Reduced C9orf72 function leads to defective synaptic vesicle release and neuromuscular dysfunction in zebrafish

The most common genetic cause of amyotrophic lateral sclerosis (ALS) and fronto-temporal dementia (FTD) is a hexanucleotide repeat expansion within the C9orf72 gene. Reduced levels of C9orf72 mRNA and protein have been found in ALS/FTD patients, but the role of this protein in disease pathogenesis is still poorly understood. Here, we report the generation and characterization of a stable C9orf72 loss-of-function (LOF) model in the zebrafish. We show that reduced C9orf72 function leads to motor defects, muscle atrophy, motor neuron loss and mortality in early larval and adult stages. Analysis of the structure and function of the neuromuscular junctions (NMJs) of the larvae, reveal a marked reduction in the number of presynaptic and postsynaptic structures and an impaired release of quantal synaptic vesicles at the NMJ. Strikingly, we demonstrate a downregulation of SV2a upon C9orf72-LOF and a reduced rate of synaptic vesicle cycling. Furthermore, we show a reduced number and size of Rab3a-postive synaptic puncta at NMJs. Altogether, these results reveal a key function for C9orf72 in the control of presynaptic vesicle trafficking and release at the zebrafish larval NMJ. Our study demonstrates an important role for C9orf72 in ALS/FTD pathogenesis, where it regulates synaptic vesicle release and neuromuscular functions.


Introduction
Amyotrophic lateral sclerosis (ALS) is a progressive and ultimately lethal neuromuscular disease involving the degeneration and loss of motor neurons. Current FDA-approved treatments for ALS are only modestly effective and the disease still results in complete paralysis and death within the ve rst years after diagnosis. GGGGCC hexanucleotide repeat expansions within the rst intron of C9or72 is the most common genetic cause of ALS and frontotemporal dementia (FTD) 1,2 . The pathogenic mechanism by which the repeat expansions cause disease may involve toxic gain-of-function (GOF) mechanisms, such as RNA toxicity 3 and protein toxicity by aberrant dipeptide repeat protein (DPR) accumulation 4,5 .
Alternatively, reduced C9orf72 mRNA and protein levels in a range of patient tissues and patient-derived cell lines 1,6,7 suggest loss-of-function (LOF) by C9orf72 haploinsu ciency may also contribute to C9orf72 ALS/FTD. The two GOF pathogenic mechanisms are extensively studied 8 , while the role of C9orf72-LOF in ALS pathogenesis remains poorly understood. Importantly, in general, how the GGGGCC hexanucleotide repeat expansions cause neurodegeneration in ALS and FTD is still uncertain. The C9orf72 protein has been shown to function in a complex with the WDR41 and SMCR proteins as a GEF for Rab8 and Rab39 9,10 . It has also been proposed to play a role in autophagic ux 9,11,12 , endosomal tra cking [13][14][15] and regulating AMPA receptor levels 16 .
Synaptic alterations at neuromuscular junctions (NMJs) have been found in ALS patients and in animal models of ALS. For instance, Killian et al observed that initial compound motor action potentials (CMAP) in ALS patients were of low amplitude but did not demonstrate early post-exercise facilitation (reduction in decrement occurred at 3 minutes post-exercise). The low baseline CMAP amplitudes with decrement may suggest a pre-synaptic transmission de cit 17 . In vitro microelectrode studies of ALS patient anconeus muscle demonstrated reduced pre-synaptic acetylcholine quantal stores, possibly explained by the diminished size of nerve terminals 17 . In mutant SOD1-expressing mice 18,19 an early retraction of presynaptic motor endings was observed long before the death of motoneurons 20 . Such an observation was also observed in tissue from patients with ALS 21 . In zebra sh, expression of mutant human TARDBP G348C mRNA or FUS R521H resulted in impaired transmission, reduced frequency of miniature endplate currents (mEPCs) and reduced quantal transmission at the NMJ 22,23 . C9orf72 is expressed presynaptically and postsynaptically 16 . The function of C9orf72 at synapses remains an interesting and largely unexplored, yet a full understanding of its synaptic function can extend its contribution to ALS pathogenesis and uncover therapeutic targets.
Zebra sh is a powerful tool for studying neurological diseases relevant to humans including ALS 24 .
Using a stable transgenic zebra sh model with reduced C9orf72 expression, we analyzed the effects of reduced C9orf72 function on the zebra sh's neuromuscular system. These zebra sh display behavioural de cits and early mortality observed in C9orf72-ALS patients. C9orf72 silencing resulted in impaired synaptic activity and downregulation of the synaptic protein, SV2a. Our ndings suggest that loss-offunction mechanisms underlie defects in synaptic function in ALS.

Generation of stable C9orf72-LOF model in zebra sh
To better understand the role of C9orf72-LOF in ALS/FTD pathogenesis, we generated a stable transgenic zebra sh gene-silencing model. A single conserved C9ORF72 ortholog is present in zebra sh on its chromosome 13. To achieve transgenic c9orf72 gene silencing in zebra sh, we used a recent miRNAbased gene-silencing approach developed for zebra sh 25 . Unlike morpholino-based knockdown approach, transgenic zebra sh lines that have been constructed to stably express miRNAs designed to target knockdown desired genes of interest have no apparent non-speci c toxic effects 26 . The miRNA knockdown technique consists in the use of transgenic DNA construct allowing the expression of synthetic miRNA targeting the 3' UTR of a gene-of-interest, here the endogenous zebra sh c9orf72 (Fig. 1a). As presented more in details in the method section, we designed 4x different miRNAs targeting speci cally c9orf72 (C9orf72-miR) that we inserted downstream of a dsRED marker and under the control of a ubiquitous promoter (Ubiquitin), the overall sequence was recombined into a mini-Tol2-R4R2 destination plasmid. To generate a transgenic line, this Tol2-DNA construct was co-injected with transposase mRNA in fertilized eggs at one-cell stage for enhanced genomic integration of the DNA construct 27 . To ease the selection of the founders/carriers, we also included an eGFP cassette under the crystallin promoter (Fig. 1b). Founders with eyes displaying GFP uorescence were selected and raised to generate a stable and heritable C9orf72-miR LOF line (hereafter referred as C9-miR). F1 transgenic sh gave a birth to a ratio of close to 50% positive GFP embryos when outcrossed with wild-type animals, suggesting the presence of a single genomic insertion.
We rst analysed C9orf72 silencing e ciency in our C9-miR line by RT-qPCR and western blotting. We showed a signi cant decrease in the level of C9orf72 mRNA (Fig. 1c) associated with a 50% decrease of C9orf72 protein (Fig. 1d, e). Altogether, these results indicate that our genetic approach e ciently reduces the C9orf72 protein levels in vivo and this C9-miR line can be used to understand the role of C9orf72 haploinsu ciency in ALS. C9orf72-LOF model shows early motor behavioural defects and reduced viability We did not observe any overt morphological abnormalities during embryonic development (0-5 dpf) in C9-miR sh (Fig. 2a). From 6-14 dpf, C9-miR larvae exhibited gradual morphological defects such as an unusual body curve and premature death (Fig. 2b,c). C9orf72 partial depletion importantly led to a signi cant decrease in survival at 10 dpf compared to wild-type controls; with a survival rate of 2-5% after 15 dpf (Fig. 2b).
We, next, examined whether normal zebra sh motor behaviour was affected in larval C9-miR zebra sh (4-11 dpf). To assess motor activity, larval zebra sh that did not display any of the abnormal morphological defects were selected and monitored using the automated Noldus Ethovision XT behaviour monitoring system. A signi cant decrease in motor activity was observed in C9-miR sh as compared to controls, as of 6 dpf (Fig. 2c,d). Such an impaired motor behaviour early on in C9-miR zebra sh is consistent with ndings that we and others have reported in several other zebra sh models of ALS 24,28−30 . C9orf72-LOF zebra sh model display adult hallmark features of ALS C9-miR sh that survive past 15 dpf were also studied at adult stages for hallmarks of ALS such as muscle atrophy, motoneuron death and paralysis. Hematoxylin & eosin (H&E) staining of cross-section of sh body trunk revealed that muscle in adult C9-miR exhibited severe atrophy (Fig. 3a), with a signi cant reduction in the thickness of the bres (Fig. 3b). Choline acetyltransferase (ChAT) staining is a hallmark feature of cholinergic motor neurons. ChAT immunostaining was performed on the spinal cord sections of adult C9-miR sh and the mature motor neurons in the C9-miR sh were reduced in size by 19.2 ± 0.02 % (Fig. 3c). At the motor behavioural level, we observed an impaired swimming ability in C9-miR compared to controls (Fig. 3d, Supplemental Videos). Prior to death, C9-miR sh spent their time in the bottom of the tank with weak movements. Adult survival was also monitored and we observed that by 16 months post-fertilisation, more than 80-90% of the adult C9-miR zebra sh die.
Cytoplasmic aggregation of Trans-activation response element (TAR) DNA-binding protein 43 (TDP-43) is a major pathological hallmark of ALS 31 . TDP-43 form aggregates in neurons, glial cells 31 and axial skeletal muscle 32 . By taking advantage of the relatively large nucleus and cytoplasm of skeletal muscle cells, we examined whether TDP-43 pathology exist in our model. Using a speci c antibody that recognizes the highly homologous human TDP-43 ortholog in zebra sh 33 , we showed that this protein is localized to the nucleus of the skeletal muscle cells in wild-type zebra sh (Fig. 4a). In contrast, in C9-miR zebra sh, we observed clusters of TDP-43 in skeletal muscles (Fig. 4b). We then analyzed these clusters further to examine their precise cellular localization and found that they are predominantly located outside of the nucleus. Altogether, our ndings provide strong evidences that C9orf72 silencing in zebra sh recapitulates key pathological hallmarks of ALS.

C9orf72 silencing affects NMJ structural integrity and quantal release
We next examined NMJ integrity by performing double-immunohistochemistry on xed embryos using speci c presynaptic (SV2) and postsynaptic markers (a-bungarotoxin). Analysis revealed no change the primary motor neuron axon architecture and in colocalization of pre-and post-synaptic signals in C9-miR sh at 2 dpf (Fig. 5a,b) and 4 dpf. However, in 6 dpf C9-miR larvae, we observed a signi cant reduction in the number of colocalizing pre-and post-synaptic puncta (Fig. 5c,d). These results indicate that, while the synaptic structures of the NMJ develop properly and are preserved at early embryonic stages in C9-miR, they do start to degenerate from 6 dpf.
To investigate if alterations in NMJ integrity had functional consequences on synaptic transmission in the 6 dpf C9-miR larvae, we recorded and analysed the spontaneous miniature end plate currents (mEPCs) that occur naturally and spontaneously at synapses and represent the unitary event during synaptic transmission (Fig. 6a). We observed that the frequency of mEPCs in C9-miR was signi cantly reduced compared to controls (Fig. 6b), suggesting a reduction in the number of functional presynaptic endings. The mean amplitude of mEPCs was also found to be smaller in zebra sh C9-miR compared to wild-type zebra sh (Fig. 6c). We observed that the mEPCs from the muscle of C9-miR larvae and controls shared similar rise time and decay time constant kinetics (Fig. 6d).
C9orf72 regulates synaptic vesicle exocytosis and synapse stability at the NMJ To gain more insights into molecular processes and pathways affected, we determined global changes at the proteomic levels by isolating total proteins at 6 dpf from C9-miR and wild-type siblings. We identi ed a total of 2602 proteins that were covered by two or more uniques peptides and were quanti able in four biological replicates (FDR≤1%). Most of the proteins in wild-type and C9-miR were at comparable expression levels. Only 24 proteins were found to be dysregulated (p<0.05; Table S2). Of these hits, 12 were upregulated and 12 were downregulated in C9-miR sh (Fig. 7a). These differentially expressed proteins (DEPs) were classi ed into functional clusters according to the PANTHER classi cation system ( Fig. 7b-e). The classi cation results revealed that many DEPs were distributed into six protein classes (Fig. 7b). These proteins are classi ed in three molecular functions namely binding (20%), structural molecule activity (20%) and catalytic activity (60%) (Fig. 7c). They are involved in biological processes, being cellular process, metabolic process and biological regulations the most represented ones with 38%, 23.1% and 15.4% of proteins respectively (Fig. 7d). Cellular component analysis revealed that the DEPs belong in majority to the organelle, membrane and synapse categories (Fig. 7e). Consistent with the synaptic dysfunction phenotype, we identi ed a strong downregulation of synaptic proteins ( Fig. 7a; Table S2). Among these proteins, the top hit of dysregulated proteins is the synaptic protein, synaptic vesicle-associated protein 2a (SV2a). Importantly, a recent study showed that SV2a is reduced in C9orf72-ALS patient-derived IPS neurons 34 . This data links the ndings in our C9orf72 loss-of-function model to ALS.
Given that SV2a is an essential component of active zones and synaptic release machinery, we next sought to further assess synaptic activity at the NMJ by measuring synaptic vesicle (SV) cycling at the NMJ in zebra sh larvae using the uorescent styryl dye, FM1-43 35,36 . C9-miR and controls larvae were exposed to FM1-43 and its uptake into NMJ presynaptic boutons was monitored. The presynaptic terminals were acutely depolarized with a high [K+] HBSS solution (45 mM) to drive the exocytotic activity, SV cycle and load FM1-43 and label synaptic clusters. In controls, we observed strong uorescence staining along terminal axon branches at individual synaptic varicosity boutons (Fig. 8a). While in C9-miR sh we found a signi cant reduction in FM1-43 loading in presynaptic terminals (Fig.  8b), indicating slowing of the exocytotic activity and the overall SV cycle. These ndings reveal a key role for C9orf72 in regulating presynaptic vesicle release at NMJ.
To assess organization of the presynaptic structure at NMJ, we examined the expression of Rab3a, a protein associated with vesicles at active zones that is essential for synaptic vesicle release and neurotransmission ( Fig. 8c-e). We found a reduced number of Rab3+ puncta in C9-miR sh compared to controls (Fig. 8c-d) as well as the area of the putative synapses were smaller in C9-miR sh (Fig. 8e).

Discussion
Despite advances in studies of C9orf72-ALS, understanding the function of C9orf72 remains a key research element that is poorly explored. We generated a C9orf72-related ALS stable zebra sh line with a reduced expression of C9orf72. These sh display motor defects, muscle atrophy, motor neuron loss and mortality in early larval and adult stages. Additionally, they exhibit TDP-43 pathology, which is a key hallmark of ALS. Analysis of the structure and function of the NMJs, revealed a signi cant reduction in the number of presynaptic and postsynaptic structures and an impaired release of quantal synaptic vesicles at the NMJ in the C9-miR line. We also identi ed a novel role of C9orf72 in controlling presynaptic vesicle tra cking and release at the zebra sh larval NMJ.
Reduced C9orf72 mRNA and protein levels in a range of patient tissues and patient-derived cell lines 1,6,7 . Our C9orf72 zebra sh model provides support to a loss-of-function mechanism underlying C9orf72dependant ALS. Our data are consistent with deletion or transient knockdown models in C. elegans 37 and zebra sh 38 respectively, showing defective motor phenotypes. However, in contrast, no motor neurons de cits were reported in C9orf72 knock-out mice [39][40][41] . In addition, these mice also did not exhibit TDP-43 proteinopathy. The model presented here, importantly, display TDP-43 pathology and replicates haploinsu ciency as a major contributor to C9orf72 ALS rather than a full ablation of C9or72 loss-offunction model. Intriguingly, the motor phenotypes observed in C9-miR zebra sh are consistent with several other zebra sh ALS models, including zebra sh model expressing C9orf72-related repeat expansions or DPR 28,42,43 . However, the presence a reduced level of C9orf72 mRNA or protein in these models, as in ALS/FTD, was not examined in these studies. Of note, the expression of GGGGCC repeat expansions or DPR in zebra sh are toxic 42,44,45 , consistent with several studies in neurons and other animals. We found that expression of GGGGCC repeat expansions in our C9-miR exacerbated toxicity and resulted in death of zebra sh by 6 dpf (Figure S1). Such a synergistic interplay between reduced C9orf72 function and repeat-dependent gain of toxicity was observed in a recent study by Zhu and colleagues 46 .
An important nding of this study is the synaptic impairments in C9-miR sh. The reduced frequencies and amplitudes of quantal neurotransmission events are consistent with observations made in several non-C9orf72 ALS models 22,23 and in tissue from patients with ALS 21 . We also report signi cant reductions in synaptic vesicle exocytosis and number and area putative synaptic puncta at NMJs. Additionally, we show a decrease in the expression of synaptic vesicle protein SV2a. These ndings provide a novel role of C9orf72 in synaptic physiology at the presynaptic level. Interestingly, consistent with our ndings, SV2a was also recently found at reduce levels in C9orf72-ALS patient-derived IPS neurons 34 . Ablation of SV2a function in knockout models resulted in reduced number of readily releasable pool of synaptic vesicles, diminished release probability and reduction in spontaneous synaptic events 47,48 . Whether C9orf72 directly or indirectly regulates the level of SV2a in presynaptic compartments remains to be investigated. Intriguingly, similar observations of loss of SV2a and synaptic dysfunction were also observed in neurons expressing the C9orf72-related glycine-alanine (GA) DPR 34 .
DPR proteins can disrupt pre-mRNA splicing in ALS/FTD patients 49 . It is possible that the expression of GA DPR in neurons reduce the level of C9orf72 transcripts leading to the synaptic phenotypes.
Rab3a is important for transport of synaptic vesicles and their docking at active zones 50 . It regulates synaptic transmission and it is associated with synaptic vesicles through GEF activity 51,52 . For instance, at rab3a-de cient terminals in mice, synaptic secretion response recovered slowly and incompletely following exhaustive stimulation 50 . In addition, the replenshiment of docked vesicles following exhaustive stimulation at these terminals was also impaired 50 . DENN domain containing proteins such as C9orf72 can function as Rab GEFs, enabling their activation, recruitment and interaction with downstream effectors 53,54 . A previous study had identi ed Rab3a as part of complex interacting with C9orf72 7 . It is plausible that in addition to the effect of reduced SV2a on synaptic dysfunction, synaptic vesicle exocytosis and quantal transmission defects in C9-miR maybe exacerbated due to the altered function of C9orf72 as a GEF for Rab3a and its recruitment to synaptic vesicles.
In conclusion, we generated a stable C9orf72 LOF model in zebra sh that recapitulated some major hallmarks of ALS and enhanced our understanding of ALS pathogenesis. Importantly, our ndings demonstrate that loss of C9orf72 function impairs synaptic function at NMJs and result in motor de cits. We postulate that synaptic de cits observed in repeat expansions or DPR models maybe the result of an indirect effect related to an impact of the repeats on C9orf72 levels.

Zebra sh Husbandry
Adult zebra sh (Danio rerio) were maintained at 28 °C at a light/dark cycle of 12/12 h in accordance with Wester eld zebra sh book 55 . Embryos were raised at 28.5 °C, and collected and staged as previously described 56 . All experiments were performed in compliance with the guidelines of the Canadian Council for Animal Care and the local ethics committee.

Anti-c9orf72 (synthetic miRNA) RNAi target site selection
We rst generated a template with c9orf72 RNA sequence, including 5'-and 3'-UTR sequence. 3'-UTR minimal sequence has been obtained from analysis of data available on ensembl (http://asia.ensembl.org/) with zebra sh GRCz11 genome iteration and on Targetscan Fish website (http://www.targetscan.org/ sh_62/). We analysed and annotated c9orf72 sequence for identifying and avoiding selecting target sequence that would run across i) potential polymorphisms in the 3'UTR sequence and ii) endogeneous miRNA. Based on these data we selected 4x unique target sites on the 3'UTR sequence of c9orf72 that do not show any off-speci c match across the zebra sh genome. Each site and corresponding mature anti-c9orf72 synthetic miRNA are presented in Table S1.

Injections for transgene integration
To integrate Tol2-UBI:dsRED:c9orf72-1234-Cryst:eGFP construct into the zebra sh genome, 1 nl of a mix of 30 ng/µl of construct and 25 ng/µl of Transposase mRNA was injected into one-cell stage embryos using the Picospritzer III pressure ejector.
Gene expression study RNA extraction was performed on 2 dpf larvae using Trizol reagent (Sigma). 30 larvae were selected and lysed on ice for 30 seconds with 250 µL of Trizol reagent, 250 µL of reagent were added before incubating the samples at room temperature for 5 minutes. 100 µL of Chloroform were added before vortexing the samples and incubated à RT° for 2 minutes and centrifuged for 15 minutes at 4 °C at 12000 rpm. To aqueous phase was transferred in a new tube and an equivalent volume of isopropanol was added. Samples were mixed by inversion, incubated 10 minutes at RT° and centrifuged for 10 minutes at 4 °C at 12,000 rpm. Pellet was then washed with 75% ethanol, left to dry for 5 minutes and resuspended in 10 to 30 µL of DEPC H 2 O. Samples were then dosed using Nanodrop device (ThermoScienti c). 1 µg of RNA from each sample was used for retro transcription performed with the superscript VILO reverse transcription mix (Invitrogen). PCR was performed on 1 µL of cDNA using BioRad mix. Elf1a was used as control gene expression.

GGGGCC expansion repeat microinjections
GGGGCC repeat constructs (p3s and p91s) were kindly provided by Dr. Ludo Van Den Bosch and Dr.
Adrian Isaacs. Synthesis of mRNAs and microinjections were performed as previously described 43 . Survival assay Zebra sh larvae were screened for GFP positive eyes at 2 dpf and split in 4 petri dishes of 25 shes. Dead shes were counted and reported everyday during 17 days. Larvae were fed starting 7 dpf and dishes were cleaned twice a day.

Behavioural assay
Zebra sh control or C9orf72-mut larvae were transferred individually into a 96-well plate and locomotor activity was recorded using Basler GenIcam camera and DanioVision recording chamber (Noldus). After 30 of dark exposition, shes were exposed to light for two hours. Analysis was performed using the Ethovision XT 12 software (Noldus) to quantify the distance swam.
Immuno uorescence staining Zebra sh at 2 and 6 dpf were xed in 4% PFA at 4 °C, overnight, washed in PBS-Tween the next day, incubated in 1 mg mL − 1 collagenase (for 30 minutes for 2 dpf shes and 180 minutes for 6 dpf shes) and collagenase was washed in PBS-Tween. Fishes were incubated in blocking solution (2% NGS, 1% BSA, 1% DMSO, 1% Triton-X in PBS) for an hour and in α-bungarotoxin (Thermo sher T1175) for 30 minutes. After washing and blocking for an hour, shes were incubated with primary antibody (SV2, 1:200; Rab3a, 1:100) overnight at 4 °C. Next day, shes were washed and incubated in secondary antibody (goat anti mouse 488, 1:1000) overnight at 4 °C. Next day, shes were washed and mounted in 80% glycerol. Slides were imaged with a Zeiss confocal microscope.

Hematoxylin & Eosin staining
For the muscle and spinal cord staining, 4 whole shes of each genotype were xed in 4% PFA for 72 h. Fishes were embedded in para n and 15 µm sections were obtained using microtome. Para n was removed by incubating slides in xylene twice and samples were rehydrated with 4 successive baths of respectively 100%, 95%, 70% and 50% ethanol in distilled water. Hematoxylin and eosin staining was then performed.

Motoneuron staining
Adult zebra sh body trunk was crossed-sectioned using a microtome at 15 µm thick slices. Para n was removed, and samples were rehydrated using same technique as described above. Samples were then rinsed in PBS several times and unmask antigen step was made using citrate buffer (1M, pH 6). Citrate was washed off with several PBS baths and incubated in 0,3% Triton-X. Samples were incubated in blocking buffer (1% DGS, 0,4% Triton-X) for an hour. Primary antibody was added at 1:500 and left overnight at 4 °C. Samples were washed on the next day in PBS, blocked again and incubated in secondary antibody at 1:750.

FM1-43 staining
Zebra sh larvae were rst anesthetized in Evans solution (134 mM NaCl, 2,9 mM KCl, 2,1 mM CaCl, 1,2 MgCl 2 , 10 mM Hepes, 10 mM glucose) containing 0,02% tricaine before being pinned on a Sylgard coated dish under the head and at the tail. Skin was then carefully removed in order to expose the muscles. Pinned sh was then exposed to Evans solution containing 10 µM of FM1-43 during 10 minutes to allow preloading penetration of the dye. Then the sh was exposed to HBSS solution with high potassium concentration (97 mM NaCl, 45 mM KCl, 1 mM MgSO 4 , 5 mM Hepes, 5 mM CaCl 2 . and 10 µM of FM1-43 in order to load the dye at the synaptic cleft during 5 minutes. 3 more minutes in Evans solution with 10 µM of FM1-43 nished the loading and nally, the sh was washed with a low calcium Evans solution (0,5 mM CaCl 2 ) three time for 5 minutes. The sh was then imaged using a 40X Examiner microscope (Zeiss).

Electrophysiology recordings
Muscle recordings were performed on shes that had been previously anesthetized in an extracellular solution (134 mM NaCl, 2,9 mM KCl, 1,2 mM MgCl, 10 mM Hepes, 10 mM glucose, pH 7,8) containing 0,02% tricaine before being pinned on a Sylgard coated dish under the head and at the tail. Skin was then carefully removed in order to expose the muscles. Fish was then washed with the extracellular solution containing 1 µM of tetrodotoxine in order to avoid action potentials ring. Patch clamp electrodes (1-2 MΩ) were lled with an intracellular solution (130 mM CsCl, 8 mM NaCl, 2 mM CaCl, 10 mM Hepes, 10 mM EGTA, pH 7,4). Calcium currents were recorded in the whole-cell con guration with an Axopatch 200B patch-clamp ampli er (Molecular Devices), and series resistance was compensated by at least 85% using the ampli er's compensation circuitry. Voltage protocol generation and data acquisition were performed using the pCLAMP10 software (Molecular Devices).

Mass spectrometry sample preparation
Proteins were extracted with the protocol used for western blot protein extraction. Then, a 1:8:1 ratio was used to precipitate proteins, 1X of cell lysates, 8X of 100% ice cold acetone and 1X of 100% tricholoroacetic acid in low binding protein tubes. 20 µg of proteins were precipitated. Proteins were incubated at -20 °C for 12 hours and centrifuged at 11,500 rpm for 15 minutes at 4 °C. Supernatant was then discarded.
A standard TCA protein precipitation was rst performed to remove detergents from the samples (or acetone precipitation). Protein extracts were than re-solubilized in 10 µL of a 6M urea buffer. Proteins acetonitrile/0.2% formic acid (buffer B). Peptides were loaded on-column at a owrate of 600 nL/min and eluted with a 2 slope gradient at a owrate of 250 nL/min. Solvent B rst increased from 2 to 40% in 120 min and then from 40 to 80% B in 20 min. LC-MS/MS data was acquired using a data-dependant top17 method combined with a dynamic exclusion window of 7 sec. The mass resolution for full MS scan was set to 60,000 (at m/z 400) and lock masses were used to improve mass accuracy. The mass range was from 360 to 2000 m/z for MS scanning with a target value at 1e6, the maximum ion ll time (IT) at 100 ms, the intensity threshold at 1.0e4 and the under ll ratio at 0.5%. The data dependent MS2 scan events were acquired at a resolution of 17,500 with the maximum ion ll time at 50 ms and the target value at 1e5. The normalized collision energy used was at 27 and the capillary temperature was 250ºC. Nanospray and S-lens voltages were set to 1.3-1.7 kV and 50 V, respectively.
The peak list les were generated with Proteome Discoverer (version 2.3) using the following parameters: minimum mass set to 500 Da, maximum mass set to 6000 Da, no grouping of MS/MS spectra, precursor charge set to auto, and minimum number of fragment ions set to 5. Protein database searching was performed with Mascot 2.6 (Matrix Science) against the Refseq Danio Rerio protein database. The mass tolerances for precursor and fragment ions were set to 10 ppm and 0.1 Da, respectively. Trypsin was used as the enzyme allowing for up to 1 missed cleavage. Cysteine carbamidomethylation was speci ed as a xed modi cation, and methionine oxidation as variable modi cations. Data analysis was performed using Scaffold (version 4.8).

Statistical analysis
All zebra sh experiments were performed on at least three replicates (N) and each consisted of a sample size (n) of 5-30 sh. Data are presented as Mean ± SEM. Signi cance was determined using either Student's t-test or One-way ANOVA followed by multiple comparisons test. A Tukey post-hoc multiple comparisons test was used for normally distributed and equal variance data. Kruskal-Walllis ANOVA and Dunn's method of comparison were used for non-normal distributions. All graphs were plotted using the Graphpad PRISM software. Signi cance is indicated as * p < 0.05, ** p < 0.01 and *** p < 0.001.

Con icts of interest
The authors declare no competing interests. Figure 1 Generation of a stable zebra sh C9orf72 knockdown line. a. Schematic representation of the technique use to silence C9orf72 in zebra sh. The transgene is designed to express four different micro-RNAs targeting C9orf72' 3'UTR and triggering knockdown by both repressing C9orf72 translation and affecting its stability. b. Snapshot demonstrating proper eGFP expression in the crystallin of the transgenic sh, a marker used to identify carrier/knockdown larvae. c. Histogram shows the relative expression of the endogenous C9orf72 gene. mRNA was normalized to elf1α mRNA levels (n=4, p value<0.0001). d. Immunoblotting of the Zebra sh protein C9orf72 and actin as control. e. Histogram shows the relative expression of the C9orf72 protein compared with actin between C9orf72 mutants and control line (n=3, p value=0.0034). *** denotes p<0.0001; ** denotes p<0.005. Data are presented as mean±SEM.  Adult zebra sh C9-miR display muscle atrophy, smaller motoneurons and behavioural de cits. a. Representative Hematoxylin and Eosin staining of adult zebra sh muscle myotomes. b. C9-miR sh display a smaller diameter of muscle bres compared to controls (N=10; p value<0.0001). c. ChAT staining in adult zebra sh spinal cord. Large motor neurons (broken circle) are reduced in size in C9-miR compared to controls. d. Representative traces of swimming activity of ve adults control and C9-miR shes during thirty seconds (left panel). C9-miR sh exhibit behavioural de cits (right panel). Data are presented as mean±SEM. *** denotes p<0.0001. Scale bar = 50 M.  Zebra sh C9-miR displayed reduced acetylcholine receptor clusters at neuromuscular junctions. a. Representative images of co-immunostaining of zebra sh neuromuscular junctions with presynaptic (SV2a) and postsynaptic (α-bungarotoxin) markers in 2dpf zebra sh. Scale bar = 50 M. b. Quanti cation of the colocalizing pre-and post-synaptic markers per somite showed no differences between C9-miR and controls at early embryonic stages (2 dpf; N=8; p value =0.064). c. Representative images of coimmunostaining of zebra sh neuromuscular junctions with SV2a and α-bungarotoxin in 6dpf zebra sh. Scale bar = 100 M. d. Quanti cation of the colocalizing pre-and post-synaptic markers per somite showed a signi cant reduction in puncta in C9-miR sh at late larval stages embryonic stages (6 dpf; N=8). *** denotes p<0.0001. Zebra sh C9-miR exhibited attenuated miniature endplate currents (mEPCs) at NMJs. a. Recordings of mEPCs, which result from spontaneous release of a quantum, were recorded in 6dpf controls and C9-miR.

Figures
Animals with reduced C9orf72 (C9-miR) displayed mEPCs with reduced frequency (b) and amplitude (c). Rise time (d) and decay time (e) constant kinetics of mEPC was not found to be signi cantly different.