Main

In cerebellar Purkinje cells, mGluR1 is linked via a Gαq protein complex to the activation of phospholipase Cβ (PLC-β), which cleaves the membrane phospholipid phosphatidylinositol-4,5-bisphosphate to yield 1,2-diacylglycerol and inositol-1,4,5-trisphosphate (InsP3). Activation of mGluR1 by burst stimulation of parallel fibres results in both Ca2+ mobilization through a PLC-β–InsP3 cascade16,17,18 and activation of a slow EPSC3 carried by a mixed-cation conductance4,5,13,14,15. The identity of the ion channel underlying the mGluR1-evoked slow EPSC has not been determined in Purkinje cells or in any other cell type. There are, however, a number of observations that constrain the identity of the ion channel underlying the slow EPSC. It is a mixed-cation conductance that reverses at about +20 mV (refs 5, 14, 15) and requires external Ca2+ (refs 5, 19). It is not blocked by antagonists of Na+/Ca2+ exchangers, purinergic receptors, hyperpolarization-activated cation channels or voltage-gated Ca2+ channels5,13,15.

There are several reasons to believe that the Ca2+ mobilization and slow EPSC signals evoked by mGluR1 activation are dependent on Gαq but subsequently involve divergent pathways. First, drugs that interfere with PLC-β, InsP3 receptors or internal Ca2+ stores completely block Ca2+ mobilization but have variable effects on the slow EPSC conductance, ranging from no blockade to weak blockade4,5,13,14,15. Second, photolysis of caged InsP3 or Ca2+ does not mimic the slow EPSC15. Third, drugs4,13 and induced mutations20 that inhibit Gαq block both Ca2+ mobilization and the slow EPSC. Fourth, weak burst stimulation of parallel fibres can evoke Ca2+ mobilization in the absence of the slow EPSC, whereas stronger burst stimulation recruits both signals (ref. 16 and S.J.K., P.F.W. and D.J.L., unpublished observations).

Group I mGluRs are organized at glutamatergic synapses through interactions with scaffolding molecules, most notably the Homer family of proteins, which form multimers capable of regulating coupling of group I mGluRs to a number of targets including InsP3 receptors, ryanodine receptor and voltage-gated calcium channels21. In recent studies we found that Homer also binds and regulates the gating of members of the TRPC family of nonspecific cation channels22. TRPC channels are activated in response to G-protein-coupled receptor activation and/or depletion of intracellular Ca2+ stores, and are believed to participate in replenishment and regulation of intracellular Ca2+ pools in a process termed capacitive calcium entry23. TRPC1 is expressed in cerebellar Purkinje cells24. Accordingly, we examined the hypothesis that a TRPC ion channel underlies the mGluR1-evoked slow EPSC.

CHO cell lines that stably express mGluR1α or mGluR5 were transfected with TRPC1, TRPC2, TRPC3, TRPC4, TRPC5 or TRPC6. mGluR1α antibody co-immunoprecipitated TRPC1 but not other TRPCs, whereas mGluR5 antibody did not co-immunoprecipitate any of the TRPC channels (Fig. 1a). mGluR1α and TRPC1 also co-immunoprecipitated using a TRPC1 antibody (Fig. 1b). These data indicate a specific physical interaction between TRPC1 and mGluR1α. A TRPC1 mutant lacking channel activity (described below) showed similar association with mGluR1α. To determine whether mGluR1α and TRPC1 associate in vivo, we assayed for co-immunoprecipitation of mGluR1α from detergent lysates of cerebellum using TRPC1 antibodies (Fig. 1c). TRPC1 antibody co-immunoprecipitated mGluR1α and this was blocked by TRPC1 peptide. TRPC1 antibody did not co-immunoprecipitate the AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) receptor subunit GluR2. To assess the ultrastructural localization of TRPC1 we performed post-embedding immuno-electron microscopy on adult rat cerebellum (Fig. 1d). Gold particles preferentially localized to dendritic spines of Purkinje neurons where they co-localized with mGluR1α along the spine membrane. Co-localization was evident in the perisynaptic region as well as at the postsynaptic membrane. These biochemical and anatomical data make TRPC1 an attractive candidate for the ion channel underlying the mGluR1-evoked slow EPSC. Indeed, this suggestion has been made previously based on the similarity of a current recorded in a heterologous cell expressing TRPC1 and either TRPC4 or TRPC5 (ref. 24) to that recorded in hippocampal25 or midbrain dopaminergic26 neurons stimulated with a group I mGluR agonist.

Figure 1: mGluR1α and TRPC1 physically associate in CHO cells, HEK293 cells and in brain, and they co-localize on Purkinje cell spines.
figure 1

a, mGluR1α- and mGluR5-expressing CHO cell lines were transfected with haemagglutinin (HA)-tagged TRPC1 - 6. HA–TRPC1, but not GluR1 or other TRPCs, co-immunoprecipitate with mGluR1α. None of the TRPCs co-immunoprecipitate with mGluR5. Markers are in kDa. b, HEK293 cells were transfected with mGluR1α and/or HA–TRPC1 (WT or F561A). TRPC1 and TRPC1(F561A) co-immunoprecipitate mGluR1α. Anti-TRPC1 antibody does not immunoprecipitate mGluR1α in the absence of TRPC1. c, TRPC1 co-immunoprecipitates mGluR1α (arrow) but not GluR2 (overexposed and displaying a nonspecific band greater than the size of GluR2) from cerebellum. Preabsorption of TRPC1 antibody with peptide antigen reduced mGluR1α by 90%. d, The left column shows immunogold labelling for TRPC1 associated with spine synapses in the molecular layer of the cerebellum. Arrowheads indicate labelling with 10 nm F(ab′)2-gold associated with the perisynaptic as well as extrasynaptic and outer synaptic membrane areas of the postsynaptic spine. Sections were labelled with antibody without (left panel, top row) and with (left panel, bottom row) preabsorption with peptide. Pre, presynaptic terminal. Scale bar: 100 nm. The middle and right columns show co-localization of immunogold labelling for TRPC1 (5 nm) and mGluR1α (10 nm) associated with spine synapses in the molecular layer of the cerebellum. Arrows indicate postsynaptic labelling (post). Peri refers to perisynaptic/extrasynaptic labelling of the spine membrane near glia processes (asterisks) adjacent to the synaptic cleft. Some labelling for TRPC1 is also seen in the presynaptic terminals (pre). Scale bar: 100 nm.

We next examined the mGluR1-evoked EPSC in Purkinje neurons of acute cerebellar slices using whole-cell patch clamp. Stimulation of parallel fibres with a brief burst gives rise to a fast component predominantly mediated by AMPA receptors and a slow component mediated by mGluR1. In some cases the falling phase of the fast component overlaps the rising phase of the slow component, making it difficult to measure them separately. To facilitate our analysis of the delayed EPSC, we applied the AMPA/kainate receptor antagonist CNQX (6-cyano-7-nitroquinoxaline-2,3-dione) at a concentration that produces an approximately 90% reduction in the amplitude of the fast EPSC (10 µM). This allowed for clearly resolvable fast and slow EPSC components (Fig. 2a, b). The amplitude of the stimulation current was adjusted to evoke a 300–500 pA fast EPSC in the absence of CNQX. A burst frequency of 100 Hz was used to maximize the slow EPSC6 and the holding potential was set to -70 mV. As previously reported3,4, single stimuli of parallel fibres did not evoke a measurable slow EPSC, and the amplitude of the slow EPSC depended on the number of pulses in the train. Bath application of an mGluR1 antagonist, CPCCOEt (100 µM), completely abolished the slow EPSC evoked by a train of five stimuli (4.9 ± 1.9% of baseline, mean ± s.e.m., n = 7 cells, within-cell comparisons). A nonselective antagonist of receptor-operated cation channels, SKF 96365 (30 µM), also strongly blocked the mGluR-evoked slow EPSC (18.6 ± 3.0% of baseline, n = 13 cells). The inhibitory effects of both CPCCOEt and SKF 96365 on the slow EPSC were selective, having no effect on the peak amplitude of the fast EPSC (93.8 ± 1.8%, n = 7 cells; 90.2 ± 6.4%, n = 10 cells, respectively).

Figure 2: Characterization of the mGluR1-evoked slow EPSC in Purkinje cells.
figure 2

a, Representative single traces show fast and slow components of the parallel-fibre-evoked EPSC which were produced by burst stimulation at 100 Hz in the presence of CNQX (10 µM) and GABAzine (5 µM). A burst consisting of five pulses was adopted as the standard stimulus. The black traces represent the pretreatment (baseline) condition whereas the blue traces represent the post-treatment condition at an asymptotic time point (typically 5–12 min after introduction of the treatment). Numbers in the top left panel indicate number of pulses. Scale bars: 100 pA, 0.5 s, except for the time-expanded traces in the lower right panel (50 ms). b, Population data for the manipulations shown in a are expressed as the percentage of baseline peak current amplitude and are presented as mean ± s.e.m. c, Fast and slow inward currents were evoked by micropressure pulses (10 p.s.i., 50 ms duration) of external solution containing DHPG (100 µM) and glutamate (40–100 µM) delivered through a patch pipette. Representative current traces are shown before (black) and after (blue) various bath-applied treatments. Scale bars: 200 pA, 0.5 s. d, Population data for the manipulations shown in c are expressed as the percentage of baseline peak current amplitude. e, Upper panel shows the DHPG-evoked slow currents at different potentials in the presence of TTX (1 µM), NBQX (20 µM), CdCl2 (200 µM) and BaCl2 (5 mM, by replacing CaCl2). Scale bars: 200 pA, 0.5 s. Lower panel shows current–voltage relations for a population of five Purkinje cells. Current has been normalized to the value at -70 mV. Linear regression line is shown in the plot.

To evaluate specifically postsynaptic responses, and to avoid possible effects of manipulations on presynaptic release, mGluRs were activated by micropressure pulses of 3,5-dihydroxyphenylglycine (DHPG, 100 µM; 10 pounds per square inch (p.s.i.), 50 ms) delivered through a patch pipette. Glutamate (40–100 µM) was included in this solution to produce a small, fast inward current, serving as a control measure for nonspecific effects (Fig. 2c, d). Similar to results with the EPSC evoked by burst stimulation of parallel fibres, CPCCOEt (100 µM) and SKF 96365 (30 µM) blocked the mGluR-evoked slow current (5.9 ± 1.0%, n = 6 cells; 18.9 ± 6.1%, n = 8 cells, respectively) but left the fast current unaffected. LaCl3, which blocks voltage-gated calcium channels and augments some, but not all, TRPC channels in heterologous expression systems24, did not significantly attenuate either the fast or slow component of the inward current (95.7 ± 4.3%; 88.0 ± 8.0%, n = 5 cells, respectively). Removal of external calcium attenuated the slow current (21.5 ± 9.0%, n = 5 cells) but did not affect the fast current (97.5 ± 6.4%, n = 5 cells). The current–voltage relation of the slow current in Purkinje cells was roughly linear and reversed at about +23 mV, consistent with a previous report15. These properties of the mGluR1-evoked slow current in Purkinje cells are consistent with a conductance mediated by TRPC1, but they do not uniquely implicate TRPC1, nor do they exclude some other candidate ion channels.

To examine the role of TRPC1 in the mGluR1-evoked current, Purkinje neurons in slice culture were transfected with an expression construct of the channel mutant TRPC1(F561A) (Fig. 1b) along with enhanced green fluorescent protein (EGFP; Fig. 3a). In control cells transfected with EGFP alone and stimulated with micropressure pulses of DHPG plus glutamate, the peak amplitude of the slow current was 640 ± 72 pA (n = 10 cells). In cells expressing TRPC1(F561A), the amplitude of mGluR-evoked slow current was reduced to 49% of that measured in an EGFP-alone control (313 ± 39 pA, n = 33 cells, P < 0.001). The amplitude of slow currents in non-transfected Purkinje cells (686 ± 95 pA, n = 9 cells) and in cells transfected with wild-type TRPC1 (802 ± 116 pA, n = 11 cells) was significantly larger than that of cells transfected with TRPC1(F561A) (P < 0.001 for both), and was not significantly different from cells transfected with EGFP alone (P > 0.10 for both).

Figure 3: Manipulations that interfere with TRPC1 function block the mGluR1-evoked slow current in Purkinje cells.
figure 3

a, Population data in four different conditions are expressed as the peak amplitude of mGluR1-evoked slow current in Purkinje cells in organotypic slice cultures. The amplitude of slow current from cells transfected with a dominant-negative pore mutant of TRPC1 (TRPC1(F561A)) was attenuated in comparison to that of cells transfected with EFGP alone, non-transfected cells and cells transfected with wild-type TRPC1. Representative single-current traces are shown to the right. Scale bars: 200 pA, 0.5 s. b, Time course of the amplitude of the slow current that was activated by micropressure pulses of DHPG together with glutamate. The current amplitude was normalized to the value recorded immediately after achieving a stable whole-cell configuration. Representative current traces are shown to the right. Scale bars: 200 pA, 0.5 s. Ab, antibody. Representative current traces are shown before (black) and after (blue) various treatments.

To provide an independent test for the role of TRPC1, we included TRPC1 antibody (30 µg ml-1) in the pipette solution and performed whole-cell recordings from Purkinje cells in organotypic culture (Fig. 3b). In a separate group of cells we applied the antibody rendered nonfunctional by preabsorption with TRPC1 peptide. In TRPC1-antibody-treated cells the amplitude of the slow current started to decrease about 15 min after achieving stable whole-cell recording. At 37.5 min, the difference between the normalized slow current with antibody preabsorbed with TRPC1 peptide (nonfunctional; 1.02 ± 0.06, n = 6 cells) and that with TRPC1 antibody (0.36 ± 0.06, n = 6 cells) was statistically significant (P < 0.01). This time course probably reflects the diffusion of TRPC1 antibody from the pipette to the dendritic site of mGluR1 activation.

As a further test of the hypothesis that TRPC1 underlies the mGluR1-evoked slow EPSC, we attempted to reconstitute this conductance in a heterologous system. A line of CHO cells constitutively expressing mGluR1 was co-transfected with EGFP and TRPC1. EGFP-positive cells were subjected to whole-cell patch-clamp recording. A command potential of -70 mV was imposed, and mGluR1 was activated with a micropressure pulse of DHPG (10 p.s.i., 100 ms). This evoked a slowly activating inward current with a peak current density of 19.3 ± 3.9 pA pF-1 (mean ± s.e.m., n = 8 cells; Fig. 4a, b). TRPC5 is reported to form heteromultimers with TRPC1 (ref. 27), and CHO cells transfected with both TRPC1 and TRPC5 displayed an even larger DHPG-evoked current (50.7 ± 7.8 pA pF-1, n = 7 cells). In contrast, when a separate population of CHO cells was transfected with TRPC3—a related channel that does not form a complex with mGluR1—no significant inward current was evoked (0.5 ± 0.5 pA pF-1 and 0.2 ± 0.3 pA pF-1, respectively, n = 6 cells per group). Transfection with TRPC1(F561A) similarly yielded no significant current (0.4 ± 0.3 pA pF-1, n = 7). When TRPC1 was transfected together with TRPC1(F561A) in a ratio of 2:1, the amplitude of the evoked slow current was reduced to about 25% of that seen with TRPC1 alone (4.8 ± 2.5 pA pF-1, n = 6). This is similar to the effect of TRPC1(F561A) transfection in Purkinje cells (Fig. 3). The current–voltage relation for TRPC1/TRPC5-transfected CHO cells (not shown) revealed that the DHPG-evoked current was roughly linear and reversed at about +23 mV, similar to the mGluR1 EPSC in Purkinje cells (Fig. 2 and ref. 22). Both CPCCOEt (100 µM) and SKF 96365 (50 µM) completely abolished the DHPG-evoked current (5 ± 4% of baseline, n = 7 cells; 3 ± 3% of baseline, n = 8 cells, respectively). The DHPG-evoked current was not significantly altered by LaCl3 (100 µM; 91 ± 9% of baseline, n = 7 cells). Removal of external calcium, previously shown to attenuate the slow EPSC in Purkinje cells7,28 (see Fig. 2c, d), produced a near-complete blockade of the DHPG-evoked current (13 ± 4% of baseline, n = 7 cells). Notably, internal application of a TRPC1-inactivating antibody (30 µg ml-1), but not a control solution consisting of antibody preabsorbed with TRPC1 peptide, produced a strong attenuation of DHPG-evoked current (26 ± 4% of baseline, n = 6; 109 ± 8% of baseline, n = 7, respectively).

Figure 4: The inward current evoked by DHPG in mGluR1/TRPC1-expressing CHO cells closely resembles the mGluR1-evoked slow EPSC in Purkinje cells.
figure 4

a, Representative current traces in a line of CHO cells that constitutively expressed mGluR1 together with the indicated ion channels. Micropressure pulses of DHPG were delivered as indicated by the horizontal bars (100 ms, 10 p.s.i.). Black traces are single sweeps and blue traces are the average of six consecutive sweeps. Scale bars: 50 pA, 1 s. b, Population data for the manipulations shown in a normalized to cell membrane capacitance (as an index of plasma membrane surface area) and presented as mean ± s.e.m. c, Representative current traces for experiments using within-cell comparisons for CHO cells expressing TRPC1 together with mGluR1. In this panel all traces are the average of six consecutive sweeps. The black trace represents the pre-treatment condition whereas the blue trace represents the post-treatment condition at an asymptotic time point (typically 5–12 min after introduction of the treatment). Scale bars: 50 pA, 1 s. d, Population data for the manipulations shown in c are expressed as the percentage of baseline peak current amplitude.

Biochemical studies demonstrate a selective interaction of mGluR1 with TRPC1 (Fig. 1a). To assess whether mGluR1 may be selectively capable of activating TRPC1, we performed simultaneous voltage-clamp and ratiometric calcium imaging of CHO cells expressing TRPC1 with other G-protein receptors, including mGluR5 or M1 muscarinic receptor. In cells expressing mGluR5 and TRPC1, DHPG application resulted in a robust calcium transient (568 ± 155 nM, peak free calcium, n = 10 cells; thereby confirming that mGluR5 was functionally expressed) but no significant activation of membrane current (0.9 ± 0.8 pA pF-1). When this experiment was repeated with cells expressing mGluR1, strong calcium responses (783 ± 145 nM, n = 10) and a robust inward current (25.3 ± 4.6 pA pF-1) were observed, similar to that previously seen using a CHO cell line engineered to stably express mGluR1 (Supplementary Fig. 1). When this experiment was repeated with cells expressing M1 muscarinic receptor in place of mGluR, and stimulation with the muscarinic agonist pilocarpine in place of DHPG, the results were similar to that seen with mGluR5 (that is, a strong calcium signal was observed (696 ± 88 nM, n = 10) but no significant membrane current accompanied this calcium transient (0.6 ± 0.5 pA pF-1)). These findings show that not all Gq/PLC-coupled receptors can activate the TRPC1-mediated slow conductance, and they suggest that the specific physical association of mGluR1 and TRPC1 is important for signal coupling.

Our observations indicate that TRPC1 is required for the mGluR1-induced inward current in Purkinje neurons. Because TRPC1 is known to co-assemble with TRPC4 and TRPC5, it is possible that the native channel is a heteromultimer24,27. Furthermore, although it is likely that the slow EPSC evoked by group I mGluR activation in Purkinje cells is similar to that in other neurons, there may be some important differences as well. For example, the slow EPSC recorded in hippocampal pyramidal neurons seems to require activation of both mGluR1 and mGluR5 to produce a maximal effect25 whereas Purkinje cells only express mGluR1.

What is the function of the mGluR-evoked slow EPSC? One possibility is that the Ca2+ influx associated with this current serves to replenish internal stores depleted by activation of the PLC–InsP3 signalling cascade. In other cell types this is accomplished by direct coupling between calcium stores and TRPC channels23. In Purkinje cells the relationship between activation of the PLC–InsP3 cascade and the TRPC1 cascade is less clear, as manipulations that interfere with PLC-β, InsP3 receptors or internal Ca2+ stores have weak and variable effects on the slow EPSC conductance4,5,13,14,15. It is also possible that the Ca2+ and/or Na+ influx mediated by the slow EPSC engages other second messenger pathways. For example, activation of mGluR1 in Purkinje cells drives the postsynaptic production of endocannabinoids, which function as diffusible retrograde messengers to transiently depress transmitter release28.

Methods

Cell biology assays

Detergent (1% Triton) lysates of rat cerebellum and cell lines were generated and analysed as described22. TRPC constructs, cell culture and transfection methods are as described22. A total of 100 µl lysates precipitated with 2 µg of anti-TRPC1 antibodies (Alomone). For the negative control, 2 µg of anti-TRPC1 antibodies were preabsorbed with equal amounts of TRPC1 peptide antigen (amino acids 523–537 TRPC1) for 6 h at 4 °C. TRPC1 point mutant F561A was generated by QuikChange site-directed mutagenesis (Stratagene).

Immuno-electron microscopy and immunogold quantification

Immuno-electron microscopy was performed by post-fixation immunogold labelling29 using a TRPC1 antibody (Alomone) at a dilution of 1:100. A monoclonal mGluR1α antibody29 was used for double labelling. Brightness and contrast of micrographs were modified using Adobe Photoshop. Results of immunogold labelling are shown in Fig. 1d. Pretreatment of the TRPC1 antibody (left panel of bottom row) with the antigen peptide removed 95% of the gold labelling. Counts of gold particles were 2.67 gold particles per synapse for the area of the postsynaptic density (PSD) and 2.80 for the remaining extrasynaptic area of the spine (63 spine synapses from two animals). As the PSD represents about one-third of the area of spine profiles, these counts indicate enrichment of labelling near the PSD. In co-localization studies a total of 96% of spines labelled with mGluR1 also displayed label for TRPC1 on the spine membrane (Fig. 1d, n = 121; two animals). The postsynaptic membrane contained 51% and 35% of TRPC1 and mGluR1α membrane labelling, respectively. The perisynaptic region (defined as one-third of the length of the postsynaptic membrane) contained 17% and 38% of TRPC1 and mGluR1α membrane labelling, respectively.

Electrophysiology

Parasagittal slices of the cerebellar vermis (200 µm thick) were prepared from P18–20 Sprague–Dawley rats using a vibrating tissue slicer and ice-cold standard artificial cerebrospinal fluid (ACSF) containing 124 mM NaCl, 5 mM KCl, 1.25 mM Na2HPO4, 2 mM MgSO4, 2 mM CaCl2, 26 mM NaHCO3 and 10 mM d-glucose, bubbled with 95% O2 and 5% CO2. After a recovery period of 30 min at 37 °C, the slices were placed in a submerged chamber that was perfused at a rate of 2 ml min-1 with ACSF supplemented with 5 µM GABAzine to block GABAA receptors. Experiments were performed using the visualized whole-cell patch-clamp technique at room temperature. The recording electrodes (resistance 4–5 MΩ) were filled with a solution containing 135 mM Cs-methanesulphonate, 10 mM CsCl, 10 mM HEPES, 4 mM Na2ATP, 0.4 mM Na3GTP and 0.2 mM EGTA (pH 7.25). Calcium-free extracellular solution was prepared by replacing CaCl2 with MgCl2 and adding 0.2 mM EGTA. LaCl3 was added to a HEPES-based external solution to prevent precipitation. All drugs were purchased from Sigma except for CPCCOEt, SKF 96365, GABAzine and CNQX, which were purchased from Tocris Cookson. Currents were filtered at 1 kHz and digitized at 5 kHz. For parallel fibre stimulation, standard patch pipettes were used and were filled with external saline and placed in the middle third of the molecular layer. Synaptic responses were evoked every 30 s using 12–16 µA pulses (100 µs duration). When burst stimulation was used, the interpulse interval was 10 ms. In some experiments membrane currents were evoked by agonist application using a pressure application system (Picospritzer; 10 p.s.i., 50 ms pulse duration).

For organotypic cerebellar cultures, 200-µm thick parasagittal slices from P10–12 rats were placed in 1 µm Millicell inserts (Becton Dickinson) within sterile 6-well plates. Medium (1.5 ml) was placed under each insert (50% Eagle's basal medium, 20% Hank's balanced salt solution, 25% heat-inactivated horse serum, 30 mM glucose, 5 mM glutamine, 5 mM HEPES, 1 µg ml-1 insulin, 0.012% ascorbic acid, and 100 U ml-1 penicillin-streptomycin; all from GIBCO BRL) to achieve an interface configuration. Plates were incubated in 5% CO2 in air at 37 °C. Cultured slices were transfected using the Helios Gene Gun (Bio-Rad) 24 h after slice preparation. To prepare ‘bullets’, 25 mg of 1 µm gold particles (Bio-Rad) were coated with DNA plasmid combinations (45 µg TRPC1 mutant plus 5 µg EGFP, 45 µg TRPC1 wild type plus 5 µg EGFP, and 45 µg pRK5 (empty plasmid) DNA plus 5 µg EGFP DNA). A minimum of 10 µl each of 0.5 M spermidine and CaCl2 was used, or this volume was increased to match the volume of DNA. Coated, rinsed gold particles were resuspended in 0.06 mg ml-1 polyvinylpyrrolidone in ethanol and then loaded into tubing. Particles were accelerated into cultured slices with the gene gun pressurized to 180 p.s.i. Plates were returned to the incubator immediately and used for recording after 24 h.

CHO cells were bathed in a solution that contained NaCl (140 mM), KCl (5 mM), CaCl2 (2 mM), MgCl2 (1 mM), HEPES (10 mM) and glucose (10 mM), adjusted to pH 7.35 with NaOH, which flowed at a rate of 1 ml min-1. The internal saline was identical to that used in Purkinje cell recordings. Patch electrodes were pulled from N51A glass and yielded a resistance of 3–5 MΩ. Membrane currents were recorded in resistive voltage-clamp mode, filtered at 2 kHz and digitized at 5 kHz. For CHO cell experiments in which voltage-clamp recording and calcium imaging were combined, EGTA was omitted from the internal saline and replaced with 0.2 mM bis-fura-2. Imaging of free calcium was performed as previously described30. Experiments were conducted at room temperature. Cells in which Rinput or Rseries varied by more than 15% were excluded from the analysis.