Sumo-dependent substrate targeting of the SUMO protease Ulp1

Cell division is a fundamental feature of all life and involves the controlled duplication and faithful segregation of an organism's genetic material from one cell to the next. In eukaryotes, each cell division cycle is therefore executed as a tightly regulated, stepwise program that relies on intact chromosomes. In humans, the consequences of faulty chromosome segregation and the inability to repair DNA damage have been implicated in cancer, aging and congenital birth defects.

Ubiquitin and small ubiquitin-like modifier (SUMO), two small proteins that can become attached to other cellular proteins in a reversible manner [1], control important aspects of the cell division program. Ubiquitin is best known for its role in the targeted, proteasome-mediated destruction of proteins, including key cell-cycle regulators, but also holds nonproteolytic functions [2]. Sumoylation, on the other hand, does not directly target proteins for degradation. Rather, modification of proteins with SUMO has been shown to modulate various cellular processes, including cell-cycle regulation, transcriptional activation, nucleocytoplasmic transport, DNA replication and repair, chromosome dynamics, apoptosis, ribosome biogenesis, and the formation of nuclear bodies [3]. Additionally, an unexpected role of SUMO in protein ubiquitination has been uncovered. Briefly, degradation of several nuclear proteins, including some that are involved in DNA repair and transcriptional regulation, are preceded by modification with SUMO. These sumoylated proteins are recognized by SUMO-targeted ubiquitin ligases (STUbLs), which mediate their ubiquitination [4].

SUMO proteins are highly conserved from yeast to humans. Yeast cells express one SUMO protein (Smt3), and vertebrates express three isoforms (SUMO1, SUMO2 and SUMO3) [5]. SUMO2, SUMO3 and yeast Smt3 can form SUMO chains. SUMO1, on the other hand, lacks the internal lysine required for polymerization and may function as a chain terminator for SUMO2 and SUMO3 chains [6]. All SUMO variants are conjugated to lysine residues of specific proteins, but only a fraction of these target proteins are modified with SUMO at any given time [7, 8]. In metazoans, the dysregulation of sumoylation adversely affects developmental processes and has been implicated in the progression of neurodegeneration, cancer and infectious diseases [9, 10]. More than 1, 000 sumoylated proteins have been identified in yeast and humans, but only in a few cases has the role of sumoylation been studied in detail [11].

In the budding yeast Saccharomyces cerevisiae, the ligation of SUMO to specific substrate proteins requires an E1 heterodimer (Aos1 and Uba2) that activates SUMO, as well as E2 (Ubc9) and E3 (Siz1, Siz2, and Mms21) enzymes that aid in the conjugation and ligation of SUMO to proper target proteins [1]. Two yeast SUMO proteases, Ulp1 and Ulp2, contain a conserved cysteine protease domain that can remove the SUMO moiety from modified proteins. Recent evidence suggests that Ulp2, and its mammalian orthologs Susp1/SENP6 and SENP7, play a role in the removal of SUMO and SUMO chains from nuclear proteins [12–16]. Ulp1, on the other hand, has two contrasting cellular functions. Ulp1 facilitates sumoylation by processing precursor SUMO into its conjugation competent form. Conversely, Ulp1 also facilitates desumoylation by removing SUMO from nuclear and cytosolic proteins after conjugation [17]. Therefore, impairment of Ulp1 results in the accumulation of SUMO conjugates and the inability to carry out de novo sumoylation. The resulting lack of mature SUMO has been shown to adversely affect cellular DNA repair processes, the processing and export of the 60S preribosomal particle, nucleus-cytoplasm trafficking and cell viability [18–21].

The substrate specificity of SUMO proteases is at least in part regulated through their localization [22]. For example, certain yeast (Ulp2) and vertebrate (SENP6 and SENP7) SUMO proteases localize within the nucleus. In contrast, both yeast (Ulp1) and vertebrate (SENP1 and SENP2) SUMO proteases reside at the nuclear envelope (NE) through their interactions with the nuclear pore complex (NPC) [23–26]. Distinct domains have been identified that are required for Ulp1 NPC localization (amino acid residues 1 to 403) and SUMO processing (amino acid residues 404 to 620) [25–28]. The Ulp1 localization domain promotes interaction with karyopherins, which are soluble proteins that mediate transport across the NE and help localize Ulp1 to the nucleoplasmic side of the NPC. The Ulp1 localization domain can be subdivided into region 1 (Kap121-binding domain) and region 2 (Kap60- and Kap95-binding domain). Juxtaposed to the NPC localization domain of Ulp1 is a coiled-coil (cc) domain with a putative nuclear export signal and region 3, the catalytically active, conserved ubiquitin-like protease domain (UD) of Ulp1 [25–27]. Only regions 1 and 2 are involved in Ulp1 localization to the NPC, and the karyopherins seem to play a redundant role. NPC association of Ulp1 requires several proteins, including the nucleoporins Nup60 and Nup84, the silencing protein Esc1 and the myosin-like proteins Mlp1/2 [18, 19, 28]. Together these proteins may provide a scaffold for the functional regulation and substrate access of Ulp1 at the NPC.

The identification of NPC localization domains in Ulp1 has done little to aid our understanding of how SUMO proteases are targeted to their respective substrates [26]. One possibility is that SUMO proteases may contain structural features which allow for noncovalent interactions with SUMO and SUMO-modified proteins as they enter the nucleus. Indeed, conserved SUMO-interacting motifs (SIMs) have been predicted to be localized in the yeast SUMO protease Ulp2, as well as in mammalian SENP1, SENP2, SENP6 and SENP7 [14, 22, 29, 30]. Even though SIMs have not been identified in Ulp1, the crystal structure of the catalytic domain (region 3) bound to Smt3 reveals that both proteins interact through multiple residues that are distributed across a SUMO-binding surface (SBS) on the SUMO protease [31]. Only the carboxy terminus of bound Smt3 is inserted into a hydrophobic tunnel that leads toward Ulp1's active site. SUMO processing and deconjugation require an active site cysteine residue that resides at the end of this tunnel (see Additional files 1 and 2). It has been suggested that this configuration may allow for the accommodation of many different sumoylated proteins as well as SUMO precursors [31].

Ulp1 and several other SUMO proteases play important roles in mitosis [17, 32]. In budding yeast, loss of Ulp1-mediated desumoylation leads to cell-cycle progression defects and cell death [17]. This observation suggests that Ulp1 plays a key role in the sumoylation dynamics of important cell-cycle regulatory proteins. Though these cell-cycle-specific targets have eluded identification, several nuclear and cytosolic proteins involved in DNA replication and mitosis have been identified as Ulp1 desumoylation substrates [33–35]. How the NPC-localized Ulp1 is targeted to these mitotic substrates, especially those that are localized in the cytosol, is not entirely clear. In budding yeast, the NE does not break down during mitosis, and access to cytosolic desumoylation targets is therefore not automatic. It has been reported that during mitosis, Kap121 blocks Ulp1's access to its NPC-binding site and thus promotes an interaction of Ulp1 with septins [27]. A deletion mutant of Ulp1 lacking region 2 (Δ2), the Kap60- and Kap95-binding domain, has previously been shown to localize to septins in a Kap121-dependent manner [27]. Curiously, it has recently been demonstrated that region 2 also plays a role in nucleolar accumulation of Ulp1 after ethanol-induced stress [36].

One set of cytosolic substrates of the Ulp1 SUMO protease are the septins [27, 35]. The septins comprise an evolutionarily conserved class of GTPases that are implicated in bud-site selection, bud emergence and growth, microtubule capture and spindle positioning [37]. Members of the septin family in yeast include Cdc3, Cdc10, Cdc11, Cdc12 and Shs1/Sep7. These proteins are unique because they can form filaments that assemble into a ring structure and mark the site of new bud formation during cell division. At the end of mitosis, this ring separates and resembles a double-collar residing at the junction between the mother and daughter cells.

The septins Cdc3, Cdc11 and Shs1 are subject to sumoylation. Sumoylation of the septins occurs very briefly from the onset of anaphase to cytokinesis, with SUMO being attached only to the mother side of the double-septin ring collar [27, 38, 39]. Cell-cycle (G₂/M) arrest with nocodazole, a microtubule depolymerizing drug, greatly increases SUMO conjugation to septins [38]. Septin sumoylation in budding yeast is mediated by the SUMO E3 ligase Siz1 [40, 41]. During most of the cell cycle, Siz1 resides in the nucleus. However, at the M phase, Siz1 exits the nucleus to sumoylate septin proteins and possibly other cytosolic substrates [42]. Deletion of SIZ1 from cells abolishes septin sumoylation while causing only mild growth and cell-cycle progression defects. At the end of mitosis, the septins are desumoylated by Ulp1, even though Ulp1 remains visibly enriched at the NPC [27, 35, 38].

In the current study, we focused on the SUMO protease Ulp1. As detailed above, Ulp1 resides at the inner face of the NPC. This enrichment at the NPC depends on direct interactions with karyopherins and two domains in the amino terminus of Ulp1. In theory, this localization is well-suited to give Ulp1 access to some nuclear substrates and those in transit across the NE. However, both nuclear and cytosolic desumoylation targets of Ulp1 have been identified, posing an interesting question. How is Ulp1 directed to its cytosolic desumoylation targets? To answer these questions, we sought to identify features of Ulp1 required for substrate targeting in vivo and in vitro. Herein we describe our finding that the carboxy terminus of Ulp1 affects the targeting and retention of this SUMO protease to sumoylated target proteins, including septins, at the bud-neck of dividing cells. Specifically, we show that the interaction with SUMO comprises an important aspect of the subcellular targeting of Ulp1 to these substrates. Our findings are confirmed by biochemical analyses that focus on the SUMO-binding properties of Ulp1(3)^C580S, a novel truncation mutant that interacts avidly with SUMO and sumoylated proteins in vivo and in vitro. Significantly, the results of this study add important new details to our understanding of how Ulp1 interacts dynamically with its substrates and also provides potentially useful new directions for the study of Ulp1-interacting proteins.

In this study, we have demonstrated that region 3 of Ulp1, the catalytic domain, contains critical information for subcellular targeting to sumoylated substrates, including the septin Cdc11. To determine how Ulp1 is targeted to its substrates, we took advantage of a catalytically inactive Ulp1 mutant (C580S) that exhibited partial redistribution from the NE to the septin ring of dividing yeast cells. We found that the relocalization of Ulp1 depended on functional Smt3 and sumoylated proteins at the septin ring of dividing cells.

Importantly, using this novel Ulp1 in vivo septin-ring localization assay, we traced the critical targeting information to two features in region 3 of Ulp1, a previously identified SBS (amino acids 418 to 447) and a SUMO-contacting residue (D451) that resides near the carboxy terminus (see structure supplied). D451 of Ulp1 has previously been shown to contact Smt3 through a salt-bridge interaction [31, 47]. Indeed, in our analysis, we provide evidence that the salt bridge-disrupting D451N mutation abolishes the SUMO-dependent targeting of Ulp1 to septins and prevents the two-hybrid interaction of the catalytic domain of Ulp1 with Smt3 (Figures 3A, B and 5).

The sole requirements for the enrichment of full-length Ulp1 at the septin ring were the catalytically inactivating C580S mutation in the catalytic domain of Ulp1 and functional Kap121. This finding has important implications for the targeting role played by the amino-terminal karyopherin-binding domains of Ulp1 (discussed below). Additionally, Smt3 processing appears to be required for substrate release. The catalytically inactive Ulp1(3)^(C580S) mutant is predominantly localized to the septin ring and nucleus of dividing yeast cells, whereas the catalytically active WT Ulp1(3) shows merely diffuse staining throughout the cell (compare Figure 3B, top). This is not due to differential expression or stability of either GFP fusion, as similar amounts of proteins were detected in cell extracts (see Figure 6D). We propose and show in our biochemical analysis that the C580S mutant may trap Smt3-modified proteins, allowing it to be observed in association with cellular desumoylation substrates. In support of this assessment, combining the D451N with the C580S mutation abolishes all visible bud-neck localization (Figure 3A). Therefore, we propose that Ulp1^(C580S) first targets and docks Smt3 through the SBS domain and the salt bridge-forming D451 residue, then traps it in place because of its inability to cleave after the di-glycine motif of Smt3. We can assume that a trapped substrate prevents further catalysis or interactions with other Smt3 molecules, an assessment that is borne by our finding that, despite its septin-targeting and SUMO-binding properties, Ulp1(3)^(C580S) interacts only weakly with Smt3 in a two-hybrid assay (Figure 5 and Additional file 4). A better understanding of how the sumoylated substrate is trapped by Ulp1(3)^(C580S) may have important implications for the rational design of inhibitors for Ulp1-like SUMO proteases, but may await the elucidation of the cocrystal structure with a trapped substrate.

The interaction of budding yeast Ulp1 with Smt3 relies on multiple hydrophobic and salt-bridge interactions between the catalytic domain (region 3) of Ulp1 and the carboxy-terminal extension of Smt3. By making multiple contacts with Smt3, Ulp1 is particularly well-suited to interact with a wide variety of sumoylated substrates [22, 31]. Other SUMO proteases, Ulp2 and several SENP proteins (Senp1, Senp2, Senp6 and Senp7), are believed to interact noncovalently with their sumoylated substrates through dedicated SIMs [14, 22]. On the basis of our structure-function analysis of region 3, Ulp1 seems to employ a unique mode of interaction with Smt3 and sumoylated substrates. Ulp1 does not appear to contain bona fide canonical SIMs, and neither of the amino-terminal domains of Ulp1 (regions 1 and 2) interact with Smt3 or become enriched at the septin ring [51]. This assessment is also underscored by the arrangement of Smt3 and Ulp1 in the cocrystal structure [31]. The hydrophobic groove of Smt3 that would interact with a SIM-containing protein is turned away from the domains of Ulp1 that interact with Smt3. Interestingly, this may suggest that Ulp1 can be recruited to proteins that are covalently or noncovalently modified with SUMO and SUMO chains.

Our research demonstrates for the first time that noncovalent interactions between Ulp1 and SUMO are important not only for SUMO binding but also for the cytosolic targeting of this SUMO protease to the bud-neck and potentially to sumoylated septins. Septins are not the only cytosolic substrates of Ulp1 and are arguably the most prevalent [35, 38]; therefore, they may be readily scored in our septin ring targeting assay (Figures 3A and 3B). We predict that Ulp1 is also targeted to other cytosolic and septin-bound sumoylated substrate proteins, such as the karyogamy protein Kar9 [33]. However, because of the low local concentrations in comparison to sumoylated septins, these proteins may be hard to detect. We propose, however, that sumoylated proteins that accumulate or aggregate in the cytosol of yeast cells may be readily detectable by Ulp1(3)^(C580S).

As detailed below, Ulp1(3)^(C580S) also provides a useful tool with which to purify these sumoylated proteins (Figure 6E), and such studies are underway (O Kerscher and Z C Elmore, unpublished data). Our findings provide strong evidence that SUMO, at least in the case of sumoylated proteins at the septin ring, is a required signal for the cytoplasmic targeting of Ulp1. Our alignments of the SBS domain and the juxtaposed salt bridge-forming D451 residue reveal that this mode of targeting may also be conserved in other metazoan Ulp1-like SUMO proteases (Additional file 1).

Though we clearly show herein that Ulp1 becomes enriched at the septin ring, we do not yet fully understand how Ulp1 arrives at this subcellular localization. Our findings support the previous observation [27] that Kap121 plays an important role in promoting Ulp1 targeting to the septin ring. Similar to a previously described Ulp1 mutant that lacks the Kap60/Kap95 binding domain (region 2) [27], the septin ring localization of the full-length Ulp1^(C580S) protein described herein is dependent on functional Kap121. It is unlikely that the association with Kap121 shuttles Ulp1 to the septins. Rather, in mitosis, Kap121 becomes associated with a transport inhibitory nucleoporin, Nup53, and may thus exclude Ulp1 access to the inner phase of the NPC [52]. This suggests that, in the absence of Kap121-binding, a fraction of Ulp1 is free to associate with sumoylated septins. Our studies confirm that Ulp1 lacking the Kap60/Kap95-binding domain (region 2) is enriched at the NPC and the septin ring (Figure 3A). We extend these observations by showing that the ability to target sumoylated septins resides in the catalytic domain (region 3) of Ulp1. We found that a Ulp1(3)^(C580S) mutant, but not WT Ulp1(3), is enriched at the septin ring in the absence of Kap121. These data suggest that both the Ulp1(3)^(C580S) mutant and WT Ulp1(3) can interact with sumoylated septins, but, unlike the subtrate-trapping Ulp1(3)^(C580S) mutant, the catalytically active WT Ulp1(3) may be quickly released after desumoylation of the target protein, giving it a diffuse appearance in the cell. How Kap121 helps Ulp1-Δ2 to be retained at the septin ring in the absence of the C580S mutation is currently unclear. Kap121 may theoretically promote the interaction with a septin ring-localized protein. However, the localization of Kap121 to septins has not previously been reported.

One intriguing aspect of our study is the results of our analysis of the substrate-trapping Ulp1(3)^(C580S) construct. Three lines of evidence reveal the avid interaction of Ulp1(3)^(C580S) with SUMO proteins and sumoylated substrates. First, this Ulp1-derived construct shows a pronounced interaction with the bud-neck, which is composed of sumoylated septins in vivo. Second, Ulp1(3)^C580S interacts with Smt3 in a two-hybrid assay, whereas WT Ulp1(3) does not. Third, the purified recombinant Ulp1(3)^(C580S) protein is a potent affinity-tag for the purification of Smt3 conjugates and SUMO-modified proteins. Furthermore, in preliminary BiaCore analyses, we determined the K_d for the interaction between Ulp1(3)^(C580S) and SUMO1 to be 12.8 nM, which is about 200 times stronger than the interaction between a SIM and SUMO, as previously reported by Hecker and co-workers [53]. A related study involving the C603S mutant of the human SENP1 protease confirms our assessment of the substrate-trapping feature. Those authors observed relocalization of their SENP1(C603S) mutant in vivo to PML nuclear bodies and domains of the HDAC4 protein, suggesting that SUMO-dependent targeting may be a conserved feature of Ulp1-like SUMO proteases [54]. The latter may also provide a useful strategy for the identification of mitotically important desumoylation substrates. Indeed, two-hybrid screens with Ulp1(3)^(C580S) in the laboratory have already identified several novel cytosolic desumoylation targets (M Donaher and O Kerscher, unpublished data). We are also exploring the ability of Ulp1(3)^(C580S) to act as a SUMO chain-binding tag that can be used to promote the interaction of putative STUbL target proteins with RNF4 and other STUbLs (Figure 7).

How Ulp1 and other SUMO proteases target specific mitotic substrates for desumoylation remains unknown. Our analysis of SUMO-dependent Ulp1 targeting to the septin ring provides important evidence that Ulp1-like SUMO proteases do not passively await their desumoylation substrates, but rather dynamically localize to them in a cell-cycle-specific manner. Future experiments that take advantage of the SUMO-binding properties of the substrate-trapping Ulp1(3)^(C580S) construct may prove useful for the identification of clinically relevant targets of conserved Ulp1-like SUMO proteases in yeast and human cells.

Ulp1 remains one of the most enigmatic SUMO proteases with an essential role in cell-cycle progression. Our work focuses on the fundamental but unresolved question of how Ulp1, an NPC-associated SUMO protease, targets mitotic desumoylation targets in the cytosol. The current study of Ulp1 reveals for the first time that (1) specific determinants of Ulp1's catalytic domain are utilized to target the septins at the bud-neck of dividing cells, (2) Ulp1 requires SUMO and intact sumoylation for correct cytosolic targeting, (3) Ulp1, unlike Ulp2, relies on specific salt bridge-forming residues for substrate targeting, (4) septin-targeting information in the catalytic domain of Ulp1 is independent of Kap121 and (5) the substrate-targeting domain of Ulp1 can be modified to identify and purify sumoylated substrates of Ulp1 in genetic screens and in biochemical analyses.