GABRB3 Mutation, G32R, Associated with Childhood Absence Epilepsy Alters α1β3γ2L γ-Aminobutyric Acid Type A (GABAA) Receptor Expression and Channel Gating

Background: A GABRB3 mutation has been associated with childhood absence epilepsy and β3 subunit hyperglycosylation. Results: The mutation altered subunit expression and reduced GABAA receptor function independent of N-glycosylation. Conclusion: The mutation introduced a charged residue predicted to alter subunit interactions. Significance: The distal N terminus of GABAA receptor subunits may play an unexpected role in receptor assembly and channel gating. A GABAA receptor β3 subunit mutation, G32R, has been associated with childhood absence epilepsy. We evaluated the possibility that this mutation, which is located adjacent to the most N-terminal of three β3 subunit N-glycosylation sites, might reduce GABAergic inhibition by increasing glycosylation of β3 subunits. The mutation had three major effects on GABAA receptors. First, coexpression of β3(G32R) subunits with α1 or α3 and γ2L subunits in HEK293T cells reduced surface expression of γ2L subunits and increased surface expression of β3 subunits, suggesting a partial shift from ternary αβ3γ2L receptors to binary αβ3 and homomeric β3 receptors. Second, β3(G32R) subunits were more likely than β3 subunits to be N-glycosylated at Asn-33, but increases in glycosylation were not responsible for changes in subunit surface expression. Rather, both phenomena could be attributed to the presence of a basic residue at position 32. Finally, α1β3(G32R)γ2L receptors had significantly reduced macroscopic current density. This reduction could not be explained fully by changes in subunit expression levels (because γ2L levels decreased only slightly) or glycosylation (because reduction persisted in the absence of glycosylation at Asn-33). Single channel recording revealed that α1β3(G32R)γ2L receptors had impaired gating with shorter mean open time. Homology modeling indicated that the mutation altered salt bridges at subunit interfaces, including regions important for subunit oligomerization. Our results suggest both a mechanism for mutation-induced hyperexcitability and a novel role for the β3 subunit N-terminal α-helix in receptor assembly and gating.

A GABA A receptor ␤3 subunit mutation, G32R, has been associated with childhood absence epilepsy. We evaluated the possibility that this mutation, which is located adjacent to the most N-terminal of three ␤3 subunit N-glycosylation sites, might reduce GABAergic inhibition by increasing glycosylation of ␤3 subunits. The mutation had three major effects on GABA A receptors. First, coexpression of ␤3(G32R) subunits with ␣1 or ␣3 and ␥2L subunits in HEK293T cells reduced surface expression of ␥2L subunits and increased surface expression of ␤3 subunits, suggesting a partial shift from ternary ␣␤3␥2L receptors to binary ␣␤3 and homomeric ␤3 receptors. Second, ␤3(G32R) subunits were more likely than ␤3 subunits to be N-glycosylated at Asn-33, but increases in glycosylation were not responsible for changes in subunit surface expression. Rather, both phenomena could be attributed to the presence of a basic residue at position 32. Finally, ␣1␤3(G32R)␥2L receptors had significantly reduced macroscopic current density. This reduction could not be explained fully by changes in subunit expression levels (because ␥2L levels decreased only slightly) or glycosylation (because reduction persisted in the absence of glycosylation at Asn-33). Single channel recording revealed that ␣1␤3(G32R)␥2L receptors had impaired gating with shorter mean open time. Homology modeling indicated that the mutation altered salt bridges at subunit interfaces, including regions important for subunit oligomerization. Our results suggest both a mechanism for mutation-induced hyperexcitability and a novel role for the ␤3 subunit N-terminal ␣-helix in receptor assembly and gating.
Three separate CAE-associated mutations were recently identified in GABA A receptor ␤3 subunits: GABRB3(P11S), GABRB3(S15F), and GABRB3(G32R) (15). Mutant subunitcontaining receptors exhibited reduced current density. Moreover, the mutant proteins all appeared to be "hyperglycosylated," because they migrated at higher molecular masses than wild type ␤3 subunits unless digested with an enzyme that removed all N-glycans. The investigators consequently hypothesized that hyperglycosylation might be responsible for the reduced current density, which might in turn lead to neuronal hyperexcitability and, ultimately, to the abnormal EEG patterns of absence seizures.
Approximately half of all eukaryotic proteins carry N-linked glycans (16). The process of N-linked glycosylation begins in the endoplasmic reticulum lumen, where standard "core" glycans are attached to the side chain nitrogen of asparagines located in the glycosylation consensus sequon, Asn-Xaa-(Ser/ Thr) (Xaa Pro) (17,18). Sequons containing threonine resi-X-100, 250 mM NaCl, 5 mM EDTA) containing protease inhibitor mixture (Sigma). Lysates were cleared by centrifugation at 16,000 ϫ g for 15 min and subsequently incubated overnight with high capacity NeutrAvidin-agarose resin (Pierce). After overnight incubation, protein was eluted and subjected to immunoblotting.
Immunoblotting-Proteins in sample buffer were separated on 4 -12% BisTris NuPAGE gels (Invitrogen) and transferred to Odyssey PVDF membranes (Li-Cor). A monoclonal antibody raised against intracellular residues 370 -433 of the GABA A receptor ␤3 subunit (4 g/ml, clone N87/25, UC Davis/NIH NeuroMab Facility) was used to detect ␤3 subunit protein, and anti-Na ϩ /K ϩ -ATPase antibody (0.2 g/ml, clone 464.6, ab7671, Abcam) was used as a loading control. Anti-mouse IRdye conjugated secondary antibodies (Li-Cor) were used in all cases. Membranes were scanned using the Li-Cor Odyssey system, and integrated intensities of bands were determined using Odyssey software.
Glycosidase Digestion-Biotinylated protein was simultaneously eluted from NeutrAvidin resin and denatured by incubation in 1ϫ glycoprotein denaturing buffer (New England Biolabs) containing 50 mM dithiothreitol for 30 min at 50°C. Eluates were divided into 15-l aliquots and digested with 1 unit of endo-␤-N-acetylglucosaminidase H (endo H) or peptide-N-glycosidase F in manufacturer-supplied buffers (New England Biolabs) at 37°C for 2 h.
For total cellular protein detection, cells were permeabilized for 15 min with Cytofix/Cytoperm (BD Biosciences) and washed twice with 1ϫ PermWash (BD Biosciences) before antibody incubation. For these experiments, all antibodies were diluted to 2.5 g/ml in PermWash. After antibody incubation, cells were washed four times in PermWash and twice in FACS buffer before fixation with 2% w/v paraformaldehyde, 1 mM EDTA diluted in PBS.
Fluorescence intensity (FI) of all samples was determined using an LSR II 5-laser flow cytometer (BD Biosciences) and analyzed off line with FlowJo 7.5 (Tree Star). For each sample, 50,000 total events were acquired; nonviable cell populations, determined in control experiments by staining with 7-aminoactinomycin D, were excluded from analysis. For all experiments, the net FI of samples was determined by subtracting the mean FI of cells transfected with blank pcDNA(3.1ϩ) vector from the mean FI of cells expressing GABA A receptor subunits.
The relative fluorescence intensity ("relative FI") for each condition was calculated by normalizing the net FI of each experimental condition to the net FI of cells expressing wild type ␤3 subunits.
Whole Cell Electrophysiology-Whole cell voltage clamp recordings were performed at room temperature on lifted HEK293T cells 24 -72 h after transfection with GABA A receptor subunits as described previously (7). Briefly, cells were bathed in an external solution containing 142 mM NaCl, 1 mM CaCl 2 , 8 mM KCl, 6 mM MgCl 2 , 10 mM glucose, and 10 mM HEPES, pH 7.4 (ϳ325 mosM), and recording electrodes were filled with an internal solution containing 153 mM KCl, 1 mM FIGURE 1. G32R mutation was predicted to be adjacent to the first of three putative glycosylation sites in ␤3 subunits and to lie at subunit interfaces in assembled GABA A receptors. A, sequences of human ␣1, ␤3, and ␥2L GABA A receptor subunits were aligned with the sequence of the human nicotinic acetylcholine receptor ␣1 subunit (AChR␣1). In the AChR␣1 sequence, ␣-helices are highlighted in orange; ␤-sheets are highlighted in blue, and the first transmembrane domain is underlined. In the GABA A receptor ␤3 subunit sequence, putative N-glycosylation sites are highlighted in green. In all sequences, signal peptides are highlighted in gray, and the cysteines forming the Cys-loop are highlighted in yellow. Sites of epilepsy-associated mutations in GABA A receptor subunits (␤3(G32R) and ␥2(R82Q)) are highlighted in red. B, model of the ␣1␤3␥2L GABA A receptor, as viewed from the synaptic cleft, is presented. The nicotinic acetylcholine receptor ␣1 subunit crystal structure (2qc1) was used to generate homology models of individual GABA A receptor subunits, which were threaded onto the L. stagnalis acetylcholine-binding protein crystal structure in the order ␥2L-␤3-␣1-␤3-␣1. The ␣1, ␤3, and ␥2L subunits are colored red, blue, and green, respectively. Glycine 32 and asparagine 33 are presented as orange and cyan space-filling models, respectively. MgCl 2 , 10 mM HEPES, 5 mM EGTA, 2 mM Mg 2ϩ -ATP, pH 7.3 (ϳ300 mosM). All patch electrodes had a resistance of 1-2 megaohms. Cells were voltage-clamped at Ϫ20 mV using an Axopatch 200B amplifier (Axon Instruments, Union City, CA). A rapid exchange system (open tip exchange times ϳ400 s), composed of a four-barrel square pipette attached to a perfusion fast step (Warner Instruments Corp., Hamden, CT) and controlled by Clampex 9.0 (Axon Instruments), was used to apply GABA to lifted whole cells. All currents were low pass filtered at 2 kHz, digitized at 5-10 kHz, and analyzed using the pCLAMP 9 software suite.
Single Channel Electrophysiology-Cell-attached single channel recording was performed as described previously (7). Briefly, HEK293T cells expressing GABA A receptor subunits were bathed in an external solution containing 140 mM NaCl, 2 mM CaCl 2 , 5 mM KCl, 1 mM MgCl 2 , 10 mM glucose, and 10 mM HEPES, pH 7.4. Recording electrodes were filled with an internal solution containing 1 mM GABA, 120 mM NaCl, 0.1 mM CaCl 2 , 5 mM KCl, 10 mM MgCl 2 , 10 mM glucose, and 10 mM HEPES, pH 7.4, and electrode potential was held at ϩ80 mV. The electrodes were polished to a resistance of 10 -20 megohms.
Single channel currents were amplified and low pass filtered at 2 kHz using an Axopatch 200B amplifier, digitized at 20 kHz using Digidata 1322A, and saved using pCLAMP 9 software (Axon Instruments). Data were analyzed using TAC 4.2, and open time and amplitude histograms were generated using TACFit (Bruxton Corp., Seattle, WA) as described previously (10). The number of components required to fit the duration histograms was increased until an additional component did not significantly improve the fit (26). Single channel openings occurred as bursts of one or more openings or clusters of bursts. Bursts were defined as one or more consecutive openings that were separated by closed times that were shorter than a specified critical duration (t crit ) prior to and following the openings (27). A t crit of 5 ms was used in this study. Clusters were defined as a series of bursts preceded and followed by closed intervals longer than a specific critical duration (t cluster ). A t cluster of 10 ms was used in this study.
Homology Modeling-Multiple sequence alignments of human GABA A receptor ␣1, ␤3, and ␥2L subunits and the human nicotinic acetylcholine receptor (nAChR) ␣1 subunit were performed using ClustalW (European Bioinformatics Institute, Hinxton, UK). Structural models of GABA A receptor N-terminal domains were generated with SWISS-MODEL (28), using the crystal structure of the nAChR ␣1 subunit (Protein Data Bank code 2qc1) (29) as a template. Point mutations were introduced into the ␤3 subunit sequence using Deep-View/Swiss-PdbViewer 4.02 (Swiss Institute of Bioinformatics, Lausanne, Switzerland), and SWISS-MODEL project files containing the mutated target sequence and the superposed template structure were submitted. For heteropentamers, subunits were threaded in the order ␥2L-␤3-␣1-␤3-␣1 onto the Lymnaea stagnalis acetylcholine-binding protein crystal structure (Protein Data Bank code 1i9b) (30) used as a template. All models were energy-optimized using GROMOS96 in default settings within DeepView/Swiss-PdbViewer, and the most likely conformations were presented here.
Statistical Analysis-Statistical analysis was performed using Prism version 5.04 (GraphPad Software, La Jolla, CA). Student's two-tailed t test or one-way analysis of variance (ANOVA) with Tukey's and/or Bonferroni's post-tests was used as appropriate to determine statistical significance among transfection conditions. Levels of significance were indicated in figure legends, and all data were expressed as mean Ϯ S.E.

RESULTS
Cotransfection of Mutant ␤3(G32R) Subunit with ␣1 or ␣3 and ␥2L HA Subunits Was Associated with Increased ␤3 Subunit and Decreased ␥2L HA Subunit Surface Expression-Because the ␤3(G32R) mutation was reported to reduce the current density of heterologously expressed ␣1␤3␥2L receptors (15), we sought to determine whether the mutation reduced surface expression of the GABA A receptor subunits under similar conditions. First, we transiently coexpressed ␣1, ␥2L HA , and either wild type ␤3 or mutant ␤3(G32R) subunit cDNAs at a 1:1:1 ratio in HEK293T cells and assessed surface expression of all subunits using surface biotinylation and Western blotting.
Because it would be highly unusual for a mutation to cause both an increase in surface receptor number and a reduction in current density, we first addressed the differences in wild type and mutant subunit expression levels. Increases in ␤3 subunit surface expression could reflect either a change in receptor subunit composition, including production of ␣1␤3 receptors and/or ␤3 subunit homomers, or an overall increase in surface ␣1␤3␥2L receptor number. To distinguish between these two possibilities, we first examined the differences in wild type and mutant partnering subunit expression levels. Immunoblotting for surface levels of partnering subunits suggested that ␣1 subunit levels did not change but ␥2L HA subunit levels decreased when coexpressed with ␤3(G32R) rather than ␤3 subunits (data not shown), indicating a potential change in receptor subunit composition. Because the reduction of ␥2L HA surface levels was subtle, we employed flow cytometry to confirm and quantify changes in subunit expression levels. In these experiments, we also included a condition modeling heterozygous expression of ␤3(G32R) subunits, in which an equimolar mixture of both ␤3 and ␤3(G32R) subunit cDNA was cotransfected with ␣1 and ␥2L HA subunit cDNAs (see under "Experimental Procedures" for exact subunit cDNA ratios and concentrations). Consistent with the prior immunoblotting data, ␣1 subunit surface levels did not differ significantly in the heterozygous or homozygous mutant conditions (Fig. 2B), but ␥2L HA subunit surface levels were decreased slightly in both heterozygous (84.8 Ϯ 2.2% of wild type, n ϭ 10, p Ͻ 0.001) and homozygous mutant (87.1 Ϯ 2.4% of wild type, n ϭ 16, p Ͻ 0.001) conditions (Fig. 2C).
We chose to coexpress ␣1 and ␥2 subunits because they are the most widely expressed subunits of their respective families in whole brain; however, subunit expression patterns vary widely among brain regions. Several studies have indicated that absence seizures frequently involve dysfunction in the thalamic reticular nucleus, where the ␣3 subunit subtype predominates, and ␤3 subunit expression is also high (31). Consequently, we also examined changes in partnering subunit expression levels using ␣3 rather than ␣1 subunit cDNA. Results resembled those obtained using ␣1 subunits, i.e. ␣3 subunit surface levels did not differ significantly among wild type, heterozygous, and homozygous mutant receptors (Fig. 2D), but ␥2L HA subunit surface levels decreased in both heterozygous and homozygous mutant conditions, although the reduction was significant only in the homozygous mutant condition (Fig. 2E).
Changes in subunit surface expression could reflect alterations in subunit production, subunit stability, or receptor trafficking. Therefore, we also assessed total cellular subunit levels ( ␤3 Subunit Mutation, G32R, Affected Subunit Surface Expression Independent of Glycosylation-To this point, we observed two principal effects of the ␤3 subunit mutation, G32R. First, mutant and wild type receptors had different surface expression patterns (specifically, a small decrease in ␥2L subunit levels and a large increase in ␤3 subunit levels); these changes suggested a partial replacement of ␣1␤3␥2L receptors by ␣1␤3 receptors and ␤3 subunit homopentamers. Second, ␤3(G32R) subunits were more likely than ␤3 subunits to migrate at the highest of three distinct molecular mass populations. However, it remained unclear if there was a causal relationship between these two phenomena. FIGURE 2. Cells expressing ␣1 or ␣3, ␤3(G32R), and ␥2L HA subunits had higher surface levels of ␤3 subunits and lower surface levels of ␥2L HA subunits compared with cells expressing ␣1 or ␣3, ␤3(WT), and ␥2L HA subunits. A1, surface protein was isolated from HEK293T cells transfected with equimolar amounts of ␣1, ␥2L HA , and either wild type (wt) or G32R mutant (hom) ␤3 GABA A receptor subunit cDNA, separated by SDS-PAGE, and evaluated using Western blots. The upper panel presents staining for Na ϩ /K ϩ -ATPase as a loading control (LC), and the lower panel presents staining for ␤3 subunits (see under "Experimental Procedures" for antibody descriptions). A2, surface protein levels of ␤3 subunits were quantified in HEK293T cells transfected with equimolar amounts of ␣1, ␥2L HA , and either wild type (wt) or G32R mutant (hom) ␤3 GABA A receptor subunits. Integrated band intensities of all ␤3 subunits were determined, summed, and normalized to integrated band intensities of Na ϩ /K ϩ -ATPase for the same sample. The normalized intensities of ␤3 subunits were then expressed as proportions of wild type ␤3 subunit intensities. Statistical significance was determined using Student's two-tailed paired t test. B and C, flow cytometry was used to evaluate surface protein levels of ␣1 (B) and ␥2L HA (C) subunits in HEK293T cells transfected with ␣1, ␥2L HA , and wild type and/or G32R mutant ␤3 GABA A receptor subunit cDNA. Wild type and homozygous mutant (hom) expression were modeled by transfecting 1 g each of ␣1, ␥2L HA , and either wild type or G32R mutant (hom) ␤3 subunit cDNA. Heterozygous mutant expression (het) was modeled by transfecting 1 g each of ␣1 and ␥2L HA cDNA together with 0.5 g each of ␤3 and ␤3(G32R) subunit cDNA. Upper panels present fluorescence intensity histograms; the abscissa indicates FI in arbitrary units plotted on a logarithmic scale, and the ordinate indicates percentage of maximum cell count (% of max). Histograms for cells transfected with blank vector (solid gray line), or WT (solid black line), heterozygous (dotted black line), and homozygous (dashed black line) subunit combinations are overlaid. Lower panels present normalized fluorescence intensities for each expression condition. Mean fluorescence intensities from cells transfected with blank vector alone (mock) were subtracted from mean fluorescence intensities of WT, heterozygous, and homozygous expression conditions. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity of the WT expression condition. D-E, flow cytometry was used to evaluate surface protein levels of ␣3 (D) and ␥2L HA (E) subunits in HEK293T cells transfected with ␣3, ␥2L HA , and wild type and/or G32R mutant ␤3 subunit cDNA. All panels are presented as described in B and C, but in all cases ␣3 subunit cDNA was substituted for ␣1 subunit cDNA. Statistical significance was determined using one-way ANOVA with Tukey's post-test. *** indicates p Ͻ 0.001 compared with WT.
Because the mutant subunit had been reported to be hyperglycosylated, we hypothesized that the multiple ␤3 subunit bands represented differently glycosylated protein populations, where individual sequons may or may not be occupied by a glycan, occupancy patterns may or may not be uniform within a protein population (e.g. among all ␤3(G32R) subunits), and the glycans themselves may contain different combinations of monosaccharides. To determine whether the multiple ␤3 subunit bands represented differently glycosylated protein populations and to characterize ␤3 subunit N-glycans, we isolated surface protein from HEK293T cells expressing ␣1, ␥2L HA , and either ␤3 or ␤3(G32R) subunits and compared the migration patterns of ␤3 and ␤3(G32R) subunits that were undigested; digested with endo H, which cleaves only high mannose, unprocessed glycans; or digested with peptide-N-glycosidase F, which removes all N-glycans regardless of modification (Fig. 4A). After digestion with peptide-N-glycosidase F, both ␤3 and ␤3(G32R) subunits migrated as one 43-kDa band, indicating that the G32R mutation did indeed increase N-glycosylation of at least one of the three ␤3 subunit sequons (Fig. 4A, 3rd and 6th lanes). Interestingly, ␤3 and ␤3(G32R) subunits also displayed different endo H digestion patterns; after endo H digestion, a substantial population of ␤3 subunits migrated at 43 kDa and thus were fully endo H-sensitive, but virtually none of the ␤3(G32R) subunits migrated at 43 kDa and thus were endo H-resistant. Therefore, the G32R mutation increased the efficiency of both addition and processing of N-glycans.
We recently established that partnering subunit incorporation could alter glycosylation patterns of ␤2 subunits (25). Thus, it was possible that the increased endo H-resistant population of ␤3(G32R) subunits reflected increased formation of ␣1␤3 and/or ␤3 receptor isoforms. To assess this possibility, we studied the digestion pattern of wild type ␤3 and mutant ␤3(G32R) subunits after transfecting ␣1 and ␤3 (Fig. 4B) or only ␤3 (Fig. 4C) subunit cDNA. We found that the increased glycosylation and glycan processing of ␤3(G32R) mutant subunits compared with ␤3 subunits persisted in the absence of ␣1 and/or ␥2L partnering subunits. Interestingly, the proportion . ␤3(G32R) mutation did not significantly affect total cellular levels of GABA A receptor subunits in cells expressing ␣1 or ␣3, ␤3, and ␥2L HA subunits. A1, total cell lysates (40 g) were obtained from HEK293T cells transfected with equimolar amounts of ␣1, ␥2L HA , and either wild type or G32R mutant (hom) ␤3 subunit cDNA, separated by SDS-PAGE, and evaluated using Western blots. The upper panel presents staining for Na ϩ /K ϩ -ATPase as a loading control (LC), and the lower panel presents staining for ␤3 subunits (see under "Experimental Procedures" for antibody descriptions). A2, total cellular levels of ␤3 subunits were quantified in HEK293T cells transfected with equimolar amounts of ␣1, ␥2L HA , and either wild type or G32R mutant (hom) ␤3 GABA A receptor subunits. Integrated band intensities of all ␤3 subunit populations were determined, summed, and normalized to integrated band intensities of Na ϩ /K ϩ -ATPase for the same sample. The normalized intensities of ␤3 subunits were then expressed as proportions of wild type ␤3 subunit intensities. Statistical significance was determined using Student's two-tailed paired t test. B and C, flow cytometry was used to evaluate total cellular levels of ␣1 (B) and ␥2L HA (C) subunits in HEK293T cells transfected with ␣1, ␥2L HA , and wild type and/or G32R mutant ␤3 subunit cDNA and permeabilized before staining with fluorescently conjugated antibodies. Wild type (wt), heterozygous (het), and homozygous (hom) expression patterns were modeled as described for Fig. 1. Upper panels present fluorescence intensity histograms; the abscissa indicates FI in arbitrary units plotted on a logarithmic scale, and the ordinate indicates percentage of maximum cell count (% of max). Histograms for cells transfected with blank vector (solid gray line), or WT (solid black line), heterozygous (dotted black line), and homozygous (dashed black line) subunit combinations are overlaid. Lower panels present normalized fluorescence intensities for each expression condition. Mean fluorescence intensities from cells transfected with blank vector alone (mock) were subtracted from mean fluorescence intensities of WT, homozygous, and homozygous expression conditions. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity of the WT expression condition. D and E, flow cytometry was used to evaluate total cellular levels of ␣3 (D) and ␥2L HA (E) subunits in HEK293T cells transfected with ␣1, ␥2L HA , and wild type and/or G32R mutant ␤3 GABA A receptor subunit cDNA and permeabilized before staining with fluorescently conjugated antibodies. All panels are presented as described in B and C, but in all cases ␣3 subunit cDNA was substituted for ␣1 subunit cDNA. Statistical significance was determined using one-way ANOVA with Tukey's post-test.
of endo H-sensitive ␤3 subunits did decrease with the number of subunits expressed, i.e. endo H sensitivity was greatest in ␣1␤3␥2L HA receptors, lower in ␣1␤3 receptors, and lowest in ␤3 receptors.
These results demonstrating altered glycosylation of ␤3(G32R) subunits contradicted in silico analysis. We used NetNGlyc 1.0 to establish that the G32R mutation did not change the occupancy potential of Asn-105 (0.7904) or Asn- Similarly, meta-analyses have concluded that sequons with Arg at the Ϫ1 position are considerably less likely to be glycosylated than sequons with Gly at the Ϫ1 position (32). Finally, although hypoglycosylation disorders frequently cause severe pathology (24), to our knowledge there are no reports of increased glycosylation adversely affecting function or trafficking of other receptors.
After confirming that ␤3 and ␤3(G32R) subunits had different glycosylation patterns, we sought to determine whether the increased glycosylation and glycan processing of ␤3(G32R) mutant subunits were indeed responsible for changes in subunit surface trafficking. To identify the occupancy of a particular sequon for both wild type and mutant receptors, we mutated each potentially glycosylated asparagine residue (N-glycosylation sites Asn-33, Asn-105, and Asn-174) individually to glutamine in wild type ␤3 and mutant ␤3(G32R) subunits, thereby creating glycosylation-defective subunits (23,25,33,34). We could not eliminate the possibility that these point mutations themselves could alter receptor assembly or function; however, we compared the characteristics of glycosylation-defective subunits bearing or lacking the G32R mutation. Furthermore, in a previous study we addressed several concerns regarding this method (25).
Consistent with previous results, ␤3 and ␤3(G32R) subunits displayed clear differences in molecular mass distribution when all three glycosylation sites remained intact. If this difference reflected increased glycosylation of ␤3(G32R) subunits at one specific site (i.e. Asn-33, Asn-105, or Asn-174), inactivating that site with an Asn to Gln mutation ("NQ mutation") should yield proteins with identical molecular mass distributions. Moreover, the 51-kDa band, which presumably represented triply glycosylated proteins, should disappear in any subunit bearing an NQ mutation. Therefore, we coexpressed ␣1 and ␥2L HA subunits with each wild type/glycosylation-deficient ␤3 subunit (N33Q, N105Q, and N174Q; labeled as 1Q, 2Q, and 3Q, respectively) and each mutant/glycosylation-deficient ␤3 subunit (G32R/N33Q, G32R/N105Q, and G32R/N174Q; labeled as GR/1Q, GR/2Q, and GR/3Q, respectively) (Fig. 4, D and E). Immunoblotting for wild type ␤3 subunit surface and total protein yielded several interesting results. Most strikingly, inactivating the second or third glycosylation site (2Q or 3Q) drastically reduced expression of wild type ␤3 subunits; indeed, expression of ␤3(2Q) subunits was nearly abolished. Although intriguing, these deficits in protein expression made it impossible to compare the glycosylation patterns of these second and third glycosylation site mutants in the presence or absence of the G32R mutation and thus to determine conclusively if the molecular mass shifts in glycosylation-competent ␤3(G32R) subunits were due to increased occupancy of the second or third glycosylation sites. Nonetheless, these data indirectly suggest that glycosylation of Asn-105 or Asn-174 was not responsible for the molecular mass shift; given that disruption of these sites so drastically reduced protein expression, it seems likely that both sites are usually glycosylated and therefore could not have their occupancy increased by the G32R mutation.
The ␤3 subunit constructs with NQ mutations allowed us to examine the effects of the G32R mutation in the absence of N-glycosylation at specific sequons. However, when all glycosylation sites were intact, the G32R mutation appeared to increase ␤3 subunit glycosylation; therefore, it was valuable to examine the effects of the mutation when both wild type ␤3 and mutant ␤3(G32R) subunits had increased glycosylation. It is not possible to force glycosylation of individual sequons, but it is well known that NXT sequons are much more likely than NXS sequons to accept N-glycans. As described above, expression of glycosylation-deficient constructs indicated that the Asn-33 site (sequon NMS) was more likely to be occupied in the presence of the G32R mutation. Therefore, we hypothesized that ␤3 subunits in which Ser-35 was mutated to Thr (␤3(S35T) subunits; ST) would exhibit a glycosylation pattern similar to that of ␤3(G32R) subunits. As shown in Fig. 4, D and E, the S35T mutation did increase glycosylation of ␤3 subunits, but the G32R mutation did not increase glycosylation further, i.e. ␤3(S35T) and ␤3(G32R/S35T) subunits exhibited similar molecular mass distributions. Taken together, these results indicated that the G32R mutation increased ␤3 subunit N-glycosylation at Asn-33.
Presence of a Basic Residue at Position 32 Reduced Surface Expression Levels of ␥2L Subunits and Increased Glycosylation at Asn-33-If the change in glycosylation was not responsible for altered subunit expression patterns seen with mutant ␤3(G32R) subunits, some other property of the point mutation itself, such as charge, must have been causative. To investigate the effects of charge at residue 32, we mutated the ␤3 subunit residue Gly-32 to lysine, glutamine, or glutamate (G32K, G32Q, and G32E, respectively). We coexpressed each of these ␤3 subunits individually with ␣1 and ␥2L subunits and evaluated glycosylation patterns of ␤3 subunits and surface levels of all subunits (Fig. 5). Interestingly, our results suggested that glycosylation of Asn-33 clearly depended upon the charge of residue 32. Thus, 32.0 Ϯ 3.9% of all wild type ␤3 subunit surface protein was fully glycosylated (i.e. migrated at 51 kDa), compared with 67.4 Ϯ 5.9% of ␤3(G32R) and 56.7 Ϯ 5.4% of ␤3(G32K) subunit proteins (Fig. 5, A1 and A2). In contrast, 42.0 Ϯ 4.9% of ␤3(G32Q) subunit surface protein and only 2.8 Ϯ 0.9% of ␤3(G32E) subunit surface protein were fully glycosylated.
As in previous experiments, ␣1 subunit surface levels changed minimally when coexpressed with any ␤3 subunit (Fig.  5B). Surprisingly, however, ␣1 subunit surface levels did decrease significantly when coexpressed with ␤3(G32K) subunits (87.4 Ϯ 3.4 of wild type; p Ͻ 0.05). Conversely, ␥2L HA subunit surface levels were correlated with charge at ␤3 subunit residue 32. Specifically, ␥2L HA subunit surface levels decreased significantly when the coexpressed ␤3 subunit had a positively charged residue (arginine or lysine) at position 32 (GK, 79.2 Ϯ 3.3% of wild type, n ϭ 12, p Ͻ 0.001), but ␥2L HA subunit surface levels decreased only slightly when the coexpressed ␤3 subunit had an uncharged residue at the same position (GQ, 93.5 Ϯ 3.4% of wild type, n ϭ 12) and did not change when the coexpressed ␤3 subunit had a negatively charged residue (GE, 102.7 Ϯ 4.0% of wild type, n ϭ 11). Taken together, these data (Figs. 4 and 5) indicated that the positive charge introduced by the G32R mutation was responsible both for increasing Asn-33 glycosylation and for decreasing ␥2L HA subunit incorporation; however, these two phenomena were not causally related to one another.
G32R Mutation Reduced Current Density Independent of Glycosylation-Up to now, we observed that the G32R mutation caused glycosylation-independent changes in subunit expression patterns that could reduce the function of ␣␤␥ GABA A receptors; however, those changes were not large enough to account for the reduction in current amplitude that was previously reported (15). This discrepancy suggested that expression of ␤3(G32R) subunits might also affect receptor gating. Furthermore, although subunit expression patterns depended upon charge at residue 32 rather than glycosylation at residue 33, any such changes in gating might still be glycosylation-dependent.
To determine how mutant ␤3(G32R) subunits affected GABA A receptor function, we used a rapid exchange system to apply 1 mM GABA for 400 ms to lifted HEK293T cells coexpressing ␣1, ␥2L, and ␤3, ␤3(G32R), ␤3(N33Q), or ␤3(G32R/ N33Q) subunits (Fig. 6A). Wild type receptors displayed a current density of 845.8 Ϯ 33.93 pA/pF (n ϭ 34), nearly 50% higher than current density of receptors containing mutant ␤3(G32R) subunits (454.3 Ϯ 57.46 pA/pF, n ϭ 32, p Ͻ 0.001 compared with wild type) (Fig. 6, A and B); this difference was consistent with previously reported data (15). When Asn-33 glycosylation was abolished by introducing the ␤3(N33Q) subunit mutation alone, current density also decreased (411.7 Ϯ 19.61 pA/pF, n ϭ 21, p Ͻ 0.001 compared with wild type). However, when the G32R mutation was introduced together with the N33Q mutation, current density decreased further (160.4 Ϯ 46.99 pA/pF, n ϭ 13, p Ͻ 0.001 compared with wild type and p Ͻ 0.01 compared with the ␤3(N33Q) subunit alone). These results suggested that although eliminating Asn-33 glycosylation by introducing the N33Q mutation itself reduced current density (because of either the absence of the glycan or the presence of the point mutation), the CAE-associated ␤3(G32R) subunit Lower panels present normalized fluorescence intensities for each condition. Mean fluorescence intensities from cells transfected with blank vector alone (mock) were subtracted from mean fluorescence intensities of cells transfected with ␣1, ␥2L HA , and the indicated ␤3 subunits. All mock-subtracted fluorescence intensities were normalized to the mock-subtracted fluorescence intensity of the cells expressing ␤3 subunits. Statistical analysis was performed using one-way ANOVA with Tukey's post-test. **, p Ͻ 0.01; ***, p Ͻ 0.001 compared with wild type. mutation also impaired receptor function independent of Asn-33 glycosylation.
In summary, both ␤3(G32R) and ␤3(N33Q) point mutations significantly reduced current densities of ␣1␤3␥2L receptors. However, the effects of these mutations were additive, indicating that the G32R point mutation reduced current density even when Asn-33 was not glycosylated. Taken together, these data suggest that the G32R mutation reduced current density by a mechanism that was independent of increasing Asn-33 glycosylation and furthermore that this region of the N-terminal ␣-helix could play a role in channel gating.
␤3(G32R) Mutation Was Predicted to Alter Salt Bridges and Conformation at ␤3-␥2 and ␤3-␤3 Subunit Interfaces-To gain insight into the mechanism by which the ␤3(G32R) mutation affected receptor assembly and channel gating, we performed homology modeling of wild type and mutant receptors using the nAChR ␣1 subunit extracellular domain structure (Protein Data Bank code 2qc1) (29) as a template (Fig. 10). In ␣1␤3␥2L receptor isoforms, the major structural changes induced by the ␤3(G32R) mutation occurred at the interface between the principal (ϩ) side of the ␥2L subunit and the complementary (Ϫ) side of the ␤3 subunit (␥2-␤3 interface). In both ␣1␤3␥2L (Fig.  9A) and ␣1␤3(G32R)␥2L (Fig. 9B) receptors, all subunits were predicted to begin with a random coil leading into an ␣-helix. However, the G32R mutation induced a conformational change in the ␤3 subunit ␣-helix, causing the random coil to project in a slightly different direction. Moreover, the side chain of Arg-32 extended across the ␥2-␤3 subunit interface, forming a salt bridge with ␥2 subunit residue Asp-123, which lies in a motif previously established to be necessary for ␥2-␤3 subunit interaction (35).
Because both ␣1␤3(G32K)␥2L and ␣1␤3(G32E)␥2L receptors also displayed reduced current densities, we performed homology modeling of these isoforms as well. These mutations also induced structural changes primarily at the ␥2-␤3 subunit interface. Interestingly, the side chain of the ␤3 subunit residue Lys-32 angled toward the cell membrane and formed a salt bridge with the ␥2 subunit residue Glu-217, which participates in a salt bridge network disrupted by the epilepsy-associated ␥2(R82Q) mutation (Fig. 9C) (36). The side chain of the ␤3 subunit residue Glu-32, conversely, extended toward the ␥2 subunit but did not come within 4 Å of any ␥2 subunit atoms (Fig. 9D).
We demonstrated that ␤3(G32R) subunits were expressed on the cell surface at much higher levels than ␤3 subunits, suggesting that mutant subunits might assemble into homopentamers. Consequently, we also created homology models of ␤3 and ␤3(G32R) homopentameric receptors. Unsurprisingly, structural changes occurred at subunit interfaces. Strong salt bridges existed at the interface of wild type ␤3 subunits (Fig.  10A), but when Arg-32 was introduced (Fig. 10B), its side chain formed three new salt bridges with Asp-94 of the adjacent ␤3(G32R) subunit. In summary, homology modeling provided a potential explanation for the changes in subunit surface expression associated with the ␤3 subunit G32R mutation. All structural changes occurred at subunit interfaces, suggesting that the point mutation could perturb subunit oligomerization.

N-terminal ␣-Helix, New Roles in Receptor Assembly and
Gating?-A large body of work exists documenting the GABA A receptor subunit domains that are responsible for receptor assembly and trafficking, GABA binding, and coupling of agonist binding to channel gating (37). However, the distal N-terminal domain, which includes a random coil followed by an ␣-helix, has not been demonstrated to be important for these processes. In fact, the helix was entirely absent in recently discovered prokaryotic nAChR homologs (38,39). Helical integrity was shown to be necessary for proper biogenesis of nAChR ␣7 subunits, but most residues could be mutated without affecting subunit expression (40). In our experiments, multiple point mutations of Gly-32 and Asn-33 residues in the distal ␣-helix of ␤3 subunits caused changes in assembly as well as gating that were not attributable fully to the glycosylation changes induced by the mutations. Our results suggest that unexpectedly the N-terminal ␣1-helix may be important for assembly and function of GABA A receptors containing ␤3 subunits.
␤3 Subunit G32R Mutation and Receptor Heterogeneity-Our data suggested that the G32R mutation promoted formation of binary ␣1␤3 receptors and ␤3 homopentamers and decreased formation of ternary ␣1␤3␥2L receptors. Wild type ␤3 subunits are known to assemble more promiscuously than most other GABA A receptor subunits in heterologous systems; ␤3 subunits reached the cell surface when expressed alone, and both ␤3 and ␥2L subunits were detected on the cell surface when coexpressed together without an ␣ subunit (14). We obtained similar results when ␤3 subunits were coexpressed with ␤, ␦, ⑀, or subunits (data not shown). In contrast, ␤2 subunits are retained intracellularly and degraded in the absence of coexpressed ␣ subunits even though ␤2 and ␤3 subunits have very similar sequences. In previous studies, four amino acid residues (Gly-171, Lys-173, Glu-179, and Arg-180) conferred the ability to form ␤3 subunit homopentamers (14). According to our model, these residues are also predicted to lie on the (Ϫ)-face of ␤3 subunits, but are much closer to the cell membrane than the Gly-32 residue. Thus, we may have uncovered a previously unknown role for the N-terminal ␣-helix in regulating ␤3 homopentamer assembly.
␤3 Subunit Glycosylation, Patterns, and Their Dependence on Receptor Subunit Composition-Although we have shown that hyperglycosylation ultimately was not responsible for the effects of the ␤3(G32R) mutation on receptor assembly and function, our studies elucidate the characteristics and importance of ␤3 subunit N-glycosylation. We demonstrated that all three N-glycosylation sites on wild type ␤3 subunits could be glycosylated in HEK293T cells, although many ␤3 subunits were not glycosylated at Asn-33. Importantly, other investigators have observed similar ␤3 subunit glycosylation patterns in mouse cortical neurons. 3 Interestingly, we also showed that ␤3 subunits retain some unprocessed high mannose glycans despite being assembled into receptors and trafficked to the cell surface. This occurred in all tested receptor isoforms (i.e. ␤3, ␣1␤3, and ␣1␤3␥2L); however, the proportion of ␤3 subunits 3 M. J. Gallagher, private communication. . ␤3(Gly-32) mutations changed conformation and salt bridge formation at the ␥2-␤3 interface of heteropentameric ␣1␤3␥2L receptors. Three-dimensional models of ␣1, ␤3, and ␥2 subunit extracellular domains were created (see under "Experimental Procedures") and threaded onto the L. stagnalis acetylcholine-binding protein structure in the order ␥2-␤3-␣1-␤3-␣1 to model ternary heteropentameric ␣1␤3␥2L receptors. Point mutations were introduced into ␤3 subunit structures and the resulting energy-minimized models were examined for structural changes. A, portion of the interface between the ␥2 subunit (yellow) and the wild type ␤3 subunit (blue) is presented. The perspective is from outside the receptor, such that the synaptic cleft would be located at the top of the figure. Residues discussed in the text, including the mutated Gly-32, the first glycosylation sequon residues Asn-33 and Ser-35, and the second glycosylation site Asn-105, are labeled, and predicted salt bridges are indicated by dotted lines. Adjacent numbers indicate the distance in angstroms between the two atoms forming the salt bridge. Two ␥2 subunit residues are also identified: Arg-125, which is predicted to form a salt bridge with ␤3 (Asn-33); and Arg-82, which was mutated to Q in a family with GEFS ϩ . Side chains are colored in the CPK scheme; carbon atoms are gray, oxygen atoms are red, and nitrogen atoms are blue. B-D, views of the ␥2-␤3 interface in ␣1␤3(G32R)␥2L (B), ␣1␤3(G32K)␥2L (C), and ␣1␤3(G32E)␥2L (D) receptors are presented as in A. Salt bridges longer than 3.5 Å are indicated by gray dotted lines, and salt bridges shorter than 3.5 Å are indicated by black dotted lines.

␤3(G32R) Impairs GABA A Receptor Expression and Gating
containing endo H-sensitive glycans was correlated with the number of different subunits expressed. We recently observed a similar phenomenon in ␤2 subunits. In heterologous systems, all ␤2 subunit bands were endo H-resistant when only ␣1 and ␤2 subunits were coexpressed, but an endo H-sensitive population appeared if ␥2 subunits were added. Furthermore, ␤2 subunits from heterozygous ␥2 subunit knock-out mice, which may form ␣1␤2 receptors due to ␥2 subunit deficiency, had a larger endo H-resistant population than ␤2 subunits from wild type mice. 4 Taken together, these findings suggest that incorporation of non-␤ subunits might alter ␤ subunits such that their glycans become accessible for modification in the Golgi apparatus. It will be interesting to determine whether glycan structure contributes to the characteristic current properties of binary and ternary receptors.
Although the G32R mutation appeared to promote increased assembly of binary ␣1␤3 and homopentameric ␤3 receptors, and those isoforms promoted glycan maturation, the G32R mutation also affected glycan processing independent of receptor stoichiometry. All wild type receptors (i.e. ␤3, ␣1␤3, and ␣1␤3␥2L) contained at least a small population of ␤3 subunits that were fully endo H-sensitive. However, that population virtually disappeared in the corresponding mutant receptor isoforms. It may be worthwhile to investigate whether microheterogeneity (i.e. sugar composition) as well as macroheterogeneity (i.e. sequon occupancy) of N-glycans can affect receptor function.

Altered Salt Bridge Formation and Receptor Conformation
May Be Responsible for Changes in Assembly, Glycosylation, and Gating, Leading to Reduced GABA A Receptor-mediated Inhibition-Because GABA A receptors have not been crystallized, homology models are limited to nAChR (29,41) and AChBP (30,42) templates, many of which have poor resolution in their N-terminal domains. Therefore, although homology models of GABA A receptors are necessarily speculative, they nonetheless provide valuable insight regarding potential interactions. In our models, mutating the Gly-32 residue to Arg-32 induced formation of new salt bridges at the ␥2-␤3 and ␤3-␤3 interfaces in ternary and homopentameric receptors, respectively. The ␤3-␤3 salt bridges were particularly strong and would likely increase the affinity of homodimer formation. This, in turn, could promote formation of isoforms containing a ␤3-␤3 interface. Such interfaces are not predicted to exist in ternary receptors, which are thought to have a ␥-␤-␣-␤-␣ orientation (anticlockwise as viewed from the synaptic cleft) (12,13). However, in binary receptors, the ␥2 subunit presumably is replaced by either an ␣1 or a ␤3 subunit; the latter would introduce a ␤3-␤3 interface. It is possible that the salt bridges introduced by the G32R mutation promote ␤3(G32R) subunit homodimerization, which in turn could "seed" the formation of (␣1) 2 (␤3) 3 and ␤3 receptor isoforms, thereby increasing ␤3 subunit surface expression. It is somewhat less clear how salt bridge formation between Arg-32 and ␥2(Asp-123) could discourage incorporation of ␥2 subunits; however, it is important to note that salt bridges can be destabilizing (43) and that ␥2 subunit (122-129) integrity was essential for ␥2-␤3 subunit interaction (35). It is also worth mentioning that the epilepsy- associated mutation ␥2(R82Q), which has been shown to disrupt receptor assembly (36,44), is located in the ␥2 subunit loop closest to the N-terminal domain of the ␤3 subunit ␣-helix. Indeed, point mutations throughout this loop impaired ␥2 subunit incorporation. Thus, it is possible that any structural changes in this area, whether on ␥2 or ␤3 subunits, could disturb an important assembly domain and result in preferential expression of low efficacy binary ␣␤3 receptors and homopentameric ␤3 receptors instead of high efficacy ternary ␣␤3␥2 receptors, thereby causing disinhibition.
How Might the ␤3(G32R) Mutation Contribute to Epileptogenesis?-The electroencephalographic signature of an absence seizure involves generalized, synchronous spikewave activity, reflecting oscillations in thalamocortical circuits. The location of the seizure discharge origin within these circuits remains a subject of debate (45), making it difficult to predict how changes in the function of a particular ion channel could initiate seizures. However, it is known that both thalamic reticular nucleus and cortex (particularly somatosensory cortex) participate in synchronized activity. Importantly, the ␤3 subunit subtype predominates in the reticular nucleus throughout life and in the cortex during development (31,46,47).
It was recently demonstrated that tonic GABAergic current is paradoxically increased in thalamocortical neurons from two rat models of absence epilepsy (48), although we discovered many changes wrought by the G32R mutation that decreased mutant receptor function. This could indicate that the G32R mutation might primarily promote hyperexcitability through cortical and/or postsynaptic (i.e. phasic current-mediating) GABA A receptors. If so, this could suggest a reason for this mutation being associated with childhood absence epilepsy, because cortical ␤3 subunit expression declines throughout childhood. Thus, it is possible that deficits in GABA A receptors containing ␤3 subunits could affect children more significantly because a larger proportion of their cortical receptors contain ␤3 subunits and that associated epilepsy syndromes might remit as ␤2 subunits displace ␤3 subunits in adulthood.