, WI. USA) with 1 μg DNase added per 1 μg RNA. The Superscript III First-Strand Synthesis kit (Life Technologies, CA, USA) was used for cDNA synthesis. Quantitative real-time PCR was performed using Platinum SYBR Green qPCR SuperMix-UDG with Rox (Life Technologies, CA, USA) on a 7900HT Fast Real-Time PCR system (Applied Biosystems, Life Technologies, CA, USA) (see Additional file 1: Table S4 for primer details). Relative gene expression was graphed as fold enrichment using the 2^−ΔΔCt method compared to RNA obtained from a single culture of pure pericytes [29].

Electrical cell-substrate impedance sensing (ECIS) is a technique used to measure the resistance to flow of current across a monolayer of cells at multiple frequencies in real time. To determine the barrier properties of isolated human brain endothelial cells or pericytes, experiments were performed on collagen I-coated (1 μg/cm²; Gibco, CA, USA) ECIS 96W20idf plates (0.33 cm² well; Applied BioPhysics, NY, USA) for the ECIS Z ϴ instrument (Applied BioPhysics, NY, USA). Endothelia were then seeded at 60,000 cells/cm² and generated raw impedance values of approximately 2000–2500 Ω when measured at 4000 Hz. Cells were cultured until transendothelial electrical resistance (TEER) had plateaued (between 96 and 120 h post plating) before treatments were performed. Barrier resistance values were obtained using ECIS software (Version 1.2.123.0; Applied Biophysics, NY, USA), using the modelling method described by Giaever and Keese [54], and applied to endothelial barrier measurements elsewhere [55]. Briefly, the changes in impedance relative to a cell-free, media-containing well were measured and used to determine the components of impedance defined by cell shape/granularity (capacitance, measured in nF), cell-substrate contacts (alpha, measured in Ω^-1/2/cm), and cell-cell contacts (R_b, given as TEER in results, measured in Ω cm²). It should be noted that ECIS does not detect comparable levels of resistance to those generated by an EVOM voltohmmeter, which is commonly used to measure TEER, where the units for are in Ω cm⁻², rather than Ω cm²; however, the measurements represent similar properties of the monolayer.

Conditioned media samples from endothelial and pericyte cases were spun at 160×g for 5 min and the supernatant collected and stored at − 20 °C. The concentration of cytokines was measured using a cytometric bead array (CBA; BD Biosciences, CA, USA) as per manufacturer’s instructions (see Table 3 for CBA kit details). CBA samples were run on an Accuri C6 flow cytometer (BD Biosciences, CA, USA). Data were analysed using FCAP-array software (version 3.1; BD Biosciences, CA, USA) to convert fluorescent intensity values to concentrations using an 11-point standard curve (0–10,000 pg/mL) and normalised to cell number as described previously [52, 56].

All experiments were performed in at least three independent cases. In general, data are expressed as mean ± standard error of mean (SEM) from at least three independent experiments. Data visualisation and statistical hypothesis testing was performed using GraphPad Prism® Version 7.00. Two-way analysis of variance (ANOVA) was used when comparing cell type response across a number of stimuli with Tukey’s post hoc adjustment for multiple comparisons, while one-way ANOVA with Dunnett’s multiple comparison adjustment was used when comparing changes in TEER in response to stimuli. For heatmap generation, including unbiased hierarchical clustering, the heatmap.2 function was utilised in ‘R’ software.

After having obtained large debris from tissue dissociation (Fig. 1a–k), blood vessels containing viable endothelia were isolated due to their ability to stick to the Matrigel®-coated plastic. Endothelia were seen to emerge from vessels within 24 h (Fig. 1i). Visual inspection of cultures indicated that although microglia (Fig. 1i, asterisks) and pericytes (Fig. 1i, stars—only observed attached to vessels) were present early in endothelial cultures, they were eliminated within a week due to the toxicity of puromycin. Earlier attempts at isolating endothelia by cell sorting, but without puromycin, had a high degree of pericyte contamination (Additional file 1: Figure S1), and so it is recommended that puromycin treatment be continued for an additional week, and only stopped 1 or 2 days prior to the first passage. Endothelial cultures have a different morphology to mixed glial cultures, with a cobblestone monolayer of spindle-shaped cells (Fig. 1i; endothelial colonies outlined in red). On the other hand, mixed glial cultures contained astrocytes (Fig. 1i, arrows), microglia (asterisks), and pericytes (stellate, phase dark cells, prominent 1 week post plating). The purity of cultures was assessed by staining for markers of endothelia (CD31), microglia (PU.1), astrocytes (GFAP), and pericytes (PDGFRβ). The purity of endothelial cultures was determined to be greater than 99% (Fig. 1m, n). In only one culture, a small population of contaminating pericytes was observed, while no microglia or astrocytes were present. In mixed glial cultures, a population of CD31-positive cells were seen, albeit with weak CD31 staining; however, these were predominantly PU.1-positive and thus categorised as microglia. CD31-positive and PU.1-negative cells were also seen in mixed cultures and denoted as endothelia, though this comprised less than 1% of the population (Fig. 1m, n).

Brain endothelial cells are notable for their high expression of tight and adherens junction proteins, and transporters, and these are of considerable interest in the study of the BBB [4]. Thus, we assessed the presence of the tight junction proteins ZO-1 and claudin-5, along with the adherens junction protein CD31, and endothelial nuclear marker ETS-related gene (ERG). All isolated endothelial cultures expressed these proteins, and staining was not present in negative control pericyte cultures (Fig. 2a). Staining patterns also indicated that ZO-1 was expressed at the junctions of cells, alongside claudin-5 staining (Fig. 2b). Likewise, CD31 expression was higher at the borders of cells, but more widely distributed across endothelia (Fig. 2c). In addition, qPCR studies revealed that genes encoding junctional proteins claudin-5, CD31, occludin, vascular endothelial cadherin (VE-cadherin), transporters glucose transporter-1 (GLUT-1), transferrin receptor (TfR), and P-glycoprotein (P-gp) and endothelial-specific von Willebrand Factor (vWF) were enriched in endothelial cultures, relative to pure pericyte cultures (Fig. 2d). Endothelial expression of the gene encoding the pericyte-specific protein for PDGFRβ was also 1000-fold lower than pericyte cultures, indicating that cultures have minimal pericyte contamination. We also observed that endothelia are proliferative in vitro, as shown by EdU-incorporation (Fig. 2e, f), and this can be increased with VEGF-A treatment (Additional file 1: Figure S2).

We assessed the response of endothelia and pericytes to a panel of immune mediators (IL-1β, TNFα, LPS, IFN-γ, TGF-β₁, IL-6, and IL-4) by examining the activation of the transcription factors NF-κB, STAT1, and SMAD2/3 after 1 h of treatment and the immunostaining of ICAM-1 and MCP-1 after 24 h of treatment. Treatment with IL-1β, TNFα, and LPS caused translocation of NF-κB to the nucleus, with no significant differences observed between endothelia and pericytes (Fig. 3a, b). IFN-γ caused the translocation of STAT1 to the nucleus, with no significant differences observed between the cell types (Fig. 3c, d). SMAD2/3 translocation occurred in response to TGF-β₁ stimulation in pericytes and endothelial cells, although endothelial cells displayed higher basal nuclear SMAD2/3 than pericytes (Fig. 3e, f). No changes in transcription factor localisation were observed following stimulation with IL-4 or IL-6. Representative images from all conditions are given in Additional file 1: Figure S3.

Using cytometric bead array (CBA), we tested the secretions of soluble ICAM-1 (Fig. 4a, k), soluble VCAM-1 (Fig. 4b, i), G-CSF (Fig. 4c, m), GM-CSF (Fig. 4d, n), CX3CL1 (Fig. 4e, o), IL-6 (Fig. 4f, p), IL-8 (Fig. 4g, q), MCP-1 (Fig. 4h, r), IP-10 (Fig. 4i, s), and RANTES (Fig. 4j, t) from endothelia and pericytes in response to the inflammatory panel. Both endothelia and pericytes have a similar response profile to the immune stimulus panel, with slightly higher levels of ICAM-1 secreted by endothelia in response to IL-1β and TNFα. We also detected soluble ICAM-1 in vehicle-treated endothelia, but not pericytes, consistent with ICC data (Additional file 1: Figure S4). On the other hand, CX3CL1 and VCAM-1 were basally secreted at higher levels in pericytes compared to endothelia, but regulated similarly in both endothelia and pericytes in response to the inflammatory stimuli (Fig. 4b, e, i, o). Pericyte secretion of CX3CL1 was also suppressed in response to TGF-β₁ and IL-4 (Fig. 4e, o). Both G-CSF and GM-CSF had similar secretion patterns, being strongly induced in endothelia, but not pericytes (Fig. 4c, d, m, n). Interestingly, in endothelia, IL-1β induced 5–10-fold more G-CSF and GM-CSF than the next strongest inducer, LPS. Both pericytes and endothelial cells secreted IL-6 in response to IL-1β stimulation, with endothelia demonstrating a stronger induction than pericytes (Fig. 4f, p). While the clearest induction of IL-6 occurred in response to IL-1β, it was also induced in both endothelia and pericytes in response to TNFα and LPS. IL-8 secretion was strongly induced by IL-1β, TNFα, and LPS in both endothelia and pericytes, though endothelial secretion of IL-8 in response to TNFα was weaker than pericytes (Fig. 4g, q). Similar levels of RANTES secretion were observed in response to LPS in both endothelia and pericytes, although endothelia secreted greater levels in response to IL-1β and TNFα (Fig. 4j, t). Interestingly, IL-4 induced a greater degree of VCAM-1 and MCP-1 secretion in pericytes than endothelia (Fig. 4h, r). Likewise, pericytes secreted higher concentrations of MCP-1 and ICAM-1 in response to IFN-γ than endothelia (Fig. 4a, k, h, r); however, endothelia secreted greater levels of IL-6 and IP-10 than pericytes (Fig. 4p, s). For a full summary of responses, see Table 4.

Corroborating CBA and immunocytochemistry data, secretions in endothelial and pericyte conditioned media were found to be strongly altered by IL-1β treatment. Both cell types demonstrated induction of a wide range of inflammatory mediators, as measured by secretome profiler array (Fig. 5a, b, d). The majority of the secretions assessed here are similar between the two cell types, with 75 secretions shared (Fig. 5d). There were 15 unique endothelial secretions including soluble CD31, the marker for endothelial cells used above (Fig. 5e). We also identified the secretion of two unique pericyte secretions, insulin-like growth factor binding protein-3 (IGFBP-3) and CD40L, the expression of which have been shown in pericytes and smooth muscle cells previously [56, 57] (Fig. 5e). Interestingly, the IGF-1 binding proteins IGFBP-2 and IGFBP-3 are highly expressed by pericytes, but not endothelia. This experiment was performed twice across two different cases; however, there was a high degree of consistency between the two sets of results (R² = 0.82; Fig. 5c). An additional file provides intensity measures for all cytokines using secretome profiler arrays (Additional file 2).

Inflammation is a critical regulator of the BBB in vivo, so continuous measurement of TEER in primary brain endothelia and pericytes using ECIS was performed to assess functional barrier integrity, alongside immunocytochemistry to test if this was accompanied by a reorganisation of tight junctions. Primary endothelial cells generated consistently high TEER, with maximal TEER between 35 and 55 Ω cm⁻² with raw impedance of approximately 2500 Ω (Additional file 1: Figure S5), unlike pericytes (range 0.1–1 Ω cm⁻², raw impedance approximately 1000 Ω) where barrier resistance was not a large component of the resistance to current (Fig. 6a, b). Measurements indicated that TEER plateaued after day 4 (Fig. 6b). The TEER of primary brain endothelial cultures was rapidly reduced by between 60 and 75% following application of IL-1β, TNFα, LPS, and IL-4 (Fig. 6d, e, g, i, j), with IL-4 and LPS-treated cells showing a rebound effect. We found that treatment of endothelia, but not pericytes, with IL-1β and TNFα resulted in toxicity (Additional file 1: Figure S4). Treatment with TGF-β₁ also reduced TEER by approximately 50% (Fig. 6d, h). Neither IFN-γ nor IL-6 had any significant effect on TEER (Fig. 6d, f, k). These data were reinforced by claudin-5 staining which revealed a reorganisation of tight junctions to granules in cells treated with IL-1β, TNFα, and LPS (Fig. 6c). In cells treated with LPS and TGF-β₁, discontinuous unsealed claudin-5 staining was present, while in IL-4-treated endothelia claudin-5-positive tight junctions appeared to have more diffuse staining (Fig. 6c). In TNFα and IL-4-treated endothelia, there was also a notable change in cell morphology to more spindle-shaped and more rounded, respectively (Fig. 6c).