Gephyrin-mediated γ-Aminobutyric Acid Type A and Glycine Receptor Clustering Relies on a Common Binding Site*

Background: Gephyrin clusters inhibitory GABAA and glycine receptors at postsynaptic sites. Results: GABAA and glycine receptor binding to gephyrin is a mutually exclusive process relying on distantly related sequence motifs. Conclusion: Clustering of GABAA and glycine receptors is mediated by a shared binding site on gephyrin. Significance: Gephyrin-dependent synaptic clustering of chloride-permeable ligand channels at synaptic sites relies on an evolutionarily conserved mechanism. Gephyrin is the major protein determinant for the clustering of inhibitory neurotransmitter receptors. Earlier analyses revealed that gephyrin tightly binds to residues 398–410 of the glycine receptor β subunit (GlyR β) and, as demonstrated only recently, also interacts with GABAA receptors (GABAARs) containing the α1, α2, and α3 subunits. Here, we dissect the molecular basis underlying the interactions between gephyrin and GABAARs containing these α-subunits and compare them to the crystal structure of the gephyrin-GlyR β complex. Biophysical and biochemical assays revealed that, in contrast to its tight interaction with GlyR β, gephyrin only loosely interacts with GABAAR α2, whereas it has an intermediate affinity for the GABAAR α1 and α3 subunits. Despite the wide variation in affinities and the low overall sequence homology among the identified receptor subunits, competition assays confirmed the receptor-gephyrin interaction to be a mutually exclusive process. Selected gephyrin point mutants that critically weaken complex formation with GlyR β also abolished the GABAAR α1 and α3 interactions. Additionally, we identified a common binding motif with two conserved aromatic residues that are central for gephyrin binding. Consistent with the biochemical data, mutations of the corresponding residues within the cytoplasmic domain of α2 subunit-containing GABAARs attenuated clustering of these receptors at postsynaptic sites in hippocampal neurons. Taken together, our experiments provide key insights regarding similarities and differences in the complex formation between gephyrin and GABAARs compared with GlyRs and, hence, the accumulation of these receptors at postsynaptic sites.

Gephyrin was initially discovered by co-purification with glycine receptors (GlyRs) 2 (1) and was found to accumulate these receptors at postsynaptic sites by simultaneous binding of the GlyR ␤ subunit (2)(3)(4) and elements of the cytoskeleton (5,6). Gephyrin is a modular protein composed of an N-terminal domain (GephG, residues 1-181) followed by a presumably unstructured linker (residues 182-317) and a C-terminal domain (GephE, residues 318 -736) that mediates interactions with the GlyR. The crystal structure of GephE in complex with a peptide derived from the large cytoplasmic loop between transmembrane helices 3 and 4 (TM3-4) of the GlyR ␤ subunit (7) defined the interactions in atomic detail. GephE forms a dimer, and residues 398 -410 of the GlyR ␤ subunit engage in critical interactions at the dimer interface; however, they primarily interact with only one monomer. Specifically, the interaction of Phe-330 and Asp-327 in GephE with Phe-398, Ser-399, and Ile-400 of the GlyR ␤-subunit was found to be a major contributor to the overall binding strength. Concurrently the F398A mutant significantly weakened the gephyrin-GlyR ␤ complex as verified by pulldown assays, isothermal titration calorimetry (ITC), and transient expression in HEK 293 cells (7).
knock-out mice demonstrated a differential subcellular effect on gephyrin and GABA A R ␣1 clustering (21). Additionally, both GlyRs and GABA A Rs can co-exist within single postsynaptic densities. Mixed glycinergic/GABAergic inhibitory synapses have been functionally identified in motoneurons of the hypoglossal nucleus (22) and abducens nucleus (23), in interneurons of the spinal cord (24,25), and the lateral superior olive (26). Which developmental role mixed inhibitory synapses play and how receptor co-clustering is mediated on the molecular level remains to be determined. Recent studies demonstrated direct interactions of GABA A R ␣1, ␣2, and ␣3 subunits with gephyrin (27)(28)(29), possibly explaining the mutual dependence of GABA A Rs and gephyrin and GlyRs for postsynaptic accumulation. Interestingly, alanine-scanning mutagenesis identified residues 325-334 of GephE to be essential for GABA A R ␣3 (28) and residues 325-343 to be essential for GABA A R ␣2 (30) interaction and hence suggest overlapping binding sites for these GABA A R subtypes and the GlyR ␤-subunit. Here, we study in detail how the GABA A R ␣1, ␣2, and ␣3 subunits mediate the proposed gephyrin interactions at the molecular level and compare it to the gephyrin-GlyR ␤ interaction.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Gephyrin variants were expressed and purified as described earlier (7). The large cytoplasmic loops between TM3-4 of the GABA A R ␣1, ␣2, and ␣3 subunits were PCR-amplified and inserted into the NcoI/NotI sites of the pETM-11 vector. Sequence numbers given in the text refer to the mature receptors without the signal sequence. Mutations were generated with the QuikChange site-directed mutagenesis kit (Stratagene). GABA A R ␣1, ␣2, and ␣3 loop variants were expressed in Escherichia coli BL21 cells (Stratagene, CA) as His 6 -tagged proteins. Cells were grown in LB medium at 30°C, and protein expression was induced after the addition of 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside for 18 h at 30°C. Cells were harvested by centrifugation (4000 ϫ g), resuspended in lysis buffer (50 mM Tris/HCl, pH 8.0, 500 mM NaCl, 5 mM ␤-mercaptoethanol), and passed through a cell disruptor (Constant Systems), and cell debris was removed by centrifugation (70,000 ϫ g). All proteins were initially purified using 5 ml of HisTrap FF crude columns according to the instructions of the manufacturer (GE Healthcare). Proteincontaining fractions were collected, concentrated, and applied to a 26/60 Superdex 200 size exclusion column (Amersham Biosciences) equilibrated with buffer (10 mM Tris/HCl, pH 8.0, 250 mM NaCl, 1 mM ␤-mercaptoethanol). Pure fractions were pooled, concentrated to 1-100 mg/ml, flash-frozen in 0.5 ml aliquots, and stored at Ϫ80°C.
Isothermal Titration Calorimetry-Before all ITC experiments, protein samples were extensively dialyzed against 10 mM Tris-HCl, pH 8.0, 250 mM NaCl, 1 mM ␤-mercaptoethanol at 4°C overnight followed by filtration and degassing. To analyze the gephyrin-receptor subtype interaction, 300 l of a solution containing 180 -1640 M concentrations of the recombinant GABA A R ␣1, ␣2, and ␣3 subtype intracellular loops, the respective variants, or the synthesized GlyR ␤-subunit-derived peptide (DFSIVGSLPRDFEL, Genscript) were titrated as the ligand into the 1.5-ml sample cell containing 9 -100 M GephE, GephG with linker, or full-length gephyrin (P2 splice variant). In each experiment a volume of 10 -15 l of ligand was added at a time with a total number of 20 -30 injections, resulting in a final molar ratio of ligand to protein of 4.5:1. All experiments were performed using a VP-ITC instrument (MicroCal, Northampton, MA) at 25°C. Buffer-to-buffer titrations were carried out as described above, so that the heat produced by injection, mixing, and dilution could be subtracted before curve-fitting. The binding enthalpy was directly measured, whereas the dissociation constant (K D ) and stoichiometry (N) were obtained by data analysis using the ORIGIN software, assuming a single site binding model. Binding parameters from singly performed measurements are given with standard derivations resulting from curve-fitting in ORIGIN. Measurements conducted several times are given as mean values and the resulting standard deviation (supplemental Table S2).
Native Gel Electrophoresis-Protein samples were mixed with OrangeG dye (Carl Roth) and 10% glycerol. 5-l samples containing 5-100 M concentrations of the respective protein were applied to 0.8% NEE0 ultra quality agarose (Carl Roth) gels containing 50 mM Tris/glycine, pH 8.4, buffer. Electrophoresis was conducted at 4°C and terminated when the dye front was leaving the gel. Gels were stained for 10 min in 5% CH 3 COOH, 10% C 2 H 5 OH, and 0.005% Coomassie Brilliant Blue R250 (Carl Roth) and destained at least 1 day in 5% CH 3 COOH and 10% C 2 H 5 OH.
Pulldown Assays-N-terminally biotinylated GABA A R ␣1 (KNNTYAPTATSYTPN)-, ␣3 (KNTTFNIVGTTYPIN)-, and GlyR ␤ (FSIVGSLPRDFEL)-derived peptides were synthesized by PANATecs. Because of excessive hydrophobicity, the corresponding GABA A R ␣2-derived peptide (QNNAYAVAVANY-APN) could not be synthesized. The biotinylated peptides were coupled to streptavidin beads and incubated with GephE (50 M) in 10 mM Tris/HCl, pH 8.0, 250 mM NaCl, and 1 mM ␤-mercaptoethanol for 1 h. After three washing steps with the same buffer, the beads were boiled with Laemmli buffer containing 10% SDS, and the supernatant was applied to an SDS-PAGE.
cDNA Constructs and Cell Culture-The full-length GABA A R ␣2 subunit (rat) used for the live imaging was modified with an N-terminal pHluorin as described previously (27). Double mutations were introduced using sequential PCR with appropriate sets of primers. Hippocampal neurons were prepared and nucleofected as described previously from E18 rat embryos of either sex (31,32). All imaging experiments were performed using 18 -21 days in vitro hippocampal cultures.
Antibodies, Immunocytochemistry, and Image Analysis-The GFP antibody (mouse monoclonal) was purchased from Roche Diagnostics, and the VIAAT (rabbit polyclonal) antibody was a generous gift from Dr. Bruno Gasnier (CNRS).
For immunostaining, cultures were lightly fixed, incubated with anti-GFP antibody followed by permeabilization with 0.4% Triton X-100, and incubated with the VIAAT antibody and subsequently with fluorescently labeled secondary antibodies (Jackson ImmunoResearch, West Grove, PA).
Images were acquired using a Nikon Eclipse Ti series confocal microscope with a 60ϫ objective (NA ϭ 1.4) and analyzed using the MetaMorph software (Molecular Devices, Downing-ton, PA) as described earlier (29,32). A receptor cluster was defined as being ϳ0.5 m 2 in size and ϳ2-3-fold more intense than background diffuse fluorescence. Synaptic clusters were co-localized with or directly opposed to VIAAT staining. Analyses are based upon the number of cells (n) as indicated in the text from at least three independent cultures. Statistical analysis (unpaired t test) was performed using Kaleidagraph (Synergy Software, Reading, PA) software.

RESULTS
Comparative Binding Studies on the GABA A R ␣1, ␣2, and ␣3 Gephyrin Interactions-After the initial observation that gephyrin directly interacts with ␣2 subunit-containing GABA A Rs (27), an interaction of gephyrin with the ␣1 (29) and ␣3 subunits (28) has been described involving the large cytoplasmic TM3-4 loop.
We cloned, expressed, and purified the corresponding loops of the GABA A R ␣1, ␣2, and ␣3 subunits and investigated their complex formation with GephE on native gels. Unlike GABA A R ␣1 and ␣3, which fully shifted GephE already at a molar ratio of 3 to 1 (Fig. 1A), only a 10-fold stoichiometric excess of GABA A R ␣2 was sufficient to complex all GephE under the experimental conditions ( Fig. 1B), suggesting a significantly lower affinity of ␣2 for gephyrin compared with ␣1 and ␣3. To analyze gephyrin interactions with GABA A R ␣1, ␣2, and ␣3 in more detail, we determined their binding parameters under similar experimental conditions by ITC (Fig. 1C). The GABA A R ␣3 subunit intracellular loop displayed the strongest interaction with GephE (K D ϭ 5.3 Ϯ 1.5 M) followed by the GABA A R ␣1 subunit intracellular loop (K D ϭ 17 Ϯ 11 M). Under the same experimental conditions, an interaction between the ␣2 subunit of GABA A R and gephyrin could not be detected by ITC. This finding is in line with our native gel experiments, which together suggest that GABA A R ␣2 and gephyrin form a rather loose complex.
Binding Site Mapping of the GABA A R ␣1 and ␣3 and GlyR ␤ Gephyrin Interaction-The binding affinities between fulllength gephyrin, different gephyrin domains, and selected GABA A R ␣1 and ␣3 variants were further investigated via ITC ( Fig. 1) and pulldown assays (supplemental Fig. S1) as summarized in Table 1. Similar to what we observed earlier for the interaction between the GlyR ␤ subunit and gephyrin, titration of full-length gephyrin with GABA A R ␣1 yielded similar binding parameters as the titration of GephE alone (Fig. 1C). Conversely, titration with GephG and the largest part of the linker region showed no detectable interaction (data not shown), thus confirming that the binding site resides in the E domain (Table  1A).
After our deletion variant mapping we studied binding of the identified minimum peptides of ␣1 (residues 334 -348, Table  1B), and ␣3 (residues 364 -378, Table 1D), which were determined to be crucial for binding and exhibit only limited homology. ITC analysis of the GlyR ␤-derived peptide containing just the critical 14 amino acids visible in the crystal structure (referred to as GlyR short) displayed a significantly lower binding affinity (K D ϭ 4.9 Ϯ 0.4 M) ( Fig. 1C and Table 2) compared with the 49-residue construct (referred to as GlyR long) used for the structural studies (7). GlyR long, in addition to binding more tightly, also necessitated the use of a two-site binding model for data analysis resulting in dissociation constants of 0.14 Ϯ 0.1 and 7.7 Ϯ 0.1 M for the high and low affinity binding sites, respectively ( Table 2). Similar to the GlyR ␤ loop, shortening of the GABA A R ␣1 and ␣3 loops lowered their affinity significantly; hence, the ␣1and ␣3-derived minimum peptides could not be studied by ITC (data not shown). Instead we performed streptavidin pulldown experiments with the corresponding peptides covalently linked to biotin (supplemental Fig. S1). The pulldown data demonstrated that the residues contained within these peptides are sufficient to mediate a specific interaction with GephE with their relative binding strengths mirroring the full-length loops (supplemental Fig. S1 and Table 1, B and D). Complex formation between GephE and the GABA A R ␣1, ␣2, and ␣3 subunits. A, native gel electrophoresis of GABA A R ␣1, ␣2, and ␣3 subunits applied in a 3-fold stoichiometric excess over GephE. Whereas GABA A R ␣1 and GABA A R ␣3 retain GephE in the gel pocket, incomplete GephE GABA A R ␣2 complex formation allows GephE (arrowhead) to enter the gel. B, native gel electrophoresis of GABA A R ␣1-3 at a 10-fold stoichiometric excess over GephE. This molar ratio does not allow any GephE (arrowhead) to enter the gel and migrate toward the anode. The migration of the receptor subunits in the absence of GephE (control) is shown below. C, ITC derived binding isotherms of GephE titrated with GABA A R ␣1 307-393 (TM3-TM4), GABA A R ␣2 307-391 (TM3-TM4), GABA A R ␣3 360 -457 (TM3-TM4), GlyR ␤ 398 -411, and full-length gephyrin titrated with GABA A R ␣1 307-393 (TM3-TM4) are shown. Measured binding enthalpies are plotted as a function of the molar ratio of the respective receptor loop to gephyrin. Data of the titration of GABA A R ␣2 to GephE could not be analyzed due to very weak binding. Titration of GephE and full-length gephyrin with GABA A R ␣1 yields very similar binding isotherms.
Binding of Gephyrin to GABA A R ␣1 and ␣3, and GlyR ␤ Is Mutually Exclusive-On the basis of the mapping experiments described here and published recently (28,29), our binding studies identified the core motifs in the GABA A R ␣1 and ␣3 and GlyR ␤ subunits, which directly interact with GephE. The wide range of affinities and the moderate overall homology observed here suggested different binding mechanisms for the respective receptor subunits. Concurrently recent studies (28, 30) did not identify individual residues within the core motifs that play a key role in binding to gephyrin. To address this issue, we conducted ITC competition experiments between GABA A R ␣1 and GlyR ␤ (Fig. 2), which exhibit the lowest overall identity in their gephyrin-interacting region. Should GABA A R ␣1, ␣2, and ␣3 and GlyR ␤ share a single binding site on gephyrin, we expected receptor binding by gephyrin to occur in a mutually exclusive fashion. Strikingly, our experiments revealed that pre-equilibration of gephyrin with a GlyR ␤-derived peptide containing only the 14-residue core motif was sufficient to significantly weaken and ultimately abolish the GABA A R ␣1 gephyrin interaction depending on the molar ratio of both loops (Fig. 2). This experiment can be rationalized as follows; upon pre-equilibration of GephE with GlyR ␤ (Fig. 2, B-D) fewer unoccupied GABA A R ␣1 binding sites are available. Although the GABA A R ␣1 loop at high concentrations will eventually replace the GlyR ␤ peptide, this will not be accompanied by a significant heat release because both binding events are exothermic reactions (Fig. 1C) and, hence, cancel each other out. Accordingly, the characteristic heat signature of the GABA A R ␣1 GephE titration ( Fig. 2A) is significantly weakened upon applying a 3-fold molar excess of GlyR ␤ peptide (Fig. 2D). Vice versa GABA A R ␣1 also interferes with GlyR ␤ binding to GephE (supplemental Fig. S2), and taken together this indicates that GABA A R ␣1 and GlyR ␤ compete for a single binding site located in the gephyrin E domain.
Comparison of the Identified GABA A R ␣1, ␣2, and ␣3 and GlyR ␤ Motifs Reveals Common Elements-On the basis of the finding of exclusive binding of the receptors to gephyrin, we   1C) and earlier (28,29) together with native gel (Fig. 1, A and B) and pulldown assays (supplemental Fig. S1) data. ϩϩϩ, K D ϭ 0.1-1 M; ϩϩ, K D ϭ 1-20 M; ϩ, no interaction detectable in ITC but verified via native PAGE or pulldown assays; Ϫ, no interaction detectable via ITC, native PAGE, or pulldown assays.  (Fig. 4, C and D) and the GlyR ␤ loop affinities determined earlier (7). Underlined residues were mutated. investigated the GABA A R ␣1, ␣2, and ␣3 binding motifs and also the gephyrin-interacting region of the GlyR ␤ subunit for common sequence features. Each gephyrin subunit is composed of four subdomains (I-IV) of which subdomains III and IV are important for the interaction with the GlyR (Fig. 3, A and  B). Several elements, summarized in Fig. 3C, appear noteworthy; our earlier structural analysis defined that Phe-398 of the GlyR ␤ subunit is of major importance for the interaction with gephyrin by forming hydrophobic interactions involving Phe-330 in subdomain III of GephE (Fig. 3A). Strikingly, the corresponding residues Phe-397 in GABA A R ␣3 and Tyr-339 in GABA A R ␣1 and ␣2 appear to be strongly conserved aromatic residues (a in Fig. 3C), thus suggesting a similar role for the respective residues. Adjacent to this type-conserved aromatic residue, the GephE-GlyR ␤ co-crystal structure displays a hydrogen bond-donating residue, Ser-399, in GlyR ␤ that interacts with the gephyrin Asp-327 followed by two hydrophobic residues, Ile-400 and Val-401, in GlyR ␤ that tightly fit into the hydrophobic pocket formed by subdomains III and IV of GephE (Fig. 3A). Noteworthy, the corresponding residues are strongly conserved in GABA A R ␣3 (Asn-370, Ile-371, and Val-372) and moderately in GABA A R ␣1 and ␣2 (Fig. 3C, b). Gly-402 in the GlyR ␤ loop points toward subdomain IV of the gephyrin E domain, and interestingly the corresponding residues in GABA A R ␣3 and ␣1 are also very small (Gly-373 and Ala-344) (Fig. 3C, c), whereas GABA A R ␣2 has a significantly bulkier residue at this position (Val-343). GABA A R ␣1, ␣2, and ␣3 subtypes share a conserved tyrosine positioned eight residues C-terminal of the first conserved aromatic residue (Fig. 3C, d), yet the corresponding residue in GlyR ␤ is a proline, thus suggesting that at least the C-terminal half of the GABA A R ␣1, ␣2, and ␣3 binding site relies on a different binding mechanism as compared with GlyR ␤. Similar Interactions Are Central to Complex Formation between GABA A R ␣1, ␣2, and ␣3 and GlyR ␤ with Gephyrin-Based on the mutually exclusive binding and the identified features shared among the receptor motifs, we assumed that GephE binding to the N-terminal motif of the GABA A R ␣1, ␣2, and ␣3 binding site is mechanistically very similar to the interaction of the homologous residues in the GlyR ␤ subunit. To test this and to explore the molecular basis of this phenomenon, we analyzed the effect on GABA A R ␣1, ␣2, and ␣3 binding of two critical gephyrin point mutants identified earlier. The mutation P713E in gephyrin was shown to completely abolish the tight GlyR ␤ binding (7), possibly by introducing its bulky and negatively charged side chain into the receptor binding pocket formed by subdomains III, IV, and IV (Fig. 3A). A less severe change is represented by the gephyrin mutant F330A, FIGURE 3. Sequence and structural determinants of gephyrin-receptor interactions. A, summary of binding interactions in the GephE-GlyR ␤ loop complex crystal structure. GephE is colored according to its subdomain architecture and is represented as a ribbon diagram with residues central for the GlyR ␤ loop binding shown as stick model (7). Residues of the GlyR ␤ loop are also displayed in stick representation with C-atoms in gray and residues critical for GephE interaction marked in red. B, subdomain architecture of the GephE dimer. Subdomains I and II of the first monomer (top) and subdomains I, II, and III of the second monomer (bottom) are not visible in this view and hence have not been assigned a color. C, sequence alignment of the different motifs recognized by gephyrin. Residues critical for the GephE-GlyR ␤ loop complex, which are strongly conserved in the GABA A R ␣3 subtype and moderately in ␣1 and ␣2, form a hydrophobic core motif marked in red. The tyrosine residue positioned in the C-terminal half of the motif (green) is conserved among the GABA A R subtypes but not in GlyR ␤. The binding sites are divided into an N-terminal region (motif 1) and a C-terminal part (motif 2). Elements a, b, and c are conserved among all receptor subtypes and together form motif 1. Motif 2 relies on a conserved tyrosine residue (element d), which is conserved among GABA A R subtypes but not in the GlyR ␤ subunit.

Receptor
which nevertheless significantly weakens GlyR ␤ binding, most likely by diminishing the contribution of the hydrophobic interactions mediated by the hydrophobic core of the GlyR ␤ loop and subdomain III of GephE. The binding of both mutants was analyzed by ITC for the GABA A R ␣1 and ␣3 subunits, and interestingly, both impaired binding so strongly that it could no longer be assessed by ITC (Fig. 4, A and B). The effect of the P713E variant supports the hypothesis that the binding site for the GlyR ␤ subunit overlaps with that of the GABA A R ␣1 and ␣3 subunits in the vicinity of Pro-713. The impaired binding of the F330A variant likewise demonstrates overlap of the different binding sites in the vicinity of this residue but also argues that interactions with the GABA A R ␣1 and ␣3 subunits are driven by a similar contribution of the hydrophobic core region (Fig. 3C, a and b). The corresponding residues interacting with Phe-330 are strongly conserved among the GABA A R ␣3 and GlyR ␤ subunits and to some extent also in GABA A R ␣1 and ␣2.
Following the idea of a similar binding mechanism, we identified several residues corresponding to GlyR ␤ residues that were identified earlier to be key contributors to the complex formation with GephE. Based on an alignment of the identified motifs, we predicted that Phe-369 of GABA A R ␣3 and Tyr-340 of GABA A R ␣1 are positioned in an analogous fashion as Phe-398 of GlyR ␤. In line with the Ͼ100-fold reduction in binding strength for GlyR ␤ F398A, binding was no longer detectable by ITC (Fig. 4, C and D) and native gel electrophoresis (Fig. 4, E and  F) for the GABA A R ␣3 F369A and GABA A R ␣1 Y340A variants. In an analogous fashion we proposed a similar role for GABA A R ␣3 Ile-371 as compared with GlyR ␤ Ile-400, which critically weakened the gephyrin interaction upon alanine mutation. Not surprisingly, the GABA A R ␣3 I371A mutant was also significantly impaired in interacting with GephE (Fig. 4D).
Puzzled by our finding of a very weak GABA A R ␣2 affinity for GephE despite the high homology to the GABA A R ␣1 and ␣3 subunits, we carefully compared the identified binding regions, also in light of the GephE-GlyR ␤ complex. This analysis focused our attention on Gly-374, which points toward subdomain IV of GephE in the GlyR ␤-GephE complex structure and is strictly conserved in GABA A R ␣3 (Gly-373) and typeconserved in GABA A R ␣1 (Ala-344) but not conserved in GABA A R ␣2, where it is replaced by a bulkier residue (Val-343) (Fig. 3C, c). In line with our suggestion of similar binding modes, we proposed this residue to repel the receptor loop by sterically clashing with subdomain IV of GephE (Fig. 3A), thus possibly explaining our observation of a rather weak GephE affinity as compared with GABA A R ␣1 and ␣3. We replaced the corresponding residue in GABA A R ␣3 with valine (G373V), and subsequent ITC analysis (Fig. 4D) revealed a significantly reduced affinity. Although we could not reconstitute binding to the level of the GABA A R ␣1 or ␣3 subtype, native gel electrophoresis of the corresponding GABA A R ␣2 mutant (V343G) visualizes an increased gephyrin affinity (supplemental Fig. S3). Taken together this indicates that this residue could at least be partially responsible for the observed weak GephE GABA A R ␣2 affinity.
Hydrophobic Interactions Mediated by Aromatic Residues Are Critical for GABA A R ␣1 and ␣3 GephE Complex Formation-After our mutational analysis we wanted to dissect the relative contributions to the overall binding strength of the conserved N-terminal hydrophobic motif (Fig. 3C, a and b) and the C-terminal parts (Fig. 3C, d) of the respective GABA A R ␣1 and ␣3 motifs. Given that ITC is not sensitive enough to display any residual binding for the respective mutants, we instead used native gel electrophoresis. As expected, the gel electrophoresis experiments verified a major impact on binding strength for all of the analyzed mutants (Fig. 4, E and F). In line with our ITC experiments, no residual binding could be detected for GephE F330A incubated with GABA A R ␣3. The Y376A variant of the GABA A R ␣3 subunit was also significantly impaired; however, it allowed for some residual binding to GephE. In the case of the GABA A R ␣1 subunit, the corresponding substitutions displayed a similarly weakened receptor binding. The simultaneous alanine substitutions of GephE Phe-330 and GABA A R ␣1 Tyr-347 synergistically weakened the interaction, yet some residual binding could still be detected. Taken together these results document major contributions of the hydrophobic motif at the N terminus (Fig. 3C, a and b) and the conserved aromatic residues for GABA A R ␣1 and ␣3 (Fig. 3C, d) but also suggest additional distinct contributions to the binding strength unique to the GABA A R ␣1 subunit.

Clustering of ␣2-containing GABA A Rs in Hippocampal Neurons Critically Depends on Two Conserved Aromatic Residues-
Our in vitro data obtained from the native gel electrophoresis and ITC experiments identified two conserved aromatic residues in the ␣1 and ␣3 subunits of GABA A Rs and GlyR ␤ subunit (a and d in Fig. 3C) to be critical for direct and exclusive binding to gephyrin. Earlier studies demonstrated that the region containing these two aromatic residues is crucial in vivo for gephyrin-mediated clustering of GABA A R subtypes ␣1 (29) ␣2 (27), ␣3 (28), and GlyR ␤ (4), and our earlier studies demonstrated that two aromatic residues (Phe-398 and Phe408 together with Ile-400 ) are critical for GlyR ␤ clustering by gephyrin in HEK 293 cells (7). Therefore, we tested whether the aromatic residues, Tyr-339 and Tyr-346, are also crucial for clustering of ␣2 containing GABA A Rs. We mutated both residues to alanine and compared (Fig. 5A) the synaptic accumulation of GABA A Rs between wild-type and the ␣2 subunit double mutant (DM), both modified with N-terminal pHluorin reporters ( pH ␣2 WT and pH ␣2 DM (Y339A/Y346A), respectively). Confocal imaging was performed on hippocampal neurons at ϳ21 days in vitro expressing pH ␣2 WT and pH ␣2 DM after cells were lightly fixed and stained with the respective antibodies. Surface staining with a GFP antibody revealed a significant loss of GABA A Rs clusters per 30 m (11.7 Ϯ 0.6 versus 4.6 Ϯ 0.4, n ϭ 30 neurons, pH ␣2 WT and pH ␣2 DM, respectively, unpaired t test; p Յ 0.001) (Fig. 5B). These clusters are primarily synaptic as they are either co-localized or opposed to VIAAT-positive puncta. Moreover, cells expressing mutant pH ␣2 subunit are positive for VIAAT puncta, suggesting that they are innervated by inhibitory presynaptic terminals.
To control for possible negative effects of mutations on receptor assembly or surface trafficking, we examined the ability of the pH ␣2 WT and pH ␣2 DM subunit to gain access to the plasma membrane on co-expression with the ␤3 subunit in HEK 293 cells. Transfected cells were surface-labeled (without permeabilization) with GFP antibodies. Fig. 5C shows that in the presence of the ␤3 subunit, both pH ␣2 WT and pH ␣2 DM subunits gain access to the surface membrane to a similar extent. This process seems specific as we observed minimal surface trafficking when expressing pH ␣2 subunit alone (data not shown). Collectively these experiments suggest that both tyrosine residues are critical for regulating the accumulation of GABA A Rs ␣2 and ␣1 subunits at inhibitory postsynaptic sites via a direct interaction with gephyrin.

DISCUSSION
Gephyrin has been implicated for quite some time in the anchoring of GABA A Rs; however, a direct interaction between this protein and specific receptor subunits has only been demonstrated recently for the ␣1, ␣2, and ␣3 subunits (27)(28)(29). Following up on these studies we investigated the interaction between the full-length intracellular TM 3-4 loops of the GABA A R ␣1, ␣2, and ␣3 subunits and the gephyrin E domain and observed that the affinities varied considerably. Moderately tight interactions that could be described by a one-site binding model were observed for the GABA A R ␣1 (K D ϭ 17 Ϯ 11 M) and ␣3 (K D ϭ 5.3 Ϯ 1.5 M) subunits, whereas the GephE interaction with GABA A R ␣2 was too weak to be analyzed by ITC.
Concurrently, the analysis via native PAGEs revealed the same order of relative binding strengths for the receptor loops, and this ranking is maintained at different pH values (pH 7-9) and salt concentrations ranging from 50 to 250 mM (data not shown). The measured moderate GABA A R ␣1 and ␣3 affinities match the binding constants determined for the interactions between the C-terminal tails of excitatory NMDA and AMPA receptors and PDZ (post synaptic density protein (PSD95)) domain-containing scaffolding proteins that play a critical role in the formation of the postsynaptic density at excitatory synapses and are characterized by K D values varying between 1 and 50 M (33).
In contrast, the much weaker gephyrin-GABA A R ␣2 affinity seems not sufficient for receptor anchoring and instead suggests that additional binding partners are involved, or that affinity is enhanced by posttranslational modifications. Interestingly, it was recently shown that GABA A R ␣2 but not ␣3 forms a tripartite complex with collybistin (30), a brain-specific Cdc42-activating GDP-GTP exchange factor that promotes submembrane clustering of gephyrin and is essential for the postsynaptic localization of gephyrin and GABA A Rs.
In contrast to the moderately tight single site interaction between GABA A R ␣1 and ␣3 subunits and gephyrin, earlier biochemical and biophysical analyses of GlyR ␤ residues 387-426, which cover the central part of the TM 3-4 loop, suggested two binding sites with K D values of 0.14 Ϯ 0.1 and 7.7 Ϯ 0.1 M, respectively (3). A subsequent co-crystal structure (7) revealed that residues 398 -411 of the GlyR ␤ subunit engage in close contacts with a binding site near the GephE dimer interface. In this study we also investigated the interaction between the core region (residues 398 -411) of the GlyR ␤ subunit and GephE by ITC and observed, in contrast to the 49 residue GlyR ␤ fragment, only a single binding site that displayed an affinity of 4.9 Ϯ 0.4 M. Because the triple alanine GlyR ␤ mutation Note that although pH ␣2-WT can form mostly synaptic clusters, as demonstrated by a predominant co-localization with VIAAT puncta (blue, merge), the ability to form clusters is greatly attenuated for pH ␣2-DM as cluster number and intensity are both strongly reduced. For each panel, a higher magnification image of the boxed area is shown on the right, with arrows pointing to clusters (scale bar ϭ 15 m). B, quantification of the average number of clusters formed by pH ␣2-WT and pH ␣2-DM. Clusters that are opposed to VIAAT-positive puncta were counted along dendrites per 30 m (n ϭ 30 cells each, unpaired t test, p Ͻ 0.001). C, to test whether attenuation of the clusters formed by pH ␣2-DM is not due to compromised assembly or surface trafficking, pH ␣2 and pH ␣2-DM were expressed along with the ␤3 subunit in HEK 293 cells. Cells were lightly fixed and surface-labeled with GFP antibody against extracellular pHluorin tag (red). Confocal images show that both pH ␣2-WT (left panel) and pH ␣2-DM (right panel) can access the surface membrane to a similar extent (scale bar ϭ 15 m).
F398A/I400A/F408A is sufficient to completely abrogate complex formation in vitro and in vivo (7), we conclude that GlyR ␤ residues 398 -411 represent the core binding site, whereas residues 387-397 and 412-426 contain elements that enhance the affinity of this core motif. In a comparable manner, applying only the homologous core motifs of either the GABA A R ␣1 or ␣3 subunits resulted in a decreased affinity as compared with the respective full-length loops. Our native gel assays, ITC, and cell culture experiments, however, provide evidence that mutation of the conserved aromatic residues (Fig. 3C, a and d) is sufficient to completely abrogate GephE binding and, hence, strongly argue against the existence of other major gephyrin binding determinants in the remaining parts of the receptor loop. How this increase in affinity for the full-length loops can be explained on the molecular level, and how it is connected to the observed second binding site in the case of the GlyR remain open questions. Possibly there are structural rearrangements in either GephE or the intracellular loops.
Our binding studies with deletion variants described here and published earlier (28,29) identified a moderately conserved core motif in GABA A R ␣1, ␣2, and ␣3 and GlyR ␤ that is essential for gephyrin binding. Recent yeast two-hybrid studies (28,30) suggested that both receptors bind to distinct, albeit partially overlapping binding sites; however, neither the role of individual residues in gephyrin-mediated receptor binding nor the question as to whether gephyrin can simultaneously bind to GABA A R ␣1, ␣2, and ␣3 and GlyR ␤ subunits has been addressed. Regarding the last issue, ITC experiments revealed competitive binding between GABA A R ␣1 and GlyR ␤. Taking into account that the GABA A R ␣1 and GlyR ␤ sequences display the lowest degree of homology among all analyzed subunits, this finding strongly suggest that, in addition to GABA A R ␣1 and GlyR ␤, GABA A R ␣2 and ␣3 also compete for the same binding site.
The mutually exclusive binding was confirmed by two GephE mutants (P713E and F330A) that abolished or weakened binding to the GABA A R ␣1 and ␣3 as well as GlyR ␤ subunits. Within the common binding pocket, GephE Phe-330 mediates conserved hydrophobic interactions with a conserved aromatic residue (Fig. 3C, a) This finding may present a molecular explanation for earlier studies that proposed GephE residues 325-334 to be essential for GABA A R ␣3 (28) and residues 325-343 to be essential for GABA A R ␣2 (30) binding. Disruption of the N-terminal helix of GephE (residues 326 -336) by introduction of poly-alanine blocks may have affected Phe-330 of GephE, which we have identified here to be a major determinant of GABA A R ␣1, ␣2, and ␣3 binding.
To verify whether a similar mechanism is also crucial for GABA A R ␣2 and gephyrin complex formation and, hence, its synaptic accumulation, we studied the respective alanine mutations in GABA A R ␣2 receptors by in vivo co-localization of the receptor and gephyrin in cultured hippocampal neurons. Significantly, compared with wild type, the clustering of the GABA A R ␣2 Y339A/Y346A double mutant revealed a significant loss of clustering in cultured hippocampal neuron. These findings support the relevance of our in vitro results for in vivo receptor clustering and demonstrate that our mutational studies can be extended to the GABA A R ␣2 subunit, which due to its weak binding to gephyrin, could not be analyzed in great detail in vitro. Additionally the major importance of the conserved aromatic residues presents a possible molecular explanation for earlier findings (29), which proposed a negative regulation of clustering upon GABA A R ␣1 Thr-348 phosphorylation. In vitro and in vivo studies verified a significantly weakened interaction for the phosphomimetic T348E mutation as compared with the wild type or the respective alanine mutant. Introduction of a negative charge at the adjacent residue apparently interfered with the critical hydrophobic interactions mediated by the conserved tyrosine (Tyr-347). Interestingly, earlier studies identified both GABA A R ␣1 tyrosines determined here as critical mediators of GephE binding to be phosphorylated in vivo and hence represent possible regulatory mechanism for the gephyrin GABA A R ␣1 interaction (34,35).
Collectively our in vitro and in vivo studies on the pleiotropic effects of the mutants define two different hotspots of GABA A R recognition by gephyrin. The first is the N-terminal half of the ϳ15-residue core binding site. This region relies on interactions that are highly similar to the GlyR ␤ subunit and involves hydrophobic contacts between the gephyrin Phe-330 and the respective aromatic residues in GABA A R ␣1, ␣2, and ␣3. This conserved interaction is also at least in part responsible for a mutually exclusive binding between the different GABA A R subtypes and the GlyR ␤ subunit. The second is the C-terminal half of the core region. This region displays a conserved tyrosine residue in the GABA A R ␣1, ␣2, and ␣3 subunits; however, it is distinct from the important second aromatic residue of the GlyR ␤ subunit. Further details of the gephyrin-GABA A R interaction, however, will have to await the structural characterization of a complex involving the E domain of gephyrin and a peptide derived from the ␣1, ␣2, or ␣3 subunits.