Alternative Splicing of a β₄ Subunit Proline-Rich Motif Regulates Voltage-Dependent Gating and Toxin Block of Ca_v2.1 Ca²⁺ Channels

Abstract

Ca²⁺ channel β subunits modify α₁ subunit gating properties through direct interactions with intracellular linker domains. In a previous report (Helton and Horne, 2002), we showed that alternative splicing of the β₄ subunit had α₁ subunit subtype-specific effects on Ca²⁺ channel activation and fast inactivation. We extend these findings in the present report to include effects on slow inactivation and block by the peptide toxin ω-conotoxin (CTx)-MVIIC. N-terminal deletion and site-directed mutagenesis experiments revealed that the effects of alternative splicing on toxin block and all aspects of gating could be attributed to a proline-rich motif found within N-terminal β_4b amino acids 10–20. Interestingly, this motif is conserved within the third postsynaptic density-95 (PSD-95)/Discs large/zona occludens-1 domain of the distantly related membrane-associated guanylate kinase homolog, PSD-95. Sequence identity of ∼30% made possible the building of β_4a and β_4bthree-dimensional structural models using PSD-95 as the target sequence. The models (1) reveal that alternative splicing of the β₄ N terminus results in dramatic differences in surface charge distribution and (2) localize the proline-rich motif of β_4b to an extended arm structure that flanks what would be the equivalent of a highly modified PSD-95 carboxylate binding loop. Northern blot analysis revealed a markedly different pattern of distribution for β_4a versus β_4bin the human CNS. Whereas β_4a is distributed throughout evolutionarily older regions of the CNS, β_4b is concentrated heavily in the forebrain. These results raise interesting questions about the functional role that alternative splicing of the β₄ subunit has played in the evolution of complex neural networks.

Voltage-gated Ca²⁺ channels participate in an extensive array of cellular activities including excitation–contraction coupling, transcription, and neurotransmitter release. Neuronal Ca_v2 channels are assemblies of up to five subunits, α₁, α₂/δ, β, and γ. The α₁ subunit consists of four homologous repeats (I–IV) of six helices (S1–S6) that arrange to form the selectivity filter and pore. The 24 transmembrane helices are connected by a series of alternating intracellular and extracellular loops. These loops are targets for a host of modifying proteins, including β subunits, G-proteins, calmodulin, and syntaxin, as well as the peptide toxins of venomous spiders and marine snails (Catterall, 2000). Interaction of these proteins with α₁subunits typically alters the voltage dependency and kinetics of channel gating, which in turn modifies Ca²⁺ entry into neurons.

Ultimately, gating behavior is determined by the interactions of individual amino acid side chains with the electrostatic forces within their microenvironments. This is especially true for the positively charged S4 helical segments that constitute the voltage sensors in Na⁺, Ca²⁺, and K⁺ channels. Biophysical studies have shown that depolarization disrupts S4 side-chain interactions of Shaker K⁺ channels to the extent that S4 helices rotate 180° along their axes (Cha et al., 1999; Glauner et al., 1999). This motion likely triggers a cascade of side-chain disruptions that ultimately leads to rotation and separation of the intracellular S6 segments that form the K⁺ channel gate (Bezanilla, 2000). Such a mechanism is supported by recent studies delineating the conformational changes associated with open and closed states of bacterial two-membrane-spanning K⁺ channels (Jiang et al., 2002b) and is generally applicable to Na⁺ and Ca²⁺ channel gating.

Attempts have been made to assign specific gating functions to individual Ca²⁺ channel homology domains. Early chimera studies indicated that the IS6 segment was critical for setting the rate of fast inactivation (Zhang et al., 1994); however, substitution of IIS6 and IIIS6 of the Ca_v2.3 channel into the slow inactivating Ca_v1.2 channel caused a leftward shift in the voltage dependence of inactivation and increased the rate of Ca_v1.2 channel inactivation to near Ca_v2.3 rates (Stotz et al., 2000). Effects on gating have been reported for amino acid substitutions in IS3 (Zhong et al., 2001), the I-II linker (Berrou et al., 2001), IIS6 (Stotz and Zamponi, 2001), extracellular linkers IIIS3-S4 (Lin et al., 1997), and IVS3-S4 (Hans et al., 1999), and IVS6 (Berjukow et al., 2001). Additive effects on Ca_v1.2 channel inactivation were reported recently for individual IS6, IIS6, IIIS6, and IVS6 substitutions (Shi and Soldatov, 2002). Together, these data support a structural model of α₁ subunits in which individual transmembrane segments are interdependently entwined (Horn, 2000).

Our results indicate that this model also applies to β subunit interactions with α₁ subunit intracellular linkers. We have shown previously that β₄subunit alternative splicing had α₁subtype-specific effects on voltage-dependent activation and inactivation (Helton and Horne, 2002). In this report, we extend these findings to include effects on slow inactivation and block by ω-conotoxin (CTx)-MVIIC; we also identify a proline-rich motif in β_4b that is responsible for the observed differences in effects.

MATERIALS AND METHODS

Deletion mutants. Truncation of the β_4b N terminus in 10 aa increments was performed using PCR and custom oligonucleotide primers [Integrated DNA Technologies (IDT), Coralville, IA]. All nucleotide and amino acid positions for primers and restriction enzymes correspond to the β_4b sequence (GenBank accession number U95020). Each forward primer sequence contained an idealized Kozak (1991)sequence and start codon corresponding to the beginning of each of the deleted 10 amino acids as follows: β_4bΔ1–10F, 5′-GCCACCATGACCGCGGACGGGCCG; β_4b Δ1–20F, 5′-GCCACCATGCAGGTGGCCCGAGGC. Both reactions included a common β₄ reverse primer, β_4b 732R (5′-TGACGGCCCCACTAACACC). Full-length β_4b was used as the template for these reactions. The β_4b Δ10–20 deletion mutant was generated with the primer β_4bΔ10–20F (5′-GCCACCATGTCCTCCTCCTCCTACGCCAAGAACTCG) paired with β_4b 732R using the β_4b Δ1–20 mutant as the template. Annealing temperature for PCR was 56°C with Gene Choice Taq DNA polymerase (PGC Scientific, Durham, NC). Correctly sized PCR fragments were cloned into the pT-Advantage vector (Clontech, Palo Alto, CA). PCR-based cycle sequencing (FS chemistry; Applied Biosystems, Foster City, CA) was used with an ABI Prism 310 Genetic Analyzer. The data were analyzed using ABI Prism DNA sequencing software (version 2.12; PerkinElmer Biosystems), and sequence alignments and restriction maps were generated using Lasergene Software (DNA Star, Madison, WI). Correct clones were digested with BamHI (Roche Molecular Biochemicals, Indianapolis, IN), and the corresponding ∼530 bp fragments were ligated into BamHI (nucleotide position 550) -digested β_4b in pBluescript II S/K⁺ (Stratagene, La Jolla, CA). Each β_4b deletion mutant was sequenced to confirm correct reading frame and proper N-terminal orientation.

Site-directed mutagenesis. For all β_4b site-directed mutants, full-length β_4b cDNA (U95020) was used as the template unless otherwise indicated. All site-directed mutagenesis reactions were performed using a QuikChange site-directed mutagenesis kit (Stratagene) and custom forward and reverse compliment oligonucleotide primers (IDT): β_4bG10A,D13A, 5′-ACGCCAAGAACGCGACCGCGGCCGGGCCGCAC; β_4bP15A,P18A, 5′-GCGGACGGGGCGCACTCCGCCACCTCGCAGGTG; β_4bG10A,D13A,P15A,P18A, 5′-ACGCCAAGAACGCGACCGCGGCCGGGGCGCAC (β_4bP15A,P18A used as template); β_4bG10A,P15A, 5′-TACGCCAAGAACGCGACCGCGGACGGGGCGCACTCCCCCACCTCGCAGGTG; β_4bH16A, 5′-ACCGCGGACGGGCCGGCCTCCCCCACCTC; β_4bG10A,P18A, 5′-TACGCCAAGAACGCGACCGCGGACGGGCCGCACTCCGCCACCTCGCAGGTG;β_4bD13A,P15A, 5′-TACGCCAAGAACGGGACCGCGGCCGGGGCGCACTCCCCCACCTCGCAGGTG; β_4bD13A,P18A, 5′-TACGCCAAGAACGGGACCGCGGCCGGGCCGCACTCCGCCACCTCGCAGGTG;β_4bH16A, 5′-ACCGCGGACGGGCCGGCCTCCCCCACCCTCG; and β_4bT11A,S17A,T19A,S20A, 5′-CAAGAACGGGGCCGCGGACGGGCCGCACGCCCCCGCCGCGCAGGTGGCC. Each of the mutant clones was sequenced to confirm reaction fidelity.

Electrophysiology. cRNAs were synthesized in vitro using an mMessage mMachine RNA transcription kit from Ambion (Austin, TX) (T3 or T7 depending on clone orientation in pBluescript II S/K⁺). StandardXenopus laevis oocyte expression methods were used to characterize β deletion and site-directed mutants. Briefly, full-length α₁, α₂/δ-1, and β₄ cRNAs were injected in equimolar ratios (5.6 ng of α_1A, 2.4 ng of α₂/δ-1, and 1.6 ng of β₄ in 46 nl) into defolliculated oocytes (stage V–VI). The BI-2 (α_1A) and α₂/δ-1 clones used in this study were provided by T. Tanabe (Tokyo Medical and Dental University, Tokyo, Japan). Calcium channel currents were recorded 2–4 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 (in mm): 5 Ba(OH)₂, 5 HEPES, 85 TEA-OH, and 2 KOH, pH adjusted to 7.4 with methanesulfonic acid. In experiments with the peptide toxin ω-CTx-MVIIC (Peptide Institute Inc. Osaka, Japan), the 5 mm Ba²⁺ solution was supplemented with 0.1 mg/ml cytochrome c to saturate nonspecific peptide binding sites. Cytochrome c at 0.1 mg/ml had no noticeable effect on recorded Ba²⁺currents. Peptides were reconstituted according to the manufacturer's instructions (100 μm stock solutions in sterile, deionized water). Fresh dilutions of the peptide were made immediately before use. Currents typically ranged between 0.8 and 2.5 μA, and leak currents were between 20 and 100 nA. 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. Curve fitting was performed with SigmaPlot version 5.0 (SSPS Inc., Chicago, IL).

Slow inactivation. Oocytes were held at −80 mV for ∼2 min before a 300 msec reference pulse (I_R) to 0 mV (β_4b) or +10 mV (β_4a). AfterI_R, the membrane potential was stepped immediately to a conditioning pulse potential ranging from −100 mV to −20 mV (20 mV increments) and held for 5 min. During the conditioning pulse, a 300 msec test pulse (I_T) to 0 mV (β_4b) or +10 mV (β_4a) was applied every 15 sec. Data were normalized as the ratio of the maximum current at time T (I_T) divided by the maximum reference current (I_R). Data were fit to the double-exponential equationI_T/I_R=A₁e^−x/τ1+A₂e^−x/τ2, where I_T equals current at time T,I_R equals maximum reference current, x equals time in seconds, and A₁ andA₂ are components for the time constants τ₁ and τ₂, respectively. The SEM is shown for each data point unless the values are smaller than the symbol.

Recovery from slow inactivation. Currents were stabilized at −80 mV or −100 mV for ∼2 min before a 300 msec reference pulse (I_R) to 0 mV (β_4b) or +10 mV (β_4a). After a 100 msec step to either −80 mV or −100 mV, the oocytes were held at a conditioning pulse potential of −30 mV for 5 min. Immediately after the conditioning pulse, a 300 msec test pulse was applied (I₁), and then the holding potential was stepped back to either −80 mV or −100 mV and 300 msec test pulses (I_2–12) were applied at 15 sec intervals starting at time 0 for a total of 3 min. Data were normalized as the ratio of the maximum current at time T (I_T) divided by the maximum reference current (I_R). Data were fit to the single exponential equationI_T/I_R =I_∞ +Ae^−x/τ, whereI_T equals current at time point T,I_R equals maximum reference current,I_∞ equals current remaining at end of protocol, x equals time in seconds, and A is the component for the time constant τ_.

Voltage dependence of activation and inactivation. Voltage dependency of activation data was generated from I–Vcurves. Maximal currents were obtained from 300 msec depolarizations from a holding potential of −80 mV to various test potentials (−40 to +10 mV in 5 mV increments). Each individual recording was then normalized, inverted, and fit to the Boltzmann equation %I_Ba = 1/[1 + exp(−(V_test −V_1/2)/k)], whereV_test equals I–V test potential, V_pre equals prepulse potential, V_1/2 equals midpoint of activation or inactivation, and k equals slope factor. The fit curves, V_1/2, and kvalues were then averaged and plotted as a function of membrane voltage.

Voltage dependency of inactivation data was obtained from peak currents elicited by a 300 msec maximal current test depolarization after a 20 sec conditioning prepulse to voltages ranging from −80 to +20 mV. Each individual recording was then normalized and fit to the Boltzmann equation %I_Ba = 1/[1 + exp([V_pre −V_1/2]/k)], whereV_test equals I–V test potential, V_pre equals prepulse potential, V_1/2 equals midpoint of activation or inactivation, and k equals slope factor. The fit curves, V_1/2, and kvalues were then averaged and plotted as a function of prepulse potential.

Pharmacology. Oocytes were held at a potential of −80 mV with maximal currents elicited by 150 msec test pulses to 0 mV (β_4b) or +10 mV (β_4a) every 15 sec for a total of 10 min. During recordings, oocytes were perfused at a constant rate of ∼0.5 ml/min. Average current sizes for β_4a and β_4b complexes were 1.9 ± 0.2 μA (4–5 d of incubation) and 2.3 ± 0.3 μA (2–3 d of incubation), respectively. The data were fit to the single-exponential equationI_T/I_R= I_∞ +Ae^−x/τ, whereI_T equals maximum current at time point T, I_R equals maximum current at time point 0, I_∞ equals residual current at end of protocol, x equals time in seconds, and A is the component for the time constant, τ. The averaged rate constants (1/τ) for the four ω-CTx-MVIIC concentrations (0.2, 0.6, 2, and 6 μm) were plotted as a log function of their concentration and were fit well by the equation (τ)⁻¹ =k_on[Tx] +k_off.

Northern blot analysis. A commercially available human neuronal tissue Northern blot [Multiple Tissue Northern (MTN) Blot brain II; Clontech] was probed with a nonspecific β₄ subunit probe (β₄ΔN; nucleotides 215–1628 plus ∼300 bp of 3′ untranslated). A ³²P-labeled β₄ subunit probe was made with a nick translation kit (Promega, Madison, WI) using the β₄ΔN mutant as the template. The β₄ΔN mutant is missing the first coding 147 bp corresponding to the 49 aa N terminus of β₄[clone from Helton and Horne (2002)]. The MTN blot was hybridized overnight at 42°C in hybridization buffer [5× SSC, 5% w/v blocking reagent (Roche), 0.1% N-lauroylsarcosine, 0.02% w/v SDS, 50% w/v formamide] plus 100 μg/ml herring sperm DNA (Promega). The probe concentration was 1 million counts/ml. The blot was washed with successive stringency washes (four washes, 15 min each at 37°C) ranging from 2× SSC/0.1% SDS to 0.1× SSC/0.1% SDS. The blot was then exposed to radiographic film for 12 hr at −80°C. One microgram of cRNA for both β_4a and β_4b was run out on a 1% denaturing formaldehyde gel along with a poly(A)-tailed cRNA mass ladder (RNA Molecular Weight Marker 1; Roche). The β_4a cRNA is longer than the β_4b cRNA because of the additional ∼400 nt of 5′ untranslated sequence.

Molecular modeling. The sequences for rat postsynaptic density-95 (PSD-95) (DLG4_rat) and human β₄ (CACNB4) were obtained (accession numbers P31016 and U95020) from the Swiss-prot database. Amino acids 10–96 of β_4b were aligned to residues 303–390 of PSD-95 based on secondary structure prediction (nnPredict) and visual inspection. For β_4a, amino acids 50–96 of β_4b were aligned to residues 345–390 of PSD-95. Using default parameters, the program MODELLER 6 (Sali and Blundell, 1993) was used to produce 50 models each of β_4b and β_4a structure based on the solved structure of the third PSD-95/Discs large/zona occludens-1 (PDZ) domain of PSD-95 (1BEF). Five models each were chosen for additional analysis based on the molecular probability density function (PDF) output from MODELLER and stereochemical analysis obtained through Ramachandran output from PROCHECK-NMR (Laskowski et al., 1993). The interactions between different atom types within these models and Cα root mean square deviation (RMSD) comparisons between the models and 1BEF were characterized with ERRAT (Colovos and Yeates, 1993). The β_4a and β_4b models chosen for comparison had the fewest disallowed residues (Ramachandran), lowest molecular PDF and RMSD values, and highest percentage of residues in acceptable conformations based on ERRAT and PROCHECK-NMR analysis. Models were visualized with the program MOLMOL (Koradi et al., 1996).

RESULTS

Alternative splicing of the β₄ subunit affects slow inactivation of Ca_v2.1 Ca²⁺ channels

In a previous study (Helton and Horne, 2002), we showed that Ca_v2.1 complexes containing the longer form of an alternatively spliced β₄ subunit N terminus, β_4b (49 aa), inactivated at more negative potentials in response to 20 sec conditioning prepulses than complexes containing a shorter form, β_4a (15 aa). To determine whether this response extended to slower types of inactivation, we examined in the present study the effects of β_4a and β_4b on Ca_v2.1 cumulative inactivation elicited by 5 min conditioning prepulses combined with stimulation at 0.25 Hz. Oocytes were stabilized at −80 mV before a 300 msec reference pulse (I_R) to potentials that were predetermined to give peak inward currents (β_4b, 0 mV; β_4a, +10 mV). The membrane potential was then stepped to and held at the conditioning prepulse potential (ranging from −100 to −20 mV) for 5 min. A 300 msec test pulse (I_T) was elicited from the conditioning prepulse potential every 15 sec (I₅ equals test pulse at 5 min). The kinetics of entry to slow inactivation for Ca_v2.1 complexes containing either β_4a or β_4b at −40 mV is shown in Figure1A. For comparison purposes, we fit the data points for both β_4aand β_4b to two exponentials (smooth curves in the figure). The time constants for the fast component of entry (τ₁) for β_4a and β_4b were 28.6 ± 2.6 and 18.9 ± 1.2 sec, respectively, and the time constants for the slow component of entry (τ₂) were 769 ± 23.6 and 384 ± 14.8 sec, respectively. Overall, theI_T/I_Rratio for Ca_v2.1 complexes containing β_4b decreased to 0.5 in ∼70 sec, whereas those containing β_4a required ∼380 sec (data not shown). This indicated that β_4b caused a more than fivefold acceleration of the kinetics of slow inactivation. Representative current traces for reference and 5 min test pulses from a conditioning potential of −40 mV for Ca_v2.1 complexes containing β_4a (top) and β_4b (bottom) are shown in Figure1B. As seen in the figure and as described in our previous study, Ca_v2.1 complexes containing β_4a underwent open-state fast inactivation faster than did complexes containing β_4b. After 5 min at −40 mV, the rate of fast inactivation was unaltered for complexes containing β_4a, and slowed only somewhat for complexes containing β_4b. The absence of any appreciable tail-current indicated that deactivation was not affected by prolonged depolarization. TheI₅/I_Rratio is plotted against the range of conditioning potentials (−100 to −20 mV) in Figure 1C. The figure illustrates that the voltage dependence of Ca_v2.1 slow inactivation is shifted to the left for complexes containing β_4b relative to those containing β_4a. Half-maximal inactivation occurred at approximately −50 mV for complexes containing β_4b and −35 mV for β_4a. These values are ∼10 mV (β_4b) and 5 mV (β_4a) more negative than were determined for inactivation in response to 20 sec conditioning prepulses (Helton and Horne, 2002). Figure1D shows that recovery from 5 min of slow inactivation at −30 mV is nearly complete when the membrane potential is stepped back to −80 mV, and that there is no difference in the time course of recovery for Ca_v2.1 complexes containing either β_4a or β_4b. Recovery was somewhat faster and more complete when the membrane potential was stepped back to −100 mV. The recovery data at both potentials fit well to single exponentials. The time constants for recovery for β_4a and β_4b at −80 mV were 28.6 ± 2.0 and 27.8 ± 1.6 sec, respectively, and at −100 mV, 18.9 ± 0.5 and 17.2 ± 0.4 sec, respectively.

Fig. 1.

Effects of β_4a and β_4bon slow inactivation and recovery from slow inactivation of Ca_v2.1 Ca²⁺ channels. Studies were performed with Xenopus oocytes expressing α_1A, α₂/δ-1, and either β_4a or β_4b. Reference (I_R) and test current (I_T) traces were generated by 300 msec step depolarizations from various holding potentials to either 0 mV (β_4b) or +10 mV (β_4a). Maximum values from 300 msec I_R andI_T current traces were used to calculateI_T/I_Rwhere indicated. Barium (5 mm) was used as the charge carrier. A, Influence of β_4a and β_4b on the development of slow inactivation at a conditioning potential (CP) of −40 mV. After a reference pulse (I_R) measured from a holding potential of −80 mV, oocytes were held at −40 mV for 5 min. During this time, 300 msec test pulses (I_T) were applied every 15 sec. Eachpoint represents the mean value ofI_T/I_R from 11 (β_4a) or 10 (β_4b) different recordings. The SEM is shown for each pointunless the values were smaller than the symbol. Thesolid lines represent double-exponential fits to the data. B, Representative reference (I_R) and 5 min (I₅) current traces from Ca_v2.1 complexes containing either β_4a(top) or β_4b (bottom) generated as described in A. C, Voltage dependence of slow inactivation. The ratio ofI₅ to I_R, generated as in A, plotted as a function of conditioning potential for Ca_v2.1 complexes containing either β_4a or β_4b. Data pointsrepresent the means of at least six determinations at a given membrane potential. Lines serve only to connect the data points. D, Influence of β_4a and β_4b on the time course of recovery from slow inactivation. After a 300 msec reference pulse (I_R) measured from a holding potential of either −80 or −100 mV, oocytes were held at −30 mV for 5 min. The membrane potential was then returned to either −80 or −100 mV, and sequential test pulses (I_T) were applied at 15 sec intervals for a total of 3 min. Eachpoint represents the mean of at least seven different recordings. Solid lines represent the single-exponential fits of the data.

Alternative splicing of the β₄ subunit affects ω-CTx-MVIIC block of Ca_v2.1 Ca²⁺channels

The results to this point indicate that changes in the structure of the β₄ subunit N terminus impact α_1A subunit structures that are important for many aspects of gating, including activation, open-state inactivation, and fast and slow closed-state inactivation. Given that recent evidence indicates that cytosolic determinants of two-membrane-spanning K⁺ channel gating are coupled to changes in outer vestibule structure (Perozo et al., 1999; Jiang et al., 2002a,b), we next sought to determine whether alternative splicing of the β₄ subunit would affect the block of Ca_v2.1 channels by a marine snail peptide conotoxin, ω-CTx-MVIIC. Conotoxin interactions with voltage-gated Ca²⁺ channels are entirely extracellular and occur through binding sites located near H5 (P) helices in several of the six helix transmembrane-spanning motifs (Ellinor et al., 1994). Figure 2A shows the effects of 2 μm ω-CTx-MVIIC on Ca_v2.1 Ca²⁺ channel complexes expressed in Xenopus oocytes in the presence of either β_4a or β_4b. The oocytes were held at −80 mV for 10 min and stimulated every 15 sec. Under these conditions, ω-CTx-MVIIC associated with Ca_v2.1 complexes containing β_4b at a faster rate (τ = 50 ± 0.75 sec) than complexes containing β_4a(τ = 200 ± 16 sec). The loss of Ca²⁺ current resulting from slow inactivation over 10 min at −80 mV (<15%) was subtracted from the data plotted in the figure. Figure 2B demonstrates that as expected for a first-order reaction, the rate constants (τ⁻¹) for toxin block were linearly dependent on toxin concentration as described by the equation (τ)⁻¹ =k_on [Tx] +k_off. Slopes of linear fits to the data for Ca_v2.1 complexes containing either β_4a and β_4b were 3.7 × 10⁻⁶m⁻¹ · sec⁻¹ and 1.1 × 10⁻⁵m⁻¹ · sec⁻¹, respectively. This indicated that the on-rate (k_on) for toxin block was approximately threefold faster for Ca_v2.1 complexes containing β_4b than for those containing β_4a.

Fig. 2.

Effects of β_4a and β_4bon the blockade of Ca_v2.1 channels by ω-CTx-MVIIC. Studies were performed with Xenopus oocytes expressing α_1A, α₂/δ-1, and either β_4a or β_4b.A, Onset and degree of block by a 10 min exposure to 2 μmω-CTx-MVIIC for Ca_v2.1 subunit combinations at a holding potential (HP) of −80 mV. Each pointrepresents the mean of seven (β_4a) or eight (β_4b) different recordings. The SEM is shown for each data point unless smaller than thesymbol. Onset of block for both subunit combinations was fit (line) to a single-exponential time course plus a constant. B, The rate constants for the time course of the onset of toxin block were determined from steady-state degree of block from single exponential fits at four different toxin concentrations (0.2, 0.6, 2, and 6 μm) for Ca_v2.1 complexes containing either β_4a or β_4b. The averaged rate constants were plotted as a function of toxin concentration (minimum of n = 7, ±SEM). The line represents a linear fit to the data.

The molecular determinants of alternatively spliced β₄ subunit differential effects on gating and pharmacology are located within amino acids 10–20 of β_4b

Having characterized many of the functional consequences of alternative splicing of the β₄ A domain, we focused next on identifying the key structural determinants underlying the observed differences in effects. It was of particular interest to determine whether or not the effects of alternative splicing on gating and pharmacology could be assigned to separate structural entities. To accomplish this, we first created a series of β_4b deletion mutants in which the N terminus was shortened by multiples of 10 aa (β_4bΔ1–10 through β_4b Δ1–49) and characterized their effects on gating and pharmacology of Ca_v2.1 complexes. Figure3A–D shows that, relative to full-length β_4b, deletion of the first 10 aa (β_4b Δ1–10) had no effect on the voltage dependence of activation (Fig. 3A), isochronal (20 sec prepulse) inactivation (Fig. 3B), onset into slow inactivation (Fig. 3C), or susceptibility to block by 2 μm ω-CTx-MVIIC (Fig. 3D). However, when amino acids 1–20 were removed (β_4bΔ1–20), both the voltage dependence of activation (Fig.3A) and inactivation (Fig. 3B) of Ca_v2.1 complexes shifted to more depolarized potentials. As shown in the figure, the acquired gating properties were essentially identical those for Ca_v2.1 complexes containing β_4a. Ca_v2.1 complexes containing β_4b Δ1–20 also had a slower onset into slow inactivation (Fig. 3C) and were less susceptible to block by 2 μm ω-CTx-MVIIC (Fig. 3D). The effects of constructs β_4b Δ1–30, β_4bΔ1–40, and β_4b Δ1–49 were identical to those of β_4b Δ1–20 (data not shown). As a first attempt at determining whether the effects of β_4b Δ1–20 were simply the result of a decreased size of the β_4b N terminus, we reintroduced amino acids 1–10 to the N terminus of β_4b Δ1–20 to create the construct β_4b Δ10–20. As shown in Figure3A,B, this did not restore theV_1/2 of either activation or inactivation to the hyperpolarized potentials characteristic of Ca_v2.1 complexes containing β_4b. Together, these results indicated that the molecular determinants responsible for the observed differences between Ca_v2.1 complexes containing β_4a versus β_4b were located in amino acids 10–20 of β_4b. Moreover, it was apparent that their influence extended to changes in both gating and pharmacology.

Fig. 3.

Localization of differential effects on Ca_v2.1 gating and pharmacology to β_4bN-terminal amino acids 10–20. The first 10 (β_4bΔ1–10), first 20 (β_4b Δ1–20), or second 10 (β_4b Δ10–20) aa of the N terminus of the β_4b subunit were removed using PCR. The deletion mutants as well as β_4a or β_4b were expressed with α_1A and α₂δ-1 in Xenopusoocytes. A, Effects of the N-terminal deletion mutants on the voltage dependency of activation of Ca_v2.1 channels. Plots were derived from averaged I–V data up to +10 mV for each β₄ subunit combination. Data pointsrepresent the means of the normalized data at a given membrane potential for a minimum of nine different recordings. Smooth lines represent single Boltzmann fits to the averaged data.B, Normalized, averaged isochronal inactivation curves for Ca_v2.1 complexes containing the various β₄ subunits. Points represent the means of the normalized data at a given membrane potential for a minimum of nine different recordings. Smooth lines represent single Boltzmann fits to the averaged data. C, Effects of β₄ N-terminal deletion mutants on the development of slow inactivation at a conditioning potential (CP) of −40 mV. Reference (I_R) and test (I_T) currents were generated as in Figure 1A. Each point represents the mean value ofI_T/I_R from 13 (β_4b Δ1–10) or nine (β_4b Δ1–20) different recordings. The solid lines represent double-exponential fits to the data. D, Onset and degree of block by a 10 min exposure to 2 μm ω-CTx-MVIIC for Ca_v2.1 complexes containing β_4b Δ1–10 or β_4b Δ1–20. Data were generated as in Figure2A. Each point represents the average of a minimum of seven recordings. The solid lines represent single-exponential fits to the data.

The β₄ A domain is a distant homolog of the third PDZ domain of PSD-95

With the results of the deletion experiments highlighting a specific location for the molecular determinants of β_4b gating and pharmacology effects, and with the observation that the β_1b A domain resembles PDZ domains (Hanlon et al., 1999), we began a systematic comparison of the β_4b sequence with similar regions of a number of PDZ domains. Unexpectedly, we found that the entire β_4b A domain was weakly homologous to the third PDZ domain of PSD-95 (Fig.4A). Of the 87 aa that have been shown by x-ray crystallography to form the modular PDZ structure of PSD-95 (Doyle et al., 1996), 27 of these (∼31%) are conserved in the β_4b sequence. Most importantly, these identities are conserved within key secondary structural elements, such as β-strand C and α-helix 2 of PSD-95. Also of note is the conservation of an RG(S/T)T motif in what would be the equivalent of the carboxylate binding loop (CBL) between β-strands A and B of PSD-95 and the loss of the GLGF motif that is extremely common among PDZ domain subtypes (Bezprozvanny and Maximov, 2001; Harris and Lim, 2001). Four of β_4b amino acids 10–20 (G10, D13, P15, and P18) were found in PSD-95. We used these as a starting point for further defining key β_4b residues involved in setting Ca_v2.1 gating parameters.

Fig. 4.

The β₄ subunit is a distant homolog of PSD-95. Identification of a conserved GXXDXPXXP motif critical to Ca_v2.1 gating. A, Amino acid alignment of the A domains of the human spinal cord β_4a (amino acids 1–63) and β_4b (amino acids 1–97) subunits with the third PDZ domain (amino acids 294–391) of PSD-95. Vertical bars denote identical amino acids between β_4b and PSD-95. Important amino acids involved in modulating the leftward shift in the voltage dependence of activation and inactivation of β_4b (GXXDXPXXP) are highlighted.Arrows and hatched bars represent predicted α-helices and β-strands of the third PDZ domain of PSD-95, respectively. B, Differences in theV_1/2 values of activation and inactivation of β_4a, β_4b, and β_4b N-terminal amino acid mutants. Solid bars represent average V_1/2 values of a minimum of nine different recordings for each β₄subunit variant. Positive or negative shifts in theV_1/2 values (in millivolts) of activation and inactivation of β_4a and β_4b mutants are relative to the V_1/2 values of activation and inactivation of β_4b. Currents were generated as described in Figure 3, A and B. The SEM for each bar is shown. Asterisks denote statistical significance (p < 0.05) by a Student's two-sample equal variance t test.

Figure 4B lists a series of site-directed mutants (left), along with their effects on the voltage dependence (V_1/2) of activation (middle) and inactivation (right) of Ca_v2.1 complexes. TheV_1/2 values for complexes containing β_4a and β_4b are included for comparison. Interestingly, the first site-directed β_4b mutant tested, G10A, D13A, P15A, P18A, in which all four of the amino acids in common with PSD-95 were altered, displayed activation and inactivation properties similar to that of β_4a. To determine whether this was a specific effect, we altered four different amino acids in the β_4b 10–20 sequence to create the mutant, T11A, S17A, T19A, S20A. As shown in Figure 4B, Ca_v2.1 gating properties changed little in response to these mutations. Ca_v2.1 complexes containing the G10A, D13A, P15A, P18A mutant also had β_4a-like slow inactivation and pharmacological properties (data not shown). This indicates that the conserved amino acids are playing a defining role in the gating motif. To delineate the structure in more detail, we subsequently characterized six of the possible G10, D13, P15, P18 amino acid pairs for their effects on gating. Surprisingly, none of the pairs were absolutely essential for maintaining wild-type β_4b gating behavior, although small but statistically significant hyperpolarizing effects on activation were noted for five of the six pairs. To complete the alanine substitution study, we created the mutant H16A, which had a small but statistically significant effect on inactivation but not activation.

One interpretation of these results is that β_4bamino acids 10–20 form a ligand motif that interacts with a binding pocket located somewhere either on the α_1Asubunit or on the β₄ subunit itself. The affinity of the ligand motif for its receptor site could be defined, for example, by the sum of the interactions of amino acids G10, D13, P15, and P18 with their individual targets. Any given pair may be capable of maintaining a binding interaction under the conditions of our experiments. As a first step toward addressing this possibility, we created three-dimensional structural models of the β_4a and β_4b A domains (Fig. 5A,B) using the real-space optimization method used in the computer program MODELLER (Sali and Blundell, 1993). The models were initiated using the distance and dihedral angle restraints derived from alignments with portions of the sequence of the third PDZ domain of PSD-95. For β_4b, amino acids 10–96 were aligned with amino acids 303–390 of PSD-95. There is 31% sequence identity over this region, which is considered minimally acceptable for this type of comparative modeling (Martí-Renom et al., 2000). For β_4a, amino acids 10–49 of β_4b were deleted from the alignment. Thus, the models do not include the first 15 aa of β_4aand the first 9 aa of β_4b. Figure 5A(left) shows that β_4a models as a compact structure containing three β-sheets and 2 α-helices. Stereochemical quality assessment of the model using PROCHECK-NMR (Laskowski et al., 1993) identified 41 residues (87.7%) in most favored regions, 5 (10.6%) in additional and generously allowed regions, and 1 (2.1%) in a disallowed region. Calculation of the electrostatic surface potential using MOLMOL (Koradi et al., 1996) reveals that the face of the molecule as oriented in Figure5A, left, contains a pocket of negative charge (red residues) between the two α-helices (Fig. 5A,right). Figure 5B, right andleft, illustrates that the overall effect of alternative splicing to form β_4b is to bury the charged pocket beneath three additional β-sheets. Interestingly, the molecule acquires as the result of splicing a positively charged binding pocket (blue residues) in what would be the equivalent of the CBL in PSD-95 (Fig. 4A). The stereochemical quality of the β_4b model as shown is not quite as good as that for β_4a. PROCHECK-NMR identified 64 residues (79%) in most favored regions, 11 residues (13.6%) in additional allowed regions, and 3 residues (3.7%) each in generously allowed and disallowed regions. Two of the three disallowed residues (R30 and K34) flank what would be the equivalent PSD-95 β-sheet B. Together with the loss of the highly conserved PDZ GLGF sequence, these results are consistent with the notion that through evolution this region of the β_4b structure has evolved away from the capacity to bind C-terminal peptide motifs. Of most importance to our present results, however, is the observation that β_4b amino acids 10–20 model as an extended arm (pointing to the left in Fig. 5B, right, left) that may serve as a ligand motif. Interestingly, the orientation of the arm appears to be dictated by the isomerization state of proline18 (data not shown).

Fig. 5.

Real-space optimization structural models of the A domains of β_4a (A) and β_4b (B) based on sequence identities with the third PDZ domain of PSD-95. Ribbon (left) and electrostatic surface potential (right) diagrams were created using MOLMOL (Koradi et al., 1996). For ribbon diagrams, arrows indicate β-strands, and helices indicate α-helices. For surface potential diagrams, red, white, and blue regions indicate negatively charged, hydrophobic, and positively charged amino acids, respectively.

Differential distribution of alternatively spliced β₄subunit mRNA

We noted in our previous study (Helton and Horne, 2002) that, based on extensive cDNA library screening, β_4awas the predominant alternatively spliced variant of the β₄ subunit expressed in human spinal cord. To confirm this observation, we performed a comparative Northern blot analysis using a commercially available multiple tissue Northern blot (Human Brain II; Clontech) and a β₄ cDNA probe containing sequence common to both β_4a and β_4b. The mRNAs for β_4aand β_4b can be readily distinguished by their distinct migration pattern in agarose-formaldehyde gels (Fig.6A). The results of the Northern blot analysis, shown in Figure 6B, were striking, revealing that not only was β_4a the predominant form of β₄ subunit in the spinal cord, but also in other “reptilian” portions of the human CNS such as the medulla and putamen. Moreover, β_4a was the predominant form of β₄ subunit expressed in evolutionarily older regions of the cerebrum, the temporal lobe, and occipital pole. In marked contrast, β_4b was highly expressed in the most recent and most highly integrative region of the cerebrum, the frontal lobe. The two forms of the β₄ subunit were equally expressed in the cerebellum.

Fig. 6.

Differential distribution of β_4a and β_4b mRNA in the human CNS. A, Electrophoresis of full-length β_4a (left) and β_4b (right) cRNAs (includes 5′ and 3′ untranslated) in a 1% agarose formaldehyde denaturing gel. RNA markers (in kilobases) are indicated on the left.B, Northern blot analysis performed with human multiple tissue blot (Human Brain II; Clontech) and a ³²P-labeled β₄ subunit probe (coding nucleotides 215–1628 plus ∼300 bp 3′ untranslated sequence). Molecular masses on theright correspond to labeled blot markers.C, A human β₄ subunit genome map depicting the lengths of intron sequences (b, bases) between alternatively spliced β_4a and β_4b N-terminal exons and the beginning of exon 2. Solid lines represent exons, and dashed lines represent introns.Numbers in parentheses below solid lines indicate position on chromosome 2. Boxesindicate protein sequence (β_4a inparentheses). The first and last two amino acids of each sequence are indicated above each box.

A basic local alignment search tool search of the human genome (Altschul et al., 1990) with β₄ sequences revealed that the exons coding for alternatively spliced forms of the β₄ subunit A domain are distributed widely on human chromosome 2. Figure 6C shows that, depending on the splice variant, the coding sequence for the β₄PDZ domain is contained within three (β_4a) or four (β_4b) exons spread out over ∼218,000 bases. The coding sequence for the GXXDXPXXP motif is included in the 5′-most exon of a pair of short exons that code for β_4b amino acids 1–49. Assembly of the β_4b mRNA requires that three RNA segments (536 bases, 214,959 bases, and 2329 bases) be spliced out. The short exon coding for the first 15 aa of β_4a lies between the β_4b N-terminal exons and the exon coding amino acids 50–89 and 16–55 of β_4b and β_4a, respectively. By comparison, the third PDZ domain of PSD-95 is encoded by two exons separated by a 200 bp intron (data not shown).

DISCUSSION

We have identified an alternatively spliced proline-rich motif in the Ca²⁺ channel β₄ subunit that has considerable influence over gating of neuronal Ca_v2.1 Ca²⁺ channels. Given that the motif also affects extracellular toxin binding, it is likely that the interaction of this motif with its binding site has wide-reaching impact on resting and open-state Ca²⁺ channel conformations. This notion is supported by recent images of the conformational changes that occur with gating of bacterial two-membrane-spanning K⁺ channels (Liu et al., 2001; Jiang et al., 2002a,b). Like eukaryotic six-membrane-spanning K⁺ channels, KcsA and MthK channels are tetramers that pack with fourfold symmetry around a central pore (Doyle et al., 1998; Jiang et al., 2002a). The principal structural elements of KcsA and MthK from N to C terminus include an outer transmembrane helix (M1), a pore helix (P), and an inner transmembrane helix (M2). These correspond to S5, H5, and S6 segments of voltage-gated Ca²⁺ channels, respectively. In the closed conformation of the KcsA structure, the four M2 helices are straight and arranged such that they form the walls of an inverted teepee that narrows from a 12 Å diameter at its center to a 4 Å pore at its tip (Doyle et al., 1998). After opening, the KcsA M2 helices tilt away from the permeation pathway and rotate about their helical axis (Liu et al., 2001). In MthK, bending and splaying of the inner helices after opening expands the diameter of the pore threefold (Jiang et al., 2002a,b). The nearly 30° bend occurs at a “gating hinge” corresponding to a glycine residue just below the selectivity filter. Applying a radial-outward force on the intracellular aspects of the inner helices places a torque on the gating hinge such that a conformational change is transmitted the full length of the M2 helix. This results in a widening of the outer vestibule [Jiang et al. (2002b), see supplementary information (movie)]. It is has been hypothesized that similar mechanical forces are at work in the gating of voltage-gated Ca²⁺ channels (Jiang et al., 2002b).

Considered in the context of this mechanical framework, our results suggest that an interaction of the β_4b ligand motif with an inner aspect of the Ca_v2.1 complex either directly or indirectly fine-tunes the torque experienced by α_1A S6 segments. In so doing, the interaction alters the conformation of the voltage sensor or gate (or both) as well as the outer vestibule. This would explain why alternative splicing of the β₄ subunit would affect both gating and toxin sensitivity. Given the potential for the ligand motif interaction to occur over a wide range (Helton and Horne, 2002), it is not possible from our results to pinpoint which S6 helices might be most affected. And although our ω-CTx-MVIIC binding results might be providing some direction, a previous study has shown that ω-CTx-GVIA binding to α_1B is affected by alterations in any of the four P loops that make up the outer vestibule (Ellinor et al., 1994). A case can be made for an indirect effect on the IS6 helix, because the primary α₁-β₄ subunit interaction occurs on the intracellular loop between homology domains I and II (I-II loop) (Pragnell et al., 1994). This is consistent with a previous study showing that IS6 is a critical determinant of voltage-dependent inactivation in Ca_v2.1 and Ca_v2.3 channels (Zhang et al., 1994). However, site-directed mutagenesis and domain-swapping studies have highlighted the equal importance of IIS6, IIIS6, and IVS6 in Ca²⁺ channel gating (for review, see Stotz and Zamponi, 2001; Shi and Soldatov, 2002), making the case for direct effects on these S6 helices equally plausible. It is interesting that many of the proteins that have evolved to modulate Ca²⁺ channel gating target α₁ I-II (β subunits, G_βγ subunits, protein kinase C) and II-III loops (syntaxin, synaptotagmin, and synaptosomal-associated protein-25). The IS6 and IIS6 (but not IIIS6 or IVS6) helices of Ca_v2.1, 2.1, and 2.3 Ca²⁺ channels have glycines in hinge positions comparable with those present in KcsA and MthK (see alignments in Horne et al., 1993).

Despite extensive binding studies with β₄subunits, there is currently no evidence to support the notion that the β_4b N terminus binds directly to α_1A subunits (Walker et al., 1998, 1999). One possible explanation for this is that the interaction is too weak to be detected in solution binding assays. The other possibility is that the α_1A subunit is not the primary binding target. The sequence of the ligand motif (GXXDXPXXP) may provide an important clue as to the nature of its own binding site. Proline-rich motifs are common within the primary structures of many ligands important for protein–protein interactions (for review, see Kay et al., 2000; Macias et al., 2002). Src homology 3 (SH3) and WW domains, for example, recognize proline-rich sequences containing a core PXXP, where X denotes any amino acid. These sequences adopt a PPII helix conformation that presents a hydrophobic surface as well as backbone carbonyls that are ideal for hydrogen bonding. Proline-rich ligands bind with low affinity, allowing for rapid modulation and added versatility in signaling pathways. In many respects, the β_4bligand motif resembles the serine/threonine proline motifs recognized by group IV WW domains (Sudol and Hunter, 2000). As is the case for the β_4b motif, the structural basis for recognition of these motifs is based on the summed contributions of a series of side-chain interactions, none of which is absolutely essential for ligand binding (Verdecia et al., 2000). It is possible that the β_4b proline-rich motif binds to its own B domain, which is structurally similar to SH3 and WW domains (Hanlon et al., 1999). This possibility is supported by what is known about related membrane-associated guanylate kinase family proteins in which intramolecular interactions are a key aspect of their functional diversity (Dimitratos et al., 1999).

To date, no specific function has been assigned to the GXXDXPXXP motif of the third PDZ domain of PSD-95. Most attention has been paid to the structure of the carboxylate binding loop to which it is immediately adjacent (Doyle et al., 1996) (Fig. 4, CBL). A partial list of the proteins that interact specifically with the third PDZ domain of PSD-95 include the cell-surface neuroligins (Irie et al., 1997), the microtubule binding protein cysteine-rich interactor of PDZ 3 (Niethammer et al., 1998), the Rho effector protein citron (Zhang et al., 1999), and the β₁ adrenergic receptor (Hu et al., 2000). It would be interesting to determine whether the GXXDXPXXP motif of PSD-95 plays a role maintaining the structure of the CBL. Such a role could also be considered for the GXXDXPXXP motif of β_4b. As is apparent in the β_4b A domain model structure (Figs.4A, 5B), the sequences of β_4b corresponding to the hydrophobic CBL and β-strand B of PSD-95 are highly positively charged. It is possible, as is true for some PDZ domains (Cuppen et al., 1998), that this region binds to an internal (as opposed to a C-terminal) negatively charged domain. Interestingly, the immediate 5′ sequence of the II-III loop of Ca_v2.1 and Ca_v2.2 Ca²⁺ channels fits this description, because it is densely packed with glutamate residues. Moreover, several lines of evidence point to the II-III loop as a critical determinant of channel gating and pharmacology. Similar to the effects of β_4b on α_1A, binding of syntaxin to a “synprint site” just downstream from this region in the II-III loop of α_1B shifts theV_1/2 of inactivation to more hyperpolarized potentials (Bezprozvanny et al., 2000) and accelerates entry into slow inactivation (Degtiar et al., 2000), and an α_1B splice variant that lacks a large portion of the II-III linker region (Δ1) (Kaneko et al., 2002) inactivates at more depolarized potentials and is less sensitive to ω-CTx-MVIIA than is the full-length Ca_v2.2 variant.

Viewed from the perspective of changing Ca²⁺ channel function, the differential distribution of β_4a and β_4b subunit mRNA displayed in Figure6B provides an unexpected snapshot of the evolution of forebrain synapses. It raises the possibility that with the introduction of β_4b to the genome, synapses acquired properties that fit better with the overall demand to organize complex neural networks. It could be that with the advent of Ca_v2.1 complexes that enter more readily into closed inactivation states (Fig. 1A,C) without perceptible gain in time required for recovery (Fig.1D), synapses inherited an enhanced mechanism for synaptic plasticity. In this regard, short-term synaptic depression has been linked to Ca²⁺ channel inactivation (Forsythe et al., 1998) through mechanisms shown to be β subunit dependent (Patil et al., 1998). This form of short-term synaptic plasticity has been implicated in cortical gain control (Abbott et al., 1997) and low-pass temporal filtering (Fortune and Rose, 2001). In addition, long-lasting Ca_v2.1 channel inhibition has been identified as the mechanism underlying certain forms of long-term synaptic depression (Robbe et al., 2002). Accordingly, our future studies will be directed toward characterizing the responsiveness of Ca_v2 Ca²⁺ channels containing alternatively spliced β₄ subunits to changes in more dynamic regulatory inputs, such as neuronal firing frequency and action potential waveform.