Nup62 is recruited to pathological condensates and promotes TDP-43 insolubility in C9orf72 and sporadic ALS/FTLD.


 Amyotrophic lateral sclerosis (ALS) and Frontotemporal Lobar Degeneration (FTLD) share clinical, neuropathological, and genetic features. This includes common genetic disease-causing mutations such as the expanded G4C2 repeat in the C9orf72 gene (C9-ALS/FTLD) and cytoplasmic and insoluble protein depositions of the TDP-43 in degenerating regions of the brain and spinal cord. Proposed mechanisms of toxicity in C9-ALS/FTLD are the production of repeat expansion transcripts and their dipeptide repeat proteins (DPRs) products which are hypothesized to drive nucleocytoplasmic transport defects. The nuclear pore complex (NPC) regulates nucleocytoplasmic trafficking by creating a selectivity and permeability barrier comprised of phenylalanine glycine nucleoporins (FG nups). However, the relationship between FG nups and TDP-43 pathology remains elusive. Here, we define two mechanisms through which TDP-43 promotes Nup62 nuclear depletion and cytoplasmic in C9-ALS/FTLD and sALS/FTLD. In C9-ALS/FTLD, poly-GR initiates the formation of TDP-43 containing stress granules (SGs) that trigger the nuclear loss and recruitment of Nup62 in vitro and in vivo. When colocalized, cytoplasmic TDP-43:Nup62 assemblies mature into insoluble inclusions through an interaction within the TDP-43 nuclear localization sequence (NLS) suggesting Nup62 promotes deleterious phase transitions. Absent of poly-GR, aberrant TDP-43 phase transitions in the cytoplasm recruits and mislocalizes Nup62 into pathological inclusions. The result of these cytoplasmic Nup62 and TDP-43 interactions are pathological and insoluble TDP-43:Nup62 assemblies that are observed in C9-ALS/FTLD and sALS/FTLD CNS tissue.


Introduction
ALS and FTLD are fatal neurodegenerative disorders that share neuropathological features and causative mutations and are considered to exist along a disease spectrum 1,2 . TDP-43 and FUS are predominantly nuclear RNA-binding proteins (RBPs) that cycle between the nucleus and cytoplasm and primarily regulate RNA metabolism 3 . However, these proteins mislocalize to the cytoplasm and form inclusions in the vast majority of ALS patients and up to half of FTLD patients [4][5][6][7][8][9] . In addition to shared neuropathology, the most common genetic cause of both ALS and FTLD is an expanded G4C2 hexanucleotide repeat sequence in the first intron of the C9orf72 gene (C9-ALS/FTLD) [10][11][12][13] . The disease-causing pathobiology underlying this mutation includes various consequences of the expanded allele: C9orf72 haploinsufficiency, the deposition of toxic repetitive RNAs, and accumulation of dipeptide repeats (DPRs) 10,[14][15][16][17][18][19][20] .
While the causative mechanism driving C9-ALS/FTLD pathobiology remains unclear, recent work suggests that C9orf72 haploinsufficiency alone is not sufficient to initiate motor or cognitive phenotypes in rodent models 21 . However, complete loss of both alleles causes mild motor impairment as shown by reduced activity on the open-field test [22][23][24] . In contrast, expression of an expanded G4C2 sequence drives neurotoxicity in vitro 25,26 and cognitive phenotypes in vivo in AAV and BAC transgenic rodent models 23,27-30 . Further, C9orf72 protein knockdown in two G4C2 repeat expansion mouse models caused a synergistic increase in cognitive defects and neurotoxicity to support the convergence of loss and gain of function mechanisms of toxicity 31,32 .
Nucleocytoplasmic transport refers to the trafficking of proteins and RNAs across the nuclear membrane through the NPC 67 . The NPC is a large multi-subunit protein complex comprised of approximately 30 different protein subunits, known as nucleoporins or nups 68 . Nups serve a variety of functions but their most well-defined role involves creating the NPC permeability and selectivity barrier [69][70][71] . Molecules smaller than 40 kDa can freely diffuse across the NPC 72 .
Facilitated nucleocytoplasmic transport of larger molecules is driven by a gradient of Ran:GTP [73][74][75][76][77] , and larger molecules are escorted through the NPC by nuclear transport receptors, or karyopherins, which traverse the pore by interacting with FG nucleoporins (FG nups) [78][79][80][81][82] . FG nups make up approximately one-third of the NPC nucleoporins and contain protein domains that are enriched in phenylalanine (F)-glycine (G) residues, amino acid residues associated with structural disorder and flexibility 83,84 . This enrichment of phenylalanine and glycine creates intrinsically disordered regions (IDRs) within FG nups that, in turn, contribute to the NPC permeability barrier by forming a hydrogel-like structure through liquid-liquid phase separation 85,86 .
The impact of G4C2 repeat expansion expression on nucleocytoplasmic trafficking is well documented in a variety of model systems. Initial studies employing RNAi and chromosomal deletion genetic screens to identify modifiers of UAS-G4C2-30 repeat and UAS-G4C2-58 repeat toxicity in Drosophila retinal neurons (GMR-GAL4 driven expression) revealed several genes of the nucleocytoplasmic transport pathway and NPC, such as FG nups, as potent modifiers of toxicity 59,63 . Furthermore, karyopherin overexpression was identified as a strong suppressor of PR50 toxicity in a yeast genetic screen 62 . Similarly, downregulation of several karyopherins and Ran gradient regulators enhanced degenerative eye phenotype in a PR25-expressing Drosophila model 61 and GR50-expressing Drosophila 53 . In addition to genetic screens using invertebrate model systems, impaired nucleocytoplasmic trafficking dynamics are well documented in C9-ALS/FTLD iPSC-derived neurons and regulators of transport were subsequently shown to exhibit abnormal staining in post-mortem tissue 59,87,88 . Non-FG nup irregularities were also observed in C9-ALS patient post-mortem motor cortex samples 59,88 . The underlying hypothesis regarding C9-ALS/FTLD and nucleocytoplasmic trafficking is that perturbations promote the mislocalization of TDP-43 or FUS proteins and while there is mounting evidence of nucleocytoplasmic trafficking impairment, the mechanisms linking FG Nups to cellular dysfunction and neuropathology remain undefined.
Here we examined the relationship between C9-ALS/FTLD pathobiology and FG nups using multiple in vitro and in vivo model systems. In this study, we show that FG nups are genetic modifiers of neurotoxicity in C9-ALS/FTLD Drosophila models and that Nucleoporin p62 (Nup62), the most abundant FG nup within the NPC that contains extensive FG-repeat tracks, is a potent enhancer of expanded GR toxicity. Cellular GR accumulation disrupts the nuclear localization of Nup62 and promotes its cytoplasmic accumulation in vitro and in vivo. Furthermore, cytoplasmic poly-GR accumulation induces the formation of TDP-43 containing SGs that recruit Nup62 protein and cytoplasmic TDP-43 inclusions, formed through aberrant liquid-liquid phase separation, also sequester Nup62 proteins independent of SGs. Cytoplasmic Nup62 and TDP-43 interactions promote soluble-to-insoluble transition of these protein assemblies that is dependent on the TDP-43 nuclear localization sequence (NLS). Analyses of post-mortem tissue validates phosphorylated TDP-43:Nup62 in C9-ALS/FTLD and sporadic ALS/FTLD cases. Collectively, this study defines two mechanisms by which cytoplasmic TDP-43 sequesters Nup62 and promotes the formation of pathological inclusions: GR-induced stress granules and phase separated cytoplasmic TDP-43 condensates.

Drosophila studies
Drosophila stocks: Fly stocks and crosses were maintained on standard cornmeal medium in light/dark controlled incubators. RNAi flies were obtained from VDRC or the Transgenic RNAi project 89 via Bloomington DGRC. The Nup62 overexpression fly was generated by the BestGene Inc methology we used previously [90][91][92] . UAS-(G4C2)30 was kindly shared with our lab by Peng Jin 34 and UAS-(G4C2)36 and UAS-GR36 flies were generous gifts from Adrian Isaac's lab 48 Nup RNAi screen: UAS-RNAi virgin females were crossed with recombinant GMR-GAL4/UAS-(G4C2)30 males or control GMR-GAL4/UAS-EGFP males at 28°C. Female progenies of the appropriate genotype were collected and their eyes were imaged with a Leica M205C digital camera at 0-1 days post-eclosion, with at least 15 eyes imaged per tested RNAi. Images of external eye phenotype were then scored as previously described 93 .
Nup62 RNAi and G4C2 toxicity characterization: GMR-GAL4/UAS-(G4C2)36 males were crossed with Nup62 RNAi virgin females at 25°C and imaged/quantified. Eclosion defects were noted in these flies, therefore we quantified the number of eclosed and uneclosed flies each evening per 24 h period, across at least 5 days of three independent sets of crosses. GMR-GAL4/UAS-(G4C2)36 males were crossed with Nup62 OE virgin females and external eyes were imaged and quantified as described above. Recombinant GMR-GAL4/UAS Nup62 RNAi and GMR-GAL4/UAS-Nup62 overexpression lines were generated and crossed with previously described UAS-GR36 94 or an UAS-EGFP control 54

Induced pluripotent stem cell maintenance and motor neuron differentiation
Induced pluripotent stem cells (iPSCs) lines studied are described in Supplemental Table 2. iPSCs were maintained in mTeSR1 medium and cells exhibiting characteristics of spontaneous differentiation were removed prior to initiating differentiation protocol. Motor neuron differentiation was conducted as previously described 65,95 . Briefly, iPSC colonies were dissociated into a single cells suspension and plated at approximately 1,000,00 cells per well on a 6-well plate. iPSCs were then differentiated towards a motor neuron phenotype over the course of two stages. Both stages of iPSC differentiation consist of daily media changes and supplementation of N2B27 base media (50% DMEM F12, 50% Neurobasal, 1x NEAA, 1x Glutamax, 1x N2, 1x B27). The first stage of neuroectoderm induction (6 days) occurs once the plated iPSCs reach 90% confluency are   treated with N2B27 base media supplemented with 10 µM SB431542, 100  supplemented with 0.2 µg/mL Ascorbic Acid, 10 ng/mL BDNF, 10 ng/mL GDNF, 10 ng/mL CNTF.

Immunohistochemistry staining and human tissue analysis
Human tissue from control, sporadic and C9orf72 ALS cases was obtained from the University of Pittsburgh Neuropathology Department Brain Bank and cases are described in Supplemental Table 3. Paraffin embedded tissue sections from the cervical spinal cord, hippocampus, and cortex were stained with Nup62 (BD Biosciences, 610497, 1:100), TDP43 1D3 (Millipore Sigma, Cat. No. MABN14) and/or FUS according as previously described 96 . Fluorescent images were captured with 60x objective on Nikon A1R confocal microscope. Nup62 secondary antibody was 488/FITC while phospho-TDP43 and FUS were labeled with 594/TRITC/Cy3 to ensure any localized signal detected was not due to bleed through between channels.

Poly-GR mouse studies
All procedures involving mice were approved by the Institutional Animal Care and Use Committee at Thomas Jefferson University. A knock-in F.A.S.T. cassette 97 at the ROSA26 locus 98,99 under the ROSA26 promoter 98,100 was used for successful integration of FLAG-GR50-GFP or FLAG-GFP. GR50 was encoded by a randomized codon sequence to allow for production of the protein product absent of repeat-rich RNA. Animals were generated at Ingenious Targeting Laboratory, with successful knock-in confirmed via PCR from tail DNA samples. To elicit GR50/GFP expression, mice were crossed with CAG-Cre to allow for excision of STOP codon. CAG-Cre mice were a generous gift from Dr. Yuichi Obata, Riken BioResource Center, Japan. All mice are on a C57BL/6 background.
Mice were anesthetized and perfused by transcardial puncture with PBS and chilled 4% paraformaldehyde (PFA). Spinal cord was immediately dissected and post-fixed in 4% PFA for 24 hours, phosphate buffer for 24 hours, and then 30% sucrose solution for a minimum of 48 hours until proper cryoprotection was ensured (ie. sample no longer floated). Spinal cord was embedded in O.C.T. Compound (Sakura, 4583) embedding medium. Samples were sectioned serially in the transverse orientation at a thickness of 30 µm and collected on glass slides. Slides were stored at -20°C until analysis. Frozen sections were processed for immunofluorescent confocal microscopy as previously described 101 with modifications. Briefly, spinal cord sections were rinsed with PBS prior to a 20-minute permeabilization with 0.5% T-X100 in HMK buffer (20 mM Hepes, pH 7.5, 1 mM MgCl2, 100 mM KCl). Sections were blocked in 10% normal donkey serum, 1% BSA in HMK buffer for 30 minutes. Sections were incubated with the following primary antibodies for approximately 18 hours at 4°C with gentle rotating: anti-nucleoporin p62 (BD Biosciences 610497, ms 1:400); anti-GFP (Millipore Sigma AB16901, chk 1:2500); and anti-NeuN (Cell Signaling Technology 24307, rb 1:200). Following 3x rinse in HMK buffer, sections were incubated with the following secondary antibodies at 1:500 each: Alexa Fluor 647 goat-anti-mouse (Thermo Fisher A32728), Alexa Fluor 546 donkey-anti-rabbit (Thermo Fisher A10040), and Alexa Fluor 488 goat-anti-chicken (Thermo Fisher A11039) for 1.5 h in HMK buffer with 1% BSA.
Sections were rinsed 4x for 10 minutes each in HMK buffer at room temperature with gentle rotating, followed by 1x rinse in water and mounted in VectaShield mounting medium with DAPI (Vector). Imaging and quantification were done on a Nikon A1R-SI confocal microscope using NIS Elements software.

Microscopy
Image acquisition of fixed samples were acquired on a Nikon A1 laser-scanning confocal system with 40X and/or 60X oil immersion objectives (CFI Plan Fluor 40X Oil; CFI Plan Apo Lambda 60X Oil, Nikon) or Nikon N-SIM super-resolution Microscope with 60X oil immersion objectives (Plan Apo TIRF 60X Oil) and Hamamatsu C11440 Orca Flash 4.0 camera. 3D SIM images were reconstructed and then deconvolved prior to analysis. Image analysis was conducted in NIS-Elements AR Analysis 4.51.
Live-cell imaging All live-cell imaging was performed on Nikon A1 laser scanning confocal microscope outfitted with Tokai HIT stage-top incubator while utilizing 40x oil immersion objective.
Stage-top incubator was allowed to equilibrate to 37 °C and 5% CO2 for 10 min prior to imaging.
HEK293 cells were transfected for 6 prior to imaging session. Images were acquired every 5 min for 15 h.
Fluorescence Recovery After Photobleaching (FRAP) Imaging FRAP studies were conduct as previously described 96 . Briefly, a 60x oil immersion objective on confocal microscope was used to monitor condensates. ROI was drawn over structure of interest and reference ROI was included in an adjacent, non-bleached cell. Four to five baseline structure images were obtained and then structure was bleached for 500 ms using 50% laser power (488nm or 594 nm laser lines).
Structures were observed for 120 seconds. Data represents the fluorescence signal recovery of 11-20 structures.

Nuclear Integrity Scoring and Analysis
Following Nup62 immunostaining and imaging by structured illumination microscopy (SIM), nuclear Nup62 integrity and continuity were measured. Nuclear Nup62 integrity was scored by a blinded, unbiased observer. Nuclei that exhibited a fragmented or irregular pattern were given a lower nuclear integrity score (described in Supplemental Table 4). Furthermore, using ImageJ analysis software, we straightened the nuclear Nup62 signal and representative examples are shown above graphs (Fig 7). The profile plot was then used to measure Nup62 signal across the length (in pixels) of the select Nup62 staining. The profile plots were normalized to maximum signal intensity to account for any variability in staining intensity and also normalized to the length of measured signal to account for nuclear size variability. The area under the curve (AUC) for these profile plots was then calculated and averaged for each group (n=7-10 nuclei).

Statistical Analysis
Experimental data was collected, and outliers determined by ROUT's outlier test (Q=1%).
Following removal of outliers, data sets are shown as the mean and standard error of the mean. Statistically significant differences between experimental groups were calculated by GraphPad Prism software (Version 7) and deemed significant when p ≤ 0.05. Statistically significant differences were determined by unpaired Student's T-test when comparing two variables or oneway ANOVA with Dunnett or Tukey's multiple comparisons test when comparing multiple.
Statistical analysis of nuclear Nup62 levels in control and C9orf72 ALS iPSC neurons was conducted by two-tailed Mann-Whitney test.

Results
Nup62 is a modifier of expanded G4C2 toxicity in Drosophila FG nups comprise the NPC selectivity and permeability barrier through liquid-liquid phase separation of the FG repeat domains to form a hydrogel within the NPC central channel 86,103,104 and these are shown to exhibit neuropathological abnormalities in ALS tissue [105][106][107] . However, the specific relationship between FG nups and the expanded G4C2 allele in C9-ALS/FTLD is unknown. To address this, we performed a genetic screen in the (G4C2)30 expressing Drosophila to determine whether FG nup depletion modulates neurotoxicity associated with the rough eye phenotype (Fig 1A) 34, 94 . The GMR-Gal4 system was used to drive (G4C2)30 and shRNA expression in the fly eye. Reduction of mammalian conserved FG nups through RNAi was validated through RT-qPCR (Supplemental Fig 1) and progeny rough eye phenotype was scored in a blinded manner 93  Downregulation of specific FG nups that increased the rough eye phenotype degeneration score were classified as enhancers while those that mitigated the rough eye phenotype degeneration were suppressors ( Fig 1A). Nup98 RNAi in (G4C2)30 flies does not alter the rough eye phenotype ( Fig 1B-C). Nup54, Nup58, or Nup153 loss suppresses the rough eye phenotype and Nup62 knockdown significantly enhances (G4C2)30 mediated eye degeneration (Fig 1B-C). Importantly, RNAi mediated knockdown of FG nups alone does not alter eye phenotype in control (UAS-eGFP) Drosophila (Supplemental Fig 1 and Fig 1B top row). This suggests that any RNAi-mediated changes were specific to a genetic interaction with expression of the (G4C2)30 transgene. These data show that specific FG nups genetically interact with the expanded G4C2 repeat and that Nup62 may play a role in disease pathogenesis as it was a potent enhancer of eye neurodegeneration when lost. Cellular DPR accumulation is a critical event in C9-ALS/FTLD associated neurotoxicity 1 .
Notably, DPRs are detected at low levels and after degeneration in the (G4C2)30 repeat fly eye 34,59 . Therefore, to validate our initial screen and determine whether DPRs might contribute to the genetic interaction seen in the (G4C2)30 Drosophila, we also tested if Nup62 similarly enhanced the phenotype of the (G4C2)36 Drosophila model which was previously shown to produce an abundance of DPRs (poly-GR and poly-GP) 94 . Surprisingly, Nup62 loss in GMR-Gal4 (G4C2)36 Drosophila causes robust pupal lethality and eclosion defects not observed with Nup62 knockdown alone (Fig 1D-E). Together, these data suggest that a genetic interaction likely exists between Nup62 and repeat RNA but the elevated DPR burden exacerbates the resulting Drosophila phenotype.

Glycine-arginine disrupts Nup62 localization in vitro and in vivo
Irregularities in nucleoporin immunostaining were previously observed in several C9-ALS/FTLD models but the direct relationship between late stage pathology (DPRs and TDP-43 inclusions) on FG nups has not been examined in depth 59,63,88 . Our genetic screens indicate that the (G4C2)36 Drosophila model, which accumulates DPRs, exhibits a higher sensitivity to Nup62 loss than lines with lower DPR production (Fig 1), suggesting cellular dysfunction involving Nup62 and DPRs. To validate these findings in human cells, we first assessed if DPRs disrupt nuclear Nup62 localization in HEK293 cells by expressing mCherry-tagged poly-DPR constructs expressing 50 repeats of GR, PR, GA, PA, or GP 54 . Immunostaining of Nup62 and quantitative analyses of confocal projection images reveal that PR50, GA50, PA50, and GP50 do not alter nuclear Nup62 signal. However, GR50 significantly reduces nuclear Nup62 intensity by 20.17% after 24 h as compared to the mCherry control (Fig 2A, Supplemental Fig 2A). Furthermore, increasing concentration of GR50-plasmid DNA transfected into the cells yields a greater reduction in nuclear Nup62 signal intensity after 24 h (Fig 2B-C) indicating a dose-dependent cellular GR50 burden and nuclear Nup62 loss. Since nuclear envelope disruption is a common event during programmed cell death 108 and Poly-GR accumulation is cytotoxic in human cell lines and rodent models 54,55,94,109-111 , we quantified LDH release to determine whether nuclear Nup62 signal loss was due to GR50-mediated programmed cell death and no measurable difference was observed in DPR50 expressing cells at the time of fixation indicating nuclear Nup62 depletion is associated with increasing cellular burden of GR (Supplemental Fig 2B). Nup62 is a short-live nucleoporin and undergoes continual turnover 112,113 , therefore, we assessed whether GR-mediated nuclear Nup62 deficits were attributed to impairment at the transcriptional level through RT-qPCR.
Consistent with previous studies in C9-ALS iPSC derived neurons 88 , there was no significant difference in Nup62 mRNA levels in eGFP-and GR50-transfected cells after 24 h of episomal expression indicating Nup62 defects are not due to transcription disruption (Supplemental Fig   2C). We next examined whether GR50 promotes the cytoplasmic mislocalization of Nup62 protein.
GR50-eGFP was expressed in HEK293 cells and nuclear/cytoplasmic Nup62 signals were quantified following immunostaining and confocal imaging. GR50-eGFP expression causes a modest but significant reduction in the nuclear/cytoplasmic ratio of Nup62 protein, indicating an enhanced relative cytoplasmic Nup62 signal (Fig 2D & E). Consistent with this, GR50 expression produced cytoplasmic Nup62 puncta that appeared more frequently ( Nup62 undergoes frequent turnover in both mitotic and post-mitotic neurons [113][114][115][116] , therefore, to ensure our findings were not limited to dividing cells, we tested whether nuclear  Table 2) were assessed at 89 days postdifferentiation 65 and immunostained for Nup62 and MAP2 ( Fig 2H). Quantification of maximum intensity projection confocal images revealed a 19.1% reduction in nuclear Nup62 ( Fig 2I) and a 279% enrichment of cytoplasmic Nup62 ( 3.07 puncta/MAP2 + neuron). Notably, no significant nuclear Nup62 defects were observed in less mature iPSC neurons (28 day differentiation) from two separate C9orf72 ALS/FTLD iPSC lines and their respective isogenic controls (Supplemental Fig 2D & E). This effect is likely to be dependent upon progressive GR accumulation reported to occur over the course of 2 months in C9-ALS/FTLD iPSC neurons 56 and coincides with DPR burden driving phenotypes in Drosophila models with higher DPR proteins (Fig 1). Together, these data suggest cellular GR deposition causes Nup62 mislocalization from the nucleus to the cytoplasm.
To determine if GR disruption of Nup62 in vitro occurs in vivo, we quantified Nup62 localization in a newly generated transgenic poly-GR mouse model (Fig 3A-B). A GR50-eGFP transgene was introduced by a Flexible Accelerated Stop Tetracycline Operator (F.A.S.T.) Cassette and driven by a ROSA26 promoter in C57BL/6 mice ( Fig 3A). Immunostaining for GR50-eGFP reveals its expression in tissue sections of lumbar spinal cord SMI32 + neurons ( Fig 3B).
Analyses of Nup62 localization in the GR50-eGFP mouse model via immunostaining showed nuclear depletion and cytoplasmic accumulation of Nup62 protein (relative to eGFP control) in NeuN + neurons of the lumbar spinal cord in 12-month-old adult animals ( Fig 3C). Total cytoplasmic Nup62 droplet surface area was significantly higher in GR50-eGFP mice than controls ( Fig 3D). These in vivo data are consistent with the in vitro findings indicating that cellular GR accumulation promotes the nuclear loss and cytoplasmic enrichment of Nup62 protein.

Elevated Nup62 rescues C9orf72 Drosophila
Reduced cellular Nup62 enhances the neurotoxicity associated with expression of the G4C2-repeat expansion (Fig 1), therefore, we tested whether Nup62 overexpression is sufficient to rescue the fly eye degeneration phenotype in (G4C2)36 Drosophila. We first generated a UAS-Nup62 expression Drosophila line which exhibited increased Nup62 mRNA by RT-qPCR (Supplemental Fig 3). Nup62 Drosophila were crossed with the GMR-Gal4 (G4C2)36 repeat expansion fly and rough eye phenotypes were then scored as a measure of neurotoxicity in the adult progeny. Nup62 expression completely abolishes the rough eye phenotype in (G4C2)36 expressing Drosophila (Fig 4A & B), suggesting that increased Nup62 levels prevent the cytotoxic effects of expanded (G4C2)36 expression in vivo. Since we observed Nup62 protein redistribution upon GR50 cellular expression in vitro and in vivo, we next tested whether elevating Nup62 mitigates phenotypes of GR-expressing Drosophila. The GMR-Gal4xGR36 Drosophila model shows a robust phenotype with reduced survival and significant retinal degeneration 94 . Similarly, we observed the GMR-Gal4xGR36 fly has extensive deterioration of eye size and an absence of ommatidial organization ( Fig 4C). When crossed with the Nup62 Drosophila, there was a modest reduction in eye deterioration (Fig 4C), though this did not rescue the rough phenotype as we observed with the (G4C2)36 Drosophila (Fig. 4A). These data provide in vivo evidence that C9-ALS/FTLD neurotoxicity may be attributed, in part, to Nup62 redistribution due to GR deposition, and that Nup62 expression, likely restores its localization to mitigates some of these neurotoxic effects in C9orf72 Drosophila models.

Poly-GR disrupts Nup62 and TDP-43 localization through stress granules
The cytoplasmic mislocalization and phosphorylation (pTDP-43) of TDP-43 is a pathological hallmark observed in the majority of ALS/FTLD patients 4,8 . Recent neuropathological studies also show that while ~4% of TDP-43 inclusions contain GR in C9orf72 ALS/FTLD postmortem tissue, most poly-GR accumulations colocalize with pTDP43 117 . Therefore, we next tested whether TDP-43 is altered by poly-GR expression. HEK293 cells expressing GR50-eGFP resulted in the cytoplasmic mislocalization of endogenous TDP-43 as observed by immunostaining and nuclear/cytoplasmic ratio analyses (Supplemental Fig 4A & B). Interestingly, GR50-eGFPexpressing cells formed droplet-like cytoplasmic condensates that colocalized with endogenous cytoplasmic TDP-43 condensates, with an average surface area of 4.90 µm 2 (Fig 5A). Analyses of these cytoplasmic poly-GR and TDP-43 structures revealed high colocalization with a Pearson's coefficient of 0.7233 (p-value: 0.0015; Fig 5A, inset). Consistent with this, orthogonal renderings showed that cytoplasmic GR50-eGFP and TDP-43 exist together within the same three-dimensional space (Fig 5A-B). Notably, this indicates that cytoplasmic GR depositions are sufficient to promote the formation of endogenous TDP-43 cytoplasmic condensates. This is supported by recent findings that show poly-GR drives aberrant phase-separation of purified or overexpressed TDP-43 118 and these data indicate that endogenous TDP-43 similarly interacts with poly-GR inclusions in vitro.
Multiple cellular pathways are thought to contribute to aberrant TDP-43 phase separation and resulting pathological inclusions. Altered dynamics in membraneless organelles formed through liquid-liquid phase separation, such as SGs, are hypothesized to promote TDP-43 proteinopathy. This is supported by work showing that chronic SGs induction promotes cytoplasmic phosphorylated TDP-43 species 119 . Recent work has also identified SG-independent mechanisms that promote aberrant TDP-43 phase transitions, aggregation, and nuclear loss 96,120,121 . Therefore, we tested whether GR-mediated sequestration of endogenous TDP-43 is associated with SG formation by immunostaining for G3BP1 and Ataxin-2 SG markers. Notably, cytoplasmic GR condensates do colocalize with G3BP1 and Ataxin-2 ( Fig 5B, Supplemental Fig   4C) suggesting GR expression induces the formation of SG-like structures. GR50-induction of Ataxin-2 assemblies is concentration-dependent, requiring high cellular levels of GR50-eGFP to form SGs (Supplemental Fig 4C). Analysis of G3BP1 and TDP-43 in the accumulated structures revealed TDP-43 surface area is significantly less than that of G3BP1 (Fig 5C & D). Intensity profile plot analysis of Figure 5C to understand protein localization within the signal indicates that TDP-43 is localized to the GR and G3BP1 condensates (Fig 5E and Supplemental Fig 4D-E).
We next characterized whether GR similarly disrupts nuclear Nup62 due to aberrant SG formation and found Nup62 to also be sequestered into these cytoplasmic GR-induced SG-like structures along with TDP-43 and Ataxin-2 ( Fig 5F, Supplemental Fig 4C, F). Notably, Nup62 consistently colocalized with SGs containing endogenous TDP-43. Profile plot analysis of cytoplasmic GR structures reveals colocalization of Nup62 within GR50-eGFP:TDP-43 accumulations (Fig 5G,   Supplemental Fig 4G-H). Together, this indicates that the cytoplasmic poly-GR burden directly correlates with the formation of SG-like structures in cells that recruit endogenous TDP-43 and disrupt Nup62 localization in C9orf72 ALS/FTD.

Cytoplasmic Nup62:TDP-43 interactions promote insolubility
To test the consequence of Nup62:TDP-43 cytoplasmic interactions on protein dynamics, as observed in GR-induced SGs, we generated a fluorescent mRuby-Nup62 expression plasmid that forms cytoplasmic Nup62 droplets when expressed in HEK293 cells. HEK293 cells were cotransfected with mRuby-Nup62 and eGFP-TDP43 (wild type) and 6 h after transfection, cells were monitored by longitudinal imaging for 15 h. This revealed that cytoplasmic mRuby-Nup62 condensates form with and without eGFP-TDP43 colocalization ( Fig 6A). Furthermore, Nup62-mRuby forms structures that exhibit characteristics with different dynamics. Nup62-mRuby signals that did not colocalize with cytoplasmic eGFP-TDP43 exhibited characteristics of liquid-liquid phase separation in that they were dynamic, fused, and dissipated over the course of minutes ( Fig 6A; "reversible"). However, when colocalized with eGFP-TDP43, mRuby-Nup62 condensates were static and formed nonspherical structures that remained present throughout the duration of the 15 h imaging session (Fig 6A, "irreversible"). We then characterized the mRuby-Nup62 signal throughout the longitudinal imaging session and found the "irreversible" structures were larger ( Fig 6B), more frequently colocalized with eGFP-TDP43 (Fig 6C), and exhibited reduced circularity ( Fig 6D) compared to reversible mRuby-Nup62 condensates. These features suggest that cytoplasmic Nup62 forms phase-separated liquid protein droplets that, when colocalized with cytoplasmic TDP-43, exhibit altered dynamics characteristic of gel-like or solid-state assemblies.
Consistent with this, FRAP analysis of mRuby-Nup62:eGFP-TDP43 colocalized signals revealed cytoplasmic eGFP-TDP43 (middle row) and mRuby-Nup62 (bottom row) signals do not recover after photobleaching (Fig 6E and F)  HEK293 cells co-expressing mRuby-Nup62 and eGFP-TDP43 ΔNLS (Fig 6G) showed an absence of colocalization between the two reporter proteins implicating the TDP-43 NLS is likely required for Nup62 this interaction potentially mediated by KPNB1. Together, these data show that cytoplasmic Nup62 and TDP-43 interactions, driven by the TDP-43 NLS, promote the soluble to insoluble phase of these proteins.

Cytoplasmic TDP-43 inclusions recruit Nup62 independent of stress granules
Here, we sought to test whether cytoplasmic TDP-43 assemblies similarly relocalize Nup62 absent of GR and SG induction. To achieve this, we utilized the optoTDP43 system that mimics key pathological hallmarks of the protein aggregates observed in ALS/FTLD in vitro and in vivo 91,96,122 . The optoTDP43 construct is comprised of Cry2olig-TDP43-mCherry and photokinetic oligomerization of optoTDP43 induces its aberrant liquid-liquid phase separation upon blue-light illumination resulting in homotypic self-interactions initiated by the C-terminal low complexity domain and later through the unbound RNA-binding domains 96 . These assemblies undergo soluble-to-insoluble maturation into aggregates through aberrant liquid-liquid phase transitions independent of SGs 96 . HEK293 cells were transfected with an optoTDP43 or Cry2olig-mCherry (Control) construct and exposed to blue-light for 18 h (Fig 7A). Nup62 was detected by immunofluorescent staining and N-SIM super-resolution microscopy and nuclear Nup62 signal was quantified by a blinded unbiased observer using a scoring criterion described in Supplemental Table 4. The presence of optoTDP43 inclusions correlates with a reduced nuclear Nup62 integrity score, indicating increased nuclear fragmentation and structural irregularities, as compared to Cry2olig-mCherry control cells exposed to blue-light (Fig 7B & C). Nuclear Nup62 integrity scores were complemented by profile plot analysis on nuclear Nup62 signal from Control (Cry2olig-mCherry) and OptoTDP43 cells exposed to blue-light (Fig 7D-F). Representative profile plots and corresponding images for each group in Figure 7D & E show that cells with optoTDP43 inclusions exhibited significantly reduced AUC compared to controls confirming reduced nuclear Nup62 localization (Fig 7F). Similar to GR-induced SGs, we observed optoTDP43 inclusions promoted the formation of cytoplasmic Nup62 droplets that colocalized with optoTDP43 by N-SIM super-resolution microscopy (Fig 7G, box). OptoTDP43 also colocalized with the FG nup, Nup153 (Supplemental Fig 6). We next validated if FG nups colocalize to optoTDP43 inclusions in motor neurons in vivo using the OK371-Gal4 optoTDP43 Drosophila 91 and the pan-FG Nup MAb414 antibody and found that the MAb414 antibody colocalized with optoTDP43 inclusions in the ventral nerve cord of third instar larvae (Fig 7H, arrow). Interestingly, while the majority of larger optoTDP43 inclusions colocalized with FG nup staining in vivo, some smaller inclusions do not, suggesting cytoplasmic TDP-43 sequestration of Nup62 likely occurs as soluble cytoplasmic condensates as they mature into more insoluble inclusions through aberrant liquid-liquid phase separation (Fig 7H, asterisks). Together, these data suggest that aberrant phase separation of cytoplasmic TDP-43 condensates and pathological insoluble inclusions promote Nup62 mislocalization independent of SGs and GR deposition.

Nup62 and TDP-43 colocalize in ALS/FTLD postmortem CNS tissue.
Neuropathological analysis was then performed to determine whether Nup62 colocalized with TDP-43 proteinopathy in ALS/FTLD postmortem spinal cord, hippocampus, and cortex by immunohistochemistry (Fig 8, Supplemental Table 3). We observe a clear, uniform nuclear Nup62 signal around DAPI nuclear signal in cells without phosphorylated TDP-43 inclusions in ALS patient spinal cord (Fig 8A, asterisk). In contrast, cells with cytoplasmic and phosphorylated TDP-43 deposits exhibit a disrupted nuclear Nup62 staining (Fig 8A, arrowhead), consistent with earlier in vitro findings (Fig 2G). Interestingly, Nup62 was frequently observed in phospho-TDP-43 + inclusions in the spinal cord, cortex, and hippocampus from both C9 ALS/FTLD and sALS/FTLD patients (Fig 8B-D). This analysis of C9orf72 ALS/FTLD and sporadic ALS/FTLD suggests TDP-43 is linked with nuclear Nup62 mislocalization both in vitro and in vivo.
Although phospho-TDP-43 pathology is present in nearly all ALS cases and nearly half of FTLD cases, another RBP, FUS, is mutated in rare forms of ALS and also forms pathological inclusions in a fraction of sporadic FTLD patients 4,5,8 . Therefore, we tested whether cytoplasmic accumulation of Nup62 is specific to TDP-43 pathology or if it is also observed in proteinopathies comprised of other RBPs, such as FUS. Immunostaining for Nup62 and FUS in two different FUS-FTLD cases did not show detectable Nup62-FUS colocalization in the hippocampus (Fig 8E). This suggests that disruption of nuclear Nup62 and its sequestration into cytoplasmic neuropathological protein inclusions is specific to the aberrant phase transitions of cytoplasmic TDP-43 in C9orf72 and sporadic ALS/FTLD.

Discussion
This study defines the pathobiological mechanisms that drive FG nup mislocalization and promote TDP-43 proteinopathy in C9orf72 and sporadic ALS/FTLD. Nucleocytoplasmic transport dysfunction is frequently characterized in C9orf72 ALS/FTLD model systems [59][60][61][62]65,123  FG nups as modifiers of the rough eye phenotype. Specifically, Nup54, Nup58, and Nup153 were suppressors, while Nup62 was an enhancer of retinal neuron degeneration (Fig 1B & C). Since Nup62 is the most potent enhancer of toxicity in the (G4C2)30 Drosophila, we validated this finding in a second C9orf72 Drosophila that expresses a slightly longer (G4C2)36 repeat and found that Nup62 loss results in a potent eclosion phenotype that does not occur with Nup62 RNAi alone (Fig 1 D & E). We propose that this difference in phenotypes in response to Nup62 downregulation is likely due to DPR mediated toxicity since the 36-repeat fly model produces higher levels of DPRs prior to degeneration compared to the 30-repeat fly model 59,94 . The role of FG nups in C9-ALS/FTLD pathobiology is supported by studies in other model systems that identified some FG nups as modifiers of toxicity in (G4C2)58 and (PR)25 Drosophila models 61,63 , and in genetic screens of (PR)50and (GR)100-expressing yeast 62,126 .
To address if DPRs contribute to the identified genetic interactions observed in the (G4C2)36 Drosophila, we tested whether DPRs alter Nup62 protein in human cells. Interestingly, nuclear Nup62 localization is specifically perturbed and redistributed to the cytoplasm in response to GR expression (Fig 2D-F). Cytoplasmic Nup62 assemblies are also highly enriched (~3 fold) in C9-ALS/FTLD patient-derived iPSC neurons compared to those derived from controls ( Fig 2G   & I). This is also observed in vivo in a transgenic poly-GR expressing mouse model and expression of a codon-optimized (GR)50 DPR results in loss of nuclear Nup62 signal and enrichment of cytoplasmic Nup62 droplets in lumbar spinal motor neurons (Fig. 3). This suggests poly-GR accumulation drives Nup62 mislocalization and its aberrant redistribution into cytoplasmic droplets and we hypothesize this contributes to enhanced eye phenotype in G4C2 expressing Drosophila upon Nup62 knockdown, since it would further reduce already low nuclear Nup62. Supportive of this, elevating Nup62 completely rescues the rough eye phenotype in (G4C2)36 Drosophila and mitigates severe eye degeneration in a (GR)36 Drosophila model (Fig.   4). Taken together, cytoplasmic GR accumulation promotes cytoplasmic mislocalization of Nup62 in vitro and in vivo. Notably, this is not the first study to implicate Nup62 in ALS since previous work found that its downregulation enhanced toxicity associated with poly-PR expression in Drosophila 61 and progressive nuclear depletion of Nup62 was also characterized in the anterior horn of SOD1 G93A mice 106 . Similarly, some fALS (SOD1 A4V) and sALS patient tissue were also found to exhibit nuclear Nup62 irregularities [105][106][107]127 .
Characterization of cytoplasmic GR droplets indicate that they colocalize with endogenous TDP-43 (Fig. 5A) and G3BP1, a protein required for the assembly of SGs [128][129][130] . This finding supports work indicating that nuclear or GR promotes the spontaneous formation of SGs with abnormal dynamics, some of which contain TDP-43 53 . Notably, the previous study did not observe cellular colocalization of GR with G3BP1 or TDP-43 proteins but did identify direct interactions between GR:G3BP1 through co-IP analyses 53 . We found cytoplasmic GR droplets colocalize with endogenous G3BP1 and TDP-43 and this correlates with a reduction in the TDP-43 nuclear/cytoplasmic ratio in a dose-dependent manner (Fig. 5B-D, S4B-C). GR and TDP-43 signal overlap with G3BP1 with reduced TDP-43 surface area compared to G3BP1 suggesting TDP-43 lies within the SG (Fig. 5). Furthermore, cytoplasmic GR:TDP-43 assemblies colocalize with endogenous Nup62 and SG markers ( Previous studies similarly show that poly-GR colocalizes with cytoplasmic pTDP-43 inclusions in patient tissue 117 , suggesting a pathological interaction between these two proteins. AAV mediated GFP-(GR)200 expression in rodents also resulted in cytoplasmic structures that colocalized with some nucleoporins, SG markers, and TDP-43 proteins in vivo 118 .
One hypothesized pathway driving deleterious TDP-43 phase separation and subsequent aggregation are through aberrant SG dynamics since chronic formation does promote TDP-43 hyperphosphorylation 119 . Notably, Nup62 was not previously reported to localize to sorbitol-and sodium arsenite-induced SGs 131 as reported with GR-induced SGs here. However, recent work in which FUS fibrils were used to induce cytoplasmic demixing of cytoplasmic TDP-43 did reveal some colocalization with Nup62 over time 120  Overall, these data suggest that cytoplasmic Nup62 droplets exhibit characteristics of proteins that undergo liquid-liquid phase separation, and it is likely that these interaction through the classical NLS promotes deleterious phase transition of cytoplasmic TDP-43 causing it to mature into insoluble inclusions.

Thus far, this work indicates that the cytoplasmic deposition of GR in C9-ALS/FTLD
induces TDP-43-and Nup62-containing SGs and that cytoplasmic interaction between mislocalized Nup62 and TDP-43 promotes their insolubility. While this highlights upstream pathways that may drive FG nup defects in C9-ALS/FTLD, it does not address whether Nup62 plays a role in TDP-43 pathology absent GR in sporadic ALS/FTLD. The absence of RNA binding allows for the aberrant liquid-liquid phase separation that mature into pathological TDP-43 inclusions 96,120 . This process can be modeled using the optoTDP43 protein to mimic end stage disease in the absence of upstream disease causing event, such as SGs or GR accumulation 91,96,122 . Therefore, we tested if aberrant TDP-43 phase transitions in the cytoplasm similarly mislocalized Nup62 as observed with poly-GR accumulation and observed that optoTDP43 do cause fragmentation of nuclear Nup62 signal (Fig. 7 A-F). Additionally, cytoplasmic optoTDP43 inclusions, which absent of SGs, colocalize with endogenous Nup62 in vitro and in vivo (Fig. 7G-H). This suggests that cytoplasmic TDP-43 phase separation, absent of GR or SGs, promote nuclear depletion and cytoplasmic deposition of Nup62 and is consistent with work in which Nup62 is can colocalize with the C-terminal TDP-43 fragment in vitro 134 . Notably, we did observe optoTDP43 inclusions absent Nup62 protein in the optoTDP43 Drosophila, raising the possibility that cytoplasmic TDP-43 phase separation may precede Nup62 sequestration (Fig 7H).
To validate physiological relevance of these findings and the relationship between TDP- Thus, these findings suggest that NPC integrity is compromised through sequestration of Nup62 into cytoplasmic TDP-43 condensates. This process is mediated by GR-induced stress granules with perturbed dynamics in C9-ALS/FTLD or through cytoplasmic RNA deficient TDP-43 condensates in sporadic disease. Future work is required to define Nup62:TDP-43 interactions and the consequence of cytoplasmic Nup62 mislocalization on specific aspects of nucleocytoplasmic transport.

Acknowledgments
We appreciated the generosity of Akiko Takedo
Representative maximum intensity projection images are shown. Nuclear Nup62 quantification is shown in Supplemental Fig 2A. B) HEK293 cells were transfected with increasing amounts of GR50-eGFP plasmid DNA and immunostained for Nup62. Image were processed by automatic deconvolution in Nikon Elements.
We observe that TDP-43 is significantly smaller than G3BP1 in these GR50-eGFP structures.        Statistically signi cant differences in y eye degeneration was determined by one-way ANOVA with Tukey's multiple comparison's test: ** p ≤ 0.01; **** p ≤ 0.0001 vs control. Statistically signi cant differences in eclosion frequency was determined by one-way ANOVA with Dunnett's multiple comparison's test: **** p ≤ 0.0001 vs UAS-EGFP x (G4C2)36 repeat expansion y.

Figure 2
Poly-GR alters Nup62 localization in vitro. A) Nuclear Nup62 (white) levels were assessed in HEK293 cells expressing mCherry-tagged poly-DPR constructs (red) by immuno uorescent staining and confocal microscopy. Representative maximum intensity projection images are shown. Nuclear Nup62 quanti cation is shown in Supplemental Fig 2A. B) HEK293 cells were transfected with increasing amounts of GR50-eGFP plasmid DNA and immunostained for Nup62. Image were processed by automatic deconvolution in Nikon Elements. Single slice images (0.2 μm) of Nup62 (white) show dosedependent reduction in nuclear Nup62 with increasing GR50-eGFP plasmid DNA. Scale bar: 20 μm C) Nuclear Nup62 signal was quanti ed in maximum intensity projection confocal images following immuno uorescent staining (n=47-93 cells/group). Data corresponds to images presented in Fig 2B. D) To assess whether GR50-eGFP alters nuclear Nup62 localization, HEK293 cells were transfected with eGFP (Control) or GR50-eGFP (green) and immunostained for Nup62 (white). Representative confocal images show nuclear Nup62 depletion that coincides with cytoplasmic Nup62 puncta accumulation. Average signal is shown by graph bars while dots and squares represent the signal in individual neurons. The control group consists of two separate iPSC lines and C9orf72 is the combination of three C9-ALS iPSC lines. Statistical signi cance in Fig 2C was determined by one-way ANOVA with Tukey post hoc analysis (Fig 2C) or two-tailed, unpaired students t-test (Fig 2E, F