C9orf72 dipeptides disrupt the nucleocytoplasmic transport machinery and cause TDP-43 mislocalisation to the cytoplasm

A repeat expansion in C9orf72 is the major cause of both frontotemporal dementia and amyotrophic lateral sclerosis, accounting for approximately 1 in 12 cases of either disease. The expansion is translated to produce five dipeptide repeat proteins (DPRs) which aggregate in patient brain and are toxic in numerous models, though the mechanisms underlying this toxicity are poorly understood. Recent studies highlight nucleocytoplasmic transport impairments as a potential mechanism underlying neurodegeneration in C9orf72-linked disease, although the contribution of DPRs to this remains unclear. We expressed DPRs in HeLa cells, in the absence of repeat RNA. Crucially, we expressed DPRs at repeat-lengths found in patients (> 1000 units), ensuring our findings were relevant to disease. Immunofluorescence imaging was used to investigate the impact of each DPR on the nucleus, nucleocytoplasmic transport machinery and TDP-43 localisation. DPRs impaired the structural integrity of the nucleus, causing nuclear membrane disruption and misshapen nuclei. Ran and RanGAP, two proteins required for nucleocytoplasmic transport, were also mislocalised in DPR-expressing cells. Furthermore, DPRs triggered mislocalisation of TDP-43 to the cytoplasm, and this occurred in the same cells as Ran and RanGAP mislocalisation, suggesting a potential link between DPRs, nucleocytoplasmic transport impairments and TDP-43 pathology.

In addition to DPR pathology, the FTD/ALS-related protein TDP-43 (transactive response DNA-binding protein 43) also aggregates in C9orf72-positive patient brain 1 . TDP-43 is mislocalised from the nucleus to cytoplasm where it becomes abnormally hyperphosphorylated and forms insoluble intraneuronal inclusions. TDP-43 pathology correlates well with neuronal loss in C9FTD/ALS, leading to the proposal of an "amyloid-like cascade" hypothesis of neurodegeneration, whereby the C9orf72 expansion produces RNA foci and DPRs which somehow trigger TDP-43 pathology and this causes neuronal death 17 . However, the link between RNA foci/DPRs and TDP-43 mislocalisation is not well established, and there are likely multiple mechanisms through which products of the C9orf72 expansion impair cellular function which converge to cause neurodegeneration in FTD/ALS. Recent literature has highlighted defects in nucleocytoplasmic transport as an important feature of C9FTD/ ALS, with large-scale genetic screening studies identifying a number of key nucleocytoplasmic transport genes as modifiers of toxicity in fly and yeast models [18][19][20][21][22] . These studies led to the novel hypothesis that the C9orf72 expansion causes neurodegeneration by disrupting nucleocytoplasmic transport, with a wide range of downstream consequences for cellular processes necessary for normal function and survival. However, the exact mechanisms underlying nucleocytoplasmic transport impairments in C9FTD/ALS remain unclear. Much of the literature published to date utilises models which express pure G 4 C 2 repeat sequences, and therefore all five DPRs as well as G 4 C 2 repeat RNA will be produced in those models, rendering it impossible to determine the specific cause of any observed phenotypes. Furthermore, those models which use alternative-codon sequences to express DPRs in the absence of G 4 C 2 repeat RNA do so at extremely short repeat-lengths which are not found in patients. We have previously shown that repeat-length determines the impact of DPRs on cellular function and toxicity, with longer repeats causing more severe phenotypes 8,11 . As such, it is essential to model DPR pathology at repeatlengths found in patients wherever technically possible, to ensure relevance to human disease.
Here we used alternative-codon sequences for expression of DPRs in the absence of G 4 C 2 repeat-RNA, to investigate the specific impact of each dipeptide on the nuclear membrane and nucleocytoplasmic transport machinery. We also investigated the link between C9orf72 dipeptides and TDP-43, which is mislocalised to the cytoplasm in the brain of expansion-carrying patients. Crucially, we used previously characterised constructs 8 to express each DPR at long repeat-lengths which are known to be found in patients (> 1000 units), ensuring our findings are relevant to human disease.

Results
DPRs cause structural damage to the nucleus and nuclear membrane. Immunofluorescence imaging was performed on HeLa cells transfected with GA 1020 , GR 1136 , PR 1100, AP 1024 or empty pEGFP-N1 vector as a GFP-only control. Cells were stained with DAPI to visualise the nucleus and an antibody against lamin-B1, a key component of the lamin matrix which forms the nuclear membrane. A large proportion of cells containing GA, GR or PR inclusions had misshapen nuclei, indicating that DPRs cause structural abnormalities in the nucleus. Rather than forming smooth, rounded structures, nuclei were excessively folded to form "horseshoe" shapes or cause severe creasing in the nuclear membrane ( Fig. 1). Quantification of this phenotype showed that DPRs significantly increased the percentage of cells with misshapen nuclei, with 39.3% of cells containing GA inclusions affected (P = 0. 0131).
43.1% for GR (P = 0.0052), and 40.6% for PR (P = 0.0096) compared to 15.7% for GFP-only control (n = 3). Lamin B1 staining of cells containing GA, GR or PR inclusions demonstrated that the lamin matrix had a frayed appearance, suggesting that DPRs caused fragmentation of the nuclear membrane ( Fig. 1). Again, quantification showed that all three DPRs significantly increased the percentage of cells with this phenotype; 52.6% of cells containing GA inclusions had fragmented lamin B1 staining (P = 0.0047), 62.6% for GR (P = 0.0006), and 50.7% for PR (P = 0. 0071) compared to 22.2% for GFP-only control (n = 3). AP 1024 did not cause structural damage to the nucleus, with only 22.5% of AP-transfected cells exhibiting misshapen nuclei (P = 0.6797) and 33.0% exhibiting nuclear membrane fragmentation (P = 0.3865; images not shown, quantification in Fig. 1E, F, respectively). Since GA, GR and PR all significantly increased the percentage of cells with either misshapen nuclei or fragmentation of the nuclear membrane, Tukey's multiple comparisons tests were also performed to compare the severity of each phenotype between each DPR treatment ( Table 1). No significant differences were observed with the exception of the percentage of cells with nuclear fragmentation caused by GR compared to AP (P = 0.0113).
Components of the Ran cycle are mislocalised in DPR-transfected cells. Next, we investigated the impact of DPR expression on the nucleocytoplasmic transport machinery. The Ran cycle is an important cellular process that provides the energy to actively transport proteins across the nuclear membrane. Immunofluorescence staining was performed for Ran and RanGAP, two key components of the Ran cycle, in HeLa cells transfected with DPR constructs or GFP-only control vector. In control cells, Ran was a predominantly nuclear protein, with some diffuse expression in the cytoplasm ( Fig. 2A). However, in many cells containing GR inclusions, Ran mislocalised to the cytoplasm where it accumulated close to the nuclear membrane ( Fig. 2C/E). The intensity of Ran staining was measured in the nucleus compared to the perinuclear region, and >90% of Ran staining was contained within the nucleus in the vast majority (86.3%) of GFP-only control cells. 90% of staining within the nucleus was therefore set as a threshold for "normal" Ran localisation. In GR-expressing cells, the percentage of cells with abnormal distribution of Ran was significantly increased from 13.7 to 59.5% (P = 0.0025), indicating that GR caused mislocalisation of Ran to the cytoplasm. Accumulation of Ran in the cytoplasm was also observed in cells containing GA inclusions, however this occurred less frequently (36.0% of cells) and was not statistically significant (P = 0.1226; Fig. 2B/E). Neither PR (Fig. 2D) Fig. 2E), demonstrating that DPRs may differentially impact the nucleocytoplasmic transport machinery. RanGAP was also predominantly a nuclear protein, with relatively diffuse nuclear staining observed in GFPonly control cells. DPR expression caused RanGAP to form punctate, granular accumulations (Fig. 3). Interestingly, all DPRs tested significantly increased the percentage of cells with abnormal accumulation of RanGAP, (GA 32.8%, P = 0.0483; GR 81.4%, P < 0.0001; PR 56.4%, P = 0.0002; AP 38.9%, P = 0.0096 compared to GFP-only control of 15.0%), although the severity of this phenotype varied between peptides with arginine-rich dipeptides causing RanGAP accumulation most frequently. The severity of RanGAP mislocalisation was also significantly different between most of the DPR treatments (Table 2). These findings demonstrate that DPRs differentially cause mislocalisation of two key components of the Ran cycle, most likely contributing to the nucleocytoplasmic transport impairments reported to occur in C9FTD/ALS.

DPRs cause mislocalisation of TDP-43 to the cytoplasm.
Since DPR expression caused both structural damage to the nucleus and nuclear membrane and disruption to key components of the system that powers nucleocytoplasmic transport, we next investigated whether TDP-43, a nuclear protein known to be mislocalised to the cytoplasm in C9FTD/ALS patient brain, was affected by DPR expression. HeLa cells were co-transfected with an expression construct for mCherry-tagged human TDP-43 and GFP-tagged DPRs or empty pEGFP-N1 control. Immunofluorescence imaging was used to determine whether DPR expression affected the nuclear localisation of TDP-43. Figure 4 shows that GA, GR and PR all caused mislocalisation of TDP-43 to the cytoplasm. In some cells, TDP-43 also formed several small inclusions. Quantification of staining intensity showed that 18.3% of TDP-43 was localised to the cytoplasm in GFP-only control cells, whereas GA, GR and PR significantly increased this to 41.0% (P = 0.0003), 37.6% (P = 0.0011) and 29.5% (P = 0.0349) respectively. AP did not cause TDP-43 mislocalisation, with only 18.7% of staining contained within the cytoplasm (P = 0.999; data not shown). Statistical comparisons between each DPR treatment are shown in Table 3.

DPR-induced mislocalisation of TDP-43 and disruption to the Ran cycle occur together.
In order to determine whether the mislocalisation of TDP-43 observed could be linked to DPR-induced disruption to the nucleocytoplasmic transport machinery, we next investigated whether these two things occurred together. Since not all cells containing DPR inclusions exhibited mislocalisation of either Ran, RanGAP or TDP-43, immunofluorescence imaging was performed to determine whether these phenotypes occurred in the same cells. HeLa cells were co-transfected with the mCherry-TDP-43 and DPR-GFP constructs, and a Cy5-conjugated antibody was used to visualise either Ran or RanGAP in these cells (Figs. 5, 6, respectively). The relative intensity of TDP-43 staining in the nucleus and cytoplasm was calculated as described above. Since an average of 18.3% of the TDP-43 staining was previously shown to be contained to the cytoplasm in GFP-only control cells, this was considered the threshold for normal TDP-43 staining, with values above this considered as TDP-43 mislocalisation. The intensity of Ran staining in the nucleus compared to perinuclear area was quantified in cells exhibiting TDP-43 mislocalisation to the cytoplasm, and <90% of Ran staining within the nucleus was considered abnormal as described above. The majority of cells exhibiting DPR-induced TDP-43 mislocalisation also had abnormal mislocalisation of Ran to the cytoplasm (GA 59.3%, P = 0.0044; GR 81.9%, P = 0.0003; PR 71.5%, P = 0.0009), indicating that these two phenotypes occur alongside each other. The vast majority of cells exhibiting TDP-43 mislocalisation also exhibited granular RanGAP staining: 85.6% of cells for GA (P < 0.0001), GR 85.5% (P < 0.0001) and PR 82.1% (P < 0.0001). Therefore, both components of the Ran cycle were mislocalised in cells with DPR-induced TDP-43 mislocalisation to the cytoplasm, highlighting the possibility that defects in the nucleocytoplasmic machinery could be the connection between DPRs and TDP-43 pathology.

Discussion
We have shown that DPRs cause structural damage to the nucleus and fragmentation of the nuclear membrane when expressed at disease-relevant repeat-lengths of over 1000 units. Nuclear membrane fragmentation was previously reported in two Drosophila models of C9FTD/ALS, however these models expressed pure G 4 C 2 repeats (at the very short repeat-lengths of 30 and 58 units), and therefore it was not possible to determine Table 1. Since the statistical analysis in Fig. 1 showed that several DPRs significantly increased the frequency of nuclear abnormalities but to different extents, Tukey's multiple comparisons tests were also performed to compare between each DPR treatment. NS not significant.     20,23 . Here, we demonstrate that breakdown of the nuclear membrane is caused by DPRs, specifically GA, GR and PR but not AP. Our finding is supported by a recent report that staining for both lamin-B1 and lamin-A/C was fragmented in a transgenic mouse expressing PR 50 13 . Destabilisation of the lamin matrix which forms the nuclear membrane is likely to have detrimental consequences for cellular function and survival in a number of ways. Firstly, impaired structural integrity of the membrane is likely to impact the transport of proteins and other cargoes across it, leading to nucleocytoplasmic transport defects and perhaps breakdown of the barrier which maintains separation of the nuclear and cytoplasmic compartments. Secondly, lamins have several crucial functions besides supporting the structural integrity of the membrane, including roles in DNA repair, chromatin remodelling, transcription and cell migration 24,25 . Furthermore, loss of function mutations in lamins are known to trigger ER stress and activate the unfolded protein response, leading to premature apoptosis 26,27 . It is therefore possible that DPR-induced mislocalisation of lamins directly contributes to neurodegeneration in C9FTD/ALS.
We have also shown that DPRs differently impact the localisation of two proteins required for the Ran cycle: the small GTPase, Ran, and its activator protein, RanGAP. The Ran cycle is an essential process which provides the energy required for active transport of proteins across the nuclear membrane via hydrolysis of RanGTP to RanGDP, creating a concentration gradient across the nuclear membrane 28 . Ran and RanGAP mislocalisation have been previously reported to occur in (G 4 C 2 ) 30 flies and human patient iPSC-derived neurons 20 , however, as with the nuclear membrane fragmentation study, it was not possible to determine the specific cause of this phenotype since these models express G 4 C 2 repeat RNA as well as all five DPRs . We have shown that both phenotypes were caused by DPRs, with all DPRs causing RanGAP to accumulate in the nucleus and primarily GR causing Ran mislocalisation to the cytoplasm. RanGAP puncta have been previously observed in brain tissue from mice expressing GA 50 , supporting our findings 14 . Of note, punctate accumulations of RanGAP are also present in post-mortem brain tissue from C9FTD/ALS patients 20 , demonstrating the relevance of our findings to human disease. Mislocalisation of these key proteins is likely to cause a loss of function phenotype, resulting in nucleocytoplasmic transport impairments. This is supported by genetic screening studies in Drosophila expressing (G 4 C 2 ) 30 which found that overexpression of RanGAP rescued toxicity in this model, while RanGAP knockdown worsened the phenotype 20 . Deletion of Ran or RanGAP also increased toxicity in separate fly models expressing (G 4 C 2 ) 58 23 and PR 50 21 , highlighting the importance of the Ran cycle in cell survival. Combined with our finding that DPRs impair the structural integrity of the nuclear membrane, our data suggests that DPRs are the driving factor causing nucleocytoplasmic transport defects in C9FTD/ALS. Of particular note, the different DPRs had distinct effects on the nucleocytoplasmic machinery, despite their previously demonstrated toxicity in cell culture models (most potently GA, GR and PR, but also AP when expressed at long repeat-lengths) [5][6][7][8] . This demonstrates that the observed phenotypes were not a by-product of DPR toxicity causing cell death through some other mechanism. Functional assays to investigate of the impact of DPR-induced Ran cycle defects on nucleocytoplasmic transport would be beneficial to fully understand their contribution to disease.
Since the nucleocytoplasmic transport machinery was impacted by DPR expression in our model, we next investigated potential downstream consequences of this by assessing the localisation of TDP-43. TDP-43 is a predominantly nuclear RNA-binding protein which is mislocalised to the cytoplasm in C9FTD/ALS patient brain, where it aggregates to form insoluble inclusions. We found that expression of GA, GR and PR, but not AP, all caused TDP-43 to be mislocalised to the cytoplasm, demonstrating a functional link between these disease proteins. This is supported by a recent report of TDP-43 mislocalisation to the cytoplasm in Drosophila expressing either pure G 4 C 2 repeats or alternative-codon DPR sequences encoding GA 64 , GR 64 or PR 64 29 . GR 200 has also recently been found to cause mislocalisation of TDP-43 in mice expressing GR 200 30 , further supporting our findings. GA 100 also caused cytoplasmic aggregation of phosphorylated TDP-43 in human neuroblastoma cells 31 . Interestingly, this study found that neither GR 100 nor PR 100 caused mislocalisation of TDP-43, indicating that longer repeats may be required to cause a phenotype in this in vitro model. Indeed, the authors note that GA triggered the formation of TDP-43 inclusions in a length-dependent manner, and we have previously shown that DPR repeat-length is strongly correlated with toxicity in human cell culture and zebrafish 8,11 .
We found that DPR-induced TDP-43 mislocalisation occurred alongside DPR-induced Ran and RanGAP mislocalisation, implying that dysfunction of the nucleocytoplasmic transport machinery and TDP-43 pathology could be linked in DPR-expressing cells, although further work would be required to definitively confirm this. Ran and TDP-43 mislocalisation also occurred together in the (G 4 C 2 ) 30 fly model and in iPSC-derived neurons from expansion-carrying patients 20 , and our findings suggest that this was most likely triggered by DPRs. Table 2. Since the statistical analysis in Fig. 3 showed that all DPRs significantly increased the frequency of RanGAP mislocalisation but to different extents, a Tukey's multiple comparisons test was also performed to compare between each DPR treatment. NS not significant.   32 . Impaired nucleocytoplasmic transport may therefore be a common mechanism leading to TDP-43 mislocalisation and neurodegeneration in different subtypes of FTD. Our findings show that expression of DPRs in the absence of G 4 C 2 repeat RNA compromises the structural integrity of the nucleus and nuclear membrane, as well as mislocalising key components of the nucleocytoplasmic transport machinery. We also highlight a potential link between DPRs and TDP-43 pathology, which is supported by recent literature from a Drosophila model. It is possible that DPRs impair nucleocytoplasmic transport by disrupting the nuclear membrane and Ran cycle, leading to TDP-43 mislocalisation to the cytoplasm. However, this remains speculative and further work is needed to fully understand the links between DPR-induced mislocalisation of nucleocytoplasmic transport proteins and TDP-43.

Materials and methods
Constructs. We previously generated and characterised expression of DPR constructs for expression of GA 1020 , GR 1136 , PR 1100 and AP 1024 under the CMV promoter and with a C-terminal GFP-tag, using the pEGFP-N1 vector 8 . To overcome the potential issue of repeat-length instability between preparations of plasmid, each individual tube of DPR construct was size-screened before use by restriction digest and agarose gel as described previously, to ensure repeat-length was correct 8 . Empty pEGFP-N1 vector was used throughout as a GFP-only control. A full length TDP-43 clone (a gift from Professor Leonard Petrucelli, Mayo Clinic, Florida, USA) was used as a template for PCR using the forward primer (5′ GCT CAA GCT TAT ATG TCT GAA TAT ATT CGG GTA 3′) and reverse (5′ GAG CGG ATC CCT ACA TTC CCC AGC CAG AAG ACT 3′) containing HindIII and BamHI sites respectively. After digestion the product was ligated in pmCherry-C1 (Clontech) in the same restriction sites to create the pmCherry-C1 TDP-43 WT construct. All constructs were verified by DNA sequencing.
Cell culture and transfection. HeLa cells (ECACC #93021013) were maintained at 37 °C with 5% CO 2 in Dulbecco's Modified Medium (Sigma) supplemented with 10% foetal calf serum (Gibco), 2 mM l-glutamine (Sigma), 100 U/ml penicillin and 100 µg/ml streptomycin (Sigma). Cells were seeded on HCl-treated glass coverslips in 6-well plates overnight and transfected with DPR constructs or empty pEGFP-N1 control vector (800 ng/well) using FuGene HD (Promega; 7.2 µl/well). Cells were incubated at 37 °C for 48 h with a media change performed at 4 h to reduce DPR toxicity. Co-transfection of DPR and TDP-43 vectors was performed by addition of 500 ng of DPR vector or empty pEGFP-N1 control and 500 ng of TDP-43 vector or empty mCherry control and 9 µl FuGene HD. Cells were incubated at 37 °C for 48 h with a media change 4 h post-transfection.
Images were collected on a Zeiss Axioimager.D2 upright microscope using a 63x/1.4 EC Plan-Apochromat objective and captured using a Coolsnap HQ2 camera (Photometrics) through Micromanager software v1.4.23. Specific band pass filter sets for DAPI, FITC, Texas red and Cy5 were used to prevent bleed through from one channel to the next. Images were processed and analysed using Fiji ImageJ (http:// imagej. net/ Fiji/ Downl oads). The relative intensity of antibody staining in the nucleus and cytoplasm was determined by calculating corrected total cell fluorescence (CTCF) values on ImageJ. DAPI staining was used to determine the boundaries of the nucleus.

Statistical analysis.
All experiments were performed in triplicate, with a minimum of 30 cells (chosen from random fields of view based on the presence of GFP-tagged DPR expression) analysed per independent replicate. One-way ANOVAs with Dunnett's test for multiple comparisons were performed for all experiments using GraphPad Prism. P values below 0.05 were considered statistically significant. Table 3. Since the statistical analysis in Fig. 4 showed that all DPRs significantly increased the frequency of TDP-43 mislocalisation but to different extents, a Tukey's multiple comparisons test was also performed to compare between each DPR treatment. NS not significant.

Data availability statement
All data generated or analysed during this study are included in this published article. The mean percentage of cells with abnormal TDP-43 that also exhibited granular accumulation of RanGAP in the nucleus was significantly increased by GA 1020 (P < 0.0001), GR 1136 (P < 0.0001), and PR 1100 (P < 0.0001). n = 3, with a minimum of 30 cells analysed for each independent replicate. Data was analysed by one-way ANOVA with Dunnett's multiple comparison test. Error bars indicate SEM.