Abstract
Ca2+ channel β subunits are important molecular determinants of the kinetics and voltage dependence of Ca2+ channel gating. Through direct interactions with channel-forming α1 subunits, β subunits enhance expression levels, accelerate activation, and have variable effects on inactivation. Four distinct β subunit genes each encode five homologous sequence domains (D1–5), three of which (D1, D3, and D5) undergo alternative splicing. We have isolated from human spinal cord a novel alternatively spliced β4 subunit containing a short form of domain D1 (β4a) that is highly homologous to N termini of Xenopus and rat β3subunits. The purpose of this study was to compare the gating properties of various α1 subunit complexes containing β4a with those of complexes containing a β4subunit with a longer form of domain D1, β4b. Expression in Xenopus oocytes revealed that, relative to α1A and α1B complexes containing β4b, the voltage dependence of activation and inactivation of complexes containing β4a were shifted to more depolarized potentials. Moreover, α1A and α1B complexes containing β4a inactivated at a faster rate. Interestingly, β4 subunit alternative splicing did not influence the gating properties of α1Cand α1E subunits. Experiments with β4deletion mutants revealed that both the N and C termini of the β4 subunit play critical roles in setting voltage-dependent gating parameters and that their effects are α1 subunit specific. Our data are best explained by a model in which distinct modes of activation and inactivation result from β-subunit splice variant-specific interactions with an α1 subunit gating structure.
Neuronal high voltage-activated Ca2+ channels (L, N, P/Q, and R) consist of at least four subunits, α1, α2/δ, and β (Liu et al., 1996), with a fifth subunit, γ, being recently described (Letts et al., 1998). Different Ca2+ channel phenotypes arise primarily from the expression of five unique α1subunit genes (α1A–α1E). These genes encode large pore-forming proteins (>2200 amino acids) that are differentially distributed throughout the nervous system (Westenbroek et al., 1990, 1998). Synaptic N-, P/Q-, and R-type channels, formed by α1B, α1A, and α1E subunits, respectively, play a principal role in regulating neurotransmitter release (Turner et al., 1992;Takahashi and Momiyama, 1993; Wheeler et al., 1994; Wu et al., 1999).
Ca2+ channel β subunits (subtypes 1–4) are highly homologous intracellular proteins with primary sequences ranging from 480 to 630 amino acids (for review, see Birnbaumer et al., 1998). The sequence can be divided into five domains on the basis of the regions of amino acid identity between subtypes. All β subunits contain a highly conserved β interaction domain (BID) in domain 4, which has been shown to interact with high affinity to an α interaction domain (AID) on the I–II linker of α1 subunits (Pragnell et al., 1994; De Waard and Campbell, 1995). Structure prediction methods using the Prodom and Pfam protein databases have established a domain structure (A–E domains) for the β1b subunit (Hanlon et al., 1999) that primarily overlaps with sequence domains 1–5. The A domain [100 amino acids (aa)] shows some homology to PDZ domains, the B domain (61 aa) to SH3 domains, and the D domain (210 aa) to guanylate-kinase, although it lacks a functional ATP-binding P-loop motif. Domains C and E were without precedent in the Prodom and Pfam protein databases; however, Domain C is rich in serine residues, suggesting that it serves a linker function between domains B and D. Thus, in many respects, Ca2+channel β subunits resemble members of the membrane-associated guanylate kinase (MAGUK) protein family, which are known to cluster ion channels, receptors, adhesion molecules, and cytosolic signaling proteins at synapses and cellular junctions (Fanning and Anderson, 1999).
Previous studies have shown that the kinetics and voltage sensitivity of α1 subunit gating are affected profoundly by β subunits (Lacerda et al., 1991; Singer et al., 1991), and the extent to which these parameters are altered varies significantly with β subunit subtype (Ellinor et al., 1993; Olcese et al., 1994). For example, although β1 and β3 subunits shift the voltage dependence of α1E subunit inactivation to more hyperpolarized potentials, β2 subunits have a marked depolarizing effect (for review, see Birnbaumer et al., 1998). Moreover, the responsiveness of α1 subunits to β subunit modulation can be modified by alternative splicing of both β (Olcese et al., 1994; Qin et al., 1996) and α1 subunits (Krovetz et al., 2000; Pan and Lipscombe, 2000). In this study, we demonstrate for the first time that alternative splicing of the N terminus of the β4 subunit alters Ca2+ channel gating and that this effect is specific to α1A and α1B subunits.
MATERIALS AND METHODS
Human spinal cord library screening. Calcium channel β4 subunits were isolated from an oligo-dT and random-primed human spinal cord λgt11 5′-Stretch Plus cDNA library (Clontech, Palo Alto, CA) using a nonradioactive digoxigenin-labeling and colorimetric detection system (Roche Molecular Biochemicals, Indianapolis, IN). The library was constructed from mRNA isolated from whole spinal cords pooled from 26 male and female Caucasians, ages 16–75 years, who died of sudden death syndrome. The insert size range of the library was 0.8–7.0 kb (average size 1.7 kb). Plaque-purified phage DNAs were isolated using a Lambda Prep Kit (Qiagen, Santa Clara, CA) and digested with the restriction endonuclease EcoRI (all endonucleases used were from Roche Molecular Biochemicals). All cDNA isolates were ligated into pBluescriptII (Stratagene, La Jolla, CA) for PCR-based cycle sequencing (FS chemistry; PE Biosystems, Foster City, CA) with universal and custom internal primers (Genosys, The Woodlands, TX). Sequences were obtained using an ABI Prism 310 Genetic DNA analyzer, and data were analyzed using ABI Prism DNA Sequencing Software (Version 2.12; PE Biosystems). Sequence comparisons, alignments, and restriction maps were performed using Lasergene Software (DNA Star, Madison, WI).
The library-screening process was initiated with a human brain β4 cDNA probe obtained from the National Center for Biotechnical Information dbEST database (1.5 kb human fetal brain β4 fragment; GenBank number R15035). Of nine first-round β4 cDNAs isolated, the 1.6 kb β4-7 clone was the largest, extending from nucleotide 216 to beyond an in-frame stop codon (the human brain β4 cDNA, GenBank number U95020, was used as a reference for all β4 nucleotide and amino acid positions). The β4-7 clone contained 134 nucleotides of 5′ untranslated sequence. A second round of screening, using a probe consisting of the N-terminal portion of β4-7 from an internal BamHI site (550) to the 5′ untranslated region, yielded seven additional β4 cDNAs, β4-15 to β4-22. Clone β4-17 possessed an in-frame start codon and novel exon 1 sequence but lacked the last 33 nucleotides of the human brain β4C-terminal coding sequence. Therefore, to create a full-length β4 cDNA, the N terminus of the β4-17 clone from the BamHI site at nucleotide position 550 to the BamHI site in the pBluescript II was ligated into a BamHI-prepared β4-7 clone. Sequence analysis was used to confirm that the β4-17/7 ligation occurred in the proper orientation. This full-length β4cDNA was referred to as β4a (GenBank numberAY054985). We used RT-PCR to isolate the previously published human brain β4 N terminus (U95020). A 694 bp fragment was obtained using a commercially available RT-PCR kit (Stratagene), custom oligonucleotide primers (β4 25F: 5′-CTCCGCCCACCGCACACG; β4 719R: 5′-CTAACACCACCGGACGCAT), and human spinal cord poly(A+) RNA (Clontech). Complete sequence analysis determined that the 694 bp fragment was identical to the U95020 N terminus, that it contained a start codon, and that it extended beyond the BamHI restriction site at position 550. Therefore, to make a second full-length β4subunit, this fragment was cloned into a BamHI-prepared pBluescriptII SK+ vector containing β4-7. Sequence analysis was used to confirm correct reading frame and proper N-terminal orientation. This full-length β4cDNA was referred to as β4b.
Construction of β4ΔN,β4aΔC,β4bΔC, and β4ΔN/ΔC deletion mutants. A β4 cDNA lacking exon 1 (β4ΔN) was obtained by using PCR to replace exon 1 of β4a with an idealized Kozak sequence (Kozak, 1991) and start codon. Custom oligonucleotide primers β4ΔNF (5′-GCCACCATGG-GTTCAGCGGATTCC), containing the Kozak sequence and start codon and beginning at nucleotide 215, and β4 719R were used in a PCR reaction with the β4-17 clone as template to generate the fragment, β4NT(−). This fragment was then cloned into the BamHI-prepared β4-7 cDNA and sequenced to confirm correct reading frame and proper N-terminal orientation. The β4aΔC, β4bΔC, and β4ΔN/ΔC cDNAs were obtained by using PCR to remove the C-terminal nucleotide sequence 3′ to nucleotide 1286 (corresponding to amino acid position 404). Custom oligonucleotide primers β4 849F (5′-GCTGACATTTCTCTTGCTAA upstream of a unique BglII site) and β4ΔCR (5′-TCAGGTTGTGTG-GGTGGCAC, which ended at β4 nucleotide 1286 and included an in-frame stop codon) were used in a PCR reaction with the β4-17 clone as template to generate the truncated fragment, β4C(−). This fragment was then cloned into the pT-Advantage vector (Clontech) and sequenced to determine correct orientation. The β4C(−) fragment was then cut with BglII and XhoI (from pT-Advantage poly-linker) and cloned into BglII- andXhoI-prepared β4a, β4b, and β4ΔN cDNAs. The resulting cDNAs were then sequenced with internal primers flanking the C-terminal deletion to confirm sequence orientation and fidelity.
The BI-2 (α1A) and α2a/δ-1 clones used in this study were provided by T. Tanabe (Tokyo Medical and Dental University, Tokyo, Japan). The rat α1B and rabbit α1C clones were kindly provided by D. Lipscombe (Brown University, Providence, RI) and E. Perez-Reyes (University of Virginia, Charlottesville, VA), respectively.
Electrophysiology and data analysis. Complementary RNAs (cRNAs) were synthesized in vitro using Ambion's mMessage mMachine RNA transcription kit [T3 or T7 depending on clone orientation in pBluescript II S/K+ or pBSTA (α1B)]. Standard Xenopus laevis oocyte expression methods were used to characterize β subunit splice variants. Briefly, full-length α1, α2/δ, and β cRNAs were injected in equimolar ratios (5.6 ng α1A or α1B, 2.4 ng α2/δ, and 1.6 ng β in 46 nl; 17 ng α1C or α1E, 7 ng α2/δ, and 5 ng β in 50 nl) into defolliculated oocytes (stage V–VI). (The α2δ-1 subunit was used in this study.) Calcium channel currents were recorded 2–8 d after oocyte injection by standard two-electrode voltage clamp using a Warner amplifier (OC-725B) at 20–22°C, and data were collected using pCLAMP6 software (Axon Instruments, Foster City, CA). Microelectrodes were filled with 3 m KCl, and the resistances of the current and voltage electrodes were 0.3–1.5 MΩ. Data were filtered at 2 kHz and sampled at 10 kHz. Currents were recorded in a chloride-free bath containing 5 mmBa(OH)2, 5 mm HEPES, 85 mm TEA-OH, and 2 mm KOH, pH adjusted to 7.4 with methansulfonic acid (α1Aand α1B), or 40 mmBa(OH)2, 5 mm HEPES, 85 mm TEA-OH, and 2 mm KOH, pH adjusted to 7.4 with methansulfonic acid (α1Cand α1E). Currents used to generate the data in this study ranged from 0.5 to 2.9 μA. For activation and inactivation experiments, the average current sizes for α1Aand α1B complexes containing either β4a or β4b were 1.2 and 1.6 μA, respectively. Leak currents were between 20 and 100 nA. Only recordings with minimal tail currents were used for each data set (see representative traces in Fig. 5). Data were analyzed using pCLAMP6 software (Axon Instruments) and Excel 7.0 (Microsoft Corp., Redmond WA). The leak and capacitive currents were subtracted on-line using a standard P/4 protocol. Boltzmann fits to the activation and inactivation data were performed using Sigma Plot version 5.0 (SSPS Inc., Chicago IL) with the equations %IBa = 1/[1 + exp(−(Vtest −V1/2)/k)] and %IBa = 1/ [1 + exp((Vpre −V1/2 )/k)], respectively, whereVtest = I–V test potential, Vpre = prepulse potential,V1/2 = midpoint of activation or inactivation, and k = slope factor. An estimate of gating charge, z, was calculated by dividing 25 (approximate value for RT/F at room temperature, whereR = gas constant, T = temperature, andF = Faraday constant) by the slope factor. Statistical analysis was performed with a Student's two-sample equal variance t test with a two-tailed distribution (Microsoft Excel 97 SR-2). Data are presented as mean ± SEM.
RESULTS
Cloning of a Ca2+ channel β4subunit with an N terminus similar to that of β3subunits
Two β4 subunit N-terminal splice variants, β4a and β4b (Fig.1), are the focus of this study. Both were isolated from a human spinal cord cDNA library using routine screening techniques. The amino acid sequence of the β4b variant is identical to a previously published sequence (GenBank number U95020), whereas this is the first reporting of the β4a sequence. The difference in the two variants lies solely in the nucleotide sequence of exon 1, the translated region of which is referred to as domain D1 (Birnbaumer et al., 1998). The remaining sequence of both β4a and β4b is composed of 1410 nucleotides that encode the 470 amino acids of domains 2–5 (data not shown). As shown in Figure 1, exon 1 of β4a encodes a 15 amino acid sequence that is highly homologous to the N-terminal sequences of several previously identified Ca2+ channel β3 subunits. This indicates that β4a exon 1 must have been present in the genome before the time that an ancestral gene duplicated to form distinct β3 and β4 genes. Interestingly, amino acids 5–11 (LYLHGIE) are identical to those found in the Xenopus β subunit, xβ32, but quite divergent from the same region of the human β3 subunit. This could imply that a particular function of this sequence has been purposely conserved throughout evolution. Also of note in the human β4asequence are two D to N conversions at positions 4 and 12 (asterisks) that eliminate two negative charges that appear to be highly conserved among β3 subunits. Figure 1 also demonstrates that D1 of β4a is not at all homologous to D1 of β4b. It can be seen, however, that D1 of β1b and β4b are more closely related than D1 of β4a and β4b. Domain 1 of β4bcontains 49 amino acids, 2 of which are negatively charged, and 8 of which are positively charged. Six of these positive charges are clustered in the center of the sequence close to consensus sites (TTR and TRR) for phosphorylation by protein kinase C. No further Prosite-listed consensus sites were found in the D1 sequences of either β4a or β4b.
Sequence comparisons of human spinal cord Ca2+ channel β4a and β4bsubunits and other β subunit subtypes. Top, The amino acid sequence of domain 1 and a short segment of domain 2 of the human β4a subunit (hβ4a) is shown aligned with comparable domains of two Xenopusβ3 subunits (xβ32 andxβ28 ) (Tareilus et al., 1997) and a human β3 subunit (hβ3 ). Amino acids identical to the hβ4a sequence are boxed. Asterisks denote D to N amino acid conversions in the human β4a sequence.Bottom, The amino acid sequence of domain 1 and a short segment of domain 2 of the human β4b subunit (hβ4b ) is shown aligned with comparable domains of the human β1b subunit. Identical amino acids are boxed. Dashed lines indicate gaps in the sequence. The bar denotes consensus sites for phosphorylation by protein kinase C.
Alternative splicing of the β4 subunit N terminus affects Ca2+ channel expression
Critical to the interpretation of our expression data is the fact that some populations of Xenopus oocytes have been shown to express low levels of an endogenous β3-like subunit that is capable of binding to and altering the gating properties of injected α1 subunits (Tareilus et al., 1997). To test for this possibility in our oocytes, we conducted experiments in which we measured the time required for α1A/α2δ, α1A/α2δ + β4a, and α1A/α2δ + β4b complexes to reach levels of expression that we thought suitable for electrophysiological recording (1 μA of peak current). Figure 2 shows that channel complexes containing β4b expressed at a much faster rate than those containing β4a, reaching adequate levels within 1–2 d. Complexes containing β4a took 3–4 d to reach similar levels, whereas complexes that did not contain a β subunit required 7–8 d to express 1 μA of current. Similarly, α1B/α2δ + β4b complexes reached adequate levels in 1–2 d, whereas α1B/α2δ + β4a complexes took 3–4 d to reach similar levels. α1B complexes expressed without β4 subunits did not reach suitable current size until day 7–8 (data not shown). α1C/α2δ and α1E/α2δ expressed with either β4a or β4breached adequate current size in 6–8 d, whereas complexes without β4 subunits showed no appreciable current even after 8 d. Expression rates and levels for α1C/α2δ and α1E/α2δ + β4a and β4b were essentially identical (data not shown). As shown in Figure 2, a sixfold increase in the amount of β subunit cRNA injected into oocytes relative to that of α1A did not affect expression rates or levels, suggesting that β subunit binding sites on α1A are saturated even when the two subunits are coinjected at a 1:1 ratio. This is consistent with the findings ofQin et al. (1996). We concluded from these experiments that the endogenous Xenopus β3-like subunit would not significantly influence the examination of exogenous currents measured in the 2–6 d time period.
Expression rates of α1ACa2+ channel complexes with different β subunit compositions. Peak currents elicited by depolarization to +10 mV (α1A/α2δ), +5 mV (α1A/α2δ + β4a), or 0 mV (α1A/α2δ + β4b) from a holding potential of −80 mV are plotted against days after injection. Barium (5 mm) was the charge carrier. Oocytes were maintained in ND96 culture media at 18°C. Comparisons between experiments in which the β4aor β4b subunits were injected at 1:1 (1×) or 6:1 (6×) ratios relative to the α1A are shown. Each data point represents a minimum of six recordings. The SEM for each point is shown unless the values were smaller than the symbol.
Alternatively spliced β4 subunits have α1-subunit subtype-specific effects on voltage-dependent activation and inactivation
To determine whether β4 N-terminal splicing affected Ca2+ channel gating properties, we expressed either β4a or β4b with rabbit α2δ and with rabbit α1A (BI-2) (Mori et al., 1991), rat α1B (Δ21 α1B) (Pan and Lipscombe, 2000), rabbit α1C (Mikami et al., 1989), or marine ray α1E (doe-1) (Horne et al., 1993) in Xenopus oocytes. (The α2δ-1 subunit is included in all experiments in this study.) Figure3A,B,E,Fshows comparisons of normalized current–voltage (I–V) curves for the four different α1 subunits expressed with either β4a or β4b. Figure 3,A and B, illustrate that the peaks of the current–voltage curves for α1A and α1B complexes containing β4b were shifted to more hyperpolarized potentials relative to complexes containing β4a. In contrast, Figure 3, E andF, shows that the I–V curves for α1C and α1E complexes containing either β4a or β4b were essentially superimposed. The difference in α1 subunit responsiveness was not caused by differences in charge carrier concentrations used in the experiments (5 mmBa2+ for α1A and α1B; 40 mmBa2+ for α1C and α1E), because we observed identical hyperpolarizing shifts for both α1A and α1B with β4b, even in 40 mm Ba2+ (data not shown). We concluded from these first experiments that alternative splicing of the β4 subunit N terminus affects activation of Ca2+ channel complexes containing α1A and α1Bsubunits but not those containing α1C or α1E. To estimate theV1/2 of activation for the different α1A and α1Bcombinations, we averaged Boltzmann fits to theI–V data generated over the range of −40 to +10 mV for α1A complexes and −40 to +20 mV for α1B complexes containing either β4a or β4b (Fig.3C,D). The results show that theV1/2 of activation for both α1A and α1B complexes containing β4b were shifted to the left relative to complexes containing β4a by ∼5 mV and ∼7 mV, respectively (Table 1). The results also show that the slopes of the β4bfits were somewhat steeper than for β4a.
β4a and β4b subunits have α1 subunit subtype-specific effects on the voltage dependence of activation. A, B,E, F, Normalized, averaged peak current–voltage (I–V) plots for α1A (A), α1B(B) α1C (E), and α1E (F) coexpressed with α2δ and either β4a or β4b.A, B, The α1A (BI-2) and α1B (Δ21) subunits used in these and subsequent experiments are those described by Mori et al. (1991) and Pan and Lipscombe (2000), respectively. Currents were activated by 300 msec depolarizations to various test potentials (−40 to +40 mV in 5 mV increments) from a holding potential of −80 mV. Barium (5 mm) was the charge carrier for both α1A and α1B. C, D, Voltage dependence of activation up to +10 mV for α1A(C) and +20 mV for α1B(D) as determined from averagedI–V data in A andB. Data points represent the means of the normalized data at a given membrane potential. The SEM for each point is shown unless the values were smaller than the symbol. Smooth curves represent a single Boltzmann fit to the averaged data. Values for V1/2 and k for α1A and α1B plus α2/δ and either β4a or β4b are listed in Table 1. E,F, The α1C (cardiac) and α1E(doe-1) subunits used in these and subsequent experiments are those described by Mikami et al. (1989) and Horne et al. (1993), respectively. Currents were activated by 300 msec depolarizations to various test potentials (−40 to +40 mV in 5 mV increments) from a holding potential of −80 mV (α1C + β4a, n = 12; α1C + β4b,n = 13; α1E + β4a, n = 9; α1E + β4b,n = 9). Barium (40 mm) was the charge carrier.
Effects of β4 subunit alternative splicing and N- and C-terminal deletions on voltage-dependent activation and inactivation of α1A and α1BCa2+ channel subunits
We next examined whether alternative splicing of the β4 subunit affected isochronal inactivation. We used a 20 sec conditioning prepulse over a wide range of potentials followed by a 300 msec test pulse to near-peak potentials to generate the data. Figure4A–D shows that, as was the case for activation, alternative splicing of the β4 subunit N terminus affects inactivation of Ca2+ channel complexes containing α1A and α1B subunits but not those containing α1C or α1E. The figure illustrates that the voltage dependence of inactivation of both α1A (Fig.4A) and α1B (Fig.4B) complexes containing β4bwas shifted to more hyperpolarized potentials relative to complexes containing β4a. In contrast, inactivation curves for α1C (Fig. 4C) and α1E (Fig. 4D) complexes containing β4a or β4b were essentially identical. The Boltzmann-derivedV1/2 for inactivation of both α1A and α1B complexes containing β4b were shifted to the left relative to complexes containing β4a by ∼10–11 mV (Table 1). Interestingly, the hyperpolarizing shift inV1/2 for α1Acomplexes (Fig. 4A) occurred as the result of a parallel shift in the voltage dependence of inactivation, whereas for α1B complexes (Fig. 4B), the shift in V1/2 occurred primarily as the result of a change in slope. Slope factors for α1B complexes containing β4a and β4b complexes were ∼14 and 7 mV, respectively (Table 1).
β4a and β4b subunits have α1 subunit subtype-specific effects on the voltage dependence of inactivation. A–D, Normalized, averaged isochronal inactivation curves for α1A (A), α1B(B), α1C (C), and α1E (D) coexpressed with α2δ and either β4a or β4b. Curves were generated from peak currents elicited by a 300 msec test depolarization to +5 mV (α1A + β4a), 0 mV (α1A + β4b), +10 mV (α1B + β4a), +5 mV (α1B + β4b), or +20 mV (α1C and α1E with β4a and β4b) after a 20 sec conditioning prepulse to voltages ranging from −80 to +30 mV (A,C, D) or −100 to +10 mV (B). Barium (5 mm for α1A and α1B; 40 mm for α1C and α1E) was the charge carrier. Data points represent the means of the normalized data at a given membrane potential. The SEM for each point is shown unless the values were smaller than the symbol. Smooth curves represent a single Boltzmann fit to the averaged data. Values forV1/2 and k for inactivation of α1A and α1B plus α2δ and either β4a or β4b are listed in Table1.
Because α1C and α1Esubunits were not affected by alternative splicing of β4 subunits, we next directed our experiments toward characterizing the α1A and α1B responses in more detail. Figure5 shows representative current traces of α1A (Fig. 5A) and α1B (Fig. 5B) complexes containing either β4a (top) or β4b (bottom) expressed inXenopus oocytes. Traces shown were generated by step depolarization to −10, 0, 10, 20, and 30 mV. The arrowsindicate that the potentials at which peak currents were reached varied with each complex. Regardless of the α1 subunit subtype, however, complexes containing β4ainactivated faster than those containing β4b,with a difference in rates being more apparent for complexes containing α1B. Figure 5, C and D, shows the averaged currents remaining after 300 msec (R300 ) step depolarizations to each potential for α1A and α1B, respectively. The results indicate that the rate of inactivation for all four complexes is voltage dependent and that the differences in rates between complexes containing β4a versus β4b become apparent primarily with depolarizations beyond 0 mV.
α1A and α1B complexes containing β4a inactivate faster than those containing β4b. A, B, Representative current traces of α1A (A) and α1B (B) plus α2δ and either β4a (top) or β4b(bottom). Currents were elicited by step depolarizations to a range of test potentials (−10 to +30 mV in 10 mV increments) from a holding potential of −80 mV. Barium (5 mm) was used as the charge carrier. Traces were fit with a single exponential from 25 msec beyond the peak inward current to the end of the depolarization. Averages of τinactivation at the peak current potential were α1A + β4a, 226.6 ± 12.5 msec (n = 12); α1A + β4b, 307.2 ± 19.2 msec (n = 10); α1B + β4a, 160.1 ± 20.0 msec (n = 10); α1A + β4b, 213.9 ± 15.6 msec (n = 10). C, D, Current remaining at the end of a 300 msec test pulse (R300), elicited as in the protocol above, for α1A (C) and α1B (D) plus α2δ and either β4a or β4b. The SEM for each bar is shown. Asterisks denote statistical significance (p < 0.05) as determined by a Student's two-sample equal variance t test.
α1 subunit-specific responses to β4subunit N- and C- terminal deletions
The results to this point indicated that the N terminus of the β4 subunit plays an important role in setting the kinetics and voltage-dependence of Ca2+ channel gating, with some differences in responsiveness noted between α1A and α1B subunits. We next sought to determine whether the β4 N terminus could be acting in concert with the β4 C terminus to exert its effects on gating. Because previous studies had shown that the β4 C terminus binds directly to the α1A subunit (Walker et al., 1998, 1999), it was of particular interest to determine whether the gating properties of α1A would change in comparison to α1B if the β4 C terminus were deleted. To address this issue, we made four β4 subunit deletion constructs that along with β4a and β4b provided us with all the possible +/− combinations of β4N- and C termini (Fig.6A). We found that all four constructs augmented Ca2+ channel expression to a level that was comparable to or exceeded (i.e., β4ΔNΔC) the expression levels we observed with β4b. The effects of these constructs on activation and inactivation of α1A and α1B subunits are shown in Figure 6,B and C, and Figure7, A and B, respectively. (Our initial results with β4a and β4b are included as dashed lines for reference in Figs. 6 and 7). Interestingly, it was readily apparent from both the activation and inactivation results shown in Figures 6and 7 that despite testing six different β4subunit constructs, our data could be grouped into two activation modes, A1 and A2(α1A and α1B), and two (α1B) or three (α1A) inactivation modes, I1–I3, on the basis of the curve position alone. As can be seen from the data, the distinction between activation and inactivation modes was most clearly delineated in experiments involving α1B(Figs. 6C, 7B). Table 1 shows that the distinction between modes is quite evident when comparing Boltzmann-derived values for V1/2 and slope factor, and along with Figures 6 and 7 reveals that the β4 subunit constructs responsible for setting each mode differ between α1A and α1B subunits.
Effects of β4 subunit N- and C-terminal deletions on the voltage dependence of activation of α1A and α1B Ca2+channels. A, Schematic diagrams of the wild-type and artificial β4 subunits used in this series of experiments. The 15 amino acid β4a and 49 amino acid β4b N termini (alternatively spliced forms of domain 1) are denoted by filled and open bars, respectively. Domains 2–4 are represented by a singlecross-hatched bar. The C terminus (domain 5) is denoted by a diagonally striped bar. B, C, Voltage dependence of activation up to +10 mV for α1A(B) and +20 mV for α1B(C) as determined from averagedI–V data. Data points represent the means of the normalized data at a given membrane potential. The SEM for each point is shown unless the values were smaller than the symbol.Smooth curves represent a single Boltzmann fit to the averaged data. Broken curves represent activation data shown in Figure 3, C and D, and are included in this figure for reference. Values forV1/2 and k for α1A and α1B plus α2/δ and each of the six β4constructs are grouped according to curve similarities in Table1.
Effects of β4 subunit N- and C-terminal deletions on the voltage dependence of inactivation of α1A and α1B Ca2+channels. A, B, Normalized, averaged steady-state inactivation curves for α1A(A) and α1B(B) coexpressed with α2δ and one of the six β4 constructs shown in Figure6A. Curves were generated from peak currents elicited by a 300 msec test depolarization to −5 mV (α1A+ β4ΔNΔC), 0 mV (α1A + β4b, β4aΔC), +5 mV (α1A + β4a, β4ΔN, and β4bΔC; α1B + β4b and β4ΔN), +10 mV (α1B + β4a, β4ΔNΔC), or +15 mV (α1B + β4aΔC and β4bΔC) after a 20 sec conditioning prepulse to voltages ranging from −80 to +10 mV (A) or −100 to +10 mV (B). Barium (5 mm) was the charge carrier for both α1A and α1B. Data points represent the means of the normalized data at a given membrane potential. The SEM for each point is shown unless the values were smaller than the symbol. Smooth curves represent a single Boltzmann fit to the averaged data. Values forV1/2 and k for inactivation of α1A and α1B plus α2δ and each of the six β4 constructs are grouped according to curve similarities in Table 1.
The details of the deletion results are best understood by examining in sequence the data that we obtained with individual β subunit constructs. Our first experiments were directed toward determining what effect deletion of both the β4 N and C termini (β4ΔNΔC) would have on α1A and α1B gating properties. Unexpectedly, both α1A and α1B complexes containing the β4ΔNΔC subunit had activation properties very similar to complexes containing full-length β4b (Fig.6B,C, modeA1 ). This indicated that α1 subunits could not distinguish β4 subunits without an N or C terminus from β4 subunits with the longer form of N terminus and the C terminus present. Relative to α1complexes containing β4a, however, β4ΔNΔC caused a 6–7 mV hyperpolarizing shift and a slight increase in slope of activation of both α1A and α1B (Table 1). Figure 7, A and B, shows that, although the inactivation curve for α1A complexes containing β4ΔNΔC fell between those for complexes containing β4a and β4b, the inactivation properties of α1B complexes containing β4ΔNΔC and β4b were also indistinguishable. For both α1A and α1B, it can be seen that relative to complexes containing β4a, β4ΔNΔC caused a qualitatively similar hyperpolarizing shift in the voltage dependence of inactivation and decrease in slope (shift from mode I2 to mode I1). As shown in Figure 7B, this effect was most dramatic for α1B complexes, where relative to β4a, β4ΔNΔC caused a ∼10 mV hyperpolarizing shift in inactivation and a nearly 50% decrease in slope (Table 1).
We next characterized the effects of the construct β4ΔN (β4ΔNΔC plus the β4 C terminus) on the gating properties of α1A and α1B subunits. Interestingly, as shown in Figure 6, A and B, the β4ΔN construct had different effects on activation of α1A as compared with α1B. Although addition of the C terminus had a depolarizing effect on α1A activation relative to β4b and β4ΔNΔC, there was no change in the activation properties of α1B. Moreover, as can be seen in Figure6B and Table 1, the activation properties of α1A complexes containing β4ΔN were essentially identical to those containing β4a (mode A2). Similarly, as shown in Figure 7, A and B, β4ΔN, like β4a, had a noticeable depolarizing effect on α1Ainactivation (mode I2) relative to complexes containing β4ΔNΔC but caused no change in the inactivation properties of α1B. These results indicated that, at least in the absence of the N terminus, the β4 C terminus has α1Asubunit-specific effects on the voltage dependence of both activation and inactivation.
To define further the role of the β4 N termini in gating, we next characterized the effects of two constructs, β4aΔC and β4bΔC, that lacked the β4 C terminus but contained the N termini of β4a and β4b, respectively (β4ΔNΔC plus β4a or β4b N terminus). Interestingly, the pattern of results that we obtained with these constructs in many respects was just the opposite of what we saw with β4ΔN. Although β4ΔN had α1Asubunit-specific effects on gating, β4aΔC and β4bΔC had, for the most part, α1B subunit-specific effects. Figure6A shows that relative to β4ΔNΔC, β4bΔC, like β4a, caused a depolarizing shift in activation of α1A subunits, but β4aΔC was without effect. Figure7A shows that relative to β4ΔNΔC, neither β4aΔC nor β4bΔC had effects on inactivation of α1A subunits. In contrast, Figures 6B and 7B show that relative to β4ΔNΔC both β4aΔC and β4bΔC caused a depolarizing shift in activation and inactivation of α1B. Moreover, the gating properties of α1B complexes containing β4aΔC or β4bΔC were essentially identical to those containing β4a(modes A2 and I2). The results of these experiments indicate that, at least in the absence of the C terminus, the β4b but not the β4a N terminus has effects on α1A activation, whereas neither affects α1A inactivation. In contrast, both the β4a and β4b N termini have effects on α1B activation and inactivation.
With the results from the deletion experiments, it was informative to reexamine the data from our initial experiments (Figs. 6 and 7,dashed lines) with the idea that full-length β4a and β4b subunits were constructed by adding back the β4a and β4b N termini to β4ΔN. As the data reveals, this also had α1 subunit-specific effects on both activation and inactivation. With respect to activation, Figure6A shows that, relative to β4ΔN, adding back the β4a N terminus had no effect on α1A activation, suggesting that the short form of N terminus could not overcome the α1A-specific β4C-terminal effect noted with β4ΔN previously. Adding back the β4b N terminus, however, did supercede the C-terminal effect and caused a hyperpolarizing shift in α1A activation relative to β4ΔN (back to mode A1). In contrast, Figure 6B shows that adding back the β4a N terminus to β4ΔN caused a depolarizing shift in the activation of α1B (back to mode A2), but adding back the β4b N terminus had no effect. This goes along with the β4ΔN data showing that with α1B there is no β4C-terminal effect to overcome, and that the β4aN terminus alone causes an α1B-specific depolarizing shift in activation. With respect to inactivation, Figure7B shows that, as was the case for activation, adding back the β4a N terminus to β4ΔN had little effect on α1A inactivation, but adding back the β4b N terminus caused a significant hyperpolarizing and, in this case, parallel shift in the curve for α1A inactivation (mode I3). It is worth noting that this shift is different from and goes beyond the curve for β4ΔNΔC, and that this effect on α1A gating is unique to β4b. Figure 7B shows that, as expected from the β4aΔC results with α1B, without a C-terminal effect to overcome, adding back the β4a N terminus to β4ΔN caused a depolarizing shift in α1B inactivation (back to mode I2). Not expected, however, was the result that adding back the β4b N terminus had no effect, recalling that the β4b N terminus alone causes a depolarizing shift in α1B inactivation. This suggests that the presence of the β4 C terminus, although not having effects on its own, interferes in some way with the ability of the β4b N terminus to influence α1B channel gating.
DISCUSSION
Our results provide the first evidence that alternative splicing of the β4 subunit alters Ca2+ channel gating and that this effect is specific to α1A and α1B subunits. The physiological relevance of our findings lies in the fact that α1A, α1B (Westenbroek et al., 1998), and β4 subunits (Wittemann et al., 2000) colocalize in nerve terminals and that α1A and β4 (Liu et al., 1996) and α1B and β4 subunits (Scott et al., 1996) are directly associated. In many respects, our experiments were similar to those of Olcese et al. (1994) and Qin et al. (1996), which characterized the effects of various β subunit splice variants, chimeras, and deletion mutants on human α1E subunit gating. Their studies yielded five results pertinent to our findings. (1) Relative to α1E alone, all β subunit constructs tested caused a nearly identical hyperpolarizing shift in theV1/2 of activation and decrease in slope factor [see also Jones et al. (1998)]. (2) Deletion of the N terminus of the β1b, β2a, and β3 subunits had no effect on the fast component of activation. (3) Alternative splicing of the N terminus, C terminus, and internal domain 3 of β1 and β2 subunits had opposing effects on the V1/2 of steady-state inactivation but did not affect slope. (4) Deletion of the N terminus of the β1b and β3 subunits caused a depolarizing shift in theV1/2 of inactivation without affecting slope, whereas deletion of the N terminus of the β2a subunit caused a hyperpolarizing shift. (5) C-terminal alternative splicing did not affect gating properties. The principal conclusion of these experiments was that, independent of effects on activation, the N terminus of the β subunit plays a dominant role in governing the voltage sensitivity of α1E subunit inactivation. This suggested toOlcese et al. (1994) that there were two separate α1 and β subunit interaction sites regulating activation and inactivation.
Our results point similarly to the N terminus of the β4 subunit as a key determinant of α1A and α1B gating properties but show some dissimilarity to the five α1E results listed above. (1) Unlike α1E with β1–β3 subunits, alternatively spliced β4 subunits had differential effects on activation of both α1Aand α1B. Relative to the short β4a N terminus, the longer β4b form caused a hyperpolarizing shift in activation of both α1A and α1B subunits (but not α1C or α1E). (2) Relative to β4b, deletion of the N terminus of the β4 subunit caused a depolarizing shift in activation of α1A but not α1B. (3) Alternative splicing of the β4 subunit affected both theV1/2 and slope of inactivation of α1B, whereas only shifting theV1/2 of α1Ainactivation without a change in slope. Alternative splicing of the β4 subunit did not affect inactivation of α1C or α1E. (4) Relative to β4b, deletion of the N terminus of the β4 subunit caused a depolarizing shift in inactivation of α1A but not α1B. (5) C-terminal deletion experiments revealed that the β4 N and C termini work in concert to set gating parameters of α1A and α1B subunits. Taken together, these results indicate that alternatively spliced β subunits can affect both activation and inactivation of Ca2+channels, and the responsiveness of Ca2+channels to β subunit splicing varies with α1subunit subtype.
To explain our results, we devised a structural model for potential α1–β4 subunit domain interactions based on a β subunit modular structure (domains A–E) (Hanlon et al., 1999) and actual molecular weights of the potential α1A and β subunit domains involved (Fig.8). Although highly speculative, the model integrates related structure–function results from a number of different laboratories that point to the β subunit D domain interaction with the α1 subunit I–II linker as a key determinant of Ca2+ channel gating properties (Herlitze et al., 1997; Bourinet et al., 1999; Stotz et al., 2000; Berrou et al., 2001). Moreover, it incorporates results showing that regulation of activation and inactivation are separable functions of β subunits (Olcese et al., 1994). Of particular relevance to our model are studies showing that the β1 subunit D domain was all that was required to reproduce the inactivation rate of L-type channels coexpressed with full-length β1(Cens et al., 1999) and that a single point mutation (R378E) in the β subunit binding site of the α1 I–II linker (AID domain) had a depolarizing effect on the voltage dependence of both activation and inactivation of an α1Esubunit (Berrou et al., 2001).
Potential α1A and β subunit domain interactions as viewed from inside the cell looking out through the pore. Top, α1A alone. Transmembrane domains I–IV are represented as gray circlesand intracellular domains as white circles. Middle, α1A + β4b. The β4b subunitA–E domains are shown as black circlessuperimposed on α1A. Bottom, α1A + β4a. The radius of eachcircle was calculated from the spherical volume (V = 4/3 πr3) of each subunit domain, where V = [(0.73 cm3/gm × 1024Å3/cm3 × molecular weight)/6.02 × 1023] and the average molecular weight of an amino acid is 120 Da. For α1A(BI-2): N terminus, 98 aa; transmembrane domains I–IV, 229–268 aa; I–II linker, 127 aa; II–III linker, 537 aa; III–IV linker, 54 aa; C terminus, 604 aa. For β4b [nomenclature as in Hanlon et al. (1999)]: A domain, 92 aa; B domain, 61 aa; C domain, 37 aa; D domain, 210 aa; E domain, 144 aa. For β4a: A domain, 44 aa. Interactions of the β4 D domain with the α1A I–II linker (Pragnell et al., 1994) and β4 E domain with α1A N and C termini have been well documented (Walker et al., 1998, 1999). Dashed arrow in the bottomdiagramindicates the potential for a conformational change when the β4a N terminus is substituted for the β4b N terminus.
Previous studies have shown that the D domain of the β4 subunit binds with high affinity to the α1A I–II linker (Pragnell et al., 1994) and that the E domain of the β4 subunit binds to both the N and C terminus of the α1A subunit (Walker et al., 1998, 1999). As shown in Figure 8, this indicates that the D and E domains likely establish the N- to C-terminal orientation of the β4 subunit relative to the α1A subunit. Although little is known about β subunit N-terminus interactions with the α1subunit, a modular structure for the β subunit A, B, and C domains suggests that the interactions could occur over a wide range. As suggested in Figure 8, a change in the size and sequence of the β subunit A domain could have an effect on the way the β subunit D domain interacts with the α1 I–II linker. Such a change might be responsible for the different gating properties observed between Ca2+ channel complexes containing β4a versus β4b.
The salient feature of the model is that a core β subunit structure encoded by exons 2–12 (β4ΔNΔC), through interactions with the α1 subunit I–II linker, sets separate default parameters for α1activation and inactivation (Table 2, mode A1I1). This mode likely represents a specific α1 I–II linker conformation that, through its connection to the α1 IS6 transmembrane domain, influences the mobility of the gating charges within α1 IS4 (Zhang et al., 1994). [The effects of β4N-terminal alternative splicing on apparent gating charge (zvalue) are shown in Table 1. Note the significant difference in calculated z values during inactivation of α1B complexes containing β4a versus β4b.] As shown in Table 2, in mode A1I1 the two presumed α1 and β subunit interaction domains are in alignment (α1, filled symbols; β, open symbols). Changes from default parameters occur when either the β subunit N or C terminus, or both, interacts with, or is acted on by, other regions of the α1 subunit such that the I–II linker changes mode conformations. For example, in mode A2I2 (row 2), the two presumed α1 and β subunit interaction domains would be out of alignment. Displacing the β subunit D domain in either the C- or N-terminal direction would enable mode 2 conformation, whereas a balance of these two forces favors mode 1.
Gating mode model describing the effects of β4 subunit constructs on α1A and α1B subunit activation and inactivation
Describing our data in terms of the model (Table 2), the default activation and inactivation parameters of α1Aand α1B complexes containing β4ΔNΔC are denoted as mode A1I1 (row 1). Steep activation and shallow inactivation typify this mode. Row 2 shows that adding back the C terminus to β4ΔNΔC (β4ΔN) shifts α1Acomplexes to mode A2I2, whereas α1B complexes remain in mode A1I1. This could be explained by the α1A–β4 C-terminal binding event described by Walker et al. (1998, 1999) causing the β4 D domain to be displaced in the C-terminal direction. Relative to mode A1I1, activation in mode A2I2 is shallower and inactivation is steeper. Rows 3 and 4 show that by adding back the N terminus, α1B complexes containing either the β4aΔC or β4bΔC construct shift to mode A2I2. This shift might be explained by β4N-terminal-α1B interactions causing the β4 D domain to be displaced in the N-terminal direction. Alternatively, steric changes resulting from the presence of the N terminus may shift the β4 D domain in the C-terminal direction (row 4, β4bΔC(α1B)). Whatever the cause, it is likely to be different for α1Bcomplexes containing β4aΔC versus β4bΔC. Adding back the C terminus to β4aΔC has no effect on α1B mode A2I2 (row 5), whereas addition of the C terminus to β4bΔC causes a shift to mode A1I1 (row 6). Row 4 also shows that β4bΔC causes an α1A subunit mode change that is limited to activation (A2I1). This was the one instance in our experiments in which regulation of activation and inactivation were separable functions. The addition of the C terminus to β4aΔC shifts α1A to mode A2I2 (row 5), which again is likely the result of a β4 C-terminal binding event. The addition of the C terminus to β4bΔC creates a distinct α1A mode, A1I3, characterized by steep activation and inactivation.
In conclusion, our results add to the developing picture of the intracellular domains surrounding the Ca2+channel pore being composed of modular “hot spots” for channel regulation by β-subunits, protein kinases, G-proteins, syntaxin, and calmodulin (for review, see Walker and DeWaard, 1998; Levitan, 1999). Our future experiments will be directed toward understanding how interactions between these diverse regulatory components might contribute to the dynamic molecular events giving rise to synaptic plasticity.
Footnotes
This work was supported by National Institutes of Health Grant R29-NS 32094, North Carolina Biotechnology Center Academic Research Grant 9905 ARG 0044, and a College of Veterinary Medicine State Research Support Grant. We thank Dr. Robert Rosenberg for advice on data analysis and manuscript preparation.
Correspondence should be addressed to Dr. William A. Horne, Department of Anatomy, Physiological Sciences, and Radiology, North Carolina State University College of Veterinary Medicine, 4700 Hillsborough Street, Raleigh, NC 27606. E-mail: bill_horne{at}ncsu.edu.