RNA-Induced Conformational Switching and Clustering of G3BP Drive Stress Granule Assembly by Condensation

Abstract

Stressed cells shut down translation, release mRNA molecules from polysomes, and form stress granules (SGs) via a network of interactions that involve G3BP. Here we focus on the mechanistic underpinnings of SG assembly. We show that, under non-stress conditions, G3BP adopts a compact auto-inhibited state stabilized by electrostatic intramolecular interactions between the intrinsically disordered acidic tracts and the positively charged arginine-rich region. Upon release from polysomes, unfolded mRNAs outcompete G3BP auto-inhibitory interactions, engendering a conformational transition that facilitates clustering of G3BP through protein-RNA interactions. Subsequent physical crosslinking of G3BP clusters drives RNA molecules into networked RNA/protein condensates. We show that G3BP condensates impede RNA entanglement and recruit additional client proteins that promote SG maturation or induce a liquid-to-solid transition that may underlie disease. We propose that condensation coupled to conformational rearrangements and heterotypic multivalent interactions may be a general principle underlying RNP granule assembly.

Keywords: G3BP; Neurodegenerative disease; RNP granules; liquid-to-solid transition; phase separation; stress granules; stress response.

Graphical abstract

Figure S1

Related to Figure 1 A: Fluorescence images from live-cell time-lapse movies of HeLa cells expressing G3BP1-mCherry filmed with lattice light sheet microscopy (LLSM, bottom) and conventional widefield microscopy (Convent., top). White arrows: intracellular G3BP1 foci. B: Fluorescence images of a LLSM movie of oxidatively stressed HeLa cells expressing G3BP1-mCherry. Time series starts 26 min after exposure to 1 mM sodium arsenate. Inset focuses on late G3BP1-positive SGs. Upon encounter, SGs coalesce with one another. C: Quantification of the number, volume and intensity of G3BP1-positive SGs as a function of time. Quantifications are shown for the mean averages of four time-lapse movies. Standard deviation is shown in gray. D: Fluorescence images of a partial photobleaching experiment of G3BP1-positive SGs in HeLa cells. Images show one granule before, the first frame and 120 s after photobleaching. Quantification shows the normalized mean fluorescence recovery curve (black dots), fit to the data (cyan) and SD (gray), n = 10. E: Still fluorescence images of a photobleaching experiment used to evaluate the internal diffusion of G3BP1 within SGs in HeLa cells (top). Kymograph of the signal intensity of the bleached G3BP1-positive SG as a function of time (x axis) along the SG distance (cyan bar indicated in the still images on top). Recovery of fluorescence occurs from the inside of the SG. F: Fluorescence images of an oxidatively stressed HeLa cell with G3BP1-mCherry SGs. The intact cell was permeabilized and treated with buffer (w/o RNase) or RNase A (w/ RNase). Scale bar 10 μm, except in D where it is 2 μm.

Figure 1

Liquid-like G3BP1 SGs Form by Heterotypic Phase Separation in Cells (A) Fluorescence images from a LLSM time-lapse movie of an oxidatively stressed HeLa cell expressing G3BP1-mCherry. Inset: G3BP1 foci. Scale bar, 10 μm. (B) Analysis of the mean averages of G3BP1 foci within an individual cell as a function of time. (C) Left: normalized number of G3BP1-positive SGs in individual cells as a function of time (every 100th data point is shown). Center: correlation of SG assembly onset as a function of G3BP1 mean fluorescence intensity. Right: correlation of G3BP1-positive SG number as a function of G3BP1 mean fluorescence intensity. Dashed lines are linear regression as a guide. See also Figure S1 and Videos S1 and S2.

Figure 2

In Vitro Reconstituted G3BP1 Condensates Recapitulate Cellular SG Properties (A) Schematic domain structure of G3BP1. (B) Phase diagram of G3BP1(WT) as a function of protein and RNA concentration. Right: fluorescence images of G3BP1(WT) with and without RNA. (C) Analysis of in vitro partial FRAP of G3BP1(WT)-RNA condensates. Mean average data (gray dots), fit (black), SD (light gray), n = 20. (D) Fluorescence images from a time-lapse video of G3BP1(WT)-RNA condensate fusion. (E) Fluorescence images of G3BP1 variants with RNA. (F) Partition coefficient of GFP-tagged RBPs in preformed SNAP (Alexa 546)-labeled G3BP1-RNA condensates. PSPC, SFPQ, and GFP served as negative controls (n = 150 fields of view [FOVs]) (Figure S2H). (G) G3BP1(WT) saturation concentration (Csat) with and without mCherry-Caprin-1 (mean, SD, fit, n = 20 FOVs). (H) Coalescence of G3BP1(WT)-RNA condensates with or without equimolar Cy3-labeled Ubc9(WT) or Ubc9(ts), measured with dual-trap optical tweezers. (I) Diffusion time of Ubc9(WT) and Ubc9(ts) within G3BP1-RNA condensates, determined by FCS. Significance levels: ^∗ < 0.05, ^∗∗ < 0.01, ^∗∗∗ < 0.001. (J) Mean average immobile fraction of Ubc9(WT) and Ubc9(ts) within G3BP1-RNA condensates as a function of time, determined by FRAP (SD, n = 5). Condensates formed with 1% PEG-20K and, when specified, 75 ng/μL of total RNA. Scale bars, 10 μm, except 1 μm in (D). See also Figure S2 and Video S3.

Figure S2

Related to Figure 2 A: Coomassie stained, non-native 4%–12% NuPage Bis-Tris gel showing ~1 μg of heterologous expressed and purified proteins. B: Phase diagram of G3BP1(WT) with control parameters KCl and protein concentration. Phase separation was scored by the absence or presence of G3BP1 condensates. C: Fluorescence images of G3BP1(WT)-RNA condensates before (top) and after (bottom) increasing the NaCl concentration from 30 mM to 280 mM. Scale bar, 10 μm. D: Phase diagram of G3BP1(ΔRG) as a function of protein and RNA concentration. (x) represent tested conditions and depicts that no condensates were present. E: Phase separated fraction of G3BP1(WT), G3BP1(ΔRG) and G3BP1(ΔNTF) as a function of PEG-20K concentration, in the absence of RNA (mean, SD, fit, n = 10 FOV). F: Fluorescence images of mCherry-Caprin-1 in the absence and presence of GFP-G3BP1. Scale bar, 10 μm. Right panel shows the quantification of the mCherry-Caprin-1 partition coefficient into G3BP1-RNA condensates as a function of G3BP1(WT) concentration. Caprin-1 concentration was 2.5 μM. G: EMSA to determine the apparent binding affinity of G3BP1 in the absence (gray) or presence of 1 μM mCherry-Caprin-1 (magenta). Quantification of one representative experiment is shown. H: Fluorescence images of indicated GFP-labeled client proteins partitioning into reconstituted SNAP(Alexa546)-labeled G3BP1-RNA condensates. Client proteins were added to preformed G3BP1-RNA condensates. As control, client proteins were tested in the absence of G3BP1-RNA condensates, showing that none of them phase separates (top panel) (scale bar, 10 μm). I: Fluorescent images of preformed FUS-GFP condensates (10 μM) in the presence of SNAP(Alexa546)-labeled G3BP1 (6 μM), in the absence of RNA. J: Fluorescence images of G3BP1(WT)-RNA condensates in presence of Cy3-labeled Ubc9(ts) and Ubc9(WT) (scale bar, 5 μm). Quantification of the mean average Cy3-Ubc9 fluorescence in G3BP1-RNA condensates is shown. K: Quantification of the diffusion times of Cy3-labeled Ubc9(ts) and Ubc9(WT) in solution, determined by FCS. Except for panels E and I, G3BP1-RNA condensates were formed in the presence of 75 ng/μl of total RNA (isolated from HeLa cells) and 1% PEG-20K.

Figure 3

G3BP1 Condensate Assembly Requires Long Unfolded RNA and Prevents RNA Entanglement (A) Fluorescence images and quantification of the phase-separated fraction of G3BP1(WT) with total RNA, mRNA, ribosomal RNA (rRNA), or NEAT1 RNA (n = 10 FOVs). (B) Fluorescence images of G3BP1(WT) with synthetic homopolymeric RNAs (STAR Methods). (C) Fluorescence images and quantification of the phase-separated fraction of G3BP1(WT) with folded and temperature-unfolded rRNA (n = 35 FOVs). (D) G3BP1(WT) phase separation with native total RNA or entangled RNA as a function of protein concentration (mean, SD, fit, n = 25 FOVs). (E) PEG-induced poly(G) RNA tangles with or without G3BP1 variants. Bright field (BF), RNA stained with F22. (F) Fusion of G3BP1-poly(G) condensates (magenta) at different time points after formation, assessed with dual-trap optical tweezers. Poly(G) tangles (dashed line) did not fuse. (G) Analysis of the A(60)-Cy5 RNA fraction bound to G3BP1 variants as a function of protein concentration, determined by EMSA. Scale bars, 10 μm, except 3 μm in (F). See also Figure S3.

Figure S3

Related to Figure 3 A: Phase separated fraction of G3BP1(WT) in the presence of 5-Loop and polyadenylated 5-Loop RNAs (n = 8 FOV). B: Fluorescence images of GFP-G3BP1(WT)-RNA condensates formed with polyadenylated and Cy5-labeled 5-Loop RNA. Scale bar, 5 μm. Condensates in A-B were formed in the presence of 1% PEG-20K and 75 ng/μl RNA. C: Quantification of the soluble fraction of RNA remaining after tangle formation (mean and SD from three independent measurements). D: Fluorescence immunostaining of oxidatively stressed U2OS wild-type cells and U2OS G3BP1/2 KO cells. G3BP1 (green) and eIF3η (magenta) are shown. Scale bars, 10 μm. E: Quantification of the number of eIF3η-positive SGs per cell in U2OS wild-type and in U2OS G3BP1/2 KO cells (mean and SD, n = 184 - 259 cells). F: Translation levels in U2OS cells (WT or G3BP1/2 KO) upon oxidative stress and recovery from stress. Translation levels were assessed by puromycin incorporation. Immunoblots against puromycin, p-eIF2a (to follow stress kinetics) and eIF2a (loading control) are shown. Harringtonine (Harr.) was used to prove specific incorporation of puromycin into nascent polypeptide chains. On the right, the normalized translation levels upon stress and recovery are shown (mean and SD from three independent experiments). G: EMSA to determine the apparent binding affinity of G3BP1(WT), G3BP1(ΔRG) and G3BP1(ΔNTF2) to A(60)-Cy5 RNA. Black arrow points toward free A(60)-Cy5 RNA. ^∗ indicates shifted RNA species due to G3BP1 binding. H: EMSA testing for the competition of unlabeled A(60) and poly(A) (500-4000 nt) RNAs for G3BP1(WT) binding to the A(60)-Cy5 probe. From left to right: Lane1: without G3BP1, Lane 2: without competitor, Lane 3: with unlabeled A(60) RNA as competitor, Lane 4: with unlabeled poly(A) as competitor. Black arrow points toward free A(60)-Cy5 RNA. ^∗ indicates shifted RNA species due to G3BP1 binding.

Figure S7

Related to Figures 4 and 7 A: Schematic representation of G3BP1 and G3BP2 domains, with the content of Arginine residues within the RG-rich region highlighted. Below, a sequence alignment of G3BP1 and G3BP2 RG-rich region. Arginine residues depicted in cyan. B: The sequence conservation of 339 G3BP1/2 orthologs was determined. The two folded domains are highly conserved, while the IDRs show virtually no absolute sequence conservation, except for a highly conserved binding motif in the PxxP domain which we speculate mediates recruitment of other key component(s). C: Despite having almost no sequence conservation, the region that would correspond to the acidic IDR (magenta) is almost uniformly highly acidic across all orthologous G3BP1/2 sequences.

Figure 4

The Acidic IDR of G3BP1 Is a Negative Regulator of Phase Separation (A) Top: schematic G3BP1 domain structure. Phosphosites S149 and S232 are depicted. IUPred, prediction of intrinsic disorder; FOLD, folding prediction using the foldindex (gray), PAPA (black), and PLAAC (magenta) algorithms; NCPR, net charge per residue with a sliding window of 10 residues; Net positive (cyan), net negative (magenta). Acidic clusters E1 and E2 are shown. (B) G3BP1(WT) phase-separated fraction as a function of pH (mean, SD, fit, n = 20–40 FOVs). Top: fluorescence images of G3BP1(WT) at the indicated pH (scale bar, 5 μm). (C) Fraction phase-separated of G3BP1 variants as a function of protein concentration under the specified conditions (mean, SD, fit, n = 20 FOVs). (D) Fraction phase-separated of G3BP1, WT or variants, under the specified conditions (n = 10 FOVs). (E) Csat of mCherry-G3BP1, WT (n = 414 cells), and ΔE1ΔE2 (n = 378 cells) overexpressed in U2OS G3BP1/2 KO cells. (F) G3BP1 assembly kinetics upon oxidative stress in U2OS G3BP1/2 KO cells expressing mCherry-G3BP1 variants (WT, n > 20; ΔRG, n > 10; ΔE1ΔE2, n > 20 cells), assessed by the number of SGs per cell as a function of time. SEM is depicted in a light color. See also Figure S4, Figure S7B, and S7C.

Figure S4

Related to Figure 4 A: Phase separated fraction of G3BP1(WT) in comparison to G3BP1(ΔE1) and G3BP1(ΔE2) variants. Phase separation was probed at pH 6 with 75 ng/μl HeLa total RNA and without PEG-20K (mean, SD, fit, n = 20 FOV). B: Quantification of the phase separated fraction of G3BP1 variants in the absence of RNA and PEG-20K, at pH 6 or 7 (mean, SD, fit, n = 20 FOV). C: Left: Quantification of the SG size in live U2O2 G3BP1/2 KO cells transfected with plasmids for expression of mCherry-G3BP1, WT (n = 1013 cells) or ΔE1ΔE2 (n = 1061 cells). Right: representative images of U2OS G3BP1/2 KO cells expressing mCherry-tagged G3BP1(ΔE1ΔE2) or G3BP1(WT), before and 60 min after addition of 1 mM sodium arsenate. Scale bar, 10 μm. D: Immunoblot for expression levels of mCherry-G3BP1 variants in G3BP1/2 KO cells. Tubulin was used as loading control. E: Left: Coomassie (CBB) stained non-native SDS-PAGE showing ~1 μg of purified GFP-labeled G3BP1(WT), G3BP1(S149A), G3BP1(S149E) and G3BP1(S149A/S232A). Right: Immunoblot (IB) against G3BP1 phospho-serine residue 149. From left to right: G3BP1(WT), G3BP1(S149A), dephosphorylated G3BP1(WT) (+PPase). Sample loading was controlled with antibodies against GFP. F: Left, phase separated fraction of G3BP1(WT) phosphorylated (gray) or dephosphorylated (green) with Lambda-phosphatase as a function of pH (mean, SD, fit, n = 16 FOV). Right, phase separated fraction of G3BP1(WT) (gray), G3BP1(S149A) (blue), G3BP1(S149E) (cyan) and G3BP1(S149A/S232A) (green) as a function of pH (mean, SD, fit, n = 25 FOV). Phase separation was tested in the presence of 75 ng/μl of poly(A) RNA and absence of PEG-20K.

Figure 5

Interactions between RG-Rich and Autoinhibitory Acidic IDRs Regulate G3BP1 Phase Separation (A) Conformational snapshot of G3BP1 from an excluded volume (EV) simulation (instantaneous R_H, ∼11 nm). (B) Normalized distance between any pair of residues in G3BP1(ΔNTF2) simulations. Cooler colors are closer together, and warmer colors are farther apart. Dashed lines delineate domains. (C) R_H for G3BP1 dimers, inferred from FCS measurements and EV simulations of full-length dimeric G3BP1 with generic, self-avoiding descriptions for the IDRs. (D) R_H of G3BP1(WT) as a function of KCl concentration, determined by DLS. Fit shown with a dashed line. (E) R_H of G3BP1 variants, determined by FCS. (F) R_H of G3BP1(WT) at pH 6 and pH 7.5, determined by DLS. (G) Oligomeric species of G3BP1(WT) at pH 6 or pH 7.5, detected by DLS. Shown is the mean average (n = 30). (H) Analytical gel filtration of G3BP1(WT) and G3BP1(ΔRG) on A(60) RNA as a function of RNA concentration. Significance levels: ^∗ < 0.05, ^∗∗ < 0.01, ^∗∗∗ < 0.001. See also Figure S5.

Figure S5

Related to Figure 5 A: Homology model sequence alignment for G3BP1-RRM. Model parameters are robust, and the resulting homology model is structurally reasonable and high-confidence. B: The intramolecular distances between the acidic tract and the RG-rich region are substantially lower in full Hamiltonian simulations (Full) than would be expected based on a generic self-avoiding description of the IDR (Excluded Volume, EV). This result highlights the importance of intramolecular attractions between the acidic tract and the RG-rich region. C: Left, comparison of the hydrodynamic radius (R_H) of phosphorylated G3BP1(WT) (indicated with ‘P’) and dephosphorylated G3BP1(WT). Right, R_H of phosphorylated G3BP1(WT) and G3BP1(S149A/S232A). R_H was determined by FCS. D: Right, EMSA to determine the apparent binding affinity of G3BP1(WT), G3BP1(ΔE1), G3BP1(ΔE2) and G3BP1(S149A) to A(60)-Cy5 RNA. Black arrow points toward free A(60)-Cy5 RNA. ^∗ indicates shifted RNA species due to G3BP1 binding. The quantification of the EMSAs is shown on the left (G3BP1(WT): KD ~2.0 μM; G3BP1(S149A): KD ~1.9 μM; G3BP1(ΔE1): KD ~1.9 μM; G3BP1(ΔE2): KD ~1.7 μM). E: R_H of G3BP1(WT) at pH 6 and 85 mM KCl and after increasing the KCl concentration to 400 mM, determined by DLS (mean average of 6 measurements). On the left, the increase of R_H of the G3BP1(WT) dimer upon increasing KCl concentration is shown. On the right, the oligomeric species of G3BP1(WT) are depicted. Note that the intensity of the oligomer decreases at 400 mM KCl.

Figure 6

Ultra-coarse-grained Simulations Reveal RNA-Dependent Clusters of G3BP1 (A) A G3BP1 dimer is coarse-grained into a two-bead model, with one bead capturing the NTF2 domain and central IDRs and the other bead capturing the RRM and RG-rich region. The RNA is coarse-grained so that each “bead” represents 8–10 nt. The interaction strength between distinct beads is shown. (B) Analysis of RNA-mediated G3BP1 oligomerization for simulations of G3BP1, WT or ΔRG, with short RNA. (C) Simulation snapshots illustrate the cooperative nature of oligomerization, demonstrating the coexistence of fully decorated short RNA molecules and predominantly unbound RNA molecules. (D) Simulations with long RNA yield clusters that form on the RNA and act as crosslinkers between RNA-RNA interactions. (E) Schematic highlighting the clusters as physical crosslinks. (F) X-Y cross-sectional slice of a refractive index tomogram of G3BP1-RNA condensates. Scale bar, 10 μm. (G) Refractive index and total concentration inside G3BP1-RNA condensates (n = 187 condensates). (H) Concentration of G3BP1 inside versus outside the condensate (n = 437 condensates). G3BP1-RNA condensates in F-H were formed with 50 ng/μl poly(A) RNA. See also Figure S6 and Video S4.

Figure S6

Related to Figure 6 A: Calibration curve of GFP fluorescence as a function of GFP concentration. B: Quantification of GFP-G3BP1 enrichment within G3BP1-RNA condensates, as determined by fluorescence intensity inside/outside condensates. C: Cryo-EM image of G3BP1-RNA droplets deposited post fixation on a holey film EM grid. Arrowheads point to droplets partially deposited in the holes of the film that are amenable for tomographic tilt series acquisition. D: Tomographic slice, 4 nm in thickness, of a G3BP1-RNA droplet. Putative naked RNA molecules are observed outside the droplet. Multiple ‘beads on a string’ are observed inside the droplet and may be interpreted as G3BP1 molecules bound to RNA. The dense beads are 3-4 nm in diameter, and regularly appear in pairs. Occasionally, larger clusters (30-40 nm diameter) of beads are observed. E: Zoom into cluster in D. For the full tomographic volume, see Video S4. F: Additional examples of clustering and pairs of ‘beads on a string’ from a different tomographic slice of the same droplet.

Figure 7

The Valence of Arginine Residues within the RG-Rich Region Determines the Efficiency of G3BP Phase Separation (A) Phase separated fraction of 5 μM G3BP1 or G3BP2 as a function of PEG-20K concentration without RNA (mean, SD, fit, n = 16 FOVs). (B) Phase-separated fraction of G3BP1 and G3BP2 as a function of G3BP concentration with poly(A) RNA and 1% PEG-20K (mean, SD, fit, n = 16 FOVs). (C) R_H of G3BP2 at pH 7 and 85 mM KCl (solid line) or after increasing the KCl to 400 mM (dashed line), as determined by DLS (mean average of 6 measurements). (D) DLS measurement of the oligomeric species of G3BP2 at pH 7 and 85 mM KCl or after increasing the KCl to 400 mM (mean average of 6 measurements). (E) Model depicting an RNA-mediated conformational transition of G3BP into a phase-separation competent state (top panel). Under physiological conditions, G3BP adopts a compact state (left) that is stabilized by intramolecular interactions between the RG-rich region and the acidic region. The compact state inhibits G3BP phase separation. Upon stress, polysomes disassemble, and mRNAs are released in an unfolded protein-free state. Binding of unfolded mRNA to G3BP outcompetes the intramolecular interactions between the RG-rich and the acidic regions. RNA-bound G3BP adopts an expanded conformation in which the RG-rich region becomes exposed to engage in protein-protein and protein-RNA interactions. These new interactions among RG-rich regions stabilize clusters of G3BP1 bound to RNA (bottom panel). RNA-mediated G3BP clusters allow physical crosslinking of RNA molecules to form inhomogeneous protein-RNA condensates. See also Figure S7.