A γ2(R43Q) Mutation, Linked to Epilepsy in Humans, Alters GABAA Receptor Assembly and Modifies Subunit Composition on the Cell Surface*

Genetic defects leading to epilepsy have been identified in γ2 GABAA receptor subunit. A γ2(R43Q) substitution is linked to childhood absence epilepsy and febrile seizure, and a γ2(K289M) mutation is associated with generalized epilepsy with febrile seizures plus. To understand the effect of these mutations, surface targeting of GABAA receptors was analyzed by subunit-specific immunofluorescent labeling of living cells. We first transfected hippocampal neurons in culture with recombinant γ2 constructs and showed that the γ 2(R43Q) mutation prevented surface expression of the subunit, unlike γ2(K289M) substitution. Several γ2-subunit constructs, bearing point mutations within the Arg-43 domain, were expressed in COS-7 cells with α3- and β3-subunits. R43Q and R43A substitutions dramatically reduced surface expression of the γ2-subunit, whereas R43K, P44A, and D39A substitutions had a lesser, but still significant, impact and K289M substitution had no effect. Whereas the mutant γ2(R43Q) was retained within intracellular compartments, αβ complexes were still targeted at the cell membrane. Coimmunoprecipitation experiments showed that γ2(R43Q) was able to associate with α3- or β3-subunits, although the stoichiometry of the complex with α3 was altered. Our data show that γ2(R43Q) is not a dominant negative and that the mutation leads to a modification of GABAA receptor subunit composition on the cell surface that impairs the synaptic targeting in neurons. This study reveals an involvement of the γ2-Arg-43 domain in the control of receptor assembly that may be relevant to the effect of the heterozygous γ2(R43Q) mutation leading to childhood absence epilepsy and febrile seizure.

␥-Aminobutyric acid (GABA) 2 is the major inhibitory neurotransmitter in the central nervous system. A modification in the GABAergic systems has been postulated in many neurolog-ical and psychiatric diseases (1). For instance, alteration or modulation of GABA neurotransmission plays an important role in epilepsy mechanisms and treatment (2)(3)(4). Furthermore, several mutations of GABA A receptor genes have recently been associated with epileptic syndromes (5)(6)(7)(8). These mutations offer an opportunity to obtain new insights into the structure and function of GABA A receptors (9,10) and may provide clues to implication of these receptors in epilepsies (4,11).
The ionotropic GABA receptors, mediating fast inhibitory transmission, are heteromeric structures composed of a combination of five out of at least 18 different subunits (␣1-6, ␤1-3, ␥1-3, ⑀, , ␦, and 1-3), grouped into several classes (12,13). Studies have demonstrated that the subunit composition determines their affinity for GABA and the specific effect of allosteric modulators, as well as receptor trafficking. The major GABA A receptor in the brain is believed to consist of two ␣1-, two ␤2-, and one ␥2-subunits (14). There is considerable evidence that the ␥2-subunit plays an essential role in the response to benzodiazepine modulators and receptor targeting (15).
An R43Q mutation in the ␥2-subunit N-terminal domain has been found to cause childhood absence epilepsy with febrile seizure (6). Analysis of the consequences of this mutation on GABA A receptors expressed in different heterologous cell types has produced conflicting data, raising the question of the subunit composition of receptors on the cell surface. It has been shown that the mutated subunit does not modify benzodiazepine sensitivity of GABA A receptors whereas most of the complexes are retained intracellularly (10,16), in contradiction with other findings showing that ␥ 2 (R43Q) specifically modifies pharmacological and functional properties (6,9,17,18). Moreover, none of these experiments were performed on neurons.
To elucidate whether ␥2(R43Q) mutation altered subunit assembly or trafficking, we decided to study the surface expression of recombinant GABA A receptors by subunit-specific labeling. Labeling of transfected neurons showed that the ␥2(R43Q) mutation prevented surface targeting. Co-expression experiments in heterologous cells showed that the ␥2(R43Q)-subunit was retained intracellularly, as it was in neurons, whereas GABA A receptor complexes reached the cell surface membrane. Finally, immunoprecipitation experiments revealed a modification in subunit assembly induced by the ␥2(R43Q) mutation. Altogether, our findings indicate a rela-tionship between the expression of GABA A receptors on the cell surface and epilepsy.
Embryonic hippocampal neurons were obtained from E18 rat embryos as described earlier (19). Briefly, dissected hippocampi were incubated in trypsin and mechanically dissociated using a Pasteur pipette. The cells were counted and plated onto poly-D-lysin-coated 12-mm glass coverslips in complete Neurobasal medium (Invitrogen) supplemented with B27 (Invitrogen), 1 mM glutamine (Invitrogen), and 0.01% penicillin/streptomycin (Invitrogen). Four hours after plating, the medium was replaced with conditioned medium obtained by incubating complete Neurobasal medium on glial cultures for 24 -48 h.
DNA Constructs-Rat ␣1, ␣3, ␤3, and ␥2 GABA A subunits were subcloned in the pcDNA3 vector (Invitrogen). We used constructs that were available from previous studies (20). The 6-Myc epitope (MEQKLISEEDLNE, repeated 6 times) was inserted between amino acids 4 and 5 of the ␥2Sor ␥2L-subunits. As previously described, an insertion within this domain does not modify the functional properties of GABA-or glutamate-gated channels (19,(21)(22)(23). Receptor mutants were generated by polymerase chain reaction with the wild-type plasmid as a DNA template, using PfuTurbo DNA polymerase (Stratagene, La Jolla, CA) to minimize artifactual mutations. Point mutations were constructed using the QuikChange site-directed mutagenesis system (Stratagene). All constructs were verified by automatic dideoxy DNA sequencing (Genome Express, Meylan, France).
Cell Transfection-COS-7 cells were transfected using the FuGENE 6 reagent (Roche Applied Science) according to manufacturer specifications. Equal amounts of ␣3and ␤3-subunit cDNAs (0.3 g/well in 24-well plates for immunochemistry, 0.6 g/well in 6-well plates for biochemistry) were mixed with different amounts (equal or 8-fold) of wild-type or mutated ␥2 cDNAs and then added to FuGENE 6. Cells were incubated for 24 h with cDNAs before analysis.
Hippocampal neurons were transfected at 7-11 days in vitro using the Lipofectamine 2000 reagent (Invitrogen) according to manufacturer specifications. Briefly, 0.5 g of cDNAs, coding for wild-type or mutated ␥2-subunits, were mixed with 1.5 l of Lipofectamine 2000 in Neurobasal medium. The cDNAs were then incubated with neurons in serum-free medium for 40 -50 min before replacing the latter with conditioned complete Neurobasal medium (see above). Neurons appeared morphologically normal and expressed the neuronal marker NeuN (not shown). The cells were analyzed 24 -60 h after transfection.
Quantitative Analysis of Fluorescence Signals-Fluorescence microscopy was performed using a Zeiss Axioplan 2 microscope, with a 63 ϫ 1.4 numerical aperture oil immersion lens. Quantification of fluorescence signals and background substraction were performed using Metamorph Software (Molecular Devices Corp., Downington, PA). For each image acquired, background levels were determined using the surface and intracellular signals measured in neighboring non-transfected cells and subtracted from the values obtained in transfected cells. Numerical data are presented as mean Ϯ S.E., and statistical significance, assessed using Kruskal-Wallis analysis of variance (Dunn's method) test (unless otherwise stated), was taken as p Ͻ 0.05 (Sigmastat; SPSS Inc.).
Coimmunoprecipitation-COS-7 cells were homogenized in buffer containing 20 mM HEPES, 0.15 mM EDTA, and 10 mM KCl, pH 8, supplemented with a mixture of protease inhibitors (Roche Applied Science). The buffer was then adjusted to 12% sucrose, and after four more strokes the cells were centrifuged at 2000 rpm for 3 min to remove genomic DNA. The supernatant was centrifuged at 15,000 rpm for 30 min. The pellet was recovered, and cell membranes were solubilized with 15 strokes in a buffer containing 20 mM Tris-HCl, 0.15 mM EDTA, 150 mM NaCl, 2% Triton X-100, and 0.5% deoxycholate, pH 8, supplemented with a mixture of protease inhibitors and then incubated for 45 min. The sample was centrifuged for 45 min at 15,000 rpm. The supernatant was incubated with 30 l of Protein-G-Sepharose (Amersham Biosciences) for 1 h and centrifuged at 2000 rpm for 3 min. The clarified supernatant was then incubated with 5 g of mouse anti-Myc 9E10 antibody for 2 h and then overnight with 30 l of anti-mouse Protein-A-Sepharose (Sigma). The resin was washed with the solubilization buffer and with the same buffer containing 500 mM NaCl. All steps were performed at 4°C. Four independent experiments were performed. The beads were resuspended in 80 l of gel loading buffer, separated by 10% SDS-PAGE, and immunoblotted with the appropriate antibodies as previously described (19). Antibodies used were anti-␣3 (1:2000; Alomone Labs), anti-␤3 (1:1000; Chemicon), and polyclonal anti-Myc (1:500; Upstate). The three antigens were detected consecutively on each Western blot membrane after stripping to selectively remove both primary and secondary antibodies (24). Each immunoprecipitate was analyzed twice. Pictures of the blots were taken with the ChemiGenius 2XE apparatus under the control of GeneSnap program. Quantitative analysis was performed using GeneTools Software (Syngene, Cambridge, UK).
Structural Building of GABA A Receptor Mutants-The coordinate file (GABA-A_expanded.pdb) of the GABA A receptor homology model (31) was used to generate mutant models. Mutations were created using DeepView/Swiss-PdbViewer software (25). Steepest descent energy minimization was performed using the GROMOS96 force field (26).

RESULTS
␥2(R43Q)-Subunit Does Not Reach the Surface of Neurons and COS-7 Cells-Analysis of ␥2(R43Q)-subunit expression in heterologous systems has produced controversial data (6, 9, 10, 16 -18). To study this mutation in a neuronal environment, hippocampal neurons at 7 to 11 days in vitro were transfected with wild-type and mutant ␥2 6myc . Subunits expressed on the cell surface were detected on living cells to avoid any intracellular labeling that could occur with fixed cells (27). Labeling the extracellular epitope with a polyclonal antibody directed against the Myc tag revealed that expression of the wild-type ␥2-subunit on the surface membrane was uniformly distributed throughout the neurons (Fig. 1, top lane). After permeabilization, labeling of the Myc tag with a monoclonal antibody revealed perinuclear and reticular structures that probably correspond to the endoplasmic reticulum, where a large majority of subunits are retained during the protein folding and assembling processes (16,23,28,29). When the arginine in position 43 of ␥2 6myc was replaced with a glutamine residue, neurons transfected with ␥2 6myc (R43Q) were not labeled at the surface ( Fig. 1, middle lane). This effect was specific to ␥2 6myc (R43Q), as ␥2 6myc (K289M), known to be responsible for generalized epilepsy with febrile seizure plus (5), was targeted to the neuron cell surface (bottom lane). We obtained similar data with wildtype or R43Q mutant when ␥2S 6myc (Fig. 1) or ␥2L 6myc (not shown) constructs were expressed in transfected neurons for 24 ( Fig. 1), 36, or 60 h (not shown). These experiments showed that the ␥2(R43Q)-subunit was not addressed to the surface when expressed in neurons.
To study the mechanisms behind this deficient GABA A receptor surface targeting, we co-expressed wild-type and mutated ␥2 6myc constructs with ␣and ␤-subunits in COS-7 cells. In ␣3-, ␤3-, and ␥2 6myc -transfected cells, anti-Myc polyclonal surface labeling of living cells revealed clusters all over the cell plasma membrane (Fig. 2, top lane). When COS-7 cells were transfected with the ␥2 6myc (R43Q) mutant, surface staining was close to background (middle lane) whereas the ␥2 6myc (K289M) was clearly targeted on the cell surface (bottom lane). Our result for ␥2(K289M) was in agreement with previous data showing that this mutant expressed in heterologous cells gave rise to GABA-induced current with unaltered amplitude (10). Moreover, as previously observed in neurons, there was no difference when experiments on COS-7 cells were conducted using the ␥2S (Fig. 2) or ␥2L splice form (not shown) or when COS-7 cells were transfected with ␣1instead of the ␣3-subunit (not shown).
␥2(R43) Domain Is a Major Determinant to Drive Cell Surface Targeting of the ␥2-Subunit-Crystal structure analysis of a soluble acetylcholine-binding protein was used to design a model of the GABA A receptor ligand-binding domain (30,31). A homology model of the whole GABA A receptor (Fig. 3A) was Neurons were transfected with either a wild-type subunit, ␥2 6myc WT, or one bearing point mutations, ␥2 6myc (R43Q) or ␥2 6myc (K289M), as indicated on the left. Surface labeling was obtained after immunostaining living neurons with a polyclonal antibody directed at the extracellular 6-Myc tag and Alexa Fluor 568-conjugated secondary antibodies (surface). Intracellular staining of the same cells was obtained after permeabilization and labeling with a monoclonal antibody directed at the same tag and FITC-conjugated secondary antibodies (intra). Merged images (overlay) revealed wild-type ␥2and ␥2(K289M)-subunits at the surface of transfected neurons whereas ␥2(R43Q) was essentially intracellular. Scale bar, 20 m.  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 Residues are shown in stick representation. The side-chain atoms of ␥2 residues mutated in this study are shown as well as ␥2(S171), ␥2(E178), and ␤2(R117). For the sake of clarity, residues around ␥2(R43) and ␤2(R117) are only represented by their main chain and residues between ␥2(S171) and ␥2(E178) have been omitted. Molecular figures were generated with PyMOL (pymol.org). F, loop region model at the ␥2␤2 interface. The loop region containing residues ␥2(D39-P44) is shown in orange and the loop ␥2(S171-E178) in red. According to the whole GABA A model, residues ␥2(S171-E178) belong to a loop that connects two ␤-sheets of an extended ␤-sandwich of the ␥2-subunit. The loss of a positive charge at position 43 could change the ␥2(E178)-␤2(R117) interaction, and this may alter the flexibility of this loop and/or the ␥2␤2 interface rigidity. also generated by combining the extracellular domain model with a model of the membrane domain constructed from the coordinates of the acetylcholine receptor pore (31,32). These homology models indicated that a bifurcated salt bridge may connect the ␥2(E178), ␥2(R43), and ␤2(R117) residues (Fig. 3B). Using the coordinate file of the whole receptor homology model, we constructed models of several Arg-43 mutants. In the R43Q mutant (Fig. 3C), ␥2(E178) only interacted ionically with ␤2(R117), whereas the ␥2(R43K) mutant model suggested that there could be a salt bridge network similar to the one found in the wild-type receptor (Fig. 3D and supplemental  movie). We also used the GABA A wild-type homology model to highlight the Asp-39 position, conserved among GABA-gated receptor channels, and Pro-44, conserved among the Cys loop ligand-gated receptor channel family (Fig. 3E). We created mutants of the residues mentioned above to study their importance in receptor assembly or trafficking. Interestingly, we noted that ␥2(S171), whose influence in GABA A receptor subunit assembly has recently been reported (33), is located at the beginning of the loop close to ␤2(R117) that ends with ␥2(E178) (Fig. 3, E and F). The ␥2K289 and ␥2R43 residues are located far apart (48 Å, Fig. 3A); thus, K289M substitution was not expected to affect GABA A receptor properties in the same way as mutations within the Arg-43 domain.

␥2(R43Q) and GABA A Receptor Trafficking
As the ␥2(R43Q) mutant behaved in a similar way when it was expressed in COS-7 cells or neurons and surface labeling is an efficient way to study GABA A receptor expression (34), we used this approach, with several ␥2 6myc mutants, to study the role of the domain surrounding the Arg-43 position. Cells were transfected with ␣3and ␤3-subunits in combination with ␥2S 6myc bearing different point mutations. Immunolabeling showed that, as described above for ␥2(R43Q), ␥2(R43A) mutant was not detected on the cell surface, whereas ␥2(R43K), ␥2(D39A), and ␥2(P44A) were consistently expressed at the surface (Fig. 4A). Quantitative analysis of intracellular ␥2 6myc revealed that there was no significant difference (p Ͼ 0.05) in expression levels of the wild-type subunit and any of the mutants in the Arg-43 domain (Fig. 4B), showing that the effect of mutations was not due to a subunit expression or stability defect. Surface targeting was analyzed by measuring the ratio of the amount of ␥2-subunit labeled on the cell surface versus ␥2 labeled within intracellular compartments (Fig. 4C). The surface versus intracellular signal ratio showed that both the R43Q and R43A mutations prevented cell surface targeting compared with wild type (0.033 Ϯ 0.002, 0.052 Ϯ 0.003, 1.163 Ϯ 0.078, respectively). Cell surface labeling was detected on cells transfected with the ␥2S 6myc (R43K) construct (Fig. 4A), showing that this conservative substitution restored ␥2 surface targeting (0.235 Ϯ 0.033) to some extent. When the conserved Asp-39 or Pro-44 residues were replaced with alanine, labeling was detected on the membrane surface (Fig. 4A), although the ␥2-subunit ratio on the cell surface was reduced (0.396 Ϯ 0.083 and 0.135 Ϯ 0.011, respectively). The significantly higher level of intracellular signals for the ␥2(K289M) mutant (p Ͻ 0.05; Fig.  4B) may be due to a different expression level or stability. This increase did not interfere with surface targeting efficacy, because the K289M substitution had no significant effect on the surface versus intracellular signal ratio (0.84 Ϯ 0.093) compared with the wild type, in agreement with previous studies (10). These experiments clearly showed the major impact of the Arg-43 residue on ␥2-subunit cell surface localization and led us to evaluate the influence of this position, either on the subunit itself or on GABA A receptor complexes associated with this subunit.

␥2(R43Q) Mutant Is Not a Dominant
Negative-Our data were in contradiction with previous reports suggesting that ␥2(R43Q) acted as a dominant negative mutant by forming a complex with ␣and ␤-subunits and retaining this complex within intracellular compartments (9,10,16). To investigate this issue, we overexpressed wild-type or R43Q-␥2S 6myc constructs with ␣3and ␤3-subunits in COS-7 cells at a 8:1:1 ratio (Fig. 6A). Compared with co-expression with wild-type ␥2S 6myc , overexpression of ␥2S 6myc (R43Q) mutant did not modify surface labeling of the ␣3-subunit (Fig. 6B). Similar data were obtained when the ␣3-subunit was replaced with ␣1 (supplemental data). As it is well established that ␣-subunits are addressed to the cell surface only in association with ␤ (with or without ␥), these data showed that, irrespective of the amount of mutated ␥2-subunit, R43Q substitution did not prevent the  assembly of ␣3-(or ␣1-) and ␤3-subunits or their cell surface targeting.
To determine whether ␥2(R43Q) behaved the same way in neurons, we investigated the possibility that recombinant subunit constructs co-assembled with native subunits in our experiments. Thus, we transfected hippocampal neurons with the ␥2 6myc (P44A) mutant that had been shown to be addressed to the COS-7 cell surface only when co-transfected with ␣and ␤-subunits (Figs. 4 and 5). Surface labeling of living neurons (Fig. 7A) clearly showed that the recombinant mutant was targeted to the neuronal cell membrane, suggesting that ␥2 6myc (P44A) co-assembled with native subunits. This was in agreement with previous findings that recombinant GABA A receptor ␣1-subunits were able to form functional complexes with native subunits in transfected hippocampal neurons (36). When wild-type and ␥2(R43Q) were expressed in neurons, there was no apparent modification in native ␣3-subunit surface labeling (Fig. 7B). Taken together, our experiments showed that recombinant ␥2 6myc subunit constructs co-assembled with native subunits and that expression of the ␥2(R43Q)subunit did not dramatically reduce the number of GABA A receptors on the cell surface in neurons, as is also the case in COS-7 cells.

␥2(R43Q) Mutation Disturbed GABA A Receptor Subunit
Assembly-Our experiments showed that, when ␣-, ␤-, and ␥2(R43Q)-subunits were co-expressed, ␣ was found on the cell surface whereas ␥2(R43Q) was absent. These findings suggested that ␥2(R43Q) did not associate with the ␣␤ complex. To challenge this hypothesis, we expressed ␣3and ␤3-subunits in COS-7 cells with either wild type or R43Q-␥2 6myc (1:1:1 ratio). Western blot analysis of membrane protein extracts and labeling, using either polyclonal anti-Myc, anti-␣3, or anti-␤3, detected bands corresponding to the expected size (Fig. 8A, left  panel). Proteins associated with the tagged ␥2-subunit were co-immunoprecipitated, and analysis showed that ␣3and ␤3-subunits were associated with both wild type and R43Q-␥2 6myc (right panel). The ␥2 6myc -, ␣3-, or ␤3-subunits in each immunopurified sample were quantified (Fig. 8B). This analysis showed that the amount of ␤3 associated with ␥2 6myc (R43Q) was not significantly different from the amount associated with wild-type ␥2 6myc (92 Ϯ 4%). Surprisingly, the amount of ␣3 associated with ␥2 6myc (R43Q) was higher than with wild-type FIGURE 7. Expression of recombinant ␥2S 6myc subunits in transfected hippocampal neurons. A, neurons were transfected with the ␥2 6myc (P44A) mutated subunit. The surface expression of ␥2-subunits was detected by double labeling with a polyclonal antibody directed at the extracellular 6Myc tag and Alexa Fluor 568-conjugated secondary antibodies (surface). Intracellular staining of the same cells was obtained after permeabilization and labeling with a monoclonal antibody directed at the same tag and FITC-conjugated secondary antibodies (intra). Merged images (overlay) revealed ␥2(P44A)-subunits on the surface of transfected neurons. B, neurons were transfected with either wild-type ␥2-subunit (␥2 6myc WT) or ␥2 6myc (R43Q). Surface labeling was obtained after immunostaining of living neurons with a polyclonal antibody directed at the extracellular domain of the native ␣3-subunit (surface). Intracellular staining of the same neurons after permeabilization and labeling with a monoclonal antibody directed at the 6Myc tag (intra). Scale bar, 20 m.  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 ␥2 6myc (186 Ϯ 18%). These data showed that the ␥2(R43Q)subunit is able to assemble with ␣3 and ␤3 and that mutation modified the subunit assembly of GABA A receptors.

DISCUSSION
Several mutations in GABA A receptor subunit genes have been implicated in epilepsy (2, 5-8, 37, 38). Functional studies of these mutations are required to elucidate the mechanisms underlying these disorders. However, analysis of the consequence of ␥2(R43Q) mutation, using the expression of recombinant subunits in heterologous cells, has produced controversial data (6, 9, 10, 16 -18). This led us to analyze the effect of ␥2(R43Q) mutation in a neuronal environment and continue in-depth studies using heterologous cells.
R43Q Mutation Prevents ␥2-Subunit Cell Surface Targeting in Neurons-We used subunit-specific labeling to monitor receptor targeting. We expressed tagged ␥2-subunits and analyzed the expression of ␥2(R43Q) in neurons, for the first time. 1 to 3 days post-transfection, we showed that, whereas both wild-type ␥2S and ␥2L isoforms were expressed at the surface of cultured neurons, surface expression of ␥2(R43Q) mutants was below detection level. It was possible that we had detected the traffic of wild-type recombinant ␥2S-subunit to the surface of neurons in monomer form, as it has been shown in heterologous cells transfected with this subunit alone that this isoform is targeted to the cell plasma membrane (Ref. 35 and this work). However, our experiments suggest that this is unlikely. We found that the ␥2(P44A) mutant was not driven to the cell membrane of transfected COS-7 cells when expressed alone, whereas it was detected on the cell surface when co-expressed with ␣and ␤-subunits (see Figs. 4 and 5). When the ␥2(P44A) mutant was transfected in cultured hippocampal neurons, it was detected on the cell surface, suggesting an association of the recombinant ␥ construct with native ␣and ␤-subunits (Fig. 7A). In addition, the ␥2L isoform, reported to have less propensity to traffic to the cell surface in monomer form than ␥2S (35), was also detected on the cell surface of transfected neurons.
As recombinant receptor expression in heterologous cells makes it possible to visualize individual subunits, we used this method to study the effect of the ␥2(R43Q) mutation on GABA A receptor subunit assembly and cell surface targeting. However, overexpression of membrane proteins in heterologous cells may alter their expression or stoichiometry (39,40), which may account, at least in part, for the contradictory previous findings (9,10,17,18). In our expression experiments in COS-7 cells, designed to avoid overexpression, when a tagged wild-type ␥2 was co-expressed with ␣3-(or ␣1-) and ␤3-subunits, the extracellular N-terminals of either the ␣or ␥2-subunits were detected on the cell surface. Most importantly, the ␥2(R43Q) mutant expressed in COS-7 cells, as in neurons, had hardly any surface expression, showing that experimental conditions were appropriate for studying the effect of mutations on GABA A receptor subunit trafficking in heterologous cells. In addition, these data are in line with biochemical analyses showing a vast decrease in ␥2(R43Q) at the surface of human embryonic kidney 293-transfected cells (18).

␥2(R43Q) Substitution Is a Decisive Factor in the Effect on GABA A Receptor Targeting-Our experiments demonstrated
that any mutation within the domain surrounding ␥2(R43) had an impact on the distinctive ability of the monomeric ␥2-subunit to traffic to the surface when expressed in heterologous cells (35). Although this property is probably not of any biological significance, the defect in cell surface targeting may reflect the influence of the Arg-43 domain on the integrity of the overall subunit structure. The ␥2(R43) domain is also implicated in intersubunit contacts, as elegantly demonstrated using a 15-residue peptide surrounding Arg-43 (9), in agreement with the current homology molecular model. This model localizes ␥2(R43) at the interface between ␥2and ␤2-subunits (30), where an ionic interaction takes place involving Arg-43, the conserved ␥2(E178), and the adjacent ␤2(R117). When ␥2(R43) was replaced by an alanine or glutamine residue, the ␥2-subunit was barely detectable on the cell surface. Substituting the lysine for glutamine partially corrected the targeting defect of the ␥2(R43Q) mutant. However, in the ␥2(R43K) mutant, receptor cell surface targeting was only partially restored, showing that a lysine residue in this position cannot fully replace an arginine residue. Analysis of ␥2(D39A) and ␥2(P44A) mutants also highlighted the influence of the Arg-43 domain on ␥2-subunit cell surface targeting. Conversely, our findings that ␥2(K289M) mutation did not affect assembly and targeting of the GABA A receptor complex are complementary to previous studies indicating that this substitution modified GABA A receptor kinetics (10).
␥2(R43Q) Mutant Does Not Reduce ␣␤ Surface Targeting-Surprisingly, the various mutations of the ␥2(R43) domain, including R43Q substitution, did not lead to a significant reduction in ␣␤ surface targeting in our co-expression experiments using COS-7 cells, even with an ␣␤3␥2(R43Q) 1:1:8 ratio. These findings support previous data showing that GABA A receptor expression in heterologous cells with wild type or R43Q-␥2 gave rise to GABA-induced currents of similar amplitude (17). It is noteworthy that upon co-expression of ␣, ␤, and ␥2(R43Q), the kinetic properties of GABA A receptors resembled those of ␣␤ receptors (17). Our experiments are also in agreement with electrophysiological data reported recently (9). Taken together, our experiments indicated that ␥2(R43Q) substitution prevented association with ␣ or, more probably, ␤, as this subunit is assumed to interact with the ␥2R43 domain. This has been reported for the ␥2(S171) residue, located in the vicinity of the Arg-43 domain in the three-dimensional structure of the GABA A receptor complex. Mutations in this position prevented association with ␣2and ␤1-subunits, resulting in intracellular retention of ␥2 and surface expression of ␣2␤1 oligomers when these subunits were co-expressed in human embryonic kidney 293 cells (33). On the contrary, our coimmunoprecipitation experiments demonstrated that no significantly different amounts of ␤3-subunit were associated with wild type or R43Q-␥2. This is in line with data showing that intersubunit contacts involve several domains along the polypeptide chains (41,42). Another surprising finding was that the amount of ␣3-subunit co-immunoprecipitated with ␥2 was higher with R43Q mutant than wild-type subunits. This finding is difficult to interpret, as there is currently no model describing the scenario driving the assembly of ligand-gated ion channels (28). Nevertheless, it points to the intriguing possibility that the conserved Arg-43 position is implicated in controlling the order of subunit assembly and/or stoichiometry.
Functional Consequences of ␥2(R43Q) Mutation-Our data suggested that ␥2(R43Q) substitution gave rise to subunit complexes with an altered assembly, leading to the targeting of ␣␤ oligomers of unknown stoichiometry. This may reconcile some of the controversial data, and, on this basis, we propose that the ␥2(R43Q) mutation modifies the subunit composition rather than the number of GABA A receptors at the neuron surface. Because the ␥2-subunit is implicated in synaptic localization of these receptors (43), this modification would lead to a decrease in the number of ␣␤␥2 receptors and an increase in ␣␤ complexes for heterozygous mutation with relevance to clinical syndromes. Thus, it is tempting to speculate that this would create an imbalance between phasic and tonic inhibition, resulting in important pathophysiological consequences for the physiology of neuronal networks (44,45). However, preliminary data in an heterozygous knock-in model or in transfected neurons showed that synaptic currents did not decrease (46,47). Interestingly, studies using ␥2 knock-out mice showed that ␥2 Ϫ/Ϫ animals had impaired GABA A receptors and died shortly after birth, unlike ␥2 ϩ/Ϫ mice that apparently had normal GABA A receptors and developed as wild types (48,49). However, a further analysis of heterozygous mice showed a limited reduction in synaptic GABA A receptor clusters in some brain areas with a parallel increase in, probably extrasynaptic, ␣␤ receptors. This subtle change was sufficient to result in an anxiety-like behavior (50). This suggests that a molecular, physiological, and behavioral comparison of ␥2 ϩ/Ϫ mice with heterozygous ␥2(R43Q) knock-in animals may provide information on the relationship between GABA A receptor subunit composition, physiology, and physiopathological states.
This work might also provide support for future study using neuronal preparations and knock-in animals (46,47). Indeed, preliminary data, using a ␥2(R43Q) knock-in mouse model, showed that, whereas mutant subunits failed to localize at the surface membrane, ␣1-subunits retained their ability to localize at the neuronal membrane. However, the clustered organization of ␣1-subunits was absent in ␥2(R43Q) knock-in mice, suggesting a loss of ability of this mutant ␥2-subunit to associate with GABA A receptors in neurons (46,47). Our findings will shed further light on this issue.