Drosophila screen connects nuclear transport genes to DPR pathology in c9ALS/FTD

Hexanucleotide repeat expansions in C9orf72 are the most common cause of amyotrophic lateral sclerosis (ALS) and frontotemporal degeneration (FTD) (c9ALS/FTD). Unconventional translation of these repeats produces dipeptide repeat proteins (DPRs) that may cause neurodegeneration. We performed a modifier screen in Drosophila and discovered a critical role for importins and exportins, Ran-GTP cycle regulators, nuclear pore components, and arginine methylases in mediating DPR toxicity. These findings provide evidence for an important role for nucleocytoplasmic transport in the pathogenic mechanism of c9ALS/FTD.

DPRs (GA, GR, PA and PR). Consistent with recent reports 8-10 , we found that expression of the arginine-rich DPRs, GR and PR, strongly reduced survival in flies expressing these DPRs in adult flies either ubiquitously or motor neuron-specific (Fig. S1). Thus, this Drosophila model recapitulates robust C9orf72 DPR toxicity, providing a tractable system to identify and characterize toxicity modifier genes.
Jovičić et al. reported results from two unbiased genome wide screens in yeast for suppressors and enhancers of PR toxicity 11 . These two screens identified a striking number of modifier genes involved in nucleocytoplasmic transport. These modifiers include karyopherins, nuclear pore complex components, and enzymes involved in generating the Ran-GTP gradient that drives nuclear transport. Since the general principles and key molecules of nucleocytoplasmic transport are highly-conserved from yeast to flies to humans 12 , we sought to validate these results in an animal model and to test the hypothesis that genes involved in nuclear transport could also modify C9orf72 DPR toxicity in vivo. We therefore performed a targeted RNAi screen in Drosophila. To facilitate the rapid identification of modifiers of DPR toxicity, we used eye degeneration as a readout. We directed expression of a single copy of the 25 PR repeat construct to the fly eye. This caused a moderate degenerative phenotype (Fig. 1a), providing the ability to identify both suppressors and enhancers.
To focus on nucleocytoplasmic transport and related processes, we compiled a library of 121 independent RNAi lines targeting 55 fly genes (Table S1), encoding nuclear pore complex proteins, importins, exportins, regulators of the Ran-GTP cycle, and arginine methylases, which affect protein localization by modulating NLS sequences 13 . We expressed each RNAi together with the PR25 construct and scored for the ability to enhance or suppress the phenotype (Fig. 1b, Fig. S2). We identified 15 enhancers and 4 suppressors of the PR25 eye  (Table 1, Table S4). Importantly, the RNAi lines did not cause a degenerative eye phenotype in a wild type background (Fig. S3a) and the RNAi lines that suppressed PR toxicity did not affect PR expression (Fig. S3b).
The strongest enhancer of PR-mediated neurodegeneration was knockdown of Trn ( Fig. 2a, Fig. S4), the fly ortholog of TNPO1, encoding transportin 1. Of notice, the yeast TNPO1 homolog Kap104 was one of the strongest suppressors of PR toxicity in yeast 11 . Interestingly, transportin 1 has been connected to another form of ALS/FTD, namely related to FUS 13,14 . ALS-causing mutations in FUS/TLS impair transportin-mediated nuclear import, resulting in FUS cytoplasmic accumulation and aggregation 15 and in FTD cases with FUS pathology transportin 1 is mislocalized to cytoplasmic FUS aggregates 14 . To determine whether PR could directly act on transportin 1 function, we performed computational docking simulations (Fig. 2b) and predicted that PR can interact with transportin 1. This suggested that PR might compete for endogenous transportin 1 cargoes. To test this hypothesis, we analyzed the effect of PR expression on the localization of a well-characterized transportin 1 cargo, the neuronal RNA-binding protein Elav. We observed increased cytoplasmic localization and decreased nuclear staining of Elav in flies expressing PR (Fig. 2c). Moreover, Elav mislocalization increased further upon Trn knockdown (Fig. 2c). Consistent with previous reports, we also observed the cytoplasmic aggregation of the transportin 1 cargo hnRNPA3 in human c9FTD (Fig. 2d, Table S2). These findings suggest a potential role of transportin 1 in the pathogenesis of c9ALS/FTD.
Finally, knockdown of four out of ten arginine methyltransferase genes enhanced PR toxicity in Drosophila (Table 1, Fig. 2e). One of these, PRMT1, has been previously implicated in ALS caused by FUS/TLS mutations 13,16 .
To determine whether this enzyme could directly affect DPR toxicity, we performed colocalization experiments in cell lines. PRMT1 colocalized with both GR and PR (Fig. 2f, Fig. S5). We subsequently tested several commercially available antibodies to detect methylation. One of them, ASYM24, was able to detect asymmetric arginine dimethylation of GR, but not PR. This was not surprising since this antibody was originally raised against a peptide showing strong sequence similarity with GR, but not PR (Fig. S5). Using this antibody we found GR accumulations to be methylated in transfected cells, but also an association of GR with other methylated proteins, as determined by super resolution microscopy (Fig. 2g). To validate this new aspect of DPR pathology in humans, we performed immunostaining on c9FTD brain samples and detected abundant methylated inclusions (Fig. 2h, Table S3). These results implicate arginine methyltransferases to c9ALS/FTD pathogenesis.

Fly gene E/S Human gene
Nuclear Pore Complex Importins/Exportins Table 1. Modifiers of PR25-mediated eye degeneration uncovered in a targeted RNAi screen. * Indicates genes which were independently discovered in genome-wide yeast screens 11 . S = suppressor; E = enhancer. In summary, we performed a targeted genetic modifier screen in Drosophila and confirm results obtained in two yeast screens 11 . Both studies consolidate DPRs as major contributing factors to c9 toxicity and point at a role for nucleocytoplasmic transport in this toxicity. Strikingly, the homologs of the strongest genes from the yeast screens were potent modifiers of DPR toxicity in Drosophila. Not only does this validate the effectiveness of modifier genes discovered in yeast in an animal model, it also suggests that DPR pathologies associated with c9ALS/ FTD disrupt a highly conserved facet of cell biology.
Depletion of RNA-binding proteins (RBPs) such as TDP-43, FUS and others from the nucleus, and cytoplasmic accumulation are pathognomonic features of both ALS and FTD 17 , including c9ALS/FTD [18][19][20] . The upstream triggers of this nuclear depletion and subsequent cytoplasmic aggregation are unresolved. Our results help to explain how disturbances in nucleocytoplasmic trafficking caused by C9orf72 repeat expansion pathology could trigger RBP mislocalization and subsequent aggregation. Most RBPs are tightly regulated in their cellular localization, and shuttle in a strictly controlled manner between nucleus and cytoplasm. Impairments, even subtle, to the nuclear pore, karyopherins, or the Ran-GTP gradient, could perturb this sensitive equilibrium providing a first hit that would eventually lead to aberrant accumulation of RBPs in the cytoplasm, setting off a cascade of aggregation and sequestration of these proteins into pathological inclusions. Interestingly, preliminary histopathological studies suggest that C9orf72 DPR accumulation predates TDP-43 pathology in FTLD patients harboring C9orf72 mutations 21,22 .
DPR pathology is one way that C9orf72 mutations might contribute to disease pathogenesis. First of all, a role for C9orf72 loss of function, owing to decreased expression of C9orf72, has not been conclusively ruled in or out. A recent report potentially even links the C9ORF72 protein to nuclear transport 23 . Secondly, sense and antisense RNA transcripts produced from the GGGGCC hexanucleotide repeat expansion accumulate in the nucleus and cytoplasm of mutation carriers 3 and could cause disease by an RNA toxicity mechanism (e.g., by sequestering important regulatory proteins like splicing factors and other RNA-binding proteins). Two recent studies report the results from genetic screens similar to ours using GGGGCC repeat fly models 24,25 . Compellingly, these studies also identify nuclear transport as a key pathogenic factor in these fly models, hereby suggesting that the repeat RNA itself could also directly perturb nuclear transport. However, two other reports have ruled out RNA toxicity in flies, at least at the short lengths used in the current models, and attribute all observed phenotypes to DPR toxicity 10,26 . Moreover, since we identify similar modifiers using a pure DPR model, this raises the question whether the repeat RNA itself is truly involved in the observed nuclear transport defects.
The three potential pathogenic mechanisms are not mutually exclusive and future studies will be required to disentangle the relative contributions of DPR proteotoxicity, RNA toxicity, and C9orf72 loss of function to c9FTD/ALS. Moreover, why the same C9orf72 mutations cause dementia in some patients, ALS in others patients, and ALS/FTD in yet others, even within the same family, is unresolved and might be influenced by modifier genes. A better understanding of the mechanisms by which C9orf72 hexanucleotide repeats cause disease will allow the identification of novel targets for therapeutic intervention.

Material & Methods
Plasmids and strains. FLAG-tagged DPR expression constructs were designed by manually codon-optimizing the sequence, and using Mfold software 27 to control for any persistent stable secondary structures (Fig.  S6, Table S5). DNA constructs were synthesized by Genscript (Piscataway, USA). DPR constructs were subcloned in CMV6 entry plasmids (Origene) for expression in mammalian cells. Subcloning to the pUAST-attB backbone allowed the generation of transgenic fly lines by targeted insertion into the 62E1 attP locus on the third chromosome (GenetiVision, USA). The PRMT1-EGFP construct was a kind gift of Dr. F. Fackelmayer (Laboratory for Epigenetics and Chromosome Biology, Ioannina, Greece).
Lifespan assay. Lifespan experiments were performed using, the TARGET system 28 . The tub-Gal4 or the D42-Gal4 driver was combined with a ubiquitously expressed temperature-sensitive Gal80 inhibitor (tub-Gal80ts). Fly crosses were grown at 18 °C and adult progeny of carrying the tub-Gal4 or the D42-Gal4 and tub-Gal80 chromosomes and the UAS-DPR responder gene were shifted to 29 °C to allow expression of the transgenes. Females were collected within 24 hr of eclosion and grouped into batches of 10 flies per food vial. Fresh food vials were provided every 2-3 days.
Dot blot analysis. The GMR PR25 screening stock was crossed with the four suppressor lines and offspring was collected. 30 flies per condition were decapitated and fly heads were homogenized in 30 μ l of RIPA buffer. 15 μ l of homogenate was spotted onto a nitrocellulose membrane (Amersham) and air dried. Loading controls were visualized using Coomassie staining (Life Technologies). Membranes were blocked in 5% milk powder (Bio-Rad) in TBS-T buffer and probed with a custom rabbit PR antibody (Thermo Scientific). HRP-labeled anti rabit secondary antibody was used (Dako) and membranes were imaged using chemiluminescence (Pierce) and an Image Quant Las 4000 imaging station (GE).
The ASYM24 ant ib o dy was or ig ina l ly generate d against an inter r upte d GR rep e at (KGRGRGRGRGPPPPPRGRGRGRG). This antigen shows strong sequence similarity with GR and hence was able to detect GR methylation, but not PR methylation. The antibody was specific for methylated residues as staining was abolished using a general methylation inhibitor Adox.
Confocal images were obtained using a Zeiss LSM 510 Meta NLO confocal microscope and SIM microscopy was performed using a Zeiss Elyra S.1 microscope (Carl Zeiss, Germany). SIM calculations were performed using default settings. Images were analyzed, formatted and quantified with FIJI and ImageJ software. For all experiments representative photographs are shown from multiple wells from transfections with at least two cell passages, or from multiple fly brains.
Autopsied brains of six C9orf72 carriers and two non-disease controls were obtained using informed consents and protocols that were approved by the Ethical Committee of University of Antwerp and Antwerp University Hospital and stored in the Antwerp Biobank of the Institute Born-Bunge. Methods were carried out in accordance with the approved guidelines. Clinical data are shown in Table S6. After a fixation period of 8 to 16 weeks in 10% buffered formalin, 5 μ m slices were cut from following regions: frontal cortex, hippocampus with dentate gyrus and parahippocampal gyrus, cerebellar cortex, medulla oblongata. Sections were deparaffinized, rehydrated and pretreated with citric acid 0.1 M. Immunohistochemical analysis was performed with anti-hnRNPA3 antibody (AV41195, Sigma) and ASYM24 antibody (07-414, Millipore). Sections were counterstained with hematoxylin and images were taken on an Axioskop 50 light microscope (Zeiss) equipped with a CCD UC30 camera (Olympus Inc.).

RNAi modifier screen.
To identify modifiers of our PR25 eye phenotype we crossed 121 RNAi lines with our screening stock. The RNAi lines were obtained from VDRC or Bloomington Drosophila stock center (USA). For each cross the collected offspring was divided by sex, and the genotypes were counted according to the balancers. The offspring ratio was determined by (expected offspring/counted offspring). For each sex we subsequently assigned an average color using following scoring scale (white = 1, yellow = 2, yellow-orange = 3, orange-yellow = 4, orange = 5, orange-red = 6, red-orange = 7, red = 8). Afterwards, each fly was individually scored for the presence of necrotic spots using following scoring scale (not affected = 0, mild = 1, medium = 2, heavy = 3, extreme = 4). We crossed each line at least two independent times. After compiling all data, two researchers assigned independently a status to each RNAi line (no effect, enhancer or suppressor) as compared to a cross of the screening stock to the RNAi w1118 background. Only whenever a gene was represented by at least two RNAi lines showing a similar effect on the phenotype, the gene was classified as an enhancer or suppressor. When shown, statistics were carried out using Prism software.
RNAi lines did not present with a degenerative eye phenotype by themselves. Two RNAi lines used had a 'greasy' eye phenotype reminiscent of mitochondrial eye phenotypes. These lines indeed had reported off target effects on genes involved in eye development: i.e. v105181 off target CG8085, v36103 off target CG4389. Importantly, both these hits were verified by independent RNAi lines without this 'greasy' eye phenotype.
Structural modeling. We employed the FoldX force field 29 to model PR and GR in the binding pocket of transportin-1 (PDB code: 2OT8). In addition, we performed an unrestrained energy minimization using the YASARA2 force field to optimize the interactions between the PR/GR peptide and the binding pocket of transportin 1 30 . The graphical representation was generated using the Yasara program (version 13.2.21) where negatively charged residues in the binding pocket, within a distance of 5 Angstrom of the peptide, were colored in blue.