04 (2007) 808–818www.elsevier.com/locate/yexnr

Experimental Neurology 2

A human anti-neuronal autoantibody against GABAB receptor inducesexperimental autoimmune agrypnia

Giovanni Frisullo a, Giacomo Della Marca a, Massimiliano Mirabella a, Marcella Caggiula b,Aldobrando Broccolini a, Marco Rubino a, Gioacchino Mennuni a,

Pietro Attilio Tonali a,b, Anna Paola Batocchi a,⁎

a Department of Neuroscience, Institute of Neurology, Catholic University, Largo Gemelli 8, 00168, Rome, Italyb Fondazione Don Gnocchi, Rome, Italy

Received 17 September 2006; revised 10 January 2007; accepted 11 January 2007Available online 23 January 2007

Abstract

In the serum and cerebrospinal fluid of a patient with recurrent acute episodes of respiratory crises, autonomic symptoms and total insomnia(agrypnia), we identified a novel anti-neural complement fixing antibody directed against GABAB receptor (GABABR). Patient purified IgGrecognized a band of approximately 110 kDa on protein extracts of mouse cerebellum, cortex and brainstem and immunolabelled cultured Chinesehamster ovary (CHO) cells, transfected with human GABABR1 and rat GABABR2 receptors. Western blot analysis of transfected CHOhomogenates showed the same band using both patient purified IgG and anti-GABABR1 antibody.

In order to verify the pathogenic role of these purified antibodies, we injected patient IgG intrathecally into cisterna magna of C57BL/6 micepre-implanted with EEG electrodes and we observed severe ataxia followed by breathing depression and total suppression of slow wave sleep, asevidenced by EEG recording, in a dose-dependent manner. Immunohistochemistry on brain sections of mice injected with patient IgG showed thesimultaneous presence of bound human IgG and C5b-9 deposits on Purkinje cells and cerebellar granular layer. After incubation with anti-GABABR antibody, a marked reduction of receptor immunostaining was found with relative sparing of neuronal architecture.

In conclusion we recognized an anti-neuronal autoantibody directed against GABABR that is associated with autoimmune agrypnia and weshowed that our patient purified IgG was able to induce in mice experimental autoimmune agrypnia characterized by a complex neurologicalsyndrome affecting several CNS functions.© 2007 Elsevier Inc. All rights reserved.

Keywords: GABA; Passive transfer; Autoimmunity; Sleep disorders; Agrypnia

Introduction

Agrypnia is a term defining a condition of severe andpersistent insomnia, with an EEG pattern characterized by lossof spindles, slow-wave and paradoxical sleep. It is usuallydescribed in subjects with vascular, traumatic or infectiouslesions of the brainstem or thalamus (Autret et al., 1995).Agrypnia excitata is a medical condition consisting of severeinsomnia with enacted dreams, mental confusion, motoragitation and autonomic hyperactivity (Lugaresi and Provini,

⁎ Corresponding author. Fax: +39 6 35501909.E-mail address: annapaola.batocchi@rm.unicatt.it (A.P. Batocchi).

0014-4886/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.expneurol.2007.01.012

2001). These clinical features can be found in various pathologicconditions including: Morvan's fibrillary chorea, a rare diseasewith possible autoimmune pathogenesis (Serratrice and Azulay,1994), familial and sporadic fatal insomnia (Montagna et al.,2003) and delirium tremens (Lugaresi and Provini, 2001).

We have described the clinical features of a 55-year-oldwoman affected by multiple cranial nerve palsy. At 3 yearsafter the onset of symptoms she developed recurrent acuteepisodes of total insomnia, dysautonomic signs and nocturnalrespiratory failure requiring mechanical ventilation (Batocchiet al., 2001). Her crises were characterized by initial beha-vioural changes with excitement and hyperactivity, musclefasciculations and cramps, heavy sweating, increased upper

809G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

airway secretion and salivation, worsening of dysphagia anddysphonia, ataxia, brisk tendon reflexes, vertical and horizontalnystagmus followed by hyperglycemia, hypertension, brady-cardia and breathing depression leading to respiratory failure,severe hypotension and coma. Polysomnographic (PSG) recor-dings (lasting from a minimum of 24 h to a maximum of5 consecutive days) showed a diffuse 8 to 11 Hz alpha activitypattern during both daytime and nighttime. Sleep EEG patterns(V-waves, K complexes, spindles and, particularly, low fre-quency delta waves) were not detectable during episodes ofagrypnia (Fig. 1). The patient benefited from serial plasmaexchange and high-dose corticosteroid treatment with improve-ment of clinical symptoms and recovery of REM and NREMsleep. At 10 years after the onset of symptoms the patient sud-denly died at home for unknown causes and therefore autopsywas not performed.

In the serum and cerebrospinal fluid (CSF) of our patientwe identified a new antibody directed against the metabotropicγ-aminobutyric acid B receptor (GABABR). Moreover, wedemonstrated that this antibody is pathogenic and causes asevere transient sleep disorder associated with other neurolo-gical symptoms in C57BL/6 mice by passive transfer.

Fig. 1. EEG spectrum diagrams and hypnograms of our patient. Patient EEG spectrumranging from 0 to 30 Hz. Absolute power log for each frequency is expressed by a cPatient EEG spectrum diagram showed a continuous and monomorphic spectral pattduring two agrypnic episodes (A; traces 1–2 and 4); traces 3 and 5 refer to remissionspectrum diagrams, showed neither recognizable NREM nor REM phases during the aa clear sleep/wake cycle and NREM/REM oscillations (B; traces 3 and 5). (For interpthe web version of this article.)

Materials and methods

IgG purification

IgG were purified from the serum of our patient and of 5healthy subjects using protein G columns (Hi-trap, AmershamBiosciences, Piscataway, NJ). Protein concentration wasmeasured by nephelometric assay. IgG from sera of our patientand of 5 different control individuals were processed togetherexactly in the same conditions, by using the same purificationmethod, analysed, put in aliquots and stored at −80 °C until theuse. The same batch of purified IgG was used for biochemicalstudies and passive transfer (12 months after purification) andcontrol IgG of each healthy subject were individually used.

Western blot analysis

Normal mouse cerebral cortex, brainstem and cerebellumwere homogenized in ice-cooled radioimmunoprecipitation(RIPA) buffer (150 mM NaCl, 10 mM Tris, pH 7.2, 0.1%SDS, 1% Triton X-100, 1% deoxycholate, 5 mM EDTA)supplemented with protease inhibitors. One hundred μg of

. Each plot represents a 24-h recording. Frequencies are expressed in ordinates,olour scale, ranging from 0 mV2 (colour: white) to 30,000 mV2 (colour: black).ern with reduction of the delta-EEG power and absence of circadian oscillationperiods. 24-h hypnograms (B), referring to the same days and presented as EEGcute phases of the disease. Remissions were characterized by the reappearance ofretation of the references to colour in this figure legend, the reader is referred to

810 G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

proteins from each sample were separated by electrophoresis ina 10% polyacrylamide gel and blotted onto a nitrocellulosemembrane (Schleicher & Schuell, Rellienhausen, Germany)

using standard protocols. Blots were blocked in 5% non-fat drymilk and incubated overnight at 4 °C with one of the followingantibodies: (i) anti-GABABR1 antibody (S-20 Santa Cruz

811G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

Biotechnology) diluted 1:500; (ii) patient purified IgG at theconcentration of 40 μg/ml; and (iii) control purified IgG at theconcentration of 40 μg/ml. After incubation with the appro-priate peroxidase-conjugated secondary antibody, blots weredeveloped using the ECL Western Blotting Analysis System(Amersham Biosciences).

CHO cells transfected with human GABABR1 and ratGABABR2 (Urwyler et al., 2001) were homogenized in RIPAbuffer and 100 μg of proteins were separated by electrophoresisin a 10% polyacrylamide gel, blotted onto a nitrocellulosemembrane using standard protocols and immunolabelled withanti-GABABR1 antibody or control or patient purified IgG asabove.

Morphological studies

Seven μm fresh-frozen sections from normal mouse brainwere incubated overnight at 4 °C with identical concentrations ofserum (1:250), CSF (undiluted, compared with six normalsubjects) or purified IgG from either the patient or 5 differentcontrols (20 μg/ml), followed by incubation with a peroxidase-conjugated anti-human IgG (Jackson ImmunoResearch Labo-ratories, West Grove, PA), diluted 1:1000 for 1 h at roomtemperature (RT). The sections were then rinsed and immuno-complexes visualized using the Vectastain ABC kit (VectorLaboratories, Burlingame, CA). Cultured Chinese hamster ovary(CHO) cells transfected with human GABABR1 and ratGABABR2 (kindly provided by Klemens Kaupmann; Urwyleret al., 2001) were fixed in 4% paraformaldehyde for 10 min andthen incubated with either anti-GABABR1 antibody diluted1:500, control purified IgG or patient purified IgG. Afterincubation with the appropriate FITC-labelled secondary anti-body anti-goat IgG and TR-labelled antibody anti-human IgG(Jackson ImmunoResearch Laboratories West Grove, PA, USA),cells were examined using a fluorescence microscope.

Immunohistochemistry on brain sections of mice receivingcontrol or patient IgG was performed using one of the followingantibodies: monoclonal fluorescein isothiocyanate (FITC)-conjugated anti-human IgG (Jackson ImmunoResearch Labora-tories) diluted 1:1000, monoclonal anti-complement C5b-9antibody (Dakocytomation, Glostrup, Denmark) visualizedwith Texas Red-conjugated anti-mouse IgG (Jackson Immu-noResearch Laboratories), goat polyclonal anti-GABABR1 IgG(1:500; Santa Cruz Biotechnology, Santa Cruz, CA) and mouseanti-SMI31 monoclonal antibody (1:1000; Sternberger Mono-clonals, Lutherville, MD) to evaluate neuronal architecture. Inparticular, we analysed five consecutive 7-μm-thick slices ofcerebellum from each animal injected with patient or control

Fig. 2. Immunoreactivity using anti-GABABR1 antibody and patient purified IgG onmouse cerebellum was observed using patient purified IgG (A) or anti-GABABR1pre-incubated with serial dilutions of patient purified IgG [1 mg/ml (C), 100 μg/antibody. The progressive reduction of patient purified IgG concentration during prcerebellum granular layer, thus suggesting that a patient antibody and the anti-GAScale bar of inserts showing figure details=50 μm. Immunohistochemistry on sectiopurified IgG (1 mg/ml) and then stained with anti-GAD antibody showed a strfigures=100 μm. (G) Western blot identified a band with a molecular weight of 110anti-GABABR1 antibody (H; lanes 4–6).

IgG. By using a low power objective (10×) we examined 6 non-overlapping microscopic fields and calculated the fluorescenceintensity by using Image-Pro Plus after subtraction of back-ground fluorescence evaluated in parallel slides only labelledwith secondary antibody.

Immunocompetition assay

In order to evaluate whether patient purified IgG and anti-GABABR1 or anti-glutamic acid decarboxylase (GAD65)antibodies bound to the same antigen, we pre-incubated mousebrain sections with different concentrations of patient purifiedIgG (1 mg/ml, 100 μg/ml, 50 μg/ml and 20 μg/ml) for 2 h at RT.Sections were then incubated with monoclonal anti-GABABR1antibody (S-20, Santa Cruz Biotechnology), or polyclonal anti-GAD65 antibody (Chemicon, Temecula, CA), both diluted1:500 (Ances et al., 2005).

Animals and passive transfer experiments

Female C57BL/6 mice (20–25 g) were housed under light–dark 12:12 h cycles and with constant temperature (25 °C).After anaesthesia with ketamine (100 mg/kg), scalps wereexposed and two EEG electrodes were tightly fixed to the skullwith dental cement. Ground electrode was a needle placedunder the skin, on the back of the animal. At 48 h after EEGelectrodes implantation, 20 μl of either control or patientpurified IgG at different concentrations (20 mg/ml, 40 mg/mland 100 mg/ml) were injected into the cisterna magna of themice, using a 27-gauge needle attached to a Hamilton syringe.Each mouse received a single injection of a given purified IgGpreparation.

In total, we tested 12 mice with patient purified IgG. In apreliminary study we tested each concentration of patientpurified IgG (20 mg/ml, 40 mg/ml and 100 mg/ml) in 2 mice(total of 6 mice) without EEG electrode implantation so as toevaluate the effect of patient purified IgG on animal behaviour.Then we tested each concentration of patient purified IgG(20 mg/ml, 40 mg/ml and 100 mg/ml) in 2 mice (total of 6 mice)after electrode implantation for EEG recording. Moreover, tomake sure that all IgG normal preparations did not affect theanimals, we tested 10 mice with the corresponding highestconcentration (100 mg/ml) of control purified IgG (5 withoutelectrodes implantation and 5 after electrode implantation forEEG recording). We also recorded EEG in 5 mice implantedwith electrodes to check the quality of EEG recording. Thenumber of mice being assigned IgG from the patient was limitedby the amount of material available.

frozen section of mouse brain. Identical immunoreactivity on granular layer ofantibody (B). Scale bar=50 μm. Adjacent sections of mouse cerebellum wereml (D), 50 μg/ml (E), 20 μg/ml (F)], and then stained with anti-GABABR1e-incubation resulted in a parallel increase of GABABR1 immunoreactivity onBABR1 antibody recognize the same antigen. Scale bar of figures=250 μm.n of mouse cerebellum pre-incubated with the highest concentration of patientong immunoreactivity on granular layer of mouse cerebellum. Scale bar ofkDa after incubation of blots with either patient purified IgG (H; lanes 1–3) or

812 G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

Observation and recording of animal behaviour was madeevery 2 h. Spontaneous motor activity and the severity of ataxiawere evaluated by scoring periods of videotaped behaviour ofexperimental and control animals according to previouslyestablished protocols (Rodgers et al., 2001). The animals'performances were videotaped for locomotor function observa-tion and the pattern of overground locomotion was also exa-mined by footprint analysis (Kunkel-Bagden et al., 1993). After20 h from injection (in correspondence with the peak of clinicalsymptoms in mice injected with the patient IgG) a subset ofanimals of each experimental group was sacrificed under deepanaesthesia and transcardially perfused with 4% paraformalde-hyde. Brains were excised, snap-frozen and processed for im-munocytochemistry with anti-human IgG. All experimentalprocedures on animals were conducted according to the guide-lines of the Italian Ministry of Health and were approved by thelocal ethic committee. Moreover, the patient gave written in-formed consent for the use of her serum.

Polygraphic recordings

PatientTwenty-five prolonged polygraphic (PSG) recordings, lasting

from 24 up to 120 consecutive hours, were performed at variousstages of the disease. Alternatively, aMicromed System II digitalpolygraph or a Micromed MS40 ambulatory recorder wereemployed. Montages included at least: 8 EEG leads (placed atthe following locations: Fp1, Fp2, C3, C4, T3, T4, O1, O2), withthe reference electrode placed on the left mastoid; EOG(electrooculography) leads, placed at 1 cm from the ocularcantus, with the reference electrode placed on the left mastoid;surface EMG of intercostal muscles; ECG. Surface electrodeswere filled with electrolytes and fixed with collodium, impe-dances were kept below 5 kΩ at the beginning of the recording,and periodically checked during prolonged PSGs. Samplingfrequencies were 128 or 256 Hz. PSG recordings were analyzedvisually, and wake and sleep stages were scored by epochs of30 s, according to the criteria established by Rechtschaffen andKales (1968), in order to draw hypnograms. Moreover, in orderto detect quantitative modifications of the delta band (0.25–4 Hz), we performed analysis of EEG frequency spectra bymeans of Fast Fourier Transform (FFT). EEG power spectrawere computed for consecutive 4-s epochs by an FFT routinewithin the frequency range of 0.25–30Hz. EEG spectral analysiswas performed on the C3-A2 and C4-A1 derivations.

MiceEEG electrodes were implanted under deep ketamine-

induced anaesthesia. Two 3 mm needles served as EEGelectrodes and were implanted 2 mm lateral from the midline,3 mm posterior from the bregma. The electrodes were anchoredto the skull with dental cement. Impedances were always keptbelow 5 kΩ at the beginning of the recording and periodicallychecked during the recording. Samples from EEG traces in onerepresentative experimental mouse are shown in Fig. 4. Themice were connected to recording leads for habituation 3 daysbefore the experiment. During the experiment procedure, mice

were kept in cages, relatively free to move around, and exposedto 12:12 h dark/light cycle. Five consecutive 24-h recordings ofthe EEG were obtained for each mouse. On day 0 electrodeswere placed, while days 1 and 2 were baseline recordings. Onday 3 the mice were injected with IgG. Spectral analysis wasperformed according to the methods described by Tobler et al.(1997): the EEG signal was amplified, digitally filtered (high-pass filter: 0.016 Hz; low-pass FIR filter: 30 Hz), stored anddisplayed on a PC monitor. Sampling rate was 256 Hz.Subsequently, EEG power spectra were computed for con-secutive 4-s epochs by an FFT routine within the frequencyrange of 0.25–30 Hz. EEG was continuously monitored for5 days in both experimental animals and controls.

Results

Immunohistochemistry on normal mouse brain sectionsusing patient purified IgG showed a strong patchy immunor-eactivity in the brainstem, hippocampus and granular layer ofthe cerebellum, suggesting a binding on GABAergic synapse-rich neuronal cells (Fig. 2A) similar to the pattern obtainedwith anti-GABABR1 antibody (1:500; Fig. 2B). Identicalresults were obtained using the patient serum (1:250) and CSF(undiluted) (data not shown). By Western blot analysis, patientpurified IgG recognized a band of approximately 110 kDa onprotein extracts of mouse cerebellum, brainstem and cerebralcortex (Fig. 2H, lanes 1, 2 and 3 respectively) suggesting apossible involvement of GABABR. A band of similarmolecular weight was detected using the anti-GABABR1antibody (Fig. 2H, lanes 4–6). GABABR1 immunoreactivitywas completely abolished after pre-incubation of sections onlywith high concentrations of patient purified IgG (1 mg/ml,100 μg/ml; Figs. 2C–F), whereas this competitive effect wasnot observed after pre-incubation with the anti-GAD antibody(Fig. 2G). Cultured Chinese hamster ovary (CHO) cells,transfected with human GABABR1 and rat GABABR2receptors (Urwyler et al., 2001), were immunolabelled byboth anti-GABABR1 antibody and patient purified IgG, but notby control purified IgG (Figs. 3A–F). Western blot analysis oftransfected CHO homogenates showed the same band wheneither patient purified IgG or anti-GABABR1 antibody wasused and no immunoreactivity when control IgG was used(Fig. 3G, lanes 1, 2 and 3, respectively). In order to evaluate invivo a possible pathogenic role of the patient antibody, femaleC57BL/6 mice were implanted with electrodes and EEG wasrecorded for more than 72 h (baseline EEG; days 1–2 in Figs.4B–C). Then 20 μl of either control or patient purified IgG atdifferent concentrations (20 mg/ml, 40 mg/ml and 100 mg/ml)were injected into the cisterna magna. Each mouse received asingle injection of a given purified IgG preparation. 20 h afterinjection (96 h after EEG implantation; day 3 in Figs. 4C–D),we observed a mild ataxia at low IgG concentration (20 mg/ml)and severe ataxia followed by breathing depression, massiveclonus, and unresponsiveness to stimulation when an IgGconcentration of 40 mg/ml was used. On average, all neuro-logical signs lasted for 10–12 h. In those mice, patient purifiedIgG administration, at both 20 mg/ml and 40 mg/ml

Fig. 3. Immunohistochemical studies and western blotting analysis on CHO GABABR1/GABABR2 transfected cells. CHO cells transfected with human GABABR1and rat GABABR2 incubated with either anti-GABABR1 antibody diluted 1:500, control purified IgG or patient purified IgG. There was a strong immunoreactivity ontransfected CHO cells after incubation with patient purified IgG (20 μg/ml; A) and with anti-GABABR1 antibody (B, E). No immunosignal was detected when controlpurified IgG at the same concentration was used (D). Scale bar of figures=25 μm. Western blot performed on transfected CHO homogenates detected a 110-kDa bandwhen either patient purified IgG (G; lane 1) or anti-GABABR1 antibody (G; lane 2) was used and no band when control IgG was used (G; lane 3).

813G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

814 G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

815G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

concentrations, induced an immediate suppression of thecircadian oscillation of EEG rhythms and a striking reductionof the delta band power (Fig. 4B, day 4). Delta activity showedan almost full recovery approximately 30 h after IgGadministration, in correspondence with the remission ofclinical signs (Fig. 4B, day 5). The reduction of delta-EEGpower and absence of circadian oscillation were also observedin our patient during agrypnic episodes (Figs. 1A–B, traces 1–2 and 4). The mice injected with the highest IgG dose (100 mg/ml) showed a severe respiratory impairment followed by deathwithin 12 h. In mice injected with control purified IgG, nobehavioural changes, neurological signs or significant mod-ification of the delta activity EEG profile were observed (Fig.4C) even when the highest IgG dose was used.

In order to evaluate the ability of autoantibodies to cross theblood brain barrier, we injected mice intraperitoneally withpatient purified IgG but we did not observe any neurologicalsigns or EEG changes (data not shown). To further understandthe pathogenic role of patient antibody, we performed immuno-histochemistry on brain sections of mice injected with eithercontrol or patient IgG, sacrificed at the peak of the neurologicalsigns. The simultaneous presence of bound human IgG andC5b-9deposits on Purkinje cells and cerebellar granular layer wasdemonstrated only in affected mice (Figs. 5A, B). No depositsof IgG or C5b-9 were observed on cerebellar cortex from miceinjected with control IgG (inserts in Figs. 5A and B,respectively). After incubation with anti-GABABR antibody,a marked reduction of receptor immunostaining was found inbrain sections from mice injected with patient IgG (Figs. 5C,D), with relative sparing of neuronal architecture or apparentmajor neuropathological abnormalities (not shown), while nochanges of GABABR immunostaining were observed in miceinjected with control IgG. The analysis of mean fluorescenceexpression showed a significant lower expression of GABABR1receptor (p<0.05) in cerebellum slides from mice injected withpatient IgG (MFI=108±24.7) than in slides from miceinjected with control IgG (MFI=340±23.3). On the contrary,no difference was found in SMI31 expression between thesetwo experimental groups (MFI=737±59.9 and MFI=697±57.8 in patient and control IgG, respectively; Fig. 5E).

Discussion

GABABR is able to control several neurotransmitter path-ways, modulating presynaptic and postsynaptic excitability(Calver et al., 2002). Multiple GABABR1 isoforms have been

Fig. 4. EEG spectrum diagrams of experimental animal model. Immunohistochemica24-h frequency plots: each plot represents a 24-h recording. Frequencies are expressedexpressed by a colour scale, ranging from 0 mV2 (colour: white) to 30,000 mV2 (colUpper trace: EEG recorded on day 2 (baseline), during behavioural wake, characterizday 2 (baseline), during behavioural sleep, characterized by low-frequency (delta), hbehavioural sleep, characterized by fast frequencies, low amplitude activity. Calibrfrequency scale (range 0–30 Hz); on the top: time scale (hours). Spectral power is exp30,000mV2 (black colour). 24-h EEG spectrumdiagrams of 5 all-day EEG recordingsshowed an EEG spectral pattern characterized by several peaks of slow wave activit(arrow, day 3) a striking suppression of delta band power occurred, persisting for aboshowed a profile closely similar to the baseline. In mice injected with control purified(For interpretation of the references to colour in this figure legend, the reader is refer

described (Holter et al., 2005) but only one GABABR2 has beenidentified (Martin et al., 1999). Both GABABR1 and GABABR2subunits have a molecular weight of about 110 kDa. In situhybridization showed that GABABR2 and GABABR1a/b co-localize throughout the brain (Kaupmann et al., 1998) but onlyGABABR1a/b is present in the rat superior cervical ganglia(Filippov et al., 2000). GABABR1a and GABABR1b variants,that differ in their N-terminal amino acid sequence (Kaupmannet al., 1997), have different distribution patterns in the CNS(Fritschy et al., 1999): GABABR1b is mainly found in themolecular layer of the cerebellum, thalamus, cerebral cortex andspinal cord dorsal horn, while GABABR1a in the hippocampus,hypothalamus, central grey, tectum, granular cell layer ofcerebellum and brainstem (Billinton et al., 1999). In this studywe detected a new anti-neural antibody directed againstGABABR in the serum and CSF of a patient with agrypnia,autonomic symptoms and respiratory crises. The immunoreac-tivity found in mouse brainstem, hypothalamus, Purkinje andgranular cells of the cerebellum after incubation with our patientpurified IgG suggested that the antibody response was alsodirected against GABABR1a isoformwhich is more expressed inthese CNS areas. The complete absence of GABABR1immunoreactivity after pre-incubation of mouse brain sectionswith high concentrations of patient purified IgG suggested thatpatient IgG and anti-GABABR1 antibody competed for the sameantigen. In addition, we demonstrated, by passive transfer, thatpatient purified IgG were indeed pathogenic, since, in C57BL/6mice, they caused a severe transient sleep disorder associatedwith other neurological symptoms, similar to the human disease.The mice injected with patient purified IgG into the cisternamagna showed a severe ataxia followed by breathing depressionand unresponsiveness to stimulation. A total suppression of thecircadian oscillation of EEG rhythms with a striking reduction ofthe delta band power was also present. These data suggest thatpatient IgG alone sufficed to induce the syndrome in the animals,thus excluding the role of specific cell-mediated immuneresponses in the effector phase.

Although several neurological diseases have been associatedwith the presence of anti-neural autoantibodies, only antibodiesagainst glutamate receptor-1 (Sillevis Smitt et al., 2000), anti-amphiphysin (Sommer et al., 2005), anti-NMDA receptor(Kowal et al., 2004) and possibly anti-GAD (Mitoma et al.,2003) and anti-recoverin (Adamus et al., 1998) have beendemonstrated as pathogenic in animal models.

Our results indicate that an autoantibody directed againstGABABR can also cause a disease of the central nervous system

l studies on brain sections of mice injected with control and patient purified IgG.in ordinates, ranging from 0 to 30 Hz. Absolute power log for each frequency is

our: black). Three samples of EEG recordings from experimental mouse #1 (A).ed by mixed frequencies, low-amplitude activity. Middle trace: EEG recorded onigh-amplitude activity. Lower trace: EEG recorded on day 4 (agrypnia), duringation is in the lower left corner. Mouse EEG spectrum diagram. On the right:ressed in a colour scale (shown under the plots) ranging from 0 (white colour) tofrommouse injectedwith patient purified IgG (B). Baseline recordings (days 1–2)y during daytime corresponding to the sleep episodes. After IgG administrationut 30 h in association with the presence of clinical signs. On day 5 the delta bandIgG, no modification of the EEG power spectrum profile was detected (C day 3).red to the web version of this article.)

Fig. 5. Immunohistochemical studies on brain sections of mice injected with control and patient purified IgG. Immunohistochemistry on brain sections of affected miceshowed the presence of bound human IgG (A) and C5b-9 deposits (B) on Purkinje cells and cerebellar granular layer. Scale bar of figures=25 μm. No binding wasobserved when brain sections of mice injected with control purified IgG were used (insert figures). Scale bar of inserts showing figure details=75 μm. A strongreduction of GABABR1 immunostaining was observed in brain sections of affected mice (D) compared with mice injected with control IgG (C). The analysis of meanfluorescence intensity of GABABR1 and SMI31 revealed a significant reduction of GABABR1 expression with preservation of neuronal architecture evidenced byanti-neurofilament antibody (SMI31) in the mice injected with patient purified IgG (E). Statistical analysis was performed on three mice injected with control IgG(100 mg/ml) and two mice injected with patient IgG (40 mg/ml). We analyzed five consecutive 7-μm-thick slices of cerebellum from each animal injected with patientor control IgG. By using a low power objective (10×) we evaluated 6 non-overlapping microscopic fields and calculated fluorescence intensity by using Image-ProPlus. Differences in variables between groups were tested by ANOVA. Asterisk indicates statistical significances between groups: *p<0.05. Scale bar offigures=25 μm.

816 G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

by binding to a neuronal receptor. In addition to the clinicalsigns, the injection of anti-GABAB antibodies induced inanimals the characteristic neurophysiological signs found inthe human disease.

It has been recently observed that GABAB receptorsmodulate hypocretin/orexin neuron activity via both pre- andpostsynaptic mechanisms, which may underlie the promotion ofnon-rapid eye movement sleep (Xie et al., 2006). Thismodulation mainly occurs at subcortical levels, specifically inbrainstem and hypothalamic structures, which were involved inthe clinical syndrome described in our patient. Therefore, aninvolvement of a hypocretinergic–GABAergic circuitry, leadingto daytime hypersomnia and nocturnal disruption of NREMsleep (loss of spindles and delta EEG activity) could behypothesized in our patient. In particular, at the onset of thedisease, she showed a sleep disorder characterized by daytimehypersomnolence and nocturnal sleep disruption, mimickingsome features of the narcolepsy syndrome, which progressivelyevolved into an overt agrypnia. In some stages of the disease, the

peculiar combination of persistence of REM sleep episodes,together with the suppression of NREM sleep, could havecaused specific symptoms, such as hallucination and delusions.

Moreover, we detected the simultaneous presence of boundhuman IgG and C5b-9 on cerebellar Purkinje cells andcerebellar granular layer cells in brain sections of affectedmice. These findings demonstrate that patient IgG were able tobind neurons and to fix complement, suggesting a complement-dependent impairment of GABAB receptor. In particular, themarked reduction of GABABR1 immunoreactivity on brainsections of affected mice confirms that patient autoantibodieswere directed against this neuronal receptor and indicates thatthe anti-GABABR1 antibody may substantially affect thereceptors perhaps by masking or reducing them. It has beendemonstrated also in other autoimmune diseases that comple-ment-fixing autoantibodies result in loss of receptors. Inmyasthenia gravis, anti-acetylcholine receptor antibodies acti-vate the complement cascade at the neuromuscular junction andcause the reduction of the receptors on the postsynaptic

817G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

membrane (Ruff and Lennon, 1998). The transient duration ofsymptoms observed in our experimental model could be due toseveral variables (i.e. time of GABABR turnover, titre of anti-GABABR antibody and affinity to the antigen), similarly towhat has been found in other passive transfer models (SillevisSmitt et al., 2000; Xie et al., 2006; Lindstrom et al., 1976;Sommer et al., 2005). Further experiments will be necessary toexplore the link between reduction of GABABR immunor-eactivity associated with complement deposition and transientduration of disease signs. We can speculate that clinical signscorrelate with functional involvement of neurons beforeirreversible structural damage arises and that recovery ofGABABR function can be due to secondary compensatorystrategies. Pharmacological studies on animals have shown thatGABABR is involved in the regulation of several CNSfunctions. GABABR seems to play a crucial role on sleep–waking cycle and slow-wave sleep (Gauthier et al., 1997) andon the central control of cardio-respiratory system by regulatingheart rate and respiratory rhythm generation (Trippenbach andLake, 1994). GABABR antagonists decrease slow wave sleep,induce a progressive increase of cortical low-voltage activity(waking and paradoxical sleep) (Gottesmann, 2002). The block-ade of GABABR by CGP55845A, a GABAB receptor antago-nist, increases the respiratory frequency in brainstem slices ofneonatal mice, whereas it causes a depression of respiratoryburst frequency in older neonatal preparations (Zhang et al.,2002). The disappearance of delta EEG activity and sleepspindles during night–day cycles and the presence of breathingdepression leading to respiratory failure both in our patient andin the passive transfer animal model can be explained by GABAcircuit impairment caused by anti-GABABR autoantibodies.GABABR involvement can also account for other clinicalsymptoms of our patient, such as dysphagia and oesophagealhypomobility. In fact, GABABR regulates the oesophagealphase of swallowing by inhibiting central circuits and reducingthe output of acetylcholine from vagal preganglionic motoneur-ons (Blackshaw, 2001). Furthermore, our patient during criseswas ataxic just like the mice injected with her purified IgG. Thisspecific symptom can be due to an attenuation of GABAB

receptor-mediated physiological inhibition of excitatory trans-mission on Purkinje cells, as recently evidenced also incerebellar ataxia mediated by anti-glutamate decarboxylaseautoantibody (Mitoma et al., 2003).

Our study demonstrates that, in our patient, GABA circuitimpairment caused a complex neurological syndrome affectingseveral CNS functions including the generators of delta sleepand the sleep–wake cycle, as demonstrated by electrophy-siology. By passive transfer, patient purified IgG was able toreproduce clinical and neurophysiological signs in animals,similar to the human disease. The anti-GABABR fraction of IgGfound in the serum and CSF seems to play a major role in thepathogenic mechanisms of our patient's syndrome but pre-absorption studies are necessary to confirm this hypothesis. Ofcourse we cannot completely rule out the possible role of someother IgG antibodies in the pathogenesis of neurological signs.Therefore, we propose that GABABR involvement should beconsidered in every case of agrypnia excitata.

Acknowledgments

Professor Klemens Kaupmann (University of Basel, andNovartis Institutes for Biomedical Research, Novartis PharmaAG, Basel, Switzerland) generously provided human GABA-BR1 and rat GABABR2 transfected CHO cells and helpfulcomments on the manuscript.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.expneurol.2007.01.012.

References

Adamus, G., Machnicki, M., Elerding, H., Sugden, B., Blocker, Y.S., Fox, D.A.,1998. Antibodies to recoverin induce apoptosis of photoreceptor and bipolarcells in vivo. J. Autoimmun. 11 (5), 523–533.

Ances, B.M., Vitaliani, R., Taylor, R.A., Liebeskind, D.S., Voloschin, A.,Houghton, D.J., Galetta, S.L., Dichter, M., Alavi, A., Rosenfeld, M.R.,Dalmau, J., 2005. Treatment-responsive limbic encephalitis identified byneuropil antibodies: MRI and PET correlates. Brain 128 (8), 1764–1777.

Autret, A., Henry-Le Bras, F., Duvelleroy-Hommet, C., Lucas, B., de Toffol, B.,1995. Agrypnia (organic insomnia). Neurophysiol. Clin. 25 (6), 360–366.

Batocchi, A.P., Della Marca, G., Mirabella, M., Caggiula, M., Frisullo, G.,Mennuni, G.F., Tonali, P.A., 2001. Relapsing–remitting autoimmuneagrypnia. Ann. Neurol. 50 (5), 668–671.

Billinton, A., Upton, N., Bowery, N.G., 1999. GABA(B) receptor isoformsGBR1a and GBR1b, appear to be associated with pre- and post-synapticelements respectively in rat and human cerebellum. Br. J. Pharmacol. 126(6), 1387–1392.

Blackshaw, L.A., 2001. Receptors and transmission in the brain–gut axis: potentialfor novel therapies: IV. GABAB receptors in the brain–gastroesophageal axis.Am. J. Physiol.: Gasterointest. Liver Physiol. 281, G311–G315.

Calver, A.R., Davies, C.H., Pangalos, M.N., 2002. GABAB receptors: frommonogamy to promiscuity. Neurosignals 11, 299–314.

Filippov, A.K., Couve, A., Pangalos, M.N., Walsh, F.S., Brown, D.A., Moss,S.J., 2000. Heteromeric assembly of GABA(B)R1 and GABA(B)R2 recep-tor subunits inhibits Ca(2+) current in sympathetic neurons. J. Neurosci. 20(8), 2867–2874.

Fritschy, J.M., Meskenaite, V., Weinmann, O., Honer, M., Benke, D., Mohler,H., 1999. GABAB receptor splice variants GB1a and GB1b in rat brain:developmental regulation, cellular distribution and extrasynaptic localiza-tion. Eur. J. Neurosci. 11 (3), 168–761.

Gauthier, P., Arnaud, C., Gandolfo, G., Gottesmann, C., 1997. Influence of aGABA(B) receptor antagonist on the sleep–waking cycle in rat. Brain Res.773 (1–2), 8–14.

Gottesmann, C., 2002. GABA mechanisms and sleep. Neuroscience 111 (2),231–239.

Holter, J., Davies, J., Leresche, N., Crunelli, V., Carter, D.A., 2005. Identificationof two further splice variants of GABABR1 characterizes the conservedmicro-exon 4 as a hot spot for regulated splicing in the rat brain. J. Mol.Neurosci. 26 (1), 99–108.

Kaupmann, K., Huggel, K., Heid, J., Flor, P.J., Bischoff, S., Mickel, S.J.,McMaster, G., Angst, C., Bittiger, H., Froestl, W., Bettler, B., 1997.Expression cloning of GABA(B) receptors uncovers similarity to metabo-tropic glutamate receptors. Nature 386 (6622), 239–246.

Kaupmann, K., Schuler, V., Mosbacher, J., Bischoff, S., Bittiger, H., Heid, J.,Froestl, W., Leonhard, S., Pfaff, T., Karschin, A., Bettler, B., 1998. Humangamma-aminobutyric acid type B receptors are differentially expressed andregulate inwardly rectifying K+ channels. Proc. Natl. Acad. Sci. U. S. A. 95(25), 14991–14996.

Kowal, C., De Giorgio, L.A., Nakaoka, T., Hetherington, H., Huerta, P.T.,Diamont, B., Volpe, B.T., 2004. Cognition and immunity: antibody impairsmemory. Immunity 21, 179–188.

818 G. Frisullo et al. / Experimental Neurology 204 (2007) 808–818

Kunkel-Bagden, E., Dai, H.N., Bregman, B.S., 1993. Methods to assess thedevelopment and recovery of locomotor function after spinal cord injury inrats. Exp. Neurol. 119 (2), 153–164 (Feb).

Lindstrom, J.M., Engel, A.G., Seybold, M.E., Lennon, V.A., Lambert, E.H.,1976. Pathological mechanisms in experimental autoimmune myastheniagravis: II. Passive transfer of experimental autoimmune myasthenia gravis inrats with anti-acetylcholine receptor antibodies. J. Exp. Med. 144 (3),739–753.

Lugaresi, E., Provini, F., 2001. Agrypnia excitata: clinical features and patho-physiological implications. Sleep Med. Rev. 5 (4), 313–322.

Martin, S.C., Russek, S.J., Farb, D.H., 1999. Molecular identification of thehuman GABABR2: cell surface expression and coupling to adenylyl cyclasein the absence of GABABR1. Mol. Cell. Neurosci. 13 (3), 180–191.

Mitoma, H., Ishida, K., Shizuka-Ikeda, M., Mizusawa, H., 2003. Dual impair-ment of GABAA and GABAB-receptor-mediated synaptic response by auto-antibodies to glutamic acid decarboxylase. J. Neurol. Sci. 208 (1–2), 51–56.

Montagna, P., Gambetti, P., Cortelli, P., Lugaresi, E., 2003. Familiar andsporadic fatal insomnia. Lancet Neurol. 2, 167–176.

Rechtschaffen, A., Kales, A., 1968. A manual of standardized terminology,techniques and scoring system for sleep stages of human subjects. U.S.Public Health Service, U.S. Government Printing Office, Washington D.C.

Rodgers, R.J., Halford, J.C., Nunes de Souza, R.L., Canto de Souza, A.L., Piper,D.C., Arch, J.R., Upton, N., Porter, R.A., Johns, A., Blundell, J.E., 2001.SB-334867, a selective orexin-1 receptor antagonist, enhances behaviouralsatiety and blocks the hyperphagic effect of orexin-A in rats. Eur. J.Neurosci. 13 (7), 1444–14452.

Ruff, R.L., Lennon, V.A., 1998. End-plate voltage-gated sodium channels are

lost in clinical and experimental myasthenia gravis. Ann. Neurol. 43 (3),370–379.

Serratrice, G., Azulay, J.P., 1994. What is left of Morvan's fibrillary chorea?Rev. Neurol. (Paris) 150 (4), 257–265.

Sillevis Smitt, P., Kinoshita, A., De Leeuw, B.,Moll,W., Coesmans,M., Jaarsma,D., Henzen-Logmans, S., Vecht, C., De Zeeuw, C., Sekiyama, N., Nakanishi,S., Shigemoto, R., 2000. Paraneoplastic cerebellar ataxia due to auto-antibodies against a glutamate receptor. N. Engl. J. Med. 342 (1), 21–27.

Sommer, C., Weishaupt, A., Brinkhoff, J., Biko, L., Wessig, C., Gold, R., Toyka,K.V., 2005. Paraneoplastic stiff-person syndrome: passive transfer to rats bymeans of IgG antibodies to amphiphysin. Lancet 365, 1406–1411.

Tobler, I., Deboer, T., Fischer, M., 1997. Sleep and sleep regulation in normaland prion protein-deficient mice. J. Neurosci. 17 (5), 1869–1879.

Trippenbach, T., Lake, N., 1994. Excitatory cardiovascular and respiratory effectsof baclofen in intact rats. Can. J. Physiol. Pharmacol. 72 (10), 1200–1207.

Urwyler, S., Mosbacher, J., Lingenhoehl, K., Heid, J., Hofstetter, K., Froestl, W.,Bettler, B., Kaupmann, K., 2001. Positive allosteric modulation of nativeand recombinant gamma-aminobutyric acid(B) receptors by 2,6-di-tert-butyl-4-(3-hydroxy-2,2-dimethyl-propyl)-phenol (CGP7930) and its alde-hyde analog CGP13501. Mol. Pharmacol. 60 (5), 963–971.

Xie, X., Crowder, T.L., Yamanaka, A., Morairty, S.R., Lewinter, R.D., Sakurai,T., Kilduff, T.S., 2006. GABAB receptor-mediated modulation of hypocretin/orexin neurones in mouse hypothalamus. J. Physiol. 574 (Pt. 2), 399–414.

Zhang, W., Barnbrock, A., Gajic, S., Pfeiffer, A., Ritter, B., 2002. Differentialontogeny of GABAB-receptor-mediated pre- and postsynaptic modulation ofGABA and glycine transmission in respiratory rhythm-generating networkin mouse. J. Physiol. 540, 435–446.