Ataxin-10, the Spinocerebellar Ataxia Type 10 Neurodegenerative Disorder Protein, Is Essential for Survival of Cerebellar Neurons*

Spinocerebellar ataxia (SCA) type 10, an autosomal dominant disease characterized by cerebellar ataxia, is caused by a novel pentanucleotide (ATTCT) repeat expansion in the SCA10 gene. Although clinical features of the disease are well characterized, nothing is known so far about the affected SCA10 gene product, ataxin-10 (Atx-10). We have cloned the rat SCA10 gene and expressed the corresponding protein in HEK293 cells. Atx-10 has an apparent molecular mass of ∼55 kDa and belongs to the family of armadillo repeat proteins. In solution, it tends to form homotrimeric complexes, which associate via a tip-to-tip contact with the concave sides of the molecules facing each other. Atx-10 immunostaining of mouse and human brain sections revealed a predominantly cytoplasmic and perinuclear localization with a clear restriction to olivocerebellar regions. Knock down of SCA10 in primary neuronal cells by small interfering RNAs resulted in an increased apoptosis of cerebellar neurons, arguing for a loss-of-function phenotype in SCA10 patients.

A number of human genetic diseases are associated with the expansion of short tandem repeats in coding or noncoding gene regions (1,2). Spinocerebellar ataxias and related genetic disorders are often caused by trinucleotide expansions of coding sequences conferring either loss-of-function as in Friedreich's ataxia (3) or gain/change-of-function by the translation of CAG repeats into extended polyglutamine tracts as in SCA1, 1 SCA2, SCA3, SCA6, SCA7, and SCA17 (4 -6). In both cases, the respective mutations lead to progressive cell degeneration in a region-specific manner (7). The cellular functions of the affected gene products are largely unknown and may greatly vary for each mutation type (8 -11), thus hampering therapeutic strategies. Spinocerebellar ataxia type 10 has been associated with an expansion of a pentanucleotide repeat (ATTCT) in intron 9 of the SCA10 gene sequence to as many as 4500 copies (22.5 kb), depending on the age of onset (12). It belongs to the group of neuronal diseases defined as "autosomal dominant cerebellar ataxias" (13). The clinical features of SCA10 involve progressing cerebellar dysfunctions combined with motor seizures and anticipation (14,15). The SCA10 gene is widely expressed in the mouse brain both in adult and juvenile animals, suggesting a basic neuronal function (12). Due to a lack of patient material, though, there has hitherto been no clear evidence of which effect the SCA10 mutation has on cellular functions and whether it leads to a loss or gain of function of the respective gene product.
In our study, we present the first biochemical data on the Atx-10 protein, which we expressed recombinantly in a mammalian cell line. Furthermore, we demonstrate the expression pattern of Atx-10 in human and mouse brain sections and show its subcellular localization in PC12 cells and primary neuronal cells. Homology searches and secondary structure analysis clearly define Atx-10 as member of the armadillo repeat protein family, suggesting a function in protein-protein interactions. siRNA experiments performed with different neuronal cell populations indicate an essential role of Atx-10 in the survival of cerebellar neurons and point to a possible mechanism of pathogenesis.

EXPERIMENTAL PROCEDURES
Construction of a PC12 cDNA Library and Isolation of the SCA10 cDNA-Polyadenylated mRNA was extracted from PC12 cells with the purification of the mRNA being performed using the QuickPrepR mRNA purification kit (Amersham Biosciences). With 6 g of poly(A) ϩ RNA as template and oligo(dT) [12][13][14][15][16][17][18] as primer, cDNA was synthesized following the instructions of the cDNA synthesis kit from Amersham Biosciences. EcoRI/NotI adaptors were ligated to each end of the bluntended cDNA, phosphorylated, and ligated into EcoRI-cleaved ZAPII vector (predigested Lambda ZAPRII/ECORI/CIAP cloning kit; Stratagene). Ligations were packaged in vitro with the GigapackRIII Gold packaging extracts from Stratagene. This cDNA library was screened during a serial sequencing project, and 13 clones were isolated. With these clones, a BLAST search was performed. One of the clones showed a sequence identity to the mouse E46 gene of ϳ95%. Because of this high sequence identity, we designated the new clone rat SCA10. The SCA10 clone we isolated from the library was not complete. To obtain the 5Ј region of the Atx-10 cDNA, a reverse transcriptase-PCR was performed according to the manufacturer's instructions (Stratagene Europe). Primers were designed in analogy to the 5Ј-untranslated region of the highly identical E46 mouse sequence. The following primers were used: RT-1-FOR (5Ј-GCT CTA GAG CCT GAG GGA AGC CAG CTA GTC TCG C-3Ј) and, as reverse primer, SCA10-Rev 2 (5Ј-TGG GAA GGC GTG GAC CCA-3Ј). The 5Ј region we obtained was restricted with PvuII and ligated with the PvuII-and NotI-digested 3Ј region of the clone into the pBluescript II KS (Stratagene Europe) cloning vector. Sequencing and sequence analysis sequencing were performed by Sequence Laboratories (Göttingen, Germany).
cDNA Constructs, Transfection Procedures, and Recombinant Expression-SCA10 cDNA was amplified by PCR using the SCA10 pBluescript vector as a template. NheI and BamHI sites were introduced in the 5Ј-and 3Ј-primers, respectively, to enable convenient cloning of the amplified DNA into the corresponding sites of the eukaryotic expression vector pCEP-Pu. Primers used were 5Ј-TTT AAG CTT GCT AGC CAT GGC GGC GCC CAG GAT G-3Ј and 5Ј-CTT GGA TCC TTA AGG CGG GGG GAT-3Ј. The resulting vector that introduces a polyhistidine tag, a Myc tag, and an enterokinase cleavage site at the N-terminal end of the protein sequence was used for transfecting HEK293 cells. For stable transfection, HEK293 cells were kept in Dulbecco's modified Eagle's medium/F-12 supplemented with 10% fetal bovine serum, 1% Gln, and penicillin/streptomycin. Cells were grown to 80% confluence in 6-well plates and transfected overnight with 1 g of vector DNA using 5 l of LipofectAMINE reagent. The selection of positive clones was performed by culturing transfected cells with 2 g/ml puromycin with frequent changes of medium until a resistant population appeared. All reagents were purchased from Invitrogen. HEK293 cells stably transfected with SCA10 cDNA were grown to high density in 125-ml cell culture flasks using complete medium. For expression of recombinant protein, cells were switched to serum-free expression medium. Cell supernatants were harvested frequently until the cells detached, pooled, and passed through a syringe filter. Conditioned medium was then purified using nickel-Sepharose chromatography according to the manufacturer's instructions (Qiagen). Eluted fractions were analyzed by SDS-PAGE and dialyzed against 20 mM Tris-HCl, pH 7.4, and 150 mM NaCl.
The P4 fragment comprising amino acids 228 -476 was expressed in Escherichia coli BL21 (DE3) pLysS cells (Stratagene Europe). The corresponding cDNA fragment was produced by PCR with 5Ј-GCG GCC ATG GCG CCG GAA CTG GTG GAA GCT-3Ј as forward and 5Ј-GGC CCA AGC TTA AGG CGG GGG GAT GTC ATT-3Ј as reverse primers and introduced into the pRSET 5d vector using NcoI and HindIII restriction sites at the 5Ј and 3Ј ends, respectively. Protein synthesis was induced at an A 600 of 0.5 with 0.4 M isopropyl-1-thio-␤-D-galactopyranoside for 2 h. Cultures were centrifuged, and pellets were resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1% Tween 20, 1 mM dithiothreitol). P4 inclusion bodies were then prepared by several freeze-thaw and sonication cycles followed by extensive washing with lysis buffer. For refolding, inclusion bodies were solubilized in 6 M GuHCl, 100 mM dithiothreitol, 50 mM Tris/HCl buffer at pH 8.0 and then submitted to dialysis against refolding buffer containing 20 mM phosphate and 200 mM NaCl, pH 6.8. The refolded protein was purified on a Superdex 75 (26/60) column (Amersham Biosciences) using refolding buffer as eluent. Fractions were analyzed by SDS-PAGE, pooled, and concentrated.
Antibody Production-An antiserum against Atx-10 was obtained by immunization of rabbits with P4 inclusion bodies at Bioscience (Göttingen, Germany) according to a standard protocol.
SDS-PAGE and Western Blot Analysis-For resolving Atx-10 oligomers by SDS-PAGE, gradient gels (3-10%) were applied with 2.5% stacker gels to ensure that oligomeric complexes were retained. Samples were not heated in this case to prevent oligomer dissociation. Monomeric Atx-10 was resolved by using 12% gels. Western blot analysis was performed using Atx-10 polyclonal antibody (rabbit) generated against P4. Primary antibody (1:1000) was detected using an antirabbit horseradish peroxidase-conjugated antibody (1:2000) and the ECL system (Amersham Biosciences).
Analytical Ultracentrifugation-A Beckman model XLA analytical ultracentrifuge equipped with absorption optics was employed. Sedimentation velocity runs were performed at 54,000 rpm for P4 and 52000 rpm for Atx-10, respectively. Sedimentation equilibrium runs were performed at 20,000 rpm for P4 at a filling height of 3 mm and 1400 rpm for Atx-10 at a filling height of 2 mm, respectively. All measurements were performed at 20°C. The molecular masses were calculated from sedimentation equilibrium runs using a floating base-line computer program that adjusts the base-line absorption to obtain the best linear fit of ln A versus r 2 (where A represents absorbance and r is distance from the rotor axis). A partial specific volume of 0.73 cm 3 /g for Atx-10 and P4 was used for the calculation, and the sedimentation velocity coefficient was corrected to standard conditions (H 2 O at 20°C).
Electron Microscopy-For rotary shadowing, purified recombinant Atx-10 was sprayed onto freshly cleaved silica, shadowed at an angle of 9°, and visualized after carbon coating and replication (25) by a Phillips 400 transmission electron microscope.
CD Spectroscopy-The CD spectrum of recombinant Atx-10 in phosphate-buffered saline (PBS) was acquired on a Jasco J720 spectropolarimeter (Japan Scientific Co.). The far-UV spectrum (200 -250 nm) was measured in a 1-mm path length quartz cell and represents an average of four accumulations. The spectrum was normalized for concentration and path length to obtain the mean molar residue ellipticity after subtraction of the buffer contribution. Helix content was calculated using the k2d program (23).
Cell Fractionation-The subcellular localization of Atx-10 in PC12 cells was analyzed after lysis of 3 ϫ 10 7 cells by passing them several times through a 0.6-mm syringe needle. Cytosolic and organelle fractions were separated by centrifugation in a tabletop centrifuge at 10,000 ϫ g for 10 min. A further separation into microsomal and cytosolic fractions was obtained by ultracentrifugation of the soluble fraction at 100,000 ϫ g for 1 h. Both fractions (200 l) were then treated with 4 l of proteinase K (1 mg/ml) in the presence or absence of 0.1% Triton X-100 for 30 min on ice. Reactions were stopped by the addition of Laemmli buffer and subsequent heat denaturation. Samples were analyzed by Western blotting using anti-Atx-10 antibody.
Primary Neuronal and Glial Cultures-Primary cerebellar and cortical neurons were prepared from postnatal Wistar rats (P1). Briefly, cerebelli and cortices were dissected, and the meninges were carefully removed. The tissue was lightly triturated with a sterile Pasteur pipette followed by trituration with a fire-polished Pasteur pipette (opening reduced to 50 -70% of its normal diameter). After 2-3 min for the segments of larger debris to settle, the supernatant containing single cells was transferred to a fresh tube; trituration procedures were repeated until no visible lumps of tissue remained. The single cell suspension was centrifuged at 200 ϫ g for 5 min. The cell pellet was resuspended in neurobasal medium, and cells were counted. The cells were diluted with appropriate culture medium and seeded into poly-Dlysine-coated 24-well plates or 8-well chamber slides. Cerebellar and cortical neurons were plated in minimum essential medium supplemented with glucose (final 25 mM), 0.5 mM L-glutamine, 0.23 mM sodium pyruvate, 1% penicillin-streptomycin sulfate, and 5% fetal calf serum. The KCl concentration was adjusted to 25 mM from the standard 5.4 mM present within minimum essential medium or neurobasal medium. The neurons were cultured at a density of 7 ϫ 105 cells/cm 2 at 37°C in an atmosphere of 5% CO 2 and 100% humidity. Three days after plating, cytosine arabinofuranoside (10 M; Sigma) was added to cultures to arrest growth of nonneuronal cells. One-half of the medium was removed every 3 days and replaced with fresh serum-free medium (neurobasal medium with 2% B-27 supplement, 0.5 mM L-glutamine, 25 mM K ϩ final, and 1% penicillin/streptomycin sulfate) to adjust to the original volume. The cells were used for experiments after 7 days in culture.
Cultures of newborn rat brain astrocytes were prepared as described previously (26). Single cell suspensions of 1-day-old rat cortex and cerebellum were prepared by enzymatic dissociation with 1% trypsin and final trituration with 0.05% DNase I. Precursor astrocytes were cultured for 5 days on poly-D-lysine-coated tissue culture flasks (250 ml) in serum-free medium (Eagle's basal medium with Earle's salts supplemented with 1% penicillin-streptomycin sulfate, 1 mg/ml bovine serum albumin, 10 g/ml insulin, 100 g/ml human transferrin, and 10 Ϫ8 M epidermal growth factor) at 37°C and 7.5% CO 2 atmosphere. Cells were subcultured until maturation in serum-containing medium (Eagle's basal medium, 10% fetal calf serum). They were then trypsinized with 0.1% trypsin and plated (5 ϫ 10 5 cells/well) in 35-mm dishes or glass coverslips coated with poly-D-lysine. Purity of the cultured astrocytes (97%) was examined by immunocytochemistry using antibodies directed against glial fibrillary acidic protein (Roche Applied Science).
RNA Isolation and Northern Blotting-RNA from PC12 cells was isolated and prepared by the method of Chomczynski and Sacchi via guanidine-phenol extraction (24). RNA concentrations were measured, and 10 g/lane were fractionated on a 1.2% agarose gel that contained 20% formaldehyde. The gel was subsequently blotted overnight on a Genescreen Plus membrane, and RNA was cross-linked to the membrane via baking at 80°C for 2 h. Hybridization took place at 68°C overnight in 120 mM NaCl, 10% dextran sulfate, 10% SDS with a 32 P-labeled NotI cDNA fragment of the clone. This fragment was prepared using the Random Primed Labeling Kit (Roche Applied Science). Membranes were washed at 60°C with 2 ϫ SCC, 1% SDS several times and subjected to autoradiography on BIOMAX x-ray films (Eastman Kodak Co.).
Immunohistochemistry-Two mouse brains and two human brains without histopathological changes were analyzed. Brain tissues were from mice transcardially perfused with 4% paraformaldehyde (in saline/sodium phosphate buffer, pH 7.4) or from human brain immersionfixed (1 week) in phosphate-buffered 4% formaldehyde. Immunohistochemistry was performed on 4-m paraffin sections from brain, cerebellum, brainstem, and spinal cord. After blocking nonspecific sites, sections were incubated overnight with the primary antibody (dilution 1:1000) at 4°C. Bound antibody was visualized with the avidin-biotin peroxidase method (Elite Standard kit SK6100; Vector Laboratories, Burlingame, CA). Peroxidase activity was revealed with a substrate solution containing 3-amino-9-ethylcarbazol (AEC kit SK 4200; Vector Laboratories). Antigenicity was enhanced by treating paraffin sections with microwave heating at 90°C for 30 min before incubation with Atx-10 antibody. All sections were counterstained with hematoxylin. For control stainings, the antibody solution was preincubated with recombinant Atx-10 at 1 mg/ml.
Neuronal survival was assayed 48 h after transfection by measuring lactate dehydrogenase release into culture supernatant using standard procedures.
Immunohistochemistry-Primary cultures of cerebellar and cortical neurons were plated on poly-D-lysine-coated 8-well CultureSlides (BD Biosciences, Belgium); PC12 cells and primary cerebellar astrocytes were plated on poly-D-lysine-coated tissue culture dishes with four compartments. Cells were fixed in 2% p-formaldehyde, 0.1% glutaraldehyde, 0.1% tannic acid for 15 min at room temperature, washed in PBS, and permeabilized in PBS plus 0.2% Triton X-100 for 5 min at room temperature, followed by three washes with PBS. Primary polyclonal antibody (rabbit) against Atx-10 was applied at 1:1000 in PBS overnight at 4°C. After washing with PBS, a 1:800 dilution of Alexa Fluor® 488 goat anti-rabbit IgG (Molecular Probes, Leiden, The Netherlands) was applied for 1 h at room temperature. Finally, cells were washed three times with PBS before being mounted with coverslips using Mowiol 4 -88 (Sigma) containing 2.5% 1,4-diazabicyclo-(2.2.2) octane (Sigma) to reduce fading. Fluorescence microscopy was performed using a DP-50 digital camera coupled to an Olympus IX51 inverted microscope.  in several matches with the mouse SCA10 gene (originally named E46 mouse brain protein, accession number BC046802) showing the highest homology. The identity between the murine sequence and the rat SCA10 cDNA is 94%. The homologous sequence of human origin (accession number BC007508) shares 86% identity with the rat gene. Related sequences identified in the Drosophila and Saccharomyces genomes show identities ranging from 24 to 35%. We speculate that the translation starts at position 201, since in the mouse E46 sequence, the translation starts at the same position. The open reading frame thus codes for altogether 476 amino acids. The nucleotide and putative amino acid sequences are shown in Fig. 1. Atx-10 does not contain a signal sequence for secretion or any subcellular compartment arguing for a cytoplasmic localization of the protein.  C, E, and G) and human  (B, D, F, and H) brain. A and B, olivary nucleus. Inset in A, cytoplasmic Atx-10 reactivity is better appreciated at higher magnification. Note the absence of nuclear staining; C and D, nuclei pontis; E and F, cerebellar cortex with Purkinje cells and upper part of the granule cell layer. Note the ring-shaped perinuclear Atx-10 reactivity in Purkinje cells and intensely stained small neuropil granules (F). Inset in F, control staining of mouse cerebellar cortex, in which antibody reactivity was completely abolished by preabsorption with recombinant Atx-10; G and H, dentate gyrus of the hippocampus. Note the almost total absence of Atx-10 staining in this structure. Paraffin sections were all counterstained with hematoxylin. A and B, ϫ 200; C-H, ϫ 400; A (inset), ϫ 800; F (inset), ϫ 400. ment containing amino acids 228 -476 (P4) was expressed in E. coli and refolded from purified inclusion bodies. Recombinant full-length Atx-10 showed an apparent molecular mass of ϳ58 kDa in SDS-PAGE and Western blots using anti-Atx-10 antibody (Fig. 2A). The difference in the molecular mass of the recombinant protein from the calculated value of 53.7 kDa corresponds to the added tag sequences, which code for an additional 5.5 kDa. The endogenous Atx-10 protein in PC12 cell lysates showed the expected molecular mass ( Fig. 2A), which was also observed for recombinant Atx-10 protein lacking a secretion signal sequence and tag sequences overexpressed in COS-7 cells (data not shown). The recombinant protein expressed in HEK293 cells exhibits a slight heterogeneity (triple band), probably due to a different behavior in SDS-PAGE caused by the added tag sequences. During purification of the P4 fragment via size exclusion chromatography, we have observed, beside the monomer peak, a pronounced second elution peak in a higher molecular mass region indicative of a distinct oligomer state (data not shown). Analytical ultracentrifugation experiments confirmed that both P4 and recombinant Atx-10 existed predominantly as oligomers with molecular masses suggesting a trimer formation (Table I). A similar result was obtained by gradient SDS-PAGE applying unheated Atx-10 samples in which the oligomeric band (ϳ150 kDa) was retained during electrophoresis (Fig. 2B). Electron micrographs of rotary-shadowed Atx-10 showed horseshoe-shaped complexes with sometimes three curved arms, suggesting a tip-to-tip association of the molecules with the concave sides facing each other (Fig. 2C). Since the P4 fragment has apparently retained the capacity for forming oligomers, we suppose that the homoassociation site is situated at the C terminus of the Atx-10 molecule.

Cloning and Sequence
Circular dichroism analysis resulted in a predominantly ␣-helical signal with characteristic minima at 208 and 222 nm (Fig. 2D). The calculated helix content is 36%. A secondary structure prediction for the Atx-10 protein sequence revealed a repetitive pattern of helical stretches (Fig. 2E) interspersed with turn structures. Fold recognition analysis using the 3D-PSSM program (16) clearly defined Atx-10 as a member of the armadillo repeat protein family with ␤-catenin and importin-␤ structures as best fitting models.
Subcellular Localization and Detection of Atx-10 in Primary Neuronal Cells-To assess the subcellular localization of Atx-10, we fractionated PC12 cell lysates by centrifugation at 10,000 ϫ g and looked for Atx-10 in the pellet and supernatant fractions by Western blotting. As shown in Fig. 3A, Atx-10 was exclusively found in the soluble cell fraction, indicating that it did not associate with organelles or the cell membrane. In a second experiment, the soluble fraction of the PC12 cell lysate obtained in the first experiment (Fig. 3A) was pelleted by ultracentrifugation at 100,000 ϫ g to separate cytosolic (supernatant) and microsomal (pellet) fractions. Both samples contained comparable amounts of Atx-10 in this experiment, indicating that the protein was not largely attached to membranes. The detection of Atx-10 in the pellet fraction in this experiment can be explained by residual cytoplasmic solution, which cannot be removed completely from the sample without loss of microsomes. Both fractions were treated with proteinase K in the absence or presence of detergent to exclude the possibility that Atx-10 was localized on the lumen side of vesicles that is protected from protease cleavage. Proteinase K treatment removed the Atx-10 signal both in the intact microsomal fraction and in the detergent-treated microsomal fraction, confirming its cytoplasmic localization (Fig. 3B).
Detection of Atx-10 in PC12 cells by fluorescence microscopy using Atx-10 antibody revealed a cytoplasmic pattern confirm-ing our biochemical results (Fig. 3C). A similar result was obtained with a preparation of neurons from rat cortex (Fig.  3D) and cerebellum (Fig. 3E), which stained strongly for Atx-10. Interestingly, in a culture of cerebellar astrocytes (95% glial fibrillary acidic protein-positive) exclusively, some remaining neurons were stained, whereas the surrounding astrocytes showed only background fluorescence intensity for Atx-10 ( Fig.  3, F and G).
SCA10 Expression Pattern in Rat Tissues and Atx-10 Immunostaining in Human and Mouse Brain Sections-The expression pattern of the rat SCA10 gene was analyzed by Northern blotting using tissues of various rat organs (Fig. 4). As shown in Fig. 4A, the SCA10 transcript showed a ubiquitous distribution. SCA10 gene expression was, however, markedly increased in tissues of testis, adrenals, and brain. For a more detailed neuronal expression pattern at the protein level, we performed Atx-10 immunostainings using mouse and human brain sections. The intensity of Atx-10 reactivity differed significantly between the different neuronal populations and between the nerve cell bodies of individual nuclei (Fig. 5 and Table II). Atx-10 immunostaining was diffusely distributed in the cell bodies of neurons and was also present in the proximal parts of dendritic processes. Only in the human cerebellum was a perinuclear ring of increased reactivity observed in some Purkinje cells (Fig. 5F). A strong Atx-10 reactivity in the cell nucleus was only observed in the spinal ganglia neurons. Strong axonal Atx-10 reactivity was observed in long tracts of the human and murine spinal cord. However, the main fiber tracts (corpus callosum and fornix) of the cerebral hemispheres remained unstained. The most intense reactivities were observed in the olivary nuclei (Fig. 5, A and B), the nuclei pontis (Fig. 5, C and D), the deep cerebellar nuclei, the Purkinje cells (Fig. 5, E and F), the vestibular and cuneate nuclei, and the sensory ganglion cells. The mouse neuronal populations showing lesser Atx-10 reactivities were the mitral cells of the olfactory bulb, pyramidal cells of the upper cortical layers of the cerebral cortex, neurons of the midbrain tectum, the raphe nuclei, the locus coeruleus, and the motor nuclei in brain stem and spinal cord. Variable intensity of staining was observed in pyramidal cells of the human cerebral cortex. Very weak staining of nerve cells was seen in the thalamus. The dentate gyrus and the pyramidal cell layer of the hippocampus remained unstained (Fig. 5, G and H). Interestingly, strong Atx-10 staining was observed in clustered small neuropil granules in the hippocampus, cerebellum, and brain stem of the mouse brain, very similar to the granules reported by Jucker et al. (17) in aging mice. Some Atx-10-positive cells were also found in the Purkinje cell layer of the human cerebellum (Fig. 5F). Functional Analysis of Atx-10 Using in Vitro siRNA Knock Down Technology-Our immunohistochemical data revealed a predominant expression of Atx-10 in neuronal subpopulations of the central nervous system. Therefore, to assess the functional role of Atx-10, we decided to use primary neurons as potential targets to interfere with SCA10 gene expression. Primary cerebellar and cortical neurons were transfected with a combination of two cognate siRNA duplexes 7 days after plating or with transfection reagent as control. To evaluate the ability of siRNA transfection to abolish target gene expression, cell lysates were analyzed 24, 48, and 72 h after transfection by immunoblotting. As shown in Fig. 6, A and B, specific silencing of the SCA10 gene product was confirmed in siRNA-transfected neurons, whereas in control transfected neurons, Atx-10 expression remained unchanged. Atx-10 protein levels were significantly reduced already after 48 h, being hardly detectable after 72 h. Equal loading was confirmed by the presence of consistent amounts of ␤-actin protein.
Phenotypic effects of Atx-10 knock down on cultured neurons were first observed by phase microscopy, indicating that neuronal survival was impaired as compared with control-transfected cells (not shown). Therefore, neuronal cell loss was evaluated measuring lactate dehydrogenase release into the culture medium of siRNA-transfected and control-transfected central nervous system neurons. To compare different neuronal target populations, lactate dehydrogenase release was analyzed 48 h after siRNA transfection in primary cortical and cerebellar neurons. Silencing of SCA10 gene expression resulted in highly significant increase in cytotoxicity in both cerebellar and cortical neurons as compared with controls (Fig.  6C). However, cerebellar neurons appeared to be significantly more sensitive to reduced Atx-10 levels (38.3 Ϯ 7.9%; mean Ϯ S.D.) than cortical neurons (9.8 Ϯ 1.9%; mean Ϯ S.D.), although Atx-10 expression was knocked down equally. These results indicate that the presence of Atx-10 is essential for the survival of this subpopulation of neurons.

DISCUSSION
In this study, we report the cloning of the rat SCA10 gene, which codes for an intracellular protein of ϳ55 kDa with clear sequence homologies to related proteins in mice and humans. Our attempts to express the full-length protein in E. coli were unsuccessful, although fragments of Atx-10 were well expressed in a bacterial system, and a large C-terminal fragment (P4) could be refolded from inclusion bodies, allowing the production of a polyclonal antibody. Full-length Atx-10 was expressed in HEK293 cells as a secreted fusion protein showing a slightly higher molecular mass in SDS-PAGE than the endogenous protein of PC12 cells due to the added tag and protease cleavage sequences (ϳ4 kDa). Homology searches and fold prediction programs classify Atx-10 as member of the armadillo repeat protein family with an all-␣-helical fold. This was confirmed by CD-spectroscopic analysis of both Atx-10 and the P4 fragment (not shown) and electron micrographs of recombinant Atx-10 showing molecules with an elongated curved shape characteristic for this protein family (Fig. 2C).
The armadillo repeat motif usually confers protein-protein interactions with diverse cellular binding partners and elicits functions in many different biological contexts (18). A tendency for homo-oligomerization as in the case of Atx-10 represents a hitherto novel feature of this domain. We have observed apparent oligomer formation both for the recombinant proteins and for endogenous Atx-10 as deduced from gel filtration analysis of PC12 cell lysates (not shown). As pointed out above, we have indications that the homoassociation site is located near the protein's C terminus. Crystal structures of complexes with armadillo repeat proteins have identified binding regions exclusively in grooves on the concave surfaces of the superhelix (19,20). As our electron micrographs of recombinant Atx-10 suggest a tip-to-tip association in a horseshoe-shaped fashion, it can be concluded that oligomerization might either open a more extended polyvalent binding region or render the binding site inaccessible, thereby inactivating the protein function. Interestingly, alignments of the SCA10 gene with homologous sequences from phylogenetically distant organisms as Drosophila (accession number AAF57806), Saccharomyces (accession number QO9888), and Arabidopsis (accession number NM116241) display a clear gradient of conservation toward the C terminus of the protein (not shown), indicating that the proposed homoassociation site is essential for its cellular function.
The subcellular localization of Atx-10 was predominantly cytoplasmic ruling out a similar function as for ␤-catenin, which is associated with the intracellular domain of E-cadherin FIG. 6. Gene silencing of Atx-10 in primary neurons of rat cerebellum and cortex. Western blot analysis of cerebellar (A) and cortical (B) neurons transfected with Atx-10 siRNA duplexes and mocktransfected neurons probed with Atx-10-specific antibody after 24, 48, and 72 h. The blot was stripped and reprobed with ␤-actin antibody. C, Atx-10 suppression by cognate siRNAs leads to significant neuronal loss within 48 h, shown as the mean percentage of lactate dehydrogenase release Ϯ S.D. (n ϭ 3) as compared with mock-transfected sister cultures (ဧ). The asterisks denote differences from controls (**, p Ͻ 0.01, analyzed with two-tailed t test). and hence attached to the cell membrane. This was additionally confirmed by co-precipitation experiments using a fusion protein in which the intracellular portion of murine E-cadherin was fused to glutathione S-transferase (not shown). We can also exclude an association of Atx-10 with major components of the cytoskeleton, since Atx-10 was exclusively found in the cytosolic fraction of cell lysates, and double stainings of PC12 cells using phalloidin or anti-tubulin antibody did not result in overlapping patterns (not shown). We are currently working on a proteomics approach to identify binding partners of Atx-10 in PC12 cells, which might point to parallels in its cellular function in cerebellar neurons. Our Northern blot experiments have shown that SCA10 expression is up-regulated to a similar degree in brain and adrenal tissues (Fig. 4A).
Spinocerebellar ataxias constitute a fast growing group of hereditary diseases caused by different genetic defects that result in a similar neurological disorder (6). The generally unspecific clinical phenotype results from tissue degeneration affecting the cerebellum itself or its afferent and efferent pathways. No autopsy studies of SCA10 patients have been published to date. Thus, we are left with conjectures concerning the distribution of histopathological changes in this particular form of SCA. However, magnetic resonance imaging data in SCA10 patients have shown severe cerebellar atrophy (21), accounting for the cardinal clinical features of the disease. The observed distribution of Atx-10 reactivity together with the high protein expression in spinal ganglia neurons and areas known to project to the cerebellum, including the nuclei pontis and the olivary nuclei, is consistent with spinocerebellar involvement and with the reported clinical symptomatology in affected families (21). In addition, the high expression of Atx-10 in spinal ganglia neurons might be related to the peripheral neuropathy known to occur in a large proportion of SCA10 patients.
The SCA10 mutation, which has been localized to 22q13.3, represents a novel class of microsatellite expansion formed by a pentanucleotide repeat within a noncoding gene region. Microsatellite expansions in noncoding regions have been described for several other dominantly inherited genetic disorders, including myotonic dystrophies 1 and 2 and SCA8, which are caused by triplet or, as in the case of myotonic dystrophy 1, tetranucleotide expansions (2). For these, different mechanisms of pathogenesis are discussed: 1) haploinsufficiency on the protein level, 2) disregulated expression of neighboring genes, and 3) toxic effects by gain-of-function at the RNA level. Although several reports have determined monocausal mechanisms responsible for the clinical phenotypes in noncoding repeat expansion disorders, there is also growing evidence for additive disease models (22).
Our siRNA data demonstrate that loss of SCA10 expression is sufficient for cell degeneration (Fig. 6C). This effect was highly specific and restricted to cerebellar neurons, excluding a general toxicity induced by the applied siRNAs (Fig. 6, A and   B). This observation would argue for a loss-of-function phenotype in SCA10 patients, although definite results will only be obtained when patient material from affected tissues is available. Matsuura et al. (15) have reported that SCA10 mRNA levels in lymphoblastoid cell lines were unchanged in SCA10 patients. However, since elevated expression levels seem to be restricted to neuronal cell lines, we suggest that a reduced expression from haploinsufficiency alone, which might not add to a basal expression, would only affect tissues in which higher concentrations of Atx-10 are essential.
Our data present the first characterization of Atx-10 and suggest a possible pathogenic mechanism involving reduced protein expression levels in affected tissues, predominantly in sensitive neurons and tracts of the cerebellum. Further studies are needed to elucidate the functional role of Atx-10 in neuronal target cells.