Research Article

The nucleolus functions as a phase-separated protein quality control compartment

See allHide authors and affiliations

Science  26 Jul 2019:
Vol. 365, Issue 6451, pp. 342-347
DOI: 10.1126/science.aaw9157

Phasing-in quality control in the nucleus

The fundamental process of protein quality control in the nucleus is not well understood. The nucleus contains several non–membrane-bound subcompartments forming liquid-like condensates. The largest of these is the nucleolus, the site of ribosome biogenesis. Frottin et al. found that metastable nuclear proteins that misfold upon heat stress enter the nucleolus. In the nucleolus, they avoid irreversible aggregation and remain competent for heat shock protein 70–dependent refolding upon recovery from stress. Prolonged stress or the uptake of proteins associated with neurodegenerative diseases prevented this reversibility. Thus, the properties of a phase-separated compartment can assist in protein quality control.

Science, this issue p. 342

Structured Abstract

INTRODUCTION

Cells have evolved quality control mechanisms that operate under normal growth conditions and during stress to maintain protein homeostasis (proteostasis) and prevent the formation of potentially toxic aggregates. Research in recent decades has identified complex quality control systems in the cytoplasm that mediate protein folding, prevent misfolding, and cooperate in protein degradation with the proteasome and autophagy pathways. Compartment-specific proteostasis networks and stress response pathways have also been described for the endoplasmic reticulum and mitochondria. In contrast, relatively little is known about protein quality control in the nucleus.

Proteins enter the nucleus in a folded state, so chaperone machinery specific for de novo folding is not required. However, the nuclear proteome is rich in stress-sensitive, metastable proteins, which suggests that effective protein quality control mechanisms are in place to ensure conformational maintenance. The nucleus contains several non–membrane-bound subcompartments. The largest of these is the nucleolus, the site of ribosome biogenesis. During stress, Hsp70 and other molecular chaperones accumulate in the nucleolus, presumably to protect unassembled ribosomal proteins against aggregation. The nucleolus consists of liquid-like phases or domains that have differential surface tension and do not intermix. The outermost of these, the granular component (GC), is rich in negatively charged proteins such as nucleophosmin and nucleolin, which, combined with RNA, can undergo phase separation into liquid droplets in vitro, as shown for nucleophosmin.

RATIONALE

Nuclear protein aggregates have been observed in various neurodegenerative disorders such as amyotrophic lateral sclerosis and Huntington’s disease, but protein quality control in the nucleus is not well understood. Here, we used a combination of fluorescence imaging, biochemical analyses, and proteomics to investigate the fate of stress-denatured and aberrant proteins in the nucleus, focusing specifically on the role of the nucleolus and its phase-separated nature in protein quality control.

RESULTS

Upon heat stress, misfolded nuclear proteins entered the liquid-like GC phase of the nucleolus, where they associated with proteins including nucleophosmin and adopted a state of low mobility. As a consequence, a fraction of nucleophosmin and nucleolin also converted to a less dynamic state. Storage in the GC phase effectively prevented the irreversible aggregation of misfolded protein species, allowing their extraction and refolding upon recovery from stress in a Hsp70-dependent manner. We identified ~200 different proteins that reversibly partitioned upon stress into the immobile substate of the GC, entering either from the nucleoplasm or from within the nucleolus. Disruption of the GC phase resulted in the formation of stable aggregates of stress-denatured proteins in the nucleoplasm, which exerted toxic effects by sequestering bystander proteins. Notably, the capacity of the nucleolus to store misfolded proteins proved to be limited. Prolonged stress or the uptake of aberrant proteins associated with neurodegenerative diseases led to a transition of the GC phase from a liquid-like to a solid state, with loss of reversibility and nucleolar dysfunction.

CONCLUSION

The liquid-like GC phase of the nucleolus functions as a non–membrane-bound protein quality control compartment. It is characterized by a remarkable chaperone-like capacity to temporarily store misfolded proteins, preventing their irreversible aggregation and maintaining them as competent for Hsp70-assisted refolding. Nucleoplasmic proteins exit the nucleolus upon refolding, and nucleolar proteins resume their functional state. Our findings provide an example of how the properties of a non–membrane-bound, phase-separated compartment can be used in protein quality control, a fundamental biological function.

Inserting misfolded proteins into the nucleolus prevents irreversible aggregation.

Upon cell stress, misfolded proteins enter the GC phase of the nucleolus to be stored in a state competent for Hsp70-dependent refolding during recovery. Potentially toxic, irreversible aggregates form when transfer into the nucleolus is prevented. A 3D-rendered high-resolution image of the nucleolus is shown: GC, granular component (red); DFC, dense fibrillar component (white); FC, fibrillar center (cyan).

Abstract

The nuclear proteome is rich in stress-sensitive proteins, which suggests that effective protein quality control mechanisms are in place to ensure conformational maintenance. We investigated the role of the nucleolus in this process. In mammalian tissue culture cells under stress conditions, misfolded proteins entered the granular component (GC) phase of the nucleolus. Transient associations with nucleolar proteins such as NPM1 conferred low mobility to misfolded proteins within the liquid-like GC phase, avoiding irreversible aggregation. Refolding and extraction of proteins from the nucleolus during recovery from stress was Hsp70-dependent. The capacity of the nucleolus to store misfolded proteins was limited, and prolonged stress led to a transition of the nucleolar matrix from liquid-like to solid, with loss of reversibility and dysfunction in quality control. Thus, we suggest that the nucleolus has chaperone-like properties and can promote nuclear protein maintenance under stress.

Cells have evolved complex quality control mechanisms that operate under normal growth conditions and during stress to maintain protein homeostasis (proteostasis) and prevent the formation of potentially toxic aggregates (14). Subcellular compartments are equipped with specialized stress response pathways (57) and vary in stress vulnerability (810). The nuclear proteome is enriched in proteins containing intrinsically disordered or low-complexity sequences (11, 12). These metastable proteins do not populate a thermodynamically stable folded state and tend to aggregate upon conformational stress (1315). Indeed, various neurodegenerative disorders associated with protein aggregation, such as amyotrophic lateral sclerosis (ALS) and Huntington’s disease, are characterized by the presence of intranuclear inclusions (1620).

The nucleus contains several non–membrane-bound subcompartments (21). The largest of these is the nucleolus, which consists of liquid-like phases that do not intermix, giving rise to distinct zones (Fig. 1A and fig. S1, A and B) (22). Embedded in the outer granular component (GC) phase is the fibrillar center (FC) for the transcription of ribosomal RNA (RNA polymerase I subunit RPA40 as marker). The FC is surrounded by the dense fibrillar component (DFC), which contains the ribonucleoprotein fibrillarin (FBL) (Fig. 1A and fig. S1, A and B). The GC phase is rich in negatively charged proteins such as nucleophosmin (NPM1) and nucleolin (23). NPM1 contains extensive unstructured regions and undergoes liquid-liquid phase separation in vitro (24, 25). During stress, Hsp70 and other molecular chaperones accumulate in the nucleolus, presumably to protect unassembled ribosomal proteins against aggregation (2628). Stress-induced transfer of a nuclear model protein to the nucleolus has also been observed (29). Here, we found that during stress, misfolded proteins enter the liquid-like GC phase of the nucleolus, where irreversible coaggregation of different misfolded protein species is prevented, allowing Hsp70-mediated extraction and refolding (or degradation) upon recovery from stress. In contrast, disruption of the GC phase causes the formation of stable protein aggregates. Prolonged stress results in a transition of the nucleolar matrix from liquid-like to solid and prevents quality control.

Fig. 1 Misfolded proteins transiently accumulate in the GC phase of the nucleolus during stress.

(A) Schematic representation and 3D-rendered DNA-PAINT (30, 31) superresolution image of a HeLa cell nucleolus under normal growth conditions. Red, NPM1 (GC); white, FBL (DFC); cyan, RPA40 (FC). See also fig. S1, A and B. (B) HEK293T cells stably expressing NLS-LG were treated with dimethyl sulfoxide (DMSO; mock) or PBT before 2 hours of heat stress (+HS), followed by recovery for 2 hours (+HS +Rec). Control cells were maintained at 37°C (–HS). Cells were stained for endogenous NPM1 (red); nuclei are marked by dashed circles. (C) Superresolution imaging of HEK293T cells expressing NLS-LG after HS treatment, with staining for GFP, endogenous NPM1, FBL, and RPA40. See fig. S1D for –HS control. (D) No-LG in the nucleolus without (–HS) and with (+HS) heat stress in the presence or absence of PBT before bleaching (Pre), immediately after bleaching (Bleach), and 2 s after bleaching. (E to G) FRAP analysis of No-LG (E), GFP-NPM1 (F), and No-GFP (G). No-LG experiments (E) show PBT treatment (open circles) or DMSO (solid circles) as a control. GFP-NPM1 experiments (F) show cotransfection of No-LS (open circles). For the –HS condition (green), cells were maintained at 37°C during acquisition. For +HS experiments (red), cells were incubated at 43°C for 1 hour before acquisition and maintained at 43°C during acquisition. For the No-LG recovery experiment [(E), right graph, blue], cells were subjected to HS and allowed to recover for 1 hour (+HS +Rec; solid circles), followed by FRAP. Hsp70 was inhibited with VER-155008 before shifting cells to recovery (+HS +Rec +VER; open circles) (32). Cycloheximide was present during recovery. The graphs display corrected and normalized FRAP curves with double-exponential fits. Curves represent means ± SD (n ≥ 4 biological repeats representing at least 12 different cells). The first 150 s after photobleaching are shown. Quantification of No-LG and GFP-NPM1 mobility is shown in fig. S3, A and B, respectively. Scale bars, 1 μm [(A), (C), (D)], 10 μm (B).

Transfer of misfolded protein to the nucleolus upon stress

To investigate the fate of a nuclear protein as it denatures during heat stress (HS), we generated human embryonic kidney (HEK) 293T cells stably expressing a fusion protein of the thermolabile firefly luciferase and heat-stable green fluorescent protein (GFP) carrying an N-terminal nuclear localization signal (NLS-LG) (fig. S1C). NLS-LG was diffusely distributed in the nucleus. Upon incubation at 43°C (2 hours), a substantial fraction of NLS-LG entered the nucleoli (Fig. 1B). Superresolution imaging (fig. S1A) (30, 31) showed that nucleolar NLS-LG localized to the NPM1-containing GC phase (Fig. 1C and fig. S1D). Transfer of NLS-LG to the nucleolus was prevented by stabilizing luciferase with the substrate analog 2-phenylbenzothiazole (PBT) (Fig. 1B and fig. S1E). Thus, unfolding was a prerequisite for transfer to the nucleolus. Upon recovery from HS, nucleolar NLS-LG redistributed to the nucleoplasm (Fig. 1B), as shown by inhibiting synthesis of new protein (fig. S1F). More than 60% of NLS-LG was degraded during HS (fig. S1G). Notably, the NLS-LG present after recovery showed a higher specific luminescence activity than during HS, indicative of refolding of misfolded protein (fig. S1G).

Hsp70 transferred to nucleoli upon HS (2729), even when NLS-LG was stabilized (fig. S1E). Thus, Hsp70 entered the nucleolus either in a complex with endogenous proteins or in free form. Inhibition of the adenosine triphosphatase activity of Hsp70 by the compound VER-155008 (32) prevented both Hsp70 and misfolded NLS-LG from exiting the nucleolus during recovery (fig. S2A). Thus, nucleolar Hsp70 is involved in refolding and repartitioning NLS-LG (and presumably other metastable proteins) to the nucleoplasm. Indeed, misfolded cytosolic carboxypeptidase Y*-mCherry (CC*) (33) also accumulated in nucleoli when its degradation was inhibited (fig. S2B). Thus, the nucleolus serves as a storage compartment for a subset of misfolded proteins under proteotoxic stress conditions, preserving them in a state competent for refolding or degradation.

Misfolded proteins in the nucleolus have low mobility

We next analyzed the mobility of NLS-LG in the GC phase of the nucleolus by recording fluorescence recovery after photobleaching (FRAP). To compare the mobility of folded and misfolded proteins within the nucleolus, we fused a nucleolar targeting sequence (34) to NLS-LG, generating the protein No-LG (fig. S1C). A large fraction of No-LG constitutively localized to the nucleolus in the absence of stress and in the presence of the luciferase stabilizer PBT (Fig. 1D and fig. S2, C and D), thus behaving as a functional nucleolar protein. No-LG in the nucleolus showed complete FRAP (Fig. 1, D and E, and fig. S3A) and a mobility similar to that of the liquid-like GFP-NPM1 (Fig. 1F and fig. S3B) (22). HS resulted in a more complete localization of No-LG to the nucleolus, an increase in the nucleolar concentration of No-LG (by a factor of 1.37 ± 0.13, n = 3), and a shift to a markedly reduced mobility (Fig. 1, D and E, and figs. S2, C and D, and S3A). In contrast, the presence of PBT during HS preserved the high mobility of No-LG (Fig. 1, D and E, and fig. S3A). Thus, unfolding changed the interaction of luciferase with the GC phase. The larger hydrodynamic radius of unfolded luciferase may also contribute to the lower mobility. Consistently, the mobility of nucleolar GFP (No-GFP) (figs. S1C and S2C) remained unchanged upon heat stress (Fig. 1G).

HS also induced the formation of an immobile fraction of GFP-NPM1 (~30% of total) (Fig. 1F and fig. S3B), which returned to normal mobility upon recovery (fig. S3, B and C). Similar observations were made for nucleolin (GFP-NCL) (fig. S3, B and C). This suggested an association with unfolded (or misfolded) proteins that altered GC mobility. In support of this notion, expression of nucleolar luciferase (as a fusion with mScarlet; No-LS) further increased the immobile fraction of GFP-NPM1 upon HS (Fig. 1F and fig. S3B), which suggests that the amount of immobile GC protein correlated with the load of misfolded protein. In contrast, folded No-LS in control conditions had no effect on GFP-NPM1 mobility (Fig. 1F and fig. S3B). Indeed, endogenous NPM1 associated (directly or indirectly) with NLS-LG or No-LG upon HS by coimmunoprecipitation, but not in the absence of stress (fig. S3D). Thus, the unfolding of luciferase enhanced the association with the GC, consistent with a fraction of liquid-like NPM1 and nucleolin adopting a less dynamic state. Inhibiting Hsp70 activity completely inhibited the stress-denatured No-LG from recovery to normal mobility (Fig. 1E and fig. S3, A and E). Because No-LG remained localized to the nucleolus after refolding, this finding suggested that refolding mediated by Hsp70 was initiated in the nucleolus and was coupled with the mobilization of luciferase. Thus, upon proteotoxic stress, misfolded proteins immersed into the nucleolus, where they associated with GC proteins, thereby converting part of the liquid-like GC phase to a state of low mobility (Fig. 1F and fig. S3, B and C). Mobility was reestablished during recovery in an Hsp70-dependent manner, concomitant with refolding.

Endogenous proteins reversibly enter the nucleolus upon stress

To identify the endogenous proteins that enter the GC phase of the nucleolus upon stress, we performed GFP-NPM1 pull-down experiments followed by quantitative proteomics. We identified ~200 proteins that associated with NPM1 specifically upon HS, including numerous proteins of the nucleoplasm and nucleolus as well as some cytosolic proteins (Fig. 2A, fig. S4, A and B, and table S1). Thus, the stress-protective GC phase is accessible to proteins from both outside and within the nucleolus.

Fig. 2 GFP-NPM1 reversibly associates with endogenous proteins.

(A) Number of GFP-NPM1–associated proteins (see table S1). GFP-NPM1 was transiently expressed in SILAC-labeled HEK293T cells before exposure to heat stress (+HS), followed by recovery (+HS +Rec) or recovery in the presence of Hsp70 inhibitor (+HS +Rec +VER). Control cells remained at 37°C (–HS). Anti-GFP immunoprecipitates from cell lysates were analyzed by mass spectrometry. Proteins that were enriched by a factor of ≥2 upon +HS over the –HS sample in at least three of four independent experiments were defined as being associated with GFP-NPM1 (see table S1). (B) Hsp70 inhibition prevents reversibility of GFP-NPM1 associations. Venn diagrams show distribution of GFP-NPM1–associated proteins under the conditions analyzed in (A). (C) Bromodomain-containing protein 2 (BRD2) reversibly accumulates in the nucleolus. HEK293T cells were treated as described above. Cells were immunostained for endogenous NPM1 and BRD2. Nuclei are marked by dashed circles. Representative images of three biological repeats are shown. Scale bar, 10 μm. (D) Partitioning of BRD2, NLS-LG, and Hsp70 between nucleoplasm and nucleoli in HEK293T cells treated as described above. Proteins were detected by immunostaining. Relative concentrations in nucleoplasm and nucleolus were quantified by measuring relative fluorescence intensities in 57 to 145 cells per condition across three biological repeats. **P ≤ 0.05, ***P ≤ 0.001 (two-sided t test).

Nucleolin was also enriched in the NPM1 pull-down, but not the DFC marker fibrillarin (fig. S4C), suggesting an enhanced association between NPM1 and nucleolin under heat stress, consistent with their reduced mobility (Fig. 1F and fig. S3, B and C). More than 400 proteins of the nucleoplasm or nucleolus were not enriched upon HS (fig. S4, A and C, and table S1). Thus, the proteins that entered the GC phase constituted a thermally sensitive subproteome. Indeed, these proteins were enriched in disordered and low-complexity sequences (fig. S4D), hallmarks of metastable structure. Their accumulation in the GC phase was reversible, whereas inhibition of Hsp70 preserved the association with NPM1 for most proteins (Fig. 2, A and B, fig. S5A, and tables S1 to S3). Additional proteins associated with NPM1 upon Hsp70 inhibition during recovery (Fig. 2, A and B, and tables S1 to S3).

We confirmed the reversible accumulation in the nucleolus for the proteins CDK1 and BRD2, which associated with NPM1 upon HS (Fig. 2, C and D, figs. S4C and S5, B and C, and table S1). A small but detectable fraction of total cellular Hsp70 also coprecipitated with NPM1 upon HS (fig. S5C), which suggests that associations with both Hsp70 and misfolded protein may contribute to forming the low-mobility GC fraction (Fig. 1, E and F, and fig. S3, B and C).

Functional relevance of the nucleolus in quality control

To explore the physiological importance of the nucleolus as a quality control compartment, we disrupted the nucleolar organization. Treating cells with a low concentration of the RNA polymerase I inhibitor actinomycin D (Act D) caused nucleolar disassembly and the release of NPM1 into the nucleoplasm (Fig. 3A) (35, 36). NPM1 lost its liquid-like properties, as judged by its fast mobility (fig. S6A). NLS-LG was diffusely distributed in the nucleus of Act D–treated cells in the absence of stress but formed aggregate foci upon HS (Fig. 3A). These foci did not colocalize with NPM1. They resolved only slowly and inefficiently during recovery (Fig. 3, A and B) and sequestered Hsp70 for hours after the removal of stress (fig. S6B). The terminally misfolded CC* also formed persistent aggregates in Act D–treated cells, when proteasome function was inhibited (fig. S6, C and D). Thus, transport to the phase-separated GC compartment of the nucleolus was required to maintain misfolded proteins in a state competent for refolding or degradation once proteotoxic stress was relieved.

Fig. 3 The nucleolar environment prevents irreversible protein aggregation.

(A) HEK293T cells expressing NLS-LG (green) were treated with actinomycin D (Act D) where indicated, followed by incubation with and without HS and recovery as in Fig. 1B. Cells were immunostained for NPM1 (red); nuclei are marked by dashed circles. (B) HEK293T cells expressing NLS-LG were treated as in (A) and recovery was monitored over 2 hours. Cells with nuclear NLS-LG foci were counted during recovery and expressed as percentage of total. Data are means ± SD; 453 to 693 cells were counted per time point and per condition across three biological repeats. *P ≤ 0.05, ***P ≤ 0.001 (two-sided t test). (C) HEK293T cells expressing NLS-LG were subjected to FRAP analysis. Cells were treated with Act D (open circles) before HS where indicated. For –HS experiments (green), the nucleoplasmic region was bleached, where NLS-LG localizes at 37°C. For +HS experiments (red), the nucleolus was bleached (see schematic). Left: Normalized FRAP curves with double-exponential fits. Curves represent means ± SD (n ≥ 3 biological repeats). Right: Mobile fraction from the double-exponential fit. ***P ≤ 0.001 (two-sided t test). (D) Cells expressing NLS-LG were subjected to Act D treatment where indicated, followed by heat stress (+HS) and stress with recovery (+HS +Rec), and stained with AmyT. Nuclei are marked by dashed circles. (E) Concentration of NLS-LG in the nucleolus and in nucleoplasmic aggregates (+Act D) upon heat stress. ***P ≤ 0.001 (Mann-Whitney test; 100 measurements per condition across three biological repeats). Scale bars, 10 μm.

The NLS-LG in nucleoplasmic aggregates of Act D–treated cells was less mobile than nucleolar NLS-LG (Fig. 3C). Moreover, the nucleoplasmic foci were positive for amyloid (cross β structure)–specific dyes, in contrast to NLS-LG in the nucleolus (Fig. 3D and fig. S6E). Consistent with an amyloid-like state, the concentration of NLS-LG in nucleoplasmic foci was higher than in the nucleolus by a factor of ~3 (Fig. 3E). When nucleoli were disrupted, HS also caused endogenous proteins to form amyloid-like foci (fig. S6, F and G). Thus, entry of misfolded proteins into the nucleolus prevented amyloid-like aggregation.

We next analyzed the effect of the nucleolar environment on the model protein β17. This small β-sheet protein undergoes amyloidogenic aggregation and forms fibrils in vitro (37). Targeting β17 to the nucleus (NLS-β17) results in its accumulation in the nucleolus and a reduced toxicity relative to cytosolic β17 aggregates (8). To determine whether the nucleolar environment was responsible for this protective effect, we targeted β17 to the nucleoplasm by expressing it with the C-terminal nuclear localization signal PY (fig. S7A) (38). β17-PY formed foci in the nucleoplasm, whereas NLS-β17 accumulated in the GC phase of the nucleolus (Fig. 4A). Note that the NLS apparently functioned as a nucleolar targeting (or retention) signal in the sequence context with β17, but not in context with LG or GFP (Fig. 1B and fig. S2C). The function of the two localization sequences was position-independent (fig. S7A). The nucleoplasmic β17-PY aggregates were more concentrated than nucleolar NLS-β17 by a factor of 3 (Fig. 4B). β17-PY was more toxic than NLS-β17 (Fig. 4C), indicating that localization to the nucleolus reduced toxicity (8). The PY sequence per se did not confer toxicity (Fig. 4C and fig. S7B). As expected, nucleolar β17 variants but not nucleoplasmic β17-PY associated with NPM1 (fig. S7C). Moreover, NLS-β17-GFP was significantly more mobile than β17-GFP-PY (fig. S7, D and E), whereas disrupting the GC phase with Act D rendered NLS-β17-GFP less mobile (fig. S7, D and E).

Fig. 4 Accumulation in the nucleolus reduces toxicity of amyloidogenic protein and prevents coaggregation with misfolded luciferase.

(A) HEK293T cells were transfected with NLS-β17 or β17-PY prior to superresolution imaging. Red, C-myc (β17); cyan, NPM1; white, RPA40. Zoomed images of NLS-β17 in the nucleolus are shown at right. (B) Density of β17 in the nucleolus (NLS-β17) and in nucleoplasmic aggregates (β17-PY) measured by superresolution imaging. Data were normalized to the average density of nucleolar NLS-β17. ***P ≤ 0.001 (Mann-Whitney test). At least 36 and 52 measurements were performed on one representative experiment out of three biological repeats for NLS-β17 and β17-PY, respectively. (C) HEK293T cells were transfected with the indicated constructs and MTT cell viability assays were performed 4 days after transfection. Data were normalized to cells transfected with empty vector. Data are means + SD (n ≥ 3). **P ≤ 0.01 (two-sided t test). (D) β17-PY or NLS-β17 were transfected into the NLS-LG–expressing HEK293T cell line; 24 hours after transfection, cells were subjected to HS (+HS) and allowed to recover for 1 hour (+HS +Rec) before fixation. Cyan, endogenous NPM1; red, c-myc (β17). Arrows show NLS-LG sequestration into β17-PY aggregates. Scale bars, 1 μm (A), 10 μm (D).

Amyloid-like aggregates exert their toxic effect in part by coaggregating and sequestering essential, metastable proteins (8, 3941). Indeed, the nucleoplasmic β17-PY aggregates sequestered NLS-LG upon HS, thereby preventing NLS-LG from entering the nucleolus (Fig. 4D). Nucleolar NLS-β17 had no such effect and did not prevent repartitioning of NLS-LG to the nucleoplasm upon recovery (Fig. 4D). Thus, the GC phase of the nucleolus has the capacity to simultaneously store different proteins and allow them to undergo selective renaturation.

Accumulation of misfolded proteins in the nucleolus did not interfere with ribosome biogenesis, as nucleolar NLS-β17 did not interfere with the assembly and export of yellow fluorescent protein (YFP)–tagged 40S ribosomal protein S2 (RPS2-YFP) to the cytosol (fig. S7, F and G) (42). In contrast, nucleoplasmic aggregates of β17-PY caused coaggregation of RPS2-YFP and nuclear retention (fig. S7, F and G).

Limitations of nucleolar quality control

To explore the capacity of the nucleolus for incorporating misfolded proteins, we exposed cells to prolonged stress. We observed a significant increase in nucleolar volume during the first 2 hours of HS (Fig. 5A), presumably reflecting the influx of misfolded proteins. The nucleoli lost their liquid droplet–like appearance and adopted irregular shapes (fig. S8, A and B), suggestive of a transition to a hardened state. Indeed, the mobile fraction of GFP-NPM1 decreased markedly during prolonged HS (Fig. 5B and fig. S8, C and D). To further assess these changes, we stained NLS-LG–expressing cells with the amyloid-specific dye AmyT and observed a distinct nucleolar staining that developed over time (Fig. 5C). The foci that formed during extended HS dissolved only slowly upon recovery (fig. S8E). Apparently, prolonged stress exhausted the storage capacity of the nucleolus for misfolded proteins, resulting in a transition to a solid, aggregated state.

Fig. 5 The nucleolus changes phase properties during prolonged stress or accumulation of dipeptide repeat protein.

(A) HeLa cells were incubated at 43°C for the number of hours indicated before staining for endogenous NPM1. The average nucleolar volume per nucleus is displayed as a bee-swarm box plot. Welch’s t test was used to assess significant differences between conditions; the resulting P values are shown. Results are from three biological repeats representing 155 to 264 analyzed cells per condition. (B) NPM1 mobile fraction from FRAP experiments performed with HeLa cells transfected with GFP-NPM1. Cells were subjected to HS for the times indicated before and during FRAP measurement, and GFP-NPM1 mobile fractions were calculated. See also fig. S8, C and D. Data are means + SD of at least three biological repeats. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 (two-sided t test). (C) NLS-LG–expressing HEK293T cells were subjected to heat treatment for the indicated times and stained with AmyT. Nuclei are marked by dashed circles. (D) HEK293T cells were cotransfected with NLS-LS and either PR175-GFP or the nucleolar control protein No-GFP. Cells were maintained at 37°C (–HS) or subjected to heat stress (+HS) and recovery (+HS +Rec). (E) Model of nucleolar protein quality control for proteins entering the nucleolus from the nucleoplasm. Misfolded proteins immerse into the liquid-like GC phase of the nucleolus, presumably as a complex with Hsp70 (green), where they associate with GC proteins such as NPM1 (dark blue). There they are stored in an immobile state within the liquid-like GC phase, accompanied by an expansion of the nucleolus. Mobility is reestablished upon recovery from stress in an Hsp70-dependent manner, allowing refolding or proteasomal degradation in the nucleoplasm. Preventing access to the GC phase results in amyloid-like aggregation in the nucleoplasm. Upon prolonged stress, the GC phase increasingly transitions toward a more solid state. Misfolded proteins are no longer dispersed but form aggregates with amyloid-like properties. Scale bars, 10 μm.

Expression of C9orf72 encoded dipeptide repeat proteins (DPRs) is a possible cause of familial ALS and frontotemporal dementia (FTD) (4345). These peptides cause nucleolar dysfunction by modulating the liquid-like properties of the nucleolus (19, 20). We expressed the DPR-protein PR175-GFP along with nuclear luciferase (NLS-LS). PR175-GFP incorporated efficiently into the GC phase of the nucleolus (Fig. 5D) (19, 20), resulting in reduced mobility of a fraction of mScarlet-NPM1 (fig. S9A). NLS-LS entered the nucleolus during HS and colocalized with PR175-GFP but failed to repartition during recovery (Fig. 5D), remaining arrested in the nucleolus for hours (fig. S9B). In contrast, control cells expressing No-GFP allowed normal NLS-LS repartitioning (Fig. 5D and fig. S9B). Thus, nucleolar DPR protein leads to a breakdown of nucleolar quality control, which may contribute to the cellular pathology in ALS and FTD.

Conclusions

The liquid-like GC phase of the nucleolus functions as a non–membrane-bound protein quality control compartment (Fig. 5E). It is characterized by a remarkable chaperone-like capacity to prevent irreversible aggregation of misfolded proteins, facilitating refolding during recovery from stress. Misfolded proteins associate with nucleolar proteins including NPM1, thereby converting a fraction of the GC phase to a less dynamic state (Fig. 5E). The association of misfolded proteins with the GC phase is regulated by the chaperone Hsp70, which is required for refolding (Fig. 5E). Nucleoplasmic proteins exit the nucleolus upon refolding, and nucleolar proteins resume their functional state. However, the capacity of the nucleolus to store misfolded proteins is limited, and prolonged stress causes aberrant phase behavior associated with the danger of irreversible aggregation (Fig. 5E). Moreover, disease-related DPR proteins impair the ability of the nucleolus to reversibly store misfolded proteins—a mechanism that may contribute to neurodegenerative pathology.

Supplementary Materials

science.sciencemag.org/content/365/6451/342/suppl/DC1

Materials and Methods

Figs. S1 to S9

Tables S1 to S3

References (4661)

References and Notes

Acknowledgments: We thank U. Kutay for the RPS2-YFP HeLa cell line; D. Edbauer for the expression plasmid PR175-GFP; B. Sperl, O. K. Wade, and S. Strauss for technical assistance; and A. Ries for support with SILAC-MS/MS. We acknowledge support by the MPIB Imaging facility and G. Cardone for providing the algorithm for image quantification. Funding: F.F. was supported by an EMBO Long Term Fellowship. The research leading to these results has received funding from the European Commission under grant FP7 GA ERC-2012-SyG_318987–ToPAG, and MolMap grant agreement 680241, the Munich Cluster for Systems Neurology, and the Max Planck Foundation. Author contributions: F.F. designed and performed most of the experiments. R.G. carried out initial experiments. F.S. and T.S. carried out the high resolution imaging. R.K. supervised the proteomic analysis and S.T. and J.C. analyzed sequence properties of NPM1 associated proteins. R.J. designed and supervised the high-resolution imaging experiments. F.U.H. and M.S.H. initiated and supervised the project and wrote the paper with input from F.F. and the other authors. Competing interests: J.C. is also affiliated with the Department of Biological and Medical Psychology, Faculty of Psychology, University of Bergen, Bergen, Norway. The authors declare no other competing interests. Data and materials availability: Data from the mass spectrometry analysis described in this manuscript can be found in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article