R-type Ca2+ channels cooperate with P/Q- and N-type channels to control neurotransmitter release at central synapses. The leading candidate as pore-forming subunit of R-type channels is the α1E subunit. However, R-type Ca2+ currents with permeation and/or pharmacological properties different from those of recombinant Ca2+channels containing α1E subunits have been described, and therefore the molecular nature of R-type Ca2+channels remains not completely settled. Here, we show that the R-type Ca2+ current of rat cerebellar granule cells consists of two components inhibited with different affinity by the α1E selective antagonist SNX482 (IC50 values of 6 and 81 nm) and a third component resistant to SNX482. The SNX482-sensitive R-type current shows the unique permeation properties of recombinant α1E channels; it is larger with Ca2+ than with Ba2+ as charge carrier, and it is highly sensitive to Ni2+ block and has a voltage-dependence of activation consistent with that of G2 channels with unitary conductance of 15 pS. On the other hand, the SNX482-resistant R-type current shows permeation properties similar to those of recombinant α1A and α1B channels; it is larger with Ba2+ than with Ca2+ as charge carrier, and it has a low sensitivity to Ni2+ block and a voltage-dependence of activation consistent with that of G3 channels with unitary conductance of 20 pS. Gene-specific knock-down by antisense oligonucleotides demonstrates that the different cerebellar R-type channels are all encoded by the α1E gene, suggesting the existence of α1E isoforms with different pore properties.
- calcium channel
- α1E subunit
- antisense oligonucleotides
- granule cells
- toxin-resistant calcium current
Most neurons of the CNS display several types of high-voltage activated Ca2+ channels, pharmacologically classified as L-, N-, P/Q, and R-type (Dunlap et al., 1995). R-type Ca2+ currents were first described by Tsien and coworkers as corresponding to the residual current observed in rat cerebellar granule neurons after pharmacological block of L-, N-, and P/Q-type Ca2+ channels (Zhang et al., 1993; Randall and Tsien, 1995). Subsequently, R-type currents with different biophysical properties were described in different types of neurons (Tottene et al., 1996; Hilaire et al., 1997; Magnelli et al., 1998; Wu et al., 1998), and single channel recordings revealed that rat cerebellar granule cells coexpress two Ca2+ channels, called G2 and G3, differing in unitary conductance and threshold for activation but both classified as R-type on the basis of their resistance to all specific Ca2+ channel inhibitors (Forti et al., 1994; Tottene et al., 1996).
The molecular basis of R-type channels is not completely settled. The leading candidate as pore-forming subunit of R-type channels is the α1E subunit. However, of the two R-type channels of cerebellar granule cells, only G2 has a unitary conductance of 15 pS, similar to that of recombinant α1Echannels (Schneider et al., 1994; Wakamori et al., 1994; Bourinet et al., 1996), whereas G3 has a conductance of 20 pS (Forti et al., 1994;Tottene et al., 1996). Whereas recombinant α1Echannels have the unique property among high-voltage activated channels of carrying more current with Ca2+ than with Ba2+ (Bourinet et al., 1996), and in addition are very sensitive to Ni2+ block (Soong et al., 1993; Schneider et al., 1994; Williams et al., 1994; Zamponi et al., 1996), native R-type currents show Ba2+ over Ca2+ current ratios ranging from 0.9 to 2 (Zhang et al., 1993; Hilaire et al., 1997; Magnelli et al., 1998), and some require relatively high concentrations of Ni2+ to be blocked (Tottene et al., 1996;Magnelli et al., 1998; Wu et al., 1998). SNX482, the first selective antagonist of recombinant α1E channels, failed to inhibit R-type currents in several types of central neurons, including rat cerebellar granule cells (Newcomb et al., 1998). However, the R-type current of these neurons was reduced by antisense oligonucleotides (ONs) against α1E(Piedras-Renteria and Tsien, 1998).
Whereas splice variants of α1 subunits with different pharmacological properties are known (Bourinet et al., 1999;Hans et al., 1999), there are to date no reports of splice variants or subunit combinations with different permeation properties and/or single channel conductance. Therefore, it remains unclear whether the different native R-type channels and in particular G2 and G3 are all encoded by the α1E gene. Indeed, it has been suggested that some R-type currents might actually be supported by channels containing α1A or α1B subunits with low affinity for their specific toxins (Nooney et al., 1997; Magnelli et al., 1998).
Using an antisense strategy, here we show that the α1E gene encodes both native SNX482-sensitive R-type channels with the unique permeation properties of recombinant α1E channels and native SNX482-resistant R-type channels showing none of these properties.
MATERIALS AND METHODS
Cell culture. Cerebellar granule cells were grown in primary culture after enzymatic and mechanical dissociation from 6- to 7-d-old Wistar rats according to the procedure of Levi et al. (1984). The cells were plated on poly-l-lysine-coated glass coverslips and kept in Basal Eagle's medium supplemented with 10% fetal calf serum, 25 mm KCl, 2 mm glutamine, and 60 μg/ml gentamycin. Cytosine arabinoside (10 μm) was added to the culture 18 hr after plating to inhibit the proliferation of non-neuronal cells. Granule cells were the large majority of the cells in the cultures and were morphologically identified by their oval or round cell body, small size, and bipolar neurites. Experiments were usually performed on granule cells grown for 5–7 d in vitro.
Transfection with oligonucleotides. One day after plating, cerebellar granule cells were transfected using polyethylenimine (PEI) (50 kDa; Sigma, St. Louis, MO) as transfecting agent (Lambert et al., 1996). DNA (1 μg) and 60 nmol of PEI (neutralized to pH 7.0 with HCl) were first separately diluted in 13 μl of 150 mm NaCl. PEI–DNA particles were thereafter obtained by gently mixing the two solutions. Cells, plated in 3.5 cm Petri dishes, were incubated for 1–2 hr in the PEI–DNA solution previously diluted to 0.8 ml with serum-free culture medium. In each culture, part of the cells were exposed to fluorescein-conjugated antisense ONs and part of the cells to fluorescein-conjugated scrambled oligonucleotides (160 nm). ONs were fully phosphorothioated and were designed and produced by Biognostik GmbH (Gottingen, Germany). Specifically, we used antisense ONs against nucleotides 81–98 of the α1E subunit (Soong et al., 1993), corresponding to the N-terminal cytoplasmic region (sequence: GCATATTTCCTGACAATG), against nucleotides 414–431 of the α1A subunit (Starr et al., 1991), corresponding to segment S2 of repeat I (sequence: CCAATGAAATAGGGTTCT), and the corresponding scrambled oligonucleotides (sequences: ACTACTACACTAGACTAC and TCAAAACGAATGCAGTTG, respectively).
Electrophysiology. Whole-cell patch-clamp recordings followed standard techniques (Hamill et al., 1981). Currents were recorded with an Axopatch-200 patch-clamp amplifier (Axon Instruments, Foster City, CA), low-pass filtered at 1 kHz (−3 dB; eight-pole Bessel filter), sampled at 5 kHz using a Digidata 1200 interface and pClamp6 software (Axon Instruments), and stored for later analysis on a computer. Compensation (typically 70–80%) for series resistance was generally used, and only data from cells with a voltage error of <3 mV were analyzed. Experiments were performed at room temperature (21–25°C).
To measure the R-type calcium current in isolation, cells were preincubated for 10 min into a recording chamber containing Tyrode's solution supplemented with 1 μm ω-conotoxin-GVIA (ω-CgTx-GVIA) (Bachem, Budendorf, Switzerland), 3 μmω-conotoxin-MVIIC (ω-CTx-MVIIC or SNX230 provided by Neurex Corporation, Menlo Park, CA), 5 μm nimodipine (gift from Dr. Hof, Sandoz, Basel, Switzerland), and 0.1 mg/ml cytochrome c (Sigma, St. Louis, MO). After attainment of the whole-cell configuration, cells were perfused with the external recording solution containing: 5 mm BaCl2, 148 mm TEA-Cl, 10 mm HEPES (adjusted to pH 7.4 with TEA-OH), 5 μm nimodipine, and 0.1 mg/ml cytochrome c. Control experiments in which ω-CgTx-GVIA and ω-CTx-MVIIC were sequentially applied at increasing times after perfusion with external solution, established that slow unblocking of the toxins could account for at most 10% of the R-type current measured after 30 min (the maximal duration of our recordings). Internal solution contained (in mm): 100 Cs-methanesulfonate, 5 MgCl2, 30 HEPES, 10 EGTA, 4 ATP, 0.5 GTP, and 1 mm cAMP (adjusted to pH 7.4 with CsOH). The perfusion system consisted of six microcapillary Teflon tubes glued together and placed inside a standard plastic pipette (Gilson Medical Electronics, Villiers-le-Bel, France) at ∼12 mm from the tip (∼1.2 mm diameter), which was cut to have a flute beak shape, and positioned close to the cell. The tubes were fed by gravity from reservoirs containing external solution with or without toxins. Switching between different solutions was controlled by solenoid valves. Delay time for complete solution change was <8 sec. Cytocrome c (0.1 mg/ml) was included in all recording solutions to block nonspecific peptide binding sites. Liquid junction potential at the pipette tip was −8 mV (pipette negative), and that between the Tyrode's solution in the experimental chamber and the external recording solution (flowing from the capillary tube) was −4 mV; these two junction potentials should be added to all voltages to obtain the correct values of membrane potential in whole-cell recordings (Neher, 1992). Isolated cells were chosen for recording. The experiment was discarded if cells showed signs of inadequate space clamping, such as notch-like current discontinuities, slow components in the decay of capacitative currents (in response to hyperpolarizing pulses), or slow tails not fully inhibited by nimodipine (Forti and Pietrobon, 1993). Barium currents were corrected on-line for leak and capacitative currents with the P/4 pulse protocol. Averages are given as mean ± SEM. The statistical significance of paired values was tested by an ANOVA, followed by a post hoc t test.
The neurotoxin SNX482, a 41 amino acid peptide present in the venom of the African tarantula Histerocrates gigas (Newcomb et al., 1998) was kindly provided by G. Miljanich and L. Nadasdi (Elan Pharmaceuticals Inc., Menlo Park, CA).
The peptide neurotoxin SNX482 selectively inhibits recombinant calcium channels containing α1E subunits with an IC50 for block of ∼30 nm(Newcomb et al., 1998). We assessed the sensitivity to SNX482 of the R-type calcium current of rat cerebellar granule cells in primary culture. To isolate the R-type calcium current, L-, N-, and P/Q-type calcium channels were completely inhibited using saturating concentrations of nimodipine, ω-CgTx-GVIA, and ω-Ctx-MVIIC (Hillyard et al., 1992; Zhang et al., 1993; McDonough et al., 1996;Tottene et al., 1996). As shown in Figure1 A (top panel), 10 nm SNX482 slowly inhibited a fraction of the R-type current. No significant further inhibition was observed with 30 nm toxin, but 100 nm toxin inhibited more, indicating the presence of at least two components of R-type current with different sensitivity to SNX482. The lack of significant further inhibition by 200 nm toxin with respect to 100 nm, shown in the bottom panel of Figure 1 A, indicates that the R-type calcium current of cerebellar granule cells actually comprises three components with different sensitivity to the toxin. The three R-type current components, Ra, Rb, and Rc, can be seen in the dose–response curve in Figure 1 B. The values of IC50for inhibition of the three components, obtained by fitting the data points with the sum of three Hill equations, were 6, 81, and 654 nm (fractional contributions: 32, 17, and 51%, for Ra, Rb, and Rc, respectively). Figure 1 then shows that the R-type calcium current of rat cerebellar granule cells comprises two components, Ra and Rb, inhibited by concentrations of SNX482 close to those that inhibit recombinant α1E channels, and in addition, a third component, Rc, that can be considered as toxin-resistant because SNX482 is not a selective blocker of α1E channels at the concentrations required to inhibit Rc (Newcomb et al., 1998). The one order of magnitude difference in IC50 for inhibition of components Ra and Rb is reminiscent of the difference in affinity for ω-agatoxin IVA (ω-AgaIVA) of P- and Q-type calcium currents (Randall and Tsien, 1995). The two components Ra and Rb differ also in the time course of inhibition by SNX482. As shown by the representative experiment in Figure4 A, the kinetics of inhibition by 30 nm toxin (component Ra) were faster (τ of 50 ± 5 sec; n = 10) than those of inhibition by 200 nm toxin, sequentially added after 30 nm (component Rb: τ of 74 ± 8 sec; n = 11).
Recombinant calcium channels containing α1Esubunits are characterized by a high sensitivity to Ni2+ block (Soong et al., 1993; Schneider et al., 1994; Williams et al., 1994; Zamponi et al., 1996) and by a larger macroscopic current with Ca2+ as charge carrier compared with Ba2+(Bourinet et al., 1996; and our unpublished observations with both human α1E-d and α1E-3subunits). Both observations point to distinct permeation properties of α1E with respect to α1A, α1B, and α1C channels. Tottene et al. (1996) have shown previously that only 50% of the R-type current of cerebellar granule cells shows a high sensitivity to Ni2+block, requiring concentrations of 30–50 μmNi2+ to be fully inhibited. The less sensitive component was inhibited with an IC50 of 153 μm. Figure 2 shows that the R-type current remaining in the presence of 30 μmNi2+ was not further inhibited by concentrations of SNX482 (30 and 100 nm) that completely blocked components Ra and Rb; however, it was partially inhibited by 500 nm SNX482, a concentration that partially inhibited component Rc. These occlusion experiments show that the two R-type current components with relatively high affinity for SNX482 are both very sensitive to Ni2+block, whereas the R-type calcium channels resistant to the toxin have a relatively low sensitivity to Ni2+block.
Figure 3 further shows that Rc, the R-type current component resistant to SNX482, has different permeation properties than the two SNX482-sensitive components. Whereas the total R-type current at the peak of the current–voltage (I–V) relationship had similar amplitude with Ba2+ or Ca2+as charge carrier (Fig. 3 A), the Rccurrent remaining in the presence of 200 nmSNX482 was larger with Ba2+ (Fig.3 B). On average, the ratioI Ca2+/I Ba2+was 0.99 ± 0.02 (n = 4) and 0.72 ± 0.09 (n = 3) for total R-type and Rccurrents, respectively (significantly different at p < 0.02). One can then argue that the two SNX482-sensitive current components (Ra and Rb) are larger with Ca2+ than with Ba2+ as charge carrier and therefore have permeation properties similar to those of recombinant α1E channels (compare also their high sensitivity to Ni2+ block). On the other hand, the Rc component has permeation properties similar to those of α1A, α1B, and α1C channels.
Tottene et al. (1996) have shown previously that at least two different calcium channels, G2 and G3, contribute to the R-type calcium current of cerebellar granule cells. The two channels differ mainly in single channel current and conductance and in voltage range for activation (Forti et al., 1994). G2 channels, with unitary conductance of 15 pS, activate at ∼15 mV more negative voltages than G3 channels with unitary conductance of 20 pS. It appears reasonable to predict that G2 channels with unitary conductance similar to that of recombinant α1E channels should support the two SNX-sensitive current components Ra and Rb (with permeation properties similar to those of recombinant α1E channels) and that G3 channels with different unitary conductance should support the SNX-resistant Rc component (with permeation properties different from those of recombinant α1E channels). If this hypothesis is correct, component Rc should activate at more positive voltages than components Ra and Rb. To verify this hypothesis and eventually obtain information on the relative sensitivity to SNX482 of G2 and G3 channels, we performed whole-cell recordings in which 30 and 200 nm SNX482 were sequentially added, and the I–Vrelationship was measured before and after each toxin addition. According to the dose–response curve in Figure 1, 30 nm toxin completely inhibits component Ra, which is then given by the difference between the R-type current measured in the presence and absence of 30 nm toxin; 200 nm toxin should completely inhibit component Rb, which can then be obtained as the difference between the R-type current measured in the presence of 200 and 30 nm toxin. Component Rc is given by the current remaining in the presence of 200 nm toxin.
As shown by the representative experiment in Figure4 A, the I–Vcurve in the presence of 30 nm toxin was shifted toward more positive voltages with respect to that in its absence, and the I–V curve in the presence of 200 nm toxin was shifted in the same direction with respect to that in the presence of 30 nm toxin. Figure 4 B shows the average normalizedI–V curves for the three components; component Ra activates at slightly more negative voltages than Rb, and both Ra and Rb activate at more negative voltages than Rc. The three R-type components had slightly different kinetics of inactivation (5 ± 2, 24 ± 3, and 17 ± 2% decay in 136 msec for Ra, Rb, and Rc, respectively;n = 10) (see Fig. 4 A). The different current–voltage relationships, the different kinetics of inactivation, and the different time course of inhibition together support the notion that different calcium channels underlie the R-type current components identified on the basis of the dose–response curve for SNX482. The data are consistent with the conclusion that the channels with the highest affinity for SNX482 correspond to G2, the R channel subtypes activating at more negative voltages. Given the large difference in half-voltage for activation (18 mV) between G2 and G3, the R-type channels resistant to SNX482 most likely correspond to G3, the R channel subtypes with unitary conductance different from that reported for α1E channels. Figure 4 C shows that the steady-state inactivation curve of the current remaining in the presence of 200 nm SNX482 was almost identical to that of the total R-type current. Steady-state inactivation of the calcium channels that underlie the three R-type current components then occurred at quite negative voltages (V 1/2 of −76 mV) in a similar voltage range. Steady-state inactivation of both G2 and G3 channels occurred in a similar negative voltage range.
Both the pharmacological and permeation properties of the R-type channels resistant to SNX482 of rat cerebellar granule cells are quite different from those of recombinant channels containing α1E subunits. α1B, α1A, or α1E subunits with low affinity for their specific toxins or an unknown α1 subunit appear equally likely as the pore-forming subunits of these channels. To investigate the molecular basis of the SNX482-resistant R-type channels and to directly show that α1E subunits are the pore-forming subunits of the SNX482-sensitive R-type channels, we turned to an antisense strategy.
Cerebellar granule cells were transfected with fluorescein-conjugated anti-α1E antisense ON 1 d after plating. R-type calcium current densities were measured at 3, 4, and 5 d after transfection. At each day, control R-type current densities were measured in cells from the same culture transfected with fluorescein-conjugated scrambled ON. Transfected neurons were identified by their fluorescent nuclei. As shown in Figure5 A, α1E subunit knock-down strongly decreased the R-type current of cerebellar granule cells. A maximum decrease of 87% was achieved 4 d after transfection (from 23 ± 4 to 3 ± 1 pA/pF; p ≪ 0.001). After 3 d, the current was already 75% reduced (p < 0.001), suggesting a turnover time of ∼3 d for these α1 subunits. Figure5 B shows that α1E subunit knock-down decreased to a similar extent both the fraction of R-type current inhibited by 30 μmNi2+ (corresponding to components Ra and Rb) and the fraction of R-type current remaining not inhibited (corresponding to component Rc). Analysis of the development of the different components of control R-type current with increasing days in culture shows that the small increase in total R-type current observed between days 3 and 4 after transfection (corresponding to the fourth and fifth day in culture) is caused by an increased expression of the SNX482-resistant R-type channels with low affinity for Ni2+. The R-type current remaining in the presence of 30 μmNi2+ increased 78% (p < 0.04) between days 3 and 4 after transfection, whereas the R-type current inhibited by the same concentration of Ni2+ did not change.
The specificity of the anti-α1E antisense ON was tested by investigating its effect on the N-type calcium current. Figure 6 A shows that the current density inhibited by 1 μmω-CgTx-GVIA was similar in cells transfected with antisense and scrambled ON, whereas in the same cells the R-type current was 73% decreased in antisense-transfected cells. As an additional test, we ascertained that an anti-α1A antisense oligonucleotide, which proved to be effective in decreasing the P-type current in Purkinje cells (Gillard et al., 1997), did not affect the R-type current of cerebellar granule cells. Figure6 B shows that the R-type current was similar in cells transfected with anti-α1A antisense and scrambled ON.
Thus, quite unexpectedly, the antisense data show that α1E subunits form the pore of the R subtypes of calcium channels of rat cerebellar granule cells with pharmacological and permeation properties different from those of recombinant channels containing α1E subunits. Moreover, they directly show that the two SNX482-sensitive R subtypes both contain α1E subunits as pore-forming subunits.
Using an antisense strategy combined with electrophysiology and the new selective pharmacological tool provided by SNX482, we show that rat cerebellar granule cells coexpress calcium channels containing α1E subunits with widely different pharmacological and biophysical (including permeation) properties.
Previous work on recombinant calcium channels containing α1E subunits has led to the identification of a few specific properties distinguishing them from high-voltage activated α1A, α1B, and α1C channels. The specific properties, common to all the different isoforms of α1E studied so far, include the following (1) a high sensitivity to Ni2+ block (Soong et al., 1993; Schneider et al., 1994; Williams et al., 1994); (2) the peptide neurotoxin SNX482 as the only selective, high-affinity antagonist (Newcomb et al., 1998, and our unpublished observations), (3) a larger macroscopic current with Ca2+ than with Ba2+ as charge carrier (Bourinet et al., 1996; and our unpublished observations with both human α1E-d and α1E-3subunits), and (4) a single channel conductance of 12–15 pS (Schneider et al., 1994; Wakamori et al., 1994; Bourinet et al., 1996).
Here, we show that these properties are shared by the native calcium channels containing α1E subunits, which account for the two SNX482-sensitive components of the R-type current of rat cerebellar granule cells. The one order of magnitude difference in IC50 for inhibition of these two components (6 and 81 nm) is reminiscent of the difference in affinity for ω-AgaIVA of P- and Q-type calcium currents (Randall and Tsien, 1995). It has been shown that α1A subunits are the pore-forming subunits of both P- and Q-type channels (Gillard et al., 1997; Piedras-Renteria and Tsien, 1998) and that alternative splicing of the α1A gene may give rise to calcium channels with pharmacological and biophysical properties similar to those of P- and Q-type channels (Bourinet et al., 1999). Likewise, alternative splicing of the α1E gene may possibly account for the different sensitivity to SNX482 of the two R-type current components. Alternatively, the different pharmacology may be attributable to different combinations with auxiliary subunits (Moreno et al., 1997).
Single channel recordings have shown previously that two calcium channels, G2 and G3, with different unitary conductance and voltage-dependence of activation, contribute to the R-type calcium current of rat cerebellar granule cells (Forti et al., 1994; Tottene et al., 1996). G2 channels have unitary conductance and current (15 pS, 0.6 pA at 0 mV) similar to those of recombinant α1E channels, whereas G3 channels have larger unitary conductance and current (20 pS, 0.8 pA at 0 mV). G2 channels activate at ∼15 mV more negative voltages than G3. A comparable difference in voltage range of activation has been found here between the component of R-type current most sensitive to SNX482 and that resistant to SNX482, the first activating at more negative voltages than the latter. Our data are consistent with the conclusion that the R-type channels with high affinity for SNX482 correspond to G2, and those with very low affinity for SNX482 correspond to G3. Given the small difference in voltage-dependence of activation of the two SNX482-sensitive components of R-type current, it appears likely that, at the single channel level, the channels underlying these two components were lumped together under the name of G2. This conclusion is further supported by the fact that, in cell-attached patches, G2 channels were observed more frequently than G3 channels (Forti et al., 1994), and R-type channels different from G2 and G3 were extremely rare (our unpublished observations).
The component of the R-type calcium current of rat cerebellar granule cells resistant to SNX482 shows none of the “specific” properties of recombinant α1E channels; it is not very sensitive to Ni2+ block, it is larger with Ba2+ than with Ca2+ as charge carrier, and, most likely, it is supported by G3 channels with conductance of 20 pS. Nonetheless, strikingly, this component is suppressed after transfection of the neurons with a specific anti-α1E antisense oligonucleotide. Therefore, in addition to calcium channels containing α1E subunits with permeation properties considered as typical of α1E subunits, rat cerebellar granule cells express calcium channels containing α1E subunits with permeation properties more typical of α1A, α1B, or α1C subunits. It is highly unlikely that these permeation properties are caused by a particular combination with auxiliary subunits, because β and α2-δ subunits do not appear to affect single channel conductance and permeation of recombinant α1 channels (Wakamori et al., 1993, 1999; and our unpublished observations). Most likely, these peculiar native R-type channels contain a novel α1E variant, whose properties have not been studied in heterologous expression systems.
The component of R-type current resistant to SNX482, accounting for ∼50% of the R-type current of our cerebellar granule cells, has many properties in common with the original R-type current described in the same neurons by Zhang et al. (1993). The common properties include the ratioI Ca/I Balower than 1, the relatively low sensitivity to Ni2+ block, the resistance to SNX482 (Newcomb et al., 1998), and the reduction by specific anti-α1E antisense oligonucleotides (Piedras-Renteria and Tsien, 1998). However, in cerebellar granule cells cultured under the conditions of Zhang et al. (1993), there was no evidence for the presence of additional components of R-type current inhibited with high affinity by SNX482 (Newcomb et al., 1998). The most likely explanation is the expression of different α1E splice variants in neurons cultured under different conditions and obtained from rats of different age. Similarly, expression of different α1A splice variants (and/or different β subunits) may explain the P-type pharmacology found by Tottene et al. (1996) in contrast with the Q-type pharmacology found by Randall and Tsien (1995) in the same neurons (Moreno et al., 1997; Bourinet et al., 1999).
Reverse transcription-PCR analysis of the mRNA isolated from our primary cultures of cerebellar granule cells has shown the expression of up to six different α1E isoforms alternatively spliced in the II–III loop and the C terminus, some of which had not been cloned before (Schramm et al., 1999). There is evidence for different relative expression of these isoforms in different brain regions and during development (Pereverzev et al., 1998; Schramm et al., 1999). Two of the isoforms do not show differences in biophysical properties (Pereverzev et al., 1998). The functional properties of the other four isoforms remain unknown. Splice variants of α1A, α1B, and α1C with different kinetics and/or voltage-dependence of the macroscopic calcium current have been described previously (Lin et al., 1997; Soldatov et al., 1997; Bourinet et al., 1999; Hans et al., 1999). However, splice variants of calcium channel α1 subunits with different permeation properties and different unitary conductance have never before been reported.
Labeling with antibodies of brain slices has shown localization of α1E subunits in both cell bodies and dendrites of many types of neurons (Volsen et al., 1995; Yokoyama et al., 1995) and more recently also in calyx-type synaptic terminals of the brainstem (Wu et al., 1999). Wu et al. (1998) have shown that R-type channels contribute to action potential-evoked transmitter release at these synapses. Most likely, R-type channels participate in controlling evoked release in many other central synapses (Turner et al., 1993), including cerebellar parallel fiber synapses (Mintz et al., 1995). The localization in cell bodies and dendrites suggest additional postsynaptic roles of R-type channels, e.g., G2-type channels with a relatively low threshold of activation might have a role in synaptic integration and the generation of calcium spikes (D'Angelo et al., 1997). It is reasonable to hypothesize that R-type channels with different voltage-dependence of activation and different unitary conductance as those coexpressed in cerebellar granule cells may have a specialized function in different neuronal processes. Alternative splicing of α1E subunits may allow different subcellular localizations, different modulation, and different binding to specific membrane proteins of the different R-type channels. Indeed, in the case of α1A subunits, it has been shown that the II–III loop and the C-terminal region, the two regions alternatively spliced in the α1E isoforms expressed in cerebellar granule cells, are involved in binding important proteins, such as syntaxin (Rettig et al., 1996) and calmodulin (Lee et al., 1999), respectively, and that alternatively spliced II–III loop isoforms have a different ability to bind syntaxin (Rettig et al., 1996) and have a different subcellular distribution (Sakurai et al., 1996).
This work was supported by Telethon-Italy Grant 720 to D.P. A.T. was supported by the Lilly Postdoctoral Research Program. We thank Drs. G. Miljanich and L. Nadasdi of Elan Pharmaceuticals Inc. for providing SNX483, and Carlo Berizzi and Francesca Bolcato for help in performing some of the experiments.
Correspondence should be addressed to Daniela Pietrobon, Department of Biomedical Sciences, University of Padova, V.le G. Colombo 3, 35121 Padova, Italy. E-mail:.