Aquaporin-4 Antibodies in Neuromyelitis Optica and Longitudinally Extensive Transverse Myelitis

Background  There is increasing recognition of antibody-mediated immunotherapy-responsive neurologic diseases and a need for appropriate immunoassays.

Objectives  To develop a clinically applicable quantitative assay to detect the presence of aquaporin-4 (AQP4) antibodies in patients with neuromyelitis optica and to characterize the anti-AQP4 antibodies.

Design  We compared a simple new quantitative fluorescence immunoprecipitation assay (FIPA) with both indirect immunofluorescence and an AQP4-transfected cell-based assay, both previously described. We used the cell-based assay to characterize the antibodies for their immunoglobulin class, IgG subclass, and ability to induce complement C3b deposition in vitro.

Setting  United Kingdom and Germany.

Participants  Serum samples from patients with neuromyelitis optica (n = 25) or longitudinally extensive transverse myelitis (n = 11) and from relevant controls (n = 78) were studied.

Main Outcome Measures  Comparison of different assays for AQP4 antibodies and characterization of anti-AQP4 antibodies in patients with neuromyelitis optica.

Results  We found antibodies to AQP4 in 19 of 25 patients with neuromyelitis optica (76%) using FIPA, in 20 of 25 patients with neuromyelitis optica (80%) using the cell-based assay, and in 6 of 11 patients with longitudinally extensive transverse myelitis (55%) with both assays; these assays were more sensitive than indirect immunofluorescence and 100% specific. The antibodies bound to extracellular epitope(s) of AQP4, were predominantly IgG1, and strongly induced C3b deposition.

Conclusions  Aquaporin-4 is a major antigen in neuromyelitis optica, and antibodies can be detected in more than 75% of patients. Further studies on larger samples will show whether this novel FIPA is suitable for clinical use. The IgG1 antibodies bind to AQP4 on the cell surface and can initiate complement deposition. These approaches will be useful for investigation of other antibody-mediated diseases.

Antibodies to different ion channels have been found in several neurologic diseases, and their measurement is clinically important.^1,2 A newly defined autoimmune channelopathy is neuromyelitis optica (NMO).^1,3 Neuromyelitis optica is an inflammatory disease that involves the spinal cord and optic nerves^4,5; it often follows a more aggressive clinical course than multiple sclerosis (MS) but responds better to immunosuppressive therapy.^5-7

A need exists for a diagnostic test to distinguish reliably NMO from MS at onset, when the full clinical features may not be apparent but early intervention may prevent disability. In 2004, Lennon et al⁸ described an NMO IgG antibody using indirect immunofluorescence on adult mouse cerebellum sections, showing a characteristic pattern of binding around microvessels, pia, and Virchow-Robin spaces; it was 58% to 73% sensitive and 91% to 100% specific for NMO.⁸ This report led to the adoption of NMO IgG positivity as 1 of 3 supportive criteria for the diagnosis of the disease.⁹ In 2005, the same group identified aquaporin-4 (AQP4) as a target antigen,³ showing binding of some NMO IgG–positive serum samples to the surface of human embryonic kidney (HEK) 293 cells transfected with human AQP4, subsequently confirmed in a study from Japan,¹⁰ and immunoprecipitation of solubilized enhanced green fluorescent protein (EGFP) AQP4 by Western blotting of the precipitates.³ In addition, 57% of NMO cases were positive by immunoprecipitation of in vitro–translated, sulfur 35–labeled methionine AQP4,¹¹ but it is unlikely that the antigen adopts its full native conformation with this technique.

We used EGFP tagging³ to demonstrate a novel, sensitive, quantitative, and highly specific fluorescence immunoprecipitation assay (FIPA) for AQP4 antibodies. In this article, we compare those results with the indirect immunofluorescence and cell-based methods and describe the characteristics of the antibodies and their pathogenic potential. Some of these data have been previously reported in abstract form.¹²

We studied the earliest serum or plasma samples available from 25 unselected patients from the United Kingdom and Germany with an eventual diagnosis of NMO (defined according to the criteria of Wingerchuk et al⁹ but discounting the criterion of NMO IgG or AQP4 assay status), 11 patients with longitudinally extensive transverse myelitis (LETM) who did not fulfill the clinical criteria for NMO, 38 patients with MS (defined according to the criteria of McDonald et al¹³; 27 with relapsing-remitting MS, 6 with secondary progressive MS, and 5 with primary progressive MS), 26 controls with other autoimmune neurologic disease (AIND) derived from routine referrals, and 14 healthy controls (HCs). Many serum samples were included in our short report using indirect immunofluorescence.¹⁴Table 1 summarizes the patients' demographic and clinical characteristics.

The NMO IgG was detected by indirect immunofluorescence on 10-μm adult mouse cerebellum cryosections as previously described.¹⁴ Sections were classified as positive if they exhibited the typical NMO immunofluorescence pattern.⁸ For IgM and IgG subclass analysis, goat anti–human IgG was replaced with antibodies to IgG1, IgG2, IgG3, IgG4, or IgM (1:100; The Binding Site, Birmingham, England) and then Alexa Fluor 488–conjugated donkey anti–sheep IgG (1:750; Invitrogen Ltd, Paisley, England) was added as a secondary antibody. Slides were mounted with standard fluorescent mounting media (DakoCytomation, Cambridge, England) that contained diamidino-2-phenylindole (DAPI) (1:1000).

Human AQP4 complementary DNA was obtained (IMAGE clone 4717755; Geneservice Ltd, Cambridge, England). Both AQP4 isoforms were cloned into plasmid EGFP-C3 (Clontech, Saint-Germain-en-Laye, France) using the reverse primer 5′-GCATCCCGGGTCATACTGAAGACAATACCTCTCCAG and either 5′-GTCACTCGAGATGAGTGACAGACCCACAGCAAG to yield plasmid EGFP-AQP4-M1 or 5′-GTCACTCGAGATGGTGGCTTTCAAAGG GGTCTG to yield plasmid EGFP-AQP4-M23.

The HEK cells were transfected with both EGFP-tagged AQP4 isoforms using standard polyethyleneimine transfection. Forty hours after transfection with EGFP-AQP4-M1, EGFP-AQP4-M23, or EGFP alone (mock transfection), coverslips were rinsed gently with Dulbecco-modified Eagle medium with 20mM HEPES buffer (4-[2-hydroxyethyl]-1-piperazineethanesulfonic acid) and incubated with patient or control serum samples (1:20) in 250 μL of buffer with 1% bovine serum albumin for 1 hour at room temperature (RT), washed 3 times with buffer, and fixed immediately with 3% formaldehyde in phosphate-buffered saline for 15 minutes at RT. Cells were then rinsed twice and incubated: (1) for the detection of IgG or IgM antibodies, with goat anti–human IgG or goat anti–human IgM Alexa Fluor 568–conjugated secondary antibody (Invitrogen–Molecular Probes, Paisley) at 1:750 in buffer with 1% bovine serum albumin for 45 minutes at RT; and (2) for the identification of IgG subclasses, with mouse anti–human IgG1 or mouse anti–human IgG4 (The Binding Site) at 1:50 in buffer with 1% bovine serum albumin for 1 hour at RT, washed 3 times, and then labeled with goat anti–mouse isotype-specific IgG Alexa Fluor 568–conjugated antibody (Invitrogen–Molecular Probes). After antibody labeling, cells were then washed 4 times in phosphate-buffered saline, nuclei were counterstained with DAPI (1:1000), and coverslips were mounted as described herein.

The HEK 293 cells transfected with AQP4 were incubated with heat-inactivated serum samples (30 minutes at 57°C) from AQP4 antibody–positive patients or controls at 1:20 dilution for 30 minutes at 37°C and then washed briefly. Fresh frozen serum from a healthy donor was used as a source of complement after incubation with HEK 293 cells on ice for 35 minutes to reduce nonspecific complement activation. The preabsorbed serum was applied to the AQP4 antibody–treated cells at 1:20 dilution for 30 minutes at 37°C. The coverslips were fixed in 3% formaldehyde, incubated with a polyclonal rabbit anti–human C3c antibody, which detects C3b (1:500, 30 minutes at 4°C; DakoCytomation, Glostrup, Denmark), or a rabbit isotype control (provided ready for use; Zymed Laboratories, South San Francisco, California); and then incubated with Alexa Fluor 568 goat anti–rabbit IgG antibody for 45 minutes (1:750; Invitrogen–Molecular Probes). All stages were performed with specimens on ice. Coverslips were mounted as described herein using standard fluorescence mounting media (Dako Cytomation) containing DAPI (1:1000). Mouse brain sections were immunostained for C3b in a similar fashion, except that the HEK 293 cell preabsorption was not required, and the Alexa Fluor goat anti–rabbit IgG antibody was applied at 1:325.

Immunofluorescence slides or coverslips were stored, light protected, at 4°C and imaged on a Zeiss fluorescence microscope (Carl Zeiss Ltd, Welwyn Garden City, England) with a MacProbe v4.3 digital image system (Perceptive Scientific Instruments Inc, Chester, England). All photographs were taken under similar conditions. Slides were coded and scored for the intensity of surface immunofluorescence and colocalization with EGFP-AQP4 by 3 of us (P.W., M.I.L., and R.G.). The binding was scored as follows: 0 to 0.5, none or slight labeling without colocalization; 1, weak labeling; 2, moderate labeling; 3, strong labeling; and 4, very strong labeling. The final score was the median of 3 independent readings (variance <1). Sample scores of 1 or more were classified as positive on the basis of control results.

Flasks of EGFP-AQP4–transfected HEK 293 cells were lysed by incubation for 1 hour at 4°C in 3 mL of extraction buffer (10mM TRIS, 100mM sodium chloride, 1mM EDTA, 1% Triton X-100, pH 7.5, with protease inhibitor cocktail [Sigma Aldrich Company Ltd, Dorset, England] added immediately before use) followed by centrifugation at 13 000 rpm in a microfuge for 6 minutes at 4°C. The EGFP-AQP4 extract was read on a fluorescence plate reader (SpectraMax GeminiXS; Molecular Devices, Sunnyvale, California).

Individual serum samples (25 μL) were added to 250 μL of cell extract and incubated at 4°C overnight before precipitation of the IgG-AQP4 complexes using 100 μL of protein A beads (Sigma Aldrich Company Ltd), previously blocked for 1 hour at 4°C in ultralow immunoglobulin fetal calf serum (Invitrogen) and washed twice in extract buffer. The bead-IgG-AQP4 complexes were rotated for 2 hours at RT, washed 3 times for 5 minutes with 1 mL of extract buffer, resuspended in 120 μL of extract buffer, and transferred to a 96-well black polymerase chain reaction plate (Abgene, Epsom, England). The fluorescence at 512 nm (excitation, 472 nm; cutoff, 495 nm) was read on a fluorescence plate reader (Molecular Devices).

We used computer software (GraphPad PRISM 4; GraphPad Software Inc, San Diego, California) to graph and analyze the data. The statistical tests used are indicated in the “Results” section.

Three serum samples previously found positive for NMO IgG at the Mayo Clinic, Rochester, Minnesota, were used to help establish our immunofluorescence assay.¹⁴ Testing all but 1 of our 36 patients with NMO or LETM by this method, we found positive results in 14 of 24 patients with NMO (58%) and in 5 of 10 patients with LETM (50%), with 1 positive serum sample from among 26 AIND controls (4%) but 0 in 38 patients with MS (Figure 1A) or 14 HCs. When tested for IgG subclass, only IgG1 was detected (Figure 1B), and in the presence of 3 NMO IgG–positive serum samples and fresh human serum as a source of complement, C3b deposition was clearly found (Figure 1C), with a distribution highly similar to that of the IgG or IgG1 antibodies.

To confirm that the antibodies bind to extracellular epitopes on AQP4, we expressed AQP4 in HEK 293 cells and used immunofluorescence to detect binding of serum antibodies. Because the EGFP-AQP4 was synthesized in the cytoplasm, green fluorescence was seen both within the cell and at the cell surface. By contrast, patients' IgG, indicated by red fluorescence, was only detected on the cell surface (Figure 2A). The median scores for each serum sample are shown in Figure 2B. No binding was detectable in HC, AIND, or MS serum samples. The results were positive (≥1.0) in 20 of 25 NMO serum samples (80%) and in 6 of 11 LETM serum samples (55%) (Table 2). More positive results were found with the cell-based assay than with immunofluorescence, and the scores were higher in the immunofluorescence-positive serum samples than in those that were negative (Figure 2C).

All positive serum samples tested, including those with both high and low AQP4 antibody reactivity, were strongly IgG1 subclass, as shown for 2 highly positive samples in Figure 3A, and weaker for IgG4 and IgM (Figure 3B). Moreover, 9 of 10 AQP4 antibody–positive serum samples showed strong complement C3b deposition on the cell membrane (Figure 4A and B), which correlated weakly (r² = 0.4, P = .05) with IgG1 AQP4 antibodies (Figure 4C).

To establish a quantitative and objective assay, we used immunoprecipitation. The EGFP-AQP4, extracted in detergent from the cells, demonstrated a sedimentation coefficient of 7.4, corresponding to that predicted for an EGFP-AQP4 tetramer (Figure 5A). Serum samples that were positive for AQP4 antibodies immunoprecipitated more EGFP-AQP4 with increasing concentrations of EGFP-AQP4 (Figure 5B), suggesting relatively low overall affinity. Serum samples that immunoprecipitated EGFP-AQP4 did not immunoprecipitate EGFP alone or other EGFP-tagged antigens (P.W., unpublished data, 2007). For testing all serum samples, we used 7500 to 9000 fluorescence units (FUs) in a volume of 250 μL, corresponding to 150nM tetrameric EGFP-AQP4 (based on the FUs detected per EGFP) (Figure 5A, insert). A total of 25 μL of the NMO and LETM serum sample immunoprecipitated between 10 and 1600 FUs, corresponding to 2.5 to 400nM of EGFP-AQP4 tetramers per liter of serum. We chose 25 FUs (equivalent to 6.25nM) as a conservative cutoff value, based on the HC results (mean + 3 SDs, 21.23 FUs); the values for the patients with AIND (mean + 3 SDs, 21.17 FUs) and the patients with MS (mean + 3 SDs, 23.85 FUs) were not different from the HCs, and all were less than 25 FUs. Positive values were found in 19 of 25 patients with NMO (76%) and in 6 of 11 patients with LETM (55%) (Figure 5C and Table 2). The FIPA gave significantly higher results in serum samples that were positive for indirect immunofluorescence (P < .001, Mann-Whitney test) (Figure 5D) and was overall more sensitive and highly specific. Results of the quantitative cell-based assay and FIPA were strongly correlated (P < .001, data not shown).

No correlation was found (P = .72) between the spinal cord magnetic resonance imaging lesion length and the AQP4 antibody values, but a trend was seen toward lower FIPA and cell-based assay values in patients immunosuppressed before serum analysis (P = .07; data not shown). However, antibody levels in patients during relapse were 100% positive, and were higher in patients during relapse than in those sampled during remission (P = .02, cell-based assay; P = .006, FIPA; Mann-Whitney test) (Figure 6A and B).

The clinical classification of the patients in Table 1 is based on their eventual clinical diagnoses and did not incorporate knowledge of their NMO IgG (or AQP4 antibody) status. In 3 patients initially diagnosed as having MS, acute disseminating encephalomyelitis, or inflammatory encephalitis, NMO was diagnosed only several years later, in 2 at the postmortem examination (M. Esiri, DPhil, I. Pomeroy, DPhil, S. Viegas, MD, unpublished data, 2007). All 3 patients were clearly positive for AQP4 antibodies by cell-based assay (scores of 3) and FIPA (>70 FUs) early in their disease course.

Neuromyelitis optica is an immune-mediated neurologic disease that can be severely disabling.^4,5 Improvement by immunosuppressive treatment and/or plasmapheresis^4,6,7 indicates the need to distinguish it from MS and other diseases associated with inflammatory lesions as early as possible, often before a “full house” of clinical features is evident. Previous reports^3,10,11 have shown that antibodies to the water channel AQP4 are specifically associated with this condition. We have established a new assay (FIPA) for potential clinical use and compared it with an AQP4-transfected cell-based assay¹⁰ and with detection of NMO IgG.⁸ Our results suggest that the antibodies detected by indirect immunofluorescence are directed principally against AQP4 and are best assayed by an antigen-specific test. In this relatively small cohort, the FIPA and cell-based assay showed superior sensitivity (76%-80%) and 100% specificity.

We were not able to confirm the results of the study by Takahashi et al¹⁰ that the AQP4 antibody titer was related to spinal cord lesion length, but our samples were not all taken at the time of magnetic resonance imaging and some were obtained after immunosuppression, which was associated with reduced values. However, antibody levels were positive in all 10 patients sampled during relapse, and higher positivity overall might be expected if serum samples were obtained during a relapse and before commencement of immunosuppression. Nevertheless, it is not yet clear whether the antibodies are detectable in all patients at presentation or how often high-risk antibody-negative patients should be retested.

The cell-based assay, as previously reported,¹⁰ demonstrates that disease-associated antibodies bind to AQP4 in a naturally folded state when it is located on the cell surface and, thus, accessible to pathogenic antibodies. Indirect evidence for a humoral pathogenesis in NMO comes from histopathologic studies that demonstrate immunoglobulin deposits and complement activation in acute NMO lesions,¹⁵ as well as a pattern-specific intralesional loss of AQP4.^16-18 Our finding that AQP4 antibody and NMO IgG belong principally to the IgG1 subclass and can activate complement in vitro provides evidence of the potential pathogenicity of this antibody, confirming a recent report¹⁹ that also demonstrated reduction of surface AQP4 expression by endocytosis or degradation and complement activation. Our detection of IgG1 and complement-activating antibodies on brain tissue by immunofluorescence supports the idea that these mechanisms could operate in vivo. Overall, these results strongly support the use of therapies that target humoral immune mechanisms^6,7,20,21 and also the possibility of anti-complement therapy.

Neuromyelitis optica is the first MS-like disease in which a major antigen has been identified, but it is one of a growing number of central nervous system diseases associated with antibodies to cell membrane proteins, particularly ion channels.^1-3 The approaches described are already being used to establish assays for antibodies to a range of different ion channels and receptors (P.W., D.B., and A.V., unpublished data, 2005-2008) to diagnose and study the widening spectrum of antibody-mediated neurologic diseases.

Correspondence: Angela Vincent, MBBS, MSc, FRCPath, Neurosciences Group, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DS, England (angela.vincent@molecular-medicine.oxford.ac.uk).

Accepted for Publication: February 19, 2008.

Author Contributions:Study concept and design: Waters, Jarius, Leite, S. Jacob, Beeson, and Vincent. Acquisition of data: Waters, Jarius, Littleton, Leite, S. Jacob, Gray, Geraldes, Vale, A. Jacob, Palace, Maxwell, and Vincent. Analysis and interpretation of data: Waters, Jarius, Littleton, Leite, and Vincent. Drafting of the manuscript: Jarius, Littleton, Leite, and Vincent. Critical revision of the manuscript for important intellectual content: Waters, Jarius, Littleton, Leite, S. Jacob, A. Jacob, Palace, and Vincent. Statistical analysis: Jarius, Littleton, and Vincent. Obtained funding: Vincent. Administrative, technical, and material support: Waters, Jarius, Littleton, and A. Jacob. Study supervision: Palace and Vincent.

Financial Disclosure: Dr Beeson has filed a patent (PCT/GB2006/050455) for the use of EGFP-tagged acetylcholine receptor, muscle-specific kinase, and AQP4 for autoantibody detection; and Dr Vincent and her department receive royalties and revenue for performing antibody assays for neurologic diseases.

Funding/Support: This study was partly supported by the National Institute of Health Research Oxford Biomedical Research Centre Programme. Dr Jarius was supported by a fellowship from the European Neurological Society. Dr Leite was supported by a fellowship from the Fundação para a Ciência e a Tecnologia in Portugal.

Additional Contributions: The Medical Research Council (Drs Waters and Beeson and Ms Maxwell), R. Voltz, MD (Dr Jarius), the Muscular Dystrophy Campaign (Drs Jarius, Beeson, and Vincent), and B. P. Morgan, PhD, provided monoclonal antibodies and advice; and M. Esiri, DPhil, and R. Hohlfeld, MD, provided comments.