A three-way inter-molecular network accounts for the CaVα2δ1-induced functional modulation of the pore-forming CaV1.2 subunit

L-type CaV1.2 channels are essential for the excitation–contraction coupling in cardiomyocytes and are hetero-oligomers of a pore-forming CaVα1C assembled with CaVβ and CaVα2δ1 subunits. A direct interaction between CaVα2δ1 and Asp-181 in the first extracellular loop of CaVα1 reproduces the native properties of the channel. A 3D model of the von Willebrand factor type A (VWA) domain of CaVα2δ1 complexed with the voltage sensor domain of CaVα1C suggests that Ser-261 and Ser-263 residues in the metal ion–dependent adhesion site (MIDAS) motif are determinant in this interaction, but this hypothesis is untested. Here, coimmunoprecipitation assays and patch-clamp experiments of single-substitution variants revealed that CaVα2δ1 Asp-259 and Ser-261 are the two most important residues in regard to protein interactions and modulation of CaV1.2 currents. In contrast, mutating the side chains of CaVα2δ1 Ser-263, Thr-331, and Asp-363 with alanine did not completely prevent channel function. Molecular dynamics simulations indicated that the carboxylate side chain of CaVα2δ1 Asp-259 coordinates the divalent cation that is further stabilized by the oxygen atoms from the hydroxyl side chain of CaVα2δ1 Ser-261 and the carboxylate group of CaVα1C Asp-181. In return, the hydrogen atoms contributed by the side chain of Ser-261 and the main chain of Ser-263 bonded the oxygen atoms of CaV1.2 Asp-181. We propose that CaVα2δ1 Asp-259 promotes Ca2+ binding necessary to produce the conformation of the VWA domain that locks CaVα2δ1 Ser-261 and Ser-263 within atomic distance of CaVα1C Asp-181. This three-way network appears to account for the CaVα2δ1-induced modulation of CaV1.2 currents.

The von Willebrand factor type A (VWA) 4 domain is found in more than 40,000 eukaryotic proteins (1) where it mediates protein-protein interaction (2). VWA superfamily is characterized by a "␤-␣-␤" fold consisting of alternating sequence of six-stranded ␤-sheets wrapped around six ␣-helical segments (3)(4)(5). Binding of a divalent cation (either Mg 2ϩ , Mn 2ϩ , or Ca 2ϩ ) is usually necessary for controlling its protein-ligand properties (6). Cations are coordinated by a maximum of five noncontiguous residues forming the metal ion-dependent adhesion site (MIDAS) motif characterized by the conserved primary sequence in the first loop Asp-Xaa-Ser-Xaa-Ser (DXSXS) followed by a threonine residue in the second loop and an aspartate residue in the third loop (7).
Human VWA-containing proteins include extracellular matrix proteins (e.g. collagens, cochlin, and matrilin) and integral membrane proteins (e.g. integrin and anthrax receptors) (3,8,9). Through the capture of a divalent ion, the MIDAS motif adopts a unique conformation that facilitates ligand association (10). Structural studies carried out in integrin receptors have shown that the nature of the ion coordinating the MIDAS motif influences the folding conformation of the protein and thus its ligand-binding properties (11,12).
L-type Ca V 1.2 channels are hetero-oligomeric protein complexes consisting of a main pore-forming Ca V ␣1C (thereafter referred to simply Ca V 1.2 for conciseness) assembled with Ca V ␤ and Ca V ␣2␦1 auxiliary subunits (13)(14)(15)(16). Like many integrin proteins, Ca V ␣2␦1 is glycosylated at multiple asparagine residues and is retained at the plasma membrane by a relatively short glycosylphosphatidylinositol anchor following the cleavage of its unique transmembrane domain (17,18). Although Ca V 1.2 is responsible for the passage of Ca 2ϩ through the pore, its association with Ca V ␣2␦1 is necessary to reproduce the biophysical properties of the native channel (17,18), in particular the channel activation at negative membrane potentials. We have recently shown that two adjacent negatively charged residues Asp-180 and Asp-181 in the S1S2 loop in the repeat I of Ca V 1.2 are required to recapitulate the complex This work was supported in part by Operating Grant 130256 from the Canadian Institutes of Health Research and Grant G-15-0009329 from the Canadian Heart and Stroke Foundation (to L. P.). The authors declare that they have no conflicts of interest with the contents of this article. 1 Recipient of a Ph.D. award from "Fonds de la Recherche du Qué bec en Nature et Technologie." 2 Junior 2 scholar from "Fonds de la Recherche du Qué bec en Santé." 3 To whom correspondence should be addressed: Dé

Electrostatic interactions are mediated by MIDAS residues at the Ca V 1.2-Ca V ␣2␦1 interface
We recently identified the negatively charged Asp-181 in the first extracellular loop of Ca V 1.2 as an essential determinant for the interaction with the extracellular Ca V ␣2␦1 protein (19). To identify the molecular determinants of the interface, we have produced a 3D model of the IS1S4 region using the molecular coordinates of the cryo-EM structure of the homologous Ca V 1.1 channel complex (PDB code 5GJV) (16). The sequence identity between IS1S4 in Ca V 1.1 and Ca V 1.2 is 69.6%, whereas VWA is perfectly conserved. Nonetheless, the molecular coordinates of the IS3S4 loop were not reported in the cryo-EM structure and were reconstructed using Modeler (see details in "Experimental procedures"). The 3D model was selected using a DOPE score of Ϫ37032 (between Ϫ37,100 and Ϫ35,578) and a molpdf score of 2147 (between 2144 and 3085). MD simulations suggest that Gly-262 and/or Ser-263 in the VWA domain of Ca V ␣2␦1 were the most likely residues to interact with Asp-181 in the first extracellular loop of Ca V 1.2. Along with Asp-259, Ser-261, Thr-331, and Asp-363, Ser-263 is one of the five polar residues that form a perfect MIDAS motif in the rat isoform of Ca V ␣2␦1 (Fig. 1A). X-ray crystallography structures of integrin proteins position a single Mg 2ϩ enclosed within MIDAS residues, although Ca 2ϩ , Zn 2ϩ , or Mn 2ϩ could also occupy the binding site (21). Ca 2ϩ and Mg 2ϩ are both plausible candidates to occupy the MIDAS at the Ca V ␣2␦1-Ca V 1.2 interface given that the concentration of Ca 2ϩ and Mg 2ϩ ions is roughly equivalent (in the millimolar range) in the extracellular medium of eukaryotic cells. We chose to perform the MD simulations in the presence of a Ca 2ϩ ion, mostly because the ion was included in the coordinates of the published cryo-EM structure (16). The Ca 2ϩ ion was virtually inserted in the geometric center of the motif, and its average position was adjusted by MD simulations.
MD simulations were carried out after the transmembrane helices of Ca V 1.2 (membrane protein) were embedded in a model dipalmitoylphosphatidylcholine (DPPC) lipid bilayer containing 128 DPPC molecules (Fig. 1B). The hydrophobic tails of the lipid molecules in the plasma membrane appear to stabilize hydrophobic residues Ala-253, Leu-254, Gly-255, and Gly-256 of the IS3S4 region within the plasma membrane. The protein-membrane system was solvated with water in the pres- 2-Ca V ␣2␦1 interface was built based using the molecular coordinates of the cryo-EM structure of the skeletal muscle Ca V 1.1 complex (PDB code 5GJV). This model differs slightly from the 3D model of the same region published previously (19) in regard to the orientation of the IS3S4 extracellular loop. A, residues forming the protein interface are zoomed in with emphasis on the MIDAS motif. The main-chain groups of amino acids forming the MIDAS motif (Asp-259, Ser-261, Ser-263, Thr-331, and Asp-363) are shown as pale rose sticks with carboxyl groups in red, amine groups are blue, and hydrogen atoms are white. Residues Asp-259, Ser-261, and Thr-331 identified within full-lined black boxes are in the foreground, and Ser-263 and Asp-363 identified within the dot-lined boxes are in the background. MIDAS residues are facing the IS1S2 loop in Ca V 1.2. Ca V 1.2 Asp-181 is shown in stick representation. B, VWA domain of the rat Ca V ␣2␦1 is shown in schematic representation in which ␣-helices appear in cyan, and ␤-strands are shown in pink. A single Ca 2ϩ ion (green) is shown as being coordinated by the MIDAS residues. The region spanning from the S1 to the S4 segments in repeat I (IS1S4) of Ca V 1.2 is shown in schematic representation, and the transmembrane helices S1 to S4 are colored from the darker (S1) to the paler shade of blue (S4). Modeled DPPC lipids are shown in stick representation with carbon, oxygen, and phosphorus atoms being shown in gray, red, and pale green, respectively. Modeling was achieved with Modeler 9.17. The figure was produced using PyMOL (DeLano Scientific).

MIDAS mutations in Ca V 1.2 channels
ence of 143 mM NaCl, 2 mM MgCl 2 , and 2 mM CaCl 2 . This 3D model reached a steady state after 1.1 ns equilibration with r.m.s.d. variations around 4.5 Å for simulations up to 25 ns, and this observation stood for three distinct trajectories. As seen, the negatively charged side chain of Ca V 1.2 Asp-181 is predicted in this model to face the hydrogen atoms from the hydroxyl groups of Ser-261 at a distances of 1.7 Å thus suggesting a key role for Ca V ␣2␦1 Ser-261 in enabling the interaction with Ca V 1.2.

Multiple mutations of the MIDAS residues prevent cell-surface expression of Ca V ␣2␦1 and physical interaction with Ca V 1.2
Coimmunoprecipitation assays were carried out to explore the contribution of MIDAS residues in promoting the interaction between Ca V 1.2 and Ca V ␣2␦1. The importance of the polar side chains was examined after substitution with the hydrophobic alanine residue at multiple MIDAS positions with Ca V ␣2␦1 D259A/S261A/S263A (DSS/A), D259A/S261A/S263A/T331A (DSST/A), and D259A/S261A/S263A/D363A (DSSD/A). Assays were performed with untagged Ca V 1.2, mCherry-Ca V ␣2␦1-HA, and Ca V ␤3c-Myc using anti-c-Myc-coated beads to bait the channel complex. For conciseness, the mCherry-Ca V ␣2␦1-HA construct will hereafter be referred to as Ca V ␣2␦1-HA. Under this protocol, Ca V ␤3c-Myc binds to the magnetic beads and pulls down the Ca V 1.2-Ca V ␣2␦1 complex ( Fig. 2A). As seen in Fig. 2B, input lane, Ca V 1.2, Ca V ␣2␦1-HA WT and mutants, and Ca V ␤3 proteins were translated at the expected molecular masses of 250, 175, and 60 kDa, respectively, after recombinant expression in HEKT cells. Ca V ␤3c-Myc was captured by the beads under all experimental conditions, and Ca V 1.2 successfully pulled down Ca V ␣2␦1 WT (Fig. 2C). The ablation of the negative or polar side chain in the MIDAS residues prevented the coimmunoprecipitation of any of the multiple mutants in Ca V ␣2␦1, even after a 200-s exposure. Two-color flow-cytometry assays (17)(18)(19) were performed to determine whether disrupting MIDAS also altered the cell-surface expression of Ca V ␣2␦1. As detailed elsewhere (17,18), the cell-surface expression of the constructs, relative to mCherry- A, schematic demonstrating the coimmunoprecipitation assay using Ca V ␤3c-Myc as the bait. B, HEKT cells were transiently transfected with pCMV-Ca V 1.2, pCMV-Ca V ␤3c-Myc, and pmCherry-Ca V ␣2␦1-HA WT or mutants as shown. Cell lysates were immunoprecipitated (IP) overnight with anti-c-Myc magnetic beads to capture Ca V ␤3, eluted in a Laemmli buffer and fractionated by SDS-PAGE using 8% gels. B, immunoblotting (IB) was carried out on total proteins (20 g) before coimmunoprecipitation assay (input lane) to confirm that all proteins have been correctly translated. Constructs are from left to right: mCherry-Ca V ␣2␦1-HA WT; mCherry-Ca V ␣2␦1-HA DSS/A (D259A/S261A/S263A); mCherry-Ca V ␣2␦1-HA DSST/A (D259A/S261A/S263A/T331A); and mCherry-Ca V ␣2␦1-HA DSSD/A (D259A/S261A/S263A/ D363A). The signal for the housekeeping protein GAPDH is shown. C, immunoblotting was carried out after eluting the protein complexes (IP-c-Myc lane) from the beads with anti-Ca V 1.2, anti-Ca V ␣2␦1, and anti-Ca V ␤3 antibodies (from top to bottom, as indicated). All immunoblots were carried out in parallel under the same transfection and extraction conditions. The signal for Ca V ␣2␦1 was revealed after three exposures times (1, 50, and 200 s). Ca V ␤3 and Ca V 1.2 proteins migrated, respectively, at 60 and 250 kDa. All Ca V ␣2␦1 proteins migrated at Ϸ175 kDa, which is consistent with the molecular mass of the mCherry-Ca V ␣2␦1-HA WT reported in previous studies (19). These assays were repeated three times over the course of 6 months with different cell preparations. Assays carried out with anti-HA-coated beads to pull down the channel complex with Ca V ␣2␦1-HA as the bait (as seen in Fig. 6) yielded similar results. D, schematic illustrating the relative position of the extracellular HA epitope and the intracellular mCherry translated after the C terminus on the mCherry-Ca V ␣2␦1-HA construct used to carry the flow cytometry assays. The construct allows for detection of intracellular and extracellular fluorescence using, respectively, a FITC-conjugated anti-HA and the constitutive fluorescence of mCherry. E, representative two-dimensional plots of mCherry versus FITC fluorescence. The cell-surface expression of the mCherry-Ca V ␣2␦1-HA construct WT and mutants from the surface fluorescence emitted by the FITC-conjugated anti-HA as measured using a flow cytometry assay (10,000 intact cells) is shown. From top to bottom, the constructs that were tested are as follows: The cell-surface fluorescence for FITC, calculated as ⌬MedFI, as explained under "Experimental procedures," was significantly smaller for Ca V ␣2␦1-HA mutants than for the WT construct (see Table 1 for numerical values). Furthermore, the mCherry signal, which reflects the total protein density, was also decreased suggesting that these mutations impaired protein stability.

MIDAS mutations in Ca V 1.2 channels
Ca V ␣2␦1-HA WT, is assessed using the fluorescence intensity of an anti-hemagglutinin tag conjugated to FITC (anti-HA FITC) antibody directed against the extracellular HA tag inserted after position Arg-676 in mCherry-Ca V ␣2␦1 (Fig. 2D). Fluorescence intensity of the anti-HA tag antibody conjugated to FITC measured in intact nonpermeabilized cells provides an estimation of the cell-surface expression of the construct, whereas the fluorescence intensity measured after cell permeabilization is used as a marker of total protein expression. As seen, multiple mutations of the MIDAS residues significantly reduced the cell-surface expression of mCherry-Ca V ␣2␦1-HA to Ϸ10% for the mutants (Fig. 2E). More importantly, multiple mutations were also seen to reduce the protein stability with a sharp decrease in the relative fluorescence for FITC measured in permeabilized cells (Table 1). This observation is in agreement with previous reports showing that coexpression of Ca V ␣2␦1 increases the total protein stability of Ca V ␤3bound Ca V 1.2 proteins, and conversely coexpression of Ca V 1.2 increases the total protein stability of Ca V ␣2␦1 in the presence of Ca V ␤3 (22). Altogether, these results indicate that the polar or negatively charged side chains of the MIDAS residues contribute to the stability of Ca V ␣2␦1, its cell-surface expression, and its interaction with the poreforming subunit of Ca V 1.2.

Functional modulation is intact with Ca V ␣2␦1 V260R
Individual contributions from the side chains of MIDAS amino acids were analyzed after substitutions to the positively charged arginine (Figs. 3 and 4) and to the neutral alanine residue (Figs. 5 and 6). This two-step strategy was meant to exclude residues that would remain functional following nonconserved substitutions predicted to alter the local electrical field. MIDAS residues as well as the intervening Val-260 and Gly-262 residues were sequentially mutated. All mCherry-Ca V ␣2␦1-HA single mutants expressed the mCherry tag and were trafficked at the plasma membrane (Table 1). Coimmunoprecipitation assays were carried out with c-Myc-coated beads to bait the channel complex. As seen in Fig. 3A, input lane, Ca V 1.2, Ca V ␣2␦1-HA WT and mutants, and Ca V ␤3 proteins were translated at the expected molecular masses of 250, 175, and 60 kDa, respectively, after recombinant expression in HEKT cells. Ca V ␤3 was captured by the beads under all experimental conditions, and Ca V 1.2 successfully pulled down Ca V ␣2␦1-HA WT (Fig. 3B). Not too surprisingly, the mutation to an arginine residue significantly impaired the ability for most Ca V ␣2␦1 mutants, save for V260R, to interact with Ca V 1.2. Longer exposure times allowed us to detect a small interaction with Ca V ␣2␦1-HA G262R and Ca V ␣2␦1-HA S263R, but no interaction was ever detected with Ca V ␣2␦1-HA D259R and Ca V ␣2␦1-HA S261R. A separate series of experiments demonstrated that protein interaction was also somewhat conserved with Ca V ␣2␦1-HA T331R and Ca V ␣2␦1-HA D363R in a fashion reminiscent of Ca V ␣2␦1-HA G262R and Ca V ␣2␦1-HA S263R (data not shown). These results indicate that the side chain at position Val-260 is not directly involved in protein Table 1 Cell-surface expression of Ca V ␣2␦1 Stable HEKT Ca V ␤3 cells were transiently transfected simultaneously with 4 g of pCMV-Ca V 1.2 and 4 g of pmCherry-Ca V ␣2␦1 HA WT or mutants. Flow cytometry was used to determine the proportion of Ca V ␣2␦1 proteins expressed at the cell surface as reported in the "intact cells" column, whereas the "permeabilized cells" column represent the total expression level. Fluorescence intensity was estimated using FlowJo software as detailed elsewhere (39). The relative cell-surface expression of Ca V ␣2␦1-HA constructs was estimated by calculating the ratio of the ⌬MedFI for the FITC fluorophore over ⌬MedFI values quantified for the FITC fluorophore for the WT construct (⌬MedFI mutant/⌬MedFI WT) in intact nonpermeabilized cells for the experiments carried out the same day under the same conditions. The relative total expression of Ca V ␣2␦1 constructs was estimated by calculating the ratio of the ⌬MedFI values quantified for the FITC fluorophore over ⌬MedFI values quantified for the FITC fluorophore for the WT construct after cell permeabilization. These ratios are expressed herein as a percentage. The total number of experiments is provided in parentheses, with each experiment being the result of a different transfection in a separate cell dish (Ͼ10,000 cells). Statistical analysis was carried out against the ⌬MedFI for FITC measured with pmCherry-Ca V ␣2␦1-HA WT. *, p Ͻ 0.05, and **, p Ͻ 0.001.
Relative ⌬MedFI (%) FITC-conjugated anti-HA Cell surface, intact cells Total, permeabilized cells anti-c-Myc magnetic beads to capture Ca V ␤3, eluted in a Laemmli buffer, and fractionated by SDS-PAGE using 8% gels. A, immunoblotting (IB) was carried out on total proteins (20 g) before the immunoprecipitation assay (input lane) to confirm that constructs were expressed at the expected molecular weight. The signal for the housekeeping protein GAPDH is shown. B, immunoblotting was carried out after eluting the protein complexes from the beads with anti-Ca V 1.2, anti-Ca V ␣2␦1, and anti-Ca V ␤3 antibody (from top to bottom, as indicated). Images for Ca V ␣2␦1 were captured after shorter (1 s) or longer exposure times until the signal was saturated (20 s). All immunoblots were carried out in parallel under the same transfection and extraction conditions. These assays were successfully repeated three times with similar results.

MIDAS mutations in Ca V 1.2 channels
interaction, whereas the side chains at Asp-259 and Ser-261 play a critical role in this process. Examination of the 3D model concurs that the main-chain atoms of Val-260 are facing the Ca 2ϩ ion and that side-chain substitution does not alter that preferred configuration.
To establish a correlation between physical interaction and functional modulation, the impact of these mutations on channel function was examined after their recombinant expression in the presence of Ca V 1.2 and Ca V ␤3 in HEKT cells (Fig. 4A). As published previously (17)(18)(19)22), coexpression of Ca V 1.2 and Ca V ␤3 with the Ca V ␣2␦1-HA WT construct increases the peak current density by 7-10-fold from Ϫ2.5 Ϯ 0.3 pA/pF (25) (no insert in the pmCherry vector) to Ϫ18 Ϯ 1 pA/pF (n ϭ 359) (Fig. 4B). Ca V ␣2␦1-HA also caused a Ϸ Ϫ15-mV leftward shift in the activation potential of Ca V 1.2 from E 0.5, act ϭ ϩ8 Ϯ 2 mV (n ϭ 25) in the absence of Ca V ␣2␦1 to E 0.5, act ϭ Ϫ8.7 Ϯ 0.4 mV (n ϭ 359) in the presence of Ca V ␣2␦1-HA. Although the negative shift was significant in our sample size, we observed some overlap between the two conditions with the corresponding free energy of activation (⌬G act ) ranging from Ϫ2 to ϩ0.2 kcal mol Ϫ1 (33% peak at Ϫ0.6 kcal mol Ϫ1 ) for the WT channel combination, whereas ⌬G act ranged from Ϫ0.2 to ϩ1.4 kcal mol Ϫ1 (40% peak at ϩ0.2 kcal mol Ϫ1 ) in the absence of Ca V ␣2␦1 ( Fig. 4C). Up-regulation of Ca V 1.2 currents was severely disrupted with all arginine mutants save for Ca V ␣2␦1-HA V260R. Peak current density measured with Ca V ␣2␦1-HA D259R, Ca V ␣2␦1-HA S261R, Ca V ␣2␦1-HA G262R, Ca V ␣2␦1-HA S263R, Ca V ␣2␦1-HA T331R, and Ca V ␣2␦1-HA D363R was significantly different from whole-cell currents measured with Ca V ␣2␦1-HA WT (p Ͻ 0.05). Furthermore, modulation by Ca V ␣2␦1-HA S261R, Ca V ␣2␦1-HA G262R, and Ca V ␣2␦1-HA D363R was indistinguishable from the current density obtained with the mock vector (Table 2). Altogether, these results rule out a significant role for the side chain at position 260. The addition of the long arginine side chain at position 262 could locally impair flexibility in a region that plays a role in the interaction with the IS1S2 loop of Ca V 1.2.

Asp-259 and Ser-261
Mutant proteins of Ca V ␣2␦1 were produced to investigate the molecular requirements for the protein interaction and functional modulation of Ca V 1.2 currents. Coexpression of Ca V 1.2/Ca V ␤3 with Ca V ␣2␦1-HA D259A, S261A, and G262A/ S263A mutants produced whole-cell currents that were not significantly different from that obtained in the presence of the  Table 2. C, bar graph reporting the distribution of the free energies of activation (⌬G act ) measured with each Ca V ␣2␦1 mutant. The distribution of the ⌬G act values for mCherry-Ca V ␣2␦1-HA D259R, S261R, G262R, and D363R was right-shifted as compared with mCherry-Ca V ␣2␦1-HA WT and overlapped with the ⌬G act values measured in the presence of mock vector (pmCherry-no Ca V ␣2␦1). mCherry mock vector (Fig. 5A), although these mutants were confirmed to form bona fide proteins that were detected at the cell surface (Table 1). In contrast, Ca V ␣2␦1-HA G262A, S263A, T331A, and D363A stimulated whole-cell currents without any alteration in the voltage dependence of activation (Fig. 5, B and C). In fact, there was a strong association between peak current density and the negative voltage dependence of activation (⌬G act ) ( Fig. 5C and Table 2) such that mutants yielding large whole-cell currents activated at negative membrane potentials. We did not identify any Ca V ␣2␦1 mutant that simply disrupted the voltage dependence of activation.
Peak current densities ranged from near-background wholecell currents measured with the mock vector in the following sequence: It should be noted that conservative substitutions Ca V ␣2␦1-HA D259E and Ca V ␣2␦1-HA S261T produced Ca V 1.2 wholecell currents that were similar to Ca V ␣2␦1 WT (Table 2) indicating that the side chains at Asp-259 and Ser-261 contribute significantly to channel modulation.
The structural requirements at position Ca V ␣2␦1 Gly-262 were further investigated ( Table 2). A significant decrease in the channel function was also observed with Ca V ␣2␦1-HA G262T, whereas coexpression with the negatively charged Ca V ␣2␦1-HA G262D, predicted to project a negatively charged side chain, failed to stimulate Ca V 1.2 currents suggesting that a polar or a negative side chain cannot compensate for the decreased flexibility of the residue at this position.
Ca V ␣2␦1-HA S263A, S263G, S263R, and S263T were all associated with whole-cell currents that activated in a range of negative voltages associated with Ca V ␣2␦1-HA WT ( Table 2). The peak current densities, significantly higher than those recorded with the mock vector, suggest that the side chain at this position does not contribute appreciably to the functional interaction with Ca V 1.2 Asp-181.

MIDAS mutations in Ca V 1.2 channels
Channel function was also characterized by combining mutations from residues in Ca V ␣2␦1-HA and Ca V 1.2 that could come into atomic proximity in the virtual model of the protein interface with Ca V ␣2␦1-HA T331A ϩ Ca V 1.2 S182A, Ca V ␣2␦1-HA T331A ϩ Ca V 1.2 K257A, and Ca V ␣2␦1-HA E363A ϩ Ca V 1.2 K257A in the presence of Ca V ␤3. These three pairs of mutants produced depolarization-activated whole-cell currents that were not significantly different from currents produced by Ca V ␣2␦1-HA WT ϩ Ca V 1.2 WT when recorded under the same conditions (Table 2). Altogether, charged mutations of MIDAS residues Ca V ␣2␦1 Asp-259 and Ca V ␣2␦1 Ser-261 were seen to impact more significantly on channel function than any other position herein tested.

Mutations of MIDAS residues impair coimmunoprecipitation of Ca V ␣2␦1 with Ca V 1.2 and Ca V 2.3
To investigate whether functional modulation requires highaffinity interaction between the two proteins at specific sites, coimmunoprecipitation assays were carried out with the single alanine mutants. Ca V ␣2␦1-HA was used to bait the Ca V 1.2- The signal for the housekeeping protein GAPDH is shown below. B, immunoblotting was carried out after eluting the protein complexes from the beads with anti-Ca V 1.2 and anti-Ca V ␣2␦1 antibodies (from top to bottom, as indicated). All immunoblots were carried out in parallel under the same transfection and extraction conditions. In these experiments, Ca V ␣2␦1-HA is bound to the beads such that the interaction is probed by investigating the signal for its Ca V 1.2 partner. The image was obtained using the automatic mode of the ChemiDoc Touch system, which optimizes the signal. As shown, the signal for Ca V 1.2 was significantly reduced for mCherry-Ca V ␣2␦1-HA D259A, S261A, and S263A. C and D, HEKT cells were transiently transfected with pcDNA3-Ca V 2.3 and pCMV-Ca V ␤3c-Myc and either pmCherry-Ca V ␣2␦1-HA WT or pmCherry-Ca V ␣2␦1-HA D259A, S261A, S263A, T331A, or D363A. Cell lysates were immunoprecipitated overnight with anti-c-Myc magnetic beads to capture Ca V ␤3c-Myc, eluted in a Laemmli buffer and fractionated by SDS-PAGE using 8% gels. C, immunoblotting was carried out on total proteins (20 g) before the immunoprecipitation assay (input lane). The signal for the housekeeping protein GAPDH is shown. D, immunoblotting was carried out after eluting the protein complexes from the beads with anti-Ca V 2.3, anti-Ca V ␣2␦1, and anti-Ca V ␤3 antibodies (from top to bottom, as indicated). In these experiments, Ca V ␤3/Ca V 2.3 is bound to the beads such that the interaction is probed by investigating the signal for Ca V ␣2␦1. As shown, the signal for Ca V ␣2␦1 was significantly reduced for mCherry-Ca V ␣2␦1-HA D259A, S261A, S263A, and D363A. All immunoblots were carried out in parallel under the same transfection and extraction conditions.

Table 2
Biophysical properties of mCherry-Ca V ␣2␦1-HA WT and mutants Ca V 1.2 WT (or mutant in a few cases) was coexpressed with Ca V ␤3 and either pmCherry-mock vector or pmCherry-Ca V ␣2␦1-HA WT or mutants using a 4:4:4 g ratio. Biophysical parameters were measured in the presence of 2 mM Ca 2ϩ as described elsewhere (17)(18)(19). Activation properties (E 0.5, act and ⌬G act ) were estimated from the mean I-V relationships and fitted to a Boltzmann equation. Only the GFP-positive cells with voltage-activated currents were kept for further analysis. Null-current cells outnumbered the cells with inward currents for mCherry-Ca V ␣2␦1-HA D259A/S261A/S263A/T331A, mCherry-Ca V ␣2␦1-HA D259A/ S261A/S263A/D363A, mCherry-Ca V ␣2␦1-HA D259A/S261A/S263A, mCherry-Ca V ␣2␦1-HA S261A/G262A/S263A, and mCherry-Ca V ␣2␦1-HA G262A/ 263A. The data are shown as the mean Ϯ S.E. of the number of cells (one cell per experiment), and the total number of experiments carried over several months is provided in parentheses. Statistical analysis was carried out against the values obtained in the presence of the mock pmCherry-N1 vector with 1 ϭ p Ͻ 0.01 and against mCherry-Ca V ␣2␦1-HA WT with 2 ϭ p Ͻ 0.05.

MIDAS mutations in Ca V 1.2 channels
Ca V ␤3 complex using anti-HA-coated magnetic beads (Fig. 6,  A and B). As shown, Ca V ␣2␦1-HA WT, V260A, and T331A successfully pulled down Ca V 1.2 and Ca V ␤3 (Fig. 6B). In contrast, coimmunoprecipitation of the Ca V 1.2-Ca V ␤3 complex was abolished with Ca V ␣2␦1-HA D259A and S261A and significantly reduced, but not eliminated, with Ca V ␣2␦1-HA S263A and D363A. It is important to stress that the pulldown assays remain a tool largely qualitative. Coimmunoprecipitation assays carried out using anti-c-Myc-coated beads to bait the Ca V 1.2-Ca V ␤3 complex yielded similar results (data not shown).
Combining the G262A with the S263A mutation in the Ca V ␣2␦1-HA G262A/S263A mutant completely eradicated coimmunoprecipitation suggesting that local conformation plays a role in these processes (data not shown). In contrast, Ca V ␣2␦1-HA E366A/E367A, situated just outside MIDAS, was easily pulled down by Ca V 1.2 (data not shown) in agreement with the patch-clamp data. The signal for the Ca V 1.2 protein obtained in the presence of Ca V ␣2␦1-HA T331A remained remarkably strong confirming that the nature of the side chain at this position does not contribute directly or indirectly to protein interaction.
The importance of MIDAS residues in mediating protein interaction was further established by carrying similar experiments with the neuronal Ca V 2.3 channel (Ca V ␣1E with auxiliary subunits) (23). Ca V ␤3c-Myc was used to bait the Ca V 2.3-Ca V ␣2␦1-HA complex using anti-c-Myc-coated magnetic beads (Fig. 6, C and D). As seen, Ca V ␤3 successfully pulled down Ca V 2.3/Ca V ␣2␦1-HA WT and Ca V 2.3/Ca V ␣2␦1-HA T331A (Fig. 6D). In contrast, coimmunoprecipitation was abolished with Ca V ␣2␦1-HA D259A, S261A, S263A, and D363A. Ca V ␣2␦1 Asp-259 and Ser-261 were herein shown to contribute to the interaction with Ca V 1.2 and Ca V 2.3. Ca V ␣2␦1 Ser-263 and Ca V ␣2␦1 Asp-363 play a more prominent role in the interaction with Ca V 2.3 than with Ca V 1.2 suggesting that the interface could be slightly different. In both channels, however, substitution of the residue Ca V ␣2␦1-HA Thr-331 did not impair protein interaction. Hence, four MIDAS residues contribute to different degrees to protein interaction in highvoltage-activated Ca 2ϩ channels.

Discussion
The glycosylphosphatidylinositol-anchored Ca V ␣2␦1 protein is an intrinsic subunit of high voltage-activated Ca 2ϩ channels of the Ca V 1 and Ca V 2 families. It is an extracellular protein that comprises several structural domains, including four CACHE domains in tandem with a single VWA domain. VWAcontaining domains mediate protein-protein interaction in a wide variety of proteins located at the cell surface, including integrins, collagens and complement factor (2), and cartilage matrix proteins (24). The VWA domain of Ca V ␣2␦1 bears high sequence identity with the domain I of integrin (15,16,25). This biological function is usually mediated via a conserved MIDAS motif that coordinates divalent cations at the protein-ligand interface (10). The recently published high-resolution cryo-EM structure of the homologous Ca V 1.1 heteromeric complex supports a close interaction between residues of the VWA domain in Ca V ␣2␦1 and extracellular loops from repeats I, II, and III of the pore-forming Ca V ␣1 subunit (16). It also shows that the VWA domain of Ca V ␣2␦1 adopts a classic ␣/␤ Rossmann fold (26) and contains a "perfect" MIDAS motif formed by residues Asp-259, Ser-261, Ser-263, Thr-331, and Asp-363 (rat numbering) (16). Multiple primary sequence alignments carried out with the four isoforms of Ca V ␣2␦ (Ca V ␣2␦1-4) in human, rabbit, and rat species, revealed that four of five MIDAS residues (Asp-259, Ser-261, Ser-263, and Asp-363) are strictly conserved with the 4th position occupied by threonine being infrequent.

N-terminal residues in MIDAS mediate interaction with Ca V 1.2
As we have previously shown, the interaction between Ca V 1.2 and Ca V ␣2␦1 requires a negatively charged aspartate or glutamate residue at position Asp-181 in the first extracellular loop in the voltage sensor domain in repeat I of Ca V 1.2 (19). A single point mutation at this position is sufficient to prevent Ca 2ϩ influx above background levels. In contrast, we are currently demonstrating that many residues in the VWA domain of Ca V ␣2␦1 play a role in the protein interaction and functional modulation of whole-cell Ca 2ϩ currents by Ca V ␣2␦1. One of the most intriguing observations is the role played by Ca V ␣2␦1 Asp-259 in mediating protein interaction with Ca V 1.2. MD simulations carried out with our 3D model of the WT complex (Ca V 1.2 IS1S4 -Ca V ␣2␦1 VWA) indicate that the minimal distance between the two ␣-carbons (C␣) of Ca V ␣2␦1 Asp-259 and Ca V 1.2 Asp-181 is never smaller than 10.5 Å during the complete course of 25-ns simulations carried out with three trajectories. Within the limitations of simulations performed with a virtual 3D model, such distance suggests that a direct interaction between the two proteins is not very likely.
In contrast to Ca V ␣2␦1 Asp-259, Ca V ␣2␦1 Ser-261 is freely exposed at the cell surface and positioned to mediate the interaction with Ca V 1.2 Asp-181. MD simulations show that Ser-261 adopts many conformations over the 25-ns period. The oxygen atom of the side-chain hydroxyl group of Ser-261 interacts with the Ca 2ϩ ion, and its hydrogen atom is seen to stabilize alternatively the carboxylate groups in the side chains of Ca V ␣2␦1 Asp-259 and Ca V 1.2 Asp-181. The hydroxyl group is a stringent structural requirement as illustrated by the observation that Ca V ␣2␦1-HA S261T mutant behaved essentially like Ca V ␣2␦1-HA WT in patch-clamp experiments. Interactions are also possible between Ca V ␣2␦1 Ser-263 and Ca V 1.2 Asp-181. MD simulations, however, show that the backbone of Ca V ␣2␦1 Ser-263 undergoes a high degree of fluctuations, which may account for the observation that Ca V ␣2␦1-HA S263A and Ca V ␣2␦1-HA S263R somewhat modulated Ca V 1.2 currents.
Without being part of the signature sequence of MIDAS, glycine residues often play essential structural role in the formation of loops (27). Mutations of Ca V ␣2␦1 Gly-262 to hydrophobic or hydrophilic residues reduced the modulation of whole-cell currents. The analysis of MD simulations suggests that the carboxylate side chain from Asp-181 interacts only occasionally with the main-chain hydrogen of Ca V ␣2␦1 Gly-262. Nevertheless, altering simultaneously the chemical properties of the side chains at positions Gly-262 and Ser-263 disrupts the functional modulation of Ca V 1.2 currents.

MIDAS mutations in Ca V 1.2 channels
Interaction between Ca V 1.2 and Ca V ␣2␦1 was also shown to mutually improve the stability of both proteins. This interaction needs to be sufficiently strong to withstand the pulldown assays. We could not however determine the strength of this interaction using isothermal calorimetric and/or surface plasmon resonance assays since it would require purifying large quantities of Ca V ␣2␦1, which were shown to be unstable when produced in regular bacterial systems (28).

Thr-331 plays a minor role in the interface in Ca V 1.2 and Ca V 2.3 channels
The polar residue Thr-331 is also predicted to participate in Ca 2ϩ binding even though the fourth position in MIDAS is not strictly conserved, often being substituted by a flexible glycine residue (29). Mutations of Thr-331 to the hydrophobic residue with T331A produced only minor changes in protein-protein interaction in Ca V 1.2 and Ca V 2.3 channels and produced modest alterations, if any, in the functional modulation of Ca V 1.2 currents. MD simulations of Ca V ␣2␦1 T331A showed that the mutation induced a large structural reorganization of the MIDAS motif that preserved, however, the interaction of Ser-261 with Ca V 1.2 Asp-181. A role for Thr-331 cannot be completely excluded because functional modulation of Ca V 1.2 currents was decreased with Ca V ␣2␦1 T331R.
These results contrast with recent data published on the integrin-like regulatory subunit CLCA1. CLCA1 up-regulates TMEM16A currents through a Mg 2ϩ -and MIDAS-dependent mechanism (30) that stabilizes TMEM16A at the cell surface by preventing its internalization (31). Although the CLCA1 VWA domain also contains a five-residue MIDAS motif (Asp-312, Ser-314, Ser-316, Thr-383, and Asp-412) similar to Ca V ␣2␦1, the molecular determinants of modulation might be slightly different between the two families of ion channels. Disruption of the MIDAS motif, in particular mutation of Ser-316 or Thr-383 to Ala (equivalent to Ser-263 and Thr-331), reduced the ability of CLCA1 VWA to modulate surface protein levels and activity of TMEM16A (32). It is suggested that the molecular determinants responsible for the modulation of channel function by MIDAS residues in integrin-like proteins might command different parameters.
Although the mutations in the N-terminal region of MIDAS disrupted protein interaction and modulation, mutations of the structurally similar Asp-363 were seen to produce a modest impact by decreasing but not abolishing interaction with the Ca V 1.2 protein and functional modulation of Ca V 1.2 currents. MD simulations suggest that this may result from the relatively high exposure of Ca V ␣2␦1 Asp-363 to the solvent providing some additional degree of freedom to the side chain.

Divalent cation binding within MIDAS could stabilize protein-protein interaction
By analogy with integrin I proteins, cation binding within the MIDAS motif of Ca V ␣2␦1 could play a role in strengthening the interaction with the pore-forming subunit of Ca V 1.2 (33). Mutations of the MIDAS motif have been shown to prevent the binding of ␣2 I domain of integrin with collagen ligand (33,34). High-resolution structures of the ␣L I domain of integrin in complex with the domain 1 of the intracellular adhesion mole-cules ICAM-1 and ICAM-3 have shown that the divalent cation within the MIDAS motif is mostly stabilized by five local residues with an additional contribution from the aspartate residue of the ligand (35,36). In the absence of the ligand, this site is filled in with a water molecule (37). To gain additional mechanistic insight, we performed 25-ns MD simulations with the minimal 3D model after virtually inserting the mutation Ca V ␣2␦1 D259A (Fig. 7). Comparison of the minimum-energy conformations in Fig. 7, A (Ca V ␣2␦1 WT) and B (Ca V ␣2␦1 D259A), shows that the substitution of Asp-259 by an alanine residue is associated with a change in the conformation of the MIDAS residues around the Ca 2ϩ ion. The hydrogen atom from the hydroxyl group contributed by the side chain from Ca V ␣2␦1 Ser-261 switches from mostly coordinating the Ca 2ϩ ion to forming a hydrogen bond with the carboxyl group of the side chain from Ca V ␣2␦1 Asp-363. As a result, the Ca 2ϩ ion is seen as being pushed away from MIDAS and away from Ca V 1.2 Asp-181 toward the extracellular aqueous medium. This structural rearrangement could account for the large adverse impact of Ca V ␣2␦1 D259A on the coimmunoprecipitation and modulation of Ca V 1.2 currents. Hence, the binding of the divalent cation within MIDAS could be an essential component in the three-part network that stabilizes the interaction between Ca V ␣2␦1 and Ca V 1.2. This proposition is congruent with structural studies showing that the transition between the open and the closed conformations of integrins is mostly driven by the interaction with a divalent ion within the MIDAS motif (38). ) and Ϸ80% of the data points deviate less than 3 Å when compared with the structure shown in A. As shown, the oxygen atoms from Ca V 1.2 Asp-181 are within a few angstroms of the side-chain hydrogen atom of Ca V ␣2␦1 Ser-261 and the main-chain hydrogen atom of Ser-263. Elimination of the negatively charged side chain at position 259 in Ca V ␣2␦1 D259A (B) produced a substantial alteration in the conformation of MIDAS, whereas the hydrogen atom from the hydroxyl group of Ca V ␣2␦1 Ser-261 switches from coordinating the Ca 2ϩ ion and Ca V 1.2 Asp-181 to facing the carboxyl group in the side chain from Ca V ␣2␦1 Asp-363. In the Ca V ␣2␦1 D259A mutant, the Ca 2ϩ ion is being more freely exposed to the solvent and becomes coordinated by five water molecules (B) instead of two water molecules in the WT protein complex (A). This conformational change disrupts the network of stabilizing interactions between Ca V ␣2␦1 and Ca V 1.2 Asp-181. PDB files are available upon request.
We remain acutely aware that this working hypothesis is contingent to our virtual 3D model and the parameters we used for the MD simulations. As mentioned earlier, the 3D model of the pore-forming subunit Ca V ␣1C was built from the atomic determinants of Ca V ␣1S with an adequate but far from perfect identity at the primary structural level (69%). In addition, the template failed to provide electronic density in some regions, notably the extracellular IS3S4 loop that was fully reconstructed. It thus remains to be seen whether longer MD simulations carried out with a virtual 3D model of the two complete proteins (Ͼ500,000 atoms) would provide a similar picture.

Recombinant DNA techniques
The rabbit Ca V 1.2 (GenBank TM X15539) and the rat Ca V ␤3 (GenBank TM M88751) were subcloned in commercial vectors under the control of the CMV promoter as described elsewhere (23,39). The human Ca V 2.3 (GenBank TM L27745) was a generous gift from Dr. Toni Schneider (40). The coding sequence (1091 residues) of the rat brain Ca V ␣2␦1 clone (GenBank TM NM_012919) (41) was subcloned in the pmCherry-N1 vector, and the hemagglutinin (HA) epitope (YPYDVPDYA) was inserted in the extracellular domain of Ca V ␣2 between Asp-676 and Arg-677, such that the HA epitope is accessible from the extracellular compartment, and the mCherry is translated after the C terminus. This construct enables the detection of the Ca V ␣2␦1 proteins expressed at the cell surface as described previously (17)(18)(19)39). Point mutations were produced with the Q5 site-directed mutagenesis kit (New England Biolabs) in the pmCherry-Ca V ␣2␦1-HA construct (or in a few rare instances in the pCMV-Ca V 1.2 construct) according to the manufacturer's instructions as described elsewhere (17,22). Briefly, substitutions were created by incorporating the desired mutation in the center of the forward primer, and the reverse primer is designed so that the 5Ј ends of the two primers anneal back-to-back. Following the PCR, a kinase/ligase/DpnI enzyme mixture was added to the amplified DNA for circularization and template removal before transformation into high-effi-ciencyDH5-␣-competentE. coli(NewEnglandBiolabs).Allconstructs were verified by automated double-stranded sequence analysis (Genomics Platform, IRIC, Université de Montréal, Quebec, Canada). The protein expression at the correct molecular weight was confirmed by standard Western blot analysis for each construct as described below (17)(18)(19)22).

Coimmunoprecipitation assay and Western blot analysis
HEKT cells were transiently transfected with 4 g of pCMV-Ca V ␤3c-Myc and pCMV-Ca V 1.2 and pmCherry-Ca V ␣2␦1-HA WT or mutants. One series of experiments was carried out (Fig. 6, C and D) with pCMV-Ca V ␤3c-Myc and pcDNA3-Ca V 2.3 and pmCherry-Ca V ␣2␦1-HA WT or mutants. Two days after transfection, cells were homogenized in 20 mM NaMOPS (pH 7.4), 300 mM NaCl, and 1% digitonin supplemented with protease inhibitors (Thermo Fisher Scientific). Homogenates were sonicated, incubated for 1 h at 4°C, and centrifuged at 16,000 ϫ g for 30 min. The supernatant was kept, and the cell pellet was discarded. The protein concentration in the supernatant was determined using BCA protein assay (Thermo Fisher Scientific). A small fraction (20 g) of the supernatant was kept as the "input fraction." The remaining Ϸ600 g of proteins were diluted with an equal volume of 20 mM NaMOPS (pH 7.4) and 300 mM NaCl to have a final concentration of digitonin at 0.5%. Two different protocols were used for the coimmunoprecipitation assays using either Ca V ␣2␦1-HA or Ca V ␤3c-Myc as the bait. In some cases (Fig.  6, A and B), Ca V ␣2␦1-HA was used as the bait to immunoprecipitate the Ca V 1.2 complex. Appropriate control experiments carried out with the three constructs confirmed that Ca V ␣2␦1-HA was the only protein pulled out by the anti-HAcoated beads (data not shown). The supernatant was then incubated overnight with 50 l of anti-HA magnetic beads (Thermo Fisher Scientific) previously washed according to the manufacturer's instructions. In some other experiments (Figs. 2, B and C, 3, and 6, C and D), Ca V ␤3c-Myc was used as the bait to immunoprecipitate the Ca V 1.2-Ca V ␣2␦1 or the Ca V 2.3-Ca V ␣2␦1 complex. In these experiments, the supernatant was incubated overnight with 50 l of anti-c-Myc magnetic beads (Thermo Fisher Scientific) previously washed according to the manufacturer's instructions. We have performed both assays with similar success, although the signal for the pore-forming subunit Ca V ␣1 of Ca V 1.2 and Ca V 2.3 was larger in the presence of the anti-c-Myc beads.

Flow-cytometry assays
Flow-cytometry experiments were carried out to evaluate the cell-surface expression of the mCherry-Ca V ␣2␦1-HA WT and mutants. Stably Ca V ␤3 cells were transiently transfected with 4 g of pCMV-Ca V 1.2 WT and 4 g of pmCherry-Ca V ␣2␦1-HA WT or mutants. Experiments were conducted and analyzed as published before (17)(18)(19)22) and are described in greater detail elsewhere (39). Briefly, the cell-surface expression of the mCherry-Ca V ␣2␦1-HA WT was detected with the FITC-conjugated mouse monoclonal anti-HA epitope tag antibody at 5 g/ml (Sigma). To determine the total quantity of both intracellular and extracellular expression of the tagged proteins, cells were fixed and permeabilized using BD Cytofix/Cytoperm TM fixation/permeabilization solution kit (BD Biosciences). Roughly 10,000 cells were counted using a FACSAria III SORP flow cytometer (BD Biosciences). The control conditions were carried out in triplicate with each series of experiments: (a) untransfected Ca V ␤3 cells without the anti-HA FITC antibody; (b) untransfected Ca V ␤3 cells with the anti-HA FITC antibody to MIDAS mutations in Ca V 1.2 channels assess the level of background staining; (c) Ca V ␤3 cells transfected with pCMV-Ca V 1.2 WT and pmCherry-Ca V ␣2␦1-HA WT, the latter serving as a quality control of transfection. Expressing mCherry-Ca V ␣2␦1-HA WT in HEKT cells produced a 1-log increase in the FITC (x axis) and a 3-log increase in mCherry fluorescence (y axis) on two-dimensional dot plots.
Flow cytometry data were analyzed using the FlowJo software, version 10 (TreeStar, Ashland, OR), as described in Ref. 39. Relative expression of Ca V ␣2␦1 was calculated based on ⌬median fluorescence intensity (⌬MedFI) for each fluorophore (mCherry or FITC). ⌬MedFI for FITC measured in intact nonpermeabilized cells was used as a relative index of the steadystate cell-surface expression of the HA-tagged Ca V ␣2␦1, whereas the ⌬MedFI for mCherry attested that the protein was translated until the C terminus. ⌬MedFI values were normalized to the maximum value measured the same day for mCherry-Ca V ␣2␦1-HA WT expressed under the same conditions. The normalized ⌬MedFI values for mCherry measured for each mutant in intact and permeabilized cells were not significantly different from one another (p Ͼ 0.1) (data not shown) suggesting that the cell permeabilization procedure did not distort significantly the relative fluorescence readout under most conditions.

Patch-clamp experiments in HEKT cells
Whole-cell patch-clamp experiments were carried out on isolated cells after transfection in HEKT cells in the presence of the peGFP vector coding for the green fluorescence protein (GFP) (0.2 g) as a control for transfection efficiency. Only the GFP-positive cells were patched. Electrodes were filled with a solution containing (in mM): 140 CsCl; 0.6 NaGTP; 3 MgATP; 10 EGTA; 10 HEPES; titrated to pH 7.3 with NaOH with a resistance varying between 2.8 and 3.2 megohms. Cells were bathed in a modified Earle's saline solution (in mM): 135 NaCl; 20 TEACl; 2 CaCl 2 ; 1 MgCl 2 ; 10 HEPES, titrated to pH 7.3 with KOH. On-line data acquisition was achieved with the Axopatch 200-B amplifier (Molecular Devices, Sunnyvale, CA) connected with the PClamp software Clampex 10.5 through the Digidata 1440A acquisition system (Molecular Devices) (19). A series of 450-ms voltage pulses were applied from a holding potential of Ϫ100 mV at a frequency of 0.2 Hz, from Ϫ60 to ϩ70 mV at 5-mV intervals. Series resistance was compensated to ϳ85% after on-line capacitive transient cancellation. Unless stated otherwise, whole-cell currents were sampled at 5 kHz and filtered at 1 kHz. PClamp software Clampfit 10.4 was used for data analysis. The mid-potential of activation values (E 0.5, act ) was estimated from the peak I-V curves obtained for each channel composition and was reported as the mean of individual measurements Ϯ S.E. (22). The free energy of activation was calculated using the mid-activation potential shown in Equation 1, where z is the effective charge displacement during activation, and F is the Faraday constant. The r100 ratio, defined as the ratio of peak whole-cell currents remaining after a depolarizing pulse of 100 ms (I 100 ms /I peak ), was calculated for each mutant.
As there was no significant change in the channel kinetics, these values were not reported herein. To assess for internal consistency, the experiments carried out with novel mutants systematically included a control experiment performed with pmCherry-Ca V ␣2␦1-HA WT (pCMV-Ca V 1.2 WT ϩ pCMV-Ca V ␤3 ϩ pmCherry-Ca V ␣2␦1-HA WT) thus explaining the larger sample size for Ca V ␣2␦1 WT. Previous experiments confirmed that mCherry-Ca V ␣2␦1-HA WT sustains the functional modulation of Ca V 1.2 currents (22). Experiments performed under the same conditions yielded peak current densities that could vary by as much as Ϯ 45% between each series of transfections. This variation appeared to be essentially linked to minor changes in the cell density at the time of transfection. Data from all experiments performed under the same conditions were pooled, and biophysical properties are reported in Table 2. Experiments were performed at room temperature (22°C).

3D homology modeling
The atomic coordinates of the Ca V ␣1 protein from Ca V 1.1 (PDB 5GJV) were used to explore the 3D structure of Ca V 1.2. The primary sequence of the transmembrane segment S1 to S4 in repeat I (residues 32-179) of the rabbit Ca V 1.2 (GenBank TM X15539) shares 69.6% identity with the same region of the rabbit Ca V 1.1. The primary sequence of the VWA domain of the rat Ca V ␣2␦1 (our construct, GenBank TM NM_012919) between amino acids 249 and 439 is perfectly conserved between both species. Amino acids 135-281 in Ca V 1.2 and amino acids 249 -439 of the rat Ca V ␣2␦1 were simultaneously aligned to the atomic coordinates of PDB 5GJV using the align2D algorithm in Modeler 9.17 (43). Modeler 9.17 was used to generate 100 3D models, and the models with the lowest values for the molpdf parameters and Discrete Optimized Protein Energy (DOPE) score (44) were selected. The DOPE parameter is a statistical potential used to access the energy of the protein model generated through many iterations by Modeler, which produces homology models by the satisfaction of spatial restraints.
The extracellular loop linking helix S3 and helix S4 was missing in the cryo-EM structure and was reconstructed using the "automatic loop refinement" protocol in Modeler with slow refinement level parameter. Of the more than 100 3D models generated that contain the S3-S4 loop, we selected the "best model" with the lowest DOPE score and molpdf parameters to conduct MD simulations.

Molecular dynamics simulations
MD simulations of the Ca V 1.2-Ca V ␣2␦1 interface were conducted using GROMACS version 2016.3, single-precision (45). We used the 3D model (Fig. 1) as the input structure, and a Ca 2ϩ ion was virtually inserted into the MIDAS site. The ion was initially positioned at the geometric center of the negatively polarized/charged groups of the five MIDAS residues. The side-chain oxygen position was used for the serine and threonine amino acids, and the midpoint of the two side-chain oxygen atoms was used for the two aspartic acid residues. This crude positioning was deemed sufficient because the ion was not constrained during energy minimization and could thus adopt a more favorable position.

MIDAS mutations in Ca V 1.2 channels
The transmembrane regions S1 to S4 from repeat I of Ca V 1.2 (IS1S4) were inserted in a membrane bilayer consisting of 128 DPPC molecules. The bilayer model was also obtained from Dr. Peter Tieleman's website in accordance with the protocol developed by Berger et al. (46). The insertion in the membrane was done by aligning hydrophobic residues in the bilayer and comparing the result with the cryo-EM electron density map to optimize the protein positioning. The membrane was then packed around the proteins using the InflateGRO protocol (47).

Minimization, equilibration, and production
The GROMOS53A6 force field (48) was used for the simulations, with the parameters for the DPPC molecules obtained from Dr. Peter Tieleman's website as described above. Solvation was done using the InflateGRO protocol with temporarily enlarged van der Waals radii for carbon atoms to prevent water molecules from being inserted in the membrane (1.5 Å is the standard value, 3.75 Å was used). A saline solution containing the physiological concentrations 143 mM NaCl, 2 mM CaCl 2 , and 2 mM MgCl 2 was used, and the protein charge was neutralized with Cl Ϫ ions. The equilibration phases and production simulations were conducted at 323 K, a value slightly above the phase transition of DPPC, which was measured to be at 315 K (42). This ensured proper fluidity as opposed to gel-like behavior under the phase-transition temperature.
Energy minimization was conducted with a target maximal force (F max ) of no more than 100 kJ⅐mol Ϫ1 ⅐nm Ϫ1 , a maximal number of steps of 50,000, and a step size of 0.1 Å. F max converged to the desired value in less than 750 steps for all the trajectories presented in this work. Equilibration was first conducted using an NVT ensemble (constant Number of particles, Volume, and Temperature) for 100 ps, followed by a second phase using an NPT ensemble (constant Number of particles, Pressure, and Temperature) for 1 ns. Production runs of 25 ns were then computed with a 2-fs time step, saving coordinates, energy, and velocities every 25 ps.
For the Ca V ␣2␦1 WT and Ca V ␣2␦1 D259A models of the Ca V 1.2-Ca V ␣2␦1 interface, the above steps were repeated for five replicates. Because large regions of the Ca V 1.2 protein are missing in our 3D model, we kept for each form of the complex the three trajectories that showed the least structural divergence across the duration of the 25-ns trajectory as determined by their r.m.s.d. values compared with the initial structure (data not shown).

Statistics
Results were expressed as mean Ϯ S.E. Tests of significance were carried out using the unpaired one-way ANOVA with the Tukey test embedded in the Origin 7.0 analysis software (OriginLab Corp., Northampton, MA). Unless stated otherwise, data were considered statistically significant at p Ͻ 0.05.