Polyglutamine Genes Interact to Modulate the Severity and Progression of Neurodegeneration in Drosophila

The expansion of polyglutamine tracts in a variety of proteins causes devastating, dominantly inherited neurodegenerative diseases, including six forms of spinal cerebellar ataxia (SCA). Although a polyglutamine expansion encoded in a single allele of each of the responsible genes is sufficient for the onset of each disease, clinical observations suggest that interactions between these genes may affect disease progression. In a screen for modifiers of neurodegeneration due to SCA3 in Drosophila, we isolated atx2, the fly ortholog of the human gene that causes a related ataxia, SCA2. We show that the normal activity of Ataxin-2 (Atx2) is critical for SCA3 degeneration and that Atx2 activity hastens the onset of nuclear inclusions associated with SCA3. These activities depend on a conserved protein interaction domain of Atx2, the PAM2 motif, which mediates binding of cytoplasmic poly(A)-binding protein (PABP). We show here that PABP also influences SCA3-associated neurodegeneration. These studies indicate that the toxicity of one polyglutamine disease protein can be dramatically modulated by the normal activity of another. We propose that functional links between these genes are critical to disease severity and progression, such that therapeutics for one disease may be applicable to others.


Introduction
The polyglutamine diseases are caused by the expansion of a CAG repeat encoding glutamine within the open reading frames of at least nine genes [1]. The disease state is thought to be due to a misfolded conformation of the pathogenic protein (termed an Ataxin in most forms of spinal cerebellar ataxia [SCAs]), which accumulates intracellularly in ubiquitinated inclusions along with chaperones and subunits of the proteasome [2]. Among the SCAs, there is a remarkable overlap of symptoms, including progressive loss of gait and limb muscle coordination, with neurodegeneration in the cerebellum and often select brainstem nuclei [3].
In addition to the shared mechanism of mutation and clinical phenotypes, there are intriguing hints of interactions between the polyglutamine genes that cause different SCAs. For example, relatively long CAG repeat lengths (still within the normal range) in the causative gene for SCA6 are correlated with an early age of onset of SCA2 [4]. In addition, there are a number of such observations suggesting links between SCA2 and SCA3 diseases. The severity of facial muscle twitch in patients with SCA3 is correlated with the length of the CAG repeat in the normal alleles of the SCA2 gene ATXN2 [5]. Both SCA2 and SCA3 are unusual among the dominantly inherited ataxias in that they can manifest with parkinsonism [6]. Finally, the normal Ataxin-2 (Atx2) protein can be detected in the pathogenic inclusions of SCA3 patients, and, likewise, normal Atx3 protein localizes to the inclusions formed in SCA2 patients [7].
Several SCAs have been successfully modeled in Drosophila [8][9][10], an organism ideally suited to screen for genes that can modify the effects of protein toxicity and neurodegeneration. Moreover, up-regulation of atx2, the Drosophila ortholog of the human gene that causes SCA2 disease, has been shown to enhance the toxicity of human disease forms of SCA1 and SCA3 in flies [11]. Many genes that modify SCA1 toxicity in the fly, including atx2, have been shown to encode proteins that interact directly with pathogenic Atx1, and further, these proteins form an interaction network that includes Atx3 [12]. Taken together with the clinical data, this work suggests that interactions between CAG-containing genes may occur that influence disease.
Despite these findings, the molecular activities of the Atx2 and Atx3 proteins do not indicate obvious functional links between these two particular CAG repeat genes. Atx2 is a cytoplasmic protein with an RNA-binding Lsm domain and a PAM2 motif that binds specifically to the PABC domain of proteins such as the cytoplasmic poly(A)-binding protein (PABP) [13,14]. PABP and Atx2 directly interact in Caenorhabditis elegans [15]; and in flies and cultured cells, PABP and Atx2 also physically associate and co-sediment with polyribosomes [16]. In contrast, Atx3 shuttles between the cytoplasm and nucleus, contains ubiquitin interaction motifs and de-ubiquitination activity [17][18][19][20], and is normally neuroprotective, an activity that depends on a fully functional proteasome [21].
In a screen in Drosophila for modifiers of degeneration induced by the human pathogenic protein causing SCA3, we identified atx2. Up-regulation of atx2 synergistically enhanced SCA3 degeneration, and strikingly, we found that the endogenous activity of atx2 modulates progression of neurodegeneration induced by pathogenic Atx3. Thus, toxicity of the polyglutamine disease protein Atx3 is critically dependent upon the normal activity of atx2, a second gene associated with polyglutamine disease in humans. These findings under-score the power of the Drosophila system, and provide a foundation for further molecular insight into human genetic and clinical studies.

Up-Regulation of Atx2 Synergistically Enhances Atx3-Induced Neurodegeneration
Expression of pathogenic human Atx3 from the transgene UAS-SCA3trQ78 in the developing eye under the control of the Gmr-Gal4 driver causes degeneration [8]. We identified a transposon insertion in atx2 conferring Gal4-dependent expression that greatly enhanced this toxicity (also reported in [11]; see Materials and Methods for details of the screen). To verify that up-regulation of atx2 was responsible for the effect, we coexpressed a UAS-Atx2 transgene [22] together with pathogenic Atx3 and confirmed strongly enhanced eye degeneration with loss of pigmentation and severe collapse of the retina (Figure 1). Coexpression of Atx2 with a nonpathogenic Atx3 construct bearing a normal-length polyglutamine repeat had no effect, such that the eyes looked the same as those expressing only Atx2, as in Figure 1C (unpublished data).
Since Atx2 expression can cause developmental effects [22], we further examined neuronal integrity in an adult-onset situation using the driver Rhodopsin-1-Gal4 (Rh1-Gal4), active selectively in differentiated photoreceptor neurons (PR). Adult flies expressing pathogenic Atx3, normal fly Atx2, or both were born with the normal complement of seven visible PR per unit eye, or ommatidium, as seen in pseudopupil preparations (see Materials and Methods). Strong expression of SCA3trQ78 induced progressive degeneration, from seven PR declining to 3.7 6 0.22 PR by 18 d (Figure 2A and 2E). Flies expressing only Atx2 also underwent retinal degener-

Author Summary
Six forms of spinal cerebellar ataxia (SCA1, À2, À3, À6, À7, and À17) are caused by dominant mutations in the respective genes. Patients suffering from these different forms of SCA have similar symptoms of progressive adult-onset neurodegeneration, and the same causative mutation, a CAG repeat expansion encoding an expanded run of polyglutamine. Although the affected proteins are distinct and share no sequence similarity beyond the polyglutamine domains, clinical and other observations hint at interactions between the genes that cause different forms of SCA. Using Drosophila as a model for human disease, we now detail an interaction between the genes associated with SCA3 and SCA2. We find that toxicity and neurodegeneration induced by pathogenic forms of SCA3 depend on the normal activity of the fly counterpart of the gene associated with SCA2, ataxin-2. This interaction depends on a conserved protein-interaction motif of Ataxin-2, and a protein that binds this motif, cytoplasmic poly(A)-binding protein (PABP), also modulates SCA3 degeneration. These results suggest that the normal roles of Ataxin-2 and PABP, potentially to regulate the translation of select target mRNAs, are critical to SCA3 disease. These studies also highlight how a fly model can serve to enhance and extend intriguing clinical findings of the human disease.
ation, but to a lesser extent (5.9 6 0.13 PR at 18 d; Figure 2B and 2F; and see below). Coexpression of Atx2 with Atx3, however, resulted in dramatically enhanced degeneration, such that even by 6 d, there were only 4.7 6 0.15 PR ( Figure  2C) compared to the near-normal eyes resulting from expression of either protein alone (Figure 2A, 2B, and 2I). By 18 d, these flies had undergone nearly complete PR loss ( Figure 2G). At either 6 d or 18 d, PR loss due to coexpression was greater than the sum of the deficits caused by each transgene alone, indicating a synergistic interaction.
These experiments were performed with a truncated version of pathogenic Atx3 [8]. In mouse models and in human SCA3 disease, the full-length pathogenic protein undergoes proteolysis, yielding a similarly sized fragment containing the polyglutamine tract that is thought to contribute to disease [23]. To determine whether Atx2 interacted with the intact as well as truncated Atx3 protein, we coexpressed Atx2 with full-length pathogenic Atx3 (SCA3Q84 [21]). Under conditions where Atx2 or SCA3Q84 alone had mild effects (6.3 6 0.16 PR and 6.97 6 0.02 PR, respectively), coexpression of Atx2 with SCA3Q84 resulted in strikingly severe degeneration (2.9 6 0.30 PR; Figure 2J). Thus, synergistic enhancement of degeneration by Atx2 was not limited to truncated versions of pathogenic Atx3, but extended to the full-length protein. To further determine the specificity of Atx2, we coexpressed Atx2 with two different Huntington disease transgenes [24,25]. Atx2 dramatically enhanced toxicity of an exon-1 Huntingtin (Htt) protein, but had no effect on toxicity of a longer Htt fragment ( Figure  S1). Cleavage of the Huntingtin protein contributes to disease [26,27]; Atx2 may play a role by selectively promoting toxicity of small Htt fragments.
For some dominantly acting disease proteins, up-regulation of the normal version of the protein is sufficient to cause neurodegeneration clinically (for example, [28]) and in animal models [9,29,30]. We were therefore interested in the extent to which normal Atx2 could induce degeneration in the adult in the absence of developmental effects. The modest photoreceptor loss due to expression of Atx2 seen in Figure 2B and 2F was more striking with stronger expression of Atx2 from the transgenic line UAS-Atx2(s) [22]; even 1-d flies had mild but significant loss of PR neurons (6.8 6 0.06, vs. 7.0 6 0.01 PR in flies expressing SCA3trQ78 only; p ¼ 0.009) that progressed rapidly to 1.5 6 0.08 PR by 18 d (Figure 2D and 2H). Therefore, up-regulation of normal Atx2 without a polyglutamine expansion causes severe, progressive degeneration.

Reduction of Atx2 Activity Mitigates SCA3 Degeneration
Our findings demonstrated that up-regulation of Atx2 activity modulates SCA3 pathogenesis. To determine whether Atx2 normally contributes to disease, we examined the effects of reducing endogenous Atx2, which is expressed abundantly in the adult fly and is enriched in the head and brain [31]. Flies lacking 50% of normal Atx2 (heterozygous for a null atx2 allele) were examined for modulated SCA3trQ78 toxicity in pseudopupil preparations with adult-onset expression. At 12 d, when Atx3 normally caused photoreceptor loss to 6.4 6 0.08 PR, reduction of Atx2 by 50% significantly mitigated degeneration (6.9 6 0.03 PR, Figure 3A). This finding was confirmed with a deletion removing atx2 (unpublished data). Similar studies showed that flies expressing full-length pathogenic Atx3 also showed mitigated PR loss when heterozygous for an atx2 deletion ( Figure S2).
To further address the role of endogenous atx2, we analyzed neuronal integrity using the MARCM method [32], whereby we could analyze small clones of neurons lacking atx2 activity. Neuronal clones selectively marked with green fluorescent protein (GFP) ( Figure 3B) were induced by mitotic recombination and were either atx2 þ or homozygous mutant for atx2. When the UAS-SCA3trQ78 transgene was present, it was expressed only in neuronal clones along with the GFP marker. GFP þ sensory neurons in the anterior wing margin were examined in young adult flies less than 8 h old and over the subsequent 6 d, following marked neurons in individual animals. In the absence of pathogenic Atx3, GFP þ neurons that were atx2 þ or homozygous for a null allele of atx2 were retained at a similar rate: 83% of the former (unpublished data) and 75% of the latter ( Figure 3C) remained at 6 d. In contrast, of GFP þ neurons expressing SCA3trQ78 that were present in young animals, only 13% were still detectable at 3 d, with all neurons lost by 6 d (Figure 3C and 3D). Strikingly, SCA3trQ78-expressing neurons were dramatically retained if they also lacked Atx2 function: 77% of these neurons were present at 3 d and 40% were retained at 6 d ( Figure 3C and 3E). Taken together, these studies indicate that normal endogenous Atx2 activity facilitates Atx3-dependent toxicity, with up-regulation of Atx2 enhancing and loss of Atx2 function dramatically slowing the progression of neurodegeneration caused by pathogenic Atx3.

Up-Regulation of Atx2 Accelerates Inclusion Formation by Pathogenic Atx3
The formation of nuclear inclusions containing pathogenic polyglutamine protein is a hallmark of polyglutamine diseases, with the size and number of inclusions typically correlating with disease severity in animal models [1][2][3]. In retinal sections of animals expressing pathogenic Atx3 driven by Rh1-Gal4, inclusions were sparse at 24 h but prominent by 4 d (Figure 4A and 4B). Up-regulation of Atx2-which enhances Atx3 toxicity (Figures 1 and 2)-also accelerated inclusion formation, such that at 24 h, Atx3-positive inclusions were now prominent ( Figure 4E). By 4 d, the inclusions were similar to those of animals expressing pathogenic Atx3 alone (compare Figure 4F to 4B).
Immunoblot experiments showed similar results ( Figure  4J). At the 24-h time point, coexpression of normal Atx2 with pathogenic Atx3 caused the early appearance of SDSinsoluble Atx3 complexes that remain within the stacking gel (compare lanes 1 and 2), presumably reflecting protein accumulations that correlate with the early appearance of nuclear inclusions in cryosections. At the latter 4-d time point, insoluble Atx3 complexes were present in extracts from both groups (lanes 4 and 5). Differences in the inclusions and high molecular weight complexes were due to post-transcriptional causes, since Atx2 coexpression did not affect the transcript levels of the SCA3trQ78 transgene ( Figure 4I).
Endogenous Atx2 protein was normally cytoplasmic throughout the retina. By 4 d, however, endogenous Atx2 was detectable in most Atx3-positive nuclear inclusions ( Figures 4B-4D and S3), reminiscent of the Atx2 localization to inclusions in SCA3 patients [7]. Up-regulated Atx2 protein was not selectively recruited to inclusions, but remained punctate along the length of photoreceptor neurons; this pattern did not change over time or with coexpression of SCA3trQ78 ( Figure S4). Notably, in both SCA2 patients and a SCA2 mouse model, pathogenic Atx2 with expanded polyglutamine is present in a similar cytoplasmic, punctate pattern [33]. Whereas inclusions of pathogenic Atx3 co-label for chaperones such as Hsp70, no such inclusions were observed in flies expressing only high levels of Atx2 ( Figure  S4). Western blot analysis confirmed that Atx2 protein levels were not affected by pathogenic Atx3 (unpublished data). In summary, endogenous Atx2 is recruited to nuclear inclusions formed by pathogenic Atx3, as in SCA3 patients.

The Interaction between Atx2 and Pathogenic Atx3 Is Dependent on the PAM2 Motif of Atx2
To provide insight into the mechanism by which Atx2 modulates polyQ toxicity, we determined whether the conserved 12-amino acid PAM2 motif of the Atx2 protein was required for enhancement of SCA3 toxicity. The PAM2 motif is required for Atx2 to associate with PABP, and together with the RNA-binding Lsm domain, it mediates the colocalization of Atx2 with PABP in polyribosomes [16]. We tested whether a construct lacking the conserved PAM2 domain, UAS-Atx2DP [16], could modulate Atx3 toxicity. Atx2DP retains partial Atx2 function, since it delays embryonic lethality caused by complete loss of atx2 function [16]. The Atx2DP transgenic line expressed the protein at a level comparable to that of the strongly expressing Atx2(s) line ( Figure S5). Expression of Atx2DP driven by Gmr-Gal4 caused a disrupted eye phenotype distinct from that caused by up-regulation of normal Atx2: black patches were present on the surface, and the underlying retina was disorganized although intact, showing no evidence of degeneration ( Figure  5A). In striking contrast to normal Atx2 protein, coexpression of Atx2DP failed to enhance SCA3trQ78 toxicity: the flies showed an eye phenotype identical to that caused by Atx2DP alone (Figure 5B; compare to Figure 1D).
We confirmed this finding by examining interactions between Atx2DP and pathogenic Atx3 in the adult, uncomplicated by any developmental effects. Coexpression of Atx2DP and SCA3trQ78 driven by Rh1-Gal4 caused a mild adult-onset photoreceptor loss at 6 d (6.7 6 0.06 PR; Figure  5D) that was similar to that caused by SCA3trQ78 alone (6.9 6 0.02 PR; see Figure 2A), and strikingly less severe than that caused by coexpression of SCA3trQ78 with normal Atx2 (p , 0.001; see Figure 2C). The accelerated appearance of inclusions caused by up-regulated normal Atx2 was also dependent on the PAM2 domain, since coexpression of Atx2DP with SCA3trQ78 did not alter the time course of inclusion formation (compare Figure 4G to 4E). Moreover, coexpression of Atx2DP did not cause the early appearance of SDS-insoluble complexes containing Atx3 ( Figure 4J, com- pare lane 3 to lane 2). Taken together, these data show that the PAM2 motif of Atx2 is required for the enhancement of pathogenic Atx3-dependent degeneration as well as for the early appearance of Atx3 inclusions and insoluble complexes. Adult-onset expression of Atx2 lacking the PAM2 motif alone also did not cause neurodegeneration: even at 18 d, flies expressing Atx2DP had a normal retinal structure (7.0 6 0.01 PR, Figure 5C; compare to normal Atx2 expressed at an equivalent level, which results in 1.5 6 0.08 PR, Figure 2H). Thus the PAM2 motif is required for the inherent toxicity of Atx2 when it is up-regulated in the adult eye.

The Atx2-Binding Protein PABP Modulates Neurodegeneration
PABP is the only known protein to date that interacts directly with Atx2 through the PAM2 motif [13,14,16]; therefore, given the important role of the PAM2 motif described above, we asked if PABP played a role in SCA3 neurodegeneration. Heterozygosity for the available pabp allele had no effect on Atx3 toxicity, although this allele is unlikely to be a complete loss of function [34]. We then tested a deletion chromosome that removed the pabp gene, comparing to appropriate control lines (see Materials and Methods). Flies expressing pathogenic Atx3 that were heterozygous for this deletion showed dramatically enhanced photoreceptor loss (1.6 6 0.07 PR, compared to the control 3.5 6 0.16 PR; Figure 6A). Control experiments confirmed that the deletion alone, in the absence of pathogenic Atx3, did not cause neurodegeneration (unpublished data). In contrast to the loss-of-function situation, overexpression of PABP significantly suppressed neurodegeneration (from 4.2 6 0.18 PR to 6.0 6 0.08 PR; Figure 6A). These observations indicated that PABP has the opposite activity as Atx2 with respect to Atx3-dependent neurodegeneration: whereas Atx2 enhances the toxicity of Atx3, PABP is protective.
We then tested whether PABP could modulate the degeneration induced by strong expression of Atx2 as shown in Figure 2D and 2H. Decreased PABP function enhanced Atx2-dependent photoreceptor loss (from 1.7 6 0.09 PR to 1.3 6 0.06 PR); likewise, up-regulation of PABP protected against photoreceptor degeneration (2.9 6 0.16 PR to 5.1 6 0.19 PR; Figure 6B). These studies suggest that the toxicity of Atx2 is mitigated by physical association with PABP, and they are consistent with PABP also playing a crucial role in the Atx2-Atx3 interaction. Together with our results demonstrating the crucial role of the PAM2 motif, these data highlight the importance of the normal biological activity of Atx2 and of PABP in modulating the toxicity of pathogenic Atx3.

Discussion
Our studies in Drosophila reveal that the toxicity of pathogenic human Atx3 is critically dependent on Atx2 activity. Reduction of endogenous Atx2 function mitigated Atx3-induced neurodegeneration, and up-regulation of Atx2 synergistically enhanced degeneration. We also reveal the roles in neural integrity played by the non-polyglutamine PAM2 motif of Atx2 and by PABP, which binds to Atx2 via this motif. These data are consistent with and expand upon clinical findings suggesting interactions between Atx2 and Atx3 in human disease [5][6][7]. In the fly, endogenous Atx2 colocalized with pathogenic Atx3 in inclusions, as seen in human patients [7], with up-regulation of Atx2 enhancing Atx3 toxicity concomitant with a faster onset of inclusions and of SDSinsoluble complexes. These findings suggest that therapeutic approaches to modulate Atx2 activity may be effective against multiple disease situations, including SCA2 and SCA3.
Interestingly, we find that normal Atx2 is toxic, causing degeneration when up-regulated. Previous animal models have demonstrated that normal protein products associated with SCA1 and Parkinson's disease-Ataxin-1 and a-Synuclein, respectively-are also toxic when expressed at sufficiently high levels [9,29,30]. Expansion of the polyglutamine domain in Ataxin-1 or Parkinson disease-associated missense mutations of a-Synuclein presumably lead to increased levels of the respective proteins, sufficiently high to elicit disease. Up-regulation of Drosophila Atx2 may cause degeneration for similar reasons. Our studies further reveal that neuronal toxicity of Atx2 depends on its PAM2 motif-an observation with an interesting parallel to Ataxin-1, the protein that causes SCA1: an expanded polyglutamine repeat in Ataxin-1 is not sufficient to cause neurodegeneration in mouse models for SCA1, but rather pathogenic Ataxin-1 also requires its AXH domain to cause disease [35].
The importance of the PAM2 motif for Atx29s toxicity and for the enhancement of Atx3 toxicity suggests a clue to the mechanism of the interaction. The PAM2 motif has been shown to bind specifically to the PABC domain [13], with PABP being currently the only known PABC-containing protein that interacts with Atx2. PABP is a ubiquitously expressed and essential protein that binds to the polyadenylated tails of mRNAs and is required for their translation. Furthermore, biochemical and genetic data support an interaction between Atx2 and PABP across many species [15,16,36]. Data from C. elegans indicate that loss of Atx2 can result in misregulated translation [15], and in yeast Atx2 negatively regulates PABP [36]. Consistent with these findings, we have shown that Atx2 and PABP have opposing activities in modulating the progression of SCA3 toxicity in flies.
Protein interaction studies indicate that Atx2 and Atx3 do not interact directly; in a survey of the interaction network of ataxia-associated proteins, Atx2 and Atx3 were separated by four nodes [12]. However, the known function of PABP and the role of the PAM2 motif in localizing Atx2 to polyribo-somes [16] together indicate that Atx2 and PABP modulate translation of specific transcripts. Since Atx2 is sufficient to cause neurodegeneration in the absence of pathogenic Atx3 (see Figure 2D and 2H), Atx3 mRNAs cannot be the sole target of Atx2-PABP interactions, and additional transcript targets must be critical to normal neuronal integrity.
Our experiments in Drosophila demonstrate that the fly provides an outstanding complement to clinical observations and to vertebrate disease models. In this case, the fly has highlighted the significance of intriguing interactions between the genes that cause SCA2 and SCA3 diseases that can be supported by molecular and genetic findings. More specifically, these data indicate striking crosstalk between the pathways of normal Atx2 function and pathogenic Atx3 activity. Further understanding of both the Atx2 and Atx3 pathways may reveal insight into maintenance of neuronal integrity in a number of distinct disease situations.
External eye microscopy. For eye pictures in Figures 1 and 5, we used a Leica Z16-Apo A motorized zoom microscope system with DFC420 digital camera and Leica Application Suite Montage module software (Leica Microsystems).
Pseudopupil preparations. Pseudopupil preparations were performed as described [40] with w þ male flies. The Rh1-Gal4 driver is active in the outer six PR neurons of each unit eye (ommatidium). As an indicator of the number of PR neurons, rhabdomeres were counted in 10-20 ommatidia of n ¼ 10 flies of each genotype. Mean PR counts 6 standard error of the mean (SEM) per ommatidium are indicated in the figures and text. Statistical analysis was performed with Prism software. The two-tailed Mann Whitney test was used to calculate the p-value when comparing the results of pseudopupil preparations between two groups. In experiments with three groups (in Figures 2I, 2J, and S2, and for a comparison of phenotypes due to SCA3trQ78 alone, with Atx2 or with Atx2DP, stated in the text discussing Figure 5D), the Kruskal-Wallis test was performed followed by Dunn post-tests.
Quantitative real-time PCR. Quantitative real-time PCR was performed in triplicate on each cDNA sample with an Applied Biosystems 7500 Fast system. Duplex PCRs were performed in single wells with human ATXN3 forward primer 59CAGGACAGAGTTCACATCCATGT, reverse primer 59GCCTTACCTAGATCACTCCCAAGT, and FAM-labeled Taqman probe 59CCAGCCACCAGTTCAG (Applied Biosystems assay by design), and with human 18s rRNA as the internal control (the Drosophila sequence is identical; primers and VIC-labeled probe, Applied Biosystems #4319413E). Template was random-primed cDNA prepared with Superscript III (Invitrogen) from RNA made from 80-115 male heads (Qiagen RNeasy). Before PCR, cDNA was treated with Turbo DNase (Ambion) to remove genomic DNA. Quantitative PCR results were analyzed with the two-tailed t-test (Prism software).
Marked wing margin clones. Marked wing margin clones were induced using previously developed reagents and procedures [32]. MARCM stock females were crossed to experimental FRT82B males and allowed to lay eggs for approximately 18 h, at which point parental flies were cleared from the vials. Progeny were of genotype Elav-Gal4 hsFLP UAS-mcd8GFP/Y; [UAS-SCA3trQ78(s)]/þ; FRT82B tubP-Gal80/FRT82B [atx2 X1 ]. The presence of UAS-SCA3trQ78 or atx2 X1 varied between experimental groups, as indicated by brackets. Mitotic recombination between the FRT sites was induced in progeny larvae 3 d later by submerging vials for 50 min in a 37 8C water bath. About half of the resulting daughter cells lacked Gal80 (a Gal4 inhibitor), and if they or their progeny were neurons, the neuronal driver Elav-Gal4 was free to drive expression of the GFP marker and (if present) of SCA3trQ78 as well. If the atx2 X1 mutation was present on the FRT82B chromosome as indicated above, these daughter cells and their progeny were homozygous for this null atx2 allele.