miR126-5p Downregulation Facilitates Axon Degeneration and NMJ Disruption via a Non–Cell-Autonomous Mechanism in ALS

Axon degeneration and disruption of neuromuscular junctions (NMJs) are key events in amyotrophic lateral sclerosis (ALS) pathology. Although the disease's etiology is not fully understood, it is thought to involve a non–cell-autonomous mechanism and alterations in RNA metabolism. Here, we identified reduced levels of miR126-5p in presymptomatic ALS male mice models, and an increase in its targets: axon destabilizing Type 3 Semaphorins and their coreceptor Neuropilins. Using compartmentalized in vitro cocultures, we demonstrated that myocytes expressing diverse ALS-causing mutations promote axon degeneration and NMJ dysfunction, which were inhibited by applying Neuropilin1 blocking antibody. Finally, overexpressing miR126-5p is sufficient to transiently rescue axon degeneration and NMJ disruption both in vitro and in vivo. Thus, we demonstrate a novel mechanism underlying ALS pathology, in which alterations in miR126-5p facilitate a non–cell-autonomous mechanism of motor neuron degeneration in ALS. SIGNIFICANCE STATEMENT Despite some progress, currently no effective treatment is available for amyotrophic lateral sclerosis (ALS). We suggest a novel regulatory role for miR126-5p in ALS and demonstrate, for the first time, a mechanism by which alterations in miR126-5p contribute to axon degeneration and NMJ disruption observed in ALS. We show that miR126-5p is altered in ALS models and that it can modulate Sema3 and NRP protein expression. Furthermore, NRP1 elevations in motor neurons and muscle secretion of Sema3A contribute to axon degeneration and NMJ disruption in ALS. Finally, overexpressing miR126-5p is sufficient to transiently rescue NMJ disruption and axon degeneration both in vitro and in vivo.


Introduction
Amyotrophic lateral sclerosis (ALS) is a lethal neurodegenerative disease that affects motor neurons (MNs) in the cortex, brains-tem, and spinal cord (SC) (Mulder et al., 1986;Peters et al., 2015). It is characterized by neuromuscular junction (NMJ) disruption, MN axon degeneration, and neuronal death (Frey et al., 2000;Fischer et al., 2004;Moloney et al., 2014). Despite some progress, currently no effective treatment is available for ALS. The diversity of ALS-related mutations has given rise to the use of numerous animal models with diverse phenotypes, ranging from no effect on MN function to severe progressive paralysis (Philips and Rothstein, 2015). Approximately 20% of familial ALS is accounted for by mutations in the superoxide dismutase 1 (SOD1) gene (Scrutton et al., 1992;Rosen et al., 1993;Gurney et al., 1994). Mutations in the SOD1 gene (mSOD1) have also been described in sporadic cases (sALS) (Rakhit et al., 2004). Other mutations found in ALS patients include hexanucleotide expansion repeats in the C9orf72 locus, which lead to various dipeptide repeats (e.g., proline-arginine or glycine-arginine repeats: PR 50 and GR 50 , respectively), and in the gene encoding the TDP43 RNA binding protein (Buratti, 2015;Wen et al., 2017). An efficient method for studying NMJ stability and health in vitro is by using the microfluidic chamber (MFC) system, which allows the culture of primary myocytes in one compartment, and MNs in the other, thus setting up conditions conducive to generating active NMJs (Zahavi et al., 2015).
The neurodegeneration that occurs in ALS is considered to be a non-cell-autonomous process involving interactions between the neuron and its diverse extracellular microenvironments via an unknown mechanism (Ilieva et al., 2009;Tsitkanou et al., 2016). Although the molecular basis for neuronal dysfunction and death in ALS is still poorly understood, it may be due to alterations in the nature of the extracellular signaling pathways that switch from prosurvival to toxic (Ilieva et al., 2009;Perlson et al., 2009). Numerous studies support the notion that multiple tissues outside the CNS, including skeletal muscle (Dupuis et al., 2006;Tsitkanou et al., 2016), astrocytes (Nagai et al., 2007), and microglia (Lee et al., 2016), contribute to ALS pathologies. Alterations in RNA metabolism and microRNAs (miRs) can contribute to, and also be part of, mechanisms that initiate the disease (Lemmens et al., 2010;Emde and Hornstein, 2014). miRs are post-transcriptional regulators that play an important role in many cellular processes, such as axon growth and retraction, and were demonstrated to be involved in many diseases, including neurodegenerative diseases, such as ALS (Hawley et al., 2017;Molasy et al., 2017). Alterations in miR expression profile were identified specifically in axons of ALS models (Rotem et al., 2017), as well in muscles leading to increasing attempts to either use or target miRs as therapeutic strategies (Di Pietro et al., 2017). Therefore, it is reasonable to assume that alterations in RNA and miRNA metabolism, of both MNs and neighboring cells, can regulate a secreted destabilizing signal, which in turn, facilitates axon degeneration and NMJ disruption.
Semaphorin3A (Sema3A) was initially identified as a repellent guidance molecule (Luo et al., 1993;Worzfeld and Offermanns, 2014). However, later works showed that it can also induce neuronal cell death of sympathetic, sensory, retinal, and cortical neurons (Nakamura et al., 2000;Shirvan et al., 2002;Ben-Zvi et al., 2008;Jiang et al., 2010). Neuropilin1 (NRP1) has been shown to be the receptor binding component for Sema3A as well as some other Type 3 Semaphorins (Kolodkin et al., 1997). Sema3A was found to be upregulated following CNS injury as well as in several neurodegenerative diseases (Kaneko et al., 2006;Van Battum et al., 2015). Importantly, Sema3A was found to be upregulated in terminal Schwann cells of the SOD1 G93A transgenic mouse model for ALS and in the motor cortex of ALS patients Körner et al., 2016), suggesting that it plays a toxic role in disease pathology and progression.
Here we demonstrated that alterations in miR126-5p result in upregulation of Type 3 Semaphorins and its cobinding receptor NRP1 in muscles and MN axons of ALS models, respectively. We further demonstrate in vitro and in vivo the contribution of this pathway to axon degeneration and NMJ disruption in ALS models.

Materials and Methods
Animals and vector injections. HB9::GFP (stock #005029) mice were originally obtained from The Jackson Laboratory. The colony was maintained by breeding with ICR mice. SOD1 G93A (stock #002726) mice were originally obtained from The Jackson Laboratory and maintained by breeding with C57BL/6J mice. Genotyping was performed using the PCR (KAPA Biosystems); DNA samples were generated from ear or tail tissue biopsies. All injection procedures were performed on presymptomatic ϳP60 mice. Mice were first anesthetized using a mixture of xylazine and ketamine. Next, 100 l of Neurobasal-containing X10 concentrated lentiviruses (6 ϫ 10 9 titer units) was injected into the gastrocnemius (GC) muscles using a 1 ml syringe and a 25G needle. A pLL-miR126-5p-GFP construct was injected into the right hind foot, whereas a pLL-miR142-GFP construct was injected into the left hind foot. All animal experimentations were approved by the Tel Aviv University Animal Ethics Committee.
MFC preparation. Polydimethylsilxane MFCs were designed and cast as described previously (Ionescu et al., 2016). After the wells were punched, a small "cave" was created in the explant well near the grooves using a 25G needle, keeping the explant in place. Microfluidic devices were cleaned of surface particles using adhesive tape and were sterilized in 70% ethanol for 15 min. Devices were completely dried under sterile conditions using UV radiation and then attached to a sterile 60 mm plastic dish (Nunc) with gentle pressure, and the margins were sealed with polydimethylsilxane before incubation at 70°C for 30 min to prevent the detachment of the chamber. Muscle channels were coated with Matrigel diluted 1:10 with DMEM containing 2.5% penicillinstreptomycin-nystatin (PSN) for 30 min at 37°C, before filling the muscle wells with 150 l of Bioamf-2 medium. The explant well and channel were filled with 150 l of 1.5 ng/ml poly-DL-ornithine (P-8638, Sigma-Aldrich) in PBS overnight and then replaced with 150 l laminin (L-2020, Sigma-Aldrich), 1:333 in deionized distilled water overnight. One day before plating the SC explant, laminin was replaced with explant medium containing Neurobasal (Invitrogen) supplemented with 2% B27 (Invitrogen), 1% penicillin-streptomycin (Biological Industries), 1% Glutamax (Invitrogen), and 25 ng/ml BDNF (Alomone Labs), until the day on which coculturing began.
Fluorescence microscopy and image analysis. All confocal images were captured using a Nikon Ti microscope equipped with a Yokogawa CSU X-1 spinning disc and an Andor iXon897 EMCCD camera controlled by Andor IQ2 software. Epifluorescence was imaged using the same microscope in bright-field mode, and images were captured with an Andor Neo sCMOS camera, or at a FLoid benchtop imaging station (Invitrogen). TIRF images were captured using a TILL photonics iMIC microscope (FEI Munich) with an Andor iXon897 EMCCD camera. All live-imaging assays were performed in a humidified 5% CO 2 incubation chamber.
Isolation and culture of human mesenchymal stem cells (hMSCs). hMSCs from healthy donors and ALS patients used in this study were obtained from bone marrow samples and were isolated, and then phenotypically characterized and cultured as described previously (Nachmany et al., 2012). All volunteers in this work signed a consent form before sample donation, according to the guidelines of the Hospital's Ethics Committee supervised by the Israeli Health Ministry Ethics Committee conforming with the Code of Ethics of the World Medical Association (Declaration of Helsinki), printed in the British Medical Journal (Rickham, 1964).
SC explants. SCs were dissected from E11.5 mouse embryos of both sexes, either using HB9::GFP or SOD1 G93A stripped of meninges and DRGs. The ventral horn was separated from the dorsal horn by longitudinal cuts along the SC, and transverse sections up to 1 mm were placed in the explant well. Before plating, the growth chambers were coated with 1.5 g/ml poly-DL-ornithine overnight at 37°C and 3 g/ml laminin overnight at 37°C. Explants were maintained in Spinal Cord Explant Medium containing Neurobasal, 2% B27, 1% PS, and 1% GlutaMAX, supplemented with 25 ng/ml BDNF. Explants were grown at 37°C in 5% CO 2 .
Primary myocyte culture. Skeletal muscle cultures were derived from the GC muscle of adult P60 female mice of either SOD1 G93A background or their littermates (LMs) using techniques previously described (Ionescu et al., 2016). Briefly, GC muscles were excised and incubated in 2 mg/ml collagenase I (Sigma-Aldrich) in DMEM containing 2.5% PSN (Biological Industries) for 3 h. Muscles were then dissociated and incubated for 3 d in Matrigel-coated (BD Biosciences) 6-well plates with Bioamf-2 medium (Biological Industries) with 1% PSN at a density of ϳ120 myofibers per well. For purification of the myoblasts, adherent cells were trypsinized and preplated in an uncoated dish for 1 h at 37°C. Nonadherent cells were then transferred into a Matrigel-coated dish with Bioamf-2 medium. Preplating was repeated for 2 d, keeping the culture at Ͻ50% confluence, before plating cells in MFC. Cultures were maintained at 37°C and in 5% CO 2 . After the final preplating, 100,000 myocytes were cultured in the precoated distal compartment of the MFC. Myocyte conditioned media (CM) were produced as follows: At the final preplating stage, myoblasts were cultured in a Matrigel-coated 100 mm dish at 80% confluence and were incubated for 2 d with Bioamf-2 medium, followed by 2 d with rich DMEM (Biological Industries) medium containing 10% FCS (Biological Industries), 10% horse serum (Biological Industries), 1% GlutaMAX, and 1% PSN. Then, once muscles reached a fully differentiated state, the culture dish was rinsed 3 times with preheated PBS and poor DMEM containing 1% GlutaMAX, and 1% PSN was applied on the cultures. CM was collected after 2 d, centrifuged for 5 min at 400 ϫ g at 25°C, and streamed through a 0.22 m PES filter.
CM preparation and application. Muscle myocytes of WT or SOD1 G93A mice were cultured as described previously (Ionescu et al., 2016). Seven days after myocytes were fully differentiated, the muscles kept growing for 3 d in complete Neurobasal containing BDNF and GDNF. The CM was refreshed with BDNF, GDNF, and glucose after its collection, as described previously (Nagai et al., 2007). CM under both conditions was applied on the axon compartment of the MFC for 48 h.
Lentiviral vectors. Genes of interest were cloned into a third-generation lentiviral pLL3.7 backbone. HEK293T cells were transfected by using calcium phosphate method and a mixture consisting of the vector of interest, vesicular stomatitis virus glycoprotein, and group antigenspolymerase (reverse transcriptase) was used. The medium was replaced after 5-8 h, and the supernatant was collected 48 h later. Next, 50 mM HEPES was added before freezing to maintain a neutral pH for long-term storage. When necessary, lentiviruses were concentrated using a PEG Virus Precipitation Kit (Abcam).
Quantification of myocyte contraction. The 1000-frame-long movies of myocytes in the distal compartment of the MFC were acquired 7 d after coculturing. Imaging was performed under bright-field conditions at a rate of ϳ33 fps while using a 20ϫ objective. A myocyte contraction plot was then profiled using an image-based method previously described (Zahavi et al., 2015;Ionescu et al., 2016). Briefly, only myocytes that came in contact with axons were plotted. Time-lapse images were taken for analysis using ImageJ. To create a time trace of contractions, high contrast (bright or dark) ROIs were selected on each myotube. Movement of the selected spot due to myotube contraction was assessed by the change in the ROI intensity over time. The number of strong contractions, as measured from the time trace, was manually validated by reexamining the time-lapse image series. The number of strong and weak contractions in innervated myotubes was compared before and after 1 M TTX was added to the neuronal compartment. A myotube with a post-to pre-TTX difference of Ͼ50% was measured as an increase or decrease in contraction, and the fraction of increased, decreased, and unchanged myotubes was calculated.
RNA extraction and cDNA synthesis. Muscle tissues were immediately frozen with liquid nitrogen. Tissue was ground to powder using a pestle and mortar. Then 700 l of TriReagent (Sigma-Aldrich) was added to the muscle powder, and the samples were further passed through a 21G needle 3 times for better homogenization. RNA from the TriReagentrinsed samples was further isolated following the TriReagent protocol. RNA quality was measured using NanoDrop3000 and a bio-analyzer. RNA purification of MN mass culture, along with transfected HeLa cells, was performed using TriReagent protocol as well. mRNAs were pooled in equal amounts and reverse-transcribed into double-stranded cDNA by using the SuperScript2 kit (QIAGEN).
NanoString Chip. A total of 100 ng RNA samples were outsourced to NanoString Technologies' facilities for an miR-Chip array assay of ϳ800 known miRs (NanoString Technologies). miR was quantified automatically by NanoString Technologies' instrumentation for miRs, which was hybridized with the template. Output data were analyzed by the nCounter analysis system. All miRs were normalized to the 100 most abundant miRs in the samples.
Primer design. Based on the consensus sequences of the desired transcripts, 2 sets of primers were designed for each gene (Table 1).
qPCR for mRNA detection. qPCR was performed on the StepOne system (Invitrogen) in a 10 l reaction containing 4 l of RNA (20 ng), 5 l Syber green master mix (Thermo Fisher Scientific), and 1 l of reverse and forward primers.
miR vectors and transfection. Mammalian expression vector pMSCV-Blast-miR constructs were generously provided by Eran Hornstein (Weizmann Institute of Science). Mammalian expression vector of C9orft72 Di-peptide PR 50 and GR 50 constructs was generously provided by David Trotti (Jefferson University) (Wen et al., 2014). Next, 50,000 HeLa/U87 human glioblastoma/muscles cells were plated in rich DMEM (1% PS, 1% GlutaMAX, 20% FBS). After 24 h, the culture medium was replaced with serum-free medium (Opti-MEM), and cells were transfected using FuGene NE 6 (Promega) protocol. Cells were collected after 48 h and used either for a functional assay or for RNA/protein extracts. Myocyte cultures were transfected using the same approach.
Semaphorin preparation. HEK293T cells were stably transfected to overexpress either Sema3A or an empty control. CM from 80% confluent cultures were collected after 3 d. We validated the purity level of the collected media using Coomassie staining and identified the stained band with a specific antibody against the desired protein using Western blot analysis.
NRP1 antibody application. A total of 5 g/ml NRP1 antibody (R&D System, AF566 dot ETH0915031) for the extracellular domain was added to the distal compartment of the MFC while maintaining a proximal-todistal volume gradient.
Histology tissue collection and fixation. GC muscles of 20 samples were harvested and fixed in 4% PFA. The samples were then outsourced for a histological assessment at Patho-Logica. All tissues were trimmed into block cassettes and sent to CDX Diagnostics for slide preparation.
Slide preparation and histological evaluation. Tissues were trimmed, embedded in paraffin sections at no more than 5 m thickness, and stained with H&E. The mean minimal muscle fiber diameter thickness was measured in microns by performing a manual count using a 10ϫ lens and analyzed by expert pathologist.
xCELLigence impedance measurement. For each experiment, 30,000 U87 cells were plated with rich DMEM in E-Plate L8 wells and incubated together with the xCELLigence system (ACEA Biosciences) at 37°C, 5% CO 2 overnight. Impedance data were collected at 5 min intervals. After 24 h, poor DMEM (1% PS, 1% GlutaMAX) with Sema3A or its control medium was replaced and recording proceeded. The data were analyzed using RTCA data analysis software 1.0 and normalized to the control sample.
CatWalk XT gait analysis. The CatWalk is a video-based analysis system used to assess gait in voluntarily walking mice (Noldus Information Technology). The principle of this method is based on an optical technique. The light of a fluorescence tube is completely internally reflected on a glass walkway floor. When the animal crosses the walkway, the light leaves the glass and illuminates only the area of contact. In this way, the different paw contacts are visualized. Based on position, pressure, and the surface area of each foot paw, multiple parameters are calculated.
Only compliant and continuing trials for each animal were analyzed, averaged, and the mean was calculated.
Experimental design and statistical analysis. Data are mean Ϯ SEM. The statistical analysis was assessed by Student's t test. In all cases, differences were considered to be statistically significant if p Ͻ 0.05.

Results
Sema3A and NRP1 levels are elevated in muscles and the MNs of ALS models ALS disease is considered to be a distal axonopathy involving axon degeneration and NMJ disruption as key processes in its pathology (Fischer et al., 2004). We therefore hypothesized that destabilizing factors secreted from adult presymptomatic ALS mutant muscles might be involved in triggering axon degeneration of MNs. Because Sema3A is known to act in such a manner, at least in development, and it was already reported to be elevated in ALS, we decided to focus on this factor Körner et al., 2016). Following this hypothesis, we first examined the expression of Sema3A in SOD1 G93A GC muscles in comparison with that of their LMs; ( Western blot analysis of muscle protein extracts revealed significant elevations in Sema3A protein levels in muscles of presymptomatic SOD1 G93A mice as early as P30 and P60, whereas testing Sema3A levels in younger animals (P7) showed no apparent differences compared with their LM controls (mean fold change over LM: To validate this difference, we also tested the transcript levels of Sema3A (Fig. 1C). qPCR analysis of total RNA extracts from muscles at presymptomatic SOD1 G93A stage and LM mice identified an ϳ1.7-fold increase in Sema3A mRNA of SOD1 G93A muscles (mean fold change over LM: SOD1 G93A , 1.72 Ϯ 0.32; LM, 1 Ϯ 0.06). Because GC muscle tissues contain heterogeneous cell types and to verify that the levels of Sema3A are indeed higher specifically in SOD1 G93A muscles fibers, we immunostained primary myocyte cultures from P60 SOD1 G93A and LM mice for Sema3A ( Fig. 1 D, E). Quantifying the mean intensity values showed a significant 50% increase in the SOD1 G93A myocytes (a mean fold change in intensity over LM: SOD1 G93A , 1.5 Ϯ 0.06; LM, 1 Ϯ 0.04). We also collected CM from myocyte cultures to determine whether the increase in Sema3A protein also results in an increase in its secretion (Fig. 1F ). Western blot analysis indicated that Sema3A levels were also elevated in P60 SOD1 G93A myocyte-CM (a mean fold change over LM: SOD1 G93A , 2.3 Ϯ 0.55; LM, 1 Ϯ 0.08). Because NMJ disruption is a primary event in ALS, we sought to examine the levels of Sema3A in NMJ in vivo (Fig. 1G,H ). Immunostaining for Sema3A in GC muscles showed a sixfold increase in the number of muscle fibers expressing Sema3A in their NMJs. Whereas only ϳ5% of NMJs stained positively for Sema3A in LM muscles, we identified its expression in ϳ30% of NMJs in P60 SOD1 G93A mice (the mean percentage of NMJs expressing Sema3A: SOD1 G93A , 30.83 Ϯ 4.73%; LM, 4.56 Ϯ 2.4%). Interestingly, a previous study described Sema3A elevation in SOD G93A mice specifically in fast fatigue NMJs expressing myosin-IIb marker . Because fast fatigue NMJs are the first to become disrupted and be eliminated in ALS pathology, we examined Sema3A levels both at P90 and P120 and hypothesized that Sema3A levels will eventually drop in later stages of the disease. We found that whereas the percentage of NMJs expressing Sema3A in SOD G93A in P90 animals is similar to P60, the end-stage animals (P120) were shown to display a reduction in Sema3A-positive NMJs, and   no apparent difference existed between WT and SOD1 G93A mice ( Fig. 1-2 A, B, available at https://doi.org/10.1523/JNEUROSCI. 3037-17.2018.f1-2). Together, these results indicate that a significant part of the MN population is exposed to high levels of Sema3A in presymptomatic stages and that this specific population is disrupted and eliminated during disease progression. We then proceeded to investigate the expression of the Sema3Areceptor binding unit, NRP1, in ALS (Fig. 1I ). Western blot analysis of GC muscle extracts revealed a significant ϳ8-fold increase in NRP1 (the mean fold change over LM: SOD1 G93A , 8.6 Ϯ 2.2; LM, 1 Ϯ 0.3). Because MNs are a primary target in ALS, we wanted to determine whether NRP1 is also overexpressed in the MNs of SOD1 G93A mice. First, we performed Western blot analysis of SNs and observed an ϳ2-fold elevation in NRP1 levels of P60 SOD1 G93A mice ( Fig. 1J; the mean fold change over LM: Western blot analysis on cell culture lysates. Importantly, we did not observe any difference in NRP1 expression after applying Sema3A, suggesting that NRP1 levels are regulated by an intrinsic mechanism in MNs ( Fig. 1-3, available at https://doi.org/ 10.1523/JNEUROSCI.3037-17.2018.f1-3). Finally, to validate our finding with other ALS models and to emphasize the impact of Sema3A in ALS, we performed Western blot analysis for Sema3A and NRP1 expression in human mesenchymal stem cells from sporadic ALS patients and healthy controls, as well as in myocyteexpressing C9orf72-PR 50 and their CM for Sema3A. In addition, we compared the results with those of a mock control. Remarkably, in all of these ALS models, we identified high expression of Sema3A and NRP1 ( Fig. 1-4, available at https://doi.org/10.1523/JNEUROSCI. 3037-17.2018.f1-4). Together, our combined in vivo and in vitro data suggest that the levels of both Sema3A and its cobinding receptor, NRP1, are presymptomatically increased in several ALS models as well as in sALS patients. These findings suggest that the Sema3A pathway is a common denominator in various ALS mutations; thus, it may contribute to MN degeneration in ALS.

Application of Sema3A on wild-type MN axons results in axon degeneration
Because our findings suggest that Sema3A is produced and secreted in excess from muscles of ALS models, and because muscles interact specifically with MN axons, we sought to test the activity of Sema3A exclusively in this distal subcellular compartment. To this end, we used an MFC that allows precise control, monitoring, and manipulation of subcellular microenvironments ( Fig. 2-1, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f2-1) (Zahavi et al., 2015). We cultured healthy ventral SC explants from transgenic mouse embryos expressing GFP under the MN-specific promoter HB9 (HB9::GFP) in one compartment of the MFC and enabled axons to extend into the opposing compartment, thus segregating axons and cell bodies into two isolated compartments. To verify that our MFCs can efficiently segregate MN axons from their somata, we stained the neuronal culture in the MFC system for the dendritic and axonal markers MAP2 and Tau, respectively ( Fig. 2-1, available at https://doi.org/ 10.1523/JNEUROSCI.3037-17.2018.f2-1). We confirmed that all neurites that traversed the distal compartment are positive for Tau staining and negative for MAP2. Next, we purified Sema3A or control media as described previously (Ben-Zvi et al., 2006), and applied them to the distal compartment, while imaging the axons for 16 h ( Fig. 2A). Our recordings reveal extensive axon degeneration in the Sema3A-treated MFCs 6 -8 h after its application ( Fig. 2B; Movies 1, 2; the mean percentage of degenerated axons: Sema3A, 83.01 Ϯ 3.54%; control, 23.94 Ϯ 7.6%). Coapplication of NRP1-blocking antibody and Sema3A on MN axons inhibits the Sema3A-dependent axon degeneration ( Fig. 2B; the mean percentage of degenerated axons: Sema3A and NRP1 antibody, 25.00 Ϯ 12%). These data indicate that Sema3A can trigger axon degeneration in MNs when applied exclusively on distal axons, and further support our hypothesis that an increase in muscle-derived Sema3A might contribute to axon degeneration in ALS. 4 (Figure legend continued.) 5m. E, Image analysis reveals an increase in Sema3A intensity in SOD1 G93A primary myocytes (Student's t test, n ϭ 3, ϳ80 cells, ***p ϭ 0.00001). F, Western blot analysis of primary myocyte-CM revealed a higher level of Sema3A in the CM of SOD1 G93A . Cultures were lysed after CM collection, and equal loading volumes of lysates were immunoblotted for ERK to validate CM, which was produced from a similar mass of myocytes (Student's t test, n ϭ 3, *p ϭ 0.018). G, H, Immunostaining of fixed whole P60 GC muscles shows distinct Sema3A expression in the NMJs of SOD1 G93A mice. White represents Sema3A. Red represents TMR-BTX labeling of acetylcholine receptors on postsynapse. Blue represents presynaptic NFH in neurons. The percentage of muscle fibers expressing Sema3A in their NMJs in P60 SOD1 G93A mice is higher (ϳ100 NMJ per 1 biological repeat; Student's t test, SOD1 G93A , n ϭ 4; WT, n ϭ 3; *p ϭ 0.011). Scale bars, 10 m. We also examined Sema3A expression in later stages of the disease (Figure 1-2, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f1-2). I, Western blot analysis of GC muscle extracts from P60 mice revealed elevated NRP1 levels in the muscles of SOD1 G93A . Tubulin was used as a loading control (Student's t test, n ϭ 3, *p ϭ 0.048). J, Western blot analysis of SN extract from P60 mice shows an elevation in the levels of NRP1 in the SNs of SOD1 G93A mice (n ϭ 3). K, Western blot analysis of primary MN lysates after 3 d in culture reveals an elevation in the NRP1 levels in SOD1 G93A MNs, which are not regulated by Sema3A binding (Figure 1-3, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f1-3). ERK was used as a loading control (Student's t test, n ϭ 3, *p ϭ 0.031). L-N, Immunostaining of primary MNs after 3 d in culture shows an elevation in the levels of NRP1 in both axons (inset, ϳ4.1-fold) and somata (ϳ1.9-fold) of SOD1 G93A MNs. White represents NRP1. Blue represents NFH. Somata (Student's t test, n ϭ 3, ϳ40 cells, ***p ϭ 0.00021); axons (Student's t test, n ϭ 3, ϳ40 cells, *p ϭ 0.012). Scale bars, 10 m. O, P, Immunostaining of fixed whole P60 GC muscles shows distinct NRP1 expression in the NMJs of SOD1 G93A mice. White represents NRP1. Red represents BTX. Blue represents NFH. The percentage of muscle fibers expressing NRP1 in their NMJs in P60 SOD1 G93A mice is higher (Student's t test, SOD1 G93A , n ϭ 4; WT, n ϭ 3; *p ϭ 0.042). Scale bars, 5 m. We further examined NRP1 expression in later stages of the disease (Figure 1-2, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f1-2). Elevations in Sema3A and its coreceptor were found also in human sALS samples (Figure 1-4  .f2-1) in the distal compartment of an MFC with no muscles after applying Sema3A to the distal compartment. After 6 h, axons in the distal compartment of chambers that were treated with Sema3A undergo degeneration, whereas axons in the control chamber or axons cotreated with NRP1 antibody and Sema3A continue growing. Scale bar, 20 m. B, Quantification of the rate of degraded axons in the distal compartment revealed a higher percentage of degradation in chambers that were exposed to Sema3A compared with either control or coapplication of Sema3A and NRP1 antibody (ϳ60 axons for Sema3A treatment, ϳ70 axons for Control; Student's t test; n ϭ 4; mean Ϯ SEM, ***p ϭ 0.00022). C, Schematic view of the experimental procedure in D-F. HB9::GFP SC explants and primary myocytes of SOD1 G93A , TDP43 A315T , C9orf72-PR 50 , C9orf72-GR 50 , or LM, GFP, and SOD1 wt as controls were cocultured in an MFC (Figure 2-2, available at https://doi.org/10.1523/ JNEUROSCI.3037-17.2018.f2-2), and the growth of HB9::GFP axons was assessed by time-lapse imaging of the microgroove compartment. D, Representative time-lapse images of the HB9::GFP axon growth when cocultured with (left to right) LM, SOD1 G93A , and SOD1 G93A ϩ NRP1 antibody. The presence of SOD1 G93A myocytes in the distal compartment triggers the retraction and degeneration of HB9::GFP motor axons growing in the groove compartment and prevents their traversing. When NRP1 antibody is applied to the distal compartment, together with SOD1 G93A -expressing myocytes, axons are less prone to degenerate. Scale bar, 5 m. E, Quantification of the rate of axons traversing the distal compartment in B shows the mean percentage of axons that traversed the distal compartment out of the total axons in each field (n ϭ 3; NRP1 antibody experiment, n ϭ 4; Student's t test, *p ϭ 0.025, *p ϭ 0.0433). F, Quantification of (Figure legend continues.)

Muscles expressing diverse ALS mutations initiate axon degeneration
To study the molecular mechanisms enabling the communication between MNs and their environment, which are essential for cell survival and synapse maintenance, we extended the use of the MFC system to coculture primary MNs and primary myocytes (Ionescu et al., 2016). Briefly, ventral SC explants from healthy 12-day-old (E12) HB9::GFP embryos were cultured in the proximal compartment, in the presence or absence of primary myocytes extracted from adult mice in the distal compartment ( Fig. 2-1, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f2-1). As we showed previously (Zahavi et al., 2015), culturing HB9::GFP explants in the presence of wild-type muscles facilitates the rapid and directed growth of axons into the distal compartment ( Fig. 2-1, available at https://doi.org/ 10.1523/JNEUROSCI.3037-17.2018.f2-1), suggesting that muscles secrete factors that support and promote the growth of motor axons. However, because ALS-mutated muscles were found to have intrinsic abnormalities throughout disease progression (Loeffler et al., 2016), we hypothesized that the nature of these factors will be altered. To study the effect of ALS muscles on MN axons in a simplified system, we plated primary myocytes from presymptomatic P60 SOD1 G93A and LM mice as well as WT myocytes transfected to express SOD1 wt in the distal compartment. Myocyte cultures were allowed to fuse and differentiate. Importantly, in all the described cases, myocyte morphology, fusion, and differentiation parameters were similar, and the culture showed no apparent difference ( Fig. 2-2, available at https:// doi.org/10.1523/JNEUROSCI.3037-17.2018.f2-2). After 7 d, HB9::GFP SC explants were cultured in the proximal compartment. Cocultures were incubated until the HB9::GFP axons began extending toward the microgroove compartment. Once the axons reached the microgroove compartment, the extension of HB9::GFP axons along the microgrooves was recorded for 16 h 4 (Figure legend continued.) the rate of axons traversing the distal compartment shows the mean percentage of axons that traversed the distal compartment out of the total number of axons in each field in coculture with TDP43 A315T , C9orf72-PR 50 , C9orf72-GR 50 myocytes, or GFP as a control. The traversing rate of HB9::GFP motor axons into the distal compartment in each of the cocultures with muscle-expressing ALS mutations is significantly reduced (n ϭ 3; Student's t test, GR 50 , *p ϭ 0.0137; PR 50 , *p ϭ 0.0374; TDP, *p ϭ 0.0304). G, Representative images of fixed and immunostained SOD1 G93A motor axons in the distal compartment of an MFC after applying LM or SOD1 G93A myocyte CM to the distal compartment. After 48 h, axons in the distal compartment of chambers that were treated with SOD1 G93A CM underwent degeneration, whereas axons that were treated with LM CM remained intact. WT MN axons remained intact after application of either CM (Figure 2-3 ( Fig. 2C). Surprisingly, HB9::GFP axons that were cocultured with the SOD1 G93A myocytes were less likely to traverse the distal side (Fig. 2 D, E). During this period, axons extending toward the SOD1 G93A myocytes were markedly incapable of traversing the distal compartment and underwent retraction, degeneration, or remained static in place, compared with the LM and SOD1 wt controls ( Fig. 2D; Movies 3, 4). Noteworthy, the addition of NRP1-blocking antibodies to the distal compartment, targeting Sema3A binding to the extracellular site of NRP1, improved the traversing rate of axons ( Fig. 2E; the mean axon traversal rate per field: LM, 33.88 Ϯ 10.40%; SOD1 wt , 52.66 Ϯ 12.7% SOD1 G93A , 11.1 Ϯ 5.5%; SOD1 G93AϩNRP1-ab , 28.18 Ϯ 5.4%). We further transfected primary myocyte cultures with several more ALSlinked mutations or aberrant toxic proteins as follows: C9orf72-PR 50 , C9orf72-GR 50 (Wen et al., 2014), and TDP43 A315T and used empty-GFP vector as a control. Transfected myocytes exhibit normal morphology and fusion in comparison with the WT muscle culture in our system ( Fig. 2-2, available at https:// doi.org/10.1523/JNEUROSCI.3037-17.2018.f2-2). Nevertheless, all ALS-causing mutations that we examined recapitulated the phenotypes we described previously in SOD1 G93A (Fig. 2F; GFP, 40.65 Ϯ 16%; GR 50 , 5.2 Ϯ 3.49%; PR 50 , 0 Ϯ 0%; TDP43 A315T , 8.75 Ϯ 6.39%). These results suggest that the dysregulated secretion of factors from ALS mutant muscles takes place, which in turn activates axon retraction and degeneration. Because muscles can either secrete positive or negative signaling molecules, we could not determine whether our observation within this assay originates from an increase in the release of destabilizing factors or the diminished release of positive factors. To this end, we collected muscle-CM from WT and SOD1 G93A muscle cultures in complete medium containing positive factors, such as BDNF and GDNF, as was previously performed in mass culture (Nagai et al., 2007), and ultimately applied it only to the distal axons of both WT and SOD1 G93A MNs (Fig. 2G,H ). Interestingly, we observed that axon degeneration occurs only when SOD1 G93A myocyte-CM is applied to SOD1 G93A axons ( Fig. 2H; LM, 3.72 Ϯ 1.15%; SOD1 G93A , 34.7 Ϯ 4%), whereas in all other combinations the axons remained intact (Fig. 2-3, available at https://doi.org/ 10.1523/JNEUROSCI.3037-17.2018.f2-3). To further determine whether Type 3 Semaphorins, such as Sema3A, contribute to MN axon degeneration in this assay, we investigated whether NRP1blocking antibody application can block this phenotype. Here again, we observed a rescue effect by this treatment (Fig. 2H; SOD1 G93A ϩ NRP1-antibody 18.6 Ϯ 7%), although the protection was incomplete. These results reinforce our hypothesis, suggesting that ALS-mutated muscles secrete destabilizing factors, such as Sema3A. Importantly, these results emphasize that SOD1 G93A MNs exhibit a higher sensitivity to degeneration, and support the MN unique vulnerability as well as the non-cellautonomous mechanism of ALS. Interestingly, previous attempts to block Sema3A signaling in SOD1 G93A mice using either a similar antibody approach or by crossing transgenic mice expressing a truncated form of Sema3A with SOD1 G93A mice also resulted in only a mild improvement or none at all of motor functions (Venkova et al., 2014;Moloney et al., 2017). This suggests that Sema3A plays a complex role in MNs and that perhaps other related proteins are involved. This also led us to investigate whether a wider deregulation of secreted factors released by the diseased muscles exists. Indeed, examining other members of the Semaphorin family, we found that the percentage of NMJs expressing Sema3B, as well as NRP2 is elevated in the SOD1 G93A ALS model ( Fig. 2-4, available at https://doi.org/10.1523/JNEUROSCI. 3037-17.2018.f2-4). Therefore, we concluded that the destabilizing effect of ALS muscles over MN axons involves more than a single factor; thus, it cannot be blocked or rescued by targeting one factor at a time. Moreover, the multiplicity of effectors suggests that a higher-order regulator, such as miRNA, might be involved in this process.
One day after transfection, cells were resuspended and plated in xCELLigence multiwell electric plates. The next day, Sema3A was added to the cultures, and any morphological or adhesive changes were monitored by the impedance readout. As demonstrated by TIRF imaging (Fig. 3H ), adding Sema3A to responsive cells, such as U87MG cells, induces their detachment from the culture dish. This detachment can be measured as a decrease in impedance (Fig. 3-2 B, available at https://doi.org/10.1523/ JNEUROSCI.3037-17.2018.f3-2). Shortly after Sema3A was added to the cultures, cells expressing miR142 exhibited decreased impedance, whereas cells expressing miR126-5p did not respond to Sema3A in the medium and kept growing with a corresponding increase in impedance (Fig. 3I). Hence, we showed that the excess production of destabilizing factors in ALS is likely to be mediated downstream of a deregulation in miR126-5p. Red and green represent a high or low abundance of miRs, respectively. *p Ͻ 0.05 (Student's t test; n ϭ 3). B, miR126-5p was the most significantly downregulated miRNA in SOD1 G93A muscles (Student's t test; n ϭ 3, **p ϭ 0.003). C, qPCR analysis of P60 GC muscle extracts further validates the decrease in miR 126-5p in SOD1 G93A (n ϭ 3). D-G, qPCR analysis of Sema3A, NRP1, Sema3B, and NRP2 transcript levels in HeLa cells overexpressing either miR126-5p or miR142 demonstrates a reduction in their expression levels specifically under miR126-5p overexpression (Student's t test; n ϭ 3, *p ϭ 0.0438, *p ϭ 0.034, *p ϭ 0.031, and *p ϭ 0.0434, respectively). H, Representative TIRF images of U87MG cells reveal a detachment of the cell membrane from the culture dish surface after Sema3A is added to the culture medium. Scale bar, 10 m. I, Impedance recording of live cells over time shows that U87MG cells overexpressing miR126-5p are unresponsive to Sema3A added to the culture medium because their impedance continuously increases, whereas the impedance of U87MG cells overexpressing miR142 decreases after treatment (Figure 3 Overexpression of miR126-5p can block SOD1 G93A muscle toxicity We overexpressed miR126-5p in SOD G93A myocyte cultures and quantified Sema3A levels in their culture extract as well as in their CM. Western blot analysis indicated that Sema3A levels in both the culture extract and CM are depleted, compared with miR142 ( Fig. 4 A, B; the mean fold change over SOD1 miR142 : Muscle extract-SOD1 miR126 , 0.24 Ϯ 0.1; SOD1 miR142 , 1 Ϯ 0.4 Muscle-CM-SOD1 miR126 , 0.63 Ϯ 0.03; SOD1 miR142 , 1 Ϯ 0.13). Next, we investigated whether overexpressing miR126-5p in both SOD1 G93A and PR 50 myocytes can rescue the negative effect on MN growth that we observed before. To this end, primary myo- Figure 4. Overexpression of miR126-5p in primary SOD1 G93A myocytes blocks motor axon degeneration and preserves NMJ activity in a compartmental coculture. A, B, Western blot analysis of transfected myocyte extract overexpressing miR126-5p or miR142 and their CM validates miR126-5p as a regulator of Sema3A specifically in muscles. ERK was used as a loading control (Student's t test, n ϭ 3, *p ϭ 0.0499 and *p ϭ 0.05, respectively). C, Schematic view of the experimental procedure in D, E. HB9::GFP SC explants and primary myocytes of SOD1 G93A mice were cocultured in a MFC. The growth of the HB9::GFP axons was assessed both by time-lapse imaging of the microgroove compartment and by imaging axons that traversed the distal compartment. D, Representative time-lapse images and quantification of HB9::GFP axon growth when cocultured with SOD1 miR126 myocytes (top) or SOD1 miR142 myocytes (bottom). SOD1 miR126 myocytes in the distal compartment enhanced the axonal traversal of the distal compartment compared with the SOD1 miR142 myocytes. The data are shown as the mean rate of axons that traversed the distal compartment out of the total number of axons in each field Ϯ SEM (Student's t test, n ϭ 3, *p ϭ 0.04216). E, Representative time-lapse images and quantification of HB9::GFP axon growth when cocultured with PR 50 miR126 myocytes (top) or PR 50 miR142 myocytes (bottom). PR 50 miR126 myocytes in the distal compartment enhanced the axonal traversal of the distal compartment compared with PR 50 miR142 myocytes. The data are shown as the mean rate of axons that traversed the distal compartment out of the total number of axons in each field Ϯ SEM (Student's t test, n ϭ 3, ***p ϭ 0.0039). F, Representative myocyte contraction plot showing the bursting contractile behavior of innervated myocytes in vitro. G, Quantification of the percentage of innervated myocytes that contract in a bursting pattern shows a diminished rate of bursting behavior in SOD1 G93A myocytes compared with LM controls. SOD1 miR126 myocytes show an increase in the rate of bursting myocytes back to the LM levels. The data are shown as the mean percentage of bursting myocytes Ϯ SEM (Student's t test, n ϭ 3, *p ϭ 0.0291, *p ϭ 0.0156, **p ϭ 0.005656). ***p Ͻ 0.001.
blasts were transfected to overexpress either miR126-5p (SOD1 miR126 ; PR 50 miR126 ) or miR142 (SOD1 miR142 ; PR 50 miR142 ) and were then plated in the distal compartment of the MFC. Myoblasts were differentiated into mature myocytes while expressing the miRs for 7 d, after which HB9::GFP explants were cultured in the proximal compartment. Once axons reached the microgrooves, their extension toward the muscle compartment was monitored for 16 h (Fig. 4C). Evidently, cocultures with SOD1 miR126 and PR 50 miR126 myocytes retained wild-type behavior and manifested a clear rescue effect on the rate of axon traversal ( Fig. 4 D, E; the mean traversal rate of axons: SOD1 miR126 , 40.77 Ϯ 6.68%; SOD1 miR142 , 12 Ϯ 7.6%; PR 50 miR126 , 45.6 Ϯ 9.4%; PR50 miR142 , 16 Ϯ 3.6%). Thus far, we have shown that myocytes expressing various ALS-linked mutations facilitate MN axon degeneration and delay their growth in a simplified compartmental coculture assay. However, while observing the cocultures for longer periods, we found that axons eventually do traverse the muscle compartment and form functional synapses with the myocytes. Using an image-based method that we recently developed to quantify contraction and assess the innervation in in vitro cocultures (Ionescu et al., 2016;Zahavi et al., 2017), we observed that the contractile behavior of innervated SOD1 G93A myocytes is significantly different from that of innervated LM myocytes, which tend to contract in a bursting pattern (Fig. 4 F, G; Movies 5, 6). Whereas 37% of innervated myocytes contract in a bursting pattern, only 18% of the innervated SOD1 G93A myocytes contract in this pattern (the mean rate of bursting innervated myocytes: LM, 37.23 Ϯ 2.8%; SOD1 G93A , 18.5 Ϯ 2.03%). Strikingly, SOD1 miR126 myocytes retain the same rate of bursting myocytes as the LM myocytes ( Fig. 4G; the mean rate of bursting innervated myocytes: SOD1 miR126 , 37.66 Ϯ 4.29%; SOD1 miR142 , 26.26 Ϯ 0.59%). Hence, miR126-5p is an effective regulator of muscle-secreted factors, such as Sema3, and can rescue the detrimental effect of destabilizing factors on MN axons, as well as on NMJ function and maintenance in vitro. miR126-5p transiently rescues early motor phenotypes of SOD1 G93A mice in vivo NMJ disruption, muscle morphology abnormalities, and hindlimb misprints are major phenotypes in SOD1 G93A mice (Gurney et al., 1994). To determine whether miR126-5p can moderate those phenotypes, we injected SOD1 G93A mice with either pLL-eGFP-miR126 (SOD1 miR126 ) or pLL-eGFP-miR142 (SOD1 miR142 ) into the right and left GC muscles of presymptomatic mice (P60), respectively (Fig. 5A). Virus expression was validated both in vitro on MNs and in muscle cultures (Fig. 5-1 A, available at https://doi.org/10.1523/JNEUROSCI. 3037-17.2018.f5-1) as well as in vivo at the transcript and protein levels ( Fig. 5-1 B, C, available at https://doi.org/10.1523B, C, available at https://doi.org/10. /JNEUROSCI.3037-17.2018.f5-1). Importantly, we observed a decrease in the number of NMJs expressing Sema3A in the pLL-eGFP-miR126-5p-injected GC muscles in comparison with the miR142 group, suggesting that miR-126 -5p is active in the injected tissue ( Fig. 5-1 D, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f5-1). Next, we performed a series of histological analyses, followed by motor behavioral tests at two time points after injection: at the age at which mice typically begin exhibiting ALS phenotypes (P90) as well as in the disease end stage (P120) (Fischer et al., 2004). Because NMJ disruption is a key process in ALS, we sought to determine whether overexpression of miR126-5p results in reduced NMJ disruption. Briefly, both the left and right GC muscles were fixed and stained for synaptic markers of the NMJ (Fig. 5B). Quantifying the percentage of intact NMJs at P90 injected mice revealed a significantly higher innervation rate in miR126-5pexpressing muscles compared with both mock-treated and SOD1 G93A muscles ( Fig. 5C; P90: WT, 71.58 Ϯ 3.32%; SOD G93A , 42.58 Ϯ 2.64%; SOD1 miR126 , 64.25 Ϯ 5.8%; SOD1 miR142 , 46.54 Ϯ 7.2%). Furthermore, careful analysis at P120 also identified a mild rescue by miR-126 -5p overexpression. (Fig. 5C; P120: WT, 74.35 Ϯ 4.74%; SOD G93A , 20.12 Ϯ 5.01%; SOD1 miR126 , 30.82 Ϯ 3.97%; SOD1 miR142 , 18.18 Ϯ 3.12%). Next, we performed histological analyses to determine muscle fiber wasting and tissue abnormalities (Fig. 5 D, E). P120 GC muscles of WT, SOD1 G93A , and both SOD1 miR126 and SOD1 miR142 were stained with H&E for histological examination, and the minimal diameter of myofibers was analyzed as described in Materials and Methods. We observed a mild, but significant, increase in the minimal fiber size of the SOD1 miR126 -injected muscle compared with the SOD1 miR142 mock control ( Fig. 5C; P120: WT, 40.25 Ϯ 2.28%; SOD G93A , 18.5 Ϯ 0.64%; SOD1 miR126 , 23 Ϯ 1.87; SOD1 miR142 , 19 Ϯ 1.47). Last, we performed a behavioral test using the CatWalk gait analysis technique. This video-based method is a computerized version of the ink bath assay and provides an objective and dynamic wide range of gait analyses (Deumens et al., 2007). Moreover, it has been used before specifically with the SOD1 G93A mouse model and displayed significant differences in several parameters (Mead et al., 2011;Gerber et al., 2012) (Fig. 5F ). One output is the Mean Stand Index (MSI), which measures the speed at which the paws detach from the walking surface. Because aged SOD1 G93A mice suffer motor defects, their MSI values for both hindlimbs are dramatically lower than their LM values. Remarkably, the MSI values of the SOD1 miR126injected limbs were significantly higher at P90 and similar to the LM control values, whereas the SOD1 miR142 -injected limb was reminiscent of typical SOD1 G93A behavior ( Fig. 5G; mean fold change over WT: SOD1 G93A , 0.68 Ϯ 0.02; SOD1 miR126 , 0.74 Ϯ 0.06; SOD1 miR142 , 0.65 Ϯ 0.04; LM, 1 Ϯ 0.04). We also examined other established parameters that have been shown to be altered in the SOD G93A model (Mead et al., 2011). We specifically focused on the percentage of single-support parameter, which indicates the relative duration of contact of all combined paws with the glass floor, and on the base of support parameter, which indicates the average width of limb spreading between both front, or both hindpaws. Remarkably, we observed a significant rescue phenotype for both parameters in the injected mice at age of P90.  . pLL-eGFP-miR126-5p injected into GC muscles of presymptomatic SOD1 G93A mice transiently rescues the early phenotype appearance in vivo. A, Schematic view of the in vivo experimental procedure. SOD1 G93A mice were injected with either pLL-eGFP-miR126-5p or pLL-eGFP-miR142 in their right or left GC muscles, respectively. Viral infection was validated ( Figure 5-1, available at https://doi.org/10.1523/JNEUROSCI.3037-17.2018.f5-1). B, Representative whole-mount NMJ immunostaining of ϳP90 SOD1 G93A GC muscles injected with either miR126-5p or miR142 lenti vectors. Red represents BTX. Green represents NFH ϩ synaptophysin in presynaptic neurons. Scale bar, 20 m. C, The percentage of innervated NMJs in miR126-5p-injected muscles is higher compared with its controls in both P90 and P120 (P90: Student's t test, n ϭ 6, *p ϭ 0.0475, **p ϭ 0.001245; P120: Student's t test, n ϭ 5, *p ϭ 0.043, **p ϭ 0.0096). D, Representative histological images of P120 WT, SOD1 G93A , miR126-5p, and miR142 GC muscle cross sections after H&E staining. Scale bar, 100 m. E, Semiquantification of a GC cross section from D shows a significant increase in the minimal muscle fiber diameter of muscles that were injected with miR126-5p (P120: Student's t test, n ϭ 4, *p ϭ 0.031). F, Illustration of the CatWalk XT gait analysis system that monitors mouse footprints. G, Gait analysis MSI parameter indicates the speed at which the paw loses contact with the surface. The MSI for the P90 miR126-5p-injected limbs was significantly higher than for miR142-injected limbs (Student's t test, *p ϭ 0.0355). H, Gait analysis percentage single-support parameter indicates the relative duration of contact of a single paw on the glass floor. The percentage in which the injected animals were used along the run with a single paw was significantly higher compared with SOD1 G93A mice and showed similarity to the WT control (Student's t test, SOD1 G93A -injected, ***p ϭ 0.0004; WT-SOD1 G93A , ***p ϭ 0.000003). I, Gait analysis base of support parameter indicates the average width between the hindpaws. The base of support of both P90-and P120-injected mice was significant higher compared with SOD1 G93A (Student's t test, P90 SOD1 G93A -injected, ***p ϭ 0.0000006; WT-SOD1 G93A , ***p ϭ 0.000007; P120 SOD1 G93A -injected, ***p ϭ 0. 0.00003; WT-SOD1 G93A , ***p ϭ 0.000000009).
Furthermore, the improvement in base of support parameter persisted also in P120 (Fig. 5 H Together, we demonstrated in vivo that miR126-5p reduces the detrimental effects of muscle-secreted destabilizing factors, such as Sema3A, on MN axons and motor function in ALS models.

Discussion
In this work, we demonstrated that the muscle toxicity in ALS is mediated by miR126-5p. We provided one specific mechanism for a well-described molecule (Sema3A), by which miR126-5p contributes to ALS pathology. We have also demonstrated that miR126-5p alterations facilitate axon degeneration and NMJ disruption in an ALS model as an outcome of presymptomatic elevations in the production and secretion of their target genes, which encode for destabilizing factors, such as Sema3 family members. Overexpressing miR126-5p in SOD1 G93A muscles inhibits the neurodegenerative process. These findings reveal how alterations in miR126-5p can be toxic to MNs, and identify a noncell-autonomous neurodegeneration process in ALS (Fig. 6).

miR126-5p as a master regulator of proper NMJ function
Our results indicate that the expression of ALS-causative mutations results in the secretion of repellent factors, including a number of Type 3 Semaphorins and potentially other factors. It is thus likely that a general gene repression mechanism, specifically miR system, is altered under such conditions. This assumption is also consistent with the fact that miR alterations are apparent in various neurodegenerative diseases, including ALS (Haramati et al., 2010). Here we identified such an miR and showed how alter-ations in this specific miR can regulate the essential signaling pathways in MNs and can trigger neurodegeneration. Intriguingly, and in line with our findings, a very recent paper demonstrated a mechanism by which miR126-5p modulates Sema3A expression through SetD5 expression, and it emphasizes its positive effect on retinal endothelial cells' survival (Villain et al., 2018). However, aside from targeting Sema3A and Sema3B, as well as NRP1 and NRP2, miR126-5p is predicted to regulate other Semaphorin signaling factors, such as Sema6D, PLXNA2, JNK2, JNK3, and PTEN. In addition, miR126-5p can regulate the ALS and motor-unit-related genes VEGF-A, SPAST, MMPs (Kaplan et al., 2014), AGRIN (Vilmont et al., 2016), and C9orf72, which are directly involved in ALS. Therefore, miR126-5p can serve as a master regulator of NMJ health by controlling multiple signaling pathways.

Sema3 alterations in ALS: settling the contradictory reports
A critical initiating event for the mechanism outlined above is the alterations in Sema3 signaling in ALS models and patients. Sema3 family members were previously found to be upregulated in terminal Schwann cells in the NMJs of SOD1 G93A mice . Recently, Sema3A was also shown to be elevated in the motor cortex of postmortem ALS patients but not in their SC (Körner et al., 2016). Consistent with this, NRP1 antibodies, blocking the obligatory binding receptor for Sema3A, were injected into SOD1 G93A mice as a potential treatment (Venkova et al., 2014). However, anti-NRP1 blocking antibody had only a modest effect. Moreover, a recent study demonstrated that crossing mice expressing a truncated form of Sema3A with SOD1 G93A mice did not result in any rescue effect (Moloney et al., 2017). An explanation for a minor effect or not at all, as a result of blocking Sema3A activity, could be based on the idea of multiple toxic factors that play a role in ALS pathology. Another explanation for this contradiction could be the fact that Sema3A plays a more complex role in the biology of MNs. Indeed, Sema3A was shown to increase survival when added to mass cultures of mouse MNs (Molofsky et al., 2014) and human MNs (Birger et al., 2018). Consistent with this, deletion of the Sema3A gene specifically in spinal astrocytes resulted in a gradual loss of spinal MNs (Molofsky et al., 2014), thus suggesting that Sema3A has a positive effect when introduced near the cell body. When these findings are together with our results, apparently Sema3A has both positive and negative effects on MNs, perhaps depending on its specific subcellular localization. When Sema3A is secreted from muscles and targets distal axons at NMJs, it mediates their destabilization; however, when it is secreted by spinal astrocytes and targets MN soma, it acts as a survival factor. Thus, it is perhaps not surprising that a genetic approach to inhibit Sema3A in all cells in a mouse model of ALS had no effect in inhibiting the symptoms. The injection of anti-NRP1 may have been a bit more beneficial possibly either because it was able to inhibit Sema3A outside the CNS more effectively, or alternatively, because NRP1 blocks other Type 3 Semaphorins as well. Figure 6. Alterations in Semaphorin3A regulation by miR126-5p trigger MN degeneration in ALS. miR126-5p is a negative regulator of Sema3 signaling in skeletal muscles. Downregulation of miR126-5p in ALS disease drives the overexpression and secretion of Sema3A and potentially other NMJ-destabilizing factors in skeletal muscles. The downregulation in miR126-5p in diseased MNs drives the overexpression of NRP1 specifically in axons. The excess binding and activation of the NRP1 receptor by its overexpressed ligand Sema3A as a result of miR126-5p alteration promote NMJ disruption and axon degeneration in a spatially confined process.