Protein disulphide isomerase (PDI) is protective against amyotrophic lateral sclerosis (ALS)-related mutant Fused in Sarcoma (FUS) in in vitro models

Mutations in Fused in Sarcoma (FUS) are present in familial and sporadic cases of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). FUS is localised in the nucleus where it has important functions in DNA repair. However, in ALS/FTD, mutant FUS mislocalises from the nucleus to the cytoplasm where it forms inclusions, a key pathological hallmark of neurodegeneration. Mutant FUS also inhibits protein import into the nucleus, resulting in defects in nucleocytoplasmic transport. Fragmentation of the neuronal Golgi apparatus, induction of endoplasmic reticulum (ER) stress, and inhibition of ER-Golgi trafficking are also associated with mutant FUS misfolding in ALS. Protein disulphide isomerase (PDI) is an ER chaperone previously shown to be protective against misfolding associated with mutant superoxide dismutase 1 (SOD1) and TAR DNA-binding protein-43 (TDP-43) in cellular and zebrafish models. However, a protective role against mutant FUS in ALS has not been previously described. In this study, we demonstrate that PDI is protective against mutant FUS. In neuronal cell line and primary cultures, PDI restores defects in nuclear import, prevents the formation of mutant FUS inclusions, inhibits Golgi fragmentation, ER stress, ER-Golgi transport defects, and apoptosis. These findings imply that PDI is a new therapeutic target in FUS-associated ALS.


Co-expression of PDI is protective against mutant FUS induced inclusion formation in a neuronal cell line and primary cells.
Firstly, the effect of PDI on the formation of mutant FUS cytoplasmic inclusions in vitro 18 was examined, and the ALS-associated R521G mutant was chosen because it is a common and highly pathogenic familial mutation. Wild-type (WT) FUS or mutant R521G tagged with green fluorescent protein (GFP) 29 , or vector pEGFP-N1 only as a control, were co-expressed with PDI tagged with V5 24 or empty vector (pcDNA3.1), in a neuroblastoma cell line, Neuro-2a. Immunoblotting of cell lysates for V5 and FUS was performed to confirm that similar levels of protein were expressed and of the expected size in each case (Fig. 1A). Quantitation of these immunoblots by densitometry confirmed that over-expression of PDI did not significantly alter the levels of FUS-GFP when normalized to actin (Fig. 1B).
The formation of mutant FUS inclusions was examined by quantifying the percentage of transfected cells bearing large compact GFP-positive aggregates, using confocal fluorescence microscopy (Fig. 1C). Cells expressing EGFP alone formed negligible inclusions (< 1%), as expected. For all the cellular phenotypes examined in this study, expression of FUS-GFP-WT gave very similar results to controls (empty vector and untransfected cells). Hence it was not possible to examine whether PDI was protective against most previously described phenotypes associated with FUS-WT over-expression in disease models, such as inclusion formation here. In the absence of PDI, significantly more cells expressing mutant FUS-GFP-R521G with empty vector pcDNA3.1 (37%, ***p < 0.001) formed inclusions compared to those expressing FUS-GFP-WT, also as expected (10%). However, co-expression of PDI with mutant FUS resulted in significantly fewer inclusions compared to cells transfected with mutant FUS and empty vector (23%, *p < 0.05, Fig. 1D), and this proportion was now similar to FUS-GFP-WT expressing cells. Hence, these data reveal that over-expression of PDI is protective against the formation of mutant FUS inclusions.
To validate these results, mouse cortical primary cultures (containing both neurons and glia) at embryonic day 16-18 were co-transfected as above (Fig. 1E). As expected, few inclusions were formed in EGFP only cells, whereas significantly more cells expressing mutant FUS-GFP-R521G with empty vector (87%, ****p < 0.0001) Figure 1. Over-expression of PDI is protective against mutant FUS-induced inclusion formation in a neuronal cell line and primary cultures. (A) Western blotting of cell lysates, in which wild-type FUS-GFP (FUS-WT) was expressed with empty vector pcDNA3.1 and mutant FUS-GFP-R521G was expressed with either empty vector pcDNA3.1 or PDI-V5; untransfected (Un) cells are represented in the first lane. The blots were probed with anti-V5 antibody to confirm the presence of PDI, and reprobed with anti-FUS antibody and anti-actin as a loading control. Approximate molecular weight markers in kilodaltons are shown on the right. Whole blots showing the position of molecular weight marker bands are represented in supplementary Figure S1. (B) Densitometric quantitation of FUS protein levels normalized to actin from the immunoblots shown in (A) confirms similar transfection efficiency in each population and that similar amounts of each protein were expressed. (C) Immunofluorescence of EGFP in cells expressing EGFP (row 1), FUS-GFP-WT (row 2) or mutant FUS-GFP-R521G with empty vector (row 3, inclusions > 1 µm, highlighted by white arrows), co-expressed with PDI (row 4), 72 h post-transfection. Scale bar = 5 µm. (D) Quantification of cells in (C) reveals significantly fewer cells formed inclusions when PDI (*p < 0.05) was co-expressed with FUS-GFP-R521G compared to empty vector. (E) Immunofluorescence detection of EGFP-positive inclusions (> 1 µm), present in mouse primary cells co-expressing EGFP only (row 1), FUS-GFP-WT (row 2) or mutant FUS-GFP-R521G with empty vector (row 3), with PDI (row 4). Scale bar = 10 µm. (F) Quantification of cells in (E) reveals significantly fewer cells formed inclusions when PDI (**p < 0.01) was co-expressed with FUS-GFP-R521G compared to vector only. Values represent mean ± SD, n = 3. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05.  www.nature.com/scientificreports/ displayed inclusions compared to those expressing FUS-GFP-WT (20%). However, this proportion was significantly reduced (69%, **p < 0.01) when PDI was co-expressed with mutant FUS (Fig. 1F). Hence, PDI is protective against the formation of mutant FUS inclusions in primary cells, thus confirming the results obtained in Neuro-2a cells.
Over-expression of PDI is protective against mutant FUS induced nuclear import defects in a neuronal cell line. We next investigated whether PDI is protective against the inhibition of nuclear import induced by mutant FUS. For this purpose, the fluorescent reporter NES-StdTomato-NLS (referred to hereafter as 'Std-Tomato'), in which StdTomato is flanked by a NES (nuclear export sequence) and a NLS (nuclear localisation sequence), was used. This reporter is widely used to examine nucleocytoplasmic transport because it accumulates in the cytoplasm in cells when the nuclear import machinery is dysregulated 30,31 . Hence, cytoplasmic accumulation of the reporter indicates inhibition of nuclear import. For this purpose, we used a N-terminal tagged FUS construct so that the PY-NLS nuclear targeting signal was retained. Neuro-2a cells were co-transfected with Std-Tomato, hemagglutinin (HA)-tagged FUS-WT or mutant FUS-HA-R521G, and either empty vector or V5-tagged PDI ( Fig. 2A). The ratio of intensity of the nuclear-tocytoplasmic (N-to-C) fluorescence of Std-tomato was quantitated using confocal microscopy and ImageJ. In untransfected cells and cells expressing FUS-HA-WT, the N-to-C ratio was 0.77 and 0.77 respectively, revealing that Std-tomato was localised predominately in the nucleus. Hence nuclear import was efficient in these cells. As a positive control we also treated cells with mifepristone a specific inhibitor of importin α/β-mediated nuclear transport which therefore perturbs nuclear import 32 . A significant reduction in the N-to-C fluorescent intensity ratio was detected in cells treated with mifepristone (0.57, **p < 0.01) compared to FUS-HA-WT. Similarly, significant reduction in the N-to-C fluorescent intensity ratio was detected in cells expressing mutant FUS-HA-R521G with empty vector compared to FUS-HA-WT (0.60, **p < 0.01, Fig. 2B). This result indicates that significantly more Std-tomato reporter was present in the cytoplasm 33 of mutant FUS expressing cells, and hence that nuclear import was inhibited. Interestingly, however, upon expression of PDI, nuclear import was restored in mutant FUS expressing cells (*p < 0.05 vs empty vector), and the N-to-C fluorescent intensity ratio detected was 0.72, similar to cells expressing FUS-WT. These results therefore imply that PDI is protective against inhibition of nuclear import induced by mutant FUS.
Over-expression of PDI is protective against mutant FUS induced ER-Golgi transport defects in a neuronal cell line. We next investigated whether expression of PDI was protective against inhibition of ER-Golgi transport in cells expressing mutant FUS. For this purpose, we used a temperature sensitive mutant of vesicular stomatitis viral glycoprotein (VSVG ts045 ), a classical ER-Golgi transport marker 21,34 . Neuro-2a cells were co-transfected with mCherry-tagged VSVG ts045 , FUS-HA-WT or mutant FUS-HA-R521G, and either empty vector or V5-tagged PDI, before being incubated first at 40 °C to accumulate VSVG ts045 in the ER, and then at the permissive temperature (32 °C) for 30 min. Immunocytochemistry was then performed using antibodies for markers of the ER (calnexin) or Golgi (GM130), followed by fluorescent microscopy (Fig. 3A). Quantification of the localisation of VSVG ts045 in either the ER or Golgi was performed using Mander's co-efficient, in the range from 0 to 1 representing 0-100% overlapping pixels, as described previously 35 .
In untransfected cells, little VSVG ts045 (13%) was retained in the ER and most (77%) was transported to the Golgi apparatus after 30 min at the permissive temperature, demonstrating efficient ER-Golgi transport. Similarly, in cells expressing FUS-HA-WT, little VSVG ts045 (20%) was retained in the ER and most was transported to the Golgi (74%). However, in cells expressing mutant FUS-HA-R521G with empty vector, as previous, inhibition of ER-Golgi transport was detected relative to the other cell populations 21 . Significantly more VSVG ts045 was retained in the ER (43%, ***p < 0.001) and less was transported to the Golgi (39%, ****p < 0.0001) compared to FUS-HA-WT (Fig. 3B). However, when PDI was co-expressed with mutant FUS-HA-R521G, transport between the ER-Golgi was restored; only 25% (**p < 0.01) of cells displayed VSVG ts045 retained in the ER and in 60% (**p < 0.01) of cells, VSVGts045 was transported to Golgi. Therefore, these data reveal that over-expression of PDI rescues inhibition of ER-Golgi transport induced by mutant FUS.

Over-expression of PDI is protective against mutant FUS induced Golgi fragmentation in a neuronal cell line.
We next investigated whether PDI is protective against Golgi fragmentation in cells transfected as above. The morphology of the Golgi apparatus was examined by immunocytochemistry using an anti-GM130 antibody and confocal fluorescence microscopy. The Golgi is normally characterised by continuous stacked membranes, mostly localized to a compact, perinuclear ribbon. In contrast, fragmentation of the Golgi is detected by the presence of punctate structures dispersed throughout the cytoplasm, displaying multiple disconnected elements or tubular-vesicular clusters, as previously described for mutant FUS (Fig. 4A 22 ). The Golgi displayed its typical morphology, with continuous staining in a twisted ribbon-like network resembling the Golgi apparatus and little fragmentation, in untransfected cells and those expressing EGFP or FUS-GFP-WT (3%, 4% and 8% respectively). In contrast, in cells expressing mutant FUS-GFP-R521G, the Golgi was markedly different, with discrete tubulovesicular structures that were dispersed into discrete vesicles, consistent with fragmentation. The Golgi apparatus was fragmented in significantly more cells (17%, *p < 0.05) expressing mutant FUS-GFP-R521G with empty vector compared to FUS-GFP-WT (Fig. 4B). However, significantly fewer mutant FUS cells co-expressing PDI displayed fragmented Golgi compared to those expressing mutant FUS with empty vector only (9%, *p < 0.05), which was now not significantly different to controls. Hence, these data reveal that PDI inhibits Golgi-fragmentation induced by mutant FUS. www.nature.com/scientificreports/ VSVGts045-mCherry and either PDI-V5 or empty vector. Immunocytochemistry was performed using antibodies against markers of the ER (calnexin) or Golgi apparatus (GM130). VSVGts045-mCherry was trapped in the ER at 40 °C, cycloheximide was added, and the temperature was shifted to the permissive temperature (32 °C) for 30 min. At 32 °C, VSVGts045 is transported to Golgi and does not colocalise with calnexin in cells expressing FUS-HA-WT (row 1). In comparison, transport is inhibited in mutant FUS-HA-R521G with empty vector expressing cells, where less VSVGts045 was colocalised with GM130 (row 2). However, overexpression of PDI in these cells resulted in more VSVGts045 colocalised with GM130 (row 3). Scale bar = 5 μm.
(B) Quantification of the degree of co-localization of VSVGts045-mCherry with calnexin or GM130 was quantified using Mander's coefficient of cells in (A). Significantly more (***p < 0.001) VSVGts045 was retained in the ER and less in the Golgi in mutant FUS-HA-R521G with empty vector cells compared to FUS-HA-WT (****p < 0.0001). In contrast, significantly more VSVGts045 was present in the Golgi (**p < 0.01) and less in the ER (**p < 0.01) when PDI was co-expressed with mutant FUS compared to cells expressing empty vector with mutant FUS. Values represent mean ± SD, n = 3. ****p < 0.0001, ***p < 0.001, **p < 0.01.  www.nature.com/scientificreports/ locates to the nucleus during ER stress, hence nuclear immunoreactivity to CHOP can be used to detect UPR induction in cells expressing mutant FUS 20 . Neuro-2a cells were transfected as above and immunocytochemistry was performed using anti-CHOP antibodies. Immunoreactivity in the nucleus was then quantified (Fig. 5A).
Over-expression of PDI is protective against mutant FUS induced apoptosis in a neuronal cell line and primary cultures. Finally, we examined whether PDI is protective against apoptosis induced by mutant FUS-GFP-R521G. Neuro-2a cells co-expressing EGFP, FUS-GFP-WT or mutant FUS-GFP-R521G with either empty vector or PDI-V5 as above were counter-stained for Hoechst 33342. The presence of fragmented or condensed nuclei, visualized using Hoechst 33342 staining, is a reliable marker of apoptosis, as examined previously [24][25][26] (Fig. 6A). Healthy cells were observed mostly as compact and round cells. In contrast, apoptotic cells displayed condensed (under ∼ 5 μm in diameter), or fragmented (many Hoechst-positive condensed structures in one cell) nuclei. Few cells expressing EGFP alone (4%) or FUS-GFP-WT (6%) displayed apoptotic nuclei, as expected, whereas 11% (**p < 0.01) of FUS-GFP-R521G cells co-expressing empty vector contained fragmented nuclei (Fig. 6B). However, when PDI was co-expressed with mutant FUS, significantly fewer cells possessed apoptotic nuclei compared to those expressing FUS-GFP-R521G with empty vector (7%, *p < 0.05). Hence, these results imply that PDI is protective against apoptosis triggered by mutant FUS-R521G.

Discussion
Mutations in FUS are present in 5% of all fALS cases 5,6 and the histopathological hallmark of FUS-ALS is the formation of FUS-positive inclusions in motor neurons and glia 5,6,36 . Moreover, protein misfolding is a central mechanism in ALS pathogenesis. Hence mechanisms to prevent protein misfolding may have potential therapeutically. We recently reported that the PDI family of proteins protect against the misfolding of mutant SOD1 and mutant TDP-43 24 . Furthermore, novel roles for PDI proteins were recently identified in neurons in mediating motor function and neuronal connectivity 37,38 . However, PDI is aberrantly modified by S-nitrosylation in ALS patients 26 which results in loss of its functional activity 24,26 Interestingly, however, PDI colocalizes with FUS in human tissues from both sALS and fALS patients and in cells expressing mutant FUS 20 , implying that it may be protective against mutant FUS given its chaperone function. Similarly, PDI is also known to immunoprecipitate with FUS 20,21 . However, it remained unclear if PDI has a protective role against mutant FUS. In this study, we demonstrate that PDI was protective against multiple cellular defects induced by mutant FUS; inclusion formation, inhibition of both nuclear import and ER-Golgi transport, Golgi-fragmentation, ER stress and apoptosis, in a neuronal cell line and primary mixed neuron-glia cultures. Thus, this study implies that PDI may be a new therapeutic target against FUS-associated ALS.
An important finding of this study is the ability of PDI to restore nuclear import in cells expressing mutant FUS. ALS-associated FUS mutations in the NLS domain impair nuclear import 33 , resulting in the mislocalisation of FUS to the cytosol and the generation of stress granules (SGs), which are linked to inclusion formation [17][18][19] . Similarly, PDI was also protective against the formation of mutant FUS inclusions. Indeed, the nucleus is now implicated as an important site of toxicity relevant to ALS pathogenesis 14,39 . Disruption of nucleocytoplasmic transport is also induced by ALS-associated TDP-43 30 and by C9orf72 mutations [39][40][41] . Together these studies highlight the importance of nucleocytoplasmic defects in ALS.
Although conventionally regarded as being ER-localised, it is now well established that PDI is also found in other cellular locations, including the nucleus and cytoplasm [42][43][44] . The protective effects of PDI against the inhibition of nucleocytoplasmic shuttling and FUS misfolding therefore imply that this is mediated from the cytoplasm. Interestingly, PDI redistributes away from the ER in punctate vesicles in ALS models, where it displays higher enzymatic activity 45,46 . Furthermore PDI inhibits ALS progression in mice in vivo when expressed in this location 45 . Moreover, a recent study demonstrated that PDI translocates away from the ER into the cytoplasm following the induction of ER stress 47 . Hence, in cells expressing mutant FUS, PDI may re-locate from the ER to the cytoplasm, from where it is protective. However, given that FUS has been detected within the ER 21 , a protective role for ER-localised PDI is also possible. Hence, further characterisation of the cellular location of PDI in cells expressing mutant FUS is warranted in future studies.
Golgi fragmentation is a well described phenomenon in ALS 48 that is known to occur when protein export from the ER is altered, or when vesicular trafficking from either to or from the Golgi is perturbed 49 . The formation of Golgi stacks requires the continuous recycling of proteins from or to the ER, therefore bidirectional vesicular transport with the ER is essential for Golgi organization 50 . Misfolded and incompletely folded proteins are excluded from transport vesicles leaving ER exit sites during the first phases of ER-Golgi transport [49][50][51][52][53] . Inhibition of ER-Golgi traffic could therefore increase the load of proteins within the ER, inducing ER stress 21  www.nature.com/scientificreports/ www.nature.com/scientificreports/ ER stress 20 . However, mutant FUS within the ER may also induce ER stress 21 . Hence PDI may be protective by both restoring ER-Golgi transport and reducing the load of misfolded proteins within the ER. We recently demonstrated that the oxidoreductase activity of PDI is necessary for its protective effects against mutant SOD1 and mutant TDP-43 24 . It remains unclear whether the chaperone or oxidoreductase activity of PDI is protective against mutant FUS. However, whilst aberrant disulphide bond formation between specific cysteine residues in both mutant SOD1 and mutant TDP-43 are implicated in protein misfolding 54,55 , no disulphide bonds have been identified as yet in FUS (native or otherwise), despite FUS containing 4 cysteines that are part of the zinc finger involved in RNA-binding 56 . FUS contains a prion-like domain and its ability to aggregate is determined by the presence of this domain 57 . Interestingly, PDI has been shown to interact with prion-like misfolded proteins, implying that PDI may interact with FUS by its prion domain 58 . Together, these studies imply that PDI is protective against mutant FUS by its chaperone, rather than its oxidoreductase activity, although further studies are required to examine this possibility.
PDI is S-nitrosylated in patients with ALS as well as in other neurodegenerative disorders 26 . This aberrant post-translational modification could therefore mask the normal protective function of PDI during disease. S-nitrosylation of PDI may compromise its protective functions, further aggravating protein misfolding and ER stress in these conditions. Protein misfolding could additionally further deplete PDI by sequestration, thus reducing the effective cellular pool of chaperones. This vicious cycle may trigger the onset of neurodegeneration due to the collapse of multiple cellular processes, thus dysregulating proteostasis. FUS is also present in the pathological inclusions of patients with FTD without TDP-43 or tau pathology, accounting for about 10% of FTLD cases. It is therefore tempting to speculate that PDI may also have a protective role against FTLD.
In summary, in this study it was demonstrated that PDI, a unique chaperone with diverse activities, is protective against important pathogenic mechanisms associated with cellular function and apoptosis in FUS-associated ALS. The findings of this study therefore provide further evidence that PDI may be an effective potential therapeutic target against multiple proteins and cellular pathologies in ALS. Immunoblotting. Cell lysates were collected in cold TN buffer (50 mM Tris-HCl pH 7.5 and 150 mM NaCl, pH 7.6) with 0.1% [w/v] sodium dodecyl sulfate (SDS), 1% Triton-X100, 1% protease inhibitor cocktail (Roche) and 1% phosphatase inhibitor (Roche) and phenylmethanesulfonyl fluoride [PMSF, Sigma #P7626-250 MG], then incubated on ice for 15 min and stored at − 20 °C overnight. Samples were centrifuged at 100,000g at 4 °C for 30 min to obtain the SDS-soluble fraction. Protein concentrations of cell lysates were determined using the BCA protein assay (Thermo Scientific) by comparison with BSA standards. Protein samples (10-20 µg) were electrophoresed through 8.5% SDS-polyacrylamide gels and transferred to nitrocellulose membranes. Membranes were blocked with 5% skim-milk in Tris-buffered saline (pH 8.0) for 30 min, then incubated with the appropriate primary antibodies at 4 °C for 16 h; anti-FUS (1:500, Abcam ab23439), anti-V5 (1:1000, Invitrogen P/N460705), or anti-actin (1:1000, Cytoskeleton #AAN01). Membranes were incubated for 1 h at room temperature with secondary antibodies (1:2000, HRP-conjugated goat anti-rabbit, or goat anti-mouse, Merck Milipore, AP130, AP132), and detected using ECL reagent (Bio-Rad). Precision Plus Protein™ Dual Colour Standard molecular weight markers were used (Bio-Rad). Quantitation of blots was performed by densitometry using ImageJ (NIH).

Methods
Immunofluorescence and microscopy. Neuro-2a cells were fixed in 4% paraformaldehyde, permeabilized with 0.1% Triton-X100 in phosphate buffered saline (PBS), blocked with 3% BSA in PBS, followed by incubation with mouse anti-CHOP (1:50, Santa Cruz Sc-7351), mouse anti-V5 (1:250, Invitrogen P/N460705), or mouse anti-GM130 (1:250, BD Transduction Lab 610823) antibodies in PBS at 4 °C overnight. The secondary antibody AlexaFluor 568-conjugated rabbit anti-mouse IgG (1:250, Invitrogen A21203) was added for 1 h and incubated in the dark at room temperature. After washing with PBS, staining of nuclei was performed using Hoechst stain 33342 (Invitrogen, nuclei). FITC (for GFP fluorescence) and TRITC (for red fluorescence) filters were used for viewing the cells and images were taken using a Zeiss Axioimager microscope or LSM 880 Zeiss confocal microscope in a single focal plane. In dual-channel imaging, photomultiplier sensitivities and offsets were set to a level at which bleed-through effects from one channel to another were negligible. Nuclear and cytoplasmic fluorescent intensity of either the Std-Tomato reporter or FUS and PDI was calculated with the same exposition time between groups, using ImageJ.  Fig. 2, the pharmacological agent Mifepristone (Cayman Chemical #10006317) was used as a positive control. The corrected total cell fluorescence (CTCF) of the whole cell and the nucleus was determined and then the ratio was obtained by dividing the CTCF of Std-tomato expression in the nucleus by that of the whole cell using Image J software 60 . In brief, 20 cells were scored for each experiment and cells of interest were selected using drawing/selection tools (freeform) from Image J. From the ' Analyze' menu, "Set measurements" was selected by marking AREA, INTEGRATED DENSITY and MEAN GRAY VALUE. The cell fluorescence was measured in the nucleus and then the whole cell by selecting "Measure" from the ' Analyze' menu. A region next to the cell lacking fluorescence was measured as background. The formula used was as follows; (CTCF = Integrated Density -(Area of selected cell × Mean fluorescence of background readings).
Cortical cell culture and transfection. Primary cultures were harvested from the cortex of C57BL/6 mouse embryos at embryonic day 16-18. The procedure for culture of primary cells was as described previously 25 . Briefly, cortical tissue was dissected under sterile conditions in Hanks' Balanced Salt solution (HBSS, Gibco) and digested in 10 units/ml papain (Sigma) in 2 mg/ml l-cysteine (Sigma) and 0.5 mM EDTA, pH 8 (Sigma) in DMEM, for 15 min at 37 °C. Cells were subsequently triturated using pipette tips so they were dissociated, then resuspended in plating medium (DMEM, 10% FBS, 100 µg/ml penicillin-streptomycin) and seeded for 1 h on 15 mm glass coverslips, previously coated with 0.1 mg/ml poly-d-lysine overnight. Cells were then incubated in neuronal medium [Neurobasal medium supplemented with 2% B27, supplement (Gibco), 1% Glutamax (Gibco) and 100 µg/ml penicillin-streptomycin] at 37 °C and 5% CO 2 .

Primary cell immunocytochemistry. Cortical primary cells transfected with FUS constructs for 48 h
were washed in PBS and fixed in 4% paraformaldehyde in PBS for 10 min. After three washes in PBS, cells were permeabilized in 0.1% Triton-X100 in PBS for 5 min and the non-specific background staining was blocked using 3% (wt/v) BSA in PBS for 45 min at room temperature with gentle rocking. Cells were then incubated overnight at 4 °C with anti-V5 antibody (1:250, Invitrogen P/N460705) in 3% (wt/v) BSA in PBS. After rinsing, cells were incubated for 1 h at room temperature with anti-mouse secondary antibody coupled to AlexaFluor 594 (Invitrogen A21203), diluted 1:250 in PBS. Cells were then washed as above and treated with 0.5 µg/ml Hoechst 33342 reagent (Sigma). After washes in PBS, coverslips were mounted onto slides in fluorescent mounting medium (Dako). At least 20 cells were then examined and photographed on a Zeiss LSM 880 inverted confocal laser-scanning microscope, equipped with a LSM-TPMT camera (Zeiss). VSVG ts045 ER-Golgi transport assay. Neuro-2a cells were transiently transfected with the appropriate construct and the VSVG ts045 -mCherry plasmid. Cells were incubated at 40 °C under 5% CO 2 for 72 h with FUS. The cells were then treated with cDMEM containing cycloheximide (20 μg/ml) and incubated at 32 °C for 30 min. Staining was performed with primary antibodies mouse, anti-GM130 (1:250) (Golgi marker) (BD Transduction 610823) or rabbit anti-calnexin (1:250) (ER marker) (Abcam ab22595) overnight. Secondary antibodies AlexaFluor 647 goat anti-mouse (1:200 Life Technology A21235) and goat anti-rabbit (1:200 Thermo Fisher Sci A27040) were used. Mander's coefficient was used to determine the degree of co-localisation between VSVG ts045 -mCherry and ER or Golgi marker. Mander's coefficient was calculated for 20 cells by JACoP (Colocalisation Plugin) 61 in ImageJ (NIH).

Statistical analysis.
The experiments were performed a minimum three times on separate days with one blinded experiment, unless specified separately. The data are represented as mean ± SD. One-way Analysis of Variance (1-way ANOVA) followed by Tukey post hoc test was used to determine between treatment differences (GraphPad Prism, San Diego, CA, USA). A p-value of 0.05 or equal was considered significant, where ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. 100 cells were scored for each population unless otherwise stated.