Expression constructs encoding the human GlyR α2 and β subunits were combined in 1α:50β ratio (heterosynapse and macropatch recordings) or 1α:100β (single-channel recordings) and transfected into HEK293 cells via Ca²⁺ phosphate-DNA co-precipitation. This resulted in a high level of expression of heteromeric α2β GlyRs (Zhang Y. et al., 2015). For heterosynapse experiments, expression constructs for the mouse neuroligin 2A (NL2A) splice variant and rat gephyrin were co-transfected along with GlyR constructs to facilitate the formation of heterosynapses. Empty pEGFP or CD4 plasmid was also transfected as expression markers. Site-directed mutagenesis was performed using the QuikChange kit (Agilent), and the successful incorporation of mutations was confirmed by Sanger DNA sequencing. Mutation position in the GlyR α2 subunit is indicated using mature subunit numbering (i.e., after signal peptide cleavage).

Primary cultures of spinal cord neurons were prepared as previously described (Dixon et al., 2015; Zhang Y. et al., 2015). E15 timed-pregnant rats were euthanized via CO₂ inhalation in accordance with procedures approved by the University of Queensland Animal Ethics Committee. Cells were plated at a density of ~80,000 cells per 18 mm poly-D-lysine coated coverslip in DMEM medium with 10% (v/v) foetal bovine serum. After 24 h, the plating medium was changed to Neurobasal medium supplemented with 2% (v/v) B27 and 1% (v/v) GlutaMAX, and a second feed after 1 week replaced half of this medium. Neurons were grown for 1–4 weeks in vitro and heterosynaptic co-cultures were prepared by directly introducing transfected HEK293 cells onto the primary neuronal cultures 1–3 days prior to recording.

Whole-cell recordings were performed on transfected HEK293 cells in voltage-clamp mode using a HEKA EPC10 amplifier (HEKA Electronics, Lambrecht, Germany) and PATCHMASTER software (HEKA), at room temperature. Cells were placed in an external solution comprising (in mM): 140 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES and 10 D-glucose, adjusted to pH 7.4 with NaOH. Patch pipettes (1–3 MΩ resistance), were pulled from borosilicate glass (GC150F-7.5, Harvard apparatus), and filled with an intracellular solution containing the following (in mM): 145 CsCl, 2 CaCl₂, 2 MgCl₂, 10 HEPES, 10 EGTA and 2 MgATP, adjusted to pH 7.4 with NaOH. Glycine-gated currents were recorded at a holding potential of −40 mV, digitized at 4 kHz and filtered at 10 kHz. For inhibitory postsynaptic current (IPSC) recordings, patch-pipette resistances were adjusted to 4–6 MΩ and filled with the same internal solution. Series resistance was routinely compensated to 60% of maximum and was monitored throughout the recording. Both spontaneous and action potential-evoked glycinergic IPSCs in HEK293 cells were recorded at a holding potential −60 mV and signals were digitally sampled at 10 kHz and filtered at 4 kHz. As these IPSCs were completely abolished by 1 μM tetrodotoxin (not shown), we infer they were induced by spontaneous action potentials.

Single-channel currents were recorded from outside-out excised patches at a clamped potential of −70 mV. Glass electrodes were pulled from borosilicate glass (G150F-3; Warner Instruments), coated with a silicone elastomer (Sylgard-184; Dow Corning) and heat-polished to a final tip resistance of 8–15 MΩ when filled with an intracellular solution containing (in mM) 145 CsCl, 2 MgCl₂, 2 CaCl₂, 10 HEPES and 5 EGTA, pH 7.4. Excised patches were directly perfused with extracellular solution by placing them in front of one barrel of a double-barrelled glass tube. Single-channel currents were either recorded while the patch was exposed to extracellular solution (without added glycine) or elicited by exposing the patch continuously to glycine (100 μM or 3 mM) containing solution. Experiments were recorded using an Axopatch 200B amplifier (Molecular Devices), filtered at 5 kHz and digitized at 20 kHz using Clampex (pClamp 10, Molecular Devices) via a Digidata 1440A digitizer. The currents were filtered off-line at 3 kHz for analysis.

Macropatch recordings were performed in the excised outside-out patch-clamp configuration. Patch pipettes were fire-polished to a resistance of approximately 10 MΩ and filled with the same internal solution. Macroscopic currents in outside-out patches pulled from transfected HEK293 cells were activated by brief (<1 ms) exposure to agonists using a piezo-electric translator (Siskiyou). The speed of the solution exchange system was regularly calibrated by rapidly switching the solution perfusing an open patch pipette between standard extracellular solution and an extracellular solution that had been diluted by 50% with distilled water. By monitoring the resulting pipette current, we were able to ensure that the solution perfusing the macropatch was completely exchanged within 200 μs (Dixon et al., 2014). Recordings were performed using a Multiclamp 700B amplifier and pClamp9 software (Molecular Devices), filtered at 4 kHz and sampled at 10 kHz.

Analyses of IPSC amplitudes, 10%–90% rise times, and weighted decay time constants were performed using Axograph (Axograph Scientific). Only cells with a stable series resistance of <25 MΩ throughout the recording period were included in the analysis. Single peak IPSCs with amplitudes of at least three times above the background noise were detected using a semi-automated sliding template. Each detected event was visually inspected and only well-separated IPSCs with no inflections in the rising or decay phases were included. To calculate macroscopic current decay time constants, averaged macroscopic traces 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). The averaged data from individual cells were then pooled to obtain group data. Statistical analysis, and plotting were performed with Prism 5 (GraphPad Software). The fitting of single Gaussian functions to IPSC amplitude and decay time constant distributions was also performed using Prism 5. All data are presented as mean ± SEM. Student’s unpaired t-tests or one-way ANOVAs, as appropriate, were employed for comparisons. For all tests, the number of asterisks corresponds to level of significance: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. Single-channel recordings were analyzed with pClamp 10 (Clampfit, Molecular Devices) or QuB software. Segments of single-channel activity separated by long periods of baseline were idealized into noise-free open and shut events using a temporal resolution of 70 μs. Single-channel activations were separated using a shut period (t_crit) ranging between 6 ms and 30 ms, however most of the t_crit values were <10 ms. Group data form current-voltage experiments were fitted to a polynomial using Sigmaplot (Systat Software), from which reversal potential was obtained. Ohm’s Law was used to find single-channel conductance (γ), in which V_hold is the holding potential (−70 mV), V_ljp is the liquid junction potential (4.7 mV for the solutions used) and V_rev is the reversal potential as follows: γ = (i)/(V_hold − V_ljp − V_rev).

Previously reported GLRA2 mutations associated with ASD include a microdeletion (GLRA2∆ex8-9; Pinto et al., 2010) and de novo missense mutations p.N109S (Iossifov et al., 2014) and p.R126Q (Pilorge et al., 2016) identified in the hemizygous state in males with non-syndromic autism. Additional clinical symptoms were noted in two of these individuals, including language delay with functional language and low average IQ (GLRA2∆ex8-9) and hyperactivity, severe language delay and tonic-clonic seizures (p.R126Q). These mutations were classified as loss of function based on impaired cell-surface expression of GlyR α2, and abolition (GLRA2∆ex8-9) or markedly reduced sensitivity to glycine (p.N109S and p.R126Q; Pilorge et al., 2016). These recessive mutations were not found to be disease causing in female carriers, due to the presence of a normal GLRA2 allele, and because GLRA2 escapes X-inactivation in the vast majority of tissues including brain (Cotton et al., 2015). However, many mutations in the GlyR α1 subunit gene (GLRA1) that cause startle disease show dominant inheritance (Harvey et al., 2008; Bode and Lynch, 2014). We therefore decided to examine the functional effects of potentially pathogenic variants found in females. In particular, we focussed on p.R323L (p.R350L in the GlyR α2 subunit with signal peptide; c.1049G>T in NM_001118885). This missense mutation was previously reported in a female with ASD (Piton et al., 2011) that was predicted to be damaging by a number of software packages (including PolyPhen-2, SNPs&GO, MutPred, PANTHER and SIFT; Pilorge et al., 2016). This mutation is also extremely rare, as it is not reported in the Genome Aggregation Database (gnomAD¹) currently comprising 126,216 exome sequences and 15,136 whole-genome sequences from unrelated individuals. The mutated arginine is located the large M3–M4 intracellular loop of the GlyR α2 subunit and is highly conserved among GlyR subunits (Figure 1A). Additional clinical symptoms reported in this case include loss of acquired words, seizures, mild motor developmental delay, macrocephaly and hypothyroidism (Gauthier, personal communication).

To determine the functional effects of the GlyR α2^R323L mutation, we examined the potency of glycine in activating recombinant homomeric α2^R323L and heteromeric α2^R323Lβ receptors. Figure 1B illustrates whole-cell currents recorded in response to increasing concentrations of glycine in HEK293 cells expressing wild-type GlyRs or those containing α2^R323L. The glycine dose-response curve of both homomeric α2^R323L and heteromeric α2^R323Lβ GlyRs were modestly right-shifted, with EC₅₀ values of 67.9 ± 13.4 (α2^R323L) and 143.0 ± 12.9 μM (α2^R323Lβ), compared with 46.6 ± 6.7 and 90.3 ± 14.6 μM for the corresponding wild-type α2 and α2β GlyRs (Table 1, Figure 1C). Thus, in both homomeric and heteromeric GlyRs, the p.R323L mutation results in a small decrease in apparent glycine sensitivity. There was no significant difference in peak whole-cell currents in all the receptors tested (range, ~2–1.5 nA, Table 1) suggesting that mutations of R323 do not alter GlyR cell-surface expression.

This modest change in dose-response relationship is insufficient to explain the pathogenic effects of the p.R323L mutation, especially since the peak glycine concentration in the synaptic cleft is thought to reach 1–3 mM in embryonic zebrafish neurons (Legendre, 1998) and 2.2–3.5 mM in adult rat spinal neurons (Beato, 2008). Clearance of glycine away from the cleft has also been estimated to occur on a 0.6–0.9 ms time scale (Beato, 2008). Based on these parameters, we examined intrinsic channel properties by rapidly applying saturating glycine (3 mM) for a period of ~1 ms to excised outside-out HEK293 cell patches expressing wild-type and mutant α2^R323Lβ receptors. Sample macroscopic currents are shown in Figure 2A. The time course of deactivation was fitted by a double exponential function with a mean time constant of 61.9 ± 3.2 ms (n = 15) for α2β GlyRs and a mean time constant of 87.9 ± 5.4 ms (n = 17) for α2^R323Lβ GlyRs (Figure 2B, Table 2). The deactivation time constant for α2^R323Lβ receptors was ~1.5-fold slower compared to wild-type GlyRs (p < 0.05). However, the rise time did not differ between wild-type and mutant receptors (α2β, 1.3 ± 0.1 ms; α2^R323Lβ, 1.4 ± 0.2 ms; p > 0.05). This suggests that the p.R323L mutation in M3–M4 loop enhances channel function by slowing the channel closing rate. An analysis of the individual components of the double exponential fit revealed that the longer time constant and the fraction of the total current it represents increased (Table 3).

To test whether the enhancement of GlyR function altered glycinergic transmission, we used a heterosynapse system that allows control over the subunit composition of GlyRs in glycinergic synapses (Zhang Y. et al., 2015). We inserted α2β and α2^R323Lβ GlyRs into heterosynapses in turn to evaluate the properties of the resulting IPSCs. Sample IPSC recordings from heterosynapses incorporating α2β and α2^R323Lβ isoforms are shown at different temporal resolution in Figure 3A with averaged normalized IPSCs. The averaged IPSC amplitudes, 10%–90% rise times and decay time constants for α2β and α2^R323Lβ GlyRs are presented in Figure 3B and Table 2. We observed that IPSCs mediated by α2^R323Lβ receptors displayed a 2-fold slower rise and decay time than those mediated by wild-type α2β GlyRs (10%–90% rise time: α2β, 2.6 ± 0.3 ms, n = 7; α2^R323Lβ, 4.9 ± 0.8 ms, n = 6, p < 0.01; decay time: α2β, 27.0 ± 1.5 ms; α2^R323Lβ, 60.8 ± 4.9 ms, p < 0.0001), whereas the mean IPSC amplitude did not differ significantly (α2β, 32.8 ± 2.5 pA; α2^R323Lβ, 34.0 ± 11.9 pA; p > 0.05). We also sought to investigate the effects of the previously described autism mutations, α2^N109S and α2^R126Q (Pilorge et al., 2016), in heterosynapses, but heteromeric GlyRs incorporating these mutations yielded no detectable synaptic currents (n > 10 cells expressing each mutant GlyR). This result was expected given that both mutations dramatically increased the glycine EC₅₀ (Pilorge et al., 2016).

We next sought to investigate whether the nature of the side chain, charge or the steric characteristics of the p.R323L substitution contributes directly to the observed increase in intrinsic channel deactivation rates. Arginine is basic and polar (positively charged) with a 3-carbon aliphatic straight chain, capped at the distal end by a complex guanidinium group. We replaced R323 by alanine (α2^R323A) which is aliphatic, uncharged and has a short side chain consisting of a single methyl group. We also examined substitutions with lysine (α2^R323K, which is basic, and has a positively charged ε-amino group) and isoleucine (α2^R323I, aliphatic, non-polar and differing from leucine only in the position of a side chain methyl group). We then characterized the effects of each GlyR α2 mutant on the kinetics of heterosynaptic IPSCs. Figure 3C shows sample recordings from heterosynapses incorporating the GlyR α2^R323A, α2^R323K and α2^R323I constructs, and averaged normalized IPSCs are presented in the right panel. All three substitutions produced heterosynaptic IPSCs that had kinetics similar to those of wild-type α2β GlyRs. Again, similar peak currents between mutant and wild-type receptors in macropatch and heterosynapse recordings suggests comparable surface expression (Table 2).

Since another pathomechanism associated with GlyR mutations is spontaneously-opening channels (Chung et al., 2010; Bode et al., 2013; James et al., 2013; Zhang et al., 2016), we examined the activity of α2β and α2^R323Lβ GlyRs in outside-out patches. No spontaneous activity was observed when patches containing α2β and α2^R323Lβ GlyRs were perfused with glycine-free solution for more than 1 min (Figures 4A,B above). However, at a saturating 100 μM glycine concentration, active periods were induced and observed as clusters of openings (Figures 4A,B). The current amplitude at −70 mV was determined for α2β and α2^R323Lβ GlyRs by plotting amplitude histograms and fitting these to Gaussian functions. GlyRs containing the α2β subunit combination displayed currents of ~3.2 pA, whereas single-channel currents for α2^R323Lβ GlyRs were ~3.8 pA at −70 mV (Figures 4A,B below). The single-channel conductance was determined for both α2β and α2^R323Lβ GlyRs by carrying out current-voltage experiments over a range of voltages from −70 mV to +70 mV and averaged from three to five patches (Figures 5A,B). Mild inward rectification was observed in the current-voltage plots of both channels (Figures 5C,D). Current-voltage plots for heteromeric α2β and α2^R323Lβ GlyRs intersected the voltage axis at ~+3 mV (Figure 5C) and ~+5.5 mV (Figure 5D), respectively. Mean single-channel conductance levels were calculated to be 41.2 pS for α2β and 47.5 pS for α2^R323Lβ and were statistically different (p < 0.05). Thus, the conductance of α2^R323Lβ GlyRs is slightly larger than that for wild-type α2β GlyRs, but less than that previously reported for homomeric α2 subunit GlyRs (60–120 pS; Wang et al., 2006; Krashia et al., 2011).

Finally, we investigated single-channel kinetics for α2β and α2^R323Lβ GlyRs. Samples of opening durations for α2β and α2^R323Lβ are shown in Figures 6A,B, respectively. The duration of activations and open probabilities (P_o) for two channel types were assessed at 100 µM and 3 mM glycine (Table 4). At both concentrations, the mean duration of active periods for α2^R323Lβ GlyRs was longer than that observed for α2β GlyRs (Figure 6C). It is noteworthy that there was a high standard deviation in the values for wild-type α2β GlyRs at 3 mM glycine. Although the P_o values of the two receptors were indistinguishable at 3 mM glycine, at 100 µM glycine α2^R323Lβ showed a significant increase in P_o compared to α2β (Figure 6D).

In summary, the increase in decay times for synaptic and macropatch currents, the small but significant increase in single-channel conductance and the prolongation of individual receptor active periods clearly demonstrates that overall, the R323L mutation confers a gain-of-function.