The metal-ion-dependent adhesion site in the Von Willebrand factor-A domain of α₂δ subunits is key to trafficking voltage-gated Ca²+ channels

Abstract

All auxiliary α₂δ subunits of voltage-gated Ca²⁺ (Ca_V) channels contain an extracellular Von Willebrand factor-A (VWA) domain that, in α₂δ-1 and -2, has a perfect metal-ion-dependent adhesion site (MIDAS). Modeling of the α₂δ-2 VWA domain shows it to be highly likely to bind a divalent cation. Mutating the three key MIDAS residues responsible for divalent cation binding resulted in a MIDAS mutant α₂δ-2 subunit that was still processed and trafficked normally when it was expressed alone. However, unlike WT α₂δ-2, the MIDAS mutant α₂δ-2 subunit did not enhance and, in some cases, further diminished Ca_V1.2, -2.1, and -2.2 currents coexpressed with β1b by using either Ba²⁺ or Na⁺ as a permeant ion. Furthermore, expression of the MIDAS mutant α₂δ-2 reduced surface expression and strongly increased the perinuclear retention of Ca_Vα1 subunits at the earliest time at which expression was observed in both Cos-7 and NG108–15 cells. Despite the presence of endogenous α₂δ subunits, heterologous expression of α₂δ-2 in differentiated NG108–15 cells further enhanced the endogenous high-threshold Ca²⁺ currents, whereas this enhancement was prevented by the MIDAS mutations. Our results indicate that α₂δ subunits normally interact with the Ca_Vα1 subunit early in their maturation, before the appearance of functional plasma membrane channels, and an intact MIDAS motif in the α₂δ subunit is required to promote trafficking of the α1 subunit to the plasma membrane by an integrin-like switch. This finding provides evidence for a primary role of a VWA domain in intracellular trafficking of a multimeric complex, in contrast to the more usual roles in binding extracellular ligands in other exofacial VWA domains.

Voltage-gated Ca²⁺ (Ca_V) channels are composed of a poreforming α1 subunit that determines the main biophysical properties of the channel. For the Ca_V1 and -2 subfamilies, this subunit is associated with an intracellular β subunit (for review, see refs. 1 and 2) and a membrane-anchored, predominantly extracellular α₂δ subunit (for review, see ref. 3). Mammalian genes encoding 10 α1, 4 β, and 4 α₂δ subunits have been identified (for reviews, see refs. 2 and 4). The topology of the α₂δ protein has been determined in detail only for α₂δ-1 but is thought to generalize to all 4 α₂δ subunits (for review, see ref. 3). All α₂δ subunits have predicted N-terminal signal sequences, indicating that the N terminus is extracellular. In early studies of α₂δ-1 purified from skeletal and cardiac muscle, it was determined that the α₂ subunit is disulfide-bonded to a transmembrane δ subunit, and both subunits are the products of a single gene, encoding the α₂δ protein, that is posttranslationally cleaved into α₂ and δ (5).

Subsequent to the identification of α₂δ subunits as stoichiometric components of skeletal muscle Ca²⁺ channels, α₂δ subunits have also been shown to be associated with native cardiac (L-type) (6) and neuronal N- and P/Q-type channels (7, 8). In coexpression studies where it has been tested, all α₂δ subunits enhance all Ca_V1 and -2 currents. The α₂δ subunits also influence the channel's biophysical properties, including inactivation kinetics and voltage-dependence (9); therefore, the effects of the α₂δ subunits are not limited to trafficking α1 subunits, but their mechanism of action remains largely unknown.

The α₂δ-1 subunit has been shown to bind to extracellular regions, including Domain III on Ca_Vα1 subunits (10, 11). It is unclear what domains of α₂δ are involved in these interactions, but all α₂δ subunits contain a Von Willebrand factor-A (VWA) domain within the α₂ moiety (12). This domain is also present in integrins and is often involved in binding extracellular matrix proteins (12, 13). VWA domains contain a sequence motif representing a metal-ion-dependent adhesion site (MIDAS) that confers divalent metal (usually Mg²⁺)-dependent binding to the ligand (14). In α₂δ-1 and -2, this motif is a perfect MIDAS motif, containing both the DxSxS motif and noncontiguous T and D residues (12), with the T in loop 3 being part of a TDG motif (15), suggesting that α₂δ-1 and -2 can both undergo an integrin-like switch and bind ligand in the presence of a divalent cation.

In this study, we have investigated the importance of the VWA domain MIDAS in the functional effects of α₂δ-2.

Experimental Procedures

Structural Modeling and CD. Suitable templates for modeling were selected from available structures by using the program fugue (16).

Construction and Heterologous Expression of cDNAs Cell Culture and Immunocytochemistry. Standard molecular biological and cell biological techniques were used, as described in refs. 17 and 18.

Biochemistry and Imaging. Cell-surface proteins in intact cells were biotinylated by using Sulfo-NHS-SS-Biotin (Pierce) for 30 min at room temperature. Standard techniques were used for immunoprecipitation, immunoblotting, and immunocytochemistry. Images were obtained by using a Zeiss LSM confocal microscope and further analyzed with imagej software (National Institutes of Health, Bethesda). The gabapentin-binding assay was performed by using a method similar to that described in ref. 19.

Electrophysiology. Standard techniques were used, essentially as described in ref. 17. Details are given in Supporting Methods, which is published as supporting information on the PNAS web site.

Further details of all methods are given in Supporting Methods. Data are mean (±SEM), and statistical significances were analyzed by using Student's t test for unpaired data.

Results

Structural Modeling of the VWA Domain of α₂δ-2. Before investigating the function of the predicted VWA domain of α₂δ-2, we first determined the likelihood that α₂δ-2 is able to fold into a VWA domain. Suitable templates for modeling of α₂δ VWA domains (residues 253–430 of α₂δ-1 and residues 294–472 of α₂δ-2) were selected from available structures. The 1.5-Å crystal structure of the VWA domain from capillary morphogenesis protein 2 (CMG2) in the open (ligand-competent) state and containing a Mg²⁺ ioninits MIDAS (20), was the best template for both α₂δ-1 and -2 VWA domains, with z scores of 15.62 and 18.85, respectively, indicating very high levels of certainty that this is an appropriate template structure. The pairwise sequence identity between each of the α₂δ subunits and CMG2 is only ≈16% within the VWA region, but the residues of the MIDAS motif are conserved (see Fig. 6, which is published as supporting information on the PNAS web site). Models (Fig. 1a) were generated according to structural alignments, based on the CMG2 structure. A Mg²⁺ ion was replaced manually within the MIDAS.

Fig. 1.

Mutation of the MIDAS in α₂δ-2 does not prevent its trafficking to the plasma membrane. (a) Model of the α₂δ-2 VWA domain. Residues comprising the MIDAS motif are labeled and shown as stick representations. The Mg²⁺ ion (purple) is coordinated within the MIDAS and is partially exposed at the surface in an appropriate manner for interaction with a second protein. α-helices are colored in cyan and β-strands in yellow. (b) The main domains in the primary sequence of α₂δ-2, showing the extent of the VWA domain (black), and the three MIDAS mutations (white asterisks). SS, signal sequence. The amino acids making up the δ subunit are indicated by hatched boxes. (c) Cos-7 cells were transfected with either α₂δ-2 or α₂δ-2 μMIDAS, as indicated, and cell-surface proteins were biotinylated. (Left) Total expression in whole-cell lysate (WCL). (Right) Pull-down of biotinylated proteins on a 3–8% Tris acetate gel. (Upper) α₂-2 (residues 102–117) Ab was used for immunoblotting (IB). (Lower) Anti-Akt Ab was used for IB to show that no intracellular proteins were biotinylated. The percentages of α₂δ-2 and α₂δ-2 μMIDAS at the cell surface were 10.0% and 9.5%, respectively (representative of three independent experiments; see Fig. 9 legend for calculation details). (d) Biotinylated α₂δ-2 (Left) or α₂δ-2 μMIDAS (Right), as in c, were treated (+) or not (–) with endoglycosidase F (5 units) and separated on a 7% Tris acetate gel. The upper arrow indicates fully glycosylated α₂δ-2, and the lower arrow indicates deglycosylated α₂δ-2. (e) Localization of α₂δ-2 and α₂δ-2 μMIDAS 48 h after transfection in Cos-7 cells that were either nonpermeabilized (Left) or permeabilized (Right) before immunostaining with the primary Abs shown. (Top) α₂-2 (residues 102–117) Ab. (Middle) δ₂-2(residues 994-1009)Ab. (Bottom) Signal peptide α₂-2 (residues 16 –29) Ab. (Scale bar, 20 μm.)

Mutation of the MIDAS Residues of α₂δ-2 Do Not Prevent Its Trafficking, Processing Glycosylation, or Expression at the Cell Surface. The results of the modeling study encouraged us to examine the role of the VWA domain in α₂δ-2 by mutating to Ala the three divalent metal-coordinating amino acids D300, S302, and S304 in the MIDAS sequence DxSxS (α₂δ-2 μMIDAS, Fig. 1b). Mutation of MIDAS motifs in integrins has been used previously to probe the cell-surface role of ligand-binding to these sites (21). We found that mutation of the MIDAS motif did not affect the total expression levels of these constructs, compared with WT α₂δ-2, when transfected into tsA-201 (data not shown) or Cos-7 cells (Fig. 1c). In addition, α₂δ-2 μMIDAS was expressed at the cell surface in Cos-7 cells to the same extent as WT α₂δ-2, as determined by cell-surface biotinylation (Fig. 1c). The biotinylated α₂δ-2 and α₂δ-2 μMIDAS had the same molecular mass and were reduced in mass to the same extent by endoglycosidase F, suggesting that they are equally glycosylated (Fig. 1d). Cell-surface immunoreactivity in nonpermeabilized cells was observed for both constructs, with not only the anti-peptide α₂-2 (102–117) Ab (Fig. 1e) but also an Ab against amino acids in δ-2 (994–1009) (Fig. 1e). No increase in intracellular retention was observed for α₂δ-2 μMIDAS, compared with WT α₂δ-2 in permeabilized cells (Fig. 1e). Further evidence of the correct processing of α₂δ-2 μMIDAS is shown by the cleavage of its N-terminal signal peptide, because no surface staining was observed with the α₂-2 (16–29) signal peptide Ab for either construct (Fig. 1e), whereas extensive staining was observed in the perinuclear region in permeabilized cells (Fig. 1e).

It is unlikely, based on previous work on integrins (21), that mutation of MIDAS amino acids will cause the α₂δ-2 VWA domain to misfold, and, indeed, we have found that GST-fusion proteins of the WT and MIDAS mutant VWA domains of α₂δ-2 expressed and purified from Escherichia coli have very similar CD spectra. The model (Fig. 1a) predicts 40% α-helix, 22% β-sheet, and 38% other structure. The values obtained from CD measurements, after subtraction of the GST spectra, were 32%, 20%, and 48%, respectively [normalized root-mean-square deviation (NRMSD), 0.073], for the WT α₂δ-2 VWA domain and 31%, 23%, and 46% (NRMSD, 0.06) for the α₂δ-2 μMIDAS VWA domain. Furthermore, WT α₂δ-2 and α₂δ-2 μMIDAS exhibit similar affinity for binding of [³H]gabapentin (K_D of 163.3 ± 8.7 nM and 166.1 ± 1.0 nM, respectively, n = 3 experiments), also indicative of correct folding of the full-length α₂δ-2 μMIDAS. The K_D for WT α₂δ-2 is in agreement with data published in ref. 19.

Effect of Mutation of the MIDAS Motif in the VWA Domain on Modulation of Ca²⁺-Channel Currents by α₂δ-2. The amplitude of the very small Ba²⁺ currents (I_Ba) through Ca_V2.2/β1b channels, expressed in tsA-201 cells, was enhanced ≈5-fold by WT α₂δ-2 (Fig. 2 a–c) and by α₂δ-2 with a C-terminal myc-His tag (Fig. 2a). In contrast, α₂δ-2 μMIDAS produced no stimulation of Ca_V2.2 currents (Fig. 2 a and b). Furthermore, steady-state inactivation was slightly hyperpolarized by α₂δ-2, with an increase in voltage-dependence compared with the Ca_V2.2/β1b combination, but this effect did not occur for α₂δ-2 μMIDAS (Fig. 2c). A similar result was obtained by using HA-tagged Ca_V1.2 (22) (see Fig. 7 a–c, which is published as supporting information on the PNAS web site). In cerebellar Purkinje cells, we have shown that α₂δ-2 is the predominant α₂δ subtype (17), and, here, the main α1 and β subunits are Ca_V2.1 and β4 (23, 24). We therefore compared the effect of α₂δ-2 and α₂δ-2 μMIDAS on Ca_V2.1/β4 currents. Whereas α₂δ-2 enhanced the Ca_V2.1/β4 current by 6.8-fold ± 0.6 (n = 10), there was no significant effect of α₂δ-2 μMIDAS (data not shown).

Fig. 2.

Comparison of the effect of α₂δ-2 and the MIDAS mutant construct of α₂δ-2 on Ca_V currents in tsA-201 cells. Ca_V2.2/β1b was expressed with the various α₂δ-2 constructs in tsA-201 cells by using 10 mM Ba²⁺as charge carrier (see Supporting Methods). (a) Representative current traces elicited between –20 and +40 mV in 10-mV steps from a holding potential (HP) of –90 mV for Ca_V2.2/β1b (Left), Ca_V2.2/β1b/α₂δ-2 (Center Left), Ca_V2.2/β1b/α₂δ-2 myc-His (Center Right), and Ca_V2.2/β1b/α₂δ-2 μMIDAS (Right). (b) Current–voltage (I–V) relationships for three of the experimental conditions. ○, Ca_V2.2/β1b (n = 8); •, +α₂δ-2 (n = 10); and ▵, +α₂δ-2 μMIDAS (n = 11). (Inset) Bar chart at +20 mV (calculated as a percentage of the mean control Ca_V2.2/β1b currents) for +α₂δ-2 (back bar) and +α₂δ-2 μMIDAS (hatched bar). The number of determinations is as described above. (c) Steady-state inactivation curves for test pulses to +20 mV from a 15-s conditioning prepulse of between –100 and 0 mV. ○, Ca_V2.2/β1b (n = 9); •, +α₂δ-2 (n = 7); and ▵, +α₂δ-2μMIDAS (n = 5). The data were fit with a single Boltzmann equation and the mean voltages at which the channel is 50% available (V₅₀) for inactivation were –42.4, –45.9, and –42.0 mV, respectively, and the slope factors (k) were 11.1, 7.2, and 9.8 mV, respectively. (d) Ca_V2.2/β1b was expressed with the various α₂δ-2 constructs and 100 mM Na⁺ was used as charge carrier (see Supporting Methods). Representative current traces were elicited between –40 and +15 mV in 5-mV steps from a holding potential of –90 mV for Ca_V2.2/β1b (Left), Ca_V2.2/β1b/α₂δ-2 myc-His (Center), and Ca_V2.2/β1b/α₂δ-2 μMIDAS (Right). (e) I–V relationships for the three conditions. ○, Ca_V2.2/β1b (n = 16); □, +α₂δ-2 myc-His (n = 17); and ▵, +α₂δ-2 μMIDAS (n = 15). (f) Bar chart at –10 mV (calculated as a percentage of the mean control Ca_V2.2/β1b Na⁺ currents) for +α₂δ-2 myc-His (open bar) and +α₂δ-2 μMIDAS (hatched bar). The number of determinations is as in e.

In Xenopus oocytes, the amplitude of Ca_V2.2/β1b currents was enhanced ≈3-fold by both WT α₂δ-2 and α₂δ-2 myc-His (see Fig. 7 d and e). In contrast, after mutation of the MIDAS amino acids, there was no enhancement of the current amplitude, and, indeed, there was a clear reduction, compared with Ca_V2.2/β1b alone (see Fig. 7 d and e). Furthermore, whereas α₂δ-2 hyperpolarized the voltage-dependence of steady-state inactivation, α₂δ-2 μMIDAS did not (see Fig. 7f). All data obtained with α₂δ-2 μMIDAS could be replicated by using an α₂δ-2 construct in which the VWA domain was deleted (data not shown).

We detected no endogenous α₂δ subunit protein in Xenopus oocytes (see Fig. 8a, which is published as supporting information on the PNAS web site) or Cos-7 cells (data not shown), whereas endogenous α₂δ-1 was detected in tsA-201 cells. The α₂δ-1 had the same molecular mass and was glycosylated to the same extent as was α₂δ-1 in brain (see Fig. 8b). No endogenous α₂δ-2 was observed in any of the cell types used (data not shown).

Does the Effect of Mutation of the MIDAS on Ca²⁺-Channel Currents Involve Binding of Extracellular Divalent Cations? Because MIDAS motifs bind divalent cations (Fig. 1a), we wished to examine whether the effect of mutation of the MIDAS affected channel permeation. We therefore removed all divalent cations from the external medium and measured Na⁺ flux through Ca_V2.2 channels in tsA-201 cells. This protocol produced the same result: the amplitude of the Ca_V2.2/β1b Na⁺ currents was enhanced >4-fold by α₂δ-2 but not by α₂δ-2 μMIDAS, whose currents were slightly smaller than those in the absence of any α₂δ-2 (Fig. 2 d–f). Therefore, if Mg²⁺ (or Ca²⁺) binding to the MIDAS in α₂δ-2 is involved in its function in relation to enhancing currents through Ca_Vα1 channels, it must be during the process of channel assembly or trafficking rather than during the permeation process, once the channels have reached the cell surface.

Does the MIDAS Mutant α₂δ-2 Affect Trafficking of Ca_Vα1 Subunits? Given these results, the complete loss of function of the α₂δ-2 μMIDAS mutant subunit, despite its normal expression compared with WT α₂δ-2, might, therefore, be due to interference with trafficking and a subsequent reduction of plasma-membrane expression of associated Ca_Vα1 subunits. Using cell surface biotinylation, we found a decrease in the percentage of Ca_V2.2 at the cell surface, when coexpressed with α2δ-2 μMIDAS, compared with WT α₂δ-2 by ≈50% (see Fig. 9a, which is published as supporting information on the PNAS web site) and also a 30% decrease in the percentage of α2δ-2 μMIDAS, compared with WT α₂δ-2 at the cell surface, when coexpressed with Ca_V2.2 (see Fig. 10b, which is published as supporting information on the PNAS web site). This result was not the case when the α₂δ subunits were expressed alone (Fig. 1c).

To examine further the basis for the reduction of surface expression of Ca_Vα1 subunits, we determined the subcellular distribution of HA-Ca_V1.2, expressed with β1b in Cos-7 cells in the presence of either α₂δ-2 or α₂δ-2 μMIDAS. When HA-Ca_V1.2 was expressed with WT α₂δ-2, its distribution was fairly uniform (Fig. 3a; and see Fig. 10 for an additional example). The immunostaining extended to the periphery of the cell, where it colocalized with α₂δ-2, as seen clearly on the line scan (representative of 11 of 11 cells analyzed in three experiments, see Fig. 3b). Further evidence of colocalization was obtained from pixel-intensity-correlation plots (Fig. 3c).

Fig. 3.

Comparison of the effect of α₂δ-2 and α₂δ-2 μMIDAS on expression of Ca_V1.2 channels and colocalization with the α₂δ construct. (a and d) Localization of α₂δ-2 or α₂δ-2 μMIDAS (α₂ (102–117) Ab (Left), cell outline shown by dotted white line), HA-Ca_V1.2 (HA Ab, Center), and overlay (Right), including nuclear stain (DAPI, blue) in Ca_V1.2/β1b/α₂δ-2 (a) or α₂δ-2 μMIDAS-transfected Cos-7 cells (d). (Scale bar, 50 μm.) (b and e) Normalized pixel intensity of the line scan given on the image. The red line is HA-Ca_V1.2, and the green line is α₂δ-2 (b) or α₂δ-2 μMIDAS (e). The arrows in e indicate regions where Ca_V1.2 immunoreactivity is absent from the periphery of the cell. (c and f) Pixel-intensity-correlation plot for nonzero pixels in the entire image, shown for α₂δ-2 (c) or α₂δ-2 μMIDAS (f) (green, y axis) vs. Ca_V1.2 (red, x axis). A pixel-number calibration bar is shown on each plot. * in f indicates an additional region of high-intensity colocalization. The diagonal dotted line indicates theoretical colocalization. (a–c) Cos-7 cells HA-Ca_V1.2/β1b/α₂δ-2. (d–f) HA-Ca_V1.2/β1b/α₂δ-2 μMIDAS.

In contrast, when HA-Ca_V1.2 was coexpressed with α₂δ-2 μMIDAS, areas of marked intracellular retention were observed, most often of both the Ca_Vα1 and the α₂δ subunit (Fig. 3d; and see Fig. 10 for an additional example). Furthermore, in 9 of 14 cells examined, HA-Ca_V1.2 immunoreactivity was observed not to extend as far as α₂δ-2 μMIDAS into the periphery of the cell (Fig. 3e). The strong intracellular retention of HA-Ca_V1.2, together with α₂δ-2 μMIDAS, can be seen clearly on the pixel-intensity-correlation plot as an additional peak of high intensity in both red and green fluorescence (Fig. 3f, marked with an asterisk). In agreement with these results, HA-Ca_V1.2 is able to coimmunoprecipitate with α₂δ-2 μMIDAS and WT α₂δ-2 (Fig. 9c).

We then used an N-terminal GFP-Ca_V2.2 construct, which we have previously shown exhibits normal functional properties in Cos-7 cells (25). We examined expression at a very early time point, 24 h after transfection, before the time at which Ca²⁺ currents could reliably be recorded (data not shown). Compared with coexpression with WT α₂δ-2, where the distribution of GFP-Ca_V2.2 was fairly uniform, when coexpressed with α₂δ-2 μMIDAS, there was a significantly greater intracellular accumulation of GFP-Ca_V2.2, particularly in the perinuclear region (Fig. 4 a, b, and d). The accumulation was also more pronounced than that seen with GFP-Ca_V2.2, in the absence of any α₂δ (Fig. 4 c and d). The same result was observed 48 h after transfection (see Fig. 11, which is published as supporting information on the PNAS web site).

Fig. 4.

Comparison of the effects of α₂δ-2 and α₂δ-2 μMIDAS on the distribution of GFP-Ca_V2.2 channels in Cos-7 cells. (Left) Localization of GFP-Ca_V2.2 in Cos-7 cells when coexpressed with β1b and either α₂δ-2 (a) or α₂δ-2 μMIDAS (b) and without α₂δ (c) 24 h after transfection. White dotted lines correspond to the positions of the line scans shown to the right. (Scale bars, 20 μm.) (Right) Line scans of GFP-Ca_V2.2 fluorescence were analyzed through the nucleus of the cells shown. The horizontal dotted lines show the 1,500 intensity threshold (12-bit images), and the line beneath the scans shows the extent of the cell. (d) Cumulative frequency histogram of the percentage of the line scan within the cell above the 1,500 intensity threshold for α₂δ-2-(n = 26, black line), α₂δ-2 μMIDAS-containing cells (n = 32, red line), and in the absence of α₂δ (n = 32, green line). The histogram shows the greater proportion of high-intensity GFP-Ca_V2.2 fluorescence regions in α₂δ-2 μMIDAS (mean 33.2 ± 3.1%), compared with both α₂δ-2-containing cells (mean 10.1 ± 2.4%; P < 0.0001) and in the absence of α₂δ (mean 21.2 ± 3.2%; P < 0.01, compared with α₂δ-2 μMIDAS).

Effect of α₂δ-2 and α2δ-2 μMIDAS on Expression of Ca²⁺-Channel Currents in Differentiated NG108–15 Cells. The NG108–15 neuroblastoma–glioma hybrid cell line represents a model for neurons, because it produces extensive neurites and shows high-voltage-activated (HVA) currents, when differentiated. These currents were isolated by maintaining a holding potential of –40 mV. Despite the presence of endogenous α₂δ-1 in NG108–15 cells (26), exogenous expression of α₂δ-2 enhanced the HVA currents substantially. I_Ba recorded from differentiated cells transfected with α₂δ-2, was increased to 264% of control (Fig. 5 a and b), indicating that the amount of endogenous α2δ is suboptimal for the expression of maximal endogenous Ca²⁺ currents. In contrast, transfection with α₂δ-2 μMIDAS produced no enhancement of I_Ba (110% of control, Fig. 5 a and b).

Fig. 5.

Comparison of the effects of α₂δ-2 and α₂δ-2 μMIDAS on endogenous Ca²⁺ currents and GFP-Ca_V2.2 distribution in NG 108–15 cells. (a) The effects of α₂δ-2 and α₂δ-2 μMIDAS on HVA I_Ba in NG108–15 cells. Current–voltage relationships for NG108–15 cells transfected with GFP-pRK5 and Kir2.1-AAA (○, control, n = 13), •, α₂δ-2 myc-His (n = 12), or ▾, α₂δ-2 μMIDAS (n = 13), all in pcDNA3.1. Holding potential = –40 mV. (b) Sample traces for the three conditions shown in a. (Top) Control. (Middle) +α₂δ-2. (Bottom) +α₂δ-2 μMIDAS. Traces shown are from –40 to +60 mV. (c and d) Localization of GFP-Ca_V2.2 48 h after transfection, cotransfected with β1b and α₂δ-2 myc-His (c) or α₂δ-2 μMIDAS (d). (Top Left) GFP fluorescence. (Top Right) Line scan of the dotted line shown on the 12-bit image. (Bottom Left) Enlarged region shown by the box, converted to 8-bit pseudocolor to show intensity in the neurite. The intensity calibration bars are shown on the right. (e) The plots represent a cumulative frequency histogram for all line scans, obtained 24 h after transfection, showing the significantly greater percentage of high-intensity GFP-Ca_V2.2-fluorescence regions in α₂δ-2 μMIDAS-containing cells (red line, mean 42.2 ± 6.1%, n = 20, P < 0.0005) in comparison with α₂δ-2-containing cells (black line, 14.6 ± 3.7%, n = 23). An additional control is included, in which a control nonfunctional protein, Kir2.1-AAA, has been included instead of α₂δ-2 (green, 15.6 ± 4.8%, n = 20, P < 0.001, compared with α₂δ-2 μMIDAS). (f) Proposed role of the MIDAS in α₂δ. ER, endoplasmic reticulum.

When NG108–15 cells were transfected with GFP-Ca_V2.2/β1b and WT α₂δ-2 and differentiated, uniform distribution of GFP-Ca_V2.2 was observed (Fig. 5c). The cells were clearly differentiated with neurites (Fig. 5d, enlarged region). In sharp contrast, coexpression with α₂δ-2 μMIDAS produced strong perinuclear retention of GFP-Ca_V2.2/β1b, although the cells also had a differentiated phenotype, showing neurites (Fig. 5d). Analysis of the amount of signal above an arbitrary threshold on line scans (Fig. 5 c and d) showed that the intracellular retention of GFP-Ca_V2.2/β1b was significantly greater in the presence of α₂δ-2 μMIDAS than in the presence of WT α₂δ-2, at both 48 h (data not shown) and 24 h (Fig. 5e). In an additional control experiment, coexpression of GFP-Ca_V2.2/β1b with a control transmembrane protein Kir2.1-AAA (27), a K⁺ channel with a mutation in the pore, such that it does not conduct current, does not produce any intracellular retention of Ca_V2.2/β1b (Fig. 5e). We also showed that GFP-Ca_V2.2 was expressed similarly under all conditions (see Fig. 12, which is published as supporting information on the PNAS web site).

Discussion

The Functional Interaction of Ca_Vα1/β Channels with α₂δ-2. Previous in vitro studies have shown that all α₂δ subunits increase the maximum conductance of a number of expressed Ca²⁺-channel α1/β subunit combinations at the whole-cell level (1, 17, 28–30). We have previously investigated the effect of α₂δ-2 on the Ca_V2.1/β4 Ca²⁺-channel subunit combination and showed that it had no effect on single-channel conductance (17, 18). This finding implies that α₂δ-2 probably has its main effect on the lifetime of the channel complex in the plasma membrane, by either enhancing forward trafficking or reducing turnover of the channel once inserted in the plasma membrane.

All integrin β subunits contain a VWA domain with a MIDAS motif (13). Based on structural studies, the inactive form of the integrin dimer is bent in a loose hairpin. When activated, the integrin dimers alter conformation and bind their ligands in the presence of divalent cations (usually Mg²⁺, see Fig. 1a; for review, see ref. 31). The single particle EM studies of Ca²⁺-channel complexes have also shown the associated α₂δ subunit to be bent (32–34). It was recently hypothesized that all VWA domains that contain a perfect MIDAS motif (DxSxS and a coordinating T and D) will both bind Mg²⁺ and subsequently undergo an integrin-like switch (15). It is tempting to speculate that the Ca_Vα1 subunit represents an endogenous ligand for the α₂δ MIDAS and that an integrin-like switch is required for trafficking the channel complex.

Mechanism of Inhibition of Trafficking by the MIDAS Mutant of α₂δ-2. One hypothesis for the role of α₂δ subunits is that they act as vestibules to increase Ca²⁺ availability at the mouth of the pore (35), and, therefore, mutating the MIDAS so that this motif is nonfunctional might reduce the conductance. If this were the mechanism, the effect should be lost when using Na⁺ as the charge carrier. However, we have shown that exactly the same effect of α₂δ-2 μMIDAS occurs under these conditions as in the presence of divalent cations.

This finding leaves two further hypotheses. First, correct binding of a divalent cation to the α₂δ MIDAS in the lumen of the endoplasmic reticulum or other intracellular compartment may be essential for the trafficking of Ca_Vα1 subunits to the plasma membrane. Several pieces of our evidence support this hypothesis, pointing to a defect in Ca_Vα1-subunit trafficking in the presence of α₂δ-2 μMIDAS, caused by an intracellular interaction between the two proteins. We observed marked intracellular retention of HA-Ca_V1.2 and intracellular colocalization with α₂δ-2 μMIDAS, whereas when α₂δ-2 μMIDAS is expressed alone, it is processed and trafficked to the cell membrane normally. A similar phenomenon of intracellular retention was observed for GFP-Ca_V2.2inlive Cos-7 and NG108–15 cells. We have also observed less Ca_V2.2 at the plasma membrane by cell-surface biotinylation when α₂δ-2 μMIDAS is coexpressed, compared with WT α₂δ-2. There was also a smaller proportion of α₂δ-2 μMIDAS at the cell surface when this subunit is coexpressed with Ca_V2.2, compared with when it is expressed alone, whereas this was not the case for WT α₂δ-2.

The second hypothesis would be that the Ca_Vα1/β heterodimers and the α₂δ subunits are trafficked separately to the plasma membrane, where they then interact, allowing the α₂δ to increase the lifetime of the channel at the plasma membrane. A prediction based on this hypothesis is that there would be no interaction in the case of α₂δ-2 μMIDAS. Evidence against this hypothesis is (i) that the colocalization and coimmunoprecipitation of α₂δ-2 μMIDAS and Ca_Vα1 subunits indicates an interaction, albeit an abortive one, and (ii) that the intracellular retention of GFP-Ca_V2.2 was more marked in the presence of α₂δ-2 μMIDAS than in the absence of any α₂δ subunit, before the time at which Ca²⁺ currents could be recorded, indicating that an interaction must occur early in the process of assembly and trafficking. Thus, even in the absence of a functional MIDAS in α₂δ-2, an interaction occurs, presumably through the mutated VWA domain and, possibly, through other sites in α₂δ-2, but this interaction does not result in trafficking. We propose that the interaction between Ca_Vα1 and an α₂δ subunit requires the high divalent cation concentration within the endoplasmic reticulum (ER) to bind to the MIDAS in α₂δ and allow forward trafficking of the complex through an integrin-like switch (Fig. 5F). This switch may serve as a trafficking checkpoint, so that, if the Ca_Vα1 subunit is not folded correctly, both α1 and α₂δ will be retained within the ER.

Mechanism of Suppression of Ca_Vα1 Currents by MIDAS Mutant α₂δ-2. There is a strong suppressive effect of α₂δ-2 μMIDAS in Xenopus oocytes, whereas in NG108–15 cells and tsA 201 cells, only a lack of up-regulation of the endogenous HVA I_Ba by α₂δ-2 μMIDAS is seen. The clear difference that we have found between these expression systems is the presence of endogenous α₂δ-1 in NG108–15 cells (26) and in tsA 201 cells, whereas no endogenous α₂δ protein was detected in the Xenopus oocytes used in this study (Fig. 7). We therefore suggest that, in Xenopus oocytes, in the absence of any endogenous α₂δ subunits, α₂δ-2 μMIDAS interacts with, and causes intracellular retention of, expressed Ca_Vα1/β complexes, which otherwise produce substantial currents in this system, although not in mammalian cells, for reasons unknown (e.g., Fig. 2 a–c, compared with Fig. 7 d and e). In contrast, the channels underlying the very small Ca_Vα1/β I_Ba observed in tsA 201 cells may reach the plasma membrane only because they have an associated endogenous α₂δ. Furthermore, in NG108–15 cells, there is evidence that the up-regulation of endogenous Ca²⁺ currents that occurs upon differentiation results, at least in part, from the stabilization of preformed channels (36), and, thus, the channels making up the endogenous HVA current in NG108–15 cells are likely to already have an associated endogenous α₂δ-1. These endogenous channels would, therefore, be protected from association with the MIDAS mutant of α₂δ-2 and, thus, from further suppression by intracellular retention.

Conclusions

Our finding that α₂δ-2 μMIDAS increased the internal retention of both Ca_V1 and Ca_V2 channels, whereas it is normally processed and trafficked itself when expressed alone, indicates that the MIDAS motif of the VWA domain has its action on trafficking of the Ca_Vα1 subunit. Our results suggest that α₂δ subunits normally interact, through the intact MIDAS, with the Ca_Vα1 subunit at a very early time point in their maturation, to enhance trafficking. This finding provides evidence for a primary role of a VWA domain in the intracellular trafficking of a multimeric complex, in contrast to their more usual role in extracellular interactions.