The α₂δ subunits of voltage-gated calcium channels form GPI-anchored proteins, a posttranslational modification essential for function

Abstract

Voltage-gated calcium channels are thought to exist in the plasma membrane as heteromeric proteins, in which the α1 subunit is associated with two auxiliary subunits, the intracellular β subunit and the α₂δ subunit; both of these subunits influence the trafficking and properties of Ca_V1 and Ca_V2 channels. The α₂δ subunits have been described as type I transmembrane proteins, because they have an N-terminal signal peptide and a C-terminal hydrophobic and potentially transmembrane region. However, because they have very short C-terminal cytoplasmic domains, we hypothesized that the α₂δ proteins might be associated with the plasma membrane through a glycosylphosphatidylinositol (GPI) anchor attached to δ rather than a transmembrane domain. Here, we provide biochemical, immunocytochemical, and mutational evidence to show that all of the α₂δ subunits studied, α₂δ-1, α₂δ-2, and α₂δ-3, show all of the properties expected of GPI-anchored proteins, both when heterologously expressed and in native tissues. They are substrates for prokaryotic phosphatidylinositol-phospholipase C (PI-PLC) and trypanosomal GPI-PLC, which release the α₂δ proteins from membranes and intact cells and expose a cross-reacting determinant epitope. PI-PLC does not affect control transmembrane or membrane-associated proteins. Furthermore, mutation of the predicted GPI-anchor sites markedly reduced plasma membrane and detergent-resistant membrane localization of α₂δ subunits. We also show that GPI anchoring of α₂δ subunits is necessary for their function to enhance calcium currents, and PI-PLC treatment only reduces calcium current density when α₂δ subunits are coexpressed. In conclusion, this study redefines our understanding of α₂δ subunits, both in terms of their role in calcium-channel function and other roles in synaptogenesis.

Voltage-gated (Ca_V) calcium channels of the Ca_V1 and Ca_V2 classes are thought to exist as heteromeric complexes. These consist of an α1 subunit that forms the pore and determines the main functional and pharmacological attributes of the channel (1 –3), which is associated with an intracellular-β subunit and an α₂δ subunit, both of which influence the trafficking of the channels and their kinetic and voltage-dependent properties. The molecular characterization and topology of the α₂δ subunits were initially determined for skeletal muscle α₂δ-1 (4 –8), but they have been suggested to generalize to all four characterized α₂δ subunits (for review, see ref. 9). When α₂δ-1 was purified from skeletal muscle, it was identified that the extracellular α₂ subunit is bonded by disulfide to a δ subunit (5, 6). Both subunits are the product of a single gene encoding the α₂δ protein, which is then posttranslationally cleaved (4). The α₂δ subunits are universally described as type-I transmembrane proteins, because they have an N-terminal signal peptide and a C-terminal hydrophobic and potentially transmembrane region. The α₂ moiety was found to play a role in enhancement of calcium currents, whereas the δ subunit modified the voltage-dependent properties (8, 10).

We noted that all of the α₂δ subunits have either a very short or nonexistent C-terminal cytoplasmic domain (11). Some of the α₂δ subunits have a C-terminal hydrophobic domain that is predicted to be rather short for a plasma membrane-spanning α-helix (11), which also contains helix-breaking prolines (e.g., α₂δ-3; Fig. 1A). For these reasons, we tested the hypothesis that some or all α₂δ proteins might be associated with the plasma membrane through a glycosylphosphatidylinositol (GPI) anchor (ω) attached to δ, rather than a transmembrane domain. Indeed, the features outlined above, among others, cause prediction programs (12) to indicate a high probability, particularly for α₂δ-3, that it is a substrate for GPI anchoring (Fig. 1A). This process occurs within the endoplasmic reticulum, and it involves cleavage of the C-terminal hydrophobic peptide at the ω-residue (Fig. 1A) and attachment of a GPI group to this residue, which then attaches the protein to the membrane through its lipid side chains (13).

Fig. 1.

Evidence that α₂δ-3 is GPI-anchored. (A) Amino acid sequence of rat α₂δ-3 C terminus showing the peptide used to generate the δ-3 (1035-1049) Ab in red, the two predicted GPI-anchor sites (potential ω amino acids arrowed and underlined in purple), and the C-terminal hydrophobic sequence (green). (B Upper) Immunoblot profile of WT α₂δ-3 in lipid-raft fractions (DRM) from transfected tsA-201 cells using α₂-3 (71–90) Ab. (Lower) Flotillin distribution in the same DRM profile. (C) Immunoblot of heterologously expressed α₂δ-3 in DRM fraction five, before and after incubation with PI-PLC and subsequent deglycosylation with endoglycosidase F (EndoF) as indicated. Deglycosylation was performed, because it enables accurate assessment of the protein MW and generally enhances immunoreactivity of the anti-peptide Abs. The Ab used was anti-δ-3 (1035-1049). The arrows show the upward shift of the δ-3 protein after PI-PLC. Similar results were obtained with PI-PLC from three different commercial sources. (D) Immunoblot profile of hippocampal α₂δ-3 with α₂-3 (71-90) Ab (Top) and flotillin (Bottom) in DRM fractions. (E) Hippocampal α₂δ-3 in DRM fraction five after deglycosylation with EndoF before and after PI-PLC as indicated. (Top) Immunoblot with α₂-3 Ab; (Bottom) immunoblot with δ-3 Ab. (Left) Equal loading of both lanes (25 μg protein); (Right) 8 μg protein in left lane and 20 μg in right lane to better compare the upward mobility shift after PI-PLC treatment, which is indicated by arrows. (F) PI-PLC-treated material from hippocampal DRM fraction five was subjected to phase separation in Triton X-114. This procedure is used to identify GPI-anchored proteins. The intact, amphipathic protein remains in the detergent-rich phase, whereas the PI-PLC-cleaved hydrophilic form is found in the detergent-poor (aqueous) phase (asterisk; see SI Materials and Methods (13). Western blots were performed on the detergent and aqueous phases to identify the presence of the different proteins. (Top) α₂-3. (Middle) Flotillin. (Bottom) δ-3 showing protein in the detergent (left two lanes) and aqueous phases (right two lanes) with and without PI-PLC treatment as indicated. The α₂-3 and δ-3 are indicated by an asterisk in the aqueous phase. Below this is shown a parallel blot that was probed with the CRD Ab, which shows a band at the level of δ-3 in the aqueous phase that was only observed after PI-PLC treatment (arrow). (G Left) Mean I-V relationship for Ca_V2.2/β1b coexpressed with α₂δ-3 in tsA-201 cells and treated for 90 min with vehicle (▪ n = 5) or with PI-PLC (4 U/mL; red • n = 11). (Right) Representative I_Ba currents (steps from −30 mV to +15 mV from a holding potential of −90 mV) are shown for the vehicle-treated control (black traces, Upper) and PI-PLC-treated (red traces, Lower) conditions; 1 mM Ba²⁺ was used as charge carrier. (H) Mean I-V relationship for Ca_V2.2/β1b expressed without α₂δ in tsA-201 cells and treated for 90 min with a vehicle control (▪ n = 11), a control in which the same amount of water was added (□ n = 20), or a control with PI-PLC (4 U/mL; red • n = 28). No significant differences were observed under any of the conditions. Note that 5 mM Ba²⁺ was used as the charge carrier, because the currents were very small in 1 mM Ba²⁺.

Results

Biochemical and Functional Evidence that α₂δ-3 Is Anchored by GPI.

To determine whether or not α₂δ subunits may be anchored by GPI, we initially focused on α₂δ-3 because of the strong prediction that this is the case. It is a well-established fact that GPI-modified proteins partition into Triton X-100-insoluble cholesterol-rich microdomains, also termed detergent-resistant membranes (DRMs) (13). We found that heterologously expressed α₂δ-3 was concentrated in DRMs (Fig. 1B). Both free α₂-3 and δ-3 were detected as well as full-length α₂δ-3 (Fig. S1), indicating that the protein is incompletely proteolytically cleaved after heterologous expression, which is the case for all α₂δ subunits (14).

The α₂δ-3 in DRMs was then treated with prokaryotic phosphatidylinositol-phospholipase C (PI-PLC), which cleaves GPI anchors. After cleavage, many GPI-anchored proteins, such as prion proteins, showed a characteristic increase in apparent molecular weight (MW) because of the anomalous interaction of the GPI anchor with SDS (13). This was observed for δ-3 after PI-PLC treatment, both before and, more clearly, after deglycosylation of the protein (Fig. 1C). We also found a reduction in the immunoreactivity of deglycosylated δ-3, which is a common occurrence after delipidation of GPI-anchored proteins (15, 16); it is thought to be indicative of a large conformational change in the protein between the soluble and GPI-anchored forms (as seen also in Fig. 1E) (17).

As an important confirmation, we found that native α₂δ-3 behaved similarly, being concentrated in DRMs from the hippocampus (Fig. 1D), with δ-3 showing an upward shift in mobility after PI-PLC treatment (Fig. 1E). PI-PLC treatment of hippocampal lysates also reduced the association of both the α₂ and δ moieties of α₂δ-3 with DRMs by more than 4-fold, whereas there was no effect on the localization of the non-GPI anchored DRM marker, flotillin (Fig. S2 A–C). Furthermore, after PI-PLC treatment, hippocampal α₂-3 and δ-3 were both lost from the detergent phase and recovered in the aqueous phase after temperature-induced phase separation in Triton X-114 (Fig. 1F, star), as described for other GPI-anchored proteins (13). Parallel blots for flotillin showed that the phase distribution of this protein, which is associated with DRMs by virtue of its myristoylation and palmitoylation (18), was not affected by PI-PLC. A similar result was obtained with trypanosomal GPI-PLC (Fig. S2D).

A generally accepted proof that a protein contains a GPI anchor is the presence of the cross reacting determinant (CRD) epitope (13), an antigenic carbohydrate determinant, including inositol 1, 2-cyclic phosphate, which is created after PI-PLC cleavage (19). In the aqueous phase after PI-PLC treatment, CRD immunoreactivity was associated with a band of the same mobility as PI-PLC-treated deglycosylated hippocampal δ-3 (Fig. 1F, arrow). This band at 17 kDa was absent when the sample was examined under nonreducing conditions and free δ-3 would not be formed (Fig. S2E).

To examine whether or not acute PI-PLC treatment would affect the properties of calcium-channel currents including α₂δ-3, we incubated tsA-201 cells expressing Ca_V2.2/α₂δ-3/β1b with PI-PLC for 90 min and found this reduced peak I_Ba by 60.7 ± 9.7% (Fig. 1G). This reduction agrees with the observation that these channels are less stable at the plasma membrane in the absence of α₂δ (20). In contrast, in the absence of coexpressed α₂δ, there was no inhibition of Ca_V2.2/β1b currents by PI-PLC treatment (Fig. 1H).

Effect of Mutation of the α₂δ-3 Predicted GPI-Anchor Motifs.

GPI-anchoring motifs have no fixed sequence, but the ω-residues are usually small and polar and ω+2 is usually Gly, Ala, or Ser (Fig. 1A) (13). We mutated residues in the amino acid motif (CGGAS), which spans the two predicted ω-sites of α₂δ-3 (Fig. 1A), to large hydrophobic or charged residues (WKW) that were poorly predicted to support GPI modification. We found that, although the mutant α₂δ-3 proteins were well expressed in the whole-cell lysate (WCL), they were poorly delivered to DRMs (Fig. 2A).

Fig. 2.

Evidence that GPI-anchoring of α₂δ-3 is required for cell-surface localization and function. (A Left) Immunoblot profile of GAS-WKW α₂δ-3 (Upper) and CGG-WKW α₂δ-3 (Lower) in DRM fractions from transfected tsA-201 cells using α₂-3 (71–90) Ab. (Right) Bar chart showing the average proportion of the total α₂δ-3 found to be in DRM (sucrose gradient fractions 4–6) for WT α₂δ-3 (black bar), CGG-WKW α₂δ-3 (blue bar), and GAS-WKW α₂δ-3 (red bar) from two independent experiments each. (B) Mean I-V relationship for Ca_V2.2/β1b coexpressed with WT α₂δ-3 (▪ n = 12), with GAS-WKW α₂δ-3 (red • n = 25), or without α₂δ (△ n = 23) in tsA-201 cells. All recordings are in 5 mM Ba²⁺. (C) Peak I_Ba (mean ± SEM) measured at +15 mV was determined from I-V relationships including those in A. Ca_V2.2/β1b is shown without α₂δ (white bar; n = 36), with WT α₂δ-3 (black bar; n = 16), with CGG-WKW α₂δ-3 (blue bar; n = 22), or with GAS-WKW α₂δ-3 (red bar; n = 25). Data were pooled from several experiments and normalized to the respective control (+WT α₂δ-3) in each experiment. The statistical significances of the differences compared to +WT α₂δ-3 were determined by ANOVA and post hoc Bonferroni test. **, P < 0.001. Example I_Ba currents (from −30 mV to +15 mV from a holding potential of −90 mV) are shown above the bars for no α₂δ (gray), WT α₂δ-3 (black), CGG-WKW α₂δ-3 (blue), and GAS-WKW α₂δ-3 (red). (D Upper) Cell-surface biotinylation of α₂δ-3 (WT), CGG-WKW α₂δ-3 and GAS-WKW α₂δ-3. The left three lanes show WCL, and the right six lanes show WT and mutant α₂δ-3 after cell-surface biotinylation, before and after deglycosylation with EndoF as indicated. Open arrow, full-length α₂δ-3; closed arrow, cleaved α₂-3. Lower shows lack of biotinylation of cytoplasmic Akt. (E) Quantification of relative amounts of full-length α₂δ-3 and cleaved α₂-3 for the CGG-WKW α₂δ-3 (blue) and GAS-WKW α₂δ-3 (red) mutants at the cell surface from two experiments, including the data shown in C, that were normalized to the amount in the WCL. (F) Confocal microscopic images showing membrane localization of α₂δ-3 using the δ-3 Ab (Left, red) for WT (Upper) and CGG-WKW α₂δ-3 (Lower) when coexpressed with green fluorescent protein Ca_V2.2 (Middle) and β1b in nonpermeabilized Cos-7 cells. (Right) Merged images show nuclear staining with DAPI (blue). Note that cell-surface staining in nonpermeabilized Cos-7 cells is seen, not as a fine ring but as a wide annulus, because of the flattened geometry (see also Fig. S3A). (Scale bar: 50 μm on merged images.) (G) Quantification of cell-surface immunofluorescence using δ-3 Ab for WT α₂δ-3 (black bars; n = 68), CGG-WKW α₂δ-3 (blue bars; n = 38), and GAS-WKW α₂δ-3 (red bars; n = 38). Statistical significance is P < 0.001 for one-way ANOVA and Tukey’s post hoc tests.

We next compared the ability of α₂δ-3 and the two α₂δ-3 GPI-site mutants to enhance calcium-channel currents. We observed that, whereas wild-type (WT) α₂δ-3 enhanced the peak Ca_V2.2/β1b I_Ba by about 4.5-fold, the CGG-WKW mutant reduced the peak I_Ba compared with currents in the absence of α₂δ (Fig. 2 B and C). A potential explanation of the reduction is that the mutant induces intracellular retention of the α1 subunit, which we suggested previously for Von Willebrand Factor A domain mutants of α₂δ-2 (21). In contrast, the GAS-WKW α₂δ-3 still produced a small increase in I_Ba and indeed, is still predicted to be a substrate, albeit a poor one, for GPI modification.

In agreement with these functional results, we found that the GPI mutant α₂δ-3 proteins, particularly the CGG-WKW mutant, showed only low expression on the plasma membrane, which was judged by cell-surface biotinylation (Fig. 2 D and E); this correlates well with the relative effect of these mutants on Ca_V2.2 I_Ba (Fig. 2C). Furthermore, immunocytochemical imaging also revealed much less of the GPI-mutant α₂δ-3 constructs on the cell surface of transfected Cos-7 cells compared with WT α₂δ-3 using both the δ-3 Ab (Fig. 2 F and G) and the α₂-3 Ab (Fig. S3). However, there was similar total expression (Fig. S3A).

Biochemical and Functional Evidence that α₂δ-2 Is Anchored by GPI.

We then examined whether or not α₂δ-2 subunits can also be anchored by GPI (Fig. 3A), because we had shown previously that the α₂δ-2 subunit was localized mainly in DRMs, both in native tissue and after heterologous expression (14). We found that in DRMs prepared from cells expressing α₂δ-2, an ∼18-kDa band representing deglycosylated WT δ-2 was recognized by two anti-peptide δ-2 Abs (1062-1079 and 1080-1094) (Fig. 3B). However, an Ab raised against the 15 C-terminal residues of δ-2 (1133-1147) did not recognize the ∼18-kDa deglycosylated δ-2 species (Fig. 3B Left, asterisk; see Fig. S4A for evidence that this Ab does recognize GPI-mutant α₂δ-2 where the C terminus is not cleaved). This indicated that the ∼18-kDa δ-2 protein does not contain the extreme C-terminal residues. Furthermore, we generated a C-terminal myc-tagged α₂δ-2 and expressed it in Cos-7 cells to find that the myc tag was cleaved from both cell surface-expressed (Fig. S4B ) and PI-PLC–released α₂δ-2 (Fig. S4C). In light of our results with α₂δ-3, an obvious possibility for further C-terminal processing would be cleavage of the C terminus associated with the formation of a GPI anchor (Fig. 3A). We then showed that incubation with PI-PLC of α₂δ-2 in DRMs resulted in a characteristic increase of apparent MW of δ-2 and a decrease in immunoreactivity, which was observed with δ-3 (Fig. 3C). A similar shift after PI-PLC treatment was seen for δ-2 concentrated in DRMs from cerebellum (Fig. 3D), and this species was also not immunoreactive with the C-terminal δ-2 (1133-1147) Ab (Fig. 3D). PI-PLC–cleaved cerebellar α₂δ-2 also partitioned into the aqueous phase after Triton X-114 phase separation (Fig. S5A, asterisk).

Fig. 3.

Biochemical and functional evidence that α₂δ-2 is anchored by GPI. (A) Amino acid sequence of mouse α₂δ-2 C terminus showing the peptides used to generate the δ-2 (1062-1079), δ-2 (1080-1094), and δ-2 (1133-1147; C-terminus, CT) Abs in blue, red, and pink, respectively. The predicted GPI-anchor site (in purple and underlined with ω indicated by the arrow) and the C-terminal hydrophobic sequence (green) are also shown. (B) Immunoblot of δ-2 from peak DRM fractions prepared from an α₂δ-2 stable cell line (14), which shows the absence of immunoreactivity to the C-terminal δ-2 (1133-1147) Ab (Left; see Fig. S4A for validation of this Ab) but immunoreactivity to other δ-2 Abs (1062-1079; Center) and (1080-1094; Right). (C) Immunoblot of δ-2 from the peak DRM fraction derived from α₂δ-2–expressing tsA-201 cells before and after incubation with PI-PLC and subsequent deglycosylation with EndoF as indicated. The Ab used was δ-2 (1080-1094). The arrows show the upward shift of the δ-2 protein after PI-PLC. (D Top) Immunoblot of α₂-2 from the peak DRM fraction derived from cerebellum before and after incubation with PI-PLC as indicated and subsequent deglycosylation with EndoF. (Middle) Immunoblot of δ-2 from the same fractions. A mixture of all anti–δ-2 Abs was used for this immunoblot. The arrows show the upward shift of the δ-2 protein after PI-PLC. (Bottom) Immunoblot of δ-2 with the δ-2 (1133-1147; CT) Ab alone from the same fractions before PI-PLC treatment, which shows the absence of immunoreactivity. (E) Immunoprecipitated HA-tagged α₂δ-2 treated or not treated with PI-PLC before purification is indicated in panel 1. Silver-stained gel was used for the affinity-purified HA-tagged α₂δ-2; full-length α₂δ-2 is the main species purified by this procedure, possibly because of the greater accessibility of the HA tag. These data also confirm that the PI-PLC used does not have any significant proteolytic activity, because no proteolytic fragments of α₂δ-2 were observed after 3 h of incubation. Panel 2 shows corresponding anti-HA blot; panel 3 shows blot for α₂-2 with α₂-2 (102-117) Ab. Panels 1–3 used 3–8% Tris acetate gel. Panel 4 shows the blot for δ-2 with δ-2 (1080-1094) Ab using a 12% Bis-Tris gel for better resolution of δ-2. Arrows indicate the position of α₂δ-2, and arrowheads show α₂-2 and δ-2. The material shown in panel 1 was then adjusted to equal protein concentration before immunoblotting for CRD. (F) Immunoblot with anti-CRD Ab. Lane 1 shows α₂δ-2 without PI-PLC treatment. Lane 2 shows α₂δ-2 after PI-PLC treatment. Lane 3 shows deglycosylated α₂δ-2 without PI-PLC treatment. Lane 4 shows deglycosylated α₂δ-2 after PI-PLC treatment that showed an ∼170-kDa band corresponding to HA-a₂δ-2 and a 150-kDa band corresponding to deglycosylated HA–α₂δ-2 (arrows). The CRD antibody has a low affinity, and we were unable to examine immunoreactivity against a band corresponding to free δ-2, because very little was purified using the internal HA tag in α₂-2 (see silver-stained gel and δ-2 blot in E). (G) Effect of acute incubation of Cos-7 cells for 1 h with PI-PLC (4 U/mL) on cell-surface localization of transfected 5′NT and α₂δ-2. Representative confocal microscopic images of α₂δ-2 (102-117 Ab; Upper) and 5′NT (Lower) cell-surface expression in transfected nonpermeabilized COS-7 cells (nuclei visualized with DAPI) were treated with either vehicle (con, left) or PI-PLC (right). (Scale bar: 20 μm.) (H) Quantification of immunofluorescence data similar to those shown in G were obtained from epifluorescence images for 5′NT incubated with vehicle (black bar; n = 25) or PI-PLC (Glyko; hatched bar; n = 31) and for α₂δ-2 incubated with vehicle (white bar; n = 30) or PI-PLC (cross-hatched bar; n = 20). The statistical significances of the differences with and without PI-PLC treatment were determined by Student's t test. *, P = 0.0083 for 5′NT; **, P = 0.0019 for α₂δ-2. (I Upper) Mean I-V relationship for Ca_V2.2/β1b coexpressed with α₂δ-2 (▪ n = 12), with α₂δ-2 GAS-WKW (red • n = 24), or without α₂δ (△ n = 18) in tsA-201 cells. All recordings are in 5 mM Ba²⁺. (Lower) Example I_Ba currents (from −30 mV to +15 mV from a holding potential of −90 mV) are shown for no α₂δ (gray), α₂δ-2 (black), and GAS-WKW α₂δ-2 (red).

To examine whether or not a CRD epitope was revealed on α₂δ-2 by PI-PLC treatment, we immunoprecipitated and concentrated α₂δ-2 from DRMs using an internal HA tag in α₂-2 (14) (Fig. 3E). As we have previously reported (14), the main species of HA-α₂δ-2 present in tsA-201 cells is uncleaved α₂δ-2, indicating that proteolytic cleavage into α₂ and δ is incomplete in heterologous-expression systems, unlike in native tissues. This is shown by the presence of both α₂ and δ immunoreactivity in the ∼190-kDa band (Fig. 3E). Importantly, only the PI-PLC–treated α₂δ-2 exhibited positive CRD immunoreactivity, and this was the case for both the glycosylated and deglycosylated α₂δ-2 (Fig. 3F). PI-PLC-treated 5′-nucleotidase (5′NT) was used as a positive-control GPI-anchored protein (Fig. S5B). A reduction in CRD immunoreactivity was seen on incubation of the deglycosylated PI-PLC-treated HA-α₂δ-2 with 1 M HCl to destroy the CRD epitope (Fig. S5C). The fact that CRD immunoreactivity was associated with the uncleaved α₂δ-2 band indicates that formation of the GPI anchor occurs before proteolysis into α₂-2 and δ-2. Furthermore, incubation of the HA-α₂δ-2 cell line with [³H]-inositol resulted in incorporation of [³H] into a band at the level of HA-α₂δ-2 (Fig. S6).

As further confirmation that α₂δ-2 is associated with the plasma membrane by a GPI anchor, we found that acute incubation of Cos-7 cells for 1 h with PI-PLC significantly reduced the amount of cell-surface expression, both of 5′NT by 31.9% and of expressed α₂δ-2 by 49.4% (Fig. 3 G and H). Similar results were found for α₂δ-3 (33.9 ± 5.5% reduction; n > 100 cells examined in each condition). Importantly, in the same system, there was no effect of a high concentration of PI-PLC (8 U/mL) on cell-surface expression of a known transmembrane protein (α7 nicotinic cholinergic receptor per 5HT3 receptor chimera; Fig. S7).

Mutational Evidence that α₂δ-2 Is Anchored by GPI.

Next, we mutated the predicted GPI-anchor site in α₂δ-2 (GAS to WKW, Fig. 3A). An ∼23-kDa deglycosylated band, the correct MW to represent a mutant δ-2 species retaining the C terminus, was recognized by both the δ-2 (1080-1094) Ab and the C-terminal δ-2 (1133-1147) Ab (Fig. S4A). However, no ∼18-kDa δ-2 band was present, suggesting that GAS-WKW α₂δ-2 was not C-terminally truncated, unlike WT α₂δ-2 (Fig. 3B; Fig. S4A). Like the α₂δ-3 GPI mutants, the α₂δ-2 GPI mutant also showed a reduction in its ability to partition into DRMs (Fig. S8A). GAS-WKW α₂δ-2 also showed a reduction in expression on the plasma membrane, which was determined by cell-surface biotinylation (30.7% of WT α₂δ-2 for the α₂-2 protein; Fig. S8B). We then examined its functionality and found that the GAS-WKW α₂δ-2 was unable to enhance Ca_V2.2 I_Ba compared with the currents observed in the absence of α₂δ (Fig. 3I).

Biochemical and Functional Evidence that α₂δ-1 Is Anchored by GPI.

Finally, we also obtained evidence that α₂δ-1 is anchored by GPI as predicted from its C-terminal sequence (Fig. 4A). Both heterologously expressed α₂δ-1 (Fig. S8C), and native α₂δ-1 from hippocampus and cardiac muscle partition strongly into DRMs (Fig. 4B for hippocampus). Furthermore, PI-PLC treatment of hippocampal lysates reduced the DRM association of α₂δ-1 from 68.3 ± 8.0% to 8.3 ± 4.6% (P = 0.0034 from three independent experiments, including the experiment in Fig. 4C), and also resulted in recovery of α₂-1 in the aqueous phase after Triton X-114 separation (Fig. 4D, asterisk).

Fig. 4.

Evidence that GPI anchoring of α₂δ-1 is required for cell-surface localization and function. (A) Amino acid sequence of rat α₂δ-1 C terminus showing the predicted GPI-anchor sites (ω) and the C-terminal hydrophobic sequence (green). (B and C) Lysates from the hippocampus were incubated with vehicle (B) or PI-PLC (C) at 37°C in the presence of protease inhibitors, and then they were subjected to sucrose gradient centrifugation. (Upper) The immunoblot profile of α₂δ-1 in DRMs from hippocampus is shown using the α₂-1 monoclonal Ab. (Lower) Flotillin distribution in same DRM profile. (D) Phase separation of α₂-1 after PI-PLC treatment. (Upper) Peak DRM fraction before phase separation showing input for α₂-1 and flotillin. (Lower) Phase separation after vehicle or PI-PLC treatment and Triton X-114 extraction shows α₂-1 in aqueous phase after PI-PLC (indicated by asterisk). Flotillin was not redistributed into the aqueous phase after PI-PLC treatment. (E Left) Immunolocalization of α₂δ-1 in nonpermeabilized DRG neurons after 2 days in culture after incubation with vehicle (Upper; confocal field contains 12 DRG neurons) or with 4 U/mL PI-PLC for 60 min (Lower; field contains 10 DRG neurons). (Scale bar: 50 μm.) (Right) Quantification of α₂δ-1 immunostaining expressed as number of pixels per cell (log₁₀ scale) in the intensity range 500-3,000 for 18 images from three different coverslips for each condition, each containing between 5 and 12 cells (Upper). Black line, vehicle-treated; red line, 4 U/mL PI-PLC; blue line, 8 U/mL PI-PLC. (Lower) Quantification of immunostaining in four separate experiments after vehicle (black bar; normalized to 100%) relative to PI-PLC (4 U/mL or 8 U/mL; red bar). **, P < 0.001 for Student's t test. (F) Immunoblot of supernatant after PI-PLC treatment of DRGs, as in E, for 4 U/mL and 8 U/mL PI-PLC. The arrow shows the presence of α₂-1. (G) Cartoon showing the structure of α₂δ subunits that is identified here. The GPI anchor consists of ethanolamine (orange), three mannose rings (blue), glucosamine (pink), and inositol (yellow). (H) Proposed scheme for processing α₂δ proteins. The N-terminal signal sequence is in gray, the C-terminal GPI signal sequence is in green, the α₂ sequence is in black, and the δ sequence is in white. The GPI anchor is drawn as in G. The position of the ω amino acid is indicated by a purple arrow in G and H.

Dorsal root-ganglion neurons (DRGs) represent a heterogeneous class of sensory neurons that express α₂δ-1 on their cell surface (22). We found that α₂δ-1 is also well expressed on the plasma membrane of DRGs in culture (Fig. 4E) and that incubation of cultured DRGs for 60 min with PI-PLC substantially reduced cell-surface α₂δ-1 immunofluorescence (Fig. 4E). However, it did not reduce that for the transmembrane p75 nerve growth-factor receptor (Fig. S9). In the same experiments, PI-PLC treatment of cultured DRGs also resulted in the appearance of α₂δ-1 in the supernatant, showing that it had been shed from the cell surface (Fig. 4F).

Discussion

Taken together, these results provide evidence that all of the α₂δ subunits studied (isoforms 1–3) are able to form GPI-anchored proteins (Fig. 4G). The α₂δ-4 subunit is also strongly predicted to be anchored by GPI. Although all α₂δ subunits have a C-terminal hydrophobic domain, which led to the original description of α₂δ-1 as a type-I single-pass transmembrane protein (4 –8), GPI-anchored proteins also require a C-terminal hydrophobic sequence. In addition, they usually have little or no cytoplasmic domain, features exhibited by the α₂δ proteins. They frequently have one or more prolines in or near the C-terminal hydrophobic domains (23), which is also the case for the α₂δ subunits. Whereas α₂δ-2 has the longest predicted C-terminal cytoplasmic domain of the α₂δ subunits (∼15 residues), there are other key determinants within the C-terminal domain that dictate GPI anchoring, even in the presence of a hydrophilic C-terminal extension (23, 24). Our results also indicate that GPI-anchoring occurs before proteolytic cleavage into α₂ and δ (Fig. 4H). Nevertheless, from the present results, we cannot rule out the coexistence of transmembrane forms of α₂δ subunits, and such alternative processing has been shown for other GPI-anchored proteins (25). However, an α₂δ-1 chimera containing the transmembrane segment of adhalin was previously found to be unable to enhance calcium currents (8).

As a consequence of the GPI-anchored structural form of α₂δ subunits revealed here, the reported role of the transmembrane segments of δ subunits to influence the biophysical properties of calcium channels (8, 10) and the limited structural information on calcium-channel complexes (26, 27) may require reinterpretation. Some important implications of GPI anchoring of α₂δ subunits include that they may direct the localization of interacting partners, including Ca_Vα1 subunits, into specific DRM domains. Indeed, we have shown that a population of Ca_V2.1 is present in DRMs prepared from the cerebellum (14). Calcium-channel complexes or α₂δ subunits playing other roles [e.g., in synapse formation (28, 29)] may participate in nanodomain signaling complexes, and thrombospondin is one identified interacting partner (28). Additionally, modification of lipid-raft composition by acute cholesterol depletion has marked effects on calcium-channel currents (14). Also, it opens the possibility that α₂δ subunits could be dynamically shed from the plasma membrane by the action of native extracellular PI-PLC enzymes (30). Therefore, it could acutely alter calcium-channel function or other functions of α₂δ proteins (28, 29) that may involve soluble forms of α₂δ (29). For α₂δ-1, this would be particularly relevant at presynaptic terminals where it is concentrated (22, 31). In relation to this, the α₂δ-1 protein is up-regulated in DRG neurons in neuropathic pain (22), and it is the main therapeutic target for gabapentinoid drugs in the alleviation of this condition (32, 33).

Materials and Methods

Standard molecular biological, biochemical, and electrophysiological techniques were used as described previously (14, 21, 33) and in SI Materials and Methods. Details of all Abs used are also provided in SI Materials and Methods. Where data are given as mean ± SEM, statistical comparisons were performed using either Student’s t test or ANOVA with a post hoc test, as appropriate.