Splicing of α_1A subunit gene generates phenotypic variants of P- and Q-type calcium channels

Abstract

P-type and Q-type calcium channels mediate neurotransmitter release at many synapses in the mammalian nervous system. The α_1A calcium channel has been implicated in the etiologies of conditions such as episodic ataxia, epilepsy and familial migraine, and shares several properties with native P- and Q-type channels. However, the exact relationship between α_1A and P- and Q-type channels is unknown. Here we report that alternative splicing of the α_1A subunit gene results in channels with distinct kinetic, pharmacological and modulatory properties. Overall, the results indicate that alternative splicing of the α_1A gene generates P-type and Q-type channels as well as multiple phenotypic variants.

Main

The electrophysiological and pharmacological diversity of native calcium (Ca²⁺) channels (L, N, P, Q and T types) is well documented¹. These subtypes have different functions; low-voltage-activated T-type channels shape action potentials and generate firing patterns², whereas L-type channels regulate Ca²⁺-dependent genes and enzymes^3,4, and N-type and P/Q-type channels contribute to neurotransmitter release^5,6,7,8.

High-threshold neuronal Ca²⁺ channels are heterotrimeric complexes composed of a pore-forming α₁ subunit associated with β and α₂δ subunits. Each of the cloned α₁ subunits has distinct functional characteristics that are modulated by co-expression of any of four β subunits⁹. Channels formed from α_1C and α_1D subunits have the properties of neuronal dihydropyridine-sensitive L-type channels, and α_1B channels encode ω-CgTx GVIA-sensitive N-type channels⁹. The α_1E channel shares some properties with both low-threshold channels (high sensitivity to nickel block, permeation Ca²⁺ > Ba²⁺) and high-threshold Ca²⁺ channels (activation at more positive potentials, single-channel conductance > 10 pS)^10,11,12, although the native counterpart of this subunit remains unclear. T-type low-threshold channels are encoded by at least three different α₁ subunits (α_1G, α_1H, α_1I) and seem not to require β and α₂δ subunits for functional expression^13,14.

P-type Ca²⁺ channels, originally described in cerebellar Purkinje cells¹⁵, are widely distributed^16,17 and mediate neurotransmitter release in the central and peripheral nervous systems^5,6,7,8. Q-type Ca²⁺ channels, first described in cerebellar granule cells^18,19, also mediate neurotransmitter release at some synapses²⁰. Native P- and Q-type channels differ in sensitivity to ω-agatoxin IVA (ω-Aga IVA; K_d ~2 nM for P-type versus > 100 nM for Q-type) and in their inactivation kinetics. (P-type currents show a non-inactivating waveform during prolonged membrane depolarization, whereas Q-type currents show a pronounced inactivation.) Some native Ca²⁺ channels have properties related to but distinct from either P- or Q-type channels, suggesting a closely related family of channel types^21,22.

The α_1A subunit^23,24 is highly expressed in both Purkinje and cerebellar granule cells and shares several properties with both native P- and Q-type Ca²⁺ currents^25,26. Mutations in the α_1A subunit result in several neuropathological conditions, including familial hemiplegic migraine, cerebellar ataxia and epilepsy^27,28,29. The properties of α_1A do not exactly match those of either P- or Q-type Ca²⁺ currents, and there has also been no account for their distinct sensitivities to ω-Aga IVA. Here we report that alternative splicing at distinct sites within the α_1A subunit gene generate multiple types of P- and Q-type conductances that exhibit distinct gating, pharmacology and modulatory characteristics. We also suggest there is little reason that α_1A channels should be selectively subdivided into distinct P- versus Q- subtypes.

Results

α_1A isoforms are generated by alternative splicing

Screening of a rat brain cDNA library identified a calcium channel α_1A subunit (called α_1A-b) that differed from the primary sequence of the original rat brain α_1A-a ( ref. 24) at three sites ( Fig. 1a ). The domain I–II linker in α_1A-b contained a single valine insertion (Val₄₂₁) located 18 residues carboxyl to the β subunit binding site³⁰. In domain IV, α_1A-b contained an insertion of two residues, Asp and Pro (N₁₆₀₅-P₁₆₀₆), in the extracellular linker separating transmembrane segments S3 and S4. Finally, α_1A-b contained ten substituted residues in a stretch of thirty amino acids in the carboxyl tail adjacent to domain IV S6. Of the ten substitutions, six are located in a region highly similar to the divalent ion binding domain (EF hand) of Ca²⁺-binding proteins³¹.

Figure 1: Alternative splicing generates multiple α_1A isoforms.

(a) Differences between α_1A-a and α_1A-b subunits (single-letter code). (b) Comparison of the α_1A genomic and cDNA sequences in the domain I–II linker identifies possible donor and acceptor sites. Exons are indicated by filled boxes and intron sequences by lines. (c) Comparison of the α_1A genomic and cDNA sequences in the domain IV S3–S4 region shows that a 162-bp intron separates two exons.

To determine the molecular nature of these differences, we sequenced clones from a rat genomic library and also analyzed rat DNA by PCR. Clones corresponding to the region flanking the α_1A I–II linker had an Asp preceding the Val₄₂₁ insertion ( Fig. 1b ). A 1.8-kb intron separated the Asp residue from downstream Gly₄₂₂ found in α_1A-a transcripts. The flanking genomic and cDNA sequences contained a common 5´ splice donor site and three possible 3´ acceptor sites: one that would result in either the inclusion or exclusion of Val₄₂₁, another that would include Gly₄₂₂, and a third that could produce an α_1A variant lacking both the Gly and Val₄₂₁ residues ( Fig. 1b ). To test this third possibility, we examined five additional α_1A cDNAs and found that one of five lacked both the Val₄₂₁ and Gly₄₂₂ residues (called α_1A-c). To confirm that the α_1A-c variant was expressed in vivo, we examined total cerebellar and hippocampal RNAs by RT-PCR followed by DNA sequencing. Of 81 PCR products, 16% were the α_1A-c variant, 79% the domain I–II-linker α_1A-a isoform, and 5% the α_1A-b isoform (data not shown).

PCR amplification of rat genomic DNA across the domain IV S3–S4 region identified a 162-bp intron that when absent resulted in α_1A channels lacking N₁₆₀₅-P₁₆₀₆ (α_1A-a variant), whereas inclusion of N₁₆₀₅-P₁₆₀₆ (the α_1Ab isoform) resulted from use of an alternative 5´ splice donor site ( Fig. 1c ). The genomic nature of the differences between α_1A-a and α_1Ab in the EF hand region were not determined, although the 10 amino-acid substitutions are contained in a contiguous stretch of 90 bp with only 53% sequence identity, suggesting a mutually exclusive alternative splicing mechanism.

Differential expression of α_1A isoforms in the rat CNS

The spatial expression patterns of α_1A-a and α_1Ab in the rat CNS were examined by RT-PCR of the domain I–II linker region and by in situ hybridization using antisense oligonucleotide probes against the distinct EF hand sequences. Stringent PCR conditions allowed the selective amplification of α_1A-b variants containing the 3-bp Val₄₂₁ insertion in the domain I–II linker ( Fig. 2a ). The expected α_1A-b 208-bp fragment was detected in all rat brain regions examined, indicating that α_1A transcripts containing Val₄₂₁ are widely expressed. Furthermore, transcripts containing Val₄₂₁ were also detected at moderate levels in kidney RNA ( Fig. 2a ). Attempts to amplify the 208-bp fragment from rat genomic DNA samples were unsuccessful, consistent with the presence of the intronic sequences in the α_1A I–II linker (see Fig. 1b ).

Figure 2: Both α_1A-a and α_1A-b variants are expressed in the rat nervous system.

(a) RT-PCR demonstrating the expression of α_1A-b transcripts. Blotting and hybridization of the PCR products with an internal oligonucleotide showed the specific amplification of an α_1A-b 208-bp fragment from the α_1A-b template cDNA control but not the α_1A-a cDNA. (b–e) In-situ hybridization to adult rat brain sections using radiolabeled oligonucleotide probes specific to the α_1A-a and α_1A-b EF hand regions. Sections show α_1A-b (b) and α_1A-a (c) in cerebellum and α_1A-b (d) and α_1A-a (e) in hippocampal CA1-CA3 pyramidal (p) and dentate granule cells (g).

In situ hybridization with the EF hand probes revealed differential expression of α_1A-a and α_1A-b transcripts. Most brain regions prominently expressed α_1A-b. Expression of α_1A-a was strong in the cerebellar cortex, with lower levels in the cortex, hippocampus and olfactory tubercule (not shown). Both α_1A-a and α_1A-b were detected in cerebellar granule cells, with α_1A-a predominantly in Purkinje cells ( Fig. 2b and c ). In contrast, α_1A-b was highly expressed in CA1-CA3 pyramidal and dentate granule cells in the hippocampus, where α_1A-a expression was significantly lower ( Fig. 2d and e ).

Transcripts encoding α_1A variants with N₁₆₀₅-P₁₆₀₆ contain an additional Tfi I restriction enzyme site. PCR amplication of rat cerebellar and hippocampal α_1A RNAs and digestion with Tfi I showed that both regions expressed both α_1A channel variants. However, in cerebellar RNA, most isoforms lacked N₁₆₀₅-P₁₆₀₆, whereas in hippocampus both forms were represented approximately equally (not shown).

α_1A splicing affects electrophysiological properties

The electrophysiological properties of α_1A-a and α_1A-b channels were compared by transient expression in Xenopus oocytes (co-expressed with rat brain α₂δ and β₄ subunits). The α_1A-b isoform had three significant functional differences from α_1A-a ( Fig. 3 ). First, α_1A-b inactivation rate was dramatically slowed, with 84 ± 3% (n = 12) of the whole-cell current remaining after a 400-ms test pulse compared to 39 ± 8% for α_1A-a (n = 8; Fig. 3a ). After a 16-s test pulse, approximately 50% of the α_1A-b current remained, whereas α_1A-a currents were completely inactivated ( Fig. 3b ). Second, α_1A-b showed a significant (p < 0.05) positive shift in the current–voltage relationship by ~6 mV (α_1A-a V_0.5 = –4.1 ± 0.5, n = 18; α_1A-b V_0.5 = 2.1 ± 0.7, n = 20; Fig. 3a ). Third, the voltage dependence of inactivation of α_1A-b was shifted more positively by ~20 mV. Furthermore, even after the holding potential was clamped to +50 mV for 15 s, ~60% of the whole-cell current was not inactivated ( Fig. 3c ).

Figure 3: α_1A-a and α_1A-b have distinct functional properties.

(a) Waveforms and current–voltage relationships for α_1A-a and α_1A-b. Waveforms of peak normalized current from a holding potential of –100 mV show that α_1A-b barely inactivates over 400 ms. The mean normalized I–V curves for α_1A-a (n = 18) and α_1A-b ( n = 20) show a positive shift for α_1A-b. (b) Normalized waveforms for α_1A-a and α_1A-b channels co-expressed with three different Ca²⁺ channel β subunits. Cells were depolarized from a holding potential of –100 mV to the peak I–V potential, and the currents remaining after 2 s or 16 s were averaged and presented as histograms (mean, n = 8–17 cells for each combination). (c) Steady-state inactivation curves for α_1A-a and α_1A-b co-expressed with β₄ and α₂. Cells were held at various potentials for 16 s before a test pulse to +10 mV. The half-inactivation potentials obtained from the fit were –17.2 ± 0.7 mV (n = 5) for α_1A-a and –1.6 ± 0.8 mV (n = 10) for α_1A-b.

Because co-expression of Ca²⁺ channel β subunits both alters channel kinetics and shifts the current–voltage relationship⁹, we asked whether these functional differences were specific for neuronal β subunit types. The inactivation rate of α_1A-b was significantly slower than α_1A-a for all β subunits tested ( Fig. 3b ), regardless of whether the β subunit normally increases (β_1b) or decreases (β_2a) inactivation rates. Furthermore, regardless of the co-expressed β subunit, the current–voltage relationship of α_1A-b was significantly (p < 0.05) more positive when compared to α_1A-a (not shown). The amino-acid differences in α_1A-b were not associated with significant changes in other channel properties examined, including ion permeation (not shown), the rate of activation (not shown) and the slope of the current–voltage relationship (α_1A-a, k = –3.5 ± 0.2, n = 18; α_1A-b, k = –3.7 ± 0.5, n = 20).

Val₄₂₁ and N₁₆₀₅-P₁₆₀₆ differentially affect properties

To identify the regions of α_1A-b responsible for its distinct gating characteristics, we constructed chimeras between α_1A-a and α_1A-b ( Fig. 4a ). The α_1A-a (+V) chimera was identical to α_1A-a except for the Val₄₂₁ insertion in the domain I–II linker, and the α_1A-b (–V) chimera differed from α_1A-a only by insertion of N₁₆₀₅-P₁₆₀₆ and substitution of the α_1A-b EF hand region. The α_1A-b (–V) chimera showed fast inactivation kinetics similar to those of α_1A-a (42 ± 3% remaining after 400 ms, n = 9) but retained a current–voltage relationship positively shifted by ~6 mV (V_0.5 = 2.7 ± 0.5, n = 19; Fig. 3a ). In contrast, the α_1A-a (+V) chimera had slow inactivation kinetics similar to that of α_1A-b (79 ± 3% remaining after 400 ms, n = 13) and a relatively negative current–voltage relationship similar to α_1A-a (V_0.5 = –5.0 ± 0.3, n = 21; Fig. 3a ).

Figure 4: Splicing of the domain I–II linker and domain IV S3–S4 loop make distinct contributions to channel kinetics and gating.

(a) Schematic representation of chimeric α_1A Ca²⁺ channels and their inactivation and activation properties in the presence of β₄ and α₂ (n = 9–21). Inactivation is measured as the percentage of whole-cell peak current remaining at the end of a 400-ms test pulse and activation as half-activation potentials obtained from fits to whole-cell current–voltage relationships. (b) Single-channel records obtained from cell-attached patch recordings from oocytes expressing α_1A-a with or without Val₄₂₁ (+ β₄ and α₂). Currents were elicited from a holding potential of –100 mV to a test potential of 0 mV. Bursts of activity separated by brief quiescent periods were observed in α_1A-a (+V). (c) Effect of Val₄₂₁ on the mean open time distributions within individual bursts. Without Val₄₂₁, the open time distribution was fitted with two exponentials with time constants of 0.34 ms (44%) and 1.33 ms (56%). With Val₄₂₁, time constants were 0.90 ms (35%), 3.26 ms (51%) and 15.74 ms (14%).

We examined two additional chimeras: α_1A-a (+NP), which was identical to α_1A-a except for the insertion of N₁₆₀₅-P₁₆₀₆, and α_1A-b (–NP), which was identical to α_1A-b except that it lacked N₁₆₀₅-P₁₆₀₆ ( Fig. 4a ). Inactivation properties of α_1A-a (+NP) were identical to those of α_1A-a, although V_0.5 was shifted similarly to α_1A-b (α_1A-a (+NP) = 3.3 ± 0.4, n = 11). In contrast, the α_1A-b (–NP) chimera had negatively shifted V_0.5 but slow inactivation kinetics typical of the wild-type α_1A-b (α_1A-b (–NP) = –3.9 ± 0.4, n = 12). Another chimera containing the α_1A-c variation in the domain I–II linker (α_1A-a missing both Val₄₂₁ and Gly₄₂₂) showed similar waveform and inactivation properties to α_1A-a and was not examined further in the present study (not shown). We could not detect any distinct contributions of the spliced EF hand regions to any of the Ca²⁺ channel properties examined in this study.

Presence or absence of Val₄₂₁ affects inactivation

Cell-attached, single-channel recordings were made in oocytes expressing either the α_1A-a or the α_1A-a (+V) chimera. At the single-channel level, the presence of Val₄₂₁ caused the appearance of multiple bursts of channel activity ( Fig. 4b ), which continued throughout test pulses as long as 24 s (not shown). Ensemble averages of sweeps with unitary activity showed a rapidly inactivating α_1A-a current, whereas α_1A-a (+V) averaged currents were sustained ( Fig. 4b ). Varying the holding potential from –100 to –20 mV completely inactivated α_1A-a channels but did not affect α_1A-a (+V) open probability, consistent with the lack of complete steady-state inactivation at the whole-cell level (not shown).

Comparison of the open-time distributions for α_1A-a and α_1A-a (+V) channels ( Fig. 4c ) shows that α_1A-a openings are characterized by two apparent brief open states (0.34 and 1.33 ms), whereas the α_1A-a (+V) distribution had three markedly longer open states (0.90, 3.26 and 15.74 ms) . Thus, Val₄₂₁ contributes to two separate effects: destabilization of the inactivated state, as indicated by the appearance of multiple bursts, and stabilization of the open state(s), as indicated by the increase in mean open time and appearance of an additional time constant. There was no significant change in the single-channel slope conductance of α_1A channels with or without Val₄₂₁ (not shown; α_1A-a, 16.5 ± 0.4 pS, α_1A-a(+V), 17.2 ± 0.7 pS; p < 0.05, n = 4).

N₁₆₀₅-P₁₆₀₆ reduces ω-agatoxin IVA sensitivity

The major distinguishing characteristic between native P- and Q-type calcium currents is their differential sensitivity to the spider peptide toxin ω-Aga IVA^{8,16,17,18,19,19}. In Xenopus oocytes, α_1A-a and α_1A-b channels were blocked approximately equally by 200 nM ω-Aga IVA (~25% block) and by 5 μM ω-CTx-MVIIC (~80% block; Fig. 5a ). In contrast to native P-type channels, ω-Aga IVA block was not reversed by application of depolarizing prepulses (data not shown). However, in transfected human embryonic kidney (HEK 293) cells, 100 to 300 nM ω-Aga IVA completely blocked the wild-type α_1A-a ( Fig. 5b and c ). As for native P-type channels, block was relieved by a series of strong depolarizations ( Fig. 5b ). The sensitivity of α_1A-a to ω-Aga IVA was markedly reduced in channels containing N₁₆₀₅-P₁₆₀₆ in domain IV S3–S4 ( Fig. 5c ). The N₁₆₀₅-P₁₆₀₆ insertion lowered the blocking rate constant, k_on, 7-fold (α_1A-a = 6.475 × 10^–3 μM^–1s^–1; α_1A-a (+NP) = 0.9198 × 10^–3 μM^–1s^–1) and increased k_off 2-fold (α_1A-a = 1.47 × 10^–3 s^–1; α_1A-a (+NP) = 3.045 × 10^–3 s^–1), resulting in an 11-fold decrease in affinity for ω-Aga IVA ( Fig. 5d and e ; K_d for α_1A-a = 14.8 nM; K_d for α_1A-a (+NP) = 167 nM). In contrast to ω-Aga IVA, the snail peptide toxin ω-CTx-MVIIC blocked both α_1A-a and α_1A-a (+NP) with a similar time course and potency ( Fig. 5f ). These pharmacological profiles in HEK cells suggest that native P-type channels lack N₁₆₀₅-P₁₆₀₆, whereas Q-type channels contain these two residues.

Figure 5: Sensitivities of α_1A-a and α_1A-a (+NP) to ω-Aga-IVA and ω-CTx-MVIIC.

Whole-cell current records obtained from α_1A-a and α_1A-b (expressed transiently in Xenopus oocytes) after application of 200 nM ω-Aga-IVA or 6 μM ω-CTx-MVIIC. Mean ± s.e.; n = 5–8. (b–f) Channels expressed transiently in HEK cells. (b) Time course of development of ω-Aga IVA (300 nM) block of α_1A-a and reversal of toxin action with repetitive depolarizations (10 × 10 ms steps to +120 mV at 1 Hz). Test pulses to 0 mV were elicited from a holding potential of –100 mV every 15 seconds. (c) Whole-cell current records obtained from α_1A-a and α_1A-a (+NP) with or without 100 nM ω-Aga IVA. (d) Average normalized time course comparing the rate of development of ω-Aga IVA (300 nM) block of α_1A-a and α_1A-a (+NP). Inset, on-rate of block for α_1A-a and α_1A-a (+NP) varies linearly with ω-Aga IVA concentration. Mean data (± s.e.; n = 3–5) were fit according to the equation 1/τ_{on =}k_on [ω-Aga IVA] + k_off. The K_d values were 14.8 nM for α_1A-a and 167 nM for α_1A-a (+NP). (e) Dose dependence of the effects of ω-Aga IVA for α_1A-a and α_1A-a (+NP). The solid lines reflect a fit with the Hill equation with the Hill coefficient arbitrarily held at 1.0. The IC₅₀ values obtained were 16.3 nM for α_1A-a and 146 nM for α_1A-a (+NP). (f) Time course of development of ω-CTx-MVIIC (6 μM) block of α_1A-a and α_1A-a (+NP). Inset, blocking rate constants obtained for α_1A-a and α_1A-a (+NP). Test pulses to +10 mV were elicited from a holding potential of –100 mV every 15 seconds.

G proteins and PKC modulate α_1A splice variants

The activation of certain neurotransmitter receptors inhibits P/Q-type Ca²⁺ channels through the binding of G-protein βγ subunits directly to the α_1A domain I–II linker³². G-protein modulation of Ca²⁺ channels characteristically involves decreased whole-cell currents, slowed activation and inactivation kinetics, and relief from inhibition by application of positive prepulses (called facilitation)³³. Because Val₄₂₁ is located in one of two G_βγ binding sites of the I–II linker and may directly alter G_βγ binding affinity, we tested whether this residue affects the G-protein-dependent modulation of α_1A channels. The μ-opioid receptor agonist DAMGO inhibited both α_1A-a and α_1A-b currents in cells co-expressing the μ-opioid receptor ( Fig. 6 ). However, DAMGO-inhibited α_1A-b activation kinetics were significantly more slowed and the overall opioid-induced inhibition more pronounced compared with α_1A-a ( Fig. 6a ). Times to peak indicated that Val₄₂₁ is responsible for the slowed kinetics of α_1A-b after G-protein activation ( Fig. 6a ). In the presence of DAMGO, the two splice variants also showed distinct magnitudes of voltage-dependent facilitation as determined by applying a 50-ms prepulse to +150 mV before a test pulse from –100 mV to + 10 mV. When measured 20 ms after the test pulse, the α_1A-b variant showed twofold more facilitation (α_1A-b, 22 ± 3%, n = 15; α_1A-a, 10 ± 2%, n = 14; data not shown). Furthermore, varying the time between the prepulse and the test pulse showed that the α_1A-a-facilitated currents were re-inhibited more rapidly compared to α_1A-b ( Fig. 6b ; τ reinhibition α_1A-a = 25 ± 2 ms, n = 15; α_1A-a(+V), 39 ± 3, n = 14). The difference in re-inhibition kinetics following DAMGO-induced G-protein inhibition is due to Val₄₂₁ in the domain I–II linker ( Fig. 6b ).

Figure 6: Differential modulation of α_1A splice variants by G proteins and PKC.

(a, b) Modulation of α_1A-a and α_1A-b by G proteins through the activation of a co-expressed μ-opioid receptor. (a) Currents elicited from a holding potential of –100 mV to a test potential of +10 mV with or without 1 μM DAMGO. Time-to-peak measurement shows that Val₄₂₁ is responsible for the slower kinetics of α_1A-b following μ-opioid receptor activation. (b) Time course of reinhibition of the channels by G-proteins after temporary relief with a strong depolarizing prepulse to +150 mV. The presence of Val₄₂₁ slows the reinhibition kinetics, suggesting a lower affinity of the channel for G_βγ. (c, d) Activation of PKC-dependent phosphorylation by application of 100 nM PMA. PKC-dependent upregulation of α_1A-a and α_1A-b follows a similar time course (c), but the presence of Val₄₂₁ causes a twofold increase in the upregulation (c, d). In (a), α_1A was co-expressed with β₄ and α₂ in Xenopus oocytes. In (b–d), α_1A was co-expressed with β₄ and α₂ in HEK 293 cells.

The stimulation of protein kinase C (PKC) with phorbol esters upregulates α_1A-a currents by ~10% ( Fig. 6c ). Because the domain I–II linker contributes to the PKC-dependent upregulation of some neuronal calcium channels³⁴, we compared the modulation of α_1A-a and α_1A-b variants. Phorbol ester (PMA, 100 nM) caused twofold more upregulation of α_1A-b channels than of α_1A-a (+PMA α_1A-a, 1.10 ± 0.01, +PMA α_1A-b, 1.24 ± 0.03; Fig. 6c ). The presence of Val₄₂₁ in the domain I–II linker was responsible for conferring the higher degree of upregulation associated with α_1A-b channels ( Fig. 6d ). Overall, the results support the notion that the domain I–II linker is a crucial determinant for PKC and G_βγ modulation³² and suggest alternative splicing as a mechanism for controlling the second-messenger-dependent modulation of neuronal Ca²⁺ channels.

Discussion

Alternative splicing alters channel properties

Our present understanding of Ca²⁺ channel structure–function relationships primarily results from chimeric cDNA and mutagenesis studies, which have identified regions of Ca²⁺ channel α₁ subunits important for permeation³⁵, activation³⁶ and inactivation³⁷, excitation–contraction coupling³⁸ and β subunit binding³⁰. The functional effect of alternative splicing is an important tool because it provides insights into native amino acids essential for controlling channel properties.

Here we show that alternative splicing of the rat α_1A gene results in variants with relatively small alterations in the domain I–II linker and domain S3–S4 regions. Two apparently distinct channel properties are affected by alternative splicing of N₁₆₀₅-P₁₆₀₆ in domain IV 53-54. The approximately 6-mV shift in current–voltage relationships associated with N₁₆₀₅-P₁₆₀₆ is consistent with studies on L-type Ca²⁺ channels showing that mutations both within and flanking S3 segments affect the voltage dependence of activation³⁶. Many native Ca²⁺ channels display variable activation characteristics, which may generally result from alternative splicing in S3–S4 regions.

The second major change in channel properties associated with N₁₆₀₅-P₁₆₀₆ is to alter the sensitivity of α_1A to the spider peptide toxin, ω-Aga IVA. ω-Aga IVA is a 48-residue positively charged peptide isolated from the venom of the funnel web spider Agelenopsis aperta, which has long been used as a diagnostic for native P/Q-type Ca²⁺ channels^16,17. As in native P-type currents, block of both α_1A-a and α_1A-b variants by ω-Aga IVA in HEK cells was highly voltage sensitive and was reversed by application of positive prepulses. The major effects of the insertion of N₁₆₀₅-P₁₆₀₆ were to decrease k_on and increase k_off, which together decreased toxin affinity 11-fold. Both ω-Aga IVA and ω-CTx MVIIC act extracellularly to block P/Q-type channels, although by distinct blocking mechanisms; ω-CTx MVIIC acts as a traditional pore blocker (S.I. McDonough, I.M. Mintz & B.P. Bean Soc. Neurosci. Abstr. 21, 140.9, 1995) and ω-Aga IVA binds to the gating machinery to stabilize channel closed states³⁹. In contrast to ω-Aga-IVA, the presence or absence of N₁₆₀₅-P₁₆₀₆ did not alter the block of α_1A channels by ω-CTx MVIIC, which is consistent with these toxins' different mechanisms of action.

ω-Aga IVA belongs to a class of polypeptide neurotoxins that act via a common mechanism to alter the gating properties of voltage-gated ion channels. These include α and β scorpion toxins (Na⁺ channels)^40,41, hanatoxin (K⁺ channels)⁴² and grammotoxin (Ca²⁺ channels)⁴³. Our data are consistent with studies showing that peptide gating modifiers interact at a conserved structural gating modifier site that primarily includes S3–S4 loops. Single amino-acid changes in the domain II S3–S4 loop of Na⁺ channels and the S3–S4 loop of K⁺ channels significantly alter toxin affinities^40,41,42,43. In this regard, the insertion of N₁₆₀₅-P₁₆₀₆ in α_1A by splicing may alter the conformation of the domain S3–S4 loop and result in an unfavorable binding site for ω-Aga IVA.

The third change in α_1A properties associated with alternative splicing is that the presence of Val₄₂₁ dramatically slowed channel inactivation kinetics and affected steady-state inactivation. The Val₄₂₁ residue is close to the site in the domain I–II linker that interacts with Ca²⁺ channel β subunits³⁰. However, Val₄₂₁ is unlikely to affect inactivation by disrupting β subunit binding because all β subunits still shifted α_1A current–voltage properties. Furthermore, co-expression with the β_2a subunit, which normally slows inactivation, caused an additional significant slowing of α_1A-b currents. The presence of Val₄₂₁ altered channel gating as reflected by these channels reopening throughout the depolarization. Assuming that Val₄₂₁ alters the prevalence of different gating modes, these channels might enter an alternate mode with an increased rate constant for returning to the open state from the inactivated state. This less-stable inactivated state would be reflected in reopening during the depolarization.

The demonstration that the domain I–II linker contributes to Ca²⁺ channel inactivation⁴⁴ ( Fig. 4 ) is distinct from studies of chimeric Ca²⁺ channels that suggested a predominant role for domain I in inactivation³⁷. Although we did not detect any functional properties that could be attributed to alternative splicing of the EF hand region, it is unlikely that the specific splicing of this highly conserved structural motif is benign. Some possibilities include the interaction of this putative divalent ion binding site with intracellular cations, or the differential interaction at this site of yet-to-be identified proteins that bind to the Ca²⁺ channel carboxyl terminus.

Differential expression of α_1A splice variants

P-type currents originally described in cerebellar Purkinje cells can be distinguished from Q-type currents in cerebellar granule cells by their higher sensitivity to ω-Aga-IVA and non-inactivating waveform^{15,16,17,18,19}. We showed that α_1A channels can exhibit the slower inactivation kinetics typically associated with P-type currents by associating with a co-expressed β_2a subunit, possessing Val₄₂₁ in the domain I–II linker, or both ( Fig. 2b ). In contrast, α_1A channels lacking Val₄₂₁ or associated with β₁, β₃ or β₄ subunits exhibit the faster inactivation kinetics typically associated with Q-type currents ( Fig. 2b ). The β_2a subunit is highly expressed in cerebellar Purkinje cells (data not shown); therefore the non-inactivating P-type currents may be encoded by the combination of α_1A-a and β_2a (and α₂δ).

α_1A transcripts both with and without N₁₆₀₅-P₁₆₀₆ in domain IV S3–S4 were expressed at approximately equal levels in the hippocampus. In contrast, most α_1A transcripts in the cerebellum lacked N₁₆₀₅-P₁₆₀₆. The presence of variants that confer the lower Q-type and higher P-type affinities for ω-Aga IVA ( Fig. 5 ) predicts that both P-type and Q-type channels are widely expressed in the hippocampus, as reported in CA1-CA3 cells^17,20. The data also predict that most α_1A transcripts in the cerebellum would have the P-type pharmacological profile, as reported in Purkinje cells²¹. The more abundant cerebellar granule cells also highly express an inactivating Ca²⁺ current (called G1) with P-type pharmacology²¹, although some controversy remains regarding the nature of Q-type currents in granule cells. The Q-type current originally described in cultured cerebellar granule cells was reported to represent up to 35% of the whole-cell Ca²⁺ current¹⁹. However, others²¹ failed to identify a Q-type current in this same primary granule cell preparation, but instead described the G1 current. One possible explanation for these conflicting results is that variations in culture conditions differentially induce expression of α_1A-a and α_1A-b splice variants in cultured granule cells.

ω-Aga-IVA affinity and expression system considerations

In Xenopus oocytes, the currents from cloned α_1A variants lacking N₁₆₀₅-P₁₆₀₆ (α_1A-a) seem pharmacologically similar to native Q-type currents (K_d for ω-Aga-IVA > 100 nM; Fig. 5a )^29,30. However, in HEK 293 cells, the K_d for ω-Aga IVA block of α_1A-a was approximately 15 nM. In another study⁴⁵, the identical rat brain α_1A-a cDNA expressed in COS cells was found to have a K_d of ~11 nM. This 15–20-fold difference in ω-Aga IVA sensitivity between mammalian cells and Xenopus oocytes may reflect cell-type-specific proteins that interact directly or indirectly with α_1A-a and alter pharmacological properties. Alternatively, cell-type-specific post-translational processing might alter channel affinity for ω-Aga IVA. For example, cloned acetycholine receptors (nAChR) expressed in Xenopus oocytes are incorrectly glycosylated compared to native nAChRs in Torpedo electroplax⁴⁶. Because ω-Aga IVA is highly positively charged, the amounts and types of oligosaccharides (for example, high mannose versus complex) in the vicinity of the binding site may significantly affect toxin binding. Small differences in glycosylation patterns between mammalian cell lines and cerebellar Purkinje cells might also account for the difference in K_d between exogenously expressed α_1A-a (K_d ≈ 11–15 nM) and native P-type currents (K_d ≈ 2 nM).

Alternative splicing: physiological consequences

It is important to gain an understanding of the contributions of α_1A channels not only to normal neuronal functions, such as neurotransmitter release^5,6,7,820, but also to pathological states such as migraine, ataxia and epilepsy^27,28,29. Immunohistochemical staining localizes α_1A calcium channels postsynaptically on cell bodies and dendrites as well as on presynaptic terminals⁴⁷. In nerve terminals, both the 6-mV shift in α_1A channel activation resulting from the splicing of N₁₆₀₅-P₁₆₀₆ and the slowed inactivation kinetics associated with Val₄₂₁ would be predicted to alter the efficacy of synaptic transmission and significantly affect excitability.

Certain G-protein-coupled receptors inhibit P/Q-type and N-type channels by direct binding of G_βγ to the domain I–II linker of α_1A and α_1B subunits³². The alternatively spliced Val₄₂₁ residue is within one of two G_βγ binding sites in the I–II linker, and its presence significantly increases slowing of the kinetics of G-protein-dependent inhibition and increases the overall degree of block ( Fig. 6a and b ). For both α_1A-a and α_1A-b, the kinetics of G-protein re-inhibition following a positive prepulse were well described by a single exponential. However, the time constant for re-inhibition was twice as long for α_1A-b (and α_1A-a + Val₄₂₁) as for α_1A-a. Within the few milliseconds of an action potential, the G-protein-inhibited α_1A-b channels with their very slow activation ( Fig. 6a ) would be more effectively inhibited, and their ability to trigger fast neurotransmitter release might be attenuated compared with α_1A-a. Similar to strong step depolarizations, bursts of action potentials can relieve G-protein inhibition⁴⁸. Thus, the slower inhibition of α_1A-b channels would be expected to result in a substantially larger amount of calcium influx compared to α_1A-a (see Fig. 6b ) and might cause more pronounced synaptic efficiency for slow neurotransmitter release.

P-, Q- or other types?

The four different types of Ca²⁺ channel β subunits are widely expressed in the rat CNS and also have subtype-specific expression patterns. Because each of the β subunits can differentially modify α_1A inactivation kinetics²⁵, inactivation kinetics are a poor diagnostic of Ca²⁺ channel subtype without knowledge of biochemical composition. Similarly, we find significant differences in the sensitivity of α_1A channels to ω-Aga IVA based on alternative splicing and the cell type used for expression. Thus, ω-Aga IVA affinity might not always be a reliable diagnostic of specific α_1A variants.

It is important to emphasize that the electrophysiological and pharmacological properties of the α_1A-a and α_1A-b variants described here do not exactly match those of either prototypical P-type or Q-type currents. Indeed, α_1A-a and α_1A-b describe two types of Ca²⁺ channels that exhibit some characteristics of both P-type and Q-type currents. Native Ca²⁺ channels with properties intermediate between P- and Q-type have been described in neurons and endocrine cells^21,22,49. For example, the cerebellar granule cell G1 Ca²⁺ channel is highly sensitive to ω-Aga-IVA (similar to P-type) yet exhibits substantial inactivation (similar to Q-type)²¹. We predict that G1 is likely to be encoded by a splice variant missing both Val₄₂₁ and N₁₆₀₅-P₁₆₀₆ (for example, α_1A-a).

The α_1A-a and α_1A-b variants examined here were each derived from single cDNAs; thus their properties likely reflect some portion of native α_1A Ca²⁺ currents. In addition, other α_1A variants are expressed that exhibit both some combination of the alternatively spliced regions examined here as well as other independent alternatively spliced sequences^{23,27,29,49,50}. Assuming a mutually exclusive splicing mechanism, our data suggest that Q-type channels represent only one possible phenotypic variant resulting from splicing of a single α_1A gene encoding the P-type Ca²⁺ channel.

Methods

Electrophysiological recording.

Preparation of Xenopus oocytes, nuclear injection with Ca²⁺ channel subunit cDNAs, two-microelectrode voltage clamp, and cell-attached single-channel, patch-clamp recordings were done as described^25,34. For macroscopic currents, the bathing medium contained 40 mM Ba(OH)_2, 25 mM TEA-OH, 25 mM NaOH, 2 mM CsOH and 5 mM HEPES (titrated to pH 7.3 with methane-sulfonic acid). During voltage-clamp recordings, oocytes were routinely injected with BAPTA to eliminate the endogenous oocyte Ca-dependent chloride current. For G-protein modulation, Ca²⁺ channels were co-expressed with the μ-opioid receptor and G-protein activation was triggered via application of 1 μM DAMGO. For PKC modulation, oocytes were perfused with 100 nM PMA as described³⁴, and BAPTA was omitted.

For the determination of current–voltage relationships, curves were fitted according to the equation I = G × (V – E)/[1 + exp((V –V_0.5)/k)], where I is the current at a given voltage V, G is the conductance, E is the extrapolated reversal potential, V_0.5 is the voltage for half activation and k is the slope factor. For steady-state inactivation, cells were held at various potentials for 15 s and currents recorded during a subsequent test pulse to the peak potential of the I–V. Steady-state inactivation was calculated with a Boltzmann equation and the potential for half inactivation determined from the best fit. All values are given as mean ± s.e. Leak currents and capacitive transients were subtracted using a P/5 procedure.

Transient transfection of human embryonic kidney cells (HEK-tsa201) was done by standard calcium-phosphate precipitation, and whole-cell patch-clamp recordings were as described⁴⁸. The external recording solution was 5 mM BaCl_2, 1 mM MgCl₂, 10 mM HEPES, 40 mM TEACl, 25 mM glucose and 65 mM CsCl (pH 7.2). The internal pipette solution was 105 mM CsCl, 25 mM TEACl, 1 mM CaCl₂, 11 mM EGTA and 10 mM HEPES (pH 7.2). Peptide toxins were perfused onto the cells using a gravity-driven microperfusion system.

Molecular biology.

To identify the α_1A-b cDNA, we screened a rat brain cDNA library²⁴ with a [³²P]deoxyoligonucleotide probe homologous to domain I of previously cloned neuronal α₁ subunits. Of seventeen clones identified as α_1A types, the complete DNA sequence of the single 7.2 kb α_1A-b cDNA was determined on both strands. The α_1A-b (–V) construct was made by substituting a 2.4-kb Xho I/Nhe I fragment of α_1A-a into α_1A-b. The α_1A-a(+V) construct was made by substituting a 1.8-kb Tth III1 fragment of α_1A-b into α_1A-a. The α_1A-a (+NP) cDNA was constructed by substituting a 2.5-kb BamHI fragment from α_1A-b into α_1A-a. The α_1A-b (–NP) construct was created by switching the 2.5-kb fragment from α_1A-a into α_1A-b. Chimeric cDNAs were originally constructed in the plasmid Bluescript and then subcloned into the Xho I/Spe I sites of the nuclear expression vector, pMT2. Each construct was verified by DNA sequencing across the chimera junctions and through the altered region.

To examine domain I–II linker splice junctions, we screened a rat genomic library (lambda DASH, Stratagene) using a [³²P]-labeled cDNA fragment to the α_1A-a domain I–II linker. A 5.2-kb EcoRI fragment that hybridized to oligonucleotide probes specific for α_1A sequences at positions 1200 and 1275 bp was selected and DNA sequencing used to identify the intron–exon boundaries across the I–II linker. Splicing in domain IV was analyzed by using PCR to amplify rat genomic sequences flanking S3 and S4. Direct sequencing of the PCR products identified the intron–exon boundaries.

For PCR detection of V₄₂₁, a sense oligonucleotide common to both α_1A-a and α_1A-b variants was used in conjunction with an antisense oligonucleotide containing the antisense sequence for Val₄₂₁ at its 3´ end. In situ localization was done using 30 μm paraformaldehyde-fixed sections from 250 to 300 gm adult male rat brains. Sections were hybridized overnight at 37°C in 50% formamide, 0.6 M NaCl, 1 mM EDTA, 100 mM DTT, 10 mM Tris pH 7.5, 1× Denhardts, 10% dextran sulphate, 0.1% SDS, 0.1% sodium thiosulphate, 0.05% denatured salmon sperm DNA and 0.05% yeast tRNA. Probes consisted of [α-³⁵S]dATP end-labeled antisense 39-base synthetic deoxyoligonucleotides derived from the unique EF hand regions of α_1A-a and α_1A-b. Sections were washed at 50°C in 0.5× SSC, 1% sodium thiosulphate and 14 mM β-mercaptoethanol, exposed to X-ray film for 1 week, then dipped in Kodak photographic emulsion NTB-2 and exposed for 2–4 weeks.