Functional genomics of epilepsy-associated mutations in the GABAA receptor subunits reveal that one mutation impairs function and two are catastrophic

A number of epilepsy-causing mutations have recently been identified in the genes of the α1, β3, and γ2 subunits comprising the γ-aminobutyric acid type A (GABAA) receptor. These mutations are typically dominant, and in certain cases, such as the α1 and β3 subunits, they may lead to a mix of receptors at the cell surface that contain no mutant subunits, a single mutated subunit, or two mutated subunits. To determine the effects of mutations in a single subunit or in two subunits on receptor activation, we created a concatenated protein assembly that links all five subunits of the α1β3γ2 receptor and expresses them in the correct orientation. We created nine separate receptor variants with a single-mutant subunit and four receptors containing two subunits of the γ2R323Q, β3D120N, β3T157M, β3Y302C, and β3S254F epilepsy-causing mutations. We found that the singly mutated γ2R323Q subunit impairs GABA activation of the receptor by reducing GABA potency. A single β3D120N, β3T157M, or β3Y302C mutation also substantially impaired receptor activation, and two copies of these mutants within a receptor were catastrophic. Of note, an effect of the β3S254F mutation on GABA potency depended on the location of this mutant subunit within the receptor, possibly because of the membrane environment surrounding the transmembrane region of the receptor. Our results highlight that precise functional genomic analyses of GABAA receptor mutations using concatenated constructs can identify receptors with an intermediate phenotype that contribute to epileptic phenotypes and that are potential drug targets for precision medicine approaches.

A number of epilepsy-causing mutations have recently been identified in the genes of the ␣1, ␤3, and ␥2 subunits comprising the ␥-aminobutyric acid type A (GABA A ) receptor. These mutations are typically dominant, and in certain cases, such as the ␣1 and ␤3 subunits, they may lead to a mix of receptors at the cell surface that contain no mutant subunits, a single mutated subunit, or two mutated subunits. To determine the effects of mutations in a single subunit or in two subunits on receptor activation, we created a concatenated protein assembly that links all five subunits of the ␣1␤3␥2 receptor and expresses them in the correct orientation. We created nine separate receptor variants with a single-mutant subunit and four receptors containing two subunits of the ␥2 R323Q , ␤3 D120N , ␤3 T157M , ␤3 Y302C , and ␤3 S254F epilepsy-causing mutations. We found that the singly mutated ␥2 R323Q subunit impairs GABA activation of the receptor by reducing GABA potency. A single ␤3 D120N , ␤3 T157M , or ␤3 Y302C mutation also substantially impaired receptor activation, and two copies of these mutants within a receptor were catastrophic. Of note, an effect of the ␤3 S254F mutation on GABA potency depended on the location of this mutant subunit within the receptor, possibly because of the membrane environment surrounding the transmembrane region of the receptor. Our results highlight that precise functional genomic analyses of GABA A receptor mutations using concatenated constructs can identify receptors with an intermediate phenotype that contribute to epileptic phenotypes and that are potential drug targets for precision medicine approaches.
GABA A receptors are essential mediators of neurotransmission in both the developing and adult brain (17). These receptors are ion channels composed of five subunits that arrange around a central ion pore (18). When GABA is released from the synapse, it binds to these receptors anchored at the postsynaptic membrane to open an ion channel, allowing chloride ions to pass, hyperpolarizing and inhibiting the cell (19).
Many genes encoding for different subunits of the GABA A receptor are present in the mammalian brain, including six ␣ (␣1-6), three ␤ (␤1-3), three ␥ (␥1-3), and a ␦, ⑀, and , and the majority of receptors are thought to contain two ␣, two ␤, and one ␥ subunit where they are anchored at the synapse, responding to high concentrations of GABA (20). Other combinations of receptors, often containing ␦ subunits, are found extrasynaptically, where they respond to low concentrations of GABA or spillover from the synapse (21). Each individual subunit consists of a large extracellular domain, four transmembrane domains where the second (M2) lines the channel pore, two short and one large loop linking transmembrane domains, and a short C terminus. At synaptic receptors, GABA binds within the ␤-␣ interface located between adjacent extracellular domains to trigger an activation pathway through a series of conformational changes that ultimately open the channel pore. These conformational changes are transmitted through interactions at the coupling region, where loops in the extracellular domain in close proximity to the membrane interact with the pre-M1 and M2-M3 loops that connect transmembrane domains (22). This results in tilt of the M2 domain to open the pore. Epilepsy-causing mutations identified in the ␣1, ␤3, and ␥2 subunits are located at different regions throughout the protein, including amino acids throughout the activation pathway from the ligand-binding pocket and extracellular structural ␤-sheets, through to the coupling and transmembrane M1 and M2 regions.
Mutations in the GABA A receptor that cause epilepsy typically impair this process, either through misfolding of protein to reduce the number of receptors at the cell surface or disturbing the ability of the receptor to open in response to GABA (8). In all cases, the mutations that cause epileptic encephalopathies are dominant, with patients carrying one copy of the WT allele and one copy of the mutant allele (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). For mutations in the ␥2 subunit, the resultant receptors will either be a WT or contain a single mutation. However, for ␤3 mutations, a more complicated mixture of receptors will be expressed. A WT containing two normal ␤3 subunits, two heteromutant receptors containing a single mutation at either of the two ␤3 subunit locations within the pentamer, and a homomutant receptor containing the mutation at both ␤3 subunit locations within the pentamer can be formed. If the surface expression is unaffected and the distribution of mutant receptors into the complex is random, some 50% of the resultant receptors will contain mutations with just a single copy of the mutation. To date, there is a lack of research that assesses the effect of a single copy of the mutation on GABA A receptor function, an important component that contributes to understanding the epilepsy phenotype of individuals. Therefore, it is vital to determine how single copies of the mutation, as well as two copies, alter the function of the receptor to properly characterize the molecular phenotype of the mutation.
We chose five mutations to investigate using the concatenated construct including one in the ␥2 subunit (␥2 R323Q ) and four in the ␤3 subunit (␤3 D120N , ␤3 T157M , ␤3 S254F , and ␤3 Y302C ) (Fig. 1A). These mutations were chosen as they were located in different regions along the activation pathway of the receptor, including the M2-M3 coupling loop of the ␥2 subunit (␥2 R323Q ). The ␤3 mutations were located in the area surrounding the ligand-binding site (␤3 D120N ), a ␤-sheet within the extracellular domain (␤3 T157M ), the M2-M3 coupling loop (␤3 Y302C ), and the M1 transmembrane region (␤3 S254F ) (Fig. 1, B and C). Previous functional genomic analysis of these mutations in Xenopus oocytes and HEK293 cells have demonstrated that the ␥2 R323Q , ␤3 D120N , and ␤3 Y302C mutations substantially reduce either the potency of GABA or the magnitude of GABAactivated currents when expressed in ␣1␤3␥2 or ␣5␤3␥3 receptors (8,11,15), whereas the ␤3 T157M mutation caused only subtle changes at ␣5␤3␥3 receptors (8). However, these experiments did not fully describe the molecular phenotype of the mutations, as they were unable to distinguish how receptors that contain one or two copies of the mutation differ from the WT receptor or from each other.
To resolve this question, we created a concatenated ␣1␤3␥2 GABA A receptor construct with five linked subunits in the sequence ␥2-␤3-␣1-␤3-␣1. For each mutation, we then created a set of receptor constructs that resembled the expressed receptors from an individual with a dominant mutation. Typically, but not always, a single copy of the mutation impaired the activation properties of the receptor, whereas a second copy intensified the effect of the mutation to be catastrophic.
We propose that precise functional genomic analysis using concatenated receptors can identify the phenotype of the individual receptors that are expressed by patients with dominant mutations. This information may ultimately assist in precision medicine approaches, where mutant GABA A receptors containing a single mutation can be targeted to treat individual patients.

Activation properties of WT concatenated receptor
To determine how receptors with either a single or two copies of an epilepsy-causing mutation differ from WT receptors, we created a concatenated receptor construct with five subunits linked by AGS repeats (Fig. 2A). Although concatenated ligand-gated ion channels have previously been created, many of these do not reliably form in the standard orientation (23) or are mixtures of dimeric and trimeric concatenated constructs (24). When injected alone, these dimeric and trimeric constructs can result in low currents that could potentially confound the analysis of mutations that impair receptor function (25). However, constructs of concatenated pentameric GABA A Figure 1. A, pentameric structure of the GABA and diazepam-bound ␣1␤3␥2 receptor (PDB code 6HUP) from above showing the orientation of the subunits. The subunits are colored with respect to their order in the pentamer: first ␥2 (green), second ␤3 (maroon), third ␣1 (blue), fourth ␤3 (red), and fifth ␣1 (dark blue). B, side view showing the first ␥2 subunit adjacent to the second ␤3 subunit. The ␥2 R323 residue on the M2-M3 loop is depicted with the side chain in black in a transparent sphere. C, side view showing the fourth ␤3 subunit adjacent to the fifth ␣1 subunit, with the GABA-binding site highlighted. The side chains of the ␤3 D120N residue at the GABA-binding site, the ␤3 T157M residue within an internal ␤-sheet, the ␤3 Y302C residue in the M2-M3 loop, and the ␤3 S254F residue in the M1 region are shown in black.

Concatenated GABA A receptor epilepsy mutations
receptors have also been described that contain ␤2 subunits (26). Therefore, the new construct was designed with five subunits in the same order so that the subunits would arrange themselves primarily counter-clockwise when viewed from the extracellular side of the membrane, forming two GABA-binding sites at the ␤3-␣1 interfaces and a benzodiazepine-binding site at the ␣1-␥2 interface (Fig.  2B). Sequence encoding for the signal peptide of the ␣1 and ␤3 subunits were removed, and four different linkers were incorporated with lengths calculated with the same methodology as previously (23). To ensure predominately counterclockwise expression, the first linker was designed to be relatively short, and subsequent linkers were longer with similar lengths when N and C termini were taken into account. These linkers contained peptide sequences of (AGS) 5 between the ␥2 and ␤3 subunits ((AGS) 5 LGS(AGS) 3 between the first ␤3 and ␣1 subunits, AGT(AGS) 5 between the ␣1 and ␤3 subunits, and (AGS) 4 ATG(AGS) 4 between the final ␤3 and ␣1 subunits) to form the DNA construct encoding the five subunits in the order of ␥2-␤3-␣1-␤3-␣1 (Fig. 2,  A and B).
cRNA (2 ng) of the WT concatenated construct was injected into Xenopus oocytes, and the oocytes were incubated for 2-4 days. Denaturing agarose gel electrophoresis was performed on the RNA to ensure that a single band at the correct size only was transcribed, and Western blotting was performed to ensure that the protein was properly translated and degradation products were not observed (Fig. S1). Peak currents were measured using two-electrode voltage clamp electrophysiology upon application of a range of GABA solutions, and the measured responses were used to construct the concentration-response curve (Fig. 2C). Injection of the WT cRNA resulted in robust GABA-activated currents, with 3 mM GABA eliciting an average current of 2 A (Fig. 2C and Table 1). GABA activated the WT concatemer with an EC 50 of 69 M, similar to our previously published value of 53 M and other previously published reports (e.g. 74 M (27)) where unlinked ␣1, ␤3, and ␥2 subunits were injected into Xenopus oocytes (28).
To ensure that the receptors were arranging in the correct orientation, the modulation of GABA-elicited currents of our concatenated receptors was measured using a benzodiazepine, clobazam, that binds selectively to the ␣1-␥2 inter- . D, representative data (above) from a single two-electrode voltage clamp experiment constructing a modulation curve to clobazam (below). Three pulses of reference 10 M GABA (closed bars) were applied prior to co-application of 10 M GABA and clobazam at concentrations shown (closed bars). Percent modulation of the control GABA response was calculated, and the mean Ϯ S.E. was plotted and fitted to the Hill equation (below). The fitted EC 50 of clobazam was 86 nM (log EC 50 ϭ Ϫ4.03 Ϯ 0.06, mean Ϯ S.E., n ϭ 10), and the fitted E max was 306% (320 Ϯ 32, mean Ϯ S.E., n ϭ 10).

Concatenated GABA A receptor epilepsy mutations
faces. When clobazam was co-applied with 10 M GABA, the response of the activated receptors was increased with increasing clobazam concentrations (Fig. 2D). We constructed a concentration-response curve of clobazam modulation of 10 M GABA-activated currents. The maximum modulation by clobazam was 306%, and the EC 50 value of clobazam was 86 nM, similar to previously published values of 256% and 132 nM at nonconcatenated ␣1␤2␥2 receptors (29) (Fig. 2D). Taken together, the GABA and clobazam concentration-response curves demonstrate that the concatenated receptor reliably replicates the activation properties of its respective unlinked receptor. We then used this construct as a backbone so that mutant ␤3 and/or ␥2 subunit(s) can be inserted at specific regions of the pentameric construct to analyze the effects of epilepsy-causing mutations. For ␥2 mutations, a single copy of the mutation was inserted into the receptor, and for ␤3 mutations, either a single copy of the mutation was inserted within different subunits or two copies of the mutation were inserted into the receptor.

Absolute expression levels of mutant receptors
We chose five mutations to investigate using concatemers, including one in the ␥2 subunit (␥2 R323Q ) and four in the ␤3 subunit (␤3 D120N , ␤3 T157M , ␤3 S254F , and ␤3 Y302C ). These mutations were chosen as they are located in different regions along the activation pathway of the receptor. In order from the extracellular to transmembrane domains, the amino acids included the area surrounding the ligand-binding site (␤3 D120N ), a ␤-sheet located in the extracellular domain of the ␤3 subunit (␤3 T157M ), the M2-M3 coupling loop of both the ␥2 (␥2 R323Q ) and ␤3 subunit (␤3 Y302C ), and the M1 region (␤3 S254F ). The ␤3 D120 and ␤3 Y302 residues are located at the interface of the ␣1 and ␤3 subunits, the ␤3 T157M residue is located within the ␤-sheet of the ␤3 subunits, and the ␤3 S254 residue is located at the interface of a ␤3 and ␣1 subunit or at the interface of a ␤3 and ␥2 subunit.
For the ␥2 mutation, we introduced a single copy of the ␥2 R323Q mutation into the first subunit of the concatenated construct. For each of the ␤3 mutations, we created a set of three constructs with a mutation in either the second or fourth ␤3 subunit and a construct with a mutation in both the second and fourth ␤3 subunits (Fig. 3A). We injected 2 ng of cRNA encoding for each of the constructs and then compared the absolute currents elicited by 3 mM GABA at the mutant receptors with the WT (Fig. 3B).
Although several concatemers containing single mutations had significant reductions in the maximum absolute currents elicited by 3 mM GABA, this crude approach is a poor measure of how mutations alter receptor properties. Variation in the maximum absolute current can be introduced in several ways Table 1 Concentration-response curves of ␣1␤3␥2 receptors Results shown are mean Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; one-way ANOVA with Tukey's post hoc test.

Concatenated GABA A receptor epilepsy mutations
that are a consequence of experimental conditions. These include large rightward shifts in the EC 50 and small changes in the RNA concentration, the incubation time, and the rate at which the individual oocytes form protein and express receptors at the cell surface. However, mutations may also cause changes in the intrinsic activation properties of the receptor that reduce the current passing across the synapse. This can occur through changes in the potency of GABA or changes in the efficacy, where GABA reverts to a more partial agonist, or both. To determine whether the mutations changed either the potency or efficacy of GABA activation, we next constructed concentration-response curves to GABA and estimated the maximum open probability of GABA to determine whether the mutations changed these intrinsic activation properties of the receptor, when either one or two copies of the mutation were present.

␥2 R323Q impairs GABA activation properties of the receptor
We therefore constructed concentration-response curves to GABA and standardized the response against an estimated maximum open probability of the receptor for the WT and each mutation. We initially compared the ␥2 R323Q -␤3-␣1-␤3-␣1 mutant receptor with the WT, as this mutation is incorporated into the receptor within the only ␥2 subunit (Fig. 4A). The potency of GABA has previously been shown to be reduced at ␣1␤2␥2 receptors when expressed in HEK293 cells, and we would expect comparable results in our concatenated construct (15).

Concatenated GABA A receptor epilepsy mutations
We first tested whether the ␥2 R323Q mutation altered the potency of GABA. We constructed concentration-response curves to GABA at ␥2 R323Q -␤3-␣1-␤3-␣1 receptors to determine the EC 50 of the mutant receptors. These experiments were run on an automated protocol, where 3 mM GABA was applied as a reference and internal standard three times during the experiment for all receptors (Fig. 4A). Similar to the results reported using this mutation with free ␣1 and ␤3 subunits in HEK293 cells (15), there was a decrease in the potency of GABA with a significant 4.5-fold decrease in the potency of GABA (EC 50 ϭ 315 M) ( Fig. 4A and Table 1), demonstrating that the activation of receptors by GABA is impaired by the ␥2 R323Q mutation.
We then determined whether the maximal efficacy of GABA was impaired by the ␥2 R323Q mutation by estimating the maximum P o (Est. P o(max) ) at WT and ␥2 R323Q -␤3-␣1-␤3-␣1 receptors using a pharmacological technique similar to Shin et al. (30). At oocytes expressing either WT or mutant receptors, we applied the 3 mM GABA reference and then co-applied 10 mM GABA with 1 M etomidate and 3 M diazepam to shift as many receptors as possible to the open state (Fig. 4B). We assumed that the combination of GABA with etomidate and diazepam opened the receptors with a probability approaching 1. The Est. P o(max) of GABA for each receptor was then calculated by dividing the current elicited by 3 mM GABA by the current elicited by GABA, etomidate, and diazepam and corrected to account for shifts in the concentration-response curves ( Fig. 4B and Table   1). We refer to Est. P o(max) as an estimated maximum open probability as the true current amplitude may be underestimated by mutations that greatly impair receptor activation or modulation or change desensitization kinetics. As expected, GABA elicited a very high Est. P o(max) of 0.95 at WT receptors, consistent with single-channel recordings where the channel enters a long-lived open state (31). The ␥2 R323Q mutation did not significantly alter the Est. P o(max) with a value of 0.85 ( Fig.  4B and Table 1).
When incorporated into the first subunit of the concatemer, the ␥2 R323Q mutation reduced the potency of GABA without causing a significant reduction in the efficacy. These changes in the activation properties of the receptor caused by the ␥2 R323Q mutation in the concatenated construct were similar to the reported effects of receptors composed of unlinked subunits. Therefore, the concatenated construct is a suitable method of analyzing the effect of mutations on the activation properties of the receptor.

A single ␤3 D120N or ␤3 T157M mutation impairs GABA potency, and two mutations are catastrophic
We next assessed the effects of two ␤3 mutations, ␤3 D120N and ␤3 T157M , that are both located in the extracellular domain at the earlier stages of the activation pathway. The ␤3 D120N mutation has previously been expressed in combination with either ␣1, ␤3, and ␥2 or ␣1 and ␥2 free subunits to determine the effects of heterozygous or homozygous expression, respec-

Concatenated GABA A receptor epilepsy mutations
tively, on receptor function (11). Both the gating properties and the absolute expression levels of the receptor were reduced in both cases, whereas in a separate study, the incorporation of the ␤3 T157M mutation with ␣5 and ␥2 subunits only made subtle changes to the activation properties of the receptor (8). To determine the effect of the mutations when they were expressed in a single subunit within the receptor, we constructed concentrationresponse curves to GABA and measured the Est. P o(max) at concatenated receptors containing a single copy of a mutation.
We created concatenated constructs by introducing ␤3 D120N or ␤3 T157M mutations in the second, the fourth or both the second and fourth subunits in the concatemer and constructed concentration-response curves to GABA (Fig. 5A). A single copy of the ␤3 D120N mutation significantly reduced the potency of GABA by 16 -20-fold, regardless of whether it was introduced at the second or fourth subunit (EC 50 ϭ 1.14 and 1.47 mM, ␥2-␤3 D120N -␣1-␤3-␣1 and ␥2-␤3-␣1-␤3 D120N -␣1, respectively) (Fig. 5B and Table 1). These EC 50 values were not significantly different from each other, strongly suggesting that the subunit location of the ␤3 D120N mutation within the pentameric structure did not affect how the mutation altered receptor activation properties.
Similarly, the ␤3 T157M mutations significantly reduced the potency of GABA by 4 -6-fold when introduced at either the second or fourth subunit (EC 50 ϭ 422 and 279 M, ␥2-␤3 T157M -␣1-␤3-␣1 and ␥2-␤3-␣1-␤3 T157M -␣1, respectively) ( Fig. 5C and Table 1). Again, the EC 50 values were not significantly different from each other, demonstrating that the subunit location of the ␤3 T157M did not affect how the mutation altered the activation properties of GABA. Neither receptor expressed measurable currents when a mutation was introduced in both ␤3 subunits, and as such, concentration-response curves could not be constructed. We next measured the Est. P o(max) of two receptors with a single copy of the mutation in the second or fourth subunit, respectively, to determine whether the efficacy of GABA had been altered (Fig. 5D). Despite the absolute current levels being reduced by the ␤3 D120N mutation, the combination of etomidate and diazepam failed to appreciably increase the maximal response to GABA at either receptor containing a single mutation (Fig. 5, D and E). This is likely due in part to the large rightward shift of the concentration-response curve where 3 mM GABA no longer elicits the maximum response. Similarly, the combination of etomidate and diazepam had little effect on the maximal GABA current elicited at the two receptors containing a single ␤3 T157M mutation (Fig. 5, D and E). Consequently, there was no significant difference in the Est. P o(max) at the four receptors (Est. P o(max) ϭ 1.12, 1.00, 0.9, and 0.96,
Taken together, single copies of the ␤3 D120N and ␤3 T157M mutations at the earlier stages of the activation pathway both impair GABA activation of the receptor by reducing the potency of GABA by ϳ20and 5-fold, respectively, without altering the maximal efficacy of GABA. There was little difference in the effect of either single mutation when located at different subunits ␤3 within a pentamer. A second copy of the ␤3 D120N or ␤3 T157M mutation intensifies the effect of the mutation and appears to be catastrophic, leading to little to no functional receptor expression.

␤3 S254F and ␤3 Y302C mutation effects are dependent on location and number of mutations
We next assessed the effects of two other ␤3 mutations, ␤3 Y302C and ␤3 S254F . The ␤3 Y302 residue is located in the M2-M3 coupling loop, a key motif in the activation pathway that links extracellular and transmembrane domains. The ␤3 S254 residue is located in the transmembrane regions within the M1 transmembrane helix that moves late in the activation process of ligand-gated ion channels (32). The ␤3 Y302C mutation has been shown to impair receptor activation when expressed with either ␣1 and ␥2 subunits or ␣5 and ␥2 receptors (8,11), whereas there are no functional data on how the ␤3 S254F mutation affects receptor activation.

Concatenated GABA A receptor epilepsy mutations
EC 50 value of the WT or the two concatemers containing a single ␤3 Y302C mutation ( Fig. 6A and Table 1).
Additionally, the introduction of the ␤3 Y302C mutation significantly reduced the Est. P o(max) compared with WT, regardless of the subunit location of the mutation or whether one or two copies of the mutation were introduced. The efficacy of GABA was least affected by one copy of the ␤3 Y302C mutation at the fourth subunit (Est. P o(max) ϭ 0.74, ␥2-␤3-␣1-␤3 Y302C -␣1), whereas a single copy of the ␤3 Y302C mutation at the second subunit significantly reduced the efficacy of GABA compared with the WT or the receptor with a single mutation at the fourth subunit (Est. P o(max) ϭ 0.4, ␥2-␤3 Y302C -␣1-␤3-␣1). Two copies of the ␤3 Y302C mutation resulted in the lowest efficacy of GABA (P o(max) ϭ 0.24, ␥2-␤3 Y302C -␣1-␤3 Y302C -␣1) (Fig. 6 (B and C) and Table 1). This demonstrates that, unique among the mutations that we have investigated, the single ␤3 Y302C mutation impairs activation of the receptor to decrease both the potency and maximum efficacy of GABA activation. Differences in the magnitude of the reduction in the maximal efficacy suggest that these residues may not be equivalent when located in different subunits.
The introduction of the ␤3 S254F mutation did not follow the same pattern as the other mutations, which reduced the potency of GABA regardless of the subunit location of the mutation. Instead, when the mutation was introduced at the second subunit, the EC 50 value of GABA was significantly increased nearly 3-fold compared with the WT (EC 50 ϭ 181 M, ␥2-␤3 S254F -␣1-␤3-␣1), demonstrating that the activation properties of this concatemer were impaired ( Fig. 6D and Table 1). However, when the mutation was introduced at the fourth subunit, the EC 50 value of GABA significantly decreased 2-fold compared with the WT, as did the EC 50 when the ␤3 S254F mutation was introduced at both the second and fourth subunits (EC 50 ϭ 34.2 and 29.6 M, ␥2-␤3-␣1-␤3 S254F -␣1 and ␥2-␤3 S254F -␣1-␤3 S254F -␣1, respectively) ( Fig. 6D and Table 1). This demonstrates that the subunit location of the ␤3 S254F mutation defines the functional effect of the receptor, determining whether the potency of GABA has increased or decreased.

Discussion
Recent advances in whole-genome sequencing have enabled the identification of a large number of de novo mutations that cause a range of severe childhood epilepsies. In all of these cases, the mutations are dominant (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16), whereby patients will contain one WT and one mutant copy of the gene. In cases where the mutations are in the ␤3 subunit of the GABA A receptor, this will lead to several potential receptors being expressed in each subtype, with heteromutant receptors containing single mutations at either of the two ␤3 subunits and a homomutant receptor containing mutations at both ␤3 subunits. Precise functional genomic analysis requires the understanding of how each of these individual receptors are affected by the mutation, as these receptors could be expressed and contribute to the pathology of the disorder or even be targeted by GABAergic drugs to treat the seizures.
The in vitro analysis of these mutations has, to date, relied solely on injection or transfection of WT and/or mutant subunits in heterologous systems and quantification of receptor expression levels complemented with whole-cell recording and, at times, single-channel analysis (8,11,15). However, this approach is inadequate to describe the entire molecular phenotype, as mixed populations of receptors with one or two mutations will form at a ratio of 2:1. This will be particularly problematic when the maximal efficacy is reduced by the mutation, as the higher efficacy of the WT receptor will dominate the signal in whole-cell recordings. Therefore, by using a concatenated receptor construct, we have derived results from heteromutant ␤3 receptors that provide significant insights into the molecular phenotypes of epilepsies caused by GABA A receptor mutations as well as knowledge about the activation mechanisms of these receptors.

Mutations impair synaptic transmission through efficacy or potency of GABA activation
There is an enormous amount of understanding of how GABA A R opening is triggered through an activation pathway initiated by GABA binding that then opens the intrinsic ion channel to mediate neuronal inhibition (22). Briefly, the agonist binds at the interface between a ␤3 and an ␣1 subunit at the extracellular domain, causing a series of conformational changes within the receptor that lead to the transmembrane domains. At the interface of the extracellular and transmembrane domains, a set of interacting loops, including the ␤1-␤2, ␤6-␤7, and ␤8-␤9 in the extracellular domain and pre-M1 and M2-M3 loops connecting the transmembrane domain, alter their conformation during receptor activation (33,34). This leads to a tilting of the M2 helices and ultimately channel opening. Additionally, the number of molecules bound is important, where two molecules of agonist are required to be bound to fully activate the receptor (35).
Synaptic receptors that contain a ␥2 subunit typically have intrinsic activation properties distinct from those of extrasynaptic receptors that contain a ␦ subunit, including a high efficacy where the maximal open probability elicited by GABA approaches 1 and lower potency (31,36). The maximal efficacy of an agonist, or whether the agonist is partial or full, is largely defined by transitional conformational states, known as preactivated or "flip" states, that precede the final conformational change that opens the channel gate (37). Residues at the earlier stage of the activation pathway (11) appear to have no influence on the maximal efficacy of GABA, with ␤3 D120N and ␤3 T157M single mutations having an efficacy similar to that of WT. At these mutations, located near the ligand-binding domain and within an extracellular structural ␤-sheet, respectively, GABA remains acting as essentially a "full" agonist, activating the receptor with a very high maximal open probability.
However, the ␤3 Y302C mutations in the M2-M3 loop change the intrinsic property of receptor activation such that the max-

Concatenated GABA A receptor epilepsy mutations
imal efficacy of GABA is significantly reduced. Essentially, GABA has become a "partial" agonist at these receptors, where the mutation in the coupling region may be destabilizing transitional conformational states. Further, the efficacy of GABA was different, depending on the subunit location of the ␤3 Y302C mutation, suggesting that the two mutations are not entirely equivalent.

Asymmetrical effects of the same mutation at different subunit locations
A notable feature of these mutations is the differential effect of the ␤3 S254F mutation when located at different subunit locations and slightly different levels of reductions in the efficacy of GABA at receptors with a ␤3 Y302C mutation in different locations. This is in contrast with the ␤3 D120N and ␤3 T157M mutations, where the location of the individual mutations had no significant effect. It is known that the pseudosymmetry of the pentameric receptor can cause positional effects of mutations (38), and there are two possible reasons for these differences. First, the local environment surrounding mutated residues may be identical at the two subunit locations for some mutations but not others. Second, the conformational changes during the activation process may not be symmetrical from ligand binding to channel opening. Furthermore, in epilepsy-causing mutations that impair surface expression, the location of the mutation can also determine the severity of the effect (39).
Recent advances in cryogenic EM (cryo-EM) have enabled the solving of many membrane-bound proteins to a very fine resolution. During the preparation of this manuscript, a cryo-EM structure of the ␣1␤3␥2 receptor with and without GABA and diazepam bound and a structure of the ␣1␤1␥2 receptor were published (40,41). The sequences between the ␤2 and ␤3 subunits are highly homologous and identical within the regions containing the mutations studied. Hence, we can utilize the cryo-EM structure to explain the positional effects, or lack thereof, of our mutations.
Within the GABA and diazepam-bound ␣1␤3␥2 structure, the ␤3 120 residue was located within the general vicinity of the ligand-binding site at the interface of the ␤3 and ␣1 subunits (Fig. 7A) (11,41). The cryo-EM structure indicated that the local environment surrounding the ␤3 120 residues is identical, making contacts with the adjacent ␣1 M141 and ␣1 P142 residues regardless of the subunit the mutation is in (Fig. 7A). Similarly, the ␤3 T157 residue is located within a ␤-sheet within the interior of the ␤3 subunit, where the amino acid side chains interact entirely with residues within the ␤3 subunit, and the local environment that surrounds the amino acid residues is identical (Fig. 7B). Therefore, the similarity of the functional changes caused by mutations at these residues suggests that both of the local environments are identical and that the conformational changes at these initial stages of the activation pathway are symmetrical.
The ␤3 Y302 residue is located in the M2-M3 loop, a key motif in the coupling of ligand binding to channel gating that moves considerably during the channel activation process. Although there are differences in the activation processes of single ␤3 Y302C mutant receptors, depending on the subunit that the mutation is located in, the local environment surrounding the ␤3 Y302 residue is similar at the two subunits (Fig. 7C). There are slightly different poses for the ␤3 Y302 residue in each of the different locations, but it is known that the M2-M3 region alters conformation during the gating process (42), and it may be differences in the conformational changes that cause subtle differences in the effect of the mutation when it is present at the different locations.
The cryo-EM structure of the ␣1␤1␥2 GABA A receptor identified asymmetry within the transmembrane regions of the receptor (40). Whereas the distance between subunits was the same, the angles in the pentamer substantially differed along with the tilt of the transmembrane helices when compared with the ␤3 subunit. This asymmetry of key secondary structures may contribute to the ␤3 Y302C mutation having different effects when introduced in different regions within the transmembrane regions of GABA A receptors, where the transitional or pre-activated states are subtly different at the coupling regions, depending on where the first GABA molecule is bound.
The ␤3 S254F mutation caused markedly different effects, depending on the subunit where the mutation was introduced. The ␤3 S254 residue is located deep within the transmembrane region of the M1 helix, where it can interact with other transmembrane helices, including the M2 helix of the adjacent subunit.
The M1 helix of one ␤3 subunit is adjacent to the M3 of an ␣1 subunit, whereas the M1 helix of the second ␤3 subunit is adjacent to the M3 of the ␥2 subunit. This difference may underlie the different functional changes when the mutation is introduced at different locations. The side chains of the ␤3 S254 residues themselves make intrasubunit interactions within the ␤3 subunit, and the introduction of the phenylalanine residue is unlikely to make different side chain interactions when introduced at the two locations (Fig. 8). However, the bulky side chain is likely to cause structural rearrangements when occupying a larger volume, where the backbone of the M1 helix backs onto the M3 helix of either the ␥2 or ␣1 subunit. At a critical part of the M3 helix, the two subunits have different amino acid sequences, and the M3 helices are in markedly different conformations in different cryo-EM structures.
The M3 helix of the ␣1 subunit is parallel to the M1 helix of the ␤3 subunit in the apo and GABA and diazepam-bound ␣1␤3␥2 structures (Fig. 8, A and B). The M3 helix of the ␥2 subunit is similarly parallel in the GABA and diazepam-bound ␣1␤3␥2 structure, but in the apo-structure, the M3 helix of the ␥2 subunit is tilted, angling away from the M1 helix of the ␤3 subunit at the extracellular end. Further, the ␥2 F343 residue has an altered side-chain conformation. A similar change in the M3 conformation of the M3 helix is also seen in the ␣1␤3␥2 cryo-EM structure (40) (Fig. 8C). A plausible reason for the different activation properties of the two single ␤3 S254F mutant receptors is that when the residue is mutated adjacent to the ␥2 subunit, the subsequent rearrangements introduce twisting or tilting of the ␣-helices to favor a closed conformation, but when adjacent to an ␣1 subunit, the interactions stabilize an intermediate or open state.

Concatenated GABA A receptor epilepsy mutations Molecular phenotype of the mutations
There is a wealth of information that suggests that impairment of GABA A receptor-mediated inhibition can lead to seizures. This includes pharmacological evidence, where antagonists of the GABA A receptor, such as bicuculline, induce seizures (43), and genetic evidence, where mutations in GABA A receptor subunits are known to reduce receptor translocation to the cell surface (44) or impair the activation properties of the receptor (8,11,13,15). In this study, we focused on five distinct mutations, four of which clearly impaired the activation by GABA when only a single copy of the mutation was present. Our approach using the concatenated receptor enabled us to assess the complexity of receptors that contain mutations in one or both ␤3 subunits.
For three of the mutations, the resulting receptors followed a predictable pattern. A single copy of the mutation, introduced at either ␤3 subunit, caused a substantial impairment of the activation of the receptor by GABA. When a second mutation was introduced to the receptor, there was a catastrophic change to the receptor such that the magnitude of currents was either too low to be measured or markedly reduced. Regardless of whether the expression of these receptors is affected by the mutation, we would expect synaptic transmission of GABAelicited currents to be impaired in the inhibitory pathways within these patients, with receptors containing single mutations mediating reduced neurotransmission and receptors with two copies of the mutation mediating little, if any, chloride current. Figure 7. A-C, enlarged view of the ␣1␤3␥2 cryo-EM structure (PDB code 6HUP) (41) showing the ␤3 mutant residues ␤3 D120 (A), ␤3 T157 (B), and ␤3 Y302 (C) in the two different subunit locations of the second ␤3 subunit (left) and fourth ␤3 subunit (right). The subunits are colored with the first ␥2 subunit in green, the second ␤3 subunit in maroon, the third ␣1 subunit in light blue, the fourth ␤3 subunit in red, and the fifth ␣1 subunit in dark blue and the GABA molecule in blue, red, and green. Residues from the adjacent ␣1 or ␥2 subunits are indicated. At the ␤1 D120 (A), ␤1 T1577 (B), and ␤1 Y302 (C) residues, the interacting partners are identical residues either on the adjacent subunit or within the ␤3 subunit itself.

Concatenated GABA A receptor epilepsy mutations
The effects of the ␤3 S254F mutation were very different, having markedly different shifts in the potency of GABA, depending on the location of the mutation. Notably, mutations in the proline residue that precedes the same serine in the ␤1 subunit also cause the potency of GABA to increase, but the same mutation in an ␣1 subunit failed to assemble in HEK293 cells (45). Although we found robust activation of the receptors in our oocyte expression system, it is possible that the mutations also impair assembly or trafficking to the cell surface in the mammalian cell. We also cannot rule out that the mutant subunit preferentially arranges itself in the location where the potency of GABA is reduced, leading to impaired GABA activation. Importantly, we have only considered the effect of the mutation on the synaptic ␣1␤3␥2, and the effect of the mutation in other GABA receptor subtypes, including ␣5␤3␥2 or ␣4␤3␦ receptors, may contribute to the overall phenotype, including seizures. Indeed, all ␤3 mutations would be expressed in these subtypes, and as such, the effects of the ␤3 D120N , ␤3 T157M , and ␤3 Y302C on extrasynaptic receptors also need to be considered, particularly when the ␤3 T157M mutation has been reported to have little effect when expressed in the ␣5␤3␥2 subtype (8). It is also possible that the ␤3 S254F mutation is not truly pathogenic in itself; however, this mutation has been subsequently identified de novo in another patient and is therefore very unlikely not to be pathogenic (46).  (40,41) showing the ␤3 S254 or ␤1 S254 mutant residues in the two different subunit locations of the second ␤3 subunit (left) and fourth ␤3 subunit (right). The subunits are colored with the first ␥2 subunit in green, the second ␤3 subunit in maroon, the third ␣1 subunit in light blue, the fourth ␤3 subunit in red, and the fifth ␣1 subunit in dark blue. Residues from adjacent ␣1 or ␥2 subunits are indicated. Although the interacting partners of the Ser-254 residue are within the ␤ subunit, the increased volume of the phenylalanine residue that substitutes for the serine at position 254 will cause the helix to occupy the space closer to the M3 helix of the adjacent subunit. When in the second position in the apo or ␣1␤1␥2 structures, the M3 helix of the ␥2 helix is kinked rather than parallel to the M1 helix of the ␤ subunit. The residues of the M1 and M3 that face each other are the ␤1 M253 , ␤1 L256 , and ␤1 I259 residues for both subunits; the ␥2 V341 , ␥2 I344 , ␥2 F345 , and ␥2 S348 residues of the first subunit (left); and the ␣1 Y321 , ␣1 F323 , and ␣1 Y325 residues of the third subunit (right). At the locations of all of these mutations, the sequence between ␤1 and ␤3 is identical.

Conclusions and future directions
Genetic epileptic encephalopathies are a devastating group of severe childhood epilepsies that are often resistant to pharmacological treatment and include patients with mutations in genes that encode for the GABA A receptor. Introducing mutations to concatenated receptors demonstrates that the number of mutations within the receptor matter. Typically, the incorporation of one mutation impairs the activation properties of the receptor, reducing the GABA potency, efficacy, or both, and the incorporation of two mutations is often catastrophic to receptor function. However, the mutations are complex, where individual mutations can also increase the potency of GABA, and the subunit location of the mutation can also determine the functional change in the mutation. The resultant molecular phenotype is likely a complex mixture of receptors with a mix of WT receptors, receptors containing a single mutation, and receptors containing two mutations. Furthermore, receptors containing one mutation that have an intermediate effect on the activation process may be a useful target for GABAergic drugs to confer the most benefit for their specific mutation.

Molecular biology
Human cDNAs for monomeric ␣1, ␤3, and ␥2 GABA A R subunits were kind gifts from Saniona A/S (Copenhagen, Denmark). The ␥2, ␤3, and ␣1 subunits were initially subcloned into five separate in-house vectors. Linker sequences were then added through standard PCRs, where the antisense oligonucleotides caused deletion of the stop codon and in-frame fusion to the AGS linker sequence, and the sense ␤3 or ␣1 oligonucleotides caused omission of the respective ␤3 or ␣1 signal peptide and in-frame fusion to the AGS linker sequence. The remaining sequences of the sense and antisense oligonucleotides were designed to match the respective WT sequences and included unique restriction sites within each linker region and at the beginning and end of each gene sequence. Standard PCRs with the ␥2, ␤3, or ␣1 sequences as template were performed using Q5 polymerase (Genesearch, Gold Coast, Australia), and PCR products were cloned into in-house vectors using restriction enzyme digestion and ligation. Correct introduction of linker sequences and fidelity of all coding sequences were verified by double-stranded sequencing. The ␥2-␤3-␣1-␤3-␣1 concatenated construct was then created by a restriction enzyme digest of the five vectors and ligation of the five subunits with linker sequences and subcloned into an in-house vector. The resulting construct contained the subunits with linker sequences in the order of ␥2-(AGS) 5 -␤3-(AGS) 5 LGS(AGS) 3 -␣1-AGT(AGS) 5 -␤3-(AGS) 4 ATG(AGS) 4 -␣1. The vector was transformed into Escherichia coli 10-␤ bacteria for plasmid amplification, and purifications were performed with standard kits (Qiagen, Chadstone, Australia). Mutations were made using the QuikChange II site-directed mutagenesis kit (Agilent Technologies, Mulgrave, Australia) in vectors containing single subunits and flanking linker sequences and then confirmed through dsDNA sequencing. A restriction digest was then performed on subunit DNA containing the mutation and the concatenated construct to remove the appropriate WT subunit and then ligated to introduce the mutant subunit. DNA gel electrophoresis was performed to ensure incorporation of the five subunits. cRNA was produced from linearized cDNA using the mMessage mMachine T7 Transcription kit (Thermo Fisher, Scoresby, Australia) according to the manufacturer's description and stored at Ϫ80°C until use.

Xenopus surgery and oocyte preparation
All procedures using Xenopus laevis frogs were approved by the animal ethics committee of the University of Sydney (AEC number 2013/5269) and are in accordance with the National Health and Medical Research Council (NHMRC) of Australia code for the care and use of animals. In brief, a section of ovarian lobe from X. laevis was surgically removed while the frog was under anesthesia induced by tricaine, cut into smaller portions, and digested with 35 mg of collagenase-A diluted in 15 ml of OR2 (82.5 mM NaCl, 5 mM HEPES, 2 mM MgCl 2 , and 2 mM KCl, pH 7.4) at 18°C for ϳ1 h until the oocytes were fully detached from the follicles and the ovary tissue. Oocytes were then injected with a total of 2 ng of cRNA per cell that encoded concatenated WT or mutant receptors and were incubated for 2-4 days on an oscillator at 18°C in ND96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES, pH 7.4) supplemented with 2.5 mM pyruvate, 0.5 mM theophylline, and 50 g/ml gentamycin.

Two-electrode voltage-clamp recording
Cell currents were recorded using the two-electrode voltageclamp method as described previously (47). Briefly, oocytes were continuously superfused at room temperature with ND96 at ϳ5 ml/min. Cells were impaled with microelectrodes fashioned from capillary glass (Harvard Apparatus, Holliston, MA) that were prepared with a micropipette puller (Narishige, Tokyo, Japan) and filled with 3 M KCl (0.3-2.0 megaohms) and then voltage-clamped at Ϫ60 mV. A semi-automated threechannel oocyte recording system was used, where the application of solution was controlled through programming of a Powerlab 8/36 data acquisition system (ADI Instruments, Sydney, Australia) that switched solutions through a VC-8 eight-channel perfusion system (Warner Instruments LLC, Hamden, CT) and then applied solution to three recording chambers. Currents were recording using a GeneClamp 500B (Axon Instruments, Foster City, CA) or OC-725C amplifier clamp (Warner Instruments) and digitized with a Powerlab 8/36 and LabChart version 8.03 (ADInstruments, Sydney, Australia).
For clobazam concentration-response curves, a 10 M concentration was applied as a reference, and the responses were normalized to the mean current of the second two GABA concentrations. For all other experiments, a 3 mM concentration of GABA was applied as a reference three times during the experiment, and for concentration-response curves, peak currents were normalized to the mean current of the second two GABA applications. When estimating the maximal P O , after three consecutive applications of the reference 3 mM GABA solution, the solution containing 10 mM GABA, 1 M diazepam, and 3 M etomidate was applied, and peak currents were normalized to the mean current of the second two GABA applications. A washout period of 10 -12 min was performed between GABA Concatenated GABA A receptor epilepsy mutations applications to prevent effects from desensitization. All experiments were performed over a minimum of two different batches of oocytes, and a minimum of 10 individual experiments were performed. The GABA applications were applied in a sequence identical to those shown in the representative data figures. Data were acquired at 1 kHz, and, for the purposes of displaying representative traces, the data were converted to 10 Hz offline through Microsoft excel.

Data analysis and statistics
Concentration-response curves were fitted using GraphPad Prism version 7 to a monophasic Hill equation of the form, where I max is the maximum current, EC 50 is the concentration that produces the half-maximum response, [A] is the concentration of ligand, and n H is the Hill slope. Individual oocytes where a complete concentration-response curve was taken are recorded as a single n. Responses were normalized to the fitted maximum response of individual concentration-response curves. The EC 50 shown is from the fitting of Hill equations to all data, whereas the log EC 50 , I max , and n H values are the mean and S.E. derived from fitting curves to individual experiments.
For clobazam concentration-response curves, the percent modulation of clobazam was derived by the equation. These data were then fitted to the Hill equation, as above, to determine the parameters of the concentration-response curves.
The estimated P o(max) for individual experiments was derived by the equation.
Est. P 0(max) ϭ correction factor ϫ I 3 mM GABA I 10 mM GABA, 1 M diazepam, 3 M etomidate (Eq. 3) The correction factor was determined for each mutation to correct for the fact that 3 mM GABA did not always elicit the maximum response to GABA. This was derived from rearranging the Hill equation, where the EC 50 was in M. When comparing the WT and mutation concentrationresponse curves, all data were transformed to the Est. P o(max).
For statistical analysis, Est. P o(max) values and parameters derived from concentration-response curves were compared with a one-way ANOVA with Tukey's post hoc test. Significance values of p Ͻ 0.05, p Ͻ 0.01, and p Ͻ 0.001 are shown under "Results."