Modification in Reverse: the SUMO Proteases

Home , Nucleoporin 37, SENP2, SUMO protein

Review TRENDS in Biochemical Sciences Vol.32 No.6

Modiﬁcation in reverse: the SUMO proteases

Debaditya Mukhopadhyay and Mary Dasso

Laboratory of Gene Regulation and Development, National Institute of Child Health and Development, National Institutes of Health, Building 18, Room 106, Bethesda, MD 20892, USA

SUMOs (small ubiquitin-like modifiers) are enzymatic mechanisms and the determinants of their ubiquitin-related proteins that become covalently con- specificity show many distinct features, which will be jugated to cellular target proteins that are involved in a the major topic of this review. variety of processes. Frequently, this modification has a key role in regulating the activities of those targets and, SUMO paralogs and pathway fundamentals thus, their cellular functions. SUMO conjugation is a Budding yeast express one SUMO protein (Smt3p), whereas highly dynamic process that can be rapidly reversed mammalian cells usually express three SUMO paralogs by the action of members of the Ubl (ubiquitin-like (SUMO-1, SUMO-2 and SUMO-3) [1]. Mature SUMO-1 protein)-specific protease (Ulp) family. The same family is 45% identical to SUMO-2 and SUMO-3, whereas of enzymes is also responsible for maturation of newly SUMO-2 and SUMO-3 are 95% identical to each other. synthesized SUMOs prior to their initial conjugation. The tails cleaved from each paralog are distinct. SUMO-2 Recent advances in structural, biochemical and cell bio- and SUMO-3 have sometimes been used interchangeably in logical analysis of Ulp/SENPs reveal their high degree of the literature. Here, we will refer to the isoform having two specificity towards SUMO paralogs, in addition to dis- amino acids (Val-Tyr) after the di-glycine motif as SUMO-2 crimination between processing, deconjugation and and that with 11 amino acids (Val-Pro-Ser-Ser-Leu-Ala-Gly- chain-editing reactions. The dissimilar sub-nuclear local- His-Ser-Phe) as SUMO-3. Where they cannot be distin- ization patterns of Ulp/SENPs and phenotypes of Ulp/ guished, SUMO-2 and SUMO-3 will be collectively referred SENP mutants further indicate that different Ulp/SENPs to as SUMO-2/3. An additional human SUMO paralog have distinct and non-redundant roles. (SUMO-4), which most-closely resembles SUMO-2 and SUMO-3, has been reported [8]. SUMO-4 is probably not Introduction conjugated under physiological conditions, so its biological SUMO (small ubiquitin-like modifier) modulates many role is unclear [9]. processes such as nuclear transport, transcription replica- SUMO conjugation occurs through a cascade of tion, recombination and chromosome segregation [1]. Like reactions that are performed by an activating enzyme ubiquitin, SUMOs are synthesized as propeptides that (E1), a conjugating enzyme (E2) and, usually, a SUMO require cleavage to reveal C-terminal di-glycine motifs ligase (E3) [1]. SUMO E1 and E2 enzymes resemble their prior to conjugation (Figure 1). Conjugation results in ubiquitin counterparts. Mammalian paralogs all become formation of an isopeptide bond between the SUMO C conjugated using the same E1 and E2 enzymes [1]. Never- terminus and an e-amino group of a lysine within the theless, there are important differences between paralogs. target protein. Enzymes responsible for SUMO processing First, individual targets show different conjugation pat- and deconjugation are called Ubl (ubiquitin-like protein)- terns: some targets are modified exclusively by SUMO-1 in specific proteases (Ulp) in yeast [2] and Sentrin-specific vivo [10], other targets are conjugated to SUMO-2/3, or proteases (SENP) in mammals [3]. Ulp/SENPs directly readily conjugated with all paralogs [11,12]. Second, the regulate the pools of free, conjugatable SUMO protein concentrations of both total and free SUMO-2/3 are higher and the half-life of conjugated species. than those of SUMO-1 [10]. Third, SUMO-1 and SUMO-2/3 Budding yeast has two Ulp/SENPs, humans have six show different in vivo dynamics [13] and responses to and Arabidopsis has seven [4,5] (Box 1). Although it is physiological stresses such as heat shock [10,13]. Finally, possible that novel Ulp/SENPs remain to be discovered, SUMO-2 and SUMO-3, like Smt3p, can form chains in vitro particularly in metazoans, the modest number of Ulp/ and in vivo, primarily through a conserved acceptor lysine SENPs seems striking by comparison to the number of (Lys15 in Smt3p, Lys11 in SUMO-2 or SUMO-3) [14–16].It enzymes in the ubiquitin pathway: there are >100 deubi- is not yet known whether individual SUMO moieties func- quitylating enzymes (DUBs) in higher eukaryotes [6] but tion distinguishably from SUMO chains in the same way only one ubiquitin polypeptide to process or deconjugate that single ubiquitin moieties and structurally distinct [7]. As an understanding of the biochemical properties polyubiquitin chains represent intracellular signals that of Ulp/SENPs emerges, it has become clear that their are functionally distinct from each other [7]. However, there is evidence that SUMO-2/3 chain formation might

Corresponding author: Dasso, M. ([email protected]). be crucial for in vivo regulation of some targets [17,18]. Available online 17 May 2007. SUMO-1 does not have an acceptor lysine at the equivalent www.sciencedirect.com 0968-0004/$ – see front matter ß 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.tibs.2007.05.002 Review TRENDS in Biochemical Sciences Vol.32 No.6 287

Figure 1. SUMO paralogs and Ulp/SENP catalyzed reactions. (a) The primary sequences of the single budding yeast SUMO protein, Smt3, are aligned with human SUMO- 1–4, to show features that are essential for their function. An essential di-glycine motif (red) is revealed by processing before conjugation. Sequences removed through the processing reaction are shown in blue. Conserved lysine residues within Smt3, SUMO-2 and SUMO-3 are the major sites of chain linkage and are shown in purple. A residue in SUMO-4 (Pro90) that prevents its processing and deconjugation by Ulp/SENPs is shown in green. (b) A schematic representation of reactions catalyzed by Ulp/SENPs. The processing reaction (broken black box) involves cleavage of a peptide bond within the SUMO pro-peptide to reveal the C-terminal di-glycine motif. Processing requires Ulp/SENPs to directly recognize newly translated SUMOs as substrates. SUMO deconjugation (broken red box) requires cleavage of the amide bond between the C terminus of the mature SUMO and the e-amine group of the target lysine within the substrate. In principle, recognition in this reaction could involve both SUMO proteins and binding surfaces contributed by the substrate. There is currently little evidence for the importance of the latter. Chain editing (solid red box) is chemically identical to deconjugation, although it is distinguished by cleavage of one or more SUMOs from a poly-SUMO chain rather than from another cellular target substrate. Enzymes of the Ulp2 sub-family are important for this reaction, and a feature(s) of the SUMO chain is likely to contribute towards substrate recognition. E1, E2 and E3 are the activating, conjugating and SUMO ligase enzymes of the conjugation pathway, respectively. For more details about this pathway, see Ref. [1]. site, although it can form chains in vitro [19] using Lys7, showed that Ulp1p bears no substantial relationship to Lys16 and Lys17 [20]. DUBs but is related to adenoviral proteases [2,21]. Sequence comparisons further predicted a 200 amino Identiﬁcation of Ulp/SENPs acid protease fold, which deﬁnes this group of enzymes Li and Hochstrasser [2] used an elegant sib-selection (C48 cysteine proteases), and showed that they diverged strategy to isolate SUMO proteases: they assayed cleavage from DUBs early in evolution [2,22,23]. C48 family mem- of a model Smt3p substrate within extracts made from bers exist in all eukaryotic species examined. Database pools of bacterial transformants expressing multiple yeast searches found one other family member in budding yeast proteins. Pools with cleavage activity were subdivided to (Ulp2p) [24] and seven related proteins in humans, which isolate individual cDNAs encoding Smt3p proteases. Using were named SENPs [3]. SENPs were originally designated this approach, they found a previously uncharacterized 72- purely on the basis of sequences found through database kDa protein, which they named Ulp1p. Sequence analysis searches. It was later discovered that the closely related SENP3 and SENP4 sequences corresponded to the same protein; as a result, there is a discontinuity within Box 1. SUMO proteases in plants the SENP nomenclature. Many mammalian Ulp/SENPs In plants, sumoylation has been implicated in abiotic stress have been given alternative names owing to their discovery response, pathogen defense, abscisic acid signaling and flower through multiple means (Table 1). It should also be noted induction [5]. Arabidopsis thaliana has eight genes that encode complete SUMO proteins [5]. Database searches using the Ulp1p that one Ulp/SENP family member, SENP8 – also known conserved domain as the search parameter yield >100 hits in the as deneddylase 1 – does not act on SUMOs. Instead, it acts Arabidopsis genome. These sequences include a family of 97 on another ubiquitin-like protein, Nedd8 [25,26]. potentially functional genes and apparent pseudogenes that arose Ulp1p and Ulp2p are not redundant. Ulp1p is encoded through selfish gene trans-duplication [68]. Currently, it seems that by an essential gene [2]. Over-expression of processed there are only eight Ulp/SENP proteases expressed in Arabidopsis [5] (Figure 2), one of which is Nedd8-specific [4]. Notably, plant Smt3p weakly rescues Dulp1 cells, but full-length Smt3p SENPs display SUMO paralog specificity [4,69], and at least one of does not [2], indicating that Ulp1p has an important role in these enzymes localizes to the nuclear envelope [70], indicating Smt3p maturation. Interestingly, mutants with decreased conservation of Ulp/SENP function between the plant and animal levels of Ulp1p activity have pronounced defects in main- kingdoms. tenance of 2 mm circle copy number [27].2mm circles www.sciencedirect.com 288 www.sciencedirect.com

Table 1. SUMO protease expression, localization and activitya Review RNSi iceia Sciences Biochemical in TRENDS o.2No.6 Vol.32

aAbbreviations: AXAM2, Axin-associating molecule 2; SMT3IP1 and 2, Smt3-specific isopeptidase 1 and 2; SSP1, SUMO-1-specific protease; SuPr-2, SUMO protease 2; SUSP1, SUMO-specific protease 1. bRed ellipse, SUMO-recognition motif (Val/Ile-Xaa-Val/Ile-Val/Ile) [67]; green box, regions required for proper nuclear localization; blue box, Crm-dependant nuclear export signal; CD, conserved catalytic domain. Review TRENDS in Biochemical Sciences Vol.32 No.6 289 exist in most budding yeast strains as multi-copy to nucleoli [39–41]. SENP6 and SENP7 fall within the Ulp2 extrachromosomal elements that propagate stably in host branch, and both localize to the nucleoplasm [35,42].All populations. ulp1 mutants show a nibbled morphology, SENPs showing unconventional domain architecture are characterized by a scalloped edge of individual colonies part of the Ulp2-containing branch. due to lineages within the colonies that possess more than Most structural and enzymatic analyses of Ulp/SENPs double normal levels of 2 mm circles. Ulp1p remains essen- have been undertaken on members of the Ulp1 sub-family. tial even in the absence of the 2 mm circle, however, The crystal structure of a transition-state analog contain- indicating that regulation of 2 mm copy number is not its ing Ulp1p linked to the C-terminal glycine residue of only crucial function [27]. Ulp1p also has important roles in Smt3p shows an extensive hydrophilic interface that does nuclear–cytoplasmic trafficking [28], and particularly in not require substantial conformational change in Smt3p 60S pre-ribosomal particle export [29]. ulp1 mutants for binding [21]. The catalytic site of Ulp1p lies within a share many phenotypes in common with mutants of other shallow, narrow cleft that comprises conserved amino SUMO-pathway components, including enzymes of the acids, which recognize the di-glycine motif of Smt3p. When conjugation machinery. Examples of this phenomenon inserted into the cleft, the di-glycine motif is clamped include the finding that mutants lacking E3 activities show within a hydrophobic ‘tunnel’ by conserved tryptophans nibbled morphology and are defective in 2 mm copy main- and other adjacent residues. Substitution of either residue tenance [30], and the finding that both E1 and E2 mutants of the Smt3p di-glycine motif would prohibit binding show defective ribosomal protein export [29]. These sim- because of a resultant steric clash within the tunnel walls. ilarities support the idea that ulp1 phenotypes arise from The active sites of SENP1 and SENP2 show a similar reduced Smt3p processing and the consequential lack of organization, with conserved tryptophans again clamping conjugation, rather than from impaired deconjugation. vertebrate SUMOs within a hydrophobic channel [43,44] By contrast, Ulp2p might be more important for (Box 2). Notably, a kink that is caused in the SUMO-4 deconjugation [31] and, specifically, in dismantling of po- chain by Pro90 should be non-permissive for interaction ly-Smt3p chains [15] (see later). Ulp2p is not essential for with the catalytic sites of either SENP1 or SENP2. Con- vegetative growth, although it is important for meiotic de- sistent with this, SUMO-4 is not processed [9]. Although velopment, accurate chromosome segregation and recovery SUMO-4 can become conjugated to cellular proteins when from DNA replication and spindle checkpoint arrest [23,24]. expressed in a mature form [8], deconjugation of these Dulp2 mutants arrest in meiotic prophase [24], and Ulp2p species has not been demonstrated; as with processing, has been implicated in remodeling of Smt3p-containing foci deconjugation of SUMO-4 should be blocked by Pro90. during the formation of synaptonemal complexes (protein- aceous structures assembled between homologous chromo- Ulp/SENPs enzymatic activities somes during prophase I of meiosis for reciprocal exchange Processing of genetic material) [32]. During mitosis, ulp2 mutants show The catalytic domains of SENP1 and SENP2 discriminate premature loss of centromeric cohesion [33] and defective between SUMO paralogs as processing substrates: SENP1 rDNA condensation [23]. processes SUMO-1 with greater efficiency than SUMO-2, but shows limited activity towards SUMO-3 [44,45]. Evolutionary relations and structure of Ulp/SENPs SENP2 processes SUMO-2 more efficiently than SUMO- The C48 catalytic domain is typically located close to the C 1, but again processes SUMO-3 poorly [43]. Because Ulp/ terminus of Ulp/SENPs, whereas N-terminal domains SENPs directly recognize the newly translated SUMOs in frequently direct their subcellular localization [34,35] these reactions, determinants of this specificity must lie (Table 1). There are two major variations on this theme: within the SUMO sequences themselves. Swapping resi- in the case of Ulp2 [34] and two plant Ulp2-like proteins dues that are C-terminal to the di-glycine motif amongst (NP_195088 and NP_172444. [5]), the C48 domain is in the paralogs demonstrates that these sequences are largely middle of the protein instead of being C-terminal; more dramatically, the C48 domain is split in half by 50–200 amino acid insertions in vertebrate SENP6 and SENP7. Box 2. Product inhibition The role of these additional sequences is unknown, but both Ulp/SENPs recognize both processing and conjugation substrates Ulp2p and SENP6 are enzymatically active [24,35].Com- through binding of the SUMO polypeptide. One remarkable parison of C48 peptidases across eukaryotic species [35] consequence of this mechanism is that mature SUMO proteins (Figure 2) shows an early branch point separating the (with a C-terminal di-glycine motif) bind to Ulp/SENPs with high SENP8/deneddylase 1-related sequences from the SUMO- affinity. In the case of SENP1, for instance, the mature SUMO reaction product binds with higher affinity than SUMO precursors specific proteases, which is consistent with the finding that do [47], raising the possibility that SUMO proteins can inhibit their these enzymes act on Nedd8. True SUMO proteases further own production and/or cleavage of conjugated species. The diverge into two major branches, one containing Ulp1 and biological importance of this inhibition is not yet understood, but the other Ulp2 (Figure 2). SENP1, SENP2, SENP3 and one intriguing speculation is that localization to nuclear sites with SENP5 fall within the Ulp1 branch. Within this subset, high substrate concentrations might help to relieve this inhibition by enabling substrates to compete more effectively for Ulp/SENP SENP1 and SENP2 are closely related to each other, and binding [71]. In this case, product inhibition would actually promote both localize within the cell at the nuclear envelope through the specificity of Ulp/SENPs by enhancing their activity within the association to the nuclear pore complex (NPC) [36–38]. correct sub-nuclear compartments. Alternatively, uncharacterized Similarly, SENP3 and SENP5 show the highest degree of cofactors might function in vivo to regulate Ulp/SENP through by facilitating product release. homology with each other and share a common localization www.sciencedirect.com 290 Review TRENDS in Biochemical Sciences Vol.32 No.6

Figure 2. Evolutionary relationship of Ulp/SENP family members. The eukaryotic genome databases at National Center for Biotechnology Information (NCBI; http:// www.ncbi.nlm.nih.gov/sutils/genom_table.cgi?organism=euk) were searched using psi-BLAST using human SENP protein sequences. The retrieved sequences were aligned and phylogenetic analysis was performed as described [35]. An early branch point separates the Nedd8-specific proteases (Den1-like) from true SUMO proteases. The SUMO proteases have further diverged into two major branches. These branches are called Ulp1-like and Ulp2-like, reflecting the assignments of the yeast enzymes (red) into different branches. The closely related SENP1, SENP2 (green), SENP3 and SENP5 (blue) are in the Ulp1-like branch, whereas SENP6 and SENP7 (gray) occupy the Ulp2-like branch. For each protein sequence, the letters indicate the shortened binomial nomenclature, and the numbers indicate the protein Genbank accession numbers. Abbreviations for species are as follows: Aga, Anopheles gambiae; Ani, Aspergillus nidulans; Ath, Arabidopsis thaliana; Cel, Caenorhabditis elegans; Cgl, Candida glabrata; Cpa, Cryptosporidium parvum; Ddi, Dictyostelium discoideum; Dme, Drosophila melanogaster; Dre, Danio rerio; Ehi, Entamoeba histolytica; Gla, Giardia lamblia; Gze, Gibberella zeae; Hsa, Homo sapiens; Lma, Leishmania major; Mgr, Magnaporthe grisea; Mmu, Mus musculus; Ncr, Neurospora crassa; Sce, Saccharomyces cerevisiae; Spo, Schizosaccharomyces pombe; Tcr, Trypanosoma cruzi; Xle, Xenopus laevis; Xtr, Xenopus tropicalis. responsible for differences in SENP1 catalysis [45]. Pro94 of Recently, structural analysis of SUMO proteins bound SUMO-3 is particularly important for its inefficient matu- to inactive SENP1 and SENP2 mutants have shown that ration [45]. In addition, SENP1 processing of SUMO-1 is both enzymes induce isomerization of the scissile peptide enhanced by a histidine residue in the SUMO-1 C-terminal bond during processing, resulting in a 908 kink in the tail (His98) [44]. SUMO C-terminal tails likewise determine SUMO C-terminal tails [46,47] (Figure 3). Remarkably, the efficiency of SENP2 processing [43] and Pro94 of SUMO- the efficiency of this isomerization largely determines the 3 again contributes towards its inefficient maturation by relativeratesofSUMO-1versusSUMO-3maturationby SENP2. SENP1 [47]:TheKm values for both substrates are www.sciencedirect.com Review TRENDS in Biochemical Sciences Vol.32 No.6 291

Figure 3. Key mechanistic steps of SUMO deconjugation by Ulp/SENP. A schematic representation of the mechanism whereby Ulp/SENPs deconjugation SUMOs from their targets is shown. (1) Free Ulp/SENPs (brown) recognize the SUMO moiety (green) conjugated to the target protein (orange). (2) Upon binding, the substrate adopts a cis conformation around the scissile amide bond (red), between the terminal glycine of SUMO (black) and lysine side-chain of a target protein (light blue). (3) The sulfide group of the active cysteine residue (dark blue) nucleophilically attacks the carbonyl carbon of the scissile amide bond. This results in release of free target protein and formation of the SUMO-Ulp/SENP thioester intermediate. (4) The thioester bond (green) is rapidly hydrolyzed. (5) The free unconjugated SUMO dissociates to release the free enzyme in a process that is thermodynamically unfavorable (see Box 2). Processing reactions of Ulp/SENPS are mechanistically similar, except that a peptide bond after the di-glycine motif gets cleaved instead of the amide bond.

similar, but the catalytic constant (kcat)forSUMO-3is Deconjugation 50-fold lower than that for SUMO-1. Re-orientation of the SUMO deconjugation involves cleavage of the amide bond scissile bond of SUMO-1 is favored because it enables between the C terminus of SUMO and the e-amine group of SUMO-1-His98 to form a hydrogen bond with SENP1- the target lysine (Figures 1 and 3). Because the binding Gly600, whereas the rigidity caused by SUMO-3-Pro94 interfaces between Ulp/SENPs and mature SUMO-2 or might inhibit isomerization [47].TheKm of SENP2 for SUMO-3 are nearly identical, it seems unlikely that there SUMO-1 is 14-fold higher than for SUMO-2, whereas its is substantial discrimination between these paralogs during kcat for both substrates is similar, indicating that SENP2 deconjugation [46,47]. Nevertheless, Ulp/SENPs can clearly binds these paralogs differentially but mediates their differentiate SUMO-1 from SUMO-2/3 [39,40,46]. Struc- maturation at comparable rates [46]. By contrast, the tural analysis of catalytically inactive SENP1 bound to a Km of SENP2 for SUMO-2 and SUMO-3 are nearly iden- model SUMO-1 deconjugation substrate shows that the tical, but SENP2 has a lower kcat for SUMO-3, indicating scissile isopeptide bond adopts a cis configuration, resulting that differential processing of these two paralogs are due in a 908 kink [47] (Figure 3). For these studies, the substrate to catalytic discrimination [46]. As with SENP1, it is likely was derived from residues 418–587 of Ran-GTPase activat- that this discrimination is due to rigidity of SUMO-3- ing protein 1 (RanGAP1), which is a major SUMO-1 conju- Pro94 [46,47]. gation substrate [10]. An inactive SENP2 mutant bound SENP5 is essentially inactive for SUMO-1 processing, similar model substrates containing either SUMO-1 or but efficiently catalyzes SUMO-3 maturation [39,40]. SUMO-2 show that the scissile bond likewise adopts a SENP6 does not function as a processing enzyme [35]. right-angled bend [46]. Notably, the four alkyl hydrocarbons www.sciencedirect.com 292 Review TRENDS in Biochemical Sciences Vol.32 No.6 of the lysine side-chain have higher degrees of freedom than SUMO and the e-amine of a conjugated lysine (Figure 1). the more-rigid peptide extension of unprocessed SUMOs. As Nevertheless, it seems that Ulp2p family members are a result of this difference, the kinked conformation required specialized for chain editing. In yeast, ULP2 deletion causes for productive catalysis might be more easily assumed by an accumulation of high-molecular-weight conjugated species isopeptide bond than by a peptide bond, explaining the fact in the presence of wild-type Smt3p, but not in smt3 mut- that SENP2 acts more efficiently in deconjugation than in ants that lack critical lysine residues [15]. Moreover, such processing [46]. non-chain-forming smt3 mutants suppress some Dulp2 phe- SENP1 deconjugates RanGAP1–SUMO-1 and notypes [15], indicating that those defects arise from Smt3p- RanGAP1–SUMO-2 with indistinguishable efficiencies chain accumulation. SENP6 is inactive for processing of all [44], reflecting the nearly identical Km for cleaving each paralogs, for deconjugation of model substrates with single of these substrates [47]. This situation is obviously remi- SUMO moieties and for disassembly of SUMO-1 chains [35]. niscent of its nearly identical Km values for cleaving of However, it readily dismantles substrates containing three SUMO-1 and SUMO-2 processing substrates, albeit with or more SUMO-2/3 moieties [35,49]. Mammalian cells show the difference that SENP1 catalytically discriminates be- subtle changes in SUMO-2/3-conjugation patterns and tween paralogs during the maturation reaction. Indeed, distribution after depletion of SENP6 by RNAi [35], perhaps such similarities are strongly predicted by the fact that the reflecting persistence of a redundant enzyme; the unchar- interfaces between SUMOs and SENP1 during processing acterized SENP7 might be the best candidate for such an and deconjugation are essentially the same outside of the activity (Figure 2). released tail or target protein [46,47]. Conversely, SENP2 is less efficient at deconjugation of RanGAP1–SUMO-1 Regulation through sub-nuclear localization and than RanGAP1–SUMO-2, reflecting its higher Km for expression patterns the former [46]. This situation again runs parallel with The requirements for localization are best understood for the efficiency of SENP2 in processing, in which its higher Ulp/SENPs associated with the NPC. Neither enzymatic Km for SUMO-1 determines the relative efficiency with function nor rescue of Dulp1 cells requires NPC localization which it discriminates between these paralogs. SENP3 and of Ulp1p, but this localization seems to control Ulp1p SENP5 are more active in deconjugating SUMO-2/3-con- specificity [34]. Truncated Ulp1p mutants that do not bind jugated targets than SUMO-1-containing species [39,40]. the NPC robustly suppress growth defects of Dulp2 cells, Future analysis of how SENP3 and SENP5 achieve such a and eliminate the majority of the Smt3p-conjugated high level of specificity will be extremely interesting. species that characterize the Dulp2 strain. These changes The proximity of the target lysine to Ulp/SENP active have been interpreted as evidence that such truncated sites offers the possibility that they might recognize features Ulp1p mutants act on nucleoplasmic Ulp2p substrates that of the conjugation targets. Modeling of Ulp1p originally would normally be inaccessible [34]. At the same time, suggested that conjugated target proteins could approach increased accumulation of other species might suggest that within 5–7 A˚ of the active site without extensive structural they are Ulp1p substrates, the efficient deconjugation of distortion [21]. This geometry would enable Ulp1p to act which requires its association to the NPC [34]. Interest- upon a wide variety of conjugated species, based primarily ingly, fragments of Ulp1p that possess only the catalytic upon recognition of Smt3p. The resultant capacity to decon- domain (residues 404–621) but lack all N-terminal jugate Smt3p promiscuously is consistent with very few Ulp/ sequences exert a dominant-lethal effect on cell growth SENPs found in yeast. In vertebrates, a similar level of [21,50]. It seems that this lethality is due to excessive target promiscuity is indicated by the fact that the Ran- amounts of catalytic activity within the nucleoplasm. GAP1 moiety makes little contribution to binding of model The dominant-negative phenotype can be reversed by substrates to either SENP1 [47] or SENP2 [46]. By contrast, restoration of a coiled-coiled domain (residues 341–403) SENP1 can discriminate between SUMO-1 conjugated to that confers nuclear export by the Crm1 transport recep- RanGAP1 and SUMO-1 conjugated to Sp100 [44],another tor, or by addition of cytoplasmic retention sequences that conjugation target that was first identified through its cause almost exclusive localization within the cytosol [50]. recognition by auto-antibodies from patients with pri- Cytoplasmically tethered forms of Ulp1p rescue Dulp1 cells mary biliary cirrhosis. Moreover, although SENP1 removes [50], arguing that little Ulp1p activity is required within SUMO-2/3 from all lysine residues within PML (promyelo- the nucleus itself, if any. cytic leukemia) protein, SENP3 and SENP5 cleave conju- Multiple NPC components have been implicated in the gates at some lysine residues more efficiently than others localization of Ulp1p. The N terminus of Ulp1p (residues 1– [40]. Finally, SENP1 efficiently cleaves SUMO-2–Reptin in 404) uses two distinct binding sequences to interact with vitro, whereas SENP2 does not [48]. Collectively, these two nuclear import receptors, Pse1p and importin-a/b [50]. findings indicate that at least some conjugated species The Ran-GTPase normally governs transport receptor are not well accommodated by all Ulp/SENPs, possibly binding and release of cargo [51]. Unlike typical cargos, because the structure of target proteins is incompatible with however, Ulp11–404 is not displaced from either receptor by Ulp/SENP binding. GTP-bound Ran protein [50]. Interference with Pse1p and importin-a/b binding to the NPC causes mislocalization of Chain editing Ulp1p, which further supports the importance of these Dismantling of SUMO chains and deconjugation from target interactions [50]. Release of Ulp1p from the NPC during proteins are chemically identical – both reactions require mitosis is required for it to act on cytosolic substrates, such cleavage of an amide bond between the C terminus of as septins [52,53]. This release is accomplished through www.sciencedirect.com Review TRENDS in Biochemical Sciences Vol.32 No.6 293 mitosis-specific rearrangements of the NPC that alter the Human SENP1 [45], rat SENP2 [61] and human SENP6 association of Pse1p with NPCs and inhibit its activity as a [62] are all expressed in a tissue-specific manner, although receptor for nuclear import [54]. Moreover, Pse1p is expression patterns of other mammalian SENPs have not required for targeting of Ulp1p to the septin ring within been reported. Moreover, human SENP1 is rapidly, but mitotic cells [54]. transiently, up-regulated during keratinocyte differen- Mlp1p and Mlp2p are two myosin-like NPC-associated tiation [63]. Because SENP1 seems to have the strongest proteins that have been implicated in the retention of activity against SUMO-1, these findings might indicate unspliced mRNAs and in spindle-pole-body formation developmental control of SUMO-1 dynamics or regulated [55,56]. Loss of Mlp1p and Mlp2p, or of a nucleoporin that deconjugation of cytoplasmic sumoylated species [64]. Con- is required for their localization (Nup60), displaces Ulp1p sistent with this, retroviral insertions within the gene from NPCs [57]. Interestingly, Dmlp1Dmlp2 mutants encoding mouse SENP1 cause placental abnormalities show a nibbled morphology similar to hypomorphic ulp1 and embryonic lethality [65]. Likewise, Xenopus laevis mutants [27,57], and epistasis analysis indicates that SENP1 (xSENP1) is regulated during development in a Ulp1p is a downstream component of the same genetic stage- and tissue-specific manner [66], with prominent pathway for 2 mm circle maintenance [57]. Like Mlp1p and expression in the central nervous system after the tail- Mlp2p, Ulp1p does not accumulate on NPCs that are bud stage. The pattern of xSENP1 expression is particu- adjacent to the nucleolus in yeast [57]. Nevertheless, Ulp1p larly interesting in light of head defects caused by its has been implicated in ribosome biogenesis [29]. Mtr2p is a overproduction. protein that associates with export-competent pre-60S particles, promoting their export [58]. The mtr2-33 allele Box 3. Outstanding issues shows a synthetic enhancement of pre-60S ribosomal sub- unit export defects when combined with temperature-sen- Recent reports have provided an increasingly sophisticated understanding of the enzymatic mechanisms of Ulp1, SENP1 and SENP2. sitive alleles of ulp1 [29]. These findings indicate that However, a similar mechanistic understanding of other Ulp/SENPs Ulp1p-mediated Smt3p processing or deconjugation is has not yet been achieved. Issues related to other aspects of Ulp/ important for a late step in pre-60S export, perhaps as SENP regulation and to SUMO paralog identity also remain to be the particle interacts with the NPC. explored. In vertebrates, the extreme N terminus of SENP2 is Outstanding issues include: SUMO-1 is processed and deconjugated by SENP1, and with required for association to the nucleoplasmic face of the somewhat lower efficiency by SENP2 [43–45]. Other SENPs seem NPC [37]. Transient over-expression of SENP2 lacking this to neither process nor deconjugate SUMO-1 [35,39,40]. These data domain in mammalian cells reduces accumulation of collectively suggest SUMO-1 might be much less dynamic than SUMO conjugates more effectively than full-length SUMO-2, which is consistent with experimental observations [13]. SENP2, suggesting that its restriction to the NPC prevents It is possible to speculate that SUMO-1 has a distinct role that is associated with or regulated by components of the NPC because its inappropriate deconjugation of nucleoplasmic sub- both SENP1 and SENP2 are bound to this structure. strates in a manner analogous to Ulp1p [37]. SENP2 Mature SUMO-2 and SUMO-3 are nearly identical in sequence and can bind to the Nup153 nucleoporin [37,38], although it behave similarly [13]. Nevertheless, SUMO-3 is inefficiently has not been demonstrated that this is the only interaction processed by SENP1 and SENP2 [43–45] and shows a more that mediates SENP2 localization in vivo. Sequences tissue-specific expression pattern than either SUMO-1 or SUMO-2 [45]. It is tempting to speculate that inefficient processing of within the N-terminal domain of SENP1 are sufficient SUMO-3 by SENP1 and SENP2 might regulate the production of for its correct targeting [36], but no binding partner respon- processed SUMO proteins, or to restrict this reaction to a subset sible for its localization has been reported. As with Ulp1p of Ulp/SENPs (possibly SENP3 and/or SENP5). [50], both human SENP1 [59] and SENP2 [60] are subject SENP3 and SENP5 show a high level of discrimination against SUMO-1 [39–41]. It will be of considerable interest to determine to export from nuclei by Crm1, a Ran-dependent nuclear the structural basis by which they discriminate between SUMO-1 export receptor. For SENP1, Crm1-mediated export might and SUMO-2/3. underlie cell-type-specific re-localization to the cytoplasm Most aspects of sub-nuclear targeting remain to be understood: it [59]. For SENP2, nuclear-cytoplasmic shuttling seems to will be interesting to determine how transport receptors, Mlp1p be important for turnover via the ubiquitin-proteasome and Mlp2p cooperate to ensure docking of Ulp1p at the NPC [29,57]. For other Ulp/SENPs, the mechanisms governing their pathway [60]. localization remain entirely unknown. Because SENP3 and SENP5 are both nucleolar Ulp1 Nothing is currently known about the structure of Ulp2-related family members [39,40], it is possible that they might SUMO proteases (i.e. Ulp2, SENP6 and SENP7) or how they have a role in pre-60S export, as Ulp1p does [29]. Defective recognize SUMO chains. It is interesting to speculate that the ribosome biogenesis would be consistent with the failure of sequences within or adjacent to their catalytic domains might contribute to their specificity [35,67]. cells to progress through the cell cycle after SENP5 Most enzymatic and structural studies on Ulp1p, SENP1 and depletion [39]. The localization of SENP3 and SENP5 SENP2 have been undertaken using only the catalytic domains of might also contribute to the observation that SUMO-2/3- these proteins [21,43–47]. It will be important to reconstitute conjugated species do not accumulate within nucleoli, processing and deconjugation reactions using full-length en- whereas SUMO-1-conjugated species are comparatively zymes to understand the contribution of sequences outside of their catalytic domains to their specificity and regulation. This will concentrated there [13]. Although localization signals have be particularly true for enzymes that possess SUMO-binding been mapped within the N-terminal domains of mouse motifs outside of their catalytic domains (see Table 1). The use of SENP3 [41] and human SENP5 [40], the interactions model deconjugation substrates based upon different SUMO responsible for their localization with the nucleolus remain conjugation targets will likewise be informative regarding the nature of target recognition by Ulp/SENPs. to be elucidated. www.sciencedirect.com 294 Review TRENDS in Biochemical Sciences Vol.32 No.6

Concluding remarks 23 Strunnikov, A.V. et al. (2001) Saccharomyces cerevisiae SMT4 encodes Ulp/SENPs strongly differentiate amongst processing and an evolutionarily conserved protease with a role in chromosome condensation regulation. Genetics 158, 95–107 deconjugation substrates, arguing that they are not highly 24 Li, S.J. and Hochstrasser, M. (2000) The yeast ULP2 (SMT4) gene promiscuous enzymes, as was once thought. Their distinct encodes a novel protease specific for the ubiquitin-like Smt3 protein. subcellular localizations, expression patterns and genetics Mol. Cell Biol. 20, 2367–2377 further argue that they have well-defined and non-redun- 25 Gan-Erdene, T. et al. (2003) Identification and characterization of DEN1, dant roles in vivo. For vertebrate cells, the capacity of Ulp/ a deneddylase of the ULP family. J. Biol. Chem. 278, 28892–28900 26 Wu, K. et al. (2003) DEN1 is a dual function protease capable of SENPs to discriminate among substrates might be particu- processing the C terminus of Nedd8 and deconjugating hyper- larly important for maintaining paralog-specific patterns neddylated CUL1. J. Biol. Chem. 278, 28882–28891 of conjugation to cellular targets. The next few years will 27 Dobson, M.J. et al. (2005) The 2 microm plasmid causes cell death in offer fascinating opportunities to understand the cellular Saccharomyces cerevisiae with a mutation in Ulp1 protease. Mol. Cell roles of Ulp/SENPs and the mechanisms through which Biol. 25, 4299–4310 28 Stade, K. et al. (2002) A lack of SUMO conjugation affects cNLS- they are regulated (Box 3). dependent nuclear protein import in yeast. J. Biol. Chem. 277, 49554–49561 References 29 Panse, V.G. et al. (2006) Formation and nuclear export of preribosomes 1 Johnson, E.S. (2004) Protein modification by SUMO. Annu. Rev. are functionally linked to the small-ubiquitin-related modifier Biochem. 73, 355–382 pathway. Traffic 7, 1311–1321 2 Li, S.J. and Hochstrasser, M. (1999) A new protease required for cell- 30 Chen, X.L. et al. (2005) Misregulation of 2 microm circle copy number in cycle progression in yeast. Nature. 398, 246–251 a SUMO pathway mutant. Mol. Cell. Biol. 25, 4311–4320 3 Yeh, E.T. et al. (2000) Ubiquitin-like proteins: new wines in new 31 Schwienhorst, I. et al. (2000) SUMO conjugation and deconjugation. bottles. Gene 248, 1–14 Mol. Gen. Genet. 263, 771–786 4 Colby, T. et al. (2006) SUMO-conjugating and SUMO-deconjugating 32 Cheng, C.H. et al. (2006) SUMO modifications control assembly of enzymes from Arabidopsis. Plant Physiol. 142, 318–332 synaptonemal complex and polycomplex in meiosis of Saccharomyces 5 Novatchkova, M. et al. (2004) SUMO conjugation in plants. Planta. 220, cerevisiae. Genes Dev. 20, 2067–2081 1–8 33 Bachant, J. et al. (2002) The SUMO-1 isopeptidase Smt4 is linked to 6 Amerik, A.Y. and Hochstrasser, M. (2004) Mechanism and function of centromeric cohesion through SUMO-1 modification of DNA deubiquitinating enzymes. Biochem. Biophys. Acta. 1695, 189–207 topoisomerase II. Mol. Cell 9, 1169–1182 7 Pickart, C.M. and Fushman, D. (2004) Polyubiquitin chains: polymeric 34 Li, S.J. and Hochstrasser, M. (2003) The Ulp1 SUMO isopeptidase: protein signals. Curr. Opin. Chem. Biol. 8, 610–616 distinct domains required for viability, nuclear envelope localization, 8 Bohren, K.M. et al. (2004) A M55V polymorphism in a novel SUMO and substrate specificity. J. Cell Biol. 160, 1069–1081 gene (SUMO-4) differentially activates heat shock transcription factors 35 Mukhopadhyay, D. et al. (2006) SUSP1 antagonizes formation of highly and is associated with susceptibility to type I diabetes mellitus. J. Biol. SUMO2/3-conjugated species. J. Cell Biol. 174, 939–949 Chem. 279, 27233–27238 36 Bailey, D. and O’Hare, P. (2004) Characterization of the localization 9 Owerbach, D. et al. (2005) A proline-90 residue unique to SUMO-4 and proteolytic activity of the SUMO-specific protease, SENP1. J. Biol. prevents maturation and sumoylation. Biochem. Biophys. Res. Chem. 279, 692–703 Commun. 337, 517–520 37 Hang, J. and Dasso, M. (2002) Association of the human SUMO-1 10 Saitoh, H. and Hinchey, J. (2000) Functional heterogeneity of small protease SENP2 with the nuclear pore. J. Biol. Chem. 277, 19961– ubiquitin-related protein modifiers SUMO-1 versus SUMO-2/3. J. Biol. 19966 Chem. 275, 6252–6258 38 Zhang, H. et al. (2002) Enzymes of the SUMO modification pathway 11 Azuma, Y. et al. (2003) SUMO-2/3 regulates topoisomerase II in localize to filaments of the nuclear pore complex. Mol. Cell Biol. 22, mitosis. J. Cell Biol. 163, 477–487 6498–6508 12 Vertegaal, A.C. et al. (2006) Distinct and overlapping sets of SUMO-1 39 Di Bacco, A. et al. (2006) The SUMO-specific protease SENP5 is and SUMO-2 target proteins revealed by quantitative proteomics. Mol. required for cell division. Mol. Cell Biol. 26, 4489–4498 Cell Proteomics 5, 2298–2310 40 Gong, L. and Yeh, E.T. (2006) Characterization of a family of nucleolar 13 Ayaydin, F. and Dasso, M. (2004) Distinct in vivo dynamics of SUMO-specific proteases with preference for SUMO-2 or SUMO-3. J. vertebrate SUMO paralogues. Mol. Biol. Cell 15, 5208–5218 Biol. Chem. 281, 15869–15877 14 Bencsath, K.P. et al. (2002) Identification of a multifunctional binding 41 Nishida, T. et al. (2000) A novel mammalian Smt3-specific isopeptidase site on Ubc9p required for Smt3p conjugation. J. Biol. Chem. 277, 1 (SMT3IP1) localized in the nucleolus at interphase. Eur. J. Biochem. 47938–47945 267, 6423–6427 15 Bylebyl, G.R. et al. (2003) The SUMO isopeptidase Ulp2 prevents 42 Cheng, J. et al. (2006) Role of desumoylation in the development of accumulation of SUMO chains in yeast. J. Biol. Chem. 278, 44113– prostate cancer. Neoplasia 8, 667–676 44120 43 Reverter, D. and Lima, C.D. (2004) A basis for SUMO protease 16 Tatham, M.H. et al. (2001) Polymeric chains of SUMO-2 and SUMO-3 specificity provided by analysis of human Senp2 and a Senp2– are conjugated to protein substrates by SAE1/SAE2 and Ubc9. J. Biol. SUMO complex. Structure (Camb.) 12, 1519–1531 Chem. 276, 35368–35374 44 Shen, L.N. et al. (2006) The structure of SENP1-SUMO-2 complex 17 Fu, C. et al. (2005) Stabilization of PML nuclear localization by suggests a structural basis for discrimination between SUMO conjugation and oligomerization of SUMO-3. Oncogene 24, 5401–5413 paralogues during processing. Biochem. J. 397, 279–288 18 Li, Y. et al. (2003) Positive and negative regulation of APP 45 Xu, Z. and Au, S.W. (2005) Mapping residues of SUMO precursors amyloidogenesis by sumoylation. Proc. Natl. Acad. Sci. U. S. A. 100, essential in differential maturation by SUMO-specific protease, 259–264 SENP1. Biochem. J. 386, 325–330 19 Pichler, A. et al. (2002) The nucleoporin RanBP2 has SUMO1 E3 ligase 46 Reverter, D. and Lima, C.D. (2006) Structural basis for SENP2 activity. Cell 108, 109–120 protease interactions with SUMO precursors and conjugated 20 Pedrioli, P.G. et al. (2006) Automated identification of SUMOylation substrates. Nat. Struct. Mol. Biol. 13, 1060–1068 sites using mass spectrometry and SUMmOn pattern recognition 47 Shen, L. et al. (2006) SUMO protease SENP1 induces isomerization of software. Nat. Methods 3, 533–539 the scissile peptide bond. Nat. Struct. Mol. Biol. 13, 1069–1077 21 Mossessova, E. and Lima, C.D. (2000) Ulp1–SUMO crystal structure 48 Kim, J.H. et al. (2006) Roles of sumoylation of a reptin chromatin- and genetic analysis reveal conserved interactions and a regulatory remodelling complex in cancer metastasis. Nat. Cell Biol. 8, 631–639 element essential for cell growth in yeast. Mol. Cell 5, 865–876 49 Uchimura, Y. et al. (2004) Generation of SUMO-1 modified proteins in 22 Barrett, A.J. and Rawlings, N.D. (2001) Evolutionary lines of cysteine E. coli: towards understanding the biochemistry/structural biology of peptidases. Biol. Chem. 382, 727–733 the SUMO-1 pathway. FEBS Lett. 564, 85–90 www.sciencedirect.com Review TRENDS in Biochemical Sciences Vol.32 No.6 295

50 Panse, V.G. et al. (2003) Unconventional tethering of Ulp1 to the 61 Kadoya, T. et al. (2000) Inhibition of Wnt signaling pathway by a novel transport channel of the nuclear pore complex by karyopherins. Nat. axin-binding protein. J. Biol. Chem. 275, 37030–37037 Cell Biol. 5, 21–27 62 Kim, K.I. et al. (2000) A new SUMO-1-specific protease, SUSP1, that is 51 Quimby, B.B. and Dasso, M. (2003) The small GTPase Ran: highly expressed in reproductive organs. J. Biol. Chem. 275, 14102– interpreting the signs. Curr. Opin. Cell Biol. 15, 338–344 14106 52 Johnson, E.S. and Blobel, G. (1999) Cell cycle-regulated attachment of 63 Deyrieux, A.F. et al. (2007) Sumoylation dynamics during keratinocyte the ubiquitin-related protein SUMO to the yeast septins. J. Cell Biol. differentiation. J. Cell Sci. 120, 125–136 147, 981–994 64 Rajan, S. et al. (2005) Sumoylation silences the plasma membrane leak 53 Takahashi, Y. et al. (1999) Smt3, a SUMO-1 homolog, is conjugated to K+ channel K2P1. Cell 121, 37–47 Cdc3, a component of septin rings at the mother-bud neck in budding 65 Yamaguchi, T. et al. (2005) Mutation of SENP1/SuPr-2 reveals an yeast. Biochem. Biophys. Res. Commun. 259, 582–587 essential role for desumoylation in mouse development. Mol. Cell 54 Makhnevych, T. et al. (2007) The role of karyopherins in the regulated Biol. 25, 5171–5182 sumoylation of septins. J. Cell Biol. 177, 39–49 66 Yukita, A. et al. (2004) XSENP1, a novel sumo-specific protease in 55 Galy, V. et al. (2004) Nuclear retention of unspliced mRNAs in yeast is Xenopus, inhibits normal head formation by down-regulation of Wnt/b- mediated by perinuclear Mlp1. Cell 116, 63–73 catenin signalling. Genes Cells 9, 723–736 56 Niepel, M. et al. (2005) The nuclear pore complex-associated protein, 67 Song, J. et al. (2004) Identification of a SUMO-binding motif that Mlp2p, binds to the yeast spindle pole body and promotes its efficient recognizes SUMO-modified proteins. Proc. Natl. Acad. Sci. U. S. A. assembly. J. Cell Biol. 170, 225–235 101, 14373–14378 57 Zhao, X. et al. (2004) Mlp-dependent anchorage and stabilization of a 68 Hoen, D.R. et al. (2006) Transposon-mediated expansion and desumoylating enzyme is required to prevent clonal lethality. J. Cell diversification of a family of ULP-like genes. Mol. Biol. Evol. 23, Biol. 167, 605–611 1254–1268 58 Nissan, T.A. et al. (2002) 60S pre-ribosome formation viewed from 69 Chosed, R. et al. (2007) Structural analysis of Xanthomonas XopD assembly in the nucleolus until export to the cytoplasm. EMBO J. 21, provides insights into substrate specificity of ULPs. J. Biol. Chem. 282, 5539–5547 6773–6782 59 Kim, Y.H. et al. (2005) Desumoylation of homeodomain-interacting 70 Murtas, G. et al. (2003) A nuclear protease required for flower- protein kinase 2 (HIPK2) through the cytoplasmic-nuclear shuttling of ing-time regulation in Arabidopsis reduces the abundance of the SUMO-specific protease SENP1. FEBS Lett. 579, 6272–6278 small ubiquitin-related modifier conjugates. Plant Cell 15, 2308– 60 Itahana, Y. et al. (2006) Nucleocytoplasmic shuttling modulates 2319 activity and ubiquitination-dependent turnover of SUMO-specific 71 Huang, D.T. and Schulman, B.A. (2006) Breaking up with a kinky protease 2. Mol. Cell Biol. 26, 4675–4689 SUMO. Nat. Struct. Mol. Biol. 13, 1045–1047

Elsevier joins major health information initiative

Elsevier has joined with scientific publishers and leading voluntary health organizations to create patientINFORM, a groundbreaking initiative to help patients and caregivers close a crucial information gap. patientINFORM is a free online service dedicated to disseminating medical research.

Elsevier provides voluntary health organizations with increased online access to our peer-reviewed biomedical journals immediately upon publication, together with content from back issues. The voluntary health organizations integrate the information into materials for patients and link to the full text of selected research articles on their websites.

patientINFORM has been created to enable patients seeking the latest information about treatment options online access to the most up-to-date, reliable research available for specific diseases.

For more information, visit www.patientinform.org