TDP43 nuclear export and neurodegeneration in models of amyotrophic lateral sclerosis and frontotemporal dementia

Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are progressive neurodegenerative disorders marked in most cases by the nuclear exclusion and cytoplasmic deposition of the RNA binding protein TDP43. We previously demonstrated that ALS–associated mutant TDP43 accumulates within the cytoplasm, and that TDP43 mislocalization predicts neurodegeneration. Here, we sought to prevent neurodegeneration in ALS/FTD models using selective inhibitor of nuclear export (SINE) compounds that target exportin-1 (XPO1). SINE compounds modestly extend cellular survival in neuronal ALS/FTD models and mitigate motor symptoms in an in vivo rat ALS model. At high doses, SINE compounds block nuclear egress of an XPO1 cargo reporter, but not at lower concentrations that were associated with neuroprotection. Neither SINE compounds nor leptomycin B, a separate XPO1 inhibitor, enhanced nuclear TDP43 levels, while depletion of XPO1 or other exportins had little effect on TDP43 localization, suggesting that no single exporter is necessary for TDP43 export. Supporting this hypothesis, we find overexpression of XPO1, XPO7 and NXF1 are each sufficient to promote nuclear TDP43 egress. Taken together, our results indicate that redundant pathways regulate TDP43 nuclear export, and that therapeutic prevention of cytoplasmic TDP43 accumulation in ALS/FTD may be enhanced by targeting several overlapping mechanisms.

A growing body of evidence indicates that nucleocytoplasmic transport dysfunction is a conserved pathway underlying neurodegeneration and the age-dependent susceptibility to neurodegenerative diseases [22][23][24][25][26][27] . Several disease-associated proteins, including fragments of the amyloid precursor protein, huntingtin, and TDP43, misfold into conformations that selectively disrupt nuclear import 28 . Furthermore, reduced expression of nucleocytoplasmic transport components is a defining feature of aging in fibroblasts and neurons directly differentiated from aged fibroblasts 26 . A subtle impairment in nucleocytoplasmic transport with age may predispose neurons to the detrimental effects of pathogenic mutations or proteotoxic stress, potentially explaining the age-dependent onset of neurodegeneration in ALS, FTD, and other neurodegenerative conditions. TDP43 contains a canonical leucine-rich nuclear export signal (NES) 29,30 that is predicted to be a substrate of exportin-1 (XPO1/CRM1), a conserved nuclear export factor in mammalian cells 31,32 . We took advantage of novel selective inhibitors of nuclear export (SINE) compounds 33 to determine whether preventing TDP43 nuclear export might slow or prevent neurodegeneration in models of ALS and FTD. Like the natural XPO1 inhibitor leptomycin B, SINE compounds interact specifically with the cargo binding pocket of XPO1, forming a covalent adduct with XPO1 at cysteine 528 34,35 . This covalent bond is slowly reversible (t 1/2 of ~24 h), likely contributing to how well the compound class is tolerated in ongoing human and veterinary clinical studies 36 . We therefore conducted a careful study of SINE compounds and their therapeutic potential in both in vitro and in vivo disease models, testing the ability of these small molecules to prevent cytoplasmic TDP43 mislocalization and extend neuronal survival.

Results
Safety of SINE compounds in rodent primary cortical neurons. We first performed dose-finding studies to identify the highest concentration of SINE compounds that can safely be used in cultured mammalian neurons. We tested two different SINE compounds, KPT-335 and KPT-350. Both are selective inhibitors of XPO1, but the former is more potent (IC50 = 33 nM and brain/plasma ratio 1.81; S.T. unpublished observations), while the latter exhibits greater blood brain barrier (BBB) permeability (IC50 = 161 nM and brain/plasma ratio = 2.24) 34 . To screen multiple concentrations of each compound in a rapid and sensitive manner, we took advantage of automated fluorescence microscopy 18,21,37,38 (Fig. 1a). This technique involves the transient transfection of rodent primary neurons with vectors encoding fluorescent proteins such as enhanced green fluorescent protein (EGFP), enabling their visualization by fluorescence microscopy. The system is fully automated, taking snapshots of thousands of neurons in culture, and is capable of returning to every neuron at any time. The images are analyzed using custom-designed programs that distinguish neurons from debris and glia, assigning each cell a unique identifying number. Cells are tracked longitudinally on a daily basis for 10 days, and cell death is determined prospectively by a set of criteria (rounding of the soma, blebbing, dendritic retraction or loss of fluorescence) that have proven to be sensitive measures of cell death in previous studies 18,39 . Because the event of cell death for the majority of neurons may not be directly witnessed by this method, the time of death is conservatively assigned as the last time the cell was classified as alive.
Using Cox proportional hazards analysis, we estimated the risk of death as a cumulative and instantaneous function for neurons expressing EGFP and treated with vehicle (DMSO) or different concentrations of KPT-335 or KPT-350. Drug was added 24 hours after transfection, after the first round of imaging. Concentrations above 150 nM caused excessive cell death (data not shown), so we focused on 150 nM or less of each compound. For KPT-335 ( Fig. 1b-d), significant toxicity was observed at 50 nM and 150 nM. The hazard ratio (HR) listed in Fig. 1c represents a relative risk of death for neurons treated with the indicated dose of SINE compound, relative to control neurons treated with vehicle. By this measure, both 50 nM and 150 nM KPT-335 increased the risk of death by ~40% (Fig. 1b,c), while no toxicity was observed for KPT-335 at 2.5 nM and 10 nM. The instantaneous risk of death, a measure of how likely neurons are to have died at the indicated time 40 (Fig. 1d), shows a prominent increase in the risk of death for neurons treated with 50 nM and 150 nM KPT-335 that peaks 1 day after application of the compound (corresponding to day 2 of imaging). In contrast, 10 nM KPT-335 slightly but significantly (Fig. 1b,c) reduced the risk of death in comparison to vehicle alone, suggesting a modest protective effect. While 2.5 nM KPT-335 exhibited a small and significant protective effect when measured by the log-rank test (Fig. 1c), we were unable to confirm statistical significance by Cox proportional hazards analysis, which accounts for multiple comparisons among the separate populations 41 .
A similar effect was observed in primary neurons treated with KPT-350 ( Fig. 1e-g). Here, however, toxicity was noted only at 150 nM. While 10 and 50 nM KPT-350 extended neuronal survival by 15-20% (Fig. 1e,f), the effect of the 2.5 nM dose was less pronounced and as before, only statistically significant by the log-rank test. The instantaneous risk of death plot (Fig. 1g) demonstrates acute toxicity peaking 1 day after drug application for high doses of KPT-350, analogous to the effect of KPT-335 (Fig. 1d). Protection, or reduced risk of death, was noted for lower doses of KPT-350 beginning approximately 72 h after application of the compound. These results suggest that SINE compounds modestly improve survival in control neurons when used at low doses (≤10 nM KPT-335 and ≤50 nM KPT-350). We therefore selected the lowest 2 doses for investigating the utility of XPO1 inhibition in disease models.

SINE compounds in models of ALS and FTD.
To gauge the therapeutic potential of SINE compounds in ALS and FTD, we employed a neuronal disease model involving overexpression of wild-type (WT) and ALS-associated mutant TDP43 18,21,38 . This model recapitulates key features of ALS and the most common pathologic subtype of FTD (frontotemporal lobar degeneration with TDP43 deposits, or FTLD-TDP), including enhanced neurodegeneration upon expression of disease-associated mutant TDP43. In this system, as in other neurodegenerative diseases characterized by accumulation of an endogenous protein 42,43 , overexpression of WT TDP43 (TDP43 WT ) is sufficient to reproduce neuron loss. Similar toxicity arising from TDP43 WT overexpression has been noted in multiple in vitro and in vivo model systems 18 protein that is also seen in other neurodegenerative diseases, including Alzheimer's and Parkinson's disease 42,50 . Versions of TDP43 carrying mutations linked to familial ALS and FTD (i.e. A315T) are more toxic at lower doses in comparison to TDP43 WT , and can form insoluble cytoplasmic inclusions 18,45,48,51 .
For these studies, we transiently transfected primary neurons with vectors encoding the red fluorescent protein mApple, as well as TDP43 WT or mutant TDP43 A315T fused to EGFP (Fig. 2a,b). This enabled us to independently track the survival of neurons using mApple as a cellular marker in the red channel, and the expression were added after imaging on day 1. Neurons were identified based on morphology (blue outlines) and assigned a unique identifier (blue #). Death (red) is determined by loss of fluorescent intensity (neuron 7, day 3) or changes in morphology (neuron 15, day 6), and time of death for each cell is used to estimate the cumulative (b,e) and instantaneous (d,g) risk of death functions. Tables showing neuron number (#), hazard ratio (HR), Cox proportional hazards p value (p), confidence interval (95% CI), and log-rank p value (LR p) are shown in (c) and (f). Doses considered "safe" are indicated with black arrows. Each condition represents 3-6 combined experiments, minimum 8 technical replicates per experiment.
Treating TDP43-expressing neurons with KPT-350 produced similar results ( Fig. 2f-h). The 2.5 nM dose of KPT-350 was the most neuroprotective, reducing the risk of death in cells transfected with TDP43 WT -EGFP by ~10% (HR 0.90, p < 0.05 by Cox proportional hazards), while higher doses of KPT-350 trended towards toxicity in these cells. The extension of neuronal survival by KPT-350 was observed by measuring the cumulative (Fig.  2f) and instantaneous risk of death (Fig. 2h) for transfected neurons. As with KPT-335 (Fig. 2c,d), 10 nM KPT-350 demonstrated slight toxicity in TDP43 WT -EGFP expressing neurons that was significant only by the log-rank test, but not Cox proportional hazards analysis (Fig. 2g). In contrast, neither KPT-335 nor KPT-350 significantly affected the survival of neurons expressing TDP43 A315T -EGFP. Together, these data show that low doses (2.5 nM)

Utility of SINE compounds in a neuron model of Huntington's disease.
Given the neuroprotective effects of SINE compounds in control neurons expressing EGFP alone ( Fig. 1) and in those transfected with WT TDP43 (Fig. 2), we surmised that SINE compounds might exert a general and non-specific beneficial effect that would extend to neurodegenerative diseases unrelated to ALS or FTD. We therefore turned to a neuronal model of Huntington's disease, in which neurons overexpress a fragment of mutant huntingtin (Htt) carrying 96 polyglutamine residues fused to EGFP (Htt 96Q -EGFP). In previous studies, neurons expressing this construct exhibit neurodegeneration and the time-dependent formation of nuclear Htt-rich inclusions reminiscent of those found in individuals with Huntington's disease 37,38,52,53 . Primary rodent cortical neurons were transfected with vectors encoding the cellular marker mApple, and either Htt 96Q -EGFP or EGFP alone, treated with KPT-335 or -350, and imaged by automated microscopy (Fig. 3a). As expected, Htt 96Q -EGFP expression increased the risk of death by ~40% in comparison to EGFP alone ( Fig. 3b-g). Neither KPT-335 nor KPT-350 effectively promoted neuronal survival in this model, suggesting that their effects are likely not generalizable to all neurodegenerative diseases. However, as toxicity in this model is more likely to correlate with increased, rather than decreased nuclear retention of mutant Htt 54 , it is possible that the modest neuroprotection afforded by SINE compounds may be offset by enhanced toxicity of nuclear mutant Htt.
In contrast to our observations, a prior study 55 noted that KPT-350 effectively prevented neurodegeneration in a Huntington's disease model. To determine the potential reasons for such discrepancies, we sought to replicate the conditions in which the previous investigations were performed. The most notable differences involved the timing of drug application and the method for determining cell death: while Grima et al. 55 (Fig. S1d,e). We suspected that this early drug-dependent toxicity might skew the results of survival analysis if not considered when assessing the effect of SINE compounds on neurodegeneration. To test this possibility, we tracked neuronal survival by automated microscopy beginning 24 h or 48 h following application of SINE compounds, which were added at the time of transfection. In doing so, we recapitulated the apparent neuroprotective effect of SINE compounds in cells overexpressing Htt 96Q -EGFP ( Fig. S1f-h). The apparent neuroprotective effect was particularly evident when the analysis began 48 h following transfection and drug addition, after the peak of SINE compound-induced toxicity had passed. Taken together, these data confirm the initial toxicity of SINE compounds when applied to neurons, an effect that may misrepresent their neuroprotective potential unless properly taken into account.

Evaluation of SINE compound efficacy in an in vivo ALS/FTD model.
We also explored whether XPO1 inhibition could prevent neurodegeneration arising from TDP43 accumulation in animals. For this purpose, we used a rat model of TDP43-induced paralysis [56][57][58][59] . Human WT TDP43 cDNA was delivered to neonatal rats via an intravenous injection of recombinant adeno-associated virus serotype 9 (AAV9), yielding consistent limb paralysis that gradually progresses over the ensuing 2-16 weeks 57,59 . To determine if SINE compounds prevented clinical manifestations of TDP43-related toxicity, transduced rats were treated at day 2 with one of two separate doses of KPT-350 (7.5 mg kg −1 and 10 mg kg −1 ) (Fig. 4a), as this compound exhibits greater blood brain barrier (BBB) permeability, and identical doses of KPT-350 proved beneficial in prior studies involving other neurological disorders 34,60 . After verifying TDP43 expression within the central nervous system of transduced rats ( Fig. 4b and Fig. S2), we measured motor function of treated and untreated animals using the hang test (Fig. 4c) and the rotarod at 3-4 weeks of age. As expected, TDP43-transduced rats exhibited a dramatically shortened latency to fall on both tests, in comparison to uninjected controls. Treatment with 7.5 mg kg −1 KPT-350 almost completely rescued the deficits in grip strength as measured by the hang test at 3 weeks (Fig. 4d). Although there was a trend towards improvement on the hang test with the 10 mg kg −1 dose, the effect was non-significant. In addition, neither the 7.5 mg kg −1 nor the 10 mg kg −1 dose affected rotarod performance at 4 weeks ( Fig. 4e) or 8 weeks (Fig. S3a). We also observed dose-dependent weight loss in KPT-350-treated animals (Fig. S3b), consistent with prior reports of anorexia and weight loss following treatment with SINE compounds 36 . Thus, SINE compounds partially rescue motor function in TDP43-overexpressing animals, although their use is limited by weight loss.

Subcellular protein localization in response to XPO1 inhibition. Inhibition of XPO1 by SINE com-
pounds is expected to enhance nuclear localization for proteins that rely upon XPO1 for nuclear export. To verify the activity of these compounds in living cells, we applied SINE compounds to HEK293 cells expressing a red fluorescent protein (mCherry) carrying a strong leucine-rich nuclear export signal recognized by XPO1, as well as a weak nuclear localization signal (NLS-mCherry-NES) 61 . Cells were cotransfected with EGFP to mark the cytoplasm and stained with the cell permeable nuclear marker Draq5. We then imaged the cells by fluorescence microscopy at regular intervals following the addition of SINE compounds (Fig. 5a). As a positive control, we used leptomycin B (LMB), a potent and irreversible XPO1 inhibitor 32,62 . Upon addition of 2.5 ng/ml LMB, the NLS-mCherry-NES reporter rapidly accumulated within the nucleus of transfected HEK293 cells; by 4 h, nearly all of the protein was found within the nucleus (Fig. 5a-c). Low-dose (2.5 nM) KPT-335 and KPT-350 produced no discernable change in reporter localization up to 12 h after drug addition, and 10 and 50 nM of the SINE compounds were equally ineffective. Only 150 nM of KPT-335 and KPT-350 elicited measurable increases in reporter nuclear localization (Fig. 5a-c). This effect was first evident 4 h after KPT-335 addition, reaching a peak by 7 h. Consistent with its enhanced XPO1 binding affinity, KPT-335 elicited more consistent and pronounced nuclear localization of the reporter than did KPT-350. These results suggest that SINE compounds are incapable of modulating nucleocytoplasmic transport at the low doses that were associated with neuroprotection in previous experiments (Figs 1, 2).
To determine if these observations extended to mature post-mitotic neurons, we transfected rodent primary cortical neurons with vectors encoding EGFP and NLS-mCherry-NES, applied LMB, SINE compounds or vehicle, and imaged the cells by automated microscopy (Fig. 6a,b). As with HEK293 cells, treating transfected neurons with LMB produced a rapid relocalization of NLS-mCherry-NES from the cytoplasm to the nucleus. At low doses (2.5 nM), neither KPT-335 nor KPT-350 provoked a significant change in the subcellular distribution of the reporter. Higher doses of the SINE compounds (150 nM) resulted in a subtle and gradual increase in nuclear reporter localization, with peak effects observed 8 h or more after drug addition. Once more, KPT-335 caused a more conspicuous change in reporter localization than did KPT-350, indicative of more potent XPO1 inhibition by KPT-335. These results are consistent with those from HEK293 cells, suggesting that high-dose but not low-dose SINE compounds effectively enhance nuclear localization of XPO1 substrates. Our observations also show that XPO1-dependent nucleocytoplasmic transport machinery is conserved between primary rodent neurons and HEK293 cells.
TDP43 carries a leucine-rich sequence of the type recognized by XPO1 29,30 . The residues comprising the putative signal sequence (amino acids 239-243 and 248-250) are surface exposed (Fig. S4), suggesting that they may serve as a valid recognition signal for XPO1. However, no investigations to date have examined the functional localization of TDP43 in response to XPO1 inhibition. We sought to determine if TDP43 undergoes XPO1-mediated nucleocytoplasmic transport through two complementary approaches. First, we treated rodent primary cortical neurons with vehicle (DMSO) or the safest dose of SINE compounds used in disease models (Fig. 2), and analyzed the amount of endogenous TDP43 within the nuclear and cytoplasmic fractions of treated neurons by immunoblotting (Fig. S5a-c). Low-dose KPT-335 and KPT-350 had no observable effect on the subcellular distribution of endogenous TDP43 or Hsc70, a documented XPO1 substrate 63,64 . Higher concentrations of SINE compounds also failed to enhance nuclear TDP43 localization, but were limited by dose-dependent cytotoxicity (data not shown). Similarly, no change in the localization of TDP43 was noted in the brains of rats transduced with AAV-TDP43 and treated with KPT-350, in comparison to those treated with vehicle alone (Fig. S5d-e).
In parallel, we also measured the effect of SINE compounds on TDP43 localization in living neurons by fluorescence microscopy. Primary rodent cortical neurons were transfected with vectors encoding EGFP and either NLS-mCherry-NES (Fig. 6a,b) or TDP43 WT -mApple (Fig. 6c,d), treated with LMB, SINE compounds or vehicle, and imaged by automated microscopy at regular intervals over a 12 h period. Despite their ability to prevent nuclear export of the NLS-mCherry-NES reporter (Fig. 6a,b), neither LMB nor the SINE compounds significantly  59 ). Full-length blots appear in Fig. S1. (c) The hang test measures the time it takes for animals to fall from grid-like scaffold when it is lifted and tilted. A single representative animal administered AAV9-TDP43 is shown; note the inability of the hind limbs to clasp the wire mesh. (d) TDP43 expression impaired motor performance on the hang test at 3 weeks, an effect that could be mitigated by 7.5 mg kg -1 KPT-350. (e) Rotarod testing at 4 weeks of age showed prominent deficits associated with TDP43 expression that were resistant to treatment with SINE compounds. n, number of animals per condition. ns, not significant, mean ± s.e.m. *p < 0.05, **p < 0.01, ***p < 0.001, one-way ANOVA with Bonferroni correction. affected TDP43 WT -mApple distribution in neurons (Fig. 6c,d). This observation suggests that TDP43 may not be an XPO1 cargo protein, or that we were unable to accurately detect changes in TDP43 WT -mApple localization with XPO1 inhibition because TDP43 is predominantly nuclear at steady state. To distinguish among these possibilities, in place of TDP43 WT -mApple we used TDP43 mNLS -mApple, which exhibits impaired but not abolished nuclear import. This construct contains point mutations disrupting the first half of the bipartite nuclear localization signal (NLS) (82KRK85 > 82GGG85), resulting in a predominantly cytoplasmic distribution 18,29 , much like NLS-mCherry-NES (Fig. 6e). However, we noted no significant increase in TDP43 mNLS -mApple nuclear localization upon application of LMB and the SINE compounds to transfected neurons, even at higher doses (Fig. 6f). Together, these data imply that TDP43 localization is unaffected by XPO1 inhibition. Subcellular protein localization in response to XPO1 depletion or overexpression. We also considered the possibility that TDP43 may utilize multiple pathways for nuclear export. To address this hypothesis, we designed a system to quickly and efficiently monitor TDP43 distribution in a large number of cells. We modified our automated survival analysis platform to i) identify both the nuclear compartment (based on the signal from a cell-permeable nuclear marker) and the cytoplasm (as identified by a diffuse fluorescent protein) for each cell (Fig. S6b-e); ii) exclude dead or dying cells based on nuclear pyknosis; iii) calculate nuclear and cytoplasmic TDP43 intensities for individual neurons; and iv) report their ratio (Fig. S6b-e,h,i). In comparison with manual analysis (Fig. S6a,f,g), the automated system consistently selects a more conservative cytoplasmic outline but is equally accurate in identifying changes in TDP43 distribution in response to different experimental conditions (compare Fig. S6f and h vs. g and i). These promising results confirm the utility of the automated system for assessing the distribution of TDP43 and other proteins in living neurons.
We took advantage of this platform to investigate alternative mechanisms of TDP43 nuclear export. We first asked if any single exporter was necessary for TDP43 nuclear egress. For these experiments, we systematically knocked down all 7 nuclear exporters and NXF1, a nuclear export factor unrelated to the karyopherins that has previously been linked to TDP43 transport 25 . We assessed TDP43 WT -mApple localization in rodent primary neurons 48 hours after treatment with pooled siRNAs targeting each exporter using our automated system for measurement of subcellular protein distribution. Verification of knockdown efficiencies revealed a 37-59% reduction in average exporter transcript abundance (Fig. S7). We noted small but significant reductions in the mean TDP43 WT -mApple nucleocytoplasmic ratio (NCR) upon knockdown of XPO1, CSE1L, XPO4, and NXF1 (Fig.  7a). Knockdown of XPO7 was the sole manipulation that significantly increased the mean TDP43 WT -mApple NCR in transfected cells (Fig. 7a,b); even so, the magnitude of the change (1%) suggested that redundant mechanisms participate in TDP43 nuclear egress.
Given previous evidence indicating a proportional relationship between cytoplasmic TDP43 localization and neurodegeneration 18 , we also tracked the survival of primary neurons transfected with TDP43 WT -mApple and siRNA targeting individual nuclear exporters (Fig. 7a-d). Consistent with the effect of XPO7 knockdown on TDP43 WT -mApple distribution, siRNA targeting XPO7 reduced the risk of death in TDP43-expressing neurons by approximately 20% (Fig. 7d). We observed a similar amelioration of TDP43-mediated toxicity upon XPO1 knockdown, despite the small decrease in TDP43 WT -mApple NCR in these cells. This modest neuroprotection mirrors that afforded by SINE compounds (Fig. 2), suggesting that genetic and pharmacologic inhibition of XPO1 are both capable of mitigating TDP43-related cell death independent of their effects on TDP43 localization. The combined knockdown of XPO1 and XPO7 showed no additive effect (HR of 0.84 compared to 0.82 and 0.81, Fig.  7d), indicating an upper threshold of neuroprotection afforded by this mechanism. XPO4 depletion was also modestly protective, reducing the risk of death by 10%. In contrast, NXF1 depletion produced a striking exacerbation of TDP43-dependent toxicity, more than doubling the cumulative risk of death.
To further explore the possibility that TDP43 undergoes nuclear egress by overlapping pathways, we overexpressed a subset of exporters with experimental or predicted effects on TDP43 nucleocytoplasmic transport, and assessed the subcellular distribution of TDP43 WT -mApple in transfected primary neurons. For these experiments, we selected 4 distinct exporters for overexpression: XPO7, because knockdown of this exporter increased TDP43 WT -mApple NCR and reduced TDP43-mediated neurodegeneration (Fig. 7); XPO1, due to the putative XPO1 recognition motif in TDP43 (Fig. S4), as well as the subtle neuroprotective effects of SINE compounds (Fig. 1) and XPO1 depletion (Fig. 7) in TDP43-overexpressing neurons; NXF1, which dramatically enhanced TDP43-related toxicity and reduced the TDP43 WT -mApple NCR in transfected neurons (Fig. 7); and XPO5 as a control, since knockdown of this protein failed to affect toxicity or localization of TDP43 WT -mApple (Fig. 7). To minimize the potential for steric hindrance from fluorescent tags, we chose to use untagged exporter proteins in our overexpression assays. XPO1, XPO5, and NXF1 were each appended to the EGFP open reading frame with a self-cleaving viral 2A sequence between the exporter and the fluorophore, enabling equimolar expression of separate polypeptides from the same mRNA transcript 65,66 . Untagged XPO7 was transfected concurrently with a plasmid containing the EGFP cell marker, and its simultaneous expression confirmed by immunocytochemistry (Fig. S8a,b).
Consistent with the hypothesis that TDP43 is transported out of the nucleus via multiple pathways, overexpression of XPO7, XPO1 and NXF1, but not XPO5, significantly decreased the mean TDP43 WT -mApple NCR in transfected primary neurons (Fig. 8a,c). Cytoplasmic TDP43 WT -mApple was clearly visible in cells expressing XPO7, XPO1 and NXF1 (Fig. 8b,d), but nuclear TDP43 WT -mApple exclusion was noted only in NXF1-transfected cells. In contrast, XPO5 had no measurable effect on TDP43 localization. As expected based on the established relationship between cytoplasmic TDP43 redistribution and neurodegeneration 18 , XPO1 and NXF1 exacerbated TDP43-dependent toxicity in transfected neurons (Fig. 8e-j). However, XPO5 overexpression also increased the cumulative and instantaneous risk of death without affecting TDP43 WT -mApple localization, suggesting TDP43-independent toxicity. Supporting this notion, exporter overexpression was significantly toxic in the absence of TDP43 (Fig. S8). Although overexpression of XPO7 alone increased the cumulative risk of death by nearly 100%, it had little effect on TDP43-mediated toxicity in primary neurons (Fig. 8g). Taken together, these data suggest that, while TDP43 is indeed a cargo of XPO1, it is also exported from the nucleus via interactions with multiple transporters, including XPO7 and NXF1.

Discussion
Here, we conducted a careful study of selective XPO1 inhibition and nuclear TDP43 export in ALS and FTD disease models. Our observations indicate that, within a defined therapeutic window, SINE compounds modestly promote neuronal survival in primary neurons, and partially rescue a motor phenotype in rats overexpressing the RNA binding protein TDP43. However, the therapeutic index for XPO1 inhibition via SINE compounds was narrow, and significant neurotoxicity was observed for the doses that effectively blocked nuclear export. TDP43 localization was largely unaffected by SINE compounds, particularly at the low doses that were safe in neurons. Our data suggest that TDP43, like other essential RNA binding proteins 67 , is trafficked through partially redundant and overlapping pathways, ensuring that TDP43 nuclear export will proceed unchecked even if one or more pathways are blocked. Furthermore, the modest beneficial effects of SINE compounds in ALS/FTD models are likely to be independent of TDP43 localization or nuclear export itself.
Although we observed no effect of SINE compounds on TDP43 distribution, low doses of these compounds moderately extended neuronal survival in primary neurons overexpressing TDP43 WT -EGFP, and KPT-350 partially rescued motor phenotypes in an animal model of ALS involving TDP43 overexpression throughout the CNS. Because cytoplasmic TDP43 deposition is characteristic of ~95% of all ALS and >50% of FTD 14,17,68,69 , in most cases reflecting dysregulation of the wild-type protein, these results may still have relevance for both ALS and FTD. The in vivo doses used here (7.5-10 mg kg −1 ) fall within the range reported by recent investigations that demonstrated neuroprotective properties of SINE compounds in animal models of inflammatory demyelination and traumatic brain injury 34,60 . These prior studies hinted at the possibility that XPO1 inhibition might be generally neuroprotective. Our results, however, suggest that KPT-350 is incapable of fully preventing neurodegeneration in TDP43-overexpressing rats. Moreover, SINE compounds elicited substantial neurotoxicity at doses associated with impaired nuclear export of a reporter protein, failed to prevent neurodegeneration in response to the ALS-associated TDP43 A315T -EGFP, and were ineffective at promoting neuronal survival in a Huntington's disease model.  Overexpression of untagged XPO7 in rodent primary neurons effectively reduces the nuclear to cytoplasmic ratio (NCR) of TDP43 WT -EGFP compared to cells transfected with empty vector (EV). (c,d) Overexpression of either XPO1 or NXF1 appended to the self-cleaving EGFP-2A peptide also results in significant reductions in TDP43 WT -mApple NCR, while XPO5 overexpression has no effect on TDP43 WT -mApple localization. Arrowheads indicate cytoplasmic TDP43 accumulation, while arrows highlight nuclear clearance of TDP43. (e-g) XPO7 overexpression has little effect on TDP43-dependent toxicity. Cumulative and instantaneous hazards shown in (e) and (f), respectively, and additional data, number of neurons (#), hazard ratio (HR), Cox proportional hazards p value (p), confidence interval (95% CI), and log rank p value (LR p) are indicated in (g). XPO1, XPO5 and NXF1 potentiate the toxicity of TDP43 WT -mApple overexpression (h-j). Data in a-j represent 3 biological replicates per condition, minimum 8 wells per replicate/condition. The NCR was normalized to control for each biological replicate before pooling, and plotted as mean normalized log (NCR) ± s.d. ****p < 0.0001, 2-way ANOVA with Tukey's adjustment. In a related study 55 , KPT-350 displayed neuroprotective effects in a primary neuron Huntington's disease model. Several differences account for this discrepancy. First, we added SINE compounds 24 h following transfection to mimic treatment of the disease rather than prevention, while Grima et al. 55 applied KPT-350 immediately following transfection with Htt 96Q -EGFP. We show that applying SINE compounds at the time of transfection elicits substantial early toxicity that, if not recognized, may inappropriately skew the results of subsequent survival analyses. Second, we directly assessed the survival of transfected neurons by visualizing cellular morphology, rather than relying on nuclear condensation of fixed cells as a proxy. Lastly, we followed neurons periodically for 10 days instead of 2, minimizing the significance of transient effects and enabling us to capitalize on the increased statistical power of longitudinal survival analysis.
XPO1 is responsible for transporting several anti-inflammatory, anti-oxidant, and cytoprotective transcription factors from the nucleus to the cytoplasm 60,63,64 . Consequently, XPO1 inhibition may affect many inflammatory/immune pathways relevant to neurological diseases, including signaling cascades involving Nrf2, NF-ĸB, and FOXOs. Low doses of SINE compounds modestly improved neuronal survival in control neurons and those expressing TDP43 WT -EGFP, but at these doses we failed to detect a change in the distribution of XPO1 target substrates, including the NLS-mCherry-NES reporter and endogenous Hsc70. Although we cannot rule out the possibility that the distribution of other XPO1 cargos was affected, the inability of low dose SINE compounds to alter the nucleocytoplasmic distribution of the NLS-mCherry-NES reporter-which carries a strong NES, as evidenced by the nuclear exclusion of this construct at steady-state-suggests that the beneficial effects of SINE compounds may be separate from XPO1's role in protein trafficking 70 . Moreover, high doses of SINE compounds that exhibited measurable effects on nuclear trafficking also elicited significant toxicity when applied to primary neurons, consistent with the conclusion that neuroprotection by SINE compounds is separable from their effects on nucleocytoplasmic shuttling.
The predicted TDP43 NES lies within residues 239-250, partially overlapping the second RNA recognition motif 29,30 . Based on homology to the hydrophobic linear sequence recognized by XPO1 32 , TDP43 was presumed to be a XPO1 substrate. Moreover, genetic disruption of the presumed TDP43 NES resulted in the formation of insoluble nuclear aggregates 29 and reduced the toxicity of cytoplasmically-mislocalized TDP43 18 . However, TDP43 was not among the list of XPO1 target proteins identified by 2 independent proteomic screens 63,64 , and no data has previously demonstrated a clear change in TDP43 localization when this domain is deleted or disrupted. These observations, together with our data showing that XPO1 overexpression results in a modest increase in cytoplasmic TDP43 WT -mApple, but XPO1 inhibition by SINE compounds and leptomycin B fail to affect TDP43 WT -mApple localization, suggest that XPO1 is not the only exporter responsible for nuclear TDP43 egress. Supporting this interpretation, overexpression of XPO7 and NXF1 as well as XPO1 effectively facilitated nuclear TDP43 WT -mApple export. Prior studies confirm that NXF1 physically interacts with TDP43 71 , consistent with its role in TDP43 nuclear export, but this is the first indication that TDP43 may be an XPO7 cargo protein. In contrast, XPO5, whose yeast homolog MSN5 was previously identified as a genetic modifier of TDP43 toxicity 71,72 , showed no effect on TDP43 localization or toxicity in our assays. In fact, among the 8 nuclear exporters tested, only XPO7 depletion significantly increased the mean TDP43 WT -mApple NCR, albeit only by 1%. Little is known about XPO7 cargo proteins 73 , and investigation into a potential XPO7/TDP43 interaction represents a promising future avenue of investigation.
Disrupted nucleocytoplasmic transport has emerged as a common underlying theme in age-dependent neurodegenerative diseases. Amyloid, mutant huntingtin 27,55 , mutant TDP43 18 and mutant FUS 74 have all been associated with abnormalities in nucleocytoplasmic transport. In addition, the most common mutation responsible for ALS and FTD -hexanucleotide expansions in the first intron of C9orf72 11,12 -also impairs nucleocytoplasmic transport [22][23][24][25]75 . While XPO1 knockdown enhanced toxicity in some models of mutant C9orf72-related ALS/FTD 24,25 , XPO1 inhibition by SINE compounds mitigated toxicity in others 22 . In light of our data demonstrating a mild neuroprotective effect of SINE compounds independent of their effects on nucleocytoplasmic transport, these observations provide additional evidence that SINE compounds likely act through alternative mechanisms to relieve toxicity in disease models. Consequently, future in vitro and in vivo studies are warranted to more closely examine the effect of SINE compounds in models of ALS, FTD and related disorders. Identifying the downstream effectors of SINE compound-mediated neuroprotection is an important next step towards the goal of developing more potent agents with a wider therapeutic index. Finally, our finding that multiple exporter proteins mediate TDP43 nuclear egress underscores the need for further characterization of novel transport motifs within TDP43; this information is crucial for precision targeting of TDP43 nucleocytoplasmic transport as a therapeutic strategy in ALS and FTD.

Adeno-associated virus preparation.
The transgene expression cassette included AAV2 terminal repeats, the hybrid cytomegalovirus/chicken beta-actin promoter, human wild-type TDP43, and the bovine growth hormone polyadenylation sequence 76 . Helper and AAV9 capsid plasmids were from the University of Pennsylvania 77 . The expression cassette was packaged into recombinant AAV9 as previously described 76 . Viral stocks were sterilized using a Millex-GV syringe filter (Millipore), aliquoted, and frozen. Viral genome copies were titered using a dot-blot assay.
SCIeNTIfIC RepoRts | (2018) 8:4606 | DOI:10.1038/s41598-018-22858-w Ethics statement. All vertebrate animal work was approved by the Committee on the Use and Care of Animals (UCUCA) at the University of Michigan, and the Louisiana State University Health Sciences Center at Shreveport's Animal Care and Use Committee (ACUC). All experiments were performed in accordance with UCUCA and ACUC guidelines. Rats (Rattus norvegicus) used for primary neuron collection were housed singly in chambers equipped with environmental enrichment. Rats used for in vivo studies were housed with the dam until weaning at three weeks of age. Thereafter, they were housed in pairs by gender. All studies were designed to minimize animal use. Rats were cared for by the Unit for Laboratory Animal Medicine at the University of Michigan or veterinary specialists at Louisiana State University; all individuals were trained and approved in the care and long-term maintenance of rodent colonies, in accordance with the NIH-supported Guide for the Care and Use of Laboratory Animals. All personnel handling the rats and administering euthanasia were properly trained in accordance with the UM Policy for Education and Training of Animal Care and Use Personnel. Euthanasia was fully consistent with the recommendations of the Guidelines on Euthanasia of the American Veterinary Medical Association.
In vivo animal studies. 40 Sprague-Dawley rats (Harlan, Indianapolis, IN) of both genders from 6 different litters were used for these studies. Animals received temporal vein injections on post-natal day 1 and were monitored for 8 weeks. Animals were either injected with AAV9 TDP43 at a dose of 1 × 10 12 vector genomes (vg; N = 23) or left uninjected (N = 17). KPT-350 was dissolved in a combination of poloxamer pluronic F-68 (6 mg ml −1 ) and plasdone PVP K-29/32 (6 mg ml −1 ) in sterile water (vehicle). For the neonatal rats, dosing began at post-natal day 2. Neonatal rats were given KPT-350 or vehicle (N = 19) twice per week via a small rubber tube attached to a 1-ml syringe. Two doses of KPT-350 were used, 7.5 mg kg −1 (N = 13) and 10 mg kg −1 (N = 8).
Performance on the rotarod was measured at 4 and 8 weeks as described (Wang et al. 57 ). The rats' grip strength was studied at 3 weeks using a hang test. Here, rats were placed on a cage cover with a metal grate and then held above a cage with ample bedding. The grate was turned to a vertical position and the amount of time the rat remained on the grate was timed (latency to fall). Trials were capped at 60 seconds. Three trials per rat were conducted and averaged.
Cell culture and transfection. Rodent primary cortical neurons were prepared as described in 18 . Briefly, neurons dissected from embryonic day 20-21 Long Evans rat pups were cultured at a density of 0.6 × 10 6 cells ml -1 in 96 well plates coated with laminin (Corning) and D-polylysine (Millipore). Euthanasia for these experiments is entirely consistent with the recommendations of the Guidelines on Euthanasia of the American Veterinary Medical Association. Primary neurons were transfected with plasmids and/or siRNA (obtained from Dharmacon) 4 days after plating using Lipofectamine 2000 (Invitrogen) 21 . SINE compounds KPT-335 and KPT-350 (Karyopharm Therapeutics), Leptomycin B (Cayman Chemicals) or vehicle (DMSO, Sigma) were added 24 h after transfection, immediately following the first imaging run. In experiments measuring subcellular localization of fluorescent proteins, the nucleus was identified using cell-permeable nuclear markers (Hoechst 33342, Sigma).
Immunoblot and Immunocytochemistry. For subcellular fractionation, primary cortical neurons from embryonic day 20-21 rats were prepared as described above, and treated with drug or vehicle 6 days after plating. Cells were harvested 7 h after treatment in resuspension buffer (10 mM Tris (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 ) supplemented with 0.6% Igepal CA-30 (Sigma) to liberate the cytoplasmic fraction, and then in RIPA buffer to lyse the nuclear compartment, following a trypan blue assay for nuclear integrity. Protein content determined by the BCA assay (Pierce) using a Nanodrop 2000. Samples normalized for protein content were run on 15% SDS polyacrylamide gels. Primary antibodies against TDP43 (Barmada et al. 21 ), Hsc70 (1:1000, Santa Cruz SC-7298), GAPDH (1:1000, Millipore MAB374), and Histone H2B (1:250 Novus Biologicals 50347) were detected using IR dye conjugated secondary antibodies (LI-COR Biosciences), imaged on a LI-COR Odyssey, and quantified using ImageStudioLite (LI-COR Biosciences). Plots were generated and analyzed using GraphPad Prism.
For detection of XPO7 by immunocytochemistry, primary cortical neurons from embryonic day 20-21 rats were prepared as described above, then transfected with plasmids encoding EGFP and XPO7 4 days after plating. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton-X-100, and treated with 10 mM glycine before blocking in PBS containing 0.1% Triton, 2% FCS and 3% BSA, for 1 h at room temperature (RT). Detection was accomplished by incubating with anti-XPO7 primary antibodies (diluted 1:500 in blocking solution, (Proteintech, 12980-1-AP) overnight at 4 °C. Samples were rinsed 3 times in PBS, incubated with Cy3 labeled anti-rabbit secondary antibodies (1:250 in blocking solution; Millipore, AP132C) for 1 h at RT, then rinsed 3 more times in PBS containing a 1:1000 dilution of Hoechst 33258 nuclear dye (Invitrogen H3569) prior to imaging. Fluorescence microscopy. Primary cortical neurons and HEK293 cells were imaged using an automated microscopy platform described previously 18,21,38 . Briefly, images were obtained with an inverted Nikon TiE-2000 microscope equipped with a Nikon PerfectFocus 3 system, a high-numerical aperture 20x objective lens, Semrock BrightLine filter sets, and a 16-bit Andor iXon digital camera with an electron multiplied ultra-cooled charge-coupled device. The Lambda XL Xenon lamp provided broad-spectrum illumination via a 5-mm liquid light guide. Stage movements were accomplished using an ASI 2000 stage with rotary encoders in the x-and y-axes. The microscope and associated components were encased in a climate-controlled environmental chamber that we built specifically for this purpose. The illumination, filter wheels, focusing, stage movements, and image acquisitions were fully automated and coordinated with publicly available (ImageJ, µManager) software. Image analysis. Neuronal survival was assessed in an unbiased and blinded manner using novel code written in Python. Twenty-five images per well of a 96 well plate were stitched together to cover a larger surface area, background fluorescence subtracted from the stitched images, and images from each timepoint assembled into a time series or stack. Individual neuronal soma and processes were segmented based on morphological characteristics, and time of death recorded as the last time a neuron was confirmed to be alive (left censoring). Cumulative and instantaneous risk of death plots were calculated from this information using the survival analysis package in R. Statistical comparisons (Cox proportional hazards analysis and the log-rank test) were also accomplished in R.
Regions of interest (ROIs) identified by blinded observers (Figs 5 and 6, S6) or by automated analysis (Figs 7 and 8, S6) corresponding to the cell body and nucleus were used to calculate the mean intensity of fluorescently-labeled proteins within specific subcellular compartments. Expression of EGFP in these experiments was necessary to outline the cell body and draw ROIs around the cytoplasm of transfected cells. For automated analysis, nuclei were identified and partitioned based upon their intensity via Baysian Gaussian Mixture modelling. Cells were removed from the analysis if the nuclear intensity distribution suggested pyknosis, or if the soma-centroid estimate was not contained within the innermost contiguous group of pixels. Data was compiled using R, and scatter plots and graphs depicting changes in protein localization were created using GraphPad Prism.
Verification of siRNA-mediated exporter knockdown. Rat PC12 cells maintained in culture were transiently transfected with 25 nM Dharmacon smartpool siRNA constructs, and harvested 48 hours post transfection, using the Qiagen RNAeasy kit, following the manufacturers protocol. Transcript abundance was determined by quantitative (q)PCR using Power SYBR Green (Applied Biosystems/Thermofisher). 4 ul of a 20 ul cDNA reaction (reverse transcribed from 1 ug of RNA with the Bio-Rad iScript kit) for each sample was used as template with the following primers: Data availability statement. The materials and datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.