Background
Neuromyelitis optica (NMO), a severe inflammatory demyelinating disorder, has gained increasing interest since the discovery of serum NMO-IgG autoantibodies targeting the aquaporin-4 (AQP4) water channel protein [
1,
2]. The detection of this highly specific biomarker resulted in the incorporation of the NMO-IgG serostatus in the diagnostic criteria of NMO [
3]. An early differentiation from multiple sclerosis (MS) is highly important, due to differences in prognosis and therapy of NMO patients. The target antigen AQP4 is localized on astrocytic endfeet [
4] and is expressed as full length M1 or shorter M23 AQP4 isoform [
5,
6]. Recently, serum anti-AQP4 antibodies were shown to bind primarily to the shorter M23 AQP4 isoform [
7‐
9], which is of high diagnostic relevance due to an increased sensitivity of NMO-IgG analysis. Antibodies to AQP4 are also frequently detected in so called "High-risk NMO" (HR-NMO) patients not fulfilling all diagnostic criteria for NMO, who present with NMO-associated symptoms like recurrent optic neuritis (ON) or longitudinally extensive transverse myelitis (LETM) extending more than three vertebral segments [
10]. NMO-IgG seropositivity was shown to be predictive for a poor visual outcome and the development of NMO in patients with recurrent ON [
11,
12]. Furthermore, the detection of AQP4-IgG in patients with a first episode of LETM extending ≥ three vertebral segments was associated with further relapses of LETM or ON, in some cases even within half a year [
13]. Therefore, NMO and HR-NMO patients (recurrent ON or monophasic/recurrent LETM) are also classified as NMO-spectrum disorders (NMOSD) [
10]. However, AQP4-IgG are missing in 5-40% of these patients, depending on the immunoassay used [
9,
12,
14‐
16]. It is not yet known whether autoantibodies to other central nervous system (CNS) specific antigens are present in patients with NMO and HR-NMO [
17].
Recent experimental studies indicated that myelin oligodendrocyte glycoprotein (MOG), a glycoprotein localized on the outer surface of the myelin sheath and oligodendrocytes [
18], might be a target antigen in NMO. Two
in vivo studies demonstrated spontaneous development of NMO-like symptoms with severe opticospinal experimental autoimmune encephalomyelitis (EAE) in a double-transgenic opticospinal EAE (OSE) mouse model expressing T cell and B cell receptors specific for MOG [
19,
20]. This mouse strain closely resembles human NMO by exhibiting prototypical inflammatory demyelinating lesions in the optic nerve and spinal cord. Furthermore, the animals were found to exhibit highly positive serum MOG-IgG1 antibodies [
19]. Additionally, there are several reports demonstrating the induction of an NMO-like disease following immunization of certain rat strains with MOG [
21‐
23].
Whereas in humans anti-MOG antibodies in MS have been extensively investigated, their role in NMO has not been adressed so far. High-titer IgG autoantibodies to conformational epitopes of MOG (MOG-IgG) were detected in a subgroup of pediatric patients with acute disseminated encephalomyelitis (ADEM) and MS, but rarely in adult-onset MS [
24‐
29]. A possible role of MOG-IgG antibodies in NMO-related diseases is supported by recent findings of our group, demonstrating an increased frequency of MOG-IgG in pediatric patients with recurrent ON compared to monophasic ON subjects (Rostasy K, Mader S, Schanda K, Huppke P, Gärtner J, Kraus V, Karenfort M, Tibussek D, Blaschek A, Kornek B, Leitz S, Schimmel M, Di Pauli F, Berger T, Reindl M: Anti-MOG antibodies in children with optic neuritis, in press). However, so far only one study using the bacterially expressed extracellular domain of MOG as antigen described the occurrence of a humoral immune response to MOG in four NMO patients [
30].
Therefore, we decided to investigate the frequency and titer levels of IgG antibodies to MOG and AQP4 in a multicenter study of patients with CNS demyelinating diseases using a live cell staining immunofluorescence assay with HEK-293A cells transfected with either AQP4 or MOG [
9,
24]. In addition, we analyzed the IgG subtypes of antibodies directed to MOG and AQP4 and their ability to activate the complement cascade in a subset of patients.
Discussion
In this multicenter study we describe for the first time the presence of serum high-titer MOG-IgG antibodies in patients with NMO and HR-NMO. Our data confirm several studies demonstrating the presence of MOG-IgG in a subgroup of patients with ADEM [
24‐
29], as well as AQP4-IgG in NMO and HR-NMO [
1,
2,
9,
10,
12,
14,
15,
32]. Moreover, we report the occurrence of MOG-IgG antibodies in AQP4-IgG seronegative patients with either NMO (two of two) or HR-NMO (seven of 21), and in monophasic ON/CIS patients (two of 32). These results suggest that MOG is a target antigen in AQP4-IgG negative patients with NMO and HR-NMO, which to our knowledge has not been described before. This is of particular relevance since AQP4-IgG is absent in approximately 5-40% of these patients. This variability of AQP4 antibody detection could depend on the antibody assay, the AQP4 isoform, as well as the study population and/or prior immunosuppressive treatment [
9,
12,
14‐
16].
However, it has been speculated whether different pathomechanisms are involved in AQP4-IgG seronegative NMO and HR-NMO patients compared to subjects with "AQP4 autoimmune channelopathies". This assumption is supported by findings showing no development of NMO-like symptoms in animals immunized with purified antibodies from AQP4-IgG seronegative NMO patients [
33]. In contrast, antibodies from AQP4-IgG positive NMO patients were shown to be pathogenic after intra-cerebral administration combined with human complement [
34], as well as following EAE induction [
33,
35,
36].
An involvement of antibodies directed against MOG in NMO and HR-NMO is encouraged by
in vivo studies demonstrating the spontaneous development of human NMO-like symptoms in a double-transgenic mouse strain with opticospinal EAE [
19,
20]. Expressing T cell and B cell receptors specific for MOG, these mice showed inflammatory demyelinating lesions in the optic nerve and spinal cord, sparing brain and cerebellum [
19]. In addition, the animals harbored a MOG-IgG1 directed humoral immune response [
19]. Several studies demonstrated the induction of an NMO-like disease in distinct rat strains following immunization with MOG [
21‐
23]. However, at present only limited information is available regarding MOG-IgG antibodies in human patients suffering from NMO or HR-NMO symptoms and related disorders. One study revealed the presence of antibodies directed to the bacterially produced extracellular domain of recombinant MOG, as investigated with ELISA and immunoblot, in four NMO patients [
30]. However, it was shown that the detection of antibodies against natively folded MOG is restricted to assays using MOG expressed on the surface of cells. In contrast, commonly applied ELISA or Western blot assays using the bacterially expressed protein failed to identify these antibodies. This might provide an explanation for the controversial results regarding serum MOG-IgG antibodies in MS patients. Furthermore, high-titer MOG-IgG was detected in a subgroup of patients with pediatric ADEM and MS, but only rarely in adult-onset MS [
24‐
27,
37]. However, these studies did not include patients with definite and HR-NMO. Most recently, we could demonstrate high-titer MOG-IgG antibodies in pediatric patients with recurrent ON (Rostasy K, Mader S, Schanda K, Huppke P, Gärtner J, Kraus V, Karenfort M, Tibussek D, Blaschek A, Kornek B, Leitz S, Schimmel M, Di Pauli F, Berger T, Reindl M: Anti-MOG antibodies in children with optic neuritis, in press). Now we describe the presence of MOG-IgG in NMO and HR-NMO. These findings expand the heterogenous spectrum of MOG-IgG mediated human demyelinating diseases from ADEM and pediatric MS to now include AQP4-IgG seronegative recurrent ON, LETM and NMO. Nevertheless, MOG might not be the only autoantigen present in AQP4-IgG seronegative patients with NMO and related disorders. Recent studies have described antibodies to NMDA-type glutamate receptors or CV2/CRMP5 in AQP4-IgG seronegative cases with NMO or ON [
38‐
40]. Therefore, our findings of MOG-IgG support a possible relevance of several specific CNS autoantigens in AQP4-IgG seronegative NMO and HR-NMO cases. Further studies are now required in order to identify potential target antigens.
Several
in vitro studies have demonstrated the pathogenic effect of AQP4-IgG in the presence of active complement [
32,
41,
42], which is confirmed by our findings of AQP4-IgG1 mediated CDC at high-titer serum levels of AQP4-IgG. Furthermore, our results show that antibodies against MOG were primarily of the IgG1 subtype and could activate the complement cascade
in vitro, resulting in the formation of the TCC on living MOG transfected HEK-293A cells. To our knowledge, these observations are novel and might provide a deeper insight into the role of high-titer serum anti-MOG antibodies. Thus, the detection of high-titer MOG-IgG might not only serve as a valuable biomarker in AQP4-IgG negative NMO and HR-NMO patients, but possibly play a role as pathogenic factor in human demyelinating diseases, although this needs to be further investigated.
There are two limitations that need to be addressed regarding our study. The first limitation concerns the usage of an immunofluorescence assay to measure AQP4-IgG and MOG-IgG antibodies and TCC formation. This is often criticised by other researchers using automated assays like flow cytometry or immunoprecipitation for the measurement of specific antibodies. However, experiences from the last decades have strongly emphasized that immunofluorescence assays are the gold standard for the detection of several autoantibodies, such as anti-nuclear antibodies. Furthermore, immunofluorescence assays were shown to yield the highest sensitivity for the detection of AQP4-IgG [
9,
14,
43‐
45]. In addition, as an important quality control in our study, all samples were evaluated by three independent investigators with 100% concordance rate. The second limitation concerns the low number of AQP4-IgG seronegative NMO patients in our study population. Nevertheless, we believe that our findings are of high importance as a substantial proportion of NMO and HR-NMO patients lack a specific biomarker. Hence, our results need to be confirmed in a larger study cohort of AQP4-IgG negative NMO and HR-NMO subjects.
Methods
Patients and serum samples
The following patients were recruited from Austria (n = 295), Germany (n = 19), Slovakia (n = 2) and Serbia (n = 19) (Table
1): (1) a total of 45 NMO patients diagnosed according to the revised diagnostic criteria from Wingerchuk et al., 2006 [
3], (2) 53 patients with a high risk of developing NMO (HR-NMO) including 28 monophasic LETM, 13 recurrent LETM and 12 recurrent ON subjects [
3,
10], (3) 33 patients fulfilling the diagnostic criteria for ADEM [
46], (4) 32 CIS patients comprising 19 myelitis (59%) and 13 ON (41%), (5) 71 patients with MS according to the revised "McDonald Criteria" 2005 [
47] including 44 patients with relapsing remitting MS, 8 patients with primary progressive MS and 19 patients with secondary progressive MS, (6) 101 controls including 24 patients with OND (stroke, Parkinson's disease, epileptic seizure, radiculopathy, insomnia, sleep apnoea syndrome, CNS lymphoma, traumatic brain injury, myasthenia gravis, chronic inflammatory demyelinating polyneuropathy, vestibular neuritis, orthostatic syncope, psychogenic neurological symptoms, CNS vasculitis, hereditary neuropathy, analgesic-induced headache, neuroborreliosis, viral encephalitis, chronic tension-type headache, glioblastoma multiforme), 27 patients with SLE and 50 healthy blood donors obtained from the central institute for blood transfusion (Central Institute for Blood Transfusion and Immunological Department, Innsbruck University Hospital).
External serum samples were shipped on dry ice to Innsbruck and all samples were stored at -20°C until analysis. The present study was approved by the ethical committee of Innsbruck Medical University (study no. UN3041 257/4.8, 21.09.2007) and all Austrian patients or parents/legal guardians gave written informed consent to the study protocol. All Serbian and Slovakian patients gave their informed consent for serum sampling and this study was approved by the Institutional Review Board of the Clinic of Neurology, Clinical Center of Serbia, Belgrade. The Slovakian patients signed the translated informed consent form of the Innsbruck Medical University. All German samples were tested anonymously as requested by the institutional review board of the University of Heidelberg.
AQP4-IgG and MOG-IgG immunofluorescence assays
Analysis of M23 AQP4-IgG was performed using a live cell staining IF assay as recently described [
9,
33,
48].
HEK-293A cells (ATCC, LGC Standards GmbH, Wesel, Germany) were transiently transfected (Fugene 6 transfection reagent, Roche, Mannheim, Germany) using the Vivid Colours™ pcDNA™ 6.2C-EmGFP-GW/TOPO plasmid (Invitrogen, Carlsbad, CA, USA), expressing M23 AQP4 fused C-terminally to an emerald green fluorescence protein (EmGFP). The AQP4-IgG IF assay was performed by blocking the transfected cells with 4 μg/ml goat IgG (Sigma-Aldrich, St. Louis, MO, USA) diluted in PBS/10% FCS (Sigma-Aldrich), subsequently incubating the cells with pre-absorbed serum samples (rabbit liver powder, Sigma-Aldrich) at a 1:20 and 1:40 dilution for one hour at 4°C. Bound antibodies were detected using Cy™3-conjugated goat anti-human IgG antibody (Jackson ImmunoResearch Laboratory, West Grove, PA, USA) for 30 minutes at room temperature. Dead cells were excluded by DAPI staining (Sigma-Aldrich). The AQP4-IgG status was determined using a fluorescence microscope (Leica DMI 4000B, Wetzlar, Germany). Each serum sample was individually evaluated by three independent, clinically blinded investigators (SM, KS and VG), yielding a concordance rate of 100%.
In order to determine AQP4-IgG titer levels, AQP4-IgG seropositive samples were further diluted until loss of specific antibody staining. AQP4-IgG was purified from a NMO patient's plasma exchange material as recently described [
33] and served as positive control for each assay.
Serum MOG-IgG was determined in pre-absorbed samples using HEK-293A cells transiently transfected with human MOG cloned into the mammalian expression vector Vivid Colours™ pcDNA™ 6.2 C-EmGFP/TOPO (Invitrogen), expressing MOG fused C-terminally to EmGFP as previously reported [
24]. Serum MOG-IgG was detected on the surface of MOG expressing cells, using Cy™3-conjugated goat anti-human IgG antibody. Titer levels were determined for MOG-IgG positive samples by serial dilution of serum until loss of signal. Based on our previous results, the cut-off value of high-titer MOG-IgG antibodies was defined as ≥ 1:160, with 100% specificity compared to healthy controls [
24]. Consequently, in this study patients with titers ≥ 1:160 are defined as being "high-titer positive". In contrast, patients with "low serum antibody reactivity" (1:20-1:80) as well as serum antibody negative samples (with no detectable antibodies, titer = 0) were summarized as "negative" cohort, to simplify the data sets. However, titer levels below 1:160 are mentioned whenever relevant and illustrated in Figure
1. Additionally, antibodies purified from the plasma exchange material of a high-titer MOG-IgG positive ADEM patient were added as a quality control to each assay. Dead cells were excluded by DAPI staining and the presence and titer levels of MOG-IgG were analyzed by three clinically blinded and experienced investigators (SM, KS and VG).
In order to exclude unspecific background staining, we additionally performed serum antibody stainings using untransfected HEK-293A cells for both IF assays, transfected cells expressing the fusion protein (EmGFP) as a control for the AQP4-IgG assay as well as CD2-EmGFP transfected cells (another protein of the immunoglobulin superfamily) for the MOG-IgG assay. Non-specific background binding was clearly distinguishable from a specific antibody staining in our immunofluorescence setting.
Furthermore, MOG-IgG and AQP4-IgG seropositive and seronegative control samples are regularly retested for antibody titer levels to ensure the quality of the testing system. Titer levels remain constant in the serum samples which are stored at -20°C, even two years after first analysis.
Determination of IgG1-IgG4 isotypes
Serum antibodies to MOG and AQP4 were analyzed in a subgroup of 15 patients for IgG1-IgG4 isotypes via our live cell staining IF assay using MOG or AQP4 transfected cells. After blocking with goat IgG, the transfected cells were incubated with the pre-absorbed serum samples (1:20 and 1:40 dilution) for one hour. Subsequently, cells were washed and incubated with mouse monoclonal anti-human IgG1-IgG4 isotype antibodies for 30 minutes (Sigma-Aldrich, 1:100 dilution in PBS/10% FCS), followed by detection using Alexa Fluor® 546 goat anti-mouse IgG (Invitrogen) for 30 minutes. Dead cells were excluded by DAPI staining and analysis was performed by three independent investigators (SM, KS and VG).
Antibody mediated terminal complement complex (TCC) in cells expressing AQP4 or MOG
Antibody mediated complement activation was investigated in 23 NMO, 33 HR-NMO, 19 ADEM, 14 CIS, 10 MS and 14 CTRL. The selection of patients for the analysis of complement-mediated cytotoxicity was based on the availability of serum samples and the use of samples which are representative for our entire study population. Briefly, serum samples and human complement (Sigma-Aldrich) were heat-inactivated at 56°C for 45 minutes. Inactivated serum samples were diluted 1:10 in serum-free X-VIVO 15 medium (Lonza, Verviers, Belgium) and pre-absorbed with rabbit liver powder. Cells expressing either MOG or AQP4 were washed three times with X-VIVO 15 medium and subsequently incubated with heat-inactivated, pre-absorbed serum samples and 20% active versus 20% heat-inactivated human complement for 90 minutes at 37°C. After washing the cells three times with 100 μl X-VIVO 15 medium, detection of TCC formation was performed by adding the murine-monoclonal anti-human SC5b-9 (Quidel, San Diego, CA, USA; diluted 1:200 in X-VIVO 15) for one hour at 4°C. Following a 30 minutes incubation with the fluorescence labelled Alexa Fluor® 546 goat anti-mouse IgG antibody (1:300, Invitrogen), cells were washed with PBS/10% FCS and dead cells were visualized by DAPI staining. All samples were assessed for the presence of the surface membrane attack complex by three independent investigators blinded for clinical information as well as the design of the assay concerning usage of active/inactive complement (SM, KS and VG).
To analyze the co-localization of the TCC and serum MOG-IgG or AQP4-IgG, we used HEK-293A cells expressing MOG or AQP4 without EmGFP fusion protein.
To obtain M23 AQP4 without EmGFP fusion protein, the M23 AQP4 isoform was cloned into the pcDNA3.1 Directional TOPO Expression vector (Invitrogen) [
9]. In order to conduct experiments with transfected cells expressing MOG without EmGFP fusion protein, we cloned MOG into the pCMV vector (Invitrogen).
Briefly, after incubating the cells with heat-inactivated samples and active versus inactive complement (90 minutes, 37°C), the cells were washed (X-VIVO 15) and stained with the murine-monoclonal anti-human SC5b-9 (Quidel, diluted 1:200 in X-VIVO 15 medium, one hour, room temperature) as described above. Following three washing steps, the Alexa Fluor
® 488 goat anti-mouse IgG antibody and Cy™3-conjugated goat anti-human IgG antibody were diluted in X-VIVO 15 medium and incubated for 30 minutes. Co-staining of AQP4-IgG and MOG-IgG antibodies (red) and TCC (green) was investigated in a blinded fashion (SM, KS and VG), and dead cells were visualized by DAPI staining. Control experiments using active complement in the absence of serum showed no TCC formation on MOG or AQP4 expressing cells (additional file
1 and additional file
2). Additionally, no cross reaction of the antibodies to other species than stated was observed.
Scanning electron microscopy (SEM)
AQP4-IgG mediated complement activation was confirmed via SEM. Briefly, HEK-293A cells were seeded on poly-L-lysine (Sigma-Aldrich) coated glass slides (Menzel, Braunschweig, Germany) and transiently transfected with the AQP4-EmGFP vector. Thereafter, serum samples supplemented with either active or inactive complement were added to the cells. Following incubation at 37°C for 90 minutes, the cells were washed with PBS and fixed in glutaraldehyde (2.5%, v/v in 0.1 m PBS, pH 7.4). After incubation for 30 minutes at room temperature, the fixative was replaced by fresh fixative, and incubated for another two hours at room temperature. Subsequently, complement activation was investigated via SEM according to standard procedures [
49]. Samples were viewed for complement activation in a blinded fashion using the field-emission SEM DSM982-Gemini (ZEISS, Oberkochen, Germany).
Statistical analysis
Statistical analysis (means, medians, range, standard deviations) and significance of group differences were done using IBM SPSS software (release 18.0, SPSS Inc., USA) or GraphPad Prism 5 (GraphPad, San Diego, USA). Between-group comparisons were performed with Kruskal-Wallis test, Dunn's multiple comparison post-hoc test, Mann-Whitney U test, Fisher's exact test and Chi-square test as appropriate. Correlation of parameters was analyzed with Spearman's non-parametric correlation. Statistical significance was defined as two-sided p-value less than 0.05 and Bonferroni's correction was applied for multiple comparisons when appropriate.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SM, VG, KS and MR conceived and designed the experiments. SM, VG and KS carried out all experiments. SM and MR analysed and interpreted the data. KP performed the scanning electron microscopy. KR, ID, AL, SJ, FDP, BK, RE, FD, FAE, MS, PK, JD, WK, TB and MR participated in serum and data collection. SM, VG, KS and MR wrote the initial manuscript. All authors have read and approved the final version of the manuscript.