Sequestration of PRMT1 and Nd1-L mRNA into ALS-linked FUS mutant R521C-positive aggregates contributes to neurite degeneration upon oxidative stress

Mutations in fused in sarcoma (FUS), a DNA/RNA binding protein, are associated with familial amyotrophic lateral sclerosis (ALS). However, little is known about how ALS-causing mutations alter protein-protein and protein-RNA complexes and contribute to neurodegeneration. In this study, we identified protein arginine methyltransferase 1 (PRMT1) as a protein that more avidly associates with ALS-linked FUS-R521C than with FUS-WT (wild type) or FUS-P525L using co-immunoprecipitation and LC-MS analysis. Abnormal association between FUS-R521C and PRMT1 requires RNA, but not methyltransferase activity. PRMT1 was sequestered into cytosolic FUS-R521C-positive stress granule aggregates. Overexpression of PRMT1 rescued neurite degeneration caused by FUS-R521C upon oxidative stress, while loss of PRMT1 further accumulated FUS-positive aggregates and enhanced neurite degeneration. Furthermore, the mRNA of Nd1-L, an actin-stabilizing protein, was sequestered into the FUS-R521C/PRMT1 complex. Nd1-L overexpression rescued neurite shortening caused by FUS-R521C upon oxidative stress, while loss of Nd1-L further exacerbated neurite shortening. Altogether, these data suggest that the abnormal stable complex of FUS-R521C/PRMT1/Nd1-L mRNA could contribute to neurodegeneration upon oxidative stress. Overall, our study provides a novel pathogenic mechanism of the FUS mutation associated with abnormal protein-RNA complexes upon oxidative stress in ALS and provides insight into possible therapeutic targets for this pathology.


Results
FUS-R521C associates more avidly with PRMT1 than with FUS-WT or FUS-P525L. To identify the proteins abnormally associated with ALS-causing FUS mutants, FLAG-FUS-R521C, FLAG-FUS-P525L, or FLAG-FUS-WT was expressed in HEK293T cells 20 . Co-IP of cell lysates using FLAG antibodies was performed 48 h after transfection, and the proteins analyzed on silver-stained polyacrylamide gels from extracts of cells overexpressing each FLAG-tagged FUS protein. Silver-stained SDS-PAGE analysis of elutes showed a 43-kDa protein that was prominent in FUS-R521C samples, but weak in FUS-WT and FUS-P525L samples (Fig. 1A). LC-MS analysis revealed that the most abundant peptides in tryptic digests of the corresponding band were derived from the human protein arginine N-methyltransferase 1 (PRMT1), the predominant type-I PRMT in mammalian cells which is involved in gene transcription, DNA repair, signal transduction, and protein translocation 21 (Supplementary Fig. 1). Our LC-MS findings further supports previous reports that FUS binds to PRMT1 in mammalian cell lines and neuronal cells in vitro and in vivo [16][17][18]22,23 .
To confirm the association between FUS and PRMT1 revealed by the LC-MS data, we conducted co-IP and western blot using anti-FLAG and anti-PRMT1 antibodies. Endogenous PRMT1 was pulled down with FLAG-FUS-WT, FLAG-FUS-R521C, or FLAG-FUS-P525L. Consistent with the silver staining results, the R521C point mutation in FUS specifically increased its interaction with PRMT1 relative to the WT or P525L mutation (Fig. 1B,C), raising the possibility that the association of FUS and PRMT1 could be altered by the specific ALS-linked mutation, R521C. We further confirmed that the R521C point mutation significantly enhanced the association between FUS and PRMT1 compared to WT or P525L mutant using cultured neuronal cell lysates infected with adeno-associated virus (AAV) expressing FLAG-FUS-WT, FLAG-FUS-R521C, or FLAG-FUS-P525L (Fig. 1D,E). Moreover, we found that the specific substitution of arginine for cysteine at position 521 in FUS (R521C) significantly enhanced its association with PRMT1, while this effect was not observed in other ALS-linked FUS mutants (R521H or R521G) (Fig. 1F,G).
Scientific RepoRts | 7:40474 | DOI: 10.1038/srep40474 Since the R521C point mutation in FUS specifically increased its interaction with PRMT1, we next examined the cellular localization of FUS-WT or FUS-R521C with PRMT1 in cultured cortical neurons expressing GFP, GFP-fused FUS-WT, or GFP-fused FUS-R521C. As expected, GFP-FUS-WT, GFP-FUS-R521C, or endogenous PRMT1 was primarily localized to the nucleus in MAP2-positive cortical neurons (Fig. 1H). However, ALS-linked FUS-R521C and endogenous PRMT1 were co-localized to the cytosol and formed cytosolic aggregates which are TIA-1-positive stress granules without any stress inducer (Fig. 1I). This finding clearly shows that the R521C point mutation in FUS causes cytosolic sequestration of PRMT1 into stress granules in cultured cortical neurons.

PRMT1 was sequestered into methylated FUS-R521C-positive stress granule compared to FUS-WT.
Various studies have reported that PRMT1 methylates arginine residues of FUS 16,18,19 . To test whether enhanced association of FUS-R521C with PRMT1 affects this process, we examined the methylation level of FUS-WT and FUS-R521C in HEK293T cells expressing FLAG-FUS-WT, or FLAG-FUS-R521C by using co-IP with anti-FLAG antibodies and western blot analysis with anti-PRMT1, anti-FLAG, and anti-ASYM24 (which recognize di-methylated proteins) antibodies. As shown in Fig. 2A, no substantial difference was observed in the global methylation of proteins associated with FUS-R521C. Furthermore, the level of methylation of FUS-R521C was not significantly enhanced compared to FUS-WT (Fig. 2B,C).
We next examined the cellular localization of methylated FUS (me-FUS) and PRMT1 in neurons expressing GFP, GFP-FUS-WT, or GFP-FUS-R521C by immunocytochemistry using anti-methylated FUS antibody, anti-PRMT1, or anti-PABP. In neurons expressing GFP or GFP-FUS-WT, me-FUS was mostly diffused in the nucleus and cytosol. However, in neurons expressing GFP-FUS-R521C, me-FUS was colocalized to FUS-R521C/ PRMT1-positive stress granules without any stress induction, indicating that PRMT1 is partially sequestered into these me-FUS-R521C-positive stress granules (Fig. 2D,E). To further characterize the abnormal interaction between FUS and PRMT1, we determined whether it requested PRMT1 activity of methyltransferase. We performed co-IP in the presence or absence of Adenosine-2′, 3′ -dialdehyde (Adox), an inhibitor of arginine methylation 24 . Incubation with Adox for 30 mins prior to transfection did not affect the abnormal association between FUS-R521C and PRMT1, indicating that such association is independent of arginine methylation (Fig. 2F). To further confirm this result, we performed co-IP using HEK293T cell lysates expressing FLAG-FUS-R521C and either N-myc vector, N-myc-PRMT1-WT, or N-myc-PRMT1-G98R (the catalytically inactive mutant PRMT1 25 ). As shown in Fig. 2G,H, there is no significant difference in the association of FUS-R521C and PRMT1 between PRMT1-WT and PRMT1-G98R, indicating that PRMT1 methyltransferase activity is not essential for the enhanced association of FUS-R521C and PRMT1.
RNA and RGG2-ZnF-RGG3 (RGG-2-3) domains are required for stable association of FUS complex with PRMT1. It has been reported that several RNA binding proteins, including FUS, require RNAs for formation of RNA granules 26 . The multi-step process of FUS aggregation has been shown to require RNA-dependent and RNA-independent processes 27 . Furthermore, it has also been reported that RNA seeds the higher-order assembly of FUS 28 . Therefore, we examined whether RNAs are required for stable FUS/PRMT1 complex formation. Cell lysates were incubated with RNase A for 2 h before co-IP with anti-FLAG antibodies was carried out. As shown in Fig. 3B,C, cell lysates treated with RNase A showed a complete loss of association of FUS-WT with PRMT1. Moreover, abnormal association of FUS-R521C with endogenous PRMT1 was significantly reduced, suggesting that RNAs are required for the association of FUS with PRMT1 ( Fig. 3B,C).
To further confirm whether RNA binding and arginine methylation of FUS affect its association with PRMT1, we examined their association using co-IP with a deletion mutant of FUS RGG2-ZnF-RGG3 domains (FUS-R521C-Δ RGG2-3; Fig. 3A), which are involved in RNA recruitment and arginine methylation of RNA binding proteins [29][30][31] . As shown in Fig. 3D,E, association of PRMT1 with FUS-R521C-Δ RGG2-3, as compared to a full length FUS-R521C, was mostly reduced, suggesting that RGG2-ZnF-RGG3 domains of FUS-R521C are involved in the stable association of the FUS-R521C/PRMT1 complex. Therefore, these data suggest that RNAs and RGG2-ZnF-RGG3 domains are essential for the stable formation of FUS-R521C/PRMT1 complexes. Oxidative stress increased cytosolic FUS-R521C/PRMT1-positive SG aggregates which are regulated by PRMT1. Previous studies including our own have shown cellular localization of ALS-linked FUS mutants into stress granules upon stress conditions 32,33 . Therefore, we examined cellular localization of PRMT1 and either FUS-WT or FUS-R521C upon oxidative stress and after the removal of oxidative stress. FLAG-FUS-WT, or FLAG-FUS-R521C was expressed in cultured cortical neurons and 48 h after transfection neurons were exposed to sodium arsenite (SA, 0.5 mM) for 1 h. As we expected, in the absence of oxidative stress PRMT1 was mostly colocalized to nuclear FUS-WT and partially localized to cytosolic FUS-R521C-positive aggregates (Fig. 4A,B). However, SA-induced oxidative stress significantly increased the Next, to examine their disassembly, cellular localization of either FUS-WT or FUS-R521C with PRMT1 was examined 2 h after the removal of SA. Interestingly, as shown in Fig. 4B,C, cytosolic FUS-R521C/PRMT1-positive SG aggregates were still present 2 h after removal of SA while cytosolic FUS-WT and PRMT1 disappeared, suggesting that their stable FUS-R521C/PRMT1 complex may contribute to stress-induced neurodegeneration during aging.
To test whether Nd1-L mRNA is sequestered into FUS-R521C complexes, we performed mRNA-protein pull-down assay with biotinylated Nd1-L mRNA and streptavidin-agarose beads using HEK293T cell lysates expressing FLAG-FUS (WT, R521C, or P525L). As shown in Fig. 6A,B, the 3′ -UTR of Nd1-L mRNA was more avidly associated with FUS-R521C than with FUS-WT or FUS-P525L. To examine the cellular localization of FUS/PRMT1/Nd1L mRNA, we performed fluorescent in situ hybridization (FISH) in neurons expressing GFP-FUS-WT, GFP-FUS-R521C, or GFP-FUS-R521C-Δ RGG2-3, using fluorescence-labelled cDNA probes against Nd1-L mRNA and anti-PRMT1 antibody. Our FISH data indicated that endogenous Nd1-L mRNA and PRMT1 are colocalized to FUS-WT and FUS-R521C in the nucleus but not to FUS-R521C-Δ RGG2-3 which fails to associates with PRMT1 (Fig. 6C). Furthermore, our data showed colocalization of Nd1-L mRNA into FUS aggregates in the cell body and in the proximal region of neurons expressing FUS-R521C compared to neurons expressing FUS-WT, or FUS-R521C-Δ RGG2-3 (Fig. 6C), thus suggesting that Nd1-L mRNA is sequestered into FUS-positive aggregates and nucleus in neurons expressing FUS-R521C but not FUS-R521C-Δ RGG2-3. Furthermore, we found that overexpressed FUS-WT and FUS-R521C enhanced the fluorescence intensity of FISH probe against Nd1-L mRNA in the nucleus and FUS-positive aggregates (Fig. 6C), raising the possibility that the sequestration of Nd1-L by FUS-PRMT1 complex affects the stability of Nd1-L mRNA as well as its cellular localization.
To investigate whether the enhanced association of FUS-R521C with the 3′ -UTR mRNA of Nd1-L indeed affects the stability of Nd1-L mRNA, we expressed FUS-WT or FUS-R521C in HEK293T cells and treated them with actinomycin D, which inhibits new transcription. To check the level of Nd1-L mRNA, we performed RT-PCR using cell lysates expressing FUS-WT or FUS-R521C and quantified the relative gene expression of Nd1-L mRNA immediately and at 4 h after actinomycin D treatment, 48 h after transfection. As shown in Fig. 6D, at 4 h after treatment, Nd1-L mRNA remained in cells expressing FUS-R521C, while its level was reduced in FUS-WT-expressing cells, suggesting that Nd1-L mRNAs might be trapped in stable FUS-RNA complexes. Therefore, these data suggest that Nd1-L mRNA is trapped and remained in FUS-R521C-complex in FUS-R521C-expressing cells compared to FUS-WT-expressing neurons.

Expression of Nd1-L ameliorates neurite shortening caused by FUS-R521C upon oxidative stress, while loss of Nd1-L further exacerbates neurite shortening.
Based on our co-IP experiments, Nd1-L mRNA was sequestered into FUS aggregates and remained trapped in FUS-R521C/PRMT1 complexes (Fig. 6). To elucidate whether this sequestration contributes to neurite shortening upon oxidative stress, we examined neurite morphology in FUS-R521C-expressing neurons compared to GFP, FUS-WT, or FUS-P525L expressing neurons while modulating Nd1-L.
We examined the effects of loss of Nd1-L on neurite shortening upon oxidative stress in neurons expressing FUS-R521C compared to the control (CTL: pSUPER-GFP-scramble) and to neurons expressing FUS-WT, or FUS-P525L with or without oxidative stress. We generated two different pSUPER-GFP-Nd1-L plasmids (#1-2) expressing shRNAs against mouse Nd1-L and checked its knockdown efficiency (Supplementary Fig. 3). We then transfected pSUPER-GFP-scramble, FLAG-FUS-WT, FLAG-FUS-R521C, or FLAG-FUS-P525L together with either pSUPER-GFP-Nd1-L (#2) or pSUPER-GFP-scramble into cultured cortical neurons. When Nd1-L expression was knocked down by transfection of pSUPER-GFP-Nd1-L (#2), neurite shortening caused by FUS-R521C was further aggravated, which was not the case in the control neurons, FUS-WT or FUS-P525L expressing neurons (Fig. 7). However, expression of Nd1-L-GFP partially rescued neurite shortening caused by FUS-R521C Furthermore, overexpression of PRMT1 did not rescue neurite shortening aggravated further by the loss of Nd1-L in FUS-R521C expressing neurons upon oxidative stress, while expression of Nd1-L partially but significantly rescued neurite length further reduced by loss of PRMT1 in FUS-R521C expressing neurons (Figs 5 and 7). These results suggest that the trap of Nd1-L mRNA into FUS-R521C-positive aggregates by the sequestration of PRMT1 can affect neurite shortening upon oxidative stress, leading to cellular pathogenesis associated with the FUS-R521C mutation.

Discussion
Despite several recent studies on the pathogenic mechanisms of ALS-associated FUS mutants, little is known about how these different mutants affect neurite degeneration and phenotypic severities. Moreover, it is still largely unknown how these specific ALS-linked FUS mutations alter protein-protein and protein-RNA interaction, which could contribute to neurodegeneration upon stress.
In our study, we identified PRMT1 and the mRNA of Nd1-L as the components more avidly associated with ALS-linked FUS-R521C mutants than with FUS-WT or FUS-P525L. Although PRMT1 has been reported as a binding partner of FUS 16,18,23,26,37 , our study is the first to demonstrate that PRMT1 is associated with FUS-R521C more than with other FUS mutants. Based on our and other cellular studies, the FUS-R521C mutant is preferentially localized to SGs, leading to the formation of FUS-positive SG aggregates [38][39][40] . The rare FUS-P525L mutation leads to a more aggressive and rapidly progressive form of ALS in young patients compared to the FUS-R521C mutation, one of the most frequent mutations 30,41 . At the cellular level, FUS-P525L, due to the mutation in the proline-tyrosine nuclear localization signal (PY-NLS) domain, is mostly localized to the cytosol and forms cytosolic aggregates associated with SGs, while FUS-R521C is localized to both the nucleus and cytosolic SG aggregates 14,33,42 . This indicates that there is a specific cellular pathogenic mechanism associated with each mutation. Using LC-MS, we found that PRMT1 has higher affinity for the R521C mutant than WT or the P525L mutant, raising the possibility that its altered association with PRMT1 may be one of the major mechanisms for severe neurodegeneration. PRMT1, which is responsible for almost 90% of cellular methylation by PRMTs in mammalian cells, is known to regulate RNA processing, transcriptional regulation, signal transduction, cellular localization, and DNA repair 21 . PRMT1 is localized in the nucleus and cytosol, and is highly mobile between these compartments 24 . Arginine methylation of FUS by PRMT1 has been known to affect its cellular localization, although the effect of this methylation on the localization of FUS depends on the cell type or the stress conditions 16,18,19,37 . In the present study, the stable complex of FUS-R521C with PRMT1 caused persistent sequestration of FUS/PRMT1 complex and accumulation of FUS-positive SG-aggregates, which lead to abnormal neurite morphology and might contribute to neurodegeneration, indicating that arginine methylation by PRMT1 regulates SG formation/assembly. A growing body of evidence also suggests that the impairment of SGs is linked to cellular pathogenesis of ALS associated with genetic mutations in C9ORF72, SOD1, TDP-43, Profilin-1 [43][44][45] .
However, PRMT1 might also affect the activity of FUS as a transcriptional regulator via arginine methylation 22 . Accumulation of human mutant FUS-R495X in the cytoplasm caused nuclear depletion of PRMT1 in vivo in FUS-R495X transgenic mice, reducing methylation of its nuclear substrates 23 . Indeed, we also found that PRMT1 was more associated with FUS-R521C compared to FUS-WT in the nucleus and the cytosol (Supplementary Fig. 4). Therefore, it is plausible that sequestration of PRMT1 by ALS-linked FUS-R521C may cause PRMT1 loss-of-function mutations and impair its nuclear functions.
Furthermore, recent studies showed that the interaction of protein arginine methyltransferase 6 (PRMT6) with androgen receptor (AR) with poly-glutamine (Q) is significantly enhanced in an AR mutant associated with spinobulbar muscular atrophy (SBMA) leading to neurodegeneration 46 . Moreover, poly-Q expanded mutant huntingtin showed altered interaction with PRMT5, which resulted in reduction of PRMT5-mediated demethylation of histones 47 . Interestingly, FUS also acts as a modifier of a poly-Q expanded disease mouse model of Huntington's Disease (HD) 48 . Therefore, it will be very interesting to investigate a molecular link between arginine methylation, FUS, cytosolic aggregation and stress granule assembly in motor neuron diseases.
The FUS-R521C mutation has indeed been reported to alter several protein-protein interactions. Neuronal aggregates formed by mutant FUS-R521C aberrantly sequestered survival motor neuron (SMN) protein through its enhanced association with SMN and concomitantly reduced SMN levels in axons, leading to axonal defects 7 . Moreover, mutant FUS proteins form stable complexes with FUS-WT and interfere with the normal interactions between FUS and histone deacetylase 1 (HDAC1) leading to DNA damage as well as profound dendritic and synaptic phenotypes 49 . Recently, it has been also reported that enhanced association between the FUS-R521C mutant and SMN reduced gems (gemini bodies) changed the steady state level of snRNA in transgenic mouse tissue and in human fibroblasts 50 .
Based on our co-IP study, specific substitution of arginine with cysteine (R -> C), but not histidine (R -> H) or glycine (R -> G), in FUS at the amino acid position 521 appears to be critical for the stable association of FUS-R521C with PRMT1. Cysteine residues are important for redox signaling 51 . The solubility of TDP-43 is regulated by direct stress-induced cysteine oxidation and disulfide bond formation 51 . Cysteine-generating ALS-linked TDP-43 mutants, such as G348C or S379C, produce abnormal disulfide cross-linking upon oxidative stress. In our study, the cysteine altered from arginine within FUS-R521C might have been affected by oxidation upon oxidative stress, which affects accumulation of SG aggregates of FUS-R521C. Therefore, it would be interesting to investigate the exact roles of disulfide species on the aggregation of FUS-R521C-positive stress granule.
A growing body of evidence has demonstrated the biochemical property of FUS as an RNA-binding protein that regulates RNA splicing 30 . Interestingly, RNA targets of FUS have long intron regions in their genes, indicating that FUS regulates alternative splicing or transcription. Although FUS RNA-binding sequences, such as GGUC/GUGGU, have been identified, recently accumulating evidence suggests that the FUS RNA-binding properties are not limited to specific target sequences, but rather depend on the secondary or tertiary structure of RNAs [52][53][54] . Indeed, it has been reported in an experiment using RNAs from the intron-exon boundary and 3′ UTR of brain-derived neurotrophic factor (Bdnf) mRNA that the association between the low-complexity domain of FUS and RNAs is important for higher-order protein-RNA complexes 49 .
According to our co-IP data, stable FUS/PRMT1 complexes seem to be dependent on RNAs and RGG domains. FUS has been known to associate with many heterogeneous ribonuclear proteins, and their stable associations are also dependent on RNA. Consistent with these findings, RNA seeds the higher-order assembly of FUS 28 . ALS-linked FUS has been shown to form more stable FUS-RNA complexes in our and other studies 26,49 . Formation and accumulation of FUS-RNA-PRMT1 complexes could perturb several cellular pathways and cause abnormal neurite morphology. Abnormal axonal or dendritic phenotypes are common cellular phenotypes found in the brain tissue of patients with ALS and in ALS animal models 30,41 . It is still unknown, however, which types of RNA targets might contribute to neurite degeneration associated with ALS.
Among several potential RNA targets, we identified Nd1-L, a Kelch protein involved in the stabilization of actin filaments, as a potential component of FUS complexes 36,55 . Kelch repeats are known to be important for actin binding, protein folding, and protein-protein interaction 56 . FUS associates with the 3′ UTR of Nd1-L, which lacks a GGUG-type motif and facilitates transport of Nd1-L mRNA intro dendritic spines upon mGluR5 (metabotropic glutamate receptor 5) activation 8,36 . Our protein-RNA co-IP and FISH data show that the FUS-R521C mutant can form stable complexes with Nd1-L mRNA and sequester it into FUS-R521C-positive aggregates. FUS can bind to Nd1-L mRNA, impairing its correct translation which leads to its sequestration into FUS aggregates. Indeed, overexpression of Nd1-L in FUS-R521C-expressing neurons rescued neurite shortening upon oxidative stress, suggesting that FUS/Nd1-L mRNA complexes contribute to abnormal neurite morphology. Nd1-L could therefore represent a molecular link between the regulation of actin cytoskeleton and neurite morphology 57 . Recent studies reported that mutations of profilin 1 (PFN1), an actin binding protein, affect cytoskeletal dynamics and aggregation of TDP-43 58 . Moreover, ALS/FTD-linked C9ORF72 expansion has been shown to dysregulate actin Scientific RepoRts | 7:40474 | DOI: 10.1038/srep40474 dynamics in motor neurons 59 . Our study showed that Nd1-L expression rescued neurite shortening enhanced further by loss of PRMT1 in FUS-R521C expressing neurons, thus supporting that improper regulation of actin cytoskeleton might contribute to cellular pathogenesis associated with ALS. According to our data (Figs 5 and 7), overexpression of FUS (WT or R521C) itself does not significantly affect neurite shortening. When oxidative stress was exposed to neurons expressing FUS-R521C but not FUS-WT, neurite length was significantly reduced. Based on our results (Fig. 4B,C), oxidative stress enhanced accumulation of FUS-R521C-positive SGs which remained in the cytosol after removal of oxidative stress. Although a small portion of FUS-R521C-positive SGs and aggregates were accumulated in cytosol, oxidative stress-induced cytosolic aggregates of FUS-R521C/ PRMT1/Nd1-L mRNA rather than the nuclear sequestration of Nd1-L mRNA into FUS-R521C/PRMT1 might impair cellular pathway involved in the regulation of neurite morphology. However, further studies are need to investigate how cytosolic aggregates of FUS-R521C/PRMT1/Nd1-L mRNA induced by oxidative stress exactly contribute to neurite degeneration.
Consistent with our data, recent evidence showed that FUS-R521C can form stable FUS-Bdnf mRNA or FUS-MECP2 mRNA complexes, which impair transcription and/or RNA splicing suggesting that FUS mutants could disrupt target gene expression at a post-transcriptional level 49,60 . Recent RNA-seq transcriptome analyses have identified more target genes that are differentially regulated in FUS-WT and ALS-linked FUS mutants in cellular and animal models 49 .
One point of high impact in the present study is the identification of FUS/PRMT1/Nd1-L RNA complexes that can contribute to altered neurite morphology via the FUS-R521C mutation. This evidence opens to further research work to determine whether these abnormal complexes can interfere with RNA splicing in target genes, RNA transport granules, or local protein synthesis by synaptic activity. An additional topic for future studies is to investigate whether PRMT8, a neural-specific PRMT, is also sequestered from axonal and dendritic membranes into FUS-positive SG aggregates, and whether its alteration can cause synaptic deficits and neurodegeneration 18 .  Table 1). The amplification products FUS-WT or FUS mutants were inserted into p3XFLAG-CMV7.1 (Sigma Aldrich, St. Louis, MO) and pEGFP vectors (Clontech, Mountain View, CA) using HindIII and BamHI enzymes restriction sites. The amplification products PRMT1-WT or PRMT-G98R were inserted into pCMV-Myc-N (Clontech, Mountain View, CA) using EcoRI and KpnI enzymes restriction sites. The amplification product Nd1-L was inserted into pcDNA ™ 3.1/myc-His vectors (Thermo Fisher Scientific) using EcoRI and KpnI restriction sites.

Cell cultures, transfection, viral infection, and immunocytochemistry. All experimental procedures
were approved by the Institutional Animal Care and Use Committee of Hannam University and in accordance with their guidelines. Preparation of neuronal cultures from ICR mice brains, transfection and immunocytochemistry were conducted as previously described 61 . Plasmid DNA was transfected into primary cortical neuronal cells using Lipofactamine 2000 (Invitrogen, part of Thermo Fisher Scientific, Carlsbad, CA), at 3-5 days in vitro (DIV) according to manufacturer's procedure.
To generate 3xFLAG-FUS fusion AAV constructs, we performed PCR using each primer sets (Supplementary Table 1). PCR products were inserted into the pAAV-CW3SL-GFP vector using Acc I and Xho I enzymes restriction sites. The titer of the virus production was quantified with real-time PCR. For the generation of the virus, 3xFLAG-FUS fusion AAV constructs and viral packaging vectors were transfected into HEK293T cells. The 3xFLAG-FUS fusion AAV virus was diluted in phosphate-buffered saline (PBS) and infected into primary neuronal cells at DIV 3 for 6 d. Silver staining and LC-MS analysis. Silver staining was performed using a PageSilver ™ Silver Staining Kit (#K0681; Thermo Fisher Scientific, Waltham, MA) according to the manufacturer's instructions. Briefly, after SDS electrophoresis, the gel was placed in 50 ml fixing solution 1 [25 ml ethanol, 5 ml glacial acetic acid, diluted to 50 ml with deionized water (DW)] and 50 ml fixing solution 2 (30 ml ethanol diluted to 100 ml with DW) for 10 min at room temperature. After rinsing the gel three times with DW, 50 ml of sensitizing solution (0.2 ml sensitizer concentrate diluted to 50 ml with DW) was added, and the sample was shaken for 1 min. The gel was stained Fluorescence in situ hybridization (FISH). Primary cultures of mouse cortical neurons on glass were fixed in 4% paraformaldehyde (PFA) at room temperature for 10 min. After washing with 1X PBS twice, cells were permeabilized with 0.1% Triton X-100 for 10 min, rinsed with 1X PBS, and treated with 1.5% BSA for 30 min before staining with primary antibodies (anti-MAP2, anti-PRMT1 and anti-FLAG). After incubation with primary antibodies, cells were incubated with secondary antibodies conjugated with Alexa 488 and DyLight 405 (Jackson Immuno Research Laboratories, West Grove, PA). Immuno-stained cells were rehydrated with rehydration buffer [2X saline sodium citrate (SSC) and 50% formaldehyde] at room temperature for 5 min. Cy3-conjugated probes (Supplementary Table1) were diluted to 500 nM in hybridization buffer (2X SSC, 25% formaldehyde, 10% dextran sulfate, and 0.005% BSA) and incubated at 37 °C for 4 h or overnight. The hybridized samples were rinsed in rehydration buffer at 37 °C for 30 min and mounted onto glass slides with Vectashield mounting medium (Vector Laboratories, Burlingame, CA).