entla, a novel epileptic and ataxic Cacna2d2 mutant of the mouse.

entla (ent) is a novel recessive phenotype of mice. The underlying mutation was mapped to chromosome 9 (60.1 centimorgans) and identified as an allele of the Cacna2d2 gene encoding the alpha2delta-2 subunit of voltage-gated calcium channels. The Cacna2d2entla allele harbors a 38-kb duplication comprising the 117 nucleotides of exon 3. The predicted duplication of 39 amino acid residues near the subunit's N terminus results in the expression of a full-length, membrane-associated protein. Western blot data were consistent with correct cleavage of the alpha2delta-2entla precursor into alpha2entla and delta2 proteins but indicated loss of the disulfide linkage between the two proteins. ent/ent mice develop ataxia by postnatal day 13-15, followed by paroxysmal dyskinesia a few days later. Two distinct types of cortical and hippocampal epileptic activity at 2 and 4 Hz were recorded, indicative of absence epilepsy. Homozygotes display reduced size and weight, increased mortality before weaning, and female infertility. No overt neuroanatomical abnormalities were detected. Ca2+ current densities recorded from acutely dissociated Purkinje cells of homozygous entla animals were reduced by 50% compared with wild type. Ligand binding assays using the antiepileptic drug [3H]gabapentin, a specific ligand of the alpha2delta-1 and alpha2delta-2 subunits, revealed a >60% reduced maximum binding to cerebellar membranes of ent/ent compared with unaffected littermates. entla is allelic to ducky and ducky2J, representing the third murine Cacna2d2 allele identified and so far the only one encoding an untruncated protein that is incorporated into membranes.

Each ␣2␦ subunit is encoded by a single gene and posttranslationally cleaved into a long, N-terminal, extracellular ␣2 protein and a shorter, membrane-anchored ␦ polypeptide; the ␣2 and ␦ proteins are covalently linked by disulfide bonds (9). Upon recombinant coexpression with ␣1, ␤, and ␥ subunits, ␣2␦ subunits modulate channel activity by increasing calcium current density, shifting the voltage dependence of activation to more negative potentials, or increasing steady-state inactivation (10,11). The ␣2␦Ϫ1 and ␣2␦Ϫ2 subunits possess high affinity binding sites for the antiepileptic drug gabapentin, which exerts an inhibitory effect on VGCC-mediated Ca 2ϩ currents (12,13). The spontaneous mouse mutants ducky and ducky 2J were recently shown to be functional Cacna2d2 null alleles encoding prematurely truncated ␣2 and no ␦2 protein (8). ducky homozygotes exhibit ataxia, absence epilepsy, and paroxysmal dyskinesia. Dysgenesis of brainstem and spinal cord as well as myelination defects and altered morphology of Purkinje cell dendrites are anatomical correlates of the phenotype (14,15).
Here, we provide a phenotypic, molecular, and functional characterization of a novel spontaneous Cacna2d2 mutant, entla (symbol used here: ent).

MATERIALS AND METHODS
Animals-Animals were maintained in a certified animal facility according to state legal guidelines and provided a commercial diet and water ad libitum. The entla phenotype was first observed in a heterogeneous genetic background (DBA/2J, CD1). Heterozygous animals were subsequently crossed into a C57BL/6J genetic background for 10 generations. For our studies, N2 to N10 animals were used. For tissue preparation, animals were anesthetized under CO 2 and decapitated, and the desired tissue was removed, shock-frozen in liquid nitrogen, and stored at Ϫ70°C until further use.
Nucleic Acid Preparation-DNA from tail tip biopsies was isolated according to standard methods. DNA from spleen or liver was isolated using the DNeasy Tissue Kit (Qiagen, Hilden, Germany) or by standard phenol extraction. Total RNA was isolated using the peqGold RNAPure system (PeqLab, Erlangen, Germany). cDNA was assembled by reverse transcription of 1-1.5 g of RNA using random hexamers or nonamers (Invitrogen).
Candidate Gene Analysis-The Cacna2d2 coding region was PCRamplified from cerebellar cDNA in four portions using the following primers (all primer sequences indicated 5Ј to 3Ј): Forward1, CGC CGC ATC TTG AAT GGA AAC; Reverse1, TGC CAC AAC AGT GTA GGG TCT TGC; Forward2, CTG CAG GAC AAC ATC AAG GAG; Reverse2, GTG TCA GGT TGA AAA CAG GGA GAG; Forward3, CCG CTC CAC ACA GGA ATA CC; Reverse3, CAT CCA CCT CAC TGA AGA ATC TGC; Forward4, TG TCA ACC AGA ACC ATC AGT GG; Reverse4, GCG TGT CTG TTT GTG TGT TCC ATC. PCRs were performed using the HiFidelity polymerase system (Roche Applied Science) with annealing temperatures of 58 -60°C. Additionally, the GC-rich 5Ј portion of the coding region was amplified using the Triple Master polymerase system for GC-rich targets (Eppendorf, Hamburg, Germany), the forward primer (GCG CCG CAT CTT GAA TGG), reverse primer (CGT GTG CTG CTG GGG GAA G), and an annealing temperature of 64°C. Amplification of the rearrangement site was performed with spleen DNA using the forward primers in intron 3 before the MfeI site at 6.6 kb (Fw1) (CCA GAT TTC AAG GAG ACA GAC AGT AAC C) or before the MfeI site at 24 kb (Fw2) (CAG GGA CCA ACA AAA CAA GAG TGT AG) with the reverse primer (GCG GGG CAT AAA AGC AGA GAT AGC) Triple Master polymerase system (Eppendorf, Hamburg, Germany) for long range PCR and an annealing temperature of 58°C. All fragments were cloned into the pCR2.1/TOPO vector (Invitrogen) and sequenced on an ABI Prism 377 sequencer (Applied Biosystems, Foster City, CA).
Genotyping and Mapping-entla homozygotes were identified at 2 weeks of age by their motor phenotype. Heterozygous and wild type animals were identified through breeding or genotyping using tail tip DNA and primers that amplify the chromosomal rearrangement site (forward primer, CAC ACC CCA GAC CAC ACT TTA TG; reverse primer, CAC TCT CCC ACC CCC ACA TTC). Using brain cDNA, ϩ/ϩ, ϩ/ent, and ent/ent animals were genotyped by amplification of Cacna2d2 from exon 1 to exon 7 using the forward primer (GGA GAT TGA CGG TGT GAT GCG) and the reverse primer (CAA ACA GGT TCC GAT TGT CCT TG). All genotyping PCRs were performed using HiFidelity polymerase mix (Roche Applied Science) and an annealing temperature of 57°C. For Southern hybridizations, 50 -100 g of DNA were digested overnight with the respective enzymes, separated on 0.3% agarose gels, transferred onto an uncharged nitrocellulose membrane, and cross-linked. Probes were [␣-32 P]dCTP-labeled using the Prime-It II Random Primer Labeling Kit (Invitrogen). Hybridizations and washes were carried out according to standard procedures. Signals were detected on a PhosphorImager (Storm 960; Amersham Biosciences). Mapping was performed using published primers and protocols (16) with tail tip DNA to detect DBA/2J-and C57BL/6J-specific simple sequence length polymorphisms.
Expression Vectors-entla Cacna2d2-cDNA was PCR-amplified as described above, joined by overlapping extension or ligation in pCR TOPO-TA (Invitrogen), and a KpnI and an XhoI site were introduced in front of the initial ATG codon and the TAA stop codon, respectively. Cacnb4 cDNA was amplified from wild type C57BL/6J animals using the forward primer AGG TAC CAT GTA TGA CAA TTT GTA CCT GC and the reverse primer ACT CGA GTC AAA GCC TAT GTC GGG A, likewise introducing a 5Ј KpnI and a 3Ј XhoI site. The constructs were cloned into pCDNA3.1V5HisTOPO (Invitrogen). Cacna2d2 wild type and Cacna1A constructs described by Hobom et al. (11) were used. Cacna2d2 constructs represent the splice variants lacking exon 23 as well as the first 6 nucleotides of exon 38.
Membrane Protein Preparations and Western Blotting-Membrane protein was prepared by repeated homogenization and centrifugation in hypotonic potassium phosphate buffer, and the protein content was determined according to the modified method of Lowry as described previously (17). Western blots were performed using polyclonal rabbit anti-␣2␦-2 antiserum (18) directed against the N-terminal epitope described by Marais et al. (12) with membrane protein preparations separated on 6% reducing or nonreducing polyacrylamide gels as described previously (19). Signals were detected on a Storm 960 fluoroimager (Amersham Biosciences).
[ 3 H]Gabapentin Binding Assays-Specific binding of [ 3 H]gabapentin (Amersham Biosciences) to membrane fractions was determined using 10 -30-g samples of membrane protein and 1 l of [ 3 H]gabapentin dilutions in a filtration assay using GF/C membranes (20). [ 3 H]gabapentin dilutions were incubated with membrane protein for 1 h at room temperature. Specific binding was determined by subtracting unspecific binding observed in the presence of 10 mM of unlabeled gabapentin from total binding. All samples were prepared in triplicate and subjected to liquid scintillation counting (Amersham Biosciences). Binding data were fitted to the following equation using Origin (Microcal, Northampton, MA), dpm spec ϭ B max ϭ [GBP]/([GBP] ϩ K D ), where dpm spec represents specific binding expressed as scintillator counts, B max is the total number of agonist binding sites, K D is the binding capacity, and [GBP] is the concentration of free [ 3 H]gabapentin.
Electroencephalographic Recording-Under general anesthesia (equithesin, 4 ml/kg intraperitoneally), mice (n ϭ 5) were stereotactically implanted with a bipolar electrode inserted into the hippocampus and two monopolar electrodes just touching the cortex. In addition, a monopolar surface electrode was placed over the cerebellum (reference electrode). The bipolar electrode was formed of two twisted enamelinsulated stainless steel wires (diameter, 170 m; distance between the tips, 0.4 mm) connected to a male connector aimed at the right dorsal hippocampus using the following coordinates (with bregma as reference): anteroposterior ϭ Ϫ1.5; mediolateral ϭ Ϫ1.8; dorsoventral ϭ Ϫ1.9 mm. The monopolar electrodes were made of the same enamelinsulated stainless steel wire (diameter, 250 m) soldered on a male connector (Wire pro, Farnell, France). They were inserted in the skull so that only the tip (0.5 mm) protruded onto the respective brain tissue. The electrodes were fixed to the skull with cyanoacrylate and dental acrylic cement. The mice were then allowed to recover from anesthesia before being placed in the EEG recording chamber. EEG activities of freely moving animals placed in a Faraday cage were recorded using a digital acquisition computer-based system (MP100WSW system; Biopac Systems Inc., Santa Barbara, CA; six channels, sampling rate of 200 Hz). Before starting the EEG recordings, a period of 1 h was allowed for the habituation of the animals to the test cage. Each animal was recorded for a total of 5 h during the resting phase of the animals (2 p.m. to 6 p.m.). Blocks of 1 h were analyzed with respect to seizure type, frequency, and duration.
Electrophysiological Recordings-Purkinje cells from P4 -P8 animals were acutely dissociated, and Ba 2ϩ currents were measured using modified protocols from (21). Briefly, the vermis region of the cerebellum was transferred into ice-cold dissociation solution (81.4 mM Na 2 SO 4 , 30 mM K 2 SO 4 , 5.8 mM MgCl 2 , 1 mM HEPES, 20.4 mM glucose, pH 7.4) and incubated while being oxygenated for 6 -8 min at 37°C in the same solution containing 3 mg/ml protease type XXIII (Sigma). After three washes in dissociation solution and one wash in trituration solution (modified Eagle's medium containing 1 mg/ml trypsin inhibitor (Sigma) and 1 mg/ml bovine serum albumin, pH 7.4), cells were dissociated by triturating 20 -30 times using a fire-polished Pasteur pipette. The cell suspension was transferred onto silanized polylysine-coated cover slips and kept on ice in trituration solution. Voltage clamp recordings were performed using an extracellular solution containing 85 mM NaCl, 20 mM triethanolamine-Cl, 10 mM BaCl 2 , 5 mM CsCl, 1 mM MgCl 2 , 5 mM HEPES, 10 mM glucose, pH 7.4, and in intracellular solution containing 120 mM CaCl 2 , 20 mM triethanolamine-Cl, 11 mM EGTA, 10 mM HEPES, pH 7.4 (11). Sodium currents were blocked by adding 300 nM tetrodotoxin to the extracellular solution. Whole cell voltage clamp recordings were carried out from a holding potential of Ϫ60 mV according to the following voltage step protocol: 30-ms steps from Ϫ60 mV to ϩ30 mV in 10-mV increments with a 10-s interval between voltage steps. Recordings were made at a temperature of 22°C using an EPC9 amplifier and the PULSE and PULSE-FIT software package (HEKA, Lambrecht, Germany).
Statistical Analyses-Limits of significance were determined using one-way analysis of variance. Errors indicate one S.D. value unless otherwise stated.

RESULTS
Phenotypic Characterization-The entla phenotype was originally observed in a heterogeneous DBA/2J/CD1 stock and follows an autosomal recessive Mendelian mode of inheritance. The mutation was crossed into a C57BL/6J genetic background for 10 generations, and no alteration of the phenotype was observed. In homozygous mutants, ataxic gait became apparent between postnatal days 13 and 15 and persisted throughout life. During motion, animals exhibited a characteristic outward pointing of the hind limbs (Fig. 1C). Overall posture of the animals was altered as well, either with a "hunchback" (Fig.  1C) or an arched back. About 2 days after the onset of ataxia, ent/ent animals started to experience multiple episodes of paroxysmal dyskinesia per day (Fig. 1, A and B), which lasted up to several minutes and were exacerbated by stress due to handling or noise. ent/ent animals were unable to remain on a rotarod but were able to swim even during episodes of paroxysmal dyskinesia. Homozygous animals were smaller than their littermates from postnatal day 10 on (Fig. 1, D and E). Lethality before weaning in homozygous animals was greater than 50%, but survivors had a normal lifespan (Fig. 1F). Whereas males were fertile albeit with a low breeding performance, homozygous females did not breed and displayed severely dysmorphic uteri with a drastically reduced diameter (Fig. 1G).
Genetic Mapping-While crossing the animals into the C57BL/6J background, we found the entla phenotype to be linked to gray coat color caused by the dilute allele of Myosin5a (22). DBA/2J mice, a contributing background in which the mutation was first observed, carry the dilute allele. Thus, this suggested a location on chromosome 9 near the Myosin5a locus. Genomic DNA samples from one N2 and 51 N5-N7 intercross entla mice were genotyped for microsatellite markers discriminating between C57BL/6J and DBA/2J at 50 -60.1 centimorgans from the centromere on chromosome 9 ( Fig. 2A). The mutation was mapped to the 1.87-megabase pair region between the markers D9Mit78 and D9Mit184 at 60.1 centimorgans with the VGCC subunit gene Cacna2d2 emerging as a likely candidate (Fig. 2B). Sequence Analysis of entla Cacna2d2 Transcript and Gene-Sequencing of the Cacna2d2 coding region cDNA revealed an exact duplication of 117 base pairs corresponding to exon 3. Using PCR primers covering the region encoded by exons 1-7, we obtained an amplification product of 598 bp from whole brain cDNA of wild type animals. Using the same primers, a 715-bp product was amplified from cDNA from ent/ent animals, corresponding to the wild type sequence plus the 117-bp duplication. Amplification of cDNA from heterozygous animals yielded both products (Fig. 2C). No apparent differences in expression level or any alternatively spliced products could be detected (data not shown). These observations characterize the entla gene as an allelic variant of Cacna2d2, implying that entla is allelic to ducky and ducky 2J . Based on the sequence duplication observed at the transcript level, it seemed plausible that the entla allele represented a genomic duplication encompassing exon 3 of the Cacna2d2 gene. Indeed, Southern hybrid-FIG. 1. Ataxia, paroxysmal dyskinesia, weight, survival and female infertility in entla animals. Homozygous ent animals are from N6 and N7 intercrosses aged 4 -6 months. A and B, animals during an episode of paroxysmal dyskinesia. C, characteristic outward pointing of the hind limbs and altered posture creating a "hunchback" appearance. D, size difference between 4-monthold ent/ent animal and wild type littermate. E, minimum, maximum, and average weights of ent/ent animals (E) as well as minimum and average weights of their unaffected littermates (f) from eight N4 -N7 intercross matings normalized to the heaviest littermate, which was never an entla homozygote. Weights were recorded in 68 animals (at day 10; entla (n ϭ 17) and healthy littermates (n ϭ 51)) aged 10 -30 days. ent/ent animals exhibited reduced weight from day 10 on, although they continuously gained weight. F, survival curve for 32 homozygous entla animals from N3-N5 intercrosses. The percentage of survivors among this group of animals is charted over a period of 50 days. In this sample of animals, 52% (17 of 32) died between the age of 17 and 37 days, and no animals died thereafter during the period analyzed. G, dystrophic uteri of 9-month-old entla females from an N9 intercross (center and bottom) and uterus of an unaffected female of the same age (top) that had never been pregnant.
entla Mouse Mutant ization analysis using a hybridization probe corresponding to the genomic region around exon 3 (probe B; Fig. 3C, nucleotide positions Ϫ322 to ϩ1383 relative to exon 3) revealed a 17-kb band in addition to the wild type 22-kb band in MfeI-digested ϩ/ent and ent/ent DNA (Fig. 3, A and D). In contrast, we saw no difference in banding pattern in ϩ/ϩ, ϩ/ent, and ent/ent MfeIrestricted DNA when using probes corresponding to the regions surrounding exon 2 (probe A, positions Ϫ1451 to ϩ258 bp relative to exon 2), exon 4 (probe D, positions Ϫ944 to ϩ920 bp relative to exon 4), or an intron 3 fragment (probe C, 1104 bp between the MfeI sites at positions 6.6 and 24 kb; data not shown). Likewise, no differences were observed using the probe B (exon 3) on EcoRI-digested DNA (data not shown). We thus concluded that the recombination locus in intron 2 resides between the MfeI site at 20 kb and the EcoRI site at 29.2 kb. Two possible recombination loci were left in intron 3: between the EcoRI site at 4.7 kb and the MfeI site at 6.6 kb or between the MfeI sites at 24 and 27.5 kb (Fig. 3C). Long range PCR using genomic DNA from ent/ent and ϩ/ϩ animals with forward primers either upstream of the MfeI sites at 6.6 kb (Fw1) or at 24 kb (Fw2) in intron 3 and a reverse primer downstream of the EcoRI site at 29 kb in intron 2 should therefore only yield amplification products in mutant DNA (Fig. 3, C and D). Indeed, we were able to amplify products of 21 kb using the Fw1 primer and of 4.6 kb using the Fw2 primer from ent/ent DNA but not from ϩ/ϩ DNA (Fig. 3B). Sequencing of the 4.6-kb amplification product revealed that it consisted of contiguous parts of intron 3 and intron 2 joined with a 3-base pair overlap (Fig. 3E). The breakpoints corresponded to positions at 27 kb in intron 2 (Ϫ12 kb relative to exon 3) and 26 kb in intron 3. A 38-kb genomic duplication between these breakpoints would thus have created a hybrid 3/2 intron consisting of the first 26 kb of intron 3 followed by the last 12 kb of intron 2 between two copies of exon 3 (Fig. 3D). Nonhomologous recombinations are often mediated by stretches of sequence homology (e.g. repetitive elements) (23). The regions in intron 2 and intron 3 immediately surrounding the recombination sites contained no apparent sequence similarities; however, sequence alignments of intron 2 and 3 revealed several short stretches of sequence homology, which were identified as partial B1 and B2 elements (Fig. 3F).
The ␣2␦-2 Subunit Is Altered in ent/ent Mice-The mutation was predicted to leave the translational frame unaffected and cause a duplication of 39 amino acid residues. Western blot analysis of membrane proteins separated under reducing conditions, using antibodies against an N-terminal epitope of ␣2, verified that mutant protein was expressed and incorporated into the plasma membrane at levels similar to wild type. The apparent molecular masses of 138 kDa for wild type and ϳ150 kDa for entla protein indicated that the ␣2␦-2 entla subunit undergoes correct post-translational cleavage into ␣2 entla and ␦ proteins. As expected, ␣2 entla possessed a slightly higher molecular mass (Fig. 4A, left panel), yet the observed 10 -12-kDa increase in molecular mass exceeded the 5 kDa calculated for the duplicated 39 amino acid residues. The additional increase of 5-7 kDa is likely to be due to glycosylation. The ␣2 protein is heavily glycosylated (12), with the 39 amino acid residues encoded by exon 3 containing several putative N-glycosylation sites. When membrane fractions containing wild type ␣2␦ subunits are separated under nonreducing conditions, the disulfide bonds linking ␣2 and ␦ remain unaffected, resulting in a shift to higher apparent molecular weight (12). Indeed, this was seen for wild type ␣2␦-2 (Fig. 4A, right panel). However, separation of entla membrane fractions under nonreducing conditions did not result in a corresponding shift. Instead, a band of the same apparent molecular weight as the ␣2 entla protein detected under reducing conditions was observed (Fig.  4A, right panel). It was therefore concluded that the ␣2 entla protein is not covalently linked to the ␦ protein via disulfide bonds.
Recombinant expression of Cacna2d2 entla cDNA in HEK293 cells showed that, when co-expressed with the ␣1A (Ca V 2.1) and ␤4 subunits as in typical P-type calcium channels, ␣2␦-2 entla immunoreactivity was localized in the plasma membrane (Fig. 4, E and J). Furthermore, the antibody recognized an epitope in nonpermeabilized cells (i.e. on the extracellular side) (Fig. 4J). Thus, localization and orientation of the entla subunit were indistinguishable from wild type (Fig. 4, C and H). Conversely, when ␣2␦-2 entla was expressed without other calcium channel subunits, immunoreactivity was localized more diffusely throughout the cell (Fig. 4D) with little staining in the plasma membrane (Fig. 4I), indicating that association with other calcium channel subunits causes ␣2␦-2 entla to efficiently move to the plasma membrane.
Overall central nervous system anatomy of entla homozygotes appeared normal. In marked contrast to ducky animals (14,15,24), no gross central nervous system abnormalities, no demyelination, and no abnormal Purkinje cell morphology could be detected (data not shown). Overall cerebellar morphology appeared unaltered, as evident from calbindin D28K immunohistochemistry (Fig. 5).
Spike Wave Discharges in Cortical and Hippocampal EEGs of entla Mice-EEG recordings from homozygous adult entla mice (n ϭ 5) revealed a pattern of 4 -5-Hz spike wave dis- FIG. 2. Genetic mapping of the entla locus and Cacna2d2 transcript analysis. A, genotypes of 51 N5-N7 and one N2 intercross entla mice. The number of animals per genotype is indicated at the bottom. For clarity, D9Mit is omitted from marker names. Light gray squares indicate DBA/2J homozygous loci; dark squares indicate C57BL/6J/DBA/2J heterozygous loci. No C57BL/6J homozygous loci were detected in the area examined. The entla locus was localized between the microsatellite markers D9Mit78 and D9Mit184. B, localization of the entla locus in the central region of mouse chromosome 9 (Chr. 9). C, Cacna2d2 transcript from exon 1 to 7 was amplified from whole brain cDNA. One 598-bp product corresponding to the wild type transcript containing one copy of exon 3 was amplified from cDNA of ϩ/ϩ animals (left lane). Using cDNA from ϩ/ent animals, the 598-bp wild type product as well as a product of 715 bp, corresponding to the mutant transcript containing two copies of exon 3, was amplified. The two amplification products displayed a similar band intensity, suggesting no drastic expression level variations (center lane). Only the 715-bp mutant transcript was amplified from ent/ent cDNA (right lane). charges ( Fig. 6A) with an average frequency of 4.6 Ϯ 1.0 Hz (n ϭ 10 frequency counts) and each seizure lasting 1-4 s ( Table  I). This type of seizure activity typically was observed synchronously in recordings taken from the intrahippocampal as well as from the cortical electrodes indicating a generalized seizure activity. Similar numbers and durations were determined for the cortical and hippocampal seizures (13.   1, left panel), as reported previously (12). Under nonreducing conditions, the ␣2 WT and ␦2 proteins remain connected via disulfide bonds, resulting in a 190-kDa band (lane 1, right panel), whereas the ␣2 entla protein is not connected to the ␦2 subunit, thus resulting in a band representing ␣2 entla protein alone (lane 3, right panel) and identical in apparent molecular mass to the band obtained under reducing conditions. Both wild type and mutant bands are seen in membrane protein from heterozyous animals (lanes 2, left and right panels). B-K, immunocytochemical stainings of HEK293 cells expressing recombinant VGCC subunits: ␣2␦-2 WT (B and G); ␣2␦-2 WT , ␣1A (Ca V 2.1), and ␤4 (C and H); ␣2␦-2 entla (D and I); ␣2␦-2 entla , ␣1A (Ca V 2.1), and ␤4 (E and J); and ␣1A (Ca V 2.1) and ␤4 (F and K). Cells were stained with a polyclonal antibody recognizing the ␣2 protein. Cells in B-F were permeabilized with Triton X-100; cells in G-K were not, allowing staining of surface antigens only. When co-expressed with other calcium channel subunits, ␣2␦-2 entla localizes to the plasma membrane with the epitope being extracellular, in a manner indistinguishable from wild type. Expression of ␣2␦-2 entla alone, however, results in less efficient membrane incorporation than in the wild type control. Virtually no immunosignals were obtained in cells expressing only ␣1A (Ca V 2.1) and ␤4 (G) or in untransfected cells (not shown). Scale bar, 10 m. (Table I)). Seizures were preceded and followed by normal hippocampal and cortical activity and did not coincide with paroxysmal dyskinesia. The spike wave discharges were accompanied by behavioral arrest of the animal and occasionally by staring and slight head nodding. In addition to the seizure type described above, in 4 of 5 animals, a second, longer lasting seizure type with a different frequency of 2.0 Ϯ 0.35 Hz (n ϭ 10 frequency counts) was observed (Fig. 6B). 10.0 Ϯ 6.5 of these seizures, each lasting 33.6 Ϯ 31.7 s, were counted per hour in cortical recordings, whereas hippocampal recordings revealed similar values of 13.0 Ϯ 9.6 seizures per hour, each lasting 31.0 Ϯ 23.9 s. These seizures were likewise accompanied by behavioral arrest of the animal, staring, and slight head nodding. No such abnormal activity could be recorded from unaffected control animals (data not shown). These types of seizures are typical for human and murine absence epilepsy (3,25,26).

s, respectively
Reduced Current in entla Cerebellar Purkinje Cells-Ba 2ϩ currents in Purkinje cells from the cerebellar vermis region of ϩ/ϩ, ϩ/ent, and ent/ent animals aged 4 -6 days were recorded in the presence of 300 nM tetrodotoxin (Fig. 7). The animals' genotypes were inferred from whole brain cDNA, thereby also demonstrating that the ␣2␦-2-subunit is expressed in these animals although their phenotypes are not yet detectable. We found a reduction of ϳ50% in Ba 2ϩ current density in entla homozygotes (p Ͻ 0.05 for ent/ent versus ϩ/ent and ent/ent versus ϩ/ϩ), whereas current densities in Purkinje cells of heterozygous animals did not differ significantly from wild type (p Ͼ 0.5). Maximum current densities at Ϫ10 mV for the three genotypes were as follows: 15.1 Ϯ 1.1 pA/pF (ent/ent; n ϭ 12); 30.5 Ϯ 2.3 pA/pF (ϩ/ent; n ϭ 12); 32.2 Ϯ 2.7 pA/pF (ϩ/ϩ; n ϭ 12). Voltage dependence of activation between the three genotypes was found to be unaltered. Purkinje cell size represented by cell capacitance did not differ significantly between the three genotypes (ϩ/ϩ, 23.5 Ϯ 2.8 pF; ϩ/ent, 22.1 Ϯ 1.5 pF; ent/ent, 26.7 Ϯ 2.8 pF; p Ͼ 0.5).
[ 3 H]gabapentin binding to ϩ/ent cerebellar membranes did not differ significantly from wild type (B max for ent/ent, 9.2 Ϯ 1.8 fmol/mg; B max for ϩ/ϩ, 27.1 Ϯ 1.5 fmol/mg; B max for ϩ/ent, 27.6 Ϯ 4.4 fmol/mg; p Ͻ 0.0005 for ent/ent versus ϩ/ent and ϩ/ϩ; p Ͼ 0.5 for ϩ/ent versus ϩ/ϩ). The K D values were similar to each other and to those reported for recombinantly expressed ␣2␦-2 (ϩ/ϩ, 87 Ϯ 9 nM; ϩ/ent, 122 Ϯ 33 nM; ent/ent, 127 Ϯ 47 nM) (Table II), indicating that entla membranes possess no drastically altered gabapentin affinity. DISCUSSION In this study, we identified the novel neurological mouse phenotype entla, and characterized the underlying mutation as an allele of Cacna2d2, the gene encoding the ␣2␦-2 subunit of   VGCCs. The entla phenotype involving ataxia, paroxysmal dyskinesia, and absence seizures resembles that of other VGCC mouse mutants, in particular the allelic Cacna2d2 mutants ducky and ducky 2J . In contrast to these, however, no gross anatomical abnormalities of the central nervous system could be detected in ent/ent mice. entla females are infertile and have severely dysmorphic uteri. Among the VGCC mutants, selectively compromised female fertility represents a unique feature of ␣2␦-2 defects. Unlike findings reported for other VGCC mouse mutants, two distinct types of generalized spike wave seizures were detected in entla. The seizures differ with respect to duration and frequency, with a shorter (ϳ2-s) type at a frequency of 4 -5 Hz and much longer (ϳ30-s) type at a slower frequency of 2 Hz. The study of the circuits, receptors, and channels involved in these two types of seizures might further our understanding into the physiology of absence epilepsy.
The entla allele of Cacna2d2 carries a hybrid intron separating two identical copies of exon 3. The mutant transcript is expressed at normal levels with no indication of alternative splicing of either one of the two exon 3 copies, indicating that all sequence elements needed for wild type-like splicing are present in the hybrid intron. Genomic duplications are typically mediated by nonhomologous recombination along stretches of sequence homology (23), ranging from 1 to 100 kb (28). Whereas the entla duplication is within these limits, no sequence homology was found in the immediate vicinity of the entla recombination site. However, partial B1 elements of the same relative orientation were found 2 and 3 kb upstream of the recombination sites within introns 2 and 3 (Fig. 3F), consistent with a nonhomologous recombination inducing a downstream breakpoint. Interestingly, the ducky allele also carries a breakpoint within intron 3, suggesting the presence of a potential recombination hot spot within the murine Cacna2d2 gene. Nonhomologous recombinations resulting in exon duplications have been frequent events throughout evolution, accounting for up to 10% of all exons in humans, Drosophila, and Caenorhabditis elegans (29). Disease-associated exon duplications, however, are rare. To our knowledge, entla is the only mouse mutant presently identified carrying a single duplicated exon.
The Cacna2d2 entla allele leads to the synthesis and posttranslational processing of a full-length protein with a 39amino acid residue duplication near the N terminus of the ␣2 protein. ␣2 immunosignals of similar intensities could be detected in wild type and entla, indicating no severe reduction in protein stability. Western blot data were further consistent with a correct cleavage of the ␣2␦-2 precursor into ␣2 entla and ␦ proteins but indicated a lack of disulfide bond formation between the two. This might be explained by alterations in the tertiary structure of ␣2 entla interfering with efficient formation of disulfide bonds at the appropriate positions. When coexpressed with the calcium channel subunits ␣1A (Ca V 2.1) and ␤4 in HEK293 cells (but not when expressed alone), ␣2␦-2 entla localizes to the plasma membrane, in an orientation indistinguishable from ␣2␦-2 WT . The ␣2␦-2 entla subunit thus does appear to be incorporated into calcium channels. The apparent reduction in maximum [ 3 H]gabapentin binding in ent/ent cerebellar membranes is consistent with the presence of two binding sites in the wild type (30), namely the ␣2␦-1 and ␣2␦-2 subunits. In the ␣2␦-1 subunit, the amino acid residues corresponding to those duplicated in the ␣2␦-2 entla subunit do not directly contribute to [ 3 H]gabapentin binding; nor do the ␣2 or ␦ proteins by themselves bind gabapentin (31,32). Unfortunately, little is known about the ligand binding domains of the ␣2␦-2 subunit. By analogy to ␣2␦-1, however, if ␣2 entla would not interact correctly with ␦, [ 3 H]gabapentin binding might be lost. This would imply that the residual [ 3 H]gabapentin bind-ing measured in ent/ent membranes might be attributable to the ␣2␦-1 subunit. We observed no compensatory up-regulation of Cacna2d1 transcript or the ␣2␦-1 subunit in homozygous entla mice (data not shown), corroborating the observation that, within the concentration range tested, the ␣2␦-2 subunit contributes substantially to overall [ 3 H]gabapentin binding in wild type cerebellum. Given the small difference in [ 3 H]gabapentin affinities between the ␣2␦-1 and ␣2␦-2 subunits in recombinant expression systems and lack of data about [ 3 H]gabapentin affinities of these subunits when coexpressed with ␣1 and ␤ subunits, correlation of the small difference in affinity between wild type and entla membranes to relative contributions of ␣2␦-1 and ␣2␦-2 subunits to [ 3 H]gabapentin binding in vivo must await further studies.
Gabapentin is used for monotherapy or adjunctive treatment of partial complex and generalized tonic-clonic seizures. Its efficacy in treating absence epilepsy could not be demonstrated (33). Gabapentin reduces calcium current densities (13) and thus has an effect similar to that of the entla mutation itself. Incidentally, it has been reported that in some patients, gabapentin therapy can in fact precipitate or aggravate absence seizures. Additionally, ataxia is one of the known side effects of gabapentin treatment (34). One might also speculate that ␣2␦ subunit polymorphisms could be a determining factor for patient response to gabapentin treatment.
In order to study a direct physiological consequence of the entla mutation, voltage-gated calcium currents were investigated in neurons from ent/ent mice and healthy littermates. Since ataxia is a hallmark of cerebellar dysfunction, cerebellar Purkinje cells, which are known to express the ␣2␦-2 subunit (8,11), were chosen as model cells. Ba 2ϩ current densities recorded from cells of ent/ent mice were significantly decreased, suggesting a loss-of-function effect of the mutation. Again, this is compatible with both a direct interference of the ␣2 entla protein with normal channel function and with impaired VGCC assembly. Altered excitability of cerebellar neurons in entla might be a direct electrophysiological correlative of ataxia.
Whereas voltage-dependent Ca 2ϩ influx into Purkinje cells represents an isolated property of a single cell type, ataxia, paroxysmal dyskinesia, and epilepsy require dysregulated complex neuronal circuits (35). Neuronal activity of thalamic T-type Ca 2ϩ channels is increased in the VGCC mouse mutants tottering (␣1A/Ca V 2.1), lethargic (␤4), and stargazer (␥2), although the respective subunits are not found in T-type Ca 2ϩ channels, and no numeric increase in T-type channel subunits was observed (36). For further studies, it might be of interest how those circuits are affected in ␣2␦-2 subunit mutants.
Murine mutations in the ␣1A (Ca V 2.1), ␤4, and ␣2␦-2 (ducky, entla) subunits result in similar phenotypes characterized by ataxia, paroxysmal dyskinesia, and spike wave seizures. P-type calcium currents were shown to be affected in ␣1A (Ca V 2.1) and ␤4 mutant mice (37,38). Electrophysiological recordings from ducky and entla Purkinje cells are consistent with reductions in P/Q-type calcium currents, and the ␣2␦-2 subunit has been shown to affect P-type calcium currents in recombinant systems (11). These data are compatible with P-type VGCCs comprising the ␣1A (Ca V 2.1) subunit along with ␤4 and ␣2␦-2 auxiliary subunits.
Given the efficient incorporation of full-length ␣2 entla into membranes, one might expect entla VGCCs to retain a higher functionality than VGCCs in the functional knockout mutants ducky and ducky 2J . Homozygous ent/ent mice are indeed affected less severely; they show a lower rate of lethality than ducky or ducky 2J (8,39), and, in contrast to ducky homozygotes, they exhibit neither gross anatomical abnormalities (14,24) nor detectable alterations of Purkinje cell morphology (15). Therefore, entla is a valuable model for the investigation of ␣2␦-2 dysfunction in isolation, without interference caused by neuroanatomical alterations.
In humans, no mutations in the ␣2␦-2 subunit are known as yet, although one coding and several noncoding polymorphisms have been identified. Mutations of other VGCC subunits, however, lead to cerebellar ataxia, generalized and absence epilepsy, clonic seizures, and psychiatric symptoms as well as hemiplegic migraine (40 -46). Mutations in the human ␤4 subunit are associated with generalized tonic-clonic seizures, absence epilepsy, or juvenile myoclonic epilepsy (47). The human symptoms thus resemble the mouse phenotypes. Thus, VGCC mouse mutants, including entla, might serve as a valuable model for human neurological disease.