Identi�cation of a Novel Interaction of FUS and Syntaphilin May Explain Synaptic and Mitochondrial Abnormalities Caused by ALS Mutations

Aberrantly expressed fused in sarcoma (FUS) is a hallmark of FUS-related amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). Wildtype FUS localises to synapses and interacts with mitochondrial proteins while mutations have been shown to cause to pathological changes affecting mitochondria, synapses and the neuromuscular junction (NMJ). This indicates a crucial physiological role for FUS in regulating synaptic and mitochondrial function that is currently poorly understood. In this paper we provide evidence that mislocalised cytoplasmic FUS causes mitochondrial and synaptic changes and that FUS plays a vital role in maintaining neuronal health in vitro and in vivo. Overexpressing mutant FUS altered synaptic numbers and neuronal complexity in both primary neurons and zebra�sh models. The degree to which FUS was mislocalised led to differences in the synaptic changes which was mirrored by changes in mitochondrial numbers and transport. Furthermore, we showed that FUS interacts with the mitochondrial tethering protein Syntaphilin (SNPH), and that mutations in FUS affect this relationship. Finally, we demonstrated mutant FUS led to changes in global protein translation. This interaction between FUS and SNPH could explain the synaptic and mitochondrial defects observed leading to global protein translation defects. Importantly, our results support the ‘gain-of-function’ hypothesis for disease pathogenesis in FUS-related ALS.


Background
Amyotrophic Lateral Sclerosis (ALS) and Frontotemporal Dementia (FTD) are both fatal neurodegenerative diseases which show a large degree of clinical, pathological and genetic overlap between patients (1).Mutations in Fused in sarcoma (FUS) are found in 5% of ALS patients while pathological FUS aggregates are found in 10% of FTD patients (2)(3)(4).The majority of FUS mutations reside within the C-terminus of the protein which contains the nuclear localising signal (NLS) and is essential for FUS to tra c between the nucleus and cytoplasm (5)(6).Under physiological conditions, FUS is a predominately nuclear protein which has a role in RNA processing (7).However, a growing body of evidence has shown that FUS has an essential cytoplasmic role in neurons involved in transport and local protein translation, especially at the synapse.FUS has been shown to be located at both the preand post-synapse (8-9) as well as interacting with GluA1, the AMPA receptor subunit, which is a postsynaptic protein (10).Evidence points towards FUS being essential for dendritic reorganisation by regulating synaptic mRNAs (11)(12) with FUS depletion leading to a decrease in GluA1 expression resulting in supressed synaptic transmission and changes in synaptic maturation (10).Notably, FUS has been implicated in the regulation of splicing and the transcription of synaptic mRNAs which could be involved in local translation at the synapse (13)(14).Mutations in the NLS lead to varying levels of accumulated mutant FUS in neuron terminals and have been shown to cause signi cant hypomethylation of arginine's which decreases new protein synthesis (15).Therefore, accumulating evidence points towards FUS having a vital cytoplasmic role for mRNA transport within axons and dendrites to facilitate local translation at the synapse (16).However, how mutated cytoplasmic FUS contributes to synaptic degeneration still needs to be elucidated.
Mitochondria are essential for synaptic plasticity, spine development and general neuronal function with ATP demand being correlated to synaptic integrity within a dendrite (17).The pathogenesis of numerous neurodegenerative diseases, including ALS, has been associated with mitochondrial dysfunction and demonstrates that mitochondrial function is essential for maintaining neuronal integrity (18)(19)(20).In ALS speci cally, there is accumulating evidence that degeneration of the NMJ, accompanied by mitochondrial abnormalities, is an early pre-symptomatic disease event (21)(22)(23)(24).FUS has been shown to interact with mitochondrial proteins (HSP60; ATP synthase β-subunit) and mutations in FUS which increase cytoplasmic FUS have been associated with mitochondrial fragmentation within neurons (25)(26).
Additionally, it has been shown that a subset of FTD patients show an increase in FUS expression within damaged mitochondrial cristae (25).It is therefore clear that healthy mitochondria are essential for synaptic functioning in the context of ALS, however the involvement of FUS is relatively unknown.
Here, we report in vitro and in vivo evidence that overexpression of mutant FUS causes differential synaptic defects which appear to depend on the level of mislocalised cytoplasmic FUS.Additionally, we have provided evidence for a relationship between synaptic and mitochondrial abnormalities due to mutant FUS by identifying a novel interaction between FUS and the mitochondrial anchor protein, syntaphilin (SNPH) which is essential for synaptic maintenance.These results indicate that FUS is intricately involved in synaptic and mitochondrial functioning and that the degree of mislocalised FUS can lead to speci c abnormalities which contribute to neurodegeneration.

Results
FUS is enriched at, and colocalises with pre-and postsynapses in rat primary neurons To investigate whether FUS was localised to the synapse, we performed immunocytochemistry using rat primary cortical neurons to co-stain for endogenous FUS and pre-and post-synaptic markers.Neurons were aged to DIV21 to ensure expression of both pre-and post-synaptic markers.Besides its expected localisation within the nucleus, FUS was also localised within puncta in neurites (Fig. 1).FUS colocalised with 75% of synaptophysin puncta (Fig. 1A, SYN, pre-synaptic) and 50.25% of post synaptic density-95 (Fig. 1B, PSD-95) puncta.This indicates that FUS is present on both sides of synaptic buttons with a preferential localisation for the pre-synapse at DIV21 (p < 0.05).

Mutant FUS leads to alterations at the synapse
To investigate whether mutations in FUS affected the synapse, we over-expressed two different mutant forms of FUS in primary neurons.R514G is an NLS point mutation originally identi ed in 2009 in a British Family and replicated in German ALS patients that results in a moderate increase of cytoplasmic FUS in cell culture models (FUS R514G ) (2,6,28).Secondly, we created a truncation mutation (K510X) that results in the loss of the entire NLS from the C-terminus of the protein showing a predominantly cytoplasmic localisation (FUS ΔNLS ) (6). eGFP-FUS WT , eGFP-FUS R514G and eGFP-FUS ΔNLS were transfected into DIV6 rat primary neurons alongside control eGFP-only to investigate if a change in pre-synaptic SYN puncta was observed (Fig. 2).As expected eGFP-FUS WT was predominantly nuclear whilst the mutant proteins misclocalised to the cytoplasm (Fig. 2A).There was no signi cant difference in the number of SYN puncta between the control and eGFP-FUS WT conditions indicating that transfection of eGFP-FUS WT did not in uence the number of SYN puncta (Fig. 2C).However, there was a signi cant increase in the number of SYN puncta of the neurons in the eGFP-FUS R514G compared to eGFP-FUS WT (p < 0.05) and a signi cant decrease in the number of SYN puncta expressing eGFP-FUS ΔNLS compared to eGFP-FUS WT (p < 0.001) suggesting potentially different effects of the two mutations (Fig. 2C).Secondly, as mutant FUS has been shown to affect dendritic branching in mouse models, an analysis of this was undertaken by analysing the MAP2 staining of transfected neurons (16).This showed that only transfection of eGFP-FUS ΔNLS led to a signi cant decrease in the ability to grow dendritic branches compared to the control (Fig. 2B, D & E).
We also investigated if a similar effect would occur on the post-synaptic side by staining for the common post synaptic marker PSD-95.DIV14 Rat primary neurons were transduced with either HA-FUS WT , HA-FUS R514G or HA-FUS ΔNLS before being grown to DIV21 to ensure full development of the post synapse (Fig. 3A).As we observed with the pre-synaptic side, expression of HA-FUS R514G resulted in a signi cant increase in PSD-95 puncta when compared to HA-FUS WT (p < 0.05), whilst HA-FUS ΔNLS led to a signi cant reduction in PSD-95 puncta compared to wildtype (p < 0.001) (Fig. 3A&C).In this case transduction of either WT or mutant FUS led to a reduction in branching compared to the control at this stage (Fig. 3B,   D&E).This suggests that perhaps any increase in FUS, in this case by overexpression, may lead to synaptic disruption.

Mutant FUS affects the NMJ in zebra sh
To investigate whether synaptic changes were also seen in an in vivo model, we developed a transient transgenic approach in zebra sh to visualise and temporally monitor the motor neuron.In our model, expression of GFP-tagged FUS was driven by an mnx1 promoter which has previously been shown to drive expression in motor and interneurons of the zebra sh spinal cord (29,30).We used the Gal4/UAS system to express eGFP-FUS WT , eGFP-FUS R514G and eGFP-FUS ΔNLS to explore changes speci cally in ventrally innervating primary motor neurons.Previous data has shown that FUS is abundant at the NMJ and that denervation of the NMJ is an early pathological hallmark of ALS-FUS (22,31).Therefore, we sought to establish whether mutant FUS disrupted the formation of the NMJ.Zebra sh co-injected with MNX1:Gal4 and either UAS: eGFP-FUS WT , UAS: eGFP-FUS R514G or UAS: eGFP-FUS ΔNLS , were xed and stained for the pre-synaptic NMJ marker SV2 and the post-synaptic NMJ marker alpha-bungarotoxin (αBTX) at 2 days post fertilisation (Fig. 4).We observed a signi cant reduction in the number of BTX and SV2 puncta in cells expressing eGFP-FUS R514G (p < 0.05) and eGFP-FUS ΔNLS (p < 0.05) compared to eGFP-FUS WT (Fig. 4B).Furthermore, investigation into the degree of colocalization between BTX and SV2 showed a signi cant reduction for both eGFP-FUS R514G (p < 0.05) and eGFP-FUS ΔNLS (p < 0.0001) compared to eGFP-FUS WT (Fig. 4B).
Following our discovery that the different mutations in FUS led to substantial synaptic defects, we sought to investigate the extent of the developmental defect of the misexpressing caudal primary motor neuron.
Expression of MNX-speci c eGFP-FUS R514G or eGFP-FUS ΔNLS resulted in a reduction in the length of the primary motor axons when compared to eGFP-FUS WT (Fig. 5A&B).Interestingly, this reduction only reached statistical signi cance in the ΔNLS mutant (p < 0.05, Fig. 5A&B).However, the expression of both mutants led to signi cant decreases in the number of secondary and tertiary branches.There was a signi cant reduction in secondary motor neuron branches in both the eGFP-FUS R514G (p < 0.05) and eGFP-FUS ΔNLS (p < 0.001) expressing zebra sh when compared to eGFP-FUS WT (Fig. 5B).Whilst there are never large numbers of tertiary branches (eGFP control averaged only 4 tertiary branches), eGFP-FUS WT motor neurons expressing cells presented an average of 2.83 tertiary branches per axon whilst none were detected in any analysed motor neurons expressing eGFP-FUS R514G or eGFP-FUS ΔNLS (Fig. 5A&B).This may indicate a delay in branching in neurons expressing higher levels of cytoplasmic FUS.Subsequently, we sought to con rm this change in axonal complexity in this in vivo system by using sholl analysis (Fig. 5C&D).This demonstrated that axonal branching was signi cantly reduced within primary motor neurons in both eGFP-FUS R514G (p < 0.05 − 0.001) and eGFP-FUS ΔNLS (p < 0.05-0.0001)expressing motor neurons compared to eGFP-FUS WT (Fig. 5C&D).

Mutant FUS affects neuronal mitochondria
We have previously shown that alterations to the morphology and a reduction in the number of mitochondria are very early disease events in a mouse model of ALS-FUS (22).We proceeded to investigate if alterations to mitochondria could explain the observed in vitro and in vivo synaptic effects caused by mutant FUS.To do this we co-transfected rat primary neurons with eGFP, eGFP-FUS WT , eGFP-FUS R514G or eGFP-FUS ΔNLS together with DS-MitoRed, which localises to mitochondria and investigated the number and size of mitochondria within the neurites of individual neurons (Fig. 6).In the eGFP-FUS R514G cells, there was a non-signi cant increase in the number of mitochondria compared to eGFP-FUS WT expressing neurons (p = 0.12) whilst in eGFP-FUS ΔNLS there was a signi cant loss of mitochondria compared to eGFP-FUS WT (p < 0.01) (Fig. 6A&B).To investigate the health of the mitochondria we investigated morphological defects by assessing the size of the mitochondria themselves.Whilst there was no difference in size when comparing eGFP-FUS ΔNLS to eGFP-FUS WT neurons (p > 0.99), there was a striking increase in the average size of the mitochondria in cells transfected with eGFP-FUS R514G compared to eGFP-FUS WT (p < 0.05) (Fig. 6A&B).To investigate whether this affected mitochondrial transport, live imaging of the mitochondria in the transfected neurons was performed (Fig. 6C&D).An analysis of the overall motility of mitochondria in the transfected cells showed that there was a signi cant reduction in movement when eGFP-FUS WT was transfected compared to the eGFP only control (Fig. 6D, p < 0.01).The presence of the eGFP-FUS ΔNLS mutant led to an even greater reduction (Fig. 6D, p < 0.01) in the overall motility of mitochondria compared to any other condition.However, in contrast, there was a non-signi cant increase in movement in the eGFP-FUS R514G transfected neurons (Fig. 6D, p = 0.34).To determine whether there were directional differences in the motility of the mitochondria, we analysed the anterograde or retrograde movement separately.Whilst both eGFP-FUS WT and eGFP-FUS R514G showed reduced anterograde movement compared to the eGFP control (Fig. 6D, p < 0.05), there was an even greater loss of movement in the eGFP-FUS ΔNLS mutant (Fig. 6D, p < 0.0001).Furthermore, when assessing retrograde movement there was no difference when comparing eGFP-FUS WT to the eGFP control (Fig. 6D, p = 0.66).There was however a signi cant increase when comparing eGFP-FUS R514G to eGFP-FUS WT or eGFP control conditions (Fig. 6D, p < 0.05).As with the anterograde movement, the presence of eGFP-FUS ΔNLS resulted in almost no retrograde transport of mitochondria.Interestingly our results demonstrate a striking mitochondrial phenotype for each respective mutation as FUS ΔNLS leads to a reduction in mitochondria and a complete loss of mitochondrial movement whereas FUS R514G appears to lead to more swollen mitochondria which are more frequently moving towards the soma.

FUS interacts with the mitochondrial anchor, Syntaphilin (SNPH)
We next examined if FUS interacted with mitochondrial proteins directly which could explain these effects of mutant FUS.We speci cally looked at Syntaphilin (SNPH) due to its signi cant role in mitochondrial anchoring and its relationship to the synapse (32)(33).In order to con rm that SNPH was localised to mitochondria, we used super resolution microscopy (iSIM) to investigate the co-localisation of endogenous SNPH and a mitochondrial marker, TOM-20.This showed that TOM20 puncta colocalised with ~ 80% of SNPH (See Additional File 1).Next, we investigated the localisation of endogenous FUS and SNPH within neurons.Results indicate that both FUS and SNPH form puncta in the soma and neurite and that FUS puncta colocalised with ~ 72% of SNPH (Fig. 7A).In order to determine whether FUS and SNPH more closely interacted, we used a Proximity Ligation Assay (PLA) which detects protein-protein interactions which are < 40nm apart.PLA indicated that FUS was in close proximity to SNPH (Fig. 7B&C) and that a stronger interaction was found in the soma when compared to neurites (Fig. 7B&C, p < 0.0001).

Mutant FUS leads to changes in interactions with SNPH
Having shown that FUS interacts with SNPH, we investigated whether overexpression of WT and mutant FUS would change the FUS-SNPH interactions as mutations in FUS are known to alter interactions with mitochondrial proteins such as HSP60 and ATP5B (25)(26).HA-FUS WT , HA-FUS R514G and HA-FUS ΔNLS were transfected into rat primary neurons, along with eGFP-SNPH to ensure there was su cient signal from both proteins, and a PLA was performed (Fig. 8).PLA indicated that there was a signi cant interaction between HA-FUS WT and eGFP-SNPH and this was predominantly focused in the soma.The presence of HA-FUS R514G led to a signi cant decrease in FUS-SNPH interactions when compared to HA-FUS WT (Fig. 8, p < 0.01) which correlates with the increased movement of mitochondria seen with the R514G mutant.Surprisingly though there was no alteration in the interaction of HA-FUS ΔNLS compared to HA-FUS WT suggesting that although this mutation almost abolishes the movement of mitochondria, it may not be due to an alteration in its interaction with SNPH as measured here.
Protein translation is impaired in the presence of mutant FUS Previously, it has been shown that mutations in FUS reduce axonal protein synthesis (34) and so to determine whether we saw a similar phenotype, we investigated whether protein translation was affected in our cellular model.We used the surface sensing of translation (SUnSET) assay in which puromycin, a structural analogue of aminoacyl tRNAs, is incorporated into nascent polypeptides and prevents elongation, allowing us to directly monitor translation (35).Primary cortical neurons were transfected with eGFP-FUS WT , eGFP-FUS R514G and eGFP-FUS ΔNLS respectively before puromycin treatment and analysed by measuring the intensity within the soma and neurites (Fig. 9A).The presence of eGFP-FUS ΔNLS led to a small non-signi cant decrease in protein translation in the soma when compared to eGFP-FUS WT (Fig. 9B, p = 0.5704).In contrast, eGFP-FUS R514G led to a small increase in translation in the soma when compared to eGFP-FUS WT (Fig. 9B, p = 0.4042).Analysis of the protein synthesis in the neurites showed that there was a similar pattern that this time reached signi cance in the eGFP-FUS R514G transfected neurons when compared to eGFP-FUS WT (Fig. 9C, p < 0.01) whilst eGFP-FUS ΔNLS still showed a small decrease when compared to eGFP-FUS WT (Fig. 9C, p = 0.5704).Interestingly, overexpression of eGPF-FUS WT alone reduced the amount of translation compared to the control which matches previous data regarding mitochondrial movement suggesting that the two processes are closely linked.

Discussion
In this study, we have shown that mutations in FUS led to alterations in synaptic protein expression and reduced the complexity of neurites and axons in vitro and in vivo and that these defects correspond to mitochondrial abnormalities observed in neurites with each respective FUS mutant.More importantly we have generated novel data showing FUS interacts with the mitochondrial anchor protein SNPH in neurons, and that mutations in FUS alter these interactions.Finally, we show that mutant FUS alters protein translation at the soma and in particular, neurites.Overall, we have presented evidence which supports a possible relationship between synaptic and mitochondrial function and neuronal health in which FUS appears to be a key player.It is well known that neuronal mitochondria are highly dynamic and transported through the neuron to regions of high metabolic demand.Therefore, it is possible that ATP demand is correlated to synaptic integrity within a neuron (17) and an increase in mitochondrial number in FUS R514G could explain the observed increase in synapses.Conversely, a decrease in mitochondrial number and therefore ATP, could explain the associated decrease in synapses in FUS ΔNLS leading to the degeneration of the neuron.Taken together, this data adds to the accumulating body of evidence that FUS plays a direct role in mitochondrial and synaptic function, and that the level of mislocalised cytoplasmic FUS can lead to varying effects on neuronal function (8; 26).One of the more interesting aspects of this work was that speci c FUS mutations can lead to different cellular phenotypes in vitro, suggesting that the degree to which FUS is mislocalised can have differential downstream consequences.Even an increase in the amount of wildtype protein was su cient to cause some cellular phenotypes such as reducing mitochondrial movement.Given that there are ALS patients who have 3' UTR mutations (36) that lead to an increase in wildtype protein and that there are mouse models in which this increase alone is su cient to cause an ALS like phenotype, it is not a surprise that we see such occurrences.Moreover, when we introduced a mutation that speci cally and moderately increases cytoplasmic FUS (FUS R514G ), this results in an increase in synapses, and mitochondrial number, size, and speed of axonal transport.Whereas FUS ΔNLS , a mutation that results in complete abolishment of the nuclear-localising signal and a very large increase in cytoplasmic FUS, contrasts FUS R514G by demonstrating synaptic and mitochondrial de cit with a complete loss of axonal transport.Patients with truncation mutations suffer from a very young onset and aggressive form of ALS (27).In comparison those with a FUS R514G mutation have a later onset and longer form of the disease though this is still often more severe than those with the sporadic form of the disease (2; 5).This suggests that the FUS R514G phenotype might represent an early disease response to the increased cytoplasmic FUS and that the FUS ΔNLS phenotype might mimic an aggressive end stage timepoint.It is worth noting that we see different phenotypes in vivo and in vitro for FUS R514G .This could be due to the increased sensitivity of motor neurons to an increase in mislocalised cytoplasmic FUS, hence why motor neurons are more selectively affected in ALS-FUS patients.
In this study, we showed that each mutation causes signi cant changes in mitochondrial transport.Mitochondria are transported anterogradely from the soma towards the synapse due to the high metabolic demand (32).We show that FUS and SNPH interact and localise together in neurites.This is of importance as SNPH acts as a stable anchor for mitochondria and is essential for synaptic modi cation and functionality by ensuring the presence of mitochondria near synapses (32).Interestingly, neurons overexpressing FUS R514G show a decrease in interactions with SNPH within the soma.This observation ts with our data, showing a greater number of motile mitochondria being transported within the neuron in cells transfected with FUS R514G .This altered interaction could explain both the potential increase in mitochondria numbers in the neurite and the increased retrograde transport we observe within our in vitro dataset.It is possible that the neuron is trying to compensate for the excess cytoplasmic FUS and stay functional by taking damaged mitochondria back to the cell body to be degraded (37).However, we observe a non-signi cant decrease in FUS-SNPH interactions in neurons expressing FUS ΔNLS within the soma and neurite.This suggests that although there are a similar number of interactions when compared to FUS WT , we also showed that there were fewer mitochondria anyway so those that exist could be trapped within the soma or stationary in the neurite and not being transported.This could explain the decrease in mitochondrial transport and the overall number of mitochondria.
Our data also con rms previous data showing that mutations in FUS affect protein synthesis (34,38).Moreover, translation defects appear to be speci c to the degree of mislocalised cytoplasmic FUS depending on the mutation present.As with our previous data, FUS R514G led to an increase in translation in affected soma and neurites whereas FUS ΔNLS led to a decrease.Previous reports have demonstrated that mutant FUS interacts with polyribosomes and that a toxic 'gain of function' in the cytoplasm affects translation (38).Therefore, it is likely that the changes in global translation we observe lead to the synaptic abnormalities exhibited by each speci c mutant.
We have presented in vivo and in vitro evidence that FUS is essential for maintenance of neuronal health and that speci c FUS mutations can cause differing mitochondrial and synaptic disruption, depending on the degree of cytoplasmic mislocalisation.This might explain some previously con icting reports on the effect of mutations in FUS on neuronal function.Future studies will be needed to better understand the interaction between FUS-SNPH to prove if changes in this interaction could partially explain the synaptic and mitochondrial defects observed in vitro and in vivo.

Cell culture
All neuronal culture techniques were performed under sterile conditions.Coverslips were coated with 1% Poly-D-lysine hydrobromide in PBS (Sigma) and incubated at 37°C overnight and washed with PBS prior to the seeding of rat primary cortical neurons at 70,000 cells/ml in 500µl/well.

Animal experiments
All animal experiments have been authorised by the KCL ethics Review Committee and under the HO license 70/7577.All animal experiments were performed in accordance with the relevant guidelines and following regulated procedures and all authors complied with the ARRIVE guidelines for animal research.
The study is reported in accordance with the ARRIVE guidelines and we have presented all details that allow for accurate follow up including group sizes, age, and detailed experimental procedures for each animal experiment.

DNA Transfection
For each transfection, 500ng DNA was mixed with 1µl of Lipofectamine 2000 (Invitrogen) in 25µl of HEPES (Gibco) and DMEM (Gibco) solution per well.Prior to addition of DNA, coverslips were removed and placed in a 350µl of fresh media (without pen/strep).DNA mix was incubated for 1 hour at 37°C before adding in a drop wise manner to DIV6 neurons.6-8 hours post transfection coverslips were replaced in the old media, before xation 48 hours later.

Viral transduction
Virus was added to rat primary neurons at DIV14 to achieve an infection rate of 1 x 107 virus particles/ml.Following the day of transduction, 250µl of cortical media was removed and 300µl of fresh cortical media was added onto the cells before xation occurred at DIV21.

Quantitative image analysis
Images were taken on a Leica TCS-SP5 laser scanning confocal microscope and imaged at x63 with a numerical aperture of 1.4 (with a digital zoom of 2.5).Images were taken at 10-14 Z-sacks with 0.5µm increment before being processed in Image J.

Colocalisation analysis
Images of selected neurites were extracted, and their length (100µm) was recorded.After splitting each channel, a Gaussian and median lter (with a radius of 10 pixels) was applied to the channel of interest, individually.The median image was subtracted from the Gaussian channel and a threshold was selected.
Particles were then analysed from the channel before overlaying puncta onto the other channel of interest to allow measurement of colocalization which was determined by subtracting the area of the overlaid puncta from the under laid puncta (colocalisation was only counted if 50% and over).Analysis was performed on three separate dendrites per cell (N = 9) from three individual experiments at DIV21.All statistical analysis to determine signi cance between groups was performed using GraphPad Prism 9 using a student's T-Test.

Analysis of synaptic puncta and mitochondria
Maximum projection images were converted to 8-bit grayscale and individual channels were then used to select the threshold which was kept consistent to the control (eGFP) channel.Puncta and mitochondria were thresholded to be bigger than ve pixels in size.Images were obtained from three independent experiments and three dendrites from ve different cells were analysed for each repeat for each condition.All statistical analysis to determine signi cance between groups was performed using GraphPad Prism 9 using a One-Way ANOVA with post-hoc tukey's multiple comparisons test.

Analysis of dendritic complexity
Maximum projection images of the MAP2 channel were then converted into an 8-bit gray scale image and dendrites were traced using an available plugin (Neuron J).After tracing, sholl analysis was performed to assess dendritic complexity from the soma at 10µm increments.Sholl analysis is a quantitative measure of the shape and/size of a dendritic tree.To measure dendritic complexity, concentric circles were drawn from the centre of the neuron and the number of times each dendritic branch intersects, branching is assessed (number of intersections divided by area against distance).Traces were analysed from 10 individual cells across three individual experiments per condition.All statistical analysis to determine signi cance between groups was performed using GraphPad Prism 9 using a Two-Way ANOVA with a post-hoc tukey's multiple comparisons test which compared the sample effects within each row.
Quantitative image analysis of mitochondrial dynamic using kymographs Cells were plated in an Ibidi 8 well plate at 50, 000 cell/ml.Following co-transfection with GFP-FUS constructs (250ng) and a Ds-Mitored plasmid (250ng) at DIV6, cells were incubated for 48hr before imaging on the Nikon Eclipse Ti Spinning disk confocal microscope at 63X.Images were taken every 30 seconds over 10 minutes.Channels were split using Image J and a line was drawn along a dendrite (100µm), to create a kymograph using an Image J plugin.After a kymograph had been created, individual particles were traced using neuron J plugin on Image J to calculate if a particle moved in a retrograde or anterograde direction.Stationary mitochondria were counted if there was no visible movement of that particle in the kymograph.Mitochondrial dynamics were performed on a single dendrite from eight individual cells across three individual experiments per condition.All statistical analysis to determine signi cance between groups was performed using GraphPad Prism 9 using a One-Way ANOVA with posthoc tukey's multiple comparisons test.

Proximity Ligation Assay
Proximity ligation assays (PLAs) were performed essentially as the manufacture instructions (Sigma-Aldrich).Brie y, neurons were xed in 4% paraformaldehyde in PBS and probed with mouse anti-FUS (1:200, Proteintech) and anti-SNPH (1:200, Proteintech), and signals developed using a Duolink In Situ Orange kit (Sigma-Aldrich).Following PLAs, neurons were immunolabeled for chicken anti-MAP2 (1:1000, AbCam).Images were taken at x60 (oil) on a Nikon Ti-E Two Camera microscope.Images were analysed in ImageJ and positive puncta counted using the cell counting tool.Five somas for each image were analysed for the soma count and 3 different neurites for each of ve cells per image were analysed with three biological replicates carried out.

Puromycin assay
Following transfection of DIV6 neurons as previously described, neurons were treated with 1mL/well of 1x ACSF (10x ACSF with H 2 O, Glucose 11mM and HEPES 5mM), MgCl 2 1.25 mM and CaCl 2 mM (pH 7.4) at DIV8.Following a 1-hour incubation, 5µL of puromycin (P8833, 10mg/mL, Sigma Aldrich) was added to each well.After 10 minutes, cells were xed and immunostained with mouse anti-puromycin (1:1000, 3RH11, Kerafast) and chicken anti-MAP2 (1:1000, AbCam).Images were taken with a Nikon iSIM super resolution microscope at x100 (oil) objective.Puromycin puncta were thresholded to be bigger than ve pixels in size and average intensity was calculated.Images were obtained from three independent experiments and ve different somas and neurites were analysed for each repeat for each condition and values were normalised to threshold.All statistical analysis to determine signi cance between groups was performed using GraphPad Prism 9 using a One-Way ANOVA with post-hoc tukey's multiple comparisons test.

Fish stock maintenance, husbandry and embryo collection
All Danio rerio lines were raised and maintained at 28°C on a 14hour light/ 10hour dark cycle in the Guy's Campus Zebra sh facility, London.Embryos were collected and incubated in dishes lled with system water with methylene blue in a 28°C incubator until experimentation.Morphological staging was used to determine embryo development (27).

Microinjection procedure
To deliver the plasmid into individual embryos, 1mm single capillary needles with lament (world precision instruments) were pulled on a model P-97 aming/brown micropipette puller (Sutter instrument Co.).Once the micropipette was created, 2.8µl of plasmid was taken up and attached to manual micromanipulator apparatus.0.5nl of plasmid solution (50ng/µl) was measured on a graticule (Pyser-SGI) and injected into a 1-cell stage embryo using the Picospritzer 111 microinjector (Parker instrumentation).

UAS: eGFP-FUS constructs
Homologous sticky-end restriction sites were used (PciI and NheI) to allow insertion of the UAS promoter.Initially, both the pN2 5UAS eGFP and the pC1 CMV eGFP-FUS were digested with PciI (NEB).Following digestion, each digested plasmid was puri ed and then digested with NheI (NEB).After each plasmid had been digested with both enzymes, the UAS insert and eGFP vector were gel extracted, and gel puri ed (Qiagen) before being ligated and transformed into competent cells (NEB).UAS: eGFP-FUS WT and eGFP-FUS R514G and eGFP-FUS ΔNLS constructs were microinjected at 25ng/µl along with 25ng/µl of MNX1:Gal4 plasmid.

Morphological analysis of motor neurons
Images were taken on a Nikon Eclipse C1 confocal microscope using a x40 water objective (N.A. 0.8).
Images were based on eGFP expression and z-stacks taken at 1µm increments.Maximum projection images of the eGFP channel were converted into an 8-bit gray scale image and axons were traced using an available plugin on Image J (Neuron J).After tracing, sholl analysis was performed to assess dendritic complexity from the soma at 10µm increments.To measure axonal length and branch numbers, axonal branches that were 0.5µm and larger were included in the axonal branching analysis.Neurons were

Figures
Figures