Arginine methylation by PRMT 1 regulates nuclear-cytoplasmic localization and toxicity of FUS / TLS harbouring ALS-linked mutations

Department of Neurology/Neurosurgery and Montreal Neurological Institute, McGill University, Montréal, Québec, Canada H3A 2B4, Terry Fox Molecular Oncology Group, The Bloomfield Centre for Research on Aging and the Segal Cancer Centre, Lady Davis Institute for Medical Research, Sir Mortimer B. Davis Jewish General Hospital, Department of Oncology and Department of Medicine, McGill University, Montréal, Québec, Canada H3T 1E2


INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease in which the dysfunction and loss of motor neurons from the brain and spinal cord leads to progressive muscle weakness and death typically within 1 -5 years (1).Approximately 80 -90% of cases occur sporadically (SALS) without known family history of the disease, while the other 5 -10% of cases are familial (FALS) (2,3).The first gene to be linked to ALS (ALS1) was SOD1 (encoding Cu/Zn-superoxide dismutase) and its mutations account for 20% of FALS cases (4).Mutations of the gene encoding fused in sarcoma (FUS), also called translated in liposarcoma (TLS), have recently been identified in a subset of FALS, termed ALS6 (5,6).
FUS is a member of the TET family of proteins that include Ewing's sarcoma and TATA-binding protein-associated factor TAF15 (7).The FUS gene was originally identified as a part of a fusion gene with transcription factor C/EBP homologous protein in myxoid and round cell liposarcomas (8,9).As a heterogeneous ribonuclear protein in its normal function, FUS has a variety of roles within the cell, including transcription, RNA processing and local translation of mRNA (7,10).In neurons, FUS is necessary for proper formation of dendritic spines and for the transport of mRNA along dendrites, including actin-encoding mRNA (11).FUS has been shown to interact with serine-arginine proteins involved in RNA splicing (12).It is also involved in microRNA processing and the formation of cytosolic stress granules (large complexes of mRNA and protein that form under conditions of stress) (13)(14)(15)(16)(17)(18)(19).
FUS WT is predominantly found in the nucleus; however, several studies have shown that ALS-linked mutations in FUS lead to a predominance of cytoplasmic versus nuclear localization (5,6,15 -19), inferring that toxicity of FUS mutants is somehow related to this imbalance.Many of the mutated amino acids are in the extreme C-terminus, which contains the nuclear localization signal (NLS) (16,17).This suggests that the abnormal nuclear/cytoplasmic distribution of FUS could result from impaired nuclear import.In addition, transportin-dependent nuclear transport of FUS mutants was reduced, with severity in proportion to the aggressivity of disease caused by that mutant (16).Another pathological hallmark in patients' autopsy tissue is the presence of FUS-immunoreactive inclusions, including granular, skeinlike, globular or flame-shaped inclusions in neurons and glia (16,(20)(21)(22).Stress granule markers colocalize with many FUS inclusions in ALS6 spinal cord and with mutant FUS expressed in cultured cells (15)(16)(17)(18)(19). FUS-immunoreactive inclusions also have been observed in the spinal cord of patients with SALS or several forms of non-SOD1 FALS (23), suggesting the more general relevance of FUS to ALS pathogenesis.
Posttranslational modifications, including methylation, can influence nuclear import.Protein arginine methylation adds mono-methyl or dimethyls to the guanidino nitrogen atoms of arginine (24).There are two major classes of enzymes responsible for protein arginine methylation: type I enzymes promote the formation of asymmetric v-N G ,N G -dimethylated arginines and type II enzymes catalyse the formation of symmetrical v-N G ,N' G -dimethylated arginines (24).Protein arginine methyltransferase 1 (PRMT1) is ubiquitously expressed and accounts for the majority of the asymmetrical dimethylarginines in the cell (25,26).PRMT1 has a preference for substrates that contain glycine and arginine-rich (GAR) sequences (26).Its substrates include histones, RNA-binding proteins and proteins that signal DNA damage (27).The PRMT1 gene has been disrupted in mice and homozygous mutant embryos die shortly after implantation, failing to develop beyond E6.5 (28).A conditional PRMT1 allele has been generated in mice; the loss of PRMT1 in mouse embryonic fibroblasts (MEFs) leads to severe genomic instability (29).We have shown that the loss of PRMT1 activity causes the accumulation of Sam68 in the cytoplasm due to hypomethylation (25), demonstrating that arginine methylation influences protein localization.
At least 20 arginine residues in the three GAR motifs are asymmetrically methylated (30).Interestingly, the majority of ALS-linked FUS mutations have been identified in the C-terminal region of the protein close to the last GAR motif (http://alsod.iop.kcl.ac.uk/Als/Overview/gene.aspx?gene_id=FUS) and the NLS (Fig. 1A).Because of the role of asymmetric arginine methylation by PRMT1 in nuclearcytoplasmic shuttling of RNA-binding proteins such as Sam68 (25), it is logical that arginine methylation by PRMT1 might also affect cellular localization of FUS and its disease-related mutants.Therefore, we hypothesized that altered arginine methylation at or near the GAR motif could contribute to the gain-of-function toxicity of ALS-linked FUS mutants.
In this study, we compared the overall level of asymmetric dimethylarginine methylation of FUS WT and ALS6 causing mutants, and assessed the effect of inhibiting this methylation on their nuclear/cytoplasmic distribution, including formation of cytoplasmic inclusion bodies; in addition, the toxicity of FUS mutants to motor neurons was examined, as they are preferentially vulnerable in the disease.

FUS WT and ALS-linked mutants are methylated by PRMT1
FUS contains three GAR motifs that harbour asymmetric dimethylarginines (Fig. 1A) (30).To determine whether PRMT1 methylates FUS, an in vitro methytransferase assay was performed using [methyl-3 H]-S-adenosyl-L-methionine.Glutathione-S-transferase (GST)-FUS, as well as a positive control (GST-GAR from fibrillarin), but not GST alone, were methylated by PRMT1 (Fig. 1B).To determine if PRMT1 methylates FUS in vivo, the methylation status of FUS was assessed in oestrogen receptor ligand-induced conditional PRMT1-knockout (PRMT1 FL/2;CreERT ) MEFs (29).The expression of CRE recombinase induced by 4-hydroxytamoxifen (OHT) led to the generation of PRMT1-null MEFs (see Materials and Methods), which harboured hypomethylated FUS as assessed by immunoblotting anti-FUS immunoprecipitates with the methyl-specific antibody ASYM24 (Fig. 1C).FUS frequently migrated as a doublet and this is likely due to additional post-translational modifications (Fig. 1C).To determine whether PRMT1 associates with FUS, co-immunoprecipitation studies were performed with HeLa cells co-transfected with expression vectors encoding myc-PRMT1 and Flag-FUS.Cell extracts were prepared and immunoprecipitations were performed using anti-myc and anti-Flag antibodies, and immunoglobulin G (IgG) as control (Fig. 1D).Myc-PRMT1 was detected in anti-Flag immunoprecipitates (Fig. 1D, upper panel) and vice versa (Fig. 1D, lower panel).These findings show that PRMT1 associates with and methylates FUS.
Next we examined whether the ALS-linked FUS mutants (R521C, R521G or R521H), the most frequently observed mutations in the ALS6 patients (5,6), were arginine methylated like FUS WT .HEK293 cells were transfected with expression vectors encoding Flag-FUS WT , Flag-FUS R521C , Flag-FUS R521G or Flag-FUS R521H and then were assessed for arginine methylation by immunoblotting with ASYM24.FUS R521C , Flag-FUS R521G , FUS R521H and FUS WT were equally recognized by ASYM24 (Fig. 1E).FUS WT and ALS-linked FUS mutants frequently migrated as a doublet and this is likely due to additional post-translational modifications.To this end, our findings show that ALS-linked FUS mutants are arginine methylated at an overall level similar to FUS WT .

Reduced cytoplasmic localization and fewer cytoplasmic inclusion bodies of ALS-linked FUS mutants in PRMT1-deficient cells
The next step was to examine the influence of PRMT1 arginine methylation on the cellular localization of FUS WT and the ALS-linked FUS mutants.For these experiments, our In vitro PRMT1 methytransferase assay.Recombinant GST-PRMT1 and substrates GST, GST-FUS and GST-GAR were purified from bacteria and subjected to PRMT1 activity assay as described in the 'Materials and Methods'.GST was used as a negative control (lane 4) and GST-GAR as a positive control (lane 6).The migration of the molecular mass markers is shown in kDa.(C) FUS is methylated in vivo by PRMT1.Inducible PRMT1-knockout MEFs (PRMT1 FL/2;CreERT ) were treated with 500 nM of the estrogen receptor ligand, 4-hydroxytamoxifen (OHT) for 4 days (+) to induce PRMT1 knockout or left non-treatment (2).The cells were lysed 2 days after treatment.Total cell lysates were subjected to immunoprecipitation and western blot analysis as described in the 'Materials and Methods' with the indicated antibodies.The migration of FUS, PRMT1 and the heavy chain of IgG is indicated.TCL represents the total cell lysate or input.(D) FUS associates with PRMT1.Co-immunoprecipitation of HeLa cells transfected with Flag-FUS and myc-PRMT1 plasmids.Cell extracts were immunoprecipitated with anti-Flag (upper panel), anti-myc (lower panel) and with control IgG.The bound proteins were analysed by immunoblotting.The migration of myc-PRMT1 and Flag-FUS is shown.(E) Arginine methylation of FUS WT and ALS-linked FUS mutants.HEK293 cells were transfected with empty plasmid vector (pcDNA3) or the vectors encoding Flag-tagged FUS WT , R521C, R521G or R521H FUS.Forty-eight hours after transfection, the cells were lysed.The protein from the input (TCL) and anti-Flag immunoprecipitation were immunoblotted with anti-Flag antibodies (left and middle panels) or with ASYM24 (right panel).The migration of FUS and the heavy chain of IgG as well as molecular mass markers is shown.
inducible conditional PRMT1 FL/2;CreERT MEFs were used (29), PRMT1-null cells being generated by treating the cells with OHT (Fig. 2A).Twenty-four hours after transfection with Flag-tagged FUS constructs, the cellular localization of FUS WT and the ALS-linked FUS mutants was examined in control (PRMT1 FL/2;CreERT ; 2OHT) and PRMT1-null (PRMT1 FL/2;CreERT ; +OHT) MEFs by indirect immunocytochemistry with anti-Flag.As reported previously (vide supra), FUS WT was localized predominantly in the nucleus, whereas the FUS mutants (R521C, R521G and R521H) were localized predominantly in the cytoplasm in inclusion bodies (Fig. 2B).The cytoplasmic/non-nuclear localization of FUS and the presence of FUS in cytoplasmic inclusion bodies were quantified.Regardless of PRMT1 status, 10% of cells had FUS WT localized in the cytoplasm (Fig. 2C) and 3% of cells had cytoplasmic inclusion bodies (Fig. 2D).The ALS-linked FUS mutants (R521C, R521G and R521H) accumulated in the cytoplasm of control MEFs (in 29, 36 and 22% of cells, respectively, Fig. 2C), and 6 -12% of cells harbouring any of the FUS mutants contained pronounced cytoplasmic inclusions (Fig. 2D).Loss of PRMT1 influenced the distribution of ALS-linked FUS mutants, promoting nuclear localization and reducing cytoplasmic accumulation and the formation of inclusions; following transfection of FUS expression plasmids into PRMT1-null MEFs, the number of cells harbouring any of the FUS mutants that exhibited cytoplasmic accumulation of the protein and cytoplasmic inclusion bodies was reduced by 50% (Fig. 2C and D).These results demonstrate that PRMT1-mediated arginine methylation is a determining factor in the accumulation of FUS mutants in the cytoplasm and the formation of cytoplasmic inclusion bodies.
A similar influence of arginine methylation on FUS localization was demonstrated in human cell lines.FUS WT or the ALS-linked FUS mutants were expressed in HEK293 cells previously treated with siRNA to knockdown PRMT1 (siPRMT1) or siLuciferase as control (siControl).A .90% knockdown of PRMT1 in siPRMT1-transfected cultures was confirmed by immunoblotting (Fig. 3A), and distribution of FUS proteins was assessed by indirect immunocytochemistry with anti-Flag antibody.Representative images of cells expressing FUS WT and ALS-linked FUS mutants are shown in Figure 3B.FUS WT was predominantly nuclear in both siControl-and siPRMT1-transfected cells and few cells contained cytoplasmic inclusion bodies (Fig. 3C).In contrast, 6 -12% of siControl-transfected cells expressing any of the ALS-linked FUS mutants contained pronounced cytoplasmic inclusions (Fig. 3C), but expressing these proteins on a PRMT1 knockdown background (siPRMT1 transfected) resulted in a .66%reduction in the percentage of cells with mutant FUS in cytoplasmic inclusions (Fig. 3C).Collectively, these findings show that in the absence of PRMT1, there is reduced formation of mutant FUS cytoplasmic inclusions in human and mouse cells.

Phenotype in primary cultured motor neurons expressing ALS-linked FUS mutants: formation of cytoplasmic inclusions and mitochondrial shortening
To examine the toxicity of ALS-linked FUS mutants in motor neurons, which are preferentially vulnerable in ALS, a primary culture model of ALS6 was developed by expressing Flagtagged FUS WT or ALS-linked FUS mutants in motor neurons of dissociated murine spinal cord-dorsal root ganglion (DRG) cultures.The methodology was the same as the one previously used to establish models of FALS due to mutations in SOD1 and TARDBP, and of Charcot -Marie-Tooth disease type 2E (CMT2E) due to mutations in NEFL (31)(32)(33).The expression of ectopic FUS WT or ALS-linked FUS mutants was achieved in motor neurons by intranuclear microinjection of expression vectors, as motor neurons in long-term spinal cord cultures are not amenable to gene transfer by transfection.The distribution of the R521G and R521H mutants in motor neurons was more cytoplasmic than nuclear compared with FUS WT , as revealed by immunolabelling with anti-Flag (Fig. 4A); the majority of motor neurons expressing these mutants exhibited predominantly cytoplasmic FUS and contained FUS-positive cytoplasmic inclusions (Fig. 4B).The preponderance of FUS inclusions colocalized with the RNA marker SYTO RNASelect TM (Fig. 4C), showing that these bodies are RNA granules, consistent with previous reports (15)(16)(17)(18)(19)34).
Expression of ALS-linked FUS mutants in motor neurons did not significantly alter their viability compared with FUS WT or empty plasmid control over a period of 7 days (data not shown).Therefore, the effect of ALS-linked FUS mutants on mitochondrial morphology was assessed as a potential phenotype in viable neurons, given that mitochondrial abnormalities have been documented in both sporadic and familial ALS (35) and that dramatic shortening of mitochondria (mitochondrial rounding/fragmentation) was an early manifestation of the toxicity of SOD1 G93A and NFL mutants in similar models of FALS1 and CMT2E (33,36,37).To image mitochondria, pOCTeGFP (eGFP with an N-terminal ornithine carbamoyltransferase signal sequence for mitochondrial targeting) was co-expressed with FUS WT or ALS-linked FUS mutants.Overexpression of FUS WT had no significant effect on mitochondrial length compared with empty plasmid (Supplementary Material, Fig. S1), whereas mitochondria were significantly shorter in motor neurons expressing either R521G or R521H FUS, compared with neurons expressing FUS WT (Fig. 5A and C).Reduction in mitochondrial length was significantly greater in motor neurons with predominantly cytoplasmic mutant FUS, compared with those with predominantly nuclear localization, whereas cytoplasmic FUS WT had no effect on mitochondrial length regardless (Fig. 5B).These findings show that the expression of the ALS-linked FUS mutants in the cytoplasm leads to mitochondrial shortening in motor neurons.

Concurrent knockdown of PRMT1 in motor neurons promotes cytoplasmic localization of FUS mutants
In order to examine the effects of PRMT1 arginine methylation on FUS WT and ALS-linked FUS mutants in primary motor neurons, shRNA constructs targeted against PRMT1 were used to knockdown expression.The difficulties with consecutive delivery of plasmids by multiple injections complicated the pre-emptive knockdown of PRMT1; therefore, expression plasmid encoding an shRNA that targets PRMT1 (or scrambled sequence as a control) was injected in the same solution as the plasmid encoding FUS WT or ALS-linked FUS mutants.Localization of FUS WT and ALS-linked FUS mutants and formation of inclusions as well as the level of PRMT1 were examined 3 days postmicroinjection by double label immunocytochemistry.PRMT1 labelling in motor neurons was found to be predominantly nuclear (data not shown), although in various cell lines, distribution in nuclear and cytoplasmic compartments is variable (38).In the case of motor neurons microinjected with shPRMT1 plasmid, only those with the loss of PRMT1 immunolabelling were included in the analysis of FUS distribution.PRMT1 knockdown did not affect the distribution of FUS WT , or its localization to inclusions (Fig. 6A).Compared with scramble control, PRMT1 knockdown led to an increase in the percentage of neurons with R521G or R521H FUS mutants found predominantly within the cytoplasm (Fig. 6B  and C), as well as an increase in the percentage of motor neurons with FUS R521H inclusions (Fig. 6C), the opposite to observations in MEFs and HEK293 cells depleted of PRMT1.This difference could have resulted from inherent differences in the properties of motor neurons or because the FUS mutants were methylated or partially methylated before the efficient knockdown of PRMT1 occurred, since we were unable to express FUS on a PRMT1 knockdown background in this system.
Given that cytoplasmic accumulation of mutant FUS reduced mitochondrial length in motor neurons (Fig. 5), the effect of manipulating PRMT1 on this phenotype was evaluated.Plasmid encoding shPRMT1 or scrambled sequence was coexpressed with plasmid encoding Flag-tagged FUS WT   or ALS-linked mutant.In neurons injected with scrambled sequence, mitochondria length was significantly shorter in neurons expressing FUS R521G or FUS R521H compared with neurons expressing FUS WT , as expected from the previous experiment.In addition, mitochondrial length was reduced by PRMT1 knockdown, even in neurons expressing FUS WT (Fig. 6D), making it difficult to determine if the further reduction in length in neurons co-expressing mutant FUS and shPRMT1 was due to increased cytoplasmic levels of the mutant protein or another effect of PRMT1 knockdown.

Prior methyltransferase inhibition reduces retention of ALS-linked FUS mutants within the cytoplasm and the formation of inclusion bodies in motor neurons
The knockdown of PRMT1 in motor neurons using shRNAs favoured the cytoplasmic localization of ALS-linked FUS mutants and the formation of inclusion bodies.As plasmids encoding the ALS-linked FUS mutants were simultaneously injected with the PRMT1 shRNA expression plasmid, it was possible that some of the mutant FUS protein was methylated or partially methylated before efficient knockdown of PRMT1 occurred.To address this possibility, spinal cord-DRG cultures were treated with adenosine dialdehyde (Adox), a general inhibitor of methyltransferase activity (39)(40)(41), prior to expression of FUS in motor neurons.Cultures were treated with 20 mM Adox or vehicle for 48 h before intranuclear microinjection of plasmid encoding Flag-tagged FUS WT or an ALS-linked FUS mutant.Pre-treatment of cultures with 20 mM Adox resulted in global reduction in asymmetric dimethylarginine, as visualized by immunoblotting with ASYM24 and ASYM25 antibodies (Fig. 7A; Supplementary Material, Fig. S2).Interestingly, the pretreatment with the methyltransferase inhibitor led to a decrease in cytoplasmic localization of the ALS-linked FUS mutants in motor neurons (Fig. 7C and D).Moreover, a large reduction in the percentage of FUS R521H -expressing motor neurons harbouring inclusion bodies was observed (Fig. 7D).Treatment with the methyltransferase inhibitor also shifted localization of FUS WT from cytoplasmic to nuclear, reducing formation of inclusions (Fig. 7B).These findings suggest that arginine methylation by PRMT1 participates in the regulation of the nuclear-cytoplasmic shuttling of FUS WT and particularly ALS-linked FUS mutants in motor neurons, a vulnerable cell type in ALS.

DISCUSSION
This study investigated the role of asymmetric arginine methylation on the intracellular compartmentalization of FUS WT and ALS6-linked mutants, using cell lines and a novel primary culture model in which these proteins were expressed in motor neurons of dissociated spinal cord-DRG cultures.The mislocalization of mutant FUS to the cytoplasm and association with cytoplasmic inclusions also was associated with altered mitochondrial morphology in motor neurons, a factor that could contribute to toxicity.The data show that FUS interacts with and is a substrate of PRMT1 and that arginine methylation by this enzyme is a determinant of the  aberrant cytoplasmic localization of ALS-linked FUS mutants, thereby contributing to their toxicity.

FUS is mislocalized to the cytoplasm and forms inclusions in both primary motor neurons and cell lines
Similar to several other studies (5,6,15 -19), mutant FUS was aberrantly localized in the cytoplasm, whereas the majority of FUS WT was nuclear.Also, cytoplasmic mutant FUS was concentrated in inclusion bodies, many of which were RNA-containing granules.
Evidence supports the importance of cytoplasmic accumulation of FUS in the pathogenesis of ALS6, and even SALS and other non-SOD1 FALS, given the prevalence of the cytoplasmic mislocalization and presence of FUS-containing inclusions in spinal motor neurons and glia in autopsy material (20 -23).The cytoplasmic localization is necessary for neurodegeneration; this was recently confirmed by Lanson et al. (42), who reported that deletion of the nuclear export signal of FUS strongly suppressed toxicity of mutants in a Drosophila model.

Cytoplasmic localization worsens the mitochondrial abnormalities caused by mutant FUS
Cytoplasmic mutant FUS, but not FUS WT , promoted mitochondrial shortening in cultured motor neurons in this study, supporting the requirement of cytoplasmic mislocalization for the toxic gain of function conferred by FUS mutations.Mitochondrial abnormalities are a hallmark of both sporadic and familial ALS (43,44); mitochondrial rounding has been observed in other forms of FALS by several groups, including our own (36,45 -49), and thus might represent a common pathogenic mechanism of ALS6 with other forms of SALS and FALS.
How cytoplasmic mutant FUS causes mitochondrial morphological abnormalities remains to be determined.Distribution of FUS did not appear mitochondrial, but this does not rule out more subtle interactions that cause an imbalance in the processes that control morphology, particularly fusion/ fission.Mitochondrial fragmentation, resulting from impaired fusion or increase in fission, is a common feature of toxicity including neuronal excitoxicity (50) and occurs in other motor neuron disorders, specifically FALS1 caused by A comprehensive analysis of the effect of FUS mutants on mitochondrial function is being conducted in a separate study to assess the contribution to toxicity and whether a primary event or secondary response to injury.

Arginine methylation by PRMT1 contributes to the toxic gain of function of FUS mutants
Recently, with the help of proteomic studies, an increasing number of proteins have been found to be methylated at arginine residues (24).These substrates represent a broadening spectrum of cellular processes (51).Therefore, it should not be a surprise that defects in arginine methylation are potentially involved in multiple human diseases (52) such as breast cancer (53) and castration-resistant prostate cancer (52).Moreover, many human disease-associated proteins, such as p53 (54), and PABPN1 (55,56) are arginine methylated.However, it currently remains unclear how the hypomethylation of cellular proteins influence human disease.In this study, pre-emptive PRMT1 knockdown lessened cytoplasmic distribution and inclusion formation of FUS mutants, implying that arginine methylation is required for the toxic gain of function of the mutated FUS protein associated with its cytoplasmic localization.
Prior reduction in PRMT1 expression or activity reduces cytoplasmic accumulation of FUS mutants in MEFs, HEK293 cells and primary motor neurons FUS was predominantly nuclear in motor neurons, MEFs and HEK293 cells, although a previous report indicated that FUS was excluded from the nucleus in hippocampal neurons (34).Knockdown of PRMT1 expression reduced cytoplasmic were expressed by intranuclear microinjection.Forty-eight hours after microinjection, cultures were fixed and immunolabelled with anti-Flag.Antibody-labelled neurons were scored according to nuclear or cytoplasmic localization of FUS and for the presence of inclusions, and the numbers presented as percent of the total number of neurons evaluated.Adox pretreatment significantly reduced the percentage of neurons with (B) FUS WT , (C) FUS R521G or (D) FUS R521H predominantly localized in the cytoplasm, and the percentage of neurons expressing (B) FUS WT or (D) FUS R521H with cytoplasmic inclusions.Shown are means + SEM; * P , 0.05, two-tailed, unpaired t-test.
accumulation of FUS mutants and retention in nuclei of MEFs or HEK293 cells.This was not apparent in cells expressing FUS WT , in which localization was already nuclear.However, an influence of arginine methylation on localization of FUS WT is supported by experiments in spinal cord-DRG cultures; in a fraction of motor neurons, FUS WT was cytoplasmic, but in cultures pretreated with Adox to inhibit arginine methylation, FUS WT was exclusively nuclear.Since Adox is a general methyltransferase inhibitor, it will be important to show that PRMT1 inhibitors such as AMI-1 (57) decrease the cytoplasmic localization of the ALS-linked FUS mutants in motor neurons.Arginine methylation has been found to influence the nuclear-cytoplasmic distribution of Sam68 (25) and several other RNA-binding proteins as well (58 -60).
Nuclear localization of FUS is not strictly dependent on arginine methylation, as recent studies have identified a NLS in the C-terminus of FUS (16,19).Deletion of the C-terminus, including the multiple ALS-linked mutation sites, caused cytoplasmic accumulation and aggregation of FUS (19,42), suggesting that the C-terminal ALS-linked FUS mutants predominantly affect protein trafficking, with subsequent aggregation as the natural consequence.Altered methylation could have a more profound effect on localization of FUS mutants when the protein is already mislocalized as a result of the mutation interfering with the function of the NLS.
Arginine methylation could also alter protein conformation and promote protein interactions leading to aggregation or sequestration in RNA granules.Addition of methyl groups to arginine residues could prevent hydrogen bonding due to steric hindrance, thus disrupting molecular interactions.In contrast, arginine methylation could also enhance molecular interactions, as the addition of methyl groups increases hydrophobicity of the arginine.However, increased hydrophobicity due to the addition of single or double methyl groups to a single arginine is limited.We propose that the increased hydrophobicity of methylated arginine residues could partially contribute to the aggregation of mutant FUS proteins in the cytoplasm through non-specific hydrophobic interactions; thus, reducing arginine methylation would prevent the aggregation by reducing the hydrophobicity.On the other hand, the C-terminal ALS-linked mutations would affect the cytoplasmic-nuclear trafficking, resulting in accumulation of the protein in the cytoplasm.The mutated residues might not be directly involved in the subsequent aggregation, but the accumulation of the protein in the cytoplasm would favour the formation of inclusions or stress granules, as can occur with FUS WT .
The FUS-methylated arginines identified by mass spectrometry are located mainly within GAR motifs (30).The ALS-linked FUS-mutated arginines (R514, R518, R521, R522 and R524) are not located within a GAR motif and were not previously identified to be arginine methylated (30).Thus, the methylation of arginines within FUS GAR motifs contributes to the toxicity of ALS-linked mutants.In the PRMT1-deficient cells, a complete lack of arginine methylation, as detected with ASYM24 (Fig. 1C), was observed, suggesting that PRMT1 is likely to be the main enzyme responsible for the in vivo arginine methylation of FUS.It is possible that FUS may also be arginine methylated by other PRMTs.
An unresolved issue is whether mutations directly affect arginine methylation of FUS.No major perturbation in arginine methylation was identified among mutants or compared with FUS WT by western analysis using an antibody recognizing asymmetric dimethylarginine residues; however, site-specific differences in methylation might occur that would not be detected by this methodology.Also relevant is that FUS is a constituent of cytoplasmic inclusions in sporadic and other familial forms of ALS (vide infra) demonstrating that FUS can be mislocalized and form inclusions as a result of factors other than mutation; e.g.abnormal post-translational modifications or protein -protein interactions.In this respect, the pathobiology of FUS resembles TDP-43: mutations in the TARDP gene cause another form of FALS, and TDP-43 WT is mislocalized to cytoplasmic inclusions in SALS (61,62).
In conclusion, this study provides evidence that aberrant cytoplasmic localization of mutant FUS is pathogenic and dependent upon asymmetric arginine methylation.These data raise the possibility of using PRMT1 inhibitors therapeutically.Caveats are that incomplete prevention of arginine methylation exacerbated this toxic property in motor neurons in the primary culture model and knockdown of PRMT1 itself reduced mitochondrial length in motor neurons.Further studies are required to identify site-specific differences in methylation of FUS mutants and how methylation normally contributes to nuclearcytoplasmic trafficking of FUS.

Culture, transfection and retroviral infection of cell lines
Conditional Cre-induced PRMT1 knockout (PRMT1 FL/2 ) MEFs were isolated from E14.5 embryos.Spontaneously immortalized PRMT1 FL/2 MEFs were created according to the standard 3T3 protocol as described previously (29).The immortalized PRMT1 FL/2 MEFs were then stably transfected by a plasmid DNA encoding oestrogen receptor -CRE fusion protein (ER-CRE).PRMT1 knockout was induced by treating the stable MEF cell line (PRMT1 FL/2;CreERT ) with 500 nM of the oestrogen receptor ligand, 4-hydroxytamoxifen (OHT), for 4 days as described previously (29).OHT was removed and the cells were then transfected with plasmid encoding Flag-tagged FUS or an ALS-linked FUS mutant using Lipofectamine TM 2000 (Invitrogen Life Technologies) according to the manufacturer's instructions.
HEK293 cells were transfected with siRNAs using Lipofectamine TM RNAi MAX (Invitrogen Life Technologies) according to the manufacturer's instructions using a final concentration of 20 nM siRNA.The siRNAs were purchased from Dharmacon (Lafayette, CO, USA).The target sequence of PRMT1 siRNA was 5 ′ -CGU CAA AGC CAA CAA GUU A-3 ′ .A Luciferase siRNA, with the sequence 5 ′ -AAC ACU UGU CAC UAC UUU CUC UU'-3 ′ , served as the control.Forty-eight hours later, the siRNA-transfected HEK293 cells were transfected with plasmid encoding Flag-tagged FUS or an ALS-linked FUS mutant using standard calcium phosphate precipitation.
HeLa cells were transfected with plasmid DNAs using Lipofectamine TM 2000 (Invitrogen Life Technologies) according to the manufacturer's instructions.The cells were cultured in Delbecco's minimum essential medium containing 10% fetal bovine serum (FBS).

Dissociated spinal cord-DRG cultures
Primary cultures of dissociated spinal cord (along with attached DRG) were prepared from embryonic day 13 (E13) CD1 mouse embryos (Charles River Laboratories, Wilmington, MA, USA) as previously described (64), and plated at a density of 350 000 per well in 12-well Nunclon culture dishes containing 18 mm round glass coverslips coated with poly-D-lysine plus Matrigelw basement membrane matrix (Invitrogen Life Technologies).Cells were maintained in the modified N3 medium [minimum essential medium (MEM) enriched with 5 g/l glucose and supplemented with 3% horse serum, 10 mg/ml bovine serum albumin, 26 ng/ ml selenium, 20 mg/ml triiodothyronine, 10 mg/ml insulin, 200 mg/ml transferrin, 32 mg/ml putrescine, 9.1 ng/ml hydrocortisone, 13 ng/ml progesterone and 10 ng/ml nerve growth factor].Cultures were used in experiments 3 -7 weeks after dissociation.At this age in vitro, motor neurons can be distinguished from other types of neurons and glia in the cultures because they develop and differentiate to resemble their counterparts in intact spinal cord, both morphologically and by expression of biological markers.They have large cell bodies (.20 mm in diameter) and dendritic trees, and express choline acetyltransferase, glutamate receptors, neurofilament proteins and the motor neuron transcription factor, Hb9 (31,65,66).

DNA transfer into primary motor neurons
Motor neurons in these long-term cultures are not amenable to liposome-mediated gene transfer; plasmids were introduced by intranuclear microinjection 3 -7 weeks after initial plating of dissociated spinal cord-DRG cultures.The injectate contained 20 mg/ml 70 kDa dextran (Invitrogen Life Technologies) in 5 mM Tris, 0.5 mM EDTA, pH 7.4 (Invitrogen Life Sciences).Each coverslip was transferred into microinjection buffer (MEM lacking NaHCO 3 and supplemented with 5 g/l D-glucose, pH adjusted to 7.4) and placed on the stage of a Zeiss Axiovert 35 microscope.Microinjection was accomplished using an Eppendorf 5246 (or Eppendorf FemtoJet transjector) and an Eppendorf 5171 micromanipulator under ×200 magnification.After microinjection, the coverslip was transferred into the modified N3 medium containing 0.75% gentamicin (Sigma-Aldrich) and returned to the incubator.First measurements were made no less than 16 h after microinjection so that any motor neurons dying from the injection procedure were not included in the experiment.

Immunolabelling and imaging of MEFs and HEK293 cells
Cells growing on glass coverslips were washed with PBS twice and fixed with 4% paraformaldehyde at room temperature for 10 min.The cells were then permeabilized in 0.5% Triton X-100 for 10 min.Following three washes with PBS, cells were blocked with 10% FBS in PBS and labelled with mouse anti-Flag antibody diluted in PBS containing 5% FBS.After three washes, the cells were labelled with Alexa Fluor 488-conjugated goat anti-mouse secondary antibody.DNA was counterstained with 4,6-diamidino-2-phenylindole; after three washes with PBS coverslips were mounted on slides using Immu-Mount (Thermo Scientific from Fisher Scientific, Nepean, ON, USA).Images were acquired using a Zeiss M1 microscope with epifluorescence optics.

Labelling of spinal cord-DRG cultures and imaging of motor neurons
For indirect immunocytochemistry, cultures were fixed with 3% paraformaldehyde in PBS for 10 min followed by permeabilization in 0.5% NP-40 in PBS for 1 min and postfixation in 3% paraformaldehyde in PBS for an additional 2 min.Cultures were blocked in 5% horse serum in PBS and then labelled with primary antibody for 30 min at room temperature followed by three washes in PBS, and then incubation with secondary antibody for 30 min followed by another three washes in PBS.
For determining RNA content of FUS-containing inclusions in motor neurons, spinal cord-DRG cultures were labelled with SYTO RNA Select (Invitrogen Life Technologies) at a concentration of 12 mM for 60 min.Coverslips were then fixed in cold 100% methanol for 4 min.
Images were viewed with a Zeiss LSM710 confocal microscopy or with a Zeiss AxioImager A1 microscope equipped with an AxioCam MRc5 digital camera (Carl Zeiss Canada, Toronto, ON, USA).Images were acquired using AxioVision software (Carl Zeiss) and processed and analysed using ImageJ (NIH; http://rsbweb.nih.gov/ij/).

Measurement of motor neuron viability and mitochondrial morphology
Viability of motor neurons was evaluated as previously described (36) following co-injection of an inert fluorescent marker (70 kDa dextran-FITC; Invitrogen Life Technologies) and by counting neurons expressing the marker by epifluorescence and phase microscopy over a period of 1 week.
Neuronal mitochondria are normally elongated and/or reticular in morphology.Mitochondrial length is a measure of rounding and loss of the reticular network.Mitochondrial length was measured as previously described (36) following co-expression of pOCT-eGFP with a mitochondrial targeting signal sequence.Mitochondrial lengths were measured using ImageJ and mean and standard error of the mean (SEM) were calculated in Microsoft Excel.Statistical analysis was accomplished using VassarStats (faculty.vassar.edu/lowry/VassarStats.html)using ANOVA or t-test where indicated.Significance was established at P , 0.05.

Treatment of motor neurons with Adox
Spinal cord-DRG cultures were treated with 20 mM Adox or vehicle as control for 48 h before microinjection of plasmid encoding Flag-tagged FUS WT or an ALS-linked FUS mutant into motor neuronal nuclei.Two days later, the cells were fixed and immunolabelled with anti-Flag antibody.Inhibition of asymmetric arginine methylation in the cultures was verified by immunoblotting (vide infra).

Immunoprecipitation and western blotting
HeLa, HEK293 or MEFs were lysed in buffer containing 1% Triton X-100, 150 mM NaCl, 20 mM Tris (pH 7.5) and proteinase inhibitor cocktail for 10 min on ice.Total cell lysates were clarified by centrifugation for 10 min at 10 000g at 48C.The lysates were incubated with the antibodies 1 h or overnight at 48C and then 20 ml of 50% protein A slurry was added.The mixture was incubated for 30 min at 48C.The Protein A Sepharose beads were washed twice with lysis buffer and once with 1× PBS.The samples were boiled and subjected to standard western blot analysis.In Figure 1E, agarose beads chemically coupled with anti-Flag antibody (Sigma) were used in the immunoprecipitation, in which the cell lysates were incubated with the beads at 48C for 2 h and then washed twice with lysis buffer and once with 1× PBS.
To examine inhibition of arginine methylation by Adox, spinal cord-DRG cultures were harvested 72 h after treatment with 0 mM, 20 mM or 40 mM Adox and lysed in 22.5 mM Tris (pH 6.8) containing 2% SDS and subjected to western blot analysis using the antibody ASYM24 and ASYM25 (Millipore Inc).

Methylation assay
Purification of the recombinant GST fusion proteins and in vitro arginine methylation assay were performed as described previously (25).

Figure 1 .
Figure1.FUS WT is methylated by PRMT1 and the ALS-linked mutations do not affect total arginine methylation of FUS.(A) Schematic representation of the FUS protein with GAR motifs and the FALS-linked C-terminal mutations.(B) In vitro PRMT1 methytransferase assay.Recombinant GST-PRMT1 and substrates GST, GST-FUS and GST-GAR were purified from bacteria and subjected to PRMT1 activity assay as described in the 'Materials and Methods'.GST was used as a negative control (lane 4) and GST-GAR as a positive control (lane 6).The migration of the molecular mass markers is shown in kDa.(C) FUS is methylated in vivo by PRMT1.Inducible PRMT1-knockout MEFs (PRMT1 FL/2;CreERT ) were treated with 500 nM of the estrogen receptor ligand, 4-hydroxytamoxifen (OHT) for 4 days (+) to induce PRMT1 knockout or left non-treatment (2).The cells were lysed 2 days after treatment.Total cell lysates were subjected to immunoprecipitation and western blot analysis as described in the 'Materials and Methods' with the indicated antibodies.The migration of FUS, PRMT1 and the heavy chain of IgG is indicated.TCL represents the total cell lysate or input.(D) FUS associates with PRMT1.Co-immunoprecipitation of HeLa cells transfected with Flag-FUS and myc-PRMT1 plasmids.Cell extracts were immunoprecipitated with anti-Flag (upper panel), anti-myc (lower panel) and with control IgG.The bound proteins were analysed by immunoblotting.The migration of myc-PRMT1 and Flag-FUS is shown.(E) Arginine methylation of FUS WT and ALS-linked FUS mutants.HEK293 cells were transfected with empty plasmid vector (pcDNA3) or the vectors encoding Flag-tagged FUS WT , R521C, R521G or R521H FUS.Forty-eight hours after transfection, the cells were lysed.The protein from the input (TCL) and anti-Flag immunoprecipitation were immunoblotted with anti-Flag antibodies (left and middle panels) or with ASYM24 (right panel).The migration of FUS and the heavy chain of IgG as well as molecular mass markers is shown.

Figure 2 .
Figure2.Reduced cytoplasmic localization and fewer cytoplasmic inclusion bodies of ALS-linked FUS mutants in PRMT1-null MEFs.PRMT1 FL/2;CreERT MEFs were left untreated (2 OHT) or treated with OHT (+ OHT) for 4 days to generate PRMT1-null MEFs.The cells were transfected 1 day later with Flag-FUS WT or ALS-linked FUS mutants.Twenty-four hours after transfection, the cells were fixed and immunolabelled with anti-Flag antibody.(A) Western blot analysis to determine PRMT1 knockout.PRMT1 FL/2;CreERT MEFs were untreated (2OHT) or treated with OHT (+ OHT) for 4 days to knockout PRMT1 and the cells were lysed after an additional 2 days.Total cell lysates (TCL) were subjected to immunoblotting with anti-PRMT1 and anti-tubulin antibodies, respectively.(B) Representative images of the typical distribution of the wild-type (WT) FUS and each mutant expressed in untreated MEFs (2OHT).Flag-FUS WT was nuclear in almost all cells, like the endogenous FUS in non-transfected cells; mutants had increased cytoplasmic localization, including in inclusion bodies.Nuclei were stained with DAPI.Scale bar ¼ 10 mm.(C and D) Anti-Flag labelled cells were scored according to whether labelling (representing plasmid-derived FUS) was localized in the nucleus or cytoplasm, and whether they contained cytoplasmic inclusion bodies; each measurement was expressed as a percentage of the total number of transfected cells evaluated.The graphs show the means + SEM from two independent experiments performed in duplicate, with .20 different microscopic fields (×200 magnification) analysed in each.Statistical significance between + OHT and 2OHT was assessed using two-tailed unpaired Student's t-test.* P , 0.05, * * P , 0.01 and * * * P , 0.001.

Figure 3 .
Figure 3. PRMT1 knockdown prevents FUS mutants R521C, R521G and R521H from localizing to the cytoplasm or forming inclusions in human embryonic kidney (HEK) 293 cells.HEK293 cells were transfected with control siRNA (siLuciferase) or siPRMT1 siRNA.Forty-eight hours after transfection, the cells were transfected with plasmid encoding Flag-tagged FUS WT or one of the ALS-linked FUS mutants.Twenty-four hours later, the cells were fixed and immunolabelled with anti-Flag antibody.Nuclei were stained with DAPI.(A) Western blot analysis to determine PRMT1 knockdown.The cells were lysed 72 h after transfection with the siRNA.Total cell lysates were subjected to western blot analysis with anti-PRMT1 and anti-tubulin antibodies, respectively.(B) Representative images of the typical distribution of FUS WT and each mutant expressed in untreated cultures.Flag-FUS WT was nuclear in almost all cells, like the endogenous FUS shown in non-transfected cells; mutants had increased cytoplasmic localization, including in inclusion bodies.Nuclei were counterstained with DAPI.Scale bar ¼ 10 mm.(C) The cells were scored for cytoplasmic localization of FUS and the presence of inclusions, and the numbers expressed as a percentage of total transfected cells evaluated.The graphs show the means + SEM of data from two independent experiments performed in duplicate, with .20 different microscopic fields (×200 magnification) analysed in each.Statistical significance between siLuciferase-treated control (siLuc) and siPRMT1-treated cells was assessed using two-tailed unpaired Student's t-test.* * * P , 0.0001.

Figure 4 .
Figure 4. Cellular distribution of FUS WT and mutant FUS in motor neurons.R521G and R521H FUS showed a significant shift from nuclear to cytoplasmic localization relative to FUS WT .Flag-FUS WT , R521G or R521H FUS were expressed in motor neurons of dissociated spinal cord-DRG cultures by intranuclear microinjection of plasmid expression vectors.Cultures were fixed on day 3 post-microinjection and labelled with anti-Flag antibody, revealing a predominance of cytoplasmic localization and cytoplasmic inclusion bodies in motor neurons expressing R521H or R521G FUS compared with FUS WT .(A) Representative confocal images of the typical distribution of FUS WT and R521H and R521G mutants in motor neurons.Nuclei were counterstained with Hoechst.Scale bar ¼ 20 mm.(B) Motor neurons were scored according to nuclear or cytoplasmic predominance of anti-Flag labelling and the presence of cytoplasmic inclusions, then the numbers were expressed as percentage of the total number of cells evaluated.Shown are means + SEM; significantly different from WT, * P , 0.05 (one-way ANOVA; standard weighted-means with Tukey post hoc analysis, n ¼ 6 cultures per condition).(C) A preponderance of FUS inclusions were colabelled by the RNA-specific stain, SYTO.Scale bar ¼ 10 mm.

Figure 5 .
Figure 5. Mitochondrial length in motor neurons expressing FUS WT or mutant FUS.Flag-tagged FUS WT , FUS R521G or FUS R521H was expressed in motor neurons in combination with pOCT-eGFP to visualize mitochondria by epifluorescence microscopy.(A) Three days following intranuclear microinjection of plasmids, mitochondrial length was significantly reduced in axons of neurons expressing either FUS mutant compared with FUS WT .Shown are means + SEM; * significantly different from nuclear WT, P , 0.05 (one-way ANOVA; standard weighted means with Tukey post hoc analysis).FUS WT had no significant effect on mitochondrial length relative to empty plasmid (see Supplementary Material, Fig. S1).(B) Mean mitochondrial length + SEM according to whether FUS was predominantly localized to the nuclear or cytoplasmic compartment, showing reduction in length particularly in neurons with cytoplasmic mutant FUS; * significantly different from nuclear WT, * * significantly different from cytoplasmic WT and from nuclear mutant, P , 0.05.(C) Representative images of mitochondria (pOCT-eGFP epifluorescence) in axons of motor neurons expressing FUS WT , FUS R521G or FUS R521H .Scale bar ¼ 10 mm.

Figure 6 .
Figure 6.PRMT1 knockdown in motor neurons concomitant with the expression of FUS constructs promotes cytoplasmic localization of mutant FUS, but does not alter distribution of FUS WT .shRNA against PRMT1, or scrambled sequence as a control, was co-expressed with plasmid encoding Flag-tagged FUS WT , FUS R521G or FUS R521H by intranuclear microinjection of plasmids.Three days later, cultures were fixed and immunolabelled with anti-Flag; the proportion of cells showing nuclear or cytoplasmic localization of FUS and the presence of inclusions was calculated for motor neurons expressing (A) FUS WT , (B) FUS R521G or (C) FUSR521H .In addition to the increased percentage of neurons with cytoplasmic mutant FUS, the percentage of motor neurons expressing FUS R521H with inclusions was increased significantly.(D) Mitochondrial length is reduced by PRMT1 knockdown.pOCT-eGFP was expressed to visualize mitochondria and mitochondrial length was measured in axons of motor neurons co-expressing FUS WT or mutant FUS with shRNA for PRMT1 (shPRMT1), or scrambled sequence as control.Knockdown of PRMT1 decreased mitochondrial length in neurons expressing FUS WT and accentuated the decrease in neurons expressing FUS mutants.Shown are means + SEM; * P , 0.05 (one-way ANOVA; standard weighted-means with Tukey post hoc analysis, n ¼ 6 -7 cultures per condition).

Figure 7 .
Figure 7. Pretreatment of spinal cord-DRG cultures with Adox to inhibit arginine methylation prevents cytoplasmic localization and inclusion formation in motor neurons expressing wild-type (WT) or mutant FUS.(A) Adox reduced the level of asymmetric arginine methylation of proteins in spinal cord-DRG cultures.Cultures were treated with 0, 20 or 40 mM Adox for 72 h, and then subjected to western analysis with antibodies against asymmetric arginine methylated residues (ASYM24 and ASYM25).Western blots probed with each antibody individually are presented in Supplementary Material, Figure S2.(B-D) Primary spinal cord-DRG cultures were pretreated with 20 mM Adox (or 0 mM as a control) for 48 h; then plasmids encoding Flag-tagged FUS WT , FUS R521G or FUS R521Hwere expressed by intranuclear microinjection.Forty-eight hours after microinjection, cultures were fixed and immunolabelled with anti-Flag.Antibody-labelled neurons were scored according to nuclear or cytoplasmic localization of FUS and for the presence of inclusions, and the numbers presented as percent of the total number of neurons evaluated.Adox pretreatment significantly reduced the percentage of neurons with (B) FUS WT , (C) FUS R521G or (D) FUS R521H predominantly localized in the cytoplasm, and the percentage of neurons expressing (B) FUS WT or (D) FUS R521H with cytoplasmic inclusions.Shown are means + SEM; * P , 0.05, two-tailed, unpaired t-test.