Blood-brain barrier hyperpermeability precedes demyelination in the cuprizone model

All animal studies were performed in compliance with the animal policies of the Max Planck Institute of Experimental Medicine, and were approved by the German Federal State of Lower Saxony. Adult male C57BL/6N mice or CX3CR1^GFP/GFP mice [26] were taken at 8–10 weeks of age. Animals were randomly assigned to an experimental group. Cuprizone (0.2% w/w, Sigma) was mixed into powder chow (V1124 ssniff).

Histological analyses were done as described [11, 49] with minor modifications. Anesthetized mice were perfused with paraformaldehyde (PFA) and brains were cut by vibratome (40 μm, Leica VT1200)) or embedded in paraffin. Brain sections (HMP 110, MICROM) at Bregma −1.58 were taken for histological characterization using standard protocols using LSAB2 (Dako) or Vector Elite ABC (Vector Labs). For occludin and ZO1 staining, animals were perfused with PFA containing 0.2% glutaraldehyde and cut on a vibratome. Tissue sections or fixed endothelial cells were processed for immunolabeling by permeabilization (0.4% Triton X-100 in PBS), blocking (4% horse serum, 2% BSA, 0.2% Triton X-100 in PBS) and incubation with first antibody (1% HS, 0.05% Triton X-100 in PBS). Incubation with fluorophore coupled secondary antibodies (ThermoFisher) and DAPI (4′,6-diamidino-2 phenylindole) were done in 1.5% HS in PBS, after which sections were mounted in AquaPolymount (Polysciences). Gallyas silver impregnation was done as described [49]. Specimens were analyzed on an Axio Imager.Z1 (Zeiss) equipped with an AxioCam MRc3, × 0.63 Camera Adaptor and the ZEN 2012 blue edition software using ×10 objective (Plan Apochromat × 10/0.45 M27) or × 20 objective (Plan-Apochromat × 20/0.8) or by confocal laser scanning microscopy (Leica SP5 HCX PL APO CS 63×/1.20) using the Leica Confocal Software (Leica Microsystems). Quantification of positive areas in the corpus callosum above the fornix were done by semi-automated analysis with ImageJ software macro and color deconvolution plug-in. Vessel paint was performed as described [34] with minor modifications. Mice were intravenously injected with 200 μl of 20 mg/ml FITC-Dextran (46,945, Sigma-Aldrich Inc., Germany). After 30 min circulation time animals were anaesthetized, flushed, stained with DID (ThermoFischer, D7757) and fixed with PFA before sectioning with a vibratome (100 μm). All Images were processed with NIH ImageJ and Adobe Photoshop CS5.1 software. Electron microscopic analysis was done as previously described [49]. Briefly, tissue was fixed in 4% PFA, 2.5% glutaraldehyde, 0.1 M Phosphate buffer and sagittal sections were cut on a vibratome (Leica VT1200). The corpus callosum with adjacent tissue was punched and embedded in epon (LYNXII, EMS). Ultrathin uranyl acetate contrasted sections were imaged with a LEO EM912 AB (Zeiss) equipped with a 2 k–CCD camera (TRS, Moorenweis).

Tracer injections were done as described [11, 49] with minor modifications. For measurement of BBB permeability, tracers were i.v. injected (Evans blue 50 mg g^-1 body weight; sodium fluorescein 200 mg g^-1 body weight). After 4 h incubation, animals were perfused with PBS to remove tracer from the circulatory system. The region of interest was dissected, and tissue samples were weighed. For lyophilization, tissues were exposed to a shelf temperature of −56 °C for 24 h under vacuum of 0.2 mBar (Christ LMC-1 BETA 1–16). Samples were weighed for calculation of water content and edema. Lyophilized tissue samples were extracted with formamide at 57 °C for 24 h on a shaker at 300 rpm (Eppendorf Thermomixer). Integrated density of tracer fluorescence was determined in triplicates after 1:3 ethanol dilution to increase sensitivity. Tracer concentration was calculated using a standard curve of tracer spiked brain samples.

Brain tissue samples were lysed in sucrose buffer (18% sucrose, 10 mM Tris/HCl pH 7.4, 1 mM sodium hydrogen carbonate, 1 mM magnesium chloride, 0.1% Triton, 0.2% lithiumdodecyl sulphate, 0.025% sodium deoxycholate) with protease inhibition (Roche) using a Precellys 24 homogenizer (Bertin technologies). Detection of immunolabeled proteins was performed with ECL detection reagent (Perkin Elmer) using ChemoCam Imager (Intas).

Glial cells were isolated according to the adult brain dissociation protocol (Miltenyi biotec). Corpus callosum and cortex were isolated using a brain matrix from Bregma +1.10 to −2.46. Antibody labeling steps were done according to the respective antibody Microbead kit protocol (Miltenyi biotec), oligodendrocytes (O4, 130–096-670); astrocytes (ACSA-2, 130–097-679), microglia (CD11b, 130–093-636), and endothelial cells (CD31, 130–091-935). Purity of cell populations was routinely determined by qPCR on extracted and reverse transcribed RNA (see below) and revealed only minimal contamination by other cell types.

Primary mouse brain endothelial cell cultures were established from 7 days old mice or rats. Briefly, cortices were digested with 1 mg/ml collagenase/dispase and 2.5 μg/ml DNAse (Roche) in dissection buffer (HBSS, 10 mM HEPES, 0.5% BSA, 5000 U/ml penicillin/streptomycin) for 45 min at 37 °C. After trituration, cells were resuspended in 25% BSA and centrifuged at 1000 g for 20 min to pellet microvessels. Isolated microvessels from individual mice were plated in Endobasal Medium (Promocell) with 0.4% puromycin for positive selection on coverslips or polyester transwell inserts (Corning). Primary astrocyte cultures were prepared from 0 to 2 days old mice as previously described [11]. Primary microglia cultures were prepared from P0-P2 old mice by differential shaking of mixed glial cultures as described [55]. Cell purity was routinely determined by immune stainings and always exceeded 95%. For coculture experiments, endothelial cells cultured in transwell inserts above astrocytes plated on the bottom of the well plate. Confluent cells cultures were treated with a final concentration of 250 μM cuprizone in 0.125% DMSO or in 0.125% DMSO alone for up to 72 h. An epithelial Voltohmmeter (EVOM2, World Precision Instruments) equipped with Endohm-12 chamber electrodes was used to measure transendothelial electrical resistance (TEER). Metabolic activity was determined using a WST1 assay (Cayman) according to the manufactures protocol, after exposure to increasing concentrations of cuprizone (0–250 μM) for up to 72 h, or 20 μM peroxide as positive control.

Expression analyses were done as described [11]. For tissue expression analyses, corpus callosum and cortex was dissected from Bregma +1.10 to −2.46. RNA was extracted using QIAshredder and RNeasy protocols (Qiagen). Concentration and quality of RNA was evaluated using a NanoDrop spectrophotometer and RNA Nano (Agilent). cDNA was synthesized with Superscript III (Invitrogen) and quantitative PCRs were done in triplicates with the GoTaq pPCR Master Mix (Promega) on a 7500 Fast Real-Time PCR System (Applied Biosystems). Expression values were normalized to the mean of two housekeeping genes, HPRT (Hypoxanthin-Phosphoribosyl-Transferase 1) and Rplp0 (60S acidic ribosomal protein P), and quantification was done by applying the ΔΔCt method, normalized to age matched untreated controls (set to 1). All primers (Additional file 1: Table S1) were intron-spanning.

CAII (Said Ghandour), Olig2 (Charles Stiles/ John Alberta), GFAP (Chemicon), MAC3 (Pharmigen), Iba1 (Wako), PECAM1 (dianova), AQP4 (Santa Cruz), occludin, ZO-1 and claudin-5 (Thermo Fisher), GAPDH (Enzo). For generation of GLUT1 antisera, rabbits were immunized with the C-terminal intracellular peptide (CDKTPEELFHPLGADSQV). Anti-GLUT1 antibody was purified by affinity chromatography.

Statistical analysis was performed using Prism software (GraphPad Software), and results are presented as the mean ± s.e.m.. Two-way ANOVA, one-way ANOVA and two-tailed unpaired Student’s t tests were performed as appropriate. Only P values <0.05 were considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001).

At the peak of demyelination after 5 weeks of cuprizone exposure, vascular permeability is increased as we showed previously [11]. To explore potential causes for this BBB instability, we performed a series of expression analyses on dissected corpus callosum samples from mice fed cuprizone for 5 weeks. As expected, histopathology (Additional file 2: Figure S1a) was reflected in reduced expression of oligodendroglial genes, e.g. Olig2 (oligodendrocyte lineage transcription factor 2) and increased expression of markers for gliosis (Gfap, glial fibrillary acidic protein; Aif1, allograft inflammatory factor 1, encoding Iba1) (Fig. 1a, Additional file 1: Table S2). When we quantified the levels of inflammatory mediators that have previously been shown to induce BBB dysfunction [42, 47, 58], Tnf (tumor necrosis factor), Il1b (interleukin 1 beta), and Ccl2 (C-C Motif Chemokine Ligand 2) were strongly upregulated (Fig. 1b). Other factors such as Il2 (interleukin 2) and Ifng (interferon gamma) remained unchanged compared to controls. Notably, BBB impairment in the cuprizone model was not associated with mRNA upregulation of HIF1α (hypoxia induced factor 1 alpha), VEGF-A (vascular endothelial growth factor A) or matrix metalloproteinases, which are jointly associated with hypoxia-induced BBB dysfunction in EAE and other disease models [5, 6] (not shown). In contrast, mRNA abundance of the nitric oxide synthases NOS2 and NOS3 was significantly increased in cuprizone treated animals, implicating reactive nitrogen species likely mediators of BBB leakage. Consistent with these findings, expression of endothelial markers including genes encoding tight and adherens junction proteins PECAM1 (platelet/endothelial cell adhesion molecule 1/CD31, Pecam1 gene), claudin-5 (Cldn5), occludin (Ocln), ZO1 (Tjp1 gene), cadherin-1 (Cdh1), and cadherin-5 (Cdh5) were strongly downregulated (Fig. 1a).

When we examined the morphology of the BBB by transmission electron microscopy, cuprizone fed animals often displayed a discontinuous electron-dense junctional area with focally increased junctional width (Fig. 1c, Additional file 2: Figure S2), implying that the physical barrier of the BBB mediated by endothelial tight and adherens junctions could be altered. This observation was confirmed by immunofluorescence analysis of selected tight junction proteins that showed strongly reduced staining intensity in cuprizone fed animals (Fig. 1d–g). Increased BBB permeability was demonstrated by extravasation of tracers such as FITC-dextran (Fig. 1f, see also below). By electron microscopy we also observed locally split endothelial and astroglial basement membranes, hypertrophic astrocytes, and sporadic endothelial cells with atypical ultrastructure in cuprizone fed animals (Fig. 1c, Additional file 2: Figure S2). Moreover, while the number of blood vessels in the entire cross-sectional area of the corpus callosum remained unchanged, vessel density was strongly reduced (Fig. 1h), likely caused by the tissue swelling due to the dramatic gliosis at this time point (Additional file 2: Figure S1). Together, these data show that the increased BBB permeability at the peak of cuprizone induced demyelination coincides with substantial upregulation of BBB-disrupting pro-inflammatory mediators. This is associated with strong downregulation of endothelial tight junction proteins and morphological changes at the endothelial barrier.

We next explored whether cuprizone directly damages cellular constituents of the NVU, namely mouse brain endothelial cells or astrocytes, as observed for mature oligodendrocytes [10]. Using primary cultures of either cell type in a WST1 assay that measures the activity of cellular dehydrogenases, cuprizone did not affect metabolic activity within 24 h, even when administered in concentrations up to 500 μM (Additional file 2: Figure S3a, b). However, after 72 h incubation with 250 μM cuprizone, the WST1 signal in endothelial cells decreased by about 14% compared to vehicle treated cultures (Additional file 2: Figure S3c, d), which were not caused by cell death but rather reflected metabolic adaptations to the toxic cuprizone insult.

To investigate the effect of cuprizone on the barrier function of endothelial cells in more detail, we analyzed transendothelial electrical resistance (TEER) in an in vitro BBB system of endothelial monocultures. Surprisingly, cuprizone significantly decreased TEER after only 48 h that dropped further to about 80 ± 3% (mean ± SD) of control values after 72 h (Fig. 2a). This effect was also observed in the advanced BBB setup, in which endothelial cells were co-cultured with primary astrocytes (Fig. 2b). Moreover, by immunostaining and expression analysis of primary endothelial cells, we observed reduced protein and mRNA abundance of the tight junction protein occludin, and reduced Abcb1a mRNA of the ABCB1 (P-glycoprotein) efflux transporter, likely explaining the decreased tightness of the barrier in vitro (Fig. 2c, d). Abundance of mRNAs for the astroglial BBB maintenance factor sonic hedgehog and for other tight junction proteins, which were strongly downregulated in the corpus callosum of cuprizone treated animals (Fig. 1a, d, e), remained unchanged. Cuprizone treatment downregulated occludin in the absence of inflammatory mediators, whose expression in primary endothelial, astrocyte, and microglial cultures were not induced by the cuprizone challenge (Additional file 2: Figure S3e-g). Together, these in vitro data suggest that cuprizone could directly affect BBB constituents in vivo, even in the absence of additional disease processes.

To test this, we treated mice for 5 weeks with cuprizone as before and then analyzed the integrity of endothelial tight junctions on cortical sections. The cortex develops the demyelinating pathology later than the corpus callosum (Additional file 2: Figure S1), as previously reported [13, 23], allowing us to test if tight junction integrity is affected before pathology exacerbates. In the cortex of cuprizone treated animals and controls, staining intensity and continuity of the tight junction proteins occludin and ZO1 was similar (Fig. 3a, b), in contrast to the strongly reduced staining intensity in corpus callosum of the same animals (Fig. 1d–g). Similarly, abundance of the tight junction protein claudin-5 (Fig. 3c) was not significantly reduced in cortex (90 ± 2% of controls in cortex lysates) of cuprizone fed mice in comparison to corpus callosum lysates (44 ± 4% of controls, n = 3, 2-way ANOVA P < 0.0001). Of note, abundance of claudin-5 in control animals was 69 ± 3% in cortex compared to corpus callosum (n = 3, n = 3, 2-way ANOVA P < 0.0001), in line with a study using porcine brain [41]. In accordance with the mild gliosis (Additional file 2: Figure S1), vessel density remained unchanged in cortex of cuprizone treated mice compared to controls (Fig. 3d). These findings suggest that in vivo, cuprizone alone is not sufficient to completely abolish BBB integrity (see also below).

To substantiate the heterogeneity of BBB pathology with another marker for NVU integrity, we compared Aquaporin 4 (AQP4) distribution in corpus callosum and cortex by immunostaining. AQP4 localization was restricted to astrocyte endfeet in control animals and the cortex of mice up to 5 weeks cuprizone (Fig. 3e, f), suggesting this brain region intact with respect to BBB integrity. In contrast, in the corpus callosum of these mice, AQP4 displayed a diffuse staining pattern, and Aqp4 mRNA expression was significantly upregulated (2.71 ± 0.01 fold, n = 4 animals per group, P < 0.0001, Student’s t-test), a condition also found in MS lesions [50] and typical for cerebral edema in EAE. Together these findings suggest that the morphological disturbances at the BBB are associated with local active disease in cuprizone fed animals. This prompted us to investigate at which disease state BBB dysfunction develops.

To determine the temporal progression of BBB impairment, we simultaneously assessed the degree of BBB permeability and vasogenic edema after 3 days to 5 weeks of cuprizone administration (Fig. 4a). Brain water as a measure of edema was already significantly evident after 3 days cuprizone and progressively increased (Fig. 4b). BBB dysfunction, as assessed by the biochemical extraction of extravasated Evans blue or sodium fluorescein (Fig. 4a), developed with a similar time course as edema, demonstrating a significant 1.25 ± 0.02 fold extravasation (mean ± s.e.m., n = 4) after 3 days of cuprizone (Fig. 4c, d). Because global edema values could be misleading, we determined water content individually, in corpus callosum and cortex of the same animals after 5 days of cuprizone exposure. While the water content in the cortex of cuprizone treated animals remained in the same range as in controls, the corpus callosum of cuprizone treated animals displayed a 16.6 ± 0.5% water increase (Fig. 4e, mean ± s.e.m., N = 4–5), a value typical for focal edema [28]. Corresponding with these local differences in edema, focal extravasation of fluorescent tracers increased more than twofold in the corpus callosum but only marginally in the cortex of cuprizone treated animals (Fig. 4f, g).

To assess how this regional heterogeneity of edema and BBB permeability correlates with local pathology and gliosis in this early disease phase, we performed a series of quantitative expression analyses on dissected corpus callosum and cortex together with histopathological evaluation of tissue sections, after 5 days of cuprizone. Expression of marker proteins for astrocytes and microglia reflected the mild histopathological changes (Fig. 4h, Additional file 1: Table S3, Additional file 2: Figure S1). Despite reduced Plp1 and Olig2 expression, mice lacked signs of demyelination and loss of oligodendrocytes, as reported previously [13, 24]. The abundance of mRNAs of endothelial tight junction proteins were only moderately downregulated in cortex compared to corpus callosum, mirroring the immunostaining data at 5 weeks of cuprizone exposure (Fig. 4h, compare Figs. 1d-e, 3a). Expression of the ABCB1 transporter and associated caveolins as well as of the BBB maintenance factor sonic hedgehog remained almost unaffected in the cortex compared to untreated wild type mice. It is possible that a direct effect of cuprizone contributed to the residual downregulation of endothelial markers and to the slight BBB hyperpermeability (Fig. 4g), similar to our observations in vitro (compare Fig. 2). In contrast, all markers for NVU integrity were strongly downregulated in corpus callosum at this early time point of cuprizone induced pathology.

Levels of the inflammatory mediators Il6, Il1b, Tnf, and Ccl2 were strongly elevated in the corpus callosum but only moderately increased in the cortex after 5 days of cuprizone (Fig. 4i). These data demonstrate that the BBB integrity is already affected within the first days of cuprizone exposure, coinciding with elevated levels of inflammatory mediators but preceding overt demyelination and oligodendrocyte loss.

These findings prompted us to test directly whether demyelination and oligodendrocyte loss or local gliosis and the secretion of inflammatory mediators correlate with BBB dysfunction. Therefore, we used type 3 CXC chemokine receptor (CXCR3) deficient mice [26] that develop demyelination in response to cuprizone as wild type mice but show strongly reduced reactive gliosis and expression of pro-inflammatory cytokines and chemokines such as TNF, IL6 and CCL2 [32].

After 5 days of cuprizone, we found attenuated expression of markers for astrogliosis and microgliosis as well as inflammatory mediators in CXCR3 deficient corpus callosum compared to identically treated wild type animals (Fig. 5a, b). Expression of the oligodendroglial transcription factor Olig2 and the myelin protein PLP1 was ameliorated in CXCR3 deficient mice (Olig2, 3.86 ± 0.23 fold; Plp1, 2.28 ± 0.05 fold in CXCR3 knockout mice compared to cuprizone fed controls), suggesting that the oligodendroglial damage was slightly less severe at this time point. Interestingly, the strong downregulation of genes indicative of BBB dysfunction such as tight junction proteins and BBB maintenance factors was also ameliorated in CXCR3 deficient animals (Fig. 5c). Reduced brain edema (Fig. 5d) and attenuated extravasation of fluorescent tracers (Fig. 5e, f) in CXCR3 deficient animals further support the hypothesis that pro-inflammatory mediators contribute to BBB disruption in response to cuprizone exposure.

Although CXCR3 is mainly expressed by microglial cells in untreated mice [26], and also when mice are challenged with cuprizone (Additional file 2: Figure S4), the cell type responsible for establishing the cytokine milieu during initial cuprizone pathology that contributes to BBB dysfunction is unknown. Therefore, we acutely isolated microglia, astrocytes, oligodendrocytes, and endothelial cells from wild type and CXCR3 deficient mice after 5 days of cuprizone treatment and from untreated wild type control animals, and analyzed mRNA abundance of Tnf, Il1b, Il6, and Ccl2. We chose these inflammatory mediators because their expression pattern correlates with the extent of BBB disturbances: after 5 days of cuprizone, their expression levels were most strongly increased in corpus callosum of wild type mice, moderately increased in the corpus callosum of CXCR3 mutant animals, and only weakly upregulated in the cortex of wild type animals compared to untreated wild type controls (compare Figs. 4i and 5b). Oligodendroglia did not significantly contribute to the cytokine and chemokine profile after 5 days of cuprizone and surprisingly, neither did microglia (Fig. 5g, Additional file 1: Table S4). Endothelial cells showed moderate upregulation of cytokine and chemokine expression. In contrast, we identified astrocytes as the major source of all tested pro-inflammatory mediators at this early disease phase (Fig. 5g). Further, the increased expression of inflammatory molecules was completely abolished in astroglia of CXCR3 deficient mice, suggesting that microglial CXCR3 signaling induces astroglial upregulation of cytokines and chemokines in response to cuprizone.