Progress in the structural understanding of voltage-gated calcium channel (Ca_V) function and modulation

Abstract

Voltage-gated calcium channels (Ca_Vs) are large, transmembrane multiprotein complexes that couple membrane depolarization to cellular calcium entry. These channels are central to cardiac action potential propagation, neurotransmitter and hormone release, muscle contraction and calcium-dependent gene transcription. Over the past six years, the advent of high-resolution structural studies of Ca_V components from different isoforms and Ca_V modulators has begun to reveal the architecture that underlies the exceptionally rich feedback modulation that controls Ca_V action. These descriptions of Ca_V molecular anatomy have provided new, structure-based insights into the mechanisms by which particular channel elements affect voltage-dependent inactivation (VDI), calcium-dependent inactivation (CDI) and calcium-dependent facilitation (CDF). The initial successes have been achieved through structural studies of soluble channel domains and modulator proteins and have proven most powerful when paired with biochemical and functional studies that validate ideas inspired by the structures. Here, we review the progress in this growing area and highlight some key open challenges for future efforts.

Electrical activity drives each heartbeat and thought. This activity arises from the action of ion channels, membrane proteins that control ion passage across cell membranes and that constitute the molecular foundations of cardiac contraction and nervous system function.¹ Voltage-gated calcium channels (Ca_V) comprise a family of multisubunit transmembrane proteins that control cellular calcium entry in response to membrane depolarization.^2,3 Calcium ions have a special signaling role as they act as agents of membrane potential change and as potent effectors of signaling pathways.⁴ Consequently, Ca_Vs form a nidus for signals that couple calcium entry in response to membrane depolarization to many physiological processes such as neurotransmitter and hormone release, cardiac action potential propagation, excitation-contraction coupling and calcium-dependent gene regulation.^1,2 Because of these roles, Ca_V regulation plays a vital role in both the normal function and pathological states of the cardiovascular, endocrine, sensory and nervous systems.^5–7 Understanding how these molecular switches work, are regulated, and are integrated into macromolecular complexes and signaling pathways, and how they become deranged in cardiovascular and neuronal diseases, remain important questions at the frontier of modern biology.

There are a number of Ca_V subtypes that have diverse physiological and pharmacological properties that depend largely on the identity of the pore-forming Ca_Vα₁ subunit. High-voltage activated channels, Ca_V1s (Ca_V 1.x, L-type^3,8) and Ca_V2s (Ca_V2.x; 2.1, P/Q-type; 2.2, N-type; and 2.3, R-type)^3,7,9 are activated by strong depolarizations (∼+30 to +50 mV), whereas low-voltage activated Ca_V3s (Ca_V3.x, T-type)^10,11 are activated by weaker depolarizations (∼−55 to −20 mV). Ca_V1s and Ca_V2s are multiprotein complexes composed of four principal components¹² (Fig. 1A): a transmembrane, pore-forming subunit, Ca_Vα₁, that sets many of the functional and pharmacological properties of the channel,² a cytoplasmic subunit, Ca_Vβ, that shapes channel inactivation properties and regulates plasma membrane expression,^13,14 a membrane anchored subunit, Ca_Vα₂δ, that is the receptor for anti-epileptic and anti-nociceptive drugs^15,16 and calmodulin, CaM.¹⁷ In contrast, Ca_V3s require only the transmembrane subunit to function.

Figure 1.

Voltage gated calcium channel structure. (A) Cartoon diagram representing a Ca_V1 or Ca_V2 channel. The four homologous transmembrane domains of Ca_Vα₁ are indicated. Ca_Vβ is shown in dark blue and interacts with its high-affinity binding site on the I–II intracellular loop known as the “α-interaction domain, AID”. CaM is shown bound to the C-terminal cytoplasmic tail at the site of the IQ domain. The membrane associated Ca_Vα₂δ subunit is shown in orange and green. (B) Topology of the pore-forming Ca_Vα₁ subunit. Positions of the Ca_Vβ/AID complex, Ca²⁺/CaM-PreIQ complex and Ca²⁺/CaM-IQ domain complex are shown. Ca_Vα₁ intracellular loops are drawn to scale according to the human Ca_V1.2 sequence.

Because they are large multicomponent membrane protein complexes (∼0.5 MDa), little is known about the details of the molecular structure of complete Ca_Vs. Presently, the best data available on complete channels are low-resolution (∼30 Å) electron microscopy reconstructions.^18–24 Although progress is being made in identifying the possible orientation of the different components,^22–24 this level of resolution reveals only the gross shape of the channel and is not sufficient for developing detailed mechanistic hypotheses.

Ca_Vα₁ pore-forming subunits are polypeptide chains of ∼1,600–2,400 amino acids in which the ion-conducting pore is formed from four homologous repeats that each bear six transmembrane segments (Fig. 1B). Interdomain intracellular loops of various sizes bridge the four repeats and serve as docking sites for auxiliary subunits and regulatory molecules that control channel activity and connect Ca_V channels to other macromolecular complexes and cellular signaling pathways.^25,26 These regions have provided the foothold for the initial progress in high-resolution structural studies that have made use of the fact that, like most proteins, Ca_Vs contain folded domains that can exist in isolation. Similar to the extensive studies of other ion channels such as voltage-gated potassium channels, TRP channels and glutamate receptors,^27,28 studies of the individual domains have been most powerful when combined with biochemical and functional studies that test hypotheses raised by the structures and that allow validation of the structures. Here, we outline the recent progress in Ca_V high-resolution studies and some of the open challenges that remain for both the near and long term.

Ca_Vβ Subunits

The cytoplasmic Ca_Vβ subunit is an essential component of Ca_V1 and Ca_V2 channel complexes.^13,14 Four isoforms are found in mammals and have a wide range of splice variants. The first high-resolution structural studies came from a set of studies that reported the X-ray crystallographic structures of the core, conserved domains of Ca_Vβ_2a,^29,30 Ca_Vβ₃,³¹ and Ca_Vβ₄.³¹ The structures showed, in line with sequence prediction,³² that Ca_Vβs have two folded, conserved domains: an SH3 domain fold and a nucleotide kinase (NK) (also called GK) fold (Fig. 2A). The Ca_Vβ domain structure is related to that seen in a class of scaffolding proteins known as membrane-associated guanylate kinases (MAGUKs),³³ but has number of key differences including the relative orientation of the SH3 and NK domains. The SH3 domain has a number of unusual features. Its structure is incompatible with the canonical mode of polyproline binding for which SH3 domains are well known as it lacks key residues required for ligand engagement³⁰ and the surface that could be used for polyproline interactions is blocked by an α-helix that leads to the V2/HOOK loop.^29–31 More strikingly, one of the β-strands that forms the central SH3 core occurs in the primary sequence after the V2 domain (also called the HOOK domain, after a similar domain in MAGUKs) and is adjacent to the NK domain. This arrangement, also found in MAGUKs, results in an interdependence of the SH3 and NK domains that is crucial for the ability of Ca_Vβ to modulate Ca_Vα₁ function.^34–36

Figure 2.

Ca_Vβ-AID interactions. (A) Structure of the Ca_Vβ-AID complex from reference ³⁰. SH3 and NK domains are shown in green and blue, respectively. The AID is shown in red with the sidechains depicted as sticks. Variable domains V1, V2 and V3 are indicated. (B) Energetics of the Ca_Vβ-AID interaction. Residues are colored according to their impact on Ca_Vβ-AID binding as follows: <0.5 kcal mol⁻¹, dark blue; 0.5–1.5 kcal mol⁻¹, light blue; 1.5–3.0 kcal mol⁻¹, orange; >3 kcal mol⁻¹, red. Select AID (L434, L438 Y437, W440 and I441) and Ca_Vβ (M245 and L352) residues are labeled. (C) View of the complementary Ca_Vβ-AID hotspots from the perspective of Ca_Vβ. Colors and labels are as in (B). (D) Disruption of the AID-Ca_Vβ hotspot interaction by the AID mutant (HotA) abolishes Ca_Vβ-dependent modulation. Data in (B–D) are from Van Petegem et al.⁴⁰

The Ca_Vα₁ I–II loop bears the high affinity Ca_Vβ binding site, known as the “AID” (α-interaction domain).^37,38 AID-Ca_Vβ co-crystal structures revealed the molecular details of this interaction^29–31 and showed that the region from Ca_Vβ previously thought to be the main AID-interaction site, the ‘BID’,³⁹ was buried in the center of the NK domain and made no direct contacts with the AID. Instead, the AID binds to a deep grove on the NK domain, termed the α-binding pocket (ABP) (Fig. 2A and B), that encompasses ∼730 Ų and that involves sidechain contacts from twenty-four residues, a large number of which are highly-conserved.⁴⁰ Isothermal titration calorimetry studies show that the AID-ABP interaction is strong, regardless of the Ca_V1 or Ca_V2 AID isoform (Kd ∼3.5–53.5 nM).⁴⁰ The good correspondence between these values and those measured using larger portions of the I–II loop^41–43 provides solid support for the idea that the AID is the only high-affinity Ca_Vβ binding site on the I–II loop. Alanine scanning mutagenesis⁴⁰ reveals that the binding energy of the AID-ABP interface is focused into two complementary energetic hotspots. Nearly all of the binding energy comes from the interaction of just four amino acids, three from the AID (Y437, W440 and I441) and one from Ca_Vβ (Met245) (Fig. 2B and C). Targeted disruption of this interaction is sufficient to prevent association and channel modulation (Fig. 2D).^40,44 The restricted area of the interaction hotspots fits a general profile⁴⁵ that makes it an attractive target for the development of small molecule or peptidomimetic strategies to control Ca_V function by blocking the AID-ABP protein-protein interaction.

The presence of the AID helix, the relatively short distance from the cytoplasmic end of the IS6 segment from the pore forming subunit, and the predicted high helical propensity of the IS6-AID sequence prompted the structure-based hypothesis that Ca_Vβ exerts control on Ca_V function through a continuous α-helix that would connect the IS6 and AID helices (Fig. 3A).^29,30 Subsequent tests of this idea have provided good support for this model.^46–48 Polyglycine replacement of residues in the middle of the IS6-AID segment uncouples the effects of Ca_Vβ on voltage-dependent inactivation (VDI) and voltage-dependent activation in exemplars from both types of high-voltage activated channels: Ca_V1.2 (Fig. 3B and C),⁴⁷ Ca_V2.1,^47,48 and Ca_V2.2.⁴⁶ While direct high-resolution structural data for IS6-AID region are still lacking, circular dichroism studies of the IS6-AID segment from Ca_V1.2 and Ca_V2.2 support the ability of this conserved region to form an α-helix and also validate the detrimental impact of the polyglycine substitutions on helix formation.^47,49 Strikingly, disruption of the Ca_Vα₁-Ca_Vβ functional coupling, not only affects VDI, but also affects calcium-dependent feedback modulation processes that traditionally have been associated with calmodulin (CaM) interactions with the Ca_Vα₁ C-terminal tail (Fig. 3D and E).⁴⁷ This unexpected functional link suggests that both Ca_Vβ-dependent and CaM-dependent channel modulation couple to the pore by a common mechanism that requires Ca_Vβ and an intact IS6-AID linker.⁴⁷ Another type of channel modulation, voltage-dependent inhibition of Ca_V2.1 by the G-protein, Gβγ, also requires the presence of Ca_Vβ and helical structure in the IS6-AID segment.⁴⁸ Together, these studies point to a situation in which Ca_Vβ relays a diverse set of modulatory signals to the pore via its action on the IS6-AID helix.

Figure 3.

Ca_Vβ controls channel function via the IS6-AID helix. (A) Model showing the proposed continuous helix between the IS6 transmembrane segment and the AID. Model is based on the likely gross similarly between the transmembrane portions of Ca_V and KV channels. A surface representation (white) of the KV tranmembrane portion (PDB 2A79)¹⁴⁵ is used to depict the Ca_V transmembrane domains. The helical IS6-AID segment (red) was modeled manually by building an α-helix corresponding to the length between the KV1.2 S6 C-terminus and the helix from the Ca_Vβ_2a-AID complex (PDB 1T0J).³⁰ The Ca_Vβ_2a SH3 and NK domains are colored green and blue, respectively. Arrow indicates the site of the depicted polyglycine and polyalanine substitutions studied in Findeisen et al.⁴⁷ Effects of IS6-AID substitutions on (B) Ca_V1.2 VDI. C, Ca_V1.2 voltage-dependent activation. (D) Ca_V1.2 netCDI and (F) Ca_V1.2 CDF. Experimental details can be found in Findeisen et al.⁴⁷ (B–E) Reprinted from Findeisen et al.⁴⁷

The IS6-AID is not the only element through which Ca_Vβs can affect Ca_V function. The various Ca_Vβ isoforms have different effects on channel properties including stereotyped effects on the rank order of VDI^50–53 that arise from the Ca_Vβ variable domains^54,55 and that persist even when the IS6-AID segment structure is disrupted.⁴⁷ Structural information for the variable V1, V2 and V3 segments is presently limited. In the crystallographic Ca_Vβ studies, the variable regions were either removed to enable crystallization, or if present, lacked sufficient order to produce clear electron density. Presently, the only direct structural data for a Ca_Vβ variable region comes from nuclear magnetic resonance (NMR) structure determination of the Ca_Vβ_4a V1 domain. These studies included the α1-helix present in the Ca_Vβ core structures and show that the preceding region can adopt a compact folded structure consisting of a short α-helix and a β-hairpin.⁵⁶ Whether the other variable regions have structures in isolation or only fold upon interaction with Ca_Vα₁ remains unknown. The persistence of the Ca_Vβ rank order effects in the Ca_Vα₁ IS6-AID polyglycine mutants and clear functional crosstalk between the different intracellular domains supports the likelihood of physical interdomain interactions between Ca_Vβ and other channel elements. Definition of the structures of the Ca_Vβ variable domains and their sites of interaction on Ca_Vα₁ remains a challenging and open question.

Interaction of Ca_Vs with Ca²⁺/CaM

Calcium influx is a potent activator of intracellular signaling pathways but is toxic in excess. Therefore, its cellular entry is tightly regulated. Because Ca_Vs are major sources of activity-dependent calcium influx, their activity is strongly controlled by self-regulatory and extrinsic mechanisms that tune Ca_V action in response to electrical activity, neurotransmitter stimulation and hormone initiated cues. This intense regulation employs multiple mechanisms and intracellular signaling molecules. Calmodulin (CaM) has a preeminent role in the intrinsic mechanisms of Ca_V1s and Ca_V2 calcium-dependent regulation and in its calcium free form (apo-CaM) is thought to be tethered constitutively to the Ca_Vα₁ C-terminal cytoplasmic tail, making an arrangement that places CaM in a privileged position as a detector of Ca_V-generated calcium signals.^57,58

CaM is a bilobed, 16.7 kDa, calcium binding protein from a large family of calcium sensor proteins that has a central role in intracellular Ca²⁺ signaling and can interact with more than 300 different target proteins in a variety of binding modes.^59–61 CaM has two independently folded lobes,^62,63 N-lobe and C-lobe, that each contain two Ca²⁺-binding motifs known as EF-hands.^64,65 The CaM lobes undergo calcium-dependent conformational changes that expose a phenylalanine and methionine-rich hydrophobic pocket that is the principal site of target engagement.^4,66 CaM is responsible for driving two opposing Ca_V feedback modulation processes: calcium-dependent inactivation (CDI) and calcium-dependent facilitation (CDF). The main site of Ca²⁺/CaM binding is an isoleucine-glutamine ‘IQ’ domain^67–80 located in the Ca_Vα₁ subunit C-terminal cytoplasmic tail ∼150 amino acids after the last transmembrane segment, IVS6 (Figs. 1B and 4A).

Figure 4.

Ca²⁺/CaM-Ca_V IQ domain interactions. (A) Ca_V cartoon showing the relative locations of the EF-hand, PreIQ, A-region, C-region and IQ domains in Ca_V1 and Ca_V2 C-terminal tails. (B) Sequence comparison of Ca_V1 and Ca_V2 IQ domains. Colors indicate residues having contacts with Ca²⁺/N-lobe, Ca²⁺/C-lobe, or both domains in the parallel (Ca_V1) and antiparallel (Ca_V2) complexes. (C) Structure of the Ca²⁺/CaM-Ca_V1.2 IQ peptide complex (PDB: 2BE6).⁷⁵ Ca²⁺/N-lobe and Ca²⁺/C-lobe are shown in green and blue, respectively. Ca_V1.2 IQ domain is shown in firebrick and selected residues are shown in stick representation. The “IQ” positions are labeled. (D) Elimination of Ca²⁺/N-lobe aromatic anchors abolishes Ca_V1.2 CDF. Experiments show the response of Ca_V1.2 to a 3-Hz pulse train (50 ms steps to +20 mV from a holding potential of −90 mV) and are done in the background of the I1624A mutant that unmasks CDF. Reprinted from Van Petegem F, et al.⁷⁵ (E) Structure of the Ca²⁺/CaM-Ca_V2.3 IQ peptide complex (PDB: 3DVK).⁷³ Ca²⁺/CaM lobes and calcium ions are colored as in (C). Ca_V2.3 IQ domain is shown in orange and the aromatic anchor positions are white and shown in stick representation. (F) Elimination of Ca²⁺/C-lobe anchors demonstrates importance for Ca_V2.1 CDF. Ca²⁺ currents in wild-type (WT) and mutant Ca_V2.1 channels elicited by trains of an action potential waveform (APW) at 100 Hz. A single exemplar depolarizing APW pulse and elicited Ca²⁺ current is shown (left). Reprinted from Kim et al.⁷³

Elegant experiments using CaM mutants impaired in calcium binding at specific lobes have demonstrated that the individual lobes control CDI in a manner that is exchanged between Ca_V1 and Ca_V2 isoforms. In Ca_V1.2, the C-lobe governs CDI,⁶⁷ whereas in Ca_V2.1,^68,69 Ca_V2.2,⁷⁰ and Ca_V2.3,⁷⁰ the N-lobe governs CDI. Further, the C-lobe has been shown to control Ca_V2.1 CDF.^68,69 The high level of sequence similarity among the Ca_V1 and Ca_V2 IQ domains (Fig. 4B) and apparent inversion of lobe specific functions, pushed the desire to understand how the same molecule, Ca²⁺/CaM, could affect separate calcium-driven processes using different lobes and yet interact with a highly conserved portion of the pore-forming subunit. This question has motivated a number of recent structural studies of Ca²⁺/CaM complexes of IQ domains from both Ca_V1 and Ca_V2s^71–75 (Table 1). To simplify comparisons in the following discussion, and because the Ca_V IQ sequences have different numbering schemes depending on the organism of origin and particular splice variant used, the IQ domain positions are designated by their relative position in the sequence with respect to the central isoleucine, Ile(0). Negative and positive numbers signify the position of each residue N-terminal or C-terminal to Ile(0), respectively.

Table 1.

Ca_V nomenclature and Ca²⁺/CaM IQ domain complexes

Ca_V1 IQ Complexes

The initial crystallographic studies of Ca²⁺/CaM-Ca_V IQ domain complexes investigated the IQ domains of the L-type channel Ca_V1.2.^72,75 These structures were not only the first descriptions of Ca²⁺/CaM-Ca_V IQ domain complexes, but were also the first description of a Ca²⁺/CaM-IQ domain complex from any source. The structures uncovered two key, unanticipated features. Ca²⁺/CaM wraps around an α-helix formed by the IQ domain in an unusual parallel binding orientation in which Ca²⁺/N-lobe and Ca²⁺/C-lobe bind to sites on the N- and C-terminal ends of the IQ α-helix, respectively (Fig. 4C). Previously, all other examples of Ca²⁺/CaM-peptide complexes, except that of the Ca²⁺/CaM dependent kinase kinase,^76,77 had been found to adopt an antiparallel arrangement with respect to Ca²⁺/CaM and the bound helical peptide. The structures also revealed that each Ca²⁺/CaM lobe interacts with three aromatic residues, termed ‘aromatic anchors’ arranged on opposite faces of the IQ helix. Ca²⁺/N-lobe interacts with Phe1618(−6), Tyr1619(−5) and Phe1622(−2). Ca²⁺/C-lobe interacts with Tyr1627(+3), Phe1628(+4) and Phe1631(+7). Ile1624(0) is completely buried by interactions with Ca²⁺/C-lobe whereas Gln1625(+1) interacts with both lobes. These principal contacts are similar in both X-ray structures (PDB: 2BE6 and 2F3Y). Additionally, Fallon et al. determined the structure of an IQ→AA mutant at positions 0 and +1 (PDB: 2F3Z). This Ca_V1.2 double mutant eliminates CDI and causes strong CDF,⁷⁸ but does not alter the ability of Ca²⁺/CaM to bind the IQ domain.^75,85 The structure of the IQ→AA mutant complex is essentially identical to the wild-type complex and leaves the reason for the strong functional effect of the IQ→AA mutation unresolved.

The interactions between Ca²⁺/N-lobe and the Ca_V1.2 aromatic anchors were unanticipated. Tests of the importance of this interaction by a triple alanine mutation (TripleA) of Phe1618(−6), Tyr1619(−5) and Phe1622(−2), abolished Ca_V1.2 CDF⁷⁵ (Fig. 4D). This result, together with the unexpected parallel binding mode raised the hypothesis that the apparent inversion of lobe-specific function between Ca_V1.2 and Ca_V2.1 originates from an exchange of binding orientations between Ca_V1 and Ca_V2.^12,57

The Ca_V1.1 IQ domain is the most divergent from the Ca_V1 family and has an H.Y change in the (+3) Ca²⁺/C-lobe anchor position (Fig. 4B). Ca_V1.1 acts in skeletal muscle as the voltage-sensing subunit for the intracellular ryanodyne receptor.¹ This function requires a physical link between Ca_V1.1 and the ryanodyne receptor but does not require calcium permeation through Ca_V1.1. Nevertheless, Ca_V1.1 subunits have been shown to have CDI.⁷⁹ The structure of a Ca²⁺/CaM-Ca_V1.1 IQ domain complex (PDB: 2VAY)⁷¹ reveals a parallel binding conformation that is very similar to that observed in the Ca_V1.2 complexes. The structure shows that His1532(+3) occupies the same Ca²⁺/C-lobe binding pocket as Ca_V1.2 Tyr1627(+3) and requires only a minor rearrangement of the sidechains that line the Ca²⁺/C-lobe pocket.⁷¹ Introduction of Y→H(+3) into the Ca_V1.2 IQ domain is sufficient to reduce Ca²⁺/CaM binding and simultaneous introduction of Y(+3)→H(+3) and K(+8)→M(+8) into Ca_V1.2 eliminates CDI.⁸⁰ Although there are no reported structures for Ca²⁺/CaM complexes with Ca_V1.3 or Ca_V1.4 IQ domains, the extremely high sequence conservation among Ca_V1 IQ domains (Fig. 4B) that includes the aromatic anchors, strongly suggests that the parallel binding mode will be found throughout the Ca_V1 family.

Ca_V2 IQ Complexes

Structure determination of Ca²⁺/CaM-Ca_V IQ domain complexes from Ca_V2.1, Ca_V2.2 and Ca_V2.3 (PDB: 3DVM, 3DVE and 3DVK, respectively), revealed that unlike the Ca_V1.2 and Ca_V1.1 Ca²⁺/CaM-IQ domain complexes, all three Ca_V2 IQ domain complexes are antiparallel (Fig. 4E). Rather than just a simple exchange of lobe binding positions, Ca²⁺/CaM is also situated further towards the N-terminal end of the IQ-helix. Consequently, the two lobes are anchored by different constellations of residues than are used in the Ca_V1.1 and Ca_V1.2 complexes. Ca²⁺/N-lobe interacts with four main residues: methionine (-1), isoleucine (0) and the aromatics at positions (+3) and (+5). This positional shift is consistent with the loss of the aromatic anchor at the (+7) position, which is a small hydrophilic residue in all of the Ca_V2 IQ domains (Fig. 4B).¹² Ca²⁺/C-lobe, which is bound to the N-terminal end of the IQ helix, is anchored by two major positions in the Ca_V2.1 and Ca_V2.3 complexes, positions (−6) and (−2) and has additional contacts with position (−9). The Ca_V2.1 and Ca_V2.3 complexes are essentially identical. The Ca_V2.2 complex has some structural differences that involve a slightly altered IQ helix pose and a relative displacement of the Ca²⁺/C-lobe towards the center of the IQ helix.

In both the parallel Ca_V1 and antiparallel Ca_V2 Ca²⁺/CaM IQ domain complexes, each Ca²⁺/CaM lobe interacts with an extensive network of IQ domain residues. Although there are differences in the details of the interactions, one common element is that the occupant of the N-terminal binding site, Ca²⁺/N-lobe for Ca_V1s and Ca²⁺/C-lobe for Ca_V2s, buries much less surface area than the occupant of the C-terminal binding site (Ca²⁺/N-lobe vs. Ca²⁺/C-lobe: Ca_V1.1 1,209 Ų, 1,630 γ²; Ca_V1.2 1,450 Ų, 1,819 Ų; Ca_V2.1 1,652 Ų, 921 Ų; Ca_V2.2 2026 Ų, 1,224 Ų; Ca_V2.3 1,673 Ų, 950 Ų; respectively).

In addition to the above structures, complexes of Ca²⁺/CaM with Ca_V2.1 and Ca_V2.3 IQ peptides shorter than those described above, (21 vs. 25 residues, Table 1) have also been determined (PDB: 3BXK and 3BXL, respectively).⁷⁴ Strikingly, these complexes are not antiparallel, but are parallel structures very similar to the Ca_V1.2 complex determined using the short IQ peptide reported by Fallon et al.⁷² The ability of Ca²⁺/CaM to bind similar peptide sequences in opposite orientations is remarkable. It should be noted, however, that the peptides in these parallel structures lack four residues that contribute to interactions with Ca²⁺/C-lobe in the antiparallel complexes, including residue (−9). Further, adoption of the antiparallel pose on the short peptide would place the positively charged IQ peptide N-terminus near the Ca²⁺/C-lobe hydrophobic pocket.

Structure-based alanine mutations designed to test the importance of the Ca²⁺/C-lobe contacts from the antiparallel orientation, show that loss of these interactions eliminates binding of Ca²⁺/C-lobe to the IQ domain in titration calorimetry experiments and greatly reduces CDF,⁷³ (Fig. 4F). This direct connection between structure, biochemistry and function suggests a clear structural basis for the swap of lobe-specific roles (Fig. 5). These data indicate that the Ca_V1 and Ca_V2 IQ domains contain dedicated sites that control CDF and that the occupant of these sites is exchanged between Ca_V1 and Ca_V2 isoforms (Fig. 5B). The observed inversion of structural polarity provides an attractive explanation for the inversion of lobe-specific roles (however, for alternative scenarios consult ref. 74). The observation that the occupant of the N-terminal site in both Ca_V1 and Ca_V2 IQ domains makes fewer contacts and binds more weakly^73,75 suggests that these physical properties are also important for the CDF mechanism. Nevertheless, the IQ domains are distant from the pore-forming subunit and it remains unclear as to how binding events in the IQ domain are coupled to changes in the behavior of the pore. Structural determination of larger portions of the C-terminal tail and clarification of the structural basis of apo-CaM binding, which is a key state in the reaction cycle that drives calcium-dependent channel modulation, will be required to gain further insight into this question.

Figure 5.

(A) Lobe-specific roles in Ca_V1.2 and Ca_V2.1 CDI and CDF. (B) Diagrams of the orientations of Ca²⁺/CaM on the Ca_V1 and Ca_V2 IQ domains and the roles of the individual lobes based on references 73 and 75.

Multiple Ca²⁺/CaMs Can Bind the C-Terminal Tail but do not Mediate Channel Dimerization

Two recent studies have succeeded at obtaining structural information about larger portions of the Ca_V1.2 C-terminal tail that contain the PreIQ region and the IQ domain^81,82 (Figs. 1B and 4A). The PreIQ region contains two Ca²⁺/CaM binding sites known as the A-region and C-region.^83,84 Surprisingly, both studies found structures that contained crystallographic dimers made from interactions between the Ca²⁺/CaM complexes. In one,⁸² the components were well enough resolved to show two C-terminal tails and four bound Ca²⁺/CaMs (Fig. 6A). The Ca²⁺/CaM-IQ domain structures are essentially identical to those solved using the individual Ca_V1.2 IQ domains and indicate that the parallel orientation is maintained in the context of a substantially longer portion of the Ca_V1.2 C-terminal tail. The PreIQ region forms a long helix (42 residues) that connects to the Ca²⁺/CaM-IQ complex by a flexible linker. In the asymmetric unit of the crystal, the two PreIQ helices interact via crossbridges formed by the binding of Ca²⁺/C-lobe to one PreIQ C-region and Ca²⁺/N-lobe to the A-region of the adjacent PreIQ segment as well as through helix-helix interactions that resemble five turns of an anti-parallel coiled coil.

Figure 6.

Structure and characterization of the Ca²⁺/CaM-PreIQ-IQ domain complex. (A) Two Ca_V1.2 PreIQ helices (red and wheat) form a crystallographic dimer cross-bridged by Ca²⁺/CaMs. Ca²⁺/CaM N-lobe and C-lobe are colored green and blue, respectively. N- and C-termini of PreIQ-IQ domain are indicated.⁸³ (B) Sedimentation equilibrium analysis of 100 µM Ca²⁺/CaM-Ca_V1.2 PreIQ-IQ complex at 11,000 rpm and 4°C and measured at 293 nm. Raw data (black open circles) and single species fit (black line) are compared to predicted curves for complexes having Ca²⁺/CaM:Ca_V1.2 tail ratios of 1:1 (red), 2:1 (yellow), 2:2 (green) and 4:2 (blue). Inset shows the distribution of residuals as a function of radial distance. Reprinted from Kim et al.⁸² (C) Cartoon diagram of the 2:1 Ca²⁺/CaM-Ca_V1.2 PreIQ-IQ complex. C-lobe_C indicates the Ca²⁺/CaM lobe bound to the C-region site. (D) TIRF image of Ca_V1.2-GFP in a live cell membrane of a X. laevis oocyte. Blue circles mark the selected spots used for subunit counting. Inset shows distribution of fluorescent spots that bleach in one or more steps. Each bleaching step indicates one Ca_Vα₁ subunit. Reprinted from Kim EY et al.⁸² (E) Model depicting the possible relative positions of Ca_V intracellular elements based on Figure 3A. “??” signifies the potential for a Ca²⁺/CaM anchored at the C-region site to make bridging interactions with other components of the Ca_V complex. Relative orientation of the C-terminal tail is chosen to display the key elements. Its orientation relative to the other intracellular components is not known. The N-terminal cytoplasmic domain, II–III loop and III–IV loop are not depicted.

Although recent studies have suggested coupled activity among Ca_V1.2s,⁸⁵ Ca_Vs are generally not thought to function as dimers. Based on the crystallographic dimer and a modest effect on CDI resulting from an E→P change in the PreIQ region, Fallon et al. proposed that Ca_V1.2s dimerize through a functionally relevant anti-parallel coiled-coil interaction augmented by bridging interactions between Ca²⁺/CaM and A-region- and C-region-binding sites.⁸¹ Discernment of oligomerization state from crystallographic contacts alone is difficult^86–88 and requires some alternative method for validating whether the observed oligomerization is robust or only a consequence of the ordered array of the crystallographic lattice. Experiments performed using a Superdex 75 size exclusion column were presented as evidence that the crystallized complex could dimerize in solution.⁸¹ However, it is worth noting that the size of the dimeric complex (85.6 kD) that forms the asymmetric unit composed of a 4:2 Ca²⁺/CaM:PreIQ-IQ complex is beyond the resolution limit of the column (3–70 kDa). Size exclusion experiments using a Superdex 200 column (resolution range 10–600 kDa), on which, in principle, 2:1 and 4:2 complexes could be resolved (42.8 kDa vs. 85.6 kD) yielded an ambiguous assessment of molecular weight for the complex (∼66 kDa).⁸² This ambiguity persisted regardless of whether the construct contained the full PreIQ-IQ segment, or a version that could not dimerize because it lacked one of the binding sites required to form the PreIQ crossbridge.⁸² These results suggested that elongated nature of the complex was a likely confounding factor for size determination by this methodology, which relies on comparison to globular proteins standards.²⁷ As analytical ultracentrifugation studies are insensitive to molecular shape,^27,89 this technique was used to measure the molecular mass of the complex in solution unambiguously. The results demonstrate that the Ca²⁺/CaM:PreIQ-IQ complex has a 2:1 stoichiometry (Fig. 6B and C).⁸² This complex shows no evidence of dimerization, even at concentrations that are on the order of that experienced by Ca_Vs in the most densely packed environment known (∼100 µM), that of the SR junction.⁹⁰

Biochemical studies of the full-length channel in cardiac microsomes were interpreted as evidence for channel dimers;⁸¹ however, these studies do not account for the contribution of the detergent to the apparent molecular weight of the protein-detergent complex and were done in the apparent absence of reducing agents. In contrast, single molecule subunit counting experiments of Ca_V1.2 channels in live cell membranes unequivocally demonstrate that the full-length channels are monomers (Fig. 6D).⁸²

Two types of mutational studies were used to test the importance of the PreIQ helix. Simultaneous alanine mutations of two residues that form the core of the asymmetric unit dimer interface, and that would be expected to preserve the helical structure while being very detrimental to the PreIQ helix interaction,^91,92 had no functional effects on VDI, CDI or CDF.⁸² The PreIQ E→P change⁸¹ had modest effects on both VDI and CDI but is at a residue that is on the opposite side of the PreIQ helix crystallographic interface. Although the proline substitution could affect the PreIQ helix structure, this change also removes a charge and thus, could produce the modest functional changes by perturbing interactions of the PreIQ helix with other channel components. Further, Ca_V1.2 channels bearing the PreIQ-IQ domain but lacking the remainder of the Ca_V C-terminal tail show no evidence of coupled activity.⁸⁵ Thus, the weight of evidence indicates that the crossbridged, dimeric structure is neither robust nor relevant for function, and is simply a consequence of crystallization.

The demonstration that the Ca²⁺/CaM:PreIQ-IQ complex has a 2:1 stoichiometry⁸² shows that multiple Ca²⁺/CaMs can bind the Ca_V1.2 C-terminal tail simultaneously (Fig. 6C), a result that is corroborated by recent independent biochemical studies.⁹³ This property is likely to be conserved among Ca_V1s and Ca_V2s as the residue central to the Ca²⁺/CaM:PreIQ C-region interaction, Trp1593, is present from mammals through the sea squirt, Ciona intestinalis. The binding of multiple Ca²⁺/CaMs along a largely helical segment is reminiscent of how CaM and CaM-like proteins are organized on myosin94 and may indicate a previously unrecognized connection between these classes of CaM-modulated proteins.

Investigation of the biochemical properties of the 2:1 Ca²⁺/CaM:PreIQ-IQ complex showed that the two bound Ca²⁺/CaMs have different properties. The one bound to the PreIQ C-region is labile, whereas the one bound to the IQ domain is not.⁸² Further, the C-region has a strong preference for binding Ca²⁺/lobes over apo-lobes. Electrophysiological experiments show that the PreIQ C-region has a previously unrecognized role in CDF.⁸² Together, the data suggest that the PreIQ region forms a long α-helix that contains a Ca²⁺/C-lobe binding site that could be used as an anchor for making bridging interactions to other channel elements via the Ca²⁺/N-lobe (Fig. 6E).⁸²

The idea that Ca²⁺/CaM could form bridges between channel elements, such as the N-terminal and C-terminal cytoplasmic domains,⁹⁵ is attractive; however, biochemical support for such interactions remains elusive.⁹⁶ The presence of two long α-helices in the Ca_Vα₁ cytoplasmic domains, the IS6-AID helix^46–48 and the PreIQ helix could position the two elements that have functional crosstalk, the Ca_Vβ/I–II loop and the Ca²⁺/CaM:PreIQ-IQ complexes, near each other (Fig. 6E). The intriguing possibilities raised by this possible convergence of key modulatory protein-protein complexes highlights the pressing need for further biochemical and structural work to identify how the various intracellular elements interact in both apo-CaM and Ca²⁺/CaM bound states and how the structural changes that underlie VDI, CDI and CDF are communicated with the pore.

Other elements from the intracellular regions of the channel are likely to have some role in both CDI and CDF. Functional data point towards potential roles for interaction of the Ca²⁺/CaM-IQ domain with the putative EF hand proximal to transmembrane segment IV6,^74,97 elements in the N-terminal^95,98,99 and the C-terminal Ca_Vα₁ cytoplasmic domains,^100–104 and the Ca_Vβ subunit.^47,105,106 Structural definition of these channel components and their interactions is an important near-term goal. Additionally, the large dodecameric protein kinase, CaMKII, has been shown to be a key component for both Ca_V1.2^107–111 and Ca_V2.1¹¹² CDF. Exactly how this large, multiprotein complex engages the channel remains unknown.

Progress on Ca_V Modulator Structures

Besides the work on conventional channel elements, the past six years have produced advances in structural studies of two channel modulators that have dramatic effects on Ca_V function. These studies form the starting point for understanding how Ca_V activity can be customized to particular cellular contexts.

CaBP1, Inhibitor of Ca_V1 CDI

In cerebellar and hippocampal neurons,^113,114 photoreceptor synapses,¹¹⁵ and auditory hair cells,^116,117 the composition of Ca_V1 complexes is changed in a way profoundly effects channel activity. Members from a family of calcium binding proteins homologous to CaM, known as CaBPs,¹¹⁸ can replace CaM on the IQ domain^114,116 and yield channels that have strikingly different functional properties than those modulated by CaM.^{113,114,116,117,119–121} CaBP1 (Fig. 7A), a CaBP that is abundant in the brain and retina,¹¹⁸ blocks CDI in Ca_V1.2,^114,119 (Fig. 7B) and Ca_V1.3,^116,117 and introduces CDF in Ca_V1.2.¹¹⁴ There are a number of features that make CaBP1 different from CaM. CaBP1 has a longer linker between the lobes and contains an N-terminal myristoylation site. As in CaM, the CaBP1 C-lobe EF-hands undergo calcium induced conformational changes.¹²² However, CaBP1 N-lobe appears insensitive to calcium due to substitutions in key calcium binding positions of EF2 (Fig. 7A) and weak calcium affinity of EF1 (Kd >100 µM).¹²² Exactly how these differences allow CaBP1 to change channel function has been unknown.

Figure 7.

CaBP1 structure and function. (A) Sequence comparison of CaBP1 and CaM. Locations of EF-hands and canonical Ca²⁺-binding positions (x, y, z, −y, −x and −z),⁶⁴ N-terminal domain (NT), linker (L) and N- and C-terminal lobes (green and blue, respectively) are shown. Cartoons indicate the loss of at least one Ca²⁺ binding site in CaBP1 due to substitutions in the canonical EF-hand positions. CaBP1 E94 is colored purple. (B) Comparison of the functional effects on CDI of CaM, CaBP1 and the CaBP1 E94A mutant. Data show normalized calcium currents of Ca_V1.2 expressed in Xenopus oocytes. Reprinted from Findeisen et al.¹²⁴ (C) Cartoon representation of the CaBP1 structure. N-lobe (green), linker (red) and C-lobe (blue) and E94, which is shown in stick representation, are indicated. Calcium ions are depicted as white spheres.

NMR studies show that similar to CaM, CaBP1 has two, independently folded lobes.¹²³ Recent X-ray crystallographic studies,¹²⁴ reveal an unanticipated interaction between a conserved glutamate, E94, in the interlobe linker and the CaBP1 N-lobe (Fig. 7C). Combined use of CaM/CaBP1 chimeras, isothermal titration calorimetry and functional studies show that this interaction is essential for the ability of CaBP1 to inhibit Ca_V1.2 CDI. The studies further demonstrate that the two CaBP1 lobes have two separable functions. The CaBP1 C-lobe serves as a high-affinity anchor that binds the Ca_V1.2 IQ domain at a site that overlaps with the Ca²⁺/CaM C-lobe site, whereas the N-lobe/linker module bears the elements required for channel modulation. Strikingly, neither the N-terminal myristoylation nor the ability of the CaBP1 lobes to bind calcium is necessary for CaBP1 to block CDI.¹²⁴

Although it is clear that the CaM and CaBP1 binding sites on the IQ domain overlap, how exactly CaBP1 blocks CDI remains unknown. This dramatic functional change likely involves interactions between the CaBP1 N-terminal domain and some other portion of the channel. One candidate site is the Ca_Vα₁ N-terminal cytoplasmic domain,¹¹⁹ but other yet to be defined elements may also be involved. Determining how exactly CaBP1 binds the IQ domain and which other portions of the channel it engages remain important open questions for future structural studies.

Kir/GEM GTPases—silencers of Ca_V function.

The RGK family of Ras-related small G-proteins has four members (Rad, Rem, Rem2 and Gem/Kir) that are capable of potently inhibiting Ca_V function in a manner that depends on the presence of Ca_Vβ^14,125–127 (Fig. 8A). Recent reports suggest that at least three mechanisms may be used:¹²⁸ (1) enhancement of dynamin-mediated endocytosis, (2) diminution of open probability without changes in voltage sensor movement and (3) voltage-sensor immobilization. The exact mechanisms of action remain a topic of extensive debate and investigation.¹⁴ Although channel modulation by RGKs requires Ca_Vβ, studies of polyglycine substitutions in the IS6-AID segment¹²⁹ rule out a mechanism that is related to how Ca_Vβ affects VDI,^47,48 CDI,⁴⁷ and voltage-dependent inhibition by Gβγ.⁴⁸

Figure 8.

RGK functional effects and Gem G-domain structure. (A) Suppression of Ca_V1.2 function by co-expression of Kir/Gem. Redrawn from Figure 2 from Beguin et al.¹²⁶ (B) Cartoon depiction of the Gem G-domain structure.¹³² Switch regions and residues implicated in Ca_Vβ interaction¹³³,¹³⁴ are indicated. Bound GDP (cyan and orange) and Mg²⁺ (yellow sphere) are also shown.

RGK proteins have a core G-domain, that binds GDP and GTP130 and that is flanked by N-and C-terminal domains of 44–88 and 29–37 residues, respectively. Structural data for the RGK family of channel modulators is limited to Mg²⁺•GDP complexes of the G-domains of Rad¹³¹ and Gem,¹³² and an Mg²⁺•GDP complex of a Gem construct bearing the G-domain and part of the C-terminal domain.¹³⁰ Compared to canonical Ras-GTPases, the RGK family has unusual sequences in two elements, Switch I and Switch II, which are important for G-protein activation and inactivation. These regions show substantial structural differences from Ras. Switch I is either disordered^130,131 or in a widely different orientation compared to Ras¹³² (Fig. 8B) and leads to a very high solvent exposure of the bound GDP. Structure-based alanine scanning identified three residues on one face of Gem that reduce binding to Ca_Vβ (Fig. 8B).^133,134 Surprisingly, a triple alanine mutant at these sites did not block Gem modulatory activity.¹³⁴ Thus, the importance of the Gem-Ca_Vβ interaction for channel modulation remains unclear. Acquisition of additional structural and biochemical data regarding the nature of RGK-Ca_Vβ interactions and nucleotide-hydrolysis driven changes should help clarify the basic mechanisms of action. Further, as engineered RGK-based Ca_V modulators offer a novel way to affect Ca_V function,¹²⁹ structural characterization of the RGK-Ca_Vβ interaction should aid protein design based efforts to make new generations of Ca_V modulators.¹³⁵

Prospectus

The protein dissection approach has proved illuminating for the initial studies of Ca_V structure. Given the power of this approach, particularly when integrated with biochemical and functional studies, such studies are likely to continue to drive the next set of advances.

Although there is good evidence supporting the model that Ca_Vβ influences the pore forming domain via a long helix, it is less clear how structural transitions of CaM and the C-terminal tail are coupled to the pore. There is a growing body of functional evidence that multiple regions of the channel interact to bring about the two processes that dominate voltage-gated calcium channel (Ca_V) inactivation, VDI^105,136 and CDI.^57,58,105 The simplest hypothesis is that, at least in one functional state, there is a direct physical interaction between the major VDI and CDI components, the Caβ/Ca_Vα₁-I–II loop and CaM-Ca_Vα₁ C-terminal tail complexes.⁴⁷ Interactions of these components with the N-terminal cytoplasmic domain may also be important for both CaM^95,98,99 and CaBP1 modulation.¹¹⁹ Nevertheless, robust evidence for such direct physical interactions remains elusive. Given the substantial amount of Ca_V mass in the cytoplasm, ∼150 kDa, it seems likely that interactions between the Ca_Vβ/Ca_Vα₁-I–II loop and CaM-Ca_Vα₁-C-terminal tail complexes may involve other yet to be determined points of contact. It will be important to determine structures of Ca²⁺/CaM bound to larger portions of the Ca_V1 and Ca_V2 tails to place the IQ domain conformations in a larger structural context. Further, the calcium-free state of CaM, apo-CaM, binds the Ca_V C-terminal tails of both L-type and non-L-type Ca_Vs^{67,84,137,138} and is thought to be essential for the ability of Ca_Vs to decode different types of calcium signals;⁹⁹ however, the exact apo-CaM binding site and interaction details remain controversial. Definition of the structural basis for apo-CaM interactions with Ca_V1s and Ca_V2s remains a major missing piece that will be required for understanding the structural transitions that underlie CDI and CDF.

Among the subunits, Ca_Vα₂δ remains the most mysterious. It is a target for drugs such as gabapentin (Neurontin) and pregabalin (Lyrica) that are used to treat epileptic seizures, pain from diabetic neuropathy, post-herpetic neuralgia, fibromyalgia and generalized anxiety disorder.^139,140 Defining the structural basis for the interaction of these drugs with Ca_Vα₂δ should provide new ideas regarding their mechanisms of action and also the general role of Ca_Vα₂δ in Ca_V function. Because it is a large, membrane-anchored glycoprotein, Ca_Vα₂δ poses a number of extra challenges for structure determination efforts.

There is no doubt that some questions about Ca_V function can only be answered by obtaining structural data on holo-channels in different states of the functional cycle (open, closed, inactivated, etc.,) and bound to compounds from the rich and the well-characterized Ca_V-directed pharmacology toolkit.^141,142 As the association of multiple channel subunits is necessary to make a properly functioning high-voltage activated Ca_Vs, effort on the structure of Ca_V1 and Ca_V2 holo-channels should benefit from the ongoing development of technologies aimed at producing full length, multiprotein membrane protein complexes.^143,144 In this regard, ability of Ca_V3s to function alone, may make them better candidates for initial work in full-length channels. The ability to produce large quantities of Ca_V holo-channels of defined composition or bearing well-studied mutations would not only provide a great advance for structural work, but would also allow the pursuit of biochemical and reconstitution experiments that could directly address questions about interdomain interactions and structural transitions. In the meantime, further studies that exploit the ‘divide-and-conquer’ approach are likely to be productive at yielding new insights and posing new questions. Such studies will be most revealing when combined with biochemical validation and functional studies of structure-based hypotheses in the full-length channels.