Incremental conductance levels of GABA_A receptors in dopaminergic neurones of the rat substantia nigra pars compacta

Abstract

1. Molecular and biophysical properties of GABA_A receptors of dopaminergic (DA) neurones of the pars compacta of the rat substantia nigra were studied in slices and after acute dissociation.

2. Single-cell reverse transcriptase-multiplex polymerase chain reaction confirmed that DA neurones contained mRNAs encoding for the α3 subunit of the GABA_A receptor, but further showed the presence of α4 subunit mRNAs. α2, β1 and γ1 subunit mRNAs were never detected. Overall, DA neurones present a pattern of expression of GABA_A receptor subunit mRNAs containing mainly α3/4β2/3γ3.

3. Outside-out patches were excised from DA neurones and GABA_A single-channel patch-clamp currents were recorded under low doses (1-5 μM) of GABA or isoguvacine, a selective GABA_A agonist. Recordings presented several conductance levels which appeared to be integer multiples of an elementary conductance of 4-5 pS. This property was shared by GABA_A receptors of cerebellar Purkinje neurones recorded in slices (however, with an elementary conductance of 3 pS). Only the 5-6 lowest levels were analysed.

4. A progressive change in the distribution of occupancy of these levels was observed when increasing the isoguvacine concentration (up to 10 μM) as well as when adding zolpidem (20-200 nM), a drug acting at the benzodiazepine binding site: both treatments enlarged the occupancy of the highest conductance levels, while decreasing that of the smallest ones. Conversely, Zn²⁺ (10 μM), a negative allosteric modulator of GABA_A receptor channels, decreased the occupancy of the highest levels in favour of the lowest ones.

5. These properties of α3/4β2/3γ3-containing GABA_A receptors would support the hypothesis of either single GABA_A receptor channels with multiple open states or that of a synchronous recruitment of GABA_A receptor channels that could involve their clustering in the membranes of DA neurones.

GABA_A receptors are multi-subunit membrane-spanning proteins that mediate inhibitory transmission by gating a flux of chloride ions. To date, at least 17 subunits of GABA_A receptors belonging to 5 different classes (α, β, γ, δ and ε) have been cloned (see Macdonald & Olsen, 1994; Whiting et al. 1995, 1997). Single-channel experiments have revealed that native GABA_A receptors could open with multiple conductance levels (Mistry & Hablitz, 1990; Newland et al. 1991; Smart, 1992; Gage & Chung, 1994; Xiang et al. 1998). Interestingly, these levels are often integer multiples of an elementary value that varies according to the preparation. The presence of multiple conductance levels has also been described for other ligand-gated receptor channels (Cull-Candy & Usowicz, 1987; Jahr & Stevens, 1987; Mathie et al. 1991; Bormann et al. 1993; Lewis et al. 1997; Legendre, 1997; Ruiz & Karpen, 1997) as well as some voltage-dependent channels (Krouse et al. 1986; Matsuda, 1988). The multiple levels have been classically interpreted as different conductance states of a single channel (Macdonald & Olsen, 1994), but also as synchronized openings of ‘co-channels’ or ‘multi-barrelled channels’ (Krouse et al. 1986; Matsuda, 1988; Gage & Chung, 1994).

One way to study the multiple levels is to look for their modulations by pharmacological agents. Indeed recently, in dissociated hippocampal neurones, Eghbali et al. (1997) showed that diazepam, an allosteric agent acting at the benzodiazepine site (Lüddens et al. 1995), induced a sequential change in the conductance of GABA_A receptors that they interpreted in terms of synchronization. However, they showed no progressive interconversion between the conductance levels. Moreover, as revealed by in situ hybridization, hippocampal neurones can express enough subunit mRNAs to give rise to GABA_A receptor diversity within a single cell (Wisden et al. 1992). To study the multiple level single-channel properties of native GABA_A receptors, it is necessary to use preparations of single neurones with a well-characterized GABA_A subunit composition. Apart from the α1-containing GABA_A receptors, a major population of GABA_A receptors encountered in the brain is that containing the α3 subunit (McKernan & Whiting, 1996). Few studies are available on α3-containing receptors which seem to be similar to a type of GABA_A receptor (type II-BZD) present in hippocampal neurones (Lüddens et al. 1995). Previous immunostaining and in situ hybridization studies have revealed that dopaminergic (DA) neurones of the pars compacta of the substantia nigra (SNc) contain essentially α3 as the α subunit (Wisden et al. 1992; Persohn et al. 1992; Fritschy & Möhler, 1995). The molecular and functional properties of the GABA_A receptor channels present in the membrane of these DA neurones have not yet been characterized and this is the purpose of the present work. These DA neurones constitute a relatively homogeneous and well-delimited population of neurones (see Hainsworth et al. 1991) and receive strong inhibitory inputs which release GABA (Haüsser & Yung, 1994). From a functional point of view, they belong to the complex network of basal ganglia connections and provide the input to movement-related areas such as striatum, thalamus and superior colliculi, contributing to a wide variety of behavioural functions. Furthermore, they are implicated in numerous mental and neurological disturbances like Parkinson's disease, psychotic disorders and seizure propagation.

In DA neurones of the SNc, the single-cell reverse transcriptase (RT)-multiplex polymerase chain reaction (mPCR) technique was used to determine the presence of mRNAs encoding for 14 GABA_A receptor subunits. Using this technique, we confirm the presence of mRNAs encoding for the α3 subunit in about 50 % of neurones tested but also find mRNAs encoding for the α4 subunit in a similar proportion. The electrophysiological properties of the GABA_A receptor channels were investigated using membrane patches in the outside-out patch-clamp configuration from two preparations (slices and acutely dissociated neurones of the SNc). Our results show that GABA_A receptor stimulation by GABA or isoguvacine induces multiple opening levels of chloride channels, each opening level being an integer multiple of an elementary conductance level of 4-5 pS. Furthermore, zolpidem and Zn²⁺ ions, two agents acting on GABA_A receptors (Lüddens et al. 1995; Whiting et al. 1995), modulate the distribution of the various levels. In Purkinje neurones, that express mainly α1β2/3γ2 subunits (Ruano et al. 1997), we also find incremental conductance levels of GABA_A receptors.

METHODS

Brain slices

Wistar rat pups (10-18 days old) were anaesthetized with 1 % halothane and decapitated. The brain was quickly removed and placed in ice-cold phosphate-bicarbonate-buffered solution (PBBS) composed of (mM): 125 NaCl, 2.5 KCl, 0.4 CaCl₂, 1 MgCl₂, 25 glucose, 1.25 NaH₂PO₄, 26 NaHCO₃ and gassed with 95 % O₂-5 % CO₂ (pH 7.4). Coronal slices (250-300 μm thick) were cut from the mid-brain in the same solution with a DTK 1000 microslicer (D.S.K., Kyoto, Japan). Three to four slices containing the SNc were transferred to an incubating chamber maintained at 32-35°C and continuously superfused with oxygenated PBBS for 1 h. These slices were then used either for cellular dissociation or directly for patch-clamp recording of DA neurones. The same experimental procedure was used to obtain coronal slices of cerebellum.

Cellular dissociation

Cells of the SNc were enzymatically and mechanically dissociated following a modified version of the methods used by Hainsworth et al. (1991) and Mintz (1994). Briefly, three to four slices containing SNc were subjected to protease XXIII (3 mg ml⁻¹) or papain (15-20 U ml⁻¹; Boehringer) digestion for 10-15 min at 35°C in Hepes-saline composed of (mM): 82 Na₂SO₄, 30 K₂SO₄, 5 MgCl₂, 10 glucose, 10 Hepes (pH adjusted to 7.4 with NaOH). The slices were rinsed in Hepes-saline supplemented with trypsine inhibitor and bovine serum albumin (1 mg ml⁻¹ each) and two segments restricted to the right and left SNc were excised from each slice. Single cells were isolated from these segments by gentle trituration with fire-polished Pasteur pipettes and then plated on culture dishes (Primaria) maintained at room temperature (20-25°C). Dissociated cells were maintained in culture in the Hepes-saline supplemented with trypsine inhibitor and bovine serum albumin for at least 1 hour before use.

Anti-tyrosine hydroxylase immunocytochemistry

The proportion of putative DA neurones obtained after cellular dissociation of SNc segments was determined with immunocytochemistry using antisera against the DA synthesizing enzyme, tyrosine hydroxylase (TH). After plating on culture dishes, isolated cells were fixed for 30 min at 4°C in 4 % paraformaldehyde in 0.1 M phosphate buffer (pH 7.6). After two washes (10 min each), they were incubated for 5 min with 10 % methanol and 3 % hydrogen peroxide in 0.25 M Tris buffer containing 0.9 % NaCl (pH 7.6) and for 2-5 days at 4°C with a 1/1000 specific anti-TH serum (Institut J. Boy, Reims, France) in 0.25 M Tris buffer supplemented with 1 % normal goat serum and 1 % normal rat serum (Gibco BRL). After rinsing, the immunological reaction was visualized using the peroxidase anti-peroxidase (UCB Pharma, Nanterre, France) method, with goat anti-rabbit IgG (Cappel, West Chester, CA, USA) coupled to horseradish peroxidase revealed by the diaminobenzidine technique (Sternberger, 1979). Before paraformaldehyde fixation, photomicrographs of putative neurones (cells larger in size than 25 μm with neurites) were taken at a magnification of ×20 under phase-contrast microscopy, whereas after immunostaining against TH, photomicrographs of the same neurones were taken under direct illumination.

Cytoplasm harvesting and reverse transcriptase (RT)-multiplex polymerase chain reaction (mPCR)

Cytoplasm harvesting and reverse transcriptase were performed as described by Lambolez et al. (1995). Briefly, using the whole-cell patch-clamp configuration, the contents of the neurones (including the nucleus in some instances) were sucked up under visual control by application of a gentle negative pressure in patch pipettes filled with 8 μl of an autoclaved internal solution containing (mM): 120 CsCl, 3 MgCl₂, 5 EGTA, 10 Hepes (pH 7.2). Harvesting was interrupted as soon as the seal was lost. The tip of the pipette was then broken and the contents of the pipette expelled by applying positive pressure into a test tube containing 2 μl of a solution made of hexamer random primers (Boehringer; final 5 μM), the four deoxyribonucleotide triphosphates (dNTP) (Promega; final 0.5 mM), and dithiothreitol (final 10 mM). Immediately after expulsion, 20 units of ribonuclease inhibitor (Promega) and 100 units of Moloney murine leukemia virus reverse transcriptase (Gibco BRL) were added. Reverse transcriptase (RT) was performed for 1 h at 37°C in a final volume of about 10 μl. Single stranded cDNAs were stored at -80°C until used for PCR amplification.

Multiplex PCR amplification of the fragments of cDNA corresponding to two isoforms of glutamate decarboxylase (GAD 65 and GAD 67), tyrosine hydroxylase (TH) and 14 subunits of GABA_A receptors were performed (see Ruano et al. 1997) using the following set of primers (from 5′ to 3′):

GAD 65 sense: TCT TTT CTC CTG GTG GTG CC;

GAD 67 sense: TAC GGG GTT CGC ACA CCT C;

GAD 65/67 antisense: CCC CAA GCA GCA TCC ACA T;

TH sense: TAA AAC TCG GGA CCA CCA CGT T;

TH antisense: CAT TCC CAT CCC TCT CCT CAA A;

GABA_Aα1 sense: CAA GCC CGT GAT GAA GAA AAG;

GABA_Aα1 antisense: GTG GAA AAA TGT ATC TGG AGT C;

GABA_Aα2 sense: TCC AAG AAG ATG AGG CTA AAA ATA AT;

GABA_Aα2 antisense: AGC CAC TGA CTT TTT CCC GTT G;

GABA_Aα3 sense: TTT CTC CTC TCT GCT TCG GG;

GABA_Aα3 antisense: AAG CCC AGG TCG CAG TCG GTT GT;

GABA_Aα4 sense: CCC TCT CCT CGC ACC CTG;

GABA_Aα4 antisense: ATC AGA AAC GGG TCC AAA GC;

GABA_Aα5 sense: CAA GAA GGC CTT GGA AGC AGC TAA;

GABA_Aα5 antisense: GGT TTC CTG TCT TAC TTT GGA GAG;

GABA_Aα6 sense: AAG CCC CCG GTA GCA AAG TCA AAA;

GABA_Aα6 antisense: TAA GCG AGG AAA ATG GAA AAT AAC C;

GABA_Aβ1 sense: TCT CTC TTT TCC TGT GAT GGT TGC;

GABA_Aβ1 antisense: GTG ATC CGT AGT CCA TAG AGA ACA GT;

GABA_Aβ2 sense: TGG GGT GCT TTG TCT TTG TCT TTA;

GABA_Aβ2 antisense: TCA GGC GAC TTT TCT TTT GTG;

GABA_Aβ3 sense: GGC TTT TCG GCA TCT TCT CG;

GABA_Aβ3 antisense: CAT CAG GGT GGA GGC GGA;

GABA_Aγ1 sense: CAG AGA CAG GAA GCT GAA AAG CAA A;

GABA_Aγ1 antisense: CGA AGT GAT TAT ATT GGA CTA AGC C;

GABA_Aγ2S sense: AAG AAA AAC CCT GCC CCT ACA ATT;

GABA_Aγ2S antisense: TTC GTG AGA TTC AGC GAA TAA GAC;

GABA_Aγ2L sense: CTT CTT CGG ATG TTT TCC TTC AAG;

GABA_Aγ2L antisense: CAT AGG GTA TTA GAT CGT TGG ACT;

GABA_Aγ3 sense: CGA ATA AGC CTT CAA GCA CCC TCT;

GABA_Aγ3 antisense: CTT CTG TCA TCC TTC AGA GCA GCA;

GABA_Aδ sense: ATG GAC TAA TGG AGG GCT ACG C;

GABA_Aδ antisense: TCG GGC TGT AGG CGG ATA AG.

A first amplification round of 20 cycles (94°C for 30 s, 60°C for 30 s, 72°C for 65 s) was performed with a programmable thermocycler (Perkin Elmer/Cetus) in a single tube containing the 10 μl product of the RT used as a matrix, 100 nM of each of the 33 primers above, 50 μM of each dNTP and 2.5 U of 1 Taq polymerase (Perkin Elmer/Cetus) in a final volume of 100 μl.

One microlitre of the resulting cDNA product was used as a template for a second round of PCR of 40 cycles (94°C for 30 s, 56°C for 30 s, 72°C for 35 s), TH and GAD markers alone being amplified in three different test tubes with their specific primer mixture in a final volume of 100 μl. Ten microlitres of the amplification reaction was run in parallel with a known amount of a molecular weight marker (ΦX174, HAE III digested) on a 1.8 % agarose gel which was then stained with ethidium bromide. The sizes of the expected fragments were (in bp) for GAD 65, 391; GAD 67, 600; TH, 299. The specificity of the bands corresponding to the fragments was checked using specific restriction enzymes. The product of the digestion was run in parallel to the undigested product and the smaller DNA fragments expected were always observed.

When TH or GAD fragments were observed, another round of PCR (40 cycles round) was performed similarly using 1 μl of the first amplification cDNA product as a template. The amplification of the 14 GABA_A markers was thus carried out in 14 test tubes each containing a specific primer mixture. The sizes of the expected fragments were (in bp) for GABA_A receptor subunit α1, 393; α2, 319; α3, 339; α4, 369; α5, 361; α6, 389; β1, 431; β2, 359; β3, 416; γ1, 384; γ2S, 359; γ2L, 413; γ3, 354; δ, 312.

The efficiency of this protocol was tested on a sample of RNA purified from total rat brain. Figure 2A shows that all specific mRNAs were detected, each generating a PCR fragment of the size predicted by its mRNA sequence. To control for contamination, RT-mPCR was performed on aspirated extracellular medium. No band was observed on the agarose gel in this condition (not shown). The specificity of the RT-mPCR was tested on cytoplasm harvested from four Purkinje cells of the cerebellum. The RT-mPCR revealed in this case (not shown) the presence of mRNAs encoding for at least one isoform of the GAD and the expression pattern of GABA_A subunits expected in this cell type, i.e. α1, β2 and/or β3, γ2 and/or γ3 (Ruano et al. 1997).

Figure 2. Identification of the TH, GAD and GABA_A subunits mRNA fragments using the RT-mPCR technique.

A, rat neocortex (10 postnatal days old) RNAs (500 pg) extracted by the Chomczynski & Sacchi method (1987) were subjected to the RT-mPCR protocol (see Methods). The amplified fragments had the size predicted by the mRNA sequences. B, RT-mPCR-generated fragments amplified from cytoplasm harvested from a single neurone dissociated from the SNc. In this neurone (No. 5 of Table 2), TH and the α3, α4, α5, β3 and γ3 subunits of the GABA_A receptor were detected. Each lane is numbered according to the enzyme (GAD 65 and 67 in the same lane) or subunit concerned. The 17 PCR products were resolved in separated lanes by agarose gel electrophoresis in parallel with ΦX 174 HAEIII as a molecular weight marker (Φ) and stained with ethidium bromide. The positions of the 603 and 310 bp bands are indicated on the right of the gel.

Patch-clamp recording

Culture dishes containing dissociated neurones were continuously superfused at a flow rate of 1 ml min⁻¹ with external solution composed of (mM): 150 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 20 glucose, 10 Hepes (pH 7.4). Slices were placed under a Nomarski optic microscope in a recording chamber superfused at a flow rate of 1 ml min⁻¹ with oxygenated PBBS containing additional CaCl₂ (Ca²⁺ 2 mM) and 0.5 μM tetrodotoxin (Latoxan).

In the outside-out configuration of the patch-clamp technique, single-channel currents due to the application of GABA or isoguvacine were measured from neurones in slices or after cellular dissociation at room temperature using an Axopatch 200A amplifier (Axon Instruments), low-pass filtered at 5 or 10 kHz (4-pole Bessel) and recorded on digital audio tape (DTR-1204; Biologic, Claix, France). Recordings were made with fire-polished borosilicate glass pipettes (impedance 10-12 MΩ) coated with Sylgard resin (Rhône-Poulenc, Vitry, France) and filled with a solution containing (mM): 137 CsCl, 5 MgCl₂, 1 CaCl₂, 10 EGTA, 4 Na-ATP, 0.4 Na-GTP, 10 Hepes (pH adjusted to 7.3 with CsOH). In these conditions, the calculated theoretical chloride equilibrium potential (E_Cl) was about -2 mV and +3 mV for patches from dissociated neurones and neurones in slices, respectively. Drugs were applied in the bath.

Each time it was posssible (i.e. when the cell survived), the cytoplasm of the cell from which a successful recording of single-channel currents had been obtained was sucked up and the RT-mPCR was performed. However, we obtained a poor efficacy of the RT-mPCR in this condition (see Results).

Electrophysiological data analysis

The data were replayed from the digital audio tape, filtered at 1-2 kHz (-3 dB, 8-pole Bessel) unless otherwise mentioned and digitized at a sampling rate of 5 kHz using an interface coupled to a microcomputer running the pCLAMP5 Fetchex software (Axon Instruments). Visual inspection of the recordings revealed a large number of current levels, with poorly distinguished values. Moreover, the lowest presumed conductance level gave rise to small unitary currents (0.2-0.3 pA) that were close to the baseline. These observations led us to take into account only the recordings with a peak-to-peak noise level < 0.1 pA measured at a filtering of 1-2 kHz.

Using the pCLAMP6 Fetchan software (Axon Instruments), two types of amplitude histograms of isoguvacine-evoked openings were successively constructed by binning the current amplitude of single-channel recordings and the analysis of amplitudes was then done by maximum-likelihood fitting of Gaussian distributions to the histogram. The mean amplitude of each current level was given by the peak of each Gaussian curve and the chord conductance was calculated using the theoretical E_Cl.

First, in order to determine all the conductance levels recorded from a patch, several all-points histograms were constructed from brief periods of recordings (1.5-2 s). For each of these histograms, at least 10 selected portions of the recording (150-200 ms each) were sampled which were chosen by inspection to contain bursts of openings in only 1-3 distinct presumed conductance levels. At least 10 bursts of openings for each presumed level were included per all-points histogram and each of the presumed levels was represented in one or more histogram(s).

Second, to determine more accurately the conductance level values, amplitude distribution histograms were then constructed from long recordings (10-300 s) using the pStat software after applying the Fetchan semi-automatic ‘event-list’ algorithm. Using this algorithm, each opening detected (with a minimum duration of 1.5 ms) is attributed to one given amplitude level (imposed by the Fetchan software) thanks to a threshold-crossing procedure. For each given level, the thresholds are equidistant from this given level to the previous one (or baseline) and to the following pre-selected level. Up to five pre-selected amplitude levels were fixed using the amplitude values obtained from the previous all-points histograms: only the 5-6 smallest conductance levels were analysed. Whatever its duration, an opening counts 1 in the histogram and the value incremented in the histogram (with a bin width of 0.03 pA) is the mean of the samples of the opening in the given level (after the exclusion of the first and last samples in order to take into account the distortion of the endpoints of the events by the filter). We used a baseline update strategy, keeping the deltas between the levels. The amplitude distribution histograms obtained by this method allow the study of much longer duration recordings which was not possible for the all-points histograms because of baseline drift. Amplitude distribution histograms obtained from stretches of real baseline noise recordings (with 5-6 fixed levels equally spaced) did not give rise to a peak.

Additionally, in the recordings for which the baseline was particularly stable, all-points histograms including every opening were constructed from long recordings (10-300 s) that were filtered at 0.2 kHz to lower the variance of the peaks. Such histograms allowed us to check that there were not other levels than those revealed by the previous method.

We wanted to determine for each level the fraction of the total time measured for which the level was observed. The occupancy of one level ‘x’ was therefore calculated as d_x/D, where d_x was the open-state duration at one level ‘x’ during the recording analysed and D was the total duration of that recording. These durations were determined by the pCLAMP6 pStat software after the amplitude distribution histogram had been constructed. In addition, the fraction of the total open-state time (open-state occupancy) was d_x/d_all where d_all represented the total open-state duration at all levels during this recording. The total occupancy of all open levels is then d_all/D.

Chemicals and drug solutions

Except where mentioned, all chemicals were obtained from Sigma-Aldrich. An isoguvacine (Tocris) stock solution (10 mM in water) was diluted with external recording solution to the indicated final concentration.

RESULTS

Experiments were carried out in parallel on dopaminergic (DA) neurones of the rat substantia nigra pars compacta (SNc) in two different preparations: brain slices and dissociated neurones. DA neurones were easier to identify in slices but patch-clamp recording of dissociated neurones gave the advantage of a better signal-to-noise ratio. Comparable results were obtained from both preparations. Additional experiments were performed on cerebellar Purkinje neurones in slices.

Molecular characterization of DA neurones TH immunostaining of SNc dissociated neurones

More than 85 % of nigrostriatal neurones express TH immunoreactivity in situ (Gerfen et al. 1987). We checked whether this proportion was kept after cellular dissociation of SNc. From three different dissociations, 28, 35 and 51 putative DA neurones were selected on the basis of the presence of elongated neurites and a cell body larger than 25 μm (Fig. 1A1 and A2). After anti-TH immunocytochemistry, 71 ± 5 % of the selected neurones were stained (Fig. 1B1, B2, C and D).

Figure 1. Morphological aspects of isolated neurones selected for electrophysiological analysis.

A and B show different types of selected neurones using phase contrast microscopy before paraformaldehyde fixation (A1 and B1), and the same neurones seen by direct light microscopy (A2 and B2) after immunocytochemistry for tyrosine hydroxylase (see Methods). The neurone in A2 and only one of the two in B2 were stained. Thus, dopaminergic neurones could not be strictly identified according to their shape. C and D show two other selected neurones showing positive immunostaining for tyrosine hydroxylase and presenting variable perikaryal morphology and elongated neurites (scale bars 30 μm).

Single-cell expression of TH and GAD (65 and 67) mRNAs in SNc

The cytoplasm of 49 neurones in slices and 47 dissociated neurones was harvested under visual control and the pipette content was used for reverse transcription (RT) and multiplex polymerase chain reaction (mPCR) (see Methods). This RT-mPCR selectively amplified at the single-cell level fragments of DNA specific for the mRNAs encoding for (i) TH, (ii) both isoforms of GAD (GAD 65 and GAD 67) and (iii) 14 subunits of the GABA_A receptor. Figure 2A shows that all specific mRNAs were detected in RNA purified from total rat brain, each generating a PCR fragment of the size predicted by its mRNA sequence. An example of a result obtained for a single neurone is presented in Fig. 2B.

After RT-mPCR for TH, GAD 65 and GAD 67, only 32 of 49 neurones in slices and 23 of 47 dissociated neurones of SNc showed the presence of mRNAs encoding for at least one of the three enzymes. The RT-mPCR had therefore an efficacy of about 65 % for the slices and only 50 % after cellular dissociation. Table 1 summarizes the percentages of neurones in which the mRNAs encoding for TH and/or at least one GAD isoform could be detected. The comparison between dissociated neurones and neurones in slices shows similar results. Respectively, 74 and 69 % of the neurones expressed the mRNA encoding for TH but not for any GAD isoform, as would be expected for the DA neurones. Indeed, the values are close to those found in situ for TH immunoreactivity (Gerfen et al. 1987), but they are higher than those measured from dissociated primary neuronal cultures obtained from the whole substantia nigra (reticulata and pars compacta) of 2-3 day old rats (Masuko et al. 1992). Therefore, these data validate our dissociation method as a way to obtain a high proportion of isolated DA neurones. The following results concerning the mRNAs coding for GABA_A receptor subunits will be restricted to this population of neurones expressing the mRNA coding for TH but not for any GAD isoform. Among this population, three dissociated neurones and three neurones in slices were first the object of an outside-out patch excision leading to a successful single-channel recording.

Table 1. Distribution of mRNAs encoding for tyrosine hydroxylase and two isoforms of glutamic acid decarboxylase in neurones of the rat substantia nigra (pars compacta).

	Dissociated neurones (n = 3) (%)	Neurones in slices (n = 3) (%)
TH+/GAD−	74	69
TH+/GAD+	17	25
TH−/GAD+	9	6

The values indicate the percentage of neurones in which the presence of mRNAs encoding for tyrosine hydroxylase (TH) and/or at least one isoform of glutamic acid decarboxylase (GAD 65 and/or 67) were detected by RT-mPCR in the aspirated cytoplasm.

In addition to this population, 9 % of dissociated neurones and 6 % of neurones in slices expressed the mRNAs encoding for at least one GAD isoform but did not express the mRNAs encoding for TH, as would be expected for GABAergic interneurones. Finally, 17 and 25 % of the neurones, respectively, expressed simultaneously the mRNAs encoding for both TH and at least one isoform of GAD (mainly GAD 65). One can argue that the presence of mRNAs encoding for TH and GAD cannot be taken as evidence that TH and GAD proteins or dopamine and GABA were present, since no quantitative conclusion about the levels of TH and GAD expression can be made from our RT-mPCR data. However, the colocalization of TH and GAD has been demonstrated by immunocytochemistry in a subpopulation of SNc neurones (Campbell et al. 1991) and could be a property of immature neurones. Hattori et al. (1991) have also shown that in the rat striatum, single DA neurones originating from the substantia nigra can make two types of synaptic contacts with striatal neurones: symmetric synapses en passant containing TH and dopamine and asymmetrical terminal boutons of non-characterized chemical nature. The neurotransmitter present in these boutons could thus be GABA.

GABA_A receptor subunit mRNAs of single DA neurones

Table 2 shows the results of the RT-mPCR for the population of DA neurones. The pattern of expression presents a large heterogeneity among the cells tested, in dissociated neurones as well as in neurones in slices. Moreover, it appears that some subunit mRNAs may have been lost during the protocol since (i) more mRNAs coding for GABA_A subunits were found in neurones in slices than in dissociated neurones, (ii) some dissociated neurones lacked mRNA for α subunit (i.e. cells 1, 11 and 12 in Table 2). It is likely that mRNAs for GABA_A subunits are not as concentrated as mRNAs for TH. In addition, a loss of mRNAs may have occurred while the cytoplasm was being sucked up, reducing the mRNA final concentrations to below the detection threshold of the RT-mPCR, and the dissociation procedure (that cut the dendrites) as well as the excision of outside-out patches could have further reduced these concentrations. However, 2 of 6 neurones in slices (i.e. cells 6 and 7 of Table 2) from which an outside-out patch-clamp recording was successfully analysed in the following section showed mRNAs for GABA_A subunits which were representative of the whole population.

Table 2. Expression pattern of the mRNAs encoding for GABA_A receptor subunits in dopaminergic neurones of the rat substantia nigra (pars compacta).

		GABAA receptor subunits of dissociated neurone
No.	Age	α1	α2	α3	α4	α5	α6	β1	β2	β3	δ	γ1	γ2S	γ2L	γ3
1	10													+	+
2	10			+										+
3	10			+	+	+			+		+
4	10				+
5	11			+	+	+				+					+
6	11								+
7	11						+			+	+			+
8	11				+										+
9	11			+									+		+
10	12	+
11	12									+					+
12	12												+
13	12			+					+	+	+
14	14			+	+
	%	7	—	43	36	14	7	—	21	28	21	—	14	21	36

		GABAA receptor subunits of single neurones in slices
No.	Age	α1	α2	α3	α4	α5	α6	β1	β2	β3	δ	γ1	γ2S	γ2L	γ3
1	10			+	+	+				+
2	10			+	+				+
3	10			+	+
4	10			+
5	11				+				+						+
6	12	+		+	+					+			+		+
7	12			+											+
8	13						+		+	+	+
9	15				+										+
10	15			+	+					+
11	16			+						+
12	16				+	+			+	+					+
13	16	+			+				+						+
14	18	+		+	+									+	+
15	18				+				+	+
16	18	+			+	+			+	+			+	+?	+
17	18			+					+	+
18	18				+	+				+
	%	22	—	56	72	22	6	—	44	56	6	—	11	11	44
Summary of the results obtained for dissociated neurones and neurones in slices (n = 32)
		α1	α2	α3	α4	α5	α6	β1	β2	β3	δ	γ1	γ2S	γ2L	γ3
	%	16	—	50	56	19	6	—	34	44	12	—	12	16	41

RT-mPCR for 14 subunits of GABA_A receptors was performed in neurones positive for TH and negative for both isoforms of GAD (TH⁺/GAD⁻, see Table 1). For each neurone tested, the presence of mRNAs encoding for a subunit is indicated + in the corresponding column. The number of the neurones as well as the age of the rats (in days) are indicated on the left. Successful patch-clamp recordings were obtained from neurones 6 and 7 in slices. The percentages at the end of each section represent the percentages of cells tested in which subunits were found.

Therefore, the results have to be considered at the level of the population of neurones rather than at the level of the single neurone and several other observations can be made: (i) the mRNAs encoding for the GABA_A subunits α2, β1 and γ1 were never found, (ii) the more frequent mRNAs detected (> 34 %) were those for α3 and α4, β2 and β3, and γ3 (α4, β2 and β3 being markedly more frequent in neurones from slices than in dissociated neurones) and (iii) mRNAs for α1, α5, α6, γ2S, γ2L and δ were more rarely observed (< 19 %).

These data are in agreement with in situ hybridization experiments (Wisden et al. 1992; Persohn et al. 1992), but due to the sensitivity of the RT-mPCR technique, also reveal the presence of mRNAs encoding for subunits that had not been found previously, as is the case for α6 (found in only two cells and thus very rare). The GABA_A receptors of the DA neurones of the SNc may thus contain α3 and/or α4 subunits. The following electrophysiological experiments focused on these GABA_A receptors. However, apart from the α3/4-containing receptors, the α1-containing receptors are another major population of GABA_A receptors (McKernan & Whiting, 1996). Additional experiments were performed on Purkinje neurones that mainly contain α1β2/3γ2 subunits (Ruano et al. 1997 and our results, see Methods).

Electrophysiological analysis of the GABA_A receptors

Outside-out membrane patches were excised from SNc dissociated neurones and from DA and cerebellar Purkinje neurones in slices. At a holding potential of -70 mV, low concentrations of GABA or isoguvacine (1-5 μM) evoked single-channel openings with multiple levels. Figure 3A and B illustrates typical recordings obtained from SNc dissociated neurones in the presence of 2 μM GABA and 1 μM isoguvacine, respectively. Comparable recordings were obtained from DA neurones in slices, and particularly for neurones 6 and 7 of Table 2. Isoguvacine is a selective GABA_A receptor agonist that has been claimed to preferentially activate subconductance states (Mistry & Hablitz, 1990). In the following results, isoguvacine has been used instead of GABA.

Figure 3. GABA_A single-channel openings recorded from outside-out patches excised from dissociated neurones of the rat substantia nigra (pars compacta).

A, multiple openings recorded in response to 2 μM GABA. B, examples of recordings obtained from different patches in response to 1 μM isoguvacine. Multiple openings with current amplitudes varying from about -0.2 to -4 pA were obtained. All the patches were held at -70 mV. The recording pipette was filled with a solution containing a chloride concentration close to that of the external solution. Recordings were filtered at 1-2 kHz (-3 dB, 8-pole Bessel), except for recordings Ba and Bb (obtained from the same patch), which were filtered at 0.5 kHz. Note that in the same recording, several amplitude levels were reached. An example of transitions between amplitude levels is illustrated in panel Bd. Dashed lines were placed at the level determined as the peak of the Gaussian components fitted to the amplitude distribution.

Isoguvacine-mediated single-channel currents

When 1 μM isoguvacine was applied to patches from SNc dissociated neurones, openings either occurred as single events of variable amplitude and duration or were grouped in prolonged bursts (Fig. 3B). In every patch (n = 42), the smallest unitary currents easily separable from the baseline had amplitudes of about -0.25 pA (Fig. 3Ba). Multiple values of current amplitude were encountered (Fig. 3Ba-i) that could occur directly from the baseline, with frequently no evidence of superimposed openings. The highest current level encountered differed from one patch to another: in several patches, it only reached about -0.5 pA (Fig. 3Bb) or -0.75 pA (Fig. 3Bc) but recordings in which at least six current levels were present (from -0.25 to -1.5 pA) were the most frequently observed (Fig. 3Bd). In some patches, the largest events could reach -4 pA: an example of an opening to a level around -3.75 pA is given in Fig. 3Bi. When a level was reached in a recording, all the smaller ones were also encountered and direct transitions from one current level to another were often observed, whatever the levels concerned. The trace in Fig. 3Bd shows an example of transitions between levels. In this study, because of the complexity of the recordings, we have chosen to focus on the analysis of the 5-6 smallest opening levels. Only the currents recorded in membrane patches excised from nine SNc dissociated neurones, six DA neurones in slices and two Purkinje neurones in slices have been analysed because of their very low noise and good stability plot (not shown) during 15 min following the excision of the patch.

The reversal potential of the currents recorded in dissociated neurones (Fig. 4A) as well as neurones in slices (not shown) approximated 0 mV which is close to the theoretical value of E_Cl (-2 mV for dissociated neurones and +3 mV for neurones in slices). The GABA_A single-channel activity was completely inhibited by 10 μM bicuculline, a specific GABA_A receptor antagonist (Fig. 4B).

Figure 4. Currents reverse at E_Cl and are blocked by bicuculline.

A, examples of single-channel current recordings at different holding potentials from an outside-out patch excised from a dissociated neurone of the SNc and perfused with 1 μM isoguvacine. B, current recordings at a holding potential of -70 mV from a patch excised from a dissociated neurone of the SNc perfused with the recording solution alone (control), or with 1 μM isoguvacine or 1 μM isoguvacine plus 10 μM bicuculline added, as indicated. In the presence of isoguvacine, the mean open probability was about 0.3. Isoguvacine plus bicuculline were applied together for at least 2 min.

Figure 5A shows single-channel recordings from a single patch excised from a SNc dissociated neurone. In the presence of 1 μM isoguvacine, currents of different amplitudes could be observed at a holding potential of -70 mV. The all-points histograms constructed from short recordings (1.5-2 s) and fitted by the sum of Gaussian curves (Fig. 5B) showed that the mean amplitudes at the peaks of the Gaussian curves seemed to increase by discrete steps, the values of which appeared to be integer multiples of about -0.25 pA. According to the calculated theoretical E_Cl value, this current corresponds to a chord conductance of about 4 pS.

Figure 5. Single-channel current amplitudes present discrete distribution levels.

A, examples of single-channel currents recorded from a single outside-out patch excised from a SNc dissociated neurone and induced by 1 μM isoguvacine. At least 5 different conductance levels (as indicated by the dashed lines) could be observed during the recording. Holding potential -70 mV. E_Cl -2 mV. B, all-points histograms constructed from short period recordings from the same patch as in A. The histograms represent 10-12 selected periods of recordings (150-200 ms) similar to those presented in A, so that the total time per histogram was 1.5-2 s. The points were fitted by the sum of Gaussian curves. For example, 10 traces like the ones presented in Aa-c were used to construct the histograms Ba-c, respectively. The values indicated at the peak of each Gaussian curve represent the mean conductance (pS). Five values were obtained which were integer multiples of about 4 pS.

To determine accurately the values of the conductance levels, a representative number of openings has to be studied. By pre-selecting current amplitude levels and updating the baseline, we constructed amplitude distribution histograms for long periods of recordings (> 10 s, see Methods). Examples of such histograms are shown in Fig. 6. The histogram shown in Fig. 6Aa, which was fitted with five Gaussian curves, was constructed from the patch excised from the SNc dissociated neurone already illustrated in Fig. 5. The current amplitude means at the Gaussian peaks give conductance values that increase by steps of 4 pS. Similar conductance values are deduced from the all-points histogram constructed from the same recording filtered at 0.2 kHz (Fig. 6Ab), which validates our pre-selection method.

Figure 6. Determination of the conductance values of the different levels.

A, from the recording obtained from the dissociated neurone already shown in Fig. 5, two different histograms were constructed. The histogram Aa is an amplitude distribution histogram constructed with 5 pre-selected levels (see Methods) for which the recording was filtered at 1 kHz (-3 dB, 8-pole Bessel) and digitized at a sampling rate of 5 kHz. The values of the pre-selected amplitude levels were chosen using the mean amplitude values of the peaks obtained in the all-points histograms presented in Fig. 5B. The histogram Ab is an all-points histogram taking into consideration all the openings of the recording after filtering at 0.2 kHz (some portions of the recording containing no openings were not included). The points were fitted by the sum of Gaussian curves. The values indicated at the peak of each Gaussian curve represent the mean conductance (pS). The recording durations were 42 and 38 s, respectively, and E_Cl was -2 mV. The current amplitude mean values (in pA), standard errors of the mean (s.e.m.) and relative areas under each peak (%) of the Gaussian components fitted to the distributions are as follows.

	Amplitude distribution histogram					All-points histogram
Mean	−1.36	−1.03	−0.80	−0.57	−0.29	−1.15	−1.02	−0.79	−0.55	−0.29
±s.e.m.	0.10	0.03	0.04	0.03	0.06	0.05	0.01	0.01	0.01	0.01
%	13.1	17.5	26.3	20.7	22.4	25.8	4.3	26.4	21.3	22.2

B and C, amplitude distribution histograms constructed with 5 pre-selected levels for a patch from a DA neurone in a slice in the presence of 2 μM isoguvacine and a patch from a cerebellar Purkinje neurone in a slice (10 μM isoguvacine). Recordings were filtered at 1 kHz (-3 dB, 8-pole Bessel) and digitized at a sampling rate of 5 kHz. The recording durations were 169 and 117 s, respectively, and E_Cl was +3 mV. The current amplitude mean values (in pA), standard errors of the means (s.e.m.) and relative areas under each peak (%) of the Gaussian components fitted to the distributions are as follows.

	DA neurone in slice					Purkinje neurone in slice
Mean	−2.12	−1.56	−1.18	−0.76	−0.38	−1.25	−0.93	−0.68	−0.46	−0.22
±s.e.m.	0.11	0.04	0.04	0.03	0.03	0.07	0.04	0.03	0.03	0.04
%	7.4	6.5	5.0	23.4	57.7	3.3	3.0	8.8	15.6	69.0

Amplitude distribution histograms presenting several sharp Gaussian peaks were also obtained from patches excised from other SNc dissociated neurones and DA neurones in slices (Fig. 6B). The presence of several conductance levels has already been described in dissociated Purkinje neurones (Smart, 1992). In Purkinje neurones in slices, our analysis also revealed the presence of several Gaussian peaks of the amplitude histogram (Fig. 6C).

For every patch analysed (n = 17), we were able to obtain comparable histograms with clearly distinguishable peaks. As seen in Fig. 6, the mean conductance values of the highest levels were approximately integer multiples of the conductance of the smallest unitary level recorded in the patch, i.e. 4 pS for the SNc dissociated neurones, 5 pS for DA neurones in slices and 3 pS for Purkinje neurones in slices. When levels of openings higher than those analysed were observed, their conductances (determined using all-points amplitude histograms) were also approximately integer multiples of the smallest conductance level.

Table 3 summarizes the mean conductance values for the various levels obtained from single-channel current recordings from outside-out patches of nine SNc dissociated neurones and six DA neurones in slices. Up to six conductance levels (I-VI) were studied for the dissociated neurones and five for the neurones in slices. The conductance values of the lowest level (level I) were found to be (means ±s.d.) 4.3 ± 0.4 pS (n = 6) and 5.1 ± 0.5 pS (n = 4), respectively, a difference which is significant (Student t test: t = 2.7, P < 0.05) but which we cannot explain. The following levels have conductances which are approximately integer multiples of the lowest value, with a mean increment of, respectively, 3.6 ± 0.8 pS (n = 31) and 5.0 ± 1.1 pS (n = 19), the difference between these two values being also significant (t = 5.2, P < 0.001). Finally, whatever the conductance value of level I (which varied depending on the neuronal preparation used), higher level conductances were always integer multiples of this smaller conductance value. Summations of openings cannot be responsible for the multiple conductance levels (although summations were also observed) since direct opening to high levels occurred even when the open probability was extremely low (< 0.01).

Table 3. Average conductance of different levels.

	Current level
	I	II	III	IV	V	VI
Dissociated neurones Conductance (pS)	4.3 ± 0.4	7.6 ± 0.1	11.1 ± 0.3	14.6 ± 0.4	18.2 ± 0.4	21.1 ± 0.8
(N = 9)	(n = 6)	(n = 5)	(n = 8)	(n = 9)	(n = 9)	(n = 7)
Neurones in slices Conductance (pS)	5.1 ± 0.5	10.8 ± 0.5	15.4 ± 0.8	19.7 ± 0.5	25.8 ± 1.1	Not
(N = 6)	(n = 4)	(n = 6)	(n = 5)	(n = 6)	(n = 5)	determined

For each patch, an amplitude distribution histogram was constructed and fitted by the sum of Gaussian curves. The mean amplitude of the single-channel currents recorded at -70 mV was determined at the peaks of the Gaussian curves. However, not all the curves showed six peaks corresponding to the levels I-VI (as evidenced by the value of n in parentheses which is in some cases lower than N, the number of patches analysed). Conductances were calculated using theoretical values of E_Cl (−2 and +3 mV for dissociated neurones and neurones in slices, respectively) and were averaged for each level. Note that the conductance mean values differ slightly from one neuronal preparation to another, but that in both cases the means of the levels II–VI are approximately integer multiples of the respective level I. Data are expressed as means ± S.D.

Effects of zolpidem and Zn²⁺ ions

We have investigated the effects of the application of zolpidem (20 and 200 nM) and Zn²⁺ (10 μM) on single-channel currents evoked by 1-2 μM isoguvacine in patches from SNc dissociated neurones. The imidazopyridine zolpidem is a positive modulator acting at the benzodiazepine site of GABA_A receptors (Lüddens et al. 1995). Zn²⁺ is a non-competitive antagonist the effects of which depend on the receptor subunit composition (see Macdonald & Olsen, 1994; Whiting et al. 1995). As seen in the recordings shown in Fig. 7A and B, the conductance mean value for each level was unchanged in the presence of zolpidem (n = 3) or Zn²⁺ (n = 5), which was confirmed by the study of the amplitude distribution histograms (not shown). We have studied the effects of these drugs on the total occupancy of all levels (d_all/D) and on the occupancy (d_x/D) and the open-state occupancy (d_x/d_all) of the various levels (see Methods). In the presence of 1-2 μM isoguvacine, the total occupancy (d_all/D) varied widely according to the patches from 0.5 % to 35 % (mean ±s.d.; 11.9 ± 12.8 %, nine patches from SNc dissociated neurones).

Figure 7. Opposite effects of zolpidem and Zn²⁺ ions on the occupancy of different current levels.

A, recordings from an outside-out patch of a SNc dissociated neurone (held at -70 mV) illustrate the increase in the occupancy of current levels induced by zolpidem (20 or 200 nM) in the presence of 1 μM isoguvacine. Notice that the occupancy of level II was particularly increased. B, examples of current recordings from a patch excised from another dissociated neurone showing the effects of Zn²⁺ ions. When 10 μM Zn²⁺ was added to isoguvacine (2 μM), the occurrence of openings of large amplitude (> 1.5 pA) was decreased. Notice also that the remaining levels are essentially the levels I and II. Ca, the occupancy of the different levels (from I to V and all the levels ≥ VI pulled together) is presented in the three pharmacological conditions illustrated in A (1 μM isoguvacine, 1 μM isoguvacine + 20 nM zolpidem, 1 μM isoguvacine + 200 nM zolpidem). The measurement durations were 180, 183 and 120 s, respectively. In the presence of 20 nM zolpidem, the total occupancy was increased from 12.9 to 24.6 % and the occupancy of levels II and IV was at least tripled, whereas, interestingly, that of level I was reduced by two. At 200 nM zolpidem, the total occupancy was further increased to 29.9 %. Overall, the distribution of the occupancy of various levels was similar to what was observed with the 20 nM treatment with, however, a larger increase of level II, the occupancy of level I remaining low. Cb, in the presence of 10 μM Zn²⁺, the total occupancy was decreased from 4.5 to 3.6 %. The occupancy of all levels ≥ IV decreased. That of level I and, to a lesser extent that of level II, increased. This effect was reversible. The measurement durations were 307, 157 and 65 s, respectively.

Zolpidem always increased d_all/D, an effect that was more important in patches where d_all/D was low. For three patches, d_all/D was increased from 7 ± 5 % to 28 ± 4 % by 20 nM zolpidem. As illustrated for one patch (Fig. 7A and 7Ca), the increase in d_all/D was due essentially to an increase in the occupancy of levels II, III, IV and ≥ VI, whereas interestingly, the occupancy of level I (4 pS) was reduced. This effect was reproduced in two other recordings. For the same patch, at 200 nM zolpidem (Fig. 7A), the distribution of the occupancies was almost unchanged compared with the 20 nM treatment, except that the occupancy of levels II and III was further increased.

Zn²⁺ reversibly decreased the total occupancy by 40 ± 16 % (n = 5), which is consistent with previous studies (Smart, 1992). Zn²⁺ effects were more pronounced in patches in which d_all/D was high than in patches with a low d_all/D. Figure 7B shows recordings obtained from a patch for which d_all/D was low. The occupancy of levels ≥ IV was decreased by 10 μM Zn²⁺, whereas the occupancy of levels ≤ II was increased despite the decrease in d_all/D (Fig. 7C b). Similar observations were made for four other patches. Therefore, Zn²⁺ appears to produce a reduction in the amount of time spent in the highest levels and a concomitant increase in the amount of time spent in the lowest levels.

Dose-dependent effect of isoguvacine

We have investigated the effect of bath application of increasing concentrations of isoguvacine on the occupancy (d_x/D) and the open-state occupancy (d_x/d_all) of the various current levels. The results shown in Fig. 8 were obtained from an outside-out patch for which the total occupancy (d_all/D) in the presence of 1 μM isoguvacine was only 0.79 %, the probability of having random summations of openings being thus very low. An increase in the isoguvacine concentration to 2 or 5 μM did not produce noticeable effects on d_all/D (respectively, of 0.51 % and 0.84 %). However, it did produce a decrease in the occupancy of level I compensated by an increase in the occupancy of levels II and ≥ V, without affecting the occupancy of levels III and IV. At 10 μM isoguvacine, d_all/D was increased to 4.7 %. In that case, the occupancy of levels ≥ II was increased, whereas the occupancy of level I was restored close to its value in 1 μM isoguvacine (Fig. 8A). However, when looking at the open-state occupancy of the various levels (Fig. 8B), it appears clear that the fraction of the total open-state time of level I (d_I/d_all) decreased gradually in 1-10 μM isoguvacine, in parallel to the gradual increase in d_{≥ V}/d_all. Finally, increasing the concentration of isoguvacine from 1 to 10 μM had a dual effect on the single-channel GABA_A currents: first it changed the distribution of the amount of time spent in various levels, decreasing time spent in the lowest level while increasing the highest levels, and second, at 10 μM, it increased the total occupancy by increasing the occurrence of the highest levels.

Figure 8. Dose-dependent effect of low concentrations of isoguvacine on the occupancy of GABA_A receptor channel current levels.

A, the occupancy of the different levels (from I to V and all the levels ≥ VI pulled together) was illustrated for an outside-out membrane patch of a SNc dissociated neurone perfused with increasing concentrations of isoguvacine (1, 2, 5 and 10 μM). The total duration of the analysed recordings was 167, 99, 271 and 141 s, respectively. In 1-5 μM isoguvacine, the total occupancy of all levels was similar (< 1 %) but the occupancy of level I decreased whereas the occupancy the other levels increased (except that of levels III and IV). At 10 μM isoguvacine, the total occupancy was increased to 4.7 %. The occupancy of all levels was increased except that of level I which did not reach the value it had in 1 μM isoguvacine. B, open-state occupancy of each level calculated for the same recordings as in A. Increasing the concentration of isoguvacine from 1 to 10 μM, the open-state occupancy of level I gradually decreased in contrast to that of levels ≥ V which gradually increased.

DISCUSSION

Subunit composition of GABA_A receptors in SNc neurones

The RT-mPCR results obtained in this study suggest that the GABA_A receptor subunit mRNAs present in the DA neurones tested were not all systematically revealed at the single-cell level, particularly in some dissociated neurones in which we were unable to detect any α subunit mRNAs. Nevertheless, considering the whole population of DA neurones, our results indicate that the main GABA_A receptor subunits expressed are α3, α4, β2, β3 and γ3, whereas α1, α5, α6, γ2S, γ2L and δ subunits could also be found in smaller proportions. We have never found mRNAs encoding for α2, β1 and γ1 subunits.

As the stoichiometry of native receptors would be a pentamer composed of 2α plus either 1β-2γ or 2β-1γ, numerous putative combinations are possible, but this number can be limited by using pharmacological data. First, GABA_A receptors of the DA neurones are sensitive to zolpidem (20 and 200 nM), which is consistent with the type II-BZD receptor (Lüddens et al. 1995). However, it has been shown that recombinant receptors containing α4 as the only α subunit are insensitive to zolpidem (Wafford et al. 1996), whereas those containing at least one α3 subunit are sensitive to zolpidem and form a type II-BZD receptor (Criswell et al. 1997). This implicates the presence of, at least, one α3 subunit in some of the GABA_A receptors of DA neurones. It has been suggested that receptors with a high affinity for zolpidem may exist that contain the γ3 subunit, and possibly the α3 subunit also (McKernan et al. 1995). It is therefore likely that, in combination with an α3 subunit, at least one γ3 subunit is present in GABA_A receptors of the DA neurones of the SNc. Second, GABA_A receptors of DA neurones always presented a sensitivity to Zn²⁺ ions and we found a small percentage of cells expressing γ2 and/or δ subunit mRNAs. That the presence of one γ2 subunit confers an insensitivity to Zn²⁺ is still controversial (Smart, 1992; White & Gurley, 1995) whereas the presence of a δ subunit was well demonstrated to confer a high sensitivity to Zn²⁺ (Saxena & Macdonald, 1994). It is therefore possible that, in addition to the γ3 subunit, some receptors also contain either a γ2 or δ subunit. Third, if the receptors contain two β subunits, it cannot be excluded that β2 and β3 co-exist in DA neurones of the SNc, as has been suggested in neurones of the substantia nigra (pars reticulata) and lateral septum (Criswell et al. 1997) and as has been shown in the rat cerebral cortex (Li & De Blas, 1997). However, other studies have also shown that β2 and β3 subunits do not co-exist in the same receptor (Connolly et al. 1996).

Finally, even though it cannot be excluded that GABA_A receptors with different subunit compositions co-exist in the membrane of a single DA neurone, the most likely subunit composition appears to be α3/4β2/3γ3.

Incremental conductance levels

In DA neurones of the SNc, we found that the conductance levels of the GABA_A receptor channels varied in an incremental manner as integer multiples of 4-5 pS. In Purkinje neurones, which contain mainly α1β2/3γ2 subunits, the conductance levels varied identically but from an elementary value of 3 pS. In numerous studies, GABA_A receptor channels present a main conductance level of about 28 pS (see Macdonald & Olsen, 1994), but most of these studies have also reported the presence of multiple conductance levels: this was the case in other cerebellar neuronal preparations (Mistry & Hablitz, 1990; Smart, 1992) as well as in hippocampal neurones (Gage & Chung, 1994) and cortex interneurones (Xiang et al. 1998), that both express a large pattern of GABA_A receptor subunits (Wisden et al. 1992). Although the smallest conductance value measured depends on the neuronal type, the discrete distribution of conductance levels is thus a common property of many GABA_A receptor channels that is not correlated to a particular subunit composition but could involve intracellular receptor-associated proteins and/or post-translational mechanisms (for example phosphorylations). In addition, other ligand-gated receptor channels, such as glutamate (Cull-Candy & Usowicz, 1987; Jahr & Stevens, 1987), acetylcholine nicotinic (Mathie et al. 1991; Lewis et al. 1997) and glycine receptors (Bormann et al. 1993; Legendre, 1997) present multiple discrete openings.

The existence of multiple conductance levels has not yet received a definitive interpretation and can be explained in several ways. There may be variable conformational states of a single receptor channel protein, each dependent on the binding of an increasing number of agonist molecules and associated with a different conductance. However, in our study, whatever the isoguvacine concentration used, we could observe direct openings to the highest conductance levels occurring with openings to the smallest levels, which argues against a model of sequential openings giving rise to higher conductance levels (Bormann et al. 1993). Moreover, we found a high number of conductance levels (more than 5), even in patches where the open probability was extremely low, whereas the native GABA_A receptor is believed to contain only two α subunits which would carry the binding sites for the agonist (McKernan & Whiting, 1996). In accordance with this, the GABA_A receptor channel complex gating behaviour has been modelled with a reaction scheme incorporating two GABA binding sites and several open, closed and desensitized states (see Macdonald & Olsen, 1994). In each recording, the existence of direct transitions from one conductance level to all other conductance levels present in that recording could suggest that the multiple levels arose from a single receptor channel with different conductance states, which oddly follow an arithmetical progression.

An alternate hypothesis stipulating that several channels could open synchronously was proposed by Gage and his collaborators (Gage & Chung, 1994; Eghbali et al. 1997). This hypothesis would explain our results better: the smallest conductance value (level I) would correspond to the open state of one single elementary conducting pathway (possibly one GABA_A receptor channel) and the other levels would be due to the synchronized co-operative opening of a variable number of elementary conducting pathways, the highest conductance level being related to the number of co-operative GABA_A receptor channels. In agreement with this hypothesis, we found that the highest levels observed in the same neuronal preparation and recording conditions varied from one patch to another. This could be due to variations in patch geometry and/or channel density of the neurone membrane. Likewise, the aptitude of the various channels to form different size assemblies of co-operating receptors could explain the broad range of GABA_A receptor channel conductance levels (8-42 pS) that have been reported in other neurones (see Xiang et al. 1998), while the main conductance level could correspond to a preferential assembly of a certain number of receptors.

Pharmacological modulation in the distribution of conductance levels

Increasing the concentration of isoguvacine (1-10 μM), induced a decrease in the occupancy of level I concomitant with the increase in occupancy of the highest conductance levels. A similar effect was obtained by applying zolpidem, as one would expect if the binding of zolpidem at the benzodiazepine site enhances the affinity of the receptor for GABA (Macdonald & Olsen, 1994). Conversely, Zn²⁺ increased the occupancy of the smallest levels relative to the highest levels, even in the absence of any marked effect on the total occupancy.

Eghbali et al. (1997) found an apparent sequential increase in conductance levels of GABA_A currents induced by the benzodiazepine diazepam. By contrast, we found no change in the conductance of the various levels when applying zolpidem in the presence of isoguvacine. However, we observed with zolpidem (as well as when increasing the concentration of isoguvacine) a progressive change in the distribution of occupancy of the various levels from low to larger conductance levels, although we did not find an abrupt switch. The difference in the action of GABA in the presence of diazepam on the one hand (Eghbali et al. 1997), and isoguvacine in the presence of zolpidem on the other hand (our results), could result from a higher dissociation rate at their binding sites of zolpidem and/or isoguvacine compared with diazepam and/or GABA, respectively (Mistry & Hablitz, 1990).

Zn²⁺ ions have been described as inducing, at low concentrations, a negative allosteric modulatory effect on GABA_A openings (Harrison & Gibbons, 1994) through a binding site that would be located on the extracellular aspect of the receptor (Gingrich & Burkat, 1998). A similar modulation has been shown for picrotoxin on the glycine receptor (Legendre, 1997), where it decreased the occurrence of a 80 pS conductance level with no apparent effect on a 40 pS state. This has been interpreted by the author as a non-competitive effect of picrotoxin promoting a desensitization state linked to the open state, and implicating a conformational change in the receptor in response to the dissociation of the agonist from its binding site. If one assumes a single channel with multiple open states, the displacement in the equilibrium from the highest to the lowest conductance levels that we found can also be interpreted as an allosteric effect of Zn²⁺ ions. Alternatively, in the case of the other hypothesis of multiple channels opening synchronously, Zn²⁺could have a de-synchronizing effect by binding at an extracellular site and impairing channel interaction. Extracellular COOH⁻ and/or NH₂⁻ terminal domains could be involved in such interactions, as is the case for other channels (Varnum & Zagotta, 1997). Moreover, GABA_A receptor channels can be clustered (Whatley & Harris, 1996; Craig et al. 1996) and synchronization of opening could also implicate clustering of channels.

Functional considerations

DA neurones receive a strong tonic inhibition from GABAergic afferences (Haüsser & Yung, 1994) as also shown by a shift in the voltage clamp-recorded mean current induced by the application of 10 μM bicuculline (L. Cathala, personal communication). This tonic inhibition could be involved, not only in normal basal ganglia function, for example in controlling dopamine metabolism (Yamada et al. 1996), but also in preventing the excessive excitation which could lead to seizure spread and neurotoxicity (Haüsser & Yung, 1994). What could be the functional relevance of the particular behaviour of the GABA_A receptor channels we describe here? It is hard to speculate what is happening at the level of the synapse, since much higher concentrations of GABA than those used in this study are released transiently. The study of the effects of high concentrations of agonist is impossible at steady state because of the high desensitization rate of the channels. Rapid applications (< 5 ms) will have to be used to determine if such behaviour plays a role at the synapse. However, it is likely that GABA spillover can tonically stimulate extrasynaptic GABA_A receptor channels at micromolar concentrations (Rossi & Hamann, 1998). In these conditions, openings in large conductance states of the receptors might slow down the decay phase of the IPSCs and increase the total charge transferred through the membrane of the DA neurone. Indeed, it has been shown that zolpidem and diazepam, which are also thought to favour the highest conductance levels (Eghbali et al. 1997), are able to slow down the decay of the IPSCs as well as the current produced by applications of GABA (Mereu et al. 1992; Rossi & Hamann, 1998). In addition, it is possible that Zn²⁺, which has been shown to be present in the synaptic terminals (see Harrison & Gibbons, 1994), could modulate the amplitude and the time course of the IPSCs and contribute to a form of plasticity of the GABA response. Finally, low doses of GABA could be sufficient to produce transient hyperpolarizations of significant enough amplitude to activate several voltage-gated currents which are de-inactivated by hyperpolarization (like transient outward K⁺ (A-)current (I_A) or low threshold Ca²⁺ current (I_T)) and could be involved in the triggering and/or the maintenance of the membrane potential oscillations of the DA neurones.