SUMO regulates p21Cip1 intracellular distribution and with p21Cip1 facilitates multiprotein complex formation in the nucleolus upon DNA damage

We previously showed that p21Cip1 transits through the nucleolus on its way from the nucleus to the cytoplasm and that DNA damage inhibits this transit and induces the formation of p21Cip1-containing intranucleolar bodies (INoBs). Here, we demonstrate that these INoBs also contain SUMO-1 and UBC9, the E2 SUMO-conjugating enzyme. Furthermore, whereas wild type SUMO-1 localized in INoBs, a SUMO-1 mutant, which is unable to conjugate with proteins, does not, suggesting the presence of SUMOylated proteins at INoBs. Moreover, depletion of the SUMO-conjugating enzyme UBC9 or the sumo hydrolase SENP2 changed p21Cip1 intracellular distribution. In addition to SUMO-1 and p21Cip1, cell cycle regulators and DNA damage checkpoint proteins, including Cdk2, Cyclin E, PCNA, p53 and Mdm2, and PML were also detected in INoBs. Importantly, depletion of UBC9 or p21Cip1 impacted INoB biogenesis and the nucleolar accumulation of the cell cycle regulators and DNA damage checkpoint proteins following DNA damage. The impact of p21Cip1 and SUMO-1 on the accumulation of proteins in INoBs extends also to CRM1, a nuclear exportin that is also important for protein translocation from the cytoplasm to the nucleolus. Thus, SUMO and p21Cip1 regulate the transit of proteins through the nucleolus, and that disruption of nucleolar export by DNA damage induces SUMO and p21Cip1 to act as hub proteins to form a multiprotein complex in the nucleolus.


Introduction
The nucleolus is a non-membrane bound subnuclear organelle in which rRNAs are transcribed, processed, and assembled with ribosomal proteins into mature ribosomes [1,2] involving an E1 activating enzyme, a single E2 conjugating enzyme (UBC9), which directly interacts with the substrate, and the substrate-specific E3 ligases [43,44]. Deconjugation of SUMO from target proteins involves SUMO-specific proteases (SENPs), of which six members have been identified, SENP1, 2, 3, 5, 6, and 7 [45]. Of these, SENP1 and SENP2 have broad specificity for the three mammalian SUMOs, while the others favour SUMO-2/3 as the substrate. SENP1, 6, and 7 localize in the nucleoplasm, but not in the nucleolus, SENP2 is nuclear and associates with the nuclear pores and nuclear speckles, while SENP3 and SENP5 are located in the nucleolus and are important for ribosome maturation and nucleolar export [46,47]. Thus, it has been proposed that SUMO-dependent subnuclear trafficking may assist in coordinating the rate of ribosome biogenesis [36]. Notably, mouse cells deficient in the UBC9 protein showed striking morphological nucleolar disruption [37]. Furthermore, SUMOylation of pre-ribosomal particles in the nucleus and subsequent deSUMOylation at the nuclear pore complex are necessary for efficient ribosome biogenesis and export in yeast [48].
Based on our previous studies on p21 and given that SUMO has been linked to nuclear and nucleolar protein export and multiprotein complex formation, we aimed to determine whether SUMOylation/deSUMOylation is involved in the nuclear/cytoplasmic distribution of p21, and investigated the role of p21 and SUMOylation in INoB generation upon DNA damage. In fact, we demonstrate that INoB biogenesis, and the concomitant accumulation in the INoB of proteins related to cell cycle regulation, DNA maintenance and chromatin organization is dependent on p21 and UBC9.
To knock down proteins using siRNA, 2.4 x 10 5 HCT116 cells in P35 dishes were transfected with the siRNA of interest at 50 nM, using HiPerFect Transfection Reagent (Qiagen) according to the manufacturer's guidelines. After transfection, the cells were cultured for 48 hours.
DNA and shRNA expression in mammalian cells was achieved by transfecting cells with the appropriate expression vector and using Lipofectamine 2000 Reagent (Invitrogen), according to the manufacturer's instructions. After transfection, the cells were cultured for 48 hours. The following cDNA constructs were used: wild-type pEGFP-p21 and pMT2HA-p21 [27,28]; pEGFP-SUMO-1 (gift from Hay RT, University of Dundee, UK [50]); GFP-SUMO-1ΔGly-GLy, obtained by PCR using the appropriate primers to introduce a stop codon before the last two glycines in SUMO-1; Orange-SUMO-1, obtained by substituting Orange for GFP in the GFP-SUMO-1 plasmid using AgeI/BsrgI restriction enzymes; pECFP-CDK2; and pSU-PER-puro-EGFP-p21, used to knock down p21 expression [51].
INoB size was quantified using the Image J program. The set scale of the image was determined and then the length of the INoB measured. Statistical analysis (the t-test) was performed using the GraphPad Prism 6 software. Values significantly different using unpaired t-test were marked with asterisks are ÃÃ p < 0.01, ÃÃÃ p < 0.001, ÃÃÃÃ p < 0.0001).

Electron microscopy
For immunogold electron microscopy, HCT116 cells were fixed with 4% paraformaldehyde in 0.1 M cacodylate buffer for 30 minutes at room temperature. Cells were scraped from the dishes, transferred to an Eppendorf tube, and centrifuged for 1 minute in a microfuge to obtain cell pellets. The pellets were washed with 0.1 M cacodylate buffer, dehydrated in increasing concentrations of methanol at -20˚C, embedded in Lowicryl K4M at -20˚C, and polymerized with ultraviolet irradiation. Ultrathin sections were mounted on nickel grids and sequentially incubated with 0.1 M glycine in PBS for 15 minutes, 5% BSA in PBS for 1 hour, and the rabbit polyclonal anti-SUMO-1 antibody (FL-101, sc-9060, Santa Cruz Biotechnology) diluted 1:50 in PBS containing 1% BSA and 0.1 M glycine for 1 hour. After washing, the sections were incubated with goat anti-rabbit IgG coupled to 10-nm gold particles (diluted 1:50 in PBS containing 1% BSA; BioCell, UK). After immunogold labeling, the grids were stained with lead citrate and uranyl acetate and examined with a Philips EM208 electron microscope operated at 60 kV. For control, ultrathin sections were treated as described above, but without the primary antibodies.
Autoradiographic analysis of rRNA processing RNA processing was analyzed as described in [14]. Briefly, 2.4 x 10 5 cells/well in 6-well plates were transfected with the appropriate siRNA for 48 hours. Newly synthesized RNA was then labeled by incubating for 1 hour in a medium containing 1.2 μCi [5.6-3 H]-uridine (Perkin Elmer) per ml of medium. Pulse-labeled cells were then washed with media containing 1 mM non-radioactive uridine (Sigma) and incubated for 4 hours at 37˚C in 5% CO 2 . Following extraction, 1 μg of total RNA was size-separated by electrophoresis on a 1% agarose-formaldehyde gel.

Measurement of new protein synthesis by 3 H-leucine incorporation
To label newly synthesized proteins, 2.4 x 10 5 cells/well in 12-well plates were incubated for 30 minutes at 37˚C with 10 μCi 3 H-Leu/ml of medium. Cells were washed with cold PBS, lysed with RIPA buffer, and then centrifuged to obtain pellets. The supernatant was precipitated with 20% cold trichloroacetic acid (TCA) for 10 minutes at 4˚C. After centrifugation, the pellet was washed twice with 5% TCA and resuspended in 0.1 M NaOH. Of this, 70% was used to count the cpm in a liquid scintillation counter and the remainder used to determine protein concentration using the BCA protein assay. The counted cpm was divided by the protein concentration measured. Cells pre-treated with 100 μg/ml of cycloheximide were used as a positive control for protein synthesis inhibition. Assays were performed in triplicates.

SUMO-1 and p21 colocalize in INoBs after DNA damage
We previously demonstrated that Adriamycin (Adr)-induced DNA damage or other types of nucleolar stress disrupted the nucleolus and induced the formation of an INoB containing p21 [25]. This structure is dynamic and present in approximately 30% of the HCT116 cells treated with Adr for 24 hours. Interestingly, as shown in S1 Fig, this structure clearly correlates with a strong reduction of RNA transcription in the nucleoli even after recovery from Adr treatment. As protein SUMOylation is important for nuclear and nucleolar architecture and for multiprotein complex formation [35,37,40,52], we determined whether SUMO-1 colocalized with p21 in the nucleolus following Adr-induced DNA damage. Immunocytochemical analysis of endogenous SUMO-1 localization revealed mainly nucleoplasmic staining in non-treated cells; however, after Adr treatment, a strong nucleolar signal for SUMO-1 was observed colocalizing p21, indicating its localization in INoBs ( Fig 1A). Similar results were obtained using a different anti-SUMO-1 antibody (S2A Fig). Immunogold electron microscopic analysis confirmed the preferential distribution of SUMO-1 immunoreactivity in the INoBs formed after the induction of DNA damage, which we have previously observed to contain p21 [25] (Fig 1B). These findings were confirmed by ectopic expression of a GFP-SUMO-1, which colocalized with p21 in the INoBs after Adr treatment, and was also found in the nuclear envelope ( Fig  1C). Nucleolar localization was not observed with GFP-SUMO-1ΔGly-Gly, a SUMO-1 mutant that cannot be conjugated to proteins, indicating that wild type SUMO-1 is conjugated to proteins in p21-positive INoBs after DNA damage. Consistent with these observations we found that UBC9, the only known SUMO E2 enzyme, was also present in the INoB following DNA damage (Fig 1D and 1E). Most interestingly, depletion of UBC9 caused a decrease in SUMO-1 in the INoBs (Fig 1E and S2B Fig). All this suggests that protein SUMOylation actively occurs in these nucleolar structures.
UBC9 and SENP2 regulate the nuclear/cytoplasmic transport of p21 As we previously demonstrated that INoB formation upon DNA damage not only correlated with p21 accumulation in the nucleolus, but also with the inhibition of the nuclear export of p21 to the cytoplasm we determined whether protein SUMOylation/deSUMOylation cycle regulated the nuclear-cytoplasmic distribution of p21. To test this possibility, we depleted UBC9 or SENP2 (the only know SUMO-1 hydrolase with nuclear and nuclear pore localization [45]). As we previously demonstrated, when HCT116 cells were transfected with HA-p21, the percentage of cells with cytoplasmic p21 significantly decreased upon Adr treatment as compared to non-treated cells (Fig 2 and S3A Fig). Importantly, this decrease was abolished in cells depleted of UBC9, indicating that a lack of SUMOylation impairs the nuclear accumulation of p21 induced by DNA damage. That this effect is dependent on DNA damage is shown by the fact that the depletion of UBC9 has no effect on p21 localization in normally growing cells (Fig 2 and S3A Fig). In contrast, SENP2 depletion in the absence of DNA insults produced a decrease in the number of cells with cytoplasmic HA-p21, suggesting that a lack of deSUMOylation impairs the cytoplasmic localization of p21 (Fig 2 and S3A Fig). Of note, depletion of either UBC9 or SENP2 did not disrupt the nucleoli in the absence of DNA damage and also did not affect Adr-induced nucleolar disruption as indicated by the presence of peripheral (segregated) fibrillarin and capped UBF in the nucleolus (S3B Fig). Last, since p21 transits through the nucleolus [25], its localization was regulated by SUMO-1 conjugation and SUMO-1 was suggested to regulate the ribosome biogenesis, we tested if p21 has any impact directly in ribosome formation. When we measure the total protein synthesis or the rRNA processing in the absence of p21, we did not observe changes comparing to control cells, suggesting that p21 does not play a direct role ribosome biogenesis in normally growing cells (S4 Fig).

Both p21 and SUMOylation favour INoB biogenesis upon DNA damage
Next, we explored the involvement of p21 and SUMO in the nucleolar disorganization and INoB formation induced by DNA damage. As shown for UBC9 and SENP2 (S3A Fig), p21 was not necessary for nucleolar disruption after Adr treatment, which was demonstrated in both HCT116 cells lacking p21 through siRNA depletion and p21-knockout HCT116 cells (HCT116 p21KO cells) (Fig 3A and 3B and S5 Fig). Furthermore, p21 was not involved in initiating INoB formation since the percentage of cells with INoBs was approximately 30% in all Adr-treated HCT116 cells after 24 hours, independent of p21 depletion. However, p21-depleted cells showed significantly smaller Adr-generated INoBs compared to those with normal p21 levels as measured by phase contrast analysis (Fig 3C and S5 Fig). This was also observed in HCT116 cells transfected with short hairpin RNA against p21 (Fig 3C and S5E  Fig). Similar effects were seen with HCT116 cells treated with UBC9 siRNA. Interfering RNA against UBC9 did not change the percentage of INoB-positive cells after Adr (30%), but they had significantly smaller INoBs compared to siRNA control treated cells, demonstrating a role of SUMOylation in INoB size (Fig 4A). Accordingly, a concomitant decrease in nucleolar p21 accumulation was observed upon depletion of UBC9 (Fig 4B). Of note, total p21 levels upon Adr treatment were also high UBC9-depleted cells, indicating that UBC9 depletion effect in INoB size was not due to a lack of p21 accumulation (Fig 4C).  Since diverse DNA maintenance and chromatin organization proteins have shown to be localized in intranucleolar structures in a proportion of non-stressed cells [33], we asked whether their accumulation in INoBs following DNA damage was dependent on SUMO-1 and p21.
We first analyzed the presence of Cyclin E, Cdk2 and PCNA, all p21-binding proteins. Cyclin E has already been reported to be present in the nucleolus, with Cdk2/Cyclin E kinase activity proposed to be involved in maintaining nucleolar functions [53,54]. Both endogenous Cyclin E and Cdk2 localized in the same nucleolar compartment as p21 in response to Adr treatment ( Fig 5A). PCNA, a protein involved in DNA replication and repair, also colocalized with p21 in the nucleolus following DNA damage (Fig 5A). Interestingly, the smaller INoBs induced by DNA in cells with low levels of p21 or UBC9, also had decreased levels of Cyclin E and PCNA (Fig 5B and 5C and S7A Fig). Accordingly, when p21 and SUMO were exogenously expressed in HCT116 p21KO cells, PCNA, Cdk2, and Cyclin E levels in INoBs after DNA damage were restored (S7B Fig). Nucleolar localization of p53 and Mdm2 was also analyzed since these proteins are linked to the DNA damage checkpoint response, they have already been described to localize at the nucleolus [9,55,56] and can also be SUMOylated [57,58]. Interestingly, both p53 and Mdm2 localized in INoBs following DNA damage ( Fig 6A) and their nucleolar accumulation was reduced under conditions that produced smaller INoBs, e.g., in HCT116 p21KO cells or UBC9-depleted cells (S8A and S8B Fig). The presence of the promyelocytic leukemia (PML) protein in INoBs was also analyzed since it is a major target for covalent modification by SUMO, and it contains SUMO-interacting-motifs (SIM) in its sequence, which allows PML to act as a scaffold protein for the formation of PML-nuclear bodies (PML-NB) and for chromatin organization [59]. Furthermore, it has also been shown to translocate to the nucleolus following DNA damage [60]. We observed PML1 in INoBs, together with p21, in Adr-treated HCT116 cells (Fig 6B).
Finally, we determined whether CRM1 accumulated in INoBs upon DNA damage in a p21and SUMOylation-dependent manner. CRM1 is an exportin that is associated with the translocation of several proteins from the nucleus to the cytoplasm, including, p21, p53, and Mdm2 [61,62], as well as specific ribosomal proteins [17]. CRM1 also regulates the nucleolar localization of proteins involved in ribosome biogenesis [63]. Furthermore, Ernoult-Lange et al (2009) [64] reported that impaired ribosome biogenesis elicited CRM1 accumulation in the nucleoli in specific foci that they termed CRM1 nucleolar bodies. Our data show that CRM1 had a diffuse nuclear localization in control HCT116 cells, but accumulated in the nuclear envelope and INoBs following DNA damage, where it colocalized with p21 (Fig 7). Moreover, CRM1 accumulation in these nucleolar structures significantly dropped in p21-or UBC9-depleted cells (Fig 7 and S8C Fig).
Taken together theses data indicate that p21, SUMO, and protein SUMOylation are important for INoB assembly after DNA damage, as well as for the accumulation of p21-binding proteins (Cyclin E, PCNA, and Cdk2), DNA damage checkpoint proteins (p53 and Mdm2), SUMO-interacting proteins (PML1), and the exportin CRM1 in INoBs.

Discussion
We have previously shown that upon DNA damage and other types of nucleolar stress, p21-containing INoBs form in a great variety of cell lines [25]. Here, we report that endogenous SUMO-1 also accumulates in INoBs. Moreover, while exogenous wild type SUMO-1 accumulates in INoBs, mutant SUMO-1 that is unable to conjugate with proteins does not. Thus, our data show that SUMOylated proteins accumulate in INoBs together with p21 following DNA damage. The presence of UBC9, the only known E2 SUMO-conjugating enzyme, in these structures suggests that SUMOylation may be actively occurring there. To the best of our  knowledge, this is the first time that UBC9 has been reported to be present in the nucleolus. Since SENP3 and 5, two SUMO isopeptidases, have also been observed in the nucleolus [65,66], it is likely that protein SUMOylation and deSUMOylation occur in the nucleolus. Our results also indicate that both p21 and SUMOylation are essential for INoB biogenesis and the concomitant nucleolar accumulation of diverse p21-and SUMO-interacting proteins that are essential for cell cycle regulation, DNA damage checkpoints and chromatin organisation.
We previously showed that p21 transits through the nucleolus during its passage from the nucleus to the cytoplasm [25]. Here, we demonstrate that SUMOylation regulates the distribution of p21 between the nucleus and the cytoplasm. Cytoplasmic p21 has an oncogenic role and has been linked to some breast tumours [30] and lymphomas [67], while nuclear p21 inhibits cell proliferation and acts as a tumour suppressor [31]. Interestingly, inhibition of nuclear export is becoming an important and promising target in cancer therapy [68]. Meanwhile, SUMOylation has been shown to regulate the transit of proteins through the nuclear pores [69], with SENP2, a deSUMOylation enzyme, being observed at nuclear pores [36]. We propose that SENP2 inhibition could be used to decrease cytoplasmic p21 localization and, consequently, promote its tumour suppressive role.
We had postulated that the transit of p21 through the nucleolus in non-stress conditions was essential for regulating ribosomal biogenesis. However, our data on protein synthesis and rRNA synthesis and processing do not support this hypothesis.
In addition to covalently conjugating with proteins, SUMO-1 can also interact with a variety of proteins containing SIMs, thereby promoting the creation of multiprotein complexes [40]. The interaction of SUMO-1 with the SIM of PML is essential for the formation of PML bodies [59]. Moreover, p21 can interact with a variety of proteins either through its N-terminal globular domain or its C-terminal domain, which is flexible and can interact with a diverse range of proteins [70,71]. Since we observed p21-and SUMOylation-dependent accumulation of several p21-binding proteins (PCNA, Cdk2, and Cyclin E) and SUMO-conjugated proteins (p53, Mdm2, and PML1) in INoBs following DNA damage, we propose that p21 and SUMO act as "hub" proteins, creating a nucleation platform to increase the concentration of specific proteins in the nucleolus. This would facilitate interactions that are important for the stress response. In this sense, SUMOylated PCNA might have a function in rDNA repair by homologous recombination [72] and p53 binds to TATA-binding proteins and TBP-associated factor 1 of the SL-1 complex, preventing its interaction with UBF and thus formation of transcription-competent Pol I complex [73]. In agreement with the formation of these platforms, we also noticed PML colocalization with p21 in INoBs. This is in agreement with an earlier study showing that nucleolar PML following DNA damage was critical for Mdm2 nucleolar localization and p53 stabilization [60].
Alternatively, although not mutually exclusive with the abovementioned model, the accumulation of some of the proteins in the INoBs may be due to DNA damage disrupting their normal translocation through the nucleolus to the cytoplasm or their shuttling between the nucleolus and the nucleoplasm, thus trapping the proteins in the INoBs. Indeed, we observed the accumulation in the INoBs of CRM1, an exportin required also to transport proteins into the nucleolus [63]. Accordingly, nucleolar CRM1 accumulation has already been described in response to RNA polymerase I inhibition [64]. The authors of that study proposed that a platform loading CRM1 onto pre-60S ribonucleoprotein particles occur in nucleoli (CNoB) in normal conditions, which, upon nucleolar disruption, causes further nucleolar CRM1 accumulation. Here, we show a similar effect in response to Adr treatment. However, we propose that the export of other proteins using the same ribonucleoprotein particle export route is also blocked. Our results indicate that p21 [74] and SUMOylation are important for the transit of these proteins through the nucleolus, since the depletion of p21 or UBC9 prevented their nucleolar accumulation following DNA damage. One possibility is that p21 and SUMOylation facilitate the loading of CRM1 on to NES-containing proteins entering the nucleolus, therefore playing an essential role in nucleoplasm-nucleolus shuttling and nuclear export through the nucleolus. The transit of some of these proteins through the nucleolus under non-stress conditions could be linked to the fundamental role of the nucleolus in monitoring and responding to different cellular stress signals that affect ribosome biogenesis [21]. Two export routes from the nucleus to the cytoplasm have been proposed for p53, one through the nucleolus, where the protein is polyubiquitinated and labeled for degradation in the cytoplasm, and another through the nucleoplasm, where the protein is monoubiquitinated and sent to the cytoplasm to perform specific functions such as inhibiting apoptosis [10]. Blocking the export from the nucleolus by nucleolar disruption facilitates p53 accumulation in the nucleus, where it may play an important role in the stress response. Furthermore, proteasome-independent, Defand calpain 3b-dependent degradation of p53 in the nucleolus has been described in nonstress conditions [75], while Cyclin E has also been shown to be polyubiquitinated and tagged for degradation in the nucleolus [76].
We propose that p21 and SUMO are involved in the active accumulation of p21-and SUMO-interacting proteins in the nucleolus following DNA damage to facilitate their specific functions in the stress response. Alternatively, they could be associated with regulating the transit of these proteins through the nucleolus under control conditions, which is impeded following DNA damage.
Dysregulated ribosome biogenesis is required for the survival of malignant cells, thus making nucleolar activity an important therapeutic target in the treatment against cancer [77,78]. The data presented here will help to reveal the mechanisms involved in regulating the transit and accumulation of proteins in the nucleolus and might contribute to the design of new cancer therapies.