Phase separation of the oncogenic fusion protein EWS::FLI1 is modulated by its DNA-binding domain

Significance

The oncogenic fusion protein, EWS::FLI1, responsible for more than 85% of Ewing sarcoma tumors combines the transactivation domain from EWS and the DNA-binding domain (DBD) from Friend leukemia integration 1 (FLI1). The fusion impacts the function of wild-type EWS and drives oncogenesis via aberrant transcriptional and splicing changes as well as defects in the DNA-damage response. Both components of the fusion are required for oncogenesis suggesting a synergistic function between the domains. Here, the authors describe the structural underpinnings of EWS::FLI1’s effect on EWS, mediated by the FLI1 DBD, that enhances the phase separation propensity of EWS and drives aberrant changes to the physical properties of condensates.

Abstract

Ewing sarcoma (EwS) is an aggressive cancer of bone and soft tissue that predominantly affects children and young adults. A chromosomal translocation joins the low-complexity domain (LCD) of the RNA-binding protein EWS (EWS^LCD) with the DNA-binding domain of Friend leukemia integration 1 (FLI1^DBD), creating EWS::FLI1, a potent fusion oncoprotein essential for EwS development and responsible for over 85% of EwS tumors. EWS::FLI1 forms biomolecular condensates in vivo and promotes tumorigenesis through mediation of aberrant transcriptional changes and by interfering with the normal functions of nucleic acid-binding proteins like EWS through a dominant-negative mechanism. In particular, the expression of EWS::FLI1 in EwS directly interferes with the biological functions of EWS leading to alternate splicing events and defects in DNA-damage repair pathways. Though the EWS^LCD is capable of phase separation, here we report a direct interaction between FLI1^DBD and EWS^LCD that enhances condensate formation and alters the physical properties of the condensate. This effect was conserved for three related E-twenty-six transformation-specific (ETS) DNA-binding domains (DBDs) while DNA binding blocked the interaction with EWS^LCD and inhibited EWS::FLI1 condensate formation. NMR spectroscopy and mutagenesis studies confirmed that ETS DBDs transiently interact with EWS^LCD via the ETS DBDs “wings.” Together these results revealed that ETS DBDs, particularly FLI1^DBD, enhance EWS^LCD condensate formation and rigidity, supporting a model in which electrostatic and structural interactions drive condensate dynamics with implications for EWS::FLI1-mediated transcriptional regulation in EwS.

The oncogenic EWS::FLI1 fusion protein is the hallmark of a group of fusion proteins associated with Ewing sarcoma (EwS), a highly aggressive pediatric bone and soft tissue cancer (1). This fusion arises from a chromosomal translocation that joins the N-terminal low-complexity domain (LCD) of the RNA-binding protein EWS (EWS^LCD) in-frame with the DNA-binding domain (DBD) of the E-twenty-six transformation-specific (ETS) family transcription factor Friend leukemia integration 1 (FLI1^DBD). The resulting EWS::FLI1 fusion is responsible for approximately 85% of EwS tumors (1). Of the 28 ETS transcription factors, FLI1, transcriptional regulator ERG (ERG), and protein FEV (FEV) are the most common EWS^LCD fusion partners identified in EwS; however, the reasons for this frequency remain unclear (2). More broadly, in the family of EwS-related primitive neuroectodermal tumors, the FET family of RNA-binding proteins, including fused-in-sarcoma (FUS), EWS, and TATA-binding protein associated factor 2 N (TAF15), are commonly found fused with the DBDs of ETS-family and other transcription factors (3).

The EWS::FLI1 fusion disrupts normal genetic programs by aberrantly binding to GGAA microsatellites and FLI1 consensus sites via the FLI1 DBD. Concurrently, the EWS^LCD facilitates the recruitment of chromatin remodelers, epigenetic modifiers, and transcriptional machinery (4–8). Specifically, the EWS^LCD plays a key role in DNA enhancer binding, and both domains are essential for recruiting ATP-dependent BRG1/BRM associated factor (BAF) chromatin remodeling complex and driving transcriptional activation (4, 9–14). However, these functions alone do not fully explain EWS::FLI1’s oncogenic role. Growing evidence suggests that it also exerts a dominant negative effect on the normal roles of EWS in transcriptional regulation and splicing (15–18). Notably, recent findings indicate that the presence of EWS::FLI1 at transcriptionally active sites interferes with the release of Breast Cancer Type 1 Susceptibility Protein (BRCA1) from DNA-directed RNA polymerase II subunit RPB1 (RNA Pol II). This disruption leads to increased transcriptional stress, homologous recombination deficiency, and the accumulation of unresolved R-loops (15).

Biomolecular condensation, in which proteins and nucleic acids phase separate from the aqueous environment to form membraneless organelles, has emerged as a key mechanism of subcellular organization (19–21). Condensates play essential roles in various cellular processes, including transcription (22, 23), splicing (24), DNA damage repair (25), and the stress response (26). The EWS^LCD is intrinsically disordered, lacking stable secondary and tertiary structure, and is characterized by a repetitive, degenerate low complexity SYGQP-rich sequence, which promotes self-association and condensate formation (5, 27–29). In cells, FET oncogenic fusions, including EWS::FLI1, form dynamic yet specific assemblies with other intrinsically disordered proteins/regions including EWS itself (28, 30), the C-terminal domain (CTD) of RNA Pol II (16, 31, 32), Bromodomain-containing protein 4, and components of the BAF complex (4) among others. Due to the physical features of the EWS^LCD, these interactions are multivalent, dynamic, and transient, making high-resolution structural analysis of the complexes challenging (29, 33). Additionally, EWS::FLI1 undergoes phase separation and is thought to require self-association to stabilize its DNA binding and promote transcriptional activation. However, the exact nature of these interactions remains unresolved (4, 5, 28).

The oncogenic function of EWS::FLI1 requires features derived from both the EWS^LCD and the FLI1^DBD (4). To investigate their interplay in self-association and condensate formation, we used biophysical approaches including turbidity and thioflavin T (ThT) fluorescence assays, microscopy, and NMR to assess how EWS::FLI1 and the isolated FLI1^DBD affect EWS^LCD condensates. We investigated the rigidification of EWS^LCD condensates by FLI1 when present as either a separate domain or in the oncogenic fusion, EWS::FLI1. We then sought to determine whether direct interactions between ETS DBDs and EWS^LCD are responsible for the altered condensate properties. This study provides mechanistic insight into EWS::FLI1’s dominant negative regulation of wild-type EWS and supports the Goldilocks hypothesis of EWS::FLI1 toxicity (34). Furthermore, these studies reveal disease-associated regulatory mechanisms of condensate dynamics, uncovering the ETS DBDs as potent modulators of condensate liquidity through the possible promotion of cross-β structure formation.

Results

EWS::FLI1 Forms Rigidified Condensates.

We generated phase diagrams for full-length EWS and EWS::FLI1 (Fig. 1A and SI Appendix, Fig. S1) using turbidity measurements. Consistent with previous findings for FUS, EWS condensate formation was inhibited at high NaCl concentrations, suggesting that electrostatic interactions between N- and C-terminal residues play a role (35, 36) (Fig. 1B). In contrast, high NaCl concentrations promoted condensate formation of EWS::FLI1 (Fig. 1B), suggesting that the hydrophobic interactions that drive EWS^LCD condensate formation also significantly contribute in EWS::FLI1 (33, 37, 38). In vivo, EWS and EWS::FLI1 function in multicomponent biomolecular condensates formed by multivalent LCD–LCD interactions with each other as well as with other RNA/DNA binding proteins with similar LCDs, such as the BAF complex components (6, 29, 30). To determine whether multicomponent biomolecular condensates comprising EWS wild-type and EWS::FLI1 have similar material properties, we prepared condensates by mixing 200 µM EWS^LCD (EWS 1 to 264, SI Appendix, Fig. S1), which is common to both EWS and EWS::FLI1, with full-length EWS, EWS::FLI1, or EWS^LCD (Fig. 1C). Full-length EWS and EWS::FLI1 readily form dual-component condensates with EWS^LCD (Fig. 1G), like in vivo observations in the EwS cell line A673 (39). We assessed condensate dynamics by monitoring fluorescence recovery after photobleaching (FRAP) of the fluorescently labeled molecules within the condensates (Fig. 1E). Condensates containing full-length EWS + EWS^LCD showed rapid FRAP, while condensates containing full-length EWS::FLI1 + EWS^LCD recovered more slowly (Fig. 1E), indicating that condensates containing EWS::FLI1 are less liquid-like and more rigid. To determine whether this observation is recapitulated in the complex cellular environment, we used a system that fuses the 52-heptad (YSPTSPS) repeat CTD of RNA Pol II (RNAPII^CTD52) to the lac repressor (LacI) and a fluorescent reporter (CFP) (40) (Fig. 1D). Expression of this CFP-LacI-RNAPII^CTD52 construct in U2OS cells modified with an array of lac operons (~50,000 copies) stably inserted into chromosome 1 creates a dense foci of RNAPII^CTD52, akin to intracellular condensates, capable of recruiting biomolecular constituents including EWS and EWS::FLI1 (Fig. 1 D and H). These cellular puncta recover fluorescence rapidly when bleached and overexpression of EWS did not reduce their liquidity (Fig. 1F). Similar to our in vitro experiments, overexpression and recruitment of EWS::FLI1 to the RNAPII^CTD52 foci (Fig. 1H) resulted in reduced FRAP (Fig. 1F), suggesting the condensate rigidification by EWS::FLI1 also affects multicomponent condensates.

Fig. 1.

Biomolecular condensation of EWS and EWS::FLI1. (A) Cartoons of EWS, FLI1, and EWS::FLI1 domains, the EWS, and FLI1 breakpoints are marked with black triangles. (B) Phase diagrams for EWS and EWS::FLI1 measured via turbidity. Experimental design for (C) in vitro and (D) in cellulo FRAP. FRAP (E and F) and micrographs (G and H) of condensates containing (E and G) 200 µM EWS^LCD plus 9 µM EWS⁴⁸⁸ (Left), 5 µM EWS::FLI1⁴⁸⁸ (Center), or 25 µM EWS^LCD,488 (Right), or (F and H) CFP-LacI-RNAPII^CTD52 foci in U2OS cells with wild-type EWS⁵⁴⁹ (Left), EWS::FLI1⁵⁴⁹ (Center), or only CFP-LacI-RNAPII^CTD52 (control, Right). Superscripts 488 or 549 indicate fluorescent labeling, (Scale bars, 25 µm.) FRAP data are means of triplicates ± SEM.

EWS::FLI1 and FLI1^DBD Alter Condensate Formation of EWS^LCD.

Next we tested the effect of EWS::FLI1 and FLI1^DBD on the phase separation propensity of EWS^LCD using turbidity measurements and microscopy (Fig. 2A). In physiological buffer conditions (100 mM NaCl), 25 μM EWS^LCD will form condensates (Fig. 2A). The addition of equimolar concentrations of FLI1^DBD (25 µM) significantly increased the turbidity of the sample and dramatically altered the morphology of the condensates which showed reduced coalescence and became fusion-impaired over the 5-h equilibration (Fig. 2A, cf. Movies S1 and S2). The addition of EWS::FLI1 to EWS^LCD condensates had a larger effect, significantly increasing the turbidity of the sample and bright-field microscopy revealed structures that appeared more aggregate-like than liquid-like droplets (Fig. 2A). At lower ratios of EWS::FLI1 to EWS^LCD, structures that resemble condensates can still be observed (Fig. 2B) indicating that the alteration of EWS^LCD condensate morphology is dependent on the concentration of EWS::FLI1. Turbidity measurements for EWS^LCD, EWS^LCD + FLI1^DBD, and EWS^LCD + EWS::FLI1 stabilize after ~2 h, with 5-h aged condensates maintaining the trends in turbidity seen in freshly prepared samples (SI Appendix, Fig. S2A). At 25 μM, full-length EWS::FLI1 alone formed rigid puncta rather than liquid-like droplets (Fig. 2A), providing further evidence that the FLI1^DBD has an adverse effect on the liquidity of the EWS^LCD condensates. To determine whether this effect was specific to the FLI1^DBD, we also tested equimolar concentrations of constructs including the RNA-recognition motif (RRM) of EWS (EWS^RRM or EWS^RRM-RGG2, 25 µM) (Fig. 2A and Table 1 and SI Appendix, Fig. S1). EWS^RRM had no significant effect on the turbidity of EWS^LCD, while EWS^RRM-RGG2 induced a slight increase in turbidity, and both EWS^RRM and EWS^RRM-RGG2 resulted in EWS^LCD condensates that appeared more spherical and retained robust coalescence (Fig. 2A and Movies S3 and S4). Therefore, the FLI1^DBD found in EWS::FLI1 increases the phase separation propensity of EWS^LCD and changes the material properties of the condensates causing them to rapidly become fusion impaired, which may drive the different effect on the morphology of the condensates compared to the RNA-binding domains that are naturally found in EWS.

Fig. 2.

FLI1^DBD alters the phase separation propensity of EWS^LCD. (A) Turbidity and micrographs of 25 µM EWS^LCD plus 25 µM of either FLI1^DBD, EWS^RRM, EWS^RRM-RGG2, or EWS::FLI1, and controls (no EWS^LCD) measured in phase separating conditions (100 mM NaCl) after 5 h equilibration. EWS::FLI1 alone is shown for comparison. ThT assay and micrographs of (B) 50 µM EWS^LCD, 5 µM EWS::FLI1 alone, and 50 µM EWS^LCD plus 0.1, 1, or 5 µM EWS::FLI1, and (C) 50 µM EWS^LCD, 5 µM FLI1^DBD (alone), and 50 µM EWS^LCD plus 1 or 5 µM FLI1^DBD in phase-separating conditions (150 mM NaCl). For (B and C) Images were taken at T > 10 h (Scale bars, 25 µm). ThT (turbidity) data are means (averages) of triplicates ± SEM. P-values: *P < 0.05; **P < 0.01; ****P < 0.0001.

Table 1.

Recombinant protein constructs of EWS, EWS::FLI1, and ETS DBDs

Construct name	Affinity Tag	MW (kDa)	Margins*
EWS (full-length)	8× His	68.5	EWS 1–656
FUS (full-length)	His-MBP	53.4	FUS 1–526
EWS::FLI1 (type 2, full-length)	8× His or 8× His-GFP	54.3	EWS 1–264 + FLI1 22 –452 (F408A)†
EWSLCD	8× His	27.9	EWS 2–264
EWSRRM	8× His	9.9	EWS 359–447
EWSRRM-RGG2	8× His	16.3	EWS 359–513
EWSRGG3	10× His-MBP	8.73	EWS 549–638
FLI1DBD	8× His	14.6	FLI1 276–399 (F362A)†
FLI1DBD C299S S390C	8× His	14.6	FLI1 276–399 (C299S, F362A, S390C)†
FLI1DBD R2L2	8× His	14.4	FLI1 276–399 (R337L, R340L, F362A)†
FLI1DBD Δα4	8× His	10.3	FLI1 276–361
ERGDBD	8× His	14.5	ERG 306–429 (F392A)†
ETV1DBD	8× His	15.1	ETV1 332–458
PU.1DBD	8× His	12.3	PU.1 168–270
PU.1DBD, ETV1 swap	8× His	12.2	PU.1 168–270 (K204I, H205H, K206P, G281del, K221P, K222A, T224N, K245A, K246G, insertE247, K247R, L248Y)
PU.1DBD, no +ve charge	8× His	12.1	PU.1 168–270 (K204I, H205H, K206P, G281del, R220N, K221P, K222A, T224N, K245A, K246G, insertE247, K247R, L248Y)

^*Canonical sequences from Uniprot.

^†Mutations EWS::FLI1 (F408A), FLI1 (F362A), and ERG (F392A) are equivalent (SI Appendix, Fig. S6).

Biomolecular condensates formed by LCD-containing proteins similar in composition to EWS^LCD age spontaneously over time, becoming less liquid-like (41, 42). Cross-β structure has been shown to stabilize the condensed state of FUS (43), and condensate rigidification is concurrent with the formation of fibrillar structures that initially form at the interface between condensed and dilute phases and sometimes protrude from the condensates (41, 42). To investigate whether the changes in condensate morphology induced by EWS::FLI1 and FLI1^DBD are due to the formation of cross-β structure within EWS^LCD condensates, we developed ThT fluorescence assays. Under phase separating conditions (150 mM NaCl), EWS^LCD condensates rigidify slowly over approximately 10 h, with a small concomitant increase in ThT fluorescence (Fig. 2 B and C). The addition of EWS::FLI1 or FLI1^DBD, even at substoichiometric concentrations (as low as 1:500 molar ratio), to EWS^LCD condensates significantly accelerated condensate rigidification in a manner dependent on the concentration of EWS::FLI1 or FLI1^DBD (Fig. 2 B and C). FLI1^DBD also increased ThT fluorescence of full-length EWS under phase separating conditions (SI Appendix, Fig. S2B). ThT-positive structures are not formed by FLI1^DBD in the absence of EWS^LCD at the low concentrations used in this assay (Fig. 2C). Under these conditions, we did not observe fibril formation by transmission electron microscopy (SI Appendix, Fig. S2C), indicating that the observed increase in ThT fluorescence arises from the formation of cross-β structure that has not had time, or is incapable of organizing into bona fide amyloid fibrils. This initial formation of protofibrillary species may be occurring at the interface between the condensed and dilute phase, as observed for FUS, which may account for the rigidification of condensates and the impaired ability of the condensates to fuse (42). The FLI1^DBD is not known to form cross-β structure, whereas FET family proteins have been shown to form cross-β structures (27, 41, 42, 44, 45); therefore, it is logical to conclude that the FLI1^DBD drives cross-β structure formation by EWS^LCD but does not bind ThT itself. Together, these results indicate that the FLI1^DBD found in EWS::FLI1 increases the rate at which EWS^LCD condensates rigidify, possibly caused via the formation of ThT-positive cross-β structure.

DNA Binding to FLI1^DBD Inhibits EWS^LCD Condensate Rigidification.

The ability of EWS::FLI1 and FLI1^DBD to rigidify EWS^LCD condensates suggests that the proteins directly interact. To determine whether the interaction site involves the FLI1 DNA-binding site, we conducted turbidity assays in the presence of DNA (Fig. 3A). At 5 µM, EWS::FLI1 undergoes rapid condensate formation and/or aggregation and the samples become turbid immediately (Fig. 3A). The addition of equimolar concentrations of double-stranded, high-affinity (HA) DNA (46) reduced the turbidity of the sample, indicating that condensate formation and/or aggregation was inhibited (Fig. 3A and Table 2). A double-stranded DNA oligonucleotide containing 10 GGAA (GGAA₁₀) repeats known to bind FLI1^DBD in vitro (14, 47), exerted the same effect on EWS::FLI1 aggregation (SI Appendix, Fig. S3A). ThT assays demonstrated that an equimolar amount of HA DNA reduced EWS::FLI1-induced rigidification of EWS^LCD condensates (Fig. 3B).

Fig. 3.

DNA binding by EWS::FLI1 or FLI1^DBD inhibits phase separation and condensate rigidification. (A) Turbidity of 5 µM EWS::FLI1 (filled) or plus 5 µM HA DNA (open), and HA DNA after 5 h equilibration. (B) ThT assay of 50 µM EWS^LCD, plus 1 µM of EWS::FLI1 (filled), or EWS::FLI1 + HA DNA (open). (C) Turbidity of 25 µM EWS^LCD, plus 25 µM of FLI1^DBD (filled), or FLI1^DBD + HA DNA (open), and HA DNA, or FLI1^DBD + HA DNA after 5 h equilibration. (D) ThT assay of 50 µM EWS^LCD plus 5 µM of FLI1^DBD (filled), or FLI1^DBD + 3 (open) or 5 (light open) µM HA DNA. Fluorescence and DIC microscopy of freshly prepared samples of (E) EWS^LCD + EWS::FLI1⁴⁸⁸ (Top) and EWS^LCD + EWS::FLI1⁴⁸⁸ + HA DNA⁶⁵⁰ (Bottom), and (F) EWS^LCD,650 + FLI1^DBD,488 (Top) or EWS^LCD,488 + FLI1^DBD + HA DNA⁶⁵⁰ (Bottom). Scale bars, 10 µm, superscripts 488 or 650 indicate fluorescent dye labeling. Fluorescence intensity ratio (I_in/I_out) and FRAP of (G and H) EWS::FLI1⁴⁸⁸ in condensates composed of EWS^LCD + EWS::FLI1⁴⁸⁸ ± HA DNA⁶⁵⁰, or (I and J) of FLI1^DBD,488 in condensates composed of EWS^LCD + FLI1^DBD,488 + increasing amounts of HA DNA⁶⁵⁰. Experiments were conducted under phase separating conditions (100 or 150 mM NaCl), turbidity, ThT, FRAP (A–C, D, H, and J), or intensity (G and H) data are shown as means of triplicates or 16 replicates ± SEM. P-values: **P < 0.01; ****P < 0.0001.

Table 2.

DNA oligonucleotides used for binding studies with ETS DBDs

Construct name	Sequence
HA DNA	TTTACCGGAAGTGTTT
10 x GGAA DNA	CGCGGATCCGGAAGGAAGGAAGGAAGGAAGGAAGGAAGGAAGGAAGGAACTGCAGTTTT
Scrambled DNA	TTGAGAGAGAGAGATT

Consistent with the findings for EWS::FLI1, the addition of HA DNA inhibited the ability of FLI1^DBD to enhance EWS^LCD condensate formation as evidenced by a reduction in turbidity relative to EWS^LCD + FLI1^DBD in the absence of DNA (Fig. 3C). Additionally, DNA binding inhibited the ability of FLI1^DBD to drive EWS^LCD condensate rigidification in a concentration-dependent manner (Fig. 3D). At the highest concentrations of HA DNA tested (50 µM corresponding to a 10:1 DNA:FLI1^DBD ratio), the effect of FLI1^DBD on EWS^LCD condensates was almost entirely abolished (SI Appendix, Fig. S3B). The GGAA₁₀ oligonucleotide also inhibited the effect FLI1^DBD exerts on EWS^LCD condensates (SI Appendix, Fig. S3C). A scrambled dsDNA control sequence (Table 2) had no inhibitory effect, demonstrating that the inhibitory effect of DNA in these ThT assays is due to DNA binding by FLI1 (SI Appendix, Fig. S3C). Collectively, the ThT assays and turbidity data demonstrate that DNA binding by EWS::FLI1 and FLI1^DBD inhibits condensate/aggregate formation of EWS::FLI1 and reduces the interaction between FLI1^DBD and EWS^LCD.

We then used fluorescence microscopy to characterize the colocalization of the protein and DNA components comprising the condensates. To quantify the colocalization of the fluorescently labeled components to the EWS^LCD, we calculated fluorescence intensity ratios inside versus outside (I_in/I_out) of the condensates. Fluorescent EWS::FLI1⁴⁸⁸ (labeled with DyLight 488) was mixed with EWS^LCD without (Fig. 3 E, Top) or with (Fig. 3 E, Bottom) HA DNA⁶⁵⁰ (labeled with Cy5) under phase separating conditions. Fluorescence corresponding to EWS::FLI1⁴⁸⁸ was localized in the condensates, indicating that EWS::FLI1 interacts with EWS^LCD in the condensates. Surprisingly, HA DNA was largely excluded from EWS^LCD + EWS::FLI1 condensates with an intensity ratio of 0.5 ± 0.1 (Fig. 3 E, Bottom). Similarly, we mixed the EWS^LCD,650 (EWS^LCD labeled with DyLight 650) with FLI1^DBD,488 under phase separating conditions and the FLI1^DBD,488 signal spatially overlapped with the EWS^LCD,650 condensates, indicating partitioning of FLI1^DBD into EWS^LCD condensates (Fig. 3 F, Top). Consistent with the behavior of multicomponent (EWS^LCD + EWS::FLI1 + HA DNA) condensates, mixing EWS^LCD,488 condensates with unlabeled FLI1^DBD and HA DNA⁶⁵⁰ revealed that HA DNA was moderately excluded from the condensates (Fig. 3 F, Bottom) with an intensity ratio of 0.8 ± 0.02. The partitioning of the FLI1^DBD to EWS^LCD condensates is specific and further reinforces that these domains interact since unrelated proteins such as green fluorescent protein (GFP) do not colocalize to EWS^LCD condensates unless fused to EWS^LCD (SI Appendix, Fig. S4A). Additionally, FLI1^DBD,488 does not form visible condensates under these conditions (SI Appendix, Fig. S4B).

The fluorescence intensity ratio for EWS::FLI1⁴⁸⁸ colocalization to EWS^LCD condensates in the absence of DNA was 5.4 ± 0.9, indicating that EWS::FLI1 is enriched in the condensates (Fig. 3G). Addition of HA DNA resulted in reduced colocalization of EWS::FLI1 to EWS^LCD condensates with a measured intensity ratio of 2.5 ± 0.1 (Fig. 3G). FRAP of EWS::FLI1⁴⁸⁸ within EWS^LCD condensates reached ~46% of the steady-state intensity 30 s after photobleaching, while in the presence of HA DNA, fluorescence recovery reached ~63% after the same interval, indicating that the addition of DNA increases the liquidity of EWS::FLI1⁴⁸⁸-containing condensates (Fig. 3H). This observation may be due to a reduction of the concentration of EWS::FLI1 in the condensates due to DNA binding (Fig. 3G), thereby reducing its adverse effect on condensate liquidity.

Analogous to the observations for EWS::FLI1, fluorescence microscopy revealed that in the absence of DNA, the I_in/I_out ratio of FLI1^DBD,488 was ~5.8 ± 2.0 (Fig. 3I). As the concentration of HA DNA was increased, the intensity of FLI1^DBD,488 in condensates became noticeably reduced (SI Appendix, Fig. S5A), and the I_in/I_out intensity ratio reduced to 1.6 ± 0.2 for EWS^LCD,650 with FLI1^DBD,488 and 50 μM HA DNA (Fig. 3I), consistent with the effect observed for EWS::FLI1 (cf. Fig. 3G). Therefore, DNA binding by EWS::FLI1 or FLI1^DBD outcompetes the interactions between EWS^LCD and FLI1^DBD that drive condensate colocalization, with DNA possibly acting as a sink for EWS::FLI1 and FLI1^DBD, precluding them from entering the condensate. The EWS^LCD,650 within condensates containing only EWS^LCD underwent rapid recovery after photobleaching, reaching ~91% of the initial fluorescence, indicating liquid-like condensate behavior (Fig. 3J). In contrast, freshly prepared condensates formed in the presence of FLI1^DBD displayed slower fluorescence recovery, reaching only ~40% recovery indicating that the dynamics of EWS^LCD molecules within the condensates change in the presence of FLI1^DBD (Fig. 3J and SI Appendix, Fig. S5B). With the addition of HA DNA to EWS^LCD + FLI1^DBD condensates, fluorescence recovered more rapidly in a DNA concentration-dependent manner, consistent with DNA binding inhibiting the effect of FLI1^DBD on EWS^LCD (Fig. 3J).

Exclusion of DNA from EWS^LCD condensates may be due to a lack of charge neutralization of the DNA phosphate backbone within the condensates since the EWS^LCD contains only 2 positively charged residues (out of 264) in contrast with 6 negatively charged residues and 27 tyrosines with delocalized π electrons in their sidechain rings. Pelleting assays revealed that although condensates are readily visible via microscopy, the majority (>95%) of the EWS^LCD remains in the dilute phase (supernatant) in the absence of the FLI1^DBD (SI Appendix, Fig. S5C). However, in the presence of the FLI1^DBD, the partitioning of EWS^LCD into the pellet fraction was noticeably higher (~50% of total EWS^LCD), consistent with the FLI1^DBD enhancing EWS^LCD condensate formation (Fig. 2A and SI Appendix, Fig. S5C). The FLI1^DBD also partitioned into the pellet fraction (~55% of total FLI1^DBD), indicating direct association with the EWS^LCD (SI Appendix, Fig. S5C). Addition of HA DNA and FLI1^DBD reduced the partitioning of both EWS^LCD and FLI1^DBD (~90% of each protein remains in the dilute phase) into the pellet further indicating that DNA inhibits the EWS^LCD–FLI1^DBD interaction (SI Appendix, Fig. S5C). The HA DNA alone had no effect on EWS^LCD condensate formation and the FLI1^DBD incubated alone remained in the dilute phase (SI Appendix, Fig. S5C).

The FLI1 DNA-Binding and Dimer Interfaces Are Not Involved in the Interaction with EWS^LCD.

The FLI1^DBD DNA recognition helix (α3) harbors two highly conserved arginine residues that contact core bases of the ETS consensus sequence in all the reported crystal structures of ETS DBDs complexed with DNA (48–52) (Fig. 4A and SI Appendix, Fig. S6A). To test whether these arginines participate in Arg-Tyr π-cation interactions with EWS^LCD as a possible mechanism that promotes rapid condensate rigidification, we mutated both arginines to leucines (FLI1^DBD,R2L2, SI Appendix, Fig. S1). This mutant is incapable of binding HA DNA (SI Appendix, Fig. S7A) and its effect on EWS^LCD condensates was assessed using ThT assays (Fig. 4B). Surprisingly, the FLI1^DBD,R2L2 mutant retained the same ability to enhance the rate of EWS^LCD condensate rigidification in the presence or absence of DNA (Fig. 4B). Together these results indicate that the positively charged arginines located on the DNA recognition helix of FLI1 are not involved in the EWS^LCD–FLI1^DBD interaction. Furthermore, in line with how HA DNA affects the ThT-positive response of EWS^LCD + FLI1^DBD condensates, the inability of FLI1^DBD,R2L2 to bind DNA correlated with its failure to inhibit the increase in ThT positivity within EWS^LCD condensates, even when DNA was present.

Fig. 4.

FLI1^DBD helices α3 and α4 do not contribute to the rigidification of EWS^LCD condensates. (A) FLI1^DBD structure highlighting the DNA recognition helix (α3), conserved R337 and R340 that contact DNA (SI Appendix, Fig. S6), and C-terminal helix α4 (5e8g). ThT assay of 50 µM EWS^LCD alone (black), and plus 5 µM of either FLI1^DBD (red), or (B) FLI1^DBD + HA DNA (red open), FLI1^DBD,R2L2 (cyan), or FLI1^DBD,R2L2 + HA DNA (cyan open), or (C) FLI1^DBD,Δα4 mutant (blue). ThT assays were conducted under phase separating conditions (150 mM NaCl), with data shown as means of triplicates ± SEM.

The structures of ETS DBDs solved to date share a common winged helix–turn–helix fold (SI Appendix, Fig. S6B). However, the reported boundaries of the ETS domains vary. Some studies have used an 85 amino acid construct containing only three α-helices (17, 53–55), while others have employed a longer construct that includes a fourth α-helix (9, 17, 49). A recent crystallographic study showed the FLI1^DBD forming a dimer, with the fourth α-helix participating in the dimer interface (49). In this study, we used a FLI1^DBD construct that includes the fourth α-helix (residues 362 to 369), but to prevent dimerization we introduced the F408A (EWS::FLI1 sequence numbering) mutation and included ~30 C-terminal disordered residues (SI Appendix, Figs. S1 and S6A). To assess whether the fourth α-helix, which contains exposed hydrophobic residues, interacts with EWS^LCD, we performed ThT assays using FLI1^DBD,Δα4, a FLI1^DBD construct lacking residues beyond 361 (Fig. 4C). As expected, this truncated construct retained DNA binding activity, as confirmed by electrophoretic mobility shift assays (SI Appendix, Fig. S7B). Moreover, in ThT assays, it affected EWS^LCD condensates similarly to the long construct (Fig. 4C). These findings define the EWS^LCD interaction site responsible for inducing EWS^LCD rigidification within condensates, narrowing it to the folded core of FLI1 (residues 276-361) while excluding the arginines of the DNA recognition helix (α3, Fig. 4A).

Enhancement of EWS^LCD Condensate Rigidification Is a General Property of ETS DBDs.

Beyond EWS::FLI1, other EWS–ETS fusions cause EwS. To determine whether FLI1^DBD condensate rigidification activity is linked to the ETS DBD fold, we performed ThT assays with three additional ETS DBDs and three constructs containing segments of EWS’s C-terminal RNA-binding domain (EWS^RRM, EWS^RRM-RGG2, EWS^RGG3, Fig. 5 and SI Appendix, Fig. S1). Two of these ETS DBDs (ERG, and ETS translocation variant 1, ETV1) are also found in oncogenic fusions with EWS^LCD, while the third, transcription factor PU.1 (PU.1) is oncogenic but not known to fuse with EWS in EwS. All constructs were properly folded, as judged from circular dichroism spectra (SI Appendix, Fig. S8). Bright-field microscopy images were captured at the end of the ThT assay (T > 10 h) (Fig. 5). Compared with EWS^LCD alone, EWS^LCD + EWS^RRM or EWS^RRM-RGG2 had minimal impact on the rate of ThT fluorescence increase, with condensates remaining well-dispersed and predominantly spherical (Fig. 5 A, C, and E). In contrast, all four ETS DBDs significantly enhanced EWS^LCD condensate rigidification (Fig. 5 B, D, F, and H), leading to irregular morphologies, particularly in the presence of PU.1^DBD and ETV1^DBD (Fig. 5 B and H). In the absence of EWS^LCD condensates none of the constructs induced a ThT signal (SI Appendix, Fig. S9A). Notably, PU.1^DBD induced the fastest ThT fluorescence increase, disrupting condensate coalescence and fusion even more severely than FLI1^DBD (Movie S5). Surprisingly, EWS^RGG3 also increased ThT positivity in a concentration-dependent manner (Fig. 5G and SI Appendix, Fig. S9B). However, unlike ETS DBD-induced condensates, these did not develop the same irregular shapes. Instead, they rapidly coalesced and spread out across the chamber bottom, resembling surface wetting (Fig. 5G and Movie S6).

Fig. 5.

Rigidification effect on EWS^LCD condensates is conserved for other ETS DBDs. ThT assays (Left) and micrographs of aged samples (T ~ 24 h, Right) of (A) 50 µM EWS^LCD incubated alone or with (B) 5 µM of PU.1^DBD, (C) EWS^RRM, (D) FLI1^DBD, (E) EWS^RRM-RGG2, (F) ERG^DBD, (G) EWS^RGG3, and (H) ETV1^DBD. Samples were prepared under phase separating conditions (150 mM NaCl). Dashed lines estimate half maximum signal. Images were taken at T > 10 h (Scale bars, 50 µm). ThT data are means of triplicates ± SEM.

The addition of HA DNA with ERG^DBD and PU.1^DBD to EWS^LCD condensates inhibited rigidification, similar to FLI1^DBD, suggesting a structurally conserved mechanism (SI Appendix, Fig. S9C). To rule out His-tag effects we tested a His-tag free version of PU.1^DBD (SI Appendix, Fig. S9D). Since His-tagged EWS^RRM and EWS^RRM-RGG2 did not induce condensate rigidification, and the His-tag free PU.1^DBD behaved identically to its tagged counterpart, the His-tag was not a contributing factor (SI Appendix, Fig. S9D). These results indicate that the impact of FLI1^DBD on EWS^LCD condensates is conserved across ETS transcription factors and is linked to the fold of the winged helix–turn–helix motif.

ETS DBDs Interact with the EWS^LCD via Residues Adjacent to the DNA-Binding Face.

We used solution NMR to map the interaction of ETS DBDs with the EWS^LCD. Given the transient nature of the interaction and the challenges of studying an aggregation-prone system, we selected the PU.1^DBD for these experiments. Backbone resonances of PU.1 were assigned using standard approaches (56). ¹H,¹⁵N heteronuclear single quantum coherence spectra were recorded for a sample containing an ~3:1 molar ratio of EWS^LCD to ¹⁵N PU.1^DBD (SI Appendix, Fig. S10). Addition of higher concentrations of EWS^LCD were not possible because PU.1^DBD induced EWS^LCD condensate formation even in low salt conditions. Nevertheless, we observed small chemical shift perturbations (CSP) (δΔ ~ 0.04 ppm) for PU.1 resonances (Fig. 6A and SI Appendix, Fig. S10). The signal intensity uniformly decreased between a sample of ¹⁵N PU.1 and the 3:1 sample likely due to the formation of complexes between PU.1^DBD and EWS^LCD that are beyond the size-limits for detection with solution NMR (Fig. 6B). However, a few peaks were differentially broadened, and coincided with or were located near residues with CSPs (Fig. 6 A and B). We plotted residues with CSPs greater than one SD from the mean and residues with differential signal intensities less than one SD from the mean onto the AlphaFold (57) structure of human PU.1 (Fig. 6C). Notably, these residues clustered to one face of the DBD, including residues in the ETS DBD “wings” (loops 3, 4, and 6) that contact the phosphate backbone of DNA (50); however, no shifts or differential broadening were observed for the DNA recognition helix (α3), consistent with the effects observed with the FLI1^DBD,R2L2 mutant (Fig. 4B). Furthermore, we observed no shifts or peak broadening on the opposite face of the DBD (Fig. 6C). The CSPs and differential broadening of residues in the disordered C-terminal tail were considered nonspecific to condensate rigidification. This is because the C-terminally truncated ETS domain (FLI1^{DBD, Δα4}) which lacks these residues, still retained condensate rigidification activity (Fig. 4C and SI Appendix, Fig. S1) and this region is poorly conserved among the ETS DBDs (SI Appendix, Fig. S6A).

Fig. 6.

Residues in the DBD wings of PU.1 are involved in the interaction with EWS^LCD. (A) CSP and (B) signal broadening of ¹⁵N-PU.1^DBD residues at a 3:1 molar ratio with EWS^LCD, SD marked with red dashed lines. Black (red) asterisks indicate overlapped/ambiguously assigned (proline) residues. Residues with CSPs (intensity differences) greater (less) than 1 SD are orange filled, and (C) mapped to the AlphaFold model of the human PU.1^DBD. (D) Sequence alignment and net charge of ETS^DBD loops 3, 4, and 6, conserved positive charges are bolded (SI Appendix, Fig. S6). (E) ThT assay and (F) micrographs of 50 µM EWS^LCD alone (black), with 5 µM of PU.1^DBD (blue), PU.1^{DBD, ETV1 swap} (ETV sequence in loops, green open circles), or PU.1^{DBD, no +ve charge} (all charged residues in loops are mutated to neutral, teal). Images were taken at T > 10 h (Scale bars, 50 µm). ThT data are means of triplicates ± SEM.

A sequence comparison of loops 3, 4, and 6 from the four ETS DBDs tested in the ThT assays (Fig. 5) revealed that the total net charge varies between +8 to 0 (PU.1 > FLI1 = ERG > ETV1, Fig. 6D). The loops in PU.1 are dynamic (SI Appendix, Fig. S11) and enriched in lysine residues, giving them a net charge of +8. Notably, loop 3 is the only loop containing positively charged residues and lacks any negatively charged residues, resulting in a net charge of +2 (Fig. 6D). Loops 3, 4, and 6 of FLI1 and ERG are identical with a net charge of +2, and in ETV1 they are depleted in Lys with a net charge of 0. Loops 4 and 6 in all the ETS DBDs tested contain one highly conserved, positively charged residue (Arg or Lys) at 220 and 247 (PU.1 sequence numbering, Fig. 6D). ThT assays showed that the PU.1^DBD rigidified condensates the fastest, followed by FLI1^DBD and ERG^DBD, and finally ETV1^DBD which had the slowest effect (Fig. 5 B, D, F, and H respectively). The net charge of loops 3, 4, and 6 directly correlates with the rate of ETS DBD-induced EWS^LCD condensate rigidification, suggesting that electrostatic interactions play a role. To test this, we constructed two PU.1^DBD mutants: one in which loops 3, 4, and 6 were replaced with those from ETV1 (PU.1^{DBD,ETV1 swap}) and another where all positive charges were removed from loops 3, 4, and 6 (PU.1^{DBD,no +ve charge}) (SI Appendix, Fig. S1). The PU.1^{DBD,ETV1 swap} mutant rigidified EWS^LCD condensates at a slower rate than wild-type PU.1^DBD and had a less severe effect on condensate morphology (Fig. 6 E and F). Subsequently, the PU.1^{DBD,no +ve charge} mutant further reduced the rate of ThT positivity and the condensates appeared even more similar to EWS^LCD-only condensates, although the difference in the rate of ThT-positivity between the two mutants appears minimal (Fig. 6 E and F). These findings support a loop-mediated condensate rigidification mechanism involving specific interactions with the EWS^LCD. However, the PU.1^{DBD,ETV swap} mutant still rigidified condensates faster than ETV1^DBD, suggesting loops 3, 4, and 6 do not completely account for the rate differences observed between PU.1^DBD and ETV1^DBD. Electrostatic interactions alone do not determine condensate rigidity. The FLI1^DBD,R2L2 mutant rigidified EWS^LCD condensates at the same rate as FLI1^DBD (Fig. 4B), and although EWS^RRM-RGG2 is rich in positively charged Arg residues, it did not rigidify EWS^LCD condensates like the ETS DBDs (Fig. 5). Similarly, EWS^RGG3, which has a net charge of +10, also induced rigidification but to a lesser extent than the ETS DBDs (Fig. 5 and SI Appendix, Fig. S9B).

To further examine the role of positive charge in the condensate rigidification, we tested 30- and 50-residue poly-L-lysine peptides for their effect on ThT positivity and condensate formation. At the same molar ratio (1:10) where ETS DBDs induced a rapid ThT response, neither peptide produced strong ThT positivity (SI Appendix, Fig. S9E). Poly-L-lysine 30 slightly impaired condensate coalescence, with some droplets failing to merge (SI Appendix, Fig. S9E and Movie S7), whereas poly-L-lysine 50 maintained robust coalescence (SI Appendix, Fig. S9E and Movie S8). These findings suggest that positive charge alone does not fully explain the rigidifying effect of ETS DBDs on EWS^LCD condensates. Instead, the structural arrangement of the three loops and their ability to adopt conformations favorable for EWS^LCD may contribute to the liquid-to-solid transition.

Discussion

EWS has roles in transcriptional regulation, homologous recombination, DNA damage response, and splicing, though its precise functions remain unclear. Notably, knockout of EWS is postnatally lethal in mice (15, 58–60). The expression of EWS::FLI1 mimics the effects of EWS knockdown in HeLa and EwS cell, supporting the hypothesis that it acts as a dominant-negative regulator of EWS function (15, 17, 18, 61). This was recently confirmed in EwS cells, where EWS::FLI1 disrupts EWS’s regulation of RNA Pol II phosphorylation by CKD7/9 (15). We demonstrated that the FLI1^DBD in EWS::FLI1 enhances EWS^LCD-mediated condensate formation, partitions into EWS^LCD condensates, and accelerates their rigidification. This effect was conserved across three other ETS DBDs, suggesting a shared mechanism common to ETS DBDs. Boulay et al. recently reported that β-isoxazole induced precipitation of EWS::FLI1 in EwS cell lines was stronger than that of EWS (4), further supporting the hypothesis that the FLI1^DBD interaction increases the intrinsic phase separation propensity of the EWS^LCD. Studies in prostate cancer cell lines identified intermolecular interactions between EWS and ERG, ETV1, ETV4, and ETV5 that were necessary and sufficient for oncogenesis, reinforcing the idea that FET–ETS interactions can have oncogenic consequences (38). These findings suggest that EWS is recruited to ETS-bound DNA enhancers through regions flanking the ETS DBD, highlighting the need for further investigation into EWS–ETS interactions (38). Furthermore, coimmunoprecipitation experiments demonstrated a direct interaction between PU.1 and FUS that inhibits the normal regulatory functions of FUS in splicing (62, 63). Therefore, mislocalization of ETS domains either as fusions or due to aberrant regulation appears to have a dual effect of enhancing self-association and negatively impacting the functions of FET-family proteins and possibly other nucleic acid-binding proteins. We found that the interaction between ETS DBDs and EWS^LCD involves residues in the “wings” (loops 3, 4, and 6) that are partially occluded by the DNA phosphate backbone in the DNA-bound state. As a result, DNA binding reduces the interactions between EWS::FLI1 or ETS DBDs and EWS^LCD that promote condensation (Fig. 7 A and B). The in vivo situation is more complicated since other proteins, posttranslational modifications, and nucleic acids surely contribute to and modulate condensate properties. We confirmed that EWS::FLI1 disrupts condensate liquidity in a complex cellular environment, but more work is required to determine how other condensate components may modulate ETS DBD-induced rigidity.

Fig. 7.

ETS DBDs enhance the rigidification of biomolecular condensates containing EWS::FLI1 or the EWS^LCD. Without ETS DBDs, EWS^LCD exists in an equilibrium between monomers, liquid-like condensates, and rigid condensates. (A) Colocalization of ETS DBDs enhances EWS^LCD phase separation and accelerates rigidification, forming ThT-positive cross-β structures. (B) DNA binding by free ETS DBDs prevents their colocalization with EWS^LCD reducing their impact condensate formation.

EWS^LCD self-association stabilizes EWS::FLI1 binding to DNA enhancers, aids in recruiting RNA Pol II and the BAF complex, and is essential for oncogenic gene activation (4, 6, 28). Chong et al. overexpressed exogenous EWS^LCD in EwS patient–derived cells to artificially increase LCD–LCD and assess their impact on EWS::FLI1 driven transcription. They found that excess EWS^LCD reduced transcriptional activation likely by increasing LCD–LCD interactions (39), suggesting transcription requires a finely tuned level of these interactions. Our findings align with this model, as excess EWS^LCD counteracted EWS::FLI1 effects and altered condensate liquidity. This raises the question of whether EWS::FLI1’s tunable transcriptional activity depends on the number of molecules in transcriptional hubs or on the regulation of intermolecular dynamics within them. These findings also provide mechanistic insight into the proposed Goldilocks effect of EWS::FLI1 toxicity where too high expression of the fusion possibly results in uncontrollable aggregation and eventual cellular death (34). Following this logic, certain ETS fusions such as EWS::PU.1 are not observed since they may produce a protein too toxic to be tolerated by the cell. Given that FLI1^DBD affects EWS^LCD condensate liquidity, further investigation is needed in the context of full-length EWS::FLI1 rather than isolated EWS^LCD constructs. We note that such experiments are extremely challenging.

The three FET family proteins can form condensates and hydrogels (27, 28). Hydrogels formed by their LCDs contain cross-β structure, which facilitate copolymerization with binding partners like the RNA Pol II CTD and hnRNPA1/2 (5, 27, 42). Given this, it is unsurprising that the EWS^LCD can form ThT-positive condensates, indicative of cross-β structures (64). High-resolution structural models of cross-β structures exist for short segments of FUS (65, 66) and for the full FUS LCD (43). Recent findings show that FET fibrilization begins at the condensate surface, forming a rigid, protofibrillary shell, stabilized by ThT-positive cross-β structures (42)—features resembling EWS::FLI1 condensates. If EWS^LCD forms cross-β structures through parallel, in-register β-strands, it would create “ladders” of identical sidechains stacked at intervals matching the interstrand distance (67). These stabilizing ladders form either through π–π stacking of aromatic tyrosine residues or by complementary hydrogen bonding of glutamine/asparagine residues (67). The exact mechanism by which ETS DBDs disrupt the intra- and intermolecular interactions governing phase separation and cross-β formation remains unclear. However, tyrosine residues in the EWS^LCD are essential for functional self-association (4, 33, 68). Indeed, removing all tyrosine residues from the LCD abolishes its ability to form condensates (29, 33), prevents EWS::FLI1 from recruiting the BAF complex (4), and eliminates the transformative activity of EWS::FLI1 (68). Our ThT assays, NMR, and mutagenesis studies indicated that residues in loops 3, 4, and 6 of the ETS DBDs contribute positively charged residues (e.g., lysine in PU.1) to the interaction with the EWS^LCD. Supporting this theory, charged residues such as lysine have been demonstrated to be important for the interaction of the CTD of RNA Pol II with FET family proteins, TAF15 and FUS (69, 70). Since the highly positively charged EWS^RRM-RGG2 and poly-L-lysine constructs had only a minimal effect on EWS^LCD condensate rigidity, the conformation of the ETS DBD loops relative to each other must also contribute to condensate rigidification. The ThT increase we observed with EWS^RGG3 suggests additional factors may contribute to ThT positivity, beyond cross-β formation. Changes in condensate viscosity, known to affect ThT fluorescence, could also play a role (71).

While fluorescence microscopy revealed, unexpectedly, that HA DNA is mostly excluded from condensates formed by EWS^LCD + EWS::FLI1 or FLI1^DBD, posttranslational modifications, or additional binding partners such as EWS or other nucleic acid binding proteins may provide sufficient charge neutralization to allow for colocalization of DNA to condensates containing EWS::FLI1. Zuo et al. reported GFP-tagged EWS::FLI1 formed condensates that colocalized with DNA (6). Here, untagged EWS::FLI1 formed ThT-positive aggregates too quickly without condensate partners, such as EWS^LCD, preventing accurate study of its condensate formation with DNA, consistent with the findings of Ryan et al. (72). The ligation of a highly soluble domain, such as GFP, to EWS::FLI1 may reduce the aggregation rate sufficiently to better study the effect of DNA or other additives on condensate formation without EWS^LCD present (73). Future studies with relevant protein and nucleic acid partners will be needed to further investigate the biophysical properties of complex multicomponent condensates containing EWS::FLI1.

EWS::FLI1 drives oncogenesis through dysregulation of genes downstream of DNA enhancers, aberrant transcription, and deregulation of alternative splicing programs, all of which are dependent on an intact DBD (1, 60, 74). As part of its oncogenic function, EWS::FLI1 acts as a dominant-negative against the wild-type copy of EWS (15, 30), and associates with a multitude of nucleic acid binding proteins such as FUS, and RNA Pol II (7, 16, 32). The structural biology and biological implications of these interactions remain poorly understood, due to their heterogenous and transient nature. Here, we demonstrated that the ETS DBDs interact transiently yet specifically with the EWS^LCD, altering the intermolecular interactions that govern its condensate formation and subsequently promoting condensate rigidification (Fig. 7). Our findings provide a mechanism that explains how EWS::FLI1 exerts a dominant negative influence over the normal functions of EWS, alters the properties of condensates and potentially interferes with the native functions of other RNA-binding proteins. In EwS, the mislocalization of the FLI1^DBD (or other ETS DBDs) in FET-fusion proteins promotes EWS self-association and aggregation, disrupting local protein network dynamics and interfering with essential intermolecular interactions.

Materials and Methods

Briefly, U2OS-lac cells were cultured and transfected with Halo-tagged EWS and EWS::FLI1 constructs, FRAP was used to assess in vitro and in cellulo condensate properties. Protein constructs were expressed in Escherichia coli, purified via immobilized metal affinity and size-exclusion chromatography, and labeled for fluorescence imaging. Turbidity and ThT assays were performed to assess condensate formation and rigidification. NMR spectroscopy was used to analyze ETS DBD and EWS^LCD interactions. Detailed descriptions of the methods used in this study are available in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Dynamics of condensates formed by 25 μM EWS^LCD in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 30 μm.

Movie S2.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM FLI1^DBD in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 30 μm.

Movie S3.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM EWS^RRM in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 20 μm.

Movie S4.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM EWS^RRM–^RGG2 in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 20 μm.

Movie S5.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM PU.1DBD in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10x magnification. Video speed is increased 6×. Scale bar = 30 μm.

Movie S6.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM EWSRGG3 in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 20 μm.

Movie S7.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM poly-L-lysine 30 in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 20 μm.

Movie S8.

Dynamics of condensates formed by 25 μM EWS^LCD + 25 μM poly-L-lysine 50 in 100 mM NaCl, 20 mM sodium phosphate buffer pH 7.4. Images were collected at 10× magnification. Video speed is increased 6×. Scale bar = 20 μm.

Data, Materials, and Software Availability

Expression plasmids data have been deposited in Addgene (188042–188053) (75–85).