GABAA Receptor α1 Subunit Mutation A322D Associated with Autosomal Dominant Juvenile Myoclonic Epilepsy Reduces the Expression and Alters the Composition of Wild Type GABAA Receptors*

A GABAA receptor (GABAAR) α1 subunit mutation, A322D (AD), causes an autosomal dominant form of juvenile myoclonic epilepsy (ADJME). Previous studies demonstrated that the mutation caused α1(AD) subunit misfolding and rapid degradation, reducing its total and surface expression substantially. Here, we determined the effects of the residual α1(AD) subunit expression on wild type GABAAR expression to determine whether the AD mutation conferred a dominant negative effect. We found that although the α1(AD) subunit did not substitute for wild type α1 subunits on the cell surface, it reduced the surface expression of α1β2γ2 and α3β2γ2 receptors by associating with the wild type subunits within the endoplasmic reticulum and preventing them from trafficking to the cell surface. The α1(AD) subunit reduced surface expression of α3β2γ2 receptors by a greater amount than α1β2γ2 receptors, thus altering cell surface GABAAR composition. When transfected into cultured cortical neurons, the α1(AD) subunit altered the time course of miniature inhibitory postsynaptic current kinetics and reduced miniature inhibitory postsynaptic current amplitudes. These findings demonstrated that, in addition to causing a heterozygous loss of function of α1(AD) subunits, this epilepsy mutation also elicited a modest dominant negative effect that likely shapes the epilepsy phenotype.

GABA A Rs 2 are ligand-gated ion channels that provide the major source of inhibitory control to the mammalian central nervous system. Each GABA A R is a pentamer whose five subunits arise from seven subunit families that contain multiple subtype isoforms. Neurons preferentially express GABA A Rs composed of distinct combinations of subunit isoforms in different brain regions at well defined times in development (1)(2)(3).
The missense AD mutation in the GABA A R ␣1 subunit gene (GABRA1) causes ADJME (8), a monogenic form of a common epilepsy syndrome that begins at a distinct developmental time point (adolescence) and confers myoclonic, generalized tonicclonic, and absence seizures as well as neuropsychiatric comorbidities (9). We demonstrated previously that the AD mutation, which substitutes a negatively charged aspartate for a neutral alanine within the M3 transmembrane domain, causes the ␣1(AD) subunit to misfold with altered topology (10). Cells rapidly degrade the misfolded ␣1(AD) subunit through both proteasome-and lysosome-mediated processes (10,11). Therefore, the ␣1(AD) subunit is expressed at substantially lower levels than the wild type ␣1 subunit. GABA A Rs that do incorporate the residual, nondegraded ␣1(AD) subunits exhibit substantially altered electrophysiological properties (8,12,13). Therefore, a major consequence of heterozygous expression of the AD mutation is a heterozygous loss of functional ␣1 subunits.
However, it is unlikely that ␣1 subunit haploinsufficiency alone explains the entire ADJME phenotype. Heterozygous Gabra1 knock-out mice do not exhibit spontaneous behavioral seizures. Moreover, a recently discovered GABRA1 frameshift mutation that causes complete elimination of the mutant ␣1 subunit protein causes a much milder form of epilepsy than juvenile myoclonic epilepsy (14,15). These findings suggest that despite being misfolded and substantially degraded, the residual ␣1(AD) subunit likely has other pathophysiological actions that lead to a more severe phenotype than would be caused by haploinsufficiency alone. Recently, an extremely small dominant negative effect caused by another epilepsy-associated GABA A R subunit mutation, ␥2(R43Q), proved to be critical for causing the epilepsy phenotype (16 -19). Therefore, we set out to determine whether the ␣1(AD) subunit altered wild type GABA A R expression, physiology, or cellular viability.

MATERIALS AND METHODS
Cell Culture-HEK293T cells were purchased from American Type Culture Collection and were cultured at 37°C in 5% CO 2 , 95% air in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) with 10% fetal bovine serum (FBS, Invitrogen) and 100 IU/ml streptomycin and penicillin (Invitrogen). Cells were replated twice weekly.
We adapted our neuron culture protocol from one described previously (20). Briefly, we dissected cerebral cortices from the brains of embryonic day 18 (E18) Sprague-Dawley rat pups. After trypsin digestion and trituration, we plated 6.7 ϫ 10 5 neurons in 35-mm dishes on polyornithine-coated coverslips. The neurons were initially cultured in DMEM containing 80 M glutamine, 8% FBS, and 8% F-12 nutrient mixture (Invitrogen). After 3 days, we added cytosine arabinoside to a final concentration of 1 M to control glial proliferation (Sigma). On the following day, we exchanged the media with Neurobasal media that contained B27 supplement (Invitrogen). One-half volumes of Neurobasal media were exchanged three times/week.
Expression Vectors and Transfection-The pcDNA3.1 plasmids containing cDNAs that encode human ␣1, ␣1(AD), ␤2S, and ␥2S GABA A R receptor subunits were described previously (10,13,21). The DNA encoding the ␣3 subunit was obtained from Origin Technologies, Inc. It was subcloned into the pcDNA3.1 vector and modified to match the coding sequences found in the GenBank TM and Swiss Protein Databases. The cDNA encoding hemagglutinin (HA) epitope-tagged transforming growth factor ␣ (TGF HA ) protein was generously contributed by Dr. Robert Coffey, Vanderbilt University. The pmaxGFP vector was from Lonza. The wild type and dominant negative K44A dynamin1 plasmids were a kind gift from Dr. Pietro De Camilli (Yale University). We purchased the pLVX-IRES-ZsGreen1 bicistronic vector from Clontech.
In this study, we numbered amino acids starting from the initiation methionine in the signal peptide. Using standard molecular biology techniques, we deleted the codon that encodes leucine 31 (leucine 4 as numbered from the putative N terminus of the mature peptide) from the human ␣1 subunit cDNA (␣1 h ) to make a cDNA encoding the rat ␣1 subunit (␣1 r ). We also introduced the nucleotides encoding the hemagglutinin (HA) epitope tag (YPYDVPDYA) between the codons encoding the 31st and 32nd amino acids (fourth and fifth amino acids as numbered from the putative N terminus of the mature peptide) of the ␣1 h subunit to form the ␣1 HA subunit. Similarly, we inserted the HA epitope in the homologous positions of the ␤2 and ␥2 sequences to form the ␤2 HA and ␥2 HA subunits. We made the AD mutation in the ␣1 HA and ␣1 r subunit cDNA using the QuikChange site-directed mutagenesis kit (Stratagene). To make constructs for use in neuronal transfections, we inserted the cDNA encoding the human ␣1, ␣1(AD), ␣1 HA , and ␣1(AD) HA subunits into the pLVX-IRES-ZsGreen1 vector.
We transfected HEK293T cells in 6-cm dishes with 3 g of total cDNA using the FuGENE 6 transfection reagent (Roche Diagnostics, 3 l/g DNA). The amount of cDNA encoding each subunit is listed in the figure legends. If the total GABA A R subunit cDNA was less than 3 g, we added sufficient empty pcDNA3.1 vector so that the total DNA transfected equaled 3 g. Neurons were transfected 10 days after plating (DIV10) using a calcium phosphate method as described elsewhere (22).
Biotinylation and Western Blot Assays-The biotinylation and Western blot assays have been described previously (13,23). Briefly, live cells were washed in PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (PBS-CM) and incubated with the membrane-impermeable, amine-reactive biotinylation reagent, NHS-SS-biotin (3 mM, Thermo), for 15 min at 4°C. The biotinylation reaction was quenched by washing the cells three times with PBS-CM containing 100 mM glycine.
We lysed the biotinylated cells in modified radioimmunoassay solution (RIPA, 20 mM Tris, pH 7.4, 1% Triton X-100, 250 mM NaCl, 0.5% deoxycholate, 0.1% SDS) that also contained 1 pellet of protease inhibitor per 10 ml (Roche Diagnostics). The lysates were centrifuged at 10,000 ϫ g for 30 min, and the protein concentration was measured using a bicinchoninic acidbased assay (Thermo). To determine total GABA A R subunit expression, we applied the unpurified lysates directly to a 10% SDS-polyacrylamide gel. To measure surface protein, we incubated equal masses of biotinylated lysates with immobilized neutravidin beads (Thermo) overnight. The neutravidin beads were washed with RIPA buffer, and the biotinylated protein was liberated by incubation with Laemmli sample buffer and fractionated on a 10% SDS-polyacrylamide gel. We confirmed that the proteins of interest did not nonspecifically adhere to the neutravidin beads by also incubating lysates from nonbiotinylated cells with the beads before applying them to SDS-PAGE (data not shown). In each experiment, we confirmed that the biotinylation reagent did not modify intracellular proteins by measuring GAPDH immunoreactivity in the neutravidin-purified lysates (data not shown).
After incubation with the primary antibody, the immunoblots were incubated with a horseradish peroxidase-coupled goat anti-mouse or anti-rabbit secondary antibody (Jackson ImmunoResearch, 1:5000 dilution) and then visualized with a chemiluminescent detection system (Amersham Biosciences) using a digital imager (Alpha Innotech). The integrated intensities of the Western blot bands were calculated using the ImageJ software (National Institutes of Health). We confirmed that the integrated intensities of all the bands remained in the linear range of detection by loading three different masses of protein on the gels and performing a linear regression through the origin of the integrated intensity versus protein mass. The integrated intensities of all proteins were normalized to the loading control (ATPase).
Flow Cytometry and Fluorescent Resonance Energy Transfer (FRET)-Flow cytometry and FRET were performed as described previously (24,25). Briefly, transfected cells were detached from the plastic plates by incubating them with trypsin for 1 min at room temperature, a process that does not cause degradation of surface GABA A R protein (24). All subsequent steps were performed at 0°C. We stopped the trypsinization by adding 5 ml of FACS buffer (PBS containing 2% FBS and 0.05% sodium azide). The cells were pelleted and placed into 96-well plates. To measure surface protein expression, the cells were stained without fixation, and to measure total protein, the cells were treated with a commercial fixation/permeabilization reagent (BD Biosciences) before staining. The cells were stained with the anti-␣1 h -647 (1:100), -␣1 r (1:50), -␣3 (1:50), and HA-555 or HA-647 (1:100) antibodies for 1 h. The cells stained with the anti-␣1 r and -␣3 antibodies were then stained with A647-conjugated goat anti-rabbit antibody (1:250, Invitrogen). In each experiment, we confirmed that these dilutions represented saturating concentrations of the respective antibodies by also staining negative control cells and wild type GABA A R cells with a 2-fold higher concentration of antibody and verifying that the fluorescence did not substantially change (data not shown). After staining, the cells were fixed with 2% paraformaldehyde in PBS.
For each sample, we analyzed 10,000 cells on an LSR II digital flow cytometer (BD Biosciences). We identified viable cells based upon their forward and side scatter properties (data not shown) that were originally determined from staining with the membrane-impermeable dye, 7-aminoactinomycin D (Invitrogen) (24,25). The A647 fluorophore was excited by a 635 nm laser and detected with a 675/20 bandpass filter, and the A555 fluorophore was excited with a 535 nm laser and detected with a 575/26 bandpass filter. GFP was excited by a 488-nm laser and detected using a 530/30 bandpass filter. In each of the samples, we measured the mean GFP, A555, and A647 fluorescence in viable cells using the FlowJo 7.2 software (Treestar) and subtracted the mean fluorescence obtained from negative control cells.
We measured the FRET between the samples stained with A555-coupled antibody (donor) and a A647-coupled antibody (acceptor) by exciting the A555 fluorophore with the 535-nm laser and detecting the FRET to the A647 fluorophore using 670/20 bandpass filter. To eliminate the fluorescence resulting from direct 535 nm stimulation to the A555 and A647 fluorophores from the FRET signal, we performed spectral compensation using singly stained samples and the FlowJo 7.2 software (Treestar). We then subtracted the mean FRET fluorescence from the samples stained with only the A647 fluorophore from those stained with both the A555 and A647 fluorophores to calculate the specific mean FRET fluorescence.
Confocal Microscopy and Image Analysis-Microscopic analyses were performed using the coverslips on which the neurons were originally plated. On DIV17, 7 days after transfection, we washed the neurons three times in PBS and then fixed them in 4% paraformaldehyde, 4% sucrose in PBS for 15 min at room temperature. We washed the neurons three times with PBS and then incubated them with 0.1% Triton X-100 in PBS for 5 min before washing three more times with PBS. We blocked nonspecific binding by incubating the fixed neurons with 0.5% BSA in PBS for 1 h at room temperature. We stained the neurons with either anti-NeuN (1:200) or anti-␣1 h antibody (1:1000) diluted in 0.2% BSA in PBS overnight at 4°C. The following day, we washed the neurons five times with PBS and then incubated them for 1 h with Alexa 546-conjugated goat anti-mouse antibody diluted in PBS containing 0.2% BSA. We washed them with PBS five more times and counterstained them with 4Ј,6diamidino-2-phenylindole (DAPI).
Confocal imaging was performed through the use of the Vanderbilt University Medical Center Cell Imaging Shared Resource. We visualized the neurons with an Olympus FV1000 confocal microscope using a 60ϫ/1.45 numerical aperture plan-apochromat objective. We used the recommended Olympus laser and filter combinations to separately detect DAPI, ZsGreen1, and A546 fluorescence and used singly stained specimens to verify that these fluorophores did not spectrally overlap. We adjusted the pinhole of all channels to obtain 3-m sections, and we obtained a single section at the level of the nucleus. In each experiment, we adjusted the laser intensity and detector sensitivity to utilize the full linear range of detection, and we used the same settings for all the samples acquired during the experiment. Images were obtained with 12-bit, 1024 ϫ 1024 pixel resolution.
To determine the percentage of cells in the culture that consisted of neurons, we stained the cells with an antibody against NeuN. In three different regions of the slide, we determined the fraction of cells stained with DAPI that were also stained with NeuN. To determine the percentage of neurons transfected, we calculated the fraction of neurons stained with NeuN that also expressed ZsGreen1.
To determine the expression of recombinant ␣1 h and ␣1(AD) h subunits, we visually identified several ZsGreen1-expressing neurons from each culture. We then acquired confocal images of each neuron. We analyzed the images off line using the ImageJ software. We determined the mean fluorescence of a 5-pixel wide line drawn either through the nucleus (for ZsGreen1 fluorescence) or though the soma excluding the nucleus (for ␣1 h fluorescence). We normalized the fluorescence of each neuron to the mean fluorescence of the neurons transfected with wild type ␣1 h that were imaged on the same day.
We visually identified transfected pyramidally shaped neurons by ZsGreen1 fluorescence using an inverted fluorescence microscope (Nikon). After making a gigaohm seal with the soma, we broke through the membrane to establish electrical contact with the cytoplasm. We voltage-clamped the neurons at Ϫ60 mV using an Axopatch 200B amplifier and using the pClamp software (Molecular Devices). We recorded more than 100 mIPSC events from each neuron. Each mIPSC recording was analyzed off line by an investigator who was blinded to the transfection conditions. The mIPSCs were chosen manually using the Mini Analysis software (Synaptosoft). The detection threshold was set at 15 pA. The inter-mIPSC intervals, mIPSC amplitude, rise time, and decay time constants were recorded.
Data Analysis-Expression levels of the GABA A R subunits were normalized to those obtained from the wild type condition and were reported as means Ϯ S.E. Statistical significance was determined using the Student's paired or unpaired t test or paired or unpaired analysis of variance test, as appropriate.
To simplify the presentation of our data, which determined the effects of co-transfecting different masses of ␣1(AD) subunit cDNA on wild type GABA A R expression, we fit the expression levels to the sigmoidal concentration-response equation as follows: Y ϭ 100%/(1 ϩ (X/IC 50 ) Hill slope ), where Y is the expression level of the GABA A R subunit normalized to that obtained in the absence of the ␣1(AD) subunit, and X is the mass of ␣1(AD) subunit cDNA. We must emphasize that although we connected our data points with the fit concentration dependence equation, we are not asserting that the ␣1(AD) subunit necessarily acted by binding competitively with wild type GABA A Rs. Therefore, we made statistical comparisons between the mean expression levels instead of the fit IC 50 and Hill slope values.
The mIPSC inter-mIPSC intervals, rise times, decay time constants, and peak amplitudes were compared using both a nonparametric analysis of variance and Kolmogorov-Smirnov tests. For the Kolmogorov-Smirnov test, the Bonferroni correction was applied to assess statistical significance among the three transfection conditions.

RESULTS
Heterozygous Cells Express Greater Amounts of Total ␣1 Subunit Protein than Hemizygous Cells-As a first step to determine whether expression of the ␣1(AD) subunit altered expression of ␣1␤2␥2 receptors, we co-transfected HEK293T cells with 0.250 g of ␤2 and ␥2 subunit cDNA and either 0.250 g of ␣1 subunit (wild type (WT)), 0.125 g of ␣1 subunit (hemizygous), 0.125 g of ␣1 subunit plus 0.125 g ␣1(AD) subunit (heterozygous), or 0.250 g of ␣1(AD) subunit (homozygous) cDNA. We use the terms hemizygous, heterozygous, and homozygous solely as a convenient naming convention to convey the ratio of masses of wild type to mutant cDNAs transfected. We used biotinylation assays and Western blots to determine the amount of surface and total ␣1 subunit expression in each of these transfection conditions (Fig. 1, A-D). As described previously, both surface and total heterozygous and homozygous ␣1 subunit expression differed substantially from each other and from wild type ␣1 subunit expression. Here, we found that both surface and total hemizygous ␣1 subunit expression was also substantially reduced compared with wild type expression. However, there was no significant difference between hemizygous and heterozygous ␣1 subunit expression.
We next used flow cytometry as a second technique to measure surface and total ␣1 subunit expression (Fig. 1, E-H). Previous studies demonstrated that flow cytometry can measure total and surface GABA A R subunit expression with high preci-FIGURE 1. Effect of heterozygous ␣1(AD) subunit expression on surface and total ␣1 subunit expression. We transfected HEK293T cells with empty plasmid (negative control) or 0.250 g of ␤2 and ␥2 subunit cDNA and either 0.250 g of wild type ␣1 (wt), 0.125 g ␣1 (hemizygous (hemi)), 0.125 g each of ␣1 and ␣1(AD) (heterozygous (het)), or 0.250 g of ␣1(AD) (homozygous (hom)) cDNA. We biotinylated surface proteins and performed Western blots to determine the surface (A and C) and total (B and D) GABA A R ␣1 subunit expression relative to that of the ATPase ␣1 subunit (loading control). Surface and total hemizygous and heterozygous ␣1 subunit expression was greater than that of homozygous expression and was significantly reduced relative to wild type expression (*), but there was no significant (ns) difference between them (n ϭ 5). We also performed flow cytometry assays (E-H) as a second method to quantify the surface and total ␣1 subunit for cells transfected with negative control (filled histogram), wild type (solid line), hemizygous (dotted line), heterozygous (dashed line), or homozygous (data not shown) GABA A R. There was no significant difference between hemizygous and heterozygous receptors in surface ␣1 subunit expression (n ϭ 7), but total heterozygous ␣1 subunit expression was larger than that of hemizygous ␣1 subunit expression (n ϭ 6, p ϭ 0.03).
sion and thus could potentially detect differences between hemizygous and heterozygous expression that could not be detected with Western blot assays (24). We found that flow cytometry did quantify ␣1 subunit expression with smaller standard deviations than Western blot experiments and was able to detect that heterozygous cells expressed 9% more total ␣1 subunit than hemizygous cells (p Ͻ 0.05). However, the flow cytometry experiments did not detect a difference between heterozygous and hemizygous cells in surface ␣1 subunit expression.
Wild Type ␣1 Subunit Accounted for 92-96% of the Surface ␣1 Subunit with Heterozygous Expression-The experiments in Fig. 1 demonstrated that compared with hemizygous cells, heterozygous cells possessed increased amounts of total ␣1 subunit protein but did not have significantly different amounts of surface ␣1 subunit protein. We thought that the increase in total ␣1 subunit protein expression in the heterozygous cells likely resulted from the residual, nondegraded ␣1(AD) subunit in intracellular compartments and that the lack of a significant change in surface expression resulted because the residual ␣1(AD) subunit did not efficiently traffic to the cell surface. Therefore, we next determined the fraction of surface ␣1 subunit with heterozygous expression that consisted of wild type or mutant ␣1 subunit.
Previously, we used ␣1 and ␣1(AD) subunits fused to different fluorescent proteins and estimated that 5-25% of surface ␣1 subunit with heterozygous expression consisted of the ␣1(AD) subunit (21). Here, we used flow cytometry with two naturally occurring ␣1 subunits to obtain a more precise estimate of the amount of wild type and mutant ␣1 subunit on the cell surface in heterozygous expression. Although fluorescent protein-and epitope-tagged GABA A R subunits incorporate into pentamers to form functional receptors, the addition of the fused fluorescent protein or epitope tag alters post-translational processing and surface expression of the GABA A R subunits (10,21). Therefore, we differentiated between wild type and mutant ␣1 subunits by transfecting cells using natural ␣1 and ␣1(AD) subunits from two different species (human and rat) and then detecting them individually with species-selective antibodies. The mature human ␣1 (␣1 h ) and rat ␣1 (␣1 r ) subunit proteins differ by only one amino acid; human ␣1 subunits possess an additional leucine at position 31 (position 4 as numbered from the putative N terminus of the mature peptide) that is absent in the rat sequence ( Fig. 2A). Two commercially available antibodies, mouse anti-␣1 h and rabbit anti-␣1 r , bind specifically to their respective subunits.
Initial flow cytometry experiments demonstrated that the species-selective antibodies were very specific for the corresponding species of ␣1 subunit (supplemental Fig. 1). Moreover, the anti-␣1 h -647 antibody demonstrated remarkable species specificity such that compared with cells transfected with 0.250 g of ␤2 and ␥2 subunits and only 0.125 g of the ␣1 h subunit, cells transfected with a 16-fold excess (2 g) of ␣1 r subunit exhibited 0 Ϯ 0% the surface fluorescence and 2 Ϯ 0.2% of the total fluorescence (data not shown). Because the anti-␣1 r antibody could not be directly coupled to a fluorophore, its detection required use of a fluorescent secondary antibody, which produced high levels of background staining in perme-abilized cells. Therefore, the anti-␣1 r antibody was only used for surface staining.
We transfected HEK293T cells with 0.250 g of ␤2 and ␥2 subunit cDNA and heterozygous ␣1 subunit consisting of either 0.125 g of cDNA each of ␣1 h and ␣1(AD) h subunits (both human, both h ), ␣1 r and ␣1(AD) r subunits (both rat, both r ), ␣1 h and ␣1(AD) r subunits (WT h /AD r ), or ␣1 r and ␣1(AD) h subunits (WT r /AD h ). We stained surface protein with the anti-␣1 h subunit antibody, quantified the fluorescence with flow cytometry, and divided the WT h /AD r and WT r /AD h fluorescence by that obtained from cells transfected with both human. These data demonstrated that with heterozygous expression, wild type ␣1 h subunits and mutant ␣1(AD) h subunits included 96 Ϯ 3 and 2 Ϯ 0.1% of the surface ␣1 subunit expression, respectively (Fig. 2B). These results were concordant with the complementary experiment in which we stained surface protein with the anti-␣1 r antibody and found that the wild type ␣1 r subunits and mutant ␣1(AD) r subunits included 92 Ϯ 2 and 5 Ϯ 0.4% of the surface ␣1 subunit expression (Fig.  2C). In permeabilized cells, wild type ␣1 h subunits and mutant FIGURE 2. Fraction of wild type ␣1 and mutant ␣1(AD) subunits in heterozygous expression. We depicted the N-terminal amino acids of the mature human (␣1 h ) and rat (␣1 r ) ␣1 subunits (A) with the amino acids numbered starting at the N termini of the signal peptides. The Gln-28 residues are located at the putative N termini of the mature peptides. The full-length mature ␣1 h and ␣1 r sequences are identical except that the rat sequence lacks leucine 31 (leucine 4 as numbered from the putative N terminus of the mature peptide). We transfected HEK293T cells with mock vector (negative control, filled histograms) or ␤2 and ␥2 subunits (0.250 g) and heterozygous ␣1 subunits in which the sequences of the wild type and mutant subunits were both human or both rat (both h , both r , solid lines and gray bars). We also transfected cells with heterozygous receptors in which the sequence of the wild type subunit was human and the mutant subunit was rat (wt h /AD r , dashed lines, white bars), and the sequence of the wild type subunit was rat and the mutant subunit was human (wt r /AD h , dotted lines, black bars). We stained surface receptors with either the anti-␣1 h (B) or ␣1 r antibody (C) and total receptors with the anti-␣1 h antibody (D). We determined the fraction of surface heterozygous subunits composed of wild type subunit by dividing ␣1 h mean fluorescence in WT h /AD r , by the ␣1 h mean fluorescence in both h cells (B, 96 Ϯ 3%). We obtained a similar result when we divided ␣1 r staining in WT r /AD h cells by that in both r cells (C, 92 Ϯ 2%). Similarly, we determined the fraction of mutant subunit in surface heterozygous receptors by dividing ␣1(AD) h staining in WT r /AD h cells by that in both h cells (B, 2 Ϯ 0.1%) and dividing ␣1(AD) r staining in WT h /AD r cells by both r cells (C, 5 Ϯ 0.4%). In total, heterozygous subunits, the wild type ␣1 h subunit included 91 Ϯ 4% heterozygous expression and the mutant ␣1 h subunit included 8 Ϯ 4% heterozygous expression. In both surface and total heterozygous expression, the contribution of the wild type subunit was modestly but significantly (*, p Ͻ 0.05) less than 100%. ␣1(AD) h subunits included 91 Ϯ 4 and 8 Ϯ 4% of total ␣1 subunit (Fig. 2D). Therefore, the preponderance (92-96%) of ␣1 subunits on the surface of heterozygous cells was wild type ␣1 subunits.
Our data also demonstrated that in both surface and total heterozygous expression, the contribution of the wild type subunit was modestly, but significantly (p Ͻ 0.05), less than 100%. This result suggests that the mutant ␣1(AD) subunit could cause a dominant negative effect either by substituting for wild type ␣1 subunit or by reducing wild type ␣1 subunit surface expression.
Finally, we compared the surface expression of wild type ␣1 h or ␣1 r protein in heterozygous cells (WT h /AD r or WT r /AD h ) with the corresponding hemizygous receptors (hemizygous ␣1 h or ␣1 r ). The presence of the mutant subunit caused small but, in the case of the ␣1 r protein, statistically significant reductions in surface wild type ␣1 subunit protein expression (8 Ϯ 2%, p ϭ 0.009, not shown). This result also suggested that the ␣1(AD) subunit could reduce wild type ␣1 subunit surface expression. Fig. 2 demonstrated that when heterozygously expressed in HEK293T cells, mutant ␣1(AD) subunits included 8% of the total ␣1 subunits and 2-5% of the surface ␣1 subunits. However, the fraction of mutant ␣1(AD) subunits expressed in neurons from ADJME patients is unknown. If the neurons of patients expressed a greater fraction of total ␣1(AD) subunit than HEK293T cells, would more ␣1(AD) subunit be expressed on the cell surface and confer a dominant negative effect by substituting mutant ␣1(AD) subunits for wild type ␣1 subunits?

␣1(AD) Subunit Did Not Substitute for Wild Type ␣1 Subunits on the Cell Surface-Experiments in
In a heterologous expression system, we can increase the amount of protein expression by increasing the mass of cDNA transfected. Therefore, we transfected cells with ␤2 and ␥2 subunit cDNA (0.250 g) and hemizygous (0.125 g) ␣1 r subunits and a range (0-2 g) of wild type ␣1 h or ␣1(AD) h ␣1 cDNA. We quantified the amount of relative total and surface ␣1 h and ␣1(AD) h subunit expression by flow cytometry and normalized all the values to the total or surface expression cells transfected with hemizygous ␣1 h ("positive control," ␣1 h ␤2␥2 0.125, 0.250, and 0.250 g).
As expected, increasing the masses of wild type ␣1 h and ␣1(AD) h subunit cDNA increased the amount of total ␣1 h and ␣1(AD) h subunit expression (Fig. 3, A and C). The total ␣1(AD) h subunit expression increased linearly with the mass of cDNA from 0 to 2 g, and the total ␣1 h subunit expression increased linearly from 0 to 0.5 g of cDNA before saturating. Because the ␣1(AD) subunit was degraded, total ␣1(AD) h subunit expression was smaller than total ␣1 h subunit expression.
Increasing the mass of ␣1 h subunit cDNA increased the amount of surface ␣1 h subunit expression. In contrast, increasing the mass of ␣1(AD) h subunit cDNA caused little change in surface ␣1(AD) h expression (Fig. 3, B and D). Replotting the surface ␣1 h and ␣1(AD) h expression versus total ␣1 h and ␣1(AD) h expression (Fig. 3E) demonstrated that at the same amount of total ␣1 h and ␣1(AD) h subunit expression there was 8-fold more ␣1 h subunit on the cell surface. Repeating these studies using cells lacking the hem-izygous ␣1 r subunit produced identical results (data not shown). These data demonstrated that in addition to reducing total ␣1(AD) expression, the AD mutation also strongly inhibited the surface expression of receptors containing the ␣1(AD) subunit. Therefore, even if neurons from ADJME patients expressed larger relative amounts of ␣1(AD) subunits than HEK293T cells, the mutant subunits would not produce a dominant negative effect by substituting for wild type subunits on the cell surface. C and D, we depicted the ␣1(AD) h (F) and ␣1 h (E) fluorescence normalized to positive control for each mass of ␣1(AD) h or ␣1 h subunit cDNA that we transfected. When measuring total ␣1(AD) h and ␣1 h expression (C), increasing both ␣1(AD) h and ␣1 h subunit cDNA caused linear increases in ␣1(AD) h and ␣1 h subunit protein expression until ␣1 h subunit expression saturated when it was greater than 300% that of positive (pos) control. When measuring surface ␣1(AD) h and ␣1 h expression (D), increasing ␣1 h subunit cDNA caused proportional increases in ␣1 h subunit expression, but increasing ␣1(AD) h subunit cDNA resulted in very little change in surface ␣1(AD) h subunit expression. E, we replotted the surface (surf) ␣1(AD) h and ␣1 h subunit expression as a function of total expression. Both surface ␣1(AD) h and ␣1 h subunit expression increased linearly with total expression (r 2 ϭ 0.82 and 0.95, respectively, n Ն 3). However, there was an 8-fold greater increased surface ␣1 h subunit expression than ␣1(AD) h subunit expression for equal increases in total expression. To determine whether the presence of the wild type ␣1 r subunit facilitated the surface trafficking of ␣1(AD) h subunits, we repeated experiments in the absence of the ␣1 r subunit and found no significant difference in the surface expression of the ␣1(AD) h subunit in the absence of the ␣1 r subunit (n ϭ 3, data not shown). Fig. 3 demonstrated that ␣1(AD) subunits did not cause a dominant negative effect by substituting for wild type ␣1 subunits on the cell surface, and experiments in Fig. 2 suggested that compared with hemizygous expression the heterozygous expression reduced the wild type ␣1 subunit on the cell surface. Therefore, we next determined the effect transfecting different amounts of ␣1(AD) r subunit on wild type ␣1 h subunit expression.

␣1(AD) Subunit Reduced Surface Expression of the Wild Type ␣1 Subunit-Experiments in
We transfected HEK293T cells with ␤2 and ␥2 subunits (0.250 g), hemizygous (0.125 g) human ␣1 h subunits, and a range (0 -2 g) of rat ␣1(AD) r cDNAs and measured surface and total ␣1 h expression by flow cytometry using the anti-␣1 h subunit antibody (Fig. 4). For comparison, we also determined the effect of wild type ␣1 r subunit on ␣1 h ␤2␥2 expression.
Because the wild type ␣1 r subunits are not detected by the ␣1 h subunit antibody, they reduce ␣1 h subunit expression by substitution rather than ␣1(AD) r subunit-mediated inhibition of expression.
The ␣1(AD) r subunit caused a concentration-dependent reduction in surface and total ␣1 h expression (Fig. 4, A-D). As expected, the ␣1(AD) r subunit caused smaller reductions in ␣1 h subunit expression than the wild type ␣1 r subunit. Nonetheless, the ␣1(AD) r -mediated reduction of ␣1 h subunit expression demonstrated that nondegraded ␣1(AD) subunit was not inert and caused a small but potentially clinically significant dominant negative effect by reducing wild type ␣1 subunit expression.
To confirm that transfection of the ␣1(AD) r subunit did not reduce ␣1 h ␤2␥2 receptor expression simply by inhibiting the transcriptional or translational capacity of the cells, we compared the effects of transfecting 2 g of ␣1(AD) r or pmaxGFP on ␣1␤2␥2 expression. Cells transfected with pmaxGFP demonstrated strong GFP fluorescence (data not shown). However, unlike the ␣1(AD) r subunit, pmaxGFP did not alter surface or total ␣1␤2␥2 receptor expression (supplemental Fig. 2, A and B).
We used biotinylation assays with Western blotting as a second method to determine the effect of the ␣1(AD) subunit on surface wild type ␣1 h subunit expression. We transfected HEK293T cells with ␤2 and ␥2 subunits (0.250 g) and hemizygous ␣1 h subunits with or without 2 g of ␣1(AD) r subunit cDNA. We performed biotinylation assays with Western blots and detected the ␣1 h subunit with the anti-␣1 h subunit antibody. Similar to the flow cytometry studies, we found that the ␣1(AD) r subunit reduced the surface expression of the wild type ␣1 h subunit (Fig. 4, E and F).
Because the ␣1(AD) subunit is grossly misfolded, it could potentially be a toxic protein and reduce ␣1␤2␥2 expression by nonspecifically decreasing cellular viability. During flow cytometry analyses, we gated viable cells based on their forward and side scatter properties that corresponded to the population of viable cells that excluded the membrane-impermeable dye, 7-aminoactinomycin D (24). We found that transfection of 2 g of ␣1(AD) subunit cDNA did not alter the percentage of "viable cells" (62 Ϯ 3%) when compared with cells transfected in the absence of the ␣1(AD) subunit (60 Ϯ 3%, n ϭ 5, p ϭ 0.710; data not shown).
To determine whether the ␣1(AD) subunit inhibited surface expression of ␣1␤2␥2 receptors by nonspecifically preventing surface trafficking of proteins processed through the secretory pathway, we measured the effect of the ␣1(AD) subunit on the surface expression of two membrane proteins not thought to interact with GABA A Rs. Using biotinylation assays and Western blots, we found that transfection of 2 g of ␣1(AD) subunit did not alter the surface expression of the endogenous HEK293T cell Na ϩ /K ϩ -ATPase ␣1 subunit ( Fig.  4E; 97 Ϯ 10%, n ϭ 5, p ϭ 0.776). Next, we determined the effect of the ␣1(AD) r subunit on surface expression of recombinantly expressed TGF HA protein. We transfected cells with ␣1␤2␥2 receptors, TGF HA , and with or without 2 g of ␣1(AD) r cDNA. We quantified expression of TGF HA by flow cytometry with the anti-HA-647 antibody and found that the ␣1(AD) subunit did FIGURE 4. Effect of ␣1(AD) subunit expression on surface and total ␣1␤2␥2 expression. We transfected HEK293T cells with empty vector (negative control, shaded histograms) or 0.250 g of ␤2 and ␥2S subunits and 0.125 g ␣1 h subunit cDNA (hemizygous) and with varying masses (0 -2 g) of ␣1 r or ␣1(AD) r cDNA. We quantified the surface (A and C) and total (B and D) ␣1 h subunit expression using flow cytometry with anti-␣1 h subunit antibodies. A and B, we depicted hemizygous fluorescence as a solid line and hemizygous ϩ 2 g of ␣1(AD) r fluorescence as a dashed line. C and D, we plotted ␣1(AD) h (F) and ␣1 h (E) subunit fluorescence normalized to that of hemizygous subunit fluorescence versus the masses of ␣1(AD) h or ␣1 h cDNA that was transfected. Increasing the masses of ␣1(AD) r subunit cDNA reduced surface and total ␣1 h subunit expression. We also determined the effect 2 g of ␣1(AD) r subunit cDNA co-transfection on surface ␣1 h ␤2␥2 receptor expression using biotinylation assays with Western blot (E and F). We stained the Western blots with the anti-␣1 h subunit antibody and quantified surface ␣1 h subunit relative to that of the Na ϩ /K ϩ -ATPase ␣ subunit (loading control (cont)). The addition of 2 g of ␣1(AD) r subunit cDNA (ϩAD) reduced surface expression of ␣1 h ␤2␥2 receptors by 46 Ϯ 6% (n ϭ 4; p ϭ 0.005). The ␣1(AD) r subunit did not alter the surface expression of the endogenous ATPase ␣ subunit (97 Ϯ 10%, n ϭ 5).
not reduce expression of TGF HA (116 Ϯ 10%, see supplemental Fig. 3). These data demonstrated that the dominant negative effect conferred by ␣1(AD) subunit did not result simply by nonspecifically inhibiting protein expression through the secretory pathway. ␣1(AD) Subunit Reduced ␣1 Subunit Expression in a Dynamin-independent Manner-Bradley et al. (11) demonstrated that total homozygous ␣1(AD)␤2␥2 expression increased by co-expression of dominant negative dynamin a mutant protein that inhibits clathrin-mediated endocytosis. These data suggested that in addition to causing degradation of the ␣1(AD) subunit by ER-associated degradation, the AD mutation also increased endocytosis of ␣1(AD)␤2␥2 receptors from the cell surface.
Here, we determined whether the ␣1(AD) subunit conferred its dominant negative effect by increasing ␣1␤2␥2 receptor endocytosis through a dynamin-dependent mechanism. We co-transfected HEK293T cells with ␤2 and ␥2 subunits and either hemizygous ␣1 h , homozygous ␣1(AD) h , or hemizygous ␣1 h subunits that also contained 2 g of ␣1(AD) r cDNA (ϩAD). In addition, we co-transfected cells with 0.375 g of cDNA expressing either wild type dynamin or a dominant negative mutant dynamin (K44A) that inhibits endocytosis.
We measured the amount of surface and total human ␣1 h or ␣1(AD) h subunit expression by flow cytometry. In agreement with Bradley et al. (11), we demonstrated that co-expression of K44A dynamin increased expression of total homozygous GABA A Rs to a greater extent than "hemizygous" GABA A Rs (Fig. 5B). In addition, we demonstrated that K44A dynamin increased surface expression of homozygous receptors to a greater extent than that of hemizygous receptors, although this latter result was not statistically significant (Fig. 5A). Although the K44A dynamin did cause a greater effect on homozygous than hemizygous expression, the magnitude of the expression of homozygous receptors was still very small, and thus we do not think that enhanced endocytosis of homozygous ␣1(AD)␤2␥2 makes an appreciable contribution to ADJME pathology. More importantly for the goals of this study, K44A dynamin did not increase the total or surface expression of wild type ␣1 h subunits when it was co-expressed with ␣1(AD) r subunits, and thus the mutant ␣1(AD) subunit did not produce the dominant negative effect by augmenting dynamin-mediated endocytosis. Interestingly, not only did the K44A dynamin fail to increase expression of the ϩ␣1(AD) r subunit samples to a greater extent than the hemizygous subunit samples, it did not cause any increase in ␣1 h subunit expression, suggesting that the ␣1(AD) subunit may actually inhibit dynamin-mediated endocytosis. Fig. 5 demonstrated that the ␣1(AD) subunit did not confer its dominant negative effect by increasing dynamin-mediated GABA A R endocytosis. Previous studies demonstrated that nondegraded ␣1(AD) subunits were retained in the ER (21,26). Could the ␣1(AD) subunit confer its dominant negative effect by associating with wild type GABA A R subunits and entrapping them in the endoplasmic reticulum (ER)? Bradley et al. (11) reported that FLAG epitopetagged ␣1(AD) subunits co-immunoprecipitated with ␥2-GFP subunit fusion proteins and thus demonstrated that the mutant ␣1(AD) protein could associate with and potentially entrap wild type GABA A R subunits (11). We extended these studies and used FRET to determine whether ␣1(AD) subunits specifically associated with wild type GABA A Rs. FRET measures the nonradiative transfer of energy between two fluorophores that are situated within at least 10 nm of one another and can be performed in situ in cells using flow cytometry (25,27).
To determine whether or not there were FRET interactions between ␣1(AD) h subunits and wild type GABA A R subunits or TGF HA proteins, we excited the FRET donor near its absorption maximum and detected the emission from the FRET acceptor at its emission maximum. In Fig. 6, A-D, we plotted flow cytometry histograms that depict the intensity of emission in the FRET channel on the abscissa and the number of cells having that intensity on the ordinate. The gray plots in Fig. 6, A-D, depict the data for cells stained with only the FRET acceptor and demonstrate the lack of nonspecific FRET signal when the donor was not present. Similar experiments demonstrated even smaller nonspecific FRET signals when the donor but not the acceptor was present (data not shown). The black plots in Fig. 6, A-D, depict the data for cells stained with both the FRET acceptor and donor. FRET signals are specific when there is  , black), or 0.125 g of human wild type ␣1 h subunit and 2 g ␣1(AD) r subunit (ϩAD, dark gray). In addition, we also co-transfected the cells with either 0.375 g of wild type (wtdyn) or dominant negative (K44A) dynamin. We measured surface and total human ␣1 subunit expression by flow cytometry and normalized each value to the expression of the hemizygous samples that were cotransfected with wild type dynamin. In the insets, we plotted the percentage change in ␣1 h and ␣1(AD) h expression because of the K44A dynamin. The K44A dynamin increased the total and surface ␣1 subunit expression in the hemizygous and homozygous conditions (p Ͻ 0.05), but not alter the dominant effect in the ϩ␣1(AD) r condition (n ϭ 5). The K44A dynamin significantly increased total homozygous ␣1(AD) h expression to a greater extent than hemizygous ␣1 h expression (p Ͻ 0.01). ns ϭ p Ն 0.05; *, p Ͻ 0.05. AUGUST 20, 2010 • VOLUME 285 • NUMBER 34 greater FRET signal in the doubly stained (Fig. 6, A-D, black) than in the singly stained (gray) samples. The regions of the histograms where one finds specific FRET signals are indicated by arrows in Fig. 6, A-D.

GABRA1 A322D Mutation Causes Dominant Negative Effects
We quantified the specific FRET signals as the difference between the mean FRET emissions of the doubly and singly stained samples. Specific FRET data are summarized in Fig.  6E (n Ն 5). Both ␤2 HA and ␥2 HA subunits, but not the negative control protein TGF HA , demonstrated specific FRET with ␣1(AD) h subunits. Although the ␣1 HA also appeared to demonstrate a specific FRET interaction with ␣1(AD) h , this value was not statistically significant (p ϭ 0.128). The lack of FRET interactions between ␣1(AD) h subunits and TGF HA protein did not result from decreased donor fluorescence. As demonstrated in Fig. 6F, detection of TGF HA protein had substantially greater HA-555 florescence than that of the GABA A R subunits.
These results demonstrated that the ␣1(AD) subunit specifically associated with other GABA A R subunits. Based on the prior subcellular localization studies (21,26), this association must occur in the ER. Because GABA A Rs containing ␣1(AD) subunits are strongly inhibited from trafficking to the cell surface (Fig. 3), this result suggests that the ␣1(AD) subunit reduced GABA A R surface expression by associating with and entrapping GABA A R in the ER. ␣1(AD) Subunit Reduced Surface Expression of ␣3 Subunitcontaining GABA A Rs-Neurons predominantly express ␣2 and ␣3 subunits early in development and then express ␣1 subunits later in development. Adult homozygous Gabra1 knock-out mice express increased amounts of ␣3 subunits, which partially compensate for the lack of ␣1 subunits (28 -31). Therefore, we determined the effect of ␣1(AD) subunits on surface expression of ␣3 subunit-containing GABA A Rs.
We transfected HEK293T cells with 0.250 g of ␣3, ␤2, and ␥2 subunit cDNA and varying amounts (0 -2 g) of either ␣1 or ␣1(AD) subunits. As in the experiments in Fig. 4, we included the wild type ␣1 subunit to serve as a basis for comparison between ␣3 subunit substitution by the wild type ␣1 subunit and ␣1(AD)-mediated inhibition of expression. To emulate the GABA A R expression in ADJME patients who have two wild type GABRA3 genes but only one wild type GABRA1 gene, we transfected 0.250 g of ␣3 subunit cDNA instead of 0.125 g as we did for the hemizygous ␣1␤2␥2 conditions. We performed flow cytometry experiments and stained for surface ␣3 subunit expression.
Co-transfecting increasing amounts of ␣1(AD) subunit cDNA caused a concentration-dependent reduction in surface ␣3 subunit expression (Fig. 7, A and B). We confirmed the inhibition of ␣3 subunit expression using biotinylation assays and Western blots. We transfected cells with ␣3, ␤2, and ␥2 subunits with or without 2 g of ␣1(AD) subunit cDNA. We performed biotinylation assays and Western blots and found that ␣1(AD) subunits significantly reduced expression of surface ␣3 subunits (Fig. 7, C and D). As described in Fig. 4, we confirmed that the ␣1(AD) subunit did not inhibit ␣3␤2␥2 receptor expression by competing for transcriptional and translational machinery by measuring the effect of pmaxGFP on ␣3␤2␥2 receptor expression (supplemental Fig. 2C).

␣1(AD) Subunits Reduced Surface Expression of ␣3
Subunitcontaining GABA A Rs to a Greater Extent than ␣1 Subunit-containing GABA A Rs-Surprisingly, when we compared the effects of ␣1(AD) subunits on surface expression of ␣1 (Fig. 4) and ␣3 (Fig. 7) subunits, we found that ␣1(AD) subunits reduced ␣3 subunit expression to a greater extent than ␣1 subunit expression (Fig. 8A). We next determined whether ␣1(AD) subunits also reduced surface expression of partnering ␤2 and ␥2 HA subunits to a greater extent when co-expressed with ␣3 subunits than with ␣1 subunits.
The addition of ␣1(AD) subunit reduced ␣3␤2␥2 HA receptor expression to a greater extent than ␣1␤2␥2 HA receptor expression. Interestingly, when cells expressing ␣1␤2␥2 receptors were transfected with 0.250 and 0.500 g of ␣1(A322D) subunits, there were small increases in ␥2 HA subunit expression, a result that may suggest that in a small fraction of receptors the ␣1(AD) subunit changes the composition of the GABA A R pentamer to one that contains more than one ␥2 HA subunit.
Unlike the transfected heterologous cells used in the previous experiments that expressed either ␣1 or ␣3 subunits, neurons simultaneously express an endogenous mixture of GABA A R subunits, including ␣1, ␣3, and other ␣ subunit subtypes. Therefore, we next determined whether or not ␣1(AD) subunits would reduce surface expression of ␣3 subunits to a greater extent than ␣1 subunits in HEK293T cells simultaneously expressing both ␣1 and ␣3 subunit cDNA.
We transfected HEK293T cells with ␣1 h , ␣3, ␤2, and ␥2 HA subunits (0.125:0.250:0.250:0.250 g of cDNA ratio) and with or without 1 g of ␣1(AD) r subunit cDNA. We quantified the relative surface expression of ␣1 h , ␣3, and ␥2 HA subunits using flow cytometry (Fig. 8, H and I). We found that ␣1(AD) r subunits reduced expression of surface ␣3 subunits (30 Ϯ 4%) to a greater extent than ␣1 h subunits (12 Ϯ 4%, p ϭ 0.010). ␣1(AD) r subunits reduced expression of surface ␥2 HA subunits by a similar amount as it reduced surface ␣1 h subunit expression, 15 Ϯ 1%. These results indicated that in cells expressing both ␣1 and ␣3 subunits, ␣1(AD) subunits produced a modest reduction in total surface GABA A R expression and conferred a greater reduction of ␣3␤2␥2 HA than ␣1␤2␥2 HA GABA A Rs. This would result in an alteration of the composition of surface GABA A Rs to those with a greater fraction of ␣1 than ␣3 subunit-containing receptors.
AD Mutation Reduced ␣1(AD) Subunit Expression in Neurons-The previous experiments determined the effect of ␣1(AD) subunits on GABA A Rs overexpressed in heterologous cells. Here, we determined the effect of ␣1(AD) subunits on endogenous GABA A Rs in cultured neurons. To enable the nondestructive identification of transfected neurons without the use of large fluorescent fusion proteins, we inserted our cDNAs of interest into a commercial plasmid upstream of an internal ribosome entry site, which was followed by the cDNA encoding the fluorescent coral protein, ZsGreen1. We made five constructs by inserting either nothing (control), ␣1 h , ␣1(AD) h , ␣1 HA , or ␣1(AD) HA subunits. Because these constructs expressed the cDNA of interest and the fluorescent protein as separate peptides, the fluorescent protein did not interfere with ␣1 subunit trafficking, post-translational modification, or function.
We obtained cortical neurons from E18 rats and established neuron-enriched cultures. We transfected neurons at DIV10 and analyzed them at DIV17, the time point at which the cultured neurons develop mature GABAergic synapses. We first characterized the composition of our neuronal cultures at DIV17. We measured the number of DAPI-positive nuclei that also stained with the neuron-specific marker NeuN. We found that 89 Ϯ 5% of the cells were neurons (n ϭ 3) and that the neuronal transfection efficiency was 4 Ϯ 0.4% (data not shown). In these characterization experiments, no non-neuronal cells were transfected. However, in subsequent experiments, we did identify cells rarely with a non-neuronal morphology that expressed ZsGreen1; these non-neuronal cells were not included in the analyses.
We next determined the effect of the AD mutation on expression of ␣1(AD) subunits in neurons. We transfected FIGURE 7. Effect of the ␣1(AD) subunit on wild type ␣3␤2␥2 GABA A R. We transfected HEK293T cells with empty vector (negative control, shaded histogram) or 0.250 g of ␣3, ␤2, and ␥2S subunits without any additional subunit (wild type) or with varying masses (0.125-2 g) of mutant ␣1(AD) or wild type (wt) ␣1 subunit cDNA. We quantified the surface (A and B) ␣3 subunit expression using flow cytometry. A, we presented a sample flow cytometry histogram with wild type fluorescence as a solid line and wild type ϩ 2 g of ␣1(AD) fluorescence as a dashed line. B, we quantified the surface ␣3 subunit fluorescence in the presence of the different amounts of ␣1(AD) (f, n Ն 4) or wild type ␣1 (Ⅺ, n ϭ 3) subunit cDNA and normalized the ␣3 subunit fluorescence to that of wild type cells. Increasing the mass of ␣1(AD) and ␣1 subunit cDNA reduced surface ␣3 subunit expression. We also determined the effect of 2 g of ␣1(AD) subunit cDNA co-transfection on surface ␣3␤2␥2 receptor expression using biotinylation assays with Western blot (C and D). We stained the Western blots with the anti-␣3 subunit antibody and quantified surface ␣3 subunit relative to that of the Na ϩ /K ϩ -ATPase ␣ subunit (loading control). The presence of the ␣1(AD) subunit reduced surface expression of ␣3␤2␥2 receptors by 47 Ϯ 15% (n ϭ 6; p ϭ 0.024). *, p Ͻ 0.05. neurons with control-, ␣1 h -, ␣1(AD) h -, ␣1 HA -, or ␣1(AD) HA -IRES-ZsGreen1 constructs. We performed Western blots on neuronal lysates and stained them with either anti-␣1 h or anti-HA antibodies. Visual inspection demonstrated that the AD mutation substantially reduced both ␣1(AD) h and ␣1(AD) HA subunit expression (Fig. 9, A and B). Because of weak signal, we could not quantify ␣1(AD) h subunit expression on Western blot. Quantification of ␣1 HA and ␣1(AD) HA subunit expression revealed that the AD mutation reduced ␣1(AD) HA subunit expression by 39 Ϯ 9%, a result consistent with what we and others reported for epitope-tagged ␣1(AD) subunits expressed in HEK293T cells (10,11) and a much lower level of reduction than we typically observe using untagged ␣1 subunits (Fig. 1).
We used immunofluorescence studies to determine the effect of the mutation on untagged human ␣1(AD) h subunit expression in neurons. We transfected neurons with control-, ␣1 h -, or ␣1(AD) h -IRES-ZsGreen1 plasmids and analyzed total expression on DIV17. We fixed, permeabilized, and stained the neurons with anti-␣1 h subunit antibody to quantify total recombinant ␣1 h and ␣1(AD) h subunit expression. We identified transfected neurons by ZsGreen1 fluorescence (Fig. 9, C and D). Compared with wild type ␣1 h subunit-transfected neurons, there was no significant difference in ZsGreen1 fluorescence in control (103 Ϯ 4%) or ␣1(AD) h subunit (99 Ϯ 12%)transfected neurons.
The anti-␣1 h antibody specifically labeled recombinant ␣1 h subunit protein. Neurons transfected with control vector (n ϭ 16) possessed 2 Ϯ 1% of the fluorescence of those transfected with ␣1 h subunits (data not shown). Both ␣1 h and ␣1(AD) h subunits were expressed diffusely throughout the cytoplasm and dendrites and did not form any definite inclusions (Fig. 9, E and F). The intensity of ZsGreen1 fluorescence correlated with the intensity of both ␣1 h and ␣1(AD) h subunit immunofluorescence (Fig. 9G). The AD mutation reduced ␣1(AD) h subunit expression to 21 Ϯ 9% that of the ␣1 h subunit (Fig. 9H).
data not shown). The ␣1 and ␣1(AD) subunit conditions did not significantly differ from each other.
In contrast to the very modest effects of ␣1 subunit transfection on inter-mIPSC intervals, both ␣1 and ␣1(AD) subunits substantially altered mIPSC current kinetics, a result consistent with altered GABA A R composition. Compared with neurons transfected with control vector, neurons transfected with either ␣1 or ␣1(AD) subunits possessed reduced mIPSC rise (data not shown) and decay times (Fig. 10, B and D, p Ͻ 0.001). Moreover, transfection of the ␣1 subunit reduced mIPSC decay time to a greater extent than the ␣1(AD) subunit (p Ͻ 0.001). The change in the GABA A R mIPSC kinetics suggested a shift in receptor composition from non-␣1 to ␣1 subunit-containing receptors, a change that could result if overexpressed wild type ␣1 subunits replaced non-␣1 subunit-containing GABA A Rs in synapses and if overexpressed mutant ␣1(AD) subunits prevented surface trafficking of non-␣1 subunit-containing GABA A Rs to a greater extent than ␣1 subunit-containing GABA A Rs. GABA-evoked currents from recombinant GABA A Rs expressed in HEK293T cells revealed that of the four ␣ subunit isoforms found in synaptic GABA A Rs in the cerebrum (␣1-3 and ␣5), only GABA A Rs expressing the ␣3 subunits possessed slower deactivation rates than ␣1 subunit-containing GABA A Rs (32). Therefore, it is likely that ␣1(AD) subunits selectively inhibited expression of ␣3 subunit-containing GABA A Rs.
In addition to altering mIPSC kinetics, the ␣1 and ␣1(AD) subunits also reduced mIPSC peak amplitudes (Fig. 10, C and D, p Ͻ 0.001). The effect of ␣1(AD) subunits on mIPSC peak amplitudes was consistent with its reduction of wild type GABA A R surface expression (Figs. 4, 7, and 8). The effect of ␣1 subunits on peak mIPSC amplitudes was unexpected because studies in heterologous cells demonstrated that ␣1␤2␥2 and ␣3␤2␥2 GABA A Rs possess similar peak current amplitudes. Different possible mechanisms could explain this unexpected finding. First, the exact composition of endogenous synaptic GABA A Rs is unknown. Although it is assumed that endogenous GABA A Rs at DIV17 consist of a mixture of ␣1␤2␥2 and ␣3␤2␥2 receptors, it is possible that endogenous GABA A Rs have a composition that demonstrates higher peak current amplitudes than the ␣1 subunitcontaining GABA A Rs that replace them. Second, it is also possible that overexpression of wild type ␣1 subunits could reduce mIPSC peak amplitudes by increasing the tonic, persistently active, currents of the neurons, which shunt synaptic currents (33). Although tonic currents are mediated by ␦ subunit-containing GABA A Rs that typically co-assemble with either ␣4 or ␣6 subunits (34,35), recent results demonstrated that they can also co-assemble with ␣1 subunits (36).

DISCUSSION
We and others reported that the AD mutation caused a loss of ␣1(AD) subunit expression and function (8, 10 -13, 21, 26). Although the heterozygous loss of ␣1 subunits causes neuronal hyperexcitability, ␣1 subunit haploinsufficiency alone does not cause the full ADJME phenotype (14,15,28,37). The main finding of this study is that residual, nondegraded ␣1(AD) subunits conferred a dominant effect that reduced GABA A R expression more than would result from haploinsufficiency alone. Unexpectedly, the ␣1(AD) subunit reduced expression of ␣3␤2␥2 receptors to a greater extent than ␣1␤2␥2 receptors. Although these effects were small, they likely contribute to the cortical excitability produced by the heterozygous reduction in functional ␣1 subunit expression and thus may shape the epilepsy phenotype.
␣1(AD) Subunit Reduced ␣1␤2␥2 Expression by Entrapping Wild Type GABA A Receptor Subunits within the ER-The disruption of wild type gene function by a mutant gene product can be accomplished by several mechanisms. A mutation may cause a nonspecific effect if it produces a toxic gene product that enhances cell death (25,38) or represses protein synthesis through the unfolded protein response (39,40). In addition, mutations in multimeric ion channels such as GABA A Rs can cause specific effects by associating with wild type subunits and either forming dysfunctional ion channels on the cell surface, causing increased endocytosis, or preventing surface trafficking of the wild type protein complex.
Our study demonstrated that despite being grossly misfolded (10), the ␣1(AD) protein did not decrease HEK293T cell viability. In addition, overexpression of the ␣1(AD) subunit did not cause visible neuronal injury or death, alter neuronal physiology, or change neuronal membrane properties. In addition, the ␣1(AD) subunit did not nonspecifically reduce expression of transmembrane proteins unrelated to GABA A Rs (Fig. 4). Our experiments also showed that even with substantial overexpression, the ␣1(AD) subunit did not produce a dominant negative effect by substituting for wild type GABA A R on the cell surface (Fig. 2) or increasing dynamin-mediated endocytosis of wild type ␣1␤2␥2 receptors (Fig. 5).
The ␣1(AD) subunit did cause a dominant negative effect by producing specific, concentration-dependent reductions in expression of surface wild type GABA A Rs (Figs. 4, 7, and 8). Because previous studies demonstrated that essentially all residual ␣1(AD) subunits were localized to the ER (21,26), and because our experiments (Fig.  6) as well as those of Bradley et al. (11) showed that ␣1(AD) subunits specifically associated with wild type GABA A R subunits, we conclude that the ␣1(AD) subunit reduced expression of surface wild type GABA A R subunits by entrapping them within the ER. Moreover, our data provided direct evidence that the ␣1(AD) subunit reduced total as well as surface GABA A R expression (Fig. 4B), a result consistent with degradation of entrapped GABA A R subunits via ER-associated degradation (Fig. 11). In five separate experiments, we transfected DIV10 cultured cortical neurons with either control-IRES-ZsGreen1 (control), ␣1-IRES-ZsGreen1 (wild type, ␣1), or ␣1(AD)-IRES-ZsGreen1 (␣1(AD)). On DIV17, we identified transfected neurons by ZsGreen1 fluorescence and recorded Ͼ100 mIPSCs events from each. A, we showed sample mIPSC current traces. Transfection of both the ␣1 and ␣1(AD) subunit altered mIPSC current kinetics (B and D). B, we plotted a histogram that depicts the mIPSC decay time constants on the abscissa and the cumulative probability (prob) of an mIPSC having that decay time on the ordinate. Neurons transfected with control vector (solid line) had the biggest decay time constants followed by those transfected with ␣1(AD) (dotted line) and ␣1 (dashed line, p Ͻ 0.001). In the inset, we plotted the median mIPSC decay times. C, we plotted a histogram depicting mIPSC peak amplitudes on the abscissa and cumulative probability of a mIPSC having that peak amplitude on the ordinate. Neurons transfected with control subunit possessed larger mIPSC amplitudes than ␣1(AD)-transfected neurons, which possessed larger amplitudes than those transfected with ␣1 subunit (p Ͻ 0.001). The median mIPSC amplitudes are plotted in the inset. D, we displayed averaged (Ͼ100 events) mIPSC traces for wild type (black) and ␣1(AD)-transfected neurons (gray) demonstrating both the accelerated decay and reduced mIPSC amplitudes in neurons transfected with the ␣1(AD) subunit. *, p Ͻ 0.05. At maturity, wild type neurons (wt, A and D) predominantly express ␣1␤2␥2 receptors (red) but also express non-␣1-containing GABA A R (depicted here as ␣3 GABA A R, green), and thus we depicted a 6:2 ratio of ␣1␤2␥2 to ␣3␤2␥2 receptors. In a purely haploinsufficient model (B and E), all the misfolded ␣1(AD) subunit (blue) is degraded without it interacting with wild type GABA A R. Therefore, haploinsufficient neurons express reduced ␣1␤2␥2 receptors and unchanged non-␣1-containing GABA A R. The time course of the IPSC current kinetics would result from the currents contributed from ␣1 and ␣3 subunit containing receptors weighted by their abundance on the neuron surface. Therefore, as depicted here, the ␣1/␣3 weighting would decrease from 6/2 in wild type neurons to 3/2 in ␣1 haploinsufficient neurons thus producing IPSCs that more resembled ␣3 subunit containing GABA A R in the haploinsufficient neurons compared with wild type neurons. Our data are more consistent with a haploinsufficient plus dominant negative model (C and F). In this model, neurons degrade some ␣1(AD) subunit but also retain some misfolded ␣1(AD) subunit (blue) in the ER, which can associate with and retain a greater fraction of ␣3␤2␥2 than ␣1␤2␥2 receptors. This is shown here as ␣1(AD) retaining 1 ⁄3 ␣1 subunit containing receptors and 1 ⁄2 ␣3 subunit containing receptors. The retention of wild type GABA A R expression in the haploinsufficient plus dominant negative case reduces IPSC peak amplitudes to a greater extent than in the ␣1 haploinsufficient neurons. In addition, the selective retention of ␣3 subunit containing receptors increases the ␣1/␣3 weighting from 3/2 to 2/1 thus producing IPSC current kinetics that are more similar to ␣1 subunit containing GABA A R than in the ␣1 haploinsufficient condition. ␣1(AD) Subunit Reduced ␣3 Subunit Expression to a Greater Extent than ␣1 Subunit Expression-Homozygous Gabra1 knock-out mice increase the protein expression of other ␣ subunits, including the ␣3 subunit (28 -30). Because increased synaptic expression of ␣3 subunits compensates for absent ␣1 subunits, we tested if ␣1(AD) subunits also reduced expression of ␣3 subunit-containing GABA A Rs. Surprisingly, we found that ␣1(AD) subunits reduced the surface expression of ␣3␤2␥2 receptors to a greater extent than ␣1␤2␥2 receptors in HEK293T cells (Figs. 7 and 8). We also found that expression of ␣1(AD) subunits in cultured cortical neurons altered the time course of endogenous GABAergic mIPSCs, a result that suggested that non-␣1 subunit-containing GABA A Rs were replaced by ␣1 subunit-containing GABA A Rs (Fig. 10). In ADJME patients, the extent to which ␣1(AD) subunits would preferentially inhibit expression of ␣3 subunit-containing receptors would depend on the amount of ␣1(AD) subunits expressed in endogenous conditions. Different mechanisms could explain why ␣1(AD) subunits reduced ␣3␤2␥2 receptors to a greater degree than ␣1␤2␥2 receptors. A simple explanation would hold that ␣1(AD) subunits associate with ␣3 subunits with higher affinity than with ␣1 subunits and thus entrap a greater percentage of ␣3 subunits in the ER. Although functional GABA A Rs that traffic to the cell surface require a defined subunit stoichiometry and assembly order without two adjacent ␣ subunits (␣-␤-␣-␥-␤) (4 -6, 41), it is possible that nonfunctional ␣1/␣3 oligomers could form, be sequestered in the ER, and undergo ER-associated degradation.
␣1(AD) subunits could also selectively reduce expression of surface ␣3␤2␥2 receptors by sequestering ␤2 and ␥2 subunits. If ␣1 and ␣1(AD) subunits have the same affinity for ␤2/␥2 subunits, and ␣3 subunits have a lower affinity for ␤2/␥2 subunits, the presence of ␣1(AD) subunits would deprive ␣3 subunits of their necessary assembly partners to a greater extent than ␣1 subunits. Previous studies identified domains within the extracellular region of ␣1 subunits that are responsible for its oligomerization with ␤2 and ␥2 subunits (42)(43)(44). The ␤2 subunit binding domains in the ␣1 and ␣3 subunits are identical except for a single conserved substitution. However, the ␥2 subunit binding domains have four differences, including nonconserved substitutions of a threonine to a lysine and an arginine to a proline. It would be of interest in future studies to determine whether ␥2 subunits oligomerize with ␣3 subunits with lower affinity than with ␣1 subunits and if either of these two amino acids explains the difference.
Differences in oligomerization affinities between the ␣1 and ␣3 subunits with ␤2 or ␥2 subunits would explain the differential effects of the ␣1(AD) subunit only if ␤2 or ␥2 subunit expression was not in excess. However, our data in Fig. 3 demonstrated that at a fixed concentration of ␤2 and ␥2 subunits, a 4-fold increase in ␣1 subunit cDNA produced corresponding increases in ␣1␤2␥2 subunit expression, evidence that ␤2 and ␥2 subunits were in excess. A mechanism that could explain a greater effect of ␣1(AD) subunits on ␣3␤2␥2 receptors than ␣1␤2␥2 receptors in conditions of excess ␤2 and ␥2 subunits would be that ␣3␤2␥2 receptors traffic to the cell surface at slower rates than ␣1␤2␥2 receptors. In this case, ␣3␤2␥2 receptors would be retained in the ER longer than ␣1␤2␥2 receptors, which would provide them with more opportunity to incorporate an ␣1(AD) subunit and be targeted for degradation.
Although our experiments focused on determining the effect of ␣1(AD) subunits on ␣1␤2␥2 and ␣3␤2␥2 receptors, interactions of ␣1(AD) subunits with other GABA A R subunits may make even more important contributions to the epilepsy phenotype. Future studies will be needed to determine the effects of ␣1(AD) subunits on expression of other GABA A R isoforms.
Model of the Dominant Effect-We depicted our model for the effect of ␣1(AD) subunits on ␣1␤2␥2 and ␣3␤2␥2 receptors in Fig. 11. In this model, wild type neurons express both ␣1␤2␥2 and ␣3␤2␥2 receptors in the ER (Fig. 11A) and on the neuron surface (Fig. 11D). Because mature cortical neurons express more ␣1 than ␣3 subunits (1-3, 45), ␣1␤2␥2 and ␣3␤2␥2 receptors are shown here in a 6:2 ratio. The amplitude and time course of GABAergic currents would result from a weighted average of ␣1␤2␥2 and ␣3␤2␥2 currents. In a haploinsufficient model, the ␣1(AD) subunits are degraded and do not associate with wild type receptors. Therefore, in a completely haploinsufficient case, neurons express reduced ␣1␤2␥2 receptors but the same amount of ␣3␤2␥2 receptors in the ER (Fig. 11B) and the neuron surface (Fig. 11E). Therefore, in a haploinsufficient model, the reduction in ␣1␤2␥2 but not ␣3␤2␥2 expression would result in a greater proportion of ␣3␤2␥2 receptors on the surface than found in wild type neurons (shown here as a 3:2 ratio). Therefore, the time course of GABAergic currents, as a weighted sum of the ␣1␤2␥2 and ␣3␤2␥2 receptors, would be more similar to that of ␣3␤2␥2 receptors in haploinsufficient than wild type neurons. Because there would be fewer GABA A R on the cell surface, inhibitory postsynaptic current peak amplitudes would be smaller in haploinsufficient than wild type neurons.
Rather than a haploinsufficient model, our data presented here suggests a haploinsufficient plus dominant negative model (Fig. 11, C and F). In this model, neurons express ␣1(AD) subunit in the ER. Some ␣1(AD) subunits degrade, but some ␣1(AD) subunits oligomerize with ␣1 and ␣3 subunit-containing GABA A Rs (Fig. 11C). Retained ␣1(AD) subunits reduce a greater fraction of ␣3␤2␥2 receptors than ␣1␤2␥2 receptors, and thus the ratio of ␣1 to ␣3 subunit-containing receptors is shown here as a 2:1 ratio (Fig. 11F). The time course of the GABAergic currents is thus more similar to that of ␣1␤2␥2 receptors in a haploinsufficient plus dominant negative model than in the pure haploinsufficient model. Because both ␣1␤2␥2 and ␣3␤2␥2 receptor expression is reduced, peak current amplitudes are reduced relative to wild type and haploinsufficient currents.
Implications of Dominant Negative Effects for ADJME and Other Dominant Epilepsy Syndromes-Although the amount of endogenous ␣1(AD) subunit expressed in neurons of ADJME patients is unknown, our data demonstrated that neurons degrade recombinant ␣1(AD) subunit with high efficiency (79%, Fig. 9). Therefore, assuming that the endogenous ␣1(AD) subunit in ADJME patients also expresses at 21% of wild type ␣1 subunit and that the ␣1(AD) subunit in neurons cause similar sigmoidal reductions in surface wild type ␣1 and ␣3 subunit expression as in HEK293T cells (Figs. 4, 7, and 8), we expect that heterozygous ␣1(AD) subunit expression in ADJME patients would reduce surface ␣1 and ␣3 subunit expression by an additional 4 and 5%, respectively, more than hemizygous ␣1 subunit expression.
Could such a seemingly small additional reduction in surface ␣1 and ␣3 subunit expression alter the epilepsy phenotype? Both empirical and theoretical systems do suggest that seizures are threshold events and that even small changes in GABA A R availability can transition an organism into a seizure-prone state. For example, even though the GABA A R antagonists, bicuculline and pentylenetetrazole, inhibit GABA-evoked currents with typical sigmoidal concentration-response curves (46,47), they produce seizures in experimental animals at specific doses with very small variances (48,49). These findings suggest that a specific threshold of functional GABA A Rs is required to prevent an organism from experiencing a seizure. Moreover, recent computational models of neural networks predicted that very small changes in ion channel function can produce seizure-like activities (50).
Small changes in GABA A R expression and composition may shape the phenotypes of other autosomal dominant GABA A R subunit mutations that are associated with epilepsy. The best studied mutation is the ␥2(R43Q) mutation associated with childhood absence epilepsy and febrile seizures (51). The ␥2(R43Q) subunit produces smaller changes in GABA A R expression and composition than the ␣1(AD) subunit. Although the R43Q mutation does not reduce the total expression of the ␥2(R43Q) subunit, it disrupts its ability to directly oligomerize with ␤2 subunits (52). The inability of ␥2(R43Q) subunits to oligomerize may explain why they do not produce a dominant negative effect on ␣1, ␣3, or ␤2 subunit expression or alter mIPSC amplitude or current kinetics (17,18). The ␥2(R43Q) subunit does reduce surface trafficking of the ␣5 subunit-containing GABA A Rs and thereby reduces tonic GABA A R currents (17). Although the effects of the ␥2(R43Q) subunit are small, they do shape the epilepsy phenotype because the heterozygous ␥2(R43Q) subunit knock-in mice but not heterozygous ␥2 subunit knock-out mice experience seizures (16,19,53).
Therefore, we think that the small dominant negative effects of the ␣1(AD) subunit could confer an epilepsy phenotype that is different from the one conferred by ␣1 subunit haploinsufficiency alone. This hypothesis needs to be directly tested by comparing heterozygous ␣1 knock-out with ␣1(AD) knock-in genetically modified mice.