Abstract
The N-type Ca channel α1B subunit is localized to synapses throughout the nervous system and couples excitation to release of neurotransmitters. In a previous study, two functionally distinct variants of the α1B subunit were identified, rnα1B-b and rnα1B-d, that differ at two loci;four amino acids [SerPheMetGly (SFMG)] in IIIS3–S4 and two amino acids [GluThr (ET)] in IVS3–S4. These variants are reciprocally expressed in rat brain and sympathetic ganglia (Lin et al., 1997a). We now show that the slower activation kinetics of rnα1B-b (ΔSFMG/+ET) compared with rnα1B-d(+SFMG/ΔET) channels are fully accounted for by the insertion of ET in IVS3–S4 and not by the lack of SFMG in IIIS3–S4. We also show that the inactivation kinetics of these two variants are indistinguishable. Through genomic analysis we identify a six-base cassette exon that encodes the ET site and with ribonuclease protection assays demonstrate that the expression of this mini-exon is essentially restricted to α1B RNAs of peripheral neurons. We also show evidence for regulated alternative splicing of a six-base exon encoding NP in the IVS3–S4 linker of the closely related α1A gene and establish that residues NP can functionally substitute for ET in domain IVS3–S4 of α1B. The selective expression of functionally distinct Ca channel splice variants of α1B and α1A subunits in different regions of the nervous system adds a new dimension of diversity to voltage-dependent Ca signaling in neurons that may be important for optimizing action potential-dependent transmitter release at different synapses.
- N-type calcium channel
- regulated alternative splicing
- S3–S4 linker
- genomic analysis
- P/Q-type calcium channel
- calcium channel α1 subunits
N-type Ca channels, together with P/Q-type channels, control calcium-dependent neurotransmitter release at the majority of synapses in the mammalian nervous system (Hirning et al., 1988; Turner et al., 1992; Takahashi and Momiyama, 1993; Olivera et al., 1994; Wheeler et al., 1994; Dunlap et al., 1995). Neuronal Ca channels are also recognized as important targets for the treatment of chronic pain and neuronal degeneration after ischemic brain injury (Miljanich and Ramachandran, 1995). As the functional core of N-type Ca channels, the α1B subunit contains domains critical for voltage-sensing, ion permeation (Dubel et al., 1992;Williams et al., 1992; Fujita et al., 1993), toxin binding (Ellinor et al., 1994), excitation–secretion coupling (Sheng et al., 1994), and second messenger-mediated modulation (Zamponi and Snutch, 1998). The α1B subunit of the N-type Ca channel is subject to alternative splicing (Williams et al., 1992; Coppola et al., 1994; Lin et al., 1997a), but information regarding the functional significance of these splicing events is limited. We recently reported that variants of the α1B-subunit that differ by four amino acids, SFMG (SerPheMetGly), in domain IIIS3–S4 and two amino acids, ET (GluThr), in domain IVS3–S4 (see Fig. 1) are differentially expressed in rat brain and sympathetic ganglia (Lin et al., 1997a). The sympathetic ganglia-dominant form of the Ca channel α1B-subunit (rnα1B-b, Δ/+; see Figs. 1, 2) activates 1.5-fold slower and at potentials 7 mV more depolarized relative to the brain-dominant form (rnα1B-d, +/Δ; see Figs. 1, 2) (Lin et al., 1997a). The relative functional impact of each of the spliced sequences (SFMG and ET) on the channel has not been determined. Furthermore, the genomic structures of these splice sites have not been characterized. It is also unclear whether alternative splicing in the functionally important S3–S4 linker region of α1B is regulated similarly in all regions of the rat brain. In this study we examine these issues and also show that the functional differences between rnα1B-b and rnα1B-d channels may under certain conditions impact the magnitude of action potential-induced calcium transients as revealed in a model neuron. Variants of the closely related Ca channel α1A subunit, which differ in the expression of two amino acids (NP) also in domain IVS3–S4, have been isolated (Yu et al., 1992; Zamponi et al., 1996;Ligon et al., 1998; Sutton et al., 1998; Hans et al., 1999). Recent studies show that these α1A NP variants differ in their gating kinetics and sensitivity to block by ω-Aga IVA (Sutton et al., 1998; Hans et al., 1999), leading to the proposal that they account for P- and Q-type Ca channels described in mammalian neurons (Wheeler et al., 1994; Sutton et al., 1998). Little is known, however, about the relative abundance of α1A splice variants of the NP locus in different regions of the nervous system. We begin to examine this issue by the use of the ribonuclease protection assay and establish that the NP exon in α1A is also differentially expressed in different regions of the nervous system.
Preliminary reports of these findings have been presented previously in abstract form (Lin et al., 1997b; Schorge et al., 1998).
MATERIALS AND METHODS
Functional assessment of the Ca channel α1BcDNA constructs
The functional properties of all Ca channel α1BcDNA constructs described in this paper were assessed in theXenopus oocyte expression system. All methods and procedures were essentially the same as described in Lin et al. (1997a). Robust N-type Ca channel currents were expressed in Xenopus oocytes after injection of α1B cRNA without coexpression of exogenous Ca channel β-subunit. Heterologous expression of the Ca channel β3 subunit along with α1B increased N-type Ca channel current expression levels but did not affect channel gating kinetics nor did the presence of β3 affect the relative differences in the time course and voltage dependence of activation of the α1B splice variants (Lin et al., 1997a). Xenopus oocytes do express, however, an endogenous Ca channel β-subunit (β3XO) (Tareilus et al., 1997) that is highly homologous to the mammalian β3. It is possible that the Xenopus β3XO associates with and modulates heterologously expressed α1B in the oocyte. This may explain why coexpression of heterologous β3 is not required for functional expression of N-type Ca channels in this system (Tareilus et al., 1997). cRNAs were in vitro transcribed using the mMESSAGE mMACHINE kit (Ambion) from the various α1B cDNA constructs subcloned into theXenopus β-globin expression vector (pBSTA) (Goldin and Sumikawa et al., 1992). A cRNA solution (46 nl of 750 ng/μl) was injected into defolliculated oocytes using a precision nanoinjector (Drummond). N-type Ca channel currents were recorded 6–7 d after injection. At least 15 min before recording, oocytes were injected with 46 nl of a 50 mm solution of 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetate (BAPTA). This we have found critical to minimize activation of an endogenous Ca-activated Cl− current, even when Ba2+ is the charge carrier (Lin et al., 1997a). Cells exhibiting slowly deactivating tail currents, indicative of the presence of Ba2+-dependent activation of the Ca-activated Cl− current, were excluded from the analyses.
N-type Ca channel currents were recorded from oocytes using the two-microelectrode voltage-clamp recording technique (Warner amplifier; OC-725b). Micropipettes of 0.8–1.5 and 0.3–0.5 MΩ resistance when filled with 3 m KCl were used for the voltage and current recording electrodes, respectively. Oocytes expressing Ca channel currents usually had resting membrane potentials between −40 and −50 mV when impaled with two electrodes. A grounded metal shield was placed between the two electrodes to increase the settling time of the clamp. Recording solutions contained 5 mmBaCl2, 85 mm tetraethylammonium, 5 mm KCl, and 5 mm HEPES, pH adjusted to 7.4 with methanesulfonic acid. The recording temperature was between 19° and 22°C.
The properties of each mutant construct were assessed by expressing it together with appropriate controls (ΔET α1B and +ET α1B). Each mutant was tested in three separate batches of oocytes, and within each batch recordings were made from at least six oocytes for each mutant construct and control. Recordings from the oocytes expressing the various Ca channel α1Bconstructs were randomized throughout the data collection period.
Data analysis. Data were acquired on-line and leak-subtracted using a P/4 protocol (PClamp V6.0; Axon Instruments). Voltage steps were applied every 10–30 sec depending on the duration of the step, from a holding potential of −80 mV. Ca channel currents recorded under these conditions showed little run-down over the duration of the recordings. Three sets of current–voltage relationships were obtained from each cell using step depolarizations of 26.3 msec, 650 msec, and 2.6 sec in duration and digitized at 25 kHz, 10 kHz, and 250 Hz, respectively. Exponential curves (activation and inactivation) were fit to the data using curve-fitting routines in PClamp (Axon Instruments) and Origin (Microcal). Inactivation time constants in the range of 70–800 msec were estimated from currents evoked by the longest depolarizations (2.6 sec). Activation time constants were best resolved from currents evoked by the shortest depolarizations (26.3 msec; sampled at 25 kHz).
Ribonuclease protection assay
The procedures are essentially the same as those described inLin et al. (1997a). Total RNA was purified from various neuronal tissues of adult rats using a guanidium thiocyanate and phenol-chloroform extraction protocol [adapted from Chomczynski and Sacchi (1987)]. 32P-labeled antisense RNA probes overlapping ET [nucleotide (nt) 4379–4836] in rnα1B-band NP (nt 4605–4930) in rbα1A (Starr et al., 1991) were constructed from linearized plasmids (pGEM-T vector) containing appropriate RT-PCR-derived subclones using the Maxi-script kit (Ambion). Probes were gel-purified and stored as ethanol precipitates. RNA (1 μg) purified from sympathetic or sensory ganglia or RNA (5 μg) isolated from various CNS tissues was precipitated with 2 × 105 cpm of probe and resuspended in 30 μl hybridization buffer containing 60% formamide, 0.4 m NaCl, 10 mm EDTA, and 40 mm PIPES at pH 6.4. Samples were denatured at 85°C and allowed to hybridize overnight at 60°C. The samples were then digested in a 350 μl reaction mix containing 0.3 m NaCl, 5 mm EDTA, 3.5 μl of the RNase mixture (Ambion), and 10 mm Tris at pH 7.5, then treated with proteinase K, extracted, and precipitated with 10 μg of tRNA as carrier. After resuspension in 30 μl formamide loading buffer, the samples were denatured and separated on a 5% polyacrylamide gel. After exposure to a phosphorimaging plate to quantify relative band intensities (Fuji BAS 1000), the gel was subsequently exposed to film with an intensifying screen for 4–5 d at −80°C.
Site-directed mutagenesis
A recombinant PCR-based technique was used to introduce mutations (QT, EA, AT, AA, NP) at the ET site in the IVS3–S4 linker of α1B-b. A pair of primers 5′-attcttgtggtcatcgccttgag (Bup 3460) and 5′-gacaggcctccaggagcttggtg (Bdw 5623) flanked a region of the clone that contained two restriction sites, RsrII (nt 3510) andBglII (nt 5465), located either side of ET (nt 4674). A second primer pair contained the desired mutation and directly overlapped the ET site (Bdwmut and Bupmut; see below). Two separate PCRs were performed with Bup 3460 and Bdwmut, and Bupmut and Bdw 5623. The PCR product then served as template for a second round of PCR using Bup 3460 and Bdw 5623, generating the final mutant PCR fragment that was subsequently subcloned into rnα1B-b at the Rsr II andBglII sites. Mutants were screened by restriction digest and confirmed by DNA sequencing. All PCR was performed using Expand High Fidelity (Boehringer Mannheim, Indianapolis, IN). The mutagenesis primers used were as follows: ET/AT: Bupmut 5′-gagattgcgGCAACGaacaacttcatc-3′; Bdwmut 5′-aagttgttCGTTTCcgcaatctccg-3′; ET/QT: Bupmut 5′-gagattgcgCAGACGaacaacttcatc-3′; Bdwmut 5′-aagttgttCGTCTGcgcaatctccg-3′; ET/EA: Bupmut 5′-gagattgcgGAAGCTaacaacttcatc-3′; Bdwmut 5′-aagttgttAGCTTCcgcaatctccg-3′; ET/AA: Bupmut 5′-gagattgcgGCAGCTaacaacttcatc-3′; Bdwmut 5′-aagttgttAGCTGCcgcaatctccg-3′; ET/NP: Bupmut 5′-gagattgcgAACCCTaacaacttcatc-3′; Bdwmut 5′-aagttgttAGGGTTcgcaatctccg-3′.
Genomic analysis
The IVS3–S4 region of the rat α1B and α1A genes were analyzed by genomic PCR. Primer pairs were directed to the IVS3 and IVS4 membrane-spanning regions that were presumed to reside in the 5′ and 3′ exons flanking the ET and NP insertions of the α1B and α1A genes, respectively. PCR was performed in a 50 μl reactions mix containing 250 ng rat liver genomic DNA, 250 μm each nucleotide, and 0.4 μm each primer. After a preincubation for 15 min at 92°C, 0.75 μl enzyme mix was added to start the amplification. The resultant gDNA products were gel-purified, cloned into pGEM-T (Promega), and sequenced. The α1B primers generated two bands of ∼11 kb and ∼900 bases. The 11 kb band was derived from the α1B gene and contained the desired ET encoding exon in IVS3–S4. The 900 base product resulted from amplification of the equivalent site in the α1E gene that contained a relatively short ∼700 bp intron and no intervening exon. The α1A primers generated a single 9 kb PCR product that was confirmed to be derived from the α1A gene by DNA sequencing (Yale University sequencing facility). Primers were as follows: α1A: Aup4737 5′-tgcctggaacatcttcgactttgtga; Adw4876 5′-cagaggagaatgcggatggtgtaacc; and α1B: Bup4599 5′-cagagatgcctggaacgtctttgac; Bdw4744 5′-ataacaagatgcggatggtgtagcc.
Modeling Ca entry
A one-compartment cell model using standard compartmental modeling techniques in NEURON (Hines and Carnevale, 1997) was used to predict the amount of Ca entering a neuron expressing either rnα1B-b or rnα1B-d N-type Ca channel currents. The cell had a total membrane area of 1250 μm2, 0.75 μF/cm2 specific membrane capacitance, and 30 kΩcm2 specific membrane resistance. For action potential simulation a fast sodium conductance (gNa) and a delayed rectifying potassium conductance (gK,DR) were included (Mainen and Sejnowski, 1995), each with densities of 300 pS/μm2. Ca2+ influx was mediated by a fast calcium conductance (gCa) (Yamada et al., 1989) with a density of 1 pS/μm2. Action potentials were evoked by a 5 msec, 400 pA current step. Resultant currents were calculated using conventional Hodgkin-Huxley kinetic schemes according to the formulae given below. The resting membrane potential was set at −70 mV, and Na and K current reversal potentials were set at +50 mV and −75 mV, respectively. The calcium current was computed using the Goldman-Hodgkin-Katz equation. Extracellular Ca concentration was 2.5 mm, and the intracellular Ca concentration was computed using entry via ICaand removal via a first order pumpd[Ca2+]i/dt= (−1 · 105 ·ICa/2F) − ([Ca2+]i − [Ca2+]∞)/τR, where [Ca2+]∞ = 100 nm and τR = 80 msec. The simulation does not incorporate mobile or immobile Ca buffers. The time constants and maximal Ca channel conductances used in this model were measured at room temperature. These values were adjusted for simulation at 37°C based on Q10 values of 1.5 and 3.0 for the conductance and activation kinetics, respectively (Cota et al., 1983). Formulae used for calculation of various currents were as follows: sodium current (INa), m3· h: αm, Na = 0.182 · (v + 25)/(1 − e−(v+25)/9); βm, Na = −0.124 · (v + 25)/(1 − e(v+25)/9); αh, Na = 0.024 · (v + 40)/(1 − e−(v+40)/5); βh, Na = −0.0091 · (v + 65)/ (1 − e(v+65)/5); h∞, Na = 1/(1 + e(v+55)/6.2); delayed recifier (IK(DR)), m: αm, K(DR) = 0.02 · (v − 25)/(1 − e−(v+25)/9); βm, K(DR) = −0.002 · (v − 25)/(1 − e(v+25)/9); high threshold, N-type rnα1B-b calcium current (Ica),m · h: m∞, Ca = 1/(1 + e−(v−3)/8); τm, Ca = 7.8/(e(v+6)/16 + e−(v+6)/16); hCa = K/(K + [Ca2+]i) with K = 0.01 mm. The brain-dominant form, rnα1B-d, was modeled by shifting the voltage dependence of the N-type Ca channel conductance activation variable (m∞, Ca) by −7 mV and decreasing the activation time constant by 1.5-fold (Lin et al., 1997a) (see Fig. 2A).
RESULTS
Alternative splicing in the putative S3–S4 extracellular linkers affects channel activation but not inactivation kinetics
In a previous study we showed that rnα1B-b(ΔSFMG/+ET) and rnα1B-d (+SFMG/ΔET) N-type currents differ with respect to their gating kinetics when expressed inXenopus oocytes [compare Δ/+ and +/Δ in Fig.2A,B; see also Lin et al. (1997a)]. We reported apparent differences in both the macroscopic rates of channel activation and the inactivation between the two splice variants, but because relatively short duration depolarizations were used, the inactivation kinetics were not fully resolved. Consequently, and as discussed in Lin et al. (1997a), it was difficult to ascribe the differences in gating kinetics between the N-type Ca channel α1B splice variants exclusively to differences in either channel activation or inactivation mechanisms. In the present study we have used both short (26 msec) and long (2.6 sec) depolarizations to resolve the time course of Ca channel activation andinactivation. rnα1B-b (Δ/+) (Fig.1) and rnα1B-d (+/Δ) (Fig. 1) subunits were expressed in Xenopus oocytes, and the resulting N-type Ca channel currents were recorded using 5 mm Ba as the charge carrier (Figs.2, 3). N-type Ca channel currents evoked by depolarizations to 0 mV or higher inactivated with a bi-exponential time course (τfast 100–150 msec and τslow700–800 msec) (Fig. 3A). The inactivation time constants of the cloned channels expressed inXenopus oocytes (rnα1B-b, Δ/+ and rnα1B-d, +/Δ) (Fig. 3A,B) were weakly voltage dependent, consistent with studies of native N-type Ca channels of bullfrog sympathetic neurons (Jones and Marks, 1989). Figure3A,B shows that the fast and slow inactivation time constants of rnα1B-b and rnα1B-d currents evoked by relatively long duration step depolarizations to between 0 mV and +30 mV were not significantly different. In contrast, the rates of channel activation of the two variants in the same cells were significantly different (Fig. 2A,B). On the basis of these observations we conclude that alternative splicing in domains IIIS3–S4 and IVS3–S4 of the α1B subunit alters the time course of N-type Ca channel activation but has no direct effect on inactivation kinetics. The apparent differences in the time courses of inactivation previously reported for rnα1B-b and rnα1B-d using relatively short duration depolarizing pulses (Lin et al., 1997a) can thus be ascribed to differences in their rates of channel activation, not inactivation. A selective effect on channel activation kinetics is consistent with the close proximity of the spliced sites, S3–S4 linkers, to their respective S4 helices that are the putative voltage sensors of the six transmembrane family of voltage-gated ion channels (Hille, 1992). In contrast, the domains of the Ca channel α1B subunit implicated in voltage-dependent inactivation of N-type Ca channels (IS6 and flanking putative extracellular and intracellular linkers) (Zhang et al., 1994) are likely to be more distant from the S3–S4 linker splice sites.
Splicing of ET in domain IVS3–S4 underlies the major functional difference between rnα1B-b and rnα1B-d
rnα1B-b and rnα1B-d differ in composition by six amino acids located in two distinct regions of the Ca channel α1B subunit (SFMG in domain IIIS3–S4 and ET in domain IVS3–S4) (Fig. 1). To separate the relative contribution of SFMG in domain IIIS3–S4 and ET in domain IVS3–S4 to the different gating kinetics observed between rnα1B-b (ΔSFMG/+ET) and rnα1B-d (+SFMG/ΔET) we constructed two additional clones, +/+ and Δ/Δ (Fig. 1) and compared the functional properties of all four clones. Figure 2A,B demonstrates that the presence of the dipeptide sequence ET in domain IVS3–S4 is directly correlated with the altered activation kinetics of rnα1B-b currents compared with rnα1B-d. Activation time constants measured from N-type Ca channel currents in oocytes expressing clones Δ/+ (rnα1B-b) and +/+ were indistinguishable and 1.5-fold slower on average than those induced by the expression of clones +/Δ (rnα1B-d) and Δ/Δ (Fig.2A,B). The presence of ET in domain IVS3–S4 also influenced the voltage-dependence of channel activation. A comparison of the midpoints of the rising phase of the peak current–voltage plots (V1/2) generated for the two ET containing clones Δ/+ (rnα1B-b; −7.8 ± 0.6 mV, n = 6) and +/+ (−9.7 ± 1.0 mV,n = 6) shows that they are not significantly different from each other (p > 0.05, Student’st test). Likewise, V1/2 values estimated from two ET-lacking constructs, +/Δ (rnα1B-d; −15.4 ± 0.4 mV,n = 7) and Δ/Δ (−13.4 ± 0.7,n = 6), were not significantly different from each other (p > 0.05) and activated at potentials that were, on average, 6 mV more negative compared with ET-containing clones Δ/+ and +/+ (data not shown). Although the presence of ET in domain IVS3–S4 dominates in regulating the voltage dependence of activation, the analysis does reveal a small contribution of SFMG. SFMG-containing clones (+/Δ and +/+) activated at potentials that were 2 mV hyperpolarized compared with those that lacked SFMG (Δ/+ and Δ/Δ). A 2 mV shift in the voltage dependence of activation was not significant at the 5% level, in a comparison ofV1/2 values from clones Δ/+ and +/+, but did reach significance in a comparison of +/Δ and Δ/Δ (p < 0.025, Student’s t test).
The pattern of expression of ET-containing Ca channel α1B mRNA in different regions of the nervous system
Figure 2 indicates that alternative splicing of ET within domain IVS3–S4 of the Ca channel α1B subunit accounts for the major functional differences between rnα1B-b and rnα1B-d. This prompted us to systematically analyze the expression pattern of the six bases in α1B mRNA that encoded ET. We had shown previously that ET-containing α1B (+ET α1B) mRNA was in very low abundance in total rat brain extracts (Lin et al., 1997a). To determine whether ET-lacking α1B (ΔET α1B) mRNA dominated throughout the CNS we used the ribonuclease protection assay and analyzed RNA isolated from spinal cord, cerebellum, cortex, hippocampus, hypothalamus, medulla, and thalamus of adult rats (Fig. 4). In all regions tested >90% of the α1B mRNA expressed in the CNS lacked the ET encoding sequence. In contrast, in sympathetic and sensory ganglia the majority of α1B mRNA contained the ET encoding sequence (Fig. 4). Together these findings suggest that +ET α1B subunits are primarily restricted to neurons of the peripheral nervous system. Consistent with this we have analyzed RNA isolated from human brain and trigeminal ganglia and observed analogous patterns of expression: low levels of +ETα1B mRNA in brain (∼10%) and high levels (∼70%) in ganglia (S. Schorge and D. Lipscombe, unpublished data).
Site-directed mutagenesis within IVS3–S4
Having shown that alternative splicing of the ET encoding sequence in the IVS3–S4 linker of α1B has a significant effect on the kinetics and voltage dependence of N-type Ca channel gating, we next used site-directed mutagenesis to determine the relative importance of each amino acid, glutamate and threonine. A series of mutants in which ET was replaced with either QT, AT, EA, AA, or NP were constructed (Fig. 5) from clone Δ/+ (rnα1B-b), which served as the background structure. The mutant constructs were then expressed inXenopus oocytes, and their properties were compared with clones +ET (100% slow; Fig. 5) and ΔET (100% fast; Fig. 5). All mutants expressed equally well in the Xenopus oocyte expression system.
The role of the glutamate in domain IVS3–S4 was of major interest because it should be negatively charged at neutral pH and consequently might influence the gating machinery of the channel via electrostatic interactions. Figure 5, however, shows that replacing glutamate with glutamine resulted in a channel that activated only slightly faster than +ET α1B (Fig. 5, QT). Substituting alanine for glutamate (AT) decreased τact, but consistent with the QT mutant, suggests that the presence of a negative charge in IVS3–S4 (glu) does not underlie the slow gating kinetics of the +ETα1B variant. Similarly, alanine substitution of either threonine alone (EA) or together with glutamate (AA) generated channels with activation kinetics that were intermediate between +ET α1B and ΔET α1B clones. Together, these results suggest that the presence of both glutamate and threonine in the IVS3–S4 linker is necessary to reconstitute the relatively slow channel opening rates characteristic of N-type Ca channel α1B-subunits that dominate in sensory and sympathetic ganglia.
Evidence for alternative splicing in the IVS3–S4 linker regions of the α1B and α1A genes
The existence of an alternatively spliced exon in the IVS3–S4 region of the rat Ca channel α1B gene has been hypothesized (Lin et al., 1997a) but not yet confirmed. Genomic analysis was therefore undertaken to locate the splice junctions in the IVS3–S4 region of the α1B gene and to pinpoint the precise location of the putative six-base, ET-encoding exon. PCR amplification from rat genomic DNA using primers designed to hybridize to the transmembrane spanning S3 and S4 helices flanking IVS3–S4 in α1B revealed the presence of a long ∼10 kb stretch of intron sequence. DNA sequencing established the location of exon/intron and intron/exon boundaries and conserved ag–gt splice junction signature sequences immediately 5′ and 3′ to the putative ET insertion site (Fig. 6A, α1B). A six-base cassette exon encoding ET was located 8 kb into the 5′ intron and establishes that ET-α1B variants are generated by alternative splicing. Sequence comparisons of several cDNAs encoding α1subunits of other voltage-gated Ca channels suggests that alternative splicing in the IVS3–S4 linker could be a general mechanism for regulating voltage-dependent Ca channel gating (Fig.6B). This has recently been demonstrated for α1A (Sutton et al., 1998; Hans et al., 1999), a Ca channel subunit that is closely related both structurally and functionally to the N-type Ca channel α1B-subunit. A comparison of the IVS3–S4 region of various mammalian α1A cDNAs derived from kidney, pancreas, and brain (Fig.6B) (Yu et al., 1992; Ligon et al., 1998; Sutton et al., 1998; Hans et al., 1999) is consistent with alternative splicing of six bases encoding asp, pro (NP) in this region. The exon/intron structure in the IVS3–S4 linker region of the closely related rat α1A gene was therefore also determined (Fig.6A). The rat α1A gene contained a long stretch of intron sequence (∼8 kb) and ag–gt splice junctions at the 5′ (gt) and 3′ (at) ends of the intronic segment (Fig.6A). We have not yet determined the precise location of the NP encoding cassette exon in the rat α1A gene but conclude that it must reside within the 8 kb of intron sequence in the IVS3–S4 linker region. Tissue-specific alternative splicing of six-base cassette exons in the IVS3–S4 linkers of both α1A and α1B explains the presence of splice variants of these subunits in the mammalian brain and underscores the high level of conservation between these two functionally related genes. We have also analyzed the genomic structure of the more distantly related rat α1E gene that encodes a pharmacologically and functionally distinct class of Ca channel (Soong et al., 1993). The α1E gene contains a ∼700 bp intron in the IVS3–S4 linker region and no obvious intervening exon (Fig.6A). The absence of an alternatively spliced cassette exon in the IVS3–S4 linker region of the α1E gene is consistent with RNase protection analysis of α1E mRNA from rat brain, which revealed no evidence of sequence variations in this IVS3–S4 linker (data not shown). The high degree of sequence homology between α1B and α1A in the IVS3–S4 linker region (Fig. 6B) together with the finding that a six-base sequence is alternatively spliced at both of these sites (Fig. 6A) suggested that ET and NP share a common functional role. To test this hypothesis we studied the functional impact on N-type Ca channel currents of replacing ET in rnα1B-b with NP. Figure 5 shows that the +NPα1B mutant gives rise to N-type Ca channel currents in oocytes with gating kinetics indistinguishable from wild type (i.e., +ETα1B).
+NPα1A and ΔNPα1A mRNAs are expressed in different regions of the rat nervous system
Although the presence of variants of the Ca channel α1A that differ in the expression of the NP site has been reported (Yu et al., 1992; Ligon et al., 1998; Sutton et al., 1998;Hans et al., 1999), the distribution of +NPα1A and ΔNPα1A mRNAs in different regions of the rat nervous system have not been quantified. We therefore used the RNase protection analysis to determine the expression pattern of the IVS3–S4 splice variants of α1A (Fig. 7). Low levels of +NP α1A mRNA were found in rat, spinal cord, striatum, and thalamus, a pattern that parallels the low levels of +ET α1B mRNA in the CNS (Fig. 4). However, the pattern of NP expression in the cerebellum, cortex, and hippocampus did not conform to this picture because mRNA isolated from these tissues contained a significant proportion of +NPα1A mRNAs. In fact, in the hippocampus +NPα1A mRNAs dominated (∼60%). Consistent with the abundance of +ETα1B mRNAs in peripheral tissue, the majority of α1AmRNA in superior cervical and dorsal root ganglia contained the six bases encoding NP in domain IVS3–S4 of α1A. The absolute level of α1A mRNA expressed in sympathetic neurons was very low as expected from the absence of P-type currents in recordings from rat sympathetic neurons (Mintz et al., 1992).
The differences in the properties of rnα1B-b(ΔSFMG/+ET) and rnα1B-d (+SFMG/ΔET) currents may influence action potential-induced Ca entry
We have demonstrated functional differences between splice variants of the N-type Ca channel α1B subunit and shown that they are differentially expressed in peripheral and central neurons. We do not as yet know whether these functional differences are sufficient to influence action potential-dependent Ca entry in neurons. As a first step toward addressing this question we have used standard modeling techniques to predict whether rnα1B-b(ΔSFMG,+ET) and rnα1B-d (+SFMG, ΔET) N-type currents can, under certain conditions, differentially affect action potential-induced Ca influx in a model neuron. A more direct comparison of the effectiveness of the two Ca channel splice variants for supporting action potential-induced Ca entry inXenopus oocytes is not feasible using conventional two-microelectrode voltage-clamp methods because of the limited temporal resolution associated with cells of this size. Simulated action potentials similar to those recorded in native sympathetic neurons (Yamada et al., 1989) (Fig. 8) were therefore used to trigger voltage-dependent Ca influx in model neurons (Na, K, and Ca current densities of 300, 300, and 1.0 pS/μm2, respectively) expressing either rnα1B-b or rnα1B-d N-type Ca channel currents. Ca channel current densities and peak intracellular Ca concentrations were modeled at room temperature (22°C) (Fig.8A) and at 37°C (Fig. 8B). Under these conditions increases in both the total charge transfer and peak intracellular Ca concentration of 45% (at 22°C) and 35% (at 37°C) were observed after an action potential in a neuron expressing rnα1B-d-type Ca channels (Fig. 8, dashed line) relative to rnα1B-b (Fig. 8, solid line). The model also predicts a slightly faster rate of rise of the intracellular calcium signal (∼1.3-fold) in a neuron expressing rnα1B-d-type Ca channels compared with rnα1B-b.
DISCUSSION
The mammalian genes encoding voltage-gated Ca channel α1 subunits are large and complex, containing approximately 50 exons, several of which are alternatively spliced (Soldatov, 1994; Yamada et al., 1995; Hogan et al., 1996; Ophoff et al., 1996). With few exceptions (Lin et al., 1997a,b; Sutton et al., 1998; Hans et al., 1999), studies that address the functional consequences of these splicing events have been limited to non-neuronal-derived L-type Ca channel subunits (Klockner et al., 1997;Soldatov et al., 1997; Welling et al., 1997; Zuhlke et al., 1998). Building on our previous studies (Lin et al., 1997a), we now focus on the importance of regulated alternative splicing in the IVS3–S4 linkers of the α1B and closely related α1Asubunits.
Two amino acids in the IVS3–S4 extracellular linker of α1B slow N-type Ca channel activation
The S3–S4 linkers of a number of voltage-gated and structurally related ion channels are important in determining the time course and voltage dependence of channel activation (Perez-Reyes et al., 1990;Nakai et al., 1994; Lin et al., 1997a; Mathur et al., 1997; Tang and Papazian, 1997; Hans et al., 1999). We now demonstrate that alternative splicing of just two amino acids, ET, in domain IVS3–S4 of the Ca channel α1B subunit underlies the different gating kinetics of two previously identified variants, rnα1B-band rnα1B-d (Lin et al., 1997a). We also demonstrate that a chemically dissimilar but functionally homologous dipeptide sequence NP (see below) can fully replace and substitute for ET in domain IVS3–S4 of α1B, showing that the side chains of ET are not unique in contributing to the relatively slow activation kinetics of +ETα1B. The importance of the IVS3–S4 linker in influencing activation of the Ca channel is supported by the high level of conservation of alternative splicing in this region of other α1 subunits (Fig. 6) (Snutch et al., 1991; Barry et al., 1995; Ihara et al., 1995; Lin et al., 1997a; Ligon et al., 1998; Sutton et al., 1998; Hans et al., 1999). Moreover, in a recent studyHans et al. (1999) compared the functional properties of two variants of the Ca channel α1A subunit that also differed by two amino acids (NP) in the IVS3–S4 linker and showed that the ΔNP variant of α1A activated at a rate 1.7-fold faster compared with +NPα1A. Remarkably, this difference corresponds precisely to that observed between ΔET α1Band +ET α1B (1.5-fold) (Fig. 2). The similar functional consequences of splicing at the ET and NP loci, combined with the high degree of conservation between the IVS3–S4 splice junctions of the α1B and α1A genes, imply a common functional role. α1A and α1B subunits are both critically important for coupling excitation to transmitter release at the majority of synapses throughout the nervous system (Dunlap et al., 1995); thus one consequence of tissue-specific expression of functionally distinct IVS3-S4 splice variants of these proteins might be to optimize the release of neurotransmitters in different regions. +ETα1B and +NPα1A mRNAs that encode relatively slow activating channels (α1B; Fig. 2) (α1A; Hans et al., 1999) dominate in peripheral neurons (Figs. 4, 7), implying that excitation–secretion coupling might be less efficient at postganglionic synapses in comparison with many synapses in the CNS. +NPα1A mRNAs, however, are not solely restricted to peripheral neurons (Fig. 7), implying that different regions within the CNS may contain functionally distinct α1A-containing Ca channels. On the basis of their different ω-aga-IVA sensitivities,Sutton et al. (1988) have proposed that the expression of ΔNPα1A and +NPα1A splice variants correlates with the presence of high (P-type) and low (Q-type) affinity ω-aga-IVA-sensitive Ca channels (Sather et al., 1993; Stea et al., 1994; Wheeler et al., 1994; Randall and Tsien, 1995). Additional studies are needed to test this hypothesis, but our ribonuclease protection analysis (Fig. 7) suggests a reasonable, although not perfect, correlation between the presence of ΔNPα1AmRNA and P-type channels on the one hand and +NPα1A mRNA and Q-type on the other. For example, in DRG neurons the majority of the α1A mRNA pool contains the NP exon (Fig. 7), and in these neurons ω-aga-IVA inhibits the Ca current with relatively slow kinetics (Mintz et al., 1992), not inconsistent with the presence of a Q-type current. In spinal cord, ΔNPα1A mRNA dominates (Fig. 7), and block by ω-aga-IVA is relatively rapid (Mintz et al., 1992) and thus not inconsistent with P-type currents. Furthermore, in the cerebellum and hippocampus, where significant levels of both ΔNPα1A and +NPα1A mRNAs are found (Fig.7), both P- and Q-type Ca channel currents have been described (Llinas et al., 1989; Wheeler et al., 1994; Randall and Tsien, 1995). The dominance of ΔNP α1A mRNA in the thalamus (Fig. 7), on the other hand, does not correlate well with the significant Q-type current component reported in this region of the brain (Kammermeier and Jones, 1997).
As yet we have no evidence that splice variants of the N-type Ca channel α1B subunit differ in their pharmacological properties. ω-conotoxin GVIA, the widely used high-affinity N-type Ca channel blocker, inhibits both variants equally well (Z. Lin and D. Lipscombe, unpublished observations), consistent with studies of native N-type Ca channels in peripheral and central neurons (Wang et al., 1998). This result is not surprising, however, considering that the IVS3–S4 extracellular linker is not the main site of ω-conotoxin GVIA binding (Ellinor et al., 1994). Drugs or toxins that bind to the S3–S4 linkers of the Ca channel α1B subunit that discriminate between S3–S4 splice variants may be identified in the future. The S3–S4 linkers of several voltage-gated channel α1 subunits, including α1A, are known to be important targets of toxin binding (Rogers et al., 1996;Swartz and MacKinnon, 1997; Cestele et al., 1998; Sutton et al., 1998;Hans et al., 1999). In the case of the voltage-gated Na channel α subunit, a single amino acid (Gly) within the IIS3–S4 linker plays a major role in β-scorpion toxin binding (Cestele et al., 1998).
Predicting the differential effect of the different Ca channel α1B splice variants on action potential-induced calcium entry
Our modeling provides a useful first step toward understanding how differences in the gating kinetics and voltage dependence of activation between the α1B splice variants might affect action potential-induced Ca entry. Under the specific condition of the model used here, the CNS-dominant Ca channel variant ΔETα1Bsupports a larger and slightly faster rising intracellular calcium signal in response to action potential-induced depolarization compared with +ETα1B (Fig. 8). However, the significance of these differences with respect to excitation–secretion coupling efficiency at synapses that use the N-type Ca channel is difficult to predict. Neurosecretion is thought to be triggered by intracellular calcium levels in the range of 500 nm to a few micromoles (Augustine and Neher, 1992; Heidelberger et al., 1994), and differences in the rate of rise of the calcium transient may be more important than differences in the absolute levels of calcium. Our model, however, does not take into account how the presence of Ca buffers might shape the Ca signal close to the Ca channel in the microdomain where excitation–secretion coupling occurs (Naraghi and Neher, 1997). Furthermore, many other factors such as (but not limited to) Ca channel density, the shape of the presynaptic action potential, and neurotransmitter and second messenger-mediated Ca and K channel modulation will affect presynaptic calcium levels. For example, the effectiveness of rnα1B-d-type currents to couple depolarization to Ca entry in a neuron is enhanced relative to rnα1B-b with relatively brief depolarizations, but declines as the stimulus duration is increased. Furthermore, if the N-type Ca channel current activates fully during the rising phase of the action potential, then differences in the rate or voltage dependence of Ca channel opening will have relatively insignificant effects on Ca entry. In contrast, a slowing of Ca channel gating kinetics, such as is observed during G-protein-mediated modulation, would serve to amplify the different effectiveness of the two splice variants to couple depolarization to Ca entry. The primary purpose of the model presented here is to emphasize that although they are small, the functional differences between the α1Bsplice variants may, under certain conditions, significantly impact action potential-induced Ca entry.
Regulated alternative splicing is an important mechanism used throughout the nervous system for generating functionally distinct products from a single gene in different regions of the nervous system and at different stages during development (Grabowski, 1998). The continued characterization of alternatively spliced loci of Ca channel α1 subunits in neurons will likely lead to a greater understanding of the scope of Ca channel diversity and the physiological importance of the expression of splice variants in the mammalian nervous system.
Footnotes
This work was supported by National Institutes of Health (NIH) grants NS 29967 and NS 01927 (D.L.) and NIH Training Grant MH19118 (Z.L.). We thank Drs. Hans and colleagues for providing a copy of their manuscript before publication.
Correspondence should be addressed to Dr. Diane Lipscombe, Department of Neuroscience, Brown University, Box 1953, Providence, RI 02912.
Dr. Lin’s current address: Cold Spring Harbor Laboratory, Beckman Building, 1 Bungtown Road, Cold Spring Harbor, NY 11724.