TDP-43, a protein central to amyotrophic lateral sclerosis, is destabilized by tankyrase-1 and -2

In >95% of amyotrophic lateral sclerosis (ALS) and ~45% of frontotemporal degeneration (FTD), the RNA/DNA-binding protein TDP-43 is cleared from the nucleus and abnormally accumulates in the cytoplasm of affected brain cells. Although the cellular triggers of disease pathology remain enigmatic, mounting evidence implicates the poly(ADP-ribose) polymerases (PARPs) in TDP-43 neurotoxicity. Here we show that inhibition of the PARP enzymes Tankyrase 1 and Tankyrase 2 (referred to as Tnks-1/2) protect primary rodent neurons from TDP-43-associated neurotoxicity. We demonstrate that Tnks-1/2 interacts with TDP-43 via a newly defined Tankyrase-binding domain. Upon investigating the functional effect, we find that interaction with Tnks-1/2 inhibits the ubiquitination and proteasomal turnover of TDP-43, leading to its stabilization. We further show that proteasomal turnover of TDP-43 occurs preferentially in the nucleus; our data indicate that Tnks-1/2 stabilizes TDP-43 by promoting cytoplasmic accumulation, which sequesters the protein from nuclear proteasome degradation. Thus, Tnks-1/2 activity modulates TDP-43 and is a potential therapeutic target in diseases associated with TDP-43, such as ALS and FTD. Given the role the role we sought to determine whether Tnks-1/2 promotes ubiquitination and degradation of TDP-43. We demonstrate that a highly selective inhibitor of Tnks-1/2 activity mitigates the neurotoxicity of TDP-43 to rodent neurons. We uncover that TDP-43 has a functional Tankyrase-binding motif, however our data show that TDP-43 is not degraded by Tnks-1/2-dependent ubiquitination. By contrast, our results suggest that Tnks-1/2 stabilizes TDP-43 and that this may occur by inhibiting degradation of TDP-43 by the nuclear proteasome. These findings provide molecular and cellular insight into the interaction between Tnks-1/2 and TDP-43 and provide a foundation for developing novel therapeutic strategies for TDP-43-associated disease. density of 5.6x10 5 cells. Cells were grown in DMEM with high glucose and L-glutamine (ThermoFisher Scientific, Waltham, MA, USA) and 10% FBS (Sigma Aldrich, St. Louis, MI, USA) with penicillin-streptomycin (ThermoFisher Scientific, Waltham, MA USA) in 6-well plates overnight. Cells were treated with 100 µg/mL Cycloheximide (Sigma Aldrich, St. Louis, MI, USA) in in DMEM with high glucose and L-glutamine (ThermoFisher Scientific, Waltham, MA, USA) and 10% FBS (Sigma Aldrich, St. Louis, MI, USA) with penicillin-streptomycin (ThermoFisher Scientific, Waltham, MA, USA) and lysed (see paragraph below)16 hr after drug treatment.

Recently, we discovered in Drosophila melanogaster that reduction of the Tnks-1/2 homologue mitigates the neurotoxicity of TDP-43, while upregulation exacerbates TDP-43-associated toxicity (McGurk et al., 2018a). Furthermore, we observed that downregulation of the Tnks-1/2 homologue led to an increase in nuclear TDP-43 and a decrease in cytoplasmic TDP-43 in Drosophila neurons (McGurk et al., 2018a). Given the role of Tnks-1/2 in protein degradation and the role of aberrant protein degradation in ALS/FTD, we sought to determine whether Tnks-1/2 promotes ubiquitination and degradation of TDP-43. We demonstrate that a highly selective inhibitor of Tnks-1/2 activity mitigates the neurotoxicity of TDP-43 to rodent neurons. We uncover that TDP-43 has a functional Tankyrase-binding motif, however our data show that TDP-43 is not degraded by Tnks-1/2-dependent ubiquitination. By contrast, our results suggest that Tnks-1/2 stabilizes TDP-43 and that this may occur by inhibiting degradation of TDP-43 by the nuclear proteasome. These findings provide molecular and cellular insight into the interaction between Tnks-1/2 and TDP-43 and provide a foundation for developing novel therapeutic strategies for TDP-43associated disease.

4
To determine if Tnks-1/2 inhibition could protect neurons from TDP-43-associated toxicity, we examined the effect of the small-molecule inhibitor G007-LK, as it is highly selective for Tnks-1/2 and has no reported effect on PARP-1 activity (Voronkov et al., 2013). To determine whether treatment with G007-LK had any effect on neurons in the absence of TDP-43-associated toxicity, we treated cultured neurons infected with HSV-LacZ with either vehicle (DMSO) or G007-LK (1M or 10M) for 7d and quantified the neuronal cell bodies immunolabelled with Tubulin -III chain. At the concentrations tested, G007-LK treatment had no significant effect on neuronal number (Fig. 1, B and C, Fig. S1, B and C, Fig. S2), indicating that G007-LK had little to no effect on general neuronal survival. By contrast, neuronal loss induced by HSV-TDP-43 was significantly reduced by treatment with 1M or 10 M G007-LK (Fig. 1, B and C, Fig S1, B and C, Fig. S2), indicating that G007-LK protects neurons from TDP-43-associated toxicity. These data demonstrate that treating rat primary cortical neurons with the Tnks-1/2 inhibitor protects against TDP-43-associated toxicity. Furthermore, they suggest that targeting Tnks-1/2 activity in TDP-43-associated disease is a potential therapeutic strategy.
To determine if there was a region within TDP-43 that may directly interact with Tnks-1/2, we computationally aligned the Tankyrase-binding motif (RxxDG) to the human TDP-43 protein sequence. This highlighted an evolutionary conserved region (amino acids 165-170) with 84 % identity to the Tankyrase-binding motif (Fig. 2, C-E and Fig. S3 A), which we call the Tankyrasebinding domain (TBD) (Fig. 2 C). To establish if the predicted Tankyrase-binding motif in TDP-43 could mediate an interaction with Tnks-1/2, we deleted the TBD from TDP-43 (TDP-43-TBD-YFP) and tested the ability of the mutant protein to co-immunoprecipitate with Tnks-1/2. This revealed that deletion of the TBD abolished the capacity of TDP-43 to co-immunoprecipitate with Tnks-1/2 in mammalian cells (Fig. 2 F). Importantly, deletion of the TBD did not affect all interactions, as it had no effect on the capacity of TDP-43-TBD to co-immunoprecipitate with endogenous TDP-43 from cellular lysates (Fig. S3 D). Collectively, these data suggest that the TBD is essential for the interaction between TDP-43 and Tnks-1/2 To further define the Tnks-1/2 interaction domain in TDP-43, we mutated each amino acid in the TBD (RxxDG) individually to alanine (Fig. 2 G) and tested the ability of the respective mutated protein variants to co-immunoprecipitate with Tnks-1/2. This demonstrated that mutation of either H166 or I168 to alanine was sufficient to abolish the interaction between TDP-43-YFP and Tnks-1/2, while mutation of R165, D169 and G170 to alanine had little to no effect (Fig. 2 H). We note that the ALS-associated mutation of TDP-43 D169G (Kabashi et al., 2008) resides within the TBD of TDP-43; however, similar to the alanine mutation in D169, the mutation to glycine had no effect on the co-immunoprecipitation of TDP-43 with Tnks-1/2 (Fig. S3 E).

Journal of Cell Science • Accepted manuscript
Analysis of the previously solved NMR structure of RRM1 and RRM2 of TDP-43 (Lukavsky et al., 2013) revealed that the TBD and the RNA-binding regions are on opposite sides of RRM1 ( Fig. 2 I). Recent studies have also demonstrated that mutation in the TBD region (D169G) has no effect on RNA-binding . The TBD spans a loop, a -strand and a second loop (Fig. 2 J) and intriguingly, the amino acids essential for the interaction with Tnks-1/2 (H166 and I168) are positioned on the internal side of the -strand (Fig. 2 J). The non-essential amino acids of the TBD are located on the unstructured loops (R165, D169 and G170) or on the external surface of the -strand (M167) (Fig. 2 J). These combined data indicate that TDP-43 and Tnks-1/2 interact and that this interaction is dependent upon H166 and I168 which are positioned in the -strand in the TBD of TDP-43.

Proteasomal turnover of TDP-43 occurs in the nucleus.
To gain an understanding of how Tnks-1/2 may lead to stabilization of TDP-43, we examined TDP-43 localization by immunofluorescence. Under normal conditions, both TDP-43-WT-YFP and TDP-43-TBD-YFP localized diffusely to the nucleus. In response to MG132 treatment, both proteins formed nuclear foci that co-labelled with ubiquitin ( Fig. 5, A and B). However, ubiquitin co-labelling occurred significantly earlier for nuclear TDP-43-TBD foci than for TDP-43-WT foci (2 hr vs 4hr of treatment) (Fig. 5, A and B). These data are consistent with our finding that TDP-43-TBD is more rapidly ubiquitinated than TDP-43-WT (see Fig. 4), and indicates that upon proteasome inhibition, both TDP-43 proteins (WT and TBD) accumulate in ubiquitinpositive foci in the nucleus.
Curiously, we observed that MG132 treatment did not lead to an increase in the percentage of cells with cytoplasmic foci of TDP-43-WT or TDP-43-TBD, or with cytoplasmic foci of the protein co-labelled with ubiquitin ( Fig. 5, A and C). This appears to be in contrast to some studies, but consistent with others, and we suggest this difference is due to the time period of MG132 treatment (see Discussion). This result, however, raised the possibility that proteasome inhibition may preferentially promote accumulation of TDP-43 selectively in the nucleus. To further explore proteasomal-turnover of TDP-43 in the context of the cellular milieu, we examined the effect of MG132 on a form of TDP-43 that cannot be imported into the nucleus and instead localizes to the cytoplasm. TDP-43 nuclear localization is dependent upon a bipartite nuclear localization sequence (NLS) that also acts as a PAR-binding motif (PBM) in the N-terminal portion of the protein ( It is important to note that under the same conditions, the cytoplasmic protein G3BP1 formed cytoplasmic foci in response to MG132 treatment (Fig. S8, A and B), which is consistent with previous reports . Thus, our data suggest that under the conditions tested, TDP-43 localized to the cytoplasm (TDP-43-NLS/PBM) does not respond to MG132.
To determine if TDP-43-NLS/PBM remained diffuse upon MG132 treatment because of its cytoplasmic localization or, alternatively, because the NLS/PBM mutation impaired the ability of the protein to respond to proteasome inhibition, we generated TDP-43-NLS/PBM-GFP with an exogenous NLS sequence. We compared the bipartite NLS/PBM from the TDP-43 protein (TDP-43-NLS/PBM-TDP-43) to the proline-tyrosine NLS (PY-NLS) from hnRNPA1 (TDP-43-NLS/PBM-A1) (Fig. S7 A). The NLS from TDP-43 differs from the PY-NLS not only in amino acid sequence but also in the transport system used to direct proteins to the nucleus. Importin / (also known as Karyopherin /1) directs TDP-43 to the nucleus while Transportin (also known as Karyopherin β2 directs hnRNPA1 to the nucleus Winton et al., 2008). Under Journal of Cell Science • Accepted manuscript normal conditions TDP-43-NLS/PBM-TDP-43 and TDP-43-NLS/PBM-A1 localized to the nucleus ( Fig. 6 A) and upon treatment with MG132, both TDP-43-NLS/PBM-TDP-43 and TDP-43-NLS/PBM-A1 formed ubiquitin-labelled nuclear foci ( Fig. 6, B, C and D). These data suggest that TDP-43 must be in the nucleus to form MG132-induced nuclear foci. These data also suggest that the NLS/PBM sequence of TDP-43 is not required for the response to MG132 treatment, because the addition of a different NLS sequence (TDP-43-NLS/PBM-A1) also rescues MG132induced accumulation of the protein.
To gain a molecular understanding of TDP-43 turnover in the context of the cell, we assessed the ubiquitination levels of immunoprecipitated TDP-43-GFP localized to the nucleus (-WT, -NLS/PBM-TDP-43) or cytoplasm (-NLS/PBM). Upon MG132 treatment, both nuclear forms of TDP-43 (-WT and -NLS/PBM-TDP-43) were ubiquitinated (Fig. 6, E and F). Although MG132 treatment led to an increase in ubiquitination of cytoplasmic TDP-43 (-NLS/PBM) compared to baseline, the levels were significantly lower than for TDP-43-WT (Fig. 6, E and F). This finding indicates that upon MG132 treatment, ubiquitinated TDP-43-GFP localized to the nucleus accumulates more rapidly than ubiquitinated TDP-43-GFP in the cytoplasm. These data further suggest that in this assay TDP-43 is preferentially degraded by the nuclear proteasome.

Tankyrase-1/2 promotes cytoplasmic accumulation of TDP-43
Previously, we found that Tnks-1/2 regulates the cytoplasmic accumulation of TDP-43 in aging neurons in Drosophila and in the cytoplasm of mammalian cells exposed to the chemical stressor arsenite (McGurk et al., 2018a). Furthermore, Tnks-1/2 has been reported to localize to the cytoplasmic face of the nuclear pore complex (Smith and de Lange, 1999). We thus considered that Tnks-1/2 may inhibit turnover of TDP-43 by promoting cytoplasmic accumulation of the protein.
8 and Tnks-1/2 interaction, could help maintain TDP-43 in the nucleus, where damaged and misfolded forms of the protein can be turned over by the proteasome. Consistent with this, we demonstrate that pharmacological inhibition of Tnks-1/2 activity protects against TDP-43associated neurotoxicity in primary rodent neurons.
The Tankyrase-binding motif is a loosely conserved motif of RxxDG . The first and last amino acid (R1 and G6) in peptides or fragments that span the Tankyrase-binding motif of 3BP2, Axin1, RNF146 and IRAP abolish the interaction with Tnks-1/2 and are considered to be invariant amino acids (DaRosa et al., 2018;Morrone et al., 2012;. Mutation analysis of a peptide spanning the TBD of 3BP2 demonstrated that at position 4, glycine, proline, alanine and cysteine are preferred while isoleucine, leucine and valine prevent Tnks-1/2 binding to the 3BP2 peptide ). This appears in contrast with our detailed mutational analysis of the Tankyrase binding domain in TDP-43, which show that mutation of R1 or G6 has little to no effect on the interaction between TDP-43 and Tnks-1/2. Furthermore, not only is position 4 of the TBD an isoleucine, mutation of this amino acid in the full-length form of TDP-43 abolishes the interaction with Tnks-1/2. Thus, we suggest that TDP-43 may harbour a Tankyrase-binding motif with non-canonical features. It is important to note that non-canonical and extended Tankyrase-binding motifs have been identified; these include an extra 2 amino acids after position 6 in 3BP2 and motifs with additional amino acids between position 1 and 2 in APC2, Axin1 and RNF146 (Croy et al., 2016;DaRosa et al., 2018;Morrone et al., 2012). Furthermore, the second Tankyrase-binding motif in Axin is not dependent on R1 (Morrone et al., 2012) and Tankyrase-binding motifs in several other proteins harbor amino acids that are not tolerated at the fourth position in the 3BP2 peptide including a valine (motif 1 of Axin 1, motif 2 in RNF146, motif 1 of APC2 and motif 3 of PEX14) and a leucine (motif 2 of MDC1) (Croy et al., 2016;DaRosa et al., 2018;Li et al., 2017b;Morrone et al., 2012;Nagy et al., 2016). Thus, in the context of different proteins or perhaps context-dependent functions of Tnks-1/2, the constraints on the consensus of the Tankyrasebinding motif differ.
The mechanism by which Tnks-1/2 stabilizes TDP-43 may include cellular localization of the protein. Using a combination of biochemistry and immunofluorescence, we observed that upon proteasome inhibition, TDP-43 primarily accumulates in the nucleus and is more rapidly Journal of Cell Science • Accepted manuscript ubiquitinated than cytoplasmic TDP-43. Our data, presented here and previously (McGurk et al., 2018a), suggest that Tnks-1/2 promotes cytoplasmic accumulation of TDP-43. We hypothesize that in doing so, Tnks-1/2 may either sequester TDP-43 from the nuclear proteasome or inhibit nuclear import of the protein which ultimately leads to stabilization (Fig. 8). Tnks-1/2 activity is known to regulate protein localization, for example by promoting Axin localization to the Wnt receptor, or alternatively by directing its degradation (De Rycker and Price, 2004;Mariotti et al., 2016;Wang et al., 2016;Yang et al., 2016). Whether the regulation of TDP-43 localization by Tnks-1/2 is direct or indirect, and whether this involves modification of TDP-43, for example by PARylation, remain to be established.
Impairment of the proteasome has been hypothesized to be a disease-causing mechanism in ALS/FTD (Scotter et al., 2015;Taylor et al., 2016), thus many studies have examined TDP-43 cellular localization upon proteasome inhibition with MG132. Treatment with MG132 for extended time periods (ranging from 12 hr to 72 hr) leads to diffuse accumulation of TDP-43 in the cytoplasm (Klim et al., 2019;van Eersel et al., 2011;Walker et al., 2013) or cytoplasmic agregates of TDP-43 (Huang et al., 2014;Li et al., 2017a;Scotter et al., 2014). Our data demonstrate that treatment with MG132 for short periods of time (3-5 hr) leads to nuclear foci formation of TDP-43-WT, while TDP-43 localized to the cytoplasm (TDP43-NLS/PBM) remains diffuse. Our findings are consistent with previous studies that show that TDP-43-WT forms nuclear foci in cells treated with MG132 for 8 hr (Wang et al., 2010) and that TDP-43 with mutation the NLS/PBM remains diffuse upon treatment with MG132 for 6 hr (Nonaka et al., 2009). Thus, we propose that upon proteasomal inhibition, TDP-43 accumulation in the nucleus occurs earlier than in the cytoplasm. Our data further suggest that turnover of TDP-43 by the nuclear proteasome is important for regulating TDP-43 degradation.
An impact on global protein levels is thought to be involved in in ALS/FTD. Many diseaseassociated mutations occur in proteins that function in protein turnover such as the UPS and autophagy, including C9orf72, Charged multivesicular body protein 2b, Optineurin, Sequestrome 1, Serine/Threonine-protein Kinase TBK1, Ubiquilin 2 and VCP, suggesting a broad impairment of protein turnover in neurodegenerative disease (Balendra and Isaacs, 2018;Gao et al., 2017;Taylor et al., 2016). Furthermore, altered TDP-43 protein levels has also been implicated in disease. For example, TDP-43 mRNA levels are upregulated in post-mortem tissue from patients with ALS/FTD as well as in a knock-in mouse model for the Q331K disease-causing mutation in TDP-43 (Gitcho et al., 2009;Mishra et al., 2007;White et al., 2018). In post-mortem tissue, TDP-43 protein abnormally accumulates in ubiquitin inclusions in the cytoplasm of affected neurons and glia, suggesting that the affected cells cannot remove and degrade TDP-43 in the cytoplasm (Arai et al., 2006;Mackenzie et al., 2007;Neumann et al., 2007). In patient fibroblasts (sporadic and those harboring a G4C2-hexanucleotide expansion in C9orf72 or mutation in TDP-43), the levels of cytoplasmic TDP-43 and levels of total TDP-43 protein compared to control cells are significantly higher (Lee et al., 2019;Sabatelli et al., 2015). Additionally, upon proteasome inhibition, TDP-43 protein levels remain unaltered in these ALS/FTD patient fibroblasts (Lee et al., 2019), suggesting that UPS turnover of TDP-43 is impaired. It is intriguing to postulate that UPS-mediated turnover of TDP-43 is impaired in the ALS/FTD patient fibroblasts because TDP-43 is accumulating in the cytoplasm and thus is sequestered from the nuclear proteasome.
There are no effective treatments for ALS/FTD and related disorders such as Parkinson's disease. In ALS, nuclear PAR is elevated in motor neurons of post-mortem spinal cord tissue and total PAR is elevated in the cerebrospinal fluid of patients with Parkinson's disease (Kam et al., 2018;McGurk et al., 2018c). These findings suggest that inhibiting PARP activity may have therapeutic potential. Similar to inhibition of Tnks-1/2, chemical inhibition of nuclear PARP-1/2 activity reduces accumulation of TDP-43 in stress foci in the cytoplasm (McGurk et al., 2018a). Thus, nuclear PARP activity may promote nuclear export of TDP-43, which may also lead to reduced turnover of the protein. Finding agents that regulate the levels of TDP-43 is important to modulate TDP-43 protein homeostasis, as well as to remove misfolded and possibly toxic forms of the protein that accumulate in affected brain regions. We propose that inhibition of Tnks-1/2, which regulates TDP-43 stability and neurotoxic properties, is a potential therapeutic target for ALS/FTD and related disorders.

Rat cortical neurons culture and neurotoxicity assays.
Rat cortical neurons were from embryos isolated from female Sprague Dawley wild rats that were 16-18 days pregnant (Neuron R Us, neuron service center, University of Pennsylvania). 100,000 neurons were plated out on poly-D-lysine coated coverslips (12mm diameter and thickness #1 (Neuvitro, Vancouver, WA, Canada) in neurobasal medium (ThermoFisher Scientific, Waltham, MA, USA) supplemented with serum-free B27 (ThermoFisher Scientific, Waltham, MA, USA), penicillin streptomycin (ThermoFisher Scientific, Waltham, MA, USA) and Glutamax (ThermoFisher Scientific, Waltham, MA, USA). Neurons were cultured at 37 o C with 5% CO2 and half of the media was removed and replaced with fresh prewarmed media 3 times per week. The primary neurons were infected at 14-17 DIV with HSV-TDP-43 or HSV-LacZ with either the Tnks-1/2 inhibitor G007-LK (SelleckChem, Houston, TX, USA) or DMSO (Sigma Aldrich, St. Louis, MI, USA). Every 2d, half of the media was removed and replaced with fresh media containing the G007-LK or DMSO. The drug containing media was made up twice concentrated so that when it was diluted 2-fold in the well it was of the appropriate concertation. 7d post infection, the neurons were fixed in 4% paraformaldehyde, blocked in 10% normal donkey serum (Sigma Aldrich, St. Louis, MI, USA) in Tris buffered saline with 0.05% Tween  20 (TBST) (ThermoFisher Scientific, Waltham, MA, USA) for 1 hr at room temperature, and immunostained overnight at 4 o C with antibodies directed to the neuronal marker Tubulin -3 chain (1 in 500, Abcam, Cambridge, UK). After 3 sets of 5 min washes in TBST (ThermoFisher Scientific, Waltham, MA, USA) neurons were incubated with mouse AlexaFluor 488 (1 in 500, ThermoFisher Scientific, Waltham, MA, USA) in the dark for 1hr at room temperature. Neurons were washed with TBST 3 times (5 min each) and counterstained with 1 µg/ml Hoechst-33342 (ThermoFisher Scientific, Waltham, MA, USA) (15 min), washed in deionized H20 and mounted in prolong diamond (ThermoFisher Scientific, Waltham, MA, USA). Five images (10X magnification) were captured from each coverslip and remaining neuronal cell bodies in each image were counted. Each condition was repeated three times, on three independent cultures each from a different pregnant rat.

Mammalian cells and culture details
COS-7 cells originally purchased by ATCC were a gift from Virginia M. Lee (University of Pennsylvania). Prior to purchase the COS-7 cells were authenticated by ATCC.
Journal of Cell Science • Accepted manuscript HEK293T cells were kindly provided by Aaron Gitler (Stanford University). COS-7 cells were routinely grown in Dulbecco's modified Eagle's medium with high glucose and L-glutamine (ThermoFisher Scientific, Waltham, MA, USA), 10% filter sterile FBS (Sigma Aldrich) and penicillin-streptomycin (ThermoFisher Scientific, Waltham, MA, USA). HEK-293T cells were grown in DMEM with high glucose, L-glutamine and sodium pyruvate (ThermoFisher Scientific, Waltham, MA, USA), 10% filter sterile FBS (Sigma Aldrich, St. Louis, MI, USA) and penicillinstreptomycin (ThermoFisher Scientific). Cells were grown at 37 o C with 5% CO2, and a water bath was used for humidification. Cells were washed with dPBS without calcium and without magnesium (ThermoFisher Scientific, Waltham, MA, USA) and trypsinized in Trypsin with 0.25 % EDTA (ThermoFisher Scientific, Waltham, MA, USA). No commonly misidentified cell lines were used.

Identification of the Tankyrase binding domain
We computationally aligned the Tankyrase-binding motif (RxxDG) to the human TDP-43 protein sequence using the PATTINPROT search engine . To map the TBD to the reported NMR structure of RRM1 and RRM2 of TDP-43 (4BS2) 43 (Lukavsky et al., 2013) we used the open source Java viewer "FirstGlance in Jmol".

Co-immunoprecipitation and Immunoblotting
To examine the interaction between endogenous TDP-43 and Tnks-1/2, 2.5 µg of control mouse IgG (Santa Cruz Biotechnology, Dallas, TX, USA) and 2.5 µg mouse anti-TDP-43 (5028) (Kwong et al., 2014) were coupled to 50 µl Protein G dynabeads TM (ThermoFisher Scientific, Waltham, MA, USA). COS-7 and HEK293T cells were each grown to confluency in 1 T75 flask overnight and lysed the following day. See below for lysis method. . For all co-immunoprecipitations lysates each from 1 T75 flask (~8.4 x 10 6 cells) , were incubated on ice for 10 min, passed through a 20G1½ 1 mL syringe (BD Biosciences, San Jose, CA, USA) 3 times, transferred to centrifuge tubes, rotated at 15 rpm for 10 min at 4°C and centrifuged for 10 min at 15000 rpm at 4°C. The supernatant was collected, and pellet discarded, 25 µl of lysate was removed for input and the remaining lysate was divided in half (500 µl) for incubation with IgG or anti-TDP-43-coupled beads. The reactions were made up to 1 mL in lysis buffer with protease inhibitor and incubated with the antibody-coupled beads for 18 hr at 4°C with 15 rpm rotation. The beads were washed 3 times by removing the supernatant, adding 500 µl lysis buffer containing protease inhibitor, briefly resuspending the beads, placing the tube on the magnet and immediately removing the buffer. Elution was performed at 95°C for 5 min in 40 µl 1X LDS Sample Buffer (ThermoFisher Scientific, Waltham, MA, USA) with 5% β-mercaptoethanol (Sigma Aldrich). Input samples (10 µl) were denatured at 95°C for 5 min in 1X LDS Sample Buffer (ThermoFisher Scientific, Waltham, MA, USA) with 5% β-mercaptoethanol (Sigma Aldrich, St. Louis, MI, USA) in a total volume of 20 µl.

Journal of Cell Science • Accepted manuscript
Eluates and inputs were electrophoresed (10 µl of the eluate and 10 µl of input was loaded for the detection of TDP-43, and 13 µl of the eluate and input for the detection of Tnks-1/2) on NuPAGE 4-12% Bis-Tris gels (   For quantifying the effect of Tnks-1/2 inhibition on TDP-43-GFP localization, 4-5 independent images at 20X were captured at the same exposure time. Cells with diffuse cytoplasmic GFP and/or GFP foci in the cytoplasm were scored as cells with cytoplasmic TDP-43-GFP. Up to 540 transfected cells were quantified per condition. All experiments were at repeated at least 3 independent times and the mean ( s.e.m.) calculated.

Statistical analysis
All data points in each graph are mean (± SEM or st. dev.) and the n is a biological repeat. T Tests, one-way ANOVA, two-way ANOVA and multiple comparison tests were performed and are described in each figure legend. Significance was set at p < 0.05, values for asterisks are presented in each legend. All statistical analyses were carried out using Graphpad prism6 software (GraphPad software, San Diego, CA, USA).

Data availably
Requests for resources, reagents and further information should be directed to, and will be fulfilled upon reasonable request. Figure 1: The Tankyrase-1/2 inhibitor G007-LK reduces TDP-43-associated loss of rat primary cortical neurons.

Figures
A. Cortical neurons isolated from Sprague Dawley embryos (E16-E18) were seeded in 24-well plate format at a density of 100,000 neurons. After 15-18 days in vitro (DIV) neurons were virally infected with either HSV-LacZ or HSV-TDP-43 and treated with DMSO or G007-LK. Neurons were fixed and immunostained 7d post infection (DPI). See Fig. S2 for expanded images.
B. Neurons were immunolabelled for the neuronal marker Tubulin -III chain (green) and counterstained with Hoechst (blue). Arrows indicate neurons.
Journal of Cell Science • Accepted manuscript A. COS-7 cells expressing TDP-43-WT-YFP and TDP-43-TBD-YFP were treated with vehicle or MG132 for the indicated time periods. Cells were fixed and immunolabelled for the cytoplasmic protein G3BP1 (yellow) and ubiquitin (magenta), and counterstained with Hoechst (blue). Both TDP-43-YFP proteins formed nuclear foci upon proteasome inhibition (green). Nuclear foci of TDP-43-TBD were co-labelled with ubiquitin by 2 hr of MG132 treatment, whereas TDP-43-WT co-labelled with ubiquitin only by 4 hr of MG132 treatment. Nuclei outlined in white dashed line. Journal of Cell Science • Accepted manuscript Model for effects of Tnks-1/2 to modulate the subcellular localization of TDP-43. The data show that Tnks-1/2 promotes cytoplasmic accumulation of TDP-43, thereby inhibiting access of TDP-43 to the nuclear proteasome. In this way, Tnks-1/2 stabilizes TDP-43 in the cytoplasm. In human disease, accumulation of TDP-43 in the cytoplasm is observed in affected brain cells of >95% of ALS cases and ~45% of FTD cases. Thus, therapeutic inhibition of Tnks-1/2 in ALS/FTD may maintain TDP-43 in the nucleus where misfolded or mutated forms of the protein can be degraded by the nuclear proteasome. B. COS-7 and HEK-293T cells expressing TDP-43-WT-GFP or TDP-43-DPBM-GFP were exposed to a control (DMSO) or 10 µM MG132 for 5 hr. TDP-43-WT-GFP formed