Structure of a complex between a voltage-gated calcium channel β-subunit and an α-subunit domain

Abstract

Voltage-gated calcium channels (Ca_Vs) govern muscle contraction, hormone and neurotransmitter release, neuronal migration, activation of calcium-dependent signalling cascades, and synaptic input integration¹. An essential Ca_V intracellular protein, the β-subunit (Ca_Vβ)^1,2, binds a conserved domain (the α-interaction domain, AID) between transmembrane domains I and II of the pore-forming α₁ subunit³ and profoundly affects multiple channel properties such as voltage-dependent activation², inactivation rates², G-protein modulation⁴, drug sensitivity⁵ and cell surface expression^6,7. Here, we report the high-resolution crystal structures of the Ca_Vβ_2a conserved core, alone and in complex with the AID. Previous work suggested that a conserved region, the β-interaction domain (BID), formed the AID-binding site^3,8; however, this region is largely buried in the Ca_Vβ core and is unavailable for protein–protein interactions. The structure of the AID–Ca_Vβ_2a complex shows instead that Ca_Vβ_2a engages the AID through an extensive, conserved hydrophobic cleft (named the α-binding pocket, ABP). The ABP–AID interaction positions one end of the Ca_Vβ near the intracellular end of a pore-lining segment, called IS6, that has a critical role in Ca_V inactivation^9,10. Together, these data suggest that Ca_Vβs influence Ca_V gating by direct modulation of IS6 movement within the channel pore.

The 1.97 Å resolution structure of the Ca_Vβ_2a core shows that Ca_Vβs comprise two well-conserved domains (Fig. 1a). The first, an SH3 fold, contains five antiparallel β-strands (β1–β5), a 3₁₀ helix (η1), and two α-helices (α1 and α2) that lie amino-terminal to β1 and carboxy-terminal to β4, respectively. The strand that completes the SH3 fold, β5 (residues 217–224), is separated in the primary structure from the core of the SH3 domain by approximately 70 residues (variable domain 2, V2, a site of splice variation and amino acid insertions and deletions²) that are absent from the structure (Fig. 1b). The second conserved domain consists of a five-stranded parallel β-sheet (β6–β10), surrounded by six α-helices (α3–α8) and two 3₁₀ helices (η2 and η3), and is related to the core of nucleotide kinase enzymes.

Figure 1.

Structure of the Ca_Vβ_2a–Ca_V1.2 AID complex. a, Ribbon diagram of the complex. Dashed lines indicate regions absent from the structures. SH3 and nucleotide kinase (NK) domains are shown in green and blue, respectively. The AID is shown in red. Ca_Vβ_2a α-helices are labelled. Variable regions V1, V2 and V3 are indicated. The Ca_Vβ_2a unbound structure is similar to that shown here for the complex. The arrow indicates where the AID connects to transmembrane segment IS6. b, Sequence alignment of representatives of each Ca_Vβ isoform. The top sequence shows residues 40–425 of rat Ca_vβ_2a. Numbers on the right denote each line’s terminal residue. Shading denotes residues identical among isoforms. The two Ca_vβ_2a domains used for crystallization are indicated in green and blue, respectively. Secondary structure elements are indicated: α, α-helix; η, 3₁₀ helix; β, β-strand. Dashed lines indicate residues present in the crystallized constructs but absent in the electron density. Location of the V2 and part of the V3 regions are shown. Asterisks identify residues that contribute side-chain contacts to the AID-binding pocket; diamonds mark side chains with direct hydrogen bonds to the AID.

Ca_Vβs share structural features with membrane-associated guanylate kinases (MAGUKs), a protein scaffold family that organizes signalling components near membranes¹¹. MAGUKs contain one or more PDZ domains N-terminal to an SH3 domain, a bridging region known as a HOOK domain and a nucleotide kinase domain^11,12. PDZ domains are approximately 100 residues long. The Ca_Vβ_2a structure indicates that Ca_Vβs lack N-terminal PDZ domains. There are too few residues N-terminal to the SH3 domain (even in Ca_Vβ_1b, the Ca_Vβ with the longest (55 amino acids) N-terminal variable region 1, V1) to fold as a PDZ domain. The absence of a PDZ domain distinguishes Ca_Vβs from canonical MAGUK proteins.

Comparison of Ca_Vβ_2a with a representative MAGUK, PSD-95 (refs ^{12, 13}), reveals other differences. Superposition of the nucleotide kinase domains shows that the relative orientations of the SH3 and nucleotide kinase domains differ by approximately 90°, an arrangement that makes Ca_Vβ_2a a more elongated structure (Fig. 2a). The nucleotide kinase domain of MAGUKs is homologous to guanylate kinases and retains guanosine monophosphate (GMP) binding, but key residues for enzymatic function are missing¹². The four-stranded β-sheet nucleotide kinase subdomain that binds GMP in MAGUKs is absent in Ca_vβ_2a (Fig. 2a). Furthermore, two Ca_Vβ_2a loops (between β7–η2 and β8–β9) occlude part of the binding site for the GMP guanosine ring. Thus, the Ca_Vβ nucleotide kinase domain seems to have lost the ability to bind nucleotides.

Figure 2.

Structural comparisons between PSD-95 (gold) and Ca_Vβ_2a (blue). a, Superposition of Ca_Vβ_2a and PSD-95 nucleotide kinase domains (RMSD_Cα = 3.9 Å). The dashed circle indicates the guanosine-monophosphate (GMP)-binding domain present in PSD-95 but absent in Ca_Vβ_2a. The guanosine monophosphate molecule bound to PSD-95 is displayed in space-filling representation. Nucleotide kinase (NK) and SH3 domains are indicated. The relative change in SH3 domain orientation is indicated. b, Superposition of PSD-95 and Ca_Vβ_2a SH3 domains (RMSD_Cα = 1.6 Å). Position of the polyproline ligand from a superposition with the Sem5 SH3 domain (Protein Data Bank code 2SEM) (RMSD_Cα = 1.8 Å) is shown in space-filling representation. The Sem5 SH3 is not shown. The DALI server generated the superpositions (http://www.ebi.ac.uk/dali/).

The structures of Ca_vβ_2a and PSD-95 SH3 domains are similar (Fig. 2b). Neither is compatible with canonical modes of proline-rich ligand binding. Both lack the aromatic residues necessary for ligand engagement¹³, and the surface that would bind polyproline ligands is blocked by the α2 helix^12,13. In PSD-95, residues C-terminal to the nucleotide kinase domain contribute an extra SH3 β-strand that is absent from canonical SH3 domains¹³ and absent in Ca_vβ_2a. The HOOK domain, present in MAGUKs and Ca_Vβ_2a, bridges SH3 β-strands β4 and β5 and comprises α2 and variable domain 2 of Ca_Vβ_2a. HOOK domains are important regulatory regions for interactions of MAGUKs with other proteins^11,12,14 and may serve a similar function for Ca_vβ protein–protein interactions.

Ca_vβs exert their effects on Ca_V function by binding the pore-forming α₁ subunit at a conserved, 18-residue sequence located between membrane domains I and II (the AID)^3,15 (Fig. 3a). Interpretation of mutagenesis and biochemical studies suggested that Ca_vβ–AID interactions occur through a 41-residue segment (Ca_Vβ_2a residues 212–252) termed the β-interaction domain (BID)^8,16. The Ca_Vβ_2a structure shows that the central region of the BID, which includes the residues previously thought to be important for BID–AID interactions⁸, is entirely buried and cannot participate directly in protein–protein interactions (Fig. 3b). Two putative BID phosphorylation sites^8,16 are also buried in this region. Given the extent of burial, mutations used to determine the relative importance of residues involved in BID–AID interactions (for example, proline to arginine)⁸ are likely to have abolished AID binding by disrupting the folded structure of the nucleotide kinase domain rather than by perturbing direct AID contacts. Although it would appear that the Ca_Vβ_2a structure conflicts with previous data, most of the supporting evidence for the BID–AID interaction relied on indirect functional experiments and direct BID–AID binding was never demonstrated^8,16.

Figure 3.

Features of the AID–Ca_Vβ_2a interaction and location of the previously described BID. a, Sequence alignment of AID domains (Ca_V1.2 residues 428–445) and neighbouring residues. The positions of the last transmembrane segment of transmembrane domain I (IS6) and the first transmembrane segment of transmembrane domain II (IIS1) are shown. Secondary structure of the AID from the co-crystal structure is indicated (red). Dashed lines indicate residues absent from the electron density. Asterisks identify side-chain contacts with Ca_Vβ_2a closer than 4 Å. b, Position of the previously described BID (residues 212–252; yellow)³,⁸. Residues previously proposed to mediate AID–BID interactions (P224, P228, P234, Y239) are indicated and have relative accessibilities of 1.4%, 0%, 0% and 32.4%. Putative PKC sites S225, S235 and S345 are also shown (magenta) and have relative accessibilities of 8.8%, 0% and 35.4%, respectively. S345 accessibility reduces to 12% in the complex. Accessibility values are relative to a tripeptide, Gly-X-Gly. c, The left panel shows F_o−F_c electron density, contoured at 2σ, for the AID–Ca_Vβ_2a complex before building the AID. The right panel shows final 2F_o−F_c density, contoured at 1σ, for the AID from the refined AID–Ca_Vβ_2a structure (right). In both panels the final AID model is shown.

If the BID does not interact directly with the AID, how do Ca_vβ_2a and the AID interact? To answer this, we solved the high-resolution (2.00 Å) structure of a complex between the conserved Ca_Vβ_2a core and an 18-residue peptide containing the AID from the L-type channel Ca_v1.2. The electron density reveals the first 16 residues of the AID and the location of the binding pocket on Ca_vβ_2a (Fig. 3c). Overall, the Ca_vβ_2a structure is very similar to the unbound form (root-mean-square deviation (RMSD)_Cα = 0.397 Å) and bears only a few conformational changes in side chains near the AID-binding pocket (Ca_vβ_2a residues M244, N390, E393 and R351). The AID forms an α-helix that is anchored to the binding pocket through a set of conserved residues (AID residues L434, G436, Y437, W440 and I441) that are important for Ca_Vβ binding and for α-subunit modulation by Ca_Vβs^8,15,17. These residues bind a deep groove that we call the α-binding pocket (ABP), formed by helices α3, α5 and α8 of the Ca_vβ_2a nucleotide kinase domain at a site distal to the SH3 domain (Figs 1a, 3b and 4a).

Figure 4.

AID–ABP interactions. a, Surface representation of the Ca_Vβ_2a ABP, bound to the AID. The AID (gold) is shown in stick representation. Y437 and W440 are white. Ca_Vβ_2a residues that contribute hydrophobic (blue) and hydrogen bond (red) side-chain contacts to the AID are labelled. Select residues of the AID are labelled to orient the reader. b, c, Slices through the AID–ABP interaction at AID positions Y437 and W440 (gold). Labels indicate the AID residues.

The complex buries approximately 730 Ų of ABP surface, of which about 350 Ų are hydrophobic. AID side chains D433, W440 and Q443 make direct hydrogen bonds to the ABP (Supplementary Fig. 1). The depth and extent of burial of the aromatic AID positions Y437 and W440 is particularly striking (Fig. 4b, c). The AID Y437 hydroxyl group is central to a buried hydrogen bond network comprising three water molecules, AID residue D433 and five Ca_Vβ_2a residues (Supplementary Fig. 1). Mutation of this tyrosine to phenylalanine greatly diminishes AID–Ca_Vβ binding¹⁵. The extensive AID–ABP interactions are consistent with biochemical experiments demonstrating strong AID–Ca_Vβ interaction (dissociation constant ~6–20 nM)¹⁸. The Ca_Vβ side chains that contact the AID are highly conserved among Ca_Vβ isoforms (see Figs 1b, 3a and 4a; see also Supplementary Fig. 1). Thus, both binding partners engage each other through conserved residues to create the AID–ABP interaction.

Interactions between the Ca_Vα₁ and Ca_Vβ subunits markedly influence the cell surface expression of functional channels^6,7. Control of Ca_V trafficking by regulating Ca_Vα₁–Ca_Vβ interactions is emerging as an important means of modulating cellular excitability⁷. Ca_V channel subtypes are major clinical targets for drugs that treat cardiovascular disease, migraine and pain¹⁹. Development of compounds that could interfere with the AID–ABP binding interactions might provide new ways to modulate Ca_V function in pathological states.

The Ca_Vβ–AID structure provides a starting point for understanding how Ca_Vβ modulates numerous channel properties. G-protein βγ subunits (Gβγ) inhibit Ca_V function^1,4; however, the sites of Gβγ–Ca_V interactions and precise inhibitory mechanisms remain highly controversial⁴. Biochemical experiments show that Gβγ binds the Ca_V2 AID and that mutations at AID positions Q1, Q2, R5, L7, G9 and Y10 (corresponding to Ca_v1.2 AID residues 428, 429, 432, 434, 436 and 437) abolish Gβγ–AID binding²⁰. In contrast, other studies suggest that the in vitro Gβγ–AID interaction is functionally irrelevant and that the relevant Gβγ-binding determinants lie elsewhere in the channel cytoplasmic domains^21,22. The structure of the Ca_Vβ_2a–AID complex shows that three of the putative Gβγ–AID interacting positions (L7, G9 and Y10), which are invariant in Ca_V1 and Ca_V2 channels (Fig. 3a), are deeply buried by interactions with Ca_Vβ (Fig. 4a, b; see also Supplementary Fig. 2 and Table 2). The extent of burial of these residues, which are critical for maintaining Ca_Vβ–AID association^3,15, suggests that Gβγ and Ca_Vβ cannot bind to the AID simultaneously. Taken together with the observation that the Gβγ–AID affinity is at least 10–20-fold weaker than the Ca_Vβ–AID²⁰ affinity, the structure also indicates that it is unlikely that Gβγ could effectively compete with Ca_Vβ for AID binding without drastic structural rearrangement of the Ca_Vβ–AID complex. Thus, our data lend support to the view that the major Gβγ interaction sites lie in other Ca_V cytoplasmic domains^21–23.

How might Ca_Vβs affect channel gating? Although the detailed mechanisms of Ca_V inactivation processes remain unknown, functional experiments show that the IS6 pore-lining segment has a key role¹⁰. Motions of pore-lining transmembrane segments are a common theme emerging in ion channel gating^24,25. Twenty-two residues separate the AID helix N terminus from the cytoplasmic end of IS6. We do not know the structure, but sequence evaluation suggests that these residues have a high helix propensity and could readily form a continuous helix between the AID helix and the presumed transmembrane helix of IS6 (Fig. 5a). AID residue 5 (E432, here) slows inactivation when negatively charged (as in Ca_V1 channels) and speeds inactivation when positively charged (as in Ca_V2 channels)^26,27. This residue is exposed on the surface of the AID helix (Fig. 4a) where it would be available to interact with other parts of the channel and could influence the rates of movement of the Ca_Vβ–AID complex and therefore IS6. The profound inactivation rate slowing caused by Ca_Vβ_2a requires anchoring of the N terminus to the membrane by palmitoylation²⁸. Orientation of the AID helix towards IS6 places the N-terminal membrane anchor on the periphery of the α₁ subunit and suggests that Ca_Vβ_2a slows channel inactivation by restricting the movement of the IS6 transmembrane domain. Ca_Vβs have a deep groove between the SH3 and nucleotide kinase domains that is on the same face as the V2 domain. These features may be used to engage other cytoplasmic parts of the channel^29,30 and allow Ca_Vβ to couple motions in other channel domains directly to IS6 through AID attachment.

Figure 5.

Cartoon of proposed model for how Ca_Vβ affects Ca_Vα₁ gating. a, Orientation of Ca_Vβ_2a with respect to the I–II loop (red), pore-forming subunit, and connection to IS6. In Ca_Vβ_2a, variable region 1 (V1) is tethered to the membrane. The I–II loop between the AID N terminus and IS6 is depicted as a helix. Ca_Vβ_2a SH3 and nucleotide kinase domains are coloured green and blue, respectively. The arrow indicates that Ca_Vβ couples to IS6 movements (rotations, translations or both). b, View from the opposite side of a. The groove between SH3 and nucleotide kinase domains (demarcated by the arrow) and two flexible Ca_vβ_2a regions, the 275–284 loop and variable region 2 (V2), are on the same Ca_Vβ face, opposite the ABP. These regions may interact with other pore-forming subunit cytoplasmic domains.

The structures presented here represent the first high-resolution view of any part of the voltage-gated calcium channel, and provide an important step towards understanding the detailed molecular mechanism models for how Ca_Vs function. This work strongly suggests that Ca_Vβ affects Ca_V gating properties by directly influencing conformational changes that are likely to occur in the channel pore¹⁰.

Methods

The crystal structures of recombinant Ca_Vβ_2a and Ca_Vβ_2a–AID complexes were solved to resolutions of 1.97 Å and 2.00 Å, respectively. The final R/R_free values are 18.55%/21.32% for Ca_Vβ_2a and 19.97%/24.15% for the Ca_Vβ_2a–AID complex. Figures were prepared with PyMOL, MOLSCRIPT and RASTER3D. The experimental details for protein expression, purification, crystallization, structure solution, statistics of data collection, phasing and refinement are available as Supplementary Information.