Muscle cells of sporadic amyotrophic lateral sclerosis patients secrete neurotoxic vesicles

Abstract Background The cause of the motor neuron (MN) death that drives terminal pathology in amyotrophic lateral sclerosis (ALS) remains unknown, and it is thought that the cellular environment of the MN may play a key role in MN survival. Several lines of evidence implicate vesicles in ALS, including that extracellular vesicles may carry toxic elements from astrocytes towards MNs, and that pathological proteins have been identified in circulating extracellular vesicles of sporadic ALS patients. Because MN degeneration at the neuromuscular junction is a feature of ALS, and muscle is a vesicle‐secretory tissue, we hypothesized that muscle vesicles may be involved in ALS pathology. Methods Sporadic ALS patients were confirmed to be ALS according to El Escorial criteria and were genotyped to test for classic gene mutations associated with ALS, and physical function was assessed using the ALSFRS‐R score. Muscle biopsies of either mildly affected deltoids of ALS patients (n = 27) or deltoids of aged‐matched healthy subjects (n = 30) were used for extraction of muscle stem cells, to perform immunohistology, or for electron microscopy. Muscle stem cells were characterized by immunostaining, RT‐qPCR, and transcriptomic analysis. Secreted muscle vesicles were characterized by proteomic analysis, Western blot, NanoSight, and electron microscopy. The effects of muscle vesicles isolated from the culture medium of ALS and healthy myotubes were tested on healthy human‐derived iPSC MNs and on healthy human myotubes, with untreated cells used as controls. Results An accumulation of multivesicular bodies was observed in muscle biopsies of sporadic ALS patients by immunostaining and electron microscopy. Study of muscle biopsies and biopsy‐derived denervation‐naïve differentiated muscle stem cells (myotubes) revealed a consistent disease signature in ALS myotubes, including intracellular accumulation of exosome‐like vesicles and disruption of RNA‐processing. Compared with vesicles from healthy control myotubes, when administered to healthy MNs the vesicles of ALS myotubes induced shortened, less branched neurites, cell death, and disrupted localization of RNA and RNA‐processing proteins. The RNA‐processing protein FUS and a majority of its binding partners were present in ALS muscle vesicles, and toxicity was dependent on the expression level of FUS in recipient cells. Toxicity to recipient MNs was abolished by anti‐CD63 immuno‐blocking of vesicle uptake. Conclusions ALS muscle vesicles are shown to be toxic to MNs, which establishes the skeletal muscle as a potential source of vesicle‐mediated toxicity in ALS.


Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal adult-onset motor neuron disorder affecting 3-5/100 000 individuals per year. 1 The cause of pathology is likely complex with onset resulting from some combination of genetic mutations, DNA damage, environmental risk factors, viral infections, or other factors, leading to diverse cellular dysfunction such as glutamate-mediated excitotoxicity, abnormal protein aggregation, and mitochondrial disorganization and dysfunction contributing to oxidative stress (refer to Vijayakumar et al. 2 and Le Gall et al. 3

for review).
Not only motor neurons (MN) are affected in ALS, but also glial cells, muscle fibres, 4 and immune cells, 5 each of which may participate actively in ALS onset and progression. In a FUS murine model of motor neurone disease, where the FUS mutation is expressed in all tissues except the MN, motor deficits still appear at a late stage of the disease. 6 In addition, when the human mutated SOD1(hSOD1) is selectively knocked-down in MN and astrocytes of newborn transgenic hSOD1 mice, their life expectancy is prolonged by only 65-70 days, and there is still significant astrogliosis and microglial activation. 7 These studies support the potential role of 'non-cell autonomous' MN death in ALS, which could involve multiple cell types. 8 Numerous cell and tissue types, including skeletal muscle, can secrete exosomes and other types of vesicle. 9,10 Such extracellular vesicles may be involved in cell-cell communication in the central nervous system, 11,12 where they can carry out intercellular transport of functional proteins, mRNA, miRNA, and lipids and may have key roles in spreading of proteinopathies 11,12 or neurotoxic elements. 13 For instance, astrocytes extracted from the SOD1 murine model of ALS secrete exosomes that contain hSOD1 and propagate this toxic protein to neighbouring MNs. 14 The present study sought to determine whether vesicle secretion is affected in muscle cells of ALS subjects, and whether ALS muscle vesicles (MuVs) could be toxic when taken up by recipient motor neurons.

Participants and ethical approvals
An open biopsy was performed on deltoid muscles of 27 ALS patients with probable or definite ALS according to the revised El Escorial criteria, 15 (who attended the Motor Neuron Diseases Center (Pitié Salpétrière, Paris). To guarantee a consistent quality of muscle tissue, deltoid samples were taken using the procedure that is consistently and routinely followed for all muscle sampling at the centre (ALS Referral Center, Pitie Salpetriere, Sorbonne University). The affectedness of each patient's deltoid muscle was assessed directly by the Manual Muscle Testing scale, against which all muscles sampled were determined to be at least at Level 3 indicating muscle movement through full range of motion against gravity. Genetic analyses were carried out on DNA extracted from blood samples for all ALS patients to screen several ALS-related genes: the C9orf72 hexanucleotide repeat expansion (using gene scan and repeat primed PCR procedures described in Millecamps et al. 16 ), ATXN2 repeat length, 17 and the coding regions of SOD1, TARDBP, FUS, UBQLN2, and TBK1 (sequences of the used primers are available upon request). Thirty deltoid muscle biopsies from healthy age-matched and gender-matched subjects were obtained from the BTR (Bank of Tissues for Research, a partner in the EU network EuroBioBank) in accordance with European recommendations and French legislation. The main demographic, clinical, and genetic characteristics of the subjects are indicated in Table 1.
The protocols (NCT01984957) and (NCT02360891) were approved by the local Ethical Committee and all subjects signed an informed consent in accordance with institutional guidelines.
Muscle stem cell extraction and culture Briefly, muscles biopsies were dissociated mechanically as previously described in Bigot et al. 18 and plated in proliferation medium [1 volume of M199, 4 volumes of Dulbecco's Subject age at time of biopsy is indicated in the column 'Age'. All amyotrophic lateral sclerosis (ALS) patients were confirmed to be ALS according to El Escorial. The ALSFRS-R and Muscle Testing results measured at the time of the biopsies are given. The manual muscle testing score is the sum of 30 measurements (15 muscle groups assessed once on each side of the body), each scored from 0, representing total paralysis, to 5, representing normal strength, according to the Medical Research Council Score.  Healthy  Healthy  Healthy  Healthy  Healthy  Healthy  x  x  x  Healthy  x  x  x  Healthy  x  x  Healthy  x  x  Healthy  x  x  x  x  Healthy  x  x  x  x  Healthy  x  x  x  x  x  Healthy  x  x  x  Healthy  x  x  x  x  x  Healthy  x  x  x  x  Healthy x Healthy x Healthy x Healthy x Healthy x modified Eagle's medium (DMEM), 20% foetal bovine serum (v:v), 25 μg mL À1 Fetuin, 0.5 ng mL À1 bFGF, 5 ng mL À1 EGF, 5 μg mL À1 insulin]. The myogenic cell population was enriched using CD56 magnetic beads, and for their myogenicity using anti-desmin antibodies as described before. 18 A minimum of 80% of the cell population was positive for desmin. After rinsing three times the proliferative myoblasts with phosphate-buffered saline (PBS), and three times with DMEM to remove any FBS residual, the human muscle stem cells were differentiated into myotubes by culturing them in DMEM for 3 days. All cell cultures were regularly checked every 3 weeks for mycoplasma test. All experiments were conducted in cells at less than 21 divisions to avoid potential cellular senescence, and thus experimental artefacts.
Muscle vesicle extraction from culture medium For each replicate, 7.5 × 10 6 primary myoblasts with less than 21 divisions were plated in 225 cm 2 Flask. After 24 h, the cells were rinsed 6 times in DMEM, then differentiated into myotubes for 3 days in DMEM. All cell cultures were checked and negative for mycoplasma. Muscle vesicles were extracted from conditioned media as previously described for large-scale isolation protocol. 19,20 Briefly the conditioned media were centrifuged at 260 g for 10 min at room temperature, then at 4000 g for 20 min at 4°C to remove any dead cells and cell debris, and finally at 20 000 g for 1 h at 4°C to remove microparticles. The subsequent supernatant was then filtered through 0.22 μm filter to remove any microparticles leftover. The filtered medium was then mixed with total exosome isolation reagent (Life technologies ™ ; 2:1, v:v), incubated overnight at 4°C, and then centrifuged at 10 000 g at 4°C for 1 h. The supernatant was discarded, and the pellet containing the MuVs was resuspended and rinsed three times in PBS using 100 K MWCO column. The 100 μL MuVs suspensions were kept at À80°C until needed. MuVs were either used for treating cell cultures (iPSC-derived motor neurons, or myotubes; treatments being always compared with untreated cells) or for protein content characterization. MuVs protein was extracted using 8 M Urea or NuPAGE buffer and quantified using BCA kit. Refer to Figure S2A and supporting information. The vesicles extracted using this protocol floated at similar density than when extracted by classic ultracentrifugation, 1.15-1.19 g mL À1 , with a better yield was observed. 20 Vesicles were positive for CD63, CD81, CD82, Flotillin, ALIX, and negative for calnexin ( Figure 2 and Le Gall et al. 20 ).

NanoSight
The MuVs pellets were resuspended in 100 μL of filtered PBS. The MuV suspension was then diluted 10× in PBS. Size and distribution of MuV secreted by primary muscle cells were evaluated by a NanoSight LM10 instrument (NanoSight) equipped with NTA analytic software (version 2.3 build 2.3.5.0033.7-Beta7). Samples were assessed three times as previously described 21,22 at temperature set to 22.5°C. The minimum particle size, track length, and blur were set to 'automatic'.

Muscle vesicles labelling
The MuVs were labelled using PKH26 kit (Sigma-Aldrich®). Briefly, after adding 100 μL of Diluent C to the MuV suspension, 100 μL of 4 μM PKH26 solution were added to the sample. After 5 min of incubation, 1 mL PBS was added, and the MuVs were washed using a 100 K concentrators, 15 000 g at 4°C for 10 min. The MuVs were washed three times in PBS using the 100 K concentrators before being mixed with the cell media for treatment. All cell cultures treated with MuVs were compared and normalized to untreated cell cultures.

Muscle vesicles added to iPSC motor neurons
The hiPSC-derived motor neurons were obtained as previously described. 23 There were 3000 MN progenitors differentiated for 9 days then plated in poly-L-ornithine (

Muscle vesicles pre-treatment with CD63 antibody
After labelling 0.5 μg MuVs with PKH26 as described earlier, the MuV suspension was incubated for 2 h at RT with 0.5 μg of CD63 antibody (TS63, Life Technologies) and then added to the culture medium of hiPSC-derived motor neurons as described in the paragraph 'Muscle vesicles added to iPSC motor neurons'. All cell cultures treated with MuVs were compared and normalized with untreated cell cultures.
Muscle vesicles added to healthy human muscle cells Labelled MuVs were added to the differentiation medium of 200 000 control cells cultured in Ibidi 35 mm μ-Dishes. MuVs absorption occurred during the first 3 days of differentiation. The myotubes were then rinsed three times with PBS, and fresh DMEM was added to the petri-dishes. The cells were fixed with 3.6% formaldehyde for 15 min at room temperature at Day 3 or Day 7 of differentiation, then washed three times in PBS and stored at 4°C until subsequent analysis: • Myonuclear domain-The myotubes were fixed and stained for DAPI and MF20 as described earlier. The myonuclear domain was calculated using the following for- • Stress blebbing-Blebs were counted on live images at 4, 24, 48, 72, 96, and 168 h.

• Cell death inducing an increase in H2Ax expression level-
The myotubes were fixed and stained for DAPI and H2AX as described earlier. To measure H2AX signal per myonucleus, the total area of H2AX signal was divided by the total number of myonuclei. • Cell loss-The total number of nuclei were counted in each field using ImageJ 1.37v and summed for each well. • Different doses tested-4 and 8 μg of MuVs were added to the culture medium of healthy myotubes. H2AX signal per myonucleus was assessed as described earlier. All cell cultures treated with MuVs were compared and normalized with untreated cell cultures.

RNA extraction
Purified muscle stem cells were differentiated for 3 days into myotubes. RNA from muscle cells was extracted as described in Bigot et al. 18 The quality of RNA samples was assessed with Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA).

Gene expression profiling
• mRNA gene expression profiling-Aliquots of high-quality total RNA from each sample (ALS n = 6 and healthy n = 6, muscle stem cells) were used for mRNA expression profiling using GeneChip Human Exon 1.0 ST arrays (Affymetrix) as previously described. 24 • Analysis of gene expression data-refer to the supporting information.

Immunolabelling
• Immunocytochemistry: 200 000 cells or 100 000 cells were respectively plated on u-dish 35 mm high ibidiTreat or 4 wells plate ibidiTreat (ibidi®) in proliferative medium. The following day, muscle stem cells were washed with PBS and myogenic differentiation was induced by cultivating the cells in DMEM only. Muscle cells were fixed at 3 days of differentiation using 4% formaldehyde. The cells were permeabilized, blocked and stained as described. 9 • Immunohistology: 8 μm muscle transverse sections were cut from human biopsies on a cryostat microtome at À20°C, permeabilized, blocked, and stained as previously described. 18 Primary antibodies used are listed in the table below and the secondary antibodies used were goat anti-mouse IgG1 or anti-mouse IgG2a, or mouse IgG2b or anti-rabbit tagged with AlexaFluor 355 or AlexaFluor 488 or AlexaFluor 555 or AlexaFluor 594 or AlexaFluor 647 (1:400, Invitrogen ™ ). The slides were washed, counter-stained with 1 μg mL À1 DAPI for 1 min, rinsed two times and mounted with ibidi mounting medium (ibidi®). Ten to twenty non-overlapping pictures were acquired in a line along the diameter of the slide with an Olympus IX70, and an Olympus UPlan FI 10×/0.30 Ph1 and an Olympus BX60 objectives equipped with a Photomatics CoolSNAP ™ HQ camera. For the human muscle sections, pictures were taken with an Olympus LCPlan FI 40×/0.60 Ph2 objective. Images were acquired using Zeiss software and analysed using either Fiji or ImageJ 1.37v. For enrichment mapping, the GSEA tool (http://software. broadinstitute.org/gsea/index.jsp) was used to assess the distribution of gene sets across the differential expression profile of ALS compared with Healthy myotubes. There were 6349 gene sets tested, including all of the Gene Ontology Biological Process and Cellular Component collections, and all of the Canonical Pathways from MSigDB. In addition, custom gene sets were created listing genes encoding the known protein binding partners of FUS 25 and TDP43. 26 Cytoscape v3 and the enrichment map plugin were used to create a graph representing as nodes the gene sets identified by GSEA to be significantly enriched with FDR < 0.05, with edges shown for those pairs of gene sets having overlap coefficient >0.5. Cytoscape's selection features were used to isolate a sub-graph of the enrichment map showing only the FUS-binding and TDP43-binding gene sets and those gene ontology or canonical pathway gene sets with which they shared genes (overlap coefficient >0.5).

Knockdown of FUS with siRNA
There were 100 000 cells plated in μ-Slide 4 Well ibiTreat (Ibibi®) one day before differentiation was induced. On Day 2 of differentiation, cells were transfected using Lipofectamine RNAiMAX Reagent with 200 nM s5401 FUS siRNA (LifeTechnologies ™ ) and treated with a low dose of PKH26-labelled ALS MuVs. MuVs were integrated by the cells for 3 days before fresh differentiation medium was added to the 5 days differentiated cells. The cells were harvested at Day 8 of differentiation to check for RNA levels by RT-qPCR and to perform cell death analysis, immunostaining for RPL5, and distribution analysis of RNA/DNA using acridine orange staining.

Western Blotting
Extracted ALS and healthy vesicles were resuspended in lysis buffer (8 M urea; 2% SDS; 10 μL/mL protease inhibitor cocktail), and protein was extracted from muscle biopsies using RIPA buffer. Extracted proteins were loaded into NuPage polyacrylamide 4-12% BisTris gels for electrophoresis under reducing conditions (Calnexin, Flotillin, ALIX, FUS, SOD1, TDP43, RPL5, and skeletal alpha-actin) and non-reducing conditions (CD63 and CD81). Transfer on polyvinylidene difluoride (PVDF) membrane was performed using the iBlot® Dry Blotting System (Life Technologies ™ ) and upon transfer polyacrylamide gels were stained with Blue Coomassie Gel Code Blue Stain Reagent (LifeTechnologies ™ ) to visualize proteins. Immunoblotting was carried out using the iBind ™ Flex Western System and primary antibodies (refer to Differences were considered to be statistically different at P < 0.05.

Amyotrophic lateral sclerosis patient muscle cells accumulate vesicles
To investigate the role of vesicles secreted by ALS muscle cells, myoblasts were extracted from biopsies of confirmed ALS patients at an early stage of pathology (refer to Table 1 for patient description, and a breakdown of which samples were used in which experiment). Among the 27 sporadic ALS subjects, genetic screening identified only four with known ALS-causative mutations (three carrying the C9orf72 hexanucleotide repeat expansion, one with aberrant ATXN2 repeat length). A consistent accumulation of extracellular vesicle markers CD63 ( Figure 1A,B) and TSG101 ( Figure S1B) in ALS myotubes was observed by immunostaining. Extracellular vesicle markers were higher by RT-qPCR ( Figure 1C), and multivesicular bodies filled with exosome-like vesicles were observed by electron microscopy (Figures 1D and S1A). In vivo, ALS patient muscle biopsies presented an increased frequency of multi-vesicular bodies (MVBs) (0.017 MVBs/sarcomere in ALS muscles, vs. 0.004 MVBs/sarcomere in healthy controls; Figure S1C,D), an accumulation of extracellular vesicle markers at the periphery of the myofibres ( Figure 1E,F), and an increased expression level of the vesicle marker CD63 ( Figure 1G,H).

Secretion of amyotrophic lateral sclerosis patient muscle cell vesicles
Importantly, because primary muscle stem cells have a limited number of divisions (~30), all experiments were carried out before the cells reached 21 divisions, thereby avoiding excessive population doublings which could lead to senescence and experimental artefacts. 18,20,30 In both ALS and healthy vesicle extracts, exosome-like vesicles with a typical cup-shaped morphology ( Figure 2A) and sizes ranging from 90 to 200 nm (Figure 2A-C) were observed. The muscle vesicles were positive for exosomal markers such as CD63, CD82, CD81, Flotillin, and ALI, and were negative for calnexin, an endoplasmic reticulum marker normally absent in the exosome fraction ( Figure 2C,D). In addition, the vesicle extracts were free of albumin and other contaminants ( Figure S2B, and proteomic data GSE122261). The MuV fraction secreted by ALS myotubes contained 1.7-fold more protein than that secreted by an equal number of healthy control myotubes ( Figures 2E  and S2B).

Secreted amyotrophic lateral sclerosis vesicles are neurotoxic
To test neurotoxicity, 0.5 μg of ALS or healthy MuVs were added to the culture medium of healthy human iPSC-derived motor neurons (hiPSC-MN). Because ALS myo-tubes secrete more MuVs than healthy myotubes, this quantity corresponded to 2 ALS myonuclei or 3.7 healthy myonuclei per MN. Following uptake of MuVs ( Figure 3A), only ALS MuVs and not healthy MuVs resulted in shorter neurites ( Figure 3B), with less branching ( Figure 3C) and a greater cell death ( Figure 3D,E) 72 h post-treatment. Cell death as a result of ALS MuVs was consistent across multiple patients ( Figure 3E) and showed a dose response when different quantities of ALS MuV protein were loaded-ALS MuV toxicity was significant at all concentrations ( Figure  3F). Conversely, when MuV uptake was decreased by preincubating MuVs with CD63 antibody (Figure S3A), hiPSC-MN death was dramatically decreased ( Figure 3G). Similarly, when added to the culture medium of healthy human myotubes, ALS MuVs induced myotube atrophy ( Figure S3B), and cell stress ( Figure S3C) leading to cell death ( Figure S3D-F). The quantity of cell death was decreased when less MuV protein was loaded, though ALS MuV toxicity remained greater than healthy MuV toxicity at either dose ( Figure   S3F), suggesting that toxicity is dependent on both the quantity and content of MuVs.

Amyotrophic lateral sclerosis muscle vesicles are enriched in proteins involved in RNA processing
Proteomic analysis of vesicle content revealed that, of 53 peptides observed at consistently higher levels in the MuVs of ALS subjects compared with healthy controls, 21 were annotated to the RNA-binding molecular function ( Figure 4A; enrichment FDR P < 1 × 10 À7 ; Table S1). Similarly, of 453 proteins detected only in ALS and not in healthy controls, 87 were involved in RNA-binding (enrichment FDR P < 1 × 10 À14 ; Table S2). ALS MuVs contained many known protein binding partners of the RNA-processing proteins FUS 25 and TDP43 26 (enrichment P < 1 × 10 À6 ; Figure 4B), with 58% (64 of 109) of FUS binding partners being detected, along with FUS itself. Similar enrichment (P from 0.001 to <1 × 10 À6 ) was observed against other protein lists of FUS-binding and TDP43-binding partners. 31,32 FUS and RPL5 -a FUS binding partner-were at a higher level in ALS MuVs by Western blotting (Figure 4C-E). Interestingly, neither SOD1 nor TDP43 were detectable in human MuVs ( Figure S4).
Transcriptomic analysis of cultured myotubes suggested that genes encoding FUS-binding and TDP43-binding proteins were upregulated in the myotubes of ALS patients and that these genes were shared with many RNA-processing pathways that were similarly upregulated ( Figure 5A). ALS myotubes presented a greater nuclear accumulation of RNA ( Figure 5B), and mislocalization of two FUS protein binding partners, RPL5 and caprin 1, that are involved in RNA processing and stress granule formation ( Figure 5C,D). 25

Secreted amyotrophic lateral sclerosis vesicles affect RNA processing in recipient motor neurons
Because a leading theory is that disruption of RNA metabolism-including RNA translation, transport, storage, and degradation-contributes to ALS physiopathology by affecting neuronal function and viability, 33 and based on the results described earlier, we hypothesized an involvement of RNA processing in ALS MuV toxicity. When human iPSC-MNs derived from healthy subjects were treated with ALS MuVs, RNA accumulated in their nuclei ( Figure 6A), which has been reported to induce cell death. 34 Similar results were obtained when ALS MuVs were added to the cultures of human myotubes from healthy subjects ( Figure S5A). We hypothesized that ALS MuV toxicity in recipient cells may be mediated through the FUS pathway. To test this, 0.5 μg of  Figure S4, Tables S1 and S2.
ALS or healthy MuVs were added to the culture medium of a human muscle cell line over-expressing a tagged form of wild-type FUS (FUS-FLAG ; previously published 27 ). This induced an increase in cell death from 8% to 42% ( Figure 6B), accompanied with greater cellular stress ( Figure S5B)-the same was not observed in cell lines that over-expressed tagged forms of wild-type TDP43 and SOD1 ( Figures 6B and S5B). Conversely, when ALS MuVs were added to the culture medium of cells in which FUS was knocked down, lower proportions of nuclei with accumulated RNA ( Figure 6C) and RPL5 granules ( Figure 6D) were observed, and the quantity of cell death was reduced ( Figure  6E). These data implicate the FUS pathway in MuV toxicity and suggest that increased levels of FUS heighten sensitivity to this toxicity, while lower levels reduce it. We note that, relative to muscle cells, high levels of FUS are observed in hiPSC-MN ( Figure 6F).

Discussion
Extracellular vesicles are suspected to carry toxic elements from astrocytes towards motor neurons, 7 and pathological proteins have recently been identified in circulating extracellular vesicles of sALS patients. 35 Here, ALS muscle vesicles are shown to be toxic to motor neurons, which establishes the skeletal muscle as a potential source of vesicle-mediated toxicity in ALS. The accumulation and over-secretion of muscle vesicles was observed as a consistent feature of sporadic ALS patients in this cohort, including patients carrying mutations in C9orf72 or ATXN2, suggesting that this may be a common feature across many or all sporadic and familial forms of ALS. The consistent observation of a pronounced RNA processing and protein mislocalization phenotype in the ALS myotubes suggests that these cells recapitulate aspects of the disease mechanism that have been observed in motor neurons.
The observation that ALS MuVs act on RNA transport, that overexpression of wild-type FUS in MuV-recipient cells resulted in increased recipient cell death, and that RNA transport protein mislocalization is partially corrected and cell death reduced when FUS expression was knocked down in recipient cells, is consistent with a body of literature suggesting an RNA processing blockade mechanism in ALS MN. 36,37 As shown, FUS expression is relatively high in iPSC MN compared with muscle cells (Figure 6F), and we also note that the cerebral cortex is among the tissues with the highest reported levels of FUS mRNA (http://www.proteinatlas.org 38 ). While here we examined the relationship of FUS to ALS MuV toxicity in recipient muscle cells, further investigation, focused on MN, will be necessary to understand the mechanistic interplay of FUS and RNA-processing with muscle vesicle contents and the potential role in ALS pathology.
The observation that FUS and many of its binding partners are present in MuVs is of interest in the context of recent observations that normal function of FUS is required for normal neuromuscular junction development in mice, and that co-culture of iPSC-derived motor neurons with myotubes from FUS mutated patients resulted in impaired endplate maturation, which was proposed to be due to intrinsic FUS toxicity in both muscle and MN. 39 Similar experiments, not only on the cells of patients with FUS mutations but also on sporadic ALS more generally, may help to explore the relationship of vesicle-mediated toxicity to the neuromuscular junction.
Several papers have described a muscle phenotype in ALS that occurs independently and prior to muscle denervation, such as metabolic imbalance, 40,41 oxidative stress, 42 and mitochondrial dysfunction. 43 However, the role of muscle in ALS is unresolved. While an ALS-like phenotype was observed in mice when exogenous human mutant SOD1 expression was restricted to the skeletal muscle, 42,44 other studies knocking down mutant SOD1 expression in skeletal murine muscle did not show any significant decreases in the progression of symptoms. 45,46 The differences observed between these studies reveal the difficulties to assess the role of muscle in ALS, as the success of targeting whole skeletal muscle requires intravenous injection of a large amount of adeno-associated virus particles. 47 The secretion of muscle cell vesicles that are toxic towards motor neurons in vitro could suggest a potential role of muscle in ALS pathology. However, further in vivo experiments in appropriate animal models are required, to test the capacity of MuVs to diffuse in vivo, and their capacity to cross the blood brain barrier or to be absorbed directly at the neuromuscular junction.