The Ducky Mutation in Cacna2d2 Results in Altered Purkinje Cell Morphology and Is Associated with the Expression of a Truncated (cid:1) 2 (cid:2) -2 Protein with Abnormal Function*

The mouse mutant ducky, a model for absence epilepsy, is characterized by spike-wave seizures and cerebellar ataxia. A mutation in Cacna2d2 , the gene encoding the (cid:1) 2 (cid:2) -2 voltage-dependent calcium channel accessory subunit, has been found to underlie the ducky phenotype. The (cid:1) 2 (cid:2) -2 mRNA is strongly expressed in cerebellar Purkinje cells. We show that du/du mice have abnormalities in their Purkinje cell dendritic tree. The mutation in (cid:1) 2 (cid:2) -2 results in the introduction of a premature stop codon and predicts the expression of a truncated protein encoded by the first three exons of Cacna2d2 , followed by 8 novel amino acids. We show that both mRNA and protein corresponding to this predicted transcript are expressed in du/du cerebellum and present in Purkinje cells. Whereas the (cid:1) 2 (cid:2) -2 subunit increased the peak current density of the Ca traces with no activity from the same voltage protocol in the same experiment and subtracting this average from each episode using pClamp6 (Axon In-struments). Statistical analysis was performed using paired or un- paired Student’s t test. For the single channel analysis, patches were only used in which three or fewer overlapping openings were detected. With an open probability of about 0.5 at (cid:1) 40 mV and at least 20 consecutive stimulations, the number of detectable multiple openings was considered to represent the number of channels active in these patches. Event detection was carried out using the half-amplitude threshold method. Single channel amplitude was determined by a Gaussian fit to the binned amplitude distributions. Mean open and closed times were determined as a single or double exponential fitted to open time distributions. Open time distributions were only collected in episodes or parts of episodes with no overlapping openings. For closed time distributions, we used either single channel patches or segments toward the end of episodes in which only one channel remains active and no further overlaps occur. Data are expressed as mean (cid:8) S.E. For steady-state inactivation, all the available patches were considered, and each was normalized to its peak current response during the prepulse to (cid:1) 40 mV. Latency to first opening was measured in 2-ms bins. First latency (FL) histograms from each experiment were divided by the number of episodes collected, and the plots were then accumu-lated and divided by the number of stimulations, to express the data as the FL probability (23, 27). Two- and three-channel patches were also corrected to 2,

Voltage-gated Ca 2ϩ (Ca V ) 1 channels have been divided functionally into L-, N-, P/Q-, R-, and T-types (1). Each Ca V channel is composed of a pore-forming ␣ 1 subunit, associated at least in the case of the Ca V 1 and -2 subfamilies with an intracellular ␤ subunit responsible for trafficking (2) and a membrane-anchored, but predominantly extracellular, ␣2␦ subunit, whose function is less well defined (2). Ca V 1.1 (␣ 1 S) is also associated with a ␥ subunit (␥ 1 ), and this may be true for other Ca V channels (3), although the neuronal ␥ subunits may also subserve other functions (4). The ␣ 1 subunit determines the main biophysical properties of the channel and is modulated by the other subunits (2). Mammalian genes encoding 10 ␣ 1 , 4 ␤, 8 ␥, and 3 ␣2␦ subunits have been identified (1,5,6).
A number of spontaneous autosomal recessive mutant mouse strains have now been identified, involving mutations in each of the four different subunits that together compose a voltagedependent calcium channel. They all have a similar phenotype that includes cerebellar ataxia and spike-wave seizures. Tottering (Cacna1a tg ) has a point mutation in Ca V 2.1 (␣ 1 A) (7), and a number of alleles of this mutant have now been identified, as summarized recently (8). Lethargic (Cacnb4 lh ) represents a truncation mutation of the ␤ 4 subunit (9). Stargazer (Cacng2 stg ) has a truncation mutation in the ␥ 2 subunit (3), although its role as a calcium channel subunit remains controversial (4,10,11). Finally, the two ducky alleles (Cacna2d2 du and Cacna2d2 du2j ) both predict truncation mutations in the ␣2␦-2 subunit (12).
Homozygotes for the ducky (du) allele are characterized by ataxia and paroxysmal dyskinesia (13). The cerebellum is reduced in size (14), but we have previously found no loss of Purkinje cell (PC) bodies at postnatal day (P) 21 (12). In this study we observed a reduction in calcium channel current in P21 PCs isolated from du/du compared with ϩ/ϩ cerebella (12). The present study provides evidence that the du mutation results in the persistence of PCs with an immature and grossly abnormal morphology, including multiple primary dendrites and a reduction in the size of the PC dendritic tree. This is associated with loss of full-length ␣2␦-2 protein and expression of a truncated mutant ␣2␦-2 protein with aberrant function.

EXPERIMENTAL PROCEDURES
Construction of du-mut1 ␣ 2 -The du mutant 1 construct (du-mut1 ␣ 2 ) was assembled by PCR (Platinum Pfx polymerase; Invitrogen) of du/du total brain cDNA using primers corresponding to the cDNA sequence (GenBank TM AF247140) containing engineered SmaI or SpeI restriction sites. The primer sequences are as follows: F, 5Ј-TTG(C-CCGGG)GAACATGGCGGTGCCCGGCT-3Ј, and R, 5Ј-TCT(CAGGT-C)AGAGTAACCAGAGACCAA-3Ј, with the recognition sites indicated in parentheses. The PCR product was digested with SmaI and SpeI and ligated into the corresponding sites of a modified pMT2 vector (Genetics Institute, Cambridge, MA). Insert sequence fidelity was determined by automated sequencing (PE Biosystems, Warrington, UK).
DNA was extracted by incubating 2 mm of tail-snip tissue in 75 l of 25 mM NaOH, 0.2 mM Na 2 EDTA at 95°C for 30 min followed by cooling to 4°C and addition of 75 l of 40 mM Tris-HCl. 5 l of the resultant solution was amplified in the PCRs. Primers 98F, 5Ј-ACCTATCAG-GCAAAAGGACG-3Ј, and 120R, 5Ј-AGGGATGGTGATTGGTTGGA-3Ј, produce a fragment of 541 bp from a region that is duplicated in the du allele. Digestion with BspHI results in two fragments of 286 and 273 bp from the du allele, whereas the wild-type allele remains uncut. Wildtype mice can be identified by the presence of a single band upon agarose gel electrophoresis. Heterozygous and du/du mice each show two bands and can by distinguished on the basis of their relative intensities.
Four cells of two genotypes (ϩ/ϩ and du/du) were scanned using identical parameters. Selected fields were optically sectioned using 1-m steps. The entire z series was projected as a single composite image by superimposition. The final image was thresholded to form a binary image for analysis by NIH Image J software version 1.62. Following formation of the skeleton, the dendritic tree was contained in a minimal rectangle, and the number of dendrites were counted crossing a horizontal and diagonal line. Branch points were determined from the soma to the end of the three longest dendrites for each cell.
Golgi-Cox Staining-P24 mice were asphyxiated with CO 2 , and their brains were removed from the cranium, immediately immersed in 20 ml of fixative (34 mM K 2 Cr 2 O 7 , 37 mM HgCl 2 , and 23 mM K 2 CrO 4 ), and left undisturbed for 12 weeks in the dark at 4°C. Vibratome sections (100 m) were developed for 20 min in a 5% Na 2 SO 3 solution, before being mounted on microscope slides and coverslipped with Vectashield (Vector Laboratories). Slides were viewed on a Leica microscope using 64ϫ magnification. Image sections were grabbed through an attached CCD camera, using Vision Explorer software (Alliance Vision, Mirmande, France), enabling the deconvolution and projection of different optical planes.
In Situ Hybridization and Immunohistochemistry-Mice (aged P21 to P24) were anesthetized by CO 2 inhalation or pentobarbitone injection (200 mg⅐kg Ϫ1 , intraperitoneal) and perfused intracardially with 4% paraformaldehyde in phosphate-buffered saline (PBS). The brain was placed in 4% paraformaldehyde at 4°C for 3 h and then transferred into 30% sucrose overnight, before embedding in Cry-M-Bed (Bright Instrument Ltd., Huntingdon, UK) and sectioning. Alternatively, the brain was removed without fixation and frozen in liquid nitrogen. 10 -25-m cryostat sections were cut and air-dried onto positively charged slides (Merck).
A cDNA fragment corresponding to Cacna2d2 5Ј (nucleotides 206 -620, using the numbering system from GenBank TM accession number AF247139) was subcloned into pBluescript SKϩ (Stratagene, La Jolla, CA). Sense and antisense RNA probes were prepared using T3 or T7 polymerase and digoxigenin RNA labeling mix and purified using Quickspin columns (Roche Molecular Biochemicals). In situ hybridization was carried out as described (16).
Two polyclonal antipeptide antibodies were raised in rabbits to amino acids 16 -29 and 102-117 of ␣2␦-2. Neither of the peptides used as antigens showed any homology with other protein sequences in the data base. They were purified by affinity chromatography using the immobilized synthetic peptide and stored at 0.5 mg⅐ml Ϫ1 in PBS, pH 7.2, at Ϫ20°C. For immunohistochemistry of ␣2␦-2, 25-m sections were incubated overnight with primary antibody at 4°C (1.25 g⅐ml Ϫ1 ). This was followed by a biotinylated anti-rabbit IgG secondary antibody (Sigma, 7.2 g⅐ml Ϫ1 ) and a Texas Red-streptavidin conjugate (2 g⅐ml Ϫ1 ). Some tissue sections or cells were also incubated for 1 min with the nuclear dye 4Ј,6-diamidino-2-phenylindole (DAPI, 300 nM, Molecular Probes). They were examined by laser scanning confocal microscopy, using 1-m optical sections. For peptide controls the appropriately diluted antibody was incubated for 1 h at 37°C with a 10ϫ higher concentration (w/v) of the immunizing peptide, before applying it to sections.
Transfection of COS-7 Cells-Transfection was performed with Ge-nePORTER transfection reagent (Gene Therapy Systems, San Diego, CA). Cells were plated onto coverslips or dishes, 2-3 h prior to transfection. The DNA and GenePORTER reagent (6 g and 30 l, respectively) were each diluted in 500 l of serum-free medium, mixed, and applied to the cells. After 3.5 h, 1 ml of medium containing 20% serum was added to the cells, which were then incubated at 37°C for 3-4 days. For electrophysiological recordings, cells were re-plated 1-6 h prior to use.
Immunocytochemistry on COS-7 Cells-The method used is essentially as described previously (17). Cells were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature. For permeabilization, cells were incubated twice for 7 min in a 0.02% solution of Triton X-100 in Tris-buffered saline; otherwise, cells were washed twice with Trisbuffered saline for the same period. The primary antibodies were used at 1.25 g⅐ml Ϫ1 . The biotinylated anti-rabbit IgG secondary antibody (7.2 g⅐ml Ϫ1 ) was then incubated for 2 h at 4°C, followed by streptavidin-Texas Red (2 g⅐ml Ϫ1 ). Cells were then incubated for 1 min with DAPI (300 nM).
Preparation of Whole COS-7 Cell Lysates-COS-7 cells were transfected with either ␣2␦-2 or du-mut1 ␣ 2 as described above. On day 4 post-transfection, the cells were resuspended in detergent-free Buffer A. Samples (2 mg of total protein) were then solubilized in Buffer A containing 1% CHAPS for immunoaffinity purification of ␣2␦-2 as described below.
Preparation of Cerebellar Tissue Homogenate-Cerebella, stored at Ϫ80°C, were thawed in ice-cold Buffer A (10 mM HEPES, pH 7.4, 150 mM NaCl, 2 mM EDTA, and protease inhibitors (Complete EDTA-free, Roche Molecular Biochemicals, 1 tablet/50 ml buffer)), plus 150 mM sucrose. The tissue was homogenized, and the homogenate was centrifuged at 5000 ϫ g for 10 min at 4°C. The resultant supernatant was diluted 4 times with Buffer A, prior to detergent treatment. The protein concentration of samples was determined using either the bicinchoninic acid (BCA) assay (Perbio, Tattenhall, UK) in the presence of 0.5% SDS, or where the samples were in SDS-PAGE buffer, by a modified filter paper dye-binding assay (18).
Immunoaffinity Purification of ␣2␦-2 and du-mut1 ␣ 2 -Affinitypurified Ab (16 -29) or Ab(102-117) (2 mg) was covalently coupled to a 1-ml column of Sepharose-NHS (Amersham Biosciences) according to manufacturer's instructions. The efficiency of coupling was assessed by SDS-PAGE. The columns were pre-washed with 200 mM glycine HCl, pH 2.4, and neutralized before first use to ensure complete removal of any residual unbound IgG. Control columns contained 2 mg of protein A-Sepharose-isolated IgG from the corresponding pre-immune rabbit sera. COS-7 cells transfected with either ␣2␦-2 or du-mut1 ␣ 2 (2 mg of total protein) or cerebellar homogenate (ϩ/ϩ or du/du, 6-20 mg of total protein) were solubilized in Buffer A (see above) containing 1% (w/v) CHAPS and placed on ice for 30 min following sonication 3 times for 10 s. The detergent extracts were cleared by centrifugation (48,000 ϫ g, 20 min, 4°C) and then applied to the relevant antibody or control IgG columns that had been pre-equilibrated with Buffer A containing 0.5% CHAPS. The lysates were recirculated through the columns using two opposing syringes for 60 min at 4°C. Unbound material was washed from the columns in Buffer A containing 0.5% CHAPS and then Buffer A ϩ 0.1% CHAPS. Bound proteins were eluted with 200 mM glycine HCl, pH 2.4, and then concentrated for SDS-PAGE and Western blot analysis by precipitation with 10% trichloroacetic acid followed by centrifugation (30,000 ϫ g, 30 min, 4°C). Samples of the precipitated proteins were separated on either 10 or 4 -20% gradient gels under reducing conditions and then electrophoretically transferred to polyvinylidene difluoride membranes for immunodetection. The polyvinylidene difluoride membranes were blocked with 3% BSA for 3 h at 55°C and incubated overnight with a 1:1000 dilution of either Ab (16 -29) or Ab(102-117) followed by a 1:1000 dilution of goat anti-rabbit IgG-horseradish peroxidase conjugate (Bio-Rad) for 1 h at 20°C and detected using ECL (Amersham Biosciences). The antibody and control IgG columns were neutralized and washed with 50 mM Tris, pH 7.4, 1 M NaCl, and 0.02% sodium azide and stored in PBS, pH 7.2, with 0.02% sodium azide at 4°C.
Heterologous Expression of cDNAs and Electrophysiology-Gen-Bank TM accession numbers of cDNAs are given in parentheses. Calcium channel expression in COS-7 cells was investigated by whole cell patch clamp recording, essentially as described previously (19), by transfection of rat Ca V 2.1 (M64373) E1686R (20), in conjunction with rat ␤ 4 (LO2315) and mouse ␣2␦-2 (AF247139, common brain splice variant, lacking exon 23 and 6 bp of exon 38 (21)) or du-mut1 ␣ 2 (AF247140) cDNAs cloned into the pMT2 vector. The cDNA for green fluorescent protein (mut3 GFP) (22) was included in the transfection to identify transfected cells from which recordings were made. Transfection was performed as described above, using the ratios for ␣ 1 , ␤, ␣2␦, and GFP of 3:1:1:0.1. In the single channel experiments cDNA for ␤-adrenergic receptor kinase 1 (␤-ARK1), G␤␥ binding domain was included in the transfections, at the same concentration as the ␤ subunit cDNA (23). For expression in Xenopus oocytes, cDNAs encoding rabbit Ca V 2.1 (X57689), rat ␤ 4 and mouse ␣2␦-2 or du-mut1 ␣ 2 cDNAs were injected intranuclearly as described previously (24), except that 4 nl of the 1:1:1 ratio cDNA mixture was injected at 1 g⅐l Ϫ1 . In control experiments where ␣2␦-2 was omitted, the ratio was made up with buffer. Recordings were made using two-electrode voltage clamp as described previously (24), using 10 mM Ba 2ϩ as charge carrier. Individual I-V relationships were fitted with a modified Boltzmann Equation 1, where G max is the maximum conductance; V rev is the reversal potential; k is the slope factor; and V 50 is the voltage for 50% current activation.
Single Channel Recording and Analysis-All recordings were taken from cell-attached patches on GFP-positive cells at room temperature (20 -22°C). Recording pipettes were pulled from borosilicate tubes (World Precision Instruments, Inc., Sarasota, FL), coated with Sylgard (Sylgard 184, Corning Glass), and fire-polished to form high resistance pipettes (ϳ10 megohms with 100 mM BaCl 2 ). The bath solution, designed to zero the resting membrane potential (25), was composed of (in mM) 135 potassium aspartate, 1 MgCl 2 , 5 EGTA, and 10 HEPES (titrated with KOH, pH 7.3), and patch pipettes were filled with a solution of the following composition (in mM): 100 BaCl 2 , 10 tetraethylammonium-Cl, 10 HEPES, 200 nM tetrodotoxin, titrated with tetraethylammonium-OH to pH 7.4. Both solutions were adjusted to an osmolarity of 320 mosmol with sucrose. Data were sampled (Axopatch 200B and Digidata 1200 interface, Axon Instruments, Foster City, CA) at 20 kHz and filtered on-line at 1-2 kHz. Voltages were not corrected for liquid junction potential (26), measured to be Ϫ15 mV in these solutions, in order to be able to compare the results to other published material. Leak subtraction was performed by averaging segments of traces with no activity from the same voltage protocol in the same experiment and subtracting this average from each episode using pClamp6 (Axon Instruments). Statistical analysis was performed using paired or unpaired Student's t test. For the single channel analysis, patches were only used in which three or fewer overlapping openings were detected. With an open probability of about 0.5 at ϩ40 mV and at least 20 consecutive stimulations, the number of detectable multiple openings was considered to represent the number of channels active in these patches. Event detection was carried out using the half-amplitude threshold method. Single channel amplitude was determined by a Gaussian fit to the binned amplitude distributions. Mean open and closed times were determined as a single or double exponential fitted to open time distributions. Open time distributions were only collected in episodes or parts of episodes with no overlapping openings. For closed time distributions, we used either single channel patches or segments toward the end of episodes in which only one channel remains active and no further overlaps occur. Data are expressed as mean Ϯ S.E. For steady-state inactivation, all the available patches were considered, and each was normalized to its peak current response during the prepulse to ϩ40 mV. Latency to first opening was measured in 2-ms bins. First latency (FL) histograms from each experiment were divided by the number of episodes collected, and the plots were then accumulated and divided by the number of stimulations, to express the data as the FL probability (23,27). Two-and three-channel patches were also corrected for the apparent number of channels in the patch, according to Equation 2, 1 and FL N are the single channel and multichannel cumulative first latency functions, respectively, and N is the apparent number of channels.

Comparison of Morphology of du/du and ϩ/ϩ PCs-
The du/du cerebellum exhibits normal foliation and laminar structure. There are no gaps in the PC layer, and their perikarya form a single row. The thickness of each layer is reduced in du/du compared with wild-type littermates as described previously (12). Two techniques were used to compare the morphology of cerebellar PCs between the different genotypes in more detail as follows: first, classical Golgi impregnation, and second, microinjection of PC somata with Lucifer Yellow and neurobiotin. Both methods revealed a changed PC cytoarchitecture in homozygous du/du cerebella at P21-26, which was the latest period in which morphology could be studied, given that the du/du mice die by P35. In the Golgi-impregnated cerebella, PCs were examined in detail in multiple cerebellar sections from one ϩ/ϩ and six du/du mice. Atypical initial lateral extensions of the primary dendrite were seen in du/du but not in ϩ/ϩ PCs (compare ϩ/ϩ in Fig. 1A to du/du in Fig. 1, B, C, and F). The primary dendrites may then bend apically in a delayed targeting of the pial surface, which they frequently fail to reach (e.g. Fig. 1, B and C). Thickened tertiary branchlets were found to bend downwards in some du/du PCs, giving a "weeping willow" appearance ( Fig. 1B, closed arrowhead). Additionally, PC somata are frequently multipolar, exhibiting up to three primary dendrites (Fig. 1, D-F), which, in the cells shown in Fig. 1, E and F, extend laterally rather than targeting the pial surface.
Similar results were obtained with the cell-filling technique, for which 4 ϩ/ϩ and 4 du/du PCs were examined in detail. The du/du PCs displayed a dendritic arbor that was significantly less complex, reduced in size, and frequently did not reach the border of the molecular layer (compare the typical ϩ/ϩ PC in Fig. 2A with three du/du PCs shown in Fig. 2B). Additionally, the shafts of the main and secondary dendritic branches of du/du PCs were often thickened (Fig. 2B). In one du/du PC, the dendrites drooped down toward the granule cell layer giving a weeping willow appearance (Fig. 2B, top right panel), similar to that seen in some of the Golgi-impregnated du/du PCs (Fig.  1B). Following formation of the skeleton of the dendritic trees (Fig. 2C), the numbers of dendrites were found to be reduced in du/du PCs (Fig. 2D, left panel), and the total length of the dendritic tree was also reduced (Fig. 2D, center panel). How- ever, the number of branch points on the longest dendrites was unchanged (Fig. 2D, right panel).
The du 5Ј Mutant Transcript Is Present and Translated in du/du Cerebellar PCs-We have shown previously, by in situ hybridization using a 3Ј antisense probe, that no full-length transcript for Cacna2d2 is present in du/du cerebellum, whereas a strong signal was obtained in ϩ/ϩ cerebellar PCs (12). In the present study we performed in situ hybridization with a 5Ј Cacna2d2 antisense RNA probe to examine whether a truncated message was present in du/du PCs. This confirmed the presence of full-length transcript for Cacna2d2 in ϩ/ϩ PCs (Fig. 3A). The data would also not be inconsistent with the additional presence of transcript in small Bergmann glial cell bodies (Fig. 3A, arrows, see also Fig. 8). The results also demonstrate a low level of message hybridizing to the 5Ј probe in du/du PCs (Fig. 3B, compared with Fig. 3A). This, together with the absence of full-length transcript in du/du PCs shown in our previous study (12), provides evidence for a low level of transcription of du mutant transcript 1 (12) in these cells.
We therefore generated an ␣2␦-2 antipeptide antibody utilizing an immunizing peptide corresponding to amino acids 102-117, with the intent of examining whether a mutant protein was expressed in du/du cerebellum. This sequence is near the N terminus of ␣2␦-2 and is also present in the predicted protein product of the du mutant transcript 1, termed du-mut1 ␣ 2 (Fig. 4A). This antibody, called Ab(102-117), was first characterized against heterologously expressed ␣2␦-2. On Western blots of gels run under reducing conditions, ␣ 2 is separated from the ␦ moiety to which it is disulfide-bonded under native conditions. Ab(102-117) specifically recognized the ␣ 2 moiety of ␣2␦-2 (as a broad band at about 150 kDa) and not the ␣ 2 moiety of ␣2␦-1 when both ␣2␦-1 and ␣2␦-2 were overexpressed in COS-7 cells (Fig. 4B). It also recognized an ϳ10-kDa protein product of the du-mut1 ␣ 2 cDNA expressed in COS-7 cells (Fig.  4B). When Ab(102-117) was used to examine the immunocytochemical localization of ␣2␦-2 in these cells, the epitope was accessible in non-permeabilized cells, orienting it exofacially (Fig. 4C, 1st and 2nd columns, upper panel). Additional intracellular staining was observed when the cells were permeabilized (Fig. 4C, 1st and 2nd columns, lower panel). When du-mut1 ␣ 2 was expressed, very little immunostaining was observed in non-permeabilized cells (Fig. 4C, 3rd column, upper panel), although a large number of cells were present in the field (Fig. 4C, 4th column, upper panel), whereas intense intracellular immunostaining, localized to intracellular organelles, was observed when the cells were permeabilized (Fig. 4C, 3rd  and 4th columns, lower panel). The lack of staining with the anti-GFP antibody in non-permeabilized GFP-positive cells provides further evidence that the plasma membrane of the cells has not been permeabilized by fixation (Fig. 4D).
The predicted protein molecular mass of the entire du-mut1 ␣ 2 -truncated protein is 16 kDa, which is larger than the ϳ10-kDa band observed here. We therefore utilized another antipeptide antibody, generated against the epitope represented by amino acids 16 -29, which is within the predicted signal sequence of ␣2␦-2, to further examine the processing of the du-mut1 ␣ 2 protein. When du-mut1 ␣ 2 was expressed in COS-7 cells, Ab(16 -29) recognized a 16-kDa band in lysates of these cells (Fig. 5A, lane 1). No smaller molecular weight bands were observed on this gel (but see Fig. 7). This suggested that the signal sequence of du-mut1 ␣ 2 is at least partly uncleaved when it is expressed in COS-7 cells. Ab (16 -29) also recognized a well defined 120-kDa protein when full-length ␣2␦-2 was expressed (Fig. 5A, lane 2) and no bands when ␣2␦-1 was expressed as a control (Fig. 5A, lane 3). This 120-kDa protein is likely to represent an immature form of ␣2␦-2 before cleavage of the signal sequence, which normally precedes glycosylation. The protein molecular mass of the ␣ 2 moiety including the 6-kDa signal sequence is calculated to be 113 kDa. Unlike Ab(102-117), Ab(16 -29) did not recognize a band of 150 kDa, indicating that, as expected, the signal sequence is cleaved from the mature glycosylated form of ␣2␦-2.
We then examined the immunolocalization of du-mut1 ␣ 2 in COS-7 cells, using Ab (16 -29). Immunostaining for this epitope was not observed at the plasma membrane in non-permeabilized cells expressing ␣2␦-2, indicating that the signal sequence is cleaved before insertion of ␣2␦-2 into the plasma membrane. Furthermore, this epitope was also generally not observed on the exofacial side of the plasma membrane when du-mut1 ␣ 2 was expressed (Fig. 5B, non-permeabilized cells). Diffuse intracellular staining was observed when the cells were permeabilized, for both ␣2␦-2and du-mut1 ␣ 2 -expressing cells (Fig. 5B).
Immunopurification of du-mut1 ␣ 2 from du/du Cerebellum-The use of an Ab(102-117) immunoaffinity column allowed the isolation of a low abundance protein of ϳ10 kDa from du/du cerebellum (Fig. 6A, lane 1), which was detected using the same antibody. This protein is very similar in molecular weight to the du-mut1 ␣ 2 protein isolated in the same way from lysates of COS-7 cells expressing du-mut1 ␣ 2 (Fig. 6A, lane 2). If a protein product were produced from du mutant transcript 2 (Gen-Bank TM accession number AF247141) in du/du cerebellum (12), its predicted molecular mass would be ϳ100 kDa. This would also be recognized by Ab(102-117), but no higher molecular weight immunoreactive bands were observed from 4 to 20% gradient gels of proteins isolated from du/du cerebellum (data not shown, n ϭ 2). A broad band of protein of ϳ150 kDa, representing the ␣ 2 moiety of ␣2␦-2, was isolated from ϩ/ϩ cerebellum using the same immunoaffinity column and detected using the same antibody (Fig. 6B, lane 1). This protein was the same molecular weight as that isolated by the same antibody from COS-7 cells transfected with ␣2␦-2, with the broad band probably representing different glycosylation states (Fig. 6B, lane 2).
By using an Ab(16 -29) immunoaffinity column, a protein of ϳ16 kDa was isolated from du/du cerebellum (Fig. 6C, lane 1). This protein is very similar in molecular weight to the du-mut1 ␣ 2 protein isolated in the same way from COS-7 cells expressing du-mut1 ␣ 2 (Fig. 6C, lane 2). Furthermore, a protein of about 120 kDa was isolated from ϩ/ϩ cerebellum using the same immunoaffinity column (Fig. 6D, lane 1). This protein was the same molecular weight as that isolated by the same antibody from COS-7 cells transfected with ␣2␦-2 (Fig. 6D, lane 2).
The basis for the difference in molecular weight between the du-mut1 ␣ 2 species recognized by the two antibodies was fur- ther examined by PAGE of a larger amount (80 g of protein) of COS-7 cell lysate expressing du-mut1 ␣ 2 , on a high percent gradient gel, followed by immunoblotting (Fig. 7A). A 16-kDa band was recognized by both antibodies, but the predominant band recognized by Ab(102-117) was ϳ10 kDa. This was not recognized by Ab (16 -29), which recognized an additional 6-kDa band. The amino acid sequence of du-mut1 ␣ 2 is shown in Fig. 7B, with the antibody recognition sites (boldface letters) and the predicted site of cleavage of the signal sequence (arrow). The predicted sizes of the peptides obtained before and after cleavage are 16 kDa for full-length du-mut1 ␣ 2 , 10 kDa for du-mut1 ␣ 2 following cleavage of the signal sequence, and 6 kDa for the cleaved signal sequence (Fig. 7C), in agreement with the experimental results.
The immunolocalization of ␣2␦-2 in the cerebellum was subsequently examined using Ab (102-117). This antibody gave a pattern of immunostaining in the Purkinje cell layer (PCL) and molecular layer (ML) in sections of ϩ/ϩ cerebellum, consistent with localization of ␣2␦-2 in PC bodies and dendrites (Fig. 8A, left  panel). The staining was lost when the primary antibody was preincubated with the immunizing peptide or when the primary antibody was not used (Fig. 8A). In cerebellar sections from du/du mice, Ab(102-117) gave a low level of immunostaining (Fig. 8A, right panel), consistent with the presence of du-mut1 ␣ 2 in du/du PCs. There was also some evidence for staining of Bergmann glia (Fig. 8A, right panel marked with *). We also used Ab (16 -29) to determine where the uncleaved form of du-mut1 ␣ 2 was expressed endogenously in du/du cerebellar sections. A low level of immunostaining was observed with this antibody in ϩ/ϩ PCs, consistent with the presence of the immature 120-kDa form of ␣2␦-2, with the signal peptide still present, which is likely to be localized to the endoplasmic reticulum (Fig. 8B, left panel). Fur-ther evidence was obtained here for staining of Bergmann glia (Fig. 8B, left panel, *). In du/du cerebellum, we observed that immunostaining with this antibody was concentrated largely in the cell bodies of PCs (Fig. 8B, right panel). This is likely to represent the 16-kDa uncleaved du-mut1 ␣ 2 species. The immunostaining was lost when the antibody was preincubated with the immunizing peptide (Fig. 8B). No differences were observed between ϩ/ϩ and du/du PCs when Bergmann glia were visualized using an anti-glial fibrillary acidic protein antibody (results not shown).
Modulation of Ca V 2.1 Ca 2ϩ Channel Currents by ␣2␦-2 and du-mut1 ␣ 2 -The possible pathological function of the du-mut1 ␣ 2 protein encoded by the Cacna2d2 du gene was investigated using in vitro expression and electrophysiology. To mimic the PC complement of calcium channel subunits, the cDNAs corresponding to rat Ca V 2.1 and ␤ 4 were transfected into COS-7 cells, with or without ␣2␦-2 or du-mut1 ␣ 2 cDNA, and the resulting Ca V currents (I Ba ) recorded. Co-expression of ␣2␦-2 increased Ca V 2.1/␤ 4 I Ba currents, inducing a 2.9-fold enhancement of amplitude at 0 mV (Fig. 9A, and I-V relationships in Fig. 9B), with no significant shift in the voltage dependence of current activation (V 50 for activation was Ϫ8.7 Ϯ 0.7 mV (n ϭ 28) for Ca V 2.1/␤ 4 and Ϫ10.7 Ϯ 0.8 mV (n ϭ 42) for Ca V 2.1/␤ 4 / ␣2␦-2). There was no significant effect of ␣2␦-2 on the activation or inactivation of the Ca V 2.1/␤ 4 combination ( Fig. 9A and results not shown).
In contrast, co-expression of du-mut1 ␣ 2 induced a consistent reduction in Ca V 2.1/␤ 4 I Ba amplitude throughout the voltage range (Fig. 9, C and D). This amounted to a 51% inhibition at 0 mV (Fig. 9E) and also resulted in a ϩ5-mV shift in V 50 for activation to Ϫ3.5 Ϯ 1.1 mV (n ϭ 12, p Ͻ 0.01 compared with Ca V 2.1/␤ 4 ). The mean current densities at 0 mV under the two FIG. 6. Immunopurification of du-mut1 ␣ 2 from du/du cerebellum and the ␣ 2 moiety of full-length ␣2␦-2 from ؉/؉ cerebellum. Immunoblot analysis of ␣2␦-2 proteins were immunocaptured from detergent-solubilized du/du or ϩ/ϩ cerebellar membranes using column-immobilized peptide antibodies. Samples were separated on either 4 -20% gradient gels (A and B), a 20% gel (C), or on a 7.5% gel (D).  (16 -29) also recognizes a band of 6 kDa, and Ab(102-117) recognizes a major band of ϳ10 kDa. B, amino acid sequence of du-mut1 ␣ 2 . The respective binding sites for Ab (16 -29) and Ab(102-117) are shown in bold. The entire signal leader sequence is underlined with its predicted cleavage site marked by an arrow (40). C shows the calculated molecular mass values for du-mut1 ␣ 2 (16 kDa) and the two proteolytic fragments as follows: du-mut1 ␣ 2 following cleavage of the signal sequence (solid line, ϳ10 kDa) and the cleaved signal sequence (dotted line, ϳ6 kDa). different conditions are compared in Fig. 9E. Similar results were obtained when these calcium channel subunits, in this case including rabbit Ca V 2.1, were expressed in Xenopus oocytes (see Ref. 12 and data not shown).
Effect of ␣2␦-2 and du-mut1 ␣ 2 on Single Ca 2ϩ Channel Currents Formed by Ca V 2.1-We compared single channel parameters between cell-attached patches of COS-7 cells transfected with Ca V 2.1/␤ 4 cDNA, either without ␣2␦-2 (Fig. 10A) or co-expressed with either full-length ␣2␦-2 (Fig. 10B) or the du-mut1 ␣ 2 (Fig. 10C). These experiments were performed in order to differentiate between a mechanism that involves changing the biophysical properties of Ca V 2.1 channels and a mechanism that involves changing the trafficking or membrane expression levels of the Ca V 2.1 channels, imposed by either ␣2␦-2 or du-mut1 ␣ 2 .
Once opened, Ca V 2.1 channels showed an average single channel conductance of 9.9 Ϯ 0.4 pS (n ϭ 8) for Ca V 2.1/␤ 4 , which was not significantly affected by co-expression of ␣2␦-2 (10.2 Ϯ 0.6 pS, n ϭ 8) or du-mut1 ␣ 2 (8.8 Ϯ 1.0 pS, n ϭ 6) (Fig.  10D, left). This conductance is similar to that of P-type channels recorded from wild-type and du/du PCs under the same conditions (12). More detailed analysis demonstrated openings to three distinct amplitude levels, as has also been shown in native Purkinje cells (28), level 2 being the most prominent in our recordings (Fig. 10D, middle and right, see legend for conductance and amplitude values). Neither the conductance nor the amplitude of the three current levels was significantly affected by expression of ␣2␦-2 or du-mut1 ␣ 2 (data not shown).
Neither ␣2␦-2 nor du-mut1 ␣ 2 caused any significant change in mean open or closed times or in the pattern of voltage dependence of Ca V 2.1 channels (Fig. 10E). We also examined the activation kinetics by measuring the latency to first opening of the channels in response to a square voltage pulse (Fig.   10F). Ca V 2.1 channel activation was not influenced by the subunits examined (Fig. 10F, left), at any voltage (Fig. 10F,  right). In addition, the voltage dependence of inactivation (Fig.  10G) was not influenced by either ␣2␦-2 or du-mut1 ␣ 2 .
Although the presence of ␣2␦-2 caused an ϳ3-fold increase in whole cell current amplitude, all the single channel parameters were indistinguishable between the three conditions. This implies that the basic active unit in the whole cell current (an individual channel) remains unchanged, and the modulation by ␣2␦-2 must involve an alteration in the number of active channels in the membrane.

DISCUSSION
PCs from du/du Mice Have a Reduced Dendritic Arbor-PC somata form a monolayer by 10 days postnatally in the mouse, and their dendrites reach the pial surface at day 20, coinciding with the completion of granule cell migration and concomitant parallel fiber production (29). The PC soma typically exhibits one primary dendrite, which emerges apically, and one axonal process projecting in the opposite direction. The PC dendritic trees develop most dramatically between postnatal day 9 and 20, reaching 80% of their adult dimension in this period (30). PCs from du/du mice appear immature, reduced both in size and complexity, with multiple primary dendrites and small arbors that often terminate well below the pial surface. Thickened secondary and tertiary dendritic trunks are also present. The multipolar appearance of some of the du/du PCs may be a remnant of their immature stage (in which the normal resorption of all somatic filopodia fails to occur), with some of these processes continuing to develop into dendrites as found in weaver and staggerer mouse mutants (31).
Thus, although we have shown that PCs are not lost in du/du cerebella at P21 (12), we now find that the PC dendritic tree is reduced in size and shows other abnormalities, such as weeping willow dendrites and dendritic thickening. Similar abnormalities have been found in a number of the spontaneously occurring Ca V 2.1 mouse mutants (8), and some of these (in particular tg la ) also show PC loss in older mice (32). The mechanism of the altered PC morphology in du/du mice may result . F, FL probability histogram, for the three conditions, respectively (E, n ϭ 8; Ⅺ, n ϭ 8; ‚, n ϭ 5). Left, mean (Ϯ S.E., shown every 10 ms for clarity) cumulative FL probability distributions at ϩ30 mV, and right, FL probability at 20 ms for all voltages examined. G, steady-state inactivation of Ca V 2.1 ensemble currents from single and multichannel patches. Left, an example of inactivation experiment. Top, the voltage protocol, holding potential, Ϫ100 mV, followed by a 2 s. Prepulse (pp) to a potential (V pp ) between Ϫ80 and ϩ40 mV, followed by a test pulse to ϩ40 mV for 100 ms. Bottom, average ensemble currents (from the data as shown in B with a V pp of ϩ40 mV (left) and test pulse responses to V pp of Ϫ80, ϩ10, and ϩ40 mV are superimposed. Right, all currents were normalized to the peak prepulse current at ϩ40 mV and the ratio plotted as a function of V pp . Data of the type shown in (A-C), respectively (E, n ϭ 11; Ⅺ, n ϭ 11; ‚, n ϭ 4). either from the reduced PC calcium channel currents, which we observed in P5-P9 PCs, before the extensive growth of the dendritic arbor (12) or more directly from the loss of ␣2␦-2, with the possible additional consequences of expression of a truncated mutant ␣ 2 protein. A number of the human genetic diseases involving Ca V 2.1, for example familial hemiplegic migraine (33), cerebellar ataxia, and PC degeneration are associated with mutations that have been shown to produce a reduction in Ca V 2.1 calcium currents in vitro (34). However, the mechanism whereby such molecular changes are translated into morphological and functional abnormalities remains to be determined.
A Truncated Mutant Protein Derived from the 5Ј Mutant Transcript of Cacna2d2 Is Expressed in du Mice-The in situ hybridization study demonstrates that although wild-type Cacna2d2 transcript is absent from the brain of du/du mice, because of the genomic rearrangement that disrupts Cacna2d2 (12), a 5Ј mutant transcript (du mutant transcript 1) is present in du/du PCs. This transcript is predicted to encode a protein (du-mut1 ␣ 2 ) that lacks most of the ␣ 2 subunit and the whole of the ␦ subunit, including its transmembrane domain. It is frequently the case that mRNA encoding mutant transcripts, where a frameshift or point mutation introduces one or more premature stop or nonsense codons, is unstable and subject to nonsense-mediated mRNA decay (35). Indeed, although a second mutant transcript 2, predicted to be formed from exons 2-39, was identified by reverse transcriptase-PCR and Northern blot in du/du mouse brain, it was not observed by in situ hybridization in du/du PCs (12). Furthermore, the 5Ј du mutant transcript 1 appeared to be present at a low level in du/du brain (12). To determine whether this mutant transcript was translated, we used two ␣2␦-2 anti-peptide antibodies, which were raised against peptides within the du-mut1 ␣ 2 sequence, Ab (16 -29) and Ab(102-117).
It has been established, from studies with site-directed antipeptide antibodies, that the topology of the ␣2␦-1 subunit is such that the ␣ 2 subunit, which has an N-terminal leader signal sequence, is entirely extracellular (36 -38). The ␣ 2 subunit is disulfide-bonded to a transmembrane ␦ subunit, and both subunits have been found to be involved in the interaction with the Ca v 1.2 subunit (38,39). Now that two other ␣2␦ subunit genes have been cloned, it is assumed that they have the same topology, and indeed, high homology is present between the N termini of ␣2␦-1 and ␣2␦-3, with the clear prediction of a cleaved signal peptide in both sequences. In contrast, although a putative signal peptide is found in ␣2␦-2, it is much longer. By using prediction analysis, it is found to have a potential cleavage site after position 64 (40) (Fig. 7B), whereas only 2% of eukaryotic signal peptides are longer than 35 residues (40). In particular, it has a longer sequence N-terminal to the putative hydrophobic signal sequence (ϳ42 amino acids) than ␣2␦-1 or ␣2␦-3 (which are ϳ3 and 11 amino acids, respectively). Such "n regions" are found to be less than 25 amino acids in 80% of secreted or transmembrane proteins where they occur (41). Therefore, it remains unclear whether this signal sequence is cleaved efficiently, as cleavage is often delayed when the signal sequence is long (42). This results in extended transit times through the endoplasmic reticulum-Golgi apparatus, which may be required for highly glycosylated proteins (42). Such an explanation is likely to be the reason for our observation using Ab (16 -29), of a 120-kDa immunolabeled protein when ␣2␦-2 was expressed in COS-7 cells. This is likely to represent the ␣ 2 moiety of full-length ␣2␦-2 (predicted protein molecular mass of 113 kDa), which is immature in that it has an uncleaved signal peptide and, judging by the molecular weight, no added carbohydrate.
It appears that in the case of du-mut1 ␣ 2 expressed in COS-7 cells, the truncated protein is processed such that the signal sequence remains at least in part uncleaved, because both Ab (16 -29) and Ab(102-117) recognized a band of ϳ16 kDa, the predicted size for the uncleaved du-mut1 ␣ 2 , and Ab (16 -29) also recognized a fainter band of about 6 kDa, which would represent the cleaved signal peptide. However, the predominant band recognized by Ab(102-117) but not Ab(16 -29) was a ϳ10-kDa protein, which is therefore likely to represent du-mut1 ␣ 2 with its signal peptide cleaved. This result corresponded exactly with the molecular weight of the native du-mut1 ␣ 2 immunocaptured from du/du cerebellum by the same antibody, indicating that it is a stable in vivo species in these mice. This study also confirmed the previous indication (12) that du mutant transcript 2, which would be recognized by Ab (102-117), is not translated. A 16-kDa protein was immunocaptured by Ab (16 -29) from du/du cerebellum, indicating that the signal sequence remains, in part, uncleaved from du-mut1 ␣ 2 . The reason that this species was not also immunocaptured by Ab(102-117) may indicate that Ab(102-117) is of lower affinity, as also suggested by the data in Fig. 8.
Immunolocalization of ␣2␦-2 and du-mut1 ␣ 2 in Cerebellum-In cerebellar sections, we found, using Ab(102-117), that ␣2␦-2 is expressed in wild-type PC somata and also in the ML of the cerebellum, suggesting localization in PCs. It is also possible that some of the immunostaining arises from cerebellar afferents or from Bergmann glia, and this will be investigated in the future. In du/du cerebellum, a low level of immunostaining was observed with the same antibody. These results support the finding that du-mut1 ␣ 2 is expressed in du/du cerebellum. Immunoreactivity in du/du cerebellar sections was also observed using Ab (16 -29), where staining, presumably representing the uncleaved du-mut1 ␣ 2 , was concentrated in PC somata. In agreement with the expression study in COS-7 cells and the immunopurification data from cerebellum, this suggests that du-mut1 ␣ 2 retains, in part, the putative signal sequence at its N terminus and does not appear to be secreted. When du-mut1 ␣ 2 was expressed in COS-7 cells both Ab (16 -29) and Ab(102-117) recognized an epitope that was only expressed intracellularly, indicating that du-mut1 ␣ 2 is unlikely to be secreted or inserted into the plasma membrane as a transmembrane protein.
The Functional Interaction of the Ca V 2.1/␤ 4 Combination with ␣2␦-2-The similarity of the ducky phenotype to that observed in mice with mutations in genes encoding the Ca V 2.1 (7) and ␤ 4 (9) subunits and their predominant PC expression pattern suggests that ␣2␦-2 contributes to the P-type current. This is reinforced by our finding that the currents formed by both rat and rabbit Ca V 2.1 co-expressed with ␤ 4 , which is the main PC ␤ subunit, were strongly enhanced by ␣2␦-2, in two expression systems (COS-7 cells and Xenopus oocytes).
Previous in vitro studies have shown that ␣2␦-1, ␣2␦-2, and ␣2␦-3 subunits act to increase the maximum conductance of a number of expressed calcium channel ␣ 1 /␤ subunit combinations at the whole cell level (2,(43)(44)(45)(46). However, this may be dependent to some extent on the specific combination of ␣ 1 and ␤ subunits expressed. Furthermore, the effects of ␣2␦ subunits on kinetics and voltage dependence of activation are more minor (2,44). We have also investigated this for the calcium channel subunit combinations used in the present study, and we show that ␣2␦-2 had no influence on voltage-dependent properties and had no effect on single channel conductance or other biophysical parameters of the Ca V 2.1/␤ 4 channels themselves. This implies that ␣2␦-2 probably has its main effect on the lifetime of the channel complex in the plasma membrane, either by enhancing trafficking or reducing turnover. In agree-ment with this proposed mechanism, it has previously been found that ␣2␦-1 increased the amount of Ca V 1.2 protein expressed in Xenopus oocytes (47).
In contrast, the protein product of du mutant transcript 1, du-mut1 ␣ 2 , produced a consistent reduction in Ca V 2.1/␤ 4 currents in COS-7 cells. Thus, whereas loss of full-length ␣ 2␦-2 is likely to be the most important contributing factor, the expression of the truncated du-mut1 ␣ 2 may also contribute to the du/du phenotype, via an additional suppressive effect, possibly by interfering with the correct trafficking of ␣ 1 subunits.