Starvation-induced proteasome assemblies in the nucleus link amino acid supply to apoptosis

Abstract

Eukaryotic cells have evolved highly orchestrated protein catabolic machineries responsible for the timely and selective disposal of proteins and organelles, thereby ensuring amino acid recycling. However, how protein degradation is coordinated with amino acid supply and protein synthesis has remained largely elusive. Here we show that the mammalian proteasome undergoes liquid-liquid phase separation in the nucleus upon amino acid deprivation. We termed these proteasome condensates SIPAN (Starvation-Induced Proteasome Assemblies in the Nucleus) and show that these are a common response of mammalian cells to amino acid deprivation. SIPAN undergo fusion events, rapidly exchange proteasome particles with the surrounding milieu and quickly dissolve following amino acid replenishment. We further show that: (i) SIPAN contain K48-conjugated ubiquitin, (ii) proteasome inhibition accelerates SIPAN formation, (iii) deubiquitinase inhibition prevents SIPAN resolution and (iv) RAD23B proteasome shuttling factor is required for SIPAN formation. Finally, SIPAN formation is associated with decreased cell survival and p53-mediated apoptosis, which might contribute to tissue fitness in diverse pathophysiological conditions.

Fig. 1. Nutrient starvation induces the formation of proteasome foci in the nucleus of mammalian cells.

a Schematic representation of the mammalian proteasome composed by the 20S catalytic particle (CP) and 19S or 11S regulatory particles (RP). b Protein levels of proteasome components and other factors following nutrient deprivation in IMR90 cells. Cells were incubated in HBSS solution and harvested at different time points for immunoblotting. 4EBP1 phosphorylation is included as a control for starvation (representative from 2 independent experiments). c IMR90 and HCT116 cells form nuclear foci following nutrient deprivation. IMR90 and HCT116 cells were incubated in HBSS for 6 hrs or 8 hrs, respectively, and harvested for immunostaining (representative from 3 independent experiments). d Fractionation of IMR90 and immunoblotting for components of the proteasome. IMR90 cells were fractionated by hypotonic lysis and cytoplasmic and nuclear fractions were used for immunoblotting. LDH and PARP1 proteins were detected as markers of cytoplasmic and nuclear fractions, respectively (representative from 2 independent experiments). e Fractionation of HCT116 cells and immunoblotting for components of the proteasome. HCT116 was fractionated by quick lysis with 0.1 % NP-40 detergent and nuclear and cytoplasmic fractions were obtained. The wash fraction corresponds to resuspension of the nuclear pellet in detergent-free buffer followed by an additional centrifugation. LDH and PARP1 were detected as markers of cytoplasmic and nuclear fractions, respectively (representative from 2 independent experiments). f Subcellular localization of PSMB4-GFP following incubation of IMR90 cells in HBSS. IMR90 cells were transduced with lentivirus particles expressing PSMB4-GFP. Following four days post-infection, cells were incubated in HBSS for 6 hrs and harvested for fluorescence microscopy (representative from 3 independent experiments).

Fig. 2. Proteasome integrity is required for foci formation upon nutrient deprivation.

a Immunostaining of proteasome components in IMR90 human fibroblasts showing that component of the CP and RP co-localize in nuclear foci following 6 hrs incubation in HBSS. The green (endogenous PSMD11 or PSMB7), red (endogenous PSMB1, PSMB2, PSMB4, or PSMD4), and blue (DAPI) signals were merged to indicate foci localization within the nucleus. The inset at the bottom right of each picture corresponds to a magnification of a small portion of the nucleus as indicated. Graphs in the bottom represent relative signal intensity of the indicated line for each component of the proteasome (representative from 3 independent experiments). b siRNA-mediated depletion of components of the RP or CP abolishes PSMD7 or PSMB4 foci formation. IMR90 cells were transfected twice with the indicated siRNAs. After three days, cells were treated with HBSS for 6 hrs and harvested for immunostaining of endogenous PSMB4 and PSMD7 proteins. The images at the bottom represent magnification of the merge as indicated by white arrows (representative from 3 independent experiments). c Cell counts of PSMD4, PSMB4, or PSMD7 foci following depletion of individual components of the proteasome (n = 3 independent experiments). d Immunostaining of PSME3 in IMR90 cells showing that this subunit of the 11S RP localizes with PSMD7 in nuclear foci following 6 hrs incubation in HBSS (representative from 5 independent experiments). e siRNA-mediated depletion of components of the RP or CP abolishes PSME3 foci formation. IMR90 cells were transfected with siRNAs of components of the RP or CP. After 3 days, cells were incubated in HBSS for 6 hrs and harvested for immunostaining of PSME3 (n = 3 independent experiments). f Depletion of PSME3 by siRNA does not affect PSMD4 or PSMB4 foci formation. IMR90 cells were transfected with siRNAs of components of the RP or CP. After 3 days, cells were incubated in HBSS for 6 hrs and harvested for immunostaining of PSME3 (n = 3 independent experiments). Data in c, e, f represent mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant; 2-sided unpaired Student’s t-test with Welch correction (c, e, f). Source data are provided as a Source data file.

Fig. 3. SIPAN are induced by metabolic stress and do not correspond to previously known nuclear structures.

a PSMD11 or PSMD4 proteasome foci do not correspond to any known nuclear foci, structure or bodies including PML bodies (PML staining), nuclei (Fibrillarin staining), nuclear speckles (SC35 staining) or DNA double strand break foci (53BP1 staining). Stainings were conducted following 6 hrs incubation in HBSS (representative from 3 independent experiments). b Confocal microscopy showing PSMB4-GFP foci in low-density chromatin regions (representative from 3 independent experiments). c Transmission electronic microscopy in conjunction with colloidal gold-based immunodetection of PSMD4. Following incubation of IMR90 cells in HBSS for 6 hrs, cells were fixed and processed for immunodetection and electronic microscopy analysis. PSMD4 condensates are membrane-less and located in regions with reduced chromatin density (representative from 2 independent experiments). d Ribosomal proteins do not co-localize with proteasome foci. Merges from Supplementary Fig. 3b–d. Graphs at the right represent relative signal intensity of the indicated line for each component of proteasome (representative from 2 independent experiments). e, f Proteasome foci are not observed in response to other stress conditions. e IMR90 cells were treated with various chemical or physical agents and endogenous PSMD4 was detected by immunostaining at the indicated times (n = 3 independent experiments). f Western blot analysis of proteins following treatments of cells shown in (e) to ensure treatment efficacy (representative from 2 independent experiments). Data in (e) represent the mean ± SD. Source data are provided as a Source data file.

Fig. 4. SIPAN are a general phenomenon common to many mammalian cell types.

a, b Immunostaining of endogenous PSMD4 or PSMD7 in diverse cell types, showing that these proteins localize in nuclear foci following incubation in HBSS. SIPAN are observed in normal primary cells (e.g., HDLF: primary human lung fibroblasts, HUVEC: human endothelial cells, mESC: mouse embryonic stem cell); immortalized cells (e.g., NIH3T3: mouse embryonic fibroblasts, C2C12: mouse myoblast cell line, 3T3L1: mouse preadipocytes); transformed (e.g., Cos7: Transformed monkey kidney cells, RAW264.7: Abelson murine leukemia virus transformed macrophage) and tumoral cells (MCF7: human breast cancer, H1299: human non-small cell lung cancer). c Immunostaining of PSMD7 in mouse pre-adipocytes 3T3L1 and differentiated adipocytes showing that this protein localizes in nuclear foci following incubation in HBSS. Right panel, Oil Red O staining was conducted to control for adipocyte differentiation (representative images from 3 independent experiments).

Fig. 5. Rapid dynamics of SIPAN assembly and resolution.

a–c Kinetics of SIPAN formation in IMR90 fibroblasts. a Cells were incubated in HBSS and harvested for immunostaining with PSMD4 antibody. Cells with more than 10 foci were counted (n = 3 independent experiments). b Signal intensity of SIPAN. Images from control and starved cells were used to estimate SIPAN signals (representative from 3 independent experiments). Arbitrary units (arb. units). c Quantification of the number of SIPAN per cell at different time points post-starvation (representative from 3 independent experiments). d Quantification of the size of the traced PSMB4-GFP foci (n = 2 independent experiments). e Time lapse from live-cell imaging indicating SIPAN fusion in vivo. IMR90 cells expressing PSMB4-GFP were incubated in HBSS and used for live imaging. Two fusion events are indicated by arrow of different colors (representative from 3 independent experiments). f Determination of the frequency of PSMB4-GFP foci fusion events over 1 hr (representative from 2 independent experiments). g Comparison of the average intensity of PSMB4-GFP foci in the corresponding cells, before and after fusion events. Note the increase in average intensity after fusion (representative from 2 independent experiments). h A representative image of a 10-min mobility trace of a PSMB4-GFP focus (representative from 2 independent experiments). i A mean-square displacement measurement plot of PSMB4-GFP foci is represented (representative from 2 independent experiments). j Quantification of the distance traveled by PSMB4-GFP foci over 10 min in two independent experiments. Each data point in the distribution plot represents a traced focus (n = 2 independent experiments). k–m SIPAN are reversible. SIPAN formation is induced for 8 hrs and then IMR90 cells were replenished with normal culture medium and harvested for immunostaining. k Cells with more than 10 foci were counted (n = 3 independent experiments). l Signal intensity of SIPAN from nutrient-starved cells and nutrient-replenished cells (representative from 3 independent experiments). m Cell nucleus showing dissipation of PSMD4 signals after medium replenishment (representative from 3 independent experiments). n Resolution of PSME3 foci after nutrient replenishment. IMR90 were incubated 6 hrs with HBSS and then incubated with culture media for 1 hr (representative from 3 independent experiments). o VCP chaperone is not required for SIPAN resolution. IMR90 cells were treated with HBSS solution for 6 hrs and then treated with various inhibitors at the indicated concentrations in complete medium for 1 hr (n = 3 independent experiments). Cells with more than 10 foci are counted. Data represent mean ± SD (a, d, f, g, i, j, k,o) or median with interquartile range for one representative experiment (b, c, l). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant; one-way ANOVA with Holm-Sidak’s (a, k) or Kruskal–Wallis test with Dunn’s test (b, c, l) or 2-sided unpaired Student t-test (g, o). Source data are provided as a Source data file.

Fig. 6. SIPAN are highly dynamic.

a SIPAN dissipate rapidly following incubation with very low concentration of Triton X-100 detergent. IMR90 cells with preformed proteasome foci were treated with Triton X-100 in HBSS for 2 min and then harvested for PSMD4 immunostaining. For 53BP1 foci, IMR90 cells were treated with ionizing radiations for 4 hrs and then used for detergent treatment in HBSS for 2 min before immunostaining. b Time lapse from live-cell imaging indicating SIPAN dissipation following incubation with 1,6-hexanediol. c SIPAN dissipate following incubation in hypotonic buffers. IMR90 cells expressing PSMB4-GFP were incubated in HBSS for 6 hrs and then treated as indicated before fixation and fluorescence microscopy. d Time lapse from live-cell imaging indicating SIPAN dissipation and recovery following incubation in Tris 10 mM pH 7.3 followed by NaCl 200 mM in Tris 10 mM pH 7.3, respectively. IMR90 cells expressing PSMB4-GFP were incubated in HBSS and used for live-imaging. e Fluorescence recovery after photobleaching (FRAP) of SIPAN. Bleaching of PSMB4-GFP foci indicate that SIPAN appear rapidly in their original foci following bleaching, while low or no recovery of fluorescence was detected for 53BP1 foci or histone H2A nuclear domains. The red arrows indicate the time of laser bleaching (graphs are from 3 independent experiments and the images are representative from 3 independent experiments). Source data are provided as a Source data file.

Fig. 7. RAD23B is required for SIPAN formation.

a Levels of ubiquitin and conjugated ubiquitin following nutrient deprivation in IMR90 cells. Cells were incubated in HBSS solution and harvested at different time points for immunoblotting with anti-ubiquitin or anti-conjugated ubiquitin FK2 antibodies (representative from 2 independent experiments). b, c Ubiquitin and K48-conjugated ubiquitin co-localize with SIPAN. Immunostaining of PSMD11 or PSMD4 and ubiquitin or K48 ubiquitin chains following nutrient starvation. IMR90 cells were incubated in HBSS for 6 hrs and harvested for immunostaining as indicated (representative from 3 independent experiments). d MG132 treatment increase SIPAN intensity and promote their formation. IMR90 cells were incubated in HBSS for 3 hrs in the presence or absence of MG132, and cells were harvested for immunostaining (representative from 3 independent experiments). Foci per cell (e) and foci intensity (f) were measured and represented by violin plot (representative from 3 independent experiments). g Schematic representation of treatment experiments with TAK-243, MG132, Bortezomib (BTZ) and b-AP15 inhibitors. h E1 ubiquitin-activating enzyme is required for SIPAN formation. Cells were incubated with TAK-243 E1 inhibitor for 6 hrs in HBSS (n = 3 independent experiments). i Continuous ubiquitination is required for SIPAN stability. Pre-formed SIPAN were treated with TAK-243 in HBSS for 1 hr (n = 3 independent experiments). j Ubiquitination is not required for SIPAN resolution. IMR90 cells were treated with TAK-243 for 1 hr in complete media (n = 3 independent experiments). k Deubiquitination is required for SIPAN resolution. Cells were incubated in HBSS solution for 6 hrs and then treated with b-AP15 deubiquitinase inhibitor, MG132, or BTZ in normal culture medium for 1 hr (n = 3 independent experiments). l RAD23B and other ubiquitin receptors are required for SIPAN formation. Following siRNA depletion of several ubiquitin-binding proteins and shuttling factors, IMR90 cells were incubated in HBSS for 6 hrs and harvested for immunostaining for SIPAN formation (n = 3 independent experiments). m Validation of RAD23B using additional siRNAs. Following siRNA transfection, IMR90 cells were incubated in HBSS for 6 hrs and harvested for immunostaining for SIPAN formation (representative from 3 independent experiments). n RAD23B is assembled in SIPAN following nutrient starvation. IMR90 cells were incubated in HBSS for 6 hrs and harvested for immunostaining for endogenous RAD23B and PSMD4 (representative from 3 independent experiments). o, p In yeast, RAD23 is important for PSGs formation under conditions of carbon depletion (n = 3 independent experiments). Data in graphs h, i, j, k, l, p represent the mean ± SD. 2-sided unpaired Student’s t-test (h, i, l, p). Data in e, f represent the median with interquartile range for one representative experiment. 2-sided Mann–Whitney test (e, f). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant. Source data are provided as a Source data file.

Fig. 8. RAD23B mediates SIPAN formation.

a Analysis of domains and intrinsically disordered regions of RAD23 proteins. Prediction of order/disorder propensity of human RAD23B (left panel, uniprot #P54727) and yeast RAD23 (right panel, uniprot #P32628) based on their protein sequences. Disorder scores were calculated using PONDR-FIT (green), VSL2B (blue), and VLXT (magenta). b, c RAD23B lacking critical domains failed to assemble in SIPAN. b Schematic representation of RAD23B mutants. c Following lentiviral infection, IMR90 cells were incubated in HBSS for 6 hrs and harvested for immunostaining for Myc-RAD23B (anti-Myc) and PSMD7. Right, Estimation of the number of cells with SIPAN is indicated (representative from 4 independent experiments). d–f RAD23B undergoes LLPS in vitro. e Purified RAD23B in 50 mM HEPES pH 7.2 and 100 mM NaCl was mixed with Ficoll 400 and phase separation was visually observed, as RAD23B solution become turbid immediately after solution mixing (representative from 3 independent experiments). f RAD23B droplets were observed by bright-field microscopy. A portion of RAD23B and Ficoll 400 mixture was added on a microscope slide and a coverslip was applied on the solution near the edge of another coverslip to create space and allow liquid movement (representative from 3 independent experiments). g Effects of Ficoll and protein concentration on His-RAD23B droplet formation. Different concentrations of His-RAD23B were mixed with various concentration of Ficoll (representative from 3 independent experiments). h RAD23B was mixed with PEG 6000 or Dextran 60,000–90,000 and was added on a microscope slide and covered with a coverslip. Crystal violet was used to stain the droplets (representative from 3 independent experiments). i Inhibition of weak hydrophobic interactions alters RAD23B droplet formation. His-RAD23B was mixed with different concentrations of 1,6-hexanediol and analyzed for droplet formation (representative from 3 independent experiments). j RAD23B droplet fusion events during phase separation in vitro. Fusion events were detected by microscopy (representative from 3 independent experiments). k Coomassie showing purification of RAD23B protein and its corresponding mutants. l His-RAD23B mutants lacking critical domains were analyzed for droplet formation. Fewer droplets are observed with UBA1/2 mutant of His-RAD23B (representative from 3 independent experiments). Data in c represent mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant; 2-sided unpaired Student’s t-test is used. Source data are provided as a Source data file.

Fig. 9. Exhaustion of non-essential amino acid is responsible for SIPAN formation.

a–c Addition of amino acids, but not glucose or pyruvate, prevents SIPAN formation in IMR90 cells. Cells were incubated with various nutrients in HBSS solution and harvested after 6 hrs for immunostaining (a) and count (b) (n = 3 independent experiments). c Measured foci intensity is represented as violin plots (representative from 3 independent experiments). d–f Availability of amino acids regulates SIPAN formation. Cells were incubated with various inhibitors in HBSS solution and harvested after 3 hrs or 6 hrs for immunostaining (d) and count (e). Inhibition of autophagy by chloroquine accelerates SIPAN formation while inhibition of protein synthesis by cycloheximide inhibits SIPAN formation. Blocking mTOR pathway, with Rapamycin or Torin 1, does not affect SIPAN formation following nutrient deprivation (n = 3 independent experiments). f Measured foci intensity is represented as violin plot (representative from 3 independent experiments). g Non-essential amino acids (NEAA) completely prevented SIPAN formation. IMR90 cells were incubated with individual amino acids in HBSS solution and harvested after 6 hrs for immunostaining (n = 3 independent experiments). h NEAA promote SIPAN resolution. SIPAN formation is induced and then cells were replenished with fresh medium containing individual amino acids and harvested after 2 hrs for immunostaining (n = 3 independent experiments). Red arrows represent NEAA (g, h). Data in b, c, e, f, g, h represent mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant; 2-sided unpaired Student’s t-test used in (b) and 2-sided ANOVA in (e). Source data are provided as a Source data file.

Fig. 10. Inhibition of SIPAN formation is associated with apoptosis.

a Proteasome is active under nutrient starvation. IMR90 cells were incubated in HBSS for 8 hrs and then treated with MG132 for 1 hr and harvested for immunoblotting (representative from 3 independent experiments). b SIPAN are catalytic active. IMR90 cells were treated with HBSS for 6 hrs and then incubated with 1 µM of Me4Bodipy-Ahx3-L3-VS proteasomal activity probe for 1 hr. Cells were then harvested for immunostaining (representative from 3 independent experiments). c–e Inhibition of RAD23B or PSME3 result in increased cell survival following nutrient deprivation. Three days following siRNA transfection, IMR90 cells were incubated in HBSS for 48 hrs and harvested for MTT viability assay (c), Western blotting (d), phase contrast imaging or FACS analysis (e) (n = 2 independent experiments). Red arrow in (e) represents subG1 apoptotic cell population. f Induction of pro-apoptotic factors during HBSS treatment. IMR90 cells were treated with HBSS and harvested at the indicated times for western blotting (representative from 3 independent experiments). g Depletion of RAD23B or PSME3 cells protects from cell death induced by nutrient starvation. IMR90 cells without RAD23B or PSME3 were treated with HBSS for 48 hrs and harvested for immunoblotting for pro-apoptotic proteins (representative from 3 independent experiments). h–j Depletion of NOXA protects from cell death induced by nutrient deprivation. IMR90 were depleted of NOXA and treated with HBSS for 48 hrs and harvested for western blotting (h), MTT assay (i) and phase contrast (j) (representative from 3 independent experiments). k Schematic representation of the procedure for the generation of IMR90 cells transformed or derived from tumors. l Primary, transformed or tumoral IMR90 cells were treated with HBSS for 8 hrs and harvested for immunostaining for PSMD4 (representative from 3 independent experiments). m, n Cells were treated with HBSS for 96 hrs and stained with Hoechst (DNA) and propidium iodide for imaging (m) and counting of dead cells (n) (n = 3 independent experiments). o Model of SIPAN formation and function. Amino acids exhaustion promotes foci formation in the nucleus with RAD23B as a driver. SIPAN formation is associated with cell death. Data in c, i, l, n represent the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant; 2-sided unpaired Student’s t-test used in (c, n). Source data are provided as a Source data file.