OUP user menu

Heterogeneity of aquaporin-4 autoimmunity and spinal cord lesions in multiple sclerosis in Japanese

Takeshi Matsuoka , Takuya Matsushita , Yuji Kawano , Manabu Osoegawa , Hirofumi Ochi , Takaaki Ishizu , Motozumi Minohara , Hitoshi Kikuchi , Futoshi Mihara , Yasumasa Ohyagi , Jun-ichi Kira
DOI: http://dx.doi.org/10.1093/brain/awm027 1206-1223 First published online: 17 April 2007

Summary

Opticospinal multiple sclerosis (OSMS) in Asians has similar features to the relapsing–remitting form of neuromyelitis optica (NMO) seen in Westerners. OSMS is suggested to be NMO based on the frequent detection of specific IgG targeting aquaporin-4 (AQP4), designated NMO-IgG. The present study sought to clarify the significance of anti-AQP4 autoimmunity in the whole spectrum of MS. Sera from 113 consecutive Japanese patients with clinically definite MS, based on the Poser criteria, were assayed for anti-AQP4 antibodies by immunofluorescence using GFP-AQP4 fusion protein-transfected HEK-293T cells. Sensitivity and specificity of the anti-AQP4 antibody assay, 83.3 and 100%, respectively, were calculated using serum samples with NMO-IgG status predetermined at the Mayo Clinic. The anti-AQP4 antibody positivity rate was significantly higher in OSMS patients (13/48, 27.1%) than those with CMS (3/54, 5.6%), other neurological diseases (0/52) or healthy controls (0/35). None of the 11 patients tested with a brainstem–spinal form of MS were positive. Among OSMS patients, the antibody positivity rate was highest in OSMS patients with longitudinally extensive spinal cord lesions (LESCLs) extending over three vertebral segments and brain lesions that fulfilled the Barkhof criteria (5/9, 55.6%). Multiple logistic analyses revealed that emergence of the anti-AQP4 antibody was positively associated only with a higher relapse rate, but not with optic–spinal presentation or LESCLs. Compared with anti-AQP4 antibody-negative CMS patients, anti-AQP4 antibody-positive MS patients showed significantly higher frequencies of severe optic neuritis, acute transverse myelitis and LESCLs while most conditions were also common to anti-AQP4 antibody-negative OSMS patients. The LESCLs in anti-AQP4 antibody-positive patients were located at the upper-to-middle thoracic cord, while those in anti-AQP4 antibody-negative OSMS patients appeared throughout the cervical-to-thoracic cord. On axial planes, the former most frequently showed central grey matter involvement, while holocord involvement was predominant in the latter. In contrast, LESCLs in anti-AQP4 antibody-negative CMS patients preferentially involved the mid-cervical cord presenting a peripheral white matter-predominant pattern, as seen in the short lesions. Anti-AQP4 antibody-positive MS patients fulfilling definite NMO criteria showed female preponderance, higher relapse rate, greater frequency of brain lesions and less frequent responses to interferon beta-1b than anti-AQP4 antibody-negative OSMS patients with LESCLs. These findings suggested that LESCLs are distinct in anti-AQP4 antibody positivity and clinical phenotypes. There were cases of anti-AQP4 antibody-positive MS/NMO distinct from CMS, and anti-AQP4 antibody-negative OSMS with LESCLs in Japanese. This indicated that the mechanisms producing LESCLs are also heterogeneous in cases with optic–spinal presentation, namely AQP4 autoimmunity-related and -unrelated.

  • opticospinal multiple sclerosis
  • neuromyelitis optica
  • aquaporin-4
  • NMO-IgG
  • Japanese

Introduction

Multiple sclerosis (MS) is a chronic inflammatory demyelinating disease of the central nervous system (CNS). It has been hypothesized to be caused by an autoimmune mechanism targeting CNS myelin. MS is rare in Asians. However, when it does appear, the selective and severe involvement of the optic nerves and spinal cord is characteristic (Kira, 2003). This form, termed opticospinal MS (OSMS), has similar features to the relapsing–remitting form of Devic's neuromyelitis optica (NMO) in Western populations (Wingerchuk et al, 1999; Cree et al, 2002; Lucchinetti et al, 2002). However, it shows features that are distinct from another form of MS seen in Asians, the conventional form of multiple sclerosis (CMS) that shows disseminated lesions in the central nervous system (CNS) including the cerebrum, cerebellum and brainstem (Kira et al, 1996). This is similar to classical MS in Westerners (Kira et al, 1996; Yamasaki et al, 1999; Kira, 2003; Ishizu et al, 2005). Early studies on NMO and MS in Japanese demonstrated that from both clinical and pathological standpoints there were many intermediate cases between NMO and MS (Okinaka et al, 1958; Shibasaki and Kuroiwa, 1973). These observations have led most Japanese researchers to consider that classical MS and NMO represent the opposite ends of a continuum rather than being two distinct diseases, with OSMS in between the two (Shibasaki et al, 1974).

Recently, a specific IgG against NMO, designated NMO-IgG, was described (Lennon et al, 2004), and its relevant antigen was found to be aquaporin-4 (AQP4) (Lennon et al, 2005). Based on the high specificity of NMO-IgG, NMO is now claimed to be a distinct disease entity with a fundamentally different aetiology from MS (Wingerchuk et al, 2006). Additionally, Pittock and colleagues (Pittock et al., 2006a) reported that asymptomatic brain lesions are common in NMO and that even the presence of symptomatic brain lesions does not exclude a diagnosis of NMO based on the presence of NMO-IgG, although NMO-IgG was originally discovered in NMO patients without brain lesions on magnetic resonance imaging (MRI).

Since Nakashima and colleagues (Nakashima et al, 2006) reported an NMO-IgG positivity rate of ∼60% in a selected series of Japanese patients with OSMS, which is similar to that for NMO in Western patients (Lennon et al, 2004), OSMS has been suggested to be NMO (Weinshenker et al., 2006a). As well, there was a significantly higher frequency of longitudinally extensive spinal cord lesions (LESCLs) in NMO-IgG-positive OSMS patients than in NMO-IgG-negative patients, suggesting that NMO-IgG may play an important role in the development of extensive spinal cord lesions in OSMS.

However, although LESCLs are more frequently reported in OSMS than in CMS, one-fourth of Asian CMS patients also have LESCLs (Chong et al, 2004; Su et al, 2006; Minohara et al, 2006), reflecting the severe spinal cord damage commonly seen in Asians. Thus, the roles of NMO-IgG or anti-AQP4 antibodies in the formation of LESCLs in each MS subtype in Asians is currently uncertain and the relationship between NMO-IgG-negative OSMS and NMO remains to be elucidated.

Therefore, we investigated the presence of NMO-IgG and anti-AQP4 antibodies in a group of Japanese MS patients in the present study. We covered the whole spectrum of OSMS and CMS disorders and studied the correlations of the antibodies with the clinical and laboratory findings in these patients. Our primary aim was to clarify the significance of these autoantibodies in the development of human demyelinating diseases of the CNS.

Material and methods

Patients

At the MS clinic in the Department of Neurology, Kyushu University Hospital during 1987–2006, 142 consecutive patients, 36 males and 106 females, were diagnosed with clinically definite MS according to the criteria of Poser and colleagues (Poser et al, 1983) and subjected to brain and whole spinal cord MRI. Of these, stock sera from 113 patients, 27 males and 86 females, were available for antibody assays and used for the present study while the remaining patients were lost to follow-up or were deceased. All patients underwent a thorough neurological examination and routine laboratory tests. All were followed up and clinically evaluated at regular intervals in the MS clinic. Their medical records and MRI films were analysed retrospectively for the present study. All of the patients were residents of Kyushu Island, the southernmost part of mainland Japan. None of the patients was seropositive for human T-cell leukaemia virus type I (during the study period two MS patients were excluded). All patients had a relapsing–remitting or relapsing–progressive course, and patients with primary progressive MS were not included in the present study (during the study period 24 patients were excluded). Patients with monophasic NMO without subsequent relapse were also excluded to avoid including patients with acute disseminated encephalomyelitis (during the study period one patient was excluded).

The MS patients were clinically classified into the two subtypes of OSMS and CMS as described previously (Kira et al, 1996). Briefly, patients who had a relapsing–remitting course and both optic nerve and spinal cord involvement without any clinical evidence of disease in either the cerebrum or the cerebellum were considered to have OSMS. Patients with minor brainstem signs, such as transient double vision and nystagmus, in addition to opticospinal involvement were included in this subtype. Patients with multiple involvement of the CNS, including the cerebrum and cerebellum, were considered to have CMS. Patients with only brainstem and spinal cord symptomatology were difficult to classify into either subtype, but were temporarily grouped together under the term ‘brainstem–spinal form of MS' (BSMS). The diagnosis of the different forms of MS was made before anti-AQP4 antibody and NMO-IgG assays; thereafter, the diagnosis remained unchanged throughout the study. The disability status of the patients was scored by one of the authors (J.K.), according to the Expanded Disability Status Scale (EDSS) of Kurtzke (Kurtzke, 1983). Severe optic neuritis was defined as grade 5 or more than 5 on Kurtzke's Visual Functional Scale (FS) (Kurtzke, 1983). Acute transverse myelitis (ATM) was defined according to Fukazawa and colleagues (1990). The response to interferon beta-1b was evaluated by the changes in the annual relapse rates in the preceding 2 years and during the therapy. All sera taken from the patients were stored at −80°C prior to the present analyses.

Magnetic resonance imaging

All MRI studies were performed using 1.5 T units, Magnetom Vision and Symphony (Siemens Medical Systems, Erlangen, Germany) as described previously (Su et al, 2006). The typical imaging parameters for the brain were as follows: axial T2-weighted turbo spin-echo imaging using TR/TE = 2800/90 ms, flip angle = 180°; axial turbo-fluid-attenuated inversion recovery (FLAIR) imaging using TI/TR/TE = 2200/9000/110 ms, flip angle = 180° and sagittal and axial precontrast and axial and coronal postcontrast T1-weighted spin-echo imaging using TR/TE range = 400–460/12–17 ms, flip angle range = 80–90°. One excitation, with a matrix of 256 × 256, slice thickness of 5 mm and slice gap of 2.5 mm, was used for all brain studies. Gadopentetate dimeglumine at 0.1 mmol/kg body weight was administered intravenously for contrast-enhanced studies. The typical imaging parameters for the spinal cord were as follows: sagittal T2-weighted turbo spin-echo imaging using TR/TE range = 2500–2800/90–116 ms, flip angle = 180°, number of excitations = 3 or 4; sagittal T1-weighted spin-echo imaging using TR/TE range =400–440/11–12 ms, flip angle range = 90–170°, number of excitations = 2 or 3; axial T2-weighted turbo spin-echo imaging using TR/TE range = 3200–5360/99–116 ms, flip angle = 180°, number of excitations = 3 or 4; axial T1-weighted spin-echo imaging using TR/TE range = 400–440/12 ms, flip angle range = 90–170, number of excitations = 2. For sagittal imaging, a matrix of 256 × 256 or 512 × 512, a slice thickness of 4 mm and a slice gap of 0.4 mm were used; for axial imaging, a matrix of 256 × 256 or 512 × 512, a slice thickness of 5 mm and a slice gap range of 1.5–5 mm were used.

MRI scans were taken at the time of clinical relapse (within 30 days of the onset of acute exacerbation) or in the remission phase. Spinal cord MRI was undertaken in 42 patients (26 OSMS and 16 CMS) at relapse presenting with exacerbation of spinal cord symptomatology and in 97 patients in remission (39 OSMS, 49 CMS and 9 BSMS). In four OSMS patients, four CMS patients and two BSMS patients, only MRI scans at relapse showing symptoms other than the spinal cord were available and these were used for evaluations of the spinal cord lesions for the entire clinical course. Brain MRI scans were examined in 69 patients (30 OSMS, 35 CMS and 4 BSMS) at relapse and in 104 patients (42 OSMS, 52 CMS and 10 BSMS) in remission. Spinal cord MRIs at relapse and in remission were analysed separately. For statistical analyses of brain MRIs, relapse MRI findings were used first, but when only remission MRIs were available, remission MRI findings were used. Brain and spinal cord MRIs were independently evaluated by two of the authors (T.M. and F.M.) who were naive to the diagnoses.

The length of the spinal cord lesions was expressed in terms of the number of vertebral segments; lesions extending for three or more than three vertebral segments in length were considered longitudinally extensive, while lesions of less than three vertebral segments were regarded as short lesions (Wingerchuk et al, 2006). Patients who showed LESCLs at either relapse or remission were classified as LESCL-positive, while those who did not have LESCLs at relapse or remission as LESCL-negative. Thus, the number of LESCL-positive patients was potentially underestimated in the present study, since those who had no LESCL at remission MRI might have had a LESCL at relapse. Cross-sectional evaluation was also done for all MRI scans of the spinal cord; lesions were classified as having a holocord pattern with or without peripheral sparing, a central grey matter-predominant pattern, or a peripheral white matter-predominant pattern, according to Tartaglino and colleagues (1996). Gadolinium-enhancement was evaluated in all available MRI scans taken at any spinal cord relapses, while fine analysis of the lesion distribution was assessed for MRI scans taken at the latest relapse or remission. Brain MRI lesions were evaluated according to either the Barkhof criteria (Barkhof et al, 1997) or the Paty criteria (Paty et al, 1988) for MS.

At the time of spinal cord MRI, immunological treatment was being received by 22 patients [14 on interferon beta-1b and 14 on high-dose (more than 40 mg/day) corticosteroids] in the 69 scans of OSMS patients, 34 patients (25 on interferon beta-1b and 13 on high-dose corticosteroids) in the 69 scans of CMS patients and one patient (on interferon beta-1b) of the 11 scans of BSMS patients. At the time of brain MRI, treatment was being received by 25 of the patients (16 on interferon beta-1b and 12 on high-dose corticosteroids) in the 72 scans of OSMS patients, 44 patients (34 on interferon beta-1b and 14 on high-dose corticosteroids) in the 87 scans of CMS patients and one patient (on interferon beta-1b) in the 14 scans of BSMS patients.

NMO-IgG

NMO-IgG was measured at the Mayo Clinic by V. A. Lennon, as previously described (Lennon et al, 2004). Serum samples from 46 OSMS patients, 51 CMS, 8 with parasitic myelitis (six with visceral larva migrans of Toxocara canis and two with visceral larva migrans of Ascaris suum; of whom four had LESCLs and two had short spinal cord lesions on MRI) and 14 with myelitis and atopic diathesis (atopic myelitis) (Kira et al, 1998), all of whom had short spinal cord lesions on MRI, were measured with the examiners blinded to the origin of the specimens. Parasitic myelitis and atopic myelitis were chosen as the disease controls because both usually show eosinophilic myelitis on pathological examination (Osoegawa et al, 2003), as observed in some cases of NMO (Lucchinetti et al, 2002).

Anti-AQP4 antibody assay

For the anti-AQP4 antibody assays, serum samples from 113 MS, 4 idiopathic recurrent transverse myelitis (IRTM) and 52 other neurological disease (OND) patients (11 parasitic myelitis, 14 atopic myelitis, 1 HTLV-1-associated myelopathy, 2 viral encephalitis, 2 neurosarcoidosis, 1 meningitis, 1 neuro-Behcet disease, 17 spinocerebellar degeneration, 2 Parkinson disease and 1 normal pressure hydrocephalus), and 35 healthy controls were evaluated.

A full-length cDNA encoding human AQP4 (AQP4 transcript variant a; GenBank accession number NM_001650) was amplified from a cDNA library generated from commercially obtained human spinal cord mRNAs (Clontech, Mountain View, CA, USA). The PCR product was cloned into the pDONR221 vector (Invitrogen, Carlsbad, CA, USA) and its sequence confirmed. After sequencing, the AQP4 cDNA was transferred to the pcDNA-DEST53 expression vector (Invitrogen). Human embryonic kidney HEK-293T cells maintained in Dulbecco's modified Eagle's medium containing 10% foetal calf serum were seeded at 5000 cells/well onto 8-well chamber slides (Becton Dickinson, Franklin Lakes, NJ, USA) at 24 h before transfection. The cells were transfected with 100 ng/well of the green fluorescent protein (GFP)-AQP4 fusion protein expression vector using FuGENE6 Transfection Reagent (Roche, Basel, Switzerland) according to the manufacturer's instructions. At 48 h after transfection, the cells were fixed in 10% paraformaldehyde in phosphate-buffered saline (PBS) for 4 min, washed in PBS and blocked by incubation in PBS containing 10% goat serum at room temperature for 1 h. Next, the cells were incubated with human serum samples (diluted 1 : 400 in accordance with the results of preliminary experiments) at room temperature overnight, washed and visualized with an Alexa 594-conjugated goat anti-human IgG antibody (Invitrogen). Fluorescence was observed with a confocal laser-scanning microscope (Fluoview FV300; Olympus Optical Co., Tokyo, Japan). With the scientists blinded to either the NMO-IgG status or the origin of the specimens, the anti-AQP4 antibody assay was carried out at least twice for each sample and those that gave a positive result twice were deemed to be positive.

Statistical analyses

Statistical analyses of ages at onset and at examination were done by an analysis of variance (ANOVA). Relapse rate was analysed by the Poisson regression analysis. EDSS score and disease duration were initially performed using the Kruskal–Wallis H test. When statistical significance was found, the Mann–Whitney U test was used to determine the statistical differences between each subgroup. Differences between two subgroups were tested for significance using Fisher's exact probability test. For multiple comparisons, uncorrelated P values (Puncorr) were corrected by multiplying them by the number of comparisons (Bonferroni–Dunn's correction) to calculate corrected P values (Pcorr). Correlation of NMO-IgG titres with various clinical parameters was analysed by Spearman's rank correlation test. Changes in the relapse rates before and after interferon beta-1b administration were analysed using the Wilcoxon signed-ranks test. Multiple logistic analyses were performed to assess possible factors contributing to the development of anti-AQP4 antibody; such as gender, age at onset, disease duration, relapse rate, OSMS, EDSS score, LESCLs and marked CSF pleocytosis (≥50/μl). In all assays, statistical significance was set at P < 0.05. The neurological (J.K.), neuroimaging (T.M. and F.M.), immunological (T.M. and Y.K.) and statistical analyses (M.O. and others) were done independently.

Results

Demographic features of total MS patients and each MS subtype

Age at onset and disease duration of all MS patients were 31.9 ± 13.2 years (mean ± SD) and 12.4 ± 9.9, respectively. Their average EDSS score was 3.8 ± 3.0. Of the 113 patients, 48 were classified as OSMS, 54 as CMS and 11 as BSMS. When clinical features were compared between OSMS and CMS, OSMS showed a significantly higher frequency of relapse rate than CMS (P < 0.0001) (Table 1). The frequencies of severe optic neuritis and ATM were also significantly higher in OSMS than in CMS (P < 0.0001). In CSF, marked pleocytosis (≥50 cells/μl) was more common in OSMS than in CMS, while IgG oligoclonal bands (OB) were more frequently present in CMS than in OSMS (P= 0.0118). Brain lesions fulfilling either Barkhof or Paty criteria were significantly more common in CMS than OSMS (P < 0.0001 and P = 0.0001, respectively), while LESCLs were more frequently observed in OSMS than CMS during the entire course (P = 0.0014) or at the time of relapse. BSMS patients showed the shortest disease duration, and mildest disease.

View this table:
Table 1

Demographic features of multiple sclerosis subgroups

Opticospinal form of multiple sclerosis (n = 48)Conventional form of multiple sclerosis (n = 54)Brainstem–spinal form of multiple sclerosis (n = 11)
No. of males/females8/40 (1 : 5.0)17/37 (1 : 2.2)2/9 (1 : 4.5)
Age at onset (years)33.6 ± 13.829.0 ± 11.438.8 ± 15.8
Disease duration (years)13.8 ± 9.512.1 ± 10.37.5 ± 8.8
Relapse rate0.9 ± 0.6*0.7 ± 0.5*0.9 ± 0.7
EDSS score4.6 ± 3.23.5 ± 2.82.3 ± 1.9
Frequency of symptoms
    Optic neuritis48/48 (100.0%)*30/54 (55.6%)*0/11 (0.0%)
    Bilateral optic neuritis7/48 (14.6%)4/54 (7.4%)0/11 (0.0%)
    Severe optic neuritis (≥FS 5)35/48 (72.9%)*16/54 (29.6%)*0/11 (0.0%)
    Myelitis48/48 (100.0%)*43/54 (79.6%)*11/11 (100.0%)
    Acute transverse myelitis28/48 (58.3%)*8/54 (14.8%)*2/11 (18.2%)
Secondary progression3/48 (6.3%)5/54 (9.3%)1/11 (9.1%)
CSF
    Marked pleocytosis (≥50/μl)7/43 (16.3%)2/51 (3.9%)0/10 (0.0%)
    Neutrophilia (≥5/μl)6/36 (16.7%)1/32 (3.1%)0/8 (0.0%)
    OB7/38 (18.4%)*21/47 (44.7%)*3/9 (33.3%)
    IgG index (≥0.658)a14/33 (42.4%)27/45 (60.0%)2/9 (22.2%)
Barkhof brain lesions13/48 (27.1%)*41/54 (75.9%)*6/11 (54.5%)
Paty brain lesions29/48 (60.4%)*50/54 (92.6%)*9/11 (81.8%)
LESCLs during entire course31/48 (64.6%)*17/54 (31.5%)*5/11 (45.5%)
LESCLs at the relapse15/26 (57.7%)5/16 (31.3%)0/0
  • * Statistically significant in comparison between OSMS and CMS (P < 0.05). Barkhof brain lesions = brain lesions fulfilling the Barkhof criteria (Barkhof et al, 1997); CSF = cerebrospinal fluid; EDSS = Expanded Disability Status Scale of Kurtzke; FS = Kurtzke's Visual Functional Scale; LESCLs = longitudinally extensive spinal cord lesions; OB = IgG oligoclonal bands; Paty brain lesions = brain lesions fulfilling the Paty criteria (Paty et al, 1988).aUpper normal rage of IgG index was derived from our previous study (Kira et al, 1996).

Validation of the anti-AQP4 antibody assay

The expressed GFP-AQP4 fusion protein in the transfected cells was mainly localized on the cell membrane, with an additional granular expression pattern in the cytoplasm (Fig. 1A–F). Positive sera stained the cell membrane of the GFP–AQP4 fusion protein-transfected cells, but not that of untransfected cells. The positive staining was colocalized with GFP on the cell membrane but not in the cytoplasm. Specificity and sensitivity of the anti-AQP4 antibody assay were calculated with the same serum samples used for the NMO-IgG assay. In total, 15 of 18 NMO-IgG positive samples were also positive for the anti-AQP4 antibody, compared to 0 of 101 NMO-IgG negative samples, indicating that the sensitivity of the anti-AQP4 antibody assay was 83.3%, specificity 100%, false-negative rate 16.7% and false-positive rate 0% (Fig. 1G). The relationships between the NMO-IgG titres and anti-AQP4 antibody positivity rates were: anti-AQP4 antibody was positive in two of two sera with an NMO-IgG titre of 1 : 61 440; two of two with a titre of 1 : 30 720; three of three with a titre of 1 : 15 360; three of three with a titre of 1 : 7 680; one of two with a titre of 1 : 3840; four of four with a titre of 1 : 1920; zero of one with a titre of 1 : 960 and zero of one with a titre of 1 : 480.

Fig. 1

Immunostaining of HEK-293T cells transfected with a GFP-AQP4 fusion protein expression vector. The expressed GFP-fused AQP4 is mainly localized to the cell membrane (green) but is also observed in the cytoplasm (A). IgG in serum from an NMO-IgG-positive OSMS patient, which combines with the cell membrane of GFP-AQP4-transfected cells (B), is detected with an Alexa594-conjugated anti-human IgG antibody (red). Merged images of the transfected cells (C) show colocalization of the serum IgG from the OSMS patient and the GFP-AQP4 protein on the cell membrane (yellow). In contrast, serum from an NMO-IgG-negative CMS patient does not stain the GFP-AQP4-transfected cells (DF). (G) Sensitivity and specificity of the anti-AQP4 antibody assay is determined using sera with NMO-IgG status predetermined by the Mayo Clinic (V. A. Lennon). AQP4 = aquaporin-4; CMS = conventional form of multiple sclerosis; GFP = green fluorescent protein; NMO = neuromyelitis optica; OSMS = opticospinal form of multiple sclerosis.

Anti-AQP4 antibody positivity rate

Among the three patients whose sera were NMO-IgG-positive but anti-AQP4 antibody-negative, one gave a positive result for the anti-AQP4 antibody with serum taken on other occasion. The remaining two NMO-IgG-positive OSMS patients with LESCLs and without Barkhof brain lesions constantly showed negative anti-AQP4 antibody assays. Therefore, anti-AQP4 antibody was positive in 13 of 48 (27.1%) OSMS, 3 of 54 (5.6%) CMS, 0 of 11 BSMS, 1 of 4 (25.0%) IRTM and 0 of 52 OND patients, including 25 with eosinophilic myelitis and 0 of 35 healthy controls (Fig. 2). The positivity rate was significantly higher in OSMS than in CMS (Pcorr = 0.0201), OND (Pcorr = 0.0069) and healthy controls (Pcorr = 0.0006). Among OSMS patients without Barkhof brain lesions, the antibody positivity rates did not differ significantly, irrespective of the presence or absence of LESCLs either during the entire clinical course (27.3 versus 15.4%) or at relapse (42.9 versus 33.3%). The anti-AQP4 antibody positivity rate was highest in patients with LESCLs and Barkhof brain lesions (55.6%). In CMS patients with LESCLs, 2 of 17 (11.8%) were anti-AQP4 antibody-positive and the antibody was found only in those with extremely long spinal cord lesions extending over 10 vertebral segments in length (2/6, 33.3%). It was never found among the remaining patients with LESCLs of 3–10 vertebral segments in length (0/11, 0%). Three LESCL-negative MS patients with anti-AQP4 antibody were all examined for spinal cord MRI at relapse. NMO-IgG gave essentially the same results as anti-AQP4 antibody (data not shown).

Fig. 2

Anti-AQP4 antibody seropositivity rates, according to the presence or absence of either LESCLs or brain lesions fulfilling the Barkhof criteria (Barkhof brain lesions) (Barkhof et al., 1997), among patients with OSMS, CMS, BSMS, IRTM and other neurological diseases including eosinophilic myelitis as well as healthy controls. Uncorrected P values are corrected by multiplying the number of comparisons by the original P value to calculate the corrected P values. AQP4 = aquaporin-4; BSMS = brainstem–spinal form of MS; CMS = conventional form of multiple sclerosis; EDSS = Expanded Disability Status Scale of Kurtzke; HC = healthy controls; IRTM = idiopathic recurrent transverse myelitis; LESCLs = longitudinally extensive spinal cord lesions; NA = not applicable; OND = other neurological diseases; OSMS = opticospinal form of multiple sclerosis. *Significant difference between the linked groups (Pcorr < 0.05).

Changes in anti-AQP4 antibody positivity during the clinical course

Anti-AQP4 antibody positivity was observed in around 20–30% of OSMS patients and 15–20% of the total MS patients over a wide range of disease durations (from <5 to ≥15 years) (Supplementary Fig. 1A). Among 13 MS patients with NMO-IgG whose sera were repeatedly sampled and examined for anti-AQP4 antibody positivity, 10 were consistently positive (ranging from 0.9 to 5.1 years), 1 was consistently negative for over 5.8 years and 2 changed from initial negativity to later positivity over 5.0 and 5.9 years, respectively, although their NMO-IgG titres were relatively low (1 : 480 and 1 : 1920, respectively). NMO-IgG positivity showed a similar trend over a wide range of disease duration (data not shown). The NMO-IgG titre had no correlations with the disease duration, number of exacerbations or relapse rate, but had an inverse correlation with both EDSS score and progression index at the time of blood sampling and at the final follow-up in NMO-IgG-positive MS patients (P = 0.0076 and P = 0.0517, respectively, for EDSS score and P = 0.0185 and P = 0.0556, respectively, for progression index, Spearman's rank correlation test) (Supplementary Fig. 1B). We further examined the effects of immunotherapies on antibody positivity and found that neither anti-AQP4 antibody nor NMO-IgG positivity was significantly affected by treatment. Specifically, the NMO-IgG positivity rates were: 3/21 (14.3%) in patients undergoing any of the immunotherapies versus 15/76 (19.7%) without; 2/16 (12.5%) with interferon beta-1b versus 16/81 (19.8%) without; and 1/6 (16.7%) with corticosteroids versus 17/91 (18.7%) without, while the anti-AQP4 antibody positivity rates were: 2/23 (8.7%) in patients with any of the immunotherapies versus 14/90 (15.6%) without; 1/18 (5.6%) with interferon beta-1b versus 15/95 (15.8%) without; and 0/5 (0.0%) with corticosteroids versus 16/108 (14.8%) without (P > 0.1 in all cases).

Relationships between anti-AQP4 antibody positivity and clinical findings

We compared the clinical and laboratory findings among anti-AQP4 antibody-positive MS, anti-AQP4 antibody-negative OSMS and anti-AQP4 antibody-negative CMS patients, excluding two OSMS patients who were positive for NMO-IgG but negative for anti-AQP4 antibody (Table 2). The anti-AQP4 antibody-positive MS patients were all females. Female preponderance was only statistically significant in anti-AQP4 antibody-positive MS compared with anti-AQP4 antibody-negative CMS patients (Pcorr = 0.0204). The age at onset was also significantly higher in anti-AQP4 antibody-positive MS than in anti-AQP4 antibody-negative CMS patients (Pcorr = 0.0331). Although disease duration did not differ significantly among the three groups, anti-AQP4 antibody-positive MS patients showed a significantly higher relapse rate than anti-AQP4 antibody-negative OSMS and CMS ones (Pcorr < 0.0001, both) while among anti-AQP4 antibody-negative MS patients, OSMS had a significantly higher relapse rate than did CMS ones (Pcorr = 0.0121). Both anti-AQP4 antibody-positive MS and the anti-AQP4 antibody-negative OSMS patients showed greater EDSS scores than anti-AQP4 antibody-negative CMS ones, but the difference did not reach statistical significance. Compared with anti-AQP4 antibody-negative CMS patients, severe optic neuritis and ATM were significantly more frequent in anti-AQP4 antibody-positive MS (Pcorr = 0.0078 and Pcorr = 0.0039, respectively) and anti-AQP4 antibody-negative OSMS (Pcorr = 0.0042 and Pcorr = 0.0006, respectively) patients. However, the frequency of bilateral optic neuritis was more than 2-fold higher in anti-AQP4 antibody-negative OSMS in comparison to anti-AQP4 antibody-positive MS patients. None of the anti-AQP4 antibody-positive MS patients showed secondary progression, compared to around 10% of those in the anti-AQP4 antibody-negative CMS group.

View this table:
Table 2

Comparison of the clinical findings among multiple sclerosis subtypes according to the anti-AQP4 antibody status

Anti-AQP4 Ab(+) multiple sclerosis (n = 16)Anti-AQP4 Ab(−) opticospinal form of multiple sclerosis (n = 33)aAnti-AQP4 Ab(−) conventional form of multiple sclerosis (n = 51)
Clinical subtype: OSMS/CMS13 : 333 : 00 : 51
No. of males/females0/16*8/25 (1 : 3.1)17/34 (1 : 2.0)*
Age at onset (years)38.2 ± 13.6*32.0 ± 13.428.6 ± 11.5*
Disease duration (years)14.0 ± 8.913.9 ± 9.812.1 ± 10.5
Relapse rate1.1 ± 0.6*,**0.8 ± 0.5*,***0.6 ± 0.5**,***
EDSS score5.1 ± 2.94.2 ± 3.23.4 ± 2.8
Frequency of symptoms
    Optic neuritis16/16 (100.0%)*33/33 (100.0%)**27/51 (52.9%)*,**
    Bilateral optic neuritis1/16 (6.3%)5/33 (15.2%)4/51 (7.8%)
    Severe optic neuritis (≥FS 5)12/16 (75.0%)*22/33 (66.7%)**15/51 (29.4%)*,**
    Myelitis16/16 (100.0%)33/33 (100.0%)*40/51 (78.4%)*
    Acute transverse myelitis9/16 (56.3%)*18/33 (54.5%)**7/51 (13.7%)*,**
Secondary progression0/16 (0.0%)2/33 (6.1%)5/51 (9.8%)
CSF
    Marked pleocytosis (≥50/μl)2/16 (12.5%)5/28 (17.9%)2/48 (4.2%)
    Neutrophilia (≥5/μl)2/14 (14.3%)4/22 (18.2%)1/30 (3.3%)
    OB1/11 (9.1%)6/28 (21.4%)21/44 (47.7%)
    IgG index (≥0.658)b4/10 (40.0%)10/24 (41.7%)26/42 (61.9%)
ANA6/16 (37.5%)5/30 (16.7%)8/45 (17.8%)
SSA/SSB5/16 (31.3%)3/24 (12.5%)4/38 (10.5%)
  • *,**,*** Significant difference between the linked values (Pcorr < 0.05). Ab = antibody; AQP4 = aquaporin-4; CMS = conventional form of multiple sclerosis; CSF = cerebrospinal fluid; EDSS = Expanded Disability Status Scale of Kurtzke; FS = Kurtzke's Visual Functional Scale; OB = IgG oligoclonal bands; OSMS = opticospinal form of multiple sclerosis. a Two OSMS patients who were positive for NMO-IgG but negative for anti-AQP4 antibody were excluded. bUpper normal rage of IgG index was derived from our previous study (Kira et al, 1996).

Marked CSF pleocytosis and CSF neutrophilia were more common in anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS than in anti-AQP4 antibody-negative CMS patients, while CSF OB were more frequently observed in anti-AQP4 antibody-negative CMS than in anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS patients, yet the differences were not statistically significant. The positivity rates for ANA and SSA/SSB were 2–3-fold higher in anti-AQP4 antibody-positive MS than in anti-AQP4 antibody-negative OSMS and CMS patients.

Relationships between anti-AQP4 antibody positivity and magnetic resonance imaging findings

Representative MRI findings of anti-AQP4 antibody-positive MS, anti-AQP4 antibody-negative OSMS and anti-AQP4 antibody-negative CMS patients are shown in Fig. 3. Brain lesions fulfilling either the Barkhof or Paty criteria were most common in anti-AQP4 antibody-negative CMS patients and the differences were significant between anti-AQP4 antibody-negative CMS and OSMS (Pcorr < 0.0001 and Pcorr = 0.0003, respectively) patients (Table 3). For all items of the Barkhof criteria, anti-AQP4 antibody-negative CMS showed significantly higher frequencies than the antibody-negative OSMS patients (Pcorr < 0.05 in all), except for juxtacortical lesions. Anti-AQP4 antibody-positive MS patients showed a similar high frequency of brain lesions fulfilling the Paty criteria to that of anti-AQP4 antibody-negative CMS ones. As well, the frequency of Barkhof brain lesions was about 2-fold higher than in anti-AQP4 antibody-negative OSMS patients. The frequency of ovoid lesions was significantly higher in anti-AQP4 antibody-negative CMS than in anti-AQP4 antibody-negative OSMS (Pcorr < 0.0001) and anti-AQP4 antibody-positive MS (Pcorr = 0.0399) patients.

Fig. 3

Representative MRIs of anti-AQP4 antibody-negative (A and B) and -positive (CE) MS patients with LESCLs. MRI of a 50-year-old female patient with OSMS at relapse (A-1–3) shows an LESCL but no Barkhof brain lesions. Disease duration was 0.7 years at the time of the MRI scan and EDSS was 7.0. The patient had neither NMO-IgG nor the anti-AQP4 antibody. The LESCL shows a holocord pattern at the thoracic level. Periventricular rim-like lesions (arrow in A-3) are seen around the anterior horns. MRI of a 37-year-old male patient with CMS at relapse (B-1–4) shows an LESCL at the C3–C7 level and Barkhof brain lesions. Disease duration was 17.2 years and EDSS was 3.5. The patient had neither NMO-IgG nor the anti-AQP4 antibody. The LESCL shows a peripheral white matter-predominant pattern (arrow in B-2). Spinal cord and brain MRI of an OSMS patient (72-year-old female with a disease duration of 16.2 years and an EDSS score of 7.5 at the time of the MRI scans) with LESCLs presenting the central grey matter-predominant pattern involving the upper-to-middle thoracic cord and subclinical brain lesions fulfilling the Barkhof criteria at relapse (C-1-5). The patient initially showed short spinal cord lesions with gadolinium-enhancement when serum was negative for both NMO-IgG and anti-AQP4 antibodies. LESCLs developed later and serum became positive for NMO-IgG (titre, 1 : 480) and anti-AQP4 antibodies. Spinal cord and brain MRI of a CMS patient (42-year-old female with a disease duration of 17.1 years and an EDSS score of 7.5 at the time of the MRI scans) with LESCLs presenting the central grey matter-predominant pattern, marked spinal cord atrophy and brain lesions fulfilling the Barkhof criteria in remission (D-1–5). The arrow (D-4) indicates a cavity-like lesion. The patient's NMO-IgG titre is 1 : 15 360. Spinal cord and brain MRI of a CMS patient (43-year-old female with a disease duration of 4.0 years and an EDSS score of 7.0 at the time of the MRI scans) with LESCLs (E-1–4). The patient had multiple gadolinium-enhanced lesions in the cerebrum and cerebellum at the initial attack, all of which have almost completely disappeared in the follow-up scans. Four years later, the patient developed severe left hemiparesis and cortical sensory impairment. Methylprednisolone pulse therapy almost completely resolved her symptoms; NMO-IgG titre is 1 : 61 440. A huge confluent brain lesion in the right cerebral hemisphere and an LESCL involving mainly the upper-to-middle thoracic cord with the central grey matter-predominant pattern were also found. No gadolinium-enhancement is visible in the spinal cord lesion or the huge brain lesion. The huge brain lesion (E-3) showed increased diffusivity (increased ADC values) on the ADC map of the diffusion-weighted sequence (Supplementary Fig. 3A and B) and increased choline and decreased N-acetylaspartate on magnetic resonance spectroscopy (Supplementary Fig. 3C). All these findings are compatible with acute demyelination. The sagittal and axial spinal cord MRI are T2-weighted images and the brain MRI are fluid-attenuated inversion recovery images except for (E-4), which is a gadolinium-enhanced T1-weighted image. AQP4 = aquaporin-4; CMS = conventional form of multiple sclerosis; EDSS = Expanded Disability Status Scale of Kurtzke; LESCLs = longitudinally extensive spinal cord lesions; MRI = magnetic resonance imaging; NMO = neuromyelitis optica; OSMS = opticospinal form of multiple sclerosis.

View this table:
Table 3

Comparison of the brain and spinal cord magnetic resonance imaging findings among multiple sclerosis subtypes according to the anti-AQP4 antibody status

Anti-AQP4 Ab(+) multiple sclerosis (n = 16)Anti-AQP4 Ab(−) opticospinal form of multiple sclerosis (n = 33)aAnti-AQP4 Ab(−) conventional form of multiple sclerosis (n = 51)
Brain MRI findings
    Axial images
        Fulfillment for Barkhof criteria8/16 (50.0%)7/33 (21.2%)*37/51 (72.5%)*
            ≥ 1 Gd-enhanced lesions or ≥9 T2 brain lesions8/16 (50.0%)9/33 (27.3%)*42/51 (82.4%)*
            ≥ 9 T2 lesions7/16 (43.8%)9/33 (27.3%)*36/51 (70.6%)*
            ≥ 1 Gd-enhanced lesions1/16 (6.3%)2/32 (2.9%)*17/50 (34.0%)*
            ≥ 1 juxtacortical lesion10/16 (62.5%)13/33 (39.4%)33/51 (64.7%)
            ≥ 3 periventricular lesions7/16 (43.8%)10/33 (30.3%)*39/51 (76.5%)*
            ≥ 1 infratentorial lesion9/16 (56.3%)10/33 (30.3%)*36/51 (70.6%)*
        Fulfillment for Paty criteria13/16 (81.3%)18/33 (54.5%)*47/51 (92.2%)*
        Atypical brain lesions5/16 (31.3%)2/33 (6.1%)9/51 (17.6%)
            Extensive white matter lesions (>3 cm)1/16 (6.3%)0/33 (0.0%)1/51 (2.0%)
            Bilateral diencephalic lesions0/16 (0.0%)0/33 (0.0%)1/51 (2.0%)
            Cavity formation3/16 (18.8%)1/33 (3.0%)6/51 (11.8%)
            Extension from cervical cord into brainstem1/16 (6.3%)2/33 (6.1%)1/51 (2.0%)
    Sagittal FLAIR images
            Ovoid lesions10/14 (71.4%)*16/27 (59.3%)**39/40 (97.5%)*,**
            Anterior rim-like lesionsb10/14 (71.4%)18/27 (66.7%)19/40 (47.5%)
Spinal cord MRI findings
    LESCLs during entire course13/16 (81.3%)*18/33 (54.5%)15/51 (29.4%)*
    LESCLs at the relapse9/13 (69.2%)6/14 (42.9%)4/14 (28.6%)
  • *,** Statistically significant in comparison between each subgroup indicated by Bonferroni–Dunn's correction (Pcorr < 0.05). Ab = antibody; AQP4 = aquaporin-4; Barkhof brain lesions = brain lesions fulfilling the Barkhof criteria (Barkhof et al, 1997); FLAIR = fluid-attenuated inversion recovery; Gd = gadolinium; LESCLs = longitudinally extensive spinal cord lesions; Paty brain lesions = brain lesions fulfilling the Paty criteria (Paty et al, 1988). a Two OSMS patients who were positive for NMO-IgG but negative for anti-AQP4 antibody were excluded.

  • bAnterior rim-like lesions are defined as anterior periventricular lesions lining along with the third and lateral ventricles (Fig. 3A–3).

Atypical brain lesions, previously reported in NMO-IgG-positive NMO patients (Pittock et al., 2006a, b), such as extensive cerebral white matter lesions (≥3 cm), bilateral diencephalic (thalamic/hypothalamic) lesions, cavity formation and extension from the cervical cord into brainstem were not significantly different among the three groups. However, anterior periventricular rim-like lesions lining along with the third and lateral ventricles, which did not deeply extend into the white matter and were not enhanced by contrast media (Fig. 3A-3), were more common in anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS patients than anti-AQP4 antibody-negative CMS ones, though the difference did not reach statistical significance. However, LESCL frequency was higher in anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS patients than in anti-AQP4 antibody-negative CMS ones during the entire course (Pcorr = 0.0012 and Pcorr = 0.0741, respectively), and the similar trend was also observed at relapse (P > 0.1, because of the small sample sizes).

Relationships between anti-AQP4 antibody positivity and distribution of spinal cord lesions on MRI

We compared the distribution and location of each spinal cord lesion between anti-AQP4 antibody-positive MS, anti-AQP4 antibody-negative OSMS and anti-AQP4 antibody-negative CMS patients in detail on MRI. On sagittal planes, in anti-AQP4 antibody-positive MS patients, 31.3% (10/22) and 39.3% (11/28) of the total lesions observed in acute and remission phases, respectively, were LESCLs, while others were short spinal cord lesions (Supplementary Fig. 2). In anti-AQP4 antibody-negative patients, 40.0% (6/15) of the lesions in acute phase and 17.9% (7/39) in remission phase were LESCLs in OSMS while in CMS it was 17.6% (6/34) and 13.3% (10/75) in the acute and remission phases, respectively.

LESCLs in anti-AQP4 antibody-positive MS patients preferentially involved the upper-to-middle thoracic cord, while those in anti-AQP4 antibody-negative OSMS patients were extremely long, extending from the upper cervical cord through to the mid-thoracic cord even during relapse and in remission (Fig. 4A and B, upper panels). Distribution of the lesions significantly differed between the two (Pcorr = 0.0333 at relapse and Pcorr = 0.0084 in remission). In anti-AQP4 antibody-negative CMS patients, both LESCLs and short spinal cord lesions most frequently affected the cervical cord either at relapse or remission (Fig. 4A and B, lower panels). On the axial plane, distribution of LESCLs in remission tended to differ between anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS patients (Puncorr = 0.0218, Pcorr > 0.1) and between anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS patients (Puncorr = 0.0274, Pcorr > 0.1) (Fig. 4C and D). In anti-AQP4 antibody-positive MS patients, LESCLs consistently showed the central grey matter-predominant pattern through relapse and remission phases; even short lesions most frequently involved the central grey matter at relapse (Figs. 3 and 4C and D). However, in anti-AQP4 antibody-negative OSMS patients, LESCLs most frequently revealed the holocord pattern at either relapse or remission, while most of the short lesions demonstrated the peripheral white matter-predominant pattern during relapse and in remission. In anti-AQP4 antibody-negative OSMS patients, distribution of the lesions significantly differed between LESCLs and short lesions in remission (Pcorr < 0.0001) although the difference was not significant due to the small sample size at relapse (Puncorr = 0.0117, Pcorr > 0.1). In anti-AQP4 antibody-negative CMS patients, LESCLs most frequently showed the central grey matter-predominant pattern at relapse but the peripheral white matter-predominant pattern in remission, while short lesions constantly had the peripheral white matter-predominant pattern in relapse and remission phases. When patients undergoing interferon beta-1b treatment at the time of the MRI studies were excluded, essentially the same profiles were seen, though differences in the relapse phase lost statistical significance due to the small sample size.

Fig. 4

A comparison of the distributions of LESCLs between anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS patients at relapse (A, upper panel) and in remission (B, upper panel), and between LESCLs and short lesions in anti-AQP4 antibody-negative CMS at relapse (A, lower panel) and in remission (B, lower panel). Comparisons of the frequencies of the patterns in the lesions on axial planes are also shown according to the presence of LESCLs and short spinal cord lesions in each group at relapse (C) and in remission (D). The distributions of the lesions are classified as central grey matter-predominant, peripheral white matter-predominant and holocord patterns, according to Tartaglino and colleagues (Tartaglino et al, 1996). Uncorrected P values are corrected by multiplying the number of comparisons with the original P value to calculate corrected P values (Mann–Whitney U test with Bonferroni–Dunn's correction). Ab = antibody; AQP4 = aquaporin-4; CMS = conventional form of multiple sclerosis; LESCLs = longitudinally extensive spinal cord lesions; Ls = lesions; MS = multiple sclerosis; n = numbers of lesions; OSMS = opticospinal form of multiple sclerosis. *Significant difference between the linked groups (Pcorr < 0.05).

Multiple logistic analyses for possible factors contributing to anti-AQP4 antibody production

To further identify possible factors contributing to the production of anti-AQP4 antibodies, we divided MS patients into anti-AQP4 antibody-positive and -negative groups and performed multiple logistic analyses. Among the clinical and laboratory parameters examined, only the relapse rate was significantly related to the occurrence of anti-AQP4 antibodies (OR = 6.612, P = 0.0229) (Table 4).

View this table:
Table 4

Multiple logistic analyses for possible factors contributing to anti-AQP4 antibody production

Possible factorsAnti-AQP4 Ab(+) multiple sclerosis (n = 16)Anti-AQP4 Ab(−) multiple sclerosis (n = 97)Odds ratio95% CIP-value
Sex (male/female)0/1627/70 (1 : 2.6)(0.000–)0.9943
Age at onset (years)38.2 ± 13.630.9 ± 12.91.060(0.996–1.127)0.0656
Disease duration (years)14.0 ± 8.912.1 ± 10.11.106(0.999–1.225)0.0528
Relapse rate1.1 ± 0.60.7 ± 0.66.612(1.299–33.664)0.0229
OSMS13/16 (81.3%)35/97 (36.1%)3.505(0.754–16.296)0.1096
EDSS score5.1 ± 2.93.6 ± 2.90.946(0.702–1.274)0.7129
LESCLs13/16 (81.3%)40/97 (41.2%)2.666(0.408–17.406)0.3057
Marked CSF pleocytosis (≥ 50/μl)2/16 (12.5%)7/88 (8.0%)0.897(0.112–7.153)0.9182
  • Ab = antibody; AQP4 = aquaporin-4; CI = confidence interval; CSF = cerebrospinal fluid; EDSS = Expanded Disability Status Scale of Kurtzke; LESCLs = longitudinally extensive spinal cord lesions; OSMS = opticospinal form of multiple sclerosis.

View this table:
Table 5

Comparison of clinical findings between anti-AQP4 antibody-positive multiple sclerosis patients fulfilling the new criteria for definite neuromyelitis optica and anti-AQP4 antibody-negative opticospinal form of multiple sclerosis patients with longitudinally extensive spinal cord lesions

Graphic
  • *Significant difference between the linked values (P < 0.05). Ab = antibody; AQP4 = aquaporin-4; CMS = conventional form of multiple sclerosis; CSF = cerebrospinal fluid; EDSS = Expanded Disability Status Scale of Kurtzke; FS = Kurtzke's Visual Functional Scale; LESCLs = longitudinally extensive spinal cord lesions; NMO = neuromyelitis optica; OB = IgG oligoclonal bands; OSMS = opticospinal form of multiple sclerosis. aTwo OSMS patients who were positive for NMO-IgG but negative for anti-AQP4 antibody were excluded. bUpper normal rage of IgG index was derived from our previous study (Kira et al, 1996). cNumber of gadolinium-enhanced lesions/number of total T2 lesions.

Comparison of clinical and laboratory features between anti-AQP4 antibody-positive MS/NMO patients and anti-AQP4 antibody-negative OSMS patients with LESCLs

Among 16 MS patients who were seropositive for both NMO-IgG and anti-AQP4 antibodies, 14 (87.5%) fulfilled the new NMO criteria (Wingerchuk et al, 2006) and could be regarded as definite NMO (MS/NMO). Therefore, we finally compared the clinical and laboratory findings between these 14 NMO-IgG- and anti-AQP4 antibody-positive MS/NMO patients and 18 anti-AQP4 antibody-negative OSMS patients with LESCLs from whom NMO-IgG-positive but anti-AQP4 antibody-negative patients were excluded. Both groups demonstrated similar clinical features, such as higher EDSS scores, higher frequencies of severe optic neuritis and ATM and lower frequencies of CSF OB, compared with the antibody-negative CMS patients (Tables 2 and 5). However, female preponderance was more marked in anti-AQP4 antibody-positive MS/NMO patients, although the difference between the two groups did not reach statistical significance. Relapse rate was significantly higher in anti-AQP4 antibody-positive MS/NMO patients than in anti-AQP4 antibody-negative OSMS ones with LESCLs (P = 0.0017). The mean age at onset was 4 years older in anti-AQP4 antibody-positive MS/NMO than in anti-AQP4 antibody-negative OSMS patients with LESCLs.

Concerning laboratory findings, ANA and SSA/SSB were more frequent in anti-AQP4 antibody-positive MS/NMO than in anti-AQP4 antibody-negative OSMS patients with LESCLs, while marked CSF pleocytosis and CSF neutrophilia were more common in the latter than in the former, although the difference was not significant. Brain lesions fulfilling either Barkhof or Paty criteria were 2-fold more commonly seen in anti-AQP4 antibody-positive MS/NMO than in anti-AQP4 antibody-negative OSMS patients (P = 0.0751 for the frequency of Paty brain lesions). Spinal cord lesions were even longer in anti-AQP4 antibody-negative OSMS with LESCLs than in anti-AQP4 antibody-positive MS/NMO patients. When the frequency of gadolinium-enhancement of the spinal cord lesions was evaluated at spinal cord relapses during the entire clinical course, it was significantly lower for ≥4 relapses than for ≤3 relapses in anti-AQP4 antibody-negative OSMS patients with LESCLs (P = 0.0172). No such trend was found in anti-AQP4 antibody-positive MS/NMO patients. The frequency of spinal cord swelling at relapse and spinal cord atrophy during remission was similar in both groups.

Although interferon beta-1b was administered for some periods during the clinical course in about half of the patients with anti-AQP4 antibody-positive MS/NMO or anti-AQP4 antibody-negative OSMS with LESCLs, the drug was frequently discontinued in all groups. The reduction in the calculated annual relapse rates was larger in anti-AQP4 antibody-negative OSMS with LESCLs than in anti-AQP4 antibody-positive MS/NMO patients, while the on-treatment relapse rate was significantly lower than the pre-treatment relapse rate only in anti-AQP4 antibody-negative OSMS patients with LESCLs (P = 0.0103) (Table 6). Moreover, the proportion of patients showing more than a 50% reduction in the relapse rate on treatment was also higher in anti-AQP4 antibody-negative OSMS with LESCLs than in anti-AQP4 antibody-positive MS/NMO patients (72.7 versus 14.3%, P = 0.0498), yet the frequency of other immunotherapies did not differ significantly between the two groups before or during interferon beta-1b therapy.

View this table:
Table 6

Comparison of responses to interferon beta-1b between anti-AQP4 antibody-positive multiple sclerosis patients fulfilling the new criteria for definite neuromyelitis optica and anti-AQP4 antibody-negative opticospinal form of multiple sclerosis patients with longitudinally extensive spinal cord lesions

Graphic
  • * Significant difference between the linked values (P < 0.05). Ab = antibody; AQP4 = aquaporin-4; IFN = interferon; LESCLs = longitudinally extensive spinal cord lesions; NMO = neuromyelitis optica; OSMS = opticospinal form of multiple sclerosis. a Two OSMS patients who were positive for NMO-IgG but negative for anti-AQP4 antibody were excluded. bAmong 11 anti-AQP4 antibody-negative OSMS patients with LESCLs who received interferon beta-1b, seven discontinued the drug (two due to exaggeration of lower limb spasticity, two because of drug eruption, one for injection site pain, one for depression and one lost to follow-up). Among seven anti-AQP4 antibody-positive MS/NMO patients, five discontinued the drug (three due to claims of exacerbation after introduction of the drug, one for injection site pain and one for the intention of getting pregnant). cAnalyses were done by intention to treat. dThe relapse rates were compared between pretreatment and on-treatment in each group. Analyses were done by intention to treat.

Discussion

The present study, by assaying anti-AQP4 and NMO-IgG antibodies in an unbiased series of Japanese MS patients, revealed: (i) anti-AQP4 antibody and NMO-IgG are positive in about 15% of total MS patients and in 30% of OSMS patients over a wide range of disease durations. Anti-AQP4 antibody emergence had a significant positive association with a higher relapse rate but with neither LESCLs nor OSMS presentation by multiple logistic analyses; (ii) although both anti-AQP4 antibody-positive MS and antibody-negative OSMS patients with LESCLs had many common clinical features, LESCLs in anti-AQP4 antibody-positive MS preferentially involved the upper-to-middle thoracic cord with the central grey matter-predominant pattern through relapse and remission phases. Whereas, anti-AQP4 antibody-negative OSMS with LESCLs had extremely long spinal cord lesions from the upper cervical to mid-thoracic cord with the holocord pattern; (iii) in anti-AQP4 antibody-negative CMS patients, LESCLs still occurred in about 30% and showed a predilection to the cervical cord with the peripheral white matter-predominant pattern, which was more evident in remission; (iv) anti-AQP4 antibody-positive MS patients who fulfilled the new NMO criteria had higher ages at onset, female preponderance, higher relapse rate, higher frequencies of severe optic neuritis and ATM and higher EDSS scores compared with antibody-negative CMS patients. While most were also common to antibody-negative OSMS patients with LESCLs, anti-AQP4 antibody-positive MS/NMO patients more commonly had brain lesions fulfilling either the Barkhof or Paty criteria and less frequent responses to interferon beta-1b.

As compared with anti-AQP4 antibody-negative CMS, LESCLs were quite distinct for not only frequency, but also distribution in anti-AQP4 antibody-positive MS and anti-AQP4 antibody-negative OSMS. Importantly, although fundamental clinical features, such as severe optic neuritis and ATM, were seen at similar frequencies in anti-AQP4 antibody-positive MS and -negative OSMS, LESCLs were significantly different between the two. There was upper to middle thoracic cord involvement in the former versus the cervical to thoracic cord involvement in the latter in a sagittal plane, while in the axial plane there was central grey matter involvement in the former versus holocord involvement in the latter. As well, there were several distinctive features of anti-AQP4 antibody-positive MS/NMO relative to anti-AQP4 antibody-negative OSMS with LESCLs; such as exclusive occurrence in females, higher age at onset, higher relapse rate and an absence of secondary progression in the former. These patients also showed poor responses to interferon beta-1b, although this was not decisive due to the limitations of the small size of the retrospective study. However, CSF inflammatory cell response was more marked and brain lesions were the least in the latter, suggesting that the latter has more restricted involvement of the CNS accompanied with graver inflammatory cell responses. These observations, together with the constant anti-AQP4 antibody negativity over the long clinical course, may indicate that a mechanism apart from anti-AQP4 antibody production is operative in anti-AQP4 antibody-negative OSMS patients with LESCLs. Various unusual features of MS seen in anti-AQP4 antibody-positive MS/NMO may indicate that the disease is fundamentally distinct from MS. However, since anti-AQP4 antibody-positive MS/NMO frequently shows brain lesions, while antibody-negative OSMS with LESCLs is compatible with NMO in Western populations, patients carrying anti-AQP4 antibody do not appear to perfectly overlap with OSMS or NMO in Asians, and so may be separately classified as an autoimmune aquaporinopathy of the CNS. Further studies are required to determine whether or not the two conditions are indeed the same, irrespective of the presence or absence of the anti-AQP4 antibody.

Earlier pathological studies (Shiraki et al, 1958; Tabira and Tateishi, 1982) and ours (Ishizu et al, 2005) disclosed variability in the degree of inflammatory cell infiltration in Asian OSMS patients; some showed marked perivascular inflammatory cell cuffing or heavy infiltration of T cells and neutrophils in the parenchyma of the spinal cord lesions, while others showed scant inflammatory cell infiltrates. In pathological studies of Western NMO cases, eosinophil infiltration has frequently been observed in lesions (Lucchinetti et al, 2002), whereas we did not detect any eosinophils in our previous studies (Ishizu et al, 2005). Such differences in pathology may suggest the heterogeneity of an effector arm in OSMS and NMO. It is plausible that anti-AQP4 antibody-negative OSMS patients with LESCLs may have predominantly T-cell-mediated immune mechanisms, while anti-AQP4 antibody-positive MS/NMO may be prone to humoral immune mechanisms, as indicated by the higher frequencies of ANA and SSA/SSB, though these are not mutually exclusive and may operate in different stages.

AQP4 is located in astrocyte foot processes surrounding capillaries (Aoki-Yoshino et al, 2005) and its defects prolong the resolution of vasogenic oedema, as shown in AQP4-knockout mice (Papadopoulos et al, 2004). Blood–brain barrier damage strongly induces AQP4 expression in perivascular and parenchymal astrocytes, which in turn contributes to efficient clearance of vasogenic oedema (Tomás-Camardiel et al, 2005). Upregulation of AQP4 in MS lesions has been reported in Japanese MS patients by some authors (Aoki-Yoshino et al, 2005), while others (Misu et al, 2006) reported the disappearance of AQP4 around capillaries in one Japanese autopsied case of OSMS. If the latter is actually a widespread occurrence, anti-AQP4 antibodies may prolong the resolution of tissue oedema associated with blood–brain barrier destruction through the perturbation of water channels in anti-AQP4 antibody-positive MS/NMO. Since NMO-IgG immunostaining produces diffuse labelling of CNS tissues (Lennon et al, 2004) and AQP-4 is present ubiquitously not only in the spinal cord, but also in the brain (Aoki-Yoshino et al, 2005), the occurrence of brain lesions in NMO-IgG-positive patients (Pittock et al., 2006a; Nakashima et al, 2006) may be a natural outcome, which explains the higher frequency of brain lesions seen in this condition compared with anti-AQP4 antibody-negative OSMS. Moreover, the preferential involvement of the central grey matter of the spinal cord, which shows abundant expression of AQP-4 (Misu et al, 2006), may also suggest the involvement of AQP-4 autoimmunity in lesion development in this condition.

In the present series, anti-AQP4 antibodies and NMO-IgG were detected at nearly constant rates over a wide range of disease durations from the early course to the late stage, when secondary destruction of optic nerves and spinal cord occurs repeatedly. These observations are in good accord with the findings of Weinshenker and colleagues (2006b) that NMO-IgG appears during the initial attack in one-third of longitudinally extensive transverse myelitis patients. Therefore, secondary production of the antibody following destruction of optic nerves and spinal cord tissue may not occur in the late stage of illness in most cases. However, this does not necessarily exclude the possibility that the antibody is produced secondarily as a response to severely disrupted optic nerves and spinal cord tissue at the very beginning of the initial insult in susceptible individuals. Since two patients in our series later developed NMO-IgG and anti-AQP4 antibody positivity, albeit with relatively low titres, a longitudinal follow-up study of the antibodies in high-risk patients would be needed to clarify this point.

If AQP-4 autoimmunity is the sole cause of NMO, it is hard to fully explain the preferential involvement of the optic nerves and spinal cord in most NMO patients. Moreover, from the results of our multiple logistic analyses, the presence of anti-AQP4 antibodies was not linked to long spinal cord lesions, disability or even OSMS presentation. Rather, it was most significantly associated with higher relapse rate. Therefore, it is also possible that the anti-AQP4 antibody is a modifying factor contributing to frequent relapses and prolongation of vasogenic oedema, but does not have a role in the severity of tissue destruction or predispose a patient to optic nerve and spinal cord involvement. Paradoxically, NMO-IgG titres tended to have an inverse correlation with the progression index. It is possible that the antibody titres may decrease in chronic burnt-out cases. However, in AQP4 deficient mice, cytotoxic oedema following either ischaemic stroke or water intoxication was decreased and neurological outcomes significantly improved (Manley et al, 2000). Thus, although anti-AQP4 antibody appears a good marker of this condition, the antibody may in some cases be neuroprotective through a reduction in cytotoxic oedemas by water channel blockade. Further studies on the in vivo actions of the antibody are necessary to clarify its exact roles in MS/NMO.

On the other hand, LESCLs in antibody-negative CMS patients showed the same predilection for peripheral white matter and the cervical cord as with short lesions. These characteristics are the same as those reported in Western MS (Tartaglino et al, 1995). In anti-AQP4 antibody-negative CMS, when clinical characteristics were compared between LESCL-positive and -negative patients, LESCL-positive ones had significantly higher relapse rate, greater EDSS scores, and higher frequencies of ATM and secondary progression than LESCL-negative ones; also, Barkhof brain lesions were more common in LESCL-negative patients (Supplementary Table 1). Therefore, anti-AQP4 antibody-negative CMS patients with LESCLs seem to have classical MS that is similar to Western MS with high disease activity. This suggests a possibility that most of the LESCLs in antibody-negative CMS patients are of a similar nature to short demyelinated lesions and may indeed be conglomerations of short ones.

In the present study, as classifications of clinical phenotypes were based on clinical symptomatology, subclinical or silent brain lesions on MRI could possibly be found in OSMS patients. However, it has repeatedly been reported in Asians, even in CMS patients, that long spinal cord lesions are not infrequent (Chong et al, 2004; Su et al, 2006; Minohara et al, 2006). This is consistent with the present result that up to 30% of anti-AQP4 antibody-negative CMS patients demonstrated LESCLs. Existence of cases intermediate between OSMS and CMS is in part attributable to the limitation of clinical classifications; however, such intermediate cases have frequently been reported in Asians clinically and pathologically (Okinaka et al, 1958; Shibasaki and Kuroiwa, 1973; Chong et al, 2004; Su et al, 2006). It remains to be determined whether or not the higher frequency of LESCLs in anti-AQP4 antibody-negative CMS patients in Japanese as well as in other Asians (Chong et al, 2004; Su et al, 2006; Minohara et al, 2006) in comparison to those in Western MS series (Tartaglino et al, 1995) is caused by genetic differences.

Finally, to exclude the possibility that anti-AQP4 antibody-positive MS/NMO and anti-AQP4 antibody-negative OSMS and CMS are a continuum, we considered the following. First, the anti-AQP4 antibody-positive MS/NMO and the antibody-negative OSMS patients with LESCLs also had the periventricular ovoid lesions that are typically seen in CMS. Also, pathologically sharply demarcated demyelinated lesions in extra-opticospinal regions have repeatedly been reported in NMO in Western (Cone et al, 1934; Balser, 1936; Stansbury, 1949) and OSMS in Asian (Shiraki et al, 1958; Okinaka et al, 1958; Shibasaki and Kuroiwa, 1973; Shibasaki et al, 1974) populations. Even a huge atypical brain lesion in one of our anti-AQP4 antibody-positive MS/NMO patients (shown in Fig. 3F-3) demonstrated increased diffusivity [increased apparent diffusion coefficient (ADC) values] on the ADC map of the diffusion-weighted sequence, along with increased choline and decreased N-acetylaspartate on magnetic resonance spectroscopy (Supplementary Fig. 3). These findings are compatible with acute demyelination. Second, we could not detect any statistically significant difference in the frequency of atypical brain lesions previously reported in NMO-IgG-positive NMO patients (Pittock et al., 2006a, b) among the three above-mentioned subgroups; yet periventricular rim-like lesions were more common in OSMS irrespective of presence or absence of the anti-AQP4 antibody. Third, some antibody-negative CMS patients with Barkhof brain lesions also had extremely long spinal cord lesions with central grey matter involvement (an intermediate case in Supplementary Fig. 4). Fourth, recent clinico-epidemiological studies showed changes in the clinical phenotypes from CMS to OSMS in Japanese (Kira et al, 1999; Nakashima et al, 1999) and NMO to classical MS in French West Indies (Cabre et al, 2005). Finally, interferon beta-1b was shown to be beneficial for OSMS in Japanese by a double blind randomized controlled trial (Saida et al, 2005). Unless we consider that these conditions constitute a spectrum, it is difficult to explain the frequent occurrence of such additional demyelinated lesions in the brain of anti-AQP4 antibody-positive MS/NMO patients, the existence of intermediate cases, and the phenotypic changes of CNS demyelinating diseases in the same races over time. However, demyelinated lesions indistinguishable from MS may tend to develop secondarily under CNS inflammatory conditions such as found with NMO (autoimmune aquaporinopathy). The beneficial effects of interferon beta on OSMS as well as the phenotypic changes from OSMS to CMS may also be explained by splitting OSMS into anti-AQP4 antibody-positive NMO (autoimmune aquaporinopathy) distinct from MS, and anti-AQP4 antibody-negative OSMS that forms part of MS.

In summary, anti-AQP4 antibodies are present in ∼30% of Japanese OSMS patients over a wide range of disease durations. Patients with these antibodies have various distinct features. However, there are also cases of anti-AQP4 antibody-negative OSMS with extremely long spinal cord lesions and few brain lesions, indicating that the mechanisms of LESCL production in OSMS are heterogeneous, i.e. both anti-AQP4 antibody-related and -unrelated. Further characterization of the precise actions of anti-AQP4 antibodies in anti-AQP4 antibody-positive MS/NMO patients, and searches for other autoantigens in anti-AQP4 antibody-negative OSMS patients may shed light on the relationships among NMO, OSMS and CMS at the molecular level.

Acknowledgements

We wish to thank Professors Allan G. Kermode and William Carroll (Australian Neuromuscular Research Institute, Sir Charles Gairdner Hospital), and Professor Brian G. Weinshenker (Department of Neurology, Mayo Clinic College of Medicine), for valuable comments on the manuscript. This work was supported in part by a Neuroimmunological Disease Research Committee grant from the Ministry of Health, Labour and Welfare, Japan.

Footnotes

  • Abbreviations:
    Abbreviations:
    ANOVA
    analysis of variance
    AQP4
    aquaporin-4
    ATM
    acute transverse myelitis
    BSMS
    brainstem–spinal form of multiple sclerosis
    CMS
    conventional form of multiple sclerosis
    CNS
    central nervous system
    CSF
    cerebrospinal fluid
    EDSS
    Expanded Disability Status Scale of Kurtzke
    FS
    Visual Functional Scale of Kurtzke
    GFP
    green fluorescent protein
    IRTM
    idiopathic recurrent transverse myelitis
    LESCL
    longitudinally extensive spinal cord lesion
    MRI
    magnetic resonance imaging
    MS
    multiple sclerosis
    NMO
    neuromyelitis optica
    OB
    oligoclonal band
    OND
    other neurological disease
    OSMS
    opticospinal form of multiple sclerosis
    PBS
    phosphate-buffered saline

References

View Abstract