Intranuclear Aggregation of Mutant FUS/TLS as a Molecular Pathomechanism of Amyotrophic Lateral Sclerosis*

Background: Abnormal accumulation of mutant FUS/TLS is a pathological change in patients with amyotrophic lateral sclerosis (ALS). Results: A pathogenic mutation, G156E, increases propensities of FUS/TLS for aggregation in vitro and in vivo. Conclusion: Intranuclear aggregation of mutant FUS/TLS is a molecular pathomechanism of ALS. Significance: A loss of functional TLS/FUS in the nucleus will lead to neurodegeneration. Dominant mutations in FUS/TLS cause a familial form of amyotrophic lateral sclerosis (fALS), where abnormal accumulation of mutant FUS proteins in cytoplasm has been observed as a major pathological change. Many of pathogenic mutations have been shown to deteriorate the nuclear localization signal in FUS and thereby facilitate cytoplasmic mislocalization of mutant proteins. Several other mutations, however, exhibit no effects on the nuclear localization of FUS in cultured cells, and their roles in the pathomechanism of fALS remain obscure. Here, we show that a pathogenic mutation, G156E, significantly increases the propensities for aggregation of FUS in vitro and in vivo. Spontaneous in vitro formation of amyloid-like fibrillar aggregates was observed in mutant but not wild-type FUS, and notably, those fibrils functioned as efficient seeds to trigger the aggregation of wild-type protein. In addition, the G156E mutation did not disturb the nuclear localization of FUS but facilitated the formation of intranuclear inclusions in rat hippocampal neurons with significant cytotoxicity. We thus propose that intranuclear aggregation of FUS triggered by a subset of pathogenic mutations is an alternative pathomechanism of FUS-related fALS diseases.


Dominant mutations in FUS/TLS cause a familial form of amyotrophic lateral sclerosis (fALS), where abnormal accumulation of mutant FUS proteins in cytoplasm has been observed as a major pathological change. Many of pathogenic mutations have been shown to deteriorate the nuclear localization signal in FUS and thereby facilitate cytoplasmic mislocalization of mutant proteins. Several other mutations, however, exhibit no effects on the nuclear localization of FUS in cultured cells, and their roles in the pathomechanism of fALS remain obscure.
Here, we show that a pathogenic mutation, G156E, significantly increases the propensities for aggregation of FUS in vitro and in vivo. Spontaneous in vitro formation of amyloid-like fibrillar aggregates was observed in mutant but not wild-type FUS, and notably, those fibrils functioned as efficient seeds to trigger the aggregation of wild-type protein. In addition, the G156E mutation did not disturb the nuclear localization of FUS but facilitated the formation of intranuclear inclusions in rat hippocampal neurons with significant cytotoxicity. We thus propose that intranuclear aggregation of FUS triggered by a subset of pathogenic mutations is an alternative pathomechanism of FUS-related fALS diseases.
Fused in sarcoma (FUS), 3 also called as TLS (1), is a DNA/ RNA-binding protein involved in physiological processes related to RNA metabolism in particular (2)(3)(4). Recently, dominant mutations have been identified in the FUS gene to cause a familial form of amyotrophic lateral sclerosis (fALS) (5,6). Wild-type FUS protein is localized mostly in the nucleus of a motor neuron, but a subset of pathogenic mutations in FUS was found to facilitate its cytoplasmic mislocalization (5,6). Upon mutations, FUS would hence lose its physiological functions performed in the nucleus, possibly contributing to the reduced viability of cells.
Notably, the C-terminal region of FUS is a hot spot for pathogenic mutations (Fig. 1), and those C-terminal mutations have been shown to weaken the affinity of FUS with Kap␤2, thereby resulting in the cytoplasmic mislocalization of mutant FUS (9). Despite this, many fALS-causing mutations have been reported also in the N-terminal SYGQ region as well as RGG1 region of FUS ( Fig. 1) and did not affect the nuclear localization of FUS at least in cultured cells (10). In addition, to our knowledge, neuropathological and biochemical analysis of spinal cords of fALS patients with those mutations have not been available so far; therefore, it remains unknown if FUS with a mutation in its SYGQ and RGG1 regions undergoes cytoplasmic mislocalization under pathological conditions. FUS is an intrinsically aggregation-prone protein even without any mutations (11). Given the sequence analysis predicting the high aggregation propensities at the N-terminal SYGQ region of FUS (12), we suspect that pathogenic mutations at the SYGQ region modulate the aggregation kinetics of FUS. In this study, therefore, we have examined the effects of pathogenic mutations at SYGQ and RGG1 regions on the aggregation propensities of FUS proteins. Among the mutations tested (G156E, G225V, M254V, and P525L), introduction of the G156E mutation at the SYGQ region was found to render FUS highly prone to aggregation in vivo as well as in vitro. Such insoluble aggregates exhibited fibrillar morphologies and were capable to triggering the aggregation of wild-type FUS through a seeding reaction. Based upon these results, an alternative pathomechanism of FUS-related fALS has been discussed in which mutations increase the aggregation propensities of FUS proteins.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant FUS Proteins-For preparation of GST-FUS proteins, cDNA of human FUS was cloned in a multiple cloning site (BamHI and SalI) of a vector, pGEX6P-2 (GE Healthcare). A plasmid for expression of GST-FUS-His was prepared by seamlessly inserting six consecutive CAT codons at the C-terminal end of the human FUS coding sequence in the above pGEX6P-2 plasmid containing the GST-FUS gene. Mutations were introduced by the In-Fusion PCR method, and all constructs examined in this study were confirmed by DNA sequencing.
After transfection of a plasmid in Escherichia coli (Rosetta TM (DE3)), expression of GST-FUS and GST-FUS-His proteins was induced by addition of 0.1 mM isopropyl ␤-D-thiogalactoside, and the cultures were shaken at 20°C for 46 h. Cells were lysed in PBS containing 5 mM MgCl 2 , 2% Triton X-100, 6.7 mg/liter of DNase I, 0.12 g/liter of lysozyme and a protease inhibitor mixture, cOmplete EDTA-free (Roche Applied Science), and centrifuged to collect soluble supernatant, which was mixed with glutathione-Sepharose 4B resins (GE Healthcare). After being washed with a TN-trehalose buffer (50 mM Tris, 100 mM NaCl, and 200 mM trehalose) at pH 8.0, the resins were treated with a TN-trehalose buffer containing 20 mM reduced glutathione to elute the bound GST-FUS/GST-FUS-His proteins. The eluted GST-FUS-His proteins were then loaded to HisTrap HP column (GE Healthcare) fitted to BioLogic LP (Bio-Rad) and washed with a buffer containing 50 mM Tris, 100 mM NaCl, and 10 mM imidazole, pH 8.0. GST-FUS-His proteins were then eluted from the column by a TN-trehalose buffer with 250 mM imidazole, pH 6.8 (or 8.0, see text), concentrated by using Vivaspin 15 (MWCO: 10 kDa, Sartorius) and checked by SDS-PAGE. The protein concentration was estimated by using Bio-Rad Protein Assay, in which bovine serum albumin was used as the standard.
Electrophoresis-A sample solution containing 10 g of total proteins was mixed with SDS-PAGE sample buffer containing ␤-mercaptoethanol and loaded on a 10% polyacrylamide gel after boiling for 5 min. Insoluble pellets were re-dissolved in buffer containing 2% SDS and then treated with ␤-mercaptoethanol before loading on a gel. Following electrophoresis, the gel was stained with Coomassie Brilliant Blue, by which the protein bands were visualized. For Western blotting, the gel was further electroblotted on a 0.2-m PVDF membrane (Bio-Rad), where the protein was detected using a mouse monoclonal anti-GST antibody (1:2,000 dilution, Wako) as a primary antibody and a stabilized goat anti-mouse IgG(HϩL) peroxidase-conjugated antibody (1:5,000 dilution, Thermo) as a secondary antibody. Blots were developed with ImmunoStar LD (Wako), and images were obtained using Limited-STAGE (AMZ System Science).
Aggregation of Recombinant FUS Proteins-To examine aggregation of FUS, 150 l of 5 M soluble GST-FUS-His proteins in a TN-trehalose buffer with 250 mM imidazole, pH 6.8, was set in a 96-well plate and left at room temperature. Turbidity was monitored at every 5 min in a plate reader (Epoch, BioTek) by measuring absorbance at 350 nm. After the aggregation reaction for 90 min, the sample was ultracentrifuged at 110,000 ϫ g for 15 min to prepare soluble supernatant (150 l) and insoluble pellets. Pellets were then re-dissolved in 150 l of an SDS-PAGE sample buffer, which contains 0.1% SDS and ␤-mercaptoethanol, and soluble supernatant was also mixed with 5ϫ SDS-PAGE sample buffer. Both soluble and insoluble fractions were then boiled and loaded on a 10% SDS-PAGE gel.
For diagnosis of amyloid-like FUS aggregates, 250 M thioflavin T was added to the reaction mixture, which was prepared by incubating 5 M GST-FUS-His in a TN-trehalose buffer with 250 mM imidazole, pH 6.8, for 2 h at room temperature. Fluorescence spectra (460 -600 nm) were then recorded by using F-4500 (HITACHI) with excitation at 442 nm.
Electron microscopic examination of in vitro FUS aggregates was also performed. GST-FUS-His aggregates collected with ultracentrifugation at 110,000 ϫ g for 15 min were first adsorbed on STEM100Cu grids coated by elastic carbon (Okenshoji), washed with pure water, and then negatively stained with 2% phosphotungstic acid. Images were obtained using an electron microscope (Tecnai TM Spirit, FEI).
For a seeding reaction, 200 l of 5 M GST-FUS G156E -His in TN-trehalose buffer with 250 mM imidazole, pH 6.8, was incubated at room temperature for 2 h, and the resultant aggregates were collected with ultracentrifugation at 110,000 ϫ g for 15 min and then resuspended in 20 l of TN-trehalose buffer, pH 6.8. After being sheared with ultrasonication, 1.5 l of the aggregates was added to 150 l of 5 M GST-FUS-His in TNtrehalose buffer with 250 mM imidazole, pH 6.8, and incubated at room temperature without any agitation. Turbidity of the reaction mixture was monitored at every 5 min in a plate reader (Epoch, BioTek) by measuring absorbance at 350 nm.
DNA Constructs, Cell Culture, and Transfection-A vector, pCMV-HA or pCMV-myc (Clontech), was used for the construction of plasmids expressing HA-tagged or myc-tagged human FUS, respectively, by utilizing its multiple cloning site (SalI/NotI). Mutations were introduced by QuikChange mutagenesis (Stratagene).
Human neuroblastoma SH-SY5Y cells were maintained in DMEM/F-12 (Invitrogen) with 10% fetal bovine serum (FBS) in 5% (v/v) CO 2 at 37°C. For immunostaining experiments, cells were seeded on a 4-well chamber slide (Lab-Tek II Chamber Slide with a cover CC2 glass slide, Nalge Nunc international). Cells were transfected with plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After transfection for 4 h, culture media were replaced with Neurobasal medium (Invitrogen) supplemented with B-27 supplement (Invitrogen), 500 M L-glutamine, and 5 mM N 6 ,2Ј-Odibutyryl cAMP (Nacalai Tesque) for differentiating SH-SY5Y cells. Following further incubation for 15 h, cells were fixed and immunostained as mentioned below.
Immunocytochemistry-For immunostaining of HA-FUS proteins, cultured SH-SY5Y cells were fixed with 4% paraformaldehyde containing 0.12 M sucrose in PBS for 10 min, permeabilized with 0.5% Triton X-100 in PBS for 5 min, and blocked with 0.1% BSA in PBS for 30 min. Cells were then incubated with anti-HA-fluorescein, High Affinity (3F10) (1:100 dilution, Roche Applied Science) in PBS containing 0.1% BSA for 1 h, washed once with 0.1% Triton X-100 in PBS and twice with 0.1% BSA in PBS. Nuclei were counterstained with DAPI (1:3000 dilution, Molecular Probes).
Transfected neurons at 9 DIV were fixed with 4% paraformaldehyde containing 4% sucrose in Mg 2ϩ and Ca 2ϩ -free Dulbecco's phosphate-buffered saline for 15 min at room temperature, and then blocked with 4% skim milk in TBS containing 0.1% Triton X-100 for 1 h. These neurons were probed with rabbit anti-HA tag polyclonal antibody (MBL), mouse anti-Myc tag monoclonal antibody (clone My3, MBL), and mouse anti-MAP2 monoclonal antibody (clone AP-20, Sigma) for 2 h at room temperature and then further incubated with the corresponding secondary antibodies conjugated with Alexa Fluor 594 or Alexa Fluor 488 (Invitrogen) for 1 h at room temperature. Nuclei were counterstained with bisbenzimide H33342 fluorochrome trihydrochloride (Hoechst 33342, Nacalai tesque). Confocal images with a slice thickness of ϳ1 m were obtained by a laser-scanning microscope of the LSM5 Exciter system (Carl Zeiss, Germany) using a ϫ40 objective lens for SH-SY5Y cells and by Axiovert 200M microscope (ZEISS) with a ϫ40 and ϫ63 objective lens for primary neurons.

Purification of Recombinant Full-length FUS Proteins in a Dual
Tag System-In a previous studies (e.g. Ref. 15), recombinant FUS was overexpressed in E. coli as a GST-fused protein, and we thus first attempted to purify GST-fused FUS proteins (GST-FUS) by affinity chromatography using glutathione (GSH)-Sepharose resins. Although most of GST-FUS proteins were detected in insoluble pellets after lysis of E. coli cells, a soluble fraction of the lysates was treated with GSH-Sepharose resins, by which a soluble form of GST-FUS was obtained ( Fig.  2A). The purified GST-FUS samples, however, exhibited the absorption peak centered at 260 nm (Fig. 2B), suggesting the contamination of nucleic acids. Indeed, washing the resins with a buffer containing 1 M NaCl can remove the species with the absorption at 260 nm (Fig. 2, B and C), which were digested with RNase but not DNase (Fig. 2D). Accordingly, GST-FUS tightly binds endogenous RNAs in E. coli, consistent with a physiological role of FUS as an RNA-binding protein.
Even after removal of bound RNAs, it should also be noted that the purified GST-FUS samples are not homogenous but contain impurities with smaller molecular weight ( Fig. 2A). Assuming that these smaller proteinaceous species were bound to GSH-Sepharose, the purified GST-FUS samples were considered to contain significant amounts of the proteins truncated in the middle of FUS. To prepare the recombinant protein containing a full-length sequence of FUS, therefore, we have introduced a polyhistidine (His 6 ) tag at the C terminus of FUS in addition to the N-terminal GST tag (GST-FUS-His). Similar to the case in GST-FUS, GST-FUS-His was mostly detected in insoluble pellets after cell lysis (Fig. 3A), and incubation of soluble supernatant with GSH-Sepharose resins isolated truncated as well as full-length GST-FUS-His proteins (Fig. 3A). By utilizing the C-terminal His 6 tag in GST-FUS-His, we have further performed Ni 2ϩ -affinity chromatography and removed the truncated proteins that lack a His 6 tag. A single major band corresponding to the molecular mass of full-length GST-FUS-His (about 81 kDa) was confirmed with SDS-PAGE (Fig. 3A) and Western blotting analysis (Fig. 3B). Furthermore, purification of GST-FUS-His using Ni 2ϩ -affinity chromatography was found to remove contaminants of nucleic acids (Fig. 3C). These results have, therefore, indicated that a full-length FUS protein without bound nucleic acids was successfully prepared with the two-step purification procedure utilizing N-terminal GST and C-terminal His 6 tags.
A Pathogenic G156E Mutation Increases the Aggregation Propensities of FUS-After being eluted from Ni 2ϩ -affinity chromatography resins with an imidazole-containing buffer at pH 8.0, purified GST-FUS-His proteins were concentrated by a centrifugal concentrator to prepare 5 M solution. The sample solution was, however, found to gradually become turbid when left at 25°C (Fig. 3D), and almost all GST-FUS-His proteins after incubation for 1 h were detected as insoluble pellets after ultracentrifugation (Fig. 3D, inset). Just after being eluted from the resins, GST-FUS-His proteins started to aggregate, but several procedures including determination of the protein concentration were required for the aggregation assay. It is, therefore, difficult to characterize the aggregation reaction with reproducible kinetic parameters. This is also the case with GST-FUS-His proteins with ALS-causing mutations. In contrast, when the pH of an elution buffer was decreased from 8.0 to 6.8, we serendipitously found no change in the solution turbidity of purified GST-FUS-His even after incubation for 1 h, and GST-FUS-His proteins remained in soluble supernatant after ultracentrifugation (Fig. 3D). Although a detailed mechanism of FUS aggregation still remains obscure, a solution pH appears to be a critical factor affecting the aggregation propensities of our FUS proteins.
We have then tested effects of pathogenic mutations (G156E (16), G225V (17), and M254V (18)) on the aggregation propensity of GST-FUS-His, pH 6.8. As shown in Fig. 4A, no change in the solution turbidity of GST-FUS-His was observed with G225V and M254V mutations (FUS G225V , FUS M254V ), but the G156E mutation in GST-FUS-His (FUS G156E ) increased solution turbidity, suggesting the formation of aggregates. GST-FUS-His with a pathogenic mutation at the C-terminal region, P525L (5), did not form aggregates and remained soluble at pH 6.8 within the incubation time examined here (Fig. 4, A and B), consistent with limited roles of C-terminal mutations in the aggregation propensities of FUS (15). It is notable that the solution turbidity of FUS G156E increased just after the sample was set in a 96-well plate; therefore, we have speculated that aggregation of FUS G156E already starts when eluted from the resins. In our experimental protocol, it usually required approximately 40 min for starting the turbidity measurement after elution of FUS proteins from the Ni 2ϩ -affinity chromatography resins. Indeed, the sample solution of FUS G156E was sometimes already turbid when the sample was set for turbidity measurements; therefore, the initial time point of the FUS G156E aggregation remains obscure. Despite this, FUS G156E was found as insoluble pellets after ultracentrifugation, whereas FUS G225V and FUS M254V remained in soluble supernatant (Fig. 4B). These data have thus clearly shown a distinct role of G156E mutation in increasing the aggregation propensities of a FUS protein at least in the physiological range of solution pH. Notably, furthermore, G144E, G154E, and G156D, which have not been reported as pathogenic mutations, also increased the aggregation propensities of FUS albeit to a lesser extent compared with G156E (Fig. 4B). It is thus likely that the introduction of a neg- FUS Aggregates Exhibit Amyloid-like Properties-Abnormal accumulation of FUS-positive cytoplasmic inclusions have been characterized as a major pathological change in FUS-related fALS cases (5, 6), albeit no report in fALS cases with G156E-mutant FUS. Ultrastructural analysis on those pathological inclusions has further revealed the presence of fibrillar FUS aggregates (19,20). Consistent with such molecular pathologies, insoluble FUS G156E aggregates in this study (Fig.  4B) were found to possess fibrillar morphologies with approximately 10 nm of the width (Fig. 4C). These structures were not observed in wild-type GST-FUS-His proteins (FUS WT ), which remained soluble at pH 6.8 (Fig. 4B). We have also found that the aggregates of FUS G156E increase the fluorescence intensity of thioflavin T, a diagnostic dye of amyloid-like protein aggregates (Fig. 4D). Again, FUS WT did not increase the thioflavin T fluorescence after incubation for 1 h. Pathological inclusions containing FUS in patients with frontotemporal lobar degeneration were not stained with thioflavin dyes (21), but there have been no detailed studies examining reactivity of thioflavin dyes to inclusions in patients with fALS-causing mutations in FUS.
Taken together, the G156E mutation is considered to trigger the formation of fibrillar amyloid-like aggregates of FUS proteins in vitro.

G156E Mutation Facilitates Formation of Intranuclear FUSpositive Foci in Cells-We next tested if the G156E mutation leads to aggregation of FUS in the intracellular environment.
Although FUS appears to shuttle between the nucleus and cytoplasm (4), immunocytochemistry has revealed intracellular localization of wild-type FUS predominantly to the nucleus (10,22,23). In differentiated human neuroblastoma SH-SY5Y cells, we have confirmed the nuclear localization of endogenous FUS proteins (data not shown) as well as the transiently transfected human wild-type FUS with an N-terminal HA tag (HA-FUS WT ); furthermore, intranuclear inclusions/foci were not observed (Fig. 5A). As reported previously, cytoplasmic staining (including diffuse and foci) of FUS was evident in almost all cells overexpressing HA-FUS with R522G and P525L mutations (Fig. 5, D and E), which is consistent with roles of the C-terminal region as a nuclear localization signal (10,22,23). In contrast, HA-FUS with the G156E mutation (HA-FUS G156E ) remained nuclear but notably formed intranuclear foci in transfected cells (approximately 40% of total cells expressing HA-FUS G156E , white arrows in Fig. 5B).
The G225V mutation appears to have minimal effects on both intracellular localization and foci formation of FUS (Fig. 5C). Given increased propensities of FUS aggregation in vitro with G156E but not the G225V mutation, these observations using cultured cells support our idea that the G156E mutation facilitates aggregation of FUS even in the intranuclear environment.  We have also examined effects of pathogenic mutations on FUS proteins in primary neurons. Rat hippocampal primary neurons (DIV 7) were transfected with plasmids for overexpression of HA-FUS proteins and immunostained with anti-HA antibodies after incubation for 2 days (DIV 9). As shown in Fig. 6A, HA-FUS WT was localized in the nucleus with no cytoplasmic staining, whereas HA-FUS P525L formed cytoplasmic foci in primary neurons (Fig. 6E), again confirming critical roles of the C-terminal region as a nuclear localization signal (10,22,23). As observed in SH-SY5Y cells, HA-FUS G156E was localized in the nucleus, but notably, abnormal intranuclear structures such as ring-like inclusions (Fig. 6B) and small dot-like foci (Fig. 6C) were evident in primary neurons at DIV 9 expressing HA-FUS G156E but not HA-FUS WT (Fig. 6A). In a subset of transfected neurons, furthermore, HA-FUS G156E was found to form cytoplasmic foci (Fig. 6D), even though Gly 156 appears to play limited roles in the intracellular localization of FUS in SH-SY5Y (Fig. 5) and HeLa cells (10). This could be described by increased aggregation propensities of FUS with the G156E mutation, which result in the formation of foci before being transported into the nucleus.
As shown in Fig. 6, A-E, primary neurons were immunostained with anti-MAP2 (microtubule-associated protein 2) antibodies for visualization, because microtubules help determine the shape and size of neurons (24). Notably, primary neurons expressing HA-FUS G156E and HA-FUS P525L appear to exhibit reduced immunoreactivity toward MAP2 compared with the neurons expressing HA-FUS WT as well as non-transfected neurons (Fig. 6, A-E). Indeed, quantification has revealed a significant decrease in the intensity of immuno-staining for MAP2 by transfecting cells to overexpress HA-FUS G156E and HA-FUS P525L but not HA-FUS WT (Fig. 6F). Loss of MAP2 has been considered as an early marker for neuronal damage following cerebral ischemia (25), spinal cord injury (26), and traumatic brain injury (27); therefore, potential toxicities of FUS G156E and FUS P525L to neurons would be reflected by reduced intensity of MAP2 immunostaining in primary neurons. In addition, primary neurons expressing HA-FUS G156E and HA-FUS P525L often exhibited abnormal cell shapes (e.g. Fig. 6D) that were characterized by less numbers of dendrites than those of non-transfected cells as well as cells expressing HA-FUS WT , albeit difficult to be quantified. These observations are considered to support toxic roles of FUS G156E and FUS P525L in neurons.
Seeded Aggregation of Wild-type FUS with G156E Mutant FUS in Vitro and in Vivo-It remains obscure how the G156E mutation in FUS exerts toxicity toward neurons, but physiological functions of FUS such as translational and splicing regulations are well expected to be retarded upon formation of aggregates in the nucleus. Notably, neurodegenerative diseases including FUS-related ALS are characterized by their rapid progression after the disease onset, which appears to be regulated by a "seeding reaction" at a molecular level (28). A seeding reaction is an important feature of amyloid-like fibrillar aggregates, in which a piece of protein fibrils can function as a structural template (called as a seed) for facilitating the fibrillation of as-yet-unaggregated protein molecules (29). Once mutant FUS proteins form aggregates in the nucleus, therefore, a seeding reaction could effectively transform the soluble functional FUS into insoluble aggregates, resulting in the progressive dysfunc- tion of FUS with cytotoxicity. Indeed, a seeded fibrillation of proteins has been increasingly noticed as a molecular pathomechanism that describes progression of several neurodegenerative diseases (28).
To test the seeding ability of our FUS G156E aggregates in vitro, we have added well sheared FUS G156E aggregates with ultrasonication to the solution containing aggregation-resistant GST-FUS WT -His at pH 6.8. As shown in Fig. 7A, no increase in the turbidity of FUS WT -containing solution was confirmed in the absence of any seeds, but addition of small amounts (10%) of sheared FUS G156E aggregates immediately made the solution turbid, suggesting the seeded aggregation of FUS WT with FUS G156E aggregates. Ultracentrifugal fractionation has further confirmed that almost all FUS WT molecules are detected in insoluble pellets by addition of FUS G156E aggregates (Fig. 7B). Based upon these results, the G156E mutation will increase the propensities of FUS proteins for fibrillar aggregation, which would further facilitate the aggregation of wildtype FUS via a seeding mechanism.
We further examined our seeding mechanism by co-transfection of rat hippocampal primary neurons with plasmids expressing Myc-tagged FUS WT (myc-FUS WT ) and HA-FUS. As shown in Fig. 7C, co-expression of myc-FUS WT and HA-FUS WT did not result in the formation of any inclusions, and both wild-type FUS proteins remained diffused in the nucleus. In contrast, when myc-FUS WT was expressed together with HA-FUS G156E in neurons, HA-FUS G156E formed intranuclear structures of ring-like inclusions (Fig.   7D) and dot-like foci (Fig. 7E), and importantly, myc-FUS WT was also detected in those abnormal structures (Fig. 7, D and  E). Based upon these results, intranuclear aggregates of FUS G156E can recruit wild-type FUS proteins into their own structures, consistent with a seeding ability of FUS G156E aggregates in vitro. A seeded aggregation would, therefore, accelerate the functional impairment of intranuclear FUS proteins.

DISCUSSION
As described in the original paper reporting the G156E mutation in FUS (16), Gly 156 located in the SYGQ domain is evolutionarily conserved among vertebrates, implying an important role in physiological functions of FUS. A low complexity region such as the SYGQ domain of FUS has been proposed as an interaction site with partner proteins (30,31); therefore, the G156E mutation might perturb protein-protein interactions, eventually exerting cytotoxic effects on motorneurons.
Alternatively, as we have shown here, the G156E mutation, but not the other mutations tested (G225V, M254V, and P525L), increases the aggregation propensity of FUS (Fig. 8).
The SYGQ domain of FUS has been predicted to exhibit a prion-forming propensity by using several computer simulations (12). Particularly, Gly 156 locates at the middle of a sequence composed exclusively of Gln, Asn, and Tyr residues (Fig. 8). Several precedents including yeast prion protein, Sup35, have shown high fibrillation propensities of polypeptides rich in Gln, Asn, and Tyr. Although it is still unclear why a Gly to Glu mutation at position 156 specifically increases the aggregation propensities of FUS, an N-terminal low-complexity (LC) region of FUS (amino acids 2-214), which covers the SYGQ domain and almost an N-terminal half of RGG1, has been recently reported to form amyloid-like fibrils in vitro with significant reversibility (32,33). Those in vitro reversible fibrils formed by an LC region of FUS have been further shown to act as seeds, from which LC regions of FUS and the other RNAbinding proteins such as heterogeneous nuclear ribonucleoprotein A1 are polymerized (33). It should also be noted that the SYGQ domain alone is resistant to aggregation in yeast cells, and certain interactions between SYGQ and RGG2 would play roles in the formation of aggregates (15). Taken together, these results support important roles of the SYGQ domain in the aggregation of FUS in vitro and in vivo. Given that the replacement of Gly to Glu adds a negative charge on FUS, the G156E mutation might modulate interactions between SYGQ and RGG2 and thereby facilitate the aggregation. An increase of the solution pH will also add negative charges on FUS, which are consistent with our observation that GST-FUS-His becomes prone to aggregation at increased pH (Fig. 3D). Charge distribution of FUS might thus be important in the description of its aggregation propensities. Once the aggregation of mutant FUS starts in the nucleus, furthermore, the aggregates will function as seeds to trigger aggregation of the wild-type protein. FUS deficiency in a mouse has been previously shown to lead to perinatal mortality (34), and loss of physiological functions of FUS appears to be detrimental. Such recruitment of wild-type as well as mutant FUS into the aggregate structures is, therefore, considered to disturb physiological processes regulated by FUS, resulting in the deterioration of cellular activities, in particular, RNA metabolisms including transcription, alternative splicing, and mRNA transport (35). Recently, it has been reported that alternative splicing of certain neurological disease-associated genes such as Mapt (encoding Tau protein) is regulated by FUS (36 -38). Given that abnormalities in the splicing events of the Tau gene are linked to frontotemporal lobar degeneration (39), intracellular amounts of alternatively spliced Tau could also be altered by aggregation of mutant FUS, which might then contribute to the appearance of neurodegenerative phenotypes.
It is clinically notable that a patient with the G156E mutation in FUS developed dementia concurrently with the presence of both upper and lower motor neuron signs (16). Although many patients with ALS have been reported to show cognitive impairment (40), cognitive dysfunction is considered to be absent or rare in ALS patients with other types of mutations in FUS (6,16,(41)(42)(43)(44)(45)(46)(47). There is a caveat that no autopsy has been reported in G156E cases and also that no animal models of G156E-FUS are available as of now. More numbers of patients with FUS mutations, especially the G156E mutation, should hence be described before conclusions, but it is tempting to speculate that increased propensities of G156E-mutant FUS for aggregation in the nucleus are responsible for its distinct clinical features.
In summary, we have successfully purified a protein containing full-length FUS that is free from truncated proteins and E. coli endogenous nucleotides. Among several pathogenic mutations, G156E did not affect functional PY-NLS at the C terminus of FUS but increased propensities of FUS for aggregation in vitro and in vivo. Furthermore, a seeding activity of FUS G156E aggregates was found to trigger the transformation of soluble, functional FUS WT into aggregated structures, which would contribute to a significant loss of physiological functions of FUS in cells (Fig. 8). Even with intact PY-NLS, aggregation of intranuclear FUS proteins caused by amino acid mutations (e.g. G156E) and/or environmental changes is an alternative pathomechanism of FUS-related ALS diseases. The amino acid sequence of the FUS SYGQ domain is also shown, and the concentration of Gln, Asn, and Tyr residues is evident near Gly 156 (underlined).