The majority of the GlyR α2 variants studied were sourced from published sources (Iossifov et al., 2014; Krumm et al., 2015; Deciphering Developmental Disorders Study, 2017. GlyR α2 p.K213E was identified by ES, GvR, and KS in diagnostic exome sequencing. Site-directed mutagenesis of the human GlyR α2 subunit cDNA in the expression vector pRK5 (Zhang et al., 2017) was performed using the QuikChange kit (Agilent, Santa Clara, CA, United States). The successful incorporation of mutations was confirmed by Sanger DNA sequencing, performed by DNA Sequencing and Services (MRC PPU, School of Life Sciences, University of Dundee, United Kingdom) and analysis using Sequencher software (Gene Codes, Ann Arbor, MI, United States). Plasmid DNAs were prepared using a HiSpeed Plasmid Maxi Kit (QIAGEN, Hilden, Germany). The damaging effects of human GlyR α2 subunit variants were assessed using Sorting Intolerant From Tolerant (SIFT; Sim et al., 2012), PolyPhen-2 (Adzhubei et al., 2013), and CADD scores (Rentzsch et al., 2019). Mutation position in the GlyR α2 subunit is indicated using mature subunit numbering (i.e., after signal peptide cleavage). The effects of the V-22L variant on signal peptide cleavage were assessed using SignalP 4.0 (Petersen et al., 2011). Molecular modeling of the p.N38K and p.K213E variants was accomplished using the recently resolved structures of the GlyR α2β pentamer in the closed (RCSB: 7L31) and glycine-bound open state (RCSB: 5BKF) (Yu et al., 2021). GlyR structures were visualized using the UCSF ChimeraX molecular visualization program (Pettersen et al., 2021). Amino acid substitutions were modeled using the swapaa command, taking into account the highest rotamer prevalence (Dunbrack backbone-dependent rotamer library, Shapovalov and Dunbrack, 2011), the highest number of H-bonds and the lowest clash score.

A homology model of the GlyR α2 homopentamer was also constructed using the Phyre2 web server (Kelley et al., 2015). GlyR α2 monomers were overlaid onto the GlyR α1 homopentamer (PDB ID: 3JAE; Du et al., 2015) to create the GlyR α2 homopentamer. The GlyR α2^T269M variant was created by introducing M269 into the GlyR α2 homopentamer using PyMOL (DeLano, 2002). Wild-type and GlyR α2^M269 homopentamer systems were embedded into a model membrane composed of 80% POPC and 20% CHOL. Each system was solvated with SPC water, neutralized with Na⁺ and NaCl was added to a concentration of 0.15 M. This led to an overall system consisting of the GlyR α2 homopentamer, in a membrane containing 80 mol% POPC and 20 mol% CHOL, surrounded by ∼155,000 water molecules and 0.15 M NaCl. Simulations were conducted in the apo state, in the absence of the ligand glycine. All systems were simulated using GROMACS 2019.4 molecular dynamics engine (Abraham et al., 2015) in conjunction with the GROMOS 54a7 force field (Schmid et al., 2011). The system was energy minimized using the steepest descent algorithm and equilibrated in five sequential 1 ns simulations with decreasing restraints on the protein (1000 kJ mol^–1 nm^–1, 500 kJ mol^–1 nm^–1, 100 kJ mol^–1 nm^–1, 50 kJ mol^–1 nm^–1, and 10 kJ mol^–1 nm^–1). Each unrestrained system was then simulated in triplicate for 500 ns. In all simulations, a 2 fs timestep was used. The pressure was maintained at 1 bar using semi-isotropic pressure coupling using the Berendsen barostat (τ_P = 0.5 ps and isothermal compressibility = 4.5 × 10^–5 bar), and the temperature was maintained at 300 K using the Bussi-Donadio-Parrinello velocity rescale thermostat (τ_T = 0.1 ps). Periodic boundary conditions were implemented. SETTLE was used to constrain the geometry of water molecules and LINCS was used to constrain the covalent bond lengths of the solute. Analysis was performed using the GROMACS tools and the trj_cavity package on the entire 1.5 μs of combined production simulation for each system. The Visual Molecular Dynamics (VMD) program was used for visualization of the simulations (Humphrey et al., 1996; Paramo et al., 2014).

HEK293 cells (CRL-1573; ATCC – Global Biosource Center, Manassas, VA, United States) were transfected with GlyR constructs using the Ca²⁺ phosphate-DNA co-precipitation method as previously described (Sontheimer et al., 1989). Cells were washed 6 h post- transfection and used for biotinylation assays (New England Biolabs, Ipswich, MA, United States) 48 h after transfection. Biotin labeling and subsequent binding to streptavidin was used to discriminate between whole-cell and cell-surface protein. Surface proteins were labeled by incubating the cells for 30 min with 1 mg/ml EZ-Link Sulfo-NHS-LC-Biotin [sulfosuccinimidyl-6-(biotinamido)hexanoate (Pierce Biotechnologies, Rockford, IL, United States)]. Following a quenching step (192 mM glycine, 25 mM Tris in PBS, pH 8.0 for 10 min), cells were detached by using ice-cold PBS buffer. After centrifugation for 10 min at 1,000 × g, cells were lysed with TBS (Tris-buffer saline with 1% Triton-X100 and protease inhibitor mixture tablet, Roche Diagnostics, Mannheim, Germany). After centrifugation for 1 min at 13,000 × g, the supernatant (whole protein fraction) was incubated with 50 μl of streptavidin-agarose beads (Pierce Biotechnologies, Rockford, IL, United States) for 2 h at 4°C. Beads were washed three times in TBS buffer. Biotinylated proteins (surface fraction) were eluted by boiling with 50 μl of 2 × SDS buffer for 5 min at 95°C. Whole cell (WC) and cell surface (SF) fractions were separated by SDS-PAGE and Western blotting on nitrocellulose membranes (GE Healthcare, Little Chalfont, United Kingdom). Membranes were blocked for 1 h with 5% BSA in TBS-T (TBS with 1% Tween 20). GlyR α2 subunits were detected with the antibody mAb4A (cat. no. 146011, 1:1,000, Synaptic Systems, Göttingen, Germany). Pan-cadherin (Cell Signaling Technology, Danvers, MA, United States, 4068, 1:1000) served as a loading control for the whole-cell fraction and cell-membrane protein fractions. Signals were detected using the ECL plus system (GE Healthcare, Little Chalfont, United Kingdom). Image quantification was performed using ImageJ (1.51)/Fiji2 (Schindelin et al., 2012, 2015; Schneider et al., 2012). Data were analyzed using Student’s t-test. For all tests, the number of asterisks corresponds to the level of statistical significance: *p < 0.05; ^**p < 0.01; ^***p < 0.001. Values are displayed as means ± standard error of the mean (±SEM) unless otherwise noted.

Spinal neurons were prepared using methods as previously described (Dixon et al., 2015; Zhang et al., 2015b,2017; Leacock et al., 2018). Briefly, E15 timed-pregnant rats were euthanized via CO₂ inhalation in accordance with procedures approved by The University of Queensland Animal Ethics Committee (Approval number: QBI/142/16/NHMRC/ARC). The spinal cords were rapidly removed, triturated and plated onto poly-D-lysine-coated coverslips in a 4-well plate at a density of 8–10 × 10⁴ cells/well, and cultured for 3–4 weeks until spontaneous inhibitory postsynaptic currents (IPSCs) could be detected. The cells were initially cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (DMEM-FBS). After 24 h, the entire DMEM-FBS medium was replaced with Neurobasal medium including 2% B27 and 1% GlutaMAX supplements. A second and final feed 1 week later replaced half of this medium with fresh Neurobasal medium. Neurons were used in co-culture experiments between 1 and 4 weeks later.

Artificial synapses were generated as previously described (Dixon et al., 2015; Zhang et al., 2015b,2017; Leacock et al., 2018). Briefly, HEK293 cells were cultured in DMEM-FBS until ∼90% confluent. One day prior to transfection, cells were trypsinized and plated onto glass coverslips in 35 mm culture dishes at a density of 5 × 10³ cells/dish. Each dish was transfected with 0.3 μg of GlyR α2 subunit DNA, plus 0.1 μg EGFP (pEGFP) was used as a transfection marker. For artificial synapses, 0.3 μg of mouse neuroligin 2A (pNice) and 0.3 μg of rat gephyrin (pCIS) were also added. Transfection was performed via a Ca²⁺ phosphate-DNA co-precipitation method for 15–20 h in a 3% CO₂ incubator and terminated by washing cells twice with divalent cation-free phosphate buffered saline. Cells were trypsinized the next day, centrifuged and re-suspended in Neurobasal medium (including 2% B27 and 1% GlutaMAX supplements) then seeded onto neurons. One 35 mm dish of HEK293 cells was typically sufficient to seed four coverslips of neurons. Once seeded with HEK293 cells, the co-cultures were returned to the incubator overnight to allow artificial synapses to form between neurons and transfected HEK293 cells. Cells were used for patch-clamp recording over the following 2–3 days.

Whole-cell patch clamp recordings were performed at room temperature (22 ± 1°C). Glycine concentration-response relationships were performed at −40 mV, whereas artificial synapse recordings were performed at −70 mV, both using a MultiClamp 700B amplifier and pCLAMP 10 software (Molecular Devices). Signals were filtered at 4 kHz and sampled at 10 kHz. Patch pipettes (4–8 MΩ resistance) were fabricated from borosilicate glass (GC150F-7.5, Harvard Apparatus) and filled with an internal solution comprising (in mM): 145 CsCl, 2 CaCl₂, 2 MgCl₂, 10 HEPES, and 10 EGTA, adjusted to pH 7.4 with CsOH. The extracellular solution comprised (in mM) 140 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, and 10 D-glucose, adjusted to pH 7.4 with NaOH.

Analyses of IPSC amplitudes, 10–90% rise times, and decay time constants were performed using AxoGraph X (AxoGraph Scientific). Only cells with a stable series resistance of <25 MΩ throughout the recording period were selected for analysis. IPSCs were detected using a semi-automated sliding template. Each detected event was visually inspected and only those with no inflections in the rising or decay phases were included. All selected events from a single cell were digitally averaged. Parameters derived from these digitally averaged waveforms were then pooled with those from other cells to obtain group data. To calculate macroscopic current decay time constants, digitally averaged macroscopic recordings were fitted with double-exponential functions in AxoGraph X, and a weighted time constant was calculated from individual time constants (τ1, τ2) and their relative amplitude (A1, A2) as follows: τ_weighted = (τ1 × A1 + τ2 × A2)/(A1 + A2). Displayed averaged data represent group means ± SEMs. The Hill equation was used to calculate the saturating current magnitude (I_max), half-maximal concentration (EC₅₀), and Hill coefficient (n_H) values for glycine activation. Individual concentration-response relationships were fitted using a non-linear least squares algorithm (SigmaPlot 11.0; Jandel Scientific, San Rafael, CA, United States). Statistical analysis and graphing were performed with SigmaPlot 11.0. Data were first tested for normality using both the Shapiro–Wilk and Kolmogorov–Smirnov tests. Via either test, all data proved normally distributed using an alpha value of 0.05. Statistical analysis was then performed using a one-way ANOVA for multiple comparisons followed by Tukey’s post hoc test. P values of <0.05 were taken to be statistically significant.

Candidate GlyR α2 subunit mutations were identified from exome sequencing studies in ASD and/or developmental disorders (Iossifov et al., 2014; Krumm et al., 2015; Deciphering Developmental Disorders Study, 2017; Figure 1A and Supplementary Table 1) and include p.V-22L (signal peptide), p.N38K (extracellular domain, ECD), p.K213E (ECD), and p.T269M (M2). These correspond to substitutions p.V6L, p.N65K, p.K240E, and p.T296M in the human GlyR α2 subunit precursor prior to signal peptide cleavage. The damaging effects of the human GlyR α2 subunit variants were assessed using SIFT (Sim et al., 2012), PolyPhen-2 (Adzhubei et al., 2013), and CADD scores (Rentzsch et al., 2019; Supplementary Table 1). All variants were found to be possibly/probably damaging with PolyPhen-2 and all had high CADD scores: p.V-22L, 15.78; p.N38K, 20.4, p.K213E, 24.3 and p.T269M 25.4, consistent with previously reported GlyR α2 subunit variants associated with ASD (p.N109S, p.R126Q, p.R323L; Supplementary Table 1). All variants were absent from gnomAD database (Karczewski et al., 2020) with the exception of p.K213E, which occurs with a low frequency of 4/177,746 alleles. However, given the high CADD score for p.K213E (24.3), we proceeded with structure/function analysis. The variant p.V-22L would not normally be expected to affect GlyR α2 subunit function, since it is located in the cleavable signal peptide found at the N-terminus of the protein. However, on analysis with SignalP 4.0 (Petersen et al., 2011) we noted that p.V-22L subtly alters the predicted signal peptide cleavage site for GlyR α2. While the wild-type protein was predicted to be cleaved between amino acids 27 and 28 AFC-KD, the GlyR α2^V–22L missense variant was predicted to be cleaved between amino acids 21 and 22: TNH-FR (Figure 1A), which could influence the efficiency of signal peptide cleavage and cell-surface expression. By contrast, GlyR α2^N38K and GlyR α2^K213E lie within the ligand-binding ECD (Figure 1A), while GlyR α2^T269M affects a highly conserved residue within the pore-forming M2 domain (Figure 1B). Interestingly, substitutions at the equivalent residues to GlyR α2^N38K, α2^K213E and α2^T269M have not been observed in the GlyR α1 subunit in startle disease (Chung et al., 2010; Bode et al., 2013; Zhang et al., 2016; Figure 1).

Initial analysis of the potential effects of GlyR α2^N38K and GlyR α2^K213E variants was conducted using the GlyR α2β pentamer in the closed (RCSB: 7L31) and glycine-bound open state (RCSB: 5BKF) (Yu et al., 2021). GlyR structures were visualized using the UCSF ChimeraX molecular visualization program (Pettersen et al., 2021). The GlyR α2^N38K variant introduces a larger charged side-chain that is predicted to result in clashes with the glycan attached to residue N45 in both closed and open states (Figures 2A,B; Yu et al., 2021). Artificial mutations of the corresponding consensus glycosylation site in GlyR α1 (N-X-T), encompassing N38 and S40, have been found to be essential for GlyR homo-oligomerization and receptor biogenesis (Griffon et al., 1999). However, GlyR α2 is now known to be glycosylated at two sites, N45 and N76 (Yu et al., 2021; Figure 1), with the second site being specific to GlyR α2. We therefore predict that while GlyR α2^N38K may not interfere with N-linked glycosylation at N76, it could negatively impact GlyR homo-oligomerization and cell-surface trafficking by interfering with glycosylation at N45. GlyR α2^K213 is located in the second dicysteine loop in the ECD and is flanked by key ligand-binding residues including GlyR α2 Y209, T211 and F214 (Figure 1, blue lettering; Figures 2C,D). GlyR α2^K213E introduces charge swap to the region and some loss of flexibility in the side chain. In the closed state, we found an obvious clash with H208 (Figure 2E), but in the open state, the glutamic acid side chain was free of clashes and made additional contacts with Y209 (Figure 2F). These changes suggest that for GlyR α2^K213E, the open state may be favored, resulting in prolonged opening of the ion channel.

For the GlyR α2^T269M variant, we used molecular dynamics simulations to examine the potential effects of this substitution in the M2 domain. Wild-type GlyR α2^T269 and mutant α2^M269 GlyR homopentamers were stable over the triplicate 500 ns simulations (Figures 3A–D). The ion channel for the GlyR α2 homopentamer is closed around T272, giving a physical occlusion to the channel pore (Figures 3A,C,E). By contrast, the M269 ion channel is open and allows movement between the extracellular and intracellular solutions (Figures 3B,D,E). This opening of the channel leads to an increase in the channel volume for the GlyR α2^M269 vs. wild-type GlyR α2^T269 and a corresponding increase in the number of water molecules within the channel. An average of 56 water molecules was found within the channel for wild-type GlyR α2^T269 across each 500 ns replicate simulation, compared to an average of 130 water molecules for the GlyR α2^M269 variant. The changes in ion channel volume and water occupancy were also coupled with an increase in Cl^– ion presence in the channel for GlyR α2^M269, compared to wild-type GlyR α2^T269. Here, one or more Cl^– ions are found in the channel for 27% of the total 1,500 ns of combined simulation time for α2^M269, compared to only 5% of the total combined simulation time of the wild-type α2^T269 system. The changes in ion channel properties between the wild-type α2^T269 and α2^M269 systems are due to altered inter-residue interactions in this region of the channel (Figures 4A,B). Specifically, in the wild-type GlyR α2, T269 is stabilized by hydrogen bonds with the adjacent polar residues T265 and Q273 located on the same M2 domain in an arrangement where the side chains are stacked (Figure 4A). Conversely, while the larger, non-polar mutant M269 also interacts with T265 and Q273, the longer side chain forms an additional interaction with T271 in the M2 domain of the adjacent GlyR α2 monomer (Figure 4B). When T271 interacts with the mutated M269, T271 is no longer able to stabilize the closed conformation of the L268 gating residue. In the GlyR α2^M269 system, the backbone of L268 forms hydrogen bonds with the sidechain of T272. Collectively, these changes in hydrogen bonding in the region surrounding M269 and L268 predict an opening of the channel for GlyR α2^M269 compared to wild-type α2^T269, increasing both the water and Cl^– occupancy of the channel.

To examine the effects of these GlyR α2 variants on cell-surface expression, we measured whole-cell versus surface GlyR expression levels by labeling of surface proteins with biotin followed by cell lysis and precipitation of biotin-labeled proteins using streptavidin beads (Figures 5A–D and Table 1). Samples from the lysate refer to the whole-cell protein pool (Figures 5A,B), samples of biotinylated proteins refer to the surface-expressed receptor protein (Figures 5C,D). Whole-cell and surface-expressed protein levels were first normalized to the expression levels of cadherin, and then relative to wild-type GlyR α2 levels which were designated as 100%. While whole-cell expression of the signal peptide variant GlyR α2^V–22L was not significantly reduced compared to wild-type GlyR α2, cell-surface expression was significantly reduced (α2^V–22L 52 ± 12% of control values, *p < 0.05). By contrast, for GlyR α2^N38K, predicted to interfere with N-linked glycosylation, both whole-cell and cell-surface expression levels were significantly reduced (α2^N38K whole-cell 26 ± 6%^**; cell surface 11 ± 7%^**; ^**p < 0.01, Figures 5A–D and Table 1). For the remaining two GlyR α2 variants α2^K213E and α2^T269M whole-cell expression levels were indistinguishable from wild-type GlyR α2 (α2^K213E 68 ± 25% and α2^T269M 86 ± 15% of wild-type values), while both showed diminished cell-surface expression (α2^K213E: 42 ± 8%*; α2^T269M: 30 ± 9%*; *p < 0.05, Figures 5A–D and Table 1). Hence, all GlyR α2 missense variants affected cell-surface expression to varying degrees, but none were completely retained in the endoplasmic reticulum or other subcellular compartments.

Consistent with cell-surface trafficking data, GlyR α2^N38K subunit homomers expressed in HEK293 cells exhibited a significantly reduced mean I_max value (α2^N38K 3.3 ± 0.7 vs. wild-type 8.9 ± 1.6 nA; n = 7 cells each; p < 0.01) and a significantly increased glycine EC₅₀ value (α2^N38K 243 ± 12 vs. wild-type 141 ± 14 μM; n = 7 cells each; p < 0.001) compared to wild-type GlyR α2 subunit homomers (Figures 6A,B and Table 2). Again, this is consistent with a loss-of-function for GlyR α2^N38K. By contrast, despite the reduced expression levels observed in cell-surface biotinylation experiments, GlyR α2^K213E subunit homomers exhibited no significant change in either I_max or EC₅₀ relative to the wild-type GlyR α2 subunit homomers (Figures 6A,B and Table 2), suggesting that this missense change has a more subtle effect on GlyR function, as predicted by molecular modeling. However, as suggested by the location of p.T269M substitution in the ion-channel pore, and molecular dynamics simulations, GlyR α2^T269M subunit homomers exhibited a dramatic phenotype (Figures 6A,B and Table 2). GlyR α2^T269M homomers not only displayed robust glycine-gated currents but also exhibited a significant leakage current, as revealed by the block of the baseline current by 100 μM picrotoxin, an inhibitor of homomeric GlyRs. Averaged from five cells, the mean magnitude of the picrotoxin-blocked current was 240 ± 35 pA, and the relative magnitude of leak current to saturating whole-cell current in individual cells was 18.7 ± 3.8% (n = 5 cells). By contrast, we did not observe any upward deflection in the baseline current when 100 μM picrotoxin was applied to cells expressing wild-type GlyR α2, α2^N38K or α2^K213E (data not shown).

Relative to wild-type GlyR α2 subunit homomers, GlyRs containing the α2^T269M subunit also exhibited a significantly reduced mean glycine-activated I_max current (2.1 ± 0.4 vs. 8.9 ± 1.6 nA; n = 7 cells; p < 0.001, Table 2) but this was counterbalanced by a significantly reduced glycine EC₅₀ value (α2^T269M 4.5 ± 1.7 vs. wild-type 141 ± 14 μM; n = 7 cells; p < 0.001, Figures 6A,B and Table 2). Thus, despite a reduction in cell surface-trafficking, the leak current and the high glycine sensitivity of GlyR α2^T269M subunit homomers are suggestive of a gain-of-function that is predicted to increase glycinergic signaling at synapses. It should be noted that tonic leak currents that impair cell viability have been previously observed for “leaky” GlyR α1 subunit mutants (Bode et al., 2013; Zhang et al., 2016). The degree of degradation in viability may well have been proportional to the functional expression level of the GlyR α2^T269M construct in individual cells. Thus, by selecting relatively healthy cells for analysis, we may have biased cell selection toward weakly expressing cells with smaller than average whole-cell current magnitudes.

For functional studies in artificial synapses, we utilized homomeric α2 subunit GlyRs, as these extrasynaptic GlyRs represent the predominant prenatal isoform that is critical for interneuron migration in the developing cortex (Avila et al., 2013, 2014). In the artificial synapse system, homomeric α2 GlyRs exhibit slow decay time constants, and are thought to be perisynaptic in location due to slow 10–90% rise times (Table 3), implying that they are located a substantial distance from presynaptic terminals (Zhang et al., 2015a). Whole-cell recordings from transfected HEK293 cells in co-culture with spinal neurons exhibited robust, spontaneous IPSCs with amplitudes up to 1000 pA. Sample recordings at low and high temporal resolution for wild-type GlyR α2 and each variant are shown in Figures 7A–D, left and center panels. After each recording, we normalized and digitally averaged all well-separated IPSCs to produce a single globally averaged waveform. We thereby obtained a single averaged 10–90% rise time, decay time constant and amplitude for each cell. Table 3 summarizes the mean values obtained for each of the three parameters. These values were averaged from 9 to 46 cells as indicated. IPSCs mediated by wild-type GlyR α2 exhibited a mean amplitude of 60.6 ± 9.7 pA, a 10–90% rise time of 6.76 ± 0.98 ms and a mean decay time constant of 105.3 ± 11.4 ms (n = 22 cells). These values are very similar to those recorded previously from wild-type GlyR α2 expressed in artificial synapses (Zhang et al., 2015b). Relative to wild-type GlyR α2 values, IPSCs mediated by GlyR α2^N38K exhibited significantly reduced amplitudes (23.4 ± 3.0 pA; p < 0.001, n = 46 cells) although IPSC rise and decay times were unchanged (Table 3). Notably, GlyR α2^K213E-mediated IPSCs were dramatically different from wild-type GlyR α2 values, with significantly larger amplitudes (α2^K213E 271.9 ± 104.2 vs. wild-type 60.6 ± 9.7 pA; n = 22 and 46 cells, respectively; p < 0.001), significantly faster rise times (α2^K213E 4.58 ± 0.35 vs. wild-type 6.76 ± 0.98 ms; n = 22 and 46 cells, respectively; p < 0.05), and significantly slower decay times (α2^K213E 240.4 ± 55.4 vs. wild-type 105.3 ± 11.4 ms; n = 22 and 46 cells, respectively; p < 0.05). Thus, although this variant appeared to have little functional effect in patch-clamp experiments, in artificial synapses GlyR α2^K213E dramatically enhanced glycinergic signaling suggesting that it causes a gain-of-function. Unfortunately, HEK293 cells expressing GlyR α2^T269M were unhealthy when maintained in co-culture for several days and this permitted only short-lasting, unstable recordings. We were able to obtain an estimate of the mean IPSC amplitude (23.2 ± 5.8 pA) from n = 9 cells despite attempted recordings from >200 cells. Moreover, due to the extraordinarily long IPSC decay times (e.g., Figure 7D, bottom center panel), it was not possible to isolate individual events, and thus we could not quantify mean IPSC rise and decay times. However, these results are consistent with the leak currents and gain-of-function observed in simple patch-clamp experiments.