µMap proximity labeling in living cells reveals stress granule disassembly mechanisms

Abstract

Phase-separated condensates are membrane-less intracellular structures comprising dynamic protein interactions that organize essential biological processes. Understanding the composition and dynamics of these organelles advances our knowledge of cellular behaviors and disease pathologies related to granule dysregulation. In this study, we apply microenvironment mapping with a HaloTag-based platform (HaloMap) to characterize intracellular stress granule dynamics in living cells. After validating the robustness and sensitivity of this approach, we then profile the stress granule proteome throughout the formation and disassembly and under pharmacological perturbation. These experiments reveal several ubiquitin-related modulators, including the HECT (homologous to E6AP C terminus) E3 ligases ITCH and NEDD4L, as well as the ubiquitin receptor toll-interacting protein TOLLIP, as key mediators of granule disassembly. In addition, we identify an autophagy-related pathway that promotes granule clearance. Collectively, this work establishes a general photoproximity labeling approach for unraveling intracellular protein interactomes and uncovers previously unexplored regulatory mechanisms of stress granule dynamics.

Extended Data Fig. 1 |. Development and application of the HaloMap platform in stress granule profiling.

a, Genetic schematic of the HaloTag–G3BP1 construct, and immunofluorescence validation of the subcellular localization of the fusion protein and its capacity to form stress-induced granules under various stressors, including oxidative, osmotic and heat shock stress. Colocalization was observed for the HaloTag–G3BP1 protein with the endogenous G3BP1 proteins after 30-min 500 μM NaAsO₂ stress treatment (AS). No aggregation was observed under native states (NT). Scale bar, 10 μm. Stress granule/cytoplasm (SG/Cyto) area ratios are calculated based on three different areas per well across three biological replicates. Data are presented as mean values ± SD. ****P < 0.0001 (two-tailed Student’s t-test). b, Cell viability assay testing cell tolerance to Ir-PEG4-hexyl-Cl ligand of different concentrations and incubation times. Measurements are performed on three biological replicates. Data are presented as mean values ± SD. Minimal cellular toxicity was observed after HaloMap workflow incubation (5 μM, 1 hour). c, Schematic of the HaloTag–NES construct, and immunofluorescence images showing no granule aggregation of the fusion protein under stress. Similar results were observed in two independent repeats. Scale bar, 10 μm. d, Venn diagram comparing enriched stress granule proteins from different methods. A list of 27 proteins consistently appears in all datasets.

Extended Data Fig. 2 |. Temporal HaloMap profiling identifies a ubiquitin-mediated granule disassembly pathway.

a, Schematic of the HaloTag conjugated with TAMRA dye (red) for live-cell imaging. Stress granule/cytoplasm (SG/cyto) area ratios are calculated based on three biological replicates. NT, native states; AS, arsenite stress, D30, 30 min disassembly; D60, 60 min disassembly; D90, 90 min disassembly; D120, 120 min disassembly; D150, 150 min disassembly. Data are presented as mean values ± SEM. Representative images at indicated time points of granule kinetics from Supplementary Video. Scale bar, 10 μm. b, Label-free quantitative proteomic analysis of the granule interactome at stressed, 30 min, 60 min and 120 min post-removal of the stressor (two-tailed Student’s t-test). Enriched granule interactors (HaloTag–G3BP1 to HaloTag–NES, log₂(fold change) > 0.5, P-value < 0.05) at each time point were combined (n = 698) for heatmap clustering in Fig. 2c. Disassembly recruited interactors (n = 203) were subjected to gene ontology analysis. c, Gene ontology results (top 10) of the disassembly-recruited interactors. Arrows highlight terms related to translation. d, Western blot validation of TAK-243 (2 μM and 10 μM, 1 hour) in inhibiting active ubiquitination of HEK293 cells. e, Immunofluorescence images of G3BP1 staining (green) showing unaffected granule formation under TAK-243 treatment. Similar results were observed in three biological replicates. Scale bar, 20 μm.

Extended Data Fig. 3 |. HaloMap profiling of (+/−)TAK-243 granule proteome identifies key ubiquitin-related modulators.

a–d, Label-free quantitative proteomic analysis of the granule interactome comparing (+/−)TAK-243 at native state (NT, a), disassembly 30 min (D30, b), 60 min (D60, c) and 120 min (D120, d) post-removal of the stressor (two-tailed Student’s t-test). Candidate disassembly-associated ubiquitin ligases and receptors are annotated in each volcano plot. e, Domain features of the NEDD4 family E3 ligases (WWP2, ITCH, NEDD4 and NEDD4L) identified in HaloMap labeling, consisting of an N-terminal C2 membrane localization domain, four WW domains for substrate recognition and a characteristic C-terminal HECT domain for catalytic ubiquitin tagging. Heclin was used as a HECT E3 ligase-specific inhibitor for subsequent experiments.

Extended Data Fig. 4 |. HECT E3 ligases facilitate granule disassembly across different cell lines.

a, Immunofluorescence validation of G3BP1 staining (green) showing heclin treatment delayed stress granule disassembly in other cell lines including SHSY5Y and HeLa. Results were validated in two independent experiments. Scale bar, 20 μm. b, Immunofluorescence images of G3BP1 staining (green) showing MLN4924 inhibition of NEDDylation has no impact on granule kinetics. Scale bar, 20 μm. c, Western blot analysis of siRNA knockdown efficiency for WWP2, NEDD4 and NDFIP2. Similar results were observed in two independent repeats. d, Immunofluorescence images demonstrating unaffected stress granule disassembly under individual siRNA knockdown of WWP2, NEDD4 and NDFIP2 at 120 min of recovery. Combinatorial siRNA knockdown of NEDD4L and ITCH shows a similar level of delay in granule disassembly compared to individual siRNA knockdown in Fig. 4d. Scale bar, 20 μm. SG/Cyto area ratios are calculated based on four different areas per well across three biological replicates. Data are presented as mean values ± SD. ****P < 0.0001, ns, not significant (one-way ANOVA with Tukey’s test). e, siRNA knockdown of ITCH, NEDD4L and NDFIP1 delayed granule disassembly in SHSY5Y cells. Results were validated in two independent experiments. Similar results were observed in two independent repeats. Scale bar, 20 μm. f, Immunofluorescence images of G3BP1 staining (green) from three biological replicates showing unaffected granule formation under siRNA knockdown of ITCH, NEDD4L and NDFIP1. Scale bar, 20 μm.

Extended Data Fig. 5 |. HECT E3 ligases potentially modify K63-specific ubiquitin chains on granule proteins.

a, Label-free quantitative proteomic analysis of HEK293 TUBE pulldown (+/−)heclin at disassembly 30 min (D30; two-tailed Student’s t-test). Gene ontology (GO) analysis of heclin-impacted ubiquitinome (log₂(fold change) > 0.5, P-value < 0.05). Black arrows highlight GO terms of interest, including regulation of K63-linked poly-ubiquitin and autophagy. b, Heatmap of K63-specific global ubiquitinome changes (+/−) heclin across native (NT), stressed (AS) and disassembly 30-min (D30) stages from proteomic analysis of K63-TUBE pulldown experiment. A diagram shows the workflow for GO analysis of the heclin-affected stress-induced ubiquitinome (n = 181). Black arrows highlight top GO terms functionally relevant to stress granules. c, A list of stress granule proteins (n = 31) exhibiting impaired stress-induced K63-specific poly-ubiquitination under heclin inhibition of HECT E3 ligases. d, Immunofluorescence images showing unaffected localization of pan-ubiquitin and K48-specific ubiquitin signals in stress granules (+/−)heclin treatment. Scale bar, 10 μm. Line profile plots depict normalized fluorescence intensity values for G3BP1 (green) and specific ubiquitin intensity (red) along the white line highlighted in the magnified ROI from left to right, showing good alignment of the two signals.

Extended Data Fig. 6 |. Ubiquitin receptor TOLLIP and autophagy machinery mediate granule disassembly.

a, Western blot analysis of siRNA knockdown efficiency for VCP, HDAC6 and SQSTM1. b, Immunofluorescence images from five different areas per well of G3BP1 staining (green) demonstrating unaffected stress granule formation (AS) and disassembly under individual siRNA knockdown of VCP, HDAC6 and SQSTM1 at 120 min of recovery (D120). Scale bar, 20 μm. c, Representative immunofluorescence images showing impaired overlap of TOLLIP signals with G3BP1/stress granule signals at stressed state (AS) and 30 min of disassembly (D30) in cells treated with either TAK-243 (left) or heclin (right). Scale bar, 10 μm. Line profile plots depict normalized fluorescence intensities of G3BP1 (green) and TOLLIP (red) along the white line highlighted in the magnified ROI from left to right, showing the misalignment of peaks between the two signals. d, Immunofluorescence validation of CYTO-ID autophagy detection reagent signal (green) overlapping with granule signal (red) upon stress treatment. Hoechst nuclei staining, blue. Scale bar, 10 μm. Line intensity profiles depict normalized fluorescence intensity values for CYTO-ID autophagy detection (green) and G3BP1 (red) along the white line highlighted in the magnified ROI from top to bottom. For a, c and d, similar results were observed in two independent repeats.

Fig. 1 |. HaloMap accurately profiles the SG proteome in living cells.

a, Schematic representation of the HaloMap approach. HaloTag is genetically fused with the target protein, enabling self-alkylation with the PEG4-hexyl chloride Ir ligand. The Ir photocatalyst locally activates biotin–diazirine probes into reactive carbenes upon blue light irradiation for interactor labeling. b, HaloMap workflow for labeling the SG interactome through photocatalyst installation, washing, probe introduction and photolabeling. c, Immunofluorescence validation of labeling of HaloTag–G3BP1 (green) before and after stress treatment. Biotinylation (red) was detected with fluorescently labeled streptavidin. Similar results were observed in two independent repeats. Scale bar, 10 μm. d, Label-free quantitative proteomic analysis of the granule interactome (HaloTag–G3BP1) against a cytosolic control (HaloTag–NES) after a 30-min treatment with 500 μM NaAsO₂ (AS; two-tailed Student’s t-test). Three biological replicates were created for each condition. Known SG interactors are marked in purple. Top interactors are annotated. Orange points denote known cytosolic proteins identified from previous peroxidase-based proximity labeling. Three new SG components verified in e are marked black. e, Representative immunofluorescence images from three different areas per well show colocalization (arrowheads) of three new SG components (red) marked in the volcano plot (d) with SGs (G3BP1). Scale bar, 10 μm. f, Label-free quantitative proteomic analysis of the prestress interactome (HaloTag–G3BP1) against cytosolic control (HaloTag–NES) under NT conditions (two-tailed Student’s t-test). Three biological replicates were created for each condition. A high-fidelity prestress interactor list is annotated. NT, native.

Fig. 2 |. Time-resolved HaloMap profiling identifies ubiquitin as a key modification in granule disassembly.

a, SG formation and disassembly kinetics plotted via immunofluorescence analysis of G3BP1 (green). Scale bar, 20 μm. SG/cyto area ratios are calculated based on four different areas per well across three biological replicates. Data are presented as mean ± s.d. b, Schematic representation of the temporal HaloMap experiment to catalog the granule proteome during disassembly. Three biological replicates were created for each condition. c, Unsupervised-clustered heatmap depicting intensities of HaloTag–G3BP1 labeling proteins significantly enriched relative to cytosolic control (log₂(fold change) > 0.5, P < 0.05) throughout granule disassembly. Highlighted proteins related to the regulation of translation (n = 17) and ubiquitination/protein quality control (n = 12) are enriched during disassembly. d, Immunofluorescence staining of pan-ubiquitin-containing puncta (red) colocalized with G3BP1 (green) granule signals after stress treatment. Scale bar, 10 μm. e, Western blot analysis of polyubiquitination levels during stress and recovery in the global proteome (left) and HaloTag–G3BP1 HaloMap-labeled and enriched lysates (right). Normalized ubiquitination level denotes densitometric quantification of lane intensities. For d and e, similar results were observed in two independent repeats. f, Mechanism of TAK-243 inhibition of ubiquitin transfer through stabilization and inactivation of the UAE–ubiquitin complex. g, Experimental schematic to test ubiquitination inhibition on granule kinetics. TAK-243 (2 μM) was differentially added (marked in teal) in separate groups, and cells were fixed at 120 min recovery (D120) time point for microscopy validation. h, Immunofluorescence analysis of delayed disassembly if TAK-243 was applied during granule formation. Representative images are for G3BP1 staining (green) from cells at D120. Scale bar, 20 μm. SG/cyto area ratios are calculated based on four different areas per well across three biological replicates. Data are presented as mean ± s.d. ****P < 0.0001 and *P = 0.0361 (one-way ANOVA with Tukey’s test). Cyto, cytoplasm; PQC, protein quality control; Ub, ubiquitin.

Fig. 3 |. Granule interactome mapping (+/−)TAK-243 identifies candidate granule modulators.

a, Experimental workflow for HaloMap profiling of the granule proteome across prestress to late disassembly state with (+) and without (−)TAK-243. Three biological replicates were created for each condition. b, Label-free quantitative proteomic analysis of the granule interactome (+/−)TAK-243 after 500 μM NaAsO₂ 30-min treatment (AS; two-tailed Student’s t-test). Candidate granule modulators are annotated. c–e, Balloon plot visualization of enrichment level of candidate modulators across the following different stages: HECT ubiquitin E3 ligases and ligase adaptors under TAK-243 treatment (c), ubiquitin receptors in the absence of TAK-243 (d) and autophagy adaptors in the absence of TAK-243 (e). The color scheme represents log₂(fold change) enrichment level. Blue, preferentially enriched in granules (+)TAK-243 and red, preferentially enriched in granules (−)TAK-243. Larger size circles indicate P < 0.05 (two-tailed Student’s t-test).

Fig. 4 |. HECT family E3 ligases regulate SG disassembly through K63-linked polyubiquitination.

a, Representative images of G3BP1 (green) exhibiting delayed SG disassembly under heclin treatment at 120 min recovery. Scale bar, 20 μm. SG/cyto area ratios are calculated based on four different areas per well across three biological replicates. Data are presented as mean ± s.d. ***P = 0.0001 (two-tailed Student’s t-test). b, SG kinetics (+/−)heclin plotted via immunofluorescence. Representative images showing granule levels at different stages. Scale bar, 20 μm. The SG/cyto area ratios are calculated based on four different areas per well across three biological replicates. Data are presented as mean ± s.d. c, Western blot analysis of siRNA KD efficiency for ITCH, NEDD4L and NDFIP1. d, Representative images showing delayed granule disassembly under individual siRNA KD at 120 min of recovery. Scale bar, 20 μm. Quantification was performed on four different areas per well from three biological replicates. Data are presented as mean ± s.d. ***P_(Scr-ITCH) = 0.0004, ***P_(Scr-NEDD4L) = 0.0001 and ***P_(Scr-NDFIP1) = 0.0005 (one-way ANOVA with Tukey’s test). e, Representative immunofluorescence images from three different areas per well reveal SG (G3BP1, green) colocalization with expressed HA-ITCH and HA-NEDD4L proteins (red). Scale bar, 10 μm. f, Schematic representation of pan-ubiquitin and subsequent K63-specific ubiquitin TUBE pulldown assays. Three biological replicates were used for each condition. g, Immunofluorescence images showing decreased K63-ubiquitin staining (red) in SGs (green) with heclin treatment. Scale bar, 10 μm. Line profile plots depict normalized fluorescence intensity values for G3BP1 (green) and K63-ubiquitin (red) along the white line highlighted in the magnified region of interest (ROI) from left to right. The alignment of the two signals is impaired under heclin treatment. For c and g, similar results were observed in two independent repeats. HA, hemagglutinin tag.

Fig. 5 |. Ubiquitin receptor TOLLIP facilitates SG disassembly.

a, Western blot analysis of siRNA KD efficiency for TOLLIP. b, Immunofluorescence images of G3BP1 staining (green) showing unaffected granule formation but delayed SG disassembly under TOLLIP siRNA KD at 120 min of recovery. Scale bar, 20 μm. c, Immunofluorescence detection of delayed SG kinetics under combined heclin treatment and TOLLIP KD. Green fluorescent signals indicate G3BP1. Scale bar, 20 μm. Quantification was performed from five different areas per well across three biological replicates. Data are presented as mean ± s.d. d, Immunofluorescence images reveal colocalization of TOLLIP (red) in SGs (G3BP1, green). Scale bar, 10 μm. Line profile plots depict normalized fluorescence intensity values of G3BP1 (green) and TOLLIP (red) along the white line highlighted in the magnified ROI from left to right. For a, b and d, similar results were observed in two independent repeats.

Fig. 6 |. Autophagy-related pathways mediate SG dynamics.

a, Western blot analysis of LC3B activation upon stress treatment. Anti-LC3B antibody stained against both LC3B-I and LC3B-II. b, Immunofluorescence detection of autophagosome formation (green) with CYTO-ID autophagy detection reagent upon stress. Blue fluorescent signals, Hoechst nuclei staining. Scale bar, 20 μm. c, Immunofluorescence detection of LC3B cytosolic puncta (red) upon AS treatment colocalized with G3BP1 granule signals (green). Scale bar, 10 μm. Line profile plots depict the normalized fluorescence intensity of G3BP1 (green) and LC3B (red) along the white line highlighted in the magnified ROI from left to right. For a, b and c, similar results were observed in two independent repeats. d, Immunofluorescence images showing delayed SG disassembly under BafA1 treatment (100 nM, 3-h pre-incubation, maintained till fixation) at 120 min of recovery. Scale bar, 20 μm. Quantification was performed on four different areas per well across three biological replicates. Data are presented as mean ± s.d. ***P = 0.0007 (two-tailed Student’s t-test). e, Representative images showing smaller SG formation after AS treatment combined with the autophagy inducer Torin 1 (250 nM) or TAT-D11 (10 μM). Scale bar, 20 μm. Quantification of SG sizes was calculated from five different areas per well across three biological replicates. Error bars represent mean ± s.d. ****P < 0.0001 (one-way ANOVA with Tukey’s test). f, Representative images displaying faster granule disassembly at 90 min recovery (D90) with autophagy inducer Torin 1 and TAT-D11 treatment. Scale bar, 20 μm. Quantification was performed on four different areas per well across three biological replicates. Data are presented as mean ± s.d. ****P < 0.0001 (one-way ANOVA with Tukey’s test).