Cytoplasmic Relocalization of TAR DNA-Binding Protein 43 Is Not Sufficient to Reproduce Cellular Pathologies Associated with ALS In vitro

Mutations in the gene TARDBP, which encodes TAR DNA-binding protein 43 (TDP-43), are a rare cause of familial forms of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). While the majority of mutations are found in the C-terminal glycine-rich domain, an alanine to valine amino acid change at position 90 (A90V) in the bipartite nuclear localization signal (NLS) of TDP-43 has been described. This sequence variant has previously been shown to cause cytoplasmic mislocalization of TDP-43 and decrease protein solubility, leading to the formation of insoluble aggregates. Since the A90V mutation has been described both in patients as well as healthy controls, its pathogenic potential in ALS and FTD remains unclear. Here we compare properties of overexpressed A90V to the highly pathogenic M337V mutation. Though both mutations drive mislocalization of the protein to the cytoplasm to the same extent, M337V produces more significant damage in terms of protein solubility, levels of pathogenic phosphorylation, and formation of C-terminal truncated protein species. Furthermore, the M337V, but not the A90V mutant, leads to a downregulation of histone deacetylase 6 and Ras GTPase-activating protein-binding protein. We conclude that in the absence of another genetic or environmental ‘hit’ the A90V variant is not sufficient to cause the deleterious phenotypes associated with ALS and FTD, despite prominent cytoplasmic protein relocalization of TDP-43.


INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a fast progressing neurodegenerative disorder associated with atrophy of upper and/or lower motor neurons leading to muscle wasting due to denervation. It is the third most frequently occurring neurodegenerative disease after Alzheimer's disease and Parkinson's disease (Hirtz et al., 2007). Death usually occurs within 2-3 years after diagnosis as a result of respiratory failure (Chio et al., 2011). Frontotemporal dementia (FTD) on the other hand is characterized by degeneration of neurons in the frontal and temporal lobes and is associated with impairment of cognitive abilities and language, as well as behavioral changes (Bang et al., 2015). Today, both diseases fall under the umbrella term of TDP-43 proteinopathies, in reference to the occurrence of cytoplasmic inclusions of the TAR DNA-binding protein 43 (TDP-43) in glial and neuronal cells: 97% of ALS as well as almost 50% of FTD cases present with such TDP-43-positive inclusions Mackenzie et al., 2007). In this pathological state, TDP-43 is found ubiquitinated, truncated, highly phosphorylated and aggregated (Arai et al., 2006;Neumann et al., 2006).
TAR DNA-binding protein 43 is a ubiquitously expressed RNA-and DNA-binding protein. In the central nervous system, TDP-43 binds over 6000 RNA targets. It has been shown to regulate the levels of 600 mRNA species and the alternative splicing of over 950 species (Polymenidou et al., 2011). RNA binding is mediated by two RNA recognition motifs, RRM1 and RRM2. In addition, the protein harbors a nuclear export signal (NES) as well as a bipartite NLS, and a C-terminal prion-like glycine-rich domain (Figures 1A,C) (Buratti and Baralle, 2001;Winton et al., 2008a). Mutations in the TDP-43 gene TARDBP in ALS were first described in 2008 (Gitcho et al., 2008;Kabashi et al., 2008;Sreedharan et al., 2008;Van Deerlin et al., 2008). The frequency of TDP-43 mutations is 1-5% in both sporadic and familial forms of ALS (Guerreiro et al., 2008;Kabashi et al., 2008;Kuhnlein et al., 2008;Iida et al., 2012;Zou et al., 2012).
The vast majority of TARDBP mutations are missense mutations, with all but three located in the C-terminal glycinerich domain (Buratti, 2015). Two mutations, P112H and D169G, are located in the RRM1 domain (Kabashi et al., 2008;Buratti, 2015;Moreno et al., 2015). The third, an alanine to valine substitution at residue 90 (A90V), is found between the bipartite NLS (Winton et al., 2008b;Chiang et al., 2012). Winton et al. (2008b) showed that the A90V mutation leads to aberrant cytoplasmic localization and decreased solubility of TDP-43, two pathological hallmarks of TDP-43 proteinopathies, in vitro. Another study found that under stress conditions, induced pluripotent stem cells with the A90V mutation obtained from an FTD/ALS patient showed increased cell death, cytoplasmic localization of TDP-43 and downregulation of microRNA-9 (Zhang et al., 2013). However, the variant has been described both in FTD and ALS patients and control individuals (Guerreiro et al., 2008;Kabashi et al., 2008;Sreedharan et al., 2008;Winton et al., 2008b;Chiang et al., 2012). Further studies described the occurrence of the A90V variant at a higher frequency in Alzheimer's disease patients compared to controls (Brouwers et al., 2010;Vanden Broeck et al., 2015). Thus the detection of the A90V variant in healthy controls combined with the biochemical findings reported by Winton et al. (2008b) leave its pathogenic nature unclear.
In this study, we compared the effects of the A90V variant of TDP-43 to both wild-type (WT) and the M337V mutation, a pathogenic mutation associated with familial ALS (Sreedharan et al., 2008), in the SH-SY5Y neuroblastoma cell line. We provide evidence that while the A90V amino acid change phenocopies the M337V mutation in terms of cytoplasmic mislocalization, it does not demonstrate such a severe phenotype in terms of formation of insoluble high molecular weight aggregates and C-terminal fragments, pathological phosphorylation or downregulation of TDP-43 targets G3BP and histone deacetylase 6 (HDAC6). We thus conclude that relocalization of TDP-43 from the nucleus to the cytoplasm alone, while a classic pathological feature of TDP-43 proteinopathies, is not in itself sufficient to cause the complete cellular disease signature. Furthermore, we propose that the A90V variant of TDP-43 could be a rare risk factor for ALS that requires another "hit" for the development of the full-blown disease pathology.

Plasmids
All plasmids were purchased from and customized by Origene. cDNA sequences were based on the accession number NM_007375.3 for human TARDBP. The WT construct was cloned into the pCMV6-AN-myc vector for expression of N-terminally tagged WT or mutant TDP-43 protein.

Cell Culture
SH-SY5Y cells were grown at 37 • C with 5% CO 2 in a humidified atmosphere in minimum essential medium (MEM; Thermo Fisher Scientific) supplemented with L-glutamine, 10% fetal bovine serum and 1% penicillin/streptomycin. Cells were subcultured every 3-4 days. For immunofluorescence analysis, cells were grown on glass coverslips coated with 1 mg/mL poly-L-lysine (Sigma) in 24-well plates (Cellstar). For Western blot analysis, cells were grown in 6-well plates (Costar). Cells were transiently transfected using Fugene HD (Promega) following the manufacturer's instructions 24 h after seeding (Fugene HD:DNA ratio 3:1). Total transfection time was 48 h.

Immunofluorescence Staining
Cells were washed in phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (in PBS) for 15 min followed by permeabilization in 0.25% Triton-X (in PBS) for 10 min. Ten percent normal goat serum (in PBS; Abcam) was used as a blocking reagent. Cells were blocked for 1 h at room temperature and incubated overnight at 4 • C in primary antibody solution.
The next day, cells were washed with PBS and incubated for 1 h in secondary antibody solution. Coverslips were mounted with Prolong Gold Antifade Mountant with DAPI (Thermo Fisher Scientific). Image acquisition was performed using a Nikon A1 confocal/Eclipse Ti inverted microscope system and NIS Elements software (Nikon).

Sequential Extraction of Insoluble Protein Aggregates
The extraction protocol was adapted from Cohen et al. (2012). Cells were grown in 6-well plates and transfected 24 h after seeding. After 48 h, transfected cells were lysed in 300 µL radio-immunoprecipitation assay (RIPA) buffer supplemented with 2 mM EDTA, 1 mM EGTA, protease inhibitors (Complete, Roche), and phosphatase inhibitors (PhosStop, Roche). Lysates were sonicated 2x 15 s with 20% maximum amplitude and centrifuged for 30 min at 100,000 × g and 4 • C (Beckman Optima TLX ultracentrifuge with TLA100.3 rotor and Delrin adaptors). The supernatant was collected as the RIPA-soluble fraction. The pellet was washed in RIPA buffer and centrifuged for an additional 30 min at 100,000 × g and 4 • C. The supernatant was discarded and the pellet was re-extracted in 100 µL urea buffer [7 M urea, 2 M thiourea, 30 mM Tris pH 8.5, 4% 3-[(3-Cholamidopropyl) dimethylammonio]-1propanesulfonate (CHAPS, Sigma)]. The samples were sonicated 2x 15 s and centrifuged at room temperature for 30 min at 100,000 × g. The supernatant was collected as the urea-soluble fraction.

Immunoblotting
Cells were lysed in RIPA buffer supplemented with 2 mM EDTA, 1 mM EGTA, protease inhibitors (Complete, Roche) and phosphatase inhibitors (PhosStop, Roche), and the protein concentration was determined by bicinchoninic acid (BCA) assay. RIPA and urea fractions obtained from sequential extractions were diluted 1:1 with RIPA buffer. Proteins were separated using standard Tris-glycine SDS-PAGE gels (polyacrylamide concentrations 8% for HDAC6, all others 10%) and transferred onto nitrocellulose (BioRad) or polyvinylidene difluoride (PVDF; Millipore) membranes. All blocking and antibody incubation steps were performed either in 5% milk in Tris-buffered saline (TBS) supplemented with 0.1% Tween-20 (TBS-T) or in 3% bovine serum albumin (BSA) in TBS-T. Western blots were developed with enhanced chemiluminescent substrates (ECL). Digital images were acquired with a ChemiDoc MP imaging system (BioRad) and quantified with Image Lab 5 (BioRad). Where necessary, blots were stripped with stripping buffer for 15 min (Restore, Thermo Fisher Scientific) and reprobed with loading control antibodies.

Statistical Analysis
All data were expressed as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism 6 software (San Diego, CA, USA). Normality was established using the Shapiro-Wilk test. Mean differences between groups were analyzed by one-way ANOVA and Tukey's multiple comparisons test. The threshold for statistical significance was p ≤ 0.05. Experiments were replicated a minimum of three times.

In silico Analysis of TDP-43 Protein and Determining Whether It Can Predict the Impact of Mutations on Protein Structure
Nuclear magnetic resonance and X-ray structural analysis of individual domains of WT and mutant TDP-43 continue to reveal insights into the relationships between disease-related mutations and the biophysical stability of the protein. We reasoned that assembling a three-dimensional model of TDP-43 may help us predict the structural impact of the A90V variant and support our in vitro efforts in elucidating its effects. Thus, we attempted to collate all structural information obtained from NMR and X-ray crystallography available in the public domain and obtain a complete 3D model of TDP-43 (Figure 1). However, structures including all domains which might enable a comprehensive examination of the effects of specific mutations on overall structure and stability are challenging and not yet realized. The human TDP-43 sequence ( Figure 1A) and NMR structures ( Figure 1B) in Figure 1 highlight the NLS, NES, RRM1, RRM2, glycinerich domain, as well as the A90V and M337V mutation sites and the S409/S410 phosphorylation site (He et al., 2004;Suzuki et al., 2005;Kuo et al., 2014;Lim and Song, 2015;Mompean et al., 2015Mompean et al., , 2016Chiang et al., 2016;Lim et al., 2016). The piecewise solutions of NMR and X-ray structures provide valuable domain-level information, but key portions of the protein remain either unresolved or missing. Iterative homology modeling and loop building techniques can be used to infer composite structures from their individual domains. However, in the case of TDP-43 a suitable template structure spanning across multiple domains does not exist, prohibiting the ability to confidently assess the overall structure (only an approximate schematic of the overall structure is shown, Figure 1C) and thus the impact of the A90V and M337V amino acid substitutions. Without a priori assumptions of the impact of the A90V variant based on a complete model, we then proceeded to investigate the consequences of expressing the A90V variant on common pathological phenotypes associated with disease: cytoplasmic relocalization, and the formation of insoluble TDP-43 and TDP-43 protein fragments. We contrasted the A90V variant with the well-characterized mutant M337V, which causes autosomal-dominant familial ALS (Sreedharan et al., 2008).

The TDP-43 A90V Variant Phenocopies the M337V Variant by Causing Cytoplasmic Relocalization of TDP-43
According to Winton et al. (2008b), transient expression of the A90V variant of TDP-43 in QBI-293 cells results in the aberrant relocalization from the nucleus to the cytoplasm. We first verified expression of N-terminally myc-tagged TDP-43 constructs in SH-SY5Y cells by Western blot using an N-terminal TDP-43 antibody (Figure 2A). Steady state levels of total TDP-43 protein were significantly increased in cells expressing M337V mutant TDP-43 compared to WT expressing cells (0.46 ± 0.19-fold increase over WT; Tukey's multiple comparisons analysis; p = 0.04), whereas there was no significant difference in cells expressing the A90V TDP-43 variant compared to WT protein (0.17 ± 0.09fold over WT; Tukey's multiple comparisons analysis; p = 0.57; Figure 2B; one-way ANOVA, F 2,15 = 3.757; P = 0.0316). We confirmed that this increase in M337V TDP-43 protein levels was not caused by increased levels of mRNA, which were comparable between genotypes (Supplementary Figure  S1).
We then performed immunofluorescence staining of transiently transfected cells in order to confirm the relocalization phenotype observed by Winton et al. (2008b) ( Figures 2C,D). In agreement with the previous report (Winton et al., 2008b), we observed a significant increase in the proportion of cells with cytoplasmic TDP-43 localization in A90V, compared to WT TDP-43 expressing cells (WT: 28.4 ± 3.0% of cells; A90V: 41.7 ± 3.7% of cells; Tukey's multiple comparisons analysis; p = 0.041). Interestingly, the proportion of transfected cells with cytoplasmic TDP-43 localization with A90V was very similar compared to the highly pathogenic mutant M337V (WT: 28.4 ± 3.0% of cells; M337V: 43.0 ± 1.6% of cells; Tukey's multiple comparisons analysis; p = 0.028; one-way ANOVA F 2,6 = 7.679; P = 0.0222; Figure 2D). In a small number of transfected cells analyzed by immunofluorescence staining we detected a punctate staining pattern for TDP-43 in the nucleus (Figure 2E). These TDP-43positive puncta were only rarely observed in SH-SY5Y cells transfected with WT TDP-43 (0.28 ± 0.10% of transfected cells), but were found in significantly higher numbers in cells expressing the pathogenic mutant M337V (2.00 ± 0.60% of transfected cells; Tukey's multiple comparisons analysis; p = 0.049; Figure 2F). However, we saw no significant difference in puncta formation between WT and A90V TDP-43 expressing cells (A90V: 1.06 ± 0.44% of transfected cells; Tukey's multiple comparisons analysis; p = 0.44; one-way ANOVA F 2,9 = 3.94; P = 0.059). In order to more closely investigate the composition of these TDP-43-containing puncta, we performed co-staining experiments with two nuclear markers: the nuclear stress body marker SAFB and the paraspeckle marker PSF (Figure 3). We qualitatively distinguished between transfected cells with small, numerous nuclear TDP-43-positive puncta (Figure 2E, asterisk; Figures 3A-C, lower panel) and cells with few, but large, puncta (Figure 2E, arrowhead; Figures 3A-C, upper panel). We observed no co-localization between TDP-43 protein and either SAFB or PSF in small puncta (Figures 3A,B). In cells harboring large TDP-43-positive puncta, both SAFB and PSF were found to be excluded from these structures. We hypothesized that these puncta represent intranuclear inclusions of insoluble, aggregated TDP-43 protein. As aggregated TDP-43 is found pathologically ubiquitinated (Neumann et al., 2006), we stained myc-TDP-43transfected cells with an anti-ubiquitin antibody ( Figure 3C). As expected, we detected diffuse ubiquitin staining in both the nucleus and cytoplasm. Both small and large TDP-43 nuclear puncta were shown to co-localize with ubiquitin, indicating that they are bona fide aggregates of ubiquitinated TDP-43 protein.

The A90V TDP-43 Variant Has Subtle Effects on Protein Solubility and Pathologic Phosphorylation Compared to the M337V Mutation
In order to further investigate how the A90V and M337V mutations affect the cell biology of TDP-43, we measured a range of cellular phenotypes which have been implicated in disease pathogenesis. Initially, we looked at the formation of insoluble TDP-43 aggregates. We fractionated lysates of transfected SH-SY5Y cells into RIPA-soluble and RIPA-insoluble, urea-soluble fractions by high speed centrifugation and separated proteins by SDS-PAGE. We used a TDP-43 antibody recognizing the N-terminus of TDP-43 to assess total TDP-43 protein levels in the RIPA and urea fractions ( Figure 4A). In line with previous reports (Johnson et al., 2009;Bilican et al., 2012;Janssens et al., 2013), we found that the M337V mutation caused accumulation of insoluble TDP-43 in the urea fraction compared to WT (3.21 ± 1.14-fold increase over WT; Tukey's multiple comparisons analysis; p = 0.023; one-way ANOVA F 2,9 = 5.38; P = 0.029). While an enrichment of TDP-43 in the urea fraction was also observed in A90V-expressing cells (1.59 ± 0.38-fold increase compared to WT), the increase was not statistically significant (Tukey's multiple comparisons analysis; p = 0.283; Figure 4B). We used a phospho-specific TDP-43 antibody raised against the disease-relevant phosphorylation sites at S409/S410 (Neumann et al., 2009) to assess the phosphorylation status of the insoluble TDP-43 protein species (Figure 4A). We normalized the phospho-signal in the urea-soluble fractions to the total TDP-43 detected in these fractions to control for the relative increase in insoluble TDP-43 protein in mutant expressing cells (Figure 4C). Immunoblotting with this antibody revealed significantly increased phosphorylation levels at the S409/S410 residues in the urea-soluble fraction of M337V TDP-43 expressing cells compared to cells expressing WT TDP-43 (6.38 ± 2.56-fold increase compared to WT; Tukey's multiple comparisons analysis; p = 0.036). We detected a nonstatistically significant increase in S409/S410 phosphorylation in cells transfected with the A90V variant (1.35 ± 0.48-fold increase compared to WT; Tukey's multiple comparisons analysis; p = 0.80; one-way ANOVA F 2,9 = 5.015; P = 0.0344). Taken together, these results indicate that while the A90V and M337V mutations induce similar levels of cytoplasmic relocalization of TDP-43, a pathology commonly associated with disease, the effects of the M337V variant on protein solubility and phosphorylation are far greater than those of the A90V mutation. In fact there looks to be a spectrum of pathology in these transfected cell models with WT and M337V TDP-43 at the opposite extremes, with the A90V TDP-43 somewhere in between, in the absence of any additional stressors.

M337V Mutant TDP-43 Readily Forms C-Terminal Fragments Compared to A90V and WT
In addition to abnormal phosphorylation as well as decreased protein solubility and aberrant cellular localization, another pathological feature of TDP-43 is the formation of C-terminal truncated protein fragments (Neumann et al., 2006;Mackenzie et al., 2007;Hasegawa et al., 2008). In order to detect C-terminal protein fragments not picked up by the N-terminal TDP-43 antibody used in Figure 4 we extracted soluble and insoluble TDP-43 from transfected SH-SY5Y cells and used a C-terminal antibody (Figures 5A,B). We identified two smaller protein species present only in TDP-43-transfected cells, but not in the vector control cells (Figure 5B). The larger fragment ran close to the 34 kDa marker, while the smaller fragment ran close to the 26 kDa marker. C-terminal TDP-43 fragments with molecular weights of 35 kDa (CTF35) and 25 kDa (CTF25) have been described in a variety of cellular and animal models (Winton et al., 2008a;Xu et al., 2011;Bilican et al., 2012;Choksi et al., 2014;D'Alton et al., 2014;Li et al., 2015), and we will thus refer to the identified truncated TDP-43 species as CTF35 and CTF25. We quantified the amount of both fragments in the RIPAsoluble and urea-soluble fractions. We normalized to the amount of full-length TDP-43 in each sample in order to account for the increase of total TDP-43 protein. In the RIPA-soluble fraction, we observed an accumulation of CTF35 in M337V-transfected cells compared to either WT or A90V-TDP-43 (2.03 ± 0.44fold increase over WT, p = 0.005; 0.68 ± 0.25-fold increase over A90V, p = 0.044; Tukey's multiple comparisons analysis; one-way ANOVA F 2,6 = 13.94; P = 0.0056; Figure 5C). However, we did not detect a significant increase of the smaller CTF25 in mutanttransfected cells (M337V: 1.09 ± 0.6-fold increase over WT, p = 0.19; 0.64 ± 0.52-fold increase over A90V, p = 0.36; Tukey's multiple comparisons analysis; one-way ANOVA F 2,6 = 2.166; P = 0.196). In the urea-soluble fractions the effect of the M337V mutant on the formation of these C-terminal fragments was even more striking: we detected a 20.0 ± 6.6-fold increase in the 25 kDa fragment in the M337V mutant compared to WT (Tukey's multiple comparisons analysis; p = 0.026) and a 4.3 ± 1.7-fold increase compared to the A90V variant (Tukey's multiple comparisons analysis; p = 0.049; one-way ANOVA F 2,6 = 7.657; P = 0.0223; Figure 5D). We could not detect any 35 kDa fragment in the urea fraction of WT-TDP-43 transfected cells and found a 5.4 ± 1.1-fold increase in CTF35 levels in M337V-compared to A90V-transfected cells (M337V compared to WT: p = 0.0008: compared to A90V: p = 0.0019; Tukey's multiple comparisons analysis; one-way ANOVA F 2,6 = 31.53; P = 0.0007).

The M337V TDP-43 Variant Down-regulates the Expression of G3BP and HDAC6
We hypothesized that the M337V mutant of TDP-43 not only impacts TDP-43 localization and post-translational modifications in a manner more detrimental than the A90V variant, but that the downstream functional consequences of expressing M337V TDP-43 might reflect this more severe phenotype. Such a functional consequence could be altered regulation of transcriptional targets of TDP-43. Previous studies have shown that both depletion of endogenous TDP-43 as well as overexpression of R361S mutant TDP-43 lead to a decrease in G3BP, a regulator of stress granule assembly (McDonald et al., 2011;Aulas et al., 2012). We investigated expression levels of G3BP following 48 h of TDP-43 overexpression and found that only M337V, but not A90V, mutant TDP-43 significantly downregulated G3BP compared to WT TDP-43 and mock control (25.8 ± 8.3% compared to WT, p = 0.01; 23.0 ± 7.8% compared to mock, p = 0.03; Tukey's multiple comparisons analysis; one-way ANOVA F 3,20 = 4.861; P = 0.0106; Figures 6A,C). Furthermore, levels of HDAC6, a known target of TDP-43 that is likewise downregulated upon , and (E) acetylated α-tubulin levels (N = 6). All results represent mean band signal ± SEM. * p < 0.05, * * p < 0.01 (One-way ANOVA followed by Tukey's multiple comparisons test).

DISCUSSION
To date, genetic studies have revealed 52 missense mutations in the TDP-43 gene TARDBP as well as one insertion/deletion and one nonsense mutation (Buratti, 2015). While most mutations cluster in the C-terminal glycine-rich domain, the A90V amino acid exchange lies in the bipartite NLS. The pathogenic nature of this variant remains uncertain; in vitro studies suggest that A90V-TDP-43 is more prone to cytoplasmic mislocalization and aggregation, two pathological changes associated with ALS and FTD (Winton et al., 2008b). Furthermore, expression of A90V TDP-43 fails to fully ameliorate the pupal lethality and neuronal loss phenotypes caused by loss of the Drosophila ortholog of TARDBP (Vanden Broeck et al., 2015). Genetic studies, however, have revealed the presence of the A90V variant both in patients suffering from ALS and FTD as well as in healthy controls (Guerreiro et al., 2008;Kabashi et al., 2008;Sreedharan et al., 2008;Winton et al., 2008b;Chiang et al., 2012). It is thus likely that A90V is not a highly penetrant disease-associated mutation, but a genetic risk factor for TDP-43 proteinopathies. In order to gain a deeper understanding of the pathogenic potential of the A90V variant, we initially conducted analyses of available structural information for TDP-43 to see if we could generate hypotheses to test based on the site of the mutations. But unfortunately, due to the lack of complete NMR and X-ray structures, we were unable to predict the impact of the A90V variant of TDP-43 in silico (Figure 1). We then resorted to more traditional in vitro studies to measure TDP-43 mislocalization, changes in protein solubility and phosphorylation, and the effects on their expression on known targets of TDP-43. We compared the impact of A90V with that of the M337V mutation, a diseaseassociated mutation frequently used for in vitro and in vivo studies of ALS pathogenesis (Liachko et al., 2010;Zhou et al., 2010;Bilican et al., 2012;Janssens et al., 2013;Mutihac et al., 2015). When we initially looked to reproduce the nuclear localization phenotype observed by Winton et al. (2008b) we found an almost identical increase in the number of cells with cytoplasmic TDP-43 with both the A90V and M337V variants. Thus, we concluded that the A90V variant of TDP-43 phenocopies the cytoplasmic relocalization phenotype associated with ALS found in the disease-associated M337V mutant. This strong effect is perhaps not surprising given the fact that the A90V variant lies within the bipartite NLS and could thus potentially interfere with nucleo-cytoplasmic shuttling of TDP-43.
In a small subset of transfected cells, we detected dot-like structures in the nucleus that stained positive for TDP-43. These TDP-43-positive nuclear structures were observed more frequently in cells transfected with the M337V, but not the A90V, mutant compared to WT. Instances of defined, patterned nuclear staining of TDP-43 in vitro have been described in the literature: overexpression of certain deletion mutants or the M337V mutant has been shown to cause the formation of nuclear inclusions of aggregated TDP-43 (Winton et al., 2008a;Nonaka et al., 2009a;Tripathi et al., 2014), while other papers describe co-localization of TDP-43 with nuclear bodies under certain circumstances, such as stress bodies under heat shock conditions, or NEAT1_2-containing paraspeckles (Nishimoto et al., 2013;Udan-Johns et al., 2014). We sought to define the nature of the observed nuclear TDP-43-positive puncta by performing co-localization experiments using antibodies against SAFB as a marker of nuclear stress bodies and PSF as a paraspeckle marker. We did not observe co-localization of either SAFB or PSF with TDP-43-positive inclusions, indicating that WT and mutant TDP-43 are not recruited into nuclear stress bodies or paraspeckles. However, we found that nuclear puncta of TDP-43 co-stain with ubiquitin, suggesting that they represent inclusions of aggregated, ubiquitinated TDP-43 in the nucleus. In this light, the increased frequency of these TDP-43-positive inclusions in the M337V mutant is probably the result of a higher aggregation propensity.
We investigated the aggregation propensity of the A90V and M337V mutants further by sequential extraction of total cell lysates. We found a stark increase in M337V in the insoluble fraction, in line with published literature showing increased aggregate formation of M337V-TDP-43 in vitro, in vivo, and in culture systems (Johnson et al., 2009;Zhou et al., 2010;Bilican et al., 2012). Compared to WT protein, the increase we observed in insoluble A90V TDP-43 was not statistically significant. Phosphorylation levels of insoluble, aggregated TDP-43 species at the pathologically relevant S409/410 residues were also markedly increased in the M337V mutant, while the effect in the A90V mutant was modest and not statistically significant. Analysis of total unfractionated lysate also revealed a significant increase in total mutant M337V TDP-43 protein compared to WT, likely due to the longer half-life of M337V-TDP-43 previously reported (Ling et al., 2010;Watanabe et al., 2013). While this increase in total M337V mutant protein could possibly help drive some of the described phenotypes, we believe the modest effect size (less than 50% compared to WT, less than 20% compared to A90V) is not sufficient to cause some of the more striking phenotypes we have observed, such as the effects on G3BP and HDAC6 expression levels.
C-terminally cleaved TDP-43 fragments are another pathological hallmark of ALS and FTD, and expression of TDP-43 in cell culture models can give rise to cleavage products of 35 and 25 kDa in the absence or presence of various stressors (Neumann et al., 2006;Hasegawa et al., 2008;Chiang et al., 2012;Yang et al., 2014;Li et al., 2015). We found an enrichment of two fragments of approximately 35 and 25 kDa in the insoluble fraction of M337V-transfected cells compared to both WT and A90V. Almost no C-terminal fragments were detected in the insoluble fraction of A90V-transfected cells, suggesting that A90V-TDP-43 is less prone to protease cleavage compared to the familial ALS mutant M337V. This phenotype is especially striking as several papers have suggested a crucial role for C-terminal TDP-43 fragments in enhancing cellular toxicity and aggregation propensity, especially the 25 kDa fragment (Johnson et al., 2008;Nonaka et al., 2009b;Zhang et al., 2009;Kitamura et al., 2016).
In addition to the effects of the A90V variant on TDP-43 aggregation, post-translational modifications and truncation, we investigated changes in known targets of TDP-43. We found that protein levels of G3BP are reduced in SH-SY5Y cells expressing M337V, but not A90V TDP-43. G3BP is a component of cytoplasmic stress granules, transient structures made from proteins and untranslated mRNAs, which form after insults such as heat shock and oxidative stress (Nover et al., 1989;Tourriere et al., 2003). Similarly, we found that only the M337V, but not the A90V mutant, leads to a reduction of the histone deacetylase HDAC6 in SH-SY5Y cells; HDAC6 has previously been claimed to be a critical component of stress granule assembly (Kwon et al., 2007). TDP-43 has been shown to be both recruited to and to regulate assembly of stress granules (Colombrita et al., 2009;McDonald et al., 2011). That the A90V variant of TDP-43, in contrast to the M337V variant, does not cause a downregulation of G3BP or HDAC6 could suggest that A90V does not interfere with stress granule assembly and dynamics, while the diseaseassociated M337V mutant does. In this regard, the M337V mutant causes a loss of normal function of TDP-43 regulation of stress granule assembly -while the A90V variant does not. Structurally, stress granules are protein-rich liquid droplets formed in a process termed liquid-liquid phase separation (LLPS). Recent studies have shown that mutations in the C-terminal domain, including the M337V mutation, inhibit this droplet formation by interrupting the interaction of α-helixes, driving the proteins to form aggregates instead (Molliex et al., 2015;Conicella et al., 2016;Schmidt and Rohatgi, 2016). Since the A90V variant is far removed from the α-helix in the C-terminal domain, it is likely that it does not interfere with the helical contact that drives droplet formation. However, further studies will be required to investigate whether A90V has more subtle effects on TDP-43 LLPS, droplet formation, and stress granule assembly.
In summary, the results of this study show that the A90V variant of TDP-43, which has been described as a rare variant found both in ALS and FTD patients and healthy controls, shows modest effects on protein solubility, phosphorylation, and truncation when compared to the disease-causing M337V mutation. In addition, in contrast to the M337V mutant, the A90V variant does not cause impairment of HDAC6 or G3BP expression. These findings are especially striking since we observed comparable levels of cytoplasmic TDP-43 relocalization in both A90V-and M337V-transfected cells, suggesting that cytoplasmic relocalization of TDP-43 protein is not by itself sufficient to cause the other molecular phenotypes associated with disease. Furthermore, we speculate that due to these modest in vitro effects observed with the A90V variant, carriers might not develop a disease phenotype in the absence of a "second hit." However, it is possible that the A90V mutant could lower the threshold required for an additional stressor or other genetic risk factor to cause the catastrophic effects on TDP-43 and its functions observed in ALS and FTD. Thus, we conclude that the A90V variant of TDP-43 could be a rare genetic risk factor for TDP-43 proteinopathies. However, further studies will be required to gain an insight into the interplay between A90V as a genetic risk factor and other genetic and environmental risk factors associated with ALS and FTD.

AUTHOR CONTRIBUTIONS
HW and NB designed experiments; HW, JC, and LD carried out experiments; HW and JC analyzed the data; SW designed the composite NMR structure; JM and TD provided technical assistance; HW, NB, SW, and SM wrote the manuscript. All authors reviewed and approved the final version of the manuscript.

FUNDING
This work was supported by funding from AstraZeneca, National Institutes of Health (NIH)-National Institute of Neurological Disorders and Stroke Grants NS051195, NS056359, NS081735, R21NS080064, and NS087662 (SM) and NIH-National Institute of Mental Health Grants MH097446 and MH106954, and DOD Grants AR140209 (SM).