Structural mechanisms of GABA_A receptor autoimmune encephalitis

Summary

Autoantibodies targeting neuronal membrane proteins can cause encephalitis, seizures, and severe behavioral abnormalities. While antibodies for several neuronal targets have been identified, structural details on how they regulate function are unknown. Here we determined cryo-electron microscopy structures of antibodies derived from an encephalitis patient bound to the γ-aminobutyric acid type A (GABA_A) receptor. These antibodies induced a severe encephalitis by directly inhibiting GABA_A function, resulting in nervous system hyperexcitability. The structures reveal mechanisms of GABA_A inhibition and pathology. One antibody directly competes with neurotransmitter and locks the receptor in a resting-like state. The second antibody targets the subunit interface involved in binding benzodiazepines and antagonizes diazepam potentiation. We identify key residues in these antibodies involved in specificity and affinity and confirm structure-based hypotheses for functional effects using electrophysiology. Together these studies define mechanisms of direct functional antagonism of neurotransmission underlying autoimmune encephalitis in a human patient.

Graphical Abstract

In Brief:

Structural analysis of antibodies from an encephalitis patient bound to the GABA_A receptor reveals mechanisms of channel inhibition and pathology that drive autoimmune encephalitis in humans.

Introduction

Autoimmunity arises when the body’s T-cells or antibodies attack self-antigens. The trigger for this improper immune response can be environmental, such as a virus, or internal, such as a tumor. The consequences of autoimmune disease can be short-lived and self-resolving, or can become chronic, debilitating, and lead to death. Autoantibodies targeting a variety of tissues have been investigated for decades (1988; Obermayer-Straub et al., 2000). However, the study of neuropsychological disorders directly mediated by anti-neuronal antibodies is in its infancy. The first case of anti-NMDA receptor encephalitis was described in 2007 (Dalmau et al., 2007), and represents one entity in the rapidly growing disease category of autoimmune encephalitis (Prüss, 2021). Autoimmune encephalitis is characterized by severe and acute-onset symptoms comprising psychosis, refractory seizures, movement disorders, impaired consciousness and dysautonomia. Established antibodies in serum or cerebrospinal fluid from affected patients can now be routinely detected in cell-based assays overexpressing the respective proteins. Novel clinically relevant antibodies are continuously discovered, and the antigenic targets identified using immunoprecipitation and mass spectrometry. Anti-neuronal targets associated with pathological disease include ionotropic and metabotropic glutamate receptors, Leucine Glioma Inactivated 1 (LGI1), contactin-associated protein-like 2 (CASPR-2), dopamine receptors, and metabotropic and ionotropic γ-amino butyric acid (GABA) receptors (GABA_B and GABA_A, respectively) (Dalmau and Graus, 2018; Prüss, 2021).

GABA_A receptors are part of the greater Cys-loop receptor family of pentameric ligand-gated ion channels. The GABA_A receptors form anion-selective channels, and generally act to dampen neuronal excitability. They are usually heteromeric, assembling from a panel of 19 subunits. The most common isoform at synapses comprises two α1 subunits, two β subunits, and one γ2 subunit (Nutt and Malizia, 2001) (Figure 1). Each subunit is positioned symmetrically or pseudo-symmetrically around the axis of the ion pore. The neurotransmitter GABA binds in the extracellular domain (ECD) at β-α subunit interfaces. This binding triggers conformational changes that open an ion channel formed by the second helix (M2) of the transmembrane domain (TMD). In the sustained presence of agonist, the channel adopts a non-conducting desensitized state (Gielen and Corringer, 2018). When GABA concentrations decrease, GABA dissociates and the channel returns to the resting, activatable state. Inhibiting this inhibitory receptor leads to nervous system hyperexcitability, which can result in anxiety, seizures, catatonia, and death. These ends may be achieved through receptor internalization, or through stabilizing one of the closed-channel states.

Figure 1: Encephalitis-associated autoantibodies bound to the synaptic GABA_A receptor.

A, Side and top views of the Fab₁₁₅: GABA_A complex cryo-EM map. Subunits and Fab densities are indicated. The constant region of the Fab is not shown as this peripheral domain is flexible and conformationally disordered. B, Same as for A, but with the Fab₁₇₅.

As with other autoimmune encephalitides, target identification of the GABA_A receptor is the essential starting point for understanding, diagnosing and treating disease (Petit-Pedrol et al., 2014). However, it reveals little about how the improper immune response causes pathology. For different forms of autoimmune encephalitis, the mechanism of the autoantibodies’ pathogenicity is mainly attributed to target receptor clustering and internalization (Dalmau et al., 2017). Direct autoantibody effects on neuronal function have only recently been appreciated (Crisp et al., 2019). Isolation of monoclonal antibodies from the CSF of an 8-year-old encephalitis patient identified at least two different autoantibodies that cause seizures and status epilepticus in the absence of receptor internalization (Kreye et al., 2021). Here, we use structural biology to reveal the pathogenic mechanisms of these patient-derived anti-neuronal antibodies bound to their physiological target, the GABA_A receptor.

Results

Identification of stable autoantibody-GABA_A receptor complexes

We tested a panel of five patient-derived antibodies (Kreye et al., 2021) for binding to the purified α1β2γ2 GABA_A receptor (Figure S1, refers to Figure 1). Purified monoclonal antibodies were first screened for receptor binding by fluorescence-detection size exclusion chromatography (FSEC) (Kawate and Gouaux, 2006). This assay detected binding by all antibodies tested except for mAb₁₉₈ (Figure S1A). Fab fragments were then generated from antibodies mAb₁₀₁, mAb₁₁₅, mAb₁₇₅ and mAb₂₀₁ for structural studies. Binding stoichiometries were estimated by FSEC with increasing molar ratios of Fab to receptor (Figure S1B–E, refers to Figure 1) and reactivity to native receptors was analyzed in murine brain tissues (Figure S1F, refers to Figure 1) before large-scale complex formation and purification. Fab₂₀₁ failed to bind detectably to the GABA_A receptor and was excluded from structural studies. Fab₁₀₁ no longer bound the GABA_A receptor with an affinity suitable for structural work; it may require bivalency to bind appreciably. We thus pursued both Fab₁₁₅ and Fab₁₇₅ in structural studies.

Structure of the GABA_A receptor bound to an inhibitory autoantibody

In contrast to mechanisms of autoantibody-induced neurological disease described thus far (Hughes et al., 2010; Kreye et al., 2016; Petit-Pedrol et al., 2014), mAb₁₁₅ and Fab₁₁₅ directly inhibit GABA_A receptor function without causing receptor internalization from the cell surface (Kreye et al., 2021). We were intrigued by this phenomenon and sought to understand its mechanism of inhibition by obtaining the structure of Fab₁₁₅ bound to the GABA_A receptor. We purified the α1β2γ2 synaptic GABA_A receptor and reconstituted it into brain lipid nanodiscs (Kim et al., 2020). Reconstituted receptors were incubated with Fab₁₁₅ before further polishing by size-exclusion chromatography (Figure S1G,H). We applied the purified complex to cryo-electron microscopy (cryo-EM) grids and collected a dataset that allowed us to reconstruct a 3.0 Å resolution map (Figures 1A and S2, Table S1). The map quality allowed us to confidently model the receptor and the variable domains of the Fab (V_H and V_L for heavy and light chain variable domains, respectively). Local resolution in the areas of greatest interest, including side-chain interactions between Fab₁₁₅ and receptor, was ~2.9 Å (Figure S2I, S3 and Video S1, refer to Figures 1, 2, 3).

Figure 2: Fab₁₁₅ competes directly with neurotransmitter binding.

A,B Side and top views of receptor and Fab (gold) atomic models. Boxes indicate expanded regions shown in panels C and D. C, Fab-receptor interactions in the neurotransmitter binding site. Key amino acids are highlighted; dashed lines indicate putative electrostatic interactions closer than 4 Å. D, Anchoring interaction between Fab115 and GABAA. Residues shown are an alignment of key interacting residues of α subunits; red letters indicate potential specificity-determining residues. E, Subunit interface contacts with Fab. One β2:α1 subunit interface is shown; red box indicates neurotransmitter binding site and spheres indicate amino acids contacted by Fab₁₁₅. Inset table shows surface area buried by Fab heavy and light chains on respective GABA_A receptor subunits.

Figure 3: Fab₁₇₅ binds the GABA_A receptor α-γ interface.

A, Overview of model with area shown in panels B and C indicated. B, Amino acid interactions between Fab₁₇₅ and the α1-γ2 interface of the GABA_A receptor. C, The interactions between the lower part of the Fab and the α1 benzodiazepine binding pocket Loop C. D, Subunit interface contacts with Fab. One β2:α1 subunit interface is shown; red box indicates neurotransmitter binding site and spheres indicate amino acids contacted by Fab₁₇₅. Inset table shows surface area buried by Fab heavy and light chains on respective GABA_A receptor subunits.

The complex structure reveals one Fab₁₁₅ bound at each of the two β-α subunit interfaces in the ECD (Figure 1A, Figure 2, Figures S3,S4). The complementarity determining region (CDR) 3 of the Fab heavy chain penetrates the neurotransmitter binding pocket and contacts conserved residues known to impact agonist binding (Figure 2B, Figures S3–5, refer to Figures 2, 3). In particular, R105 of the V_H CDR3 is positioned to make electrostatic and cation-π interactions with several residues on the β2 and α1 subunits (Figure 2C). Of special interest is the contact between V_H R105 and β2 E155. E155 forms a salt bridge with the primary amine of the neurotransmitter GABA and is critical for binding and gating (Jatczak-Sliwa et al., 2020; Mortensen et al., 2014; Newell et al., 2004). Other residues lining the neurotransmitter binding site that interact with CDR3 of Fab₁₁₅ have also been implicated in binding to GABA, including the following on the β subunit: Y97 (Boileau et al., 2002), Y157 (Amin and Weiss, 1993; Newell and Czajkowski, 2003; Newell et al., 2004), S156 (Newell et al., 2004), F200 (Zhu et al., 2018), and Y205 (Amin and Weiss, 1993; Zhu et al., 2018). Thus, the Fab₁₁₅ V_H CDR3 mimics an aspect of neurotransmitter chemistry and likely competes with GABA binding in a manner analogous to a competitive antagonist (Kim et al., 2020; Masiulis et al., 2019).

In the initial antibody characterization, mAb₁₁₅ was found to be selective for binding among different GABA_A receptor subunits (Kreye et al., 2021). The structure reveals the basis of this exquisite selectivity. First, although GABA_A β subunits can form a homopentamer, β alone did not bind antibody despite its numerous contacts with Fab₁₁₅ in the neurotransmitter binding pocket (Kreye et al., 2021). Examination of the Fab₁₁₅:receptor structure reveals that most of the interactions between Fab₁₁₅ and receptor are with the α1 subunit (Figures 2E, S5); contacts with β alone cannot support binding. Furthermore, mAb₁₁₅ is only able to bind GABA_A receptors containing the α1, but not α2–5 subunits. This specificity may arise from interactions between Fab₁₁₅ and four key residues on the α1 subunit: E170, R173, E174, and R177 (Figure 2D). While only R177 has been studied for functional importance (Newell and Czajkowski, 2003), evidence for the contribution of all these residues to antibody binding specificity is found in an examination of contacts between Fab₁₁₅ and the receptor (Figure S5, yellow highlights). Most contacted residues are conserved among α1-α6; as mAb₁₁₅ can only bind α1, residues shared across subunits would not be expected to be sufficient for antibody binding. The exceptions are T122-D124 and E170-V180. In the case of T122-D124, Fab₁₁₅ only makes electrostatic interactions with backbone carbonyl oxygens, and therefore the identity of the residues present in the receptor are less important. The string of 11 residues beginning at E170 exhibits variable conservation (Figure 2D); however, the tetrad of E170, R173, E174 and R177 is only found in α1, and all four residues are involved in salt bridges. These electrostatic contacts are poised to stabilize the Fab-receptor interaction and likely determine the selectivity of Fab₁₁₅ for α1-containing receptors.

Structural and sequence analyses suggest the ways in which mAb₁₁₅ evolved to increase not just its selectivity but also its affinity for the GABA_A receptor. Somatic hypermutation is a process unique to B cells and antibody secreting cells that diversifies immunoglobulins in their variable gene regions as an adaptive immune response to enhance affinity for their antigenic target. We compared the Fab₁₁₅ antibody sequence to an immunoglobulin gene database to derive a germline ancestor with highest sequence identity. This approach identifies somatically hypermutated residues within the variable V(D)J gene segments that have evolved for affinity maturation (Figure S6, (Reincke et al., 2020)). Both Fab₁₁₅ V_H and V_L contain somatically hypermutated residues that are involved in binding the GABA_A receptor (Figure S6, refers to Figures 1–3, S3). These include R97 and K98 of the Fab₁₁₅ V_L, which are encoded in the germline sequence as an asparagine and a glycine. These mutations contribute two new positive charges at each interface that are stabilized at the interface through salt bridges to help anchor Fab₁₁₅ to the α1 subunit (Figure 2D). Furthermore, D52 of the Fab₁₁₅ V_H is an asparagine in the germline antibody sequence (Figures 2D, S6). This mutation from a polar residue to a charged residue would generate a stronger molecular interaction, changing it from a hydrogen bond to a salt bridge. Thus, somatic hypermutations are predicted to enhance the affinity of Fab₁₁₅ for its target.

Structure of the GABA_A receptor in complex with a non-competitive autoantibody

The mAb₁₇₅ autoantibody also triggered seizure activity in rodents and was shown to require both the α1 subunit and the γ2 subunit for binding to the GABA_A receptor in cell-based assays (Kreye et al., 2021). The α-γ interface is of special interest because it forms the binding site for benzodiazepines, a modulatory class of GABA_A drugs that includes anxiolytics and sedative-hypnotics (Sigel and Ernst, 2018), and is an important site for endogenous modulators, termed endozepines (Tonon et al., 2020). In accordance with requirement of the α-γ interface for mAb₁₇₅ binding, FSEC experiments indicated a stoichiometry of 1:1 for receptor:Fab₁₇₅ (Figure S1D, refers to Figures 3,S2,S3). We used cryo-EM to obtain a structure of the Fab₁₇₅:GABA_A complex at an overall resolution of 3.0 Å (Figure 1B, Figure S2H, Table 1). The map quality allowed us to confidently model the receptor and the variable domains of the Fab (Figure S3G–J).

Fab₁₇₅ binds at the α1-γ2 subunit interface, consistent with cell-based and FSEC experiments (Figures 1B, 3A, S3G–J, S4C–D). In a manner similar to Fab₁₁₅, the majority of surface area contacted by Fab₁₇₅ is with the α1 subunit (Figures 3D, S5). Fab₁₇₅ V_H contains a methionine (M103) that buries itself into a hydrophobic pocket composed of residues from both sides of the α1:γ2 interface, above the benzodiazepine site (Figures 3B, S3H,J). The Fab₁₇₅ V_L docks onto Loop C of the α1 subunit; this loop caps the benzodiazepine-binding pocket of the receptor (Figures 3C, S3G). This V_L attachment site centers upon α1 R164 of the GABA_A receptor, which forms a network of five electrostatic interactions connecting the two proteins (Figure 3C). This arginine is unique to α1; while some of the other α subunits contain a lysine at this position, a primary amine would not recapitulate the elaborate network of electrostatic interactions. This arginine likely contributes to subunit specificity, as most other receptor residues contacted by Fab₁₇₅ are conserved (Figure S5, refers to Figures 1–3)). Fab₁₇₅ underwent somatic hypermutation to a lesser degree than Fab₁₁₅, changing six amino acids in the heavy chain and one in the light chain (Figure S6, refers to Figures 1–3)).

Autoimmune encephalitis Fabs act as competitive and allosteric antagonists

The conformational states observed in the receptor structures shed light on how autoantibody binding triggers disease pathology. We first compared the two autoantibody complexes with the receptor bound to GABA alone (Figure 4A, PDB 6X3Z, (Kim et al., 2020)). Fab₁₁₅, through direct interactions, stabilizes the neurotransmitter site Loop C in a widely open conformation at β-α interfaces akin to its conformation in the presence of competitive antagonists (Figure 4B). Unlike the Fab₁₁₅-bound receptor, the Fab₁₇₅-bound receptor contains density consistent with GABA in its binding site, even though the agonist was not added to the preparation. The presence of GABA induces a partial closure of Loop C in the neurotransmitter binding pockets (Figure 4B). Normally, closure of Loop C by agonist triggers a rotation and contraction of the ECD, which is then translated into opening of the transmembrane pore (Kim and Hibbs, 2021; Nemecz et al., 2016). After channel opening, the agonist remains bound but a lower gate in the pore closes, a process termed desensitization. Indeed, in a similar preparation, the presence of endogenous GABA caused the transmembrane domain to adopt a desensitized-like conformation (Laverty et al., 2019). However, the Fab₁₇₅-bound GABA_A structure reveals an interruption in this state transition (Figure 4E, F). Although Loop C is mostly closed, the ECD of the Fab₁₇₅ complex is not contracted. Fab₁₇₅ blocks this contraction by wedging between the α1 and γ2 subunits in the upper part of the ECD (Figure 3B). Superposition of the Fab175 bound structure on the GABA-bound structure suggests many atomic clashes would prevent the ECD contraction that initiates channel gating in the pentameric receptor superfamily (Figure 4G, (Lefebvre et al., 2021). The consequence is that GABA is positioned farther from the residues of the neurotransmitter binding pocket. Specifically, the GABA amino nitrogen is 4.1–4.5 Å from E155, as opposed to 2.7–3.1 Å in the GABA-bound structure in the absence of Fab (range covers the two GABA sites). The GABA-E155 interaction is required for converting agonist binding into channel gating (Newell et al., 2004), and is weakened in the presence of Fab₁₇₅. Together, the consequences of both Fab₁₁₅ and Fab₁₇₅ binding are stabilization of the receptor pore in a resting-like, non-conducting conformation, with the classical activation gate at the 9ʹ position closed (Figure 4D, F, H). The structural findings thus suggest a direct competition mechanism for Fab₁₁₅ and an allosteric antagonism mechanism for Fab₁₇₅, with both stabilizing a resting-like state of the ion channel.

Figure 4: Conformational Differences between Fab₁₁₅-bound, Fab₁₇₅-bound, and GABA-bound receptors.

A, Side view of Fab₁₁₅-bound (bright green, dark blue, magenta), Fab₁₇₅-bound (light green, light blue, pink), and GABA-bound (gray, PDB 6X3Z (Kim et al., 2020)). Boxes indicate areas represented in indicated panels. B, Detail of Loop C displacement at a β2-α1 interface; mesh indicates residual density for GABA. C, Top view of the ECD comparing Fab₁₁₅-bound vs Fab₁₇₅-bound. Arrows indicate binding sites of Fab₁₇₅ (red) and Fab₁₁₅ (gold). D, Same as in C, but comparing cylinder representations of the TMD. E, Top view of the ECD comparing Fab₁₇₅-bound vs GABA-bound. F, same as in E but comparing cylinder representations of the TMD. G, Fab₁₇₅ docked into GABA-bound receptor structure (6X3Z); spheres indicate side chains that clash. H, Pore radius analysis for a panel of GABA_A receptor structures. GABA_A + Fab₁₇₅ and GABA_A + Fab₁₁₅, this study. 6I53 and 6X3Z are agonist-bound (Kim et al., 2020; Laverty et al., 2019). 6HUJ and 6X3S are antagonist-bound (Kim et al., 2020; Masiulis et al., 2019).

We sought to probe the structure-based autoantibody mechanisms using functional measurements. Initial electrophysiological studies on autaptic neurons found that mAb₁₁₅ but not mAb₁₇₅ directly inhibited GABA-activated currents (Kreye et al., 2021). We performed whole cell patch clamp experiments on transfected cells expressing the α1β2γ2 receptor to test the concentration dependence of Fab inhibition. These experiments revealed, first, that Fab₁₁₅ binds effectively irreversibly (Figure 5A). Application of GABA after exposing cells to increasing concentrations of Fab revealed a clear dose-dependence in inhibition (Figure 5B). Tests with Fab₁₇₅ revealed no substantial inhibition of GABA activation, consistent with the earlier results testing inhibition by mAb₁₇₅ on autaptic neurons. Nonetheless, when the antibody was injected into mouse CNS, it triggered spontaneous epileptic events (Kreye et al., 2021). The structural information suggested that Fab₁₇₅, through binding at the α-γ interface near the benzodiazepine site, could affect the activity of modulators acting there. We thus tested the effect of Fab₁₇₅ on diazepam potentiation (Figure 6). We found that Fab₁₇₅, in a concentration dependent manner, inhibits diazepam potentiation of GABA activation (Figure 6B). A speculative mechanism for Fab₁₇₅ antagonism of receptor activity may thus involve altering the way endogenous modulators (endozepines) act through this site, with effects on GABA activity too subtle to detect in our patch clamp experiment. One prominent endozepine is the diazepam binding inhibitor (DBI), which is a benzodiazepine site ligand that was recently found to be an anti-epileptic positive modulator in the thalamic reticular nucleus (Christian et al., 2013). Antagonism of this interaction, as well as other positive modulator endozepines (Tonon et al., 2020), may be a component of how this antibody triggers disease pathology; studies with GABA_A subunit knockout mice could further substantiate this hypothesis. Taken together, these findings reveal distinct mechanisms leveraged by the two antibodies to cause autoimmune encephalitis with seizures.

Figure 5: Fab₁₁₅ irreversibly antagonizes GABA_A receptor activity.

A, Representative whole cell patch-clamp electrophysiology recordings from transfected HEK cells. Response to GABA activation is shown before and after perfusion of Fab₁₁₅. B, Bar graph showing mean responses with standard deviation for the GABA response in the absence (control) and presence of increasing concentrations of Fab₁₁₅. T-test **** p<0.0001, *** p<0.001. Biological replicate numbers (individual cells) are shown in the graph. C, Cartoon model representing a proposed mechanism of GABA_A receptor antagonism by Fab₁₁₅.

Figure 6: Fab₁₇₅ inhibits benzodiazepine-mediated potentiation of GABA_A receptors.

A, Representative whole cell patch-clamp electrophysiology recordings from transfected HEK cells. Response to GABA activation in the absence and presence of diazepam is shown, followed by the response to GABA plus diazepam following application of Fab₁₇₅. B, Bar graph showing mean responses with standard deviation for the fractional diazepam potentiation before (control) and after application of Fab₁₇₅. T-test * p=0.64, not significant; *** p=0.0006. Biological replicate numbers (individual cells) are shown in the graph. C, Cartoon model representing a proposed mechanism of action of Fab₁₇₅ at the GABA_A receptor.

Discussion

The extent to which humoral autoimmunity in the CNS triggers disease has only recently begun to be appreciated. To date there have been no structural studies on how disease-causing autoantibodies from patients act on their neurotransmitter receptor targets. Here we determined structures of the GABA_A receptor bound to two patient-derived pathogenic antibody fragments. The structures reveal how the antibodies bind at distinct subunit interfaces to interfere with receptor function. Mapping of the interactions identified residues important for the specificity of the human antibodies for restricted subtypes of GABA_A receptors. We defined conformational changes that are stabilized or inhibited in the presence of autoantibodies by comparison to established GABA_A receptor structures. These structural analyses combined with electrophysiological measurements provide mechanisms for direct and allosteric antagonism of inhibitory neurotransmission by autoimmune antibodies and can explain the clinical symptoms of patients with GABA_A receptor encephalitis at the atomic level.

Limitations of the Study

In this study we determined two structures of receptor-bound recombinant antibodies that were isolated from an individual patient in the acute phase of an autoimmune encephalitis. While we have shown functional effects of both Fabs that correlate with structural predictions, systematic mutagenesis would be important to confirm the relative importance of individual determinants of antibody-receptor binding. Further, we cannot draw broad conclusions about effects from other GABA_A receptor autoantibodies from this individual and other patients. In a previous study, these antibodies recapitulated a severe clinical phenotype reminiscent of autoimmune encephalitis with the predominant symptom of seizures in rodent models without causing receptor internalization. It is possible that other GABA_A receptor autoantibodies in this patient and in separate autoimmune encephalitis cases work through receptor cross-linking and internalization. Our findings support that cross-linking and internalization are not the only mechanisms of autoimmune antibodies targeting neuronal receptors and draw attention to a spectrum of alternate mechanisms contributing to autoimmune pathology.

STAR Methods

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ryan E. Hibbs (ryan.hibbs@utsouthwestern.edu).

Materials Availability

All unique reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Data and Code Availability

• Cryo-EM maps and atomic model coordinates have been deposited in the EMDB and RCSB, respectively, under the accession codes 7T0W/EMD-25583 and 7T0Z/EMD-25585. They will be publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental Model and Subject Details

Cell Lines

The Sf9 insect cells were obtained from the ATCC (Cat# CRL-1711) and cultured in an InnovaR shaking incubator at 27°C at 130 rpm with Sf900 III media (Thermo Fisher Scientific). HEK293S GnTI-suspension cells were obtained from the ATCC (Cat# CRL-3022) and cultured in a Thermo Scientific Reach-In CO₂ Incubator at 37°C with 8% CO2 at 130 rpm. All cell lines used in this study are female in origin.

Method Details

Fab production and purification

Fab fragments were produced as previously described (Kreye et al., 2021). cDNA encoding the heavy and light chains of the immunoglobulins was isolated via reverse transcription and specific amplification from single cerebrospinal fluid cells sampled from a pediatric patient in the acute phase of GABA_A receptor encephalitis. These sequences were cloned into mammalian expression vectors, from which monoclonal antibodies were recombinantly expressed in HEK293T cells and purified from cell culture supernatant for initial characterization. Fab fragments were generated by replacing the Fc domains CH2 and CH3 of the heavy chain expression vector with a FLAG-Tag and a His-Tag following the amino acids PKSCDKTH of the hinge region. Fab fragments were expressed, purified using affinity chromatography via the His-Tag and controlled for purity using SDS Page by an external provider (inVivo BioTech, Germany).

Immunohistochemistry

Immunohistochemical stainings of murine brain tissue were performed as previously described (Kreye et al., 2016; Kreye et al., 2021). In brief, 20 μm sagittal cryosections were fixed with paraformaldehyde (PFA), first blocked in PBS supplemented with 2% BSA (Dianova) and 5% Normal Goat Serum (Abcam) at room temperature for one hour, then incubated with 20 μg/ml of Fab at 4°C overnight, then with an anti-human Fab-specific secondary antibody (Dianova, 1:500) at room temperature for 2 hours, and finally with DRAQ5 (Thermo Fisher Scientific, 1:500) at room temperature for 10 minutes. Confocal images were recorded at an inverted fluorescence microscope (Leica DMI8) with identical conditions for all Fabs.

Receptor expression, purification, and reconstitution into nanodiscs

To express the GABA_A receptor, we generated bacmam virus from a tri-cistronic construct encoding the α1, β2 and γ2 subunits of the GABA_A receptor (Kim et al., 2020). These subunits contained a deletion in the intracellular domain between transmembrane helices M3 and M4. These amino acids were replaced with an SQPARAA linker; this deletion does not affect functional behaviors in electrophysiology and functional assays (Kim et al., 2020; Zhu et al., 2018). Bac-mam viruses were made in Sf9 cells (Vaughn et al., 1977), and the tric-cistronic virus was transduced at an MOI of 1 into HEK293S GnTI⁻ suspension cells (Reeves et al., 2002) at a density of 3–4 × 10⁶/ml. The cells were cultured at 30 °C for 72 h before harvesting by centrifugation. Cells were washed once in 20 mM Tris, pH 7.4, 150 mM sodium chloride (TBS) and recentrifuged. After resuspension in TBS + 1 mM phenylmethanesulfonyl fluoride (PMSF; Sigma), cells were lysed by passage through an Avestin Emulsiflex at 5,000–15,000 psi. Lysed cells were first centrifuged at low speed (10,000 g) to remove nuclei and unlysed cells, then centrifuged at 186,000 g for 2 h to pellet membranes. Membranes were stored at −80 °C until used. To purify proteins, approximately 5 g of membranes were Dounce homogenized in TBS + 1 mM PMSF. The detergent n-dodecyl-β-D-maltoside (DDM; Anatrace) was added to a final concentration of 40 mM, and solubilization was performed at 4 °C while nutating. Insoluble material was removed by centrifugation at 186,000 g for 40 min. The supernatant containing solubilized protein and its native lipids was bound to Strep-Tactin Superflow high-capacity resin (IBA Life Sciences) via gravity flow. The resin was washed with TBS plus 1 mM DDM, supplemented with 1% Porcine Brain Polar Lipids (Avanti) to promote receptor stability. The GABA_A receptor protein was eluted in the same wash buffer supplemented with 5 mM desthiobiotin (Sigma).

Purifications of the GABA_A receptor were done in the absence of exogenous ligand. For autoantibody FSEC experiments shown in Supplemental Figure 1A, purified receptor reconstituted in lipidic nanodiscs was aliquotted and flash frozen in liquid nitrogen until required. A GABA_A receptor aliquot was thawed on ice, and ~2 μg of receptor was incubated with ~0.5–2 μg of each Fab or mAb at room temperature for 10 min. Protein was diluted in TBS (20 mM Tris, pH 7.4, 150 mM NaCl) and centrifuged to pellet aggregated receptor before analysis by FSEC. The Fab stoichiometry screening shown in Figure S1B–E was performed in a similar manner, with increasing molar ratios of Fab to receptor.

Cryo-EM Sample Preparation

For Cryo-EM, the GABA_A receptor protein reconstituted in nanodiscs was concentrated to ~500 μl. The concentrated protein was divided into 50μl aliquots and preincubated with porcine brain polar lipids for 10 minutes at room temperature. Saposin (Lyons et al., 2017) was then added for two minutes at room temperature. Porcine brain polar lipids and saposin were added to the purified protein at a molar ratio of 1 GABA_A protein: 30 saposin protein: 230 lipid. After the addition of the saposin the mixture was diluted to 500 μl for each aliquot with TBS to initiate reconstitution. This mixture nutated at 4°C overnight in the presence of bio-beads (Bio-Rad) to remove detergent at a concentration of 100mg/ml. The removal of detergent was continued the next morning with a fresh batch of bio-beads at 4 °C for one hour, after which incorporation was assayed by fluorescence-detection size-exclusion chromatography (Kawate and Gouaux, 2006). Fab was added to the receptor at molar ratios of 1:3 (for Fab₁₁₅) or 1:2 (for Fab₁₇₅). The complex was further polished by size exclusion chromatography using TBS as a mobile phase. The best fractions, as assessed by FSEC, were pooled and concentrated to ~4–6 mg/ml before freezing for cryo-electron microscopy (cryo-EM).

For preparation of cryo-EM grids, 200-mesh Cu 1.2/1.3 grids from Quantifoil were glow-discharged in a Pella glow discharge apparatus for 30 mA/80s on top of a metal grid holder (Ted Pella). To induce random particle orientations, protein was supplemented with 0.5 mM fluorinated fos-choline-8 (Anatrace) immediately before freezing. Grids were frozen with a Mark IV Vitrobot (FEI/ThermoFisher). Three μl of protein were applied to grids before blotting for 3.5 s at 100% humidity and 4 °C. Frozen grids were stored in liquid nitrogen until used.

Cryo-EM data collection and processing

The electron microscopy was performed on a 300kV Titan Krios (FEI/Thermo-Fisher) equipped with a K3 direct electron detector and a GIF energy filter (20 e⁻V) using super-resolution mode (Gatan) and collected via SerialEM (Mastronarde, 2005). The approximate physical pixel size was 1.079 Å/pixel. The total dose was 36–86 e⁻/Ų.

All data were processed within the Relion 3.0.8 suite of software (Zivanov et al., 2018). Dose-fractionated images were collected using SerialEM (Mastronarde, 2005). Data processing is described in Supplemental Figure 3. In brief, micrographs were motion corrected and gain normalized using MotionCor2 (Zheng et al., 2017). Contrast transfer function (CTF) and defocus values were estimated using GCTF (Zhang, 2016). Particles were picked using CrYOLO (Wagner et al., 2019). Following 2D and 3D classification, particle orientations were refined and subjected to Bayesian polishing, all performed in Relion (Zivanov et al., 2018). Focused 3D classification was performed using the refined angles on polished particles. For the Fab₁₇₅ dataset, we performed a further 3D classification focused on the γ-TMD, which is more mobile in some purified sample preparations (Zhu et al., 2018). The Fab₁₁₅ γ-TMD classes had equivalent density, and this extra step was unnecessary. Programs were compiled by SBGrid (Morin et al., 2013).

Model building and refinement

Statistics for each dataset are shown in Supplemental Table 1. The bicuculline-bound α1β2γ2 structure (PDB:6X3S) was used as a starting model for the GABA_A receptor. To generate starting models for the Fabs used in this study, amino acid sequences were submitted to the SwissModel (Waterhouse et al., 2018) and PIGS (Waterhouse et al., 2018) servers. The top hits for Fab₁₇₅ and Fab₁₁₅ were modeled on PDB IDs 4HH9 and 5WL2.1 (Bryson et al., 2016), respectively. These homology models were manually docked and then fitted into maps using UCSF Chimera (Pettersen et al., 2004). Real-space refinement was performed via iterations of local and global adjustment in Coot (Emsley et al., 2010) and Phenix (Liebschner et al., 2019) respectively. Model quality was assessed in MolProbity (Williams et al., 2018). For placement of important side chains such as E174 and K98 (of the receptor α1 and Fab₁₁₅ Light Chains, respectively), cryo-EM maps were high-pass filtered to 6.0 Å using Relion Image Handler (Gharpure et al., 2019). The pore profiles shown in Figure 4 were created using HOLE (Smart et al., 1996); hydrophobicity plots were created with CHAP (Klesse et al., 2019).

Analysis of Antibody Sequences to Germline

Variable antibody sequences were compared using custom BASE software (Reincke et al., 2020) to human immunoglobulin germline V(D)J gene segments to identify genes with highest sequence identity.

Electrophysiology

Whole cell voltage-clamp recordings were made from adherent HEK293S GnTI⁻ cells (ATCC CRL3022) transiently transfected with the tri-cistronic pEZT construct used for structural analysis. Each well of cells in a 12-well dish was transfected with 0.1–0.3 μg plasmid using Lipofectamine 2000 Reagent (Invitrogen). The transfected cells were incubated at 30°C. 24 h post-transfection, cells were re-plated on 35 mm dishes and allowed to settle for at least 2 hours. Recordings were made 24–96 hours after transfection. Bath solution contained (in mM): 140 NaCl, 2.4 KCl, 4 MgCl₂, 4 CaCl₂, 5 HEPES and 10 glucose pH 7.3. Borosilicate pipettes were pulled and polished to an initial resistance of 2–4 MΩ and filled with the pipette solution containing (in mM): 100 CsCl, 30 CsF, 10 NaCl, 10 EGTA, and 20 HEPES pH 7.3. The recordings were made with an Axopatch 200B amplifier, sampled at 5 kHz, and low-pass filtered at 2 kHz using a Digidata 1440A (Molecular Devices) and analyzed with pClamp 10 software (Molecular Devices). Cells were held at −75 mV. Solution exchange was achieved using a gravity driven RSC-200 rapid solution changer (Bio-Logic). For short application (60–80 s pre-application), Fab solutions were prepared in bath solution from a concentrated stock solution.

Quantification and Statistical Analysis

Peak currents in whole cell patch clamp electrophysiology experiments were measured in the absence or presence of antibody fragments. Statistical analysis was performed using GraphPad Prism 9.2.0 software (GraphPad software, Inc, La Jolla, CA). Data are expressed as means ± S.D, and Welch’s t-test was used. A p-value of ≤ 0.05 was considered statistically significant (**** p ≤ 0.0001; ***, p ≤ 0.001; *, 0.01 ≤ p ≤ 0.05) (Figure 5 and 6). Replicate numbers n = 7–9 for Figure 5 and n = 4–12 for Figure 6, from independent cells.

Supplementary Material

1

Supplemental Figure 1: Patient-derived antibodies bind to the α1β2γ2 GABA_A receptor and representative purification, related to Figure 1. A, Purified GABA_A receptor was reconstituted in nanodiscs in the absence or presence of each monoclonal antibody (mAb) indicated and analyzed by fluorescence-detection size exclusion chromatography (FSEC). A shift to the left indicates an increase in size associated with binding. B-E are screens of selected Fab fragments produced from the mAbs. Increasing amounts of Fab were added to GABA_A receptor in nanodiscs to determine at which point the elution volume shift is saturated. F, Immunofluorescence staining of selected GABA_A Fabs (green) on PFA-fixed murine brain tissue sections shown in the CA3 region of the hippocampus with DRAQ5 stained nuclei (red). Representative scale bar indicates 50 μm. G, FSEC traces of receptor alone (black trace) and receptor plus Fab₁₁₅ (blue trace). H, FSEC trace of the final material applied to EM grids; inset shows Coomassie-stained SDS-PAGE gel of the complex.

2

Supplemental Figure 2: Cryo-EM processing workflow for Fab₁₇₅ + GABA_A, related to Figures 1, 3. Each panel represents a different stage of the data collection and processing. A, Representative micrograph. Scale bar indicates 100 nm. B, 2D classification with selected classes boxed in red. C, First 3D classification with selected class indicated. D, First refinement. E, Second 3D classification using the refined angles from D to perform fine local angular searches. F, Second refinement. G, 3D classification on signal subtracted particles to focus alignment on the γ-TMD. H, Final refinement, sharpening, local resolution estimation and FSC Curve for the Fab₁₇₅+GABA_A dataset. I, as in H but for the Fab₁₁₅+GABA_A dataset.

3

Supplemental Figure 3: Quality of Cryo-EM maps at regions of interaction between Fab and receptor, related to Figures 2, 3. A, Overview of Fab₁₁₅:GABA_A complex map from view parallel to membrane. Panel D inset is indicated. B, As in panel A, but turned 180° to show other Fab₁₁₅:GABA_A interface. Inset indicated is shown in panel E. Mesh in panels D and E is the result of high-pass filtering the density maps to 6 Å in order to emphasize higher resolution features. C, Overview of Fab₁₁₅:GABA_A complex map from view of outside the cell looking in. Panel F inset is indicated, highlighting the interaction between the Fab CDR3 and neurotransmitter binding domain. G, Overview of Fab₁₇₅:GABA_A complex map from view parallel to membrane. Panel I inset is indicated. H, Overview of Fab₁₇₅:GABA_A complex map from view of outside the cell looking in. Panel J inset is indicated, highlighting CDR3 of the antibody heavy chain, which inserts into the ECD α1:γ2 interface.

4

Supplemental Figure 4: Overview of receptor:Fab models, related to Figures 1–4. A, Fab₁₁₅: GABA_A model. ECD, extracellular domain. TMD, transmembrane domain. B, Fab₁₁₅: HC, heavy chain; LC, light chain. CDR, complementarity determining region. The CDRs are colored in red-orange for contrast. C, D Same as for A, B but CDRs are colored green. E, F Subunit interface contacts with Fab. E, One β2:α1 subunit interface is shown; red box indicates neurotransmitter binding site and spheres indicate amino acids contacted by Fab₁₁₅. Inset table shows surface area buried by Fab heavy and light chains on respective GABA_A subunits. F, Same as for E but for Fab₁₇₅.

5

Supplemental Figure 5: Alignment of GABA_A receptor subunits contacted by autoantibodies, related to Figures 2–3. Amino acid sequence alignments of GABA_A subunits contacted by Fab₁₁₅ and Fab₁₇₅ are shown. On the right-hand side of the figure are cartoons indicating the location of the antibodies bound (yellow is Fab₁₁₅, red is Fab₁₇₅). Subunits shown in the alignment for each box are colored.

6

Supplemental Figure 6: Somatic hypermutations of Fab₁₁₅ and Fab₁₇₅ compared to receptor contacts, related to Figures 2–3. A, Alignments between sequences of Fab₁₁₅ and its presumed germline ancestor. Only the variable domain is shown. CDR1, CDR2, and CDR3 are indicated in bold for each antibody segment. Somatic hypermutations are indicated by noting the parental sequence in the germline. Residues of the V(D)J junctions are underlined. Note, these residues cannot reliably be differentiated from germline ancestors. Hence assignment of somatic hypermutations is limited to V and J genes only, and thus not for most part of the highly variable CDR3. Residues of the heavy and light chains that contact the GABA_A receptor are indicated and color coded based on subunit. The NG-KR mutations discussed in the text are indicated by asterisks (*). B, Same as in A but for Fab₁₇₅.

7

Supplemental Table 1: Cryo-EM Data Collection, Processing and Model Statistics, refers to Figures 1, S1, S2.

8

Supplemental Video 1: Overview and zooms of Fab₁₁₅:GABA_A map density, modeling, and quality, refers to Figures 1–2, S1–S3

9

Supplemental Video 2: Overview and zooms of Fab₁₇₅:GABA_A map density, modeling, and quality, refers to Figures 1, 3, S1–S3

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Viral Strains
DH5α	Thermo Fisher Scientific	Cat# C404003
DH10Bac	Thermo Fisher Scientific	Cat# 10361012
Chemicals, Peptides, and Recombinant Proteins
n-Dodecyl-β-D-Maltopyranoside (DDM)	Anatrace	Cat# D310
Cholesterol	Sigma-Aldrich	Cat# C8667
Porcine brain extract polar lipids	Avanti	Cat# 141101C
Fos-choline-8, fluorinated	Anatrace	Cat# F300F
Recombinant Fab101	InVivo BioTech, This Paper	This Paper
Recombinant Fab115	InVivo BioTech, (Kreye et al., 2021)	(Kreye et al., 2021)
Recombinant Fab175	InVivo BioTech, This Paper	This manuscript
Recombinant Fab201	InVivo BioTech, This Paper	This manuscript
Recombinant mAb101	(Kreye et al., 2021)	(Kreye et al., 2021)
Recombinant mAb115	(Kreye et al., 2021)	(Kreye et al., 2021)
Recombinant mAb175	(Kreye et al., 2021)	(Kreye et al., 2021)
Recombinant mAb198	(Kreye et al., 2021)	(Kreye et al., 2021)
Recombinant mAb201	(Kreye et al., 2021)	(Kreye et al., 2021)
d-Desthiobiotin	Sigma-Aldrich	Cat# D1411
Lipofectamine 2000	Thermo Fisher Scientific	Cat# 11668019
Freestyle 293 Expression Medium	Thermo Fisher Scientific	Cat# 12338-018
Saposin	(Lyons et al., 2017)	(Lyons et al., 2017)
Sf-900 III SFM	Thermo Fisher Scientific	Cat# 12658-027
DMEM Medium	Corning	Cat# 10-013-CV
PBS	Corning	Cat# 21-040-CV
Fetal Bovine Serum	EMD Millipore	Cat# TMS-013-B
Bovine Serum Albumin (IgG-free, Protease-free)	Dianova	Cat# 001-000-161
Normal Goat Serum	Abcam	Cat# ab138478
Phenylmethanesulfonyl fluoride	Sigma	Cat# PMSF-RO
Bio-beads SM-2 Adsorbent Media	Bio-Rad	Cat# 1523920
Murine Brain Tissue	(Kreye et al., 2016; Kreye et al., 2021)	(Kreye et al., 2016; Kreye et al., 2021)
Paraformaldehyde (PFA)	Carl Roth	Cat# 0335.1
Anti-human IgG (F(ab’)2-fragment-specific Antibody	Dianova	Cat# 109-545-097
DRAQ5	Thermo Fisher Scientific	Cat# 62254
Critical Commercial Assays
Superose 6 Increase 10/300 GL	GE Healthcare	Cat# 29091596
Strep-Tactin Superflow high capacity resin	IBA Life Sciences	Cat# 2-1208-500
Sepax SRT-500 5μM 4.6×300mm	Sepax	Cat# 215500-4630
Deposited Data
Coordinates of GABAA in complex with GABA	(Kim et al., 2020)	6X3Z
Coordinates of GABAA in complex with GABA	(Kim et al., 2020)	6X3S
Coordinates of GABAA in complex with Megabody 38	(Laverty et al., 2019)	6I53
Coordinates of GABAA in complex with GABA and picrotoxin	(Masiulis et al., 2019)	6HUJ
Coordinates of GABAA in complex with Fab175	This manuscript	7T0Z
Coordinates of GABAA in complex with Fab115	This manuscript	7T0W
Cryo-EM map of GABAA in complex with Fab175	This manuscript	EMD-25585
Cryo-EM map of GABAA in complex with Fab115	This manuscript	EMD-25583
Experimental Models: Cell Lines
HEK GnTI−	(Reeves et al., 2002)	CVCL_A785
Sf9	(Vaughn et al., 1977)	CVCL_0549
Recombinant DNA
pEZT-α1β2γ2 tricistronic GABAA receptor	(Kim et al., 2020)	(Kim et al., 2020)
Software and Algorithms
Serial EM	(Mastronarde, 2005)	http://bio3d.colorado.edu/SerialEM/
MotionCor2	(Zheng et al., 2017)	https://emcore.ucsf.edu/cryoem-software
GCTF	(Zhang, 2016)	https://www.mrc-lmb.cam.ac.uk/kzhang/Gctf/
CrYOLO	(Wagner et al., 2019)	https://cryolo.readthedocs.io/en/stable/
Relion 3.0.8	(Zivanov et al., 2018)	https://www3.mrc-lmb.cam.ac.uk/relion/index.php/Main_Page
UCSF Chimera	(Pettersen et al., 2004)	https://www.cgl.ucsf.edu/chimera/
Phenix	(Afonine et al., 2018)	https://www.phenix-online.org/
BASE Software	(Reincke et al., 2020)	https://github.com/automatedSequencing/BASE
Coot	(Emsley et al., 2010)	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Pymol	PyMOL	http://www.pymol.org
CHAP	(Klesse et al., 2019)	https://www.channotation.org/
HOLE	(Smart et al., 1996)	http://www.holeprogram.org/
pClamp 10	Molecular Devices	https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite
Prism 8	GraphPad	https://www.graphpad.com/scientific-software/prism/
Swissmodel	(Waterhouse et al., 2018)	https://swissmodel.expasy.org/
PIGS	(Waterhouse et al., 2018)	http://arianna.bio.uniroma1.it/pigs
MolProbity	(Williams et al., 2018)	http://molprobity.biochem.duke.edu/index.php
Acquire	Bruxton	https://www.bruxton.com/products.html
TAC 4.2.0	Bruxton	https://www.bruxton.com/products.html
Other
Quantifoil Holey Carbon Grids, 1.2/1.3 Cu 200 Mesh	Electron Microscopy Sciences	Q2100CR1.3

Highlights of “Structural mechanisms of GABA_A receptor autoimmune encephalitis”.

• GABA_A receptor structures with pathogenic antibodies from an encephalitis patient.

• Direct or allosteric receptor inhibition reflect clinical symptoms.

• Contacts include antibody residues that underwent somatic hypermutation.

• Electrophysiology supports structure-predicted inhibition mechanisms.