Visualizing liquid-liquid phase transitions

Abstract

Liquid-liquid phase condensation governs a wide range of protein-protein and protein-RNA interactions in vivo and drives the formation of membrane-less compartments such as the nucleolus and stress granules. We have a broad overview of the importance of multivalency and protein disorder in driving liquid-liquid phase transitions. However, the large and complex nature of key proteins and RNA components involved in forming condensates such as stress granules has inhibited a detailed understanding of how condensates form and the structural interactions that take place within them. In this work, we focused on the small human SERF2 protein. We show here that SERF2 contributes to the formation of stress granules. We also show that SERF2 specifically interacts with non-canonical tetrahelical RNA structures called G-quadruplexes, structures which have previously been linked to stress granule formation. The excellent biophysical amenability of both SERF2 and RNA G4 quadruplexes has allowed us to obtain a high-resolution visualization of the multivalent protein-RNA interactions involved in liquid-liquid phase transitions. Our visualization has enabled us to characterize the role that protein disorder plays in these transitions, identify the specific contacts involved, and describe how these interactions impact the structural dynamics of the components involved in liquid-liquid phase transitions, thus enabling a detailed understanding of the structural transitions involved in early stages of ribonucleoprotein condensate formation.

Fig.1. SERF2 colocalizes with stress granules upon various stress conditions.

a, Immunofluorescence images show that endogenous SERF2 is prevalently distributed in the nucleolus of fixed U2OS cells, as evidenced by fibrillarin staining. b, Colocalization analysis plot of fibrillin and SERF2 obtained from (a). c, SERF2 forms cytoplasmic foci and co-localizes with the core stress granule marker protein G3BP1 in different stress conditions, suggestive of its involvement in stress granule formation or stability. d, The plot shows the quantification of stress granules retrieved from (c) under various stress conditions containing both SERF2 and G3BP1. e, Fixed U2OS immunofluorescence cell images showing oxidative stress-induced granules containing SERF2, FUS, and rG4 quadruplexes detected by specific antibodies indicated in green, purple, and red, respectively. f, Plot showing SERF2 colocalization with FUS and rG4 quadruplexes retrieved from (e) as a measure of Pearson’s coefficient. The scale bars in (a, c, and e) are 10 μm.

Fig. 2|. SERF2 regulates stress granule formation and dynamics.

a, Immunofluorescence of SERF2 and G3BP1 in fixed U2OS cells treated with 0.5 mM sodium arsenite for 1 hour. b, Plot shows percentage of stress granule positive cells under sodium arsenite treatment calculated from images shown in (a). Error bars were calculated from four independent experiments. c, Live-cell imaging of EGFP-FUS HeLa Kyoto cells with siCTRL or siSERF2 conditions, treated with different stressors (Sodium arsenite, 0.5 mM; Sorbitol, 0.4 M; MG132, 10 μM) for 1 hour. Scale bars in (a and c) are 10 μm. d, Plot shows percentage of stress granule positive cells under different stress treatments calculated from images shown in (c), **** shown in (b and d) indicates P < 0.0001. e,f, The graphs show FRAP recovery curves in EGFP-FUS HeLa Kyoto cells with siCTRL or siSERF2 conditions treated with 0.4 M sorbitol and (e) 0.5 mM sodium arsenite (f).

Fig. 3|. High-throughput screening of SERF2 binding substrates.

a, Fluorescence polarization assay measuring the binding affinity of SERF2 with 6-FAM labeled random ribo-polynucleotides and rG4 quadruplex forming sequences as indicated. b, A schematic representation of RNA bind-n-seq experiments with RNA pools. A randomized DNA oligo was transcribed to RNA and folded in buffer containing KCl or LiCl. An additional RNA pool was made by replacing guanines (G) with 7-deaza (7dG) to limit rG4 quadruplex folding. These pools were mixed with GST-SBP-SERF2, and bound RNA was isolated and sequenced to ~10–20+e6 reads. c, RNA bind-n-seq analysis of the 6-mers enrichment in KCl versus RNA made with 7dG in a sample mixture containing 50 nM SERF2. The guanine-rich 6mers in the top 5 kmers are labeled. d, Enrichment of rG4 quadruplex patterns in different conditions and varied G4 quadruplex strengths containing 50 nM SERF2. Sequences containing 3 or more guanines in the G-tetrad are referred to as strong G4 quadruplexes, while sequences with ≥8 guanines and lacking a defined G4 forming motif are referred to as non-G4 quadruplexes. e, FOREST analysis average binding intensities of SERF2 containing 1800 folded human pre-miRNA and 10 rG4 quadruplex structures. The p-value was determined by the two-tailed Brunner-Munzel test. f, Fluorescence polarization plot shows the binding affinity between SERF2 and three different 6-FAM labeled rG4 quadruplexes measured at room temperature. The binding assays in a and f were done with varied protein concentrations mixed with 20 nM rG4 quadruplex or polynucleotides at 20 nM in 20 mM NaPi (pH 7.4) and 100 mM KCl.The standard deviations are calculated from three independent replicates.

Fig. 4|. RNA interaction drives liquid-liquid phase transition in SERF2.

a, Phase regimes illustrating phase transition in SERF2 as a function of total RNA concentration extracted from HeLa cells. b, Fluorescence imaging shows gel-like structures in 50 μM SERF2 mixed with 200 ng/μL of total RNA containing 10% (w/v) PEG8000, incubated for 30 minutes at room temperature. c-e, 50 μM SERF2, dissolved in 20 mM NaPi (pH 7.4) and 100 mM KCl, readily undergoes a phase transition (c-e, top) when mixed with equimolar concentrations of different rG4 quadruplexes, that include TERRA, (G4C2)₄, and (UG4U)₆. The sample mixture contains 1/200^th Cy-5 labeled SERF2 (purple signals) and 6-FAM rG4 quadruplex (green signals), as indicated in the figure inset. Two-component FRAP analysis (c-e, bottom) was done to measure the recovery rates of SERF2 and rG4 quadruplex in the SERF2-rG4s droplets. On the top of each FRAP plot, the pre-bleached, after-bleached (0 s), and recovered droplets (300 s) are shown. The FRAP data were fitted in GraphPad Prism, using a non-linear regression, one-phase association model, to obtain the recovery halftime (t1/2). f,g, Schematics showing SERF2 and TERRA rG4 sample mixtures phase separation, in 10% PEG8000, at varied protein to RNA concentrations (f), and salt to PEG8000 concentrations (g). h, Dynamics and recovery of the Cy5-labeled proteins, in SERF2-total RNA droplets, obtained by FRAP analysis suggesting that the gel-like structures are dynamic and reversible. Standard errors were calculated by analyzing 8 isolated droplets subjected to FRAP. i, DIC and fluorescence images showing co-phase separation of SERF2 (purple) and G3BP1 with 12.5 ng/μL HeLa total RNA. j, DIC and fluorescence images show SERF2 facilitates G3BP1-RNA condensation in samples containing variable SERF2 and G3BP1 concentrations as indicated and 12.5 ng/μL HeLa total RNA. k, Phase diagram showing G3BP1 phase transition in non-crowding and crowding conditions containing PEG8000 in 20 mM NaPi, pH 7.4, 100 mM KCl buffer. G3BP1 phase transition was measured in the absence or presence of SERF2 or a 36-nucletide long UG4U6 rG4 quadruplex or mixture of SERF2 and rG4s as indicated. l, Time-series images of 50 μM G3BP1 condensates after photobleaching in the absence or presence of SERF2, total RNA, and TERRA rG4 quadruplexes prepared in 20 mM NaPi, 100 mM KCl (pH 7.4) containing 5% PEG8000. m, FRAP recovery plots of G3BP1-Cy5 (left) and SERF2-AF488 (right) obtained from (l), and the graph colors correspond to the sample mixture shown in (l).

Fig. 5|. High-resolution NMR structure of SERF2 reveals the disordered and dynamic domain binds rG4 quadruplex.

a, 20 best NMR ensemble model structures of SERF2 calculated using Cyana and refined using Crystallography and NMR System. The average converged helical structure spanning residues 33–47 in SERF2 is shown in orange. b, 15N/1H amide resonance assignment (red spectrum) of 100 μM human SERF2 mixed with 20 μM (yellow) and 50 μM (purple) TERRA rG4 quadruplex at 4 °C. c, The 15N/1H chemical shift perturbations (CSPs) were calculated from (b) and plotted for each assigned residue in SERF2 with increasing TERRA rG4 quadruplex concentrations. The color codes correlate to the spectrum shown in (b) and unassigned peaks are denoted with asterisk. The heteroNOE values of SERF2 in the absence of TERRA rG4 quadruplex were plotted along with the CSPs on the right y-axis to highlight the correlation between SERF2 dynamics and TERRA rG4 interaction. d, 15N relaxation rates R2/R1 demonstrating a significant change in 200 μM SERF2 dynamics interacting with 100 μM TERRA rG4 as a function of residues. Protein and RNA samples are dissolved in a buffer containing 8% D₂O and spectra are recorded at 4 °C on a Bruker 600 MHz spectrometer. e,f, 2D analysis plots derived from AUC for 4.7 μM TERRA rG4 quadruplex mixed without (e) or with 2-molar excess SERF2 (f). The partial concentration shown in color on the right y-axes represents the abundance of individual species in the sample solution. g,h, Cartoon shows the top- and side-view of a quadrupole-like (ellipses, g) and planar (vertical slab, h) interaction in SERF2 and TERRA rG4 quadruplex complex. Residues generating the quadrupole-like interaction and TERRA rG4 quadruplex structure distortion are labeled, and hydrogen bonds are indicated with dashed lines. i, 3D structure of TERRA rG4 quadruplex complex with SERF2 before and after 0.5 μs MD simulation shows G-tetrad distortion in TERRA:SERF2 1:1 (center) and 1:2 (right) complex. G-tetrads are indicated by red arrows and each TERRA unit in the tetrameric structure (PDB ID: 2M18) is represented with different colors. A G-quartet in TERRA rG4 quadruplex is formed by guanines in i and i+6 as shown on the top. j, EMSA gel-shift assay of 5 μM TERRA rG4 quadruplex mixed with equimolar SERF2 and different lysine to alanine SERF2 mutants in a non-crowding condition (20 mM NaPi, pH 7.4, 100 mM KCl) as indicated in the figure. k, Phase transition of SERF2 and its mutants mixed with equimolar TERRA rG4 quadruplexes in a crowding condition (20 mM NaPi, pH 7.4, 100 mM KCl, 10% w/v PEG8000).

Fig. 6|. Integrated single-molecule fluorescence microscopy and MD simulation approach to study high-resolution structure of SERF2 – TERRA rG4 quadruplex phase separated condensate.

a, A cubic all-atom MD simulation box encapsulating randomly distributed 30 molecules SERF2 (orange), 30 molecules of TERRA rG4 quadruplex (blue), 10% PEG (pink), Cl⁻ (green), and K⁺ (grey). The downward arrow on top indicates the transition of a dilute phase low-ordered structure to a more droplet-like condensed phase structure in 0.5 μs large-scale MD simulation. b,c, Surface representation of the structure of SERF2-TERRA rG4 ring-shaped condensed droplet-like large-structure (b), and lower-ordered oligomers (c) obtained at time 0.5 μs MD simulation. The enlarged all-atom cartoon structures of the condensed droplet-phase and the lower-ordered 1:2 SERF2:TERRA rG4 dilute-phase oligomer are shown on the top. The three distinct contact sites in SERF2 in the lower-ordered complex structure (c, top) is shown in ball-stick format and the TERRA rG4 interacting residues are labeled. The high-resolution images in (b, top) show a representative interaction network that involves the three critical binding sites in SERF2 located in the disordered N-terminus. The SERF2 interacting residues are represented with an uppercase letter and TERRA rG4 quadruplex nucleotide in a smaller-case letter (e.g. R11-U14 denotes Arg3 and uracil 14 in SERF2 and TERRA rG4 quadruplex, respectively). d, (left) Single-molecule fluorescence microscopy shows a single droplet of 50 μM SERF2 (spiked with picomolar Cy5 labeled SERF2) mixed with equimolar TERRA rG4 quadruplexes. The corresponding single-particle tracking image of SERF2 molecules within the single droplet generated using ImageJ is shown on the right. e, Single-molecule tracking image constructed using the 0.5 μs all-atom MD trajectory in the SERF2-TERRA all-atom MD simulation system. Each sphere represents the center of mass of individual SERF2 molecules, and their trajectories were retrieved at every 20 nanoseconds interval from the 0.5 μs MD simulation to construct the tracks using ImageJ. The mean square displacement (MSD) of a set of representative SERF2 molecules in the lower-ordered dilute and droplet-like condensed phase are indicated by arrows. f, MSD plot of SERF2 (average MSD 0.95±0.23 μm²/s) obtained from the single-particle tracking fluorescence microscopy experiment. g,h, FRET distance histogram between donor and acceptor was determined based on the FRET efficiency in SERF2 (T2C and A51C) and TERRA rG4 phase-separated droplets containing picomolar Cy3-Cy5 labeled SERF2 in dilute-phase (g) and condensed-phase (h) samples separated after centrifugation (see methods). i, Mean distance between T2 and A51 in SERF2 was computed from 0.5 μs MD simulation in a system containing 30 SERF2 molecules and 30 TERRA rG4 molecules. The dashed line indicates molecules with a mean distance of 1.5 nm.