Distinct proteostasis circuits cooperate in nuclear and cytoplasmic protein quality control

Abstract

Protein misfolding is linked to a wide array of human disorders, including Alzheimer's disease, Parkinson's disease and type II diabetes^1,2. Protective cellular protein quality control (PQC) mechanisms have evolved to selectively recognize misfolded proteins and limit their toxic effects^3-9, thus contributing to the maintenance of the proteome (proteostasis). Here we examine how molecular chaperones and the ubiquitin-proteasome system cooperate to recognize and promote the clearance of soluble misfolded proteins. Using a panel of PQC substrates with distinct characteristics and localizations, we define distinct chaperone and ubiquitination circuitries that execute quality control in the cytoplasm and nucleus. In the cytoplasm, proteasomal degradation of misfolded proteins requires tagging with mixed lysine 48 (K48)- and lysine 11 (K11)-linked ubiquitin chains. A distinct combination of E3 ubiquitin ligases and specific chaperones is required to achieve each type of linkage-specific ubiquitination. In the nucleus, however, proteasomal degradation of misfolded proteins requires only K48-linked ubiquitin chains, and is thus independent of K11-specific ligases and chaperones. The distinct ubiquitin codes for nuclear and cytoplasmic PQC appear to be linked to the function of the ubiquilin protein Dsk2, which is specifically required to clear nuclear misfolded proteins. Our work defines the principles of cytoplasmic and nuclear PQC as distinct, involving combinatorial recognition by defined sets of cooperating chaperones and E3 ligases. A better understanding of how these organelle-specific PQC requirements implement proteome integrity has implications for our understanding of diseases linked to impaired protein clearance and proteostasis dysfunction.

Extended Figure 1 ∣. Puncta formation assay distinguishes between misfolded versus natively-folded proteins

a, WT cells expressing natively-folded Ubc9^WT-GFP or temperature-sensitive Ubc9^ts-GFP from a galactose-inducible promoter for 5-6 h at 30 °C were shifted to glucose-containing media for 1 h at 30 °C or 37 °C to shut off expression. Cells were fixed and imaged by fluorescence microscopy. 300 cells were counted per condition, and the percentage of cells with GFP-positive puncta is shown (mean ± SEM from 3 biologically independent experiments). Only cells expressing temperature-sensitive Ubc9^ts-GFP showed a statistically significant change in the percentage of puncta-positive cells compared with WT (one-way ANOVA followed by Dunnett’s multiple comparisons test, ****p < 0.0001). b, Deletion of individual E3 ligases does not increase puncta formation. Experiment performed as in a, but in strains with endogenous deletions of the genes denoted on the x-axis. E3 ligases that are previously implicated in PQC (and shown in Fig. 1d) are grouped to the right hand side. Bars represent mean ± SEM from 3 biologically independent experiments, with the exception of rad5, hel1, etp1, irc20, hel2, apc11, hrt1, tfb3, cdc24, prp19, upf1, upf3, itt1, and rad18 (where bars represent the mean from 2 biologically independent experiments), as well as WT (where bars represent the mean ± SEM from 7 biologically independent experiments). No strains showed statistically significant differences compared with WT by one-way ANOVA followed by Dunnett’s multiple comparisons test. c, Deleting certain pairs of E3 ligases increases misfolded protein stability. Cyclohexamide-chase followed by immunoblot to assess stability of GFP-VHL, CPY^‡-GFP. Ubc9^ts-TAP or Ubc9^wt-TAP in E3 ligase double deletion strains. For the WT + bortezomib condition, 50 μM bortezomib was added to the glucose-containing media 10 minutes before cyclohexamide treatment. Graphs represent densitometric quantification of bands relative to t = 0 timepoint (mean ± SEM from 3 biologically independent experiments). d, Multiple misfolded proteins are sequestered in the same subcellular location. Δubr1Δsan1 or Δdoa10Δhrd1 strains co-expressing VHL with temperature-sensitive Ubc9^ts (top) or natively-folded Ubc9^WT (bottom) from galactose-inducible promoters for 5-6 h at 30 °C were shifted to glucose-containing media for 1 h at 37 °C (for Ubc9^ts). Fluorescence microscopy images are representative of at least 100 cells from each of 3 biologically independent experiments. e, Deletion of certain pairs of E3 ligases increases puncta formation. Experiment performed as in a, but in strains with endogenous deletions of pairs of E3 ligase genes. The right hand panels represents experiments where cells were shifted to 37 °C for the 1 h of galactose shut-off. Bars represent mean ± SEM from 3 biologically independent experiments. Strains for which statistically significant differences were observed by one-way ANOVA followed by Dunnett’s multiple comparisons test compared with WT are indicated with the adjusted p-value, or **** for p < 0.0001. f, Overexpressing a single E3 ligase does not compensate for loss of others. Ubr1, San1 or Hrd1 were overexpressed alongside GFP-VHL in the indicated strains. The rest of the experiment was performed as in a. Bars represent mean ± SEM from 3 biologically independent experiments.

Extended Figure 2 ∣. San1 forms a complex with Doa10 but not with Hrd1.

a-b, San1-V5His6 co-immunoprecipitates with Doa10-GFP but not with Hrd1-GFP. Yeast cells co-expressing Doa10-GFP (a) or Hrd1-GFP (b) from their endogenous promoters with San1-V5His6 from a galactose-inducible promoter for 16 h were shifted to 37 °C for 1 h, and immediately lysed by cryo-grinding. Native complexes were immunoprecipitated with GFP-Trap®-MA nanobodies before immunoblotting with the indicated antibodies. Immunoblots are representative of 3 biologically independent experiments. c-d, Flag-Ubr1 does not co-immunoprecipitate with Doa10-GFP or Hrd1-GFP. Experiment was performed as above, but with cells expressing Flag-Ubr1 (from the constitutive ADH promoter) instead of San1-V5His6. Immunoblots are representative of 3 biologically independent experiments.

Extended Figure 3 ∣. K48 and K11 Ub linkages are reduced in Δubr1Δsan1 and Δdoa10Δhrd1 strains, respectively.

a, Schematic illustrating Ub linkage ELISA used to quantify Ub linkages. Flag-VHL from a yeast lysate was immunoprecipitated in an α-Flag-conjugated 96-well plate (4 wells used per sample), and incubated with antibodies against GFP (negative control), Flag, K11-Ub, or K48-Ub. Following incubation with a secondary antibody (α-rabbit-HRP), the strength of each signal was detected by electrochemiluminescence at 450 nm. To quantify the K11 or K48 Ub linkages on Flag-VHL, the α-K11 or α-K48 signal was subtracted of the negative control (α-GFP) and normalized to the total Flag-VHL signal for each sample. b, Ub linkage ELISA confirms that K48 and K11 Ub linkages are reduced on Flag-VHL in Δubr1Δsan1 and Δdoa10Δhrd1 strains, respectively. WT or E3 double deletion strains expressing Flag-VHL at 30 °C for 5-6 h were lysed after 1 h bortezomib treatment, also at 30 °C. Ub linkage ELISA was then performed as described in c. Bars represent Flag-normalized values from each strain (mean ± SEM from 3 biologically independent experiments) expressed as a proportion of the Flag-normalized WT values. Strains with statistically significant differences compared with WT by one-way ANOVA followed by Dunnett’s multiple comparisons test are indicated (****p < 0.001). c, GFP-VHL denaturing immunoprecipitation (8M Urea +1% SDS) followed by immunoblot for K48-Ub or K11-Ub in WT or E3 double deletion strains. Immunoblots representative of 3 independent experiments are shown. d, Relative amounts of K11 and K48 Ub linkages present on GFP-VHL in Δubr1Δsan1 or Δdoa10Δhrd1 strains compared with WT. WT or E3 double deletion strains expressing GFP-VHL at 30 °C for 5-6 h were lysed in denaturing conditions (8M Urea + 1% SDS) after 1 h bortezomib treatment, also at 30 °C. Ub linkage ELISA was then performed using GFP multiTrap® plates. Bars represent GFP-normalized values from each strain (mean ± SEM from 3 biologically independent experiments) expressed as a proportion of the GFP-normalized WT values. Strains for which statistically significant differences were observed by one-way ANOVA followed by Dunnett’s multiple comparisons test compared with WT are indicated with the adjusted p-value, or **** for p < 0.0001.

Extended Figure 4 ∣. K11-Ub linkages are not necessary for proteasomal degradation of all cytoplasmic substrates.

WT or Ub-K11R cells expressing stable Ub-M-GFP (a), the N-End Rule substrate Ub-R-GFP (b), the ubiquitin fusion degradation (UFD) substrate Ub^G76V-GFP (c) or GFP fused to the artificial degron CL1 (d) from galactose-inducible promoters for 5-6 h at 30 °C were shifted to glucose-containing media for 1 h at 30 °C or 37 °C to shut off expression. Cells were fixed and imaged by fluorescence microscopy. 300 cells were counted per condition, and the percentage of cells with GFP-positive puncta is shown (mean ± SEM from 3 biologically independent experiments). There was a statistically significant increase in puncta when GFP-CL1 (which contains a short amphipathic CL1 helix that could mimic a partially unfolded protein) was expressed in K11R-Ub cells compared with WT cells by one-way ANOVA followed by Dunnett’s multiple comparisons test (p = 0.0127). The differences for all other substrates were not significant (ns, p > 0.05).

Extended Figure 5 ∣. Misfolded VHL is modified with branched K11/K48 ubiquitin chains.

a-b, Both K11 and K48 Ub linkages are present on the same VHL molecule. a, Experiment was designed to determine if both K48 and K11 Ub linkages are present in the same VHL population. Sequential immunoprecipitation, first with a FLAG antibody, followed by a K11 or K48 ubiquitin antibody. The resultant negative control (no Flag: mock Flag + K11 or K48 immunoprecipitation with lysate from cells expressing GFP-VHL instead of Flag-VHL), Eluate and Flow-Through, in addition to samples with just the first Flag immunoprecipitation (Input), were subjected to SDS-PAGE and immunoblotted for the presence of the other Ub linkage (b). Immunoblots representative of 3 biologically independent experiments are shown. c, Bispecific K11/K48 Ub antibody designed to bind ubiquitin chains with K11 and K48 linkages branching off the same ubiquitin moiety. d, Misfolded VHL co-localises with K11/K48 Ub branched chains. WT cells expressing GFP-VHL from a galactose-inducible promoter for 5-6 h at 30 °C were shifted to glucose-containing media with 50 μM bortezomib for 1 h to shut off expression. Cells were fixed, spheroplasted and detergent-permeabilized before immunostaining with an antibody designed to recognize ubiquitin that had K11 and K48 linkages emanating from the same moiety (K11/K48). Confocal fluorescence microscopy images are representative of at least 100 cells from each of 3 biologically independent experiments. Scale bars = 2 μm. e, VHL is modified with branched K11/K48 Ub chains. GFP-VHL denaturing immunoprecipitation followed by immunoblot for K11/K48 Ub or GFP (VHL) in WT or E3 double deletion strains. Immunoblots representative of 3 biologically independent experiments are shown.

Extended Figure 6 ∣. Nuclear and cytoplasmic proteins require different PQC pathways for clearance.

a, NLS-GFP-VHL and NES-GFP-VHL form a single punctum in the nucleus or cytoplasm, respectively, upon proteasome inhibition. WT cells expressing NLS- or NES- GFP-VHL from a galactose-inducible promoter for 5-6 h at 30 °C were shifted to glucose-containing media with 50 μM bortezomib for 1 h at 30 °C to shut off GFP-VHL expression. Fixed and spheroplasted cells were immunostained for the nuclear pore complex protein Nsp1 prior to imaging by fluorescence microscopy. Representative cells from 3 biologically independent repeats are shown. b-d, Misfolded Luciferase^ts confined to the nucleus can be cleared by San1-mediated K48- ubiquitination. b, The increase in percentage of NLS- and NES- Luc^ts-GFP puncta-containing cells across the E3 ligase single and double deletion strains is similar to the pattern observed with NLS- and NES- GFP-VHL in Fig. 4. Percentage of cells (mean ± SEM from 3 biologically independent experiments, each with n = 300) containing NLS- or NES- Luc^ts-GFP puncta in WT, single or double deletion strains after 5-6 h expression of the protein at 30 °C followed by 1 h shut-off at 37 °C. Strains for which statistically significant differences were observed by one-way ANOVA followed by Dunnett’s multiple comparisons test compared with WT are indicated with the adjusted p-value, or **** for p < 0.0001). c, Misfolded nuclear Luciferase^ts has severely reduced K11-Ub ubiquitin linkages (****p < 0.0001 by one-way ANOVA followed by Dunnett’s multiple comparisons test). Ubiquitin linkage ELISA was performed on lysates of WT yeast expressing NLS-, NES- or unaltered GFP-Luciferase^ts at 37 °C as described in Fig. 2c, but in GFP-multiTrap® 96-well plates instead of α-FLAG-conjugated 96-well plates. α-FLAG was used instead of α-GFP as the ELISA negative control. Bars represent mean ± SEM from 3 biologically independent experiments. d, Misfolded Luciferase^ts confined to the nucleus does not require K11-Ub linkages for clearance. Experiment was performed as in b, but with yeast strains expressing WT or mutant K11R-Ub as their sole source of ubiquitin. 300 cells were counted per condition, and the percentage of cells with GFP-positive puncta are shown in (mean ± SEM from 3 biologically independent experiments). Only NES-GFP-Luciferase^ts had a statistically significant change in puncta-positive cells in the K11R strain when compared with WT (one-way ANOVA followed by Dunnett’s multiple comparisons test, ****p < 0.0001; ns = p > 0.05). e, VHL confined to the nucleus (NLS) or cytoplasm (NES) requires different chaperones for clearance. Experiment was performed as in b, but with the indicated chaperone deletion strains. Bars represent mean ± SEM from 3 biologically independent experiments. Strains for which statistically significant differences were observed compared with WT by one-way ANOVA followed by Dunnett’s multiple comparisons test are indicated with the adjusted p-value, or **** for p < 0.0001.

Extended Figure 7 ∣. Mass spectrometry of VHL interactome identifies distinct PQC circuitries for nuclear and cytoplasmic VHL.

a, Triple SILAC-base mass spectrometry of VHL immunoprecipitates. WT yeast cells transfected with one of NLS-GFP-VHL, NES-GFP-VHL or Flag-VHL were grown overnight at 30 °C in raffinose-synthetic media supplemented with Light Lys-0, Heavy Lys-8, or Medium Lys-4, respectively. Growth of VHL was induced in galactose for 4-5 h before shut off in glucose for 90 min. 1.5 mg of protein from each of the three lysed samples were mixed prior to immunoprecipitation using GFP-TRAP_MA magnetic beads on-bead restriction digestion and peptide clean-up. Peptides were identified using LC-MS analysis prior to analysis using MaxQuant. b, Strong correlation between the four biological repeats. Raw intensities for Light (NLS-GFP-VHL, top), Heavy (NES-GFP-VHL, middle) and Medium (Flag-VHL Control, bottom) were log₁₀ transformed and plotted as scatter plot matrices. The Pearson correlation coefficient for each pairwise comparison is indicated, and the density distribution of intensities within for repeat is shown in the diagonal axis of the matrices. c, Enriched PQC proteins in NLS-GFP-VHL and NES-GFP-VHL interactomes. Normalized median Light/Medium (NLS-GFP-VHL) and Heavy/Medium (NES-GFP-VHL) SILAC ratios were log₂ transformed. Proteins with log₂(SILAC ratio) > 0.5 were considered as enriched, yielding 49 and 56 proteins for the NLS and NES interactomes, respectively. Enriched proteins known to play a role in PQC are shown. Both nuclear and cytoplasmic VHL share enrichments in proteasomal subunits, the Hsp70 chaperones Ssa1, Ssa2, Ssa4 and Ssb2, and the thioredoxins Trx1, Trx2 and Tsa1 (previously implicated in misfolded protein management). All enriched proteins are shown in Extended Table 1. d, Enriched PQC pathways in NLS-GFP-VHL and NES-GFP-VHL interactomes. The enriched proteins from each interactome (median values from four biologically independent experiments) were subjected to pathway analysis to search for enriched GO Terms, KEGG Pathways and PFAM Protein Domains in either interactome using the STRING database. Selected enriched PQC pathways are shown (p < 0.05 using Fisher’s exact test followed by Benjamini-Hochberg multiple testing correction).

Figure 1 ∣. Combined deletion of certain E3 ligases impairs misfolded protein clearance.

a, Clearance of proteins misfolded for different reasons involves conserved pathways, with initial recognition by Hsp70-Hsp90 family chaperones, ubiquitination by one or more E3 ligases, and targeting for proteasomal degradation. Blocking any step triggers misfolded protein sequestration into puncta. b-c, VHL and CPY^‡ form puncta upon proteasome inhibition. WT cells expressing galactose-inducible GFP-VHL or CPY^‡-GFP were shifted to glucose media with 50 μM bortezomib (Bz) or vehicle control (Ctrl) for 1 h to shut off expression. Fixed cells were imaged by fluorescence microscopy, c, % cells with GFP-VHL puncta. d, Deleting individual E3s implicated in PQC does not increase puncta formation. Experiment performed as in b, but in strains with endogenous deletions of E3s. e-f, Misfolded proteins are stabilized in Δubr1Δsan1 and Δdoa10Δhrd1 strains. Cycloheximide-chase and immunoblot to assess stability of GFP-VHL in E3 double deletion strains, or following 50 μm bortezomib treatment. f, Densitometric quantification of bands relative to t = 0 (mean ± SEM from 3 biologically independent experiments). g, Multiple misfolded proteins are sequestered in the same subcellular location. Experiment performed as in b, but in strains co-expressing VHL with CPY^‡. Images represent over 100 cells from each of 3 biologically independent experiments. Scale bars = 2 μm. h, Deleting certain E3 pairs increases puncta formation. Experiment performed as in b, but in strains with endogenous deletions of E3 pairs. On the right, cells were shifted to 37 °C for the shut-off. i, Misfolded protein clearance requires soluble E3s Ubr1 or San1 and membrane-bound E3s Doa10 or Hrd1. c, d, h, 300 cells were counted per condition. Bars represent mean ± SEM from 3 biologically independent experiments—except for WT in d (7 biologically independent experiments). Statistically significant differences vs. WT by one-way ANOVA + Dunnett’s multiple comparisons test are indicated (adjusted p-value, or ****p < 0.0001).

Figure 2 ∣. Cytoplasmic misfolded proteins are modified with both K11- and K48-linked ubiquitin chains.

a, VHL ubiquitination is impaired in Δubr1Δsan1 and Δdoa10Δhrd1 strains. Flag-VHL denaturing immunoprecipitation followed by immunoblot for ubiquitin (Ub) in WT or E3 double deletion strains. Deletion of the co-translational E3s LTN1 & HEL2 served as a control. b, K48 and K11 Ub linkages are reduced on Flag-VHL in Δubr1Δsan1 and Δdoa10Δhrd1 strains, respectively. Experiment performed as in a, but K48 or K11 linkage specific Ub antibodies were used for immunoblot. c, K48 and K11 Ub co-localise with GFP-VHL puncta. Experiment performed as in Fig. 1b. Fixed cells were spheroplasted and immunostained before imaging by confocal fluorescence microscopy. Images represent >100 cells from each of 3 independent experiments. Scale bars = 2 μm. d-f, VHL clearance is impaired in the absence of K11-Ub linkages. Cells co-expressing galactose-inducible GFP-VHL, and either WT or Lys-to-Arg mutant Ub (KnR) as their only ubiquitin source, were shifted to glucose media for 1 h to shut off expression. 300 cells were counted per condition. e, % cells with GFP-VHL puncta (mean ± SEM from 3 biologically independent experiments). Only K11R cells significantly altered the percentage of puncta-positive cells vs. WT (one-way ANOVA + Dunnett’s multiple comparisons test, ****p < 0.0001). f, Cycloheximide-chase and immunoblot to assess GFP-VHL stability in Ub mutant strains. Graphs represent densitometric quantification relative to t = 0 (mean ± SEM from 3 biologically independent experiments). g, Doa10/Hrd1 and Ubr1/San1 E3 ligases collaborate to ubiquitinate misfolded proteins with branched K11/K48 chains, thereby targeting them for proteasomal clearance. Inhibition of either type of linkage results in the sequestration of the misfolded proteins into puncta. a, b, f, Immunoblots represent 3 biologically independent experiments.

Figure 3 ∣. K11- and K48- linked ubiquitination of misfolded proteins involves different chaperones.

a, Molecular chaperones are involved in the ubiquitination of misfolded proteins by E3s. b, Relative amounts of K11 and K48 Ub present on Flag-VHL in chaperone deletion strains compared with WT. WT or chaperone deletion strains expressing Flag-VHL were lysed after 1 h bortezomib treatment. Ub linkage ELISA was then performed as described in Extended Fig. 3a. Bars represent normalized values from each strain (mean ± SEM from 3 biologically independent experiments) expressed as a proportion of normalized WT. Strains with statistically significant differences vs. WT by one-way ANOVA + Dunnett’s multiple comparisons test are indicated with the adjusted p-value, or **** for p < 0.0001. c, K11 and K48 ubiquitination of misfolded proteins proceeds through the action of different chaperone pathways.

Figure 4 ∣. Confining misfolded proteins to the nucleus or cytoplasm alters their PQC requirements.

a-b, NLS-GFP-VHL and NES-GFP-VHL accumulate in the nucleus or cytoplasm, respectively, upon proteasome inhibition. Expression of NLS- or NES- GFP-VHL in WT cells was shut off in glucose media with 50 μM bortezomib for 1 h (b). Cells were immunostained for nuclear pores (Nsp1, red) and imaged by fluorescence microscopy. Images represent 3 biologically independent experiments. Scale bars = 2 μm. c, NLS- and NES- GFP-VHL are cleared by the proteasome. Cycloheximide-chase and immunoblot to assess stability of NLS- and NES- GFP-VHL in WT cells treated with (WT + Bz) or without (WT) 50 μM bortezomib. Immunoblots represent 3 biologically independent experiments. d, Confining VHL to the nucleus or cytoplasm alters its E3 ligase requirement. % cells with NLS- or NES- GFP-VHL puncta in deletion strains following 1 h shut-off. e, Nuclear VHL has severely reduced K11-Ub ubiquitin linkages. ELISA performed as in Fig. 2c, but in GFP-multiTrap® instead of α-FLAG-conjugated plates. f, Nuclear VHL clearance is unaffected by K11-Ub linkages. Experiment performed as in d, but in cells expressing WT or K11R mutant Ub as its only ubiquitin source. g-h, Clearance of NLS-GFP-VHL requires Dsk2. Cycloheximide-chase performed as in c, but in WT or Δdsk2 cells, g, Densitometric quantification relative to t = 0 (mean ± SEM from 3 biologically independent experiments). i, Nuclear and cytoplasmic misfolded proteins have distinct clearance requirements. Cytoplasmic misfolded proteins require tagging with both K11- and K48- Ub, by cooperating chaperones and E3s, for proteasomal degradation. In the nucleus, tagging with K48-Ub is sufficient for recognition by Dsk2 and subsequent proteasomal degradation. d-f, Bars represent mean ± SEM from 3 biologically independent experiments. Statistically significant differences vs. WT (d, f) or GFP-VHL (e) by one-way ANOVA + Dunnett’s multiple comparisons test are indicated (adjusted p-value, or ****p < 0.0001; ns = p > 0.05).