The fruit fly Drosophila melanogaster has been proven to constitute an excellent model organism for scientific research for more than a century now (reviewed in [228]).
Since roughly 75 % of the known disease-associated genes in humans also have orthologues in flies (annotated genome with roughly 16,000 protein-coding genes [229]), it might be reasonable to draw conclusions from investigations on molecular mechanisms in the fly to those in humans. Drosophila melanogaster was one of the first multicellular organisms whose genome has been sequenced completely and the corresponding genetic knowledge is well-established. Creation of transgenic animals allows for the modelling of human diseases by expressing toxic gene products. Besides these rationales, Drosophila also conjoins additional advantageous properties especially for high-throughput approaches. Due to the fast replication cycle and the high number of offspring, experiments can be conducted within short time periods with a reasonable number of individuals, allowing for drug and genetic modifier screening [230]. Although being an invertebrate organism, experimental findings are gained from an in vivo situation and conclusions about molecular mechanisms in higher animals can be drawn without raising ethical issues. Last but not least, several powerful genetic tools have been introduced in the past years in order to render research with the fruit fly even more feasible, precise and easy to handle.
Some of these tools and a number of respective models and studies for polyQ disease are reviewed below.
2.4.1 The UAS/GAL4 expression system
The bipartite UAS/GAL4 ectopic expression system is frequently used in Drosophila as a means of overexpression of transgenes [231-233]. It makes use of the yeast transcriptional activator GAL4. Enhancer trap constructs (designed to facilitate GAL4 expression) were randomly inserted into the fly genome. If the insertion took place in the vicinity of an endogenous gene, GAL4 expression might mimic the expression pattern of this particular gene. To date there are plenty of so-called GAL4 driver lines available, mediating GAL4 expression in virtually every tissue at different time points throughout fly development. The gene of interest is introduced into a different fly line and put under the control of a GAL4 target, the upstream activation sequences (UAS). Upon crossbreeding of
these two fly lines, both moieties of the system are conjoined. GAL4 is produced under the control of the endogenous enhancer and able to bind to the UAS flanking the previously silent transgene of interest. Thus, expression of the gene of interest is enabled and directed in a spatiotemporal manner in the offspring (Figure 4). This renders the UAS/GAL4 system a valuable tool in fly genetics, although caution has to be taken since high GAL4 expression levels can have detrimental effects during development [234].
Figure 4. Model of the UAS/GAL4 expression system.
A tissue-specific endogenous enhancer binds to the promoter (grey) of the enhancer trap construct, thereby enabling gal4 expression (red) in the driver fly line. Association of GAL4 with the upstream activation sequence (UAS) of the fly line transgenic for the gene of interest activates expression (green). The bipartite nature of the system allows for tissue specificity and temporal restriction of activation of the gene of interest.
2.4.2 RNA interference (RNAi)
The gene silencing effect induced by double stranded RNA (dsRNA) termed RNA interference (RNAi) was fully established after experiments in Caenorhabditis elegans in 1998 [235]. It was the final step in a series of fundamental findings in plants [236, 237] as well as animals [238, 239]. Originally an endogenous mechanism involved in translational repression [240], development [241] and defence against parasitic genes [242], RNAi quickly evolved to be a powerful technique in scientific research, e.g. in mimicking knockout experiments without the extensive work effort of creating classical knockout animals. The effectors of RNAi are diverse small interfering RNAs (siRNAs) categorised according to their origin, biogenesis, mode of action and size [243, 244]. The source of siRNAs used in this work are transgenes coding for short hairpin RNAs (shRNAs). These transgenes consist of 100-400 base pairs present as an inverted repeat (IR) separated by miscellaneous nucleotides. Following expression of the IR, it will form a short hairpin RNA
(shRNA) which is exported from the nucleus to the cytoplasm. The shRNA is bound and cleaved by the ribonuclease protein Dicer-2, resulting in a double stranded structure without loop and RNA tails [245]. This small interfering RNA (siRNA) is then bound and translocated to the RNA-induced silencing complex (RISC) by the RISC loading complex (RLC) protein R2D2 which additionally discriminates between guide and passenger strands of the siRNA [246, 247]. The RLC recruits Argonaute2 (Ago2) and transfers the dsRNA to it, resulting in decay of the passenger strand by this endonuclease [248].
Subsequent to the release of the passenger strand and disassembly of R2D2, the active RISC is formed. The complex is capable of recognising and binding the messenger RNA (mRNA) target by base pairing with the guide strand. Eventually, this leads to cleavage of the mRNA and effectively to silencing of gene expression (Figure 5) [249].
Based upon this mechanism, Dietzl et al. were the first to establish a Drosophila RNAi library covering ~90 % of the entire fly genome. It utilises the conditional UAS/GAL4 expression system for induction of shRNAs under UAS control, leading to RNAi for the respective gene upon crossbreeding with a GAL4 driver line [250].
Figure 5. Mechanism of RNAi with shRNA.
The inverted repeats of the shRNA transgene are transcribed and the RNA is assembled into an shRNA. Following export from the nucleus, the shRNA is processed into double-stranded siRNA by Dcr-2.R2D2 forms the RLC together with the siRNA and discriminates the guide and the passenger strand, the latter is degraded upon binding of the RLC to Ago2. The guide strand and Ago2 form the RISC, eventually binding to and cleaving the target mRNA.
Partially adapted from Dan Cojocari, Dept. of Medical Biophysics, University of Toronto, 2010.
2.4.3 Rough eye phenotype (REP)
The Drosophila compound eye is a highly ordered structure made up of about 800 single unit eyes termed ommatidia. Each ommatidium consists of eight photoreceptor cells arranged in an asymmetric trapezoid pattern accompanied by cone and pigment cells [251]. Cellular dysfunction and cell death as well as perturbation of crucial developmental pathways during compound eye formation lead to disturbances in this exact lattice and a so-called rough eye phenotype (REP). Consequently, a REP can be induced by the overexpression of toxic gene products. Expression of a disease gene can be targeted to postmitotic cells, including photoreceptor neurons, of the compound eye with the driver line glass multiple reporter (GMR)-GAL4 in combination with the UAS [252]. Glass expression starts at day one of larval stage L3 in all cells posterior of the morphogenetic furrow of the eye disc [253] as well as in a minor population of cells in the brain. The severity of the REP is directly correlated to the loss of underlying photoreceptor neurons reflected by vacant, interstitial or fused ommatidia and disordered sensory bristles. Since the fly compound eye is a neuronal structure easily accessible by light microscopy, the REP is an easy readout to assess changes in the decline of photoreceptor neurons caused by eye-specific expression of neurotoxic proteins. Thus, changes in REP have been successfully used in genetic screens set to identify modifiers of neurodegenerative disorders [28, 136, 254-256]. However, neurodegeneration in the fly eye cannot completely mimic the complex processes leading to disease in the human brain.
2.4.4 Drosophila models of polyglutamine disease
Disease models for polyQ disorders in Drosophila mostly involve overexpression of the common pathogenic feature of the causative proteins, concentrating on the expanded polyQ tract itself. Several fly lines have been introduced containing proteins entirely composed of normal or mutated polyQ stretches of different length (20Q, 22Q, 108Q, 127Q;
[28, 257]). The expanded polyQ peptides in these models were sufficient to cause neurotoxicity despite the absence of their disease gene context. Additionally, studies have shown that a pure polyQ domain is much more toxic than a polyQ domain flanked by even relatively small protein sequences [258]. The detrimental intrinsic cytotoxic effects could be modified by genetic factors or modulations of the polyQ tract alone. These pure polyQ
approaches naturally neglect disease gene-specific characteristics and do not explain cell type specificity of distinct polyQ diseases. However, the previous work utilising such polyQ peptides revealed valuable novel insights into disease mechanisms.
The utilisation of truncated disease gene models is a feasible approach to study polyQ toxicity since it is assumed that causative proteins are also cleaved and thus truncated in vivo prior to oligomerisation and noxious effects. Several fly models for polyQ diseases are described. These models rely on expression of polyQ repeats either embedded in the C-terminal region of the human Ataxin-3/MJD protein (SCA3tr-Q27, SCA3tr-Q78;
[219]), in an N-terminal truncated fragment of human Htt (Q2, Q75, Q120, Q128; [259, 260]) or exon 1 of Htt only (Q93; [75]). The distinction between the pathologies of polyQ diseases is more apparent at the level of truncated polyQ proteins. For instance, expression of the viral antiapoptotic protein p35 mitigated the REP in the SCA3tr-Q78 model, but failed to do so in Htt-trQ120 [219, 259]. Most of these truncated disease gene models exhibit progressive protein aggregation, forming nuclear inclusions in neurons, and late-stage neurodegeneration. On the contrary, flies with a normal number of repeats show a diffuse and cytoplasmic protein distribution and no overt neurotoxicity.
Expression of human versions of the full length polyQ proteins in Drosophila is an approach to investigate the pathogenic potential of elongated polyQ tracts in their native protein context. Studies show that high levels of wild-type full length Ataxin-1 (Q30) exert disturbances in eye morphology and expansion of the polyQ tract (Q82) leads to detrimental effects and a rough eye phenotype that can be modified by genetic interactors [136]. Investigations on Huntington’s disease involve generation of a fly line with a full-length Huntingtin containing 128Q [261], presenting a neurodegenerative eye phenotype due to demise of photoreceptor neurons. Similar approaches have been described for the androgen receptor in SBMA [262] and for full-length Ataxin-3 with polyQ expansion of 84 repeats in SCA3 [209]. In the latter model, flies showed severe and adult-onset neural degeneration when expression of the toxic disease protein was restricted to the compound eye or the nervous system, which was not observed for the wild-type protein. The full-length mutated protein is more selectively toxic to the nervous system compared to the truncated isoform and accumulated in ubiquitinated inclusions. Coexpression of wild-type full-length Ataxin-3 on the contrary is able to ameliorate the detrimental effects of the toxic variant even in models of SCA1 and HD.
2.4.5 Previously implemented modifier studies
Several studies were conducted in order to reveal genes that modify the cytotoxicity and deleterious consequences of polyQ expansion in the diverse causative disease proteins.
As already mentioned above, the baculovirus antiapoptotic gene p35 suppressed truncated mutant Ataxin-3-dependent degeneration in the eye, as does the human heat shock protein HSP70 [30, 219]. Kazemi-Esfarjani et al. published results of one of the first large scale modifier screens in flies expressing prolonged polyQ peptides only [28]. They utilised a set of 7,000 P-element insertions for crossbreeding with the disease flies and assessed the offspring for suppression or enhancement of the polyQ-induced rough eye phenotype. Out of a number of potential candidates they presented two chaperone-related gene products, dHDJ1 (equivalent to human HSP40/HDJ1) and dTPR2 (equivalent to human TPR2) as potent genetic suppressors of polyQ toxicity. Fernandez-Funez et al. utilised a fly model based on full-length Ataxin-1 Q82 expression for two screens with 1,500 lethal P-elements and 3,000 EP insertions respectively, also evaluating the change of REP in the F1 generation [136]. They identified 18 genes that, if altered in expression, enhanced or suppressed Ataxin-1 toxicity. Among these genes were some coding for ubiquitin-related proteins, chaperones, RNA-binding molecules and transcription factors. Bilen et al.
described the crucial involvement of microRNAs (miRNAs) pathways in the modulation of polyQ toxicity induced by Ataxin-3 after screens in Drosophila and human cells [263]. The same group conducted a genome-wide EP element-based screen for modifiers of Ataxin-3-induced neurodegeneration in Drosophila [256]. They identified 25 modifiers representing 18 genes that are mainly involved in biological processes affecting protein misfolding and ubiquitin-related pathways. Among others, this experiment was designed as a genetic high-throughput screen based on the misexpression of endogenous genes [264]. Despite their general feasibility, these screening strategies create artificial expressions states potentially masking the native influence of the respective gene product. Furthermore, they are mostly confined to a small portion of the genome.
These disadvantages can be overcome by mimicking classical knockout experiments with the help of RNA interference-mediated gene silencing. This powerful technique has been successfully implemented into high-throughput screening for modifiers of Huntingtin aggregation in yeast [265] and Drosophila cells [266]. Genome-wide studies utilising RNAi revealed regulators of polyQ and Huntingtin aggregation in C. elegans [267] and Drosophila [268] respectively.
3 Aim of the Study
The comprehensive understanding of molecular mechanisms and cellular pathways resulting in polyQ neurotoxicity and pathogenesis are a prerequisite for the development of effective treatments for the corresponding disorders. In an attempt to contribute to this process we intended to conduct a genome-wide high-throughput modifier screen in order to identify genetic modifiers of polyQ neurotoxicity in Drosophila melanogaster. This should be accomplished first of all by characterisation of a feasible disease model exhibiting a readily accessible readout for the large scale experiments. Expression of a human variant of truncated Ataxin-3, harbouring a stretch of 78 glutamines, in the compound eye results in an REP. This REP combines pathological involvement of neuronal photoreceptor cells with an easy exterior observability of neurodegeneration. By means of RNA interference we planned to knockdown a genome-wide set of potential modifier genes, thereby evaluating the impact of the gene silencing on the REP. For this purpose, we obtained a set of fly RNAi strains from the VDRC representing all Drosophila genes known to have an orthologue in humans. A genetic modifier screen with these ~7,500 genes would represent the most comprehensive endeavour in this field and setting so far. The gene knockdown approach, the in vivo situation and the easy assessment of neurotoxicity modification mark the advantages of this work in comparison to previously conducted modifier screens. By subsequent thorough analysis and processing of our results and the obtained modifiers we hope to aid in conceiving polyQ diseases better and opening avenues for therapeutic approaches.