Development of a multicellular in vitro model of the meningeal blood-CSF barrier to study Neisseria meningitidis infection

Human induced pluripotent stem cell (iPSC) line IMR90-4 (WiCell) was maintained on Matrigel (Corning) coated 6-well plates (Sarstedt) in StemFlex medium (Gibco), passaging twice a week and changing medium daily as previously described [33, 44‐46, 48]. Meningioma cells of the meningothelial histological subtype were derived from tumor biopsies and characterized as previously described [36]. Meningioma cells were grown in DMEM + GlutaMAX (Gibco) with 10% FCS (Gibco) and 1% Pen/Strep (Gibco) and passaged twice a week with a seeding density of 4 × 10³ cells/cm². The cells were expanded and used for experiments up to passage 10 or until showing signs of senescence. hCMEC/D3 cells (Sigma, SCC066) were cultured in EndoGRO-MV Complete Culture Media (Sigma) splitting as needed. hCMEC/D3 and meningioma cells were grown on 30 µg/ml collagen 1 (rat tail, Gibco) coated tissue culture flasks.

iPSC-derived brain microvascular endothelial-like cells (iBECs) were differentiated as previously described [33, 44‐46, 48]. Briefly, iPSCs were seeded from a single cell suspension onto Matrigel coated cell culture flasks (Sarstedt) at a density of 1 × 10⁴ cells/cm2 and expanded in StemFlex medium for 3 days with daily medium changes. Following iPSC expansion, the cells were differentiated in unconditioned medium [UM; DMEM/F-12 (Gibco), 20% Knockout serum replacement (Gibco), 1% minimal essential medium-nonessential amino acids (Gibco), 0.5% GlutaMAX (Gibco), and 0.07% beta-mercaptoethanol (Sigma)] for 6 days, changing medium daily. After 6 days, the medium was changed to basic EC medium [human endothelial cell serum-free media (Gibco) and 1% platelet-poor plasma derived serum (Fisher) or 1% B-27 supplement (Gibco)], supplemented with 20 ng/ml basic fibroblast growth factor (bFGF, ReproTech), and 10 μM all trans-retinoic acid (RA, Sigma), for 2 days. Finally, the differentiated iBECs were purified onto collagen IV (Sigma) and Fibronectin (Sigma) coated plates and transwells inserts (Corning; Greiner). Basic fibroblast growth factor and retinoic acid were removed one day after iBEC purification. Quality of differentiated iBECs was assessed via TEER measurements and immunofluorescence staining of characteristic BEC markers.

On day 8 of iBEC differentiation, meningioma cells were seeded at a density of 3.6 × 10⁴ cells/cm² on the underside of transwell inserts [0.4 µm inserts (Corning), or 3 µm inserts (Greiner)] coated with collagen IV and fibronectin on top, and 150 µg/ml collagen 1 (rat tail, Gibco) on the bottom (direct co-culture). For indirect co-culture, 2.2 × 10⁴ cells/cm² were seeded onto collagen 1-coated cell culture plates, to which collagen IV and fibronectin-coated transwell inserts were added later. After 4 h of incubation at 37 °C and 5% CO₂, transwell inserts were flipped into the designated wells and iBECs were purified on top of the transwell membrane, seeding differentiated cells at a density of approximately 9 × 10⁵ cells/cm² onto the collagen IV and fibronectin matrix. Co-cultures were cultivated in EC medium with Pen/Strep for 1 day, before changing media to basic EC medium without bFGF and RA. Experiments were conducted on day 2 of co-culture (i.e., day 10 of the iBEC differentiation protocol). For co-culture of hCMEC/D3 cells and meningioma cells, transwell inserts were coated entirely with 150 µg/ml collagen 1. Meningioma cells were seeded as described for the iBEC model. After 4 h, hCMEC/D3 cells were seeded on top of the membrane at a density of 9 × 10⁴ cells/cm2. Co-cultures were cultivated for 3 days in EndoGRO medium before performing experiments.

Neisseria meningitidis serogroup B strain MC58—sequence type (ST)-74 [ST-32 clonal complex (cc)], kindly provided by E. R. Moxon [55]—was used in this study. Meningococci were grown on Columbia agar with 5% sheep blood (bioMérieux) at 37 °C and 5% CO₂ overnight. The following day, bacteria were subcultured in liquid proteose peptone medium (PPM)—reshly supplemented with 1% Kellogg’s supplement, 10 nM MgCL₂ and 10 nM NaHCO₃ (ROTH)—at 37 °C and 200 rpm for 60 to 90 min.

One day before conducting infection experiments, antibiotics were removed from the cellular co-culture medium by washing twice with Dulbecco's phosphate-buffered saline (DPBS, Gibco). Medium was changed once on the day of infection, at least 1 h before adding bacteria. Bacterial infection of cells was prepared as described previously [33, 46, 56]. Bacteria grown in liquid subculture were spun down, washed in DPBS, and diluted in cell culture medium prior to infection. A multiplicity of infection (MOI) of 10 was used unless specifically noted otherwise.

Bacterial adherence and invasion of the model endothelial cell layer was evaluated using gentamicin protection assays. BEC layers were infected with N. meningitidis strain MC58 for 2 h or 8 h using an MOI of 100, which is within the range of MOIs used in previously published studies to investigate the interaction of the pathogen with BECs [15, 16, 24, 32, 57]. For enumeration of intracellular CFU per monolayer at the indicated times post-infection, cells were washed once with DPBS and incubated in cell culture medium with 200 µg/ml Gentamicin (Biochrom) for an additional 2 h. To evaluate the total number of cell-associated bacteria, samples were processed immediately at the indicated time points. Briefly, the cells were dissociated and lysed with 0.05% Trypsin–EDTA (Gibco) at 37 °C for 5 min and 1% saponin (SERVA) for 15 min at room temperature (RT). Samples were plated in a dilution series and numbers of bacteria were calculated from CFU counts of countable dilutions. Data are presented as total CFU recovered/monolayer.

Rate of bacterial transmigration was determined by enumeration of CFU that had migrated from the apical to the basolateral compartment of the models within an hour after each indicated time point, using similar methodology as described previously [33]. Transwell inserts with a 3 µm pore size (Greiner) were used to ensure meningococci could pass through the membrane. Inserts were washed twice in PBS and transferred to fresh medium, only replacing the medium on the bottom. After 1 h of incubation at 37 °C and 5% CO₂, samples from the basolateral medium were plated in a dilution series to determine bacterial loads.

Barrier tightness was assessed via transendothelial electrical resistance (TEER) using a volt-ohm meter (Millicell ERS-2, Merck) and sodium fluorescein (NaF) permeability measurements as described previously [48]. Briefly, medium on top of the transwell inserts was replaced with 10 µM NaF (Sigma) and samples were taken from the basolateral medium every 15 min for 1 h. At the final time point, an additional sample was collected from the top compartment, all samples were analyzed in a fluorescence plate reader (Tecan), and NaF permeability (Pe) was calculated as described previously [48].

Immunofluorescence staining of co-culture models was adapted from previously described protocols for iBEC and LMC staining [36, 46, 48]. Cells on transwell inserts were fixed in ice-cold methanol for 15 min, washed with DPBS, and incubated in 10% FCS for 1 h at RT. For detection of epithelial membrane antigen (EMA/Mucin 1) and E-Cadherin, cells were fixed in 4% paraformaldehyde and blocked with 10% FCS containing 0.1% saponin. Transwell membranes were cut out using a scalpel and transferred to wells of a cell culture plate for primary antibody staining at 4 °C overnight (Table 1). The following day, samples were washed in DPBS and secondary staining was conducted for 1 h at RT using Alexa Fluor 488 goat anti-mouse (Invitrogen, ref. A11001), and Alexa Fluor 555 donkey anti-rabbit (Invitrogen, ref. A31572) antibodies at a dilution of 1:200 in 10% FCS. After final washing steps and DAPI staining (Invitrogen, ref. D1306; 1:5000 in DPBS), membranes were mounted on glass slides (mounting medium: Fluoroshield, Sigma) and cellular staining was visualized on an Eclipse Ti2 confocal microscope (Nikon). Image acquisition and analysis was performed using NIS Elements image software version 5.02 (Nikon). Junction coverage was determined for occludin expression using the JAnaP junction analyzer program [58].

After 24 h of infection with GFP-expressing wildtype strain MC58 [26], iBEC-LMC as well as hCMEC/D3-LMC in vitro co-culture models were washed once in DPBS, permeabilized in cytoskeleton buffer (10 mM MES, 150 mM NaCl, 5 mM EGTA, 5 mM glucose, and 5 mM MgCl₂ adjusted to pH 6.11) with 0.25% glutaraldehyde (Sigma) and 0.25% Triton X-100 (ROTH) for 1–2 min at 37 °C, and fixed in cytoskeleton buffer with 2% glutaraldehyde for 10 min at RT. Samples were washed twice and quenched in 0.1% sodium borohydride (Sigma) in PBS for 7 min at RT. Following a third washing step, transwell membranes were cut out and stained with 5 U/ml phalloidin-Alexa Fluor 543 (Invitrogen, ref. A22283) or 100 nM phalloidin-ATTO 643 (ATTO-Tec, ref. AD643-81) overnight at 4 °C. The next day, samples were stained with DAPI (Invitrogen, ref. D1306; 1:5000 in DPBS) and washed in ddH₂O. Transwell membranes were cut into two halves and mounted in ProLong^™ Glass Antifade Mountant (TermoFisher) on clean high precision coverglass slides (BRAND) one half of the membrane with the endothelial cell side facing the coverslip the other with the meningeal cell side facing the coverslip. Mounted samples were cured at RT overnight, then stored at 4 °C.

Structured illumination microscopy (SIM) was performed on a Zeiss Elyra 7 with Lattice SIM² using either a 40x (Plan-Apochromat 40x/1.4 Oil DIC M27) or a 63x (Plan-Apochromat 63x/1.4 Oil DIC M27) oil immersion objective and four different excitation lasers (405 nm diode, 488 nm OPSL, 561 nm OPSL, and 642 nm diode laser). Z-stacks of endothelial or meningeal cell layers were captured in 3D Leap mode with optimal z-steps, controlled by a piezo stage, using appropriate band-pass and long-pass filters sets. SIM imaging was performed with 9 or 13 phase-shifts of the lattice SIM pattern. Reconstruction of the super-resolved images was performed using ZEN 3.0 SR FP2 (black) (Version 16.0.10.306; Zeiss) with SIM and SIM² processing modules. Final images were processed using Imaris 9.2.1. (Bitplane).

Samples were prepared for transmission electron microscopy (TEM) at the indicated time points post-infection. Transwell membranes were washed in DPBS and fixed in 2.5% glutaraldehyde (Sigma) for 1 h at RT, washed three times in 50 mM cacodylate buffer (Roth), cut out using a scalpel and transferred to a glass vial containing cacodylate buffer for further processing. Samples were processed for EM as previously published [59], with the following modification: samples were infiltrated and embedded in Durcupan (Sigma). Electron micrographs were recorded on a JEM-1400Flash transmission electron microscope (JEOL) equipped with a Matataki camera system.

After the indicated time of incubation with bacteria, cells from transwell models were collected using Accutase (Sigma) dissociation for 15 min (iBEC) or 5–7 min (hCMEC/D3). Uninfected cells were used as control. Cell lysis and RNA purification was performed using a NucleoSpin RNA kit (Machery-Nagel). cDNA was generated using LunaScript RT (New England BioLabs). Quantitative PCR (qPCR) was conducted on a StepOnePlus real-time PCR thermocycler (Applied Biosystems) using PowerUp SYBR Green master mix (Applied Biosystems). Primer sequences are supplied in Table 2. Primers were validated by primer efficiency analysis and agarose gel electrophoresis of the qPCR product. qPCR data are presented as fold change over 18S using the cycle threshold (ΔΔC_t) calculation [60].

hCMEC/D3 cells were transfected with siRNA using TransIT-siQUEST^® Transfection Reagent (Mirus) according to manufacturers instructions. hCMEC/D3 were seeded at a density of 2 × 10⁵ cells/well onto a 24-well plate on the day prior to transfection. SNAI1 siRNA (Santa Cruz) or scrambled, FITC conjugated siRNA (Santa Cruz) was added at a final concentration of 50 nM to a mix of EndoGRO basal medium (50 µl/well) and TransIT-siQUEST® transfection reagent (1.5 µl/well) and incubated for 30 min at RT before transfection. After 24 h of transfection, the medium was changed, and the cell layers were infected with Neisseria meningitidis serogroup B strain MC58 at an MOI of 10 for 24 h. Finally, cell lysates were collected for qPCR analysis. To assess cell viability, whole cells were collected, washed in PBS, and stained with 1 mg/ml Propidium iodide (PI) for 10 min before flow cytometry analysis. Non-transfected cells treated with or without 2.5% Triton-X during PI staining were used as controls.

After the indicated time of incubation with bacteria, supernatants were collected from the transwell models and used for detection of IL-8 using the Human IL-8 ELISA Set (BD Biosciences) and following the manufacturers instructions. Briefly, 96-well plates (clear, flat bottom) were coated with capture antibody (1:250 in 0.1 M sodium carbonate, pH 9.5) at 4 °C o/n. After blocking with 10% FCS for 1 h at RT, samples and standard were diluted appropriately in 10% FCS and added to the plate in technical duplicate. After 2 h of incubation at RT in the sealed plate, a 1:1 mix of detection antibody and Streptavidin–horseradish peroxidase conjugate (both 1:250 in 10% FCS) was added for 1 h at RT. Between all mentioned steps, multiple washes were performed using 0.05% Tween in PBS. Finally, the wells were incubated with tetramethylbenzidine and hydrogen peroxide (Thermo Fisher) at a 1:1 ratio for 15–30 min at RT, before the reaction was stopped with 2 N H₂SO₄. Wavelength corrected absorbance (450 nm–570 nm) was measured using an absorbance plate reader (Tecan) and protein concentrations were calculated using a standard curve.

GraphPad Prism version 6.01 (GraphPad Software Inc.) was used for all statistical analysis. 2-tailed Student’s t test was used for pairwise comparison, where appropriate. Analysis of variance (ANOVA) followed by Dunnett’s multiple comparisons test was used for multiple comparisons. Statistical significance was accepted at a P value of less than 0.05.

Brain endothelial-like cells (iBECs) were differentiated from iPSC line IMR90-4 according to previously describe protocols [44‐46, 48] and co-cultured with model leptomeningeal cells, derived from meningioma [36], (LMCs) for two days immediately upon purification (Fig. 1). The co-culture models were characterized using transmission electron microscopy (TEM) and confocal immunofluorescence imaging. At the optimized seeding densities, TEM and imaging of semithin cross-sections revealed morphologically distinct monolayers of iBECs and LMCs on either side of the transwell membrane (Fig. 2a). Using TEM, we observed various electron dense regions between adjacent cells of the dense iBEC monolayer, most likely representing cellular junctions including tight junctions, adherens junctions, and desmosomes (Fig. 2a). The much larger LMCs on the basolateral side displayed a spread morphology with long cellular processes extending out to other cells, and cytosolic components including Golgi, endoplasmic reticulum, vesicles, lamellar bodies, and mitochondria could be visualized via TEM (Fig. 2a). Immunofluorescence staining showed characteristic expression of brain endothelial markers such as CD31, VE-cadherin, and tight junction proteins ZO-1, occludin, and claudin-5 at the cell junctions within the iBEC layer (Fig. 2b). These markers were not observed in the LMC layer apart from a distinct expression of ZO-1 (Fig. 2b; Additional file 1: Video S1). The meningioma derived cells of the meningothelial subtype retained their characteristic morphology and expression of histopathological markers vimentin, desmoplakin, and epithelial membrane antigen (EMA) in co-culture (Fig. 2b). Despite immunopositivity for E-Cadherin, which is reported to be expressed in arachnoid barrier forming leptomeningeal cells [64], these cells did not show junctional localization of the protein. Finally, as current research has provided insight into the importance of the extracellular matrix on BBB development and maturation, particularly focusing on laminin [65], we have staining the iBEC-LMC co-culture model for this class of ECM proteins and found expression on both sides of the transwell membrane (Additional file 2: Fig. S1). Together these results demonstrate that iBECs can be co-cultured with the LMC cells, and staining patterns suggests that both cell types retain their characteristics and remain distinct in culture.

For reference, BEC-LMC co-culture models using the established BEC cell line hCMEC/D3 [42, 43] were developed in parallel and characterized using the same methods. After 3 days of co-culture, confluent layers of both cell types were observed on either side of the transwell membrane and expression of BEC makers such as VE-Cadherin and ZO-1 was detected (Additional file 2: Fig. S2a, b). However, within the hCMEC/D3 layer, we were unable to achieve clear and uniform immunofluorescence staining of junctional BEC markers at the cell–cell borders (Additional file 2: Fig. S2b). While hCMEC/D3s can reportedly lack certain barrier properties such as continuous junctional expression of tight junction components, the model does retain much of the BEC phenotype, especially compared to other human primary or immortalized cell lines, and has been widely used to study N. meningitidis-BEC interaction in vitro [11, 14‐16, 28, 41‐43]. Therefore, we continued to use hCMEC/D3s for reference in this study, reporting on hCMEC/D3 co-culture with LMCs for the first time.

To evaluate effects of LMC co-culture on barrier tightness and integrity of the iBEC model, we measured transendothelial electrical resistance (TEER) and sodium fluorescein (NaF) permeability. TEER measurements were conducted over a period of 14 days after iBEC purification and start of co-culture (without media changes after the switch to EC medium on day 1; i.e., day 9 of differentiation) (Fig. 2c) and NaF permeability was determined on indicated days (Fig. 2d and Additional file 2: Fig. S3b). Both, iBEC monoculture and iBEC-LMC co-culture reached TEER levels of > 2000 Ω x cm² on day 2 of co-culture [40] (Fig. 2c), which corresponds to day 10 of the iBEC differentiation protocol when iBEC monoculture typically reaches peak TEER [46, 48]. At that time, TEER was increased (3389 ± 354 Ω x cm²) and NaF permeability was reduced (1.96 ± 0.38 × 10^–7 cm/s) in the iBEC-LMC direct co-culture models, compared to iBEC monoculture (2820 ± 384 Ω x cm²; 2.85 ± 0.73 × 10^–7 cm/s) (Fig. 2c, d). After that, TEER remained stable and significantly higher in both direct and indirect co-culture for 7 days (Fig. 2c). LMC cells alone, cultured on the transwell underside, did not generate TEER higher than 37 ± 17 Ω x cm², and the increase in TEER continued to be accompanied by a reduction in permeability to NaF (day 4; Additional file 2: Fig. S3). TEER of iBEC monoculture dropped after peaking on day 2 before rising again to near peak levels and, finally, continuously descending after day 8 (Fig. 2c), which is comparable to previously reported data [66]. Continuous reduction of TEER beginning on day 8 was also observed in both co-culture models in parallel to monoculture. However, while TEER reached near-monoculture levels in indirect co-culture on day 12, TEER remained significantly higher with direct co-culture for the remainder of the experiment (day 14; Fig. 2c). In our hCMEC/D3 based models, peak TEER was reached after 3–4 days of co-culture and did not exceed 30 ± 8 Ω x cm² within 14 days (Additional file 2: Fig. S2c), remaining substantially below physiological range [40]. Permeability to NaF was measured at 2.25 ± 0.36 × 10^–5 cm/s for direct co-culture (Additional file 2: Fig. S2d). Elevated TEER was also detected in hCMEC/D3-LMC direct co-culture compared to hCMEC/D3 monoculture, accompanied by lower NaF permeability (Additional file 2: Fig. S2d). However, the maximum TEER was comparable to TEER from LMC cells alone (37 ± 17 Ω x cm²; Additional file 2: Fig. S3a). In summary, LMC co-culture increased BEC barrier tightness and stability over time as reflected by lower NaF permeability and higher TEER. This effect was most pronounced in the high TEER iBEC models and was observed in both direct and indirect co-cultures, suggesting that LMC co-culture improves the barrier function of iBEC monolayers.

To validate the mBCSFB co-culture models for infection studies with N. meningitidis, we first examined N. meningitidis interaction at the BEC layer. BEC layers were infected with N. meningitidis serogroup B strain MC58 after 2 (iBEC models) or 3 days (hCMEC/D3 models) of direct co-culture with LMCs. For comparison, infection experiments were also performed on iBEC and hCMEC/D3 monocultures. Bacterial adherence and invasion were determined 2 and 8 h post-infection (p.i.) using gentamicin protection assays. N. meningitidis strain MC58 significantly adhered to iBECs in iBEC-LMC co-cultures already after 2 h of infection (3.58 ± 1.53 × 10⁶ cfu/monolayer) and adherence increased marginally when infection was carried out for 8 h (1.75 ± 0.80 × 10⁷ cfu/monolayer; Fig. 3a). The number of recovered intracellular bacteria was comparatively low (1.26 ± 0.60 × 10³ cfu/monolayer, 2 h p.i.) and increased over time (2.06 ± 1.12 × 10⁴ cfu/monolayer, 8 h p.i.). The average counts of adherent and invasive bacteria per monolayer were comparable to iBEC monoculture (Fig. 3a). Compared to iBECs, similar numbers of adherent bacteria were detected on hCMEC/D3 layers of hCMEC/D3-LMC co-culture 2 h (5.19 ± 2.25 × 10⁶ cfu/monolayer) and 8 h p.i. (1.39 ± 0.54 × 10⁷ cfu/monolayer) (Additional file 2: Fig. S4a). Counts of recovered intracellular bacteria were lower after 2 h (42 ± 33 cfu/monolayer) and comparable after 8 h of infection (4.43 ± 5.38 × 10⁴ cfu/monolayer). No significant difference in adherence or invasion was observed between hCMEC/D3-LMC co-culture and hCMEC/D3 monoculture (Additional file 2: Fig. S4a). Next, we examined the interaction between N. meningitidis and iBECs of the iBEC-LMC direct co-culture model using TEM. In parallel, iBECs grown in co-culture with LMC and infected with a green fluorescence protein (GFP) expressing mutant of strain MC58 for 24 h, were stained for F-actin and analyzed using structured illumination microscopy (SIM). We detected bacteria tightly associated with iBECs as well as previously described cellular structures such as cytoplasmic protrusions around the adherent bacteria [29, 57] (Fig. 3bI, II), and observed clusters of meningococci, commonly described as microcolonies (Fig. 3c, left panels). Taken together, these results demonstrate that BEC-LMC co-cultures serve as novel, suitable cellular mBCSFB models for infection studies with N. meningitidis.

Next, we investigated the traversal of N. meningitidis across the mBCSFB. After infecting the BEC layers with N. meningitidis strain MC58, we analyzed iBEC-LMC co-culture models using TEM and SIM. Using TEM, we were able to observe few bacteria traversing the iBEC-LMC co-culture model and interacting with the LMC layer at 24 and 30 h p.i. (Fig. 3bIII, IV). SIM analysis of iBEC-LMC models, stained for F-actin after 24 h of infection with GFP expressing N. meningitidis strain MC58, revealed small numbers of meningococci on the apical side of or in the process of crossing the LMC layer on the basolateral side of the model (Fig. 3c, right panels). To quantify bacterial transmigration, we conducted bacterial transmigration assays on BEC-LMC co-culture and BEC monoculture models. 2, 6, 24 and 30 h p.i., bacterial transmigration rates were determined by enumeration of cfu in the basolateral compartment after 1 h of incubation in fresh basolateral media. While bacterial traversal of the iBEC-LMC model was mostly undetectable 2 h p.i., bacteria crossing the barrier were detected more frequently and in greater numbers (> 100 cfu/h) after 6 h (Fig. 3d). At 24 h p.i., transmigration rates had increased substantially (0.16–1.76 × 10⁶ cfu/h) and continued to increase over the 30 h time course of the experiment (0.94–2.92 × 10⁶ cfu/h). Similar results were obtained from infected iBEC monocultures, although measured bacterial traversal was higher at the early infection timepoints (0.43–4.84 × 10³ cfu/h at 6 h p.i.) (Fig. 3d). In the hCMEC/D3-LMC models, high rates of bacterial transmigration were already detected 2 h p.i. (0.02–1.08 × 10⁶ cfu/h) and increased to a lesser degree up to 6 h p.i. (1.52–5.28 × 10⁶ cfu/h) before remaining at approximately the same level (1.64–8.40 × 10⁶ cfu/h, 24 h p.i.) (Additional file 2: Fig. S4b). The numbers of bacteria traversing hCMEC/D3s grown in monoculture were comparable to hCMEC/D3-LMC co-culture data, although slightly higher at 2 h (0.06–1.8 × 10⁶ cfu/h) and slightly lower at 24 h p.i. (0.24–4.8 × 10⁶ cfu/h) (Additional file 2: Fig. S4b).

In previous research, it has been a matter of debate, whether N. meningitidis crosses the brain endothelium at the mBCSFB via a transcellular [17, 29, 31] or a paracellular route [25, 28] or both [33]. To gain more information on the mechanism of barrier traversal in our models, we measured TEER over a 32 h time course p.i. as an indication of barrier integrity and restriction to paracellular transport. In our iBEC-LMC model, TEER remained high for 24 h p.i. (Fig. 3e), although we detected increasing amounts of bacterial transmigration in that time frame (Fig. 3d), suggesting transcellular traversal by N. meningitidis. Additionally, we still observed intact cell-junctions after 24 h of infection using TEM (Fig. 3bII). However, after 24 h of infection, TEER began to drop leading to significant loss of TEER compared to the uninfected control by 30 h p.i. (Fig. 3e), which has also been previously described for iBEC monoculture [33]. We observed similar effects analyzing infected iBEC monocultures, in this study, although TEER was lower overall compared to iBEC-LMC co-culture (Fig. 3e). TEER of hCMEC/D3-LMC co-culture and hCMEC/D3 monoculture models also remained stable for up to 24 h of infection before dropping to significantly lower levels, although total TEER was substantially lower compared to the iBEC model (Additional file 2: Fig. S4c). These data indicate that, while the barrier remains intact for some time, prolonged N. meningitidis infection eventually compromises barrier integrity of BECs at the mBCSFB. Taken together, these results suggest a transcellular route for N. meningitidis traversal of the mBCSFB, although loss of barrier integrity during the later course of infection (> 24 h of infection) may also enable paracellular crossing.

To further investigate the mechanisms behind barrier deterioration upon prolonged N. meningitidis infection, we analyzed expression of cell-junction components in our BEC-LMC co-culture models. First, we examined expression of adherens and tight junction genes in infected BECs from our BEC-LMC co-culture and BEC monoculture models. In the iBEC-LMC co-culture model, expression of genes encoding for endothelial adherens junction protein VE-Cadherin (CDH5) and tight junction proteins ZO-1 (TJP1) and claudin-5 (CLDN5) was downregulated in infected iBECs compared to the uninfected control, especially after 24 h and 30 h p.i. (Fig. 4a). The strongest effect was observed with CLDN5. Expression of CD31 (PECAM1) and occludin (OCLN) was not altered (Fig. 4a). Expression of SNAI1, which encodes for Snail1, a repressor of tight junction gene expression [67‐71], was significantly increased with infection at the indicated time points (Fig. 4a). The same expression profiles were observed with iBEC monocultures (Fig. 4a), consistent with previously reported data on iBECs cultured in cell culture plates [33]. qPCR data from hCMEC/D3 samples grown in co-cultures with LMC are consistent and show reduced PECAM1 expression in addition to downregulation of CDH5, TJP1, and CLDN5 (Additional file 2: Fig. S5). Interestingly, expression of OCLN was increased in hCMEC/D3s with infection, and we observed a tendency towards upregulation of TJP1 and CLDN5 together with downregulation of SNAI1 after 8 h of N. meningitidis infection. This observation could not be correlated with functional changes such as increased TEER, a phenomenon that was, however, occasionally observed at the early infection time points both in iBEC and hCMEC/D3 models (Fig. 3e; Additional file 2: Fig. S4c). Together, these data demonstrate downregulation of tight and adherens junction genes upon N. meningitidis infection.

To assess whether Snail1 plays a role in this mechanism, we performed siRNA knockdown of SNAI1 and analyzed its effect on expression of cell-junction genes in hCMEC/D3 cells with and without N. meningitidis infection (Additional file 2: Fig. S6). SNAI1 knockdown lead to increased expression of CLDN5, CDH5, and PECAM1, which are all Snail1 targets [72]. However, we observed no significant rescue effect of N. meningitidis induced downregulation of these genes. Together, these findings indicate that other mechanisms instead or in addition to Snail1 mediated repression may be responsible for the downregulation of cell-junction genes after N. meningitidis infection.

Previous studies with immortalized BECs grown in monoculture showed that N. meningitidis modulates the localization of adherens junction proteins together with the polarity complex and affects tight junction expression in brain endothelial cells [25, 28]. To assess the effects of N. meningitidis on the localization and expression of occludin, which was not transcriptionally regulated, we analyzed BEC layers of our iBEC-LMC co-culture model after 24 and 30 h of infection with strain MC58 using confocal microscopy and the recently developed junction analyzer software JAnaP [58]. Here, we observed reduced junction coverage 24 h and 30 h p.i. (Fig. 4c). In summary, these results demonstrate that N. meningitidis leads to barrier disintegration in BECs at late time point of infection, although barrier function of BECs was enhanced by LMCs in the mBCSFB model.

Finally, we examined immune activation of BECs in our mBCSFB models in response to N. meningitidis infection, considering that BEC activation presumably contributes to recruitment of leukocytes into the CSF and progression of bacterial meningitis [73]. Using qPCR, we investigated gene expression of neutrophilic chemoattractants and activators IL8 (CXCL8), C-X-C motif chemokine 1 and 2 (CXCL1, CXCL2), C–C motif chemokine 20 (CCL20), and the broad cytokine Interleukin-6 (IL6). In both iBEC-LMC and iBEC monocultures, CXCL8, CXCL1, CXCL2, and CCL20 were upregulated during the time course of infection, with higher upregulation of CXCL8 and CXCL1 in the monoculture model at 24 h p.i. (Fig. 5a). Despite significant increases in mRNA expression, IL-8 was barely detectable by ELISA in the cell culture supernatants (Fig. 5b). Slightly higher protein concentrations were measured in the apical medium of iBECs co-cultured with LMCs. However, this effect may be due to IL-8 secretion by the meningioma derived LMCs, which has previously been described [34]. These observations, including the disconnect between cytokine expression and secretion in iBECs are consistent with previous studies that have used iBECs cultured as monolayers in cell culture plates to model infection with N. meningitidis and Group B Streptococcus [33, 50]. Highly significant upregulation of all cytokines and chemokines examined, including CXCL8, CXCL1, CXCL2, CCL20, as well as IL6, was observed in hCMEC/D3 co-culture with LMC (Additional file 2: Fig. S7a). This effect was consistent with data from hMCEC/D3 monoculture, although the relative fold change of CXCL8, CXCL1, and IL6 expression was a bit lower at certain time points. In contrast to the iBEC models, high protein levels of IL-8 were detected in the cell-culture supernatants of the hCMEC/D3 models (Additional file 2: Fig. S7b). In summary, these results demonstrate that relevant proinflammatory cytokines are upregulated in iBECs and hCMEC/D3s co-cultured with LMCs, comparable to iBEC and hCMEC/D3 monocultures, and suggest that the molecular response of BECs to N. meningitidis infection plays a role in immune activation at the mBSCFB, as previously described [33, 50, 73].