Charged residues in the beta2 subunit involved in GABAA receptor activation.

Fast synaptic inhibition in the mammalian central nervous system is mediated primarily via activation of the gamma-aminobutyric acid type A receptor (GABAA-R). Upon agonist binding, the receptor undergoes a structural transition from the closed to the open state. This transition, known as gating, is thought to be associated with a sequence of conformational changes originating at the agonist-binding site, ultimately resulting in opening of the channel. Using site-directed mutagenesis and several different GABAA-R agonists, we identified a number of highly conserved charged residues in the GABAA-R beta2 subunit that appear to be involved in receptor activation. We then used charge reversal double mutants and disulfide trapping to investigate the interactions between these flexible loops within the beta2 subunit. The results suggest that interactions between an acidic residue in loop 7 (Asp146) and a basic residue in pre-transmembrane domain-1 (Lys215) are involved in coupling agonist binding to channel gating.

Fast synaptic inhibition is a major determinant of network dynamics in the central nervous system (1). In the mammalian brain, this is mediated primarily via activation of the ␥-aminobutyric acid type A receptor (GABA A -R) 1 (2,3). Each GABA A -R is composed of five subunits arranged around a central ion-conducting pore (4), with each subunit consisting of a large intracellular domain, four transmembrane domains (TM1-TM4), and a larger N-terminal extracellular domain (2). Experimental evidence suggests that the GABA-binding site lies within an asymmetric pocket formed at the interface between the ␣ and ␤ receptor subunits (5)(6)(7)(8). The channel "gate" in the GABA A -R is believed to be formed by charged residues in the TM1-TM2 loop (9 -11). The binding of agonist triggers a complex structural transition that results in the opening of the gate, allowing ions to flow through the channel. The mechanisms by which this occurs remain poorly defined.
The nature of the coupling between binding and channel opening in this receptor family has been recently investigated in several laboratories. Interactions between charged residues in the flexible loops 2 and 7 in the extracellular domain and those in the short linker between the second and third transmembrane domains (TM2-3L) of the GABA A -R ␣ 1 subunit have been implicated in the process of receptor activation (12). In addition, a recent report suggests that the pre-TM1 region is critical for receptor activation in the closely related serotonin (5-HT 3 ) receptor (13). In this study, we used site-directed mutagenesis and a number of GABA A -R agonists of varying efficacies to examine the contribution of the corresponding flexible loops in the GABA A -R ␤ 2 subunit to receptor activation. Based on previous studies (12)(13)(14) and the presence of highly conserved charged residues (see Fig. 1), we hypothesized that interactions between charged residues in these domains are crucial for coupling agonist binding to channel gating. We tested this hypothesis using a charge reversal double mutant approach, by inverting the hypothetical interacting pair of residues (e.g. Asp 146 -Lys 215 to Lys 146 -Asp 215 ), and then assaying receptor function. A strong restoration of function relative to either of the single mutant receptors is consistent with a strong interaction between the charged residues.

EXPERIMENTAL PROCEDURES
Mutagenesis-Full-length cDNAs encoding the GABA A -R ␣ 1 (human), ␥ 2 (human), and ␤ 2 (rat) subunits were expressed via the pCIS2 vector, which contains one copy of the strong promoter from cytomegalovirus and a polyadenylation sequence from simian virus 40. Point mutations in the GABA A -R ␤ 2 subunit were created using the QuikChange kit (Stratagene) and confirmed as described previously (12). 2 GABA A -Rs were expressed in HEK293 cells, maintained, and transfected as described (15).
Electrophysiology and Data Analysis-Recordings were made 48 -72 h after transfection at room temperature (20 -22°C) using the wholecell patch-clamp technique (15). Concentration-response curves were determined for GABA in wild-type and mutant GABA A -Rs as described (12). The relative efficacies (⑀) of piperidine 4-sulfonate (P4S) and taurine were defined as ⑀ ϭ I max(partial) /I max(GABA) , where I max(partial) is the maximal current elicited by a saturating concentration of either P4S or taurine, and I max(GABA) is the maximal current elicited by a saturating concentration of GABA. All data are reported as means Ϯ S.E. Statistical significance was assessed using either Student's two-tailed unpaired t test or, where appropriate, one-way analysis of variance (ANOVA) with the appropriate post-test. Cross-linking experiments were performed essentially as described previously (16). Cu:phen solutions were prepared by diluting stock solutions of CuSO 4 and 1,10phenanthroline (Sigma) to 100 and 400 M, respectively, in buffer. In experiments in which dithiothreitol (DTT) was used, a 10 mM solution was made fresh daily in buffer. When GABA was co-applied with Cu:phen, a saturating GABA concentration was used to ensure that all of the channels were in the activated state during the period of reagent application. The percent modulation was ((I GABA(after) Ϫ I GABA(before) )/ (I GABA(before) )) ϫ 100, where I GABA(before) is the peak current of the initial GABA applications, and I GABA(after) is the peak current of the GABA test pulses after the application of reagent.
Molecular Modeling-The molecular model of a GABA A -R ␤ 2 subunit * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: GABA A -R, ␥-aminobutyric acid type A receptor; TM, transmembrane domain; TM2-3L, TM2-TM3 linker; P4S, piperidine 4-sulfonate; ANOVA, analysis of variance; Cu:phen, CuSO 4 : 1,10-phenanthroline; DTT, dithiothreitol; AChBP, acetylcholine-binding protein; nAChR, nicotinic acetylcholine receptor. 2 Mutagenic primer sequences are available upon request.
was built using a previously described homopentameric GABA A -R ␣ 1 subunit model as a template (12,17). The primary sequence of a GABA-R ␤ 2 subunit was threaded onto the backbone coordinates of one of the ␣ subunits in a model of the GABA A -R ␣ 1 homopentamer using the Homology module of Insight II (Version 2000.1, Accelrys, San Diego, CA). Short loops were generated where gaps occurred in the alignment between the two sequences. The ␤ 2 subunit was optimized in the presence of the remaining ␣ 1 subunits to preserve appropriate intersubunit contacts. We refined the entire model using the autorotomer feature in the Biopolymer module of Insight II. Backbone atoms (carbon, ␣-carbon, and nitrogen) of each residue were then tethered to their initial coordinates with a force constant of 100 kcal/A 2 ; coulombic interactions were calculated with the default cell multipole method with a constant dielectric of 1; and the structure was optimized with the Discover module of Insight II. The structure was relaxed by performing 5000 1-fs steps of molecular dynamics at 400 K and was then re-optimized with the Discover module to a derivative of 1 kcal/Å 2 .

Characterization of Mutations in the TM2-TM3
Linker of the GABA A -R ␤ 2 Subunit-The amino acid sequences of the GABA A -R ␤ subunits were aligned and compared with those of the GABA A -R ␣ 1 and ␣ 2 subunits and the acetylcholine-binding protein (AChBP) (Fig. 1). Previous reports have suggested that TM2-3L of the GABA A -R ␣ subunit is involved in receptor activation (12,18). To determine whether this domain in the GABA A -R ␤ 2 subunit plays a similar role, we created point mutations R269A, R269D, L272A, K274D, and K279D at positions corresponding to mutations in the GABA A -R ␣ subunit that are important for receptor function (12,18). The mutant receptors were then characterized by determining GABA concentration-response curves. A comparison of the GABA responses of the wild-type ␣ 1 ␤ 2 ␥ 2s GABA A -R ( Fig. 2A) and the mutant ␣ 1 ␤ 2 (R269A)␥ 2s GABA A -R (Fig. 2B) demonstrates the decreased sensitivity of the mutant receptor to GABA. The ␤ 2 (R269A) and ␤ 2 (K274D) mutations caused a significant reduction in GABA sensitivity, whereas the GABA sensitivity of the ␤ 2 (L272A), ␤ 2 (R269D), and ␤ 2 (K279D) mutations was indistinguishable from that of the wild-type receptor ( Fig. 2C and Table I).
Determination of Agonist Efficacy in Mutants-The potency of the agonist (measured here as the EC 50 for GABA) can be influenced at either step in the simplified receptor activation pathway: agonist binding or isomerization from the closed to open state (18,19). With high efficacy agonists (E Ͼ 10), slight reductions in the ability of the agonist-bound receptor to isomerize from the closed to open state typically produce a reduction in sensitivity to agonist, but little reduction in maximal response. This can create a problem in the interpretation of the effects of receptor mutations, as the effects of modest changes in efficacy can closely resemble the effects of mutations that alter agonist binding affinity (20). The use of low efficacy agonists ("partial agonists") circumvents this problem. Mutations that alter agonist efficacy will reduce the maximal current elicited by a saturating concentration of partial agonist (20). We used the low efficacy GABA A -R agonist P4S (7,11,17) to investigate the effects of these mutations. A comparison of the saturating P4S responses of the wild-type ␣ 1 ␤ 2 ␥ 2s GABA A -R ( Fig. 2D) and the mutant ␣ 1 ␤ 2 (R269A)␥ 2s GABA A -R ( Fig. 2E) in HEK293 cells demonstrates the reduction in the relative efficacy (⑀) of P4S for the mutant receptor. The ␤ 2 (R269A) and ␤ 2 (K274D) mutations caused a significant reduction in P4S relative efficacy, whereas the relative efficacy of P4S for the ␤ 2 (L272A), ␤ 2 (R269D), and ␤ 2 (K279D) mutations was indistinguishable from that for the wild-type receptor (Fig. 2F).
Characterization of Mutations in the Pre-TM1 Region of the GABA A -R ␤ 2 Subunit-A previous report has suggested that a cluster of conserved basic residues in the pre-TM1 region of the 5-HT 3 receptor are involved in receptor activation (13). To determine whether these residues have a similar role in the GABA A -R ␤ 2 subunit, we created the following ␤ 2 subunit mutations: K215D, R216D, and N217D. GABA sensitivity was significantly reduced by the ␤ 2 (K215D) mutation and was unchanged by the ␤ 2 (R216D) and ␤ 2 (N217D) mutations (Table I). The ␤ 2 (K215D) and ␤ 2 (R216D) mutations caused a significant reduction in P4S relative efficacy, whereas the ␤ 2 (N217D) mutation had no effect (Fig. 3A). Taurine relative efficacy was significantly reduced in the ␤ 2 (K215D) mutation, but was unchanged in the ␤ 2 (R216D) and ␤ 2 (N217D) mutations (Fig. 3B).
Characterization of Charge Reversal Double Mutants-Using charge reversal double mutants, we previously demonstrated that specific residues in loops 2 and 7 interact with TM2-3L of the GABA A -R ␣ 1 subunit (12). To determine whether similar interactions might occur in the GABA A -R ␤ 2 subunit, we created and characterized the following double mutants: E52K,K274D, D56K,K274D, D139K,K274D, D146K,K274D, and E147K,K274D. All of these double mutant receptors were less sensitive to GABA compared with the wild-type receptor (Table II). We also created and characterized a second series of double mutants in which charges were exchanged between loops 2 and 7 and the pre-TM1 region in the GABA A -R ␤ 2 subunit. In several of these double mutant receptors, notably ␤ 2 (D146K,K215D), the sensitivity to GABA was substantially FIG. 1. Amino acid sequence alignment of GABA A -R ␤ subunits and AChBP. The amino acid sequences of rat GABA A -R ␤ subunits are aligned with that of AChBP. The sequences shown represent four regions: loops 2 and 7, the pre-TM1 region, and TM2-3L. Note that AChBP contains neither transmembrane nor TM2-3L domains. Acidic and basic residues located within these regions are colored in red and blue, respectively. The amino acid residues mutated in this study are labeled with asterisks. The nomenclature for the loops was adopted from Brejc et al. (27). Numbering corresponds to the mature GABA A -R subunit sequences. The sequences of the human GABA A -R ␣ 1 and ␣ 2 subunits are included for comparison. Underlined residues indicate charged residues implicated in the coupling of agonist binding to channel gating. restored, approaching the EC 50 for the wild-type receptor, and significantly improved relative to the corresponding single mutants (Table II). In addition, the relative efficacy of P4S for the ␤ 2 (D146K,K215D) double mutant receptor was enhanced relative to that for ␤ 2 (D146K) (Fig. 4B).
Disulfide Trapping Experiments-To examine the relative proximity and mobility of the domains investigated in this study, a disulfide bond trapping technique was utilized (16,21,22). We analyzed the interaction between cysteine residues inserted at potential contact points by examining the effects of oxidation and reduction on the function of wild-type and double mutant receptors. The ability to induce disulfide bond formation was assayed by studying the effect of the oxidizing reagent Cu:phen on receptor function. In the wild-type receptor, application of Cu:phen or the reducing agent DTT had no effect on receptor function in the absence or presence of GABA (Fig. 5A). Similarly, the single mutant receptors incorporating D56C, D139C, D146C, K215C, or K274C were also unaffected by FIG. 2. Mutations in the TM2-TM3 linker of the GABA A -R ␤ 2 subunit reduce both sensitivity to GABA and the relative efficacy of the partial agonist P4S. A and B, representative recordings of GABA responses from HEK293 cells transfected with the ␣ 1 ␤ 2 ␥ 2s and ␣ 1 ␤ 2 (R269A)␥ 2s GABA A -Rs, respectively. C, GABA concentration-response curves for the ␣ 1 ␤ 2 ␥ 2s (q), ␣ 1 ␤ 2 (L272A)␥ 2s (), ␣ 1 ␤ 2 (R269A)␥ 2s (ࡗ), and ␣ 1 ␤ 2 (K274D)␥ 2s (OE) GABA A -Rs. D and E, representative recordings of P4S responses from HEK293 cells transfected with the ␣ 1 ␤ 2 ␥ 2s and ␣ 1 ␤ 2 (R269A)␥ 2s GABA A -Rs, respectively. The gray bars above the recordings indicate the periods of GABA application, and the numbers above each bar indicate the micromolar concentrations of GABA applied. The black bars above the recordings indicate the periods of P4S application, and the numbers above each bar indicate the micromolar concentrations of P4S applied. The calibration bars represent 200 pA and 10 s. F, bar graph denoting P4S efficacy for wild-type (WT) and mutant receptors as I max(P4S) /I max(GABA) . Values represent means of multiple cells (n ϭ 5-25). *, p Ͻ 0.01 (values that are significantly different from the wild-type receptor value calculated using one-way ANOVA with Dunnett's post hoc test). Cu:phen or DTT (data not shown). Application of Cu:phen alone and in the presence of GABA significantly inhibited receptor function in all of the loop 2 and 7 and TM2-3L double cysteine mutants (Fig. 5D). This effect was reversed following incubation with 10 mM DTT (Fig. 5B). In double cysteine mutants between loops 2 and 7 and the pre-TM1 region, Cu:phen alone had no effect, whereas Cu:phen plus GABA significantly inhibited receptor function (Fig. 5, C and E). All of the mutants investigated in this study were unaffected by treatment with the reducing agent DTT (10 mM), indicating there were no spontaneously formed disulfide bonds (data not shown).

TM2-TM3 Linker
Mutations-Recent work has begun to illuminate the mechanisms of receptor activation in the Cys loop ligand-gated ion channel superfamily. In particular, studies of inherited mutations in the glycine receptor (23) and the nicotinic acetylcholine receptor (nAChR) (24) first revealed the importance of the TM2-3L region. Naturally occurring mutations in TM2-3L have been reported to have strong inhibitory (23) or facilitatory (24) effects on receptor function that are unrelated to agonist binding, leading to the idea that this loop is intimately involved in the coupling of ligand binding to ion channel opening. In this study, we have identified two mutations in TM2-3L of the ␤ 2 subunit (R269A and K274D) that interfere with receptor function by reducing agonist efficacy. Furthermore, we investigated the role of other basic residues located in TM2-3L of the ␤ 2 subunit (Arg 269 and Lys 279 ) and found that only the basic residue at position 274 is sensitive to a charge reversal mutation, suggesting that it is uniquely involved in electrostatic interactions important for receptor activation. This study also confirms the importance of TM2-3L of the GABA A -R ␤ 2 subunit in receptor activation, as previously reported for the ␣ 1 , ␣ 2 , ␤ 2 , and ␤ 3 subunits (12,14,18,25). An interesting difference between the ␣ 1 and ␤ 2 subunits was observed here because ␤ 2 (L272A) failed to alter receptor function, whereas the ␣ 1 ortholog (L277A) reduces GABA sensitivity dramatically (EC 50 ϭ 278 M) (26).
Pre-TM1 Mutations-Of the three mutations in the pre-TM1 cluster of basic residues in the ␤ 2 subunit, only K215D had a significant effect on receptor function, producing a large decrease in GABA sensitivity. This mutation also reduced the relative efficacy of both taurine and P4S, suggesting that the major effect is a reduction in agonist efficacy. It is also worth noting that, although there are residues homologous to the pre-TM1 region in AChBP, they are not included in the structure because they are disordered (27). Our model (Fig. 5A) predicts that those residues should be in a ␤ strand, as an ␣ helix would be less than half as long and would not reach the top of TM1.
Loop 2 and 7 Mutations-Three charge reversal mutations in loops 2 (D56K) and 7 (D139K and D146K) inhibited receptor function, apparently by reducing agonist efficacy. These data confirm an early report suggesting that mutations in loop 7 of the ␤ 2 subunit interfere with gating/coupling rather than with the GABA-binding site (28). The role of these residues appears to be conserved across the receptor superfamily, as homologous mutations in the glycine receptor ␣ 1 subunit have similar ef-  fects on receptor function (29). We noted with interest that the E147K mutant had no effect on receptor function, whereas the neighboring residue Asp 146 appeared to be critical for receptor function. This result appears less surprising upon inspection of a model of the ␤ 2 subunit. The molecular model provides a plausible explanation for this, as the Asp 146 side chain appears to be positioned close to a group of positively charged residues, whereas the Glu 147 side chain is located on a hydrophilic surface of the subunit, facing the extracellular space (Fig. 6B).

Double Mutant Experiments Reveal Asp 146 -Lys 215
Interactions-After establishing that the charged residues located in these domains appeared to be involved in receptor activation, we used a charge reversal double mutant approach to examine the interactions between these domains. This approach has been used successfully to identify intramolecular interactions in transmembrane (30) and soluble (31) proteins. Our results with the D146K,K215D double mutant strongly suggested that an interaction between these residues is critical for optimal receptor activation. Although we did observe small differences in function between the D146K,K215D mutant and the wildtype GABA A -R, this is not surprising, as the reversal of a charged pair may alter the energetics of the interaction (30) and hence perturb the functional properties of the protein (31). FIG. 4. Select charge reversal double mutants in the GABA A -R ␤ 2 subunit restore sensitivity to GABA and the relative efficacy of the partial agonist P4S to levels similar to those of the wild-type receptor. A, GABA concentration-response curves for the ␣ 1 ␤ 2 ␥ 2s (E), ␣ 1 ␤ 2 (D146K)␥ 2s (f), ␣ 1 ␤ 2 (K215D)␥ 2s (OE), and ␣ 1 ␤ 2 (D146K,K215D)␥ 2s (q) GABA A -Rs. B, bar graph denoting P4S efficacy for wild-type (WT) and mutant receptors as I max(P4S) /I max(GABA) . Values represent means of multiple cells (n ϭ 6-25). *, p Ͻ 0.05; **, p Ͻ 0.001 (values that are significantly different from the wild-type receptor value calculated using one-way ANOVA with Bonferroni's multiple comparison post hoc test).

FIG. 5. Cross-linking experiments reveal the proximity of domains of interest.
A, a 1-min application of Cu:phen alone or Cu:phen plus 3 mM GABA does not alter the response to a 10 M pulse of GABA in the wild-type receptor. Black bars indicate the duration of GABA application; gray bars highlight currents that were recorded after a 1-min application of Cu:phen; striped bars highlight currents that were recorded after a 1-min application of Cu:phen plus 3 mM GABA; and checkered bars indicate currents that were recorded after application of DTT. Scale bars indicate 100 pA and 5 s on all traces. B, a 1-min application of Cu:phen alone inhibits the response to a 100 M test pulse of GABA in the D146C,K274C double mutant. This inhibition was reversed by application of 10 mM DTT. C, a 1-min application of Cu: phen plus 3 mM GABA inhibits the response to a 30 M test pulse of GABA in the D56C,K215C double mutant. This inhibition was reversed by application of 10 mM DTT. D, bar graph denoting changes in receptor function in the loop 2 and 7 and TM2-3L double cysteine mutants measured by response to a submaximal GABA test pulse following treatment with oxidizing reagent alone (Cu:phen) or oxidizing reagent plus maximal agonist (Cu:phen plus 3 mM GABA). Values represent means of multiple cells (n ϭ 6 -15). *, p Ͻ 0.05; **, p Ͻ 0.01 (values that are significantly different from the corresponding wild-type (WT) treatment calculated using one-way ANOVA with Dunnett's multiple comparison post hoc test). E, bar graph denoting changes in receptor function in the loop 2 and 7 and pre-TM1 double cysteine mutants measured by response to a submaximal GABA test pulse following treatment with oxidizing reagent alone (Cu:phen) or oxidizing reagent plus maximal agonist (Cu:phen plus 3 mM GABA). Values represent means of multiple cells (n ϭ 6 -13). *, p Ͻ 0.05; **, p Ͻ 0.01 (values that are significantly different from the corresponding wild-type treatment calculated using one-way ANOVA with Dunnett's multiple comparison post hoc test).
To determine the proximity of the domains investigated in this study, we performed engineered disulfide trapping experiments, the results of which indicate that loops 2 and 7 are located in close proximity to TM2-3L, independent of the presence of agonist. One potential explanation is that interactions between loops 2 and 7 and TM2-3L are involved in stabilizing the closed state of the receptor. By contrast, loops 2 and 7 appear to be located within close proximity of the pre-TM1 region only in the presence of GABA, suggesting that these domains move relative to one another during the open or desensitized state. It is of interest to note that a recent study that examined state-selective incorporation of hydrophobic probes in the nAChR identified loop 7 of the nAChR ␣ 1 subunit as a region that undergoes a conformational change during receptor activation (32). Based on these results, we propose that there are intramolecular interactions between loop 7 and the pre-TM1 region that are important for receptor activation. This idea is structurally reasonable based on our molecular model (Fig. 6A), which incorporates both the extracellular and transmembrane domains of the GABA A -R. We also note with interest the high degree of similarity between the molecular model we present here and the recently published structure of the nAChR (33).
Functional Asymmetry between ␣ and ␤ Subunits-An intriguing result was the failure of charge reversal double mutants in the ␤ 2 subunit between loops 2 and 7 and TM2-3L to restore receptor function. This finding contrasts with our previous data from studies of the ␣ 1 subunit, where we found evidence for coupling between orthologous residues in loops 2 and 7 (Asp 57 and Asp 149 ) and Lys 279 in TM2-3L (12). These differences in intradomain interactions between ␣ and ␤ subunits may reflect local dissimilarities in structure or an asymmetry in the propagation of conformational change within the individual subunits. A possible origin for such differences could be the asymmetric nature of the agonist-binding site itself, which is believed to constrict upon binding GABA (7). The constriction of this cleft between adjacent subunit polypeptides implies that the adjacent domains of the ␣ and ␤ subunits are essentially pulled in opposite directions, thereby resulting in an asymmetric change in subunit structure. This would be consistent with reports of asymmetric motion within the extracellular domains of nAChR subunits following receptor activation (34). Our latest results suggest that there may also be asymmetric conformational change in the extracellular domain following agonist binding in the GABA A -R and that, within the ␤ subunit, the structural changes initiated upon agonist binding are coupled to the transmembrane domain via interactions between loop 7 and the pre-TM1 region.