γ-Aminobutyric Acid (GABA) and Pentobarbital Induce Different Conformational Rearrangements in the GABAA Receptor α1 and β2 Pre-M1 Regions*

γ-Aminobutyric acid (GABA) binding to GABAA receptors (GABAARs) triggers conformational movements in the α1 and β2 pre-M1 regions that are associated with channel gating. At high concentrations, the barbiturate pentobarbital opens GABAAR channels with similar conductances as GABA, suggesting that their open state structures are alike. Little, however, is known about the structural rearrangements induced by barbiturates. Here, we examined whether pentobarbital activation triggers movements in the GABAAR pre-M1 regions. α1β2 GABAARs containing cysteine substitutions in the pre-M1 α1 (K219C, K221C) and β2 (K213C, K215C) subunits were expressed in Xenopus oocytes and analyzed using two-electrode voltage clamp. The cysteine substitutions had little to no effect on GABA and pentobarbital EC50 values. Tethering chemically diverse thiol-reactive methanethiosulfonate reagents onto α1K219C and α1K221C affected GABA- and pentobarbital-activated currents differently, suggesting that the pre-M1 structural elements important for GABA and pentobarbital current activation are distinct. Moreover, pentobarbital altered the rates of cysteine modification by methanethiosulfonate reagents differently than GABA. For α1K221Cβ2 receptors, pentobarbital decreased the rate of cysteine modification whereas GABA had no effect. For α1β2K215C receptors, pentobarbital had no effect whereas GABA increased the modification rate. The competitive GABA antagonist SR-95531 and a low, non-activating concentration of pentobarbital did not alter their modification rates, suggesting that the GABA- and pentobarbital-mediated changes in rates reflect gating movements. Overall, the data indicate that the pre-M1 region is involved in both GABA- and pentobarbital-mediated gating transitions. Pentobarbital, however, triggers different movements in this region than GABA, suggesting their activation mechanisms differ.

␥-Aminobutyric acid (GABA) binding to GABA A receptors (GABA A Rs) triggers conformational movements in the ␣ 1 and ␤ 2 pre-M1 regions that are associated with channel gating. At high concentrations, the barbiturate pentobarbital opens GABA A R channels with similar conductances as GABA, suggesting that their open state structures are alike. Little, however, is known about the structural rearrangements induced by barbiturates. Here, we examined whether pentobarbital activation triggers movements in the GABA A R pre-M1 regions. ␣ 1 ␤ 2 GABA A Rs containing cysteine substitutions in the pre-M1 ␣ 1 (K219C, K221C) and ␤ 2 (K213C, K215C) subunits were expressed in Xenopus oocytes and analyzed using two-electrode voltage clamp. The cysteine substitutions had little to no effect on GABA and pentobarbital EC 50 values. Tethering chemically diverse thiol-reactive methanethiosulfonate reagents onto ␣ 1 K219C and ␣ 1 K221C affected GABA-and pentobarbital-activated currents differently, suggesting that the pre-M1 structural elements important for GABA and pentobarbital current activation are distinct. Moreover, pentobarbital altered the rates of cysteine modification by methanethiosulfonate reagents differently than GABA. For ␣ 1 K221C␤ 2 receptors, pentobarbital decreased the rate of cysteine modification whereas GABA had no effect. For ␣ 1 ␤ 2 K215C receptors, pentobarbital had no effect whereas GABA increased the modification rate. The competitive GABA antagonist SR-95531 and a low, non-activating concentration of pentobarbital did not alter their modification rates, suggesting that the GABA-and pentobarbital-mediated changes in rates reflect gating movements. Overall, the data indicate that the pre-M1 region is involved in both GABA-and pentobarbital-mediated gating transitions. Pentobarbital, however, triggers different movements in this region than GABA, suggesting their activation mechanisms differ.
Ligand-gated ion channels (LGICs) 2 are integral membrane proteins that mediate fast synaptic transmission between cells in the brain and at the neuromuscular junction. The type A ␥-aminobutyric acid receptor (GABA A R) is the main inhibitory LGIC in the brain and is the target for a wide range of therapeutic agents such as benzodiazepines, barbiturates, and anesthetics. Barbiturates, such as pentobarbital (PB), have three distinct effects on GABA A R activity. At low concentrations, PB modulates GABA-mediated Cl Ϫ current (I GABA ). At higher concentrations, PB directly activates the GABA A R in the absence of GABA, and at still higher concentrations, PB blocks channel activity (1). Little is known, however, about the structural rearrangements underlying these functional effects.
Single channel studies from mouse spinal neurons (2)(3)(4) and from rat hippocampal neurons (5) have shown that currents evoked by PB are similar in conductance as those evoked by GABA, suggesting that the open state structures stabilized by PB binding are similar to those stabilized by GABA. However, GABA and PB bind to distinct sites on the GABA A R (Fig. 1). The GABA binding site is located at the interfaces of the ␣ 1 and ␤ 2 subunits in the extracellular domain, whereas the PB/general anesthetics binding site(s) are believed to be located ϳ50 Å below the GABA binding site in a water-accessible pocket located between the four transmembrane helices (M1-4) of the receptor (Fig. 1). Mutational analyses as well as photolabeling studies have identified positions in the GABA A R transmembrane helices that are important for mediating the effects of PB/anesthetics, with a proposed binding pocket involving residues in M1 (␣ 1 M236), M2 (␤ 2 N265), and M3 (␤ 2 M286) (6 -8).
The structural machinery associated with coupling agonist binding to channel gating in the Cys-loop family of LGICs likely involves distributed movements of, and interactions between, several discrete domains. Recent evidence suggests that binding of neurotransmitter in the extracellular domain triggers a series of molecular motions (conformational wave) that initiates in the ligand binding pocket, followed by movements in Loop 2, Loop 7 (Cys-loop), the pre-M1 region, the M2-M3 linker, and finally the transmembrane domains to gate the channel (9 -11). Because PB-and GABA-activated channel open state structures are alike (3,12) but PB and GABA bind to different sites, we were interested in determining whether gating motions induced by PB are similar to those induced by GABA.
We previously demonstrated that the ␣ 1 and ␤ 2 pre-M1 regions of the GABA A R, which connect the extracellular domain of each subunit with the transmembrane domain, undergo structural rearrangements during GABA activation (13). In this study, we measured PB-mediated changes in the accessibility of cysteines engineered into the pre-M1 region to monitor structural movements induced by activating concentrations of PB and compared these changes to those induced by GABA. Our data indicate that the pre-M1 region is part of a common gating pathway used by both ligands. PB, however, triggers different movements in this region than GABA, suggesting that structural transitions evoked and/or stabilized by PB binding/ channel activation differ from those triggered by GABA.

EXPERIMENTAL PROCEDURES
Mutagenesis-Rat cDNAs encoding ␣ 1 and ␤ 2 GABA A R subunits were used for all molecular cloning and functional studies. Cysteine mutants were made as previously described (13).

Two-electrode Voltage Clamp-
Oocytes were continuously perfused at a rate of ϳ5 ml/min with ND96 while being held under twoelectrode voltage clamp at Ϫ80 mV. The bath volume was ϳ200 l. Stock solutions of GABA (Sigma-Aldrich) and PB (Research Biochemicals, Natick, MA) were prepared fresh daily in ND96. Borosilicate electrodes (Warner Instruments, Hamden, CT) were filled with 3 M KCl and had resistances between 0.7 and 2 M⍀. Electrophysiological data were acquired with a GeneClamp 500 (Axon Instruments, Foster City, CA) interfaced to a computer with an ITC16 analog-to-digital device (Instrutech, Great Neck, NY) and recorded using Whole Cell Program 3.2.9 (kindly provided by J. Demspter, University of Strathclyde, Glasgow, Scotland).
Concentration Response Analysis-PB concentration responses were measured, and the resulting data were fit to the following equation: where I is the peak response to a given concentration of PB, I max is the maximum amplitude of current, EC 50 is the concentration of PB that evokes half-maximal response, [A] is the agonist concentration, and n is the Hill coefficient. At high PB concentrations, currents were partially blocked during PB application. Thus, peak PB currents were measured immediately after PB wash out, when a prominent tail current appears (21). GraphPad Prism 4 software (San Diego, CA) was utilized for data analysis and fitting.
Modification of Introduced Cysteine Residues by MTS Reagents-Three derivatives of methanethiosulfonate (CH 3 SO 2 X; MTS) were used to covalently modify the introduced cysteines: MTS-N-biotinylaminoethyl (X ϭ SCH 2 CH 2 NH-biotin; MTSEA-biotin), MTS-ethyltrimethylammonium (X ϭ SCH 2 CH 2 N(CH 3 ) 3 ϩ ; MTSET ϩ ), and MTSethylsulfonate (X ϭ SCH 2 CH 2 SO 3 Ϫ ; MTSES Ϫ ) (Biotium, Hayward, CA). MTSET ϩ is positively charged whereas MTSES Ϫ is negatively charged at neutral pH. Stock solutions (100 mM) were made in DMSO for all MTS reagents, aliquoted into microcentrifuge tubes, and rapidly frozen on ice before storage at Ϫ20°C. For each application of MTS reagent, a new aliquot was thawed, diluted in ND96 to the working concentration, and used immediately to avoid hydrolysis of the MTS compound. The final DMSO concentrations were Յ2%, which had no effect on PB-mediated current responses.
MTS modifications of the engineered cysteines were assayed by measuring changes in PB-evoked current (I PB ). The effects of MTSEA-biotin, MTSET ϩ , and MTSES Ϫ were studied using the following protocol: PB (EC 40 -60 ) current responses (10 s) were measured from oocytes expressing wild-type (␣ 1 ␤ 2 ) or mutant receptors and stabilized. Stability was defined as Ͻ10% variance of peak current responses to PB on two consecutive applications. After stabilization, the MTS reagent (2 mM) was bathapplied for 2 min, followed by a 5-min wash, and then I PB was measured at the same concentration as before the MTS treatment. The effect of the MTS application was calculated as: [((I after /I initial ) Ϫ1) ϫ 100], where I after is the peak PB current elicited after the MTS application and I initial is the peak current before MTS.
Rate of MTS Modification-The rates at which the various MTS reagents modified the engineered cysteines were determined by measuring the effect of sequential applications of low concentrations of MTS reagents on I GABA as described previously (17). The protocol is described as follows: EC 40 -60 GABA was applied for 10 s every 3-5 min until I GABA stabilized (Ͻ3% variance). After a 40-s ND96 wash out, MTS reagents were applied for 5-20 s, and the cell was washed for an additional 2.5-4.5 min. The procedure was repeated until I GABA no longer changed, indicating that the reaction had proceeded to apparent completion. Concentration of MTS reagent and time of application varied as follows: For all rate experiments, the decrease or increase in GABAinduced current was plotted versus cumulative time of MTS exposure. Peak current at each time point was normalized to the initial peak current (t ϭ 0) and fit to a single exponential function using GraphPad Prism software to obtain a pseudofirst-order rate constant (k 1 ). The second-order rate constant (k 2 ) was calculated by dividing k 1 by the concentration of the MTS reagent used (18).
Statistical Analysis-Log (EC 50 ) values, changes in PB EC 50 after MTS modification, and second-order (k 2 ) rates were analyzed using a one-way analysis of variance, followed by a posthoc Dunnett's test to determine the level of significance between wild-type and mutant receptors.
Structural Modeling-A model of the entire GABA A receptor was built as previously described (13).

Functional Characterization of Pre-M1 Mutant Receptors-
We previously showed (13) that cysteine substitutions in the pre-M1 region of the ␣ 1 (K219C, K221C) and ␤ 2 (K213C, K215C) subunits had no effects on GABA EC 50 (␣ 1 ␤ 2 receptors, EC 50 ϭ 6.9 Ϯ 0.7 M). To determine whether the same cysteine substitutions altered PB activation, we measured PB concentration responses using two-electrode voltage clamp (Fig. 2). Cysteine substitutions of ␣ 1 K219, ␣ 1 K221, and ␤ 2 K215 had little (2-fold) to no effect (␤ 2 K213) on PB EC 50 values relative to ␣ 1 ␤ 2 receptors (EC 50 ϭ 271 Ϯ 18 M; Fig. 2; Table 1). Mutant receptors had Hill coefficients for PB activation that were not significantly different from wild-type ␣ 1 ␤ 2 receptors (Table 1). PB Peak PB-activated currents were measured after PB wash out (tail current) and used for concentration response fitting. Data points represent the mean Ϯ S.E. from four to six independent experiments. Data were fit by nonlinear regression analysis as described under "Experimental Procedures." PB EC 50 and n H values are reported in Table 1.
maximal macroscopic currents elicited from mutant receptors were similar to ␣ 1 ␤ 2 receptors, ranging from 600 nA to 18 A. The fact that these mutations had little effect on GABA and PB EC 50 values, Hill coefficients, or surface expression suggests that the side chains of the introduced cysteines are in similar positions as the side chains of the native residues, making the introduced cysteines at positions ␣ 1 K219, ␣ 1 K221, ␤ 2 K213, and ␤ 2 K215 ideal candidates to probe the dynamics of the pre-M1 region induced by PB binding/gating.

Effect of PB on MTS Reaction
Rates-To determine whether PB binding/gating induces different structural rearrangements in the pre-M1 regions than GABA, we measured the rates of MTS modification of ␣ 1 K219C, ␣ 1 K221C, ␤ 2 K213C, and ␤ 2 K215C in the presence of a directly activating concentration of PB (500 M) and compared these to rates measured in the presence of GABA (Figs. 4 and 5). The rate of modification of a cysteine by a MTS reagent mainly depends on the ionization of the thiol group and the access pathway of the reagent. Thus, changes in rates measured in the presence of PB or GABA provide a measure of structural changes in the receptor that are triggered by their binding. PB had no effect on the MTSET ϩ rate of modification of ␣ 1 ␤ 2 K215C whereas GABA increased the rate by 4-fold. For ␣ 1 K221C␤ 2 receptors, PB decreased the MTSEA-biotin rate by 2-fold whereas GABA had no significant effect. For ␣ 1 ␤ 2 K213C receptors and ␣ 1 K219C␤ 2 receptors, GABA increased and decreased their rates of modification by 3-   Table 2. Data are the mean Ϯ S.E. from at least three independent experiments. * and ** indicate values significantly different from control at p Ͻ0.05 and p Ͻ0.001, respectively. NFE (no functional effect) MTS reagent reacts with the cysteine mutant in the presence of PB but has no functional effect on subsequent GABA responses. and 4-fold, respectively. In contrast, when PB was present during the MTS reaction, the MTS treatment no longer altered subsequent GABA current responses for ␣ 1 ␤ 2 K213C and ␣ 1 K219C␤ 2 receptors (Fig. 4C). A subsequent application of MTS reagent in the absence of PB had little effect on GABA current, indicating that the thiol was modified in the presence of PB but modification now resulted in no detectable effect on I GABA . Although the mechanism underlying this loss of functional effect is unknown, one can conclude that cysteine modification in the presence of 500 M PB is different from modification in the presence of GABA. A low concentration of PB that does not activate the receptor but potentiates GABA responses (50 M) decreased the rate of modification of ␣ 1 K219C by ϳ3-fold and had no effect on the rates of modification of ␣ 1 K221C, ␤ 2 K213C, and ␤ 2 K215C (Fig. 4, Table 2), indicating that the effects of PB (500 M) on ␣ 1 K221C␤ 2 and ␣ 1 ␤ 2 K213C receptors likely reflect gating-associated motions. Overall, these data indicate that GABA and PB induce different structural rearrangements in the pre-M1 regions of the ␣ 1 and ␤ 2 subunits.
Effect of SR-95531 on MTS Reaction Rates-To explore whether the structural rearrangements in the pre-M1 regions induced by GABA reflect conformational movements associated with gating, we measured the rates of MTS modification of ␣ 1 K219C, ␣ 1 K221C, ␤ 2 K213C, and ␤ 2 K215C in the presence of the GABA binding site competitive antagonist SR-95531 (10 M) (Fig. 5). Because GABA and SR-95531 bind to the same site, but GABA promotes channel opening/desensitization whereas SR-99531 does not, co-application of SR-95531 and MTS should capture motions associated with stabilization of a closed state whereas co-application of GABA and MTS should capture motions associated with open/desensitized states (i.e. gating). SR-95531 had no significant effect on the rate of modification of ␣ 1 K221C␤ 2 , ␣ 1 ␤ 2 K213C, and ␣ 1 ␤ 2 K215C receptors compared with control ( Fig. 5 and Table 2), suggesting that occupancy of the GABA binding site alone does not induce structural rearrangements in or near these positions. SR-95531 slowed the MTSEA-biotin rate of modification of ␣ 1 K219C␤ 2 receptors 3-fold ( Fig. 5 and Table 2), indicating that binding of a competitive antagonist can induce structural rearrangements in the pre-M1 region of the ␣ 1 subunit.

DISCUSSION
The Monod-Wyman-Changeux allosteric theory has been used with great success to model LGIC gating behavior (19). In this theory, channel gating is accomplished by a concerted quaternary movement of all the subunits switching from an inactive to an active conformation. GABA and PB activation both evoke the same single channel conductances (3) despite binding to different parts of the receptor (20). Thus, a key question is whether binding of an allosteric modulator such as PB triggers similar allosteric gating transitions as GABA.
For Cys-loop LGICs, it has been suggested that the pre-M1 region acts as a central hub that couples neurotransmitter-induced motions in the ligand binding site to movements in Loop 2 and the M2-M3 linker, which then ultimately triggers movements in the M2 channel region that opens the channel (Fig. 1)  (10, 13). A key element in the transduction pathway coupling neurotransmitter binding to gating of the channel is believed to be a salt bridge between two highly conserved residues present in all Cys-loop LGIC subunits: an arginine in the pre-M1 region with a glutamic acid in Loop 2 ( Fig. 1) (10). Previously, we showed that cysteine substitution of this highly conserved arginine (Arg-216) in the ␤ 2 pre-M1 region of the GABA A R abolished channel gating by GABA without altering binding of the GABA agonist [ 3 H]muscimol (13), suggesting that this residue plays a key role in allosterically coupling GABA binding to gating. Interestingly, the ␤ 2 R216C mutation also abolished channel gating by PB, suggesting that this residue and the pre-M1 region may also play a role in PB activation (13).
Here, we provide evidence that the ␣ 1 and ␤ 2 pre-M1 regions move in response to PB activation of the GABA A R and that PB triggers different movements in this region than GABA. Cysteine substitutions of the conserved pre-M1 lysine residues had little to no effect on PB EC 50 (Fig. 2 and Table 1); thus, the positions occupied by the cysteine side chains in the mutant receptors are likely similar to the native lysine positions. The rates of modification of ␣ 1 K221C␤ 2 , ␣ 1 K219C␤ 2 , ␣ 1 ␤ 2 K213C, and ␣ 1 ␤ 2 K215C receptors differ depending upon whether GABA or PB is present. For ␣ 1 K221C␤ 2 receptors, only 500 M PB caused a significant change in the rate of MTS modification whereas for ␣ 1 ␤ 2 K215C receptors only GABA caused a change (Fig. 5, Table 2). For ␣ 1 K219C␤ 2 and ␣ 1 ␤ 2 K213C receptors, GABA altered their rates of modification, whereas in the presence of PB the mutant receptors were modified but MTS modification no longer altered subsequent GABA-induced current (Fig. 5, Table 2), demonstrating that PB stabilizes the receptor in a different conformation(s) than GABA. Moreover, structur- ally perturbing the pre-M1 regions by tethering chemically diverse thiol-reactive groups onto these mutant cysteines had different effects on PB and GABA current responses (Figs. 2 and  3). Based on these data, we infer that the structural transitions evoked and/or stabilized by PB channel activation differ from those triggered by GABA in this region of the receptor. Alternatively, one could argue that some of the differences measured in the effects of modifying these cysteines on GABA and PB current responses result from the pre-M1 residues being located near the PB binding site pocket. Several lines of evidence indicate that ␣ 1 K219, ␣ 1 K221, ␤ 2 K213, and ␤ 2 K215 do not form part of the PB binding site. First, mutations of these residues to cysteine had little to no effect on PB EC 50 . If these residues directly formed part of the PB binding site, one would expect bigger shifts in PB EC 50 , especially because of the nonconservative cysteine for lysine substitution. Second, modification of ␤ 2 K213C and ␤ 2 K215C with a variety of MTS reagents all increased PB-induced current and modification of ␣ 1 K219C had no effects on PB-induced currents. If these residues were lining the PB binding site, one would expect that tethering bulky/charged groups at these positions would sterically inhibit the ability of PB to bind and would decrease PB-mediated current. Although modification of ␣ 1 K221C with MTSEA-biotin caused a decrease in PB-mediated current, modification with MTSET and MTSES had no effect on PB-mediated currents, again suggesting that this residue is not part of the PB binding site. Third, PB caused an increase in the rate of MTS modification of ␤ 2 K215C. If ␤ 2 K215C were part of the PB binding site, one would expect that PB would decrease the rate of modification. Moreover, based on our homology model of the GABA A R and given the size of PB (ϳ7 Å), it seems unlikely that ␣ 1 and ␤ 2 pre-M1 region residues are forming part of the general anesthetic binding site. Residues in the ␣ 1 and ␤ 2 pre-M1 regions are separated by 20 Å or more from the residues that have been previously identified as forming the potential PB/general anesthetic binding site ( Fig. 1) (8,(21)(22)(23)(24). Mutations in M2 (Asn-265) and M3 (Met-286) in the ␤ subunit eliminate PB activation of the receptor (8,25) and the actions of the related general anesthetics etomidate and propofol (26 -28). More recently, ␣ 1 M236 (in M1) and ␤M286 (in M3) have been directly identified as being part of a general anesthetic binding site by photolabeling with an etomidate analog (6). Propofol blocks covalent modification of ␤M286C by sulfhydryl-specific reagents, indicating that this residue forms part of a general anesthetic binding site (29). Furthermore, a knock-in mouse for ␤N265M removes the immobilizing and hypnotic actions of PB as well as the actions of etomidate and propofol (30). Taken together, the data indicate that general anesthetics, including PB, likely share a similar binding site, which is located in a water-filled pocket ϳ50 Å below the GABA binding site between M1, M2, and M3.
In the presence of GABA, the receptor undergoes transitions between an ensemble of open and desensitized states (31,32). If PB were stabilizing similar states as GABA, one might expect similar changes in the rate of modification in the presence of PB and GABA. Because this was not the case, we infer that PB binding/gating induces structural rearrangements near the pre-M1 region that are structurally distinct from movements induced by GABA. Interestingly, a recent report using disul-fide-trapping experiments demonstrated that GABA and PB induce a similar open state structure at the 6Ј-position in M2 (12). This is consistent with functional studies that have shown similar single-channel conductances (2-4) regardless of whether GABA A R channels are opened by GABA or by PB. We speculate that the unique movements induced by PB and GABA in the pre-M1 regions are the result of their binding to different sites and triggering different activation pathways that lead to their functional effects.
We envision that PB binding between transmembrane helices initiates a conformational change in the M2-M3 linker that propagates to the pre-M1 region via Loop 2 (Fig. 1). Mutational studies have implicated the ␣ 1 and ␤ 2 M2-M3 linker as involved in PB activation (21,33). The movements in the pre-M1 region triggered by PB are then likely to be transmitted to various regions of the GABA A R extracellular ligand binding domain as well as channel membrane domain. The pre-M1 region may be the conduit by which the actions of PB are propagated to the GABA binding site. Binding studies have shown that PB enhances GABA apparent affinity (34), suggesting that the structure of the GABA binding site changes in the presence of PB. Moreover, we have identified 13 positions in the GABA binding site interface that change accessibility during pentobarbital binding/gating (␤ 2 T160C, ␤ 2 D163C, ␤ 2 G203C, ␤ 2 S204C, ␤ 2 R207C, ␤ 2 S209C, ␤ 2 D62C, ␣ 1 S68C, ␣ 1 E122C, ␣ 1 R131C, ␣ 1 V180C, ␣ 1 A181C, and ␣ 1 R186C) (17,(35)(36)(37)(38), indicating that the extracellular domain undergoes conformational rearrangements during PB binding/gating.
In summary, we have shown that the ␣ 1 and ␤ 2 pre-M1 regions of the GABA A Rs are structural elements involved in both GABA-and PB-mediated channel activation. PB binding, however, induces different structural movements in this region than when the receptor binds GABA, suggesting that PB stabilizes a different state or ensembles of states than GABA. These differences reveal distinct molecular mechanisms of action of these two ligands.