Endoplasmic Reticulum Retention and Associated Degradation of a GABAA Receptor Epilepsy Mutation That Inserts an Aspartate in the M3 Transmembrane Segment of the α1 Subunit*

A GABAA receptor α1 subunit epilepsy mutation (α1(A322D)) introduces a negatively charged aspartate residue into the hydrophobic M3 transmembrane domain of the α1 subunit. We reported previously that heterologous expression of α1(A322D)β2γ2 receptors in mammalian cells resulted in reduced total and surface α1 subunit protein. Here we demonstrate the mechanism of this reduction. Total α1(A322D) subunit protein was reduced relative to wild type protein by a similar amount when expressed alone (86 ± 6%) or when coexpressed with β2 and γ2S subunits (78 ± 6%), indicating an expression reduction prior to subunit oligomerization. In α1β2γ2S receptors, endoglycosidase H deglycosylated only 26 ± 5% of α1 subunits, consistent with substantial protein maturation, but in α1(A322D)β2γ2S receptors, endoglycosidase H deglycosylated 91 ± 4% of α1(A322D) subunits, consistent with failure of protein maturation. To determine the cellular localization of wild type and mutant subunits, the α1 subunit was tagged with yellow (α1-YFP) or cyan (α1-CFP) fluorescent protein. Confocal microscopic imaging demonstrated that 36 ± 4% of α1-YFPβ2γ2 but only 5 ± 1% α1(A322D)-YFPβ2γ2 colocalized with the plasma membrane, whereas the majority of the remaining receptors colocalized with the endoplasmic reticulum (55 ± 4% α1-YFPβ2γ2S, 86 ± 3% α1(A322D)-YFP). Heterozygous expression of α1-CFPβ2γ2S and α1(A322D)-YFPβ2γ2S or α1-YFPβ2γ2S and α1(A322D)-CFPβ2γ2S receptors showed that membrane GABAA receptors contained primarily wild type α1 subunits. These data demonstrate that the A322D mutation reduces α1 subunit expression after translation, but before assembly, resulting in endoplasmic reticulum-associated degradation and membrane α1 subunits that are almost exclusively wild type subunits.

GABA A 2 receptors are pentameric ligand-gated chloride ion channels that are the major inhibitory neurotransmitter receptors in the mammalian central nervous system (1). The five subunits arise from seven subunit families that contain multiple subtypes and assemble in a limited number of subunit combinations, with the most prevalent consisting of two ␣1 subunits, two ␤2 subunits, and one ␥2 subunit (2)(3)(4)(5). Each subunit contains four hydrophobic segments (M1-M4) that are homologous to the four membrane spanning helices of the Torpedo marmorata nicotinic acetylcholine receptor (AChR) subunits whose three-dimensional structure has been determined to 4 Å (6).
A nonconserved missense mutation in the GABA A receptor ␣1 subunit gene (GABRA1, ␣1(A322D)) that codes for an aspartate in place of an alanine at position 7 of the M3 transmembrane segment is present in a form of autosomal dominant juvenile myoclonic epilepsy (7), an idiopathic generalized epilepsy syndrome that accounts for ϳ10% of all cases of epilepsy (8). When expressed in heterologous cells, this mutation affects both the function and expression of GABA A receptors. Expression of the ␣1(A322D) subunit with ␤2 and ␥2 subunits ("homozygous expression") reduced peak currents by ϳ90%, substantially altered whole cell current kinetics, and reduced mean single channel open times (7,9,10). We recently reported that the ␣1(A322D) mutation reduced ␣1 subunit expression by 94%, and that it produced asymmetrical, subunit position-dependent reduction of heterozygous receptor currents and ␣1 subunit protein expression (10). Heterozygous receptors constructed from concatamers with the ␣1(A322D) subunit positioned between two ␤2 subunits had 35% of peak current amplitudes and 70% of protein expression relative to wild type receptors, whereas heterozygous receptors with ␣1(A322D) positioned between ␤2 and ␥2 subunits had 1% of peak current amplitude and 51% of protein expression of wild type receptors. To our knowledge, this is the first naturally occurring missense mutation that reduces expression of a ligand-gated ion channel subunit.
Because ␣1(A322D) substantially reduced the amount of ␣1 subunit, it is likely that it is this reduction in total ␣1 subunit expression, and not the alteration of GABA A receptor current kinetics, that is the predominant mechanism by which this mutation causes disinhibition and epilepsy. Here we determined the mechanism by which this single missense mutation reduced ␣1 subunit expression.

MATERIALS AND METHODS
Expression of recombinant GABA A Receptors-pcDNA3.1 plasmids containing cDNAs that encode human ␣1, ␤2S, and ␥2S GABA A receptor subunits were a gift from Dr. Mathew Jones (University of Wisconsin, Madison, WI). ␣1-YFP and ␣-CFP cDNAs were constructed by first inserting HpaI and SacII restriction sites between the codons encoding amino acids four and five of the mature ␣1 subunit. DNA encoding the fluorescent protein (FP) from the corresponding ␥2S-FP subunit (11) was removed by HpaI and SacII digestion and then ligated into the ␣1 subunit-containing plasmid between the codons encoding amino acids four and five of the mature subunit. Cycle 3 GFP-tagged ␣1 subunit was constructed by performing a blunt ligation of the ␣1 subunit cDNA into cycle 3 GFP-containing pcDNA3.1 plasmid (Invitrogen). The ␣1(A322D) mutation was made using the QuikChange site-directed mutagenesis kit (Stratagene). All cDNA sequences were confirmed by DNA sequencing.
Transfections were similar for the electrophysiology experiments but also included 1 g of pHOOK (Invitrogen), which was used for immunomagnetic separation (12). Twenty hours after transfection, cells were trypsinized, centrifuged (400 ϫ g), and incubated with hapten-coated magnetic beads for 30 min at 37°C. The cells that bound the magnetic beads were isolated via a magnetic stand and plated on 35-mm culture dishes.
For the confocal microscopy experiments, cells were transfected in 35-mm dishes using equal amounts (1 g) of ␣1-FP, ␤2S, and ␥2S subunit cDNAs. For the ER colocalization experiments, the cells were also transfected with 30 ng of a cyan fluorescent protein-tagged endoplasmic reticulum marker (CFP-ER, BD Biosciences).
Immunoblots-The Western blot protocol has been described (10). Transfected cells were lysed in modified radioimmunoassay solution (RIPA, 20 mM Tris, pH 7.4, 1% Triton X-100, 150 mM NaCl, 0.25% deoxycholate) that contained one pellet of Complete Mini TM protease inhibitor (Roche Diagnostics) per 10 ml. The lysates were centrifuged at 10,000 ϫ g for 30 min. Lysates were fractionated by SDS-PAGE at the acrylamide concentrations given in the figure legends. After SDS-PAGE, the proteins were electrotransfered to polyvinylidene fluoride membranes (Millipore Inc.). All primary antibodies were monoclonal and were purchased from Chemicon Inc. (Temecula, CA). In experiments in which the cells were transfected with untagged ␣1 and ␣1(A322D) subunits, the membranes were first incubated with an antibody against the ␣1 subunit (5 g/ml, clone BD24) in addition to an antibody against glyceraldehyde-3-phosphate dehydrogenase (GAPDH, 0.1 g/ml, clone 6C5), which was used to control for the amount of protein loaded on the gel. For experiments in which the cells were transfected with FP-tagged ␣1 subunit, the membranes were incubated with a monoclonal antibody to green fluorescent protein (1:2500) as well as to an antibody to ␤-actin (clone C4, 1 g/ml). After incubation with the primary antibody, all immunoblots were incubated with a horseradish peroxidase-coupled goat anti-mouse secondary antibody (Jackson Immunoresearch Laboratories, 1:6000 dilution) and then visualized with a chemiluminescent detection system (Amersham Biosciences) using a digital imager (Alpha Innotech, San Leandro, CA). The integrated density volume (IDV, pixel intensity ϫ mm 2 ) of each band was calculated using the Alpha Innotech software. Background IDVs were obtained from regions adjacent to the bands of interest and subtracted from the total IDV. All protein bands were normalized to the loading control (GAPDH or actin).
Endoglycosidase Digestion-Protein lysates were prepared as described above and their protein concentrations were measured using the Micro BCA Protein Assay TM (Pierce). Endoglycosidase H (endo-H) and peptide N-glycosidase-F (PNGaseF) were obtained from Sigma. Endo-H and PNGaseF digestion of nicotinic AChR ␦-subunit has been described (13). Both endoglycosidase digestions were performed for 3 h at 37°C. Endo-H digestions were performed in 50 mM sodium citrate, pH 5.5, 1% Triton X-100, 0.1% SDS, 50 mM ␤-mercaptoethanol with 0.2 units/ml endo-H. PNGaseF digestions were performed in 50 mM Tris-HCl, pH 7.4, 1% Triton X-100, 0.1% SDS, 50 mM ␤-mercaptoethanol with 0.2 units/ml PNGaseF. The reactions were terminated by addition of Laemmli sample buffer, and the reaction products were detected by SDS-PAGE and Western blot as described above.
Confocal Microscopy and Image Analysis-Approximately 24 h before the confocal microscopy experiments, the transfected cells were plated in collagen-coated 35-mm glass-bottom dishes (MatTek, Ashland, MA). The dishes were then coded, and thus the microscopy experiments were performed in a single-blinded fashion. Immediately before imaging, N-(3-triethylammoniumpropyl)-4-(6-(4-(diethylamino)phenyl)hexatrienyl)pyridinium dibromide (FM 4-64, 0.3 g/ml, Invitrogen) was added to the culture media to stain the cells' plasma membranes without staining the ER (14). Cells were chosen at random and were imaged using a Zeiss 500 META confocal microscope with a ϫ40, 1.3 numerical aperture Pan neofluoar objective. For all cells except those transfected with ␣1-CFP, the pinhole of all channels was adjusted so that the images were obtained as a single 2-m slice through the middle of the cell. To collect more emitted light from the cells that were transfected with ␣1-CFP, the pinhole of all channels was adjusted so that the images were obtained as a single 3-m slice through the middle of the cell. Excitation wavelengths were 455, 514, and 543 nm for CFP, YFP, and FM 4-64, respectively. For each excitation wavelength, the laser power was adjusted as necessary to utilize the full dynamic range of the detector for each respective fluorophore. CFP emission was detected with a 475-525-nm band pass filter, and FM 4-64 emission was detected using a 560-long pass filter. YFP emission was spectrally separated from CFP and FM 4-64 by reflecting the emitted light off a NFT545 dichroic mirror and then filtering it through a 530 -600-band pass filter. The digital images were obtained with 2.8 times scanning zoom and 8 bit, 512 ϫ 512 pixel resolution. All images presented in the figures are unprocessed.
The confocal image files were coded and processed in a single-blind fashion using the ImageJ software (National Institute of Health, Bethesda, MD). Background FP fluorescence for each image was obtained from cells that stained with FM 4-64 but did not express FP. Background CFP-ER fluorescence was defined as the CFP fluorescence in the nucleus. The background fluorescence was subtracted from the images at the 99th percentile. For cells cotransfected with CFP-ER, membrane YFP fluorescence was defined as the YFP outside the boundary of CFP-ER and FM 4-64 staining. For cells not cotransfected with CFP-ER, membrane FP fluorescence was defined as the FP outside the inner portion of FM 4-64 staining. ER YFP fluorescence was defined as intracellular YFP that colocalized with ER fluorescence. The integrated fluorescence from each subcellular region was calculated by summing all the background-subtracted fluorescence values from each pixel in that area.
Fluorescence Spectroscopy-Lysates of cells expressing ␤2 and ␥2S subunits and wild type, heterozygous, or homozygous cycle 3 GPPtagged ␣1 subunit (␣1-GFP) were prepared as described above. Cycle 3 GFP was used to tag the ␣1 subunit for fluorescence spectroscopy rather than enhanced fluorescence proteins (CFP or YFP) because of its high fluorescence efficiency (15); it was not used for microscopy because its absorbance maximum is in the ultraviolet region of the spectrum. Aliquots (200 l) of the lysates were placed in microtiter plates, and the fluorescence was determined in a Flexstation TM fluorescence spectrometer (Molecular Devices, Sunnyvale, CA) with an excitation of 395 nm and emission of 507 nm. Background fluorescence was defined as the fluorescence from lysates from untransfected cells; the background was subtracted from the fluorescence of the experimental lysates.
Electrophysiology-Electrophysiological recordings were performed 48 h after transfection. The intrapipette solution contained (in mM): 153 KCl, 1 MgCl 2 , 5 EGTA, 10 HEPES, and 2 MgATP (pH 7.3, osmolarity ϭ 305-310 mOsm). The external recording solution consisted of (in mM): 142 NaCl, 8 KCl, 6 MgCl 2 , 1 CaCl 2 , 10 glucose, 10 HEPES (pH 7.4, osmolarity ϭ 319 -325 mOsm). Recording pipettes were pulled on a Sutter P-2000 micropipette electrode puller (Sutter Instrument Co., San Rafael, CA) from borosilicate capillary glass (Fisher). GABA was applied to the cells via gravity using a system consisting of pulled four-barrel square glass (200 -300 m) connected to a Perfusion Fast-Step (Warner Instrument Corp., Hamden, CT). The solution exchange time was determined by stepping a 10% dilute external solution across the open electrode tip to measure a liquid junction current; the 10 -90% rise times for solution exchange were consistently Յ0.4 ms. GABA A receptor currents were recorded in voltage clamp mode with cells clamped at Ϫ50 mV using a lifted whole cell patch-clamp technique (16) using an Axon 200B amplifier (Molecular Devices). The signals were sampled at 10 kHz and written to a computer hard drive.
Homology Modeling and Secondary Structure Prediction-Multiple amino acid sequences were aligned using the ClustalW software (www.ebi.ac.uk/clustalw/). We identified amino acids of the M3 segments of the Torpedo marmorata AChR subunits that were homologous to the Ala-322 residue of the GABA A receptor ␣1 subunit in two steps. First, the following amino acid sequences were aligned in four groups: 1) Torpedo marmorata nAChR ␣, ␤, ␥2, and ␦ subunits; 2) human GABA A receptor ␣1-6 subunits; 3) human GABA A receptor ␤1-3 subunits; and 4) human ␥1-3 subunits. Next, the 16 sequences in these four groups were aligned together in conjunction with the human GABA A receptor ␦-subunit, and the results of the 17-sequence alignment were visually compared with the small group analyses to make certain the sequences that were most similar with one another remained in alignment.
Identification and scoring of transmembrane helices were performed using the relevant ExPASY Proteomics tools (us.expasy.org/tools/) topology prediction algorithms (DAS, TopPred, TMPred, and TMHMM). The predictions were performed first on the native AChR sequences and then on the sequences in which the amino acid in the M3 domain homologous to the GABA A ␣1 subunit Ala-322 was changed to an aspartate. The scores from the topology prediction programs, which predicted a transmembrane helix for the 25-amino acid segments beginning at ␣Tyr-277, ␤Tyr-283, ␥Tyr-291, and ␦Tyr-286 (corresponding to the M3 transmembrane helices in the Torpedo marmorata electron diffraction data) for the sequences with or without the aspartate substitution, were compared.
Data Analysis-Values are reported as mean Ϯ S.E. Statistical significance for the endoglycosidase, confocal, and fluorescence spectroscopy experiments were determined using the Student's unpaired t test and the significance of expression differences in the Western blot experiments and the effect of aspartate substitution on topology prediction scores were determined using the paired t test (GraphPad, San Diego, CA).
The ␣1 subunit has two sites of N-linked glycosylation that reside on its extracellular N terminus (17,25). Membrane proteins are N-linked glycosylated with high mannose carbohydrates co-translationally within the ER, but upon trafficking to the trans-Golgi the high mannose carbohydrates are replaced with low mannose carbohydrates (26). Digestion with endo-H removes high mannose N-linked carbohydrates, whereas digestion with PNGaseF removes all carbohydrates. Therefore, FIGURE 1. A and C, cells were transfected with equimolar ratios of ␣1, ␤2, and ␥2S subunits; or B and D, ␣1 subunit and a 2-fold excess of empty pcDNA3.1 plasmid that was used to transfect equivalent masses of DNA. A and B, Western blots on 10% SDS-PAGE gels from whole cell lysates were probed with anti-␣1 subunit antibody as well as an antibody against GAPDH, which was used to control for the amount of lysate protein loaded on the gel. C and D, the fraction of the ␣1 subunit band relative to the GAPDH band from each gel was quantified and averaged.
Undigested cell lysates and those digested with endo-H or PNGaseF were analyzed by Western blot (Fig. 2, A and B). Because the wild type ␣1 subunit expresses more efficiently than heterozygous and homozygous ␣1 subunits, protein was loaded in the ratio WT:heterozygous: homozygous, 8:15:50 g, to balance the amount of ␣1 subunit detected on the blot. Undigested wild type, heterozygous, and homozygous proteins ran at 50 kDa, and those digested with PNGaseF (removal of all N-linked carbohydrates) ran at 46 kDa, a reduction in molecular weight similar to PNGaseF digestion of Torpedo AChR ␦-subunit (13) and tunicamycin inhibition of GABA A receptor 9E10-tagged ␣1 subunit glycosylation (17). With endo-H digestion, wild type and heterozygous receptors ran in two bands at 46 (endo-H sensitive) and 48.4 kDa (endo-H resistant), whereas endo-H digestion products of homozygous receptors ran in a single band at 46 kDa (Fig. 2, C and D). The 1.6-kDa shift in the endo-H-resistant band is similar to endo-H digestion of the AChR ␦-subunit (13). Quantification of the 46-and 48.8-kDa digestion products demonstrated that 74 Ϯ 5% of wild type (n ϭ 6), 64 Ϯ 4% of heterozygous (n ϭ 6, p ϭ 0.13), and 9 Ϯ 4% (n ϭ 6, p Ͻ 0.0001) of homozygous receptors were endo-H resistant. This shows that essentially all the homozygous ␣1(A322D) subunits retained ER processing and provides evidence that the defect in ␣1(A322D) subunit trafficking occurs after translation. Given the results of endo-H digestion of wild type and homozygous ␣1 subunits, with independent processing of these subunits, 42, not 64%, of heterozygous ␣1 subunit should have been endo-H sensitive. This suggests that other processes (such as pro-tein degradation) may alter the relative ratios of wild type ␣1 and ␣1(A322D) subunits. Another alternative is that this assay lacks the resolution and statistical power to detect a difference in the extent of glycosylation between wild type and heterozygous receptors.
As a second method to determine the effect of ␣1(A322D) on the expression of the ␣1-YFP subunit, we performed Western blots using an anti-GFP antibody on whole cell lysates from cells transfected with wild type, heterozygous, and homozygous ␣1-YFP subunits with cotransfection of ␤2 and ␥2S subunits or an equivalent amount of empty pcDNA3.1 plasmid (Fig. 4B). For both wild type and heterozygous ␣1␤2␥2S receptors as well as ␣1-YFP subunits transfected alone, there  were two specific immunoreactive bands at 64 and 72 kDa (n ϭ 6). For homozygous ␣1(A322D)-YFP␤2␥2S receptors, there was only one specific immunoreactive band at 64 kDa.
These data demonstrate that similar to the non-YFP-tagged ␣1 subunits, the ␣1(A322D) mutation reduced expression of ␣1-FP subunits. It had been reported recently that Western blots of whole cell lysates from cells expressing GABA A receptors containing ␣1(A322D)-GFP subunits demonstrated the presence of the ␣1(A322D)-GFP subunit. From this, it was concluded that ␣1(A322D) does not preclude ␣1 subunit expression (30). Here we also show that ␣1(A322D)-YFP is expressed. However, our quantification of the Western blots demonstrated that its expression was reduced, a result confirmed by measurement of ␣1(A322D)-GFP fluorescence.
The addition of YFP to the N terminus of the ␣1 subunit did change some aspects of its processing. First, ␣1(A322D)-YFP expression relative to wild type expression (46 Ϯ 4%) was greater than that of the non-FP-tagged receptors (22 Ϯ 6%, Fig. 1). Second, on Western blot, ␣1-YFP ran as bands of two different molecular weights. One possibility is that the presence of the large YFP epitope tag on the N terminus interfered with N-glycosylation of the subunit.
Endo-H digestion products of wild type ␣1-YFP␤2␥2S receptors (n ϭ 4) ran as two bands with apparent molecular masses of 72 and 64 kDa, thus demonstrating that these receptors contained low mannose glycosylation and were thus processed in the trans-Golgi. In contrast, endo-H digestion of cell lysates transfected with ␣1-YFP subunits alone (n ϭ 4) ran as a single band at 64 kDa, indicating maturation arrest in the ER.
Interestingly, although non-YFP-tagged ␣1 and ␣1(A322D) subunits had substantially different fractions of endo-H-sensitive N-linked glycosylation, each was 100% glycosylated (Fig. 2). In contrast, only 85% of wild type ␣1-YFP and 15% of ␣1(A322D)-YFP subunits were glycosylated. This highlights the fact that even though fluorescently tagged ␣1 subunits form GABA A receptors with similar functional properties as native receptors (Fig. 3), and that the A322D mutation causes qualitatively similar reductions in ␣1-YFP subunits as with non-fluorescently tagged ␣1 subunits (Fig. 4), there do exist some differences in posttranslational glycosylation Assembled, but Not Unassembled, ␣1-YFP Subunits Localized to the Membrane-It has been shown that 9E10-tagged ␣1 subunit homomers do not traffic to the plasma membrane, but, instead, are sequestered in the ER (17) and degraded (24). Likewise, ␣1 subunit homomers tagged with GFP at the C terminus remain in the ER of oocytes (29). We tested the ability of ␣1-YFP subunits to traffic to cell membranes with and without coexpression of ␤2 and ␥2S subunits. Cells were transfected with CFP-ER, YFP-␣1 subunit, and either ␤2 and ␥2S subunits or an equivalent amount of pcDNA3.1. The cells were imaged by confocal microscopy (Fig. 6A). Although the ER extends to the plasma membrane in these cells, dual labeling of the ER and membrane compartments allowed reliable separation of the ER component from the membrane component. Thus, the fraction of ␣-YFP subunit colocalized to the membrane and ER could be quantified (Fig. 6B). For receptors transfected with ␣1-YFP subunit and empty pcDNA3.1, 80 Ϯ 3% of the ␣1-YFP subunit colocalized to the ER and 6.5 Ϯ 0.8% colocalized to the membrane (n ϭ 11). In contrast, for receptors transfected with ␣1-YFP, ␤2, and ␥2S subunits, 55 Ϯ 4% of the ␣1-YFP subunit colocalized in the ER (p Ͻ 0.001) and 36 Ϯ 4% colocalized with the plasma membrane (p Ͻ 0.001, n ϭ 13). Thus, assembled, but not unassembled, ␣1-YFP subunits were localized to the membrane.
Membrane ␣1-YFP Subunits with Heterozygous Expression Are Derived from Wild Type ␣1-YFP, but Not Mutant ␣1(A322D)-YFP, Subunits-Individuals with autosomal dominant juvenile myoclonic epilepsy are heterozygous for the ␣1(A322D) mutation (7), and thus it was of interest to determine the relative amounts of wild type ␣1 and mutant ␣1(A322D) subunit that were trafficked to the plasma membrane with heterozygous expression. We constructed CFP-tagged wild type (␣1-CFP) and mutant (␣1(A322D)-CFP) subunits. Cells that were heterozygous for the ␣1(A322D) subunit mutation were formed by cotransfecting either wild type ␣1-YFP and mutant ␣1(A322D)-CFP subunits or wild type ␣1-CFP and mutant ␣1(A322D)-YFP subunits. Visual analysis of confocal microscopic images demonstrated that with FIGURE 6. A, cells were transfected with ␣1-YFP and either ␤2 and ␥2S subunits or an equivalent mass of pcDNA3.1. Confocal microscopy demonstrated that for both transfections, ␣-YFP (R, green) receptors co-localized with the CFP-ER (ER, blue). However, only the receptors co-transfected with ␤2 and ␥2S subunits localized to the plasma membrane (M, red). Colocalization (Co) of ␣1-YFP and CFP-ER is cyan and colocalization of ␣-YFP and the membrane marker is yellow. B, the fraction of total ␣1-YFP/␤2/␥2 and ␣1-YFP/pcDNA that colocalized to the membrane or ER is graphed.  64 (M, red). The majority of wild type, heterozygous, and homozygous ␣1-YFP subunit colocalized (Co) with the ER (cyan). Only wild type and heterozygous ␣1-YFP subunit colocalized with the plasma membrane FM 4-64 marker (yellow). B, the fraction of total wild type, heterozygous, and homozygous ␣1-YFP that colocalized to the membrane or ER is graphed.
With both heterozygous ␣1-YFP␣1(A322D)-CFP␤2␥2S and ␣1-CFP␣1(A322D)-YFP␤2␥2S transfections, there was an apparently greater percentage of mutant ␣1(A322D)-FP subunit associated with the plasma membrane than in cells transfected with homozygous ␣1-YFP␤2␥2S receptors and CFP-ER (Fig. 7). In the cells transfected with ␣1-YFP␣1(A322D)-CFP␤2␥2S and ␣1-CFP␣1(A322D)-YFP␤2␥2S receptors, CFP-ER could not be used to label the ER. Thus, because the ER abuts the plasma membrane in these cells, a distinct boundary between ER and membrane could not be identified. Therefore, we thought it was likely that in cells not transfected with CFP-ER, a portion of ␣1-FP subunit that was deemed to be colocalized with the membrane was actually associated with the ER, and thus the measured percentage of ␣1-FP and ␣1(A322D)-FP subunit associated with the membrane was overestimated. To test this hypothesis, we analyzed cells transfected with ␣1(A322D)-YFP␤2␥2S receptors and CFP-ER, and when processing the images, we defined membrane-associated ␣1-YFP subunits by using only the inner leaflet of the FM 4-64 marker (as was done for cells transfected with ␣1-YFP␣1(A322D)-CFP␤2␥2S and ␣1-CFP␣1(A322D)-YFP␤2␥2S receptors) and not by using the distinct CFP-ER/FM 4-64 boundary as was done in Fig. 7. We found that without defining the CFP-ER/FM 4-64 boundary, the calculated amount of ␣1-YFP subunit colocalized to the membrane was 19 Ϯ 3%, a significantly greater value than was calculated when using the CFP-ER/FM 4-64 boundary (5.1 Ϯ 0.3%, n ϭ 5, p ϭ 0.002). Thus, the molar ratio of the membrane-associated ␣1-FP subunit to the ␣1(A322D)-FP subunit with heterozygous expression is likely larger than 5:1 as calculated above and may be as high as 19:1. This result also highlights the advantage of using, when possible, both membrane and ER markers in cells such as HEK293T cells, whose ER is diffusely distributed and thus abuts the membrane.
These results demonstrate that in the cells transfected with heterozygous ␣1 and ␣1(A322D) subunits, the majority of the mutant protein remained sequestered in the ER and only wild type ␣1 subunit formed GABA A receptors at the membrane surface, a result consistent with the nearly identical current kinetic profiles of wild type and heterozygous ␣1␤2␥2S GABA A receptors (10).

DISCUSSION
We previously demonstrated that the ␣1(A322D) mutation reduces ␣1 subunit expression in a subunit position-dependent manner. Here, we report that the trafficking of the mutant ␣1(A322D) subunit is altered between translation and oligomerization within the ER. Previously identified GABA A receptor epilepsy mutations altered the function (31,32) or cell surface trafficking (11,33,34) of GABA A receptor subunits. In contrast, the GABA A receptor ␣1(A322D) epilepsy mutation that codes for a simple substitution of a charged for an uncharged amino acid in the M3 transmembrane helix has a novel mechanism by reducing ␣1 subunit expression at a post-translational and pre-oligomerization trafficking step. Imaging of the wild type ␣1-FP subunit is labeled WT and imaging of the ␣1(A322D)-FP subunits are labeled AD. ␣1-CFP that colocalized (Co) with FM 4-64 is colored purple and ␣1-YFP that colocalized with FM 4-64 is colored yellow. B, the fraction of each ␣1-FP that colocalized with the plasma membrane marker was graphed. ␣1(A322D) Alters ␣1 Subunit Trafficking after Translation but Before Receptor Assembly-We previously asserted that it was unlikely that the GCC to GAC mutation in the GABRA1 mutant subunit gene reduced ␣1 subunit expression by disrupting transcription or translation. Because GAC codes 11 of the 23 aspartates in the wild type ␣1 subunit, a codon bias against GAC is improbable. Here, we demonstrate that post-translational trafficking is altered with this mutation; the mutant ␣1(A322D) subunit has only immature N-linked glycosylation that is characteristic of ER-associated glycosylation. In addition, confocal microscopy demonstrated that essentially all of the ␣1(A322D)-YFP subunit was sequestered in the ER, whereas a substantial portion of wild type ␣1-YFP subunit was trafficked to the cell membrane. These trafficking differences between ␣1 and ␣1(A322D) subunits cannot be explained by reduced transcription or translation. However, the possibility of reduced translation in addition to altered trafficking cannot be excluded.
The trapping of unassembled GABA A receptor subunits in the ER identified that the ␣1(A322D) mutation occurred before subunit oligomerization. It has been shown that GABA A receptors require at least an ␣ and a ␤ subunit to assemble into pentamers; homomeric GABA A receptor subunits are subject to ERAD (17,24,29). We extended these studies by using confocal microscopy to show that unassembled ␣1-YFP subunits are also retained in the ER (Fig. 6), and that unassembled ␣1(A322D)-YFP subunits lack Golgi-associated glycosylation (Fig. 4). Because the A322D mutation reduced ␣1 subunit expression to approximately the same extent in the absence of cotransfected ␤2 and ␥2 subunits as in ␣1␤2␥2 receptors (Fig. 1), it is clear that this mutation reduces ␣1 subunit expression prior to subunit oligomerization. Whereas it is possible that the ␣1(A322D) mutation may also affect other trafficking processes for the small fraction of ␣1(A322D) subunits that escape ERAD, oligomerize, and traffic to the cell surface, the bulk of ␣1(A322D) subunit expression reduction can be explained by ERAD alone.
Membrane Expression of Heterozygous Receptors Is Composed of Wild Type ␣1 Subunits-By making heterozygous transfections using dually labeled ␣1-YFP/␣1(A322D)-CFP␤2␥2S and ␣1-CFP/␣1(A322D)-YFP␤2␥2S receptors, we demonstrated that the plasma membrane expression of ␣1-FP␤2␥2S receptors resulted almost exclusively from wild type ␣1-FP subunits and not from ␣1(A322D)-FP subunits. In one respect this result was expected given that the ␣1(A322D) mutation reduced ␣1 subunit expression after translation but before oligomerization and transport out of the ER. However, this result was also surprising given that we previously showed that heterozygous receptors with the ␣1(A322D) positioned between the two ␤2 subunits (Het ␤␣␤ ) produce relatively large GABA-evoked peak currents (335 Ϯ 87 pA) (10)), a result that implied that substantial the ␣1(A322D) subunit trafficked to the plasma membrane in heterozygous receptors. One explanation for this apparent discrepancy was that even though Het ␤␣␤ produced relatively large GABA-evoked currents, they were still only 35% of wild type receptor currents. Therefore, one would expect that approximately only 25% of the ␣1 subunit on the cell surface contained the mutant ␣1(A322D) subunit, a result that could be consistent with our confocal microscopy studies (5-17%). A second explanation for this apparent contradiction is that the construction of Het ␤␣␤ and Het ␤␣␥ receptors were forced using subunit concatamers. The mechanisms that target untethered ␣1(A322D) subunits for degradation might not apply to larger concatamer subunits. Finally, it is possible that the ␥␤␣(A322D) concatamer may fold more efficiently than the ␣1(A322D) monomeric subunit and thus not even activate ERAD mechanisms. ␣1(A322D) Subunit Misfolding May Cause ERAD-We hypothesize that the mechanism by which the A322D mutation reduced ␣1(A322D) expression was by misfolding and ERAD (35,36). Could a single point mutation cause misfolding of the majority of translated ␣1 protein? It has been shown in model peptides that at neutral or high pH, a single aspartate positioned near the center of a transmembrane helix would disrupt the helix to allow the aspartate to reside near the aqueous surface (37). Therefore, at physiological pH, it may be expected that a substantial number of ␣1(A322D) M3 segments would fail to form stable transmembrane helices after translation. These misfolded subunits could then be targeted for degradation via ERAD.
The Torpedo AChR three-dimensional structure has been determined and each of its four subunit M3 segments have been demonstrated to be ␣ helical (6). Because the primary structure of the Torpedo marmorata AChR four subunits are homologous to GABA A receptor subunits, we determined the effect on predicted helix formation of placing an aspartate position 7 of the M3 helices of the Torpedo AChR ␣, ␤, ␥, and ␦ subunits. Alignment of the GABA A receptor subunits and Torpedo AChR subunits demonstrated that the Torpedo ␣Ile-283, ␤Ile-289, ␥Phe-297, and ␦Ser-292 residues (numbering from the mature protein) were homologous to the GABA A ␣1(A322) residue (Fig. 9). Each of the transmembrane helix prediction algorithms (DAS, TopPred, TMPred, and TMHMM) accurately identified with good confidence that the M3 segments from each Torpedo AChR subunit were transmembrane helices. Interestingly, when the sequences of each subunit were reanalyzed with ␣Ile-283, ␤Ile-289, ␥Phe-297, or ␦Ser-292 residues changed to an aspartate, the transmembrane helix prediction algorithms identified M3 as a transmembrane helix with substantially lower confidence than without the aspartate substitution. Not surprisingly, the confidence reduction was most significant when using the ToPred algorithm (p Ͻ 0.001), a method that is based upon the clustering of hydrophobic residues. The reduction of confidence of forming a transmembrane helix was also significant using the TMPRED (p ϭ 0.005) and DAS (p ϭ 0.02), but not the THMM (p ϭ 0.1) algorithms. These theoretical analyses suggest that the substitution of an aspartate at position 7 of an M3 segment destabilizes, but does not preclude, formation of a transmembrane helix, a hypothesis that needs to be verified empirically.
We hypothesize that the misfolding and subsequent elimination of the mutant ␣1(A322D) subunit in ADJME is similar to the misfolding and elimination of the cystic fibrosis transmembrane regulator protein (38,39). The majority of the cases of cystic fibrosis are caused by a deletion mutation, ⌬Phe-508, in its cytoplasmic loop. Like the ␣1(A322D) mutation, the cystic fibrosis transmembrane regulator ⌬Phe-508 retains high mannose core glycosylated and sequestered in FIGURE 9. The Torpedo marmorata ␣, ␤, ␥, and ␦ subunit amino acid sequences were aligned with the human GABA A receptor ␣1-6, ␤1-3, ␥1-3, and ␦ subunit sequences. Only the Torpedo subunits and human GABA A ␣1 subunit M3 segment sequences are shown here. This alignment demonstrated that the Torpedo ␣Ile-283, ␤Ile-289, ␥Phe-297, and ␦Ser-292 residues (shaded) were homologous to the GABA A ␣1(A322D) residue (shaded). the ER. Cystic fibrosis transmembrane regulator ⌬Phe-508 has been shown to misfold and undergo rapid ERAD via the 26 S proteasome.
It should be emphasized that our studies were performed in fibroblasts and not in neurons. Although protein trafficking in neurons differs from that in fibroblasts, we suggest that because the trafficking error occurs before oligomerization, it is likely that in neurons the mutant ␣1(A322D) subunit would also undergo ERAD. However, it is also possible that overexpression of the ␣1(A322D) subunit (either in fibroblasts or cultured neurons) may overwhelm protein folding mechanisms and that increased ERAD of the ␣1(A322D) subunit protein would not occur to such an extent in a more physiologic expression system. It would be of substantial interest to create this mutation in a transgenic animal. We predict that in a heterozygous ␣1(A322D) knock-in mouse, the wild type ␣1 subunit would likely be trafficked to the cell surface and the mutant ␣1(A322D) subunit would undergo ERAD, producing a mouse that, like the heterozygous ␣1 knock-out mouse, would have fewer total brain GABA A receptors (40 -42).