Ligand- and Subunit-specific Conformational Changes in the Ligand-binding Domain and the TM2-TM3 Linker of α1 β2 γ2 GABAA Receptors*

Cys-loop receptor ligand binding sites are located at subunit interfaces where they are lined by loops A–C from one subunit and loops D–F from the adjacent subunit. Agonist binding induces large conformational changes in loops C and F. However, it is controversial as to whether these conformational changes are essential for gating. Here we used voltage clamp fluorometry to investigate the roles of loops C and F in gating the α1 β2 γ2 GABAA receptor. Voltage clamp fluorometry involves labeling introduced cysteines with environmentally sensitive fluorophores and inferring structural rearrangements from ligand-induced fluorescence changes. Previous attempts to define the roles of loops C and F using this technique have focused on homomeric Cys-loop receptors. However, the problem with studying homomeric receptors is that it is difficult to eliminate the possibility of bound ligands interacting directly with attached fluorophores at the same site. Here we show that ligands binding to the β2-α1 interface GABA binding site produce conformational changes at the adjacent subunit interface. This is most likely due to agonist-induced loop C closure directly altering loop F conformation at the adjacent α1-β2 subunit interface. However, as antagonists and agonists produce identical α1 subunit loop F conformational changes, these conformational changes appear unimportant for gating. Finally, we demonstrate that TM2-TM3 loops from adjacent β2 subunits in α1 β2 receptors can dimerize via K24′C disulfides in the closed state. This result implies unexpected conformational mobility in this crucial part of the gating machinery. Together, this information provides new insights into the activation mechanisms of Cys-loop receptors.

GABA A receptor (GABA A R) 4 chloride channels are pentameric Cys-loop receptors that mediate most of the fast in-hibitory synaptic transmission in the brain (1). This family also includes nicotinic acetylcholine receptors (nAChRs), glycine receptors (GlyRs), and serotonin type-3 receptors (2,3). Individual Cys-loop receptor subunits consist of a large ligand-binding domain (LBD) and a transmembrane domain comprising four ␣-helices (TM1-TM4). LBDs themselves consist primarily of a "sandwich" of inner and outer ␤-sheets. The ligand binding site is situated at the interface of adjacent subunits and is formed by binding domain loops A-C from one subunit and loops D-F from the neighboring subunit (4).
The activation mechanism of Cys-loop receptors is the subject of considerable interest as it is important for understanding how this receptor family functions under normal and pathological conditions. Unwin (5) and co-workers (6) originally proposed that an agonist binding to its site induces loop C in the outer ␤-sheet to clasp tightly around the agonist. This is accompanied by a rotation of the inner ␤-sheet (5,(7)(8)(9). Abundant structural and functional evidence indicates that these conformational changes are transmitted to the transmembrane domain via a differential movement of loops 2 and 7 (both part of the inner ␤-sheet) and the pre-TM1 domain (which connects to the outer sheet) (3, 10 -12). As loops 2 and 7 intercalate with the extracellular TM2-TM3 linker domain, agonist-induced movement of these loops alters the conformation of this linker, leading to channel gating.
Crystallography studies on acetylcholine-binding protein have shown that loop F of the outer sheet also moves upon ligand binding (13)(14)(15). However, there is as yet no consensus as to whether conformational changes in this loop represent local conformational changes to lock the ligand onto the binding site, simple distensions of the site to accommodate the ligand, or global conformational changes that are transmitted to the channel gate (16 -19). Similarly, although agonists invariably induce loop C closure (5,(13)(14)(15), a direct link between loop C movement and channel opening has proved difficult to establish. Indeed, a recent study on the homomeric ␣1 GlyR demonstrated that loop C conformational changes occur at high agonist concentrations only, raising the question of whether loop C closure is essential for receptor activation (18).
The aim of the present study was to clarify the roles of loops C and F in channel activation using voltage clamp fluorometry (VCF). This technique correlates conformational changes occurring at the gate with those occurring in some other protein domain of interest in real time (20,21). It takes advantage of the fact that rhodamine fluorescence exhibits an increase in quantum efficiency as the hydrophobicity of its environment is increased. Thus, rhodamine fluorescence intensity reports receptor conformational changes that alter its immediate chemical microenvironment. The unique advantages of VCF are that it can resolve movements in domains that are distant from the gate and detect conformational changes that are electrophysiologically silent. Indeed, VCF has provided valuable insights into the conformational rearrangements of various Cys-loop receptors (16 -19, 22-29).
Attempts to date to define the roles of loops C and F using VCF have focused on homomeric 1 GABA A Rs (16,19,30) and homomeric ␣1 GlyRs (18,29). However, with homomeric receptors, it is not possible to discern whether or not fluorescence changes of fluorophores attached to loop C or F residues are restricted to those sites that contain bound ligands. In the present study, we used ␣1 ␤2 and ␣1 ␤2 ␥2 GABA A Rs in an attempt to compare conformational changes in loops C and F of each subunit that are induced by the agonists GABA and ␤-alanine, the antagonist gabazine (SR-95531), and the allosteric modulators diazepam and methyl-6,7-dimethoxy-4ethyl-␤-carboline-3-carboxylate (DMCM). The crucial advantage of GABA A Rs over homomeric Cys-loop receptors is that GABA, ␤-alanine, and gabazine bind at LBD ␤-␣ interfaces only, whereas diazepam and DMCM bind with high affinity at LBD ␣-␥ interfaces (31) and possibly to LBD ␣-␤ interfaces (see "Discussion"). In addition, a low affinity benzodiazepine site is also present in the transmembrane domain region of GABA A Rs. Thus, by attaching fluorescent labels to loop C or F residues in one subunit only, we can discriminate between those conformational changes imposed by the immediate proximity of the ligand and those that are propagated to adjacent subunits and binding sites. We also investigated conformational changes in TM2-TM3 domain of each subunit to correlate conformational changes in loops C and F with those occurring at a more distant site along the gating pathway.
Molecular Biology-Rat ␣1, ␤2, and ␥2 GABA A R subunit cDNAs were subcloned into the pGEMHE plasmid vector. To eliminate uncross-linked extracellular cysteines, subunits incorporated the following background mutations: ␣1 subunit, C233S and C292S; ␤2 subunit, C287S; and ␥2 subunit, C244S and C303A. QuikChange (Stratagene, La Jolla, CA) was used to generate all cysteine mutants used in this study. The successful incorporation of the mutations was confirmed by the automated sequencing of the entire coding sequence. Capped mRNA for oocyte injection was generated using mMessage mMachine (Ambion, Austin, TX).
Oocyte Preparation, Injection, and Labeling-Oocytes from female Xenopus laevis (Xenopus Express) were prepared as described previously (26) and injected with 10 ng of mRNA. The oocytes were then incubated at 18°C for 3-7 days in an incubation solution containing 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES, 0.6 mM theophylline, 2.5 mM pyruvic acid, and 50 g/ml gentamycin (Cambrex Corp., East Rutherford, NJ), pH 7.4. On the day of recording, oocytes were transferred into ND96 containing 10 -20 M dye. Typical labeling times were 30 s for MTS-R and MTS-TAMRA (on ice), 30 min for TMRM (on ice), and 45 min for AF546 (at room temperature). Oocytes were washed thoroughly after labeling and stored in ND96 on ice for up to 6 h before recording. All four sulfhydryl-reactive fluorophores used in this study respond with an increase in quantum efficiency as the hydrophobicity of their environment is increased (22,24,30). Each cysteine mutant was incubated with all four fluorophores in turn, and the one yielding the largest change in fluorescence (⌬F) upon agonist application was analyzed. As unmutated ␣1 ␤2 dimeric or ␣1 ␤2 ␥2 trimeric GABA A Rs never exhibited a ⌬F or a change in electrophysiological properties following fluorophore incubation, we can rule out nonspecific effects of the labels.
VCF and Data Analysis-The experimental setup comprised an inverted fluorescence microscope (IX51, Olympus, Tokyo, Japan) equipped with a high-Q tetramethylrhodamine isothiocyanate filter set (Chroma Technology, Rockingham, VT), a LUCPlanFLN 40ϫ/numerical aperture 0.6 objective (Olympus), and a PhotoMax 200 photodiode (Dagan Corp., Minneapolis, MN) with a 12-V/100-watt halogen lamp (Olympus) as light source. The recording chamber is similar to those described previously (22,30). Cells were voltageclamped at Ϫ40 mV, and currents were recorded with an OC-725C oocyte amplifier (Warner, Hamden, CT). Current and fluorescence traces were acquired at 200 Hz via a Digidata 1322A interface using pClamp 9.2 software (Axon Instruments, Union City, CA). Fluorescence signals were digitally filtered at 1-2 Hz with an eight-pole Bessel filter for analysis and display. Half-maximal concentrations (EC 50 ) and Hill coefficient (n H ) values for ligand-induced activation of current and fluorescence were obtained using the Hill equation fitted with a non-linear least square algorithm (SigmaPlot 9.0, Systat Software, Point Richmond, CA). All results are ex-pressed as mean Ϯ S.E. of three or more independent experiments. Unless otherwise indicated, statistical analysis was performed using unpaired Student's t test with p Ͻ 0.05 representing significance.
Western Blotting-Oocytes were injected with 1 g total of ␣1 and ␤2-K24ЈC subunit RNA in a 1:3 ratio. After 3 days of incubation, oocytes were washed once with ND96, and then some oocytes were treated with ND96 containing a 10 mM concentration of the reducing agent dithiothreitol (DTT). After 30 min of incubation with gentle shaking at 18°C, DTTtreated oocytes were rinsed twice with ND96. After washing, DTT-treated and control oocytes were biotinylated in ND96 solution containing 1.22 mg/ml EZ-Link NHS-SS-biotin (succinimidyl 2-(biotinamido)-ethyl-1,3Ј-dithiopropionate; Pierce) for 60 min on ice accompanied by slow shaking. Oocytes were then washed twice with ND96 and gently pipetted at 4°C in 20 l of oocyte lysis buffer (comprising 0.1 M Tris-HCl pH 7.4 buffer, 1% Nonidet P-40, 0.3 M NaCl, 15 mM EDTA, 1 mM phenylmethanesulfonyl fluoride, and Complete Mini protease inhibitors, EDTA-free (Roche Diagnostics)) followed by several passages through a hypodermic syringe fitted with a 30-gauge needle. The resultant lysis mixture was sonicated three times for 20 s each on ice. To solubilize proteins from the membrane, Triton X (Sigma-Aldrich) was added to a final concentration of 1% to the sonicated cell mixture. The mixture was solubilized by rotating for 1 h at 4°C. The lysate was then centrifuged at 50,000 ϫ g for 30 min at 4°C to remove debris and yolk. The resultant supernatant containing biotinylated proteins was transferred to streptavidin beads (Pierce) and rotated for 12-16 h at 4°C. Beads were then washed with lysis buffer, and protein was eluted with 2ϫ SDS loading buffer (no bromphenol blue/no DTT) plus 10 M urea. After concentration analysis using BCA Protein Assay kit (Pierce), bromphenol blue was added to both the control and DTT-treated samples, and 10 mM DTT was re-added to the DTT-treated sample. A total of 100 g of protein from each sample was heated at 65°C for 15 min before loading on a NuPAGE 4 -12% Bis-Tris gel (Invitrogen) to separate. Protein bands were subsequently transferred to PVDF membrane by electrotransfer for 1 h using a Bio-Rad Mini Trans-Blot Electrophoretic Transfer Cell. PVDF membranes were blocked for 1 h with 3.5% skimmed milk in Tris-buffered saline containing 0.05% Tween 20 (Sigma-Aldrich) and processed with a primary rabbit anti-GABA A R ␤2 polyclonal antibody (Millipore Bioscience Research Reagents) and subsequently with goat anti-rabbit HRP-conjugated secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA). PVDF membranes were developed using Super Signal West Pico (Pierce).

RESULTS
␣1 Subunit Loop F-The GABA binding site lies at the interface of ␤2 and ␣1 subunits. Loop F of the ␣1 subunit is an unstructured domain that lines the base of the ligand binding pocket. In an attempt to clarify the role of this domain, we compared the conformational changes induced by agonists, antagonists, and a positive allosteric modulator in the loop F domains of ␣1, ␤2, and ␥2 subunits. None of the drugs tested in this study produced a detectable ⌬F response at unmutated ␣1 ␤2 or ␣1 ␤2 ␥2 GABA A Rs following incubation with any of the four fluorophores used in this study. We focused our investigation on the following cysteine-substituted residues: ␣1-R186C, ␤2-I180C, and ␥2-S195C. Although the precise alignment between subunits in the loop F region is uncertain, any plausible alignment reveals that these three residues correspond closely. The locations of these residues on a GABA A R structural model are shown in Fig. 1. Although we also attempted to investigate a range of other labeled sites in loop F, including A181C, E182C, D183C, G184C, and S185C in the ␣1 subunit and T175C, G176C, V177C, T178C, and K179C in the ␤2 subunit, mutant receptors incorporating these mutations did not produce significant ligand-induced ⌬F responses following labeling with MTS-R, MTS-TAMRA, TMRM, or AF546.
The mean EC 50 of the GABA-activated change in current (⌬I) in ␣1-R186C ␤2 receptors was significantly higher (p Ͻ 0.01, unpaired t test) than that of unmutated ␣1 ␤2 GABA A Rs (Table 1). Labeling with MTS-R resulted in a further rightward shift in the mean GABA ⌬I EC 50 value and a significant (p Ͻ 0.001) reduction in peak current (⌬I max ), although the n H did not change (Table 1). A saturating (30 M) GABA concentration produced a peak ⌬F (⌬F max ) of 4.9% in ␣1-R186C ␤2 GABA A Rs ( Fig. 2A and Table 1). As MTS-R labeling of unmutated ␣1 ␤2 GABA A Rs had no effect on the GABA ⌬I EC 50 or ⌬I max values (Table 1) and resulted in no detectable GABA-mediated ⌬F, we conclude that R186C was specifically covalently labeled by rhodamine. We were unable to detect a significant ⌬F following labeling of ␣1-R186C ␤2 receptors with AF546 or TMRM, although a small GABA-evoked ⌬F was observed following labeling with MTS-TAMRA. The on and off time courses of ⌬F responses generally correlated well with agonist application and removal ( Fig. 2A). However, because currents are recorded from the entire oocyte and fluorescence is recorded only from a small surface area, slight discrepancies between current and fluorescence on and off rates are expected. After normalization, the ⌬I and ⌬F concentration-response relationships of MTS-R-labeled ␣1-R186C ␤2 GABA A Rs were almost superimposed ( Fig. 2B) with the averaged ⌬I and ⌬F EC 50 values showing no significant difference (p Ͼ 0.05; Table 1). ␤-Alanine is a low affinity full agonist of the GABA A R (32). As with GABA-mediated responses, MTS-R labeling of the ␣1-R186C ␤2 GABA A R resulted in a rightward shift in the mean ␤-alanine ⌬I EC 50 value (Fig. 2C) that was also accompanied by a significant reduction in ⌬I max (p Ͻ 0.001; Table  1). Saturating (10 mM) ␤-alanine concentrations produced ⌬F max values comparable in magnitude with those produced by GABA ( Fig. 2A and Table 1). As with GABA-mediated responses, the averaged ␤-alanine ⌬I and ⌬F dose-response relationships were largely overlapping in these receptors ( Fig.  2C and Table 2).
Gabazine is a high affinity competitive antagonist of GABA A Rs (33). One previous VCF study showed that gabazine and GABA elicited different conformation changes in the GABA A R ␣1 subunit loop E domain (24), although the F loop has yet to be investigated. Another showed that agonists and competitive antagonists evoked distinct ⌬F max responses at several labeled loop F residues in homomeric 1 GABA A Rs (16). Our own VCF study on homomeric ␣1 GlyRs showed that the agonist glycine and the competitive antagonist strychnine elicited distinct ⌬F max responses in loops C, D, and E but identical ⌬F max responses at two labeled loop F sites (18). Consistent with our previous GlyR results, in the present study, we found that gabazine produced a ⌬F max that was similar in magnitude and direction to that elicited by the agonists ( Fig. 2A). The mean gabazine ⌬F EC 50 was 0.4 Ϯ 0.02 M, the n H was 1.1 Ϯ 0.1, and ⌬F max was 5.8 Ϯ 0.7% (all averaged from n ϭ 4 oocytes). A comparison of mean ⌬F max and ⌬I max values for GABA, ␤-alanine, and gabazine at the ␣1-R186C ␤2 GABA A Rs reveals no significant difference in their ⌬F max responses (Fig. 2D). This result provides no evidence for GABA, ␤-alanine, and gabazine inducing different conformational changes in this domain. Although ␣1 ␤2 GABA A Rs contain a low affinity benzodiazepine binding site (34), we found that concentrations of diazepam up to 100 M, applied in the presence of 3 M (EC 20 ␤2 Subunit Loop F-The ␤2-I180C mutation had no significant effect on GABA sensitivity of ␣1 ␤2 GABA A Rs (Table 1). Although incubation with MTS-R did not result in a detectable GABA-mediated ⌬F, MTS-TAMRA-labeled ␣1 ␤2-I180C GABA A Rs elicited a GABA-mediated ⌬F max of about 1.5% (Fig. 3A). MTS-TAMRA labeling had no significant effect on the GABA ⌬I EC 50 , n H , or ⌬I max values ( Fig. 3B and Table 1). The GABA ⌬I and ⌬F concentration-response relationships were largely overlapping with no significant difference in their respective EC 50 values (p Ͼ 0.05; Fig. 3B and Table 1). There was, however, a dramatic reduction in the mean n H of the ⌬F relative to the ⌬I and concentration-response relationship (p Ͻ 0.01; Table 1).  Receptor sensitivity to ␤-alanine was significantly (p Ͻ 0.01) decreased in the ␣1 ␤2-I180C mutant relative to the unmutated GABA A R ( Table 2), but there was no ␤-alanine ⌬⌱ or ⌬F EC 50 difference between unlabeled and MTS-TAMRAlabeled mutant receptors ( Fig. 3C and Table 2). The ⌬F max elicited by ␤-alanine reached 1.3%, which is comparable with the value induced by GABA (Fig. 3A). As with the GABAmediated responses, the ␤-alanine-evoked ⌬I and ⌬F concentration-response relations were largely overlapping with no significant difference in their respective EC 50 values (p Ͼ 0.05; Table 2). Notably, however, the ␤-alanine ⌬F n H value was dramatically reduced relative to that of ⌬I (p Ͻ 0.001; Table  2). A ⌬F signal was also observed following application of gabazine, although the ⌬F max was significantly smaller than that induced by GABA (p Ͻ 0.01; Fig. 3, A and D). The EC 50 of the gabazine ⌬F concentration-response curve was 0.67 Ϯ 0.03 M, n H was 1.3 Ϯ 0.1, and ⌬F max was 0.8 Ϯ 0.1% (all n ϭ 4 oocytes). As loop F of the ␤2 subunit does not contribute to the agonist binding site, the ligand-induced conformational rearrangements seen in this domain cannot be due to a direct ligand-fluorophore interaction. Thus, we conclude that all three ligands produce allosteric conformational changes to this domain with those of GABA and ␤-alanine being different from that of gabazine. Although ␣1 ␤2 GABA A Rs contain a low affinity benzodiazepine binding site (34), we found that concentrations of diazepam up to 100 M, applied in the presence of 0.3 M (EC 20 ) GABA, elicited no detectable ⌬F response in the MTS-TAMRA-labeled ␣1 ␤2-I180C GABA A R.
␥2 Subunit Loop F-When co-expressed with ␣1 and ␤2 subunits, the ␥2 subunit increased the ⌬I EC 50 values of GABA and ␤-alanine dramatically, providing strong evidence for the incorporation of these subunits into trimeric ␣1 ␤2 ␥2 GABA A Rs (Table 1). Arg-197 in loop F of the ␥2 subunit is homologous with ␣1-Arg-186 and is thought to be important for diazepam binding (35).
Unfortunately, ␣1 ␤2 ␥2 GABA A Rs incorporating the mutation R194C, W196C, or R197C to the ␥2 subunit showed no evidence for labeling by any of the four tested fluorophores. However, the ␣1 ␤2 ␥2-S195C GABA A R was productively labeled by TMRM. The mutation itself produced no significant change in the GABA or ␤-alanine ⌬I EC 50 values (Tables  1 and 2). Labeling with TMRM significantly increased the GABA ⌬I EC 50 value (p Ͻ 0.01; Table 1) and gave rise to a small (1.7%) GABA-induced ⌬F max (Fig. 4, A and B). The GABA ⌬F EC 50 value was slightly left-shifted relative to that of ⌬I (Table 1). In contrast, TMRM labeling resulted in a modest decrease in the mean ␤-alanine ⌬I EC 50 value ( Table  2). The ␤-alanine-mediated ⌬F max value was 1.0%, which was significantly smaller than that produced by GABA (p Ͻ 0.01; Table 2). The small size of this signal precluded quantitation of the full ␤-alanine ⌬F concentration-response relationship. We did not detect a significant gabazine-induced ⌬F. Taken together, several observations, including the non-overlapping GABA ⌬I and ⌬F concentration-response relationships and the differing effects of the ␥2-S195C mutation on GABA and ␤-alanine ⌬I EC 50 values coupled with their significantly different ⌬F max values (Fig. 4B), suggest that GABA and ␤-alanine produced different conformational changes at this residue. As the ␥2 subunit loop F contributes to the binding site for diazepam and DMCM (31), we investigated the actions of both compounds on the ␣1 ␤2 ␥2-S195C GABA A R. Because diazepam has both a high affinity site that requires the ␥2 subunit and a low affinity site that does not (34), we applied 10 and 100 M concentrations of diazepam both alone and together with 3 M (EC 30 ) GABA in an attempt to distinguish between the effects of diazepam occupying the low and high affinity sites. Application of diazepam alone at either concentration produced no detectable ⌬I or ⌬F. However, when coapplied with GABA, both diazepam concentrations produced a small ⌬F (0.33 Ϯ 0.05%, n ϭ 5 oocytes) that was opposite in sign to those produced by GABA and ␤-alanine at the same labeled site (Fig. 4, C and D). Control experiments confirmed that all ⌬F signals reported in this study were not mediated by the DMSO used to dissolve the drugs (not shown). A 10 M concentration of DMCM blocked 42 Ϯ 3% (n ϭ 5 oocytes) of the ⌬I induced by 3 M GABA (e.g. Fig. 4D), and the accompanying ⌬F max was of the same sign as that produced by diazepam but was slightly larger (0.50 Ϯ 0.01%, n ϭ 5 oocytes). These results suggest that GABA and ␤-alanine induce an allosteric conformational change in the ␥2 subunit F loop region, whereas diazepam and DMCM elicit a distinct local conformational change. We also investigated the effects of 100 M diazepam on MTS-R-labeled ␣1-R186C ␤2 ␥2 and MTS-TAMRA-labeled ␣1 ␤2-I180C ␥2 GABA A Rs but observed no significant ⌬F at either receptor (n ϭ 5 oocytes for each experiment; not shown).
Loop C-As ␤2 and ␣1 loop C residues are directly involved in coordinating GABA and benzodiazepines, respectively (31), we introduced cysteine mutations one at a time only to those loop C residues that are not involved in ligand binding. The mutations we introduced were as follows: ␣1 subunit: Q203C, S204C, G207C*, and E208C; ␤2 subunit: S200C, T201C, G202C*, and S203C; and ␥2 subunit: K214C, T216C, and G228C. Asterisks denote those mutations that prevented the expression of GABA-mediated currents. Mutated ␣1 subunits were expressed as dimers with unmutated ␤2 subunits or as trimers with unmutated ␤2 and unmutated ␥2 subunits. Similarly, mutated ␤2 subunits were expressed with unmutated ␣1 subunits with or without unmutated ␥2 subunits. Mutated ␥2 subunits were expressed as trimers with unmutated ␣1 and ␤2 subunits. Oocytes injected with each of these subunit combinations were incubated with each of the four fluorophores in turn. Unfortunately, none of the incubated receptors responded with a detectable ⌬F when challenged by saturating concentrations of GABA, ␤-alanine, gabazine, or diazepam.
␣1 Subunit TM2-TM3 Domain-The TM2-TM3 linker is an extracellular domain that interacts closely with loops 2, 7, and 9 of the ligand-binding domain (3,11). To allow for comparison of TM2 residues among different Cys-loop receptors, a residue numbering system is used that assigns 1Ј to the innermost TM2 residue and 19Ј to the residue at the external lipid-water interface that forms the start of the TM2-TM3 linker. As the TM2-TM3 domain lies distant from the ligand binding sites and forms a critical component of the channel opening mechanism (36 -38), we compared the actions of the agonists, allosteric modulators, and antagonists in inducing conformational changes in this region. Although we attempted to label all cysteine-substituted residues from 19Ј to 27Ј in the ␣1 subunit M2-M3 domain using each of the four dyes, robust ligand-mediated ⌬F responses were observed only when N20ЈC (Fig. 1) was labeled with TMRM (Fig. 4A). The ␣1-N20ЈC ␤2 GABA A R exhibited a small increase in the GABA and ␤-alanine ⌬I EC 50 values relative to the unmutated receptor, although for both agonists, this increase was reversed after labeling with TMRM (Tables 1 and 2). Oocytes expressing ␣1-N20ЈC ␤2 GABA A Rs produced saturating GABA-and ␤-alanine-evoked ⌬F max of values near 2% ( Fig.  5A and Tables 1 and 2). It was not possible to quantitate agonist ⌬F concentration-response relationships as the ⌬F signal appeared to run down dramatically after an initial agonist application. Possible explanations for this phenomenon are considered below. The ␤-alanine-induced ⌬I max and initial ⌬F max values were similar in magnitude to those elicited by GABA (Fig. 5, A and B). Application of saturating (30 M) gabazine did not produce a detectable ⌬F signal (Fig. 5, A and  B). Thus, the conformational change reported by this ⌬F signal is restricted to agonists, implying that this receptor region may be involved in gating.
TMRM labeling of trimeric ␣1-N20ЈC ␤2 ␥2 GABA A Rs significantly increased the agonist sensitivity to ␤-alanine (p Ͻ 0.001) but not to GABA (p Ͼ 0.05; Tables 1 and 2). The ⌬F max values induced by either GABA or ␤-alanine were significantly reduced in magnitude (p ϭ 0.01 and p Ͻ 0.001, respectively) following ␥2 subunit incorporation (Tables 1 and  2), whereas saturating gabazine still elicited no ⌬F response (Fig. 5C). We next investigated whether diazepam alters the conformation of the ␣1 subunit TM2-TM3 domain. As shown in the example in Fig. 5D, the magnitude of the ⌬I increase produced by 100 M diazepam in the presence of 3 M (EC 10 ) GABA in the ␣1-N20ЈC ␤2 ␥2 GABA A R was 1.8-fold larger (representative of n ϭ 4 recordings) than that produced in the presence of an equivalent (EC 10 ) GABA concentration in the ␣1-N20ЈC ␤2 GABA A R. As shown in the sample traces in Fig.  5D, in the ␣1-N20ЈC ␤2 GABA A R and ␣1-N20ЈC ␤2 ␥2 GABA A R, 100 M diazepam produced a mean increase in ⌬F of 0.66 Ϯ 0.01 and 0.30 Ϯ 0.01% (both n ϭ 4 oocytes), respectively. However, in the ␣1-N20ЈC ␤2 ␥2 GABA A R, 10 M diazepam produced a mean decrease in ⌬F of 0.55 Ϯ 0.13% (n ϭ 4 oocytes). Application of either diazepam or DMSO alone did not produce a detectable ⌬F signal. This result suggests that diazepam induces a different conformational change depending on whether it is binding to its high or low affinity sites.
␤2 Subunit TM2-TM3 Domain-To investigate the role of the ␤2 subunit TM2-TM3 domain, we were able to successfully label K24ЈC with MTS-R (Fig. 1). Application of GABA or ␤-alanine to ␣1 ␤2-K24ЈC GABA A Rs produced a robust increase in ⌬F (Fig. 6A). Relative to the unmutated receptor, the ␣1 ␤2-K24ЈC mutant GABA A R exhibited a dramatic increase in GABA and ␤-alanine ⌬I EC 50 values and a dramatic decrease in their respective ⌬I max values (p Ͻ 0.001 for both; Tables 1 and 2). Labeling with MTS-R modestly increased the GABA ⌬I EC 50 value (Fig. 6B) but not the ␤-alanine ⌬I EC 50 value (Fig. 6B) and reduced the ⌬I max values even further by around 50% for both agonists (p Ͻ 0.001; Tables 1 and 2). MTS-R labeling also produced significant decreases in the n H values for both agonists (p Ͻ 0.001; Tables 1 and 2) Saturating concentrations of GABA and ␤-alanine induced ⌬F max values of 1.1 and 0.9%, respectively (Tables 1 and 2). The concentration-response curves of GABA-and ␤-alanine-mediated ⌬I and ⌬F responses were almost overlapping (Fig. 6, B and C) with their respective ⌬I and ⌬F EC 50 values showing no significant difference from each other (p Ͼ 0.05 for both GABA and ␤-alanine; Tables 1 and 2). After labeling, ⌬I max values induced by ␤-alanine were significantly larger than the current induced by saturating GABA (p Ͻ 0.05), but there was no significant difference in their respective ⌬F max values (p Ͼ 0.05; Tables 1 and 2). As suggested by the example in Fig. 6A and confirmed by the averaged results summarized in Fig. 6D, a 30 M concentration of gabazine evoked a negative ⌬F signal. Together, these results suggest that the labeled ␤2-K24ЈC residue reports a conformational change associated with channel opening and that gabazine produces an allosteric conformational change distinct from that induced by agonists.
When incorporated into the ␣1 ␤2-K24ЈC GABA A R, the ␥2 subunit produced the expected rightward shift in the GABA and ␤-alanine ⌬I EC 50 values that was accompanied by a dramatic increase in their respective ⌬I max values (Tables 1 and  2). Labeling with MTS-R induced modest increases in ␣1 ␤2-K24ЈC ␥2 GABA A R sensitivity to GABA and ␤-alanine (Fig. 6, E and F), significant decreases in the n H values for both agonists (Tables 1 and 2), and no change in their respective ⌬I max values (p Ͼ 0.05 for GABA and ␤-alanine; Tables 1 and 2). The ⌬F max responses activated by GABA and ␤-alanine were significantly smaller than those observed in the dimeric ␣1 ␤2-K24ЈC GABA A R (p Ͻ 0.01 for GABA and p Ͻ 0.05 for ␤-alanine; Tables 1 and 2), probably reflecting the reduced abundance of ␤2-K24ЈC subunits. As summarized in Fig. 6G, the ⌬F max responses of ␤-alanine and GABA were not significantly different from each other. Incorporation of ␥2 subunits resulted in small but significant (*, p Ͻ 0.05) increases in GABA and ␤-alanine ⌬F EC 50 values relative to their corresponding ⌬I EC 50 values (Fig. 6, E and F, and Tables 1 and 2). Because of their small size, we were unable to analyze the ⌬F signals induced by gabazine and diazepam in ␣1 ␤2-K24ЈC ␥2 GABA A Rs (e.g. Fig. 6H).
Because insertion of the ␥2 subunit dramatically increased ⌬I max (e.g. Fig. 7A), we hypothesized that two adjacent ␤2-K24ЈC subunits might cross-link to form a disulfide bond that holds the channel closed. Disulfide cross-links between adja-   cent ␤2 subunits via cysteines introduced at the 6Ј and 20Ј positions have previously been shown to prevent GABA A R activation (39 -42). To test this hypothesis, we applied a 10 mM concentration of the reducing agent, DTT, for 3 min to MTS-R-labeled ␣1 ␤2-K24ЈC GABA A Rs. As shown in Fig. 7B, a dramatic increase in GABA-activated current was observed (from 0.7 Ϯ 0.1 to 6.5 Ϯ 0.5 A, n ϭ 4 oocytes) that was accompanied by the complete loss of the ⌬F signal. The loss in ⌬F was most likely due to the DTT-mediated reduction of the disulfide bond attaching rhodamine to K24ЈC. The increased current was most likely due to the reduction of an inter-␤ subunit disulfide bond in a different population of receptors from those labeled by rhodamine. We tested this by applying 10 mM DTT for 3 min to unlabeled ␣1 ␤2-K24ЈC GABA A Rs (Fig. 7C) whereupon the current activated by 100 M GABA increased Ͼ10-fold (from 0.34 Ϯ 0.02 to 4.4 Ϯ 0.5 A, n ϭ 4 oocytes), which is significantly more than the increase observed in labeled receptors (p Ͻ 0.05). We interpret the enhanced current increase as being due to a smaller initial control current caused by the cross-linking of a larger proportion of receptors. Following DTT exposure, the current recovered slightly following a 2-min wash in control solution (Fig. 7C). However, application of 100:400 M copper phenanthroline (an oxidizing agent) accelerated the current recovery (Fig.  7C). The opposing effects of oxidizing and reducing agents on current magnitude suggests that two adjacent ␤2-K24ЈC subunits form a disulfide bond in the closed state that holds the channel closed. As shown in Fig. 7D, Western blot analysis supports this conclusion. The unmutated ␣1 ␤2 GABA A R exhibited a dominant monomer band (55 kDa) and a weaker dimer band (110 kDa) with the relative quantity of monomer increasing following DTT treatment (Fig. 7D). In contrast, the ␣1 ␤2-K24ЈC GABA A R displayed a much more dominant dimer band and a weaker monomer band that also increased in intensity following DTT treatment (Fig. 7D). This Western blot result was replicated in four independent experiments. In all four experiments, monomer labeling appeared much weaker than dimer labeling possibly due to reduced epitope exposure. Together, these data provide strong evidence for the dimerization of adjacent ␤-K24ЈC subunits. ␥2 Subunit TM2-TM3 Domain-Cysteines introduced at 20Ј and 24Ј positions of the ␥2 subunit produced no significant GABA-or diazepam-dependent ⌬F responses, whereas P23ЈC was productively labeled with TMRM. Its location is shown in Fig. 1. Mutant ␣1 ␤2 ␥2-P23ЈC receptors exhibited GABA and ␤-alanine ⌬I EC 50 and ⌬I max values that were indistinguishable from those of unmutated ␣1 ␤2 ␥2 GABA A Rs (Tables 1 and 2). Following incubation with TMRM for 30 min, the mean ⌬I EC 50 values for both GABA and ␤-alanine were slightly increased relative to those of unlabeled ␣1 ␤2 ␥2-P23ЈC GABA A Rs (Tables 1 and 2). The ⌬F max values induced by saturating GABA and ␤-alanine were both near 1% ( Fig. 8A and Tables 1 and 2), but gabazine produced no detectable ⌬F (Fig. 8B). As shown in Fig. 8A, 10 M diazepam produced an increase in both ⌬I and ⌬F when co-applied with 3 M (EC 35 ) GABA. In this experiment, GABA alone produced a mean ⌬I of 2.4 Ϯ 0.3 A and a ⌬F of 0.17 Ϯ 0.01% (both n ϭ 4), whereas diazepam co-application increased the ⌬I to 3.0 Ϯ 0.2 A and the ⌬F to 0.37 Ϯ 0.01% in the same four oocytes. Using a paired t test, the increase in ⌬F was statistically significant (p Ͻ 0.05), whereas the increase in ⌬I was not (p Ͼ 0.05).

DISCUSSION
General Considerations for Data Interpretation-An agonist-induced ⌬F implies that the microenvironment of an at-  Tables 1 and 2 for averaged ⌬I max and ⌬F max values. B, all traces were recorded from the same oocyte expressing ␣1 ␤2-K24ЈC GABA A Rs. A 3-min application of 10 mM DTT dramatically increases saturating current magnitude. Note that the ⌬F signal disappears presumably due to reduction of the disulfide bond linking rhodamine to K24ЈC. C, all traces were recorded from the same oocyte expressing ␣1 ␤2-K24ЈC GABA A Rs. The DDT-mediated current increase reverses spontaneously, and this is enhanced by application of 100:400 M copper phenanthroline (Cu:phen). D, Western blot of ␣1 ␤2 and ␣1 ␤2-K24ЈC GABA A Rs with molecular masses as indicated. Lanes 1 and 2 show that ␣1 ␤2 GABA A Rs exist primarily as monomers with a small dimer population with the relative quantity of monomer increasing with DTT treatment. Lanes 3 and 4 show that ␣1 ␤2-K24ЈC GABA A Rs exist mainly as dimers with a weak monomer band that increased in intensity following DTT treatment. This result was representative of four independent experiments using different protein samples. In all four experiments, monomer labeling appeared much weaker than dimer labeling possibly due to reduced epitope exposure.
tached fluorophore has been chemically altered as a result of agonist binding (20). This response could be due to a direct fluorophore-ligand interaction, an agonist-induced conformational change associated with channel opening, or an agonistinduced conformational change that is unrelated to channel opening. A standard assumption we used in this study is that ligands that produce significantly different ⌬F max values produce different microenvironmental changes (16,18,24). In cases where this occurs, we discuss whether such changes may represent distinct conformational changes and, if so, whether these may be relevant to receptor activation. We also assume that if an agonist and an antagonist produce a similar ⌬F max then that microenvironmental change does not represent a conformational change essential for gating. While recognizing that different conformational changes may elicit similar ⌬F responses, the most conservative interpretation is that a ⌬F max of defined direction and magnitude represents a unique conformational change.
If an agonist-mediated ⌬F concentration response is rightshifted relative to the ⌬I concentration response, then we infer that ⌬F is unlikely to be reporting a conformational change essential for gating because at low agonist concentrations the channel can be activated without a ⌬F response (18). It has previously been argued that if a ⌬F concentration response is modestly left-shifted relative to the ⌬I concentration response then ⌬F may be reporting an agonist binding event (16). The rationale for this is that multiple bound agonists per receptor are required for complete receptor activation, whereas a single bound agonist per receptor should be sufficient to elicit a significant ⌬F but not a significant ⌬I. In the present study, we extend this interpretation to assume that such a ⌬F may report either a direct steric agonist interaction or a local conformational change induced by the binding of a single agonist. We define "local" as a conformational change that may extend between subunits but does not propagate to the gate.
␣1 Subunit Loop F-The GABA binding pocket is formed by ␤2 subunit loops A, B, and C (which form the principle or ϩ side of the interface) and ␣1 subunit loops D, E, and F (which form the complementary or Ϫ side of the interface). Crystallography studies on acetylcholine-binding protein demonstrated that loop F distends when ligands bind to this site (13)(14)(15). However, functional studies have not reached consensus on whether such movements form an essential component of the gating mechanism (17).
In the present study, we identified an ␣1 subunit residue (␣1-R186C) that, following labeling with MTS-R, produced a ⌬F max that was similar in magnitude for the full agonist, GABA; the low efficacy agonist, ␤-alanine; and the antagonist, gabazine. We also showed that at low agonist concentrations the ⌬F signal was proportionately larger than the ⌬I signal (Fig. 2, B and C). The most conservative explanation for all these results is that the attached label is either directly interacting with the bound ligands or, more likely, is reporting a local conformational change that occurs as a consequence of ligand binding. Given that both agonists and a structurally distinct antagonist produced identical ⌬F max values, such a conformational change is unlikely to be associated with channel gating or ligand recognition.
These results are generally consistent with those obtained previously at structurally related homomeric 1 GABA A Rs and homomeric ␣1 GlyRs (16,18,19). ␣1-R186C aligns with S223C in the 1 GABA A R subunit. A fluorescent label attached to 1-S233C produced a negative ⌬F max that was largest for GABA, smaller in magnitude for two partial agonists, and smaller again but still significant for an antagonist (16). However, pooling results from a range of labeled residues in the 1 GABA A R loop F, there proved to be no correlation between agonist efficacy and the size of the corresponding ⌬F max throughout this domain (16,19). A similar situation was observed at two labeled F loop sites in the ␣1 GlyR where agonists and an antagonist produced identical ⌬F max responses (18,29). On the basis of all these results, it must be concluded that ⌬F max of a label attached to ␣1-R186C or corresponding residues in other Cys-loop receptors reports a local conformational change that is not part of the channel activation mechanism.
It has been reported that picrotoxin, which binds in the pore (43,44), can affect ⌬F responses of labels attached to 1 GABA A R loop F residues (19). Although this demonstrates that perturbations of distant sites can allosterically affect loop F conformation (see also the next section), it does not necessarily imply that such conformational changes are involved in channel activation. ␤2 Subunit Loop F-This domain is not part of the agonist binding pocket, and hence, ligand-mediated ⌬F responses of labels attached to this domain must be allosterically mediated. In the present study, we investigated ⌬F responses of MTS-TAMRA attached to ␤2-I180C in loop F. We demonstrated here that the GABA-and ␤-alanine-evoked ⌬F max values were identical to each other but that the gabazine-induced ⌬F max was significantly smaller. Moreover, at low concentrations of both GABA and ␤-alanine, the ⌬F signal was proportionately larger than the ⌬I signal (Fig. 3, B and C). We also found that n H values were decreased dramatically for ⌬F versus ⌬I concentration-response relationships. Together, these results are consistent with agonists and antagonists inducing distinct allosteric conformational changes in the ␤2 subunit loop F domain. However, our observations do not necessarily imply a role for the agonist-induced conformational change in receptor activation. Loop F is physically connected to loop C of the same subunit (4). Because the ␤2 subunit loop C is part of the agonist binding site, but loop F is not, changes in loop C conformation induced by ligand binding could produce conformational changes in loop F. We suggest that the conformational change induced in ␤2 subunit loop F is allosteric and was probably elicited by loop C of the same subunit "clasping" around the bound agonist (6). Although this interpretation has previously been invoked (17), it was made on the basis of data recorded in homomeric Cys-loop receptors where it was not possible to determine whether ⌬F signals emanated from ligand-bound subunits, ligand-unbound subunits, or both.
␥2 Subunit Loop F-Loop F of the ␥2 subunit does not contribute to the binding sites for GABA, ␤-alanine, or gabazine, although it does contribute to the high affinity benzodiazepine and DMCM binding sites (34,(45)(46)(47)(48)(49). On the basis of their significantly different ⌬F max values, our results indicate that GABA and ␤-alanine induce distinct allosteric conformational changes in this domain. In contrast, gabazine produced no detectable ⌬F. The positive correlation between agonist efficacy and ⌬F max magnitude for all three ligands suggests that this conformational change may be associated with activation. This is reasonable as GABA does not bind to either interface involving the ␥2 subunit and thus can only induce a conformation in it via a concerted allosteric conformational change.
A 10 M concentration of diazepam applied to TMRMtagged ␣1 ␤2 ␥2-S195C GABA A Rs produced a ⌬F of 0.33%. Because a higher (100 M) diazepam concentration produced no additional ⌬F response and is insufficient to saturate the low affinity site (34), we conclude that the observed ⌬F response represents the ⌬F max in response to diazepam binding to the high affinity site. We also found that a strongly inhibiting (10 M) concentration of DMCM produced a ⌬F max response similar in magnitude to that elicited by diazepam. As the two ligands both bind at the ␣1-␥2 interface, we conclude that the ⌬F max signals a common conformational change in response to ligand occupation of this site.
There is evidence for a second high affinity benzodiazepine site located at the ␣4-␤3 LBD interface. This is based largely on experiments on ␣4 ␤3 ␦ GABA A Rs where it was shown that a ␦ subunit mutation increased diazepam binding to the ␣4-␦ interface without affecting the already present Ro15-4513 binding site (50). Unfortunately, the pharmacological selectivity profile of this site is not known. The possible existence of such a site at the LBD ␣1-␤2 interface would not affect the conclusions of this study as this site would still remain physically distinct from the GABA site at the ␤2-␣1 subunit interface.
Summary Loop F-We demonstrated that ligand occupation of a ␤2-␣1 interface GABA binding site can produce allosteric conformational changes to the ␤2 subunit loop F that faces the adjacent non-ligand-binding subunit interface. As loop F and loop C are physically connected by a short strand within a single subunit, the most likely explanation is that loop C closure around the bound ligand at one interface directly alters loop F conformation at the adjacent interface. As suggested previously (17), this provides a mechanism to explain how adjacent agonist binding sites can interact and thus provides a mechanism to explain cooperative activation in homomeric pentameric Cys-loop receptors (51). Of course, as heteromeric GABA A Rs do not have adjacent GABA binding sites, this mechanism does not necessarily apply. We also found that high and low efficacy agonists and an antagonist induce identical loop F conformational changes in ␣1 subunits. This suggests that loop F conformational changes observed in GABA A R and GlyR binding sites are not involved in ligand discrimination. However, because a labeled GABA A R ␤2 subunit loop F residue responded differently to agonists and an antagonist, it is possible that loop C movements may mediate ligand discrimination and transmit these to loop F of the adjacent non-ligand-binding interface. Consistent with this, it has previously been concluded that loop C is involved in ligand discrimination on the basis of structural data (13)(14)(15) and VCF studies on the homomeric 1 GABA A R and ␣1 GlyR (18,30). Finally, mediated ⌬F responses in the labeled ␥2 subunit induced by GABA, ␤-alanine, and gabazine were likely to have been caused by concerted local conformational changes as this subunit does not bind these ligands.
␣1 Subunit TM2-TM3 Domain-As this domain lies distant from the ligand binding sites, ligand-mediated ⌬F responses observed in this domain must be mediated allosterically. The initial ⌬F max responses elicited by saturating concentrations of GABA and ␤-alanine at TMRM-tagged ␣1-N20ЈC ␤2 GABA A Rs were indistinguishable from each other, whereas gabazine produced no detectable ⌬F. This suggests a common agonist-mediated conformational change in this region. We do not know why ⌬F responses declined dramatically with successive agonist applications. One possibility is that the fluorophore orientation was irreversibly altered by the first agonist application. Alternatively, the fluorophore could be irreversibly quenched by photoinduced electron transfer in the open state. Unfortunately, the inability to compare ⌬F and ⌬I agonist concentration-response relationships limited our ability to interpret the ⌬F responses observed in this region. Nevertheless, as gabazine induced no ⌬F response, the observed ⌬F is consistent with a role in activation.
The TMRM-labeled ␣1-N20ЈC ␤2 GABA A R also displayed an increased ⌬F response following application of 100 M diazepam. Dimeric ␣1 ␤2 GABA A Rs contain only low affinity diazepam sites that are thought to be located in the transmembrane region (34). We interpret this ⌬F as a conformational change mediated following occupation of the low affinity site. Interestingly, incorporation of the ␥2 subunit resulted in a ⌬F that was reversed in magnitude. As this was activated by a low (10 M) concentration of diazepam, we conclude that it occurred in response to occupation of the high affinity site ␣1-␥2 interface site. Thus, occupation of high and low affinity diazepam sites produces different conformational changes in the ␣1 subunit TM2-TM3 domain. This is the first evidence for different conformational changes mediated by different concentrations of diazepam.
␤2 Subunit TM2-TM3 Domain-The MTS-R-labeled ␣1 ␤2-K24ЈC GABA A R exhibited similar GABA-and ␤-alaninemediated ⌬F max responses. Moreover, the respective GABA and ␤-alanine ⌬I and ⌬F concentration-response relationships overlapped closely (Fig. 6). These results suggest that GABA and ␤-alanine produce similar conformational changes in this domain that may be important for activation. This interpretation is strengthened by the observation that gabazine produced a ⌬F response opposite in sign to that produced by the agonists. The reason this effect was not observed in a previous study (25) may have been due to the use of a different fluorophore and/or an insufficiently high gabazine concentration. Our result provides evidence for gabazine, supposedly a classical competitive antagonist, inducing conformational changes in the gating domain distant from its binding site. We note, however, that incorporation of the ␥2 subunit abolished this ⌬F response. It is unlikely this can be explained by a reduction in the number of labeled ␤2-K24ЈC subunits per receptor as incorporation of ␥2 subunits would have only halved the number of ␤2-K24ЈC subunits but at the same time would have increased the number of uncross-linked ␤2-K24ЈC subunits available for labeling. Thus, the most likely explanation for the reduction in magnitude of the gabazinemediated ⌬F is a structural change imposed by the ␥2 subunit.
Incorporation of ␥2 subunits resulted in small increases in GABA and ␤-alanine ⌬F EC 50 values relative to the corresponding ⌬I EC 50 values. This provides further evidence for a structural change imposed by the ␥2 subunit. Beyond this, however, it is difficult to interpret to this result. One possibility is that in the dimeric receptor the K24ЈC label is detecting conformational change associated with gating, but in the trimeric receptor, it may be detecting a conformational change unrelated to gating (28).
Cross-linking of Adjacent ␤2 Subunits-Several observations prompted us to conclude that adjacent ␤2 subunit residues dimerize efficiently via disulfide bonds between their respective K-24ЈC residues. First, current magnitude in ␣1 ␤2-K24ЈC GABA A Rs was increased by a reducing agent and irreversibly reduced by an oxidizing agent. Second, Western blot analysis showed a dramatically increased dimer formation in ␣1 ␤2-K24ЈC GABA A Rs relative to ␣1 ␤2 GABA A Rs. Finally, incorporation of ␥2 subunits, which prevents the formation of GABA A Rs containing adjacent ␤2 subunits, also dramatically increased current magnitude. This rules out the possibility of dimer formation between non-adjacent ␤2-K24ЈC subunits. It is also unlikely that dimers may have formed between ␤2-K24ЈC and ␣1 or ␥2 subunits due to the lack of nearby free cysteine groups on those subunits. ␤ subunit dimerization has previously been reported to inhibit receptor activation in dimeric ␣1 ␤1-T6ЈC and ␣1 ␤2-E20ЈC GABA A Rs and in trimeric ␣1 ␤1-E20ЈC ␥2 GABA A Rs (39,40,42). These studies concluded that either spontaneous ␤ subunit movements, asymmetric movements of adjacent ␤ subunits, or both were essential for channel activation. The present study demonstrates that dimerization of ␤ subunits via residues extracellular to TM2 can also inhibit receptor activation. Cysteine ␣-carbon atoms must move to within a distance of 5-6 Å of each other to form a disulfide bond (52). Of course, this may represent a minimum distance reached at the limits of their respective mobility ranges and may be much less than their average separation. The secondary structure of the GABA A R ␤ subunit has not been unequivocally established. According to both the Torpedo nAChR structure (38) and substituted cysteine accessibility studies on both ␣1 ␤1 GABA A Rs (53) and muscle nAChRs (54), the TM2 ␣-helical structure extends beyond 24Ј. In the nAChR structure, adjacent subunit 24Ј ␣-carbons are located ϳ17 Å apart. On the other hand, the prokaryotic GLIC Cys-loop receptor structure shows the 24Ј residue to be located in an unstructured loop that extends radially outward from the 19Ј residue at the top of the TM2 ␣-helix (7,8). In this model, adjacent subunit Lys-24Ј ␣-carbons are located 23 Å apart. Our finding of a ␤2-K24ЈC cross-link therefore provides evidence for exceptionally large thermal motions in the 24Ј region of ␤ subunits. However, this result does not unequivocally favor one structural model over the other.
␥2 Subunit TM2-TM3 Domain-TMRM-tagged ␣1 ␤2 ␥2-P23ЈC GABA A Rs displayed identical GABA-and ␤-alanineevoked ⌬F max values, whereas gabazine elicited no ⌬F response. This pattern is similar to that observed in labeled TM2-TM3 residues in ␣1 and ␤2 subunits. It thus suggests that GABA and ␤-alanine produce similar conformational changes in this region, whereas gabazine does not. This is consistent with a role for this conformational change in receptor activation. We also observed a small diazepam-induced ⌬F, suggesting that ␥2-P23ЈC may be reporting a conformational change associated with diazepam modulation.
Summary TM2-TM3 Domain-Gabazine, a putative competitive antagonist, produced TM2-TM3 domain conformational changes that were markedly different from those produced by agonists. As ␤-alanine and GABA produced identical ⌬F max responses at labeled TM2-TM3 residues in all three subunits, it appears that conformational changes were similar for the two agonists. Together, these results suggest that the agonist-mediated ⌬F responses may be reporting conformational changes involved with gating. The detection of conformational changes in the TM2-TM3 domains of all three subunits implies cooperativity at this point in the activation "conformational wave" (36). Although we were unable to detect robust diazepam-or DMCM-mediated ⌬F responses at most tested TM2-TM3 sites, our results do indicate that occupation of the high and low affinity diazepam binding sites results in strikingly different conformational changes in the ␣1 subunit TM2-TM3 domain. Finally, we were surprised to discover that adjacent ␤ subunits could be cross-linked via ␤2-K24ЈC disulfide bonds. In the nAChR and GLIC receptor structures, adjacent 24Ј ␣-carbons are separated by 19 and 23 Å, respectively. As disulfide bond formation requires a cysteine ␣-carbon separation of 5-6 Å, our results demonstrate surprisingly large thermal motions in this region that, when prevented by disulfide formation, can inhibit gating.
Conclusions-This study provides evidence for both ligandand subunit-specific conformational changes in both loop F and the TM2-TM3 linker. We provide no evidence that the loop F conformational changes are important for channel gating. However, conformational changes in the loop F domains of ␤2 and ␥2 subunits are allosterically induced by ligands binding to ␤2-␣1 interface binding sites. We also demonstrate that gabazine, a putative competitive antagonist, induces conformational changes in the TM2-TM3 domain. Finally, we demonstrate that TM2-TM3 loop 24Ј residues from adjacent ␤2 subunits can move into close physical proximity to each other, indicating unexpected conformational freedom in this crucial part of the gating machinery. Together, this information provides new insights into the activation mechanisms of Cys-loop receptors.