To construct the plasmid for expression of TIA-1 RRM2,3 (93-274), pGEX-4T-2 (GE Healthcare) was modified to bear a 6xHis tag and a thrombin cleavage site. TIA-1 RRM2,3 was inserted directly after the thrombin site using oligonucleotides TIA-1 RRM2,3-For (GCG​TGG​ATC​CCC​AGG​AAT​TCC​CAA​TCA​TTT​CCA​TGT​CTT​TGT​TGG​TG) and TIA-1 RRM2,3-rev (CAG​TCA​CGA​TGC​GGC​CGC​TTA​TTT​GCC​CCA​ATA​GCA​TTT​CAC​AAC). To construct the plasmid TIA-1 RRM2,3 bearing mutations H94A, H96 A and H94A H96A, site-directed mutagenesis by inverse PCR (Liu and Naismith, 2008) was performed using the oligonucleotide pairs:

TIA-1 RRM2,3-H94A-For: CCA​ATg​ctT​TCC​ATG​TCT​TTG​TTG​GTG​ATC​TCA​GCC​CAG

TIA-1 RRM2,3-H94A-Rev: ATG​GAA​agc​ATT​GGG​AAT​TCC​TGG​GGA​TCC​ACG

TIA-1 RRM2,3-H96A-For: TTT​Cgc​tGT​CTT​TGT​TGG​TGA​TCT​CAG​CCC​AG

TIA-1 RRM2,3-H96A-Rev: AAA​GAC​agc​GAA​ATG​ATT​GGG​AAT​TCC​TGG​GGA​TCC​ACG

TIA-1 RRM2,3-H94A H96A-For: CCA​ATg​ctT​TCg​cTG​TCT​TTG​TTG​GTG​ATC​TCA​GCC​CAG

TIA-1 RRM2,3-H94A H96A-Rev: AAA​GAC​agc​GAA​agc​ATT​GGG​AAT​TCC​TGG​GGA​TCC​ACG​CGG

TIA-1 protein short isoform (P31483-2) was expressed from plasmid pETM11 as a 6xHis-tagged protein in BL21 pLysS E. coli cells and purified as previously reported (Loughlin et al., 2021). In brief, this involved initial purification using Co-TALON metal affinity resin (Takara), followed by cleavage of the 6xHis-tag by using TEV protease and its removal using Ni-NTA resin. The final purification was by size exclusion chromatography in a buffer established to maintain TIA-1 solubility (20 mM sodium phosphate, 60 mM KCl, 0.5 M arginine-HCl, 1 mM MgCl₂, 2 mM DTT, 0.5 mM EDTA, pH 7.0). Aliquots were concentrated to a maximum of 100 μM, filtered and stored at −80°C in size exclusion buffer. TIA-1 protein used in Zn²⁺ experiments was prepared in the same manner, with the exception that EDTA was not included in the final size exclusion step. The A_280/260 ratio for each preparation was <0.65 AU and protein was quantified using a theoretical molar extinction coefficient of 80,330 M ⁻¹ cm ⁻¹ _.

GST-TIA-1 RRM1 (amino acids 1-81; in plasmid pGEX-4T-2) and 6xHis-TIA-1 RRM2,3 (amino acids 93-27; in plasmid pETtrx-1a) variants (Waris et al., 2017) were produced in E. coli BL21 (DE3) cells grown at 37°C to an OD₆₀₀ of 0.6–0.8. LB medium was used if proteins were intended for turbidity measurements, whereas ¹⁵N-labelled TIA-1 constructs for NMR titrations were produced in M9 minimal medium with ¹⁵NH₄Cl as a nitrogen source. Expression was induced by the addition of 1 mM IPTG and cultures were incubated at 30°C for 16–18 h. Cell pellets were resuspended in lysis buffer (20 mM potassium phosphate, 500 mM KCl, pH 7.4) with 20 μg/ml DNase I, 1 mM phenylmethanesulfonyl fluoride (PMSF) and 100 μg/ml lysozyme. Cells were disrupted by sonication (cycles of 30 s at 40% of amplitude, 60 s of rest, 6 min total time, on ice), and cell debris was removed by centrifugation at 28,000 × g (4°C for 45 min). The supernatant from GST-TIA-1 RRM1 was loaded onto a 5 ml Profinity GST column (BioRad) and eluted in a single-step with lysis buffer containing 20 mM GSH in a NGC Chromatography System Quest 10 (BioRad). The supernatant from 6xHis-TIA-1 RRM2,3 WT or its mutant variants was loaded onto a Ni-NTA matrix (ThermoFisher), previously equilibrated with lysis buffer containing 5 mM imidazole, and incubated for 1 h at 4°C. Recombinant proteins were eluted using a non-continuous imidazole gradient and purity was further checked by SDS-PAGE. The GST- and 6xHis-tag were removed by overnight incubation (4°C) of the protein with 2.5 U/mg of thrombin protease (Cytiva). Cleaved TIA-1 constructs were then isolated with Ni-NTA resin (for RRM2,3) or a two-step chromatography with ENrich SEC 650 (BioRad) and Profinity GST columns (for RRM1). Purified proteins were dialyzed twice against 5 L of either HEPES buffer (20 mM HEPES, 50 mM NaCl, pH 6.9) or citrate buffer (20 mM citrate, 50 mM NaCl, pH 5.5) at 4°C overnight. For complete disulphide bond reduction, buffered TCEP was added to dialyzed proteins up to a concentration of 1 or 5 mM for turbidity assays and NMR titrations, respectively. Then, samples were concentrated using Amicon Ultra-15 centrifugal filters (Merck-Millipore). Proteins were quantified by spectrophotometry at 280 nm using extinction coefficients of 8,480 M⁻¹cm⁻¹ for TIA-1 RRM1 and 30,940 M⁻¹cm⁻¹ for TIA-1 RRM2,3.

Two constructs of FUS were expressed and purified as per Loughlin et al. (Loughlin et al., 2019). FUS-RBD (amino acids 242-526) comprising FUS RGG1-RRM-RGG2-ZnF-RGG3 domains was produced from pET24b consisting of a N-terminal GB1 solubility tag followed by a 6xHis-tag. FUS-RBD-ΔRGG1/3 (amino acids 269-454) was produced from pET28a consisting of a N-terminal 6xHis tag. Each construct was expressed in E coli BL21 Rosetta2 cells grown at 37°C, followed by protein expression induction with 0.5 mM IPTG and overnight expression at a reduced temperature of 25°C. Cell pellets were resuspended in lysis buffer (50 mM Tris-HCl, 1 M NaCl, 0.5% Triton-X, 5 mM imidazole, 0.5 mM ß-mercaptoethanol, pH 8.0) and lysed by sonication. Lysate was clarified at 48,000 × g for 30 min, loaded onto Ni-NTA beads, washed with 20 and 40 mM imidazole in wash buffer (50 mM Tris-HCl, 1 M NaCl, 0.5 mM ß-mercaptoethanol, pH 8.0) and eluted with 200 mM imidazole in wash buffer. For FUS-RBD short, the 6xHis-tag was cleaved with TEV protease in 50 mM Tris-HCl, 1 M NaCl, 10 mM imidazole, 0.5 mM ß-mercaptoethanol, pH 8.0, and removed with Ni-NTA beads. Proteins were dialysed into storage buffer (50 mM Tris-HCl, 1 M NaCl, 20 mM imidazole, 0.5 mM ß-mercaptoethanol, pH 8), concentrated to 0.4 mM and stored at −80°C. Proteins were quantified using extinction coefficients of 36,883 M⁻¹cm⁻¹ for GB1-6xHis-FUS-RBD fusion and 20,970 M⁻¹cm⁻¹ for FUS-RBD-ΔRGG1/3.

Synthetic single stranded DNA oligonucleotides containing three or five TC-rich TIA-1 binding sites were synthesized and HPLC purified commercially (IDT, Australia) and quantified using molar extinction coefficients supplied by IDT.

TC3 (34 nt): TTT​TTA​CTC​CAA​TTT​TTA​CTC​CAA​TTT​TTA​CTC​C

TC5 (58 nt): TTT​TTA​CTC​CAA​TTT​TTA​CTC​CAA​TTT​TTA​CTC​CAA​TTT​TTA​CTC​CAA​TTT​TTA​CTC​C

Turbidity measurements of full length TIA-1 were used to measure the relative amounts of LLPS under set conditions. Measurements were taken within 20 min of instigating LLPS, a timeframe in which we do not observe aggregation either by microscopy or ThT fluorescence. TIA-1 protein in size exclusion buffer was diluted either alone, or in the presence of nucleic acids in aggregation buffer. In general, TIA-1 protein was diluted to 2.5 μM in the presence of 0.5 μM oligonucleotides unless otherwise stated. Final aggregation buffer was 20 mM HEPES, 50 mM NaCl, pH 7.2, including residual 15 mM arginine-HCl. As controls, nucleic acid or ZnCl₂ alone were diluted in aggregation buffer. Triplicate samples of 150 µL were set-up at room temperature and incubated for 10 min at 25°C in 96-well clear bottom non-binding black plates (Greiner), then analysed at 385 nm using a CLARIOstar plate reader (BMG Labtech). Assays included 3 replicates with error bars representing S.D.

Turbidity measurements of TIA-1 RRM2,3 WT, H94A, H96A and H94A H96A were performed at a final concentration of 20 μM in either HEPES buffer (20 mM HEPES, 50 mM NaCl, pH 6.9) or citrate buffer (20 mM citrate, 50 mM NaCl, pH 5.5). Each construct was measured at 25°C every min for 30 min after the addition of ZnCl₂ or TPEN at the indicated ratios. Independent samples of 100 µL were prepared at room temperature and measured at 385 nm in a 96-well plate with a Varioskan LUX microwell plate reader (ThermoFisher). Graphs represent the average of the last 10 values of turbidity measurements, after signal stabilization. Assays included 3 replicates with error bars representing S.D. Boxplots were generated using R version 4.0.5 (http://www.r-project.org).

Samples of full length TIA-1 with or without ZnCl₂ or oligonucleotides were prepared fresh as per samples used in turbidity measurements and imaged within 20 min of mixing unless otherwise stated. A 10 μL-aliquot of each sample was spotted onto a glass microscope slide with double sided tape and covered with 0.17 mm HP glass coverslip (Zeiss). Solutions were imaged at room temperature with differential interference contrast (DIC) on an Inverted Olympus IX81 x 2UCB microscope (Olympus, Tokyo, Japan) with a 60 × objective. Images were processed using FIJI (Schindelin et al., 2012).

Thioflavin-T (ThT) fluorescence assays were used to monitor ThT positive aggregate formation of full length TIA-1 alone and in the presence of nucleic acids. Samples were prepared as per turbidity assays in aggregation buffer. After turbidity measurements, a 1 mM ThT stock solution was diluted to a final concentration of 5 μM and equilibrated to 30 ^oC for several minutes prior to measurements commencing. Samples were then agitated through orbital shaking at 500 r.p.m. and 30°C over 12–16 h, and ThT fluorescence was monitored with an excitation wavelength of 425 nm and an emission wavelength of 485 nm, using a CLARIOstar plate reader (BMG Labtech). Measurements were baseline corrected using ThT fluorescence of aggregation buffer/ThT alone. Assays included 3 replicates with error bars representing S.D.

To analye the morphology of the aggregates present at the endpoint of the ThT assay, samples were imaged by Transmission Electron Microscopy (TEM). A 5 µL of endpoint sample was applied to the surface of glow-discharged continues carbon grids and stained with 2% uranyl acetate solution. The excess stain was removed using filter paper and grids were air-dried. All samples were imaged on an FEI Tecnai F2 F20 TWIN electron microscope with a Gatan 4k x 4k CCD, under a working voltage of 200 kV at the Ramaciotti Centre for Cryo-Electron Microscopy of Monash University.

Nuclear Magnetic Resonance (NMR) titrations involving TIA-1 constructs were recorded and monitored at 25°C by 1D ¹H and 2D [¹H-¹⁵N] Heteronuclear Single Quantum Correlation (HSQC) spectra in a Bruker Avance-III 700 and 500 MHz equipped with a 5 mm TCI cryoprobe. Samples were dialyzed in 20 mM citrate, 50 mM NaCl, 5 mM TCEP, pH 5.5, or 20 mM HEPES, 50 mM NaCl, 5 mM TCEP, pH 6.9, for NMR titrations. 5% D₂O was added to all samples to adjust the lock signal of the NMR spectrometer. Samples were prepared at 200 or 300 μM for TIA-1 RRM1 and TIA-1 RRM2,3, respectively, in a final volume of 350 μL. Samples were loaded into Shigemi tubes (Shigemi Inc.). Data were acquired and processed using TopSpin 3.5pL7 software (Bruker). Linewidth broadening and chemical-shift perturbation analysis were performed using NMRFAM-SPARKY software distribution (National Magnetic Resonance Facility, Madison). The NMR assignment of TIA-1 RRM1 and TIA-1 RRM2,3 was already available (Biological Magnetic Resonance Bank [BMRB] accession numbers 34144 and 19,735, respectively) (Wang et al., 2014; Sonntag et al., 2017). Chemical-shift perturbations (Δδ_AVG) were calculated as previously described (Rivero-Rodríguez et al., 2021).

Circular Dichroism (CD) spectra were recorded in the far-UV range (195–250 nm) at 20°C on a Jasco J-815 CD spectropolarimeter equipped with a Peltier temperature control system. 10 µM of each TIA-1 RRM2,3 construct in phosphate buffer (10 mM sodium phosphate, 1 mM TCEP, pH 6.9) was measured in a 1 mm quartz cuvette. The final spectra were an average of 20 scans.

To explore the effect of Zn²⁺ on TIA-1 LLPS and potential subsequent aggregation, we first assessed the effect of Zn²⁺ on the LLPS of TIA-1 protein in the absence of nucleic acid. In this study, purified monomeric TIA-1 protein was prepared in the presence of 500 mM arginine, then diluted into a low salt buffer to a concentration just below its saturation (Csat), at which TIA-1 shows minimal LLPS (Loughlin et al., 2021). The effect of ZnCl₂ on the LLPS of TIA-1 protein was analysed by DIC microscopy and quantified using turbidity (absorbance at 385 nm). TIA-1 protein alone at 2.5 μM showed minimal LLPS when imaged by DIC (Figure 1A). Addition of 0.1–50 μM ZnCl₂ resulted in the appearance of spherical droplets that escalated in number on increasing Zn²⁺ concentration, demonstrating ZnCl₂ enhanced LLPS of TIA-1 (Figure 1A). The turbidity of TIA-1 also increased in the presence of ZnCl₂ in a concentration-dependent manner from 0.2 to 0.8, reflecting increased LLPS (Figure 1B). Taken together, these results show that ZnCl₂ alone enhances LLPS of TIA-1, in agreement with previous studies of TIA-1 fusion protein (Rayman et al., 2018).

Our previous work showed that nucleic acid harbouring multiple binding sites enhances LLPS of TIA-1, lowering the concentration at which the protein spontaneously phase separates (Loughlin et al., 2021). To determine whether nucleic acid and Zn²⁺ enhance LLPS of TIA-1 cooperatively or act in a competitive manner, we analysed the effect of ZnCl₂ on ssDNA-induced LLPS of TIA-1. TIA-1 (at 2.5 μM) in the presence of ssDNA harbouring five TIA-1 binding sites (“TC5”) was assessed for LLPS using DIC microscopy and quantified by turbidity. DIC showed the presence of LLPS droplets (∼2–5 μm in diameter) that were slightly larger than those instigated only by ZnCl₂ and gave rise to turbidity measurements of ∼0.3 (Figures 1A,C). Addition of 0.1 μM ZnCl₂ had minimal effect on the turbidity of TIA-1:ssDNA; however, 1–10 μM ZnCl₂ enhanced the number of droplets and increased turbidity to 0.4-0.75, showing that Zn²⁺ and ssDNA together enhance LLPS of TIA-1. Addition of 50 μM ZnCl₂ resulted in a further increase in turbidity to ∼1.0, and DIC showed the presence of a large number of smaller droplets, suggesting that at these concentrations Zn²⁺-induced TIA-1 LLPS dominates the phase separation of these samples. It was also confirmed that ZnCl₂ alone did not give rise to sample turbidity (Figure 1D). These results show that Zn²⁺ further enhances nucleic acid-induced LLPS of TIA-1.

In order to explore the molecular basis for Zn²⁺-induced LLPS of TIA-1, NMR ¹H-¹⁵N-HSQC titration experiments were undertaken for RRM1 (amino acids 1-81) and RRM2,3 (amino acids 93-274) domains. ZnCl₂ was titrated into ¹⁵N-labelled RRM1 or RRM2,3 at 200 and 300 μM, respectively, at pH 6.9. Addition of Zn²⁺ into either sample resulted in some turbidity observed by eye, suggesting a decrease in solubility at these concentrations. The RRM1 amide resonances showed no significant changes in either chemical-shift perturbations or linewidths; therefore, no specific Zn²⁺ binding site was identified (Supplementary Figure S1). In contrast, when Zn²⁺ was added to a sample of RRM2,3, the average broadening was enhanced from 25.5 ± 4.8 Hz (free TIA-1 RRM2,3) to 37.4 ± 18.3 Hz (Zn²⁺:protein ratio of 1:1), indicative of multimerization of TIA-1 (Supplementary Figure S2). Furthermore, our analysis also revealed a subset of amide resonances undergoing specific line broadening (>250 Hz). Interestingly, some of the affected residues—namely N93 (ß ₁), F95 (ß ₁), H96 (ß ₁), V97 (ß ₁), S122 (ß ₂), A124 (ß ₂) and F143 (ß ₃)—are positioned in a cluster adjacent to and including RRM2 ß-sheet residues of the canonical RNA binding site (Figures 2A,B and Supplementary Figure S2A). Moreover, other residues at RRM2 α₂-helix (K146 and D148) and those located at a flexible RRM2 loop and the interdomain linker (M130 and A171, respectively) were also substantially broadened (>250 Hz; Figures 2A,B). Moderate, but still significant, broadening (> mean + 2σ Hz) was also observed for resonances H94 (ß ₁), F98 (ß ₁), V99 (ß ₁), I111 (α₁), R125 (ß ₂), S135 (ß ₃), S142 (ß ₃), A149 (α₂), N151 (α₂) and A152 (α₂) in RRM2 (Figures 2A,B, Supplementary Figure S2). Several residues from the linker and RRM3 also broadened above the threshold, but not to the extent of that observed in RRM2. Remarkably, both substantial line broadening of RRM2 in RRM2,3 and sample turbidity were fully reversible upon addition of the Zn²⁺-chelator TPEN (average linewidth of 26 ± 4 Hz), which confirms that these effects are unequivocally induced by Zn²⁺ (Figure 2A, Supplementary Figures S2E,G). It was also confirmed that TPEN alone had no effect on TIA-1 RRM2,3 resonances (Supplementary Figures S2F,G). Together, these results suggest that Zn²⁺ mediates the multimerization of TIA-1 RRM2,3 domains independently of the PrLD at high protein concentration (300 µM).

From these data, H94 and H96 in RRM2 stand out as potential Zn²⁺ ligands of TIA-1 as histidine residues are often involved in Zn²⁺ coordination (Kocyla et al., 2021). Based on this premise, the NMR titration was repeated under lower pH conditions at which histidine residues could not interact with Zn²⁺ due to protonation (Cruz-Gallardo et al., 2013; Cruz-Gallardo et al., 2015). Titrating TIA-1 RRM2,3 with Zn²⁺ at pH 5.5 resulted in neither enhanced line broadening nor visible turbidity, consistent with the lack of histidine coordination to Zn²⁺ in TIA-1 RRM2,3 (Supplementary Figures S2B–D). It should be noted that Zn²⁺ addition to a solution of TIA-1 RRM1 at pH 5.5 also did not produce any visible turbidity and the linewidth of the signals remained unaltered (Supplementary Figures S1B–D). Thus, histidines in RRM1, that occur at positions 54, 56 and 58, could also transiently coordinate Zn²⁺, but without forming stable complexes. Altogether, NMR titrations identified TIA-1 RRM2 H94 and H96 residues as potential candidates for specific Zn²⁺ binding, possibly mediating the formation of Zn²⁺-induced multimers underlying the observed solution turbidity for TIA-1 RRM2,3.

To test this hypothesis we first quantified the effect of Zn²⁺ titration into 20 µM TIA-1 RRM2,3 on solution turbidity—as measured by absorbance at 385 nm. Increasing the Zn²⁺ enhanced the turbidity of RRM2,3 in a concentration dependent manner (Supplementary Figure S3A). Furthermore, under these conditions, the suppressing effects of TPEN and pH 5.5 in reducing TIA-1 RRM2,3 turbidity were also observed, as previously seen in the more concentrated TIA-1 RRM2,3 NMR solutions (Supplementary Figures S3B,C). Thus, in order to confirm the role of H94 and H96 in Zn²⁺-induced multimerization, TIA-1 RRM2,3 mutants were designed so that one or both histidine residues were substituted by alanine. The correct protein folding was confirmed using CD spectropolarimetry, revealing the same secondary structure content for TIA-1 RRM2,3 wild type and mutants (Supplementary Figure S4). While addition of Zn²⁺ to a ratio of 8:1 increased the turbidity of TIA-1 RRM2,3 (20 µM), minimal or negligible effects were observed for H94A, H96A or H94A H96A mutants (Figure 2C). This finding suggests a structural role for Zn²⁺ in specifically coordinating H94/H96 residues in RRM2, underlying Zn²⁺-induced multimerization of TIA-1 RRMs.

Whilst LLPS of TIA-1 contributes to stress granule formation, aberrant LLPS of disease-associated TIA-1 variants and enhanced propensity to form fibrillar aggregates are associated with impaired stress granule dynamics facilitating pathological inclusions linked to ALS (Mackenzie et al., 2017; Zhang et al., 2019). Thus, we next probed the effect of Zn²⁺ on fibrillar aggregation of TIA-1 over time, in and out of the presence of nucleic acid, as monitored by Thioflavin T (ThT) fluorescence and TEM. At concentrations at which TIA-1 shows minimal LLPS (2.5 μM), no significant TIA-1 aggregation was detected (Supplementary Figure S5A). Addition of 10 μM ZnCl₂, that robustly induces LLPS of TIA-1, did not enhance aggregation of TIA-1 and even appeared to show a slight suppression of ThT fluorescence (Figure 3A). TEM measurements did not detect any fibrillar aggregation (Figure 3B), showing that Zn²⁺-induced LLPS of TIA-1 does not produce fibrillar aggregates.

We then undertook the same experiment in the presence of the TC-rich dsDNA harbouring 5 TIA-1 binding sites (“TC5”), that enhances TIA-1 LLPS and fibril formation (Loughlin et al., 2021). As previously observed, addition of TC5 to TIA-1 (2.5 μM) resulted in fibrillar aggregation as detected by ThT fluorescence over time, and fibrils as observed by TEM (Figures 3C,D). To test the effect of Zn²⁺ on nucleic acid-induced fibrilization of TIA-1, we monitored aggregation of TIA-1 in the presence of TC5 and 10 μM ZnCl₂. No enhancement in ThT fluorescence was observed over time and TEM of final samples did not show numerous fibrils, but small aggregates with potential oligomeric species (Figures 3E,F). Interestingly, we also observed inhibition of ThT fluorescence at substoichiometric concentrations of ZnCl₂ (Supplementary Figure 5B). These results suggest that the presence of Zn²⁺ impedes fibrilization of TIA-1, maintaining it within a reversible LLPS state.

As part of our study of stress granule co-factors, we also analysed the effects of the RGG-rich RNA-binding domain (RBD) of FUS—as a representative RGG-rich protein and potential molecular modulator of TIA-1 LLPS. FUS localises to stress granules via its RGG-rich intrinsically disordered region RGG3 (Bentmann et al., 2012). Moreover, in vitro TIA-1 protein partitions into phase separated FUS protein mediated in part by the RGG rich RBD (Wang et al., 2018). Interestingly, although RGG IDPs can drive homotypic LLPS, unlike PrLD-mediated LLPS, they do not readily mature into irreversible fibrillar aggregates (Gui et al., 2019). Thus, interactions of RGG intrinsically disordered regions may also influence TIA-1 LLPS and potential fibrillar aggregation. We therefore tested the effect of RGG-rich RBD of FUS on TIA-1 LLPS and aggregation, by DIC and turbidity measurements, both alone and in the presence of nucleic acids. FUS-RBD (amino acids 242-526) comprising RGG1-RRM-RGG2-ZnF-RGG3 (Figure 4A) was purified in high salt to homogeneity, free from RNA contamination. TIA-1 and FUS-RBD diluted to 2.5 µM in aggregation buffer showed minimal LLPS in DIC microscopy, with no apparent increase upon their addition (Figure 4B), and only a very small increase in turbidity was observed upon their addition (Figure 4C). Together, this suggests that FUS-RBD does not greatly enhance LLPS of TIA-1 at these concentrations.

Next, we tested the effect of FUS-RBD on TIA-1 LLPS in the presence of a ssDNA comprising three tandem TIA-1 binding sites (“TC3”). The addition of sub-stoichiometric amounts of TC3 to TIA-1 dramatically enhanced LLPS, as measured by increased solution turbidity (Figure 4C) and by droplets observed by DIC that were stable over 2 h (Figure 4D). Addition of equimolar FUS-RBD to TIA-1 and TC3 further enhanced LLPS, resulting in greatly increased turbidity and substantially larger LLPS droplets (Figures 4C,D). Unlike the stable TIA-1:TC3 samples, TIA-1:TC3:FUS-RBD condensates fused together over 2 h. To test whether these changes were due to RGG regions of FUS, a short version of FUS RBD (amino acids 269-454) without RGG1 or RGG3 regions (FUS-RBD-ΔRGG1/3) was mixed with TIA-1:TC3. On addition of FUS-RBD-ΔRGG1/3, no change in turbidity or appearance of LLPS was observed, with droplets appearing relatively stable over a 2 h time period (Figures 4C,D). Although TC3 was designed with TIA-1 binding sites, addition of TC3 ssDNA to FUS-RBD resulted in increased turbidity, with a small increase in the number of LLPS droplets (Figures 4C,E). In contrast, the addition of TC3 to FUS-RBD-ΔRGG1/3 did not increase turbidity or droplet formation (Figures 4C,E). In order to check whether FUS-RBD could be forming promiscuous interactions by displacing TIA-1 from the TC3 ssDNA, we determined its direct binding affinity to a TC DNA sequence in comparison to that of TIA-1 RRM2,3. This showed that the FUS-RBD affinity (K _d = 30 nM) was significantly lower than that of TIA-1 RRM2,3 (K _d = 2 nM) (Supplementary Figure S6) and is unlikely to displace TIA-1 under the conditions of the LLPS assay. These results show that the addition of an RGG-rich protein can modulate the nucleic acid-induced LLPS of TIA-1, potentially via interactions with RRM-bound nucleic acid and/or with TIA-1 PrLD.

Following the observation of changes in LLPS of TIA-1:TC3 condensates, we next examined the fibril formation capacity of TIA-1 in the presence of RGG-rich FUS-RBD as monitored by ThT fluorescence and TEM. Whereas 2.5 µM TIA-1 showed minimal aggregation with only a small increase in ThT fluorescence after ∼8 h, TC3 ssDNA efficiently enhanced ThT fluorescence over time, resulting in amyloid-like fibrils, as observed under TEM (Figures 5A,B). In contrast, the addition of FUS-RBD to TIA-1:TC3 resulted in a dramatic reduction in ThT fluorescence, suggesting a substantial decrease in fibril formation, confirmed by TEM imaging (Figures 5C,D). Interestingly, the small increase in ThT fluorescence of TIA-1 alone is also inhibited in the presence of FUS-RBD (Figure 5C). To assess whether suppression of TIA-1 fibril formation was due to RGG repeats of FUS-RBD, we tested TIA-1:TC3 aggregation in the presence of FUS-RBD-ΔRGG1/3. FUS-RBD-ΔRGG1/3 did not suppress TIA-1:TC3 fibril formation, with a sigmoidal increase in ThT fluorescence occurring and amyloid-like fibrils observed in TEM (Figures 5E,F). Similarly, FUS-RBD-ΔRGG1/3 did not suppress the small increase in ThT fluorescence of TIA-1 alone. Finally, neither FUS-RBD nor FUS-RBD-ΔRGG1/3, alone or in the presence of TC3, showed evidence of fibril formation (Figures 5C,E). These results indicate that RGG intrinsically disordered domains of FUS-RBD prevent the fibril formation of TIA-1 under these conditions.