Objective To report the clinical and immunological characteristics of 22 new patients with glial fibrillar acidic protein (GFAP) autoantibodies.
Methods From January 2012 to March 2017, we recruited 451 patients with suspected neurological autoimmune disease at the Catholic University of Rome. Patients’ serum and cerebrospinal fluid (CSF) samples were tested for neural autoantibodies by immunohistochemistry on mouse and rat brain sections, by cell-based assays (CBA) and immunoblot. GFAP autoantibodies were detected by immunohistochemistry and their specificity confirmed by CBA using cells expressing human GFAPα and GFAPδ proteins, by immunoblot and immunohistochemistry on GFAP-/- mouse brain sections.
Results Serum and/or CSF IgG of 22/451 (5%) patients bound to human GFAP, of which 22/22 bound to GFAPα, 14/22 to both GFAPα and GFAPδ and none to the GFAPδ isoform only. The neurological presentation was: meningoencephalomyelitis or encephalitis in 10, movement disorder (choreoathetosis or myoclonus) in 3, anti-epileptic drugs (AED)-resistant epilepsy in 3, cerebellar ataxia in 3, myelitis in 2, optic neuritis in 1 patient. Coexisting neural autoantibodies were detected in five patients. Six patients had other autoimmune diseases. Tumours were found in 3/22 patients (breast carcinoma, 1; ovarian carcinoma, 1; thymoma, 1). Nineteen patients were treated with immunotherapy and 16 patients (84%) improved. Histopathology analysis of the leptomeningeal biopsy specimen from one patient revealed a mononuclear infiltrate with macrophages and CD8+ T cells.
Conclusions GFAP autoimmunity is not rare. The clinical spectrum encompasses meningoencephalitis, myelitis, movement disorders, epilepsy and cerebellar ataxia. Coexisting neurological and systemic autoimmunity are relatively common. Immunotherapy is beneficial in most cases.
- autoimmune neurology
- autoimmune encephalitis
Statistics from Altmetric.com
Autoimmune encephalitis and encephalomyelitis are inflammatory diseases of the central nervous system (CNS) associated with IgG antibodies binding to intracellular or plasma membrane neural antigens. Neural autoantibody detection in serum or cerebrospinal fluid (CSF) aids the diagnosis allowing the initiation of appropriate treatment.1 Autoantibodies binding to neural antigens can arise idiopathically or in the context of a primary or recurrent cancer (ie, paraneoplastic).2
Several autoantibodies have been recently identified in patients with neurological syndromes presenting with cognitive impairment, behavioural disturbances, seizures, movement disorders and ataxia. The molecular characterisation of the target antigens and the availability of reliable antibody assays have been critical to the characterisation of CNS diseases that often benefit from immunotherapy. Compelling evidence suggests that the autoantibodies binding to the extracellular domains of cell-surface-expressed neuronal or glial proteins such as N-methyl-D-aspartate receptor, leucine-rich glioma inactivated protein 1 and aquaporin-4 have pathogenic potential,3–6 while antibodies specific to intracellular antigens are not pathogenic but rather surrogate markers of an underlying cytotoxic T cell-mediated autoimmune response.7
Recently, a novel meningoencephalomyelitis associated with IgG autoantibodies binding to astrocytic glial fibrillar acidic protein (GFAP) isoforms has been described.8 9
Here, we report the clinical and immunological characteristics of 22 new patients with GFAP autoimmunity. We studied two GFAP isoforms, GFAPα and GFAPδ/ε (in this manuscript we use the term GFAPδ).10–12
Patients gave informed consent before participating to the study. The Ethic Committee of the Catholic University of Rome approved the study. We enrolled 451 patients with suspected neurological autoimmune disease admitted to the Catholic University of Rome between January 2012 and March 2017. We included patients with acute or subacute encephalopathy, subacute cognitive impairment, subacute cerebellar ataxia, cryptogenetic epilepsy, movement disorders of unknown origin, myelitis, subacute myelopathy, optic neuritis. Exclusion criteria were: patients with multiple sclerosis (MS), neurodegenerative, metabolic or genetic neurological disorders, brain or spinal cord tumours.
Patients’ serum and CSF when available, were tested for neural autoantibodies by indirect immunofluorescence assay on a substrate of mouse brain, kidney and stomach, by cell-based assays (CBA) and by immunoblot for antibodies binding to cell-surface and intracellular neural antigens.
In 22 patients (5%), immunohistochemical analysis on a substrate of mouse brain kidney and stomach revealed IgG binding to astrocytes in the subventricular zone, the subpial regions and to myenteric plexus. These patients are the object of this study. The final diagnoses in the remaining 429 patients were: encephalomyelitis, 22; autoimmune encephalitis, 43; viral encephalitis, 23; subacute encephalopathy, 51; cerebellar ataxia, 78; epilepsy, 92; neuromyelitis optica spectrum disorders, 22 (20 of whom were AQP4-IgG+ and 1 MOG-IgG+); myelitis, 43; optic neuritis, 26; acute disseminated encephalomyelitis, 12 (5 of whom were MOG-IgG+); movement disorder, 17.
Control samples included sera from 100 age-matched and sex-matched healthy subjects and 158 sera and 44 CSF samples from patients with other neurological diseases (patients’ diagnoses were: MS, 65; chronic inflammatory demyelinating polyneuropathy, 33; amyotrophic lateral sclerosis, 28; Parkinson’s disease 18; myopathy, 14). All patients underwent brain and spinal cord MRI, radiological screening for a systemic neoplasm and serological or CSF studies to rule out other aetiologies for their disorder.
Modified Rankin scale score (mRss) was employed to assess the treatment outcome.
All experiments were approved by the Animal Experimentation Committee of the Royal Netherlands Academy of Arts and Sciences, in accordance with the European Community Council Directive of 24 November 1986 (86/609/EEC). All animals were housed under standard conditions with ad libitum access to water and food.
Heterozygous mice with null mutations in the GFAP genes were maintained on a mixed genetic background of C57Bl6/129Sv/129Ola.13 14 Genotyping was performed as described by Pekny et al 13 and allowed distinguishing between homozygous, heterozygous and wild-type (wt) mice. Brains of homozygous GFAP-/- and wt mice littermates were used in the current study.
Indirect immunofluorescence assays on mouse and rat brain sections
Patients’ serum and CSF were evaluated for the presence of IgG binding to neural antigens by an indirect immunofluorescence assay (IFA) on frozen sections of mouse brain, kidney and stomach or rat brain sections. Mouse and rat frozen sections were fixed with 10% buffered formalin (4 min) and permeabilised with 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfate (CHAPS) 1% (4 min). After blocking for 30 min with 10% normal goat serum, sections were incubated with patients’ serum (1:200) or CSF (1:3) for 40 min. After washing with phosphate-buffered saline (PBS), the substrate was incubated (1:200) with Alexa Fluor 488 conjugated goat antihuman IgG antibody (Thermo Fisher Scientific) for 30 min. After washing with PBS, slides were mounted with ProLong Gold antifade medium without DAPI (4',6 diamidino-2-phenylindole)(Invitrogen) and visualised by fluorescence microscopy.
Five representative sera of GFAP-IgG+ patients and one control serum were tested on brain sections of GFAP-/- mice and wt littermates. Mice (aged 6–9 months) were perfusion-fixed by transcardial perfusion with 4% paraformaldehyde in PBS (pH 7.4), their brain was dissected, postfixed for 2 hours and rinsed in PBS. The brains were placed in 20% sucrose-PBS overnight and frozen over dry ice. Cryosections (10 mL) were mounted on Superfrost Plus slides (Thermo Fisher Scientific), fixed for 10 min with 4% paraformaldehyde in PBS, washed and heated in 10 mM citrate buffer, pH 6.0 in a steamer for 20 min. After cooling down, the brain sections were blocked in Tris-buffered saline (TBS, pH 7.2) containing 2% v/v normal horse serum, 1% w/v BSA, 0.1% Triton X-100, 0.05% Tween-20 for 30 min at room temperature. Then the sections were incubated with human serum (1:100 in TBS with 1% BSA) or rabbit anti-pan-GFAP (Dako 1:4000) at 4°C overnight. After washing 3×5 min with TBS, the sections were incubated with Alexa 488-conjugated antihuman IgGs (1:200) or antirabbit IgGs (1:1400) in TBS with 1% BSA. The sections were 3x in TBS and mounted in Mowiol.
Immunohistochemistry on patient’s brain specimens
One patient with GFAP-IgG (no. 19, see table 1) underwent a meningeal biopsy.
Routine diagnostic immunohistochemistry on brain and meningeal specimens was performed on a Ventana automated immunostainer (Roche-Ventana, Darmstadt, Germany) with antibodies against CD8 (Dako, Milan, Italy), epithelial membrane antigen (Dako), CD68 (Dako), GFAP (Dako), CD45 (Dako), CD3 (Dako) and CD20 (Dako).
Human embryonic kidney-293 (HEK, derived from ECACC) cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% Fetal Calf Serum (FCS) (Sigma-Aldrich) and 1% each of penicillin, streptomycin and glutamine penicillin, streptomycina and glutamine ((PSA), Invitrogen) at 37°C in an atmosphere of 5% CO2. The human astrocytoma cell line, U251MG was grown in DMEM with high glucose (Invitrogen, Paisley, UK) supplemented with 10% FCS and 1% each of penicillin and streptomycin and maintained at 37°C in an atmosphere of 5% CO2.
HEK-293 cells were transfected with plasmids encoding GFAP homo sapiens transcript variant 1 (SC118873, pCMV6-XL6-GFAPα) and variant 2 (SC325916, pCMV6-Entry-GFAPδ, OriGene) using Lipofectamine 3000 (Invitrogen). Thirty-six hours after transfection cells were fixed (4% paraformaldehyde, 10 min), permeabilised (0.2% Triton X-100, 15 min) and incubated for 1 hour at room temperature with patient serum (1:50 dilution), CSF (1:2), rabbit pan-GFAP-specific IgG (Invitrogen) (1:500), rabbit GFAPδ-specific IgG (AbCam) (1:100). After washing with PBS, cells were incubated with 10% normal goat serum for 30 min and then with secondary antibodies (1:200) for 30 min. Cells were mounted in ProLong Antifade medium with DAPI (Invitrogen) and images were acquired by fluorescence microscopy.
U251 cells were transfected with GFAPα pcDNA3 or GFAPδ pcDNA315 using polyethylenimine (PEI). For the U251 cells, western blots were performed as described by Kamphuis et al with a slight modification.16 Proteins were isolated by homogenisation in lysis buffer (0.1 mol/L NaCl, 0.01 mol/L Tris-HCl (pH 7.6) and 1 mmol EDTA (pH 8.0)) supplemented with a protease inhibitor cocktail (Roche). The samples were dissolved in 2x loading buffer (2x: 100 mmol/L Tris, 4% sodium dodecyl sulfate (SDS), 20% glycerol, 200 mmol/L dithiothreitol and 0.006% bromophenol blue) and boiled for 5 min.
Then, the samples were run on a 7.5% SDS-polyacrylamide gel electrophoresis gel and blotted semi-dry on nitrocellulose. The patient sera and the pan-GFAP (Dako) and GFAPδ (rabbit polyclonal (bleeding 27-11-2003)15 antibodies were diluted in Supermix (0.05 mol/L Tris, 0.9% NaCl, 0.25% gelatin and 0.5% Triton X-100, pH 7.4). The next day, the blots were washed with TBS-T (100 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, with 0.2% Tween-20) and incubated with peroxidase conjugated antirabbit or antihuman secondary antibodies in Supermix for 1 hour at room temperature. After washing three times in PBS plus 0.05% Tween and once in PBS, blots were developed with Enhanced Chemiluminescent Western Blotting Substrate (Pierce).
Coexisting autoantibodies testing
All patients’ sera and CSF samples were tested for neural antibodies as previously reported.17 18 All samples were also tested by indirect immunofluorescence on non-permeabilised/non-fixed rat hippocampal neurons as described previously.19 GABAA receptor antibodies were tested by a CBA employing HEK-293 cells transfected with cDNA encoding the human α1 (Origene: SC119668), β3 (Origene: SC125324) and γ2 subunits (Origene: SC121595). MOG-IgG was tested by an in-house CBA using both flow cytometry and immunocytochemistry employing as substrate, HEK-293 cells transfected with the cDNA coding for the human full isoform of MOG.
In sum, the antibodies specific for the following antigens were tested: NMDA, AMPA, GABAA, GABAB, glycine receptors, AQP4, MOG, Lgi1, Caspr2, DPPX, IGLON5, mGluR1, mGluR5, GAD65 and onconeural proteins (Hu, Yo, Ri, amphiphysin, CRMP5, Tr). The different assays performed for neural autoantibodies testing are summarised in online supplementary figure 1.
Supplementary file 1
Graphs and statistical analysis were carried out with GraphPad Prism V.7.0. Summary statistics were reported as median (range, minimum-maximum) for continuous variables and as frequencies and percentages for categorical variables. Clinical data were compared using the Wilcoxon matched pair test and two-tailed Fisher’s exact test.
Patients’ IgG binding to GFAP isoforms
Patients’ sera and CSF (when available) showed a characteristic immunoreactive pattern reminiscent of the intermediate filament cytoskeleton protein GFAP of astrocytes in pial, subpial and subventricular regions of the mouse brain (figure 1A-C). Enteric glia was also stained (not shown). The immunostaining pattern on rat brain confirmed that the CSF and sera were binding to GFAP (Figure 1D-F). None of the control sera or the CSF showed any immunoreactivity on the brain sections of mice and rats, indicating that antibodies binding to astrocytic cytoskeletal proteins were not present in control subjects. To prove that the antibodies were specifically interacting with GFAP, we tested five representative sera of patients with GFAP-IgG on brain sections of GFAP-/- mice brain. Patients’ IgG as well as a commercial antibody specific for all GFAP isoforms stained astrocytes on wt mouse brain sections, while no immunoreactivity was observed on GFAP-/- mouse brain sections (figure 2A-I).
CBAs revealed that IgG in the sera of all patients bound to GFAPα isoform and in 14/22 patients (64%) to both GFAPα and GFAPδ isoforms (figure 3A-F). Immunoblots revealed IgG binding to GFAPα in all patients and to GFAPα and GFAPδ in 13/22 patients (59%) (figure 3G). None of the patients’ IgG bound only to the GFAPδ isoform. Sixty-six sera from healthy subjects, 25 sera from patients with other neurological diseases (MS, 15; chronic inflammatory demyelinating polyneuropathy, 5; amyotrophic lateral sclerosis, 5) and CSF samples from 12 patients with MS served as control for CBA and immunoblot assay (figure 3H-J). IgG in the serum of one patient with other neurological diseases bound to GFAPα expressing cells. However, the serum did not yield any immunoreactivity on mouse or rat brain sections. The CSF of 12 patients was tested for neural antibodies. GFAP-IgG and was detected in eight patients (67%). None of the CSF samples from MS patients yielded any positive results on CBA or immunoblot.
Clinical characteristics of patients with GFAP-IgG
The demographics, clinical characteristics, CSF and serologic findings of the patients are summarised in table 1. The median age of the 22 subjects was 52 years (range: 6–80), 13 patients were females (59%).
Eleven patients presented with acute neurological symptoms, eight with a subacute onset (see table 1), while three patients with epilepsy had a chronic disease. The most common clinical presentation was meningoencephalitis with or without myelitis (10 patients, 45%; see table 1). Patient no. 9 presented with headache and rigour nucalis. Tuberculous meningitis was initially suspected because hypoglicorrachia and pleocytosis were observed on CSF analysis, and subsequently excluded on the basis of negative results of QuantiFERON Tb test and CSF molecular and microbial culture analyses for mycobacterium.
Other clinical presentations included: movement disorders in three patients; AED-resistant epilepsy in three; cerebellar ataxia in three, myelitis in two; optic neuritis in one.
Coexisting neural autoantibodies were detected in five patients (GABAAR-IgG, 1; Yo-IgG, 1; IgG binding to unclassified antigens (UNCA), 3).
All three patients with cerebellar ataxia had a coexisting neural antibody (Yo-IgG, 1; UNCA-IgG, 2). Six patients (27%) had a coexisting autoimmune disorder: rheumatoid arthritis, three; ulcerative colitis, two; psoriatic arthritis, one.
CSF analysis was performed in 14 patients and revealed abnormalities in 8 (54%). Increased protein concentration was observed in seven patients, pleocytosis in three, low glucose in one and oligoclonal bands in one case.
MRI and histopathological findings
Brain MRI was performed in all patients and spinal cord MRI in nine patients. Hyperintense lesions on T2-weighted images consistent with inflammation were present in 10 of 22 (45%) (figure 4A-F); 9 of 22 (41%) showed gadolinium enhancement. T2-hyperintense lesions of the spinal cord were detected in four patients; three of them had a lesion that extended longitudinally for more than three contiguous vertebral segments. In two patients with cerebellar ataxia, MRI showed atrophy of the cerebellum without T2-hyperintense lesions.
None of the patients with GFAP-IgG had MRI findings fulfilling the Barkhof criteria.20
In one patient (no. 19, see table 1), a leptomeningeal biopsy was performed. Histopathological analysis revealed a necrotising inflammatory process with infiltrate characterised by the presence of CD8+ lymphocytes, macrophages and some multinucleated giant cells (figure 4G, H).
Nineteen patients were treated with immunotherapy (intravenous methylprednisolone 1 g/day for 5 days, 15; intravenous immunoglobulins, 2; oral steroids, 1; plasma exchange, 1) and 16 patients (84%) improved. After treatment, the mRss decreased significantly (pretreatment mean mRss: 3.26 (95% CI 2.78 to 3.74); last follow-up mRss: 1.47 (95% CI 0.82 to 2.12); p=0.0001) (Figure 4I). Patients’ median follow-up was 8.5 months. Two patients refused immunotherapy and died from complications related to the underlying neoplasm.
Prodromal symptoms and oncological associations
Six patients (27%) presented with prodromal symptoms (fever of unknown origin, 4; dengue fever, 1; interstitial pneumonia, 1) and were initially diagnosed as having an infectious disease. Patient no. 4 developed chronic severe diarrhoea resistant to symptomatic treatment that later subsided after immunotherapy; in this case, extensive gastroenterological evaluation failed to detect any infection or an underlying inflammatory bowel disease. Neoplasms were found in 3/22 patients (14%): breast carcinoma, 1; ovarian carcinoma, 1; thymoma, 1.
We described the clinical characteristics of 22 patients with GFAP-IgG autoantibodies and we confirmed antigen specificity with IFA on brain sections of the GFAP-/- mouse. Our study provides several new insights on the spectrum of GFAP autoimmune astrocytopathy. To our knowledge, this is the first report from a European tertiary referral hospital. The 5% detection rate of GFAP-IgG in patients with suspected autoimmune neurological diseases indicates that GFAP autoimmunity is not rare. The most frequent clinical presentation was an acute-onset meningoencephalitis with or without spinal cord involvement, presenting with altered consciousness, seizures, psychosis or headache with rigour nucalis in line with the case series recently described.8 9 In these patients, brain MRI may reveal T2-hyperintense lesions of the meninges with perivascular or leptomeningeal contrast enhancement.
Interestingly, three patients in our cohort had concomitant rheumatoid arthritis. Leptomeningeal inflammation has been described in patients with rheumatoid arthritis and referred as rheumatoid meningitis, which may occur, although rarely, both in active or inactive phases of the disease. The clinical presentation can be as pachymeningitis with headache and cranial nerve palsies, or as leptomeningitis with altered consciousness, psychiatric disturbances and seizures.21 However, the pathogenetic mechanisms of the meningeal inflammation in rheumatoid arthritis remain to be clarified.
The recent identification of GFAP autoantibodies in the synovial fluid of patients with rheumatoid arthritis may support the hypothesis22 of a possible role of the GFAP autoimmunity in rheumatoid meningitis.
However, GFAP-IgG was also detected, although less frequently in patients presenting with subacute or chronic CNS symptoms such as choreoathetosis, dystonia, myoclonus and AED-resistant epilepsy. No coexisting neural antibodies were detected in these patients suggesting that GFAP autoimmunity may present with epilepsy refractory to AED or with movement disorders as the sole neurological symptoms. Patients with encephalomyelitis associated with MOG autoantibodies may have seizures as presenting symptom. However, MOG-IgG testing yielded negative results in all patients with GFAP-IgG.23
All patients with cerebellar ataxia had a coexisting autoantibody suggesting that GFAP-IgG may not be primarily associated with autoimmune cerebellar ataxia. Further studies are needed to assess the role of GFAP autoantibodies in autoimmune cerebellar ataxia.
The clinical heterogeneity of GFAP astrocytopathy is not surprising and it resembles the diverse neurological manifestations that can be observed in patients with other autoimmune or paraneoplastic syndromes that are associated with antibodies binding neural intracellular antigens like GAD65 or Hu.1 However, further studies are needed in order to assess the frequency of the diverse neurological manifestations unified by GFAP-IgG seropositivity and the long-term outcome of patients with GFAP autoimmunity. Interestingly, a different immunoreactivity of patients’ IgG was observed in mouse and rat brain sections. Differences in GFAP expression or post-translational modifications may explain the different immunostaining patterns.
We verified whether the antibodies would recognise a specific isoform of GFAP. GFAPδ differs from the more abundant GFAPα isoform only in its C-terminal tail, it has a unique 41 amino acids long C-terminal tail and is one amino acid shorter than GFAPα.24 25 It can be expected that IgG binding to GFAPα but not to GFAPδ recognise an epitope in the C-terminal tail.
The intracellular location of GFAP and its inaccessibility to circulating antibodies predict a non-pathogenic role for GFAP-IgG. The GFAP antibodies could indeed be markers of a T-cell-mediated autoimmune response.
Histopathological findings from the leptomeningeal biopsy performed in one patient with GFAP-IgG supports a CD8+ T-cell-mediated type of autoimmune response. Interestingly, it has been shown in a mouse model that GFAP-specific CD8+ T cells can spontaneously infiltrate the grey matter and white matter of the CNS, causing a relapsing-remitting encephalomyelitis.26
However, the response to immunotherapy observed in the patients described is not typical of paraneoplastic or idiopathic neurological autoimmune diseases associated with antibodies specific for intracellular neural antigens. Some of the cases herein reported resemble a previously described steroid-responsive encephalopathy, namely the non-vasculitic autoimmune inflammatory meningoencephalitis.27
Dendritic cells (DC), the most potent antigen-presenting cells are highly sensitive to corticosteroids28 29 and are particularly abundant in the meninges,30 31 the perivascular space and the juxtavascular parenchyma, sites that seems to be particularly affected in GFAP autoimmunity. A direct effect of corticosteroids on DC presenting GFAP peptides to lymphocytes may in part explain the good response to this drug observed in our patients. In 23% of patients with GFAP-IgG, other coexisting neural antibodies were detected. An extensive evaluation failed to detect the presence of coexisting antibodies binding to classified or unclassified neuronal plasma-membrane antigens in most patients with GFAP-IgG except in one case, in whom GABAA receptor antibodies were detected. However, the presence of other antibodies specific for a yet uncharacterised astrocytic plasma-membrane antigen in patients with GFAP-IgG remain to be established.
In sum, GFAP autoimmunity may present with a spectrum of neurological manifestations that encompasses seizures, cognitive and behavioural disturbances, movement disorders and myelitis. Importantly, the detection of GFAP-IgG in patients presenting with meningeal irritation without evidence of CNS infection may lead to the identification of an autoimmune meningoencephalitis that can be immunotherapy-responsive.
Contributors RI: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript. VD, AE, MG, SG, VDL, GS, JAS: acquisition of data, analysis and interpretation of data. EMH: acquisition of data, analysis and interpretation of data, critical revision of manuscript for intellectual content.
Competing interests None declared.
Ethics approval Ethic Committee of the Università Cattolica del Sacro Cuore, Rome Italy.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.