Modulation of neuronal and recombinant GABA_A receptors by redox reagents

Abstract

1. The functional role played by the postulated disulphide bridge in γ-aminobutyric acid type A (GABA_A) receptors and its susceptibility to oxidation and reduction were studied using recombinant (murine receptor subunits expressed in human embryonic kidney cells) and rat neuronal GABA_A receptors in conjunction with whole-cell and single channel patch-clamp techniques.

2. The reducing agent dithiothreitol (DTT) reversibly potentiated GABA-activated responses (I_GABA) of α1β1 or α1β2 receptors while the oxidizing reagent 5,5′-dithio-bis-(2-nitrobenzoic acid) (DTNB) caused inhibition. Redox modulation of I_GABA was independent of GABA concentration, membrane potential and the receptor agonist and did not affect the GABA EC₅₀ or Hill coefficient. The endogenous antioxidant reduced glutathione (GSH) also potentiated I_GABA in α1β2 receptors, while both the oxidized form of DTT and glutathione (GSSG) caused small inhibitory effects.

3. Recombinant receptors composed of α1β1γ2S or α1β2γ2S were considerably less sensitive to DTT and DTNB.

4. For neuronal GABA_A receptors, I_GABA was enhanced by flurazepam and relatively unaffected by redox reagents. However, in cultured sympathetic neurones, nicotinic acetylcholine-activated responses were inhibited by DTT whilst in cerebellar granule neurones, NMDA-activated currents were potentiated by DTT and inhibited by DTNB.

5. Single GABA-activated ion channel currents exhibited a conductance of 16 pS for α1β1 constructs. DTT did not affect the conductance or individual open time constants determined from dwell time histograms, but increased the mean open time by affecting the channel open probability without increasing the number of cell surface receptors.

6. A kinetic model of the effects of DTT and DTNB suggested that the receptor existed in equilibrium between oxidized and reduced forms. DTT increased the rate of entry into reduced receptor forms and also into desensitized states. DTNB reversed these kinetic effects.

7. Our results indicate that GABA_A receptors formed by α and β subunits are susceptible to regulation by redox agents. Inclusion of the γ2 subunit in the receptor, or recording from some neuronal GABA_A receptors, resulted in reduced sensitivity to DTT and DTNB. Given the suggested existence of αβ subunit complexes in some areas of the central nervous system together with the generation and release of endogenous redox compounds, native GABA_A receptors may be subject to regulation by redox mechanisms.

The role of γ-aminobutyric acid (GABA) in causing activation of GABA_A receptors leading to inhibition of neuronal excitability is widely recognized. GABA_A receptors are structurally diverse since they are formed from a variety of subunits selected from the following five distinct families of which three contain multiple isoforms; α(1-6), β(1-3), γ(1-3), δ(1) and ε(1) (Sieghart, 1995; Rabow et al. 1995; Davies et al. 1997). Following selected co-assembly of these polypeptides, each receptor complex is postulated to form a pentameric hetero-oligomer (Nayeem et al. 1994). Although this diversity in subunit composition offers considerable potential for variation in physiological and pharmacological properties, there are several structural features that are common to many of the individual subunits, including four transmembrane domains, a large intracellular loop, external N-termini, and two cysteine residues present near the N-terminus that are conserved in all GABA_A receptor subunits and are presumed to participate in disulphide bridge formation (Barnard et al. 1987). Such a structure is expected to be susceptible to disulphide reducing agents, such as dithiothreitol (DTT), which cleave the bridge forming independent cysteine moieties with intact sulfhydryl groups.

The importance and even the existence of the disulphide bridge for GABA_A receptors has not yet been resolved. Previous functional studies with redox reagents report apparently different results. In spinal cord and retinal neurones, DTT treatment modestly enhanced GABA-activated membrane currents (Porter et al. 1991; Pan et al. 1995) whilst oxidation of the sulfhydryl groups with 5,5′-dithio-bis-(2-nitrobenzoic acid) (DTNB) resulted in inhibition (Pan et al. 1995). The involvement of the cysteine residues was deduced because the alkylating agent N-ethylmaleimide occluded the action of the reducing agent DTT (Pan et al. 1995). In contrast, recurrent inhibition mediated by GABAergic synapses in guinea-pig hippocampus was unaffected by DTT (Tolliver & Pellmar, 1987) and interestingly, in invertebrates, GABAergic synaptic function was also unaffected by redox reagents (Ben-Haim et al. 1973; Sato et al. 1976). Recently, using site-directed mutagenesis to replace one of two cysteines, believed to participate in disulphide bridge formation, with serine residues in either α1, β2 or γ2 subunits resulted in an apparent failure of subunit expression on the cell surface (Amin et al. 1994). This suggested a possible role for the cysteine loop structure in subunit assembly and/or transport to the surface membrane.

The regulation of GABA_A receptor function by redox reagents has the potential to be physiologically important considering the presence of endogenous redox compounds, e.g. glutathione and ascorbic acid, in the central nervous system (CNS) (Slivka et al. 1987; Yudkoff et al. 1990). Moreover, the reducing agent glutathione can be released in a Ca²⁺-dependent fashion (Zangrle et al. 1992) upon stimulation and the levels achieved could be sufficient to modulate GABA_A receptor function and also inhibitory synaptic transmission.

The present study addresses the issue of modulation of native neuronal and recombinant GABA_A receptor function with redox agents. We have elucidated that redox modulation of GABA-activated responses depends on the subunit composition of the receptor. In particular, the presence of the γ2 subunit appears to diminish the sensitivity of GABA_A receptors to exogenous and naturally occurring redox reagents. Thus, redox modulation at GABAergic synapses is predicted to vary throughout the CNS depending on the subtypes of GABA_A receptors that are expressed.

A preliminary account of some of this work has been published previously (Amato et al. 1996).

METHODS

Expression vectors

Murine cDNAs, encoding GABA_A receptor subunits, were cloned as Eco RI fragments and inserted into the mammalian expression vector pGW1 as described previously (Krishek et al. 1993).

Cell preparation

Human embryonic kidney (HEK) cells

Transient expression of recombinant GABA_A receptors was achieved by electroporation of HEK cells (American Type Culture Collection CRL 1573). Cells were grown in a 10 cm culture dish at 37°C in a humidified 95 % air-5 % CO₂ atmosphere and bathed in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10 % (v/v) fetal calf serum (Gibco), 2 mM glutamine, 100 u ml⁻¹ penicillin G and 100 μg ml⁻¹ streptomycin, until they reached 70 % confluence. Cells were then harvested by trypsinization, and washed once in Optimem medium (Gibco) before resuspension in 0.5 ml Optimem. The cell suspension was transferred into an electroporation cuvette containing a total of 10 μg of cDNA for the GABA_A receptor subunits present in equal ratios. Cells were electroporated twice using a Gene Electropulser II (Bio-Rad) with the following parameters: 0.4 kV, 125 μF capacitance and infinite resistance. Electroporated cells were briefly triturated with polished Pasteur pipettes and plated onto 35 mm culture dishes or glass coverslips pretreated with poly-L-lysine and used for electrophysiological recording 18-72 h after transfection. Transfection efficiency was approximately 90 % as assessed by electrophysiological recording of GABA_A currents from cells preselected for displaying a granular morphology.

Superior cervical ganglionic neurones

Cultured rat sympathetic neurones were obtained as described previously (Smart, 1992). Briefly, ganglia were removed from day 1 postnatal Sprague- Dawley rats (P1), killed by exsanguination, and subjected to enzymatic and mechanical dissociation. Dissociated neurones were plated onto a laminin substratum and cultured in Leibovitz medium (Gibco) supplemented with 10 % (v/v) fetal calf serum, 2 mM glutamine, 0.6 % (w/v) glucose, 0.19 % (w/v) NaHCO₃, 100 u ml⁻¹ penicillin G, 100 μg ml⁻¹ streptomycin and 50 ng ml⁻¹ 7S-nerve growth factor (Calbiochem). Neurones were incubated at 37°C in 95 % air-5 % CO₂. Neurones were used after 4-7 days in vitro and possessed membrane potentials of -40 to -65 mV and action potential amplitudes of 62 to 105 mV.

Cerebellar granule neurones

Sprague-Dawley postnatal P3-P6 rats were killed by decapitation and the cerebella removed into ice-cold Hanks' Ca²⁺-Mg²⁺-free balanced salt solution (HBSS). Cerebella were transversely sectioned into four pieces and incubated in 0.25 % (w/v) trypsin and 0.5 mg ml⁻¹ DNase I in HBSS for 9 min. Tissue chunks were washed by sedimentation through 5 ml of HBSS (× 3) and then resuspended in MEM 10/10 medium containing 0.5 mg ml⁻¹ DNase I. The chunks were then triturated using a gradation of flame-polished Pasteur pipettes and the supernatants pooled. The cell suspension was centrifuged at 140 g (Denley Bench centrifuge) and the cells were resuspended by gentle pipetting in MEM 10/10 and plated onto poly-L-lysine-coated glass coverslips maintained at 37°C in a 95 % air-5 % CO₂ atmosphere. After 2 days in vitro, the media were replaced with MEM 10 supplemented with 20 mM KCl. After 18 h, the cultures were treated with 10 μM cytosine arabinoside for 24 h. All media were replenished 2, 5 and 8 days after plating. MEM 10/10 medium contained: MEM Earle's salts (Gibco), 10 % (v/v) horse serum, 10 % fetal calf serum, 600 mg l⁻¹ glucose, 2 mM glutamine, 100 u ml⁻¹ penicillin G and 100 μg ml⁻¹ streptomycin. MEM 10 medium had a similar composition but lacked the fetal calf serum.

Patch-clamp electrophysiology

Whole-cell recording

Membrane currents were recorded from HEK cells, sympathetic neurones and cerebellar granule cells using the whole-cell patch-clamp configuration in conjunction with an Axopatch-1C amplifier (Axon Instruments). Patch pipettes (resistance, 1-5 MΩ) were pulled from thin-walled borosilicate glass (Clark Electromedical) and filled with a solution containing (mM): 140 KCl, 2 MgCl₂, 1 CaCl₂, 10 Hepes, 11 EGTA and 2 adenosine triphosphate; pH 7.1. The cells were continuously perfused with Krebs solution containing (mM): 140 NaCl, 4.7 KCl, 1.2 MgCl₂, 2.5 CaCl₂, 10 Hepes and 11 glucose; pH 7.4. Membrane currents were filtered at 5-10 kHz and displayed on a Gould 2200S ink-pen chart recorder and stored on a Racal Store 4 FM tape recorder (DC to 5 kHz) for analysis.

Single channel recording

Single channel currents were recorded from outside-out patches bathed in Krebs solution using thick-walled patch pipettes and an Axopatch 200B amplifier. The patch pipettes were coated with insulating varnish (Radio Spares) and possessed resistances of 5-10 MΩ. All single channel currents were stored on a Racal tape recorder, filtered at 1-2 kHz (-3 dB; 6-pole Bessel) and digitized at 20 kHz prior to analysis using PAT version 7.2 (John Dempster, University of Strathclyde, UK). Open and closed channel durations were allotted on the basis of a 50 % threshold detection cursor to the main conductance state of 16 pS. The transition detection of open and closed states was used to form an idealized record from which individual open durations were measured and collated into frequency distributions. The distributions were fitted with the sum of exponential functions using a non-linear least-squares Levenberg-Marquardt fitting routine as described previously (Smart, 1992). The number of exponential functions required for fitting was increased until the overall fit was not significantly improved according to the χ² test. Single channel currents were accepted for analysis if the number of multichannel openings never exceeded 2 % of the total number of open channel events.

Drugs and solutions

Drugs and solutions were rapidly applied to the cells via a U-tube manufactured from glass electrode tubing and positioned approximately 300 μm from the recorded cell. Solutions travelled around the U-tube due to the creation of a vacuum. Drug application was achieved by isolating the vacuum using a solenoid valve enabling the solution to pass out of the tube over the recorded cell by gravity. A complete exchange of solution was achieved in 20-60 ms. When recording NMDA-activated currents from cerebellar granule neurones, the extracellular solution contained nominally zero Mg²⁺ and 10 μM glycine. All drugs were prepared in the external Krebs solution and the pH was adjusted to 7.4 with 1 n NaOH or HCl. The dissolution of DTNB was aided by sonication.

Analysis of whole-cell current data

GABA-activated membrane currents were analysed by measuring the amplitudes of the peak response and the response obtained 10 s after the onset of the GABA-activated current. Responses to GABA were normalized to the amplitude of the current induced by 10 μM GABA and used to construct equilibrium concentration-response curves. Curves were fitted to the logistic equation:

where I and I_max are the currents induced by a GABA concentration, A, and by a saturating concentration of GABA, respectively. The best fit was determined by using a Marquardt non-linear least-squares routine. The concentration producing a half-maximal response is given by the EC₅₀ and n_His the Hill coefficient. Data are reported as means ± s.e.m. Significance was assessed using Student's unpaired t test; P < 0.05 was considered significant.

Kinetic modelling

Macroscopic membrane currents were modelled using Modelmaker (version 3; Cherwell Scientific, Oxford, UK) simulation software on a Viglen Pentium PC.

Quantification of GABA_A receptor cell surface expression

Untransfected HEK cells and HEK cells expressing α1β1, α1β2, α1β1γ2S and α1β2γ2S GABA_A receptors were exposed to 2 mM DTT for 1 h. The cells were then cooled to 4°C and incubated with saturating levels of ¹²⁵I-labelled 9E10 antibody (Connolly et al. 1996). Following extensive washing, the degree of labelling of surface cell membrane GABA_A receptors was assessed by gamma counting.

RESULTS

Modulation of I_GABA by redox reagents: αβ subunit GABA_A receptors

The effects of redox compounds were studied in single HEK cells expressing α1βi GABA_A receptor subunits (where i = 1 or 2). GABA activation of these receptors was recorded as rapidly declining inward membrane currents using whole-cell voltage clamp at a holding potential of -40 mV with E_Cl (chloride equilibrium potential) at approximately 0 mV (Fig. 1A). After pre-incubation with 2 mM DTT for a minimum of 3 min, which did not alter the resting membrane current or conductance, 10 μM GABA was co-applied in the presence of DTT. The peak (I_peak) GABA-activated current (I_GABA) was enhanced by DTT but the steady-state response at the end of the GABA application (typically 10-20 s) was apparently unaffected (Fig. 1A). DTT enhanced I_peak by 71 ± 18 % over the control GABA-activated response (n = 5 cells) for α1β1, and by 32 ± 10 % (n = 8) for α1β2 constructs. In comparison, I_GABA measured 10 s after the onset of the response to GABA (I₁₀) was not enhanced over controls (9 ± 9 and 4 ± 2 % for α1β1 and α1β2 receptors, respectively; Fig. 1B). The EC₅₀ value for enhancement of I_peak by DTT was 51.8 ± 5.03 μM (n = 3-11). This potentiation by DTT was readily reversed in control Krebs solution.

Figure 1. Modulation by redox reagents of I_GABA in α1β2 subunit GABA_A receptors.

A, whole-cell currents activated by 10 μM GABA and recorded at a holding potential of -40 mV from HEK cells expressing α1β2 receptor constructs. GABA was rapidly applied for the periods indicated by the horizontal bars in the presence and absence of 2 mM DTT or 0.5 mM DTNB. B, bar graph showing the amplitudes of the 10 μM GABA-activated current at the peak of the response (I_peak) and 10 s after the start of the GABA application (I₁₀) in the presence of 2 mM DTT or 0.5 mM DTNB (n = 8 cells). Control 10 μM GABA I_peak and I₁₀ amplitudes were designated 100 %. In this and subsequent bar graphs, each bar represents the mean ± s.e.m.C, membrane currents induced by 4 μM muscimol recorded from α1β2 constructs before, during and after application of 2 mM DTT. D, bar graph showing the enhancement by 2 mM DTT of the peak currents activated by 4 μM muscimol (I_Musc) and 10 μM GABA (I_GABA) (n = 3). Both control I_Musc and I_GABA in the absence of DTT were designated 100 %.

To assess the potential for modulation of GABA_A receptor function by oxidation, transfected HEK cells were exposed to the strong oxidizing reagent DTNB (0.5 mM). This agent did not affect the resting membrane current or conductance but the peak response to 10 μM GABA was reduced with little effect on the steady-state GABA-induced current (Fig. 1A). DTNB inhibited the GABA-activated I_peak by 44 ± 15 % (n = 3) for α1β1 and by 43 ± 4 % (n = 5) for α1β2 receptors, while I₁₀ was inhibited by only 6 ± 6 and 17 ± 6 %, respectively (Fig. 1B). This inhibitory effect was fully reversible following removal of DTNB with control Krebs solution.

To discount whether DTT was modifying the GABA molecule in preference to the GABA_A receptor, the activity of DTT was assessed with the GABA_A receptor agonist muscimol. The activity of DTT appeared to be independent of the receptor agonist since 4 μM muscimol-activated responses were also enhanced by 68 ± 4 % over control by 2 mM DTT (Fig. 1C and D; n = 3). The concentration of muscimol used was selected since this was equipotent with 10 μM GABA for the activation of α1β2 subunit receptors. To also eliminate whether only the GABA recognition site was sensitive to redox modulation, α1β2 receptors were directly activated with 200 μM pentobarbitone which will bind to a discrete site (see Rabow et al. 1995; Sieghart, 1995). The pentobarbitone-activated peak currents (I_PB) were clearly enhanced by 2 mM DTT (I_PB(+DTT)/I_PB(control) = 5.44 ± 0.87; n = 4). This suggested that either both binding sites are redox sensitive or that DTT is affecting the allosteric mechanisms controlling ion channel activation.

GABA concentration-response curve analysis: effect of DTT and DTNB

Equilibrium concentration-response curves were constructed for GABA on HEK cells expressing α1βi subunit constructs and exposed to DTT or DTNB. For either α1β1 (data not shown) or α1β2 (Fig. 2A), GABA-activated responses were increased by 2 mM DTT at each concentration of GABA resulting in an enhanced maximum response and the curve was displaced upwards along the ordinate. The degree of enhancement showed little variation with the GABA concentration. In comparison, DTNB (0.5 mM) produced an inhibition at each GABA concentration in accordance with non-competitive antagonism (Fig. 2A). The EC₅₀ values and Hill coefficients determined from the curve fits for the equilibrium dose-response curve data for α1β1 (not shown) and α1β2 recombinant receptors were not altered by DTT or DTNB (EC₅₀ values: control, 2.51 ± 0.37 μM; + 2 mM DTT, 1.99 ± 0.44 μM; + 0.5 mM DTNB, 2.61 ± 0.47 μM) (P > 0.05; Fig. 2B).

Figure 2. GABA equilibrium concentration-response curves for α1β2 receptors.

A, concentration-peak response curves constructed for GABA in the absence (Control) and presence of 2 mM DTT or 0.5 mM DTNB for HEK cells expressing α1β2 subunits. The data (means ± s.e.m.) were normalized to the peak response to 10 μM GABA in control Krebs solution and fitted according to the logistic equation as described in Methods. B, bar graph of the GABA EC₅₀ values and Hill coefficients (n_H) for control, + 2 mM DTT and + 0.5 mM DTNB (n = 4-8 cells).

Modulation of GABA_A receptors by endogenous redox compounds: chemical specificity

To establish whether modulation by redox compounds could be physiologically important, recombinant α1β2 GABA_A receptors were exposed to the endogenous antioxidant glutathione, in its reduced form (GSH). At concentrations of 1-10 mM, GSH enhanced the peak 10 μM GABA-activated response in a concentration-dependent manner (Fig. 3A). GSH (5 mM) potentiated I_peak by 67 ± 20 % (n = 3), which is comparable to the enhancement induced by 2 mM DTT. Conversely, 5 mM glutathione in its oxidized form, GSSG, which is also present in the CNS though at a much lower concentration than that of GSH (Slivka et al. 1987), had a slight inhibitory effect on I_peak (13 ± 3 %, n = 3; Fig. 3B).

Figure 3. Modulation of GABA-activated currents by glutathione in α1β2 receptors.

A, membrane currents induced by 10 μM GABA on HEK cells expressing α1β2 subunits in control Krebs solution and following pre-incubation for 3 min and subsequent co-application with 1-10 mM glutathione, in its reduced form (GSH), at a holding potential of -40 mV. B, bar graph of peak I_GABA amplitudes activated by 10 μM GABA in control (100 %), following application of 2 mM oxidized DTT (OxDTT), + 2 mM reduced DTT (RedDTT, equivalent to DTT), + 5 mM oxidized glutathione (GSSG) and + 5 mM reduced glutathione (GSH) (n = 4 cells). Individual cells were exposed to up to 3 redox agents separated by full GABA response recoveries from preceding treatments.

To ascertain whether the redox potential of DTT or a non-specific interaction with the receptor protein was important for the modulatory action on the GABA_A receptor, recombinant α1β2 receptors were exposed to an oxidized form of DTT (OxDTT). Unlike the reduced form of DTT, the oxidized congener failed to enhance GABA-activated responses but instead induced a modest inhibition of I_peak of 25 ± 12 % (n = 4; Fig. 3B). The level of inhibition was weaker than that observed with the stronger oxidizing agent, DTNB. This suggested that the redox status of DTT was more important than the basic chemical structure and also implied that DTT was probably interacting with a redox-sensitive site on the GABA_A receptor complex.

Redox modulation of GABA_A receptors: importance of the γ2 subunit

Although binary complexes of α and β subunits may form functional GABA_A receptors in some areas of the CNS (i.e. thalamic nuclei; Kawahara et al. 1993), the majority of native functional GABA_A receptors are thought to exist as ternary subunit complexes incorporating one or more γ subunits. The ability of redox agents to modulate γ2 subunit-containing receptors was assessed with α1β1γ2S or α1β2γ2S constructs. The overwhelming predominance of αβγ receptor subtypes expressed after transfection in HEK cells was monitored by observing the sensitivity of I_GABA evoked by 10 μM GABA to 1 μM flurazepam or 10 μM Zn²⁺, as these two pharmacological agents have proved diagnostic for the presence or absence of the γ2 subunit (Prichett et al. 1989; Draguhn et al. 1990; Smart et al. 1991).

In contrast to the actions of DTT and DTNB on receptors comprising only αβ subunits, receptors containing the γ2 subunit were relatively insensitive to either DTT (2 mM) or DTNB (0.5 mM), irrespective of whether I_peak or I₁₀ was being measured. This apparent insensitivity was consistent at all concentrations of GABA between 0.3 and 300 μM resulting in overlapping concentration-response curves with similar maxima for GABA in the presence and absence of DTT or DTNB (Fig. 4A and B). Consequently, there were no alterations by DTT or DTNB in the GABA EC₅₀ values (α1β2γ2S: control, 14.3 ± 0.83 μM; + DTT, 14.9 ± 1.97 μM; + DTNB, 15.4 ± 0.58 μM; P > 0.05) or Hill coefficients (control, 1.3 ± 0.08; + DTT, 1.1 ± 0.14; + DTNB, 1.47 ± 0.07; P > 0.05). Moreover, exchanging β2 for β1 in the receptor complex did not alter the sensitivity to DTT or DTNB (data not shown).

Figure 4. Redox modulation of γ2 subunit-containing GABA_A receptors.

A, whole-cell membrane currents activated by 10 μM GABA in two HEK cells expressing α1β2γ2S receptor constructs, prior to, during co-application and following recovery from 2 mM DTT (upper panel) or 0.5 mM DTNB (lower panel). B, equilibrium concentration-peak response curves for GABA constructed for α1β2γ2S receptors in control Krebs solution, and in the presence of 2 mM DTT or 0.5 mM DTNB. The data (means ± s.e.m.) were normalized to the peak response to 10 μM GABA and the points were fitted as described in Methods (n = 3-5).

Voltage dependence of redox modulation

The site of action for the redox reagents on the GABA_A receptor may be affected by the membrane electric field, particularly if a redox site is located within the ion channel. Moreover, the apparent inability of the redox reagents to affect the response to GABA of α1βiγ2S subunit constructs may be a feature of particular membrane potentials. These aspects were addressed by analysing the current-voltage (I-V) relationships for α1β2 and α1β2γ2S subunit receptors over the membrane potential range -70 to +30 mV. The peak GABA-activated currents for α1β2 receptors were enhanced by 2 mM DTT in a voltage-independent manner (Fig. 5A). The GABA I-V relationship in the presence of DTT exhibited an approximately 2-fold increase in chord conductance (DTT/control chord conductance ratio of 1.89 ± 0.2; n = 4) but neither a change in the reversal potential (control, 3.3 ± 2.8 mV; + DTT, 4.7 ± 2.4 mV; n = 5) nor an area of overt rectification (Fig. 5A). This suggests that the redox modulatory site of the α1βi receptor is not affected by membrane voltage and is probably located outside the membrane ion channel presumed to be dominated by the second transmembrane domain of the receptor subunits. This notion is in accordance with the N-terminal location of the presumed disulphide bridge in GABA_A receptors.

Figure 5. Voltage sensitivity of redox modulation of recombinant GABA_A receptors.

A, typical membrane currents were recorded in response to 10 μM GABA in a single HEK cell expressing α1β2 receptors at different holding potentials between -70 and +30 mV in control Krebs solution and in the presence of 2 mM DTT (top). These data were used to construct current-voltage (I-V) relationships for the peak response to GABA (bottom). The chord conductance measured between -30 and -70 mV was 0.03 μS in control and 0.05 μS in 2 mM DTT. B, membrane currents evoked by 10 μM GABA recorded from a HEK cell expressing α1β2γ2S receptors at different membrane potentials. The I-V relationships for control and in the presence of DTT yielded chord conductances of approximately 0.015 μS. V_h, holding potential.

For the α1β2γ2S construct, a similar variation of membrane potential failed to unveil a modulatory effect of 2 mM DTT on the 10 μM GABA-activated response (Fig. 5B). Both I-V relationships in the presence and absence of DTT were coincident suggesting that the inclusion of the γ2 subunit in the receptor complex resulted in either the loss of the redox modulatory site or at least a reduction in the sensitivity to redox reagents.

Native neuronal nicotinic acetylcholine, NMDA and GABA_A receptors: modulation by DTT and DTNB

The precise subunit composition and stoichiometry is unknown for native neuronal GABA_A receptors. Apart from selected loci such as the thalamus (Kawahara et al. 1993), it is the consensus opinion that the majority of neuronal GABA_A receptors will contain a γ subunit (Rabow et al. 1995). The results obtained with DTT and DTNB on recombinant GABA_A receptors containing γ2 subunits were compared with those for native neuronal GABA_A receptors using cultured rat sympathetic ganglionic and cerebellar granule neurones. These preparations were selected since ganglia express nicotinic acetylcholine (nACh) receptors and granule neurones express NMDA receptors, which are both sensitive to redox modulation (nACh: Derkach et al. 1991; NMDA: Janaky et al. 1993; cf. Aizenman et al. 1989).

For sympathetic neurones, I_GABA was potentiated by 1 μM flurazepam to 40 ± 13 % over control responses (n = 4) suggesting the presence of a receptor population containing γ subunits. In the presence of 2 mM DTT or 0.5 mM DTNB, I_peak and I₁₀ were not significantly different from their respective controls (Fig. 6A). This result was consistent for concentrations of GABA between 0.3 and 300 μM (Fig. 6B) producing coincident concentration-response curves and similar EC₅₀ values (control, 12.51 ± 0.61 μM; + DTT, 13.09 ± 1.08 μM; + DTNB, 12.55 ± 0.42 μM; P > 0.05, n = 6-11). In addition, the peak GABA-activated responses recorded from sympathetic neurones were also insensitive to 5 mM GSH (100 ± 1 % compared with control; n = 4, data not shown). Increasing the DTT concentration to 10 mM also failed to significantly enhance the response to GABA (121 ± 18 %; n = 4, P > 0.05). Direct activation of the GABA_A receptor by 2 mM pentobarbitone was also unaffected by 2.5 mM DTT (106 ± 2.9 %, n = 4, P > 0.05; Fig. 6C and D). As a positive control for the efficacy of DTT on sympathetic neurones, the response evoked by the nACh receptor agonist 1,1-dimethyl-4-phenylpiperazinium (DMPP) was assessed for sensitivity to redox modulation. Exposure of neurones to 2.5-5 mM DTT markedly inhibited the response to 10 μM DMPP in an apparently irreversible manner (Fig. 6C and D) (+ 10 mM DTT reduced the response to 5.8 ± 5.1 % of control, n = 4).

Figure 6. Sensitivity of GABA_A and nicotinic acetylcholine receptors in sympathetic neurones to redox agents.

A, whole-cell currents activated by 10 μM GABA in a cultured sympathetic neurone recorded at a holding potential of -50 mV in the presence and absence of 0.5 mM DTNB or 2 mM DTT. B, equilibrium concentration-peak response curves for GABA in control, 2 mM DTT and 0.5 mM DTNB from 6-11 cells. Data (means ± s.e.m.) were normalized to the control 10 μM GABA-activated peak response. C, membrane currents recorded from a single sympathetic neurone at a holding potential of -50 mV in response to 10 μM GABA, 2 mM pentobarbitone (PB) and 10 μM 1,1-dimethyl-4-phenylpiperazinium (DMPP) in the presence and absence of 2.5 or 5 mM DTT. D, bar graph showing the mean current amplitudes normalized with respect to the first response to 10 μM GABA, 2 mM PB or 10 μM DMPP in each cell, in the presence and absence of 2.5 or 5 mM DTT (n = 4 cells).

We also examined the effect of redox reagents on I_GABA recorded from rat cerebellar granule neurones. This cell type was selected because of the expression of GABA_A receptors with a pharmacological profile suggestive of the presence of γ2 subunits (Wisden et al. 1996) in addition to NMDA receptors that can be modulated by redox compounds (Janaky et al. 1993). Consistent with the previous results, 2 mM DTT had little effect on I_GABA (4 ± 3 % enhancement over control, n = 6; Fig. 7A and B); however, as expected, DTT markedly potentiated I_NMDA (by 77 ± 1 %, n = 4; Fig. 7A and B). Oxidation by 0.5 mM DTNB produced only a small inhibition of I_GABA in cerebellar granule neurones (16 ± 4 %, n = 6) whilst the NMDA-activated current was substantially reduced (to 35 ± 8 % of control; n = 4; Fig. 7A and B). The potentiation and inhibition of the NMDA-activated current were both readily reversible.

Figure 7. Redox modulation of native GABA_A and NMDA receptors on cerebellar granule neurones.

A, whole-cell membrane currents recorded from 2 cerebellar granule neurones activated by 10 μM GABA (left) and 20 μM NMDA (right) before, during and after application of 2 mM DTT (top) or 0.5 mM DTNB (bottom). Both I_GABA and I_NMDA were recorded at a holding potential of -40 mV. B, bar graph showing the effect of 2 mM DTT and 0.5 mM DTNB on I_GABA and I_NMDA in 4 cells. Currents are normalized to control values (100 %) in the absence of DTT and DTNB.

Redox modulation of α1β1 GABA_A receptors: single channel analysis

The underlying mechanism by which DTT was potentiating GABA-activated responses of recombinant α1β1 GABA_A receptors was examined using single GABA-activated channel analysis from outside-out patches at holding potentials of -70 to -90 mV. Patches were exposed to 0.1-2 μM GABA in the presence or absence of 1 mM DTT (Fig. 8 and Table 1). The single channel current and chord conductances at -90 mV patch potential were unaffected by DTT. In contrast both the mean open time and the probability of channel opening (NP_o) were significantly increased (P < 0.05; Table 1). Analysis of the distribution of all the open durations for patches devoid of many simultaneous channel openings (< 2 %) required two exponential functions to fit the dwell time distribution (Fig. 8B and Table 1). Both the time constants that represent the short and long open time components to the distribution (τ_o1 and τ_o2, respectively) were unaffected by DTT exposure (Fig. 8B and Table 1). The major cause of the increased mean open time resulted from a shift in the relative areas of the open time distribution with more long openings appearing at the expense of shorter openings.

Figure 8. Single channel analysis of DTT modulation of α1β1 receptors.

A, single channel currents recorded in outside-out patches at -90 mV excised from a HEK cell expressing α1β1 GABA_A receptors. The patch was exposed to 0.1 μM GABA in the absence (Control) and presence of 1 mM DTT. B, open time histograms for GABA-activated ion channel currents at a holding potential of -90 mV. The time constants and associated areas were: control: τ_o1, 0.21 ± 0.02 ms (58 ± 10 %) and τ_o2, 1.15 ± 0.4 ms (42 ± 5 %); 1 mM DTT: τ_o1, 0.27 ± 0.06 ms (31 ± 5 %) and τ_o2, 1.35 ± 0.21 ms (69 ± 10 %).

Table 1.

Single channel properties of α1β1 GABA_A receptor subunits

Parameter	Control	+ 1 mm DTT
Single channel current (pA)	−1.465 ± 0.022 (5)	−1.416 ± 0.036 (4)
Conductance (pS)	16.27 ± 0.24	15.73 ± 0.4
NP0	0.524 ± 0.126 (5)	1.133 ± 0.206 (4)*
Mean open time (ms)	0.781 ± 0.055 (3)	1.045 ± 0.066 (3)*
To1 (ms)	0.17 ± 0.03 (5)	0.26 ± 0.06 (4)
Area 1 (%)	65.41 ± 14.78	37.55 ± 2.39
To2 (ms)	1.05 ± 0.26 (5)	1.43 ± 0.28 (4)
Area 2 (%)	30.92 ± 11.96	65.56 ± 2.89*

Data were obtained from outside-out patches (numbers in parentheses) excised from HEK cells expressing α1β1 subunits. Values are means ± s.e.m.

^*P < 0.05.

DISCUSSION

Redox modulation is dependent on GABA_A receptor subunit composition

This study investigated the ability of redox reagents to modulate native neuronal and recombinant GABA_A receptor function and addressed whether a presumed disulphide bridge, which appears as a ubiquitous structural feature of GABA_A receptor subunits, is susceptible to oxidation- reduction mechanisms. Other ligand-gated ion channels can also be regulated by redox reaction, notably, NMDA (Aizenman et al. 1989; Levy et al. 1990; Gilbert et al. 1991; Tang & Aizenman, 1993; Janaky et al. 1993; Sullivan et al. 1994; Köhr et al. 1994; Omerovic et al. 1995), nACh (Ben-Haim et al. 1973; Derkach et al. 1991; Xie et al. 1992; Sorenson & Gallagher, 1993; Servent et al. 1995) and glycine (Pan et al. 1995; Ruiz-Gomez et al. 1991) receptors. Our study suggests that redox reagents modulate GABA_A receptor function in a manner dependent upon the receptor subunit composition. Clearly the addition of the γ2 subunit to α1β1 or α1β2 receptors significantly decreased the sensitivity to DTT, glutathione and also to DTNB. This could arise either by the γ2 subunits occluding the site of redox regulation or by disruption of allosteric regulation. Interestingly, redox modulation of recombinant NMDA receptors is also dependent on receptor subunit composition (Köhr et al. 1994; Omerovic et al. 1995) suggesting that the structural arrangement of the whole receptor-channel complex is probably involved in determining the degree of redox sensitivity.

Site of action of redox reagents

The rapidity of onset and reversibility of redox modulation of GABA_A receptors, the partial permeability of DTT and membrane-impermeant nature of glutathione (Lauriault & O'Brien, 1991) suggested that a redox site is probably located on the external face of the GABA_A receptor. The ability of DTT, a specific sulfhydryl reducing agent, to reduce disulphide bridges to component sulfhydryl groups suggests that similar structures are being targeted on the GABA_A receptor, although an action elsewhere on the receptor cannot be discounted completely. The molecular structure reveals the presence of two conserved N-terminal cysteines separated by 13 amino acids which are presumed to participate in disulphide bridge formation, and from hydropathy plot analysis, to have an external location. This putative structural feature is also found in other members of the ligand-gated ion channel superfamily, notably nACh receptors and glycine receptors (Barnard et al. 1987), but it is only for the nACh receptor that a disulphide bridge has been demonstrated to form (Kao & Karlin, 1986). For GABA_A receptor subunits cysteine residues can also be found in the first and third transmembrane domains in addition to five cysteines in the large intracellular loop between transmembrane domains 3 and 4. Given the likelihood of an external site of action for the redox reagents and the postulated transmembrane topology of the GABA_A receptor, it appears likely that the two cysteines thought to form a disulphide bridge are likely candidates for the site of action of the redox agents. Indeed, intracellular application of 1 mM DTT via diffusion from the patch pipette electrolyte did not prevent the ability of extracellularly applied DTT (1 mM) to enhance GABA-activated currents for α1β1 constructs (data not shown).

The susceptibility of the α1βi subunit construct to both reducing and oxidizing reagents suggests that, under our experimental conditions, the cysteine residues are in equilibrium between fully oxidized and fully reduced forms in one or more subunits of the GABA_A receptor. The experiments reported here cannot discern between the possibility of the formation of disulphide bridges within each subunit or between neighbouring subunits. Indeed, a cross-link between adjacent cysteine residues does exist in recombinant nACh receptors (Kao & Karlin, 1986); however, immunoprecipitation of GABA_A receptor subunits under non-reducing conditions which presumably do not perturb the putative disulphide bridges still results in the isolation of single subunits (data not shown) tending to discount the possibility that cysteines cross-link neighbouring subunits. Unfortunately, site-directed mutagenesis on these cysteines to locate the redox regulatory site is precluded since mutation of one cysteine residue furthest from the N-terminal resulted in failure of that subunit to assemble (Amin et al. 1994), supporting the notion that a disulphide bridge actually forms in GABA_A receptors. This is interesting and further suggests that whilst a single oxidized cysteine or absent disulphide bond prevents successful assembly or transport to the cell surface, once the subunit is incorporated into the receptor structure, function is affected in a reversible manner by redox reaction without the loss of receptors from the surface membrane (see below).

Mechanism of action for redox modulation

Redox regulation by DTT and DTNB appeared to be independent of the GABA concentration used to activate the receptor. The concentration-response curves for α1βi constructs were not laterally displaced and the GABA EC₅₀ values were unchanged, which would be in accordance with DTT and DTNB not affecting the agonist binding site. However, redox modulation of the receptor clearly resulted in changes to the concentration-response curve maxima which could result from changes in single channel conductance, channel open and closed times, probability of channel opening or in the number of expressed receptors on the membrane surface. Single channel analysis revealed that DTT affected neither the single channel conductance nor the distribution of all open times suggesting no major changes to the underlying gating kinetics of the GABA_A receptor ion channel. However, the mean open time was increased, apparently due to an increased probability of channel opening. We can discount a change in the relative number of cell surface receptors following quantification of cell surface labelling with the 9E10 antibody to 9E10-epitope-tagged α1β2 receptors expressed in HEK cells (Table 2). Exchanging the β2 subunit for β1, or also including the γ2 subunit, did not affect cell surface receptor expression after exposure to DTT (Table 2).

Table 2.

Quantification of cell surface labelling of GABA_A receptors with 9E10 antibody

	Cell surface membrane binding (c.p.m.)
HEK cell treatment	− DTT	+ DTT
Untransfected HEK cells	121.78 ± 9.93	132.33 ± 26.57
α1β1	393.13 ± 24.66	343.83 ± 31.56
α1β2	701.80 ± 21.74	649.83 ± 25.50
α1β1γ2S	327.13 ± 13.32	337.23 ± 24.66
α1β2γ2S	522.68 ± 30.84	480.68 ± 23.45

Analysis of binding of¹²⁵I-labelled 9E10 antibody to untransfected HEK cells and HEK cells expressing α1β1, α1β2, α1β1γ2S and α1β2γ2S GABA_A receptors in the absence (− DTT) and presence (+ DTT) of 2 mm DTT. Values are means ± s.e.m. from 4 experiments. There are no significant differences between counts determined in the presence and absence of DTT (P > 0.05).

Kinetic model of redox modulation

To assess the likely mechanism of action of DTT and DTNB on the GABA_A receptor complex, a kinetic model was constructed to account for the whole-cell and single channel current data (Scheme 1). A cyclic scheme evolved incorporating oxidized forms of the receptor (Ox) in equilibrium with reduced forms (Red), and both forms could be activated from closed (C) to open (O) states by agonist (A) binding. The single channel analyses suggested that the distribution of all open times was best fitted by two exponential components resulting in the open states O₁ and O₂. Agonist binding was assumed to induce the open states such that O₁ was a monoliganded and O₂ a biliganded state. Since the whole-cell currents clearly exhibited desensitization, these open states were allowed to enter into desensitized states D₁ and D₂. The actions of DTT were accounted for by assuming that reduction of the receptor caused increased conversion of C(Ox) to C(Red) by increasing the rate constant k₁₁ and reducing k₁₂. Since the GABA EC₅₀ and individual open times from single channel analyses were unaltered, the reactions C(Red) or C(Ox) to O₁(Red), O₁(Ox) and O₂(Red), O₂(Ox) were considered to be unaffected by redox reactions. Lastly, the enhanced onset of desensitization by DTT was accounted for by DTT increasing k₃ enabling more rapid entry into desensitized state D₂. This model implicitly assumed that both the oxidized and reduced forms of the receptor are able to gate the ion channel. This is supported by GABA-activated responses for α1βi constructs being enhanced by subsequent application of DTT during the exposure to GABA (data not shown) and also by high concentrations of DTNB failing to completely abolish GABA-activated responses. The theoretical currents generated by the model illustrate the salient features of DTT action. The peak and steady-state currents are enhanced at low concentrations of GABA, but as the GABA concentration is increased only the peak current is enhanced, including the rate of desensitization (Fig. 9). The lack of effect of DTT on the steady-state current paradoxically occurs by DTT increasing k₄ to promote exit from D₂ to O₂(Red). Thus DTT, by increasing k₃, and to a lesser extent k₄, is essentially causing the receptor to achieve a rapid equilibrium with a desensitized state. The model accounted for the DTNB action by simply reversing the changes caused by DTT, as expected for a redox reaction. Thus DTNB reduces k₁₁ and increases k₁₂ to shift the receptor equilibrium towards C(Ox) slightly reducing k₃ and leaving k₄ unaffected. Since the cyclic model proposes that O₁(Ox) can convert directly to O₁(Red), and O₂(Ox) to O₂(Red), it is conceivable that DTT will influence k₂₁ and k₂₃ similarly to k₁₁. The model predicts that DTT, acting on the former rate constants, has only minimal effect on the whole-cell current profiles. The most significant change is to influence k₁₁ and k₁₂ to account for the data.

Scheme 1.

Figure 9. Kinetic simulation of whole-cell GABA-activated currents and modulation by redox reagents.

Theoretical computer simulated GABA-activated currents using Scheme 1, representing the effects of redox reagents DTT and DTNB on an α1βi GABA_A receptor. Microscopic association (M s⁻¹) and dissociation (s⁻¹) transition rate constants were chosen partly empirically and also according to experimental observation. Currents were activated by 0.01, 0.07, 0.6, 5, 38 and 300 μM GABA with k₁ = 5 M s⁻¹; k₂ = 0.7 s⁻¹; k₃ = 0.7 s⁻¹; k₄ = 0.04 s⁻¹; k₇ = 0.06 M s⁻¹; k₈ = 0.1 s⁻¹; k₉ = 0.6 s⁻¹; k₁₀ = 0.001 s⁻¹; k₁₁ = 0.1 s⁻¹; k₁₂ = 0.1 s⁻¹; k₁₃ = 0.006 M s⁻¹; k₁₄ = 0.01 s⁻¹; k₁₅ = 0.6 s⁻¹; k₁₆ = 0.001 s⁻¹; k₁₇ = 0.5 M s⁻¹; k₁₈ = 0.07 s⁻¹; k₁₉ = 0.7 s⁻¹; k₂₀ = 0.04 s⁻¹; k₂₂ = 0.01 s⁻¹; k₂₄ = 0.01 s⁻¹; k₂₁ and k₂₃ were calculated as functions of the other rate constants according to the law of microscopic reversibility. In the presence of 2 mM DTT, the following rate constants were altered to: k₃ = 1 s⁻¹; k₄ = 0.1 s⁻¹; k₁₁ = 1 s⁻¹; and k₁₂ = 0.01 s⁻¹. In the presence of 0.5 mM DTNB, these rate constants were altered to: k₃ = 0.7 s⁻¹; k₄ = 0.04 s⁻¹; k₁₁ = 0.01 s⁻¹; and k₁₂ = 100 s⁻¹.

Comparative neuronal studies

There are few reports of redox modulation of GABA_A receptors. In retinal ganglion cells, I_GABA was potentiated, though modestly, by DTT and inhibited by DTNB (Pan et al. 1995). In the present study both sympathetic and cerebellar granule neurones exhibited only minor if any sensitivity to DTT or DTNB, despite the parallel controls of DTT enhancing NMDA receptor-mediated currents in granule neurones and virtually abolishing nACh receptor-mediated responses in sympathetic neurones. Previous studies on GABA_A receptors suggested that DTT did not affect recurrent inhibition mediated by GABAergic synapses in the guinea-pig hippocampus (Tolliver & Pellmar, 1987). In addition, DTNB did not affect GABA receptor-mediated synaptic responses in rat hippocampus (Bernard et al. 1997). The insensitivity of GABA_A receptors in sympathetic and cerebellar granule neurones to DTT could be a consequence of large populations of GABA_A receptors containing γ subunits.

Physiological context

The potentiating effect of I_GABA mediated by α1βi subunit-containing GABA_A receptors by GSH has potentially important physiological implications. GSH is a natural antioxidant compound normally present at 1-2 mM external concentration in the CNS (Slivka et al. 1987; Yudkoff et al. 1990). The present study indicates that in the presence of GSH in vivo α1βi subtypes of GABA_A receptors reside in a reduced state until some pathological circumstance such as ischaemia alters the GSH/GSSG ratio (Rehncrona et al. 1980) making the receptor highly vulnerable to oxidation. The reduction in whole-cell responses by oxidation in cells expressing α1βi might result in reduced synaptic inhibition and could accelerate neurotoxicity previously ascribed to the condition of oxidative stress (Murphy et al. 1989). This scenario is pertinent to the thalamic nuclei where α1βi receptor subtypes are purported to exist. The apparently ubiquitous αβγ subunit construct found elsewhere in the CNS is clearly less sensitive to redox regulation despite the presence of cysteine residues in all these three subunits suggesting that the γ2 subunit-containing recombinant receptors are structurally organized in such a way that redox-sensitive sites are inaccessible.

Note added in proof

During the course of this work, we have exchanged views and results with Drs Z.-H. Pan and S. Lipton who were working independently on a similar project. The data are largely in accord and are in preparation for future publication.