Pli1PIAS1 SUMO Ligase Protected by the Nuclear Pore-associated SUMO Protease Ulp1SENP1/2*

Background: SUMO conjugation to the proteome critically controls cell growth, but mechanisms for regulating SUMO ligases are poorly defined. Results: Desumoylation of the major fission yeast SUMO ligase by a nuclear pore-associated protease protects it from proteolysis. Conclusion: Desumoylation of a SUMO ligase antagonizes autoinhibition of the SUMO pathway. Significance: These data demonstrate cooperation between STUbL and Cdc48(p97) in proteasome-mediated degradation of SUMO conjugates. Covalent modification of the proteome by SUMO is critical for genetic stability and cell growth. Equally crucial to these processes is the removal of SUMO from its targets by the Ulp1 (HuSENP1/2) family of SUMO proteases. Ulp1 activity is normally spatially restricted, because it is localized to the nuclear periphery via interactions with the nuclear pore. Delocalization of Ulp1 causes DNA damage and cell cycle defects, phenotypes thought to be caused by inappropriate desumoylation of nucleoplasmic targets that are normally spatially protected from Ulp1. Here, we define a novel consequence of Ulp1 deregulation, with a major impact on SUMO pathway function. In fission yeast lacking Nup132 (Sc/HuNUP133), Ulp1 is delocalized and can no longer antagonize sumoylation of the PIAS family SUMO E3 ligase, Pli1. Consequently, SUMO chain-modified Pli1 is targeted for proteasomal degradation by the concerted action of a SUMO-targeted ubiquitin ligase (STUbL) and Cdc48-Ufd1-Npl4. Pli1 degradation causes the profound SUMO pathway defects and associated centromere dysfunction in cells lacking Nup132. Thus, perhaps counterintuitively, Ulp1-mediated desumoylation can promote SUMO modification by stabilizing a SUMO E3 ligase.

Covalent modification of the proteome by the small ubiquitin-like modifier SUMO 2 regulates most aspects of cell growth, including cell cycle transitions and genome stability (1)(2)(3)(4). The addition of SUMO and its removal from target proteins must be tightly regulated to avoid the pathological consequences of defects in pathway homeostasis (5)(6)(7).
SUMO is attached to lysine residues in target proteins via an enzymatic cascade of E1 activating, E2 conjugating, and E3 ligase factors (4). E3 ligases provide target specificity and enhance SUMO conjugation. A major family of SUMO E3 ligases is the PIAS (protein inhibitor of activated STAT) family, which has several members in mammalian cells e.g. PIAS1-4, two in budding yeast SIZ1/2, and one called Pli1 in fission yeast (8,9). Pli1 catalyzes the majority of sumoylation (Ͼ90%) including SUMO chain formation, and its deletion causes meiotic defects, centromeric heterochromatin dysfunction, and telomere elongation (8,10).
Sumoylated proteins can also be ubiquitinated by a SUMOtargeted ubiquitin ligase (STUbL) to promote their degradation at the proteasome (24 -26). Correlative evidence suggests that SUMO chains act as targeting signals for STUbLs. Consistent with this, high molecular weight SUMO chains accumulate in STUbL mutant cells (24 -26), a phenotype also caused by Ulp2 inactivation (10,27). Moreover, in fission yeast, the growth and genome stability defects caused by both STUbL and Ulp2 inactivation are suppressed by blocking SUMO chain formation (10,28). In contrast, preventing SUMO chain formation in budding yeast is lethal to STUbL mutants but suppresses some ulp2⌬ defects (27,29,30). These findings highlight the importance of SUMO pathway homeostasis and that the "wiring" of the STUbL-SUMO interface differs notably between these yeasts (see above and Refs. 10 and 27-30).
The nuclear pore has emerged as a broadly conserved hub of SUMO-mediated signaling, impacting key processes such as transcription, chromosome segregation, and DNA repair (31,32). However, despite functional overlap, mutants of orthologous nuclear pore components cause different phenotypes in fission and budding yeast, the latter of which has been used in most studies (33,34). Therefore, to robustly model how the nuclear pore could impact SUMO pathway homeostasis in higher eukaryotes, it is important to determine how the nuclear pore impacts sumoylation in fission yeast.
Herein we reveal that sumoylation is globally reduced in fission yeast deleted for Nup132, which contrasts with the relatively mild and selective sumoylation defects of analogous budding yeast mutants (16,35,36). Nevertheless, we show that as in budding yeast (36), Ulp1 is both delocalized and less abundant in nup132⌬ (Sc nup133⌬) cells. What then causes the profound SUMO conjugation defect in nup132⌬ cells? Intriguingly, we found that Ulp1 dysfunction in nup132⌬ cells allows an accumulation of SUMO chains on the major SUMO E3 ligase Pli1, which promote its degradation in a STUbL-, Cdc48-Ufd1-Npl4-, and proteasome-dependent manner. Moreover, this novel phenomenon allowed us to execute a detailed mechanistic dissection of SUMO pathway and cofactor requirements in the turnover of a specific STUbL substrate. Overall, our data define how the activity of a PIAS family SUMO ligase, and thus its physiological impact, e.g. centromere function, hang in the balance between its autosumoylation and desumoylation by a nuclear pore localized SUMO protease.

Experimental Procedures
Yeast Strains and Growth Conditions-Standard yeast methods were performed as described previously (37). The strains used in this study are listed in Table 1.
Spot Assays-The cells were grown at 25°C to logarithmic phase (A 600 of 0.6 -0.8), spotted on agar plates in 5-fold dilutions from a starting A 600 of 0.5 and then incubated at 25-35°C for 3-5 days.
Fluorescence Microscopy-GFP fusion proteins and DAPI staining were imaged in live cells using an Eclipse E800 microscope (Nikon Metrology) with a 100ϫ Plan Differential Interference Contrast H oil immersion objective. The images were captured through an INFINITY 3 charge-coupled device camera using the INFINITY ANALYZE software (Lumenera Corporation).
Western Blotting-Exponentially growing cells (ϳ15 A 600 units) were washed with STOP buffer (10 mM EDTA, 50 mM NaF, 150 mM NaCl, 1 mM NaN 3 ). The pellets were flash-frozen in liquid nitrogen, and the cells were lysed by beating twice at 5.0 m/s for 20 s in a FastPrep-24 instrument (MP Biochemicals) in 200 l of 20% TCA supplied with 100 ml silica-zirconia beads (BioSpec Products). After bead beating, 400 l of 5% TCA was added, and the total cell lysate was centrifuged at 16,000 ϫ g for 5 min at 4°C. The pellet was washed twice with 0.1% TCA. The precipitated proteins were resuspended in 8 M urea, 50 mM Tris, pH 8.5, 150 mM NaCl. Protein was quantitated by measuring absorbance at A 280 , and 60 g of proteins were resolved on a gradient SDS-PAGE gel and transferred to a nitrocellulose membrane. The membrane was blocked in 1% (w/v) nonfat milk in phosphate-buffered saline solution with 0.1% (v/v) Tween 20, probed with peroxidase anti-peroxidase and then detected using an ECL Dura system (Pierce) or probed with antibodies against ␣-tubulin, Pmt3 (SUMO) and then an IRDye-conjugated secondary antibody and imaged on an ODYSSEY scanner (Li-Cor).
Nickel-Nitrilotriacetic Acid Pulldown-Proteins from cell pellets were extracted by TCA precipitation as described above, and 20 mg of proteins were incubated with 50 l of nickelnitrilotriacetic acid beads (Qiagen) in the presence of 15 mM imidazole at room temperature for 1 h. The beads were then washed three times with 8 M urea, 50 mM Tris, pH 8, 150 mM NaCl, and 20 mM imidazole and eluted with 2ϫ LDS sample loading buffer (Invitrogen) at 70°C for 10 min. Input, flow through, and one-third of the nickel pulldown were resolved by SDS-PAGE and detected by peroxidase anti-peroxidase Western blotting.
RNA Extraction and RT-Quantitative PCR-Total RNA was extracted from ϳ5 A 600 units of exponentially growing cells. The cells were lysed by bead beating in 0.8 ml of TRIzol (Invitrogen), using the equipment and settings described above. RNA extraction with TRIzol was performed by following the manufacturer's instructions. The extracted RNA was treated with DNase I (New England Biolabs) and cleaned up with the RNeasy Plus minikit (Qiagen). A total of 1 g of DNase-treated RNA was reverse transcribed using the ProtoScript II first strand cDNA synthesis kit (New England Biolabs). The completed reactions were diluted 10-fold, and 5 l of the dilutions was used as the template in 20-l quantitative PCR mixtures using the SensiFAST SYBR No-ROX kit (Bioline), and the quantitative PCRs were carried out using the Chromo4 system (Bio-Rad).
Plasmid Construction-Unless otherwise stated, cDNAs were cloned by PCR amplification using specific primers and PrimeStar DNA polymerase (Clontech) followed by ligation into the pREP series of yeast shuttle vectors (38). Details regarding plasmid construction are available on request.

Nuclear Pore Mutant Suppresses Phenotypes of STUbL
Dysfunction-To assess roles of fission yeast Nup132 in SUMO pathway homeostasis and genome stability, we analyzed genetic interactions between its deletion (nup132⌬) and alleles of STUbL (slx8-29) and Ulp2 (ulp2⌬). As in previous reports, nup132⌬ cells grew similarly to wild type and exhibited little to no sensitivity to genotoxins or elevated temperatures (Fig. 1A) (33,34). These subtle phenotypes are consistent with the fact that protein import/export and RNA export pathways are intact in nup132⌬ cells and that Nup132 has an expressed paralog called Nup131 (33,34).
At permissive temperature, hypomorphic slx8-29 cells have increased levels of high molecular weight (HMW) SUMO conjugates as compared with wild type (Fig. 1B) (28). In contrast, both nup132⌬ and slx8-29 nup132⌬ cells had less HMW SUMO conjugates than wild type (Fig. 1B). Moreover, at restrictive temperature, the strong accumulation of HMW SUMO conjugates in slx8-29 cells was absent in slx8-29 nup132⌬ cells (Fig. 1B). There was also a lower level of HMW SUMO conjugates in ulp2⌬ nup132⌬ double mutant versus ulp2⌬ single mutant cells (Fig. 1B). Therefore, for as yet undefined reasons, nup132⌬ cells exhibit reduced global sumoylation that correlates with the rescue of STUbL and Ulp2 SUMO protease dysfunction.
Altered Ulp1 Activity in nup132⌬ Cells-Ulp1 endogenously tagged with GFP (Ulp1-GFP) gives a weak fluorescence signal but shows that fission yeast Ulp1 exhibits a perinuclear localization in live as well as fixed cells ( Fig. 2A) (14). Based on analyses of Ulp1 orthologs in other species, this likely reflects its nuclear pore association (32). Budding yeast Ulp1 is anchored at the nuclear pore via a complex set of interactions between the Ulp1 N terminus and several nuclear pore-associated factors, and both its anchoring and protein stability are disrupted in cells lacking Nup133 (fission yeast Nup132; reviewed in Ref. 32). Similarly, we found that both the perinuclear signal of Ulp1 and the level of full-length protein were reduced in nup132⌬ cells (Fig. 2, A and B). Thus, the anchoring mechanism of Ulp1 at the nuclear pore is broadly conserved, but the consequences of Ulp1 delocalization and destabilization for SUMO pathway homeostasis are different.
Because Ulp1 functions in SUMO maturation, we next tested whether reduced sumoylation in nup132⌬ cells was due to inefficient SUMO maturation. We ectopically expressed either fulllength (SUMO FL ) or the mature form (SUMO GG ) of SUMO, which both ran at the size of mature SUMO (Fig. 2C, black arrowhead) in wild type cells, indicating that normal Ulp1 activity can fully process overexpressed full-length SUMO. Ulp1 maturation activity is slightly compromised in nup132⌬ cells, because they show a small amount of residual immature full-length SUMO (Fig. 2C, white arrowhead). However, because SUMO conjugation in nup132⌬ cells is low and unchanged whether they express full-length or mature SUMO (Fig. 2C), we conclude that reduced sumoylation in nup132⌬ cells is not caused by a SUMO maturation defect.
Interestingly, nup132⌬ slx8-29 cells exhibited a major increase in SUMO conjugates, to a level exceeding that of slx8-29 single mutant cells (Fig. 2C). Therefore, the inactivation of STUbL, when combined with higher levels of SUMO, restores global sumoylation in nup132⌬ cells. Enhanced global sumoylation in nup132⌬ slx8-29 as compared with slx8-29 cells is consistent with reduced Ulp1-mediated desumoylation in nup132⌬ mutants.
The stability of endogenous Pli1 was monitored over time in cells treated with cycloheximide, an inhibitor of protein synthesis. Intriguingly, we found that Pli1 was both less abundant and more rapidly degraded in nup132⌬ versus wild type cells (Fig.  3A). Given that STUbL inactivation can restore global sumoylation in nup132⌬ cells (Fig. 2C), we tested whether the slx8-29 mutation affected Pli1 stability. Strikingly, whereas Pli1 was only weakly detectable in nup132⌬ cells, Pli1 abundance and a series of lower mobility species were dramatically increased in slx8-29 nup132⌬ cells (Fig. 3B). Neither Pli1 levels nor its gel migration were affected in slx8-29 single mutant cells (Fig. 3B).
Direct Evidence that SUMO Chains Are Required for STUbL Targeting-The role of STUbL in targeting SUMO-conjugated proteins is now well established, and notably, slower migrating forms of Pli1 accumulate in slx8-29 nup132⌬ cells (Fig. 3B). To confirm that these species are SUMO-conjugated Pli1, we purified SUMO conjugates under denaturing conditions from slx8-29 nup132⌬ cells and probed for Pli1 (Fig. 4A). This approach enriched only the slower migrating Pli1 species and revealed higher order polysumoylation of Pli1 (Fig. 4A, white arrowheads).
STUbLs are thought to target proteins that are modified by SUMO chains, and strong correlative data have been published (see Refs. 25,26,[46][47][48]. To test the role of SUMO chains more directly, we used two SUMO mutants that we previously showed block SUMO chain formation in different ways (10).

Ulp1 Desumoylates Pli1
In light of the above data, it is interesting to note that the absence of SUMO chains in SUMO K14,30R , SUMO D81R , or pli1⌬ cells suppresses all tested STUbL mutant phenotypes (10,28). Explaining this epistatic relationship, as demonstrated for Pli1, STUbL is only engaged when SUMO chains are present. Overall, these data provide direct mechanistic insight into the role of SUMO chains in STUbL targeting.
Ulp1 Protects Pli1 from Sumoylation and STUbL-mediated Degradation-Given that the Ulp1 SUMO protease is deregulated in nup132⌬ cells, we tested whether increased Ulp1 dos-age would impact the SUMO conjugation state and stability of Pli1. Strikingly, overexpression of Ulp1 in nup132⌬ cells restored normal Pli1 levels (Fig. 4C). Moreover, Pli1 was restored in a hyposumoylated form, consistent with Ulp1 desumoylating and protecting it against STUbL-dependent degradation (Fig. 4C). Because budding yeast Ulp1 modulates transcription of the GAL1 gene (50), we monitored transcript levels of Pli1 upon Ulp1 overexpression. Whereas Ulp1 overexpression yields the anticipated increase in Ulp1 transcript levels, no difference in Pli1 transcription was detected (Fig. 4D). These results suggest that Ulp1 is normally optimally positioned to desumoylate and stabilize Pli1 and that increased Ulp1 dosage can compensate for this loss of spatial regulation.
Pli1 Overexpression Reverses SUMO-related Phenotypes of nup132⌬ Cells-Pli1 acts at the end of the SUMO conjugation pathway to attach activated SUMO to its substrates. Therefore, if mature SUMO or E1/E2 factors were limiting, this would render the SUMO pathway unresponsive to the presence of Pli1. Therefore, to test whether Pli1 is the major or sole SUMO pathway factor deregulated by deleting Nup132, we overexpressed Pli1 in pli1⌬, nup132⌬, and nup132⌬ slx8-29 cells. As anticipated, expression of Pli1 in pli1⌬ control cells restored global SUMO conjugates to a level higher than that seen in wild type (Fig. 5A). Increased Pli1 dosage also restored SUMO con- FIGURE 3. Pli1 SUMO ligase is destabilized in nup132⌬ cells. A, wild type or nup132⌬ cells grown at 25°C to mid log phase and then shifted to 35°C for 4 h. Cycloheximide was added to cell cultures at 200 g/ml, and cells were harvested at 0, 15, 30, 60, or 120 min after cycloheximide addition and analyzed by Western blotting. The levels of Pli1-TAP proteins (unmodified and all SUMO modified species) were normalized to the levels of tubulin in each sample and quantified with National Institutes of Health ImageJ software. The graph is representative of results from three experiments. B, Western blots of Pli1-TAP or tubulin of indicated strains grown at 25°C to log phase (top panels) or to mid log phase then shifted to 35°C for 6 h (bottom panels). Black and white arrowheads mark the positions of unmodified and SUMO modified Pli1-TAP bands, respectively.

Ulp1 Desumoylates Pli1
jugates in nup132⌬ cells but to levels that were lower than those in pli1⌬ cells (Fig. 5A). This is expected based on the constitutive degradation of Pli1 in nup132⌬ but not pli1⌬ cells. In slx8-29 nup132⌬ double mutant cells, SUMO conjugates were higher than those in the nup132⌬ single mutant upon Pli1 overexpression (Fig. 5A). Again, this is consistent with the role of STUbL in degrading Pli1 and antagonizing HMW SUMO conjugates. Overall, these data indicate that the SUMO pathway is largely intact upstream of Pli1 in nup132⌬ cells.
Thus far, our data indicate that the limited Pli1 activity present in nup132⌬ cells rescues the lethality of slx8-29 cells. To directly test this, we assayed the effect of restoring Pli1 expression on the growth of slx8-29 nup132⌬ cells. The growth of wild type or slx8-29 cells was similar whether they carried an empty vector control or Pli1 plasmid (Fig. 5B). As expected for the vector control (i.e. nup132⌬ suppresses slx8-29), slx8-29 nup132⌬ double mutant cells grew well compared with the slx8-29 single mutant (Fig. 5B). Notably, however, expression of Pli1 in slx8-29 nup132⌬ cells caused poor growth and drug sensitivity that was similar to that of slx8-29 cells (Fig. 5B). Therefore, consistent with our biochemical analyses, restoring Pli1 expression is sufficient to nullify the rescue of slx8-29 phenotypes by nup132⌬.
We also tested whether, as was the case for cells lacking STUbL activity (28), deletion of Pli1 could suppress the growth and genome instability phenotypes of ulp2⌬ cells. Indeed, the pli1⌬ ulp2⌬ double mutant phenocopies pli1⌬ cells, indicating that pli1⌬ is epistatic to ulp2⌬ (Fig. 5C). We previously demonstrated that Pli1-dependent SUMO chain formation is toxic to both slx8-29 and ulp2⌬ cells (10). Therefore, in keeping with our biochemical analysis, Pli1 degradation and reduction of HMW SUMO conjugates can explain the suppression of both slx8-29 and ulp2⌬ phenotypes.
Compromised Centromere Function in Cells Lacking Nup132-Centromere function is disrupted in pli1⌬ cells, whereby a ura4ϩ marker inserted at the heterochromatic inner repeats (imr) is constitutively expressed, versus the variegating phenotype in wild type cells (40). We therefore assessed chromatin function at imr in nup132⌬ cells. As expected, wild type cells carrying the ura4ϩ insertion at imr (imr:ura4ϩ) exhibited variegated expression, being able to grow both in the absence of uracil (leucine adenine histidine) and in the presence of 5-fluoroorotic acid (51). However, both pli1⌬ imr:ura4ϩ and nup132⌬ imr:ura4ϩ cells grew poorly in the presence of FOA but robustly in the absence of uracil (Fig. 5D). This result is consistent with the constitutive degradation of Pli1 in nup132⌬ cells, causing hyposumoylation of key epigenetic regulators.

Discussion
Herein, we reveal a key novel consequence of nuclear pore dysfunction and subsequent spatial deregulation of Ulp1, which has a major impact on global sumoylation and centromere function. In budding yeast that lack Nup133 (SpNup132), Nup60, or Nup120, Ulp1 is destabilized and mislocalized in the nucleoplasm, causing altered global SUMO conjugate patterns and associated defects in genetic stability (16,31,35,36,52). A gain of Ulp1 function that promotes the desumoylation of normally spatially protected nucleoplasmic SUMO conjugates is proposed to cause these phenotypes. Such promiscuous desumoylation of nucleoplasmic SUMO conjugates by delocalized Ulp1 in nup132⌬ fission yeast is also likely.
Interestingly, however, our results demonstrate that the residual Ulp1 activity in nup132⌬ cells is unable to counteract Pli1 sumoylation. This leads to Pli1 degradation and associated hyposumoylation phenotypes. It will be interesting to determine whether this novel mechanism contributes to the SUMO conjugation defects reported for certain budding yeast nucleoporin mutants (16,31,35,36,52). In this regard, STUbL degrades the nuclear pool of a budding yeast PIAS family ligase, Siz1, particularly when its nuclear export is inhibited (53). Therefore, although untested, it is possible that nuclear pore-associated Ulp1 desumoylates Siz1 and protects it from degradation. If so, given the evolutionary divergence of fission and budding yeast, conservation of this mechanism in human cells seems likely.
How Ulp1, which is normally tethered at the nuclear periphery, accesses nucleoplasmic Pli1 is an interesting question for the future. We determined that increasing Ulp1 dosage in nup132⌬ cells stabilized Pli1. This is consistent with increased Ulp1 activity compensating for the loss of a normally more coordinated interaction with Pli1. For example, fission yeast centromeres cluster at the nuclear periphery and are subject to functionally critical Pli1-dependent sumoylation (40,54). Therefore, Pli1 acting at centromeres is also spatially organized at the nuclear periphery, more proximal to Ulp1.
In addition, budding yeast Ulp1 was recently shown to desumoylate transcription factors on chromatin to facilitate transcriptional activation, providing an example of recruitment of Ulp1 targets to the nuclear pore (50). Interestingly, a subcomplex of the human nuclear pore including Nup133 transiently localizes to centromeres and contributes to their mitotic function (55). Therefore, it will be interesting to determine whether Nup133-associated SENP2 (56) also modulates centromere/ kinetochore sumoylation to support its function. Relocalization of the human pore complex and SUMO protease, rather than the target, may be a result of the open mitosis in human cells versus the closed mitosis of yeast.
Based on high throughput analysis, nup132⌬ cells share another phenotype of pli1⌬ cells, that is, highly elongated telomeres (40,43). Notably, like centromeres, telomeres are also clustered at the nuclear periphery and are subject to Pli1dependent regulation (40,41). Pli1 sumoylates the telomere protein Tpz1 to support a mechanism for inhibiting excessive telomerase activity at chromosome ends (57). Therefore, based on the synonymous centromere silencing roles of Pli1 and Nup132, we envisage that Tpz1 will be hyposumoylated in cells lacking Nup132, leading to telomere elongation.
Identification of Pli1 as a STUbL substrate has enabled a comprehensive analysis of SUMO pathway and ancillary factor requirements. The role of SUMO chains in STUbL targeting has been largely inferred from (i) the accumulation of SUMO chain species when STUbL activity is compromised, (ii) the multivalent SUMO-interacting motif arrangement in STUbLs, and (iii) SUMO chain-induced autoubiquitination/degradation of RNF4 (25,47,58). We now demonstrate directly that the STUbL pathway is dependent on the chain forming lysine residues of SUMO, as well as the noncovalent SUMO-Ubc9 interface that promotes SUMO chain formation. These data fit well with the epistatic relationship of SUMO chain and STUbL mutants in fission yeast (10).
In addition, Cdc48-Ufd1-Npl4 was recently identified as a cofactor for STUbL in managing global SUMO conjugate levels (44,45). However, evidence for a cooperative role in degrading a specific substrate was lacking. Herein, we show that both STUbL and Ufd1 are indeed involved in the proteasome-dependent turnover of Pli1 in nup132⌬ cells.
As in fission yeast (44,45), budding yeast Cdc48 also acts as a SUMO-targeted segregase that removes sumoylated DNA repair factors from chromatin (59). Given this degree of functional conservation, it seems likely that the human STUbL RNF4 and p97 cooperate in the remodeling of chromatin e.g. at DNA repair sites, as reported independently for each factor (60 -63).
Author Contributions-M. N. and M. N. B. were both involved in study design, experimental execution, and writing of the manuscript.