Basal ganglia are interconnected subcortical nuclei, connected to the thalamus and all cortical areas involved in sensory motor control, limbic functions and cognition. The striatal output neurones (SONs), the major striatal population, are believed to act as detectors and integrators of distributed patterns of cerebral cortex inputs. Despite the key role of SONs in cortico-striatal information processing, little is known about their local interactions. Here, we report the existence and characterization of electrical and GABAergic transmission between SONs in rat brain slices. Tracer coupling (biocytin) incidence was high during the first two postnatal weeks and then decreased (postnatal days (P) 5–25, 60%; P25–30, 29%; n= 61). Electrical coupling was observed between 27% of SON pairs (coupling coefficient: 3.1 ± 0.3%, n= 89 at P15) and as shown by single-cell RT-PCR, several connexin (Cx) mRNAs were found to be expressed (Cx31.1, Cx32, Cx36 and Cx47). GABAergic synaptic transmission (abolished by bicuculline, a GABAA receptor antagonist) observed in 19% of SON pairs (n= 62) was reliable (mean failure rate of 6 ± 3%), precise (variation coefficient of latency, 0.06), strong (IPSC amplitudes of 38 ± 12 pA) and unidirectional. Interestingly, electrical and chemical transmission were mutually exclusive. These results suggest that preferential networks of electrically and chemically connected SONs, might be involved in the channelling of cortico-basal ganglia information processing.
The basal ganglia are mainly involved in the control of movement and cognitive functions (Graybiel, 1995). The striatum, the main input nucleus of the basal ganglia, acts functionally as a detector of distributed patterns of cortical inputs (Wickens, 1993; Graybiel et al. 1994; Wilson, 1995). In fact, the striatum receives convergent inputs from the entire cerebral cortex. Cortico-striatal information appears to be processed through parallel functional circuits (Alexander et al. 1986; Middleton & Strick, 2000). These parallel pathways seem to be segregated all along neuronal loops connecting the cerebral cortex, the basal ganglia and the thalamus. However, some transverse interactions exist between these loops.
SONs represent the main neuronal subpopulation (95%) of the striatum. These GABAergic neurones are characterized by a very hyperpolarized resting membrane potential during weak and non-correlated cortical inputs (Calabresi et al. 1987; Nisenbaum & Wilson, 1995; Mahon et al. 2001). To fire, SONs need to receive at the same time a convergence of strong cortical activity (Kita, 1996; Wilson & Kawaguchi, 1996). SONs represent a key step in the highly complex process of cortico-striatal signal integration, and more broadly in motor control and habit formation (Graybiel, 1995). SON discharges induce an inhibition of basal ganglia output structures (the substantia nigra pars reticulata and the entopedoncular nucleus) which in turn leads to disinhibition of basal ganglia target cells within some thalamic and brain stem nuclei (Alexander et al. 1986; Alexander & Crutcher, 1990). In spite of their key roles, little is known about the local interactions (electrically and chemically mediated synapses) of SONs, i.e. the organization and modes of communication between these cells.
In the present study performed on rat brain slices, we report the existence and characterize GABAergic and gap junction-mediated transmission between SON pairs. In contrast with previous studies, GABAergic transmission appears to be reliable, strong, fast and strictly unidirectional. In addition, we demonstrate that none of the chemically connected SON pairs were electrically coupled. Conversely, electrically coupled SONs did not appear to interact through GABAergic synapses. Such peculiar features of exclusive communication could participate in the channelling of cortico-basal ganglia information processing.
Whole-cell recordings in oblique horizontal (Kawaguchi et al. 1989) or sagittal brain slices from Wistar rats (postnatal days (P) 5–35) were performed as described (Venance & Glowinski, 2003). Animals were killed by decapitation and brains were immediately removed. All experiments were performed in accordance with local ethical committee and EU guidelines (directive 86/609/EEC). Patch-clamp whole-cell recordings were made using borosilicate glass pipettes of 5–7 MΩ resistance containing (mm): 105 potassium gluconate, 30 KCl, 10 Hepes, 10 phosphocreatine, 4 ATP-Mg, 0.3 GTP-Na, 0.3 EGTA (adjusted to pH 7.35 with KOH). A similar intracellular solution, but with potassium gluconate (125 mm) and KCl (10 mm), was used to determine the voltage dependence of the postsynaptic current. The composition of the extracellular solution was (mm): 125 NaCl, 2.5 KCl, 25 glucose, 25 NaHCO3, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2, plus 10 μm pyruvic acid and 10 μm 6,7-dinitroquinoxaline-2,3-dione (DNQX), bubbled with 95% O2 and 5% CO2. All whole-cell recordings were performed at 34°C using a temperature control system (Bioptechs ΔTC3, Butler, PA, USA). Signals were amplified using an EPC9-2 amplifier (HEKA Elektronik, Lambrecht, Germany). Current-clamp recordings were filtered at 2.5 kHz and sampled at 5 kHz; voltage-clamp recordings were filtered at 5 kHz and sampled at 10 kHz using the program Pulse-8.50 (HEKA Elektronik). Off-line analysis was performed using Igor (Wavemetrics, Lake Oswego, OR, USA). Series resistance compensation was set to 75–90% in whole-cell configuration (13.7 ± 0.9 MΩ, n= 145). Liquid junction potential error (−13.6 mV) was corrected. Neuronal recordings were performed equally in a putative striosomal-enriched area (antero-median area) and in the matrix (lateral posterior area). The distance between two recorded neurones never exceeded 50 μm.
All chemicals (bicuculline methiodide, carbenoxolone, halothane) were purchased from Sigma (St-Quentin, France), except DNQX (Tocris).
All results were expressed as mean ±s.e.m. and statistical significance was assessed using Student's t test when appropriate or the non-parametric Mann-Whitney U test at the significance level (P) indicated. When specified, χ2 or Fisher's exact tests were used. Neuronal input resistance (Rinput) and coupling coefficients (CC) were estimated from the membrane potential responses to the injection of current steps into either neurone. CCs were determined from the ratio of presynaptic to postsynaptic membrane potential changes. The junctional conductance (Gj) was estimated from Rinput and CC (Bennett, 1977): Gj1−2=Rinput1CC1−2/((Rinput1Rinput2) − (Rinput1CC1−2)2), where Rinput1 and Rinput2 are the Rinput values of the injected and receiving neurones, respectively, and CC1−2 the CC between the injected and receiving neurones. The K ratio is the ratio of the higher to the lower CC of an electrically coupled SON pair (K ratio = CC1−2/CC2−1 with CC1−2 > CC2−1).
Biocytin filling and histochemistry
Biocytin (Sigma), 5 mg ml−1, was dissolved into the pipette solution and cells were filled during at least 45 min of recording. Subsequently, slices were fixed overnight in 2% paraformaldehyde at 4°C. Biocytin-filled cells were visualized using the avidin–horseradish peroxidase (HRP) reaction (ABC Elite peroxidase kit; Vector Laboratories, Burlingame, CA, USA) according to the manufacturer's instructions.
Single-cell RT-PCR was performed as previously described (Monyer & Jonas, 1995; Venance et al. 2000, 2001). Patch-clamp pipettes used for RT-PCR experiments (2–2.5 MΩ) were filled with autoclaved internal solution containing (mm): 140 KCl, 3 MgCl2, 5 EGTA and 5 Hepes (pH 7.35 adjusted with KOH). Extensive electrophysiological analysis was not carried out on these cells since ATP, GTP and phosphocreatine could not be used due to the required autoclave treatment of the intracellular solution. After recordings, the neuronal cytoplasm was aspirated by gentle suction into the patch pipette under visual and electrophysiological controls. The harvested material was subsequently expelled into an autoclaved tube containing deoxyribonucleoside triphosphate, dithiothreitol, ribonuclease inhibitor and Superscript reverse transcriptase II (Gibco/BRL, Cergy Pontoise, France), and incubated for 1 h at 37°C. Two rounds of PCR amplification were performed, with 4 μl of the first PCR as a template for the second PCR. PCR conditions were the same for both amplification rounds. After a hot start at 94°C for 5 min, 30 cycles (94°C for 30 s; 53°C for 30 s; 72°C for 40 s) and a final elongation step at 72°C for 10 min were performed. As previously described, in each experiment contamination artifacts were excluded both for contamination of the PCR and the inadvertent harvesting from surrounding material in the slice preparation (Monyer & Jonas, 1995; Venance, 2001). All primers were intron-spanning (except those for Cx33, Cx50 and Cx57). Two controls were systematically performed for each experiment: slice controls and an intracellular solution control. For the former, a patch pipette was lowered into the slice close to a neurone for 1–2 min but without patching it. For the latter, 8 μl of internal solution was used. In both cases, the material was processed similarly to the harvested cell content. The analysis entailed mixed nested PCRs for μ-opioid receptors (MOR), met-enkephalin (ENK), substance P (SP) and one Cx. For MOR, PCR primers sat on exons 1 (5′-primer (P5′): 5′-CACAGCCATTACCATCATGG-3′) and 2 (P3′: 5′-AGACAGCAATGTAGCGGTCC-3′ and 3′-nested-primer (P3′n): 5′-ATCTTGCAGAGGATGGTTCC-3′). For SP, PCR primers were placed on exons 3 (P5′: 5′-GCATCTTCTTCAGAGAATCGC-3′) and 7 (P3′: 5′-TAAAAGCAACCAAGGGAAGC-3′, and P3′n: 5′-TCGCTGGCAAACTTGTACAAC-3′). For ENK, PCR primers were placed to span the first exon (P5′: 5′-ACAGGATGAGAGCCACTTGC-3′, P3′: 5′-CTTCATCCGAGGGTAGAGAC-3′, and P3′n: 5′-TGTTGGTGGCTATCTTTCGC-3′). For specific amplification of Cxs the following primers were used: Cx26 (P5′: 5′-CGGACCTGCTCCTTACAGG-3′, P3′: 5′-CATGATCAGCTGCAGAGCC-3′, and P3′n: 5′-CGTAGCACACATTCTTACAGCC-3′), Cx30 (P5′: 5′-TTCCAGTTCACCTCACACGG-3′, P3′: 5′-ACCACGAGGATCATGACTCG-3′, and P5′n: 5′-TGACTGCCAGAGGAGTAGAAGG-3′), Cx30.3 (P5′: 5′-TTCTTGCCTTCATCCAGTCC-3′, P3′: 5′-TTCATGTTGGCTGTGTGTCC-3′, and P5′n: 5′-CTCTGAATCACTGCGTATGAGG-3′), Cx31 (P5′: 5′-GATGCCTCCTTAATGAGTAGGG-3′, P3′: 5′-TGGAGTACTGGTTCACACCG-3′, and P3′n: 5′-CCTGAAGCTTCTTCCAATCC-3′), Cx31.1 (P5′: 5′-TCTGATGCTTGCTGAACCC-3′, P3′: 5′-AAGTCCTTCTGGTCGTCACC-3′, and P3′n: 5′-GCACACGGAAGACAAAGACC-3′), Cx32 (P5′: 5′-CACAGACATGAGACCATAGG-3′, P3′: 5′-AACCAAGATGAGTTGCAGG-3′, and P3′n, 5′-TAGCAGACGCTGTTACAGCC-3′) Cx33 (P5′: 5′-GAGGCAGATTGCTGCTAACC-3′, P3′: 5′-ACACCACCAACATGAAGAGG-3′, and P5′n: 5′-ATGGATTCACTCTGAGTGCC-3′), Cx36 (P5′: 5′-AATGGACCATCTTGGAGAGG-3′, P3′: 5′- GATCTGGAAGACCCAGTAACG-3′, and P3′n: 5′-CTGCTCATCATCGTACACCG-3′), Cx37 (P5′: 5′-TAGGAGGAGCTGAGAAAGGC-3′, P3′: 5′-GAAATCAGACTGCTCGTCGC-3′, and P5′n: 5′-TGAGAAAGGCATTGTGCCC-3′), Cx40 (P5′: 5′-CTGGACAGTTGAACAGCAGC-3′, P3′: 5′- TATCACACCGGAAATCAGCC-3′, and P3′n: 5′-AATGAACAGGACGGTGAGC-3′), Cx43 (P5′: 5′-TCCTTTGACTTCAGCCTCC-3′, P3′: 5′-GACTGTTCATCACCCCAAGC-3′, and P3′n: 5′-TATGAAGAGCACTGACAGCC-3′), Cx45 (P5′: 5′-CTCTAAACCACTGCACCAGC-3′, P3′: 5′-AGATGGACTCTCCTCCTACAGC-3′, and P3′n: 5′-TCGAATGGTTGTGGATCTCC-3′), Cx47 (P5′: 5′-TCCCATGACCAACATGAGC-3′, P3′: 5′-AGCAGACGTTGTCACAACCC-3′, and P3′n: 5′-GAATAGATGGACTCACCACCG-3′), Cx50 (P5′: 5′-TCTTGGAAGAGGTGAATGAGC-3′, P3′: 5′-TGGAGACGAAGATGATCTGC-3′, and P5′n: 5′-CAGTGCTCTTCATCTTCCGC-3′), Cx57 (P5′: 5′-CATCCTAGAGGAAGTCCACTCC-3′, P3′: 5′-CAGATATTGTTGCAACCGG-3′, and P5′n: 5′-GAAGATCTGGCTGACCATCC-3′). The molecular identity of the PCR amplicons was systematically confirmed by direct sequencing (Genset, Montreuil, France).
To test the presence of electrical and chemical synaptic local interactions between SONs, single whole-cell recordings and double recordings were made on 380 single SONs and 131 pairs of SONs, respectively. All recorded SONs were identified by their typical electrophysiological characteristics (Fig. 1A) (Wilson & Groves, 1980; Kawaguchi et al. 1989). Based on their spiking pattern and basic membrane properties, SONs were easily distinguished from the two other neuronal subtypes of the striatum: the long lasting after-hyperpolarization (LA) and fast-spiking (FS) interneurones. Briefly, SONs showed a very hyperpolarized resting membrane potential (SONs: −87.1 ± 0.6 mV, n= 201; LA interneurones: −62.7 ± 3.1 mV, n= 24; FS interneurones: −77 ± 1.3 mV, n= 9), a fast inward rectification (instead of a linear current–voltage (I–V) relationship for LA and FS cells), a long depolarizing ramp to spike threshold, a long delay to first spike (SONs: 375.3 ± 25 ms; LA: 210 ± 45 ms; FS: 205 ± 94 ms), a relatively slow membrane time constant (SONs: 33.7 ± 1.0 ms; LA: 24.3 ± 2.1 ms; FS: 55.1 ± 2.7 ms) and, finally, a medium range discharge frequency in response to depolarizing current pulses (SONs: 23.7 ± 1.3 Hz for +50 pA injected current above action potential (AP) threshold; LA: 10.0 ± 0.5 Hz; FS: 63.2 ± 3.4 Hz).
Tracer coupling between SONs
To assess the occurrence of gap junctional communication, 61 SONs were filled with biocytin during early (P5–10) and juvenile (P10–30) development (Fig. 1B and C). The incidence of tracer coupling was relatively high until P15 and then significantly decreased (5 ≤P < 10: 67%, n= 15; 10 ≤P < 15: 64%, n= 21; 15 ≤P < 25: 52%, n= 14; 25 ≤P < 30: 29%, n= 11; r=−0.97, P < 0.05). The number of coupled neurones to which tracer spread showed a non-significant decrease between P5 and P15 and then reached a plateau between P15 and P30 (5 ≤P < 10: 3.5 ± 2.4; 10 ≤P < 15: 2.3 ± 1.0; 15 ≤P < 25: 1.9 ± 1.2; 25 ≤P < 30: 2.0 ± 1.7) (Fig. 1C). No close apposition was observed between the soma or primary dendrites of coupled cells suggesting that tracer coupling sites were located on secondary dendrites (spiny dendrites). Gap junctional communication was blocked after 5 min pretreatment of the slices before whole-cell recordings with either halothane (2 mm, n= 10) or carbenoxolone (150 μm, n= 9) since 10% and 0% of tracer coupling incidence, respectively, were observed under treatment with these uncoupling agents in P10–12 rats.
Electrical coupling between SON pairs
Using double patch-clamp recordings, 24% of SON pairs (P15–18 animals) were found to be electrically coupled (n= 91) (Fig. 2). These coupled SON pairs displayed a mean coupling coefficient (CC) of 2.9 ± 0.2% (range 1.3–6.4%) and a junctional conductance (Gj) of 102.2 ± 8.4 pS (range 29.9–232.6 pS). The relationship between the amplitude of the pre- and postsynaptic membrane changes was linear (Fig. 2Ac and Bc), denoting a lack of rectification in electrical coupling. While a large proportion of these coupled pairs (68%) displayed symmetrical coupling (Fig. 2A, C and D), partial (0.001 < P < 0.05) or complete asymmetrical coupling was observed in 9% and 23% of the remaining pairs, respectively (Fig. 2B, C and D). However, as shown by the lack of significant difference between CC values of asymmetrical (3.3 ± 0.5%, n= 7) and symmetrical (2.9 ± 0.2%, n= 15) SON pairs, this asymmetry (partial and complete) was not related to the CC value.
According to several observations, this asymmetrical coupling could not be attributed to a difference in Rinput between these coupled cells (Fig. 2Ab, Bb and E). First, as calculated from voltage-clamp recordings, Gj values displayed a distribution similar to those of CCs (Fig. 2C and D). Second, no relationship was found between the ratio of CC values (K ratio) and the ratio of Rinput values of coupled neurones (Fig. 2E). Finally, no significant difference was found between asymmetrical (n= 5) and symmetrical (n= 17) SON pairs for either Rinput values (360 ± 38 MΩversus 333 ± 18 MΩ with P= 0.51, t test) or Rinput ratios (1.59 ± 0.40 versus 1.24 ± 0.07 with P= 0.65, Mann-Whitney U test). This indicates that a difference in non-junctional conductances does not account for the incidence of asymmetrical electrical coupling. In addition, since the mean value of input conductances (3333 ± 182 pS, n= 44) was much larger than the junctional conductance values, the asymmetry of the junctional conductances does not affect input conductances.
Connexin expression analysis by single-cell RT-PCR
Firstly, primers for each Cx were tested and optimized for the two PCR steps on either P17 rat brain or P20 rat lens cDNAs (indeed, Cx46 and Cx50 are expressed massively in the lens). Surprisingly, PCR amplicons of most tested Cxs were found in total cDNAs from the rat brain (except Cx30.3 and Cx33) (Table 1).
Table 1. Connexin expression in total rat brain cDNA and in single-cell electrophysiologically identified SONs
Total rat brain cDNA
Percentage of Cx amplicons
Investigations were performed first on total rat brain cDNA (the presence (+) or absence (−) are listed (n= 2–5)), and then on single SONs. a Connexin (Cx) analysis performed from P20 rat lens. b Since the primers used for Cx50 and Cx57 amplifications are not intron-spanning, it is impossible to differentiate a genomic from a mRNA amplification. n.d.: not determined, because of the absence of amplicons after two rounds of PCR on total rat brain cDNAs.
Secondly, to determine the expression of Cx mRNAs in SONs, single-cell RT-PCR was performed using individual harvested cytoplasms from 319 SONs identified by their firing pattern (Table 1). Due to the very low expression of neuronal Cx mRNAs (Venance et al. 2000) and for optimum conditions for PCR amplification (which differ among pairs of primers), RT-PCR was separately performed for each Cx type. cDNAs coding for four different Cxs were detected: Cx31.1, 32, 36 and 47.
Interestingly, all Cx32 amplicons showed a particular cDNA structure since an additional intronic part of 336 nucleotides (in the 3′ side of the intron) was found (see figure available online as Supplementary Material). Direct sequencing of these Cx32 amplicons indicated that this additional intronic part was located upstream of the start codon. Therefore, it should not affect the mRNA coding sequence but could contain some regulatory sequences. Since oligodendrocytes are known to express Cx32 (Rash et al. 2001), similar experiments were performed on oligodendrocytes from the corpus callosum (14 cells) that were identified by their tail currents. In contrast to observations made in SONs, the ‘classical’ Cx32 form was exclusively found (detection rate of 57%) in these oligodendrocytes (details of the amplified sequences in Supplementary Material).
In addition to the functional territories (limbic to motor), two main compartments (the striosomes and the matrix) are distinguished in the striatum (Graybiel, 1990; Gerfen, 1992). SON subpopulations are identified by the differential expression of three markers: the enkephalin (ENK), substance P (SP) and μ-opioid receptor (MOR). Indeed, SP is expressed in matrix SONs projecting mainly to the substantia nigra pars reticulata and the internal globus pallidus (the so-called direct pathway). ENK is present in matrix SONs projecting mainly to the globus pallidus (the indirect pathway). MOR is present in striosomal but not in matrix SONs. We have investigated if Cxs were expressed preferentially in some SON subpopulations. For this purpose, coexpressions of MOR, SP and ENK, together with Cxs were analysed in 191 SONs (Table 2). SONs expressing Cxs were preferentially located in the matrix (MOR negative, 24%, n= 59) rather than in striosomes (MOR positive, 13%, n= 40). In addition, in the matrix, SONs expressing Cxs appeared to belong mainly to the indirect pathway (ENK positive, 26%, n= 47) rather than the direct pathway (SP positive, 0%, n= 7) (Table 2).
Table 2. Single-cell RT-PCR analysis of the marker expression (met-enkephalin, substance P and μ-opioid receptor cDNAs) among SONs and among connexin-expressing SONs
SONs expressing the different marker patterns (%) (n= 191)
SONs coexpressing a Cx and the different marker patterns (%) (n= 191)
The molecular identity of the amplified Cxs (Cx31.1, 32, 36 or 47) was not taken into account in this analysis in order to have reasonable samples of SONs. Among these SONs, 14% were positive for a Cx (Cx31.1, 32, 36 or 47), 21% for μ-opioid receptor (MOR), 13% for substance P (SP) and 38% for met-enkephalin (ENK) cDNAs.
Frequency dependence of electrical synaptic transmission
Neuronal electrical synapses act as low-pass filters (Bennett, 1977; Galarreta & Hestrin, 2001; Bennett & Zukin, 2004), therefore to investigate this characteristic in SONs, subthreshold sinusoidal current stimuli were applied to electrically coupled pairs (n= 5) within a frequency range of 1–50 Hz (Fig. 3). Electrical coupling was rapidly reduced with increasing frequencies. Most cells displayed spontaneous irregular oscillations in the 1–2 Hz range which could interfere with the electrical coupling observed at 2 Hz. Indeed, as estimated in three representative cells, these oscillations reached 0.8 ± 0.5 mV and 4 ± 3 pA, in current or voltage clamp at resting membrane potential (see the very large s.e.m. for this value in Fig. 3). Electrical synapses of SONs display properties of low-pass filtering as all mammalian neuronal electrical synapses already described (Galarreta & Hestrin, 2001).
Electrical coupling, via spikelets, can modulate the shape of postsynaptic potentials
Due to their low-pass filter properties, electrical synapses only partially transmit an evoked presynaptic AP since slow depolarization and after-hyperpolarization AP phases are more efficiently transmitted than the spike itself. Such evoked spikelets occurred in 4 out of 22 coupled SON pairs (Fig. 4A and B). The mean CC for AP (ratio of presynaptic spike to spikelet amplitude) in these pairs was 0.90 ± 0.10% (range 0.32–1.76%; n= 27 spikelets). Spikelet latency and mean decay time (τ) were 1.15 ± 0.12 ms (n= 21) and 19.3 ± 6.4 ms (n= 11), respectively. Smaller spikelets with a mean amplitude of 0.30 ± 0.05 mV (n= 25) were also observed (in 7 out of 22 coupled SON pairs) (Fig. 4B). The mean CC for AP was only 0.43 ± 0.03% (n= 25) for these spikelets. This CC value was much higher (2.7 ± 0.9%; n= 25) when calculated from the maximal amplitude of the initial depolarization phase preceding the spike instead of the peak amplitude.
Postsynaptic electrotonic events interact with chemical synaptic signalling. Indeed, the decay of a spontaneous PSP evoked by a third cell (Fig. 4C and D) could be modified by a spikelet evoked from a presynaptic AP. In all cases (n= 5), the shape of the slowest part of the mono-exponential decay of the PSP was altered suggesting that the spikelet was more effective on the slower than on the faster part of the PSP decay. Such a spikelet-induced modification of the PSP decay should amplify the PSP summation when another PSP hits the postsynaptic cell and therefore facilitate the response of the target cell.
IPSPs and IPSCs at SON–SON synapses
Inhibitory synaptic transmission between SON pairs was also examined (n= 72, P15–18 animals). In 17% of the recorded SON pairs, APs evoked in one cell triggered a depolarizing IPSP or inward IPSC in the other cell held at −80 mV (Fig. 5). Such IPSPs and IPSCs at SON–SON synapses are illustrated in Fig. 5 and their kinetic properties are detailed in Table 3. The mean IPSC peak amplitude was 29.6 ± 7.2 pA (n= 12) when the potential of the postsynaptic cell was held at −80 mV. Most IPSP and IPSC decays were fitted with a mono-exponential function, with mean time constants of 27.1 ± 4.8 and 9.2 ± 0.7 ms, respectively. However, in two SON pairs, some IPSPs (5/143 and 3/64) and IPSCs (8/135 and 19/164) were better fitted with a bi-exponential function (Table 3).
Table 3. Properties of unitary IPSPs and IPSCs generated at the SON–SON synapse (n= 12 pairs)
Values are mean ±s.e.m. (range). a Range of raw data (n= 8 for IPSPs and 27 for IPSCs); elsewhere values were determined from averaged IPSPs and IPSCs.
0.84 ± 0.05 (0.58–1.1)
0.58 ± 0.03 (0.43–0.81)
10–90% rise time (ms)
2.77 ± 0.40 (1.34–4.58)
1.03 ± 0.10 (0.60–1.69)
Peak amplitude (pA; mV)
2.99 ± 0.91 (0.57–7.2)
29.6 ± 7.2 (8.18–95.9)
Decay τ (ms)
27.1 ± 4.8 (10.0–48.4)
9.23 ± 0.67 (6–12.32)
Decay τ1 (ms)a
16.7 ± 0.8 (14.1–18)
8.7 ± 0.7 (5.3–14.3)
Decay τ2 (ms)a
64.9 ± 5.8 (55–78.1)
14.2 ± 0.1 (8.8–23.8)
39.6 ± 6.7 (17.3–72.8)
19.9 ± 2.3 (12.9–35.4)
19.9 ± 3.2 (8.4–36.3)
7.0 ± 0.6 (4.3–9.9)
12.4 ± 7.0 (0–55.2)
25.1 ± 7.8 (0–67.4)
The occurrence of evoked IPSCs among SON pairs was much higher in sagittal (n= 7/22) than in horizontal (n= 5/50) brain slices. In addition, the mean unitary IPSC (uIPSC) amplitude was much higher and the failure rate lower when SONs were recorded from sagittal (37.9 ± 11.9 pA and 6.2 ± 2.5%, respectively) compared with horizontal slices (17.9 ± 5.7 pA and 51.7 ± 7.1%, respectively). Due to the antero-posterior orientation of SONs (Walker & Graybiel, 1993), these results could be explained by a better preservation of the dendritic arborization of these cells in sagittal slices. According to the reduced failure of chemical transmission between SON pairs recorded from sagittal slices (6.2%, range 16–0%) (see also Fig. 5C), this chemical synaptic transmission appears to be very reliable. Indeed, in response to single or double presynaptic APs, postsynaptic events were detected without any failure in two SON pairs (n of stimulations = 578 and 623) and with only 4 failures out of 783 stimulations in another pair of cells (Fig. 6E).
To investigate the voltage dependency of the IPSC, a 10 mm KCl intracellular solution (see Methods) was used to obtain a theoretical equilibrium potential for Cl− ions of −65.3 mV instead of −37.5 mV with the standard 30 mm KCl intracellular solution. In this condition, the current–voltage relationship reversed close to −66 mV (Fig. 5D), indicating that chloride conductances are exclusively responsible for these IPSCs. Finally, demonstrating that evoked IPSCs in SON pairs involved GABAA receptors (Fig. 5E), the bath application of the GABAA antagonist bicuculline (20 μm, n= 4) totally abolished IPSCs. Partial recovery of the evoked IPSC was observed after bicuculline washing (72 ± 7%, n= 4).
IPSPs and IPSCs time courses
The SON-mediated GABAergic inhibition was of short latency, rapid onset and relatively short duration and half-width (Table 3). These features were even more pronounced when SON pairs were recorded from sagittal slices.
As illustrated by typical results obtained from a SON pair (Fig. 6A and D), unitary IPSCs (uIPSCs) evoked by a presynaptic AP exhibited a low variability. This was further indicated by the low variation coefficients (CVs) of the IPSC amplitude (0.40), latency (0.37), 10–90% rise (0.40) and decay (0.43) time constants. Further analysis of the IPSC amplitudes (Fig. 6A) revealed a unimodal distribution of events centred on 15 pA occurring well above the background noise (0.87 ± 0.05 pA, 161 measures).
The relationships between the IPSC amplitudes and the percentage of failures and CV are illustrated in Fig. 6E and F. The reliability of SON–SON transmission increased for higher IPSC amplitudes since the percentage of failure significantly decreased as a function of the IPSC mean peak amplitude (r=−0.59, P < 0.05) (Fig. 6E). In fact, the percentage of failures was relatively high (up to 67%) when unitary IPSCs were small (8.1 ± 0.9 pA), but sharply decreased for larger IPSC amplitudes to finally reach a ‘null failure’ plateau. Similarly, the CV was inversely related to the amplitudes of IPSC (r=−0.84, P < 0.001, Fig. 6F). In simple binomial models of synaptic transmission these relationships are to be expected when IPSC amplitude is primarily determined by the release probability. Latency, rise time, decay, half-width and duration of IPSCs were also plotted as a function of IPSC peak amplitudes (Fig. 6G and H). A positive correlation was observed between the IPSC rise time and duration with the mean IPSC amplitudes (r= 0.66, P < 0.05, and r= 0.91, P < 0.0001, respectively), but this was not the case when the IPSC latency, half-width or decay were considered. Although the IPSC latency appeared to be relatively stable whatever the IPSC amplitude, the s.d. for the mean latency (0.29 ± 0.32 ms, n= 12) was larger than expected for electrotonically equidistant sites (Feldmeyer & Sakmann, 2000). Altogether, large IPSC amplitudes were associated with slow rise times and long duration while small IPSC amplitudes were associated with faster kinetics for the rise time and duration.
Short-term plasticity at SON–SON chemical synapses
We investigated the frequency dependence of synaptic transmission between SON pairs. IPSP and IPSC amplitude ‘ratios’ (i.e. ratio of the second event relative to the first) were first plotted as a function of the interspike interval (ISI) (Fig. 7A and B). Second IPSP (IPSP2) and IPSC (IPSC2) amplitudes were measured by either their absolute values (A2, amplitude from the baseline preceding IPSP1 to the IPSP2 peak or from that of IPSC1 to the IPSC2 peak) or their relative values (ΔA2, from the initiation point to the peak of IPSP2 or IPSC2). A2 is similar to ΔA2 when the ISI is longer than the decay times of IPSP1 or IPSC1. For both the IPSP and IPSC, A2 represents the total number of opening channels during the second spike while ΔA2 provides an estimation of the net increase in the number of opening channels activated by the second spike.
A marked depression of evoked IPSPs and IPSCs was observed for most ISIs and as illustrated in Fig. 7A and B, the depressions for IPSPs and IPSCs were −38 ± 9% and −34 ± 9%, respectively, for a 100 ms ISI (n= 4 SON pairs). Such a depression persisted for 1 s ISIs and was never observed for ISIs longer than 3 s. Similar results were obtained for IPSCs with the two estimation procedures (A2/A1 or ΔA2/A1), while this was not the case for IPSPs (Fig. 7A and B). This discrepancy resulted from differences in IPSC (9.2 ± 0.7 ms) and IPSP (27.1 ± 4.8 ms) decay kinetics. IPSP summations were thus observed for ISIs shorter than 30 ms corresponding to a presynaptic firing of 40 Hz, a SON frequency range recorded in vivo (Wilson & Kawaguchi, 1996; Stern et al. 1998). In contrast, such summations were not observed for IPSCs since this would require ISIs shorter than 10 ms, corresponding to a 100 Hz spike frequency hardly reached in SONs.
To further characterize the reduction of synaptic efficacy, kinetic parameters (amplitude, latency, rise time, half-width, duration and decay time constant) of IPSP2s and IPSC2s were then compared to those of IPSP1s and IPSC1s using 50, 100 and 150 ms ISIs (n= 4 SON pairs) (Fig. 7C and D). For both IPSPs and IPSCs, besides the decrease in amplitude reflecting the depression itself, longer latency and shorter duration were observed. This longer latency could be attributed to a reduced amount of transmitter release from the presynaptic element while the shorter duration of IPSP2s or IPSC2s probably results from their reduced amplitude.
Electrical and unidirectional chemical couplings are mutually exclusive
As illustrated in Fig. 8A, the inhibitory transmission at SON–SON synapses appeared to be strictly unidirectional. Indeed, when SON recordings were made either on sagittal or horizontal slices, IPSPs and IPSCs in all synaptically connected SON pairs (n= 12) were only observed in one direction.
Finally, the combined occurrence of electrical coupling and GABAergic transmission was investigated in 72 SON pairs (Fig. 8B). Surprisingly, in both sagittal and horizontal slices none of the chemically connected SON pairs (17%) were electrically coupled and reciprocally, none of the electrically connected SON pairs (25%) were found to be chemically connected.
The understanding of information processing along cortico-basal ganglia circuits requires precise knowledge of local interactions in the striatum, the main input relay nucleus of basal ganglia. The demonstrations in the present study, that electrical coupling and unidirectional GABAergic transmission occur between SONs and that these two communication modalities are mutually exclusive, provide an additional level of complexity for intrastriatal circuits and cortico-basal ganglia information processing.
Tracer coupling between SONs
Gap junction-mediated pathways could interconnect SONs belonging to either the same or different cortico-basal ganglia channels. SONs display some heterogeneity of membrane receptors linked to second messenger pathways (μ-opioid, D1 and D2 dopaminergic receptors, for example; Gerfen, 1992; Nicola et al. 2000). As already shown for calcium signalling in striatal astrocytes such ‘pharmacological’ heterogeneity of SONs could be attenuated or unmasked by the opening or closing of gap junction channels, respectively (Venance et al. 1998). Indeed, in addition to electrical coupling, gap junctions allow metabolic coupling by intercellular diffusion of energetic metabolites (glucose and fructose), second messengers (IP3 and cAMP) and amino acids (glutamate and glycine) (Harris, 2001). Therefore, through intercellular calcium signalling or metabolic homeostasis, this metabolic coupling could participate in intercellular neuromodulation.
Electrical coupling between SONs
CCs between SON pairs depend not only on the number and elementary properties of gap junction channels but also on non-junctional dendritic conductances varying with the distance between the coupling sites and the soma. A large proportion of SONs exhibit symmetrical coupling. The causes of the electrical coupling asymmetry observed in a minor proportion of cells remain to be solved.
According to our single-cell RT-PCR experiments, several mRNA Cxs, including Cx31.1, 32, 36 and 47 are expressed in SONs. This profile differs from the one determined with this technique in cortical and hippocampal interneurones in which Cx36 was solely expressed (Venance et al. 2000). The diversity of Cx expression in SONs allows the formation of heteromeric and/or heterotypic channels and thus accounts for the asymmetry of electrical coupling observed in a subpopulation of SONs. Electrical coupling has been recorded between striatal GABAergic interneurones (Koòs & Tepper, 1999), but tracer coupling is restricted to SONs (Cepeda et al. 1989; Onn & Grace, 1994; this study) suggesting that interneurones and projecting neurones correspond to two independent electrically coupled networks in the striatum. Such an independence could be attributed (besides the fact that neuronal subpopulations could interact specifically with their own kind) to the difference in the types of Cxs present in these two cell populations since each Cx exhibits specific biophysical and regulatory properties as well as a specific profile of compatibility with other Cxs (Harris, 2001; Bennett & Zukin, 2004).
SON populations differ by some specific markers, such as membrane receptors or peptides and their structure(s) of projection. Single-cell RT-PCR for Cxs and specific markers (ENK, SP and MOR) of SON subpopulations were performed. Cxs appeared to be expressed in larger amounts in SONs from the matrix compartment belonging to the indirect output pathway. Indeed, these cells, which relay in the external segment of the globus pallidus are characterized by the presence of ENK and the absence of MOR and SP.
GABAergic transmission between SONs
Although numerous synaptic contacts have been observed between SONs (Wilson & Groves, 1980; Somogyi et al. 1981), functional inhibitory synapses were not initially found between these projecting neurones (Jaeger et al. 1994). Recently, a weak, relatively slow, but unreliable GABAergic transmission was reported by Tunstall et al. (2002) and Czubayko & Plenz (2002) from striatal slices and organotypic cultures, respectively. In contrast, according to our recordings performed on sagittal slices, in response to single APs, SON–SON GABAergic transmission was strong (∼7 mV), fast (latency < 0.7 ms), precise (latency CV ≈ 0.06) and reliable (failure 6%). Moreover, this GABAergic transmission was unidirectional.
According to anatomical studies, synaptic contacts between SONs mainly occur on proximal aspiny dendrites (Wilson & Groves, 1980; Smith & Bolam, 1990). In agreement with these observations, the fast postsynaptic events recorded in our study favour a low dendritic filtering and therefore suggest that GABAergic synapses between SONs are located on proximal dendrites.
In our recording conditions, due to the very hyperpolarized resting membrane potential of SONs, GABAergic responses between SONs are depolarizing at rest but cannot trigger a spike. Indeed, since ECl was below the AP threshold, postsynaptic responses resulted in a membrane shunt, which inhibited firing and spiking activity. Supporting this statement, the blockade of GABAA receptors increased the spike frequency of SONs evoked by cortical stimulation (Kita, 1996). Similarly, according to Koòs & Tepper (1999), GABAergic interneurones exert an inhibitory influence on the activity of SONs.
This local inhibitory interaction between SONs displays a short-term plasticity: a summation of IPSPs when the presynaptic spike interval was shorter than 25 ms and a depression in the postsynaptic cell for presynaptic spike intervals longer than 25 ms. As observed occasionally in vivo, cortical inputs may trigger a firing activity of SONs at a frequency above 40 Hz (Wilson & Kawaguchi, 1996; Stern et al. 1998). In these conditions, the summation of IPSPs could regulate the efficacy of the unidirectional GABAergic synapse between SONs. However, in most cases the discharge frequency of SONs is lower than 40 Hz thus allowing the occurrence of the depression.
Functional implications of local interactions between SONs
The activity of SONs is generated by cortical excitatory inputs delivered at distal dendrites (Somogyi et al. 1981) and controlled by GABAergic inhibitory inputs on proximal dendrites from either FS interneurones (Koòs & Tepper, 1999) or other SONs through their axon collaterals (Tunstall et al. 2002). Very likely, due to their proximal location, inhibitory GABAergic synapses modulate cortical inputs efficiently. As shown in the cerebral cortex and the hippocampus, electrical coupling contributes to the synchronization of GABAergic interneurones (Fricker & Miles, 2001; Galarreta & Hestrin, 2001; Bennett & Zukin, 2004) and may increase the ability of coincidence detection (Galarreta & Hestrin, 1999; Bartos et al. 2001). However, there is yet no evidence in vivo for the synchronization of action potentials between SONs (Stern et al. 1998) and according to Laurent (2002) such a lack of synchronization could allow more ‘efficient information compression’. Electrical coupling could thus preferentially increase the ability of coincidence detection. In favour of this statement SONs have been shown to act as coincidence detectors of coherent cortical activity (Wickens, 1993; Graybiel et al. 1994; Wilson, 1995).
The unidirectional GABAergic transmission between SONs could contribute to the lack of synchronization of these cells and to the role of electrical synapses in their coincidence detector function. As proposed in Fig. 8C, electrical and GABAergic transmissions are mutually exclusive between pairs of SONs. However, the combination of their actions in neuronal subpopulations may facilitate or decrease output signals of SONs triggered by cortical inputs. Due to the membrane and spiking properties of SONs, the collateral inhibition described in this study should restrict the channelling of the cortico-striatal activity to SONs receiving strong and highly correlated cortical inputs, according to the ‘winner-takes-all’ mechanism proposed by Fukai (1999). In addition, it is expected that GABAergic lateral inbition, in terms of conductance, should be much more efficient than the gap junction conductance. Needless to say, a higher degree of complexity should be expected when considering differences in the strength and timing of cortical inputs between SONs as well as their different neuromodulatory inputs.
The critical role of electrical and GABAergic synapses between SONs in the local circuits of the striatum should also be taken into consideration in pathophysiological states such as Parkinson's disease, which mainly results from the progressive degeneration of the nigro-striatal dopaminergic neurones, or schizophrenia. Indeed, dye coupling between SONs was shown to be increased both in 6-OHDA-lesioned animals and after chronic treatment with neuroleptics (Cepeda et al. 1989; Onn & Grace, 1994) suggesting an increase in electrical coupling (i.e. in terms of the strength of the CC, or the number of coupled cells, or both). Whether the respective importance of gap junctions and unidirectional GABAergic synapses do or do not affect the overall activity of striatal local circuits and therefore their influence on striatal output messages has still to be demonstrated.
The authors wish to thank A.-M. Godeheu for excellent technical assistance with the biocytin experiments and Professor J.-M. Deniau for careful reading of the manuscript. This work was supported by the Institut National de la Santé et de la Recherche Médicale (INSERM), ACI ‘Jeune Chercheur’ from the Ministère de la Recherche and the Fondation de France grant 20020111943.
The online version of this paper can be accessed at: