To identify disease-specific gene expression patterns associated with poly-GR, we performed RNA-seq on heads of adult flies expressing ATG-driven GR100 specifically in neurons, using the Elav-GS RU486-inducible driver. RNA-seq was performed three days after induction of GR100 expression, in order to identify early changes prior to overt neurodegeneration. These flies showed strong alterations in their transcriptome when compared with the control line. We identified 2754 genes significantly differentially regulated (adjusted p<0.05) (Figure 1—figure supplement 1A and GEO: GSE151826). To gain insight into the potential function of the differentially expressed genes, we performed gene ontology (GO) enrichment analysis using TopGO on the 2754 up- and down-regulated genes (Figure 1A). Flies expressing poly-GR100 showed altered expression of genes involved in pathways previously implicated in C9orf72 pathology, including translation, DNA damage and repair, proteasome, and RNA metabolism (Figure 1A and Figure 1—figure supplement 1B). Interestingly, neuropeptide hormone activity was the most enriched category of down-regulated genes (Figure 1A). Among these neuropeptides, Drosophila insulin-like peptides (dilps) 2, 3, and 5 were highly down-regulated (3.7-, 9.5-, and 3.2-fold change, respectively), adjusted p value<0.05 – Figure 1—figure supplement 1A. RT-PCR analysis of brains of flies expressing poly-GR confirmed the lowered expression of dilp 2, 3, and 5 (Figure 1B). To test whether these dilps were also down-regulated in a GGGGCC repeat model that generates poly-GR via RAN translation, we performed RT-PCR in flies expressing (G4C2)36. Lower expression of dilps was also observed in this model, although the expression change in the dilp5 gene did not reach statistical significance (Figure 1—figure supplement 2).

Dilps 2, 3, and 5 are neuropeptides secreted by insulin-producing cells (IPCs) in the Drosophila brain (Brogiolo et al., 2001; Ikeya et al., 2002). To investigate whether reduced expression of dilps was simply due to the death of IPCs, we imaged brains of flies expressing poly-GR in all neurons. Dilp2 immunostaining showed no alteration in the number of IPCs (Figure 1—figure supplement 3A and B). However, consistent with our RNA-seq data, Dilp2 protein levels were reduced in IPCs expressing poly-GR (Figure 1—figure supplement 3C). In addition, expression of poly-GR specifically in IPCs using the dilp3-Gal4 driver was not sufficient to induce neuronal cell death (Figure 1—figure supplement 3D). Reduction of dilp2, dilp3, and dilp5 expression in GR100 flies was therefore not due to loss of IPCs.

Dilps signal via a single insulin/IGF receptor (InR), through which they regulate the main signalling pathways that modulate 4E-BP1 phosphorylation (Figure 1—figure supplement 4A), a well-described read-out of insulin signalling in Drosophila (Bhandari et al., 2001; Wang et al., 2003; Blancquaert et al., 2010). To assess insulin/IGF pathway activity, we measured the phosphorylation state of 4E-BP1 in (G4C2)36 flies, and found that they had significantly decreased ratio of phosphorylated 4E-BP1 to the non-phosphorylated form, indicative of reduced insulin signalling (Figure 1—figure supplement 4B). Co-expressing a constitutively active insulin receptor (InR^Active) significantly increased this ratio, indicating a rescue of insulin/IGF signalling immediately downstream of the dilps (Figure 1—figure supplement 4B). Taken together, these data indicate a reduction in insulin/IGF pathway activity in flies expressing either G4C2 repeats or poly-GR as a result of a reduction in expression of the dilps.

We next asked whether restoring insulin signalling in neurons could ameliorate G4C2 repeat toxicity. To assess this, we monitored the survival of flies co-expressing (G4C2)36 and InR^Active, specifically in neurons, and found that their lifespan was significantly extended (p<0.001) (Figure 2A). In contrast, reducing insulin signalling through the expression of dominant-negative InR (InR^DN) significantly reduced lifespan (p=0.027) (Figure 2A). Activating insulin signalling in neurons reduced lifespan of wild-type flies (InR^Activep<0.0001), as previously reported (Ismail et al., 2015), whilst its reduction led to increased lifespan (p=0.035) (Figure 2B), again as previously reported (Augustin et al., 2018). These observations indicate that increasing insulin/IGF signalling specifically suppresses C9orf72 repeat toxicity.

Induction of the neuron-specific driver requires flies to ingest RU486. To exclude the possibility that the rescue effect was a consequence of decreased fly feeding, and therefore reduced induction of the (G4C2)36 transgene, we measured food intake. There was no significant difference between the amount ingested across the different experimental groups (Figure 2—figure supplement 1A). To rule out the possibility that expression of either InR^Active or InR^DN had a direct effect on the transcription of the G4C2 transgene, we measured the transcript repeat levels by RNA dot blot analysis in flies ubiquitously expressing the two constructs. Neither expression of InR^Active or InR^DN altered the G4C2 transcript levels (Figure 2—figure supplement 1B). To further investigate whether expression of InR^Active or InR^DN indirectly affected the inducible protein expression system, we measured levels of GFP driven by the Elav-GS RU486-inducible driver and found that they were unaltered in neuronal cells expressing InR^Active or InR^DN (Figure 2—figure supplement 1C). Activation of insulin/IGF signalling therefore ameliorated C9orf72 repeat toxicity, rather than simply reducing expression of the (G4C2)36 transgene.

To confirm that increasing insulin signalling could also ameliorate a second, distinct (G4C2)36-induced neuronal phenotype, we recorded motor activity and day sleep using a Drosophila activity monitor system. Flies expressing (G4C2)36 in adult neurons exhibited decreased locomotor activity (p=0.0003) and extended sleep periods during the day and night (Figure 2C and Figure 2—figure supplement 1D and E), and these phenotypes were abolished by expression of InR^Active (Figure 2C and Figure 2—figure supplement 1D and E).

We next determined whether activation of insulin/IGF signalling also rescued a molecular signature of (G4C2)36 toxicity. Increased levels of p53 have been observed in C9orf72 patient iPSC-neurons and fly models, and have been suggested to be a downstream marker of repeat-induced toxicity (Lopez-Gonzalez et al., 2016; Lopez-Gonzalez et al., 2019). We therefore tested if co-expression of InR^active in (G4C2)36 flies was associated with decreased levels of p53 and found that it led to a significant reduction (Figure 2D), whereas no difference was observed in uninduced flies (Figure 2D). These results show that increased insulin/IGF signalling can ameliorate multiple read-outs of G4C2 repeat-induced toxicity.

As toxicity in the (G4C2)36 flies is mediated by poly-GR (Mizielinska et al., 2014), and increased p53 has been suggested to be driven by poly-GR (Lopez-Gonzalez et al., 2016Lopez-Gonzalez et al., 2019), we next assessed whether increased insulin/IGF signalling could specifically rescue poly-GR toxicity.

We took advantage of the rough eye phenotype and degeneration of eye tissue caused by expression of 36 poly-GR repeats (GR36) in the Drosophila eye (Mizielinska et al., 2014). To examine whether this pathology could be ameliorated by increased insulin/IGF signalling, we co-expressed GR36 with either InR^Active or InR^DN (Figure 3A). As previously reported (Mizielinska et al., 2014), ectopic expression of GR36 resulted in a mild rough eye and decreased eye size (Figure 3A). Co-expression of GR36 with InR^Active increased the size and decreased the roughness of the eyes, whereas co-expression with InR^DN exacerbated the rough eye phenotype and further decreased eye size (Figure 3B). Although insulin/IGF signalling also influenced eye cell growth in flies not expressing the dipeptide repeats (Figure 3B), the effect in diseased flies was much larger (p<0.0001, two-way ANOVA), indicating a specific additional interaction of insulin signalling with C9orf72 repeat-induced toxicity. Expression of dilp2 had no effect (Figure 3C), likely because of a negative feedback system that acts to coordinate Dilp expression levels in the central nervous system (Grönke et al., 2010). To better understand how insulin/IGF signalling rescued the toxic effects of poly-GR, we interrogated the effect of downstream effectors of insulin signalling. We co-expressed GR36 with activated PI3K or Akt, which function downstream of the InR, and found that over-expression of either partially rescued the eye size of GR36 flies (Figure 3C). Together, these observations show that increased insulin/IGF signalling through the InR/PI3K/Akt pathway can rescue neurotoxicity associated with poly-GR expression.

We next investigated the mechanism by which increased insulin signalling reduced C9orf72 repeat toxicity. We tested whether activation of insulin signalling could alter the level of poly-GR present in the heads of flies expressing (G4C2)36. Using a quantitative Meso Scale Discovery (MSD) immunoassay, we found that expression of InR^Active significantly decreased poly-GR levels (Figure 4A).

In order to investigate whether the effect of InR^Active on poly-GR levels occurs at the level of RAN translation, we investigated whether InR^Active could reduce poly-GR levels in GR100 expressing flies, which generate poly-GR but do not undergo RAN translation. Neuronal expression of InR^Active in GR100-expressing flies reduced poly-GR levels (Figure 4B), with a concomitant extension of lifespan (Figure 4C), further confirming the protective effect of increased insulin signalling. This indicates that InR^Active is acting downstream of RAN translation to reduce poly-GR levels, and suggests that insulin/IGF signalling activation ameliorates toxicity by decreasing poly-GR levels.

In order to determine whether the insulin-PI3K/AKT signalling pathway regulates DPR levels in mammalian cells, we utilized a nanoluciferase (NLuc) reporter which contains 92 seamless G4C2 repeats, which are preceded by 120 nucleotides of the endogenous human upstream sequence and followed by NLuc, lacking a start codon, in frame with poly-GR, termed 92R-NL. Thus the NLuc signal reports on RAN translated poly-GR levels. 92R-NL was co-transfected into HEK293T cells with a control plasmid expressing an ATG-driven firefly luciferase (FLuc) as a transfection efficiency and cell number control. To modulate the insulin-PI3K/Akt pathway, cells were treated for 2 days with either the pan-Akt inhibitor MK2206 (Hirai et al., 2010) or the PTEN inhibitor SF1670 (Rosivatz et al., 2006); insulin leads to the phosphorylation and activation of Akt, thus inhibiting Akt negates the effects of insulin on the PI3K/Akt pathway (Figure 1—figure supplement 4A). PTEN inhibition facilitates the phosphorylation and activation of Akt, thus activating the pathway (Figure 1—figure supplement 4A). MK2206 increased NLuc/polyGR levels (1.77-fold ± 0.54 SD, p=0.0168) (Figure 5A), while SF1670 decreased NLuc/polyGR levels (0.24-fold ± 0.09 SD, p=0.041) (Figure 5B). Therefore, consistent with our data in flies, increasing insulin signalling via Akt decreased poly-GR levels, while inhibiting the pathway increased poly-GR.

Finally, we tested if treatment with insulin itself could also mitigate toxicity in flies. We injected 0.03 mg/ml insulin into fly haemolymph, equivalent to blood, at the second and seventh day post-induction of (G4C2)36 expression, and determined the effect on fly survival. Insulin treatment significantly extended lifespan in three independent cohorts of flies (Figure 6 and Figure 6—figure supplement 1A and B), while modestly reducing lifespan in control, uninduced flies, again indicating a beneficial effect of insulin signalling specific to C9orf72 repeat toxicity. Higher concentrations of insulin became toxic (Figure 6—figure supplement 1A), indicating there is a therapeutic window within which insulin treatment is beneficial.

Drosophila stocks were maintained on SYA food (15 g/L agar, 50 g/L sugar, 100 g/L autolysed yeast, 30 ml/L nipagin [10% in ethanol], and 3 ml/L propionic acid) at 25°C in a 12 hr light/dark cycle with constant humidity. The Elav-GS stock was generously provided by Herve Tricoire (Paris Diderot University). The dilp3-Gal4 (#52660) driver and the over-expression InR constructs lines (UAS-InR^DN #8252; UAS-InR^Active #8263), PI3K^CA and Akt^CA (UAS-PI3K^CA #25908; Akt^CA #8194) were obtained from the Bloomington Drosophila Stock Centre. The GMR-Gal4 and UAS-(G4C2)36, UAS-GR100, and flies have been previously described in Mizielinska et al., 2014. The da-GS was kindly provided by Veronique Monnier (Tricoire et al., 2009), and the UAS-mCD8::GFP was a kind donation from Dr. Luo (Lee and Luo, 1999).

The parental generation of the genotype used in each lifespan assay was allowed to lay for 24 hr on grape agar plates supplemented with yeast. Eggs were placed at a standard density into bottles containing SYA medium. Adult experimental flies were allowed to emerge and mate for 2 days before being lightly anaesthetised with CO₂, and females randomly allocated onto SYA containing RU486 (200 μM) at a standard density per vial (n = 15), with a minimum 150 flies per condition. Flies were tipped onto fresh food every two days and dead flies counted. Escaping flies were censored from the data.

To detect differential gene expression upon adult-onset, neuron-specific over-expression of poly-GR100, flies carrying the UAS-GR100 transgene were crossed with Elav-GS driver flies. As a control, Elav-GS driver flies were crossed with wild-type flies. Experimental flies were generated as described above. Female flies were fed for three days with 200 µM RU486, and subsequently snap frozen. Total RNA was isolated from 25 fly heads using Trizol and treated with DNase. For sequencing total RNA was depleted of ribosomal RNA and libraries were generated at the Max Planck Genome Centre Cologne (Germany). This experiment was performed in triplicate. RNA sequencing was performed with an Illumina Hi-Seq2500 and 35 million single-end reads/sample and 100 bp read length at the Max-Planck Genome Centre Cologne. Raw sequence reads were quality-trimmed using Trim Galore! (v0.3.7) and aligned using Tophat2 (Kim et al., 2013) (v2.0.14) against the Dm6 reference genome. Multi-mapped reads were filtered using SAMtools (Li et al., 2009). Data visualization and analysis was performed using SeqMonk, and the following Bioconductor packages: Deseq2 (Love et al., 2014), topGO and org.Dm.eg.db. For visualization of functional enrichment analysis results, we further used the CellPlot package. Genes were considered to be significantly differentially expressed with an adjusted p value<0.05 and no cut-off for fold change was used. Unless stated otherwise, the set of expressed genes was used as background for all functional enrichment analyses involving expression data. The data have been deposited in NCBI's Gene Expression Omnibus (GEO) (Edgar et al., 2002) under the accession number GSE151826.

Adult female flies were induced on SYA medium containing 200 μM R4486 for 7 days before being flash-frozen in liquid nitrogen. RNA from 12 to 15 heads per replicate was extracted using TRIzol reagent (Thermo Fisher Scientific) following the manufacturer’s protocol. Approximately 1 μg of RNA per sample was treated with TURBO DNase (Thermo Fisher Scientific), followed by reverse transcription using the SuperScript II system (Invitrogen) with random hexamers (Thermo Fisher Scientific). Quantitative RT-PCR was conducted on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) using SYBR Green Master Mix (Applied Biosystems). Relative mRNA levels were calculated relative to alphaTub84B expression by the comparative Ct method. Primer sequences used are described in key resource table.

Brains from 4-day-old female flies were dissected in PBS and immediately fixed in 4% PFA in /PBS at 4°C for 2 hr and washed for 4 × 30 min in PBST (0.5% Triton X-100 in PBS). Fly brains were then blocked in PBST + 5% fetal bovine serum (FBS; Sigma #F524) for 1 hr at RT and incubated with 5H9 rat anti-poly GR (1:50) (Mori et al., 2013) or rabbit anti-dilp2 (Okamoto et al., 2012) in blocking buffer for 48 hr at 4°C. The tissues were washed 4× 30 min in PBST at RT and incubated with anti-rat IgG-Alexa Fluor 647 (ThermoFischer, catalog #A-21247) or anti-rabbit IgG-Alexa Fluor 488 (ThermoFischer, catalog #A-32731), diluted 1:1000 in blocking solution for 2 hr at RT, and washed 4× 30 min with PBST. The brains were then incubated 50% glycerol-PBS and mounted in Vectashield mounting medium (Vectorlabs, catalog #H-1200), and confocal stacks were taken with a 2 µm step size using a Leica SP8X confocal and a dry 20× (for whole brains) or a glycerol IMM 60× (for IPC zooms) objectives. The mean dilp2 immunofluorescence within each cluster of brain IPCs was calculated using the FIJI package (2.0.0-rc-43/1.51 p; NIH) software. Z projections (SUM projection) of image stacks were created. The mean fluorescence within a region adjacent to the IPCs served as background and was subtracted from the mean dilp2 fluorescence within the IPCs. Finally, a mean value representing each genotype/condition was calculated. Total numbers of IPCs were counted from each Z brain projection.

Total RNA of 25 female flies per genotype was extracted using Trizol and the Qiagen RNeasy Mini kit. For the dot blot analysis, 5 µg of RNA per sample were spotted onto a positively charged nylon membrane (GE Healthcare). The membrane was briefly washed with 10× SSC and RNA was then cross-linked to the membrane surface using a UVC 500 crosslinker (Amersham Biosciences). A (GGCCCC)₅ oligonucleotide probe was 5' labelled with γ[32P]-ATP using polynucleotide kinase to detect sense repeats. The membrane was prehybridized with ULTRAhyb-Oligo hybridization buffer (Thermo Fisher Scientific) for 1 hr at 42°C before adding the oligonucleotide probe. Hybridisation was carried out over night at 42°C. The membrane was washed twice for 30 min in 2× SSC/0.1% SDS and then exposed to X-ray films. After autoradiography the membrane was stripped by boiling in 0.1% SDS for 30 min. For normalization the blot was re-hybridized with a probe detecting ribosomal protein RpL32 transcripts. Dot intensities were quantified in Fiji.

Flies carrying UAS-InR constructs, UAS-PI3K^CA and UAS-Akt^CA were crossed to the GMR-GAL4; UAS-(G4C2)36 driver line. The progeny were allowed to develop and eclose at 25°C; female eyes were imaged 2 days after emergence. All eye images were obtained under the same magnification; eye area was calculated from each image using ImageJ (Schneider et al., 2012).

Two-day-old mated female flies (n = 32) developed and eclosed under 12 hr:12 hr light:dark cycle conditions (12L:12D) were fed with food containing either 200 µM RU486 or ethanol vehicle for 12 days. After transferring into tubes, locomotor activity and sleep behaviour were recorded over 4 days in 12L:12D using the Drosophila Activity Monitor (DAM, TriKinetics Inc, MA) system within the experimental incubator (Percival), set at 25°C and 65% humidity. Fly activity is measured by infra-red beam crosses in the DAM tube. After 2 days of acclimatisation, data were acquired from a 24 hr period on the third day (beginning at the onset of lights-on). A custom Microsoft Excel workbook (Chen et al., 2019) was used to calculate total activity counts per fly in the day and night periods, and to calculate sleep minutes during the day period (continuous periods of fly inactivity lasting 5 min or longer were classified as sleep). Flies with more than 12 hr of continuous inactivity at the end of the experiment were excluded as potentially dead.

Heads from female flies (n = 15) induced on SYA medium containing 200 μM RU486 for 7 days were collected and processed as described previously to measure poly-GR levels (Moens et al., 2018).

Heads from female flies (n = 15) induced with 200 μM RU486 for 7 days were collected and processed as previously described (Mizielinska et al., 2014). Membranes were incubated overnight at 4°C with primary antibodies: mouse anti-GFP (Ab#11814460001; MilliporeSigma) (1:10.000 in TBS-T); mouse anti-actin antibody (ab8224, Abcam – 1:10.000 in TBS-T); mouse anti-tubulin (T6199, Sigma-Aldrich – 1:2000); mouse anti-p53 (dmp53-H3, DSHB – 1:200); rabbit anti-non-P 4E-BP1 (4923, Cell Signaling – 1:1000); and rabbit anti-P 4E-BP1 (2855, Cell Signaling – 1:1000). HRP-conjugated anti-mouse and anti-rabbit secondary antibodies (ab6789 and ab6721, Abcam – 1:10.000) were used for 1 hr at room temperature.

Injections were performed twice, on the second day of repeat induction and 5 days later, by anesthetising the flies with CO₂. For each experiment, adult female flies (n = 80) were injected into the thorax with 32 nl of insulin (0.03 mg/ml) in PBS (pH 7.5) using a nanoinjector (Nanoject III; Drummond Scientific). Injection of the same volume of PBS acted as a control. Injected flies were then maintained at 25°C and transferred to fresh vials every third day throughout the experiment. They were collected at the indicated time points and directly processed for western blot analysis.

In the capillary feeder assay (CAFE), a single female fly was presented with liquid food using one 10 µl calibrated capillary per chamber (n = 15 per condition). Changes in liquid meniscus height were measured over 3 days at each capillary change. Feeding volume was calculated after background subtraction of measurements from control chambers without flies.

For dual-luciferase assays, mycoplasma-free HEK293T cells were used and maintained in DMEM media supplemented with 10% FBS, 4.5 g/L glucose, 110 mg/L sodium pyruvate, and 1× GlutaMAX and kept at 37°C with 5% CO₂. HEK293T cells were plated at a density of 30,000 cells per well in a 96-well plate. The following day, the cells were transiently transfected with 12.5 ng of firefly luciferase expression plasmid, and 2.5 ng of RAN translated poly-GR nanoluciferase reporter plasmid (92R-NL) using Lipofectamine 2000 according to manufacturer’s instructions. One hour post-transfection, cells were treated with 1 µM of either MK2206 (Cayman Chemicals, #11593), SF1670 (Merck, # SML0684), or a DMSO control. Each experiment consists of three technical replicate wells per condition, with experiments repeated three times independently. 48 hr post-transfection both firefly and nanoluciferase signals were measured using the Nano-Glo Dual Luciferase Assay according to manufacturer’s instructions, on the FLUOstar Omega (BMG Labtech) with a threshold of 80% and a gain of 2000 for both readings. The nanoluciferase reading was normalised to the firefly luciferase reading for each well to control for variable transfection efficiencies and this normalised value was further normalized to the control DMSO treatment.

Statistical analyses were performed with Prism6 (GraphPad Software). Normality of data was tested using the D’Agostino-Pearson omnibus normality test. When data were normally distributed, a Student’s t-test, one-way ANOVA, or two-way ANOVA was performed followed by multiple comparison test. For all data figures, the n values can be found in the figure legends and correspond to the number of biological repeats used in the analysis. Results were presented as mean ± 95% confidence intervals unless stated otherwise. Statistical differences were considered significant at p<0.05. Log-rank test on lifespan data were performed in Microsoft Excel (template available at http://piperlab.org/resources/) and data was plotted using Prism6.