Antibodies to GABAA receptor &agr;1 and &ggr;2 subunits: Clinical and serologic characterization

Objective:To search for antibodies against neuronal cell surface proteins. Methods:Using immunoprecipitation from neuronal cultures and tandem mass spectrometry, we identified antibodies against the &agr;1 subunit of the &ggr;-aminobutyric acid A receptor (GABAAR) in a patient whose immunoglobulin G (IgG) antibodies bound to hippocampal neurons. We searched 2,548 sera for antibodies binding to GABAAR &agr;, &bgr;, and &ggr; subunits on live HEK293 cells and identified the class, subclass, and GABAAR subunit specificities of the positive samples. Results:GABAAR-Abs were identified in 40 of 2,046 (2%) referred sera previously found negative for neuronal antibodies, in 5/502 (1%) previously positive for other neuronal surface antibodies, but not in 92 healthy individuals. The antibodies in 40% bound to either the &agr;1 (9/45, 20%) or the &ggr;2 subunits (9/45, 20%) and were of IgG1 (94%) or IgG3 (6%) subclass. The remaining 60% had lower antibody titers (p = 0.0005), which were mainly immunoglobulin M (IgM) (p = 0.0025), and showed no defined subunit specificity. Incubation of primary hippocampal neurons with GABAAR IgG1 sera reduced surface GABAAR membrane expression. The clinical features of 15 patients (GABAAR &agr;1 n = 6, &ggr;2 n = 5, undefined n = 4) included seizures (47%), memory impairment (47%), hallucinations (33%), or anxiety (20%). Most patients had not been given immunotherapies, but one with new-onset treatment-resistant catatonia made substantial improvement after plasma exchange. Conclusions:The GABAAR &agr;1 and &ggr;2 are new targets for antibodies in autoimmune neurologic disease. The full spectrum of clinical features, treatment responses, correlation with antibody specificity, and in particular the role of the IgM antibodies will need to be assessed in future studies.

Antibodies to the a1 and b3 subunits of GABA A R, 2 subunits of the heteropentameric ligand gated ion channel that mediates the majority of inhibitory neurotransmission in the brain, were recently reported in 18 patients. 12 The 6 patients with high serum and CSF GABA A R-Abs presented mainly with seizures and refractory status epilepticus, whereas lower serum titers, without CSF antibodies, were observed in 12 patients with broader neurologic diagnoses including stiff-person syndrome and adult-onset opsoclonus myoclonus syndrome.
We independently identified the GABA A a1 subunit and the novel g2 subunit as antibody targets and, using a live cell-based assay, detected them in 45 of a total of 2,548 sera referred for other CNS antibody tests. GABA A R-Abs fell into 2 broad groups defined by their subunit specificity, titer, and immunoglobulin class or subclass.
METHODS Standard protocol approvals, registrations, and patient consents. The research use of referred sera is approved by the Oxfordshire Research Ethics Committee A (07 Q160X/28). When GABA A R was identified, and a cell-based assay (CBA) established, 502 sera with voltage-gated potassium channel (VGKC) complex, NMDAR, or other antibodies, 92 healthy and 112 disease control sera (table 1), and a further 2,046 referred sera negative for the requested antibodies were tested for the presence of GABA A R-Abs. Brief clinical data were requested from referring neurologists of positive sera. Specific written consent was obtained from patient 2 for inclusion of his case report and videos.
Immunoprecipitation from cortical neurons. To identify new neuronal antibodies, sera were tested for binding to cultured primary rat hippocampal and cortical neurons. A serum with very strong binding was chosen for further study. The patient's IgG was bound to rat cortical neurons, and the immune complexes solubilized with 2% digitonin and captured using Protein G-Sepharose beads (Sigma, Dorset, UK). The immunoprecipitate was separated by gel electrophoresis and the GABA A R a1 subunit was identified as the target by mass spectroscopy from a sample of digested bands from the patient, but not healthy control, immunoprecipitate.
Expression of GABA A R in transfected human embryonic kidney cells. As for other antibody tests, [13][14][15] individual GABA A R subunits (a1, b2, b3, or g2) were individually or coexpressed in human embryonic kidney 293 (HEK293) T cells and cell surface expression examined (e-Methods, tables e-1 and e-2 on the Neurology ® Web site at Neurology.org). Antibody reactivity was initially assessed using HEK293 cells coexpressing a1b2g2 GABA A R subunits and binding detected with Alexa Fluor 568 goat antihuman IgG (H 1 L) (1:750, A-21090, Invitrogen, Paisley, UK). All sera were scored (0: negative, 1: low positive, 2-4: positive) and colocalized with a commercial antibody against the a1 subunit of the GABA A R (1:500, clone N95/35, Antibodies Inc., Davis, CA). Endpoint dilution titers were established by determining the last dilution at which binding was scored as 1.
To determine subunit specificities, all positive sera were tested by CBAs for binding to different GABA A R subunit combinations.
Effects of patient antibodies on GABA A R expression in vitro. Primary P0 rat neuronal cultures (DIV 7) were incubated for 3 days with patient or healthy control serum (1:100; heated at 56°C for 30 minutes to inactivate complement). Subsequently, surface proteins were biotinylated, the cells lysed, and biotinylated surface proteins isolated on a NeutrAvidin agarose column (89881, Pierce Biotechnology, Rockford, IL). Isolated membrane proteins were eluted in sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) buffer (Invitrogen) containing 50 mM dithiothreitol; equal amounts of samples were then analyzed by SDS-PAGE and Western blot probing for the a1 and g2 subunits of GABA A R. Antibody to the transferrin receptor (13-6800, Invitrogen) was used as a cell surface fraction loading control. Quantification of GABA A R receptor loss was determined by densitometric analysis of the Western blots using ImageJ software, and calculated as the ratio of a1: transferrin receptor and g2:transferrin receptor.

(2)
Abbreviations: AMPAR 5 a-amino-3-hydroxy-5-methyl-4isoxazol-propionic acid receptor; CASPR2 5 contactinassociated protein-like 2; GABA A R 5 g-aminobutyric acid A receptor; LGI1 5 leucine-rich, glioma-inactivated 1; disorder and increased anxiety but without psychosis. She was seen by a neurologist but there was no objective evidence of encephalitis and she returned to her care home. Subsequently, serum VGKC complex antibodies were reported (1938 pM), but all other antibody tests (antibodies to LGI1, CASPR2, NMDAR, AMPAR, GABA B R, glycine R) were negative. However, her serum IgG bound intensely to the surface of both hippocampal and cortical neuronal cultures, indicating the presence of a potentially pathogenic antibody against a neuronal surface protein ( figure 1A). Using immunoprecipitation and mass spectrometry (see Methods), her serum antibodies were found to bind to the a1 subunit of GABA A R (figure e-1A). GABA A R in the immunoprecipitate from the patient, but not from a healthy individual, was confirmed by Western blotting (figure 1B).
Detection of GABA A R-Abs in patient sera. In vivo, GABA A R is composed of multiple subunits (a1-6, b1-3, g1-3, p, e, u), which combine to form heteropentamers with a central pore; the a1b2g2 is the most abundant neuronal GABA A R subtype. 16 Individual homomeric GABA A R subunits and heteropentameric GABA A Rs (a1b2g2 subunits) were expressed in HEK cells and their cell surface expression assessed. Immunostaining of permeabilized fixed cells showed intracellular pools of all of the GABA A R subunits (figure e-1B), but surface GA-BA A R expression was only found with cotransfection of all 3 GABA A R subunits, and we used a1b2g2 to establish the CBA. Patient 1's antibody bound to the surface of live GABA A R-transfected cells, colocalizing with commercial GABA A R a1 subunit antibody (figure 1C).
GABA A R-Abs in patients and controls. Sera from healthy and disease controls (table 1) did not bind to GABA A R-transfected cells (healthy control mean 1 3 SD 5 0.28, figure 1D). Only 2 of 108 (1.9%) additional sera positive for VGKC complex antibodies were positive for GABA A R-Abs, and adsorption of patient 1's serum showed that GABA A R was not a After identification of g-aminobutyric acid A receptor (GABA A R) peptides by tandem mass spectroscopy of the immunoprecipitate from cultured neurons, its presence was confirmed by Western blotting using a commercial antibody against the a1 subunit of GABA A R (52 kDa); cortical brain homogenate (Cx) was used as positive control. (C) A cell-based assay was developed using human embryonic kidney 293 cells cotransfected the a1, b2, and g2 subunits of GABA A R. Antibody binding to GABA A R was demonstrated with serum from patient 1 (red), which colocalized with commercial antibody against the a1 subunit (green; upper row). Immunoglobulin G (IgG) immunoreactivity to GABA A R was not observed with control serum (lower row).
(D) GABA A R-Abs were identified in 5/502 sera with known antibodies (3 voltage-gated potassium channel complex, 2 NMDAR-Abs), and 40/2,046 sera previously found negative in other routine antibody tests. Samples scoring above 1 (dotted line) are considered positive. GABA A R-Abs were not present in healthy (n 5 92) or disease (n 5 112) control sera. Scale bars are 30 mm.
component of the VGKC complex (figure e-2). GABA A R a1b2g2 antibodies were detected in only 2 of 393 (0.5%) sera positive for other known neuronal surface antibodies but were present in 40 of 2,046 (2%) sera previously found negative for NMDAR, AMPAR, or GABA B R antibodies (table 1). Serum endpoint titers were between 1:80 and 1:4,860, and all 45 GABA A R-Abs-positive sera bound to live hippocampal neurons (as in figure 1A). There were no CSF samples available for testing from these patients.
IgG class and subclass of GABA A R-Abs. The antihuman IgG (H 1 L, A-21090, Invitrogen) is widely used for Specific GABA A receptor subunit reactivities and immunoglobulin classes (A) In the index patient 1, substitution of the a1 subunit with the a2, a3, or a5 subunit ablated binding to the g-aminobutyric acid A receptor (GABA A R)transfected cells, illustrating that the a1 subunit was the antigenic target. a1-Specific antibodies were observed in a further 8 patients (20% of total 45). Case 8 illustrates 1 of 9 sera (20%) that bound only to GABA A Rs containing the g2 subunit, but not the g1 subunit (second row). The third row shows sera from patient 15, which bound to all GABA A Rs without a defined subunit specificity. (B) Sera with subunit-specific GABA A R-Abs (a1 and g2) had significantly higher antibody titers than sera without a distinct subunit reactivity (Mann-Whitney, p 5 0.0005). (C) Sera with specific a1 (n 5 9, red) or g2 subunit (n 5 8, blue) antibody reactivities had IgG1 (16) or IgG3 (1) antibodies, compared to only 2/20 sera without a defined GABA A R subunit (green) antibody reactivity (both IgG1), p , 0.0001, whose antibodies were predominantly immunoglobulin M (IgM), p 5 0.025 (18/20). IgG 5 immunoglobulin G. detection of human IgG binding but as it is directed against total IgG, including both heavy and light chains, it can crossreact with the light chains shared with other classes of antibody. Using anti-IgG or anti-IgM subclassspecific secondary antibodies only 19/37 (51.4%) available sera contained IgG1 (n 5 18) or IgG3 (n 5 1) GABA A R-Abs. Seventeen of the IgG antibodies were specific for GABA A R a1 or g2 subunits; 6 of these also had IgM antibodies. By contrast, 18 of the 20 remaining sera, without a defined subunit specificity, were exclusively IgM-GABA A R-Abs (p 5 0.0025, Mann-Whitney; figure 2C).
GABA A R-Abs reduce surface GABA A R expression on cultured neurons. As patient sera bound to primary cortical neurons, we investigated the effects of 4 sera with high titers (endpoint dilutions 1:540-1:4,860) of IgG1 GABA A R a1 (n 5 2, patients 1 and 2) or g2 subunits (n 5 2, patients 8 and 9) on GABA A R expression in vitro, as described by others. 17 Neuronal cultures were exposed to the heat-inactivated sera (1:100) for 72 hours, and GABA A R expression assessed by Western blot. All 4 patient sera caused a reduction in the surface expression of both a1 and g2 subunits when compared to neurons treated with 2 control sera (p 5 0.0023 and p 5 0.0067, respectively, figure 3, A and B). We subsequently obtained brief clinical data for 15 representative patients, 6 with a1 and 5 with g2 GABA A R-Abs specificity, and 4 with undefined subunit specificity (table 2). The most common presenting features were seizures (n 5 7, 47%), memory impairment (n 5 7, 47%) with confusion or disorientation (n 5 4, 27%), or psychiatric features (n 5 5, 33%) with hallucinations (n 5 2, 33%) or anxiety (n 5 4, 27%). One 13-year-old girl had a dysembryoplastic neuroepithelial tumor resected earlier in life, with established neurodevelopment problems, but presented with unexplained onset of behavioral disturbance.
CSF had been examined at presentation in only 4/15 patients; 3/4 were normal. MRIs were normal in 4/9 or not performed (6/15). In the few with informative MRIs, patient 3 (a1-specific antibodies) had unilateral hippocampal high signal but this was thought at the time to be due to temporal lobe seizures rather than autoimmune encephalitis. Patient 10 (g2-specific antibodies), with a small head of caudate on MRI, subsequently was diagnosed with Huntington disease. In 2 other patients there were nonspecific white matter lesions. Patient 14 (subunit undefined) was in remission from non-Hodgkin lymphoma after treatment when she presented with personality changes, memory loss, and confusion. CSF showed lymphocytic pleocytosis and oligoclonal bands, and there was temporal lobe high signal on MRI. The changes were consistent with a paraneoplastic limbic encephalitis rather than direct infiltration. However, there was mediastinal recurrence of the lymphoma and she failed to respond to intrathecal chemotherapy or immunotherapies, dying soon after. At further follow-up (either verbal or at clinic visit, up to 12 months after reporting the antibodies), 4 patients had done well on symptomatic treatment only, and a further 2 appeared to have improved spontaneously. Of the 3 patients who received some immunotherapy, one (patient 2, a1-specific antibodies) clearly improved (see below) and one showed modest improvement but one was declining after a short course. The others were either lost to follow-up or had died (see table 2).    Six months after PEX, the patient relapsed with bizarre and unpredictable behavior (e.g., sudden onset intense handwashing for a few days followed by new-onset praying) and GABA A R-Abs were once again detected in his serum (1:180), but were undetectable in his CSF. There were subtle motor symptoms of catatonia and subtle frontal symptoms (BFCRS 5/FAB 11), and he received PEX again; symptoms of catatonia resolved within 2 weeks. He continued to have reduced verbal fluency, severe apathy, emotional and social withdrawal, blunted affect, and difficulty in abstract thinking (FAB 9/18). He received 5 days of methylprednisolone 1,000 mg IV and 5 days of immunoglobulins in January 2014 and was started on 1 mg/kg/OD prednisolone, which was weaned down to 10 mg/day in August 2014. His FAB returned to 17/18 in April 2014 (video 2). These symptoms improved slowly in the following months but did not disappear, perhaps due to the continued use of olanzapine and fluoxetine and an underlying diagnosis of mild Asperger syndrome made at age 11. His last GABA A R-Abs levels in April 2014 were undetectable. At this time his FAB was 17/18. His symptoms of catatonia and frontal dysfunction twice improved, strongly linked to disappearance of GABA A R-Abs with immunotherapy. DISCUSSION We identified a new antibody target, GABA A R, established a CBA using HEK cells expressing heteropentameric GABA A Rs, and identified a total of 45 patients with GABA A R-Abs. The antibodies in 40% of patients were IgG1 or IgG3, bound to GABA A Rs containing a1 or g2 subunits, and all 4 sera tested were able to reduce GABA A R expression on live cortical neurons. In the remaining 60%, however, the titers were lower, the antibodies were mainly IgM, and they did not show subunit specificity, although the sera also bound to hippocampal neurons in culture. The clinical features of 15 representative patients included seizures, psychiatric and cognitive problems, and only one had a relevant malignancy. GABA A R-Abs are relatively common (up to 2% of referred sera compared with around 4% identified with NMDAR-Abs over the same time period), are potentially pathogenic, and associate with seizure and behavioral phenotypes. However, although the clinical features were variable and the paraclinical findings often normal, one boy with severe catatonia twice improved substantially following immunotherapy in parallel with normalization of his GABA A R-Abs.
GABA A Rs are ionotropic cell surface receptors that predominantly mediate the fast-inhibitory neurotransmission in the brain, and are usually assembled as heteropentamers. On activation, influx of chloride ions results in hyperpolarization and stabilization of the neuronal membrane potential. The GABA A Rs are the therapeutic target of many clinically important drugs, such as barbiturates, benzodiazepines, and topiramate, with anticonvulsant, anxiolytic, sedative, cognitive, and mood-altering properties (reviewed in reference 20). In 18 patients, we identified a1 or g2 subunits as the main targets and showed that 4 of these were able to reduce GABA A R complexes from the neuronal surface in vitro, most likely through antibody cross-linking and internalization, as described for NMDAR and AMPAR antibodies, 2,16 supporting the idea that these antibodies are pathogenic.
This study was initially designed to identify the target for antibodies in a patient with VGKC complex antibody of 1,938 pM. Despite preadsorption of GABA A R-Abs, the patient sera still bound to cultured neurons, indicating the presence of a second cell surface neuronal antibody. Thus VGKC complex antibodies that are negative for binding LGI1/CASPR2/ Contactin-2 but bind cultured neurons require further study to identify their specific targets and to explore their pathogenicity.
The sera positive for GABA A R-Abs had all been sent for other CNS antibody tests. Although many of the patients had seizures, or cognitive or neuropsychiatric problems, they were given a range of tentative diagnoses (table 2). In most there was little to suggest a classical immune-mediated disease such as limbic encephalitis or NMDAR-Abs encephalitis, and in 2 patients a functional or psychogenic condition was suspected initially. Nevertheless, the large number of referrals for CNS autoantibodies (over 6,000 per year from the United Kingdom) and heterogeneity of the patients described here illustrates the increasing interest in identifying antibodies in patients with subacute onset of unexplained seizures or cognitive or psychiatric features. GABA A R-Abs, binding the a1 or b3 subunits, were identified recently in 6 patients with refractory status epilepticus or epilepsia partialis continua and a change in cognition/behavior with extensive imaging abnormalities 12 and in another 12 with a variety of phenotypes and lower titers. The authors did not report g2 subunit specificity or examine the immunoglobulin classes and subclasses. IgG GABA A b3 antibodies were also recently reported in 2 patients with thymoma-associated encephalopathies. 21 Both IgM and IgA NMDAR-Abs have previously been reported to be pathogenic in vitro, but their clinical relevance is not clear 22,23 ; however, the serum GABA A R-IgM-Abs identified here, although low titers, were not observed in 92 healthy control sera, and they also bound to live hippocampal neurons. This suggests that they could be pathogenic in vivo if they are able to reach the brain parenchyma, or are synthesized intrathecally. However, these possibilities clearly need further study.
As this study was retrospective in design, there are several limitations, in particular the lack of available CSF samples and limited or no immunotherapy intervention in all but 2 of the patients. Nevertheless, this study, in suggesting that a potentially pathogenic antibody can associate with clinical features that are less characteristic of the well-known autoimmune encephalitis syndromes, could have implications for the field. Future prospective studies, detecting GABA A R-Abs at onset and testing CSF, with judicious use of immunotherapy, and in vitro and in vivo experiments comparing the effects of IgG and IgM antibodies, will be important in determining their clinical relevance.

AUTHOR CONTRIBUTIONS
Philippa Pettingill: drafting/revising the manuscript, study concept or design, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/ patients, acquisition of data, statistical analysis. Holger Kramer: drafting/revising the manuscript, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/patients, acquisition of data. Jan Adriaan Coebergh: drafting/revising the manuscript, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, acquisition of data. Rosemary Pettingill: study concept or design, accepts responsibility for conduct of research and final approval, acquisition of data. Susan Maxwell: analysis or interpretation of data, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/patients. Anjan Nibber: analysis or interpretation of data, accepts responsibility for conduct of research and final approval, acquisition of data. Andrea Malaspina: drafting/revising the manuscript, analysis or interpretation of data, accepts responsibility for conduct of research and final approval. Anu Jacob: drafting/revising the manuscript, accepts responsibility for conduct of research and final approval, acquisition of data. Sarosh R. Irani: drafting/revising the manuscript, study concept or design, accepts responsibility for conduct of research and final approval, obtaining funding. Camilla Buckley: drafting/revising the manuscript, study concept or design, accepts responsibility for conduct of research and final approval, study supervision, obtaining funding. David Beeson: study concept or design, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/patients, study supervision, obtaining funding. Bethan Lang: drafting/revising the manuscript, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/patients, acquisition of data. Patrick Waters: drafting/revising the manuscript, study concept or design, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/patients, acquisition of data. Angela Vincent: drafting/ revising the manuscript, study concept or design, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, contribution of vital reagents/tools/patients, study supervision, obtaining funding.