GABA A Receptor (cid:1) 2 Tyr 97 and Leu 99 Line the GABA-binding Site INSIGHTS INTO MECHANISMS OF AGONIST AND ANTAGONIST ACTIONS*

The identification of residues that line neurotransmit-ter-binding sites and catalyze allosteric transitions that result in channel gating is crucial for understanding ligand-gated ion channel function. In this study, we used the substituted cysteine accessibility method and two-electrode voltage clamp to identify novel (cid:2) -aminobu-tyric acid (GABA)-binding site residues and to elucidate the secondary structure of the Trp 92 -Asp 101 region of the (cid:1) 2 subunit. Each residue was mutated individually to cysteine and expressed with wild-type (cid:3) 1 subunits in Xenopus oocytes. GABA-gated currents ( I GABA ) were measured before and after exposure to the sulfhydryl reagent, N -biotinylaminoethyl methanethiosulfonate (MTS). V93C, D95C, Y97C, and L99C are accessible to derivatization. This pattern of accessibility is consistent with (cid:1) 2 Val 93 -Leu 99 adopting a (cid:1) -strand conformation. Both GABA and SR95531 protect Y97C and L99C from modification, indicating that these two residues line the GABA-binding site. In D95C-containing receptors, application of MTS in the presence of SR95531 causes a greater effect on I GABA than MTS alone, suggesting that binding of a competitive antagonist can cause movements in the binding site. In addition, we present evidence that (cid:1) 2 L99C homomers form spontaneously open

Central to the understanding of neurotransmission is the nature of allosteric transitions, such as the movements required for an apoprotein to grasp its ligand, or the movements that govern the shift from a closed to an open conformation in an ion channel protein.Among members of the superfamily of ligand-gated ion channels (LGICs 1 ), both allosteric transitions are required, and must be linked together in a coordinated fashion, for the ligand binding to effect channel gating.The association of the neurotransmitter with the receptor repre-sents the initial step in the induction of conformational movements that result in channel activation, deactivation, and desensitization.Therefore, an understanding of the structure of the ligand-binding site is necessary to comprehend these processes.
Multiple amino acid residues from both ␤ and ␣ subunits have been identified that are important determinants for GABA binding.A widely accepted model of ligand recognition is based on the nicotinic acetylcholine receptor (nAChR) (10).The ligand-binding site is formed at subunit-subunit interfaces with clusters of binding site residues (which are, by arbitrary convention, designated "loops" A-F) found in several distinct regions of neighboring subunits.In the GABA A R-binding site, these residues include: Phe 64 , Arg 66 , Ser 68 (loop D), Arg 119 and Ile 120 (loop E) of the ␣ 1 subunit (11)(12)(13)(14)(15) and Tyr 157 , Thr 160 (loop B), Thr 202 , Ser 204 , Tyr 205 , Arg 207 , and Ser 209 (loop C) of the ␤ 2 subunit (16,17).
To date, no residues in the region defined by Trp 92 -Asp 101 (loop A) of the ␤ 2 subunit have been implicated in the formation of the GABA-binding site.A residue in this domain, ␤ 2 Leu 99 , is found in the homologous position as His 101 of the ␣ 1 subunit, which participates in the formation of the GABA A R benzodiazepine-binding site (Fig. 1) (18 -20).In addition, ␤ 2 Leu 99 aligns with Tyr 93 of the nAChR ␣ subunit, which has been identified as forming part of the acetylcholine-binding site (21)(22)(23)(24).On the basis of LGIC-binding site homology, we reasoned that the ␤ 2 Trp 92 -Asp 101 region is likely to contribute to formation of part of the GABA-binding site.To test this hypothesis, we used the substituted cysteine accessibility method (SCAM) to identify potential binding site residues and examine the secondary structure of this domain.
The development of SCAM (25) has proved to be a powerful tool to gain information about the structure and dynamics of protein domains and has been utilized to study LGICs (26,27), G-protein coupled receptors (28), voltage-gated ion channels (29), and transporters (30).Specifically, it been used to identify amino acid residues that line ion channels (31,32) as well as residues that line binding sites (12,17,24,28,33,34).We have previously used this technique to 1) gain information about the secondary structure of benzodiazepine-binding and GABAbinding domains of specific GABA A R subunits (12,17,34), 2) identify residues that line these binding sites, and 3) examine binding site conformational dynamics (17,34).SCAM entails introducing cysteine residues at individual positions within the protein and testing the effect (if any) of modifying these introduced cysteines with sulfhydryl-specific reagents.We use two criteria to determine if a cysteine-substituted residue lines part of the binding site: (i) reaction with a sulfhydryl-specific reagent alters function and (ii) binding site ligands (agonists and antagonists) protect the engineered cysteine from covalent modification.
In this report, using SCAM analysis of the ␤ 2 Trp 92 -Asp 101 region of the GABA A R, we found that this domain adopts a ␤ strand conformation and identified two novel amino acid residues, Leu 99 and Tyr 97 , that participate in the formation of the GABA-binding site.These observations mark the initial structural description of this region of the GABA A R ␤ 2 subunit.Using the recently determined crystal structure of the related acetylcholine binding protein (35) as a template, we present a model of this domain of the GABA-binding site.Furthermore, we illustrate that the binding of a competitive antagonist, SR95531, triggers conformational movement near the binding site and that mutation of a binding-site residue (Leu 99 ) can perturb channel gating.

L99C.
Oocytes were prepared as previously described (41).GABA A receptor ␣ 1 and ␤ 2 or ␤ 2 mutant subunits were expressed by injection of cRNA (0.7-2.7 ng/subunit cRNA) in a 1:1 ratio (␣:␤) with the exception of L99C, which was injected with excess ␣ subunit cRNA in a 4:1 ratio (see "Results" and "Discussion").The maximal currents for the majority of the mutant receptors were similar to wild-type receptors (1-5 A).Expression of P94C and F98C produced receptors in which the peak current amplitudes were significantly reduced (Ͻ100 nA), even with 7-ng cRNA injections.The oocytes were maintained in modified ND96 recording solution (in mM: 96 NaCl, 2 KCl, 1 MgCl 2 , 1.8 CaCl 2 , and 5 HEPES, pH 7.4), that had been supplemented with 100 g/ml gentamicin and 100 g/ml bovine serum albumin.Oocytes were used 2-14 days after injection for electrophysiological recordings.
Two-electrode Voltage Clamp Analysis-Oocytes under two-electrode voltage clamp (V hold ϭ Ϫ80 mV) were perfused continuously with ND96 at a rate of ϳ5 ml/min.The volume of the recording chamber was 200 l.Standard two-electrode voltage clamp procedures were carried out using a GeneClamp500 Amplifier (Axon Instruments, Inc.).Borosilicate electrodes were filled with 3 M KCl and had resistances of 0.5-3.0M⍀ in ND96.Stock solutions of GABA (Sigma Chemical Co., St Louis, MO) and SR95531 (Sigma) were prepared in water, whereas a 100 mM stock of N-biotinylaminoethyl methanethiosulfonate (MTSEA-biotin, MTS hereafter; Biotium, Hayward, CA) was dissolved in Me 2 SO.The MTS molecule (14.5 Å, unreacted; 11.2 Å reacted moiety) is longer than GABA (4.5 Å) but similar in length to SR95531 (13.5 Å).Furthermore, MTS is a relatively membrane impermeant compound (42,43), allowing only for extracellular reactions.Taken together, these properties make MTS a reasonable choice for experimentation on water-soluble ligandbinding regions.All compounds were diluted appropriately in ND96 such that the final concentration of Me 2 SO in MTS solutions was Յ3%.This concentration of solvent did not affect recombinant GABA A R properties.
To measure the sensitivity to GABA, the agonist (0.0001-30 mM) was applied via gravity perfusion (ϳ5-8 s) or by pipette application (150 -200 l) with a 3-to 15-min washout period between each application to ensure complete recovery from desensitization.Peak GABA-activated current (I GABA ) was recorded at each application.To correct for slow drift in the maximum amplitude of the response, concentration-response data were collected by a prior application of a low, non-desensitizing concentration of GABA (EC 3 -EC 5 ) between which there was an approximate 20-s wash-out period and to which each test concentration of agonist was normalized.Concentration-response curves were generated for each recombinant receptor, and the data were fitted by nonlinear regression analysis using GraphPad Prism software (San Diego, CA, www.graphpad.com).Data were fitted to the equation: where I is the peak amplitude of the current for a given concentration of GABA ([A]), I max is the maximum amplitude of the current, EC 50 is the concentration required for half maximal receptor activation, and n is the Hill coefficient.
To measure the sensitivity to SR95531, GABA (EC 50 ) was applied via gravity perfusion followed by a brief (20 s) wash-out period before concomitant application of GABA (EC 50 ) and increasing concentrations of SR95531 by pipette application.The response to the application of SR95531 and GABA was normalized to the response elicited by the agonist alone.Concentration-inhibition curves were generated for each recombinant receptor, and the data were fitted by non-linear regression analysis using GraphPad Prism software.Data were fitted to the equation: 1 Ϫ 1/(1 ϩ (IC 50 /[Ant]) n ), where IC 50 is the concentration of antagonist ([Ant]) that reduces the amplitude of the GABA-evoked current by 50% and n is the Hill coefficient.K I values were calculated using the Cheng-Prussof correction: K I ϭ IC 50 /(1 ϩ ([A]/EC 50 )), where [A] is the concentration of GABA used in each experiment and EC 50 is the concentration of GABA that elicits a half-maximal response for each receptor (44,45).
MTS Reactions and Agonist/Antagonist Protection Assays-Oocytes expressing either wild-type or mutant receptors were pulsed with solutions containing GABA concentrations corresponding to approximately EC 50 for the particular receptor being studied every 5 min until I GABA varied by Ͻ5% for two consecutive pulses.Then, cells were exposed to a high concentration of MTS (3 mM exposure for 3 min) and washed for 10 min, and GABA at EC 50 was applied again to determine accessibility to the modified cysteine residue by the MTS reagent.Preliminary experiments showed that this concentration and time of MTS exposure were sufficient to achieve maximum inhibition of I GABA in all accessible mutant residues.Percent inhibition was calculated as (1 Ϫ (I GABA after MTS/I GABA before MTS)) ϫ 100%.Currents were further examined by multiple pulses of EC 50 GABA after MTS reaction, and responses remained stable.All accessible residues were also tested for the specificity of the MTS reaction by treating the oocytes with the reducing reagent dithiothreitol (20 mM, 3 min), which reversed the inhibition caused by covalent modification.
To determine whether the accessibility of the mutated cysteine could be altered by the presence of agonist (GABA) or antagonist (SR95531), the following protocol was used: After I GABA at EC 50 was stabilized, cells were exposed to a concentration of MTS that yielded approximately 50% of the full inhibitory effect seen with maximal MTS exposure.This concentration was determined separately for each accessible mutation.In a different set of cells, accessible residues were "protected" from reaction with the half-maximal concentration of MTS by co-incubation with either GABA (ϳ500 ϫ EC 50 ) or SR95531 (ϳ10 ϫ K I ).After measurement of percent inhibition in the presence of protecting ligand, the same cells were then exposed to maximal MTS to ensure that the previously protected mutant cysteines were still accessible.Control experiments with these high concentrations of agonist or antagonist in FIG. 1. Partial amino acid sequence alignment of selected LGIC rat subunits and the L. stagnalis acetylcholine binding protein (AChBP) from the loop A domain of putative binding sites.The numbering reflects the position of the residues in the mature GABA A R ␤ 2 subunit.Residues in boldface were found to be accessible to modification by MTS after cysteine substitution.Circled residues have been implicated in the formation of recognitions sites for acetylcholine (Torpedo nAChR ␣ (21,22,24,55)), benzodiazepines (GABA A R ␣ 1 (18 -20)), and serotonin (5HT 3A (56,57)).Underlined residues indicated the ␤ strand structure determined in the AChBP crystal structure (35).
the absence of the MTS reagent were performed in cells prior to the co-application, and necessary wash times were determined to obtain a Ͻ5% change in I GABA .In all cases, a 15-min wash in ND96 was sufficient to ensure complete recovery from desensitization.
Statistical comparisons for accessibility or protection employed oneway ANOVA with Dunnett's post-test to determine levels of significance.Accessibility of mutant receptors using maximal MTS exposure was compared with the same treatment for wild-type ␣ 1 ␤ 2 GABA A receptors.For protection assays, the inhibition of peak I GABA after half-maximal MTS exposure served as the control measurement for the Dunnett's post-test.This average was compared with half-maximal MTS plus the protectant GABA, half-maximal MTS plus SR95531, and to the maximal MTS exposures in the same cells after the respective protection assay was performed.All significant results had p values Ͻ 0.001.

RESULTS
Cysteine substitutions were made individually at ten positions in the ␤ 2 subunit (Trp 92 , Val 93 , Pro 94 , Asp 95 , Thr 96 , Tyr 97 , Phe 98 , Leu 99 , Asn 100 , and Asp 101 ) and tested for functional expression in Xenopus oocytes by co-expression with the wildtype ␣ 1 subunit.GABA activated wild-type ␣ 1 ␤ 2 receptors in a concentration-dependent manner (Fig. 2A), with an apparent affinity of 1.6 M (Table I).All mutants formed functional channels, with the exception of W92C.This tryptophan residue is invariant in all LGIC subunits and may play an essential structural role for receptor assembly.Most cysteine substitutions shifted the EC 50 values for GABA Ͼ10-fold, with the exceptions of V93C and L99C, which were not significantly different from the wild-type value (Table I).The largest shift in the concentration dependence of GABA activation occurred on expression of Y97C (Ͼ100-fold).The Hill coefficients for all mutants were not significantly different from wild-type receptors (Table I).
The competitive antagonist SR95531 reduced I GABA in a concentration-dependent manner (Fig. 2B) with a K I value of 330 nM at wild-type receptors (Table I).Only the Y97C substitution increased SR95531 K I .For all other mutations, there was either no change in the K I value for SR95531 or there was a moderate decrease.The K I value was not determined for receptors carrying the F98C mutation, because small currents elicited by GABA at EC 50 (Ϸ25 nA) precluded an accurate determination (see "Experimental Procedures").None of the SR95531 Hill coefficients were significantly different from wild-type (Table I).Overall, these results indicate that cysteine substitution in this region is tolerated.
Wild-type and cysteine mutant receptors were then tested for modification by MTS (Fig. 3).After exposure to 3 mM MTS for 3 min, I GABA peaks were measured and compared with current peaks prior to MTS exposure.I GABA was significantly inhibited in only four mutant receptors (Fig. 3B); ␣ 1 ␤ 2 V93C was inhibited by 71.4 Ϯ 2.1%, ␣ 1 ␤ 2 D95C by 98.9 Ϯ 0.4%, ␣ 1 ␤ 2 Y97C by 66.9 Ϯ 1.9%, and ␣ 1 ␤ 2 L99C by 98.2 Ϯ 0.4%.MTS had no effect on wild-type GABA A Rs and receptors containing P94C, T96C, F98C, N100C, and D101C.It should be noted that these receptors may have reacted with MTS, but the reactions had no functional effect.Nevertheless, a change in current amplitude following MTS treatment and reversibility of the sulfhydryl-specific reaction by dithiothreitol proves that cysteine modification has occurred.The pattern of accessibility (i.e.every other residue) is indicative of a ␤ strand conformation.
MTS inhibition of I GABA could be the result of a direct effect such as steric block and/or an indirect allosteric effect on the binding site.To determine if the accessible residues contribute to the lining of the GABA binding pocket, mutant receptors were tested for "protection" from the MTS reaction using coapplication of agonist (GABA) or antagonist (SR95531) as "protectants."First, we determined a concentration and time of MTS exposure that resulted in an approximately half-maximal inhibition for each of the accessible mutants (Fig. 4).For ␣ 1 ␤ 2 V93C, normalized inhibition of 41.9 Ϯ 3.6% (29.9% mean inhibition divided by mean maximal inhibition of 71.4%) required 1-min exposure to 3 mM MTS; ␣ 1 ␤ 2 D95C required 67 M, 1 min (normalized inhibition 47.3 Ϯ 3.6%); ␣ 1 ␤ 2 Y97C required 3 mM, 1.25 min (54.8Ϯ 2.9%); ␣ 1 ␤ 2 L99C required 100 M, 1 min (56.2Ϯ 2.6%).Then, either GABA (at ϳ500ϫ the EC 50 value for that mutant) or SR95531 (at ϳ10ϫ the K I value) was co-applied with the half-maximal MTS in previously unexposed cells, and the change in I GABA was measured.Mutant receptors containing L99C were completely protected from MTS reaction (0% MTS inhibition) by both agonist and antagonist (Fig. 4, A and  B), even at concentrations of MTS (1 mM) that produce maximal inhibition (data not shown).␣ 1 ␤ 2 Y97C receptors were significantly protected by both GABA and SR95531 as well (Fig. 4A).V93C-and D95C-containing receptors were not protected by either ligand, indicating that these residues do not line the binding site.Interestingly, SR95531 enhanced MTS inhibition of I GABA in D95C-containing receptors (Fig. 4, A and C, 93 Ϯ 1.7% inhibition in the presence of SR95531 versus 47.3 Ϯ 3.6% inhibition in the absence).To test whether maximal MTS effects could still be attained, "protected" receptors were exposed to a high concentration of MTS after each protection assay (Fig. 4, B and C).In all cases, I GABA in mutant receptors could be further inhibited to a level indistinguishable from maximal MTS exposure in naı ¨ve cells (data not shown).In summary, these data indicate that Tyr 97 and Leu 99 contribute to the lining of the GABA binding pocket, and Asp 95 is a nearby FIG. 2. A, concentration-response curves for  I. B, IC 50 curves for SR95531 demonstrate that this compound is able to antagonize I GABA in a concentration-dependent manner at ␣ 1 ␤ 2 (f), ␣ 1 ␤ 2 V93C (Ⅺ), ␣ 1 ␤ 2 D95C (q), ␣ 1 ␤ 2 Y97C (ࡗ), and ␣ 1 ␤ 2 L99C (E) GABA A receptors.Data represent the mean Ϯ S.E. for at least three independent experiments.Concentration-inhibition data for SR95531 for wild-type receptors and those carrying mutations of the ␤ 2 subunit are summarized in Table I.
residue that undergoes a change in environment in response to SR95531 binding.
It was interesting to note that, for studies using receptors containing L99C, it was necessary to express this mutant cRNA with an excess of ␣ 1 subunit (4:1 ratio).Receptors ex- pressed in a 1:1 ␣:␤ mutant ratio exhibited a large "leak" current at the holding potential of Ϫ80 mV.To determine the source of this leak current, we expressed the cRNA encoding the ␤ 2 L99C alone.␤ 2 L99C homomers consistently produced a leak current (500 -8000 nA) that appeared to slowly desensitize and was "blocked" by picrotoxin (Fig. 5A).To extricate mutant homomer spontaneously open current from measurements of I GABA in ␣ 1 ␤ 2 L99C receptors, we biased the mutant receptors to assemble with ␣ 1 subunits by increasing the ␣:␤ ratio.
The ␤ 2 L99C leak current was always much larger than the leak current in wild-type ␤ 2 homomers (50 -400 nA) injected even at 5-to 10-fold higher cRNA concentrations.Furthermore, like ␤ 2 homomers, the ␤ 2 L99C channels formed were activated by pentobarbital (Fig. 5, B and C).The leak and pentobarbitalactivated currents are both blocked by picrotoxin, providing further evidence that both the wild-type ␤ 2 and mutant ␤ 2 L99C subunits form homomeric GABA A receptors.The data suggest that mutation of Leu 99 within the GABA-binding site is capable of initiating the initial allosteric transitions necessary to gate the ion channel open.

DISCUSSION
In this study, we used SCAM to investigate the structure and function of the GABA A R ␤ 2 Trp 92 -Asp 101 region, which overlaps the putative "loop A" agonist-binding domain.Because alternating residues from ␤ 2 Val 93 -Leu 99 reacted with MTS, we pre- dict this region forms a ␤-strand.Our secondary structure prediction is consistent with the recently solved crystal structure of the acetylcholine binding protein (AChBP) (35).The AChBP is a homologue of the extracellular amino-terminal domain of the nAChR and binds acetylcholine; thus, its structure can be used as a homology model for the agonist-binding sites of the LGIC superfamily of receptors.Residues in the AChBP aligned with the loop A domain of the GABA A R form a ␤-strand (Figs. 1 and 6).Based on our accessibility data, we would predict that the ␤ strand in the GABA A R extends from ␤ 2 Val 93 to ␤ 2 Leu 99 , which is longer than the length of the homologous ␤ strand in the AChBP structure.This small difference may reflect variation in the structure of the binding sites of these different proteins or may reflect limitations of the SCAM approach.The relative placement of the residue side chains overlain onto the AChBP crystal structure reveals that the homologous ␤ 2 residues may still be accessible in an alter- nating pattern (Fig. 6), even if the underlying structure were not a strict ␤ strand.Surprisingly, in the aligned region of the nAChR ␣ 1 subunit, the accessibility to sulfhydryl modification does not predict a ␤-strand conformation (20).The pharmacological specificity of these receptors reflects differences in structure and side-chain chemistry of the amino acid residues lining the binding pocket and thus may account for this lack of secondary structural homology between members of the LGIC c Experiments carried out only twice due to unstable expression.Note that expression of ␤ 2 W92C with ␣ 1 subunits did not produce functional channels.

FIG. 3. Summary of the inhibitory effect of the application of MTS on the amplitude of I GABA .
A, current traces for selected mutant-containing receptors demonstrating the effect of maximal MTS application (3 mM, 3 min).The response to GABA at EC 50 was stabilized by repeated application at regular time intervals (5 min).Stability is defined as Յ5% variation of the peak amplitude of the current for two or more successive trials.The peak amplitude of the current at the same GABA concentration was measured after MTS application (arrow).MTS had no effect on wild-type or D101C-containing receptors.B, inhibition of I GABA in receptors carrying engineered cysteines that were accessible to derivatization by MTS ( ) was significant compared with control (*, p Ͻ 0.001).Percent inhibition was calculated as (1 Ϫ (I GABA after MTS/I GABA before MTS)) ϫ 100%.Data represent the mean Ϯ S.E. of at least three independent experiments.

superfamily.
GABA EC 50 values for seven of the cysteine mutations were Ͼ10-fold higher than wild-type receptors (Fig. 2, Table I), demonstrating that GABA EC 50 is quite sensitive to perturbation of this domain and points to the possible importance of this domain for GABA binding and/or activation.As in all mutagene-sis experiments, one complication with this interpretation is that the overall structure of the mutant receptors and/or the orientation of the cysteine side chain may be unlike wild-type receptor structure.However, because the majority of the cysteine mutations caused either no change in the K I value for SR95531 or decreased SR95531 K I Ͻ 6-fold, we believe that the overall structures of the mutant receptors are not significantly compromised by the cysteine substitutions.If any ligand for a FIG. 4. Summary of the effects of co-application of agonist (GABA) or antagonist (SR95531) on MTS inhibition of I GABA .A, peak responses to EC 50 concentrations of GABA in receptors with accessible cysteine mutations (␣ 1 ␤ 2 V93C, D95C, Y97C, or L99C) were stabilized, and cells were exposed to a concentration of MTS and application time that inhibited currents by ϳ50% of the maximal MTS effect (control, black bars).Note that the half-maximal inhibition was normalized to that which produced the maximum effect.Parameters for the mutants were as follows: 3 mM, 1 min for ␣ 1 ␤ 2 V93C; 67 M, 1 min for ␣ 1 ␤ 2 D95C; 3 mM, 1.25 min for ␣ 1 ␤ 2 Y97C; 100 M, 1 min for ␣ 1 ␤ 2 L99C.The same experiments were repeated with co-application of high concentrations of GABA (ϳ500ϫ EC 50 ; ϩGABA, white bars) or SR95531 (ϳ10ϫ K I ; ϩSR95531, gray bars).Concentrations of agonist or antagonist were as follows: 2 mM GABA, 0.5 M SR95531 for ␣ 1 ␤ 2 V93C; 27.5 mM GABA, 3 M SR95531 for ␣ 1 ␤ 2 D95C; 87.5 mM GABA, 265 M SR95531 for ␣ 1 ␤ 2 Y97C; 1 mM GABA, 1 M SR95531 for ␣ 1 ␤ 2 L99C.Data represent the mean Ϯ S.E. of at least three independent experiments.*, p Ͻ 0.001 compared with half-maximal MTS exposure.B, current traces for ␣ 1 ␤ 2 L99C receptors demonstrating protection from the inhibitory effect of half-maximal MTS application (arrow) by GABA (middle traces) or SR95531 (bottom traces) following completed washout of protecting ligands.Following the "protection assay," cells were subjected to maximal MTS exposure (generally 3 mM, 3 min) to determine whether residues were still fully accessible to the MTS reagent.C, current traces for ␣ 1 ␤ 2 D95C receptors demonstrating an increased effect of MTS application by co-application of SR95531.Compare inhibition of current following MTS exposure in the absence (top traces) or presence (bottom traces) of SR95531.FIG. 6. Model of two apposed putative ␤ strands identified in the GABA-binding site.Side-chain residues from this study on the GABA A ␤ 2 subunit found to be accessible (magenta, Val 93 and Asp 95 from top to bottom) or accessible and protected by agonist/antagonist (blue, Tyr 97 and Leu 99 ) are shown superimposed on the crystal structure of two adjacent subunits of the AChBP (35).Gray ribbons represent ␤ strand chains, and white loops are connecting residues.Accessible residues from a previous study (12) from the ␣ 1 subunit are depicted in red.From bottom to top, these are Asp 62 , Phe 64 , Arg 66 , and Ser 68 in the rat sequence.Phe 64 and Arg 66 were determined to be accessible and protected by GABA.Illustration was created using SwissPdbViewer (Glaxo Wellcome).GABA A protein sequence was aligned with the AChBP sequence, and side chains shown were mutated in the AChBP protein data bank (pdb) format background to the corresponding GABA A homologous residues, but not energy minimized.particular recognition site binds equally well to both wild-type and mutant receptors, it is unlikely that gross structural rearrangement has occurred (46).Consistent with this conclusion, maximal peak currents of all the accessible mutant receptors were similar to wild-type ␣ 1 ␤ 2 receptors (1-5 A).
Evidence that Tyr 97 participates in formation of part of the GABA-binding site is derived from multiple approaches.Both agonist and antagonist protect Y97C from MTS reaction (Fig. 4).We cannot entirely exclude the possibility that GABA and SR95531 reduce MTS access to this residue due to conformational changes.However, because agonist and antagonist binding likely induce different changes within the binding site, the observation that both ligands protect suggests that they sterically block the reaction and that Tyr 97 faces into the binding site.Expression of Y97C produced the largest rightward shift in GABA responsiveness (109-fold) and the largest shift in SR95531 sensitivity (81-fold rightward shift).A large shift in agonist EC 50 with parallel slopes and no reductions in current amplitude is consistent with the mutation disrupting the microscopic binding affinity of GABA; however, one needs to be cautious in this interpretation, because it is difficult to compare current maxima between mutant receptors expressed in different oocytes and a change in gating can cause shifts in agonist sensitivity (47).The parallel rightward shift in the sensitivity for the competitive antagonist, SR95531 (lacking efficacy and therefore incapable of gating the channel), lends additional credence to the idea that mutation of Tyr 97 disrupts binding affinity and ␤ 2 Tyr 97 is a GABA-binding site residue (47).
We also have evidence that Leu 99 lines part of the GABAbinding site.Protection from MTS reaction at L99C was robust.Both GABA and SR95531 reduce access completely in ␣ 1 ␤ 2 L99C receptors, even at high MTS concentrations, suggesting that regardless of whether these ligands exert this effect by overlying the binding site or by some allosteric mechanism, the L99C residue is in a position that allows for little or no access by MTS.To test for the possibility that the protectants did not actually impede the MTS reaction but rather caused an undetected conformational change in the mutant receptor, each mutant receptor assayed for protection with agonist or antagonist was further exposed to maximal MTS (Fig. 4, B and C).For example, a protected cysteine-like L99C (Fig. 4B) may have reacted with MTS in the presence of ligand differently than in the unliganded state and irreversibly altered the receptor such that peak I GABA was increased (e.g. with a leftward shift in the GABA concentration response, or an increase in efficacy for GABA).Similarly, in mutant receptors that were accessible but did not show protection (e.g.V93C), a shift in GABA responsiveness due to MTS modification may have masked any increase or decrease in accessibility.In all cases, I GABA in mutant receptors could be further inhibited to a level indistinguishable from maximal MTS exposure in naı ¨ve cells, indicating that protection merely impeded access by MTS to the substituted cysteine, either by direct steric hindrance, or by blocking a local access route for MTS to the substituted cysteine residue.GABA and SR95531 concentration response characteristics for ␣ 1 ␤ 2 L99C receptors differ little from wild-type, suggesting that although Leu 99 is in the binding site this residue may not be a contact residue for GABA binding.
SR95531 has long been considered a classical competitive antagonist at the GABA A receptor (48), and little evidence is available suggesting any allosteric effects of the antagonist on the GABA-binding site.However, there have been conflicting reports of allosteric block by SR95531 on compounds, which, at higher concentrations, can directly activate the GABA A receptor, such as pentobarbital and alphaxolone (49,50).We speculate that SR95531 binds and causes movements in the agonist-binding site that stabilize a closed channel conformation.
The increase in accessibility of ␣ 1 ␤ 2 D95C receptors in the presence of SR95531 was unexpected (Fig. 4).In previous studies (17,34), we have not observed increases in MTS reactivity using antagonists.The mechanism by which SR95531 binding allows MTS better access to D95C is unclear.We speculate that D95C (or residues nearby) move in response to antagonist binding.This movement causes D95C to be in a different environment (i.e. less sterically hindered or more ionized), which increases MTS reaction.Regardless of the mechanism, these data provide evidence that SR95531 binding causes local conformational rearrangements near the GABA-binding site that are different than the movements induced by GABA.A recent crystallographic study of the agonist binding core domain of the glutamate receptor (GluR2) demonstrated that the antagonist 6,7-dinitroquinoxaline-2,3-dione causes a small contraction of the binding site (51) and provides support for the conclusion that antagonist binding can induce structural changes in a ligand-binding site.An alternative explanation for our results is that, once SR95531 is bound within the site, the hydrophobic ring structures of SR95531 may help stabilize the long carbon chain of MTS and increase MTS modification of D95C.However, the major factor in the reactivity of methanethiosulfonate reagents is the ionization of the sulfhydryl side chain, which more likely occurs in an aqueous environment.
The observation that mutation of Leu 99 (a binding site residue) produces receptors that are spontaneously open is novel.Several studies of GABA A receptors have reported mutations that cause the channels to remain in an open state (32,(52)(53)(54), but to date, these phenomena have been attributed to placement of the mutations in or very near the channel.It is tempting to speculate that the unusual, large leak currents produced by ␤ 2 L99C homomers, which are blocked by PTX (Fig. 5), reflect an allosteric change in channel gating.In the case of L99C, a change in the GABA binding region may create an "at a distance" change in gating, at least when the apposing interface is another ␤ subunit.We speculate that Leu 99 may play a role in coupling agonist binding to channel gating.Alternatively, the L99C mutation may somehow impinge on subunit aggregation, allowing for the surface expression of loosely aggregated ␤ 2 L99C subunits.This "leaky" conformation may be tightened by PTX closure, allowing the homomers to resemble wild-type ␤ 2 homomers more closely.
In summary, using SCAM, we have identified two novel residues, ␤ 2 Tyr 97 and Leu 99 , that contribute to forming the GABA binding pocket.In addition, we provide evidence that the ␤ 2 Val 93 -Leu 99 region of the receptor forms a ␤ strand.A residue in this region, Asp 95 , responds differently to GABA and SR95531 binding, and thus, this region may play a role in distinguishing between agonist and antagonist action.Furthermore, we identify a residue within the GABA-binding site, Leu 99 , that when mutated perturbs channel gating.The availability of the crystal structure of the AChBP provides a useful template to overlay residues of the extracellular N termini of the GABA A R subunits (Fig. 6) and to make predictions about binding site residues and distances and interactions between various amino acid residues of interest.Although our experimental data are consistent with the AChBP structure, determining the exact role residues in the loop A region play in GABA A R function will necessitate not only a high resolution structure of the GABA ligand-binding domain but also knowledge of the relative movements of the binding site when the receptor undergoes microscopic binding and gating transitions.

FIG. 5 .
FIG. 5. "Leak" currents in cells injected with ␤ 2 L99C alone appear to be the result of open channel homomers.Dashed lines represent the zero current level.A, trace from a cell expressing ␤ 2 L99C alone demonstrates a large (several A) desensitizing current at Ϫ80mV holding current that is blocked by 100 M PTX.B, activation of ␤ 2 L99C receptors by 500 M pentobarbital (PB).The residual current from PB activation can also be blocked by 100 M PTX and reactivated by 500 M PB.C, activation of ␤ 2 homomeric receptors by 500 M PB.The second peak represents unblocking of the channels at this high PB concentration.

TABLE I
Concentration-response data for 1) I GABA peak measurements and 2) their antagonism by SR95531 at recombinant GABA A receptors expressed in Xenopus laevis oocytes Data represent the mean Ϯ S.E. for n independent experiments performed with similar results.Log (EC 50 ), log (K I ), and n H values were analyzed using a one-way analysis of variance followed by a post-hoc Dunnett's test to compare mutant and wild-type receptors.
a p Ͻ 0.001.Although some of the Hill numbers appear to differ from wild-type, these values were not significantly different.b ND, not determined, due to low levels of expression.