Serum from five patients (Pat1-5) with anti-GlyR autoantibodies was used for the experiments. Clinical data of two patients (Pat1 with SPS, Pat5 with PERM) were described previously (Rauschenberger et al., 2020). Serum from five healthy individuals served as negative control (HC), a serum from a patient with multiple sclerosis was used as disease control (DC).

Patients 2 and 3 did not have a history of cancer, SPS or PERM but another neurological disorder. Their serum and CSF samples did not exhibit antibodies directed against other neuronal surface antigens (NMDAR, AMPAR, LGI1, CASPR2, GABA_AR, GABA_BR) but autoantibodies against the GlyR assessed by appropriate cell-based assays (Ekizoglu et al., 2014). Sera of patients 2 and 3 were included in the present study to examine general molecular mechanisms for autoantibody binding and detection.

Patient 2 was a 33-year old woman with a 16-year history of focal epilepsy of unknown cause, characterized with focal temporal seizures evolving to bilateral convulsive seizures and loss of consciousness. Her CSF and MRI examinations were normal. She favorably responded to antiepileptic medications (oxcarbazepine and topiramate) and thus immunotherapy was not considered.

Patient 3 was a 14-year old girl with a 6-year history of unclassified epileptic encephalopathy characterized with generalized tonic–clonic seizures and atypical absences. She showed mildly progressive cognitive decline with normal metabolic screening, normal CSF findings and mild cerebral atrophy on MRI. EEG showed generalized discharges and focal spikes over the right frontotemporal region. She favorable responded to IVIG treatment and anti-epileptic medications (carbamazepine and valproate).

Patient 4, a 71-year old male at blood withdrawal, developed symptoms of SPS (PERM) at age 65. He suffered from stiffness and spasms of the right arm, falls with bone fractures and a pronounced startle reaction, and was finally wheelchair bound. Extensive search for a malignoma and for other autoantibodies than GlyR antibodies was negative. Under i.v. steroid pulse therapy every 4 months and oral clonazepam at a dose of 0.25 mg in the morning and 0.375 mg in the evening he recovered the ability to walk, and his condition remained mostly stable.

Site-directed mutagenesis was used to create the non-glycosylation mutant GlyRα1^N38Q. The human GlyRα1 cDNA in pRK5 vector (gift †P. Seeburg) was used as template. Sequence correctness was verified (Eurofins Genomics Germany GmbH, Ebersberg, Germany). For transfection with GFP, the eGFP-N1 plasmid (Takara Bio Europe, Saint-Germain-en-Laye, France) was used.

HEK293 cells (human embryonic kidney, ATCC^® CRL-1573™, Wesel, Germany) were cultured in Minimum Essential Media (MEM, Life Technologies, Darmstadt, Germany) supplemented with 10% fetal calf serum (FCS), glutaMAX (200 mM), sodium pyruvate (100 mM), penicillin (10,000 U/ml), and streptomycin (10,000 μg/ml). The cells were incubated at 37°C and 5% CO₂.

HEK293 cells were transfected using a calcium phosphate precipitation method. Plasmid DNA (1 μg/μl; 1 μg of each plasmid was used in cotransfections of GFP and GlyRα1 variants or the empty pRK5 vector (MOCK control) for 3 cm dishes containing four coverslips and 180,000 cells seeded for immunocytochemical experiments and electrophysiological recordings; transfections in 10 cm dishes containing 1.8 × 10⁶ HEK293 cells for biotinylation experiments used 10 μg of plasmid per dish), 2.5 M CaCl₂, 0.1× TE buffer (Tris/EDTA) and 2× HBS (50 mM HEPES, 12 mM glucose, 10 mM KCl, 280 mM NaCl, 1.5 mM Na₂HPO₄) were incubated for 20 min at 20°C. The mixture was applied to the cells. Cells were washed after 5–6 h and used for experiments after 24–72 h.

Pregnant CD1 mice were euthanized with CO₂ and cervical dislocation. Spinal cords were extracted from embryonic day 12.5 (E12.5) embryos. Spinal cord tissue fragments were triturated three times. Supernatants with cells were centrifuged at 400 rpm for 20 min. Cells were counted and plated on poly-D-lysine-covered coverslips using a density of 150,000 cells/3 cm dish. Neurons were grown in Neurobasal medium (Life Technologies, Darmstadt, Germany) containing 5 ml of L-glutamine (200 mM) and B27 supplement (Life Technologies, Darmstadt, Germany) with an exchange of 50% medium after 7 days in culture. Neurons were used for experiments at day in vitro 14 (DIV14).

The crystal structure of human glycine receptor α3 homopentamer in complex with strychnine (PDB: 5CFB) (Huang et al., 2015) was used as template to introduce the mutation and to analyse possible structural rearrangements or interaction changes in the surrounding region of the mutation. In silico site-directed mutagenesis (N38Q) was carried out in Coot taking basic geometry of the mutated residue and clashes of the surrounding residues into account (Emsley and Cowtan, 2004). The figures were prepared by using PyMOL¹.

Cell lysates of transfected HEK293 cells were prepared using the CytoBuster Protein Extraction Reagent (Merck Millipore, Billerica, MA, United States). To verify GlyR N-glycosylation, lysates of GlyRα1^WT and GlyRα1^N38Q were incubated with peptide N-glycosidase F (PNGase F) or endoglycosidase H (EndoH) according to manufacturer’s instructions (P0704S and P0702S, New England BioLabs, Ipswich, MA, United States).

Biotinylation experiments were performed as described previously (Atak et al., 2015).

40 μg protein/lane were loaded on a 11% polyacrylamide gel. Proteins were transferred to nitrocellulose membrane (GE HealthCare, Freiburg, Germany). Membranes were washed in TBST (TBS with 1% Tween 20), blocked for 1 h in 3% bovine serum albumin (BSA, Carl Roth, Karlsruhe, Germany) and incubated in MAb4a (1:500; 146,011, RRID:AB_887722, Synaptic Systems, Göttingen, Germany), anti-gephyrin (1:1,000; 147,111, RRID:AB_887719, Synaptic Systems, Göttingen, Germany), anti-GFP (1:5,000; SC8334, RRID:AB_641123, Santa Cruz Biotechnology, Dallas, TX, United States) or anti-pan-cadherin antibody (1:1000; 4,068, RRID:AB_2158565, Cell Signaling, Danvers, MA, United States) overnight at 4°C. As secondary antibodies horse radish peroxidase (HRP) conjugated goat anti-mouse (1:15,000, 115–035-146, RRID:AB_2307392, Dianova, Hamburg, Germany) or goat anti-rabbit secondary antibodies (1:15,000, 111–036-003, RRID:AB_2337942, Dianova, Hamburg, Germany) were used. Signal detection was done using the SuperSignal West (Thermo Fisher Scientific, Waltham, MA, United States). Western Blot Images were taken using the Chemostar Touch Imager (Intas Science Imaging Instruments, Göttingen, Germany).

Electrophysiological recordings using the patch-clamp method in a whole-cell configuration were performed. Transfected HEK293 cells were used for recordings at room temperature 21°C. Used borosilicate capillaries had open resistances of 3.5–5.5 MΩ and were filled with internal buffer (120 mM CsCl, 20 mM N(Et)₄Cl, 1 mM CaCl₂, 2 mM MgCl₂, 11 mM EGTA, 10 mM HEPES; pH 7.2, adjusted with CsOH). The external buffer contained (137 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 5 mM HEPES; pH 7.35, adjusted with NaOH). For determination of maximal current amplitudes (I_max) and EC₅₀ values, glycine was used in a concentration series of 10 μM, 30 μM, 60 μM, 100 μM, 300 μM, 600 μM, 1 mM glycine The agonist was applied using the Octaflow II system (ALA Scientific Instruments, Farmingdale, NY, United States). Current responses were amplified with an EPC-10 amplifier (HEKA, Elektronik GmbH, Lambrecht/Pfalz, Germany) and measured at a holding potential of −60 mV using Patchmaster Next software (HEKA Elektronik GmbH, Lambrecht/Pfalz, Germany). The analysis included cells with input resistances above 100 MΩ-1 GΩ, leak currents smaller than −300 pA and capacitances of 11–13 pF. Recorded maximal currents were plotted and fitted using the hill 1 function in Origin 2019 (Originlab Corporation, Northampton, United States) to examine EC₅₀ values.

Lysis, preparation of the inclusion bodies (IB), refolding and ECD purification were performed according to the established protocol by Breitinger et al. with some minor changes (Breitinger et al., 2004). In brief, GlyR ECDs (GlyRα1 and GlyRα2; both in the vector pET30) were expressed in Escherichia coli as IBs. After lysis, the lysates were centrifugated (20,000 g, 4°C, 30 min). The pellets, including the IBs, were washed twice with Tris/HCl-buffer, pH 7.4, 0.5% Triton-X-100. A third washing step was done with a Tris/HCl buffer containing 1 M urea instead of Triton-X-100. The IB-pellet was dissolved in 8 M urea in 100 mM Na₂PO₄, 10 mM Tris/HCl pH 8. Preparative refolding was done by dialyzing 10 ml (1 mg/ml) IB-solution against a volume of 5 l buffer with decreasing concentrations of urea, starting from 5 M urea in 50 mM Tris/HCl pH 8 in nine subsequent steps, each lasting at least 4 h at 4°C. The last dialysis buffer contained 5 mM Na-phosphate, pH 7.4 and 150 mM NaCl. To get rid of aggregated and non-refolded ECDs, the preparation was ultracentrifugated for 40 min at 100,000 g, 4°C. To exclude aggregated proteins further, a size exclusion chromatography using a Superose column was performed. The purity and size of the refolded GlyRαl or α2 ECDs was verified on 11% PAA gels and stained using Coomassie solution (0.1% Coomassie brilliant blue R250, 25% methanol, 7.5% acetic acid). De-staining was performed until protein bands became visible (40% Methanol, 10% acetic acid). GlyRα2 ECD refolding was proven by Circular-dichroism (CD) spectroscopy.

CD spectroscopy was conducted with a Jasco J-810 spectropolarimeter. Far-UV spectra from 190 to 270 nm were recorded at a scanning speed of 50 nm/min with a response time of 1 s and a bandwidth of 1.5 nm. Spectra were baseline corrected by subtracting buffer runs. Ten individual scans were taken and averaged. The buffer was exchanged to 50 mM N-phosphate buffer, pH 8 with ultrafiltration units (Sartorius Vivaspin 500, Göttingen).

ELISA plates (Sarstedt, Nümbrecht, Germany) were coated overnight with 2 μg/ml purified and refolded GlyRαl or α2 ECD. ELISA plates were washed 5x with H₂O_dest. and blocked with 10% BSA diluted in PBS containing 0.05% Tween 20 for 1 h at 37°C. Wells were incubated with patient serum (1:100 in blocking solution) or GlyR pan-α antibody MAb4a (1:500) for 1 h at 37°C. The supernatants were transferred three times into fresh coated wells subsequent to incubation with GlyRα1/α2 and GFP co-transfected HEK293 cells or primary spinal cord neurons. The ELISA plates were further processed with washing steps in PBS and secondary antibody goat anti-mouse (1:20,000 in blocking, 115–035-146, RRID:AB_2307392, Dianova, Hamburg, Germany) or goat anti-human HRP (1:20,000 in blocking, 109–035-088, RRID:AB_2337584, Dianova, Hamburg, Germany) incubation for 1 h at 37°C. TMB solution (00–4,201–56, Thermo Fisher Scientific, Waltham, MA, United States) was added and reaction was stopped after 15 min with 1 M H₃PO₄. Absorbance was read with a Wallac 1,420 Victor2 Microplate Reader (Perkin Elmer, Waltham, MA, United States) at 450 nm. As controls, uncoated ELISA plates have been used with the same experimental procedure to test for specificity. Absorbance data upon binding to plates without the target protein have been subtracted from values exhibited from ELISA plates with GlyR ECDs bound.

Data sets were tested for statistical significance by using unpaired t-test with the Holm-Šidák correction for multiple samples or the one-way ANOVA with Šidák’s multiple comparison post hoc test: *value of p<0.05; **value of p<0.01; ***value of p<0.001, ****p < 0.0001. Bar diagrams indicate mean values ± standard error of the mean (SEM) or as noted otherwise. Statistical analyses have been performed using GraphPad Prism version 9.3.1.

The proposed common epitope of GlyR autoantibodies between amino acid residues 29–62 (refers to residues 1–33 in the mature protein) (Rauschenberger et al., 2020) is close to the GlyR glycosylation site (Asn38, number refers to mature GlyR protein, Figure 1A). The region surrounding the glycosylation site might be important for autoantibody binding as it has been shown for other autoantibody subtypes (Gleichman et al., 2012; Castillo-Gomez et al., 2017). The N-glycosylation position 38 of the mature GlyRα1 is conserved in GlyRα2 and GlyRα3 (Figure 1A). First, we investigated the impact of the mutation on surrounding interactions using in silico mutational analysis. The crystal structure of human GlyRα3 homomer in complex with strychnine was used as template as it displays the best resolution among available glycine receptor structures (Figure 1B; Huang et al., 2015). In the wildtype structure, N-acetyl-D-glucosamine covalently linked to Asn38 interacts with Asn31 residing in the loop connecting the first α-helix and the first β-strand via a hydrogen bond (Figures 1C,D). In the mutant version (N38Q), this interaction was lost, while generating a possible interaction between the side chain of N38 with the main chain carboxylate of the P36 (Figure 1D). However, the overall structure in this region is unchanged between the glycosylated and the non-glycosylated form (Figure 1D, compare left and right image).

To verify the glycosylation state of the mutant GlyRα1^N38Q, cell lysates of GlyRα1^WT and GlyRα1^N38Q transfected HEK293 cells were treated with the glycosidases PNGaseF and EndoH. In contrast to GlyRα1^WT, PNGaseF and EndoH treated GlyRα1^N38Q showed no reduction in molecular weight compared to untreated samples (Figure 2A), thus demonstrating that the mutant GlyRα1^N38Q is present in a non-glycosylated isoform. The two bands observed after EndoH digestion of GlyRα1^WT result most probably from additional glycosylation modifications making the GlyR complex partially resistant to EndoH digestion (Bormann et al., 1993).

To verify the amount of non-glycosylated GlyRα1^N38Q at the cellular surface, biotinylation assays were performed discriminating between whole cell, intracellular and surface protein fractions. Signal intensities were quantified relative to pan-cadherin signals and normalized to GlyRα1^WT. Whole cell expression of GlyRα1^N38Q was reduced but not significantly changed compared to GlyRα1^WT (relative whole cell expression GlyRα1^WT 1.0 ± 0.16; GlyRα1^N38Q 0.57 ± 0.13; p = 0.1118). A similar decrease was observed for expressed GlyRα1^N38Q receptors at the cell surface (35% compared to GlyRα1^WT; relative surface expression GlyRα1^WT 1.0 ± 0.2; GlyRα1^N38Q 0.65 ± 0.12; p = 0.3888; Figures 2B,D). Gephyrin-transfected cells were used as control. As expected for such an intracellular protein, gephyrin was not detectable in the surface fraction (Figure 2C). Hence, the GlyRα1^N38Q is expressed at the cell surface, a prerequisite to test for binding of patient GlyR autoantibodies. First, patient sera were subjected to untransfected HEK293 cells and cells transfected with the empty pRK5 vector (MOCK control) showing no binding to the cells (Figure 3A). In contrast, patient sera bind to GlyRα1 transfected cells (Figure 3B). An endpoint titration of the sera on transfected HEK293 cells revealed variability in titres between 1:50–1:1000 (Pat1 1:1000; Pat2 1:500; Pat3 1:50 (titration experiment could not be performed due to lack of material); Pat4 1:500, Pat5 1:50; Figure 3B). Interestingly, all five patient sera (Pat1-5), which recognized the glycosylated GlyRα1^WT, were also able to bind the non-glycosylated GlyRα1^N38Q isoform (Figures 4A,B), thus arguing that receptor glycosylation is not essential for GlyR autoantibody binding.

To investigate whether the GlyRα1^N38Q mutant shows functional alterations, we performed electrophysiological recordings. The dose–response curve of GlyRα1^N38Q revealed an increased glycine EC₅₀ value of 206 μM compared to GlyRα1^WT with an EC₅₀ of 85 μM (p = 0.00029; Figures 5A,D). Comparing the absolute currents of GlyRα1^WT and GlyRα1^N38Q at a glycine concentration of 100 μM, a significant reduction of the I_max values was obvious for the mutant (WT: 3.6 ± 0.5 nA; N38Q: 1.0 ± 0.3 nA; p = 0.0006) which was absent at saturating concentrations of glycine (1 mM; GlyRα1^WT: 6.0 ± 0.6 nA; GlyRα1^N38Q: 5.0 ± 0.8 nA; p = 0.398) (Figures 5B,C). These data indicate a decreased glycine potency of GlyRα1^N38Q whereas the efficacy is unaltered.

As a first life cell-based approach, GlyRα1 transfected HEK293 cells were incubated with patient autoantibodies for 1 h. This approach presumes that GlyR autoantibodies bind to the proposed N-terminal epitope A²⁹-G⁶² and are adsorbed by access to the native GlyR ECD. Afterwards, the supernatant was transferred three times to another dish with transfected cells for further incubation. For all patient samples a decrease in signal intensity was seen following two rounds of incubation with transfected cells (Figure 6A; controls shown in Supplementary Figure S1A). The sera of patients 2 and 5 required three rounds of transfer until the adsorption was almost complete (Figure 6A). For verification that autoantibodies are not degraded during experimental procedure, coverslips without cells were used for transfers 1 and 2, and finally incubated with transfected HEK293 cells. Normal binding of autoantibodies to expressed GlyRα1 was detected (Supplementary Figure S1B). Comparing the first and second labelling, the ratio of mean signal intensity of patient sera to the GFP signal was significantly reduced in 4 of 5 patients (Figure 6B). The reduction was significant for all patients when comparing the first to the last labelling (Pat1 p = 0.003; Pat2 p = 0.020; Pat3 p = 0.0214; Pat4 p < 0.0001; Pat5 p = 0.0085). The commercial GlyRα1-specific antibody MAb2b showed intense labelling even after four rounds of labelling due the high concentration of used antibody (2 μg/ml) (Supplementary Figure S1A). The mean ratio of signal intensities was unaltered between the first and the second as well as first and last labelling when cells were incubated with either MAb2b, HC, or DC (first to last transfer MAb2b p = 0.3142, HC p = 0.7650, DC p = 0.8859; Figure 6B). In sum, autoantibody binding to the native glycosylated GlyR ECD was verified, which consequently represents a suitable tool to efficiently adsorb autoantibodies against the GlyR from patient serum samples.

We determined that GlyR glycosylation is not essential for autoantibody binding (Figure 4). Therefore, a GlyRα1 ECD expressed, purified, and re-folded from E. coli (Breitinger et al., 2004) may serve as a valuable tool for GlyR autoantibody screening due to efficient binding of GlyR autoantibodies from patient sera. E. coli cells are unable to generate glycosylated protein, therefore the verification of GlyR autoantibody binding to non-glycosylated GlyRs was a prerequisite for the use of GlyR ECDs as tool for autoantibody detection. The GlyRα ECD constructs contain a 6xHis-tag and harbours the ligand binding and receptor assembly domains. Purity and folding of the GlyRα ECD protein (29 kDa, dimer at 58 kDa) was verified with Coomassie stain and Western blot using MAb4a or an anti-His-tag antibody as well as CD spectroscopy (Figures 7A,B).

Following coating of ELISA plates with GlyRα1 ECD, the patient autoantibody containing samples were incubated for 1 h with coated wells and transferred twice. The binding of autoantibodies to the GlyRα1 ECD in the well plate was analysed by determining the absorbance at 450 nm (Figure 7C). MAb4a showed the highest absorbance and served as positive control. Patient sera 1–4 showed increased absorbance values compared to healthy and disease control (compared to DC: p-values Pat1 p = 0.0757, Pat2 p = 0.0556, Pat3 p = 0.0007, Pat4 p = 0.0427, Pat5 p = 0.639), indicating that patient autoantibodies bound to the purified non-glycosylated GlyRα1 ECD from E. coli. To control for efficient patient autoantibody adsorption, the ELISA supernatants were subjected to primary motoneurons where GlyRs are highly abundant and to GlyRα1 transfected HEK293 cells. In motoneurons, GlyR-labelling of patient autoantibodies was largely reduced whereas the synapsin signal was constant independent of the ELISA adsorption of autoantibodies (Figures 7D,E). The same autoantibody signal reduction was identified when the suspension was transferred to GlyRα1 transfected HEK293 cells (Supplementary Figure S2). It has been shown that GlyR autoantibodies do not only target the α1 subunit but also the α2 and α3 subunits (Carvajal-Gonzalez et al., 2014). Using a similar approach with the purified ECD of the GlyRα2 subunit in an ELISA, patient autoantibodies against the α2 subunit were also efficiently bound and thus verified by two independent measures using a cell-based experiment and an ELISA-based assay (Figures 8A,B). Pat3 was excluded from this experiment due to a lack of sufficient serum material. Interestingly, the serum of Pat4 showed still binding to with GlyRα2 transfected HEK293 cells in the cell-based assay following the ELISA. This result demonstrates that GlyR autoantibodies remained in the diluted serum sample which further suggests that the serum titre of GlyR autoantibodies from Pat4 might be higher than that of the other samples or an inter-subunit epitope between two adjacent GlyR subunits exists which can only be presented by native pentameric GlyRα2 receptors but not by the GlyRα2 ECD. The titre of Pat4 was with 1:500 higher than for Pat5 but similar to Pat2 and Pat1 (Figure 3B).

In sum, the non-glycosylated GlyRα1 did not exhibit large structural alterations, was to a significant degree transported to the cell surface and reveals an increased glycine potency while glycine efficacy was unaffected. As glycosylation is not a prerequisite for GlyR autoantibody binding, GlyR ECD domains purified from E. coli represent a valuable tool to screen for patient autoantibodies with high efficiency in addition to the so far used cell-based assays on living cells.