Genetic and Chemical Modifiers of a CUG Toxicity Model in Drosophila

Non-coding CUG repeat expansions interfere with the activity of human Muscleblind-like (MBNL) proteins contributing to myotonic dystrophy 1 (DM1). To understand this toxic RNA gain-of-function mechanism we developed a Drosophila model expressing 60 pure and 480 interrupted CUG repeats in the context of a non-translatable RNA. These flies reproduced aspects of the DM1 pathology, most notably nuclear accumulation of CUG transcripts, muscle degeneration, splicing misregulation, and diminished Muscleblind function in vivo. Reduced Muscleblind activity was evident from the sensitivity of CUG-induced phenotypes to a decrease in muscleblind genetic dosage and rescue by MBNL1 expression, and further supported by the co-localization of Muscleblind and CUG repeat RNA in ribonuclear foci. Targeted expression of CUG repeats to the developing eye and brain mushroom bodies was toxic leading to rough eyes and semilethality, respectively. These phenotypes were utilized to identify genetic and chemical modifiers of the CUG-induced toxicity. 15 genetic modifiers of the rough eye phenotype were isolated. These genes identify putative cellular processes unknown to be altered by CUG repeat RNA, and they include mRNA export factor Aly, apoptosis inhibitor Thread, chromatin remodelling factor Nurf-38, and extracellular matrix structural component Viking. Ten chemical compounds suppressed the semilethal phenotype. These compounds significantly improved viability of CUG expressing flies and included non-steroidal anti-inflammatory agents (ketoprofen), muscarinic, cholinergic and histamine receptor inhibitors (orphenadrine), and drugs that can affect sodium and calcium metabolism such as clenbuterol and spironolactone. These findings provide new insights into the DM1 phenotype, and suggest novel candidates for DM1 treatments.


Introduction
Myotonic dystrophy 1 (DM1) is an autosomal dominant neuromuscular disease involving the expansion of unstable CTG repeats in the 39 untranslated region (UTR) of the DM protein kinase (DMPK) gene. DM1 is multisystemic and characteristic features include myotonia, muscular dystrophy, iridescent cataracts, cardiac arrhythmias, and signs of neuropathology [1]. A biochemical hallmark of DM1 is misregulated alternative splicing of specific skeletal muscle, heart and brain pre-mRNAs, which explain defined DM1 symptoms such as myotonia (reviewed in [2]).
In mice, expression of 250 CUG repeats within a heterologous RNA gives rise to DM1-like phenotypes thus demonstrating that expanded CUG repeat transcripts are themselves toxic to cells [3]. Results in Drosophila, however, are less clear cut. Expression of 162 pure CTG repeats in the context of the 39UTR of a Green Fluorescent Protein (GFP) reporter gene has been reported not to cause signs of pathology [4] whereas larger, interrupted, CTG repeats induced muscle degeneration [5]. Several RNA binding proteins, most notably human Muscleblind-like proteins MBNL1, MBNL2 and MBNL3, are sequestered by mutant DMPK transcripts. MBNL1 proteins co-localize with distinctive CUG ribonuclear foci within muscle and neuron nuclei in DM1 patients [6][7][8]. Drosophila model flies, though, demonstrate that ribonuclear foci are not pathogenic per se. RNA containing 162 CUG repeats accumulates in numerous nuclear foci together with Drosophila Muscleblind, but no evident pathogenic phenotype is detected [4]. DM1-associated defects are remarkably similar to those observed in Mbnl1 knockout mice and include myotonia, ocular cataracts, histological abnormalities, and the abnormal use of specific alternative exons [9], [10]. muscleblind (mbl) loss-of-function mutations in Drosophila provide additional examples of DM1-like phenotypes such as missplicing of the Z-band-associated transcripts a-actinin and CG30084 [11], [12].
Mbnl1 regulates a fetal to postnatal developmental switch that controls the splicing pattern of a set of murine skeletal muscle transcripts [10]. CUG-binding protein 1 (CUG-BP1) forms an RNA-dependent complex with hnRNP H that antagonizes the activity of MBNL1 proteins [13]. Both CUG-BP1 and hnRNP H are upregulated in DM1 muscle cells [13], [14] thus further contributing to the splicing pathology. Significantly, rescue experiments in DM1 model mice demonstrate that loss of Mbnl1 function is the key event of missplicing and myotonia [15].
Additionally, overexpression of normal DMPK 39UTR mRNA in mice induced up-regulation of CUG-BP1 and also reproduced cardinal features of DM1 [16].
Great effort has been put to ameliorate myotonia and abnormal cardiac conduction in DM1, which are currently treated with sodium channel inhibitors (e.g. mexiletine). Muscular weakness and wasting, or daytime somnolence, however, show little or no improvement in pharmacological trials [17]. A number of genotoxic agents suppress somatic CTG expansion mosaicism in a cell culture model [18]. PC12 neuronal cell lines expressing 250 CTG repeats exhibit cell death after cell differentiation in vitro that is specifically inhibited by flavonoids [19].
We previously established that Drosophila Mbl and human MBNL1 proteins are functional homologs [20]. Haro et al. (2006) have reported that expression of 480 interrupted CTG repeats is toxic to Drosophila muscle cells, that CUG RNA and human MBNL1 accumulate into ribonuclear foci, and that human MBNL1 suppresses a CUG-induced eye phenotype. Here we describe similar transgenic flies in which we confirm muscle degeneration, ribonuclear formation, and genetic interaction with muscleblind gene dosage. We show that CUG expressing flies reproduce additional key features of the DM1 disease including misregulated alternative splicing of muscle genes, CUG tract length dependence of phenotypes, and CUGdependent central nervous system alterations. Furthermore, model flies were used in genetic screens and functional assays to identify new components of the pathogenesis pathway and chemical suppressors of DM1-like phenotypes, respectively.

Continued expression of expanded CUG repeats in Drosophila reduces lifespan and causes muscle degeneration
To understand the molecular and cellular mechanisms underlying the DM1 pathology we generated transgenic Drosophila lines that express 60 uninterrupted or 480 interrupted CUG repeats as a noncoding transcript under the control of the Gal4/UAS system. 480 repeats consisted of synthetic CTG repeats interrupted every 20 units by the CTCGA sequence (hereafter referred to as i(CTG) 480 ).The effect of expressing CUG repeat RNA in the Drosophila muscles or ubiquitously in the fly was studied with Myosin heavy chain (Mhc)-Gal4 and daughterless (da)-Gal4 lines, respectively. First we analyzed whether expression of i(CUG) 480 RNA in Drosophila tissues had any impact in their lifespan. Average survival of flies expressing i(CUG) 480 repeat RNA was lower than their corresponding control flies heterozygous for the UAS transgene or Gal4 driver. Furthermore, differences in survival curves were statistically significant except for the Mhc-Gal4.UAS-i(CTG) 480 and Mhc-Gal4/ + survival curves, possibly due to a dominant effect of the Mhc-Gal4 insertion, as this is a particularly weak stock, or the small population of flies tested (n = 40) (Figure 1). Continued expression of CUG repeat RNA in the fly musculature, or throughout the animal body, was therefore detrimental to fly survival.
Flies expressing i(CUG) 480 RNA in muscles additionally showed an age-dependent tendency to position wings upheld. These flies were flightless (n = 274) and showed alterations in indirect flight muscles (IFM), whereas those expressing (CUG) 60 RNA did not (0% flightless, n = 204). Both UAS-(CTG) 60 and UAS-i(CTG) 480 transgenes expressed repeat RNA to similar levels ( Figure S1A). 2-3 day old flies expressing i(CUG) 480 RNA developed muscle histopathology, including vacuolization and reduction in fiber size (Figure 2A-F). We measured cross-sectional area of dorso longitudinal muscle 45e ( Figure 2G). Average size of muscle 45e decreased to approximately 45% of normal when expressing 480 CUG repeat transcripts. The phenotype was degenerative as 38day old flies had smaller IFM packages, muscles were occasionally missing, and vacuoles increased in average volume ( Figure 2H). (CUG) 60 RNA did not appreciably affect muscle organization.
Degeneration of the pigmentary retina and loss of photoreceptor neurons has been described in DM1 patients [1]. To investigate whether the retina of Drosophila was also susceptible to CUG repeatmediated toxicity, we expressed i(CUG) 480 transcripts ubiquitously in the eye-antennal imaginal disc under the control of the glass multiple reporter (gmr)-Gal4 line. These flies showed eyes that were smaller and acutely rough. Tangential and frontal sections revealed severe alterations in the retina, including detachment of subretinal cells, thinning of fenestrated membrane and lack of photoreceptor rhabdomeres ( Figure 2I-L). Expression of (CUG) 60 RNA under the same conditions did not appreciably affect eye morphology in tissue sections (data not shown). Thus, accumulation of CUG repeat RNA in Drosophila muscle and eye tissue produces degenerative phenotypes that are dependent on the CUG tract length.

CUG repeat RNA co-localizes with Muscleblind in nuclear foci
Nuclear inclusions containing CUG repeats and MBNL proteins are characteristic of DM1. We investigated whether i(CUG) 480 RNA similarly forms nuclear foci that include Drosophila Mbl. mbl encodes protein isoforms MblA, B, C and D, of which MblC has been shown to regulate alternative splicing [12]. We co-expressed i(CUG) 480 RNA and the MblC isoform fused to the GFP (MblC:GFP) under the control of a heat shock (hs)-Gal4 line. Simultaneous fluorescence detection in fly thorax sections showed nuclear co-localization of i(CUG) 480 RNA and MblC ( Figure 3A-C). This was not observed in controls expressing i(CUG) 480 RNA or the fusion protein alone. (CUG) 60 RNA did not form nuclear foci when targeted with Mhc-Gal4 to adult musculature (data not shown). Therefore, Drosophila MblC incorporates into expanded CUG repeat RNA-containing foci like its human MBNL counterparts. muscleblind dose modifies CUG toxicity phenotypes sevenless (sev)-Gal4 driven expression of i(CUG) 480 repeats (sev-Gal4.UAS-i(CTG) 480 ) disorganizes ommatidia and mechanosensory bristles, and reduces eye size, which generates an externally rough eye ( Figure 3H). Introduction of the weak mbl 7103 or strong hypomorphic mbl E27 mutant alleles in this genetic background did not significantly modify eye morphology ( Figure 3I, J; mbl E27 may reduce size slightly). However, a clear enhancement was observed in mbl 7103 /mbl E27 trans heterozygous flies simultaneously expressing i(CUG) 480 RNA ( Figure 3K). Conversely, targeted expression of human MBNL1 to Drosophila eye precursors expressing 480 interrupted CUG repeat transcripts strongly suppressed the rough eye phenotype, whereas expression of the unrelated GFP protein under the same conditions showed no effect ( Figure 3F, G). 60 CUG repeat RNA caused a milder effect on external eye morphology, only altering mechanosensory bristles ( Figure 3D, E). From these experiments we conclude that CUG repeat RNA compromises mbl function in vivo as similarly shown in DM1 model mice and patients [21], [15], [10].

CUG repeat RNA induces spliceopathy in Drosophila
Sequestration of MBNL1 correlates with missplicing events in DM1 patients. To assess whether long CUG repeat transcripts in the fly produce analogous alterations, we studied the splicing pattern of muscle genes CG30084, a described target of Mbl activity in embryos [11], and Drosophila troponinT (TnT) in embryos, pupae and adult flies expressing 60 CUG and 480 interrupted CUG repeat RNAs ( Figure 4). Missplicing of CG30084 pre-mRNA was conspicuous with a strong upregulation of reverse transcriptase (RT) PCR band E in adult flies expressing either 60 or 480 CUG repeat RNA ( Figure 4A, C; see also Figure S2). TnT was similarly affected. Two-day old pupae failed to show RT-PCR band D, which was not expressed in younger pupae (data not shown), when 480 interrupted CUG transcripts were targeted to the musculature and significantly lowered its levels with 60 CUG repeat RNA ( Figure 4B, D). Thus, CUG transcripts induce spliceopathy in the Drosophila musculature.

Dominant genetic modifiers of a CUG-induced rough eye phenotype
Once cardinal aspects of DM1 were confirmed in flies, we sought to identify new components of the pathogenic pathway. We performed a genetic screen of enhancer/suppressors of the sev-Gal4.UAS-i(CTG) 480 rough eye phenotype using a collection of 695 lethal P-element insertions and several candidate genes (Table 1, Figure 5A-H).
Some modifiers are regulators of gene expression. The suppressor cap-n-collar (cnc) encodes a bZIP protein involved in oocyte axis determination and head segment identity [22], [23]. Three Cnc protein isoforms have been described, of which CncC has been suggested to play a role in redox homeostasis [24]. We tested the ability of alleles cnc 03921 (disrupts all cnc isoforms), cnc EP3258 and cnc EP3633 (interrupts cncC) to modify the CUG toxicity phenotype. Only cnc 03921 dominantly suppressed the eye phenotype thus suggesting a limited or null implication of CncC in CUG toxicity. Halving the pyrophosphatase component of the Nucleosome remodelling factor (Nurf-38) improved eye morphology but did not suppress unrelated overexpression phenotypes in the eye (data not shown).
Additional modifiers identified genes and pathways not previously implicated in CUG-induced toxicity. Mutations in the regulators of cell adhesion and actin cytoskeleton coronin (coro) [25] and fear of intimacy (foi) [26] suppressed the phenotype. Reduction of the major structural component of basement membrane a2-chain type IV collagen (vkg k00236 ) enhanced the sev-Gal4.UAS-i(CTG) 480 phenotype.
Some modifiers of CUG toxicity control cell number. Csk negatively regulates the Src family of cytoplasmic tyrosine kinases. Mutations in Csk, which enlarge organs due to increased cell proliferation [27], suppressed i(CUG) 480 RNA toxicity (Csk j1D8 ). Mutations in the pro-and anti-cell death genes spinster and thread were suppressors and enhancers, respectively. Drosophila inhibitor of apoptosis protein (Diap), encoded by the thread (th) gene, showed complex interactions. Of the three alleles tested, loss-of-function th 4 and th 5 and gain-of-function th 6-3s , only th 4 strongly enhanced the CUG toxicity phenotype ( Figure 5F). Nevertheless, sev-Gal4 driven overexpression of th (th EP3308 ) in eyes simultaneously expressing i(CUG) 480 significantly suppressed the phenotype ( Figure 5G) whereas expression of a control GFP transgene (UAS-GFP) under comparable conditions did not modify eye morphology. A similar suppression was observed upon expression of the closely related Diap2 protein (gmr-diap2 fusion construct). Furthermore, th 4 and th 5 dominantly enhanced mblC overexpression in the Drosophila eye [28].
The sev-Gal4.UAS-i(CTG) 480 eye phenotype was enhanced by halving the genetic dose of the mRNA export factor Aly. Several observations indicate a close relationship between mRNA export factors and exon junction complex (EJC) components [29]. However, when we tested a lethal mutation in EJC core component tsunagi (tsu EP567 ) we found no effect. In summary we identified four cellular processes likely altered by CUG repeat RNA: gene transcription, cell adhesion, programmed cell death and export of nuclear transcripts.
Targeted expression of expanded CUG repeats to the mushroom bodies produces a temperature-sensitive pupal lethal phenotype Mushroom bodies (MBs) are brain structures involved in learning, sleep and memory. Because of the central nervous system involvement in DM1, we targeted expression of (CUG) 60   and i(CUG) 480 RNA to the Drosophila MBs ( Figure 5I). Expression with the X-linked 103Y-Gal4 driver was not deleterious at 25uC. However, an increase in the level of expression (by raising the temperature) originated a female-specific semilethal phenotype in F 1 mature pupae expressing 480 interrupted CUG repeat RNA ( Figure 5A). 28uC offered a threshold to CUG toxicity since only about 20% of emerged F 1 individuals were females (versus 50% expected) and some died during eclosion.
Reducing the genetic dose of mbl in a background expressing CUG repeats in the MBs (103Y-Gal4/+; mbl E27 /+; UASi(CTG) 480 /+) reduced the number of F 1 females six fold compared to control flies that expressed CUG repeats only (p,0.001; Figure  S3). Hence, targeted expression of i(CUG) 480 RNA to the MBs sensitizes flies to the genetic dose of mbl supporting that the expression of CUG RNA in neurons reproduces a pivotal aspect of the DM1 pathogenesis, namely partial loss of mbl function.

Chemical modifiers of a CUG-induced neuronal phenotype
At 28uC the semilethal phenotype of 103Y-Gal4.UAS-(CTG) 480 flies was highly sensitive to small changes in expression of CUG RNA and was easy to quantify. It therefore provided a tool to screen chemical suppressors of the neuronal toxicity to CUG RNA. To this end, we assayed the ability of 400 compounds from the Prestwick Chemical Library (PCL), a collection of drugs selected for their biological activity, to increase viability of female flies expressing i(CUG) 480 RNA in their MBs.
Drugs were tested individually diluted in nutritive media to <5 mM, which carried along the maximum amount of Dimethyl Sulfoxide (DMSO) that flies could tolerate ( Figure S4), and the number of adult females was compared to controls ( Figure 5J). Statistical analysis identified ten molecules (p,0.01; 2.5% of total tested) that significantly suppressed CUG-induced lethality (Table S1).
Chemical suppressors were classified into five categories according to their mechanism of action (MOA), including nonsteroidal anti-inflammatory agents, and drugs showing activity on sodium and calcium metabolism ( Table 2). Dopaminergic and cholinergic neurons enervate motor neurons, which are among the most abundant neuron populations in the Drosophila MBs [30], [31]. Two classes of compounds identified specifically acted on dopaminergic and cholinergic neurons, which suggests that i(CUG) 480 RNA is toxic to these cell types. Genetic evidence supports this hypothesis; targeted expression of i(CUG) 480 RNA to dopaminergic (Ddc-Gal4) and cholinergic (Cha-Gal4) neurons caused lethality (data not shown). Significantly, sodium channel blocker clenbuterol, which has been suggested effective to treat membrane excitability disorders including myotonic syndromes [32], [33], improved viability.
Compounds inhibiting Gal4 activity would lower transgene expression thus reducing toxicity to CUG RNA. Similarly, drugs might be working by stabilizing or degrading the CUG repeat RNA. To address these issues we first drove expression of the reporter UAS-lacZ with the 103Y-Gal4 line and measured ß-galactosidase activity in flies taking suppressor drugs and controls (Table S1). None of the chemical suppressors tested significantly altered reporter expression. Second, we measured the level of expression of 480 interrupted CUG repeat RNA under the same conditions used for the chemical screen in flies taking suppressor compounds and controls taking DMSO. Levels of expression were comparable for all tested drugs ( Figure S1B). Taken together these results suggest that candidate drugs did not significantly alter expression or stability of CUG repeat RNA and thus act through alternative mechanisms.

Discussion
Drosophila flies expressing 162 pure CTG repeats in the context of a 39UTR reporter gene show no detectable pathological phenotype despite forming discrete ribonuclear foci in muscle cells [4]. This suggests that ribonuclear foci are not directly pathogenic but also that Drosophila might be refractory to CUG-induced toxicity since 162 pure CTG repeats are well within the pathogenic range in humans. In an attempt to express larger CTG repeat expansions, we and others [5] used synthetic, interrupted, CTG repeat minigenes [34] to model DM1 in flies. This was necessary because manipulation of large CTG repeat expansions is difficult due to their intrinsic instability and failure to amplify by PCR. Interrupted minigenes have been shown to reproduce molecular alterations characteristic of DM1, in particular missplicing of cardiac troponin T [34] and colocalization with Muscleblind in the cell nucleus ( [35], [5]; this work). In the fly, targeted expression of 480 interrupted CTG minigenes to the eye precursors generated phenotypes sensitive to the genetic dose of muscleblind and in the adult musculature produced muscle degeneration ( [5], this work). Furthermore, we describe missplicing of muscle transcripts (CG30084 and troponin T). Although these are all alterations consistent with interrupted CTG repeats reproducing the behavior of pure CTG repeats, it remains formally possible that interrupting CTCGA repeats initiate molecular alterations unrelated to those of pure CUG repeat RNA, or somehow modify CUG-dependent toxicity. In this regard recent evidence shows that CGG trinucleotide repeats in permutation alleles of the fragile6gene (FMR1) cause neurodegeneration in Drosophila [36], [37] and involve disruption of RNA-binding protein function (hnRNP A2, Pura and CUG-BP1) as similarly described for alternative splicing regulators Muscleblind and CUG-BP1 in DM1. Thus, trinucleotide repeats similar to CTG have the capacity to cause RNA gain of function effects through mechanisms distinct from those described for CTG repeats. DM1 was the first example of spliceopathy, i.e. expression of splice products that are developmentally inappropriate for a particular tissue. CUG repeat RNA effectively misregulated alternative splicing of Z-band component CG30084 in Drosophila, leading to a strong increase of a transcript isoform we detect as RT-PCR band E ( Figure 4A), whereas such isoform was almost absent in control adult flies. Similarly, expression of a Drosophila TnT transcript isoform we detect as RT-PCR band D ( Figure 4B) was repressed in pupae expressing CUG repeat RNA, also leading to a developmentally abnormal alternative splicing. Expression of 60 CUG repeats altered alternative splicing of CG30084 and TnT transcripts although these repeats did not appreciably affect muscle morphology and did not accumulate in ribonuclear foci. The apparent mismatch between molecular and cellular markers of pathology merits further consideration. First, we detect a mild eye phenotype in flies expressing 60 CUG repeats ( Figure 3E) thus suggesting that 60 CTG repeats are indeed toxic to Drosophila cells but the phenotypes may be too weak to detect. Second, because the role of the ribonuclear foci in the disease state is currently unclear (foci are not pathogenic per se, at least in Drosophila [4]), absence of foci is not evidence that 60 CTG repeats are not toxic to Drosophila cells. Finally, the relevance of the alternative splicing alterations we detect in the TnT and CG30084 genes is currently unknown. However, we do note that all normal alternative splicing products are detected in CG30084 and appearance of band D is only delayed in TnT splicing. Therefore, we suggest that the apparent lack of match between phenotype and molecular defects in flies expressing 60 CUG repeat RNA might stem from the very different sensitivities of molecular methods and standard phenotypic assessment methods. Expectation was that flies expressing toxic RNA would show splice abnormalities typical of mbl loss-offunction [10]. However, we can not verify this prediction because no loss of mbl function phenotypes have been described in pupae and adults so far. We do notice, nevertheless, that expression of CUG repeat RNA in Drosophila embryos does not mimic molecular alterations described for mbl mutants [11], but we found inconsistencies in such description ( Figure S2). It is also likely that sequestration of Mbl by CUG RNA is incomplete, thus not generating a mbl null-like molecular phenotype. Indeed, the splicing of CG30084 was unaffected in mbl heterozygous embryos [11] demonstrating that even a reduction of 50% in Mbl protein is insufficient to interfere with splicing of CG30084.
Splicing of defined pre-mRNAs is defective in DM1, but the cellular readout of those changes is only beginning to be understood. The isolation of genetic enhancers and suppressors of a CUG-induced phenotype provides an unbiased approach for their identification. Our genetic screen recovered transcription and chromatin remodelling factors as modifiers. Previous observations have linked CUG toxicity to altered gene transcription [38]. Weakened cell adhesion due to impaired basement membrane, cell adhesion receptors, or both, might explain detachment of subretinal cells and sensitivity to the genetic dose of basement membrane component vkg and genes also influencing cell adhesion and cytoskeleton dynamics such as cnc [23], coro [25], and foi [26]. CUG repeat RNA might impair cell adhesion and sensitize cells to programmed cell death thus accounting for the reduction in eye size, and interaction with pro-apoptotic spin and apoptosis inhibitor th. Cell loss has been reported in specific brain areas of DM1 patients [1]. Cultured DM1 lens cells also show increased cell death, although the triggering event appears to be high intracellular Ca 2+ levels [39]. Isolation of mutations in mRNA export factor Aly as enhancers, finally, possibly underscores the relevance of changes in nuclear accumulation of (CUG) 480 transcripts for toxicity. Out of 400 drugs tested we identified ten that notably alleviated neuronal toxicity to CUG RNA. Assuming that the known MOA of the suppressor drugs apply to Drosophila, we found a number of molecules that inhibit neuron excitation through distinct mechanisms. These include dopamine D2 receptor antagonists (metoclopramide), inhibitors of monoamine reuptake (nefopam), and muscarinic and histamine receptor blockers (orphenadrine). Mutations that decrease or increase membrane excitability are known to trigger neurodegeneration to varying degrees in Drosophila [40]. Expanded CUG repeats might similarly induce excitotoxicity to MB neurons. Alternatively, neuronal hyperactivation may affect motor neurons in the brain, because pupae failed to emerge but were viable if released from puparium manually.
Using our CUG RNA fly model we identified mutations and drugs that significantly modified CUG toxicity phenotypes. These results advance our understanding of the cellular processes altered by CUG RNAs and provide a proof-of-concept data that Drosophila DM1 models can be successfully utilized for chemical screens.

Drosophila transgenics
Construct UAS-(CTG) 60 was generated by subcloning 54 uninterrupted CTG repeats from the pCTG54 plasmid [41] into the EcoRI/BamHI sites of the Drosophila expression vector pUAST. Sequencing of the construct revealed that repeats expanded to 60 during cloning. Because DM1 alleles carrying longer expansions probed intractable we decided to use synthetic CTG repeats interrupted every 20 CTG units by the sequence CTCGA [34]. CTG repeats in sp72 (Promega) were digested with XhoI and cloned into the same site in pUAST to generate the UAS-i(CTG) 480 construct. Both transgenes were injected into y 1 w 1118 embryos and independent lines established (6 UAS-(CTG) 60 and 14 UASi(CTG) 480 ). Nine out of 14 UAS-i(CTG) 480 lines were crossed to T80-Gal4, sev-Gal4, gmr-Gal4 and Mhc-Gal4 (see below for a description of these drivers) at different temperatures of culture. Of these, seven (1.1, 2.2, 3.3, 6.4, 7.1, 9.2, 13.1) revealed externally similar phenotypes in eyes, thorax/wing positioning, or ability to fly. Subsequent experiments were carried out with transgenic line 1.1, except for the assessment of nuclear CUG repeat RNA foci formation, which was also performed with transgenic line 2.2 giving the same qualitative result. Transgenic flies UAS-mblC:GFP will be described elsewhere. Briefly, the coding region of mblC was amplified by PCR and cloned in frame with GFP into peGFP-N3 (Clontech). The entire fusion gene was excised with BglII/NotI and subcloned into pUAST digested with the same enzymes. Transgenic flies were generated as above.

Scanning electron microscopy (SEM) and histology
Adult Drosophila eyes and thoraces were dissected out and embedded in Epon for transversal semi-thin sectioning [49] or processed for SEM [50]. Alternatively, thoraces were embedded in OCT and transversal sections (12 mm) were taken with a Leica CM 1510S cryomicrotome. Sections were processed for in situ hybridization with a Cy3-labeled (CAG) 10 probe and fluorescent detection of the MblC:GFP fusion protein as described [4]. SEM images were from a HITACHI S-2500. Image Manager Leica IM50 software was used to acquire cross-sectional muscle and vacuole areas.

Reverse transcription-polymerase chain reaction (RT-PCR) assay
Total RNA was extracted using Tri-Reagent (Sigma). To analyze the splicing patterns, 5 mg of total RNA were treated with DNase I and reverse transcribed (RT) with SuperScriptII RNase H 2 RT following instructions from the provider (Invitrogen). 10 ml of a 1:25 dilution (CG30084), 1 ml (Drosophila TnT) or 1 ml of a 1:100 dilution (Rp49) of the RT reaction were used as template in a standard 50 ml (CG30084) or 20 ml (TnT, Rp49) PCR using TaKaRa LA Taq (CG30084) or Thermus thermophilus DNA polymerase (Netzyme, NEED) (TnT, Rp49) polymerases. For cycling conditions, primer sequences and annealing temperatures see supplementary materials and methods (Text S1) and Table S2.

Compound administration and screen
Laying pots from en masse crosses (yw; +; UASi(CTG) 480 1.16103Y-Gal4/Y; +; +) were periodically checked for first instar larvae. Ten male larvae of the genotype yw/Y; +; UASi(CTG) 480 /+ and 20 female larvae with the genotype yw/103Y-Gal4; +; UAS-i(CTG) 480 /+ were hand-picked and transferred to vials with 1 ml of Instant Drosophila Medium (SIGMA) containing 5 mM of compound or 0.1% DMSO in controls. 400 compounds of the PCL (Tables S3 and S4) were individually tested in triplicate. Cultures were grown at 28uC and the sex of adults scored. Males were used as internal controls to discard unviable cultures or toxic drugs. Compounds showing activity in the initial screen (p,0.01; 30 drugs) were independently tested two more times in triplicate as above. For ß-galactosidase activity readings and DMSO toxicity assays see supporting materials and methods.

Statistical analysis
For the chemical screen, the following modification of the Fisher's exact test (a = 0.01) was used to analyze data from small size samples: where a is emerged females from control; b is dead females from control; c is emerged females form drug treated culture; d is dead females from drug treated culture. The number of emerged and dead females after drug administration was compared to that in control cultures. Data from all replicates was summed up and treated all together in order to increase the power of the test giving a final n = 60 (initial screen), or 180 for those drugs that were retested. We note that because the test we developed is exact, meaning by that we know the probability of the first species error and the potency of the test, the number of false positives does not increase with the continued use of the test. A t-student test was applied to all other comparisons between two groups.

Supporting Information
Text S1 It contains supplementary material and methods and supplementary reference list  Figure 3A. Bands A to D correspond to bands a to d in [11]. According to our results exon 12b is not detected in any RT-PCR product and exon composition of all bands shows inconsistencies with the published description [11]. Note that the presence of the transgene alone does not affect survival as the O/E ratio in males is still 1. These results show that muscleblind function is compromised in CUG-expressing MB neurons, thereby confirming the relevance of this phenotype to study DM1 defects in the brain. *** indicates p-value ,,0.0001. Found at: doi:10.1371/journal.pone.0001595.s004 (0.72 MB TIF) Figure S4 Toxicity of DMSO carrier. yw larvae were fed food containing increasing concentrations of DMSO and the number of individuals that reached adulthood was scored. Ten larvae were tested per replicate and up to five replicates were analyzed for each concentration. DMSO was not toxic up to 0.1% whereas concentrations of 0.15% or higher reduced viability in a doseresponsive manner when compared to controls; *** indicates pvalue ,, 0.001. Bars represent standard deviations. Found at: doi:10.1371/journal.pone.0001595.s005 (0.04 MB TIF) Table S1 Complete list of chemical suppressors of a CUGdependent semilethal phenotype. Drugs are listed alphabetically along with their main known activity in human cells, effect on expression of the UAS-lacZ reporter (measured by the enzymatic activity of b-galactosidase), and chemical structure. b-galactosidase activity comparisons between drug-treated flies and controls were only performed when total protein quantifications found no significant differences between samples. Found at: doi:10.1371/journal.pone.0001595.s006 (0.08 MB DOC)