Allosteric modulators induce distinct movements at the GABA-binding site interface of the GABA-A receptor

Abstract

Benzodiazepines (BZDs) and barbiturates exert their CNS actions by binding to GABA-A receptors (GABARs). The structural mechanisms by which these drugs allosterically modulate GABAR function, to either enhance or inhibit GABA-gated current, are poorly understood. Here, we used the substituted cysteine accessibility method to examine and compare structural movements in the GABA-binding site interface triggered by a BZD positive (flurazepam), zero (flumazenil) and negative (3-carbomethoxy-4-ethyl-6, 7-dimethoxy-β-carboline, DMCM) modulator as well as the barbiturate pentobarbital. Ten residues located throughout the GABA binding site interface were individually mutated to cysteine. Wild-type and mutant α₁β₂γ₂ GABARs were expressed in Xenopus laevis oocytes and functionally characterized using two-electrode voltage clamp. We measured and compared the rates of modification of the introduced cysteines by sulfhydryl-reactive methanethiosulfonate (MTS) reagents in the absence and presence of BZD-site ligands and pentobarbital. Flurazepam and DMCM each accelerated the rate of reaction at α₁R131C and slowed the rate of reaction at α₁E122C, whereas flumazenil had no effect indicating that simple occupation of the BZD binding site is not sufficient to cause movements near these positions. Therefore, BZD-induced movements at these residues are likely associated with the ability of the BZD to modulate GABAR function (BZD efficacy). Low, modulating concentrations of pentobarbital accelerated the rate of reaction at α₁S68C and β₂P206C, slowed the rate of reaction at α₁E122C and had no effect at α₁R131C. These findings indicate that pentobarbital and BZDs induce different movements in the receptor, providing evidence that the structural mechanisms underlying their allosteric modulation of GABAR function are distinct.

1. Introduction

GABA_A receptors (GABARs), besides being activated by the neurotransmitter GABA, are modulated by numerous therapeutically important drugs, including barbiturates, anesthetics, and benzodiazepines (BZDs). These compounds are allosteric modulators as they bind to sites distinct from the orthosteric GABA binding sites to potentiate or inhibit GABA-evoked current. Significant strides have been made in identifying the binding sites for these modulators (see (Hemmings et al., 2005; Mitchell et al., 2008; Sigel, 2002) for reviews). However, the structural mechanisms that couple the binding of these modulators to the allosteric modulation of GABA-gated current are less well understood.

BZDs have varying effects on GABAR function. BZD-site agonists (e.g. diazepam and flurazepam), act as positive modulators to enhance GABA-gated current (I_GABA). BZD-site inverse agonists (e.g. DMCM) act as negative modulators to inhibit I_GABA, while BZD-site antagonists (e.g. flumazenil) bind but have no effect on I_GABA and thus act as “zero” modulators. The BZD binding site is located in the extracellular amino terminal domain, at the interface between the GABAR α and γ subunits (Pritchett et al., 1989; Smith and Olsen, 1995). Previous studies have identified residues in Loop F/9, the M2-M3 extracellular loop and the pre-M1 region of the γ2 subunit that are required for enhancement of I_GABA by BZD positive modulators (Boileau and Czajkowski, 1999; Boileau et al., 1998; Hanson and Czajkowski, 2008; Jones-Davis et al., 2005; Padgett and Lummis, 2008). Interestingly, these γ2 subunit residues/regions are not critical for inhibition of I_GABA by the BZD-site negative modulator, DMCM (Boileau and Czajkowski, 1999; Hanson and Czajkowski, 2008), suggesting that the structural pathways mediating BZD negative allosteric modulation are distinct from those involved in positive modulation.

Depending on the electrophysiological approach and kinetic models used, investigators have asserted that BZDs exert their allosteric effects on I_GABA by either altering the microscopic binding of GABA at the orthosteric site (Goldschen-Ohm et al., 2010; Lavoie and Twyman, 1996; Mellor and Randall, 1997; Rogers et al., 1994; Thompson et al., 1999; Twyman et al., 1989), or by shifting the closed to open state channel equilibrium of the agonist-bound receptor (Campo-Soria et al., 2006; Downing et al., 2005; Rusch and Forman, 2005). Regardless of the mechanism, because GABA binding and channel gating are energetically coupled (i.e. the open channel state of the GABAR has a higher affinity for GABA than the closed state), one would predict that BZD actions would result in structural rearrangements of the GABA binding site. Recently, we demonstrated that the GABA binding site undergoes structural rearrangement upon binding the positive BZD modulator flurazepam (Kloda and Czajkowski, 2007).

Here, we tested whether positive, negative, and zero allosteric modulators trigger distinct conformational movements within the GABA-binding site interface that are correlated with their efficacy. We mutated ten residues throughout the GABA binding site interface to cysteine (Fig. 1). We then measured the rate at which the introduced cysteines were modified by sulfhydryl-specific reagents in the absence and presence of flurazepam (BZD positive modulator), DMCM (BZD negative modulator), and flumazenil (BZD zero modulator) to probe the dynamics of the GABA binding interface upon binding BZD modulators that elicit different pharmacological effects. We also examined whether low concentrations of the positive allosteric modulator, pentobarbital, which binds at a site distinct from both the BZD and GABA binding sites (Amin, 1999; Belelli et al., 1999; Serafini et al., 2000), induces similar structural rearrangements at the GABA binding site interface as a BZD positive modulator.

Figure 1. Location of residues substituted to cysteine at the GABA binding interface of the GABAR.

(A) Homology model of the α₁ (cyan), β₂ (green) and γ₂ (light grey) subunits of the GABAR with residues substituted to cysteine labeled and highlighted in red. The region in the α₁ subunit that links the BZD and GABA binding interfaces (denoted LINKER) is highlighted in dark blue. (B) Partial sequences of α₁ (cyan) and β₂ (green) GABA binding site loop regions A (β-strand 4), B (β-strand 7), C (β-strand 10), D (β-strand 2), and E (β-strand 5–6), with residues substituted to cysteine in red (and denoted with the letter C below the residue). Numbering is based on mature protein sequences. LINKER residues are in dark blue, and residues previously implicated in lining the GABA binding pocket are underlined.

2. Materials and Methods

2.1. Site directed mutagenesis

Rat cDNAs encoding for the GABAR α₁, β₂ and γ₂ subunits were used in this study. α₁ and β₂ cysteine mutants were engineered using the Altered Sites II in vitro Mutagenesis Systems (Promega Corp., Madison, WI) or by recombinant PCR, as described previously (Boileau et al., 1999; Kucken et al., 2000). Cysteine substitutions were made in the rat α₁ subunit at positions T60, S68, E122, R131 and the rat β₂ subunit at positions L99, T160, G203, P206, R207 and S209 (Fig. 1) (Boileau et al., 1999; Boileau et al., 2002; Holden and Czajkowski, 2002; Kloda and Czajkowski, 2007; Newell et al., 2004; Wagner and Czajkowski, 2001). The α₁ and β₂ cysteine mutants were subcloned into pGH19 (Liman et al., 1992; Robertson et al., 1996) for expression in Xenopus laevis oocytes. All mutants were verified by restriction digest and double-stranded cDNA sequencing. All α₁ and β₂cy steine mutants have been named, using the single letter code, as wild-type residue, residue number (in the mature subunit protein), and mutated residue.

2.2. Expression in oocytes

Xenopus laevis oocytes were prepared, and then injected with cRNA as described previously (Boileau et al., 1998). In brief, cRNA encoding for the wild-type (WT) and mutant subunits were synthesized by in vitro transcription from NheI linearized cDNA template using the mMessage mMachine T7 kit (Ambion, Austin, TX). Oocytes were prepared by incubation in a collagenase solution (2mg/ml in ND96/Ca²⁺-free media containing 96mM NaCl, 2mM KCl, 1mM MgCl₂ and 5mM HEPES, pH7.6), followed by several washes in recording solution. Within 1–2 days, WT or mutant α₁ and WT or mutant β₂ subunits were expressed with WT γ₂ subunits by injecting 25nl of αβγ cRNA mix into oocytes (at 5–25ng/ml cRNA for α₁ and β₂, and 50–250ng/μl cRNA for γ₂ to achieve a final ratio of 1:1:10/α₁β₂γ₂). Oocytes were stored at 17–19 °C in recording solution (ND96 with 1.8mM CaCl₂; ND96/Ca²⁺) Supplemented with 100μg/ml gentamicin and 100μg/ml bovine serum albumin for 2–14 days and used for electrophysiological recordings.

2.3. Voltage Clamp analysis

Oocytes were held at −80mV under two-electrode voltage clamp while continuously perfused with ND96/Ca²⁺ at a rate of 5ml/min in a bath volume of 200μl. Stock solutions of Flurazepam (RBI, Natick, MA), pentobarbital (Sigma, St. Louis, MO), and GABA (Sigma, St. Louis, MO) were dissolved in ND96/Ca²⁺. Stock solutions of DMCM (RBI, Natick, MA), flumazenil (RBI, Natick, MA), N-biotinylaminoethyl methanethiosulfonate (MTSEA-biotin) and 2-sulfonatoethyl methanethiosulfonate (MTSES) (Toronto Research Chemicals, Toronto, Ontario, Canada) were prepared in DMSO and subsequently diluted in ND96/Ca²⁺ for working concentrations where the final [DMSO] (≤ 2%) did not affect GABAR function. Borosilicate electrodes (Warner Instruments, Hamden, CT) were filled with 3M KCl and had resistances between 0.5 and 2MΩ. Electrophysiological data were acquired with a GeneClamp 500 (Molecular Devices, Sunnyvale, CA) interfaced to a computer with a Digidata®1200 analog-to-digital device (Axon Instruments/Molecular Devices, Sunnyvale, CA) and recorded using the WinWCP Strathclyde Whole Cell Program, version 3.5.5 (kindly provided by J. Dempster, University of Strathclyde, Glasgow, UK).

GABA EC₅₀ values for mutant and wild-type GABARs were previously established (Boileau et al., 1999; Boileau et al., 2002; Holden and Czajkowski, 2002; Kloda and Czajkowski, 2007; Newell et al., 2004; Wagner and Czajkowski, 2001). As such, for each mutant receptor, GABA EC₅₀ was estimated by examining the ratio of response to a low concentration of GABA (1–100μM, dependent on mutant) versus a high or maximal concentration of GABA (10mM-600mM, dependent on mutant). We then calculated GABA EC₅₀ for each mutant (with previously measured Hill coefficients and the low/high GABA current ratios determined) using the standard Hill equation:

$$\theta ={[{\text{GABA}}_{\text{low}}]}^{\text{nH}}/({{\text{EC}}_{50}}^{\text{nH}}+{[{\text{GABA}}_{\text{low}}]}^{\text{nH}}),$$

where nH is the Hill coefficient, GABA_low is the low test concentration of GABA, and θ is the ratio of current response to a low concentration of GABA versus a high or maximal concentration of GABA.

Flurazepam and pentobarbital potentiation, DMCM inhibition, or flumazenil (Ro15–1788) antagonism of BZD-potentiation of IGABA were recorded at GABA EC₁₅. Modulation of GABA current is reported in fold-changes, defined as I_GABA+DRUG(s)/I_GABA, where I_GABA+DRUG(s) is the current response in the presence of the drug(s) tested, and I_GABA is the control GABA-induced current. Maximal potentiation of I_GABA was measured with 10μm flurazepam or 25–50μM pentobarbital. Maximal inhibition of I_GABA was measured with 100μM or 1μM DMCM. Flumazenil antagonism of BZD-induced modulation of I_GABA was measured by co-applying 1μM flumazenil with 10μm flurazepam or 1μM DMCM.

2.4. Rate of MTS modification

We previously established that all of the introduced cysteines were accessible to MTS modification (Boileau et al., 1999; Boileau et al., 2002; Holden and Czajkowski, 2002; Kloda and Czajkowski, 2007; Newell et al., 2004; Wagner and Czajkowski, 2001). GABA current (I_GABA) for each oocyte was stabilized before addition of MTS reagents by applying successive pulses of GABA, separated by 2min washes, until I_GABA varied by <5% to ensure that the observed changes in I_GABA after MTS application were due to MTS modification of the introduced cysteine. The rate of MTS covalent modification of the introduced cysteines in the absence of drug was determined by measuring the outcome of sequential applications of MTS reagents on I_GABA. The protocol was as follows: A concentration of EC_20–30 GABA (for MTS potentiation of I_GABA) or EC_50–65 (for MTS inhibition of I_GABA)was applied for 5 sec; the cell was washed for 1min 40sec-1min 53sec; MTS reagent was applied for 7–20 sec; the cell was washed for 1 min; and the procedure was repeated until I_GABA no longer changed indicating that the reaction reached completion. Time and concentration of MTS application for each mutant was chosen such that at least six I_GABA measurements were obtained before the reaction reached completion, where most importantly the rise/decay phase of the exponential curve was defined by at least 3 data points. As such, concentrations of MTS reagent and time of application at each cysteine varied as follows: α₁T60C-500μM MTSEA-biotin, 10 sec; α₁S68C-100μM MTSES, 7 sec; α₁E122C-1μM MTSEA-biotin, 10 sec; α₁R131C-100μM MTSEA-biotin, 10 sec; β₂L99C-300μM MTSEA-biotin, 10 sec; β₂T160C-200nM MTSEA-biotin, 10 sec; β₂G203C-400nM MTSEA-biotin, 10 sec; β₂P206C-200μM MTSEA-biotin, 10 sec; β₂R207C-200μM MTSEA-biotin, 10 sec; β₂S209C-200μM MTSEA-biotin, 20 sec.

The effects of modulators on the rates of MTS modification were tested by co-applying 10μM flurazepam, 100nM-1μM DMCM, flumazenil (1μM or 25–50μM pentobarbital with MTSEA-biotin, with the exception of α₁S68C, in which case they were co-applied with MTSES. Pentobarbital concentrations used in rate experiments were chosen such that they did not elicit more than 5% of maximal I_GABA (i.e., were not activating). DMCM concentrations were chosen based on the effectiveness of DMCM-induced I_GABA inhibition and wash-out. Thus, the lower concentration of DMCM (100nM), used in rate experiments at α₁E122Cβ₂γ₂ and α₁P206Cβ₂γ₂ receptors, was saturating and significantly inhibited I_GABA (0.45-fold) but more effectively washed-out relative to 1μM DMCM. For these studies, I_GABA was stabilized before MTS application as follows: A fixed concentration of EC₂₀-EC₆₅ GABA was applied for 5 sec; the cell was washed for 1min 40 sec-1min 53 sec; flurazepam, DMCM, flumazenil, or pentobarbital was applied for 7–20 sec; the cell was washed for 1–5 min; and the procedure was repeated until I_GABA was stabilized (e.g. I_GABA varied <5%). This procedure was subsequently repeated for MTS rate measurements while successively co-applying MTS with drug until the reaction reached completion (i.e. no additional change in I_GABA). For each mutant, application times were kept constant between control (MTS alone) and experimental (MTS+drug) conditions, while MTS concentrations were varied as necessary. At least two different concentrations of MTS were used for rate measurements, to verify that the rates were independent of the MTS-reagent concentration. In all cases, the second order rate constants were independent of the reagent concentrations.

For all rate experiments, the change in I_GABA (decrease or increase) after each reagent application was plotted versus cumulative time of MTS exposure. The change in I_GABA (ΔI_GABA) at each time-point was defined as: ΔI_GABA= (I_GABA/I_GABAo)−1, where I_GABA is the peak GABA current measured after each successive MTS application and I_GABAo refers to peak stabilized GABA current prior to MTS application at time 0. The ΔI_GABA versus cumulative time of MTS exposure was fit with a single-exponential association or decay function: y= span × e^−kt + plateau, where k is the pseudo-first-order rate constant of the reaction (Prism, GraphPad Software Inc., San Diego, CA). Plateau is the ΔI_GABA at the end of the reaction when I_GABA is no longer changing and is defined by the curve fit. Because ΔI_GABA is defined relative to I_GABA at time zero (i.e. ΔI_GABA= (I_GABA/I_GABAo)−1), span=0-plateau. The second-order rate constants (k₂) were determined by dividing k by the concentration of MTS reagent used in the rate measurement (Pascual and Karlin, 1998) and are reported in Table 1. For easier visual comparisons between control and experimental data sets, the data in Figures 2, 3, and 4, were normalized to the maximum effect of the MTS reagent, where the ΔI_GABA at each time point was normalized to the plateau: [(I_GABA/I_GABAo)-1]/plateau.

TABLE 1. Rates of MTS modification of cysteines introduced in the GABA binding site interface of the GABAR in the presence of BZD modulators Flurazepam and DMCM, and low concentrations of modulatory Pentobarbital.

Rate of MTS reaction with introduced cysteines were measured, and second-order rate constants (k₂; M⁻¹s⁻¹) were calculated as described under Materials and Methods Second-order rate constants reflect the means±S.E.M

Receptor	Loop	Control		Flurazepam		DMCM		Pentobarbital
		k2	n	k2	n	k2	n	k2	n
α1T60Cβ2 γ2	D	74.4±5.2	4	85.5±3.3	5	70.6±12	3	68.8±0.73	3
α1S68Cβ2γ2	D	407±8.1	4	432±44	3	429±23	3	526±4.7*	3
α1R122Cβ2γ2	E	42703±4466	3	25.507±I730*	3	25,267±2392*	3	28,680±3396*	4
α1R13lCβ2γ2	E	455±49	3	953±78**	4	S45±100**	3	575±27	3
α1β2L99Cγ2	A	163±14	4	151±32	3	202±22	3	157±14	3
α1β2T160Cγ2	B	154525±9643	4	131570±15325	5	132456±12409	3	164800±12127	3
α1β2G203Cγ2	C	103.800±10018	3	103.967±8714	3	85138±2709	3	93.558±789
α1β2P206Cγ2	C	159±6.4	4	255±29	5	260±35	5	300±20**	5
α1β2R207Cγ2	C	283±30	5	305±16	3	343±37	4	358±31	3
α1β2S209Cγ2	C	87.8±6.8	4	80.4±5.5	3	94.8±4.3	3	106.7±3.7	3

^*Rates of reaction in the presence of ligand that were statistically significant from control, where p < 0.5.

^**Rates of reaction in the presence of ligand that were statistically significant from control, where p< 0.01.

Figure 2. Effects of allosteric modulators on the rate of MTSEA-Biotin modification of α₁ subunit cysteine substitutions in loop E.

(A) Representative GABA evoked (~EC_20–30) currents following successive 10 second MTSEA-biotin applications (arrows) in the absence (top traces) and presence (bottom traces) of 10μM flurazepam at α₁E122Cβ₂γ₂ receptors. (B) Representative rates of reaction of MTSEA-biotin in the absence or presence of 10μM flurazepam, 100nM DMCM, or 50μM pentobarbital at α₁E122β₂γ₂mutant receptors. Allosteric modulators significantly slowed the rate of reaction at position 122. (C) Representative GABA-evoked (~EC₅₀) currents following successive MTSEA-biotin applications (arrows) in the absence (top traces) and presence (bottom traces) of 10μM flurazepam at α₁R131Cβ₂γ₂ receptors. (D) Representative rates of reaction of MTSEA-biotin in the absence and presence of 10μM flurazepam or 25μM pentobarbital at α₁R131Cβ₂γ₂ receptors. Rate experiments were preformed as described under Materials and Methods. Increases or decreases in I_GABA (panels B & D) were plotted versus cumulative time of MTS exposure. Data were normalized and fit to a single-phase exponential decay as described in Materials and Methods. As multiple concentrations of MTS were used for the rate experiments, only representative rate plots, all obtained using the same concentration of MTS reagent, are shown for graphical clarity and comparative accuracy. For comparison of averaged normalized rates, see Figure 5.Se cond-order rate constants for MTS modification of α₁ subunit cysteine substitutions are summarized in TABLE 1.

Figure 3. Effects of allosteric modulators on the rate of MTS modification of α₁ subunit cysteine substitutions in loop D.

(A) Representative rates of reaction of MTSES in the absence and presence of 10μM flurazepam, 1μM DMCM, or 50μM pentobarbital at α₁S68Cβ₂γ₂ mutant receptors. (B) Representative rates of reaction of MTSEA-biotin in the absence and presence of 10μM flurazepam, 1μM DMCM, and 50μM pentobarbital at α₁T60Cβ₂γ₂ mutant receptors. Allosteric modulation by pentobarbital, but not BZD-site ligands DMCM or flurazepam, significantly slowed the rate of MTS reaction at S68C. The rate of MTS reaction at T60C was not significantly altered in the presence of any allosteric modulator tested. Data were acquired, normalized, and fit to a single-phase exponential decay as described in Fig 2 and Materials and Methods. Second-order rate constants for MTS modification of α₁ subunit cysteine substitutions are summarized in TABLE 1.

Figure 4. Effects of allosteric modulators on the rate of MTSEA-biotin modification of β₂ subunit cysteine substitutions in loop C.

(A) Representative rates of reaction of MTSEA-biotin in the absence and presence of 10μM flurazepam, 100nM DMCM, and 50μM pentobarbital at α₁β₂P206Cγ₂ receptors. Of the loop C residues tested, only position 206 displayed altered rates of reaction in the presence of allosteric modulators. (B) Representative rates of reaction of MTSEA-biotin in the absence and presence of 10μM flurazepam, 1μM DMCM, and 25μM pentobarbital at α₁β₂R207Cγ₂ receptors. Modulators had no effects on the rates of MTSEA-biotin modification at position 207. Data were acquired, normalized, and fit to a single-phase exponential decay as described in Fig 2 and Materials and Methods. Second-order rate constants for MTS modification of β₂ subunit cysteine substitutions are summarized in TABLE 1.

2.5. Statistical analysis

All data were from at least three different oocytes from at least two different frogs. Data from MTS modification experiments were fit using nonlinear regression analysis included in the Prism software package (GraphPad Software Inc., San Diego, CA). Significant differences in k₂ (second order rate) values were determined by one-way ANOVA, followed by a post hoc Dunnett’s test (GraphPad Software Inc., San Diego, CA). We also used FDR (false discovery rate) analysis to examine statistical differences in MTS rates of reaction between control and experimental conditions at β₂P206C (Herrington, 2002).

3. Results

The aim of this study was to test the following distinct, yet related, hypotheses: (1) Allosteric modulators induce conformational movements in the GABA binding site interface, (2) Positive and negative BZD modulators induce distinct movements within the GABA binding site interface, which encode a BZD’s efficacy, and (3) Positive allosteric modulators that bind to different sites on the GABAR initiate similar movements at the GABA binding site interface. To test these hypotheses, we introduced cysteine residues throughout the GABA binding site interface and measured their rates of modification by sulfhydryl specific reagents in the absence or presence of flurazepam (BZD-site positive modulator), DMCM (BZD-site negative modulator), flumazenil (BZD-site zero modulator) or pentobarbital (PB, barbiturate-site positive modulator). The rate of covalent modification depends on the local physical and electrostatic environment near the introduced cysteine. As such, any alteration in the modification rate of a cysteine in the presence of ligand is indicative of a change in the local environment of the cysteine and thus a ligand-induced conformational movement in the protein.

3.1. Expression and functional characterization of cysteine mutant receptors

The GABA binding site is comprised of six non-contiguous regions, arbitrarily designated loops A (β-strand 4), B (β-strand 7), and C (β-strand 10) on the β₂ subunit and loops D (β-strand 2), E (β-strand 5–6) and F (loop 9) on the α₁ subunit (Galzi and Changeux, 1995). We individually mutated residues in Loop D (T60, S68) and loop E (E122, R131) of the α₁ subunit and, loop A (L99), loop B (T160) and loop C (S209, R207, P206 and G203) of the β₂ subunit to cysteine (Fig. 1). Mutant α₁ and β₂ subunits were co-expressed with WT α₁ or β₂ subunits and the γ₂ subunit in Xenopus laevis oocytes to form α₁β₂γ₂ GABARs, which were then analyzed using two-electrode voltage clamp. All of the mutant subunits formed functional channels (I_GABAmax = 1–20μA). Previously, we determined the GABA EC₅₀ values for these mutant receptors (Boileau et al., 1999; Boileau et al., 2002; Holden and Czajkowski, 2002; Kloda and Czajkowski, 2007; Newell et al., 2004; Wagner and Czajkowski, 2001). In this study, we estimated the GABA EC₅₀ of each oocyte we recorded from by measuring the ratio of the amplitude of current induced by a low versus maximal concentration of GABA (see Materials and Methods). Our GABA EC₅₀ estimates replicated our data previously reported, where β₂ mutants T160C, G203C, R207C, S209C and α₁ mutants S68C, E122C, R131C significantly increased GABA EC₅₀ values 12-, 2700-, 29-, 5-, 6-, 3-, and 23-fold relative to WT (WT EC₅₀= 10.7 +/− 1.3 μM, n = 6), respectively. Cysteine substitution of α₁T60, β₂L99 and β₂P206 had no significant effect on GABA EC₅₀ compared to WT receptors.

The mutant receptors were all responsive to BZD and pentobarbital (PB) modulation of I_GABA. Flurazepam maximally potentiated GABA (EC_10–15) currents in mutant receptors 2.2 to 3.3-fold, comparable to WT receptors (WT potentiation = 2.6; (Boileau and Czajkowski, 1999; Hanson and Czajkowski, 2008)). DMCM maximally attenuated GABA (EC_10–15) current in mutant receptors 0.27 to 0.65-fold, similar to WT GABARs (0.62-fold; (Boileau and Czajkowski, 1999; Hanson and Czajkowski, 2008)). The BZD antagonist flumazenil had no effect on GABA-gated current in mutant receptors tested (α₁E122Cβ₂γ₂ and α₁R131Cβ₂γ₂) and attenuated the potentiating and inhibiting actions of flurazepam and DMCM respectively, as expected for a BZD-site antagonist. Low, modulating concentrations (25–50μM) of PB, potentiated GABA (EC_10–15) current from mutant receptors between 3 and 6-fold, comparable to its effects on WT receptors.

3.2 Effects of flurazepam, DMCM, Flumazenil, and PB on rates of modification of introduced cysteines in the GABAR α₁ subunit

Previously, we demonstrated that flurazepam altered the rate of modification of α₁E122C and α₁R131C (Kloda and Czajkowski, 2007). α₁E122 and α₁R131 are located on β-strands 5 and 6 (loop E), respectively, at the GABA binding interface (Fig 1). Here, we examined and compared the effects of DMCM, flurazepam, flumazenil and PB on the rate of modification of these residues. Flurazepam, DMCM and PB each significantly slowed MTSEA-biotin modification of α₁E122C (Fig. 2; Fig. 5; Table 1), suggesting that this residue and/or residues in close proximity undergo conformational rearrangement in the presence of these allosteric modulators. Despite the fact that flurazepam is a positive modulator and DMCM is a negative modulator, both induced a similar 1.7-fold reduction (p<0.05) in the rate of reaction of MTSEA-biotin at α₁E122C (Table 1; Fig. 5A). At α₁R131Cβ₂γ₂ receptors, MTSEA-biotin reacted 2.1- and 1.9-fold faster in the presence of BZD-site ligands flurazepam and DMCM, respectively (p<0.01; Fig. 2; Fig. 5A; Table 1). Interestingly, unlike flurazepam and DMCM, flumazenil (BZD-site antagonist) did not significantly alter the rate of modification at α₁R131C (k₂ = 307±4.3 M⁻¹s⁻¹; n=3) or α₁E122C (k₂ = 31,430±3617 M⁻¹s⁻¹; n=4) indicating that simple occupation of the BZD site is not sufficient to induce structural rearrangements at the GABA binding interface. Low, non-activating concentrations of PB significantly slowed MTSEA-biotin modification of α₁E122C 1.5-fold (Fig. 2; Fig. 5A; Table 1) but had no effect on the rate of modification of α₁R131C indicating that conformational movements induced by PB and BZDs are distinct.

Figure 5. Summary of the effects of allosteric modulators on second-order rate constants (k₂) of MTS reaction at α₁ and β₂ subunit cysteine substitutions.

Data were normalized to control second-order rate constants (the rate measured in the absence of allosteric modulators) and represent the mean±SEM of at least 3 experiments. Statistical significance was determined using a one-way ANOVA with Dunnett’s post-test. (A) Rates of modification of cysteines on the α₁ subunit. Flurazepam and DMCM significantly slowed the rate of reaction at position α₁E122C (*p<0.05), while increasing the rate of reaction at position α₁R131C (**p<0.01). Pentobarbital, but not flurazepam or DMCM, significantly increased the rate of reaction at α₁S68C (*p<0.05). (B) Rates of modification of cysteines on the β₂ subunit. With the exception of β₂P206C, allosteric modulators did not alter the rate of reaction at any of the positions tested on the β₂ subunit. Modulatory pentobarbital significantly increased the rate of reaction at β₂P206C (**p<0.01) and, while not significant by ANOVA, false discovery rate analysis suggests that both flurazepam and DMCM speed the rate of reaction.

Two additional α₁ subunit cysteine mutations (T60C, S68C), located on either end of β-strand 2 (loop D), were also examined (Fig 1). Based on comparisons between the agonist bound crystal structure of the AChBP (acetylcholine binding protein) and the 4Å resolved cryo-EM structure of the nicotinic acetylcholine receptor (Unwin, 2005), as well as crystal structures of bacterial LGIC (ligand-gated ion channel) homologues, ELIC and GLIC, in presumed closed and open channel conformations (see (Corringer et al., 2010) for review), it has been postulated that the β-strands in the extracellular domain rotate upon agonist-mediated channel activation. If BZD binding initiates similar movements in the GABAR as agonist-mediated activation of structurally related pentameric LGICs, we might expect residues in β-strand 2 (loop D), as well as those in β-strands 5 and 6 (loop E), to undergo structural rearrangements upon binding of BZDs. Unlike residues on β-strands 5 and 6 (α₁R131C and α₁E122C), the rates of MTS modification of α₁S68C and α₁T60C were not significantly altered in the presence of flurazepam or DMCM (p>0.05; Fig. 3; Fig 5A; Table 1) suggesting that BZD binding does not induce a concerted movement of the inner β-strands of the α₁ subunit. We cannot, however, rule out the possibility that BZD-induced movements occurred but were not detected by our experimental approach. Low modulatory concentrations of PB significantly accelerated the rate of MTSES modification of α₁S68C (p<0.05; Fig. 3; Fig. 5A; Table 1), consistent with our previous data that showed that this residue undergoes a change in environment in the presence of PB when expressed in the background of α₁β₂ receptors (Holden and Czajkowski, 2002). Since the rate of modification of α₁S68C was altered in the presence of PB but not flurazepam or DMCM, the data provide additional evidence that the binding of PB and BZDs induce distinct movements at the GABA binding site interface.

3.3. Effects of flurazepam, DMCM, and PB on the rate of modification of cysteine substitutions in the β₂ subunit

We were particularly interested in examining the rate of modification of cysteine mutations in loop C (β-strand 10) of the β₂ subunit, as loop C is postulated to be a key region involved in agonist-mediated receptor activation in members of the cys-loop LGIC super family (see (Sine and Engel, 2006) for review). Upon agonist binding, Loop C is proposed to cap the site and trap the agonist, which then triggers a series of down-stream conformational events that ultimately leading to channel opening (Celie et al., 2004; Gao et al., 2005; Law et al., 2005; Lee and Sine, 2005; Unwin, 2005; Wang et al., 2009). If binding of positive allosteric modulators triggers some of the same conformational movements at the orthosteric site that are induced by agonist binding, we might expect to detect structural rearrangements in loop C upon binding of flurazepam and potentiating concentrations of PB. Due to its speculated role in coupling agonist binding to channel gating (Boileau et al, 2002), we also examined residue β₂L99C in Loop A (β-strand 4). Moreover, we examined β₂T160C in loop B (β-strand 7), since previous work had shown that this region undergoes structural rearrangement in the presence of GABA and modulatory concentrations of PB (Newell et al, 2004).

Flurazepam, DMCM and PB had no significant effects on the rates of MTSEA-biotin modification of β₂L99C (Loop A), β₂T160C (Loop B), β₂G203C, β₂R207C, or β₂S209C (loop C) (Fig. 4B; Fig. 5B; Table 1). MTSEA-biotin reacted significantly faster at β₂P206C (1.9-fold) in the presence of PB (p<0.01; Fig. 4A; Fig. 5B; Table 1). In the presence of flurazepam and DMCM, the rates of MTSEA-biotin reaction at β₂P206C were increased 1.6-fold (Table 1, Fig. 5B), but the data did not reach statistical significance using an ANOVA. False discovery rate analysis of the data (FDR; Herrington, 2002), however, suggested that flurazepam and DMCM significantly increased MTSEA-biotin reaction rates.

4. Discussion

4.1. BZD agonist and inverse agonist actions induce structural rearrangement at the GABA binding site interface

Both a BZD-site agonist and a BZD-site inverse agonist changed the rate of modification of two residues in Loop E of the GABAR α₁ subunit, α₁E122C (on β-strand 5) and α₁R131C (on β-strand 6) (Fig. 2; Fig 5A; Table 1). The data demonstrate that BZDs induce a conformational movement near these residues and that structural changes initiated at the α/ γ subunit BZD binding site interface can extend over considerable distance. In the α₁ subunit, β-strands 5 and 6 (Loop E of the GABA binding site) are connected to the BZD-binding pocket (Loop A) by a six-residue linker element (Fig. 1). We envision that BZD-induced conformational movements at the α/γ BZD binding site interface are propagated across the α₁ subunit via this linker to the β/α GABA binding site interface. Consistent with this idea, it has been proposed that Zn²⁺ binding to this linker region in the structurally related glycine receptor mediates Zn²⁺-mediated allosteric inhibition of glycine receptor function by hindering agonist-induced movement at the adjacent glycine binding site interfaces (Miller et al., 2008).

Flurazepam (BZD site agonist) and DMCM (BZD site inverse agonist) both induced nearly identical changes in the rates of modification of α₁E122C and α₁R131C (Fig. 5; Table 1), suggesting that the movements being detected near these residues are similar and are associated with BZD binding site occupation and do not reflect whether a BZD is a positive or negative modulator. Alternatively, it is possible flurazepam and DMCM trigger structurally distinct movements that are not detectable by our experimental approach. For example, if flurazepam and DMCM initiate different movements such that the resulting physico/chemical environments around the introduced cysteines were coincidentally nearly identical, then the change in the rate of modification in the presence of either BZD-ligand would be similar. To help distinguish between these possibilities, we measured the rate of modification of only those cysteine substitutions conformationally sensitive to modulatory BZDs, α₁E122C and α₁R131C, in the presence of the BZD-site antagonist flumazenil. We reasoned that if the movements induced by flurazepam and DMCM carry no information regarding efficacy, a BZD-site antagonist (zero modulator) would alter the rate of modification of α₁E122C and α₁R131C to the same extent as flurazepam and DMCM. However, flumazenil did not alter the rate of modification of α₁R131C or α₁E122C. This finding indicates the movements induced by flurazepam and DMCM do not simply reflect BZD binding site occupation, but instead correlate with the ability of these BZD-site ligands to modulate GABAR function (efficacy). Further studies will be necessary, using different experimental approaches (e.g. voltage clamp fluorimetry) and multiple BZD-site ligands with different structures and efficacies, to distinguish the specific movements induced by these modulators. Nonetheless, while our methodological approach cannot distinguish whether a BZD-site ligand is a positive versus negative modulator, the data demonstrate that BZD efficacy is, at least in part, encoded at the GABA binding site interface. Although the temporal resolution of these conformational movements initiated by BZDs has yet to be determined, we hypothesize that BZD-induced structural perturbations at the GABA binding interface trigger structural movements at or near the transmembrane domains, which ultimately alter GABAR gating machinery. In support of our hypothesis, previous studies have shown that BZDs induce structural rearrangements in the α₁ M3 transmembrane helix of the receptor (Williams and Akabas, 2000) and in γ₂ Loop F(9) near the transmembrane region (Hanson and Czajkowski, 2008).

Surprisingly, flurazepam and DMCM did not significantly alter the rates of modification of the majority of introduced cysteines in the β₂ subunit, suggesting that BZD-induced conformational movements at the α/γ BZD binding site interface do not extend to these regions on the β₂ subunit. However, we cannot rule out the possibility that BZD-induced movements may have occurred such that the physico-chemical environment surrounding the cysteines was unchanged, thus rendering these movements undetectable using the substituted cysteine accessibility method. While not statistically significant using an ANOVA analysis, flurazepam and DMCM each increased the rate of modification of β₂P206C 1.6-fold (Fig. 4A; Fig. 5; Table 1), which when analyzed using a less stringent FDR analysis were significant, hinting that BZDs may induce movements at or near this residue. We previously demonstrated that GABA significantly increased the rate of modification at β₂P206C by 2.6-fold (Wagner and Czajkowski, 2001) indicating that GABA binding and/or GABA-mediated channel activation triggers structural rearrangements at or near this residue. We speculate that BZD-induced movements may not be as large or as wide spread as those induced by GABA since BZD binding in the absence of GABA does not under normal conditions gate the GABAR channel. Consistent with this idea, GABA induced a 3-fold decrease in the rate of modification of α₁E122C (Kloda and Czajkowski, 2007), while flurazepam and DMCM induced a more modest 1.7-fold decrease in rate (Table 1). Moreover, diazepam similarly induced a more limited conformational change in the α₁ M3 transmembrane helix relative to GABA (Williams and Akabas, 2000).

4.2. Flurazepam and PB induce different structural rearrangements at the GABA binding site interface

BZD-site ligands and low potentiating concentrations of PB had different effects on the rates of modification of α₁S68C and α₁R131C. At α₁S68C, flurazepam and DMCM had no effect on its modification rate, while PB significantly increased its rate of modification (Fig. 3B; Fig 5; Table 1). At α₁R131C, BZD-site ligands increased the modification rate, while PB had no effect (Fig 2C&D; Fig 5; Table 1). Furthermore, at Loop C residue β₂P206C, a larger increase in the rate of reaction in the presence of PB relative to flurazepam was observed. These results demonstrate that PB and flurazepam induce different structural rearrangements at or near these residues, indicating that the structural mechanisms underlying positive allosteric modulation by the general anesthetic PB is different than the mechanisms underlying BZD positive modulation. This divergence in structural mechanisms is consistent with the differential effects these drugs have on GABAR activation and single channel kinetics. The fact that PB can act as an agonist, while flurazepam cannot, may explain the differential effects of PB and flurazepam in loop C (β-strand 10), as closure or movement of loop C is associated with channel activation (see (Sine and Engel, 2006) for review). Additionally, general anesthetics such as PB and propofol increase channel open time durations whereas BZD-site positive modulators increase the frequency of channel opening (MacDonald et al., 1989; Mellor and Randall, 1997; Rogers et al., 1994; Steinbach and Akk, 2001; Twyman et al., 1989). Given that PB and BZDs bind to distinct sites on the GABAR (Amin, 1999; Belelli et al, 1999; Serafini et al, 2000), one might predict that the receptor would undergo different structural rearrangements upon binding of each drug. Our findings provide experimental evidence for this idea. Additional supporting evidence comes from studies that demonstrate propofol and BZDs initiate different structural rearrangements in the GABAR α₁ M3 transmembrane helix (Williams and Akabas, 2000, 2002). Taken together, the data suggest that the allosteric pathways governing positive modulation of the GABAR by different classes of modulators are distinct. Mapping these structural pathways is an important goal for future studies.

Abbreviations

BZD

benzodiazepine

GABAR

GABA-A receptor

MTSEA-Biotin

N-biotinaminoethyl methanethiosulfonate

MTSES

2-sulfonatoethyl methanethiosulfonate

DMCM

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

Ro15–1788

flumazenil

PB

pentobarbital

LGIC

ligand-gated ion channel

AChBP

acetylcholine binding protein

GLIC

bacterial Gloeobacter violaceus pentameric LGIC homologue

ELIC

bacterial Erwinia chrysanthemi LGIC homologue