dAtaxin-2 Mediates Expanded Ataxin-1-Induced Neurodegeneration in a Drosophila Model of SCA1

Spinocerebellar ataxias (SCAs) are a genetically heterogeneous group of neurodegenerative disorders sharing atrophy of the cerebellum as a common feature. SCA1 and SCA2 are two ataxias caused by expansion of polyglutamine tracts in Ataxin-1 (ATXN1) and Ataxin-2 (ATXN2), respectively, two proteins that are otherwise unrelated. Here, we use a Drosophila model of SCA1 to unveil molecular mechanisms linking Ataxin-1 with Ataxin-2 during SCA1 pathogenesis. We show that wild-type Drosophila Ataxin-2 (dAtx2) is a major genetic modifier of human expanded Ataxin-1 (Ataxin-1[82Q]) toxicity. Increased dAtx2 levels enhance, and more importantly, decreased dAtx2 levels suppress Ataxin-1[82Q]-induced neurodegeneration, thereby ruling out a pathogenic mechanism by depletion of dAtx2. Although Ataxin-2 is normally cytoplasmic and Ataxin-1 nuclear, we show that both dAtx2 and hAtaxin-2 physically interact with Ataxin-1. Furthermore, we show that expanded Ataxin-1 induces intranuclear accumulation of dAtx2/hAtaxin-2 in both Drosophila and SCA1 postmortem neurons. These observations suggest that nuclear accumulation of Ataxin-2 contributes to expanded Ataxin-1-induced toxicity. We tested this hypothesis engineering dAtx2 transgenes with nuclear localization signal (NLS) and nuclear export signal (NES). We find that NLS-dAtx2, but not NES-dAtx2, mimics the neurodegenerative phenotypes caused by Ataxin-1[82Q], including repression of the proneural factor Senseless. Altogether, these findings reveal a previously unknown functional link between neurodegenerative disorders with common clinical features but different etiology.


Introduction
Inherited ataxias are a genetically heterogeneous group of neurodegenerative diseases characterized by loss of motor coordination and balance. They can be caused by loss-offunction or gain-of-function mechanisms; some ataxias are triggered by missense mutations, while others by triplet repeat expansions, which may occur either in coding or noncoding sequences. Furthermore, the gene products implicated in the different ataxias do not share obvious functional or structural relationships to each other. In spite of this genetic heterogeneity, many ataxias show striking similarities. In particular, it is often difficult to distinguish between Spinocerebellar ataxias (SCAs) based only on clinical and pathological observations, and their differential diagnosis often requires genetic testing. In addition, a common neuropathological feature of SCAs is the atrophy of the cerebellar module (reviewed in [1][2][3]). These similarities suggest that SCAs, and perhaps other ataxias, may also share common mechanisms of pathogenesis. In support of this hypothesis a recent study reported a network of physical protein-protein interactions among many factors associated with ataxia and Purkinje cell degeneration in humans and mice [4]. However, no specific molecular mechanisms are known that can account for the clinical and neuoropathological similarities among SCAs and other ataxias. SCA1 is caused by the expansion of a CAG repeat encoding a polyglutamine tract in the protein Ataxin-1 that induces a toxic gain of function [5]. The expanded protein accumulates in neuronal nuclear inclusions (NIs) that also contain transcription factors, chaperones, proteasome subunits, and other components of the protein quality control/degradation machinery like CHIP or Ataxin-3 [6][7][8][9][10][11]. Abnormally long polyglutamine tracts are the common cause of pathogenesis in at least five other SCAs (SCA2, 3, 6, 7 and 17) and three additional neurodegenerative diseases including Huntington's disease (HD) [1,12]. Protein quality control machinery as well as transcriptional dysregulation are general mechanisms that have been implicated in the pathogenesis of these polyglutamine disorders [13][14][15].
Although the polyglutamine expansion triggers the toxicity of Ataxin-1, experiments in Drosophila and mouse SCA1 models have shown that protein context plays a key role in expanded Ataxin-1-induced neurodegeneration (reviewed in [15]). The nuclear localization signal [16] and phosphorylation [17] influence the toxicity of expanded Ataxin-1. In addition, certain interacting partners of unexpanded Ataxin-1 are critical to expanded Ataxin-1 toxicity [9,18,19]. In this context, expanded Ataxin-1 was recently found to induce a decrease in the levels of Senseless (Sens) and its murine orthologue growth factor independent 1 (Gfi1) [18]. These are transcription factors that interact with unexpanded Ataxin-1 and are necessary for Purkinje cell survival in mice [18] and for sensory organ development in Drosophila [20]. The importance of the protein framework has also been shown in models of other polyglutamine diseases [15,21]. Genetic screening in Drosophila models of neurodegenerative diseases is a powerful approach to identify modifier genes and pathways implicated in pathogenesis [22][23][24]. We previously reported an unbiased genetic screen with a Drosophila model of SCA1 [25]. Here we report the identification of the Drosophila homolog of Ataxin-2 (dAtx2) as a major modifier of expanded Ataxin-1-induced toxicity. Ataxin-2 is a widely expressed cytoplasmic protein with no similarity to Ataxin-1 except for the polyglutamine domain. The normal function of Ataxin-2 remains unclear, although it has been implicated in mRNA processing [26][27][28] and translational regulation in yeast [29,30], C. elegans [31] and Drosophila [32], where it is also required for actin filament formation [33]. However, expansion of its polyglutamine domain leads to SCA2 [34][35][36]. The functional interactions between Ataxin-1 and Ataxin-2 described here mechanistically tie these two proteins and point to previously unknown pathogenic links between two inherited ataxias.

Results
Increased/Reduced Levels of Ataxin-2 Enhance/Suppress Expanded Ataxin-1 Toxicity in the Drosophila Eye Expression of Ataxin-1 [82Q] in the eye of SCA1 82Q flies causes external and internal abnormal phenotypes [25].
Externally, the eyes of these animals show severe ommatidial disorganization as well as interommatidial bristle loss when compared with control eyes (Figure 1, compare A and A9 with B and B9). Internally, examination of the retina reveals tissue loss and shortened and curved photoreceptor neurons ( Figure 1, compare F with G). In a screen for genetic modifiers of Ataxin-1[82Q]-induced toxicity we recovered EP(3)3145 as an enhancer of the eye phenotype (data not shown). This is an insertion of an EP transposable element [37] in the 59 end of dAtx2, the Drosophila orthologue of human Ataxin-2. The Drosophila and human proteins share 23% amino acid identity and 36% amino-acid similarity over the entire protein with the most conserved sequences corresponding to the ATX2-N and ATX2-C domains (43% and 62% identity, respectively) [33]. Molecular analysis revealed that the EP element is inserted 3121 bp upstream of the ATG and in the same orientation as the dAtx2 transcription unit (data not shown and [33]). These data suggested that EP(3)3145 over-expresses the dAtx2 transcription unit to enhance the SCA1 82Q eye phenotype. As described below, this possibility was confirmed using a transgene that overexpresses the dAtx2 cDNA.
Co-expression of a wild-type dAtx2 transgene (dAtx2 OE ) at low levels enhances the Ataxin-1[82Q]-induced eye phenotype. Externally, the eyes of SCA1 82Q /dAtx2 OE animals show no bristles and increased ommatidial disorganization when compared with the eyes of SCA1 82Q controls (compare Figure  1D and D9 with B and B9). Internally, photoreceptor cells are considerably shorter (compare Figure 1I with G). Expression of the same low levels of dAtx2 alone in the eye causes relatively mild external disorganization and reduction of the retinal width ( Figure 1E, E9 and J). Overexpression of dAtx2 from EP(3)3145 and UAS-dAtx2 also aggravates the phenotypes of other fly models of neurodegenerative diseases besides SCA1 [38,39]. However, since overexpression of dAtx2 causes an eye phenotype by itself ( Figure 1E, E9 and J) and it is toxic in many other tissues [33], it is difficult to make strong conclusions about the specificity of these genetic interactions.
To test the specificity of the genetic interaction, we investigated if decreasing the levels of endogenous dAtx2 modifies expanded Ataxin-1-induced toxicity. For this, we used a 1.4 kb deletion in the dAtx2 locus (dAtx2 X1 ) that removes part of the dAtx2 promoter, the ATG codon and extends into the first intron [33]. We find that flies expressing Ataxin-1[82Q] and heterozygous for the dAtx2 X1 mutant allele show a strong suppression of the eye phenotype, with much improved arrangement of the ommatidia and bristles compared to eyes from flies expressing Ataxin-1[82Q] with normal dAtx2 levels (compare Figure 1C and C9 with B and B9). This suppression is also evident in the retinas of SCA1 82Q / dAtx2 X1 flies that show elongated photoreceptors and very little tissue loss (compare Figure 1H with G). To further test the specificity of this interaction, and to exclude potential genetic background artefacts, we asked whether adding back dAtx2 to SCA1 82Q /dAtx2 X1 flies eliminates the suppression effect. Figure S1 shows that SCA1 82Q /dAtx2 X1 /dAtx2 OE flies show an eye phenotype that is very similar to the phenotype of SCA1 82Q flies. The effects of the dAtx2 X1 and dAtx2 OE alleles decreasing/increasing dAtx2 levels are demonstrated in Figure S2.
Since dAtx2 is an RNA binding protein, we investigated if the observed suppression of Ataxin-1[82Q] toxicity was the

Author Summary
The spinocerebellar ataxias (SCAs) are a group of ;30 neurodegenerative disorders caused by different types of mutations in a variety of unrelated genes. For example, SCA1 and SCA2 are caused by mutations in Ataxin-1 and Ataxin-2, two proteins related in name only. Despite these differences, most SCAs share a number of striking clinical and neuropathological similarities, such as ataxia, tremor, speech difficulties, and atrophy of the cerebellum and brainstem. In addition, many ataxia-causing proteins share interacting protein partners. Together, these observations suggest that many SCAs also share common mechanisms of pathogenesis. Here, we report previously unknown functional interactions between the genes and proteins responsible for SCA1 and SCA2. We find that Ataxin-1 and Ataxin-2 physically interact, and that mutant Ataxin-1 forces Ataxin-2 to accumulate in the nucleus instead of the cytoplasm. Most importantly, using an animal model, we discovered that the Drosophila Ataxin-2 gene is a strong suppressor of Ataxin-1induced neurotoxicity. Thus, neuronal degeneration may take place through common mechanisms in different SCAs. These findings open the possibility of future common therapies for these neurodegenerative disorders for which there is no effective treatment.
result of dAtx2 affecting the levels of the SCA1 82Q mRNA transcript or the levels of Ataxin-1 [82Q] protein. As shown in Figure 1K and L neither the levels of SCA1 82Q mRNA nor the levels of the Ataxin-1[82Q] protein are affected by changing the levels of dAtx2.

Reduced Levels of Ataxin-2 Suppress Expanded Ataxin-1-Induced Neuronal Dysfunction in Drosophila
To verify that the genetic interaction between Ataxin-1 and dAtx2 is not limited to the eye, we analyzed the effect of altering dAtx2 levels on expanded Ataxin-1-induced neuronal dysfunction.
The motor performance of flies as a function of age can be quantified using a climbing assay [40]. This assay has been used to analyze the effects of toxic proteins on neurons in other Drosophila models of neurodegenerative diseases [41,42]. Control flies show no significant decrease in their motor performance until late in life. Figure 2A shows that ;74% of control flies still climb after thirty-six days (black triangles).
Flies expressing Ataxin-1[82Q] specifically in the nervous system (using nrv2-GAL4) display a progressive impairment of their motor performance (Figure 2A, blue circles). In the context of the Drosophila life span, this is a late onset and progressive phenotype as compared to the performance of control flies in the same period of time. We then analyzed the   Figure 2A (red squares) in SCA1 82Q flies also heterozygous for the dAtx2 X1 mutation climbing performance is significantly improved compared to flies with normal dAtx2 levels (p,0.0001 for repeated measures anova (rma) between genotypes). Figure 2A show that while all SCA1 82Q animals fail to climb after 26 days, SCA1 82Q / dAtx2 X1 flies continue to climb until later in life. Thus the impairments in motor performance caused by neuronal expression of Ataxin-1[82Q] are suppressed by decreased dAtx2 levels.
We also studied the effect of Ataxin-1[82Q] expression in a life span assay. Figure 2B shows that expression of Ataxin-1[82Q] in the nervous system leads to premature death in SCA1 82Q flies in comparison to GFP controls ( Figure 2B, compare blue circles with black triangles). While SCA1 82Q animals do not survive past 30 days, this early lethality phenotype is suppressed in SCA1 82Q / dAtx2 X1 animals ( Figure  2B, red squares).

Decreased Levels of dAtx2 Do Not Suppress Neurodegeneration Caused by Expanded Huntingtin
To investigate whether dAtx-2 also modulates neurodegeneration in other models of polyglutamine disease, we tested the effect of altering the dAtx2 levels in a Drosophila model of Huntington's disease [7,43]. Adult flies expressing an expanded N-terminal fragment of human huntingtin (N-Htt 128Q ) in the eye show a progressive retinal degeneration which becomes obvious at day 5 after eclosion [7]. N-Htt 128Q retinas show disorganized and missing photoreceptors ( Figure 2D, compare with control in 2C). N-Htt 128Q flies also overexpressing dAtx2 present a more degenerated retina than flies expressing N-Htt 128Q with normal levels of dAtx2 (data not shown). However, since overexpression of dAtx2 is sufficient to cause retinal degeneration ( Figure 1J), this result is not conclusive by itself. Therefore, we tested the effect of decreasing dAtx2 levels on the N-Htt 128Q induced retinal degeneration. As shown in Figure 2D-E, decreasing the levels of dAtx2 (N-Htt 128Q /dAtx2 X1 ) does not obviously alter degeneration induced by N-Htt 128Q in the Drosophila eye.
We also investigated a possible genetic interaction between dAtx2 and N-Htt 128Q in the motor performance assay described above. Like with Ataxin-1[82Q], expression of N-Htt 128Q in the nervous system leads to motor performance impairments ( Figure 2F) where N-Htt 128Q animals stop climbing before day 20. The climbing performance of animals expressing N-Htt 128Q with decreased levels of dAtx2 (N-Htt 128Q /dAtx2 X1 ) is not significantly different from that of animals expressing N-Htt 128Q with normal levels of dAtx2 ( Figure 2F). Therefore decreasing the levels of dAtx2 fails to suppress N-Htt 128Q induced degeneration both in the retina and in the nervous system.

Ataxin-2 Levels Modulate Expanded Ataxin-1-Induced Loss of Mechanoreceptors Caused by Loss of Senseless Protein
Senseless (Sens) is a proneural factor that is expressed and required in the sensory organ precursor (SOP) cells of the peripheral nervous system [20]. Expression of high levels of Ataxin-1[82Q] in the thoracic SOPs using scabrous-GAL4 (sca-GAL4) leads to a reduction in Sens protein levels in these cells and loss of large mechanoreceptors (macrochaetae) in the thorax of adult flies [18]. Therefore, scoring adult thoracic macrochaetae in SCA1 82Q animals provides a quantitative phenotype with a known molecular foundation.
In the wing imaginal disc, Sens expression includes the precursors of the large thoracic mechanoreceptor bristles, and the two parallel rows of bristles at each side of the wing margin (Figs. 3F-G). We quantified the amount of fluorescence detected after anti-Sens immunostaining in wing margin cells expressing Ataxin-1[82Q] with normal or increased levels of dAtx2. Expression of either low levels of Ataxin-1[82Q] or dAtx2 with sca-Gal4 (at 25C) produces no detectable decrease in the levels of Sens in the wing margin ( Figure 3H and I compare with G and quantification in K). However co-expression of both Ataxin-1[82Q] and dAtx2 in the same conditions induces a strong decrease in the levels of Sens ( Figure 3J). Quantification of the Sens signal in SCA1 82Q / dAtx2 OE animals ( Figure 3K, fourth column) reveals a decrease of ;50% in the amount of Sens signal when compared to either wild type, SCA1 82Q or dAtx2 OE controls (p,0.001, Tukey-Kramer HSD).
In addition, we investigated the consequences of decreasing the amount of dAtx2 (using the dAtx2 X1 allele) in conditions where Ataxin-1[82Q] reduces Sens levels in the wing margin (i.e. at 278C and 298C). However, we did not detect a significant modification (data not shown). Since it is possible that we are not able to detect small changes on Sens levels in wing discs from late third instar larvae, or this changes might happen later in the SOP development, we also investigated the effect of decreasing the amount of Ataxin-2 on Ataxin-1[82Q]-induced mechanoreceptor loss. Expression of Ataxin-1[82Q] at high levels (278C) in the SOP cells results in the loss of ;20% of thoracic macrochaetae when compared to controls ( Figure 3L compare columns 1 and 2 p,0.0001, Tukey-Kramer HSD; see Figure 3M). Reducing the levels of endogenous dAtx2 with the heterozygous dAtx2 X1 mutation partially rescues the Ataxin-1[82Q]-induced bristle phenotype. Loss of macrochaetae in SCA1 82Q / dAtx2 X1 animals is approximately half of that seen in animals with normal levels Ataxin-2 ( Figure 3L, compare columns 2 and 3 p,0.005, Tukey-Kramer HSD, and see Figure 3N).

Physical Interaction between Human Expanded Ataxin-1 and Ataxin-2 Proteins
To further characterize the interactions between Ataxin-1 and Ataxin-2, we investigated possible protein-protein interactions. Lysates from cells expressing Drosophila or human Ataxin-2 and GST-Ataxin-1 [82Q] were subjected to co-affinity purification (co-AP) glutathione-S-transferase (GST) pulldown assays. As shown in Figure 4A (lanes 1 and 2) and Figure  4B (lanes 1 and 4), GST-Ataxin-1[82Q] is able to pull down the Drosophila and human Ataxin-2 proteins, which indicates that both proteins are able to physically interact. We also asked whether this interaction is polyglutamine dependent. As shown in Figure 4B lanes 2-4 we did not detect significant differences in the interactions between GST-Ataxin-1 with 2, 30 or 82 glutamines and human Ataxin-2 in this co-AP assay.
In addition, we investigated if Ataxin-1 can pull down endogenous hATX2. We find that unexpanded Ataxin-1 is able to precipitate endogenous hATX2 from human cells, suggesting that the two proteins may be functional interactors in vivo ( Figure 4C).
We also investigated whether the interaction between Ataxin-1 and Ataxin-2 is cytoplasmic or nuclear. We carried out co-AP assays with Ataxin-1 and dAtx2 after nuclear/ cytoplasmic fractionation of cultured cells. Figure 4D shows that we were not able to detect differences in protein interactions between these cellular compartments using this assay.
Lastly, we investigated whether specific domains of the hAtaxin-1 protein are responsible for the interaction with hAtaxin-2. Co-AP experiments were carried out with lysates   Figure 4E lane-5). All three fragments pull-down Myc-hAtaxin-2, indicating that each Ataxin-1 fragment can interact independently with hAtaxin-2 ( Figure 4E lanes 3-5). However, the interaction of hAtaxin-2 is stronger with the N-terminal Ataxin-1 fragment ( Figure 4E

Ataxin-2 Accumulates in the Nucleus of Expanded Ataxin-1-Expressing Drosophila Cells and Human Neurons
The co-AP assays with expanded hAtaxin-1 and hAtaxin-2 indicate that the two proteins are able to interact. However, Ataxin-1 normally localizes to the nucleus in Drosophila and many human cell types, while Ataxin-2 is a cytoplasmic protein. To address whether the interaction observed in cultured cells is relevant in vivo, we monitored the localization of Ataxin-2 in Drosophila cells expressing Ataxin-1[82Q]. Since Ataxin-1[82Q]-induced toxicity is suppressed by dAtx2 loss of function in the Drosophila eye; we first analyzed the localization of dAtx2 in retinal cells. dAtx2 is not normally detected in the nuclei of retinal cells from either control eyes or eyes overexpressing dAtx2 ( Figure 5A and C respectively). In contrast, we find that endogenous dAtx2 localizes to the nuclei of retinal cells expressing Ataxin-1[82Q]. Furthermore, nuclear dAtx2 signal is detected both diffusely in the nucleoplasm as well as in nuclear inclusions (NIs) ( Figure 5B). To confirm this unexpected result, we examined the localization of dAtx2 in other Ataxin-1[82Q]expressing neurons. Similar to the results obtained in the retina, endogenous dAtx2 normally localizes to the cytoplasm  Figure 5 F-F99). Next we investigated whether dAtx2 and Ataxin-1 colocalize. Costaining of Ataxin-1 and dAtx2 is not possible since the available antibodies for both proteins were raised in rabbit; however, Ataxin-1 NIs in Drosophila neurons are positive for Ubiquitin [25], so we performed a double staining for dAtx2 and Ubiquitin. Figure 5G-G99 shows that the dAtx2 NIs present in the neurons of SCA1 82Q flies are positive for Ubiquitin. These results indicate that expanded Ataxin-1 causes dAtx2 to localize to the nucleus and suggest that both proteins co-aggregate in NIs.
To validate these results and investigate their relevance for SCA1 pathogenesis, we analyzed the localization of hAtaxin-2 in SCA1 neurons from human postmortem brain samples by anti-Ataxin-2 immunohistochemistry. Pontine neurons from control samples consistently show cytoplasmic localization of hAtaxin-2 ( Figure 5H). Pontine neurons from SCA1 brain samples display frequent Ataxin-1 NIs that are clearly visible with Hematoxylin staining. We found that approximately twenty percent of these NIs are positive for hAtaxin-2 ( Figure  5I); see also [10]. These findings, together with the co-AP data using human Ataxin-1 and Ataxin-2 proteins, suggest that the Ataxin-1-Ataxin-2 interactions observed in Drosophila may be relevant for SCA1 pathology.  Figure 6A-D shows that no nuclear dAtx2 signal is detected in control neurons ( Figure  6A), or neurons expressing Ataxin-1[2Q] or [30Q] ( Figure 6B and C). In contrast, accumulation of dAtx2 is detected in the nucleus of Ataxin-1[82Q]-expressing neurons ( Figure 6D). Although this result does not rule out an interaction between wild-type Ataxin-1 and dAtx2, it indicates that abundant nuclear accumulation of Ataxin-2 is specific to the expanded Ataxin-1 form responsible for SCA1 pathogenesis.

Nuclear Accumulation of Ataxin-2 Causes Severe Eye Toxicity
The observations that expanded Ataxin-1 induces nuclear accumulation of Ataxin-2 and that decreasing the levels of endogenous Ataxin-2 suppresses toxicity in SCA1 82Q flies, suggest that nuclear accumulation of Ataxin-2 may lead to neurotoxicity. To test this hypothesis, we generated dAtx2 transgenic flies with an exogenous nuclear localization signal (NLS) engineered on its C-terminus end (dAtx2 NLS ) (Figure 7 compare A-A9 with C). Although wild-type Ataxin-2 is only detected in the cytoplasm in Drosophila and human neurons, it is difficult to rule out the possibility that some Ataxin-2 may be present in the nucleus. Therefore, we also generated flies carrying a dAtx2 construct with an exogenous nuclear export signal (NES) (dAtx2 NES ) to use as an additional control ( Figure  7B-B9). Ten transgenic lines were recovered for each dAtx2 construct, with a wide range of expression levels for each transgene. We selected transgenic lines expressing wild-type dAtx2 (dAtx2 OE ), dAtx2 NES or dAtx2 NLS at similar levels ( Figure 7D compare lanes 2-4), and compared their toxicity in the eye. As shown in Figure 7E-H, dAtx2 NLS is more toxic than wild-type dAtx2 or dAtx2 NES . While both wild-type dAtx2 and dAtx2 NES induce a relatively mild eye phenotype compared to controls (Figure 7 E-G), expression of dAtx2 NLS results in strong eye toxicity ( Figure 7H). Eyes of dAtx2 NLS flies show a severely disorganized ommatidial lattice and a In summary, expression of dAtx2 NES and wild-type dAtx2 in the eye cause similar mild phenotypes. This is consistent with dAtx2 being mainly cytoplasmic, and with observations of SCA2 pathogenesis in the cytoplasm [44,45]. Interestingly, increasing the levels of Ataxin-2 in the nucleus is sufficient to cause a much more severe eye phenotype. These observations suggest that the toxicity of expanded Ataxin-1 is mediated in part, by the nuclear accumulation of Ataxin-2.

Nuclear Ataxin-2 Decreases Sens Protein Levels and Induces Loss of Mechanoreceptor Bristles
To further test the hypothesis that nuclear Ataxin-2 contributes to expanded Ataxin-1-induced toxicity, we analyzed the effect of expressing the different dAtx2 transgenes on Sens distribution in the wing margin SOPs. Expression of Ataxin-1[82Q] in the antero-posterior compartment boundary (using dpp-GAL4) induces a cell autonomous decrease of Sens only in the dpp-GAL4 expressing area of the wing margin ( Figure 8A, A9 and 8B, B9, arrowhead).
Expression of dAtx2 NES from dpp-GAL4 does not reduce the levels of Sens, whose distribution in imaginal wing discs is unchanged ( Figure 8C and C9, arrowhead). Next, we tested the effect of dAtx2 with a nuclear localization signal (dAtx2 NLS ). Like Ataxin-1[82Q], expression of dAtx2 NLS induces loss of Sens in the wing margin SOPs ( Figure 8D and D9, arrowhead). dAtx2 NLS -induced loss of Sens is also observed in other cell types. Salivary gland cells express Sens at high levels [20], which localizes to the nucleus ( Figure 8E-H). Salivary gland cells expressing dAtx2 NES show no detectable change in Sens levels or distribution ( Figure 8I-L). In contrast, expression of dAtx2 NLS causes a dramatic decrease in the amount of Sens ( Figure 8M-P). The nuclei of dAtx2 NLS cells are still present and their morphology is similar to controls, indicating that Sens loss is unlikely a consequence of cell death. Therefore, nuclear dAtx2 mimics Ataxin-1[82Q] in causing loss of Sens protein accumulation.
Expression of Ataxin-1 in the thoracic SOPs (using sca-GAL4) leads to loss of macrochaetae in the adult thorax (Figure 3 and ref. [18]). Therefore, we also investigated the consequences of nuclear or cytoplasmic Ataxin-2 accumu- Figure 7. Nuclear Accumulation of dAtx2 Increases Its Toxicity In Vivo (A-C) Immunofluorescence staining to reveal the localization of dAtx2 in neurons overexpressing CD8:GFP together with wild-type dAtx2 (dAtx2 OE ) or dAtx2 with a nuclear export or a nuclear localization signal (dAtx2 NES or dAtx2 NLS , respectively). (A and A9) dAtx2 localizes to the cytoplasm even when overexpressed. (B and B9) dAtx2 NES distribution is similar to that of wild-type dAtx2. (C) nuclear localization of dAtx2 NLS . (D) Anti-dAtx2 immunoblot demonstrating similar expression levels of the different dAtx2 transgenes used in this work. The transgenes were driven by gmr-GAL4. Larval eye imaginal discs (ten per genotype) were used to avoid artifact potentially caused by cell degeneration or loss. dTubulin was blotted for loading control and larvae cultured at 258C. Relative protein expression levels are indicated below normalized to the wild-type dAtx2 overexpressing line (dAtx2 OE   lation on the development of adult macrochaetae. As with wild-type dAtx2 (Figure 3), expression of dAtx2 NES in the thoracic SOP cells causes no visible change in the number of macrochaetae in the adult thorax ( Figure 8Q and Figure 8S column-1). In contrast, expression of dAtx2 NLS induces a significant decrease in the number of macrochaetae, with an approximate twenty percent reduction in comparison to control animals ( Figure 8R and Figure 8S column-2). These results indicate that increasing the levels of nuclear dAtx2 mimics expanded Ataxin-1 in inducing loss of mechanoreceptors and reducing the levels of Sens protein.

Discussion
Here we report functional interactions between the proteins causing two distinct Spinocerebellar ataxias. We use a Drosophila model of SCA1 to show that wild-type dAtx2 (the fly homolog of the protein that when expanded causes SCA2) mediates, at least in part, neuronal degeneration caused by expanded Ataxin-1 (the protein triggering SCA1). Ataxin-1[82Q]-induced toxicity is worsened by increasing the levels of dAtx2. More significantly, decreasing the levels of dAtx2 suppresses expanded Ataxin-1-induced neuronal degeneration as shown in several independent assays. The suppression of Ataxin-1[82Q] phenotypes by partial loss of function of dAtx2 argues against a possible mechanism by which sequestration and depletion of Ataxin-2 contributes to expanded Ataxin-1-induced neurodegeneration. This is further supported by lack of cerebellar or other neuronal abnormalities in mice that are deficient for Ataxin-2 [46].
We find that the human expanded Ataxin-1 interacts with the dAtx2 and human Ataxin-2 proteins in co-AP assays. Furthermore, overexpressed Ataxin-1 pulls down endogenous hAtaxin-2 in cultured cells. These results suggest that Ataxin-1 and Ataxin-2 may be functional interactors in vivo. Consistent with this, we find that expanded Ataxin-1 induces accumulation of Ataxin-2 in the nucleus, where the two proteins localize in NIs both in Drosophila neurons and SCA1 human brain tissue. These are surprising observations since Ataxin-2 is normally a cytoplasmic protein both in humans and Drosophila. Interestingly, wild-type Ataxin-1 can cause neurotoxicity when overexpressed, although to a much lesser extent than expanded Ataxin-1 [25]. However, nuclear accumulation of dAtx2 is triggered by pathogenic but not wild-type forms of Ataxin-1, at least in detectable amounts. Taken together these data suggested that accumulation of Ataxin-2 in the nucleus contributes to the exacerbated toxicity of expanded Ataxin-1, and is an important mechanism of pathogenesis in SCA1. To investigate this hypothesis we targeted dAtx2 to the nucleus by means of an exogenous NLS signal. We find that dAtx2 NLS is sufficient to cause a dramatic increase of its toxicity, when compared to either wild-type dAtx2 or dAtx2 with an exogenous nuclear export signal (dAtx2 NES ) expressed at similar levels.
To further test the hypothesis that nuclear accumulation of Ataxin-2 contributes to neurodegeneration caused by expanded Ataxin-1 we investigated Sens levels. Sens and its murine orthologue Gfi1 are proneural factors whose levels are decreased in the presence of expanded Ataxin-1 [18]; thus providing a molecular readout for the neurotoxicity of Ataxin-1. In Drosophila, reduction of Sens levels leads to the loss of mechanoreceptors [18], so we monitored Sens in the context of flies expressing either dAtx2 NLS or dAtx2 NES but not carrying the Ataxin-1[82Q] transgene. We find that nuclear targeted, but not cytoplasmic, dAtx2 mimics both the Sens reduction and mechanoreceptor loss phenotypes caused by Ataxin-1[82Q].
Expanded Ataxin-2 accumulates both in the cytoplasm and the nuclei of SCA2 postmortem brains [2,[47][48][49]. In mouse and cell culture models of SCA2, expanded Ataxin-2 accumulates in the cytoplasm and its nuclear accumulation is not necessary to induce toxicity [44,45]. However, nuclear accumulation of expanded Ataxin-2 also occurs in cultured cells [45], and is consistently observed in human SCA2 postmortem brainstem neurons [2,[47][48][49]. These observations suggest that both nuclear and cytoplasmic mechanisms of pathogenesis contribute to neurodegeneration in SCA2, as it is known to occur in other polyglutamine diseases like HD and SCA3 [50][51][52]. One possibility is that Ataxin-2 shuttles between the nucleus and the cytoplasm although the protein is normally detected only in the cytoplasm. Our data show that accumulation of dAtx2 in the nucleus is more harmful than in the cytoplasm. Thus, neurons with nuclear Ataxin-2 in SCA2 patients may be relatively more compromised than neurons where Ataxin-2 accumulates in the cytoplasm. In agreement with this possibility, expanded Ataxin-2 is found in the nuclei of pontine neurons of SCA2 brains, one of the neuronal groups and brain regions with prominent degeneration in SCA2 [2,[47][48][49].
Reducing Ataxin-2 levels suppresses expanded Ataxin-1 toxicity, strongly arguing against a mechanism of pathogenesis by loss of function of Ataxin-2 in the cytoplasm. Studies of the normal function of Ataxin-2 and its yeast [29,30], C. elegans [31], and Drosophila [32,33] homologs suggest a role in translational regulation. Thus, an attractive possibility is that Ataxin-1 [82Q] requires dAtx2 to impair Sens translation and induce the loss of mechanoreceptors. Consistent with this hypothesis is the finding that partial loss of function of dAtx2 suppresses the loss of mechanoreceptors phenotype caused by expanded Ataxin-1.
The data described here uncover unexpected functional interactions between proteins involved in two different SCAs. Nuclear accumulation of Ataxin-2, normally a cytoplasmic protein, is a common denominator of SCA1 and SCA2, and leads to reduced levels of at least one important proneural factor; i.e. Sens, whose mammalian orthologue Gfi1 is required for Purkinje cell survival [18]. Thus neuronal degeneration may take place through common mechanisms in different ataxias, and one of these mechanisms may involve the abnormal accumulation of Ataxin-2 in neuronal nuclei.

Methods
Drosophila strains. The cDNA GH27029 containing dAtx2 was obtained from the BDGP repository. SV40 nuclear localization signal and PKI nuclear export signal were engineered on the 39 end of dAtx2 cDNA by PCR. Both constructs were then subcloned first in pGEM t -T (Promega) and then in a previously generated pUAST-flag expression vector [53,54]. The Drosophila transgenic lines UAS-dAtx2 NES and UAS-dAtx2 NLS were obtained by injecting both constructs following standard procedures. EP(3)3145 was obtained from the Szeged Drosophila Stock Center in Hungary. The wild type UAS-dAtx2 (dAtx2OE) and mutant dAtx2-X1 lines [33] were kindly provided by Dr. Pallanck, L.J.. The UAS-SCA182Q, UAS-SCA130Q and UAS-SCA12Q lines have been previously described [7,25]. N-Htt128Q flies have been previously described [7,43]. All other Drosophila strains were obtained from the Bloomington Drosophila Stock Center at Indiana University.
Scanning electron microscopy and retinal paraffin sections. We used previously published procedures [7,25].
Scoring of macrochaetae in adult thoraxes. Flies were raised at either 258C for low Ataxin-1[82Q] expression levels or 278C for high Ataxin-1[82Q] expression levels. Flies were collected at day 1 and the number of macrochaetae per thorax of same sex flies was counted for 20 animals per genotype. The percentage of lost macrochaetae over a total of 26 was calculated, and the average per genotype was plotted in the chart.
Motor performance assay. Between 25 and 30 adult females per genotype are collected for periods no longer than 24 hours. Flies are transferred to vials containing new food every day. The assay is carried out in an empty vial. The vial is tapped so all flies fall to the bottom then we score flies that climb past a line 5cm high in 18 seconds, and this procedure is repeated ten times for each day shown in the chart. The average percentage of flies climbing per day is then calculated and plotted in Microsoft Excel. Experiments are always performed at the same time in the day to ensure no circadian rhythm effects. Four replicas (25,n,30) are analyzed per genotype. Motor performance in each replica is measured until all animals failed to reach the 5 cm line.
Quantification of Sens signal. Sens quantification in the SOP and bristle precursor cells of the wing margins was carried as follows: animals were raised at 258C. immunofluorescence was done on wing discs of the different genotypes (gp anti-Sens 1:1000, Hugo Bellen). Confocal images were obtained throughout the thickness of the wing margin (20 sections per wing disc with 2 lm interval) and then the signal was stacked and summed. Quantification was carried out by selecting the area covered by each wing margin (dorsal or ventral) separately and calculating fluorescence and area. Data corresponding to 20 wing discs per genotype was analyzed by Tukey-Kramer HSD and plotted in a chart.