Introduction

Frontotemporal dementia (FTD) is one of the most common forms of presenile dementia and involves the progressive degeneration of the frontal and temporal lobes of the brain. Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder that affects the upper and lower motor neurons leading to muscle weakness and paralysis1. Increasing genetic evidence, neuropathological and clinical observations have identified a substantial overlap between these two disorders2. Several genes have been identified where mutations can cause either FTD or ALS (FTD-ALS genes) including C9orf72, VCP, OPTN, SQSTM1, TBK13 and, most recently, CYLD4. In addition, FTD and ALS share neuropathological similarities: ~ 95% of ALS and ~ 50% of FTD patients show cytoplasmic inclusions of TAR DNA-binding protein 43 (TDP-43) in the brain5,6, while ~ 5% of ALS and ~ 10% of FTD patients show cytoplasmic inclusions of fused in sarcoma protein (FUS)6,7,8. In healthy cells, TDP-43 and FUS are largely expressed within the nucleus. In FTD and ALS, affected neurons can display a redistribution of TDP-43 or FUS to the cytoplasm, as well as insoluble TDP-43 or FUS aggregates6,9,10. The importance of TDP-43 and FUS is further demonstrated by the fact that mutation of their encoding genes TARDBP and FUS is sufficient to cause ALS7,11,12 or, rarely, FTD13.

Identified genetic mutations cause 20% of FTD and 5-10% of ALS cases14,15; however, for the remainder of patients the origin of the disease is uncertain. Despite TDP-43 being a prominent feature in the majority of FTD and ALS cases, including all those carrying FTD-ALS gene mutations13, this fact has not been applied to the screening of FTD and ALS candidate gene variants on a large scale. In previous studies, TDP-43 and FUS mislocalisation has largely been assessed by subcellular fractionation, immunohistochemistry and manual analysis of confocal microscopy7,16,17,18,19,20,21,22,23,24 or high-content imaging25,26,27. Whilst studies utilising these techniques have been informative, these assays are labour intensive and low-throughput or, expensive and in some cases produce qualitative data. In addition, some microscopy-based techniques could be subject to bias due to manual selection of the cells to be analysed.

Here we report the development of a rapid, cheap and unbiased technique for the in vitro study of TDP-43 and FUS cytoplasmic mislocalisation. TDP-43 or FUS staining in nuclear and cytoplasmic subcellular compartments was measured in thousands of cells from confocal microscope images, using the freely available analysis software CellProfiler28. We validated this technique using known genetic drivers of TDP-43 or FUS mislocalisation. We were able to detect increased cytoplasmic localisation of exogenously expressed FUS and TARDBP mutations relative to wild-type (WT) sequence. We also observed mislocalisation of endogenous FUS upon treatment with the apoptosis inducer staurosporine. Importantly, we could also detect more subtle changes in the cellular distribution of endogenous TDP-43, arising from mutation in a secondary gene (CYLD p.M719V). This validated methodology can now be utilised for rapidly prioritising newly discovered FTD and ALS gene variants by quantifying their effect on TDP-43 and/or FUS localisation.

Results

Detection of FUS cytoplasmic mislocalisation with exogenous expression of FUS mutations

In order to validate our method for the unbiased quantification of FUS and TDP-43 mislocalisation we initially utilised known pathogenic FUS mutations that drastically alter FUS subcellular localisation7,29. Expression of green fluorescent protein (GFP)-tagged FUS protein with missense mutation FUSR521C or truncation mutant FUSR495X in the neuronal-like neuroblastoma cell line, SH-SY5Y, led to a significant increase in the fluorescence intensity of cytoplasmic FUS (Fig. 1e–k) when compared to FUSWT (Fig. 1a–c,h). This increase in cytoplasmic FUS was accompanied by a corresponding decrease in the fluorescence intensity of nuclear FUS (Fig. 1d). Together, these resulted in a 9.5- and 12.5-fold increase in the FUS cytoplasmic/nuclear ratio for cells expressing FUSR521C (0.97 ± 0.04; p < 0.0001) and FUSR495X (1.28 ± 0.04; p < 0.0001), respectively, relative to FUSWT (0.10 ± 0.00; Fig. 1l). Similar results were found when overexpressing the FUSR521C (1.04 ± 0.05; p < 0.0001) and FUSR495X (1.81 ± 0.10; p < 0.0001) mutants in human embryonic kidney (HEK293) cells (Supplementary Fig. S1), although the decrease in nuclear fluorescence intensity of FUS was not as prominent (Supplementary Fig. S1).

Figure 1
figure 1

Detection of FUS cytoplasmic mislocalisation with exogenous expression of FUS mutations. Representative images of SH-SY5Y cells overexpressing GFP-tagged (a–c) FUSWT, (e–g) FUSR521C or (i–k) FUSR495X. Nuclei were visualised with DAPI (blue). (d) Quantification of the fluorescence intensity of the nucleus shows a significant reduction in nuclear FUS expression in cells expressing FUSR521C or FUSR495X, when compared to FUSWT. (h) Quantification of the fluorescence intensity of cytoplasmic FUS shows significantly higher cytoplasmic FUS expression in FUSR521C and FUSR495X, when compared to FUSWT. (l) Quantification of the cytoplasmic/nuclear ratio of exogenous FUS shows a marked increase in FUSR521C- and FUSR495X-expressing cells when compared to FUSWT. Scale bars = 5 μm. Data is represented as mean ± SEM. a.u. = arbitary units. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Full size image

Detection of endogenous FUS cytoplasmic mislocalisation following staurosporine treatment

To test our ability to quantify the subcellular distribution of endogenous FUS we utilised the apoptosis inducer staurosporine, which has been previously reported to cause cytoplasmic mislocalisation of FUS16. Application of increasing doses of staurosporine showed an increase in the fluorescence intensity of cytoplasmic FUS in cells treated with the two highest doses of staurosporine (1 and 10 μM; Fig. 2a–d,f). A corresponding decrease in the nuclear intensity of FUS was also seen in SH-SY5Y cells (Fig. 2e). There was a significant increase of cytoplasmic/nuclear FUS ratio for all treatment groups (0.1 μM: 0.13 ± 0.01; p = 0.0405; 1 μM: 0.33 ± 0.05; p = 0.0152; 10 μM: 1.53 ± 0.12; p = 0.0002) relative to vehicle-treated cells (0.09 ± 0.00; Fig. 2g). HEK293 cells treated with staurosporine also exhibited a dose–response effect on FUS subcellular localisation. In general, HEK293 cells were more tolerant of the staurosporine treatment (Supplementary Fig. S2) as shown by the lower cytoplasmic/nuclear ratios observed when compared to the SH-SY5Y cells (Fig. 2g and Supplementary Fig. S2). Treatment with 10 μM staurosporine caused the majority of FUS to be present in the cytoplasm rather than the nucleus in SH-SY5Y cells (1.53 ± 0.12; p = 0.0002; Fig. 2d,g) whilst the same treatment in the HEK293 cells caused only a small proportion of FUS protein to be mislocalised (0.28 ± 0.01; p < 0.0001; Supplementary Fig. S2). Despite these differences in effect size, we could detect a significant effect of staurosporine treatment on mislocalisation of endogenous FUS in both cell lines.

Figure 2
figure 2

Detection of endogenous FUS cytoplasmic mislocalisation following staurosporine treatment. (a–d) Representative fluorescence images of SH-SY5Y cells treated with increasing concentrations of staurosporine. FUS was detected by immunofluorescent staining (red) and nuclei were visualised with DAPI (blue). (e) Quantification of nuclear and (f) cytoplasmic FUS fluorescence intensity show significant changes in 1 μM- and 10 μM-treated cells. (g) Quantification of the FUS cytoplasmic/nuclear ratio shows an increase in all treatment groups when compared to those treated with DMSO alone. Scale bars = 20 μm. Data is represented as mean ± SEM. a.u. arbitrary units. *p < 0.05; **p < 0.01; ***p < 0.001.

Full size image

Detection of TDP-43 cytoplasmic mislocalisation with exogenous expression of TARDBP mutations

We took the same initial approach used for FUS to validate this method for the quantification of TDP-43 cytoplasmic mislocalisation: i.e. exogenous expression of known pathogenic mutations in TARDBP. We examined p.A315T, one of the earliest detected and thus most extensively examined mutations30,31,32, and two of the mutations most commonly observed in ALS patients: p.M337V and p.A382T11,31,32,33.

In SH-SY5Y cells, a significant increase in the cytoplasmic/nuclear ratio of GFP-tagged TDP-43A315T (0.87 ± 0.01; p = 0.0016; Fig. 3b,f,i–k) and TDP-43A382T (0.86 ± 0.02; p = 0.0032; Fig. 3d,h–k) was observed when compared to TDP-43WT (0.75 ± 0.01; Fig. 3a,e,i–k). In contrast, there was a significant decrease in the cytoplasmic/nuclear ratio of GFP-tagged TDP-43M337V (0.63 ± 0.01; p = 0.0028; Fig. 3c,g,i–k) relative to TDP-43WT. Similar to the experiments with FUS, expression of TDP-43 mutations had a lesser effect in HEK293 cells (Supplementary Fig. S3). Once again, there was a significant increase in the cytoplasmic/nuclear ratio of TDP-43A315T (0.57 ± 0.02; p = 0.0021; Supplementary Fig. S3) and TDP-43A382T cells (0.66 ± 0.03; p = 0.0026; Supplementary Fig. S3) when compared to TDP-43WT (0.50 ± 0.02; Supplementary Fig. S3). However, no significant effect was observed in TDP-43M337V cells (0.52 ± 0.01; Supplementary Fig. S3).

Figure 3
figure 3

Detection of TDP-43 cytoplasmic mislocalisation with exogenous expression of TARDBP mutations. Representative images of SH-SY5Y cells overexpressing GFP-tagged (a,e) TDP-43WT, (b,f) TDP-43A315T, (c,g) TDP-43M337V or (d,h) TDP-43A382T. Nuclei were visualised with DAPI (blue). (i) Quantification of the fluorescence intensity of the nucleus and (j) the cytoplasm shows no significant difference in TDP-43 expression. (k) Quantification of the cytoplasmic/nuclear ratio of exogenous TDP-43 shows a significant increase in TDP-43A315T and TDP-43A382T, and a decrease in TDP-43M337V, when compared to cells expressing TDP-43WT. Scale bars = 5 μm. Data is represented as mean ± SEM. a.u. arbitrary units. **p < 0.01.

Full size image

Detection of endogenous TDP-43 cytoplasmic mislocalisation in cells expressing a causative FTD/ALS mutation

To demonstrate the potential of this methodology for identifying potentially pathogenic variants in other FTD/ALS genes, we expressed a known familial FTD/ALS-causative mutation in the CYLD gene, p.M719V4, and observed the changes in endogenous TDP-43 localisation. We previously determined that overexpression of CYLDM719V increases the proportion of cytoplasmic TDP-43-positive mouse primary cortical neurons by ~ 20% when compared to CYLDWT4. When compared to GFP-only vector control (0.079 ± 0.001; Fig. 4a–d), expression of CYLDWT-GFP (Fig. 4e–h) in SH-SY5Y cells led to a significant 1.2-fold increase in the TDP-43 cytoplasmic/nuclear ratio (0.095 ± 0.001; p = 0.0002; Fig. 4q–s). In turn, expression of CYLDM719V-GFP (Fig. 4m–p) caused a further 1.3-fold increase in cytoplasmic/nuclear TDP-43 relative to CYLDWT-GFP (0.121 ± 0.002; p < 0.0001; Fig. 4q–s). The CYLDM719V mutation had the same effect in HEK293 cells, causing a 1.3-fold increase in the cytoplasmic/nuclear ratio relative to CYLDWT (0.145 ± 0.007; p = 0.0115; Supplementary Fig. S4). We also transfected cells with the CYLDD681G mutation (Fig. 4i–l and Supplementary Fig. S4) which is catalytically inactive and known to cause CYLD cutaneous syndrome, a skin tumour disorder34,35. Our data corroborated previous results4 that the CYLDD681G mutation had no effect on TDP-43 localisation when compared to the GFP-only vector control in either cell type (Fig. 4s and Supplementary Fig. S4).

Figure 4
figure 4

Detection of endogenous TDP-43 cytoplasmic mislocalisation in cells expressing CYLD mutations. Representative images of SH-SY5Y cells overexpressing (a–d) GFP or GFP-tagged (e–h) CYLDWT, (i–l) CYLDD681G or (m–p) CYLDM719V. TDP-43 was detected by immunofluorescent staining (yellow) and nuclei were visualised with DAPI (blue). (q) Quantification of the fluorescence intensity of the nucleus and (r) cytoplasm shows no difference in TDP-43 expression.  (s) Quantification of the cytoplasmic/nuclear ratio of exogenous TDP-43 shows a marked increase in CYLDWT, when compared to cells expressing GFP. The cytoplasmic/nuclear ratio of TDP-43 in CYLDM719V-expressing cells was significantly higher than CYLDWT. Scale bars = 5 μm. Data is represented as mean ± SEM. a.u. arbitrary units. ***p < 0.001; ****p < 0.0001.

Full size image

Discussion

In this study we have validated a novel methodology for the unbiased quantification of cytoplasmic TDP-43 and FUS in human cell lines under three different experimental paradigms: determining localisation of exogenously expressed mutant TDP-43 or FUS; detecting the effect of a chemical modulator on endogenous FUS localisation; and detecting the effect of a secondary gene on endogenous TDP-43 localisation. This rapid, cheap, quantitative assay can now be utilised for rapidly assessing drug treatments and the potential pathogenicity of newly discovered FTD and ALS gene variants by quantifying their effect on TDP-43 and/or FUS localisation.

Previous studies examining TDP-43 and FUS have quantified cytoplasmic mislocalisation in different ways. Many studies have reported the proportion of the cell population that display cytoplasmic TDP-43 or FUS expression12,18,19,21,24. This may present a problem with reproducibility, since fluorescence detection thresholds for considering a cell as ‘positive' for cytoplasmic protein likely differ between labs and will vary according to the microscopy equipment used. We selected the ratio of cytoplasmic to nuclear expression as our primary measure, having observed reduced variability between experimental replicates in comparison to individual nuclear and cytoplasmic measurements. Some studies have reported similar measurements using techniques such as high-content screening confocal microscopy to study TDP-43 or FUS mislocalisation under different conditions25,26. Whilst high-content screening can allow for a more high-throughput study design than our methodology, the extensive infrastructure required is expensive, not readily available and can require substantial optimisation. We note that while we selected the CellProfiler program for post-imaging analysis, other software platforms are capable of the same type of analysis e.g. Imaris (RRID:SCR_007370), ImageJ36 and Huygens (RRID:SCR_014237). We chose CellProfiler based on its ease of use, the ability to perform batch analysis of images, cost and accessibility.

Another general observation in our experiments was that SH-SY5Y cells often displayed a greater degree of cytoplasmic mislocalisation than that in HEK293 cells under the same conditions. This may be due to SH-SY5Y cells being a neuronal cell line and thus more disease relevant, or due to the longer transfection time required to achieve optimal transfection efficiency in SH-SY5Y cells (48 h, versus 24 h in HEK293), causing additional stress on the cells. It should also be noted that transfection efficiency also dictated the number of images required to reach the desired cell number per replicate (300 cells). SH-SY5Y cells, which had a lower transfection efficiency, required more images to achieve the same number of cells for quantification.

The cytoplasmic mislocalisation and/or aggregation of FUS in neurons and glia of post-mortem tissue from patients with ALS caused by FUS mutations has been widely established7,12. This pathology has been recapitulated by expression of FUS mutations in mammalian cells7,12,29. In this study, exogenous expression of the severe truncation mutation FUSR495X caused the majority of FUS to be mislocalised to the cytoplasm as previously demonstrated by Bosco et al.29 using live-cell confocal microscopy and 3D images (11–28 cells). Our quick and simple methodology was able to duplicate these results using 2D images and a larger number of cells (1800 cells total). While we recognise the merits of quantifying the entire cell volume, measuring such a small number of selected cells is not representative of the population and may not be able to detect more subtle or variable effects in protein localisation.

Expression of the FUSR521C missense mutation and treatment with staurosporine (1–10 μM) also reproduced previous qualitative microscopy and subcellular fractionation results7,16,37 showing a significant increase in the FUS cytoplasmic/nuclear ratio. Confirmation of these results demonstrates the effectiveness of this new methodology in providing quantitative data where previously qualitative or labour-intensive experiments were required. Our ability to quantify FUS cytoplasmic mislocalisation following staurosporine treatment also demonstrates the possibility of this technique for use in testing the ability of novel FTD and ALS drug treatments to modulate FUS or TDP-43 mislocalisation.

The majority (~ 95%) of ALS and ~ 50% of FTD cases are characterised by the abnormal accumulation of TDP-43 in the cytoplasm of neurons and glia, even though in most cases mutations in TARDBP are absent6. Mutations in TARDBP have been associated with both FTD and ALS and lead to TDP-43 cytoplasmic mislocalisation11,13. Unlike FUS, the expression of TARDBP mutations in primary cells and cell lines does not reproduce the dramatic cytoplasmic mislocalisation seen in post-mortem tissue38,39,40. In addition, overexpression of exogenous TDP-43 in vitro is sufficient to induce mislocalisation even for WT sequence19,23. Any differences between WT and mutant TDP-43 for this characteristic are therefore harder to discern and represent a significant limitation for all TDP-43 exogenous expression assays, including the one described here, and may be responsible for differing results being reported for the same mutations. Our technique was able to detect a significant increase in the TDP-43 cytoplasmic/nuclear ratio in TDP-43A315T and TDP-43A382T cells when compared to TDP-43WT. This confirms previous results for TDP-43A315T18,19 and TDP-43A382T41,42. The TDP-43M337V variant, which displayed a significant decrease in cytoplasmic TDP-43 in SH-SY5Y cells, has previously been shown to increase cytoplasmic TDP-43 in some19,21,24, but not all studies22. We note that TDP-43M337V has previously been shown to aggregate into nuclear puncta in SH-SY5Y cells21, which we observed in our experiments. As these puncta fluoresce at a higher intensity than diffuse TDP-43 expression, this may lead to an overall decrease in cytoplasmic/nuclear fluorescence ratio relative to TDP-43WT. Thus, in some cases quantification of nuclear and cytoplasmic puncta may be a useful adjunct to assess the pathogenicity of a given mutation. However, the disease relevance of nuclear inclusions in in vitro studies of TDP-43M337V is unclear: although both intranuclear and cytoplasmic inclusions were reported in a transgenic TDP-43M337V mouse model43 and it is reminiscent of intranuclear inclusions observed in post-mortem tissue from sporadic FTD patients9, such inclusions are rare in human disease. In the only neuropathological report to our knowledge that concerns a patient with a TARDBP M337V mutation, no mention was made of intranuclear inclusions44. We also note that mislocalisation of TDP-43 is not the only mechanism by which TARDBP mutations lead to disease: for example, changes in protein stability or interaction partners have also been described31. However, we envisage that the mislocalisation assay as implemented here would be a useful tool in a battery of assays to screen novel TARDBP variants.

Lastly, we were able to recapitulate the effect of WT CYLD, the FTD-ALS-causative mutation p.M719V and the catalytically inactive mutation p.D681G on subcellular localisation of endogenous TDP-43, which we previously observed in mouse cortical neurons4. In comparison to the other experiments used to validate this quantification method, the degree of change in the endogenous TDP-43 cytoplasmic/nuclear ratio between the empty vector, the CYLDWT and the CYLD mutations was very small: e.g., absolute increase in ratio of ~ 0.03–0.04 between CYLDWT and CYLDM719V (Fig. 4s and Fig. S4s). Our ability to detect significant differences for such a subtle change demonstrates the sensitivity of this technique for detecting changes in TDP-43 and FUS localisation in vitro. We note that this assay relies on overexpression of the secondary gene and it would therefore be judicious to perform follow-up studies to examine whether the effect is still observed when the gene is expressed at physiological levels (e.g. by CRISPR/Cas9-editing to introduce candidate gene mutations) and/or in more biologically relevant cells (e.g. induced pluripotent stem cell-derived neurons). However, the utilisation of our quantification method in this format demonstrates what we believe to be its primary use, prioritising potentially pathogenic newly identified FTD- or ALS-associated variants in genes other than FUS and TARDBP for more in-depth investigation.

An important parameter to consider when visualising mislocalisation of TDP-43 and FUS in exogenous expression assays is the positioning of the fluorescent tag in the TDP-43 or FUS fusion proteins. In this study we utilised a N-terminal GFP tag for FUS constructs, due to the location of its nuclear localisation signal (NLS) at the C-terminus (aa497-526)45. Conversely, we used a C-terminal GFP tag for TDP-43, as its NLS is located near the N-terminus of the protein (aa82-98)46, and TDP-43 forms homodimers via its N-terminus47,48. By tagging the protein at the opposite end to its NLS or dimer-binding region we hoped to minimise any effect on the normal localisation pattern of TDP-43 and FUS. Additionally, aggregates of TDP-43 contain C-terminal fragments49, which may have been undetectable had we used a N-terminal GFP tag. We note that the cytoplasmic/nuclear ratio of exogenous GFP-FUSWT (0.10 ± 0.00; Fig. 1l) was almost identical to that of endogenous FUS in vehicle-treated cells (0.09 ± 0.00; Fig. 2g), implying that this tag does not affect FUS localisation. For TDP-43, the cytoplasmic/nuclear ratio of the GFP-tagged protein (0.75 ± 0.01; Fig. 3k) was much higher than that of endogenous TDP-43 (0.079 ± 0.001; Fig. 4s). This is seen for exogenous expression of TDP-43 even with small peptide tags19,23. Nevertheless, we acknowledge that the presence of the larger GFP tag may further affect TDP-43 localisation and represents a limitation of this assay.

A parameter of equivalent importance for detection of endogenous FUS and TDP-43 mislocalisation is selection of an appropriate detection antibody. This is particularly important for TDP-43, where C-terminal TDP-43 fragments observed in FTD and ALS patients9 contribute to its mislocalisation and aggregation49. Antibodies raised against the extreme N-terminus of TDP-43 would therefore not be suitable for a mislocalisation assay, as they would not detect these cleavage products. In contrast, the anti-TDP-43 antibody we used (Proteintech; #10782-2-AP) recognises all major truncated TDP-43 forms9.

In summary, this study demonstrates a simple methodology for unbiased quantification of TDP-43 and FUS cytoplasmic mislocalisation in vitro. This technique can easily be added to studies utilising qualitative data to strengthen clearly noticeable phenotypes in an unbiased manner, as well as to highlight subtle changes that may have not been previously identified. Utilisation of this fast and cost-effective technique to aid in assessing pathogenicity of gene variants observed in patients with FTD or ALS will help to prioritise these variants for more intensive research efforts into how they cause disease.

Methods

DNA constructs

CYLD, TARDBP and FUS mutations were introduced into the pCMV6-CYLD, pCMV6-TARDBP and pCMV6-FUS constructs (Origene) by site-directed mutagenesis using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent). Mutated CYLD and TARDBP cDNA sequences were then subcloned, using AsiSI and MluI restriction sites, into the pCMV6-AC-GFP vector (Origene) for expression of C-terminal GFP-tagged CYLD and TDP-43 proteins, respectively. Mutated FUS cDNA sequences were subcloned similarly into the pCMV6-AN-GFP vector (Origene) for expression of N-terminal GFP-tagged FUS protein, to avoid interference of the GFP tag with the C-terminal NLS of the FUS protein45. All clones were verified by restriction digestion and sequence analysis.

Cell culture

HEK293 and SH-SY5Y were purchased from ATCC and authenticated by STR analysis prior to use in this study. HEK293 cells were maintained in Eagle’s Minimum Essential Medium (EMEM; Gibco) and SH-SY5Y cells in Dulbecco’s Modified Eagle’s Medium/Nutrient Mixture F-12 (DMEM/F12; 1:1 mixture; Gibco) each containing 10% heat-inactivated fetal calf serum (Sigma-Aldrich). For immunocytochemistry, 8-well chamber slides (Ibidi) were coated with 0.005% poly-L-lysine (Sigma-Aldrich) for 1 h and then washed twice with Dulbecco’s phosphate-buffered saline (DPBS; Gibco). SH-SY5Y cells were seeded at 5–7.5 × 104 cells/well and HEK293 cells were seeded at 8 × 104 cells/well. After 24 h, cells were transfected with GFP constructs (250 ng/well) for 24 (HEK293) or 48 (SH-SY5Y) hours using Lipofectamine 3000 (0.75 μL/well; Invitrogen) as per the manufacturer’s protocol. For the staurosporine experiment, cells were incubated for 48 h after seeding, then treated for 5 h with 0.1, 1 or 10 μM staurosporine, to induce apoptosis. Staurosporine dilutions were prepared from a 1 mM staurosporine stock solution dissolved in dimethyl sulfoxide (DMSO; Sigma-Aldrich). All vehicle and staurosporine treatment groups were adjusted to a final concentration of 1% DMSO.

Immunocytochemistry

Transfected or drug-treated cells were fixed in 4% paraformaldehyde (PFA) in PBS for 20 min in the dark. For visualisation of exogenous TDP-43 and FUS, cells were then washed twice with DPBS and mounted using DAKO fluorescent mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI; Agilent). For visualisation of endogenous TDP-43 and FUS, non-specific background staining was blocked after cell fixation using 1% bovine serum albumin (BSA) in PBS for 30 min. Cells were then incubated overnight at 4 °C with rabbit anti-TDP-43 (1:500; Proteintech #10782-2-AP) or rabbit anti-FUS (1:500, Proteintech #11570-1-AP) antibodies. The next day, cells underwent 2 × 5-min washes with DPBS, incubation with rabbit Alexa Fluor-647 secondary antibody (1:500; Invitrogen #A-21244) for 1 h, followed by 2 × 5-min washes and mounting as above. Cells were imaged for fluorescence intensity quantification with a 40 × objective on a Nikon A1R confocal microscope. Images of representative cells were obtained with a 100 × objective for figure preparation. Laser intensity, gain and offset settings were kept constant and at least five random fields of view were imaged for each experimental replicate.

Fluorescence intensity quantification

Quantification of changes in the cellular localisation of TDP-43 or FUS between the nucleus and the cytoplasm was done by utilising CellProfiler 3.1.9 software (Broad Institute, Cambridge, MA)28. To prepare images for analysis, ND2 microscope files were converted to TIFFs, blinded and split into individual wavelength channel images using ImageJ (National Institutes of Health, Bethesda MD)36. In CellProfiler, the DAPI channel was analysed first using the Identify Primary Objects module and a global, three-class Otsu thresholding method. This allowed for identification and separation of each nucleus within an image. Strict filtering for cell diameter (30–60 pixels) eliminated any irregular or overlapped nuclei that were present. TDP-43- or FUS-positive cells were also identified using the Identify Primary Objects module. Cell diameter filtering for this step had a much larger range and was dependent on experimental conditions and cell type (20–150 pixels). The Relate and Filter modules were used to link filtered nuclei to TDP-43- or FUS-positive cells. Centroid distance filtering, which limits the distance between the centre of the nucleus and the centre of the cytoplasm (maximum 40 pixels) was also applied to remove any debris or abnormal cells that may be fluorescing in the TDP-43/FUS channel. For the experiments quantifying endogenous TDP-43 in cells expressing GFP-tagged CYLD constructs, additional steps in the analysis were required. GFP-positive CYLD cells were identified in the same way as TDP-43- or FUS-positive cells using the Identify Primary Objects module, described above. However, prior to this, TDP-43-positive cells were identified by association with their corresponding nuclei, using the Identify Secondary Objects module. The watershed-image method was used with global, three-class Otsu thresholding. Size filtering was also applied (maximum 3000 pixels) to remove any clumped cells. The Relate and Filter modules were used to select only TDP-43-positive cells that were also positive for GFP. TDP-43- and GFP-positive cells were then linked to their corresponding nuclei using the Relate and Filter modules described above. For all experiments, in each TDP-43/FUS-positive cell, the nucleus and cytoplasm were separated into individual objects using the Identify Tertiary Objects module. Integrated (total) fluorescence intensity for both cellular compartments were then measured from the TDP-43/FUS channel image using the Measure Object Intensity module. Using these measurements, the cytoplasmic to nuclear ratio of TDP-43 or FUS was calculated for 300 cells per group, for each of the six replicates of the experiment. All filtering and thresholding steps in each experimental pipeline were kept consistent throughout all experimental replicates. Filtering values were specific to each experiment and require adjustment for different cell types and microscope parameters.

Statistical analysis

Data are presented as mean ± standard error of the mean (SEM) from at least six independent experiments of n = 300 cells. Mean fluorescent intensity of TDP-43 or FUS across six biological replicates was in most cases compared by repeated measures one-way ANOVA and Dunnett’s multiple comparisons test. The effect of CYLD on endogenous TDP-43 was analysed using repeated measures one-way ANOVA and Sidak’s multiple comparisons test. All statistical analyses were performed using GraphPad Prism 9. Significance for all tests was set at p < 0.05.