Oncogene (2011) 30, 790–805 & 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11 www.nature.com/onc ORIGINAL ARTICLE inhibitors induce nucleolar aggregation of proteasome target and polyadenylated RNA by altering availability

L Latonen1,3, HM Moore1, B Bai2,SJa¨a¨maa1 and M Laiho1,2

1Molecular Cancer Biology Program and Haartman Institute, University of Helsinki, Helsinki, Finland and 2Department of Radiation Oncology and Molecular Radiation Sciences, Sidney Kimmel Comprehensive Cancer Center, The Johns Hopkins University School of Medicine, Baltimore, MD, USA

The ubiquitin–proteasome pathway is essential for most Keywords: proteasome inhibitor; proteasome; ubiquitin; cellular processes, including quality control, cell polyadenylated RNA; nuclear export; cycle, , signaling, protein transport, DNA repair and stress responses. Hampered proteasome activity leads to the accumulation of polyubiquitylated proteins, endoplastic reticulum (ER) stress and even cell Introduction death. The ability of chemical proteasome inhibitors (PIs) to induce apoptosis is utilized in cancer therapy. During Ubiquitin is a small conserved protein that is covalently PI treatment, misfolded proteins accrue to cytoplasmic linked to its target proteins by ubiquitylation (Hershko . The formation of aggresome-like structures and Ciechanover, 1998; Weissman, 2001; Finley, 2009). in the nucleus has remained obscure. We identify here Ubiquitin is essential for many if not most cellular a nucleolus-associated RNA-protein aggregate (NoA) processes, including protein quality control, cell cycle formed by the inhibition of proteasome activity in and transcriptional control, cellular signaling, protein mammalian cells. The aggregate forms within the transport, DNA repair and stress responses. These nucleolus and is dependent on nucleolar integrity, yet is diverse functions are governed by attachment of mono- a separate structure, lacking nucleolar marker proteins, and polyubiquitin chains on the targets through ribosomal RNA (rRNA) and rRNA synthesis activity. The ubiquitin lysine residues. Polyubiquitylation generally NoAs contain polyadenylated RNA, conjugated ubiquitin leads to proteolytic degradation of the target by the and numerous nucleoplasmic proteasome target proteins. proteasome. Polyubiquitylation occurs predominantly Several of these are key factors in oncogenesis, including through Lys48 residues, although recently ubiquitin transcription factors p53 and retinoblastoma protein (Rb), chains based on most lysine residues have been several cell cycle-regulating cyclins and cyclin-dependent implicated in proteasomal targeting (Ikeda and Dikic, kinases (CDKs), and stress response kinases ataxia- 2008; Saeki et al., 2009; Xu et al., 2009). Monoubiqui- telangiectasia mutated (ATM) and Chk1. The aggregate tylation regulates, for example, membrane transport, formation depends on ubiquitin availability, as shown by nucleocytoplasmic protein localization, protein kinase modulating the levels of ubiquitin and deubiquitinases. activation, DNA repair and chromatin dynamics (Chen Furthermore, inhibition of region main- and Sun, 2009). Conjugated ubiquitin is released during tenance 1 protein homolog (CRM1) export pathway proteasomal processing and by deubiquitylating aggravates the formation of NoAs. Taken together, we enzymes (Reyes-Turcu et al., 2009), and reused. identify here a novel nuclear stress body, which forms The ubiquitin–proteasome pathway is a major upon proteasome inactivity within the nucleolus and is proteolytic system of all eukaryotic cells (Hershko detectable in mammalian cell lines and in human tissue. and Ciechanover, 1998; Weissman, 2001; Finley, These findings show that the nucleolus controls protein 2009). Inhibition of proteasome activity leads to the and RNA surveillance and export by the ubiquitin accumulation of polyubiquitylated and misfolded pathway in a previously unidentified manner, and provide proteins, ER stress and eventually apoptosis (Kopito, mechanistic insight into the cellular effects of PIs. 2000; Navon and Ciechanover, 2009). Proteasome Oncogene (2011) 30, 790–805; doi:10.1038/onc.2010.469; activity can be inhibited by chemical inhibitors, which published online 18 October 2010 fall into several classes based on their chemical structure, mechanism of inhibition and specificity (Kisselev and Goldberg, 2001). Several in vitro studies Correspondence: Dr M Laiho, Department of Radiation Oncology and Molecular Radiation Sciences, Sidney Kimmel Comprehensive utilize peptide aldehydes (such as MG132, ALLN), Cancer Center, The Johns Hopkins University School of Medicine, which are substrate analogs, or a non-peptide inhibitor 1550 Orleans Street, CRB II, Room 444, Baltimore, MD 21231, USA. lactacystin. The ability of proteasome inhibitors (PIs) to E-mail: [email protected] induce cell death is exploited in clinical cancer treatment 3Current address: Institute of Medical Technology, University of Tampere, Tampere, Finland. with bortezomib (Velcade; PS-341; MG-341), which is a Received 13 May 2010; revised 21 August 2010; accepted 7 September Food and Drug Administration-approved PI in clinical 2010; published online 18 October 2010 use in mantle cell lymphoma and multiple myeloma Nucleolar aggregates L Latonen et al 791 (Navon and Ciechanover, 2009). The potential use of nucleoli in PI-treated cells (Klibanov et al., 2001; PIs against solid tumors is under active investigation. Latonen et al., 2003; Karni-Schmidt et al., 2007). To Upon PI treatment, polyubiquitylated proteins accu- study how the nucleoli are affected upon proteasome mulate at sites of , forming organized inhibition, we treated WS1 human skin fibroblasts aggregates. Most prominently, this occurs in the and HEL-299 human lung fibroblasts with MG132 pericentrosomal area in the cytoplasm, in structures (10 mM) for 12 h. At this time, cell death is not yet called aggresomes (Wo´jcik et al., 1996; Johnston et al., prevailing (Supplementary Figure S1a). We immuno- 1998). In the nucleus, chemical inhibition of the stained nucleolar substructures using antibodies against proteasome induces transient nuclear stress granules nucleophosmin (NPM; for granular component), fibril- containing stress response proteins such as heat-shock larin (FBL; for dense fibrillar component) and upstream factors (Holmberg et al., 2000), but no aggresome binding factor (UBF; for fibrillar center). Treatment formation in the nucleus has been reported. We and of the cells with MG132 led to drastic changes in others have shown that proteasome inhibitor MG132 the localization of the nucleolar markers (Figure 1a). causes translocation of certain stress response-related NPM formed a ring-shaped structure, which partially nuclear proteins (p53, , PML, ) to the overlapped by FBL and UBF (Figure 1a). Under phase- nucleolus (Klibanov et al., 2001; Mattsson et al., 2001; contrast illumination, a dense structure became visible Latonen et al., 2003; Kurki et al., 2004; Karni-Schmidt that appeared to localize in the center of the reorganized et al., 2007). As these proteins are degraded by the nucleolus. Similar relocalization was observed by using proteasome, the question has arisen as to whether the antibodies against nucleolar antigens Ki-67 and ARF nucleolus has a role in the degradation of these proteins. (data not shown). To study this effect in more detail, The nucleolus is membrane-less nuclear organelle, we performed transmission EM (TEM) of mock- and which is responsible for the assembly of ribosomal MG132-treated WS1 cells. Although the nucleolar subunits (Leary and Huang, 2001; Fatica and Tollervey, substructures were visible in the mock-treated cells, 2002). The nucleoli are composed of fibrillar centers, MG132 treatment caused alterations in the nucleolar dense fibrillar component and granular component morphology (Figure 1b). This was apparent by inter- (Olson et al., 2000; Hernandez-Verdun, 2006; Boisvert ruption of the nucleolus by electron-dense material et al., 2007), in which ribosomal RNA (rRNA) (Figure 1b, arrowhead). transcription, maturation of pre-RNA transcripts, Previously, Stavreva et al. (2006) have shown that assembly of pre-ribosomal particles and late RNA a high dose of MG132 (100 mM) inhibits pre-RNA processing occur (Lafontaine and Tollervey, 2001; Leary processing. However, a much smaller, more commonly and Huang, 2001). Proteomic analyses have revealed used dose of MG132 (10 mM), is sufficient to inhibit that besides proteins involved in synthesis, the proteasome activity. Hence, we analyzed the effect of nucleolus contains a large number of proteins associated MG132 at this lower concentration, used throughout the with cell cycle and mitotic division regulation, DNA study, on nascent rRNA synthesis using fluorouridine repair and control of tumor suppressor proteins and (FUrd) incorporation. As shown in Figure 1c, rRNA oncogenes (Andersen et al., 2005; Boisvert et al., 2007). synthesis was readily detectable in the nucleoli following Here, we investigated the possible role of the nucleolus MG132 treatment for 12 h, and colocalized with FBL, as in the ubiquitin–proteasome pathway. We show that expected. However, the dense aggregate observed within nuclear proteasome targets accumulate to, and are the nucleolar marker was devoid of any rRNA synthetic immobilized in, aggregates forming in the nucleoli. These activity. To further assess the effect of PI treatment on aggregates accumulate polyadenylated RNA (polyA( þ )), rRNA synthesis, we performed quantitative PCR using but not rRNA, possibly reflecting defective RNAs targeted primers for rRNA precursors. Although RNA poly- for degradation (Houseley and Tollervey, 2009). Forma- merase I (RNA pol I) inhibitor actinomycin D (Act D) tion of these structures depends on the nucleolar rRNA potently suppressed rRNA transcription, MG132 did synthetic activity. We show that an increase in ubiquitin not (Supplementary Figure S1b). We conclude that while pool overcomes the formation of the aggregate, and MG132 treatment causes nucleolar reorganization, the conversely, that inhibition of nuclear export aggravates it. nucleoli retain RNA pol I transcriptional activity. We propose that the formation of the aggregates is a As MG132 may inhibit other enzyme activities in consequence of the decreased availability of ubiquitin, addition to the chymotrypsin-like activity of the which leads to the accumulation and immobilization of proteasome (Kisselev and Goldberg, 2001), we tested multiple proteins and polyA( þ ) RNA to the aggregate. the effect of other proteasome and lysosomal inhibitors. These findings define a novel nucleolus-dependent struc- We treated WS1 and HEL-299 cells with MG132, ture that forms following severe overload of ubiquitylated, ALLN (a peptide aldehyde inhibitor of the proteasome), and likely misfolded, proteins. lactacystin (a highly specific proteasome inhibitor) and two inhibitors of lysosomal proteases, E64 and leupep- tin. We assessed nucleolar changes under phase-contrast Results and FBL staining, and included p53 as marker for a protein undergoing nucleolar translocation. Similarly to Proteasome inhibition alters nucleolar morphology MG132 treatment, ALLN caused nucleolar transloca- We and others have earlier shown that tumor suppressor tion of p53, and this was accompanied by an altered p53, a target protein of the proteasome, localizes to FBL localization in a ring-shaped pattern (Figure 1d).

Oncogene Nucleolar aggregates L Latonen et al 792

Figure 1 Proteasome inhibition alters nucleolar morphology. (a) WS1 cells treated with MG132 (10 mM) for 12 h or left untreated (ctrl) were stained for NPM (green) and co-stained for FBL and upstream binding factor (red) as indicated and imaged using confocal microscopy. Merged images of cells stained with Hoechst (blue), and phase-contrast images are shown. (b) TEM images of WS1 cells treated with or without MG132 (10 mM) for 12 h. Asterisks show the segregation of the nucleolar structures in MG132-treated cell intervened by an electron-dense structure (arrowhead). (c) In vivo labeling of MG132-treated (10 mM) WS1 cells with fluorouridine (green). Cells were co-stained for FBL (red) and DNA (blue). Differential interference contrast (DIC) image is provided. (d) Cells were treated with proteasome inhibitors MG132 (10 mM), ALLN (50 mM) and lactacystin (10 mM), and lysosomal inhibitors E64 (10 mg/ml) and leupeptin (100 mM) for 12 h. Cells were fixed and stained for p53 (green), FBL (red) and DNA (blue), and imaged using confocal microscopy. (e) Quantification of cells with dense nucleolar aggregates formed by the treatments in (d)(n ¼ 3 experiments, error bars denote s.d.). Scale bars, 10 mm. (f) Nucleolar integrity is required for aggregate formation. Cells were pre-treated with Act D (1 mM) for 1 h before the addition of MG132 and incubated for 10 h. Cells were stained for p53 (green) and FBL (red). Arrowheads indicate p53 staining surrounding the nucleolar remnant. Asterisks indicate nucleolar caps. Scale bars, 10 mm. As shown by projection of z-stacks throughout the cell detectable following lactacystin treatment (Figure 1d). by confocal imaging, FBL was observed surrounding The formation of nucleolar aggregates was quantified the central p53 aggregate (Supplementary Movie S1). under phase contrast based on the dense structures p53 aggregation and FBL reorganization were also (Figure 1e). The degree of aggregate formation was

Oncogene Nucleolar aggregates L Latonen et al 793 concentration-dependent for each PI (data not shown). varied, suggesting that the targeting may depend on Inhibition of lysosomal proteases did not lead to factors inherent for each protein, possibly protein half- alterations in FBL or p53 distribution(Figures1dande). life. Certain proteins, like cyclin D, which is expressed In addition, the nucleolar morphology was undisturbed both in the cytoplasm and in the nucleoplasm, was by the lysosomal inhibitors under phase contrast, as detected both in the nucleolar aggregate and in cyto- compared with the PIs, which all caused formation of plasmic aggresomes (Supplementary Figure S2a). To test dense aggregates surrounded by nucleolar structures whether proteasomally degraded cytoplasmic proteins (Supplementary Figure S1c). The aggregates were also undergo translocation to nucleolar aggregates, we detected irrespective of the fixation method used, as analyzed IkBa localization following proteasome inhibi- they were visible in both paraformaldehyde and tion. IkBa did not accumulate to nucleolar aggregates methanol-fixed cells (data not shown). Aggregate for- (Supplementary Figure S2b). Glyceraldehyde 3-phosphate mation was detected in several mammalian cell types, dehydrogenase (GAPDH) and a-smooth muscle actin regardless of their transformation stage or tissue type of (SMA), which are proteins degraded by the lysosomal origin (Supplementary Table S1). In addition, analysis pathway, did not accumulate to the aggregates (Supple- of normal and malignant cell lines in their propensity to mentary Figure S2b, data not shown). These results undergo p53 and MDM2 nucleolar localization follow- indicate that the accumulation of nuclear proteins to the ing proteasome inhibition showed that the response was nucleolar aggregates is not restricted to stress response irrespective of tissue type or pathological status of the proteins, and instead represents a common feature of a cells (Supplementary Table S1). Furthermore, treatment multitude of nuclear proteasome targets. of ex vivo cultured human prostate tissues (Kiviharju-af Ha¨llstro¨m et al., 2007) with MG132 led to the formation of p53 aggregates in epithelial cells of the prostate gland, Polyadenylated RNA is entrapped in nucleolar aggregates indicating that the aggregate formation is not only an The presence of dense nucleolar material in both phase artifact of cells cultured in monolayer, but is also contrast and TEM suggested that the aggregates may, in observed in the context of normal tissue architecture addition to proteins, contain RNA. We first detected (Supplementary Figure S1d). RNA using SYTO12 green, and co-stained for p53. As the aggregates form in the nucleoli, we next asked As shown in Figure 3a, SYTO12 positivity was detected whether nucleolar integrity is essential for the formation in the aggregate and is colocalized with p53. We then of aggregates. To this end, we pretreated the cells for 1 h used in situ hybridization using highly specific locked with Act D, which leads to the inhibition of RNA pol I nucleic acid probes to detect localization of processed activity, disintegration of the nucleolar protein content and precursor rRNAs, and used NPM and p53 as and formation of nucleolar caps (Andersen et al., 2005; co-staining markers for the nucleoli and nucleolar Hernandez-Verdun, 2006). Act D treatment abrogated aggregates, respectively. 28S, 18S and 5.8S rRNA the aggregate formation by MG132, although a little and 50 external transcribed spacer (50ETS) 47S pre- p53 still accrued to the nucleolar remnants (Figure 1f, cursor rRNA largely overlapped with NPM in both arrowheads). This finding indicates that the aggregate mock- and MG132-treated cells (Figure 3b and Supple- formation is dependent on nucleolar structures, and mentary Figure S3). In situ hybridization for 50ETS possibly nucleolar activity. and 28S rRNA and co-staining for p53 showed that their staining patterns were distinct (Figures 3c and d). These results indicate that the nucleolar rRNAs do Proteasome inhibitors cause accumulation of nuclear not colocalize with the aggregate formed in the proteasome targets to the nucleolar aggregate nucleolus by proteasome inhibition, and that they instead Previous studies have shown nucleolar translocation of colocalize with bona fide nucleolar structures and certain proteasome-targeted stress response proteins proteins. (Klibanov et al., 2001; Mattsson et al., 2001; Latonen To resolve the identity of RNA accumulated to the et al., 2003; Kurki et al., 2004; Karni-Schmidt et al., nucleolar aggregates, we tested for the presence of 2007). As 20S proteasome is also localized to the polyA( þ ) RNA. In situ hybridization using a locked nucleolus upon PI treatment (Mattsson et al., 2001) nucleic acid (LNA) probe for polyA( þ ) showed that (Supplementary Figure S2), we asked whether the proteasome inhibition led to a positive signal within the accumulation of proteins to the nucleoli is restricted to nucleolar aggregate, and that p53, but not NPM, nuclear stress response proteins, or whether this repre- overlapped with the signal (Figure 3e and Supplemen- sents a more general phenomenon for proteasome target tary Figure S3d). Furthermore, the polyA( þ ) signal proteins. We tested several other nuclear and cytoplasmic colocalized with the prominent aggregate observed proteins for their response to PI treatment by immuno- under phase contrast (Figure 3e). To test whether the fluorescent microscopy. All 22 nuclear proteins that aggregate components are soluble, we exposed MG132- produced distinct immunofluorescent signals were found treated cells to treatments with TritonX-100 (TX100) to accumulate to the nucleolar aggregate in the MG132- to extract soluble proteins, or to TX100 and RNase. treated cells (Figure 2 and Table 1). These, in addition TX100 could not disperse polyA( þ ) or p53 (Figure 3f). to stress response proteins, included several cell cycle RNase treatment effectively obliterated the polyA( þ ) proteins, transcription factors and transcriptional reg- signal, whereas p53 staining was still detectable ulators. The extent of translocation between the proteins (Figure 3f). Taken together, these results show that

Oncogene Nucleolar aggregates L Latonen et al 794

Figure 2 Inhibition of proteasome causes nucleolar aggregation of nucleoplasmic proteins. Confocal microsopy of cells mock treated or treated with MG132 for 12 h and stained for the antigens shown (red). Cells were co-stained for p53 (green) and DNA (blue). Scale bars, 10 mm.

Table 1 Proteins detected in nucleolar aggregates p53 and MDM2 are immobilized within the nucleolar aggregates Protein Function If the proteins accumulated to the nucleolar aggregates CDK2 Cell cycle upon PI treatment are forming inactive precipitates, CDK4 Cell cycle they should have decreased motility. To test this, we Cyclin A Cell cycle performed fluorescence recovery after photobleaching Cyclin B1 Cell cycle Cyclin D1 Cell cycle (FRAP) experiments on representative proteins accu- Cyclin E Cell cycle mulating to these structures. Fluorescent fusion proteins Rb Cell cycle for p53 and MDM2 were generated using yellow-variant p21 Cell cycle/CDK inhibitor Venus (Nagai et al., 2002) and for NPM using cyan p27 Cell cycle/CDK inhibitor WEE1A Cell cycle/stress response enhanced green fluorescent protein (Sawano and Miya- SUMO-3 Protein regulation waki, 2000) (Venus-p53, MDM2-Venus and NPM- Ubiquitin Protein regulation ECGPF, respectively). As a test for functionality, we SMAD1/2/3a Signal transduction confirmed the ability of Venus-p53 to cause p53-reporter HDMX E3 Ub-ligase transactivation and the ability of p53 and MDM2- Ku80 Stress response/transcription Chk1 Stress response/signal transduction Venus fusion proteins to form complexes (data not ATM Stress response/signal transduction shown). We also verified that NPM-ECGPF underwent p300 Transcription/histone modification expected localization changes similar to the endogenous p73 Transcription factor protein in response to RNA pol I inhibition and ultra- Sp1 Transcription factor Sp3 Transcription factor violet radiation-induced stress (data not shown). The YY1 Transcription factor constructs were expressed in WS1 cells, after which cells were either mock treated or treated with MG132 for Abbreviations: CDK, cyclin-dependent kinases; SMAD1, Sma- and 12 h. Proteins localized to different compartments of the Mad-related protein 1; SUMO-3, small ubiquitin-related modifier-3; cell were photobleached, and the fluorescence recovery Ub, ubiquitin. in the bleached area was recorded. As expected, Venus- aAntibody recognizes SMADs 1, 2 and 3. p53 and MDM2-Venus localized to the nucleoplasm in control cells and both proteins were observed in the polyA( þ ) RNA is entrapped within the nucleolar nucleolus following proteasome inhibition (Figures 4a aggregate, and that both protein and RNA are in an and b). Nucleoplasmic Venus-p53 and MDM2-Venus insoluble state in this structure. in control cells were highly mobile as indicated by

Oncogene Nucleolar aggregates L Latonen et al 795

Figure 3 Nucleolar aggregates accumulate polyA( þ ) RNA. WS1 cells were mock treated or treated with MG132 (10 mM) for 12 h. (a) Cells were stained for p53 (red) and RNA using SYTO 12 (green). Confocal images are shown. Arrowheads indicate nucleolar aggregates. (b) In situ hybridization of 28S rRNA (red) and co-staining for NPM (green). (c) In situ hybridization of 50ETS rRNA precursor and co-staining of p53. Merged image shows cells co-stained for DNA (blue). (d) In situ hybridization of 28S rRNA and co-staining of p53 (green). (e) In situ hybridization of polyA( þ ) RNA (red) and co-staining for p53 (green). Arrowheads indicate nucleolar aggregates. (f) Cells were treated with TX-100 and RNase before fixation and in situ hybridization for polyA( þ ) RNA and p53 immunostaining. DNA is shown in blue. Scale bars, 10 mm.

fluorescence recovery within 2 min after photobleaching localize to the nucleoli in cells treated with MG132 (Figures 4a and b). Inverse FRAP experiments of (Stavreva et al., 2006). To test whether ubiquitin MG132-treated cells, performed by photobleaching the colocalizes with aggregate proteins, we performed whole nucleus leaving a single aggregate intact, revealed immunofluorescence analysis of ubiquitin and conju- that Venus-p53 and MDM2-Venus were not able to gated ubiquitin with p53 (Figure 5) and cyclin D (data diffuse from the aggregates, indicating that the mobility not shown). As shown in Figures 5a and b, ubiquitin of the proteins is significantly decreased in the aggre- and conjugated ubiquitin accumulated to cytoplasmic gates (Figures 4a and b). In contrast, based on similar aggresomes and within the nucleolar aggregate, the photobleaching experiments of nucleolar NPM- latter of which overlapped with p53. This suggested that ECGPF, the mobility NPM-ECGPF was only slightly the aggregates contain conjugated ubiquitin. Thus, we decreased following proteasome inhibition (Figure 4c). hypothesized that ubiquitylation may affect the locali- The latter finding is consistent with that published zation of the proteins to the aggregates. We first earlier (Stavreva et al., 2006). These results show that confirmed successful extraction of nucleolar prepara- proteins in nucleolar aggregates, but not bona fide tions in control and PI-treated cells. We treated WS1 nucleolar proteins, are immobilized upon PI treatment, cells with MG132 followed by preparation of total confirming that the aggregate proteins are in an inactive, cellular extracts or nucleolar extracts and performed immobile state. western analysis for p53. p53 levels were increased in both total and nucleolar extracts, whereas there was no change in the levels of FBL serving as a nucleolar Aggregate formation is overcome by an increase marker (Figure 6a). We then expressed hemagglutinin in ubiquitin pool (HA)-tagged ubiquitin in HeLa cells and isolated total Upon PI treatment, the levels of polyubiquitylated and nucleolar proteins. Although proteasome inhibition proteins increase (Kisselev and Goldberg, 2001; Navon markedly increased the levels of HA-ubiquitin in the and Ciechanover, 2009). Ubiquitin has been shown to cellular extracts, there was no increase of HA-ubiquitin

Oncogene Nucleolar aggregates L Latonen et al 796

Figure 4 Proteins in nucleolar aggregates are immobile. WS1 cells were transfected with fluorescence-tagged proteins (green) and were treated with MG132 or left untreated. FRAP analysis was performed by selecting a region of interest as indicated. Representative images are shown. Normalized intensities are shown to the right (n ¼ 3 for each analysis, error bars, s.d.). (a) Venus-p53, (b) MDM2- Venus and (c) NPM-ECGFP. Scale bars, 10 mm. A full colour version of this figure is available at the Oncogene journal online.

in the nucleolar fraction (Figure 6b, middle panel). The decreased p53 (Figure 6c) and cyclin D (data not shown) level of nucleostemin, used as a nucleolar loading accumulation to the aggregates. We also expressed a control, did not change (Figure 6b, bottom panel). mutant ubiquitin with conjugation site in However, probing with an antibody that detects three of seven possible lysines (K29,48,63R), including conjugated ubiquitin (FK2) showed that its levels the canonical polyubiquitylation site for proteasome increased also in the nucleolar fraction (Figure 6b, recognition (K48) (Haglund et al., 2003). Expression of upper panel). This suggested that, although nucleolar the mutant HA-Ub-K29,48,63R reduced the aggregate aggregates contain conjugated ubiquitin, ectopic ubi- formation in MG132-treated cells as effectively as HA- quitin is, surprisingly, not retained in the aggregates. ubiquitin (Figure 6d). These results show that an overall We then performed immunofluorescence staining on increase in the ubiquitin pool can overcome aggregate HA-ubiquitin and p53 (Figure 6c). Ectopically expressed formation, and that there may be extensive flexibility of HA-ubiquitin was present throughout the cell, but potential ubiquitin conjugation sites involved. This following proteasome inhibition the localization of indicates that aggregates form, at least in part, due to HA-ubiquitin was primarily cytoplasmic (Figure 6c). the lack of free ubiquitin. Concomitant with the lack of accumulation of ectopic To further address the relevance of ubiquitin con- ubiquitin in the nucleoli, we observed that the forma- jugation in the aggregate formation, we used an tion of aggregates was significantly decreased in cells inhibitor against ubiquitin-like modifying activating expressing significant levels of HA-ubiquitin (Figures 6c enzyme E1 ligase (UBE). If the availability of ubiquitin and d). Furthermore, ectopic expression of ubiquitin determines the aggregation of the proteins, the E1 ligase

Oncogene Nucleolar aggregates L Latonen et al 797

Figure 5 p53 aggregates colocalize with ubiquitin and conjugated ubiquitin. WS1 cells were mock treated or treated with MG132 (10 mM) for 12 h. Cells were stained for (a) ubiquitin (green) and (b) conjugated ubiquitin using FK2 antibody detecting both mono- and polyubiquitin conjugates (green) and p53 (red). Merged images show cells co-stained for DNA (blue). Arrowheads indicate nucleolar aggregates. Scale bars, 50 mm. inhibitor should further increase the aggregate forma- cells and treated the cells with MG132. Quantification of tion. UBE inhibitor used in combination with MG132 p53-containing aggregates in the control and HAUSP- caused such a prominent formation of aggregates that expressing cells indicated that HAUSP expression virtually all cells exhibited nucleolar alterations in phase significantly decreased the accumulation of p53 to the contrast (data not shown). In order to quantify the aggregates (Figure 7a). Conversely, the absence of effect of UBE inhibitor, we used lactacystin treatment of HAUSP should potentiate the formation of p53 the cells, which causes aggregates in approximately 10% aggregates following proteasome inhibition. We hence of the cells at the concentrations used (Figure 1e). As used HCT116 cells with knockdown of HAUSP (USP7) shown in Figure 6e, UBE inhibitor enhanced the expression by homologous recombination (HAUSP/ formation of aggregates by lactacystin by threefold. cells; Cummins et al., 2004). Treatment of HAUSP/ These results indicate that lack of conjugation-compe- cells with MG132 resulted in increased accumulation tent ubiquitin can promote the formation of nucleolar of p53 to the nucleolar aggregates compared with aggregates. the parental HCT116 (Figure 7b and Supplementary To further examine the role of ubiquitin conjugation Table S1). Furthermore, accumulation of MDM2 to in the aggregate formation, we tested whether herpes- the aggregates was markedly enhanced in HAUSP/ virus-associated ubiquitin-specific protease (HAUSP) cells as compared with the parental cell line following deubiquitinase, which cleaves ubiquitin moieties on MG132 treatment (Figure 7c). These results indicate p53 and MDM2 (Li et al., 2002), affects p53 and that HAUSP-targeted ubiquitin conjugation regulates MDM2 accumulation to the aggregates induced by PI p53 and MDM2 accrual to the nucleolar aggregates. treatment. We ectopically expressed HAUSP in HeLa Taken together, these results indicate a role for ubiquitin

Oncogene Nucleolar aggregates L Latonen et al 798

Figure 6 Ubiquitin is rate limiting in the formation of nucleolar aggregates. (a) WS1 cells were treated with MG132 (MG) or left untreated (ctrl). Cellular lysates (total) or nucleolar (No) fractions were prepared, boiled in Laemmle sample buffer and equal amounts were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and blotted for p53 and FBL. (b) HeLa cells were transfected with HA-ubiquitin (Ub) and were treated with MG132 (MG) or left untreated (ctrl). Cellular lysates (total) or nucleolar (No) fractions were prepared and resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and blotted for HA-Ub or antibody detecting conjugated ubiquitin (FK2). Loading, 15:1 nucleolar:total extract. Nucleostemin (GLN3) was used as a loading control. (c) WS1 cells were transfected with HA-Ub and were mock treated or treated with MG132 for 12 h. Cells were fixed and stained for HA-Ub and p53. Arrowheads indicate aggregates. Scale bars, 10 mm. (d) Quantification of nucleolar aggregates of WS1 cells expressing HA-Ub (wt) or HA-Ub-K29,48,63R (Ub-3mut) and treated with MG132 for 12 h (n ¼ 3, error bars, s.d.). **Po0.01; ***Po0.001. (e) Quantification of nucleolar aggregates in WS1 cells treated with lactacystin (LC), UBE1-inhibitor (UBE1-i, 10 mM)or both for 12 h (n ¼ 3, error bars, s.d.). A full colour version of this figure is available at the Oncogene journal online.

in aggregate formation, and suggest that proteins are cells with nucleolar CRM1 expression decreased fol- retained in the nucleolar aggregates due to lack of lowing either MG132 or LMB treatment, and their sufficient pool of ubiquitin. combination (Figures 8a and b). A similar effect was observed following lactacystin and combinatory lactacystin/LMB treatment of the cells (Figure 8b). Formation of nucleolar aggregates by proteasome These results indicate that inhibition of either pro- inhibition is enhanced by inhibition of nuclear export teasome or CRM1 hampers the normal nucleolar Conjugated ubiquitin, especially in the form of mono- localization of CRM1. ubiquitin, has been previously linked to cellular To assess whether inhibition of nuclear export could trafficking (Li et al., 2003; Chen and Sun, 2009). Given affect the formation of aggregates, we treated WS1 cells that many proteasomally targeted nuclear proteins with lactacystin either alone or in combination with are exported to the cytoplasm for degradation via LMB, and assessed aggregate formation under phase CRM1/exportin-1-mediated nuclear export pathway, contrast. Treatment of cells with LMB alone did not and that the CRM1 pathway is involved in the induce nucleolar aggregates (Figures 8c–e), indicating export of and certain mRNAs (Ho et al., that inhibition of nuclear export is insufficient to induce 2000; Gadal et al., 2001; Thomas and Kutay, 2003; nucleolar aggregates under conditions where nuclear Hutten and Kehlenbach, 2007), we tested whether proteasome activity is maintained. However, LMB CRM1-mediated nuclear export is involved in the significantly enhanced the formation of the aggregates aggregate formation consequent to proteasome inhi- by PI treatment. As nearly all cells treated with MG132 bition. First, we studied the effect of proteasome and LMB exhibited nucleolar aggregates (data not inhibition on CRM1. We treated HeLa cells with shown), the potency of LMB to increase nucleolar MG132 either alone or in combination with CRM1 aggregates in combination with lactacystin was quanti- inhibitor leptomycin B (LMB, 5 ng/ml), and stained the fied by phase-contrast imaging and was 2.5-fold cells for CRM1. In control cells, CRM1 was localized (Figure 8c). We further analyzed whether LMB, in com- to the nucleoplasm and nucleoli, but the number of bination with PI treatment, affected the accumulation of

Oncogene Nucleolar aggregates L Latonen et al 799

Figure 7 p53 and MDM2 accumulations are modulated by HAUSP deubiquitinase. (a) Quantification of nucleolar p53 in HeLa cells transfected with control vector (ctrl) or Flag-HAUSP (HAUSP), and treated with MG132 for 12 h (n ¼ 2, error bars, s.d.). *Po0.05. (b, c) HCT116 (wt) and HCT116 HAUSP/ (HAUSP/) cells were treated with MG132 (10 mM) for 12 h or left untreated (control), and were co-stained for p53 (b), MDM2 (c) and DNA. Arrowheads indicate nucleolar aggregates. Scale bars, 10 mm. A full colour version of this figure is available at the Oncogene journal online. p53 and polyA( þ ) RNA to the aggregates. The number as model proteins, that the aggregate represents of cells containing p53 and polyA( þ ) aggregates by immobilized and insoluble proteins. We further show MG132 was substantially enhanced by LMB (Figures 8d that these aggregates form owing to the lack of and e). Similar results were obtained using antibody availability of conjugation-competent ubiquitin under against cyclin D (data not shown). Thus, impaired proteasome inhibition. Free ubiquitin is required to nuclear export has an additive effect on the formation of prevent formation of nucleolar aggregates, likely aggregates by proteasome inhibition. through promoting nucleocytoplasmic export. These results provide mechanistic insight into the events taking place during proteasome inhibition, and identify a Discussion common mechanism for the inactivation of several cancer-related proteins during PI treatment. Proteasome inhibition leads to an increase in polyubi- Several stress-associated proteins (Hsp70, PML, p53, quitylated and misfolded proteins, induces cell death MDM2) have previously been described to undergo and is considered an effective strategy in cancer therapy. nucleolar localization following proteasome stress In this work, we studied the molecular events taking (Klibanov et al., 2001; Mattsson et al., 2001; Latonen place in the nucleus of PI-treated cells. We identify here et al., 2003; Kurki et al., 2004; Karni-Schmidt et al., a nuclear aggregate, NoA, which forms in the nucleolus. 2007). Here we significantly extend the previous find- We show that the nucleolus undergoes major reorgani- ings, and expand the number and type of proteins zation following inhibition of the proteasome, as the to include cyclins, CDKs, transcription factors, stress- nucleolar structures and activity surround a dense inducible kinases and signaling molecules. These repre- aggregate that contains proteins and polyA( þ ) RNA. sent central, highly regulated proteins functioning in Strikingly, we observe aggregation of over 20 proteaso- stress responses, cell cycle and transcription. Their mally degraded nucleoplasmic proteins in this structure, common denominator is that all are nucleoplasmic several of which are key cellular regulators and relevant proteins that can be targeted to proteasomal degradation, for cancer. We provide evidence, using p53 and MDM2 and most of them have been shown to be ubiquitylated.

Oncogene Nucleolar aggregates L Latonen et al 800

Figure 8 Inhibition of nuclear export enhances accumulation of p53 and polyA( þ ) RNA to aggregates. (a) HeLa cells were treated with MG132 (10 mM), LMB (5 ng/ml) or both, or left untreated. CRM1 was detected by immunostaining. Scale bars, 10 mm. (b) HeLa cells were treated with lactacystin (LC, 10 mM) or MG132 (MG, 10 mM) either alone or in combination with LMB (5 ng/ml). Cells with the expression of nucleolar CRM1 were quantified (n ¼ 2–3, error bars, s.d.). *Po0.05; **Po0.01; ***Po0.001. (c) WS1 cells were treated with lactacystin (LC, 10 mM), LMB (5 ng/ml) or both. Cells with aggregates were quantified (n ¼ 4, error bars, s.d.). (d) WS1 cells treated as in (a) were stained for p53. Scale bars, 20 mm. (e) In situ hybridization for polyA( þ ) RNA (red) of WS1 cells treated as in (a). Cells were co-stained for NPM (green) and DNA (blue). Merged images are shown. Scale bars, 10 mm. (d, e) Arrowheads indicate nucleolar aggregates.

The NoAs resemble cytoplasmic aggresomes, as both (r) proteins are ubiquitylated or translated as fusion contain aggregated proteasome target proteins. How- proteins with ubiquitin (Finley et al., 1989; Spence et al., ever, whereas the cytoplasmic aggresomes accumulate 2000; Matsumoto et al., 2005). Defects in ubiquitin cytoplasmic proteins (Johnston et al., 1998), the NoAs conjugation or reduced deubiquitylation alter the seem specific for nuclear proteins. Thus, it seems that the nucleolar structure (Sudha et al., 1995; Endo et al., NoAs are a nuclear counterpart of cytoplasmic aggre- 2009). Ubiquitin conjugation is also an essential step in somes, and that both bodies can coexist in the same the quality control of rRNA (Fujii et al., 2009). cells. The most striking difference between these stress- Inhibition of the proteasome activity results in the induced bodies is that RNA, detected in NoAs, is not accumulation of ribosomal proteins in the nucleus, present in cytoplasmic aggresomes. As the proteins in suggesting that ribosomal proteins that have failed to NoA are virtually immobile and insoluble, proteasome assemble into ribosome subunits undergo proteasomal inhibitor treatment may result in the loss of function of degradation (Lam et al., 2007). We find here that many cellular activities through irrevocable retention of ubiquitin serves as a highly central modifier of NoA key proteins in the NoA. formation and protein mobility. NoAs contained The ubiquitin–proteasome system has several pre- conjugated ubiquitin as determined both by immuno- viously identified links to nucleoli and ribosome fluorescence and biochemical analysis of the nucleoli, biosynthesis. Ubiquitin is present in the nucleoli suggesting the accumulation of ubiquitin-tagged (Stavreva et al., 2006; this work), and several ribosomal proteins either as mono- or polyubiquitin conjugates

Oncogene Nucleolar aggregates L Latonen et al 801 or both. However, it is possible that some of the NoA studies (Fornerod et al., 1997; Daelemans et al., 2005; proteins are not modified by ubiquitylation, but directed Ernoult-Lange et al., 2009). CRM1 pathway is involved to these structures via protein-protein interactions. By in the export of ribosomes and certain mRNAs (Ho the use of HAUSP deubiquitinase, we further showed et al., 2000; Gadal et al., 2001; Thomas and Kutay, that HAUSP-driven deubiquitylation of p53 decreases 2003; Hutten and Kehlenbach, 2007), and for transport its localization to NoAs. Thus, at least in the case of of small nucleolar RNA from Cajal bodies to the p53, the accumulation to nucleolar aggregates is likely nucleolus (Boulon et al., 2004). Inhibition of CRM1 mediated by ubiquitylation. leads to 40S and 60S ribosome subunit retention in the Ubiquitin, however, seems to have a dual role in the nucleoli (Ho et al., 2000; Gadal et al., 2001; Thomas and formation of nucleolar aggregates. Ectopic expression of Kutay, 2003). Based on these results, the nucleoli may ubiquitin effectively rescued NoA formation, an effect serve as a platform promoting CRM1 associations with that we linked to enhanced cytoplasmic export of certain cargo, including ribosomal particles. However, ubiquitin. Interestingly, p53 targeted to degradation inhibition of CRM1 alone did not induce the formation has previously been suggested to be exported via the of NoAs, indicating that the inability to transport ribo- nucleolus (Sherr and Weber, 2000; Rubbi and Milner, some particles alone is insufficient for NoA formation. 2003), and the stress-induced nuclear accumulation of When CRM1 is inhibited, but the nuclear proteasomes p53 could depend on the loss of nucleolar integrity remain functional, protein degradation and ubiquitin (Rubbi and Milner, 2003). In addition, p53 trafficking is recycling can take place in the nucleus despite defective known to rely on ubiquitylation; monoubiquitylation nucleocytoplasmic export. When nuclear proteasomes promotes nuclear export of p53, whereas high levels of are inhibited, ubiquitylated proteins may be directed to MDM2 activity promote polyubiquitylation and nuclear the export pathway; however, exhaustion of ubiquitin degradation of p53 (Li et al., 2003). In this work, prevents the later stages of export and induces forma- ectopically expressed ubiquitin caused dissolution of p53 tion of nucleolar aggregates. Concomitant inhibition of and cyclin D from the aggregates. These results indicate nuclear export with proteasome inhibition could thus that as proteins may need conjugated ubiquitin to be promote aggregate formation by accelerating the accu- targeted to the nucleolus, more ubiquitin is needed for mulation of degradation targets. proteins to exit. This suggests that complex, sequential NoA contained polyA( þ ) RNA, but not mature or ubiquitylation events are required for export of protea- precursor rRNA, and lacked nascent rRNA synthesis. some targets, at least under conditions where nuclear Further studies are required to resolve the identity of the proteasomes are inactive. We cannot currently identify polyA( þ ) RNA accumulating in the nucleolar aggre- the type or types of ubiquitylation patterns involved, gates, as polyA( þ ) RNA may represent either nascent and whether mono- or polyubiquitylation, or both, are mRNA exported via the CRM1 route or RNA destined required. Expression of a triple-mutant ubiquitin shows for degradation by polyA( þ ) tag. However, the finding that the ‘classical’ proteasome-targeting K48 link is not that ubiquitin is rate limiting for the aggregate formation essential for ubiquitin to mitigate the aggregate forma- suggests that ubiquitin-mediated licensing is required for tion, nor are the K63 and K29 links. However, four RNA surveillance and export also in mammalian cells, additional lysines on ubiquitin are competent of ubi- as previously described in yeast (Thomsen et al., 2008; quitylation, in addition to the C-terminal glycine which Houseley and Tollervey, 2009). mediates monoubiquitylation. Numerous possibilities Impairment of the proteasome function and aggre- thus exist as to how ubiquitin could serve to relieve somes are linked with aging-related neurodegenerative nucleolar aggregate formation. In addition, recent diseases, and aggresome-like structures are formed in literature describes increasing complexity of possible the nucleoplasm in relation to Huntington’s disease ubiquitylation chains, and even heterologous chains (Klement et al., 1998; Saudou et al., 1998). Certain involving ubiquitin and ubiquitin-like modifiers, such neurodegenerative diseases, such as polyglutamine as small ubiquitin-related modifier (SUMO) (Ikeda (polyQ) repeat diseases, are characterized by the and Dikic, 2008). We and Mattsson et al. (2001) have aggregation of misfolded proteins to nuclear aggregates detected SUMO accumulating to the nucleoli upon in DNA-sparse areas resembling the nucleoli (Bennett proteasome inhibition, and a recent proteomic study et al., 2005). PolyQ variant of ataxin-1 forms nuclear provides evidence of SUMO conjugates accumulated in aggregates that co-stain with ubiquitin (Bennett et al., the nucleoli upon PI treatment (Matafora et al., 2009). 2005). Ataxin-1 binds RNA depending on the polyQ Furthermore, SUMO-1 modification can promote cyto- tract length and has been proposed to participate in plasmic export of p53 in concert with ubiquitylation RNA metabolism (Yue et al., 2001). Moreover, in (Carter et al., 2007; Carter and Vousden, 2008). In the Drosophila, RNA increases toxicity of the ataxin-3 future, it will be interesting to determine whether a polyQ repeat (Li et al., 2008). Lastly, HAUSP has been specific ubiquitin and/or ubiquitin-like modification shown to bind ataxin-1 and to lose this interaction with signature exists targeting proteasome targets from the the polyQ variant (Hong et al., 2002). Although these nucleus to the cytoplasm via the nucleolus. are presently circumstantial observations, we raise the Inhibition of CRM1-dependent nuclear export possibility that these neuronal aggregates may represent potentiated NoA formation. Moreover, proteasome the nucleolus-associated aggregates described here. inhibition caused exclusion of CRM1 from the nucleoli. Taken together, these findings may indicate that the CRM1 nucleolar localization has been noted in previous underlying causes in protein aggregation diseases may

Oncogene Nucleolar aggregates L Latonen et al 802 be exacerbated by defects in RNA processing in addition CRM1); Bethyl Laboratories (Montgomery, TX, USA; to ubiquitin turnover. HDMX), Epitomics (Burlingame, CA, USA; SMA, E184); In conclusion, we define here a novel nucleolus- Europa Bioproducts Ltd (Cambridge, UK; GAPDH, 9.B.88), dependent structure that arises consequent to severe DAKO (Glostrup, Denmark; Ki-67, MIB-1); Neomarkers overload of nuclear proteasome targets. We propose (Thermo Scientific, Fremont, CA, USA; cyclin B1, Ab-3; that these nucleolar aggregates accumulate polyA( ) cyclin D1, Ab-4; p73, Ab5); Pharmingen (Franklin Lakes, þ NJ, USA; Ku80, M040337; retinoblastoma, 14001, 14006); RNA destined to degradation and ubiquitylated nuclear Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA; proteins targeted to nucleocytoplasmic export. The data ATM, 2C1; CDK2, M2; CDK4, C-22; Chk1, G-4; cyclin A, presented here identify a remarkable number of cancer- C-19; cyclin E, C-19; IkB, C-21; MDM2, 2A10, SMP-14; related proteins accumulating in these PI treatment- p21, M-19, C-19; p27, C-19; p300, N-15; p53, DO-1, FL-393; induced stress bodies. This set of factors most likely PML, H-238; SMAD1/2/3, H-2; Sp1, 1C6; Sp3, D-20; represent only a fraction of proteins targeted to ubiquitin, P4D1; Wee1A, H-300; YY1, H-414); and Zymed nucleolar aggregates upon proteasome inhibition, and (San Francisco, CA, USA; SUMO-1, 33–2400; SUMO-3, suggest that the event is common among nuclear 51–9100). For quantification, a minimum of 300 non- proteasome targets. Our results identify a close link transfected or 100 transfected cells were analyzed per sample. between nuclear export, nucleoli and the proteasome, The nucleolar aggregates were quantified by counting the percentage of cells with dense nucleolar aggregates visible and provide mechanistic insight into how PI treatment under phase-contrast imaging. All analyses were performed affects several of the key regulators in cancer. from two to four independent experiments. Results were scored according to their statistical significance using Student’s one-tailed t-test (*Po0.05; **Po0.01; ***Po0.001). The fluorochromes were visualized either with Zeiss Materials and methods Axioplan 2 Imaging MOT (Zeiss, Jena, Germany) equipped with 20/0.5 NA or 400.75 NA Plan-Neofluar objectives Cell culture, chemicals and treatments and Chroma 31000v2, Chroma 41001, and Chroma 41004 WS1 skin fibroblasts (CRL-1502, ATCC, Manassas, VA, filters (Chroma, Bellows Falls, VT, USA), or Zeiss AxioImager USA) and HEL-299 lung fibroblasts (CCL-137, ATCC) were MOT Z1 (Jena), equipped with EC-PlanNeofluar 40/0.75 maintained in Dulbecco’s modified Eagle’s medium supple- objectives, and Zeiss filters sets 38, 45 and 49. Images were mented with 10% fetal calf serum and non-essential amino captured with Zeiss AxioCam HRm 14-bit grayscale CCD acids, HeLa cells in high glucose Dulbecco’s modified Eagle’s camera and AxioVision program version 4.6. medium and 10% fetal calf serum, and HCT116 cells (wild Confocal imaging was performed with Zeiss LSM510 type and HAUSP / ) in Dulbecco’s modified Eagle’s medium META microscope equipped 63/1.25 NA Plan-Neofluar and 10% fetal calf serum. HAUSP/ cells were a kind gift objective, diode, argon and HeNe lasers. Emissions were from Dr B Vogelstein (Johns Hopkins University, Baltimore, detected with the following filter settings: BP 420–480 for 1 MD, USA). Cells were kept at þ 37 C in a humidified Hoechst, BP 505–530 for Alexa 488 and LP 585 for Alexa 594. atmosphere containing 5% CO2. All cell culture reagents were HFT405/488/543 was used as dichroic beam splitter and obtained from Gibco-BRL (Rockville, MD, USA). Chemicals NFT545 as emission splitter. used in the treatments were MG132 (Z-Leu-Leu-Leu-CHO; Biomol, Enzo Life Sciences Inc., Farmingdale, NY, USA), ALLN (MG101/calpain inhibitor I/N-Ac-Leu-Leu-norleucinal, TEM Calbiochem, San Diego, CA, USA), lactacystin (clasto- Cells were harvested by pelleting and fixed by 2.5% lactacystin-b-lactone, Calbiochem), E64 (Sigma, Sigma- glutaraldehyde. Samples were post-fixed by 1% osmium Aldrich, St Louis, MO, USA), leupeptin (Sigma), Act D (Sigma), tetroxide for 1 h at room temperature, dehydrated in graded LMB (Calbiochem) and UBEI-41 (Biogenova, Rockville, MD, ethanol and embedded in epoxy resin LX-112. Serial sections USA). For combination treatments, cells were pre-incubated with were cut at 60 nm and stained with uranyl acetate and lead Act D, UBEI or LMB of 1 h, after which PIs were added. citrate using Leica EM STAIN automatic stainer (Leica Microsystems GmbH, Wetzlar, Germany). The sections were examined with a JEM-1400 (Jeol) TEM at 80 kV at 5000 Immunofluorescence analysis magnification. Images were captured using Olympus-SIS Cells grown on glass coverslips were fixed with 3.5% Morada digital camera (Olympus, Mu¨nster, Germany). paraformaldehyde and permeabilized with 0.5% NP-40 in phosphate-buffered saline (PBS) at room temperature or, alternatively, were fixed with methanol and permeabilized with Transient transfections acetone at –20 1C. For in vivo labeling of nascent RNA Transfections to WS1 cells were performed using electro- synthesis, cells were treated with 1 mM 5-fluorouridine (Sigma) poration essentially as described earlier (Latonen et al., 2003). for 30 min, fixed with 1% paraformaldehyde for 5 min, Cells were trypsinized, suspended to Opti-MEM (Gibco BRL), permeabilized with 0.5% Triton-X100 for 10 min and detected DNA was added and the mixture was electropulsed (300V, using anti-bromodeoxyuridine antibody (B2531, Sigma). 800 mF) with Pulser II (BioRad, Hercules, CA, USA). Primary antibodies were detected using fluorescence-conju- HeLa cells were transfected using JetPei transfection reagent gated secondary antibodies (Molecular Probes, Invitrogen, (Polyplus-transfection Inc., New York, NY, USA) according Carlsbad, CA, USA). DNA was stained with Hoechst 33258 to the manufacturer’s protocol. (Sigma). RNA was detected with SYTO 12 (Molecular NPM-ECGFP fusion protein was generated by excision of Probes). The antibodies were obtained from the following NPM1 cDNA from B23-GFP (a kind gift from Dr M Olson, sources, Abcam (Cambridge, UK; ARF, 14PO2; FBL); University of Mississippi Medical Center, MS, USA; Dundr Affiniti Research Products and Biomol (now Enzo Life et al., 2000) and ligation to ECGFP-pRSETb (a kind gift from Sciences; FK2, S20); BD Transduction Laboratories (BD Dr A Miyawaki, Brain Science Institute, RIKEN, Saitama, Transduction Laboratories, Franklin Lakes, NJ, USA; Japan; Sawano and Miyawaki, 2000). The construct was

Oncogene Nucleolar aggregates L Latonen et al 803 further subcloned to pCDNA3.1 þ (Invitrogen) to yield NPM- CGAACCTCCGA; 50ETS, GACGTCACCACATCGATCG ECGFP. Venus-p53 was generated by ligating p53 cDNA to AAG; 5.8S, TTCTTCATCGACGCACGAGCCG; and C-terminus of Venus/pCS2 (a kind gift from Dr A Miyawaki; polyT(25)Vn, TTTTTTTTTTTTTTTTTTTTTTTTTVN. All Nagai et al., 2002). MDM2-Venus was generated by ligating probes were from Exiqon (Vedbaek, Denmark). In indicated MDM2 insert to N-terminus of Venus/pCS2 to generate experiments, the cells were incubated with 0.05% TX-100 and MDM2-Venus in-frame fusion. Expression vectors for ubiqui- 0.05% Tween-20 in PBS for 10 min, followed by RNase tin (HA-Ub-wt/pcDNA3 and HA-Ub-K29, 48, 63R/pcDNA3) treatment (1 mg/ml, Roche) for 30 min at room temperature were a kind gift from Dr I Dikic (Goethe University, before fixation and hybridization as above. Frankfurt, Germany; Haglund et al., 2003), and pCIneo- HAUSP-Flag (USP7) vector was kindly provided by Dr B Immunoblotting Vogelstein (Johns Hopkins University, Baltimore, MD, USA; Cells were lysed in RIPA buffer (50 mM Tris–HCl (pH 7.5), Cummins et al., 2004). 150 mM NaCl, 1% NP-40, 0.1% sodium dodecyl sulfate, 1% sodium deoxycholate containing 1 mM Na3VO4,1mM phenyl- FRAP methanesulfonyl fluoride, 1 mM dithiothreitol, 20 mM N- Before imaging, normal growth medium was replaced with ethylmaleimide and 10 mg/ml of each E64, leupeptin and Dulbecco’s modified Eagle’s medium without phenol red SBTI) to obtain total cellular lysates. Nucleolar fractionation (Gibco-BRL). In MG132-treated cells, MG132 was present was performed as described previously (Andersen et al., 2005). throughout imaging. FRAP was performed using LSM 510 N-ethylmaleimide (20 mM) was added to extraction buffers in META confocal laser scanning microscope (Zeiss) equipped experiments assessing ubiquitin conjugation. Proteins were with a heating stage. Argon laser line (488 nm) was set at 100% separated by 5% sodium dodecyl sulfate–polyacrylamide gel during the bleaching and at 2% during imaging with 50% electrophoresis, and immunoblotting was performed as de- output using Plan-Neofluar 63/1.25 NA oil-immersion scribed previously (Latonen et al., 2003). The following objective. The region of interest was bleached after three pre- antibodies were used: nucleostemin (H-270; Santa Cruz bleach scans with 30 iterations, and 97 post-bleach images Biotechnology) and HA.11 (16B12; Covance, Princeton, NJ, were captured. For each fusion protein, at least three USA). Horse radish peroxidase-conjugated secondary anti- independent experiments were performed. Fluorescence in- bodies were from DAKO. tensities were measured using the LSM 510 Physiology Software. Data from three representative cells for each fusion protein and from each condition were used for the FRAP analysis. Recovery curves from untreated, nucleoplasmic Abbreviations protein intensities were corrected to total cell intensity and to scan controls. Act D, actinomycin D; ATM, ataxia telangiectasia mutated; CDK, cyclin-dependent kinase; CRM1, chromosome region maintenance In situ hybridization 1 protein homolog; ER, endoplastic reticulum; ETS, external Cells grown on coverslips were fixed in 4% paraformaldehyde transcribed spacer; FBL, fibrillarin; FUrd, fluorouridine; HA, with 10% acetic acid in diethylpyrocarbonate-treated water for hemagglutinin; HAUSP, herpes virus-associated ubiquitin-specific 20 min. The cells were washed three times in PBS and protease; LMB, leptomycin B; LNA, locked nucleic acid; NoA, nucleolar aggregate; NPM, nucleophosmin; PI, proteasome immersed in cold 70% ethanol overnight. Cells were then rehydrated in PBS for 10 min and pre-hybridized in 40% inhibitor; Rb, retinoblastoma; RNA pol I, RNA polymerase I; polyA( ), polyadenylated RNA; polyQ, polyglutamine; TEM, formamide (Sigma) in 2 SS (sodium chloride–sodium phos- þ phate–ethylenediaminetetraacetic acid buffer) for 20 min. transmission; TX100, Triton X-100; UBF, upstream binding factor. Locked nucleic acid probes were diluted in hybridization buffer (50% formamide, 5 SSC, 250 mg/ml Escherichia coli tRNA (Roche, Basel, Switzerland), 500 mg/ml salmon sperm Conflict of interest DNA (Invitrogen), 2% Roche blocking reagent (Roche), 0.02% Tween-20, 0.05% CHAPS (Sigma) in diethylpyrocar- The authors declare no conflict of interest. bonate-treated water) and incubated at 37 1C for 5 h. Cover- slips were washed with 5 SSC for 15 min at 37 1C, twice for 35 min each at 37 1C in 0.2 SSC and then once in PBS for Acknowledgements 15 min atroom temperature. Coverslips were blocked in 4% sheep serum and 3% bovine serum albumin in PBS for 1 h and We thank Drs M Olson, A Miyawaki, I Dikic, B Vogelstein incubated overnight in anti-digoxigenin-AP (Roche) solution and A Salminen for providing reagents and cell lines. M Salo at 4 1C followed by washing seven times in PBST (0.01% and H Liu are thanked for excellent technical assistance. Tween in PBS) for 5 min and once in PBS. For detection of Members of Laiho lab at Helsinki and Hopkins, O Matilainen AP, Fast Red tablets (Roche) were dissolved in 0.1 M Tris–HCl and C Holmberg are thanked for helpful discussions. University (pH 8.2) and the color reaction was carried out in dark for of Helsinki Molecular Imaging Unit is thanked for assistance in up to 5 h. Finally, the coverslips were stained for DNA image acquisition. University of Helsinki Advanced Imaging using Hoechst 33258 and mounted in VECTASHIELD Unit is thanked for assistance in TEM imaging and sample mounting medium (Vector Laboratories, Burlingame, CA, preparation. This work was supported by Academy of Finland USA). The following LNA probe sequences were used: 28S, (ML Grant No. 129699, LL Grant No. 108828), Biocentrum CCTTAGAGCCAATCCTTATCCC; 18S, CTGATCGTCTT Helsinki and Helsinki Biomedical Graduate School (HMM).

References

Andersen JS, Lam YW, Leung AK, Ong SE, Lyon CE, Lamond AI Bennett EJ, Bence NF, Jayakumar R, Kopito RR. (2005). Global et al. (2005). Nucleolar proteome dynamics. Nature 433: 77–83. impairment of the ubiquitin–proteasome system by nuclear or

Oncogene Nucleolar aggregates L Latonen et al 804 cytoplasmic protein aggregates precedes inclusion body formation. Ikeda F, Dikic I. (2008). Atypical ubiquitin chains: new molecular Mol Cell 17: 351–365. signals. ‘Protein Modifications: beyond the usual suspects’ review Boisvert FM, van Koningsbruggen S, Navascue´s J, Lamond AI. series. EMBO Rep 9: 536–542. (2007). The multifunctional nucleolus. Nat Rev Mol Cell Biol 8: Johnston JA, Ward CL, Kopito RR. (1998). Aggresomes: a cellular 574–585. response to misfolded proteins. J Cell Biol 143: 1883–1898. Boulon S, Verheggen C, Jady BE, Girard C, Pescia C, Paul C et al. Karni-Schmidt O, Friedler A, Zupnick A, McKinney K, Mattia M, (2004). PHAX and CRM1 are required sequentially to transport U3 Beckerman R et al. (2007). Energy-dependent nucleolar localization snoRNA to nucleoli. Mol Cell 16: 777–787. of p53 in vitro requires two discrete regions within the p53 carboxyl Carter S, Bischof O, Dejean A, Vousden KH. (2007). C-terminal terminus. Oncogene 26: 3878–3891. modifications regulate MDM2 dissociation and nuclear export of Kisselev AF, Goldberg AL. (2001). Proteasome inhibitors: from p53. Nat Cell Biol 9: 428–435. research tools to drug candidates. Chem Biol 8: 739–758. Carter S, Vousden KH. (2008). p53-Ubl fusions as models of Kiviharju-af Ha¨llstro¨mTM,Ja¨a¨maa S, Mo¨nkko¨nen M, Peltonen K, ubiquitination, sumoylation and neddylation of p53. Cell Cycle 7: Andersson LC, Medema RH et al. (2007). Human prostate 2519–2528. epithelium lacks Wee1A-mediated DNA damage-induced check- Chen ZJ, Sun LJ. (2009). Nonproteolytic functions of ubiquitin in cell point enforcement. Proc Natl Acad Sci USA 104: 7211–7216. signaling. Mol Cell 33: 275–286. Klement IA, Skinner PJ, Kaytor MD, Yi H, Hersch SM, Clark HB Cummins JM, Rago C, Kohli M, Kinzler KW, Lengauer C, Vogelstein et al. (1998). Ataxin-1 nuclear localization and aggregation: role in B. (2004). Tumour suppression: disruption of HAUSP gene polyglutamine-induced disease in SCA1 transgenic mice. Cell 95: 41–53. stabilizes p53. Nature 428: 486–487. Klibanov SA, O’Hagan HM, Ljungman M. (2001). Accumulation of Daelemans D, Costes SV, Lockett S, Pavlakis GN. (2005). Kinetic and soluble and nucleolar-associated p53 proteins following cellular molecular analysis of nuclear export factor CRM1 association with stress. J Cell Sci 114: 1867–1873. its cargo in vivo. Mol Cell Biol 25: 728–739. Kopito RR. (2000). Aggresomes, inclusion bodies and protein Dundr M, Misteli T, Olson MO. (2000). The dynamics of postmitotic aggregation. Trends Cell Biol 10: 524–530. reassembly of the nucleolus. J Cell Biol 150: 433–446. Kurki S, Peltonen K, Laiho M. (2004). Nucleophosmin, HDM2 and Endo A, Matsumoto M, Inada T, Yamamoto A, Nakayama KI, p53: players in UV damage incited nucleolar stress response. Cell Kitamura N et al. (2009). Nucleolar structure and function are Cycle 3: 976–979. regulated by the deubiquitylating enzyme USP36. J Cell Sci 122: Lafontaine DL, Tollervey D. (2001). The function and synthesis of 678–686. ribosomes. Nat Rev Mol Cell Biol 2: 514–520. Ernoult-Lange M, Wilczynska A, Harper M, Aigueperse C, Dautry F, Lam YW, Lamond AI, Mann M, Andersen JS. (2007). Analysis of Kress M et al. (2009). Nucleocytoplasmic traffic of CPEB1 and nucleolar protein dynamics reveals the nuclear degradation of accumulation in Crm1 nucleolar bodies. Mol Biol Cell 20: 176–187. ribosomal proteins. Curr Biol 17: 749–760. Fatica A, Tollervey D. (2002). Making ribosomes. Curr Opin Cell Biol Latonen L, Kurki S, Pitka¨nen K, Laiho M. (2003). p53 and MDM2 14: 313–318. are regulated by PI-3-kinases on multiple levels under stress Finley D. (2009). Recognition and processing of ubiquitin–protein induced by UV radiation and proteasome dysfunction. Cell Signal conjugates by the proteasome. Annu Rev Biochem 78: 477–513. 15: 95–102. Finley D, Bartel B, Varshavsky A. (1989). The tails of ubiquitin Leary DJ, Huang S. (2001). Regulation of ribosome biogenesis within precursors are ribosomal proteins whose fusion to ubiquitin the nucleolus. FEBS Lett 509: 145–150. facilitates ribosome biogenesis. Nature 338: 394–401. Li M, Brooks CL, Wu-Baer F, Chen D, Baer R, Gu W. (2003). Mono- Fornerod M, van Deursen J, van Baal S, Reynolds A, Davis D, Murti versus polyubiquitination: differential control of p53 fate by Mdm2. KG et al. (1997). The human homologue of yeast CRM1 is in a Science 302: 1972–1975. dynamic subcomplex with CAN/Nup214 and a novel nuclear pore Li M, Chen D, Shiloh A, Luo J, Nikolaev AY, Qin J et al. (2002). component Nup88. EMBO J 16: 807–816. Deubiquitination of p53 by HAUSP is an important pathway for Fujii K, Kitabatake M, Sakata T, Miyata A, Ohno M. (2009). A role p53 stabilization. Nature 416: 648–653. for ubiquitin in the clearance of nonfunctional rRNAs. Dev Li LB, Yu Z, Teng X, Bonini NM. (2008). RNA toxicity is a component 23: 963–974. of ataxin-3 degeneration in Drosophila. Nature 453: 1107–1111. Gadal O, Strauss D, Kessl J, Trumpower B, Tollervey D, Hurt E. Matafora V, D’Amato A, Mori S, Blasi F, Bachi A. (2009). Proteomics (2001). Nuclear export of 60s ribosomal subunits depends on Xpo1p analysis of nucleolar SUMO-1 target proteins upon proteasome and requires a nuclear export sequence-containing factor, Nmd3p, inhibition. Mol Cell Proteomics 8: 2243–2255. that associates with the large subunit protein Rpl10p. Mol Cell Biol Matsumoto M, Hatakeyama S, Oyamada K, Oda Y, Nishimura T 21: 3405–3415. et al. (2005). Large-scale analysis of the human ubiquitin-related Haglund K, Sigismund S, Polo S, Szymkiewicz I, Di Fiore PP, Dikic I. proteome. Proteomics 5: 4145–4151. (2003). Multiple monoubiquitination of RTKs is sufficient for their Mattsson K, Pokrovskaja K, Kiss C, Klein G, Szekely L. (2001). endocytosis and degradation. Nat Cell Biol 5: 461–466. Proteins associated with the promyelocytic leukemia gene product Hernandez-Verdun D. (2006). Nucleolus: from structure to dynamics. (PML)-containing nuclear body move to the nucleolus upon Histochem Cell Biol 125: 127–137. inhibition of proteasome-dependent protein degradation. Proc Natl Hershko A, Ciechanover A. (1998). The ubiquitin system. Annu Rev Acad Sci USA 98: 1012–1017. Biochem 67: 425–479. Nagai T, Ibata K, Park ES, Kubota M, Mikoshiba K, Miyawaki A. Ho JH, Kallstrom G, Johnson AW. (2000). Nmd3p is a Crm1p- (2002). A variant of yellow fluorescent protein with fast and dependent adapter protein for nuclear export of the large ribosomal efficient maturation for cell-biological applications. Nat Biotechnol subunit. J Cell Biol 151: 1057–1066. 20:87–90. Holmberg CI, Illman SA, Kallio M, Mikhailov A, Sistonen L. (2000). Navon A, Ciechanover A. (2009). The 26 S proteasome: from basic Formation of nuclear HSF1 granules varies depending on stress mechanisms to drug targeting. J Biol Chem 284: 33713–33718. stimuli. Cell Stress Chaperones 5: 219–228. Olson MO, Dundr M, Szebeni A. (2000). The nucleolus: an old factory Hong S, Kim SJ, Ka S, Choi I, Kang S. (2002). USP7, a ubiquitin- with unexpected capabilities. Trends Cell Biol 10: 189–196. specific protease, interacts with ataxin-1, the SCA1 gene product. Reyes-Turcu FE, Ventii KH, Wilkinson KD. (2009). Regulation and Mol Cell Neurosci 20: 298–306. cellular roles of ubiquitin-specific deubiquitinating enzymes. Annu Houseley J, Tollervey D. (2009). The many pathways of RNA Rev Biochem 78: 363–397. degradation. Cell 136: 763–776. Rubbi CP, Milner J. (2003). Disruption of the nucleolus mediates Hutten S, Kehlenbach RH. (2007). CRM1-mediated nuclear export: to stabilization of p53 in response to DNA damage and other stresses. the pore and beyond. Trends Cell Biol 17: 193–201. EMBO J 22: 6068–6077.

Oncogene Nucleolar aggregates L Latonen et al 805 Saeki Y, Kudo T, Sone T, Kikuchi Y, Yokosawa H, Toh-e A et al. cycle arrest owing to the temperature-sensitive ubiquitin-activating (2009). Lysine 63-linked polyubiquitin chain may serve as a enzyme E1. Chromosome Res 3: 115–123. targeting signal for the 26S proteasome. EMBO J 28: 359–371. Thomas F, Kutay U. (2003). Biogenesis and nuclear export of Saudou F, Finkbeiner S, Devys D, Greenberg ME. (1998). ribosomal subunits in higher eukaryotes depend on the CRM1 Huntingtin acts in the nucleus to induce apoptosis but death export pathway. J Cell Sci 116: 2409–2419. does not correlate with the formation of intranuclear inclusions. Thomsen R, Saguez C, Nasser T, Jensen TH. (2008). General, rapid, Cell 95: 55–66. and transcription-dependent fragmentation of nucleolar antigens in Sawano A, Miyawaki A. (2000). Directed evolution of green S.cerevisiae mRNA export mutants. RNA 14: 706–716. fluorescent protein by a new versatile PCR strategy for site-directed Weissman AM. (2001). Themes and variations on ubiquitylation. Nat and semi-random mutagenesis. Nucleic Acids Res 28: E78. Rev Mol Cell Biol 2: 169–178. Sherr CJ, Weber JD. (2000). The ARF/p53 pathway. Curr Opin Genet Wo´jcik C, Schroeter D, Wilk S, Lamprecht J, Paweletz N. (1996). Dev 10: 94–99. Ubiquitin-mediated proteolysis centers in HeLa cells: indication Spence J, Gali RR, Dittmar G, Sherman F, Karin M, Finley D. (2000). from studies of an inhibitor of the chymotrypsin-like activity of the Cell cycle-regulated modification of the ribosome by a variant proteasome. Eur J Cell Biol 71: 311–318. multiubiquitin chain. Cell 102: 67–76. Xu P, Duong DM, Seyfried NT, Cheng D, Xie Y, Robert J et al. Stavreva DA, Kawasaki M, Dundr M, Koberna K, Mu¨ller WG, (2009). Quantitative proteomics reveals the function of unconven- Tsujimura-Takahashi T et al. (2006). Potential roles for ubiquitin tional ubiquitin chains in proteasomal degradation. Cell 137: 133–145. and the proteasome during ribosome biogenesis. Mol Cell Biol 26: Yue S, Serra HG, Zoghbi HY, Orr HT. (2001). The spinocerebellar 5131–5145. ataxia type 1 protein, ataxin-1, has RNA-binding activity that is Sudha T, Tsuji H, Sameshima M, Matsuda Y, Kaneda S, Nagai Y inversely affected by the length of its polyglutamine tract. Hum Mol et al. (1995). Abnormal integrity of the nucleolus associated with cell Genet 10: 25–30.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene