Altered Channel Conductance States and Gating of GABAA Receptors by a Pore Mutation Linked to Dravet Syndrome

Abstract We identified a de novo missense mutation, P302L, in the γ-aminobutyric acid type A (GABAA) receptor γ2 subunit gene GABRG2 in a patient with Dravet syndrome using targeted next-generation sequencing. The mutation was in the cytoplasmic portion of the transmembrane segment M2 of the γ2 subunit that faces the pore lumen. GABAA receptor α1 and β3 subunits were coexpressed with wild-type (wt) γ2L or mutant γ2L(P302L) subunits in HEK 293T cells and cultured mouse cortical neurons. We measured currents using whole-cell and single-channel patch clamp techniques, surface and total expression levels using surface biotinylation and Western blotting, and potential structural perturbations in mutant GABAA receptors using structural modeling. The γ2(P302L) subunit mutation produced an ∼90% reduction of whole-cell current by increasing macroscopic desensitization and reducing GABA potency, which resulted in a profound reduction of GABAA receptor-mediated miniature IPSCs (mIPSCs). The conductance of the receptor channel was reduced to 24% of control conductance by shifting the relative contribution of the conductance states from high- to low-conductance levels with only slight changes in receptor surface expression. Structural modeling of the GABAA receptor in the closed, open, and desensitized states showed that the mutation was positioned to slow activation, enhance desensitization, and shift channels to a low-conductance state by reshaping the hour-glass-like pore cavity during transitions between closed, open, and desensitized states. Our study revealed a novel γ2 subunit missense mutation (P302L) that has a novel pathogenic mechanism to cause defects in the conductance and gating of GABAA receptors, which results in hyperexcitability and contributes to the pathogenesis of the genetic epilepsy Dravet syndrome.


Introduction
Dravet syndrome (also known as severe myoclonic epilepsy in infancy, OMIM: 607208) is an epileptic encephalopathy of childhood that is characterized by multiple types of seizures that are often prolonged and particularly fever sensitive. With onset in the first year of life, de novo heterozygous missense mutations in SCN1A, which encodes the pore-forming ␣1 subunit of the sodium channel, are found in about 95% of patients with Dravet syndrome (Vadlamudi et al., 2010). Heterozygous nonsense mutations in the ␥-aminobutyric acid type A (GABA A ) receptor ␥2 subunit gene, GABRG2, and missense mutations in the GABA A receptor ␣1 subunit gene, GABRA1, are associated also with Dravet syndrome (Harkin et al., 2002;Carvill et al., 2014). The GABRG2 mutation, ␥2(Q40X), caused truncation of the predicted mature ␥2 subunit in the first amino acid, resulting in translation of no mature subunits and to activation of nonsense-mediated mRNA decay (NMD) that reduced truncated signal peptide production (Huang et al., 2012;Ishii et al., 2014). In contrast, a nonsense mutation in the last exon of GABRG2, ␥2(Q390X), which was located in the intracellular loop of the ␥2 subunit between transmembrane segments M3 and M4, produced mRNA that was stable and not degraded by nonsense-mediated mRNA decay, and truncated subunits that were degraded slowly and incompletely by endoplasmic reticulum-associated degradation (Kang et al., 2009). The heterozygous Gabrg2 ϩ/Q390X knock-in mouse had a severe epilepsy phenotype that included spontaneous generalized tonic-clonic seizures and sudden death after seizures and accumulation and aggregation of mutant ␥2(Q390X) subunits that may contribute to the chronic progressively declining clinical course of the epileptic encephalopathy (Kang et al., 2015).
GABA A receptor ␥2 subunits, which belong to the Cysloop ion channel superfamily, are formed by an ϳ200 residue N-terminal extracellular domain and four transmembrane segments, M1 to M4, that are homologous to the glutamate-gated chloride channel (GluCL) (Hibbs and Gouaux, 2011), and the glycine receptor (GlyR) ␣1 subunit (Du et al., 2015). Pentameric assembly of GABA A receptor ␣1, ␤3, and ␥2 subunits form the most abundant GABA A receptor subtype in the brain (Farrant and Nusser, 2005) that mediates the majority of fast inhibitory neurotransmission and controls network excitability in the brain. To date, only 8 missense mutations in the ␥2 subunit are associated with mild forms of epilepsy syndromes and are located mainly in the receptor N-terminal extracellular domain and in the outermost region of the M2 transmembrane segment (Baulac et al., 2001;Wallace et al., 2001;Audenaert et al., 2006;Shi et al., 2010;Lachance-Touchette et al., 2011;Carvill et al., 2013;Reinthaler et al., 2015). Using targeted next-generation sequencing, we identified a novel de novo GABRG2 missense mutation, P302L, in a patient with Dravet syndrome. Multiple sequence alignments among GABRG genes showed that this mutated proline was located in the cytoplasmic entrance of the transmembrane segment M2 that contributes to the formation of the pore region of the channel and was completely conserved in the three ␥ subunits in humans and other species (Fig. 1A). In vitro analysis showed that the ␥2(P302L) subunit mutation caused slight changes of surface expression, but mainly disrupted GABA A receptor gating by reduction of GABA-evoked singlechannel current amplitudes, slowing activation, enhancing desensitization, and decreasing single-channel open probability, all of which caused reduction of GABA A receptor-mediated mISPCs. We propose a structural mechanism that leads to dysfunction of this channel. Through modeling of the channel structure in the closed, open, and desensitized states that govern the gating of the receptor, we found that the mutation reshaped the pore of the GABA A receptor by altering the features of both outer and inner vestibules of the hour-glass-like pore cavity, and thus altering the conduction pathway. The findings presented here shed light into the molecular mechanisms of the pathogenesis of Dravet syndrome.

Patient, targeted next-generation, and Sanger sequencing
Genomic DNA was extracted from peripheral leukocytes from the patient (case 1007) and parents for segregation analysis. Written informed consent was obtained from all individual subjects of this study. The study was approved by the Peking University First Hospital Medical Ethics Committee. Custom-designed panels capturing the coding exons of GABRG2 were synthetized using the Agilent SureSelect Target Enrichment technique. Targeted next-generation sequencing was subsequently performed on an Illumina GAIIx platform (Illumina) using a paired-end sequencing of 110 bp to screen for mutations as described previously (Kong et al., 2015). We used Sanger sequencing to confirm the origin of the mutation as being de novo.

cDNA constructs and expression of recombinant GABA A receptors
cDNAs encoding human ␣1, ␤3, and ␥2L GABA A receptor subunit subtypes (NM000806, NM021912, and NM198904, respectively) were subcloned into the plasmid Figure 1. The de novo GABRG2(P302L) missense mutation was identified in a patient with Dravet syndrome. A, Sequence alignments of the transmembrane segment M2 of GABA A receptor ␥1-3 subunits with the D. rerio GlyR receptor ␣1 subunit, highlighting the evolutionary conservation of the proline (shown in red) at the -2' position. The residues in gray were conserved across all of the expression vector pcDNA3.1 (Thermo Fisher Scientific) using standard techniques. The ␥2L(P302L) subunit mutation was generated by site-directed mutagenesis using the QuikChange Site-Directed Mutagenesis kit (Agilent Technologies) and verified by sequencing. Human embryonic kidney cells (HEK293T) were grown in 100-mm tissue culture dishes (Corning) with DMEM, supplemented with 10% fetal bovine serum at 37&cenveo_unknown_entity_ timesnewroman_00B0;C in 5% CO 2 /95% air and passaged every 3-4 d. For surface biotinylation experiments, 4 ϫ 10 5 cells were plated onto 60-mm diameter culture dishes. Twenty-four hours after plating, cells were transfected using polyethyleneimine (PEI,MW 40,000;Polysciences). For mock or single subunit expression, empty pcDNA3.1 vector was added to make a final cDNA transfection amount to 1.8 g. For electrophysiology experiments, cells were plated onto 12-mm cover glass chips at 4 ϫ 10 4 in 35-mm diameter culture dishes, transfected after 24 h with 0.3-g cDNA of each ␣1, ␤3, and ␥2L subunits and 0.05 g of EGFP (to identify transfected cells) using X-tremeGENE9 DNA transfection Reagent (Roche Diagnostics; 1.5 l/g cDNA). Recordings were obtained 48 h after transfection.

Electrophysiology
Whole-cell recordings from lifted HEK293T cells and cell attached single-channel recordings were obtained as previously described (Hernandez et al., 2011;Janve et al., 2016). For whole-cell recordings the internal solution consisted of: 153 mM KCl, 10 mM HEPES, 5 mM EGTA, 2 mM Mg-ATP, and 1 mM MgCl 2 .6H 2 O; pH 7.3, ϳ300 mOsm. The external solution was composed of: 142 mM NaCl, 8 mM KCl, 10 D(ϩ)-glucose, 10 mM HEPES, 6 mM MgCl 2 .6H 2 O, and 1 mM CaCl 2 ; pH 7.4, ϳ326 mOsm. This combination of external (1Na-external) and internal solutions produced a chloride equilibrium reversal potential (V rev ) of ϳ0 mV, and cells were voltage clamped at -20 mV. For the ionic selectivity current-voltage (I/V) experiments, the 142 mM NaCl of the external recording medium was replaced with 71 mM NaCl (a 50% reduction of Na concentration), and 142 mM sucrose was added to maintain isoosmolar conditions. This external solution (0.5Na-external) produced a chloride V rev of ϳ13 mV. Liquid junction potentials were calculated using Clampex's Junction Potential Calculator and corrected by the Pipette Offset circuitry of the amplifier. The I/V experiments were performed by holding the cell membrane potential (in mV) at: -80, -60, -40, -20, 0, ϩ20, ϩ40, ϩ60, and 4 s GABA A receptor currents evoked by 1 mM GABA were recorded at each membrane potential. External solutions and drugs were gravity fed to a four-barrel square glass pipette connected to a SF-77B Perfusion Fast-Step system (Warner Instruments). The solution exchange time across the open electrode tip was ϳ200-400 s, and the exchange around lifted cells (ϳ8-12 pF) occurred within 800 s, which was sufficiently fast for these experiments (Bianchi and Macdonald, 2002) and guaranteed rapid solution exchanges and accurate measurement of the kinetic properties of the receptor currents. Although series resistance errors were not compensated in this study, we rule out the possibility of underestimating the "true" peak amplitude and desensitization kinetics by using low resistance electrodes (ϳ1 M⍀) and cells that showed current amplitudes ϳ5 nA (ϳ8-12 pF) at Ϫ20 mV holding potential as previously reported (Bianchi and Macdonald, 2002). Single-channel currents were recorded in an external solution containing: 140 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 10 mM glucose, and 10 mM HEPES; pH 7.4. During recording, 1 mM GABA was added to the internal solution that consisted of: 120 mM NaCl, 5 mM KCl, 10 mM MgCl 2 , 0.1 mM CaCl 2 , 10 mM glucose, and 10 mM HEPES; pH 7.4. The intramicropipette potential was ϩ80 mV. ␣␤-GABA A receptor currents were blocked by adding 100 M zinc. Single-channel conductance in the presence of 1 mM GABA and 100 M zinc was determined by holding the transmembrane potential at ϩ80 mV. All experiments were performed at room temperature (22-23°C).
Macroscopic activation and deactivation current times () were measured by application of 1 mM GABA for 10 ms, while desensitization and peak current amplitude by application of 1 mM GABA for 4 s. Activation, desensiti-zation and deactivation current time () courses were fitted using the Levenberg-Marquardt least squares method with up to four component exponential functions of the form Αa nexp(-t/ n ) ϩ C, where n is the number of the exponential components, t is time, a is the relative amplitude, n is the time constant, and C is the residual current at the end of the GABA application. Additional components were accepted only if they significantly improved the fit, as determined by an F test on the sum of squared residuals. The time course of deactivation was summarized as a weighted time constant, defined by the following expression: Αa n n /Αa n . The extent of desensitization was measured as (fitted peak current -fitted steady-state current)/ (fitted peak current).
GABA peak current-voltage plots were fitted to a first order polynomial function, and V rev values were read directly from the fitted current-voltage plots for each cell. Peak current responses obtained with 0.5Na solutions and recorded with holding potentials from -80 to -40 mV were excluded from the fitting analysis as they showed outward rectification. GABA A receptor current concentrationresponse curves were fitted using GraphPad Prism version 6.07 for Windows (GraphPad Software). We used a nonlinear regression Hill equation of the form E ϭ E basal ϩ (E MAX -E basal )/(1 ϩ 10^((LogEC 50 -X)‫ء‬Hill slope)), where E is the fractional response of the GABAR-gated currents, E max is the maximal response, X is [GABA], EC 50 is the [GABA] at which response ϭ 50% of maximal response, and the Hill slope (nH). Thus, the nH for wt and ␥2(P302L) GABA A receptors were 1.17 Ϯ 0.25 and 2.60 Ϯ 0.88, respectively. Inhibition of 1 mM GABA A receptor evoked currents by 10 M zinc was measured by preapplication for 10 s followed by coapplication with GABA for 4 s. GABA and zinc were obtained from Sigma.
Single-channel open and closed events were analyzed using the 50% threshold detection method and visually inspected before accepting the events. Single-channel openings occurred as bursts of one or more openings or cluster of bursts. Bursts were defined as one or more consecutive openings that were separated by closed times that were shorter than a specified critical duration (t crit ) prior to and following the openings (Twyman et al., 1990). A t crit duration of 5 ms was used in the current study. Clusters were defined as a series of bursts preceded and followed by closed intervals longer than a specific critical duration (t cluster ). A t cluster of 10 ms was used in this study. Open and closed time histograms as well as amplitude histograms were generated using TACFit 4.2 (Bruxton). Single-channel amplitudes (i) were calculated by fitting all-point histograms with single-or multi-Gaussian curves. The difference between the fitted "closed" and "open" peaks was taken as i. Duration histograms were fitted with exponential components in the form: Α(A i / i )exp(-t/ i ), where A and represent the relative area and time constant of the I component, respectively, and t is the time. The mean open time was then calculated as follows: ΑA i i . The number of components required to fit the duration histograms was increased until an additional component did not significantly improve the fit (Fisher and Macdonald, 1997).
Single-channel current-voltage plots were fitted to a linear regression function and conductance values were extracted from the slopes of the best-fit line using GraphPad Prism version 6.07 for Windows (GraphPad Software).

Primary cortical neuron culture
Mouse cortical neurons were obtained from embryonic day 17.5 mouse pups (four to eight) of either sex. C57BL/6 mice ages 2-4 months were used in schedule breeding 17 days prior to the day of neuron isolation . Two female mice were used after confirmation of pregnancy by detection of a vaginal plug. Briefly, dissociated cells were plated at a density of 6.5 ϫ 10 4 cells/ cm2 onto 12-mm round coverslips in 24-well plates coated with poly-L-ornithine (0.5 mg/ml; Sigma). Cortical neurons were incubated at 37°C in 5% CO 2 incubator and maintained in serum-free Neurobasal medium (Gibco) supplemented with B27 supplement (Gibco), glutamine (Gibco), and penicillin/streptomycin (Gibco, 20 U/ml). Cultured neurons were transfected at DIV12 with 2 g of ␥2L or ␥2L(P302L) and 0.5 g of EGFP using X-tremeGENE9 DNA transfection reagent. Recording were obtained at DIV 19-21. All animal procedures were performed in accordance with Vanderbilt University Medical Center animal care committee's regulations.
Analysis of GABA A receptor-mediated mIPSCs were obtained using the Clampfit 10.4 data analysis module. Detection of mIPSCs was determined from all automatically detected events in a given 200-s recording period. For kinetic analysis, the mIPSCs were automatically detected by the program initially and then manually analyzed based on the criteria that only single-event mIPSCs with a stable baseline, rising phase (10-90% rise time), and exponential decay were chosen during visual inspection of the recording trace. Double-and multiple-peak mIPSCs were excluded. Events of low amplitude (Ͻ15 pA) were discarded from this analysis. For each neuron, 1000 individual mIPSC events were recorded. Cumulative histograms of the peak currents were compared using a Kolmogorov-Smirnov (K-S) test with a significance value of p Ͻ 0.05.

Structural modeling, simulation, and channel pore characterization
Three-dimensional structural models of the GABA A receptor in the open, closed, and desensitized states were modelled using the electron cryo-microscopy of the Danio rerio GlyR subunit ␣1 structures (Du et al., 2015) in the open (3JAE), closed (3JAD), and desensitized (3JAF) conformations as templates. GABA A receptor ␣1, ␤3, and ␥2 subunit raw sequences in FASTA format were uploaded to the Swiss-PdbViewer 4.10 server (Schwede et al., 2003) for template searching against the ExPDB database (ExPASy, http://www.expasy.org/). The initial sequence alignments between GABA A receptor ␣1, ␤3, and ␥2 subunits and D. rerio GlyR ␣1 subunits in the open (3JAE), closed (3JAD), and desensitized (3JAF) states were generated with full-length multiple alignments using ClustalW. Sequence alignments were inspected manually to assure accuracy among structural domains solved from the template. Because the long M3/M4 cytoplasmic loop of the GABA A receptor subunits was absent in the solved GlyR structures, the corresponding fourth transmembrane segments (M4) were misaligned onto the template. Consequently, the M3/M4 cytoplasmic loop was excluded from the modelling, and separate alignments were generated for the transmembrane segments M4. Then full-length multiple alignments were submitted for automated comparative protein modelling implemented in SWISS-MODEL (http://swissmodel.expasy.org/SWISS-MODEL. html). Before energy minimization, resulting 3D models of human GABA A receptor ␣1, ␤3, and ␥2 subunits in the three conformational states were inspected manually, their structural alignments confirmed, and evaluated for proper h-bonds, presence of clashes, and missing atoms using Molegro Molecular Viewer (CLC bio). Pentameric 3D GABA A receptor models were generated by combining ␣1, ␤3, and ␥2 structural models in the stoichiometry 2␤:2␣:1␥ with the subunit arrangement ␤3-␣1-␤3-␣1-␥2 in a counterclockwise order by superposition onto the D. rerio GlyR receptor in the open (3JAE), closed (3JAD), and desensitized (3JAF) conformation states. Neighborhood structural conformational changes within a radius of 6 Å of the mutated residue P302L in the ␥2 subunit in the 3D structural models of the GABA A receptor in the open, closed, and desensitized conformation states were modelled using Rosetta 3.1 (Smith and Kortemme, 2008) (https://kortemmelab.ucsf.edu). Up to 20 of the bestscoring structures were generated at each time by choosing parameters recommended by the application. Root mean squared (RMS) deviation was calculated between the initial (wt) structures and superimposed modelled (mutated) structures. For each 3D GABA A receptor conformational state, the average RMS deviation over 10 low-energy structures was computed and conformational changes displayed among neighborhood structural domains. The detection and pore shape and visualization of the 3D structural models of the wt and mutated (P302L) GABA A receptor in the open, closed, and desensitized conformation states were determined using PoreWalker (Pellegrini-Calace et al., 2009), a computational automated method available as a web-based resource (http:// www.ebi.ac.uk/thornton-srv/software/PoreWalker/). From the PoreWalker outputs, a list of the identified pore-lining residues with their ␤-carbon coordinates along the pore axis (i.e., x-coordinates) and the given pore diameter profile of a channel were used to plot of pore diameters as a function of distance along the pore axis. In addition, we implemented ChExVis (Masood et al., 2015) (http:// vgl.serc.iisc.ernet.in/chexvis/) to determine the pore radius profile and the list of the identified pore-lining residues of the transmembrane pore of the 3D structural models of the wt and the P302L in the open conformation state. The models were rendered using UCSF Chimera version 1. 10 (Pettersen et al., 2004).

Statistical analysis
Numerical data were expressed as mean Ϯ SEM. Statistical analysis was performed using GraphPad Prism version 6.07 for Windows (GraphPad Software). Statistical significance was taken as p Ͻ 0.05, using unpaired two-tailed Student's t test, one-way ANOVA with Dunnett's and Tukey's multiple comparisons test, or two-way ANOVA with Sidak's multiple comparisons test as appropriate.

Results
A de novo mutation in GABRG2 that encodes the GABA A receptor ␥2 subunit was associated with Dravet syndrome Using a designed targeted sequencing panel, we identified a heterozygous de novo missense mutation in GABRG2, c.905CϾT (encoding p.Pro302Leu), which affected the highly conserved transmembrane segment M2 that lines the pore domain of the GABA A receptor ␥2 subunit (Fig. 1A,B). The affected male (patient 1007) carrying the de novo missense ␥2 subunit mutation c.905CϾT; p.Pro302Leu ( Fig. 1B) was diagnosed with Dravet syndrome. The phenotype included early psychomotor and language developmental delay and multiple seizure types (partial seizures, generalized tonic-clonic seizures, atypical absence seizures, and myoclonus) that were often prolonged. Neither parent had cognitive impairment or epilepsy. During his first year of life, he had seizures with febrile episodes, and his EEG was normal. However, he had frequent afebrile seizures after his first year. When he was 4 years old, the patient still had frequent atypical absence seizures with EEG showing bursts of bilateral 2to 3-Hz spike waves and multispike and slow waves (Fig.  1C). The patient was treated with more than five antiepileptic drugs, which did not control the seizures. Therefore, his epilepsy syndrome was consistent with a diagnosis of Dravet syndrome.
The GABA A receptor ␥2(P302L) subunit mutation substantially reduced GABA-evoked currents, enhanced their desensitization, and reduced GABA potency To gain insight into the molecular mechanisms underlying the epilepsy syndrome, we determined how GABA A receptor function was affected by the presence of the ␥2(P302L) subunit mutation. Whole-cell currents were evoked from lifted HEK293T cells cotransfected with ␣1 and ␤3 subunits and wt ␥2L or mutant ␥2L(P302L) subunits by applying a saturating GABA concentration (1 mM) for 4 s and 10 ms using a rapid exchange system (Fig. 2). Peak GABA-evoked current amplitudes recorded from cells coexpressing ␣1␤3␥2L(P302L) subunits were reduced to ϳ10% of those from wt ␣1␤3␥2L receptors (p Ͻ 0.0001; Fig. 2A, Table 1) .
To further determine whether the ␥2L(P302L) subunit mutation altered the gating of ␥2L(P302L) subunit-containing GABA A receptors, we examined the desensitization, activation, and deactivation rates of macroscopic whole-cell currents, properties that represent transitions among the closed, open, and desensitized states of the receptor. Currents from cells coexpressing ␣1␤3␥2L(P302L) subunits with ␣1 and ␤3 subunits were strongly desensitized (ϳ80%; Fig. 2B,C), slowly activated (ϳ2 times slower than wt receptor currents; Fig. 2D), and rapidly deactivated (ϳ2 times faster than wt receptor currents; Fig. 2E, Table 1). Moreover, while wt currents desensitized with three exponential components (1, 2, 3) as reported before (Bianchi et al., 2007), ␥2L(P302L) currents desensitized with Figure 2. The de novo ␥2(P302L) subunit mutation reduces GABA-activated currents and enhances desensitization. A, B, Representative GABA evoked-current traces obtained following rapid application of 1 mM GABA for 4 s to lifted HEK293T cells expressing wt ␥2L and mutant ␥2L(P302L) subunit-containing ␣1␤3␥2L GABA A receptors. Traces on the right (B) were normalized to illustrate the differences in desensitization between wt and mutant receptor currents. Bar graphs summarized the effects of wt and mutant GABA A receptors on peak current amplitudes and extent and weighted of desensitization. C, Bar graphs summarized the effects of wt and mutant GABA A receptors on desensitization time constants (1, 2, and 3) and relative areas (a1, a2, and a3). D, E, Representative current traces showed activation (D) and deactivation (E) obtained following rapid application of 1 mM GABA for 10 ms to cells coexpressing ␣1 and ␤3 subunits with wt ␥2L or mutant ␥2L(P302L) subunits. The traces were normalized for clarity. Bar plots summarized the differences in activation and deactivation between wt and mutant receptors. F, GABA A receptor concentrationresponse curves for wt ␣␤␥ (solid blue line), ␣␤ (dashed green lines), and mutant ␥2L(P302L) subunits (dashed red lines) were obtained. The top left inside panel was a comparison between currents from receptors containing wt ␣␤ and mutant ␣␤␥2L(P302L) subunits. Values were expressed as mean Ϯ SEM (n ϭ 5-6 cells for each experimental condition). The data represented the summary of 17 cells with comparable capacitances (8-12 pF) recorded from three independent transfections. Values were expressed as mean Ϯ SEM. Unpaired two-tailed Student's t test relative to wt ␥2L. ‫‪p‬ءءءء‬ Ͻ 0.0001, ‫‪p‬ءء‬ Ͻ 0.01, respectively. See Table 1 for details. New Research only two exponential components (1 and 2; Fig. 2C). The first desensitization time constant (1) was not different from wt, but its relative contribution increased from 6 to 72% (p Ͻ 0.0001; Table 1), which may account for the acceleration of the desensitization weighted . Interestingly, the second time constant (2) of the mutant receptor was not different from the longest wt time constant (3) (p Ͻ 0.2103; Table 1), which suggested the mutant lacks the second or intermediate time constant (2) (p Ͻ 0.0020; Table 1).
To determine whether the changes observed in receptor gating altered GABA A receptor efficacy and/or potency, we measured the effects of mutant ␥2L(P302L) subunits on ␣1␤3␥2L GABA A receptor concentration-response curves (Fig. 2F). Thus, macroscopic peak currents were evoked by applying various concentrations of GABA for 4 s to wt and mutant receptors. For wt receptors, the EC 50 for current stimulation was 7.50 Ϯ 0.77 M, and the maximal current was 7038 Ϯ 302 pA (n ϭ 5-6). The ␥2L(P302L) subunit mutation caused an ϳ6-fold right shift of the EC 50 (50.26 Ϯ 0.82 M), with a substantial reduction of peak current to ϳ13% of the maximal response to GABA (931.2 Ϯ 57.6 pA, n ϭ 5-6). When comparing the maximal response to GABA between mutant ternary ␣1␤3␥2L receptors (see above) and wt binary ␣1␤3 receptors (1204 Ϯ 235 pA, n ϭ 5-6), their efficacy was similar, but binary ␣1␤3 receptor potency (see above) was 10-fold higher than mutant ternary ␣1␤3␥2L receptors (5.20 Ϯ 0.47 M), which was comparable with previous results (Wooltorton et al., 1997). These results suggest that, although the ␥2L(P302L) subunit mutation conferred characteristics to ternary receptors that resembled those of binary receptors, their kinetic defects were distinguishable.
cannot totally exclude the possibility of the presence of a small proportion of ␣1␤3 receptors, these results suggested that the mutated ␥2L(P302L) subunit was effectively assembled in the receptor and, once in the surface membrane, drastically reshaped the kinetic behavior of the receptor.
Three-dimensional structural modeling of pentameric ␣1␤3␥2 GABA A receptors showed that the ␥2(P302L) subunit mutation occurred where the desensitization-closed gate of Cys-loop receptors resides Three-dimensional structural modeling of the pentameric ␣1␤3␥2 GABA A receptor showed that the ␥2(P302L) subunit mutation occurred at the -2' position of the transmembrane segment M2 (Figs. 1A and 4B ), which was at the intracellular entrance of the pore (Fig. 4A). Recent x-ray structures of the GluCL (Hibbs and Gouaux, 2011) and the electron cryo-microscopy of the GlyR (Du et al., 2015) demonstrated that the desensitization-gate of Cysloop receptors resides at the intracellular entrance of the pore (Gielen et al., 2015), where the -2'P302 mutation was located, which represented the second site of constriction within the pore. These findings suggest a molecular mechanism by which the ␥2(P302L) subunit mutation mainly reduced GABA-evoked currents and induced fast desensitization of GABA A receptors.
Considering that residues lining the transmembrane segment M2 of the receptor contribute to gating transitions between closed and open states, we propose that the mutant ␥2(P302L) subunit might have different impacts on receptor structure in the closed, open, and desensitized states. Taking advantage of the electron cryo-microscopy of the D. rerio GlyR structures (Du et al., 2015) in the open (3JAE), closed (3JAD), and desensitized (3JAF) conformational states, we modelled wt ␣1␤3␥2 and mutant ␣1␤3␥2(P302L) receptors in the open (Fig.  5A,D), closed (Fig. 5B,E), and desensitized (Fig. 5C,F) states (see Materials and Methods for details). As described for the GlyR (Du et al., 2015), the ␣1␤3␥2 GABA A receptor pore exposed two sites of constriction, one at position 9' (Fig. 5A-C, top panels), which corresponded to the activation/open gate, and the other in the cytoplasmic interface at position -2' (Fig. 5A-C, lower panels), which represented the desensitization gate (Gielen et al., 2015). The ␣1␤3␥2 GABA A receptor modeling showed that these two gates were open or closed depending on the conformational state of the receptor. Structural modeling of ␣1␤3␥2(P302L) receptors in the open (Fig. 5D), closed (Fig. 5E), and desensitized (Fig. 5F) states showed that the ␥2(P302L) subunit mutation caused rearrangements of the side chain residues confined to the transmembrane segment M2 of the ␥2 subunit and propagated to the ␣1 Figure 3. The ␥2(P302L) subunit mutation causes slight changes in surface levels of GABA A receptor subunits. A, Wild-type (wt) ␥2L or mutant (mt) ␥2L(P302L) subunits were coexpressed with ␣1␤3 subunits in HEK293T cells. Surface protein samples were biotinylated, isolated, separated by SDS-PAGE, and probed by anti-GABA A receptor subunit and anti-ATPase antibodies. B, Band intensities of the ␥2, ␤3, and ␣1 subunits were normalized to the ATPase signal, and the summarized data were shown in the bar graphs. C, Wild-type (wt) or mutant (mt) ␥2L(P302L) subunits were cotransfected with ␣1␤3 subunits into HEK293T cells, and total cell lysates were collected, analyzed by SDS-PAGE, and blotted by anti-␣1, anti-␤3, or anti-␥2 subunit and anti-ATPase antibodies. D, Band intensity of ␥2L subunits was normalized to the ATPase signal. Values were expressed as mean Ϯ SEM. An unpaired t test was used to determine significance. and ␤3 neighboring subunits. We computed that the rearrangements of the subunit's secondary structure between the wt and mutant structural modeling had RMS deviation values Ն0.5 Å, which were shown in rainbow colors in Figure 5D-F. In addition, the ␥2(P302L) subunit modeling showed a repositioning of the side chains of the pore-lining residues of neighboring ␤3 and ␣1 subunits in between -2' and 9' positions ( Fig. 5G-I), where the closeddesensitized and open gates are located.
The rearrangements observed in the side chains of the pore-lining residues were dependent on the conformational state of the receptor. In addition, the residues were perturbed differently depending on their localization within the pentameric receptor when taking into account its counterclockwise arrangement as ␤3(chain A)-␣1(chain B)-␤3(chain C)-␣1(chain D)-␥2(chain E) subunits (chains refer to the subunits in the pentameric structure). Thus, in the open state (Fig. 5G), the perturbations were located mainly below the 9' position, perturbing pore-lining residues in the M2 helices of ␣1D-␥2E-␤3A neighbor subunits (␣1V279-V287, ␥2A300-T310, ␤3A271-L278), and in the M3 helix of ␥2(L350-V360). In contrast, in the closed and desensitized states, all five ␤3A-␣1B-␤3C-␣1D-␥2E subunits were altered (␤3A and ␤3C, I267-T281; ␣1B and ␣1D, S278-F285; ␥2E, S293-T310), compromising a greater number of residues but mainly located in M2 helices (Fig. 5H,I, only showing ␣1D-␥2E-␤3A neighbor subunits for clarity). The latter may be due to the predicted projection of L302 towards the channel pore, which was absent in the open state ( Fig. 5G-I, lower panels). It is noteworthy that ␥2T310 is a residue that was perturbed independent of the conformational state of the receptor, whereas ␤3T281 was solely perturbed in the closed and desensitized states. These threonines correspond to the 6' position in the M2 helix that outlines the channel pore. Highly conserved through all GABA A subunits, T310 and T281 are located just below the gate of the channel (L9' position) and shape the inner mouth of the channel pore.

The pore mutation P302L destabilized the channel gate in the open state
To further investigate whether the ␥2(P302L) subunit mutation altered the ion channel pore, we determined the structural characteristics of the channel cavity and the pore-lining residues along the axis of the channel pore through the implementation of two computational fullyautomatic methods (Pellegrini-Calace et al., 2009;Masood et al., 2015). Six views of the transmembrane domain (M1 to M4) of the GABA A receptor structure ( Fig. 6A-C, top panels), which represent approximately 32 Å along the pore axis (i.e., xz-plane section), show the variation of pore diameters at 3-Å steps among wt and mutant structures in the open, closed, and desensitized states (Fig.  6D). In these pore visualization views, the top of the pore axis corresponds to the outermost part of the pore (ϳ20' position), which is the extracellular vestibule of an asymmetric pore divided by the narrow open gate (9' position). On the other hand, the bottom of the pore axis corresponds to the innermost part of the pore (-2' position), which is the intracellular vestibule of the pore. As previously reported for GlyR (Du et al., 2015) and GluCL channels (Hibbs and Gouaux, 2011; Althoff et al., 2014), these A, A 3D structural model of the ␣1␤3␥2 GABA A receptor was displayed with the ␤ subunits in red, ␣ subunits in blue, and the ␥ subunit in gray. The ␥2(P302L) subunit mutation was mapped onto the structure and represented in orange. The dashed box represented the transmembrane domain of the receptor and transmembrane segments M1 to M4 were labeled in the ␥2 subunit. B, The transmembrane domain of 3D structural model of the ␣1␤3␥2 GABA A receptor, with residues at positions -2= (dashed circle) in each transmembrane segment M2 was displayed as spheres and colored by subunit (␥2 in orange, and ␣1 and ␤3 in black). Subunits ␣1 (blue), ␤3 (red), and ␥2 (gray) were labeled, and transmembrane segments M1 to M4 were labeled in the ␥2 subunit. A 3D model was viewed from the extracellular side as shown in the lower left corner (dashed box); for clarity, the N-terminal extracellular domain was not shown. , and desensitized (C) conformational states were viewed from the extracellular side and displayed the ␤ subunits in red, ␣ subunits in blue, and ␥ subunit in gray. Side chains of the pore-lining residues at 9'(in black) and -2' (P302 in orange, other residues in black) positions were shown within the ion channel pore of the receptor. D-F, Superimposed 10 best-scoring 3D transmembrane domains of GABA A receptors in the open (D), closed (E), and desensitized (F) states modelled between the initial wt and P302L mutated structures were viewed from the cytoplasmic side. The modelled structures were in stick representation. The wt GABA A receptor structure was in gray. The structural rearrangements in side chain residues that differ among the wt and the mutated structures (RMS Ն 0.5 Å) were represented as a different color. G-I, observations support the concept that the shape of the pore of the GABA A receptor also shows an asymmetrical hour-glass-like cavity. Considering the centers of the pore at 3-Å steps along the pore axis ( Fig. 6A-C, red spheres), and their sizes being proportional to the pore diameter measured at that point, the pore diameter profile shows a great variability of both outer and inner vestibules among the structures (Fig. 6D). Taking into account the presence of these two vestibules, we compared the diameters of the cavity between the outermost position (21'-18' position) and the open gate (9' position), which represent the outer vestibule, and the latter and the innermost position (-2') that represents the inner vestibule, between wt and mutant structures for each conformational state. Figure  6D shows the differences of the pore diameter profiles along the pore axis of both outer and inner vestibules. Comparisons made against the wt structure (solid lines) revealed that although in the open state the mutated structure narrowed the outer vestibule, but increased the inner vestibule, in the closed state, the mutated structure narrowed the inner vestibule whereas increased the outer vestibule (dashed lines). Conversely, in the desensitized state, the mutated structure solely increased the inner vestibule. Moreover, cross-sectional views of the pore at the open gate (9' position) ( Fig. 6A-D, bottom panels) showed that in both open and closed states the open gate was reduced, whereas in the desensitized state, it was increased.
These data suggested that in the open state, while the channel gate was constricted, the desensitized gate was enlarged. By analyzing the radius along the channel pore ( Fig. 7A,C), the pore radius at the 9' position was reduced to 5.18 Å in the P302L model in comparison with the wt structure model (5.28 Å). These subtle differences were accompanied by the rearrangement of two residues, not present in the wt structure, as outlining the channel pore. Both residues of the ␤3 subunit, H292 (17' position) and A273 (-2' position), were located at the outer and inner entrances of the pore, respectively. Although GABA A receptors were formed by arrangement of ␣, ␤, and ␥ subunits, small differences in the entrances of the pore caused by a single substitution in the ␥2 subunit might account for differences in how the ions pass through the channel. These observations indeed support our hypothesis that the occurrence of mutations at the intracellular entrance of the pore is relevant to the structural mecha-continued Superimposed 10-best-scoring transmembrane domains of P302L structures in the open (G), closed (H), and desensitized (I) states were seen parallel to the membrane (top panels) and from the cytoplasmic side (bottom panels). Two subunits were removed for clarity. Perturbed neighborhood side chains within of 10 Å of L302 at the -2' position were shown within the ion channel pore, and the structural perturbations that differ among the wt and mutated structures (RMS Ն 0.5 Å) were represented in different colors. The wt-␥2 subunit structure was in gray. The channel gate at the 9' position was represented as a dashed line in the top panels. The ␤ and ␣ subunits were in red and blue, respectively. In the bottom panels, dashed black circles represented the channel pore, and the location of the L302 residue was shown in red dashed circles. Lists of the residues perturbed by the L302 mutation were detailed in the text. , and desensitized (C) states were shown parallel to the membrane (top panels) and from the extracellular side (bottom panels) of wt and P302L structures. In the top panels, the section of each structure was obtained by cutting the protein structure along the pore-axis. Red spheres represented pore centers at given pore heights and their diameters correspond to 1/10 of the pore diameter calculated at that point. Bottom panels represent transverse sections of the pore-axis at the 9' position (dashed lines in top panels), where the channel gate was located. Pore-lining side chains and residues at 3-Å steps within the cavity were colored in orange and blue, respectively. The remaining part of the transmembrane domain was shown in green. D, Pore diameter profiles at 3-Å steps corresponding to the pore-lining residues in the open (A), closed (B), and desensitized (C) states of wt and P302L structures. Horizontal black dotted lines show positions of pore-lining residues in M2 (see Figure 1A).  . The pore mutation P302L had no effect on GABA A receptor anion selectivity. A, C, Two-dimensional representations of channel radius (in Å) along the pore-axis for both P302L (A) and wt (C) structures in the open state, which was implemented by ChExVis. The analysis showed the pore-lining residues (colored boxes) numbered according to protein sequence and position in M2 of pentameric 3D GABA A receptor structures with the subunit arrangement ␤3(chain A)-␣1(chain B)-␤3(chain C)-␣1(chain D)-␥2(chain E). Above, alignments of the transmembrane M2 of GABA A receptor ␥2, ␣1, and ␤3 subunits showed the location and the conservation of the residues within M2 [identical ‫,)ء(‬ conservative (:), semiconservative (.)]. The residues were in concordance with the boxes below the 2D views. In red were common pore-lining residues that formed part of the inner face of the cavity predicted for both wt and P302L structures, while in blue were those only predicted for the P302L structure. In green, it was the site of the substitution. Below the 2D views, boxes were labeled and colored based on amino acid type as follows: E, glutamic acid (orange); T, threonine (light green); S, serine (yellow); A, alanine (red); I, isoleucine (rose); L, leucine (purple); V, valine (dark green); P, proline (brown); K, lysine (red orange); H, histidine (blue). The suffix after the residue number indicates the subunit to which the residue belongs. The dotted lines connecting the boxes represented the position of the residues along the pore-axis, as indicated on the top of the panels. B, D, Current-voltage (I/V) plots and corresponding GABA-gated peak current responses of mutant ␥2L(P302L) (C) and wt (D) ␣1␤3␥2L GABA A receptors obtained from membrane potential range from -80 mV to ϩ60 mV following rapid application of 1 mM GABA for 4s. The traces represented representative individual cells recorded in control (1 Na) or dilute (0.5 Na) solutions to determine the relative permeability of anions of mutant P302L and wt receptors. nisms that govern the gating of GABA A receptors, as discussed in the next sections.

The pore mutation P302L at the -2' position did not affect GABA A receptor anion selectivity
It is assumed that the movement of ions through the channel pore obeys the physical-chemical transport properties into barriers and wells (Bormann et al., 1987). Similar to other channels, GABA A receptors have wide extracellular and intracellular entrances that lead to the channel gate. Thus, amino acid substitutions at these entrances, such as that caused by P302L, may cause a loss of the selectivity of ions entering the channel as discussed above. To gain insights into whether the predicted perturbations at the entrances of the channel pore could result in loss of selectivity to anions, GABA-gated peak current responses of ␣1␤3␥2L(P302L) and wt ␣1␤3␥2L GABA A receptors at membrane potentials from -80 to ϩ60 mV were measured in physiological (1Na) and diluted (0.5Na) extracellular solutions (Fig. 7B,D). In our experimental conditions, at the extracellular NaCl concentration of 142 mM (1Na), the theoretical V rev for chloride given by the Goldman-Hodgkin-Katz equation is approximately -1.63 mV. The dilution of NaCl to 72 mM (0.5Na) predicts a relative right shift of the zero chloride current to V rev approximately ϩ13.04. Figure 7 shows representative I/V plots recorded from four different cells expressing receptors containing ␥2L(P302L) subunits (Fig. 7B) and wt (Fig. 7D) receptors in physiological (P302L, 0.62 Ϯ 1.07 mV, n ϭ 3; wt, 0.02 Ϯ 1.26 mV, n ϭ 5; p ϭ 0.7560, unpaired two-tailed Student's t test) and diluted NaCl concentrations. Noticeably, I/V plots for P302L mutant and wt receptor currents showed a significant rightward shift in V rev (p Ͻ 0.0001, one-way ANOVA with Dunnett's and Tukey's multiple comparisons test) with decreasing extracellular NaCl (P302L, 14.73 Ϯ 0.52 mV, n ϭ 7; wt, 13.99 Ϯ 1.12 mV, n ϭ 5; p ϭ 0.5222), which demonstrated that both currents were carried by chloride ions and remained anion selective.

Single-channel gating properties of GABA A receptors were impaired by the ␥2(P302L) subunit mutation
Our studies showed that depending on the state of the channels, whether they were in the closed or open states, the occurrence of the ␥2(P302L) subunit mutation in the channel pore increased or decreased, in an asymmetrical manner, both outer and inner vestibules. Conversely, in the desensitized state, the mutation produced only an increase of the diameter of the inner vestibule of the pore, which widened the cytoplasmic side of the pore. In the open state, a decrease in the outer vestibule of the pore can affect the anticlockwise rotation of the transmembrane M2 segments for channel activation (Hibbs and Gouaux, 2011;Althoff et al., 2014;Du et al., 2015). In this regard, our data showed that the ␤3(H292) residue was among the pore-residues outlining the inner face of the cavity in the P302L model (Fig. 7A). It seems that H292, mapped at the ␥2ϩ/␤3-interface in the wt model, is accessible to the pore in the P302L model, which might predicted by a clockwise tilt towards the channel pore, perturbing the anticlockwise rotation during channel acti-vation. In addition, the narrowing of the inner vestibule of the pore in the closed state suggested that the channel was transiently trapped in a nonconducting state (Gielen et al., 2015). On the other hand, a failure of pore closure of the inner entrance in the desensitized state could lead to a partially nonconducting desensitized state, thus favoring occurrence of subconductance openings (Dani, 1986;Bormann et al., 1987;Root and MacKinnon, 1994). To determine whether the structural changes were correlated with the macroscopic kinetic defects and resulted in changes in the gating of the receptor, we measured the properties of GABA-evoked (1 mM) single-channel currents recorded from HEK293T cells coexpressing ␣1 and ␤3 subunits with wt ␥2L or mutant ␥2L(P302L) subunits.
Single channels from coexpressed ␣1␤3␥2L subunits opened into brief bursts and frequent prolonged (Ͼ500 ms) burst clusters (Fig. 8A) and opened to a high mainconductance level of ϳ21-28 pS and to a lowconductance level of ϳ14-18 pS in all seven patches (Table 2), as described previously (Twyman et al., 1990;Fisher and Macdonald, 1997). In contrast, single-channel burst clusters from coexpressed ␣1␤3␥2L(P302L) subunits occurred at three distinct conductance levels in 14 patches. Combination of clusters of main-conductance (ϳ22-25 pS) and low-conductance (ϳ13-15 pS) openings were observed in seven of the patches (Fig. 8A, Table 2). Low-conductance openings were reported in patches with ␣␤ receptors (Lo et al., 2010). In addition, a novel sublow-conductance cluster type of openings of ϳ7 pS (6.61 Ϯ 0.15) occurred in combination with clusters of openings of ϳ13-15 pS in seven of the ␣1␤3␥2L(P302L) patches recorded (Fig. 8A). It was important to note that this type of sublow-conductance openings did not occur in wt receptors (Bormann et al., 1987;Twyman et al., 1990;Fisher and Macdonald, 1997). With the intent to simplify the kinetic analysis, those patches with sublowconductance openings were excluded from further singlechannel measurements. Therefore, we measured the singlechannel properties of seven of the patches from cells coexpressing wt ␣1␤3␥2L or mutant ␣1␤3␥2L(P302L) receptors that exhibited only main-and low-conductance openings. When the openings were compared between wt and ␣1␤3␥2L(P302L) receptors, no significant differences in conductance levels were observed (Fig. 8B, top panel; Table 2). Conversely, the relative distribution of main-and low-conductance openings that contributed to the behavior of ␣1␤3␥2L(P302L) receptors was significantly different from the wt receptor. The ␥2L(P302L) subunit mutation caused a significant shift in the contribution of these levels, increasing the contribution of the lowconductance openings (Fig. 8B, lower panel), while reducing the contribution of the main-conductance openings ( Table 2). Both mutant and wt receptors opened to at least three different types of openings (O1, O2, and O3) with three open time constants ( o1 , o2 , and o3 ) ( Table 2), and open time distributions that were fitted best by three weighted (a o1 , a o2 , and a o3 ) exponential functions (Fig.  9A,B) (Table 2). In addition, mutant receptors reduced channel open probability (p ϭ 0.002) and single-channel opening frequency (p ϭ 0.039) and increased singlechannel mean closed time intraburst (p ϭ 0.029) compared with wt receptors (Fig. 9C-F, Table 2).
The low-conductance ␥2(P302L)-GABA A receptors are zinc insensitive As described above, the presence of low-conductance cluster type of openings suggested the formation of ␣1␤3 receptors in the membrane of wt and ␣1␤3␥2L(P302L) patches and support the notion that the observed differ- Figure 8. The de novo ␥2(P302L) subunit mutation shifted the openings of GABA A receptors to low-conductance states. A, Representative single-channel current traces from cell-attached patches of wt ␣1␤3␥2L and mutant ␣1␤3␥2L(P302L) GABA A receptors recorded from HEK293T cells. Patches were voltage clamped at ϩ80 mV and continuously exposed to 1 mM GABA. Openings were downward and each representative trace was a continuous 1600-ms recording. Divisions in the x-axis corresponded to 200 ms, and in the y-axis to 0.5 pA. Note that the cluster of openings fell into the three distinct types of conductance: main (3), low (2), and sublow (1) openings. C refers to the closed state. B, Means of the main and low conductance (pS) and the relative contribution (%) of each conductance level were calculated from patches of wt ␣1␤3␥2L and mutant ␣1␤3␥2L(P302L) GABA A receptors displaying both levels of openings. Values were expressed as mean Ϯ SEM. Two-way ANOVA with Sidak's multiple comparisons test was used to determine significance. ‫‪p‬ءءءء‬ Ͻ 0.0001, and ns, not significant (p ϭ 0.7287), respectively. See Table 2 for details. Kinetic parameters of single-channel currents in cell-attached patches held at ϩ80 mM with 1 mM GABA in the glass electrodes were obtained. The s and a s refer to the time constants and fractions of the three exponential components (O1, O2, and O3), which best represent the distributions of the single-channel openings. Values refer to combined data of low-and main-conductance openings reported are mean Ϯ SEM (n ϭ 7). a Unpaired two-tailed Student's t test relative to ␣1␤3␥2L. b Two-way ANOVA with Sidak's multiple comparisons test relative to ␣1␤3␥2L.
ence in the gating of ␣1␤3␥2L(P302L) receptors was the result of the kinetic behavior of a mixture of two types of channels. Thus, we computed the kinetic properties of each component separately to gain insight into the contribution of both main-and low-conductance openings in the gating of wt and mutant ␣␤␥ receptors and then compared with those that express only ␣␤ receptors. As shown in Table 3, it was obvious that wt and mutant ␣␤␥ receptors were not composed of ␣␤ receptors due the presence of prolonged type 3 openings (O3) in both mainand low-conductance openings, which is characteristic of receptors assembled from ␣␤␥ subunits. Conversely, the lack of O3 openings together with the presence of brief O1 type of openings are distinct properties of ␣␤ receptors, which makes the single-channel behavior of these receptors faster than ␣␤␥ receptors. These comparisons clearly showed that the type of receptor containing the ␥2L(P302L) subunit mutation had kinetic properties that differed from receptors assembled with ␣␤ subunits alone. When the two types of conductance openings between wt and mutant receptors were compared, it was notable that low-conductance openings were contributing Values were expressed as mean Ϯ SEM. Statistical differences were determined using unpaired t test relative to wt. ‫‪p‬ءء‬ Ͻ 0.01, ‫ء‬p Ͻ 0.05, and ns p Ͼ 0.05, respectively. See Table 2 for details.
largely to the kinetic behavior displayed by mutant receptors. Notwithstanding the presence of the three types of openings in the low-conductance state, these openings were briefer than in the wt receptor, which decreased by 70% the mean open time in the mutant receptor. Unexpectedly, main-conductance types of openings were more prolonged in the ␥2L(P302L) mutant receptors than wt receptors, but this was not deemed significant, because of the small contribution (ϳ13%) of total channel openings. We can conclude that ␣1␤3␥2L(P302L) receptors open mainly through clusters of bursts of low-conductance openings.
Further, to confirm that the low-conductance openings resulted from tertiary ␣1␤3␥2L(P302L) receptors, we analyzed the properties of GABA-evoked single-channel openings measured in the presence of 100 M zinc (Fig.  10). Zinc is a well-known GABA A receptor antagonist whose potency is affected by the subunit composition of the receptor (Smart et al., 1991;Hosie et al., 2003). Thus, receptors lacking ␥ subunits are strongly inhibited by zinc. Whether low-conductance openings occurred due to the presence of binary ␣␤ receptors, it is expected to be block by zinc. Noteworthy, we found that, in the presence of zinc and at a membrane potential of ϩ80 mV, the activation of ␣1␤3␥2(P302L) receptors occurred mainly as clusters of low-conductance openings (12.30 Ϯ 0.92 pS, n ϭ 5, p ϭ 0.240 unpaired t test) (Fig. 10A,B), which was comparable with the results obtained previously (Table 2). In addition, no differences were found for wt ␣1␤3␥2 receptors (24.84 Ϯ 0.46 pS, n ϭ 3, p ϭ 0.863 unpaired t test).
Up to this point, we have shown that most of mutant channels open with a lower conductance than wt channels, and that this was not due either to alterations in channel permeability for chloride ions or the presence of binary receptors on the cell surface. These findings also suggested that the conductance of the mutant channel is expected to be a linear assuming that the chloride V rev was ϳ0 mV. Figure 10B shows GABA-evoked singlechannel openings at various membrane potentials in the presence of 100 M zinc. At membrane potentials of ϩ100 mV and ϩ120 mV, the ␥2L(P302L) mutant receptors displayed openings with a conductance of 13.38 Ϯ 0.35 pS (n ϭ 5, p Ͻ 0.0001 unpaired t test) and 23.72 Ϯ 1.58 pS (n ϭ 5, p ϭ 0.004 unpaired t test), which were distinguishable from the wt receptor (29.38 Ϯ 0.97 pS, n ϭ 3; and 35.83 Ϯ 2.63 pS, n ϭ 3, respectively). As a result, low-conductance openings for mutant receptors had a slope conductance of 14.24 Ϯ 1.72 pS, which was signif-icantly different from wt receptors (24.17 Ϯ 1.18 pS, p ϭ 0.000165).
In line with the structural modeling, these results demonstrated that mutant ␣1␤3␥2L(P302L) receptors might perturb the conduction pathway and destabilize the open conformation while trapping the receptor in nonconducting closed and desensitized conformations. This is in accord with reductions in maximum currents, the increased relative contribution of low (and sublow) conductance openings over the main-conductance openings, the enhanced desensitization, slowed activation, and accelerated deactivation in functional experiments.

GABA A receptor-mediated mIPSCs were reduced by overexpressing ␥2(P302L) subunits in cultured neurons
Our data demonstrated that the ␥2(P302L) subunit mutation significantly decreased the function of GABA A receptors expressed in HEK293T cells, also suggesting that there must be an impairment of their function at GABAergic synapse. As previously described (Swanwick et al., 2006), widespread expression of clusters of GABA A receptors containing ␣␤␥ subunits at GABAergic synapses was reached in mature cultured neurons around DIV 19-21. To determine whether the mutant ␥2(P302L) subunit altered the function of GABAergic synapses, we overexpressed wt ␥2L or mutant ␥2L(P302L) subunits in cultured cortical neurons at DIV 12 and measured current amplitude and kinetic properties of GABA A receptor-mediated mIPSCs at DIV 19-21 (Fig. 11). Thus, GABA A receptormediated mIPSCs recorded after overexpression of wt ␥2L subunits displayed current amplitude (-33.94 Ϯ 3.08 pA, n ϭ 4), rise time (1.70 Ϯ 0.08 ms), and decay (25.63 Ϯ 2.23 ms) values that suggested the transfected wt ␥2L subunits formed functional receptors with endogenous ␣␤ subunits, resembling mIPSCs yielded at the GABAergic synapse (Fig. 11). In contrast, overexpression of mutant ␥2L(P302L) subunits decreased current amplitude (-21.16 Ϯ 0.72 pA, n ϭ 4, p ϭ 0.0156) and slowed rise time (2.13 Ϯ 0.14 ms, p ϭ 0.049) and decay (69.11 Ϯ 8.96 ms, p ϭ 0.009). These results suggest that mutant ␥2L(P302L) subunits were incorporated into GABAergic synapses and caused a dominant negative effect, reducing GABAergic function. Although the type of receptor expressed in the membrane was unknown since multiple GABA A receptor subunits were expected to be coexpressed (Swanwick et al., 2006), these findings were comparable with the gating deficiencies found in ␣1␤3␥2L(P302L) receptors coexpressed in HEK293T cells.

Discussion
The findings presented here shed light onto a novel mechanism in the pathogenesis of Dravet syndrome. This study identified a missense de novo mutation in the GABA A receptor ␥2 subunit, P302L, in a patient with Dravet syndrome. The first question to consider was whether a single missense mutation was associated with this catastrophic epileptic encephalopathy. It is well known that missense mutations in coding sequences in the GABA A receptor ␥2 subunit gene GABRG2 are asso- Figure 10. The low-conductance ␥2(P302L)GABA A receptors were zinc insensitive. A, Single-channel current traces from cellattached patches of mutant ␥2L(P302L) and wt recorded at ϩ80 mV and continuously exposed to 1 mM GABA and 100 M zinc. Cluster openings were downward. Upper panels were representative traces continuously recorded for 4000 ms. Red lines indicated that the channel was closed (C), while the gray lines indicated when the channel was open (O) in steps of 1 pA. Dashed boxes enclosed an expanded section of 1600 ms, which was shown below as indicated. B, Representative low-conductance current traces from cell-attached patches of mutant ␥2L(P302L) receptors recorded at various membrane potentials in the presence of 1 mM GABA and 100 M zinc. For the linear regression analysis of conductance, the membrane potentials were indicated as negative values. The traces were recorded from the same patch at the membrane potentials indicated. The red lines represented the closed state of the channels. In the right panel, single-channel current-voltage (I/V) plots of mutant ␥2L(P302L) and wt receptors were shown. The amplitude of the most frequently occurring opening was used for the construction of the I/V plots. Thus, low-conductance openings were used for the mutant receptor, whereas main-conductance openings were used for wt receptors. Data points were values expressed as mean Ϯ SEM (n ϭ 3-5 cells for each experimental condition). Solid lines represented fits of the linear regression equation, and the doted lines represented 95% confidence intervals of the best-fit line. Zn 2ϩ , zinc. ciated with relatively mild epilepsy phenotypes, including childhood absence epilepsy and febrile seizures (Wallace et al., 2001;Audenaert et al., 2006;Shi et al., 2010), and with the genetic epilepsy with the febrile seizures plus (GEFSϩ) spectrum (Baulac et al., 2001). Intriguingly, missense mutations in the GABA A receptor ␣1 subunit gene have been reported in cases of Dravet syndrome (Carvill et al., 2014). Taking into account the location of these mutations in the ␣1 subunit, it is significant that all three mutations (R112Q, G251S, and K306T) share the same topological domain in the receptor, which is directly related to the ligand-binding coupling mechanism (Bianchi et al., 2001;Bera et al., 2002;Keramidas et al., 2006;Wang et al., 2010;Calimet et al., 2013;Althoff et al., 2014). Thus, ␣1(R112Q) and ␣1(G251S) subunit mutations face the ␤ϩ/␣subunit-subunit GABA-binding interface within the N terminus and M1 transmembrane helix, respectively, whereas the ␣1(K306T) subunit mutation is Figure 11. GABA A receptor-mediated mISPCs were altered by overexpression of ␥2(P302L) subunits in cultured neurons. A, Representative GABA A receptor-mediated mISPCs traces recorded at -60 mV in cultured cortical neurons overexpressing mutant ␥2L(P302L) subunits or wt ␥2L subunits. B, A segment of each raw trace in A (red dashed lines) was expanded for better visualization of the differences in mISPC properties. C, Cumulative histograms of GABA A receptor-mediated mISPCs current amplitudes of neurons overexpressing mutant ␥2L(P302L) subunits or wt ␥2L subunits show the Kolmogorov--Smirnov (K-S) comparison test with a statistical value of 61% (p Ͻ 0.00001). D, Overlapped ensemble average traces of mISPCs recorded from receptors containing wt-␥2L (blue) and mutant ␥2L(P302L) (red) subunits showed reduced amplitude and kinetic changes. E, F, Bar graphs summarize the effects of overexpression of wt and mutant ␥2L(P302L) subunits on current amplitude (E), rise time (F), and decay (G) of GABA A receptor-mediated mISPCs. Values were expressed as mean Ϯ SEM. Statistical differences were determined using unpaired t test relative to wt. ‫‪p‬ءء‬ Ͻ 0.01 and ‫ء‬p Ͻ 0.05, respectively. located in the M2-M3 extracellular loop, which is part of the coupling interface of the GABA A receptor. In line with these findings, the next question to consider was whether the location of the ␥2 subunit mutation in the GABA A receptor may account for the severity of the epilepsy phenotype. While most of the missense ␥2 subunit mutations are preferentially distributed across the N-terminal domain and in the outermost region of the transmembrane segment M2 not facing directly ␤ϩ/␣-GABAbinding coupling interface (Baulac et al., 2001;Wallace et al., 2001;Audenaert et al., 2006;Shi et al., 2010;Lachance-Touchette et al., 2011;Carvill et al., 2013;Reinthaler et al., 2015), the ␥2(P302L) subunit mutation was mapped in the pore-forming region of the GABA A receptor. This was probably the most significant feature that distinguished the defects caused by the occurrence of this mutation as discussed below.

A mechanism of gating impairment of GABA A receptors by the ␥2(P302L) subunit mutation
For Cys-loop receptors (Hibbs and Gouaux, 2011;Althoff et al., 2014;Du et al., 2015), pore-lining residues of the transmembrane M2 domain of GABA A receptors form the axis of the channel pore. The pore exposes two gates that are open or closed depending on the conformational state of the receptor. One gate corresponds to the activation/opening gate, which is at position 9', and the other in the cytoplasmic interface representing the desensitization gate at position -2' (Fig. 5). Thus, the conducting pathway shows an asymmetrical hour-glass-like cavity divided by the narrow open gate at 9' position. The top of the pore axis corresponds to the outermost part of the pore, which is the extracellular vestibule, and the bottom of the pore axis (-2' position) corresponds to the innermost part of the pore, which is the intracellular vestibule, where the ␥2(P302L) subunit mutation occurs. Through structural simulation of the different states that the GABA A receptor adopts during the activation and deactivation of the channel, we found that the shapes of the vestibules that determine the features of the pore were deeply affected by the mutation, thereby altering the gating of the receptor. Further, we propose a structural mechanism to explain the deficiencies observed in the gating of receptors containing the ␥2(P302L) subunit mutation. Thus, the decrease in the outer vestibule of the pore in the open state can affect the anticlockwise rotation of the transmembrane segments M2 for channel activation (Hibbs and Gouaux, 2011;Althoff et al., 2014;Du et al., 2015) and may account for the slowing in the activation of the channel and the occurrence of low-conductance openings. The narrowing of the inner vestibule of the pore in the closed state instead suggests that the channel is transiently trapped in a nonconducting state (Gielen et al., 2015), explaining the profound desensitization of the currents. Finally, the pore closure failure at the inner entrance in the desensitized state could lead to a partially nonconducting desensitized state, favoring the occurrence of sublow-conductance openings (Dani, 1986;Bormann et al., 1987;Root and MacKinnon, 1994), corroborating the experimental observations of clusters of sublow openings. Further, slowing of recovery from desensitization and reduction in the activation caused by the ␥2(P302L) subunit mutation may be responsible for facilitating the allosteric modulation by zinc (Barberis et al., 2000) and increase of zinc sensitivity.

Missense epilepsy mutations in the M2 domain of Cys-loop receptors cause hyperexcitability by altering current desensitization and channel conductance
Among Cys-loop family receptors, structural studies revealed that the nicotinic acetylcholine receptor (nAChR) (Unwin, 2005), GABA A receptor (Miller and Aricescu, 2014), GlyR (Du et al., 2015), and GluClR (Hibbs and Gouaux, 2011) share a common ion-conducting pore scaffold composed of four transmembrane segments (M1 to M4), with the transmembrane segment M2 lining the pore. Thus, it has been suggested that these receptors share a similar structural basis for the ligand bindingchannel gating coupling mechanism (Unwin, 2005;Hibbs and Gouaux, 2011;Althoff et al., 2014;Miller and Aricescu, 2014;Du et al., 2015). Missense mutations of the nAChR ␣4 subunit associated with autosomal dominant nocturnal frontal lobe epilepsy affected highly conserved residues in M2, specifically those residues predicted to be part of the M2-axis that rotates when the agonist binds and opens the pore. Similar to the defects caused by the GABA A receptor ␥2(P302L) subunit mutation, nAChR ␣4(S248F) (Weiland et al., 1996;Kuryatov et al., 1997) and ␣4(S252L) (Matsushima et al., 2002) subunit mutations caused faster desensitization and decreased singlechannel conductance. In addition, in homomeric GlyR ␣1 subunits, autosomal dominant mutations associated with hyperekplexia were also found clustered in and around the pore-lining transmembrane segment M2 impairing GlyR ␣1 subunit gating. At the extracellular end of the M2, GlyR ␣1(R271Q) and GlyR ␣1(R271L) subunit mutations reduced single-channel conductance (Langosch et al., 1994), while the GlyR ␣1(V280M) subunit mutation in the M2-M3 loop enhanced glycine sensitivity and spontaneous channel activity (Bode et al., 2013). The GABA A receptor ␥2(P302L) subunit missense mutation occurred in the evolutionarily conserved proline residue in the pore region of GABA A receptors, GlyRs (Du et al., 2015) and GluCL receptors (Hibbs and Gouaux, 2011). Remarkably, the autosomal dominant hyperekplexia GlyR ␣1(P250T) subunit mutation occurred at a position homologous to that of the ␥2(P302L) subunit mutation. Resembling the ␥2(P302L) subunit mutation, the GlyR ␣1(P250T) subunit mutation reduced glycine-activated current amplitudes, induced fast desensitization and reduced single-channel conductance (Saul et al., 1999), which accounted for the hyperekplexia phenotype. None of these mutations impaired cell surface expression. Thus, the main mechanisms by which these mutations cause hyperexcitability appear to be through impairment of receptor gating, thereby contributing to the epilepsy phenotype.