Hsp90 and its co-chaperone Sti1 control TDP-43 misfolding and toxicity

Protein misfolding is a central feature of most neurodegenerative diseases. Molecular chaperones can modulate the toxicity associated with protein misfolding, but it remains elusive which molecular chaperones and co-chaperones interact with specific misfolded proteins. TDP-43 misfolding and inclusion formation is a hallmark of amyotrophic lateral sclerosis (ALS) and other neurodegenerative diseases. Using yeast and mammalian neuronal cells we find that Hsp90 and its co-chaperones have a strong capacity to alter TDP-43 misfolding, inclusion formation, aggregation, and cellular toxicity. Our data also demonstrate that impaired Hsp90 function sensitizes cells to TDP-43 toxicity. We further show that the co-chaperone Sti1 specifically interacts with and modulates TDP-43 toxicity in a dose-dependent manner. Our study thus uncovers a previously unrecognized tie between Hsp90, Sti1, TDP-43 misfolding, and its cellular toxicity.


Introduction
Protein misfolding and aggregation are hallmarks of protein conformational diseases, including neurodegenerative diseases, such as Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), Frontal Temporal Dementia (FTD), and amyotrophic lateral sclerosis (ALS) (1)(2)(3). ALS is characterized by death of upper and lower motor neurons causing progressive loss of muscle function, eventually leading to fatal paralysis of the respiratory system (4). In the majority of ALS cases (~97%), the transactive response element DNA/RNA binding protein of 43 kDa (TDP-43) mislocalizes from the nucleus to the cytoplasm (5)(6)(7). Misfolded and hyperphosphorylated TDP-43 forms inclusions in the cytoplasm of neurons in the brain and spinal cord of ALS patients (8). This TDP-43 proteinopathy is also found in the majority of FTD patients and over 50% of AD patients and 60% of PD patients (9,10), indicating a broader role (mechanism?) of TDP-43 in the pathogenesis of multiple neurodegenerative diseases. A recent report suggests that accumulation of TDP-43 in limbic structures defines a new type of dementia (11). TDP-43 has a high propensity to aggregate due to its carboxyterminal prion-like domain (12,13). Accordingly, prediction of prion-like and aggregation-prone proteins in humans by an algorithm ranks TDP-43's propensity to aggregate as 69 th out of the entire human proteome (14).
Cells combat the accumulation of damaged or misfolded proteins by the action of molecular chaperones and heat shock proteins, which facilitate protein refolding or degradation to prevent the toxic consequences associated with protein misfolding (15). Hsp90 is a cytosolic molecular chaperone that is expressed at high levels under normal conditions and strongly induced by stress.
It is highly conserved within eukaryotes and regulates many cellular pathways, such as cellular signaling, cell cycle control, and cell survival (16).
Hsp90 co-localizes with many disease-related protein aggregates, such as tau tangles and amyloid beta (Aβ) plaques in AD (17,18). It has also been shown to inhibit tau and A aggregation (19,20), which may represent a defense mechanism. Yet, it has also been proposed that Hsp90 binds and stabilizes toxic aberrant protein species, leading to the toxic accumulation of these proteins and thereby contributing to neurodegeneration (18). For instance, epichaperomes, i.e. metastable chaperone networks stabilized by Hsp90, seem to promote the accumulation and toxicity of misfolded tau (21). Experiments using purified proteins have shown that Hsp90 has the capacity to both inhibit TDP-43 aggregation and to convert TDP-43 aggregates into more soluble species (22).
The binding and hydrolysis of ATP is essential for Hsp90 binding to client proteins. Co-chaperones, such as activator of Hsp90 ATPase protein 1 (Aha1), cell division cycle protein 37 (Cdc37), and stress inducible phosphoprotein 1 (Sti1, also known as STIP1 and Hsp70/Hsp90 organizing protein, HOP in humans) inhibit (Sti1 and Cdc37) or stimulate (Aha1) Hsp90 ATPase activity and facilitate the folding or activation to specific client proteins (23).
Aha1 stimulates the ATPase activity of Hsp90 by up to twelve fold over its basal level (24,25) and regulates the activation of Hsp90 client proteins (26). Cdc37 regulates the activity of protein kinases; it can act as a molecular chaperone independently of Hsp90 and its over expression can compensate for decreased Hsp90 function (27). Cdc37 together with Hsp90 participate in the nuclear localization of TDP-43, which is typically associated with reduced TDP-43 toxicity; the disruption of the Cdc37/Hps90 complex triggers degradation of TDP-43 via autophagy (28).
Unlike other Hsp90 co-chaperones, Sti1 can bind to both Hsp70 and Hsp90 simultaneously to allow for client protein transfer between these two major chaperone systems (29). Sti1 can function independently of Hsp90 and is involved in protein folding of protein kinases and the curing of yeast prions through interaction with Cdc37 (30,31). Sti1 also interacts with many different aggregation prone proteins (32)(33)(34), including the mammalian prion protein (35).
Collectively, these studies document the capacity of Hsp90 and its co-chaperones to modulate protein misfolding, aggregation, and the ensuing toxicity. Although Hsp90 has been shown to interact with TDP-43, it remains unclear whether Hsp90 and its co-chaperones modulate TDP-43 toxicity. Here we report for the first time that Hsp90 and its co-chaperones, particularly Sti1, regulate TDP-43 aggregation and toxicity in yeast and mammalian cell models.

Hsp90 regulates TDP-43 toxicity
We used well-established yeast models to explore how Hsp90 levels and its ATPase activity modulate TDP-43 proteinopathy and toxicity (13,(36)(37)(38). In a slight variation to previously published TDP-43 yeast models, we used TDP-43-GFP expressed from a low-copy plasmid (CEN) under control of the inducible galactose (GAL-1) promotor, to provide low level overexpression, which produces moderate growth defects ( Fig 1A). This model allows us to study factors that can either enhance or suppress TDP-43 toxicity. We first sought to determine the effect of inhibiting Hsp90's ATPase activity by treating yeast cells expressing TDP-43-GFP with two established Hsp90 inhibitors, radicicol and 17AAG (17-(Allylamino)-17-demethoxygeldanamycin) (39,40).
Growth assays on agar plates ( Fig 1B) and liquid cultures ( Fig 1C) document an increase in TDP-43 toxicity even at low concentrations of both Hsp90 inhibitors. Of note, these low concentrations of radicicol and 17AAG do not cause growth defects in wild type control cells or elicit the heat shock response (HSR) (41). Fluorescence microscopy ( Fig 1D) shows no major difference in TDP-and more diffuse localization of TDP-43-GFP in radicicol-and 17AAG-treated cells compared to untreated cells.
If Hsp90 regulates TDP-43 toxicity, reducing cellular Hsp90 levels should reproduce the effects of pharmacological inhibition. Deletion of both Hsp90 alleles together in yeast (Hsc82, the constitutive allele and Hsp82, the stress induced allele) is lethal, whereas the deletion of either one allele alone does not result in adverse growth phenotypes under normal conditions (42). TDP-43-GFP was expressed in yeast cells that lack either one of the yeast Hsp90 encoding genes (hsc82Δ or hsp82Δ yeast strains). Our results show an increase in TDP-43 toxicity in both Hsp90 deletion strains when compared to wild type controls (Fig 2A). Fluorescence microscopy did not reveal any striking differences in TDP-43-GFP localization and aggregation between wild type and either Hsp90 deletion strain ( Fig 2B). Occasionally, Hsp90-depleted cells displayed a slight increase in smaller cytosolic puncta, which was more pronounced in the hsc82Δ than in the hsp82Δ strain.
These results indicate that inhibiting Hsp90 ATPase activity or reducing cellular Hsp90 levels increases TDP-43 toxicity.

Over-expression of Hsp90 exacerbates TDP-43 toxicity
We next tested whether Hsp90 over-expression alters TDP-43 toxicity by over-expressing Hsc82 or Hsp82 using a high copy plasmid (2). When co-expressed with TDP-43-GFP, over-expression of Hsc82 or Hsp82 caused a substantial increase in TDP-43 toxicity (Fig 2C). This synthetic toxicity is especially apparent in liquid culture growth (Fig 2D-E), where the doubling times of Hsp82 or Hsc82 over-expressing cells that co-express TDP-43-GFP are significantly increased.
Fluorescence microscopy in cells over-expressing either Hsc82 or Hsp82 are very similar to control cells with some cells showing slightly increased diffuse cytoplasmic localization of TDP-43-GFP ( Fig 2F).

Sti1, Aha1, and Cdc37 and TDP-43 toxicity
Given the robust effects of changing Hsp90 on increasing TDP-43 toxicity, we tested whether modulating Hsp90 co-chaperones could also affect TDP-43 aggregation and toxicity. We focused on three co-chaperones that have been previously implicated in regulating protein misfolding (16,28,43,44). First, we tested the effect of deletion of Sti1 or Aha1 (sti1Δ or aha1Δ yeast strains respectively), and reduced expression of the essential Cdc37 (Cdc37-DAmP yeast strain). We also included two essential genes, Ess1 and Sgt1, in our studies to test whether reduced expression of these genes had an effect on TDP-43 toxicity. Ess1 is an essential gene that encodes a peptidylprolyl cis/trans isomerase that plays a role in transcription regulation. Sgt1 is an essential Hsp90 co-chaperone that has been implicated in kinetochore complex assembly and in neurodegeneration (45,46).
Our results show that deletion of Sti1, which did not have any adverse effect on the growth of control cells, drastically exacerbated TDP-43 toxicity. In contrast, deletion of Aha1 had no effect on TDP-43 toxicity (Fig 3A). Due to the innate toxicity associated with reduced expression of Cdc37, we chose a construct that expresses TDP-43 at a very low level and exhibits little to no toxicity when expressed in wild type cells to test for TDP-43 toxicity in the Cdc37-DAmP strain.
We observed an increase in TDP-43 toxicity when TDP-43 is expressed in the Cdc37-DAmP strain ( Fig 3B). However, it should be noted that reducing Cdc37 also affects the wild type strain, hence the effect of Cdc37 may not be selective to TDP-43. Synthetic toxicity is not seen in reduced expression of the other two essential Hsp90 co-chaperones, Ess1 and Sgt1 (S1A- B Fig).
We then assessed how Sti1, Aha1, and Cdc37 over-expression affects TDP-43 toxicity. Given the surprising effects of high Hsp90 over-expression, we used two different expression systems-high over-expression from high copy plasmids (2) and moderate over-expression from low copy plasmids (CEN). High over-expression of Sti1 resulted in a slight reduction of growth in control cells, but high over-expression of Aha1 and especially Cdc37 showed significant reduced growth in control cells (Fig 3C, S2A Fig). In TDP-43-expressing cells, high over-expression of these cochaperones produced an increase in toxic phenotypes (Fig 3C, S2A Fig). However, moderate overexpression of Sti1 showed no toxicity, whereas moderate over expression of Aha1 or Cdc37 was toxic ( Fig 3C). This increased toxicity is also seen when TDP-43 is co-expressed with Aha1 and Cdc37 ( Fig 3C). By contrast, moderate Sti1 over-expression rescued TDP-43 toxicity (Fig 3C).
Fluorescence microscopy of TDP-43-GFP in the sti1 strain showed an increase in cytosolic aggregates that appear to be smaller in size, while TDP-43-GFP localization in the aha1 strain and the strain expressing reduced levels of Cdc37 were indiscernible from wild type ( Fig 3D).
Notably, moderate over-expression of Sti1 increase diffuse cytosolic TDP-43, whereas high overexpression shows more inclusions of TDP-43 ( Fig 3E). Moderate Aha1 and Cdc37 over-expression did not produce obvious changes in TDP-43-GFP localization when compared to control cells ( Fig   3E). Conversely, high over-expression of Aha1 and Cdc37 both increased diffuse cytosolic TDP- Hsp90 and Sti1 deletion show synthetic lethality (47) and there is ample genetic evidence for their co-regulation. We therefore sought to investigate how the over-expression of Hsp90, Aha1, or Cdc37 in sti1 yeast strains controls TDP-43 toxicity (Fig 3F), to determine whether overexpression of other molecular chaperones or co-chaperones can compensate for the loss of

Hsp90 and its co-chaperones have no effect on TDP-43 Q331K
Mutations in the gene encoding TDP-43 are associated with familial forms of ALS (fALS). The majority of fALS TDP-43 mutations result in amino acid substitutions in the prion-like domain at the carboxy-terminus of TDP-43 (48). These mutations are thought to result in increased TDP-43 proteinopathy and toxicity (13). Yeast models, e.g. for the fALS mutant TDP-43 Q331K, indeed show increased aggregation and toxicity (13). Our experiments document that unlike wild-type TDP-43, manipulating Hsp90 and Sti1 activity in yeast had no significant effect on TDP-43 Q331K toxicity and aggregate localization and solubility (Fig 4A-D), suggesting that the conformation assumed by this mutant makes it independent of the Hsp90 chaperone machinery in yeast.

Higher Hsp90 levels increase TDP-43 protein levels
Our next experiments aimed to decipher the cellular mechanism by which Hsp90 and its cochaperone affects TDP-43. We asked whether changes in TDP-43 toxicity and localization in yeast cells expressing different levels of Hsp90 and Sti1, Aha1, or Cdc37 may be the result of altered TDP-43 protein levels. We thus monitored TDP-43-GFP levels in yeast cells bearing deletions or over-expression of Hsp90 and its co-chaperones. Our Western blot analyses reveal that Hsp90 over-expression (Hsp82 or Hsc82) increased TDP-43 levels at steady-state, whereas TDP-43 remained unaltered in the hsc82Δ and hsc82Δ yeast strains (Fig 5A and 5B). Interestingly, TDP-43 protein levels were reduced in the sti1 strain even though TDP-43 toxicity was increased ( Fig   5C). Deletion of Aha1, the moderate and high over-expression of Sti1, Aha1, and Cdc37 had no effect on TDP-43 protein levels (S4 Fig). These results indicate that increased Hsp90 levels increases TDP-43 protein levels and thus exacerbated its toxicity. By contrast, decreasing Hsp90 did not alter TDP-43 levels. In addition, while over-expression of Sti1 decreases TDP-43 toxicity, these effects were not associated with altered TDP-43 protein levels.

Sti1 modulates TDP-43 solubility
The aggregation and mislocalization of TDP-43 from the nucleus to the cytoplasm is a central pathological hallmark of ALS and other TDP-43-associated neurodegenerative diseases (8,49) that can be faithfully recapitulated in yeast models. By contrast, the biophysical properties of these TDP-43 inclusions seems more variable. TDP-43 can form both detergent-and denaturing-resistant aggregates and more soluble amorphous protein deposits (50). These different biophysical properties may depend on the experimental system, expression levels, and growth conditions. Therefore, following a previously established protocol (51, 52), we devised a sedimentation assay to assess TDP-43 solubility in the presence of mild detergents. These sedimentation assays did not reveal any difference in solubility under conditions of increased Hsp90 expression levels ( Fig 6A   and 6B). Neither the over-expression of Aha1 nor Cdc37 showed any changes in TDP-43 solubility in sedimentation experiments (Fig 6C and 6D). Comparatively, sti1 strains showed an increase in the relative proportion of insoluble TDP-43 ( Fig 6E). Both moderate and high over-expression of Sti1 also resulted in increased levels of soluble TDP-43 ( Fig 6F and 6G).
These Sti1-KO cells showed a markedly reduced level of endogenous TDP-43 protein compared to wild type SN56 cells, without any change in TDP-43 mRNA levels ( Fig 7A). In addition, tissue from Sti1-KO mouse embryos (55) showed that endogenous TDP-43 levels were found to be reduced in Sti1-KO embryos (Fig 7B and 7C). These results suggest a conserved role for Sti1 in modulating TDP-43 levels in yeast, mammalian cells, and mouse embryos.

Hsp70, Hsp90, and Sti1 form a complex with TDP-43 and regulate its toxicity in neuronal cell
Hsp90 and Hsp70 can form protein complexes with TDP-43 (28,56,57). To test the possibility that Sti1 can also form a complex with TDP-43, we immunoprecipitated HA-Sti1 and tested for specific interaction with TDP-43 in HEK cells ( Fig 7D). The results show that Sti1 precipitates TDP-43. We then performed the reciprocal experiment and immunoprecipitated GFP-TDP-43 from wild type SN56 cells and tested for co-immunoprecipitation of a complex containing Sti1, Hsp70, and Hsp90 ( Fig 7E). Indeed, we observed that TDP-43 is part of a complex with Sti1, Hsp70 and Hsp90, suggesting that the regulation of TDP-43 misfolding may depend on the formation of a complex with Hsp90, Hsp70 and Sti1. We repeated these experiments in Sti1-KO

SN56 cells and tested whether endogenous Sti1 is required for the interaction of TDP-43 with both
Hsp90 and Hsp70 ( Fig 7E). Interesting, even in the absence of Sti1, TDP-43 could still precipitate both Hsp90 and Hsp70, suggesting that these molecular chaperones can interact with TDP-43 in the absence of Sti1. In sum, our results obtained with SN56 cells recapitulated key results from yeast and demonstrated a direct biochemical interaction between Sti1 and TDP-43. Importantly, in STI1-KO cells, transfection of TDP-43 increased cell death and toxicity (Fig. 7F). Conversely, increased expression of Sti1 in SN56 cells decreased TDP-43 toxicity (Fig. 7G).
Using a second cell model, we performed experiments in partially differentiated N2a cells to confirm these results in a second cell line and further evaluate the stoichiometry of Sti1 expression levels. These cells, particularly following serum withdrawal and limited glucose supply (see Material and Methods for details), quantitatively recapitulated TDP-43 proteinopathy and toxicity Hsp90 are part of a complex with Sti1 by co-immunoprecipitation in the N2a cells ( Fig 8D).

Discussion
Hsp90 and its co-chaperones have been suggested to have a highly select suite of client proteins, such as hormone receptors, kinases, and other signaling molecules (58,59). Rather than regulating their de novo folding and preventing or reversing protein misfolding, Hsp90 and its co-chaperone supposedly regulate a final step in the maturation of its clients, which is essential for attaining their active conformation (44). Yet, Hsp90 and Sti1 also modulate the aggregation and toxicity of polyglutamine expansion proteins (60), tau (61) A (34) and other misfolded proteins. In addition, we have shown increased expression levels of Sti1 in human AD brains and AD mouse models and that in vitro Sti1 decreases the toxicity of Aβ oligomers in cultured neurons and brain slices (62), whereas in vivo it favors amyloid aggregation (54). Our results suggest that the Hsp90 cochaperone, Sti1, has critical roles in the regulation of TDP-43 misfolding and the ensuing toxicity, possibly by connecting the Hsp90 and Hsp70 chaperone systems to misfolded proteins (54). TDP-43 toxicity has important implications not only to ALS, but also to FTD, AD and possibly other neurodegenerative diseases. Our work in combination with published work thus contributes to an emerging view of a crucial role of Hsp90 and its co-chaperones in protein homoeostasis in the context of protein misfolding associated with neurodegenerative diseases (54,63).
We find that decreased Hsp90 chaperone activity and decreased levels of Sti1 increase TDP-43 toxicity. This action of Sti1 seems to be selective. Altered levels of another Hsp90 cochaperone, Aha1, does not seem to affect TDP-43 toxicity, whereas increased Cdc37 and Hsp90 expression exacerbates TDP-43 toxicity. Other essential Hsp90-associated genes in yeast also does not affect TDP-43 toxicity. Notably, moderate over-expression of Sti1 reduces TDP-43 toxicity, whereas high levels of Sti1 expression and its deletion increases TDP-43 toxicity. Moreover, we found that Sti1 forms a physical complex with TDP-43, Hsp90 and Hsp70 and that it regulates TDP-43 toxicity in a dose-dependent manner. Our results suggest that TDP-43 toxicity depends on properly balanced function of Hsp90 and its co-chaperones, particularly Sti1, and that disturbances of this balance sensitizes cells to TDP-43 toxicity. Changes in Sti1 levels can regulate the levels of classical clients of Hsp90 (54) and disturbances on Sti1 can contribute to abnormal chaperone function (35). Sti1 is also a critical node, which can disrupt epichaperomes, abnormal complexes of chaperone proteins that present increased connectivity due to protein misfolding (21).
Remarkably, TDP-43 seems to behave as a client of Hsp90 to maintain its stability in cells.
Both, reduction of Hsp90 levels and, even more so, reduction of Sti1 levels, decreases TDP-43 protein. This result was observed in yeast, cultured cells and tissue of knockout Sti1 mice, suggesting that these interactions are conserved. It should be noted that TDP-43 has important developmental roles in regulating mRNA metabolism (12), and it is possible that the Hsp70 and Hsp90 chaperones are required for these physiological functions. Indeed, both Hsp90 (64)  Given that Hsp90, Hsp70, and Sti1 form a complex with TDP-43 we cannot dismiss the possibility that misfolded and aggregated TDP-43 sequesters Hsp90 and its co-chaperones preventing them from properly performing their many essential functions, which can indirectly contribute to TDP-43 toxicity. This is less likely for the highly expressed Hsp90, but quite plausible for all co-chaperones, which are expressed at much lower levels (67). Finally, our data support the notion that a strong interaction between Hsp90 and some of its co-chaperones increase TDP-43 toxicity by accelerating TDP-43 export from the nucleus or stabilizing toxic TDP-43 conformations in the cytoplasm. All these different possible mechanisms are by no means mutually exclusive and may act together in regulating TDP-43 misfolding and toxicity.
Our data clearly does not provide the basis for a simplistic model whereby increased Hsp90 and its co-chaperone activity ameliorate TDP-43 toxicity and their reduced function exacerbate it.
Rather, we show that the interaction between Hsp90, its co-chaperones, and TDP-43 is much more complex and depends on specific co-chaperones and their expression levels. This complexity plausibly originates from the multitude of different cellular functions of Hsp90 and the specific biophysical properties, cellular localization, and function of the misfolded protein, and the affected neuronal cell types. Future in vitro studies using purified protein and studies in animal models can help to clarify the exact mechanism by which Hsp90 and its co-chaperones' regulation of TDP-43 proteinopathy and toxicity.

Mouse Embryo
Wild type and Sti1 -/mouse embryos were developed and collected as previously described (55).
Super-ovulation was induced by injecting female mice with 5 IU of pregnant mare serum. 48 h after, females were further injected with 5 IU human chorionic gonadotropin. Immediately after this injection, female mice were mated with male mice overnight; the males were removed the following day. Embryos were collected and genotyped at successive time intervals.

Antibodies
The antibodies used in this study are shown in Table 1.

Yeast Transformation
Yeast transformations were performed using the standard PEG/lithium acetate method (57). to monitor the growth of the yeast colonies by taking photographs using a digital camera.

Quantification of yeast growth assays
Agar plates were photographed, and converted to black and white images, which were quantified by ImageJ. Unless indicated otherwise, the pixel count for the third dilution of growth assays under the different conditions were used for quantification after subtraction of the background signal.
Measured values were input into GraphPad prism was used to generate bar graphs and statistical analyses by applying One Way Analysis of Variance, Tukey Post Hoc using IBM SPSS. Error bars represent standard errors. A minimum of three biological replicas were used for quantification.

CRISPR-Cas9 Sti1 knock out
Sti1 knockouts of SN56 cells were generated as previously described (54 Several clones with decreased Sti1 levels were obtained but only clones with complete elimination of STIP1 expression were expanded and used to investigate the role of Sti1 in TDP-43 mediated toxicity and pathology.

Mammalian Cell Viability Assay
The CellTiter-Glo Luminescent Cell Viability Assay (Promega) along with the Live/Dead TM Viability/Cytotoxicity Kit (ThermoFisher Scientific, Cat#L3224) were used in this study.
The CellTiter-Glo Luminescent Cell Viability Assay delivers a highly sensitive read out of cellular fitness and is particularly useful for studies on the toxicity associated with misfolded proteins.
Transfected cells are split into 96 well plates and grown in low glucose and low serum DMEM conditions. This minimal medium increases sensitivity to the toxic effects of protein misfolding in many other systems (71), as well as differentiating N2a cells (69). The cells are incubated at 37°C for 20-24 h. The cell viability assay is then performed according to the supplier's instructions.
Plates were then measured using the Cytation 5 Cell Imaging Multi-Mode Reader (Biotek).

Co-immunoprecipitation
The N2a cells were lysed according to the lysing protocol described in the sedimentation assay. We add lysate and antibody mixture to the washed beads and incubate lysate-antibody-bead mixture at room temperature on a rocker for 2-3 hours. Following incubation, we wash the pelletbead mixture three times in PBS and then remove the supernatant. We add 30 μL of 1X SDS sample buffer. We boil our sample for 5 mins and centrifuge bound sample to remove magnetic beads. We also reboil the frozen total lysate. The lysates are loaded onto a polyacrylamide gel. The gel is then run according to the western blot procedures described above. We probe blot with Rabbit α-HA (Sigma) antibody to Co-IP Sti1. We then strip the blot using the Antibody Stripping Buffer (Gene Bio-Application) according to the manufacturer's protocol and reprobe with α-Hsp90 and α-TDP-43 antibody.

Statistical Analysis
Statistical analysis of viability assays and western blots were done using the GraphPad Prism 6 software. Statistical significance was obtained by performing unpaired t-tests to compare the means and standard deviations between the control data set and the experiment data set (each at a minimum of three biological replicas). Significance levels are indicated using asterisks, where **** is p<0.0001, *** is p<0.001, ** is p<0.01, * is p<0.05. Error bars represent standard errors of the mean.

Competing interests:
The authors declare that they have no competing interests.