Yin and yang regulation of stress granules by Caprin-1

Significance

Stress granules (SGs) are membraneless organelles regulating many vital biochemical processes. G3BP1 plays a critical role in regulating SG dynamics through its interaction with Caprin-1 and USP10. The crystal structures of the NTF2L domain of G3BP1 in complex with the G3BP1-interacting motif (GIM) of Caprin-1 and USP10 showed that both GIMs interact with the same pocket of G3BP1. Both GIMs suppressed the liquid–liquid phase separation (LLPS) of G3BP1, suggesting that Caprin-1 facilitates SG formation via other mechanisms. The C-terminal domain underwent spontaneous LLPS and promoted SG formation, whereas the N-terminal domain and GIM of Caprin-1 suppressed LLPS and SG formation. We propose that Caprin-1 regulates SG dynamics by a “yin and yang” mechanism of its N- and C-terminal domains.

Abstract

Stress granules (SGs) are cytoplasmic biomolecular condensates containing proteins and RNAs in response to stress. Ras-GTPase–activating protein binding protein 1 (G3BP1) is a core SG protein. Caprin-1 and ubiquitin specific peptidase 10 (USP10) interact with G3BP1, facilitating and suppressing SG formation, respectively. The crystal structures of the nuclear transport factor 2-like (NTF2L) domain of G3BP1 in complex with the G3BP1-interacting motif (GIM) of Caprin-1 and USP10 show that both GIMs bind to the same hydrophobic pocket of G3BP1. Moreover, both GIMs suppressed the liquid–liquid phase separation (LLPS) of G3BP1, suggesting that Caprin-1 likely facilitates SG formation via other mechanisms. Thus, we dissected various domains of Caprin-1 and investigated their role in LLPS in vitro and SG formation in cells. The C-terminal domain of Caprin-1 underwent spontaneous LLPS, whereas the N-terminal domain and GIM of Caprin-1 suppressed LLPS of G3BP1. The opposing effect of the N- and C-terminal domains of Caprin-1 on SG formation were demonstrated in cells with or without the endogenous Caprin-1. We propose that the N- and C-terminal domains of Caprin-1 regulate SG formation in a “yin and yang” fashion, mediating the dynamic and reversible assembly of SGs.

Biomolecular condensation is a subcellular compartment without a surrounding membrane (also called a membraneless organelle) that dynamically regulates a variety of vital biochemical processes by aggregating multiple biomolecules (1–4). Stress granules (SGs) are cytoplasmic condensations formed in eukaryotic cells in response to a variety of stressors, such as heat shock, osmotic pressure, and drug-induced stress (5–7). Responding to stress, protein translation is halted, and SGs form with abundant uncoated mRNAs, RNA-binding proteins, and many translation initiation factors (8–11). SGs are dynamic and reversible under physiological conditions. They protect mRNAs from damage and prevent the accumulation of defective proteins caused by misfolding or premature termination (12, 13). SGs also play an important role under pathological conditions. For example, SGs are induced during viral infection, impacting viral replication, viral life cycle, and host–pathogen interaction (14–16). SGs are also implicated in cancer pathogenesis and progression (17, 18) and neurodegenerative diseases. In the latter, gene mutations implicated in amyotrophic lateral sclerosis and frontotemporal dementia can delay SG disassembly and promote the transition of SGs to irreversible protein aggregates (19–21).

Ras GTPase-activating protein-binding proteins (G3BPs) regulate the dynamic assembly and disassembly of SGs (22–24). G3BPs can mediate both protein–protein and protein–RNA interactions via the nuclear transport factor 2-like (NTF2L) domain and RNA-binding domain (RBD) (5, 24–27). Two other SG component proteins, Caprin-1 and ubiquitin specific peptidase 10 (USP10), are vital in regulating SG formation via binding to the NTF2L domain of G3BPs (28, 29). Caprin-1 and USP10 have antagonistic effects on SGs: Caprin-1 promotes SGs, whereas USP10 suppresses SGs (5, 22, 23). These studies indicate that Caprin-1 acts through its RBD in the C terminus as a scaffold to enhance the binding of RNAs to the SG complex, thus promoting SG formation. In contrast, USP10 lacks an RBD and acts as a cap, preventing additional RNA binding to G3BPs and suppressing SG formation (23). With these processes substantiated in prior research, two questions arise: 1) What is the structure of G3BP1/Caprin-1/USP10 complexes? and 2) How do different domains of Caprin-1 contribute to the regulation of SGs?

In this study we identified the G3BP1-interacting motif (GIM) in Caprin-1 and determined the crystal structures of the NTF2L domain of G3BP1 (G3BP1 NTF2L) in complex with the Caprin-1 GIM and the USP10 GIM. The crystallographic results showed that the Caprin-1 GIM and USP10 GIM bound to the same hydrophobic pocket on G3BP1 NTF2L. Furthermore, both the Caprin-1 GIM and USP10 GIM suppressed liquid–liquid phase separation (LLPS) of G3BP1, leading us to examine the effect of other domains of Caprin-1 on LLPS. We found that the N-terminal domain (NTD) of Caprin-1 suppressed LLPS, whereas the C-terminal domain (CTD) of Caprin-1 induced spontaneous LLPS in vitro. Moreover, the Caprin-1 CTD promoted SG formation even in the absence of the arsenite (As) treatment, whereas the Caprin-1 NTD repressed SGs induced by As in cells. These results suggest a dual regulation model in which the NTD and CTD of Caprin-1 have opposing effects on G3BP1 phase transition and SG formation in vitro and in vivo, which we term as a “yin and yang” regulation.

Results

Structure Characterization of the G3BP1 NTF2L/Caprin-1 GIM and G3BP1 NTF2L/USP10 GIM Complexes.

To elucidate the mechanism by which Caprin-1 and USP10 regulate SG formation, we started with the structural biology of G3BP1/Caprin-1/USP10 complexes. Previous studies suggested that Caprin-1 interacted with G3BP1 NTF2L via residues 352 to 380 with a low affinity (30), and USP10 interacted with G3BP1 NTF2L via residues 1 to 40 (5) (Fig. 1A). We defined the minimal region of Caprin-1 required for the binding to G3BP1 NTF2L using NMR spectroscopy and isothermal titration calorimetry (ITC) assays. A truncated Caprin-1 (330 to 432) was expressed, purified, and subjected to NMR analysis in the absence and presence of G3BP1 NTF2L (residues 1 to 139) (Fig. 1B). On the ¹H-¹⁵N HSQC spectrum of Caprin-1, ∼40 ¹⁵N-HSQC peaks decreased or disappeared upon titrating with G3BP1 NTF2L, suggesting that ∼40 amino acids in Caprin-1 interacted with G3BP1 NTF2L. The Gly368 residue, which showed two peaks in the upfield ¹⁵N region on the HSQC spectrum due to proline (Pro369) cis-trans isomerization, disappeared completely upon titration with G3BP1 NTF2L (Fig. 1B). In addition, ITC measurement showed a dissociation constant between full-length G3BP1 and the full-length WT Caprin-1 at 0.68 μM, whereas the F372A mutation abolished the interaction (SI Appendix, Fig. S1 A and B). These results suggest that Gly368 and Phe372 are critical residues for the interaction between G3BP1 and Caprin-1.

Fig. 1.

GIM of Caprin-1 and USP10. (A) Schematic domain structure of G3BP1, Caprin-1, and USP10. (B) Two-dimensional ¹H-¹⁵N HSQC of Caprin-1_330–432 (blue) and Caprin-1_330–432 bound with G3BP1 NTF2L (red). Approximately 40 signals decreased or disappeared after the titration with G3BP1 NTF2L, suggesting ∼40 amino acids in Caprin-1 are involved in binding to G3BP1 NTF2L. Signals of Gly368 (outlined in green) completely disappeared upon titration. (C) ITC measurement of the binding affinity of Caprin-1_347–386 with G3BP1 NTF2L.

We further tested four peptides with varying lengths adjacent to Gly368 and Phe372 and measured their binding with G3BP1 NTF2L using ITC (Fig. 1C and SI Appendix, Fig. S1 C–E). The results revealed that the peptide with residues 347 to 386 in Caprin-1 has the highest affinity binding to G3BP1 NTF2L with a dissociation constant of 1.3 μM (Fig. 1C). Thus Caprin-1_347–386 was used for crystallographic studies. USP10 residues 1 to 40 were reported to suppress SG formation (5), and we measured the binding between USP10_1–40 and G3BP1 NTF2L using ITC with a dissociation constant at 10 μM (SI Appendix, Fig. S1F). Thus, USP10_1–40 was used for crystallographic studies.

The crystals of the complex of G3BP1 NTF2L/Caprin-1_347–386 and G3BP1 NTF2L/USP10_1–40 were cocrystallized and diffracted to 2.4 Å and 2.7 Å, respectively. The crystal structures of the two complexes were determined using molecular replacement (SI Appendix, Table S1) and the ribbon diagrams of the overall structures are shown in SI Appendix, Fig. S2. Each asymmetric unit of NTF2L/Caprin-1_347–386 (PDB ID code 7XHG) contains four NTF2L molecules, three of which are bound with a Caprin-1 peptide. Each asymmetric unit of G3BP1 NTF2L/USP10_1–40 (PDB ID code 7XHF) consists of two NTF2L molecules with two USP10 peptides (Fig. 2A and SI Appendix, Fig. S2). The G3BP1 NTF2L molecule is arranged as dimers and the overall structure is similar to the published structures (27, 29). Each NTF2L molecule is composed of three α-helices and five β-strands (Fig. 2A). Visible segments of Caprin-1 and USP10 correspond to residues 369 to 378 and residues 6 to 21, respectively (Fig. 2 B–D). The electron density maps of Caprin-1_369–378 and USP10_6–21 are shown in Fig. 2 E and F. Based on these structural results, Caprin-1_369–378 and USP10_6–21 are defined as the GIM of Caprin-1 and USP10, respectively.

Fig. 2.

Crystal structure of G3BP1 NTF2L in complex with Caprin-1_347–386 and USP10_1–40. (A) Ribbon diagram showing the overall structures of NTF2L/Caprin-1_347–386 and NTF2L/USP10_1–40 complexes. Two G3BP1 NTF2L molecules are shown in light blue and pale cyan. Caprin-1_347–386 and USP10_1–40 are shown in yellow and red, respectively. (B) Exploded view on the interaction interface of G3BP1 NTF2L with Caprin-1_347–386 and USP10_1–40. The two GIMs bind to the same hydrophobic pocket of G3BP1 NTF2L. Tyr370 and Phe372 residues in Caprin-1 adopt a nearly identical conformation as Tyr8 and Phe10 residues in USP10. (C and D) Interactions between G3BP1 NTF2L and Caprin-1_369–378 (C) and USP10_6–21 (D). Residues involved in the interactions are highlighted. The hydrogen bonds are shown as dotted lines. (E and F) Electron density of Caprin-1_369–378 and USP10_6–21 in the crystal structure.

Structural comparison of the NTF2L/Caprin-1_347–386 and NTF2L/USP10_1–40 revealed that the two peptides bind to the same hydrophobic pocket between helices αI and αII on G3BP1 NTF2L (Fig. 2 A and B) and both form hydrogen bonds with the residues N122, K123, F124, and R32 of G3BP1 (Fig. 2 B–D). Strikingly, Tyr370 and Phe372 residues in Caprin-1 adopt a nearly identical conformation as Tyr8 and Phe10 residues in USP10 (Fig. 2B), indicating that the conserved aromatic rings at these two positions are critical to the binding of G3BP1. The significance of Phe372 in the interaction revealed by the structural analysis is supported by the ITC data (SI Appendix, Fig. S1 A and B).

The Effect of the NTD and CTD of Caprin-1 on G3BP1 LLPS In Vitro.

We next examined how GIM and other domains (Fig. 3A) of Caprin-1 affected the LLPS of G3BP1 in vitro. The double-label fluorescence microscopy showed that, in the presence of total RNA isolated from U2OS cells (50 ng/μL), G3BP1 formed more and larger droplets in the presence of full-length Caprin-1 as compared to PBS (Fig. 3B), suggesting that Caprin-1 facilitated G3BP1 LLPS, as previously reported (5, 22, 24). Moreover, Caprin-1 (mCherry-labeled) and G3BP1 (GFP-labeled) were colocalized in the same droplets. The GIM of USP10 drastically reduced the number and sizes of G3BP1 droplets (Fig. 3B), suggesting that the inhibitory effect of USP10 on G3BP1 LLPS is likely caused by the interaction between USP10 GIM and G3BP1. The GIM of Caprin-1 also suppressed G3BP1 LLPS in a similar fashion as the GIM of USP10 (Fig. 3B), suggesting that Caprin-1 likely promotes G3BP1 LLPS via other mechanisms involving other domains. The GIM is located in the middle of the Caprin-1 sequence; thus, we decided to determine the effect of the NTD and CTD on G3BP1 LLPS.

Fig. 3.

The effect of different domains of Caprin-1 on the LLPS of G3BP1 in vitro. LLPS was monitored by droplets in the fluorescence microscopy images. (A) Diagram of Caprin-1 constructs used in the experiment. (B) Purified full-length G3BP1-EGFP was mixed with PBS control or mCherry-tagged full-length Caprin-1, Caprin-1 GIM, USP10-GIM, Caprin-1 NTD, NTD+GIM, CTD, or CTD+GIM in the presence of 50 ng/μL U2OS cellular total RNA. LLPS was monitored by fluorescence protein droplets. (Scale bars, 20 μm.) (C) G3BP1 (50 μM) was incubated with untagged proteins and 50 ng/μL U2OS cellular total RNA. LLPS was measured by turbidity assay at 600 nm. Data are shown as mean ± SE from three independent experiments. A two-tailed unpaired t test was used for statistical analysis. ***P < 0.001; ****P < 0.0001. Results in B and C consistently support that GIM and the NTD suppressed G3BP1 LLPS, whereas the CTD and full-length Caprin-1 facilitated G3BP1 LLPS. (D) LLPS of purified full-length Caprin-1 at different protein and RNA concentrations as measured by the turbidity assay. LLPS of full-length Caprin-1 increased with increasing concentrations of protein and RNA concentrations. (E) LLPS of mCherry–Caprin-1 CTD and CTD+GIM alone with no other protein or U2OS cellular total RNA. (Scale bars, 20 μm.) (F) LLPS of G3BP1-GFP mixed with mCherry–Caprin-1 CTD and CTD+GIM in the absence of U2OS cellular total RNA. (Scale bars, 20 μm.)

The NTD domain of Caprin-1 reduced the number of droplets (i.e., suppressed G3BP1 LLPS in vitro). Caprin-1–NTD+GIM also suppressed G3BP1 LLPS and nearly inhibited any droplets. It is noted that neither the NTD nor NTD+GIM of Caprin-1 formed any droplets (mCherry channel in Fig. 3B). In contrast, both the CTD (379–709) and GIM+CTD (369–709) facilitated G3BP1 LLPS with significantly more droplets with overlapping mCherry and GFP signals, in a similar fashion to the full-length Caprin-1. Similar results were obtained when LLPS experiments were carried out using differential interference contrast (DIC) microscopy (SI Appendix, Fig. S3A). We also quantified LLPS using a previously reported turbidity assay (31). As compared to the PBS control, the NTD and NTD+GIM suppressed LLPS of G3BP1 (50 µM), whereas the CTD and CTD+GIM promoted G3BP1 LLPS (Fig. 3C). Similar results were obtained with 100 µM G3BP1 used in the turbidity assay (SI Appendix, Fig. S3B). The results from three independent assays consistently demonstrate that the NTD suppresses the LLPS of G3BP1, whereas the CTD facilitates LLPS.

With the opposing effects of the NTD and CTD of Caprin-1, we characterized LLPS of the full-length Caprin-1 without G3BP1. Different concentrations of purified full-length Caprin-1 and total RNA isolated from U2OS cells were mixed, and LLPS was measured using the turbidity assay. In the presence of RNAs, Caprin-1 underwent LLPS in a concentration-dependent fashion (Fig. 3D). Similar protein and RNA concentration-dependent LLPS of Caprin-1 were also observed using DIC microscopy (SI Appendix, Fig. S3 C and D).

Since the Caprin-1 CTD facilitated G3BP1 LLPS (Fig. 3 B and C) and the full-length Caprin-1 underwent LLPS in the absence of G3BP1 (Fig. 3D), we next examined whether the CTD alone could undergo LLPS. In the absence of exogenous RNA, both the CTD and GIM+CTD of Caprin-1 underwent spontaneous LLPS (Fig. 3E), further supporting that the Caprin-1 CTD is the main factor promoting G3BP1 LLPS. The CTD without GIM also facilitated G3BP1 LLPS and colocalized with G3BP1 in droplets in the presence of exogenous RNA (Fig. 3B), suggesting that the CTD likely interacted with G3BP1 indirectly through RNA and promoted G3BP1 LLPS. To test this hypothesis, we examined LLPS of the CTD with G3BP1 without exogenous RNA. Without the addition of total RNA isolated from U2OS cells, mCherry-CTD underwent robust LLPS with a large number of droplets, whereas G3BP1-GFP did not form any droplets (Fig. 3F). The results support that the CTD without GIM can facilitate LLPS of G3BP1 in an RNA-dependent fashion. We also tested CTD+GIM and G3BP1 without exogenous RNA and found that both CTD+GIM and G3BP1 colocalized in droplets (Fig. 3F), suggesting that CTD+GIM can facilitate G3BP1 LLPS via GIM in the absence of RNA.

The Effect of the NTD and CTD of Caprin-1 on the Formation of SGs in Cells.

We next examined the effect of different domains of Caprin-1 on SG formation in live cells. GFP-tagged Caprin-1 domains or GFP vector control was expressed in U2OS cells, and SG formation was visualized through G3BP1 staining (Fig. 4). Without As treatment, no G3BP1⁺ puncta were observed in cells expressing GFP vector or Caprin-1 constructs, with the exception of full-length Caprin-1 (SI Appendix, Fig. S4). Upon As treatment, the majority of cells expressing the GFP vector (∼90%) showed more than five G3BP1⁺ puncta (i.e., SGs), whereas ∼2% of cells showed no puncta (see a representative image in Fig. 4A and the quantitative result in Fig. 4B). Cells expressing wild-type full-length Caprin-1 showed ∼10% cells with no puncta, ∼30% cells with one to five G3BP1 puncta, and ∼60% cells with more than five puncta (Fig. 4 A and B). The expression of Caprin-1 GIM or Caprin-1 NTD showed a similar distribution: ∼10% cells with no puncta, ∼20% cells with one to five G3BP1 puncta, and ∼70% cells with more than five puncta (Fig. 4 A and B). Compared to the GFP vector control, the percentage of cells with no puncta significantly increased. Most strikingly, the expression of Caprin-1 NTD+GIM dramatically suppressed SG formation with ∼80% of transfected cells with no puncta, ∼20% with one to five puncta, and no cells with more than five puncta (Fig. 4 A and B). This is consistent with the in vitro results that the NTD and NTD+GIM suppressed G3BP1 LLPS (Fig. 3 B and C). As a control, expression of USP10 GIM also suppressed SG formation: ∼35% cells with no puncta, ∼40% cells with one to five puncta, and ∼25% cells with more than five puncta (Fig. 4 A and B).

Fig. 4.

The effect of the NTD and CTD of Caprin-1 on SG formation. (A) U2OS cells transiently transfected with the GFP vector control, GFP-tagged full-length Caprin-1, GIM, NTD, NTD+GIM, or USP10 GIM. All cells were treated with 500 μM As for 1 h, fixed, stained for G3BP1, and imaged with a confocal microscope. No G3BP1⁺ SGs were observed in cells transfected with NTD+GIM. Images of untreated cells transfected with the same set of Caprin-1 constructs are shown in SI Appendix, Fig. S4. (Scale bars, 20 μm.) (B) Quantification of SG formation based on the number of SGs in cells: no SG; 1–5 SGs; and >5 SGs. Approximately 200 GFP⁺ cells were counted for each construct. Boxplot presents the percentages of cells of each level of SG formation from three independent experiments. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. (C) U2OS cells were transiently transfected with GFP-tagged full-length Caprin-1, CTD, or GIM+CTD and subjected to fluorescence microscopy. SGs were observed in all three cell populations without As treatment. (Scale bars, 20 μm.) (D) CAPRIN1 knockout U2OS cells were transiently transfected with GFP-tagged full-length Caprin-1, CTD, or GIM+CTD. Cells without and with As treatment were fixed, stained for G3BP1, and imaged. (Scale bars, 20 μm.) (E) Quantification of SGs in CAPRIN1 knockout U2OS cells in the absence and presence of As treatment as in D. Boxplot represents the percentages of cells with SGs in each population from three independent experiments. *P ≤ 0.05; NS: not significant. The CTD and GIM+CTD promoted SG formation and did not respond to As treatment. (F) Diagram of full-length Caprin-1 constructs with three RGG motifs mutated to AGG. (G) Quantification of SG formation in cells in the presence of As treatment. Boxplot represents the percentages of cells in three categories: no SGs, 1–5 SGs, and >5 SGs. Approximately 100 to 200 GFP⁺ cells were counted for each construct and the results shown are from three independent experiments. *P ≤ 0.05; **P ≤ 0.01; ****P ≤ 0.0001; NS: not significant. Loss of RNA binding capability of full-length Caprin-1 reduced SG formation.

Since the CTD and CTD+GIM underwent spontaneous LLPS in vitro in the presence and absence of RNA (Fig. 3 B and C and E and F), we hypothesize that the CTD would promote SG formation in live cells. GFP-tagged full-length Caprin-1 nucleated G3BP1⁺ SGs in cells without As treatment (Fig. 4C). Similarly, overexpression of GFP-tagged Caprin-1–GIM+CTD or Caprin-1–CTD alone induced the SG assembly without As treatment (Fig. 4C). Notably, all three GFP-tagged Caprin-1 proteins were colocalized with G3BP1 in SGs, even the CTD alone that lacks the GIM.

To eliminate the possible contribution of the endogenous Caprin-1 protein, we examined SG formation in CAPRIN1 knockout cells expressing full-length Caprin-1, GIM+CTD, or CTD (Fig. 4D). The quantitative results showed that ∼10% of cells expressing full-length Caprin-1 contained SGs. In response to As treatment, ∼30% of cells showed SGs (Fig. 4E), demonstrating a robust stress response. More interestingly, ∼40% of cells showed G3BP1⁺ SGs when the CAPRIN1 knockout cells expressed the GIM+CTD or CTD alone. No additional SGs were observed upon As treatment (Fig. 4E). The results suggest that the CTD of Caprin-1 induced spontaneous SG formation and did not respond to stress induced by As treatment. The spontaneous SG formation by CTD in CAPRIN1 knockout cells was consistent with the spontaneous LLPS by CTD observed earlier (Fig. 3). In summary, the CTD of Caprin-1 underwent LLPS and formed SGs spontaneously, likely due to its own low complexity sequence and RNA binding capability.

The arginine-glycine-glycine (RGG) repeat sequence has been reported as an RNA binding motif in many proteins involved in LLPS (32–34). To assess the significance of the three RGG repeats in the Caprin-1 CTD in facilitating SG formation, we mutated arginine to alanine in three RGG repeats (termed AGG) (Fig. 4F). Upon As treatment, cells expressing wild-type Caprin-1 showed ∼10% cells with no puncta, ∼30% cells with one to five G3BP1 puncta, and ∼60% cells with more than five puncta. In contrast, AGG mutation of Caprin-1 showed ∼30% cells with no puncta, ∼40% cells with one to five G3BP1 puncta, and ∼30% cells with more than five puncta (Fig. 4G). Representative images are shown in SI Appendix, Fig. S5. The RNA binding ability of the RGG motif is critical to SG formation.

Discussion

The goal of this study was to elucidate the mechanism by which Caprin-1 regulates SG formation using structural biology, biochemical, and cellular experiments. First, we identified the GIM in Caprin-1 responsible for direct protein–protein interaction by NMR and ITC (Fig. 1). Subsequently, we solved the crystal structures of the complex of G3BP1 NTF2L with Caprin-1 GIM and USP10 GIM, respectively (Fig. 2). A comparison of the two structures revealed that Caprin-1_369–378 and USP10_6–21 occupy the same hydrophobic pocket of G3BP1 with Tyr370 and Phe372 residues in Caprin-1 and Tyr8 and Phe10 residues in USP10 adopting an identical conformation (Fig. 2). Two previously published crystal structures also showed that the same hydrophobic pocket of G3BP1 interacted with nsP3_449–471 and SARS-CoV-2 N_1–25 (27, 35). Moreover, Caprin-1 GIM and G3BP1 NTF2L domains are the primary direct protein–protein interaction sites between Caprin-1 and G3BP1 because: 1) the binding affinity of Caprin-1_347–386 and G3BP1 NTF2L domains (Fig. 1C) was approximately half of that between full-length Caprin-1 and G3BP1 proteins (SI Appendix, Fig. S1A), and 2) the F372A mutation completely abolished the interaction between two full-length proteins (SI Appendix, Fig. S1B).

We also determined that Caprin-1 GIM suppressed G3BP1 LLPS in vitro (Fig. 3B) in a similar fashion to USP10 GIM, suggesting that other domains of Caprin-1 contribute to its regulation of G3BP1 LLPS and SG formation. The results led us to examine the effect of the NTD and CTD of Caprin-1 on SG formation. Strikingly, the NTD+GIM of Caprin-1 almost completely inhibited the G3BP1 LLPS in vitro (Fig. 3). Consistently, the expression of NTD+GIM of Caprin-1 dramatically inhibited As-induced SG formation in cells (Fig. 4). It is likely that the NTD of Caprin-1 affects G3BP1-mediated SG formation by influencing the protein–protein and protein–RNA interaction of G3BP1.

In contrast, the CTD of Caprin-1 underwent spontaneous LLPS in vitro (Fig. 3 B and C), likely due to the low complexity sequence and RNA binding motifs in the CTD. Similarly, the Caprin-1 CTD nucleated SG formation both in wild-type and CAPRIN1 knockout cells without As treatment (Fig. 4 C and D). The SG-promoting effect was independent of Caprin-1 GIM, suggesting that the self-assembly of the Caprin-1 CTD likely recruits G3BP1 indirectly in an RNA-mediated fashion to form SGs. This study showed that the Caprin-1 CTD facilitated G3BP1 LLPS in the presence of RNA (Fig. 3B) but failed to do so in the absence of RNA (Fig. 3F). In addition, the RGG mutation in the CTD reduced SG formation (Fig. 4G), supporting the significance of RNA-mediated interaction between Caprin-1 CTD and G3BP1. Similar RNA-dependent assembly of SGs was previously observed (24, 34). The results also suggest that, in addition to the direct protein–protein interaction via GIM, Caprin-1 can also interact with G3BP1 via RNA-dependent interaction through the CTD.

Unlike the CTD that formed spontaneous SGs at similar levels with and without As treatment, the full-length Caprin-1 responded to As treatment, with SG-containing cells increasing approximately three fold in the presence of As (Fig. 4E). We propose that the NTD may interact with the CTD and mask the spontaneous SG-promoting effect of the CTD under physiological conditions without stress. In the presence of stress, it is conceivable that the Caprin-1 CTD does not interact with the NTD anymore, but rather interacts with G3BP1 in an RNA-dependent manner and promotes SG formation to enable a robust stress response. Based on this model, in the presence of stress the CTD is a major factor driving SG formation with its RNA-mediated self-assembly, and we predicted that reducing the RNA binding ability of the CTD would reduce SG formation. Mutating the RNA-binding RGG motifs dampened the stress-induced SG formation of full-length Caprin-1 protein in response to As treatment (Fig. 4G), supporting the model. The opposing effect of the NTD and CTD resembles a yin and yang mechanism for the full-length Caprin-1 to regulate SGs under various conditions. It is noted that the structure of full-length Caprin-1 is yet to be determined to better understand how the NTD and CTD interact. It is also unknown how stress signals are transduced to Caprin-1 to induce conformational changes to allow the CTD to dissociate from the NTD. It has been reported that protein translation is halted under stress conditions and RNA levels increase (20, 24); thus, RNA can potentially play an important role in modulating the NTD/CTD interaction as well as the Caprin-1/G3BP1 interaction. Future studies are needed to elucidate how the Caprin-1 NTD and CTD interplay under physiological and stress conditions.

The proposed dual-regulation model of Caprin-1 is illustrated in Fig. 5. The NTD of Caprin-1 binds to G3BP1 through GIM and suppresses G3BP1 LLPS and SG formation. The CTD provides the primary driving force of SG formation by intrinsic self-assembly, RNA binding ability, and RNA-dependent indirect interaction with G3BP1. The yin and yang interplay of the NTD and CTD provides a mechanism for fine-tuning the overall effect of the full-length Caprin-1 on SG formation under different physiological and pathological conditions. Maintaining a delicate balance of SG dynamics under physiological conditions is critical. Dysregulation of SGs can contribute to various human diseases in a complex fashion (14–17, 20, 21). The mechanistic insights and the structural details revealed in this study enable us to better modulate SG dynamics as potential therapeutic strategies for disease treatment.

Fig. 5.

Model of yin and yang dual function of the NTD and CTD of Caprin-1. The crystal structure of Caprin-1 GIM and G3BP1 NTF2L was determined. The NTD of Caprin-1 suppressed G3BP1 LLPS in vitro and G3BP1-mediated SG formation in cells (Left). The C-terminal domain underwent spontaneous LLPS in vitro and SG formation in cells (Right). The full-length Caprin-1 formed basal levels of SGs without stress (Left) and showed a robust response to As treatment with higher levels of SGs (Right). This stress-induced SG formation was suppressed by the NTD of Caprin-1 (Left). The CTD promoted SG formation and showed high levels of SGs even without stress, thus showing no response to stress (Right).

Materials and Methods

Detailed materials and methods are described in the SI Appendix, Materials and Methods. Protocols are briefly described below.

Cloning, Protein Expression, and Purification.

Recombinant DNA encoding full-length G3BP1 (Human, UniProt: Q13283), USP10 (Human, UniProt: Q14694), Caprin-1 (Human, UniProt: Q14444), and various mutations and domains were generated using standard methods. Detailed information on strains and plasmids used in this study is listed in SI Appendix, Table S2. Recombinant proteins were expressed and purified as previously reported (36).

NMR Titration and ITC.

¹⁵N-labeled Caprin-1_330–432 was produced in an M9 medium containing ¹⁵NH₄Cl as the sole nitrogen source. The ¹H-¹⁵N HSQC spectra of Caprin-1_330–432 and titration with unlabeled G3BP1 NTF2L were recorded at room temperature on a Bruker 600 MHz NMR spectrometer.

ITC measurements were carried out at room temperature using an ITC 200 (MicroCal). PEAQ-ITC (MicroCal), or Program Origin 7.0 was used to analyze the titration data and the curves were fitted by the one-site binding model.

Crystallization and Structure Determination of NTF2L/Caprin-1_347–386 and NTF2L/USP10_1–40.

The G3BP1 NTF2L/Caprin-1_347–386 and G3BP1 NTF2L/USP10_1–40 complexes were generated and subjected to crystallization using the sessile drop vapor-diffusion method at 16 °C (37). The diffraction data were collected at the wavelength of 0.979 Å in BL17U1 at the Shanghai Synchrotron Radiation Facility (SSRF China) (38) and processed using the HKL2000 package (39). The two complex structures were solved by Phaser Molecular Replacement (40) in the CCP4 suite (41) with the G3BP1 NTF2L structure (PDB ID code 4FCJ) as the searching model. Refmac5 was used for structure refinement (42). Coot was used for model building (43). Further manual model adjustment and refinement were carried out by Coot. All structure figures were prepared using PyMol (https://pymol.org/2/).

LLPS.

LLPS experiments were performed in vitro using three assays modified from previously published protocols (22, 24, 31). For the double-fluorescence microscopy assay (24), purified recombinant mCherry-tagged Caprin-1 or individual domains or mCherry-USP10 GIM was incubated with full-length G3BP1-EGFP in the presence of U2OS cellular total RNA. LLPS droplet formation was observed under a Zeiss LSM 880 laser-scanning confocal microscope. DIC microscopy was employed to observe droplet formation (22) when untagged Caprin-1, USP10, and G3BP1 proteins/peptides were mixed with U2OS cellular total RNA. LLPS was also measured by turbidity assay at 600 nm, as previously reported (31), using Nanodrop.

Cell Culture, Transfection, Immunofluorescence Imaging, and SG Quantification.

Cell culture, transfection, immunostaining, and confocal microscopy were performed as previously published (44). Images were captured using a Zeiss LSM 880 laser-scanning confocal microscope with a 25× or 40× oil objective, or a Nikon A1 confocal microscope with a 60× oil objective. SGs were quantified based on the percentage of GFP⁺ cells containing SGs and the number of SGs in cells.

Data, Materials, and Software Availability

The atomic coordinates have been deposited in the Protein Data Bank, www.wwpdb.org (PDB ID codes 7XHG (45) and 7XHF (46)).