Identification of a disulfide bridge essential for structure and function of the voltage-gated Ca²⁺ channel α₂δ-1 auxiliary subunit

Abstract

Voltage-gated calcium (Ca_V) channels are transmembrane proteins that form Ca²⁺-selective pores gated by depolarization and are essential regulators of the intracellular Ca²⁺ concentration. By providing a pathway for rapid Ca²⁺ influx, Ca_V channels couple membrane depolarization to a wide array of cellular responses including neurotransmission, muscle contraction and gene expression. Ca_V channels fall into two major classes, low voltage-activated (LVA) and high voltage-activated (HVA). The ion-conducting pathway of HVA channels is the α₁ subunit, which typically contains associated β and α₂δ ancillary subunits that regulate the properties of the channel. Although it is widely acknowledged that α₂δ-1 is post-translationally cleaved into an extracellular α₂ polypeptide and a membrane-anchored δ protein that remain covalently linked by disulfide bonds, to date the contribution of different cysteine (Cys) residues to the formation of disulfide bridges between these proteins has not been investigated. In the present report, by predicting disulfide connectivity with bioinformatics, molecular modeling and protein biochemistry experiments we have identified two Cys residues involved in the formation of an intermolecular disulfide bond of critical importance for the structure and function of the α₂δ-1 subunit. Site directed-mutagenesis of Cys404 (located in the von Willebrand factor-A region of α₂) and Cys1047 (in the extracellular domain of δ) prevented the association of the α₂ and δ peptides upon proteolysis, suggesting that the mature protein is linked by a single intermolecular disulfide bridge. Furthermore, co-expression of mutant forms of α₂δ-1 Cys404Ser and Cys1047Ser with recombinant neuronal N-type (Ca_V2.2α₁/β₃) channels, showed decreased whole-cell patch-clamp currents indicating that the disulfide bond between these residues is required for α₂δ-1 function.

1. Introduction

Calcium is an essential and versatile intracellular messenger. Normally submicromolar levels of Ca²⁺ within the cytoplasm are disrupted by Ca²⁺ increases that trigger diverse Ca²⁺-dependent responses [1]. Transient increases in cytosolic Ca²⁺ can be the result of entry from the extracellular space or release from intracellular stores. Ca²⁺ influx may be mediated by several molecules [1], including voltage-gated (Ca_V) channels which have been subdivided according to their activation threshold into (i) low voltage-activated (LVA or T-type) channels and (ii) high voltage-activated (HVA) channels, a class that includes the L-, N-, P/Q- and R-types [2,3]. Biochemical purification has revealed that HVA channels are composed of four proteins, including the pore-forming α₁ and auxiliary α₂δ, β and γ subunits [3–5].

The α₂δ subunit is a protein of ~170 kDa encoded by a single gene that is translated as a precursor polypeptide. The δ domain anchors the α₂ protein to the cell membrane via a single putative transmembrane segment. The α₂ domain is entirely extracellular [3,5–7], is heavily glycosylated and contains several functional domains. These include a ~180 amino acid von Willebrand factor-A (vWFA) domain similar to extracellular matrix-binding regions of integrins, a 5 amino acid metal ion-dependent adhesion site (MIDAS) motif [8] and a poorly understood 92 amino acid Cache (Ca²⁺ channels, chemotaxis receptors) domain. The transmembrane segment of δ and some extracellular regions of α₂ are thought to associate with the α₁ subunit [9].

Even though the physiological relevance of α₂δ is not well defined, there is evidence indicating a role in trafficking and also in the modulation of Ca_V channel properties. Heterologous expression of α₂δ with various α₁ and β subunits results in an increase in current density, acceleration of current activation and inactivation and hyperpolarizing shifts in the current–voltage relationship [5–7]. The mechanisms for these effects are as yet incompletely defined, although the increase in current density could be explained by improved targeting of Ca_V channels to the plasma membrane [8,10] and significant reduction in the rate of entry of surface resident channels into degradative pathways [11]. Additionally, Ca_V channels appear to be segregated into cholesterol-rich membrane microdomains (i.e. lipid rafts), and it has been found that the α₂δ subunit is necessary and sufficient for targeting channels to lipid rafts [12,13].

Four subtypes of the α₂δ protein (1–4) encoded by four separate genes with several known splice variants have been described [3,5–7]. Interestingly, some insights into the functional role of the α₂δ auxiliary subunits in vivo come from a series of spontaneously occurring mutations in mice. For example, mutations in α₂δ-2 (ducky, entla) that result in similar phenotypes characterized by ataxia, paroxysmal dyskinesia, and spike wave seizures [14–16], and mutations in α₂δ-4 that presumably underlies a channelopathy that leads to cone-rod dysfunction in the murine visual system [17]. Last, α₂δ-1 and α₂δ-2 subunits have been shown to be a target of gabapentinoid drugs, which are small molecules with anticonvulsant and analgesic effects that exert chronic inhibitory actions on Ca_V channel functional expression [16,18–22], although the role of α₂δ in epilepsy and neuropathic pain remains unclear.

As mentioned earlier, during biosynthesis α₂δ subunits are generated as large precursor proteins, which undergo posttranslational cleavage, oxidation and glycosylation to yield mature proteins comprised of disulphide-linked α₂ and δ glycopolypeptides [3,5–7,23,24]. Although we have learned much about the posttranslational modification of α₂δ [23–26] the precise role of these events for Ca_V channel function is only partially understood. In particular, the exact role of disulphide bonding between α₂ and δ has not been well established and the exact cysteine residues involved in this intermolecular interaction are unknown. Therefore, in the present study, we aimed to identify the residues responsible for the disulfide bridging in the α₂δ auxiliary subunit.

2. Materials and methods

2.1. Cell culture and transfection

Human embryonic kidney (HEK) 293 cells (ATCC, Manassas, VA) were maintained in Dulbecco’s modified essential medium-high glucose supplemented with 10% horse serum, 1% L-glutamine, 110 mg/L sodium pyruvate and antibiotics, at 37 °C in a 5% CO₂-95% air humidified atmosphere. Gene transfer was performed using Lipofectamine Plus reagent (Invitrogen, Carlsbad, CA). Briefly, for a 35-mm Petri dish of HEK293 cells, 2 μg of the plasmid cDNA encoding the rabbit brain N-type Ca²⁺ channel Ca_V2.2 pore-forming subunit (GenBank accession number D14157) [27]; in combination with 2 μg cDNA of the rat brain β₃ (M88751) [28]; and 2 μg cDNA coding the wild-type rat brain α₂δ-1b cDNA (M86621) [29]; or its mutants were premixed with 6 μL of Lipofectamine in 100 μL serum-free medium according to the manufacturer’s instructions. The solution was then added to the dish and cells grown at 37 °C for 24 h, when medium was changed. The Ca_V2.2α_1B/α₂δ-1b/β₃ subunit composition was selected to emulate the structure of the native N-type Ca²⁺ channels [30–32].

2.2. Site directed mutagenesis

The α₂δ-1 cDNA was inserted into the recombinant bicistronic expression plasmid pIRES-hr-GFP1A [23,24] and expressed under the control of a cytomegalovirus promoter. The point mutations were introduced with ~40-mer synthetic oligonucleotides using the Quik-Change XL-mutagenesis kit (Stratagene, La Jolla, CA). Initially, single amino acid mutations changing a cysteine residue to a methionine or serine were created. Next, multiple mutations were done using suitable mutagenic primers. cDNAs of all mutant channel α₂δ-1 subunit were sequenced on an automated sequencer ABIPrism310 (PerkinElmer Applied Biosystems, Foster City, CA).

2.3. Electrophysiology

Ionic currents from HEK293 cells were recorded at room temperature (~22 °C) using the whole-cell configuration of the patch-clamp technique [33] 48 h after transfection. Ba²⁺ was used as the charge carrier. The extracellular solution contained (in mM): BaCl₂, 10; TEA-Cl, 125; HEPES, 10; glucose, 10 (pH 7.3). The intracellular solution contained (in mM): CsCl, 110; MgCl₂, 5; EGTA, 10; HEPES, 10; Na-ATP, 4; GTP, 0.1 (pH 7.3). Recordings were performed using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA). Currents were digitized at a sampling rate of 5.7 kHz and filtered at 2 kHz (four-pole Bessel filter). Linear leak and parasitic capacitance components were subtracted on-line using a P/4 protocol. Membrane capacitance (C_m) was determined as previously described [34] and used to normalize currents.

Current–voltage (I–V) relationships were obtained by step depolarization between −50 mV and +60 mV in 10 mV increments, from a typical holding potential (V_h) of −80 mV. To assess steady-state inactivation properties, a 1-s conditioning pulse to various holding potentials preceded a test depolarization to +10 mV. The peak current values were converted to peak conductance values using the expression G = I/(V_m − V_rev), where I is current, G is conductance, V_m is the test potential and V_rev is the extrapolated reverse potential. Conductance–voltage (G–V) curves for activation were fit with a Boltzmann equation of the form G = G_max/(1 + exp[(V_m − V_1/2)/k]⁻¹), where G_max is maximum conductance, V_m is the test potential, V_1/2 is the potential for half-maximal activation of G_max and k is a slope factor. Steady-state inactivation curves were fitted with a Boltzmann function I_Ba = I_max/(1 + exp[(V_m − V_1/2)/k]), where the current amplitude I_Ba has decreased to a half-amplitude at V_1/2 with an e-fold change over k mV. Time-to-peak (TtP) was measured using a trough-seeking function in pClamp within the test pulse duration. Current remaining was measured from an average of three individual sweeps per recording using the equation: I_rem = 100 × (I_end/I_peak), where I_rem is the current remaining at the end of a 140-ms test pulse, I_end is the value at the end of the test pulse, and I_peak is the maximum inward current measured during the test pulse.

2.4. SDS-PAGE and western blotting

Microsomes from transfected HEK293 cells were prepared as previously described [25,35]. Briefly, cells were harvested, washed twice with PBS and homogenized during 15 min in cold lysis buffer (50 mM Tris–HCl [pH 7.4], 0.8 μM aprotinin, 640 μM benzamidin, 1.1 μM leupeptin, 0.7 μM pepstatin and 230 μM PMSF). Homogenized was clarified at 3000 rpm/4 °C during 5 min. After centrifugation at 38,000 rpm/4 °C during 37 min pellet was suspended in sucrose buffer (0.3 M sucrose, 20 mM Tris–HCl [pH 7.4] and protease inhibitors) and stored at −70 °C. SDS-PAGE analysis was performed under reduced or non-reduced conditions as follows. Aliquots of 100 μg of protein were mixed with sample buffer (50 mM Tris–HCl [pH 6.8], 1.7% SDS, 5% glycerol, 0.002% bromophenol blue) containing or not the reducing agent dithiothreitol (DTT) at 100 mM and boiled for 5 min. Samples were subjected to 8% SDS-PAGE electrophoresis. After electrophoresis, proteins were transferred to a nitrocellulose membrane (Hybond-N; GE Healthcare, Buckinghamshire, UK). After blocking with non-fat milk (5%) supplemented with 0.2% Tween 20, membranes were incubated overnight with the primary anti α₂δ-1 antibody 1:200 in TBS-T with 5% non-fat milk (Alomone labs, Jerusalem, Israel) washed in TBS-T (10 mM Tris–HCl, 0.15 M NaCl, 0.05% Tween 20), incubated with horseradish peroxidase goat anti-rabbit secondary antibody and developed with the Amersham ECL reagent according to the manufacturers’ instructions.

2.5. Bioinformatics

Multiple protein sequences were aligned using the modified Clustal W algorithm [36] of Vector NTI 8 software package (Invitrogen). The disulfide bonds prediction was performed using DiAminoacid Neural Network Application (DiANNA 1.1), which is a recent state-of-the-art web-based software determining the cysteine oxidation state and disulfide connectivity of a protein [37,38] available at http://clavius.bc.edu/(clotelab/DiANNA/. Bonds with a score >0.9 were considered for further analysis. Homology modeling was performed by analyzing the α₂ and the δ domains separately. The sequence from residues 1–944 (α₂) and 945–1091 (δ) were modeled by the Robetta [39] and the I-TASSER [40] full-chain protein structure prediction servers (http://robetta.bakerlab.org and http://zhanglab.ccmb.med.umich.edu/I-TASSER, respectively). Models were assessed using Swiss-Model server and visualized using the Chimera package available from the Resource for Biocomputing, visualization and Informatics at UCSF (http://www.cgl.ucsf.edu/chimera).

3. Results

As a first step to identify the molecular determinants of the interchain disulfide bond(s) between the two domains of the Ca_V channel α₂δ-1 auxiliary subunit, we searched for conserved cysteine residues in the protein using multiple sequence alignments. Since the α₂ domain is a 944 amino acid polypeptide characterized by a large number of cysteine residues, we decided to start our analysis by performing the sequence alignment on the extracellular region of the δ domain, a 117 amino acid region that showed to possess only six conserved Cys. A selection of the δ peptide sequences aligned to show the pattern of the Cys residues is shown in Fig. 1A.

Fig. 1.

Mutation of the conserved Cys residues in the extracellular domain of the δ subunit. (A) Multiple alignment of δ proteins using Vector NTI 8 software. Sequences used were downloaded from GenBank. Residues are colored according to homology. Yellow highlights residues that are identical among species. Unshaded residues are unique, while blue and green indicate residues shared among a subset of variants or species. Sequences are from the top: human (Homo sapiens), rat (Rattus norvegicus), and mouse (Mus musculus). Numbers in parentheses indicate the position of the first Cys in the sequence. (B) Bar plot of current densities after a 140-ms depolarization pulse to +10 mV for the Ca_V2.2α₁/β₃ recombinant channels co-expressed with the α₂δ-1 mutants (n = 19–131). The left panel shows a schematic representation of the point mutations in δ: gray bars represent the protein with its TM region in white. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

We next investigated the functional role of the six conserved Cys in the extracellular region of δ by preparing single mutant variants. Initially, we constructed the Cys962Met mutation to preserve the sulfur atom in the amino acid. However, changing a hydrophilic residue by a more hydrophobic one could alter the properties of the protein, and therefore the rest of the conserved Cys were replaced by serine (Ser) residues (Fig. 1B). This change produces a smaller structural alteration and is likely to maintain the polarity and size of the amino acid side chains as close as possible to the original protein. The mutants were then transiently co-expressed with the N-type (Ca_V2.2) HVA channel α₁ pore-forming and the β₃ auxiliary subunit in HEK293 cells, and whole-cell patch clamp recording was then used to examine the effects of the mutations on functional channel expression. Inward Ba²⁺ current (I_Ba) through Ca_V channels containing the wild-type (WT) α₂δ-1 auxiliary subunit displayed the hallmark characteristics of α₂δ-mediated channel regulation such as increased current density and hyperpolarizing shifts in the voltage dependences of activation and inactivation, and acceleration of inactivation kinetics (Figs. 2 and 3). Interestingly, when examining current densities obtained in the presence of WT and mutant subunits, five out of the six Cys mutants showed no loss of α₂δ-mediated regulation (Fig. 1B). In sharp contrast, a subsequent electrophysiological analysis using the same protocol showed that the degree of current upregulation was significantly less than that of WT when the Cys1047Ser mutant was co-expressed with the N-type (Ca_V2.2α₁/β₃) channel (Fig. 2A).

Fig. 2.

Effect of the Cys1047 residue substitution on the α₂δ-1 subunit function. (A) Representative macroscopic currents from HEK293 cells co-expressing Ca_V2.2α₁/β₃ channels with the wild-type (WT) or the Cys1047Ser (C1047S) α₂δ-1 subunits during a 140-ms depolarizing pulse to +10 mV from a V_h of −80 mV. (B) Bar plot of current density after a 140-ms depolarization pulse to +10 mV for the Ca_V2.2α₁/β₃ recombinant channels co-expressed with the α₂δ-1 mutants shown in the left as in Fig. 1B (n = 32–131). Asterisks denote statistically significant differences between WT and α₂δ-1 mutant subunits. (C) Average current versus voltage plots (I–V) for the Ca_V2.2α₁/β₃ subunits in the presence of the α₂δ-1 indicated variants. Currents were evoked by 140-ms depolarizing pulses ranging from −50 mV to +60 mV in 10-mV increments from a V_h of −80 mV.

Fig. 3.

Effect of α₂δ-1 WT and the Cys1047 mutant proteins on the properties of the currents through N-type channels recorded in HEK293 cells. (A) Bar chart comparing current time-to-peak recorded in response to a depolarizing pulse to +10 mV from a V_h of −80 mV in cells expressing the α₂δ-1 WT, the Cys1047Ser or the C6 mutant subunits. (B) Proportion of non-inactivated current remaining at +10 mV in cells expressing the WT and the α₂δ-1 mutant subunits as listed. (C) Voltage dependence of activation of Ca_V2.2/β₃ channels co-expressed with α₂δ-1 WT (filled circles), the Cys1047Ser or the C6 mutant proteins (open circles and filled squares, respectively). The normalized data, obtained from recordings such as those shown in Fig. 2C, are plotted against the test pulse. Relative conductance (G/G_max) values were calculated and the data were fitted as described in Section 2. (D) Voltage dependence of steady-state inactivation of Ca_V2.2/β₃ channels co-expressed with α₂δ-1 WT (filled circles), Cys1047Ser (open circles) or the C6 mutant proteins (filled squares). Normalized data obtained from recordings such as those shown in Fig. 2C were plotted against the conditioning potentials. The mean data were fitted with a Boltzmann function.

To further characterize the role of Cys1047, additional multiple mutants of the α₂δ-1 subunit containing the Cys1047Ser mutation along with Cys962Met, Cys984Ser, Cys987Ser, Cys1032Ser and Cys1059Ser substitutions (C5 and C6) were engineered and co-expressed with the N-type (Ca_V2.2α₁/β₃) channel. As expected, neither of the latter mutations resulted in the upregulation of current density normally observed after co-expression of the WT α₂δ-1 subunit (Fig. 2B) measured at +10 mV. These data thus further support the notion of residue 1047 being a crucial determinant of α₂δ-1 protein modulation of N-type channels. A similar pattern was observed in whole cell recordings from HEK293 cells in experiments where macroscopic currents were evoked by a series of depolarizing steps ranging from −50 mV to +60 mV from a V_h of −80 mV. Mean currents at all command potentials were 2- to 4-fold smaller in cells expressing α₂δ-1 subunits containing the Cys1047Ser mutant compared with cells expressing the WT protein (Fig. 2C). In summary, recombinant N-type (Ca_V2.2α₁/β₃) channel function was not affected by the substitution of Cys962 by Met or Cys984, 987, 1032, 1059 by Ser in the α₂δ-1 auxiliary subunit, but it was impaired by the substitution of Cys1047 by Ser. Therefore, the later cysteine residue is required for α₂δ-1 function.

The loss of channel function as a result of the Cys1047Ser amino acid substitution could be attributed to different mechanisms that affect channel activity and/or channel membrane expression. To get insight into the effect of this mutation, a kinetic analysis was next performed where we measured the time-to-peak of I_Ba and the current remaining (I_rem) at the end of the command pulse as indices of channel activation and inactivation, respectively (Fig. 3). TtP was measured at a test potential of +10 mV from a V_h of −80 mV. As can be seen in Fig. 3A, TtP was significantly slower for currents recorded in cells transfected with α₂δ-1 subunits containing the Cys1047Ser mutation. In addition, mutating the six conserved Cys residues within the extracellular loop of the δ domain or the substitution of the Cys1047 alone affected the voltage dependence of channel activation. Fig. 3C shows that there is an overlap in the activation curves independently of what α₂δ-1 mutant was co-expressed with of Ca_V2.2α₁/β₃ channels. In contrast, in Ca_V2.2α₁/β₃ channels containing the WT α₂δ-1 subunit the activation curve appeared shifted 5–10 mV toward more negative potentials, suggesting that Cys1047 within the δ sequence may participate in the α₂δ-mediated regulation of the coupling between the voltage sensor and the channel gate.

The participation of α₂δ-1 in the regulation of Ca_V channel voltage-dependent inactivation is well documented. This auxiliary subunit tends to accelerate inactivation kinetics and produces a hyperpolarizing shift in half-inactivation potential of all Ca_V channels subtypes [35,41,42]. Consistent with this, in the absence of α₂δ-1 the fraction of channels that inactivated after a 140-ms depolarizing pulse did not exceed 50% over a wide range of voltages, and no negative shifts in half-inactivation potential was observed. In contrast, as seen from the raw current records in Fig. 2A and from the data shown in Fig. 3B, co-expression with α₂δ-1 WT or its mutants resulted in a significant change in the inactivation kinetics of Ca_V2.2α₁/β₃ channels. It should be noted however, that the percentage of residual current remaining at the end of the pulse was different for channels containing the WT and the Cys1047Ser mutation. The WT α₂δ-1 subunit preferentially accelerated inactivation kinetics of the Ca_V2.2α₁/β₃ channels, while inactivation kinetics of channels was not effectively regulated by the mutant α₂δ-1 subunits (Fig. 3B). Likewise, Fig. 3D illustrates the steady-state inactivation behavior of the Ca_V channels in the presence of the WT and mutant α₂δ-1 auxiliary subunits. Ca_V channels containing the WT subunit, but not the Cys1047Ser or the C6 mutations, exhibited a discernible ~10 mV hyperpolarizing shift in half-inactivation potential.

To further address the role of the conserved Cys residues within the transmembrane δ domain in the formation of a disulfide bridge within α₂δ-1, the WT auxiliary subunit and its mutated versions were expressed in HEK293 cells and analyzed by immunoblotting. In these experiments, proteins were exposed to a reducing agent (DTT) to disrupt endogenous disulfide bridges between Cys residues. The result of this analysis showed that under non-reducing conditions the WT protein migrates as a high molecular mass protein of ~170 kDa. Given that the epitope recognized by the α₂δ-1 antibody used in these experiments is located in the N-terminal of α₂, upon reduction with DTT the majority of the protein migrated as a product of the predicted size for the α₂ peptide (~140 kDa). This result confirmed that under our experimental conditions the WT α₂δ-1 subunit is proteolytically cleaved into separate α₂ and δ peptides (Fig. 4A). Likewise, expression of the α₂ protein alone (lacking the δ domain) as a control, resulted in the presence of high molecular mass species of ~140 kDa under both reducing and non-reducing conditions, as expected (Fig. 4A).

Fig. 4.

Role of Cys1047 in the formation of an intermolecular disulfide bridge in the Ca²⁺ channel α₂δ-1 subunit. (A) Total cell lysates from HEK293 cells transiently transfected with plasmid cDNA encoding the full length α₂δ-1 or the α₂ peptide were incubated with (+) or without (−) 100 mM DTT as described in Section 2. The samples were analyzed by SDS-PAGE and visualized by western blotting using the anti-α₂δ-1 antibodies. Cells transfected with empty vector were used as the control (pIRES). (B) Cell lysates from HEK293 cells expressing different auxiliary subunits mutated in the extracellular loop of δ were analyzed by SDS-PAGE and western blotting under reducing (+) and non-reducing (−) conditions using the same anti-α₂δ-1 antibodies. (C) HEK293 cells were transiently transfected with plasmid cDNA coding the α₂δ-1 WT and the mutant proteins Cys1047Ser and C6 and analyzed by Western blot under reducing and non-reducing conditions.

To confirm that the conserved Cys 962, 984, 987 and 1032 present in the extracellular domain of the δ protein are not involved in disulfide bonding as predicted from the functional data, we compared the mobility of the previously genetically engineered δ single mutants expressed in HEK293 cells with that of the WT α₂δ-1 protein under reducing and non-reducing conditions. As shown in Fig. 4B, single mutants Cys962Met, Cys984Ser, Cys987Ser and Cys1032Ser exhibited an identical mobility pattern as the WT α₂δ-1 protein. This result is consistent with the predicted absence of intermolecular disulfide bridges formed by these Cys residues.

In sharp contrast, expression of the Cys1047Ser and the C6 mutations resulted in proteins which size remained unchanged upon reduction, suggesting that the recombinant proteins harboring a mutation in the Cys residue at position 1047 were not able to form the disulfide bond that maintains the α₂ and the δ peptides linked in vivo (Fig. 4C). It should be noted that a small fraction of the C6 mutant α₂δ-1 protein migrated as a diffuse ~170 kDa band under non-reducing conditions, suggesting the formation of aberrant intermolecular disulfide bonds that cause the protein to migrate as diffuse smear. However, the majority of the C6 mutant protein appeared to be the expected ~140 kDa form of the proteins, confirming that Cys1047 participates in the disulfide bond necessary for covalent assembly of the α₂δ-1 complex.

A bioinformatic study was next conducted using DiANNA 1.1 web server [37,38] to predict the counterpart of the Cys1047 involved in the formation of the disulfide bond occurring in the α₂δ-1 subunit. The result of this analysis showed one cysteine residue at position 404 in the α₂ domain potentially involved in this process (Suppl. Table 1). The multi-sequence alignment (using Vector NTI 8 software) demonstrated that Cys404 is conserved in the three species analyzed (Fig. 5A), which strongly supports the hypothesis of a potential role of this amino acid in the α₂δ-1 disulfide bridge formation.

Fig. 5.

Effect of the Cys404 residue substitution on the α₂δ-1 subunit function and its role in disulfide bond formation. (A) Multiple alignment of different α₂ proteins. As in Fig. 1, yellow highlights residues that are identical among species, unshaded residues are unique, while blue and green indicate residues shared among variants or species. Numbers in parentheses indicate the position of the first amino acid in the sequence. (B) Superposition of representative traces to illustrate current amplitude for the subunit combinations indicated. Currents were recorded in response to a depolarizing pulse to +10 mV from a V_h of −80 mV in cells co-expressing Ca_V2.2α₁/β₃ channels with the WT or the mutant (C404S) subunits. (C) Bar plot of current density after a 140-ms depolarization pulse to +10 mV for the Ca_V2.2α₁/α₃ recombinant channels expressed in the absence (−α₂δ-1) or the presence of the α₂δ-1 WT or the C404S mutant. Asterisks denote statistically significant differences with respect to the control value (WT subunit). (D) Western blot of total lysates of HEK293 cells transfected with the α₂δ-1 WT subunit or the C404S mutant under non-reducing or reducing conditions (− and +, respectively). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

We then examined the hypothesis that a specific disulfide bond formed by Cys404, located in the von Willebrand factor-A region (vWFA) of α₂ extracellular domain of the protein, could be required to generate and maintain the structure that determines the function of the α₂δ-1 subunit. Hence, whole-cell current recordings were carried out following transient transfection of HEK293 cells with the WT or the mutant Cys404Ser α₂δ-1 cDNA constructs, together with the N-type (Ca_V2.2) channel α₁ pore-forming and the β₃ auxiliary subunit. In contrast to the robust currents observed following expression of the WT channels, the Cys404Ser mutant subunit failed to induce the upregulation of channel functional expression observed after co-expression of the α₂δ-1 subunit (Fig. 5B and C).

To confirm the role of the Cys residue at position 404 in the intermolecular disulfide bond formation of α₂δ-1, we examined the electrophoretic mobility characteristics of the Cys404Ser mutant expressed in HEK293 cells. Consistent with the results in Fig. 4, when samples were not treated with the reducing agent, the mobility of the WT α₂δ-1 protein decreased, such that the ~170-kDa band was replaced by a band at ~140 kDa. In contrast, the mobility shift under non-reducing conditions, indicative of the presence of a disulfide bond, was eliminated by mutation of Cys404Ser. As can be seen in Fig. 5D, the apparent molecular mass for the mutant α₂δ-1 protein under both reducing and non-reducing conditions was ~140 kDa. Taken as a whole, these results strongly suggest that Cys404 is capable of interacting with Cys1047 to form a disulfide bond in the Ca²⁺ channel α₂δ-1 auxiliary subunit. Assuming a disulfide bond between Cys1047 and Cys404, we speculated that the mutation of Cys404 should present the same functional phenotype as observed for Cys1047. The results of the functional analysis show that indeed substitution of Cys404 to Ser results in significant changes in TtP and voltage dependence of channel activation and steady-state inactivation (Fig. 6), as expected.

Fig. 6.

Effect of the α₂δ-1 Cys404 mutant protein on the properties of Ca_V2.2/β₃ (N-type) channels expressed in HEK293 cells. (A) Comparison of current time-to-peak recorded in response to a depolarizing pulse to +10 mV from a V_h of −80 mV in cells expressing the α₂δ-1 WT and the Cys404Ser mutant subunits. (B) Voltage dependence of activation of recombinant N-type channels co-expressed with α₂δ-1 WT subunit (filled circles) or the Cys404Ser mutant protein (open circles). The normalized data are plotted against the test pulse. Relative conductance (G/G_max) values were calculated and the data were fitted as described in Section 2. (C) Voltage dependence of steady-state inactivation of recombinant N-type channels co-expressed with α₂δ-1 WT (filled circles) or the Cys404Ser (open circles). Normalized data were plotted against the conditioning potentials. The mean data were fitted with a Boltzmann function.

Last, to further determine the assignment of these two Cys residues in disulfide bonding, a predicted three-dimensional structure of the α₂δ-1 protein was generated by homology modeling (Fig. 7). Given that the two putative domains of the protein were outside the size and complexity limits of Robetta, development of the α₂δ-1 homology model was an iterative process in which the problem was resolved by analyzing both peptides (α₂ and δ) independently. In the first step, a homology model for the α₂ peptide was generated (Fig. 7A). To this end, fold recognition algorithms provided by the Protein Homology/AnalogY Recognition Engine V 2.0 PHYRE2 available at URL http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index [43] were used to score sequence compatibility of proteins that incorporate the vWFA domain fold. Hence, suitable templates for modeling of α₂δ-1 vWFA domain (residues Asp253-Leu430) were selected from available structures. The crystal structure of the vWFA region from the I-domains of integrins CD11b and CD11a with Mn²⁺ bound to its metal ion-dependent adhesion site (MIDAS), were the best templates. The three-dimensional structure of the vWFA domain in α₂ is represented by a doubly wound, open, twisted β-sheet flanked by α-helices. The domain adopts a classic α/β Rossmann fold and contains the MIDAS region (Suppl. Figs. 1 and 2).

Fig. 7.

Three-dimensional model of the Ca²⁺ channel α₂δ-1 subunit showing the potential interaction of Cys residues at positions 404 and 1047. (A) Model of the α₂ peptide showing the vWFA domain (orange). (B) Ribbon representation of the δ peptide showing the transmembrane helix (cyan). (C) The α₂ and δ partial models were then assembled into a combined model, which was built using PDBePISA, a web-based interactive software tool widely used for prediction of probable quaternary structures. (D) Close up of the α₂δ-1 subunit region showing the distance between the Cys pair that participates in the proposed disulfide bridge. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

Another domain found in our prediction was the Cache domain. This region consists of an N-terminal with three predicted strands and a α-helix, and a C-terminal with a strand dyad followed by a relatively unstructured region (Suppl. Fig. 3). The three strands in the N-terminal region of the Cache domain form a sheet analogous to that present in the core of PAS, a multifunctional domain found in Drosophila period clock protein, aryl hydrocarbon receptor and single-minded proteins [44].

In the next step, a three-dimensional model for the protein structure of the δ subunit was generated. To this end, the known NMR structure of the plasma membrane sarcolipin was used as template [45] for the putative transmembrane domain of δ, while the rest of the model was generated using the experimental structure of a non-related lipase [46]. The model shows that δ adopts a highly defined α-helical conformation from a Val1062 through a Leu1082. The nine amino acid C-terminal (Val1083 through Trp1091) is mostly unstructured, and N-terminal region is characterized by several turn-and-bend regions (Fig. 7B).

Likewise, to model the α₂ region in complex with the δ domain, the above individual three-dimensional models of the two proteins were assembled together using the protein interfaces surfaces and assemblies (PISA) web server (http://www.ebi.ac.uk/msdsrv/protint/pistart.html). The quality of the final model of the α₂δ-1 complex, as assessed by PROCHECK, Mfold, Qmean, Z score (Suppl. Table 2). Stereochemical main-chain and side-chain parameters were better than those for structures determined at 2 Å and 94.1% of residues of the complex were located in the most energetically favored regions. The Dfire value was −1160.64, which suggest a stable conformational structure. The quality of the final model reflected the quality of the crystal structures used as templates for the modeling, and the final three-dimensional model of the α₂δ-1 subunit is shown in Fig. 7C. Using this model we determined the distance between the residues that participate in the disulfide bond (Fig. 7D). The distance between the two Cys residues at positions 404 and 1047 is 5.28 Å. This measurement is within the range of bond distances for the formation of disulfide bridges [47]. In addition, the analysis by PROCHECK showed that the model has the side-chain environment of these Cys is such that they are conformationally able to permit the disulfide bond to adopt a torsional angle of 90° or −90°.

4. Discussion

The topology of the α₂δ-1 subunit was originally determined for the skeletal muscle variant. Three transmembrane segments were suggested for the protein based on hydropathy and biochemical analyses [48,49]. Later studies using site-directed antibodies revealed that the α₂δ-1 auxiliary subunit was actually a type I transmembrane (TM) protein with an extracellular α₂ domain anchored to the channel complex by a single membrane-located δ domain [25,50,51]. The early molecular studies also suggested that the α₂ and δ peptides were encoded by a single gene and that the α₂δ mature protein is post-translationally processed with the formation of disulfide bonds, before proteolytic cleavage into the two peptides. However, the exact cysteine residues involved in the formation of the putative disulfide bonds in the mature protein were not identified.

The main finding of our study, which combines bioinformatics, protein biochemistry and site-directed mutagenesis with molecular modeling and electrophysiological recordings, is that the conserved Cys404 and Cys1047 residues located in the vWFA region of α₂ and the extracellular domain of δ, respectively, form a single intermolecular disulfide bridge obligatory for normal α₂δ-1 subunit function. Hence, we used bioinformatics for predicting Cys residues in δ that might form direct contacts with Cys residues in α₂. These residues were mutated to Ser, and a possible disulfide bond formation between mutated subunits was investigated. Most of the Ser residues introduced into the extracellular domain of δ did not cause a change in the electrophoretic mobility of the protein following in vitro reduction of disulfide bonds with DTT. As mentioned earlier, under non-reducing conditions the α₂δ-1 WT protein is a diffuse ~170 kDa band. This band was absent in the presence of DTT, becoming a more discrete band of material that migrated at ~140 kDa. The mutation Cys1047Ser, however, migrated as a protein with an apparent molecular mass of about 140 kDa under both reducing and non-reducing conditions, suggesting a crucial role for this Cys residue in disulfide bond formation. Qualitatively similar results were observed when the conserved Cys residue at position 404 in the α₂ extracellular peptide was mutated to Ser, indicating the formation of a disulfide bond between Cys404 and Cys1047.

The above mentioned mutants were also transiently co-expressed in HEK293 cells, and whole-cell patch clamp recording was then used to examine their effects on functional channel expression. Previous studies in heterologous expression systems have shown that α₂δ-1 increases the maximum conductance of different recombinant Ca_Vα₁/β channels [3,4,6]. In the present work, we show that whereas there was a ~4-fold increase in current density through Ca_V2.2α₁/β₃ channels upon co-transfection with WT α₂δ-1, both Cys404Ser and Cys1047Ser mutations had little influence in whole-cell current density, implying that formation of an intra-subunit disulfide bond between these residues is essential for α₂δ-1 functional enhancement of Ca²⁺ currents.

Interestingly, it has been proposed that an important fraction of the α₂δ-2 subunit is not proteolytically processed into the α₂ and the δ peptides when expressed in the HEK293-derived cell line tsA201 [26]. Since it is acknowledged that co-expression of α₂δ-2 increases current density through different types of Ca_V channels [8,52], the possibility exists that cleavage of α₂δ might not be essential for functional enhancement of currents. Although the discrepancy between this finding and the data obtained for the α₂δ-1 subunit might lie in the lack of sequence similarity between the two α₂δ variants around the site of proteolysis [24,49], the fact that the protein is correctly processed post-translationally in the cerebellum [26] as well as in a HEK293 cell line stably expressing it [53] suggests that the proteolytically cleaved may be also the active form of the protein.

Last, there is now compelling evidence for a major involvement of the α₂δ-1 subunit for targeting the Ca²⁺ channel complex to lipid rafts [12,13,54]. Although the mechanism for this action remains unclear, α₂δ-1 co-localizes with lipid raft marker proteins when expressed in heterologous systems and is also necessary and sufficient for the targeting of neuronal channels to the rafts [12,13,54]. Recently, a new mechanism has been proposed for Ca_V channel localization to lipid rafts, which also challenges the conventional structural model of the α₂δ-1 subunit. This model suggests that the α₂δ protein associates with the plasma membrane via a glycosylphosphatidylinositol (GPI) anchor attached to the δ peptide [54]. However, studies by Jones and co-workers [55] have demonstrated more recently that the raft localization of α₂δ-1 is preserved after replacement of the reported GPI anchoring motif with the transmembrane domain of the Type I TM-spanning protein PING. This finding is consistent with the idea that the α₂δ-1 subunit resides in lipid rafts via protein-protein and/or specialized lipidprotein interactions, and argues against any involvement of the putative GPI-anchor mechanism. In line with this, the predicted model for the structure of the δ protein (Fig. 7B) is consistent with a model where the α₂δ-1 subunit retains its type I transmembrane topology. Using a selection of membrane topology prediction methods (Uniprot, MEMSAT-SVM, MEMSAT3 and Kyte-Doolitle) different models were produced for the δ peptide. There was a consensus between all prediction methods that the protein has one TM spanning helix in the region comprising amino acid residues Val1062 through a Leu1082. It is worth emphasizing that final three-dimensional structure presented here for δ (Fig. 7B) is in general agreement with existing experimental data, and that the confidence scores for estimating the quality of the predicted model (Suppl. Table 2), indicate adequate levels of certainty that the templates used were appropriate for modeling the proteins.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ceca.2011.10.002.