See Irani (doi: 10.1093/awv364 ) for a scientific commentary on this article.
Stiff-person syndrome is the prototype of a central nervous system disorder with autoantibodies targeting presynaptic antigens. Patients with paraneoplastic stiff-person syndrome may harbour autoantibodies to the BAR (Bin/Amphiphysin/Rvs) domain protein amphiphysin, which target its SH3 domain. These patients have neurophysiological signs of compromised central inhibition and respond to symptomatic treatment with medication enhancing GABAergic transmission. High frequency neurotransmission as observed in tonic GABAergic interneurons relies on fast exocytosis of neurotransmitters based on compensatory endocytosis. As amphiphysin is involved in clathrin-mediated endocytosis, patient autoantibodies are supposed to interfere with this function, leading to disinhibition by reduction of GABAergic neurotransmission. We here investigated the effects of human anti-amphiphysin autoantibodies on structural components of presynaptic boutons ex vivo and in vitro using electron microscopy and super-resolution direct stochastic optical reconstruction microscopy. Ultrastructural analysis of spinal cord presynaptic boutons was performed after in vivo intrathecal passive transfer of affinity-purified human anti-amphiphysin autoantibodies in rats and revealed signs of markedly disabled clathrin-mediated endocytosis. This was unmasked at high synaptic activity and characterized by a reduction of the presynaptic vesicle pool, clathrin coated intermediates, and endosome-like structures. Super-resolution microscopy of inhibitory GABAergic presynaptic boutons in primary neurons revealed that specific human anti-amphiphysin immunoglobulin G induced an increase of the essential vesicular protein synaptobrevin 2 and a reduction of synaptobrevin 7. This constellation suggests depletion of resting pool vesicles and trapping of releasable pool vesicular proteins at the plasma membrane. Similar effects were found in amphiphysin-deficient neurons from knockout mice. Application of specific patient antibodies did not show additional effects. Blocking alternative pathways of clathrin-independent endocytosis with brefeldin A reversed the autoantibody induced effects on molecular vesicle composition. Endophilin as an interaction partner of amphiphysin showed reduced clustering within presynaptic terminals. Collectively, these results point towards an autoantibody-induced structural disorganization in GABAergic synapses with profound changes in presynaptic vesicle pools, activation of alternative endocytic pathways, and potentially compensatory rearrangement of proteins involved in clathrin-mediated endocytosis. Our findings provide novel insights into synaptic pathomechanisms in a prototypic antibody-mediated central nervous system disease, which may serve as a proof-of-principle example in this evolving group of autoimmune disorders associated with autoantibodies to synaptic antigens.
The discovery of autoantibody-mediated disorders in the CNS is one of the major achievements in neurology during the past decades. The list of identified target antigens is continuously growing, but our understanding of the pathomechanisms of disease and of characteristic disease symptoms is still at the beginning ( Irani et al. , 2014 ; Leypoldt et al. , 2015 ). Stiff-person syndrome is the prototype of an autoantibody-mediated CNS disease with autoantibodies to a presynaptic antigen. In paraneoplastic stiff-person syndrome the dominant IgG autoantibodies are directed at the synaptic protein amphiphysin (encoded by AMPH ; De Camilli et al. , 1993 ; David et al. , 1994 ). Amphiphysin is a N-BAR domain protein involved in clathrin-mediated endocytosis ( Lichte et al. , 1992 ). It is important in various stages of clathrin-mediated endocytosis including membrane bending, clathrin coating, and recruitment of dynamin, then mediating membrane fission of newly retrieved presynaptic vesicles ( Takei et al. , 1996 ; Arkhipov et al. , 2009 ).
The pathogenic role of autoantibodies to amphiphysin was previously demonstrated by systemic and intrathecal passive transfer to rats ( Sommer et al. , 2005 ; Geis et al. , 2010 ). Using super-resolution stimulation emission depletion (STED) microscopy we could provide evidence that specific autoantibodies to amphiphysin bind to their antigen in spinal cord presynapses at the site of vesicle endocytosis ( Geis et al. , 2010 ). In neuronal cell culture, preabsorption and autoantibody competition assays revealed an epitope-specific process of pathogenic autoantibody internalization. Symptoms of patients with stiff-person syndrome have been attributed to decreased GABAergic neurotransmission based on the observation that drugs enhancing GABAergic transmission, e.g. benzodiazepines induce clinical improvement ( Vasconcelos and Dalakas, 2003 ; Murinson, 2004 ). Neurophysiological studies in patients and in experimental animals showed compromised tonic GABAergic inhibition as an underlying pathomechanism ( Sandbrink et al. , 2000 ; Geis et al. , 2010 ). Despite progress in understanding the stiff-person syndrome pathophysiology, the molecular events at the ultrastructural level leading to autoantibody-induced dysfunction remain unclear. Acute blocking of the amphiphysin SH3 (Src homology 3) domain with inhibitory peptides injected into reticulospinal lamprey synapses in vitro led to stimulus-dependent endocytic dysfunction with accumulation of endocytic intermediates and smaller releasable vesicle pools, thus resulting in defective neurotransmission at high synaptic activity ( Shupliakov et al. , 1997 ; Evergren et al. , 2004 ). Congenital deficiency of amphiphysin in a mouse null mutant induces memory deficits and increases seizure susceptibility as a consequence of impaired vesicle recycling ( Di Paolo et al. , 2002 ).
Synaptic vesicles differ in their molecular composition determining their affinity to the distinct vesicle pools within the presynapse. The essential vesicular soluble N -ethylmaleimide-sensitive-factor attachment receptor (v-SNARE) proteins synaptobrevin 2 (syb2, encoded by VAMP2 ) and synaptobrevin 7 (syb7, encoded by VAMP7 ) are known to be directed to the readily releasable and to resting pool vesicles, respectively ( Hua et al. , 2011 ). This provides an opportunity to differentiate the vesicle pools by their v-SNARE composition ( Fig. 1 ). The synaptic proteins endophilin and synaptojanin are direct binding partners of amphiphysin. Endophilin is involved in membrane bending, vesicle uncoating and has also been shown to be important in endocytic pathways independent of clathrin ( Milosevic et al. , 2011 ; Boucrot et al. , 2015 ). Synaptojanin interacts with amphiphysin and endophilin ( Dong et al. , 2015 ) and serves as an uncoating factor at the later steps in clathrin-mediated endocytosis ( Fig. 1 ). In the context of the hypothetical anti-amphiphysin autoantibody-induced dysfunctional clathrin-mediated endocytosis, we aimed at elucidating if these interacting synaptic proteins are affected and if they may act as compensatory factors.
Using ultrastructural analyses we here provide evidence that affinity-purified pathogenic human autoantibodies targeting the SH3 domain of amphiphysin lead to a stimulus-sensitive reduction of releasable vesicles and clathrin-coated vesicles in spinal cord presynapses in vivo .
In primary neurons, super-resolution microscopy (direct stochastic optical reconstruction microscopy, d STORM) revealed anti-amphiphysin autoantibody-induced changes of v-SNARE composition on presynaptic vesicles. Furthermore, molecular targeting of endophilin is affected by human anti-amphiphysin autoantibodies within GABAergic presynaptic terminals.
Materials and methods
Patients and therapeutic plasma exchange
The clinical details of the patients with paraneoplastic stiff-person syndrome and high titres of anti-amphiphysin autoantibodies and a control patient suffering from peripheral neuropathy without CNS disease and without specific autoantibody reactivity to neuronal antigens have been reported ( Wessig et al. , 2003 ). A commercial enzyme immunodot assay was used to determine titres of anti-amphiphysin autoantibodies with rabbit antisera raised against recombinant amphiphysin I as a positive control (H.P. Seelig). Serum titres before plasma exchange were 1–2 × 10 8 .
Amphiphysin expression and IgG preparation
In this study, we exclusively used affinity purified human IgG specifically directed at the amphiphysin SH3 domain. Affinity purification of human IgG obtained from patient plasma filtrates was performed by capturing anti-amphiphysin autoantibodies from IgG fractions by affinity chromatography and reconstitution of specific anti-amphiphysin autoantibody eluates (specAmph) as previously described ( Sommer et al. , 2005 ; Geis et al. , 2010 ). Expression and purification of human recombinant glutathione S-transferase (GST)-amphiphysin and of GST-SH3 domain fusion protein by using the gene encoding for amphiphysin and the construct containing its wild-type SH3 domain was performed as described ( Grabs et al. , 1997 ). The IgG fractions were then dialyzed separately against distilled water, freeze dried, and stored at −20°C. Lyophilized IgG was dissolved in normal saline just before use and checked by western blotting. Affinity purification resulted in an IgG fraction with anti-amphiphysin specificity of >99%. To further test the binding specificity, we performed preabsorption experiments. SpecAmph autoantibodies (0.4 µg/ml) were preincubated using recombinant SH3 domain/GST fusion protein in ascending concentrations overnight at 4°C in blocking buffer. Mouse brain lysate (100 µg total protein) was separated by gel electrophoresis and blotted. The blot was cut in 3-mm stripes and each was incubated overnight with the preabsorbed specAmph autoantibodies and GAPDH (1:10 000, Cell Signaling Technology). For detection secondary HRP-conjugated antibodies (1:2000, Dako and Santa Cruz) were incubated for 2 h at room temperature and chemiluminescence images of stripes was captured at the same time (LAS 3000, FujiFilm). Analyses of preabsorbed specAmph autoantibodies by western blot showed complete deletion of the specific band detecting amphiphysin at 128 kDa when SH3 domain fusion protein was used in excess ( Supplementary Fig. 1 ).
Intrathecal autoantibody delivery and tibial nerve stimulation
Female Lewis rats were obtained from Harlan-Winkelmann. All conducted experiments were approved by the respective State authorities and animal experiments were performed according to the ARRIVE guidelines ( Kilkenny et al. , 2010 ). Catheters for intrathecal injection of IgG were placed as described ( Geis et al. , 2010 ). Following a recovery period of 8 days, either specAmph autoantibodies or control IgG (1 mg) was injected intrathecally. Injections of 10 µl IgG solution following a flush of 10 µl saline were repeated daily for a period of 5 days, every second day for the following five injections, and every third day for the final two injections. This infusion protocol ensures fast saturation with IgG solutions and provides continuously high IgG concentration in the subarachnoid space over a longer time period as shown in previous studies ( Geis et al. , 2010 , 2012 ). Animals were deeply anaesthetized with Narcoren® (Merial) and Ia afferents of the right tibial nerve were stimulated supramaximally (8–9 V, 10 Hz, 1 min) using a Grass S88 stimulator (Grass Technologies). Immediately at the end of stimulation, animals were perfused with 2.5% glutaraldehyde and 1% paraformaldehyde. After perfusion, lumbar spinal cord was removed post-fixed and washed with 0.1 M PBS.
Tissue processing for electron microscopy
Chemicals were obtained from Sigma Aldrich if not stated otherwise. The spinal cord was cut in hemi-segments using a razorblade. At the level L4-L5 from each hemi-segment 45-µm sections were cut with a vibratome (Leica VT1000S) in 0.1 M PBS. Sections were further processed according to a published protocol with minor modifications ( Weinberg and van Eyck, 1991 ). After fixation in 1% osmium tetroxide, sections were rinsed in maleate buffer (0.05 M; pH = 6.0) and stained en bloc with uranyl acetate (1% in 0.05 M maleate buffer). Dehydration included a rising ethanol concentration series with a final rinse in 100% propylenoxide. Sections were flat embedded in epoxy resin on aclar foil (Serva). Ultrathin sections (70 nm; Ultracut E, Leica) were mounted on formvar-coated nickel grids (Plano).
Post-embedding immunolabelling for electron microscopy
GABA post-embedding immunolabelling procedures were adapted from established protocols ( Watson and Bazzaz, 2001 ; Ranson et al. , 2006 ) using an affinity purified rabbit polyclonal autoantibody to GABA (1:1500, Sigma) detected with silver-enhanced subnanometer (Ultra Small) gold-conjugated secondary antibody (1:80, Aurion). Briefly, ultrathin sections were subjected to etching in periodic acid and sodium-metaperiodate ( Watson and Bazzaz, 2001 ; Ranson et al. , 2006 ) for 30 s each. Rinsing in sterile PBS was followed by blocking of unspecific binding, primary antibody incubation (overnight at 4°C), secondary antibody incubation (2 h at room temperature), post-fixation in 2% glutaraldehyde and silver enhancement with R-Gent-SE-EM according to the manufacturer’s recommendations (Aurion). In preparatory experiments, a silver enhancement time of 45 min was found to yield recognizable individual silver-gold particles while preventing particle fusion. After enhancement, sections were contrasted with uranyl acetate (2% in 70% ethanol) and Reynold’s lead citrate ( Reynolds, 1963 ). Each GABA-immunostaining procedure included a negative control omitting the primary antibody.
Electron microscopy imaging
Electron micrographs were captured on a LEO 912 AB electron microscope or on a Leo 906 E electron microscope (Zeiss) with a ProScan Slow Scan CCD (ProScan) by a blinded investigator. Image acquisition and storage was performed with corresponding software iTEM (Olympus Soft Imaging Solutions).
For quantitative analyses, multiple square grids of 100 µm 2 were randomly placed on a photograph taken next to a ventral horn motor neuron at the spinal cord L5 level. These squares were screened for presynaptic boutons identified by the presence of presynaptic vesicles of 40–50 nm size and/or pre- and postsynaptic membrane specializations of the boutons and adjacent postsynaptic profiles. Silver–gold particles and presynaptic structures were counted by an experimenter blinded to the experimental conditions with MacBiophotonics ImageJ (Wayne Rasband, www.macbiophotonics.ca ). Densities of silver-gold particles (representing GABA-immunolabelling intensity) and of vesicular structures were determined by dividing the respective counts by the area of the bouton (minus the area of mitochondria). For each GABA-immunolabelled section, bouton particle density was corrected for minor background labelling determined over non-synaptic areas (perikarya of motor neurons) of the same section. For analysis, particle density was related to quantitative GABA immunoreactivity. In addition, boutons were ranked according to particle density and the resulting histograms were divided into thirds. The upper third (threshold ≥ rank 165) characterized ‘GABA+’ boutons with high density of GABA-immunogold labelling and the lower third (threshold ≤ rank 82) ‘GABA−’ boutons with low immunolabelling intensity.
Primary neuronal cell culture
Reagents were obtained from Life Technologies if not stated otherwise. Cells were prepared from hippocampi of embryonic Day 18 embryos of pregnant C57Bl/6 mice (Harlan-Winkelmann). Hippocampi were dissected and separated from meninges and surrounding tissue before enzymatic digestion with 0.25% w/v trypsin EDTA for 5 min at 37°C. Following enzymatic treatment, trypsin was inactivated by two generous flushes of Hank’s Balanced Salt Solution (HBSS) supplemented with penicillin/streptomycin and 10 mM HEPES. Hippocampi were subsequently triturated in Neurobasal® medium supplemented with glutamine (1%), B27® (2%) and penicillin/streptomycin (1%) with a narrowed glass pipette for further mechanical separation. Cells were counted with a haemocytometer (Hartenstein) and plated at a density of 50 000 cells on 18-mm diameter coverslips (Langenbrinck). Primary neurons were used for the experiments at in vitro Day 14.
In a separate set of experiments we tested specificity of specAmph autoantibody-induced pathomechanisms in amphiphysin-deficient neurons from knockout mice. Amphiphysin knockout mice were bred and genotyped as described ( Geis et al. , 2010 ). Hippocampal cell cultures were prepared from embryonic Day 18 embryos of heterozygous breeding pairs. Genotyping was performed for each embryo and amphiphysin protein deficiency was confirmed in neuronal cultures at in vitro Day 14 using immunohistochemistry with a commercial antibody to amphiphysin (Acris, 13379-1-AP) and revealed complete loss of immunoreactivity in knockout cultures. All experiments were performed in triplicates.
Autoantibody treatment and stimulation of primary neurons
Primary hippocampal neurons were incubated with affinity purified specific patient autoantibodies targeting the SH3 domain of amphiphysin (specAmph) and control IgG for 6 h at 37°C (100 µg/ml). The experimenter was blinded as to treatment conditions (IgG application and stimulation protocols). Stimulation of cells was performed with a customized stimulation chamber RC-49FS (Warner Instruments). Platinum electrodes with a spacing of ∼10 mm delivered electric fields of ∼10 V/cm at a frequency of 10 Hz for 90 s (pulse duration = 1 ms). A Grass stimulator S88 (Grass Technologies) was used to deliver voltage at the stated parameters. Cells were stimulated in artificial CSF containing 119 mM NaCl, 2.5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 25 mM HEPES and 30 mM glucose.
For brefeldin A, treatment cells were first incubated with 100 µg/ml of specAmph for 6 h at 37°C and then brefeldin A (Sigma Aldrich, B5936-200UL) was added for the last 2 h of incubation time at a concentration of 10 µg/ml as described previously ( Shetty et al. , 2013 ). The same volume of the solvent dimethyl sulphoxide (DMSO) served as control. After total incubation time of 6 h cells were processed for stimulation, immunostaining and imaging as described. All experiments were performed in triplicate.
Immediately after stimulation neurons were fixed with ice-cold 4% paraformaldehyde (PFA) for 20 min and permeabilized with 0.1% Triton™ X-100 for 30 min at room temperature. Primary antibodies were incubated for 1 h at room temperature in PBS containing 10% normal bovine and 10% normal goat serum (blocking solution), coverslips were washed six times for 10 min and subsequently incubated with secondary antibodies in blocking solution overnight at 4°C. Incubation was followed by washing steps as described above. Samples were kept in PBS until imaging in 100 mM mercaptoethylamine (MEA; pH 7.9) buffer.
Primary antibodies against synaptobrevin 2 (1:2000, #104211), synaptobrevin 7 (1:2000, #232011), endophilin (1:1000; #159002), synaptojanin (1:500, #145003), and vgat (1:1000, #131004) were obtained from Synaptic Systems. Secondary antibodies were from Life Technologies (Alexa Fluor® 647, goat anti-mouse, 1:500, A21237; Alexa Fluor® 647, goat anti-rabbit, 1:500, A21246) and Dianova (Cy3, goat anti-guinea pig, 1:500, #106-165-003).
Coverslips of immunostained neurons were mounted tightly on a custom built imaging chamber and placed on a customized Olympus IX71 inverted microscope. Samples were imaged in phosphate buffer containing 100 mM MEA adjusted to pH 7.9. Fluorophores were excited with an Ibeam Smart 640 s Laser (Toptica Photonics) at 640 nm and a Nano laser (Quioptiq photonics) at 532 nm while keeping laser powers [75 mW (640 nm) and 1 mW (532 nm)] constant during the experiment. Photons were collected using two EMCCD (Andor Ixon Ultra, BFI Optilas) cameras keeping detector gain and frame rate constant throughout the experiment. Resulting images were processed in rapidSTORM software ( Wolter et al. , 2012 ) and saved as density matrices with z -dimensions representing localization counts. Matrices were rearranged by a custom written Python script kindly provided by Thorge Holm and fed into ImageJ as text image (Wayne Rasband, www.macbiophotonics.ca ) for further analysis. For signal quantity calculations a region of interest was created in ImageJ by applying a constant threshold of 45 bits (of maximum 256 bits on the colour scale) on the epifluorescent signal of GABAergic vesicular transporter (vgat). It has been shown previously that vgat serves as a reliable normalization factor without relevant fluctuations ( Werner et al. , 2015 ). Variations of vgat thresholds were tested in subsets of presynaptic boutons resulting in confined region of interest leading to inaccuracy and missing signals in d STORM localization count quantification ( Supplementary Fig. 2 ). A binary mask was created and converted to regions of interest for quantification of integrated density in ImageJ. The reliability of density analysis in ImageJ was additionally checked by a parallel computation of raw values in Microsoft Excel. Localization count analysis (represented as signal quantity) is regarded as a direct measure of immunofluorescent molecules located in a distinct pre-defined area. It is not intended for counting molecules but serves as a proportional quantification of signals arising from fluorescent molecules.
Cluster analyses were performed by calculating the distances between each maximum residing in every singular vgat region of interest. Distances were calculated by a custom written script in Sigmaplot (Systat). Thus, shorter distances represent enhanced clustering and longer distances dispersed signal. Maxima were defined by ImageJ (FIJI distribution package) embedded ‘find maxima’ operation using a noise tolerance level of 2. Density of maxima within a presynaptic bouton, and density of the endophilin signal within a maximum were calculated by ImageJ using constant parameters for detection. Maxima within a presynaptic bouton are regarded as spots of several endophilin proteins and are used to evaluate distribution (signal distance) and the amount (maxima quantity) of these spots within a presynaptic compartment. Signal composition of endophilin within a single maximum is measured by localization counts and is regarded as a proportional measure of endophilin proteins located in this specific spot. Localization counts residing in endophilin maxima were calculated using the area from detected maxima defined as region of interest and applying the ‘analyze particles’ function in ImageJ.
For electron microscopy data, calculations were performed in Excel, d STORM data were processed as described above. Statistical analysis was done in Sigmaplot 12. The non-parametric Mann-Whitney U-test was applied for comparing individual groups and one-way ANOVA with Tukey’s post hoc test was used for statistical comparison of multiple groups.
Sustained stimulation increases synaptic vesicle pool size in GABAergic spinal cord interneurons in vivo
First we examined the consequences of long-term high-frequency unilateral stimulation (10 Hz for 60 s) of sciatic nerve Ia afferents on the presynaptic vesicle pool of naïve spinal cord interneurons. Besides monosynaptic excitation of motor neurons, this stimulation paradigm leads to a heterosynaptic activation of local GABAergic interneurons mediating presynaptic inhibition [ Fig. 2 A(i)]. We analysed structural changes in the presynaptic vesicle pool of boutons in the neuropil surrounding ventral horn motor neurons at the lumbar spinal cord (L4–5 level) by electron microscopy. Spinal cord presynaptic nerve endings are not restricted to a single type of neurotransmitter and may therefore contain both excitatory and inhibitory neurotransmitters ( Somogyi, 2002 ). Boutons were identified as primarily GABAergic if they exceeded a preset upper threshold of GABA immunoreactivity as measured by post-embedding immunogold staining, whereas low GABA containing synapses had GABA immunoreactivity below a predefined lower threshold [ Fig. 2 A(ii) and ‘Materials and methods’ section]. These subgroups of boutons were compared in addition to analyses showing the dependence of vesicular structures on a continuous scale of GABA immunoreaction intensity in the respective boutons ( Supplementary Fig. 3 ). Further, we quantified all boutons regardless of transmitter specificity. All investigated boutons were located in the vicinity of ventral horn motor neurons. Vesicle pool size, appearance of clathrin-coated vesicles and endosome-like structures, endocytic intermediates as recognized from studies on ultrafast endocytosis ( Kittelmann et al. , 2013 ; Watanabe et al. , 2013 ) were then compared in presynaptic boutons in the two respective spinal cord hemi-segments at lumbar levels (L4–5; stimulated versus unstimulated). In control animals receiving non-reactive control patient IgG intrathecally, sustained stimulation profoundly increased the size of presynaptic vesicle pools ( Fig. 2 B). This effect was mediated primarily by GABAergic synapses whereas vesicle density was unchanged in synapses with low GABA content ( Fig. 2 B and Supplementary Fig. 3A ). At resting conditions, only few endocytic intermediates (clathrin-coated vesicles and endosome-like structures) were observed. After prolonged stimulation, we found an increase of these intermediates in all boutons without preference for one of the subclasses analysed ( Figs 2 C, 4A and Supplementary Fig. 3B ). Collectively, the findings yield ultrastructural evidence that spinal cord presynapses keep up with long-duration high-frequency stimulation by activity-induced formation of a large bulk of presynaptic vesicles including endocytic intermediates.
Pathogenic anti-amphiphysin autoantibodies induce activity-dependent presynaptic vesicle depletion
In contrast to control conditions, the vesicle density in spinal presynaptic boutons was markedly reduced upon sustained stimulation in rats that had been chronically injected intrathecally with specific affinity-purified pathogenic human autoantibodies to amphiphysin (specAmph) and that showed characteristic disease signs of stiff-person syndrome ( Geis et al. , 2010 ) ( Fig. 3 A). This reduction at stimulated synapses was seen in primarily GABAergic and non-GABAergic boutons. We found a higher incidence of presynaptic boutons with larger vesicle content in unstimulated compared to stimulated synapses ( Fig. 3 A). We then analysed vesicle density dependent on GABA immunolabelling intensity in these presynaptic boutons and found a reduction of vesicle density in boutons with low and high GABA density. However, in presynapses containing highest GABA density, vesicles were almost completely depleted ( Supplementary Fig. 3A ). Notably, the reduction of endocytic intermediates after stimulation was even more pronounced. Here, frequency distribution showed an almost complete depletion of clathrin-coated vesicles in boutons of stimulated synapses ( Fig. 3 B) and a reduction of endosome-like structures ( Fig. 4 B). Similar to the reduction of presynaptic vesicles, clathrin-coated vesicle depletion was apparent in boutons with low and high GABA density. Again, presynapses with the highest GABA density showed no remaining clathrin-coated vesicles after stimulation and preincubation with specAmph autoantibodies ( Supplementary Fig. 3B ). Thus, sustained high-frequency stimulation had an opposite effect on vesicle pool size and on endocytic intermediates in spinal boutons after application of pathogenic specAmph autoantibodies. Moreover, at resting conditions with only basal synaptic activity, presynaptic boutons of rats showed an increased density of vesicles (54.0 ± 3.8 µm −2 ) and clathrin-coated vesicles (6.1 ± 0.9 µm −2 ) after intrathecal application of specAmph autoantibodies as compared to control conditions (32.5 ± 3.6 µm −2 ; P < 0.01 and 3.2 ± 0.6 µm −2 ; P < 0.05, respectively). Increased vesicle density may reflect compensatory vesicle recruitment including endocytic intermediates at basal activity levels, which is then nearly completely decompensated during sustained stimulation.
Altered v-SNARE composition of GABAergic vesicle pools induced by human anti-amphiphysin autoantibodies
Clathrin-mediated endocytosis depends on adaptor protein complex AP-2 (encoded by AP2A1 to AP2M1) ( Kim and Ryan, 2009 ) and blockade of this endocytosis pathway involving the membrane interaction function of amphiphysin may lead to activation of compensatory pathways, e.g. clathrin-independent vesicle formation via AP-3 (encoded by AP3B1 to AP3S2), favouring production of different vesicle identities ( Hua et al. , 2011 ; Shetty et al. , 2013 ). The endocytosis defect characterized by our electron microscopy findings led to the hypothesis that following exposure to anti-amphiphysin autoantibodies, presynaptic vesicle pools may consist of vesicle subtypes differing in their composition of v-SNAREs from those under physiological conditions.
To test this hypothesis we used d STORM in primary neuronal cell cultures since this approach offers low complexity in synaptic connections and low background for quantifying fluorescent signals. Neurons were preincubated with purified IgG fractions and field stimulation paradigms were adapted to in vivo experiments. To determine vesicle identity and to quantify the size of the respective vesicle pools, we analysed the quantity of v-SNAREs syb2 and syb7 signals that are known to be directed to readily releasable and to resting pool vesicles, respectively ( Hua et al. , 2011 ). We aimed at maximizing the accuracy of quantification by recording syb2 and syb7 signals as the sum of localization counts using a defined and constant amount of primary and secondary antibodies (for details see ‘Materials and methods’ section). To focus the analyses on GABAergic terminals, we defined the area of GABAergic presynaptic boutons by a fixed threshold on fluorescence signal of the presynaptic vesicular GABA transporter (vgat). This was tested for accuracy and then applied for all analysed boutons ( Fig. 5 A and Supplementary Fig. 2 ). Under control conditions, quantification of syb2 signal revealed nearly identical levels in GABAergic presynapses at high stimulation paradigms and basal activity ( Fig. 5 B). In contrast, sustained stimulation of GABAergic synapses pretreated with pathogenic specAmph autoantibodies led to an increase of syb2 quantity per unit of vgat positive area ( Fig. 5 B). This increase of quantitative syb2 signals was also significantly different from stimulated control presynapses by group comparisons. Moreover, this enhanced syb2 signal mainly arose from the border areas of presynaptic boutons located near the plasma membrane, since analysis of the central area of presynaptic boutons revealed lower signals of syb2 that were similar to control groups and to unstimulated conditions ( Supplementary Fig. 2 ).
Next, we tested whether syb7, a v-SNARE protein typically present in vesicles of the presynaptic reserve pool, is deregulated by defective endocytosis. Analysis showed low overall syb7 signal densities, which is in line with reports of lower amounts of syb7 copies on synaptic vesicles in comparison to the syb2 isoform ( Takamori et al. , 2006 ). Quantitative analysis revealed that sustained stimulation does not significantly change syb7 signals in GABAergic presynapses under control conditions. However, after pretreatment with specAmph autoantibodies, syb7 quantity was markedly reduced upon sustained stimulation ( Fig. 5 C).
Specificity of autoantibody-induced changes in syb2 and syb7 synaptic regulation was tested with neurons deficient for amphiphysin. In amphiphysin knockout neurons preincubated with control IgG we found a slight but not significant upregulation of syb2 after sustained stimulation and no change of syb7 quantity ( Supplementary Fig. 4 ). This change after stimulation was similar to that seen in wild-type neurons after incubation with specAmph ( Fig. 5 B and C), thus indicating analogous pathophysiological processes in the amphiphysin knockout situation per se and in wild-type neurons after acute specAmph autoantibody incubation. Importantly, in amphiphysin knockout neurons preincubation with specAmph autoantibodies did not further increase syb2 and has no impact on syb7 quantity as compared to control IgG preincubation ( Supplementary Fig. 4 ), thus confirming that autoantibody-induced dysregulation in wild-type neurons is indeed mediated by autoantibodies to amphiphysin.
To further test the hypothesis of an autoantibody-induced switch to an AP-3 dependent pathway, we used brefeldin A, a fungal metabolite and inhibitor of protein transport from the endoplasmic reticulum that is known to block the AP-3 dependent pathway ( Faundez et al. , 1998 ; Voglmaier et al. , 2006 ). Conversely to neurons only preincubated with specAmph autoantibodies, additional pretreatment with brefeldin A not only reversed the stimulation-induced effect of syb2 increase but also induced even reduced amounts of syb2 signal ( Fig. 6 A). Concordantly, the syb7 signal was increased in GABAergic boutons upon brefeldin A pretreatment in addition to specAmph incubation ( Fig. 6 B), indicating that the switch to the AP-3 dependent pathway may be blocked and AP-2 mediated vesicle formation increasingly reactivated.
Hence, endocytic dysfunction induced by anti-amphiphysin autoantibodies may influence synaptic vesicle pool dynamics and v-SNARE composition of GABAergic vesicle pools. The altered vesicle pool properties may underlie the high-frequency, stimulus-dependent defect of GABAergic transmission in the animal model of stiff-person syndrome reported previously ( Geis et al. , 2010 ) and in characteristic clinical findings with motor hyperexcitability in patients with stiff-person syndrome ( Meinck et al. , 2001 ; Dalakas, 2009 ).
Human autoantibodies to amphiphysin alter endophilin density and clustering in GABAergic presynaptic vesicle pools
Repetitive intrathecal application of specAmph autoantibodies in the animal model led to enhanced density of clathrin-coated vesicles at basal synaptic activity. This finding led us to investigate the direct amphiphysin interaction partners endophilin and synaptojanin. Unlike dynamin, these proteins are not directly involved in membrane fission and the respective knockout mice show similar synaptic pathology as observed here in our passive-transfer model ( Geis et al. , 2010 ; Milosevic et al. , 2011 ).
We first quantified endophilin signals over GABAergic vesicle pools in primary neurons. Under control conditions, sustained stimulation led to a decrease of total endophilin signal in individual GABAergic presynaptic boutons as revealed by quantification of the localization counts ( Fig. 7 A). In addition to alterations of the amount of endophilin protein, changes in its localization may lead to disturbed endocytic function or may account for compensatory mechanisms. Endophilin signal distance of single maxima within an individual presynaptic bouton and localization counts within these signal maxima were unchanged under control conditions at both activity stages ( Fig. 7 B and D). However, sustained stimulation led to a significant reduction of individual maxima in single GABAergic presynaptic boutons ( Fig. 7 C), consistent to the reduction of total endophilin signal as revealed by quantification of localization counts ( Fig. 7 A). In neurons pretreated ex vivo with specAmph autoantibodies, the overall quantity of endophilin signal was not reduced throughout the individual GABAergic boutons upon stimulation ( Fig. 7 A). However, a significant reduction of endophilin localization count was evident in single maxima within these boutons ( Fig. 7 D). Accordingly, in comparison to stimulated synapses under control conditions, the dispersion of endophilin signal was increased upon pretreatment with specAmph autoantibodies ( Fig. 7 B). Individual endophilin clusters were reduced in unstimulated synapses after pretreatment with specAmph autoantibodies to a level similar to that of stimulated control synapses ( Fig. 7 C). Against these findings regarding quantity and distribution of endophilin in GABAergic presynapses, analyses of the uncoating factor synaptojanin showed no changes in signal quantity or synaptic localization ( Supplementary Fig. 5 ).
These findings indicate that endophilin localization is stimulus-dependent and more widespread with decreased clustering in GABAergic presynapses due to dysfunctional clathrin-mediated endocytosis on treatment with anti-amphiphysin autoantibodies.
Stiff-person syndrome is an enigmatic autoimmune disease with key symptoms of motor hyperexcitability and increased anxiety. Previous experimental studies on paraneoplastic stiff-person syndrome with autoantibodies to amphiphysin have provided functional evidence for reduced GABAergic transmission ( Sommer et al. , 2005 ; Geis et al. , 2010 , 2012 ). Synaptic neurotransmission depends on reliable replacement of synaptic vesicles by clathrin-mediated endocytosis or budding of new vesicles from bulk endosomes. This is very important in GABAergic inhibitory synapses as these often serve as tonic synapses with a high turnover rate of presynaptic vesicles and therefore need for proper endocytic machinery ( Ferguson et al. , 2007 ). Studies on clathrin-mediated endocytosis have so far focused on genetic knockout of clathrin-mediated endocytosis proteins ( Milosevic et al. , 2011 ; Raimondi et al. , 2011 ; Soda et al. , 2012 ), on acute interference by inhibitory peptides, or on the effects of small molecules blocking clathrin-mediated endocytosis proteins ( Shupliakov et al. , 1997 ; Macia et al. , 2006 ; von Kleist et al. , 2011 ). Here we show that depending on synaptic activity, specific human pathogenic autoantibodies to amphiphysin may have striking effects upon morphological and molecular architecture of presynaptic vesicle pools. These autoantibody-mediated effects are best explained by dysfunctional clathrin-mediated endocytosis and may severely afflict the machinery for protein sorting and efficient preparation of vesicles for successive transmitter exocytosis. This consequently may impede fast replenishment of synaptic vesicles in presynaptic boutons, which is indispensable during sustained activity. We document that presynaptic vesicle pools and clathrin-coated vesicles are much reduced upon stimulation in vivo in spinal boutons. Together, these structural abnormalities support the proposition that application of specific human anti-amphiphysin autoantibodies in vivo induces slower endocytosis rates and faster synaptic exhaustion during high frequency firing of inhibitory neurons as reported previously ( Geis et al. , 2010 ). Thus, these findings may represent the ultrastructural correlate of characteristic disease symptoms in patients with stiff-person syndrome consisting of reduced spinal inhibitory regulation ( Sandbrink et al. , 2000 ; Wessig et al. , 2003 ).
Super-resolution d STORM of GABAergic presynaptic vesicle pools in primary neurons revealed that specAmph autoantibodies induced pathomechanisms include changes in the equipment of synaptic vesicles with v-SNARE isoforms as essential components of vesicle exocytosis. d STORM offers precise antigen localization and accurate determination of signal components within small neuronal structures as shown here in presynaptic compartments. Recent reports could demonstrate that d STORM is useful for characterizing synaptic dysfunction ( Andreska et al. , 2014 ; Esbjorner et al. , 2014 ) and for investigating synaptic organization including quantification of synaptic proteins ( Dani et al. , 2010 ; Ehmann et al. , 2014 ). In the present study we used the high localization precision of d STORM to quantify protein distributions and to determine cluster properties. The protein composition of synaptic vesicles is generally believed to be different depending on which endocytic route is prevalent at a particular synapse ( Voglmaier and Edwards, 2007 ). v-SNAREs were previously shown to characterize different vesicle pools. Syb2 is predominantly associated with the readily releasable pool whereas syb7 is mainly present on vesicles of the resting pool ( Hua et al. , 2011 ). We found that syb7 was strongly reduced in synapses upon exposure to specAmph autoantibodies during sustained stimulation. This is in line with our ex vivo observations of activity-dependent depletion of vesicles and clathrin-coated vesicles in spinal presynapses after specAmph autoantibody passive transfer. Besides changing location on presynaptic vesicles, v-SNAREs may also be trapped at the synaptic plasma membrane after fusion, which is caused by a slowed or inhibited endocytic machinery, as observed in a recent report investigating endocytic adaptor proteins ( Shetty et al. , 2013 ). Differential distribution with preferential location at presynaptic bouton border areas and increase of syb2 quantity may also result from its sorting by AP-180 (encoded by SNAP91 ; Koo et al. , 2011 ) for which we have previously shown a more intense clustering in the presence of specAmph autoantibodies ( Geis et al. , 2010 ). Taken together, disturbed clathrin-mediated endocytosis may lead to maldistribution of essential v-SNAREs and this may challenge sustained exocytosis at high activity levels.
In this dysfunctional synaptic condition we provide evidence that the direct amphiphysin interaction partner endophilin ( Micheva et al. , 1997 ) is differentially regulated within GABAergic vesicle pools and is abnormally localized. Similar to our present ultrastructural observations in spinal cord presynaptic boutons at basal activity levels after intrathecal specAmph autoantibody passive transfer, triple knockout of all endophilin isoforms was reported to lead to accumulations of clathrin-coated vesicles in resting conditions ( Milosevic et al. , 2011 ). Endophilin seems especially important for high frequency neurotransmission ( Llobet et al. , 2011 ). It is not yet clear if the more widespread distribution of endophilin in the presynapse during sustained stimulation and exposure to pathogenic specAmph autoantibodies as shown here is due to general reorganization of endocytic proteins or if it represents an early compensatory mechanism initiated by anti-amphiphysin autoantibody-induced clathrin-mediated endocytosis dysfunction. Similarly, the synaptic density of amphiphysin protein levels has been reported to be increased in endophilin triple knockout mice ( Milosevic et al. , 2011 ).
In conclusion, our observations provide insights into molecular events of synaptic dysfunction underlying the effects of anti-amphiphysin autoantibodies in human stiff-person syndrome. This finding may serve as a proof-of-principle example for future research into the molecular pathophysiology of synaptopathies of autoimmune ( Lai et al. , 2009 ; Gleichman et al. , 2012 ) and neurodegenerative aetiology ( Trempe et al. , 2009 ; De Jesus-Cortes et al. , 2012 ) using electron microscopy and super-resolution microscopy in combination.
We thank B. Broll, B. Dekant, S. Hellmig, K. Reinfurt-Gehm, S. Schenk and C. Sommer (Jena) for providing expert technical assistance in animal experiments, immunohistology, IgG and electron microscopy preparations. The authors declare no competing financial interest.
This work was supported by the Deutsche Forschungsgemeinschaft (SFB 581 [TP A7], SFB/TR 166 [TP B2], GE2519_3-1), by the IZKF and CSCC Jena (E-3.3), and by intramural University Research Funds (Würzburg and Jena).
Supplementary material is available at Brain online.
- d STORM
direct stochastic optical reconstruction microscopy
specific anti-amphiphysin antibody eluates
GABAergic vesicular transporter
vesicular soluble N -ethylmaleimide-sensitive-factor attachment receptor