Calcium regulates a wide spectrum of physiological processes such as heartbeat, muscle contraction, neuronal communication, hormone release, cell division, and gene transcription. Major entryways for Ca2+ in excitable cells are high-voltage activated (HVA) Ca2+ channels. These are plasma membrane proteins composed of several subunits, including α1, α2δ, β, and γ. Although the principal α1 subunit (Cavα1) contains the channel pore, gating machinery and most drug binding sites, the cytosolic auxiliary β subunit (Cavβ) plays an essential role in regulating the surface expression and gating properties of HVA Ca2+ channels. Cavβ is also crucial for the modulation of HVA Ca2+ channels by G proteins, kinases, and the Ras-related RGK GTPases. New proteins have emerged in recent years that modulate HVA Ca2+ channels by binding to Cavβ. There are also indications that Cavβ may carry out Ca2+ channel-independent functions, including directly regulating gene transcription. All four subtypes of Cavβ, encoded by different genes, have a modular organization, consisting of three variable regions, a conserved guanylate kinase (GK) domain, and a conserved Src-homology 3 (SH3) domain, placing them into the membrane-associated guanylate kinase (MAGUK) protein family. Crystal structures of Cavβs reveal how they interact with Cavα1, open new research avenues, and prompt new inquiries. In this article, we review the structure and various biological functions of Cavβ, with both a historical perspective as well as an emphasis on recent advances.
Calcium is arguably one of life's most important elements. Intracellular Ca2+ concentration ([Ca2+]i) is kept at very low levels (∼100 nM) under resting conditions, but it rises sharply (to tens or hundreds of μM) upon stimulation. This allows Ca2+ to play a crucial role in numerous biological processes, including neurotransmitter and hormone release, muscle excitation-contraction coupling, cell division, tumorigenesis, differentiation, migration, and cell death. In addition, Ca2+ influx across the plasma membrane causes changes in cellular excitability. Mechanisms that rigorously control intracellular Ca2+ levels are therefore essential for eukaryotic cell function. [Ca2+]i is maintained at low levels by Ca2+-ATPases through active extrusion of cytosolic Ca2+ to the extracellular milieu or into intracellular organelles. On the other hand, Ca2+ entry into cells is mediated primarily by passive flow through voltage-, ligand-, temperature-, and mechanical stretch-gated ion channels.
The principal Ca2+ entryways of nerve, muscle, and some endocrine cells are voltage-gated Ca2+ channels (VGCCs). They were discovered in 1953 with the unexpected observation that crab muscle action potentials (APs) persist in the absence of external Na+, unlike squid nerve APs (145). Muscle APs were then found to increase with increasing extracellular Ca2+ concentration ([Ca2+]o), consistent with a Ca2+ conductance (200). Similar currents were later found in nerve, endocrine, and other tissues in diverse organisms (12, 221, 225, 253, 291, 304). Based on the membrane voltage required for activation, VGCCs were subsequently classified into high-voltage activated (HVA) and low-voltage activated (LVA) channels (65, 66, 146, 293). Later studies further classified Ca2+ currents into L-, N-, P/Q-, R-, and T-type currents, which exhibit distinct biophysical and pharmacological properties (127, 137, 292, 335, 359, 408, 444, 446, 500).
Molecular characterization of VGCCs began with the purification and cloning of the skeletal muscle Ca2+ channel (also called dihydropyridine receptor or DHPR) (107, 430, 434). The purified channel complex is composed of five subunits, termed α1 (175 kDa), α2 (143 kDa), β (54 kDa), δ (24–27 kDa), and γ (30 kDa). α2 and δ are linked posttranslationally by disulfide bonds into a single subunit referred to as α2δ (430). Subsequent research showed that L-, N-, P/Q- and R-type channels are made up of α1, α2δ, β, and, in some tissues, γ subunits (Fig. 1A). T-type channels, on the other hand, appear to require only an α1 subunit (351, 352).
The α1 subunit (Cavα1) is the principal component of VGCCs and is responsible for their unique biophysical and pharmacological properties. However, proper trafficking and functioning of L-, N-, P/Q- and R-type channels require the auxiliary subunits. In particular, the β subunit (Cavβ) plays a crucial role in trafficking the channels to the plasma membrane, fine-tuning channel gating, and regulating channel modulation by other proteins and signaling molecules. Crystal structures of the core region of three distinct Cavβs have opened up new avenues for investigating the molecular basis of Cavβ's actions. There is also emerging evidence that Cavβ may possess functions unrelated to VGCCs. This review focuses on the molecular biology, structure, function, and channelopathy of Cavβ, beginning with a brief overview of all VGCC subunits. Summaries of classical and recent work on VGCC electrophysiology, pharmacology, biochemistry, molecular biology, modulation, cell biology, and pathophysiology can be found in numerous excellent reviews (20, 22, 72–74, 108, 126, 138, 189, 216, 220, 234, 237, 246, 318, 351, 389, 423, 440, 445, 495).
A. The α1 Subunit
Cavα1 is the principal subunit of VGCCs. It is a 190- to 250-kDa protein containing four homologous repeats (I–IV) connected through cytoplasmic loops (Fig. 1B). Each repeat has six predicted transmembrane segments (S1–S6) and a reentrant pore-forming loop (P-loop) between S5 and S6. The four P-loops form the ion-selectivity filter, where four highly conserved negatively charged amino acids (glutamate or aspartate), one from each P-loop, form a signature locus that is essential for selecting and conducting Ca2+ (256, 266, 389, 482). Similar to K+ channels (128, 243, 290), the S6 segments form the inner pore (505), and the S4 segments' positively charged amino acids form part of the voltage sensor. The voltage-dependent movement of this sensor results in channel opening and closing. Furthermore, the majority of drug and toxin binding sites are located on Cavα1 (72). Thus Cavα1 possesses all the key features that define a VGCC, including pharmacological and biophysical properties such as gating, ion selectivity, and permeation.
Mammalian Cavα1 are encoded by 10 distinct genes. Based on amino acid sequence similarity, Cavα1 are divided into three subfamilies: Cav1, Cav2, and Cav3 (reviewed in Refs. 10, 72, 141, 486). The Cav1 subfamily includes channels that conduct L-type Ca2+ currents; the Cav2 subfamily includes channels that conduct N-, P/Q-, and R-type Ca2+ currents; and the Cav3 subfamily includes channels that conduct T-type Ca2+ currents (Fig. 1C).
B. The α2δ Subunit
The Cav1 and Cav2 subfamilies contain an auxiliary α2δ subunit (reviewed in Ref. 112). To date, there are four known α2δ subunits, each encoded by a unique gene and all possessing splice variants (Fig. 1D). Each α2δ protein is encoded by a single messenger RNA and is posttranslationally cleaved and then linked by disulfide bonds (259, 367). The δ peptide, originally presumed to be transmembrane but recently shown to be attached to the membrane through a glycosylphosphatidylinositol linker (113), anchors the larger extracellular α2 peptide in place (Fig. 1A). α2δ subunits can modify channel biophysical properties (63, 406, 459), but their main role is to increase Ca2+ channel current (63, 111, 174, 259, 260, 322, 406, 459) by promoting trafficking of Cavα1 to the plasma membrane and/or by increasing its retention there (32, 64, 194, 385). More recently, it was reported that α2δ functioned as a thrombospondin receptor to regulate excitatory synpatogenesis, independently from its regulation of VGCC activity (140, 267).
In two different mouse strains, naturally occurring mutations that lead to the loss of the full-length α2δ2 protein cause the ducky phenotype. This is characterized by shortened life spans, absence epilepsy, spike wave seizures, cerebellar ataxia, and decreased Purkinje cell dendritic arborization and firing rates (112, 260). α2δ2 knockouts also have abnormalities in the cardiovascular, immune, respiratory, and nervous systems. Irregularities in the cardiovascular system are also found in α2δ1 knockouts (169). α2δ3-null Drosophila are not viable, and the mutants have significantly impaired synaptic transmission (123, 267). Upregulation of α2δ1, on the other hand, is associated with neuropathic pain (283, 284). Importantly, α2δ1 is the main target of the antiepilepsy and antineuropathic pain drugs gabapentin and pregabalin, respectively (150, 169, 254).
C. The γ Subunit
There are eight different γ subunit genes, all yielding proteins with four transmembrane segments and intracellular amino (NH2) and carboxy (COOH) termini. γ1 was the first cloned γ subunit (182, 238, 430) and was copurified with muscle VGCCs, consistent with its primary role as a VGCC subunit. γ1 knockout mice are viable, morphologically indistinguishable from wild-type (WT) mice, but have larger Ca2+ currents with altered inactivation kinetics (168). γ2, γ3, and γ4 also associate with VGCCs (250, 399). γ1–4 subunits have been shown to produce varying effects on VGCC activity, depending on the partnered Cavα1 and Cavβ (134, 168, 207, 250, 258, 277, 379, 406, 467). The most consistent effect is a small reduction of current, caused mainly by a hyperpolarizing shift of the voltage dependence of inactivation and/or a positive shift of the voltage dependence of activation. However, unlike α2δ subunits, whose primary role is regulating VGCCs, γ subunits have more diverse biological functions. Since the discovery that mutations in γ2 underlie the stargazer mouse phenotype (277), which includes absence epilepsy and defects in the cerebellum and inner ear, it has become clear that γ2 and three other closely related γ subunits (γ3, γ4, and γ8) regulate the trafficking, localization, and biophysical properties of α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors (41, 82, 249, 343, 442). They are therefore referred to as transmembrane AMPA receptor regulatory proteins (TARPs). Indeed, acting as TARPs seems to be the primary role of γ2, γ3, γ4 ,γ8, and probably γ7 (252). While the function of γ5 remains unknown, γ6 is suggested to inhibit Cav3.1 channels (288), and γ7 is involved in the turnover of the mRNA of Cav2.2 and other proteins (149, 323). For recent reviews on γ subunits, see References 41, 82, 249, 320, 343, 350, 382.
D. The β Subunit
Purified Cav1 and Cav2 channels contain a tightly bound cytosolic Cavβ protein. There are four subfamilies of Cavβs (β1-β4), each with splice variants, encoded by four distinct genes. All four Cavβs can dramatically enhance Ca2+ channel currents when they are coexpressed in heterologous expression systems along with a Cav1 or Cav2 α1 subunit (268, 319, 322, 361, 405, 450, 467, 470). Cavβs also change the voltage dependence and kinetics of activation and inactivation (247, 268, 322, 332, 406, 412, 418, 450, 495); however, they do not affect ion permeation (183, 405, 458; but see Ref. 390). Furthermore, Cavβ either regulates or is indispensable for the modulation of Cav1 and Cav2 channels by protein kinases, G proteins, and small RGK (Rem, Rem2, Rad, Gem/Kir) proteins. Not surprisingly, Cavβ knockouts are either nonviable (in the case of β1 and β2) or result in a severe pathophysiology (in the case of β3 and β4).
The rest of this review is devoted to Cavβ.
II. CLONING OF CaVβ
Molecular studies on Cavβ can be traced back to the first purification and identification of the components of the skeletal muscle DHPR (107). With the use of a combination of chromatography, sucrose gradient sedimentation, and labeling with a high-affinity DHPR-specific ligand, three noncovalently attached subunits were purified: the largest 160-kDa subunit was named α, a 53-kDa subunit was named β, and a 32-kDa subunit was named γ (107). Subsequent purification studies of skeletal and neuronal Ca2+ channel complexes showed the presence of similar protein bands (4, 59, 114, 162, 278, 322, 381, 430, 434) and established that the DHPR actually consisted of five subunits, including α1 (175 kDa), α2 (143 kDa), β (54 kDa), δ (24–27 kDa), and γ (30 kDa) (430, 434).
Cloning of the first Cavβ was accomplished by Ruth et al. (381) using a classical approach based on peptide sequences derived from a purified skeletal muscle β subunit. This β subunit is now referred to as β1a. This cloning paved the way for the identification of other β subunits, their genes, and splice variants. Using a labeled skeletal muscle β1a cDNA, Pragnell et al. (362) screened a rat brain cDNA library and cloned a new β subunit, which later turned out to be a splice variant of β1 named β1b (360) (see sect. iv). Perez-Reyes et al. (353) also screened a rat brain cDNA library with β1a and, using low-stringency hybridization, uncovered another new β subunit, which was encoded by a different gene and named β2 (now named β2a). Screening a cardiac cDNA library, Hullin et al. (230) found β2a and two other β2 splice variants (β2b and β2c); in addition, they isolated the cDNA for β3. Meanwhile, using degenerate primers corresponding to the conserved domains of β1 and β2 to perform reverse-transcription PCR, Castellano et al. (67, 68) cloned β3 and β4 from a rat brain cDNA library.
The cloning of Cavβs subsequently led to the mapping of the four Cavβ genes (55, 94, 143, 347, 438) to chromosomes 17, 10, 12, and 2 for β1, β2, β3, and β4, respectively, and to the discovery of many other splice variants (see sect. iv).
III. STRUCTURE OF CaVβ
Prior to the determination of the crystal structure of Cavβ, it was already well recognized, based on amino acid sequence alignment, biochemical and functional studies, and molecular modeling, that Cavβ has a modular structure consisting of five distinct regions (40, 93, 119, 203, 342, 361). The first, third, and fifth regions are variable in length and amino acid sequence, whereas the second and fourth regions are highly conserved and are homologous to the Src homology 3 (SH3) and guanylate kinase (GK) domains, respectively. The SH3 domain is a common protein interaction module present in diverse groups of proteins (reviewed in Ref. 307). The GK domain, originally found in guanylate kinase from baker's yeast (416), is also engaged in protein-protein interactions (136, 170, 431). The middle three regions of Cavβ constitute the so-called Cavβ core, which is able to reconstitute many key functions of Cavβ (83, 84, 119, 176, 206, 313, 342, 502). In addition, early studies determined that Cavβ binds with high affinity to Cavα1. This high-affinity site is located in the cytoplasmic loop connecting the first two homologous repeats (i.e., the I–II loop) of Cavα1 and was named the α-interaction domain or AID (121, 361, 472) (Fig. 1E).
In 2004, three groups simultaneously and independently reported the crystal structure of the core of β2a, β3 and β4, either alone or in complex with the AID (84, 341, 447). The structures show that the Cavβ core indeed contains an SH3 domain and a GK domain, which are connected by a so-called HOOK region (Fig. 2A).
The existence of an SH3-HOOK-GK module places Cavβ in a family of proteins called the membrane-associated guanylate kinases (MAGUKs). MAGUKs, which include proteins such as PSD95, SAP97, CASK, Shank, and Homer, function as scaffold molecules that play a key role in organizing multiprotein complexes at functionally specialized regions such as synapses and other cellular junctions (136, 170, 431). MAGUKs contain an SH3-HOOK-GK module; in addition, they also contain one or more PDZ domains in the NH2 terminus, which serve protein-protein interaction and oligomerization functions. Cavβ is only partially related to MAGUKs structurally, however, because it does not contain a well-defined PDZ domain. Not surprisingly, the functions of Cavβ are markedly different from those of MAGUKs.
A. The GK Domain
Guanylate kinases are members of the nucleotide monophosphate kinase family that exists in organisms ranging from bacteria to humans. They catalyze the reversible phosphoryl transfer from ATP to GMP to produce ADP and GDP. Crystal structures of yeast guanylate kinases show that these enzymes have a compact structure with well-defined domains and folds and a catalytic site harboring the GMP- and ATP-binding pockets (43, 415, 416). The general structural features of yeast guanylate kinases are preserved in the Cavβ GK domain (Cavβ-GK), but large structural variations exist in the catalytic site, and many key catalytic residues are absent in Cavβ-GK (84, 341, 447). Thus Cavβ-GK is catalytically inactive. Similarly, the GK domain of MAGUKs does not possess catalytic activity, as indicated by the structural changes in the catalytic site and the lack of critical catalytic residues (285, 312, 437). Instead, the GK domains in these proteins have evolved into a protein interaction module. The Cavβ structures show that Cavβ-GK binds tightly to the AID in Cavα1 (84, 341, 447) (Fig. 2A), an interaction that will be further discussed in detail.
B. The SH3 Domain and the HOOK Region
Classical SH3 domains have a well-conserved and compact fold consisting of five sequential β-strands (βstrand 1–5) assembled into two orthogonally packed sheets (271). They mediate specific protein-protein interactions by binding to PxxP-containing motifs in target proteins, through a surface formed by a cluster of highly conserved hydrophobic residues. The Cavβ SH3 domain (Cavβ-SH3) has a similar fold as canonical SH3 domains do, but its last two β sheets are noncontinuous, separated by the HOOK region (84, 341, 447) (Fig. 2A). This split configuration is also shared by the SH3 domain of PSD-95, a MAGUK (312, 437). Cavβ-SH3 contains a well-preserved PxxP motif-binding site and therefore has the potential to bind PxxP motif-containing proteins. However, in the crystal structures, this binding site is partly shielded by the HOOK region and a long loop connecting two of the four continuous β sheets. Thus access to this site requires movement of these two regions. Such conformational changes are conceivable when Cavβ is bound to full-length Cavα1 and/or when it interacts with other partners, but are yet to be demonstrated. In contrast, the PxxP motif-binding site of the SH3 domain of PSD-95 is unobstructed (312, 437), consistent with the observation that the SH3 domain of MAGUKs can associate directly with PxxP motif-containing proteins (177, 306).
The HOOK region is variable in length and amino acid sequence among the Cavβ subfamilies (Fig. 3). In the crystal structures, a large portion of the HOOK is unresolved due to poor electron density, indicating that it has a high degree of flexibility (84, 341, 447). As will be discussed below, the HOOK region plays an important role in regulating channel inactivation.
C. The NH2 Terminus
The NH2 and COOH termini of Cavβ (abbreviated as Cavβ-NT and Cavβ-CT) are highly variable in length and amino acid composition (Fig. 3). There is yet no structure available for Cavβ-CT. However, an NMR structure of the NH2 terminus of β4 was solved recently, revealing a fold consisting of two α-helices and two antiparallel β sheets (451). This structure also shows that, unlike previously thought (203), Cavβ-NT does not have a PDZ fold, which consists of five β sheets (380). Incidentally, one of the two α-helices in the NMR structure is equivalent to the very first α-helix in the Cavβ core structures. Superposition of this helix in the two structures reveals that the NH2 terminus is oriented away from the core (Fig. 4).
D. The SH3-GK Intramolecular Interaction
The crystal structures of the Cavβ core show that the SH3 and GK domains interact intramolecularly (84, 341, 447). The affinity of this interaction is unknown, but the interaction is strong enough such that hemi-Cavβ fragments containing the NT-SH3βstrand 1–4-HOOK module and the SH3βstrand 5-GK-CT module can associate biochemically in vitro and reconstitute the functionality of full-length Cavβs when they are coexpressed in cells (298, 313, 342, 431, 432, 447). In fact, one of the β2a structures was obtained from cocrystals of two β2a hemifragments truncated at the HOOK region (447).
The last β sheet of Cavβ-SH3 (SH3βstrand 5), which is separated from the rest of the SH3 domain by the HOOK region, is critical for the strong intramolecular SH3-GK interaction (83, 298, 313, 342, 431, 432). This β sheet is directly connected to the GK domain, and it interacts extensively with both the GK domain and the rest of the SH3 domain (83). As a result, SH3βstrand5 glues the NT-SH3βstrand1–4-HOOK module and the SH3βstrand5-GK-CT module together and strengthens the otherwise weak interactions at the SH3-GK interface (83).
As in MAGUKs, the SH3-GK intramolecular interaction is important for the function of Cavβ (83, 298, 313, 431, 432). Weakening this interaction by mutating the SH3-GK interface or by inserting flexible linkers between the SH3 domain and the GK domain severely compromises the gating effects of Cavβ (83). Thus mutations, modifications, or protein-protein interactions that alter the SH3-GK intramolecular interaction may produce significant functional consequences.
E. The AID-Cavβ Interaction
Which regions anchor Cavβ to Cavα1? By screening an epitope library of 20,000 Cav1.1 fragments, Pragnell et al. (361) identified a region in the I–II loop that binds β1b. This region, known as the AID, is comprised of 18 residues, with a conserved consensus motif (QQxExxLxGYxxWIxxxE) in all Cav1 and Cav2 α1 subunits (Fig. 1E). The AID binds to all four Cavβs (121). The affinity of the AID-Cavβ interaction ranges from 2 to 54 nM, depending on the AID/Cavβ or Cavα1/Cavβ pair and the method of affinity measurement (30, 56, 62, 120, 121, 179, 342, 371, 395, 448). Single mutations of several conserved residues in the AID, including Y10, W13, and I14, greatly weaken the AID-Cavβ interaction, as indicated by in vitro binding experiments and by the reduction or abolishment of Cavβ-induced stimulation of Ca2+ channel current in heterologous expression systems (33, 34, 56, 120, 181, 185, 206, 218, 276, 361, 448). Thus the role of the AID as the principal interacting domain with Cavβ is firmly established.
Which region(s) of Cavβ interact with the AID? In an influential study, De Waard et al. (119) described a 31-amino acid segment of Cavβ, referred to as the β-interacting domain or BID, as the main binding site for the AID. The BID was able to slightly enhance Ca2+ channel current and modulate gating (119), and several BID point mutations were able to weaken the Cavβ/Cavα1 interaction and reduce BID-stimulated Ca2+ channel currents (119, 120).
For the next decade, it had been generally accepted that Cavβ interacted with Cavα1 primarily through the BID. Surprisingly, however, the crystal structures of two different AID-Cavβ core complexes reveal that the AID does not bind the BID (84, 341, 447). Indeed, the AID and the BID do not come into direct contact (Fig. 2B). Instead, the AID binds to a hydrophobic groove in the GK domain termed the AID-binding pocket (ABP; Fig. 2C) (84, 447, 448). The AID occupies only a tiny fraction of the Cavβ surface area, raising the possibility that other domains of Cavβ are involved in interactions with other regions of Cavα1 or with other proteins. As will be discussed later, both are indeed the case.
The AID-GK domain interactions are extensive and predominantly hydrophobic (Fig. 2C). These interactions account for the 2–54 nM affinity of the AID-Cavβ binding. Functional studies show that mutating two or more key residues in the ABP severely weakens or completely abolishes the AID-Cavβ interaction (206, 502).
The binding of the AID with Cavβ does not significantly change the Cavβ structure, except for some small and localized changes near the ABP. Importantly, however, the AID undergoes a dramatic change in secondary structure when it is engulfed by the ABP. When alone, the AID exists as a random coil in solution, as determined by circular dichroism spectrum measurements (341). When bound to Cavβ, the AID forms a continuous α-helix, as shown in the crystal structures. Together with the observation that the 22-amino acid linker between the AID and the first S6 segment of Cavα1 (i.e., IS6) also forms an α-helix (9), a picture emerges that the entire region encompassing IS6 and the AID adopts a continuous α-helical structure in the presence of Cavβ (Fig. 4). This structural hallmark is crucial for the regulation of Ca2+ channel gating by Cavβ, as will be discussed later.
Since the publication of the Cavβ structures, some investigators have been continuing to perform or interpret experiments based on the notion that the BID interacts with Cavα1 (85, 281, 302, 388, 441, 506), so before leaving this section, we briefly revisit the BID. The crystal structures show that the BID spans three different regions of Cavβ (SH3, HOOK and GK) and that most of it is completely buried (Fig. 2B). Thus the BID does not directly interact with Cavα1; rather, it is crucial for maintaining the SH3-GK intramolecular interaction and the structural integrity of Cavβ. Of the four residues in the BID whose mutations weakened the Cavβ/Cavα1 interaction, three were proline and one was tyrosine (119, 120). Mutating these residues most likely alters the folding and/or structure of Cavβ, which explains its inability to bind Cavα1.
But how could the BID enhance Ca2+ channel current (119)? While the mechanism of this action remains unclear, it reminds us of an experiment of our own in which a random 43-amino acid peptide (which has no sequence similarity in GenBank) was coexpressed with Cav2.1 and α2δ in Xenopus oocytes. This random peptide significantly increased Ca2+ channel currents (compared with no Cavβ), to ∼50% of β3-induced current. This obvious nonspecific effect, reported in 2004 (84), suggests that the BID-induced current increase may also be a nonspecific effect. Given these structural and functional information, it is prudent to exercise caution when interpreting experimental data concerning the BID.
IV. CaVβ SPLICE VARIANTS AND THEIR TISSUE DISTRIBUTION
Mammalian Cavβs are encoded by four distinct genes, Cacnb 1-4. They all have 14 exons except Cacnb3, which has 13, and each Cavβ has 2 or more splice variants. Figures 5 and 6 show most of the human Cavβ splice variants found thus far. The five distinct domains and regions of Cavβ are mapped onto their corresponding exons and protein sequences. Alternative splicing occurs in those exons that encode the variable domains or regions, namely, the NH2 and COOH termini and the HOOK region. The four different Cacnb genes utilize different alternative splicing sites. Cacnb1 and Cacnb2, which produce β1 and β2, respectively, exhibit alternative splicing in exon 7, giving rise to divergent HOOK regions. Cacnb1 is also alternatively spliced in the COOH terminus, with exon 14 either included or excluded. On the other hand, Cacnb2 is alternatively spliced extensively in the NH2 terminus, yielding highly diversified NH2 termini. Cacnb4 has no alternative splicing in the HOOK region but has NH2- and possibly COOH-terminal alternative splicing. Cacnb3 has no alternatively spliced exons, but like all other Cacnb genes, produces a truncated isoform.
Table 1 shows the tissue distribution of some Cavβ splice variants. As expected, Cavβs are abundantly expressed in excitable tissues such as the brain, heart, and muscles. While some splice variants (e.g., β1b and β2b) are widely expressed, others (e.g., β1a, β2d, and β2e) have a more restricted expression. The expression of some splice variants is developmentally regulated. For example, β1b and β4 expression increases with development, whereas β2c, β2d, and β2e expression decreases with development. It is important to note that, in most cases, protein but not mRNA expression was listed. Immunolocalization experiments should be interpreted cautiously since an antibody may recognize several splice variants, for example, an antibody against the β2 COOH terminus will recognize all β2 splice variants.
Since the association between Cavβ and Cavα1 is promiscuous (i.e., any full-length Cavβ can associate with any Cav1 or Cav2 α1 subunit), alternative splicing greatly increases the molecular diversity and functionality of HVA Ca2+ channels. Furthermore, some splice variants may take on functions other than regulating HVA Ca2+ channels (see sect. xii). Thus a major future challenge (and a fruitful area of research) is to determine how alternative splicing is regulated in various tissues and at different developmental stages.
V. CaVβ REGULATES THE SURFACE EXPRESSION OF HIGH-VOLTAGE ACTIVATED Ca2+ CHANNELS
The α1 subunit of Cav1 and Cav2 channels cannot reach the membrane by itself; it shows no surface expression and produces very small or no currents when expressed without auxiliary subunits. Coexpression of Cavβ with Cavα1 increases currents by orders of magnitude, depending on factors such as the expression system, DNA or RNA concentration, VGCC turnover rate, inhibitory factors present, the α1/β combination, etc. (reviewed in Refs. 10, 40, 125, 237, 495). The current increase reflects enhanced channel expression on the plasma membrane and also an increase in channel open probability. In this section we discuss the evidence and mechanisms of increased channel surface expression.
A. Cavβ Is Required for Normal Channel Expression
It has been well established that Cavβ can function as a chaperone to dramatically increase the surface expression of Cav1 and Cav2 channels. This is observed in various heterologous expression systems with all four subfamilies of Cavβ and all Cav1 and Cav2 subunits (14, 50, 88, 93, 191, 245, 247, 248, 276, 322, 353, 458, 481, 495). The increased surface expression can be detected by Cavα1 epitope tag staining, surface biotinylation, gating charge measurements, or increased Ca2+ channel current. An important point to mention is that Xenopus oocytes, a widely used expression system for studies of VGCCs, have two endogenous β subunits that share 98% homology with β3 (436). These endogenous subunits are expressed at sufficient levels to transport a small number of exogenously expressed Cavα1 to the plasma membrane and hence lead to small Ca2+ channel currents in the absence of an exogenous Cavβ. Antisense oligonucleotides against endogenous β3 are able to suppress these currents (62, 436). Little or no endogenous Cavβ was detected in widely used mammalian cell lines such as HEK 293 cells, COS cells, and CHO cells (276, 315). Nevertheless, expression of Cavα1 alone in these cells can produce measureable, albeit miniscule, Ca2+ channel currents (245, 247, 248, 303, 407, 418, 436), suggesting that either a very small fraction of Cavα1 can be trafficked to the plasma membrane in the absence of Cavβ or these cells contain low levels of endogenous Cavβs.
Cavβ also enhances Ca2+ channel surface expression in vivo. For example, β1 and β2 knockout mice have severely reduced Ca2+ currents in muscle and heart (see sect. xiii). Knockdown of Cavβ also decreases endogenous Ca2+ currents in neuronal cells (35, 279). Conversely, overexpression of Cavβ using adenoviruses increases Ca2+ channel current density in native cardiac cells, suggesting that Ca2+ channel surface expression may be limited by the availability of Cavβ (336, 465).
Binding of Cavβ to the AID of Cav1 and Cav2 is essential for its chaperone effect. Point mutations in the AID that weaken or abolish the AID-Cavβ interaction severely reduce or abolish Cavβ-stimulated Ca2+ channel current (33, 34, 56, 120, 181, 185, 206, 218, 276, 336, 361, 448). Deleting the AID altogether, not surprisingly, abolishes Cavβ-induced current enhancement (185, 298). Likewise, mutations in the ABP that weaken or abolish the AID-Cavβ interaction also reduce or abolish Cavβ-stimulated Ca2+ channel expression and current (206, 502). Recent studies show that the GK domain itself can largely recapitulate the chaperone function of full-length Cavβs, greatly increasing Ca2+ channel surface expression and current in Xenopus oocytes and mammalian cells (129, 206).
How does Cavβ enhance Ca2+ channel surface expression? One hypothesis is that Cavβ shields or disrupts one or more ER retention signals on the I–II loop of Cavα1 (39), and several lines of evidence support this hypothesis. The I–II loop of Cav1.2 and Cav2 can trap α1 subunits in the ER (except Cav1.1), but the I–II loop of Cav3.1 (a T channel) fails to do so. Also, tagging a Shaker K+ channel with the I–II loop of Cav1.2 or Cav2.1 decreases its expression by approximately sevenfold, while coexpression of Cavβ prevents this downregulation (39). Moreover, deleting the I–II loop from Cav1.2 (Δ389–423) increases its surface expression in the absence of Cavβ (39).
However, some results are inconsistent with this hypothesis. 1) The I–II loop of Cav1.1 does not cause ER retention of a CD8 peptide (99). 2) CD4 fusion constructs of the I–II loop of Cav1.2 and Cav2.2 are trafficked efficiently to the plasma membrane, rather than being retained in the ER (5). 3) Transplanting the I–II loop of Cav2.2 into Cav3.1 causes Cavβ-independent current upregulation instead of downregulation (9).
An alternative possibility is that additional trafficking signals exist in the NH2 and COOH termini of Cavα1 (99, 163, 175, 262, 466). However, the NH2 and COOH termini of Cav1 and Cav2 are not conserved, and yet, the chaperone function of Cavβ is universal, suggesting that any ER retention signals in the NH2 and COOH termini may only be modulatory.
Recently, a new study suggested that Cavβ increases Cavα1 expression on the plasma membrane by preventing its ubiquitination and proteasomal degradation (5). Thus Cavβ may simply be required to help Cavα1 escape the degradation pathway.
B. Membrane Association and Subcellular Targeting of Cavβ
Cavβs are expected to have a cytosolic localization based on analyses of their amino acid sequence (353, 381). This is true for the majority of Cavβ splice variants when they are expressed alone, without a Cavα1 (with a few exceptions discussed below; Refs. 176, 181). However, some Cavβs, most notably β2a, can be localized to the plasma membrane on their own. β2a is linked to the plasma membrane through palmitoyl groups that are covalently attached to two cysteines (Cys 3,4) in the NH2 terminus (86, 87). When palmitoylation is abolished, in a double Cys→Ser mutant, membrane localization disappears (87). Importantly, β2a palmitoylation can be dynamically regulated in vivo, adding a layer of physiological control (232, 464). However, palmitoylation alone may not be sufficient for membrane localization because implanting the β2a NH2 terminus into other Cavβs does not yield membrane localization (87, but see Ref. 369). Thus β2a probably possesses additional determinants that help target it to the plasma membrane. Another β2 subunit, β2e, is not palmitoylated but is found at the plasma membrane (433). The underlying mechanism is yet unknown. Finally, β1b is localized to the plasma membrane in COS-7 cells (44, 50), but this is not observed in tsA201 cells (87) or primary cardiomyocytes (93). The reason for the discrepancy is unclear, but in COS-7 cells, the membrane localization is attributed to a COOH-terminal acidic motif (WEEEEDYEEE) whose deletion diminishes membrane localization. When this motif is fused to β3, which is normally cytosolic, it migrates to the plasma membrane (44, 50). As will be discussed in section vi, membrane localization of Cavβ coincides with many functional effects, especially slowed inactivation.
In the presence of Cavα1, all Cavβs localize to the plasma membrane through their association with Cavα1; however, they may be targeted to different subcellular locations depending on which Cavα1 they associate with. For example, β3 and β4, which predominantly associate with presynaptic Cav2 channels, can be found in axons, whereas β1 is scarce in this compartment (336); instead, β1 is found in postsynaptic compartments (soma and dendrites). In skeletal muscle, β1a is targeted to the triads through its association with Cav1.1 (333). When exogenously expressed in epithelial cells, β1b is localized on the apical membrane with Cav2.1 but on the basolateral membrane with Cav1.2 (44). Conversely, Cavβ may affect the subcellular localization of Cavα1. For example, β1a helps arrange L-type Ca2+ channels as tetrads in the t tubules of skeletal muscles (see sect. xiA; Refs. 164, 191, 394, 507), and β4 is implicated in the synaptic localization of P/Q-type channels in cultured hippocampal neurons (474). Furthermore, through interactions with different proteins, Cavβ helps attach Ca2+ channels to synaptic vesicles (257), the cytoskeleton (223), or the surface of sarcoplasmic reticulum (333). These examples illustrate the role of Cavβ as a scaffold protein.
VI. CaVβ REGULATION OF Ca2+ CHANNEL GATING
Once the Ca2+ channel complex reaches the plasma membrane, Cavβ powerfully modulates its gating. The main features of gating modulation are the enhancement of voltage-dependent activation (VDA) and voltage-dependent inactivation (VDI). β2a is unique in that it inhibits VDI. This section describes these Cavβ effects and their mechanisms.
A. Cavβ Enhances Voltage-Dependent Activation
All Cavβs shift the voltage dependence of activation to more hyperpolarized voltages (by ∼10–15 mV, Table 2 and Fig. 7). This was shown for both Cav1 and Cav2 channels in various expression systems (61, 116, 229, 245, 268, 322, 406, 412, 414, 443, 495). The shift can also be observed in vivo in some knockout mice (191, 325, 468), while in some other cases, it is probably obscured by the compensatory effects of other Cavβ genes (31, 330, 331). In addition, the speed of activation is increased in general (268, 412), but it could appear slower depending on the stimulus voltage (353) and the particular α1/β pair (184, 245, 443, 459).
These effects are also visible at the single-channel level. Thus channels without a Cavβ tend to open less frequently, open for a shorter duration, and require more positive activation voltages. Cavβ coexpression greatly increases channel open probability (Po) and shortens the latency to first opening (93, 125, 215, 229, 295, 457). Notably, β2a produces the most dramatic increase in Po (76, 93, 132).
Normal VDA is largely reconstituted by the core region of Cavβ (206). Deleting the entire Cavβ COOH terminus has no effect on VDA, at least for Cav2.1 channels expressed in Xenopus oocytes (206). The NH2 terminus, however, appears to have a small role in modulating VDA. For example, β4b, which has a longer NH2 terminus compared with β4a, induces a larger hyperpolarizing shift in the activation of some Cavα1 (208).
B. Cavβ Promotes Voltage-Dependent Inactivation, Except β2a
VDI reduces the amount of Ca2+ entering the cell following depolarization and decreases the number of channels responsive to subsequent depolarizations. Cavβ is a key modulator of VDI, as first demonstrated in 1991 (268, 406, 450) and subsequently confirmed for various α1/β combinations in different expression systems (116, 137, 245, 340, 348, 414, 417, 458). Several aspects of VDI are affected by Cavβ. 1) β1, most β2 splice variants, β3, and β4 shift the voltage dependence of inactivation to more hyperpolarized voltages (by ∼10–20 mV; Table 2 and Fig. 7), whereby weaker depolarizations are able to inactivate the channels. β2a, however, causes a shift to more depolarized voltages (by ∼10 mV) (40, 93, 119, 132, 206, 218, 245, 276, 314, 340, 369). 2) Cavβs (except β2a) promote the process of “closed state” inactivation exhibited by Cav2 channels when they rapidly transition between closed and open states, such as during a train of action potentials (β3 > β1b = β4 >> β2a; Refs. 348, 491). Similarly, a large hyperpolarization of steady-state inactivation (approximately −40 mV) is observed when β3 is overexpressed with N- and R-type channels, dramatically increasing the population of inactivated channels at resting conditions (491). 3) β1, most β2 splice variants, β3, and β4 speed up the inactivation kinetics, whereas β2a and β2e slow down inactivation (Table 2 and Fig. 7).
The unique effects of β2a on VDI are largely abolished when palmitoylation of β2a is disrupted by mutating its two NH2-terminal cysteine residues to serine (β2a C3,4S) (365, 369). WT β2a-like properties can be restored when a transmembrane segment of an unrelated membrane protein is fused to this mutant, suggesting that membrane anchorage rather than palmitoylation per se is critical for β2a's unique functions (369). Supporting this idea, the nonpalmitoylated but membrane-attached β2e has properties similar to β2a (433).
Multiple domains and regions of Cavβ are involved in the regulation of VDI. The GK domain alone, when expressed together with Cav2.1 and α2δ in Xenopus oocytes, has been shown to speed up VDI and hyperpolarize the voltage dependence of VDI (206). The GK domain of all four subfamilies of Cavβ produces the same effects (Fig. 7, D and E; Ref. 206), as expected from its high degree of amino acid conservation. Similarly, the GK domain of β2a greatly accelerates VDI and hyperpolarizes the voltage dependence of VDI of Cav2.2 channels expressed in oocytes and tsA-201 cells (129, 372). On the other hand, it has been reported that refolded and purified proteins of β2a and β1b GK domains slow down VDI and depolarize the voltage dependence of Cav2.3 channels expressed in oocytes (187). The discrepancy between these studies may result from the use of different Cavα1 or from RNA versus protein injection, but it should be noted that the refolded and purified GK domains appear to be dimerized proteins (187), and it is unknown whether and how dimerization changes the function of the GK domain.
The HOOK plays an important role in regulating VDI, as first suggested by chimeric studies between different Cavβs (364, 420). Two recent studies based on structurally defined Cavβ domains provide more definitive evidence. 1) Swapping the HOOK between the core regions (SH3-HOOK-GK) of β1b and β2a, which have opposite effects on VDI, also swaps their effects on VDI (206). 2) Deleting the HOOK in either β2a core or full-length β2a results in increased VDI (372). These studies, in conjunction with those discussed earlier, indicate that both membrane attachment through palmitoylation and a long HOOK region contribute to the unique effects of β2a on VDI.
The role of the NH2 terminus of Cavβ in regulating VDI has long been established. Deleting or shortening the NH2 terminus, or swapping the NH2 terminus of different Cavβs markedly alters VDI (236, 340, 364, 420). β2 or β4 splice variants differing in the NH2 terminus exhibit markedly different VDI (208, 209, 215, 433). As discussed above, the palmitoylation site of β2a is in the NH2 terminus.
Surprisingly, the COOH terminus of Cavβ seems to play a very limited or no role in regulating VDI, even though it is highly variable among the four Cavβ subfamilies. Thus, although a very small change in the inactivation kinetics of Cav2.1 channels is observed when the COOH terminus of β4 is deleted (460), exchanging the COOH terminus between β3 and β4 or deleting the entire COOH terminus of any of the four CaVβs has little effect on VDI of Cav2.2 channels or Cav2.1 channels (206, 420). It remains to be determined whether the COOH terminus exerts a more prominent effect on VDI under other conditions and for certain combinations of Cavα1 and CaVβ. Intriguingly, a β4 COOH-terminal truncation mutant missing the last 38 amino acids, which causes slightly faster inactivation of Cav2.1 channels at moderate depolarizations, was identified in a juvenile myoclonic epilepsy patient (142). Whether the very subtle change in Ca2+ channel inactivation underlies the disease is unclear.
C. A Unified Model for Cavβ Regulation of Ca2+ Channel Gating
How does Cavβ regulate VDA and VDI of Cav1 and Cav2 channels? Before addressing this question, we first briefly discuss the pore structure, the location of the activation gate, and the mechanism of VDI of VGCCs.
The external pore, including the ion selectivity filter, of VGCCs is formed by the pore loop between the S5 and S6 transmembrane segments of each of the four homologous repeats of Cavα1; point mutations in this region, especially of the four conserved glutamate or aspartate residues, drastically alter ion selectivity, permeation, and pore blockage (256, 266, 389, 482). The inner pore is formed by all four S6 segments of Cavα1, as demonstrated by the substituted cystine accessibility method (505). Cystine accessibility studies also indicate that the activation gate is located at the cytoplasmic end of the S6 segments (476). The S6 segments, together with the I–II loop and the NH2 and COOH termini of Cavα1, are involved in controlling or regulating VDI (for review, see Refs. 212, 422). Although the precise molecular mechanism of VDI is unknown, a prevalent model is that the I–II loop of Cavα1 functions as a “hinged lid” to physically occlude the pore by binding to the cytoplasmic ends of the S6 segments (421, 422), reminiscent of VDI of voltage-gated Na+ channels (71). Which amino acids form the inactivation gate and its receptor site remain unknown. An alternative model is that VDI is produced by a constriction of the pore (151). Either way, the S6 segments constitute a converging point through which both VDA and VDI are controlled and regulated.
The biochemical, functional, and structural studies presented above support a unified model for Cavβ regulation of VDA and VDI of VGCCs (9, 125, 151, 206, 298, 341, 448, 455, 461). This model has two central components.
First, the high-affinity AID-GK domain interaction and a rigid IS6-AID linker are essential for Cavβ regulation of VGCC gating. As mentioned in section iiiE, in the presence of Cavβ, through the AID-GK domain interaction, the entire region encompassing the IS6 segment and the end of the AID becomes a continuous α-helix (9, 151, 341). Via this rigid structure, Cavβ gains a lever with which to regulate both activation and inactivation (Fig. 4). Thus Cavβ binding adds mass and tension to IS6 and the I–II loop, which most likely affects the energetics of voltage-dependent movement of both IS6 and the inactivation gate, thereby directly changing the voltage dependence and kinetics of activation and inactivation. This explains why the GK domain alone is capable of affecting both activation and inactivation (129, 206, 372). Equally important, the AID-GK domain interaction anchors Cavβ to Cavα1, thereby enabling interactions between Cavβ and other parts of Cavα1 that are of intrinsic low affinity but are important for Cavβ's gating effects (see below). Supporting an essential role of the AID-GK domain interaction, many studies show that Cavβ regulation of gating is abolished by mutations in the AID (33, 34, 56, 120, 181, 185, 206, 218, 276, 361, 448) or in the ABP (206, 502). However, one difficulty in interpreting these and similar experiments is that those mutations dramatically reduce or abolish Cavβ-stimulated Ca2+ channel surface expression, leaving minuscule currents to be scrutinized. This problem is circumvented in several recent studies where the rigid α-helical structure of the IS6-AID linker was disrupted by substituting linker residues with glycines, or inserting multiple glycines in the linker, while leaving the AID-GK domain interaction intact. These substitutions or insertions do not affect Cavβ-enhanced Ca2+ channel surface expression, but they severely compromise or eliminate the ability of Cavβ to regulate Cav1 and Cav2 channel activation and inactivation (151, 455, 502). These results underscore the essential role of a rigid IS6-AID linker in Cavβ regulation of VGCC gating.
An additional factor that is important for Cavβ regulation of gating is the orientation of Cavβ relative to Cavα1 (455, 502). Inserting five alanine residues in the IS6-AID linker, which is expected to maintain the α-helical structure of the linker but induce a 180° rotation of Cavβ with respect to Cavα1, markedly diminishes Cavβ regulation of activation, while insertion of seven alanines, which produces two full turns, has no significant detrimental effect (502). Similarly, deleting one or three residues in the IS6-AID linker totally abolishes Cavβ regulation of both activation and inactivation (455). These studies are consistent with the notion that additional contacts between Cavβ and Cavα1 besides the AID-GK domain interaction are critical for Cavβ regulation of VGCC gating.
Second, intrinsically low-affinity interactions between Cavβ and Cavα1 are crucial for Cavβ regulation of VGCC gating (especially VDI), and these interactions confer each Cavβ its distinct modulatory effect and α1/β pair-specific gating properties. Besides the AID-GK domain interaction, other direct contacts between Cavβ and Cavα1 have been observed in vitro. For example, the Cavβ SH3 domain interacts with the I–II loop, but at a region different from the AID (298), and a COOH-terminal region conserved only in β2 binds to a COOH-terminal region of Cav1.2 where calmodulin (CaM) also binds (270). The same Cav1.2 COOH-terminal region also binds to a β2a construct containing the N-SH3βstrand1–4-HOOK module (501). Other regions of Cavα1, including the NH2 and COOH termini and the III-IV loop, have also been shown to interact directly with Cavβ (366, 436, 460, 461). It remains to be determined which regions of Cavβ they associate with, but the Cavβ NH2 terminus and HOOK are prime candidates since they are critically involved in regulating VDI. These additional α1/β interactions have intrinsic low affinity, and on their own, do not produce significant gating effects. However, the strength of these interactions increases dramatically when Cavβ is anchored to Cavα1 by the AID-GK domain interaction. These notions are supported by the aforementioned mutagenesis/insertion studies in the AID, the ABP, and the IS6-AID linker. Further supporting these ideas, Chen et al. (83) reported that, without changing the AID-GK domain interaction, splitting β2a into two connected modules (N-SH3-HOOK and GK-C) through the insertion of increasingly longer flexible linkers between the SH3 and GK domains leads to a gradual diminishment of the effect of the N-SH3-HOOK module on VDI (83). This result indicates that keeping the N-SH3-HOOK module near Cavα1 is essential for its modulatory effect. A future challenge is to develop ways to precisely map the interface of intrinsically low-affinity Cavα1/Cavβ interactions, which might be too weak to be identified biochemically and might require more than one Cavα1 region.
How low-affinity α1/β interactions regulate gating is unclear. These interactions could pull on Cavβ and thereby modulate the movement of IS6 and the presumed inactivation gate in the I–II loop. They may also interfere with intramolecular interactions between the I–II loop and other parts of CaVα1, such as the NH2 and COOH termini and the III-IV loop, where point mutations and deletions cause marked changes of VDI (for review, see Refs. 212, 422). These intramolecular interactions, as well as the low-affinity α1/β interactions, are α1 or α1/β pair specific (2, 99, 178, 262, 369, 386, 404, 436, 460, 501). Thus, to fully appreciate the physiological importance of Cavβ regulation of VGCC gating, it is crucial to examine the pairing of Cavα1 and Cavβ in different tissues and cell types, in different subcellular locations, and at different developmental stages.
A final point that should be mentioned here is that many proteins that interact directly with Cavβ have been shown to regulate VGCC gating, such as RGK proteins (see sect. ix), Best1 (493), and RIM1 (257) (see sect. xi).
D. Can Cavβ Produce AID-Independent Gating Effects?
Several reports, which at first seemed to contradict the model presented above, are in fact in accord with the model upon closer examination. It has been shown that β2a is able to modulate VDA and VDI of Ca2+ channels formed by a mutant Cav2.1 subunit (Cav2.1_ΔAID) whose AID is deleted (298). This result led the authors to conclude that essential Cavβ modulatory properties are AID independent. This result, however, has an alternative explanation: β2a can be anchored to the plasma membrane through palmitoylation, and this membrane tethering might mimic, at least partially, the anchoring role of the AID-GK domain interaction, bringing β2a near Cav2.1_ΔAID subunits and promoting the functionally important low-affinity α1/β interactions alluded to above. Indeed, this result lends strong support to the second part of the model discussed above, i.e., there are low-affinity interactions between Cavβ and Cavα1 that are crucial for Cavβ regulation of VGCC gating.
Several studies reported that a 41-amino acid β2 COOH-terminal fragment and some Cavβ splice variants, including β2f, β2Δg, β1d, and chicken β4c, all of which lack most or the entire GK domain (and hence cannot bind the AID), are all able to enhance Ca2+ channel currents and/or regulate their gating (92, 204, 270). However, these effects are much weaker than those produced by full-length Cavβs. Moreover, the specificity of these effects is called into question by the clear nonspecific effects of two short peptides that do not exist in nature: a 35-amino acid peptide containing the BID and a 43-amino acid peptide with a random sequence, both of which are able to stimulate Ca2+ channel expression and weakly modulate gating (84, 119, 120). Nevertheless, given that β2f and β2Δg are found in native cells (204, 270), they could affect Ca2+ channel gating through low-affinity α1/β interactions if they are expressed at very high levels. At present, the physiological role of β2f and β2Δg remains unknown.
E. Cavβ Regulation of Ca2+-Dependent Inactivation and Facilitation
HVA Ca2+ channels are strongly regulated by another type of inactivation that depends on Ca2+ influx, namely, Ca2+-dependent inactivation (CDI) (for reviews, see Refs. 53, 78, 201, 358), which serves as a negative-feedback mechanism. CDI is mediated by the ubiquitous Ca2+-sensing protein CaM, which is constitutively bound to the Cavα1 COOH terminus (494, 509). The exact molecular mechanism of CDI is unclear, as is the relationship between CDI and VDI, but a recent study shows that two of the elements critical for VDI, Cavβ and a rigid IS6-AID linker, are also essential for CDI (151). Glycine (but not alanine) substitutions that disrupt the α-helix of the IS6-AID linker dramatically slow CDI. The absence of Cavβ binding to Cavα1, ensured by mutating the AID, produces similar results (151). Thus CDI and VDI appear to share a common mechanism by which conformational changes caused by CaM-Cavα1 interactions or Cavβ-Cavα1 interactions are transmitted to the pore through the rigid IS6-AID linker.
HVA Ca2+ channels also undergo Ca2+-dependent facilitation (CDF), which occurs during repetitive channel activation, such as during a train of action potentials (for reviews, see Refs. 53, 201, 358). This process, which is dependent on CaM binding to the Cavα1 COOH terminus, also requires Cavβ binding to the AID and an intact IS6-AID α-helix (151). Interestingly, CDF is readily observed with β2a but not with β1b or β4 (80, 272). The main reason for this difference is probably that channels with β2a inactivate much slower; slow inactivation not only allows the unmasking of CDF but also further stimulates CDF by permitting a larger Ca2+ influx.
F. CaVβ Regulation of Voltage-Dependent Facilitation
L-type Ca2+ channels exhibit voltage-dependent facilitation (VDF) (47). VDF is manifested as a gradual increase in L-type current during high-frequency action potentials, and it partly explains activity-dependent enhancement of L-type currents in skeletal muscle, brain, and heart. VDF can be differentiated from CDF by using Ba2+ as the charge carrier; it is accompanied by an increase in high Po gating (357) and may be dependent on phosphorylation (273). Like CDF, VDF depends on the presence of Cavβ (47, 75; but see Ref. 274); it is supported by β1 and β3 but not β2a (47, 75, 365; but see Ref. 109). Some of the discrepancies in the literature may result from the following reasons. 1) Differences in L-type channel splice variants and the α2δ subunits used affect the results. For example, α2δ1 and α2δ3 seem to mask VDF by increasing inactivation (109). 2) The β2a-containing channels already have a high Po, so VDF is harder to observe in these channels. 3) Nonpalmitoylated β2a mutants can restore VDF (365), suggesting that different levels of palmitoylation may contribute some variations in the results.
The GK domain alone appears to be necessary and sufficient to confer VDF; deleting other domains, including the SH3 domain, separately or in combination, spares VDF (77). Hence, it is possible that VDF, just like CDF, CDI, VDI, and VDA, relies on the rigid IS6-AID linker and Cavβ to affect gating. It would be of interest to investigate whether glycine substitution or insertion in the IS6-AID linker also affects VDF.
VII. STOICHIOMETRY AND REVERSIBILITY OF THE CaVα1-CaVβ INTERACTION
How many β subunits need to bind to each Cavα1 to bring about the aforementioned trafficking and gating effects? Is the Cavα1-Cavβ interaction reversible? This section discusses these two important issues.
A. Cavα1 and Cavβ Are Paired With a 1:1 Stoichiometry
Early biochemical studies suggest that skeletal and neuronal VGCCs contain a single Cavα1 and a single Cavβ (430, 473). This remains the prevalent view today, but it comes after a brief competition with the idea of a 1 Cavα1:2 or more Cavβ stoichiometry (62, 436).
As mentioned in section vA, Xenopus oocytes express two endogenous β3-like subunits, called β3xo (436). When Cavα1 cRNA is injected into Xenopus oocytes alone, a small fraction of the Cavα1 is transported to the plasma membrane by β3xo (436). Coinjection of a mammalian Cavβ or either of the two Xenopus β subunits greatly increases Ca2+ channel current and changes its gating properties. These results led to the proposal that the “Cavα1-alone” channels in fact contained a β3xo and that one or more exogenous Cavβ bind the Cavα1/β3xo complex to form a higher order complex with modulated gating (436). Subsequently, by varying the concentration of coexpressed β3, it was found that β3 produced the trafficking effect with a sevenfold higher apparent affinity than it did gating modulation (17 vs. 120 nM) (62). This result was initially explained by one of two hypotheses: either two Cavβs bind a single Cavα1 or the mature Cavα1 on the plasma membrane has a lower affinity for Cavβ than the nascent Cavα1 does (62).
Subsequent extensive studies indicate that Cavβ associates with Cavα1 in a 1:1 stoichiometry and that this stoichiometry is determined by the AID-GK domain interaction. 1) Channels coexpressed with a mixture of β2a and β3 form two biophysically distinct channel populations, rather than a single population of “mixed”-channel type (245). 2) Colecraft and colleagues (110) covalently linked a single β2b to the COOH terminus of Cav1.2 (creating Cav1.2-β2b) and found that the channels formed by Cav1.2-β2b exhibited the same gating properties as channels formed by the coexpression of Cav1.2 and β2b did. Moreover, coexpression of β2a and Cav1.2-β2b did not further change channel gating. 3) The crystal structures of the AID-Cavβ core complexes clearly show that each Cavβ binds a single AID (84, 341, 447). 4) Mutations of key residues in the AID or the ABP abolish both Cavβ-mediated Ca2+ channel surface expression and gating modulation (33, 34, 56, 120, 181, 185, 206, 218, 276, 361, 448, 502).
B. The Cavα1-Cavβ Interaction Is Reversible
The affinity of the AID-Cavβ interaction measured in vitro is very high, with a Kd ranging from 2 to 54 nM (30, 56, 62, 120, 121, 179, 342, 371, 395, 448). The affinity of Cavα1-Cavβ interactions in cells is less certain but seems to be lower (218), probably partly due to competition for Cavα1 and Cavβ binding by other proteins. The lower affinity likely permits a more dynamic Cavα1-Cavβ interaction. Indeed, several studies support the notion that the Cavα1-Cavβ interaction is reversible in intact cells. 1) Injection of β3 protein into oocytes expressing L-type Cavα1 alone quickly alters Ca2+ channel gating properties, suggesting that some channels on the plasma membrane are devoid of Cavβ (480). 2) A synthetic AID peptide can significantly reduce the Po of channels formed by L-type Cavα1 and β2a in HEK 293 cells when it is applied to the cytoplasmic side of inside-out membrane patches, but it has no effect on channels containing no β2a, suggesting that the AID peptide can compete off bound β2a (224). 3) Injection of β2a protein into oocytes expressing Cav2.3 and β1b results in a dramatic inhibition and slowing down of inactivation, consistent with β2a replacing previously bound β1b and overtaking the channel (218).
That Xenopus oocytes have endogenous Cavβs and that the Cavα1-Cavβ interaction has a 1:1 stoichiometry and is reversible provide a straightforward explanation for why Ca2+ currents can be recorded in oocytes expressing Cavα1 alone, and why the gating properties of these currents can be modulated by exogenous Cavβ: the endogenous β3XO subunits are expressed at high enough levels to interact with a small fraction of the nascent Cavα1 in the ER and transport them to the plasma membrane; there, β3XO eventually dissociates from Cavα1, leaving most of the channels devoid of a β subunit because the cytoplasmic concentration of β3XO is too low to rebind these β-less channels. The β-less channels, however, can associate with exogenously overexpressed Cavβ to form a stable Cavα1/Cavβ complex, as long as the cytoplasmic concentration of Cavβ is a few fold higher than the Kd of the Cavα1-Cavβ interaction. It is likely that this is also the scenario in mammalian expression systems.
A dynamic and reversible Cavα1-Cavβ association might play an important role in regulating Ca2+ channel activity, especially during development when changes in the expression level of different Cavβ isoforms occur (311, 435, 449). It has been shown that the Cavβ component of N-type Ca2+ channels changes during postnatal development, from β1b > β3 >> β2 at P2 to β3 > β1b = β4 at P14 and adult age (449). This study further shows that although no N-type channels associate with β4 at P2, 14 and 25% of N-type channels contain β4 at P14 and adult age, respectively.
VIII. ROLE OF CaVβ IN G PROTEIN INHIBITION OF CaV2 CHANNELS
VGCCs are susceptible to negative-feedback inhibition by hormones and neurotransmitters through the activation of G protein-coupled receptors (GPCRs). An extensively studied form of inhibition is the G protein-mediated, membrane-delimited, and voltage-dependent inhibition of members of the Cav2 channel family (i.e., N-, PQ-, and R-type channels). It is believed that this inhibition contributes to presynaptic inhibition and short-term synaptic plasticity (36, 52, 126, 219, 440, 471). This inhibition is mediated by the direct binding of G protein Gβγ subunits to the channel (213, 233), and it demonstrates three hallmarks: 1) it shifts channel activation to more depolarized potentials (23); 2) it is accompanied by a slowing of channel activation (23), resulting from latent Gβγ unbinding from the channel (139, 244, 349); and 3) it can be reversed by a strong conditioning depolarizing prepulse, which accelerates Gβγ dissociation from the channel in a phenomenon known as prepulse facilitation, or PPF (23, 139, 251). Below we discuss the role of Cavβ in the Gβγ-mediated, voltage-dependent inhibition. For in-depth reviews on other aspects of this inhibition, see References 117, 126, 138, 424, 440, and 497.
A. Cavβ Is Required for Voltage-Dependent Gβγ Inhibition
It has long been observed that some effects of Gβγ on VGCCs, such as the slowing of activation and the depolarizing shift of the voltage dependence of activation, are opposite to those of Cavβ, raising the possibility that Gβγ and Cavβ compete with each other (48, 60). Supporting this idea, early studies found that knockdown of endogenous Cavβ in neurons increased GPCR-induced inhibition of Ca2+ currents (60), and coexpression of Cavβ with Cavα1 in oocytes decreased G protein-mediated inhibition (48, 366). However, later studies showed that in COS-7 cells G protein inhibition of N-type Ca2+ channels was markedly enhanced by coexpressed Cavβs (315), and that in tsA-201 cells, a mutant Cav2.2 that contained a point mutation in the AID (W391A) and was unable to associate with Cavβ could no longer display voltage-dependent G protein inhibition (276). The latter studies indicate that Cavβ is essential for voltage-dependent G protein inhibition of N-type Ca2+ channels.
The discrepancy among these studies could arise from many factors. In particular, in the early studies (48, 60, 366), G protein inhibition was examined at a single voltage, which could complicate the interpretation because Gβγ and Cavβ both shift the voltage dependence of channel activation, but in opposite directions. Another factor could be the difficulty of 1) characterizing inhibition of tiny Ca2+ channel currents typically recorded in the absence of coexpressed Cavβ, and 2) excluding the contribution of endogenous Cavβs. To overcome these difficulties, a mutant β2a subunit (named β2a_Mut2) was created by mutating two key AID-binding residues (M245 and L249) to alanine (502). When coexpressed with Cav2.1 in Xenopus oocytes, β2a_Mut2 is still capable of promoting channel trafficking, but owing to its reduced affinity for the AID, it can be washed off from the surface of Ca2+ channels in excised membrane patches (502). With the use of this approach, large populations of Ca2+ channels devoid of Cavβ can be generated on the plasma membrane. Such β-less channels are still inhibited by purified Gβγ protein applied to the cytoplasmic side of the channels; however, all the hallmarks of voltage-dependent inhibition are absent (502). This finding strongly supports the notion that Cavβ is indispensible for voltage-dependent Gβγ inhibition.
Although β-less channels do not display voltage-dependent G protein inhibition, they can still be inhibited by G proteins in a voltage-independent manner. For example, the mutant Cav2.2 harboring the W391A mutation is still susceptible to voltage-independent G protein inhibition (276). Likewise, the β-less Cav2.1 channels (produced by washing off the bound β2a_Mut2 in inside-out membrane patches) are also still inhibited by Gβγ, but without any voltage-dependent features (502). These findings indicate that Gβγ can bind Cavα1 in the absence of Cavβ. Thus the essential role of Cavβ is to enable voltage-dependent dissociation of Gβγ from the inhibited channels, a process that gives rise to the voltage dependence of Gβγ inhibition (45).
B. Gβγ Does Not Displace Cavβ
Another important question is whether Cavβ and Gβγ coexist on Cavα1 during voltage-dependent inhibition. The apparent opposing actions of Gβγ and Cavβ prompted the hypothesis that Gβγ displaces Cavβ from Cavα1 (48, 60, 366, 375). This conclusion was also reached in a study showing that Förster resonance energy transfer (FRET) signals between Cavβ and Cavα1 change during Gβγ inhibition (387). However, functional antagonism does not necessarily indicate direct competition, and, while FRET signal changes are indicative of protein conformational changes, they are inadequate in demonstrating protein dissociation (3, 231, 490).
On the contrary, several lines of evidence indicate that Cavβ remains associated with Cavα1 during Gβγ modulation. 1) Different subfamilies of Cavβ have different effects on the magnitude and properties of voltage-dependent Gβγ inhibition, with β2a being the least effective in promoting this inhibition (61, 129, 147, 316, 376). 2) Cavβ increases the rate of Gβγ dissociation (as determined by the time constant of PPF) from the inhibited channels, but the efficacy of the four Cavβs is different (with a rank order of β3 > β4 > β1b > β2a; Refs. 61, 147). These observations (1 and 2) are most easily explained if Cavα1, Cavβ, and Gβγ form a tripartite complex during Gβγ modulation. 3) Since Cavβ critically affects VDA and VDI, these properties are expected to change if Cavβ were dislodged from the channel. However, the voltage dependence and kinetics of VDI remain unchanged before, during, and after Gβγ modulation (23, 316, 502). Similarly, the voltage dependence of activation is unchanged before and after Gβγ modulation (502). These results support the notion that Cavβ is not dislodged by Gβγ from the inhibited channels.
C. The Rigid IS6-AID α-Helix Is Necessary for Voltage-Dependent Gβγ Inhibition
The antagonistic effects of Cavβ and Gβγ on channel activation suggest that their actions are related structurally and mechanistically. As mentioned in section viC, disruption of the α-helical structure of the IS6-AID linker (by inserting 3–7 glycine residues in the linker or by substituting 3 linker residues with glycine) abolishes Cavβ modulation of VGCC gating (151, 455, 502). The same maneuver also completely eliminates voltage-dependent Gβγ inhibition but spares voltage-independent Gβγ inhibition (502), indicating that a rigid IS6-AID helix is not necessary for Gβγ binding to Cavα1 but is essential for voltage-dependent dissociation of Gβγ. Strikingly, both voltage-dependent and -independent Gβγ inhibition are abolished when five alanine residues are inserted into the IS6-AID linker of Cav2.1, which is likely to maintain the α-helical structure of the linker but produce a ∼180° rotation of Cavβ with respect to Cavα1 (502). It is possible that Gβγ can no longer bind to this mutant Cav2.1, but further investigation is necessary to confirm this speculation.
Consistent with the requirement of a rigid IS6-AID linker, two recent studies show that the GK domain alone is sufficient to support voltage-dependent G protein inhibition in both N- and P/Q-type channels (129, 502). On the other hand, the observation that different isoforms of Cavβ differentially modulate voltage-dependent G protein inhibition indicates that other Cavβ domains and regions can fine-tune this process. Indeed, the HOOK region has been suggested to play a role in enhancing the voltage-dependent dissociation of Gβγ (129).
On another note, the differential effect of different Cavβs on Gβγ-mediated inhibition can be further exposed by the expression of other proteins. For example, RGS2, which is a member of the regulators of G protein signaling that catalyze GTP hydrolysis and terminate G protein signaling (498), can unmask differences in G-protein modulation of P/Q-type channels containing different types of Cavβ (300).
D. Model for the Voltage Dependence of Gβγ Inhibition
Before presenting a model for voltage-dependent Gβγ inhibition, we first consider the molecular components involved in this process. 1) Several distinct regions in Cavα1, all of which bind Gβγ in vitro, play a role in voltage-dependent Gβγ inhibition, including the NH2 terminus (3, 345), the I–II loop (118, 346, 439, 496), and to a lesser extent, the COOH terminus (3, 189, 231, 282, 366, 499). The NH2 terminus of Cavα1 binds directly to the I–II loop, and together they form a Gβγ-gated inhibitory module (3). 2) The I–II loop has two Gβγ binding sites: one extends from the COOH-terminal end of IS6 to the NH2-terminal end of the AID and contains a signature Gβγ-interacting QxxER motif (QQIER in Cav2.1), and the other is located further downstream of the AID (118, 439, 496). The downstream site, termed the G protein interaction domain or GID, is likely to be an anchoring site for Gβγ. It has a ∼20 nM affinity for Gβγ (118, 496), and when applied as a 21-amino acid peptide, it can prevent PPF. The upstream site containing the QxxER motif has a ∼60 nM affinity for Gβγ, and its mutations attenuate Gβγ modulation (214). However, this site may serve only as a secondary Gβγ-binding site in the holo-channel for three reasons. 1) It is partially buried by Cavβ, as shown by the AID-Cavβ core crystal structures (84, 341, 447). Hence, Gβγ binding to the upstream site is significantly weaker in the presence of Cavβ than in the absence of Cavβ (502). 2) The QxxER motif is unlikely to become completely available, since Cavβ does not vacate from the Gβγ-bound channels (61, 231, 315, 502), as discussed above. When seven alanine residues are inserted in the upstream site, which is expected to prevent Gβγ binding to this site, voltage-dependent Gβγ inhibition remains intact (502). 3) As discussed in section iiiE, Cavβ binding to the AID results in the formation of an α-helix extending continuously from IS6 to the AID. The integrity of this α-helix is critical for voltage-dependent Gβγ inhibition, as mentioned above.
Figure 8 depicts an allosteric model proposed recently for the origin of the voltage dependence of Gβγ inhibition of Cav2 channels (502). This model links voltage-dependent dissociation of Gβγ to the voltage-dependent movement of IS6 and to the obligatory role of Cavβ. The pocket where Gβγ binds in the holo-channel to produce the voltage-dependent inhibition is still unknown, but it is postulated to be downstream of the COOH-terminal end of the AID and is formed collectively by portions in the NH2 terminus, the I–II loop, and the COOH terminus (and possibly yet unknown regions). Under the resting condition and in the presence of Gβγ, the channel is inhibited (Fig. 8A, left). Upon depolarization, the S6 segments of Cavα1 move, and owing to the continuous rigid α-helical structure of the IS6-AID linker, this movement is transmitted to and beyond the AID, resulting in a movement of the distal I–II loop and, consequently, a conformational change of the Gβγ-binding pocket. Such a chain of events ultimately leads to the disassembly of the NH2 terminus-I–II loop inhibitory module and the dissociation of Gβγ from the channel (Fig. 8A, right), which account for the slowing of the activation kinetics and prepulse facilitation. In the absence of Cavβ, Gβγ can still bind to the holo-channel but cannot be discharged by the depolarizing potential, because the Gβγ-binding pocket is uncoupled from IS6 as a result of the unwinding of the AID into a random coil (Fig. 8B). Such uncoupling can also be produced by glycine insertions in the IS6-AID linker (Fig. 8C). In summary, this model postulates that the voltage dependence of Gβγ inhibition of Cav2 channels arises from the voltage-dependent movement of IS6 and that Cavβ and a rigid IS6-AID linker play a pivotal role in translating this movement to Gβγ dissociation.
IX. ROLE OF CaVβ IN RGK INHIBITION OF HIGH-VOLTAGE ACTIVATED Ca2+ CHANNELS
The RGK (Rad, Rem, Rem2, Gem/Kir) family of Ras-related monomeric small GTP-binding proteins has emerged as potent inhibitors of HVA Ca2+ channels (27, 155). There are four members in this family: Rad (Ras associated with diabetes; Ref. 370), Rem (or Ges = human ortholog; Ref. 153), Rem2 (157), and Gem/Kir (91, 296). They share a conserved Ras-like core but differ from other Ras members in that their GTP/GDP-binding domains have nonconserved mutations that alter or abolish the GTP/GDP cycle (410, 489). They also contain extended NH2 and COOH termini. The COOH terminus has a motif that can anchor them to the membrane (reviewed in Refs. 81, 104, 211, 296) and is critical for their function (81, 104, 105, 488). RGK proteins have two known functions: shaping cytoskeletal dynamics and inhibiting HVA Ca2+ channels (24, 27, 104, 255, 324). These two functions can be differentially regulated; for example, RGK modification of cytoskeletal reorganization, but not inhibition of HVA Ca2+ channels, is attenuated by dephosphorylation of certain RGK residues (152, 463). The physiological importance of RGK inhibition of HVA Ca2+ channels is illustrated by recent in vivo studies that manipulate endogenous Rad levels with consequences for the heart (79, 462, 478). For example, dominant negative suppression of endogenous Rad in the heart increases L-type Ca2+ channel currents and action potential duration in cardiac cells and causes longer QT intervals and arrhythmias (478). Here, we only discuss RGK inhibition of HVA Ca2+ channels and the role of Cavβ in this process. For more comprehensive reviews of RGK proteins and their functions, see References 104 and 255.
A. Cavβ Is Essential for RGK Inhibition of HVA Ca2+ Channels
All members of the RGK family are able to inhibit, in a voltage-independent manner, HVA Ca2+ channels when expressed in various heterologous expression systems (15, 24–27, 81, 101, 102, 154–156, 165, 281, 397, 478, 487, 488). This inhibition depends on Cavβ, as RGK proteins do not affect Ca2+ channel currents recorded in cells expressing only Cavα1 (27, 155, 397 but see Ref. 106). Consistent with this notion, RGK proteins do not affect the activity of T-type Ca2+ channels, which do not associate with Cavβ nor require Cavβ for their activity (81, 155). Furthermore, RGK proteins interact directly with Cavβ, both in vitro and in cells (24–28, 101, 102, 154–156, 165, 281, 487), and this interaction seems to be promiscuous whereby any RGK protein can interact with any full-length Cavβ. A structural model of Gem-β3 interaction has been recently developed (28) based on systematic mutagenesis analysis and homology modeling based on the crystal structure of the AID-β3 core complex (84) and a crystal structure of GDP-bound Gem (PDB 2G3Y). This model shows that Gem binds to the β3 GK domain at a site distinct from the AID-binding pocket and that residues D194, D270, and D272 in β3 and R196, V223, and H225 in Gem are critical for this interaction. Supporting this model, mutating these residues individually or in combination severely weakens or abolishes in vitro binding of Gem and β3 (28, 144).
Three mechanisms of RGK inhibition have been reported. The first mechanism, reported in the first study on RGK regulation of HVA Ca2+ channels (27) and advanced in subsequent studies mostly from the same groups (24–26, 28, 388, 478), is that all RGK proteins disrupt the trafficking of HVA Ca2+ channels to the plasma membrane and hence reduce the number of surface Ca2+ channels, as determined primarily by imaging the subcellular localization of epitope-tagged Cavα1. It was hypothesized that RGK proteins compete with Cavα1 by sequestering Cavβ in the cytoplasm and/or the nucleus, thus leaving Cavα1 trapped in the ER. Several observations are consistent with this hypothesis. 1) Nuclear targeting of Rem and Rad causes nuclear sequestration of Cavβ (24). 2) In addition to Cavβ, all RGK proteins also interact with CaM and 14–3-3 (24–26, 158, 297). Abolishing CaM and 14–3-3 binding or CaM binding alone results in nuclear accumulation of RGK proteins (24–26), suggesting that these interactions regulate the subcellular localization of RGK proteins. Moreover, Rem2 and Gem (but not Rad and Rem) mutants deficient in CaM binding are unable to inhibit HVA Ca2+ channel currents (24–27, 463).
On the other hand, other findings are inconsistent with the aforementioned mechanism (i.e., sequestration of Cavβ and disruption of channel surface expression). 1) Cavβ-Cavα1 binding through the AID-GK domain interaction is much stronger than RGK-Cavβ binding (154), making it unlikely for RGK proteins to compete off Cavβ from Cavα1. 2) Several studies show that Cavα1, Cavβ, and RGK proteins form a trimeric complex in vitro and in cells (28, 101, 144, 154, 487). 3) Work from several groups show that RGK proteins can inhibit HVA Ca2+ channels without affecting their surface expression, with the latter determined by surface binding of a radioactive toxin (81), surface biotinylation (144, 154, 156), or gating charge measurement (487). These observations suggest that cytoplasmic and/or nuclear sequestration of Cavβ by RGK proteins occurs via a mechanism other than competition with the AID and may, if at all, only partially account for the observed RGK inhibition of HVA Ca2+ channels.
The second mechanism, closely related to the first one, is that RGK proteins decrease the number of surface Ca2+ channels by increasing their internalization. A recent study shows that Rem enhances dynamin-mediated endocytosis of L-type channels expressed in HEK 293 cells (488).
The third mechanism of RKG inhibition is the suppression of the activity of channels already on the plasma membrane. Such direct inhibition has been observed for different RGK proteins and different HVA Ca2+ channels in a variety of expression systems. For example, Rem2 inhibits endogenous surface N-type channels in native neurons (81), Rem inhibits surface L-type channels expressed in pancreatic β-cells (154, 156) and HEK 293 cells (488), and rapid translocation of a recombinant Rem derivative acutely inhibits L- and N-type channels expressed in tsA201 cells (487). Recently, Gem inhibition of P/Q-type channels was reconstituted in inside-out membrane patches by direct application of a purified Gem protein domain (144). Furthermore, it was found that this acute inhibition was completely abolished when Cavβ was washed away from the patch (using the strategy mentioned in sect. viiiA), leaving P/Q channels β-less, but it was fully restored after application of a purified Cavβ protein, demonstrating that Cavβ is indispensable for Gem inhibition of surface P/Q channels. Finally, in agreement with a recent study (281), it was found that the Cavβ GK domain alone was sufficient to support Gem inhibition of P/Q channels expressed in Xenopus oocytes (144).
The absolute requirement of Cavβ for RGK-mediated inhibition of HVA Ca2+ channels is reminiscent of such a requirement for voltage-dependent Gβγ inhibition of N- and P/Q-type channels. The latter process further requires a rigid α-helical structure of the IS6-AID linker. It was shown, however, that disrupting the IS6-AID α-helix by glycine insertions did not affect Gem inhibition of P/Q channels (144). This result suggests that Gem inhibition uses a fundamentally different mechanism from voltage-dependent Gβγ inhibition and may not involve IS6-transmitted conformational changes in the pore.
The three mechanisms of RGK inhibition discussed above operate on different time scales and are not mutually exclusive. For example, in the case of Rem inhibition of L-type channels expressed in HEK 293 cells, both reduction of surface channel density and direct inhibition of surface channels occur (488). Which mechanism dominates or is utilized in native cells remains to be determined, but it likely depends on the type and expression level of RGK proteins as well as Ca2+ channel subunits.
B. A New Paradigm for RGK Inhibition of HVA Ca2+ Channels
As mentioned above, all members of the RGK family can interact with all four subfamilies of Cavβ. The RGK-Cavβ interaction has been widely presumed to be essential for RGK inhibition of HVA Ca2+ channels (15, 24–27, 81, 101, 102, 104, 154–156, 165, 206, 281, 397, 478). However, it was recently reported that while Cavβ is required for Gem inhibition of surface P/Q-type channels, the interaction between Cavβ and Gem is not (144). In this study, residues D194, D270, and D272 in β3 and R196, V223, and H225 in Gem were simultaneously mutated to alanine; these residues are predicted and shown to be critical for the Gem-β3 interaction (28, 144). This combination of mutations completely abolished binding of Gem and β3, and yet, the mutant Gem (named Gem_Mut3) was still fully capable of inhibiting P/Q channels containing either the mutant or WT β3 (144). Meanwhile, it was found that Cav2.1 could be coimmunoprecipitated by WT Gem or Gem_Mut3, either in the presence or absence of β3, suggesting that Gem directly interacts with Cav2.1. Chimeric studies with PQ- and the RGK-insensitive T-type channels indicate that the IIS1-IIS3 region of Cavα1 is essential for gem inhibition (144).
Based on these results and those discussed above, we propose a “Cavβ-priming” model for Gem inhibition of P/Q-type Ca2+ channels on the plasma membrane (Fig. 9; Ref 144). (This model may be applicable to Rem and Rem2 since they also inhibit surface Ca2+ channels.) A distinct feature of this model is that the interaction between Gem and Cavβ is not necessary for Gem's inhibitory effect, but a direct association between Gem and Cav2.1 is essential. In this model, Gem interacts directly with Cav2.1 through an anchoring site, with or without Cavβ being present. In the presence of Cavβ and Gem, Cav2.1 forms a multimeric complex with both proteins on the plasma membrane (Fig. 9A). Binding of Cavβ to Cav2.1 produces a conformational change, resulting in the formation of an inhibitory site in Cav2.1 where Gem binds to produce inhibition (Fig. 9A). When Cavβ dissociates or is washed off from surface Cav2.1, the inhibitory site disappears, rendering Gem unable to inhibit Cav2.1, even though it can remain attached to Cav2.1 via the anchoring site (Fig. 9B). When Cavβ and Gem mutants that cannot bind to each other are used (Fig. 9C), inhibition can nevertheless proceed since the ability of Cavβ and Gem to bind Cav2.1 is not compromised. Thus the essential role of Cavβ is to convert Cav2.1 into a state permissive for Gem inhibition.
At present, this model remains speculative, and many questions remain unanswered. For example, where is the anchoring site for Gem on Cav2.1? Where is the inhibitory site? How does Cavβ binding to Cav2.1 create the inhibitory site? How does Gem binding to the inhibitory site lead to channel inhibition? Does the Gem-Cavβ interaction play any role at all? With regard to the last question, we speculate that in native cells, with physiological levels of Gem protein, the Gem-Cavβ interaction may increase the effective concentration of Gem near surface Ca2+ channels and thereby facilitate Gem inhibition.
X. ROLE OF CaVβ IN PHOSPHO- AND LIPID REGULATION OF HIGH-VOLTAGE ACTIVATED Ca2+ CHANNELS
Phosphorylation allows dynamic regulation of protein functions, including those of HVA Ca2+ channels. During the “fight-or-flight” response, for example, β-adrenergic stimulation leads to PKA-dependent upregulation of L-type Ca2+ channel currents, which results in a faster and stronger heartbeat. The activity of HVA Ca2+ channels is regulated by a variety of protein kinases and phosphatases. While Cavα1 is often the target of phosphorylation, Cavβ can nevertheless modulate the effect of such phosphorylation, and in some cases, Cavβ itself is the target. HVA Ca2+ channels can also be regulated by membrane lipids and their metabolic products. This section discusses the role of Cavβ in phospho- and lipid regulation of HVA Ca2+ channels.
A. Ca2+/Calmodulin-Dependent Kinase II
Ca2+/calmodulin-dependent kinase II (CaMKII) is among the most abundant enzymes in many cell types. Recent studies show that CaMKII can interact directly with Cavα1 of HVA Ca2+ channels and regulate their activities (242, 273, 358). In cardiac and smooth muscle cells, CaMKII plays a role in the facilitation of L-type Ca2+ channels (133, 192, 308). In cardiomyocytes, the molecular mechanism partly involves CaMKII-mediated phosphorylation of the β2a subunit (192). CaMKII binds to the COOH terminus of β2a and phosphorylates it at T498, which leads to an upregulation of L-type currents (192). This upregulation is not observed in the absence of β2a or when a nonphosphorylatable β2a mutant, β2a(T498A), is coexpressed in tsA201 cells. This mutant can also act as a dominant negative, preventing CaMKII-mediated facilitation of endogenous Ca2+ currents. Moreover, T498 phosphorylation promotes the dissociation of CaMKII from β2a, which may serve as a negative-feedback mechanism (193). Since most β2 splice variants have a common COOH terminus identical to that of β2a (Fig. 6), it would be interesting to examine whether CaMKII also interacts with and phosphorylates these β2 variants.
CaMKII has also been shown to associate with β1b in vitro, but not with β3 and β4 (193). A recent study further shows that CaMKII coimmunoprecipitates with forebrain L-type Ca2+ channel complexes containing β1 or β2 but not β4 (1). β1b, β3, and β4 have also been shown to be phosphorylated by CaMKII (193), but the physiological consequences remain to be determined.
B. Mitogen-Activated Protein Kinase
Mitogen-activated protein kinase (MAPK) is a member of a signaling network that responds to extracellular stimuli and induces diverse physiological and pathological processes. The small monomeric G protein, Ras, can upregulate Ca2+ currents in dorsal root ganglion neurons through the activation of the MAPK signaling pathway (160). In COS-7 cells, this upregulation is shown to require Cavβ, because in the absence of Cavβ, MAPK-dependent upregulation is abolished (159). Furthermore, different Cavβs support different degrees of upregulation. For example, in the presence of β2a, but not other Cavβs, Ca2+ channels are partially resistant to inhibition by an antagonist of MAPKK, the exclusive activator of MAPK (159). It is speculated that MAPK directly phosphorylates the channel complex (159). While both Cavα1 and Cavβ have consensus MAPK phosphorylation sites, it is unclear which, if any, are phosphorylated in cells.
C. Phosphoinositide 3-Kinase and Protein Kinase B (or Akt)
Some external stimuli (e.g., insulin-like growth factor) activate receptors that are associated with tyrosine kinases and upregulate L- and N-type Ca2+ channel currents (42). The mechanism likely involves phosphoinositide 3-kinase (PI3K) activation and subsequent production of phosphatidylinositol 4,5-bisphosphate (PIP2) and phosphatidylinositol 3,4,5-trisphosphate (PIP3), known regulators of HVA Ca2+ channels (for reviews, see Refs. 122, 318). Increased PIP3 levels recruit protein kinase B (PKB) to the membrane, which can phosphorylate β2a at S574 (454). This results in an upregulation of currents conducted by channels containing β2a and Cav1.2 or Cav2.2, mainly by increasing surface expression (454). Ca2+ channels containing an unphosphorylatable β2a (bearing the S574A mutation) are resistant to upregulation by PI3K/PKB, whereas those containing a phosphorylation-mimic β2a (bearing the S574E mutation) exhibit tonically increased current and are insensitive to upregulation by PI3K. The permissive role of β2a is not shared by other Cavβs and does not depend on palmitoylation (454).
A similar PI3K/PKB signaling pathway may play a role in maintaining normal cardiac function. Thus, in the absence of active PKB (created in a cardiac-specific conditional knockout), L-type Ca2+ channel surface expression is greatly reduced, leading to severe cardiomyopathy (70). This deficit results from lysosomal degradation of L-type channels, initiated by conserved PEST sequences (signals for rapid protein degradation) in Cav1.2. When PKB is active, it binds and phosphorylates β2, which in turn masks the degradation signals and leads to an increased channel surface expression (70).
D. cAMP-Dependent Protein Kinase
cAMP-dependent protein kinase (PKA)-mediated upregulation of cardiac L-type Ca2+ channel currents was one of the first examples of ion channel modulation. During the “fight-or-flight” response, β-adrenergic stimulation, through G proteins and adenylate cyclase, results in increased levels of cAMP and subsequent activation of PKA, eventually leading to dramatic increases of cardiac L-type Ca2+ currents and the consequent faster and stronger heartbeat (for reviews, see Refs. 72, 309). In spite of intense research, the target of PKA phosphorylation that underlies L-type current upregulation still remains obscure, partly owing to the difficulty of reconstituting this regulation in heterologous expression systems. In vivo, this upregulation is accompanied by increased phosphorylation of both Cav1.2 and β2 (54, 72, 114, 115, 180, 197). A PKA phosphorylation site on Cav1.2 (S1928) has been proposed to mediate the β-adrenergic response (115). However, a recent study using the Cav1.2 (S1928A) knock-in mice shows that basal L-type currents and the upregulation of L-type currents by PKA and β-adrenergic receptor stimulation are unchanged (275), indicating that PKA phosphorylation of S1928 is not the underlying cause for L-type current upregulation.
On the other hand, it has been shown that, in vitro, PKA phosphorylates three sites on β2a: S459, S478, and S479 (180). It was initially proposed that β2a phosphorylation at S478 and S479 is critical for L-type channel upregulation (54, 173). This was based on the lack of or reduced upregulation of currents conducted by L-type channels containing β2a(S478A/S479A) (54, 173). A caveat is that these studies were done either with a truncated Cav1.2 (54) or without comparative experiments with WT β2a (173). The role of β2a phosphorylation in L-type current upregulation has been strongly challenged by a recent study, which shows that in cardiac muscle cells L-type channels containing β2a(S459A/S478A/S479A) exhibit the same degree of PKA-mediated upregulation as channels containing WT β2a (321). This study further demonstrates that the extent of PKA modulation is influenced by the associated Cavβ. Thus channels containing β1b show the strongest upregulation, followed by those containing β3 and β4, whereas channels containing β2a show the least modulation (probably because β2a dramatically increases channel Po in the first place, as mentioned in sect. viA). Since β1b, β3, and β4 do not share the aforementioned PKA phosphorylation sites in β2a, the latter observation further strengthens the notion that PKA phosphorylation of β2a does not play an essential role in the upregulation of cardiac L-type currents. Thus identifying the site(s) of PKA phosphorylation that give rise to this event remains a stubborn challenge.
E. Protein Kinase C
Protein kinase C (PKC) can enhance some neuronal L- and N-type Ca2+ channel currents (483). In Cav2 channels, the enhancement partly results from the disruption of Gβγ inhibition (19, 21, 96, 202, 426, 496). This is achieved by phosphorylating the I–II loop of Cavα1, which may block Gβγ binding, considering that the I–II loop contributes to form the Gβγ-binding pocket (see sect. viiiD). Since Cavβ also binds to the I–II loop, it is conceivable that its binding is sensitive to PKC phosphorylation, or vice versa. In fact, pharmacological activation of PKC can increase Cav2.2 and Cav2.3 channel currents, but only in the presence of Cavβ, and transferring the I–II loop from these channels to the unresponsive Cav2.1 and Cav1.2 channels can transfer PKC sensitivity (49, 413). Thus it seems that Cavβ is permissive for PKC modulation of certain Cavα1. However, the effects of PKC phosphorylation of Cavβ itself (e.g., β2a; Ref. 180) are unknown.
F. cGMP-Dependent Protein Kinase
cGMP-dependent protein kinase (PKG) is activated by cGMP and can phosphorylate both Cav1.2 and β2a-Cav1.2 at S1928 (the same residue that can also be phsophorylated by PKA) and β2a at S496 (484, 485). Activation of PKG results in inhibition of channels containing Cav1.2 and β2a in HEK cells (484). This inhibition is abolished by the β2a S496A mutation, suggesting that PKG phosphorylation of β2a is critical for this process. The physiological consequences of this inhibition remain to be determined.
G. Arachidonic Acid
Like many other types of ion channels, the activity of HVA Ca2+ channels can be regulated by membrane lipids and their metabolic products (for reviews, see Refs. 46, 122, 172, 318, 373). One example is the voltage-independent inhibition of HVA Ca2+ channels upon PIP2 depletion from the plasma membrane (171, 475). Another example is channel inhibition by arachidonic acid (AA), an unsaturated fatty acid released from phospholipids (including PIP2) by the action of some phospholipases (18, 289, 374, 403). The magnitude of AA inhibition depends on the partnered Cavβ in the channel (374). In particular, β2a seems to dampen AA inhibition of Cav1.3 channels. This unique effect of β2a is abolished when palmitoylation is eliminated in a mutant β2a (β2a C3,4S). Attaching a transmembrane segment of an unrelated membrane protein to the NH2 terminus of this mutant β2a does not restore the dampening effect, suggesting that palmitoylation rather than membrane anchorage per se is responsible for the antagonizing effect of β2a on AA-mediated inhibition. It is proposed that the palmitoyl groups of β2a compete with AA for a common binding site on Cav1.3 (374). This unique ability of β2a has also been proposed to underlie the enhancement of Cav2.2 channels containing β2a by the stimulation of Gq-coupled receptors, which, in contrast, causes inhibition of channels containing β1b, β3, or β4 (210).
XI. INTERACTION OF CaVβ WITH OTHER PROTEINS
For many years, Cavα1 was the only known interacting partner for Cavβ. In recent years, however, a growing number of proteins have been found to interact with Cavβ, in some cases with significant functional impact. This section reviews some of these interactions and their functional consequences. Two more examples are discussed in section xii, in the context of a potential role of Cavβ in transcriptional regulation.
A. Ryanodine Receptors
In skeletal muscle, Ca2+ channel complexes (DHPRs), which are made up of Cav1.1, β1a, α2δ1, and γ1 subunits, are arranged on the plasma membrane of t tubules in tetrads, which are in contact with ordered RyRs in the membrane of adjoining sarcoplasmic reticulum (SR) (for reviews, see Refs. 167, 363). This physical arrangement is required for efficient excitation-contraction (EC) coupling. It has been shown that β1a, which is expressed exclusively in skeletal muscle, is indispensible for EC coupling (for reviews, see Refs. 100, 164). This is because β1a is essential not only for the surface expression of DHPRs (37, 38, 191, 393, 394, 402) but also for the tetrad formation (393, 394). It seems that β1a allosterically primes Cav1.1 to properly interact with RyRs to form the tetrads (393). Expression of exogenous Cavβ in skeletal muscle cells isolated from β1a-null mice or zebrafish can fully rescue L-type Ca2+ channel current, but only β1a (and β1c) can normalize EC coupling (38, 393, 402). This unique ability is due to the presence of a heptad repeat (L478-V485-L492) in the distal COOH terminus of β1a (393, 401, 402). When this region is deleted from β1a, EC coupling is lost, and when it is transferred to β2a, EC coupling is observed. It is unclear, however, where the heptad repeat binds (402). What is certain is that β1a can bind directly to the RyR (85, 164). On the RyR, β1a binds to a highly charged region (KKKRRxxR), whose mutation attenuates EC coupling (85). Interestingly, two pathogenic mutations, R3348H and P3527S, occur in this region of the RyR and cause malignant hyperthermia susceptibility (384) and multi-minicore congenital myopathy (148), respectively. It remains to be determined whether the loss of interaction with β1a is the underlying cause.
Ahnak, a ubiquitous large (700 kDa) signaling and scaffolding protein involved in diverse aspects of cell physiology and pathophysiology (reviewed in Refs. 7, 195), has been shown to interact with Cavβ in several distinct cell types, including cardiac cells, osteoblasts, and T lymphocytes (6, 199, 223, 305, 398). This interaction provides a potential link between Ca2+ channels, the cytoskeleton, and cellular organelles. Multiple regions in the COOH terminus of Ahnak can bind β2a in vitro, with apparent affinities ranging from 50 to ∼300 nM (196, 199, 223). The Ahnak-interacting region on β2a is unknown, but since Ahnak coimmunoprecipitates with β1b, β3, and β2a (6, 199, 398), it is likely to be in the conserved GK or SH3 domains.
It has been proposed that the Ahnak-β2a interaction plays a role in PKA-mediated upregulation of cardiac L-type currents (195, 196). Both β2a and Ahnak can be phosphorylated by PKA, and this phosphorylation weakens the binding between the two proteins by ∼50% (196). This effect is accompanied by L-type current upregulation, a hyperpolarizing shift in the voltage dependence of activation, and an occlusion of subsequent PKA effects (196, 199). Thus it was suggested that under basal conditions, L-type channel activity is suppressed by the Ahnak-β2a interaction and that PKA phosphorylation of both proteins disengages β2a from Ahnak and hence relieves this tonic inhibition (195). However, a challenge to this proposal is that PKA phosphorylation of β2a does not appear to play a role in the β-adrenergic receptor stimulation-induced upregulation of cardiac L-type currents, as discussed in section xD.
In the immune system, T cells are central in cell-mediated immunity. T cells are nonexcitable cells, yet they express all four Cav1 channels and all four Cavβs (241, 419). Activation of T-cell antigen receptors (TCR) causes Ca2+ influx, which is key for T-cell activation. This Ca2+ influx is thought to be mediated in part by Cav1 channels, although the mechanism of channel activation remains unclear (11, 241, 305, 419). CD4+ T lymphocytes isolated from β3- and β4-null mice and CD8+ T cells isolated from β3-null mice display impaired TCR-triggered Ca2+ response (11, 241), presumably because of deficient surface expression of Cav1 channels. T cells from Ahnak1 knockout mice also respond poorly to TRC stimulation and have impaired Ca2+ influx (305). It is proposed that this deficit results from decreased plasma membrane expression of Cav1 channels, owing to the lack of the Ahnak1-Cavβ interaction (305); however, this hypothesis needs further testing.
C. BKCa Potassium Channels
Large-conductance Ca2+-activated K+ channels (BKCa) are synergistically activated by membrane depolarization and intracellular Ca2+ and play important roles in various physiological processes (383). Even though BKCa channels have their own auxiliary subunits, it has been shown recently that BKCa channels bind directly to the Ca2+ channel β1 subunit (508). In HEK 293T cells, this interaction dampens the Ca2+ sensitivity of BKCa channels and slows their activation and deactivation. The GK domain of β1 is necessary and sufficient for these effects, suggesting that other Cavβs may share this β1 function, but this remains to be determined. It is also unknown whether the Cavβ-BKCa channel interaction occurs in native cells, and whether and how this interaction affects HVA Ca2+ channels.
Bestrophin (Best1) is a 585–604 amino acid chloride channel expressed in the retinal pigment epithelium (RPE) whose mutations cause Best's and other retinopathies (see review in Ref. 205). It modulates L-type Ca2+ channel gating and blocks L-type channel-mediated rises in [Ca2+]i in RPE (301, 378). Recently, it was shown that Best1 binds the Ca2+ channel β4 subunit and that its effects disappear in the absence of β4, suggesting that Best1 may be acting through β4 on Cav1.3 channels (493). Befittingly, β4 knockout mice also have rethinopathies (301). The COOH terminus of Best1, which on its own does not generate Cl− currents, can also inhibit L-type channels. It contains a predicted proline-rich domain (PRD) whose mutation abolishes the effects of Best1. It is proposed that this PRD binds to the SH3 domain of β4, disrupts the GK-SH3 interaction, and causes L-type channel inhibition (493). However, direct evidence for such a mechanism is still lacking. It would be interesting to examine whether the PxxP-binding region of β4 binds Best1. As discussed in section iiiB, although this region is occluded in the Cavβ crystal structures (84, 341, 447), it could conceivably become accessible when Cavβ is bound to Cavα1 and other proteins.
A recent study reported that full-length β2a interacts in vitro with dynamin, a multi-partner GTPase involved in endocytosis (186). This interaction was presumed to involve a PRD of dynamin and the SH3 domain of β2a, since a purified β2a fragment (amino acids 24–136) containing the four contiguous β sheets of the SH3 domain was found to interact with dynamin and this interaction was partially blocked by the dynamin PRD. This β2a fragment was able to markedly suppress the surface expression of Cav1.2 channels, and this suppression depended on dynamin. It was proposed that the dynamin-SH3 domain interaction links HVA Ca2+ channels to the endocytotic machinery (186). It should be noted, however, that the β2a fragment used in this study lacks the fifth (i.e., the last) β sheet of the SH3 domain and the HOOK region, which hinders access to the PxxP-binding region of Cavβ (84, 341, 447). To determine whether the PxxP-binding region of β2a is involved in the interaction between dynamin and full-length β2a, it would be useful to examine whether this interaction is abolished by selective mutations of β2a residues that are presumably directly involved in binding PRDs.
F. Synaptic Proteins: Synaptotagmin I and RIM1
The α1 subunit of presynaptic HVA Ca2+ channels physically interacts with presynaptic proteins, including syntaxin, SNAP-25, and synaptotagmin I; these interactions are important for synaptic vesicle docking and fusion (reviewed in Ref. 400). Recent studies show that Cavβ can also interact with synaptic proteins (257, 452). For example, the NH2 terminus of β3 and β4a (but not β4b) binds to synaptotagmin I, and this interaction is abolished by a high concentration (10 mM) of Ca2+ (452). However, the physiological importance of this interaction is yet unknown. Another study shows that RIM1, a presynaptic protein critical for synaptic transmission and plasticity (69, 392), binds directly and with a high affinity (35 nM) to β4b and β2a. This interaction is mediated by the COOH terminus of RIM1, and the SH3-HOOK-GK module of Cavβ is sufficient for binding to occur. The most prominent effect of the RIM1-Cavβ interaction on HVA Ca2+ channels is the slowing of VDI and a hyperpolarizing shift of the voltage dependence of inactivation. This effect is observed on recombinant L-, N-, P/Q-, and R-type channels containing β4b and on recombinant P/Q-type channels containing β1a, β2a, β3, or β4b. RIM1 may also play a role in anchoring synaptic vesicles to presynaptic VGCCs through binding to the synaptic vesicle protein Rab3. Consequently, overexpression of a mutant Cavβ that is unable to bind Cavα1 can attenuate vesicle docking at the presynaptic membrane in PC12 cells, presumably by competing with the WT Cavα1/β complex for RIM1 binding. Furthermore, overexpression of RIM1 in PC12 cells and cultured cerebellar neurons enhances neurotransmitter release. This study establishes a direct role of Cavβ in the physical organization of the synaptic vesicle release machinery (257).
G. Zinc Transporter 1
Zinc transporter 1 (ZnT-1), a ubiquitous transmembrane protein involved in zinc transport and metabolism, binds directly to β2a (280) and has been shown to inhibit L-type Ca2+ channels in heterologous expression systems and native cells (29, 280, 337, 396). When coexpressed with Cav1.2 and β2a in Xenopus oocytes, ZnT-1 reduced Cav1.2 channel currents (280). This inhibitory effect disappeared in the absence of β2a or in the presence of an excess amount of β2a. ZnT-1 reduced the surface expression of Cav1.2 without changing its total expression level when they were coexpressed in HEK 293T cells (280). The authors proposed that ZnT-1, through direct binding to Cavβ, inhibits L-type channels by reducing their trafficking to the plasma membrane (280). It remains to be examined whether ZnT-1 physically interacts with other types of Cavβs and whether it inhibits Cav2 channels.
XII. Ca2+ CHANNEL-INDEPENDENT FUNCTIONS OF CaVβ
Until recently, the functions of Cavβ have been exclusively linked to VGCCs. However, a stream of recent studies suggests that Cavβ may possess functions independent of their association with VGCCs. This line of inquiry began with the cloning of various short isoforms of Cavβ, some of which lacked the GK domain. This inability to engage in the high-affinity AID-GK domain interaction with Cavα1 raised questions about their functions (92, 166, 204, 217, 229). The first study examining possible alternative functions of truncated splice variants of Cavβ centered on a β4 splice variant expressed in chicken cochlea and brain, termed β4c (217); to avoid confusion, we refer to it as cβ4c. cβ4c is truncated after exon 8 and thus lacks 90% of the GK domain and the entire COOH terminus (Fig. 6). As expected, cβ4c barely affects Cav2.1 channels coexpressed with α2δ in Xenopus oocytes. However, cβ4c interacts directly with the scaffolding domain of heterochromatin protein 1 (HP1), a nuclear protein involved in gene silencing and transcriptional regulation. Both proteins are colocalized in the nuclei of cochlear hair cells, and their coexpression in tsA201 cells causes translocation of cβ4c from the cytoplasm to the nucleus. Moreover, cβ4c attenuates the repressor function of HP1 in a dose-response manner. The effects on HP1 are specific since a longer isoform, β4a, has no effect. These findings suggest that cβ4c may function as a transcription regulator.
In a recent study, Zhang et al. (504) reported that full-length β3 could directly interact with a new splicing isoform of Pax6, a transcription factor critical for the development of the eye and nervous system. The new isoform, named Pax6(S), has a truncated COOH terminus with a unique serine-rich tail. The interaction between Pax6(S) and β3 is conferred mainly by the S tail and the SH3-HOOK-GK module of β3. Since the other three subtypes of Cavβ can also interact with Pax6(S), the binding site for the S tail likely resides in the conserved SH3 or GK domain. Coexpression of Pax6(S) with Cav2.1 channels containing β3 in Xenopus oocytes does not alter channel properties; however, the in vitro transcriptional activity of Pax6(S) is markedly suppressed by β3. Furthermore, coexpression of β3 and Pax6(S) in HEK 293T cells results in the translocation of β3 from the cytoplasm to the nucleus. These results suggest that full-length Cavβs may function as transcription regulators (504).
This notion is further supported by other recent studies (16, 427), which show that, upon neuronal differentiation, full-length β4a physically interacts with B56δ, a nuclear regulatory subunit of phosphatase 2A (PP2A). The β4a/B56δ complex relocates to the nucleus, where it associates with nucleosomes and regulates the dephosphorylation of histones, a key mechanism in transcriptional regulation. A mutant β4 lacking the last 38 COOH-terminal residues and associated with a case of juvenile myoclonic epilepsy (142), can neither associate with B56δ nor translocate to the nucleus, and its in vitro transcriptional regulation activity is different from that of WT β4 (16, 427). Formation of the β4a/B56δ complex requires an intact intramolecular SH3-GK interaction.
Consistent with the notion that full-length Cavβs may function as transcription regulators, it has been shown that full-length Cavβs can be targeted to the nucleus in native cells. For example, β4, and to a lesser extent, β1b and β3, are translocated into the nucleus when they are exogenously expressed in cardiac cells (93). A recent study reports that endogenous β4 is present in the nuclei of cerebellar granule cells and Purkinje cells (425). When heterologously expressed in skeletal myotubes or cultured hippocampal neurons, β4b is robustly targeted to the nucleus, whereas other Cavβs are not. Nuclear localization of β4b is dependent on an Arg-Arg-Ser motif in the NH2 terminus, which is necessary since deleting this motif decreases nuclear targeting of β4b. This motif is also sufficient since fusing it to β4a increases nuclear targeting of the resulting chimera. Importantly, nuclear targeting of β4b is diminished upon increased electrical activity and Ca2+ influx through L-type Ca2+ channels, suggesting a potential physiological function (425).
Another possible VGCC-independent function for β4 is demonstrated by a recent study in zebrafish (135), which express all four subtypes of Cavβs (506). Morpholino knockdown of zebrafish β4 abolishes or retards epiboly, an early development process, due to disturbances in mitotic and postmitotic cytoskeletal rearrangements. Epiboly can be rescued by coinjecting full-length human β4a or β4b cRNA. Interestingly, epiboly can also be rescued upon coinjection of a mutant β4a, which contains a triple mutation (M204A/L208A/L350A) in its AID-binding pocket and cannot bind Cavα1 or enhance Ca2+ channel currents in Xenopus oocytes. These results suggest an involvement of β4 in zebrafish early development, probably through VGCC-independent actions. It remains to be determined how β4 is involved and whether this function is shared by other Cavβs.
In a study using β3 knockout mice, high glucose conditions caused pancreatic β cells to produce twofold more insulin than their WT counterparts (31). No change in VGCC currents was detected, but the β3-deficient cells exhibited a higher frequency of glucose-induced intracellular [Ca2+] oscillations, accompanied by increased IP3 production and increased Ca2+ release from intracellular stores. On the basis of the colocalization of these proteins in pancreatic β cells, it was hypothesized that β3 may directly interact with IP3 receptors to cause some of these effects (31).
In snails (Lymnaea), the expression of the sole Cavβ (LCavβ) is temporally and spatially uncoupled from the expression of LCav2, a Lymnaea homolog of the mammalian Cav2 family of VGCCs (409). Functionally, LCavβ does not modulate the surface expression or gating of LCav2 when they are coexpressed in tsA201 cells, even though they show current upregulation and gating modulation when they are partnered with rat Cavα1 and Cavβ, respectively (409). Furthermore, knockdown of LCav2, but not LCavβ, alters neurite morphology. These results suggest that LCavβ may have VGCC-independent functions, which remain to be elucidated. This work illustrates that studies in simple model organisms might be beneficial to our understanding of the full spectrum of Cavβ functions.
XIII. CaVβ KNOCKOUTS AND PATHOPHYSIOLOGY
As expected, because of the essential role of Cavβ in the surface expression and functional modulation of HVA Ca2+ channels, Cavβ knockouts or mutations can produce severe functional deficits and, in some cases, are lethal. The phenotype of Cavβ knockout mice depends on the ability of the remaining three Cavβ genes to compensate. This section discusses the phenotypes and pathophysiology of Cavβ knockouts and mutations.
A. Gene Knockouts and Mutants
As mentioned in section xiA, β1a is irreplaceable in partnering with Cav1.1 channels to enable skeletal muscle EC coupling. Thus β1 knockout mice, similar to Cav1.1 knockouts, are born motionless and die immediately from asphyxiation (191). Skeletal muscles isolated from β1-null mice are twitchless upon electrical stimulation, and action potentials do not elicit Ca2+ transients. L-type Ca2+ channel currents and the surface expression of Cav1.1 subunits are much reduced in these muscles, but caffeine can still cause contractions, indicating that internal Ca2+ stores are intact (191). Transgenic expression of β1a exclusively in the skeletal muscle of β1-null mice rescues the mice, which exhibit no obvious phenotype, suggesting that the remaining Cavβ genes can compensate for the functions of β1 in other tissues (14).
Zebrafish β1 knockouts also exist (393, 394). They have the Relaxed phenotype and die paralyzed days after hatching, with completely deficient EC coupling. Skeletal muscles from β1-null zebrafish have no tetrads and show reduced depolarization-induced Ca2+ transients, but exhibit normal caffeine-induced Ca2+ transients (394). Unlike in β1-null mice, targeting of Cav1.1 subunits to t tubules and the formation of triads are preserved (394), suggesting a nonessential role of β1 in these processes in zebrafish. It remains to be determined whether other Cavβs are expressed in β1-null zebrafish skeletal muscles or whether zebrafish Cav1.1 is able to traffic to the plasma membrane on its own.
Several β2 splice variants are the predominant Cavβs expressed in the heart (Table 1). Thus it is no surprise that β2 knockouts have no cardiac contractions and are nonviable beyond embryonic day 10.5 (14, 468). This is due to diminished L-type Ca2+ channel currents in cardiomyocytes and cardiac failure-associated defective remodeling of blood vessels. The β2-null phenotype can be rescued by the expression of β2 under a cardiac muscle-specific promoter (14). These partial knockouts revealed an essential role of β2 in tissues besides the heart: such mice (lacking β2 in all but cardiac tissues) are deaf due to a dramatic reduction in the membrane expression of Cav1.3 channels in inner hair cells, coupled with decreased exocytosis, improper hair cell development, and defective cochlear amplification (331). These “rescued” mice also have defects in vision with a phenotype similar to human patients with congenital stationary night blindness (13).
Given the knockout results, genetic mutations in β1 and β2 are expected to affect mainly skeletal and cardiac muscles, respectively. While no β1 mutations have been associated with genetic diseases thus far, β2 mutations have. Thus a mutation in the COOH terminus of β2b (CACNB2b), S481L, contributes to a type of sudden death syndrome characterized by a short QT interval and an elevated ST segment (8), which are categorized into a group of genetic heart diseases called the Brugada syndrome. This mutation decreases Cav1.2 currents by ∼75% in an expression system (CHO-K1 cells). Another mutation, in the β2b NH2 terminus (T11I), causes accelerated inactivation of cardiac L-type channels and is also linked to the Brugada syndrome (97). This mutation occurs in exon 2C of the CACNB2 gene and only affects β2b, the most abundant Cavβ isoform in the heart (93). A recent study suggests that variations in CACNB2 may be also associated with a heightened risk for Alzheimer's disease (286).
β3 Knockouts are viable and were initially found to be normal (328, 330). Later studies, however, uncovered a wide spectrum of abnormalities, especially under stress conditions. For example, at high glucose concentrations, the frequency of [Ca2+]i oscillations and the resulting insulin secretion from pancreatic β cells are potentiated (31). This is likely due to the attenuation of β3-mediated inhibition of IP3 production (31, 339). Also, a high-salt diet causes elevation of blood pressure, reduction in plasma catecholamine levels, and a hypertrophy of heart and aortic smooth muscle (328, 329). The latter effects, as well as observations from mice overexpressing β3 (327), are consistent with a function of β3 in sympathetic control. In this regard, β3 knockouts resemble N-type channel (Cav2.2) knockouts (428, 429), which is not surprising since N-type channels preferably partner with β3 (294, 325–327, 395). Indeed, sympathetic neurons from β3-null mice have reduced N- and L-type channels activity (330). N-type current is also decreased in dorsal root neurons, which dampens inflammatory pain, but not mechanical or thermal pain (325). In the brain, N-type channel expression is reduced by ∼40% (326); in the hippocampus, expression of NR2B (an NMDA receptor subunit), NMDA receptor currents, and long-term potentiation are all increased (239). Some forms of hippocampus-dependent learning and memory appear to be enhanced, but working memory is impaired (239, 326). Furthermore, pruning of visual retinocollicular pathways is developmentally reduced and delayed (98). Behaviorally, β3-null mice have lower anxiety, increased aggression, and increased nighttime activity (326). Finally, β3-null mice show abnormal signaling in CD4 T-cells, where receptor-mediated Ca2+ responses, nuclear translocation of NFAT, and cytokine production are all attenuated (11).
Mutated β4 was first reported in lethargic mice (55, 130, 131). The naturally occurring null mutation is a four nucleotide insertion in Cacnb4, causing a translational frame shift and a premature stop codon. Lethargic mice have ataxia, seizures, absence epilepsy, and paroxysmal dyskinesia (17, 55, 227). The abnormal phenotype appears after postnatal day 15, a time when WT animals have an increase in β4 expression in the brain (311). The upregulation of β4 in WT mice is particularly robust in cerebellar granule and Purkinje neurons, which likely explains the ataxia in null mice (55). T-type Ca2+ channel upregulation (by ∼50%) in thalamic neurons of lethargic mice likely contributes to the seizures (503). It is not clear why the remaining Cavβs fail to compensate for the lack of β4, but this could be partly because of the unique interactions between the NH2 and COOH termini of β4 with other proteins (for examples, see Refs. 51, 121, 420, 460, 461, 474). Nevertheless, in lethargic mice, there is increased pairing of Cav2.2 and Cav2.1 with other Cavβs; in particular, both β1b and Cav2.2/β1b complexes are upregulated, similar to what is found in the developing brain (310, 311). Some other characteristics of lethargic mice include lower N-type channel expression in the forebrain and cerebellum (310), reduced excitatory neurotransmission in the thalamus (57), a modified electro-oculogram (301), splenic and thymic involution (130, 131), and renal cysts (130). Similar to β3 knockouts, CD4+ T-cells have attenuated receptor-mediated Ca2+ responses, nuclear translocation of NFAT, and cytokine production (11).
Since β4 is the predominant partner for P/Q-type (Cav2.1) channels in brain (311), it is no surprise that tottering mice (161), with mutations in Cav2.1, have a phenotype very similar to lethargic mice (356). Both tottering and lethargic mice are also models for epilepsy (17, 226). Indeed, there are examples where mutations in β4 precipitate epilepsy and ataxia in humans. In one case, an R468Q mutation in CACNB4, which enhances Cav2.1 current, was associated with a history of febrile seizures (338). In another, truncated β4 (R482x) that only has a very minor effect on HVA Ca2+ channel properties was found in a juvenile myoclonic epilepsy patient (142). In yet another case, a mutation in the SH3 domain of β4 (C104F) causes different symptoms in two different families: episodic ataxia in one and generalized epilepsy and praxis-induced seizures in the other, presumably as a result of different genetic backgrounds (142).
B. Cavβ in Pathophysiology
Changes in the expression level of various Cavβs have been reported in certain pathological conditions. For example, in hypertrophic obstructive cardiomyopathy, β2 is upregulated, which likely drives the observed increase in Ca2+ channels (198, 465). Downregulation of Cavβ is observed in allografts from diastolically failing hearts (228), in pancreatic islets from type 2 diabetic rats (235), and during atrial fibrillation (188). However, these observations are only correlative, and it remains unclear whether these changes are causative or incidental to the disease.
In the Lambert-Eaton myasthenic syndrome (LEMS), an autoimmune disease, autoantibodies against the extracellular loops of presynaptic Ca2+ channels disrupt channel arrays at the neuromuscular junction and impair synaptic transmission (269, 299). However, antibodies against β3 and β4 are also common in sera from LEMS patients (55% of the time), including in five of five LEMS patients who also had small-cell lung carcinoma (368). In some instances, the Cavβ autoantibodies can prevent Cavα1-Cavβ binding (368, 377). It is unclear, however, how Cavβ autoantibodies contribute to the disease, since Cavβ is an intracellular protein and is unlikely to be a target of the Cavβ autoantibodies in the intact muscle. Indeed, immunization of rats with a purified Cavβ protein causes no neuropathy in spite of the induction of high antibody titers (453). These observations and considerations suggest that Cavβ autoantibodies do not directly contribute to the pathology of the disease, but their presence can serve as an additional diagnostic tool (368).
Finally, schistosomiasis, or bilharzia, is a parasitic disease caused by Schistosoma flatworms, which infect ∼200 million people in the developing world, damaging the nervous system and internal organs. It is relatively successfully cured with Praziquantel (PZQ). The exact mechanism of action is still unclear, but PZQ seems to target a variant of Schistostoma Cavβ (reviewed in Refs. 124, 190, 240). Schistosomas have one “conventional” Cavβ and one, named βvar, with a long COOH terminus and nonconserved changes in both the SH3 and GK domains (240, 263, 264, 334). When expressed with a mammalian Cavα1, βvar modulates gating as expected for a Cavβ, but it causes a decrease in current amplitude (263). PZQ recovers current amplitude (263), consistent with results showing that PZQ causes a Ca2+ influx into worms, followed by a sustained muscular contraction and paralysis (240). It is not clear, however, whether βvar associates with Schistosoma Cavα1 since expressing them has been difficult (190). A new study shows that Cavβ knockdown in Schistosomas, using siRNA, confers resistance to PZQ, further implicating Cavβ (334). Thus PZQ likely targets Schistosoma Cavβ, but the downstream events remain to be elucidated (124, 190). They may involve an increase in Ca2+ channel currents but may also include other pathways. Recently, it was suggested that Ca2+ influx on its own is not sufficient to kill Schistosomas, because cytochalasin D, an inhibitor of actin polymerization, can rescue Schistosomas from PZQ, in spite of cellular Ca2+ overload (354).
Great strides have been made in the last two decades in our understanding of the molecular biology, structure, function, and regulation of Cavβ. An emerging theme is that Cavβ is a multifunction protein, acting primarily as a Ca2+ channel regulatory subunit but also performing Ca2+ channel-independent functions. Although much is known, many important questions and issues remain to be elucidated, some of which we highlight here.
1) Although it is well established that Cavβ is essential for the surface expression of HVA Ca2+ channels, it is yet unclear why Cavβ is required. The traditional view is that Cavβ facilitates the export of Cavα1 from the ER. However, no definitive ER retention signals have been found on Cavα1 that are blocked by the binding of Cavβ. An alternative possibility is that Cavα1 can traffic to the plasma membrane on its own, but its continued presence there requires Cavβ. Further studies on the role of Cavβ in Cavα1 internalization, ubiquitination, and proteasomal degradation may shed light on this issue.
2) Given that many isoforms of Cavβ exist, that the association between Cavβ and Cavα1 is promiscuous, and that Cavβ regulates channel gating in a Cavα1-Cavβ pair-specific manner, there is enormous combinatorial complexity. Furthermore, the reversible nature of Cavα1-Cavβ association provides a means for dynamic regulation of HVA Ca2+ channel activity. Thus, to better understand the function and regulation of HVA Ca2+ channels in native cells, it is necessary to examine the spatial and temporal expression of different Cavβ isoforms, not only at cellular levels but also at subcellular levels, as exemplified by the work in neurons (317, 336). Currently, we know very little about the molecular mechanisms governing the splicing and expression of Cavβ in different tissues, cell types, and subcellular locations (e.g., soma vs. dendrites vs. axon terminals).
3) Although the high-affinity interaction between the Cavβ GK domain and the AID is essential for Cavβ regulation of HVA Ca2+ channel gating, it is the interactions among other regions of Cavβ and Cavα1 that confer distinct Cavα1-Cavβ pair-specific characteristics to Cavβ regulation. We do not yet have a full grasp of the molecular determinants involved in these interactions, owing to their intrinsically low affinity. Characterizing these low-affinity interactions will remain a difficult challenge, as conventional biochemical approaches may not be sufficient to uncover the underlying molecular components and mechanisms.
4) Useful knowledge has been gained from Cavβ knockouts; however, lethal phenotypes and/or compensation by other Cavβs have limited the amount of information gleaned from systemic knockouts. It would be desirable to achieve inducible and tissue-specific knockout, knockdown, or overexpression of a particular Cavβ. A recent study (90) suggests that such approaches may even become useful venues for gene therapy of certain forms of cardiovascular or neurological disorders.
5) The list of Cavβ-interacting proteins continues to grow, but in most cases, the physiological importance of their interactions with Cavβ in native cells remains unclear. Some VGCC-independent functions of Cavβ are presented in this review, but almost certainly more are to be discovered. In this regard, it would be particularly interesting to investigate whether, and under what conditions, full-length Cavβs participate in regulating gene expression in native cells.
6) To better understand the function of Cavβ, it would be valuable to obtain high-resolution structures of full-length Cavβs, by themselves and in complex with their various interacting partners. These are clearly long-term goals, and there will undoubtedly be technical challenges in such endeavors, but the successful determination of the crystal structure of three different Cavβ cores' and two different AID-Cavβ core complexes warrants optimism.
This work was supported by National Institute of Neurological Disorders and Stroke Grants NS-53494 and NS-45383 (to J. Yang).
No conflicts of interest, financial or otherwise, are declared by the authors.
We thank Kathryn Abele for reading and commenting on a draft of this article, Minghui Li and Yong Yu for helping with figures, and Linling He, Yun Zhang, and Mingming Fan for discussion.
Address for reprint requests and other correspondence: J. Yang, Dept. of Biological Sciences, 917 Fairchild Center, MC2462, Columbia University, New York, NY 10027 (e-mail:).
- Copyright © 2010 the American Physiological Society