Functional Consequences of Neuromyelitis Optica-IgG Astrocyte Interactions on Blood-Brain Barrier Permeability and Granulocyte Recruitment¹

Neuromyelitis optica (NMO)³ or Devic’s disease is a recurrent disease of the CNS that affects mainly the optic nerves and spinal cord (1). NMO pathology is characterized by necrotic lesions with perivascular inflammatory infiltrates enriched in granulocytes. Ig deposition associated with products of complement activation suggests a pathogenic role of humoral-mediated immune mechanisms (1). A disease-specific serum autoantibody (NMO-IgG) has been shown to bind to structures adjacent to CNS microvessels, pia, subpia, and Virchow-Robin spaces (2). The target Ag has been identified as being aquaporin-4 (AQP4), the main water channel protein in the CNS (3). AQP4 is mainly expressed on astrocyte end-feet at the blood-brain barrier (BBB) and brain-cerebrospinal fluid barrier (4). Analyses of AQP4-deficient mice revealed its involvement in cerebral water and ion homeostasis regulation, astrocyte migration, neural signal transduction, and demonstrated its role for the maintenance of BBB integrity (5, 6).

A series of clinical and pathologic observations suggest that NMO-IgG plays a central role in the physiopathology of NMO: 1) plasma exchange can produce an early therapeutic response in NMO-IgG-positive patients (7, 8, 9); 2) the Ig deposition and activated complement components in NMO lesions are in a vasculocentric distribution corresponding to sites of high AQP4 expression (10, 11, 12); 3) AQP4 and glial fibrillary acidic protein (GFAP) immunoreactivity are lost in most of the NMO lesions, especially in active perivascular regions (12, 13); and 4) anti-AQP4 Ab titers correlate with the clinical and radiological severity of the disease and decrease after efficient treatment (14). Nevertheless, the mechanisms by which NMO-IgG contributes to disease pathogenesis remain to be defined. In this study, we assessed the functional effects of sera from NMO-IgG-positive patients on human primary astrocytes and brain endothelial cells (ECs), and defined novel functions on BBB permeability and granulocyte recruitment that could contribute to the development of the pathological features described in NMO.

We selected sera from seven adult patients (22–81 years of age) in whom NMO-IgG was detected by indirect immunofluorescence on primate cerebellum slides (Binding Site, Birmingham, U.K.), five patients with NMO, and two patient with longitudinally extensive transverse myelitis. Four of the NMO sera samples were assayed by the classical test on mice tissues described by the Mayo Clinic Neuroimmunology Laboratory (Rochester, MN), and NMO-IgG positivity was confirmed in all samples tested. Nine age-matched patients (age 29–55) with clinically definite multiple sclerosis (MS) and eight healthy subjects (age 24–65) were selected as control. Although meeting the diagnostic criteria for clinically definite MS, MS4 only presented bilateral brain stem lesions with no spinal cord or supratentorial lesions on one patient’s magnetic resonance imaging scan and could possibly still develop into clinical NMO. All MS and healthy control sera samples were negative by indirect immunofluorescence on primate cerebellum slides. All patients gave informed consent, and all studies were approved by the ethics committee of McGill University following Declaration of Helsinki principles. Anti-AQP4 was purchased from Alomone Laboratories, FITC-conjugated goat anti-human IgG from Jackson Immunoresearch Laboratories, PE-conjugated anti-CD107a, FITC-conjugated anti-MHC class I, and goat anti-rabbit Ig from BD Biosciences, anti-von Willebrand factor VIII from DakoCytomation, Alexa Fluor 488-conjugated anti-GFAP, Alexa Fluor 594 goat anti-rabbit, and Alexa Fluor 633 goat anti-mouse Abs from Molecular Probes, PE-conjugated anti-human, Cy3-conjugated and nonconjugated anti-GFAP Abs from Sigma-Aldrich and FITC-conjugated anti-rabbit Ab from Chemicon International.

Human fetal astrocytes (HFAs) were isolated as previously described (15) from human CNS tissue from fetuses at 17–23 wk of gestation obtained from the Human Fetal Tissue Repository (Albert Einstein College of Medicine, Bronx, NY) following Canadian Institutes for Health Research-approved guidelines. Cultures were determined to be >90% pure (16). Human BBB-derived ECs (BBB-ECs) were isolated from CNS tissue obtained from temporal lobe resections from young adults undergoing surgery for the treatment of intractable epilepsy, as previously described (17, 18).

Primate cerebellum sections (The Binding Site) or primary cells grown on Thermanox coverslips (Nunc) were fixed in 4% paraformaldehyde for 10 min, washed in PBS and blocked for 1 h in HHGB (HBSS containing 10% goat serum, 2% horse serum, 2% FBS, 1 mM HEPES buffer, and 0.1% sodium azide). For intracellular staining, cells were permeabilized in PBS containing 0.5% Triton X-100 at room temperature for 15 min. After incubation with primary Abs and diluted sera (1/50) overnight at 4°C in blocking solution, the substrates were washed in PBS, labeled at room temperature with species specific fluorescent secondary Abs in blocking solution, washed in PBS, incubated 30 min with TO-PRO-3 (Invitrogen) for nuclear staining, mounted in Prolong antifade mounting medium (Molecular Probes), and examined with a Leica TCS 4D confocal scanning microscope.

HFAs were detached using PBS containing 1 mM EDTA, incubated with anti-MHC class I Ab or test human sera (1/50) in FACS buffer (PBS containing 1% FCS and 0.1% sodium azide), washed and incubated with PE-labeled goat anti-human Ab. Intracellular staining was performed by fixing and permeabilizing the cells for 10 min at room temperature with paraformaldehyde/saponin buffer (4% (w/v) paraformaldehyde, 0.1% (w/v) saponin, and 10 mM HEPES in HBSS). After washes with saponin buffer (0.1% (w/v) saponin in FACS buffer), cells were incubated with Alexa Fluor 488-conjugated anti-GFAP Ab. Isotypes matched for concentration of primary Abs were used for all stainings. Cells were washed, fixed in 1% formaldehyde containing PBS, and analyzed on a FACSCalibur (BD Biosciences).

Venous blood was drawn from normal volunteer donors in agreement with institutional guidelines. PBMCs were isolated using Ficoll-Hypaque density gradient centrifugation (GE Healthcare), and NK cells were subsequently purified using allophycocyanin-conjugated anti-NKP46 Abs and anti-allophycocyanin MACS beads (Miltenyi Biotec) following the manufacturer’s instructions. Cells obtained (CD3⁻ NKP46⁺) were >95% pure. Granulocytes were prepared by density centrifugation using Polymorphprep (Axis-Shield), according to the manufacturer’s instructions. Cells obtained were >90% pure.

CD107a mobilization assay was done as previously described (19). NK cells (effector) were used at an E:T ratio of 1:1 with HFAs (target) preincubated for 1 h in culture medium containing 10% of patient or control sera. After washes, HFAs were left in 100 μl of medium to which was added 100 μl of NK cells, 1 μl of 2 mM monensin, and 1 μl of PE-conjugated CD107a Ab or concentration-matched isotype control IgG1-PE. After 5 h at 37°C, cells were washed and analyzed on a FACSCalibur.

HFAs were plated in 96-well plates and incubated overnight at 37°C with 1–10 μCi/well of ⁵¹Cr (MP Biomedicals). After washes, HFAs were preincubated 1 h in culture medium containing 10% of patient or healthy control sera. After washes, NK cells were added at an E:T ratio of 1:1 and incubated for 4 h at 37°C. HFA lysis was assessed by radioactivity measurement in a gamma counter (Matrix 9600; Canaberra Packard) as previously described (20). Fc receptor blockade on effector cells was performed by preincubating NK cells for 1 h at 37°C with 50 μg/ml polyclonal human IgG (Caltag Laboratories). Serum IgG depletion was performed using HiTrap protein G HP columns (GE Healthcare) according to the manufacturer’s instructions.

BBB-ECs and HFAs were used to generate an in vitro model of the human BBB. HFAs were plated on the bottom side of the Transwell filter insert of a gelatin-coated 3-μm pore size Boyden chambers (BD Biosciences) positioned upside down in EC culture medium and allowed to adhere for 1 h. The insert was inverted and placed in a 24-well cell culture plate. On day 3, BBB-ECs were plated on the topside of the insert. On day 6, 5% patient or control sera were added to both chambers. FITC-BSA (Invitrogen) (50 μg/ml) was added to the upper chambers and its diffusion across the BBB model was monitored using a FL600 microplate fluorescent reader (BioTek). BBB permeability was calculated using the following formula, as previously published (21): (BSA lower chamber) × 100/(BSA upper chamber).

HFAs were plated in 24-well plates and BBB-ECs were plated on the Transwell filter insert of gelatin-coated 3-μm pore size Boyden chambers. On day 3, HFAs were incubated 1 h in culture medium containing 10% of patient or control sera. After washes, 10% of fresh serum was added in the lower chamber as a source of complement and purified granulocytes were added into the upper chamber. When mentioned, complement was inactivated in the added fresh serum by heating 1 h at 56°C. After 2 h, granulocytes that had migrated through the EC was counted in the lower chamber, stained with an anti-CD107a Ab and analyzed by flow cytometry. Sera used as a source complement were tested in our FACS binding assay to exclude the presence of unspecific anti-astrocyte Abs.

Data handling and analysis were performed using Prism 3.0 (GraphPad Software). Due to small sample size, non-parametric tests (two-tailed Mann-Whitney tests) were used to compare the reactivity of sera derived from the total NMO, MS, and healthy controls (see Figs. 3, 5, and 6). For remaining experiments (see Figs. 4 and 7) in which serum samples were selected from a homogeneous larger population of healthy controls or MS patients (i.e., all negative for anti-AQP4 Abs) and NMO patients (i.e., all positive for anti-AQP4 Abs), parametric statistic tests were applied. Permeability curves (see Fig. 4) were evaluated with repeated-measures two-way ANOVA followed by a Bonferroni posttest comparing all pairs of data points. All data points are expressed as mean from at least n = 3 experiments, each conducted in duplicate. Granulocyte recruitment (see Fig. 7) results were compared using a Student’s t test.

To ascertain that NMO-IgGs bind primate CNS tissues in similar pattern as in rodent tissues (2, 3), we tested five NMO and two longitudinally extensive transverse myelitis patients’ sera by indirect immunofluorescence on sections of normal primate cerebellum. We observed the characteristic perivascular staining pattern previously described for rodent tissues (2); staining was prominent in the white matter and the granular layer where microvascular elements are the most abundant (Fig. 1,A). Dual immunostaining with anti-AQP4 Ab and NMO sera showed that the Ag recognized by NMO-IgGs colocalized with the AQP4 at the BBB around the capillary endothelium (Fig. 1,B). No specific staining was observed with MS or healthy control subject sera (Fig. 1, A and C).

We first demonstrated by confocal microscopy that primary HFAs express AQP4 and are recognized by NMO-IgG (Fig. 2). The NMO-IgG binding colocalizes with AQP4 Ab labeling at the membrane of HFAs. Sera from healthy control subjects or MS patients produced only nonspecific intracellular staining. Flow cytometry analysis confirmed that HFAs, identified as GFAP- and MHC class I-positive cells, are recognized both by a commercially available anti-AQP4 Ab and by NMO-derived sera, but not by MS sera (Fig. 3,A). We found that sera from all seven NMO-IgG-positive patients stained the surface of HFAs (Fig. 3,B); mean percentage of HFAs stained was significantly increased in the NMO sera group when compared with MS sera or to healthy control sera (61 ± 12% vs 6.8 ± 10.7% and 4.8 ± 2.2% respectively; p < 0.0001) (Fig. 3 C). Under similar staining conditions, there was no significant binding of NMO sera or anti-AQP4 Ab to neurons (data not shown).

To address whether NMO-IgG binding could alter BBB function, we used an in vitro model of the BBB in which HFAs were grown above a monolayer of human BBB-derived ECs (BBB-ECs). In contrast to HFAs grown on a plastic substrate where such cells are large with few and short extensions (Fig. 4,A, left), HFAs cultured in contact with BBB-ECs present a smaller cell body and extend long processes (Fig. 4,A, right). We could show by confocal immunofluorescence that whereas AQP4 appeared regularly distributed on the membrane of HFAs in monoculture, it displayed a polarized expression at the end of HFA processes when cocultured with BBB-ECs (Fig. 4,B). To investigate the effect of NMO-IgG binding on AQP4 expression, HFA BBB-EC cocultures were incubated for 24 h in the presence of 10% of NMO-IgG-positive sera. Although the level of AQP4 expression on astrocytes did not change significantly (assessed by FACS, data not shown), the pattern of expression was markedly altered, giving rise to an intracellular and punctate immunostaining pattern with decreased AQP4 expression on the astrocyte processes (Fig. 4,B), probably indicating an internalization of AQP4 molecules in intracytoplasmic vacuoles. Incubation with healthy control sera did not alter the polarized AQP4 expression (Fig. 4 B).

To assess the effect of altered AQP4 expression on BBB permeability, we applied NMO-IgG-positive sera to cocultures of human BBB-ECs and HFAs, grown on alternate sides of a Boyden chamber membrane, our in vitro BBB model. Five percent (final concentration) of test sera were added in both the upper and the lower chambers and fluorescent-BSA permeability was measured at different time points over 72 h. Incubation with NMO-IgG sera induced a significant increase in BBB permeability to FITC-BSA at 48 and 72 h compared with MS sera (p < 0.001 and p < 0.001, respectively; n = 3) (Fig. 4 C).

To test whether AQP4 targeting by NMO-IgG could induce ADCC of astrocytes, we assessed the expression of the degranulation marker CD107a on the surface of human NK cells (effector cells) cocultured with HFAs (target cells) in the presence or absence of NMO-IgGs. The FACS profile presented in Fig. 5,A illustrates the increased percentage of NK cells expressing CD107a following contact with HFAs preincubated with NMO-IgG-positive sera. As shown in Fig. 5, B and C, we found that sera from six of seven NMO-IgG-positive patients induced significant CD107a surface expression by NK cells compared with only one of nine MS sera (p < 0.0001) and none of the four healthy control sera (p = 0.0002). We found a correlation between the ability of serum IgGs to bind HFAs (NMO-IgG positivity) (Fig. 3,B) and their efficiency in inducing NK cell degranulation (p = 0.0002) (Fig. 5 D).

We directly measured the ability of NK cells to kill HFAs in a chromium release assay. After 4 h, preincubation with 10% NMO-IgG-positive sera induced 10 times more NK-mediated astrocyte killing when compared with MS sera (14.7 ± 6.7% vs 1.4 ± 3.7; p = 0.006) (Fig. 6,A). Serial dilution of NMO-IgG serum showed a clear dose response correlation (Fig. 6,B). This lytic response was inhibited when Fc receptors were saturated by preincubation of NK cells with irrelevant polyclonal human IgG (“Fc block”) (6.2 ± 0.75%; p = 0.002) (Fig. 6,C). Furthermore, serum IgG depletion using a Sepharose-protein G column completely suppressed the astrocyte killing (0 ± 0.4%; p = 0.001). Conversely, the NMO-IgG-concentrated eluate fraction restored the killing (Fig. 6,C). Interestingly, the only MS serum able to induce astrocyte killing by NK cells was one of the two sera shown to slightly bind HFAs by FACS analysis (serum MS4) (Figs. 3,B and 6,A). We further found a significant positive correlation between serum IgGs binding to HFAs (NMO-IgG positivity) (Fig. 3,B) and their ability to induce NK-cell mediated astrocyte-directed ADCC (p < 0.0001) (Fig. 6,D). Finally, we demonstrated in this model that cell surface mobilization of CD107a by NK cells is strongly correlated with their cytotoxic activity (p < 0.0001) (Fig. 6 E). We could not induce HFA lysis by incubation of the cells only with NMO-IgG containing serum and fresh serum as a source of complement (data not shown), suggesting that effector cells are required in situ to propagate Ab-dependent cytotoxicity of astrocytes.

To examine whether NMO-IgG binding to astrocytes would lead to complement-dependent granulocyte attraction, we used a BBB-EC-coated Boyden chamber based assay in which the bottom chamber contained HFAs preincubated with either NMO-IgG-positive or healthy control sera together with fresh serum as a source of complement. Granulocytes were added to the upper chamber and allowed to migrate across BBB-ECs for 2 h. Compared with healthy control sera, HFAs preincubated with NMO-IgG-positive sera induced a significant increase in granulocyte migration through BBB-ECs (total migrated cells = 86 × 10³ ± 1.7 × 10³ and 161 × 10³ ± 23 × 10³ respectively; p < 0.01) (Fig. 7,A). Heat inactivation of the serum inhibited the chemoattraction induced by NMO-IgGs (total migrated cells = 48 × 10³ ± 10 × 10³; p = 0.001) (Fig. 7,A). Granulocytes that had migrated toward NMO-IgG-incubated HFAs exhibited increased CD107a mobilization compared with migration toward HFAs incubated with healthy control sera (percentage of CD107a-positive cells = 23.8 ± 8.9 and 12.8 ± 4.8 respectively; p < 0.01) (Fig. 7, B and C). Although complement inactivation inhibited granulocyte attraction, it did not suppress their degranulation (percentage of CD107a-positive cells = 22 ± 5.8) (Fig. 7 C), indicating that granulocyte degranulation was not related to complement activation, but rather to the contact with NMO-IgG-coated astrocytes at the bottom of the well.

We initiated our study by identifying sera, obtained from NMO and longitudinally extensive transverse myelitis patients, which showed in situ immunoreactivity using primate cerebellar slices. The vasculocentric pattern of immunoreactivity obtained was similar to that previously reported in studies using rodent cerebellar tissue sections and to that obtained using anti-AQP4 Ab (2, 3). We used these sera to demonstrate that NMO-IgGs bind to primary human astrocytes in vitro. We further show that NMO-IgGs have functional effects on the cellular components of the neurovascular unit, promoting BBB opening and granulocyte recruitment, both of which are features of NMO.

CNS microvasculature ECs need constant input from the glial cells to maintain their BBB-related properties (22, 23). Our data suggest at least two different ways by which interaction of NMO-IgGs with astrocytes could alter BBB function. The first would be to interfere with AQP4 expression or function. To test this hypothesis we used HFA/BBB-EC cocultures as a model of BBB. In these cocultures, astrocytes present long processes with AQP4 molecules concentrated at their end-feet, reproducing the polarized expression at the BBB seen in situ. This polarized expression was completely disrupted after incubation with NMO-IgG-containing sera; these sera seemed to induce AQP4 internalization, similarly to a phenomenon observed by Lennon and colleagues (24) using non-neural tissue-derived cell lines. As AQP4 is known to regulate water fluxes in the brain, it was likely that this altered expression would influence BBB permeability. Accordingly, altered BBB integrity associated with hyperpermeability has been observed in AQP4 knockout mice (6). To directly test this hypothesis, we cultured HFAs and BBB-ECs on alternate sides of a Boyden chamber membrane and found that NMO-IgG-containing sera induce a significant increase in BBB permeability. AQP4 has recently been implicated in the adhesion and the migration of astrocytes providing additional potential mechanisms by which AQP4 internalization could alter BBB integrity (25, 26, 27). These direct effects of NMO-IgGs on the neurovascular unit could account for the relatively rapid clinical efficacy of plasma exchanges when used to treat NMO (7, 8, 9).

We then explored a second mechanism by which AQP4 targeting by NMO-IgGs could promote a breach in the neurovascular unit, by sensitizing astrocytes for ADCC. As a proof of principle for these studies, we used NK lymphocytes as the effectors, as these cells are known to secrete cytotoxic granules and be mediators of ADCC following Fc receptor engagement. Although the exact influence of NK cells in the development of NMO lesion has not been defined, presence of Fc receptor bearing effector cells (e.g., macrophage-microglia, granulocytes) within NMO lesions is well known (28). Our data confirm that when NK cells are cocultured with HFAs in the presence of NMO-IgGs, NK cells degranulate, as measured by expression of surface CD107a, and astrocyte death ensues. Taken together, our findings are consistent with neuropathological features of NMO demonstrating necrotic inflammatory lesions associated with a loss of AQP4 and GFAP immunoreactivity, vasculocentric deposition of immune complexes, and macrophage-microglia activation (12, 13, 29, 30). We did not observe any binding of NMO sera IgGs with AQP4 negative cells such as neurons and mature oligodendrocytes (data not shown). We found a strong correlation between the ability of serum IgGs to bind HFAs as assessed by FACS analysis, and the capability of the serum to potentiate ADCC, demonstrating the functional relevance of this assay. When NMO-IgG were quantified by indirect immunofluorescence, we did not find a significant correlation between immunohistochemical reactivity and HFA binding or NK cell degranulation (data not shown), suggesting that our FACS assay reflect more closely the functional potential of NMO-IgG than the classical immunofluorescence assay. However, we did not observe astrocyte cytotoxicity of NMO-IgGs along with complement, in the absence of cellular effectors. This observation is consistent with previous studies showing astrocyte resistance to complement-induced lysis, due to their membrane expression of complement regulatory proteins (31).

We further used the astrocyte/BBB-EC BBB coculture model to demonstrate that astrocyte-bound NMO-IgGs may be involved in granulocyte recruitment, another pathologic feature that discriminates NMO lesions from typical MS lesions (32). We observed increased granulocyte recruitment only in the presence of a source of fresh complement. This event could be related to the release of chemotactic factors, i.e., the anaphylatoxins C3a and C5a, in the classical complement activation pathway. C5a is a chemoattractant for neutrophils and eosinophils, whereas C3a is a chemotaxin for eosinophils and can stimulate neutrophils cells after eosinophil activation (33). Consistently, a recent study has demonstrated that transgenic expression of C3a in the CNS markedly exacerbates experimental autoimmune encephalomyelitis and results in massive perivascular recruitment of inflammatory cells to the spinal cord (34). The tetrameric structure of AQP4 molecules favors IgG oligomerization, making NMO-IgG-coated astrocyte membranes ideal surfaces for complement classical pathway activation leading to release of C3a and C5a fragments and consequently, increase in microvascular permeability and granulocyte chemo-attraction. Although we observed CD107a mobilization on granulocytes after their migration through the BBB, this degranulation was not prevented by complement inactivation suggesting that this response was mediated by direct contact with the Ab-coated astrocytes. Although ADCC has been described with polymorphonuclear cells (35), we failed to demonstrate significant cytotoxicity of granulocytes on NMO-IgG-coated HFAs even at high E:T ratio (our unpublished data). However, it is likely that the release of hydrolytic granular enzymes, bioactive lipids, and superoxide anions after granulocyte degranulation could contribute to brain injury (36).

In conclusion, we provide experimental demonstration of the pathogenic mechanism by which serum-derived NMO-IgGs induce CNS necrotic lesion formation. We have demonstrated that HFA-bound NMO-IgGs are able to induce 1) decreased AQP4 expression on astrocyte end-feet and increased BBB permeability, 2) NK cell degranulation and astrocyte killing by ADCC, and 3) complement-dependent granulocyte attraction through the BBB. These data bring new physiopathological explanations for BBB breakdown, loss of AQP4 immunoreactivity, and perivascular granulocyte infiltration described in active NMO lesions. A better understanding of the physiopathological processes underlying NMO lesion formation should help the development of therapies for this severe and disabling disease.

We thank Nathalie Arbour and Salvatore Carbonetto for highly stimulating discussion. We thank M. Blain and M. Bernard for excellent technical assistance in the laboratory and all the clinical staff involved in this study.

The authors have no financial conflict of interest.