Cloning and Deletion Mutagenesis of the α2δ Calcium Channel Subunit from Porcine Cerebral Cortex

The anti-epileptic, anti-hyperalgesic, and anxiolytic agent gabapentin (1-(aminomethyl)-cyclohexane acetic acid or Neurontin) has previously been shown to bind with high affinity to the α2δ subunit of voltage-dependent calcium channels (Gee, N. S., Brown, J. P., Dissanayake, V. U. K., Offord, J., Thurlow, R., and Woodruff, G.N. (1996)J. Biol. Chem. 271, 5768–5776). We report here the cloning, sequencing, and deletion mutagenesis of the α2δ subunit from porcine brain. The deduced protein sequence has a 95.9 and 98.2% identity to the rat and human neuronal α2δ sequences, respectively. [3H]Gabapentin binds with a K D of 37.5 ± 10.4 nm to membranes prepared from COS-7 cells transfected with wild-type porcine α2δ cDNA. Six deletion mutants (B–G) that lack the δ polypeptide, together with varying amounts of the α2 component, failed to bind [3H]gabapentin. C-terminal deletion mutagenesis of the δ polypeptide identified a segment (residues 960–994) required for correct assembly of the [3H]gabapentin binding pocket. Mutant L, which lacks the putative membrane anchor in the δ sequence, was found in both membrane-associated and soluble secreted forms. The soluble form was not proteolytically cleaved into separate α2 and δ chains but still retained a high affinity (K D = 30.7 ± 8.1 nm) for [3H]gabapentin. The production of a soluble α2δ mutant supports the single transmembrane model of the α2δ subunit and is an important step toward the large-scale recombinant expression of the protein.

Voltage-dependent Ca 2ϩ channels (VDCCs) 1 are heteromultimeric complexes present in both neuronal and non-neuronal tissues, including heart and skeletal muscle (1,2). VDCCs are minimally composed of three subunits: a pore-forming transmembrane ␣ 1 subunit, a hydrophilic intracellular ␤ subunit, and a membrane-associated ␣ 2 ␦ subunit; a transmembrane ␥ subunit is also found in skeletal muscle tissue. Multiple subtypes and/or splice variants of the ␣ 1 , ␤, and ␣ 2 ␦ subunits have been found (3). In heterologous expression studies, the ␣ 2 ␦ subunit has been shown to increase ␣ 1 currents both by facilitating the assembly of ␣ 1 subunits at the cell surface (4) and by stimulating the peak ␣ 1 current (5,6). The modulatory effects of ␣ 2 ␦ are more pronounced if the ␣ 1 and ␣ 2 ␦ subunits are co-expressed with the ␤ subunit (7,8). However, the functions of the ␣ 2 ␦, ␤, and ␥ subunits in vivo are not known.
A novel high affinity binding site for [ 3 H]gabapentin in rat, mouse, and pig brains has been characterized (18,28). Recently, the [ 3 H]gabapentin-binding protein was isolated from pig brain and identified as the ␣ 2 ␦ subunit of VDCCs (29). None of the prototypic anticonvulsant drugs displace [ 3 H]gabapentin binding from the ␣ 2 ␦ subunit (18,30). [ 3 H]Gabapentin is stereospecifically inhibited by two enantiomers of 3-isobutyl GABA. The rank order of potency of gabapentin, and S-and R-isobutyl GABA, at the [ 3 H]gabapentin binding site mirrors their anticonvulsant activity in mice (31). However, electrophysiological studies have yielded conflicting data on the action of gabapentin at VDCCs (32,33), and the relevance of the interaction of gabapentin at the ␣ 2 ␦ subunit to the clinical utility of the drug is presently unclear.
The ␣ 2 ␦ subunit is derived from a single gene (34), the product of which is extensively post-translationally modified. The polypeptide is cleaved to form disulfide-bridged ␣ 2 and ␦ peptides (35), both of which are heavily glycosylated (36). Hydropathy analysis of the ␣ 2 ␦ subunit from rabbit skeletal muscle predicts two putative transmembrane domains in the 145,000 M r ␣ 2 polypeptide component and one in the 25,000 M r ␦ component (37). However, a number of biochemical studies lend support to a single transmembrane model for the ␣ 2 ␦ subunit in which the protein is anchored through a domain near the C terminus of the ␦ polypeptide (35, 38 -40). The structural requirements for binding of [ 3 H]gabapentin to the ␣ 2 ␦ subunit are not known.
Here, we describe the cloning of the porcine cerebral cortical * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF077665.
FIG. 1. Primary sequence and translation of the porcine cerebral cortical voltage-dependent calcium channel ␣ 2 ␦ subunit cDNA. Primary sequence with amino acid translation. Numbering in the margin is for the primary DNA sequence, where 1 is the start of the cDNA. Numbering in italics represents the residue number, where ϩ1 is the start of the mature processed ␣ 2 ␦. The location of the four putative transmembrane domains ((I), residues 229 -250; (II), residues 426 -445; (III), residues 885-918; and (IV), 1035-1060) are indicated by the solid boxes. The putative zinc finger motif (residues 960 -988) is marked and labeled ZF (49,54). The solid bar represents sequence that conforms to the motif Cys-X 2 -Cys-X 4 -His-X 4 -Cys (48), and the dashed line indicates that the gap between the His and the final Cys is 19 rather than 4 residues. Putative N-linked glycosylation sites are identified with an asterisk and phosphorylation sites by the letters P, C, or T, indicating putative protein kinase C, cAMP-and cGMP-dependent protein kinase, and tyrosine kinase sites, respectively. ␣ 2 ␦ cDNA and the expression of a series of C-terminal deletion mutants of the protein in COS-7 cells. We have examined the ability of these mutants to bind [ 3 H]gabapentin, and we report on the expression of a soluble ␣ 2 ␦ mutant that retains high affinity for [ 3 H]gabapentin.
Cloning of the Porcine Cerebral Cortical ␣ 2 ␦ cDNA-An oligo dTprimed gt10 porcine cerebral cortical cDNA library was screened by ECL (Amersham) using a 2,381-bp HindIII fragment (coding sequence 268 -2649) of the rabbit skeletal muscle ␣ 2 ␦ clone (pcDNA3-Rab-␣ 2 ␦-(ϩ); supplied by Dr. Offord, Parke-Davis Pharmaceutical Research, Ann Arbor, MI) as the probe. A positive insert was identified and subcloned into pBluescript-SK-(ϩ) to generate pB-PC-␣ 2 ␦-1.1. The clone was sequenced on both strands, except for a 711-bp stretch at one end of the clone, which had a high degree of homology to mitochondrial C oxidase (41). The ␣ 2 ␦ coding region was homologous to the 3Ј region of the human neuronal ␣ 2 ␦ sequence (42) but lacked 926 bp of 5Ј coding sequence. The missing sequence was obtained by 5Ј-RACE using total RNA prepared from porcine cerebral cortex. RACE was performed across a BglI site unique in known ␣ 2 ␦ sequences (rabbit (accession no. M21948) (37), rat (accession number M86621) (43), human (accession no. M76559) (42)). The sequence derived from the 5Ј RACE product was used to design a primer (JB042, 5Ј-GGGGATTGATCTTCGATCGCG-3Ј) specific for the 5Ј-untranslated end of the cDNA. PCR was then performed with Pfu DNA polymerase using JB042 and a primer downstream of the BglI site (JB040, 5Ј-CTGAGATTTGGGGTTCTTTGG-3Ј). The PCR product was ligated to EcoRI linkers (5Ј-GGAATTCC-3Ј) and then digested with EcoRI and BglI. The 1,564-bp fragment (5Ј portion of the ␣ 2 ␦ cDNA) was gel-purified. Similarly, a 2,303-bp fragment (3Ј portion of the ␣ 2 ␦ cDNA) was isolated after digestion of pB-PC-␣ 2 ␦-1.1 with BglI and EcoRI. The two fragments of ␣ 2 ␦ cDNA were then ligated to EcoRI-digested pcDNA3 in a three-way ligation. A clone was picked with the full-length ␣ 2 ␦ sequence in the positive orientation with respect to the cytomegalovirus promoter (pcDNA3-PC-␣ 2 ␦-(ϩ)). The PCRderived 5Ј ␣ 2 ␦ sequence in this plasmid was sequenced on both strands. Any sequence within this Pfu-amplified region that resulted in the coding of a residue not present in the rat (43), rabbit (37), or human (42) sequence was checked against the sequence of the Taq-derived 5Ј RACE fragment. All differences were confirmed as genuine by this method.
Generation of Anti-␣ 2 and Anti-␦ Polyclonal Antibodies-The ␣ 2 ␦ subunit was purified from porcine brains as described by Gee et al. (29) up to, and including, the Sephacryl S400 step. The sample of partially purified ␣ 2 ␦ subunits was then further purified on a 1-ml CuSO 4 charged iminodiacetic acid-Sepharose column. Prior to each use, the column was recharged with CuSO 4 following a modified version of the protocol described by Brown et al. (44). Briefly, the column was stripped of metal ions with 3 ml of 0.5 M EDTA/NaOH, pH 8.0 (at 21°C), washed with 20 ml of H 2 O, and then charged with 20 ml of 0.3 M CuSO 4 before a second wash with 20 ml of H 2 O and equilibration in buffer A (750 mM NaCl, 0.08% Tween 20, 10 mM HEPES/KOH, pH 7.4 (at 21°C)). The partially purified ␣ 2 ␦ subunits obtained from the S400 chromatography was applied to this column at 0.5 ml/min. Breakthrough material was concentrated to ϳ100 l by ultrafiltration (10,000 M r cut-off membrane) before separation by SDS-polyacrylamide gel electrophoresis on an 8% preparative gel. The 145-kDa band was excised, and the peptide recovered from the gel by electroelution. Rabbits were immunized by Serotec (Oxford, UK). Anti-␦ antibodies were raised by immunizing rabbits with a keyhole lympet hemocyanin-conjugated peptide, VEMEDDDFTASL-SKQSC, corresponding to the start sequence of the ␦ polypeptide (residues 922-938, relative to the first residue of the mature ␣ 2 polypeptide). Peptide synthesis and immunization protocols were performed by Genosys Biotechnologies Inc. (The Woodlands, TX).
Affinity Purification of Antibodies-Purified pig brain ␣ 2 ␦ (125 g) was electrophoresed under reducing conditions on a single wide track 4 -20% gradient SDS-polyacrylamide gel. After transfer onto nitrocellulose membrane, two thin horizontal strips corresponding to the ␣ 2 and ␦ polypeptides were excised with a razor blade. The strips were incubated with blocking buffer (2% milk powder, 150 mM NaCl, 0.1% Tween 20, 50 mM Tris-Cl, pH 7.5) for 30 min. Immune serum (1 ml) was diluted 5-fold in blocking buffer and incubated with the appropriate strip for 2 h at 4°C. Strips were then washed three times (15 min each) with blocking buffer and eluted with 2 ml of 50 mM glycine/HCl, pH 2. , and PCR-WT (3Ј-untranslated region deleted) amplifications were performed with an anchored 5Ј primer (JB055, 5Ј-TGGCTTATC-GAAATTAATACG-3Ј), which anneals at position 849 -869 in pcDNA3-PC-␣ 2 ␦-(ϩ). For mutants A (⌬111-1067) and B (⌬229 -1067), the anchored 5Ј primer was 5Ј-AACTCCGGGGATTGATCTTCG-3Ј (JB115), which anneals at position 947-967. The 3Ј primer was designed to anneal internally to the ␣ 2 ␦ coding sequence to generate the specified C-terminally truncated ␣ 2 ␦ mutant. All 3Ј primers had the following tail structure: a double stop codon followed by an EcoRI site (5Ј-CAGAATTCCTCATCA-N (18 -21) -3Ј), where N is the in-frame sitespecific sequence complementary to the ␣ 2 ␦ cDNA. Pfu DNA polymerase was used in the PCR reactions, and the products amplified with JB055 were blunt-end cloned into pBluescript-SK(ϩ). The insert was then subcloned into the EcoRI site of pcDNA3. Products generated with JB115 were cloned directly into the EcoRV site of pcDNA3. Clones were partially sequenced to confirm primer regions and a positive orientation with respect to the cytomegalovirus promoter.
Construction of a ␦-Only Mutant-The ␣ 2 sequence (residues 1-921) was deleted utilizing the two-round PCR method (45) employing Pfu DNA polymerase. The product was blunt-end cloned into pBluescript-SK-(ϩ) and then directionally subcloned into pcDNA3 as described above.
Transient Expression in COS-7 Cells-All media contained 50 units/ml penicillin and 50 g/ml streptomycin. COS-7 cells were maintained in Dulbecco's modified Eagle's medium ϩ glutamax, 10% fetal bovine serum (gamma irradiated) in a 37°C/5% CO 2 incubator and passaged by trypsinization. For transient expression experiments, 150-mm culture dishes were seeded with 3.9 ϫ 10 6 cells and incubated for 16 h. Cells were then washed twice with 30 ml of optiMEM-1 and transfected (t ϭ 0 h) with 30 g of plasmid DNA by lipofectaminemediated transfection in 21 ml of optiMEM-1. At t ϭ 6 h, a further 21 ml of optiMEM-1 was added. At t ϭ 24 h, the medium was replaced with 42 ml of optiMEM-1. At t ϭ 48 h, the cells were washed twice with 30 ml of phosphate-buffered saline and then harvested in 3 ml of buffer A (1 mM EDTA, 1 mM EGTA, 20% glycerol, 10 mM HEPES, pH 7.4, at 4°C) plus 0.1 mM phenymethylsulfonyl fluoride using a cell scraper. All subsequent operations were performed at 4°C. The cells were tumbled on a Spiramix (Denley Instruments) for 30 min, centrifuged at 20,000 ϫ g for 5 min, resuspended in 1 ml of buffer A, recentrifuged at 20,000 ϫ g for 5 min, and finally resuspended in 400 l of buffer A. Membrane preparations were stored at Ϫ70°C until required.
Processing of Tissue Culture Media-Spent tissue culture medium recovered at t ϭ 24 and 48 h was ultracentrifuged at 100,000 ϫ g for 1 h and then concentrated by ultrafiltration (10,000 M r cut-off) to approximately 1 ml. The concentrated sample was then extensively dialyzed against buffer A and stored at Ϫ70°C until required.
Extraction of COS-7 Membranes-Samples of membranes (3 g in 48 l) were agitated for 2 h on a Spiramix at 4°C in a total volume of 60 l with a final concentration of either 1 M NaCl or 10% ethylene glycol. Samples were ultracentrifuged at 100,000 ϫ g for 2 h, and 20 l of supernatant was removed for SDS-polyacrylamide gel electrophoresis. The pellet was washed again for 10 min at 4°C in 1 ml of the same buffer before ultracentrifugation at 100,000 ϫ g for 30 min. The supernatant was discarded, and the pellet was resuspended in 120 l of SDS-polyacrylamide gel electrophoresis loading buffer and boiled for 20 min; 40 l was loaded onto the gel.
Miscellaneous Methods-Protein concentrations were determined by the method of Bradford (46) using bovine serum albumin as a standard. [ 3 H]Gabapentin binding assays were performed as described previously (29). For saturation analysis, incubations were performed in duplicate. All other incubations were performed in triplicate. SDS-polyacrylamide gel electrophoresis and Western blotting were performed using the Novex gel and buffer system (Novex Europe, Frankfurt, Germany). Molecular weights were determined by reference to Kaleidoscope markers (Bio-Rad). Detection was performed using the ECL system (Amersham).

Cloning and Characteristics of the Porcine Cerebral Cortical
␣ 2 ␦ cDNA-A single positive clone was isolated from a gt10 porcine cerebral cortical library. The EcoRI insert derived from the plaque comprised a 711-bp fragment with homology to mitochondrial C oxidase (41) and a truncated form of the ␣ 2 ␦ cDNA lacking 926 bp of the 5Ј coding sequence (according to an alignment with the human ␣ 2 ␦ sequence (42)). The missing 5Ј-coding sequence was obtained by a combination of 5Ј RACE and standard PCR. A 3,273-bp open reading frame coding for a 1,091 residue protein was identified (Fig. 1). The deduced amino acid sequence contains the partial N-terminal protein sequence of the purified porcine brain ␣ 2 subunit (29). The identity of the protein to the two published neuronal ␣ 2 ␦ sequences from rat (43) and human (42) was 95.9 and 98.2%, respectively. The predicted ␣ 2 /␦ cleavage site (34) was identified in the deduced sequence along with 16 putative N-linked glycosylation sites (Fig. 2). The calculated M r of the ␣ 2 ␦ subunit was 120,737, consisting of an ␣ 2 polypeptide of 104,337 M r and an ␦ polypeptide of 16,400 M r . Four putative hydrophobic domains were identified in the porcine sequence (Fig. 2) using the Kyte and Doolittle algorithm (47) with an integration span of 7. Three of the domains (II to IV; see Fig. 2) coincide with those identified by Ellis et al. (37) for the rabbit skeletal muscle ␣ 2 ␦ employing an integration span of 19. The protein sequence in the additional domain (I) is identical across species and corresponds to residues 229 to 250.
Analysis of the ␣ 2 ␦ C-terminal Deletion Mutants-An initial set of seven C-terminal deletion mutants (A-G) of the porcine cerebral cortex ␣ 2 ␦ subunit were generated by PCR and expressed in COS-7 cells. The expression of ␣ 2 ␦ was confirmed by SDS-polyacrylamide gel electrophoresis and Western blotting using ␣ 2 -specific polyclonal antibodies (Fig. 3, A and B). Mutants B-G were found to be associated with membrane fractions. Mutant A (⌬111-1067) was not detected in either the membrane fraction (Fig. 3, A and B) or a concentrated sample of spent culture medium (data not shown). The apparent M r values for the mutant polypeptides are greater than those predicted from primary sequence data; this discrepancy is probably due to the effects of glycosylation (Fig. 3B). Under nonreducing conditions, multiple high molecular weight products were visible for mutants D, E, F, and G, in addition to weaker products of the predicted size (Fig. 3A). Expression of the WT and PCR-WT ␣ 2 ␦ clones (Fig. 3A) also resulted in the presence of a high molecular mass species (Ͼ200 kDa) under nonreducing conditions. Upon reduction, the majority of the high molecular weight bands resolve to their predicted M r (Fig.  3B); the remaining high M r species seem resistant to the reducing conditions used. These products are most likely derivatives of the expressed ␣ 2 as their size decreases in parallel with the observed lower molecular weight products. Thus, these bands are likely to be residual ␣ 2 products, which for unknown reasons are resistant to the reducing conditions used. The multiple low molecular weight bands observed for mutants B and C (Fig. 3B) are most likely degradation products of the ␣ 2 mutants, perhaps due to inappropriate folding, making the protein more susceptible to proteolysis. A ladder of aggregation products was apparent when ␦ was expressed alone (Fig. 3C,  mutant N); under reducing conditions, all of these products resolved to ϳ25 kDa. No specific binding of [ 3 H]gabapentin (90 nM) was detected to any of the ␣ 2 deletion mutants or with ␦ expressed alone (n ϭ 2). In parallel experiments, [ 3 H]gabapentin bound to both wild-type ␣ 2 ␦ (WT) and a similar clone lacking the 3Ј-untranslated region (PCR-WT) with K D values of 37.5 Ϯ 10.4 nM and 30.8 Ϯ 4.5 nM, respectively. No specific [ 3 H]gabapentin binding was detected to membranes from COS-7 cells transfected with pcDNA3.
As these data suggested that both ␣ 2 and ␦ sequences were required for correct assembly of the [ 3 H]gabapentin binding pocket, a further six mutants with C-terminal deletions of the ␦ subunit were constructed. No [ 3 H]gabapentin binding to membranes from COS cells transfected with mutants H and I was observed (See Fig. 2). Specific [ 3 H]gabapentin binding was detected to mutants J (K D 40.9 Ϯ 6.9 nM), K, L, and M. Under reducing conditions, ␣ 2 ␦ from porcine cerebral cortical membranes yielded the predicted ␣ 2 and ␦ peptides, and no 170-kDa band (bridged ␣ 2 and ␦) was apparent (see Fig. 4, A and B). However, under the same conditions, the apparent molecular weight of the ␣ 2 ␦ deletion mutants, as determined using the anti-␣ 2 antibody (Fig. 4A) or the anti-␦ polyclonal antibody (Fig. 4B), increases as more ␦ is co-expressed. Thus, the deletion mutants are not proteolytically cleaved into separate ␣ 2 and ␦ peptides. The full-length ␣ 2 ␦ clones expressed in COS cells seem to be partially cleaved (Fig. 4, A and B; compare PCR-WT and WT with porcine cerebral cortical membranes) since, under reducing conditions, the ␦ antibody detects bands at both 170 kDa (single chain ␣ 2 ␦) and 25 kDa (␦) (see Fig. 4B, PCR-WT and WT). Deletion of the last seven residues of ␦ (mutant M) seems to inhibit proteolytic cleavage of ␣ 2 ␦ (Fig.  4B, compare mutant M with PCR-WT and WT). Because mutants J to M bind [ 3 H]gabapentin (see Fig. 2), it follows that proteolytic cleavage is not crucial to the formation of the [ 3 H]gabapentin binding pocket.
Production of a Soluble Form of the ␣ 2 ␦ Subunit-A protein at ϳ160 kDa is recognized by the anti-␣ 2 antibody in the media sample from COS-7 cells transfected with mutant L (⌬1040 -1067) (Fig. 5B). No protein is detected by this antibody in a similar sample from cells transfected with pcDNA3 (Fig. 5, A  and B). Very low levels of ␣ 2 ␦ are detected in the media after transfection with PCR-WT (Fig. 5A). Mutant L seems to be present in both secreted and membrane-associated forms in approximately equal amounts (Fig. 5, A and B; compare mutant L membranes and media samples). The particulate form remains associated with membranes after incubation with ei-ther 1 M NaCl or 10% ethylene glycol. Similar results were obtained with PCR-WT and mutants B to G expressed in COS-7 cells (data not shown). The membrane-associated form of mutant L comprises both cleaved and uncleaved ␣ 2 ␦, whereas the soluble form exists as a single polypeptide chain (Fig. 5D). Saturation [ 3 H]gabapentin binding gave K D values of 44.7 Ϯ 12.5 nM (Hill slope ϭ 0.92 Ϯ 0.038) and 30.7 Ϯ 8.1 nM (Hill slope ϭ 0.943 Ϯ 0.009) for the membrane-associated and soluble forms, respectively. As described previously (44), it is possible to assay ␣ 2 ␦ subunits using the ligand pair [ 3 H]leucine/ isoleucine. Binding was observed to both membrane-associated and soluble forms of ␣ 2 ␦ expressed in COS cells using this method (data not shown). DISCUSSION We recently identified the [ 3 H]gabapentin binding protein as the ␣ 2 ␦ subunit of voltage-dependent calcium channels (29). To enable a parallel biochemical and molecular biological study of ␣ 2 ␦, we have cloned the porcine cerebral cortical ␣ 2 ␦ subunit cDNA. The full-length ␣ 2 ␦ clone was obtained by a combination of library screening, 5Ј RACE, and standard PCR techniques. The amino acid sequence shows a high degree of identity with ␣ 2 ␦ from human (42) and rat brain (5) as well as with the rabbit skeletal muscle splice variant (37).
To investigate the topology of ␣ 2 ␦ and to determine the minimal N-terminal fragment of ␣ 2 ␦ required for [ 3 H]gabapentin binding activity, we constructed a total of 14 deletion mutants. Wild-type ␣ 2 ␦ expressed in COS cells was membraneassociated and displayed a high affinity for [ 3 (49,54). Potential N-linked glycosylation sites are identified with an asterisk. Two positive control transfections were performed. The pcDNA3-PC-␣ 2 ␦-(ϩ) construct is the full-length ␣ 2 ␦ cDNA. The PCR-WT construct codes for the full-length ␣ 2 ␦ coding sequence with a deleted 3Ј-untranslated region. expressing either ␣ 2 or ␦ alone was observed. This indicates that at least some of the ␦ polypeptide must be coexpressed with ␣ 2 to form the [ 3 H]gabapentin binding pocket. In the absence of the ␦ component, the ␣ 2 polypeptide was tightly associated with COS membranes. More extensive deletions to remove the two putative transmembrane domains identified by Ellis et al. (37) and an additional domain suggested by this study (see domain I in Fig. 2; residues 229 -250) also yielded mutants that were associated with particulate fractions (see Fig. 3A).
The association of mutants lacking ␦ sequences with membranes is at odds with the single transmembrane model of the ␣ 2 ␦ subunit. This model predicts an anchor region in the ␦ polypeptide and is supported by a number of studies. Antipeptide polyclonal antibodies directed to a region of ␣ 2 ␦ predicted to be intracellular, according to the three-transmem-brane domain model of Ellis et al. (37), recognized the antigen on intact nonpermeabilized rat dorsal root ganglions (38). Biochemical evidence for the single transmembrane model includes the release of ␣ 2 from membranes using dithiothreitol in combination with either urea (40) or alkaline extraction (35). Our studies provide a possible explanation for this discrepancy. Mutants that are devoid of cysteinyl residues migrate to similar extents under reducing and nonreducing conditions. Multiple aggregation products are seen under nonreducing conditions for mutants D, E, F, and G. These mutants contain cysteinyl residues that, in wild-type ␣ 2 ␦, may form disulfide bridges with residues in the deleted region of the protein. Thus, the association of full-length ␣ 2 and truncated forms of ␣ 2 with membranes could, in part, be an artifact caused by the disruption of the disulfide bridges found in the WT protein. Similarly, a ladder of aggregation products was also observed when ␦ was FIG. 3. Expression of C-terminal ␣ 2 ␦ deletion mutants and ␦ in COS-7 cells. Membranes were prepared from COS-7 cells transfected with ␣ 2 ␦ expressing constructs or a vector (pcDNA3) only control as described under "Experimental Procedures." Cell membrane proteins (3 g) were resolved on 4 -20% gradient SDS gels and transferred to nitrocellulose. ␣ 2 ␦ was identified by affinity-purified polyclonal antibodies and ECL (Amersham) detection. A sample (3 g) of porcine cerebral cortical membranes (p.c.c.m.) was run as a positive control. Markers are Kaleidoscope markers (Bio-Rad).
FIG. 4. Expression of C-terminal deletions through the ␦ subunit. Membranes were prepared and ␣ 2 ␦ detected as described in Fig. 3, except protein samples (3 g) were resolved on a 6% (A) or 4 -20% gradient (B) SDS gel. expressed alone under reducing conditions. However, mutants A, B, and C, which lack cysteinyl residues, are not detected in the media. The association of mutants B and C with membranes is not disrupted with either buffers of high ionic strength or with a polarity-reducing agent. The nature of this interaction is not known but could be a result of incomplete or defective post-translational processing of the truncated proteins, perhaps resulting in incomplete cleavage of the signal sequence.
To determine the minimum fragment of ␦ required for [ 3 H]gabapentin binding, we examined six mutants with Cterminal deletions of the ␦ component. C-terminally truncated mutants extending to residue 942 (mutant H) and 959 (mutant I) both failed to bind [ 3 H]gabapentin. However, mutants extending to residue 994 (J) and those with smaller truncations (K, L, and M) displayed gabapentin binding activity. Thus, we have identified a 35-residue stretch (residues 960 -994), the deletion of which results in a loss of specific [ 3 H]gabapentin binding. From our data, we cannot say whether this region is directly involved in the formation of the [ 3 H]gabapentin binding pocket or is essential for the structural integrity of the subunit. The two pairs of cysteinyl residues at positions 960/ 963 and 988/990 may contribute to the tertiary structure of the protein by disulfide bridging. In addition, the histidyl residue at 968 and the first pair of cysteinyl residues described above form a sequence similar to the zinc finger motif found in transcription factors (Cys-X 2 -Cys-X 4 -His-X 4 -Cys) (48), although the final cysteinyl residue is separated from the histidyl by 19 residues. This motif is also found in the rat (43) and human (42) neuronal ␣ 2 ␦ sequence as well as the rabbit skeletal muscle splice variant (37). Sequences similar to the one described above make up a growing family of Zn 2ϩ finger motifs (49,50), some of which are found in membrane-associated proteins (51)(52)(53). However, residues 960 -990 in the porcine ␣ 2 ␦ sequence are predicted to be extracellular, based on either the single (37) or the three (40) Once the minimal fragment for [ 3 H]gabapentin binding had been defined, we attempted to design a soluble ␣ 2 ␦ mutant that would be secreted into the medium. To preclude the formation of inappropriate disulfide bridges, we designed mutant L (⌬1040 -1067) to include a conserved Cys residue (Cys-1035) followed by four residues at the start of the putative transmembrane domain in ␦ (see Fig. 2). However, the rest of the hydrophobic anchor and C-terminal tail was deleted. We reasoned that Cys-1053 was unlikely to have a critical bridging role in ␣ 2 ␦ as it is absent in the rat (43), human (42), and rabbit (37) sequences. All of the ␣ 2 ␦ deletion mutants whose expression could be confirmed by Western blotting were detected in particulate fractions. However, large scale expression of mutant L (⌬1040 -1067) yielded a soluble as well as a membrane-associated form (Fig. 5). The latter form could not be extracted with either 1 M NaCl or 10% ethylene glycol; the nature of the interaction of this protein with membranes is unclear. Soluble mutant L was not processed into separate ␣ 2 and ␦ polypeptides. Nevertheless, the soluble form bound [ 3 H]gabapentin with a K D of 30.7 Ϯ 8.1 nM, similar to that for wild-type ␣ 2 ␦ FIG. 5. Expression of a soluble form of ␣ 2 ␦ lacking the ␦ transmembrane domain. Membranes were prepared as described in Fig. 3. The transfection media were also recovered (See "Experimental Procedures" for details), and aliquots (equivalent percentage of total sample) were analyzed. Cell membrane proteins (3 g) and transfection media were resolved on 6% (A and B) or 4 -20% gradient (C and D) SDS gels and transferred to nitrocellulose. ␣ 2 ␦ was detected as described previously (see Fig. 3).
subunits. The membrane-associated form of mutant L consists of both proteolytically cleaved and uncleaved ␣ 2 ␦ protein; the K D for this mixed population was also similar to that for wildtype ␣ 2 ␦ subunits. Although the formation of a soluble mutant is consistent with the single TM model, the model cannot explain the association of a proportion of mutant L with membranes. Other groups, while generally supporting the single TM model, have also noted inconsistencies. Jay et al. (35) found that only half of the ␣ 2 immunoreactivity was removed from muscle microsomal membranes after alkaline extraction in the presence of reducing agents (35). Furthermore, using antipeptide antibodies directed to a region of ␣ 2 predicted to be intracellular according to the three-transmembrane domain model, Brickley et al. (38) showed that immunostaining of dorsal root ganglion cells was enhanced after membrane permeabilization.
In summary, we have identified a region within ␦ (residues 960 -994) that is crucial for the formation of the [ 3 H]gabapentin binding pocket. Mutant L, which lacks most of the ␦ transmembrane domain but includes conserved cysteine residues, is found in both soluble and membrane-associated forms. The soluble form is not proteolytically cleaved into separate ␣ 2 and ␦ peptides but retains high affinity for [ 3 H]gabapentin. Our data broadly support the single transmembrane model of the ␣ 2 ␦ subunit, although the model cannot explain all of the properties of mutant L and the heavily truncated forms of ␣ 2 (mutants B and C), which remain associated with the membrane. The expression of a soluble form of ␣ 2 ␦ that retains high affinity for [ 3 H]gabapentin opens the way for large scale production of mutant L for detailed biochemical and crystallographic studies.