Alternative Splicing as a Molecular Switch for Ca²⁺/Calmodulin-Dependent Facilitation of P/Q-Type Ca²⁺ Channels

Abstract

Alternative splicing of the P/Q-type channel (Ca_V2.1) promises customization of the computational repertoire of neurons. Here we report that concerted splicing of its main α_1A subunit, at both an EF-hand-like domain and the channel C terminus, controls the form of Ca²⁺-dependent facilitation (CDF), an activity-dependent enhancement of channel opening that is triggered by calmodulin. In recombinant channels, such alternative splicing switches CDF among three modes: (1) completely “ON” and driven by local Ca²⁺ influx through individual channels, (2) completely “OFF,” and (3) partially OFF but inducible by elevated global Ca²⁺ influx. Conversion from modes 1 to 3 represents an unprecedented dimension of control. The physiological function of these variants is likely important, because we find that the distribution of EF-hand splice variants is strikingly heterogeneous in the human brain, varying both across regions and during development.

Introduction

Alternative splicing can vastly diversify the function of single genes, potentially tailoring molecular phenotype to local environment (Modrek and Lee, 2002). In the brain, this diversity can impact the behavior of neural systems (Grabowski and Black, 2001). Within individual neurons, splicing is amply documented for molecules constituting major Ca²⁺ entry pathways, such as the Ca_V1-2 family of Ca²⁺ channels (Lipscombe et al., 2002; Lipscombe and Castiglioni, 2003). In particular, P/Q-type (Ca_V2.1) Ca²⁺ channels, the predominant triggers of CNS neurotransmitter release (Wheeler et al., 1994; Dunlap et al., 1995; Wheeler and Tsien, 1999), possess principal α_1A subunits that are spliced extensively (Mori et al., 1991; Starr et al., 1991; Ophoff et al., 1996; Zhuchenko et al., 1997; Bourinet et al., 1999; Soong et al., 2002; Tsunemi et al., 2002). Splicing at certain loci has been seen to impact P/Q-type channel pharmacology and gating (Bourinet et al., 1999), as well as expression (Soong et al., 2002). However, the functional consequences of variation have only been examined at a restricted subset of sites (Fletcher et al., 1996; Ligon et al., 1998; Bourinet et al., 1999; Hans et al., 1999; Krovetz et al., 2000; Toru et al., 2000; Soong et al., 2002). In numerous cases, the potential neurobiological impact of P/Q-type channel splice variation remains uncharted.

This study concerns the potential for splicing of P/Q-type channels to affect their regulation by calmodulin (CaM). CaM exerts two opposing effects, initially promoting Ca²⁺-dependent facilitation (CDF) of channel opening and then inducing Ca²⁺-dependent inactivation (CDI) of opening (Lee et al., 1999; DeMaria et al., 2001). CDF and CDI rapidly alter the amplitude of recombinant P/Q-type Ca²⁺ currents evoked by repetitive physiological stimuli (Lee et al., 2000; DeMaria et al., 2001). These regulatory processes may thus contribute to short-term synaptic plasticity (Borst and Sakmann, 1998; Cuttle et al., 1998; Forsythe et al., 1998; Borst and Sakmann, 1999) and thereby the computational potential of the brain (Abbott et al., 1997; Tsodyks and Markram, 1997; Tsodyks et al., 1998).

Of particular interest are α_1A splice variants within structural determinants of CaM-channel regulation. Specifically, splicing of exon 37 produces two variants of an EF-hand-like domain (see Fig. 1, EFa and EFb) (Zhuchenko et al., 1997; Bourinet et al., 1999; Krovetz et al., 2000; Soong et al., 2002), a region implicated by mutagenesis as critical for CaM regulation of related L-type (Ca_V1.2) channels (de Leon et al., 1995; Zuhlke and Reuter, 1998; Peterson et al., 2000; Kim et al., 2004). Unlike classic EF hands (Tufty and Kretsinger, 1975), the L-type EF hand (Babitch, 1990) might not bind Ca²⁺ but may instead transduce Ca²⁺-CaM binding into channel modulation (Peterson et al., 2000; Kim et al., 2004). Because sequence alignments of the corresponding segments of the L- and P/Q-type channels clearly suggest a common heritage (de Leon et al., 1995), could splicing of the analogous α_1A EF-hand-like region impact CaM regulation of P/Q-type channels? Here, we find that such EF-hand variation acts as a molecular switch for CDF while leaving CDI unaffected. Additionally, we show that the distribution of α_1A EF-hand splice variants is highly variable across the human brain in both space and time, suggesting widespread usage of this molecular switch.

Figure 1.

Structural determinants underlying CaM regulation of P/Q-type Ca²⁺ channels. The diagram depicts major elements of the P/Q-type channel carboxyl tail implicated in CaM-channel interactions underlying channel regulation. Such interactions span an IQ-like domain (IQ), a ∼70 aa region upstream of the IQ (pre-IQ), and a putative CaM-binding domain (CBD) (Lee et al., 1999; DeMaria et al., 2001; Erickson et al., 2001; Lee et al., 2003). The IQ-like domain is particularly important in the CDF of channel opening. Our results here show that two additional regions, an EF-hand-like region (EF-hand) within exon 37 and the last exon (exon 47), can each modify CDF. Splice variants at these sites have been identified (Soong et al., 2002), and amino acid sequences and corresponding nomenclature for these variants are detailed in the boxes below the diagram. Differences between the EFa and EFb splice variants are in bold.

Materials and Methods

Transient transfection. Methods were adapted from previous studies (DeMaria et al., 2001). To generate functional P/Q-type channels, we cotransfected plasmids encoding the human α_1A (α₁2.1) pore-forming subunit and β_2a and α₂δ accessory subunits, along with the simian virus 40 T antigen, into human embryonic kidney 293 (HEK293) cells by a calcium phosphate protocol (Dhallan et al., 1990). The β_2a subunit reduces voltage inactivation, allowing a closer examination of Ca²⁺-dependent processes. The plasmid encoding the β_2a subunit also encoded a green fluorescent protein sequence that followed an internal ribosomal entry site. This construct thus allowed successfully transfected HEK293 cells to be identified by fluorescence. The T antigen permits plasmid replication in HEK293 cells, thereby enhancing expression of products encoded by transfected plasmids. The α₁2.1 clone was a gift from Dr. Terry Snutch (University of British Columbia, Vancouver, Canada) (Sutton et al., 1999), and its splice variant content has been engineered to the following (Soong et al., 2002): Δ10A (+G); 16⁺/17⁺; 17 (-VEA); -31 ^*(-NP); 37a (EFa) or 37b (EFb); 43⁺/44^-; Δ47 or 47+, where +G, -VEA, and -NP refer to splice variants with the inclusion (+) or omission (-) of amino acids specified by the single-letter code.

Electrophysiology. Two to 4 d after transfection of HEK293 cells, whole-cell current traces were recorded at room temperature. To examine CDI and CDF, we used a bath solution containing (in mm): 140 TEA-MeSO₃, 10 HEPES, and 5 CaCl₂ or BaCl₂, pH 7.4 (300-310 mOsm). We used the following internal solution (in mm): 135 Cs-MeSO₃, 5 CsCl₂, 0.5 EGTA, 1 MgCl₂, 4 MgATP, and 10 HEPES, pH 7.4 (290-300 mOsm). For certain experiments (see Fig. 6), we used a 10 mm concentration of the fast Ca²⁺ chelator BAPTA instead of 0.5 mm EGTA. All reagents were purchased from Sigma (St. Louis, MO).

Figure 6.

Current-density-dependent increase in CDF in [EFb/Δ47] channels. A-E, Plot of CDF strength (g) versus Ca²⁺ current density for the splice variant indicated at the top of each panel (top). Each symbol corresponds to an individual cell. Exemplar Ca²⁺ traces are displayed from cells with large (middle) or small (bottom) current densities, except in D and E. Shown for each splice variant are 5 mV test-pulse currents in the absence (light trace) or presence (dark trace) of a 20 mV prepulse. In D and E, a single exemplar (middle) and summary RF relationships (bottom) are depicted. g values for exemplar cells are shown with traces (middle and bottom). A, Although cells expressing [EFa] channels all show robust CDF, g does not correlate with current density, consistent with CDF driven exclusively by local Ca²⁺ entry through individual channels. B, [EFb/47+] channels show negligible CDF across all current densities. C, For cells expressing [EFb/Δ47] channels, only those with large current densities manifest appreciable CDF. Smooth curve drawn by eye. D, CDF is preserved in cells expressing [EFa/Δ47] channels even with increased intracellular Ca²⁺ buffering (10 mm BAPTA). Preservation of CDF is evident from the unchanged RF relationships for Ca²⁺ and Ba²⁺ (bottom) compared with cells recorded with minimal Ca²⁺ buffering (0.5 mm EGTA) (Fig. 4B). Data in D are reformatted from previously published data (Soong et al., 2002). E, Elevated intracellular Ca²⁺ buffering (10 mm BAPTA) eliminates CDF of [EFb/Δ47] channels (for current densities >45 pA/pF, p < 0.05 compared with [EFb/Δ47] channels recorded in 0.5 mm EGTA). Strong attenuation of CDF is confirmed by near superposition of averaged RF relationships for Ca²⁺ and Ba²⁺ (bottom). For reference, the top-most curve reproduces the corresponding Ca²⁺ RF relationship for [EFb/Δ47] channels during weak buffering of intracellular Ca²⁺ (0.5 mm EGTA, from Fig. 5B). Dotted lines show zero levels here and throughout.

Currents were recorded and filtered at 2 kHz bandwidth using an Axopatch 200A (Axon Instruments, Foster City, CA) amplifier and custom data-acquisition programs written by our laboratory. Data were analyzed off-line using custom programs written in Matlab (Math-Works, Natick, MA) and Excel (Microsoft, Seattle, WA). For statistical analysis, we used the t test, applying the Bonferroni correction when comparing across multiple groups (Glantz and Slinker, 1990).

Transcript scanning. According to previously described methods (Soong et al., 2002), we used PCR to amplify the exons flanking the EF hand from the following eight human cDNA libraries: whole brain (7187-1; Clontech, Palo Alto, CA), fetal brain (7129-1), amygdala (7190-1), cerebellum (7120-1), cerebral cortex (7110-1), hippocampus (7169-1), substantia nigra (7193-1), and thalamus (7188-1). The forward primer hEF5580U20 (5′-GGGAAACCGTGTGATAAGAA-3′) and the reverse primer hEF5975L15 (5′-GCTGTGCGGATCAGA-3′) were used for the amplification of a 410 bp fragment from the cDNA libraries. The reaction mix consisted of a 200 μm concentration of deoxy-ribose-NTPs (dNTPs), a 100 nm concentration of primers, 2.5 U of Taq DNA polymerase (Roche, Hertfordshire, UK), 1× PCR buffer with Mg²⁺ (Roche), and 0.5 μl of QUICK-clone human cDNA (Clontech) in a final volume of 25 μl. The cycling profile was set as follows: initial denaturation of template DNA at 94°C for 2 min, five touch-down cycles of denaturation at 94°C for 30 sec, annealing at 58-54°C for 40 sec, and extension at 72°C for 1 min, followed by 30 cycles of denaturation at 94°C for 30 sec, annealing at 53°C for 40 sec, and extension at 72°C for 1 min. The final extension was performed at 72°C for 10 min.

The amplicon represents the pool of both EFa and EFb cDNA forms. The fragment was gel purified, cloned into pDrive (Qiagen, Hilden, Germany), and transformed into Escherichia coli DH10B cells by electroporation. We identified clones by blue-white selection and then cultured them in 96 well microtiter plates. In general, a number of plates of white transformants were collected for each cDNA library.

Next, nested PCR was performed directly on the clones using an exon 36-specific forward primer, hEF5702U18 (5′-CGTCATCATGGACAACTT-3′), and either the exon 37a-specific primer EF1a.2-L20 (5′-ATATTACTCGTAATAAACTG-3′) or the exon 37b-specific primer EF1b.1-L20 (5′-GGGCGGAGACATGTGTCTCA-3′). The expected sizes of the EFa-specific and EFb-specific amplicons are 154 and 175 bp, respectively. The reaction mix consisted of a 50 μm concentration of dNTPs, a 50 nm concentration of each primer, 0.625 U of Taq polymerase (Promega, Madison, WI), 1× PCR buffer (Promega), 2 μm (for EFa screening) or 1 μm (for EFb screening) MgCl₂, and 1 μl of culture in a final volume of 25 μl. The cycling profile was set as follows: initial denaturation of template DNA at 94°C for 5 min, five touch-down cycles of denaturation at 94°C for 30 sec, annealing at 59-55°C for 40 sec, and extension at 72°C for 30 sec, followed by 30 cycles of denaturation at 94°C for 30 sec, annealing at 53°C for 40 sec, and extension at 72°C for 1 min. The final extension was performed at 72°C for 10 min

All PCR products were visualized in a 1.5% agarose gel. Only clones that were positive for EFa and negative for EFb, or vice versa, were taken to be confirmative results. All ambiguous results were disregarded. Accuracy of the exon-specific PCRs was verified by spot-checking clones by direct sequencing. Tabulation of the number of positives for each splice-specific PCR produced a ratio of EFa-versus EFb-containing transcripts in the various libraries.

Results

Splicing of the EF-hand domain as a molecular switch for CDF

For reference, we first considered a familiar variant of P/Q-type channels, composed of α_1A subunits containing both the EFa version of exon 37 and the longer carboxyl tail encoded by exon 47 [EFa/47+] (Fig. 1). Here, and for all recombinant channel expression experiments, α_1A was coexpressed with accessory β_2a and α₂δ subunits in HEK293 cells. This genre of P/Q-type channels, closely similar to those investigated previously (DeMaria et al., 2001; Soong et al., 2002), would be expected to manifest robust CDF and CDI, characterized as follows. Indeed, with Ca²⁺ as the charge carrier, CDF was readily apparent as a slow phase of increasing Ca²⁺ current seen during short (50 msec) test depolarizations (Fig. 2A, arrow, gray trace). When channels were facilitated by Ca²⁺ entry during a preceding voltage prepulse, ensuing currents activated rapidly to the facilitated level during the test pulse (black trace). To quantify the facilitation produced by the prepulse, we integrated the difference (ΔQ) between normalized test-pulse currents in the absence and presence of a prepulse, and this integral was used to determine the relative facilitation (RF) induced by the voltage prepulse (see supplemental information, available at www.jneurosci.org/cgi/content/full/24/28/6334/DC1). Reassuringly, RF demonstrated a bell-shaped dependence on prepulse voltage (Fig. 2B, filled circles), as expected for a genuine Ca²⁺-driven process (Brehm and Eckert, 1978). In contrast, as expected from the strong preference of CaM for Ca²⁺ over Ba²⁺ (Chao et al., 1984), there was little evidence of such prepulse facilitation with Ba²⁺ as the charge carrier, either in exemplar traces (Fig. 2A) or in population averages (Fig. 2B, open circles). The small, monotonic increase in RF seen with Ba²⁺ (Fig. 2B) reflected weak background G-protein modulation (DeMaria et al., 2001). Thus, to index pure CDF, we used the difference between Ba²⁺ and Ca²⁺ RF relationships (g value of ∼0.23) after a 20 mV prepulse.

Figure 2.

The [EFa/47+] splice variant shows robust CDF and CDI. A, Prepulse protocol characterizing CDF of [EFa/47+] channels, as expressed in HEK293 cells (top). The 50 msec test-pulse traces collected in Ca²⁺ show a slow phase of activation (arrow, middle) in the absence of a prepulse (light trace). The slow phase is abolished when the test pulse follows a prepulse (dark trace). The trace with a prepulse is scaled slightly to match the terminal phase of the test-pulse current without a prepulse. Test-pulse traces in Ba²⁺ from the same cell (bottom), showing a similar profile with (dark trace) or without (light trace) a prepulse. Format and normalization are as with Ca²⁺ traces (middle). Tail currents are clipped to frame for clarity. B, Fractional increase in rapidly activating test-pulse current produced by prepulse (RF), expressed as a function of prepulse voltage. RF is calculated from the area between test-pulse currents in A, after traces have been normalized to unity at the end of test-pulse depolarization. Symbols show averages from multiple cells, and error bars are displayed when larger than symbol size. The difference in RF obtained with Ba²⁺ and Ca²⁺, as observed after a 20 mV prepulse, provides an index of pure CDF (g). Average g values ± SEM for n cells are shown. The g value for exemplar cell CDF is displayed in A. See supplemental information (available at www.jneurosci.org/cgi/content/full/24/28/6334/DC1) for a complete description of the analysis. C, CDI during a 1 sec test-pulse depolarization, seen as faster decay of Ca²⁺ versus Ba²⁺ currents. To improve visual comparison of decays, Ba²⁺ currents are scaled ∼2× downward to match peak Ca²⁺ current amplitudes, here and throughout. D, Fraction of peak current remaining after 800 msec depolarization, r₈₀₀, plotted as a function of test-pulse voltage. Because CDI depends on global Ca²⁺ concentrations (Soong et al., 2002), cells with small current densities show minimal CDI. To index CDI without bias attributable to the fraction of cells with small total Ca²⁺ currents, we analyzed CDI only in cells in which Ca²⁺ current density was >40 pA/pF. Symbols are averaged from multiple cells, and error bars are displayed when larger than symbol size. To index pure CDI, we measured the difference in r₈₀₀ obtained with Ca²⁺ and Ba²⁺ during depolarization to 10 mV (f). Average f values ± SEM for n cells are shown. The f value for exemplar cell CDI is displayed in C. Tail currents are clipped to frame for clarity. Calibration, 1 nA. Dotted lines show zero levels here and throughout.

When viewed over longer (1 sec) depolarizations, the same P/Q-type channels [EFa/47+] gave strong evidence of an opposing CaM-mediated regulatory process, CDI. This manner of regulation could be readily appreciated from a significantly greater decrease in current with Ca²⁺ rather than Ba²⁺ as the charge carrier (Fig. 2C). To quantify CDI, we measured the fraction of peak current remaining at 800 msec after the start of the voltage step (r₈₀₀). With Ca²⁺, graphs of r₈₀₀ plotted against the step voltage traced a classic U-shaped trajectory, characteristic of a Ca²⁺-dependent process (Fig. 2D). The slower decay of Ba²⁺ currents (Fig. 2C) reflected a separate voltage-dependent inactivation process (Stotz and Zamponi, 2001), consistent with the monotonic decline of the Ba²⁺ r₈₀₀ with voltage (Fig. 2D). To index pure CDI, then, we calculated the difference between r₈₀₀ values for Ca²⁺ and Ba²⁺ currents evoked by 10 mV voltage steps (f value of ∼0.39). Overall, the robust CDI and CDF of [EFa/47+] channels were very similar to those exhibited by previously characterized splice variants of P/Q-type channels (DeMaria et al., 2001; Soong et al., 2002).

In striking contrast, CDF was essentially abolished in P/Q-type channels containing α_1A subunits that differed only in regard to the presence of the EFb version of exon 37, [EFb/47+] (Figs. 1, 3). For this construct, Ca²⁺ currents in the absence of a prepulse lacked the slow phase of increase, showing mainly the rapidly activating component of current (Fig. 3A). Moreover, the test-pulse current was not significantly enhanced by a voltage prepulse, mirroring almost completely the activation pattern of Ba²⁺ currents through the same channels (data not shown). These trends were corroborated entirely by the near equivalence of averaged Ca²⁺ and Ba²⁺ RF relationships (g value of ∼0.05) (Fig. 3B), both of which lacked the bell-shaped signature of Ca²⁺-driven regulation. Thus, just as man-made mutations in the EF-hand region of the L-type channel could eliminate CaM-dependent regulation of that channel (Peterson et al., 2000), here a naturally occurring splice variant exerted a similar effect, acting as a molecular switch to extinguish CDF of P/Q-type channels.

Figure 3.

CDF is abolished but CDI is preserved in [EFb/47+] channels. A-D, Format as in Figure 2, except CDF traces for Ba²⁺ are omitted. A, Exemplar Ca²⁺ currents showing little or no CDF. B, Lack of CDF is significant compared with [EFa/47+] channels (p < 0.01) and confirmed by virtually identical RF plots for Ca²⁺ and Ba²⁺, averaged from multiple cells. For reference, corresponding [EFa/47+] fit is reproduced as the top-most curve. C, D, CDI preserved at levels comparable with that in [EFa/47+] channels (p > 0.05).

Examination of CDI in these same channels [EFb/47+] provided another valuable perspective on the functional impact of splice variation at the EF-hand locus. Despite common reliance of CDI and CDF on an IQ-like element in the α_1A carboxyl tail (Fig. 1) (DeMaria et al., 2001), the presence of the EFb version of exon 37 entirely spared the CDI mechanism (f value of ∼0.28) (Fig. 3C,D). Hence, splice variation in the EF-hand segment of P/Q-type channels switches channel function with exquisite selectivity, dissecting apart CDI and CDF, two processes that have heretofore appeared to be closely linked (Lee et al., 1999; DeMaria et al., 2001).

Combinatorial effects of alternative splicing at exons 37 and 47

The effects of splice variation could be enriched if a specific function of a molecule were specified, not by the outcome of splicing at one locus, but collectively by the outcome at multiple loci (Mittman et al., 1999a,b). We thus considered whether the identity of the terminal stretch of the α_1A carboxyl tail might modulate the effects of EF-hand splicing, given that the terminus probably interacts with various adaptor proteins to organize channels in variable clusters and local environments (Maximov et al., 1999; Maximov and Bezprozvanny, 2002; Spafford et al., 2003; Weick et al., 2003). For P/Q-type channels, the terminus of the α_1A carboxyl tail is determined by the choice of splice acceptor site near the beginning of exon 47 (Fig. 1) (Soong et al., 2002). Splicing just before the intron 46/exon 47 border yields a transcript in which exon 47 is translated to produce a long C terminus, [47+]; splicing right at the border places an in-frame stop codon right after exon 46, yielding a shorter C terminus, [Δ47]. In the data presented thus far, EF-hand splicing was always examined in the context of the longer carboxyl tail, [47+]. Would substitution of the shorter carboxyl tail, [Δ47], affect our findings?

For P/Q-type channels composed of α_1A subunits with the EFa version of exon 37 and the shorter carboxyl tail [EFa/Δ47], robust CDF and CDI were clearly evident (Fig. 4) and were no different from those found in the presence of the longer carboxyl tail (Fig. 2). Moreover, the initial impression was that channels with the EFb version of exon 37 and the shorter carboxyl tail [EFb/Δ47] appeared to be unaffected by the identity of the distal C terminus (Fig. 5). On average, CDF was appreciably suppressed (g value of ∼0.09) (Fig. 5A,B), and CDI was just as strong as seen with any of the other constructs (Fig. 5C,D).

Figure 4.

Robust CDF and CDI in [EFa/Δ47] channels. A-D, Format as in Figure 2, except CDF traces for Ba²⁺ are omitted. p > 0.05 compared with [EFa/47+] for CDF and CDI.

Figure 5.

CDF is reduced and CDI is preserved in [EFb/Δ47] channels. A-D, Format as in Figure 2, except CDF traces for Ba²⁺ are omitted. A,Ca²⁺ currents in the absence of a prepulse (light trace) show a small facilitating component. B, Modest CDF is demonstrated by a diminished RF relationship for Ca²⁺ currents, averaged from multiple cells. For reference, the top most curve shows the corresponding RF relationship for [EFa/Δ47] channels. CDF is significantly less than in [EFa/Δ47] channels (p < 0.01) but comparable with that in [EFb/47+] channels (p > 0.05). C, D, CDI preserved at levels comparable with that in [EFa/47+] channels (p > 0.05).

However, on closer examination of the CDF manifest by [EFb/Δ47] channels, we uncovered a remarkable effect. For reference, consider the behavior of EFa splice variants when represented as a plot of facilitation strength (Fig. 6A, g) as a function of Ca²⁺ current density. This plot confirms robust CDF that is independent of current density, an outcome suggestive of modulation driven by local Ca²⁺ influx through individual channels rather than by global influx through many channels (Soong et al., 2002). At the other extreme, the corresponding plot for [EFb/47+] channels is consistent with complete elimination of CDF, as is evident from small g values across all current densities (Fig. 6B). With these reference behaviors in mind, we scrutinized [EFb/Δ47] channels from a similar perspective. To start, CDF of these channels, although strongly reduced compared with constructs with EFa, was undeniably present (Fig. 5A,B), with an average strength intermediate between the virtual elimination seen in [EFb/47+] channels and the robust baseline of [EFa] channels. Underlying this average was considerable heterogeneity in facilitation (Fig. 6C, middle and bottom). Certain cells showed minimal CDF, whereas others showed rather impressive CDF. Intriguingly, the heterogeneity was not random, but reflected a positive correlation between the strength of CDF (g) and Ca²⁺ current density in individual cells (Fig. 6C). This positive correlation suggests that CDF is now driven by elevations in global Ca²⁺ influx, as represented by overall current density (Soong et al., 2002). In addition, the small g values at lower current densities imply that local Ca²⁺ influx through individual channels fails to induce substantial CDF in this splice variant. This profile of Ca²⁺ requirements underlying CDF is especially notable, because it suggests a novel capability of splicing: the ability to tilt the preference of a Ca²⁺ regulatory process between local and global Ca²⁺ influx. In fact, presynaptic P-type Ca²⁺ current densities can easily extend beyond 100 pA/pF (Sun and Wu, 2001), suggesting that native currents could readily induce substantial CDF driven by global Ca²⁺ activity of [EFb/Δ47] channels (Fig. 6C).

To substantiate further the possibility that [EFb/Δ47] channels possess CDF preferentially driven by global versus local Ca²⁺ influx, we investigated the effects of intracellular dialysis with the rapid Ca²⁺ chelator BAPTA (10 mm). This buffer should strongly suppress Ca²⁺ elevations throughout the cell interior, except within a nanodomain near the channel mouth (Naraghi and Neher, 1997; Augustine et al., 2003). Thus, CDF would persist in the presence of BAPTA only if CDF were driven preferentially by local Ca²⁺ activity; such an outcome has been observed previously for EFa splice variants (Fig. 6D) (Soong et al., 2002). Alternatively, if CDF is driven by global Ca²⁺ activity, little or no facilitation should be observed with BAPTA. As a check on the effectiveness of BAPTA to chelate global Ca²⁺ in the experiments here, we confirmed the complete elimination of CDI (f = -0.02 ± 0.02; n = 12), a process previously established to require global Ca²⁺ activity (Soong et al., 2002). Under these conditions, we also observed essentially complete elimination of CDF in [EFb/Δ47] channels (g value of ∼0.06) (Fig. 6E), as tested across the entire range of current densities. Overall, then, the CDF of [EFb/Δ47] channels is not switched off, but rather reprogrammed to be sensitive to global Ca²⁺ activity. In this regard, it is interesting that the CDF of presynaptic P-type currents in the calyx of Held is similarly sensitive to Ca²⁺ buffering (Borst and Sakmann, 1998; Cuttle et al., 1998), and this effect may well reflect the prevalence of [EFb/Δ47] channels at this synapse.

Prospects for EF-hand switching in native neurons

The rich functional diversity produced by alternative splicing of recombinant P/Q-type channels at exons 37 and 47 made us wonder how prevalent such splice variation might be in the nervous system. Indeed, there are indications that the rat brain shows differential expression of P/Q-type channel EF-hand variants across certain regions and developmental stages (Bourinet et al., 1999; Vigues et al., 2002). To expand this portrait of differential EF-hand expression, particularly with reference to the human context, we examined the distribution of transcripts possessing either the EFa or EFb form of exon 37 in human cDNA libraries (Fig. 7), using a PCR-based strategy. Across the entire adult brain, EFa and EFb transcripts had an almost equal distribution. This lies in stark contrast to the situation early in development, in which the brain was dominated by channels possessing the EFb variant. Additional inspection revealed that there were regional differences in the distribution of EF-hand transcripts across the adult brain. Certain areas, such as the cerebellum and hippocampus, possessed essentially equal amounts of P/Q-type channels with EFa or EFb domains. A completely new finding, unappreciated in previous screens of P/Q-type channel splice variation, concerned regions in which there was an extreme bias toward EFa or EFb. In the thalamus and substantia nigra, the EFa splice variant predominated, whereas in the amygdala, an opposite and even more extreme bias toward the EFb form was present.

Figure 7.

Prospects for neuronal control of CDF through alternative splicing. Prevalence of EF-hand splice variation in human brain, as manifest across different regions and developmental stages. The EF-hand region was amplified by PCR from the listed cDNA libraries and sub-cloned (see Materials and Methods); specific clones were subsequently identified as EFa or EFb. The bar graph tabulates these results, displaying the percentage of clones containing EFa versus EFb splice variants. The number of clones tallied for each library is listed in parentheses.

In regard to alternative splicing of exon 47, our previous work with the adult human brain indicates that approximately two-thirds of α_1A transcripts elaborate [47+] channels, whereas the remainder of transcripts direct [Δ47] channels (Soong et al., 2002). Overall, the evidence strongly suggests that splicing of exons 37 and 47 modulates P/Q-type channel CDF across temporal and spatial dimensions in the human brain.

Discussion

Custom switching of Ca²⁺/CaM regulation of P/Q-type channels by alternative splicing

We establish that the functional profile of a particular P/Q-type channel phenotype, Ca²⁺-dependent facilitation, is specified by the outcome of alternative splicing at more than one locus of variation. In particular, the combinatorial effect of splicing at both exons 37 and 47 enables CDF to be modulated with exquisite selectivity among three functional settings: (1) “ON” and sensitive to local Ca²⁺ activity ([EFa/47+] or [EFa/Δ47]), (2) “partially OFF” but inducible by heightened global Ca²⁺ activity ([EFb/Δ47]), and (3) “completely OFF” ([EFb/47+]). This finding merits discussion in two regards.

First, concerning the regulatory potential of alternative splicing, functional diversity could be enriched if a particular molecular phenotype were specified by splice outcomes at more than one locus (Mittman et al., 1999a,b; Emerick et al., 2003). For example, if a single functional characteristic were specified by the outcome of splicing at a single locus, then the number of potential operational settings would be limited by the possible splice outcomes at this locus (n). Alternatively, if molecular function were determined by splice outcomes at two loci, the number of potential functional profiles would be multiplicatively expanded to n× m, where m is the number of variants at the second locus. Involvement of additional loci would entail additional nonlinear augmentation of the possible modes of molecular operation. The control of P/Q-type CDF by splicing at both exons 37 and 47 provides a clear-cut example of multilocus control, one with likely biological impact, as discussed later.

Second, the manner in which splice variation impacts CDF, switching its Ca²⁺ responsiveness between a preference for local versus global Ca²⁺, represents a new dimension of modulation, as follows. It has been established recently that CaM regulation extends across virtually the entire Ca_V1-2 family of voltage-gated Ca²⁺ channels. In every case, CaM is preassociated with channels as a resident Ca²⁺ sensor (Liang et al., 2003), and each form of channel regulation is triggered preferentially by Ca²⁺ binding to one lobe of CaM but not the other. Furthermore, regardless of channel context or whether the observed Ca²⁺-driven modulation facilitates or inactivates channel opening, a consistent pattern of CaM lobe-specific functionality has prevailed. Regulation triggered by the C-terminal lobe of CaM appears invariably insensitive to internal Ca²⁺ buffering, whereas that initiated by the N-terminal lobe is eliminated by elevated Ca²⁺ buffering with EGTA or BAPTA. From Ca²⁺ diffusion arguments (Naraghi and Neher, 1997; Song et al., 1998; DeMaria et al., 2001; Augustine et al., 2003), the implication is that the C-terminal lobe of CaM responds preferentially to the opening of its associated Ca²⁺ channel (sensitive to local Ca²⁺ influx), whereas the N-terminal lobe detects aggregate activity of more distant sources of Ca²⁺ (sensitive to global Ca²⁺ influx). For previously studied P/Q-type channels (all with EFa), this pattern has manifested as CDF driven by the C-terminal lobe of CaM and CDI induced by the N-terminal lobe of CaM (DeMaria et al., 2001). Also consistent with these findings, 10 mm BAPTA has spared P/Q-type channel CDF while eliminating CDI (Soong et al., 2002). In the present study, we encounter an initial exception to the rule, in that the CDF of [EFb/Δ47] channels (triggered by the C-terminal lobe of CaM) becomes sensitive to global Ca²⁺ influx. This may be the first example in which the preference of a resident Ca²⁺ sensor for local versus global Ca²⁺ influx has been transformed.

EF-hand domain is required for regulation by the C- but not N-terminal lobe of CaM

Regarding structure-function mechanisms underlying CaM modulation of Ca²⁺ channels, [EFb/47+] channels provide a tantalizing clue; CDF is eliminated, whereas CDI appears entirely preserved. When CDF and CDI are viewed in relation to the lobe of CaM that triggers each regulatory form, such structural selectivity hints at a general design principle for the Ca_V1-2 family. Specifically, although Ca²⁺-CaM interactions with pre-IQ and IQ channel segments (Fig. 1) seem to be early triggers of channel regulation (Peterson et al., 2000; DeMaria et al., 2001; Liang et al., 2003), could the EF-hand domain be the general downstream transduction element for modulation by the CaM C-terminal lobe (CDF) but not the N-terminal lobe (CDI) (Fig. 8)?

Figure 8.

EF-hand domain may selectively transduce modulation by the CaM C-terminal lobe. The diagram depicts interactions proposed to underlie channel regulation by CaM. Preassociation of Ca²⁺-free CaM with the channel IQ region is permissive for the standard open state of the channel (conducting). Ca²⁺ binding to the C-terminal lobe of CaM induces altered interaction with the IQ domain, an event that is sensed and transduced into channel modulation by the EF-hand domain (yielding CDF in P/Q-type channels and CDI in L-type channels). In contrast, Ca²⁺ binding to the N-lobe of CaM induces conformational changes that are sensed and transduced into channel modulation (CDI of N-, P/Q-, and R-type) by means that may not involve the EF-hand element.

Telling support comes from L-type (Ca_V1.2) channels, in which Ca²⁺ binding to the C-terminal lobe of CaM selectively produces CDI and man-made mutations in the EF-hand region of these channels eliminate such regulation (Peterson et al., 2000; Kim et al., 2004). No splice variation of the EF-hand region has been reported (Lipscombe and Castiglioni, 2003), because disruption of L-type channel CDI would likely have a lethal cardiac effect (Alseikhan et al., 2002). Whether the EF hand is also unimportant for N-terminal lobe signaling in L-type channels cannot currently be assessed, because the N-terminal lobe of CaM is not presently linked to an identifiable form of channel modulation (Peterson et al., 1999). For N- (Ca_V2.2) and R-type (Ca_V2.3) channels, however, both of which exhibit CDI triggered by the N-terminal lobe of CaM (Liang et al., 2003), the prediction is that CDI would not be affected by manipulations of their corresponding EF-hand segments. Indeed, the impact of alternative splicing of the EF-hand segment of N-type channels (Mittman and Agnew, 2000; Lipscombe et al., 2002) seems related to the overall amplitude of currents (Bell et al., 2004) rather than to a detectable change in CDI.

Beyond concerns of structure-function mechanisms, channel splice variants and mutants that selectively eliminate C- or N-terminal lobe signaling will also prove useful for neurobiological discovery. For example, to understand the biological impact of P/Q-type channel CDF, it would be advantageous to have targeted genetic means to selectively eliminate CDF without disrupting other forms of CaM regulation. Replacement of [EFa] channels by their [EFb/47+] counterparts, certainly by transgenic approaches, would serve nicely in this capacity.

Spatial and developmental control of P/Q-type channel regulation by alternative splicing

Our analysis of human mRNA transcripts (Fig. 7) suggests that both EFa and EFb versions of P/Q-type channels are likely to be prevalent and subject to considerable spatial and developmental control. We demonstrated that the distribution of splice variants encoding these channels varies systematically across different regions of the brain, as well as developmentally over time. These findings are consistent with those from rats (Bourinet et al., 1999; Vigues et al., 2002). However, our more extensive screen in humans not only amplifies these themes but also reveals some remarkable new trends. Consistent with findings in rat, human fetal whole brain shows an overwhelming presence of EFb transcripts, which transitions to an approximately equal distribution of EFa and EFb by adulthood. In adulthood, we unexpectedly discovered that certain subcortical areas (the amygdala, substantia nigra, and thalamus) manifest highly skewed distributions of EFa versus EFb transcripts (Fig. 7), suggesting that the presence of one or the other version of the EF hand is critical for proper function in these regions. Finally, although coarse developmental and regional variations in splicing are certainly observed, an even finer scale of control may be present at the level of individual neurons. For example, whereas EFa and EFb transcripts are seen in approximately equal amounts in the adult cerebellum, Purkinje cells from adult rats express predominantly EFa transcripts (Bourinet et al., 1999).

Given our functional characterization of channel splice variants, it seems likely that the purpose of variations in EFa and EFb transcripts would be to modulate channel CDF, thereby impacting neuron responsiveness to repetitive stimulation. In particular, because P/Q-type channels are enriched at presynaptic terminals (Dunlap et al., 1995) and neurotransmitter release increases with Ca²⁺ concentration by a third-to-fourth power relationship (Dodge and Rahamimoff, 1967), CDF (and preference for local or global Ca²⁺) could significantly alter the profile of short-term synaptic plasticity. As an estimate, consider that Ca²⁺ currents through recombinant EFa channels facilitate ∼20% when driven by trains of simulated action potentials (DeMaria et al., 2001), yielding a predicted enhancement of synaptic efficacy of ∼200%. In contrast, the corresponding [EFb/47+] channels hardly facilitate when driven identically (data not shown), suggesting large differences in the short-term plasticity of synapses dominated by EFa versus EFb P/Q-type channels. Such increased short-term synaptic facilitation may convert synapses from detectors of changes in input firing rate (“novelty detectors”) to more faithful conduits of absolute input firing rate (Tsodyks and Markram, 1997; Tsodyks et al., 1998). Enhanced short-term facilitation would also augment the efficacy of burst stimuli, a robust form of input signaling that enhances the precision of mapping location in hippocampal place cells, and of detecting stereotyped movements in visual cortex (Lisman, 1997). Beyond information processing, P/Q-type channel activity is particularly important for the pruning of multiple climbing fibers that innervate single cerebellar Purkinje cells in neonates, which leads to the one-to-one pairing of climbing fibers and Purkinje cells in adulthood (Miyazaki et al., 2004). Because this pruning process occurs contemporaneously with the increase in EFa (Fig. 7B) (Vigues et al., 2002) and P/Q-type channels are unique in manifesting rapid CDF, one could hypothesize that CDF is essential for this sculpting of neuronal architecture. Testing for neurobiological consequences such as these now emerges as a critical challenge for the future.