The GABA_A receptor α1 subunit epilepsy mutation A322D inhibits transmembrane helix formation and causes proteasomal degradation

Abstract

A form of autosomal dominant juvenile myoclonic epilepsy is caused by a nonconservative missense mutation, A322D, in the GABA_A receptor α1 subunit M3 transmembrane helix. We reported previously that the A322D mutation reduced total and surface α1(A322D) subunit protein and that residual α1(A322D) subunit resided in the endoplasmic reticulum. Here, we demonstrate that the reduction in α1(A322D) expression results from rapid endoplasmic reticulum-associated degradation of the α1(A322D) subunit through the ubiquitin–proteasome system. We provide direct evidence that the α1(A322D) subunit misfolds and show that in at least 33% of α1(A322D) subunits, M3 failed to insert into the lipid bilayer. We constructed a series of mutations in the M3 domain and empirically determined the apparent free energy cost (ΔG_app) of membrane insertion failure, and we show that the ΔG_app correlated directly with the recently elucidated transmembrane sequence code (ΔG_Lep). These data provide a biochemical mechanism for the pathogenesis of this epilepsy mutation and demonstrate that ΔG_Lep predicts the efficiency of lipid partitioning of a naturally occurring protein's transmembrane domain expressed in vivo. Finally, we calculated the ΔΔG_Lep for 277 known transmembrane missense mutations associated with Charcot–Marie–Tooth disease, diabetes insipidus, retinitis pigmentosa, cystic fibrosis, and severe myoclonic epilepsy of infancy and showed that the majority of these mutations also are likely to destabilize transmembrane domain membrane insertion, but that only a minority of the mutations would be predicted to be as destabilizing as the A322D mutation.

GABA_A receptors, the major inhibitory neurotransmitter receptors in the mammalian central nervous system, are pentameric ligand-gated ion channels whose five subunits originate from seven families containing multiple subtypes but most commonly assemble with two α1 subunits, two β2 subunits, and one γ2 subunit (1–3). In mammalian cells, expression of an α-subunit is required to form functional receptors (4). GABA_A receptors are homologous to other types of ligand-gated ion channels in the cys-loop receptor ion channel superfamily, which also includes glycine, serotonin type three, and nicotinic acetylcholine receptors (AChR). Cryoelectron microscopy experiments elucidated the three-dimensional structure of the transmembrane domains of the Torpedo AChR to a resolution of 4Å (5), and it is assumed that the GABA_A receptor structure is similar (6). Each GABA_A receptor subunit contains four transmembrane helices (M1–M4). One M2 domain from each subunit lines the ion channel pore and is surrounded by M1, M3, and M4 helices.

A nonconservative missense mutation (A322D) in the M3 domain of the GABA_A receptor α1 subunit gene (GABRA1) causes autosomal dominant juvenile myoclonic epilepsy (ADJME) (7). Electrophysiological experiments demonstrated that the A322D mutation altered GABA_A receptor function (7–9). In addition, surface trafficking experiments demonstrated that the A322D mutation reduced receptor cell surface expression (9–11). Biochemical experiments revealed that the A322D mutation reduced the total cellular amount of the α1(A322D) subunit posttranslationally before subunit assembly and that the residual α1(A322D) subunits resided in the endoplasmic reticulum (ER) (9, 10). These data suggested that the α1(A322D) subunit underwent ER-associated degradation (ERAD). The mechanism by which ERAD would target a transmembrane protein with only a single missense mutation is unknown, but we hypothesized that, because the A322D mutation inserts a highly polar aspartate into the M3 domain, it could destabilize M3 helix lipid insertion. Here, we directly determined whether the A322D mutation reduced α1(A322D) subunit expression by means of ERAD through the ubiquitin proteasome system and tested whether the addition of aspartate or other highly polar amino acid residues to the M3 domain reduced the efficiency of M3 insertion into the lipid bilayer.

Results

A322D Reduced Total α1 Subunit Expression.

The effect of the A322D mutation on total cellular α1 subunit expression is disputed. Immunoblots of lysates from cells transfected with rat α1β2γ2 or α1(A322D)β2γ2 receptors probed with a polyclonal anti-rat α1 subunit antibody did not differ by visual inspection (7), but immunoblots from cells transfected with human α1β2γ2 and α1(A322D)β2γ2 subunits and stained with a monoclonal anti-human α1 subunit antibody demonstrated that α1(A322D) subunit expression was reduced (9). To address this discrepancy, we transfected cells with cDNAs encoding rat α1β2γ2, α1(A322D)β2γ2, α1-FLAGβ2γ2, or α1(A322D)-FLAGβ2γ2 subunits and performed immunoblots with the same polyclonal ant-rat α1 antibody used by the investigators who found no difference in α1 and α1(A322D) expression (7) (Fig. 1A) or an antibody directed against the FLAG epitope (Fig. 1B). We developed the immunoblots with either chemiluminescence (Fig. 1) or infrared fluorescence [supporting information (SI) Fig. 6] and determined the linear range of detection by quantifying the integrated band density (IBD) at each protein mass as described in Methods.

Fig. 1.

The A322D mutation reduced total cellular expression of the α1 subunit. (A and B) Total lysate protein (2.5–40 μg) from mock-transfected cells (m) or those cotransfected with β2 and γ2 subunits and either α1 (a) or α1(A322D) (d) subunit (A) or α1-FLAG (a) or α1(A322D)-FLAG (d) subunit (B) were analyzed by immunoblot (n = 4). The blots were probed with a primary antibody directed against the α1 subunit (A) or the FLAG epitope (B). (C) Each point represents the mean normalized IBD ± SEM for cells transfected with α1 (●), α1(A322D) (■), α1-FLAG (○), or α1(A322D)-FLAG (□) subunits. The lines are linear regressions through the linear portions of each plot and the connected data points for the untagged wild-type α1 subunit. (D) The mean apparent ratios of α1(A322D) to α1 (▴) were plotted.

Each immunoblot showed two to three bands ranging from 45 to 51 kDa, which corresponded to differentially glycosylated forms of the α1 subunit (10). Wild-type α1 and α1-FLAG subunits were expressed at greater levels than their α1(A322D) and α1(A322D)-FLAG counterparts. However, for the α1 subunit (Fig. 1 A and C), chemiluminescent detection saturated at protein masses >15 μg, and thus, the apparent ratio of α1(A322D) to α1 subunit expression increased from 9 ± 3% at 2.5 μg to 45 ± 5% at 40 μg (Fig. 1D). We also developed the immunoblots using an infrared fluorescent detection system (SI Fig. 6), a nonenzymatic immunoblot imaging system not susceptible to signal saturation from limited chemiluminescent substrate availability. The fluorescent method had substantially less curvature (P < 0.0001) than the chemiluminescent method, and therefore, by using fluorescence, the ratio of α1(A322D) to α1 subunit mass was unchanged from 2.5 to 25 μg. These data (i) confirmed that the A322D mutation reduced α1(A322D) subunit expression and (ii) suggested that nonlinear detection of the α1 subunit by chemiluminescence could explain why previous investigators (7) did not observe a difference between rat α1 and α1(A322D) subunit expression.

α1(A322D) Subunits Were Degraded Faster Than α1 Subunits.

We performed [³⁵S]methionine pulse–chase labeling experiments to determine whether the A322D mutation changed the rate of α1 subunit degradation (Fig. 2A). Immediately after the 10-min pulse labeling, there was no significant difference in the amount of α1(A322D)-FLAG subunit compared with wild-type α1-FLAG subunit (72 ± 17%, P = 0.163). However, the α1(A322D)-FLAG subunit degraded faster than wild-type subunit; from 30 to 180 min of chase, the α1(A322D)-FLAG subunit was significantly reduced relative to the α1-FLAG subunit (P < 0.023). The amount of α1(A322D)-FLAG subunit at each chase time fit well to a single exponential function with a half-life of 23 min. In contrast, the amount of wild-type α1-FLAG subunit did not change for times ≤90 min. Because of its relevance to ADJME patients, we verified that untagged α1(A322D) subunits also degraded faster than wild-type untagged α1 subunits (SI Fig. 7).

Fig. 2.

The A322D mutation caused rapid degradation of the α1(A322D)-FLAG subunit. Pulse-labeled mock-transfected (m) cells or those transfected with either human α1-FLAG (●, a) or α1(A322D)-FLAG (■, d) subunits were lysed, immunopurified, and fractionated by SDS/PAGE. (A) After a 10-min labeling, cells were chased for the indicated times (n = 5). The amount of α1-FLAG and α1(A322D)-FLAG subunits differed significantly at points labeled with an asterisk. (B) Cells were pulse-labeled for 5, 10, 15, or 20 min before lysis (n ≥ 3).

Next, we determined whether the A322D mutation altered the biosynthesis rate of α1 subunits. We pulse-labeled cells transfected with α1-FLAG or α1(A322D)-FLAG subunits with [³⁵S]methionine for various time periods before lysis (Fig. 2B). The A322D mutation did not alter the biosynthetic rate of the subunits; there were no significant differences in the amounts of α1-FLAG and α1(A322D)-FLAG subunits at 5, 10, or 15 min (P > 0.131). Most likely because of the onset of α1(A322D)-FLAG subunit degradation, at 20 min, α1(A322D)-FLAG subunit was 69 ± 14% the density of wild-type subunit. However, even this value was not statistically significant (P = 0.079).

α1(A322D) Subunits Were Degraded via the Ubiquitin–Proteasome System.

Using a standard methodology (12), we tested whether the ubiquitin–proteasome system participated in the degradation of the α1(A322D) subunit. We incubated cells transfected with α1(A322D)-FLAG subunit cDNA with or without the cell-permeable proteasome inhibitor lactacystin (10 μM). The cells were then [³⁵S]methionine pulse-labeled and chased (Fig. 3A). These data demonstrated that proteasome inhibition reduced α1(A322D)-FLAG subunit degradation; bands from treated samples were significantly more intense at times >30 min (172-297%, P ≤ 0.003). Because degradation proceeded even during the 10-min pulse-labeling, at time 0, the IBDs of treated samples were larger than the IBDs of untreated samples. The degradation of both treated and untreated samples could be fit with single exponential functions with half-lives of 157 and 19 min, respectively.

Fig. 3.

The α1(A322D) subunit was degraded through the ubiquitin–proteasome system. (A) Mock-transfected cells (m) or those transfected with the α1(A322D)-FLAG subunit were incubated for 30 min in the absence (−, ■) or presence (+, □) of 10 μM lactacystin (LAC). The cells were then [³⁵S]methionine pulse-labeled for 10 min and then chased for the indicated time periods (n = 4). Statistical significance was indicated by an asterisk. (B) Mock-transfected cells or cells transfected with α1-FLAG (a) or α1(A322D)-FLAG (d) subunits were incubated in the absence or presence of 10 μM lactasystin (n ≥ 5) and analyzed by immunoblot that was probed with an antibody directed against polyubiquitin (B1) or FLAG (B2). The relative amounts of polyubiquitinated subunits were plotted (B3).

To determine whether the subunits were polyubiquitinated, we incubated transfected cells for 4 h in the presence or absence of 10 μM lactacystin before cell lysis. The subunits were then immunopurified and analyzed by an immunoblot (80% sample) probed with an antibody against polyubiquitin (Fig. 3B1) and another immunoblot (20%) probed with an antibody against the FLAG epitope (Fig. 3B2). Lactacystin increased the polyubiquitinated wild-type (336%, P = 0.008) and mutant (395%, P = 0.023) subunit. The absolute amounts of polyubiquitinated subunits did not differ, but the relative amounts (Fig. 3B3) of polyubiquitinated α1(A322D)-FLAG subunit was greater than the polyubiquitinated α1-FLAG subunit in both the presence (median increase 239%, P = 0.031) and absence of lactacystin, although this latter result was not statistically significant (Fig. 3B3, median increase 324%, P = 0.125).

α1(A322D) Receptors Adopted an Altered Transmembrane Topology.

Because ERAD degrades misfolded transmembrane proteins, we determined whether the α1(A322D)-FLAG subunit was misfolded. We hypothesized that the addition of a negative charge to the M3 transmembrane domain by the A322D mutation inhibited M3 insertion into the lipid bilayer. To test this, we determined the glycosylation state of the putative glycosylation site at N365, a site that would be cytoplasmic and unglycosylated if M3 were transmembrane, but in the ER lumen and potentially glycosylated if M3 failed to insert in the membrane (Fig. 4A). The extent of N365 glycosylation corresponded to the fraction of M3 domains that failed membrane insertion. We calculated the free-energy cost of M3 membrane insertion failure, ΔG_app, from the ratio of glycosylated to unglycosylated N365. It should be noted that the extent of N365 glycosylation represented the lower limit of the amount of M3 that failed to insert in the membrane; other factors such as the identity of the amino acids flanking N365 and the rates of both endogenous deglycosylation and degradation of misfolded proteins could reduce the amount of recovered N365 glycosylated α1 subunit. To minimize these factors, we metabolically labeled cells to evaluate only the newly biosynthesized subunits before they underwent substantial degradation. In addition, we labeled in the presence of lactacystin and used subunits containing the mutations K364D and N366T, mutations that computer algorithms predicted would enhance N365 glycosylation. Finally, to separate the effects of N365 glycosylation from glycosylation of N38 and N138 residues, we mutated these two N-terminal asparagines to glutamines (N38Q, N138Q).

Fig. 4.

The A322D mutation altered α1 subunit transmembrane topology. (A1 and A2) Schematic diagrams demonstrate that if the α1 subunit's M3 domain inserts into the ER membrane, N365 will be cytoplasmic and cannot be glycosylated (A1), but, if M3 fails membrane insertion, N365 will be within the ER lumen and could be glycosylated (A2). Although M4 was shown to be transmembrane, these experiments do not determine the M4 transmembrane status. In all constructs, the N-terminal glycosylation sites N38 and N138 were mutated to glutamines, and amino acids flanking N365 were mutated (K3654D + N366T) to enhance glycosylation. Mock-transfected cells (m) or cells transfected with wild-type (a) or mutant (d) subunits were [³⁵S]-labeled and fractionated by SDS/PAGE. (B) With N365 intact, mutant, but not wild-type, subunit migrated as two distinct bands (n = 7). With N365 mutated to glutamine, both wild-type and mutant subunits migrated as a single band. (C) Mutant subunits were untreated (−) or treated (+) with, peptide N-glycosidase F before SDS/PAGE. (D) Untagged mutant subunits, but not wild-type subunits, migrated as two bands (n = 4); a protein that coimmunopurified with the untagged proteins and migrated with the same molecular mass as the unglycosylated subunit is marked (→).

The α1 subunits containing the A322D mutation migrated as two distinct bands. The higher-molecular-mass band comprised 33 ± 2% of the IBD (Fig. 4B), a result consistent with at least 33% of the subunits containing a glycosylated N365 and, thus, adopting an altered transmembrane topology. In contrast, the intensity of the upper band for wild-type subunits was much less distinct and accounted for only 10 ± 3% of the total IBD (P < 0.001). These values corresponded to ΔG_app of 0.63 ± 0.06 kcal/mol and 1.81 ± 0.26 kcal/mol for the mutant and wild-type proteins, respectively.

We verified that the mutant subunit band with increased mass represented glycosylated N365 by two methods. (i) We performed ³⁵S labelings using constructs with N365 mutated to a glutamine (Fig. 4B). (ii) We digested [³⁵S]-labeled mutant subunits with peptide N-glycosidase F before SDS/PAGE (Fig. 4C).

Because of its relevance to a human disease, we confirmed that the presence of the FLAG epitope did not influence the topology by repeating the glycosylation experiments with untagged subunits and immunopurifying using antibodies directed against the native subunit. Like its FLAG-tagged counterpart, the untagged mutant α1(A322D) subunit migrated in two distinct bands, indicating substantial N365 glycosylation and thus altered transmembrane topology (Fig. 4D). Although a nonspecific protein that coimmunopurified with the untagged subunit and migrated at the same mass as the unglycosylated α1 subunit confounded precise quantification, we demonstrated that the untagged mutant α1(A322D) subunit contained significantly more (26 ± 6%) glycosylated N365 than wild-type α1 subunit (5 ± 2%, P = 0.024).

Altered Topology of α1-FLAG Subunits Correlates with M3 Hydrophilicity.

To determine the structural basis underlying the reduction of M3 membrane insertion efficiency by the α1(A322D) subunit mutation, we mutated residues A322 and I329, an M3 residue two helical turns below A322, to a variety of amino acids of varying size, charge, and polarity and determined the ΔG_app for all these constructs (Fig. 5). The ΔG_app did not correlate with amino acid volume; mutating A322 to tryptophan, an amino acid residue with a larger side-chain volume than aspartate, had a substantially larger ΔG_app than that for aspartate (2.29 ± 0.48, 0.63 ± 0.06 kcal/mol). Likewise, inhibition of M3 membrane insertion did not depend strictly on the introduction of charged residues. Introduction of two uncharged, but polar, residues (A322N, I329N) produced an M3 segment with a lower ΔG_app (0.09 ± 0.05 kcal/mol) than when M3 contained the charged aspartate residue. The M3 ΔG_app did correspond with the M3 polarity; it correlated well with both the published transmembrane sequence code ΔG_Lep (13) (Fig. 5B; r² = 0.85) as well as the Wimley–White hydrophobicity scale ΔG_ww (r² = 0.76, Fig. 5C). Our data also correlated well (r² = 0.87, data not shown) with the Goldman–Engelman–Steitz (GES) hydrophobicity scale (14), although the slope was only 0.10, indicating that hydrophilic amino acid substitutions in M3 were not as energetically disfavored as would be predicted by the GES scale.

Fig. 5.

M3 membrane insertion efficiency correlated with its hydrophobicity. We pulse-labeled mock-transfected cells (1) or cells transfected with cDNA encoding α1-FLAG subunits that contained the following M3 substitutions: no substitution (2); A322W (3); A322G (4); A322S (5); A322N (6); A322Q (7); A322N, I329N (8); A322Q, I329Q (9); A322K (10); A322R (11); A322E (12); A322D (13); A332K, I329K (14); and A322D, I329D (15). We listed the mean ΔG_app below the corresponding gel lane. (B) For each mutation, the empirical M3 (ΔΔG_app) relative to alanine was plotted relative to that from the published transmembrane sequence code derived from the Lep protein (ΔΔG_Lep). (C) The M3 ΔG_app was plotted (± SEM, n ≥ 4) relative to the calculated Wimley–White hydrophobicity free-energy change (ΔG_ww). All the data points were plotted with error bars, although for some, the error bars were smaller than the symbol size and thus were obscured.

Discussion

The A322D Mutation Reduced Total α1(A322D) Subunit Expression.

Others reported that the A322D mutation did not reduce total expression of α1 subunits, a conclusion based on (i) visual inspection of fluorescence microscopic images of cells expressing α1-GFPβ2γ2 and α1(A322D)-GFPβ2γ2 receptors (11) and (ii) immunoblots of cellular lysates containing rat α1β2γ2 or α1(A322D)β2γ2 receptors stained with a polyclonal anti-rat α1 antibody (7). Previously, using quantitative fluorescence spectroscopy, we showed that the A322D mutation reduced α1(A322D)-GFP subunit fluorescence (10), and, using immunoblots with a monoclonal antibody, we showed that A322D reduced total cellular α1(A322D) subunit expression (9, 10). Immunoblot experiments depicted in Fig. 1 using two additional antibodies confirmed our previous result that A322D reduced total α1(A322D) subunit expression and suggested that nonlinear detection using the anti-rat α1 subunit polyclonal antibody may explain why the other investigators failed to observe a difference in total α1 and α1(A322D) subunit expression.

The A322D Mutation Altered α1 Subunit Topology.

Popot and Engelman (15, 16) postulated that integral membrane protein folding occurred in two stages: first, transmembrane topology is established as individual transmembrane helices insert into the lipid bilayer, and second, individual helices interact with each other to form tertiary structures. Our experiments demonstrated that the A322D mutation caused 33 ± 2% of the α1(A322D) subunits to fail the first stage, lipid insertion (Fig. 4). Although many disease-causing mutations result in protein misfolding, the misfolding typically alters function, assembly, or subcellular trafficking; protein topology is not affected (17). Recently, using an in vitro bacterial protein leader peptidase (Lep) translation system, Milenkovic et al. (18) suggested that mutations in the macular degeneration-associated bestrophin-1 protein inhibit transmembrane segment insertion. However, we are only aware of one mutation proven to alter topology in a native protein; the prion–protein (PrP) A117V mutation found in Gerstmann–Straussler–Scheinker disease increases the fraction of the transmembrane form of the PrP (19). Therefore, the altered topology of the α1(A322D) subunit represents either a rare or unrecognized structural consequence of a disease-causing mutation.

To determine whether altered topology could contribute to the pathogenesis of other well studied monogenetic diseases associated with missense mutations in integral membrane proteins, we calculated the ΔΔG_Lep for all of the transmembrane domain missense mutations in proteins associated with Charcot–Marie–Tooth disease (1A and X-linked), nephrogenic diabetes insipidus, retinitis pigmentosa, cystic fibrosis, and severe myoclonic epilepsy of infancy (Table 1). Our calculations demonstrated that, for these diseases, the majority of mutations (66 ± 2%, P < 0.001 relative to chance) reduced ΔΔG_Lep and, thus, would be expected to reduce insertion efficiency of the affected transmembrane domain in the lipid bilayer. However, only 3.3% of these “destabilizing” transmembrane mutations reduced ΔΔG_Lep to at least the same extent as the A322D mutation (−3.36 kcal/mol). Therefore, we hypothesize that, although membrane insertion failure may contribute the pathogenesis of several genetic diseases associated with membrane proteins, it plays a substantial role in only a minority of them.

Table 1.

Transmembrane domain missense mutations in disease-associated proteins

Disease	Protein	No. of mutations	Avg. ΔΔG, kcal/mol	Mutations destabilizing, %	Avg. ΔΔG destabilizing mutations, kcal/mol
CMT 1A/HMSN 3	PMP-22	30	−1.2 ± 0.3	73	−2.0 ± 0.2
NDI	Aquaporin-2	7	−0.5 ± 0.5	57	−1.2 ± 0.6
NDI	Vasopressin receptor	62	−0.8 ± 0.2	69	−1.8 ± 0.2
Retinitis pigmentosa	Rhodopsin	29	−1.0 ± 0.3	72	−1.9 ± 0.3
CMT-X	Connexin-32	76	−0.3 ± 0.2	61	−1.1 ± 0.2
Cystic fibrosis	CFTR	43	−0.5 ± 0.3	63	−1.5 ± 0.2
SMEI	SCN1A	30	−0.6 ± 0.3	63	−1.5 ± 0.2
Summary		277	−0.7 ± 0.1*	66 ± 2**	−1.6 ± 0.1

CMT, Charcot–Marie–Tooth 1A, X-linked; HMSN3, hereditary motor-sensory neuropathy-3; NDI, nephrogenic diabetes insipidus; SMEI, severe myoclonic epilepsy of infancy; PMP-22, peripheral myelin protein-22; CFTR, cystic fibrosis transmembrane conductance regulator; SCN1A, sodium channel protein type 1 α -subunit. *, avg ΔΔG_app < 0, P = 0.002; **, percentage of mutations destabilizing >50%, P < 0.001.

The A322D Mutation's Effect on M3 Insertion Corresponded to Biophysical Hydrophobicity Scales.

Secretory and integral membrane proteins cotranslationally traverse the ER membrane via the Sec61 translocon complex (20). Recently, an x-ray structure of the bacterial Sec61 homolog (SecY) revealed that the translocon consists of a gated pore that, during translation, allows the nascent protein to enter the ER lumen from the cytoplasm and a lateral gate that allows transmembrane domains to enter the lipid and avoid crossing the polar phospholipid head groups (21).

A protein domain's amino acid composition determines whether it will insert through the translocon to enter the ER lumen, cross the translocon's lateral gate to become a transmembrane segment, stop the transfer of protein to become a cytoplasmic domain (22), or allow the protein to assume one of two different topologic orientations (23). Hessa et al. (13) used 19 amino acid test peptides fused to the bacterial Lep protein to elucidate the transmembrane sequence code that determines whether a protein domain resides in the membrane bilayer. Our experiments demonstrated that the M3 ΔG_app correlated well with the transmembrane sequence code and the Wimley–White and GES hydrophobicity scales (Fig. 5).

Although our results correlated well with the transmembrane sequence code, the slope was less than unity (0.44), and the intercept was >0 (0.43 kcal/mol). This result may be attributed to the position of the mutated amino acid (A322) within the M3 helix. Hessa et al. derived the sequence code by inserting the test amino acid in position 10, whereas the A322D mutation is at position 4 of M3. During elucidation of the transmembrane sequence code, it was determined that hydrophilic amino acids at the ends of the helix had a larger ΔG_Lep than those placed in the middle of the helix, presumably because at the ends of the helices, they are able to interact with the phospholipid head groups. A positional effect also explains why, in contrast to the transmembrane sequence code, the A322W ΔΔG_app was >0, consistent with electrostatic interactions of the aromatic groups of tryptophan with solvent at the lipid–aqueous interface (24) and preferential placement of tryptophan at the ends of transmembrane domains in integral membrane proteins (25).

Helix Packing May also Contribute to α1(A322D) Subunit Misfolding.

Although at least 33% of the α1(A322D)-FLAG subunits adopt an altered topology, altered topology alone cannot entirely explain the accelerated degradation of α1(A322D)-FLAG. At equilibrium, the A322D mutation reduced α1(A322D)-FLAG expression by 43% (Fig. 1). We hypothesize that other types of misfolding besides altered topology could also account for reduced expression, and, thus, the A322D mutation would inhibit both the first and second of Popot and Engelman's proposed stages of integral membrane protein folding (15, 16). Analyses of transmembrane proteins with known three-dimensional structures revealed that interhelix packing determines the folding of the transmembrane domains and that small, weakly polar amino acids are found at helical interfaces (26). Therefore, substitution of an aspartate for an alanine may disrupt the helix–helix packing of M3 with one of the other three transmembrane helices, causing protein misfolding and degradation. Although poor sequence homology between GABA_A receptor and AChR M3 domains prevents identification of the M3 face occupied by A322, biochemical evidence suggests that it is oriented toward the other helices (6), and, from this position, A322D could disrupt helix-helix packing.

Degradation of the α1(A322D) Subunit May Be the Mechanistic Basis for ADJME.

Here we demonstrated that the α1(A322D) subunit was degraded rapidly via ERAD through the ubiquitin–proteasome system. We hypothesize that in ADJME patients, who are heterozygous for the A322D mutation, α1(A322D) degradation results in haploinsufficiency. Previously, we demonstrated that cells transfected with “heterozygous” α1α1(A322D)β2γ2 receptors (i.e., a 0.5:0.5 α1:α1(A322D) cDNA mixture) produced significantly smaller GABA_A receptor peak-current amplitudes and contained fewer surface receptors compared with cells transfected with wild-type α1β2γ2 receptors (9). In addition, we reported that in “heterozygous” cells, the majority of surface GABA_A receptors contained wild-type α1, but not mutant α1(A322D), subunits (10). These previous results coupled with the data presented here provide a model to explain how the A322D mutation causes ADJME. We propose that in heterozygous patients, both wild-type α1 subunit and α1(A322D) subunit genes are transcribed and translated, but misfolded subunits are degraded, resulting in a posttranslational haploinsufficiency. In neurons, haploinsufficiency could result in reduced total GABA_A receptor expression or unchanged total GABA_A receptor expression with compensation by up-regulation of other α subunit subtypes (α2-α6), which would confer different electrophysiological properties than α1-containing receptors.

Of the 17 known genes associated with monogenetic idiopathic epilepsy syndromes (27), only the A322D mutation in the α1(A322D) subunit causes misfolding and accelerated degradation. Three epilepsy genes are GABA_A receptor subunits: α1, γ2, and δ subunits. The γ2(K289M), γ2(R139G), δ(E177A), and δ(R220H) epilepsy mutations/polymorphisms alter receptor function (28–30), whereas the γ2(R43Q) and γ2(Q351X) epilepsy mutations reduce cell surface expression without altering total cellular subunit expression, likely through altering oligomer assembly and cell surface trafficking (31–33). Recently, Maljevic et al. (34) identified a 2-bp deletion in the GABRA1 gene, α1(S326fs328X), in a patient with childhood absence epilepsy. Were this gene transcript translated in vivo, it would introduce a premature stop codon in the M3 domain, which could cause α1 subunit misfolding and degradation and may provide a second example of the mechanism of epileptogenesis typified here by the mutant α1(A322D) subunit, namely, misfolding and degradation of the mutated protein.

Methods

Expression of Recombinant GABA_A Receptor Subunits.

The cDNAs encoding human GABA_A receptor α1, β2, and γ2 subunits were expressed in pcDNA3.1 plasmids, and those encoding rat α1, β2, and γ2 subunits were expressed in pCMV plasmids. Rat α1 subunit cDNA that contained the sequence coding for the FLAG epitope (DYKDDDDK) inserted between the seventh and eighth amino acids of the mature protein was expressed in the pcDNA3.0 plasmid and was a gift from J.-H. Steinbach (Washington University, St. Louis, MO). We used PCR to construct the cDNA encoding the human α1 subunit that also contained the sequence coding for the FLAG epitope between the seventh and eighth amino acids. All missense mutations were made by using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). All cDNA sequences were confirmed by DNA sequencing.

HEK293T cells were purchased from American Type Culture Collection (Manassas, VA) and were maintained at 37°C, 5% CO₂/95% air in DMEM (Invitrogen, Carlsbad, CA) containing 10% FBS (Invitrogen) and 100 units/ml of penicillin and streptomycin (Invitrogen). Cells were transfected with 6 μg of total plasmid DNA by using the Fugene 6 reagent (Roche Diagnostics, Indianapolis, IN).

Cell Lysis.

Cells lysis buffer consisted of 50 mM Tris·HCl, 250 mM NaCl, 5 mM EDTA, 1% Triton X-100, one pellet per 10 ml of Complete Miniprotease inhibitor (Roche Diagnostics) and, for the polyubiquitination experiments, also contained 5 mM N-ethylmaleimide (Sigma–Aldrich, St. Louis, MO). Protein concentrations were determined by using a colorimetric bicinchoninic acid method with BSA as a standard (Pierce, Rockford, IL).

Immunopurification.

The α1-FLAG and α1(A322D)-FLAG subunits were purified by incubating the cell lysates (1 mg/ml) overnight with 50 μl of agarose-immobilized anti FLAG M2 antibody (Sigma–Aldrich). The antibody resin was pelleted by centrifugation and washed three times with lysis buffer, and the subunits were liberated by either incubation with 30 μl Laemmli sample buffer or by incubation for 30 min with the 3X FLAG peptide (Sigma–Aldrich). Untagged α1 subunits were immunopurified by using a mouse monoclonal anti-α1 antibody (clone BD24, 2 μg; Millipore, Billerica, MA) and protein A/G agarose (Santa Cruz Biotechnology, Santa Cruz, CA).

Metabolic Labeling.

Forty-eight hours after transfection, the cells' DMEM was exchanged with 3.0 ml of starving media, DMEM that lacked methionine and cysteine (Invitrogen), and incubated at 37°C for 30 min. The starving media was then exchanged for 1.5 ml of labeling media, DMEM that also contained [³⁵S]methionine [100–250 μCi/ml (1 Ci = 37 GBq); PerkinElmer, Wellesley, MA], and incubated at 37°C in for the times indicated in the figure legends. For the degradation experiments, the labeling media were exchanged for nonradioactive DMEM, and incubation was continued for the indicated times.

For the proteasome inhibition experiments and the glycosylation experiments, both the starving media and the labeling media contained 10 μM lactacystin (Boston Biochem, Cambridge, MA) or an equivalent concentration (0.1%) of DMSO for the controls. For glycosylation experiments using the untagged α1 subunits, both the starving and labeling media also contained 30 μM N-carboxybenzyl-Val-Ala-Asp(O-Me) fluoromethyl ketone (Sigma–Aldrich), a cell-permeable inhibitor of endogenous deglycosidases (35). The peptide N-glycosidase F digestion has been described previously (10). Immunopurified subunits were fractionated by 10% SDS/PAGE, which were exposed on a digital PhosphorImager (GE Healthcare, Piscataway, NJ).

Immunoblots.

Equal masses of total protein from each condition were fractionated by 10% SDS/PAGE and electrotransfered to nitrocellulose membranes, which were blocked with 0.1% Tween 20 in Tris-buffered saline that contained 0.5% BSA. Immunoblots were probed with a primary rabbit polyclonal antibody directed against the rat GABA_A receptor α1 subunit (1:1,000, lot 026K4848; Millipore), FLAG epitope (1:1,000, lot 31775; Sigma–Aldrich), or a mouse monoclonal antibody (clone P4G7) directed against mono- and polyubiquitin (1:1,000; Covance, Princeton, NJ). To verify equivalent lysate loading, blots were also probed with a monoclonal antibody directed against the sodium–potassium ATPase α-subunit (1:5,000; Abcam, Cambridge, MA). Secondary antibodies included either HRP-conjugated (1:5,000; Jackson ImmunoResearch, West Grove, PA) or fluorophore-conjugated (1:20,000, Li-Cor, Lincoln, NE) goat anti-mouse or goat anti-rabbit antibodies. Chemiluminescence was detected by a digital imager (Alpha Innotech, San Leandro, CA), and fluorescence was detected by using an infrared fluorescence imaging system (Li-Cor).

Data Analysis.

IBDs were quantified by using Quantity One software (Bio-Rad, Hercules, CA). Values are reported as means ± SEM, with the exception of polyubiquitin experiments (non-Gaussian), which are reported as median values. Statistical significance was determined by using Student's unpaired t test or, if appropriate, single-value t test, with the exception of the polyubiquitin experiments, which were evaluated by using the Wilcoxin signed-rank test (GraphPad, San Diego, CA).

We determined the protein concentrations at which the immunoblots' protein concentration vs. IBD curves maintained linearity by the following method: (i) we performed linear-regressions (through the origin) of each successive three data points in the protein concentration vs. IBD curves, and (ii) we used Student's t test to determine the protein concentration at which the slopes became significantly different. The degree of curvature was determined by calculating the second derivatives at the point of nonlinearity.

The fraction of M3 segments that failed to insert into the bilayer (ƒ) was calculated as the ratio of the IBDs of the glycosylated N365 divided by the IBD of unglycosylated N365; only IBDs >0 were included in the analyses. The ΔG_app was calculated as ΔG_app = −RT ln(ƒ), where T is the temperature (310 K), and R is the gas constant, 1.99 cal mol⁻¹K⁻¹. The theoretical M3 (V319-F338) ΔG_ww was determined by using the MPEx software (36) using the Wimley–White hydrophobicity scale (37, 38). We also used the GES hydrophobicity scale (14) and calculated the theoretical free-energy change in lipid/aqueous partitioning expected from substituting an alanine (for A322) or an isoleucine (for I329) for the amino acids that we analyzed.

Abbreviations

ADJME

autosomal dominant juvenile myoclonic epilepsy

IBD

integrated band density

Lep

leader peptidase protein

ERAD

endoplasmic reticulum-associated degradation

AChR

nicotinic acetylcholine receptor.