Cytokine/chemokine-induced leakage of immunoglobulins across the blood–brain barrier

In first experiments, we injected several different cytokines and chemokines (IL-1β, TNF-α, IFN-γ, CCL7, CX3CL1, CXCL1, CXCL2, and IL-6) into the striatum of juvenile Lewis rats, and analyzed the integrity of the BBB 18–24 hrs later, using rat IgG leakage into the CNS parenchyma as surrogate marker for barrier dysfunction. We found that the injection of IL-1β, TNF-α, IFN-γ, CXCL2, and IL-6 caused profound leakage of rat IgG (data not shown). These findings raised the question whether pathogenic serum antibodies against AQP4 could also enter in sufficient concentrations to initiate damage to astrocytes. To address this point, we next injected these cytokines into the striatum, and at the same time provided NMO-IgG (J0) or human control IgG by intraperitoneal injections as described [4]. 18–24 hrs after injections, the brains of these animals were examined. We found clear evidence for wide-spread leakage of both rat and human IgG (Figure 1), which was not only confined to the injected striatum, but was also seen throughout the entire ipsilateral hemisphere, affecting cortex, corpus callosum, striatum and thalamus (Figure 1). In all cases, the contralateral sides did not show any evidence for IgG leakage (data not shown).

Entry of inflammatory cells and AQP4 loss at the needle tract

We then studied the effects of this treatment on AQP4 reactivity within the tissue. Only the area immediately adjacent to the needle tract revealed AQP4 loss, but there were no statistically significant differences in the extent of AQP4 loss between the different treatments indicating that these pathological changes at the needle tract were rather caused by the wounding itself than by the action of NMO-IgG at this site (Figure 2). Since all cytokines/chemokines were injected at the same concentration, we next made sure that this concentration was high enough to show biological activity, i.e. to recruit immune cells to the area around needle tract. We found that injection of IL-1β and CXCL2 was associated with a significant recruitment of neutrophils, injection of TNF-α with significant recruitment/activation of macrophages and microglia, and injection of IFN-γ with slight increase in the number of CD3+ T cells, although this did not reach significance (Figure 2). Hence, the cytokine/chemokine concentrations were sufficiently high to reveal effects. Moreover, the effects of IL-1β and CXCL2 were remarkably similar in NMO-IgG seropositive and control IgG positive animals, indicating that, for changes observed at the needle tract, the action of the cytokines was more important than the presence of NMO-IgG.

Formation of additional, perivascular lesions

Although all the animals described above had comparable AQP4-specific antibody titres in their serum (data not shown), they displayed remarkable differences in the formation of additional lesions:

After intrastriatal injection of CXCL2 and supplementation of the immune system with NMO-IgG, 2/9 animals displayed small perivascular lesions with AQP4 loss and neutrophil infiltration. It remained unclear whether they had formed de novo, or whether they represented a continuum with the actual site of injury, since they were found in immediate proximity to the needle tract.

However, after intrastriatal injection of IL-1β into NMO-IgG seropositive animals, 3/5 juvenile and 3/5 adult rats showed small perivascular lesions with neutrophilic infiltrates and AQP4 loss, which developed clearly independently from the needle tract-associated area of AQP4 loss (as revealed by careful analysis of consecutive tissue sections). The frequency of these neutrophil-infiltrated lesions with AQP4 loss was highly variable, ranging from 1–8 per lesion-positive rat (Figure 3), which might explain why we also had some animals in each group where we could not detect them at all. In general, these lesions were scattered throughout the entire ipsilateral hemisphere, and were found in the cortex, striatum and thalamus (Figure 3, Additional file 1), but not in the corpus callosum, possibly due to the fact that AQP4 expression is higher in gray matter than in white matter. In all these lesions, the number of CD3⁺ T cells was extremely low (on average 12 cells/mm²). Neutrophils (on average 732 cells/mm²) and ED1⁺ microglia/macrophages (on average 594 cells/mm²) were the dominating cells (Additional file 1). Complement deposition in these lesions was variable (Additional file 2), correlated well with the loss of GFAP⁺ astrocytes or astrocytic foot processes, but was always less pronounced than the complement deposition seen at the needle tract.

There was no evidence for the infiltration of eosinophils, for demyelination or the presence of myelin degradation products in macrophages within these lesions (data not shown), which is in line with the fact that eosinophil recruitment is not a feature of NMO-like lesions in Lewis rats [4], and with the lack of demyelination within a similar time window in previously published NMO models [3, 4].

All the data described above have been obtained with NMO-IgG derived from one single patient (J 0). To further substantiate our findings, we repeated these experiments with intrastriatal injection of IL-1β, and intraperitoneal application of IgG from 3 additional anti-AQP4 antibody-positive NMO patients, 2 different anti-AQP4 antibody-negative NMO patients, 3 anti-AQP4 antibody-negative MS patients, and from 2 different AQP4 antibody-positive NMO patients after depletion of AQP4-specific antibodies. Lesions with AQP4 loss and neutrophil infiltration were only observed in the presence of anti-AQP4 antibodies of seropositive NMO patients, but not when AQP4-specific antibodies had been depleted from the NMO-IgG preparations, and also not with any other anti-AQP4 antibody-negative human control IgG (Figure 3).

When IL-1β had been injected into the striatum of human control IgG seropositive animals, neutrophil-infiltrated lesions with AQP4 loss were absent, but small blood vessels with large numbers of intraluminal neutrophils were observed (Additional file 1).

The effects of IL-1β on endothelial cells of the BBB

Cytokines/chemokines have only short half-life times [7] and are rapidly redistributed from the parenchyma to the vasculature [8]. Yet, only IL-1β was able to trigger de novo formation of perivascular lesions with neutrophilic infiltration and AQP4 loss distant from the needle tract, which indicated that the BBB became permeable for NMO-IgG and complement. To learn more about the mechanisms involved in this process, we cultured rat brain microvascular endothelial cells and confronted these cells for 22 hours with IL-1β, or with other cytokines/chemokines or vehicle as control. We observed that IL-1β was much more efficient in inducing a strong upregulation of mRNA for the granulocyte-recruiting chemokines Cxcl1 and Cxcl2, and for granulocyte colony stimulating factor Csf-3 than any other molecule tested. In addition, IL-1β also triggered enhanced expression of transcripts encoding the monocyte/macrophage-recruiting chemokines Ccl2 and Ccl5, and the adhesion molecules ICAM-1 and VCAM-1 (Figure 4, Table 1). All these effects were identical in endothelial cells derived from juvenile and adult Lewis rats (Figure 4). The increased production of Cxcl1 and Ccl2 by IL-1β could be further confirmed at protein level, using endothelial cell lysates prepared 12 hours after IL-1β treatment (Figure 4), and the expression of ICAM-1 was confirmed by immunocytochemistry on IL-1β-treated endothelial cells (Figure 4).

The effects of IL-1β on astrocytes and microglia

We next tested whether IL-1β has also effects on astrocytes and microglial cells in vitro. We observed that incubation of astrocytes with IL-1β increased the expression of Cxcl1, Cxcl2, Ccl2, Ccl5 and VCAM-1, while it had little effects on the constitutive expression of ICAM-1 in these cells. In contrast, microglia constitutively expressed Cxcl1, Cxcl2, Ccl2 and ICAM-1 transcripts, possibly due to the baseline activation (and IL-1β production) of these cells under culture conditions. CSF-3 mRNA, however, was not detected in IL-1β treated astrocytes or microglia (Additional file 3).

The effects of IL-1β on the local availability of complement components

As described above, lesions with AQP4 loss and neutrophil recruitment showed a variable extent of complement deposition. Therefore, we studied the effects of intrastriatal cytokine injections on perivascular C1q reactivity in NMO-IgG seropositive animals. The largest number of blood vessels with perivascular C1q⁺ cells was observed in animals subjected to IL-1β treatment, followed by animals injected with TNF-α (Figure 5, Table 1). After the injection of IFN-γ, IL-6 or CXCL2, the number of such vessels was similar to PBS-injected control animals (Figure 5, Table 1).

IL-1β reactivity in active lesions of NMO patients

We analyzed the expression of IL-1β in human NMO lesions of different stages in comparison to multiple sclerosis lesions and controls. In active NMO lesions, which were characterized by immunoglobulin and complement deposition (Figure 6), massive granulocyte infiltration (Figure 6), AQP4 loss and acute astrocyte and tissue injury (data not shown), we found profound expression of IL-1β in activated macrophages and microglia (Figure 6). This IL-1β expression was restricted to active lesions, while no IL-1β immunoreactivity was detected in more advanced NMO lesions, which lacked complement activation and granulocyte infiltration (Table 2). Furthermore, IL-1β expression was below the level of detection in our immunohistochemical staining of active demyelinating lesions of acute (Figure 6) or chronic MS, or in control brains (Table 2). These differences in IL-1β expression were not a feature of lesion sites (in spinal cords/medullas of NMO patients vs. brains of MS patients), because IL-1β expression was essentially absent in initial/active spinal cord and medullary lesions of MS cases 10 and 11, and in late/active lesions in the spinal cords and optic chiasm of patients with progressive MS (cases 12–17) (Table 2). Moreover, also in EAE, lesions in brain and spinal cord essentially show the same pattern of IL-1β reactivity (data not shown).

What could all these findings mean for NMO patients?

We observed that patients with NMO have significantly more IL-1β expressing macrophages in early active lesions than MS patients with stage-matched lesions do. Although the presence of IL-1β in active MS lesions had been reported elsewhere [40, 41], it was below the detection level in a broad spectrum of MS material available in our laboratory, using well-established immunohistochemical methods. On the other hand, the same immunostaining procedure revealed a high number of strongly IL-1β -positive microglia (considerably different from those reported in [40]) in active lesions of NMO patients (Figure 6), as well as in HSE-affected brain tissue (data not shown). The observed pattern of IL-1β reactivity in human CNS material was also confirmed at mRNA level, by in situ hybridization (data not shown). Moreover, other groups also described elevated levels of IL-1β or IL-1β/TNF-α induced molecules like IL-1ra, IL-6, IL-8, IL-13, G-CSF, and Cxcl10 (IP-10) in the CSF of NMO patients, compared to the CSF of patients with multiple sclerosis or other neurological diseases [7, 42].

Enhanced IL-1β expression most likely derives from the inflammatory condition per se [14, 43], where IL-1β is produced by activated microglia cells/macrophages and most likely – in much lower concentrations – also by neutrophils, the most abundant cell type in acute NMO lesions. Although it was for a long time a matter of debate, it is now firmly established that neutrophils do not only transcribe IL-1β mRNA, but are also able to produce and release this protein, essentially only upon stimulation by IL-1β itself or TNF-α [44].

There might be additional amplification mechanims: NMO patients have elevated expression of two key cytokines in their cerebrospinal fluid: IL-6 [7, 16] and IL-17 [45]. Both the production of IL-6 by astrocytes, and the stabilization of IL-6 mRNA are triggered by IL-1β [46, 47]. IL-17, in turn, prolongs the half-life of the mRNA for CXCL1 [48], a molecule induced in endothelial cells by the action of IL-1β. In addition, IL-17 synergizes with TNF-α to cause an enhanced endothelial expression of neutrophil recruiting chemokines and adhesion molecules [49].

Finally, IL-1β can induce further expression of IL-1β [50, 51], which could provide a powerful amplification mechanism to drive neuroinflammatory changes in the brain [13], and it can induce the expression of AQP4 in astrocytes [52], which could increase the amount of target molecules for pathogenic NMO-IgG.

Human CNS samples

Autopsy CNS tissues of patients and control cases archived at the Center for Brain Research, Medical University of Vienna, Austria were used. They included 4 cases with NMO (cases 1–4), 6 cases with acute MS (cases 5–10), 1 case with relapsing remitting MS (case 11), 1 case with primary progressive MS (case 16), 6 case with secondary progressive MS (case 12–15, 17), 1 case with herpes simplex encephalitis (case 18), and 2 cases without evidence of CNS pathology (cases 19 and 20). Studies on archival autopsy tissue were approved by the Ethics Committee of the Medical University of Vienna (EK No. 535/2004/2012).

Animals

All Lewis rats used in this study were obtained from Charles River Wiga (Sulzfeld, Germany). They were housed in the Decentral Facilities of the Institute for Biomedical Research (Medical University Vienna) under standardized conditions. The experiments were approved by the Ethics Committee of the Medical University Vienna and performed with the license of the Austrian Ministery for Science and Research.

Injections into the striatum

3-week old (juvenile) and 7-week old (adult) wildtype Lewis rats were anesthetized with Ketanest S/Rompun and injected into the striatum as described [53], using 0.3 μl solution containing 100 ng/μl of the respective cytokines and chemokines in sterile endotoxin-free PBS. The needle was left in place for additional 10 min before it was removed. Immediately afterwards, some rats were left untreated, while the others additionally received an intraperitoneal injection of patient-derived IgG or control IgG (10 mg/ml; 0.5 ml injected in juvenile animals, 1 ml injected in adult animals). The animals were sacrificed 18–24 hrs later for histological analyses.

Sources of cytokines

The following cytokines were used: rat recombinant IL-1β, IL-6, interferon gamma (IFN-γ), tumor necrosis factor alpha (TNF-α), chemokine (C-X3-C motif) ligand 1 (Cx3cl1) (all from R&D Systems, Minneapolis, MN, USA), rat recombinant chemokine (C-C motif) ligand 7 (Ccl7) and rat recombinant chemokine (C-X-C motif) ligands 1 and 2 (CXCL1 and CXCL2) (all from PreproTech, Rocky Hill, NJ, USA).

Sources and characterization of patient-derived immunoglobulin preparations

Unless otherwise indicated, experiments were performed with the anti-AQP4 antibody containing human NMO-IgG derived from patient J 0 [4]. We further used human immunoglobulin preparations from patients I GF [4], J NMO-IgG6, J NMO-IgG7, J NMO-IgG8 (all AQP4 antibody-positive NMO patients), J 3 and J 4 (AQP4 antibody-negative NMO patients, [4]) and J 5, J 6, and J 7 (AQP4 antibody negative MS patients, [4]). The use of the patients´ plasma for this study was approved by the Ethics Committee of Tohoku University School of Medicine (No. 2007–327) and the Ethics Committee of Innsbruck Medical University (No. UN3041, 257/4.8).

Removal of AQP4-specific antibodies from NMO-IgG preparations

NMO-IgG of 2 NMO patients (I GF and J NMO-IgG6) was depleted of AQP4-specific antibodies as described before [4]. This led to a drop in AQP4-specific antibody titers from 1:2560 (untreated NMO-IgG) to 1:1280 (NMO-IgG absorbed with HEK-EmGFP cells) to 1:320 (NMO-IgG absorbed with HEK-AQP4/EmGFP cells) for I GF, and from 1:1280 (untreated NMO-IgG) to 1:640 (NMO-IgG absorbed with HRK-EmGFP cells) to 1:320 (NMO-IgG absorbed with HEK-AQP4/EmGFP cells) for NMO-IgG6.

Tissue sampling

The animals were sacrificed by inhalation of an overdose of CO₂. Blood was drawn for the analysis of antibody titers in the serum. Then, the animals were perfused with 4% paraformaldehyde (PFA), the brains were carefully dissected, post-fixed for 24 hrs in 4% PFA, and paraffin embedded.

Histological analysis

2–4 μm thick adjacent serial sections were cut on a microtome. All stainings were done as described before [54], using the following primary antibodies: polyclonal goat anti-human IL-1β (1:2000, Santa Cruz Biotechnology, Heidelberg, Germany), polyclonal rabbit anti-rat AQP4 (1:250, Sigma-Aldrich, Vienna, Austria), biotinylated sheep anti-human IgG (1:200, Amersham GE Healthcare, Vienna, Austria), donkey anti-rat IgG (1:1500, Jackson Immunoresearch, West Grove, PA, USA), rabbit anti-rat C9neo (1:2000, [55]), rabbit anti-rat C1q (1:100, kindly provided by Sara Piddlesden), polyclonal rabbit anti-cow glial fibrillary acidic protein (GFAP, cross-reactive with rat; 1:3000; DakoCytomation), monoclonal mouse anti-rat ED1 (1:10000), monoclonal rabbit anti-human CD3 (cross-reactive to rat CD3, 1:2000, Thermo Scientific, Vienna, Austria), monoclonal mouse anti-rat W3/13 (1:50, Harlan Sera-Lab), and rabbit anti-Iba1 (1:3000, Wako Chemicals, Neuss, Germany).

Immunohistochemistry was completed by incubation with corresponding biotinylated secondary antibodies (donkey anti-rabbit, 1:2000, sheep anti-mouse, 1:500, both antibodies from Jackson ImmunoResearch; donkey anti-sheep/goat, 1:200, Amersham GE Healthcare), followed by exposure to avidin-peroxidase complex (1:100 in DB/FCS; Sigma). Enhancement of the CD3 staining was performed using biotinylated tyramine amplification [56]. Labeling was visualized with the AEC system (in case of C9neo or C1q) or with 3,3’ diaminobenzidine-tetra-hydrochloride (DAB, Sigma) containing 0,01% hydrogen peroxide. All sections were counterstained with Meyer’s hematoxylin, dehydrated and mounted in geltol (sections developed with the AEC system) or Eukitt (Sigma; all other sections).

For conventional staining, the sections were dewaxed in xylol for 30 min, rehydrated, and stained with hematoxylin/eosin.

For immunofluorescent stainings, the sections were heated for 1 hr in a commercial food steamer using 10 mM EDTA buffer pH 9.0. The sections were then blocked with DAKO Antibody Diluent (DAKO), and goat polyclonal anti-IL-1β (1:125, Santa Cruz Biotechnology) and rabbit anti-Iba1 (1:1500, Wako Chemicals, Neuss, Germany) were applied in the same solution overnight, at 4°C. This was followed by incubation with biotin-conjugated donkey anti-sheep/goat (1:200, Amersham Biosciences), streptavidin-Cy2 (1:75, Jackson ImmunoResearch) and donkey anti rabbit-Cy3 (1:100, Jackson ImmunoResearch) in DB/FCS.

Quantitative analysis

All histological measurements and cell counts were made using standardized microscopic fields defined by an ocular morphometric grid. To determine the area of AQP4 loss around the injection site, a final magnification of 200 x was used. The grid was positioned over the widest area of the lesion, perpendicular to the needle tract. For final calculation, the area of AQP4 loss caused by wounding (i.e. the needle tract proper) was subtracted from the total area of AQP4 loss. The number of cells recruited to the parenchyma adjacent to the needle tract was also determined at a final magnification of 200 ×. For quantification of cell numbers in newly formed perivascular lesions, a final magnification of 400 × was used, and cross sections of blood vessels were centered in the grid.

Endothelial cell cultures

Rat brain endothelial cells of 7-9-week old (adult) or 3-week old (juvenile) Lewis rats were cultured essentially as described [57]. To reach high purity of these cultures, 3 μg puromycin (Sigma)/ml culture medium were added [58] for the first 3 days in culture. Afterwards, the culture medium was replaced by endothelial cell medium (PAA) supplemented with 2 ng/ml recombinant human basic fibroblast growth factor (R&D Systems) and 500 ng/ml hydrocortisone (Sigma).

Immunocytochemical characterization of endothelial cells

Immunocytochemical analysis were done as described [54], using polyclonal rabbit anti-rat zonula occludens 1 (ZO-1, 1:50, Invitrogen by Life Technologies, Vienna, Austria) and monoclonal mouse anti-rat ICAM-1 (1:200, AbD Serotec, Kidlington, UK) as primary, and donkey anti-rabbit Cy5 (1:200, Jackson Immunoresearch) or biotinylated donkey anti-mouse (1:1500, Jackson ImmunoResearch) as secondary antibodies, the latter followed by incubation with DyLight™ 488-conjugated streptavidin (1:75, Jackson ImmunoResearch).

Cultures of astrocytes and microglial cells

These cells were isolated from neonatal Lewis rats and propagated/purified essentially as described [54].

Chemokine/cytokine treatment of cells

5-7-day old primary rat brain endothelial cell cultures were washed once with PBS and transferred to endothelial medium without supplements. Since this medium did not contain serum, serum starvation prior to chemokine/cytokine treatment was not necessary. 10 ng/ml of cytokines/chemokines were added, and the culture was continued for 12 (antibody array), 22 (gene expression analysis), or 24 hrs (immunocytochemistry). For antibody array analyses, 1 μl of protein transport inhibitor solution (BD GolgiPlug™, BD Biosciences) was included in the medium. After cytokine treatment, cells were washed three times with PBS and subjected either to RNA or protein isolation, or to immunocytochemical staining.

Microglia and astrocytes were cultured under serum-free conditions over night, and were then incubated with 10 ng/ml IL-1β for 22 hrs.

Antibody array analysis

Cells were washed with PBS. Proteins were isolated and analysed using the RayBio® Rat Cytokine Array 2 kit (RayBiotech, Norcross, GA, USA) according to the manufacturer’s instructions.

RNA isolation and cDNA synthesis

RNA was isolated with the RNeasy kit and QIAshredder (both from Qiagen, Vienna, Austria) according to the instructions of the manufacturer. First strand cDNA was synthesized using M-MLV Reverse Transcriptase (Promega, Mannheim, Germany), as suggested by the manufacturer. Afterwards, the cDNA was used directly for polymerase chain reactions (PCR).

PCR analysis

The following primer pairs were used: Cxcl1 (forward 5^′-AAGGGTGTCCCCAAGTAATGG-3^′; reverse 5^′-CCTTCTTCCCGCTCAACACC-3^′), Cxcl2 (forward 5^′-CACCAACCATCAGGGTACAGG-3^′; reverse 5^′-GAGGCACATCAGGTACGATCC-3^′), CSF-3 (forward 5^′-TTGCCACCACCATCTGGC-3^′; reverse 5^′-ACTGCTGTTTAAATATTAAACAGGG-3^′), Ccl2 (forward 5^′-CACTCACCTGCTGCTACTCATTCA-3^′; reverse 5^′-GCTTGAGGTGGTTGTGGAAAAG-3^′), Ccl5 (forward 5^′-CTGCTGCTTTGCCTACCTCTCC-3^′; reverse 5^′-GATAGCATCTATGCCCTCCCAGG-3^′), Icam1 (forward 5^′- GGGTTGGAGACTAACTGGATGA-3^′; reverse 5^′-GGATCGAGCTCCACTCGCTC-3^′), Vcam1 (forward 5^′-GAGACAAAACAGAAGTGGAAT-3^′; reverse 5^′-AGCAACGTTGACATAAAGAGT-3^′) and GAPDH (forward 5^′-GGCATTGCTCTCAATGACACC-3^′; reverse 5^′-TGAGGGTGCAGCGAACTTTAT-3^′). The FastStart Taq DNA Polymerase kit (Roche Applied Science, Vienna, Austria) was used for amplifications. One reaction consisted of: 5 μl 10x PCR buffer (200 mMTris–HCl, pH 8.4, 500 mM KCl), 1 μl 10 mM dNTP mix, 1 μl forward primer (100 pmol/μl), 1 μl reverse primer (100 pmol/μl), 0.4 μl polymerase (5 U/μl), 1 μl cDNA and 40.6 μl H₂O. The reaction mixture was subjected to an initial denaturation step (11 min, 95°C), and then to 25, 28, 30 or 35 cycles of denaturation (30 s, 95°C), annealing (30 s; 55°C for Cxcl1, 57°C for Cxcl2 and Csf3, 56°C for Ccl2 and Icam1, 59°C for Ccl5, 49°C for Vcam1, and 53°C for GAPDH) and elongation (30 s, 72°C). The reaction was terminated with final extension for 10 min at 72°C, and PCR products were detected by agarose gel electrophoresis.

Statistics

The following statistical evaluations were performed using the PASW statistics 18 software system (SPSS Inc., Chicago, USA): One-way ANOVA, followed by Dunett T3 post-hoc test, and Kruskal-Wallis followed by Mann–Whitney U test and Bonferroni-Holm correction.