TFEB/Mitf links impaired nuclear import to autophagolysosomal dysfunction in C9-ALS

Disrupted nucleocytoplasmic transport (NCT) has been implicated in neurodegenerative disease pathogenesis; however, the mechanisms by which disrupted NCT causes neurodegeneration remain unclear. In a Drosophila screen, we identified ref(2)P/p62, a key regulator of autophagy, as a potent suppressor of neurodegeneration caused by the GGGGCC hexanucleotide repeat expansion (G4C2 HRE) in C9orf72 that causes amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). We found that p62 is increased and forms ubiquitinated aggregates due to decreased autophagic cargo degradation. Immunofluorescence and electron microscopy of Drosophila tissues demonstrate an accumulation of lysosome-like organelles that precedes neurodegeneration. These phenotypes are partially caused by cytoplasmic mislocalization of Mitf/TFEB, a key transcriptional regulator of autophagolysosomal function. Additionally, TFEB is mislocalized and downregulated in human cells expressing GGGGCC repeats and in C9-ALS patient motor cortex. Our data suggest that the C9orf72-HRE impairs Mitf/TFEB nuclear import, thereby disrupting autophagy and exacerbating proteostasis defects in C9-ALS/FTD.


Introduction
A GGGGCC (G4C2) hexanucleotide repeat expansion (HRE) in chromosome nine open reading frame 72 (C9orf72) is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), accounting for up to 40% of cases of familial ALS (DeJesus-Hernandez et al., 2011;ITALSGEN Consortium et al., 2011). ALS and/or FTD caused by mutations in C9orf72 (C9-ALS/FTD) is inherited in an autosomal dominant manner, suggesting that the HRE causes disease through gain-of-function or haploinsufficiency (DeJesus-Hernandez et al., 2011;Ling et al., 2013). Loss of C9orf72 function has been linked to disruption of autophagy and lysosome function, though neurodegeneration is not observed in C9orf72 knockout mice (Liu et al., 2016;Shi et al., 2018;Webster et al., 2016), suggesting that C9-ALS/FTD is primarily caused by toxicity of the HRE. Furthermore, expression of G4C2 repeats causes neurotoxicity in Drosophila and cell culture models of C9-ALS (Goodman et al., 2019a;Kramer et al., 2016;Tran et al., 2015). This toxicity has been proposed to occur through either G4C2 repeat RNA-mediated sequestration of RNA-binding proteins or translation of the G4C2 repeats into dipeptide-repeat proteins (DPRs) through non-canonical repeat-associated non-AUG (RAN) translation (Donnelly et al., 2013;Goodman et al., 2019a;Mori et al., 2013;Tran et al., 2015).
We previously conducted a Drosophila screen of candidate proteins that bound with moderateto-high affinity to G4C2 RNA and identified modulation of the nucleocytoplasmic transport (NCT) pathway as a potent modifier of G4C2 toxicity in both fly and iPS neuron models of C9-ALS (Zhang et al., 2015a), a finding that has also been made by other groups (Freibaum et al., 2015;Jovicˇic´et al., 2015). The mechanisms by which the G4C2 HRE disrupts NCT remain unclear, but potential mechanisms include G4C2 RNA binding to the master NCT regulator RanGAP (Zhang et al., 2015a), DPRs disrupting the nuclear pore complex (Boeynaems et al., 2016;Shi et al., 2017;Zhang et al., 2016), stress granules sequestering NCT factors , or cytoplasmic TDP-43-dependent dysregulation of karyopherin-a (Chou et al., 2018;Gasset-Rosa et al., 2019;Solomon et al., 2018). Recently, a role for NCT disruption in Huntington's disease and Alzheimer's disease has been proposed, indicating that NCT disruption may be a common mechanism in several neurodegenerative diseases (Eftekharzadeh et al., 2018;Gasset-Rosa et al., 2017;Grima et al., 2017). However, the pathways affected by NCT disruption that cause neurodegeneration have not yet been elucidated.
Although autophagy and nucleocytoplasmic transport have both been implicated in neurodegeneration, it is unclear whether or how these two pathways interact in disease pathogenesis (Gao et al., 2017;Thomas et al., 2013). Here, we show that expression of expanded G4C2 repeats is sufficient to disrupt autophagy in Drosophila, leading to an accumulation of p62 and ubiquitinated protein aggregates. We find that autophagolysosomal defects are caused by loss of nuclear localization of the transcription factor Mitf (the Drosophila homolog of TFEB), which regulates transcription of genes involved in autophagolysosome biogenesis (Bouché et al., 2016;Palmieri et al., 2011;Sardiello et al., 2009;Zhang et al., 2015b). Furthermore, suppressing this NCT defect is sufficient to rescue Mitf nuclear localization, restoring autophagy and lysosome function and rescuing neurodegeneration. These findings suggest a pathogenic cascade in C9-ALS/FTD whereby NCT disruption causes a failure of autophagosome biogenesis and lysosome dysfunction that ultimately leads to neuronal death.

Ref(2)P/p62 knockdown suppresses G4C2-mediated neurodegeneration
Expression of 30 G4C2 repeats (30R) in the eye using GMR-Gal4 results in progressive photoreceptor degeneration and visible ommatidial disruption by day 15 ( Figure 1A; Xu et al., 2013;Zhang et al., 2015a). In a genetic modifier screen of over 800 RNAi lines, UAS-ref(2)P RNAi was among the strongest of 32 suppressors of G4C2-mediated eye degeneration (Zhang et al.,    2015a; Figure 1A). ref(2)P is the Drosophila homolog of p62/SQSTM1, and this modifier is of particular interest because SQSTM1 mutations that cause loss of selective autophagy cause ALS/FTD (Cirulli et al., 2015;Goode et al., 2016;Le Ber et al., 2013), and p62 aggregates are pathological features of both familial and sporadic ALS (Al-Sarraj et al., 2011;Cooper-Knock et al., 2012). Knockdown of ref(2)P suppresses eye degeneration, whereas overexpression of ref(2)P enhances this phenotype ( Figure 1A (2)P also rescued eye degeneration in a second G4C2 model expressing 36 G4C2 repeats (36R) (Mizielinska et al., 2014;Figure 1C-D). We next assessed the ability of ref(2)P RNAi to rescue toxicity of G4C2 repeats in motor neurons using the 30R model and a new G4C2 model expressing 44R (Goodman et al., 2019b). As shown in Figure 1E-F, while expression of either 30R or 44R in motor neurons with vGlut-Gal4 leads to paralysis and lethality during pupal development, knockdown of ref(2)P partially rescues this phenotype, whereas overexpression of ref(2)P enhances the pupal lethality observed with 44R expression. These data suggest that ref(2)P is required for G4C2-mediated toxicity during Drosophila development.

GMR-Gal4
To determine whether ref(2)P knockdown is able to suppress age-dependent neurodegeneration, we used a pan-neuronal, inducible 'GeneSwitch' driver (elavGS) in which 30R-expression leads to a marked reduction in climbing ability after 7 days ( Figure 1G). This climbing defect is suppressed with coexpression of ref(2)P RNAi , suggesting that ref(2)P contributes to G4C2-mediated neurotoxicity in the adult nervous system. Since RAN-translation of arginine-containing DPRs have been implicated in G4C2-mediated toxicity in Drosophila (Kwon et al., 2014;Mizielinska et al., 2014), we next tested whether ref(2)P knockdown rescues poly-glycine-arginine (GR) repeat-mediated toxicity. As shown in Figure 1-figure supplement 1D, ref(2)P RNAi partially rescues the severe eye degeneration phenotype caused by poly(GR) 36 expression. Together, these data indicate that ref(2)P, the Drosophila orthologue of p62/SQSTM1, modulates G4C2-mediated neurodegeneration.

G4C2 repeat expression impairs autophagic flux
p62/SQSTM1-positive inclusions are a common pathologic feature seen in brains of C9-ALS/FTD patients where they colocalize with ubiquitin and DPRs (Al-Sarraj et al., 2011). We next investigated the localization of Ref(2)P protein (hereafter referred to as p62) in motor neurons. Expression of 30R leads to the formation of many large p62:GFP puncta in cell bodies compared to controls that strongly colocalize with poly-ubiquitinated proteins (Figure 2A . These data show that G4C2 repeat expression in fly models recapitulates the p62 accumulation with ubiquitinated protein aggregates seen in C9-ALS/FTD patient tissue and iPS neurons (Almeida et al., 2013;Mackenzie et al., 2014).
Increased p62 levels can be due to either increased transcription and/or translation or insufficient protein degradation (Korolchuk et al., 2010). Using qRT-PCR, we find that ref(2)P transcript levels are unchanged in G4C2 repeat-expressing larvae (Figure 2-figure supplement 1F), suggesting that G4C2 repeats cause p62 upregulation by inhibiting p62 degradation. Since p62 is degraded by autophagy and disrupted autophagic flux is known to cause p62 upregulation, we assessed autophagy in G4C2-repeat-expressing flies. We first co-expressed the tagged autophagosome marker mCherry:Autophagy-related 8 (Atg8, the fly orthologue of mammalian Microtubule-associated protein 1A/1B-light chain 3 (LC3)) with 30R in fly motor neurons and found a marked reduction in mCherry:Atg8 autophagic esicles (AVs) when compared to wild-type controls ( Figure 2D

G4C2 repeat expression causes lysosome defects
To further study lysosomal morphology and function, we expressed Lysosome-associated membrane protein 1 (Lamp1) with luminally-tagged GFP in our control and G4C2-expressing flies. Since GFP is largely quenched by the acidity of lysosomes in control animals (Pulipparacharuvil et al., 2005), the accumulation of GFP:Lamp-positive vesicles in 30R-expressing motor neurons suggests a defect in lysosomal acidity or targeting of GFP:Lamp to mature lysosomes ( Figure 2G-H). Furthermore, we observe a marked increase in size and number of late endosomes and lysosomes using genomically tagged Ras-related GTP-binding protein 7, Rab7:YFP, throughout 30R-expressing motor neurons (Figure 2-figure supplement 2A) without alterations in early endosomes labeled with Rab5:YFP (data not shown). Together, these data demonstrate a marked expansion of the late endosome/lysosome compartment in G4C2-expressing neurons.
Though accumulation of p62 and ubiquitinated proteins could be caused by a failure of autophagic vesicles to fuse with the degradative endolysosomal compartment, we did not detect a decrease in mCherry:Atg8+, Rab7:GFP+ amphisomes in G4C2-expressing motor neuron cell bodies ( Mauvezin et al., 2014). GFP is degraded more slowly than the rest of the mCherry:Atg8 protein, leaving a population of free GFP in functioning lysosomes. Free lysosomal GFP is not observed in G4C2-expressing larvae, suggesting an impairment in GFP:mCherry:Atg8 degradation by the lysosome ( Figure 2F). To directly probe lysosome enzymatic activity, we performed Western analysis of Drosophila cathepsin Cp1. Whereas pro-Cp1 is normally cleaved to its mature form by acid hydrolases in lysosomes (Kinser and Dolph, 2012), larvae ubiquitously expressing 30R show an increase in the ratio of pro-Cp1 to Cp1, indicating a decrease in pro-Cp1 cleavage efficiency ( Figure 2I-J).
Together, these data suggest that lysosomes are expanded and dysfunctional in G4C2 repeatexpressing animals.
To investigate whether the autophagic pathway defects precede neurodegeneration in G4C2 repeat-expressing neurons, we performed transmission electron microscopy (TEM) on Drosophila eyes. As GMR-Gal4 is expressed throughout the development of the eye, we chose to perform electroretinograms (ERGs) of fly eyes selectively expressing 30R in photoreceptor neurons (PRs) using Rh1-Gal4, which turns on during adulthood. Rh1 >30R PRs show only a mild reduction of ON transient amplitude at 28 days, but a complete loss of ON and OFF transients and a decrease in ERG amplitude by 56 days (Figure 3-figure supplement 1A-D), indicating a slow and progressive loss of synaptic transmission and impaired phototransduction respectively. These changes also correspond to a marked loss of photoreceptors and synaptic terminals by 54 days which are not observed at 28 days ( Figure 3A-B; Figure 3-figure supplement 1E). We therefore examined autophagic structures by TEM at 28 days, prior to cell loss. Strikingly, we observe a marked increase in the size and number of multilamellar bodies (MLBs) ( Figure 3C-D). MLBs are commonly observed in lysosomal storage diseases and result from a deficiency of lysosomal hydrolases and accumulations of lysosomal lipids and membranes (Hariri et al., 2000;Weaver et al., 2002). Though we did not detect an alteration in the number of autophagosomes, lysosomes, or multivesicular bodies, we did see a significant increase in the number of autolysosomes ( Figure 3C-D). These data suggest that autophagolysosomal function is disrupted in G4C2-expressing photoreceptor neurons at early stages of degeneration.
Given the impairment in autophagic flux, we hypothesized that genetic or pharmacologic manipulations that accelerate autophagy may suppress neurodegenerative phenotypes, whereas those that further impede autophagy would enhance the phenotypes. Indeed, in a candidate-based screen, activation of early steps in the autophagic pathway (e.g. by Atg1 overexpression) suppresses eye degeneration and blocking autophagosome/lysosome fusion (e.g. by Snap29 knockdown) enhances eye degeneration (Supplementary file 1). Similarly, pharmacologic activation of autophagy via inhibition of mTor with rapamycin or mTor-independent activation via trehalose (Sarkar et al., 2007) rescues neurodegenerative phenotypes and p62 accumulation (Figure 2-figure supplement 2G-K). Together, these data show that promoting autophagy or lysosomal fusion are potent suppressors of G4C2-mediated neurodegeneration.

Nucleocytoplasmic transport impairment disrupts autophagic flux
A diverse array of cellular pathways including autophagy, RNA homeostasis, and NCT are implicated in the pathogenesis of ALS and FTD (Balendra and Isaacs, 2018;Evans and Holzbaur, 2019;Gao et al., 2017;Lin et al., 2017;Ling et al., 2013;Ramesh and Pandey, 2017). However, the sequence of events in the pathogenic cascade remains unknown. Cytoplasmic protein aggregates or RNA stress granule formation is sufficient to disrupt nucleocytoplasmic transport (Woerner et al., 2016;Zhang et al., 2018). We therefore tested whether defects in autophagy are upstream, downstream, or in parallel with defects in NCT.
We first tested whether knockdown of ref(2)P rescues the mislocalization of the NCT reporter shuttle-GFP (S-GFP) containing both a nuclear localization sequence (NLS) and nuclear export sequence (NES). G4C2 repeat expression causes mislocalization of S-GFP to the cytoplasm (Zhang et al., 2015a), but knockdown of ref (2)    overexpressing the G4C2 repeats ( Figure 5-figure supplement 1C), suggesting that NCT disruption is sufficient to disrupt autophagic flux in Drosophila motor neurons.

Mitf is mislocalized and inactivated in Drosophila models of C9-ALS/ FTD
Because we observed a reduction in autophagosomes and expansion of lysosome-related organelles, we hypothesized that transcription factors regulating autophagolysosomal function may be mislocalized to the cytoplasm due to disrupted nuclear import. The MiT/TFE family of transcription factors (TFEB, TFE3, MITF, and TFEC) regulates multiple steps of autophagy from autophagosome biogenesis through lysosome acidification via a network of genes called the Coordinated Lysosome Expression And Regulation (CLEAR) network (Settembre et al., 2011). These transcription factors are regulated by localization between the cytoplasm and nucleus . In Drosophila, this conserved transcription factor family is represented by a single homolog called Mitf (Bouché et al., 2016;Zhang et al., 2015b). Mitf knockdown in the nervous system causes lysosomal defects similar to those observed in G4C2-expressing flies (Bouché et al., 2016;Hallsson et al., 2004;Sardiello et al., 2009;Song et al., 2013). Additionally, TFEB levels are reduced in superoxide dismutase 1 (SOD1) mutant cell culture and mouse ALS models (Chen et al., 2015) as well as in ALS and Alzheimer's patient brain tissue . Therefore, we hypothesized that impaired Mitf nucleocytoplasmic transport might underlie the autophagolysosomal phenotypes in fly models of C9-ALS. Indeed, both salivary gland cells and motor neurons expressing 30R show a reduction in percent nuclear Mitf ( Figure 4A-D). To assess whether disrupted Mitf NCT alters CLEAR gene expression in adult heads, we expressed 30R using a ubiquitous inducible driver, daughterless-Gen-eSwitch (daGS). In control flies, a mild (~1.75 fold) overexpression of Mitf mRNA resulted in a significant upregulation of 3 of the 7 Mitf targets tested (the vesicular ATPase (v-ATPase) subunits Vha16-1, Vha68-2, and Vha44) and a trend towards upregulation of 4 others ( Figure 4E). Importantly, coexpression of 30R with daGS >Mitf led to a similar~2 fold increase in Mitf transcripts but did not induce Mitf target genes ( Figure 4E). This lack of Mitf target induction in 30R flies suggests that decreased nuclear import of Mitf suppresses the ability of 30R-expressing flies to upregulate CLEAR genes in order to maintain or induce autophagic flux.
We next examined whether rescue of nucleocytoplasmic transport defects in 30R-expressing animals can rescue Mitf nuclear import and autophagolysosomal defects. Exportin-1 has recently been demonstrated to regulate Mitf/TFEB nuclear export Silvestrini et al., 2018). Knockdown of exportin-1 (Drosophila emb) rescues G4C2-mediated cytoplasmic Mitf mislocalization in the salivary gland ( Figure 5A-B) and GFP:Lamp accumulation in motor neurons ( Figure 5C-D). Importantly, emb knockdown increases the total number of autophagosomes in G4C2-expressing motor neuron cell bodies by~3 fold ( Figure 5E-F), suggesting that nuclear retention of Mitf rescues autophagolysosomal defects. However, emb knockdown caused a slight elevation of p62:GFP puncta intensity in controls and did not rescue the accumulations of p62:GFP in 30R-expressing motor neurons ( Figure 5G-H). Together, these data indicate that autophagolysosomal dysfunction in 30Rexpressing animals occurs downstream of nucleocytoplasmic transport disruption, whereas inhibition of nuclear export is not sufficient to rescue p62 accumulation.

Mitf rescues G4C2 repeat-mediated degeneration
Since Mitf mislocalization contributes to autophagolysosome defects in a fly C9-ALS model, we hypothesized that increasing total levels of Mitf might compensate for impaired nuclear import. While high level Mitf overexpression is toxic in Drosophila (Hallsson et al., 2004), a genomic duplication construct containing the Mitf gene lacking the DNA repetitive intron 1 (Mitf Dp) (Zhang et al., 2015b), is sufficient to partially rescue 30R-mediated eye degeneration, while Mitf knockdown enhances eye degeneration ( Figure 6A-B). Furthermore, pupal lethality caused by 30R expression in motor neurons and climbing impairment in elavGS >30R flies are also partially rescued by Mitf Dp (Figure 6C-D). In contrast, Mitf Dp did not rescue the severe rough eye phenotype observed with GMR-Gal4 overexpression of poly(GR) 36 ( Figure 6-figure supplement 1A-B), suggesting that Mitf Dp rescues toxicity caused by the G4C2 repeat RNA rather than the DPRs alone.
To determine whether increased levels of Mitf rescue G4C2-mediated neurodegeneration through effects on the autophagolysosomal pathway, we examined GFP:Lamp and p62:GFP expression in   30R-expressing motor neurons. Indeed, Mitf Dp rescues increased GFP:Lamp1 expression ( Figure 6E-F) and reduces p62:GFP accumulation in motor neurons of vGlut >30R larvae ( Figure 6G-H). Thus, increasing Mitf levels in multiple neuronal subtypes in Drosophila suppresses G4C2-mediated neurotoxicity, consistent with our hypothesis that loss of nuclear Mitf is a key contributor to G4C2-mediated neurodegeneration. If the impaired lysosomal function we observe in our Drosophila model is contributing to neurodegeneration downstream of NCT defects, we would predict that genetic upregulation of key regulators of lysosome function may suppress degenerative phenotypes. Indeed, overexpression of Rab7, the small GTPase required for fusion of autophagosomes with lysosomes, or Trpml, a lysosomal calcium channel, suppress eye degeneration ( Figure 6-figure supplement 1C-D). Furthermore, overexpression of key lysosomal v-ATPase subunits whose expression is regulated by Mitf also suppresses neurodegeneration in the Drosophila eye, while RNAi-mediated knockdown enhances degeneration ( Figure 6-figure supplement 1C-D). Interestingly, loss of the ALS-associated gene ubqn in Drosophila was also rescued by increase in key lysosomal v-ATPase subunits or by nanoparticle mediated lysosome acidification (Ş entürk et al., 2019). Overexpression of these Mitf-regulated genes also showed partial rescue of pupal lethality in animals expressing 30R in motor neurons Figure 8. A proposed model of GGGGCC repeat expansion pathogenesis. G4C2 repeat expansion causes nucleocytoplasmic transport disruption through multiple proposed mechanisms including G4C2 RNA binding of RanGAP and stress granule recruitment of nucleocytoplasmic transport machinery. Transport disruption leads to a blockage in the translocation of autophagy-mediating transcription factors such as Mitf/TFEB to the nucleus in response to proteotoxic stress. Failure to induce autophagic flux leads to autophagy pathway disruption such as the accumulation of large, nondegradative lysosomes and MLBs. Loss of autophagic flux leads to accumulation of Ref(2)P/ p62 and ubiquitinated protein aggregates, leading to chronic protein stress signaling and eventually neuronal cell death.
( Figure 6-figure supplement 1E). These findings suggest a model whereby downregulation or cytoplasmic retention of Mitf targets leads to lysosomal disruption in G4C2-repeat-expressing flies.

Nuclear TFEB is reduced in human cells and motor cortex with GGGGCC repeat expansions
In humans, TFEB is the homolog of Drosophila Mitf that is best characterized for its role in autophagy and has been implicated in neurodegenerative disease (Cortes and La Spada, 2019;Martini-Stoica et al., 2016). Interestingly, a previous study showed nuclear TFEB was selectively depleted in the motor cortex of a sample of five ALS patients compared to five controls . To test the relevance of our findings in Drosophila models to human disease, we next examined whether G4C2 repeat expression impairs nuclear import of TFEB in HeLa cells stably expressing TFEB:GFP (Roczniak-Ferguson et al., 2012) using a 47-repeat (47R) G4C2 construct that expresses tagged DPRs (see Materials and methods). In control cells, TFEB:GFP is predominantly localized to the cytoplasm, whereas induction of autophagy by 3 hr starvation leads to robust nuclear translocation of TFEB ( Figure 7A-B). In contrast, while 47R-expressing cells have a mild basal elevation of nuclear TFEB, the nuclear translocation of TFEB in response to starvation is significantly impaired relative to control cells ( Figure 7A-B). We then tested the effect of expression of DPRs produced by alternate codons (i.e. in the absence of G4C2 repeats): poly-glycine-alanine (poly-GA 50 ), poly-glycine-arginine (poly-GR 50 ), and poly-proline-arginine (poly-PR 50 ) (Figure 7-figure supplement 1A -B). While poly-GA 50 causes a mild decrease in TFEB nuclear translocation, expression of poly-GR 50 or poly-PR 50 does not disrupt TFEB:GFP nuclear translocation. These data suggest that human cells expressing an expanded G4C2 repeat, but not DPRs, are unable to efficiently import TFEB into the nucleus in response to stimuli.
To further investigate the relevance of loss of TFEB nuclear import to C9-ALS patients, we obtained human motor cortex samples from four non-neurological controls and four C9-ALS patients (Supplementary file 2). These samples were fractionated into cytoplasmic and nuclear-enriched fractions and assayed for TFEB using Western analysis. TFEB is reduced by an average of 76% in the nuclear fraction and by about 50% in the cytoplasm in C9-ALS compared to controls ( Figure 7C-D, Figure 7-figure supplement 1C). These data suggest that TFEB protein is downregulated in C9-ALS/FTD motor cortex, but the greatest depletion occurs in the nucleus. Therefore, we propose a model whereby disruption of protein nuclear import by the C9orf72-HRE results in a failure of Mitf/ TFEB to translocate to the nucleus to regulate the autophagic response to protein stress (Figure 8).

Discussion
Our work has revealed that the ALS-associated G4C2 hexanucleotide repeat is sufficient to disrupt multiple aspects of autophagy. In Drosophila, G4C2 repeats cause loss of autophagosomes and disrupt lysosomal structure and function. This accumulation of autolysosomes and lysosome-related organelles (MLBs) has been observed in lysosomal storage disorders and has been reported in spinal cord tissue from sporadic ALS patients (Bharadwaj et al., 2016;Parkinson-Lawrence et al., 2010;Sasaki, 2011). Regulation of protein and lipid homeostasis by the lysosome may be particularly important in neurons since they are post-mitotic and have high energy demands (Fraldi et al., 2016). Loss of function of C9orf72 also disrupts autophagy and lysosomal function in multiple cell types (Farg et al., 2014;Ji et al., 2017;O'Rourke et al., 2015;Sellier et al., 2016;Shi et al., 2018;Sullivan et al., 2016;Ugolino et al., 2016;Webster et al., 2016;Yang et al., 2016;Zhu et al., 2020), suggesting a mechanism whereby G4C2 repeats may have synergistically detrimental effects with haploinsufficient C9orf72 in C9-ALS/FTD patients. Additionally, multiple forms of familial ALS are caused by mutations in genes in autophagy and lysosome function (Evans and Holzbaur, 2019;Lin et al., 2017;Ramesh and Pandey, 2017), and upregulation of lysosome function has been proposed to be beneficial in multiple preclinical models of ALS (Donde et al., 2020;Mao et al., 2019;Ş entürk et al., 2019;Shi et al., 2018). Thus, our findings suggest that, as has been shown in other forms of ALS, neurotoxicity of G4C2 repeats in C9 ALS-FTD is at least partially caused by disrupted autophagolysosomal function.
The finding that ref(2)P knockdown prevents or delays G4C2-mediated neurodegeneration is surprising, as p62/SQSTM1 is thought to link toxic ubiquitinated aggregates to LC3 to remove aggregates via selective autophagy (Cipolat Mis et al., 2016;Levine and Kroemer, 2008;Saitoh et al., 2015). However, other studies have also suggested that p62 may contribute to (rather than ameliorate) toxicity of ubiquitinated proteins. For example, Atg7 -/mice display severe defects in autophagy and accumulation of p62-positive protein aggregates in the liver and brain, and knockout of p62 in these mice prevents the formation of ubiquitinated aggregates and rescues liver dysfunction via suppression of chronic oxidative stress signaling (Komatsu et al., 2007). Additionally, Ataxia Telangiectasia Mutated-mediated DNA double stranded break repair is impaired in cultured neurons expressing the C9orf72-HRE, and this phenotype is rescued by p62 knockdown (Walker et al., 2017). These findings suggest that increases in p62 may contribute to DNA damage previously described in C9-ALS. Further, p62 is found to co-localize with DPRs in C9-ALS patients (Al-Sarraj et al., 2011;Mackenzie et al., 2014;Mori et al., 2013) and may promote protein aggregation. We hypothesize that p62-positive aggregate or oligomer formation in C9-patients contributes to neurotoxicity by activating downstream signaling pathways that are alleviated by autophagymediated clearance.
While many groups have reported nucleocytoplasmic transport dysfunction in ALS, it has remained unclear how NCT disruption causes ALS. Stress granules can recruit nuclear pore proteins to the cytoplasm and cause nucleocytoplasmic transport defects, suggesting that the disruptions in phase separation of RNA-binding proteins may lie upstream of nucleocytoplasmic transport defects . Recently, Ortega et al. discovered that hyperactivity of nonsense-mediated decay may lie downstream of nucleocytoplasmic transport, indicating that multiple proteostasis pathways may be disrupted (Ortega et al., 2020). Additionally, selective autophagy is required for nuclear pore turnover (Lee et al., 2020), implying that autophagy defects may contribute to the cytoplasmic nuclear pore pathology found in C9-ALS patients and animal models. Our data show that in Drosophila, HeLa cells, and human tissue, nucleocytoplasmic transport defects lead to an inability to activate TFEB translocation to the nucleus, causing widespread autophagy defects and accumulation of protein aggregates (Figure 8). Interestingly, genetic inhibition of nuclear export or increase in Mitf expression are able to strongly rescue autophagosome and lysosome phenotypes and neurodegeneration, but do not result in complete clearance of p62 accumulations (Figures 5-6). Additional studies will be needed to better understand the relationship between p62 accumulation, autophagy, nucleocytoplasmic transport, and neurodegeneration. Overall, these findings place nucleocytoplasmic transport defects in ALS upstream of proteostasis defects.
Importantly, TFEB has been previously proposed as a therapeutic target in ALS and other neurodegenerative disease (Cortes and La Spada, 2019). Upregulation of TFEB signaling helps clear multiple types of proteotoxic aggregates found in Alzheimer's disease, Parkinson's disease, Huntington's disease, ALS and FTD (Decressac et al., 2013;Parr et al., 2012;Polito et al., 2014;Torra et al., 2018;Vodicka et al., 2016). Our study suggests that modulation of TFEB nucleocytoplasmic transport may be an additional therapeutic target, and that targeting both nucleocytoplasmic transport and autophagy may act synergistically in ALS and FTD.

Drosophila genetics
Drosophila were raised on standard cornmeal-molasses food at 25˚C. For eye degeneration, GMR-GAL4, UAS-30R/CyO, twi-GAL4, UAS-GFP were crossed to UAS-modifier lines or background controls and GMR-GAL4, UAS-30R/UAS-modifier or GMR-GAL4, UAS-30R/+ were selected (where UASmodifier can be on any chromosome) from the offspring and aged at 25˚C for 15 days. Eye degeneration is quantified using a previously described method (Ritson et al., 2010). Briefly, points were added if there was complete loss of interommatidial bristles, necrotic patches, retinal collapse, loss of ommatidial structure, and/or depigmentation of the eye. Eye images were obtained using a Nikon SMZ 1500 Microscope and Infinity 3 Luminera Camera with Image Pro Insight 9.1 software.
For the climbing assay, UAS-30R; elavGS were crossed to experimental or genetic background controls. Adults were transferred 3-5 days after eclosion to vials containing 200 mM RU486 food or ethanol vehicle alone and transferred to new vials every 2-3 days. After aging 7-10 days, groups of 10 flies were placed into empty food vials and were tapped to the bottom and then locomotor function assessed by their negative geotaxis (flies reflexively crawl against gravity) response as measured by ability to climb 8 cm in 10 seconds. Each cohort of 10 flies was tested 10 times to obtain an average. N represents individual cohorts of 10 flies.

Drosophila drug feeding
Cornmeal-molasses-yeast fly food was melted and then cooled for 5 min before being mixed with concentrations of mifepristone (RU486), rapamycin, or trehalose and cooled to room temperature. Ethanol or DMSO was used as a vehicle control. Parent flies were crossed on normal food, and then they were transferred to food containing drug every 2-3 days such that their offspring would develop in food containing drug or adult offspring were transferred to drug food once eclosed as noted. Wandering third-instar larvae were selected for immunostaining or western blot analysis. Adult flies were aged on the drug-containing food for 15 days before analyzing their eye morphology or assessed for climbing ability on the day noted.

Quantitative RT-PCR
For each genotype, mRNA was collected from 5 flies or 30 heads using the TRIzol reagent following the manufacturer's protocol. Reverse transcription was performed using SuperScript III First-Strand synthesis kit following the manufacturer's protocol. Quantitative PCR was performed using SYBR Green PCR system on a 7900HT fast Real-Time PCR system (Applied Biosystem). The primers for G4C2 repeats were designed to amplify a 3' region immediately after the repeats in the UAS construct.

Immunofluorescence staining and imaging
For Drosophila ventral nerve cords, wandering third-instar larvae were dissected in HL3 (Stewart et al., 1994) using a standard larval fillet dissection then fixed in 4% paraformaldehyde (or Bouin's fixative for UAS-mCherry:Atg8 experiments) (Sigma) for 20 min, followed by wash and penetration with PBS 0.1% Triton X-100 (PBX) for 3 Â 20 min washes. The tissues were blocked for 1 hr at room temperature in PBS with 5% normal goat serum (NGS) and 0.1% PBX, then stained with primary antibodies at 4C overnight (16 hr). Tissues were washed three times for 20 min each with 0.1% PBX. Secondary antibodies (Goat antibodies conjugated to Alexa Fluor 568, 488, 633) diluted in 0.1% PBX 5% NGS and incubated for 2 hr and then washed three times for 20 min each with 0.1% PBX. During one wash, DAPI was added to the prep at a final concentration of 1 mg/mL. Larvae were mounted in Fluoromount-G (Invitrogen).
Drosophila salivary glands were dissected using a standard protocol and stained as above excepting for stronger solubilization with 0.3% PBX. Fixed cells or tissues were analyzed under an LSM780 or LSM800 confocal microscope (Carl Zeiss) with their accompanying software using Plan Apochromat 63 Â, NA 1.4 DIC or Plan Apochromat 40Â, 1.3 Oil DIC objectives (Carl Zeiss) at room temperature. Images were captured by an AxioCam HRc camera (Carl Zeiss) and were processed using ImageJ/Fiji. To quantify fluorescent intensities, after opening the images in ImageJ/Fiji, certain areas/bands were circled and the intensities were measured. Puncta were counted using the Analyze Particles function in Image J using the same thresholding across experiments. Images are representative and experiments were repeated two to five times.

Western blotting
Tissues or cells were homogenized and/or lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, and 1% Triton X-100) supplemented with protease inhibitor cocktail (Complete, Roche) using microcentrifuge pestles, and then were incubated in RIPA buffer on ice for 20 min. Samples were spun down at 100 g for 5 min to remove carcass and unbroken cells. For protein quantification, solution was diluted and measured by BCA assay (Thermo Fischer Scientific).
For nucleocytoplasmic fractionation of autopsy tissue, fractionation was performed with the NE-PER Nuclear and Cytoplasmic Extraction Kit according to the manufacturer's protocol. For detection of proteins in the whole fraction, Drosophila larvae were solubilized in 8M urea. For the soluble and pelleted fraction, larvae were first solubilized in RIPA buffer as described above. The samples were spun down at 15000 rpm for 20 min and the soluble supernatant was set aside. Freshly prepared 8M urea buffer (Sigma) was added to the pellet and dissolved through vortexing. Samples were spun again at 15000 rpm for 20 min and urea-soluble pellet fraction was collected. A small amount of sample buffer dye was added and urea-buffered protein samples were run immediately on SDS-PAGE without heating. For immunoblot, 10-50 mg of total protein sample was mixed with 4x Laemmli buffer (Bio-Rad) and heated at 98˚C for 10 min. The protein samples were run on 4-15% SDS Mini-PROTEAN TGX Precast Gels (Bio-Rad) and transferred to nitrocellulose membrane. TBST (50 mM Tris-HCl pH 7.4, 1% Triton X-100) with 5% non-fat milk (Bio-Rad) was used for blocking.

Electroretinogram (ERG) Assay
For ERG recordings, Rh1-GAL4/UAS-LacZ and Rh1-GAL4/UAS-30R flies were aged at 25˚C in 12 hr light/12 hr dark cycle. ERG recordings were performed as described (Ş entürk et al., 2019). In brief, adult flies were immobilized on a glass slide by glue. A reference electrode was inserted in the thorax and a recording electrode was placed on the eye surface. Flies were maintained in the darkness for at least 2 min prior to 1 s flashes of white light pulses (LED source with daylight filter), during which retinal responses were recorded and analyzed using WinWCP (University of Strathclyde, Glasgow, Scotland) software. At least five flies were examined for each genotype and timepoint.

Transmission Electron Microscopy (TEM)
Rh1-GAL4/UAS-lacZ and Rh1-GAL4/UAS-30R flies were aged at 25˚C in 12 hr light/12 hr dark cycle. Retinae of adult flies were processed for TEM imaging as previously described (Chouhan et al., 2016). Three flies were examined for each genotype and timepoint.
TFEB:GFP HeLa cell culture, transfection, and immunofluorescence analysis HeLa cell line with stable expressing TFEB:GFP was a gift from Dr. Shawn Ferguson at Yale University. Hela cells were grown in DMEM media (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone Laboratories Inc). The cell line was authenticated by observing nuclear translocation of TFEB:GFP in the presence of starvation (Figure 7). Absence of mycoplasma contamination was confirmed by staining with DAPI. Transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Briefly, 1-2 mg of cDNA was diluted into 100 ml of Opti-MEM I Medium (Invitrogen) and mixed gently. Lipofectamine 2000 mixture was prepared by diluting 2-4 ml of Lipofectamine 2000 in 100 ml of Opti-MEM I Medium. The ratio of DNA to Lipofectamine 2000 used for transfection was 1:two as indicated in the manual. The DNA-Lipofectamine 2000 mixture was mixed gently and incubated for 20 min at room temperature. Cells were directly added to the 200 ml of DNA-Lipofectamine 2000 mixture. After 48 hr, transfected HeLa cells were treated with EBSS medium for 3 hr for starvation. HeLa cells were fixed with 4% PFA at room temperature for 15 min, washed three times with PBS, permeabilized for 10 min with 1% PBTX, washed another three times with PBS, and blocked for 1 hr at room temperature with 10% normal goat serum (Sigma) diluted in 0.1% PBTX. Cells were then incubated overnight at 4˚C with primary antibody mouse anti-Flag antibody. After three washes in PBS (5 min each), cells were incubated for 1 hr at room temperature with secondary antibodies (goat anti-Alexa Fluor 568) diluted in the blocking solution. Cells were washed three times in PBS and mounted with Prolong Gold anti-fade reagent with DAPI (Cell Signaling).

Collection of human autopsied tissue
Human autopsied tissue used for these data are described in detail in Supplementary file 2. The use of human tissue and associated decedents' demographic information was approved by the Johns Hopkins University Institutional Review Board and ethics committee (HIPAA Form five exemption, Application 11-02-10-01RD) and from the Ravitz Laboratory (UCSD) through the Target ALS Consortium.

Statistics
All quantitative data were derived from independent experiments. Each n value representing biological replicates is indicated in the figure legends. Statistical tests were performed in Prism version 8.3.1 or Microsoft Excel 16.34 and were performed as marked in the figure legends. All statistical tests were two-sided. Results were deemed significant when the P value a = 0.05. No statistical methods were used to predetermine sample size. The investigators were not blinded during experiments.

Additional information
Competing interests Hugo J Bellen: Reviewing editor, eLife. The other authors declare that no competing interests exist. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. . Supplementary file 1. Candidate Screen of autophagy-related genes. Flies expressing 30 G4C2 repeats in the eye under control of GMR-GAL4 were crossed to the indicated UAS line and scored for enhancement (<0) or suppression (>0) as described (Zhang et al., 2015a).

Funding
. Supplementary file 2. Demographics of human patients. Motor cortex from postmortem brain autopsies from four C9-ALS patients and four non-neurological disease controls were used is this study (Figure 7). Patient ID, cause of death (diagnosis), gender, age at death, post-mortem interval (PMI) in hours, and presence of C9orf72 expanded hexanucleotide repeat are indicated.
. Transparent reporting form

Data availability
All data generated or analysed during this study are included in the manuscript and supporting files.

Sequencebased reagent
ref (2) Continued on next page