Disulphide trapping of the GABA_A receptor reveals the importance of the coupling interface in the action of benzodiazepines

Abstract

BACKGROUND AND SIGNIFICANCE

Although the functional effects of benzodiazepines (BZDs) on GABA_A receptors have been well characterized, the structural mechanism by which these modulators alter activation of the receptor by GABA is still undefined.

EXPERIMENTAL APPROACH

We used disulphide trapping between engineered cysteines to probe BZD-induced conformational changes within the γ₂ subunit and at the α₁/γ₂ coupling interface (Loops 2, 7 and 9) of α₁β₂γ₂ GABA_A receptors.

KEY RESULTS

Crosslinking γ₂Loop 9 to γ₂β-strand 9 (via γ₂S195C/F203C and γ₂S187C/L206C) significantly decreased maximum potentiation by flurazepam, suggesting that modulation of GABA-induced current (I_GABA) by flurazepam involves movements of γ₂Loop 9 relative to γ₂β-strand 9. In contrast, tethering γ₂β-strand 9 to the γ₂ pre-M1 region (via γ₂S202C/S230C) significantly enhanced potentiation by both flurazepam and zolpidem, indicating γ₂S202C/S230C trapped the receptor in a more favourable conformation for positive modulation by BZDs. Intersubunit disulphide bonds formed at the α/γ coupling interface between α₁Loop 2 and γ₂Loop 9 (α₁D56C/γ₂L198C) prevented flurazepam and zolpidem from efficiently modulating I_GABA. Disulphide trapping α₁Loop 2 (α₁D56C) to γ₂β-strand 1 (γ₂P64C) decreased maximal I_GABA as well as flurazepam potentiation. None of the disulphide bonds affected the ability of the negative modulator, 3-carbomethoxy-4-ethyl-6,7-dimethoxy-β-carboline (DMCM), to inhibit I_GABA.

CONCLUSIONS AND IMPLICATIONS

Positive modulation of GABA_A receptors by BZDs requires reorganization of the loops in the α₁/γ₂ coupling interface. BZD-induced movements at the α/γ coupling interface likely synergize with rearrangements induced by GABA binding at the β/α subunit interfaces to enhance channel activation by GABA.

Introduction

GABA_A receptors are allosteric proteins that couple GABA binding to the opening of a chloride-conducting channel. They are members of the pentameric family of ligand-gated ion channels that also includes glycine, 5-HT type 3 and nicotinic acetylcholine receptors (receptor and channel nomenclature follows Alexander et al., 2009). GABA_A receptors are comprised of five homologous subunits arranged pseudo-symmetrically around a central transmembrane channel. The majority of GABA_A receptors in the brain consist of two α, two β and one γ subunit arranged in a clockwise orientation αβαβγ when viewed from the extracellular side of the membrane (Baumann et al., 2002) (Figure 1A). Each subunit has a large N-terminal extracellular domain, a similarly sized C-terminal domain consisting of four α-helical membrane-spanning segments (M1-M4), and a large cytoplasmic loop between M3 and M4. The channel is formed by residues from the M2 helices of each of the subunits (Xu and Akabas, 1996). Each receptor has two GABA binding sites located in the extracellular domain at the β/α subunit interfaces (reviewed in (Akabas, 2004)(Figure 1A). In each subunit, flexible loops (Loop 2, Loop 7, Loop 9, pre-M1 and the M2-M3 linker) lie at the interface between the extracellular domain and the transmembrane domain (the ‘coupling interface’) (Figure 1A), which functionally couple the binding domain to the channel domain for all members of the pentameric family of ligand-gated ion channel receptor family (see Sine and Engel, 2006).

Figure 1.

GABA_A receptor structure and the γ₂ subunit coupling interface. (A) Left, homology model of the α₁β₂γ₂ GABA_A receptor pentamer (Mercado and Czajkowski, 2006) as seen from the extracellular membrane surface. The α₁, β₂ and γ₂ subunits are highlighted in red, yellow and blue respectively. Arrows indicate that GABA binds at the β₂/α₁ interfaces whereas benzodiazepines (BZDs) bind at the α₁/γ₂ interface of the receptor. Right, side view of the α₁ (red) and γ₂ (blue) extracellular domains. Note most of the transmembrane region has been removed for clarity. The general location of the BZD binding site is indicated by an arrow. Several loops at the coupling interface are highlighted as follows: γ₂Loop 9, purple; γ₂pre-M1, yellow; γ₂Loop 7, red; α₁Loop2, green; α₁Loop7, blue. (B) and (C) close-up view of the γ₂ subunit coupling interface showing residues mutated to cysteine in stick representation. Loops are coloured as indicated above. β-strand 9 is highlighted in orange.

GABA_A receptor function is modulated by a number of clinically important drugs including the benzodiazepines (BZDs), barbiturates, anaesthetics and ethanol. BZDs are widely prescribed drugs and exert their anxiolytic, muscle relaxant, sedative and anti-epileptic actions by binding to GABA_A receptors and allosterically modulating GABA-induced current (I_GABA) responses. The BZD binding site is located extracellularly at the α/γ interface (Figure 1A) at a position homologous to the GABA binding sites at the β/α interfaces (see Sigel, 2002). Many structurally diverse ligands bind to the BZD site, including BZD and non-BZD agonists that potentiate I_GABA (positive modulators such as flurazepam and zolpidem), BZD inverse agonists that inhibit I_GABA (negative modulators such as DMCM), and BZD antagonists that bind at the BZD site but have no effect on I_GABA.

The exact mechanism by which these modulators exert their influence on the GABA_A receptor is still debated. Previous studies have led to different conclusions including; that BZDs alter the conductance of the ion channel (Eghbali et al., 1997); that the rate of desensitization is affected (Mellor and Randall, 1997); that the GABA binding rate (Rogers et al., 1994; Lavoie and Twyman, 1996) or dissociation rate (Mellor and Randall, 1997) is changed, and that the channel closed-to-open gating equilibrium is altered (Downing et al., 2005; Rusch and Forman, 2005; Campo-Soria et al., 2006). Many of the identified BZD binding site residues are in homologous positions to residues on the β and α subunits that participate in GABA binding. These observations suggest that the structural mechanisms that couple BZD binding to receptor modulation may be similar to movements induced by GABA binding to promote channel gating.

The therapeutic value of BZDs depends upon how they modulate I_GABA, hence elucidating the structural elements that underlie BZD efficacy is crucial. Previously, using γ₂/α₁ chimeric subunits, we identified residues located in the γ₂ subunit pre-M1 region, extracellular end of M2 and M2-M3 linker that were required for modulation of I_GABA by BZD agonists (Boileau et al., 1998; Boileau and Czajkowski, 1999). We also recently identified γ₂Loop F/9 as a key element in an allosteric pathway that relays positive BZD modulator-induced structural changes at the BZD binding site to the coupling interface (Hanson and Czajkowski, 2008). Here, we tested the hypothesis that the coupling of BZD binding to modulation of I_GABA is mediated, at least in part, via the coupling interfaces in the α₁ and γ₂ subunits. We introduced targeted pairs of cysteines to disulphide trap and constrain movement in the γ₂ subunit and between the α₁ and γ₂ subunits. Both intrasubunit and intersubunit disulphide crosslinking at the α₁/γ₂ coupling interface altered modulation by positive but not negative BZD modulators. Moreover, constraining movement of α₁Loop 2 at the α₁/γ₂ coupling interface significantly impaired GABA-activated channel gating as well as BZD potentiation of I_GABA. These results demonstrate that GABA-induced channel activation and positive modulation by BZD-site agonists require reorganization of the loops in the α₁/γ₂ coupling interface.

Methods

Site-directed mutagenesis

Rat cDNA encoding α₁, β₂ and γ_2L GABA_A receptor subunits in the pUNIV vector (Venkatachalan et al., 2007) were used for all molecular cloning and functional studies. All γ_2L and α₁ cysteine mutants were made by recombinant PCR and verified by double-stranded DNA sequencing.

Expression in Xenopus laevis oocytes

All animal care and experimental protocols were in accordance with the guidelines of the National Institutes of Health and were approved by the Animal and Use Committee of the University of Wisconsin. X. laevis were housed in fully accredited University of Wisconsin Animal Care facilities. Oocytes were harvested and prepared as described previously (Boileau et al., 1998). Capped cRNA was transcribed in vitro from NotI-linearized cDNA using the mMessage mMachine T7 kit (Ambion, Austin, TX, USA). Oocytes were injected within 24 h of treatment with 27 nL (1–15 pg·nL⁻¹ per subunit) in the ratio 1:1:10 (α:β:γ) (Boileau et al., 2002) and stored at 16°C in ND96 buffer [ (in mM) 96 NaCl, 2 KCl, 1 MgCl₂, 1.8 CaCl₂, 5 HEPES, pH 7.2] supplemented with 100 µg·mL⁻¹ gentamycin and 100 µg·mL⁻¹ BSA until used for electrophysiological recordings.

Two-electrode voltage clamp

Oocytes were perfused continuously (5 mL·min⁻¹) with ND96 while held under two-electrode voltage clamp at −80 mV in a bath volume of 200 µL. Borosilicate glass electrodes (0.4–1.0 MΩ) (Warner Instruments, Hamden, CT, USA) used for recordings were filled with 3 M KCl. Electrophysiological data were collected using GeneClamp 500 (Axon Instruments, Foster City, CA, USA) interfaced to a computer with a Digidata 1200 A/D device (Axon Instruments), and were recorded using the Whole Cell Program, v.3.8.9 (kindly provided by J. Dempster, University of Strathclyde, Glasgow, UK).

Concentration–response analysis

For the single-cysteine mutants, six to eight concentrations of GABA were used for each determination of GABA EC₅₀. Each response was scaled to a low, non-desensitizing concentration of GABA (EC_1-5) applied just before the test concentration to correct for any drift in I_GABA responsiveness over the course of the experiment. All concentration–response data were fitted to the following equation: I = I_max/(1 + (EC₅₀/[A]ⁿ), where I is the peak response to a given drug concentration, I_max is the maximum amplitude of current, EC₅₀ is the drug concentration that produces a half-maximal response, [A] is drug concentration, and n is the Hill coefficient using Prism v.5.02 (GraphPad, San Diego, CA, USA). For all mutant and wild-type (WT) receptors, the maximal current responses were ≥5 µA and were achieved with 10 mM GABA. For double cysteine mutant receptors, GABA EC₅₀ was estimated by examining the ratio of response to a sub-maximal concentration of GABA (1–50 µM) (GABA_sub-max) versus 10 mM GABA (GABA_max). Using a standard Hill equation to model the GABA dose–responses and the determined (GABA_sub-max)/(GABA_max) current ratio, we calculated GABA EC₅₀ values (Boileau and Czajkowski, 1999). This measurement also required using an estimated n_H value, which was based on the experimentally determined Hill coefficients of the single-mutant receptors, all of which (except one: D56C) fell within a very tight range: 1.2–1.5. Given this range, the GABA EC₅₀ value using the lowest measured n_H for either single mutant and the highest measured n_H for either single mutant fell within 10% of each other and the average of these values is the reported EC₅₀. The reported EC₅₀ values for these mutants are the average of estimates performed on multiple oocytes.

Benzodiazepine modulation was defined as: [ (I_GABA+BZD/I_GABA) − 1], where I_GABA+BZD is the current response in the presence of GABA and BZD, and I_GABA is the current evoked by GABA alone. BZD modulation was measured at GABA EC₅ for each mutant. For each oocyte, GABA EC₅ was determined prior to each BZD modulation experiment as described above using a (GABA_sub-max)/(GABA_max) current ratio and a standard Hill equation to model the GABA dose-responses.

Dithiothreitol (DTT) and H₂O₂ treatment

Dithiothreitol (Fisher) was dissolved in water to make a 1 M stock solution and stored at −20°C. DTT and hydrogen peroxide (H₂O₂) (3%; Fisher) were diluted in ND96 buffer to final concentrations of 10 mM and 0.3%, respectively, before each experiment.

Before application of DTT or H₂O₂, oocytes were stabilized by applying GABA (EC₅₀) at 3 min intervals until the I_GABA peak current amplitude varied by <5% (generally 2–3 pulses). After achieving current stability, 10 mM DTT was applied for 2 min, followed by a 2 min wash period. GABA (EC₅₀) was applied again. In each case, current amplitude remained stable with multiple GABA applications. Oocytes were then treated with 0.3% H₂O₂ for 2 min, followed by a 2 min wash and GABA (EC₅₀) test pulses. A second treatment of 10 mM DTT for 2 min, followed by GABA (EC₅₀) application was used to assess the reversibility of the H₂O₂ effect.

Dithiothreitol and H₂O₂ effects on BZD modulation of I_GABA was measured similarly. After stabilization of I_GABA (EC₅) current, maximum BZD modulation was measured. This was followed by an 8 min (flurazepam) or 10 min (zolpidem) wash before the 2 min application of DTT or H₂O₂, wash and subsequent measurement of BZD modulation of I_GABA.

Methanethiosulfonate (MTS) modification

Three derivatives of MTS were used to covalently modify the introduced cysteines: MTS-ethylammonium biotin (MTSEA-Biotin), MTS-ethyltrimethylammonium (MTSET) and N-biotinylcaproylaminoethyl-MTS (MTSEA-Biotin CAP) (Toronto Research Chemicals, Toronto, Ontario, Canada). All GABA responses were stabilized by applying GABA (EC₅₀) at 3 min intervals until the peak currents varied by <5%. After achieving current stability, 10 mM DTT was applied for 2 min, followed by a 2 min wash period. I_GABA was measured again at 3 min intervals until stable. Oocytes were then treated with 2 mM MTS for 2 min, washed for 2 min, and I_GABA was remeasured. The effect of the MTS reagent was calculated as: [ ( (I_GABAafter /I_GABAbefore) − 1) × 100].

Structural modelling

A homology model, based on the crystal structure of the Lymnaea acetylcholine-binding protein (Brejc et al. 2001) for the extracellular domain, and the 4 Å structure of the Torpedo nicotine acetylcholine receptor (Miyazawa et al. 2003) for the transmembrane domain, was constructed for the rat GABA_A receptor as described (Mercado and Czajkowski, 2006). In brief, amino acid sequences of the GABA_A receptor were aligned and threaded onto the parent structures, then energy minimized with SYBYL (Tripos Inc. St. Louis, MO, USA). The two domains were then physically docked and again energy-minimized in SYBYL.

The GABA_A receptor model images were developed using PyMOL (DeLano Scientific, Palo Alto, CA, USA).

Statistical analysis

All data were obtained from at least three different oocytes from at least two different frogs. The data were analysed by one-way anova with Dunnett's post-test for significance of differences using Prism v.5.02 (GraphPad, San Diego, CA, USA).

Materials

We used the BZD-site agonists, flurazepam (Research Biochemicals, Natick, MA, USA) and zolpidem (Sigma-Aldrich, St. Louis, MO, USA) and the BZD inverse agonist 3-carbomethoxy-4-ethyl-6,7-dimethoxy-β-carboline (DMCM) (Sigma-Aldrich). Stock solutions of flurazepam and zolpidem (10 mM; H₂O) and DMCM (10 mM; DMSO) were stored at −20°C and diluted in ND96 buffer prior to use. The concentrations of flurazepam and zolpidem used for maximum modulation of I_GABA were independently determined for wild type and each mutant receptor to be 10 µM flurazepam and 10 µM zolpidem. Concentrations below this level (1 and 3 µM) did not evoke maximum potentiation and concentrations above this level (30 µM) did not increase potentiation beyond that achieved at 10 µM for any mutant.

Results

Disulphide trapping γ₂Loop 9 to γ₂β-strand 9 reduces flurazepam potentiation of I_GABA

In order to test the hypothesis that binding of positive BZD modulators induces movement of γ₂Loop 9 towards β-strand 9 and the membrane (Figure 1), we engineered pairs of cysteines in γ₂Loop 9 and γ₂β-strand 9 with the idea that a disulphide bond would constrain movement and alter BZD modulation. Residues that possess a C_β-C_β distance between 4 and 8 Å are ideal candidates for disulphide trapping. Disulphide bond formation becomes less likely as the C_β-C_β distance exceeds 12 Å (Careaga and Falke, 1992). Using a structural homology model of the GABA_AR (Mercado and Czajkowski, 2006), we chose three pairs of residues in Loop 9 and β-strand 9: γ₂S187C/L206C and γ₂S195C/F203C, with C_β-C_β distances of ∼ 5 Å each, and γ₂D192C/L206C (C_β-C_β∼ 13 Å) to serve as a negative control (Figure 1B). The five single-cysteine mutant γ₂ subunits as well as the three double mutant γ₂ subunits were co-expressed with WT α₁ and β₂ subunits in Xenopus oocytes, and characterized using two-electrode voltage clamp.

All of the mutant subunits assembled into GABA_A receptors that responded to GABA. Only three out of eight mutations (γ₂F203C, γ₂S187C/L206C and γ₂S195C/F203C) significantly decreased GABA EC₅₀ values, and these were reduced <3.5-fold compared with WT α₁β₂γ₂ receptors (GABA EC₅₀ = 29.6 ± 4.0 µM; Table 1), indicating that the cysteine substitutions were tolerated. Maximal flurazepam (10 µM) potentiation of GABA EC₅ currents was unaltered for the single-cysteine mutant receptors and α₁β₂γ₂D192C/L206C as compared with WT receptors (Table 1, Figure 2). In contrast, flurazepam potentiation was significantly decreased for α₁β₂γ₂S187C/L206C and α₁β₂γ₂S195C/F203C receptors by 3.6- and 1.9-fold, respectively (Table 1, Figure 2), suggesting that disulphide bonds between these pairs of cysteines may be the underlying cause for the reduced potentiation.

Table 1.

Summary of GABA, flurazepam, and zolpidem data from wild-type (WT) and γ₂ cysteine mutant receptors used for intrasubunit crosslinking

		GABA			Flurazepam		Zolpidem
Mutant position	Receptor	EC50 (µM)	nH	n	Maximum potentiation	n	Maximum potentiation	n
	αβγ	29.6 ± 4.0	1.2 ± 0.1	6	2.76 ± 0.27	8	4.74 ± 0.50	5
Loop 9	αβγS187C	21.6 ± 3.5a	1.6 ± 0.1	3	2.72 ± 0.34	3	4.94 ± 0.39	3
Loop 9	αβγD192C	17.3 ± 3.5a	1.5 ± 0.2	3	2.19 ± 0.15	3	4.12 ± 0.26	3
Loop 9	αβγS195C	29.6 ± 2.1a	1.3 ± 0.1	3	2.54 ± 0.19	5	5.03 ± 0.33	3
β-strand 9	αβγF203C	8.7 ± 1.8**	1.3 ± 0.2	3	2.42 ± 0.26	3	4.52 ± 0.29	3
β-strand 9	αβγL206C	21.2 ± 1.4	1.3 ± 0.2	3	2.71 ± 0.36	3	3.94 ± 0.25	3
9/β-strand 9	αβγS187C/L206C	16.2 ± 4.3*b	ND	12	0.76 ± 0.15**	5	3.99 ± 0.25	3
9/β-strand 9	αβγS195C/F203C	8.4 ± 2.7**b	ND	12	1.44 ± 0.13**	3	4.44 ± 0.57	3
9/β-strand 9	αβγD192C/L206C	19.3 ± 2.8b	ND	4	2.90 ± 0.08	3	5.22 ± 0.34	5
Loop 7	αβγD161C	12.4 ± 3.9**	1.2 ± 0.2	3	4.30 ± 0.61**	4	6.56 ± 0.45**	3
β-strand 9	αβγS202C	23.2 ± 1.0	1.2 ± 0.1	3	2.55 ± 0.15	4	4.71 ± 0.42	3
β-strand 9	αβγV204C	26.5 ± 4.7	1.2 ± 0.1	3	2.45 ± 0.39	4	4.87 ± 0.21	3
pre-M1	αβγS230C	25.6 ± 3.9	1.3 ± 0.1	3	2.84 ± 0.15	4	4.82 ± 0.18	3
7/β-strand 9	αβγD161C/V204C	3.8 ± 1.8**b	ND	5	2.21 ± 0.50	5	4.21 ± 0.20	3
β-strand 9/pre-M1	αβγS202C/S230C	33.3 ± 7.0b	ND	8	6.23 ± 0.76**	5	7.63 ± 0.81**	3

^aHanson and Czajkowski, 2008.

^bGABA EC₅₀ estimated as described in the Methods.

GABA EC₅₀ values were derived by non-linear regression of the concentration-response data as described in the Methods. Flurazepam and zolpidem potentiation of I_GABA(EC5) was calculated as: [ (I_GABA+BZD/I_GABA) − 1]. Maximum potentiation values were established for each mutant receptor. The values reported are in the presence of 10 µM flurazepam and 10 µM zolpidem. Data represent mean ± SD from n experiments. n_H, calculated Hill coefficient. Values significantly different from WT receptors are indicated (^*P < 0.05, ^**P < 0.01). The loop or β-strand where the cysteine mutation is located is indicated in the leftmost column.

ND, not determined.

Figure 2.

Flurazepam potentiation of I_GABA is significantly reduced for α₁β₂γ₂S187C/L206C and α₁β₂γ₂S195C/F203C receptors. Potentiation of GABA EC₅ current by 10 µM flurazepam (FZM) from oocytes injected with wide-type (WT) or mutant receptors. (A) Representative current traces from WT and α₁β₂γ₂S187C/L206C mutant receptors. In order to easily see changes in potentiation, traces are scaled so that the GABA EC₅ current amplitude is the same for each receptor and condition. DTT and H₂O₂ treatment of WT and double cysteine mutant receptors was performed as described in the Methods. Current responses for α₁β₂γ₂S195C/F203C mutant receptors were similar to α₁β₂γ₂S187C/L206C receptors for each condition. Potentiation of GABA current was calculated as [ (I_GABA+FZM/I_GABA) − 1] and is graphed in (B) and reported in Table 1. (B) Potentiation of I_GABA by flurazepam for WT and γ₂-mutant receptors. Black bars indicate mutants in which the change in potentiation was significantly different from WT receptors and the corresponding single-cysteine mutants (**P < 0.01). Data represent mean ± SD from at least three separate experiments. In each case, DTT and H₂O₂ treatment of the WT and double cysteine mutant receptors had no significant effects on flurazepam potentiation of I_GABA.

Initially, we assayed for disulphide bonds by measuring the effects of 10 mM DTT (a reducing agent) and 0.3% H₂O₂ (an oxidizing agent) on BZD modulation of I_GABA. DTT and H₂O₂ had no effect on flurazepam potentiation of I_GABA for WT receptors (Figures 2 and S1). If disulphide bonds were responsible for the decreases in maximum flurazepam potentiation measured for γ₂S187C/L206C and γ₂S195C/F203C mutant receptors, one might expect DTT would restore flurazepam potentiation. Surprisingly, application of DTT or H₂O₂ had no effect on flurazepam potentiation or I_GABA responses for the three double cysteine mutant receptors (Figures 2, S1, and data not shown). Increasing the concentration of DTT, time of DTT application, or treatment with another reducing agent, Tris(2-carboxyethyl)phosphine, also had no effect on flurazepam potentiation or I_GABA responses for the three double cysteine mutant receptors (data not shown), suggesting that these pairs of cysteines may not be disulphide-linked. Non-reducible disulphide bonds between engineered cysteines have been reported in the GABA_A receptor (Horenstein et al., 2001) and between native cysteines in other proteins (Shelness and Thornburg, 1996; Negroiu et al., 2000). Disulphide reduction by DTT requires access to the disulphide bond at the appropriate orientation. Thus, it is possible that the introduced cysteine pairs formed disulphides that were resistant to reduction by DTT.

In order to help distinguish between free and disulphide-linked cysteines, we examined if the introduced single cysteines and cysteine pairs were accessible to modification by sulphydryl-specific MTS reagents. MTS reagents preferentially modify water-accessible ionized cysteine residues (Karlin and Akabas, 1998). We reasoned that if the single engineered cysteines were modified by MTS reagents but the cysteine pairs were not, then this would provide evidence that the cysteine pairs were participating in a disulphide bond. Similar approaches have been used successfully by Bali et al. (2009) to detect whether two cysteines participate in a disulphide bond in the GABA_A receptor transmembrane domain and in voltage-gated K⁺channels to detect free versus disulphide-bonded cysteine residues (Lu and Deutsch, 2001; Kosolapov and Deutsch, 2003).

The single- and double-cysteine mutant receptors were pretreated with DTT and then we measured I_GABA (EC₅₀) and/or flurazepam potentiation of I_GABA before and after exposure to MTSET, MTSEA-Biotin and MTSEA-Biotin CAP. These reagents have no functional effects on WT GABA_A receptors (Figure 3, bar graphs and see Hanson and Czajkowski, 2008). Thus, any changes measured in I_GABA or flurazepam potentiation of I_GABA following MTS exposure indicate that the introduced cysteines were modified. The MTS reagents had no significant effects on I_GABA or flurazepam potentiation of I_GABA for α₁β₂γ₂D192C and α₁β₂γ₂L206C receptors (data not shown) indicating that these introduced cysteines were not accessible to MTS modification or their modification had no functional effect. MTSEA-Biotin CAP significantly increased I_GABA in α₁β₂γ₂S187C receptors and MTSEA-Biotin significantly increased I_GABA in α₁β₂γ₂S195C and α₁β₂γ₂F203C receptors (Figure 3). MTSET also significantly increased I_GABA in α₁β₂γ₂F203C receptors (39.3 ± 2.5%) (data not shown). These data indicate that in single-cysteine mutant receptors γ₂S187C, γ₂S195C and γ₂F203C are ionized and accessible to modification.

Figure 3.

Methanethiosulfonate (MTS) modifies single-cysteine mutant receptors but not double cysteine mutant α₁β₂γ₂S187C/L206C or α₁β₂γ₂S195C/F203C receptors. (A) Representative GABA EC₅₀ currents from oocytes expressing α₁β₂γ₂S187C and α₁β₂γ₂S187C/L206C receptors before and after treatment with 2 mM N-biotinylcaproylaminoethyl-MTS (MTSEA-Biotin CAP). Receptors were pre-treated with 10 mM dithiothreitol (DTT) for 2 min, except where noted. Changes in I_GABA following treatment with MTSEA-Biotin CAP are summarized in the bar graph and were calculated as: [ ( (I_GABAafter /I_GABAbefore) − 1) × 100]. Note that MTS treatment of α₁β₂γ₂S187C receptors significantly increases I_GABA compared to wide-type (WT) receptors (**P < 0.01), whereas the change in I_GABA following MTS treatment of α₁β₂γ₂S187C/L206C receptors both with and without DTT pretreatment is not statistically different from WT receptors. Data represent mean ± SD from at least three separate experiments. (B) Representative GABA EC₅₀ currents from α₁β₂γ₂S195C, α₁β₂γ₂F203C, and α₁β₂γ₂S195C/F203C receptors recorded before and after treatment with 2 mM MTSEA-Biotin. Changes in I_GABA following treatment with MTSEA-Biotin are summarized in the bar graph. Note that MTS treatment of single-cysteine mutant receptors significantly increases I_GABA compared to WT receptors (**P < 0.01), whereas the change in I_GABA following MTS treatment of α₁β₂γ₂S195C/F203C receptors is not significantly different from WT receptors. Data represent mean ± SD from at least three separate experiments.

In contrast, receptors containing the double-cysteine mutation, γ₂S187C/L206C, showed no significant changes in I_GABA compared with WT GABA_A receptors following treatment with MTSEA-Biotin CAP either without or with DTT pretreatment (Figure 3). The MTSEA-Biotin CAP effects on γ₂S187C/L206C were significantly less than that observed with γ₂S187C (Figure 3) and not statistically different from WT receptors. MTSEA-Biotin and MTSET also had no significant effects on γ₂S195C/F203C receptors compared with WT receptors, in contrast to the increases in I_GABA observed for α₁β₂γ₂S195C and α₁β₂γ₂F203C receptors following treatment with the same MTS reagents (Figure 3). The observation that MTS treatment had no significant effects on I_GABA when γ₂S187C was paired with γ₂L206C (γ₂S187C/L206C) and when γ₂S195C was paired with γ₂F203C (γ₂S195C/F203C) demonstrates that these cysteines are not free sulphydryls and indicates that the cysteines are disulphide-linked and are unavailable to react with the MTS reagents. While one could argue that introduction of two cysteines induces a conformational change in the receptor that renders both cysteines resistant to MTS modification, which is not observed for the single-cysteine substitutions, this seems highly unlikely. Moreover, while not significant, MTSEA-Biotin CAP modification of receptors containing γ₂S187C/L206C showed a small but consistent increase in I_GABA following DTT pretreatment (Figure 3A), suggesting reduction of a resistant disulphide bond.

To determine whether modulation of I_GABA by other BZD ligands was also affected by the single- and double-cysteine mutations, we measured maximum potentiation of I_GABA by zolpidem, a structurally distinct BZD-site agonist, and inhibition of I_GABA by the BZD inverse agonist, DMCM. Interestingly, for all of the above single- and double-cysteine mutant receptors, zolpidem (10 µM) maximally potentiated I_GABA to a similar extent as WT receptors (Figure 4, Table 1). In addition, the maximal inhibition of I_GABA by 1 µM DMCM was not significantly different for the single- and double-cysteine mutant receptors as compared with WT receptors (DMCM inh = −0.66 ± 0.02) (Figures 4 and S2). The observations that zolpidem potentiation and DMCM inhibition of I_GABA were unchanged indicate that γ₂S187C/L206C and γ₂S195C/F203C do not impair γ₂-subunit assembly or incorporation into functional GABA_A receptors. Moreover, these results demonstrate that the effects of the mutations on flurazepam potentiation are not due to changes in GABA EC₅₀ values but are specific to flurazepam actions and suggest the coupling of flurazepam binding to potentiation involves unique movements of γ₂Loop 9 relative to γ₂β-strand 9.

Figure 4.

Zolpidem and 3-carbomethoxy-4-ethyl-6,7-dimethoxy-β-carboline (DMCM) modulation of I_GABA is unchanged for γ₂Loop 9/β-strand 9 double cysteine mutant receptors. Modulation of GABA EC₅ currents by 10 µM zolpidem (ZPM) (top) or 1 µM DMCM (bottom) from oocytes expressing wild-type (WT) and mutant receptors. In order to easily see changes in modulation, traces are scaled so that the GABA EC₅ current amplitude is the same for each receptor. Modulation of GABA current was calculated as [ (I_GABA+BZD/I_GABA) − 1]. Values for zolpidem potentiation are reported in Table 1. Zolpidem potentiation and DMCM inhibition for the mutant receptors were not significantly different from WT.

Disulphide trapping γ₂β-strand 9 to γ₂ pre-M1 increases flurazepam and zolpidem potentiation of I_GABA

Two flexible loops, the pre-M1 region and Loop 7, lie between γ₂Loop 9, β-strand 9 and the transmembrane domain (Figure 1C). To test the hypothesis that movements in γ₂Loop 9 and γ₂β-strand 9 initiated by flurazepam binding get transduced to the transmembrane domain via γ₂ pre-M1 we engineered a pair of cysteines in γ₂β-strand 9 and γ₂ pre-M1 (γ₂S202C/S230C, C_β-C_β∼ 4 Å), to restrict local movements in this region (Figure 1C).

Mutations in γ₂S202C, γ₂S230C and γ₂S202C/S230C had no effect on GABA EC₅₀ values compared with WT receptors (GABA EC₅₀ = 30 µM; Table 1). Flurazepam and zolpidem potentiation of I_GABA was also unaltered for α₁β₂γ₂S202C and α₁β₂γ₂S230C receptors as compared with WT receptors, whereas flurazepam and zolpidem potentiation of I_GABA for the double mutant α₁β₂γ₂S202C/S230C receptor was increased by 2.3- and 1.6-fold respectively (Figure 5, Table 1). None of the mutations significantly affected inhibition of I_GABA by DMCM (Figures 5C and S2). Thus, the effects of the γ₂S202C/S230C mutation are specific for modulation by BZD-site agonists.

Figure 5.

Flurazepam and zolpidem potentiation of I_GABA is significantly increased for α₁β₂γ₂S202C/S230C receptors. Modulation of GABA EC₅ currents by 10 µM flurazepam (FZM; A), 10 µM zolpidem (ZPM; B), or 1 µM DMCM (C), from oocytes injected with wild-type (WT) or mutant receptors. In order to easily see changes in potentiation, traces are scaled so that the GABA EC₅ current amplitude is the same for each receptor and condition. Dithiothreitol (DTT) treatment of α₁β₂γ₂S202C/S230C receptors was performed as described in the Methods. Values for flurazepam and zolpidem potentiation are reported in Table 1. For α₁β₂γ₂S202C/S230C receptors, flurazepam and zolpidem potentiation was significantly increased compared to WT and the corresponding single-cysteine mutants, whereas DMCM inhibition was not altered. In each case, DTT and H₂O₂ (not shown) treatment of α₁β₂γ₂S202C/S230C had no significant effect on potentiation.

As observed with α₁βγ₂S187C/L206C and α₁β₂γ₂S195C/F203C receptors, neither DTT nor H₂O₂ had any effect on flurazepam or zolpidem potentiation of I_GABA or on I_GABA alone for α₁β₂γ₂S202C/S230C receptors (Figure 5 and data not shown). To assess whether γ₂S202C/S230C existed as free or disulphide-linked cysteines, we examined the ability of MTS reagents to modify the introduced single- and double-cysteine residues. MTSET significantly increased I_GABA in α₁β₂γ₂S202C and α₁β₂γ₂S230C receptors compared with WT GABA_A receptors (Figure 6). For the double cysteine mutant α₁β₂γ₂S202C/S230C receptor, the effect of MTSET on I_GABA (Figure 6) was significantly less than its effects on the two single-cysteine mutants and was not statistically different from WT receptors. These data suggest that γ₂S202C and γ₂S230C, when expressed together, are disulphide-linked and that crosslinking γ₂β-strand 9 (S202C) to γ₂ pre-M1 (S230C) enhances potentiation of I_GABA by flurazepam and zolpidem.

Figure 6.

Methanethiosulfonate (MTS) modifies α₁β₂γ₂S202C and α₁β₂γ₂S230C receptors but not α₁β₂γ₂S202C/230C receptors. Representative GABA EC₅₀ traces from α₁β₂γ₂S202C, α₁β₂γ₂S230C, and α₁β₂γ₂S202C/230C receptors before and after treatment with 2 mM MTS-ethyltrimethylammonium (MTSET). All receptors were pre-treated with 10 mM dithiothreitol (DTT) for 2 min. Changes in I_GABA following treatment with MTSET are summarized in the bar graph and were calculated as: [ ( (I_GABAafter /I_GABAbefore) − 1) × 100]. Note that MTS treatment of single-cysteine mutant receptors significantly increases I_GABA compared to WT receptors (**P < 0.01), whereas the change in I_GABA following MTS treatment of α₁β₂γ₂S202C/S230C receptors is not significantly different from WT receptors. Data represent mean ± SD from at least three separate experiments.

Disulphide trapping γ₂β-strand 9 to γ₂ Loop 7

To test the hypothesis that movements in γ₂Loop 9 and γ₂β-strand 9 initiated by flurazepam binding get transduced to the transmembrane domain via γ₂Loop 7, we also engineered a pair of cysteines in this region (γ₂D161C/V204C, C_β-C_β∼ 5 Å) (Figure 1C). Maximum flurazepam and zolpidem potentiation of I_GABA was significantly increased for α₁β₂γ₂D161C receptors by 1.6- and 1.4-fold respectively, but was unaltered for α₁β₂γ₂V204C and α₁β₂γ₂D161C/V204C receptors (Table 1). None of the mutations significantly affected inhibition of I_GABA by DMCM (Figure S2). Neither DTT nor H₂O₂ had any effect on flurazepam or zolpidem potentiation of I_GABA or on I_GABA alone for α₁β₂γ₂D161C/V204C receptors and none of the MTS reagents had significant effects (<10%) on I_GABA for α₁β₂γ₂D161C and α₁β₂γ₂V204C receptors (data not shown). Thus, we could not assess whether a disulphide bond between the two residues had formed.

Intersubunit disulphide trapping at the α₁/γ₂ coupling interface reduces flurazepam potentiation

Recently, we demonstrated that movements in γ₂Loop 9 at the α/γ coupling interface are triggered specifically by positive BZD modulators (Hanson and Czajkowski, 2008). We speculated that γ₂Loop 9 movements are propagated to residues in the adjacent α₁ subunit via α₁Loops 2 and 7 due to their proximity in structural models (Figures 1A and S3). To test this hypothesis, we made seven individual cysteine mutations in γ₂Loop 9 (R194C-Q200C), and combined them with single-cysteine mutations in α₁Loop 2 (D56C, M57C), and α₁Loop 7 (E137C, P139C) to create 15 α/γ double-cysteine mutant receptors in order to probe the proximity and dynamics of these regions (Figure S3). Our model also predicted a close interaction (C_β-C_β 5.8 Å) between α₁Loop 2 (D56) and γ₂β-strand 1 (P64) and thus, we also examined this double cysteine mutant.

Six out of the 11 single-cysteine mutations significantly decreased flurazepam maximal potentiation of I_GABA (α₁M57C, α₁E137C, α₁P139C, γ₂P64C, γ₂R197C, γ₂Q200C) and one mutation, α₁D56C, increased flurazepam potentiation (Tables 2 and S1) indicating that residues distributed throughout the α/γ coupling interface are important for regulating BZD efficacy. Changes in potentiation were not correlated to changes in GABA EC₅₀ values. For example, flurazepam potentiation was decreased for α₁M57C, α₁E137C, and γ₂R197C but the GABA EC₅₀ values for these mutants was increased, decreased and unchanged, respectively, compared with WT receptors. For four of the double-cysteine mutants (α₁D56C/γ₂R194C, α₁D56C/γ₂S195C, α₁D56C/γ₂W196C, α₁D56C/γ₂Q200C), flurazepam potentiation was in-between that of the two corresponding singles (Figure S1 and Table S1), suggesting a simple additive effect of the mutations. For the other double mutants, flurazepam potentiation was either not statistically different from one of the single-cysteine mutants (α₁D56C/γ₂R197C, α₁D56C/γ₂Y199C, α₁M57C/γ₂Y199C, α₁D56C/γ₂P64C, α₁E137C/γ₂R194C, α₁E137C/γ₂W196C) or it was significantly less than the two singles (α₁D56C/γ₂L198C, α₁M57C/γ₂S195C, α₁M57C/γ₂W196C, α₁M57C/γ₂L198C, α₁M57C/γ₂Q200C, α₁P139C/γ₂L198C), suggesting the cysteines interact and are potentially disulphide-linked (Figure S1 and Table S1).

Table 2.

Summary of GABA, flurazepam, and zolpidem data from wild-type (WT) and α₁/γ₂ cysteine mutant receptors that form intersubunit disulphide bonds

		GABA			Flurazepam		Zolpidem
Mutant position	Receptor	EC50 (µM)	nH	n	Maximum potentiation	n	Maximum potentiation	n
	αβγ	29.6 ± 4.0	1.2 ± 0.1	6	2.76 ± 0.27	8	4.74 ± 0.50	5
2	αD56Cβγ	123.4 ± 38.2**	0.7 ± 0.1	12	5.65 ± 0.30**	4	6.48 ± 0.70**	4
9	αβγL198C	4.7 ± 0.6**	1.4 ± 0.1	3	2.36 ± 0.06	3	3.67 ± 0.70	3
2/9	αD56CβγL198C	17 ± 3*b	ND	6	1.33 ± 0.44**	5	1.84 ± 0.44**	3
β-strand 1	αβγP64C	6.6 ± 2.6**	1.5 ± 0.1	12	1.82 ± 0.16**	5	3.05 ± 0.06**	3
2/β-strand 1	αD56CβγP64C	56 ± 23b	ND	5	1.61 ± 0.36**	4	2.39 ± 0.35**	5

^bGABA EC₅₀ estimated as described in the Methods.

GABA EC₅₀ values were derived by non-linear regression of the concentration-response data as described in the Methods. Flurazepam and zolpidem potentiation of I_GABA(EC5) was calculated as: [ (I_GABA+BZD/I_GABA) − 1]. Maximum potentiation values were established for each mutant receptor in the absence of any dithiothreitol treatment. Values reported are in the presence of 10 µM flurazepam and 10 µM zolpidem. Data represent mean ± SD from n experiments. n_H, calculated Hill coefficient. Values significantly different from WT receptors are indicated (^*P < 0.05, ^**P < 0.01). The loop or β-strand where the cysteine mutation is located is indicated in the leftmost column.

ND, not determined.

To probe for potential disulphide bonds, we examined the effects of DTT on flurazepam potentiation of I_GABA. Only two out of the 16 double α₁/γ₂ mutant receptors were significantly affected by DTT. DTT significantly increased flurazepam potentiation for α₁D56Cβ₂γ₂L198C (α₁Loop 2/γ₂Loop 9) and α₁D56Cβ₂γ₂P64C (α₁Loop 2/γ₂β-strand 1) receptors (Figure 7A). DTT also significantly increased zolpidem potentiation for α₁D56Cβ₂γ₂L198C receptors (data not shown). DTT had no effect on WT and the corresponding single-cysteine substitutions (Figure 7B). The DTT-induced increase in flurazepam potentiation was reversed by H₂O₂ and a subsequent application of DTT increased flurazepam potentiation to a similar extent as that measured after the first DTT treatment (Figure 7A). From these results, we infer that disulphide bonds between α₁D56C and γ₂L198C and between α₁D56C and γ₂P64C form spontaneously, which inhibit flurazepam potentiation of I_GABA. DMCM inhibition of I_GABA was unaffected by the α₁D56C/γ₂L198C or α₁D56C/γ₂P64C mutations (Figures 7C and S2) indicating that the receptors assembled into functional αβγ receptors and that the effects exerted by the single- and double-cysteine mutations on flurazepam and zolpidem potentiation were specific for agonists at the BZD site.

Figure 7.

Intersubunit disulphide trapping at the α₁/γ₂ coupling interface reduces flurazepam potentiation of I_GABA. Modulation of GABA EC₅ currents by 10 µM flurazepam (FZM; A), or 1 µM 3-carbomethoxy-4-ethyl-6,7-dimethoxy-β-carboline (DMCM) (C), from oocytes injected with wild-type (WT) or mutant receptors. Traces are scaled so that the GABA EC₅ current response is the same for each receptor or condition to compare differences and changes in potentiation. Dithiothreitol (DTT) and H₂O₂ treatment of WT and double cysteine mutant receptors was performed as described in the Methods. Modulation of GABA current was calculated as [ (I_GABA+BZD/I_GABA) − 1]. Values for flurazepam potentiation prior to DTT treatment are reported in Table 2. (B) Changes in flurazepam potentiation following treatment with DTT for each receptor and the corresponding single-cysteine mutants are shown in graphical form. The effect of DTT treatment was calculated as: [ ( (flurazepam potentiation_after/flurazepam potentiation_before) − 1) × 100]. Note that DTT treatment significantly increased flurazepam potentiation for both α₁D56Cβ₂γ₂L198C and α₁D56Cβ₂γ₂P64C receptors compared with WT and the corresponding single-cysteine mutants (**P < 0.01). Data represent mean ± SD from at least three separate experiments. (C) Modulation of GABA EC₅ currents by 1 µM DMCM are shown for WT and mutant receptors.

For the rest of the α/γ double-cysteine mutants, DTT and H₂O₂ had no effect on flurazepam potentiation compared with WT receptors (data not shown). It is possible that these cysteines formed non-reducible disulphide bonds. For these mutations, we could not use MTS accessibility to probe whether the cysteines were free or disulphide-linked because one of the engineered cysteine residues is in the α subunit. Because each GABA_A receptor has two α subunits (Figure 1A), even if the introduced cysteine at the α/γ interface was trapped in a disulphide bond, the cysteine at the α/β interface would still be freely accessible to react with MTS, which would complicate the analysis.

Disulphide trapping α₁Loop 2 to γ₂β-strand 1 inhibits I_GABA

We also tested the effect of DTT and H₂O₂ on I_GABA (EC₅₀) for the 16 α/γ double-cysteine mutants. Only one double-cysteine mutant, α₁D56C/γ₂P64C (Figure 8A) was significantly affected by DTT, which increased the amplitude of I_GABA-EC50 (Figure 8C). H₂O₂ completely reversed this effect, restoring I_GABA to its initial level. A second application of DTT increased I_GABA again about 50%. DTT had no effect on the corresponding single-cysteine mutants (Figure 8C). We conclude that the effects of DTT and H₂O₂ on I_GABA are due to the formation and reduction of a disulphide bond between α₁D56C and γ₂P64C.

Figure 8.

Disulphide trapping α₁Loop 2 to γ₂β-strand 1 inhibits GABA activation. Representative current traces from wide-type (WT) and α₁D56Cβ₂γ₂P64C receptors elicited by GABA EC₅₀ (A) or 10 mM GABA (∼GABA EC₉₉) (B) before and after treatment with dithiothreitol (DTT) and H₂O₂. (C) Changes in GABA EC₅₀ current following treatment with DTT are shown. The effect of DTT treatment was calculated as: [ ( (I_GABAafter /I_GABAbefore) − 1) × 100]. Note that DTT treatment significantly increased I_GABA for α₁D56Cβ₂γ₂P64C receptors compared with WT and the corresponding single-cysteine mutants (**P < 0.01). Data represent mean ± SD from at least four separate experiments.

The peak amplitudes elicited by a saturating concentration of GABA (10 mM) from oocytes expressing α₁D56Cβ₂γ₂P64C receptors were relatively small (<5 µA) compared with WT receptors (>15 µA) (Figure 8). To determine whether the increase in I_GABA following treatment with DTT was due to a shift in the GABA concentration–response curve and/or changes in maximum current amplitude, we measured I_GABA at low and saturating concentrations of GABA to estimate GABA EC₅₀ values and to determine maximum current before and after treatment with DTT. We found that DTT had a small effect (<20% change) on the calculated GABA EC₅₀ value but that the maximum current amplitude from saturating GABA was significantly increased by DTT (Figure 8B), suggesting that crosslinking α₁Loop 2 (D56C) to γ₂β-strand 1 (P64C) reduces the efficacy of GABA to activate the channel.

Discussion

Although the functional effects of BZDs have been well characterized, the protein movements underlying BZD modulation of GABA_A receptor current are largely undefined. In a previous study, we demonstrated that potentiation of I_GABA by positive BZD modulators initiates distinct movements in γ₂Loop 9 (Hanson and Czajkowski, 2008). Here, we used disulphide trapping to examine how movements in Loop 9 and the nearby α₁/γ₂ coupling interface influence BZD modulation of I_GABA. Disulphide trapping is a powerful approach for probing protein dynamics of both soluble and membrane-bound proteins in their native environment (Bass et al., 2007) and has been used to study movements in the GABA_A receptor transmembrane segments triggered by both GABA and anaesthetic channel activation (Horenstein et al., 2001; Bera and Akabas, 2005; Horenstein et al., 2005; Jansen and Akabas, 2006; Rosen et al., 2007; Yang et al., 2007; Bali et al., 2009).

Reorganization of the α₁/γ₂ coupling interface accompanies BZD potentiation of I_GABA

Based on our electrophysiological assays, disulphide bonds formed between γ₂S187C/L206C, γ₂S195C/F203C, γ₂S202C/S230C, α₁D56C/γ₂P64C and α₁D56C/γ₂L198C (Figures 2,3,5–7). Intrasubunit disulphide trapping γ₂Loop 9 to β-strand 9 in α₁β₂γ₂S187C/L206C and α₁β₂γ₂S195C/F203C receptors significantly reduced maximal flurazepam potentiation (Figures 2 and 9 and Table 1), indicating that modulation of I_GABA by flurazepam involves movements of γ₂Loop 9 relative to γ₂β-strand 9. Because Loop 9 sits near the BZD binding pocket (Figures 1 and 9), tethering Loop 9 to β-strand 9 likely impairs flurazepam potentiation by preventing initial structural rearrangements of Loop 9 induced by flurazepam binding. Interestingly, crosslinking γ₂Loop 9 to β-strand 9 reduced flurazepam potentiation without affecting zolpidem potentiation (Figure 4). These data suggest that zolpidem binding initiates movement in a different part of Loop 9 (i.e. unaffected by crosslinking γ₂S195C/F203C and γ₂S187C/L206C). Consistent with this idea, in a previous study, we showed that mutations in γ₂Loop 9 differentially affect flurazepam and zolpidem potentiation of I_GABA without altering flurazepam and zolpidem binding (Hanson and Czajkowski, 2008). In addition, unlike other BZD-site ligands, zolpidem appears to be able to adopt several favourable orientations in the BZD binding pocket (Hanson et al., 2008). Thus, depending on its orientation, the zolpidem binding site movements that are transduced to the rest of the protein may differ.

Figure 9.

Summary of the effects of disulphide bond formation at the GABA_AR α/γ interface. Homology model of the α₁ (red) and γ₂ (blue) extracellular domains of the GABA_AR. The general location of the benzodiazepine (BZD) binding site is indicated. The loops at the coupling interface are highlighted as in Figure 1: γ₂Loop 9, purple; β-strand 9, orange. γ₂pre-M1, yellow; γ₂Loop 7, red; α₁Loop2, green; α₁Loop7, blue. Disulphide bonds that inhibit (-) and enhance (+) BZD potentiation of I_GABA are shown. Bonds that affect both flurazepam and zolpidem potentiation are indicated by a blue (-) (α₁D56Cβ₂γ₂L198C) or blue (+) (α1β₂γ₂S202C/S230C), while those that only affect flurazepam potentiation are indicated by a red (-)(α₁β₂γ₂S187C/L206C and α₁β₂γ₂S195C/F203C).

In contrast, intrasubunit disulphide trapping γ₂β-strand 9 (S202C) to the γ₂ pre-M1 region (S230C) significantly enhanced both flurazepam and zolpidem potentiation of I_GABA (Figures 5 and 9, Table 1), indicating that tethering β-strand 9 to the pre-M1 region (β-strand 10) traps the receptor in a more favourable conformation for BZD positive modulation. In related glycine receptors, Zn²⁺ binding to a site between β-strand 9 and β-strand 10 enhances receptor function supporting the idea that tethering β-strand 9 to β-strand 10 stabilizes the protein in a conformation that favours activation (Miller et al., 2005). Intersubunit disulphide bonds formed at the α/γ coupling interface between α₁Loop 2 and γ₂Loop 9 (α₁D56C/γ₂L198C) prevent flurazepam and zolpidem from efficiently modulating I_GABA (Figures 7 and 9). Based on these data, we speculate that binding of BZD-site agonists destabilizes intersubunit interactions at the α/γ subunit coupling interface allowing movement of γ₂β-strand 9 towards β-strand 10 and the pre-M1 region.

As none of the above disulphide bonds affected negative modulation of I_GABA by DMCM (Figures 4,5,7, and S2), rearrangements in the α₁/γ₂ coupling interface are likely not to be responsible for the actions of negative BZD modulators. This is consistent with our previous findings that residues in the γ₂ pre-M1 region and M2-M3 loop, which are required for BZD potentiation of I_GABA, are not critical for inhibition of I_GABA by DMCM (Boileau and Czajkowski, 1999), and that only positive BZD modulators induce structural rearrangements in the γ₂Loop 9 region near the coupling interface (Hanson and Czajkowski, 2008).

Changes in maximal BZD potentiation of I_GABA can arise via several mechanisms, including alterations in BZD binding, changes in coupling BZD binding to potentiation, and/or shifts in GABA EC₅₀. While we cannot completely rule out effects on BZD binding, we believe that the effects are likely not due to changes in BZD binding. In previous studies, we demonstrated that residues in the γ₂Loop 9 region are not involved in BZD binding using radioligand binding assays (Hanson and Czajkowski, 2008; Hanson et al., 2008). In addition, the mutations are located at the coupling interface and are not near the BZD binding site (Figure 1). Moreover, because our experiments were done at saturating concentrations of BZD, the data demonstrate that the mutations alter efficient coupling of BZD binding to modulation of I_GABA (BZD efficacy). The increases in flurazepam and zolpidem potentiation for γ₂S202C/S230C are also not due to changes in GABA dose–response, because the mutation had no effect on GABA EC₅₀ (Table 1). Some mutations caused small but significant changes in GABA EC₅₀ raising the possibility that the changes in BZD potentiation observed are a consequence of GABA EC₅₀ alterations. BZD positive modulators enhance GABA_A receptor current by decreasing GABA EC₅₀, whereas negative modulators, like DMCM, inhibit GABA_A receptor current by increasing GABA EC₅₀. Using a simple model, if a mutation and/or disulphide bond shifts the GABA dose–response curve to the left, one would predict that the perturbation would decrease both flurazepam and zolpidem potentiation and would enhance DMCM inhibition. Intrasubunit disulphide bonds between γ₂S187C/L206C and γ₂S195C/F203C decreased GABA EC₅₀ (1.8- and 3.5-fold respectively). While we observed a significant decrease in flurazepam potentiation of I_GABA for both mutants, zolpidem and DMCM modulation were unchanged (Figure 4, Table 1) indicating that the shift in GABA EC₅₀ was not responsible for the observed changes in flurazepam potentiation. The intersubunit disulphide bond between α₁Loop 2 and γ₂Loop 9 (α₁D56C/γ₂L198C) also decreased GABA EC₅₀ (1.7-fold, Table 2). In this case, the disulphide bond decreased flurazepam and zolpidem potentiation but had no effect on DMCM inhibition, again suggesting that a simple shift in GABA EC₅₀ is not responsible. In many cases, the mutations affects on GABA EC₅₀ and BZD potentiation were not correlated; some mutations significantly altered BZD potentiation without affecting GABA EC₅₀ whereas others affected GABA EC₅₀ without changing BZD potentiation (Tables 1, 2, and S1). Together, these observations support the idea that the mutations and disulphide bonds formed in this region specifically affect coupling of BZD binding to positive modulation of I_GABA and that the α/γ interface is an important part of the transduction pathway for BZD potentiation. This is consistent with our previous data demonstrating that flurazepam and zolpidem significantly slowed covalent modification of γ₂R197C in Loop 9, whereas DMCM, GABA and the allosteric modulator pentobarbital had no effects (Hanson and Czajkowski, 2008), indicating that movements in this region are specific for positive BZD modulators.

GABA-induced channel activation requires movements at the α₁/γ₂ coupling interface

Disulphide trapping α₁Loop 2 (α₁D56C) to γ₂β-strand 1 (γ₂P64C) significantly decreased maximal GABA-induced current (Figure 8). Changes in GABA I_max can be attributed to changes in GABA binding (k_on, k_off), channel gating (α or β), and/or single-channel conductance. At high saturating agonist concentrations, the maximal current activation rate is limited by the channel opening step or β. Thus, an increase in I_max following DTT treatment is consistent with changes in channel gating or channel conductance. Because the disulphide bond is only at a single-subunit interface and is localized away from the channel vestibule, a change in gating is the simplest explanation. Detailed kinetic analyses are required to quantitatively tease apart the effects of this mutation on microscopic binding affinity and channel gating properties. Nonetheless, the data indicate that although the γ₂ subunit does not directly participate in GABA binding, movements at the single α₁/γ₂ coupling interface are required for efficient coupling of GABA binding to channel activation. Comparison of the crystal structures of the related bacterial pentameric ligand-gated ion channels, ELIC and GLIC, in presumed closed and open channel conformations suggests that channel activation is accompanied by a rearrangement of loops in the coupling interfaces including a downward motion of Loop 2 and outward movements of the M2-M3 loop (Bocquet et al., 2009; Hilf and Dutzler, 2009). Thus, tethering α₁Loop 2 to γ₂β-strand 1 may prevent this downward motion of Loop 2, whereas tethering α₁Loop 2 to γ₂ Loop 9 (α₁D56C/γ₂L198C) does not. α₁D56C/γ₂P64C also decreased maximal flurazepam and zolpidem potentiation of I_GABA (Table 2) consistent with this region being involved in BZD positive modulator actions.

Summary and conclusions

In summary, the data in this study provide new insights into the transduction pathway for BZD allosteric modulation of the GABA_A receptor. We found that mutations throughout the α/γ coupling interface reduce BZD potentiation of I_GABA, supporting the idea that this region is important for efficient coupling of BZD binding to modulation of I_GABA. We demonstrate that restricting local movements in the α/γ coupling interface through disulphide trapping specifically affects potentiation of I_GABA by BZD-site agonists and does not affect negative modulation by BZD inverse agonists, demonstrating that the allosteric pathways for positive and negative modulation of I_GABA by BZDs are different. Our data support a mechanism of BZD action in which coupling between BZD-site agonist binding and the potentiation of I_GABA occurs by movements of γ₂Loop 9, γ₂β-strand 9, the γ₂ pre-M1 region and Loop 2 in the adjacent α₁ subunit (Figure 9).

Based on the crystal structures of ELIC and GLIC in presumed closed and open conformations (Bocquet et al., 2009; Hilf and Dutzler, 2009) as well as comparison of low-resolution nicotinic acetylcholine receptor structures with and without agonist (Unwin, 2005), channel activation appears to be accompanied by a reorganization of subunit–subunit interfaces. We envision that neurotransmitter binding destabilizes subunit–subunit interface interactions, which help maintain the resting/closed channel state, and that this may be a common mechanism underlying pentameric LGIC activation. In a related fashion, we speculate that the actions of allosteric drug modulators are also driven by realignments at subunit–subunit interfaces. While BZDs do not activate GABA_A receptors directly, our data demonstrate that positive BZD-modulator binding at the α/γ interface induces rearrangements of the α/γ coupling interface. Moreover, our data suggest that constraining movement of the α/γ coupling interface inhibits GABA-induced channel activation. Thus, BZD-induced movements at the α/γ coupling interface are likely to synergize with the rearrangements induced by GABA binding at the β/α subunit interfaces to enhance channel activation by GABA.

Glossary

Abbreviations

BZD

benzodiazepine

DMCM

3-carbomethoxy-4-ethyl-6,7-dimethoxy-β-carboline

DTT

dithiothreitol

WT

wild-type

Conflicts of interest

None to declare.

Supporting information

Additional Supporting Information may be found in the online version of this article:

Teaching Materials; Figs 1–9 as PowerPoint slide.

Figure S1 DTT does not significantly affect FZM potentiation of double cysteine mutants. Changes in FZM potentiation following treatment with DTT is graphed for WT and mutant receptors. The effect of DTT treatment was calculated as: [ ( (FZM potentiation_after /FZM potentiation_before) − 1) × 100]. Note that DTT treatment did not significantly change FZM potentiation for the mutant receptors compared to WT receptors.

Figure S2 DMCM modulation of WT and mutant GABA_A receptors. Maximal inhibition of GABA EC₅ responses by DMCM (1 μM) is graphed for WT and mutant receptors. Inhibition of GABA current was calculated as [ (I_GABA+DMCM/I_GABA) − 1]. DMCM inhibition for the mutant receptors was not significantly different from WT.

Figure S3 FZM potentiation of I_GABA is significantly affected by mutations in the α₁/γ₂ coupling interface. Left panels (A,D,F), Models of the GABA_AR α₁ (red) and γ₂ (blue) subunit coupling interface are shown. Loops and β-strands are coloured as in Figure 1: γ₂Loop 9, purple; γ₂ β-strand 9, orange; γ₂pre-M1, yellow; γ₂Loop 7, red; α₁Loop2, green; α₁Loop7, blue. Residues mutated to cysteine are shown in stick representation except in panel (A) where only every other Loop 9 residue is shown for clarity. Right panels (B,C,E,G), Potentiation of GABA EC₅ current by 10 μM FZM is graphed for WT (×), single-cysteine mutant receptors (•), and double cysteine mutant receptors (○). A dashed line indicates the level of potentiation for the single α₁ subunit cysteine mutant that is expressed with multiple γ₂ mutant subunits. Data represent mean ± SD from at least three separate experiments. (**P<0.01), FZM potentiation of the double cysteine mutant receptor was significantly reduced compared to both single-cysteine mutants. Potentiation values were calculated as [ (I_GABA+FZM/I_GABA) − 1] and are reported in Table S1.

Table S1 Summary of GABA, flurazepam, and zolpidem data from WT and α₁/γ₂ cysteine mutant receptors used for inter-subunit disulphide bonds.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.