Three-Dimensional Architecture of the L-Type Calcium Channel: Structural Insights into the CaVα2δ1 Auxiliary Protein

L-type calcium channels (LTCC) are responsible for Ca2+ influx into muscle and neurons. These macromolecular complexes minimally comprise the main poreforming CaVα1 and auxiliary subunits CaVβ and CaVα2δ1. The ultrastructure of the oligomeric LTCC complexes from heart and skeletal muscle has been reported previously at ≈ 20 Å, a resolution that prevent identification of structural domains. Recent improvements in cryo-electronic microscopy (EM) methods made it possible to obtain a three-dimensional structure of the rabbit skeletal muscle LTCC CaV1.1 complex at a resolution of 4.2 Å and recently at 3.6 Å. This technique requires only nanograms of purified proteins and circumvents crystallization as a means for structure determination. The high resolution cryo-EM structure shows the molecular architecture of the subunits comprising the oligomeric complex and for the first time, a high-resolution glance of the largely extracellular CaVα2δ1 protein with its extracellular domains (Cache1, VWA, and Cache2). Although the CaVα2δ1 protein is a single-pass transmembrane protein, the complex topology of its extracellular domain represents a technical challenge for structure determination using conventional purification approaches. Herein we show the merits of a strategy based upon the purification of small structural domains that can be elucidated individually before these domains are reassembled into the quaternary structure. A structural model was derived using ab initio structure prediction constrained by small angle X-ray scattering profile of the refolded Cache2 domain. The excellent agreement between the predicted structure and the available cryo-EM structure suggests a novel and rapid procedure to discover structural information of protein domains.


Introduction
L-type calcium channels (LTCC) form a large family of structurally related channels expression in skeletal muscle (Ca V 1.1), working myocardium (Ca V 1.2), neuroendocrine cells (Ca V 1.3), and the retina (Ca V 1.4). In cardiac cells, calcium ions entering into the cell through Ca V 1.2 channels during the plateau phase of the action potential are essential to initiate the excitation-contraction coupling [1,2]. Together with voltage-gated sodium and potassium channels, Ca V 1.2 contributes to the heart rhythm and its activity can be derived from the measure of the QT interval on the electro cardiogram [3]. Gain-of-function and loss-of -function genetic mutations of sodium and potassium channels have been associated with many forms of cardiac arrhythmias [4]. To a smaller extent, mutations in the genes encoding for Ca V 1.2 channels have been associated with Timothy, Brugada, and early This article is available in: http://biochem-molbio.imedpub.com/archive.php

Biochemistry & Molecular Biology Journal ISSN 2471-8084
after depolarization syndromes [5][6][7]. Some of these cardiac dysfunctions are characterized by an increase in the QT interval, whereas others are manifested by a shorter QT interval and an elevated ST segment [4,8].
In cardiomyocytes, Ca V 1.2 is an oligomer consisting of a main pore-forming Ca V α1 (≈ 250 kDa) and additional auxiliary subunits: Ca V β2 (≈ 55 kDa) and Ca V α2δ1 (≈ 150 kDa). The Ca V α1 subunit confers the biophysical and pharmacological properties of the channel and is the molecular target of the class IV antiarrhythmic drugs, among which dihydropyridine, verapamil, and diltiazem compounds are most widely used. This subunit is essential and targeted disruption of Ca V α1 is embryonically lethal in mice By contrast, there is little structural information on Ca V α2δ1. The reason can be found in the complexity of the Ca V α2δ1 protein topology that results from multiple co-and post-translational modifications including the addition of N-glycans at 16 Asn sites that is required for the folding and stability of Ca V α2δ1 [16,17]. Furthermore, Ca V α2δ1 is encoded by a single gene and is posttranslationnally cleaved into the large extracellular Ca V α2 and the putative transmembrane Ca V δ proteins bound by disulfide bridges [18][19][20]. In fact, the rat Ca V α2δ1 protein includes 20 cysteine residues and it has been proposed that intra-molecular disulfide bonds are required to stabilize its higher order structure [21]. These features represent significant hurdles for expressing and purifying the protein complex. in bacterial systems and account for the limited structural information on eukaryotic LTCC channels

Three-Dimensional Structure of the Mammalian Ca V α2δ1 Proteins
Structural data on the pore-forming subunit Ca V α1 has been mostly inferred from the high-resolution crystal structures (≈ 2.7 -3.1 Å) of bacterial homologs of voltage-gated sodium channels and a modified variant referred to as Ca 2+ -selective Ca V Ab channels [22][23][24]. Unlike their mammalian homologues, the bacterial channels form symmetrical channels with 4 identical subunits forming the pore region and appear to be functional without specific auxiliary subunits. Low-resolution electron microscopic (EM) models of the LTCC complex purified from skeletal muscle at a concentration of 90-140 µg LTCC complex for 400 g of skeletal muscle were first solved at 27 Å in 2004 [25]. The relatively low resolution provided a general outline of the protein complex and confirmed that Ca V α2 was mostly extracellular. More recently improvements in the electron detection and image processing algorithms made it possible to reconstruct the 3-D structure of the endogenous Ca V 1.1 channel complex from rabbit skeletal muscle membranes [17] without the need to obtain crystals. Instead of photoaffinitylabeling the protein complex with radioactive dihydropyridine receptor ligands [26], the authors chose the brilliant strategy to overexpress the cytosolic Ca V β1a subunit as a fusion protein in a bacterial system and use it as a bait to pull down with nanomolar affinity the entire LTCC complex [17,27]. Aliquots (4 µl) of the digitonin-purified Ca v 1.1 complex at 0.1 mg/ml were examined by cryo-EM and more than 10 6 particles were selected for further analysis. A three-dimensional structure of the LTCC complex with dimension of 170 x 100 Å was obtained at 4.2 Å [17] and then at 3.6 Å in the presence of 10 mM Ca 2+ [27]. The electron density map revealed the position of the three subunits (Ca V β1a, Ca V γ1, and Ca V α2δ1) in relation to the pore-forming Ca V α1 subunit of the Ca V 1.1 complex. In particular, the 3-D structure demonstrates for the first time the position of the transmembrane Ca V γ1 protein, which is the major isoform expressed in the skeletal muscle [28]. More importantly, the authors provide the first description of the extracellular structural domains within Ca V α2δ1 and their position relative to Ca V α1.
A BLASTP search conducted with the "conserved domain" tool in NCBI [29] revealed four structural domains in the extracellular region of the rabbit (NP_001075745.1) and the rat (NP_037051.2) Ca V α2δ1 proteins: i.e., VWA-N, VWA, Cache2, and VGCC, the latter being included in the Cache2 domain identified in the cryo-EM structure of the rabbit Ca V α2δ1 (Figure 1). The VWA domain, believed to be the molecular target of anti-epileptic drugs of the gabapentin family [30], appears to be positioned just above the voltage sensor of the Ca V α1 subunit suggesting that Ca V α2δ1 could modulate the channel function by stabilizing the channel voltage sensor domain [31]. There was however insufficient electron density to support amino acid assignment for the region between amino acids 627 to 950 within the unstructured C-terminal domain of Ca V α2 as well as for the Ca V δ transmembrane domain between 1065 and 1106 in the first structure at 4.2 Å. Unfortunately, this relatively low resolution precluded assignment of the side chains even in regions where the backbone has been solved confidently such as in the N-terminal of Ca V α2 (between 40 and 77 and between 112 and 178).
The Cache domains face the extracellular environment and projects approximately 60 Å away from the membrane where it could anchor an extracellular networking hub for LTCC. Structural information could help identify crucial partners for protein trafficking and/or function [32]. of 11 residues with the former protein which accounts in part for the lower sequence identity. The online server I-TASSER [34][35][36] nonetheless confirmed these two proteins as the most appropriate templates for Cache2 domain of Ca V α2δ1 suggesting that these proteins may share similar folding patterns.
The structural complexity of Ca V α2δ (disulfide bonds, multiple glycosylation sites, and one transmembrane domain) makes purification of the whole protein in a bacterial system quite challenging. To bypass these limitations, we implemented the "Divide and Conquer" approach [37] whereby the Cache2 domain, one of the basic building units of Ca V α2δ1 [38], was cloned into a bacterial expression vector and purified for structure determination.

SAXS Structure of the Cache2 Domain of Ca V α2δ1
The primary sequence of the rat Cache2 domains (between residues 446 and 636) display 93% identity with the rabbit isoform (between residues 448 and 651) ( Figure 1). However, the rabbit Cache2 domain in the cryo-EM structure includes residues numbered 76 to 91 that are absent in our protein [38]. We overexpressed the Cache2 protein (rat residues 446-636) as a 25-kDa (HIS)6-tagged protein in E. coli BL21(DE3)pLYS. The protein was purified using a three-step purification procedure including affinity, DEAE and size exclusion chromatography. Most of the expressed protein found in inclusion bodies, was solubilized using 8 M urea and purified under denaturing conditions by immobilized metal ion affinity chromatography (IMAC) using a nickel resin followed by diethylaminoethanol anion-exchange chromatography. Refolding was achieved by removing urea by overnight dialysis at 4°C in the presence of 0.5 M of L-arginine, a stabilizing agent [39] and 1 mM DTT [40]. The Cache2 protein eluted as a single symmetric peak at 15.9 ml thus having an apparent molecular weight of 21 kDa according to our gel filtration calibrations. This validated the monomeric state of the purified Cache2 protein. The secondary structure of the purified domain contained 8% alpha helix and 41% beta strands. The protein was stable at 4°C at a pH of 7.4 for up to 2 weeks although N-glycosylation of eukaryotic proteins is generally not adequately performed in bacteria [41].
Structural characterization of the rat Cache2 protein was carried out using SAXS (small angle X-ray scattering) to assess the authenticity of the structure of the refolded protein. SAXS is a powerful method that provides structural information of proteins in solution [42], and requires small volumes (20-50 μl) of sample at relatively low protein concentration (0.1-1 mg/ml). SAXS provides information as to the folded/unfolded state of a protein, its aggregation, flexible domains, oligomeric state, shape and limited conformational data without any mass limitation [43]. When combined with biochemical knowledge and/or known atomic structures of component domains, SAXS provides an overall solution structure using flexible linkers to connect the structural domains [44] thus delivering a first approximation of the molecular shape, protein assembly and structural dynamics of biological macromolecules in their native state [44][45][46]. The program SAXSTER [47], an on-line service of the I-TASSER server, was used to build an ab initio model structure using the amino acid sequence of the rat Cache2 domain and was combined with the SAXS data as a constraint. The resulting model was then assessed against the coordinates of the Cache2 domain identified in the Ca V α2δ ΕΜ structure. SAXS data were collected on the refolded rat Cache2 protein at protein concentrations of 3.2, 1.6, 0.8 and 0.4 mg /ml (Figure 2). The data collected at 3.2 mg /ml showed evidence of protein-protein interactions at very small angles and were thus excluded from further analysis. The Guinier plots (lnI(q) versus q 2 ) were linear for the remaining concentrations at very small scattering angle (q*R g <1.3) indicative of sample monodispersity. The ab initio molecular envelopes of the rat Cache2 protein were reconstructed and averaged in DAMMIN [48]. The resulting envelope was bean-shaped and slightly asymmetrical (Figure 2) with dimensions of 71× 51× 20 Å corresponding to the D max of 70 Å obtained from the pair distance probability plot [49]. The atomic coordinates of the rabbit Cache2 domain (rabbit amino acids from 448 to 651) were superimposed onto the experimental envelope using SUPCOMB program. CRYSOL was then used to calculate the solution scattering of the atomic structure of the rabbit Cache2 which was then used to fit our experimental SAXS data [50]. The discrepancy was evaluated with a chi-square value of 1.14 ( Figure 3A). The analysis was performed using a Cache2 model from SAXSTER and when superimposed onto the experimental envelope, an improved chisquare value of 1.05 as illustrated by comparing panels A and B in Figure 3.
The refolded rat structure is shown to superimpose well with the rabbit EM with a Root Mean Square Deviation for C-alpha carbons = 1.73 Å (Figure 4). In particular, the structured N-terminal β-strands 1 and 2 as well as the α-helix 1 and 2 located downstream display the best fits. The "cores" of the rabbit Cache2 and the rat Cache2 proteins are identical. Structural differences occur in the intervening loop between the amino acids 532 and 551 of the rabbit Cache2 and the rat Cache2 proteins presumably due to differences in the primary sequence between the two isoforms and due to seven additional amino acids in the primary sequence of the rat Cache2 a C-terminus a TEV cleavage site, and 6 histidine residues. The SAXS envelope from the rat Cache2 refolded protein was also well fitted by the molecular coordinates of the KinD/methyl-accepting chemotaxis protein and consistent with an ambiguous score of 2.3 obtained when performing the AMBIMETER calculation [51].

Conclusion
Ca V α2δ1 is an integrin-like protein that belongs to the LTCC complex. It promotes LTCC activation and as a result it enhances heart contractility [11]. Ca V α2δ1 undergoes many co-and posttranslational modifications that create a sizable challenge for its purification. In this short commentary, we have shown that the Cache2 domain of Ca V α2δ1 can be purified and refolded from bacterial cultures at a yield of 2 mg per liter. SAXS data measured for the refolded protein enabled an ab initio prediction of a model structural whose fold was identical to the native state. The refolded Cache2 protein conserved the overall folding of the Cache2 protein purified from rabbit skeletal muscle, even though Ca V α2δ1 exists mostly as a large extracellular domain loosely organized around multiple β-sheets [52]. Altogether, this validates an experimental strategy based upon the purification of isolated domains of LTCC subunits in a bacterial system and paves the way for implementing a "building block" approach in studying the structural biology of complex membrane proteins.