Distinctive Modulatory Effects of Five Human Auxiliary β₂ Subunit Splice Variants on L-Type Calcium Channel Gating

Abstract

Sequence analysis of the human genome permitted cloning of five Ca²⁺-channel β₂ splice variants (β_2a–β_2e) that differed only in their proximal amino-termini. The functional consequences of such β₂-subunit diversity were explored in recombinant L-type channels reconstituted in HEK 293 cells. β_2a and β_2e targeted autonomously to the plasma membrane, whereas β_2b–β_2d localized to the cytosol when expressed in HEK 293 cells. The pattern of modulation of L-type channel voltage-dependent inactivation gating correlated with the subcellular localization of the component β₂ variant—membrane-bound β_2a and β_2e subunits conferred slow(er) channel inactivation kinetics and displayed a smaller fraction of channels recovering from inactivation with fast kinetics, compared to β_2b–β_2d channels. The varying effects of β₂ subunits on inactivation gating were accounted for by a quantitative model in which L-type channels reversibly distributed between fast and slow forms of voltage-dependent inactivation—membrane-bound β₂ subunits substantially decreased the steady-state fraction of fast inactivating channels. Finally, the β₂ variants also had distinctive effects on L-type channel steady-state activation gating, as revealed by differences in the waveforms of tail-activation (G-V) curves, and conferred differing degrees of prepulse facilitation to the channel. Our results predict important physiological consequences arising from subtle changes in Ca²⁺-channel β₂-subunit structure due to alternative splicing and emphasize the utility of splice variants in probing structure-function mechanisms.

INTRODUCTION

High-voltage-activated (HVA) Ca²⁺ channels transduce electrical signals into diverse and essential processes such as muscle contraction, synaptic transmission, and gene expression. HVA Ca²⁺ channels are hetero-multimers comprised minimally of pore-forming α₁, and auxiliary β and α₂δ subunits. The rich variety of physiological responses mediated by HVA Ca²⁺ channels is mirrored by an equally impressive molecular diversity of individual subunits—seven genes encoding α₁ subunits (α_1A–α_1F, α_1S), four encoding β subunits (β₁–β₄), and three encoding α₂δ subunits (α₂δ-1 to α₂δ-3), have been identified (for review, see Catterall, 2000).

Auxiliary β subunits are potent determinants of Ca²⁺ channel behavior. Qualitatively, all four β-subunit isoforms act similarly to markedly increase surface membrane expression of co-expressed α₁-subunits (Chien et al., 1995; Brice et al., 1997; Gao et al., 1999; Yamaguchi et al., 2000), dramatically enhance Ca²⁺ current amplitude over that obtained with α₁ alone (Singer et al., 1991; Jones et al., 1998), and produce hyperpolarizing shifts in the voltage-dependence of channel activation (Singer et al., 1991; Perez-Reyes et al., 1992; De Waard et al., 1994). Just as important, there are significant distinctions in the properties of the different β-subunit isoforms with respect to their tissue distribution (Castellano et al., 1993a,b; Ludwig et al., 1997; Pichler et al., 1997), subcellular localization (Colecraft et al., 2002), and impact on channel inactivation kinetics (Olcese et al., 1994; De Waard and Campbell, 1995; Wei et al., 2000; Colecraft et al., 2002). Such functional diversity of Ca²⁺ channel β subunits is a likely contributor to the broad repertoire of physiological responses supported by HVA Ca²⁺ channels.

Alternative splicing generates an even greater molecular diversity of β subunits, potentially extending the physiological dimensions of HVA Ca²⁺ channel functions (Collin et al., 1993; Cahill et al., 2000; Helton and Horne, 2002). Presently, however, knowledge of the full complement of β-subunit splice variants that are expressed, and the biophysical and physiological traits that they impart, remains relatively obscure. Two key reasons for this impasse can be identified. First, discovery of β-subunit variants has been driven primarily by traditional cloning methods that usually identify only one or a few members of a splice variant population at a time (Hullin et al., 1992; Perez-Reyes et al., 1992; Castellano et al., 1993a,b; Helton and Horne, 2002). Second, even when splice variants of a given β subunit are known, the biophysical properties conferred by the distinct β-splice forms to a given channel are rarely rigorously compared.

Research relating to splice variation in the Ca²⁺ channel β₂ subunit epitomizes these deficiencies. β₂ is the dominant β subunit expressed in heart (Perez-Reyes et al., 1992; Pichler et al., 1997; Haase et al., 2000) and retina (Ball et al., 2002), but is also an important component of HVA Ca²⁺ channels expressed in brain (Ludwig et al., 1997; Pichler et al., 1997), smooth muscle (Reimer et al., 2000), and pancreas (Iwashima et al., 2001). Five distinct β₂ subunits, varying only in the proximal amino terminus, have been cloned from different species including rat (Perez-Reyes et al., 1992), rabbit (Hullin et al., 1992), human (Rosenfeld et al., 1993), and mouse (Massa et al., 1995). To date, however, no more than three of the β₂ forms have been identified in any one species (Qin et al., 1998), rendering it ambiguous whether they represent bona fide splice variants of the β₂ gene. Also, aside from the first β₂ subunit cloned, β_2a, the biophysical properties of the different β₂ variants have not been rigorously explored. This constitutes a critical impediment to understanding the physiological implications of having distinct β₂ variants expressed in different tissues.

The recently published draft of the human genome sequence provides new opportunities to identify and clone novel splice variants of a given gene. Here, guided by genomic sequence analysis, we cloned all five β₂ splice variants from human heart and brain cDNA libraries. The biophysical properties of the different β₂ variants were explicitly compared in recombinant L-type (Ca_V 1.2; Ertel et al., 2000) Ca²⁺ channels (α_1C, β₂, α₂δ) reconstituted in HEK 293 cells. We found fundamental qualitative and quantitative distinctions among the subtly divergent β₂-splice variants with respect to their subcellular localization and influence on L-type channel gating, providing new insights into structure-function mechanisms underlying α_1C–β₂-subunit interactions.

METHODS

Molecular cloning

To identify novel human Ca²⁺ channel β₂ splice variants, we searched the HTGS database using TBLASTN algorithm with peptide query sequences derived from human (β_2a and β_2d) or other vertebrate species (β_2b, β_2c, and β_2e). In some cases, either because of short peptide length or sequence differences, the TBLASTN search was not sensitive enough to find orthologous human sequences. In those cases, more sensitive similarity searches were performed on the draft BAC sequences identified in the first round of searches to locate the remaining sequences. The complete open reading frames of the five Ca²⁺ channel β₂ splice variants were amplified from human brain (β_2a, β_2c, β_2d, and β_2e) and heart (β_2b) cDNA libraries (Clontech) by PCR. Upstream primers for the PCR reactions were designed using sequence information from public databases. β_2a, β_2b, and β_2e were amplified with the respective sense primers TCTGCCCCGCTAGCATGCAGTGCTGCGGGCTGGT, GGAATTGGGCTAGCATGCTTGACAGACGCCTTAT, GCGCTGTCGCTAGCATGAAGGCCACCTGGAT, and the antisense primer ACAAAAGGGCAGAATTCATTGGGGGATGTAAACA. Reactions contained 1 μl cDNA, 400 nM each primer, 200 μM each dNTP, 1.8 mM Mg²⁺ and 1 unit eLONGase enzyme mix with its recommended buffer system in a total volume of 25 μl. Amplification was with a touchdown protocol with tubes held at 94°C for 30 s, a variable annealing temperature for 30 s, and extension at 68°C for 120 s. The initial cycle annealing temperature was 64°C; in each of the subsequent five cycles the annealing temperature was 0.5°C lower. In all subsequent cycles, the annealing temperature was 61°C. β_2a and β_2b PCR products were TA cloned into pCR2.1, and then cloned by NheI and EcoRI sites into the bicistronic expression vector pAd IR (Alseikhan et al., 2002), which contains a CMV promoter followed sequentially by a multiple cloning site, an internal ribosomal entry site sequence, and GFP cDNA. In this configuration, upon transfection into mammalian cells, β₂-subunits and GFP are translated as independent protein entities from the same mRNA transcript. This permits GFP fluorescence to be directly used to verify cell transfection.

For β_2c and β_2d, a first round of amplification was performed with the following sense primers AGGGGAGTGGACTGGACCTG and AAAGGGACATGTCCAAGTCGCCTCCCACAG, and antisense primer GTTTTGGGGATGCTGTTAGT. Touchdown during the first 20 cycles was followed by 12 cycles with annealing temperature of 54°C. These amplicons were then gel-purified and reamplified with the respective sense primers CACATGGGCTAGCATGAATCAGGGGAGTGGACTGGACCTG and CCGACTTTGCTAGCATGGTCCAAAGGGACATGTCCAAGT, and antisense primer ACAAAAGGGCAGAATTCATTGGGGGATGTAAACA. Due to the relatively high rate of misincorporation with PCR amplification, most clones contained errors. Therefore, once one error-free clone of the β_2b-splice variant (in pAd IR) had been identified, it served as the backbone for substitution of error-free versions of the four other D1 sequences via NheI and HindIII restriction sites. For GFP fusion constructs of β₂ subunits, their complete coding sequences up to the penultimate codon were PCR amplified with high fidelity Pfu polymerase using as template the β₂ subunits in pAd IR described above. Sense primers (as above) were used in conjunction with the antisense primer AGGATCCTTGGGGGATGTAAACATCCCTG. The PCR products were sequenced and cloned into the expression vector pAdCMV EGFP-N3 (Colecraft et al., 2002) using NheI and BamHI sites, yielding pAd CMV β_2a–e-GFP. The cloned β₂ variant sequences have been deposited in GenBank with the following accession numbers: β_2a AF423189, β_2b AF285239, β_2c AF423190, β_2d AF423191, and β_2e AF423192.

Transfection of HEK 293 cells

Low-passage number HEK 293 cells were maintained as previously described (Brody et al., 1997). The calcium-phosphate precipitation method was used to transfect 10 μg each of cDNA encoding wild-type rabbit α_1C (Wei et al., 1991), rat α₂δ (Tomlinson et al., 1993), and the appropriate β₂ splice variant (subunits).

Confocal imaging

β₂-GFP fusion constructs were transiently transfected into HEK 293 cells, and confocal images were acquired 48 h posttransfection using a Zeiss laser-scanning confocal microscope. Exemplar intensity images were low-pass filtered and analyzed offline using custom-written software in MATLAB 6.0, Release 12 (Mathworks, Natick, MA).

Electrophysiology

Whole-cell recordings were obtained at room temperature 2–3 days after transfection using an Axopatch 200A patch-clamp amplifier (Axon Instruments, Union City, CA). Cells were continuously perfused with external solution containing (in mM): 140 TEA-MeSO₃, 10 HEPES, and 5 BaCl₂ (pH 7.3, adjusted with TEA-OH). The internal solution contained (in mM): 135 Cs-MeSO₃; 5 CsCl₂; 0.5 EGTA; 1 MgCl₂; 4 MgATP; and 10 HEPES (pH 7.3, adjusted with CsOH). Solution osmolarities were adjusted to 290–300 mOsm with glucose. Series resistances were typically 1–2 MΩ after >70% compensation.

Currents were sampled at 10 kHz and filtered at 2 kHz, except for tail-activation protocols, in which currents were sampled at 25 kHz and filtered at 10 kHz. Voltage protocols were delivered at 60-s intervals, and leak and capacitive transients were subtracted by P/8 protocol.

Data and statistical analysis

Data were analyzed using custom-written software in MATLAB and Microsoft Excel, and are displayed as mean values ± SE. Recovery from inactivation, steady-state inactivation, and steady-state activation were well-described by the following equations, with parameters (Table 1) determined using least-squares criteria.

TABLE 1.

Comparison of Boltzmann parameters among β₂ splice variants

	h(20 s) V1/2 (mV)	Recovery Ffast	Activation Flow
β2a	−41	0.51	0.70
β2b	−45	0.73	0.36
β2c	−41	0.66	0.51
β2d	−41	0.72	0.61
β2e	−38	0.52	0.99

Values were determined using least-squares criteria. For steady-state inactivation, k = 8.6 mV. For recovery from inactivation, τ₁ = 91.6 ms and τ₂ = 20 s. For activation, V_1/2,low = −2 mV, V_1/2,high = 38 mV, k_low = 10 mV, and k_high = 20 mV.

Recovery from inactivation data were fitted by a biexponential function of the form: RF = 1 − [F_fast × exp(−t/τ₁) + (1 − F_fast) × exp(−t/τ₂)], where F_fast is the fraction of the fast-recovery component, t is the interpulse duration, and τ₁ and τ₂ are the fast- and slow-recovery time constants, respectively.

Steady-state inactivation data were fitted by a Boltzmann function of the form: h(20 s) = I₂/I₁ = 1/(1 + exp((V − V_1/2)/k)), where I₂ is the test-pulse current, I₁ is the normalization current, V is the membrane potential of the conditioning pulse, V_1/2 is the potential for half-inactivation, and k is the slope factor.

Steady-state activation data were fitted by a dual-Boltzmann function of the form: G/G_max = F_low/(1 + exp(V_1/2,low − V)/k_low) + (1 − F_low)/(1 + exp(V_1/2,high − V)/k_high)), where G is the tail current, G_max is the tail current evoked by a depolarization to +110 mV, F_low is the fraction of the low-threshold component, V is the membrane potential of the test pulse, V_1/2,low and V_1/2,high, and k_low and k_high are the half-activation potentials and slope factors for the low- and high-threshold components, respectively.

Kinetic modeling

Stochastic simulations were performed in MATLAB using the built-in matrix exponential function (expm) to solve an 11-state Markov chain for transition probabilities (Colquhoun and Hawkes, 1981). Repeated matrix exponential calls were made with a step size of 400 μs, and ionic currents were calculated assuming a linear conductance for Ba²⁺ with a reversal potential of +132 mV. Voltage-dependent transitions were modeled with rate constants of the form: k_ij(V) = k_ij × exp(z_ij × V × F/RT), where k_ij is a rate constant (in ms⁻¹) at zero membrane potential associated with the i-to-j transition, z_ij is the valence of the gating charge (set to one elementary charge for simplicity), V is the membrane potential (in mV), and RT/F is the gas constant and absolute temperature divided by Faraday's constant (equal to +25 mV). The final transition between closed and open states within each mode was modeled as voltage-independent (Zagotta and Aldrich, 1990; Colecraft et al., 2002) and set to the approximately ms timescale of rapid openings and closings during L-type Ca²⁺ channel bursts (Colecraft et al., 2002). The maximum P_o was set to 0.6. Parameters were optimized by eye. Fig. 8 C, Scheme 2, and Table 2 provide kinetic structure and detailed rate constants used in the simulations.

FIGURE 8.

Quantitative descriptions of L-type channel inactivation gating profiles as conferred by distinct β₂ splice variants. (A–C) Alternative kinetic schemes that could account for the existence of two kinetically distinct inactivation states, which are suggested experimentally by the biexponential kinetics of channel recovery from inactivation. In Fig. 8 A, Scheme 1, inactive states were separated into those connected to the activation pathway with either fast or slow entry/exit rates (I_f and I_s, respectively). In Fig. 8 C, Scheme 2, inactivation proceeds from the open state by only one pathway. In this case, the biexponential kinetics of recovery from inactivation arise from the existence of two channel populations (or modes) displaying kinetically distinct fast and slow modes of inactivation. (D–F) Simulations of inactivation gating of β_2a and β_2b channels using Fig. 8 C, Scheme 2 showed marked agreement between the model (black curves) and experimental data (gray curves, and ▪, ○). Values of rate constants used in simulations are presented in Table 2. In simulations, the fraction of fast channels was given by the corresponding F_fast values, as determined from recovery from inactivation protocols (Table 2). Note that for simulations of β_2a and β_2b channels, rate constants were identical in the two cases. Hence, all the differences between the channel types were completely accounted for by the differing values of the fraction of fast inactivating channels, F_fast. (G) In the case of steady-state inactivation, simulations using the dual-population model as described above (dotted lines) diverged from the experimental data (▪, ○). Remarkably, when it was assumed that all the channels were gating in the slow mode, model predictions of steady-state inactivation (black curves) superimposed the experimental data. This suggested that during the ultra-long 20-s conditioning pulses used in steady-state inactivation protocols, channels consolidate into the slow inactivating mode. Evidence of such consolidation has been experimentally observed.

TABLE 2.

Parameters for quantitative kinetic model of β₂ splice variants

	Activation constants				Allosteric factors		Inactivation constants
	α(V)	β(V)	γ	δ	f	h	ki	kr
Fast (ms−1)	1.0	0.8	1.5	1.0	0.25	0.25	3.E-03	1.E-03
Slow (ms−1)	1.0	0.8	1.5	1.0	0.25	0.25	4.E-04	4.E-06

Rate constants correspond to transitions diagrammed in Fig. 8 A, Scheme 1. α(V) and β(V) are voltage-dependent rate constants using Hodgkin-Huxley formalism. In simulations, relative contributions of fast to slow channels were given by F_fast values determined from the recovery from inactivation protocol.

RESULTS

Genomic perspective permits in silico cloning of novel human β₂-splice variants

Primary sequence homology between β subunits suggests a modular domain structure (domains D1–D5; Fig. 1 A) in which three variable regions (D1, D3, and D5) are interspersed by two conserved domains (D2 and D4; see De Waard et al., 1994; Birnbaumer et al., 1998). A search of public databases revealed the existence of five β₂-subunit forms, varying only in the D1 domain, distributed among different species (Colecraft et al., 2002). Presently, no more than three of the β₂ forms have been cloned from any one species, making it uncertain whether they constitute genuine splice variants, or simply reflect species variation. The availability of human genomic sequence in the form of draft BAC sequences originally provided a unique opportunity to resolve this core ambiguity. A search of the HTGS database using TBLASTN (as described in Methods) yielded positive outcomes that explicitly confirmed the presence of sequences for all five D1 domains within the human β₂ gene. Therefore, we have designated them β_2a–β_2e, with the alphabetical order corresponding to their chronological identification in the different species and not their order in the genome (Fig. 1 B; see also Colecraft et al., 2002). This nomenclature supplants previous assignments that were primarily species-based and suffered from lack of a clear recognition that there exist at least five distinct D1 domain splice variants of β₂.

FIGURE 1.

Genomic perspective clarifies basis of Ca²⁺ channel β₂-subunit molecular diversity. (A) Modular domain structure proposed for Ca²⁺-channel β subunits based on primary sequence identity. Two conserved domains (D2 and D4, gray boxes) are interspersed by three variable domains (D1, D3, and D5). D1 is an important determinant of channel inactivation, and D4 contains the β-interaction domain, important for interactions with pore-forming α₁-subunits. Alternative splicing of β₂ is known to occur in at least two regions, D1 and D3 (black boxes). (B) Schematic structure of human CACNB2 based on analysis of human genomic DNA sequence. Exons are represented by vertical lines and introns by horizontal lines; introns are approximately to scale, whereas exons are not to scale. (C) Schematic of mutually exclusive splicing patterns of exons 1–6 that result in the generation of five alternative β₂-subunit D1 domains. The corresponding amino acid sequences are represented. Exons 1, 2, and 4–6 encode the amino-termini of the distinct β₂ splice variants; each has their own initiation codon and 5′ untranslated region (white box). Exon 3 is common to β_2c and β_2d. Exon 7 (gray box, with underlined sequence) is a constitutive exon that encodes the beginning of the D2 domain common to all β₂ variants.

The existence of a contiguous genomic DNA sequence, in which all five D1 variants could be identified, fully illuminated the relative orientation of the corresponding exons in the β₂ gene (Fig. 1, B and C). As shown in Fig. 1 C, the five D1 variants result from the alternative splicing of six exons. Exon 7, a constitutive exon that encodes the beginning of the D2 domain common to all β₂ variants, is the acceptor exon in four mutually exclusive splicing pathways. β_2a arises from splicing between donor exon 5 and acceptor exon 7. For variants β_2b and β_2e, exons 4 and 6 serve as donor exons, respectively. β_2c and β_2d share a common 31-amino-acid sequence just upstream of domain D2; for these variants, splicing occurs between exons 3 and 7. Exon 3 lacks an initiation codon. Splicing occurs between either of donor exons 1 or 2 and acceptor exon 3 to give rise to the β_2d and β_2c transcripts, respectively. Although not addressed in this study, β₂ transcripts also exhibit variability in domain D3, arising from mutually exclusive splicing of exons 11 (134 nt), 12 (20 nt), and 13 (62 nt) (Colecraft et al., 2002). β_2a–β_2e transcripts described herein all contain exon 11 and lack exons 12 and 13.

Despite the predicted existence of at least five human β₂ D1 splice variants, only β_2a and β_2d have thus far been cloned from human tissue (Rosenfeld et al., 1993; Yamaguchi et al., 2000). Armed with sequence information from the human genome, we designed primers that permitted the successful amplification of the open reading frames of all five variants from human heart (β_2b) and brain (β_2a, β_2c–β_2e) cDNA libraries. Having thus established the presence of β_2a–β_2e transcripts in human tissues, we next sought to determine the functional consequences of such β₂-subunit diversity.

Distinctive subcellular targeting of β₂-splice variants

The bulk of published work on Ca²⁺ channel β₂-subunits has centered on β_2a, the earliest cloned splice variant. A key distinguishing trait of β_2a is that, when expressed alone in heterologous systems, it targets to the plasma membrane—an observation at odds with predictions that it would localize to the cytosol (Chien et al., 1995). This anomalous behavior is due to palmitoylation of two cysteines residing in the D1 domain (Chien et al., 1996; see also Fig. 1 C) and is the underlying basis for several unique functional hallmarks of β_2a. None of the other β₂-splice variants (β_2b–β_2e) retain consensus sequences for palmitoylation (Milligan et al., 1995; Resh, 1999), raising expectations that they would differ from β_2a with respect to membrane targeting and any attendant functional sequelae. Nevertheless, since it is difficult to predict with certainty the subcellular localization of a protein, and because this feature is a vital prognosticator of β-subunit function (Qin et al., 1998), we first sought to compare the subcellular targeting of heterologously expressed β₂ variants.

To directly visualize β₂-subunit localization in living cells, we expressed GFP-fusion constructs of β_2a–β_2e in HEK 293 cells and viewed cellular GFP fluorescence by confocal microscopy (Fig. 2). Cells expressing GFP alone displayed bright green fluorescence throughout the cell, with equal intensity in the nuclear and cytoplasmic compartments (not shown). Reassuringly, β_2a-GFP localized to the periphery of the cell, a clear indication that fusion of GFP to the carboxyl terminus did not disrupt the well-known membrane targeting of β_2a. In accord with their lack of consensus sites for palmitoylation and relatively high molecular weights (>90 kDa), β_2b–β_2d were localized to the cytosol and excluded from the nucleus (molecular weight cutoff for passive diffusion into the nucleus is 40 kDa; Davis, 1995). By contrast, β_2e-GFP displayed plasma membrane targeting, akin to that observed with β_2a (Fig. 2). This novel observation was completely unexpected given the hydrophilic character of the β_2e D1 domain and serves to underscore uncertainties associated with a priori prediction of β-subunit subcellular location.

FIGURE 2.

Recombinant human β₂ splice variants target differentially in HEK 293 cells. (A) Confocal, and (B) line-scan images of β₂-GFP fusion proteins. β_2a and β_2e splice variants were localized to the plasma membrane; membrane targeting of β_2a is caused by palmitoylation of two cysteines in D1. In contrast, β_2b–β_2d variants were distributed throughout the cytosol but excluded from the nucleus. F is the normalized fluorescence intensity through the midpoint of images in A.

These results firmly established that β₂-splice variants divided into two camps with respect to subcellular localization—β_2a and β_2e target to the plasma membrane, while β_2b–β_2d are present in the cytosol. Co-expression of GFP-tagged β_2b–β_2d proteins with an untagged Ca²⁺ channel α_1C-subunit did not perceptibly change their subcellular localization over that observed with the β₂-GFP subunits alone (not shown). This result might be expected given the likelihood that the β₂-subunits are overexpressed with respect to the α₁-subunit. To more directly demonstrate an interaction between our cloned β₂ variants and α₁-subunits, we turned to functional electrophysiological studies. The strict dichotomy in targeting we observed presaged functional distinctions among the β₂-splice variants as presented below.

Divergent modulatory effects of β₂-splice variants on L-type channel inactivation gating

Tuning of Ca²⁺ channel inactivation kinetics is a well-recognized and important function of auxiliary β subunits. Therefore, we investigated this dimension of β₂-splice variant function in recombinant L-type Ca²⁺ channels (α_1C, β₂, α₂δ) reconstituted in HEK 293 cells. Reconstitution of robust whole-cell Ba²⁺ currents (Figs. 3–6) amply demonstrated the viability of the cloned β₂ subunits. In the absence of co-expressed β subunits, reconstituted L-type channel currents are exceedingly small and only infrequently observed. Distinctive effects of β₂ splice variants on key L-type channel inactivation parameters were measured as described below.

FIGURE 3.

β₂ splice variants confer distinctive inactivation kinetics to recombinant L-type channels. (A) Normalized and averaged Ba²⁺ currents evoked by 1-s depolarizations to 0 mV. All currents had biexponential inactivation kinetics; those reconstituted with β_2a and β_2e had markedly slow inactivation kinetics compared to β_2b–β_2d. (B) Voltage-dependence of fast inactivation. The fraction of current remaining 300 ms into the test pulse, r₃₀₀, was computed as an index of inactivation. Higher r₃₀₀ values reflect a slower rate of inactivation. Inactivation conferred by β_2a and β_2e was markedly less voltage-dependent compared to β_2b–β_2d. In both A and B, β_2a data are reproduced as gray lines in β_2b–β_2e panels to permit direct visual comparison.

FIGURE 6.

Different β₂ splice variants impart distinctive steady-state activation behavior to L-type channels. (A) Exemplar current traces from β_2a and β_2b channels evoked by a traditional tail-activation protocol in which channels were steady-state activated by a variable-voltage family of 20-ms test pulses, and tail currents measured after repolarization to −50 mV. Traces shown were obtained with test-pulse voltages of −20, 0, +40, and +100 mV. (B) Normalized (G/G_max) G-V curves for channels incorporating the different β₂ variants. Smooth black curves through the data were generated from dual-Boltzmann functions as described in Methods, with parameter values given in Table 1. To generate fits, V_1/2 and k values were constrained to the values given in the legend to Table 1, and the amplitudes of the low- and high-threshold components determined by least-squares criteria. The reproduced fits from β_2a data (gray curves) displayed varying degrees of difference from the data for other β₂ variants, visually affirming that distinct β₂ variants gave rise to differing fractions of channels that operated in the low threshold of activation regime.

Kinetics and voltage-dependence of current inactivation

With each β₂ variant, 1-s depolarizing steps to 0 mV produced L-type channel currents that activated quickly and then decayed with biexponential kinetics (Fig. 3 A). However, there were quantitative distinctions among the β₂ variants with regard to the degree of inactivation conferred to the channel. This effect was immediately apparent by comparing the normalized mean current waveform for β_2a channels with those of channels reconstituted with β_2b–β_2e subunits, respectively (Fig. 3 A; gray traces are reproductions of β_2a current waveform). β_2b–β_2d channels exhibited clearly faster inactivation kinetics compared to β_2a channels as reported visually by the discordance between corresponding current waveforms. By contrast, β_2e channels inactivated with virtually identical slow kinetics as β_2a. Hence, membrane-targeted β₂ variants (β_2a and β_2e) conferred slower inactivation kinetics compared to variants localized in the cytosol (β_2b–β_2d). This simple correlation is in accord with the notion that the characteristic slowed inactivation kinetics imparted by β_2a to HVA Ca²⁺ channels is due to its anchoring to the plasma membrane (Restituito et al., 2000).

We examined the voltage-dependence of channel inactivation by utilizing a family of 1-s voltage steps from −20 to +30 mV. To compare β₂-splice variant signatures on the voltage-dependence of inactivation in a straightforward manner, we used the fraction of current remaining 300 ms after depolarization (r₃₀₀) as an index of channel inactivation (Fig. 3 B). For β_2a channels, r₃₀₀ values barely changed in the voltage range from −20 mV (r₃₀₀ = 0.70 ± 0.03) to +20 mV (r₃₀₀ = 0.64 ± 0.03). Continuing the trend of functional similarities between these two β₂ variants, r₃₀₀ values for β_2e channels were mostly superimposed on values obtained for β_2a channels (Fig. 3 B) over this voltage range. In sharp contrast, r₃₀₀ values for β_2b–β_2d channels dropped appreciably with increasing voltage, indicating a stronger voltage-dependence of inactivation for these channels compared to those reconstituted with β_2a and β_2e. Moreover, r₃₀₀ values for β_2b–β_2d channels were significantly depressed compared to values obtained for β_2a and β_2e channels over a wide range of voltages, reflecting the greater rate of macroscopic current decay observed with cytoplasmically localized β₂ subunits (Fig. 3 B).

Recovery from inactivation

Next, we investigated the effects of β₂-splice variants on the kinetics of L-type-channel recovery from inactivation. After a 1-s conditioning pulse to 0 mV, test-pulse currents recovered with biexponential kinetics for all the β₂ forms (Fig. 4). In fact, across β₂ variants, the two components of recovery from inactivation were well described as the sum of two exponential functions with identical time constants (τ_fast = 91.6 ms, and τ_slow = 20 s), but differing amplitudes (Fig. 4). Here again, there was a clear demarcation in the functional properties of membrane-bound versus cytosolic β₂ variants—β_2a and β_2e displayed a similarly lower fraction of channels recovering from inactivation with fast kinetics, compared to β_2b–β_2d channels. This is visually apparent from the virtual congruence of recovery curves between β_2a and β_2e channels, and their divergence from curves obtained with β_2b–β_2d channels (Fig. 4). Mean values of the fractional recovery from inactivation (RF) measured at interpulse durations of 100 ms and 4.1 s were significantly different between β_2a/e and β_2b–d channels (RF₁₀₀ values were: β_2a, 0.35 ± 0.03, n = 8; β_2b, 0.53 ± 0.02, n = 7; β_2c, 0.46 ± 0.05, n = 6; β_2d, 0.47 ± 0.03, n = 5; β_2e, 0.37 ± 0.05, n = 7; P < 0.01, one-way ANOVA). Our results are in qualitative agreement with previous work showing that L-type channels recover from inactivation with biexponential kinetics (Zuhlke and Reuter, 1998; Jeziorski et al., 2000). Moreover, distinct β-subunit isoforms were found to differentially modulate L-type channel recovery from inactivation (Jeziorski et al., 2000). We now show that more subtle sequence variations conferred by alternative splicing within a single β-subunit isoform are sufficient to confer differential modulation of L-type channel recovery from inactivation. The biexponential nature of the kinetics of recovery from inactivation strongly indicates the presence of two kinetically distinct forms of inactivation (Kass and Sanguinetti, 1984; Hering et al., 2000). We defer consideration of the possible physical processes underlying these forms of inactivation, and the mechanism of their differential modulation by β₂ splice variants, to the Discussion.

FIGURE 4.

L-type channel recovery from inactivation occurs with distinct fast and slow components, but with differing amplitudes, across β₂ splice variants. (A) Exemplar current traces from β_2a and β_2b channels evoked by a traditional recovery from inactivation protocol in which a 1-s prepulse to 0 mV is followed by a 50-ms test pulse to 0 mV at variable interpulse durations. Here and throughout, β_2a and β_2b exemplars are shown as representative slow- and fast-inactivating channels, respectively. (B) Plots of the fractional recovery from inactivation, RF. Data points were calculated as RF = (I_test − I_end)/(I_prepulse − I_end). Peak prepulse currents decreased by ∼20% for all β₂ variants by the end of the recovery protocol. This rundown was systematically accounted for in the analysis by normalizing all currents to the peak prepulse current from the first sweep. Smooth black curves through the data are biexponential fits generated from the equation presented in Methods. Curves were constrained to have the same fast- and slow- time constants, while the amplitudes of fast and slow components were determined by least-squares criteria (parameter values are presented in Table 1). Biexponential fits to β_2a data are reproduced in the other panels (gray curves) to facilitate visual comparison of the kinetics of recovery from inactivation among the β₂ splice variants. β_2a and β_2e had a smaller fraction of channels recovering from inactivation with fast kinetics compared to β_2b–β_2d channels.

Steady-state inactivation

Finally, to complete our examination of modulation of L-type channel inactivation gating by β₂-splice variants, we probed quasi-steady-state inactivation (hereafter referred to as steady-state inactivation) using a three-pulse protocol in which a 10-ms prepulse to 0 mV was followed in turn by a family of 20-s conditioning pulses (−80 to +20 mV), and a 100-ms test pulse to 0 mV (Fig. 5). By contrast with the trends established above regarding clear-cut functional distinctions between membrane-bound and cytosolic β₂ variants, there were only minor distinctions in steady-state inactivation curves (Fig. 5; Table 1). All the curves were well described by single-Boltzmann functions with identical values for k, and nearly identical values for V_1/2 (Table 1). This finding is in agreement with previous studies indicating a relative resistance of L-type channel steady-state inactivation curves to manipulation by distinct Ca²⁺ channel β subunits (Jones et al., 1998). This is opposite the situation in neuronal Ca²⁺ channels (N-, P/Q- and R-type), where β-subunit identity is a major determinant of steady-state inactivation profiles (Jones et al., 1998; Stea et al., 1994).

FIGURE 5.

L-type channel steady-state inactivation properties are relatively insensitive to the identity of the component β₂-subunit splice variant. (A) Exemplar current traces from β_2a and β_2b channels evoked by a traditional steady-state inactivation protocol in which a 10-ms normalizing pulse to 0 mV is followed sequentially by a variable-voltage family of 20-s conditioning pulses and a 100-ms test pulse to 0 mV. Only the first 200 ms of the currents during the conditioning depolarizations is shown. Traces shown were obtained with conditioning pulse voltages of 0, −30, −50, and −80 mV. (B) Plots of steady-state inactivation, h(20 s), as a function of voltage of the conditioning pulse. The normalized data points were obtained from the peak values of the prepulse and test currents, repectively, as h(20 s) = I_test/I_prepulse. Smooth black curve through the β_2a data was generated from a single Boltzmann function as described in Methods, with parameter values given in Table 1. Reproduction of the β_2a Boltzmann fit onto the panels for other β₂ splice variants (gray curves) demonstrated marked agreement with these other data, indicating little difference among the distinct β₂ variants.

Overall, our in-depth study of modulation of L-type channel inactivation gating by β₂ splice variants has revealed fundamental functional differences between membrane-bound (β_2a and β_2e) and cytosolic (β_2b–β_2d) β₂ subunits. Further, we find that the functional distinctions imposed by membrane-bound β₂ subunits on L-type channels are more pervasive than previously realized.

Distinctive effects of β₂-splice variants on L-type channel steady-state activation and prepulse facilitation

One prominent effect of Ca²⁺ channel β subunits is that they shift the voltage-dependence of Ca²⁺ channel activation gating in the hyperpolarizing direction. By contrast with their markedly divergent influences on channel inactivation kinetics, β subunits are traditionally viewed as having more homogeneous effects on steady-state activation profiles. Moreover, while the D1 domain of β subunits is appreciated to be an important determinant of Ca²⁺-channel inactivation, it is presumed to play only a minor role in steady-state activation. Recently, however, alternative D1 splice variants of the β₄ subunit were found to differentially modulate the voltage-dependence of activation of N- and P/Q- type channels, thereby challenging this simple interpretation (Helton and Horne, 2002). Therefore, we investigated possible distinctions in β₂ splice variant modulation of L-type channel steady-state activation utilizing traditional tail-activation (G-V) protocols (Fig. 6).

To permit first-order comparison of G-V curve waveforms among the β₂ variants, we generated dual-Boltzmann fits by constraining V_1/2 and k values for both high- and low-threshold components, and determined the amplitudes of the two components by least-squares criteria (values in Table 1 and its legend). The impressive fits obtained for each β₂ variant authenticated this strategy (Fig. 6 B). Explicit overlay of the fits to data obtained from β_2a channels (gray traces in Fig. 6 B) onto the graphs for other splice variants provided a qualitative indication of distinctions in L-type channel steady-state activation among the different β₂ variants. Most notably, β_2b and β_2e channels displayed G-V curves with opposite shifts in the relative amplitude of low- versus high-threshold components compared to β_2a channels (Fig. 6 B). Specifically, whereas β_2b decreased the fraction of channels activating at a lower voltage threshold (F_low = 0.36 for β_2b versus 0.70 for β_2a), β_2e increased this parameter (F_low = 0.99). By contrast, β_2c and β_2d channels displayed G-V curves that were relatively unchanged, or only subtly different from β_2a (Fig. 6 B). These results revealed an unappreciated role of the D1 domain of Ca²⁺-channel β₂ subunits as an important determinant of L-type channel steady-state activation gating in addition to their more recognized function in modulating inactivation. The distinctions in activation gating did not extend to the kinetics of channel activation, as single exponential fits to the activation phase of currents evoked by 0-mV test pulses produced time constants that were not significantly different among the β₂ variants (not shown).

A potential explanation (suggested by reviewers) for the increasing activation observed at positive voltages, evident as the high-threshold component of G-V curves (Fig. 6 B), could be voltage-dependent facilitation of L-type channels. If so, then the distinct β₂ variants might be expected to differentially modulate this physiologically important property of L-type Ca²⁺ channels. This proposal was directly tested using a protocol in which test-pulse currents at a family of voltages were recorded with or without a 50-ms prepulse to +100 mV (Fig. 7 A). An interpulse duration of 50 ms at −100 mV was chosen, being long enough to ensure complete deactivation of channels after the prepulse, but short enough that any facilitation that was evoked by the prepulse would not have completely decayed. Exemplar currents evoked from β_2a and β_2b channels both displayed prepulse facilitation, although to different extents (Fig. 7). The larger degree of L-type channel facilitation conferred by β_2b resulted in a more dramatic prepulse-induced hyperpolarizing shift in the G-V curve for β_2b compared to β_2a (Fig. 7 B), fitting with the suggestion that differences in voltage-dependent facilitation may play a role in establishing the observed distinctions in waveforms of steady-state activation curves (Fig. 6). Explicit comparison of L-type channel facilitation across all five β₂-splice variants revealed that, compared to β_2a, only β_2b supported a significantly higher degree of facilitation (23.9 ± 4.8%, n = 8 for β_2b; 10.4 ± 2.5%, n = 9 for β_2a; P < 0.05, two-tailed Student's unpaired t-test; see also Fig. 7 C). Although β_2d supported a high mean value of facilitation (28.4 ± 9.7%, n = 8), primarily due to a single cell that exhibited an unusually large degree of facilitation (86%), this value fell short of significance when compared to β_2a (P = 0.08, two-tailed Student's unpaired t-test).

FIGURE 7.

β₂ splice variants differentially affect prepulse facilitation of recombinant L-type channels. (A) Alternating prepulse voltage protocol (above) and exemplar Ba²⁺ test-pulse currents (below) recorded from L-type channels reconstituted with β_2a (left) or β_2b (right). Currents were evoked by a 20-ms test-pulse to 0 mV in the absence of prepulse (black traces) or after a 50-ms prepulse to +100 mV (gray traces); the interpulse duration was 50 ms. Tail currents were analyzed at −50 mV repolarization potential. (B) Normalized and averaged G-V curves generated from alternating-prepulse protocol for L-type channels reconstituted with β_2a (left, ▪, □) and β_2b (right, •, ○). Tail-current amplitudes were normalized to the current evoked at +110 mV in the absence of prepulse and plotted as a function of test-pulse potential, which was varied from −40 to +120 mV in 10-mV increments. Solid symbols (▪, •) denote currents evoked in the absence of prepulse; open symbols (□, ○) denote currents evoked after a prepulse. (C) Comparison of facilitation measured at a test-pulse potential of 0 mV across all five β₂ splice variants. Facilitation was calculated as the tail-current amplitude after a prepulse, normalized to the tail-current amplitude in the absence of prepulse. The difference between β_2a and β_2b was statistically significant (asterisk denotes P < 0.05, Student's t-test).

Overall, these results demonstrate that the D1 domain of the Ca²⁺ channel β₂ subunit is an important determinant of prepulse facilitation of L-type channels, in agreement with previous observations (Cens et al., 1998; Qin et al., 1998). However, we did not observe a strict segregation of the degree of L-type channel facilitation according to whether the supporting β₂ subunit was membrane-bound (β_2a, β_2e) or localized to the cytosol (β_2b–β_2d). Such a result might have been envisioned given a previous report that the lower degree of facilitation conferred by β_2a compared to other β-subunit isoforms was completely accounted for by the posttranslational palmitoylation of β_2a (which confines it to the membrane; see Qin et al., 1998). Hence, our results indicate that the role of the β₂-subunit D1 domain in specifying L-type channel prepulse facilitation may be more complicated than previously suspected.

DISCUSSION

The important findings presented in this study are: 1), human genomic sequence analyses permitted, for the first time, the cloning of five distinct amino-terminal β₂-subunit variants from the same species, unambiguously identifying them as genuine splice variants of the β₂ gene, CACNB2. 2), A second β₂-subunit variant, β_2e, targeted unexpectedly to the plasma membrane when expressed in HEK 293 cells, in a manner akin to that previously described for β_2a. 3), The subcellular localization of the β₂-subunit forms was correlated with the kinetics of L-type current decay—membrane-anchored subunits (β_2a and β_2e) produced slowly inactivating currents, whereas cytoplasmic subunits (β_2b–β_2d) imparted relatively faster inactivation kinetics. 4), For all subunit combinations, L-type channels displayed a biexponential recovery from inactivation. However, membrane-anchored β_2a and β_2e channels exhibited a smaller fraction of channels recovering with fast kinetics compared to their cytosol-localized counterparts. 5), L-type channel steady-state activation curves displayed low- and high-threshold components with distinct β₂ splice variants differentially determining the relative amplitudes of the two components. 6), The distinct β₂ variants supported varying degrees of prepulse facilitation of L-type channels.

We discuss these results in relation to previous studies in terms of their biophysical and physiological implications.

Historical perspective on cloning and nomenclature of Ca²⁺ channel β₂-subunit splice variants

The first Ca²⁺ channel β₂ subunit was cloned from rat brain (Perez-Reyes et al., 1992) and has commonly been referred to as β_2a. Shortly thereafter, two β₂-subunit homologs were cloned from rabbit heart (Hullin et al., 1992). These differed from each other and from the rat brain β_2a subunit only in the amino-terminal D1 domain, all other parts of the proteins being essentially identical. Nevertheless, the rabbit clones were dubbed rabbit β_2a and rabbit β_2b. A fourth β₂ variant with a unique D1 domain was subsequently cloned from human fetal brain (Rosenfeld et al., 1993) and is commonly referred to as β_2c. Finally, a fifth β₂ subunit with a distinctive D1 domain was cloned from mouse brain and was originally denoted as mouse brain β_2a (Massa et al., 1995), although this variant has more recently been referred to as β_2d (Yamaguchi et al., 2000). Before the present study, no more than three of the β₂ forms had been cloned from any one species, making it uncertain whether all five were genuine splice variants, or simply reflected species variation. Recently, based on human genome sequence analysis, we reported that all five described β₂ D1 domains were represented within the human β₂ gene (Colecraft et al., 2002). Here, by cloning all five β₂ forms from human cDNA libraries, we explicitly confirm their collective expression in a single species, unequivocally establishing that they are true splice variants. This insight validates our proposal to rename the five distinct β₂ D1 variants β_2a–β_2e (Colecraft et al., 2002), as we have adopted in this study.

The fragmentary nature of the history of discovery of the distinct β₂ variants can be attributed to limitations in traditional cloning methods, which rarely provide an exhaustive representation of splice variants within a given gene. The availability of the human genome sequence, and other ongoing mammalian genome sequencing efforts, provides a more panoramic perspective from which to study molecular diversity generated by alternative splicing. Genomic sequence information is particularly advantageous in identifying splice variants with different first or last exons. By contrast to splice variations that occur in internal exons, such first- and last-exon splice variants cannot be isolated by the technique of transcript-scanning PCR (Mittman et al., 1999). Overall, our study powerfully demonstrates the advantages of genome-based cloning over more traditional methods in identifying and cloning alternative splice variants of a gene.

Importance of Ca²⁺ channel β-subunit subcellular localization

Hydropathy analysis indicates a lack of membrane-spanning domains in Ca²⁺-channel β subunits (Perez-Reyes et al., 1992; Castellano et al., 1993a), predicting that they would localize to the cytosol when expressed in cells. This prediction is met for the majority of Ca²⁺-channel β subunits so far examined. However, β_2a bucked this trend when it was discovered that it targeted to the plasma membrane when exclusively expressed in HEK 293 cells (Chien et al., 1995). Subsequently, this anomalous localization was attributed to palmitoylation of β_2a, in perfect agreement with the identification of a consensus motif for such modification in the amino terminus of the protein (Milligan et al., 1995; Chien et al., 1996; Resh, 1999). Importantly, such membrane localization of β_2a has been associated with a constellation of distinguishing functional effects including: slowing of Ca²⁺-channel inactivation kinetics (Qin et al., 1998; Restituito et al., 2000); depolarizing shifts in the voltage-dependence of steady-state inactivation in neuronal Ca²⁺ channels (Jones et al., 1998); abolition of voltage-dependent facilitation of L-type channels expressed in Xenopus oocytes (Qin et al., 1998); and modulation of G-protein inhibition of neuronal Ca²⁺ channels (Qin et al., 1997). Hence, the subcellular location of β subunits is an important determinant of function.

Here, we have identified β_2e as a second β₂ splice variant that targets anomalously to the plasma membrane when expressed alone in HEK 293 cells. The mechanism underlying β_2e targeting to the membrane is unclear, although the molecular determinants of this localization must reside in its unique D1 domain. Unlike β_2a, visual inspection of the β_2e D1 domain does not immediately reveal likely sites for posttranslational lipid modification (Milligan et al., 1995; Resh, 1999). Further work will be needed to delineate the mechanism of β_2e targeting to the membrane. Interestingly, β_2a and β_2e were functionally similar, lending support to the hypothesis that physical association with the membrane underlies the characteristic effects of β_2a on Ca²⁺ channels (Restituito et al., 2000). The Ca²⁺-channel β_1b subunit has also been reported to be membrane-localized when expressed in COS-7 cells (Brice et al., 1997) due to an acidic motif in the carboxyl terminus of the protein (Bogdanov et al., 2000). However, this property is not reproduced in other cell types (Chien et al., 1998; Colecraft et al., 2002) and does not result in any distinguishing electrophysiological phenotypes (Bogdanov et al., 2000). Hence, β_2a and β_2e are, thus far, clearly unique in this regard.

Our results contribute to a growing awareness that the subcellular distribution of Ca²⁺-channel subunits represents an important dimension of their function (Brice and Dolphin, 1999; Colecraft et al., 2002). The approach of viewing such distribution by tagging subunits with GFP is useful for such studies and offers the distinct advantages of: 1), permitting viewing of subunit localization in living cells, and 2), avoiding artifacts due to inaccessibility of epitopes or nonspecificity of antibodies. Indeed, we recently used this approach to reveal unexpected targeting of β₄ subunits to the nucleus and transverse elements when expressed in heart cells (Colecraft et al., 2002). Nevertheless, the results of such experiments need to be interpreted with caution because of potential confounding factors arising from fusion of the relatively large GFP molecule to the target protein. For example, fusion of GFP to the amino-terminus of β_2a abolishes membrane targeting, presumably by interfering with the palmitoylation of the cysteine residues in this region (H. M. Colecraft, unpublished observations).

Quantitative mechanism of β₂-subunit modulation of L-type channel inactivation gating

L-type Ca²⁺ channels undergo at least three types of inactivation—Ca²⁺-dependent inactivation, and fast and slow forms of voltage inactivation. Ca²⁺-dependent inactivation is mediated by binding of Ca²⁺ to calmodulin that is tethered to the channel complex (Peterson et al., 1999; Qin et al., 1999; Zuhlke et al., 1999; Erickson et al., 2001; Pitt et al., 2001) and displays a characteristic “U-shaped” response to voltage that is a direct reflection of its current dependence. Our use of Ba²⁺ as charge carrier eliminates Ca²⁺-dependent inactivation; therefore, we will consider this form of inactivation no further. Instead, under our experimental conditions, fast and slow forms of voltage-dependent inactivation are prevalent. Mechanistically, it has been speculated that fast inactivation of HVA Ca²⁺ channels may be analogous to fast N-type inactivation in K⁺ channels (Hoshi et al., 1990; Zagotta et al., 1990) and to the fast hinged-lid inactivation mechanism in Na⁺ channels (West et al., 1992), both of which involve physical occlusion of the pore by cytoplasmic domains of the channel (Hering et al., 2000; Stotz and Zamponi, 2001). A leading candidate for the tethered plug in HVA Ca²⁺ channels is the cytoplasmic loop between domains I and II (I–II loop; Cens et al., 1999b; Stotz et al., 2000; Stotz and Zamponi, 2001), although the possible existence of inactivation gates on other parts of the pore-forming α₁ subunit, or accessory β subunits, has not been excluded. Similarly, slow inactivation of Ca²⁺ channels may be analogous to C-type inactivation of K⁺ channels (Shi and Soldatov, 2002), which involves a constriction of the channel pore (Choi et al., 1991; Kukuljan et al., 1995). Here, the biexponential kinetics of current decay and recovery from inactivation authenticated the prevalence of both fast and slow inactivation across all five β₂-splice variants. Nevertheless, there were key distinctions in L-type channel gating that segregated into two categories, depending on whether the associated β₂ subunit was membrane-anchored, or localized in the cytosol. Specifically, membrane-anchored β₂ subunits (β_2a and β_2e) produced a slower rate of channel inactivation (Fig. 3 A), exhibited elevated r₃₀₀ values over the range of voltages (Fig. 3 B), and displayed a smaller fraction of channels with fast recovery from inactivation kinetics (Fig. 4). By contrast, steady-state inactivation was not very different among distinct β₂ variants (Fig. 5).

To gain mechanistic insights into the variable effects of β₂ variants on L-type channel inactivation gating, we turned to quantitative modeling (Fig. 8). Since inactivation characteristics conferred by the β₂ variants segregated into two major divisions, we modeled β_2a and β_2b channels as representatives of the two functional phenotypes. As a starting point, the biexponential kinetics of recovery from inactivation suggested two kinetically distinct inactivation states. Therefore, we first considered alternative schemes that could give rise to such behavior. In this regard, a three-state model consisting of one open and two inactive states was recently invoked to explain the biexponential kinetics of inactivation and recovery in recombinant P/Q-type Ca²⁺ channels (Sokolov et al., 2000; Berjukow et al., 2001). In those studies, inactivation from the open state could occur by two separate pathways with either a fast or slow rate. By itself, this minimal three-state model cannot describe features such as the voltage-dependence of inactivation or steady-state inactivation. Therefore, we incorporated the three-state inactivation mechanism into a modified Hodgkin-Huxley kinetic model (Fig. 8 A, Scheme 1), similar in structure to those previously used for voltage-gated K⁺ (Zagotta and Aldrich, 1990), Na⁺ (Armstrong and Bezanilla, 1977), and Ca²⁺ channels (Boland and Bean, 1993; Colecraft et al., 2002). Here, the activation pathway consisted of three voltage-dependent transitions (C → C) followed by a distinct voltage-independent step immediately preceding channel opening (C → O). Transitions from the activation pathway to the inactive states were state-dependent but voltage-independent; allosteric inactivation and recovery factors, f and h, respectively, served to hasten inactivation from the open state and enhance recovery to closed states (Klemic et al., 1998; Serrano et al., 1999). Additionally, inactive states were separated into those connected to the activation pathway with either fast or slow entry/exit rates (I_f and I_s respectively). Although simulations using Fig. 8 A, Scheme 1 could qualitatively predict aspects of the inactivation properties conferred by β₂ splice variants, quantitative agreement with experimental data was elusive (simulations not shown). Moreover, it proved difficult to obtain simultaneous good fits to inactivation and recovery kinetics.

Consequently, we considered a second scheme that could describe the biexponential behavior of L-type channel inactivation and recovery from inactivation kinetics. Such behavior could arise from the existence of two channel populations displaying kinetically distinct fast and slow modes of inactivation (Fig. 8 B). To implement this dual-population scenario, fast and slow channels were both simulated independently using a modified Hodgkin-Huxley model (Fig. 8 C, Scheme 2). In contrast to Scheme 1, Fig. 8 A, inactivation could occur from the open state by only one pathway. Biexponential recovery-from-inactivation kinetics emerged from the existence of fast and slow channel populations rather than two exit pathways from the open state. For simplicity, the fast and slow channels employed identical rate parameters except for the microscopic inactivation rate constants k_i and k_r (Table 2). To simulate β₂-subunit-specific effects on inactivation gating by this scheme, we set the fraction of fast channels to be the corresponding F_fast values calculated from recovery from inactivation protocols (Table 1) and computed the total current as the weighted sum of fast and slow channels.

This strategy furnished a strong agreement between model simulations and experimental data for inactivation and recovery kinetics (Fig. 8, D and F), a clear improvement over fits generated with Scheme 1. Less impressive initially was the fit to steady-state inactivation data generated by the model (Fig. 8 G, dotted trace). Remarkably, however, the model produced steady-state inactivation curves congruent with experimental data when it was assumed that all the channels were in the slow-inactivating mode (Fig. 8 G, solid curve). This finding suggested that over the exceptionally long 20-s prepulse duration employed to generate steady-state inactivation curves, initially fast-inactivated channels are consolidated into a slow-inactivated mode. In fact, such consolidation with increasingly prolonged depolarizations has been experimentally observed (Sokolov et al., 2000). For voltage-dependence of inactivation, simulations showed qualitative agreement, although they predicted a steeper voltage-dependence of r₃₀₀ for β_2a than was experimentally determined (Fig. 8 E). This discordance may indicate that β_2a may act to slow the microscopic propensity for inactivation in fast channels. Such a scenario would be consistent with the conceptualization that membrane tethering of β_2a would serve to slow the rate of engagement of an inactivation gate (presumably the α₁-subunit cytoplasmic I–II loop) with its receptor site (Restituito et al., 2000; Stotz and Zamponi, 2001). Such predictions could be tested within the framework of the currently proposed model.

Overall, the differing effects of β₂ variants on L-type channel inactivation gating are well described by a model in which L-type channels distribute reversibly between fast and slow inactivating modes, with membrane-bound β₂ forms displaying an ∼50% lower fraction of fast channels compared to cytosolic β₂ subunits. At the single-channel level, fast and slow channels may be equivalent to the rapidly and slowly inactivating modes observed in N-type channels recorded from rat superior cervical ganglion neurons (Plummer and Hess, 1991). Finally, although this model is preliminary, since rate parameters were determined empirically and we have not systematically ruled out alternative models, it provides a framework for the design of future experiments aimed at investigating mechanisms of channel gating (Serrano et al., 1999).

Possible physiological implications of β₂-subunit splice variation

Beyond the biophysical implications discussed above, our results also raise issues of physiological importance. First, Ca²⁺ currents recorded from many native preparations frequently display a slower inactivating phenotype than expected. It is often tacitly assumed that such behavior may indicate an association of the α₁ subunit with β_2a. Our results indicate that this interpretation is not necessarily correct. Here, we identify β_2e as a second β₂-splice variant that gives rise to comparably slow currents as β_2a. This finding gives reason to wonder whether there are as-yet-unidentified membrane-bound splice variants of other β-subunit isoforms present in native cells. Such subunits could potentially also give rise to slow inactivating currents upon association with native α₁ subunits. Our studies provide a firm foundation to investigate this intriguing notion.

Second, our results indicate that alternatively spliced β₂ subunits would be expected to differentially tune critical functional properties of native voltage-dependent Ca²⁺ channels ranging from modulation of inactivation kinetics to determining the extent of prepulse facilitation. The influence of such β₂-subunit splice variation would be expected to be especially significant in tissues or cells in which β₂ is the sole or predominant β subunit, as is the case in heart (Perez-Reyes et al., 1992; Colecraft et al., 2002) and outer plexiform layer of the retina (Ball et al., 2002). Different β₂-splice variants have been identified in heart cDNA libraries from different species (Hullin et al., 1992; Yamada et al., 2001; Colecraft et al., 2002), raising the possibility that multiple β₂-splice variants exist either in single heart cells, or in cells from different regions of the heart. Our results predict this would lead to functionally heterogeneous L-type channels in heart, with possible implications for cardiac function. Further studies are required to explore such physiological dimensions of β₂-subunit splice variation.

A full appreciation of the likely physiological impact of β₂-subunit splice variation must await studies determining the relative expression levels and tissue distributions of the different β₂ forms. Such studies would benefit from the isolation of the full complement of β₂ splice variants in other mammalian species from which cells and tissue would be more readily available. To this end, we have recently used a genome-based cloning approach to amplify the rat orthologs of β_2a–β_2e subunits, two of which were previously undescribed (F. Chan and H. Colecraft, unpublished data).

CONCLUSIONS

Much work has focused on the use of artificially constructed chimeras to study the structure function of Ca²⁺-channel β subunits (Olcese et al., 1994; Cens et al., 1998, 1999a). Molecularly diverse β subunits generated by alternative splicing constitute a potential treasure trove to gain new insights into the structure-function of α₁–β-subunit interactions and the full scope of β-subunit functions in vivo. Furthermore, studies of such natural chimeras have the added benefit of direct physiological relevance. To these ends, the availability of the human and other mammalian genomic sequences in public databases represents a key enabling asset as demonstrated in this work.