Nucleation and dissolution mechanism underlying amyotrophic lateral sclerosis/frontotemporal lobar dementia-linked fused in sarcoma condensates

Summary Fused in sarcoma (FUS) is a nuclear RNA-binding protein. Mutations in FUS lead to the mislocalization of FUS from the nucleus to the cytosol and formation of pathogenic aggregates in neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) and frontotemporal lobar dementia (FTLD), yet with unknown molecular mechanisms. Using mutant and stress conditions, we visualized FUS localization and aggregate formation in cells. We used single-molecule pull-down (SiMPull) to quantify the native oligomerization states of wildtype (WT) and mutant FUS in cells. We demonstrate that the NLS mutants exhibited the highest oligomerization (>3) followed by other FUS mutants (>2) and WT FUS which is primarily monomeric. Strikingly, the mutant FUS oligomers are extremely stable and resistant to treatment by high salt, hexanediol, RNase, and Karyopherin-β2 and only soluble in GdnHCl and SDS. We propose that the increased oligomerization units of mutant FUS and their high stability may contribute to ALS/FTLD pathogenesis.


INTRODUCTION
Liquid-liquid phase separation (LLPS) underlies the formation of biomolecular condensates in cells which helps to spatially organize and sequester specific cellular components for efficient processing. [1][2][3][4][5][6] LLPS is governed by multivalent interactions of proteins and nucleic acids which form intra-and intermolecular interactions. [7][8][9][10][11][12][13] Multivalent interactions promote high concentrations of macromolecules and drive rapid transitions from single molecules of proteins and nucleic acids to large oligomeric ribonucleoprotein (RNP) complexes. 14,15 The mechanisms underlying LLPS have been widely studied as aberrant condensate formation is implicated in neurodegenerative diseases. Fused in sarcoma (FUS) is a nuclear RNA-binding protein (RBP) involved in many cellular processes such as transcription, splicing, and mRNA export. 13,[16][17][18] FUS can undergo LLPS by forming transient multivalent interactions with RNA and other RBPs. [19][20][21][22] FUS mutations have been linked to amyotrophic lateral sclerosis (ALS) and frontotemporal lobar dementia (FTLD) in which pathological FUS inclusions cause neuronal degeneration. [23][24][25][26] Aberrant LLPS in ALS/FTLD-linked FUS mutants results in the formation of cytotoxic irreversible aggregates. [27][28][29] Current tools for investigating LLPS measure the formation and material properties of phase-separated condensates and include fluorescence recovery after photo bleaching (FRAP), Forster resonance energy transfer (FRET), fluorescence correlation spectroscopy, optical tweezers, microrheology and turbidity assays. 30,31 While such tools are useful for probing the global properties of condensates, they are blind to molecular-level details, such as the status of oligomeric RNP complexes which may be the building blocks for the larger assemblies of condensates in cells. To address this question, we employed the single-molecule pulldown (SiMPull) assay to measure the oligomerization status of FUS-RNA complexes in neuroblastoma cells (SH-SY5Y). SiMPull enables capturing and quantifying soluble protein complexes directly from cell lysates. Previously, SiMPull has been used to determine the oligomerization of protein complexes including the monomeric vs. dimeric states of mTOR kinase. 32,33 We chose to investigate wildtype (WT) FUS and several ALS/FTLD-linked FUS mutants to test the cellular oligomerization status that may drive the formation of FUS granules and aggregation. While ALS/FTLD-linked FUS mutants have been characterized in vitro at a droplet level 34 and in cellulo and in vivo at a condensate level, 35 FUS mutants have not been studied at a single-molecule level in cells. We asked if cellular condensates consist of smaller units of soluble oligomers of proteins. In conjunction, we sought to assess if the oligomerization pattern is correlated with the propensity of condensate formation with and without stress. Using the SiMPull platform, we also interrogated the pulled-down complexes by applying a diverse set of dissolving reagents to examine what molecular interactions are required for sustaining the oligomers.
We demonstrate that while FUS wildtype exists primarily as monomers, ALS/FTLD-linked FUS mutants display higher oligomers. Although osmotic stress induced by sorbitol drives mislocalization and puncta formation of FUS in cells, the basal oligomers remain unchanged for both wildtype and FUS mutants, suggesting that the oligomers are inherent basal units of FUS in cells. Strikingly, the oligomers remain intact even when treated with harsh chemical conditions such as high salt, hexanediol, and treatment by RNase A and Karyopherin-b2, indicating highly stable FUS complexes. Together, our findings suggest that the inherently stable and higher order complexes of FUS mutants may contribute to pathogenic aggregation.

Validation of wildtype and mutant fused in sarcoma-GFP constructs
We prepared green fluorescent protein (GFP)-tagged FUS constructs to assess the subcellular localization and FUS granule formation for WT and mutant FUS under normal and stressed cellular conditions. GFP tags were inserted at either the N or the C terminus of FUS ( Figure 1A). Insertion of a GFP tag at the N terminus of FUS resulted in proper nuclear localization, consistent with endogenous FUS localization in the nucleus iScience Article probed by immunofluorescence ( Figure 1B top row, c). Conversely, the insertion of a GFP tag at the C terminus of FUS resulted in the mislocalization of FUS to the cytosol and appearance of FUS puncta in the cytosol ( Figure 1B, bottom row). This effect is likely due to the GFP tag interfering with the C-terminal NLS which is required for nuclear transport by Karyopherin-b2. Others have shown that the insertion of GFP at the C terminus of FUS does not disrupt localization when a long linker separates the NLS and the GFP tag. 25 To avoid the mislocalization and aggregation induced by the C-terminal GFP tag, WT, and all mutant FUS constructs were designed with an N-terminal GFP tag for subsequent studies. A GFP-only construct was also used as a control for all experiments.
Overexpression of FUS upon the transfection of FUS-GFP constructs was a potential concern because high concentration increases the condensation and aggregation propensity of FUS. To minimize overexpression, we applied the lowest possible plasmid concentration that still allows for fluorescence imaging (0.1 mg cDNA plasmid per transfection). We assessed endogenous and exogenous FUS protein levels via western blot ( Figure 1D). We observed consistent endogenous FUS expression among cells transfected with WT and mutant FUS-GFP and the GFP only construct ( Figure 1E). Importantly, exogenous and endogenous FUS expression was consistent among WT and all mutant FUS constructs ( Figure S1), indicating a similarity in stability and expression of exogenous WT and mutant FUS constructs. When comparing endogenous and exogenous FUS protein levels, overexpression via FUS-GFP transfection resulted in a 2-fold increase of total cellular FUS protein ( Figure 1F). Nevertheless, exogenous WT FUS-GFP did not cause a mislocalization of FUS nor increased aggregate formation typically seen with overexpression of FUS 23,36,37 ( Figure 1C).
The selected ALS/FLTD-linked FUS mutants were chosen to represent three classes of FUS mutants: R (R216C, R244C), G (G156E, G187S), and NLS (R495X, and P525L) mutants. R mutants of FUS exhibit defective binding to RNA leading to increased FUS aggregation, which may be reversible upon interaction with WT FUS. 34,35 G mutants of FUS result in the formation of homotypic gel-like condensates and do not interact with WT FUS. 35 Lastly, NLS mutants of FUS result in mislocalization from the nucleus to the cytoplasm. 29,38 NLS mutants are the most deleterious as they result in a loss of FUS function in the nucleus coupled with a gain of toxicity in the cytoplasm. This gain of toxicity results from the FUS aggregation which is promoted by low RNA concentration in the cytoplasm. 12 High concentrations of RNA in the nucleus have been described to help maintain the solubility of RNA-binding proteins (RBPs) such as FUS and prevent condensate formation. 12 Therefore, a reduction in RNA concentration in cellular compartments such as the cytoplasm would promote condensate formation and could lead to aggregation.
Localization and aggregation induced by amyotrophic lateral sclerosis/frontotemporal lobar dementia mutation and sorbitol stress Cellular localization of WT and mutant FUS-GFP was determined via immunostaining. Cells were fixed and stained with nuclear dye while GFP signal from the GFP-only plasmid or FUS-GFP constructs were used to determine localization and aggregation patterns. The GFP-only plasmid control displays diffuse green signal both in the nucleus and cytoplasm as GFP can shuttle freely between both compartments ( Figure 2A, top row). Under normal conditions, WT FUS displays a diffuse nuclear localization as expected ( Figure 2A, second row). R and G FUS mutants display a nuclear localization, but exhibit increased punctate pattern in contrast to WT FUS, suggesting an increased propensity for condensation ( Figure 2A, third-sixth rows). As expected, both NLS mutants are mislocalized to the cytoplasm due to defective NLS (Figure 2A, last two rows). The NLS mutants also display a punctate pattern in both nuclear and cytoplasmic compartments. Although the punctate pattern in the cytoplasm could be explained by the lower RNA concentration, the nuclear puncta suggest increased condensation propensity for the NLS mutants in addition to disrupted localization. The FUS granule sizes are comparable in all conditions, but the NLS mutants significantly increase the number of cells with FUS puncta ( Figure 2C, top) in agreement with the higher pathogenicity of the NLS mutants. Overall, R and G mutations maintain proper localization but have increased aggregation while NLS mutants exhibit both improper localization and aggregation.
We next sought to examine the effect of stress on FUS localization and punctate formation. Osmotic stress has been known to cause prominent mislocalization of WT FUS and incorporation into stress granules. 39,40 We asked if our FUS mutants showed the same pattern following osmotic stress. iScience Article 20% WT and 30-40% R and G mutant FUS from the nucleus to the cytosol ( Figure 2A, right column). NLS mutants, which already display a $50% mislocalization under normal conditions, remained mislocalized to a similar degree ( Figure 2B). As a control, our GFP-only construct displayed no change in localization pattern upon osmotic stress treatment. Osmotic stress also increased the granule size and cells with puncta in both WT and mutant FUS conditions ( Figure 2C, bottom). We later correlate this puncta size to the cellular oligomerization status of FUS.
Several studies have detailed the effect of oxidative stress via sodium arsenite treatment in causing cytoplasmic FUS incorporation into stress granules under mutant FUS conditions that already perturb FUS localization such as NLS mutants. 41 Temperature has also been described as a critical regulator of phase separation. 42,43 Therefore, we induced oxidative stress and temperature variations to cells expressing WT and NLS mutant FUS-GFP to assess the potential shift in localization or punctate pattern. Sodium arsenite treatment (1 mM for 1 h) to induce oxidative stress did not cause a robust mislocalization for either WT or mutant FUS ( Figures S2A and S2D), but induced a slight increase in a punctate pattern, similar to the osmotic stress (Figures 2A and 3A). Next, we assessed the heat (42 C for 2 h) and cold (  iScience Article Amyotrophic lateral sclerosis/frontotemporal lobar dementia -linked fused in sarcoma mutants induce higher order oligomerization We next sought to measure the oligomeric status of cellular FUS by using the SiMPull assay. We hypothesized that FUS mutants and/or stress conditions would result in larger and more stable oligomers based on the increased condensation propensity observed at the cellular level ( Figure 2) SiMPull enables efficient and robust isolation of native FUS-GFP protein complexes from cells to a single-molecule surface. Briefly, we transfect SH-SY5Y neuroblastoma cells with WT or mutant FUS-GFP constructs for $24 h and gently lyse the cells with NP40 cell lysis buffer to maintain FUS complexes (see STAR Methods). Cell lysates containing FUS-GFP protein are flowed onto a pre-assembled single-molecule surface coated with anti-GFP antibody ( Figure 3A). GFP signals do not appear unless the surface is treated with the anti-GFP antibody, signifying a specific capture of GFP complexes ( Figure 3B). The surface is illuminated with a laser beam (488 nm) and the GFP photobleaching steps ( Figure 3C) are counted at each single-molecule spot to determine the number of FUS molecules in each complex. To preserve the native protein complexes, we minimized the processing time from lysing cells to setting up the imaging surface to acquiring images within 10 min. The GFP-only control showed successful capture of $70% monomers and $30% dimers ( Figure 3D), consistent with previous reports 33 showing a fraction of GFP forms self-dimers. The GFP oligomer pattern was used as a background to subtract dimerization states for subsequent photobleaching analysis i.e 30% of dimers in all conditions is subtracted due to GFP self-dimerization ( Figure 3E). First, we compared the oligomerization status of WT to mutant FUS. WT FUS displays predominantly monomeric and dimeric states with an average photobleaching step of $1.5 after subtracting the GFP self-dimer fraction as stated above ( Figures 3E and 4A). R and G FUS mutants display higher order oligomerization states with increased dimeric or trimeric steps with average photobleaching steps of $1.7-2.4 ( Figures 4B-4E and 4H). Strikingly, NLS FUS mutants R495X and P525L ( Figures 4F and 4G) display significantly higher oligomerization states including 4-6 steps with an average photobleaching step of $3 ( Figure 4H). These findings reveal inherent differences in FUS complexes even at a basal level. Therefore, interactions seeding these small units of FUS likely differ among WT and mutant FUS.

Cellular stress does not alter the oligomerization status
The increased oligomerization observed for FUS mutants aligns with their increased propensity to form puncta at the cellular level, suggesting a potential link between oligomerization and granule formation ( Figure 5I, left). To test this link, we tested if osmotic stress by sorbitol would increase the oligomers of FUS and FUS mutants since it significantly increased the puncta formation in cells (Figure 2A). Surprisingly, while WT FUS and P525L displayed a moderate increase toward higher-order oligomers with an increase in average photobleaching steps by $0.5 ( Figures 5A and 5G), all the other FUS mutants showed similar levels of oligomers before and after the sorbitol stress although wildtype and G156E exhibited slightly increased photobleaching steps in the stressed condition ( Figures 5B-5F). This result does not align with the drastic increase of FUS puncta formation under sorbitol stress shown in Figure 2 ( Figure 5H). In addition, arsenite stress which induced slight aggregation of nuclear FUS did not increase photobleaching steps ( Figure S3). Therefore, cellular condensation or aggregation may not be correlated with the oligomeric state of the protein or protein complexes ( Figure 5I, right). Instead, oligomers that we measure by SiMPull are reflective of more fundamental aberrant interactions that occur at the molecular level for FUS mutants. Therefore, our result indicates that although the interaction between FUS monomers in FUS oligomer differs between wildtype and FUS mutants, the interaction between FUS oligomers induced by stress is likely weak, resulting in a facile dissolution into intrinsic oligomer units upon being pulled down.

Cytosolic mislocalization alone does not cause higher fused in sarcoma oligomerization
We next sought to understand if cytosolic mislocalization caused the higher oligomerization of the two NLS mutants, R495X and P252L. We took advantage of the fact that C-terminally GFP-tagged FUS results in iScience Article improper cytosolic mislocalization and coacervation ( Figure 1B, bottom row). We created constructs for WT, G156E, and R244C FUS with a C-terminal GFP tag to induce mislocalization even under non-stressed conditions and performed SiMPull on these constructs ( Figure 6A, left panel). C-term constructs did not result in an increase in oligomerization relative to N-term constructs ( Figures 6B and 6C). Osmotic stress only slightly increased mislocalization ( Figure 6A, right panel, 6D) since FUS was already mislocalized due to the C-term GFP tag. We note that the R244C displayed less mislocalization than the WT possibly because the aggregated state of R244C in the nucleus prevented mislocalization into the cytoplasm. In addition, stress did not increase the oligomerization of C-term FUS constructs ( Figures 6E-6H). Therefore, cytosolic localization itself is not sufficient to promote higher order oligomerization and the increased oligomerization pattern of FUS mutants, particularly NLS mutants, is potentially due to other complex interactions.   12,34,35 interactions are thought to underlie the formation of ribonucleoprotein (RNP) granules. To better understand the interactions maintaining WT FUS or driving enhanced oligomerization of mutant FUS, we systematically perturbed these multivalent interactions with dissolving agents and compared the number of photobleaching steps before and after dissolving agents. We expect that effective dissolving agents will disrupt basal FUS-GFP complexes, resulting in primarily monomeric states of WT and mutant FUS.
As a control, a strong protein denaturant (8 M guanidine hydrochloride (GdnHCl) and protein denaturant (0.1% sodium dodecyl sulfate (SDS)) was applied to dissolve FUS-GFP complexes and perturb oligomeric interactions. GdnHCl treatment resulted in a markedly increased monomeric population of WT and mutant FUS following a 5 min incubation ( Figure 7A). SDS resulted in a mild increase in the monomeric population of WT and mutant FUS ( Figure 7B). We next applied 10% 1,6-hexanediol ( Figure 7C), 500 mM NaCl (Figure 7D), and 1 mg/mL RNase A ( Figure 7E) to probe the importance of hydrophobic, electrostatic, and RNA-dependent interactions, respectively. Strikingly, none of these treatments significantly disrupted the oligomers of WT or mutant FUS ( Figure S4). Thus, these basal-level FUS oligomers are highly stable nano clusters that are not maintained by the same weak, multivalent interactions characteristic of condensates. Even WT FUS oligomers did not decrease or trend toward a predominantly monomeric pool of FUS, highlighting the presence of stable basal FUS oligomers that are resistant to dissolution.
We next assessed whether Karyopherin beta 2 (Kapb2) could dissolve FUS oligomers. KapB2 is a known nuclear importin of FUS which binds the NLS of FUS and helps to promote solubility by preventing iScience Article interactions with FUS or other RBPs. [48][49][50][51] Kapb2 also prevents phase separation and incorporation of FUS into condensates by importing FUS back into the nucleus where RNA can buffer phase separation. 49, 51 Mutations in FUS that prevent FUS-Kapb2 interactions result in enhanced FUS aggregation and pathological inclusions. 29,38 Because Kapb2 binds the NLS of FUS, mutations in this region are especially prone to aggregation. Thus, we expected Kapb2 may be able to bind and help disaggregate WT and other mutant FUS oligomers, although we expected less of an effect for NLS mutants, R495X and P525L. Surprisingly, we found Kap b2 did not solubilize WT or mutant FUS at 10-1000 nM Kapb2 and roughly equimolar or higher isolated FUS-GFP on the single-molecule surface, ( Figure 7F). Kapb2 was ineffective in dissolving WT or mutant FUS oligomers, which may be due to the FUS-FUS or FUS-RBP interactions that mask the NLS domain. In addition, Kapb2 combined with RNase A ( Figure 7G) to perturb FUS-FUS, FUS-RBP, and FUS-RNA interactions was unable to dissolve oligomers.
Overall, dissolution experiments revealed that our isolated FUS-GFP oligomers captured from cells in their native contexts were highly stable and resistant to dissolution except in strong protein denaturation conditions.

DISCUSSION
Mislocalization and aggregation of many RBPs such as FUS and TDP43 are major hallmarks of degenerating neurons. 52 The same pattern was recapitulated in cell-based studies in which disease-linked mutant proteins or stress conditions were applied. While mislocalization has been induced in NLS mutants or under severe stress or aged conditions, mislocalization has not previously been correlated with oligomerization pattern. Condensation or aggregation in aged or severed ALS/FTLD-linked FUS mutants has also been observed in vitro and in cells but has not been correlated with oligomerization. Other studies also fail to capture basal RNP complexes in a cellular context and focus either on full condensates in cells or RNP complexes with minimal RNA and protein components using purified protein in vitro.
SiMPull enables the capture of native FUS complexes from cells within minutes using minimal cell lysate. This promotes the preservation of multivalent interactions maintaining small FUS complexes. With SiMPull, we were able to detect the oligomerization patterns of WT and ALS/FTLD-linked FUS. We confirm that FUS WT exists primarily as monomers in cells. Strikingly, we observed an increase in FUS oligomerization states with NLS mutants R495X and P525L. Such increased oligomerization seen in NLS mutants is not solely due to the mislocalization since the C-terminally GFP-tagged FUS WT which is also mislocalized does not exhibit increased oligomerization. FUS mutants G187S, R216C, and R244C also showed increases in oligomerization states compared to WT FUS. The increased oligomerization under sorbitol stress we report here is likely underestimated because a considerable amount of FUS remained soluble i.e as monomers or dimers in the nucleus which was mixed with the higher oligomers of cytoplasmic condensates for the experiment.
FUS complexes captured by SiMPull reveal the oligomerization state of WT and mutant FUS. Although we cannot directly relate the oligomerization to cellular toxicity, we show that the oligomer state of FUS mutants is highly correlated with the cellular features often associated with neurodegenerative diseases including the level of FUS granule formation, granule size, and mislocalization ( Figure S5). However, we note that these complexes may contain other components such as proteins and RNA that maintain the stability of oligomers. In the future, we aim to discern the components of isolated complexes by tagging other proteins or RNAs with different fluorophores. This will further enable us to visualize the mixing of WT and mutant FUS with each other and with other molecules, providing insight into the composition and resulting properties of such oligomers.
Basal FUS oligomers rapidly isolated from cells and preserved in their native contexts gives us an indication of how these units differ in oligomer states and which interactions may underlie the formation of these iScience Article oligomers. Differences between WT and mutant FUS oligomers may indicate potential issues with ALS/ FTLD-linked FUS mutants whose multimerization at a single-molecule level is already perturbed and underlies the formation of pathological FUS aggregates. This suggests that SiMPull enables the isolation of basal FUS oligomeric units from their native states. Thus, SiMPull captures FUS oligomers as the basic building units of FUS condensates and enables us to further study the interactions that promote FUS oligomerization at a single-molecule level. While we are not pulling down FUS condensates, we are able to isolate native, soluble FUS oligomers in their basal states. There already exists a difference between WT and ALS/FTLD-linked FUS mutants in their oligomeric states under normal conditions. Interestingly, stress does not increase FUS oligomerization. Therefore, stress potentially acts on FUS condensates by promoting FUS aggregation via multivalent interactions among basal oligomers that lead to the formation of pathological aggregates over time.

Limitations of the study
The FUS oligomers that we pulldown from cell lysate and apply to SiMPull measurement may not capture the full extent of cellular oligomers of FUS as we may lose some loosely bound proteins during the process i.e obtaining cell lysate and making dilutions required for the SiMPull experiment. The ectopic overexpression of FUS-GFP may impact the oligomerization status of FUS although the same condition applies to all measurements (WT, FUS mutants, stress and so forth) equally.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Immunocytochemistry
Cells were grown on glass bottom chambers prior to immunostaining. Cells were fixed $24 hrs after transfection in formaldehyde diluted to 4% in 1X PBS for 10 min at RT. Cells were permeabilized and blocked in buffer (0.3% Triton TM X-100 and 1% BSA in 1X PBS) for 10 min. Cells were incubated with primary antibodies in buffer overnight at 4 C. Secondary antibodies combined with Hoechst dye were diluted in buffer and incubated for 1 hr at RT. Washing steps after antibody incubation were performed with 1X PBS. Cells were maintained in 1X PBS at 4 C after staining and imaged the same or next day via confocal microscopy.

Confocal microscopy
Confocal microscopy was performed at the Integrated Imaging Center at the Johns Hopkins Homewood Campus with a Zeiss LSM 700 microscope equipped with lasers for 405, 488, 568, and 647 nm excitation. Images were acquired using four-fold frame averaging with a 40 3 1.4 oil objective. The same laser and acquisition settings were maintained for imaging all samples within one experiment.

Single molecule pulldown
For single molecule pulldown, a methoxy polyethylene glycol (PEG) coated microscope slide and coverslip were prepared following a previously described method. 53 $2% biotinylated PEG was added during slide preparation to allow for specific immobilization of biotinylated molecules to the surface. Holes were drilled on opposite ends of the slide to allow for rapid flow of cell lysate and exchange of solution. Flow chambers were constructed using double sided tape to seal the coverslip to the slide, and edges of the slide and coverslip were secured with epoxy. $20 mL of solution is enough to saturate each chamber. Once prepared, 30 mL neutravidin was flowed into the chamber at a 20 mg/mL dilution and incubated for $2 min 50 mL of T50 buffer (10 mM Tris-Cl pH 8.0, 50 mM NaCl) was flowed through to wash off unbound neutravidin, then 50 mL biotin conjugated anti-GFP was flowed onto the surface at a 1:200 dilution for $2 min followed by another 50 mL T50 buffer wash. Cells transfected for $24 hrs were lysed with NP40 cell lysis buffer supplemented with cOmplete TM EDTA-free protease inhibitor cocktail for 1 min. Cell lysates containing soluble protein were then diluted 1:100-1:10,000 in T50 buffer and 50 mL of diluted cell lysate was flowed onto the single molecule surface to obtain a surface density of $100-300 molecules in the 2,500 mm 2 imaging area. A 50 mL ll OPEN ACCESS iScience 26, 106537, April 21, 2023 iScience Article T50 buffer wash was performed to remove unbound protein. A prism type total internal reflection fluorescence microscope equipped with an electron-multiplying charge-coupled device was used for single molecule imaging. GFP-tagged proteins were excited at 488 nm. All single molecule experiments were performed at room temperature. Single molecule fluorescence time traces of surface immobilized GFPtagged proteins were manually quantified for number of photobleaching steps. Cell lysates were diluted accordingly to obtain $100-300 molecules on the single molecule imaging surface to avoid oversaturation of GFP molecules.

Dissolution experiments
Using the same single molecule pulldown protocol described above, cell lysates were isolated for imaging on the single molecule surface. Single molecule imaging was performed to obtain photobleaching steps for GFP-only, WT, and mutant FUS complexes. Then, 50 mL dissolving agent was applied to the chamber and incubated for $5 min. Two washes of 50 mL T50 buffer was used to remove dissolving agent and any unbound, dissolved GFP-only or FUS-GFP protein. Single molecule imaging was performed again to obtain photobleaching steps following dissolving agent addition.

Image analysis
Confocal images were acquired in ZEN (Zeiss). Images were processed using ImageJ/FIJI software. Artificial colors were applied for individual channels as depicted in figures.

Quantification of FUS localization in cells
FUS subcelluar localization quantification was performed in ImageJ. A boundary was drawn around total GFP signal within each cell. Nuclear GFP signal was then subtracted to obtain cytosolic GFP signal. Nuclear and cytosolic GFP signal were divided by total cellular GFP signal to obtain ratios.

Quantification of FUS granule area
FUS granule area quantification was performed in ImageJ. A boundary was drawn around each individual granule to obtain the area.

Quantification of photobleaching steps
Photobleaching steps were analyzed via a MatLab script Trace_Viewer_Categorizer which enables stepwise traces to be visualized. Photobleaching steps were quantified manually and counted if a stepwise decrease in fluorescent signal intensity was observed.