Single-molecule imaging reveals dynamic biphasic partition of RNA-binding proteins in stress granules

Abstract

Stress granules (SGs) are cytosolic, nonmembranous RNA-protein complexes. In vitro experiments suggested that they are formed by liquid-liquid phase separation; however, their properties in mammalian cells remain unclear. We analyzed the distribution and dynamics of two paradigmatic RNA-binding proteins (RBPs), Ras GTPase-activating protein SH3-domain-binding protein (G3BP1) and insulin-like growth factor II mRNA-binding protein 1 (IMP1), with single-molecule resolution in living neuronal cells. Both RBPs exhibited different exchange kinetics between SGs. Within SGs, single-molecule localization microscopy revealed distributed hotspots of immobilized G3BP1 and IMP1 that reflect the presence of relatively immobile nanometer-sized nanocores. We demonstrate alternating binding in nanocores and anomalous diffusion in the liquid phase with similar characteristics for both RBPs. Reduction of low-complexity regions in G3BP1 resulted in less detectable mobile molecules in the liquid phase without change in binding in nanocores. The data provide direct support for liquid droplet behavior of SGs in living cells and reveal transient binding of RBPs in nanocores. Our study uncovers a surprising disconnect between SG partitioning and internal diffusion and interactions of RBPs.

Figure 1.

G3BP1 and IMP1 exhibit different dynamics of protein exchange between SGs. (A) Protein interaction (gray) and RNA-binding domains (black) of human G3BP1 and IMP1 according to SMART analysis for identification of signaling domains (Schultz et al., 1998). RRM, RNA recognition motif; KH, K homology domain. Right: Western analysis of cellular lysates after transfection with PAGFP-tagged G3BP1 and IMP1 (arrowhead). Lysates were analyzed using size-based capillary electrophoresis, and electropherograms are represented as pseudoblots as described in Materials and methods. Molecular mass standards are indicated. The respective endogenous proteins are indicated by an arrow. Please note that analysis by size-based capillary electrophoresis can yield protein mobilities that differ from separation by standard SDS-PAGE. (B) Colocalization of exogenously expressed mCherry-G3BP1 with the SG marker TIA-1. Bar, 10 µm. (C) Colocalization and dynamic exchange of fluorescence-tagged G3BP1 and IMP1 in SGs of living PC12 cells. Granules were labeled with mCherry-IMP1 and PAGFP-G3BP1 was activated in one granule (dashed square). The outline of the cell and the nucleus are indicated in the red fluorescent micrographs. Fluorescence distribution was followed over time. After some seconds, photoactivated PAGFP-G3BP1 appears in SGs outside of the activated region indicating dynamic exchange of G3BP1 between SGs. Bar, 10 µm. (D) Bar plot showing the size distribution of SGs as determined from the area of mCherry-IMP1 positive granules. The box represents 50% of the population, whiskers range from 5% to 95% and crosses correspond to the minimal and maximal values (n = 34). (E) Schematic representation showing the FDAP approach to determine dynamics and binding of PAGFP-tagged G3BP1 and IMP1 in granules. Photoactivation and fluorescence recording was performed in a 3 × 5 µm region containing cytosol and granules between nucleus and plasma membrane (red box). The decay curves were fitted with model FDAP functions. (F) FDAP curves for PAGFP-IMP1, PAGFP-G3BP1, and 3×PAGFP (mean ± SEM, n = 13 [PAGFP-IMP1], n = 17 [PAGFP-G3BP1], n = 20 [3×PAGFP]) showing the different dynamics of IMP1 and G3BP1. (G) Residence time of PAGFP-tagged IMP1 and G3BP1 in granules as determined by the model FDAP function (mean ± SEM, n = 19 [PAGFP-IMP1], n = 20 [PAGFP-G3BP1], n = 20 [3×PAGFP] from two [3×PAGFP], four [PAGFP-G3BP1], and five [PAGFP-IMP1] independent experiments). Comparison between the constructs involved one-way ANOVA followed by post-hoc Tukey's test. **, P < 0.01; ***, P < 0.001 (compared with control [3×PAGFP]); ⁺, P < 0.05. For all experiments, stress had been induced by a 20-min treatment with 0.5 mM sodium arsenite before imaging.

Figure 2.

Effect of G3BP1 and IMP1 deletions on protein exchange between SGs. (A and B) FDAP curves for PAGFP-G3BP1_C (A) and PAGFP-IMP1_C (B; mean ± SEM, n = 22 and n = 13, respectively). Schematic representations of the deletion constructs are shown on top. Protein interaction (gray) and RNA-binding domains (black) are indicated. Residence time of PAGFP-tagged G3BP1_C (A) and PAGFP-IMP1_C (B) in granules as determined by the model FDAP function (mean ± SEM, n = 22 from four and n = 13 from six independent experiments, respectively) are shown on the right. Values for full-length PAGFP-G3BP1 and PAGFP-IMP1 from Fig. 1 are indicated for comparison by dotted lines (green and red, respectively). The deletion constructs were not showing statistically significant differences to their respective full-length counterparts, but were statistically significantly different from the control construct (3×PAGFP vs. PAGFP-G3BP1_C, P < 0.05; 3×PAGFP vs. PAGFP-IMP1_C, P < 0.001). Statistics involved one-way ANOVA followed by post-hoc Tukey's test. Bars, 20 µm. For all experiments, stress had been induced by a 20-min treatment with 0.5 mM sodium arsenite before imaging.

Figure 3.

G3BP1 and IMP1 are enriched in distributed nanocores within SGs. (A) Single-molecule imaging of SiR-labeled SNAP-G3BP1 and TMR-labeled HaloTag-IMP1. Snapshots before image acquisition confirmed the presence of SGs and showed that G3BP1 and IMP1 colocalized in the same granules (left). The outline of the cell and the nucleus are indicated. Localizations of single molecules in SGs of the indicated region (yellow box in the snapshot images) are shown (right). Bars: (left, middle) 10 µm; (right) 1 µm. (B) Spatial clustering of single-molecule localizations of an area within a single SG after combined expression of G3BP1 and IMP1. Raw localizations (left), the respective cluster patterns (middle), and cluster pattern overlap (right) are shown. (C) Quantitation of the fraction of the granular area, which is occupied by clusters, and weighted overlap for different combinations of expressed RBPs. The value for “n” represents the number of individual granules analyzed. Granules were obtained from 5–19 cells from three to eight independent experiments. For all experiments, stress had been induced by a 20-min treatment with 0.5 mM sodium arsenite before imaging.

Figure 4.

Nanocores are relatively immobile within SGs and contain multiple binding sites. (A and B) Top view (left) and kymographic representation (right) of a typical single binding event of Halo-IMP1 on SNAP-G3BP1 background (A) and Halo-G3BP1 on SNAP-IMP1 background (B) demonstrating spatial restriction of nanocores within SGs. (C) Bar plots showing the jump distance distributions of G3BP1 and IMP1 in different combinations indicating the presence of multiple binding sites within a nanocore. 14–27 transition events for each protein per condition were analyzed and the jump distance distributions determined by calculating the pairwise distances of all linked cluster centers. The box represents 50% of the population, whiskers range from 5% to 95%, and the horizontal line shows the median value.

Figure 5.

G3BP1 and IMP1 exhibit a biphasic partition in a bound and mobile fraction within SGs. (A) Example of a representative granule in a double-transfected cell. The granule was identified based on SiR staining (left) and localizations and trajectories based on TMR staining of HaloTag-G3BP1 (middle and right) are shown. The border of the SG is indicated by a dashed line. Bars, 1 µm. (B) Trajectory of a single G3BP1 molecule (as indicated by an arrowhead at time point 0) showing alternating phases of fast movement and immobile time periods. The trajectory is indicated in green. Bar, 250 nm. (C) High-mobility trajectories (mobile fraction, left) and low mobility trajectories (bound fraction, right) of HaloTag-G3BP1 within the granule shown in A. Bar, 1 µm. (D) Probability density of single-molecule diffusion coefficients from trajectories of a single granule expressing HaloTag-G3BP1 showing a biphasic partition in a bound (red) and mobile fraction (green). (E) Quantification of bound fractions of the respective Halo-tagged constructs expressed in different combinations. Calculations were performed on the total trajectories of 18–28 SGs per experimental condition as indicated by “n.” SGs were obtained from 12–15 cells from 3–10 independent experiments. For all experiments, stress had been induced by a 20-min treatment with 0.5 mM sodium arsenite before imaging.

Figure 6.

G3BP1 and IMP1 have a short lifetime in the bound fraction and display anomalous subdiffusion in the mobile fraction. (A) Lifetime determination of different combinations of HaloTag- and SNAP-tagged G3BP1 and IMP1 in the bound fraction of SGs. Trajectories of the respective Halo-tagged constructs (shown in bold) are evaluated. (B) Example of the trajectories of the mobile fraction of HaloTag-G3BP1 within a single granule (left) and plot of the MSD against the elapsed time showing anomalous diffusion. Mean diffusion constants (Γ) and α values of the respective HaloTag construct in the mobile fraction as determined for the different combinations. Bar, 500 nm. Calculations were performed on the mobile trajectories of 8–11 SGs per experimental condition as indicated by n. SGs were obtained from 8–11 cells from two to eight independent experiments. For all experiments, stress had been induced by a 20-min treatment with 0.5 mM sodium arsenite before imaging.

Figure 7.

Reduction in the number of LC regions of G3BP1 results in less detectable mobile molecules within the liquid phase without changing binding properties to nanocores. (A) Example of a representative granule after expression of a deletion fragment of G3BP1 (G3BP1_C) lacking two LC domains fused to HaloTag on a SNAP-IMP1 background. The granule was identified based on the SiR staining (left) and localizations and trajectories based on TMR staining of HaloTag-G3BP1_C (right) are shown. The border of the SG is indicated by a dashed line. Note that most of the trajectories remained locally very restricted. Bar, 1 µm. (B) Probability density of single-molecule diffusion coefficients from trajectories of a granule expressing HaloTag-G3BP1_C showing the presence of a substantial fraction of essentially immobile molecules. (C) Lifetime determination of HaloTag-G3BP1_C on SNAP-IMP1 background in the bound fraction of SGs. Quantification of granules from six cells from two independent experiments was performed. For all experiments, stress had been induced by a 20-min treatment with 0.5 mM sodium arsenite before imaging.

Figure 8.

Schematic representation visualizing the major findings of the study. The study reveals the presence of distributed nanocores within the mobile, liquid droplet-like phase of SGs. The two RBPs, G3BP1 (green arrows) and IMP1 (red arrows), exhibit markedly different SG–cytosol interconversion dynamics. Within granules, both RBPs exhibit alternating binding in the nanocores and liquid droplet-like diffusion in the mobile phase.